Sp8 mutant brains exhibit multiple malformations
At E8.0-E8.5, Sp8 is strongly expressed in the anterior neural ridge (Figure 1a). At stage E9.5, Sp8 mRNA is apparent in the entire forebrain anlage in a rostral-ventral/high, caudal-dorsal/low gradient. Additionally, Sp8 transcripts were detected in the olfactory placode (Figure 1b; data not shown). After E10.5, cortical Sp8 mRNA levels decrease, although keeping the typical mediolateral expression gradient in cortical progenitors (Additional data file 1). During later developmental stages, Sp8 transcription is further down-regulated within the cortical ventricular zone (VZ), but remains strongly evident throughout adulthood in the septum (SE), dLGE and olfactory system  (data not shown). To gain more insights into the role of Sp8 during forebrain development, we generated mice with a floxed Sp8 locus. Sp8 floxed mice were crossed with mice expressing the Cre-recombinase under the control of the Foxg1 promoter  (Additional data file 1) and the floxed allele was bred to homozygozity (termed cKO).
Homozygous Sp8 mutants died at birth. Mutant E10.5 forebrains lacked detectable levels of Sp8 mRNA (Figure 1d). At midgestation cKO embryos showed strong craniofacial abnormalities (data not shown). Nissl-stained histological sections revealed that cKO embryos displayed a dysgenesis of the olfactory bulbs (data not shown) and SE, including an almost complete absence of the midline (15% penetrance, n = 25) that resulted in a mild rostral holoprosencephaly (Figure 1e, e', e''; Additional data file 2). In addition, the thickness of the cortical plate of the cKO cortex was reduced (Figure 1g, g'; 63.5% ± 5.1% of control, n = 10). The basal ganglia consisted of a single eminence (Additional data file 2) with a barely discernable constriction between the LGE and MGE. Corticofugal fiber tracts did not cross the midline and instead formed probst bundles (Figure 1e, e', e''). Caudally, neuronal fibers formed bundles between the internal and external capsules (arrows in Figure 1f, f', f'').
Sp8 modulates the D/V patterning of the medial telencephalon
The loss of Sp8 activity in the forebrain provokes structural perturbations of the SE and MGE and prompted us to look for molecular markers associated with forebrain patterning. Whole mount in situ hybridization (WMISH) at E9.5 (after the initiation of Cre activity ) revealed that the mean expression levels of Pax6 and Emx2 are, conversely, down- and up-regulated in cKO (Figure 2c, c', d, d') compared to wild-type embryos. Importantly, the activity of both Fgf8 and Shh seemed unchanged in the absence of Sp8 (Figure 2a, a', e, e'). Remarkably, Nkx2.1 mRNA  was reduced in cKO. Rostrally, the septum anlage was free of Nkx2.1 transcripts (red arrow in Figure 2b, b'), and the caudal portion of the MGE anlage was devoid of Nkx2.1 activity (white arrow in Figure 2b, b'). At E10.5, Fgf8 mRNA was evident in the telencephalic midline, including the septum anlage (Figure 2f, f'). Interestingly, at E12.5, we found that the Fgf8 expression domain in the SE of the mutants was specifically lost (Figure 2g, g'), while the expression of Shh [4,7] appeared to be unaffected (data not shown). This suggests that Sp8 function might be required for the maintenance of the late expression of Fgf8 in the telencephalic midline.
Next, we examined the patterning of the expression interface at the border of the pallium/subpallium in the medial telencephalic wall, designated medial pallial-subpallial boundary (mPSB). At the mPSB, both the Emx2+ and Pax6+ expression domains were expanded ventrally (arrows in Figure 3c, c', d, d'). Similarly, the Pax6 target gene , Ngn2, displayed ectopic expression in the presumptive territory of the SE (Figure 3e, e'). In accordance with the reported inhibition of the ventral marker Mash1 by Ngn2 , the Mash1+ territory at the mPSB was reduced in cKO (Figure 3j, j'). A similar down-regulation of the expression of the ventral markers Gsh2 [20-22] and Dlx1  (arrows in Figure 3h, h', g, g') was also evident in the dorsal SE. At E15.5, the Ngn2 expression domain was still massively enlarged in the medial telencephalic wall, and thus, the expression domain of Dlx1 remained severely shrunken (Figure 3l, l', m, m'). Intriguingly, despite the partially preserved expression domain of Nkx2.1 in the presumptive MGE territory at E9, examination of E12.5 and E15.5 mutant embryos revealed that Nkx2.1 expression was almost absent from the subpallium (Figure 3f, f', n, n'). The only exception was a small stripe in the putative caudal ganglionic eminence (data not shown). Furthermore, the Nkx6.2+ domain that normally marks the MGE/LGE junction  was shifted medially and reached the ventrally expanded Emx2+, Pax6+, and Ngn2+ midline territories in cKO (Figure 3k, k' ; see also Figure 3c, c', d, d', e, e'). Given the loss of substantial parts of the MGE during embryonic development, where the neuronal progenitors mainly generate interneurons , cKO cortices contained a strongly reduced amount of Gad67+ interneurons at E18.5+ (20.3 ± 2.7% of wild type, n = 3; Figure 3o, o'). Our analysis further indicated that the expression boundaries of Pax6 (Figure 3b, b'), Ngn2 (Figure 3e, e'), Dlx1 (Figure 3g, g'), Mash1 (Figure 3j, j'), and Gsh2 (Figure 3h, h') at the lateral pallial-subpallial boundary (PSB) were not altered in Sp8-deficient brains. This demonstrates that the molecular identity of the lateral pallium and LGE [17,20-22] is not affected by the absence of Sp8. In contrast, our data are consistent with a role for Sp8 in the D/V patterning in the medial telencephalic wall.
To study whether the Sp8 mutation might affect pattern formation through the Wnt signaling pathway , we examined the expression of Wnt3a, Wnt5a and Wnt7b in the cortical hem and the Wnt antagonist Sfrp2 in the antihem. No change in the expression of these markers was apparent (data not shown). We concluded that Sp8 might act downstream or independently of Wnt signaling [5,7,10].
Abnormal cortical arealization and thalamic innervation in Sp8 mutants
In the embryonic brain, Sp8 and Emx2 show similar expression characteristics along the medial-lateral axis (high at caudomedial levels and low at rostrolateral levels; Figure 3; Additional data file 1). Interestingly, the pallial Emx2 gradient is clearly up-regulated in mutant embryos at E12 (arrow in Figures 3a, a' and 4a, a'). Conversely, Pax6 normally displays an expression gradient that is high in the rostrolateral and low in the caudomedial pallium, opposing that of Emx2. These two genes mutually control their activities in the cortical neuroepithelium . In accordance with the enhanced expression level of Emx2, Pax6 mRNA is clearly reduced in cKO (arrow in Figure 3b, b' ; Figure 4b, b'). Given the crucial roles played by Emx2 and Pax6 in the arealization of the neocortex [1-3,10,24,25], their altered expression in Sp8 mutants prompted us to analyze cortical arealization. The expression of the EphA7 receptor and the EphrinA5 ligand specifically demarcates the regions of the motor/visual and somatosensory cortex, respectively [24,26]. In the cKO, the EphrinA5 somatosensory domain extends rostrally (Figure 4c, c'). Accordingly, the motor cortex area, which normally expresses EphA7 at high levels, shrinks (Figure 4d, d'). A similar alteration was observed for the characteristic expression of Coup-TF1, normally showing a prominent caudal/high to rostral/low expression gradient [21,27]. In Sp8 mutants the strong caudal expression domain of Coup-TF1 expanded much more rostrally, reaching the presumptive motor cortex (arrows in Figure 4e, e'). A caudalization of the molecular properties along the A/P axis of the cortex was also evident, when ID-2 expression was examined. High ID-2 levels in upper cortical layers (arrowhead in Figure 4f) of the motor cortex and in layer V in the caudal cortex normally highlight the border between the motor and somatosensory domains [2,3] (red arrow in Figure 4f). In Sp8 mutants the caudal ID-2 expression territory extends ectopically into the rostral cortex. Additionally, in the presumptive motor cortex, ID-2 expression in upper cortical layers was abolished (Figure 4f').
To study whether the observed molecular caudalization of the Sp8cKO cortex reflects alterations of the cortical area identity, we performed retrograde labeling of thalamocortical (TCA) projections by placing crystals of the lypophilic dies DiI (red) and DiO (green) in the presumptive visual cortex or somatosensory area, respectively (insets in Figure 4g, g'). In controls, the red dye was exclusively present in cells within the dorsal lateral geniculate nucleus (dLGE; dorsal from the dashed line in Figure 4g), while the DiO labels (in green) cells in the ventroposterior complex (VP; ventral from the dashed line in Figure 4g). In the Sp8KO brain, the dLGE was labeled by TCA projections coming from both visual and somatosenory cortex (Figure 4g'). These results suggest that the molecular caudalization of the Sp8KO cortex causes a partial change in the cortical area identity (somatosensory to visual fate).
To elucidate whether the regulation of Emx2 by Sp8 might be direct or indirect, we performed biochemical in vitro experiments. Glutathione S-transferase (GST)-pull down assays revealed that truncated Emx2 protein, lacking the homeobox, does not interact with GST-Sp8 or GST-Sp8 without zinc fingers (Figure 4h, lanes 3 and 4). However, we found that GST-Sp8 as well as GST-Sp8 lacking zinc fingers are able to bind the full-length Emx2 protein in vitro (Figure 4h, lanes 7 and 8). Taken together, this indicates that the Sp8 expression gradient in cortical progenitors plays an important role in the correct positioning of distinct cortical domains along the A/P axis of the developing cortex by modulating the expression level of Emx2. Furthermore, our findings support the idea of a direct interaction between Sp8 and Emx2 proteins.
Sp8 controls cell survival in the developing forebrain
The hypoplasia provoked in the forebrain of cKO led us to speculate that this may be related to defects in cell proliferation and/or apoptosis [28,29]. Such alterations were additionally observed in Foxg1 knockout mice . However, Foxg1 activity was not affected in Sp8 mutants (data not shown). We therefore determined the bromodeoxyuridine (BrdU) labeling index (BrdU pulse), M-Phase index (pH3 staining), and cell cycle exit [30,31] (BrdU/iododeoxyuridine (IdU) double labeling) parameters in the forebrain of cKO at E12.5. None of these parameters were significantly altered in mutant forebrains (Additional data file 3). Furthermore, neither ectopic mitosis nor premature differentiation seems to occur in cKO (Additional data file 3). However, using a TUNEL assay at E12.5, we detected a dramatic increase of apoptosis in forebrain sections (Figure 5a, a'). TUNEL+ cells were found randomly distributed in the dorsal and basal telencephalon (Figure 5a, a', b, b' ; data not shown). The apoptotic cells formed clusters consisting of three to six individual nuclei (arrow in Figure 5c). Counting of TUNEL+ cells on E10.5, E15.5 and E18.5 sections revealed that the Sp8-deficient forebrains contained six times more apoptotic cells at E10.5 and three times more from E12.5 to E18.5 compared to controls (Figure 5e). TUNEL+ nuclei were found in putative proliferative regions (white arrows in Figure 5d) and in the cortical plate proper (CP; blue arrow in Figure 5d). Sp8 deficiency possibly affects the survival of some post-mitotic neurons (Figure 5d; 78.4% TUNEL+/Tuj- at E15.5, n = 2), but mainly early (E10.5-E12.5) neuronal progenitors. Thus, the forebrain hypoplasia in cKO appears as a consequence of the apoptosis-induced loss of progenitors.
We next wanted to know whether the observed loss of progenitors might affect the generation of specifically early- or late-born cortical neurons . By injecting BrdU at E12.5 or E15.5 and sampling at E18.5, we created specimens that had BrdU exclusively incorporated in early- (E12.5) or late-born (E15.5) neurons. Analysis of the samples injected at E12.5 identified cells populating deep and intermediate positions of the CP (Figure 5f). In accordance with the detected apoptosis, the number of BrdU+ nuclei in mutant brains was reduced (Figure 5f' ; 64.7 ± 7.8% of wild type, n = 3), pointing to a diminished progenitor pool. Cells labeled with BrdU at E15.5 populated mainly deep compartments (VZ, subventricular zone (SVZ), intermediate zone (IZ)) of the cortex of both genotypes (Figure 5g, g'). However, the amount of BrdU+ cells within the putative SVZ, which were recently shown to generate exclusively upper cortical layers [33-35], appeared reduced in the mutants (yellow arrow in Figure 5g'). Conversely, more BrdU-labeled cells were detected in superficial positions in the CP (white arrow in Figure 5g'), most probably reflecting the reduction in the distance from the ventricular to the marginal zone in the mutant cortex.
The loss of Sp8 results in defective preplate splitting
The apoptotic cell death detected in the forebrain of Sp8 mutant embryos may also affect neurogenesis and early cortical layer development. The cortical preplate (PPL) consists of early-born (E10-E11.5) neurons of pallial origin. Later-born neurons will progressively split this domain into the marginal zone (MZ), CP and subplate (SP) [36,37]. We used the Tbr1 riboprobe to label early PPL and SP populations [38,39] at mid-gestation. While in controls Tbr1+ SP cells are well separated from Tbr1+ CP cells (Figure 6a), this does not occur in cKO embryos (Figure 6a').
Furthermore, we used a cell labeling approach consisting of injecting BrdU at E11 to label SP cells [36,40] and then harvesting tissue after a visible separation of SP and CP at E15 (Figure 6a). In addition, the co-detection of BrdU and Tbr1 enabled us to follow the laminar position of the double positive cells, which in controls were found in the SP (arrow in Figure 6b). In contrast, in cKO these cells were located in virtually the most superficial part of the cortical plate (compare Figure 6b, b'). We conclude that proper PPL splitting does not occur in Sp8cKO.
In accordance with defective SP formation [37-40], Gap43 antibody staining revealed that the subcortical connectivity and the axonal wiring appeared abnormal in mutants. In contrast to controls, Gap43+ fibers in Sp8 mutants formed aberrant bundles within the internal capsule, with some axons projecting ectopically towards the MZ (arrows and arrowheads in Figure 6d, d') and basal telencephalon (Additional data file 2 (d, d'')).
Perturbed specification of distinct cortical layer neurons in Sp8 mutants
Because the detected PPL defect may provoke additional abnormalities, we investigated whether the generation of infra-/supragranular layers and the differentiation of specific neuronal layers were altered in cKO. We therefore labeled late-born/upper cortical neurons with the Cux2 riboprobe [34,35], combined with immunolabeling for Tbr1, tracing early-born/deep layer neurons . We could not observe co-localization (Figure 7a, a') of both markers, demonstrating that the basic laminar organization of the cortex and the switch from early to late neuronal fate is not altered in mutants. To examine whether Sp8 might be required for the specification of distinct cortical neuron subtypes, we assayed layer-specific marker genes at E18.5. Recent findings [36,41] indicate that Tbr1 activity is able to promote the specification of the SP and layer VI. In accordance with defective preplate splitting and the apoptotic cell death of progenitors in cKO, the analysis of Tbr1 expression by immunohistochemistry or in situ hybridization revealed ectopic Tbr1+ cells in the uppermost region of the CP (arrows in Figure 7l'), and a less thick band corresponding to layer VI of the deep CP (Figure 7k, k'). In addition, ER81 was used to trace a subpopulation of layer V neurons , but we could not detect ER81 mRNA in CP neurons at E18.5 (Figure 7j, j'). However, the expression of another marker of layer V neurons (Robo1 ) appeared to be only diminished in cKO (Figure 7h, h'), suggesting that only a subset of lower cortical layers might not be correctly specified.
The orphan nuclear receptor RzR-β was utilized to follow layer IV genesis . At E18.5, RzR-β transcripts were completely missing in the CP of mutants (compare Figure 7f, f'). Moreover, the expression of Cux proteins in SVZ progenitors was recently shown to promote the fate specification of late-born neurons [34,35]. Accordingly, we found that in the mutant, although Cux2 expression was reduced (Figure 7d, d'), Cux1 mRNA could not be detected at E18.5 (Figure 7c, c'). Along the same line of evidence, the expression of an additional upper layer neuron marker, Lhx2, is also highly down-regulated in the cortical plate of Sp8 mutants (Figure 7e, e'). This suggests that a reduction in the generation of late-born/upper cortical layer neurons occurs. We assayed Tbr2 immmunoreactivity. Tbr2 is a specific marker for basal/SVZ progenitors , which predominantly generate the upper cortical layers [34,35]. In Sp8 mutants, the population of Tbr2+ (basal) progenitors was significantly reduced at E18.5 (49.4 ± 4.3% of controls, n = 3; Figure 7g, g'). This is consistent with a diminished pool of late progenitors, resulting in a diminished generation of upper cortical layer neurons in conditional Sp8 mutants.
The MZ mostly consists of Reelin+ Cajal-Retzius cells [1,9,37]. Using in situ hybridization (ISH) for Reelin mRNA (and a Reelin antibody for quantification) we found more Reelin+ neurons in cKO than in control littermates (142.2 ± 6.4% of controls; Figures 6c, c' and 7b, b'). In summary, these findings support the idea that the lack of Sp8 function during early neurogenesis is responsible for a severe depletion of the early and the late cortical progenitor pool, resulting in a misspecification of distinct cortical neuron subtypes, such as Cux1+, Lhx2+, RzR-β+, and ER81+ lineages.