Gene expression profile of single Müller glial cells
Adult murine retinae were dissociated to single cells by using papain digestion. Within less than 2 hours of dissection, individual Müller glial cells were identified by their distinctive morphology (Fig. 1) and picked from a dish of dissociated cells with a micropipette. They were then washed, lysed, and subjected to reverse transcription. After a terminal transferase reaction to add A's to the 3′ end of the cDNA, 35 rounds of PCR were carried out by using modified oligo dT primers. cDNA preparations from five individual Müller glial cells were hybridized to Affymetrix Genechip® Mouse genome 430 2.0 arrays. These arrays provide almost complete coverage of the mouse genome with 45,000 probe sets. The identity of cells as Müller glial cells was confirmed by expression of known marker genes for Müller glia. Glutamine synthetase (Glul), clusterin (Clu), dickkopf homolog 3 (Dkk3) (Blackshaw et al.,2004), and S100 calcium binding protein A16 (Seigel et al.,1996) were successfully detected (Fig. 2 and Supplemental Fig. S1). All of the cells share expression of these key marker transcripts.
Figure 1. Prospective identification of single Müller glial cells by morphology. Micrograph of cells dissociated from an adult mouse retina. A: Immunostaining for glutamine synthetase (Glul). B: Dissociated cell in situ hybridization for clusterin. C: Brightfield image. D: Merged image. Scale bar = 10 μm in A (applies to A–D).
Download figure to PowerPoint
Figure 2. Comparison of highly expressed transcripts in Müller glia with other retinal cell types. The 100 most highly expressed transcripts, averaged over the five single Müller glial cells (MGs) and sorted in decreasing order, are illustrated in a heat map and compared with their expression in other cell types. Values above 10,000 are represented in dark purple. Rows correspond to different genes (with abbreviated name and Affymetrix ID), and columns represent the single cell probes. Shown are 19 immature amacrine (ACs) and ganglion cells (RGCs), 2 immature rod photoreceptors (PRs) from postnatal day 0 (P0), and 2 mature rod photoreceptors (PRs) (Trimarchi et al.,2007).
Download figure to PowerPoint
The values for the Affymetrix signals from the five Müller glial cells were averaged (Supplemental Table T2). There were 7,377 Affymetrix IDs with averages > 1,000, of which 871 had values > 10,000. Many known Müller glial specific genes, such as Glul, aquaporin 4 (Aqp4), and Clu/ApoJ were some of the most highly expressed genes (Fig. 2), perhaps facilitating their previous discovery as Müller glial markers (Jenne and Tschopp,1992). Additionally, this comparison of highly expressed genes with the data in the literature (Nagelhus et al.,1998; Blackshaw et al.,2004) validates the general performance of the assay.
Biological processes and molecular functions were systematically assigned to genes with average expression values > 10,000 by using IPA (Ingenuity Systems, Redwood City, CA). Genes with the highest significance can be categorized into protein synthesis, cell-to-cell signaling, cellular assembly and organization, cellular function and maintenance, carbohydrate metabolism, cell morphology, and lipid metabolism (Supplemental Table T6).
Comparison of expression of Müller glia enriched transcripts to expression in other retinal cells
The Affymetrix signal values from the Müller glial cells were compared with those of several different types of single retinal single cells in order to gain an appreciation of how Müller glial cells differ from neurons and also to identify markers for Müller glial cells. This comparison included 2 immature rod photoreceptors, 2 adult rod photoreceptors, and 19 immature amacrine and ganglion cells (Trimarchi et al.,2007). The data were sorted according to decreasing average values from the five Müller glia samples (Supplemental Table 1) and are represented in Figure 2 as a heat map. The 100 transcripts expressed with the highest averaged values show an almost uniform general expression pattern across all Müller glial samples. Many genes that had high expression values in Müller glia were also highly expressed in rod photoreceptors and immature amacrine and ganglion cells. Some of these genes are involved in generic cellular functions, e.g., ribosomal subunits, eukaryotic translation elongation factor 1 α 1 (Eef1a1), ubiquitins (Ubb and Ubc), and glyceraldehyde-3-phosphate dehydrogenase (Gapd).
This initial comparison revealed a dozen genes that appear to be preferentially expressed by Müller glia, including the tRNA ligase BING4 and the transcription factor Sox2. They also include the lipid transporter ApoE, the major apolipoprotein in the central nervous system (Pitas et al.,1987). Interestingly, ApoE has also been implicated in cell proliferation (Mahley,1988) and in modulation of the innate and acquired immune response (Laskowitz et al.,2000). Another gene involved in lipid metabolism that was found enriched in Müller glia is the diazepam binding inhibitor (Dbi), which is also known to be expressed in glial cells of both the central and peripheral nervous systems (Yanase et al.,2002). Other Müller glia enriched genes are the water channel, Aqp4, and Abca8a, a transporter with ATPase activity, the isoprenoid binding protein Rlbp1, and the receptors Gnai2 and Gpr37. Gnai2 (also called Gi2) mediates signaling from the endothelin-B receptor to maintain mitogenic activity of neural progenitors in the brain (Sinohara et al.,2004).
Expression of the Gpr37 and Dkk3 transcripts in Müller glia was shown in a previous study using serial analysis of gene expression (SAGE; Blackshaw et al.,2004). Dkk3 is a gene involved in cell differentiation and negative regulation of Wnt signaling. Further, caveolin1 (Cav), which is important in nitric oxide metabolism and vesicle-mediated transport, Spbc25, a component of the Ndc80 kinetochore complex (McCleland et al.,2004), and genes with unknown biological function such as Itm2b, Jagn1, and 2310032F03Rik can be classified as enriched in Müller glia. The data suggest that these genes are specific for Müller glia in the adult murine retina. However, some genes are also expressed in retinal progenitor cells, as has been shown for Sox2 and Dkk3 (Blackshaw et al.,2004), and for Spbc25 (data not shown; Trimarchi and Cepko, in preparation). Müller glia enriched transcripts for synaptoporin (Synpr) and type IX procollagen alpha1 (Col9a1) might also be Müller glia specific, although both of these are also detectable in one immature ganglion cell (P0 A6). Some of the potentially Müller glia specific transcripts, such as BING4, Prss2, Ube1c, and Gnb1l could only be detected in three of five Müller glial cells. This might be an indication of heterogeneity within Müller glial cells, or it might be due to a technical issue.
The Müller glia marker clu, expected to be specific to Müller glia in this comparison, is also detected in one of the rod photoreceptors. Because the retinal pigmented epithelium (RPE) expresses clu as well, it is possible that this expression occurred through RPE contamination of the rod single-cell preparation (Trimarchi et al.,2007). A summary of these data can be found in Table 1. Verification of the expression pattern of these genes by another method is required, such as in situ hybridization on retinal sections, as was done for a subset in this study (below).
Table 1. Summary of Müller Glia Specific and Enriched Genes1
|Category||Gene name||Reference||Verification; Specificity||Expression level|
|Classical markers||ApoE||Amaratunga et al.,1996||ISH||High|
| ||Aqp4||Nagelhus et al.,1998||ISH, DISH||High|
| ||Clu||Blackshaw et al.,2004||ISH, DISH||High|
| ||Vimentin||Greenberg et al.,2007||ISH||High|
| ||Kir2.1||Newman and Reichenbach,1996||N/A||High|
| ||Kir4.1||Newman and Reichenbach,1996||N/A||High|
| ||S100a16||Seigel et al.,1996||N/A||High|
| ||Glul||Blackshaw et al.,2004||ISH||High|
| ||CRLBP-1||Matsuda et al.,2004||ISH, DISH||High|
| ||Dkk3||Blackshaw et al.,2004||ISH, DISH||High|
| ||Chx-10||Rowan et al.,2004||N/A||High|
|New markers||Spbc25|| ||ISH||High|
| ||Itm2b|| ||ISH, DISH||High|
| ||Apg4b|| ||ISH, DISH||Low|
| ||Dbi|| ||ISH||High|
| ||GPR37||Blackshaw et al.,2004||ISH, Xgal||High|
| ||Car2|| ||ISH||High|
|Müller glia enriched||Abca8a|| ||N/A; high||High|
| ||Sox2||Blackshaw et al.,2004||N/A; high, also prog||High|
| ||Gnai2|| ||N/A; high||High|
| ||Cav|| ||N/A; high||High|
| ||Jagn1|| ||N/A; high||High|
| ||2310032F03Rik|| ||ISH; high||High|
| ||Synpr|| ||N/A; medium||High|
| ||Col9a1|| ||N/A; medium||High|
| ||Prss2|| ||N/A; high, subset||High|
| ||Ube1c|| ||N/A; high, subset||High|
| ||Gnb1l|| ||N/A; high, subset||High|
| ||Bing4|| ||N/A; high, subset||High|
| ||Kcnj10|| ||N/A; high||High|
| ||Cts0|| ||N/A; high||High|
| ||Ctsh|| ||N/A; high||High|
| ||Rhpn|| ||N/A; high||High|
Identification of genes characteristic for Müller glial cells
To identify Müller glial enriched or specific genes, the analytical tool Cluster 3.0 (Eisen et al.,1998) and Treeview software (Eisen et al.,1998) were used for clustering the single cell profiles, as well as to create a visual map of the data. The five Müller glial cells were once again compared with immature amacrine and ganglion cells and with rod photoreceptors. However, instead of considering only the most highly expressed transcripts, as was done in the previous analysis (Fig. 2), the complete data set obtained from the microarrays was used for comparison. The data were filtered such that any probe set that failed to reach a signal of 1,000 in any single cell was eliminated. The data for the remaining probe sets was log transformed and normalized according to Eisen et al. (1998).
A representative part of the resulting expression level matrix is shown in Supplemental Figure S1 and a portion of that same cluster around Aqp4 is shown enlarged in Figure 3A. Interestingly, tight clustering of genes in a given cell type can be observed, with Figure 3A showing a portion of a Müller glial cluster, and Figure 3B showing a rod photoreceptor cluster. All Müller glial cells shared a similar gene expression profile over a wide range of genes (Supplemental Fig. S1). Other cell types showed expression values that were considerably distinct from those of the Müller glia. The data in Figure 3A of genes enriched in Müller glia suggest a Müller glia core transcriptome, which likely comprises genes encoding some of the key functions of Müller glia. One of the most prominent examples is the inwardly rectifying potassium channel (Kcnj10). Müller glia have been previously shown to play an active role in regulating retinal functions by allowing transport of small molecules through specific ion channels. Other potassium channels expressed in Müller glial cells include Kir4.1 and Kir2.1 (Newman and Reichenbach,1996; Bringmann et al.,2000). Genes included in this cluster are also ApoE and Abca8a, which have been identified to be among the most highly expressed genes in the previous analysis (Fig. 2). Genes not previously identified in Müller glia, such as two members of the lysosomal cystein protease family, cathepsin O (Ctso) and cathepsin H (Ctsh), and rhophilin, which is involved in actin cytoskeleton reorganization, have been found to be present in this cluster as well.
Figure 3. Cluster analysis of retinal single cells. A hierarchical cluster analysis was run on the Affymetrix signals from 19 developing amacrine (ACs) and ganglion cells (RGCs), 2 immature rod photoreceptors (PRs) from postnatal day 0 (P0), 2 mature rod photoreceptors, and 5 mature Müller glial cells (MGs). Selected portions of clusters are shown as a heatmap. Columns represent the samples, and rows represent the transcripts (with abbreviated name and Affymetrix ID). A: Part of the cluster around the gene Aqp4, revealing genes enriched in Müller glia. B: A cluster of photoreceptor specific genes.
Download figure to PowerPoint
The expression of genes known to encode functions that might be relevant for Müller glia physiology was examined. For example, as noted above, there are retinal progenitor genes expressed in Müller glia, such as Dkk3 and Chx10 (Blackshaw et al,2004; Rowan and Cepko,2004), perhaps in keeping with their ability to produce neurons in some species under some conditions, or respond to injury by re-entering the cell cycle.
Expression of the paired-type homeobox, Pax6, was noted in four out of five of the single Müller glial cells (Fig. 5). Pax6 is a marker for progenitor cells and also for ganglion cells, amacrine cells, and horizontal cells in the mature retina. It has not been reported to be expressed in mature Müller glia. However, expression is consistent with the expression of other progenitor genes in Müller glia.
Examination of the expression of cell cycle genes by Müller glia showed that cyclinD3, Cdc14A, Cdk10, and Spbc25 were enriched in Müller glia. Several genes of the Notch pathway were expressed in significant amounts in Müller glia, such as Notch1 and Notch2, and might regulate the re-entry of Müller glia into the cell cycle under specific (e.g., pathological) conditions (Fig. 4). Growth factors such as Fgfs have been suggested to be important for short-range interactions among retinal cells. The facts that aFGF can be released from rod outer segments by a phosphorylation-dependent mechanism, and that apical processes of Müller glial cells surround the photoreceptors, suggest that these factors may mediate communication between glial cells and neurons (Mascarelli et al.,1991). Fgfs, neurotrophic factors, and growth factor receptors are also suggested to be important in Müller glial proliferation (Milenkovic et al.,2003). Activation of several growth factor- and G-protein-coupled receptors has been shown to induce proliferation in cultured Müller glial cells (Ikeda and Puro,1994). FGF r1, Ngf R, and Ogfr were highly expressed among Müller glia, whereas Eph and Igfb4 were expressed among a few Müller glial cells.
Figure 4. Heatmap of Müller glia enriched genes illustrating potential functions of Müller glia. Expression of genes with potentially relevant functions (such as growth factors, cell cycle components, and Notch pathway components) for Müller glia physiology was examined. The data are summarized as a heatmap. RGCs, retinal ganglion cells; ACs, amacrine cells; PRs, photoreceptors; MGs, Müller glia cells.
Download figure to PowerPoint
Figure 5. Section in situ hybridization of genes enriched in single Müller glia. Top: Heatmap of Müller glia enriched genes. A–N: ISH signal of candidate Müller glial transcripts on mouse retinal sections. Gene name is as indicated. Probes used are listed in the supplementary data. O: Xgal histochemistry on GPR37tm1Dgen retinal sections. P: Recombinase-dependent nuclear lacZ expression in Müller glial cells of Pdgfra-Cre; RC::PFwe retinal sections. Q: Immunohistochemistry on mouse retinal sections with anti-Pax6 (red) and anti-Glul (green) antibodies. R: Digital magnification of image shown in Q. RGCs, retinal ganglion cells; ACs, amacrine cells; PRs, photoreceptors; MGs, Müller glia cells; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. See Supplemental Figure S3 for magenta-green versions of Q and R. Scale bar = 50 μm in A (applies to A–P); 20 μm in Q.
Download figure to PowerPoint
Under pathological conditions, Müller glia might additionally act as modulators of the immune and/or inflammatory response. Retinal glial cells have been shown to be a major source of cytokines after retinal detachment (Nakazawa et al.,2006). Recently, MCP-1 (monocyte chemoattractant protein 1) expression was shown in Müller glia after retinal detachment (Nakazawa et al.,2007). Components of the chemokine system that could be detected in Müller glia were Xcr1 and Cxcl16 (Fig. 4). Müller glia might also be involved in the prompt clearing of photoreceptor debris under pathological conditions. During Drosophila development, it has been shown that glial cells engulf degenerating axons through recognition mediated by drpr/ced-1 and ced-6/CED-6 (Awasaki et al.,2006). It has been proposed that apoptotic cells and degenerating axons of mature neurons are removed by a similar mechanism (Awasaki et al.,2006; MacDonald et al.,2006). The ability of Müller glia to phagocytose retinal cell debris and foreign substances has been previously shown in vitro (latex beads) and in vivo (melanin) after retinal detachment (Mano and Puro,1990; Stolzenburg et al.,1992; Francke et al.,2001). However, two of the genes, Megf10 (multiple EGF-like-domains 10)/ced-1 and Abca1 (ATP-binding cassette, subfamily A, member 1)/ced-7 were observed to be expressed in a subset of Müller glia, whereas others, Gulp1 (GULP: engulfment adaptor PTB domain containing 1)/ced-6, Lrp1 (low-density lipoprotein receptor-related protein 1)/ced-1, and Rac1 (RAS-related C3 botulinum substrate 1)/ced-10, could not be detected in the Müller glial cells studied here.
Heterogeneity of Müller glia
Although the focus of this study was to define the transcriptome of Müller glia, it was noted that certain genes were expressed heterogeneously among the Müller glial cells analyzed (Fig. 4, Supplemental Figs. S1, S2), including common housekeeping genes such as cytoplasmic β-actin, β-2-microglobulin, heat shock protein 1 β, and TATA box binding protein. This might be due to technical or real differences. Indeed, other studies employing Affymetrix microarrays to examine housekeeping gene expression among different tissues, as well as our SAGE study of the retina during development, similarly found that housekeeping genes vary considerably in their expression levels (Warrington et al.,2000; Blackshaw et al.,2004; Lee et al.,2007).
Little is known about the significance and extent of Müller glial cell heterogeneity, although it is clear that molecular differences exist because the homeodomain-containing transcription factor Chx10 is expressed in only a subset of all Müller glial cells (Rowan and Cepko,2004). Heterogeneous expression of Chx10 could also be observed in the current study (Fig. 4). Heterogeneity of gene expression among three of the Müller glial cells was observed across a large cluster (Supplemental Fig. S1). This was further analyzed for a few of these genes by using in situ hybridization (see below). A recent expression analysis of astrocytes has revealed extensive molecular heterogeneity among that class of glial cells as well (Bachoo et al.,2004). Furthermore, morphological studies have identified extensive anatomical differences between Müller glial cells in the central versus peripheral chick retina (Anezary et al.,2001). Further characterization of the glial genes identified in this study will help clarify the diversity of Müller glial cells and perhaps shed light upon a potential functional diversity.
Validation of expression in Müller glia
To validate the expression of genes expressed in Müller glial cells, in situ hybridization on retinal sections was performed for a subset of genes (Fig. 5). The corresponding Affymetrix signal values are represented in the top panel of Figure 5 as a heatmap. Because the cells used for comparison were mainly immature cells, some of the genes that are indeed specific to Müller glia in the adult retina are also expressed in those immature cells that still contain a certain expression profile of (late) progenitor cells. The similarity of Müller glia gene expression to progenitor cell gene expression has been discussed previously (Fig. 4 and Blackshaw et al.,2004). Müller glia transcripts were localized most prominently in the middle portion of the inner nuclear layer (INL), consistent with the location of the transcript in the cell body. Transcripts with this predicted restriction in expression pattern are Dkk3, ApoE, Spbc25, Apg4B, Pak3, Car2, Rlbp1, deltex 2 homolog (Dtx2), and pleiotrophin (Ptn) (Fig. 5A–F,H,L,M). Certain genes, such as Dbi, GPR37, and Glul (Fig. 5G,N,K, respectively), gave punctuate staining in the middle portion of the INL, as well as diffuse expression in the outer nuclear layer (ONL), consistent with expression in ascending glial processes, and in the ganglion cell layer (GCL), consistent with expression in Müller glial descending processes and endfeet. Expression in the INL was either throughout the INL, as is the case for clu (Fig. 5J), or in a subset of cells in the middle of the INL, as is the case for Dbi (Fig. 5G). The pan INL pattern is suggestive of a gene being expressed in several cell types in the INL, whereas the more specific middle of the INL pattern is more consistent with expression restricted to Müller glia. Even transcripts expressed in only a subset of Müller glial cells profiled, such as Apg4b, Dtx2, and Ptn could be verified and detected by in situ hybridization (Fig. 5D,M,L, respectively).
It was of interest to determine whether some of the genes with moderate to high levels of Affymetrix signals in profiled Müller glia were specific to Müller glia. Specificity might give some insight into function and would indicate that such genes are suitable for use as specific markers of Müller glia. Although analysis of candidate Müller glial genes by section ISH is a fairly rapid and informative assay, it was unclear for some genes whether their expression was restricted to Müller glial cells. To determine specificity of expression more precisely, ISH on dissociated cells was carried out in combination with immunocytochemistry for the known glial marker glutamine synthetase (Glul). Dissociated cell in situ hybridization (DISH) for Itm2b revealed that it was exclusively expressed in Glul-positive cells. A modulator of the Notch pathway, Apg4b, as well as Dkk3, Aqp4, and ApoE were shown to be fairly specific to Müller glia as well (Fig. 6). DISH also allowed an assessment of heterogeneity of expression within Muller glial cells. The Rlbp1 transcript, expressed in only three of five cells in the microarray analysis, was found in only a subset of Müller glia, in 70% of the cells marked with the anti-Glul (Supplemental Fig. S2). In contrast, there was a complete overlap of Itm2b and clusterin transcript expression with anti-Glul staining (Fig. 5, top panel, Supplemental Fig. S2).
Figure 6. Dissociated cell ISH confirms expression in Müller glia. A–J: Retinae from mature mice were dissociated, plated on slides, and processed for detection by ISH for the indicated probes (ApoE, Dkk3, Apg4b, Itm2b, Rlbp1). Merged images show DAPI stain and immuno-staining for Glul and ISH. Insets are digital magnifications. Scale bar = 25 μm in A (applies to A,B), C (applies to C,D), E (applies to E,F), G (applies to G,H), and I (applies to I,J).
Download figure to PowerPoint
Additional validation of GPR37 as a Müller glia specific gene in the murine retina was possible through the use of a transgenic mouse line, GPR37tm1Dgen, available from the Jackson Laboratory. The GPR37tm1Dgen mice express a transgene consisting of lacZ driven by the endogenous GPR37 promotor. Xgal staining on retinal sections showed lacZ expression in Müller glial cells (Fig. 5O).
Pax6 expression was further investigated by immunocytochemistry by using anti-Pax6 and anti-Glul antibodies (Fig. 5Q,R; see Supplemental Fig. S3 for magenta-green versions of images). Analysis of the sections with confocal microscopy showed that Pax6 is indeed expressed in adult Müller glial cells, as well as in some types of retinal neurons, as previously reported for neurons (de Melo et al.,2003), but not previously reported for Müller glial cells.
Creation of a mouse line expressing Cre for genetic access to Müller glia
To create a line of mice for future experiments in which gene expression specifically in Müller glial cells could be manipulated, a BAC was engineered to express Cre from a promoter that was predicted to be enriched or specific to glial cells (Kessaris et al.,2006). This BAC was used to create a transgenic mouse. Mice that successfully transmitted the BAC were assessed for Cre activity by crossing to mice carrying a nuclear lacZ floxed indicator allele, RC::PFwe (Farago et al.,2006). Strong nuclear lacZ expression was detected in all or nearly all Müller glial cells in the progeny of this cross (Fig. 5P). This may be a fortuitous finding, as inspection of the microarray data for Müller glia did not show expression of Pdgfra. Some Xgal staining was observed in the ONL and GCL, but it was not distinctly nuclear. Control Xgal staining of littermates did not show this lacZ activity, and thus the origin of the ONL and GCL staining is unclear.