It is well established that canonical Wnt signaling plays a central role in the formation of the hippocampus and the dentate gyrus (DG; Grove et al.,1998; Lee et al.,2000; Galceran et al.,2000; Tole and Grove,2001; Machon et al.,2003; Li and Pleasure,2005). Recently, it has been reported that the canonical Wnt pathway is also important for cellular proliferation in the ventricular zone and for fate determination in the cortex (Chenn and Walsh,2002; Backman et al.,2005; Zhou et al.,2006; Woodhead et al.,2006; Machon et al.,2007). Furthermore, it was recently shown that neurogenesis in the cortex is directly correlated with a gradual decrease of canonical Wnt signaling (Machon et al.,2007; Faedo et al.,2007).
During embryogenesis, the hippocampus develops from the medial wall of the dorsal telencephalon, the hippocampal primordium. The ventral margin of the medial wall, the hem, serves as an important signaling centre which expresses several Wnt genes, including Wnt3a, -2a, -5a, and -7a (Grove et al.,1998; Lee et al.,2000). Among these, Wnt3a has a crucial role because a targeted deletion of Wnt3a leads to the absence of the hippocampus and the DG (Lee et al.,2000). Evidence suggests that the predominant role of Wnt-3a during hippocampal and DG development is mediated through the canonical Wnt pathway, in particular since a dominant-negative form of Lef1 or a conditional inactivation of β-catenin yielded a similar phenotype (Galceran et al.,2000; Machon et al.,2003).
The dentate neuroepithelium is located dorsal to the hem, in the ventral part of the hippocampal primordium, and generates the progenitor cells of the DG that expresses Prox1 as early as at E12.5 (Oliver et al.,1993; Zhou et al.,2004). Later, from embryonic day (E) 14.5 onward, the dentate neuroepithelium is transformed into the premigratory primary neuroepithelium (pne) of the hippocampal ventricular zone that further forms the secondary matrix (sm) near the dentate notch (Altman and Bayer,1990a,b). Thus, the precursor pool expands in the sm and the first granule neurons migrate radially from the sm through the dentate gyrus migration (dgm) to form the primordial granule cell layer of the DG. During this process, a neurogenic gradient is formed such that the earliest cells that went through the dgm form the external (upper) dentate blade (DGe) at around E17.5, while later cells settle at the tip of the internal (lower) dentate blade (DGi) that starts to be visible from around the day of birth (P0). The majority of adult granule neurons are born postnatally, substantially during the first week, in the tertiary matrix of the hilus, and in the proliferative subgranular zone (SGZ) of the dentate blades located beneath the granular cell layer (Altman and Bayer,1990a; Pleasure et al.,2000; Muramatsu et al.,2007).
In Lef1−/− and LRP6−/− double mutants, the number of proliferating cells measured by bromodeoxyuridine (BrdU) incorporation was significantly lower in embryonic premigratory, migrating and postmigratory progenitor pools (Zhou et al.,2004; Lie et al.,2005). This resulted in an almost complete lack of dentate granule neurons in the DG, which again supports the importance of the canonical Wnt signaling in the DG development. In the adult brain, DG precursor cells reside in the SGZ where they receive instructive cues from the surrounding microenvironment to promote neurogenesis. One of these signals is Wnt3 which is produced by astrocytes present in the stem cell niche in the DG (Lie et al.,2005).
Dickkopf1 (Dkk1) is a secreted protein that functions as a negative regulator of canonical Wnt signaling (Glinka et al.,1998). It binds to the Wnt co-receptors Lrp5 and Lrp6 (Mao et al.,2001), or to the Kremen receptor, which leads to down-regulation of Lrp6 on the cell surface (Mao et al.,2002). At early embryonic stages, Dkk1 is expressed in the neural tissue where it participates in the establishment of a gradient of Wnt activity, which is crucial for cellular specification in the neural tube along its rostrocaudal axis (Mukhopadhyay et al.,2001). Later at E15.5, expression of Dkk1 is only seen in a small population of cells in the anterior hippocampal ventricular zone near the fimbria, before it vanishes from the telencephalon (Diep et al.,2004).
Although the role of Wnt signaling in the development of the hippocampus and the DG has been addressed, several questions remain unanswered. For example, it is unclear when the progenitor pool requires active Wnt pathway for its expansion. Furthermore, it is unclear whether canonical Wnt signaling affects differentiation rate, neurogenic division, or both. In this study, we analyzed phenotypic changes in the dorsal telencephalon after a specific down-regulation of the Wnt activity in transgenic D6-Dkk1 mice that ectopically express Dkk1 in cortical and hippocampal progenitors after E11. Thus, the D6-Dkk1 mouse allowed us to study the impact of canonical Wnt signaling in the neurogenic cortex at a stage when the canonical Wnt signaling rapidly declines, and in the neurogenic hippocampus and the premigratory primary neuroepithelium of the DG that is strongly Wnt-positive. We show that down-regulation of canonical Wnt signaling after E11 did not lead to significant alterations in morphology and gene expression in the developing cortex. In contrast, the ectopic expression of Dkk1 in the medial cortical wall impaired development of hippocampal fields and the DG. We found significant reduction of neurogenic division and longer cell cycle of progenitors that resulted in decreased neurogenesis and severely reduced DG.
Expression of Dkk1 in D6-Dkk1 transgenic animals
For area-specific expression of Dkk1 in the dorsal telencephalon, we used the promoter/enhancer D6 derived from the mouse Dach1 gene that is specifically expressed in the developing cortex and hippocampus. D6 activity is first detected at E10.5 in the most dorsal telencephalon from where it spreads to the lateral and medial cortical primordium by E11.5 (Machon et al.,2002) and thus provides a valuable tool for manipulating the levels of canonical Wnt signaling in the dorsal forebrain. D6-Dkk1 transgenic mice were created by pronuclear injections. From seven D6-Dkk1 founders, five showed a phenotype in the hippocampus that resembled knockouts of Wnt3a, Lef1, or β-catenin conditional knockouts, and the one with the most severe phenotype was selected for further analysis.
In situ hybridization was used to confirm expression of the transgenic Dkk1 under the control of the D6 enhancer at various embryonic stages (Fig. 1A–D). The expression of D6-Dkk1 clearly reflected D6 enhancer activity (Machon et al.,2002), and only minor variations in the expression were observed among embryos. As we have previously reported (Diep et al.,2004), endogenous Dkk1 is hardly present in the developing mouse brain after E12.5, and only a minor population of cells with a moderate expression of endogenous Dkk1 was observed in the anterior hippocampal ventricular zone of wild-type animals at E15.5, which means that all expression of Dkk1 in the transgenic animals was derived from the D6-Dkk1 transgene.
As seen in Figure 1A, strong expression of Dkk1 was seen in the cortex and hippocampal primordium of D6-Dkk1 mutants. The medial boundary of the ectopic Dkk1 expression was sharply marked within the cortical hem at E13.5, while laterally the expression gradually decreased and vanished at the lateral margin of the cortical primordium (antihem). Dkk1 presence was particularly strong in the ventricular (VZ) and subventricular zone (SVZ) of the prospective hippocampus. In the germinal zone of DG at stages E15.5 to postnatal day (P) 0 (Fig. 1B–D), the ectopic expression of Dkk1 was strongly detected in the pne, the sm, and in the migratory route until P0. A few scattered cells expressing Dkk1 were also found in the fimbria. No Dkk1 expression was seen in postmigratory cells within the DG. During all neurogenic stages, Dkk1 was expressed in the hippocampal VZ, SVZ, CA1–CA3 layers, as well as in the lateral cortical VZ, SVZ and cortical layers II–VI. Thus the D6-Dkk1 mice provided a precise tool for the analysis of effects of canonical Wnt signaling in premigratory precursor cells of the DG, the hippocampus and the cortex.
Dkk1 Down Regulates Canonical Wnt Signaling in the Developing DG
The capability of Dkk1 to inhibit canonical Wnt signaling in cortex and hippocampus was tested by crossing D6-Dkk1 to BAT-Gal reporter mice (Maretto et al.,2003). In these mice, the β-galactosidase (β-gal) gene is under the control of a promoter containing multiple β-catenin/TCF responsive elements and it, therefore, serves as a sensitive tool for monitoring canonical Wnt responsive cells in vivo. Due to the high stability of the β-gal protein, Wnt activity was visualized by in situ hybridization using an anti–β-gal RNA probe, instead of a standard β-gal enzymatic assay. In BAT-Gal control animals without ectopic Dkk1 expression, β-gal mRNA was detected in a gradient manner at E.14.5, that is, weak in the lateral cortical wall and strong in the hippocampal primordium. In contrast, D6-Dkk1 mutants showed a clear reduction of β-gal–positive cells in cortex and hippocampus (Fig. 1E–E′). At E15.5, a strong expression of β-gal was detected in the wild-type dgm that was lost in the D6-Dkk1 mutants (Fig. 1F,F′). At E17.5 (Fig. 1G–G′), β-gal expression was detectable in the wild-type hippocampal VZ, including the sm emerging from the pne, but the Wnt activity was generally weaker at this stage. In the D6-Dkk1 mutants, however, β-gal staining was dramatically reduced to a very few cells in the sm and dgm. It is noteworthy that the BAT-Gal activity was only weakly detectable in granule neurons of the wild-type DG at the analyzed embryonic stages. In summary, D6-Dkk1 effectively reduced, but did not entirely turn off the canonical Wnt activity in the germinal zones of the prospective hippocampus and DG.
Inhibition of Canonical Wnt Signaling in the Neurogenic Phase Did not Affect Patterning of Cortical Layers in D6-Dkk1 Mice
The developing mouse shows a gradient of canonical Wnt signaling from the early neurogenic phase and onward (Fig. 1E–G). Previously it has been shown that expression of a constitutively active form of β-catenin (Chenn and Walsh,2002) as well as its targeted deletion (Machon et al.,2003) causes substantial cortical changes. This may be explained by the role of β-catenin in mediating canonical Wnt signaling but could also be a consequence of its early role in cell adhesion, which is vital for the development of a proper neuroepithelium (Junghans et al.,2005). Because expression of Dkk1 specifically inhibits canonical Wnt signaling, it allows the two roles of β-catenin to be distinguished.
Possible alterations in cortical development in D6-Dkk1 embryos were analyzed using several markers that identify cortical structures. Sfrp1 is a marker of the VZ of the lateral and medial cortex throughout development of the forebrain, as well as of the cingulate cortex at prenatal stages (Augustine et al.,2001). In situ RNA hybridization of D6-Dkk1 mutants at E13.5 revealed that Sfrp1 expression was present and correctly positioned in the lateral cortex of the mutants although the length of the medial wall (hippocampal primordium) was remarkably shorter and the lateral wall was thicker than in wild-type controls (Fig. 2A–A′). At P0, the size of the subiculum expressing Sfrp1 was moderately smaller, but the thickening of the lateral wall was no longer evident (Fig. 2B–B′). Adherens junctions in the VZ and Cajal Retzius cells in the outer layer are central for maintenance of neuroepithelial character and organization of the cortical architecture. Immunostaining of N-cadherin in adherens junctions or Reelin, which is present in Cajal Retzius cells in cortical layer I, did not reveal significant changes in the cortical scaffolding in D6-Dkk1 mutants (Fig. 2C–C′,D–D′). Furthermore, in situ hybridization with a SCIP riboprobe, marking layers II/III+V (Frantz et al.,1994), and Tbr1 that is present in layers II/III+V and VI (Bulfone et al.,1995; Hevner et al.,2001), showed normal cortical organization in D6-Dkk1 mutant embryos (Fig. 2E–E′,F–F′). This is not surprising because canonical Wnt signaling is naturally down-regulated in this area in a gradual movement to initiate a neurogenic program in the cortex (Machon et al.,2007). The presence of intact adherens junctions shows that Dkk1 does not affect the cytoplasmic role of β-catenin in the cytoskeleton.
Inhibition of Canonical Wnt Signaling Leads to Severe Defects in the Development of the Hippocampus and the DG
Several lines of evidence show a crucial role of canonical Wnt signaling during development of the hippocampus, an area which, in contrast to the cortex, shows a substantial Wnt activity during development. Prevented canonical Wnt signaling in the Wnt3a mutants (Lee et al.,2000), in the Lef1 mutants (Galceran et al.,2000) or in conditional β-catenin mice (Machon et al.,2003) led to substantial proliferation defects in the hippocampus. Accordingly, D6-Dkk1 embryos exhibited a truncated hippocampal primordium as early as at E13.5 (Fig. 2A–A′). At newborn stage, Tbr1 in situ hybridization (a marker for neurons originating from the cortical ventricular zone; Hevner et al.,2001) showed that the hippocampal fields CA1–CA3 were greatly reduced and that granule neurons in the DG were hardly present (Fig. 3A–A′). The presence of specific hippocampal fields was examined by in situ hybridization with riboprobes against Neuropilin-2 (NP2, expressed in CA1–CA3 fields and the DG; Chen et al.,1997), SCIP (expressed in pyramidal neurons of the CA1–CA2 fields; Tole et al.,1997), and KA1 (expressed in CA2–CA3 and DG; Tole et al.,1997). In D6-Dkk1 newborns, a severe reduction in the size of all CA1–CA3 fields was observed (Fig. 3B–B′,C–C′,D–D′). Furthermore, the expression of the homeobox prospero-like protein Prox1, an early and specific granule cell marker in the DG (Liu et al.,2000), exhibited a very strong reduction in size of the DG in mutants (Fig. 3E–E′), a finding that was also confirmed by the Tbr1, NP2, and KA1 in situ hybridizations. While Figure 3E′ represents an average size of the DG in D6-Dkk1 mutants (n = 10), some extreme cases were observed where the DG was completely missing (n = 2, not shown). Furthermore, as expected from the anterior–posterior gradient of D6 activity (van den Bout et al.,2002), the alterations in the mutant hippocampus were more dramatic in posterior than anterior areas (not shown).
The hippocampal neuroepithelium, which is located dorsally from the cortical hem and is marked by expression of Prox1 at E13.5, also serves as the dentate neuroepithelium from which the future dentate granule cells and their precursors arise (Altman and Bayer,1990b; Bagri et al.,2002; Zhou et al.,2004). During embryonic development of the DG, expansion of precursor cells occurs predominantly in the premigratory pne and in the migratory route, the dgm, to the DG. In contrast, the hilus of the DG contains fewer proliferating cells as measured by BrdU incorporation (Zhou et al.,2004). To test which of the germinal zones at various developmental stages can be affected by the D6-Dkk1 activity; Prox1 immunohistochemistry was performed on sections from D6-EGFP mice (Machon et al.,2002). We analyzed co-expression of Prox1 and enhanced green fluorescent protein (EGFP), which is driven by the same D6 enhancer as Dkk1 in D6-Dkk1 mice. D6-EGFP mice were used because available anti-Dkk1 antibodies failed to visualize the presence of Dkk1 on tissue sections. In addition, EGFP expression was more reliable from the D6-EGFP compared with a D6-Dkk1-IRES-EGFP transgenic line.
At E13.5, Prox1 was expressed at the tip of the hippocampal primordium adjacent to the hem and largely overlapped with the D6 activity. Only a few scattered Prox1+ but D6-EGFP− cells were found in the hem (Fig. 4A–A‴). Thus, the premigratory precursor pool of the DG can be affected by the D6-Dkk1 transgene. At E15.5, all Prox1+ cells in the primary neuroepithelium of the hippocampal ventricular zone were also D6-EGFP+, while in the dgm, D6-EGFP was detected in roughly 50% of the Prox1-positive cells (Fig. 4B–B‴). The Prox1+/D6-EGFP− cells were found in the medial part of the dgm. At P0, the majority of Prox1+ cells were found in granule cells of the DG, where the D6 enhancer is also active at this stage (Fig. 4C–C‴). Prox1+ cells in the dgm are also co-labeled with D6-EGFP at P0. In adults, patches of D6-EGFP activity were seen in the hilus of the DG, and only a few cells below the SGZ showed expression of EGFP (Fig. 4D–D′). Thus D6- activity was strong in embryonic progenitor zones of the DG and weaker in granule cells, while at postnatal stages, it was found to be minimal in the hilus and SGZ. To reveal that the majority of DG cells originate from the D6-positive lineage, we crossed D6-Cre mice to R26R reporter mice (Soriano,1999) in which the β-gal reporter gene is activated upon Cre-mediated recombination. The vast majority of granule cells in the adult DG were β-gal–positive (Fig. 4E–E′) confirming that most of the granule cells in the adult DG are derived from the D6-positive precursor pool. To investigate the impact of ectopic expression of Dkk1 on the dentate precursor population, anti-Prox1 immunofluorescence was performed on coronal brain sections from D6-Dkk1 embryos or infants and corresponding wild-type controls. Already at E13.5, the premigratory Prox1+ population was greatly reduced in the Dkk1 expressing animals (Fig. 4F–F′). Between E15.5 and P0, the emerging dentate blade was roughly half the size of wild-type controls (Fig. 4G–I′), and at P10, when the dentate blades have reached their developmental endpoint, the inner blade of the DG (DGi) was completely missing and the external blade (DGe) was reduced in size to less than one half of that found in the wild-types (Fig. 4J–J′). Quantifications of Prox1+ cells in the DG of embryos and infants are summarized in Figure 4K.
Several studies have shown that Dkk1 is capable of promoting apoptosis, also independently of canonical Wnt signaling (Grotewold and Ruther,2002; Lee et al.,2004; Busceti et al.,2007). To rule out the possibility that the developmental defects seen in D6-Dkk1 mutants are not due to an increase in apoptosis, we performed a terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) assay on E13.5, E15.5, and E17.5 embryos. However, we could not detect any measurable increase in apoptosis in mutant animals (data not shown), indicating that the developmental defects are not caused by increased cell death. In contrast to the cortex, where down-regulation of the canonical Wnt signaling by D6-Dkk1 did not yield any apparent fate alterations, the same genetic change led to a smaller hippocampus and a drastically reduced DG. This finding is consistent with the requirement of Wnt signaling for expansion of the precursor pool in the medial cortical wall (Lee et al.,2000; Galceran et al.,2000).
Reduced Proliferation and Neurogenesis in Germinal Zones of the DG in D6-Dkk1 Mice
Furthermore, the effect of the ectopic expression of Dkk1 on proliferating progenitor populations that contribute to the developing dentate gyrus was investigated. At E15.5, acute labeling of proliferating cells with BrdU revealed a strong reduction of BrdU-labeled cells both in the premigratory pne, in migrating progenitors in the dgm, and also in the DG (Fig. 5B–B″). Reduced proliferation was in particular observed in Prox1-positive cells as quantified by BrdU/Prox1 double-labeled cells (Fig. 5C–C″). The number of Ngn2+ cells (neuronal progenitors) was at this developmental stage only moderately reduced in the pne (Fig. 5A–A″), while Tbr2+ cells (Tbr2 also labels intermediate neuronal progenitors; Hevner et al.,2006) were reduced both in the dgm and DG (Fig. 5D–D″). In the pne, only a few cells were observed. The number of postmitotic neurons labeled with Tbr1 was found to be lower in the dgm but normal in the DG (Fig. 5E–E″). The relatively low decrease in the number of neuronal precursors (Ngn2+ or Tbr2+) or mature Tbr1+ neurons in the DG in comparison to remarkable decrease of dividing progenitors (BrdU+) might indicate premature differentiation.
At E17.5, only a modest reduction in BrdU+ cells was seen in the sm (i.e., the pne in earlier stages), whereas 50% reduction in the dgm and almost 70% reduction in the DG was observed in the number of proliferating cells (Fig. 5G–G″). Similarly at P0, the number of Ki67+ progenitors was not much reduced in the sm but dropped to less than half in the dgm (Fig. 5K–K″). This is best explained by a shift of proliferating dentate granule cells from the sm to the dgm at this stage as suggested by Zhou et al. (2004).
To measure the length of the cell cycle of progenitor cells in D6-Dkk1 animals; a labeling index analysis was performed. The labeling index quantifies the fraction of proliferating cells in the S-phase by counting the percentage of progenitor cells labeled with Ki67 that were also co-labeled with a single pulse of BrdU (Kee et al.,2002). A decrease in the labeling index indicates an increase in the cell cycle length, which is further associated with neurogenic division and neurogenic differentiation (Chenn and Walsh,2002). Examination of the labeling index in the sm, dgm, and DG of wild-type and D6-Dkk1 mutants at E17.5 by double labeling sections with anti-BrdU and anti-Ki67 antibodies, showed a significant decrease of the labeling index in the mutants in all areas examined (Fig. 5F–F″). This finding suggests that with a reduced canonical Wnt signaling, progenitor cells increase the length of the cell cycle. This is also reflected by decreased neurogenic division documented by lower number of BrdU/Prox1 double-positive cells in the pne/sm and the dgm (Fig. 5C–C″ [E15.5] and 5H–H″ [E17.5]) as well as by almost complete disappearance of Tbr2+ neuronal intermediate progenitors at E17.5 (Fig. 5I–I″). The overall decrease in neurogenic proliferation led to severe reduction of Prox1+ dentate granule neurons (shown above in Fig. 4H–J′) that also expresses markers of postmitotic neurons such as Tbr1+ (Fig. 5J–J″). To further confirm that the reduction of canonical Wnt signaling reduces neurogenic division, dividing cells were labeled with BrdU at E15.5 and their fate was analyzed at P0 by double staining with anti-NeuN antibody together with BrdU (Fig. 5 L–L″). In wild-type controls 70–90% of BrdU+ progenitors at E15.5 differentiate into NeuN+-positive postmitotic neurons in the sm, dgm, and DG. In contrast, only approximately 17–25% of the BrdU/NeuN co-labeled cells were found in the same areas in mutants. Severely reduced neurogenesis was further documented by the absence of Ngn2 mRNA in the dgm (Fig. 5M–M′) and reduced NeuroD (Fig. 5N–N′) and Tbr1 (Fig. 5O–O′) mRNA in the DG. These data show that inhibition of the canonical Wnt signaling by the ectopic expression of Dkk1 extends the cell cycle and reduces proliferation of neurogenic progenitor cells.
Impaired DG in Newborn and Adult D6-Dkk1 Mice
We analyzed whether inhibition of the canonical Wnt signaling in the premigratory and migratory zones also affected the general structure and cell fate in the postmigratory cells of the DG. Cajal Retzius cells express Reelin that functions as a guiding signal for radially migrating cells and thereby maintains a normal cell integration in the DG (Gong et al.,2007). Reelin was clearly expressed in the external blade of wild-type DGs, while in mutants its staining reflected the truncation of the external blade (Fig. 6A–6A′). However, although reduced in extension, the Reelin-positive cells appeared well structured in the mutants. The structured layer of Cajal Retzius cells in the mutants is also reflected in the organized presence of Prox1 that shows correctly organized presence of early granules cells (Fig. 4). In contrast, structural damages were observed in the scaffold of radial glia of DGs in mutant animals. Postnatal the DG develops its own scaffold of radial glia cells that can be visualized with anti- glial fibrillary acidic protein (GFAP) antibody. In wild-type P0 mice, GFAP+ cells are present in the dgm and in the hilus of the DG. While GFAP+ cells in the dgm of the mutants appeared similar to GFAP+ cells in the wild-type littermates, scaffolding of GFAP+ cells in the dentate blades was strongly reduced and appeared disorganized (Fig. 6B–B′). Musashi1 (Msi1) is normally expressed in progenitor cells in the hilus of newborn DG, and in a few cells in the dgm (Sakakibara et al.,1996; Kaneko et al.,2000; Yagita et al.,2002). In D6-Dkk1 mutants, Msi1+ cells were hardly visible in the hilus, whereas its expression was maintained in the truncated dgm (Fig. 6C–C′), indicating depletion of neural stem cell in the DG. NP2 is a class 3 semaphorin receptor that is involved in guidance of axons toward the appropriate synaptic targets during the construction of neuronal networks in the hippocampus and DG (Kolodkin et al.,1997), while Calretinin is a marker for newly generated post mitotic granule cells (Brandt et al.,2003). Similarly to the other analyzed markers, NP2 and Calretinin expression are greatly reduced, but not entirely absent in the DG of D6-Dkk1 mutants at P0 (Fig. 6D–E′). The immunostaining of Calretinin at P10 (Fig. 6F–F′) also shows that the few cells expressing this marker are in a correct position within the DG as it is also seen with Prox1 at the same stage (Fig. 4J–J′). In summary, our results indicate that, in the D6-Dkk1 embryos, DG progenitor cells that enter the migratory stream in D6-Dkk1 mutants retain the potential for correct differentiation program in the DG.
Since the D6-Dkk1 activity was found to be absent in the progenitor SGZ of the adult DG (Fig. 4D–D′), we asked whether defects observed in the neonate DG of D6-Dkk1 mice could be rescued in adult animals and a normal size DG could be restored. However, also in adult animals immunostaining of Prox1 showed a strong reduction of dentate granule neurons in the DG (Fig. 7A–A′). The DGe was less than half the size compared with wild-type and the DGi was severely reduced. The pool of GFAP+ progenitor cells was much smaller and more scattered in the mutants compared with wild-types (Fig. 7B–B′). Expression of NeuN, reflecting the presence of newborn or earlier born post-mitotic neurons, showed the same pattern as Prox1 (Fig. 7C–C′), but the neuronal layer was thinner than in wild-types (Fig. 7D–D′). Acute BrdU labeling of proliferating cells revealed a slight, but not significant, decrease in their number (data not shown). Thus, we demonstrate that the decreased progenitor pool in the embryonic germinal zones was not restored effectively enough in the adult DG. Although Wnt signaling could not be sufficiently inhibited in the adult DG of D6-Dkk1 mice, the affected progenitor pool was not able to rebuild a normal size of the DG.
Our study confirms the central role of canonical Wnt signaling in the expansion of the precursor pool in developing hippocampal fields and the DG, and also for the proper timing of neurogenesis in the embryonic telencephalon. Although the pool of DG progenitor cells in D6-Dkk1 mutants was highly diminished the remaining progenitors retained their ability to differentiate into proper cell types.
Our results are in accordance with published work. Wnt3a knockout mice showed reduced cell proliferation of hippocampal progenitors that resulted in the absence of the entire hippocampus including the rostral portion of the DG, with a slight presence of CA1 and CA3 caudally (Lee et al.,2000). A conditional knockout of β-catenin using the same D6-driven Cre recombinase also led to the disappearance of the anlage between the hem and the cingulate cortex at early embryonic stages (Machon et al.,2003), but in addition to a reduced CA2, CA3, and DG, these mice showed a complete lack of CA1 at P0. In LRP6 (Zhou et al.,2004) and Lef1 (Galceran et al.,2000; Zhou et al.,2004) knockout mice, a smaller medial wall was seen, which was similar to our observations in D6-Dkk1 mice. Furthermore, a reduced number of both BrdU+- and Prox1+ cells in LRP6−/− mice resulted in a decreased production of dentate granule neurons, an underdeveloped DG and a defect in the normal horseshoe-shaped scaffolding of glia cells that was accompanied with a loss of the DGe (Zhou et al.,2004). Both the LRP6 and Lef1 knockouts retained all CA fields intact, which was probably due to compensation from LRP5 and Tcf genes, respectively. Because both Lrp5 and Lrp6 can be inhibited by Dkk1 (Bafico et al.,2001; Semenov et al.,2001) these two membrane proteins cannot compensate each other in D6-Dkk1 animals. We could detect severe but not complete reduction of the DG in D6-Dkk1, which might be explained by several reasons. (1) The level of D6-Dkk1 expression was not sufficient to completely inhibit the canonical Wnt pathway in the key progenitor zones. Indeed, we detected some, albeit lower expression of the BAT-Gal Wnt reporter in D6-Dkk1/BAT-Gal crosses. (2) Timing of inhibition of Wnt signaling by D6-Dkk1 was improper. From around E9.5, Wnt signals from the hem provide patterning and proliferation signals to the developing hippocampus and choroid plexus (Grove et al.,1998). Because the D6 enhancer becomes active between E10.5 and E11.5, the initial specification and proliferation of hippocampal cells in the hippocampal primordium could not be inhibited in D6-Dkk1 mutants. Because Wnt3a−/− mice lack the entire DG one could also exclude the possibility that a subpopulation of DG precursor cells is not under the control of Wnt signaling. Indeed, in all Wnt pathway mutants mentioned above, the DG was always strongly affected, whereas the hippocampal fields show various degrees of defects.
The dentate progenitor cells, which are situated closest to the Wnt producing hem within the hippocampal primordium, will be the most sensitive to down-regulation of the canonical Wnt signaling because they may need a high dose of Wnt activity to develop into dentate progenitors expressing Prox1. The hippocampal fields might also develop along such a Wnt gradient (stronger in CA3 and weaker in CA1) and may, therefore, require a lower dose. This presumption may explain an almost complete absence of the hippocampus in the Wnt3a−/− mice that entirely lack this specification signal. Along this line, failure to develop a proper DG in the LRP6 or Lef1 mutants where compensation may occur with Lrp5 or TCFs, respectively, the compensation may not be sufficient for development of DG progenitor cells but strong enough for hippocampal layers. This “gradient” hypothesis is in agreement with the analysis of D6-CLEF mice in which ectopic activation of the Wnt pathway led to over proliferation and altered cellular specification in the cortex where Prox1+ ectopic cells required a higher level of Wnts than hippocampal specific cells (Machon et al.,2007).
We observed a dramatic reduction in dentate progenitor cells expressing Prox1 in D6-Dkk1 mutants already at E13.5, which resulted in a smaller area of hippocampal neuroepithelium. At this stage, markers of mature neurons such as Tbr1 and NeuN in the hippocampal neuroepithelium are not yet expressed (data not shown), suggesting that Prox1 labels DG progenitors. This finding indicates that the reduction in canonical Wnt signaling at this stage resulted in a reduced proliferation of cells rather than premature neurogenesis and differentiation of progenitors. On the other hand, relatively low decrease of Ngn2+ neuronal progenitors and negligible change in Tbr2+ and Tbr1+ population in the DG at E15.5 in comparison to remarkable decrease of BrdU+ cells in all progenitor zones, would suggest a temporary increased neuronal differentiation that leads to depletion of the neurogenic progenitor pool. A longer length of the cell cycle revealed by the labeling index supports this explanation. The depletion of neurogenic progenitor pool leads to decreased neurogenic division that was clearly documented by counting double labeled BrdU+/Prox1+ cells at E17.5 and NeuN+ cells at P0 that were prelabeled with BrdU at E15.5 (See Fig. 8 for schematic representation of the phenotypes).
At late gestational and postnatal stages, the production of granule cells from their precursors are relocated to the migratory route and the dentate hilus (Altman and Bayer,1990a; Bagri et al.,2002). This finding may explain why we only observed reduction in progenitor proliferation (labeled with BrdU or Ki67) in the dgm and DG from E17.5 onward. Canonical Wnt signaling is, therefore, a central requirement for the initial size and proliferation of the premigratory DG progenitor population at early embryonic stages, and for the proliferation of migrating dentate progenitor cells, as well as for the timing of onset of neurogenesis that was prematurely initiated when canonical Wnt signaling was reduced.
Although reduced in size, a cellular specification of various cell types in the DG in D6-Dkk1 animals was investigated at postnatal stages. All analyzed cell types were present and properly situated but only in lower numbers. Msi1, Ngn2, or Calretinin showed a great loss of progenitor cells, immature neuronal precursor cells and also more mature dentate granule cells as a result of ectopic expression of Dkk1. The severe reduction of progenitor cells in this area is supported by a complete loss of NeuroD, because the NeuroD knockout animals show dramatic defects in the proliferation of precursor cells once they reach the dentate gyrus (Liu et al.,2000). Immunostaining of radial glial cells with anti-GFAP showed both reduction and disorganization of the DG, which was accompanied by the presence of a highly reduced DGe and a missing DGi. Both Lef1 and LRP6 mutants lacked GFAP+ radial glia scaffolding resulting in loss of the DGe but presence of DGi and this was also associated with insufficient expansion of the DG precursor pool (Galceran et al.,2000; Zhou et al.,2004). The opposite effect in D6-Dkk1 mutants with present DGe and absent DGi is most probably explained by the proper organization of GFAP and Reelin that correctly positions the dentate cells, but that the reduced population of these cells depletes before they are able to properly populate the entire DG. These results again suggest that Wnts specify and regulate the size of the initial region of the dentate neuroepithelium and also its maintenance in the neurogenic zones at later stages. In addition, it supports the idea that canonical Wnts are required for radial glia cells that may serve as neural progenitors in the DG in a manner similar to the cortex.
In adult stages, the DG was still significantly smaller in D6-Dkk1 mice, despite the fact that Dkk1 expression under the D6 enhancer was very weak at this stage. In addition, the number of proliferating cells labeled with BrdU was not significantly lower in the mutants. Nevertheless, the size of the DG was not restored to normal, indicating that the depleted progenitor pool from embryonic stages cannot repopulate the adult DG. Neurogenesis in the adult DG requires active Wnt signaling from astrocytes surrounding the germinal SGZ (Lie et al.,2005). Increased Wnt signaling by means of a lentivirus led to higher number of neuronal progenitors while active inhibition decreased it. Our results at embryonic stages are in principle in line with this report. In adult DG of D6-Dkk1, BrdU incorporation was restored to normal but the astrocytic population labeled with GFAP was much reduced, and thereby also the level of Wnt signaling was probably lower. Therefore, the progenitor pool, which was depleted from embryonic stages and was in addition surrounded by fewer Wnt-expressing astrocytes, could not rescue the abnormalities.
The cortical development in D6-Dkk1 mutants was found to be normal when analyzed in newborns. This finding contrasts to the findings that were seen in the LRP6 mutants (Zhou et al.,2006) that exhibit lower proliferation in the cortical neuroepithelium and thinner cortical plate. Furthermore, Woodhead et al. (2006) reported a premature cell cycle exit and differentiation of cortical progenitors upon elimination of β-catenin expression of dominant-negative Tcf4. The expression of Dkk1 under D6 may not be sufficient to inhibit canonical Wnt signaling in the cortical area where the Wnt pathway is only moderately active.
Generation and Genotyping of D6-Dkk1-IRES-EGFP Transgenic Mouse Line
PCR with the primers p18-Dkk1f (5′-GTCAGTCGACGCGGCCGCCATGATGGTTGTGTGTGCAGC-3′) and p19-Dkk1r (5′-TGACGGATCCTTGGGATTGAGCTGACAAATC-3′) was carried out on a Dkk1 cDNA (a gift from C. Niehrs) to create a 0.9-kb fragment covering the entire open reading of Dkk1. This fragment included a SalI and NotI site at the 5′ end and a BamHI at the 3′ end. The PCR product was cut with SalI and BamHI and cloned into the corresponding sites of pIRES2-EGFP (Clontech). The resulting plasmid was further cut with NotI, and a 2.2-kb fragment containing Dkk1-IRES-EGFP was ligated into the NotI sites of D6-EGFP (Machon et al.,2002). This resulted in D6-Dkk1-IRES-EGFP. SalI digestion deleted the dispensable sequence of the plasmid, and the 8.3-kb fragment containing D6-Dkk1-IRES-EGFP and a SV40-poly(A) signal was injected into the pronuclei of mouse zygotes (strain C57Bl6/CBA F2) according to standard protocols. Analysis was performed on animals from crosses between D6-Dkk1+/− (F0, F1, or F2) and wild-types (wt) (C57Bl6/CBA). In the text, wt refers to homozygous null littermates in these crosses.
Animals or embryos were genotyped by PCR using primers p18-Dkk1f and p19-Dkk1r for the Dkk1 gene, or by GFP expression in newborn animals. Out of seven transgenic animals, six transmitted to progeny, and five of these yielded a specific phenotype. Only animals with the most severe phenotype were analyzed. Canonical Wnt signaling was mapped by crossing heterozygous D6-Dkk1 transgenic mice to homozygous BAT-gal mice (Maretto et al.,2003). Lineage tracing of D6-positive cells was carried out as described in (van den Bout et al.,2002). The day of vaginal plug was considered as embryonic day 0.5 (E0.5), and the day of birth as P0. All experiments on animals were carried out in accordance with the National Institute of Health guide for the care and use of laboratory animals. All efforts were made to minimize the number of animals used and their suffering.
In Situ Hybridization and Synthesis of Digoxigenin-Labeled Riboprobes
In situ hybridization on 10–12 μm coronal cryosections was carried out as previously described (Machon et al.,2002) with overnight hybridization at 68°C. Synthesis of digoxigenin-labeled riboprobes were performed as described by Diep et al. (2004; Dkk1) and by Machon et al. (2003; SCIP, NP2, KA1, Tbr1, Prox1, SFRP1, and Ngn2). NeuroD (provided by F. Guillemot) was linearized with KpnI and transcribed with T7 RNA polymerase, and LacZ riboprobe was transcribed with T7 RNA polymerase from the plasmid pBS-βgalNE linearized with NotI.
TUNEL assay was used to assess the presence of apoptotic cells on 10–12 μm coronal cryosections. The procedure was performed according to manufacturer's instructions (In Situ Cell Death Detection Kit, POD, Roche Applied Science).
Immunohistochemistry and BrdU Labeling
Immunohistochemistry on 10–12 μm coronal cryosections was executed as previously described (Machon et al.,2003). Staining with anti-Ki67, anti-Tbr2, and anti-Tbr1 required an additional step with 0.1M citrate buffer (pH 6.0) boiling in the microwave oven at full effect for 1 min. To test adult neurogenesis in the DG, adult mice were injected intraperitoneally daily with BrdU (150 μl of 10 mg/ml; Sigma) for 5 consecutive days before harvest. Short-pulse BrdU injections was performed with 150 μl of 10 mg/ml 2 hr before harvesting. Detection of BrdU incorporation required an additional treatment with 2 M HCl for 10 min at 37°C and subsequently neutralized with 1× TBE buffer and phosphate buffered saline.
Primary antibodies: mouse monoclonal anti-Reelin (Chemicon) 1:200 for Cajal Retzius cells, mouse anti-N-cadherin (BD Biosciences) 1:500 for cell adhesion, rabbit polyclonal anti-Ki67 (Novo Castra) 1:1,000 for dividing cells, rabbit polyclonal anti-Prox1 (Chemicon) 1:2,000 for dentate granule cells, rat polyclonal anti-Musashi1 (a gift from H. Okano) 1:200 for neural progenitors, mouse monoclonal anti-GFAP (Sigma) 1:400 for radial glial cells, rabbit polyclonal anti-calretinin (Chemicon) 1:200 for immature granule cells, mouse monoclonal anti-NeuN (Chemicon) 1:400 for postmitotic neurons, rat monoclonal anti-BrdU (Abcam) 1:50 for prelabeled proliferating cells, rabbit anti-Tbr21 1:500 (Chemicon) for intermediate neuronal progenitors, rabbit anti-Tbr11:500 (Chemicon) for postmitotic neurons, mouse anti Neurogenin2 1:20 (a gift from D. Anderson), and DAPI 1 mg/ml (Roche) 1:1,000 labeling cell nuclei. Ki67, Tbr2, Tbr1 and Reelin immunohistochemistry were enhanced with goat anti-rabbit or mouse biotin-xx (Molecular Probes) 1:500. Secondary antibodies: anti-mouse, anti-rabbit, or anti-rat Alexa 594 (Molecular Probes). For some antibodies, anti-rabbit biotin followed by streptavidin-Alexa594 (Molecular Probes) incubation was used to enhance fluorescence.
Counting of cells was performed on at least three animals from each genotype at comparable section levels in areas of the same size to derive average values, except for Prox1 staining at P10 that only had contained one counting of each genotype. To count different anatomic compartments, standardized boxes were drawn on the images and positive cells were counted within that compartment. Error bars indicate standard deviation.
We thank the Norwegian Transgenic Centre for creating the D6-Dkk1-IRES-eGFP transgenic mice. The authors also thank Dzung Diep, Line Mygland, Olga Machonova, and James Booth for excellent technical help, and Benedicto Geronimo for technical help with mouse line maintenance.