Research Article

Foxg1 Suppresses Early Cortical Cell Fate

See allHide authors and affiliations

Science  02 Jan 2004:
Vol. 303, Issue 5654, pp. 56-59
DOI: 10.1126/science.1090674


During mammalian cerebral corticogenesis, progenitor cells become progressively restricted in the types of neurons they can produce. The molecular mechanism that determines earlier versus later born neuron fate is unknown. We demonstrate here that the generation of the earliest born neurons, the Cajal-Retzius cells, is suppressed by the telencephalic transcription factor Foxg1. In Foxg1 null mutants, we observed an excess of Cajal-Retzius neuron production in the cortex. By conditionally inactivating Foxg1 in cortical progenitors that normally produce deep-layer cortical neurons, we demonstrate that Foxg1 is constitutively required to suppress Cajal-Retzius cell fate. Hence, the competence to generate the earliest born neurons during later cortical development is actively suppressed but not lost.

In both invertebrate (1, 2) and vertebrate (3, 4) central nervous system development, neuronal progenitors produce specific cell types in a characteristic temporal order. Analysis in the mammalian brain (57) and retina (810) suggests a general rule governing this process: Neural progenitors can produce cells characteristic of later but not earlier points in development. The mechanism behind this progressive restriction in progenitor potential is not understood. The laminar cell fate in the mammalian cortex provides an excellent model for studying these changes in progenitor potential. The mammalian cerebral cortex comprises six layers of neurons that are generated in an orderly sequence during development (11, 12). With the exception of the Cajal-Retzius (CR) cells, which reside in layer 1, the cerebral cortex is produced in an inside-out manner. The deeper layer cells exit the ventricular zone (VZ) first, followed by more superficial cells at later periods. Hence, the birthdate of a cortical neuron is predictive of its fate (1315). Furthermore, cell transplantation studies suggest that early-born classes of neurons can adopt later cell fates but not the converse (57). Thus, during cortical development, there appears to be a ratcheting mechanism by which the potential of early progenitors is progressively restricted.

The first restriction in the neuronal cell types that cortical progenitors can generate is the transition from the production of CR cells to the production of deep-layer neurons. CR cells, in addition to being the first postmitotic population, are of particular importance for the development of a properly organized cerebral cortex (16, 17). CR cells reside in the subpial region of layer 1 and secrete the extracellular glycoprotein Reelin (18, 19), which provides a critical signal for the guidance of later born cells that populate the cortical laminae. One of the few genes known to affect this early phase of cortical development is Foxg1, which encodes a winged helix transcriptional repressor (2022). Foxg1 controls the number of cells produced in the cortex and the loss of this gene results in hypoplasia of the cerebral hemispheres (Fig. 1A) (23, 24). However, the principal mechanism by which Foxg1 regulates this early step in cortical development has remained elusive.

Fig. 1.

Laminar specification is lost in the Foxg1–/– cortex. (A) Coronal sections through the forebrain stained with cresyl violet of E18.5 Foxg1+/– heterozygote (left) and Foxg1–/– (right) mice. The boxed regions in the upper left of each panel indicate the regions shown in (B) [left panel in (A) corresponds to top panel in (B); right panel in (A) corresponds to bottom panel in (B)]. The Foxg1–/– mutant telencephalon is substantially reduced in size, but appears thicker in lateral cortical regions as a result of the accelerated neuronal differentiation. tel, telencephalon; di, diencephalon. Scale bar, 1 mm. The MZ (mz), cortical plate (cp), intermediate zone (iz), subventricular zone (svz), and VZ (vz) of Foxg1+/– animals are morphologically indistinguishable by Nissl staining from wild-type mice. In the Foxg1–/– cortex, the VZ appears thinner and the cortical cytoarchitecture is disrupted. (C) Expression of the layer-specific markers (Foxp2, ER81, Otx1, Foxp1, RORβ, Reelin) in the cortex (top panels are Foxg1+/–; bottom panels are Foxg1–/–). Markers are aligned from the deep layers (Foxp2) to superficial layers (Reelin). In Foxg1–/– mutants, all examined markers of cortical laminae are lost. However, the number of reelin-expressing CR neurons in the telencephalon of Foxg1–/– mice is increased, and CR neurons are dispersed throughout the cortex. Scale bar, 100 μm.

Most cortical neurons in Foxg1 null mutants become CR cells. To address which cortical cell types are generated in the Foxg1–/– mutants, we examined the expression of layer-specific markers at embryonic day 18.5 (E18.5), the latest time point at which these mutants are viable. In the E18.5 wild-type cortex, only the deeper layers (layers 4 to 6) have achieved their mature laminar organization. These layers are characterized by the expression of ER81, Otx1, Foxp1, Foxp2, and RORβ (Fig. 1C) (2528). In situ analysis of the telencephalon revealed that the Foxg1–/– mice failed to express any of these genes, demonstrating that these cortical laminae are absent in these mutants.

In contrast, we observed that the earliest born CR neurons were not only present but supernumerary in these mice (Fig. 1C, bottom right panel). CR neurons can be identified by a number of criteria. Foremost among these is their expression of Reelin. In addition, CR cells express both calretinin (29) and CXCR4 (30). Consistent with the Foxg1 telencephalon's possessing increased numbers of CR cells, we observed the widespread expression of both of these markers in the cortex of the mutant mice (fig. S1).

To determine how this overproduction of CR neurons occurs, we examined progressive stages of corticogenesis in these mutants. The cortex of Foxg1–/– mice appeared phenotypically indistinguishable from wild-type animals from E8.5 through E10.5. Furthermore, at E10.5, when the normal production of CR cells is occurring, we observed that the distribution and number of these cells is comparable in Foxg1 null mice compared with wild-type littermates. During normal cortical development, although all postmitotic cells express microtubule-associated protein 2 (MAP2), only those cells in the most superficial layer of the cortex express CR-50. In contrast, in the Foxg1 null mutant cortex, the entire MAP2 population appears to express CR-50 (fig. S2 and Fig. 2).

Fig. 2.

Supernumerary production of CR neurons in the Foxg1–/– cortex. Boxed areas in the cresyl violet–stained coronal sections (panels on the left) represent the regions enlarged in (A to C) and (D to F), respectively. The CR cell population is visualized by CR-50 immunoreactivity (green). At E14.5, a substantial population of MAP2-positive neurons (red) has been generated in both Foxg1+/– (A) and Foxg1–/– cortex (D). CR cells are restricted to the superficial layer in the Foxg1+/– heterozygotes (B). In Foxg1 mutant mice, all postmitotic cells appear to coexpress CR-50 and MAP2 (F). These results indicate that CR cell production progresses concomitantly with neurogenesis in the absence of Foxg1 gene.

Foxg1cell-autonomously represses CR cell identity. To evaluate whether the repression of CR cell production is cell-autonomously regulated by Foxg1, we assessed the normal expression of Foxg1 in this population with a Foxg1lacZ/+ transgenic. Except for the presence of β-galactosidase–positive cells, Foxg1lacZ/+ heterozygous mice are phenotypically indistinguishable from wild-type mice (23). Previous studies have demonstrated that Foxg1 is expressed in progenitor cells (20). We found that expression persisted in postmitotic neurons in the cortical plate (Fig. 3A). However, we observed that LacZ expression was excluded from the early-born CR neurons. At E18.5, individual CR cells were seen within the marginal zone (MZ), a region only sparsely populated with cells (Fig. 3C). Although a few LacZ-positive cells were also detected in the MZ of Foxg1lacZ/+, these likely represent cells that migrated from the ventral telencephalon during later neurogenesis (31). Consistent with this notion, these cells never coexpressed CR-50 (Fig. 3C).

Fig. 3.

Foxg1-LacZ expression is absent in CR cells. Coronal sections of the lateral cortex at E14.5 (A and B) or E18.5 (C and D) double-labeled to visualize CR-50 (green) and LacZ (red) [(A) and (C)] and corresponding 4′,6′-diamidino-2-phenylindole nuclear stain [(B) and (D)] in the Foxg1lacZ/+ cortex. Foxg1-lacZ is expressed in the majority of the cells in the cortical plate (cp) but is excluded from both E14.5 (A) and E18.5 (C) MZ (mz) CR cells. Arrows in (C) and (D) indicate CR cells in the MZ in the E18.5 cortex, all of which are LacZ-negative.

Our results suggest that Foxg1 cell-autonomously represses CR cell identity. Conversely, we found evidence that Foxg1 expression was repressed in CR cells. Specifically, the complementary expression of LacZ and CR-50 cells persisted in Foxg1lacZ/lacZ null cortex. In these mice, LacZ was expressed in neural progenitor cells before CR cell differentiation but was absent in mature CR cells in the cortex (fig. S3). Taken together, these results suggest that early neuronal fate is suppressed by Foxg1 in later progenitors in the wild-type cortex.

The absence of Foxg1 expression in CR cells in the cortex raises the question of whether these two populations are segregated from the onset of neural development. To address this issue, we crossed Foxg1Cre/+ mice (32) onto a ROSA26 reporter (R26R) line (33). This allowed us to identify cells that expressed Foxg1 at any point of their development by evaluating whether β-galactosidase was produced. Examination of the Foxg1Cre/+:R26R cortex revealed that the majority of the CR cells were LacZ-positive (Fig. 4, A and B). This suggests that most CR cells express Foxg1 at some point during their development. We envisioned two scenarios: (i) The total complement of CR cells may be committed before the onset of Foxg1 expression, making them refractory to the effects of Foxg1; or (ii) continued Foxg1 expression may be required to suppress neurons born later in cortical development from adopting an early neuronal fate. This raises the question of whether the inability of later progenitors to give rise to CR cells reflects a loss of competence or suppression by Foxg1.

Fig. 4.

LacZ expression in the Foxg1Cre/+:R26R cortex at postnatal day 0. Double-labeling of CR-50 (brown) with β-galactosidase staining (blue) demonstrates colabeled CR neurons [arrowheads in (A)] in the MZ. Scale bar, 100 μm. (B and C) Enlarged views of the cortex reveal that the majority (∼90%) of CR cells express LacZ (B). LacZ-negative CR cells [arrow in (C)] were, however, occasionally detected in the MZ, indicating that small numbers of CR cells never express Foxg1.

To differentiate between these possibilities, we used the tet-transactivator (tTA) system (24) to conditionally remove the Foxg1 gene function in the progenitor cells after the normal birthdate of CR cells had passed. To achieve this, we replaced the endogenous Foxg1 gene with a tTA-regulated Foxg1 gene by generating Foxg1lacZ/tTA:tetOFoxg1-IRESlacZ mice (Foxg1-rescued mice, designated as Foxg1tetO-foxg1; IRES, internal ribosome entry site). We confirmed that in mice where Foxg1 expression was rescued, cortical lamination was restored (fig. S4) and the CR neurons were confined to the superficial layer as in wild-type animals.

Removal ofFoxg1function at E13 reinitiates CR cell production. We next selectively removed Foxg1 gene function in these mice at E13 by administering doxycycline (2 mg/ml daily in drinking water). We refer to these mice as Foxg1tetO-foxg1-E13Doxy. In wild-type mice (and in Foxg1tetO-foxg1 mice), this is the time point when layer 5 neurons are being generated. Interestingly, the removal of Foxg1 expression at E13 resulted in the resumption of CR cell production in the cortex (Fig. 5A, compare the left and middle panels with the right panel). Indeed, the number of CR cells in the MZ of these animals was increased by over 50% on the basis of the numbers of Reelin and CXCR4 cells in this region (fig. S5).

Fig. 5.

CR cell production is reinitiated in Foxg1tetO-foxg1-E13Doxy mice (i.e., mice in which Foxg1 gene function is removed at E13). (A) The expression of reelin in the E16.5 cortex of Foxg1+/– heterozygotes, Foxg1tetO-foxg1, and Foxg1tetO-foxg1-E13Doxy mice. Expression is restricted to the marginal layer in Foxg1tetO-foxg1 cortex [center panel in (A)]. Ectopic and supernumerary CR cells in the cortex are observed in Foxg1tetO-foxg1-E13Doxy mice [right panel in (A)] (fig. S5). (B) Reelin (green) and BrdU (red) double-labeling demonstrates coexpression of reelin in BrdU-labeled cells in Foxg1tetO-foxg1-E13Doxy mutants, indicating that the ectopic CR cells are newly born. In the Foxg1+/– control littermates, which were fed with doxycycline, no reelin-expressing cells were detected in the cortical plate.

Newborn E13 neurons become CR cells in the absence ofFoxg1gene function. To determine whether CR cells in Foxg1tetO-foxg1-E13Doxy mice arose from newborn neurons, we used 5-bromo-2′-deoxyuridine (BrdU) to pulse-label dividing cells in these mice subsequent to the administration of doxycycline. The majority of BrdU-labeled cells in these animals expressed reelin (Fig. 5B, bottom right panel), indicating that these CR neurons likely represent cells that exited the cell cycle subsequent to removal of Foxg1 gene function. These results suggest that in Foxg1tetO-foxg1-E13Doxy mutants, neural progenitors that are normally committed to generating deeper layer neurons revert to producing the earliest born CR neurons. To confirm this, we followed the fate of deep-layer neurons in Foxg1tetO-foxg1-E13Doxy mice. We found that fewer cells expressed ER81, and those that were generated were not colabeled with BrdU (fig. S6). This suggests that ER81 neurons observed in these animals were born before doxycycline administration. In control mice, many ER81 cells were BrdU-positive, showing that under normal conditions they are generated at this stage. Moreover, ER81 cells and reelin cells never overlapped in the Foxg1tetO-foxg1-E13Doxy mutants (fig. S6). This suggests that the neurons rising from progenitor cells in which Foxg1 gene function is removed do not adopt a hybrid deep-layer/CR cell fate.

In summary, we have provided evidence that Foxg1 functions cell-autonomously during early phases of cortical neurogenesis to prevent the generation of CR neurons after their normal birthdate. Hence, although cortical progenitor cells are competent to produce CR neurons during later neurogenesis, Foxg1 normally acts to suppress this early-born cell fate during the subsequent generation of the cortical laminae. It will be interesting to assess at which time point during development the progenitor cells lose their competence to revert to earliest born neurons in the absence of Foxg1.

Supporting Online Material

Materials and Methods

Figs. S1 to S6


References and Notes

View Abstract

Navigate This Article