The cortical hem was first described as a potential signaling center at the telencephalic midline because of an enriched expression of multiple members of the Wnt and Bmp families of morphogens, and its position at the border between the presumptive cortex and the choroid plexus. There is now definitive evidence that the cortical hem is an organizing center in the telencephalon, and that it instructs the formation of the hippocampus. In this review, we present an analysis of the molecular and cellular events that lead to the formation of the cortical hem, and define its position and extent in the telencephalon. This directly controls the positioning of the hippocampus within the telencephalon. We conclude with a summary of the current understanding of the role of the hem as the hippocampal organizer.
During the development of the telencephalon, the dorsal midline region invaginates to give rise to paired telencephalic vesicles. Concomitantly, the medial tissue becomes reorganized to generate two structures, the cortical hem and the choroid plexus epithelium (CPe). The CPe is a simple cuboidal epithelial sheet containing the secretory vasculature that produces the cerebrospinal fluid, whereas the cortical hem becomes a highly restricted source of several morphogens of the Bmp and Wnt families (Furuta et al. 1997; Grove et al. 1998). The cortical hem appears similar to and contiguous with the pseudostratified cortical neuroepithelium, expresses dorsal telencephalic proneural genes such as Ngn 1 and 2 (Sommer et al. 1996; Imayoshi et al. 2008), and generates Cajal–Retzius cells which are the earliest born neurons of the cortex (Takiguchi-Hayashi et al. 2004;Yoshida et al. 2006). Functionally, however, the cortical hem is distinct from the rest of the cortical neuroepithelium and in fact functions as a classical developmental organizer that can pattern the naïve dorsal telencephalon into distinct functional domains. The mechanisms that regulate the formation of the cortical hem itself are therefore of great interest. We present different approaches which have each contributed to the current understanding of how the cortical hem forms, and how it is localized to its restricted position at the caudomedial edge of the telencephalic vesicle.
Positioning the Hem
A signaling center regulates not only the induction, but also the appropriate structural organization of target tissue. It is therefore critical that the signaling center itself be correctly positioned in the brain with respect to the target tissue it acts on. It appears that the same molecular gradients that act to establish and regionalize the early neural tube and determine the position of the telencephalon also regulate the positions of the signaling centers that are located within it (Hebert 2005).
Many studies have revealed key insights into the mechanisms that confine the hem as a whole or specific morphogens within it, to the caudomedial domain of the telencephalon. Here we describe four models for hem positioning, which are by no means mutually exclusive. In fact, there is considerable overlap between the inductive and cross-regulatory mechanisms that are central to each of the models described below. However, each model approaches this problem using a unique molecular or cellular event to probe the regulatory interactions leading to hem specification. We have discussed each of these models separately in some detail, and attempted to combine some insights into a schematic which summarizes the current understanding in the field.
Model 1: The Hem Arises from the Telencephalic Roof Plate in a 2-Stage Process
Interactions which define the position and character of the hem begin prior to neural tube formation, when interaction with adjacent non-neural ectoderm causes Bmp expression in the lateral edge of the open neural plate (Liem et al. 1995). When the plate folds into the neural tube, this lateral edge now becomes the roof plate, extending along the entire length of the neuraxis. At the rostral end, the roof plate region invaginates and begins to reorganize to form the CPe and the cortical hem.
Currle et al. (2005) used lineage tracing to determine where the hem cells originate, using a Gdf7 driver that partially labels the telencephalic roof plate. This study reported that some cells of the choroid plexus and the hem appeared to arise from, or share a common lineage with, roof plate cells. These experiments also uncovered an interesting dichotomy within each of these midline derivatives. Only cells of the anterior choroid plexus appeared to belong to the Gdf7 lineage. However, ablation of the Gdf7 component of the roof plate eliminated the entire choroid plexus. This suggested that the posterior portion was induced by cell nonautonomous inductive signals, which could arise either from the anterior portion of the choroid plexus or from the roof plate. Similarly, a few cells of the Gdf7 lineage were also found to be localized to the anterior cortical hem. Ablation of the Gdf7 component of the roof plate also resulted in loss of the cortical hem, based upon which the authors suggested that the caudal portion of the hem arises due to inductive effects of its anterior portion. However, roof plate ablation causes significant midline patterning defects and perturbs molecular gradients in the medial telencephalon (Cheng et al. 2006). These disruptions could be the primary cause of the loss of the hem upon roof plate ablation. These mechanisms are discussed in subsequent sections.
An intriguing twist to this model comes from studies that map the Wnt3a lineage (Yoshida et al. 2006; Louvi et al. 2007), demonstrating that it includes the hem as well as the choroid plexus. Wnt3a expression in the cortical hem appears as early as E9.5, in bilateral domains flanking a Wnt3a negative roof plate (Parr et al. 1993; Louvi et al. 2007). The lineage relationship of the cortical hem with the telencephalic roof plate remains unclear. But because the appearance of the cortical hem precedes the appearance of the choroid plexus (identified by TTR expression, from E11.5), it is likely that the hem produces the choroid plexus.
Model II: Interactions between Multiple Signaling Centers Define the Position of the Hem
In this model, an interaction of molecules from multiple telencephalic signaling centers is responsible for hem induction and positioning (Shimogori et al. 2004; Storm et al. 2006). The telencephalon is induced at the anterior end of the neural tube by signals from the anterior neural ridge (ANR). Ablation of the ANR in mice prevents the induction of FoxG1 whose expression defines the future telencephalic vesicles (Shimamura and Rubenstein 1997). Studies in zebrafish indicate that secreted Wnt inhibitors from the cells of the anterior neural border, a structure analogous to the mouse ANR repress posterior fates and thus “allow” the telencephalon to form (Houart et al. 2002). The cells of the mouse ANR also express members of the fibroblast growth factor family (Fgf) including Fgf8, possibly as a consequence of Wnt inhibition (Houart et al. 2002; Hebert, 2005). Fgf8 can itself induce FoxG1 expression in the anterior neural plate and is therefore a positive inducer of telencephalic character. In addition, Fgf8 regulates midline fate in the telencephalon by inducing transcription factors Lhx5, Zic2, Vax1, and repressing Lhx2 (Okada et al. 2008).
The Fgfs not only regulate the early induction of the telencephalon but also act to morphogenetically pattern this neuroepithelium into functional areas (Fukuchi-Shimogori and Grove 2001; Garel et al. 2003). Thus, the ANR takes on the identity of a rostral telencephalic signaling center in addition its role of inducing the telencephalon itself. Fgfs from the rostral signaling center and Bmps from the roof plate appear to cross regulate each other. In chick embryos, Bmp beads placed near the Fgf8 source limit its extent (Ohkubo et al. 2002). Studies using hypomorphic alleles of Fgf8 have shown that Fgf8 also regulates Bmp expression at the dorsal midline in a dose dependent manner (Shimogori et al. 2004; Storm et al. 2006). This confines the Bmp expressing cells to a more caudal location away from the Fgf-enriched rostral signaling center. Similarly, Fgfs can also repress the expression of Wnt genes in the dorsal midline region and confine them to a caudal domain of expression (Shimogori et al. 2004). These interactions serve to define the anterior limit of the hem in the telencephalon.
In summary, a finely choreographed sequence of mutual repression (Fig. 1) helps to localize hem-specific signaling molecules to a posterior domain in the medial telencephalon.
Model III: Instructive Transcription Factor Gradients Define the Hem Domain
Kimura et al. (2005) focused on the graded expression of transcription factors known to be important in telencephalic development, particularly the very early expression patterns of Emx2 and Pax6 in the telencephalon. They analyzed embryos mutant in either or both of these genes, in the background of heterozygous Otx2 mutants. Compound mutants of Emx2 and Pax6 are unable to develop the entire caudomedial telencephalic domain including the archipallium, the cortical hem, and the choroid plexus (Kimura et al. 2005), although the roof plate is present (Muzio et al. 2002; Kimura et al. 2005). A similar result is seen in embryos that are null for Emx2, and lacking one copy of Otx2. Because the prospective archipallium, hem, and choroid plexus are disrupted together, their data support the interpretation of the caudomedial telencephalic domain as a unique developmental compartment.
The Emx2 mutant itself shows a specific decrease in proliferation in the caudomedial domain (Muzio et al. 2005), which is consistent with a shrinkage of the hippocampal fields (Tole et al. 2000). Emx2 regulates the expansion of this domain by regulating the effectors of canonical Wnt signaling (Muzio et al. 2005). Interestingly, Emx2 also plays an indirect role in positioning the source of Wnt signals, the hem, by restricting the Fgf expression domain to the rostral midline. This contains the inhibitory effects of Fgf on Wnt expression, and thus participates in localizing the cortical hem to its caudomedial position (Shimogori et al. 2004). Thus, Emx2 may first act indirectly to permit formation of the cortical hem, and then function within the prospective archipallial domain to regulate the response to hem (Wnt) signaling. The hem is necessary (Yoshida et al. 2006) and sufficient (Mangale et al. 2008) to induce hippocampal fields in the cortical primordium. The action of Emx2 in positioning the hem may in fact explain its role in controlling specification of the archipallial domain, uniting the hem and the archipallium within a common compartment as proposed by Kimura et al (2005).
Intriguingly, the Emx2 enhancer can be activated by both Tcf and Smad transcription factors (Theil et al. 2002). It is tempting to speculate that early, graded Wnt and Bmp signals from the roof plate themselves act to set up the Emx2 gradient that will later help establish the hem domain.
The Otx and Emx genes were first identified as a nested series of development control molecules (Simeone et al. 1992), analogous to the nested hox genes which pattern the hindbrain rhombomeres (reviewed in Kiecker and Lumsden 2005). New understanding of the specification of the caudomedial domain fits well with and extends this model of molecular parcellation of the telencephalon.
Model IV: A “Hem-Competent Zone” along the Mediolateral Axis
The models described above primarily describe the mechanisms which act to position the hem along the rostro-caudal axis of telencephalon. An analysis of telencephalic patterning along the mediolateral axis gives rise to a fourth model, described below.
Boundary Formation along the Mediolateral Axis of the Telencephalon
The cortical hem is bounded by the CPe on the medial side and the cortical neuroepithelium on the lateral side. A simple explanation for the medial location of the cortical hem is that it arises from invaginated midline tissue (Monuki et al. 2001; Currle et al. 2005). Yet, it is now clear that the extent of the cortical hem is controlled by multiple interactions of signaling molecules and transcription factors. These mechanisms act to define the cortical hem as a unique tissue with distinct cytological and gene expression boundaries with the choroid plexus and cerebral cortex.
The Hem–Choroid Plexus Boundary
The cortical hem and choroid plexus are easily distinguishable by E11.5–E12.5 in the mouse due to the distinct differences in their cytoarchitecture, as well as molecular signatures. Prior to this stage, the hem displays Wnt3a expression from E9.5 (Parr et al. 1993; Louvi et al. 2007), in bilateral domains flanking the roof plate (Louvi et al. 2007). How is the lineage relationship between the hem and the choroid plexus (Louvi et al. 2007) regulated to produce a distinct boundary between these two structures? Bmp signals have been implicated in the development of the choroid plexus (Furuta et al. 1997). Animals with defective Bmp signaling in the dorsal midline fail to form the choroid plexus (Hebert et al. 2002), but significantly, these animals have an apparently normal cortical hem and other medial structures. This indicates that the specification of the choroid plexus uniquely requires BMP signaling. Mutants which lack the cortical hem also lack the choroid plexus (Grove et al. 1998; Lee et al. 2000; Yoshida et al 2006), reinforcing the idea that the choroid plexus is derived from the hem.
An elegant study by Imayoshi et al. (2008), has examined molecular events regulating these dorsal midline structures, and revealed interesting parallels with mechanisms that act to define neural/non-neural cell fate specification in other systems (Ross et al. 2003). Imayoshi et al. (2008) report that the dorsal midline at E9.5 appears to function as a bipotential field capable of producing Cajal–Retzius neurons or choroid plexus cells, subject to the differential action of the Hes and Neurogenin genes (Ngn1 and 2) of the bHLH family. Hes gene expression is enriched in the Cpe, whereas Ngn1 and 2 expression excludes the CPe, but is present in the hem and in the adjacent cortical neuroepithelium. High levels of Hes gene expression could thus potentially inhibit the expression of the Neurogenins in the prospective CPe. Supporting this idea, compound mutants for Hes1, Hes3, and Hes5 display an expanded expression of Ngn1 and 2 in the prospective CPe region. Likewise, misexpression of Ngn2 into the prospective CPe region is sufficient to prevent the formation of the choroid plexus and promotes Cajal–Retzius fate (Imayoshi et al. 2008). Intriguigingly, this study suggests that the E9.5 dorsal midline tissue can produce either of the hem derivatives—Cajal–Retzius cells or choroid plexus—without transiting through an intermediate hem fate. However, hem-specific markers were not examined, raising the possibility that dorsal midline cells may transiently acquire hem identity before differentiating into hem derivatives. Alternatively, the Hes and Ngn gene perturbations may cause dorsal midline cells to bypass this step, directly producing CPe or Cajal–Retzius cells.
An important aspect of this study is that it integrates a role for Bmp signals from the roof plate in CPe development (Imayoshi et al. 2008). High levels of Bmp signaling can enhance activation of Hes genes in neuroepithelial cells (Takizawa et al. 2003). Complementarily, expression of Bmp targets is abrogated in the compound Hes mutants. Together, these studies suggest that an autoregulatory loop involving Bmp and Hes genes may control CPe differentiation (Fig. 1). The location and extent of the CPe may be determined by the portion of the dorsal midline in which this loop is active. The adjacent tissue would continue to maintain Neurogenin expression and take on hem fate. Cross suppression between Hes and Ngn genes would lead to the formation of a boundary between these two structures, which is the final step in the differentiation of two unique tissues adjacent to each other.
The Hem–Cortex Boundary
The cortical hem is defined not only by the expression of selected signaling molecules and transcription factors but also by the marked exclusion of this structure in the expression patterns of other molecules. Prominent among these are the transcription factors FoxG1 and Lhx2 (Bulchand et al. 2001; Monuki et al. 2001; Hanashima et al. 2007). At early ages both are expressed throughout the cortical neuroepithelium and eventually their expression delineates a sharp boundary between the cortical tissue and the hem. Further, in mice mutant in these genes, the cortical hem expands, suggesting their role in actively confining the lateral extent of the hem. An important difference between these two molecules is that in FoxG1 mutants both the hem and the medial part of the cortical neuroepithelium expand (Muzio and Mallamaci 2005); in Lhx2 mutants however the cortical hem expands at the expense of the cortex (Bulchand et al. 2001; Monuki et al. 2001). This suggests a more specific role for the LIM homeodomain transcription factor Lhx2 in defining the hem–cortex boundary, whereas FoxG1 probably regulates the broad medial telencephalic domain.
Further evidence for a role of Lhx2 in defining the hem–cortex boundary is derived from studies on the brains of embryos that are mosaic for the Lhx2 mutation (Mangale et al. 2008). These mosaic animals were generated in two different ways. Lhx2 null embryonic stem cells were aggregated with wild-type morulae to produce chimeric embryos. Alternatively, low-dose tamoxifen injections in CreER-conditional Lhx2 knockouts produced mosaics of wild-type and Lhx2 null cells. Both types of mosaic embryos showed distinct homophilic affinity within control and mutant cell groups, leading to sorting of Lhx2-on and Lhx2-off cells into clusters. How might Lhx2-off cells be created in vivo? Bmp activity from the roof plate regulates Lhx2 expression in a bi-modal manner, such that Bmp signaling represses Lhx2 at high concentrations but activates it at lower concentrations (Monuki et al. 2001). Thus, high Bmp signaling from the roof plate creates the domain of Lhx2-off cells that will eventually become the hem. Cells located in the medial neuroepithelium adjacent to the hem experience lower levels of Bmp signaling and upregulate Lhx2 expression. The affinity difference between Lhx2-on and Lhx2-off cells causes them to sort separately and gives rise to the sharp boundary between the hem and the cortical neuroepithelium. The absence of Lhx2 is sufficient to induce cells to take on a hem fate within the caudomedial telencephalic domain (Mangale et al. 2008). Together these studies define a mechanism by which the hem–cortex boundary is generated (Chen and Price 2008; Grove 2008; Mangale et al. 2008).
A Cryptic Hem–antihem Boundary Revealed in the Lhx2 Mutant
In the Lhx2 standard knockout, the expanded hem is complemented by an expanded putative lateral signaling center, the antihem (Assimacopoulos et al. 2003), which together comprise the entire dorsal telencephalon (Mangale et al. 2008). What mechanisms drive prospective cortical cells to take on either of these two distinct alternative fates in the absence of Lhx2? Both, in the Lhx2 standard knockout, as well as in chimeric brains containing Lhx2 null patches, medially positioned Lhx2 null cells become hem, whereas more laterally localized cells instead take on the fate of the antihem (Fig. 2; Mangale et al. 2008). These findings raise the interesting concept of a “hem-competent zone” within the dorsal telencephalon, within which loss of Lhx2 leads to hem fate.
Chimeric embryos present a system in which the extent of this zone may be analyzed. The expression of Lmo2, a LIM-only protein expressed in a graded manner in the medial neuroepithelium (Bulchand et al. 2003), appears to mark the limit of the hem competent zone in chimeras, extending up to the most lateral patch of ectopic hem (Fig. 2; Mangale et al. 2008). What is the nature of the molecular cues that define this zone? One obvious possibility is that the extent of this zone is determined by early Bmp and Wnt gradients from the roof plate. However, the ectopic hem patches which themselves express Wnts and Bmps would then reposition these gradients and extend the hem competent zone farther into the lateral neuroepithelium. This does not seem to occur and therefore questions the role of these cues in determining the hem competent zone. Another possibility is based on the presence of Emx2 very early in the neuroepithelium, in high medial, low lateral gradient, which may define the hem competent zone. A background of high Emx2 may be required for hem to form in response to Lhx2 downregulation or loss of function. Pax6, expressed in an opposite gradient, and required for antihem fate (Assimacopoulos et al. 2003), may function in a complementary manner, suppressing the hem competent zone and promoting an antihem competent zone. Thus graded expression of transcription factors may regulate the cryptic boundary between hem and antihem forming ability in the cortical primordium.
The Cortical Hem as a Patterning Center
A major event in the mediolateral patterning of the dorsal telencephalon is the determination of medial and dorsal pallial identity; which will in turn determine the development of the hippocampus and the neocortex, respectively. The hippocampus develops immediately adjacent to the hem, motivating the hypothesis that the hem functions as a source of patterning cues for the hippocampus (Grove et al. 1998). Several studies that analyzed mutants defective in Wnt signaling supported this hypothesis (Lee et al. 2000; Galceran et al. 2000; Zhou et al. 2004). However, these studies were unable to separate the actual specification of hippocampal cell fate from a role for hem signals in regulating the expansion of the precursor population (Li and Pleasure 2005). A clear role was established for canonical Wnt signals in regulating the expansion of the hippocampal progenitors (Muzio et al. 2005). Supporting this role, reduced clusters of hippocampal precursors were detected in these mutants (Galceran et al. 2000; Lee et al. 2000; Zhou et al. 2004).
An important insight about Wnt-mediated hippocampal specification came from a study that activated the canonical Wnt signaling effector Lef1 in the lateral telencephalon, which resulted in upregulation of a subset of hippocampal markers in prospective neocortical territory (Machon et al. 2007). Definitive evidence that the hem induces hippocampal fate came from an analysis of the effects of ectopic hems in Lhx2 null chimeric embryos (Mangale et al. 2008). Multiple hippocampal fields developed in appropriate spatial order adjacent to each patch of ectopic hem tissue, demonstrating that the hem is sufficient to induce detailed hippocampal field specification, and in fact acts as a hippocampal organizer (Grove 2008; Mangale et al. 2008).
Together, these findings reveal a cascade of intertwined, cross-regulatory events, beginning from signaling at the lateral edges of the neural plate and the ANR, which lead to the localizing of an important telencephalic signaling center, the cortical hem. A puzzling feature of the current literature is that to date no signaling molecule or transcription factor has been discovered to play a positive role in specifying the hem. Foxg1 and Lhx2 each repress hem fate, because mice lacking either of these transcription factors display an expanded cortical hem (Muzio and Mallamaci 2005; Mangale et al. 2008). The Foxg1 mutant also contains hippocampal tissue (Muzio and Mallamaci 2005), suggesting that Foxg1 activity represses a broader caudomedial telencephalic compartment not limited to the cortical hem, similar to the domain missing in the Emx2/Pax6 double mutant, or the Emx2 null/Otx2 heterozygous mutant (Kimura et al. 2005).
Loss of function of Lhx2, or overexpression of Bmps which repress Lhx2, each expands the extent of the hem (Monuki et al. 2001). Fgf overexpression, which shrinks the hem (Shimogori et al. 2004), is likely to act via the Bmp-Lhx2 pathway, because Fgfs repress Bmps. The absence of Emx2 increases the expression of noggin and expands the Fgf8 expression zone (Shimogori et al. 2004). Because each of these proteins can decrease Bmp expression, loss of function of Emx2 potentially also regulates the hem via the Bmp-Lhx2 pathway.
A cell autonomous factor that instructs hem fate remains elusive. Emx2 is an attractive candidate for a permissive or instructive early role defining the hem competent zone, but as yet concrete evidence for such a function is lacking. Ngn2 is another potential candidate for specifying hem identity, because misexpression of Ngn2 into the prospective CPe is sufficient to produce Cajal–Retzius cells (Imayoshi et al. 2008), suggesting a transdifferentiation into hem fate. However, Ngn2 null mice have normal hem and hippocampal phenotypes, and the Ngn1,2 double knockout appears to have some hippocampal tissue, suggesting the presence of the hem, although this issue has not been specifically addressed (Schuurmans et al. 2004). In summary, known early acting instructive mechanisms all promote nonhem fates, those of the choroid plexus and the cortex respectively. Hem fate appears to be “released” by the removal of hem suppressing mechanisms, such as those controlled by Lhx2. An as yet unidentified downstream target repressed by Lhx2 may be sufficient to instruct hem identity. Together, the cell autonomous and cell nonautonomous regulatory mechanisms that create the hem at the medial edge of the cortical neuroepithelium eventually determine the caudomedial position of the hippocampus in the telencephalon.
Work in S. Tole's lab supported by the following: Swarnajayanti Fellowship (Department of Science and Technology, Government of India); a grant from the Department of Biotechnology, Government of India; and a grant from the Department of Science and Technology, Government of India. L.S. was funded by a Kanwal Rekhi Career Development Award from the TIFR Endowment Fund.
Conflict of Interest: None declared.