Abstract
The mammalian neocortex undergoes dramatic transformation during development, from a seemingly homogenous sheet of neuroepithelial cells into a complex structure that is tangentially divided into discrete areas. This process is thought to be controlled by a combination of intrinsic patterning mechanisms within the cortex and afferent axonal projections from the thalamus. However, roles of thalamic afferents in the formation of areas are still poorly understood. In this study, we show that genetically increasing or decreasing the size of the lateral geniculate nucleus of the mouse thalamus resulted in a corresponding change in the size of the primary visual area. Furthermore, elimination of most thalamocortical projections from the outset of their development resulted in altered areal gene expression patterns, particularly in the primary visual and somatosensory areas, where they lost sharp boundaries with adjacent areas. Together, these results demonstrate the critical roles of thalamic afferents in the establishment of neocortical areas.
Introduction
The neocortex is one of the most complex regions in the mammalian brain, and the formation of functionally and anatomically discrete neocortical areas during development is crucial for our perception, motor control, learning, and memory. Area formation is a multistep process. In early embryonic stages, neuroepithelial and radial glial cells show graded expression patterns of transcription factors regulated by secreted signaling molecules including fibroblast growth factors and Wnt proteins. These transcription factors specify the positional identity of neural progenitor cells and their postmitotic areal fate (O'Leary et al., 2007; Hoch et al., 2009; Rakic et al., 2009; Sansom and Livesey, 2009; Borello and Pierani, 2010).
Later in embryogenesis, topographically organized, afferent axons from the thalamus [thalamocortical axons (TCAs)] reach the cortex and start to invade the cortical plate in their target cortical regions. By the end of the first postnatal week, discrete tangential and laminar gene expression patterns are established in postmitotic neurons, delineating future functional areas in the neocortex. Ablation and transplantation studies have long suggested that TCAs are required for the establishment of area-specific cytoarchitecture (Wise and Jones, 1978; Dehay et al., 1991; Rakic et al., 1991; Schlaggar and O'Leary, 1991; Gitton et al., 1999). However, genetic mutations in mice that cause missing (Miyashita-Lin et al., 1999; Nakagawa et al., 1999) or shifted (Garel et al., 2002) TCA projections did not result in major changes in gene expression patterns in the cortical plate at late embryonic or neonatal stages. These studies suggested that intrinsic patterning mechanisms are predominantly responsible for regionalization of the cortex during embryogenesis. However, these mutant mice died before area boundaries in the neocortex were well defined. Thus, it is currently unknown whether TCAs exert any roles in the establishment of discrete neocortical areas.
To resolve this issue, we genetically altered thalamic neurons from the outset of their axonal projections to the neocortex. Specifically, we generated postnatally viable mutant mice in which, (1) the number of thalamocortical projection neurons was increased or decreased in the principal visual nucleus [dorsal lateral geniculate (dLG)], or (2) almost no TCAs project to the neocortex. These genetic manipulations were specific to the thalamus so that changes observed in the neocortex could be attributed to changes in TCA projections. When we increased or decreased the size and cell number of the dLG nucleus by manipulating Sonic hedgehog (Shh) signaling in thalamic progenitor cells, TCA projections to the primary visual area were expanded or reduced, respectively. These mice still retained clear gene expression boundaries between areas, but the size of the primary visual area was altered in accordance with the change in geniculocortical projections. Furthermore, when TCA projections to the neocortex were nearly eliminated in mice lacking the homeobox gene Gbx2 in the thalamus, gene expression patterns in the neocortex, most evident in the prospective primary sensory areas, were disrupted, leading to less discrete boundaries between prospective sensory areas and their surrounding areas. These findings demonstrate that TCAs play a crucial role in early establishment of area identity in the neocortex.
Materials and Methods
Mice
Care and experimentation on mice were done in accordance with the Institutional Animal Care and Use Committee of the University of Minnesota. Noon of the day on which the vaginal plug was found was counted as embryonic day 0.5 (E0.5), and the day of birth was designated as postnatal day 0 (P0). To generate Olig3Cre/+;R26stopSmoM2/+ mice, we bred Olig3Cre/+ mice (Vue et al., 2009) with Rosa26stopSmoM2/stopSmoM2 mice (Jeong et al., 2004) (obtained from The Jackson Laboratory), and used Cre− progeny as controls. For conditional deletion of Shh, we bred Olig3Cre/+;ShhN/+ mice with ShhC/C mice (Lewis et al., 2001) (obtained from The Jackson Laboratory). All of the non-Olig3Cre/+;ShhN/C littermates showed normal phenotypes. For conditional deletion of Gbx2, we bred Olig3Cre/+;Gbx2C/+ mice with Gbx2C/C mice (Li et al., 2002) (obtained from The Jackson Laboratory). Only Olig3Cre/+;Gbx2C/C showed detectable phenotypes. We found similar phenotypes in TCA pathfinding in Gbx2 conditional mutants in which a null mutant allele was used in the breeding scheme to generate Olig3Cre/+;Gbx2N/C mice. The Gbx2N allele was produced by germline deletion of Gbx2 exon 2 by breeding Gbx2c/c mice and heterozygous CMV-Cre mice. These Olig3Cre/+;Gbx2N/C mice were used for Fig. 7M–T. All mice were kept in C57B/L6 background and embryos or pups of either sex were used.
In situ hybridization and immunostaining
In situ hybridization and immunostaining were performed based on Vue et al. (2007). In this study, goat anti-Netrin-G1 (R&D Systems, 1:100), rabbit anti-serotonin (Immunostar, 1:20,000 or Millipore, 1:500), rabbit anti-RORβ (Diagenode, 1:100), and goat anti-Lmo4 (Santa Cruz Biotechnology, 1:100) were used. Prenatal brains were cut with a cryostat at 20–35 μm thick. Postnatal brains were cut with a sliding microtome at 50 μm thick and the sections were mounted on slides before in situ hybridization or immunostaining was performed. Each set of data shown in Figs. 1A–H, 3, 4, 5A–F, 7M–T, 8, and 9 A–F were obtained from adjacent sections.
Axon tracing with NeuroVue dyes
NeuroVue axon tracers (FS-1002 NeuroVue Red, FS-1001 NeuroVue Maroon) were inserted to primary visual area (V1) and primary somatosensory area (S1) of P8 cortex and the brains were incubated in 4% paraformaldehyde for ∼4 weeks, before sectioning with a vibrating microtome (150 μm). Sections were counterstained with bisbenzimide H 33342 (Sigma-Aldrich).
Quantification
dLG cell count and size measurement.
Each 20 μm cryosection was collected and sequentially mounted onto six slides for in situ hybridization analysis of P0 thalamus. Two slides were analyzed for cell count and size measurement and other slides were used for in situ hybridization for RORα. Because some Olig3Cre/+;R26stopSmoM2/+ mice had clustered RORα-positive areas, the region to be counted or measured as dLG was defined by the presence of RORα mRNA on adjacent sections. Although RORα is also expressed in the nearby ventral posterior (VP) nucleus, distinction between VP and dLG was easy with a histological boundary. All the DAPI-positive cells in the dLG nucleus (7–10 sections/thalamus) were counted using ImageJ software. The dLG volume was measured by summing the area of RORα-positive regions. Four thalami were analyzed for each genotype. Statistical analysis was done using two-tailed Student's t test.
Area measurement on tangential sections.
The area and length of V1 with respect to the area and length of the entire neocortex of P7 tangential sections were measured in Olig3Cre/+;ShhN/C mice, Olig3Cre/+;R26stopSmoM2/+ mice and their respective littermate controls using ImageJ software. Eight to 13 hemispheres were analyzed for each genotype. Statistical analysis was done using two-tailed Student's t test.
Area measurement on sagittal sections.
The gap between S1 and V1 was measured on P7 sagittal sections of Olig3Cre/+;ShhN/C mice, Olig3Cre/+;R26stopSmoM2/+ mice and their respective littermate controls using ImageJ software. The low-Cdh8 region in layer 4 was used as a landmark of S1 and V1. Four hemispheres were analyzed for each genotype. For each hemisphere, six sagittal sections at distinct mediolateral levels were compared between genotypes. Statistical analysis was done using two-tailed Student's t test.
Area measurement on coronal sections.
The medial-lateral position of V1 was measured on P7 coronal sections of Olig3Cre/+;ShhN/C mice, Olig3Cre/+;R26stopSmoM2/+ mice and their respective littermate controls using ImageJ software. Lmo4 expression in layer 4 is absent in V1, forming a sharp medial border of V1. In higher-order visual area as well as retrosplenial cortex, Lmo4 is strongly expressed in upper layers (see Fig. 4S). We measured the length of Lmo4 expression domain along the line that separates Lmo4-high upper layers from the lower layer (line segments in yellow in 4S shows an example). Six hemispheres were analyzed and statistical analysis was done using two-tailed Student's t test.
Results
Conditional manipulations of Shh signaling in the thalamus cause bidirectional changes in the size and cell number of the dorsal lateral geniculate nucleus
The basic-helix-loop-helix (bHLH) transcription factor Olig3 is expressed specifically in thalamic progenitor cells and the zona limitans intrathalamica (ZLI) (Vue et al., 2007). The ZLI is located at the rostral boundary of the thalamus, which in addition to the basal plate, produces the morphogen Shh to regulate diencephalic development (Scholpp and Lumsden, 2010; Vieira et al., 2010; Nakagawa and Shimogori, 2012). Previous studies demonstrated that different levels of Shh signaling is responsible for setting up the positional identity of neural progenitor cells within the thalamus, which in turn is essential for generating the proper number and size of thalamic nuclei (Szabo et al., 2009; Vue et al., 2009; Jeong et al., 2011). Ectopic expression of SmoM2, a constitutively active form of the Shh effector molecule Smoothened (Smo), expanded the molecular identity of rostroventral thalamic progenitors at the expense of caudodorsal identity (Vue et al., 2009). This expansion results in enlargement of rostro-ventrally located nuclei including the dLG nucleus, which project to the V1 (Vue et al., 2009). Quantitative analysis of these mice showed that the number of DAPI-positive cells in the expanded dLG was increased by twofold (Fig. 1A,B,I), whereas its volume showed a threefold increase (Fig. 1J) compared with control littermates. Interestingly, lateral posterior (LP) nucleus, which projects to higher-order visual areas and expresses Calb2 (calretinin), was strikingly shrunken (Fig. 1E,F).
We next generated mice in which Shh signaling was reduced in the thalamus. Shh is expressed both in the ZLI and the diencephalic basal plate, but only the ZLI express both Shh and Olig3 (Vue et al., 2007). Thus, with the Olig3Cre allele, Shh was conditionally knocked out in the ZLI (Bluske et al., 2012). At E12.5, Olig3Cre/+;ShhN/C embryos specifically lacked the expression of Shh in the ZLI (Bluske et al., 2012). In these embryos, Olig2 expression in rostral thalamic progenitor cells was reduced (data not shown), reflecting a reduction of rostroventral identity of thalamic progenitor cells (Vue et al., 2007). Dbx1, a marker expressed highly in the caudo-dorsal part of the thalamus (Vue et al., 2007), was rostrally expanded in the same embryos (data not shown). These changes were opposite to those found in mice expressing ectopic SmoM2 (Olig3Cre/+;R26stopSmoM2/+ mice) (Vue et al., 2009), and resulted in a reduction of the expression domain of RORα in the dLG nucleus and reciprocal expansion of the LP nucleus (Fig. 1C,D,G,H). Quantitative analysis of the RORα expression domain demonstrated that the number of cells within dLG was decreased by ∼50%, whereas the volume of dLG was decreased by ∼65% in Olig3Cre/+;ShhN/C mice compared with littermate controls (Fig. 1C,D,K,L).
By using Netrin-G1 as a TCA marker (Nakashiba et al., 2002), we further found that altering Shh signaling in either direction within the thalamus does not affect gross patterns of TCA projection at E12.5 and E15.5 (Fig. 2A–H). Combined with a strong Cre reporter line Ai6 that expresses ZSGreen (Madisen et al., 2010), Olig3Cre allele showed no sign of recombination within the neocortex until at least P7 (Fig. 2I–K). We thus expected that the intrinsic patterning of the cortex will be intact in both Olig3Cre/+;R26stopSmoM2/+ and Olig3Cre/+;ShhN/C mice. As predicted, regional expression patterns of RORβ, Cdh8, Id2, EphA7, and Lmo4 in E17.5 cortical plate were similar to control mice (Fig. 2P–Y). These results are consistent with earlier reports showing that TCAs are not critically required for neocortical regionalization during embryogenesis (Miyashita-Lin et al., 1999; Nakagawa et al., 1999). Both Olig3Cre/+;ShhN/C and Olig3Cre/+;R26stopSmoM2/+ mice survived postnatally, and the overall morphology of the P8 forebrain was not different between the genotypes (Fig. 2L–O).
In summary, we established two postnatally viable mouse models in which the size and number of neurons in the dLG nucleus are bidirectionally altered due to changes in Shh signaling in the thalamic ventricular zone. Changes in the dLG size were accompanied by reciprocal changes of the LP nucleus, which normally project to higher-order visual areas. In these two mutants, timing of TCA projections and cortical patterning were not altered, thereby allowing us to assess the specific roles of geniculocortical afferents in the formation of primary visual cortex.
Expansion or shrinkage of geniculocortical projections alters the size and boundaries of the primary visual area in postnatal cortex
At P7 in control mice, axon terminals from principal sensory nuclei of the thalamus show strong immunoreactivity of serotonin mainly in layer 4 of primary sensory areas (Fig. 3A,C) (Lebrand et al., 1996). In Olig3Cre/+;R26stopSmoM2/+ mice, the serotonin-positive region in the caudal cortex, which corresponds to the geniculocortical axon terminals, was expanded, whereas it was shrunken in Olig3Cre/+;ShhN/C brains (Fig. 3E,G,I,K). This indicated that altered cell numbers in the dLG nucleus results in expansion or shrinkage of TCA projections to V1.
We next tested whether the altered geniculocortical projections are accompanied by matching changes in gene expression patterns in V1. In control animals at P7, the distribution of RORβ, which is known to be strong in layers 4 and 5 of primary sensory areas at this stage as well as in the adult brain (Schaeren-Wiemers et al., 1997), closely matched with that of serotonin staining on adjacent sections (Fig. 3A–D). In Olig3Cre/+;R26stopSmoM2/+ and Olig3Cre/+;ShhN/C mice, expression of RORβ was altered in the same way as the serotonin signal (Fig. 3F,H,J,L), demonstrating that V1 is expanded or shrunken in response to the altered size of geniculocortical projections.
In addition to RORβ, other markers for V1 borders including Lmo4, Cdh8, and Id2 were analyzed on adjacent sections (Fig. 4). In this analysis, we used serially adjacent sections for the four molecular markers (Fig. 4A–D, E–H, I–L, and M–P, from adjacent sections). Lmo4 was expressed in layers 2, 3, and 4 of the medial cortex (Fig. 4B,J). The lateral border of this Lmo4 expression domain (Fig. 4B,J, medial arrow) matched with the medial border of RORβ (Fig. 4A,I, medial arrow) on adjacent sections, delineating the medial border of V1. We further confirmed the matching gene expression borders by double immunostaining of Lmo4 and RORβ on the same section (Fig. 4Q,R). Cdh8 expression was detected in layers 2–5 of visual areas, but layer 4 in V1 showed specifically low expression (Fig. 4C,K, between black arrows). Expression of Id2 in layer 6 also marked V1. In Olig3Cre/+;ShhN/C mice, the medial border of RORβ expression in both layers 4 and 5 was shifted laterally compared with controls at all rostrocaudal levels examined (Fig. 4A,E). The same shift was found for Lmo4 (Fig. 4F), Cdh8 (Fig. 4G), and Id2 (Fig. 4H). In contrast, in Olig3Cre/+;R26stopSmoM2/+ mice, a medial shift of the medial border of V1 was found for all of the above markers (Fig. 4I–P). The distance from the medial border of V1 to the midline was quantitatively evaluated by measuring the length of the medial expression domain of Lmo4, which starts at the medial border of V1 and ends at the medial border of retrosplenial cortex, abutting the subiculum (Fig. 4S). We found that this distance was decreased in Olig3Cre/+;R26stopSmoM2/+ mice and increased in Olig3Cre/+;ShhN/C mice in comparison with their respective control littermates (Fig. 4T,U). In contrast, the lateral border of V1, which is indicated by the lateral border of low Cdh8 expression in layer 4, did not seem to be significantly shifted (Fig. 4C,G,K,O). These results show that along the mediolateral axis, V1 is contracted in Olig3Cre/+;ShhN/C mice and expanded in Olig3Cre/+;R26stopSmoM2/+ mice. Similar changes in gene expression for these markers were already evident at P3 at least in Olig3Cre/+;R26stopSmoM2/+ mice (data not shown). These data suggest that the size and shape of V1 are significantly altered in Olig3Cre/+;R26stopSmoM2/+ and Olig3Cre/+;ShhN/C mice, and the influence of geniculocortical axons on V1 in these mice was not limited just to layer 4, but was also detected in layers 2, 3, 5, and 6.
Finally, to determine how the overall size and shape of V1 are altered in correspondence to the changes in TCA terminations, the dimensions of V1 were measured on tangential sections by comparing serotonin staining and RORβ expression on adjacent sections between the different genotypes. As on coronal sections, we found that serotonin staining and RORβ expression in V1 precisely match in Olig3Cre/+;R26stopSmoM2/+, Olig3Cre/+;ShhN/C, and their respective control mice, and the relative size of V1 was expanded in Olig3Cre/+;R26stopSmoM2/+ and shrunken in Olig3Cre/+;ShhN/C mice, both in area and rostrocaudal length ratios against the total cortex (Fig. 5A–H). The gap between V1 and S1 contains higher-order sensory areas. This gap was narrowed in Olig3Cre/+;R26stopSmoM2/+ and widened in Olig3Cre/+;ShhN/C mice (Fig. 5A–F, between bars). Quantitative analysis of Cdh8 expression in layer 4 on sagittal sections at P8 confirmed the change in the gap between V1 and S1 (Fig. 5I–N). The altered gap width between V1 and S1 indicates that the shift of V1 boundaries is accounted for by a switch of areal fate from primary to higher-order sensory areas or vice versa. The reciprocal changes in the size of the dLG and LP nuclei in both Olig3Cre/+;R26stopSmoM2/+ and Olig3Cre/+;ShhN/C mice (Fig. 1) imply that thalamic afferents from principal sensory nuclei and other nuclei play distinct roles in conferring areal identity to their respective target region in the neocortex.
Thalamus-specific Gbx2 mutant mice survive postnatally and show severe deficiency of thalamocortical axon projections to the neocortex
In both Olig3Cre/+;R26stopSmoM2/+ and Olig3Cre/+;ShhN/C mice, the relative sizes of dLG and LP nuclei were changed as a result of altered Shh signaling in thalamic progenitor cells, whereas the overall TCA trajectory to the cortex was not significantly affected. In this light, determining the default state of cortical development without any influence by TCAs would provide insights into the nature of the specific roles of primary sensory and other types of TCA afferents. We therefore generated thalamus-specific Gbx2 mutant mice, in which TCA projection to the neocortex was found to be nearly absent from the outset of its development.
Gbx2 encodes a homeodomain-containing transcription factor expressed in nearly all cortex-projecting postmitotic neurons in the thalamus at E12.5 (Bulfone et al., 1993; Chen et al., 2009), although its expression later becomes restricted to several nuclei (Nakagawa and O'Leary, 2001; Jones and Rubenstein, 2004; Li et al., 2012). Gbx2 is also expressed at the midbrain-hindbrain boundary, in the spinal cord and in cholinergic neurons of the ventral telencephalon. Germ-line deletion of Gbx2 in mice resulted in the lack of TCA projections to the cortex (Miyashita-Lin et al., 1999; Hevner et al., 2002). Although these mutant mice die at birth, differential gene expression in the neonatal cortical plate showed normal patterns, which indicated the sufficiency of intrinsic cortical patterning in early phases of area formation (Miyashita-Lin et al., 1999).
To avoid the neonatal lethality so that we could determine the roles of TCAs in cortical development during both embryonic and postnatal stages, we generated conditional Gbx2 mutant mice using the Olig3Cre allele (Olig3Cre/+;Gbx2C/C mice or Gbx2 cko mice). We first found that Gbx2 cko mice survive postnatally, up to 5–6 weeks after birth. At E12.5 and E14.5, the mutant mice showed near-complete deletion of the floxed exon2 of the Gbx2 gene in the thalamus, whereas its expression in the ventral telencephalon was preserved (Fig. 6A,B,D,E and data not shown). Expression of RORα, which mainly labels principal sensory nuclei of the thalamus (Nakagawa and O'Leary, 2003), is severely reduced in Gbx2 cko mice (Fig. 6C,F).
Due to the early expression of Gbx2 at midbrain-hindbrain boundary and the contribution of brainstem Raphe neurons in providing serotonergic innervation to the cortex, we tested the effects of using the Olig3Cre allele on these neuronal populations. Although the Olig3Cre allele mediates recombination in dorsal hindbrain region that is fated to produce precerebellar nuclei including the inferior olive (Liu et al., 2008; Storm et al., 2009), we did not detect an overlap between recombined cells and serotonergic neurons at E12.5 (Fig. 6G). Furthermore, serotonergic neurons in dorsal and medial Raphe nuclei seemed intact in Gbx2 cko mice at E14.5 (Fig. 6H,I), and serotonergic projections from the Raphe neurons appeared normal through the septum into the cortex at E16.5 (Fig. 6J,K). Therefore, in Gbx2 cko mice, the cortex is specifically deprived of afferent projections from the thalamus but not from the brainstem Raphe nuclei.
Detailed analysis of TCAs of Gbx2 cko mice showed profound deficiency from the initial phase of their projections toward the cortex. Already at E12.5, TCAs failed to reach the diencephalon-telencephalon border (Fig. 7A–D). By E14.5, when TCAs have normally reached the neocortex, only a small axon bundle was found in the ventral telencephalon of Gbx2 cko mice (Fig. 7E–H). At E16.5, a faint labeling of Netrin-G1-expressing TCAs was barely detected in the neocortex (Fig. 7I,J). These results demonstrate that Gbx2 expression in the thalamus is required for TCA projection to the cortex. Placement of axon tracers in putative S1 and V1 at P8 showed a severely decreased connections between the neocortex and the thalamus (Fig. 7K,L). Thus, the defects in TCA projections persist into postnatal stages. At P21, the thalamus was further disorganized. In Nissl staining, we could not detect dLG or VP, and there was a severe deficiency of internal capsule, indicating that there is no recovery of TCA defects in Gbx2 cko mice (Fig. 7U,V).
In summary, Gbx2 cko mice provide us with a robust model system to test the requirement of thalamic afferents in the formation of neocortical development both during embryonic and postnatal stages.
Distinct areal and laminar gene expression patterns are disrupted in postnatal neocortex of conditional Gbx2 mutant mice
At E17.5, overall regional gene expression patterns in the cortical plate appeared normal in Gbx2 cko mice (Fig. 7M–T), which is similar to a previous study of germline Gbx2 mutants (Miyashita-Lin et al., 1999). However, by the end of the first postnatal week we found specific defects in areal and laminar gene expression patterns (Fig. 8).
As in V1 (Fig. 4), RORβ expression in the control cortex was strong in layer 4 in S1 and A1, and faint in the motor area as well as areas adjacent to primary sensory areas (Fig. 8A–E). In Gbx2 cko cortex, such distinction was no longer evident; RORβ expression was decreased in primary sensory areas and increased in the motor cortex (Fig. 8F–J). In particular, the robust thickening of RORβ-positive layer 4 in the barrel field of S1 (S1BF) was dramatically reduced (Fig. 8G), whereas the gap between strong RORβ expression in the hindlimb region (S1HL) and S1BF (Fig. 8B, near arrow) was not detected in the mutant cortex (Fig. 8G). Tangential sections through layer 4 of the Gbx2 cko cortex showed severely disorganized pattern of RORβ expression, reflecting the blurred borders of its expression domains and the thinning of layer 4 (Fig. 9G,H).
Lmo4 showed largely complementary expression patterns to that of RORβ. In the control motor cortex, Lmo4 was strong not only in layer 4 but also in layers 2, 3, 5, and 6 (Fig. 8K). In the adjacent S1 (forelimb, lip, and jaw areas),which is located lateral to the motor area, Lmo4 was faint and restricted to layers 4 and 6 (Fig. 8K). At more caudal levels containing S1BF, Lmo4 expression was almost undetectable except in layer 6, forming a sharp medial border of S1BF (Fig. 8L–N). At the caudal-most level, expression of Lmo4 in layer 4 was excluded from the primary sensory areas SIBF and V1, clearly demarcating their borders (Fig. 8N,O). In Gbx2 cko mice, Lmo4 was ectopically expressed in layer 4 of S1BF and V1, which made the primary sensory areas resemble the adjacent, higher-order sensory areas (Fig. 8Q–T). At more rostral levels, we found the motor area showed reduced expression of Lmo4, making the distinction from the adjacent S1 less clear (Fig. 8P).
Cdh8 expression in layers 5 and 6 was broad across the entire neocortex, whereas its expression in layers 2,3, and 4 was generally higher medially than laterally (Fig. 8U–Y). We found that layer 4 of S1 and V1 shows specifically low expression of Cdh8, clearly delineating their boundaries (Fig. 8U–Y), which was also confirmed on sagittal sections (Figs. 5I,J, 9E). In Gbx2 cko cortex, this Cdh8-low “hollows” in V1 was no longer detectable and the pattern of Cdh8 expression was indistinguishable between the putative V1 and its surrounding areas (Figs. 8c,d, 9F).
In summary, Gbx2 cko mice showed altered patterns of laminar and areal gene expression patterns in the neocortex at the end of the first postnatal week. As a result, discrete areal boundaries, particularly the boundaries of primary sensory areas, were much less clearly defined. These results demonstrate the requirement of TCA projections in the formation of distinct gene expression patterns across neocortical areas.
Discussion
In this study, altering the size and cell number in the thalamic dLG nucleus resulted in corresponding changes in the size of its target area, V1, at postnatal stages. In addition, genetic removal of a majority of TCAs in the neocortex prevented the formation of discrete gene expression boundaries between areas. These results collectively demonstrate both the sufficiency and requirement of TCAs in defining neocortical area boundaries.
Previous studies addressing the roles of TCAs in the formation of neocortical areas relied largely on postnatal manipulations of afferent innervation of the cortex. For example, surgical ablation of the thalamus in neonatal rats and mice resulted in the lack of characteristic barrel-like histology and lack of expression of the transgene H-2Z1 in S1 (Wise and Jones, 1978; Windrem and Finlay, 1991; Gitton et al., 1999). Transplantation of late embryonic tissue of the rat occipital cortex to the parietal cortex of the neonatal host brain resulted in the barrel-like morphological features in the graft (Schlaggar and O'Leary, 1991), supporting the idea that thalamic afferents have influence on the areal fate of the neocortical cells. In these studies, manipulations were done on the tissue after TCAs had started the invasion of the cortical plate. Therefore, it was not clear whether TCAs are required for the initial establishment of areal features or their maintenance. Bilateral eye removal in monkey fetuses before the arrival of TCAs at the occipital cortex resulted in partial replacement of the primary visual area with an area in which layer 4 was much thinner; designated as area 18 (Dehay et al., 1989, 1991,1996) and “area X” (Rakic, 1988, 1991). Although these studies demonstrate the requirement of the afferent input in characteristic cytoarchitecture of the primary visual cortex, the indirect nature of afferent manipulations and the difficulty of in utero surgeries on primate fetuses prevented further elucidation of when and how the geniculocortical axons affect V1 formation.
Mutant mouse lines that lack TCA projections to the neocortex did not show gross changes of regional gene expression patterns in the neocortex (Miyashita-Lin et al., 1999; Nakagawa et al., 1999). Furthermore, in mice where subcortical defects caused a shift in the topographic projections of TCAs in the neocortex, regional gene expression in the cortical plate was not affected at late embryonic stages (Garel et al., 2002). These studies indicated the sufficiency of intrinsic patterning mechanisms in regional differentiation of the neocortex, which is independent of thalamic afferents. However, all of these mutant mice died at birth, preventing the analysis of detailed area-specific and layer-specific patterns of gene expression that become evident only postnatally. A more recent study reported that gene expression patterns in postnatal cortex look grossly normal in mice lacking both TCAs and corticofugal axons (Zhou et al., 2010), leading the authors to propose that arealization in early postnatal neocortex is independent of TCAs. Furthermore, grafting of embryonic stem cell-derived cortical neurons harboring molecular signatures of the visual area established subcortical and intracortical connectivity characteristic of visual cortical neurons even when these cells were grafted into the frontal cortex, again indicating the sufficiency of intrinsic mechanisms in the specification of cortical area identity (Gaspard et al., 2008). Thus, it has remained elusive whether or how TCAs regulate the early establishment of area identity.
In this study, we genetically altered early thalamic patterning and TCA pathfinding without affecting the intrinsic patterning mechanisms within the cortex. When dLG and its cortical projections were expanded or shrunken, discrete area boundaries still formed but their positions shifted according to the altered size of the geniculocortical projections (Fig. 10A,C,D). The LP nucleus, which normally projects to higher-order visual areas, showed reciprocal size changes to those of dLG. When the entire TCA projections were severely disrupted regardless of their nuclear origin, the postnatal cortex no longer exhibited distinct gene expression patterns that delineate boundaries of primary sensory areas (Fig. 10B). The disruption was most pronounced in the nascent primary sensory areas, where layer 4 was significantly affected both in gene expression patterns and in its thickness. These results indicate that TCAs from principal sensory thalamic nuclei provide specific information to instruct the target cortical regions to take on the fate of primary sensory areas, which would otherwise become higher order sensory areas as a default fate. The gross patterns of gene expression that we observed in Gbx2 cko neocortex were still partially reminiscent of those in the normal mice. This provides the first detailed picture of the default state of area specification, which is set up by early intrinsic pattering and non-thalamic afferents that are intact in Gbx2 cko mice.
It is currently unknown which populations of cortical cells are directly influenced by TCAs during the formation of areas. A likely scenario is that immature cortical plate neurons are exposed to molecules that are provided by the incoming TCAs and undergo dynamic changes in gene expression, which are evident by comparing the patterns between E17.5 (Figs. 2, 7) and P8 (Figs. 4, 8).
The gradients of gene expression in embryonic cortical plate are smooth, and their subtle changes could be difficult to detect. Therefore, although we did not find gross changes in gene expression in the cortical plate at E17.5 when TCAs were altered (Figs. 2, 7), this result does not completely exclude the possibility that TCAs have a role in embryonic cortex.
Because of the reduced thickness of layer 4 in S1BF, it is possible that TCAs have influence on the neuronal number in layer 4 of the primary sensory areas. We did not observe significant increase in cell death in layer 4 at E17.5 or P8 in Gbx2 cko mice (data not shown), which suggests that generation of layer 4 neurons might be under the regulation of TCAs. Therefore, it would be worth investigating the roles of TCAs in embryonic cortical development, and our model mice would be ideal for this purpose because of the early and specific manipulations of TCAs.
Currently, little is known about the cellular and molecular mechanisms responsible for the roles of TCAs in early neocortical development. Results of our study indicated that specific factors produced by principle sensory thalamic nuclei induce the characteristic gene expression and cytoarchitecture of primary sensory areas, and without the TCA afferent, the neocortical areas resemble adjacent, higher-order sensory areas as a default state. Our transgenic mouse models will be useful in future studies to identify potential candidate molecules and reveal the cellular and molecular mechanisms of area formation in the neocortex.
Footnotes
This study was supported by NINDS Grant R01 NS049357, the Whitehall Foundation, and Wallin Neuroscience Discovery Funds and Autism Initiative Funds from the University of Minnesota, as well as Doctoral Dissertation Fellowship (T.Y.V.) and Undergraduate Research Opportunities Programs (Y.E.T., L.W., and Z.W.) from the University of Minnesota. We thank Geoff Ghose, Glenn Giesler, Paul Letourneau, Steven McLoon, and members of Nakagawa Laboratory for comments and discussion; Dino Leone for advice on immunostaining; Lauren Bolopue, Andrew Hanson, David Scott, and Lyhun Tan for excellent technical assistance; and Yasuhiko Kawakami for ShhN/+ and CMV−Cre mice.
- Correspondence should be addressed to Dr. Yasushi Nakagawa, Department of Neuroscience, University of Minnesota, 6-145 Jackson Hall, 321 Church Street SE, Minneapolis, MN 55455. nakagawa{at}umn.edu