SEARCH

SEARCH BY CITATION

Keywords:

  • retina;
  • progenitor;
  • microarray;
  • cell fate;
  • Notch

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. ACKNOWLEDGMENTS
  8. REFERENCES
  9. Supporting Information

Background: The vertebrate retina comprises sensory neurons, the photoreceptors, as well as many other types of neurons and one type of glial cell. These cells are generated by multipotent and restricted retinal progenitor cells (RPCs), which express Notch1. Loss of Notch1 in RPCs late during retinal development results in the overproduction of rod photoreceptors at the expense of interneurons and glia. Results: To examine the molecular underpinnings of this observation, microarray analysis of single retinal cells from wild-type or Notch1 conditional knockout retinas was performed. In situ hybridization was carried out to validate some of the findings. Conclusions: The majority of Notch1-mutant cells lost expression of known Notch target genes. These cells also had low levels of RPC and cell cycle genes, and robustly up-regulated rod precursor genes. In addition, single wild-type cells, in which cell cycle marker genes were down-regulated, expressed markers of both rod photoreceptors and interneurons. Developmental Dynamics, 242:1147–1159, 2013. © 2013 Wiley Periodicals, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. ACKNOWLEDGMENTS
  8. REFERENCES
  9. Supporting Information

The vertebrate retina is an excellent model system for understanding how regulatory pathways control gene expression during development. It consists of six major neuronal cell types and one glial cell type that can be readily identified by molecular markers, gene expression, and morphology. These cell types arise in a temporal, overlapping order from a pool of multipotent retinal progenitor cells (RPCs) (Livesey and Cepko, 2001), as well as from terminal divisions from some more restricted RPCs (Godhinho et al., 2007; Rompani and Cepko, 2008; Hafler et al., 2012). During retinal neurogenesis, ganglion cells are generated first, followed by horizontal cells, cone photoreceptors, and amacrine cells. Rod photoreceptors, bipolar cells, and Müller glial cells are the last cell types to be produced (Young, 1985b; Wong and Rapaport, 2009).

Previous studies have determined that the Notch signaling pathway regulates both cell cycle exit and cell fate specification during retinal development (Austin et al., 1995; Dorsky et al., 1995; Tomita et al., 1996; Henrique et al., 1997; Furukawa et al., 2000; Hojo et al., 2000; Satow et al., 2001; Scheer et al., 2001; Silva et al., 2003; Jadhav et al., 2006a, 2006b; Nelson et al., 2006, 2007; Yaron et al., 2006; Riesenberg et al., 2009; Zheng et al., 2009). Genetic removal of a Notch1 conditional allele from early RPCs resulted in cell cycle exit and the premature onset of neurogenesis (Jadhav et al., 2006a; Yaron et al., 2006). Furthermore, overproduction of cone photoreceptors at the expense of other cell types was observed in these Notch1 mutant retinas. Deletion of Notch1 by viral delivery of Cre during later, postnatal stages of retinal development led to the overproduction of rod photoreceptors (Jadhav et al., 2006a), in keeping with the birth order of rod and cone photoreceptor cells. Furthermore, Notch1 mutant cells generated in the postnatal environment acquired their phenotype in a cell autonomous manner. Therefore, Notch signaling is crucial for maintenance of the progenitor state, as well as for the repression of the photoreceptor fate.

Despite our knowledge of several factors involved in cell fate specification, it is currently unknown when and how cells become locked into their respective identities. Lineage analyses in the postnatal retina have shown that individual RPCs can give rise to two very different cell types: a rod and Müller glial cell, a rod and a bipolar cell, or a rod and an amacrine cell, as demonstrated by the composition of two cell clones (Turner and Cepko, 1987). These cells may be sorting out their fates as they exit the cell cycle, or perhaps after entering a newly postmitotic state. Previous single cell transcriptional profiling showed that cycling cells are very heterogeneous in terms of gene expression (Trimarchi et al., 2007, 2008). They must lose this heterogeneity as they transition into differentiated neurons, because even in the wild-type (WT) case, most take on the rod fate. We wished to further explore the newly postmitotic state where these processes were taking place, and exploit the differences among WT and Notch1 conditional knockout (N1-CKO) cells for insight into these events. To this end, we examined the individual transcriptional profiles of 13 WT cells and 13 N1-CKO cells by single cell microarray analysis. Comparisons between the two sets of cells led to the identification of a large number of genes that were either up or down regulated in the absence of Notch1. From this dataset, we identified Notch dependent genes that may regulate or be markers of cell cycle, progenitor state, and cell fate determination. By post hoc classification, we were able to identify WT and N1-CKO individual cells at different stages of the progenitor to postmitotic neuron continuum, revealing the transcriptional profile of cells during this transition. Finally, we observed that single WT cells expressed early differentiation genes of both interneurons and photoreceptors. These expression profiles may indicate that there is plasticity regarding cell fate, and/or that certain types of genes are de-repressed transiently during this phase of retinal development.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. ACKNOWLEDGMENTS
  8. REFERENCES
  9. Supporting Information

Profiling Single WT and N1-CKO Retinal Cells

We aimed to generate cells that had lost Notch1 by electroporation, and thus first sought to confirm that electroporation into the Notch1fl/fl background recapitulated the phenotype observed in the previous viral experiments (Jadhav et al., 2006a). Retinas of Notch1fl/fl postnatal day (P) 0 pups were electroporated in vivo with plasmids encoding Cre driven by a broadly active promoter, CAG, along with a Cre-responsive green fluorescent protein (GFP) reporter, also driven by the CAG promoter (CALNL-GFP) (Matsuda and Cepko, 2004) to generate GFP+ N1-CKO cells. For controls, the retinas of sibling P0 Notch1fl/fl pups were electroporated with CAG:GFP to generate WT GFP+ cells. Retinas were harvested after maturation (>P21) and the fate of GFP+ cells was assessed by morphology and location in the retinal layers. For example, rod photoreceptors are exclusively localized in the outer nuclear layer, which is distinct from the location of the other neurons and Müller glial cells. In accord with previous studies, the majority of cells that lost Notch1 took on a rod photoreceptor fate, whereas WT cells took on a variety of fates (Fig. 1A,B).

image

Figure 1. Removal of Notch1 from the postnatal retina by electroporation. Plasmids encoding CAG:GFP (green fluorescent protein) or CAG:Cre with a Cre reporter (CALNL-GFP) were electroporated in vivo into Notch1fl/fl postnatal day (P) 0 mouse retinas. The fates of electroporated cells were analyzed in the mature retina after P21 by histology. A: Electroporation of CAG:GFP alone (WT) into Notch1fl/fl retinas labeled GFP+ photoreceptors (located in the ONL), and interneurons and Müller glial cells (located in the INL). B: Electroporation of CAG:Cre and CALNL-GFP (Notch1 conditional knockout, N1-CKO) into Notch1fl/fl retinas labeled GFP+ photoreceptors and some amacrine cells, but not bipolar cells or Müller glial cells. C: Notch1fl/fl retinas were electroporated at P0 in vivo with CAG:GFP to mark WT cells, or CAG:Cre and CALNL-GFP, to mark N1-CKO cells. After 3 days in vivo, retinas were dissociated and single GFP+ cells were harvested for profiling. Each single cell was subjected to reverse transcription and PCR, with the resulting probes hybridized to Affymetrix arrays. The average signal levels for selected Notch target and pro-neurogenic genes are shown. D: Notch1fl/fl retinas were electroporated at P0 in vitro with CAG:GFP to mark WT cells, or CAG:Cre and CALNL-GFP, to mark N1-CKO cells. After 3 days in culture, GFP+ cells were collected by flow cytometry and used to prepare cDNA. Samples were subjected to semi-quantitative real-time PCR to detect differences in expression of Notch target and pro-neurogenic genes between N1-CKO and WT cells. All expression values were normalized to actin expression levels in each sample. n = 3 retinas per condition. P value < 0.05. r, rod photoreceptor; bp, bipolar cell; Mg, Müller glial cell; ac, amacrine cell. Cellular laminae are denoted: ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer.

Download figure to PowerPoint

After confirming that Cre electroporation into the Notch1fl/fl background produces a Notch1 loss of function phenotype, we profiled individual P3 N1-CKO and WT cells on microarrays to examine their transcriptomes at earlier stages of development. For single cell profiling, Notch1fl/fl P0 pups were electroporated as described above, but the retinas were harvested at P3 to allow time for Notch1 to be deleted by Cre and downstream gene expression changes to occur, but before the electroporated cells could fully differentiate into mature cell types. At this time point, the majority of electroporated cells was either exiting the cell cycle or had recently exited the cell cycle. We know this for several reasons. First, our previous studies of cell fates marked by electroporation at P0 showed that cell types born in the embryonic period were only rarely labeled, with the vast majority instead being cell types generated in the postnatal period (Matsuda and Cepko, 2004). Our interpretation of this finding is that the cells facing the subretinal space where the DNA is introduced are the most likely to be electroporated, and are RPCs or cells that are exiting cell cycle. By P3, Young showed that proliferation is almost over in the center of the retina and waning in the periphery (Young, 1985a, 1985b), so these P3 electroporated cells are likely RPCs or newly exited cells. In corroboration of these results are the clone sizes of retrovirally marked clones from P0 infections in the mouse. Gammaretroviruses can only mark mitotic cells and their progeny, and marking is initiated after the first or second M phase following viral infection (Roe et al., 1997). In previous work from our lab, we measured the clone size for two different retroviruses, one encoding alkaline phosphatase (AP) and one encoding lacZ. The average clone sizes for these two vectors were 1.8 cells/clone and 1.9 cells/clone, respectively, following infection at P0 (Fields-Berry et al., 1992). Also, the clone size distribution indicates that most RPCs produce postmitotic daughters or RPCs with very limited proliferation capacity. Taken together, all of these data indicate that the majority of cells electroporated at P0, and analyzed at P3, will be cells in transition from cell cycle to the newly postmitotic state.

Retinas electroporated with either CAG:Cre and CALNL-GFP or CAG:GFP alone were dissected and dissociated to individual cells, which were then harvested under a dissecting microscope on the basis of their GFP signal. In total, 13 N1-CKO cells and 13 WT cells were harvested and profiled on Affymetrix microarrays. These methods have been used previously for profiling individual retinal cells, with the results validated by several methods, primarily in situ hybridization (Brady and Iscove, 1993; Tietjen et al., 2003; Trimarchi et al., 2007, 2008).

To confirm that Notch1 signaling was indeed depleted in N1-CKO cells, the average levels of the direct Notch target genes, Hes1, Hes5, and Nrarp were assessed in N1-CKO vs. WT cells (Ohtsuka et al., 1999; Krebs et al., 2001). Expression levels of these genes were markedly lower in N1-CKO cells as compared to WT cells (Fig. 1C). As RPCs divide to generate neurons, these newly born cells turn off genes associated with cell cycle and progenitor status and begin to express genes that regulate neuronal identity. Some early transcription factors that influence neuronal fate include Math3, NeuroD1, and Blimp1 (Morrow et al., 1999; Chang et al., 2002; Inoue et al., 2002; Brzezinski et al., 2010). These genes were found to be strongly up-regulated in N1-CKO cells as compared to WT cells (Fig. 1C).

To validate the single cell profiling method as a means to assess changes in gene expression after the removal of Notch1, we performed a quantitative polymerase chain reaction (qPCR) assay on populations of N1-CKO and WT cells. Retinas of Notch1fl/fl P0 pups were electroporated in vitro with plasmids encoding CAG:Cre, along with a Cre-responsive GFP reporter (CALNL-GFP). For controls, the retinas of sibling Notch1fl/fl pups were electroporated with CAG:GFP. Electroporated retinas were cultured for three days and dissociated to single cells. GFP+ cells (pooled from two retinas for each sample) were sorted by flow cytometry and collected. RNA was extracted from each sample and cDNA was generated. Samples were subjected to real-time qPCR to detect expression of actin (as a control), Hes1, Nrarp, Math3, NeuroD1, and Blimp1 (Fig. 1D). In accord with the changes observed by microarray analysis, Hes1 and Nrarp were down-regulated in N1-CKO cells, as compared to WT cells (Fig. 1D). Additionally, Math3, Blimp1, and NeuroD1 were up-regulated in N1-CKO cells as compared to WT cells (Fig. 1D). Changes in expression of these key genes in a population of N1-CKO cells as compared to WT were highly similar to those observed in individual cells.

Classification of Cells Using Their Molecular Signatures

We wanted to understand the transitions that cells undergo as they exit the cell cycle and choose their fate, both in the WT and N1-CKO cells. To investigate if there are indications of an eventual fate choice during this transition, we classified each N1-CKO and WT cell according to its expression of cell type-specific markers. The classification scheme, as devised previously, is based upon the normalized values of genes coexpressed with known markers of each of the retinal cell types (Trimarchi et al., 2007, 2008). As an example, to determine if a cell has characteristics of an amacrine cell, we used the expression levels of genes co-regulated with well-validated markers of the amacrine fate. These co-regulated genes had been identified using microarray data from single cells profiled previously, as genes whose expression was strongly associated with the expression of the known amacrine-specific genes, Tcfap-2β, Gad1, and Glyt1. The associations were derived from 194 single retinal cells that were profiled in our lab and encompassed most retinal cell types (Trimarchi et al., 2007, 2008; Kim et al., 2008; Roesch et al., 2008; Cherry et al., 2009). A Fisher's exact test was used to determine the P-value for correlations between any given gene and Tcfap-2β, Gad1, and Glyt1. Only associated genes that had P values of <0.01 were considered to be highly associated. The relative expression level for each associated gene in each P3 N1-CKO or WT cell was calculated by dividing a cell's signal level by the maximum signal level found in all of the single cells within the entire dataset of 194 cells. These scaled values for all of the amacrine associated genes in each cell were summed, and then the sums were scaled, such that the maximum score was 10. This classification procedure was repeated with markers of other retinal cell types to generate scores for each cell type. For an RPC score, genes associated with FGF15, Sfrp2, and μ-crystallin were used; for retinal ganglion cells, those associated with NF68 and Ebf3; for Müller glia, those associated with Apoe; and for bipolar cells, those associated with Og9x (Trimarchi et al., 2008). To generate a developing photoreceptor score, the gene Blimp1 was used to find associated genes. This gene has been shown to be expressed early in photoreceptor development and its expression tapers off as these cells mature (Chang et al., 2002; Brzezinski et al., 2010; Katoh et al., 2010). This was considered a more appropriate marker for newly postmitotic cells that would likely achieve the rod fate, instead of a more typical rod photoreceptor marker, such as rhodopsin, whose expression is later in development.

Using this classification scheme, the majority of the N1-CKO cells scored highly as incipient rod photoreceptors (Fig. 2, see N1-CKO cells 1–10). Three of the N1-CKO cells scored as amacrine precursor cells (Fig. 2, see N1-CKO cells 11–13). Most of the WT cells were classified as RPCs (Fig. 2, see WT cells 2–6, 8, 12), while some cells had high rod (Fig. 2, see rod WT cells 9–11, 13) or amacrine scores (Fig. 2, see amacrine WT cell 7). Some WT cells had intermediate scores for RPC, rod, and amacrine cell types (discussed below; Fig. 2, see WT cells 1, 2, 6). These outcomes are in keeping with Notch1 depletion in N1-CKO cells, as none of these mutant cells scored as RPCs and the majority were classified as rod precursor cells. For comparison, the classification scores of postnatal WT cells which were picked for other studies and classified as either amacrine precursors (Fig. 2, see cells P0 A4, P0 B1, P0 D1, P0 G3; Cherry et al., 2009) or rod precursors (Fig. 2, see cells P0 E1, P5 C4, P5 D2, P5 C2; unpublished data), are depicted.

image

Figure 2. Classification of single profiled cells. Profiled cells were classified as retinal progenitor cells (RPCs), rod photoreceptors, bipolar cells, amacrine cells, ganglion cells, or Müller glia based on the expression levels of genes associated with known markers of that particular cell type. Genes associated with one or several cell type-specific markers were determined by a Fisher's exact test and the relative expression level for each associated gene was calculated by dividing the cell's signal level by the maximum signal level found in a large collection of single cells (Trimarchi et al., 2008). These values were summed and normalized to generate a cell type score for each cell, with 10 being the maximum score for each cell type. Genes associated with FGF15, Sfrp2, and μ-crystallin were used to generate a RPC score, genes associated with Blimp1 for rod photoreceptors, genes associated with Og9x for bipolar cells, genes associated with Tcfap-2β, Gad1, and Glyt1 for amacrine cells, genes associated with NF68 and Ebf3 for ganglion cells, and genes associated with Apoe for Müller glia. For comparison, previously profiled cells, which were classified as amacrine (cells P0 A4, P0 B1, P0 D1, P0 G3; Cherry et al., 2009) or rod precursors (cells P0 E1, P5 C4, P5 D2, P5 C2; Trimarchi and Cepko, unpublished data), are shown. The highest score for each cell is boxed in red.

Download figure to PowerPoint

Additionally, it is worth noting that none of these postnatal cells had a ganglion cell signature. This gives confidence in the classification scheme, and is in keeping with the idea that primarily mitotic or newly postmitotic cells adjacent to the subretinal space were electroporated (Matsuda and Cepko, 2004). Ganglion cells would have already been produced and would have migrated away from this surface by P0 when the retina was electroporated. Overall, these molecular data support the observed cell fate changes, as well as provide a source of gene expression changes that are likely informative with regard to the network that is regulated by Notch1.

Changes in RPC and Cell Cycle Gene Expression in Single Cells That Have Lost Notch1

In contrast to whole tissue microarray analysis, single cell profiling affords the ability to examine changes in gene expression at the resolution of individual cells. Cell-by-cell analysis is especially important in the retina, as RPCs and the neurons they produce are highly heterogeneous in the types of genes they express (Trimarchi et al., 2007, 2008; Cherry et al., 2009). We anticipated that the downstream changes in gene expression in WT cells vs. N1-CKO cells would include genes relevant to Notch signaling, the progenitor state, and cell fate choices. We visually inspected the microarray data for changes in key genes that represent these stages and whose expression patterns may be different from cell to cell.

Although on average N1-CKO cells lost expression of Notch target genes as compared to WT cells (Fig. 1C), inspection of gene expression in individual cells showed that this group of cells was in various stages of maturation, as suggested by the classification scheme (Fig. 2). For example, a few N1-CKO cells still expressed some, but not all, downstream target genes, suggesting that they had not lost all of their Notch signal and were in the process of down-regulating Notch signaling (Fig. 3, see N1-CKO cells 1–3). Some WT cells expressed high levels of Notch targets, indicating that they still had active Notch signaling (Fig. 3, see WT cells 1–6, 13). As discussed above, most of these cells were classified as RPCs (Fig. 2, see WT cells 2–6). Similar to the majority of the N1-CKO cells, several of the WT cells exhibited low levels of Notch target genes (Fig. 3, see WT cells 7–12). These WT cells most likely were in a transitional state during which they were down-regulating Notch activity to exit the cell cycle, as occurs normally, especially at this time in development (Young, 1985b).

image

Figure 3. Gene expression of selected Notch target, cell cycle, progenitor, and pro-neurogenic genes in wild-type (WT) and Notch1 conditional knockout (N1-CKO) single cells. The microarrays performed using cDNA from single WT and N1-CKO cells (described in Fig. 1) were analyzed for the expression of selected genes. The signals for different types of genes are shown as a heatmap that was generated using Treeview software. The expression levels of Notch target, cell cycle, progenitor, and pro-neurogenic genes are shown. See also Supplementary Tables S1 and S2. The signal intensity from Affymetrix microarray chips has been scaled and is represented by a gradation in color, from bright red to black. Signals below 3,000 are black and signals from 3,000 to 10,000 are a scaled shade of red. For Affymetrix identifier, Unigene number, and full gene name, see Table S3.

Download figure to PowerPoint

We also examined genes predicted to regulate or drive cell cycle, as loss of Notch1 has been reported to lead to cell cycle exit (Jadhav et al., 2006a; Yaron et al., 2006). The levels of cell cycle genes, Geminin, Ccna2, CyclinB1, Cdc20, and CyclinB2 (Trimarchi et al., 2008), were thus assessed in N1-CKO and WT cells (Fig. 3). Indeed, the N1-CKO cells showed a reduction in the expression levels of these cell cycle genes (Fig. 3, see N1-CKO cells 4–13). Moreover, the cells with the most comprehensive reduction in cell cycle genes were the same cells that showed the most significant loss of Notch target genes (Fig. 3, see N1-CKO cells 4–13). The WT cells with low levels of Notch targets, also had reduced levels of these same cell cycle genes (Fig. 3, see WT cells 7–13). In contrast, WT and N1-CKO cells that expressed Notch target genes had high levels of cell cycle genes (Fig. 3, see WT cells 1–6 and N1-CKO cells 1–3).

In addition to cell cycle genes, expression of other previously identified RPC genes was assessed. Some of these genes are in all or most RPCs, but are also expressed in subsets of neurons (e.g., Pax6). We chose to analyze expression of genes that are expressed in most RPCs, but are not expressed in many neurons. These included Lhx2, Mik67, Cdca8, Cdc2a, Fgf15, Ttyh1, and μ-crystallin (Fig. 3; Blackshaw et al., 2004; Trimarchi et al., 2008). Most N1-CKO cells had low levels of these genes and were the same cells that had scored highly as neurons in the classification scheme described above (Figs. 2, 3, see N1-CKO cells 4–13). Of interest, the WT cells that showed low levels of Notch target genes and cell cycle genes retained expression of some RPC genes (Fig. 3, see WT cells 7–13), even though they were classified as neurons when the wider range of marker genes was scored in the classification scheme (Fig. 2, see WT cells 7–13). This observation may indicate that these WT cells had a slower pace of exiting the RPC state and executing their differentiation process. In contrast, several WT cells that expressed high levels of Notch target genes and cell cycle genes, also expressed high levels of most RPC genes (Fig. 3, see WT cells 1–6). This group of cells was classified as RPCs, with the exception of WT cell 1, which had an intermediate RPC and rod score (Fig. 2). These data reveal which genes are sensitive to Notch1 signaling and provide examples of single cells at various stages in the transition between RPC and determined states.

Unbiased Search for Genes With Expression Changes Following Loss of Notch1

An unbiased search for significantly down-regulated genes was conducted by comparing gene expression levels in cells classified as RPCs (WT cells 2–6, 8, 12) to those in cells classified as rod precursor cells (N1-CKO cells 1–10). These particular cells were selected, because it was anticipated that RPCs express different sets of genes than rod precursor cells. T-test analysis with a cutoff P value of <0.05 was performed to find significantly down-regulated genes (Supplementary Table S1, which is available online). We observed that Notch target genes such as Hes1 and Hes5 were found to be significantly down-regulated (Supplementary Table S1), in accord with the trend observed for the average signal levels across all the profiled cells (Fig. 1).

As loss of Notch signaling leads to the overproduction of rod photoreceptors at this stage of retinal development, it was anticipated that genes involved in photoreceptor development would be up-regulated in N1-CKO cells. Again, gene expression levels were compared between WT RPC cells and N1-CKO rod precursor cells. T-test analysis with a cutoff P-value of <0.05 was performed to find significantly up-regulated genes (Table S2). NeuroD1, Math3, and Blimp1 are three such genes that were significantly up-regulated in rod precursor N1-CKO cells (Figs. 1, 3; Table S2), similar to the trends observed when all the single cells were taken under consideration (Fig. 1C). NeuroD1 and Math3 encode pro-neurogenic bHLH transcription factors that can lead to overproduction of rods when misexpressed (Morrow et al., 1999; Inoue et al., 2002; Cherry et al., 2011). Interestingly, Math3 was up-regulated in the N1-CKO cells that were classified as incipient rods, but not in the cells classified as amacrine precursor cells (Figs. 2, 3 see rod N1-CKO cells 1–10 and amacrine N1-CKO cells 11–13), while NeuroD1 was up-regulated in N1-CKO cells classified as amacrine cells, as well as in those classified as rods (Figs. 2, 3, see N1-CKO cells 1–13). This is in keeping with the expression of NeuroD1 in amacrine cells and the induction of amacrine cells, along with rods, following NeuroD1 misexpression (Morrow et al., 1999; Inoue et al., 2002; Cherry et al., 2011).

Blimp1, a gene that has been demonstrated to positively regulate the production of photoreceptor cells through repression of the bipolar cell fate (Brzezinski et al., 2010; Katoh et al., 2010), was also up-regulated in N1-CKO cells (Figs. 4, 5, discussed below). Because its expression is transient in retinal development (Chang et al., 2002; Brzezinski et al., 2010), Blimp1 is thought to demarcate the early period of photoreceptor formation. The robust up-regulation of these key photoreceptor genes provides additional support for the validity of the single cell microarray approach in defining the genes responding to the loss of Notch1 and inducing the rod fate.

image

Figure 4. Analysis of Blimp1 associated genes. The microarrays performed using cDNA from single wild-type (WT) and Notch1 conditional knockout (N1-CKO) cells (described in Fig. 1) were analyzed to identify genes whose expression patterns significantly correlated with Blimp1 (P-value listed next to gene row). A: A heatmap was generated using Treeview software to visualize expression levels of Blimp1 and its associated genes in N1-CKO and WT cells. The signal intensity from Affymetrix microarray chips has been scaled and is represented by a gradation in color, from bright red to black. Signals below 3,000 are black and signals from 3,000 to 10,000 are a corresponding shade of red. For Affymetrix identifier, Unigene number, and full gene name, see Table S3. B–G: In situ hybridization for a novel gene, Epha8, correlated with Blimp1 expression was performed at P3 (B,E), P9 (C,F), and adult (D,G) stages. B–G: Probes used were Blimp1 (B–D) and Epha8 (E–G). Cellular laminae are denoted: ONBL, outer neuroblastic layer; INBL, inner neuroblastic layer; ONL, outer nuclear layer; INL, inner nuclear layer. Scale bar = 50 μm.

Download figure to PowerPoint

image

Figure 5. Expression of selected rod and amacrine marker genes in Notch1 conditional knockout (N1-CKO) and wild-type (WT) cells. The microarrays performed using cDNA from single WT and N1-CKO cells (described in Fig. 1) were analyzed for the expression of genes that are expressed in rod photoreceptors and amacrine cells. A heatmap was generated using Treeview software to visualize the expression of these genes. For comparison, expression levels of selected genes in previously profiled cells, which were classified as amacrine (Cherry et al., 2009) or rod precursors (Trimarchi and Cepko, unpublished data), are shown. The signal intensity from Affymetrix microarray chips has been scaled and is represented by a gradation in color, from bright red to black. Signals below 3,000 are black and signals from 3,000 to 10,000 are the appropriate shade of red. For Affymetrix identifier, Unigene number, and full gene name, see Table S3.

Download figure to PowerPoint

Identification of Genes Associated With Blimp1

Blimp1 is expressed during embryonic time points to early postnatal stages in the developing retina in a temporal and spatial pattern highly correlated with incipient photoreceptors (Chang et al., 2002; Brzezinski et al., 2010; Katoh et al., 2010). Double immunohistochemistry experiments showed that Blimp1+ cells do not express RPC markers, but do coexpress NeuroD1 (Brzezinski et al., 2010). Because cells expressing Blimp1 have the molecular characteristics of early rods, the genes that are co-regulated with Blimp1 are candidates for genes involved in photoreceptor differentiation. Using the pairwise comparison described above to find co-regulated genes, genes with expression patterns similar to Blimp1 with P-values of <0.01 were identified (Fig. 4A). Some known factors that closely tracked with Blimp1 included Rax, Math3, and Rbp3. These genes are either markers of developing rods (Rbp3) or have been shown to play a role during photoreceptor genesis (Rax, Math3; Chen and Cepko, 2002; Inoue et al., 2002; Jin et al., 2009; Muranishi et al., 2011; Fig. 4A). In addition, Epha8, a gene not previously identified as associated with rod development, was identified. This may be a novel marker of rod photoreceptors, perhaps playing a functional role during rod specification and/or differentiation (Fig. 4A).

To validate if Epha8 has a similar expression pattern to Blimp1, in situ hybridization (ISH) was performed on retinal sections at P3, P9, and adult stages. Detection of Blimp1 expression by ISH matched previous reports of Blimp1 expression by ISH, immunohistochemistry, and transgene expression (Chang et al., 2002; Brzezinski et al., 2010; Katoh et al., 2010). At P3, Blimp1 was expressed in the scleral outer neuroblastic layer (ONBL), where incipient photoreceptors are located (Fig. 4B). A similar expression pattern was observed at P9 (Fig. 4C). Very faint staining was detected at adult stages (Fig. 4D). The expression pattern of Epha8 was investigated at the same stages. At P3, staining in the ONBL was observed, similar to the pattern of Blimp1 expression (Fig. 4B,E). Epha8 expression was at a very low level throughout the retinal layers at P9 and adult stages (Fig. 4F,G). These results corroborated the microarray analysis, as Epha8 expression was very similar to Blimp1 expression at postnatal stages (Fig. 4). Further study is necessary to identify a functional role for Epha8 during retinal development.

Markers of Cell Types Expressed by Profiled Single Cells

In addition to learning about genes involved in rod development, it was of interest to mine the microarray data from WT and N1-CKO cells for the expression of genes that are markers of amacrine cells, bipolar cells, and Müller glia. These fates are the ones normally taken by approximately 30% of the postnatally generated cells, and which are greatly reduced in the N1-CKO population. Expression values of known amacrine and rod marker genes were assessed in WT and N1-CKO cells, as well as in cells previously analyzed by our lab, which had been classified as either rod or amacrine precursor cells (Fig. 5). As expected from the classification results, P3 N1-CKO cells, which had been classified as rod precursor cells, expressed rod marker genes robustly, but did not express amacrine marker genes (Figs. 2, 5, see N1-CKO cells 1–10). The transcriptional profiles of these cells resembled those of previously profiled rod precursor cells (Fig. 5, see P0 cell E1, P5 cell C4, P5 cell D2, P5 cell C2; Trimarchi et al., 2007). Conversely, N1-CKO cells classified as developing amacrine cells expressed amacrine marker genes and not rod marker genes (Figs. 2, 5, see N1-CKO cells 11–13). The expression profiles of these cells were similar to cells classified as amacrine precursor cells in our previous studies (Fig. 5, see P0 cell A4, P0 cell B1, P0 cell D1, P0 cell G3; Trimarchi et al., 2007; Cherry et al., 2009).

Of interest, some WT cells expressed marker genes specific to mature amacrine cells, as well as genes specific to rod photoreceptors. This is in keeping with the scores on the classification scheme, as some WT cells did not exhibit scores indicating clear cell type identities. Examples of “mixed identity” cells that expressed several amacrine marker genes (such as Tcfap-2β, Fgf13, Nhlh2) and rod marker genes (such as Crx, Otx2, Nrl) include WT cells 1, 2, and 6 (Fig. 5). These cells did not score highly in the classification scheme for any of the potential cell types (Fig. 2, see WT cells 1, 2, and 6) and retained expression of cell cycle and RPC marker genes (Fig. 3, see WT cells 1, 2, and 6).

To independently validate the observation that WT cells can coexpress marker genes of two different cell types, we performed two-color fluorescent dissociated cell in situ hybridization. This was done on dissociated cells to remove any ambiguity of signals overlying more than one cell. The probes were applied simultaneously to detect expression of an amacrine marker (Tcfap-2β) and a rod marker (Crx) in individual P3 retinal cells (Fig. 6A). We found that 47.9 ± 1.3% of all cells were Crx+, 10.3 ± 1.0% were Tcfap-2β+, and 2.2 ± 0.5% of cells were positive for both markers (Fig. 6C). In addition, we probed for the expression of the ganglion cell marker, NF68, and the rod marker, Crx, in dissociated P3 cells (Fig. 6B). Because ganglion cells are not produced postnatally, we did not anticipate that any cells would be in a transitional state in which they would be double positive for Crx and NF68. In this experiment, 54.2 ± 1.5% of all cells were Crx+ and 1.3 ± 0.15% of all cells were NF68+ (Fig. 6C). We found that a negligible amount of cells (0.03 ± 0.02%) were double positive for both Crx and NF68 (Fig. 6C). The coexpression of an amacrine marker gene and a rod marker gene in a subset of individual P3 WT cells supported our microarray findings.

image

Figure 6. Simultaneous detection of RNA for two different cell type-specific marker genes by two-color fluorescent in situ hybridization. Wild-type (WT) postnatal day (P) 3 retinas were probed for expression of different cell type-specific markers by double fluorescent in situ hybridization on dissociated cells. A,B: Cells from dissociated retinas were probed for the expression of Tcfap-2β (amacrine marker) and Crx (photoreceptor marker; A), or for Crx (photoreceptor marker) and NF68 (ganglion cell marker) expression (B). Arrow indicates a double marker+ cell. C: Quantification of single and double marker+ cells. n = 2047 total cells examined for Tcfap-2β and Crx expression. n = 6278 total cells examined for NF68 and Crx expression.

Download figure to PowerPoint

The profiled cells were also examined for expression of Müller glial and bipolar cell marker genes. Previous studies have shown that several genes expressed by mature Müller glial cells, such as Sox2, μ-crystallin, and Dkk3, were also expressed in WT late RPCs (Blackshaw et al., 2004; Roesch et al., 2008; Trimarchi et al., 2008). Some of these shared genes were down-regulated in N1-CKO cells (Fig. 3; Supplementary Table S1). Additionally, marker genes thought to be specific for Müller glial cells, such as Apoe and clusterin (Blackshaw et al., 2004; Roesch et al., 2008), were not expressed above detectable levels in almost any of the profiled cells (Supplementary Fig. S1). For these reasons, it was difficult to determine by direct inspection of the heatmaps if cells were becoming Müller glia. Inspection of the profiled cells for expression of bipolar genes did not yield many positives, which may not be surprising, as known bipolar marker genes are not robustly expressed at P3 (e.g., Lhx3, Car8, Car10, and Nfasc; Supplementary Fig. S2; Kim et al., 2008). The absence of these marker genes does not preclude the possibility that some of these cells may later express bipolar genes or be on the path to becoming bipolar cells. In addition, using the classification scheme, which relies on a large number of genes, none of the P3 cells were classified as Müller glial or bipolar cells.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. ACKNOWLEDGMENTS
  8. REFERENCES
  9. Supporting Information

In this study, we profiled single N1-CKO and WT retinal cells on Affymetrix microarrays to examine changes in gene expression that occur in the absence of Notch1 and as cells transition from the progenitor to the neuronal state. At early postnatal stages, the majority of RPCs produce postmitotic cells, which then differentiate into functional neurons over the course of several weeks (Young, 1985b). Levels of Notch1 signaling are likely being read out during the time period when mitotic cells are producing postmitotic cells, and the fates of these cells are being established. This expectation is based on the expression of Notch1 RNA using in situ hybridization (Lindsell et al., 1996; Bao and Cepko, 1997), SAGE (Blackshaw et al., 2004), and single cell microarrays (Trimarchi et al., 2008). Furthermore, the loss of function studies performed in several labs support a role for Notch1 in cycling cells (Dorsky et al., 1995; Henrique et al., 1997; Jadhav et al., 2006a; Nelson et al., 2007; Yaron et al., 2006). Our previous studies of electroporated cells in the P0 retina have shown that mitotic cells or cells exiting mitosis take up plasmids (Matsuda and Cepko, 2004, 2007). These observations, coupled with the analysis of retroviral clone sizes (Fields-Berry et al., 1992) and Young's birthdating studies (Young, 1985b), lead to the conclusion that the electroporation of Cre into Notch1fl/fl mice at P0, with harvest at P3, resulted in the removal of Notch1 from this transitioning population of cells. Because the Notch signaling pathway is a potent regulator of many cellular processes, it is tightly regulated to prevent sustained activation (Kopan and Ilagan, 2009). For example, it is known that the activated form of the receptor, NICD, does not accumulate or linger in the cell, due to a PEST sequence that targets it for degradation (Öberg et al., 2001). From these observations, we anticipated that the rate of Notch protein turnover in retinal cells was relatively rapid, likely faster than the rate at which retinal cells take on their various fates. Indeed, the transcriptomes of the 13 N1-CKO cells that were profiled supported this expectation. The majority of these cells lost expression of Notch target, cell cycle, and progenitor genes, while a few cells appeared to be in the process of down-regulating these genes. Furthermore, these cells expressed early marker genes of rods (such as Blimp1, Crx, Otx2), but not markers of differentiated rods (such as rhodopsin), providing evidence that the loss of Notch1 did not dramatically accelerate their differentiation program. However, the state of these cells did appear further advanced down the rod development pathway as compared to the WT cells that had also turned down Notch target genes, but which had retained expression of some cell cycle and progenitor genes.

Using this unbiased method of single cell profiling, we identified a large number of genes that were either up or down-regulated in the absence of Notch1. The cohort of down-regulated genes included cell cycle regulators or progenitor markers, some of which were not yet appreciated to be Notch1 sensitive (e.g., Fgf15, Cdc20, Crym). Up-regulated genes included known regulators or markers of rod photoreceptor development, such as NeuroD1, Math3, Rbp3, and Blimp1. Future experiments need to be performed to test the novel Notch responsive genes identified by this study for their roles in retinal development.

The majority of the profiled N1-CKO cells were classified as incipient photoreceptors using Blimp1 as a marker. As described above, Blimp1 is expressed in rod precursor cells. Genes whose expression patterns were highly correlated with Blimp1 included genes known to be expressed in rods, such as Math3, Rbp3, and Rax, in addition to the newly identified gene, Epha8. The expression pattern of Epha8 at P3 suggests that this gene is a good marker of early rod photoreceptors, similar to Blimp1. Future experiments will determine this specificity, as well as elucidate whether Epha8 plays a functional role in retinal development.

Single cell microarray analysis and in situ hybridization on dissociated retinal cells simultaneously using probes for two genes revealed that newly postmitotic retinal cells coexpressed amacrine and rod marker genes. This result is consistent with the idea that cells may transition through a plastic phase shortly after exiting the cell cycle during which they can express marker genes of different cell types. Alternatively, the coexpression of markers of two cell types may indicate that certain loci are de-repressed, independently of whether a cell is still plastic enough to choose more than one cell fate. It is important to note that the coexpression of rod and amacrine markers does not appear to be an artifact of the single cell profiling method. Only certain genes were coexpressed, and the same ones were seen in multiple cells. If, for example, coexpression was the result of contamination of a single cell's RNA preparation with RNAs from another cell, one might predict random patterns, as opposed to consistent genes found in such profiles. In addition, many cells that appeared more mature, as assessed by higher levels of specific genes and classification scores that indicated a more definitive fate, did not coexpress genes of two cell types (Fig. 2).

It is unclear what population of cells is represented by the cells that coexpress these markers. If the majority of RPCs are determined to give rise to stereotyped progeny, and coexpression of amacrine and rod genes occurs in the RPCs that will give rise to a rod and an amacrine, then only a few single cells should coexpress amacrine and rod marker genes. This prediction is based upon Young's birthdating data, as amacrine cells are only a small percentage (1–2%) of the progeny of P0 RPC (Young, 1985b). However, most of the profiled WT cells in this study coexpressed amacrine and rod marker genes. If there are determined subsets of RPCs that produce a rod and an amacrine, a rod and a bipolar, or a rod and Müller glial cell, then single cells coexpressing rod and bipolar genes, as well as single cells coexpressing rod and Müller genes would have been predicted. In fact, Müller glial genes are expressed in the majority of P0 RPCs, though only a small percentage of the progeny of postnatal RPCs (<10%) will be Müller glial cells (Young, 1985b; Blackshaw et al., 2004; Roesch et al., 2008; Trimarchi et al., 2008). Single WT cells did coexpress marker genes shared by late RPCs and Müller glial cells, but not marker genes exclusive to Müller glial cells. In addition, the single cells did not express most known markers of bipolar cells, which is likely because these genes are not expressed as early as P3 (Kim et al., 2008). There may be active repression of most bipolar genes as a result of Blimp1 activity, which is expressed in these cells and has been shown to inhibit bipolar fate, likely by means of Chx10-mediated represssion (Brzezinski et al., 2010; Katoh et al., 2010). This repression may be transient, as Blimp1 expression wanes late in the first postnatal week, when many bipolar cells are born and/or begin to differentiate (Young, 1985b; Kim et al., 2008). Despite this, however, there was expression of Gnb3, a bipolar gene in the mature retina, in the majority of N1-CKO cells. Further analysis of these genetic relationships will be required to understand how these cells sort out their fates, and to understand the meaning of the coexpression of two different cell type marker genes.

An intriguing possibility is that coexpression of different marker genes represents a plastic stage through which cells transition after cell cycle exit as they take on their identities. Although newly postmitotic cells likely receive fate determining factors from progenitor cells, it is currently unknown whether cell fate determination occurs in cycling progenitor cells or their postmitotic daughter cells. Of interest, removal of Notch1 from newly postmitotic cells results in the overproduction of rods at the expense of other cell types at postnatal stages, demonstrating that input of this signaling pathway is necessary for specification of non-rod fates even after cell cycle exit (Mizeracka et al., 2013). Together with the microarray data presented here, these findings point to the idea that some newly postmitotic cells are not locked in their fate choices, and that fate acquisition in non-rod cells is a Notch1-dependent process that may occur over the course of several days.

EXPERIMENTAL PROCEDURES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. ACKNOWLEDGMENTS
  8. REFERENCES
  9. Supporting Information

Animals

Notch1fl/fl were maintained as homozygotes (Radtke et al., 1999). WT CD-1 mice were obtained from Charles River Laboratories. All experiments were approved by the Institutional Animal Care and Use Committee at Harvard University.

In Vivo and in vitro Electroporation

In vivo and in vitro electroporation were performed as previously described (Matsuda and Cepko, 2004, 2007). DNA constructs used include CAG:GFP, CAG:Cre, CALNL-GFP (Matsuda and Cepko, 2007).

Single Cell Probe Preparation and Affymetrix Array Hybridization

Single cells were isolated and profiled as described previously (Trimarchi et al., 2007, 2008). Cells were chosen based on GFP expression. Probe reactions were performed as described previously, and Affymetrix microarrays (Mouse 430 2.0 arrays) were hybridized and processed using standard Affymetrix protocols (Trimarchi et al., 2007, 2008; Roesch et al., 2008; Cherry et al., 2009). Global scaling was performed using the Affymetrix Microarray Software (MAS 5.0) and the target intensity was set to 500. The signal data for each probe set was exported for further analyses in Microsoft Excel. To eliminate probesets called marginal or absent and to reduce the false-positive rate, only probesets with a RS > 1000, as determined by MAS 5.0 were considered in this analysis. Previous reports suggest that this threshold corresponds to transcripts that are present at between 10 and 100 copies per cell (Tietjen et al., 2003). Treeview software was used to view the microarray signal data. Previously profiled cells that were classified as either amacrines or rod photoreceptors were chosen to provide examples of WT cells that were further along in their differentiation (from a previous study, Trimarchi and Cepko, unpublished, and Cherry et al., 2009). The raw and processed Affymetrix data files have been deposited in the NCBI Gene Expression Omnibus (GEO). GEO submission: GSE35682.

Fluorescence-Activated Cell Sorting Purification and Semi-quantitative PCR

Fluorescence-activated cell sorting was performed on BD Aria II sorter, gated for GFP detection. A total of 3–5 × 105 GFP+ cells were collected from two dissociated retinas for each sample. After sorting, GFP+ cells were lysed in Trizol (Invitrogen) and stored at −80°C. Phenol-chloroform extractions were performed to isolate total RNA. cDNA was generated using Accuscript High Fidelity (Agilent Technologies) according to manufacturer's guidelines. Semi-quantitative real-time PCR was performed and gene expression was normalized according to actin expression in each sample. Primers used included: actin - accaactgggacgacatggagaa, tacgaccagaggcatacagggac; Nrarp - agggccagacagcactacac, cttggccttggtgatgagat; Hes1 - acaccggacaaaccaaagac, atgccgggagctatctttct; Blimp1 - cacacaggagagaagccaca, ttgtgacactgggcacactt; Math3 - attcagggctcgaagagtca, gttccttgccagtcgaagag; NeuroD1 - gtgtcccgaggctccagggt, gggaccttggggctgaggct.

Immunohistochemistry and In Situ Hybridization

Retinas were fixed either as whole-mounts for 30 min or as eyeballs for 2 hr in 4% PFA at room temperature in 1× phosphate buffered saline (PBS). Retinas were equilibrated in sucrose/PBS solutions of increasing sucrose concentrations (5, 20, 30%), a 1:1 solution of OCT (Tissue-Tek) and 30% sucrose/PBS, and frozen on dry ice. Twenty micrometer cryosections were cut using a disposable blade on a Leica CM3050S cryostat.

For immunohistochemistry, retinal cryosections were blocked for 1 hr in 0.1% Triton, 0.02% sodium dodecyl sulfate, 1% bovine serum albumin in 1× PBS. Sections were then incubated in a humidified chamber at 4°C overnight with chicken anti-GFP (1:2,000; Abcam) diluted in blocking solution. Sections were washed in 1× PBS and incubated for 2 hr with fluorescently coupled secondary antibodies (Jackson ImmunoResearch) and DAPI (4′,6-diamidine-2-phenylidole-dihydrochloride; Sigma-Aldrich). Slides were mounted in Fluoromount-G (Southern Biotechnology Associates).

Retinas were collected at various developmental time points for in situ hybridization. Section in situ hybridization and double fluorescent in situ hybridization on dissociated cells was performed as previously described (Trimarchi et al., 2007).

ACKNOWLEDGMENTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. ACKNOWLEDGMENTS
  8. REFERENCES
  9. Supporting Information

We thank members of the Cepko, Tabin, and Dymecki labs for helpful discussions and advice. C.L.C. is an Investigator of the Howard Hughes Medical Institute.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. ACKNOWLEDGMENTS
  8. REFERENCES
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. ACKNOWLEDGMENTS
  8. REFERENCES
  9. Supporting Information

Additional Supporting information may be found in the online version of this article.

FilenameFormatSizeDescription
dvdy24006-sup-0001-suppfig1.pdf500KSupplementary Figure 1.
dvdy24006-sup-0002-suppfig2.pdf467KSupplementary Figure 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.