A goal of this study was to use recombinant adenovirus (AdV) to ectopically express Emx2 in the embryonic neocortex as a gain-of-function approach to study its role in the area-specific targeting of thalamocortical axons (TCAs), using the rat as a model. First, we cloned the cDNA for the full-length coding region of rat Emx2, a homologue of Drosophila empty spiracles. We also used this sequence to define the full-length coding region of mouse Emx2 from genomic DNA. Our analysis of Emx2 expression shows that in rat, as reported in mouse, Emx2 is expressed in high caudal to low rostral, and high medial to low lateral, gradients across the cortex throughout cortical neurogenesis, and expression is primarily restricted to progenitors in the neuroepithelium. We also carried out an analysis of the distribution of cells infected with a replication defective recombinant type 5 adenovirus (AdV) containing a CAG/LacZ expression construct, following an injection into the lateral ventricle of the cerebral hemisphere at different stages of embryonic cortical neurogenesis. AdV-infected cells are broadly distributed tangentially, but their laminar distribution is differentially restricted and reflects the temporal sequence of generation of cortical neurons. This finding indicates that the AdV predominantly infects progenitors in the ventricular zone, which leads to a preferential labeling of their immediate progeny, and infects cells that have recently become postmitotic and have yet to move far from the ventricular surface. We then show that AdV-mediated ectopic Emx2 expression results in aberrant intracortical pathfinding and areal targeting of TCAs from the dorsal lateral geniculate nucleus. These findings indicate that EMX2 imparts positional information normally associated with caudal cortical areas, such as the primary visual area, that influences the area-specific targeting of TCAs. These results are consistent with a role for EMX2 in areal specification of the neocortex as suggested by recent analyses of Emx2 null mutants.
The mammalian neocortex is parceled along its tangential extent into functionally specialized areas that are distinguished, both radially and tangentially, by unique distributions of neuronal phenotypes, as well as characteristic efferent and afferent axonal projections. Layer-specific properties of neocortical neurons are determined genetically within the ventricular zone shortly before neurons become postmitotic (McConnell and Kaznowski, 1991; Chenn et al., 1997). Along the tangential dimension of the neocortex, area-specific properties appear to develop as a result of the early action of regulatory genes distributed in graded patterns across the neocortex, and extrinsic influences, such as thalamocortical axon (TCA) input, which act later within the constraints of the genetic framework (Rakic, 1988; O’Leary, 1989; Chenn et al., 1997; O’Leary and Nakagawa, 2002).
TCAs originate in the principal sensory nuclei of dorsal thalamus, are the sole source of modality-specific sensory information relayed to the neocortex, and project to the primary sensory areas in an area-specific manner. The functional specializations of the primary sensory areas are defined by, and dependent upon, TCA input. TCA projections exhibit area-specificity throughout their development, and the gradual developmental differentiation of cytoarchitecture and connections that distinguish cortical areas depends to a large extent upon area-specific TCA input. In addition, experimental manipulations have demonstrated that the neocortex exhibits considerable plasticity in the development of area-specific features, and have implicated TCA input as a major influence controlling this plasticity (Chenn et al., 1997).
Evidence for the genetic regulation of arealization has begun to emerge over the past few years (Monuki and Walsh, 2001; Ragsdale and Grove, 2001; O’Leary and Nakagawa, 2002). Indirect evidence included descriptions of graded or restricted patterns of gene expression across the ventricular zone or the cortical plate that are established and maintained in the embryonic neocortex independent of TCA input, and therefore likely controlled by mechanisms intrinsic to the dorsal telencephalon (Miyashita-Lin et al., 1999; Nakagawa et al., 1999). Genes that regulate arealization presumably confer positional identities to cortical cells and regulate the expression of axon guidance molecules that control the area-specific targeting of TCAs. Two genes proposed to regulate arealization are the homeodomain transcription factor Emx2 and the paired-box transcription factor Pax6 (O’Leary et al., 1994). In mouse, Emx2 is expressed in a low rostrolateral to high caudomedial gradient (Gulisano et al., 1996) and Pax6 in a high rostrolateral to low caudomedial gradient (Stoykova and Gruss, 1994; Stoykova et al., 1996) across the ventricular zone of the embryonic neocortex.
Recent loss-of-function studies have provided evidence for a role for EMX2 and PAX6 in arealization by analyzing Emx2 and Pax6 (small eye; sey) mutant mice (Bishop et al., 2000, 2002; Mallamaci et al., 2000). Changes in marker expression and patterns of area-specific TCA projections suggested that rostral–lateral areas are expanded, whereas caudal–medial areas are reduced in Emx2 mutants (Bishop et al., 2000, 2002; Mallamaci et al., 2000). Marker analyses of Pax6 mutants show the opposite changes to those in Emx2 mutants, suggesting that rostral–lateral areas are reduced and caudal–medial areas are expanded in Pax6 mutants (Bishop et al., 2000, 2002). Thus, Emx2 appears to preferentially impart caudal and medial area identities, and Pax6 preferentially imparts rostral and lateral identities. Analyses of Emx1 single mutants and Emx1/Emx2 double mutants suggest that EMX1 does not regulate arealization (Bishop et al., 2002), despite its close similarity in sequence and expression to EMX2 (Simeone et al., 1992a,b; Gulisano et al., 1996). To date, gain-of-function analyses of the roles of EMX2 and PAX6 in arealization have not been reported.
A goal of this study was to use a replication defective recombinant adenovirus (AdV) to ectopically express Emx2 in the embryonic neocortex as a gain-of-function approach to study its role in the area-specific targeting of TCAs, using the rat as a model. We chose to use an AdV because it infects both mitotically active and postmitotic cells in vivo and transcription of the virally transferred genes begins almost immediately after infection (Moriyoshi et al., 1996; Tamamaki et al., 2001). Cells infected in vivo with adenoviral vectors express very high levels of the transgene, typically for 5–10 days postinfection, with a gradual decline in expression level (Verma and Somia, 1997). As a prelude to this experimental phase of our study, we cloned the cDNA for the full-length coding region of rat Emx2, since when this project began, coding sequence had been reported for only the homeodomain region of mammalian Emx2 and some 3′ and 5′ flanking sequence (Simeone et al., 1992a,b). In addition, we examined the expression patterns of Emx2 in rat, focusing on the cerebral cortex during the period of cortical neurogenesis, to demonstrate its patterned expression as reported in mouse [for a review, see Cecchi (Cecchi, 2002)]. We also carried out an analysis of the tangential and laminar distribution of AdV-infected cortical cells following an AdV injection into the lateral ventricle of the cerebral hemisphere at specific stages during embryonic cortical neurogenesis. Finally, we analyzed the effect of AdV-mediated ectopic Emx2 expression on the intracortical pathfinding and areal targeting of TCAs.
Materials and Methods
Timed-pregnant Sprague–Dawley rats were obtained from Harlan. The day of implantation is considered embryonic day (E) 0. The first 24 h postnatal is postnatal day (P) 0.
Isolation and Sequencing of Emx2 cDNA Clones
Primers were designed to the conserved regions of the published human and mouse Emx2 homeodomain sequence (5′ GCGGATCCAAAGCGG-ATTCGAAC and 3′ CCGGAATTCTGAGCCTTCTTCCTC) deposited in GenBank (Simeone et al., 1992b). These primers had introduced restriction enzyme sites, BamHI and EcoRI, respectively. To obtain template for PCR reactions, RNA from embryonic day 16.5 (E16.5) whole rat brain was treated with DNase1, poly-A selected with Oligo-dT latex beads (TaKaRa), and then cDNA synthesized using Superscript II (Gibco) following the manufacturer’s protocols. PCR reactions (50 μl) were prepared containing 1X PCR buffer (15 mM NH4SO4, 60 mM Tris pH 8.5, 2.0 mM MgCl2), 0.5 mM of each primer, 0.25 mM each dNTP, and 0.8 U of Taq polymerase. The mix was heated to 90°C, then 0.5 μl of cDNA template added and cycled 35X (95°C 30 s, 54°C 30 s, 72°C 2 min), with a final 72°C extension step for 10 min before cooling to 4°C. PCR products were analyzed on 1.4% agarose gels containing ethidium bromide. The 238 bp PCR product was digested with BamHI and EcoRI, subcloned into pBluescript (Stratagene), and sequenced from both directions using T3 and T7 primers, confirming it as Emx2. Preparative quantities of the insert were isolated, labeled with 32P, and used to screen 1 × 106 plaques of a rat E18 whole brain cDNA library (courtesy of Prof. H. Nawa, CSH Beckman Laboratory) constructed in a l ZapII vector (Stratagene). Nitrocellulose filters were hybridized in 5× SSPE, 5× Denhardt’s solution, and 0.5% SDS at 55°C overnight. The filters were washed twice in 2× SSPE with 0.5% SDS at 55°C for 30 min and exposed to X-ray film (X-OMAT, Kodak) overnight. Eight positive clones were obtained, one of which contained the entire open reading frame. The 1.5 kb insert from this clone was ligated into the EcoRI and XhoI sites of pBluescript and sequenced entirely in both directions using the dideoxy chain termination method.
Total-RNA samples were isolated from CNS tissue derived from different embryonic and postnatal ages of rats using the method of Chomczynski and Sacchi (Chomczynski and Sacchi, 1987). RNA (30 mg per well) was electrophoresed on a 0.8% agarose gel, blotted onto a nitrocellulose membrane overnight (Sambrook et al., 1989), and UV cross-linked. Blots were stained with 0.04% (w/v) methylene blue in 0.5 M sodium acetate, pH 5.2 for 45 s, followed by destaining in water for 2 min. Probes were generated by excising full-length coding regions of Emx2 from cDNA plasmids, gel-purifying the fragments with Geneclean III (Bio 101), and labeling with 32P using the Prime-a-Gene labeling kit (Promega). Pre-hybridization and hybridization were performed in a solution of 0.5 M Na2HPO4, 1% bovine serum albumin, 1 mM EDTA, 5% SDS and 20% formamide at 55°C. Probes were added at a concentration of 106 dpm/ml, hybridized overnight, and washed in decreasing concentrations of sodium succinate with 0.1% SDS (Sambrook et al., 1989) The blots were exposed to X-ray film (X-OMAT, Kodak) for 2.5 days. Films were scanned and figures were prepared directly from these scans using Adobe Photoshop software (Adobe).
In Situ Hybridization and Densitometry Analysis
Serial sections of brains were cut either in sagittal or coronal planes at 15 mm on a cryostat, and four to six adjacent series thaw-mounted on 3-aminopropyltriethoxysilane-coated slides (Sigma) and allowed to dry briefly before storing at –70°C prior to in situ hybridization. Emx2 sense and antisense 35S- and digoxigenin-labeled (DIG) cRNA probes were produced using a 780 bp template of the full-length coding region. Emx1 sense and antisense digoxigenin-labeled (DIG) cRNA probes were produced using a 771 bp template of the coding region (a kind gift from J.J.A. Contos and J. Chun). Plasmids containing the cDNA templates were linearized and transcription reactions with T7 or T3 polymerase (Promega) were carried out in the presence of [35S]UTP (Amersham) or digoxigenin-UTP. The template was degraded with RNase-free DNase (Promega) and extracted with phenol–chloroform. For in situ hybridization using radiolabeled probes, tissue sections were processed as described by Bishop et al. (Bishop et al., 2002). Sections were apposed to autoradiographic film (Hyperfilm, Amersham) and exposed for durations ranging from 6 h to 3 days. All slides for each individual probe were apposed to the same film for each exposure duration (6 h, 12 h, 1 days, 2 days, 3 days). Films were examined using a stereomicroscope and photographed with a digital camera (Kodak DCS 420) mounted on a Nikon Optiphot enlarger. Autoradiographic images were captured digitally using a Kodak DCS 420 camera and densitometric analyses performed using Image software (NIH, v1.6). Measurements of specific binding within individual cortical regions or laminae were obtained by subtracting non-specific binding, determined in near-adjacent sections incubated with sense probes, and were expressed in relative terms to reveal both temporal and spatial profiles in the levels of gene expression in each plane of section. Sections processed for in situ hybridization using DIG-cRNA probes were as described by Nakagawa et al. (Nakagawa et al., 1999). Sections were examined and photographed under an optical microscope (Zeiss Axiophot) and photographed with a digital camera (Kodak DCS 420).
Construction, Propagation and Injection of Recombinant Adenoviruses
We used a human adenovirus type 5 (AdV) deleted of sequences in the non-essential E3 region, and in the E1A/E1B region, impairing the ability to replicate in non-permissive cells (Miyake et al., 1996). Construction and propagation of the adenoviral vectors was done essentially as described in Moriyoshi et al. (Moriyoshi et al., 1996), with the exception that we used an internal ribosome entry site (IRES) from the encephalomyocarditis virus to generate dicistronic expression constructs that produce independent translation of two proteins from the single transgene transcript (Ghattas et al., 1991; Martinez-Salas, 1999). Expression was under the control of the strong and ubiquitous cytomegalovirus/chick β-actin hybrid promoter (CAG) (Niwa et al., 1991). Expression constructs were sub-cloned into the unique Swa1 site of the cosmid shuttle vector pAdex1w or pAdex1CAwt. To create recombinant adenoviruses, 8 μg of the cosmid shuttle vector carrying our expression construct together with 1 mg of adenoviral genomic DNA tagged at both ends with a 55 kDa terminal protein were transfected into 70–80% confluent HEK 293 cells using the Lipofectamine method (Life Technologies) in serum-free medium and, 12–16 h later, 1 ml of DME supplemented with 10% FBS was added. Twenty-four hours after transfection, the cells were plated with varying dilutions of untransfected 293 cells in DMEM 5% FBS. Cells were cultured for 2–3 weeks and observed for the presence of dying cells, indicating viral propagation. Positive wells were selected, amplified, and analyzed for the presence of the LacZ reporter gene. Restriction enzyme mapping on the viral DNA confirmed that the recombinant AdV contained the appropriate expression construct. Recombinant viruses were purified over a CsCl concentration gradient, dialyzed and stored at –70°C in 10% glycerol. Recombinant adenoviruses were injected in utero in the lateral ventricle as described (Luskin et al., 1988; Austin and Cepko, 1990).
Reporter Gene Detection and Axon Labeling
We used two reporter genes, green fluorescent protein (GFP) and LacZ, which encodes the enzyme, β-galactosidase (β-gal). GFP is a bio-luminescent protein that fluoresces green when illuminated with blue light (Chalfie et al., 1994). For X-gal histochemical detection of β-gal, brains were perfused in 2% PF, 0.3% gluteraldehyde and either sections or whole embryos were incubated overnight at 37°C with the substrate 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside at 0.1% (XGal; Diagnostic Chemicals), 2 mM MgCl2, 5 mM EGTA, 0.01% (w/v) sodium desoxycholate, 0.2% (w/v) Nonidet P-40, 5 mM K3Fe(CN)6 and 5 mM K4Fe(CN)6.6H2O. For immunohistochemical localization of β-gal, brains were fixed in 4% PF, cryoprotected in 30% sucrose and cryosectioned. Sections were blocked in PBT, 10% normal calf serum, 0.1% thimosol, 6% H2O2, washed in PBS, 1% GS and incubated in a β-gal polyclonal antibody (Cappel) overnight at 4°C. Sections were washed three times in PBS, 1% GS, incubated with a biotinylated goat anti-rabbit antibody for 2 h, washed in PBS and incubated in a fluorochrome or in a streptavidin-peroxidase-conjugated complex. For the latter, staining was visualized with DAB and nickel intensification.
The fluorescent, lipophilic dye, DiI (1–5% wt/vol in DMF, Molecular Probes) was used as an anterograde axon tracer to label TCAs in PF-fixed brains as described (De Carlos and O’Leary, 1992; Braisted et al., 1999). For this, the brainstem was bisected at the midbrain/diencephalic junction and a crystal of DiI was targeted to the dLG or VP. Crystal placement was later verified in bisbenzimide counterstained sections through the dorsal thalamus.
Sequence and Expression Analyses of Rat Emx2
Cloning and Northern Blot Analysis
Using oligonucleotides designed to the published partial Emx2 sequence for mouse (Simeone et al., 1992a,b), we generated PCR products specific for Emx2, and used them as probes to screen a rat E18 whole brain cDNA library to obtain cDNA for the full-length coding region (Fig. 1A), as well as additional 3′ and 5′ untranslated regions (data not shown). Northern blots show a single band for Emx2 transcripts of ∼2.8 kb (Fig. 2), as previously reported (Simeone et al., 1992b). Based upon the rat sequence, we were able to define a predicted cDNA sequence for the coding region of mouse Emx2 (Fig. 1A) through analysis of mouse genomic DNA sequence in the ‘Mouse HTGS Database’ (accession no. AC098733.1). Figure 1A shows a comparison of the cDNA sequences for the full-length coding regions of Emx2 for rat and mouse, as well as published sequence for human (Noonan et al., 2001) and chick (Bell et al., 2001). Figure 1B shows comparisons between these species of the predicted amino acid sequences for EMX2 protein.
The predicted amino acid sequence of the homeodomain region of rat EMX2 (Fig. 1) is 100% identical to that for EMX2 of mouse (Simeone et al., 1992b), human (Noonan et al., 2001), chick (Bell et al., 2001), zebrafish (Morita et al., 1995) and dogfish (Derobert et al., 2002). Comparison of predicted amino acid sequences for the full protein shows that rat and mouse EMX2 are 100% identical. Thus, the predicted amino acid sequences indicate that the EMX2, and in particular the homeodomain, is highly conserved across species.
Laminar Expression of Emx2 Compared to Emx1
We examined the laminar expression patterns of Emx2 compared to Emx1 in rat brain throughout cortical neurogenesis (Fig. 3). At E12, when cortical neurogenesis begins with the generation of preplate neurons (i.e. the Cajal–Retzius neurons of the marginal zone and subplate neurons), the expression of both genes appears limited to the ventricular neuroepithelium. At later stages (i.e. after E14), when cortical plate neurons are being generated and are migrating through the intermediate zone to the cortical plate, the spatial and temporal expression patterns of Emx1 and Emx2 become distinct. The expression of Emx2 appears confined to the neuroepithelium throughout embryonic development [but see Mallamaci et al. (Mallamaci et al., 1998)], Emx1 is expressed at high levels in the transitional layers (sub-ventricular zone and intermediate zone) and cortical plate between E15 and E18. As the neuroepithelium thins near the end of cortical neurogenesis, Emx2 expression diminishes and becomes confined to the deeper aspect of the neuroepithelium, and later to the ependymal lining of the lateral ventricle (data not shown). At E19, Emx1 remains expressed most highly in the neuroepithelium, but in a broader radial domain than Emx2, and at moderate levels in the cortical plate and deeper aspects of the subplate.
Graded Expression of Emx2 in the Embryonic Rat Neocortex
In mice, Emx2 expression begins in the rostrolateral neural plate around E8.5 (corresponding to about E10 in rat) (Simeone et al., 1992a,b; Shimamura et al., 1995). However, we limited our analyses to the period of cortical neurogenesis, which in rat begins on E12 and is completed around E20 (Bayer and Altman, 1991). Emx2 expression was examined in sagittal (Fig. 4) and coronal (Fig. 5) sections of embryonic rat brain, and found to exhibit graded expression along both rostral–caudal and medial–lateral axes of the neocortex. At E12, levels of Emx2 were already graded, being higher in caudal cortex relative to rostral cortex; the graded expression continued throughout embryonic cortical neurogenesis. A modest high caudal to low rostral graded expression was observed at E20 (Fig. 6B) despite the substantial reduction in the absolute level of Emx2 expression, and its restriction to the deepest part of the neuroepithelium (Fig. 3). Emx2 is also expressed in a high medial to low lateral gradient in the embryonic cortex during neurogenesis (Fig. 5). At E20, near the end of cortical neurogenesis (Bayer and Altman, 1991), the M–L gradient is no longer evident (Fig. 6C).
We performed a densitometric analysis to quantify the relative temporal changes in expression levels of Emx2 at successive embryonic ages, as well as its graded expression. Levels of Emx2 expression in the neuroepithelium were highest at E12, exhibited a precipitous reduction between E13 and E14, then a gradual reduction thereafter (Fig. 6A). Densitometry measurements confirmed the graded expression patterns along both the rostral–caudal axis (Fig. 6B) and medial–lateral axis (Fig. 6C) of the embryonic rat neocortex. Thus, the strongest graded expression of Emx2 is during the generation of subplate, marginal zone, and the deep layers of the cortical plate.
Expression Efficiency of Dicistronic Vectors
To ectopically express Emx2 in embryonic neocortex, we constructed a replication defective recombinant adenovirus containing an expression cassette with a CAG promoter, cDNA for the coding region of Emx2, an IRES sequence, and as a reporter, LacZ, which encodes β-galactosidase (Fig. 7A). Although a single RNA transcript is generated, the IRES enables the mRNA for both cDNAs to be independently translated into two distinct proteins (Ghattas et al., 1991; Martinez-Salas, 1999). Reportedly, the mRNA upstream to the IRES is translated at roughly twice the level of the downstream mRNA. As a control to show that both gene products are efficiently translated in the adenoviral constructs that we used, we made a dicistronic expression construct with two distinct reporters, GFP and β-gal, by inserting cDNA encoding GFP in place of Emx2, upstream to the IRES, and leaving the downstream LacZ intact. In 293T cells transfected with the CAG/GFP/IRES/LacZ pA vector, the transfected cells express both GFP (Fig. 7B) and β-gal (Fig. 7C), indicating that the GFP/IRES/LacZ mRNA transcript is efficiently translated into two distinct proteins. Similarly, the CAG/Emx2/ IRES/LacZ pA AdV gives strong expression of β-gal in both 293T cells (data not shown) and in vivo in neurons (Fig. 9). Although we do not directly localize EMX2 protein, because we have been unable to obtain a specific antibody, our control experiments, and those of others, show that the mRNA encoding the protein upstream to the IRES is translated more efficiently that the one downstream to the IRES (Ghattas et al., 1991), leaving no doubt that EMX2 protein is produced at least as abundantly, and likely at twice the levels, of β-gal in the infected cells.
Temporal–Spatial Parameters of Adenoviral Infection in Developing Cortex
We have previously shown that a replication defective recombinant AdV containing a CAG/GFP expression construct injected into postnatal rat cortex results in the infection and intense labeling of neural cells, including neurons and glia (Moriyoshi et al., 1996). However, in the present study, our aim was to overexpress Emx2 in embryonic neocortex, preferably in progenitor cells within the neuroepithelium and their progeny. As a prelude to our experimental AdV studies, we assessed the effectiveness and spread of cortical AdV infections. For these studies, we injected a replication defective recombinant AdV containing a CAG/LacZ pA expression construct unilaterally into the lateral ventricle of the cerebral hemispheres at sequential times over embryonic cortical neurogenesis. Injections were done at E13.5, E15.5, E16.5 and E17.5, and the brains were fixed at P4 or P12, and reacted for β-gal to localize AdV-infected cells. The distribution and density of labeled cells across the tangential axes of the cortex varied among cases, likely due to the effectiveness and placement of the AdV injection in the lateral ventricle. However, an injection at each of these ages often resulted in a high density of labeled cells broadly distributed tangentially across the cortex (Fig. 8). In contrast, the laminar distribution of infected cells varied in a reproducible manner that depended upon the age of AdV injection (Fig. 8).
An AdV injection at E13.5 resulted in a high density of infected cells in the subplate and the deeper layers of the cortical plate, and moderate to low density of infected cells in more superficial layers of the cortical plate (Fig. 8A–D). Interestingly, a low density of infected cells was seen in the marginal zone (Fig. 8B,C). These cells are likely to be Cajal–Retzius neurons, the only postmitotic cortical neurons reported to express Emx2 (Mallamaci et al., 1998), which together with subplate neurons, are the earliest generated cortical neurons. These two populations of early-generated neurons initially form the preplate, which is later split by the cortical plate into the marginal zone and subplate. The pattern of infected cells in brains injected at E15.5 had similarities to the E13.5 cases, but the density of labeled cells in the subplate was lower, and few if any labeled cells were detected in the marginal zone. An AdV injection done at E16.5 (Fig. 8H–K) or E17.5 (Fig. 8L–N) resulted in labeling patterns very distinct from the earlier injections. In both, a high density of infected cells were localized to the superficial layers of the cortical plate, whereas the subplate, marginal zone and deep layers of the cortical plate were virtually devoid of infected cells. The high density band of infected cells in the superficial layers was somewhat broader with an E16.5 injection (Fig. 8J,K) than an E17.5 injection (Fig. 8N).
The laminar distribution of cells infected with AdV injected at successive ages reflects the inside-out generation of cortical neurons, and is reminiscent of the labeling patterns observed using birthdating markers injected at the same ages (Bayer and Altman, 1991). The early injections preferentially label the earliest generated cortical neurons, those that form the preplate (marginal zone and subplate neurons) and the deepest layers of the cortical plate, and later injections label the later generated superficial layer neurons but not the earlier generated cells. These findings indicate that AdV injected into the lateral ventricle has limited spread, if any, into the cortical wall. In addition, they suggest that such AdV injections preferentially infect progenitors in the ventricular zone, which leads to a preferential labeling of their immediate progeny, and possibly cells that have recently become postmitotic but have yet to move far from the ventricular surface.
Role for EMX2 in Regulating Areal Targeting of Thalamocortical Axons
To address the role of EMX2 in regulating the area-specific targeting of TCAs, we used a replication defective, recombinant AdV containing the CAG/Emx2/IRES/LacZ pA dicistronic expression construct (Emx2–AdV), and as a control, recombinant AdVs containing either a CAG/GFP/IRES/LacZ pA dicistronic expression construct or a CAG/LacZ pA expression construct (Fig. 7A). We hypothesized that ectopic domains of Emx2–AdV expression would perturb the areal targeting of TCAs in a manner that would reflect that high levels of EMX2 ‘caudalize’ rostral areas of neocortex. We focused on the TCA projection of the dorsal lateral geniculate nucleus (dLG), which normally projects specifically to the primary visual area (V1), a caudal-medially located area.
Recombinant AdV injections were targeted in utero to the lateral ventricle of E13.5 rats (developmentally equivalent to E11.5–E12 mice), during the generation of subplate and deep layer cortical neurons (Bayer and Altman, 1991; Koester and O’Leary, 1993). The infected animals were fixed between hours before birth (E22/P0) to P2, and processed for X-gal histochemistry to visualize the β-gal reporter that marks the AdV-infected cells. Cortical domains of AdV infection were correlated with the intracortical pathfinding and targeting of TCAs revealed by anterograde labeling using crystals of DiI placed into the dLG, or into the ventroposterior (VP) thalamic nucleus that projects to the primary somatosensory area (S1). In some AdV-infected cases, the cortex was reduced in radial thickness and the lateral ventricles were expanded. This effect was independent of which of the three recombinant AdVs was injected, and likely non-specific, since it was not evident in most of the brains heavily infected with any of the three recombinant AdVs used. These brains were not included in this analysis.
Infection with the control recombinant AdVs had no effect on the pathfinding and area-specific targeting of TCAs. We analyzed 11 control infected cases with good DiI labeling of TCA axons; six of these had high levels of AdV infection in the cortex and five had moderate levels of cortical AdV infection. Laminar patterning and differentiation of the cortex appeared unaffected in these brains (data not shown). An example of a control case with high levels of AdV infection in the cortex and good DiI labeling of the TCA projection from the dLG is illustrated in Figure 9A–D. The labeled TCAs passed through the striatum within the internal capsule, entered the cortex as a relatively tight bundle, turned dorsally, accumulated in the subplate and extended tangentially within it caudally to the occipital cortex (Fig. 9B). At the rostral–lateral location where they first entered the cortex, the labeled dLG axons did not invade the cortical plate of presumptive S1 (Fig. 9B,D). However, within the occipital cortex, the labeled TCAs heavily invaded the cortical plate of presumptive V1 in an area-specific manner indistinguishable from normal (Fig. 9C).
In contrast, infection with the Emx2–AdV had a substantial effect on the pathfinding and area-specific targeting of TCAs. We analyzed 15 cases with high levels of Emx2–AdV infection in the cortex; four of these cases had good DiI labeling and 11 had sparse DiI labeling of the TCA projection. All four cases with good DiI labeling exhibited aberrant TCA projections within the cortex; two were strongly aberrant and two had similar defects but less pronounced. Nine of the 11 sparsely labeled cases had similarly aberrant TCA projections, but again less pronounced than the strongly aberrant cases; we suspect that at least in some, if not all, of these instances the milder appearance of the aberrancy may be due to fewer TCAs labeled. As in the control AdV-infected cases, no obvious defects were apparent in the laminar patterning and differentiation of the cortex in the Emx2–AdV infected brains (data not shown).
An example of a case with high levels of Emx2–AdV infection in the cortex and good DiI labeling of the TCA projection from the dLG is illustrated in Figure 9E–I. In this case, the Emx2–AdV infection was particularly high in lateral cortex (Fig. 9E) at the position that TCAs pass from the striatum into the cortex (Fig. 9F,H). The appearance of the labeled TCAs as they extend through the striatum within the internal capsule appeared normal (Fig. 9F). However, upon entering the cortex, their path-finding became highly aberrant in several respects (Fig. 9F–I). First, rather than turning dorsally and accumulating on the subplate as a tight bundle, the TCAs turned but splayed out into several bundles of fascicles that extended tangentially with an abnormally broad distribution that included not only their appropriate pathway centered on the subplate, but also layer 6 and deep into the intermediate zone (Fig. 9F,H,I). In addition, many of the labeled TCAs left these aberrant bundles, turned into the cortical plate, extended radially through it without branching, and accumulated in the marginal zone. Within the marginal zone, the labeled TCAs turned dorsally, and extended tangentially within it (Fig. 9H,I). These aberrancies were coincident with dense domains of deep layer cells infected with the Emx2–AdV (Fig. 9H). Only a proportion of the labeled TCAs continued caudally toward their target area, V1, and only a small proportion appeared to reach it (Fig. 9G). These findings indicate that TCAs labeled from the dLG aberrantly invade an inappropriate cortical area coincident with high levels of deep layer Emx2–AdV infection. This targeting inaccuracy is not seen during normal development, and none of these intracortical pathfinding defects in the Emx2–AdV infected brains are seen at any stage of development for any population of TCAs in normal or control AdV-infected brains.
In six additional rats injected with the CAG/Emx2/IRES/LacZ pA AdV, good DiI labeling of TCAs was obtained and Emx2–AdV infection was detected in the cortex, but the density of β-gal positive, Emx2–AdV infected cells was low relative to the moderate to high Emx2–AdV infected cases described above with aberrant TCA targeting. The pathfinding and areal targeting of the labeled TCAs in these cases were indistinguishable from control infected and normal rats (data not shown). In conclusion, our findings show that dLG TCAs aberrantly invade the cortical plate of rostral–lateral areas (S1) coincident with domains of moderate to high levels of ectopic Emx2–AdV expression that they pass beneath en route to V1; they do not invade domains of low levels of ectopic Emx2–AdV expression or domains of high levels of control AdV infection. These results suggest that higher levels of EMX2 in the ectopic domains in rostro-lateral cortical areas can specify them to have characteristics, such as thalamic innervation from the dLG, that are normally associated with caudo-medial cortical areas, such as V1.
A major goal of our study was to use a gain-of-function test of the role of EMX2 in regulating the area-specific development of TCA projections. For this, we used a replication defective recombinant AdV with a dicistronic expression construct under the control of the strong and ubiquitous CAG promoter (CAG/ Emx2/IRES/LacZ pA AdV), to ectopically express Emx2. Similar viral constructs lacking cDNA for Emx2 were used as controls. Partial cDNA sequence for Emx2, and the related gene Emx1, the vertebrate homologues of the Drosophila empty spiracles gene, were originally reported by Simeone et al. (Simeone et al., 1992a,b). Because at the time we initiated this study, cDNA for the full-length coding region of mammalian Emx2 was not available, we cloned and sequenced rat Emx2 cDNA. The clone was ∼1500 kb and contained an open reading frame of 780 bp, as well as 5′ and 3′ untranslated regions. Northern analysis of total RNA from E16 rat cortex using radiolabeled probes of the full-length coding region for Emx2 reveals a transcript size of ∼2.8 kb, as previously reported for mouse (Simeone et al., 1992b). Based upon the rat sequence, we defined a predicted sequence for the coding region of mouse Emx2 (Fig. 1A) through analysis of a mouse genomic DNA database. Sequence has been recently published for the full-length coding region of Emx2 for human (Noonan et al., 2001), zebrafish (Morita et al., 1995), chicken (Bell et al., 2001) and dogfish (Derobert et al., 2002). Within the homeodomain of these six species, the predicted amino acid sequence is 100% identical; they are 80% identical to the homeodomain of Drosophila empty spiracles. For the full-length coding region, the predicted amino acid sequence of rat and mouse EMX2 is 100% identical; each is 98.8% identical to human, 96.8% to chick, 93.1% to zebrafish and 86.6% to dogfish.
Our in situ hybridization analyses in embryonic rat confirm previous reports in mouse that Emx2 is predominantly expressed in the forebrain (Simeone et al., 1992a,b), and that within the cortex, it exhibits a graded expression pattern (Gulisano et al., 1996; Mallamaci et al., 1998). Throughout cortical neurogenesis in rat and mouse, Emx2 is expressed in a gradient with highest levels caudally and medially, and lowest levels rostrally and laterally. As cortical neurogenesis ends, Emx2 expression ceases. In addition, the laminar expression patterns in rat neocortex of both Emx2 and Emx1 are similar to reports in mice [for a review, see Cecchi (Cecchi, 2002)]. Emx2 is restricted to progenitors in the neocortical neuroepithelium (except for Cajal–Retzius neurons), and Emx1 is expressed in progenitors as well as in the postmitotic neurons that they generate (present study) (Gulisano et al., 1996; Mallamaci et al., 1998). Thus, Emx2 is expressed highest in progenitors in the ventricular zone that will generate caudal and medial areas of neocortex, such as visual areas, and lowest in rostral and lateral areas, such as sensorimotor areas. Interestingly, the strongest graded expression of Emx2 appears to be at earlier stages of cortical neurogenesis, during the generation of subplate, marginal zone, and the deep layers of the cortical plate. This finding suggests that EMX2 regulated positional information may be more prominent in the neurons of these layers. Intuitively, these layers would seem to most require such information because the subplate is likely involved in establishing area-specific TCA projections, and layers 6 and 5 are the sources of area-specific cortical efferent projections to subcortical targets.
In the adult rodent, TCAs from the principal dorsal thalamic sensory nuclei project to specific primary sensory neocortical areas (Hohl-Abrahao and Creutzfeldt, 1991). The development of area-specific TCA projections occurs in two phases: (i) the targeting phase: after entering the cortex, TCAs grow tangentially along an intracortical pathway centered on the subplate layer to reach their appropriate cortical area (Ghosh and Shatz, 1993; Miller et al., 1993; Bicknese et al., 1994); and (ii) the invasion phase: after reaching their appropriate target area, TCAs extend collaterals superficially into the overlying cortical plate (Catalano et al., 1991; Ghosh and Shatz, 1993). TCAs from the principal sensory thalamic nuclei target and invade their appropriate cortical areas in a precise area-specific manner; TCAs rarely overshoot their appropriate cortical areas or make gross directional errors, and only invade the cortical plate of the appropriate cortical area (Crandall and Caviness, 1984; Miller et al., 1993). In addition, TCAs grow past, rather than invade, the cortical plate overlying regions of the subplate pharmacologically depleted of neurons (Ghosh et al., 1990; Ghosh and Shatz, 1993). These findings suggest that the axon guidance molecules that control the area-specific targeting of TCAs are associated with the subplate.
In addition, the ‘handshake hypothesis’ postulates a subplate axon-mediated mechanism for TCA guidance, whereby orderly arrays of subplate axons and TCAs meet in the internal capsule, interact with their corresponding area-specific partners, and serve as topographic scaffolds for one another (Molnar et al., 1998). Recent evidence supporting this possibility comes from the analysis of Tbr1 and Gbx2 mutants (Hevner et al., 2001, 2002), as well as Emx1/Emx2 double mutants (Bishop et al., 2003). Gbx2 expression in the dorsal thalamus is required for the production of a normal TCA projection (Miyashita-Lin et al., 1999). Despite the lack of Gbx2 expression in the cortex, corticothalamic axons fail to reach the dorsal thalamus in Gbx2 mutants (Hevner et al., 2002). Emx1 and Emx2, as well as Tbr1, are expressed in the embryonic cortex and not in the dorsal thalamus. In Emx double mutants (Shinozaki et al., 2002; Bishop et al., 2003) and Tbr1 mutants (Hevner et al., 2001) cortical axons do not reach the dorsal thalamus and TCAs fail to reach the cortex. Regardless of whether the guidance information for TCAs is present on subplate axons, within the subplate layer itself, or both, the precise mapping of TCA connections is likely to involve guidance molecules expressed in graded or areal patterns. Regulatory proteins such as EMX2 are candidates to control the differential expression of these guidance molecules.
The modified AdV5 used in our studies infects cells through a receptor-mediated mechanism that involves the Coxsackie Adenovirus Receptor (CAR) and the avβ3 or avβ5 integrin receptors (Nemerow, 2000). CAR is highly expressed in progenitors in the cortical ventricular zone, and is moderately expressed by postmitotic cortical neurons (Tamamaki et al., 2001). Interestingly, it has been reported that only one daughter cell of an AdV-infected progenitor in the cortical ventricular zone will inherit the AdV vector (Tamamaki et al., 2001), which does not replicate during cell division due to the episomal localization of the AdV vector (Verma and Somia, 1997). This is consistent with our finding that the laminar distribution of cells infected with the replication defective recombinant AdV injected at successive embryonic ages reflects the temporal sequence of generation of cortical neurons (Bayer and Altman, 1991). The early AdV injections done at E13.5, the age that we injected the Emx2–AdV to study the area-specific targeting of TCAs, preferentially labeled the earliest generated cortical neurons, including subplate neurons, neurons in the marginal zone, and the deepest layers of the cortical plate. These findings indicate that AdV injected into the lateral ventricle preferentially infects cells that have recently become postmitotic and have yet to move far from the ventricular surface, and/or progenitors in the ventricular zone, which leads to a preferential labeling of their immediate progeny. AdV vectors can infect cells in vivo, causing them to express very high levels of the transgene within hours after infection (Moriyoshi et al., 1996; Verma and Somia, 1997). In somatic tissue, this expression usually lasts for 5–10 days postinfection due in part to a virally mediated immune response. Thus, the time course of high expression of the Emx2 transgene would be appropriate to alter area-specific information that EMX2 may regulate in subplate and cortical plate cells.
We hypothesized that the establishment of area-specific TCA projections is controlled by the graded position-dependent expression of regulatory genes, including Emx2. Higher levels of EMX2 expression impart more caudal cortical positional values, such as those normally associated with visual areas. Our findings show that TCAs labeled from the dLG, which would normally target the primary visual area (V1), aberrantly invade an inappropriate, rostral–laterally located cortical area (S1) coincident with high levels of deep layer Emx2–AdV infection. These results suggest that higher levels of EMX2 in the ectopic domains of Emx2–AdV expression in rostral–lateral cortical areas specify them to have properties normally associated with caudal–medial cortical areas, such as V1. This finding is consistent with loss-of-function analyses of the role of EMX2 in arealization in mutant mice deficient for Emx2 (Bishop et al., 2000, 2002; Mallamaci et al., 2000). Changes in the patterns of gene expression suggest that rostral–lateral areas are expanded, whereas caudal–medial areas are reduced in the mutant. Alterations in the organization of area-specific TCA projections are also consistent with this interpretation. Retrograde labeling from the neocortex of Emx2 mutants indicates an orderly expansion and caudal shift of the topographic TCA projection of VP, indicative of an expansion of its target area, S1, a caudal shift of the S1 border, and a contraction of V1 (Bishop et al., 2000; Mallamaci et al., 2000). Unfortunately, Emx2 mutant mice die soon after birth; thus, the adult anatomical and functional organization of the mutant cortex cannot be studied in these mice.
These findings suggest that during normal development, the graded expression of Emx2 controls through a regulatory cascade the differential expression across the neocortex of genes encoding cell surface or ECM molecules that target TCAs to their appropriate areas. Our findings suggest that the ectopic expression of Emx2 alters this regulation of guidance molecules for TCA targeting. In the Emx2–AdV infected brains, not only do dLG TCAs aberrantly invade rostral–lateral areas that ectopically express Emx2, but they exhibit other aberrancies as well. For example, within the ectopic domain of Emx2 expression, TCAs extend radially directly through the cortical plate and turn and extend dorsally within the marginal zone. This behavior is never seen during normal development. It may be due to persistent Emx2 expression in postmitotic deep layer neurons, such as layer 5 and 6 neurons, all of which at the ages analyzed extend their apical dendrites radially through the cortical plate and into the marginal zone (Koester and O’Leary, 1992). Thus, the ectopic expression of Emx2 within these neurons may lead to the overexpression of cell surface proteins that promote the growth of TCAs and provide a substrate for their extension through the cortical plate, which at early ages normally has inhibitory effects on the growth of TCAs (Tuttle et al., 1995).
EMX2 regulates arealization presumably by conferring positional identities to cortical cells and regulating their expression of axon guidance molecules that control the area-specific targeting of TCAs. A potentially analogous scenario is the graded expression in the developing chick optic tectum and rodent superior colliculus of the vertebrate engrailed genes, En-1 and En-2 (Gardner et al., 1988; Martinez and Avarado-Mallart, 1990; Martinez et al., 1991), which, like EMX2, are homeodomain transcription factors. Ectopic expression of En-1 and En-2 using recombinant retrovirus in embryonic chick tectum shows that these genes regulate the topographic targeting of retinal axons along the rostral–caudal axis of the tectum (Friedman and O’Leary, 1996; Itasaki and Nakamura, 1996), in part through their regulation of two membrane associated proteins, ephrin-A2 and ephrin-A5 that have a graded expression that parallels that of the engrailed genes (Logan et al., 1996; Shigetani et al., 1997). Ephrin-A2 and ephrin-A5 act as repellents for retinal axons and branches (Nakamoto et al., 1996; Monschau et al., 1997; Yates et al., 2001) and are required for the proper topographic mapping of retinal projections (Frisén et al., 1998; Feldheim et al., 1998, 2000). Although considerable progress has been made in recent years in defining the genetic and molecular control of TCA pathfinding from dorsal thalamus to the neocortex (Kawano et al., 1999; Miyashita-Lin et al., 1999; Tuttle et al., 1999; Zhou et al., 1999; Braisted et al., 2000; Bishop et al., 2003; Hevner et al., 2002; Lopez-Bendito et al., 2002), characterization of the area-specific targeting of TCAs within the neocortex has lagged. As in the visual system (O’Leary et al., 1999), area-specific TCA targeting is likely primarily controlled by graded guidance molecules, and may also be influenced by neural activity, since blockade of neural activity in cats results in aberrant areal targeting of TCAs (Catalano and Shatz, 1998).
Members of the cadherin family of cell adhesion molecules have been suggested to influence the development of area-specific TCA projections because the principal sensory thalamic nuclei and their target primary sensory areas show matching expression of cadherin-6, -8 and -11 (Korematsu and Redies, 1997; Suzuki et al., 1997; Inoue et al., 1998). Other candidates to control TCA targeting are the ephrins and their Eph receptors, which as described above, act as axon guidance molecules in many systems, in ways that resemble the mapping of area-specific TCA projections (O’Leary and Wilkinson, 1999). In the rhesus monkey, primary (V1) and secondary (V2) visual areas form reciprocal connections with adjacent dorsal thalamic nuclei, the dLG and pulvinar, respectively. In embryonic monkeys before TCA projections are established, EphA3, EphA6 and EphA7 and ephrin-A5 are expressed at higher levels in V1 than V2; these three EphA receptors are also highly expressed in overlapping graded patterns in pulvinar and to a lesser extent in dLG, whereas ephrin-A5 is highly expressed in the ventrolateral complex that projects to S1 rather than visual areas (Sestan et al., 2001). Similarly, in mice, ephrin-A5 is expressed in a medial to lateral gradient across S1, and EphA4 is expressed in a matching gradient across VP, which provides TCA input to S1 (Mackarehtschian et al., 1999; Vanderhaeghen et al., 2000). In vitro, ephrin-A5 repels VP axons. The ephrin-A5 expression in S1 has also been suggested to repel axons from medial dorsal thalamic nuclei, which express EphA5 and pass under S1 en route to their target areas in limbic cortex, which do not express ephrin-A5 (Gao et al., 1998). Another candidate TCA targeting molecule is the neurotrophin receptor (NTR), p75, which is expressed by subplate and layer 6 neurons throughout the period of TCA targeting in a graded pattern that resembles that of Emx2 (Mackarehtschian et al., 1999).
Interestingly, the restricted or graded patterns of expression of p75, EphA7, ephrin-A5, and cadherin-6 and cadherin-8, are all shifted or altered in Emx2 mutant mice in a manner consistent with a role for them in controlling the area-specific targeting of TCAs (Bishop et al., 2000, 2002; Mallamaci et al., 2000). The appropriate analyses have not been done to confirm the roles of some of these candidate TCA targeting molecules. Surprisingly, though, in ephrin-A5 knockout mice, VP axons properly target S1 and form an orderly map of the body, albeit one that exhibits a graded, topographic distortion with medial representations contracted and lateral representations expanded (Vanderhaeghen et al., 2000). In contrast, mice lacking p75NTR have diminished or absent innervation of V1 by TCAs from the dLG, a defect that is coincident with an altered morphology of subplate growth cones, and the mistargeting of some subplate axons (McQuillen et al., 2002). TCA projections to auditory and somatosensory areas of the cortex are normal in the p75NTR mutants, consistent with their lower levels of graded p75NTR expression.
Further studies will be required to confirm the roles of EMX2 in controlling not only the area-specific targeting of TCAs, but also the process of arealization of the neocortex. An important aspect of these future studies will be to identify the downstream targets of EMX2, and in particular cell surface ligands and receptors that control the pathfinding and area-specific targeting of TCAs.
This work was supported by NIH grant NS31558. A.L. was supported by a DFG fellowship, C.A. by the Sankyo Foundation of Life Science, and L.J.R. was a Lucille P. Markey Postdoctoral Fellow. We thank J.J.A. Contos and J. Chun for comments, advice, and Emx1 cDNA. We thank Inder Verma and members of his laboratory for advice on adenoviral vectors.