Presenilin-1 regulates neuronal differentiation during neurogenesis

PS1 is a major gene responsible for familial Alzheimer’s disease (FAD), and mutations in PS1 account for approximately 50% of early-onset FAD cases (Selkoe, 1998). Understanding the normal physiological functions of PS1 may shed light on the pathogenic mechanism of FAD-linked PS1 mutations. Identification of the PS1 homologue in C. elegans, sel-12, which facilitates signalling mediated by the Notch homologue LIN-12, provided the first evidence that PS1 interacts with the Notch signalling pathway (Levitan and Greenwald, 1995). The LIN-12/Notch family of receptors mediates cell-cell interactions that specify cell fate during development. SEL-12 affects LIN-12 activity by regulating its processing or trafficking (Levitan and Greenwald, 1998). PS1 shares functional homology with SEL-12, based on the finding that the wild-type human PS1 cDNA complements the sel-12 mutant phenotype (Baumeister et al., 1997; Levitan et al., 1996). However, PS1 containing FAD-linked mutations exhibited reduced ability to rescue sel-12 mutations, suggesting that mutant PS1 has reduced biological activity (Baumeister et al., 1997; Levitan et al., 1996).

Recent identification of the Drosophila Presenilin (PS) provided further evidence for the interaction between PS1 and the Notch signalling pathway (Struhl and Greenwald, 1999; Ye et al., 1999). Fly mutants lacking both maternal and zygotic PS exhibit a neurogenic phenotype and are virtually indistinguishable from the Notch-null mutant, suggesting that PS function is required for normal Notch signalling in Drosophila (Struhl and Greenwald, 1999; Ye et al., 1999). PS is also required for the proteolytic cleavage of Notch to release its intracellular effector domain (ICD) (Struhl and Greenwald, 1999). The involvement of Presenilins in Notch processing is further supported by in vitro studies using truncated Notch1 and primary cell cultures derived from PS1^−/− mice (De Strooper et al., 1999; Song et al., 1999). Levels of the ICD fragment were reduced in cultured PS1^−/− neurons and fibroblasts, indicating that PS1 is important for the efficient proteolytic release of the ICD. PS1 bearing FAD-linked mutations exhibits reduced ability to rescue this Notch processing defect in cultured PS1^−/− cells, suggesting that mutant PS1 has reduced activity in Notch processing (Song et al., 1999).

To characterize the normal physiological role of PS1 in mice, we previously generated mice with a targeted germ-line disruption of the PS1 gene (Shen et al., 1997). Mutant mice homozygous for the resulting null allele exhibited similar defects in somitogenesis to those observed in Notch1-null mutant mice (Conlon et al., 1995; Swiatek et al., 1994), as well as severe malformation of the axial skeleton and cerebral hemorrhage (Shen et al., 1997; Wong et al., 1997). Furthermore, we showed that lack of PS1 results in a reduction in the neural progenitor population and a subsequent reduction in the neuronal population, indicating a critical role for PS1 in murine neurogenesis (Shen et al., 1997). We also observed symmetric, region-specific loss of neural progenitor cells in the diencephalon and telencephalon of the PS1^−/−brain beginning at E12.5 and 14.5, respectively. Loss of Cajal-Retzius neurons in late embryonic development has also been reported (Hartmann et al., 1999).

Here, we report that the reduction in the neural progenitor population in PS1^−/− mice is caused by premature differentiation of neural progenitor cells. To investigate further the mechanism by which PS1 regulates neuronal differentiation, we examined the expression of genes involved in the Notch signalling pathway. We found a reduction in the level of HES-5 transcripts as well as an increase in the level of Dill transcripts in the PS1^−/− brain, indicating that absence of PS1 function leads to reduced Notch signalling during neural development. We also found that expression of Dll1 is increased in the presomitic mesoderm of PS1^−/− embryos, whereas Notch1 expression is unchanged, in contrast to a previous publication reporting that expression of both Notch1 and Dll1 is markedly reduced in the presomitic mesoderm of PS1^−/− embryos (Wong et al., 1997). Taken together, our findings provide the first evidence that PS1 controls neuronal differentiation and the expression of the downstream target genes regulated by Notch signalling.

Timed matings between heterozygous PS1 knockout mice were set up, and the morning of the day when a vaginal plug is seen is designated as embryonic day 0.5 (E0.5). Embryos at the desired developmental stages were dissected from pregnant females, and the tail and limbs were removed from each embryo for genotyping. Embryonic heads were then removed and fixed in 4% paraformaldehyde for 2-3 hours at 4°C, cryoprotected overnight at 4°C in 30% sucrose, embedded in OCT medium (Tissue-Tek) and frozen on dry ice. Serial transverse sections (7-10 pm) were cut on a cryostat and collected on coated slides. Embryos with cerebral hemorrhages, regardless of the severity, were discarded.

Brain sections were blocked for 3 hours at room temperature with 3% nonfat milk, rinsed briefly in PBS and incubated overnight at 4°C with an anti-microtubule-associated protein 2 (MAP2) monoclonal antibody (1:1000, kindly provided by Dr Ken Kosik) or an anti-Nestin monoclonal antibody (1:1000, PharMingen). For Notch1 immunostaining, sections were incubated with a blocking solution containing 5% goat serum, 1% BSA, 0.1% glycine, 0.1% L-lysine and 0.4% Triton X-100 in PBS for 1 hour at room temperature followed by incubation with a polyclonal antibody (IC) against mouse Notch1 amino acid residues 1759-2306 (1:250, kindly provided by Dr Alan Israel) overnight at 4°C. Sections were then rinsed twice in PBS and incubated at room temperature with a rhodamine-conjugated goat anti-mouse secondary antibody (Jackson Immunoresearch) for MAP2 (1:400) and Nestin (1:100). For Notch1 immunostaining, the sections were rinsed three times in PBS and incubated with a fluorescein-conjugated goat anti-rabbit secondary antibody (1:500, Alexa) for 1 hour at room temperature. Images were viewed and collected on a Zeiss Axioscop connected with a SPOT 2 Cooled Color Digital camera (Diagnostic Instruments) or a BioRad MRC 1024 confocal laser scanning microscope.

Pregnant females were injected intraperitoneally with BrdU (10 mg/ml) at 100 pg/g body mass. After 30 minutes or 96 hours of labelling, embryos were removed, processed and serially sectioned. The brain sections were incubated in 2 N HCl for 1 hour at 37°C, neutralized in borate buffer (pH 8.5) for 1 hour at room temperature, rinsed twice in PBS and blocked with 3% nonfat milk for 3 hours at room temperature. The sections were then incubated with an anti-BrdU monoclonal antibody (1:10, Becton Dickinson) overnight at 4°C, rinsed twice in PBS, followed by incubation with a rhodamine-conjugated goat anti-mouse secondary antibody (1:100, Jackson Immunoresearch) for 1 hour at room temperature.

The percentage of BrdU-labelled cells was determined by dividing the number of BrdU₊ cells by the total number of progenitor cells. For both PS1^−/− and littermate control brains, the number of BrdU⁺ cells was counted in a comparable area of the anterior telencephalon and in half of the diencephalon. The total number of progenitor cells in the telencephalon was determined by counting the number of nuclei stained with Sytox Green in the comparable area on the adjacent sections, since there are few postmitotic neurons in the telencephalon at E10.5. Due to the presence of postmitotic neurons in the diencephalon at this stage, the total number of progenitor cells in the diencephalon was determined by subtracting the number of MAP2-positive cells from the total number of Sytox Green-stained cells in half of the diencephalon on the adjacent sections. A Zeiss Axioscop equipped with a reticular eyepiece was used to count the number of cells in an area of 5x 10⁴ pm² in the telencephalon and half of the diencephalon under 63x magnification.

The TUNEL (terminal-deoxynucleotidyl-transferase-mediated deoxy uridine triphosphate nick-end labelling) assay was used to label cells that are undergoing apoptosis. Brain sections were rinsed in PBS for 5 minutes and then 50-75 pl of TUNEL reaction mixture (In situ cell death detection kit, Boehringer Mannheim) was applied to each slide. The sections were incubated at 37°C for 1 hour and 15 minutes, rinsed twice in PBS and then incubated in bisbenzimide (2 pg/ml) for 10 minutes at room temperature. Apoptotic cells were identified by their TUNEL-reactivity as well as their fragmented and condensed nuclei stained by bisbenzimide. The number of apoptotic cells in each brain was quantified by counting three serial sections under 40x magnification. Statistical significance was determined by Student’s t-test.

Sense and antisense riboprobes were synthesized using an in vitro transcription kit (Boehringer Mannheim) following the manufacturer’s instructions and then hydrolyzed to approximately 100-200 nucleotides (nt) fragments. In situ hybridization was carried out as previously described (Schaeren-Wiemers and Gerfin-Moser, 1993) with minor modifications. Brain sections were incubated with the prehybridization solution (50% formamide, 5x SSC, 0.02% SDS, 0.1% N-lauroylsarcosine and 2% blocking reagent) at room temperature (RT) for 6 hours, then hybridized together in hydrolyzed riboprobes (40 ng/ml) at 70°C overnight.

For color detection, the sections were placed upside down on NBT/BCIP reaction mixture containing 2 mM freshly prepared levamisole. The slides containing brain sections were elevated with two spacers (0.2 mm) to allow sufficient and equal contact with the reaction mixture. Precautions were taken to ensure identical conditions for all procedures of in situ hybridization involving the PS1^−/− and littermate control brain sections. All images were collected with a cooled CCD camera with identical exposure time between the PS1^−/− and control brain sections. The brain sections hybridized with the sense probe showed no staining.

Heads and tail buds from PS1^−/− and littermate control mice at E11.5 were collected for the preparation of total RNA using Tri Reagent (Sigma). Poly(A)Pure kit (Ambion) was used to enrich poly(A)₊ RNA from total RNA. Approximately 30 gg of total RNA and 5 |ag of poly(A)₊ RNA were used for the Northern blotting. The blots were hybridized with ³²P-labelled probes, which were synthesized with Prime-It II random labelling kit (Stratagene), in the hybridization solution (1% SDS, 6x SSC and 10% dextran sulfate) at 67°C overnight. The same blot was then hybridized with a G3PDH probe to normalize the amounts of mRNA loaded in each lane. NIH Image software was used to quantify the level of transcripts by comparing the intensities of the bands after subtracting the background.

In situ hybridization was performed on comparable sections from PS1^−/− and littermate control brains at E11.5 under identical conditions. The images of brain sections and a micro-ruler were collected with a CCD camera under the same magnification and resolution. The large number of Dill -expressing cells in the diencephalon, particularly in the PS1^−/− diencephalon, makes it difficult to count the number of Dll1⁺ cells accurately in the diencephalon; therefore, our analysis was focused on the telencephalon. D ll1-positive cells in half of the telencephalon were counted on the collected brain sections, and the area of half of the telencephalon was measured using NIH Image software. The density of Dll1 -expressing cells was calculated by dividing the number of Dll+ cells by the measured area. The data shown represent an average of one section from three brains per genotype.

Whole-mount in situ hybridization was performed essentially as described (Hogan et al., 1994). Briefly, embryos at day 9.5 were dissected and fixed in 4% paraformaldehyde at 4°C overnight. Yolk sacs were used for genotyping. After genotyping, PS1^−/− and littermate control (PS1^−/−) embryos were processed either individually or together during subsequent steps. The same probes were used as those used for in situ hybridization on brain sections. For those embryos that were processed together, the PS1^−/− and control embryos were later identified by the morphological differences in somites (Shen et al., 1997).

To understand the normal function of PS 1 in the brain, we examined the PS1^−/− embryonic brain for defects in neural development. Beginning at embryonic day 12.5 (E12.5), it becomes evident that the ventricular zone in the PS1^−/− brain is thinner than in the control brain, reflecting a reduction in the population of neural progenitor cells during neurogenesis (Shen et al., 1997). Cortical neurogenesis in mice begins at approximately E11 and continues through E17, comprising 11 cell cycles (McConnell, 1995; Takahashi et al., 1995). During early neurogenesis, very few progenitor cells in the neuroepithelium exit the cell cycle to become postmitotic neurons, while the vast majority of progenitor cells reenter the cell cycle after mitosis, resulting in a steady expansion of the neural progenitor population. As neurogenesis progresses, a fraction of neural progenitor cells exit the cell cycle, differentiate into postmitotic neurons and migrate to form the developing neocortex. Therefore, a reduction in the population of neural progenitor cells could occur via several possible mechanisms: premature differentiation of progenitor cells, decreased proliferation, or increased apoptotic cell death. To distinguish among these possibilities, we compared the PS1^−/− and control brains for differences in neuronal differentiation and migration, cell proliferation and survival. Due to severe cerebral hemorrhage often associated with PS1^−/− mice at later embryonic stages, our analysis is focused on the examination of the PS1^−/−and control brains that are free of hemorrhages between E10.5 and 13.5.

To assess neuronal differentiation in the PS1^−/− and control brains, we performed immunostaining for MAP2, a marker specific for postmitotic neurons (Crandall et al., 1986). Serial transverse sections of the PS1^−/− and littermate control brains at E10.5, 11.5, 12.5 and 13.5 were stained with anti-MAP2 antibody, and corresponding sections of the PS1^−/− and control brains were compared. Although few MAP2-immunoreactive neurons are present in the telencephalon at E10.5 (data not shown), as many as 4-6 layers of MAP2⁺ neurons were detected in the PS1^−/− diencephalon, whereas only 1-2 layers of MAP2⁺ neurons were detected in the control (Fig. 1A). The ratio of MAP2₊ cells to total cell population in the PS1^−/− diencephalon is two-to threefold higher than in the littermate control (data not shown).

At E11.5, increases in MAP2 immunoreactivity in the PS1^−/− brain relative to the control are found in the anterior telencephalon (Fig. 1Ba,b), and more substantially in the posterior telencephalon (Fig. 1Bc,d). In the anterior telencephalon, 2-4 layers of MAP2⁺ neurons were detected in the PS1^−/−brain in comparison to a single layer of MAP2 immunoreactive neurons in the littermate control (Fig. 1Ba,b). The ventricular zone in this region contains similar numbers of MAP2-negative progenitor cell layers in the PS1^−/− brain and the control. The posterior telencephalon of the PS1^−/− brain contains more layers of MAP2⁺ neurons and a smaller ventricular zone in comparison to the control (Fig. 1Bc,d). The diencephalon exhibits the most substantial differences in the populations of MAP2-immunoreactive neurons and MAP2negative progenitor cells. In the diencephalon of the PS1^−/− brain, the MAP2-immunoreactive neurons encompass numerous cell layers and have largely replaced the MAP2negative progenitor cells in the ventricular zone (Fig. 1Be-h).

At E12.5, these differences are still prominent in the telencephalon and diencephalon. More layers of MAP2⁺ neurons are present in the anterior telencephalon in the PS1^−/− brain (Fig. 1Ca,b), and the diencephalic ventricular zone is largely replaced by the MAP2₊ neurons (Fig. 1Ce,f). By E13.5, however, MAP2 immunoreactive neurons comprise a similar number of cell layers in the anterior telencephalon of the PS1^−/−and littermate control brains (Fig. 1Da,b).

In summary, markedly increased numbers of postmitotic neurons identified by MAP2 immunoreactivity accumulate in the telencephalon and diencephalon of the PS1^−/− brain during early neural development, accompanied by a progressive reduction in size of the ventricular zone. These observations indicate that neural progenitor cells differentiate into postmitotic neurons prematurely in the absence of PS1, resulting in early depletion of the neural progenitor population.

Interestingly, the severity of this phenotype varies in the brain subregions examined. The diencephalon exhibits the most striking difference in MAP2 immunoreactivity, whereas the anterior telencephalon shows the least difference. Thus, during neurogenesis, PS1 controls neuronal differentiation in a regionspecific manner by regulating the cell-fate choice between neural progenitor cells and postmitotic neurons.

Although premature neuronal differentiation provides an explanation for the reduction in the number of neural progenitor cells observed in the PS1^−/− brain, it remained possible that a decrease in cell proliferation and/or an increase in apoptotic cell death could be contributing factors as well. Cell proliferation in the neuroepithelium was examined by BrdU labelling. The fraction of cells labelled with BrdU during a short pulse reflects the fraction of cells in S phase during that period, permitting an assessment of the proliferation rate of a population of cells (Gratzner, 1982). After 30 minutes of BrdU pulse-labelling, serial transverse sections of the PS1^−/− and littermate control brains were stained with anti-BrdU antibody. Comparable sections of the PS1^−/−brains at E10.5, 11.5, 12.5 and 13.5 were compared to their littermate controls.

We observed no significant differences in the BrdU labelling patterns of the PS1^−/− and littermate control brains at E10.5 (Fig. 2A). The ratio of BrdU-labelled cells to the total number of progenitor cells in the ventricular zone of the telencephalon and diencephalon is similar in the PS1^−/− and control brains (Table 1). This result shows that lack of PS1 does not lead to a reduction in the proliferation rate of neural progenitor cells.

We did, however, observe differences in BrdU labelling in the diencephalon at E11.5. The BrdU labelling pattern in the telencephalon of the PS1^−/− brain appears similar to that of the control (Fig. 2Ba,b), but fewer BrdU-labelled cells are detected in the diencephalon of the PS1^−/− brain (Fig. 2Bc-f). The location of these BrdU-labelled cells in the PS1^−/− brain corresponds to the thin remnant of the diencephalic ventricular zone revealed by the absence of MAP2 immunostaining at this age (Fig. 2Be,f; compare with unstained cells in Fig. 1Be,f). The BrdU labelling patterns of the anterior telencephalon are indistinguishable at E12.5 (Fig. 2Ca,b) and E13.5 (Fig. 2Da,b), but fewer BrdU-labelled cells appear to be present in the ganglionic eminence of the PS1^−/− brain (Fig. 2Cc,d). The simplest explanation for the reduced BrdU labelling in the ganglionic eminence and diencephalon of the PS1^−/− brain is that it is a secondary effect, reflecting early and progressive depletion of progenitor population in the ventricular zone of these brain regions due to premature differentiation of neural progenitor cells. The corresponding presence of increased numbers of postmitotic neurons in these particular regions supports this explanation.

To determine whether increased apoptotic cell death might contribute to the reduction in the neural progenitor population observed in the PS1^−/− brain, we assessed the number of apoptotic cells in the PS1^−/− and control brains at E11.5 and E12.5. Apoptotic cells were detected by the TUNEL assay, which labels the 3’-OH termini of DNA strand breaks generated during DNA fragmentation (Gavrieli et al., 1992). Bisbenzimide staining was also used to confirm the identity of apoptotic cells, which contain condensed, clumped and/or fragmented nuclei (Deckwerth and Johnson, 1993). Similar numbers of apoptotic cells labelled by the TUNEL assay were detected in the telencephalon (Fig. 3) and diencephalon (data not shown) of the PS1^−/− and littermate control brains at E11.5 (Fig. 3a,b) and E12.5 (Fig. 3c,d).

Quantitative comparison of the number of apoptotic cells in multiple comparable sections of the PS1^−/− and littermate control brains at E11.5 and E12.5 revealed no significant differences (Table 2). A total of 42-50 apoptotic cells were detected in transverse sections of the entire telencephalon and diencephalon of both PS1^−/−and control brains. In addition, the small percentages of apoptotic cells (approx. 0.6% for E11.5 and 0.3% for E12.5) cannot account for the marked reduction of the progenitor population in the PSP∼ brain. Likewise, no significant differences in the number of apoptotic cells were seen in cultured primary neural progenitor cells derived from the PS1^−/− and littermate control telencephalons at E12.5 (data not shown). These results indicate that increased apoptosis is not a primary cause of the defective neurogenesis in PS1^−/− brain, and that PS1 is not involved in the regulation of apoptosis during early neurogenesis.

The laminar architecture of the developing cerebral hemisphere is disrupted in the PS1^−/− brain, as indicated by the absence of the characteristic demarcations among the developing cortical plate, the intermediate zone, the ventricular and subventricular zones in the PS1^−/− brain at E14.5 (Fig. 4b). The cortical plate in particular appears poorly defined and generally thinner in the PS1^−/−brain (Fig. 4b), compared to wild type (Fig. 4a) (Shen et al., 1997). In addition, occasional neuronal heterotopias are present in the marginal zone of the developing cortex in the PS1^−/−brain at E17.5 (data not shown), as previously reported (Hartmann et al., 1999). These results indicate that cortical neuronal migration is abnormal in the PS1^−/− brain.

To investigate further the pattern of cortical neuronal migration in PS1^−/− mice, we pulse-labelled newly generated progenitor cells and neurons with BrdU at E10.5 and analyzed their migration pattern at E14.5. Comparison of corresponding transverse sections of the PS1^−/− brain (Fig. 4d) and the littermate control (Fig. 4c) revealed an increased number of brightly stained cells in the PS1^−/− telencephalon (Fig. 4d). These brightly stained cells represent differentiated neurons that exited the cell cycle immediately following the pulse, thereby retaining all of the incorporated BrdU. In contrast, progenitor cells in the ventricular zone that have undergone multiple rounds of S and M phase following the pulse stain rather faintly due to dilution of the incorporated BrdU (Fig. 4c,d). This result indicates that greater numbers of neuronal cells in the PS1^−/− brain exit the progenitor pool early to become postmitotic neurons, which is consistent with our results obtained with MAP2 immunostaining (Fig. 1). Furthermore, the BrdU-labelled neurons in the control telencephalon are localized appropriately within a well-defined cortical plate (Fig. 4c), whereas the labelled neurons in the PS1^−/−telencephalon are scattered diffusely across multiple cell layers (Fig. 4d), forming a disorganized cortical plate lacking a clear boundary with the adjacent intermediate and ventricular zones. These results support a role for PS1 in cortical layer formation and neuronal migration.

To understand the mechanism underlying the premature neuronal differentiation observed in the PS1^−/− brain, we examined the expression of genes in the Notch signalling pathway, which is known to be involved in cell-fate determination during neurogenesis in Drosophila (Artavanis-Tsakonas et al., 1999). The role of Notch receptors in murine neurogenesis is poorly understood. Mice lacking Notch1 or Notch2 function die at approximately E9 or E11, respectively, before cortical neurogenesis begins (Conlon et al., 1995; Hamada et al., 1999; Swiatek et al., 1994). An increased number of cells expressing proneural transcription factors, however, was identified in the midbrain and hindbrain of Notch1^−/−mice (de la Pompa et al., 1997). The basic helixloop-helix transcription factors Hes1 and Hes5 (mammalian hairy and Enhancer-of-split homolog), which are downstream effectors of the Notch signalling pathway, have been shown to regulate murine neuronal differentiation (Ishibashi et al., 1995, 1994; Kageyama and Nakanishi, 1997; Ohtsuka et al., 1999; Tomita et al., 1996). We therefore first examined expression of Hes1 and Hes5 transcripts by in situ hybridization and northern analyses to determine whether their expression patterns and levels are affected in PS1^−/− mice.

In situ hybridization analysis of comparable sections of PS1^−/− and littermate control brains at E11.5 showed that expression of Hes5 is reduced in the PS1^−/− brain (Fig. 5A). Consistent with previous studies (Akazawa et al., 1992), Hes5 expression is localized to the ventricular zone within each brain region. In the control brain, the expression level of Hes5 transcripts is higher in the ganglionic eminence (Fig. 5Ac) and diencephalon (Fig. 5Ad) than in the anterior telencephalon (Fig. 5Ab). In the PS1^−/− brain, the expression level of Hes5 is similar in the anterior telencephalon, ganglionic eminence and diencephalon, but is reduced relative to the control in each of these regions (Fig. 5Ab’,c’,d’). Therefore, the reduction in the level of Hes5 transcripts is greater in the ganglionic eminence and diencephalon than in the anterior telencephalon. These regional differences in reduced Hes5 expression parallel the regional differences in premature neuronal differentiation observed at the same developmental stage, with a more severe phenotype in the ganglionic eminence and diencephalon than in the anterior telencephalon of the PS1^−/− brain. In addition, the region of Hes5 expression is thinner, with fewer cell layers in the PS1^! diencephalon (Fig. 5Ad’), corresponding to the thinned ventricular zone in this region of the PS1^−/− brain (Fig. 1Bf,h).

Despite the precautions we took to perform in situ hybridization of PS1^−/− and littermate control brain sections simultaneously and under identical conditions, it is still a less quantitative method than northern analysis for comparing the expression level of mRNA transcripts in PS1^−/−and control mice. Therefore, we also performed northern analysis using poly(A)⁺ RNA derived from the PS1^−/−and littermate control heads at E11.5. The northern result showed an approximately 25% reduction in the level of the Hes5 transcript in the PS1^−/− brain, confirming the downregulation of Hes5 expression in the absence of PS1 (Fig. 5B).

In situ hybridization analysis of Hes1 expression revealed a similar pattern and level of expression in the PS1^−/−and control brains at E11.5 (data not shown). Consistent with a previous report (Sasai et al., 1992), Hes1 appears to be expressed in all cells of the embryonic brain with the highest expression in the progenitor cells of the ventricular zone. We also performed Northern analysis using poly(A)⁺ RNA derived from the PS1^−/− and littermate control heads at E11.5, and the result showed similar levels of Hes1 transcripts in the PS1^−/− and control brains, providing further confirmation that expression of Hesl is unaffected in the PS1^! brain (Fig. 5B).

In summary, the reduced expression of the Notch1 downstream target gene Hes5 suggests a downregulation of Notch signalling in the absence of PS1. These results also provide a molecular basis for the premature neuronal differentiation observed in the PSl^−/−brain, as suggested by the finding that neural progenitor cells prematurely differentiate into postmitotic neurons in Hes5^−/−mice (Ohtsuka et al., 1999). The fact that expression of Hesl is unaltered in the PSl^−/−mouse suggests that Hesl and Hes5 expression might be regulated via different mechanisms.

It has been postulated in Drosophila that Notch and its ligand Delta are linked by a regulatory negative feedback loop under the transcriptional control of the Enhancer-of-split and achaete-scute complex gene products (Heitzler et al., 1996; Heitzler and Simpson, 1993, 1991). Expression of Dlll, which encodes a ligand of Notch1, is upregulated in the Notch1^−/− mouse embryo (de la Pompa et al., 1997). We therefore examined expression of Dlll to determine whether it is upregulated in the PSl^−/−brain. In situ hybridization analysis revealed that Dlll expression is confined to isolated cells in the ventricular zone, consistent with previous reports (Bettenhausen et al., 1995; Henrique et al., 1995). At E11.5, the number of Dlll-expressing cells is increased in the PS1^_/∼ brain in all brain regions examined, including the anterior telencephalon (Fig. 6Aa,b) and diencephalon (Fig. 6Ac,d). Quantitative comparison of Dlll-expressing cells in the telencephalon revealed a 40% increase in the density of Dlll-expressing cells in the PSl^−/−neuroepithelium (Fig. 6B). Northern analysis of total RNA derived from the PSl^-/- and control heads at E11.5 showed a 30% increase in the level of Dlll transcripts in the PSl^−/−brain relative to the control (Fig. 6C). These results indicate that lack of PS1 results in an increase in the number of Dlll-expressing cells and the overall expression level of Dlll transcripts, providing further evidence for the downregulation of Notch signalling.

To determine whether PS1 regulates Notch signalling at the level of transcription, translation and/or post-translational activation, we examined the PSl^−/−and control brains for differences in the levels of Notchl expression and subcellular localization (Fig. 7). In situ hybridization analysis of comparable brain sections of the PSl^−/−and littermate control at E11.5 revealed no significant differences in the expression pattern and the level of Notchl transcripts in the anterior telencephalon (Fig. 7Aa,b), ganglionic eminence (Fig. 7Ac,d) and diencephalon (Fig. 7Ae,f). Northern analysis confirmed that the level of the Notchl transcript is unaltered in the PS1^−/− brain (Fig. 7B). Immunohistochemical analysis using a polyclonal antiserum (Logeat et al., 1998) raised against the ICD of mouse Notch1 showed intense Notch1 immunoreactivity in the ventricular zone, particularly in the ependymal cell layer and the adjacent progenitor cell layers (Fig. 7C). Notch1 immunoreactivity is localized predominantly to the cytoplasm and associated with the plasma membrane, although very low levels of nuclear Notch1 immunoreactivity could be detected in some progenitor cells (Fig. 7C). Comparison of corresponding sections of the PS1^−/− and control brains revealed no differences in the intensity of cytoplasmic and plasma membrane associated Notchl immunoreactivity in the telencephalon and diencephalon (Fig. 7C). These results indicate that Notch signalling is unlikely to be regulated by PSl at the levels of transcription and translation, thus most likely to be regulated at the level of post-translational activation. These findings are consistent with previous studies by us and other groups using cultured PS/^−/− cells, which showed a requirement for PS1 in the normal production of NICD (De Strooper et al., 1999; Song et al., 1999).

Previously, Wong et al. (1997) reported that expression of Notchl and Dlll transcripts is markedly reduced in the presomitic mesoderm of PSl^−/− embryos. We observed increased D lll expression and unchanged Notchl expression in the PSl^−/−brain, however, raising the possibility that expression of Notchl and Dlll is regulated differently in the developing CNS and mesoderm. To address this issue, we examined the expression of Notchl and Dlll in the presomitic mesoderm in the PSl^−/− and littermate control embryos by whole-mount in situ hybridization and northern analyses.

Whole-mount in situ hybridization analysis of PSl^−/− and littermate control embryos at day 9.5 showed that Notchl and Dlll transcripts are expressed at highest levels in the presomitic mesoderm (Fig. 8A), consistent with previous reports (Bettenhausen et al., 1995; Franco Del Amo et al., 1992; Reaume et al., 1992). As we previously described (Shen et al., 1997), the somites in the caudal region of the PSl^−/− embryo are disorganized and the boundaries between these somites are blurred (Fig. 8Ab,d). However, similar levels of Notchl and Dlll transcripts were detected in the presomitic mesoderm, which appears to be at its appropriate position (Fig. 8A). Because whole-mount in situ hybridization does not permit quantitative comparisons of mRNA expression levels, we carried out northern analysis to examine the levels of Notchl and Dlll transcripts. Northern analysis using total RNA derived from the tail buds of PSl^−/− and control embryos at day 11.5, where Notchl and Dlll are expressed at their highest level, revealed similar levels of Notchl transcripts and increased levels of Dlll transcripts in the PSl^−/− embryo relative to the control (Fig. 8B). These results indicate that regulation of Notch signalling does not differ in the CNS and paraxial mesoderm.

Our previous characterization of PSi^−/− mice documented specific defects in central nervous system development, revealing a function for PS1 in the mammalian brain (Shen et al., 1997). Here we characterize the mechanisms underlying the progressive reduction in neural progenitor cells that we observed in the PSi^−/− brain, as depicted schematically in Fig. 9. Analysis of the PSi^−/− brain between E10.5 and 13.5 revealed increased numbers of postmitotic neurons detected by MAP2 (Fig. 1) and TuJ1 (data not shown) immunoreactivity in the telencephalon and diencephanlon. The premature neuronal differentiation in the PSi^−/− brain leads to a reduction in the neural progenitor population and thinning of the ventricular zone. This is particularly evident in the diencephalon, where increased MAP2 immunoreactivity at E10.5 (Fig. 1Ab) is followed by decreased BrdU immunoreactivity at E11.5 (Fig. 2Bf). These results demonstrate that lack of PS1 function leads to an increase in the number of differentiated postmitotic neurons at the expense of progenitor cells during early neurogenesis, resulting in a partial depletion of the neural progenitor population. These observations provide an explanation for the progressive reduction of the neural progenitor population and the subsequent reduction of neuronal population previously observed in the PSi^−/− brain (Shen et al., 1997). Taken together, these findings support a novel role for PS1 in the regulation of neuronal differentiation and a cell-fate decision between neural progenitor cells and postmitotic neurons in the developing brain.

We further investigated the role of PS1 in neuronal proliferation, survival and migration during early neural development. The similar proliferation rates of neural progenitor cells in the PSi^_/∼ and control brains, as measured by the percentage of BrdU-labelled cells to total progenitor cells, suggest that PS1 is not involved in the regulation of neural progenitor proliferation (Fig. 2 and Table 1). Our analysis of the PSi^_/∼ brain at E11.5 and 12.5 also shows that lack of PS1 does not result in increased apoptotic cell death at these early neural developmental stages (Fig. 3 and Table 2). Severe neuronal loss was, however, observed in specific regions of the PSi^_/∼ telencephalon beginning at E16.5 (Shen et al., 1997). Thus, it remains possible that PS1 influences neuronal survival during later stages of neurogenesis.

An involvement of wild-type and FAD-linked mutant presenilins in regulation of apoptosis has been suggested by previous investigations in various cell culture systems (Deng et al., 1996; Guo et al., 1996; Kim et al., 1997; Loetscher et al., 1997; Roperch et al., 1998; Vito et al., 1996; Wolozin et al., 1996; Zhang et al., 1998). The Notch signalling pathway has also been implicated in the regulation of apoptosis, with Notch activity appearing to promote cell survival. Loss of Notch activity in Drosophila is spatiotemporally correlated with increased levels of apoptosis (Kim et al., 1996; Van Doren et al., 1992). In mice lacking Notch2 function, increased apoptosis was observed in neural tissues (Hamada et al., 1999). Finally, a connection between Presenilin and Notch activity in apoptosis was suggested by the finding that mutant flies either overexpressing or lacking PS exhibited increased levels of apoptotic cell death; this phenotype was suppressed in both cases by expression of constitutively active Notch (Ye and Fortini, 1999). It is not yet clear whether the loss of neural progenitor cells and neurons observed in the PSi^−/− brain is cell-autonomous, and whether downregulation of the Notch1 signalling activity plays a direct role. We are currently investigating the mechanism underlying this region-specific neuronal loss by generating conditional PS1^_/∼ mice lacking PS1 function specifically in neuronal cell types.

The results of our neuroanatomical and BrdU birthdating analysis support a role for PS1 in the regulation of neuronal migration during cortical neurogenesis. Examination of the PS1^−/− brain stained with Hematoxylin and Eosin suggested disruption of the normal architecture of the developing forebrain, with an indistinct boundary between the ventricular and intermediate zones and thinning of the cortical plate (Fig. 4). The BrdU birthdating study revealed disorganization of postmitotic neurons in the developing cortical plate (Fig. 4). However, our birthdating analysis was limited to PS1^−/− brains at E14.5, due to severe cerebral hemorrhage often associated with PS1^−/−brains at later developmental stages. Therefore, whether cortical lamination defects in the PS1^−/− brain are similar to those seen in the reeler mouse is unclear. Extensive BrdU birthdating analysis on PS1 conditional knockout mice, in which PS1 inactivation is restricted to the neuronal population, should address this question. The mechanism by which PS1 regulates cortical layer formation and neuronal migration is unknown. The perturbed laminar organization may be associated with the progressive loss of Cajal-Retzius neurons observed in the PS1^−/− brain; however, the cause of the progressive loss of these neurons is unclear. It is also possible that PS1 is an upstream regulator or a downstream target of Cdk5/p35 or mDab1, both of which play an important role in neuronal migration (Chae et al., 1997; Howell et al., 1997; Ohshima et al., 1996). Reduced Notch signalling in PS1^−/− mice may also contribute to the neuronal migration defects, as suggested by the findings in Drosophila that Notch genetically interacts with Abl and its intracellular domain binds directly to Dab in vitro (Giniger, 1988).

To understand the mechanism by which PS1 controls neuronal differentiation during neurogenesis, we examined the expression of genes involved in the Notch signalling pathway: Notch1, Dll1 and the Notch downstream effector genes Hes1 and Hes5. Previous studies have suggested a connection between Notch signalling and the regulation of neuronal differentiation. Premature neuronal differentiation has been observed at E10.5 in mutant mice lacking Hes1, Hes5, or both Hes1 and Hes5 (Ishibashi et al., 1995; Ohtsuka et al., 1999). Continuous overexpression of Hes1 by retroviral infection inhibits neuronal differentiation (Ishibashi et al., 1994). Here we have shown that expression of Hes5 transcripts is downregulated in the PS1^−/− brain, particularly in the ganglionic eminence and diencephalon, whereas expression of Hes1 is unaffected (Fig. 5). These results suggest that Notch signalling is compromised in the absence of PS1, as evidenced by the downregulation of Hes5 expression. In addition, regulation of Notch downstream target genes appears to be complex, since PS1 regulates specifically the expression of Hes5 but not Hes1. It is not yet clear whether Hes5 is the only downstream target of PS1 in the regulation of neuronal differentiation during neurogenesis. The extent of the premature differentiation of neural progenitor cells observed in the PS1^−/− and Hes5^−/−brains at E10.5 appears to be similar (Fig. 1; Ohtsuka et al., 1999). Examination of the PSi^−/− and Hes5^−/− brains at later embryonic stages for differences in neuronal differentiation and neurogenesis might reveal whether PS1 controls neuronal differentiation and cell fate decision between progenitor cells and postmitotic neurons through the regulation of Hes5.

Further supporting evidence for the downregulation of Notch signalling includes the upregulation of Dili expression in PSi^−/− mice. Our results demonstrated that the level of Dili transcripts is elevated in the brain and the presomitic mesoderm of PS1^−/− mice and more Dili-expressing cells are present in the PS1^−/− brain (Figs 6, 8). Studies inXenopus and the chick embryo have shown that Dili homologues are expressed in the prospective neurons and their expression precedes expression of early neuronal markers (Chitnis et al., 1995; Henrique et al., 1995). An increase in Dlli -expressing cells is consistent with the presence of more differentiated neurons identified in the PSi^−/−brain.

Lack of PS 1 function was previously reported to result in reduced transcription of Notchi and Dlli in the presomitic mesoderm of the PSi^−/− embryo, suggesting a role for PS1 in the regulation of Notchi and Dlli at the transcriptional level (Wong et al., 1997). We detected unaltered levels of Notchi transcripts (Figs 7, 8) and elevated levels of Dlli transcripts (Figs 6, 8) both in the PSi^−/− embryonic brain and presomitic mesoderm. Furthermore, we observed no differences in the intensity of cytoplasmic and plasma membrane associated Notch1 immunoreactivity in the PSi^−/− and control brains (Fig. 7). Therefore, PS1 is unlikely to be involved in the regulation of Notch1 transcription and translation in the developing mammalian brain and paraxial mesoderm. However, PS1 is involved in the activation of Notch signalling, as indicated by the alterations of Hes5 and Dlli expression in PSi^−/− mice (Figs 5, 6, 8). These results are consistent with recent studies showing that similar levels of furin-cleaved Notch1 fragment were detected in the PSi^−/− and control mouse brains by immunoprecipitation-western analysis (De Strooper et al., 1999). Analysis of the proteolytic processing of truncated Notch1 proteins in cultured cells derived from PSi^−/− mice has shown that PS 1 is required for efficient release of the Notch1 ICD (De Strooper et al., 1999; Song et al., 1999). The present study supports such a role for PS1 in the regulation of Notch signalling, and further provides evidence that the reduced Notch processing observed in PSi^−/−cells is functionally significant in neural development.

Finally, our results indicate an important difference in the consequences of reduced Notch signalling during neurogenesis in Drosophila and mice. In Drosophila neurogenesis, Notch controls a cell-fate decision between two cell types produced from a multipotent common precursor, promoting epidermal production at the expense of neuronal production. Loss of function mutations in Notch thereby lead to excessive neuronal production (Artavanis-Tsakonas et al., 1999). Our findings suggest that Notch1 regulates a cell-fate choice between neural progenitor cells and differentiated neurons early in murine neurogenesis, promoting regeneration of neuronal precursor cells at the expense of differentiation of postmitotic neurons. Therefore, although Notch functions to suppress the production of postmitotic neurons in both mice and Drosophila, downregulation of Notch activity in mice and Drosophila results in a reduction in neuronal population and a neurogenic phenotype, respectively.

In summary, we demonstrate that lack of PS1 during neural development results in a reduction in Notch signalling activity, as indicated by reduced Hes5 expression and increased Dlli expression, which alters the cell-fate decision of neural progenitor cells in favor of differentiation into postmitotic neurons (Fig. 9). The roles of PS1 and the Notch signalling pathway in the adult brain are currently unknown. A recent report documented the existence of Notch1-expressing neural stem cells residing in the ependymal cell layer of the adult rat brain (Johansson et al., 1999), which suggests that Notch signalling may be involved in the maintenance of a neural stem cell population in the adult brain. Understanding the roles of PS1 and Notch signalling in the adult brain may therefore lead to novel therapeutic strategies for combating the neuronal and synaptic loss that occurs in neurodegenerative illnesses such as Alzheimer’s disease.

We thank Drs Spyros Artavanis-Tsakonas, Bruce Yankner, Dennis Selkoe, Ken Kosik, Chris Walsh, Rudy Tanzi and anonymous reviewers for comments on the manuscript; Drs Ryoichiro Kageyama, Gerry Weinmaster and Randy Johnson, for plasmids; Dr Alain Israel for Notch1 antiserum; and Dr Ken Kosik for the anti-MAP2 antibody. We are grateful to Ricardo Soto and Huakui Yu for their excellent technical assistance. This research was supported in part by the American Federation for Aging Research, the Alzheimer’s Association and NINDS.