GDF11 Controls the Timing of Progenitor Cell Competence in Developing Retina

See allHide authors and affiliations

Science  24 Jun 2005:
Vol. 308, Issue 5730, pp. 1927-1930
DOI: 10.1126/science.1110175


The orderly generation of cell types in the developing retina is thought to be regulated by changes in the competence of multipotent progenitors. Here, we show that a secreted factor, growth and differentiation factor 11 (GDF11), controls the numbers of retinal ganglion cells (RGCs), as well as amacrine and photoreceptor cells, that form during development. GDF11 does not affect proliferation of progenitors—a major mode of GDF11 action in other tissues—but instead controls duration of expression of Math5, a gene that confers competence for RGC genesis, in progenitor cells. Thus, GDF11 governs the temporal windows during which multipotent progenitors retain competence to produce distinct neural progeny.

The vertebrate neural retina comprises seven neural cell types, all derived from one population of multipotent progenitors (1, 2). Retinal cell types do not arise synchronously but are generated in a stereotyped sequence (3, 4). In vitro results imply that retinal progenitors at different stages differ in their competence to produce distinct cell types (57). Such changes in potential are likely dictated by changes in expression of the transcription factors encoded by proneural genes (8, 9), but mechanisms of proneural gene regulation are poorly understood. An important role for cell-cell signaling is suggested by the fact that production of at least two retinal cell types, retinal ganglion cells (RGCs) and amacrine cells, can increase to compensate for losses of mature cells in either population (10, 11). This process has been postulated to be mediated by a feedback signal produced by mature cells (12), but the identity of the signal(s) is unknown.

GDF11, a member of the transforming growth factor-β superfamily of secreted signaling molecules, is expressed in several regions of a developing nervous system, including the retina (13). In olfactory epithelium (OE), GDF11 negatively regulates neuron number by causing cell-cycle arrest of the progenitor cells that give rise to olfactory receptor neurons (ORNs) (14). Here, we demonstrate that GDF11 is also a negative regulator of neuron number in neural retina, but through a completely different mechanism: GDF11 controls the period during which retinal progenitor cells are competent to produce certain progeny, thus governing the relative numbers of neural cell types that arise.

In mouse retina, Gdf11 expression begins about embryonic day (E) 12.5, when RGCs begin to differentiate (Fig. 1A). Gdf11 mRNA is observed throughout the retina, including the neuroblastic layer (NBL), until at least the first postnatal day (P0), although by E15.5, expression is highest in the developing ganglion cell layer (GCL). Expression of follistatin (Fst), which encodes a secreted GDF11 antagonist (15), is first detected at E13.5. From E15.5 on, Fst expression is highest in the nascent GCL but also evident in the NBL and presumptive amacrine cells. Putative receptors for GDF11 (14, 1618) are also expressed in appropriate patterns in the neural retina from E12.5 to 13.5 onward (fig. S1).

Fig. 1.

Retinal abnormalities in Gdf11 mutants. (A) In situ hybridization (ISH) for Gdf11 and Fst in developing mouse retina. nbl, neuroblastic layer; gcl, ganglion cell layer. Arrow in inset indicates Fst expression in presumptive amacrine cells. Scale bars, 200 μm. (B) Left, hematoxylineosin-stained paraffin sections of retina. Right, ISH for Brn3b. Insets, higher magnification of Brn3b+ gcl. Scale bars, 100 μm. (C) Top, increased cell number (P < 0.01, Student's t test) in Gdf11tm2/tm2 retinas. Total cell nuclei in GCL IPL were counted in 300 μm of + central retina in P0 cryosections stained with Hoechst. Bottom, no significant change in central retina thickness. Histograms show mean ± SEM of measurements from 4 to 5 animals of each genotype. (D) B-galactosidase (X-gal) staining of sections of Gdf11tm2/tm2;Tattler-1 and Gdf11+/+;Tattler-1 littermate embryos (33). Scale bars, 200 μm. on, optic nerve; oc, optic chiasm. Arrows point to optic tract. (E) Cross sections of dissected optic nerves stained with antibodies to neurofilament. Scale bar, 50 μm.

To investigate the role of Gdf11 in retinal development, we examined mice homozygous for the null allele Gdf11tm2 (14). Gdf11tm2/tm2 retinas show obvious changes as early as E14.5, when closure of the optic fissure is incomplete (fig. S2). By E17.5, the presumptive GCL of mutant embryos has an abnormally high cell density, and the inner plexiform layer (IPL), well demarcated in wild-type littermates, is not observed (Fig. 1B). Increased cell density in the mutant GCL is accompanied by widening of the cell layer expressing Brn3b [Gdf11tm2/tm2, 49.5 ± 3.3 μm (SD); wild-type, 38.5 ± 0.4 μm (SD)], which encodes a POU-domain transcription factor specific for differentiated RGCs (19, 20). By P0, the latest time at which the mutant is viable, Gdf11tm2/tm2 GCLs contain ∼50% more cells than wild types (Fig. 1C). The excess RGCs that form in Gdf11tm2/tm2 animals appear to differentiate normally, extending axons through the optic chiasm and tracts, which also appear abnormally thick (Fig. 1D). By neurofilament immunohistochemistry, we estimate a 37% increase in the cross-sectional areas of optic nerves in Gdf11tm2/tm2animals (Fig. 1E and fig. S2).

These changes, observed in all mutant mice examined (>32), imply that Gdf11 is a negative regulator of RGC genesis. In this respect, the changes in Gdf11tm2/tm2 retinas recall those in OE, in which Gdf11tm2/tm2 mice also have excess differentiated ORNs (14). However, unlike the situation in OE (14), Gdf11tm2/tm2 retinas display no increase in overall thickness, nor are the distribution or number of proliferating cells significantly altered (fig. S3). These observations suggest that the mechanism by which Gdf11 regulates neurogenesis in the retina differs from that in the OE.

Because Fst is known to antagonize GDF11 function in vivo and in vitro (14, 15), we also examined Fst–/– and Gdf11tm2/tm2;Fst–/– retinas. Fst–/– retinas showed a 26% reduction in the number of cells in the GCL and a large decrease in thickness of the Brn3b+ cell layer (Fig. 2, A and B), which indicates that Fst is a positive regulator of RGC development. Gdf11tm2/tm2;Fst–/– retinas showed an expanded Brn3b+ GCL, comparable to that observed in Gdf11tm2/tm2 retinas, consistent with the primary role of Fst as an inhibitor of GDF11 (Fig. 2B). Just as in Gdf11tm2/tm2 retinas, the level and pattern of progenitor cell proliferation was unaltered in Fst–/– retinas (fig. S3). The fact that cell proliferation is normal in Gdf11tm2/tm2 and Fst–/– retinas suggests that the size of the progenitor pool is not regulated by GDF11. Moreover, expression of several genes involved in early eye specification, patterning, and expansion is also normal in Gdf11tm2/tm2 mice (fig. S4).

Fig. 2.

The period of RGC genesis is altered in Gdf11tm2/tm2 and Fst–/– retinas. (A) Total cell nuclei per 300 μm GCL of Fst–/– central retina counted after Hoechst staining. (B) ISH for Brn3b on cryosections of Fst–/–, Gdf11tm2/tm2, and Gdf11tm2/tm2;Fst–/– retinas. Insets show superabundance of Brn3b+ cells in NBL of Gdf11tm2/tm2 and Gdf11tm2/tm2;Fst–/– retinas. Scale bar, 200 μm. (C) Dividing progenitor cells were labeled in vivo by BrdU injection into pregnant dams at E13.5 or E15.5, and retina cryosections (20 μm) were processed 2 days later for Brn3b ISH and BrdU immunostaining. Scale bar, 50 μm. (D) Quantitative analyses of experiments illustrated in (C). Histograms show mean ± SEM of total Brn3b,BrdU double-positive cells in GCL per 200 μm of retina. N = 2 to 3 animals of each genotype at each age. P < 0.05 (Student's t test, all comparisons). C, central retina; P, peripheral retina (200 μm from peripheral margin). (E) Quantitative analysis of RGC birthdating following BrdU injection at E13.5 and analysis at E15.5. No significant differences were observed.

During development, RGCs are born at the outer margin of the neural retina and migrate inward to the GCL during a defined period (21). Detailed examination of Gdf11tm2/tm2 and Gdf11tm2/tm2;Fst–/– retinas at E17.5 revealed that the NBL of these mutants contains three times as many Brn3b+ cells (migrating RGCs) as do wild types (Fig. 2B, insets, and fig. S5). This suggested that, in Gdf11tm2/tm2 retinas, RGC production may be prolonged beyond its normal period. To test this, we performed birthdating experiments. The results, shown in Fig. 2C, show an abnormally large number of bromodeoxyuridine-positive (BrdU+) cells in the GCL of Gdf11tm2/tm2 animals pulsed with BrdU from E15.5 to E17.5. Conversely, BrdU+ cells in the GCL of Fst–/– animals pulsed over this same time course were strongly decreased in number, as expected if Fst acts to inhibit GDF11. These differences were not seen when pulse labeling was done at earlier ages (Fig. 2E). Thus, although onset of RGC production appears unaffected by loss of Gdf11 or Fst, its down-regulation is delayed in Gdf11tm2/tm2 retinas (and accelerated in Fst–/–). A lengthened period of RGC production likely explains why Gdf11tm2/tm2 retinas accumulate abnormally large numbers of RGCs.

To determine whether Gdf11 regulates production of other retinal cell types, we examined rod photoreceptors and amacrine cells, two cell types whose peak periods of differentiation follow that of RGCs. Crx1,amarkerfor early photoreceptors, is normally up-regulated around birth when rod photoreceptor production peaks, and expands to cover much of the NBL (22). In Gdf11tm2/tm2 retinas, upregulation and expansion of Crx1 expression are not observed (Fig. 3A). Amacrine cells may be visualized by expression of syntaxin (23), as well as Pax6 and Prox1 (24, 25). In the amacrine cell layer of Gdf11tm2/tm2 retinas, expression of all three markers was reduced (Fig. 3A). Altogether, these results suggest that prolonged production of RGCs in Gdf11tm2/tm2 retinas occurs at the expense of cell types (amacrine cells and photoreceptors) that normally differentiate after RGC production has declined. We observed no obvious increase in amacrine cell or photoreceptor production in Fst–/– animals, possibly because excess GDF11 activity in Fst–/– retina is mitigated by the reduction in RGC cells, which express the highest levels of Gdf11 (Fig. 1).

Fig. 3.

GDF11 regulates development of photoreceptors and amacrine cells. (A) ISH for Crx1, immunofluorescence for syntaxin, and ISH for Pax6 and Prox1 in Gdf11tm2/tm2 retinas versus wild types. Blue counterstain in syntaxin panels is Hoechst. Arrow indicates subset of Prox1+ amacrine cells absent in Gdf11tm2/tm2 retinas. Arrowhead indicates Prox1+ horizontal cells, which are unaffected. Scale bar, 100 μm. (B) Retinal explants were grown for 3 or 4 days with or without GDF11 (50 ng/ml). GDF11 treatment reduced the number of Brn3b-expressing cells, but increased Crx1- and syntaxin-expressing cells. Scale bars, 100 μm.

We further tested the idea that GDF11 controls amacrine and photoreceptor cell number, as well as RGC number, by using retinal explant cultures to examine effects of exogenous GDF11 on wild-type retinas (26). E13.5 retinal explants grown in GDF11 exhibited a large reduction in Brn3b+ RGCs, whereas expression of both Crx1 and syntaxin were increased with GDF11 treatment (Fig. 3B). These findings support the idea that GDF11 is an important regulator of all three retinal cell types.

The finding that RGC genesis is increased, whereas amacrine and rod production are decreased, in Gdf11 nulls led us to hypothesize that Gdf11 regulates induction of cell-intrinsic changes by which progenitor cells lose competence to produce RGCs and acquire competence to produce later-born cell types. If GDF11 directly controls progenitor cell competence, Gdf11 mutants might exhibit changes in expression of factors that determine competence states. Math5 is among the first such factors expressed during retinal neurogenesis and is required for competence to produce RGCs (2729). Math5 expression is initiated normally in Gdf11tm2/tm2 retinas, but mutants maintain high levels of expression in the NBL for an abnormally long period. Normally, Math5 expression is down-regulated in central NBL by E16.5 and is essentially absent by E18.0; in Gdf11tm2/tm2 retinas, however, Math5 expression is still evident at these ages (Fig. 4A). Conversely, down-regulation of Math5 expression occurs prematurely in Fst–/– retinas (Fig. 4B) and is accelerated when retinal explants are cultured in GDF11 (Fig. 4C). The prolonged period of Math5 expression in Gdf11tm2/tm2 retinas corresponds to the period of prolonged RGC genesis (Fig. 2).

Fig. 4.

GDF11 and Fst regulate temporal changes in Math5 expression. (A) ISH for Math5 in retinas of Gdf11tm2/tm2 and wild-type littermates. (B) ISH for Math5 expression in Fst–/– and wild-type littermate retinas. (C) E13.5 retinal explants grown in presence or absence of GDF11 (50 ng/ml) for 2 days (E13.5 + 2 days), then hybridized with probe to Math5. (D) Expression of Mash1 and NeuroD in retina in Gdf11tm2/tm2 mice and wild-type littermates. oe, olfactory epithelium. (E) Early onset of Mash1 expression in Fst–/– retinas. Asterisks in (A) to (E) indicate reproducible differences from wild-type controls. Scale bars, 200 μm.

The alteration in the period of Math5 expression in Gdf11tm2/tm2 retinas is accompanied by a shift in onset of expression of two other proneural genes, Mash1 and NeuroD, which are involved in the development of bipolar and amacrine cells (30, 31). In Gdf11tm2/tm2 embryos, expression of both genes is barely detectable at E14.5, when substantial levels are seen in wild types (Fig. 4D). Conversely, Mash1 expression occurs prematurely in Fst–/– retinas, at E13.5, when wild-type littermates express only low levels of Mash1 (Fig. 4E). By E17.5, both Mash1 and NeuroD expression recover to normal levels in Gdf11tm2/tm2 retinas (Fig. 4D), which suggests that progenitor cells can acquire competence to produce later-born cell types even though Math5 expression (and RGC genesis) remain elevated. Altogether, these observations suggest that GDF11 regulates the timing of progenitor competence by controlling the expression of genes crucial for progenitor cell fate determination.

Does GDF11 regulate generation of all retinal cell types, or only some? Because Gdf11tm2/tm2 animals die at birth (14), this question cannot yet be answered with certainty. Expression of Lim1, a horizontal cell-specific transcription factor (32), appears to be normal in Gdf11tm2/tm2 retinas (fig. S6), although changes in expression of a number of other regulatory genes expressed by retinal progenitors are observed (fig. S7). However, the absence of an effect on horizontal cells indicates that GDF11 signaling does not regulate production of all cell types in the retina. Instead, it must govern either a specific subprogram of retinal neurogenesis or act on only a subset of multipotent progenitor cells. This last idea suggests that early retinal progenitors, despite possessing the potential to give rise to all retinal cell types, are nonetheless heterogeneous, at least with respect to their capacity to respond to GDF11.

Our finding that GDF11 governs retinal progenitor cell fate without altering proliferation supports the idea—suggested by retroviral lineage studies—that regulation of cell division and cell-type determination occur independently in the retina (2). Moreover, our results highlight the difference in feedback mechanisms employed in different regions of the developing nervous system to effect proper neuron number. In retina, feedback regulation of neural cell number, mediated by GDF11 expressed by the earliest born neurons, is accomplished by altering the fates of multipotent progenitor cells independent of proliferation. In other regions, such as OE, neuronal GDF11 feeds back to regulate progenitor cell proliferation, independent of changes in cell fate (14). Finally, these studies demonstrate the diversity of action of GDF11 itself. In OE, GDF11 exerts its antineurogenic action by inducing reversible cell-cycle arrest in committed progenitors through increased expression of the cyclin-dependent kinase inhibitor p27Kip1 (14). In retina, by contrast, GDF11 controls the time course of expression of genes that regulate competence to produce RGCs, but neither p27Kip1 levels nor cell proliferation are affected (figs. S3 and S8). Thus, GDF11 acts as a negative feedback regulator of neurogenesis during development by altering either progenitor cell proliferation or progenitor cell fate in different tissues.

Supporting Online Material

Materals and Methods

Figs. S1 to S8


References and Notes

Stay Connected to Science

Navigate This Article