Author Affiliations

• ^aDepartment of Biological Structure, and
• ^bMedical Science Training Program, 357420 Health Science Center, University of Washington, School of Medicine, Seattle, WA 98195
  • Edited by Fred H. Gage, The Salk Institute for Biological Studies, San Diego, CA, and approved October 20, 2008 (received for review August 1, 2008)

Abstract

Müller glia can serve as a source of new neurons after retinal damage in both fish and birds. Investigations of regeneration in the mammalian retina in vitro have provided some evidence that Müller glia can proliferate after retinal damage and generate new rods; however, the evidence that this occurs in vivo is not conclusive. We have investigated whether Müller glia have the potential to generate neurons in the mouse retina in vivo by eliminating ganglion and amacrine cells with intraocular NMDA injections and stimulating Müller glial to re-enter the mitotic cycle by treatment with specific growth factors. The proliferating Müller glia dedifferentiate and a subset of these cells differentiated into amacrine cells, as defined by the expression of amacrine cell-specific markers Calretinin, NeuN, Prox1, and GAD67-GFP. These results show for the first time that the mammalian retina has the potential to regenerate inner retinal neurons in vivo.

• amacrine Gabaergic neuron glia

It is well established that the retina of cold-blooded vertebrates regenerates very well after damage (1–3). The avian retina is also capable of limited regeneration of new neurons following neurotoxic damage (4). In both systems, damage to retinal neurons causes Müller glial cells to re-enter the cell cycle, after which they dedifferentiate into retinal progenitors, and ultimately differentiate into neurons. In fish, all types of neurons are regenerated. However, in chicks, only a limited number of different types of inner retinal neurons (amacrine, bipolar, and ganglion cells) are produced; few, if any, photoreceptors are regenerated.

In the mammalian retina, by contrast, the proliferative response of Müller glia to injury is even more limited than in chicks. In response to injury in mouse or rat retina, the Müller glia may become reactive and hypertrophy, but few re-enter the mitotic cell cycle. Due to the lack of a spontaneous regenerative response in the mammalian retina, several groups have attempted to stimulate regeneration with intraocular injections of growth factors and/or transcription factors, in vitro or in vivo (5–10). Taken together, the studies of the mammalian retina indicate that Müller glia have a very limited proliferative response to injury, but can be stimulated to re-enter the cell cycle after photoreceptor or inner retinal neuron injury. These studies also reported that some of the progeny of Müller glial mitotic divisions go on to differentiate characteristics of rod photoreceptors. However, these studies failed to detect regenerating inner retinal neurons after damage, unless they were transfected with genes that specifically promote amacrine fate (9). This is puzzling, since in the chick, amacrine cells are the primary neuronal cells that are regenerated after injury. In an attempt to resolve this issue, we have carried out a systematic analysis of the response to injury in the mouse retina, and the effects of growth factor stimulation on Müller glial proliferation. The previous studies have used rats because of the relative ease of intraocular injections, but mice have the advantage that it is possible to use a variety of genetically manipulated animals to verify neural regeneration. We find that as in the rat, very few Müller glia re-enter the cell cycle after N-methyl-D-aspartic acid (NMDA) mediated damage to inner retinal neurons; however, the proliferation can be greatly stimulated by intraocular injection of either EGF, FGF1, or the combination of FGF1 and insulin. Although most progeny of the Müller glia maintain their expression of Müller glial markers, a small number of bromodeoxyuridine (BrdU) labeled cells differentiate into amacrine cells, as indicated by their expression of NeuN, Calretinin, Pax6, and GAD67-GFP. Thus, our results show for the first time that the mammalian retina has the potential to regenerate inner retinal neurons in vivo.

Results

Neurotoxic Damage and Stimulation of Proliferation. To test whether Müller glia can regenerate inner retinal neurons in the mouse retina, we elicited retinal damage with NMDA. Overstimulation of NMDA-activated ionotropic glutamate receptors causes cell death of retinal ganglion cells (RGC) and amacrine cells (11–13). We confirmed in adult mouse retina that NMDA induced damage leads to loss of Brn3-positive RGCs and GAD67-GFP GABAergic neurons, but not bipolar cells. (supporting information (SI) Fig. S1 A and B).

Previous studies have found only limited Müller glia proliferation after various forms of retinal injury in the mammalian retina; however, a recent report shows that retinal damage causes increased EGF receptor expression in Müller glia (5). We therefore reasoned that NMDA induced damage to the retina might elicit a similar increase in the sensitivity of Müller glia to EGF. Previous studies in chick retina demonstrated that the combination of FGF and insulin also stimulated proliferation of Müller glia. Therefore, we made intraocular injections of either EGF or the combination of FGF and insulin in adult mice that had received an injection of NMDA two days previously (Fig. 1A). At this time, the mice also received BrdU to identify any cells that re-entered the cell cycle. We found that there were few BrdU+ cells 24 h following an injection of EGF in undamaged retina (Fig. 1B; 24 h after EGF), but there was a large, dose dependent increase in proliferation in animals that received EGF after prior NMDA-induced damage (Fig. 1 C and D).

Fig. 1. Müller glia proliferation in adult mouse retina. (A) Treatment scheme: NMDA induced damage at day 0 (D0) with subsequent EGF injection at day 2 (D2). BrdU was applied at D2 to identify cell proliferation. (B–F) NMDA damage followed by EGF treatment induces significant proliferation (C, D, F), while EGF (B) or NMDA neurotoxic damage (E) alone do not (E-F, BrdU = black pseudosquares). RGCs are shown in retinal flatmounts of adult mice labeled for neurofilament M (NFM, red) without (B) and after cell loss due to NMDA neurotoxic damage (C and D). (E) NMDA alone produces some proliferating cells around the optic nerve head and occasional BrdU+ cells appear at the very periphery. NMDA damage followed by EGF (F) results in significant proliferation and correlates with the amount NMDA given (C and D). (G) There are only slightly more BrdU labeled cells in the central retina compared with the peripheral retina after NMDA damage and EGF treatment (see also Fig. S2A). (H) The number of BrdU+ cells in NMDA-damaged retinas decreases with time after treatment (see also Fig. S2 B–D). Images show single confocal planes of 2 μm (B-D red channel) and z axis projections of 70 × 1 μm (B-D green channel). (Scale bars: 10 μm (B–D)).

We analyzed the distribution of the proliferating cells (BrdU+) in NMDA/EGF-treated retinas in flatmount preparations at various time points after treatment. Systematic quantitative analysis of cell proliferation in mouse retina over time was conducted by counting BrdU+ nuclei on confocal images taken across the entire depth of flatmounted retina (Fig. S1C). Injection of either NMDA or EGF alone did not induce substantial proliferation in the retina (Fig. 1 B, E, and G). By contrast, NMDA at day zero (d0) followed by a single injection of EGF 48 h later (d2) resulted in a striking increase in proliferation across the retina (day 3: Fig. 1F–H). Even as early as 4 h after the EGF injection there were up to 1728 BrdU+ cells/mm² (1168 ± 203 BrdU+ cells/mm², n = 4, Fig. 1H). The number of BrdU+ cells declined over the next week (Fig. 1H, S2 B–D). By 24 h (d3) after the EGF injection, 57% of the BrdU+ cells remained and at d 7 there was a further decline to 169 ± 17 BrdU+ cells/mm². However, these remaining BrdU+ cells persist for at least 30 days (see below).

Several previous reports have found that in fish, birds, and mammals, Müller glia make up the majority of cells that re-enter the mitotic cycle after retinal damage, although vascular cells, astrocytes and microglia also proliferate after damage (14–16). To confirm that Müller glia were among the cells that take up BrdU+ after NMDA treatment in our experiments, we carried out double-label immunofluorescence for BrdU and Sox2 or Sox9, both nuclear markers for Müller glia (17–19). The Müller cell specific expression was confirmed by co-labeling sections with one of two well-characterized Müller glia markers: cellular retinaldehyde binding protein (CRALBP; Fig. 2 A and B) or a glial high affinity glutamate transporter, called solute carrier family 1 member 3 (GLAST; data not shown). Moreover, GFAP (glial fibrillary acidic protein) is up-regulated in many Müller glia after damage and therefore these cells can be labeled in mice with Cre-recombinase driven by GFAP crossed to ROSA-EYFP reporter mice (= GFAP-Cre::EYFP) (Fig. 2 C and D). Damaging adult mouse retina with NMDA and subsequently stimulating proliferation with various growth factors showed that as early as 4 h after the mitogen, many of the BrdU+ cells were also Sox2+ or Sox9+, residing in the presumptive Müller glia sublayer of the INL (Fig. 2 E and G). Many of the GFAP-Cre::EYFP positive Müller cells are PCNA positive after damage and growth factor injections (Fig. 2F). We also observed GFAP-Cre::EYFP cells co-labeled for other mitotic markers: BrdU, Ki67, and PH3 (Fig. 2 H–J).

Fig. 2. Müller glia proliferation is inducible after retinal damage. Sox2 (A) and Sox9 (B) label Müller glia, as shown by colabeling with CRALBP. After NMDA damage, GFAP-Cre::EYFP labels Müller glia, as shown by double-labeling with Sox9 (C) and CRALBP (D, D′,D″). (E) Confocal image of retina flatmounts show BrdU+/Sox2+ double-labeled cells at depths corresponding to Müller glia as early as 4 h after induction of proliferation by various factors following NMDA damage and EGF. The xz- (E′) and yz-views (E″) represent single optical sections through the indicated area (crosshair). (F) After NMDA damage followed by 4 consecutive injections of FGF1/insulin, Müller glia up-regulate PCNA and treatment (F), BrdU (G and H), Ki67 (I) and phospho-histone3 (J). (K) GFAP-Cre::EYFP Müller glia express also express Pax6. Interestingly, at this time point (D6 after damage and treatment) some GFAP-Cre::EYFP Müller glia can be found in the ONL (C and D). (Scale bars: 10 μm.)

To determine if mouse Müller glia dedifferentiate and to assess whether genes relevant to neurogenesis were re-expressed in these cells when they are stimulated to proliferate with injections of EGF after NMDA damage, we carried out several types of studies. Using immunohistochemistry we found that Müller glia acquire characteristics of retinal progenitors after NMDA/growth factor treatment, shown by their expression of the transcription factor Pax6. Pax6 is normally expressed by retinal progenitors in developing retina and in inner retinal neurons (amacrine/RGCs). Pax6 re-expression is a well-documented step in the dedifferentiation of Müller glia after damage of the fish and chick retina in vivo(20–22) as well as adult rat retina in vitro (10). Following NMDA induced retinal damage and intraocular injections of either EGF or the combination of FGF1 and insulin, the majority of GFAP-Cre::EYFP positive Müller glia are also Pax6 positive (Fig. 2K).

We have recently reported in the chick retina that many genes normally expressed only in developing retina are re-expressed following NMDA induced damage (23). To assess whether the same genes are up-regulated in mouse retina, we used a similar approach. We harvested mouse retinas 4 h after EGF injection (d2), 24 h (d3), 48 h (d4) and seven days later (d9) (Fig. S3A). The left eyes were injected, while the right eyes served as uninjected controls; we used semiquantitative real-time PCR to measure changes in gene expression. Over the time course studied,Notch1 and one of its ligands, Delta1, were both significantly increased, as well as other progenitor specific genes, FoxN4 and Pea3. Two additional progenitor-specific genes, Ascl1 and Neurogenin2 were not up-regulated to levels detectable by RT-PCR (Fig. S3 B and C); however, the RT-PCR changes we observe are likely to miss small changes in expression or even large changes if they are in a small number of cells. Nevertheless, these experiments show that at least a part of the developmental program relevant to neurogenesis is re-expressed in the mouse retina following NMDA damage and stimulation of Müller glial proliferation.

Amacrine Cell Differentiation of Müller Glial Progeny. Pax6 is expressed in both progenitors and postmitotic amacrine cells and ganglion cells. We observed Pax6 expression in BrdU+ cells up to 12 days after injection of various factors (Fig. 3A), suggesting that some of the BrdU+ cells differentiate into amacrine cells. To investigate whether BrdU+ cells were differentiating into amacrine or ganglion cells, we immunolabeled flatmounts or sections from NMDA-damaged and growth factor-treated retinas with antibodies against BrdU and one of two markers of amacrine and ganglion cells: NeuN or Calretinin. Fig. 3shows examples of cells expressing both BrdU and each of these markers. The double-labeled cells were found in either the ganglion cell layer (GCL) or the inner nuclear layer (INL) and had morphologies similar to the other neurons in these layers. We found BrdU+/NeuN+ cells five days following the growth factor injection in both the EGF and FGF1/insulin treatment groups (Fig. 3B). In either of those two treatments between one and three double-labeled NeuN+/BrdU+ cells were present in z-stacks across a 0.05 mm² retinal area (Table 1). Using a second marker for amacrine cells and RGC, Calretinin, we also found double-labeled cells five days after EGF or FGF1/insulin treatment. These experiments show that some of the cells that incorporated BrdU after retinal damage and subsequent growth factor treatment developed characteristics of inner retinal neurons.

Fig. 3. BrdU+ cells express Pax6, NeuN and Calretinin (CalR) and GAD67-GFP. (A–C) After NMDA induced retinal damage and EGF or FGF1 treatment, we found BrdU+ neurons labeled with Pax6 (A), NeuN (B), and Calretinin (C) 8 to 14 days later. Images show single focal planes of 2 μm. (Scale bars: 5 μm in orthogonal and inset views; 10 μm all other views.) GAD67-GFP mice express GFP in immature neurons, a subset of mature RGCs, horizontal and all GABAergic amacrine cells in adult mouse retina. NMDA/4xFGF1+insulin treatment of GAD67-GFP mouse retinas leads to BrdU+ cells in both the ganglion cell layer (F) and the amacrine cell layer (D, E, G and H). (H) At D30 some BrdU+ GAD67-GFP+ cells are also labeled for Prox1. (I) Graphs showing the number of BrdU+/GAD67-GFP+ cells as a function of days after NMDA damage. (J) Most of these cells are located in the inner nuclear layer, but some cells are found in the ganglion cell and horizontal cell layers. Images show zaxis projections of 20 × 0.5 μm, i.e., 20 planes, 0.5 μm apart (D), 1 × 0.5 μm (d), 3 × 0.5 μm (E, e), 16 × 0.25 μm (F, f), 2 × 0.25 μm (G, g), 2 × 1 μm (H), and 1 × 1 μm (h). (Scale bars: 5 μm in orthogonal and inset views; 10 μm all other views.)