Retinal Stem Cells in the Adult Mammalian Eye

See allHide authors and affiliations

Science  17 Mar 2000:
Vol. 287, Issue 5460, pp. 2032-2036
DOI: 10.1126/science.287.5460.2032


The mature mammalian retina is thought to lack regenerative capacity. Here, we report the identification of a stem cell in the adult mouse eye, which represents a possible substrate for retinal regeneration. Single pigmented ciliary margin cells clonally proliferate in vitro to form sphere colonies of cells that can differentiate into retinal-specific cell types, including rod photoreceptors, bipolar neurons, and Müller glia. Adult retinal stem cells are localized to the pigmented ciliary margin and not to the central and peripheral retinal pigmented epithelium, indicating that these cells may be homologous to those found in the eye germinal zone of other nonmammalian vertebrates.

The two functional components of the rodent retina, the inner neural retina (NR) and the outer retinal pigmented epithelium (RPE), are largely developed by the early postnatal period and show no evidence for adult regeneration (1–3). These developmental characteristics differ from other vertebrate species in which NR cells are produced throughout life (4) and show a remarkable regenerative capacity following injury (5). Thus, it has been generally assumed that the adult mammalian eye is devoid of retinal stem cells (self-renewing and multipotential cells) (6,7) and is incapable of substantial neural regeneration. To determine if a stem cell exists in the mouse eye, we examined the ability of single NR cells and single pigmented cells from the entire RPE layer (including pigmented cells from the ciliary margin) (Fig. 1A) to clonally proliferate in vitro using a neural stem cell colony–forming assay (6, 8). Dissociated cells from the RPE and the NR were obtained from adult (2 to 3 months old) and embryonic day 14 (E14) mouse eyes and were cultured independently in serum-free media either without exogenous growth factors or in the presence of epidermal growth factor (EGF) or basic fibroblast growth factor (FGF2) (6, 8). A small number of individually identified single adult RPE cells (8) clonally proliferated in 1 to 2 days to initially form small sphere colonies composed of darkly pigmented cells, independent of exogenous growth factors (Fig. 1B). These small colonies continued to proliferate with or without exogenous growth factors; this is in contrast with the isolation of forebrain stem cell colonies, which require exogenous growth factors (6, 7). After 5 to 7 days, each colony was composed of a mixture of cells ranging from darkly pigmented to nonpigmented (Fig. 1C). To confirm that the increase in colony size was due to proliferation, we added bromodeoxyuridine (BrdU), a thymidine analog, between days 2 and 3 of the culture period (8). After 7 days, many cells within the expanded colony were immunolabeled for BrdU, indicating that cells were dividing in the colonies. The single sphere colonies consisted of ∼13,000 cells (8), indicating that cells within the colony could sustain extended cell division in our culture conditions.

Figure 1

(A) Schematic representation of a sagittal section of the ciliary margin region of the adult mouse eye. The boxed area on the left is magnified on the right. The ciliary margin consists of pigmented cells (thick black line) overlying the smooth ciliary muscle (light gray) of the inner eye facing the lens. (B) A clonally derived PCM sphere colony after 2 days in vitro from low-density culture conditions with no exogenous growth factors. All cells within the small colony are heavily pigmented and derived from a single heavily pigmented PCM cell from either primary or subcloned cultures. (C) Cells within the small sphere colony continue to proliferate and, after 7 days, generate relatively large colonies containing both pigmented and nonpigmented progeny. The granular appearance of the extracellular space is due to pigment granules derived primarily from the degradation of nonviable PCM cells or the extrusion of pigment. The sequence of colony formation was examined in cultures with a single cell per well (n = 960 wells from each of two separate experiments), landmarking single hypertrophic cells at 5 cells per microliter (n = 24 wells from each of two separate experiments) and cultures at 20 cells per microliter (Fig. 2). Scale bar, 100 μm.

To identify the origin of the adult pigmented colony-forming cells more precisely, we cultured separately the RPE cells (central and peripheral) and pigmented cells from the ciliary margin (PCM) (Fig. 1A). Only PCM cells (and not RPE cells proper) proliferated to form sphere colonies. In preliminary experiments, a low frequency of sphere colony–forming PCM cells was similarly isolated from postmortem adult bovine and human ciliary margin tissue, indicating that the colony-forming ability of PCM cells is conserved across mammalian species. In addition, colonies did not arise from cells of the adult iris, ciliary muscle, or NR or from nonpigmented ciliary process cells in any of the species tested. A similar rate of sphere colony formation was observed from pigmented cells isolated from the entire mouse E14 RPE (including the presumptive PCM region at the peripheral RPE margin). However, adult PCM cells generated ∼10 times more sphere colonies per eye than did the cells isolated from the entire E14 RPE (Fig. 2A) (9). Thus, cells capable of forming sphere colonies expand between E14 and adulthood. Nonetheless, the adult colony-forming PCM cell is rare. In cultures with a single cell per well, we observed six sphere colonies out of 960 wells (∼0.6%). To estimate the frequency of colony-forming PCM cells, we performed a limiting dilution analysis (8, 10), which revealed that the minimum frequency of the PCM sphere-forming cells in the adult eye was ∼0.2% of pigmented cells in the ciliary margin (Fig. 2B). To document further that PCM sphere colonies were generated by the proliferation of a single PCM cell in low-density cultures, we examined colonies generated from mixtures of dissociated PCM cells from green fluorescent protein (GFP)–expressing mice and from mice lacking the GFP transgene (11). Separate GFP-positive and non-GFP PCM sphere colonies were observed with no evidence of cell mixing (8), demonstrating that the colonies were not derived by cell aggregation but arose clonally at low cell densities, as they do in single-cell cultures.

Figure 2

Adult and E14 PCM sphere colonies are generated at a very low frequency. (A) Number of colonies [mean ± S.E.M (error bars)] generated from either E14 or adult PCM cells in no exogenous growth factors, EGF, or FGF2 after 7 days in vitro. There is no significant difference between growth factor condition within an age group. However, a greater number of sphere colonies per eye can be isolated from adult PCM cells in comparison to the entire E14 RPE in all growth factor conditions. The graph represents the average from four to six separate experiments. (B) Limiting dilution analysis (8) revealed a linear relation between the number of PCM cells plated (no exogenous growth factor) and the formation of a colony [mean as percent control ± SEM (error bars)], indicating that a single cell type caused the formation of a colony with a frequency of ∼1 in 600 cells (∼0.2%) (dashed lines). The graph represents the average from three separate experiments. (C) Primary PCM sphere colonies can be subcloned, and in the presence of FGF2, the number of secondary and tertiary sphere colonies [mean ± SEM (error bars)] generated is greater than the number generated in the absence of exogenous growth factor. Subcloned sphere colonies have a similar appearance to primary sphere colonies. The graph represents the average of at least four sphere colonies per group from three separate experiments.

Stem cells are capable of self-renewal (7, 12). To demonstrate that the colony-forming PCM cells exhibit this property, we dissociated and then recultured individual PCM colonies in the presence of FGF2 or in the absence of exogenous growth factors (8). A small number of pigmented cells from dissociated primary PCM colonies generated new secondary sphere colonies regardless of growth factor condition, indicating that the initial colony-forming PCM cell had the capacity to self-renew (Fig. 2C). Although adult PCM stem cells can proliferate in the absence of exogenous growth factors, the release of endogenous growth factors may facilitate colony formation. Consistent with this notion, antibodies to FGF2, known to block the formation of FGF-dependent sphere colonies from embryonic forebrain neural stem cells (10), caused a 50% reduction in the number of PCM colonies in the absence of exogenous growth factors. The second and third subcloning of single PCM sphere colonies revealed that they generated an average of two new sphere colonies in the absence of growth factors. In contrast, colonies generated in FGF2 and subcloned in FGF2 produced an average of six to eight new sphere colonies (Fig. 2C). Sphere colonies could be subcloned for at least six generations (longest passage attempted). Thus, the ability of PCM stem cells to repeatedly generate a sphere colony is partially dependent on endogenous FGF2, and exogenous FGF2 can influence the proliferation or survival of the retinal stem cell or its progeny. Sphere colonies from each generation invariably arose from single pigmented cells and then proliferated to contain similar proportions of pigmented and nonpigmented cells after 7 days in culture, indicating that the colony-forming stem cells did not lose their pigmentation. Nonpigmented colony-forming stem cells, however, could be isolated from the ciliary margin of adult albino CD1 mice, suggesting that the biochemical components of pigment formation are not required for stem cell function.

To determine if some of the nonpigmented cells generated within the PCM sphere colonies represented undifferentiated retinal progenitor cells, we examined the expression of CHX10 and nestin (13,14). Immunolabeling of free-floating PCM sphere colonies harvested after 5 to 7 days in culture (8, 10) revealed that many nonpigmented cells within a PCM colony expressed CHX10 (Fig. 3A). Thus, a single adult PCM cell proliferated to produce cells expressing a marker for embryonic NR precursors, which is not expressed in the RPE or in the ciliary margin (13). A similar proportion of cells within a PCM colony also expressed nestin, indicating that many sphere colony cells remained undifferentiated (Fig. 3B).

Figure 3

Immunostaining of whole adult PCM sphere colonies harvested after 7 days in culture (A andB) and differentiated PCM colonies 21 days after plating on laminin [top panels in (C) through (H) and (I)]. Many cells within a colony expressed CHX10 (A) and nestin (B) before differentiation. The average percentage of cells (±SD) with distinct immunoreactivity is given as observed for the following: rod photoreceptor markers were (C) 51.1 ± 24.0% (309L), (D) 12.8 ± 6.7% (Rho1D4), (E) 56.2 ± 17.6% (D2P4), and (F) 14.2 ± 4.3% (ROM-1); bipolar neurons were (G) 11.6 ± 5.4% (CHX10) and (H) 11.4 ± 0.4% (PKC); and Müller glia were (I) 41.8 ± 14.1% (10E4). Scale bar, 70 μm (A and B) and 20 μm (C through I). Phase-contrast images [bottom panels in (C) through (H)] identify cells that are immunoreactive for retinal markers (arrows) or adjacent cells that are unlabeled (arrowheads). There was considerable variation in the observed frequencies of different retinal neuronal phenotypes; however, each retinal immunolabeling phenotype was consistently observed in all sphere colonies examined. All antibodies labeled their cell-specific antigens in cryosections of the adult mouse retina. Cultured cells and retinal sections processed simultaneously with primary antibodies omitted were negative.

Multilineage potential is a hallmark feature of stem cells (7, 12). To determine whether the PCM stem cells were multipotential, we cultured single PCM sphere colonies under conditions known to promote retinal cell differentiation (6, 8). Immunolabeling of PCM colonies cultured for up to 21 days revealed that many cells that migrated away from the sphere colonies expressed the pan-neuronal marker microtubule-associated protein 2, whereas other separate cells that differentiated expressed the astrocytic marker glial fibrillary acidic protein. Differentiated sphere colonies contained a small number of nestin-positive cells that remained confined to the centers of the colonies. PCM cells that were cultured as an adherent monolayer (not in the colony-forming assay) for the same 21-day period in identical conditions did not express any neural markers and remained as large, flattened pigmented cells. Thus, PCM cells do not directly transdifferentiate into neurons; proliferation of PCM stem cells is correlated with CHX10 and nestin expression and the development of neuronal and glial phenotypes. In peripherally localized cells of PCM colonies, we were able to demonstrate the presence of markers for different differentiated retinal cells (13, 15). These included 309L (rod cyclic guanosine 3′,5′-monophosphate–gated channel), Rho1D4 (rhodopsin), D2P4 (peripherin), ROM-1 (rod outer disk protein), CHX10 and protein kinase C (PKC) (bipolar neurons), and 10E4 (Müller glia) (Fig. 3, C through I). The morphological profiles of the immunolabeled cells observed after 21 days were cell type specific. For example, CHX10- and PKC-labeled bipolar neurons (Fig. 3, G and H) had a relatively small soma that contained very little perinuclear cytoplasm, with short leading and trailing processes. Cells labeled with rod photoreceptor markers (Fig. 3, C through F) were also small and often had a slightly flattened appearance (some with small numbers of processes) resembling rod morphology in low cell density cultures (16). In contrast, 10E4-labeled Müller glia (Fig. 3I) were relatively large cells resembling protoplasmic astrocytes. To further examine photoreceptor differentiation, we analyzed RNA from differentiated PCM colonies for the presence of transcripts for the photoreceptor-specific homeobox gene Crx (17). Using reverse transcriptase–polymerase chain reaction (RT-PCR) (8), we detected Crx in differentiated sphere colonies (Fig. 4A). In situ hybridization ofCrx (8) also was performed on differentiated sphere colonies. Expression of Crx mRNA was detected in 37.7 ± 22.5% of cells (resembling those that were immunoreactive for rod-specific antibodies), whereas other nearby cells exposed to the antisense probe remained unlabeled (Fig. 4C). Differentiation of ganglion, horizontal, and amacrine cells was not observed under the conditions described above. However, in preliminary experiments with high-density differentiation pellet cultures (centrifugation of multiple-PCM sphere colonies, which maximizes cell contact), the amacrine-specific expression of syntaxin was detected with the HPC-1 antibody (15). Thus, differentiation of all retinal cell types will likely require conditions that are more comparable to the in vivo environment. These diverse retinal cell types that were observed could be differentiated from both primary and subcloned PCM colonies, which suggests that self-renewing retinal stem cells maintain their multipotentiality. Furthermore, although most cells extrude or stop producing pigment and differentiate (as described above), a few heavily pigmented cells in PCM colonies maintain pigmentation, even after 21 days of differentiation. Thus, a proportion of cells may be differentiated PCM or RPE cells or self-renewing retinal stem cells (that also maintain pigmentation).

Figure 4

Expression of the homeobox geneCrx in 21-day differentiated PCM sphere colonies. (A) RT-PCR analysis of total RNA isolated from adult mouse retina control samples (lane 1) and differentiated adult PCM colonies (lane 2). Nested Crx RT-PCR resulted in a 207–base pair product (arrow). No product was seen in an adult liver control sample (lane 3) or in adult retinal RNA with RT omitted (lane 4). Data are representative of ∼75 ng of adult retinal or liver RNA (controls). We used one differentiated PCM colony (∼13,000 cells) isolated from two separate cultures; we estimate that less RNA (more than one order of magnitude) was used from the PCM colony in comparison with the adult retinal and liver controls. (B andC) In situ hybridization analysis of Crxon differentiated PCM colonies exposed to antisenseCrx probes. Cells expressing CrxmRNA [arrowheads, top panel in (B)] and cells with noCrx mRNA expression [arrowheads, top panel in (C)] could be identified in the same culture well exposed to antisenseCrx probes using phase (Ph3) contrast and focusing specifically on the silver grains in the emulsion layer. Bottom panels in (B) and (C) show the position of the same cells indicated by arrowheads under phase (Ph2) contrast focusing specifically on cell morphology. Culture wells exposed to sense Crxprobes were identical in appearance to (C). Data are representative of four adult PCM colonies exposed to antisense Crxprobes and three adult PCM colonies exposed to sense probes from two separate cultures. The same probes detected Crx mRNA in sections of the adult mouse eye. Scale bar, 20 μm.

To determine whether the multilineage potential of PCM-derived colonies was distinct from forebrain-derived colonies in similar culture conditions, we immunolabeled cells from forebrain-derived colonies with retinal-specific antibodies. No (or very infrequent) immunoreactivity was observed. Although O4-positive oligodendrocytes differentiated from E14 forebrain-derived colonies (9.8 ± 1.6% of differentiating cells), oligodendrocytes were not observed from adult or E14 PCM colonies. Oligodendrocytes are not generated from within the embryonic or adult eye in vivo; thus, retinal stem cells may be specified to generate only retinal cells. Alternatively, distinct culture environments may determine the extent of oligodendrocyte differentiation.

The NR was similarly tested for the presence of a stem cell. In culture conditions including EGF or FGF2, only E14 (but not adult) NR cells generated nonpigmented sphere colonies. In contrast to the PCM stem cells, the E14 colony-forming NR cells did not demonstrate the ability to self-renew, strongly suggesting that the NR does not contain a stem cell and that NR-derived colonies have been formed by the well-described multipotential progenitor found during retinal development (1). NR sphere colonies could not be isolated from the adult eye, indicating that retinal progenitors have a limited life-span, unlike the PCM stem cells (9). Longevity is suggested to be an important criterion for stem cells (12), and forebrain neural stem cells have been shown to exist in senescent mice (18).

Genes that regulate the proliferation of retinal progenitor cells in vivo might be expected to influence the proliferation of progenitor cells derived from retinal stem cells, and the number of retinal progenitor cells may regulate the number of adult retinal stem cells. For example, a complete loss of function of the Chx10homeobox gene in mice, which is expressed in NR progenitors but not in the RPE or PCM during development, is associated with a greatly reduced proliferation of retinal progenitors (13). To test the hypothesis that the number of adult retinal stem cells in vivo is regulated by the number of NR progenitors produced during development, we examined PCM colony formation in homozygous Chx10 null (orJ /orJ ) mice. Consistent with previous reports (13), the PCM of the adultorJ /orJ eye was expanded compared to the wild type, whereas the central and peripheral RPE layer appeared relatively unperturbed (9). There was an approximate fivefold increase in the number of primaryorJ /orJ PCM colonies (per eye) compared to wild-type (mouse strain 129/sv) PCM colonies [orJ /orJ , 195.7 ± 11.1 (n = 9); wild type, 42.5 ± 5.6 (n = 6); data are from two separate experiments in the presence of FGF2]. No colonies were generated fromorJ /orJ NR cells. The average diameter of primaryorJ /orJ sphere colonies was 36% smaller than primary wild-type colonies [orJ /orJ , 225 ± 2.9 μm (n = 3); wild type, 350 ± 17.3 μm (n = 3)], confirming that Chx10is required for full progenitor cell proliferation (13). SingleorJ /orJ PCM colonies generated similar numbers of secondary sphere colonies as controls and contained both neurons and glia when differentiated. Thus, the ultimate size of the retinal stem cell population in the ciliary margin of the adult eye may be determined by the numbers of NR progenitor cells, which inhibit the in vivo expansion (through symmetric division) of the stem cell population.

Important differences exist between this study and earlier work in nonmammalian vertebrate species demonstrating regeneration of the NR through transdifferentiation of the RPE (5, 19). First, the mammalian retinal stem cell is relatively rare and localized to the ciliary margin. Previous studies on transdifferentiation show that most or all of the RPE cells differentiate into neurons (5, 19). Second, transdifferentiation is thought to entail a direct phenotypic change from RPE cell to neuron without cell division. Retinal stem cells divide to produce their neuronal and glial progeny. Finally, the ciliary margin in some nonmammalian vertebrate species is thought to harbor stem cells that facilitate neurogenesis throughout life (20). Thus, the mammalian ciliary margin may represent an evolutionarily homologous region that has independently acquired a mechanism for repressing stem cell division. Because isolated stem cells can proliferate in the absence of exogenous growth factors and the size of the stem cell population may be regulated in vivo by the number of NR progenitor cells during development, this quiescence is likely due to an inhibitory environment in the adult eye. Once freed from the inhibition (or if inhibitory factors can be overcome in vivo), the stem cells have the potential to generate new retinal cells.

  • * To whom correspondence should be addressed. E-mail: derek.van.der.kooy{at}


Stay Connected to Science

Navigate This Article