Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

6.1 Regulation of Cortical Circuit Formation

The mammalian neocortex is a complex, highly organized structure that contains hundreds of different neuronal cell types and diverse types of glial cells (Guillemot 2007; Molyneaux et al. 2007). It is the most anterior part of the telencephalon and is responsible for sensory perception, high cognitive functions, and consciousness; as such, it has undergone pronounced expansion during evolution, with maximal representation in the human cortex (Selzer 1990). The complex cortical functions rely on the formation of precise networks and subnetworks among the many neuron types during development. These networks form in a stereotyped manner able to create a reproducible human cortex and to facilitate common human behavior. Certain cortical circuits are also preserved among species throughout evolution, while new circuits and functions have been added to the more primitive existing structures (Innocenti 2011; Molnar 2011).

The accuracy and complexity of cortical circuitry predicts that the developmental mechanisms that direct each of these neurons to connect with its siblings must necessarily be precise. Several processes are conserved during evolution, and certain mechanisms are added or modified to create new networks that expand the cognitive capabilities of the cortex. Remarkable advances have been made in recent years in our understanding of these mechanisms and their spatial and temporal coordination, but we still know only a fraction of them. An additional difficulty is that the intricate connectivity and physiology of these circuits is far from being definitively untangled. Much of the knowledge comes from relatively simple animal models, including mice, which have a lissencephalic (smooth) cortical surface, whereas the close resemblance and evolutionary distance of the gyrencephalic brain of ferrets and cats provide excellent tools for deciphering processes exclusive to higher mammals. Relevant information is also derived from the study of human genetic conditions that affect cognitive capabilities, such as schizophrenia, autism, micro- and macroencephaly, and other syndromic and non-syndromic forms of mental retardation (Clowry et al. 2010; Manzini and Walsh 2011). In the near future, the field will benefit from the use of these approaches combined with new technologies and computer modeling to make a decisive step forward.

Neurons of the cerebral cortex can be classified into two broad classes, excitatory and inhibitory neurons. Inhibitory GABAergic, locally connecting neurons are born in the basal telencephalon and have modulatory functions. Excitatory neurons are of dorsal origin and are pyramidal neurons (most abundant) and spiny stellate excitatory interneurons of layer IV. Pyramidal neurons are projecting neurons; some extend their axons to distant subcortical and subcerebral targets, and others project to local and distant intracortical targets (Selzer 1990). This chapter will focus mainly on the connectivity of excitatory neurons, which integrate and transmit information between different neocortex regions and to other regions of the brain (subcortical targets).

6.2 General Structure of Cortical Connectivity

The cerebral cortex is a laminated structure, and each lamina or layer contains neurons with similar morphologies, connectivity patterns (Selzer 1990), and molecular identities (Molyneaux et al. 2007) that originate sequentially during development from radial precursors (Caviness et al. 1996; Takahashi et al. 1996; Heins et al. 2002; Malatesta et al. 2003; Hansen et al. 2010). The number of layers, their thickness, cell composition, and architecture varies throughout the tangential surface of the cortex and among the different functionally specialized areas. The neocortex, and by extension most of the cortex, is composed of six layers, numbered I to VI, which show further expansion and subdivisions in human. Most sensory information is routed to the cerebral cortex from the thalamus (Selzer 1990) and is conveyed to extracortical targets via corticofugal projections (Fig. 6.1). Nevertheless, the vast majority of cortical neuron connections are from one cortex region to another (intracortical) rather than to subcortical targets, allowing complex processing and integration (Fig. 6.2).

Fig. 6.1
figure 1

The cortical column. Scheme showing the connectivity of a column in the somatosensory cortex. The precise connectivity of columns shows some variations on this general pattern among functional areas. Circular grey cells represent inhibitory interneurons; diamonds indicate excitatory interneurons in layer IV

Full size image
Fig. 6.2
figure 2

The corpus callosum. (a) Myelinated axons of the CC project from neurons in layers II and III (~80 % in mouse) and in layer V (~20 %), and a very small population from cells in layer VI (not shown). (b) Confocal micrographs showing somas and CC axons of GFP-expressing layer II–III neurons in the P21 cortex. Neuronal morphology was analyzed at P21 after in utero electroporation at E15.5. Axons of CC neurons projecting from layers II–III invade the cortical plate at homotypic areas (six layered cortex), where they branch and profusely innervate layers II–III and V (a and b). (c) Layer specific pattern of innervation in the contralateral site. Magnification from b

Full size image

Cortical connectivity can be visualized in a simple scheme that reflects its hierarchical organization and the mechanism of origin during development. Radial inter-laminar connectivity establishes the most essential intracortical circuit, the so-called cortical columns (Fig. 6.1). These columns, composed of neurons from different layers, were described from early electrophysiological recordings in the sensory cortex demonstrating that neurons inside the column respond with similar activity to precise stimuli. The thalamic input is preferentially distributed vertically in columns to superficial and deep layers, rather than horizontally (Mountcastle et al. 1957). In the sensory cortex, neurons in a cortical column all process sensory information from the same peripheral location and submodality (Feldmeyer et al. 2013). These studies were extended in the visual cortex by Hubel and Wiesel, who showed among other things that innervation of the visual cortex from the two eyes is also organized in columns (ocular dominance) and discovered orientation columns (Hubel and Wiesel 1962, 1963). Columnar organization is also the result of the common precursor foundation and the migration mode of cortical pyramidal neurons during development (Rakic 1988; Heins et al. 2002; Malatesta et al. 2003; Torii et al. 2009; Jones and Rakic 2010).

Columns communicate tangentially through laminar connectivity, essentially through layers II–III and V, to form the functionally specialized areas of the cortex generally classified as sensory, motor, and association areas (Rakic 1988). In the adult, the transition from one neocortical area to another can be defined by differences in cytoarchitecture, gene expression patterns, input projections, and by the specific mode of connections between neurons of the column. These properties determine the physiology and connectivity of specific circuits to allow the functional specializations that distinguish areas. For example, the somatosensory area in the mouse is defined by a thicker layer IV, expression of markers such as Rorβ, input from the whiskers and barrel formation. Finally, areas are interconnected, facilitating integration and complex behavior. Interhemispheric commissural axons permit information exchange between the cerebral hemispheres, whereas other axons that do not cross the midline, but run along the anterior posterior axis, connect areas from the same hemisphere.

In essence, the mechanisms that control cortical circuit formation during development select axon pathways and influence formation of dendritic structures and synapses, as will be discussed below. Studies in recent years have shown remarkable coordination between intrinsic molecular programs that specify neuronal cell identity and those regulating their connectivity. In the last two decades, numerous studies have reported examples of transcription factors (TF) expressed only by selected neuronal subtypes that regulate discrete aspects of connectivity (Hevner et al. 2001; Molnar et al. 2003; Jacobs et al. 2007). The pattern of overlapping functions of these TF creates cell diversity and acts as a genetic code that encrypts the rules that govern cortical networks. These intrinsic programs regulate fundamental aspects such as neurotransmitter expression, cell morphology, and the ability to respond selectively to external cues, including soluble factors and membrane-bound molecules. These mechanisms are discussed separately in this chapter. Finally, during postnatal stages, experience- and activity-mediated mechanisms involved in plasticity ultimately shape the circuits and give rise to the final stereotypical networks (Metin et al. 1997; Molnar and Cordery 1999).

6.3 Corticofugal Neurons

Projection neurons extend their axons to distant subcortical targets to transmit information to other brain regions. They are located mainly in the deeper layers of the cerebral cortex and are generally referred to as extracortical projection neurons or corticofugal neurons, and are subdivided in subcerebral and corticothalamic. Subcerebral projection neurons reside mostly in layer V and innervate different parts of the brain stem and cerebellum, as well as the higher-order thalamic nuclei (HOTN) through secondary collaterals (Fig. 6.3). The neurons of the HOTN relay cortico-cortical information by projecting excitatory fibers to layers I, IV, and VI of a cortical area distinct from that from which they receive input. Subcerebral projecting layer V neurons can be subdivided into three major subpopulations, corticotectal, corticospinal, and corticopontine. Corticotectal neurons are located in the visual cortex; they send their primary axon to the superior colliculus and secondary collaterals to the rostral pons. Corticospinal motor neurons reside in the sensorimotor area of the cortex; they send primary projections to the spinal cord and secondary collaterals to the striatum red nucleus, caudal pons and medulla. Finally, corticopontine neurons are in charge of transmitting information to the pons (Molyneaux et al. 2007) (Fig. 6.3).

Fig. 6.3
figure 3

The development of corticofugal axons. Scheme of the axonal pathway of corticofugal neurons of cortex layers V and VI. The different anatomic and genetic regions these axons encounter are depicted. Coronal (a) and sagittal views (b). DTB dorsal telencephalic boundary, FOTN first order thalamic nuclei, HOTN high-order thalamic nuclei, PSPB pallial-subpallial boundary, RTN reticular thalamic nuclei, SC superior colicullum, SPC spinal cord

Full size image

Corticothalamic neurons are located in layer VI and enable cortical processing of peripheral data. They project axons to and receive input from the first-order thalamic nuclei (FOTN) (Fig. 6.3). These nuclei receive peripheral sensory input and relay it to layer IV and VI neurons. Corticothalamic primary axons generate numerous small synapses with thalamic neurons, providing signals for peripheral information. Corticothalamic neurons projecting from layer VI primary visual cortex (V1) send axons to the dorsal lateral geniculate nucleus (dLGN); those in the auditory cortex (A1) project to the medial geniculate nucleus (MGN) and those in the primary somatosensory cortex (S1), to the ventrobasal nucleus (VB). The axon collaterals of these neurons innervate the reticular thalamic nucleus (RTN).

6.3.1 Development of Corticofugal Tracts

Development of corticofugal tracts follows a complex process by which distinct neuron subpopulations innervate specific extracortical regions in a temporal pattern with characteristic axon outgrowth kinetics. The subset of TF expressed by each neuron confers a unique identity, essential for its connectivity pattern and behavior. This identity would nonetheless be worthless in the absence of long- and short-range guidance cues that follow spatiotemporal dynamics. The development of corticofugal tracts is also closely associated with thalamocortical tract formation, since axons that form both tracts establish the physical association necessary to guide each other and to complete their development. Considerable controversy nonetheless remains regarding the relative importance of this interaction and of other intrinsic and extrinsic mechanisms. This will not be discussed here in detail, as the reader can find many complete reviews (Cang et al. 2005; Torii and Levitt 2005; Rash and Grove 2006; Rubenstein 2011).

The preplate contains the first subsets of cortical differentiated neurons and gives rise to Cajal-Retzius and to subplate cells. The latter are the first cortical neurons to extend their axons into the internal capsule, the natural path to extracortical territories. These initial projections act as a scaffold for subsequent corticofugal axons; the majority will disappear in the early postnatal period, correlating with a wave of cell death that eliminates their somas (Hevner et al. 2001; Jacobs et al. 2007). Neurons that form permanent connections between the cortex and extracortical regions will begin to extend their neurites at around embryonic day (E)10. Depending on their location and identity, their axons take a lateral, medial, rostral, or caudal trajectory, and grow until they reach the region adjacent to the lateral internal capsule. The distinct populations arrive at this zone at slightly different times between E13 and E15.5, depending on the position of their somas; the lateral fibers are the first to arrive and the dorsally derived fibers, the last (Fig. 6.3a). At this point, temporal synchronization requires axons to align in order to continue their journey together. The first incoming axons await the arrival of the others before continuing growth; this is termed the first waiting period. All the axons then cross the pallial-subpallial boundary (PSPB) and enter the internal capsule. The PSPB is a major boundary that expresses a very specific subset of TF (high Pax6, null Emx1, Dlx1). This territory has modulatory potential, making early corticofugal projections turn sharply from their original ventrolateral to a medial trajectory, to cross the subpallium. The internal capsule is the site at which early corticofugal axons emitted from the subplate and thalamocortical projections first meet and establish a close interaction that will be maintained throughout the intermediate zone, PSPB, and the lateral sector of the internal capsule; this interaction is needed for guidance (Hevner et al. 2001, 2002; Lopez-Bendito et al. 2007; Chen et al. 2012; Grant et al. 2012) (23).

Once the axons exit the internal capsule, they arrive at the diencephalon-telencephalon boundary (DTB), where they enter the prethalamus and encounter the cells of the perireticular nuclei (PRN) and RTN at E16 (Fig. 6.3a). The extension will undergo a second pause that lasts until E17.5 (second waiting period). At this time, corticofugal projections continue through different pathways (Fig. 6.3a, b). Layer V primary axons continue to grow and cross the cerebral peduncle to the brainstem and spinal cord. Layer VI primary axons and layer V collaterals change direction to enter the thalamus, a process that takes several days and results in postnatal innervation of most thalamic nuclei. In higher mammals, this correlates with the functional establishment of behaviors associated with the relevant sensory systems; somatosensory and motor functions mature before visual and auditory functions. For example, in mice, somatosensory ventrobasal and motor ventrolateral nuclei are innervated earlier (E18.5 and P0.5) than auditory MGN and visual dLGN, which are not fully innervated until postnatal day (P)8 (O’Leary and Koester 1993; Metin et al. 1997; Molnar and Cordery 1999; Molnar et al. 2003; Jacobs et al. 2007; Grant et al. 2012; Lickiss et al. 2012).

One of the most fascinating characteristics of layer V and VI axons is therefore that they must navigate through several distinct territories until they reach their target. This requires dynamic recognition of territory-specific signals and modulation of axon responses. It has become apparent that several neuron populations and their axons, such as the thalamic afferents discussed above, provide structural support essential for crossing these anatomic regions and their boundaries. Pioneer axons are those of neurons (in this case, subplate neurons) that, thanks to their intrinsic electrical activity, can navigate without the help of preexisting axons and pave the way for follower axons. Voltage-gated ion channels, which in subplate neurons are voltage-gated K + 3.4 (Kv3.4), are responsible for the intrinsic electrical activity patterns of neurons, and are thus necessary for corticofugal development (Huang et al. 2012). The corridor cells, a population derived from the lateral ganglionic eminence, also illustrate these cooperative interactions. These cells are needed to generate a permissive substrate for cortical axon growth across the medial ganglionic eminence (MGE). The axon guidance functions of corridor cells overlap with the guidance and sorting functions of PRN neurons, thought to have a role in directional change in the internal capsule (Lopez-Bendito et al. 2006; Grant et al. 2012).

Following spatiotemporal dynamics, axons respond differently to distinct sets of cues in the environment they traverse. These specific behaviors enable correct navigation and innervation of their targets. These guiding factors include intrinsic factors at the neuron that emits the axon (e.g., cell surface receptors or molecules that influence intracellular signaling) as well as extrinsic factors (membrane-bound and soluble factors presented or secreted by intermediate or final targets); the latter act at short and long range, and affect growth cone extension as well as orientation by generating repulsive or attractive responses. Soluble molecules often establish concentration gradients critical for precise axon guidance of corticofugal neurons.

6.3.2 Guidance Factors and Receptors that Direct Corticofugal Axons

Although gaps remain in our knowledge, several families of guidance molecules are known to determine the trajectory of corticofugal axons. We summarize a series of illustrative examples. The semaphorin family provides early context-dependent cues. Pioneer explant experiments showed that Sema3A expression in the most superficial cortical plate, the marginal zone (MZ), is responsible both for repelling axons toward the VZ (Polleux et al. 1998) and for attracting apical dendrites (Polleux et al. 2000). Further complementary studies demonstrated that combinations of Sema3 molecules have a specific effect on the corticofugal axon pathway. For example, in addition to the superficial cortical plate, Sema3A is expressed throughout the ventricular zone and lower subventricular zone, and Sema3C is expressed in the intermediate and the subventricular zones. Although cortical axons are exposed to Sema3A and Sema3C concurrently, Sema3A has a repulsive effect that overrides Sema3C attraction, even at very low concentrations. As a result, corticofugal axons grow over the corridor generated at the intermediate zone and the upper SVZ, where Sema3C is expressed alone (Ruediger et al. 2012). Likewise, Sema5B is expressed in many regions of the corticofugal pathway, including the ventricular zone and the ventrolateral cortices, and inhibits axon entry into these territories (Bagnard et al. 2001; Lett et al. 2009). Sema molecules bind to neuropilins, whose expression and differential association with plexins also critically modulate cortifugal axon responses and dynamics (Pasterkamp 2012). Several pathways involving Sema signaling alone can thus explain many of the corticofugal axon turns and trajectories.

Netrin-1 is expressed in the internal capsule and mediates long-range attraction of corticothalamic axons at E12.5–13.5. The attractive effects of netrin-1 can induce axon turning and thus appears to be responsible for corticofugal growth cone reorientation toward the ventral telencephalon. Slit1 and 2 have a major role in corticothalamic and thalamocortical axon guidance within the ventral telencephalon and diencephalon, mainly through binding to Robo1 and Robo2 receptors, which appear to have partially redundant functions. In Robo mutant mice, and more markedly in Robo1 and Robo2 double mutants, corticothalamic axons do not grow through the internal capsule but are aberrantly directed to cross the midline. In addition, Robo1 (but not Slit) appears to act as a slowing signal, since both corticothalamic and thalamocortical axons grow faster in Robo1 knockouts (ko) than in WT mice. This deceleration might be relevant in the developmental control of the temporal dynamics of these tracts, specifically in the regulation of the two waiting periods (Andrews et al. 2006; Lopez-Bendito et al. 2007; Grant et al. 2012).

Finally, the EphA family of tyrosine kinase receptors and their ligands are essential for the initial establishment of corticothalamic targeting. Neocortical neurons express an EphA7 gradient that controls the topography of corticothalamic projections, through local interactions within individual thalamic nuclei. Other EphA proteins, such as EphA5, also have a role in the correct patterning of corticothalamic and thalamocortical wiring (Sestan et al. 2001; Cang et al. 2005; Torii and Levitt 2005; Torii et al. 2013).

Further studies are needed to better delineate the elements that determine corticofugal connectivity. As these neurons are characterized by their long-distance journeys, the challenge is not only to understand what these signals are and how they are transduced, but also the nature of the spatiotemporal mechanisms that regulate them.

6.4 The Formation of Intracortical Circuits

6.4.1 The Development of Callosal Projecting Neurons

Interhemispheric connections are essential components of intracortical circuits and contribute to the integration ability and high associative function of the mammalian brain. The corpus callosum (CC) and the anterior commissure formed by axons of layer V are the main commissures that connect the hemispheres. The CC is the major commissural track of the mammalian brain. Partial or total CC agenesis is associated with many human developmental syndromes that affect the brain (Fame et al. 2011). Most myelinated axons of the CC project from neurons in layers II and III (~80 % in the mouse) and in layer V (~20 %), and a very minor population from cells in layer VI. A number of callosal neurons also send axonal collaterals to the same hemisphere (ipsilateral) and communicate cortical areas. There are also dual connections to the contra- and ipsilateral striatum. Axon guidance cues and synaptic maturation mechanisms that target callosal neurons and their projections are critical in the development of this important cortical circuitry.

In the several steps of axon routing involved in CC formation, different glial and neuronal cells act as intermediate guideposts and present secreted and membrane-bound navigation signals. Defects in hemisphere fusion cause partial or total CC agenesis; fusion occurs simultaneously as callosal neurons are born, just before they extend their axons, and is necessary for axons to cross the midline. Early studies showed that CC axons are guided across the cerebral midline by a glial population, then termed sling-like glial and now known as the glial sling. These astroglial populations form a bridge-like structure at the midline between the two lateral ventricles (Hankin et al. 1988; Silver et al. 1993). It was shown early on, that in acallosal mice midline crossing could be restored postnatally when this glial scaffold was presented artificially (Silver and Ogawa 1983). More recent observations in mice and humans nonetheless show that many neurons are also present within the glial slings (Shu et al. 2003a; Ren et al. 2006). Semaphorin 3C expression in one of these transient neuronal populations helps to attract callosal axons to and through the midline (Niquille et al. 2009). Additional glial structures in the CC are considered relevant for axon navigation, including radial glial cells in the glial wedge (GW) and astrocytes in the indusium griseum (IG) (Shu and Richards 2001; Shu et al. 2003b). In the developing CC, GW-expressed Slit2 guides callosal axons to the corticoseptal boundary (Bagri et al. 2002; Shu et al. 2003c). Robo receptors bind to Slit proteins; callosal axons express Robo1, and mice deficient in this protein (Robo1 −/−) have defects in CC formation (Shu and Richards 2001; Andrews et al. 2006; Lopez-Bendito et al. 2007). Once axons cross the midline, the same signal repels them from this boundary (Bagri et al. 2002; Shu et al. 2003c). Other long-range molecules such as Wnt are necessary for the guidance of callosal axons. Wnt5a is expressed by the GW and the IG cells, and stimulates both outgrowth and repulsion of developing callosal axons via Ryk receptors (Keeble et al. 2006; Li et al. 2010). Other signals such as ephrins and their receptors (EphA5, EphB1 and EphrinB3) act at a shorter range and are essential not only for callosal formation, but also have a broader effect on other commissures (Mendes et al. 2006; Lindwall et al. 2007).

CC formation is also highly dependent on the earlier extensions emitted by a population of pioneer callosal neurons. This is the earliest neuron population to extend axons across the midline, at around E17 in the mouse. The cell bodies of these neurons are located in the most medial part of the cortical plate and the cingulate cortex, and their axons appear to guide the neocortical callosal projections (Koester and O’Leary 1994; Rash and Richards 2001; Fame et al. 2011). Short-range signals such as neuropilin 1 (Nrp1) regulate crossing of these early axons (Hatanaka et al. 2009; Piper et al. 2009).

Callosal axons initiate their journey guided by this plethora of signals. After midline crossing, they travel along the CC; they make a sudden turn in their trajectory and invade the contralateral cortical plate at homotypic areas. Little is known about the mechanisms that trigger this turn, but it might imply changes in axon capacity to respond to cortical cues, similar to those that occur when they cross the midline. Recognition of the correct contralateral territories might also imply recognition of lateral gradients at the cortical plate, although these mechanisms remain unclear.

Axons are able to branch and extend many synapses along their length, which allows neurons to send information to various cells simultaneously. Callosal axons branch at several points during their trajectory; most branches profusely innervate layers II–III and V in the ipsilateral and contralateral columns (Fig. 6.2), although some neurons (termed dual projecting) also send collaterals to other areas and regions. Despite their probable importance in human cognition, the patterns of these branched connections are not fully resolved, although they are likely to be responsible for certain associative properties of the cortex. For example, an undetermined number of callosal projecting neurons from the sensory cortex simultaneously extend exuberant projections to the contralateral homotypic cortex and to both contralateral and ipsilateral areas of the motor cortex. Laterally located superficial neurons can also extend dual axons toward the midline and the internal capsule, although in the latter case, they apparently retract at P11 (Garcez et al. 2007). Similar schemes of dual projections are found in certain callosal neurons of the motor cortex, which send dual axonal projections to sensory areas (Mitchell and Macklis 2005). In mice, these dual projections show maximum numbers at P8; they are refined until approximately P21, probably through activity-dependent mechanisms, but many persist into adulthood (Innocenti and Price 2005; Mitchell and Macklis 2005). Little is currently known of the molecular control of these double connections.

6.4.2 Factors that Regulate Selectivity of the Synapse: From Intra-Columnar and Intra-Laminar Connectivity to Microcircuits

Based on the work discussed above, it is clear that scientists have successfully identified several crucial regulatory mechanisms responsible for delivering axons to the vicinity of their targets. After this arduous journey, however, only half the job is done. Axons do not establish synapses without a pattern. The nervous system shows considerable specificity at this level, and connections are made only with certain neurons; there is even selection of specific cell compartments. This is extreme in the case of cortical circuits, which implicate hierarchical organization in layers: axons selectively establish connections with certain layers, certain cells within the layers, and even choose between apical or basal dendrites. The cellular and genetic mechanisms responsible for the assembly of specific connections in the nervous system are the subject of intense study. These mechanisms involve coordinated expression of homophilic adhesion molecules by both pre- and postsynaptic partners, including the diverse cadherins and immunoglobulin superfamily (IgSF) proteins. Repulsive signals also prevent abnormal innervation (Shen and Scheiffele 2010; de Wit et al. 2011).

Few of the mechanisms known to select synaptic targets in other parts of the nervous system have been reported or tested in the cortex; there is an intriguing relative lack of knowledge about the elements that implement the beautiful patterns of cortical laminar connectivity. Barrels, which are prominent sensory units in the rodent somatosensory cortex, have been examined in detail. Data suggest that the initial gross formation of the barrel map relies on molecular cues, while refinement of its topography depends on neuronal activity. Temporal and cell-specific expression of cadherins contributes to the barrel-like distribution of thalamic axonal inputs into layer IV (Huntley and Benson 1999; Inan and Crair 2007). The development of excitatory synapses between axons emitted from layer II–III neurons with dendrites in layers II–III and V, but not those in layers IV and VI, is another perfect paradigm of layer-specific synaptic organization. Activity has a role in determining the relative innervation of layers II–III and V by contralateral CC afferent connections. Reduced firing results in increased innervation of superficial layers at the expense of layer V innervation (Mizuno et al. 2007). Recent work identified an unexpected molecular regulators of innervation of layers II–III and V in Shh, a secreted molecule known mainly for its patterning and axon guidance effects, and in its high-affinity receptor Brother of CDO (Boc) (Okada et al. 2006). The restricted Shh expression in layer V promotes synaptic formation with Boc-bearing axons; these axons are precisely those of neurons in layers II–III. Genetic manipulation of mice showed that conditional Shh deletion in the dorsal telencephalon mimics Boc ko phenotypes of layer V neurons. Boc-depleted layer V neurons show reduced dendritic complexity, spine density and synaptic strength as a result of decreased innervation from layer II–III callosal projecting neurons (Harwell et al. 2012). Although alteration of activity or the Shh-Boc pathway did not result in layers being ectopically innervated, these studies open the path to understand layer-specific connections and the possible implications of other patterning molecules in cortical wiring, perhaps in conjunction with activity.

These studies of synaptic specificity mechanisms are also extremely important when considering the existence and formation of microcircuits and subnetworks embedded within cortical circuits. There is cellular and molecular heterogeneity not only between layers and cortical areas, but also within the neurons of the same layer (Fame et al. 2011); this results in the expression of different membrane and secreted proteins that might contribute to generating networks in the cortex. In layers II–III, microcircuits have been described functionally by the characterization of neuron firing patterns (Burgalossi et al. 2011). They have also been identified genetically, through visualization of GFP-labeled neurons that express high c-fos levels, and are highly interconnected, as shown by electrophysiology studies (Yassin et al. 2010). Common neuronal birth origin might be implicated in the formation of these microcircuits and in columnar formation. A common progenitor increases the probability of synapse between neurons, the probability to form strong electrical coupling with each other rather than with adjacent non-sister excitatory neurons, and the likelihood of producing similar excitatory responses (Yu et al. 2009, 2012; Li et al. 2012).

6.4.3 The Regulation of Dendritic Structures

Another facet of the regulation of cortical circuitry is the modulation of postsynaptic structures: dendrites, spines, and synapses. Dendritic branching specifies connectivity with selected axonal input and determines neuron morphology (Shen and Scheiffele 2010). Morphology in turn influences the way information is processed, amplifies or silences presynaptic input depolarization signals (Mainen and Sejnowski 1996), and even affects plasticity (Feldman 2012). Spine density and spine morphology determine the number, strength, and stability of synaptic contacts (Tada and Sheng 2006; Edbauer et al. 2010; Shen and Scheiffele 2010).

Developmental mechanisms that target regulation of postsynaptic structures and compartments have considerable importance in cortical function and circuit modulation, and are critical for the acquisition of higher intellectual abilities. Alterations in dendritic morphology and in spine number and structure are defects that often correlate with cognitive disorders and mental retardation (Tada and Sheng 2006; Bourgeron 2009; Jan and Jan 2010; Kulkarni and Firestein 2012). Many of the mechanisms involved in the control of dendritic structures and synapses were thus identified during the study of human intellectual disabilities, including autism and fragile X syndrome, the most frequent cause of mental retardation. Analysis of human mutations linked to autism often shows alterations in genes that regulate the cytoskeleton and synaptic scaffold (Segal 2001); this is the case of Shank proteins (Bourgeron 2009), kalirin (Penzes and Remmers 2012), and mutations that affect the Ras/Epac2 pathway (Srivastava et al. 2012). Autism-related genes also appear to target postnatal mechanisms of plasticity and synaptic refinement. Human mutations linked to fragile X syndrome (Zhang et al. 2001) affect FMR1, a gene that encodes the RNA-binding protein FMRP (fragile mental retardation protein), which regulates transport and local translation to axons and dendrites (Tada and Sheng 2006; Napoli et al. 2008; Boda et al. 2010; Darnell et al. 2011; Penzes et al. 2011; van Bokhoven 2011; De Rubeis et al. 2012).

Human and mouse genes that encode TF also control dendrite and synapse development. In mice, Mef2a controls activity-dependent dendritogenesis (Fiore et al. 2009) as well as activity-dependent spine deletion (Flavell et al. 2006), which involves downstream use of FMRP (Pfeiffer et al. 2010). Neurog2 regulates early neuritogenesis and alters neuron migration via phosporylation of the small GTPase Rnd2 (Hand et al. 2005), and by forming a DNA-binding complex with the LIM-only protein LMO4 (Asprer et al. 2011). Calcium signals and calcium-binding TF such as CREB are also involved in migration and dendritogenesis in the cortex (Redmond et al. 2002; Redmond and Ghosh 2005). Of the several cortical layer-specific TF described so far, the expression in mice of Fezf2/Zfp312 in layer V neurons (Chen et al. 2005) and of Cux1 and Cux2 in layers II-IV regulate dendrite formation, and also synaptogenesis in the case of Cux proteins (Chen et al. 2005; Cubelos et al. 2010). Cux TF functions might be linked to evolution; the number of superficial layers in mammals expands together with brain volume and is maximal in humans (Hill and Walsh 2005). This correlates with the fact that upper layer neurons participate in highly associative circuits and tasks, and show an extreme degree of interconnectivity. It is thus possible that Cux optimize these neurons to increase their connectivity and their capacity to integrate information.

In a similar conceptual line, the two human-specific duplications of SRGAP2 are proposed to be a delay mechanism for synaptic maturation which expands the temporal window of neonatal plasticity in humans. Mice bear one copy of the SRGAP2 gene, while humans have three alleles (A, B, and C). In the mouse neocortex, SRGAP2 promotes spine maturation and limits spine density. The human SRGAP2B and SRGAP2C duplications are partial and encode truncated forms that dimerize with the ancestral SRGAP2 (SRGAP2A) protein. Surprisingly, this dimerization inhibits normal SRGAP2 function. Thus, experiments in mice show that ectopic expression of hSRGAP2C phenocopies SRGAP2 deficiency; in both cases, mice have abundant, immature long spines. These findings suggest that inhibition of SRGAP2 function by its human-specific paralogs has contributed to evolution of the human neocortex (Charrier et al. 2012). In sum, these studies suggest that specific mechanisms that target dendritic structures and synapses contribute to the evolution of cerebral cortical circuits and the definition of human intellectual capacity.

6.5 Molecular Identity of Cortical Neurons: Layer and Area Identity as Determinants of Connectivity

Molecular identity is broadly defined by the subset of genes expressed by each neuron. Subtype-specific TF ultimately determine the molecular identity of neurons by initiating and maintaining specific genetic programs. Expression of these TF is often interconnected through gene expression cascades (Molyneaux et al. 2007; Leone et al. 2008; Fame et al. 2011). Neuron identity programs are initiated early in dividing cells by progenitor-specific TF and passed on to neuronal progeny through expression of the same or other subtype-specific TF (Molyneaux et al. 2007; Leone et al. 2008; Fame et al. 2011). Because laminar organization of the cortex coincides with the segregation of neuron subpopulations, many of the TF that specify neuronal identity have been identified as layer specific (Fig. 6.4). In the last two decades, genetic studies in mice have shown how several of these layer-specific TF modulate different aspects of connectivity during development and indicate that they are related to almost every process the neurons undergo. This is a fascinating and dynamic field, as indicated by the ongoing identification of genes essential for determination of each neuron’s fate and behavior and, thus, its connectivity. The most recent studies clearly established that there is even further molecular diversity within the layers, which could explain the instructive signals that direct formation of cortical circuits and microcircuits.

Fig. 6.4
figure 4

The molecular identity of cortical neurons. Molecular identity is defined by the subset of TF expressed by each neuron. Many of the TF that specify neuronal identity have been identified as layer-specific factors. Neuron identity programs are initiated early in dividing cells by progenitor-specific TF and passed on to neuronal progeny through expression of the same or other subtype-specific TF. This identity determines the connectivity pattern of these neurons. Schematic representation of reported molecular and genetic interactions that inter-regulate the expression of subclass-specific TF

Full size image

It is increasingly apparent that there are many TF genes common to all projection neurons, which would explain the common pattern of initial development. A smaller group of TF defines closely related subtypes of projection neurons and an even smaller group is characteristic of each neuron population (Arlotta et al. 2005; Molyneaux et al. 2007; Leone et al. 2008). Most studies analyze the phenotypes of neurons with loss and gain of function of specific genes. More research is needed to fully understand the specification of all these neuron subtypes and the molecular mechanisms underlying their integration into selected circuitries. We can nonetheless begin to define some mechanisms that are quite illustrative of the extreme importance of the TF selective mode of control.

6.5.1 Transcription Factors in Lower Layers

Sox5, Ctip2 (COUP-TF-interacting protein 2), and Tbr1 expression patterns selectively mark distinct subtypes of corticofugal populations (Fig. 6.4). Subplate neurons express an intermediate level of Sox5, high Tbr1, and low Ctip2 levels; corticothalamic neurons in layer VI express Sox5 and Tbr1 strongly and little Ctip2, and subcerebral projection neurons in layer V show high Ctip2 levels, intermediate Sox5, and little Tbr1. These expression patterns prompt the hypothesis that these proteins form a coregulatory network that governs the adoption of neuronal fates (Fig. 6.4) (Arlotta et al. 2005; Molyneaux et al. 2007).

Tbr1, a T-box family TF gene, is expressed soon after cortical progenitors begin to differentiate (Fig. 6.4). It is found at high levels in early-born neurons of the preplate and layer VI and is necessary for their correct differentiation, as it is for cortical laminar organization and guidance of cortical afferent and efferent axons (Bulfone et al. 1995; Hevner et al. 2001). Several studies suggest that its functions overlap partially with those of Sox5, although defects in Tbr1 ko mouse cortex are more severe. In the absence of Tbr1, the corticothalamic tract disappears and there is greater upregulation of neuronal markers than in Sox5 ko mice. Chromatin immunoprecipitation and luciferase assays demonstrated that Tbr1 binds to and inhibits Fezf2 promoter (McKenna et al. 2011).

Studies of Sox5 ko mice and of its overexpression demonstrate that Sox5 is critical for generation of diversity in extra-cortical projecting neurons, as it regulates and coordinates timing of sequential emergence of the different corticofugal neuron types (subplate, corticothalamic, and subcerebral) during early corticogenesis. Sox5 expression is essential for correct differentiation of corticothalamic and subplate neurons, and blocks premature emergence of subcerebral neurons. When Sox5 is absent, subplate and corticothalamic neurons locate to more superficial areas, while subcerebral neurons accumulate within layer VI and the white matter. This is interpreted as an anomalous overlap in the generation of the three principal corticofugal neuron subtypes. In addition, in the Sox5 ko mouse cortex, subplate neurons aberrantly express molecular hallmarks and connectivity patterns of subcerebral projection neurons, resulting in the appearance of additional subcerebral projection tracts. Differentiation of corticothalamic neurons is imprecise, and that of subcerebral projection neurons is accelerated. In contrast, Sox5 gain of function at later stages of corticogenesis causes reemergence of neurons with corticofugal features (Lai et al. 2008).

Ctip2 is one of the molecular targets of Sox5 that is upregulated in the subplate of Sox5 ko mice (Lai et al. 2008). Ctip2 is also a major downstream effector of Fezf2; it is expressed at high levels in layer V corticospinal and corticotectal neurons, and at much lower levels in layer VI corticothalamic neurons. Ctip2 expression begins once neurons reach the cortical plate and is not implicated in early specification of cortical precursors (Arlotta et al. 2005). Ctip2 participates in directing the extension, fasciculation, and refinement of subcerebral axonal projections, particularly the ability of corticospinal neurons to extend projections to the spinal cord during formation of the corticospinal tract. Thus, Ctip2 ko axons fail to extend past the pons to reach the spinal cord (Arlotta et al. 2005; Lickiss et al. 2012).

Fezf2 represses callosal neuron identity, is sufficient for specification of layer V subcortical projection neurons, and is needed for layer VI neuron maturation (Rouaux and Arlotta 2010). Ctip2- and Fezf2-null mice have very similar phenotypes. In Fezf2 ko mice, the corticospinal tract disappears; corticotectal and pontine projections are also greatly reduced; inappropriate new projections appear instead (Chen et al. 2005; Molyneaux et al. 2005). In Fezf2 ko mice, Ctip2 expression is absent, whereas forced expression of Fezf2 by in utero electroporation induces upregulation of Ctip2 in neurons that would not normally express it (Chen et al. 2005, 2008). This suggests that these two genes might act in a common pathway and that Fezf2 is a key upstream regulator of corticospinal projection neuron differentiation.

Although the genetic regulatory pathways of the TF described above are relatively well characterized, there are many other TF that define lower layer identities or are involved in axon extension and pathfinding. Otx1 is expressed in 40–50 % of subcerebral neurons, primarily those of the visual cortex, as well as by a number of cells in layer VI; it is essential for development of the corticotectal projection neurons and controls the refinement and pruning of their axon collaterals (Weimann et al. 1999). Opn3 is a marker of layer V and Foxp2 of layer VI. Er81 is expressed in layer V cortico-cortical and subcerebral projection neurons; Nfh and Pou3f1 are expressed primarily in layer V subcerebral projection neurons (Frantz et al. 1994; Ferland et al. 2003; Hevner et al. 2003; Voelker et al. 2004; Arlotta et al. 2005).

6.5.2 Transcription Factors in Superficial Layers

Several TF define the molecular identity of the superficial layers. We will mention some that exemplify distinct roles in neuron differentiation. Brn1 and Brn2 are two POU domain transcriptional regulators expressed in superficial cortical neurons and are necessary for correct migration and cortical lamination (McEvilly et al. 2002; Sugitani et al. 2002). Genetic loss of Brn1 and Brn2 in mice thus abrogates the appearance of late-born superficial neurons (Sugitani et al. 2002). Other TF directly implement programs that regulate connectivity. Satb2 (AT-rich sequence-binding protein 2) is a chromatin-remodeling TF expressed in a broad subset of layer II–III neurons and in a smaller subpopulation of layer V neurons. Loss of Satb2 expression in mice results in agenesis of the corpus callosum and reorientation of axons toward subcortical targets through the internal capsule. This abnormal wiring scenario is explained by the observation that Satb2 represses expression of Ctip2, which regulates corticofugal identities; Satb2-deficient neurons also have other molecular features of corticofugal projecting neurons (Alcamo et al. 2008; Britanova et al. 2008). An epigenetic regulator, the proto-oncogene Ski, cooperates with Satb2 for callosal axon guidance (Baranek et al. 2012).

Cut-like homeobox proteins Cux1 and Cux2 also mark layers II–III and IV specifically (Nieto et al. 2004; Zimmer et al. 2004). As mentioned above, in cortical layers II–III, both genes regulate dendritogenesis, spine formation, and synaptogenesis in a non-redundant manner and act in combination to specify the final dendritic tree and the synapses of these neurons (Cubelos et al. 2010). CUX2 also defines the upper layers of the human cerebral cortex (Arion et al. 2007), and a possible association of CUX1 polymorphisms with failure of antidepressant response is reported (Sasayama et al. 2012). Additional TF, including Id2, act as markers of the molecular identity of superficial layers. The functions of Bhlhb5, which marks superficial layers but is also found in layer V, are described below.

6.5.3 Area-Specific TF

Neocortical areas are characterized by unique molecular profiles and cyto-architecture, which ultimately reflect specific modes of axonal and dendritic connectivity. A strong deterministic function of TF expressed in the progenitor pools was demonstrated in relation to cortical area formation. Four murine TF, Coup-TFI (Armentano et al. 2007; Faedo et al. 2008), Emx2, Pax6 (Bishop et al. 2000; Mallamaci et al. 2000), and Sp8 (Sahara et al. 2007), all of which are expressed in gradients across the embryonic cortical axis, determine cortical area sizes and positions by specifying or repressing area identities within cortical progenitors. Early expression of area-specific progenitor TF is modulated by morphogens and signaling molecules secreted by patterning centers that are positioned at the perimeter of the dorsal telencephalon. These centers generate graded TF expression in cortical progenitors. Two major patterning centers are the commissural plate, which expresses Fgf8 and Fgf17, and the cortical hem, which expresses Bmps and Wnts (O’Leary and Nakagawa 2002). Progenitor area-specific TF also interact genetically, thus modifying the expression of one another; for example, Pax6 and Emx2 are mutually exclusive (Bishop et al. 2000; Mallamaci et al. 2000). There is interplay between intrinsic genetic mechanisms and extrinsic information conveyed by thalamocortical input to the cortex, especially to layer IV. The relative contribution of each of these early mechanisms to area formation is still debated, and has been reviewed extensively (O’Leary et al. 2007; O’Leary and Sahara 2008).

Expression of progenitor area-specific TF can be downmodulated (Emx2, Pax6) or maintained in postmitotic neurons (Coup-TFI). Area-specific TF generally inhibit or promote expression of other area-specific genes including Cadherin8, Eph receptors and other layer-specific TF such as Satb2, Rorβ, and Id2 (O’Leary et al. 2007; O’Leary and Sahara 2008). Coup-TFI is expressed as a gradient and, during corticogenesis, is needed to maintain the balance between frontal/motor and sensory areas (Armentano et al. 2007). This factor temporally inhibits generation of corticospinal motor neurons, which in large numbers characterize motor areas (Tomassy et al. 2010), and regulates axon outgrowth as well as the formation of the CC and other brain commissures (Armentano et al. 2006), and governs neuronal migration (Alfano et al. 2011).

Arealization is closely linked to the identity of the postmitotic neurons. Moreover, certain layer-specific TF have a role in this process. Bhlhb5 is selectively expressed in layers II–IV and V and regulates area identity; during embryonic development, it shows a transient high caudomedial to low rostrolateral gradient. It is gradually downmodulated in the postnatal brain to produce a sharp boundary between sensory and caudal motor cortices around P4, and practically disappears at P14. Bhlhb5-null mice show aberrant expression of layer-specific markers and disorganization of vibrissal barrels, and those layer V corticospinal motor neurons of the motor cortex that normally express this TF also show aberrant development (Joshi et al. 2008).

Our picture of arealization mechanisms is still incomplete. Fortunately, considerable research is ongoing to further our understanding of this process. These studies include the contribution of other TF expressed in postmitotic neurons to area specification and how they might coordinate with the action of thalamocortical input, as well as with activity and experience. Unraveling circuit formation in the cerebral cortex will help us to comprehend the precise modes of connections in the cortex and that are altered in many human conditions that affect cognition, from mental retardation to neurodegeneration.