Laminar specificity is one of the most striking features of neocortical circuitry. To explore the molecular basis of this specificity, particularly in relation to thalamocortical connectivity, we searched for the genes expressed in the upper cortical layers by constructing a subtraction cDNA library that was enriched for genes expressed in layer 4 of perinatal rat somatosensory cortex. Differential screening, sequence analysis and in situ hybridization demonstrated that a new unc5 family member (unc5h4), deltex-like gene, stem cell factor (SCF) and myocyte-specific enhancer factor-2C (MEF-2C) were specifically expressed in layer 4 or layers 2/3–4 at postnatal day 7, by when laminar organization and fundamental cortical circuitries have been established. In terms of regional specificity, unc5h4 and SCF signals were stronger in sensory cortices, whereas MEF-2C and deltex-like gene were expressed rather uniformly in all neocortical regions. Analysis during development demonstrated that expression of these genes was pronounced between late embryonic and early postnatal developmental stages, except for MEF-2C expression, which continued in later stages. These results demonstrate that certain types of molecules including transcription factors, receptor and ligand molecules, are expressed specifically in the upper layers of the developing neocortex, suggesting a role in laminar specification of cortical cells and circuitry.
The neocortex is fundamentally composed of six cell layers, which are distinguishable by cellular morphology and the extrinsic and intrinsic connections they make (McConnell, 1989). An intriguing question is how cortical neurons differentiate into a particular laminar type, and are connected with a specific population of subcortical and cortical neurons. A plausible mechanism is that a set of transcriptional factors expressed in a given layer regulates laminar fate, and expression of downstream molecules including ligand and receptor molecules, is required for cell type specification and axonal guidance of cortical afferents and efferents. Therefore, a key approach is to identify the molecules that are expressed with laminar specificity.
Several molecules with lamina-specific expression patterns in the cortex have been identified. For instance, transcription factors of Otx1 and Id2 are primarily distributed in layer 5, and Tbr1, a T box gene, in layer 6 (Bulfone et al., 1995; Frantz et al., 1994; Hevner et al., 2001). Moreover, retinoid Z receptor (RZR-β) and chick ovalbumin upstream transcription factor 1 (COUP-TF1) are expressed rather specifically in layer 4 (Becker-Andre et al., 1994; Park et al., 1997; Liu et al., 2000). It has also been demonstrated that cadherin-6 and rCNL3 — homologous to G-protein-coupled receptors — are expressed in the upper layers (Suzuki et al., 1997; Chenn et al., 2001). Although molecular identification has progressed, a further understanding of gene expression in the developing cortex is necessary to pursue the mechanisms of laminar specification.
In the present study, we attempted to identify lamina-specific genes, focusing on thalamocortical connectivity. The thalamocortical projection exhibits typical laminar specificity (Jones, 1981; Gilbert, 1983) and has also been well characterized during development (Lund and Mustari, 1977; Ghosh and Shatz, 1992; Agmon et al., 1993; Kageyama and Robertson, 1993; Catalano et al., 1996; Molnár et al., 1998). To date, in vitro studies using organotypic co-cultures of the cortex with the thalamus have demonstrated that there is a target recognition mechanism by which thalamocortical axons recognize layer 4, their target (Yamamoto et al., 1989, 1992; Molnár and Blakemore, 1991, 1999; Bolz et al., 1992). Our previous investigations have further suggested that thalamocortical axon branching is induced by membrane-associated molecules in layer 4, whereas termination of axonal growth is regulated by growth-inhibitory molecules in layers 2/3 and 4 (Yamamoto et al., 1997, 2000a, 2000b). These findings suggest that the factors regulating thalamocortical axon targeting and differentiation of target layer cells are expressed specifically in the upper layers. Based on this hypothesis, we searched for the molecules expressed specifically in these layers, in particular, layer 4, by constructing a subtraction cDNA library. Four genes, including a new unc5 family member, were obtained. Their expression patterns were further investigated to gain an insight into their possible roles in thalamocortical connectivity and/or cortical cell identity.
Materials and Methods
Sprague–Dawley rats were used for all experiments (Nihon Animals, Osaka, Japan). The day of vaginal plug detection was designated as E0, and the day of birth as postnatal day 0 (P0).
Construction of a Subtraction cDNA Library
The whole brain was removed from P7 rats. Coronal slices (250 µm thickness) were cut with a microtome in ice-cold Hanks’ solution. Layer 4 strips, ∼500 µm in length, were dissected from the somatosensory cortex with a small scissors (Fig. 1). The cortical barrel structures, which were visible under a trans-illuminating microscope, were used as a landmark for layer 4 (Fig. 1). Virtually the same size of layer 5 strip was dissected beneath layer 4 (Fig. 1). Three or four pieces for each layer were collected, from which total RNAs were extracted (RNeasy Mini Kit, Qiagen, Tokyo, Japan). Approximately 10–20 ng/µl of total RNA (30 µl) was obtained. Layer 4 and layer 5 cDNAs were synthesized from ∼50 ng of these RNAs with reverse transcriptase (SMART PCR cDNA Synthesis Kit, Clontech, Tokyo, Japan). These cDNAs were amplified with a primer (AAGCAGTGGTAACAACGCAGAGT), and subjected to RsaI digestion for the following hybridization. The layer 4 DNA fragments were hybridized with an excess amount of the layer 5 DNA fragments (PCR Select cDNA Subtraction Kit, Clontech, Tokyo, Japan). Thereafter, unhybridized DNA fragments were further amplified, cloned into pGEM-T vectors (Promega, Tokyo, Japan), and stored as a subtraction cDNA library (layer 4 minus layer 5). Likewise, a reverse subtraction library (layer 5 minus layer 4) was produced.
Differential Screening and Sequence Analysis
To eliminate false-positive clones a differential screening was performed prior to in situ hybridization. After transformation of the subtraction cDNA library, colonies were picked randomly and grown in a 96-well plate. Each insert was amplified by polymerase chain reaction (PCR) from the bacterial culture solution. Amplified DNAs were denatured with 0.6 M NaOH and duplicated onto two nylon membranes. Two kinds of DNA probes (layer 4 and layer 5 probes) were produced with Digoxigenin (DIG)-labeled nucleotides (Roche, Tokyo, Japan) by amplifying inserts from the two subtraction cDNA libraries (layer 4 minus layer 5 and layer 5 minus layer 4). The duplicate membranes were subjected to hybridization with each probe and detected with chemilluminescence reaction. The clones showing >2-fold intensity with layer 4 probes were selected and subjected to in situ hybridization.
The selected clones were applied to sequence analysis with the plasmid-specific sequences. In some cases, extended DNA fragments were obtained by 5′ race method (Smart race kit, Clontech, Tokyo, Japan). In a few cases, a human brain cDNA library was used to obtain full-length cDNAs (Nagase et al., 1997).
In Situ Hybridization
DIG-labeled RNA probes were used for hybridization. To produce linearized templates for the synthesis of riboprobes, inserts in pGEM-T vectors were amplified by PCR using oligonucleotides that contain T7 and SP6 promoter sequences. The PCR products were purified (PCR purification kit, Qiagen, Tokyo), and in vitro transcription was carried out (DIG-RNA Synthesis Kit, Roche, Tokyo). Finally, these probes were purified with gel filtration columns (Quick Spin Columns, Roche, Tokyo) and kept at –80°C.
Rats were decapitated after anesthesia to obtain whole brains from postnatal animals (P0, P3, P7 and P14). Embryonic brains were taken from fetuses under deep anesthesia. The brains were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer for 1 h at room temperature and then 2 h at 4°C. After overnight incubation in PBS containing 20% sucrose, the brains were frozen and then cut into 10 µm sections (frontal or sagittal sections) with a cryostat.
Sections were refixed in 4% paraformaldehyde in 0.1 M phosphate buffer, washed with distilled water and 0.1 M triethanolamine, then acetylated in 0.25% acetic acid in 0.1 M TAE, followed by a final wash in PBS. Prehybridization was carried out for 1 h in hybridization buffer (50% formamide, 5% SDS, 5× SSPE, 1 mg/ml tRNA), followed by hybridization overnight at 60°C in hybridization buffer containing 1µg/ml DIG-labeled RNA probe. After three washes in 50% formamide and 2 × SSC at 60°C, these sections were subjected to blocking (blocking regent, Roche, Tokyo, Japan) for 1–4 h at room temperature, and then incubated overnight at 4°C with alkaline phosphatase-conjugated anti-DIG antibody (1:2000, Roche, Tokyo, Japan). After washing five times at room temperature, the color reaction was carried out at room temperature or 4°C in BM Purple (Roche, Tokyo, Japan). The reaction was terminated by immersing the sections in 4% paraformaldehyde in 0.1 M phosphate buffer for 15 min. Then sections were treated in 70%, 80%, 90% and 100% ethanol, and xylene, and then embedded.
For Nissl staining, adjacent sections were used. These sections were immersed in 0.1% methylene blue for 15–60 s and then subjected to the ethanol series and embedding.
Expression of Upper Layer-specific Genes in the Neocortex
To identify the genes that are expressed in layer 4 or layer 2/3–4, we constructed a subtraction cDNA library in which cDNAs derived from layer 4 strips of P7 rat somatosensory cortex was enriched by subtracting cDNAs from layer 5 strips (see Materials and Methods). Approximately 1000 clones from the subtraction library were subjected to the differential screening in the first-round screening. As a result, we obtained 130 positive clones, which showed stronger signals to the layer 4 probe than the layer 5 probe. In situ hybridization was then performed with each of the 130 clones, in order to examine laminar specificity on P7 rat brain, in which neocortical laminar configuration is established. Although most of the clones tested showed virtually no or very faint signals, four clones (571, 585, 746 and 846) exhibited specific expression in layer 4 or layer 2/3–4.
Figure 2 shows the laminar expression pattern of these genes in sensory cortices. The expression of 571 was highly restricted to layer 4 of the visual and somatosensory cortices, with almost no expression in any other layers (Fig. 2A,E). The expression of 585 was also strong in layer 4, although it expanded slightly to the adjacent layers (Fig. 2B,F). On the other hand, the expression of 746 was distributed in layers 2/3–4, with no detectable expression in the deep layers (Fig. 2C,G). The expression of 846 was also localized in the upper layers, but faint signals were also detected in the deep layers (Fig. 2D,H).
Identification of the Lamina-specific Genes
Sequence analysis was performed for the four DNA fragments. The sequences of 746 and 846 have high homology to mouse stem cell factor (85 %), and mouse myocyte-specific enhancer factor (95 %), respectively. Taking into consideration the difference between species, 746 and 846 were thought to be rat homologs of these two genes.
On the other hand, DNA fragment 571 showed no homology to any known gene, though there were expression sequence tags (ESTs) that exactly matched these clones. However, an extended product of 571 from the 5′ race method was homologous to an unreported human cDNA, which was obtained from a human brain cDNA library (Nagase et al., 1997, 1999). The transcript contained an open reading frame (2847 bp) and 3′ untranslated region (UTR) of >4000 bp. The deduced amino acid sequence (948 aa) revealed its features to be a transmembrane protein, including a signal peptide sequence, two immunoglobulin and thrombospondin domains (Fig. 3). Its cytoplasmic region consists of ZU5 and death domains, which are common to unc5-like netrin receptors (Ackerman et al., 1997). We hereafter designate this novel member of the unc5 family as unc5h4. The nucleotide sequence of human unc5h4/KIAA1777 was deposited into DDBJ/GenBank/EMBL DNA databases (accession no. AB055056). KIAA1777 is an alias for this new gene in the human brain cDNA database.
DNA fragment 585 was also extended by the 5′ race method and found to highly match a coding region of KIAA0937, which was obtained from the sequence analysis of human cDNAs (Nagase et al., 1999). The deduced amino acid sequence (653 aa) contained two WWE domains, which are predicted to mediate specific protein–protein interactions (Aravind, 2001), and had high homology (58 %) to human deltex. Therefore, KIAA0937 is referred to as deltex-like gene.
Regional Specificity and Expression in Other Brain Regions
Expression patterns of the four obtained genes were further studied in P7 rat brain. In particular, areal specificity was examined in order to understand whether these genes are involved in more generalized features of cortical laminar organization.
Unc5h4/KIAA1777 (571) exhibited a highly restricted expression in layer 4 across all neocortical regions. However, its expression was particularly strong in sensory cortices including somatosensory and visual areas (Fig. 4A,B). Another characteristic of this gene was its prominent expression in the amygdala and hippocampus, especially in CA3 and dentate gyrus. Moderate expression was also found in the hypothalamus. In the thalamus, weak expression was observed in the ventral part of the lateral geniculate nucleus. In other brain regions, it was expressed in mitral cell layers of the olfactory bulb and Purkinje cells of the cerebellum.
The deltex-like gene (585) was also expressed in layer 4 of all neocortical areas. However, in the somatosensory area, the signal was not only distributed in layer 4 but also in the upper part of layer 2/3 (Fig. 4C,D). No apparent signal was found in areas other than the neocortex, except for weak expression in the granule cell layer of the cerebellum.
SCF (746) showed strong expression between layers 2/3 and 4 across all neocortical regions (Fig. 5A,B). As observed for unc5h4/KIAA1777, SCF expression was strong in the somatosensory and visual areas. Its strong expression in the limbic region was another characteristic it shared with unc5h4/KIAA1777. The most striking aspect was the highly specific and strong expression in the thalamus. The expression was located in the dorsal lateral geniculate nucleus and the lateral posterior nucleus, which project to visual areas, and in the ventral basal thalamic nucleus, which projects to somatosensory areas. Strong signals were also found in the habenula, central lateral nucleus, central medial nucleus, parafascicular nucleus and intermediodorsal nucleus. In addition, this gene was expressed in the granule cells of the olfactory bulb and Purkinje cells of the cerebellum.
The expression of MEF-2C (846) was highly restricted to the neocortex (Fig. 5C,D). Its expression in layers 2/3–4 was uniform across the whole neocortex, but no expression was observed in other brain regions.
Developmental Changes of Laminar Expression Patterns
To gain an insight into how each gene is associated with laminar property, cellular differentiation and afferent invasion, their expression patterns were studied in the developing somatosensory cortex.
The expression of unc5h4/KIAA1777 (571) was observed in the subventricular and intermediate zones at E18 (Fig. 6A). At P0, the expression appeared just beneath the marginal zone, and then in slightly lower layers at P3. At P7, the message was restricted to layer 4 (Fig. 6A). Thus, the expression pattern of this gene during development was closely related to the laminar locations of the cells destined to layer 4, but the signal was rather weakened after P14.
The message of deltex-like gene (585) was expressed in both the ventricular zone and the cortical plate (CP) at E18, but was distributed in the upper part of the CP from P0 to P3 (Fig. 6B), which is similar to unc5h4 expression pattern. At P7, the expression was observed in layer 4 although it was slightly diffuse. Moreover, the message was also observed in the upper part of layer 2/3 in the somatosensory cortex. At P14, almost no expression was retained.
The message of SCF (746) was localized in the deepest part of the CP at E18 (Fig. 7A). At P0, the expression in layer 6 was weaker, and even more so at P3, although the signal began to appear in the upper part of the CP. At P7, the expression was strong in layers 2/3 and 4, whereas it had virtually disappeared in layer 6. At P14, almost no expression was observed.
MEF-2C (846) was strongly expressed in the upper part of the CP from E18 to P3 (Fig. 7B). At P7, the expression was primarily observed in layers 2/3 and 4, and slightly in the upper part of layer 5. Unlike the former three clones, the message was retained at P14 to a great extent.
We obtained four genes that were expressed in the upper layers in P7 rat cortex when laminar configuration is established. Unc5h4/KIAA1777 and deltex-like gene/KIAA0937 were expressed specifically in layer 4, whereas SCF and MEF-2C were expressed in layers 2/3–4. As for area specificity, unc5h4 and SCF signals were stronger in sensory cortices, but the messages of MEF-2C and deltex-like gene were distributed rather broadly in the neocortex. All of the genes were strongly expressed in embryonic and early postnatal stages, except that MEF-2C messages were still present in later stages. Thus, the present study using systematic screening of the subtraction cDNA library clearly demonstrates that certain types of molecules including extracellular proteins and transcription factors, which might be involved in cellular differentiation and neural circuit formation, are expressed in the upper layers of the developing cortex.
Expression and Molecular Properties of Upper Layer-specific Genes
Unc5h4, a new member of unc5 family, was expressed most specifically in layer 4 of P7 rat cortex. After the sequence of human unc5h4/KIAA1777 had been registered in GenBank/DDBJ database, its mouse homolog was reported (Engelkamp, 2002). However, the present study is the first to demonstrate its molecular characteristics and expression pattern in the brain. The expression profile during development further showed a migrating behavior of unc5h4 expression. Layer 4 cells are born at E16–17 in the ventricular zone and migrate to the subventricular and intermediate zones at E18. At P0, they move to the most superficial part of the CP, and gradually settle in layer 4 by P6 (Berry and Rogers, 1965; Lund and Mustari, 1977). Unc5h4 expression closely resembles this migration pattern, suggesting that unc5h4 is expressed during development by the cells destined to form layer 4 of the cortex. Weak expression in the frontal lobe is consistent with this view, as granular cells, the major population in layer 4, are scarce in the motor cortex.
It has been shown that unc5 family members are involved in axonal elongation and cell migration as netrin-1 receptors (Leonardo et al., 1997). In addition, these members have been shown to suppress neuronal apoptosis, upon receiving a ligand signal (Hofmann and Tschopp, 1995; Llambi et al., 2001). Although the molecules that interact with unc5h4 have not yet been identified, it is possible that the ligand may influence neuronal survival of layer 4 cells expressing unc5h4. The survival effect might further influence neuronal connectivity. Another possibility is that unc5h4 acts as a ligand molecule. The extracellular domains containing immunoglobulin and thrombospondin domains might directly influence ingrowing thalamic axons. Identification of the molecules that interact with unc5h4 would be necessary to elucidate its role to a great extent.
We cloned and characterized another layer-4-specific molecule, deltex-like gene/KIAA0937, although its expression was not as localized as the unc5h4 message. Deltex has been shown to mediate the Notch signaling pathway in the nucleus as well as in the cytoplasmic region (Matsuno et al., 1995; Kishi et al., 2001; Yamamoto et al., 2001). Therefore, deltex-like protein may be involved in the differentiation of layer 4 cells by regulating Notch activity transcriptionally and/or through direct binding.
As for MEF-2C, Leifer et al. (1993) have shown that MEF-2C mRNA is expressed in the upper layers of the developing cortex. Our results not only confirm their findings but also demonstrate that this lamina-specific expression is present throughout the neocortex, suggesting that MEF-2C may be involved in the upper cortical neuron identity. Although it has been demonstrated that MEF-2C supports cell survival of postmitotic neurons (Mao et al., 1999), the prolonged and rather uniform expression in the neocortex indicates the possibility that MEF-2C may contribute to the maintenance and differentiation of upper cortical cells rather than simply to the differentiation of postmitotic neurons.
On the other hand, SCF expression during cortical development was more complicated, but seems to be related to thalamic axon invasion. Since the tips of thalamic axons start to invade the CP at E18, extend into layer 5 at P0, and reach immature layer 4 at P3 (Lund and Mustari, 1977; Kageyama and Robertson, 1993; Catalano et al., 1996; Molnár et al., 1998), a correlation might exist between the expression of SCF and ingrowth of thalamic axons in the neocortex. In accordance with thalamic axon extension, SCF expression gradually became weaker in the deep layers and increased in the upper layers. Indeed, immunohistochemistry with an antibody against c-kit, the receptor of SCF, showed that c-kit was expressed in thalamic nuclei and axons (not shown), indicating the possibility that thalamic axons respond to SCF distributed in the developing cortex. The weaker expression of SCF in the frontal cortex might also be related to the lack of sensory thalamic axons in this area.
Implications of the Presence of Upper Layer-specific Genes
To date, it has been shown that putative regulatory genes, RZR-β (Becker-Andre et al., 1994; Park et al., 1997) and COUP-TF1 (Liu et al., 2000) are expressed in layer 4 of the neocortex, although the functions of these genes are not clear. Thalamocortical projections are disrupted in COUP-TF1 mutants, but it is likely that the phenotype is attributable to the lack of subplate neurons rather than its direct influences on thalamocortical axon targeting in layer 4 (Zhou et al., 1999). The present results further demonstrate the presence of another putative regulatory gene, deltex-like gene, which is expressed primarily in layer 4 and might act as a regulator of cellular differentiation (see above).
It is also important to reveal extracellular molecules such as cell surface or extracellular matrix molecules, in order to understand the molecular basis of cellular interactions and circuit formation (Yamamoto, 2002). In this sense, unc5h4 is the gene that encodes a cell surface protein and is expressed rather specifically in cortical cells destined for layer 4 during development. Although how unc5h4 contributes to cellular differentiation and connectivity in the cortex is unknown, its molecular characteristics raise the possibility that it is involved in the interactions between layer 4 cells, or between layer 4 cells and thalamocortical fibers. Recent studies have shown that ephrin-A5, an Eph ligand, affects thalamocortical axon behavior by its expression in layer 4 (Mann et al., 2002), but this is unlikely to be the case with the entire sensory thalamocortical projections (Donoghue and Rakic, 1999; Mackarehtschian et al., 1999; Vanderhaeghen et al., 2000; Yabuta et al., 2000). In contrast, localization of unc5h4/KIAA1777 in layer 4 is found throughout the sensory cortices.
Cadherin-6 and rCNL3 are known cell surface molecules that are expressed in the upper layers of the developing cortex (Suzuki et al., 1997; Chenn et al., 2001). Phosphacan, a proteoglycan (Maeda and Noda, 1996), and SemaK1, a GPI-anchored semaphorin (Xu et al., 1998), are also expressed in the upper layers of early postnatal stages, though their expression patterns during development are not well characterized. It would be worthwhile to examine whether these molecules affect thalamic axonal growth, as previous investigations have indicated that thalamocortical axon termination is governed by molecules distributed in layers 2/3–4. (Yamamoto et al., 2000a,b; Noctor et al., 2001; Yamamoto, 2002). In addition, the present findings of SCF expression pattern raises the possibility that a developmental change of laminar expression may influence afferent fiber ingrowth (see above). A similar laminar transition has been reported in the expression of Sema3A, a potential molecule, which might be responsible for the formation of thalamocortical projections (Skaliora et al., 1998).
Several transcriptional factors have been shown to be expressed in a lamina-specific fashion: Otx1 (Frantz et al., 1994) and Id2 (Bulfone et al., 1995) in layer 5 and Tbr-1 (Bulfone et al., 1995; Hevner et al., 2001) in layer 6. As for layer 4, COUP-TF1 and RZR-β are expressed in layer 4, as described above. The present results show that deltex-like gene and MEF-2C are expressed in layer 4 and layer 2/3–4 as regulatory factors, respectively. Determining the relation between these regulatory genes and their downstream molecules would also be useful in elucidating the molecular basis of cortical cell differentiation and neural circuit formation.
We thank Dr R. Carney for critical reading of this manuscript. We also thank Drs Y. Zhu and Y. Hatanaka for helpful comments on this manuscript, and Dr M. Mieda for helpful advice on construction of the subtraction cDNA library. This work was supported by Grants-in-Aid for Scientific Research Projects 12050227 and 13041038 from Japanese Ministry of Education, Culture, Science and Sports, by Core Research for Evolutional Science and Technology (CREST) from the Japan Science and Technology Corporation (JST), and by Human Frontier Science Project (RGP0107/2001).
Y. Zhong and M. Takemoto contributed equally to this study.