Formation of the eye is a complex and coordinated process that results from the interaction of cells from several distinct sources during the early embryonic period (Chow and Lang,2001). Among these cellular interactions, neuroectoderm-derived cells in the eye field, located at the anterior/ventral sides of the neural tube, undergo extensive cell proliferation and morphogenetic changes, and develop into the neural retina, as well as non-neural tissues, such as the retinal pigmented epithelium (RPE), ciliary body, and iris (Chow and Lang,2001; Davis-Silberman and Ashery-Padan,2008). The ciliary body and iris are part of the anterior segment structures of the eye and the ciliary body is required for accommodation, the process by which the lens changes shape, allowing for changes in focal length. The ciliary body and iris comprise the inner bilayered epithelia and the overlying stroma. The inner two layers originate from the inner and outer layers of the optic cup, which is derived from the ventral forebrain. The inner non-pigmented layer of the ciliary epithelium is continuous with the inner pigmented iris epithelium and the retina. The outer pigmented layer of the ciliary epithelium is continuous with the outer iris epithelium and RPE (Beebe,1986). The third layer derives primarily from the neural crest, which contributes melanocytes and connective tissue cells to the iris, as well as connective tissue and muscle to the ciliary body. However, this layer is a mixture of different cells, some of which are derived from the mesoderm (Gage et al.,2005).
The primary role of the ciliary body is the production and secretion of the aqueous humor of the posterior chamber, which then flows to the anterior chamber where it maintains the intraocular pressure (IOP) of the eye. Maintenance of the proper IOP is important both for the normal functioning of the adult eye (Gould et al.,2004) and the continued growth of the developing embryonic eye (Coulombre and Coulombre,1957). In addition to its important role in IOP establishment and maintenance, the ciliary body also is responsible for synthesizing many proteins of the inner limiting membrane, a structure that is crucial for the survival of retinal ganglion cells (Halfter et al.,2005). Intriguingly, recent studies suggest that the ciliary body and iris harbor cells with stem cell/progenitor characteristics (Ahmad et al.,1999; Tropepe et al.,2000; Sun et al.,2006).
The iris is responsible for controlling the amount of light that passes through the lens and strikes the retina, analogous to the role of the shutter in a camera. It is also involved in the circulation of the aqueous humor that is secreted by the ciliary body (Davis-Silberman and Ashery-Padan,2008). The iris is composed of several cellular layers. The pigmented epithelium of the iris is on the lens/retinal side of the tissue, abutting the outer layers, which are composed of the iris muscles and stroma. The iris root is attached directly to the ciliary body. An intriguing and unique developmental plasticity is exhibited by the iris. The pigmented, postmitotic cells of the immature iris transdifferentiate into the sphincter and dilator muscles, the muscles that open and close the iris (Volpe et al.,1993). During development, the iris cells lose their pigment, enter mitosis, and become these peripheral muscles. These iris-derived muscles are the only known ectodermally derived muscles of the body (Imaizumi and Kuwabara,1971).
Developmental abnormalities of the anterior segment structures are readily observed in humans (Idrees et al.,2006). Notably, for many of these disorders involving the anterior segment, there is a high occurrence of associated glaucoma (Gould et al.,2004). Glaucoma, a disease characterized by the degeneration of the optic nerve, is usually linked to an increase in IOP in the anterior chamber, possibly controlled by aqueous humor secretion of the ciliary body (Gould et al.,2004). Another disorder, aniridiae, is commonly caused by the mutations in the Pax6 gene and is characterized by a partial or complete loss of the iris (Davis-Silberman and Ashery-Padan,2008). While the disorders described above are the most common, there exist other ocular diseases that have been linked to developmental defects in the peripheral tissues (Gould et al.,2004; Idrees et al.,2006; Davis-Silberman and Ashery-Padan,2008), although the precise mutations causing these disorders have yet to be identified.
Although the ciliary body and iris play critical roles in vision and might harbor the source of retinal stem cells, our understanding of the molecular mechanisms governing the development of these tissues remains limited. It has been shown that the perturbation of several distinct signaling pathways can disrupt the formation of the ciliary body and iris. In particular, BMP signaling was the first to be demonstrated to have an important role in this process (Zhao et al.,2002). The ciliary body was completely absent in transgenic mice engineered to overexpress Noggin, a BMP antagonist, using a lens-specific promoter (Zhao et al.,2002). BMP signaling appears to involve FGFs as forced expression of Fgf4 from the RPE showed that there is an interplay between FGF and BMP in the generation of the ciliary body domain (Dias da Silva et al.,2007). Finally, the canonical Wnt signaling pathway has also been shown to play a crucial role in the determination of peripheral ciliary/iris epithelial fates during early optic vesicle/cup development (Liu et al.,2003,2007; Cho and Cepko,2006).
In addition to the links between ciliary body development and several signaling pathways, there are a handful of transcription factors that have been shown to be necessary for formation of the anterior segment structures of the eye (Napier and Kidson,2007). Mice that lack Otx1, for instance, have smaller irises than their wildtype counterparts and are devoid of ciliary processes (Acampora et al.,1996). Similarly, mice mutant for Lmx1b, a LIM homeodomain transcription factor, also display iris and ciliary body hypoplasia (Pressman et al.,2000). Interestingly, in humans, haploinsufficiency at the LMX1B locus results in nail-patella syndrome, an inherited condition characterized by developmental defects in the nails, kneecaps, and elbows (McIntosh et al.,1998; Vollrath et al.,1998). In many of these patients, there is also an associated rise in IOP and an increased risk for developing open angle glaucoma (Dreyer et al.,1998; McIntosh et al.,1998; Vollrath et al.,1998).
To identify candidate genes expressed in a variety of ocular tissues during development, we have profiled single cells from the developing eye. A post hoc classification scheme was used to identify the single-cell profiles based upon clusters of gene expression (Trimarchi et al.,2008). During this analysis, it became apparent that two of the single cells expressed sets of genes that distinguished them from the other profiled cells. By comparing the expression profiles of these cells to many other retinal cell types, candidate genes to determine the identity of these cells were chosen. Section in situ hybridization (ISH) both in the mouse and the chick was used to demonstrate the expression of these candidate genes in the developing ciliary body/iris. Therefore, these cells most likely originate in the developing ciliary body/iris and provide sets of marker genes expressed during the development of this tissue. In addition, we examined the expression of these genes in a chick retina overexpressing a constitutively activated β-catenin (CA-β-catenin) (Capdevila et al.,1998). Previously, CA-β-catenin was shown to convert retinal progenitor cells to more peripheral fates (Cho and Cepko,2006; Liu et al.,2007). Therefore, ectopic expression of the newly identified genes after overexpression of CA-β-catenin confirmed that these genes were useful markers of peripheral tissues and revealed those genes that could potentially act downstream of Wnt signaling.
RESULTS AND DISCUSSION
Single Cell Isolation and Retrospective Identification
Retinas were isolated from several stages of mouse development, ranging from embryonic day 12.5 (E12.5) to postnatal day 0 (P0) (Trimarchi et al.,2007). These time points were chosen to maximize the number of retinal progenitor cells (RPCs) harvested and profiled (Trimarchi et al.,2008), as well as to provide additional cells such as newborn neurons for comparison. Subsequently, the retinas were dissociated into single cells and individual cells were collected and lysed. The mRNAs from these lysed cells were reverse transcribed, and the resulting cDNAs were PCR amplified (Brady and Iscove,1993; Tietjen et al.,2003; Trimarchi et al.,2007). The cDNA quality was assessed through visual inspection on an agarose gel and those cDNA preparations that displayed a smear centered at 1 kb were chosen for further analysis (data not shown). Ten micrograms of cDNA from each single cell were biotinylated and hybridized to an Affymetrix mouse 430 2.0 oligonucleotide array using standard Affymetrix protocols (Trimarchi et al.,2007). These arrays allowed for the sampling of the expression of over 40,000 probesets corresponding to approximately 20,000 genes (www.netaffx.com), or almost the entire mouse transcriptome. The Affymetrix signal data were collected for each single cell and normalized using the Affymetrix MAS 5.0 software.
Since the retinal cells isolated during these developmental times had not yet adopted their mature morphologies, a post hoc strategy was employed to classify each cell based upon the expression of clusters of genes (Trimarchi et al.,2008). These gene clusters were generated using a Fisher's exact test to identify genes that are associated with previously characterized marker genes of the different retinal cell types. This classification method allowed for the identification of 21 single cells as developing retinal ganglion (RGC), amacrine (AC) or photoreceptor cells (PR) (Trimarchi et al.,2007) and an additional 42 cells as cycling RPCs (Trimarchi et al.,2008). However, a number of the single cells did not score significantly for any of the retinal cell types. Also, in the course of classifying these single-cell profiles, one of the additional methods used to evaluate the data involved a visual inspection in Microsoft Excel. During this examination of the single-cell expression data, it became apparent that two cells (E12 cell A6 and P0 cell E8) expressed numerous transcripts that were not detected in the majority of the other single-cell profiles (Fig. 1). Given these readily apparent differences in gene expression, the profiles of these two cells were subjected to further scrutiny and validation to assess precisely which type of cell(s) produced these profiles.
Analysis of Transcripts Expressed in the Mouse Peripheral Margin of the Optic Cup and Developing Ciliary Body/Iris
Visual inspection of the microarray data revealed that several genes were present in both E12 cell A6 and P0 cell E8 and largely absent from the previously examined developing RGCs and ACs (Fig. 1). However, a further examination of the single-cell profiles from these two cells revealed that they shared some expression in common with RPCs (Fig. 1 and data not shown). This encompassed a range of expression levels such that some genes were broadly expressed in RPCs (for example, see Sox2), some were only expressed in a small subset of RPCs (for example, see Igf2), and some were not found in the RPC profiles (for example see RIKEN cDNA B130021B11) (Fig. 1). Despite sharing some gene expression in common with RPCs, these two cells did not score significantly for RPC character using our previously developed classification scheme (data not shown). Additionally, for many of the RPC expressed genes, the expression levels found on the Affymetrix microarrays were lower in these two cells than in the cells that scored highly as RPCs in the classification scheme (data not shown).
Included among the genes whose expression was enriched in these two cells were cell cycle regulators (Cyclin D2), signaling molecules (Igf2 and Notch2), and transcription factors (Zic1 and Zic2). To identify additional genes that were associated with these markers, hierarchical clustering was performed using E12 cell A6, P0 cell E8, and a set of 21 previously characterized developing RGCs, ACs, and PRs (Trimarchi et al.,2007). An examination of the genes that clustered with cyclin D2 with a correlation coefficient of >0.75 revealed several of the genes (Igf2, Notch2, and Zic1) that had been identified by visual inspection (Fig. 1 and Supp. Fig. S1A, which is available online). Many genes that were not readily apparent by visual inspection were also identified through clustering. These genes included, for example, a cell surface molecule (CD38 antigen) that has been shown to be present in both the pigmented and non-pigmented ciliary body (Khoo and Chang,1999) and many uncharacterized RIKEN cDNAs (Supp. Fig. S1A).
To more fully explore the identity of these single cells and to validate the microarray findings, ISH was performed on cryosections of E12.5, E16.5, and P0 mouse eyes. Prominent expression was detected either in the peripheral margin of the optic cup or in the developing ciliary body for each of the genes from this list that was tested (Fig. 2 and Supp. Fig. S2). It was somewhat surprising that cells from the developing ciliary body/iris would have been isolated given that they represent a small percentage of the total cells at any given stage. However, in situ hybridizations performed on dissociated cells have shown that cells from the ciliary body/iris are present in these preparations from all three stages examined (data not shown). Also, previous single-cell profiling analyses from these time points examined developing ganglion and amacrine cells, which would also have been present at a low percentage in the developing retina at these time points (Trimarchi et al.,2007).
At all three stages, Cyclin D2 expression was observed in the lens and in the developing ciliary body (Fig. 2A–C). No significant expression was detected in the central portion of the retina at E12.5 and E16.5, while some faint staining was observed in the outer neuroblastic layer (ONBL), where the cycling progenitor cells are located, at P0 (Fig. 2C). Likewise, the transcript for the growth factor Igf2 was found enriched at the peripheral margin of the optic cup at E12.5 and then in the developing ciliary body at E16.5 and P0 (Fig. 2D–F). Staining was also detected for Igf2 in the lens (E12.5, Fig. 2D) and the optic nerve head (E16.5, Fig. 2E). H19, a non-coding RNA that is located in the same chromosomal region as Igf2 and is reciprocally imprinted with Igf2 (Zemel et al.,1992), was similarly found in the lens, the developing periphery of the optic cup (Fig. 2G), and the developing ciliary body (Fig. 2H and I). However, unlike Igf2, robust expression of H19 also was observed in a subset of cells in the retinal ONBL at E16.5 and P0 (Fig. 2 HI, Supp. Fig. S3A,B). These cells were predicted to be RPCs based upon the presence of H19 transcripts in the profiles of many RPCs at E12.5 and P0 (Fig. 1). To fully characterize the identity of these additional H19+ cells, double dissociated ISH was performed for H19 and Fgf15, a previously characterized marker of RPCs (Blackshaw et al.,2004; Trimarchi et al.,2008), on dissociated P0 retinas that had been labeled with [3H]-thymidine for 1 hr. There was no overlap observed between H19+ cells and [3H]-thymidine+ cells (Supp. Fig. S3C,D), while a robust overlap was seen between Fgf15 and [3H]-thymidine (Supp. Fig. S3F). These data indicate that the H19+ cells in the ONBL were likely not progenitor cells in the S phase of the cell cycle. Intriguingly, many of the H19+ cells were also Fgf15+ (Supp. Fig. S3E) suggesting that these cells may be RPCs in either the G1 or G2 phase of the cell cycle.
Transcription factors represented another category of genes detected in these two single-cell profiles. Zic1, for example, was observed in both E12 cell A6 and P0 cell E8, not in any of the developing RGCs or ACs, and in a subset of RPCs (Fig. 1). ISH experiments on retinal cryosections confirmed that Zic1 mRNA was found in the peripheral margin of the optic cup at E12.5 (Supp. Fig. S2A) and then in the developing ciliary body at E16.5 and P0 (Supp. Fig. S2B,C). Additionally, transcripts for the related family member, Zic2, were also enriched in the developing ciliary body/iris, but unlike Zic1, staining for Zic2 was also detected throughout the ONBL (Supp. Fig. S2D–F). These data are consistent with previous reports detailing the retinal expression of these zinc finger transcription factors (Nagai et al.,1997).
Uncharacterized transcripts also were observed in these two single-cell profiles (Fig. 1 and Supp. Fig. S1). ISH for one of these (RIKEN cDNA 5730449L18/riboprobe AW489153) showed expression in the optic nerve head at E12.5 (Fig. 2J) and subsequent expression in the developing ciliary body/iris at E16.5 (Fig. 2K) and P0 (Fig. 2L). Additional expression was noted in a subset of cells on the vitreal side of the developing inner neuroblastic layer (INBL), consistent with expression in developing RGCs (Fig. 2K and L).
Hierarchical clustering produced a group of genes that associated significantly with cyclin D2 (Supp. Fig. S1A). To validate this clustering and more fully examine the expression patterns of some of the cyclin D2-associated genes, ISH were performed for ninjurin1 and rhomboid family 1 (Rhbdf1). Ninjurin1 is an adhesion molecule originally identified through its role in nerve regeneration (Araki and Milbrandt,1996), which could have a similar adhesive role in the developing ciliary body. The expression of ninjurin1 was very weak at E12.5 in the peripheral margin of the optic cup, but was not found in the retina itself (Fig. 2M). At E16.5 and P0, ninjurin1 transcripts were detected in the developing ciliary body/iris and in the ONBL (Fig. 2N and O). Rhbdf1 is an atypical member of a family of serine proteases, the Drosophila homolog of which has been demonstrated to activate growth factors (Urban,2006). However, while Rhbdf1 has been shown to associate with growth factors, it lacks key residues responsible for protease activity and, therefore, its function remains to be fully determined (Nakagawa et al.,2005). Similar to ninjurin1, Rhbdf1 was detected weakly in the peripheral margin of the optic cup at E12.5 (Supp. Fig. 2G) and then stronger signal was found in the developing ciliary body/iris at both E16.5 and P0 with weaker staining in the ONBL at E16.5 (Supp. Fig. 2H and I). These ISH experiments validate that the cluster of genes that associate with cyclin D2 are expressed in the developing ciliary body/iris and provide new markers for this tissue.
There were transcripts that were enriched in both E12 cell A6 and P0 cell E8 that also were present in subsets of the developing RGCs and ACs (Fig. 1) or in subsets of RPCs (data not shown). These genes included a cell-cycle regulator (cyclin D3), signaling molecule (Tgfb2), an IGF-binding protein (Igfbp5), and transcription factors (Sox2, Myc). ISH performed for several of these transcripts confirmed that they were expressed in the developing ciliary body/iris (Fig. 2 and data not shown), while also having other domains of expression. Tgfb2 mRNA, for instance, was detected in both the ONBL and the developing ciliary body/iris at E12.5, E16.5, and P0 (Fig. 2P–R). The staining in the ciliary body/iris was consistently more intense than that in the ONBL at E16.5 and P0 (Fig. 2P and R). Additionally, at P0 the staining for Tgfb2 mRNA was more pronounced in the outer layer of the developing ciliary body/iris than the inner layer (Fig. 2R). Cyclin D3 displayed a very similar staining pattern as Tgfb2 in that it was also significantly enriched in the ciliary body/iris at E16.5 and P0 while being present in the ONBL as well (data not shown). Myc, however, was expressed in the peripheral margin of the optic cup at E12.5 and then at E16.5 and P0 was more strongly expressed in subsets of cells in the most vitreal portion of the INBL, consistent with expression in developing RGCs (data not shown). Transcobalamin 2 (Tcn2) was found in the profiles for both of the developing ciliary body single cells and was absent from developing RGC and AC profiles (Fig. 1). However, Tcn2 was detected in a subset of the RPC single-cell profiles (Fig. 1). ISH performed on retinal cryosections showed intense staining for Tcn2 in the retinal pigment epithelium (RPE) at E12.5 and P0 and weaker staining in the ONBL (Fig. 3A and B). The ISH also revealed Tcn2 expression in the developing ciliary body/iris, which was especially pronounced at E16.5 and P0 (Fig. 3B and data not shown).
Additionally, transcripts were observed that were present in only one or the other of the putative developing ciliary body single cells (Fig. 1). As observed with the sets of genes characterized above, the categories of these transcripts included transcription factors (Foxp2, Otx1, Tcf3, Etv1), cell-cycle regulators (Wfdc1, cyclin E1), and signaling molecules (Fgf14, Bmp7, Ltbp1). Finding associated genes with a significant correlation coefficient through hierarchical clustering was more difficult for these genes since they were only expressed in one of the single-cell profiles. Nonetheless, one cluster did emerge around the genes Foxp2 and Aldh1a3 (Supp. Fig. S1B). As was the case with the clusters of genes observed above, many of the genes associated with Foxp2 were uncharacterized RIKEN cDNAs and other expressed sequence tags (Supp. Fig. S1B).
ISH on retinal cryosections from three developmental stages confirmed the expression of the genes that clustered with Foxp2 and Aldh1a3 in the peripheral margin of the optic cup at E12.5 and in the developing ciliary body/iris at E16.5 and P0 (Fig. 3 and data not shown). A riboprobe for the growth inhibitor Wfdc1 (Larsen et al.,1998) showed strong staining in the peripheral margin of the optic cup at E12.5, and while additional staining was detected in cells located in the ONBL, the most central part of the retina was not stained (Fig. 3C). At subsequent developmental stages, Wfdc1 mRNA became localized only to the developing ciliary body (Fig. 3D and data not shown). An additional growth inhibitory gene, Gas1 (Del Sal et al.,1992), was also found in the single-cell profile for E12 cell A6 (Fig. 1). ISH confirmed that Gas1 was expressed in the peripheral margin of the optic cup at E12.5 (Fig. 3M) and then became restricted to the ciliary body/iris at E16.5 (Fig. 3N). It is curious that several growth inhibitory genes (Gas1, Wfdc1, and p57) are expressed in the same cell as genes that stimulate cell growth (several cyclin genes, Myc, and Notch2) (see Fig. 1; Rowan et al.,2004). While it is possible that the explanation is simply a lack of correlation between mRNA expression and protein expression, this observation warrants further study to understand the interplay between genes that drive the cell cycle and those that impede it in the developing ciliary body/iris.
Aldh1a1, which codes for an enzyme that catalyzes the second step of the conversion of retinoic acid to retinol (Duester,2000), was found to be strongly expressed in the lens and in the ONBL of the dorsal retina at E12.5 and E16.5 (Fig. 3E and data not shown), as previously reported (Fan et al.,2003). Additionally, this gene was observed in both the dorsal and ventral developing ciliary body at all three stages (Fig. 3E and F). Aldh1a3, a related member of the aldehyde dehydrogenase family, was observed in E12 cell A6, much like Aldh1a1 (Fig. 1). ISH on retinal cryosections for Aldh1a3 showed that while there was robust expression in the peripheral margin of the optic cup at E12.5, no expression was detected in the developing ciliary body/iris at P0 (data not shown), unlike in the case of Aldh1a1. Another enzyme, Dhrs8, a member of the alcohol dehydrogenase family, was initially observed in a subset of cells in the dorsal portion of the ONBL at E12.5 (Fig. 3G), but then became expressed in the ciliary body at E16.5 and P0 (Fig. 3H and data not shown).
Several transcription factors were found in the single-cell profile of E12 cell A6, including Foxp2, Otx1, and Tcf3 (Fig. 1). Otx1 has been reported previously to be expressed in the developing ciliary body (Martinez-Morales et al.,2001) and ISH on retinal cryosections demonstrated robust expression of this transcript in both the E12.5 peripheral margin of the optic cup and the E16.5 ciliary body (Fig. 3I and J). Foxp2 also was detected in the peripheral margin of the optic cup at E12.5 and the developing ciliary body at E16.5, but this gene also displayed additional staining in a subset of cells in the INBL (Fig. 3K and L). The location of these cells to the scleral side of the INBL indicated that they might be developing ACs (Fig. 3L). Uncovering the distinct transcription factors expressed in the developing ciliary body/iris will enable future studies dissecting the specific transcription factor networks driving the differentiation of this region.
Several additional genes were found to be expressed in one of the putative two ciliary body cells. Insulin receptor substrate proteins (Irs1–4) dock with the insulin receptor upon its activation and phosphorylation and act as bridging molecules with the downstream signaling pathway (Giovannone et al.,2000). While the other members of this family play roles in growth and glucose metabolism, the function of Irs4 is not well understood (Fantin et al.,2000). Irs4 was found in the transcriptional profile for E12 cell A6 (Fig. 1) and ISH at E12.5 showed robust expression in the peripheral margin of the optic cup (Supp. Fig. S2J) and the developing ciliary body/iris at E16.5 (Supp. Fig. S2K). The expression of Irs4 was absent from both the retina and the developing ciliary body at P0, suggesting this gene may play a role in early ciliary body development. The transcript for Ltbp1, which encodes a protein that is part of a larger complex that binds Tgfβ family members and keeps them inactive while assisting in their assembly and targeting (Oklu and Hesketh,2000), was also found in the profile for E12 cell A6 (Fig. 1). ISH for Ltbp1 revealed that this gene was expressed in the peripheral margin of the optic cup at E12.5 and then in the developing ciliary body/iris at E16.5 and P0 (Supp. Fig. S2M-O). Vinculin, which encodes an actin-binding protein that plays a role in cell/cell and cell/extracellular matrix adhesion (Ziegler et al.,2006), also was observed in E12 cell A6 (data not shown). ISH further demonstrated that this gene was expressed in a subset of cells in the ONBL at E12.5 (Fig. 3O), but then was detected in the outer layer of the developing ciliary body at E16.5 and P0 (Fig. 3P and data not shown). Additional strong staining was observed in the developing iris tip, both in the mouse at P0 and in the chick at E6 and E8 (Fig. 3P, Supp. Fig. S2P and R).
Nephronectin (Npnt) is an EGF-repeat-containing protein that has been identified in tissues undergoing an epithelial-mesenchymal transition during development (Brandenberger et al.,2001; Morimura et al.,2001). Transcripts for Npnt were found in the profile for P0 cell E8, but not E12 cell A6 (Fig. 1). ISH on E12.5 retinal cryosections revealed strong staining for Npnt in the RPE (Fig. 2S). At E16.5, the RPE staining decreased considerably and the most prominent signal was found in the tip of the developing ciliary body and in the developing iris (Fig. 2T). This pattern continued at P0 with weak staining for Npnt in the ciliary body and more pronounced signal in the developing iris (Fig. 2U). The expression pattern for Npnt suggests that P0 cell E8 may originate from a different portion of the developing ciliary body/iris than E12 cell A6 and, therefore, provides distinct marker genes for this region.
BMP7, a potent signaling molecule that has been shown to play an important role in the development of the mouse ciliary body (Zhao et al.,2002), was found in the single cell profile for P0 cell E8 (Fig. 1). ISH on retinal cryosections showed that BMP7 was expressed in the peripheral margin of the optic cup at E12.5 (Fig. 3Q). In addition, BMP7 was found in the lens and in a subset of cells in the central retina, located in a position consistent with newborn RGCs at this stage (Fig. 3Q). At E16.5, BMP7 was expressed robustly in the ciliary body/iris, but now was also found in a subset of cells in the ONBL (data not shown). By P0, BMP7 retinal expression was confined to the developing ciliary body/iris (Fig. 3R). These microarray-profiling experiments coupled with the ISH validation demonstrate the dynamic nature of gene expression in the developing mouse peripheral margin of the optic cup and ciliary body/iris and the extensive diversity of the classes of genes that are expressed in this region of the developing eye.
Analysis of Transcripts Expressed in the Developing Chick Peripheral Margin of the Optic Cup
It has been suggested that the intricate morphogenetic events that occur during ciliary body morphogenesis may be quite different between the mouse and the chick (Napier and Kidson,2007). Given this potential for both similarities and differences in how the anterior segment structures form in these organisms, a comparison in gene expression between mouse and chick peripheral retina might be informative. To identify the genes that also were expressed in the periphery of the developing chick eye, ISH was performed on cryosections from E4, E6, and E8 (Fig. 4A). These stages were chosen to most closely reflect E12.5, E16.5, and P0 stages of mouse development. The genes examined included those identified from the single-cell mouse microarrays (Fig. 1), and genes previously shown to be expressed in the developing peripheral mouse eye (Rowan et al.,2004). Among the genes identified from mouse, the transcripts of the signaling molecules BMP4 and BMP7 (Fig. 4B and C), the Wnt receptor, Frizzled 4 (Fig. 4D), transcription factors such as FoxP2, Zic1, Otx1, and TFEC (Fig. 4E–G,I), cell-cycle regulators such as GAS1, WFDC1, SGK, c-Myc, and Cyclin E1 (Fig. 4H,K,L, and data not shown), and other genes such as Transcobalamin 2, Clusterin, and Collagen IX (Fig. 4J and data not shown) were enriched in the periphery of the optic cup between E4 to E8. In general, ISH signals specific to the periphery of the retina were detected with two different probes corresponding to different parts of the gene (Supp. Table 1 and data not shown). The expression was highest at the tip of the optic cup and decreased toward the center of the retina. In addition, the intensity of the signal decreased over time, so that the expression at E8 was very low in most cases. It is not clear whether this was due to the profound thinning of the ciliary and iris epithelia at this time, or a reduction in transcript levels.
The kinetics of the expression was variable during these developmental stages as well. For some genes, the ISH signal intensified over time, whereas in most cases the expression at E4 and E6 was stronger than that of E8. The extent of expression in the periphery varied at a given stage. For example, TFEC and SGK (Fig. 4I, I1, I2,L, L1, L2) displayed a narrow expression domain located at the tip of the optic cup, which is thought to develop into the iris, while most other genes showed a much broader expression pattern in the periphery of the optic cup at E4.
Although the majority of the genes tested were expressed in both the mouse and chick, some of the genes (cyclin D2, Igf2, cyclin D3, IGFBP5, Crhbp, REST, FGFR2, and CD63) expressed in the mouse were either weakly expressed, not expressed, or expressed in different parts of the developing optic tissues of the chick embryo (Supp. Table 1). Because several genes of this class showed expression in other tissues surrounding the developing optic cup, the absence of expression is likely to be the result of species differences, rather than technical issues. For example, cyclin D2, CD63, and FGFR2 showed strong expression in the lens and periocular mesenchyme (data not shown), whereas IGF2 showed broad expression in the ventral optic cup (data not shown). This may indicate that there are molecular distinctions in the ciliary body and iris development between mouse and chick. For example, it is known that the mechanism of periocular mesenchyme development is significantly different between mouse and chick. In this tissue, the neural crest- and mesoderm-derived cells differentially contribute to the periocular mesenchymal cells surrounding the eye in mouse and chick (Gage et al.,2005).
Ectopic Expression of Ciliary Body Markers Upon Induction of the Wnt Signaling Pathway
The canonical Wnt signaling pathway plays an important role during early specification of the ciliary body and iris in the periphery of the developing optic vesicle/cup (Liu et al.,2003,2007; Cho and Cepko,2006). In order to investigate whether the peripheral genes might be downstream of Wnt signaling, the expression of these genes was examined in the retina following overexpression of a constitutively active form of beta-catenin (CA-β-catenin) (Capdevila et al.,1998; Cho and Cepko,2006). In previous experiments, an avian replication-competent retroviral vector encoding an N-terminal deletion mutant of β-catenin (RCAS:CA-β-catenin), which rendered it constitutively active, was introduced into the chick retina. Infection with this virus led to retinal progenitor cells adopting a more peripheral retinal fate. Additionally, downregulation of retinal genes and upregulation of several previously known markers of the ciliary body/iris was observed (Cho and Cepko,2006; Fu et al.,2006).
To examine which of the newly identified ciliary body markers was potentially downstream of Wnt signaling, RCAS:CA-β-catenin was used to infect the optic vesicle at stage 10. The infected retinas were harvested at E9.5, and sectioned and stained with an anti-gag antibody (3C2) to determine the extent of the infected region that also expressed CA-β-catenin (Fig. 5A–D). Retinal epithelia without 3C2 staining displayed both a normal thickness and a normal pattern of neurogenesis as determined by anti-β-tubulin staining (Fig. 5C), whereas 3C2+ areas showed a gradual diminution of β-tubulin staining and a thinning of the retina (Fig. 5A–D). Since the correlation between the retinal thinning and the induction of ciliary body/iris fate was highly reproducible, these sections were examined for the expression of the newly identified ciliary body/iris markers.
To insure that the viral overexpression assay was functioning properly, the ectopic induction in the retina of a few previously identified iris/ciliary body marker genes was examined. These genes included the secreted signaling molecule (BMP7, Fig. 5G,H), the cell-cycle regulator (WFDC1, Fig. 5W,X), and the matrix protein (Collagen IX, Fig. 5 AA, AB). The expression of a Wnt receptor (Fzd4) and a Wnt signal transducer (Lef1) have been characterized previously in the peripheral retina (Jasoni et al.,1999; Kubo et al.,2003). To determine whether the expression of Fzd4 and Lef1 was potentially downstream of Wnt signaling, Stage-10 optic vesicles were infected with RCAS:CA-β-catenin and eyes were harvested at E9.5. Both Fzd4 (Fig. 5Q,R) and Lef1 (Fig. 5Y,Z) were ectopically upregulated in the central retina upon activation of the Wnt pathway. This upregulation of Lef1 and Fzd4 may be caused by a positive feedback mechanism of β-catenin-mediated Wnt signaling (Kubo et al.,2003; Liu et al.,2003).
The newly characterized genes identified from the combined microarray analysis/section ISH from the mouse, described above, and additional periphery-enriched genes reported previously (Rowan et al.,2004) also were examined for their responsiveness to Wnt signaling. All of the results were observed in ISH on sections from multiple RCAS:CA-β-catenin-infected retinas (minimum of two). In a similar fashion as Fzd4 and Lef1, infection with RCAS:CA-β-catenin led to an upregulation of the secreted signaling molecule BMP4 (compare Fig. 5E to F), the transcription factors FoxP2, Otx1, Zic2, and TFEC (compare Fig. 5I to J, K to L, M to N, and O to P, respectively), and the cell-cycle regulators Gas1 and SGK (compare Fig. 5S and U to T and V, respectively). In general, the induction of these markers was present across the entire range of the thinner retina except in a few cases where the upregulation was confined to only a small area of the thinner epithelium (see Fig. 5F,J,N). Additionally, in any given retinal section, a similar upregulation was observed in any part of the epithelium where thinning had occurred. The absence of a larger induction in these cases may be due to a requirement of higher levels of Wnt signaling to attain full transcriptional upregulation. Ectopic induction of negative cell-cycle regulators, such as SGK, Gas1, and WFDC1 suggests that canonical Wnt signaling in the developing periphery of the optic cup may participate in the suppression of cell proliferation. This interpretation is consistent with previous results demonstrating that the cell proliferation index of the peripheral retina is lower than that of the central retina, and is lower in the central retina following infection with the RCAS:CA-β-catenin (Cho and Cepko,2006).
Since infection with the RCAS:CA-β-catenin virus leads to a repatterning of the eye, such that the central retina takes on the peripheral fate (Cho and Cepko,2006), the observed ectopic expression in the central retina of these candidate genes may be either downstream of Wnt signaling directly, or simply be a consequence of the fate change. Either way, these experiments provide further validation of these newly identified as bona fide peripheral markers. They have now also been shown to be potential downstream targets of the Wnt signaling pathway. Their exact position in the hierarchy of gene expression downstream of Wnt signaling will need to be established by further analysis.
Comparison of Ciliary Body Marker Studies
Several strategies have been employed to identify the genes specifically expressed in the peripheral structures of the eyes from various animals. These approaches have included ISH screens, subtractive hybridization strategies, and microarrays using mouse, human, and chick anterior segment structures (Thut et al.,2001; Escribano and Coca-Prados,2002; Kubota et al.,2004; Diehn et al.,2005). In the current study, a combination of single-cell transcriptional profiling and ISH was used to identify markers of the ciliary body in the developing mouse and chick retina. A previous high-throughput ISH screen in the developing mouse eye identified 6 genes as having specific expression in the optic cup margin, the location of the developing ciliary body (Thut et al.,2001). Three of these markers (Tgfb2, Col9a1, and Slc16a1) were expressed in both ciliary body single-cell profiles and observed in very few single RGC, AC, or PR profiles (Supp. Fig. S4A). Tac1 was not observed in either putative single ciliary body cell and Ptmb4 could not be located on the Affymetrix microarray (Supp. Fig. S4A). The strongest marker for the developing ciliary body found in the previous study of Thut et al. was Tgfb1i4 (Thut et al.,2001). Another group further characterized the expression of this marker in more mature peripheral structures (Napier and Kidson,2007). Tgfb1i4 was not identified as a ciliary body marker in our screen since robust expression of this gene was also detected in all of the single RGCs and ACs profiled between E12.5 and P0 (Supp. Fig. S4A). To examine whether Tgfb1i4 was expressed outside of the developing ciliary body, ISH were performed on E12.5, E14.5, E16.5, and P0 retinas. Strong staining was observed at each of these stages in the developing ciliary body/iris (Supp. Fig. S4B-E). However, at E14.5, E16.5, and P0, significant signal also was present in the INBL (Supp. Fig. S4C–E), where the developing RGCs and ACs reside, consistent with the single-cell microarray results. While Tgfb1i4 is strongly expressed in the developing ciliary body, this gene does not mark this area specifically and is expressed in other nearby retinal cells as well.
Additional studies searching for markers of human ciliary body structures have been performed. These analyses utilized either a microarray-based approach (Diehn et al.,2005) or EST database mining (Escribano and Coca-Prados,2002) to produce lists of candidate ciliary body–specific genes. To examine the expression of these sets of marker genes in our single cell transcriptional profiles, heat-maps were generated with the genes identified by EST mining (Supp. Fig. S5A) and by microarray analysis (Supp. Fig. S5B). In general, the markers found by EST mining were expressed in many more single cells than merely the putative ciliary body cells (Supp. Fig. S4A). Additionally, not all of these markers were expressed in either E12 cell A6 or P0 cell E8. Conversely, the majority of the markers identified in the microarray screen were not expressed in either of the two single cells from this study that may have originated from the developing ciliary body or in any of the other single-cell profiles (Supp. Fig. S5B). The two exceptions were Tpm2 and Bmp7. The principal reason behind these differences is most likely that these other studies were conducted using adult human tissue, whereas the current study focused solely on the developing mouse and chick ciliary body.
Finally, one study was conducted to find genes specifically expressed in the chick ciliary epithelium. A combination of subtractive hybridization and cDNA microarrays were used to identify candidate genes and ISH was employed subsequently to validate the specificity of expression of these candidate markers. The majority of the genes identified as specific markers of the ciliary epithelium by Kubota et al. (2004) did not overlap with those from our study. Only three genes (Zic2, Tgfb2, and Collagen IX) were commonly isolated by the two approaches. This is most likely due to the differences in stages of the tissue samples taken for the comparison. In our experiments, embryonic and early postnatal cells were used for the analysis in mice, and E4 to E8 eyes were used for the chick analysis, whereas Kubota et al. (2004) used posthatch day-1 chick eyes. This stage is after most of the development of the eye structures is complete. Alternatively, the failure to identify known markers of the ciliary body could be due to some technical shortcomings in the protocols used. For example, the Msh homeobox-containing genes Msx1 and Msx2 are expressed in the developing ciliary body at multiple developmental time points (Monaghan et al.,1991). However, these genes were not detected in the transcriptional profiles of the two presumptive ciliary body cells in this study. In fact, these two genes were never detected in any of the single cells profiled to date, suggesting that either they are expressed at levels that are too low to be sampled by the single-cell profiling method or that the target sequences on the Affymetrix microarray fail to hybridize with the cDNAs for these genes.
Single-Cell Harvesting and cDNA Generation
Single cells were isolated from CD-1 mice (Charles River Laboratories) exactly as described in (Trimarchi et al.,2007). Briefly, retinas were dissected, dissociated in a papain solution, and plated in PBS (pH 7.4) containing 0.1% BSA. Each individual cell was harvested into a drawn glass pipette using capillary action, washed, and collected into lysis buffer (10 mM Tris-HCl [pH 8.3], 50 mM KCl, 1.5 mM MgCl2, 5 mM DTT, 0.5% NP-40, 200 ng/ml modified Oligo dT primer [TATAGAATTCGCGGCCGCTCGCGATTTTTTTTTTTTTTTTTTTTTTTT] [Oligos etc.], 250 U/ml RNAguard [Amersham], 15 U/ml Prime RNase inhibitor [Eppendorf], and 40 μM each of dATP, dGTP, dCTP, and dTTP). The mRNAs were reverse transcribed with Superscript II (Invitrogen) for 30 min at 37°C and a poly(A) tract was added to the first-strand synthesis using TdT (Roche). The second strand was generated and the subsequent mixture PCR amplified using the following program: 95°C for 2 min, 37°C for 5 min, 72°C for 16 min, 93°C for 40 sec, 67°C for 1 min, 72°C for 6 min, with a 6-sec extension each cycle for 35 cycles.
Affymetrix Microarray Hybridizations
Ten micrograms of single-cell cDNA was RNase treated (Roche) for exactly 13 min at 37°C. The samples were heated to 99°C for 15 min and end labeled with 25 μM Biotin N6 ddATP (Enzo Biosciences). Mouse Genome 430 2.0 Genechip® oligonucleotide arrays were hybridized with the labeled single-cell cDNAs using standard Affymetrix protocols as detailed in Trimarchi et al. (2007). Global scaling of each array dataset was performed using the Affymetrix Microarray software (MAS 5.0) and the target intensity was set to 500. The resulting signal data for each probe set was exported as a tab delimited text file and subsequent analyses were performed using Microsoft Excel. The raw and processed Affymetrix data files have been deposited in the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) under the accession number GSE15566.
In Situ Hybridizations and Immunohistochemistry
In the mouse, both ISH on retinal cryosections and dissociated cell ISH were performed exactly as detailed in Trimarchi et al. (2007,2008).
For the chick ISH, freshly fertilized chicken eggs were purchased from SPAFAS (CT) and incubated at 37°C for 4, 6, and 8 days to embryonic day (E) 4, 6, and 8 for section ISH. Twenty-micron frozen cryosections were generated from either whole heads or eyes that had been fixed in 4% paraformaldehyde. ISH were performed as previously described (Murtaugh et al.,1999).
Immunohistochemical detection of RCAS:CA-β-catenin-infected cells and β-tubulin was done as described previously (Cho and Cepko,2006). Primary antibodies (α-β-Tubulin III, Upstate Biotechnology, and 3C2 mAB, Developmental Studies Hybridoma Bank; DSHB) were used at 1:250 and at 1:50, respectively. Secondary antibodies were Cy2- and Cy3-conjugated anti-rabbit antibodies or anti-mouse (Jackson Immunoresearch Laboratories) used at 1:300.
Chicken Embryos and In Ovo Injection
For the in ovo injections, concentrated viral particles encoding CA-β-catenin (RCAS:CA-β-catenin) (Capdevila et al.,1998) were injected into the right optic vesicle of the Hamilton-Hamburger (HH) stage-10 (Hamburger and Hamilton,1992) chicken embryo as described previously (Cho and Cepko,2006) and eggs were further incubated to E9.5. All experimental procedures with embryos were approved by the Institutional Animal Care and Use Committee at Harvard University.
This work was supported by the Howard Hughes Medical Institute (C.L.C.).