During development, a cerebral cortex appears in the wall of the telencephalic vesicle in reptiles and mammals. It arises from a cell-dense cortical plate, which develops within a primordial preplate. The neurons of the preplate are essential for cortical development; they regulate the neuronal migration of the cortical plate neurons and form the first axonal connections. In the reptilian cortex and in the hippocampus of the mammalian cerebral cortex, most ingrowing afferent axons run above the cortical plate, in the zone where the receptive tufts of apical dendrites of the cortical pyramidal neurons branch extensively. In the mammalian neocortex, however, axons enter mainly from below the cortical plate where they do not encounter the apical tufts of these pyramidal neurons. In this paper, we discuss the idea that this difference in cortical development has relieved a functional constraint in the expansion of the cortex during evolution. We hypothesize that the entrance of axons below the cell-dense cortical plate, together with the inside-out migration of cortical neurons, ensures that the neocortex remains an ‘open’ system, able to differentiate into new (sub)layers and more cortical areas.
The cerebral cortex is a thin sheet of nervous tissue in the telencephalic roof and can be observed in reptiles and mammals. Ontogenetic studies have demonstrated that the cerebral cortex develops from a layer that is densely packed with cells: the cortical plate (Uylings et al., 1990; Bayer and Altman, 1991). This cortical plate is formed within a primordial structure (preplate), which contains the first post-mitotic neurons, here called pioneer neurons (Fig. 1) (Rickmann et al., 1977; Marín-Padilla, 1978,1984). This preplate is subsequently split into a superficial layer (marginal zone or layer I) and an underlying layer (subplate).
Although most studies on cortical development have focused on the neocortex, recent reports demonstrate that these ontogenetic processes also occur in the developing hippocampal formation of the mammalian cerebral cortex (Soriano et al., 1994; Supèr et al., 1998a) and during the development of the cerebral cortex of reptiles (Blanton and Kriegstein, 1991a,b). In the hippocampus, the first neurons to be born reside outside the cortical plate in the marginal zone and subplate layer (Bayer and Altman, 1991; Supèr et al., 1998a). The subsequently produced hippocampal cortical neurons form a cortical plate between the marginal zone and the subplate, following an inside-out sequence in a process similar to the formation of the neocortical plate (Bayer and Altman, 1991; Supèr et al., 1998a). The formation of the dentate gyrus of the hippocampal complex, however, is an exception in the sense that the granular cell layer follows an outside-in pattern of histogenesis (Bayer, 1980). In the cerebral cortex of reptiles, the neurons outside the developing cell-dense cortical plate show more mature morphologies and an earlier peptide expression than the cortical plate neurons. This indicates that they have an earlier birth date than the neurons in the cortical plate (Goffinet, 1983; Blanton and Kriegstein, 1991a,b; Nacher et al., 1996), an observation that has been confirmed by findings on cortical neurogenesis in reptiles (Goffinet et al., 1986). These findings suggest a separate preplate stage in the cortical development of reptiles that occurs prior to the formation of the cell-dense layer. Once this cell-dense layer has been formed, a marginal zone appears above this cortical plate. The origin of the neurons in this zone is not clear, but it could be that they derive from the preplate layer (Blanton and Kriegstein, 1991a,b). The existence of a separate subplate layer in reptilian cortical development, however, has not yet been established. As with the granular cells of the dentate gyrus in mammals, the cortical neurons in reptiles follow an outside-in gradient where the newly generated neurons settle in the cell-dense cortical layer beneath the already settled cortical neurons. Thus, in the neocortex as well as in the hippocampus, and probably also in the reptilian cortex, a cell-dense cortical plate develops within the preplate layer (Fig. 2). These preplate neurons are subsequently split into marginal zone neurons above and subplate neurons below the cortical plate.
At more advanced stages, these developmental layers are renamed (Fig. 2). In reptiles, the cortical plate becomes the cellular layer, which remains a cell-dense layer. The reptilian marginal zone layer is known as the molecular layer. The exact origin of the subcellular layer below the cortical plate in adult reptiles, however, is not clear. In the mammalian hippocampus, the cortical plate remains a cell-dense hippocampal plate, i.e. the pyramidal layer. The marginal zone of the hippocampus becomes the stratum radiatum and stratum lacunosum-moleculare, and the hippocampal subplate becomes the stratum oriens (Soriano et al., 1994; Supèr et al., 1998a). In the neocortex, the cell-dense neocortical plate expands and differentiates into cortical layers II–VI. The marginal zone becomes neocortical layer I, and the subplate transforms into the neocortical layer VIb (Uylings et al., 1990; Reep, 2000) or disappears largely (Kostović and Rakic, 1990; Mrzljak et al., 1990; Allendoerfer and Shatz, 1994).
In addition to the horizontal layering of the cerebral cortex, it is parcellated ‘vertically’. The reptilian cerebral cortex consists of medial, medio-dorsal, dorsal and lateral parts. The medial and medio-dorsal cortices are believed to be homologous to the mammalian hippocampal complex and the lateral cortex to the pyriform cortex. The dorsal reptilian cortex is believed to be homologous to the dorsal neocortex (Medina and Reiner, 2000) and contains several basic cortical areas (e.g. visual, somatic and motor areas). The mammalian hippocampus contains the subicular areas, areas CA1, 2 and 3, hilus and dentate gyrus (Amaral and Insausti, 1990), which can be further divided into a few subfields (Witter et al., 2000). The mammalian neocortex contains sensory and motor areas (primary, secondary) and, in more advanced mammalian species, it also contains so-called associational areas (Felleman and Van Essen, 1991). Many of these neocortical areas can be further subdivided into vertically arranged cortical columns.
Thus, both the reptilian cerebral cortex and the mammalian cerebral cortex (hippocampus and neocortex) derive from the ‘preplate’ layers and the cortical plate layer. However, the reptilian cortex and the mammalian hippocampus (here together referred to as the allocortex) become three-layered cortices whereas the neocortex develops into a six-layered cortical structure. Furthermore, both reptilian and mammalian cortices are functionally segregated into several regions/areas, where the neocortex shows a higher degree of area segregation than the allocortex, especially in anthropoid primates.
Similar differences in developmental patterns between the reptilian cortex/mammalian hippocampus on the one hand and the neocortex on the other, can be observed to have taken place in the course of evolution. The cortex of some reptiles has expanded laterally in the dorsal part (e.g. pleurodite turtles) and the exact cortical area numbers of extant reptiles may vary across species. However, based on variations in cell density and the configuration of the three layers, the reptilian cortex can be divided into regions typical for representative species of all four reptilian orders, i.e. the medial cortex, the dorso-medial cortex, the dorsal cortex and the lateral cortex. Thus, comparison of the cortices from different reptiles indicates that the cortex of reptiles is a relatively ‘conservative’ structure where no large expansion or differentiation of the cell-dense cortical plate occurred (Reiner, 1991; Butler, 1994; Ten Donkelaar, 1998). However, the data regarding area diversification of ‘advanced’ reptiles currently available are far from complete and we cannot exclude the possibility of considerable area diversification in extinct reptiles. As in the reptilian cerebral cortex, the evolution of the mammalian hippocampus is not characterized by a large expansion of the cortical plate. The hippocampal plate remained as one cellular layer without differentiating into several cortical laminae and did not expand laterally into numerous areas, but retained its structure containing about five basic areas — dentate gyrus including the hilus, CA1, 2 and 3, and subiculum (Stephan and Manolescu, 1980; Gall, 1990; Frahm and Zilles, 1994). The mammalian neocortex, in contrast, has evolved rapidly into a relatively large and complex brain structure containing several primary, secondary and association areas, leading to improved information processing. The evolution of the neocortex is reflected in its expanded size and differentiation of the cortical plate (Rakic, 1988,1995; Marin-Padilla, 1992; Butler, 1994; Nieuwenhuys, 1994; Northcutt and Kaas, 1995; Karten, 1997). For example, primitive mammals have few neocortical areas (~10–20), whereas human primates have ~50–70 cortical areas (Felleman and Van Essen, 1991; Krubitzer, 1995; Northcutt and Kaas, 1995; Rakic, 1995).
Thus, the initial allo- and neocortical evolutionary development is characterized by the appearance of a cell-dense cortical layer. Eventually, however, these two structures emerge with very different sets of cortical differentiation and organization, where the neocortical plate shows a remarkable area and layer differentiation compared to the allocortical plate. Still, there is no satisfactory explanation for this evolutionary differentiation between the allocortex and the neocortex (Supèr et al., 1998b).
Hypothesis on Cortical Expansion in Evolution
Here we will hypothesize that two developmental processes, both controlled by pioneer neurons, changed during evolution and allowed the differentiation of the neocortex. The first change is the way in which cortical neurons settle in the cortical plate. During the formation of the cortex, the majority of neurons migrate vertically from the ventricular zone into the cortical plate. In the reptilian cortex, neurons settle in the cortical plate following an outside-in sequence of radial migration, i.e. the first migrated neurons lie in the upper part of the cortical plate (Goffinet et al., 1986). In contrast, in the mammalian cerebral cortex (except the dentate gyrus of the hippocampal formation), neurons follow an inside-out sequence, where newly arrived neurons settle in the upper level of the developing cortical plate, passing the formerly migrated cortical neurons. This vertical (radial) migration of cortical neurons is guided by radial glial cells and is controlled by pioneer neurons (e.g. Cajal–Retzius cells).
The main important change is the location of ingrowth of the main afferent axonal systems into the cortex. In the reptilian cerebral cortex, axons are observed in the marginal zone above the cortical plate (Hall and Ebner, 1965; Ten Donkelaar, 1998). In the hippocampus, afferents grow mainly into the marginal zone and are guided by preplate cells, which are the transient targets for these ingrowing afferents (Supèr and Soriano, 1994; Supèr et al., 1998a). Ingrowing afferents into the neocortex are guided by cells in the subplate (Allendoerfer and Shatz, 1994; Molnár, 2000). Therefore, in the allocortex afferents enter and branch mainly in the marginal zone above the cortical plate, whereas in the neocortex afferents run mainly below the cortical plate, from where they ascend and terminate vertically into the neocortex. We postulate that this difference in axonal distribution pattern, together with the inside-out migration of neurons, may have given the neocortex the potential to expand and differentiate.
The Main Ingrowth of Axonal Fibers
As mentioned above, an obvious difference between the allocortex and the neocortex is the way axonal afferents enter the cortex. In the allocortex, axons enter through the marginal zone, whereas in the neocortex the main afferents enter via the subplate (Fig. 3). This indicates that during evolution the distribution pattern of ingrowing axons into the cortex changed from ‘above’ the cortical plate towards ‘below’ the cortical plate. This change may have occurred gradually, since in some lower mammals relatively many afferents are seen both above and beneath the cortical plate (Valverde and Facal-Valverde, 1986; Ten Donkelaar, 1998). What caused this change of ingrowing afferents during evolution? One of the mechanisms for axonal growth into the cortex is the guidance by efferent axons of pioneer neurons. In the neocortex and hippocampus, neurons in the subplate and marginal zone cells are the first neurons to project outside the cortex and these early projections may guide growing afferents into the cortex (Allendoerfer and Shatz, 1994; Supèr and Soriano, 1994; Supèr et al., 1998a; Soria and Fairen, 2000). Likewise, in the reptilian dorsal cortex, axonal ingrowth may be guided by the marginal zone neurons (Cordery and Molnár, 1999). The specific ingrowth of thalamic axons into the neocortical subplate ‘in vivo’ may be guided by the subplate cells through the ‘handshake’ principle in the internal capsule where thalamic fibers become intermingled with subplate axons (Molnár, 1998). Besides the anatomical evidence for this proposal, genetic data may support the ‘handshake’ principle, since in mutant Tbr1 mice subplate projections are defective and thalamocortical fibers fail to reach the neocortex (Dwyer and O'Leary, 2001; Hevner et al., 2001). In addition, the neocortical subplate cells may not only guide afferents from sub-cortical regions, but also ingrowing cortical afferents by the initial projections to other neocortical structures. These latter roles are observed in the neocortex of more ‘advanced’ mammals, such as domestic cats and ferrets (McConnell et al., 1989). That the subplate became progressively the target zone for developing axons instead of the marginal zone may be reflected by the thickness of the subplate. The subplate is larger in cats than in rodents, and larger in primates than in cats. For example, in humans the subplate develops to approximately six times the thickness of the cortical plate around 29 weeks of gestation (Mrzljak et al., 1990), whereas in rodents it consistently remains a relatively thin layer (Uylings et al., 1990). Furthermore, subplate cells may participate in cortical column formation by their transient connections with particular cortical zones and with layer IV, the main target layer of thalamocortical axons (Ghosh and Shatz, 1992; Allendoerfer and Shatz, 1994). In addition, the ganglionic eminence, a structure that is situated between thalamus and neocortex, may act as an intermediate target for growing corticofugal and thalamocortical axons (Metín and Godement, 1996). Thus, early guidance by subplate cells and guidepost cells in the ganglionic eminence may have rerouted the entrance of growing afferent axons into the neocortex from above the cortical plate towards below the cortical plate in the course of evolution. Recently, Molnár proposed a molecular basis for this evolutionary switch of the entrance of dorsal thalamic fibers into the dorsal cortex (Molnár, 2000).
Once in the subplate, the afferents may have a particular chemoaffinity for molecular cues expressed in specific regions of the neocortex, resulting in a tangential mosaic pattern of afferent fiber termination (Allendoerfer and Shatz, 1994; O'Leary et al., 1994; Rubenstein et al., 1999; Bishop et al., 2000; Mallamaci et al., 2000). In their specific cortical region, the afferent fibers ascend vertically from the subplate into the cell-dense neocortical plate. The precise mechanism that regulates this vertical ingrowth of axons into the cortical plate is unclear, but includes guidance molecules and neural activity (Allendoerfer and Shatz, 1994; Killackey et al., 1995; Sanes and Yamagata, 1999). For example, the transient connections of subplate cells with cortical layer IV, which is the target zone of the thalamic afferents, may be involved in the establishment of the layer-specific connection of the thalamic axons (Ghosh and Shatz, 1992). After reaching their appropriate level in the cortical plate, axonal fibers can branch horizontally into a specific layer without any significant sprouting into other layers (Fig. 3D), and a layer-specific connection can be established. The cortical target layer produces a ‘stop’ signal for the ascending axons and should create a permissive environment for axonal sprouting (Molnár, 1998), whereas the surrounding cortical layers should repel these afferent axons. Chemorepellent factors such as semaphorins are most prominent in the developing cortex and are likely candidates for controlling vertical ingrowth of axonal fibers and layer-specific axonal sprouting (Giger et al., 1996; Sanes and Yamagata, 1999). In conclusion, in the hippocampus and reptilian cortex the afferents are guided by cells in the marginal zone and remain confined to this zone without descending into the cortical plate, whereas in the neocortex afferents ascend from the subplate and grow into the cell-dense cortical plate (Monuki and Walsh, 2000; Skutella and Nitsch, 2001). This may have relieved a functional constraint limiting early expansion of the cortex, as will be discussed below.
Laminar Expansion of the Neocortex
During their radial migration, the cortical neurons have very small and bipolar elongated morphologies (Mrzljak et al., 1990; Uylings et al., 1990; Auladell et al., 1995; Uylings and Delalle, 1997), which may facilitate their way up through the cell-dense cortical plate. After migration, when neurons settle in the cortical plate, they start to mature and a major outgrowth of dendritic and axonal branches occurs. The main neuronal cell type of the cerebral cortex, the pyramidal cell, grows a characteristically apical dendrite towards the pial surface that ramifies abundantly, resulting in an apical tuft in the marginal zone. These neurons maintain their connection with the marginal zone during development and only during late developmental periods do some of the neurons retract their apical dendrite from the marginal zone (O'Leary and Koester, 1993). This suggests that the connection with the marginal zone is needed for the initial maturation, survival and/or processing capacities of these cortical cells.
The cortex has expanded radially in the course of evolution (Stephan and Manolescu, 1980; Hofman, 2001), although this radial expansion is limited compared to the lateral expansion, especially in the neocortex. The radial expansion occurred by adding extra neurons to the cortical plate, which could be the result of tangential migration of non-pyramidal cells from the lateral ganglionic eminence into the cortex (Parnavelas, 2000; Pleasure et al., 2000). However, alterations in the proliferation kinetics in the ventricular zone resulting in an expanded size of the precursor pool seem to be essential for the expansion of the neocortex (Caviness et al., 1995; Kornack and Rakic, 1998). The radial expansion of the neocortical plate in the course of evolution by the addition of extra neurons is supported by the finding that during development the radial migration appears to consist of two waves of cohorts of migrating neurons, one for the lower cortical layers and one for the upper cortical layers (Bayer and Altman, 1991; McConnell, 1995; Frantz and McConnell, 1996). Moreover, Ebner proposed that the dorsal cortex of reptiles lacked the neuronal types found in the superficial layers of the mammalian neocortex (Ebner, 1976). This is supported by immunocytochemical studies from Reiner, who suggested that the neocortex in mammals was characterized by the addition of neurons to layer II–IV in the reptile–mammal lineage (Reiner, 1991).
In addition to the extra neurons, the number of axons in the cerebral cortex also increased during evolution, resulting in a radial expansion of the cortex. In the primate hippocampus, for example, the marginal zone contains more axons than the hippocampus in insectivores (Stephan and Manolescu, 1980) and the primate neocortical subplate contains many more axonal fibers than the subplate of lower mammals (Mrzljak et al., 1990; Ten Donkelaar, 1998). In the neocortex, the increase of axonal fiber numbers is likely to be related to the increase of cortical areas with their feedforward and feedback connections to other cortical areas. Apart from the evolutionary radial expansion of the cortical plate by the addition of more neurons and axons, the cortex also increased in thickness by the formation of extra cortical sub-layers. This, however, occurred in the neocortex and not in the allocortex. This laminar differentiation is related to the growth of axonal afferents into the cortical plate terminating into particular cortical layers. For example, the developing neocortical plate does not differentiate into well-segregated layers when thalamo-cortical fibers are absent (Dehay et al., 1991; Windrem and Finlay, 1991). Thus, in the course of evolution, the cortical plate increased in thickness by the addition of extra neurons and axons. The neocortical plate, however, also differentiated into cortical sublayers, whereas the hippocampal plate and reptilian cerebral cortical plate remained essentially rather undifferentiated cell layers.
We postulate a mechanistic view where the extra addition of both cortical neurons and axonal fibers have been crucial factors in the laminar differentiation of the cortical plate in the course of evolution. When the axon fiber bundles run above the cortical plate, as they do in the hippocampus, the later-generated neurons destined for the upper level of the hippocampal cortical plate may have too limited space to expand radially and to form an additional cortical layer. These neurons are able to migrate through the cell-dense cortical plate due to their elongated and slender shape. However, upon their arrival in the superficial part of the hippocampal cortical plate they are ‘tightly’ enclosed by the axonal strata above them and by the more differentiated lower cortical zones of the hippocampal plate below them. This may result in a limited space for the cell-dense cortical plate to be able to create additional layers, especially during evolution, when extra neurons and axons are added to the cortex. Thus, the increasing numbers of afferent axons in the marginal zone may have been an important limitation for the radial expansion of the cortical plate during evolution (Fig. 3A,C). This may be supported by the finding that the hippocampal marginal zone expanded in the course of evolution due to the increasing number of incoming fibers, whereas radial expansion of the hippocampal plate only happened because cortical neurons dispersed towards the ventricular zone and not by a large radial extension into different laminae (Stephan and Manolescu, 1980; Gall, 1990; Frahm and Zilles, 1994). In the reptilian cortex, cortical neurons follow an outside-in sequence, where late -generated neurons settle in the lower level of the cortical plate. Thus, earlier-generated cortical neurons and axons do not enclose these late-generated cortical neurons, but both are on top of them.
A further important factor in cortical layering is the functional segregation of the different cortical inputs. The laminar differentiation of the cortical plate is linked with the specific ingrowth and termination of axons into a particular cell layer without entering other cortical layers. To establish layer-specific connections, axons must therefore grow into the cortical plate and make contact specifically with a subset of cortical neurons without synapsing onto other neurons. However, in the reptilian cortex as well as in the hippocampus, the axons in the marginal zone are in the same zone as the apical dendrites of almost all cortical pyramidal neurons, which is a main receptive site of the neuron (see Fig. 3). Thus in the allocortex, the physical segregation into discrete axonal-dendritic systems is not feasible. Therefore, axons entering the same zone as the receptive fields of the cortical neurons form a limitation on cortical laminar differentiation of the cell-dense cortical plate of the mammalian hippocampus and reptilian cortex.
In contrast, in the neocortex, the newly arrived neurons that settle in the upper level of the cortical plate are not enclosed between major fiber bundles in the subplate and the expanding lower cortical layers. The ingrowing afferents run below the cortical plate and thus the axons do not have to avoid numerous apical dendrites from cortical neurons. Afferent fibers guided by the subplate cells can ascend vertically into the developing cortical plate after their arrival at the correct areal position (Catalano et al., 1996). Thus, both the radial inside-out migration of cortical neurons and the ascending axons from the subplate follow the same direction of the radial maturation of the cortex. The neocortex therefore remains an ‘open system’ where, during evolution, extra neurons and axons can be added without obstructing the formation of new (sub)layers.
Cajal–Retzius Cells and the Inside-out Genesis of the Neocortical Plate
To develop the cortical plate, the majority of post-mitotic neurons migrate radially outward from the ventricular zone into the telencephalic pallium (Rakic, 1971; Bayer and Altman, 1991; Parnavelas, 2000). During evolution, this cortical migration has become more radially aligned (Rakic, 1974; Goffinet et al., 1986; Ten Donkelaar, 1998) and a better radial alignment has been associated with the inside-out migration pattern (Goffinet et al., 1986; Butler, 1994). Recently, it has become clear that pioneer neurons (in particular Cajal-Retzius cells) play an important role in the control of the radial histogenesis of the cortical plate (Ogawa et al., 1995; Supèr et al., 1997a,2000; Frotscher, 1998; Aboitiz, 1999a). Cajal-Retzius neurons organize the radial glial scaffold (Soriano et al., 1997; Supèr et al., 2000) and support the inside-out order of neuronal migration (Ogawa et al., 1995). Here we will discuss observations suggesting that the Cajal– Retzius cells achieved phenotypes that are more complex during evolution, which may have contributed to the inside-out corticogenesis.
Cajal-Retzius cells are present in the marginal zone throughout the entire cerebral cortex. However, the morphology of the Cajal-Retzius cell is different for different species and cortical areas. In reptiles, Cajal-Retzius cells are simple bipolar horizontal neurons in the outer part of the marginal zone. In the dentate gyrus of the hippocampal formation, which chiefly resembles the reptilian cortex since neurogenesis is prolonged and neurons here settle in the dentate cortical plate following an outside-in pattern of migration, Cajal-Retzius cells also display simple bipolar shapes (Von Haebler et al., 1993) (Mrzljak and Uylings unpublished observations in human brain). The morphology and peptide expression of Cajal-Retzius cells tend to be more complex in the hippocampus proper and even more so in the neocortex (Uylings et al., 1990; Soriano et al., 1994; Berger and Alvarez, 1996; Supèr et al., 1997b). Furthermore, Cajal-Retzius cells have morphologies that are more intricate in the primate neocortex than in, for example, rodents. In addition, they are divided into several subtypes (Marín-Padilla, 1988; Huntley and Jones, 1990; Del Río et al., 1995; Uylings and Delalle, 1997; Meyer et al., 1998,1999). Thus, in the course of evolution, Cajal-Retzius cells may have become more specialized in their contribution to the radial cortical organization, resulting in the inside-out histogenesis of the cortical plate.
Lateral Expansion of the Neocortex by Areal Segregation
The evolutionary increase in cortical size is mainly due to a lateral expansion (Rakic, 1988; Hofman, 2001). This lateral expansion of the cortex is believed to be the result of the enlargement of existing areas by the production of extra cortical neurons, by the segregation of existing areas into additional ones and by further specification of functional cortical domains (Ebbesson, 1984; Rakic, 1988, 1995; Krubitzer et al., 1993; Caviness et al., 1995; Krubitzer, 1995; Northcutt and Kaas, 1995). The proposed evolutionary mechanisms by which new cortical fields are created are related to the progressive number of axonal fiber bundles, where fiber bundles become segregated and terminating axons become spatially restricted by pruning after exuberant growth (Ebbesson, 1984; Krubitzer et al., 1993; Krubitzer, 1995; Innocenti, 1995). These events can be achieved by a further specialization of molecular guiding cues that attract or repel axons and by the segregation of overlapping axonal branches, for example by competitive neuronal activity during development. To create new areas during evolution, the axonal input has to be segregated vertically. We will argue that the development of topographic maps is facilitated when axons enter the cortical plate from below.
Targeting fiber bundles to discrete cortical areas involves the expression of a tangential mosaic pattern of positional cues and neural activity (Catalano and Shatz, 1998). It has been shown that a tangential pattern of molecular cues is expressed in the subplate and marginal zone (Allendoerfer and Shatz, 1994; Soria and Fairen, 2000). In order to reach their appropriate cortical area in the allocortical design, the guided afferent fibers in the marginal zone inevitably cross the fields of the apical terminal dendrites in other cortical regions (Fig. 3A,C). Axons enter the marginal zone branch at several points on their path and can easily form synaptic contacts with the cortical apical dendrites and induce their sprouting (Mattson et al., 1988; Fletcher et al., 1994). Cortical neurons must therefore express a multitude of molecular cues in order to establish and maintain area- and layer-specific axon-dendritic connections. This is inefficient from a molecular point of view, as well as rather implausible, especially since specificity of axonal-dendritic connections depends on competitive interactions. Thus, the parcellation of the cortex into additional areas by the functional segregation of new incoming fiber bundles in the marginal zone would involve a complicated spatio-temporal expression pattern of molecular cues by growing apical dendrites and axons.
In the neocortical design, the main afferent systems guided by subplate neurons enter and run below the cortical plate, where there are no extensive dendritic receptive fields of cortical plate neurons (Fig. 3B,D). In order to reach a particular cortical region, the ingrowing axons can pass below other cortical areas in the subplate without encountering receptive pyramidal cell dendrites belonging to these areas. In this situation, numerous distinct fiber systems can be directed specifically to different cortical regions without having the problem of avoiding a-specific connections. Differential expression of guiding cues in the subplate can specifically target axons and, after reaching their appropriate area, the axons can ascend vertically into the cortical plate and synapse more specifically onto the target cells in their cortical area, as discussed above. This radial arrangement of ascending axons in the neocortex may have permitted the development of multiple instances of reciprocal convergent and divergent cortical pathways, which produces more complex processing. Moreover, the cortical areas can be further divided into functional segregated columns. These columns are formed by the physical segregation of axons due to the elimination of exuberant axonal fiber growth. This occurs under the influence of competitive interactions based on environmental cues and neuronal activity (Jhaveri et al., 1991; Killackey et al., 1995). The physical segregation of axons into vertically defined columns is almost inconceivable in the situation in which all axons run and terminate in the same zone as cortical dendrites (see Fig. 3). Active segregation of cortical areas into distinct columns is facilitated if axons can enter separately from below the cortical plate, via the subplate, as happens in the neocortex. Therefore, we propose that area specification and formation of new cortical domains by axonal segregation of afferents is more limited in the allocortical design than in the neocortical one and thus did not allow the allocortex to expand laterally.
In addition to the proposed potential of cortical differentiation, the entrance of the axonal fibers via the subplate guarantees that the axons travel over a relatively shorter distance in a laterally expanding cortex and provides an immediate benefit in the form of increased processing velocities. The paths that the afferents follow via the marginal zone would become too long in a laterally expanding cortex, especially in the case of a gyrencephalic cortex (Aboitiz, 1999b). Moreover, to keep the increase in cortex size in a relatively compact 3D space (Thompson, 1966), the neocortical expansion in brains >3–4 cm3 is coupled with cortical folding (gyrification) of the cerebral cortex (Hofman, 2001). Recently, Van Essen (Van Essen, 1997) suggested that tension along the axonal afferents under the cortical layers is a driving force for cortical convolution. Therefore, axons running below the cortical plate may have relieved a functional constraint of the major axon bundles running in the marginal zone that limited lateral expansion of the neocortex.
The roles of pioneer neurons in the preplate are essential in cortical development and alterations in the roles of subplate pioneer and marginal zone pioneer cells can lead to different cortical developments. We hypothesize that the shift from the marginal zone to the subplate as the main, initial target for ingrowing afferents, and the change from the outside-in to the inside-out histogenesis of the cortical plate enabled the specific evolution of the mammalian neocortex. Further examination of the development of the differentiation of other laminar structures is necessary to reveal the extent to which our hypothesis is applicable.
We thank Drs Mar Fernández Borja, Michael Frotscher, Michel Hofman, John Parnavelas and Wil Smeets for their helpful comments on an earlier version of the manuscript.