Regulation of Area Identity in the Mammalian Neocortex by Emx2 and Pax6

See allHide authors and affiliations

Science  14 Apr 2000:
Vol. 288, Issue 5464, pp. 344-349
DOI: 10.1126/science.288.5464.344


The contribution of extrinsic and genetic mechanisms in determining areas of the mammalian neocortex has been a contested issue. This study analyzes the roles of the regulatory genesEmx2 and Pax6, which are expressed in opposing gradients in the neocortical ventricular zone, in specifying areas. Changes in the patterning of molecular markers and area-specific connections between the cortex and thalamus suggest that arealization of the neocortex is disproportionately altered in Emx2 andPax6 mutant mice in opposing manners predicted from their countergradients of expression: rostral areas expand and caudal areas contract in Emx2 mutants, whereas the opposite effect is seen in Pax6 mutants. These findings suggest thatEmx2 and Pax6 cooperate to regulate arealization of the neocortex and to confer area identity to cortical cells.

The mature mammalian neocortex is tangentially organized into distinct areas that serve specialized functions, such as sensory processing and motor control. These areas can be distinguished by differences in their input and output connections and architecture. It is assumed that the specification and differentiation of neocortical areas are controlled by an interplay between epigenetic (extrinsic) and genetic (intrinsic) mechanisms. Numerous studies suggest that diverse parts of the neocortex initially share potentials to develop features unique to a specific area (1) and have implicated thalamocortical input as a major epigenetic influence in this process (2, 3). Evidence for genetic regulation of arealization is limited to descriptions of graded or restricted patterns of gene expression in the embryonic neocortex (4–6) that are established independent of thalamocortical input (5, 6). Remaining unidentified, though, are genes that regulate arealization, presumably by conferring area identities to cortical cells and controlling the differential expression of downstream genes that govern, among other things, the area-specific targeting of thalamocortical afferents.

The regulatory genes that control neocortical arealization are assumed to be expressed in graded or restricted patterns in the embryonic cortex (3). Two genes that meet this criterion are the homeodomain transcription factor Emx2 and the paired-box–containing transcription factor Pax6. Both genes are expressed in the dorsal telencephalic neuroepithelium, the proliferative zone that gives rise to cortical neurons (7–9). Emx2 is expressed in low rostral to high caudal and low lateral to high medial gradients (7, 8), whereas Pax6 is expressed in low caudal to high rostral and low medial to high lateral gradients (8, 9). On the basis of these expression patterns, we hypothesized that Emx2preferentially imparts caudal and medial area identities to cortical neurons, whereas Pax6 preferentially imparts rostral and lateral identities (Fig. 1).

Figure 1

Predicted changes in neocortical arealization in Emx2 and Pax6 mutant mice. Dorsal views of the surface of mouse neocortex. (A) Organization of the mouse neocortex, showing the auditory area (A1), the motor areas (M1 and M2), the somatosensory areas (S1 and S2), and the visual areas (V1 and V2). M, medial; L, lateral; R, rostral; C, caudal. (B) Predicted change in area identity in Emx2 mutants. The gradient of Emx2expression is high in the caudal and medial neocortex and low in the rostral and lateral neocortex. Thus, it is predicted that, inEmx2 mutant mice, rostral and lateral neocortical areas are expanded caudally and medially, whereas caudal and medial areas are contracted. (C) Predicted change in area identity in Pax6 mutants. The graded expression ofPax6 is opposite to that of Emx2, being high in rostral and lateral neocortex and low in caudal and medial neocortex. Thus, it is predicted that, in Pax6 mutant mice, caudal and medial neocortical areas are expanded rostrally and laterally, whereas rostral and lateral areas are contracted.

We have used molecular and connectional markers of area identity to test these hypotheses in mutant mice lacking eitherEmx2 (10) or Pax6 [Small eye (Sey)] (11) functional alleles. We predicted that, in the Emx2 mutant neocortex, caudal and medial identities would be lost, resulting in a caudal and medial shift in rostral and lateral area identities (Fig. 1B). Conversely, we predicted that, in the Pax6 mutant neocortex, rostral and lateral identities would be lost, resulting in a rostral and lateral shift in caudal and medial area identities (Fig. 1C). Emx2and Pax6 (Sey/Sey) mutant mice die within hours after birth (10, 11), limiting our analysis to the late embryonic period, before the mice die but after some features of arealization have become evident.

The type II classical cadherins, Cadherin6(Cad6) and Cadherin8 (Cad8), are expressed in areal patterns in the late embryonic mouse neocortex:Cad8 has been reported to mark the rostrally located motor area, and Cad6 marks the somatosensory area, located immediately caudal to the motor area, and the auditory area, located in the caudolateral neocortex (12). Because the cortex is reduced in size in both Emx2 and Pax6 mutants compared to wild type (Tables 1 through 5), we determined the proportion of the cortical surface covered by domains of cadherin expression as well as their absolute size (Fig. 2, B and D). No change in proportional sizes of cadherin expression domains would indicate that arealization per se has not been affected, suggesting that the full range of putative area identities is present in the smaller cortex and that all areas are uniformly reduced in size. A change in proportional sizes would indicate that areas are disproportionately affected in the mutant neocortex and therefore thatEmx2 and/or Pax6 has a role in regulating arealization of the neocortex. This finding and interpretation would be most strongly supported by a change in absolute sizes of cadherin expression domains.

Figure 2

Areal patterns of cadherin expression are altered in opposing manners in Emx2 and Pax6mutants. Dorsal and lateral surfaces of the cortex of E18.5Emx2 and Pax6 wild-type (+/+) and mutant (−/−) mice. (A) Cad8 expression is expanded caudally, and Cad6 expression is expanded caudally and medially in theEmx2 mutant neocortex. Arrowheads indicate borders of the cadherin expression domains. (B) Quantification of cadherin expression in the Emx2 wild-type, heterozygote, and mutant cortices. Arrows show the axis for the width and length measurements of cadherin expression domains presented in graphs and in Tables 1 through 5. In proportion to the size of the cortex, Cad8 area and length are increased, whereas Cad6 area and width are increased on the dorsal surface and decreased in area and length on the lateral surface of the Emx2 mutant cortex. (C) InPax6 mutants, cortical Cad8 expression is contracted rostrally, and Cad6 expression is contracted rostrally and laterally. Arrowheads indicate borders of the cadherin expression domains. (D) Quantification of cadherin expression in the Pax6 wild-type, heterozygote, and mutant cortices indicates that, in proportion to overall cortical size,Cad6 and Cad8 expression are decreased in area, length, and width in Pax6 mutants. Scale bars in (A) and (C), 2.0 mm.

Table 1

Cad8 Quantification in the Emx2cortex [mean ± SD (in parentheses)].

View this table:
Table 2

Cad6 dorsal cortex quantification in theEmx2 cortex [mean ± SD (in parentheses)].

View this table:
Table 3

Cad6 lateral cortex quantification in theEmx2 cortex [mean ± SD (in parentheses)].

View this table:
Table 4

Cad8 quantification in the Pax6cortex [mean ± SD (in parentheses)].

View this table:
Table 5

Cad6 quantification in the Pax6cortex [mean ± SD (in parentheses)].

View this table:

The areal pattern of cadherin expression in the Emx2homozygous mutant cortex (13) is substantially different than that in wild type (Fig. 2). The domain of Cad8expression is expanded caudally (Fig. 2A). This expansion appears to be greater along the medial edge of the cortex than farther laterally, suggesting that Cad8 expression may also be expanded medially. The domain of Cad6 expression is shifted caudally and medially, as seen on the dorsal and lateral surfaces of the cortex (Fig. 2A). No changes in the cortical patterns of cadherin expression are observed in Emx2 heterozygous mice (Tables 1 through 3). The domain of Cad8 expression is significantly larger in theEmx2 homozygous mutant neocortex, in both proportional and absolute area, as well as in its proportional and absolute linear extent across the cortical surface (Table 1 and Fig. 2B). In fact, the absolute area of the Cad8 expression domain inEmx2 homozygous mutants is almost double that in wild type, even though the surface area of the dorsal cortex is reduced by one-third. In Emx2 homozygous mutants, the domain ofCad6 expression on the dorsal cortical surface is significantly increased in proportional area and in both proportional and absolute width, compared to that in wild type (Table 2). In contrast, the area and length of the Cad6 expression domain on the lateral cortical surface of Emx2 homozygous mutants each exhibit both proportional and absolute reductions compared to that in wild type (Table 3 and Fig. 2B). These results suggest that areas located in rostral and lateral parts of the neocortex are expanded and shifted caudally and medially in the Emx2 mutant neocortex.

Because Pax6 is expressed in a countergradient toEmx2, we predicted that changes in domains of cadherin expression in Pax6 mutant mice (Sey/Sey mutants) would be in the opposite direction of those observed in Emx2 mutants (14). The domain of Cad8 expression is contracted rostrally in the Sey/Sey cortex compared to that in the wild type (Fig. 2C). The Cad6expression domain is contracted both laterally and rostrally in theSey/Sey cortex (Fig. 2C). The domains ofCad8 and Cad6 expression on the dorsal surface of the Sey/Sey cortex, which is about three-quarters the size of that in the wild type, show both proportional and absolute reductions in area (Tables 4 and 5 and Fig. 2D). Heterozygous mice (Sey/+) show no changes in cortical cadherin expression compared to wild-type mice (Tables 4 and 5). These results suggest that areas located in rostral and lateral parts of the neocortex contract rostrally and laterally in Sey/Sey mutants (15).

Changes in the expression domains of Cad6 andCad8 in Emx2 and Pax6 mutants suggest corresponding changes in neocortical arealization. The primary neocortical areas (motor, visual, somatosensory, and auditory) receive area-specific inputs from the principal motor and sensory thalamic nuclei [ventrolateral (VL), dorsal lateral geniculate (dLG), ventroposterior (VP), and medial geniculate (MG), respectively]. During normal development, thalamocortical axons target and invade their neocortical areas in a precise manner (16). Therefore, as an additional assay for changes in area identity in theEmx2 mutant neocortex, we used retrograde and anterograde axon tracing to map thalamocortical projections (17). This analysis was not done inSey/Sey mutants because thalamic axons do not reach the cortex in these mice (18).

Retrograde labeling from the cortex of Emx2homozygous mutants indicates a caudal shift in the border between the somatosensory and visual areas compared to that of the wild type (Fig. 3A). Injections confined to the cortical plate of the occipital cortex, the location of the primary visual area, normally backlabel neurons in the dLG nucleus. However, inEmx2 mutants, similarly placed injections label cells in the VP nucleus, which normally projects to the primary somatosensory area located rostral to the visual area (Fig. 3A). Deeper injections made into the subplate of the occipital neocortex in Emx2 mutants backlabel neurons in both dLG and VP nuclei, indicating that dLG thalamic axons extend through the subplate below the occipital cortex but fail to invade the overlying cortical plate (19). Retrograde tracing from anterior and posterior portions of the occipital cortex reveals the expected topography in wild-type mice but again indicates a caudal shift in the border between the somatosensory and visual areas in Emx2 mutants (Fig. 3B). In wild-type mice, injections into the anterior occipital cortex backlabel cells in the posterior dLG nucleus, and injections into the posterior occipital cortex backlabel cells in the anterior dLG nucleus (Fig. 3B). In contrast, in Emx2 mutants, injections into the anterior occipital cortex do not label cells in the dLG nucleus but instead label cells in the VP nucleus; injections in the very posterior occipital cortex do label cells in the dLG nucleus, but their number is significantly reduced compared to the wild type (wild type = 319 ± 89, n = 3; mutant = 35 ± 29,n = 3; P < 0.05) (Fig. 3B). These findings suggest that the visual area in Emx2 mutants is contracted and restricted to the extreme caudal part of the occipital cortex.

Figure 3

Changes in area-specific thalamocortical and corticothalamic projections in Emx2mutants indicate a caudal shift of the somatosensory cortex and a contraction of the visual cortex. Sagittal sections through E18.5Emx2 wild-type and mutant brains show retrograde and anterograde DiI (red) and DiA (green) tracing and bisbenzimide counterstain (blue). Rostral is to the left in all panels. (A) Injections into the somatosensory cortex (DiA) backlabel VP thalamic neurons in both wild-type and mutant brains (medial sections). Injections into the occipital cortex at the normal location of visual cortex (DiI) label dLG neurons in Emx2 wild types but label VP neurons in Emx2 mutants (lateral sections). (B) In Emx2 wild types, injections into anterior (DiA) and posterior (DiI) parts of the occipital cortex backlabel neurons in posterior (medial section) and anterior (lateral section) parts of dLG, respectively. However, in Emx2 mutants, injections into the anterior occipital cortex (DiA) backlabel neurons in the VP nucleus (medial section), whereas injections into the posterior occipital cortex (DiI) backlabel neurons in the dLG nucleus (lateral section). (C) DiI injections into the dLG nucleus label axonal branching and backlabel cortical plate neurons (arrowheads) in the occipital cortex in Emx2 wild types. InEmx2 mutants, axons are found in the subplate below the occipital cortex and invade the cortical plate only sparsely. Backlabeled cells (arrowheads) are located more caudally than in wild types. (D) Axonal branches and backlabeled neurons (arrowheads) labeled by DiI injections into VP are shifted caudally inEmx2 mutants compared to wild type. Scale bars, 0.4 mm.

Anterograde tracing of thalamocortical projections is consistent with the retrograde tracing results. Injections into the dLG nucleus of Emx2 mutants label axons in the subplate beneath the caudal occipital cortex, but in comparison to the wild type, few invade the cortical plate (Fig. 3C). Injections into the VP nucleus ofEmx2 mutants label axons that extend farther caudally than in the wild type and aberrantly invade the cortical plate of occipital cortex, whereas in the wild type, VP axons invade the cortical plate of the more rostrally located parietal cortex (the location of the primary somatosensory area) (Fig. 3D). Thalamocortical projections in heterozygous Emx2 mutants resemble those in the wild type (19). Overall, anterograde and retrograde tracing of thalamocortical projections in Emx2 mutants provides evidence for a contraction of the visual area and a caudal shift in the border between the somatosensory and visual areas.

Area-specific connections between thalamic nuclei and neocortical areas are reciprocal. Injections into thalamic nuclei backlabel cortical neurons in wild-type mice and Emx2mutants (Fig. 3, C and D). Injections into the dLG nucleus backlabel significantly fewer cells in the occipital cortex of Emx2mutants compared to the wild type (wild type = 139 ± 4,n = 3; mutant = 19 ± 13, n = 3; P < 0.05). The few labeled cells in Emx2mutants are restricted to the very caudal cortex (Fig. 3C). Injections into the VP nucleus in wild-type mice backlabel cells in the parietal cortex but backlabel none in the occipital cortex (Fig. 3D). In contrast, VP injections in Emx2 mutants backlabel a substantial number of cells in the occipital cortex (Fig. 3D). These findings suggest that, in Emx2 mutants, corticothalamic neurons located in the occipital cortex have acquired a somatosensory area identity instead of their usual visual area identity. These changes in area-specific corticothalamic projections in Emx2mutants are consistent with the changes observed in area-specific thalamocortical projections. Together, they suggest that the primary visual area is substantially reduced and restricted to the very caudal part of the neocortex, with a corresponding caudal shift in the border between visual and somatosensory areas.

Emx2 is reported to be expressed in a small patch of neuroepithelium in the ventral-most part of the dorsal thalamus (20). Although this part does not generate cells of the principal sensory and motor thalamic nuclei (21), we used several markers to confirm the normal development and organization of the dorsal thalamus in Emx2 mutants (22). Patterns of acetylcholinesterase (AChE) staining and Gbx2expression and general morphology revealed by nuclear 4′,6-diamidino-2-phenylindole staining are all normal in the dorsal thalamus of Emx2 mutants (Fig. 4, A and B) (19). Thus, alterations in thalamocortical and corticothalamic projections inEmx2 mutants can be presumed to be due to changes intrinsic to the neocortex.

Figure 4

The dorsal thalamus appears normal in Emx2 mutants. Coronal sections through the E18.5 Emx2 wild-type and mutant thalamus; medial is to the left. (A) AChE histochemical staining indicates that thalamic nuclei that project to primary motor and sensory cortical areas appear normal. (B) Gbx2 expression in the dLG (mid section) and the MG (caudal section) thalamic nuclei is the same in Emx2 mutants as compared to wild type. Scale bars, 0.5 mm.

Our findings implicate Emx2 andPax6 in the genetic control of neocortical arealization. They cannot be explained by a potential delay in neocortical development (23), nor are the expansions and contractions of the cadherin expression domains secondary to a loss of thalamocortical input (15). In addition, the observed changes in cadherin expression domains are not simply a by-product of the reduced overall size of the cerebral cortex in Emx2 and Pax6mutants, because the changes are disproportionate in each mutant and opposing in the two mutants as predicted from the countergradients ofEmx2 and Pax6 expression. Similarly, the observed changes in thalamocortical and corticothalamic projections in theEmx2 mutants are not due simply to a caudal truncation of the neocortex with an associated loss of the visual area. Arguing against this possibility are the changes in cadherin expression inEmx2 mutants, especially the expansion of theCad8 expression domain in the frontal cortex (the motor area), in both proportional and absolute size. Instead, the findings presented here indicate a disproportional, but orderly, arealization of the Emx2 mutant neocortex reflected by an expansion of rostral areas and a contraction of caudal areas, and an opposite effect on arealization in the Pax6 mutant neocortex.

Changes in the areal expression patterns of the cadherins and the area-specific distribution of corticothalamic neurons in the mutants suggest that Emx2 and Pax6 confer area identities to cortical cells, including projection neurons. The changes in cadherin expression and, presumably, receptors for axon guidance molecules that control corticothalamic axon targeting may be indicative of a direct role for Emx2 andPax6 in their regulation, or they may be an indirect effect of the regulation of area identity. Similarly, the changes in area-specific thalamocortical projections suggest thatEmx2 and Pax6 are involved either directly or indirectly in the regulation of axon guidance molecules within the cortex that control thalamocortical axon targeting. The restricted cortical expression of Eph receptor tyrosine kinases and their ligands, the ephrins, which act as axon guidance molecules in several systems, makes them candidates for controlling the development of area-specific projections between the thalamus and cortex (4, 24).

Emx2 and Pax6 may act independently or in a combinatorial manner (possibly with other transcription factor genes) to specify neocortical areas. Because areas in the neocortex have sharp borders, it is likely (but not required) that the graded expression patterns of Emx2 andPax6 are translated to regulate some downstream genes in restricted patterns with abrupt borders that relate to areas. Studies in Drosophila have shown that gradients of transcription factors can be translated through thresholding mechanisms and combinatorial actions into sharply bordered expression patterns of downstream genes. For example, a single regulatory protein, such as Dorsal, distributed in a graded manner across several tissues can establish sharply patterned, tissue-specific expression of downstream genes through a thresholding mechanism based on concentration-dependent differences in binding efficacy to promoter and repressor elements (25). Analysis of the regulation of theeven-skipped gene has shown that multiple transcriptional activators and repressors expressed in graded patterns act combinatorially to generate sharp borders of even-skippedexpression (26). Although the downstream targets ofEmx2 and Pax6 in the cortex have yet to be identified, transcription factors such as T-brain1 andId2 are expressed in the neocortex in discrete patterns with abrupt borders (27) that may be controlled by upstream regulatory genes expressed in gradients.

Emx2 and Pax6 appear to be independently regulated (28). The opposing gradients ofEmx2 and Pax6 may be induced by signals secreted from the poles of the cortex. Several secreted proteins are candidates for these inductive signals, including the BMP, WNT (2b, 3a, 5a, and 7a), and FGF8 proteins (29). In addition, cortical expression of the transcription factor Gli3 is required forEmx2 expression (30). Thus, combinations of inductive signals and upstream transcription factors may specify gradients of Emx2 and Pax6. A better understanding of the roles of Emx2 and Pax6 in regulating neocortical arealization will require identifying the patterning mechanisms that establish their differential expression, identifying downstream targets, and defining the mechanisms by which they, in combination with other factors, intrinsic and extrinsic, control the process of arealization of the neocortex.

  • * To whom correspondence should be addressed. E-mail: doleary{at}


View Abstract

Navigate This Article