The homeobox-containing gene, Emx1, a mouse homologue of Drosophila empty spiracles, is specifically expressed in the developing telencephalic cortex. It has been reported that Emx1 transcripts and the protein product are localized in most cells of the cerebral cortex during the process of proliferation, migration, differentiation and maturation. We provide evidence here, based on a multitude of experimental approaches in developing rats, in support of the hypothesis that the expression of this gene is restricted to pyramidal neurons. Specifically, we found that, similar to pyramidal neurons, cells expressing Emx1 are distributed in all cortical layers, except layer I. Using in situ hybridization and immunocytochemistry at the light and electron microscope levels, we have shown that the density, distribution, soma shape and ultrastructural features of these cells were identical to those of pyramidal neurons. Double-labelling experiments confirmed that the vast majority of Emx1-expressing cells also contained glutamate, a marker of pyramidal neurons. We also found that this gene is expressed by most glutamate-containing neurons in dissociated cortical cell cultures and the vast majority of cells in radially arranged clones of pyramidal cells in the cortices of chimeric mice. Thus, the homeobox gene Emx1 can be reliably used as a marker of the pyramidal cell lineage.
The mammalian cerebral cortex is composed of an enormous number of neurons and glia organized into cytologically and functionally distinct areas. All areas share a common basic structure with neurons arranged in six layers. The majority of cortical neurons are pyramidal cells found in all layers except layer I (Lorente de Nó, 1949). These are the projection cells of the cortex that utilize the excitatory amino acid l-glutamate as a neurotransmitter (Parnavelas et al., 1989). The remaining neurons, scattered in all layers, are the nonpyramidal cells. They are the cortical interneurons that contain the inhibitory neurotransmitter GABA (Parnavelas et al., 1989).
Recent evidence suggests that the two neuron types are generated in distinct proliferative zones. Pyramidal cells are derived from the germinal ventricular zone that lines the telencephalic ventricles and migrate to their positions in the cortex guided by radial glia (Mione et al., 1997; Tan et al., 1998). While few nonpyramidal cells are generated in the ventricular zone (Parnavelas, 2000), the vast majority is derived from the ganglionic eminence of the ventral telencephalon (Anderson et al., 1997; Lavdas et al., 1999; Parnavelas, 2000).
Little is known about the molecular mechanisms underlying the complex changes that occur during the development of the forebrain and about the genes that are involved in the control of cell fate in the cortex. In the past decade, several families of genes coding transcription factors, including homeobox-containing genes, have been cloned in Drosophila and shown to be expressed with overlapping patterns in the developing brain of rodents (Rubenstein and Puelles, 1994). Several of these genes, including members of the empty spiracles (Emx1 and Emx2) and orthodenticle (Otx1 and Otx2) families, have been found to be expressed in the cerebral cortex (Simeone et al., 1992a,b; Boncinelli et al., 1995). Among these four genes, only Emx1 shows widespread expression both in proliferating and postmitotic neurons of the cerebral cortex. The fact that this gene is expressed both in the developing and mature cortex hints for a possible role of Emx1 in the initiation and maintenance of the ‘cortical neuron’ phenotype. The spatial and temporal pattern of expression of Emx1 coincides with the pattern of distribution of pyramidal cells from the time of their origin in the ventricular zone to their final settling in the mature brain. In contrast, nonpyramidal cells, which originate in the ganglionic eminence, express the LIM-homeobox gene Lhx6 and appear to acquire their morphological and neurochemical identity prior to their arrival in the cortex (Anderson et al., 1997; Lavdas et al., 1999).
Here, we describe the pattern of expression of Emx1 in the cerebral cortex of rats at different stages of postnatal life and in adult animals. Using in situ hybridization and immuno-cytochemistry at the light and electron microscope levels, we show that the expression of this homeobox gene is restricted to pyramidal neurons, the projection cells of the cerebral cortex.
Materials and Methods
Sprague–Dawley albino rats of various ages, postnatal day (P) 4 (n = 2), P7 (n = 2), P14 (n = 2), P21 (n = 2), P35 (n = 4) and 8 weeks (n = 2), were anaesthetized with ether and perfused through the heart with 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4 (PB). For electron microscopy (one additional animal at P21 and at P35), 0.2% glutaraldehyde was added to the fixative solution. Following the perfusion, the brains were immersed in fixative for 3 h before being transferred into PB. Sections were cut in the coronal plane at 50 μm using a Vibratome and collected in PB. They were rinsed in PB, incubated in 5% normal goat serum in PB for 45 min, and then transferred into EMX1 antiserum [(Briata et al., 1996); 1:500 in PB with 5% normal goat serum and 0.3% Triton X-100] at 4°C for ~40 h. They were then processed with the avidin–biotin–peroxidase method using the ABC-Elite kit (Vector Laboratories, Burlingame, CA) and reacted with diaminobenzidine and hydrogen peroxide. Sections were mounted on slides and coverslipped with DPX. Sections destined for electron microscopy were postfixed in 1% OsO4 in PB, rinsed in acetate buffer, stained in 1% aqueous uranyl acetate, dehydrated through an ethanol series, passed through propylene oxide, and embedded in Araldite on microscope slides.
In Situ Hybridization
Two animals for each of the postnatal ages used for EMX1 immuno-cytochemistry were perfused with 4% paraformaldehyde in PB and the brains removed and immersed in fixative for ~3 h. They were then washed in PB, immersed in 20% sucrose, sectioned in the coronal plane at 15 μm using a cryostat, and collected on Superfrost Plus slides (BDH, Poole, UK).
Emx1 sense and antisense DIG-labelled (Roche, Mannheim, Germany) RNA probes were generated by in vitro transcription using a 330 bp sequence as template (Simeone et al., 1992a,b). Sections were treated with proteinase K (5 μg/ml), washed in phosphate buffer saline (PBS), acetylated, washed in distilled water, and pre-hybridized for 3 h at 50°C. Each slide was then hybridized overnight in hybridization buffer containing 500 ng/ml DIG-labelled cRNA. Slides were washed in stringent conditions (60°C, 2 × SSC, 50% formamide twice, 0.2× SSC, 50% formamide, 0.1 × SSC). The in situ signal was visualized using alkaline phosphatase conjugated to anti-DIG antibody (Roche, diluted 1:2000 in TBST, 1% blocking serum), and developed with NBT/BCIP substrate (Roche). Sections were examined and photographed using a light microscope (Leica DMRB) with brightfield illumination.
In order to identify the cell types that express Emx1, double-labelling experiments were carried out. Cryostat sections were first processed for immunocytochemistry using antibodies against glutamate (1:500; Sigma, St Louis, MO; or 1:50; Biogenesis, Poole, UK) or GABA (1:1000; Sigma), markers of pyramidal and nonpyramidal neurons, respectively, and then for in situ hybridization as described above. Sections were fixed in 4% paraformaldehyde after the first procedure and prior to processing for in situ hybridization.
Cortical Cell Cultures
Rat fetuses at embryonic day (E) 16 and E18 (E1, day vaginal plug was found) were removed from Sprague–Dawley dams, culled by cervical dislocation, and the cerebral cortices of their brains were dissected and placed in Dulbecco's modified Eagle's medium (DMEM; Sigma), containing 0.25% trypsin (Sigma) and 0.01 units of DNAse I (Roche), at 37ºC for 30 min. Cortices were dissociated with trypsin in Ca2+, Mg2+ free S-MEM medium supplemented with 0.6% BSA, 0.038% MgSO4 and 0.6% glucose (Sigma). The reaction was terminated with the addition of 10% heat-inactivated fetal calf serum (HI-FCS; Gibco-BRL, Rockville, MD). The cells were plated on poly-l-lysine coated 10 mm coverslips at a concentration of 3.5 × 105 cells/well in defined medium (DMEM/HAM F12), supplemented with 2 mM glutamine, 5 mM HEPES, 25 mM KCl, 0.6% glucose, 100 μM putrescine, 20 nM progesterone, 30 nM selenium, 25 μg/ml insulin, 100 μg/ml transferrin and 5% FCS (Sigma). After 3 days in culture, cells were fixed with 4% paraformaldehyde and 0.1% glutaraldehyde in PB for 20 min, and kept in PBS in preparation for in situ hybridization and immunocytochemistry.
For in situ hybridization, cells were dehydrated by passing through an ascending series of ethanol. After drying briefly, coverslips were incubated in pre-hybridization buffer at 45°C for 90 min. The buffer contained 50% formamide, 5 × SSC, 5 × Denhardt's, 5 mM EDTA, 25 μg/ml yeast t-RNA and 0.2 mg/ml fish sperm DNA. Hybridization took place at the same temperature for 18 h, with a probe concentration of 30 ng/ml of hybridization solution (50% formamide, 5 × SSC, 5 × Denhardt's, 5 mM EDTA). Washes were done with 2 × SSC (pre-warmed) and 0.2 × SSC at 45°C. This procedure was followed by glutamate (1:200, Sigma) or GABA (1:200; Sigma) immunocytochemistry in order to identify the cells that expressed the homeobox gene in culture. Cells were incubated in antibodies in blocking buffer, diluted 1:10 in maleic acid buffer (100 mM maleic acid, 150 mM NaCl, pH 7.5) at room temperature for 2 h. They were then incubated with anti-DIG FITC (1:500) and TRITC conjugated secondary antibodies at room temperature for 2 h, mounted in Citifluor, and observed with an epi-fluorescence microscope.
EMX1 in Chimeric Mice
We wanted to study the expression of EMX1 in pyramidal cell clones generated by injecting a single embryonic stem cell carrying a lacZ reporter into wild-type blastocysts as described by Tan and colleagues (Tan et al., 1998). Both primary antisera, anti-EMX1 and anti-β-galactosidase, were raised in rabbit. Using a method originally developed by Shindler and Roth (Shindler and Roth 1996), both antigens were detected simultaneously using secondary antibodies with discriminating fluorophores. This method works on the basis that one of the two primary antisera is used at such dilute levels that only tyramide signal amplification allows for its detection. The second primary antiserum is used at normal concentration and is detected by conventional immuno-cytochemistry. The secondary antibody used to detect the second primary antibody is incapable of detecting the first primary antiserum by virtue of the latter's extreme dilution. Here, tyramide signal amplification was used to detect the anti-EMX1 antibody while the β-galactosidase antigen was detected by conventional means.
Mice were killed by intracardiac perfusion with 4% paraformaldehyde in PB. Brains were removed and fixed for a further 10 min in the same fixative. They were then cryoprotected, frozen and sectioned at 20 μm with a cryostat. Sections were collected on AES-coated slides, dried for 2 h, and then either stored at –20°C or processed immediately.
Sections were blocked first in 10% normal horse serum (CSL, Melbourne, Australia) in PBS for 1 h and then in 0.5% Blocking Reagent (Renaissance TSA Indirect, NEN Life Science Products, Boston, MA) in a buffer containing 0.1 M Tris-Cl and 0.15 M NaCl (pH 7.6) at room temperature for 30 min. Incubation with rabbit anti-EMX1 (1:6000 in PBS, 0.3% TX-100) was performed over 36 h, followed by biotinylated goat anti-rabbit IgG (1:200, Vector Laboratories) and streptavidin-HRP (1:100, Renaissance TSA Indirect, NEN Life Science Products) for 1 h each. This was followed by a 3 min amplification of biotinyl tyramide diluted 1:100 in Amplification Diluent (Renaissance TSA Indirect, NEN Life Science Products) and a 1 h incubation in Fluorescein–Avidin D (1:200, Vector Laboratories). Following several washes in PBS, the tissue was processed for the detection of the β-galactosidase transgene product using a rabbit polyclonal antibody (1:500 in PBS, 0.3% TX-100, 5′ → 3′, Boulder, CO). Overnight incubation in primary antibody was followed by incubation in Alexa FluorTM 594 conjugated goat anti-rabbit IgG (1:400, Molecular Probes, Eugene, OR). Sections were mounted with Fluoro-mount-G (Southern Biotechnology, Birmingham, AL).
Controls included the omission of each of the primary antibodies while maintaining the rest of the immunocytochemical protocol. This allowed for the assessment of non-specific background. Additionally, tissue was processed for the detection of EMX1 as indicated above followed by an incubation in Alexa FluorTM 594 conjugated goat anti-rabbit IgG (1:400), without the addition of the second primary antibody to β-galactosidase. This ensured that the first primary antiserum was sufficiently dilute to escape detection by the conventional secondary antibody.
Expression of Emx1 in Developing and Adult Cerebral Cortex
Previous studies have shown that Emx1 is first expressed at the early stages of the embryonic rodent brain (E9.5 mouse), and expression continues into early postnatal life (Briata et al., 1996; Gulisano et al., 1996). Here, we found that the expression of this homeobox gene persists through all stages of postnatal development and in the adult cerebral cortex. Cells expressing Emx1 were numerous in all layers of the cortex, except layer I. Although at first glance cells appeared to be evenly distributed throughout these layers, careful examination clearly showed that the density, size and morphology of labelled cells was characteristic for each cortical layer. Layers II/III showed the highest density of Emx1-expressing cells, most of which had features characteristic of pyramidal neurons. Layer IV showed the lowest density of Emx1-expressing cells, most showing small round or ovoid cell bodies. Layer V contained mainly a complement of medium and large labelled neurons pyramidal in shape. The medium-sized cells were predominantly in the superficial portion of this layer, whereas the large cells were found chiefly in the deep part of layer V. Labelled cells in layer VI were generally small, showing a range of shapes and orientations (Fig. 1). This pattern of expression appeared consistent throughout different cortical areas along the rostro-caudal axis. Also, the distribution of Emx1 transcripts correlated closely to the distribution of EMX1 protein (Fig. 2A,B), an observation consistent with previous studies in embryonic cerebral cortex (Briata et al., 1996).
Emx1 is Expressed in Pyramidal Neurons
We used a variety of experimental approaches to characterize and identify the cortical cell types that express Emx1. Firstly, the in situ hybridization preparations suggested that Emx1 transcripts were present in cells whose density, distribution and soma shapes were identical to those of pyramidal neurons as described for the developing and adult rat neocortex (Parnavelas et al., 1977; Miller, 1988). Their pattern of distribution was also very similar to that of neurons labelled in the cortex with an antibody to neurotransmitter glutamate, a marker of pyramidal cells (Conti et al., 1987; Dori et al., 1989, 1992). Secondly, analysis of ultrathin sections cut from tissue taken from one P35 animal, and processed for electron microscope immunocyto-chemistry for EMX1, showed that immunoreactive neurons displayed ultrastructural features characteristic of pyramidal cells (Peters and Kara, 1985). The analysis was based on the observation of 200 labelled neurons: the first 10 cells encountered in each of layers II/III, IV, V and VI, in each of five Vibratome sections. Although some cells only revealed nuclear staining, a considerable number of neurons showed some degree of cytoplasmic labelling. As is typical of pyramidal neurons, the nuclei of labelled cells were round or slightly oval, and their nuclear envelopes showed smooth contours (Fig. 2C). Consistent with what is known about the synaptic organization of cortical pyramidal neurons (Feldman, 1984), all EMX1-labelled neurons received only symmetrical axosomatic synapses.
Thirdly, to further confirm the identity of the Emx1- expressing cells, we performed double-labelling experiments. Specifically, we processed sections from each postnatal age first for immunocytochemistry with an antibody against glutamate and then for Emx1 in situ hybridization. We found that, at all ages examined, the vast majority of Emx1-expressing cells were also glutamate immunoreactive (Fig. 3A,B). Emx1 transcripts were localized in the perinuclear cytoplasm of the cells, whilst glutamate immunoreactivity was typically present throughout the cell body. However, double-labelling experiments that involved Emx1 in situ hybridization and GABA immunocyto-chemistry showed only a very small percentage of double-labelled cells (Fig. 3C,D). Cell counts were performed in 300 μm wide strips spanning the thickness of the motor cortex in each of five sections taken from each of four animals at P35. The counts showed that 84.6% of the total number of labelled cells were Emx1 positive, 14.5% were GABA positive, and 0.9% expressed both GABA and Emx1.
We also used highly unbalanced mouse stem cell chimeras (Tan et al., 1998) to study the expression of EMX1. In these mice, single embryonic stem (ES) cells carrying the lacZ reporter gene are injected into host blastocysts, occasionally resulting in a small number of marked progenitor cells in the embryonic cerebral wall. Some progenitors generate radial columns of pyramidal neurons characterized by their expression of glutamate (Tan et al., 1998), whilst other progenitors give rise to widely scattered GABAergic neurons. Using these chimeric mice, we showed that every radially arranged β-galactosidase-positive cell also expressed EMX1 (Fig. 4A–D). This observation was confirmed for three separate columns seen in two chimeras. In contrast, EMX1 protein was not localized in scattered cells, the GABAergic interneurons (Fig. 4E–H).
Emx1 is Co-localized with Glutamate in Neurons in Cortical Cell Cultures
Primary cortical cell cultures were prepared from E15 and E18 rat embryos and maintained for 3 days in the presence of 5% FCS. Consistent with earlier studies (Pappas and Parnavelas, 1997), these cultures contained predominantly neurons and only a small percentage (~1%) of cells of glial origin. Double-labelling experiments, using in situ hybridization for Emx1 and immuno-fluorescence for either glutamate or GABA, indicated that Emx1 mRNA was present in the majority (85% in the E18 cultures) of glutamate-containing neurons. Only a small percentage (10% in the E18 cultures) of the GABAergic neurons in cultures prepared from cortices of either age were observed to express Emx1 (Fig. 5).
Emx1 is a gene that contains a homeobox analogous to that present in empty spiracles, a Drosophila gene expressed in the developing head (Walldorf and Gehring, 1992). Its expression is exclusively confined to the dorsal telencephalic neuroepithelium from the time of its first appearance in mouse embryos at E9.5 (Simeone et al., 1992a,b; Boncinelli et al., 1995; Briata et al., 1996; Gulisano et al., 1996). Thus, Emx1 is a gene specifically restricted to the developing telencephalic cortex as no hybridization signal is detectable in the basal telencephalon.
It has been reported that Emx1 transcripts and the protein product are localized in most cells of the cerebral cortex during the process of proliferation, migration, differentiation and maturation (Briata et al., 1996; Gulisano et al., 1996). However, these analyses have focused on the period of corticogenesis during embryonic life and in the early stages of postnatal life in mice. During the early stages of neurogenesis, Emx1 is expressed in nearly all cells in the proliferative zone and in the other forming layers of the developing cortex. However, there are some unlabelled cells present at the later stages of corticogenesis. The proportion of such cells increases as development proceeds through the stages of differentiation and maturation in postnatal life (Briata et al., 1996; Gulisano et al., 1996). This gradual increase in Emx1-negative cells coincides with the increasing number of nonpyramidal neurons and glial cells in the cortical plate (Parnavelas, 1999, 2000). These observations have led us to suggest that Emx1 is expressed only by pyramidal neurons from very early in embryonic life. There are several lines of evidence that support this hypothesis.
In situ hybridization experiments have shown that, similar to pyramidal neurons, cells expressing Emx1 were distributed in all layers except layer I. This pattern of distribution was also similar to that of cortical neurons stained for the neurotransmitter glutamate, a marker of pyramidal cells (Dori et al., 1989, 1992). Further, in situ hybridization and immunocytochemistry at the light and electron microscope levels showed that the density, distribution, soma shape and ultrastructural features of these cells were identical to those of pyramidal neurons (Parnavelas et al., 1977; Peters and Kara, 1985). Double-labelling experiments confirmed that the vast majority of Emx1-expressing cells were also glutamate immunoreactive. We also observed that Emx1 is expressed by most of glutamate-containing neurons in dissociated cortical cell cultures and the vast majority of cells in radially arranged clones of pyramidal neurons in the cortices of chimeric mice. Based on these observations, we suggest that Emx1 is expressed almost exclusively by pyramidal neurons from the early stages of embryonic life to adulthood and, therefore, can be reliably used as a marker of the pyramidal cell lineage.
Although we have demonstrated that the vast majority of Emx1-expressing cells were pyramidal, a small percentage was observed to contain GABA, conventionally a marker of non-pyramidal neurons. This observation suggests that some of the cells that express this homeobox gene may be cortical inter-neurons. Electron microscopical examination of Emx1-stained sections taken through the cortex of a 35-day-old animal showed no labelled cells with features characteristic of nonpyramidal neurons (Fairén et al., 1984). However, double-labelling experiments in animals of the same age showed that slightly less than 1% of the Emx1-expressing cells were GABAergic. This does not necessarily imply that these cells are nonpyramidal neurons. Evidence suggests that a small proportion of neurons in the developing and adult cortex express both neurotransmitters glutamate and GABA (Lavdas et al., 1996; Hill et al., 2000). The percentage is highest at the early stages of development (3.7% at P7, declining to 0.8% in the adult) (Lavdas et al., 1996). It is not known why some cells in the cortex and in other areas of the brain [see Lavdas et al. for references (Lavdas et al., 1996)] express both neurotransmitters. However, it is interesting to note that some GABAergic cells in the cortical plate of rhesus monkey were found to exhibit pyramidal morphology (Schwartz and Meinecke, 1992), prompting speculation that some projection neurons and interneurons of the cortex share phenotypic traits in early development.
The possibility cannot be excluded that the GABA-containing neurons expressing Emx1 represent the relatively few inter-neurons generated, together with pyramidal cells, in the cortical ventricular zone (Mione et al., 1997; Parnavelas, 2000). These cells, most likely produced by progenitors of mixed potential, have the tendency to be diffusely scattered in the cortex and, similar to their pyramidal relatives, may express Emx1. Why does the proportion of these cells decline as development proceeds in postnatal life? The decline in their frequency parallels the decrease in the percentage of mixed pyramidal/ nonpyramidal clones observed in the rat cerebral cortex during postnatal development, and may be due to selective cell death [see Lavdas et al. for a discussion (Lavdas et al., 1996)]. On the basis of the present and previous findings, it may be concluded that the homeobox gene Emx1 is expressed by pyramidal neurons and, possibly, by the relatively few GABA-containing cells generated in the cortical ventricular zone. This is in concurrence with the recent finding that the two neuronal types of the cerebral cortex are generated in distinct proliferative zones, with pyramidal cells produced in the cortical ventricular zone whilst nonpyramidal cells are derived from the ganglionic eminence.
The expression of Emx1 exclusively in the telencephalic cortex has prompted the suggestion that this gene plays a role within the dorsal part of the forebrain. The appearance of Emx1 in the dorsal forebrain even before the generation of the first cortical neurons may suggest that this transcription factor is essential of the pyramidal phenotype. However, mutant mice that lack functional Emx1 protein show only subtle differences in cortical cytoarchitecture and no change in the expression of other transcription factors, such as Otx1, Tbr1 and Id2 (Qiu et al., 1996). Further, expression of the partially deleted Emx1 gene in mutant mice is normal, suggesting that EMX1 is not important in regulating its own expression at the transcriptional level (Briata et al. 1996). One noticeable difference in the mutant mice is the reduction or absence of the corpus callosum formed by the axons of cortical pyramidal neurons located predominantly in layers II/III, V and VI (Jacobson and Trojanowski, 1974). This observation suggests that although Emx1 is not required for the acquisition of the pyramidal phenotype, it may be essential for the formation and maintenance of projections to cortical and subcortical targets.
We thank Jane Pendijky for help with the preparation of the figures. The work was supported by the EU Fifth Framework Programme (QLRT-1999-30 158).