The adult primary motor cortex (area 6) is characterized by a stronger projection to the spinal cord than the primary somatosensory cortex (area 3). Here we have explored the progressive and regressive phenomena that determine these areal differences in the number of corticospinal neurons (CSNs). CSNs were birthdated with [3H]thymidine and subsequently retrogradely labeled from the spinal cord. The time window of CSN production is identical in both areas. The probability that a cohort of neuroblasts project to the spinal cord is indicated by the percentage of [3H]thymidine-positive neurons that can be back-labeled from the spinal cord. In the neonate this fate index is significantly higher in area 6 compared with area 3, indicating that early regionalization of cell fate contributes to areal differences in CSN number. In neonatal reeler mice, an increase in CSN number was accompanied by an increased fate index, showing that the integrity of the post-mitotic environment is required for the specification of the appropriate number of neurons expressing a given connectional phenotype. Postnatal development in reeler and normal is characterized by an area-specific elimination of CSN axons, which reduces areal differences in CSN number. These results show a progressive acquisition of CSN fate in the neocortex and indicate that both early regionalization and late environmental signals contribute to determining areal differences of connectional phenotype.
While there have been considerable advances in our understanding of the determination of the phenotype of cortical neurons, very little is known concerning the developmental mechanisms controlling areal differences in the numerical values of neuronal phenotypes. This is an important issue because the distinctive physiological roles of individual neocortical areas are thought to be largely determined by the patterns of their afferent and efferent connections (Gilbert, 1983; Douglas and Martin, 1991). In the adult sensorimotor cortex of rodents the number of corticospinal neurons (CSNs) is significantly higher in the primary motor cortex, area 6, compared with the primary somatosensory cortex, area 3 (Miller, 1987; Schreyer and Jones, 1988a, b; Oudega et al., 1994). In the present study we have used the areal differences of CSN number in these two areas in the normal and the reeler mutant as a model system for studying the developmental mechanisms that specify neuron number of a precisely defined projection phenotype.
Mechanisms of areal specification of CSN number have been at least partially explored in investigations of the developmental control of cortical differentiation. It is generally agreed that the timing of the specification events of different cortical traits are staggered throughout corticogenesis (Levitt et al., 1993; Eagleson and Levitt, 1999). A number of studies have shown an early regionalization of the neuroepithelium in terms of molecular marker expression (Barbe and Levitt, 1991; Arimatsu et al., 1992; Ferri and Levitt, 1993; Cohen-Tannoudji et al., 1994; Barbe and Levitt, 1995; Tole and Patterson, 1995; Miyashita-Lin et al., 1999; Bishop et al., 2000). However, the notion that molecular phenotype is specified early in development has recently been refined by evidence that molecular fate of cortical neurons is determined progressively during proliferative and post-mitotic phases of neuron production (Arimatsu et al., 1999a, b; Gitton et al., 1999b). In contrast to molecular phenotypes, neuropeptide expression is thought to be specified relatively late (Parnavelas and Cavanagh, 1988; Obst and Wahle, 1995).
There is evidence that the specification of connectional phenotypes like the specification of the molecular phenotype occurs at different stages of corticogenesis (Innocenti, 1981; Schwartz et al., 1991). Transplantation studies show that there is an early regionalization and that newly born neurons are committed to the CSN phenotype shortly after their final mitosis (McConnell, 1988; McConnell and Kaznowski, 1991; Ebrahimi- Gaillard et al., 1994; Ebrahimi-Gaillard and Roger, 1996; Frantz and McConnell, 1996). These findings contrast with previous transplantation experiments showing that the CSN connectional phenotype continues to be specified late in development (Stanfield et al., 1982; O'Leary and Stanfield, 1985, 1986; Stanfield and O'Leary, 1985). Elsewhere, it has been speculated that in general terms, pruning of exuberant connectivity constitutes a late specification event (Innocenti, 1995).
During cortical development many connections are lost, raising the possibility that the adult number and distribution of neurons displaying a given connectional phenotype results from a process of selective elimination of connections (Bates and Killackey, 1984; O'Leary and Stanfield, 1985; Schreyer and Jones, 1988a, b; O'Leary, 1992). This issue has been extensively studied in the corticospinal pathway where, compared with the adult, the immature brain has been shown to have a more widespread distribution of CSNs across the cortex (Bates and Killackey, 1984; O'Leary and Stanfield, 1985). However, in the immature brain the distribution of CSNs is non-uniform so that in the neonate the location of peak densities corresponds to the location of peak densities in the adult (Meissirel et al., 1993; Oudega et al., 1994). This could mean that regressive phenomena are not selective and in which case they would not contribute significantly to determining adult areal differences in CSNs. The alternative is that areal differences in CSN phenotype are established earlier in development as a result of regional differences in CSN number produced by the ventricular zone. This would imply that regions of the ventricular zone that are fated to produce CSNs show local variations in the expression of this fate.
At the present time we do not know the relative contributions of early regionalization and late elimination of connections to the determination of areal differences in CSN number. This is conceptually important because the relative timing of specification events determines which environmental factors can influence cortical development (Kennedy and Dehay, 1993; Levitt et al., 1993). Because connectional phenotype plays such a crucial role in cortical physiology the timing of the specification of these traits is an important issue. One approach to monitoring the specification of connectional phenotype is to study the mitotic history of numerically and ontogenetically defined populations of neurons in the adult.
Such an approach is provided by pulse injections of tritiated thymidine ([3H]thymidine) at regular intervals throughout the generation of CSNs (Polleux et al., 1997a). Precursors that quit the cell cycle immediately after the pulse contain maximum grain counts, correspond to first generation (FG) neurons and can be accurately distinguished from precursors that return to the cell cycle and dilute out their radioactivity (Polleux et al., 1998). CSNs that underwent their final mitosis immediately after the [3H]thymidine pulse (i.e. FG neurons) are double-labeled (DL) and the ratio DL/CSN defines a generation index that accurately indicates the timetable of CSN production (Carter-Dawson and LaVail, 1979; Mustari et al., 1979; Reznikov, 1990; Valverde et al., 1995; Polleux et al., 1997b). The fraction of DL with respect to FG (DL/FG) defines a fate index that reflects the probability that a neuron born on a given date becomes a CSN (Polleux et al., 1998). In the present study, measurement of this fate index reveals that the increased number of CSNs produced in area 6 compared with area 3 is due to a higher proportion of precursors being specified to express the corticospinal phenotype.
It is known that signals at the level of the ventricular zone determine neuronal phenotype (McConnell, 1988; McConnell and Kaznowski, 1991; Ferri and Levitt, 1993; Cohen-Tannoudji et al., 1994; Frantz and McConnell, 1996; Polleux et al., 1998; Arimatsu et al., 1999a, b; Miyashita-Lin et al., 1999; Bishop et al., 2000). The present results show that signals at the ventricular zone also contribute to specifying areal differences in CSN number. However, at this stage we could not exclude the possibility that the cortical post-mitotic environment might also contribute to areal differences in CSN number. To address this issue we have examined the CSN fate index in the perturbed cortical environment of the reeler mutant mouse (Caviness and Sidman, 1973; Caviness, 1976, 1982; Terashima et al., 1983, 1985; Inoue et al., 1991; Hoffart et al., 1995). The reeler returns considerably higher fate indices compared with the normal and this increase in the fate index in the mutant correlates with a large increase in CSN number. This difference between normal and reeler suggests that the cortical environment does contribute to the specification of CSNs and that cell fate in normal development occurs in a two-step fashion in response to signals found in the ventricular zone and in the cortical plate (Arimatsu et al., 1999a, b; Gitton et al., 1999a).
Postnatal development is characterized by an important reduction in cortical connectivity (Cragg, 1975; Finlay et al., 1987; Schreyer and Jones, 1988a, b; Oudega et al., 1994), which could be expected to influence the final number of connections and which importantly could be influenced by evoked neural activity [reviewed in Katz and Shatz (Katz and Shatz, 1996)]. Monitoring postnatal changes in DL/FG provides an efficient and sensitive measure of axon loss. This shows that there is an important postnatal elimination of CSN projections which actually reduces the areal differences in CSN number found transiently in the neonate. The fact that the rate of loss differs in areas 3 and 6 and that it is greatly increased amongst the expanded CSN population of the reeler suggests that the protracted postnatal pruning of connections serves to adjust CSN number to physiological needs.
Materials and Methods
BALB/c mice (Iffa-Credo, Lyon, France) were maintained under a constant photoperiod of 12 h of light (08:00–20:00 h). The offspring of homozygous reeler mutant males and heterozygous females with BALB/c background carrying the Orleans allele of the reeler gene (rlOr) were used. In this study, we used heterozygous littermates (+/rlOr) as controls and refer to these animals as normal. Comparison between hetero- and homozygous animals were made within litters. The morning of the vaginal plug was designated E1.
Pregnant females received an intraperitoneal injection of [3H]methyl thymidine (Amersham, UK, 5 μCi/g body wt, sp. act. 25 Ci/mmol) between 12:00 and 14:00 h on four different dates of embryonic development. The spinal cord was exposed by section of ligaments and the tectorial membrane between the first and the second cervical vertebrae in adults (2–4 months old) under ketamine/xylazine anesthesia (Ketalar 10 mg/kg, Rompun 4 mg/kg) and hypothermia in neonates. The corticospinal tracts, visualized under the dissection microscope, received a large pressure injection of Fluorogold (2.5%) (Schmued and Fallon, 1986) (0.4–0.8 μlin adults; 0.05–0.1 μl in neonates). The wounds were sutured and the animals returned to their cages for a survival period of 5 days in the adults and 2 days in the young pups.
Perfusion and Tissue Processing
After the survival period, all animals received an overdose of sodium pentobarbital and were perfused intracardially with 0.9% NaCl solution including a vasodilator (procaine 1 g/l) followed by a phosphate-buffered (PB; pH 7.4, 0.1 M) fixative solution of 4% paraformaldehyde. The brain was removed and post-fixed in the same solution for 1 week, rinsed in running tap water for 24 h and dehydrated in ethanol, cleared with toluene and embedded in paraffin. The brains were cut in the coronal plan (10 μm) and mounted on glass slides from an albumin–gelatine mixture. Sections were processed for autoradiography as described elsewhere (Polleux et al., 1997a). Spinal cord were cryoprotected in 30% sucrose and sectioned horizontally at 40 μm using cryostat in order to check systematically the size and the position of Fluorogold injection sites (see Fig. 3).
The Fluorogold labeling was observed under epifluorescence (D type filter; Leitz, UV epifluorescence 325–455 nm). After analysis of the fluorescent and autoradiographic labeling, selected sections were counterstained with cresyl violet (0.1%, 10 min), dehydrated and coverslipped with Depex in order to observe with greater precision the autoradiographic labeling over individual cortical cells and to count and measure neuron profiles.
Neuronal Profile Counts
All soma measurements were made using an interactive plotting system (Biocom), which accurately records the position and size of individual retrogradely labeled and unlabeled neurons as well as the number of autoradiographic silver grains per cell nucleus. Measures of the soma size were made on images from a CCD Cohu camera and projected on a Barco CD233 screen. The average cell profile diameter (d) was corrected using the equation described by Schüz and Palm (Schüz and Palm, 1989).
Concerning the autoradiographic labeling, the number of silver grains overlying individual nucleus profile has been corrected for different nuclear profile diameters using the Appleton formula (Appleton et al., 1969) [see also Polleux et al. (Polleux et al., 1997b)]. Autoradiographic background was estimated by counting the number of grains over neocortical tissue which does not present any nuclear profiles. This showed background levels to be inferior to one grain per 400 μm2, i.e. much less than one grain per mean nuclear surface (~250 μm2).
Counts of neuron number per unit area of cortical surface were corrected for split cell error (Abercrombie, 1946) using a cytoarchitectonic analysis based on categorization of cell body diameters (Clarke, 1992) as described earlier (Polleux et al., 1998). Using this procedure, the maximal error attributable to profile splitting is <0.5% for neurons such as CSNs with soma diameters exceeding 16 μm. Statistical analysis of CSN counts was performed using a non-parametric Mann–Whitney test. Finally, glial and endothelial cell profiles were identified using the morphological criteria described by Heumann and Leuba (Heumann and Leuba, 1983) and were excluded from the profile counts.
Estimation of Total Cortical Volume (Vc) and Injection Site Volume in Normal and Reeler Mice
The total cortical volume was estimated in three animals of each genotype and age (newborn and adult) from three different litters using Cavalieri's principle (Cavalieri, 1966). Briefly, the neocortical surface (from midline to pyriform cortex) was measured using an image analysis system (Biocom) on coronal paraffin sections (thickness 10 μm) at 300 μm intervals throughout the entire rostro-caudal extent. Then the following formula derived from the Cavalieri's principle was applied to estimate cortical volume (Vc):
The average volume of Fluorogold injection sites was also estimated using the same approach (Cavalieri, 1966) by measuring the surface occupied by the tracer on 40 μm thick horizontal sections of the spinal cord (spacing 120 μm) of four animals of each genotype and age group.
Exhaustive analysis of the relationship between the intensity of [3H]thymidine labeling of cortical neurons and their radial height in the cortex shows that, provided the Appleton corrections (Appleton et al., 1969) are applied, there is a halving principle of the radioactive signal with each round of mitosis (Polleux et al., 1997a, b). This means that, following a single [3H]thymidine pulse, each generation of labeled neurons can be unambiguously identified. Neurons that show >50% of the maximum labeling have exited the cell cycle at the first mitosis following the [3H]thymidine pulse and are first generation neurons (FG neurons) (Rakic, 1973; Polleux et al., 1997a, b). The timing of the [3H]thymidine pulse that generates a cohort of FG neurons defines the birthdate of the FG neurons. Neurons that, following the pulse, complete two or more divisions before quitting the cycle have <50% maximum labeling and are subsequent generation neurons.
Statistical analysis of the two ratios DL/CSN and DL/FG was performed using a chi-square analysis as described earlier (Polleux et al., 1997b, 1998). Briefly, variability in proportions among all animals of a given genotype or among all values within a given area (intra-group variability) was systematically tested, and failed to show any statistical significance. Then inter-group variability (between normal and reeler animals or between areas 6 and 3) was analyzed on pooled values obtained from different animals (see text for numbers). Throughout this study, DL/CSN and DL/FG values were obtained by examining 300–600 neurons in three or four animals at each given time point.
In the present study we have used [3H]thymidine pulse injections to investigate the developmental processes that specify CSN number. This approach can be used to describe accurately the mitotic history of specific subsets of neurons (Polleux et al., 1997b, 1998). There are numerous technical difficulties associated with the autoradiographic process and the quantitative use of [3H]thymidine labeling which globally result in an underestimation of FG labeling (Polleux et al., 1997a). However, because in the present study we are examining ratios of FG neurons, the underestimation of the autoradiographic signal can not be expected to influence our principal conclusions.
The percentage of FG neurons within a given population of adult neurons defines a generation rate which quantifies the rate of neuron production of that population (Carter-Dawson and LaVail, 1979; Mustari et al., 1979; Reznikov, 1990; Vaccarino et al., 1995; Valverde et al., 1995; Polleux et al., 1997b) (see Fig. 1A). The generation rate of the CSN population is determined by the number of CSNs that are FG neurons (i.e. double-labeled neurons — DL) and expresses this number as a percentage of the total CSN population. DL/CSN describes the relative rate at which the daughter cells of neuroblasts undergoing their final mitosis are incorporated into the CSN population. More precisely, changes in DL/CSN reflect the changes in labeling index (LI, i.e. percentages of [3H]thymidinelabeled precursors) in the ventricular zone and in frequencies of differentiative divisions of the cohort of precursors that generate the CSN population (Schultze et al., 1974; Schultze and Korr, 1981; Miller, 1988; Reznikov, 1990; Polleux et al., 1997b). Given the constancy of S-phase, variation of the LI in the ventricular zone reflects changes in the duration of the cell cycle (Kaufmann, 1968; Waechter and Jaensch, 1972; Hoshino et al., 1973; Schmahl, 1983). Because differences in numbers of precursors do not contribute to variations of DL/CSN, this means that variation of this ratio reflects changes in rates of neuron production due to modulation of cell-cycle parameters (cell-cycle duration and mode of division) (Polleux et al., 1997b). Importantly, DL/CSN is independent of the size of the precursor pool.
The probability that a neuron born on a specific date forms and maintains a projection to the spinal cord is determined by the ratio of DL/FG neurons, which is defined as the fate index (Polleux et al., 1998) (see Fig. 1A,B).
Areal and Laminar Boundaries in the Neonate
Cytochrome oxidase histochemistry in the post-natal day (PND) 6 neonate reveals a well defined primary sensorimotor cortex (Fig. 2A). At this age the barrel field has clearly defined borders and individual barrels are present. The present study critically requires that the counts of birthdated neurons are accurately performed in areas 3 and 6. This requires a method of positive identification of areas 3 and 6 in the neonates and adults in 10 μm thick paraffin sections (see Materials and Methods). The location of areal boundaries on such sections is achieved by tracing out the profiles of all neurons encountered in a 250 μm wide stripe of cortex using the image analysis system (see Methods). This procedure reveals the laminar organization of the cortex and shows that areas 3 and 6 in the neonate can be reliably distinguished as shown in Figure 2B,C. While some indication of areal identity can be obtained from observation of pooled neuronal profiles (far left panel, Fig. 2B,C), areal and laminar localization is further enhanced by allocating profiles to one of six categories according to soma size (right-hand side panels in Fig. 2B,C). This shows that in the neonate, area 3 has a well developed layer 4, which contrasts with a very thin layer 4 in area 6 (compare panels 2 in Fig. 2B,C) (Morin and Beaulieu, 1994). The large pyramidal neurons of layer 5 are more numerous in area 6 than in area 3 (compare section 5 in Fig. 2B,C). This procedure enabled us to make our analysis of labeling in the neonate in identified areas which validates the inter-areal comparisons.
Single and double-labeled neurons can be readily identified in the 10 μm thick sections by making simultaneous observations under fluorescent- and bright field illumination (Fig. 3C). In the present experiments, the exposure time for the autoradio- graphic procedure was chosen in order to standardize maximum grain counts at ~80 per profile (Fig. 3C). Elsewhere we have shown that neurons with >50% maximum labeling correspond to the first generation (FG) neurons leaving the cell cycle only one division following the [3H]thymidine pulse (Polleux et al., 1997a, b). Thus this double labeling method allows the identification of CSNs that have undergone their final division on the date of the [3H]thymidine pulse (Polleux et al., 1998) (double-labeled neurons DL, see Fig. 3C).
Early Areal Differences of CSN Density
Injections performed at PND4 reveal a CSN distribution stretching from the medial limit of the cingulate cortex across to the pyriform cortex (Fig. 4A). High labeled cell densities as well as the immature widespread distribution are maintained at PND6 (Fig. 4B). Following injection at PND8, however, CSNs in the cingulate cortex and in the lateral part of area 3 are less numerous than at PND6, and in the adult these regions are devoid of CSNs (Fig. 4C,D). These observations suggest that the peak of corticospinal projection occurs around PND6, in line with earlier studies in rat (Bates and Killackey, 1984; O'Leary and Stanfield, 1985; Joosten et al., 1987; Schreyer and Jones, 1988a, b; Joosten and van Eden, 1989; Oudega et al., 1994) and mouse (Uematsu et al., 1996; Gianino et al., 1999). Because PND6 injections are performed after the completion of target invasion and prior to noticeable elimination of connections, they are expected to label optimally the CSN population.
Our quantitative analysis shows that in normal mouse cortex at PND6, the areal density of CSNs in area 6 is nearly twice that in area 3 (Fig. 5A,B E;, P < 0.001, Mann–Whitney test). During late postnatal development there is a reduction of areal density that is non-significantly different between area 6 (–41%) compared with area 3 (–27%) (Fig. 5E). These results show that there are significant areal differences in CSN number in the neonate. The postnatal decrease in the CSN density in normal and reeler is greater in area 6 than in area 3 (Fig. 5, see also Fig. 7). This suggests that there is an area-specific decrease in CSN density. However, it needs to be pointed out that this decrease reduces areal differences and does not create them.
The differences of CSN density between area 6 and 3 are conserved in the reeler mouse newborn (P < 0.001, Mann– Whitney test) and adult cortex (P < 0.05, Mann–Whitney test) (Fig. 5C–E). Importantly, the density of CSNs in reeler newborn is largely increased compared with normal both in area 6 (40% increase; P < 0.001, Mann–Whitney test) and in area 3 (52% increase; P < 0.001, Mann–Whitney test).
The measurement of total cortical volume in normal newborn (mean ± SD: 18.56 ± 0.41 mm3; n = 3) and reeler newborn (18.20 ± 2.27; n = 3) revealed no statistically significant difference according to an ANOVA analysis (PLSD Fisher test, P = 0.883). Therefore, the observed differences in CSN densities between normal and reeler cortex are not the reflection of differences of total cortical volume that would artefactually affect the estimation of cell packing density.
Finally, in order to rule out possible differences in the size and position of Fluorogold injections, we systematically checked the spinal cord injection sites and estimated the volume occupied by the tracer using Cavalieri's principle (Cavalieri, 1966) applied to serial sections (see Materials and Methods; Fig. 3A). We found no statistical differences in the average injection volumes between normal and reeler in the newborn (n = 8, P > 0.27, Mann–Whitney test; Fig. 3B) or in the adult injections (n = 8, P > 0.50, Mann–Whitney test; Fig. 3B).
Areal Timetable of CSN Generation — Normal Cortex
Since postnatal regressive phenomena are not responsible for establishing adult CSN areal differences in density an alternative possibility is that areal differences are the consequence of CSNs being generated over a longer time period in area 6 than they are in area 3. To investigate this issue we have examined the incorporation of FG neurons (i.e. neurons that underwent their final round of mitosis shortly after the [3H]thymidine pulse) into the CSN population.
Successive injections of [3H]thymidine reveal the inside-first outside-last histogenic gradient of cortical layers (Fig. 6A). Quantification of the percentage of DL neurons with respect to the population of CSNs defines the generation rate (Carter-Dawson and LaVail, 1979; Mustari et al., 1979; Reznikov, 1990; Valverde et al., 1995; Polleux et al., 1997b) (Fig. 6B). Although DL/CSN is expected to be immune to developmental regressive events (since cell death or axon elimination will influence identically both terms of the fraction), this index needs to be defined in the neonate, since a transient population of CSNs generated at either early or late stages of development can not be excluded. FG neurons returned from the E13.5 pulse are concentrated in layer 6b and a few are found at the bottom of layer 5. At this stage in both areas ~3% of the CSN population are FG neurons, indicating that generation of the CSN population is initiated simultaneously in both areas. By E14.5, there is an increase in the generation rate of CSNs and this rate of increase is greater in area 3 than in area 6. Hence, while the overall numbers of CSNs are lower in area 3 compared with area 6, pulse injections of [3H]thymidine at E14.5 double-labeled larger numbers of CSNs in 3 compared with 6. At E15.5 CSN production in area 6 is greatly increased compared with area 3. This illustrates the use of the generation index to describe the timetable of CSN production.
By E16 the production of CSNs is terminated in both areas. These results show negligible differences in the profiles of CSN generation rate calculated in the adult and neonate, confirming that regressive phenomena do not influence the generation index (Polleux et al., 1998). This means that the adult values obtained with eight different time points (E12.5, E13.5, E14, E14.5, E15.5, E16, E16.5, E17.5) provide a high temporal resolution of the timetable of CSN production (Polleux et al., 1998). The important finding is that both the onset and duration of CSN generation is identical in the two areas.
This result provides further evidence that the timing of the final mitosis is an important constraint in the determination of cortical neuron phenotype (Caviness and Sidman, 1973; Caviness, 1982; McConnell, 1988; McConnell and Kaznowski, 1991; Arimatsu et al., 1992, 1994; Frantz and McConnell, 1996; Polleux et al., 1998). However, a single [3H]thymidine pulse leads to a relatively wide vertical spread of FG neurons across several cortical layers (Fig. 6A). At later stages of corticogenesis occasional FG neurons take up a deep position and we have investigated whether those late generated neurons that occupy layer 5b can form projections to the spinal cord albeit with a reduced frequency. We examined large numbers of FG neurons resulting from a [3H]thymidine pulse on E16 and which occupy layer 5b in the neonate. This showed that there was 0% doublelabeling in both areas (n = 163 FG neurons, 27 sections). Together these results show that timing of the final mitosis is critically important for the specification of cortical neuron phenotype.
Areal Differences in Fate Index — Normal Cortex
The percentage of FG neurons that are retrogradely labeled from the spinal cord (fate index — see Fig. 1) indicates the probability of a neuroblast of a given birthdate acquiring the CSN phenotype. In Figure 6C the fate index for each time point is calculated from a minimum of 12 sections taken from two or three different pups (two pups from one litter injected at E13.5 and E16 and three pups taken randomly from two different litters injected at E14.5). At E13.5 both areas return identical fate indices (χ2 = 0.41; d.f. = 1, P > 0.50). At E14.5 there is a substantial increase in the fate indices in both areas and importantly the rate of increase in fate index is substantially greater in area 6 compared with area 3 (χ2 = 8.23; d.f. = 1, P < 0.01). This result indicates that at E14.5 the probability that a neuroblast undergoing its final round of mitosis acquires the CSN phenotype is significantly greater in area 6 compared with area 3 (Fig. 6C). These areal differences in DL/FG are not likely to be the expression of a mitogenic gradient because elsewhere we have shown that the areal differences in cell-cycle characteristics are not part of a continuum but are area-restricted (Polleux et al., 1997a, b).
DL/FG is sensitive to late elimination of corticospinal axons by pruning (Bates and Killackey, 1984; O'Leary and Stanfield, 1985; Joosten et al., 1987; O'Brien et al., 1987; Joosten and van Eden, 1989; O'Leary and Stanfield, 1989; Oudega et al., 1994). Postnatal decrease in DL/FG values reflect the extent of axon elimination, making it possible to examine whether selective elimination of CSN axons contributes to the adult differences in CSN number. In fact the postnatal decrease in DL/FG is higher (51%) in area 6 than in area 3 (10%) showing that this regressive factor, although area specific, actually reduces the initial areal differences in CSN number observed in the adult (Fig. 7B).
Influence of the Cortical Environment on CSN Fate — DL/FG in Reeler Cortex
So far the results show that there is a regionalization of cell fate that contributes to the specification of areal differences in CSN number. Does the commitment to a CSN fate take place entirely in the ventricular zone during the final mitosis? If not, does the cortical environment also contribute to the determination of CSN phenotype? Because the reeler shows a highly disorganized cortex (Caviness and Sidman, 1973; Caviness, 1977; Caviness and Rakic, 1978; Terashima et al., 1983, 1985; Inoue et al., 1991), the difference in cell fate index between normal and reeler would indicate that the cortical environment does contribute to cell-fate decisions. Note that because layering is absent in the reeler mutant the counts of DL/FG used to compare normal and reeler neonate are based on FG neurons throughout the full thickness of the cortex. This shows that the nearly 40–50% increase in CSN number in reeler compared with normal neonates (Fig. 6) is accompanied by a 55–58% increase in DL/FG ratio both in area 6 (χ2 = 7.21; d.f. = 1, P < 0.01; Fig. 7A) and in area 3 (χ2 = 8.43; d.f. = 1, P < 0.01; Fig. 7A).
The massive increase in the CSN number and fate index in the reeler mutant provides a challenge to the postnatal pruning of CSN axons that occurs during normal development. In the reeler, the postnatal decrease in DL/FG is area-specific as in the normal, but is considerably higher than in the normal (–22% in reeler area 3 versus –10% in normal area 3 and –84% in reeler area 6 versus –51% in normal area 6; Fig. 7B,C), suggesting that axon elimination is a developmental adjustment that serves to counterbalance excess generation of CSNs.
Significance of Indices with Respect to Cell-cycle Parameters
DL/CSN and DL/FG provide complementary information on the mitotic history of the population of CSN precursors. Variations of the generation index (DL/CSN ratio) mainly reflect variations of the dynamics of cell-cycle parameters underlying the production of a given phenotype (Polleux et al., 1997b, 1998). In the present study, we have restricted the use of DL/CSN to estimate quantitatively the timetable of CSN production. In contrast, changes in DL/FG are not related to changes in the kinetics of proliferation. At early stages of postnatal development, prior to the phase of axon elimination, DL/FG reflects the fraction of neuroblasts born on a given date that acquires a CSN phenotype and can be thought of as a probability factor governing cell fate (Polleux et al., 1998). During the period of peak production of layer 5b, DL/FG is ~30% higher in area 6 compared with area 3.
Cell Fate and Axon Elimination — Normal Cortex
The present results show that during normal development there are important areal differences in cell fate that correlate with differences in CSN number. This indicates that regionalization of cell fate contributes to determining the inter-areal differences in CSN phenotype in the adult. McConnell and Kaznowski challenged laminar identity by performing heterochronic grafts (McConnell and Kaznowski 1991). Neuroblasts which were destined to migrate to layer 5 continued to do so when grafted into an older host which is generating supragranular layers. This work demonstrated that signals encountered in the ventricular zone during S-phase play a role in cell-fate determination, suggesting that an important aspect of cell fate is determined at the final mitosis.
DL/FG is exquisitely sensitive to loss of CSN axons and so during postnatal development can be used to monitor refinement of the CSN pathway. During normal postnatal development we found that the decrease in DL/FG was significantly different in the two somatomotor areas — whereas there is a 51% reduction in DL/FG in area 6 there was only a 10% reduction in area 3. These findings show that although the magnitude of axon elimination is area-specific, this phenomenon does not contribute to creating the areal difference in CSN number. Previous studies have shown regional elimination of axon collaterals (O'Leary et al., 1981; O'Leary and Stanfield, 1985; O'Leary and Terashima, 1988) and here we show that it can be area-specific. Because the rate of axon elimination is different in neighboring areas 3 and 6, and because it reduces the areal differences, it would seem that axon elimination could serve to finely tune numbers of projection neurons within a given cortical area.
Cell Fate and Axon Elimination — Reeler Cortex
The reeler mutant provides an alternative experimental paradigm to the heterochronic transplants used by McConnell for challenging cell fate during corticogenesis. The possibility that the environment of the cortex contributes to the specification of phenotype is directly addressed by birthdating experiments in the reeler (Caviness and Sidman, 1973; Caviness, 1982; Polleux et al., 1998). These experiments show that in the absence of cortical lamination, the timing of the final mitosis continues to determine the CSN phenotype as stringently as in the normal cortex (Polleux et al., 1998). The importance of the timing of the final mitosis, however, has to be reconciled with the fact that the radial distribution of FG neurons in the normal cortex from a pulse at E14.5 is considerably wider than layer 5b, suggesting either that the ventricular zone at a given moment is generating a range of phenotypes destined for different layers (Price and Thurlow, 1988; Fishell et al., 1990; Walsh and Cepko, 1992; Luskin et al., 1993; Kornack and Rakic, 1995) or that FG neurons that do not terminate in layer 5b are inhibited from forming a projection to the spinal cord. This last possibility would mean that the cortical environment contributes to specifying the CSN phenotype.
If the cortical environment contributes to specifying CSN fate then one would predict that CSN number should be modified in the perturbed environment of the reeler mutant. This in fact is the case, the increase in CSNs is accompanied by an increase in DL/FG, indicating that the increase in CSNs in the mutant is the result of a fate change and is not due to some other mechanisms such as an increased survival or reduced elimination. Although the contribution of the cortical environment to the specification of cell fate is clearly established by the present results in the reeler, we do not know if the cortex exerts its influence postmigratory, or whether it signals down to the ventricular zone and influences cell fate during mitosis. Such a possibility would be in line with studies suggesting that post-migratory cortical neurons exert a feedback control on neurogenesis as has been suggested in the retina (Reh, 1987; McConnell and Kaznowski, 1991; Gillies and Price, 1993; Polleux et al., 1998).
If the normal cortical environment outside of layer 5b inhibits formation of projections to the spinal cord (consistent with the fact that FG neurons resulting from a pulse injection of [3H]thymidine at E14.5 have a wider radial distribution than the CSN population) then the removal of this inhibition in the non-laminated cortex found in reeler could lead to the observed increase of DL/FG values found in both areas of the mutant. In normal cortex, the present finding that the extremely rare late FG neurons that come to occupy layer 5b fail to project to the spinal cord, even at a very low frequencies, shows that whereas the timing of the final mitosis is a necessary factor governing cell fate, laminar location is not a sufficient one. This result shows that birthdates discriminate between subclasses of layer V neurons.
The fact that the final cellular environment can contribute to the specification of neuronal connectivity is supported by recent results showing that in the developing spinal cord, the appropriate specification of the phenotype of late-generated motor neurons pool depends on the interaction they perform while migrating through previously generated motor neurons (Sockanathan and Jessell, 1998). Our results in the reeler are compatible with the fact that early-born neurons can regulate the expression of the connectional phenotype of later-generated neurons through cell–cell interactions.
Theories of Areal Specification and Numerical Control of Connectional Phenotype
The present study sets out to explore if the connectional phenotype of cortical neurons is specified according to similar mechanisms as other phenotypic traits such as the neurotransmitter or the laminar phenotype. Recently, it has been shown that in the developing spinal cord, the connectivity of motor neurons is specified early at the level of dividing progenitors by the expression of specific subsets of transcription factors by inductive cues that restrict the competence of precursors to produce a limited set of neuronal phenotypes [reviewed by Edlund and Jessell (Edlund and Jessell, 1999)]. However, it remains to be determined if the developmental mechanisms governing the acquisition of neuronal phenotype in the spinal cord can be applied to higher order, laminated structures of the forebrain such as the cerebral cortex.
There are two long-standing views of cortical development that are relevant to understanding area-specific connectional fate. Protomap theory views the neuroepithelium as being regionalized prior to the arrival of the thalamic afferents (Rakic, 1988). Protocortex theory views the early formed cortex as having a wide developmental potential (O'Leary, 1989). Two recent studies using in vitro approaches were able to show that while there is early regionalization of molecular markers expression prior to arrival of thalamic afferents, environmental signals continue to contribute to the acquisition of the molecular phenotype in the post-mitotic population (Arimatsu et al., 1999a, b; Gitton et al., 1999b). In both studies late environmental signals were found to influence the extent that the early specified molecular phenotype is expressed. The studies of Arimatsu et al. and Gitton et al. (Arimatsu et al., 1999a, b; Gitton et al., 1999b) are conceptually important on several accounts. Firstly, they reconcile protomap and protocortex theories and provide the conceptual basis required by the fact that there is extensive mixing of cortical precursors (Walsh and Cepko, 1992; Fishell et al., 1993; O'Rourke et al., 1995; Tan et al., 1995) and an early regionalization of the cortex at least with regard to the expression of molecular markers (Barbe and Levitt, 1991; Arimatsu et al., 1992; Cohen-Tannoudji et al., 1994; Miyashita-Lin et al., 1999; Gitton et al., 1999a; Bishop et al., 2000). The present results provide in vivo evidence for a similar progressive specification of a cardinal feature of neuronal identity, i.e. the efferent connectivity. The areal differences in CSN cell fate at E14.5 bare witness to early regionalization prior to the arrival of thalamic afferents (E14.5). This early regionalization of cell fate can be challenged by the perturbed cortical environment of the reeler, which the present study shows, leads to a radical increase in CSN fate index.
The progressive acquisition of CSN phenotype reconciles apparently contradictory findings in the literature. Fetal heterotopic grafting experiments of somatosensory cortex originally were thought to show that regional locus in the host determined the connectional CSN phenotype (O'Leary and Stanfield, 1989). However, quantification of results showed that heterotopically transplanted neurons retain some of the connectional features corresponding to their site of origin (Ebrahimi-Gaillard et al., 1994; Ebrahimi-Gaillard and Roger, 1996). One could expect that the environmental signals in the heterotopic graft would lead to reduced levels of acquisition of the phenotype so that while it remains true that site of origin determines the connectional phenotype, signals in the final location will modulate the expression of this phenotype.
The large excess of CSNs in the neonatal reeler allows us to investigate the control mechanism underlying the postnatal elimination of axons from the corticospinal pathway. The postnatal decrease of DL/FG is greatly increased in reeler compared with normal and the decrease is area-specific, being significantly larger in area 6 compared with area 3. One consequence of this is that the final numbers of CSNs in adult reeler and normal mice are actually very similar (Fig. 5E). These findings suggest that the elimination of axons during postnatal development reflects an adaptive process that allows the numerical value of CSN in different areas of the cortex to be adjusted to the physiological needs of the developing organism. We do not exclude that cell death can also play a role in the acquisition of area-specific number of CSNs. In fact, the role of cell death has been already well documented in the control of cell number elsewhere in the nervous system (Williams and Herrup, 1988; Finlay and Pallas, 1989; Finlay, 1992). However, our results in the normal and in the reeler show that the final numbers of CSNs are specified by an early two-step cell-fate determination as well as by late elimination of corticospinal axons during postnatal development.
Our results underscore that laminar and areal specification are interrelated (Eagleson et al., 1997; Polleux et al., 1997a, b). They provide in vivo evidence of a progressive specification of connectional phenotype due to signals originating from the ventricular zone, cortical plate and maturing cortex. In this way they suggest that connectional phenotype is determined by early regionalization of the neuroepithelium as well as environmental signals in the post-migratory cortex (Jensen and Killackey, 1984; McConnell, 1988; Rakic, 1988; O'Leary, 1989; O'Leary and Stanfield, 1989; Barbe and Levitt, 1991; McConnell and Kaznowski, 1991; Arimatsu et al., 1992, 1999a, Arimatsu et al.b; Ferri and Levitt, 1993; Gillies and Price, 1993; Gitton et al., 1999a, b). While the mitotic history of CSNs shows that adult values of the number of neurons expressing a given connectional phenotype are achieved by multiple events occurring at the level of the ventricular zone, the cortical plate and in the maturing cortex, there are a number of key events that remain unclear. For example, it is not known whether the component of cell-fate control that is cortical in origin is pre-migratory or postmigratory [as in the studies of Arimatsu et al. and Gitton et al. (Arimatsu et al., 1999a, b; Gitton et al., 1999b)] or both.
We are grateful to Pascale Giroud and Christelle Merrouche for excellent technical assistance and to Ghislaine Clain for animal care. Financial support from Region Rhône Alpes 97021178, EC QLG3-2000-00158 and from European Economic Community Grant BIOMED BMH4 CT961604.