During neurogenesis, a subset of progenitor cells undergoes neural differentiation, and these new neurons are thought to feedback and inhibit neighboring progenitor cells from differentiating. This process is called lateral inhibition, and serves to regulate the number of neurons born at a given time and to maintain a pool of progenitor cells for generation of subsequent neurons and glia. Lateral inhibition is mediated at the molecular level by the intercellular Notch signal transduction pathway (Lewis,1996; Lowell,2000). Notch signaling is composed of a cell-surface bound ligand from the Delta/Serrate/Lag (DSL) gene family that binds to its cognate cell-surface bound receptor Notch on a neighboring cell. Binding of a DSL ligand to Notch activates its signaling and results in a series of proteolytic events, ultimately releasing the internal cytoplasmic domain (NICD). Free NICD then forms a transcriptional activation complex with Mastermind and CSL, which in turn activates transcription of Notch target genes. One of the major target gene families activated by Notch signaling is the hairy/enhancer of split (HES) genes, such as Hes1 and Hes5. Hes genes negatively regulate the expression and action of members of the proneural bHLH transcription factor family, such as Ascl1 and Neurog2 (Nelson et al.,2007b; Ohsawa and Kageyama,2008), which induce neural differentiation and specification (Bertrand et al.,2002).
In Drosophila, one function of the proneural bHLH genes is to induce expression of Delta, thereby completing the molecular circuitry underlying the basis of lateral inhibition (Skeath and Carroll,1994; Technau et al.,2006). Several DSL-ligands have been identified in vertebrates, including members of the Delta-like (Dll) and Serrate (also called Jagged) gene family. Three Dll and two Jagged (Jag) genes have been identified in mice, four to five Dll and three Jag genes have been identified in zebrafish, and two Dll and two Jag genes have been identified in chick (Linsdell et al., 1995; Bettenhausen et al.,1995; Henrique et al.,1995; Shawber et al.,1996; Myat et al.,1996; Hayashi et al.,1996; Dunwoodie et al.,1997; Valsecchi et al.,1997; Dornseifer et al.,1997; Appel and Eisen,1998; Haddon et al.,1998; Shutter et al.,2000; Zecchin et al.,2005; Nimmagadda et al.,2007; Pintar et al.,2007). While most studies of vertebrate proneural bHLH genes have focused on their role in neural subtype specification, recent evidence has revealed that at least Ascl1 and Neurog2 can directly regulate Dll1 expression in certain regions of the nervous system (Castro et al., 2007). Thus, the underlying transcriptional network controlling lateral inhibition is evolutionarily conserved.
In the vertebrate retina, Notch and its ligands maintain the progenitor pool during the course of retinal development, and regulate the evolutionary conserved sequence of progenitor cell differentiation into the six types of neurons and one type of glia (Dorsky et al.,1995,1997; Austin et al.,1995; Tomita et al.,1996a; Henrique et al.,1997a; Furukawa et al.,2000; Hojo et al.,2000; Satow et al.,2001; Silva et al.,2003; Takatsuka et al.,2004; Nelson et al.,2006,2007a; Jadhev et al.,2006; Yaron et al.,2006). Notch activity is also necessary during the earliest events in eye specification to expand the population of retinal founder/stem cells in the newly specified eye-field (Onuma et al.,2002; Hatakeyama et al.,2004; Lee et al.,2005). The vertebrate retina expresses several Notch ligands during development. In mice, Jag1 is expressed in the distal optic cup, ciliary epithelium, lens, and in the late postnatal ganglion cell layer, Jag2 is expressed early in the newly differentiated ganglion cell layer, while Dll1 expression is confined to the neuroblastic layer in the central retina (Lindsell et al.,1996; Bao and Cepko,1997; Valsecchi et al.,1997). In the chick, Jag1 (Serrate1) is apparently not expressed, while Dll1 is present in both the peripheral and central retina (Myat et al.,1996; Henrique et al.,1997a). Work in mice has revealed that Jag1 is important for peripheral eye and optic nerve development (Wang et al.,1998; Xue et al.,1999; Kiernan et al.,2007), and studies in chick have shown that Dll1 is important for maintaining the progenitor pool (Henrique et al.,1997a).
Recently, Dll4 expression was reported in both mouse and chick retina (Benedito and Duarte,2005; Nimmagadda et al.,2007). We also observed both Dll1 and Dll4 expression in our microarray analysis of differentiating mouse retinas (Nelson et al.,2007a). To better understand how Dll1 and Dll4 coordinately function within the general framework of the Notch pathway, we first determined their expression pattern with respect to progenitor cells or newly differentiating neurons in the chick, and then how their expression changes during synchronized progenitor cell differentiation. We then compared these expression patterns and kinetics with those of the proneural bHLH genes, revealing that one bHLH in particular, Ascl1, may have a potential role in regulating Delta-like genes in the retina. We found that Ascl1 overexpression leads to an up-regulation of Delta-like gene expression and Notch signaling. These data are the first to link a proneural bHLH gene with the regulation of Delta-like genes and Notch signaling in the retina, and suggest that multiple ligands of Notch maintain the progenitor pool and coordinate retinal histogenesis.
Identification of Chick Delta-Like 4 (Dll4)
Our previous microarray/quantitative polymerase chain reaction (QPCR) analysis revealed that in addition to Dll1, another Delta-like family member, Dll4, was up-regulated early after DAPT (γ-secretase inhibitor N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S- phenylglycine t-butyl ester) -induced differentiation of retinal progenitor cells (Nelson et al.,2007a). We searched the chick expressed sequence tag (EST) databases for clones to chick Dll4 that could be used for in situ hybridization. We used mammalian Dll1 (mouse Dll1, NM_007865; human Dll1, NM_005618), Dll3 (mouse Dll3, NM_007866; human Dll3, NM_016941), and Dll4 (mouse Dll4, NM_019454; human Dll4, NM_019074) sequences to identify several candidate chick clones with greater sequence homology to mammalian Dll4 than chicken Dll1 (NM_204973). Candidate ESTs were sequenced and aligned to the chicken genome (Ensemble Genome Browser, http://www.ensembl.org/index.html; Evolutionary Conserved Regions (ECR) Browser, http://ecrbrowser.dcode.org, Ovcharenko et al.,2004), all of which aligned to the Dll4 locus (XM_421132), indicating that these clones represent a set of overlapping sequences of the same gene. Sequences were assembled and compared with a computationally predicted chick Dll4 coding sequence and the genomic locus (Sequencher, http://www.genecodes.com, Gene Codes), and a full-length chick Dll4 consensus sequence was generated. Clustal-W (http://www.ebi.ac.uk/Tools/clustalw2/index.html, EMBL-EBI) was used to align the peptide sequences of mouse Dll1 (NP_031891), Dll3 (NP_031892.), Dll4 (NP_062327), and human Dll1 (NP_005609), Dll3 (NP_058637), and Dll4 (NP_061947), with chick Dll1 (NP_990304) and the predicted chick Dll4 peptide (XP_421132), revealing that the new chick Delta family member was most similar to mouse and human Dll4 (Fig. 1A). Within the DSL domain, peptide alignment confirms that this new chick Delta-like gene is most similar to mouse and human Dll4, rather than to Dll3 (Fig. 1B), although its sequence diverges somewhat in the C-terminus (Supplementary Figure S1, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat). The similarity in the DSL domain is a key distinction, since recent evidence demonstrates that mammalian Dll1 and Dll4 can bind and activate Notch at the cell surface, but that the DSL domain of Dll3 cannot (Ladi et al.,2005; Geffers et al.,2007), and that Dll3 is not actually present at the cell surface (Geffers et al.,2007). Thus, the chick Dll4 is likely a functional activator of the Notch signaling pathway.
We performed a similar search to specifically identify a chick Dll3 homolog. While further analyses of multiple whole-genome alignments revealed a syntenic arrangement of the Dll loci across multiple species, we found that the Dll3 locus was not apparent in the chick genome (Supplementary Figure S2). This does not rule out the possibility that a Dll3 ortholog exists in the chick, however, a more thorough phylogenic analysis failed to reveal a chicken Dll3 (Pintar et al.,2007).
Dll1 Is Expressed in Mitotically Active Progenitor Cells, While Dll4 Is Expressed Within Newly Differentiating Neurons
Because both Dll1 and Dll4 are present in the chick retina, we sought to determine which retinal cell types express these genes. Previous observations revealed that Dll1 was not expressed in differentiating Islet1+ neurons, suggesting that Dll1 was expressed during an earlier stage of neuronal differentiation than that marked by Islet1 (Henrique et al.,1997a). Neurofilament (NF) is among the earliest markers of neuronal differentiation in the chicken retina, and is observable during the final mitosis of differentiating progenitors at the ventricular (scleral) surface (Supplementary Figure S3; McCabe et al.,1999). We used in situ hybridization followed by NF immunolabeling to re-examine whether Dll1 is expressed by newborn neurons. We found that surprisingly few Dll1+ cells were NF+ (12%, 83/688 total Dll1+ cells), even at the ventricular surface (Figs. 2A–C, 4A), suggesting that Dll1-expressing cells were progenitors. Indeed, Henrique et al. (1997) reported that a small fraction of Dll1-expressing cells incorporated bromodeoxyuridine (BrdU) in a 30-min pulse (13%, Henrique et al.,1997a). We found that over half of the Dll1+ cells had incorporated BrdU (Figs. 2D–F, 4A: 52%, 452/875 total Dll1+ cells) with a slightly longer pulse (1–2 hr).
We found a different pattern of expression for Dll4. In situ hybridization for Dll4 expression (Fig. 3A) followed by immunolabeling for BrdU incorporation (Fig. 3B) and NF (Fig. 3C) revealed that the majority of Dll4+ cells were NF+ (Fig. 4A: 40%, 117/292 total Dll4+ cells), while a smaller proportion were BrdU+ (Fig. 4A: 30%, 89/292 total Dll4+ cells). Dll4 was expressed in NF+ cells at the ventricular surface, but was down-regulated as neurons differentiate and migrate to the ganglion cell layer (Fig. 3C,E).
To more precisely examine the relationship between Notch, Notch-ligand expression, and progenitor cell differentiation status, we took advantage of the ability in chick to identify the newest born neurons, which can be identified by the up-regulation of NF during their final mitosis at the ventricular surface (Waid and McLoon,1995; McCabe et al.,1999). Although Notch activity is down-regulated just before neural differentiation (Nelson et al.,2006), newborn NF+ neurons at the ventricular surface have yet to down-regulate Notch1 expression: ∼26% of Notch1+ cells were also NF+ (25/98 Notch1+ cells at the ventricular surface, Supplementary Figure S3) representing the approximate number of progenitor cells undergoing differentiation at the ventricular surface at this stage of retinogenesis. Of interest, we observed a similar number of Dll4-expressing cells at the ventricular surface as the number of progenitor cells undergoing differentiation. The majority of these Dll4+ cells were also NF+ (Fig. 4B: 80%, 27/36 Dll4+ cells at the ventricular surface) compared with the number of Dll1+ cells at this surface that were NF+ (Fig. 4B: 25%, 24/97 Dll1+ cells at the ventricular surface). Thus, this analysis demonstrates that Dll4, rather than Dll1, is up-regulated in newly differentiating neurons.
Dll1 and Dll4 Are Expressed During Different Phases of Progenitor Cell Differentiation
The difference in Dll1 and Dll4 expression patterns suggests that these two different Notch ligands might function at different stages in the progenitor cell differentiation program. To further analyze the sequence of expression of these two genes, we used a method to synchronize the differentiation of progenitor cells (Nelson et al.,2007). We treated retinas with the soluble presenilin/γ-secretase inhibitor DAPT, which rapidly blocks the presenilin/γ-secretase-dependent cleavage of the Notch intracellular domain and subsequent canonical Notch signaling (Geling et al.,2002; Michelli et al.,2003). Blockade of Notch signaling in progenitor cells commits them to differentiate, thereby synchronizing cell behavior and enriching for molecular changes in this population. QPCR is used to analyze ensuing changes in gene expression levels over time, which reveals progenitor cells' normal differentiation program.
In a previous study, we found that interruption of Notch signaling led to a rapid decline in Hes5 expression, and an immediate and transient up-regulation of the proneural bHLH genes Ascl1 and Neurog2 (Nelson et al.,2007). This was followed by transient increases in other bHLH genes, such NeuroD4 and Atoh7, and ultimately the concomitant down-regulation of progenitor genes such as Chx10, Pax6, and Pea3, and up-regulation of markers of terminal differentiation like RXR-γ, visinin, Isl1, and class-III β-tubulin (Tuj1; Nelson et al.,2007).
When we analyzed changes in Dll1 and Dll4 expression levels with this same approach, we found that Dll1 expression was significantly increased by as little as 3 hr of Notch inactivation, peaked by 6 hr, and was down-regulated after 12 hr (Fig. 5A). By contrast, changes in Dll4 expression occur later. Dll4 is slightly up-regulated by 6 hr, peaks at 12 hr, and declines to that of control levels by 48 hr (Fig. 5B). These data are consistent with the hypothesis that Dll1 and Dll4 are used sequentially during the molecular network regulating neurogenesis.
Dll1 and Dll4 Expression Correlates With Different Proneural bHLH Transcription Factors
The fact that Dll1 and Dll4 peak at different times in the differentiation process suggested to us that they might be driven by different proneural bHLH transcription factors. To determine which proneural genes correlate best with Dll1 and Dll4, we carried out in situ hybridizations for Ascl1 (Cash1), Neurog2 (Ngn2), NeuroD4 (NeuroM), and Atoh7 (Cath5) followed by immunolabeling for BrdU incorporation and NF (we now use the new conventional naming system for these genes). Ascl1 expression was observed in clusters of BrdU+ cells, but not in NF+ neurons (Fig. 6A), confirming Jasoni et al. (1994). Similarly, Neurog2 expression was observed in BrdU+ progenitor cells, and not in NF+ neurons (Fig. 6B). Neurog2+ progenitors seemed to be less clustered than Ascl1+ progenitors. By contrast, NeuroD4 and Atoh7 were expressed in cells that also labeled for NF, many of which at this stage would be retinal ganglion cells (Fig. 6C,D, respectively; Prada et al.,1991). We quantified the double-labeled cell populations (Fig. 6E). The majority of Ascl1+ (83%, 112/135 cells, Fig. 6A) and Neurog2+ (65%, 111/171 cells, Fig. 6B) -expressing cells incorporated BrdU within a 1- to 2-hr pulse (Fig. 6G), similar to Dll1-expressing cells. However, few NeuroD4 (16%, 12/74 cells, Fig. 6C) or Atoh7 (10%, 27/212 cells, Fig. 6E) -expressing cells had incorporated BrdU (Fig. 6G). Instead, the majority of NeuroD4 (62% 111/179 cells, Fig. 6D) and Atoh7 (58%, 203/352 cells, Fig. 6F) -expressing cells were labeled with NF, compared with fewer Ascl1 (4%, 6/135 cells, Fig. 6A) and Neurog2 (8%, 13/171 cells, Fig. 6B) -expressing cells (Fig. 6G). It is also noteworthy that NeuroD4 and Atoh7 expression becomes down-regulated in maturing ganglion cells after they had migrated to the ganglion cell layer.
As noted above, we can synchronize the differentiation process by blocking the Notch pathway with DAPT (Nelson et al.,2007). In a previous study, we found that Ascl1 expression was rapidly up-regulated within 3 hr, and peaked after 6 hr of DAPT treatment. Although Neurog2 expression was also rapidly up-regulated, its expression peaked later at 12 hr, similar to the expression profile of NeuroD4, NeuroD1, and Atoh7. As described above, the expression profile of Dll1 following DAPT-treatment most closely follows the profile of Ascl1, while Dll4 more closely corresponds to that other the other downstream bHLHs (Fig. 6F). These expression profile kinetics support the double-labeling studies, and together suggest a potential regulatory role for Ascl1/Dll1, while one or more of the other proneural bHLH transcription factors could potentially regulate Dll4 expression.
Ascl1 and Dll1 Expression Precedes Expression of Dll4 and the Other Proneural bHLH Genes
We further explored the relationship between Delta-like and proneural bHLH gene expression by taking advantage of the central-to-peripheral gradient in maturation of the retina to analyze their onset of expression (Prada et al.,1991). Within the embryonic avian eye, a non-neurogenic peripheral domain is specified early and gives rise to the iris and ciliary epithelium (CE), while a neurogenic central domain gives rise to the mature neural retina (NR; Kubota et al.,2004). Juxtaposed between these domains is a presumptive preneurogenic zone that likely gives rise to the ciliary marginal zone (CMZ; see Supplementary Figures S3, S4), which contains a population of persistent, stem-like progenitor cells (Moshiri et al.,2004; Reh and Fischer,2006). The CE and presumptive CMZ can be visualized by using a combination of markers such as Coll-IX, Lef1, and NF (Supplementary Figures S3, S4; Kubo et al.,2003; Kubota et al.,2004; Dias de Silva et al.,2007).Within the neural retina, one can observe the onset of neural differentiation marked by the edge of NF immunolabeling (McCabe et al.,1999), preceding Isl1 expression (Supplementary Figure S3), another marker that has been used to identify differentiating neurons (Henrique et al.,1997a).
To determine which Dll and proneural bHLH gene(s) correlated with the onset of neurogenesis, we examined whether their expression extended beyond the margin of the neural retina into the presumptive CMZ and CE. We found that Dll1 expression likely extended through the preneurogenic region of the presumptive CMZ, and into the non-neurogenic region of the presumptive CE (Fig. 7A), similar to a previous observation (Henrique et al.,1997a). Furthermore, while Dll1 expression seemed somewhat diffuse in the most peripheral regions, its expression became restricted to isolated cells just ahead of the front of neural differentiation in the presumptive CMZ (Fig. 7). In contrast, Dll4 expression almost exactly coincides with the region of neural differentiation (Fig. 7B). Analysis of the proneural bHLH genes revealed that only Ascl1 was expressed ahead of the front of neural differentiation (Fig. 7C), likely extending through most of the Dll1 domain, while Neurog2, NeuroD4, and Atoh7 were restricted to the region of neural differentiation (Fig. 7D–F). These data also suggest a potential role for Ascl1 as a candidate regulator of Dll1 in the chick, whereas Neurog2, NeuroD4, and/or Atoh7 are better candidates for regulators of Dll4. Expression of Ascl1 and Dll1 in the preneurogenic zone suggests that Notch signaling is active in the region where the CMZ will ultimately form. Indeed, we found that Notch1 expression extended through this specific region, and interestingly, that Hes1 expression was increased in the prospective CMZ compared with the neural retina, while strong Hes5 expression was restricted to a region of neural differentiation (Supplementary Figure S3). It is also interesting that the highest levels of Dll1 and Hes1 expression are actually observed in the most peripheral non-neurogenic zone of the presumptive CE, where Notch1 expression is not observed (Fig. 7A; Supplementary Figure S3; Henrique et al.,1997a), indicating that potentially a different Notch family member may be used in this domain. The expression patterns of all of these markers are summarized with respect to these domains in Figure 8.
Increased Ascl1 Expression Up-regulates Delta-like and Hes5 Gene Expression
The expression analyses described above indicate a likely specific relationship between Ascl1 and Dll1 in the preneurogenic zone of the presumptive CMZ. However, Ascl1, Neurog2, Dll1, and Dll4 are all expressed in progenitor cells in the central neurogenic zone of the presumptive retina, although Dll4 expression also marks the newly differentiating neurons. To better understand the relationship between Delta-like and proneural bHLH genes, we tested whether Ascl1 or Neurog2 overexpression specifically influenced Dll1 or Dll4 gene expression in the central neurogenic zone of the presumptive retina. We cotransfected E4.5 retinal explants with Ascl1, Neurog2, or CXM (control Myc epitope) expression plasmids, along with expression plasmids for their cofactor E12, and GFP to mark the transfected region (Nelson et al.,2006;2007b). Explants were cultured for 24 hr to allow ectopic DNA expression, and similar-sized transfected regions were collected and individually prepared for QPCR. We found that both Dll1 and Dll4 expression levels were up-regulated in Ascl1 transfected regions compared with CXM controls, while Neurog2 transfection appeared not to change Dll1 or Dll4 levels (Fig. 9). Increased Dll1 and Dll4 gene expression levels might also lead to increased levels of Notch signaling activity, because both of these ligands would likely activate Notch receptors present in adjacent cells. We found that Hes5 gene expression levels were also up-regulated in Ascl1 transfected regions compared with CXM control and Neurog2 transfections (Fig. 9), consistent with this hypothesis. These gain-of-function results demonstrate that, in the central neurogenic zone of the presumptive retina, Ascl1 can positively influence both Dll1 and Dll4 expression and hence Notch signaling activity levels. Thus, Ascl1 may play a more general role by initiating and/or regulating neurogenesis in the different zones of the embryonic eye.
In this report, we identify chicken Dll4, and describe its expression and relationship with Dll1 and the proneural bHLH genes in the retina. Dll4 is expressed in newly differentiating neurons compared with Dll1, which by contrast, is primarily expressed in mitotically active progenitor cells. We also describe that expression of several proneural bHLH genes distinguishes progenitor cells from differentiating neurons, revealing a potential upstream role for Ascl1 in the regulation of Delta-like genes and Notch signaling activity, which were confirmed by overexpression assays. These data suggest that the current model of Notch signaling in the retina may be too simple. We propose that Dll4 mediates lateral inhibition between newborn neurons and progenitor cells, while Dll1 mediates mutual inhibition among progenitor cells themselves.
Different Roles for Dll1 and Dll4 in Retinal Development?
It is well established that Notch signaling serves to maintain the progenitor pool during the period of retinogenesis, and regulate the conserved sequence of neuronal and subsequent glial differentiation (Dorsky et al.,1995,1997; Austin et al.,1995; Tomita et al.,1996a; Henrique et al.,1997a; Furukawa et al.,2000; Hojo et al.,2000; Satow et al.,2001; Silva et al.,2003; Takatsuka et al.,2004; Nelson et al.,2006,2007; Jadhev et al.,2006; Yaron et al.,2006). Early experiments in Xenopus and chicken showed that Dll1 plays a fundamental role in this regard, serving as an activating Notch ligand that signals to progenitors, maintaining a pool for later retinal differentiation events (Dorsky et al.,1997; Henrique et al.,1997a). These data led to the model that Dll1 was up-regulated in differentiating neurons and functioned to laterally inhibit adjacent progenitors from differentiating by activating their Notch receptor, an extension of the lateral inhibition model from Drosophila. However, the presence of Dll4 complicates this simple model (Benedito and Duarte,2005; Nimmagadda et al.,2007; Nelson et al.,2007; this report).
To understand the relationship between Dll1 and Dll4 in the retina, we systematically compared their expression patterns with that of newly differentiating RGCs and proliferating progenitor cells. While initial descriptions of general Dll4 expression pattern identified early retinal expression in mouse and chick (Benedito and Duarte,2005; Nimmagadda et al.,2007), the cell types expressing Dll4 were not characterized. We took advantage of NF and BrdU immunoreactivity after in situ hybridization to identify the newly differentiating neurons and proliferating progenitor cells (as in Nelson et al.,2006). NF is among the earliest markers of neural differentiation in the avian retina and can be observed within the final mitosis of progenitors undergoing neural differentiation at the ventricular (scleral) surface (McCabe et al.,1999). We found that, while Dll4 expression was observed in some progenitor cells, its expression also matched the pattern of newly differentiating neurons, compared with Dll1, which was primarily expressed in progenitor cells. These data suggest that Dll4 mediates lateral inhibition between newly differentiating neurons and neighboring progenitor cells, while Dll1 mediates mutual inhibition within the progenitor pool itself.
Whether progenitor cells expressing Dll1 are selected from the pool for subsequent cell fate choices or remain cycling is not clear. We found that the peak in Dll1 expression following DAPT-induced progenitor differentiation occurs before that of Dll4, consistent with the conclusions from the BrdU labeling studies: that Dll1 is primarily expressed by mitotically active progenitors, while Dll4 is expressed in newly postmitotic neurons. However, Notch activity levels in progenitors normally oscillate during the cell cycle (Tokunaga et al.,2004; Nelson et al.,2007), and Dll levels could do so as well. Thus, an alternative hypothesis would be that progenitors mutually inhibit themselves through reciprocal oscillations in Notch and Ascl1/Dll1 activity levels. Extrinsic signals could increase Ascl1/Dll1 activity beyond a critical threshold in subsets of these progenitors, activating the neurogenic cascade of downstream bHLH and Dll4 genes that would promote their differentiation and laterally inhibit neighboring progenitors by driving high levels of Notch signaling. It may also be that two separate populations of Dll1 and Dll4 mitotically active progenitors exist, but that only Dll4 progenitors differentiate early while Dll1 progenitors differentiate later. In either case though, progenitors express a Delta gene and signal back to remaining progenitors in a mutual inhibition mode.
In addition to these two Notch ligands, there are potentially several other ligands that could activate this pathway in the retinal progenitors. A member of the Jagged family of Notch ligands (Jag2) is expressed in the postmitotic ganglion cell layer at early stages (Valsecchi et al.,1997). Furthermore, an additional Notch receptor (Notch3) is also expressed in the mouse retina (Lindsell et al.,1996; Kitamoto et al.,2005). Thus, in the neurogenic zone of the presumptive retina, multiple types of Notch-mediated signaling pathways potentially operate cooperatively during development to maintain progenitors and regulate retinal histogenesis (Nelson et al.,2007b).
Distinct Domains of Expression of Notch Signaling Components Correlate With Non-neurogenic, Preneurogenic, and Neurogenic Zones in the Chick Retina
By double- and triple-labeling individual retinal sections with combinations of regional-specific and cell type-specific markers, we were able to define distinct non-neurogenic (presumptive ciliary epithelium), preneurogenic (presumptive CMZ), and neurogenic (presumptive retina) domains in the developing eye. We found that Ascl1 and Dll1 are expressed in the preneurogenic domain of the presumptive CMZ, while the other proneural bHLH genes and Dll4 are restricted to the neurogenic zone of the presumptive retina. Notch1 expression extends through both the neurogenic and preneurogenic zones. High levels of Hes1 expression were observed in the preneurogenic zone, while Hes5 expression was restricted to the neurogenic domain. These results are reminiscent of the spatial organization observed in the developing Xenopus eye (Perron et al.,1998). Retinogenesis in Xenopus occurs by the sequential addition of differentiated retinal neurons from the CMZ, creating a spatial gradient of temporal development (Perron et al.,1998). Using double-labeling techniques, these authors demonstrated expression of Ascl1, Dll1, Notch1, and Hes genes (ESR1/3) in the most peripheral domain, while potential downstream bHLH genes, such as Atoh7, NeuroD4, and NeuroD1, were expressed more centrally (Perron et al.,1998). Thus, evolutionarily conserved expression of Ascl1/Dll1/Notch1/Hes-family genes specifically defines a preneurogenic domain of cells that prefigures a region of persistent progenitors in both frogs and chicks (Fig. 8 and Perron et al.,1998).
The pattern of expression of proneural genes and Notch signaling components in distinct zones suggests that the preneurogenic zone may be the source of the CMZ in posthatch chicks. The domains of expression in the anterior ocular epithelium are already distinct before the time when we have carried out our analysis. Kubota et al. (2004) and Dias de Silva et al. (2007) show that the ciliary epithelium becomes specified quite early, by the optic cup stage. Kubota et al. (2004) show that by colabeling with Coll-IX and NF on the same section that a nonlabeled gap exists between the presumptive ciliary epithelium and neural retina at these early stages. This region, which corresponds to the preneurogenic zone, in which Ascl1, Hes1, and Dll1 are highly expressed, likely contains cells that will contribute to the retina during the embryonic phase (and, therefore, resemble the early optic vesicle cells), as well as cells that will persist into posthatch retina as the ciliary marginal zone. The anatomical location of these cells and the continued expansion of the retinal neuroepithelium make it very likely that both types of cells are present in this zone. In principle, there may be no difference between the cells of the very early optic vesicle and the presumptive CMZ, because both must contain the most primitive progenitors or stem cells of the retina.
A similar model was recently proposed for the stem zone located in the caudal region of Henson's node (Akai et al.,2005). Intriguingly, another member of the proneural bHLH transcription factor family Ascl4 is strongly and uniformly expressed within mitotically active stem zone cells (Henrique et al.,1997b; Storey et al.,1998), along with Dll1, Notch1, and Hes5 (Akai et al.,2005; Hämmerle and Tejedor,2007). Ascl4/Dll1-mediated Notch activity results in mutual inhibition between stem cells, which maintains proliferation in the spinal cord stem zone (Akai et al.,2005). As spinal cord stem cells move out of the caudal stem zone and experience different extrinsic signals, they up-regulate different proneural genes such as Neurog1 and Neurog2 (Akai et al.,2005). This transition switches the output of Notch signaling from mutual inhibition between progenitors to lateral inhibition, mediating the transition to neurogenesis (Hämmerle and Tejedor,2007).
Recently, Gaiano and colleagues reported that differences in the level of Notch signaling discriminate between neural stem cells and more limited intermediate neural progenitor cells in the developing telencephalon (Mizutani et al.,2007). We observed a similar difference in the levels of Hes expression in the different zones of the retinal epithelium: Hes1 is expressed more highly in the presumptive CE (non-neurogenic) and CMZ (preneurogenic) than in the retina; Hes5 is expressed exclusively in the presumptive retina. Within this neurogenic zone, not only are different levels of active Notch signaling observed during the cell cycle, with higher levels in progenitors during S-phase compared with lower levels during mitosis (Tokunaga et al.,2004; Nelson et al.,2007), differences between Hes1 and Hes5 expression patterns are also apparent (Nelson et al.,2006). These differences may reflect differences in progenitor potential, such that Hes1+ progenitors are more “stem-like” than Hes5+ progenitors. In this regard, it is interesting that Hes1 mutation leads to microphthalmia, while Hes5 mutation only results in a partial loss of Muller glia (Tomita et al.,1996a; Hojo et al.,2000). Because the context of Notch signaling is critical to its output, it may be that different Notch-ligands can activate Notch differently, contributing to differences in downstream effects.
Proneural bHLH Transcription Factors Distinguish Progenitor Cells From Newborn Neurons
To understand the relationship between Dll1/4 and their potential upstream transcriptional activators, the bHLH transcription factors, we re-analyzed the expression of proneural genes together with immunolabeling for progenitor or neuronal markers. We confirmed that Ascl1 is expressed in mitotically active progenitor cells (Jasoni et al.,1994; Jasoni and Reh,1996; Marquardt et al.,2001; this report). We also found that Ascl1 is expressed in the preneurogenic zone, potentially marking the least mature retinal progenitors (Perron et al.,1998; Matter-Sadzinski et al.,2001,2005); while Ascl1 expression is highest during later stages of chick and mouse retinal development, a low level is observed even at the earliest optic cup stages (Philips et al.,2005; Lee et al.,2005). We also found that Neurog2 expression is restricted to the central neurogenic zone, consistent with previous observations (Yan et al.,2001; Marquardt et al.,2001; Matter-Sadzinski et al.,2001,2005; Le et al.,2006; Ma and Wang,2006). Interestingly, Ascl1 and Neurog2 progenitor cells in the central neurogenic zone of the chick retina exhibit distinct expression patterns: Ascl1 progenitors are more clustered than Neurog2 progenitors, confirming Jasoni et al. (1994). In the mouse retina, Ascl1 and Neurog2 are also expressed in distinct subpopulations of progenitor cells, (Marquardt et al.,2001), confirming that retinal progenitors are heterogeneous with respect to at least two upstream proneural bHLH genes (Jasoni and Reh,1996). Similar to Ascl1, Neurog2 is also expressed from the earliest stages of mouse retinal development in proliferating progenitor cells that can differentiate into all types of retinal neurons (Lee et al.,2005; Le et al.,2006; Ma and Wang,2006).
Our analysis of other bHLH transcription factors confirms and extends the results of earlier reports. Expression of NeuroD4, Atoh7, and NeuroD1 is confined to the central neurogenic zone (this report; Roztocil et al.,1997; Liu et al.,2001; Matter-Sadzinski et al.,2001,2005; Ma et al.,2005). Within this neurogenic zone, we found that few NeuroD4- and Atoh7-expressing cells incorporated BrdU, confirming previous observations (Matter-Sadzinski et al.,2001,2005; Ma et al.,2005). NeuroD4 and Atoh7 expression begins early at E2 and peaks at E6 (Roztocil et al.,1997; Matter-Sadzinski et al.,2001,2005). Although Matter-Sadzinski et al. (2005) argue that BrdU Atoh7-expressing cells must be progenitors in a different phase of the cell cycle, we find that both NeuroD4- and Atoh7-expressing cells are already postmitotic NF+ neurons, many of which would be retinal ganglion cells at E4.5 (Fig. 6). These results are in excellent agreement with the birthdating studies in the chick demonstrating 10% of retinal ganglion cells left the cell cycle at E2, but that their peak period of differentiation is later from embryonic day (E) 5 to E7, along with most of the other retinal neuronal cell types (Prada et al.,1991). In the Xenopus retina, NeuroD4 and Atoh7 expression commences in the most central neurogenic zone of the CMZ, along with NeuroD1 (Kanekar et al.,1997; Perron et al.,1998). Although some NeuroD4- and Atoh7-expressing cells were observed to have incorporated BrdU, they did not express Notch1, and were thus likely to be newly differentiating retinal neurons (Perron et al.,1998). In the mouse, NeuroD4 is expressed in the early retina, detected in a band of cells positioned near the ventricular zone, and then maintained in adult inner retinal neurons (Takebayashi et al.,1997). Atoh7 is also expressed early during mouse retinal development. Although there has been some disagreement over whether Atoh7 is expressed exclusively in mitotically active progenitors or in postmitotic neurons, there is an emerging consensus that in mouse, its expression is in a population of postmitotic “transition” cells that can subsequently differentiate into multiple types of retinal neurons (Brown et al.,1998,2001; Wang et al.,2001; Yang et al.,2003; Mu et al.,2005; Brzezinski,2005; Le at al.,2006). Thus, these observations altogether demonstrate that an evolutionarily well-conserved pattern of proneural bHLH gene expression distinguishes proliferating progenitor cells from differentiating neurons in the retina.
Relationship Between Dll Genes and Proneural bHLH Transcription Factors
To begin to understand the relationship between Notch-ligands and proneural bHLH genes in the retina, we compared their expression with combined in situ/immunolabeling and reverse transcriptase (RT) -PCR during DAPT-induced progenitor cell differentiation. Dll1 is expressed primarily in BrdU+ cells and its gene expression was up-regulated early in the process of differentiation, consistent with its expression in progenitor cells. By contrast, Dll4 was expressed in NF+ cells and up-regulated later than Dll1, consistent with its expression in newly differentiating neurons. Comparing the kinetics of expression of bHLH genes with those of Dlls, we find that the expression kinetics for Dll1 most closely matches that of Ascl1, while the expression kinetics for Dll4 matches that of proneural genes expressed later in the differentiation process. A relationship between Dll1 and Ascl1 is further strengthened by the observation that Ascl1 and Dll1 are coexpressed in the preneurogenic zone of both Xenopus (Perron et al.,1998) and chick, while Dll4 and the other proneural bHLH genes are expressed only in the neurogenic domain. Therefore, Ascl1 is a strong candidate for regulating Dll1 expression in the retina, while one or a combination of the other proneural genes likely regulates Dll4 expression.
Within the neurogenic domain, bHLH genes such as Ascl1 and Neurog2 are expressed in separate populations of progenitor cells (Marquardt et al.,2001; this report). Our data suggest a potential link between Ascl1 and Dll1, but it is clear that in other regions of the nervous system that Neurog2 can also regulate Dll1 expression (Fode et al.,1998; Castro et al.,2006). While our kinetic assay cannot distinguish differences in Dll1 regulation between Ascl1 or Neurog2 progenitor cells, it does demonstrate that rapid expression changes in these three genes are observed in progenitors, which was confirmed by our expression analyses. To test whether Ascl1 or Neurog2 specifically influenced Dll1 or Dll4 expression in the central neurogenic region, respectively, we overexpressed Ascl1 or Neurog2 and assayed for changes in Dll1 and Dll4 expression levels with QPCR. These results reveal that Ascl1 has a positive input into both Dll1 and Dll4 expression compared with Neurog2 and control experiments. Thus, although the kinetic profile observed for Dll4 is delayed with respect to Ascl1 and Dll1, it seems that Ascl1 may be important nonetheless for expression of both Delta-like genes in the neurogenic retina. This idea is consistent with the observation that some Dll4 cells incorporated BrdU, and that Hes5 expression levels were also increased with Ascl1-induced up-regulation of Dll1 and Dll4. This result indicates that Ascl1 may play a more general role in regulating Delta-like gene expression and Notch signaling activity during retinal development by initiating and maintaining neurogenesis.
While the role of proneural bHLH genes in the retina have been primarily discussed with regard to cell-autonomous specification of neural subtype identity (reviewed by Cepko,1999; Vetter and Brown,2001; Hatakeyama and Kageyama,2004; Yan et al.,2005; Harada et al.,2007; Ohsawa and Kageyama,2008), changes in cell nonautonomous Delta-Notch signaling may complicate the interpretation of these experiments. Further studies into the contribution of proneural bHLH genes to Notch signaling using more sophisticated knockout and reporter lines combined with live-cell imaging approaches should further elucidate the role of Notch signaling in experimental manipulations of proneural bHLH gene expression on retinal cell fate decisions.
In Situ Hybridizations and Immunolabeling
Fertilized white leghorn eggs were obtained (Hyline, Seattle, WA) and incubated to embryonic day E4.5 (Hamburger and Hamilton,1951) according to approved protocols at the University of Washington: some embryos received a 1- to 2-hr pulse of BrdU in ovo before killing (15 μl of a 15 mg/ml stock, Sigma). E4.5 chick embryos were collected. Some were fixed in 4% paraformaldehyde for 1–2 hr at room temperature, and prepared for cryosectioning or whole-mount immunolabeling, while others were fixed overnight in a modified Carnoy's solution (60%:30%:10% ratio of 100% EtOH, 37% formaldehyde, 100% Glacial Acetic Acid stocks), dehydrated and prepared for paraffin sectioning (6 μm, and processed for in situ hybridization for genes as previously described (Nelson et al., 2004,2006). Dll1 (cDelta1), Notch1 (cNotch1), Ascl1 (Cash1), Neurog2 (Ngn2), NeuroD4 (NeuroM), Atoh7 (Cath5), Hes1 (cHairy1), and Hes5 (cHes5-1) digoxigenin (DIG) -labeled antisense riboprobes for in situ hybridization detection were previously described (Jasoni et al.,1994; Henrique et al.,1995,1997a; Myat et al.,1996; Roztocil et al.,1997; Perez et al.,1999; Matter-Sadzinski et al.,2001; Nelson et al.,2006). To obtain a chick Dll4 probe, we used mammalian Dll1/4/3 sequences to BLAST the Chick EST Databases (University of Delaware, www.chickest.udel.edu; BBSRC, chick.umist.ac.uk) to identify ESTs with sequence homology. Several candidate clones were obtained, sequenced (SBRI, Seattle, WA), assembled, and aligned to genomic and predicted cDNA chicken Dll4 sequences XM_421132 (Sequencher, http://www.genecodes.com, Gene Codes; Ensemble Genome Browser, http://www.ensembl.org/index.html; Evolutionary Conserved Regions (ECR) Browser, http://ecrbrowser.dcode.org, Ovcharenko et al.,2004; ClustalW http://www.ebi.ac.uk/Tools/clustalw2/index.html, EMBL- EBI): clones pgn1c.pk015.f16, pgf2n.pk001.e7, pgp2n.pk004.o4 were from the University of Delaware (www.chickest.udel.edu), and ChEST714c11 from the BBSRC (chick.umist.ac.uk). We generated DIG-labeled probes from pgn1c.pk015.f16 for in situs, which contained ∼1.9 kb of sequence encompassing most of the C-terminal and 3′-untranslated region of chick Dll4, similar to the partial sequence from ChEST714c11 (Nimmagadda et al.,2007).
Following in situ hybridizations, sections were post-fixed in 4% paraformaldehyde, rinsed in phosphate buffered saline (PBS) and blocked in 10% goat serum PBS-0.1% Triton X-100. For BrdU detection, slides were first treated with 4N HCl for 7min, rinsed 4 times with PBS and then blocked. Primary antibodies include rabbit anti-neurofilament M 145 kDa (NF, 1:1,000 dilution, Chemicon), rat anti-BrdU (1:80 dilution; Accurate Chemical), and mouse anti-Islet1 and collagen-IX (1:20, Developmental Hybridoma Studies Bank). Secondary antibodies were species-specific Alexa Fluor 488 or 568, depending on the desired wavelength (1:500, Molecular Probes), some sections were counterstained with DAPI (4′,6-diamidino-2-phenylindole, 1:5000, Sigma), and all were mounted in Fluoromount-G (Southern Biotechnology Associates). Individual Nomarski and/or fluorescent images were acquired with a Zeiss Axioscope equipped with a Spot Camera. Whole-mount retinas stained for NF and Isl1 immunoreactivity were cleared and mounted in 50% glycerol PBS, and imaged by means of laser scanning confocal microscopy (Zeiss Pascal LSCM). Images were combined and assembled in Photoshop. Over 3,000 total cells were analyzed for the expression of Dll1, Dll4, Notch1, Ascl1, Neurog2, NeuroD4, and Atoh7 with respect NF immunolabeling and/or BrdU incorporation: multiple sections were analyzed for each marker from at least two different embryos (see the Results section for actual cell numbers).
We used QPCR to analyze the kinetics of Dll1 and Dll4 gene expression changes during timed inactivation of Notch signaling as described (Nelson et al.,2007a,b). Briefly, pairs of E4.5 chick retinal explants cultured for 3, 6, 12, 24, and 48 hr; one explant was cultured in 10 μM DAPT (γ-secretase inhibitor N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester, DAPT; Sigma), while the sister served as the dimethyl sulfoxide (DMSO) vehicle control. Retinas were lysed in Trizol (Invitrogen), total RNA was extracted, treated with RQ1 Rnase-free DNase (Promega) to remove genomic DNA, and converted to cDNA with SuperScript II reverse transcriptase (except RT- controls). Sample cDNA concentrations were diluted and then normalized to the respective ratios of glyceraldehydes-3-phosphate dehydrogenase (GAPDH) levels per retinal pair, with three pairs of retinas analyzed per time point (SYBR Green QPCR Master Mix, Applied Biosystems and an Option DNA Engine Real-Time QPCR machine, Bio-Rad). Student's t-test was used to determine significance differences between control and DAPT treated retinas at each time point, analysis of variance (ANOVA) was used to determine significant differences between time points, and changes of P < 0.05 were considered significant. QPCR primers are as follows: Dll1 forward CACTGACAACCCTGATGGTG, Dll1 reverse TGGC- ACTGGCATATGTAGGA; Dll4 forward CAAATTGCCGATTCTGTCCT, Dll4 reverse TGCTGTCGATGCTTGGTAAG; GAPDH forward CATCCAA- GGAGTGAGCCAAG, GAPDH reverse TGGAGGAAGAAATTGGAGGA.
Retinal Explant Transfection
We used electroporation to transfect retinal explants as previously described (Nelson et al.,2006,2007b). Briefly, E4.5 chick retinas were collected in HBSS+ and the retinal pigmented epithelium was removed. The central neurogenic region of the explants was targeted for electroporation: conditions were 25 V, 5 pulses, 50 msec-interval (BTX ECM 830). Plasmids expressing mouse Ascl1, Neurog2, and E12 were in pCS2+ driven by the CMV promoter (gift of D. Turner). Plasmids for expressing enhanced green fluorescent protein (GFP) and the Myc epitope (CXM) were in pCAGGS/pCAX driven by the CMV immediate early enhancer and chick beta-actin promoter. DNA for co-transfection was mixed at a ratio of 2× to 1× to 1× for Ascl1/Neuorg2/CXM to E12 to GFP, respectively: total DNA concentration for each mix was maintained at 4 μg/μl, and 1 μl was used per explant. After transfections, explants were cultured with gentle nutation for 24 hr at 37°C as previously described (Nelson et al.,2006,2007b). Explants were checked for GFP expression (Stemi SV11 fluorescent stereomicroscope, Zeiss), and the transfected fluorescent regions were individually microdissected and lysed in Trizol (Invitrogen), and cDNA was prepared as described above. QPCR was performed as described above to assay for changes in Dll1, Dll4, and Hes5 (Hes5.1) expression levels (see Nelson et al.,2006 for chick Hes5 QPCR primers). Student's t-test was used to determine significance differences between control CXM and Ascl1 or Neurog2 transfected explants, n = 4 per condition, and changes of P < 0.05 were considered significant.
We thank D. Henrique, D. Anderson, D. Turner, and J.M. Matter for use of reagents, and J. Brzezinski for critical reading of the manuscript. The monoclonal anti-Isl1 and anti–Coll-IX antibodies were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. B.R.N. was supported by NRSA postdoctoral fellowship and T.A.R. was funded by a NIH grant.