Mutations in the BMP pathway in mice support the existence of two molecular classes of holoprosencephaly

The cerebral hemispheres, the seat of higher cognitive function in mammals,develop from the embryonic telencephalon. The telencephalon becomes divided into morphologically distinct left and right hemispheres shortly after neural tube closure. Dorsally, the hemispheres are separated by two midline cell types: the choroid plexus, which produces the cerebrospinal fluid, and the cortical hem, which lies between the choroid plexus and hippocampal primordium. The molecular pathways that generate these midline cell types are poorly understood.

Holoprosencephaly (HPE), the most common congenital forebrain defect in humans (Wallis and Muenke,2000; Muenke and Beachy,2000), is characterized by a loss of midline tissue. The large spectrum of phenotypic severity and midline areas affected in inherited cases of human HPE are consistent with a complexity of genetic pathways regulating midline formation (Nanni et al.,1999; Ming and Muenke,2002). In only a minority of HPE cases have mutations at known genetic loci been identified (Wallis and Muenke, 2000; Muenke and Beachy, 2000). In most identified cases, the mutated genes normally function in the ventral telencephalon. Disruption of the ventrally acting SHH or Nodal signaling pathways leads to HPE(Wallis and Muenke, 2000; Muenke and Beachy, 2000; Hayhurst and McConnell, 2003)and although not yet linked to HPE in humans, loss-of-function mutations in the mouse genes Cdo (Cdon - Mouse Genome Informatics; which encodes a multifunctional transmembrane protein), Lrp2 (a low-density lipoprotein-related receptor) and Chrd and Nog (secreted BMP antagonists), exhibit loss of SHH expression and HPE(Anderson et al., 2002; Zhang et al., 2006; Spoelgen et al., 2005).

Of the genes known to cause HPE in mice, only mutations in Lrp2,Chrd;Nog, Zic2 and Fgf8 have been reported to display clear alterations in dorsal midline development(Anderson et al., 2002; Spoelgen et al., 2005; Nagai et al., 2000; Storm et al., 2006). Whereas Lrp2 and Chrd;Nog mutants exhibit excessive dorsal midline features such as BMP expression and apoptosis(Furuta et al., 1997; Anderson et al., 2002; Spoelgen et al., 2005), the dorsal midline appears to be lost in the holoprosencephalic Zic2 and Fgf8 hypomorphic mutants (Nagai et al., 2000; Storm et al.,2006). ZIC2 is also the only locus implicated in a subclass of HPE in humans, the midline inter-hemispheric HPE (MIH), which is associated with impaired dorsal midline formation, whereas in other classes of HPE the ventral medial area is primarily affected(Barkovich and Quint, 1993; Simon et al., 2002; Marcorelles et al., 2002; Nagai et al., 2000). Note that most ZIC2 mutations also affect ventral midline formation(Brown et al., 2001).

Bmp2, 4, 5, 6 and 7 are expressed in the dorsal midline,prior to separation of the hemispheres(Furuta et al., 1997). The receptors are expressed broadly in telencephalic precursor cells. BMP4-soaked beads placed on explants of lateral telencephalon induce midline features(apoptosis and Msx1 expression) and inhibit lateral features[proliferation and Foxg1 expression(Furuta et al., 1997)]. BMP signaling is also crucial in specifying choroid plexus development(Panchision et al., 2001; Hebert et al., 2002). These findings have led to the hypothesis that BMP signaling is required to form the dorsal midline.

Breeding and genotyping of Foxg1^Cre,Bmpr1a^fx/fx and Bmpr1b^-/- mice have been described previously (Hebert and McConnell, 2000; Mishina et al., 2002). Control embryos were Foxg1^Cre/+;Bmpr1a^fx/+;Bmpr1b^-/+.

RNA in situ hybridization, TUNEL and immunolabeling were carried out on 10-12 μm cryosections from fresh frozen embryos. For RNA situ analysis,sections were hybridized with ³⁵S-labeled RNA probes as previously described (Frantz et al.,1994). Plasmids were used to make probes: Bmp4(Jones et al., 1991), Fgf8 (Crossley and Martin,1995), Foxg1 (Tao and Lai, 1992), Lhx2(Rodriguez-Esteban et al.,1998), Msx1 (A. W. Helms), Nkx2.1(Shimamura et al., 1995), Pax6 (Walther and Gruss,1991), Wnts (Grove et al.,1998), Gsh2 (Gsx2 - Mouse Genome Informatics)(Hsieh-Li et al., 1995), Sfrp2 (Rattner et al.,1997), Ttr (gift of W. S. Blaner, Columbia University,NY, USA), Rfx4-v3 (Blackshear et al., 2003), Zic2 (gift of M. Hayhurst, Stanford University, CA, USA), Actr1 (ATTC, Image Clone 5336738). TUNEL analysis was performed according to the manufacturer's specifications (Roche,catalog #2156792). Immunolabeling for phosphohistone H3 was performed with a polyclonal (Upstate, 1:200) and a Texas Red-conjugated secondary antibody. Cell counts and data analysis were performed as previously described(Gutin et al., 2006).

To test the role of BMP signaling in telencephalon development, mouse embryos lacking both Bmpr1a and Bmpr1b were generated using the cross illustrated in Fig. 1A. Double-mutant embryos were obtained until E11.5 without signs of necrosis. However, at this age, the expected ratio of double mutants(6.25%) was not obtained (3.5% of 202 embryos), indicating some lethality. The dorsomedial area of the E11.5 double-mutant telencephalon appeared rounded and less defined, whereas a midline remained visible rostrally. Severe reduction of the eyes was also observed (Fig. 1C). These phenotypes are 100% penetrant with essentially invariable severity. RNA in situ hybridization analysis for the deleted segment of Bmpr1a, as previously performed(Hébert et al., 2002),revealed that expression begins to disappear at E8.5 (10 somites) and is absent by E9.5 in the telencephalon, whereas expression is unaffected in other tissues, such as the head mesenchyme, foregut and heart (see Fig. S1 in the supplementary material).

At E10.5 and E11.5, although a midline still formed rostrally in the mutant, it did not show the characteristic thinning observed in controls(Fig. 1D-E′,H,I). More posteriorly in controls, the choroid plaque was identifiable as a thin neuroepithelium protruding inward from the dorsal midline creating the medial walls that separate the ventricles, whereas in the double mutant this thin inward protrusion failed to form, resulting in HPE(Fig. 1F-G′,J,K). The worsening phenotype at E11.5 indicates that the phenotype observed at E10.5 is not due to a delay in midline development. Even more posteriorly, the diencephalon of the mutant was rounded and displaced, presumably owing to the lack of an invaginated dorsal telencephalic midline in which to intercalate(data not shown). These results indicate that disrupting BMP signaling prevents the formation of a morphologically defined dorsal midline and induces HPE.

The telencephalic dorsal midline is composed of the choroid plexus and the cortical hem. The choroid plexus is marked at E11.5 and onward by the expression of Ttr. In the double mutant, no Ttr-positive cells were detected anywhere along the rostrocaudal axis of the telencephalic dorsal midline (Fig. 2A,B). From E10/E10.5 onward, Wnt2b and Wnt3a were strongly expressed in the cortical hem (Grove et al., 1998; Lee et al.,2000). In the double mutant at E10.5 and E11.5, expression of these genes was greatly diminished or absent throughout the rostrocaudal axis(Fig. 2D and data not shown). Moreover, the choroid plaque and cortical hem normally express Msx1at E9.5 and 10.5. In the mutant, however, Msx1 expression was lost(Fig. 2F and see Fig. S4 in the supplementary material). These results demonstrate that BMP signaling is required for formation of the choroid plexus and cortical hem.

Other markers that are also expressed in the medial area of the dorsal telencephalon were, however, still readily detected in Bmpr1a;Bmpr1b mutants. Bmp4 and Wnt8b,whose expression at E10.5 normally extends beyond the choroid plaque and cortical hem into the adjacent neuroepithelium, were still expressed in the double mutant, although the length of the expression domain appeared diminished, presumably owing to the loss of midline cell types(Fig. 2G,H and data not shown). Interestingly, Zic2, a gene implicated in forms of human and mouse HPE in which the dorsal midline is lost(Nagai et al., 2000), was expressed in double mutants (Fig. 2I,J), demonstrating that BMP signaling is not required for the maintenance of Zic2 expression.

From E9.5 onward, the midline is marked by high levels of apoptosis. In control E10.5 brains, TUNEL labeling was clearly visible at the dorsal midline, whereas in the double mutant no midline apoptosis was observed(Fig. 2K,L). This demonstrates that BMPR signaling is required in vivo for midline apoptosis in the telencephalon. Surprisingly, the double mutant also exhibited a loss of apoptosis in the rostroventral area of the midline. One potential explanation is that the strong dorsal expression of certain Bmp genes, such as Bmp4 and Bmp5, extends far enough through the rostral midline to act ventrally at E10.5 (Fig. 2G, arrow, and data not shown).

Previous reports have suggested that BMPs inhibit the proliferation of neural progenitors in vitro (Furuta et al., 1997; Mehler et al.,2000). Bmpr1a and Bmpr1b have also been hypothesized to have opposing functions, with Bmpr1a promoting and Bmpr1b inhibiting telencephalic precursor cell proliferation(Panchision et al., 2001). In the E10.5 Bmpr1a;Bmpr1b double mutant, no significant changes in the percentage of mitotic (phosphohistone H3-positive) cells were observed in medial or lateral areas (for pooled areas: control,5.6±1.5% s.d.; mutant, 5.9±1.2%; P=0.6), suggesting that at least at this age, BMP signaling is not required to regulate proliferation. Overall, the phenotypes observed demonstrate functional overlap between Bmpr1a and Bmpr1b in dorsal midline development.

Overexpression of a constitutively active BMPR1A transforms cortical precursors into choroid plexus cells(Panchision et al., 2001),suggesting that BMP signaling has the potential to induce the choroid plexus cell fate. Deletion of Bmpr1a, although demonstrating a requirement for BMP signaling in choroid plexus development, failed to distinguish between a requirement to induce precursor cells to adopt a choroid plaque fate or to promote the differentiation of already specified precursors(Hebert et al., 2002). However, in the Bmpr1a;Bmpr1b double mutant, expression of Msx1, a marker for both choroid plaque and cortical hem precursors,was absent (Fig. 2E,F),consistent with a failure to induce these cell types. The alternative possibilities that BMP signaling is required to maintain proliferation or survival of specified precursors are less likely because no change in these processes was observed in the double mutant.

Compelling evidence suggests that BMPs pattern the cortical neuroepithelium(Monuki et al., 2001; Shimogori et al., 2004; Hirabayashi et al., 2004; Crossley et al., 2001; Ohkubo et al., 2002; Cheng et al., 2006; Theil et al., 2002). To address this hypothesis, we analyzed E10.5 and E11.5 double mutants for the expression domains of transcription factors that specify cortical fates: Lhx2, Foxg1, Emx2 and Pax6. Surprisingly, the expression of these factors, normally present in cortical precursors and excluded or greatly reduced in the dorsal midline, was not affected in the Bmpr1a;Bmpr1b mutant(Fig. 3A-D and data not shown). This suggests either that BMP signaling does not regulate the expression of these genes or that there is another signal, possibly Wnt8b (see above) (Gunhaga et al., 2003),that compensates. In embryos in which the dorsal midline is ablated (and thus a primary source of both BMPs and WNTs is removed), the expression of Emx2, Lhx2, Pax6 and Ngn2 (Neurog2) is also maintained; although for Emx2 and Lhx2, their normal gradients of expression are flattened(Cheng et al., 2006). Subtle changes in the slopes of expression gradients might not be detected with the RNA in situ approach used here. Expression of the v3 variant of Rfx4, which marks the dorsal telencephalon and is essential for medial cortex and hem development(Blackshear et al., 2003), also appeared unperturbed in the double mutant (see Fig. S2 in the supplementary material).

It is also possible that BMP signaling through a more distantly related type I receptor or prior to recombination and loss of Bmpr1a (i.e. earlier than ∼E8.5) is sufficient to convey grossly normal patterning to cortical regions. Actr1, which encodes a type I receptor with affinity for BMP2, 4, 6 and 7 (ten Dijke et al., 2003; Derynck and Zhang, 2003), was not expressed in the telencephalon until after E10.5 (see Fig. S3 in the supplementary material and data not shown). Nevertheless, the results presented here demonstrate that Bmpr1a and Bmpr1b are not required for maintaining the domains of expression of cortical transcription factors.

In most cases of HPE caused by a mutation in a known gene, ventral telencephalic development is greatly impaired in mice and humans (e.g. Shh mutants) (Muenke and Beachy,2000; Hayhurst and McConnell,2003; Chiang et al.,1996). However, expression of Nkx2.1, Gli1 and Fgf8, which depends on SHH signaling(Ohkubo et al., 2002; Pabst et al., 2000; Park et al., 2000), and of Shh, Gsh2 and Sfrp2 was maintained in the E10.5 Bmpr1a;Bmpr1b double mutant, although Fgf8expression was lost in the rostral midline(Fig. 3E-L and data not shown). These results demonstrate that ventral cell specification is preserved in the Bmpr1a;Bmpr1b mutant, unlike in other mouse models for HPE(Muenke and Beachy, 2000; Hayhurst and McConnell,2003).

Although some dorsal midline features have been reported in the Shh mutant (Ohkubo et al.,2002), a more detailed examination of the dorsal midline in these mutants is lacking. To examine the dorsal midline in Shh mutants,expression of Msx1, Bmp5, Bmp4 and Foxg1 was analyzed at E10.5. Msx1, Bmp5 and Bmp4 were still expressed in the dorsal midline, whereas Foxg1 expression remained excluded(Fig. 4C-H and data not shown). Moreover, apoptosis was maintained (Fig. 4A,B). This indicates that the initial formation of the dorsal midline occurs, unlike the HPE phenotype observed in the Bmpr1a;Brmp1b mutant. Moreover, the lack of hemispheric separation dorsally in the Shh mutant at E10.5 could at least in part be due to a failure of the expansion of lateral areas, as compared with control embryos (Fig. 4).

The present study demonstrates that mutations in the BMP pathway can lead to HPE with a lack of the dorsal midline. In contrast to the majority of HPE cases described in humans and mice, ventral patterning in Bmpr1a;Bmpr1b mutants is not affected and thus mimics the MIH subclass of HPE. These results support the existence of at least two developmental mechanisms leading to HPE: one involving loss of ventral medial cell types without a loss of early dorsal midline features, as in mutants in which the SHH pathway is disrupted; and a second involving a lack of dorsal midline development without impaired cell specification in the ventral telencephalon, as in the Bmpr1a;Bmpr1b mutant(Fig. 4I).

Recent independent findings confirm the presence of early dorsal midline features in the Shh^-/- mutant mice(Rash and Grove, 2007).

Supplementary material for this article is available at http://dev.biologists.org/cgi/content/full/134/21/3789/DC1

This work was supported in part by the Sinsheimer Foundation, the James S. McDonnell Foundation, NIH MH075779 and MH70596 (to J.M.H.).