The mammalian neocortex comprises two major neuronal subtypes; interneurons derived from the ganglionic eminence (GE) and projection neurons from the cortical ventricular zone (VZ). These separate origins necessitate distinct pathways of migration. Using mouse genetics and embryonic forebrain slice culture assays, we sought to identify substrates and/or guidance molecules for nonradial cell migration (NRCM). Mice carrying a mutation in Pax6 (Sey−/−), a paired domain transcription factor, are reported to have increased numbers of cortical inhibitory interneurons, suggesting that Pax6 could induce inhibitors of interneuron development or alternatively play a repressive role in guiding NRCM and/or specifying interneurons. Unexpectedly, we found a cell nonautonomous reduction in the distance Sey−/− neurons migrated, reflecting a disorganized migration, with frequent changes in direction. In contrast, no difference in the number of nonradially migrating GE cells was observed in Sey−/− mice. Our data indicate that the increased numbers of interneurons observed in Sey−/− do not result from an increased rate or number of nonradially migrating cells; instead, loss of Pax6 results in the ectopic specification of interneurons in the cortical VZ. Further, our data indicate that the known axonal disorganization in Sey−/− mice contributes to the observed reduced distance of NRCM.
The mammalian neocortex consists of approximately 80% excitatory projection neurons, the remainder being inhibitory interneurons (Xu et al. 2004). In rodents, these 2 neuronal classes are derived from distinct progenitor zones. Cortical interneurons are predominantly specified ventrally and migrate nonradially to reach the cortical plate (CP), whereas projection neurons migrate radially from their dorsal origin in the cortical ventricular zone (VZ) (Parnavelas 2000; Marin and Rubenstein 2001; Nery et al. 2002; Stuhmer, Puelles, et al. 2002). Thus, normal cortical development requires integration of neuronal populations migrating from distinct sites.
Discrete signaling events in the dorsal and ventral forebrain lead to the generation of projection neurons and interneurons, respectively (Campbell 2005). Transcription factors play essential roles in cell specification and, somewhat surprisingly, in cell migration. For example, loss of Dlx1/2 or Nkx2.1, transcription factors with ventrally restricted domains of expression, results in cell autonomous defects in nonradial cell migration (NRCM) and interneuron specification (Anderson et al. 1997; Sussel et al. 1999). Deficits in NRCM are also observed with loss of the dorsally expressed Emx1/2 genes, although this migration phenotype is cell nonautonomous and may be related to a primary axon defect (Shinozaki et al. 2002). In contrast to these mutations that result in interneuron deficits, loss of Pax6 is the only mutation reported to cause an increase in the migration and number of cortical interneurons (Chapouton et al. 1999).
Pax6 is a paired-type homeodomain transcription factor that is highly expressed in the dorsal–rostral telencephalic neuroepithelium (Callaerts et al. 1997). It plays an essential role in dorsal–ventral patterning (Toresson et al. 2000), cell migration (Schmahl et al. 1993; Caric et al. 1997; Chapouton et al. 1999; Nomura and Osumi 2004), and cortical arealization (Bishop et al. 2000). Sey−/− mice have a point mutation in Pax6 resulting in a nonfunctional truncated protein (Hill et al. 1991). These mice exhibit increased interneuron numbers and subcortical heterotopia containing interneurons. The pathogenesis of these heterotopia is unknown but may reflect ectopic interneuron specification in the cortical VZ (Kroll and O'Leary 2005) and/or increased NRCM (Chapouton et al. 1999), though the latter hypothesis was not directly tested. Furthermore, the relative contribution of NRCM to the increased numbers of cortical interneurons remains unclear, with reports of contradictory findings (Chapouton et al. 1999; Jimenez et al. 2002).
Based on these data, we hypothesized that Pax6 regulates the expression of permissive or guiding molecules for NRCM. In addition, since loss of Pax6 causes a dorsal expansion of ventral transcription factors (Toresson et al. 2000), this hypothesis predicts an increase in medial ganglionic eminence (MGE) cell migration into the Sey−/− neocortex. Here we report disorganized NRCM and a cell nonautonomous defect in the distance neurons migrate nonradially in Sey−/− mice but no change in the rate of NRCM or in the number of MGE cells specified to migrate into the cortex. Instead, we find that the increase in numbers of cortical interneurons reflects ectopic specification of these cells in the ventrolateral cortical VZ.
Materials and Methods
Mouse Strains and Genotyping
Sey mice were obtained from Dr K. Campbell (Children's Hospital Medical Center, Cincinnati, OH) and maintained on a C57BL/6 background. Timed-pregnant mice were considered embryonic day 0.5 (E0.5) on the morning a vaginal plug was identified. All embryos were morphologically staged (Theiler 1989) and genotyped. Genotyping of Sey mice was performed by polymerase chain reaction (primers: 5′-GCAGATTACCCAGTCCTCGGAGTT-3′; 5′-TCACCTTTCTCCAGAGCCTCAATC-3′) and DdeI digest, as previously described (Grindley et al. 1995). For transplantation experiments (see below), Sey+/− mice were crossed with a transgenic line that constitutively expresses GFP in all tissues (kindly provided by Dr A. Nagy, Samuel Lunenfeld Research Institute, Toronto, Canada) to generate a line of Sey+/−; GFP mice. The institution's animal care and use committee approved all animal breeding, handling, and experimental procedures.
Embryonic Slice Cultures
Sey+/− × Sey+/− matings were used to generate Sey+/+ (wild type) and Sey−/− littermate embryos. E14.5 embryonic brains were dissected and embedded in 4% low–melting point agarose for coronal sectioning on a vibrating microtome (250 μm; LeicaVT1000S vibrotome; Leica Microsystems, Nussloch, Germany). The brain slices were transferred to Millicell-CM membranes (Millipore Corporation, Billerica, MA) precoated with laminin (10 μg/mL; BD Biosciences, San Jose, CA) and poly-L-lysine (10 μg/mL; Sigma, St Louis, MO). A DiI crystal (Molecular Probes, Eugene, OR) was implanted in the ganglionic eminence (GE) of each slice, as previously described (McManus, Nasrallah, Pancoast, et al. 2004). The cultures were incubated in 1:1 DMEM:F12 (Invitrogen Corporation, Carlsbad, CA) with 10% fetal bovine serum (Invitrogen), 1 mM penicillin/streptomycin (Invitrogen), and 6.5 mg/mL glucose for 1 h and then transferred to DMEM plus N2 supplement (1:50, Invitrogen), 1 mM penicillin/streptomycin, and 6.5 mg/mL glucose. Slices were grown in culture for 2 days and then fixed in 4% paraformaldehyde. Each slice was analyzed for the straight-line distance traveled by nonradially migrating cells from the DiI crystal into the cortex using ImagePro software (Media Cybernetics Bethesda, MD), as previously described (McManus, Nasrallah, Pancoast, et al. 2004). Time-lapse imaging of DiI-labeled cells in live slices to measure the migration rate of individual nonradially migrating cells was performed as previously described (Nasrallah et al. 2006). All migration analyses were conducted without knowledge of the animal's genotype. Comparisons were made using a 2-tailed Student's t-test.
Angle Analysis of Cell Migration
Using ImagePro software, a line approximating the orientation of each nonradially migrating cell was determined for each time point by placing one point (x1, y1) in the center of the cell soma and a second point (x2, y2) near the origin of the leading process. The slope of the line of orientation was calculated for each movie frame using these 2 points (slope, m = (y2 − y1)/(x2 − x1)). The change in the orientation of the cell (angle, Δθ) between successive movie frames (10-min intervals) was calculated using θ = arctan [(m2 − m1)/(1 + m1m2)]. After observing the migration of several wild-type (outbred) or Sey+/+ cells, it was empirically determined by 2 independent observers that a θ ≥ 15° between frames defines a change in direction or turn. Additionally, the sum of 2 successive θ's in the same direction that equals or exceeds 15° was also defined as a change in direction. The θ's between time points were also summed to calculate the net change in direction for each cell. Comparisons between Sey+/+ and Sey−/− cells were made using a 2-tailed Student's t-test.
Transplantation Slice Culture
Sey+/−; GFP mice were mated to Sey+/− mice to obtain Sey−/−; GFP, Sey+/+; GFP, and non-GFP littermates. E14.5 brains were removed and embedded in 4% low–melting point agarose and processed as described above to generate 250-μm coronal slices. Using an ocular micrometer, 100 × 100 μm pieces of MGE were dissected from Sey−/−; GFP and Sey+/+; GFP slices. As above, slices from Sey−/− and Sey+/+ non-GFP littermates were transferred to membranes. At the transplantation site, a 100-μm cut was made between the MGE and lateral ganglionic eminence (LGE) in these non-GFP slices. The 100 × 100 μm pieces of GFP-expressing GE were transplanted into Sey−/− and Sey+/+ non-GFP slices, at the cut site between the MGE and LGE. Transplanted slices were allowed to grow in culture for 2 days and were fixed in 4% paraformaldehyde. The number of GFP+ cells that had migrated into the cortex was counted manually. Analysis of migration distance was performed as stated above and cell numbers were similarly compared, using a 2-tailed Student's t-test.
Brains of Sey−/− and Sey+/+ littermate embryos, ages E12.5, E14.5, E16.5, or E18.5, were fixed for 4–24 h in 4% paraformaldehyde alone or with 0.25% gluteraldehyde at 4 °C. Fixed brains were embedded in 2% agarose and sectioned coronally on a vibrotome at 60–75 μm for immunostaining as free-floating sections, were processed through paraffin and sectioned coronally at 5 μm, or were cryoprotected in 30% sucrose and cryosectioned coronally at 30 μm. For immunostaining of paraffin-embedded tissue, nonadjacent sections (2–4 sections per animal) representing different rostrocaudal levels of the GE were chosen for immunostaining, and antigen retrieval with a citric acid–based Antigen Unmasking Solution (Vector Laboratories, Burlingame, CA) was performed. For all subsequent steps, free-floating sections, paraffin sections, and cryosections were immunostained using the following method. Sections were blocked in 10% normal goat serum for 1 h at room temperature. Primary antibodies against calretinin (1:400, Swant, Bellonzona, Switzerland), calbindin (1:400, Swant), GABA (1:1000, Sigma), intermediate weight neurofilament protein (NFP) (2H3) (1:50, Developmental Studies Hybridoma Bank (DSHB), University of Iowa), NFP 3A10 (1:50, DSHB), or Lhx6 (1:300, Abcam, Cambridge, MA) were diluted in 5% normal goat serum and incubated with sections for 1.5 h at room temperature or overnight at 4 °C. Secondary antibodies used were goat anti-rabbit Texas Red or FITC (1:200, Jackson ImmunoResearch, West Grove, PA), goat anti-mouse IgG Texas Red or FITC (1:200, Jackson), and biotinylated goat anti-rabbit (1:500, Jackson). Biotinylated secondary antibodies were subsequently incubated with streptavidin-Cy3 (1:500, Jackson) or streptavidin–horseradish peroxidase (Jackson) and reacted with 3′-3′-diaminobenzidine (Sigma). Immunostained free-floating sections were scanned using a Leica DM IRE2 HC fluo TCS 1-B-UV microscope coupled to a Leica TCS SP2 spectral confocal system. Immunostained paraffin sections were viewed on a Nikon E400 equipped with a Leica D480 digital camera or were scanned using a BLISS imaging system (BACUS Laboratories, Lombard, IL). All immunopositive cells that had migrated into the pallium (defined by the corticostriatal notch ventrolaterally to the apex dorsomedially) were counted. Binning of cells was performed by separately counting cells in the deep layers (VZ to intermediate zone [IZ]) and superficial layers (CP to marginal zone [MZ]).
Whole-Brain Electroporation and Fate Mapping
Brains of Sey−/− and Sey+/+ littermate embryos, ages E14.5 to E15.5, were dissected and embedded in 4% low–melting point agarose. An eGFP expression construct in pNIT (1.4 μg/mL; kindly provided by Dr F. Gage, Salk Institute) was pressure microinjected (approximately 1 μL) into one of the lateral ventricles using a heat-pulled glass capillary on a Hamilton syringe (100 μL). Electroporation (EP) was performed by discharging three 40-V pulses, each 50 ms in length, with1 s intervals between each pulse (Electro Square Porator ECM 830; BTX, Holliston, MA). Electroporated brains were sectioned coronally at 225–250 μm and cultured as above for 2 days. Fixed slices were immunostained with the anti-Distalless (Dll) antibody that recognized murine Dlx family members (a gift of Dr G. Boekhoff-Falk, University of Wisconsin) or for calbindin and/or calretinin. Sections were blocked in 5% normal goat serum, 0.3% Triton-X, 0.3% Tween-20, and 0.05% sodium azide/PBS for 4 h at room temperature. The primary antibodies, Dll (1:200; Dr G. Boekhoff-Falk), or calretinin (1:400, Swant) and calbindin (1:400, Swant), were diluted in 5% normal goat serum, 0.05% sodium azide/PBS and incubated with sections for 24 h at 4 °C. After several hours of washing, sections were incubated with Texas red conjugated goat anti-rabbit secondary antibodies (1:150, Jackson) for 3 h at 4 °C. Nuclei were counterstained with DAPI (1:1000, Molecular Probes). Sections were scanned with a Leica DM IRE2 HC fluo TCS 1-B-UV microscope coupled to a Leica TCS SP2 spectral confocal system.
The Distance of NRCM Is Reduced in Sey−/− Mice
As a result of impaired radial cell migration (RCM), neuroblasts accumulate in the subventricular zone (SVZ) and IZ of Sey−/− mice (Caric et al. 1997). How the loss of Pax6 affects NRCM is less clear. An increased number of cells migrating nonradially from the LGE have been reported in Sey−/− mice (Chapouton et al. 1999). Given this observation and that loss of Pax6 causes a dorsal expansion of transcription factors normally expressed ventrally (Toresson et al. 2000), we hypothesized that there may also be expansion of motogens or migration attractants expressed in the ventral telencephalon. Alternatively, downregulation of molecules that guide or restrict migration could also explain these data. Thus, we predicted that Sey−/− mice would show an increased rate of NRCM and/or increased numbers of migrating cells.
To determine the population average rate of NRCM into neocortex, we labeled migrating GE cells in embryonic mouse forebrain slice cultures with DiI at E14.5 and compared the nonradial distance migrated over a defined period of time (average rate of migration) in Sey+/+ and Sey−/− forebrain slices. By E14.5, interneuron migration from the GE has begun in both Sey+/+ and Sey−/− animals; therefore, any observed difference in distance traveled by DiI-labeled cells in Sey+/+ and Sey−/− slices should be due to varied rate of migration and not to a delay in the start of migration. After 2 days in culture, the forebrain slices were fixed, and the distance cells that had migrated in Sey+/+ and Sey−/− slices were calculated as previously described (McManus, Nasrallah, Pancoast, et al. 2004). Labeled cells oriented radially from the DiI crystal were excluded. Unexpectedly, we found that Sey+/+ cells migrated significantly farther on average than did Sey−/− cells (748 ± 86 μm versus 517 ± 75 μm, P < 0.001; 1723 cells from 8 Sey+/+ embryos and 1428 cells from 6 Sey−/− embryos, taken from at least 4 litters; Fig. 1).
Frequent Changes in Direction Reduce the Distance of NRCM in Sey−/− Mice
Though the reduced distance of migration we observed reflects a slower average rate of NRCM in Sey−/− mice, these data do not indicate whether individual Sey−/− cells migrate more slowly or whether they migrate at a normal rate but take a more circuitous route through cortex. To directly test the rate of NRCM of individual cells, we performed time-lapse imaging of DiI-labeled cells as previously described (Nasrallah et al. 2006). Sey−/− cells migrated at an average speed of 0.88 ± 0.33 μm/min (n = 12 cells, from 4 embryos derived from 3 litters), similar to the 0.87 ± 0.21 μm/min (n = 14 cells, from 5 embryos derived from 3 litters) found for Sey+/+ littermate cells. However, direct observations indicated the Sey−/− cells migrated irregularly, often “zigzagging” through the IZ, following a leading process with one or two branches. These directional changes were quantified by measuring the change in orientation (θ) of a migrating cell between successive time points of time-lapse images. Angle analysis revealed that Sey−/− cells made more changes in direction, θ ≥ 15 (defined as a turn) than Sey+/+ cells, suggesting a defect in guidance cues (Fig. 2; 0.96 ± .0.79 turns/h for Sey+/ cells, n = 15, and 2.2 ± 0.78 turns/h for Sey−/− cells, n = 18; P < 0.001). These data indicate that there is a defect in NRCM, although not a defect in the ability of the cells to move. Instead, these findings indicate that a defect in guidance underlies the reported delay in interneurons reaching the cortex in Sey−/− animals compared with Sey+/+ animals of the same age (Jimenez et al. 2002). Because our data are inconsistent with the overmigration reported by Chapouton et al. (1999), we sought to further characterize the NRCM defect observed in Sey−/− forebrain slices.
NRCM Deficits in Sey−/− Are Cell Nonautonomous
The RCM defect in Sey−/− is the result of incomplete differentiation of radial glia, the scaffold for RCM (Gotz et al. 1998). Given that Pax6 is not expressed in the MGE, where cortical interneurons are generated, we hypothesized that the observed defects in NRCM would also be cell nonautonomous. To test this hypothesis, we used transplantation slice cultures, similar to those of Metin et al. (1997). The 100 × 100 μm pieces of Sey+/+; GFP and Sey−/−; GFP MGEs were excised from coronal forebrain sections, transplanted into the GE (between the LGE and MGE) of Sey+/+ or Sey−/− non-GFP forebrain slices, and then cultured for 48 h (Fig. 3A). Migration of MGE cells on slices of the same genotype served as controls. Comparing the distance of migration of Sey+/+; GFP or Sey−/−; GFP MGE cells on slices of opposite genotype allowed us to assess whether the decreased distance of migration is cell autonomous or nonautonomous. All comparisons were made using a Student's t-test.
To compare the distance of NRCM in these 4 transplantation combinations, 2 measures were used: average distance traveled by all cells and average distance traveled by the leading 25 cells. GE-derived cells, independent of genotype, migrated equally well on Sey+/+ slices, whereas they migrated an equally reduced distance on Sey−/− slices (Fig. 3A,B). In addition, the absolute difference between distance of migration on Sey+/+ and Sey−/− cortices is greater when we assay the distance of migration for only the farthest 25 cells (Fig. 3C). Presumably, the leading 25 cells have had the most time to migrate and therefore most clearly exhibit the defects in the migratory substrate (McManus, Nasrallah, Pancoast, et al. 2004). Further, no significant difference is observed in the ability of Sey+/+ and Sey−/− GE cells to migrate on Sey+/+ cortex; thus, there is no intrinsic defect in the ability of Sey−/− GE cells to migrate or to follow appropriate migratory cues. Instead, these data suggest that the Sey−/− cortical substrate has defects in molecular and cellular substrates and/or guidance cues because neither Sey+/+ nor Sey−/− GE cells can migrate a normal distance on Sey−/− cortex. These data indicate that increased rates of NRCM cannot account for the increased numbers of cortical interneurons in the Sey−/− neocortex.
Disorganization of Axon Tracts in Sey−/− May Contribute to the Reduced Distance of NRCM
The cell nonautonomous defect in the distance of NRCM in Sey−/− suggests that the mutant cortex lacks NRCM guidance molecules and/or substrates. Although roles for Neuregulin-ErbB4 signaling as a chemoattractive interaction (Flames et al. 2004) and semaphorin–neuropilin signaling as a guidance cue for NRCM (Marin et al. 2001) have recently been identified, less is known about substrates for NRCM. Axons are possible substrates for NRCM (Golden et al. 1997; Denaxa et al. 2001), and our laboratory has directly shown axonophilic migration for a population of interneurons from the GE (McManus, Nasrallah, Gopal, et al. 2004).
Thalamocortical and corticofugal axon outgrowth and guidance defects are well characterized in Sey−/− mice (Mastick et al. 1997; Hevner et al. 2002; Jones et al. 2002; Pratt et al. 2002). Thus, we hypothesized that the nonautonomous NRCM defect in Sey−/− mice results from disorganization or absence of axons used as a migration substrate. Similar to results described by others, thalamocortical fiber tracts, marked by intermediate weight NFP and neurofilament 3A10, are disorganized and qualitatively reduced in Sey−/− mice (Fig. 4A,B and data not shown; Mastick et al. 1997; Pratt et al. 2000; Jones et al. 2002). In addition, confocal microscopy for GABA and NFP double immunofluorescence revealed many nonradially oriented GABA-expressing cells with migratory morphology (Anderson et al. 1997) closely apposed to NFP-expressing fibers in Sey+/+ or wild-type (CD1) mice (Fig. 4C,E,G; McManus, Nasrallah, Gopal, et al. 2004). This close association of migrating GABA immunopositive interneurons with NFP fibers was not observed in Sey−/− animals; instead, GABAergic cells appeared to be randomly interspersed among the disorganized NFP immunopositive axons (Fig. 4D,F,H). These data are consistent with the hypothesis that the nonautonomous NRCM defect in Sey−/− mice results from a lack of substrate for migration.
We next sought to test whether the observed axon defects contribute to the reduced distance of NRCM. Recognizing NRCM occurs in both the IZ where axons are present and in the VZ/SVZ where axons are absent (McManus, Nasrallah, Gopal, et al. 2004), we postulated that the distance cells migrated in the VZ/SVZ would be normal, whereas the distance cells migrated in the IZ would be reduced. To test this hypothesis, we immunolabeled cryosections from E14.5 Sey−/− and wild-type or heterozygote littermates with an Lhx6 antibody. Lhx6 is expressed in the MGE and cells expressing Lhx6 migrate nonradially to the neocortex (Lavdas et al. 1999). We verified the specificity of the Lhx6 antibody by immunostaining cortical sections from E14.5 Dlx5/6 cre GFP transgenic embryos for Lhx6. As expected, greater than 95% of Lhx6 immunopositve cells in the VZ/SVZ and IZ were also GFP positive (data not shown; Wonders and Anderson 2006). Importantly, Lhx6 expression is not dorsally expanded into the neocortex of Sey−/− mice (Stoykova et al. 2000); therefore, any Lhx6 immunopositive (hereafter designated as +) cells observed in the neocortex must have migrated nonradially from the MGE. Fewer Lhx6+ cells were found in the IZ of Sey−/− mice (n = 3 animals), compared with Sey+/+ or Sey+/− littermates (n = 5 animals), whereas similar numbers of Lhx6+ cells were found in the SVZ of Sey−/− and Sey+/+ littermates (Fig. 5A–C; 9.4 ± 2.4 Lhx6+ cells per IZ of dorsal cortex in Sey+/+ compared with 4.5 ± 1.1 Lhx6+ cells per IZ of dorsal cortex in Sey−/− animals, P < 0.02; 1.6 ± 1.1 Lhx6+ cells per SVZ of dorsal cortex in Sey+/+ animals compared with 1.3 ± 0.50 Lhx6+ cells per SVZ of dorsal cortex in Sey−/− animals, P > 0.05). These data are in agreement with our hypothesis that the axon defect present in Sey−/− mice results in disorganized NRCM guidance and therefore reduced overall migration distance. However, we cannot exclude the possibility that the loss of extracellular matrix molecules or other signals in the IZ may also contribute to the documented defects.
Number of Nonradially Migrating Interneurons Is Comparable in Sey−/− and Sey+/+
Rather than an increased rate of NRCM to explain the increased numbers of interneurons in Sey−/− cortex, we next postulated that loss of Pax6 increases the number of cells migrating from the GE into the neocortex. This hypothesis predicts that the GE normally produces a set number of cells that migrate nonradially to the cortex versus cells that migrate to the striatum or other areas of the forebrain. Because loss of Pax6 disrupts dorsal–ventral patterning, an increased proportion of cells may be specified to migrate to the cortex in Sey−/− mice. To investigate this hypothesis, we determined the number of cells derived from the 100 × 100 μm transplanted piece of Sey+/+; GFP MGE or Sey−/−; GFP MGE that migrate into Sey+/+ and Sey−/− forebrain slices. Similar numbers of Sey+/+; GFP and Sey−/−; GFP MGE cells migrated into the cortices of Sey+/+ and Sey−/− non-GFP brains (P > 0.05 for all pairwise combinations; Fig. 3D). Thus, from a defined volume of GE, no difference in the number of interneurons specified to migrate to neocortex was found between Sey+/+ and Sey−/− animals. Taken together, our slice culture and transplantation data refute the hypotheses that an increase in the rate of NRCM or number of nonradially migrating cells contributes to excess inhibitory interneurons in the Sey−/− cortex.
In Vivo Numbers of Inhibitory Interneurons Reflect Decreased Migration Only at Early Stages
Migration studies have shown that interneurons follow spatially and temporally distinct routes to the neocortex: a superficial route through the MZ and CP early in neurogenesis (E12) and a deep route through the IZ at the peak of NRCM, between E13.5 and E15.5 (Anderson et al. 2001; Marin and Rubenstein 2001; Wichterle et al. 2001). Because of the dynamic nature of migration, we examined the number and distribution of interneurons expressing calbindin and calretinin, calcium-binding proteins expressed by 2 subpopulations of cortical GABAergic interneurons as well as a subset of molecular layer neurons (Kubota et al. 1994; Meyer et al. 1999; Xu et al. 2004). Sey+/+ and Sey−/− littermate embryos were collected from at least 3 litters at E12.5, E14.5, E16.5, and E18.5. The number of calbindin or calretinin immunoreactive cells in the entire dorsal cortex (corticostriatal notch to apex) was counted in coronal sections representing different rostrocaudal planes through the GE. In addition, the number of calbindin immunopositive cells was further subdivided to show the distribution of cells in the deep layers (VZ to IZ) and the superficial layers (CP to MZ) in the neocortex.
We found that Sey+/+ and Sey−/− cortex have similar numbers of calretinin and calbindin immunopositive (+) cells early in development (E12.5 and E14.5; Fig. 6E,F). Despite the similar total numbers of calbindin+ cells at E14.5, the decreased number of calbindin+ cells in the IZ to VZ, a main route of nonradially migrating interneurons at this age, reflects the NRCM defect we and others have observed (Fig. 6C,D,G and Figs 1 and 5; Jimenez et al. 2002). However, as neurogenesis progresses, the number of calretinin+ and calbindin+ cells increases disproportionately in Sey−/− cortex. Similar to previous studies, the number of calretinin+ cells is increased at E16.5 and E18.5 (Fig. 6E; Stoykova et al. 2003). In addition, increased numbers of calbindin+ cells were found in the subcortical layers (VZ through IZ) at E18.5 (Fig. 6G).
Fate of Neurons Generated in the Cortical VZ in Sey−/−
Given that our data support neither an increased rate of NRCM nor an increased number of migrating MGE cells to account for the increased numbers of interneurons we observed and reported by others in Sey−/− mice (Chapouton et al. 1999), we invoked an alternate mechanism to explain the increased numbers of cortical interneurons. Our data on the distribution of calbindin+ cells in the Sey−/− neocortex suggest that the increase in interneurons occurs primarily subcortically, and interestingly, some of these cells have migratory morphology but are oriented radially (see Fig. 6 and data not shown). Normally, the neocortical VZ of the dorsal telencephalon produces projection neurons (Parnavelas 2000); however, recent studies have shown ventralization of the Sey−/− cortex. For example, cells of the Pax6−/− dorsolateral telencephalon show loss of VGLUT2 (Schuurmans et al. 2004) but express GAD67 transcript (Yun et al. 2001), and interneuron heterotopia in the Sey−/− cortex are derived from the cortical VZ (Kroll and O'Leary 2005). Based on these data, we hypothesized that neurons in the cortical VZ are misspecified as interneurons and migrate radially toward the CP.
To test this hypothesis, we performed whole-brain EP with an expression vector encoding GFP to label cells of the cortical VZ in Sey+/+ and Sey−/− brains, ages E14.5 and E15.5. Past experience with this technique in our laboratory and others has shown that cortical VZ can be selectively targeted (Fig. 7A; Bai et al. 2003). We followed the fate of GFP electroporated cells by culturing electroporated forebrain slices for 48 h. Slices were then fixed and immunostained for Dlx or calbindin and calretinin. We chose to immunostain for Dlx rather than GABA or GAD65/67 because Dlx expression correlates highly with GAD gene immunoreactivity (Stuhmer, Anderson, et al. 2002), and in our experience, Dll antibody produces stronger and more reliable signal than GABA and GAD antibodies. We used confocal microscopy to identify cells that express both GFP and an interneuron marker in Sey+/+ and Sey−/− slices. Although we never observed GFP-positive, Dlx+ cells in Sey+/+ slices (n = 632 cells analyzed from 3 animals), 13 ± 2.4% of GFP-positive cells coexpress Dlx in the cortex of Sey−/− animals (n = 758 cells analyzed from 4 animals; Fig. 7B–D). However, we did not observe GFP-positive, calbindin, or calretinin immunopositive cells in either Sey+/+ or Sey−/− forebrain slices (data not shown). These data support a model in which the excess cortical interneurons are derived from misspecified dorsolateral cortical VZ. Interestingly, our results suggest that the expansion of Dlx mRNA expression into the dorsoventral neocortex (Toresson et al. 2000; Schuurmans et al. 2004) is not sufficient to respecify all these neocortical cells as interneurons.
The reported increase in cortical interneurons in Sey−/− animals (Chapouton et al. 1999; Kroll and O'Leary 2005) prompted us to study this mutant mouse in an effort to identify regulators of NRCM as well as mechanisms of interneuron specification. Our data indicate that Sey−/− mice have a cell nonautonomous defect in the distance neurons migrating nonradially, but Sey−/− interneurons migrate at a normal rate. This demonstration of reduced distance of NRCM with an unchanged migration rate in a mutant animal has not been previously reported. In addition, the novel finding that Sey−/− interneurons migrate via a circuitous route, as shown by the angle analysis, further indicates that a disruption of substrates and/or guidance cues exists for NRCM in the Sey−/− mutant mouse. In contrast to the findings of Chapouton et al. (1999), we find no alteration in the number of GE cells specified to migrate into the pallium. Thus, interneuron migration from the ventral forebrain cannot account for the increased numbers of cortical interneurons observed in the Sey−/− neocortex. However, we do find increased numbers of cortical interneurons specified in the cortical VZ and SVZ, indicating these cells are born in the pallium. These data are consistent with the ventralization of the Sey−/− cortex based on molecular markers (Toresson et al. 2000; Yun et al. 2001; Schuurmans et al. 2004) and cortical VZ–derived interneurons populating the subcortical heterotopia (Kroll and O'Leary 2005). Thus, our data invoke a shift from a reported overmigration defect in Sey−/− mice to a deficiency of migration reflecting a pathfinding defect. We further determine that the increased number of cortical interneurons reflects misspecification of a subset of cells in the ventrolateral cortical VZ as interneurons.
Disorganized Axon Tracts May Account for the Cell NonAutonomous NRCM Defect in Sey−/−
Radial migration is dependent on radial glia, which serve as scaffolds for neuronal migration (reviewed in Rakic 1990; Hatten 1999). Furthermore, physical interactions between radial glia and migrating neurons are necessary to maintain the radial glial profile (e.g., astrotactin, Edmondson et al. 1988; Adams et al. 2002). This symbiotic relationship is also dependent on reciprocal signaling between the cells, through Neuregulin-ErbB2 and possibly other molecules (Anton et al. 1997; Rio et al. 1997).
Axons have been proposed as substrates for NRCM in the chick (Golden et al. 1997; Heffron and Golden 2000) and in the rodent (Phelps and Vaughn 1995; Phelps et al. 1996; Denaxa et al. 2001; McManus, Nasrallah, Gopal, et al. 2004). However, whether axons and adhesion molecules expressed on axons guide NRCM remains controversial. For example, TAG-1, an immunoglobulin cell adhesion molecule on axons, was found to be necessary for axon-dependent NRCM in vitro (Denaxa et al. 2001), yet not in mice deficient for this same protein (Fukamauchi et al. 2001; Denaxa et al. 2005). It is unclear whether other adhesion molecules expressed on axon tracts also guide NRCM and compensate for lack of TAG-1 expression in the knockout animals. Additionally, although the majority of interneurons travel through the IZ, where axons are present, some cells migrate multidirectionally through the VZ/SVZ and MZ, presumably on other cellular and molecular substrates (Tanaka et al. 2003; Tanaka et al. 2006). Thus, although axons may not be the only substrate for NRCM, several lines of evidence, including direct observations of axonophilic migration for interneurons in the mouse (McManus, Nasrallah, Gopal, et al. 2004), indicate that a subset of migrating interneurons do use axons as migration guides.
Sey−/− mice have well-characterized axon defects (Mastick et al. 1997; Pratt et al. 2000; Jones et al. 2002 and data herein). Sey−/− embryos exhibit axon pathfinding defects during the early development of the prosencephalon (Mastick et al. 1997) and impaired axon outgrowth with disorganization of corticofugal and thalamocortical tracts (Hevner et al. 2002; Jones et al. 2002; Pratt et al. 2002). In agreement with these data, we observed disorganization of NFP immunopositive axon fascicles as well as several other markers of axon tracts in the forebrain (Fig. 4 and data not shown).
Given that Pax6 is not expressed in the GE and the known axon defect in Sey−/− mice, a defect in NRCM would be reasonably attributed to cell nonautonomous mechanisms. Our data clearly fit this prediction. Like the Sey−/− mouse, Emx1/2−/− mice also have corticofugal and thalamocortical axon defects (Lopez-Bendito et al. 2002) and a cell nonautonomous NRCM defect (Shinozaki et al. 2002). Interestingly, Pax6 and Emx2 are expressed in complementary domains in the pallium (Bishop et al. 2000) and regulate each other's expression to pattern the cortex (Muzio et al. 2002). However, whether the NRCM phenotypes in Sey−/− and Emx1/2−/− mice are caused by a common mechanism is unknown. In addition to the axon defects in Sey−/−, there may also be a lack of other NRCM regulatory molecules in the cortex, which contribute to the migration defect. For example, Sey−/− cortex has reduced expression of Neuregulin1, an NRCM attractant (Flames et al. 2004), but interestingly, expression of Neuregulin1 in the LGE is conserved (Assimacopoulos et al. 2003).
The forebrain of Sey−/− mice has numerous structural and morphological abnormalities that could also contribute to or be the primary cause of the NRCM defect. For example, reduced cellularity and disorganization of the CP (Schmahl et al. 1993) as well as subpial and paraventricular ectopias (Kroll and O'Leary 2005) disrupt the normal pallial architecture; however, we believe these are unlikely contributing factors. Studies in Lis1+/− mice, which also have disorganized cortices (Hirotsune et al. 1998), show cell autonomous defects in NRCM with minimal cell nonautonomous defects (McManus, Nasrallah, Pancoast, et al. 2004). Thus, specific deficits in the axonal substrate discussed above and/or guidance molecules in Sey−/− mice are more likely to be responsible for the robust cell nonautonomous reduction in distance of NRCM.
Interneurons Are Ectopically Specified in Sey−/− Cortical VZ
Although we found a defect in NRCM and no alteration in the number of GE cells migrating to the neocortex (Fig. 3D), normal to increased numbers of interneurons were located in the pallium (Chapouton et al. 1999 and Fig. 6). The distribution of calbindin+ cells in the Sey−/− pallium suggests that normal numbers of interneurons eventually reach the CP of Sey−/− animals, whereas the increase in interneurons occurs primarily subcortically. Increased numbers of calretinin+ cells at E16.5 and later were also observed; however, at least some of the additional calretinin+ cells in the CP and MZ are Cajal-Retzius cells (Stoykova et al. 2003) and may or may not also reflect increased numbers of interneurons. Interestingly, some of these cells have a radially oriented migratory morphology. This distribution of interneurons suggested different origins for the interneurons in the cortical and subcortical layers of the Sey−/− pallium. Indeed, recent work has shown that ectopic accumulations of interneurons in the cortical VZ of Sey−/− mice express Sp8, a marker for LGE-derived interneurons, whereas interneurons in the CP express Lhx6, a marker of MGE-derived interneurons (Kroll and O'Leary 2005).
Although the majority of cortical interneurons originate ventrally and migrate nonradially to reach the cortex in rodents, it has been suggested that5–10% of cortical interneurons are generated from the dorsal cortical VZ in mice (Stuhmer, Puelles, et al. 2002). Because progenitors in the Sey−/− telencephalon have decreased expression of dorsal transcription factors and cortical markers, Ngn2, Emx1, and VGLUT2, and gain expression of ventral markers, Dlx1/2 and GAD67, particularly in the lateral neocortex (Toresson et al. 2000; Yun et al. 2001; Schuurmans et al. 2004), we hypothesized that neurons in the cortical VZ are misspecified as interneurons and migrate radially but do not necessarily reach the cortex. Our data (Fig. 7) and genetic lineage–tracing experiments (Kroll and O'Leary 2005) confirm that the increased numbers of interneurons in the Sey−/− pallium can be explained by misspecification of interneurons, rather than an increase in NRCM. These data raise the question as to whether the Sey−/− neuroepithelium increases its production of interneurons from 5% or whether the Sey−/− cortical neuroepithelium is redefined. Ectopic expression of Dlx, Mash1 (Toresson et al. 2000), and GAD67 mRNA transcript (Yun et al. 2001) in the ventrolateral cortical VZ of Sey−/− animals suggests that the Sey−/− cortex is molecularly redefined. In addition, electroporation of Dlx2 or Dlx5 in the neocortex of wild-type animals induces GAD65 immunoreactivity in cells that receive Dlx (Stuhmer, Anderson, et al. 2002). Taken together, these data suggested that loss of Pax6 and gain of Dlx expression shifts the pallio–subpallial boundary dorsally and expands the interneuron progenitor zone into the ventrolateral neocortex (Toresson et al. 2000). Paradoxically, however, our data indicate that not all cells derived from the ventrolateral cortical VZ are respecified to an interneuron fate, suggesting that loss of Pax6 and gain of Dlx, Gsh2, and Mash1 transcript in the Sey−/− cortex (Toresson et al. 2000) are not sufficient to respecify all cells in the ventralized cortex as interneurons. Alternatively, all cells in the ventralized cortex may not retain competence to change specification by E15. Additionally, in our EP experiments, we never observed colocalization of GFP with calbindin or calretinin, markers of differentiated interneuron subtypes. Interestingly, ectopic Dlx2 or Dlx5 expression in wild-type neocortex also failed to induce markers of cortical interneurons such as NPY, nNOS, calretinin, or calbindin (Stuhmer, Anderson, et al. 2002). These data suggest that either additional time is required to express more mature markers or that gain of Dlx expression is not able to induce the complete cascade of interneuron differentiation. Again, we speculate that cortical VZ cells are not competent to express all genes of GABAergic interneurons or additional transcription factors are required in parallel or in conjunction with Dlx to complete interneuron specification.
Recent studies have found a correlation between the developmental origin of cortical interneurons and their molecular and electrophysiological properties (Xu et al. 2004; Butt et al. 2005). The misspecification of interneurons in the cortical VZ of Sey−/− animals raises the question as to whether these cortically derived interneurons have distinct molecular and electrophysiological properties from interneurons derived ventrally. For example, it has been suggested that a Mash1+, Dlx+ population of interneurons arise in the cortex of the developing primate, distinct from the Mash1−, Dlx+ interneurons derived from the GE. However, the authors cannot exclude the possibility that some of the Mash1+, Dlx+ cells might have migrated from the GE (Letinic et al. 2002). Therefore, a GABAergic interneuron derived from the pallium can be unequivocally identified molecularly by showing that GABA+ (or Dlx+) cells are of the Emx1 lineage, using Emx1-Cre; R26R mice (Kroll and O'Leary 2005). Further studies will be necessary to determine the electrophysiological properties of these Emx1 lineage interneurons and how they differ from ventrally derived interneurons.
Pax6 Regulates Multiple Migration Pathways
The reduced distance of NRCM we observed in Sey−/− mice appears to contradict the previously reported increase in NRCM from the LGE (Chapouton et al. 1999). This reported increase in migration from the LGE likely represents a distinct population of cells. More recent lineage-tracing experiments have shown that cells of the LGE do not migrate nonradially to the neocortex and instead give rise to olfactory bulb interneurons (Anderson et al. 2001; Wichterle et al. 2001; Nery et al. 2002). Based on our interpretation of the data presented by Chapouton et al. (1999), we believe that they focused on migration of LGE cells to the lateral cortex and olfactory bulb–like structures in the lateral cortex of Sey−/− animals (Jimenez et al. 2000), rather than ventral to dorsal migration into the pallium, which we have specifically studied.
In summary, our data indicate that mutations in Pax6 have no cell autonomous effect on nonradially migrating neurons. Instead, we find 2 important defects in Sey−/− mice to account for the observed interneuron phenotypes. First, disorganization of axon tracts likely accounts for the abnormal NRCM that we have observed. Second, our data, along with those from Kroll and O'Leary (2005), show that ectopic specification of interneurons in the ventrolateral cortical VZ results in subcortically positioned interneurons within heterotopia in Sey−/− mice.
National Institutes of Health (NS45034, HD26979).
We would like to thank MacLean Pancoast, Haiyan Gu, and William Shapiro for maintaining the Sey+/− and Sey+/−; GFP lines and setting up animal breedings. We thank all members of the Golden lab, past and present, for their support, discussions, and assistance. Conflict of Interest: None declared.