Adult Human Müller Glia Cells Are a Highly Efficient Source of Rod Photoreceptors§


  • Serena G. Giannelli,

    1. Stem Cell and Neurogenesis Unit, Division of Neuroscience, San Raffaele Scientific Institute, Milan, Italy
    2. Eye Repair Unit, Division of Neuroscience, San Raffaele Scientific Institute, Milan, Italy
    Search for more papers by this author
  • Gian Carlo Demontis,

    1. Department of Psychiatry, Neurobiology, Pharmacology and Biotechnology, University of Pisa, Pisa, Italy
    Search for more papers by this author
  • Grazia Pertile,

    1. Department of Ophthalmology, Ospedale Sacro Cuore, Negrar, Verona, Italy
    Search for more papers by this author
  • Paolo Rama,

    1. Eye Repair Unit, Division of Neuroscience, San Raffaele Scientific Institute, Milan, Italy
    2. Ophthalmology Unit, San Raffaele Hospital, Milan, Italy
    Search for more papers by this author
  • Vania Broccoli

    Corresponding author
    1. Stem Cell and Neurogenesis Unit, Division of Neuroscience, San Raffaele Scientific Institute, Milan, Italy
    • San Raffaele Scientific Institute, Via Olgettina 58, 20132 Milan, Italy
    Search for more papers by this author
    • Telephone: 39-02-26434616; Fax 39-02-26434621

  • Author contributions: S.G.G.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, final approval of the manuscript; G.C.D.: conception and design, collection and assembly of data, final approval of manuscript; G.P.: provision of study material, final approval of manuscript; P.R.: conception and design, provision of study material, financial support, final approval of manuscript; V.B.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, financial support, final approval of the manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS Month 00, 2010.


There is growing evidence that Müller glia cells (MGCs) might act as regenerative elements in injured retinas of fishes and amniotes. However, their differentiation potential in humans is yet unknown. We isolated Müller glia from adult human retinas and propagated them in vitro revealing for the first time their ability to differentiate into rod photoreceptors. These results were also confirmed with mice retinas. Here, we describe conditions by which human MGCs adopt a rod photoreceptor commitment with a surprising efficiency as high as 54%. Functional characterization of Müller glia-derived photoreceptors by patch-clamp recordings revealed that their electrical properties are comparable to those of adult rods. Interestingly, our procedure allowed efficient derivation of MGC cultures starting from both injured and degenerating and postmortem human retinas. Human transplanted Müller glia-derived photoreceptors integrate and survive within immunodeficient mouse retinas. These data provide evidence that Müller glia retains an unpredicted plasticity and multipotent potential into adulthood, and it is therefore a promising source of novel therapeutic applications in retinal repair. STEM CELLS 2011;29:344–356


Müller Glia cells (MGCs) are organized in a glial scaffold network, which safeguards the complex architecture of the adult retina and its functional homeostasis. Following retinal injury or degeneration, MGCs undergo reactive gliosis, which is a complex process characterized by sustained cell proliferation and migration accompanied by both morphological and molecular changes [1, 2]. During this event, MGCs dedifferentiate and acquire the features of retinal progenitors and at times give rise to new retinal neurons. Indeed, MGC-induced neurogenesis is exceptionally robust in fish, where it can fully regenerate wide areas of damaged retinas [3–7]. Unfortunately, this regenerative activity dwindles rapidly with vertebrate evolution [8, 9] to the point that in rodents a very limited number of Müller cells can re-enter the cell cycle [10]. Nonetheless, the use of exogenous growth factors was seen to increase MGC proliferation [11]. Notably, Karl et al. [12] demonstrated the generation of new retinal neurons in murine retinas from reactivated MGCs. Although few studies, so far, have attempted to define the developmental potential of in vitro adult reactivated MG, Das et al. suggested that mouse MGCs exhibits features of stem cells when cultured in sphere aggregates being able to generate retinal neurons, astrocytes, and oligodendrocytes [13]. These data indicate that the differentiative capability of MG can be enhanced to a large extent when removed from its natural environment. Similar conclusions seem to apply to humans as well, even though these studies are limited to immortalized MG cell lines [14]. Despite these data, the potential of human MG to generate functional photoreceptors has been poorly characterized and therefore remains uncertain thus far.


Isolation and Collection of Human Retinal Samples

Fresh human retinas were obtained from vitreoretinal surgical procedures. The three main surgical procedures that permit the collection of retina samples without increased intraoperative or postoperative complications (removal was in any case necessary) are the following:

  • (a)Vitrectomy for complicated retinal detachment (RD). In some cases of recurrent or long-standing (RD) part of the retina contracts and loses elasticity, and it cannot be reattached without removal of the part causing traction.
  • (b)Macular translocation in age-related macular degeneration (AMD) with choroidal neovascularization (CNV). This procedure is indicated in eyes that cannot benefit from less invasive therapies, such as intravitreal anti-VEGF (Vascular Endothelial, Growth Factor) injections. The aim of this operation is to relocate the macula on an area of healthy retinal pigment epithelium (RPE). On completion (RD), the peripheral retina is cut at the ora serrata, CNV is then removed, and the retina is rotated by about 45° around the optic nerve. In this way, a part of the temporal retina is displaced to the nasal part. As the temporal retina usually has a longer radius then the nasal one, the excess covers the pars plana and needs to be removed to reduce the risk of complication.
  • (c)Vitrectomy following traumatic sclera laceration. In the case of sclera lacerations involving the retina, recent surgical guidelines suggest the removal of 1–2 mm of retina surrounding the wound to prevent scar tissue from the injured area from pulling the surrounding retina into the lesion.

The undesirable part of the retina can be removed using a pneumatic vitreous cutter. To collect the retina, a syringe with a special connector was interposed along the aspiration line of the vitreous cutter. This procedure allowed collection of the retinal tissue without any change in surgical technique. Collected retinal samples were transferred into a test tube then kept in sterile balanced salt solution at 4°C before proceeding with tissue manipulation and cell culture.

Postmortem human eye bulbs were obtained from cornea donors collected by the Italian Eye Bank (Fondazione Banca degli Occhi del Veneto, Venice, Italy). The research followed the tenets of the Declaration of Helsinki. Retinas were dissected out of the bulbs to separate them from the RPE and the ciliary margin. Dissected retinas were then processed for primary cell culture.

Primary Cell Cultures of Human Müller Glia

Human retinal samples were digested with papain (10 U/ml, Worthington, Lakewood, NJ, USA, in Hanks' balanced saline solution and containing 1 mM cystein and 0.5 mM EDTA and triturated with glass pipettes. Cells were washed twice and finally plated on Matrigel (BD, Franklin Lakes, NJ, USA, dishes in neural stem cell medium (NSC medium) containing DMEM-F12 (Invitrogen, Carlsbad, CA, USA, supplemented by 20 ng/ml bFGF (basic Fibroblast Growth Factor, 20 ng/ml EGF (Epidermal Growth Factor), 4 mg/ml heparin, 0.1 mg/ml apo-transferrin, 25 mg/ml insulin, 1 mg/ml putrescine, 20 ng/ml progesterone, 30 ng/ml sodium selenite, 0.2% BSA (Bovine Serum Albumine) (Sigma-Aldich, St. Louis, MO, USA, at a cell density ranging from 0.1 to 2 × 106 cells per centimeter squared. Clones appeared about a week after seeding, reached confluence in 2–3 weeks, and could then be amplified 5–7 times by trypsin enzymatic digestion. Differentiation was achieved by plating cells in NSC medium, either alone or on a cell feeder layer composed of mitomycin-inactivated PA6 mouse bone marrow stromal cells. Twenty-four hours after cell seeding, mitogens were sequentially removed and, thereafter, B27 supplemented (Invitrogen). Small molecules such as taurine (1 mM, Sigma-Aldich) or the γ-secretase inhibitor L685,458 (1–50 nM, Sigma-Aldich) were added to media. For cell transplantation, MGCs were first infected with a hPGK-GFP lentivirus, and 24 hours later were plated on PA6 cells and primed for differentiation in presence of taurine (1 mM, Sigma-Aldich) and bFGF. Four days later, cells were detached and resuspended in phosphate-buffered saline (PBS) for retinal injection.


Cells were fixed with 4% paraformaldehyde (30 minutes, room temperature [RT]). Fixed cells and cryosections (20 mm) were treated with unspecific binding blocking solution (10% donkey serum, 0.1% Triton X-100 in PBS) followed by primary antibody incubation (4°C, overnight), after which secondary antibody incubation (1 hour at RT) was performed. Finally, slides were mounted using DakoCytomation mounting medium (Dako, Carpinteria, CA, BrdU immunodetection needed a DNA denaturation step (2 N HCl, 10 minutes, RT) before the blocking reaction. Primary antibodies are listed in Supporting Information Table 2. Hoechst was used for nuclei counter staining. For testing senescence, cells were fixed in 1% formaldehyde/0.1% glutaraldehyde (5 minutes, RT) and then incubated (10 hours, 37°C) with X-Gal (0.75 mg/ml, 5-bromo-4-chloro-3-indolyl-b-D-galactosidase, Promega, Madison, WI, USA,, in 5 mM K4Fe(CN)6, 5 mM K3Fe(CN)6, 0.02% NP-40, 2 mM MgCl2) acidic solution (Na2PHO4; pH = 4) to visualize the endogenous β-galactosidase activity. Fluorescent immunosignals were imaged with a laser-scanning confocal microscope (SP5, Leica Microsystems, Wetzlar, Germany, and with fluorescent microscopes (Eclipse TE200 and E600, Nikon, Tokyo, Japan, Images were processed with PhotoshopCS2 (Adobe, San Jose, CA, USA,

Gene Expression Assays

Total RNA was extracted from retinas with TRIZOL reagent (Invitrogen), whereas cell RNA was extracted with RNaesy Micro kit (Qiagen, Hilden, Germany, cDNA was synthesized using Transcriptor HF RT-PCR system (Roche Molecular Biochemicals, Basel, Switzerland, following the supplier's protocol and used as template for either polymerase chain reaction (PCR) reactions or quantitative (q) PCR reactions. PCR reactions were performed in 25 μl containing cDNA, 500 nM primers (as listed in Supporting Information), 1× PCR master mix, 200 mM dNTPs, 0.025U TaqGold (Applied Biosystems, Carlsbad, CA, USA,, and 50 ng/ml cDNA. Quantitative PCR analysis was carried out with MX3000P System (Stratagene-Agilent Technologies, Santa Clara, CA, USA, in Clear Thin Wall Tubes (Axigen, Union City, CA, USA, PCR mix included 1× SYBR®Green PCR Master Mix (BioRad, Hercules, CA,, 100 nM primers (as listed in Supporting Information Table 7) and 10 ng/μl cDNA. Triplicate amplifications were carried out for each target gene. The output data were transferred to Microsoft Excel for analysis. Quantification was performed through a standard curve-based method and normalized to the expression of the housekeeping gene β-actin.

Electrophysiological Recordings

For patch-clamp recordings, cells were stained by Syto-13 (Invitrogen) and human MGC (hMGC)-derived cells were identified by their condensed nuclei. Alternatively, cells were infected with a vesicular stomatitis virus-pseudotyped lentivirus (LV) expressing the GFP under the control of the rhodopsin minimal promoter. The dish was mounted on the stage of a DMI-4000B Leica microscope and a picture of the recorded cell was taken using a Leica DFC-350FX camera. Perforated-patch recordings were carried out as previously reported for acutely dissociated adult mouse rods [15].

Statistical Analysis

All counts are presented as average and SEM. Statistical analysis for correlation and significance was carried out using Graphpad Prism software (GrphPad Software Inc., La Jolla, CA, USA,

Retinal Cell Transplantation

All experiments involving the use of animals were approved by the local Institutional Animal Care and Use Committee. For in vivo transplantation, ∼0.3 μl of PBS cell suspension (4,000 cells per microliter) was injected into the subretinal space of Rag2/g(c) immunodeficient mice at postnatal day 2. Briefly, mouse pups were anesthetized through hypothermia; the eyelid was opened with an incision and the sclera punctured with a 30-gauge needle. Through this incision a 33-gauge needle was inserted and cell suspension was injected in the subretinal space. Injected retinas were harvested 3–4 weeks later.


Derivation of Primary Cultures of Human and Mouse Adult MGCs

We thus set out to investigate the regenerative potential of hMGCs in vitro, isolating proliferative cells from surgical adult retina explants. Retinal specimens were obtained from patients whose pathology required surgical treatment, such as vitrectomy. These pathologies were varied and clinically different, but the prescribed surgical procedures for them all implicated the removal of part of the peripheral retina (see Materials and Methods section). In particular, samples were collected from retinas affected by trauma, pathological RD, and AMD (Supporting Information Table S1A).

Vitreoretinal specimens were enzymatically digested, triturated to single cells and cultured with NSC medium on Matrigel-coated plates. Adherent populations of proliferative cells emerged 2 weeks after plating (Fig. 1A, 1B) and were propagated by single-cell dissociation (Fig. 1C). Primary cells were highly homogenous, with an elongated morphology, and the majority expressed MGC markers such as glutamine synthase (GS), GFAP (Glial Fibrillary Acidic Protein), vimentin, and CD44 (Fig. 1D–1I). Moreover, markers of reactive Müller glia such as nestin (Figs. 1I and 3E) and Pax6 when in association with other MG markers (Fig. 1G) or with Chx10 (Fig. 3F) were also present. Conversely, none of them were found immunoreactive for βIII-tubulin, aPKC, calbindin, recoverin or Pmel17 indicating the absence in our cultures of retinal neurons, photoreceptors and RPE cells, respectively (Fig. 1F). Likewise, Iba-1+ and CD31+ cells were not scored, indicating the lack of microglia and endothelial cells, respectively (Fig. 1I).

Figure 1.

Efficient generation and characterization of human Müller glia cell (hMGC) cultures from human vitreoretinal explants. (A–C): Bright-field images of three different human MGC primary cultures taken at different stages of culture progression. (D–I): The majority of the cultured cells result immunoreactive to MGC markers such as GS (D), nestin (D, I), GFAP (E, H), vimentin (F), Pax6 (G), and CD44 (H), but not photoreceptor (Rec) (F), or microglial (Iba1) (I) markers. Assessment of active cell cycle progression of hMGCs is assessed by GFAP and Ki67 coimmunolabeling (E). (J): Reverse transcriptase polymerase chain reaction analysis confirming GFAP, GS, and Cralbp expression in the derived cultures as well as the other MGC-specific markers Sox2 and Rx. On the contrary, the RPE marker RPE65 does not result expressed. (K): Graph showing three cell growth curves from different representative hMGC cultures. Once cell growth plateau was reached, human culture showed positive for acidic LacZ staining (L) indicating senescence state. Scale bar = 250 μm (B), 100 μm (A, C, I), 50 μm (D, E, F, G, H). Abbreviations: GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFAP, glial fibrillary acidic protein; GS, glutamine synthase; hMC, human Muller glia Culture; hRet, human retina; Rec, recoverin; RSC.

Accordingly, reverse transcription polymerase chain reaction (RT-PCR) analysis confirmed the expression of MGC-restricted transcripts such as GS, GFAP, Nest, Cralbp, Rx, Sox2, and conversely, the absence of the RPE-specific marker RPE65.

These in vitro conditions allowed us to stringently select a pure population of hMGCs. Although morphologically comparable overall, only a subset of cultured hMGCs (25% ± 6%) was actively proliferating, as tested by both BrdU incorporation and Ki67 immunoreactivity, suggesting different ability in re-entering the cell cycle after in vitro plating (Fig. 1E). Most of the established primary lines exhibited an exponential growth curve for about five to eight passages corresponding to around 19 population doublings and were done on average once a week after reaching 90% confluence (Fig. 1K). Nonetheless, all cultures invariably ended proliferation and undertook cell senescence as assessed by acidic-LacZ staining (Fig. 1L).

Fifty-two explants were processed, 32 of which gave rise to proliferating hMGC cultures, estimating the overall success rate at 61% (Supporting Information Table S1A). The major critical constraint limiting the successful establishment of the cultures was the initial amount of cells available from each specimen. In fact, starting with more than 2 × 106 cells the efficiency of cell line derivation was above 77%, which fell to 40% with less than 1 × 106 cells (Supporting Information Table S1B).

Donor age also influenced the positive outcome of the procedure, as the success rate declined abruptly in patients over 60-year old from 79% (<60 years) to 47% (>60 years) (Supporting Information Table S1C). Conversely, the gender and the interval between explant isolation and its processing (limited to 7 days) did not particularly affect the outcome of the procedure (Supporting Information Tables S1D and S1E). Proliferating hMGCs were obtained with high efficiency from trauma (65%, 7 of 11) and RD (89%, 17 of 19), whereas AMD-affected retinas gave rise to primary cell lines with a rate of only 41% (9 of 22; Supporting Information Table S1A). However, the relatively high mean age of the patients suffering from this disease (> 50 years) might also contribute to this poor outcome.

Similar findings were obtained when we used postmortem retinas from cornea donors instead of surgical explants (see Materials and Methods section). In fact, primary cultures of homogeneous MGCs were successfully obtained from autoptic tissue harvested within 48 hours from death, with an overall efficiency of 57% (4 of 7; Figs. 2A–2E).

Figure 2.

Generation, characterization, and differentiation of human Müller glia cell (hMGC) cultures from human autoptic tissue. (A–E): Characterization of hMGC primary cultures derived from postmortem retinas (A, B). Bright-field images of two different primary cultures (C, D). Immunofluorescence staining of hMGC culture shows positivity for proliferation markers: Ki67 (D); and MGC reactivation: Nestin (C), GFAP (C), Vime (D). (E): Graph showing growth curves from different representative cultures. (F–M): Differentiation of hMGC primary cultures derived from postmortem retinas. (F, G): Nuclear counterstaining of hMGC differentiated either without (F) or with PA6 feeder layer (G) and its corresponding recoverin immunostaining (F′, G′) reveal PA6 cell ability to promote photoreceptor commitment. Quantification of this phenomenon is shown (M). (H–L): Higher magnification of recoverin-immunolabeled hMGC-derived photoreceptors in absence (H, I) or presence (J–L) of PA6 cells. No major difference in morphology is appreciable. Scale bar = 100 μm (A, B), 20 μm (F, F′, G, G′), 20 μm (C, D), 10 μm (L), 5 μm (H–K). Abbreviations: GFAP, glial fibrillary acidic protein; hMC, human Muller glia Culture; Nest, nestin; Vime, vimentin.

We next sought to derive primary MGC cultures from adult mouse retinas (mMGCs), applying the same procedures established for human tissue. Dissociated retinas from P30-40 mice were plated at a cell density of 1 × 106 per centimeter square and cultured as previously described. Two weeks after plating, a monolayer of cells with bipolar elongated morphology was obtained (Fig. 3A, 3B). Primary cultures showed immunoreactivity for many MGC markers, such as GS, GFAP, nestin, CD44, and Rx and coexpression of Pax6 and Chx10 (Fig. 3C–3H). In contrast, no cells were found positive for βIII-tubulin, pMel17, rhodopsin, and Iba1, whereas only a few S100+ astrocytes (2% ± 0.5%) were identified, indicating the presence of a negligible fraction of other cell types in these preparations. RT-PCR analysis confirmed these conclusions (Fig. 3M). To further corroborate these findings, we also established primary cultures from GFAP-GFP transgenic retinas. Accordingly, most of the adherent cells (97%) at P1 expressed GFP (Green Fluorescent Protein), (97% ± 2%) further supporting the establishment of a highly homogenous culture of MGCs (Fig. 3I).

Figure 3.

Derivation of homogeneous mMGC cultures from adult murine retinas. Proliferation of mMGCs was characterized using 24 hours BrdU staining at different time points of primary culture growth ([A], 48 hours, small clones; [B], 8 days, bigger clones), Ki67 immunostaining in secondary cultures (E) and determining cell growth curves of different lines, three of which are represented in (N). Müller glia identity of this proliferating culture is ascertained by their expression of a number of MGC key markers: GS ([C], [M]), GFAP ([D], [J], [M]), nestin (Nest, [E], [G], [L]), Sox2 (G), and CD44 (H) both by immunofluorescence (C–H) and reverse transcriptase polymerase chain reaction (RT-PCR) analysis (M). Coexpression of Pax6 and Chx10 (F) also confirms the MGC phenotype. On the contrary, absence of immunoreactivity for βTubIII (J), pMel17 (K), and Iba1 indicates absence of neuronal, retinal pigment epithelium, and microglial cells. (I): Cells obtained by sorting of GFAP-GFP retinas gave a more homogeneous and comparable cell population. (M): RT-PCR analysis reveals the presence of MGC-specific molecular genes such as Cralbp, GFAP, nestin, Rx, and Sox9 and the absence of genes associated with RPE (Mitf and Tyr), endothelium (vWF), and microglia (CD45). (N): Senescence as tested by acidic-Xgal staining (O) occurred in all MGC cultures on extended in vitro culture. Scale bar = 250 μm (I), 100 μm (A, B, O), 50 μm (C, E, G, K, L), 20 μm (D, F, J), 10 μm (H). Abbreviations: BrdU, 5-bromo-2-deoxyuridine; GFAP, glial fibrillary acidic protein; GFP, Green Fluorescent Protein; mMC, mouse Muller glia Culture; mRet, mouse retina; MGC, Muller glia cell; Rx, Retinal Homeobox; Mitf: microphthalmia-associated transcription factor; Tyr, tyrosinase; vWF, von Willebrand factor.

In adulthood, MGCs remain in an enduring resting state unless challenged by retinal damage or disease. This promotes morphological and biochemical changes coupled with cell cycle re-entry, a general complex process known as glial reactivation [16, 17]. We, therefore, wondered whether in vitro proliferating mMGCs exhibited features of this reactivated state. We found, with both immunohistochemistry and qPCRs, that cells kept in culture for at least 12 days modified their expression profile activating GFAP, Sox2, CyclinD3, ceruloplasmin, and nestin and downregulating GS, Kip1, Kir4.1, and acquaporin-1 (Supporting Information Fig. S1O–S1P), as previously described for murine-reactivated MGCs in vivo [18, 19]. These findings indicate that in vitro culture conditions promote mMGCs to gain an activated proliferative state.

Induction of Mouse Adult MGCs into Neurons and Photoreceptors

We then sought to determine the differentiation potential of both mouse and human in vitro cultured MGCs. We initially established differentiation conditions, starting from mouse cells, with the aim to subsequently translate those conditions to human cell cultures. Differentiation was induced by sequential withdrawal of growth factors with B27 supplementation for at least 5 days (Fig. 4A). After this period, the cells changed morphology acquiring a multibranched shape and organizing in an intricate cellular network. Importantly, although most of the cells turned into postmitotic GFAP+ and GS+ mMGCs (Fig. 4B–4E), we noted a small but constant fraction of βIII-tubulin+ neurons that accounted for 8% of the entire differentiated population (Fig. 4D, 4F, 4I). In line with this finding, G0α/Chx10 as well as NF200/Calb were identified indicating the generation of bipolar and horizontal cell interneurons (Fig. 4D, 4N, 4O). Surprisingly, in some circumstances, we identified rare recoverin+/phosducin+ cells accounting for differentiation along the photoreceptor cell lineage (Fig. 4D, 4G). Together, these observations revealed that mMGCs retained some neurogenic potential that was sufficient to generate neurons and photoreceptors, although in a smaller fraction, even when extensively cultured in vitro. Thereafter, we sought to enhance neuronal fate by stimulating Müller glia differentiation with exogenous factors, such as retinoic acid (1 mM), Noggin (500 ng/ml), and the TGFβ chemical inhibitor SB4131542 (10 μM). However, we were unable to appreciate major improvements in any of these conditions. In contrast, mMGC differentiated in a substantially higher percentage of both neurons and photoreceptor (8% ± 2% to 16% ± 4% and 2% ± 1% to 8% ± 3%, respectively) in presence of the γ-secretase inhibitor L685,458 (4 nM; Fig. 4E–4J). Of the neuronal fractions, G0α and Chx10 double-positive bipolar neurons were the most represented, whereas fewer neurons expressed NF200 and calbindin. No Syntaxin+ and Brn3a+ neurons were observed whatever the conditions used to elicit cell differentiation. On the contrary, the photoreceptor fraction displayed recoverin, phosducin, and rhodopsin coexpression and striking bipolar morphology with a highly condensed nucleus (Fig. 4P–4W). These results are consistent with the crucial role that Notch signaling plays in the suppression of photoreceptor differentiation, during both the retinal development and the in vitro differentiation of mouse embryonic stem cells [20–22]. To confirm the glial origin of the observed neuronal progenies, we took advantage of a genetic tracing system capable of labeling the MGC and its cell progeny that consisted of a human GFAP-Cre; Rosa26rEYFP double transgenic mice, where Cre-mediated recombination allows EYFP (Enhanced Yellow Fluorescent Protein) expression only in GFAP+ cells and their descendents. Surprisingly, mMGC primary cultures derived from double transgenic retinas once differentiated in vitro gave rise to EYFP+ neurons, providing direct evidence of their glial derivation (Fig. 4K–4M).

Figure 4.

In vitro differentiation potential of murine Müller glia cells (mMGCs). (A): Outline showing the procedure for murine MGC differentiation. (B, C): Two populations of glial cells were identified by immunofluorescence. The first was positive for GS (C) and the second for GFAP (B, E, H) and nestin (B, C), both glial subsets were quantified in (D). Inhibition of Notch pathway using 4 nM L685,458 ([H–J], NI) compared with control (E–G) considerably reduced gliogenesis (D, E, H) and improved neurogenesis, but most importantly photoreceptor formation as assessed by βTubIII staining (F, I) and recoverin/phosducin coimmunolabeling (G, J), respectively, and quantified in (D). (K–M): Two different examples of neurons differentiated from cultured MGCs deriving from GFAP-Cre;Rosa26rEYFP retinas. βTubIII+ neurons displayed EYFP immunoreactivity, thus confirming glial origin. (N, O): In vitro-derived neurons mostly exhibit features of bipolar neurons as indicated by Chx10 and G0α colocalization. (P–W): Four different MGC-derived photoreceptors are shown in single-channel image (P–S) or four-channel merge picture (T–W) indicating colabeling of recoverin, phosducin, and rhodopsin and highlighting their bipolar morphology. Scale bar = 100 μm (E–J), 50 μm (B, C), 20 μm (K, L, M), 5 μm (N, O, U, V), 4 μm (P, Q, R, S, T, W). Abbreviations: bFGF, basic Fibroblast Growth Factor; Cont, control; EGF, Epidermal Growth Factor; GFAP, glial fibrillary acidic protein; GS, glutamine synthase; mMC, mouse Muller glia Culture; NI, Notch inhibitor.

Thus, adult mMGC is endowed with an unexpectedly significant potential toward neuronal and photoreceptor cell lineage commitment. Notably, its full mMGC neurogenic potential is enhanced on Notch inhibition, as occurs with retinal embryonic progenitors, thereby strengthening the close relationship between these two cell types.

Optimized Photoreceptor Commitment in Human Adult MGCs

We used a similar procedure for the induction of hMGC differentiation. Human cells were serially deprived of both mitogens and supplemented with B27 ± L685,458. Under these conditions cells inevitably detached during differentiation, regardless of the substrate used (Matrigel, fibronectin, polylisine, and polylisine/ornithine), causing the loss of the entire culture. Nonetheless, despite their prematurely arrested differentiation, we noted the presence of immature neurons. Therefore, to extend human cell differentiation, we opted for a feeder layer of cells that would provide stronger physical support for our cells. Using inactivated murine embryonic fibroblasts (MEFs) as a feeder layer, we completed MGC differentiation revealing the development of bipolar-shaped recoverin+ photoreceptors (Supporting Information Fig. S2A–S2C′). In addition, we also tested PA6 or MS5 feeder cells, which have a strong neuronal inducing activity in other cellular contexts [23, 24]. Recognition of hMGC from mouse PA6 cells was based on murine-specific heterochromatic nuclear foci and occasionally on Lamin immunostaining selective for human nuclei (Fig. 5E–5J′ and data not shown). On cocultures with stromal cells, hMGC photoreceptor commitment (scored as recoverin+ cells) was enhanced by roughly 57% with respect to MEF cocultures (Supporting Information Fig. S2D–S2G). PA6 cells were effective also in enhancing mMGC photoreceptor differentiation with respect to the basal conditions previously reported (6.5% ± 1.4% vs. 3.1 ± 1.1; Supporting Information Fig. S3). We thereby continued to employ PA6 feeders for hMGC differentiation, which could be maintained for up to 20 days (Fig. 5A). Under these conditions, Nrl+ and Crx+ photoreceptor progenitors were identified as early as 4 days after coculture, whereas recoverin+ photoreceptors appeared from 7 days onward (Fig. 5A). After 20 days of coculturing, recoverin+ photoreceptors accounted for 15% ± 5%, whereas βIII-tubulin+ neurons were 11% ± 4% of all the plated hMGCs (Fig. 5B–5D′). Recoverin+ immature photoreceptors appeared as very small bipolar cells with extremely condensed nuclei and a small cytoplasm/nucleus ratio. Virtually all of them resulted immunoreactive for Crx, the photopigment rhodopsin, and for the α-subunit of the heterotrimeric G protein transducin (GNAT1; Fig. 5E–5J′). Differentiated hMGC resulted also immunoreactive for CNGA1 and Peripherin-2 (Prph2, Rds), two proteins normally detectable in differentiated photoreceptors only (Supporting Information Fig. S4).

Figure 5.

PA6 feeder layers and taurine promote efficient human MGC differentiation into photoreceptor cells. (A): Outline showing the procedure for human Müller glia cell (hMGC) differentiation resulting in the induction of Nrl+ photoreceptor progenitors as early as 4 days (as tested by reverse transcriptase polymerase chain reaction [RT-PCR], boxed) and derivation of recoverin+ photoreceptor after 20 days. (B): Analysis of βTubIII (neurons) and Rec (photoreceptor precursors) immunoreactivity at day 20. (C–J′): Characterization of photoreceptors differentiated from hMGCs on PA6 at day 20. In vitro-derived Rec+ photoreceptors coexpress Crx, a nuclear factor essential for photoreceptor development (C, E, E′, F, G, H, I) but also specific markers for rods, such as rhodopsin (Rho, [D, D′]) and GNAT1 (J, J′). Prevalence of rod fate with respect to cone fate was confirmed by RT-PCR analysis in 20-day differentiated MGCs. (P): Indeed, rod-specific genes (Pde6c, Gnat1, Cnga1, Grk1, Rho) are more robustly expressed than cone-specific genes (G-OPS and R-OPS). Specific and nonspecific photoreceptor transcripts (such as Rec and Tulp1) indicate the high degree of maturation and phototransduction viability in these cells. (E–J′): Higher magnification pictures highlight the bipolar-elongated shape with a small soma of the in vitro-derived photoreceptors. (K–N): Effect of 1 mM taurine (M–N) compared with control (K–L) on photoreceptor differentiation from cultured hMGCs. (K, M): Representative magnification images of Hoechst nuclei counterstaining and (K′, M′) recoverin immunostaining clearly illustrates photoreceptor differentiation induced by taurine. (L, N): Morphology of control (L) and taurine-promoted (N) photoreceptors at high-power view. Samples in (D, E, E′, H, I, J, K, L, M, N) are counterstained with DAPI to label cell nuclei. (O): Quantification for each differentiated cell type (GFAP+, Müller glia; recoverin+, rods; G0α-Chx10, bipolar neurons; Syntaxin+, amacrine cells) comparing conditions with (blue bars) or without (red bars) TAU. *, p <.05. (P): Gene expression profiling by RT-PCR analysis of genes coding for key molecular components of the transduction pathway. As positive control tissue, adult hRET was used. Scale bar = 100 μm (K, K′, M, M′), 50 μm (B, C, D, D′), 20 μm (L, N), 10 μm (H), 5 μm (E, E′, F, G, I, J). Abbreviations: βTubIII, βtubulin-III; bFGF, basic Fibroblast Growth Factor; EGF, Epidermal Growth Factor; GFAP, glial fibrillary acidic protein; GNAT1, Guanine Nucleotide binding protein, Alpha Transducing 1; hMC, human Muller glia Culture; hRET, human retina; Rec, recoverin; Rho, rhodopsin; TAU, taurine.

G0α+/Chx10+ bipolar, Syntaxin+ amacrine, and NF200/calbindin+ horizontal cells at a relative ratio of 4.9% ± 1.0%, 1.3% ± 0.5%, 3.5% ± 7%, respectively, were also identified among the differentiated neuronal progenies (Supporting Information Fig. S5). We obtained similar findings when MGC cultures derived from autoptic tissue were differentiated, generating recoverin-positive photoreceptors that were morphologically identical to their counterparts from living-donor cells. Once again, we observed an evident increase in photoreceptor differentiation employing a PA6 feeder layer compared with basal conditions (Fig. 2F–2M). In the attempt to further enhance photoreceptor differentiation, we chose to combine cocultures with soluble factors. Initially, the treatment with the γ-secretase inhibitor L685,458 did not modify photoreceptor induction in the hMGC setting either at the concentration applied to murine cells (4 nM) or at other concentrations (1 nm, 10 nm, 50 nm). We, then, tested taurine, a rod-inducing molecule secreted from retinal cells that has positive effects on the in vitro differentiation of embryonic stem (ES) cells and induces pluripotent stem cells in photoreceptors [11, 25]. Although our results indicate that human MGCs, unlike mMGCs, are resistant to the neurogenic stimulus induced by the inhibition of Notch signaling, taurine supplementation (1 mM), together with stromal feeder cocultures, displayed an additive effect that led to significant enhancement of the photoreceptor commitment (44% more of recoverin+ cells) in the global differentiated progeny (Fig. 5K–5N). Conversely, taurine failed to elicit an evident increase of the neuronal population, suggesting an inductive activity limited to the photoreceptor lineage (Fig. 5O).

We performed RT-PCR analysis confirming GNAT1, Rec, and Rho expression on MGC-differentiated progenies isolated from surgical samples and revealed a plethora of genes coding for other elements of the phototransduction cascade, including PDE6B, GRK1, CNGA1, green OPSIN, and red OPSIN cone photopigments (Fig. 5P). Finally, the expression of TULP1, a critical gene for photoreceptor synapse formation, was revealed; in vivo, this has been described as one of the last to appear during human photoreceptor differentiation [26, 27] (Fig. 5P).

To evaluate the functionality, we initially attempted to record light responses from hMGC-derived cells after allowing rhodopsin regeneration by priming with a commercially available analog of retinal (9-cis retinal), as previously reported for Xenopus tadpoles [28]. We used infrared light to avoid bleaching of regenerated rhodopsin by visible light, but found visualization of hMGCs growing on the feeder layer impossible. Discrimination of MGCs from the feeder PA6 cells, required their labeling using either the nuclear dye Syto13 (Supporting Information Fig. S6A–S6C) or a lentivirus expressing the GFP under the control of the rhodopsin promoter [29] (see Materials and Methods section), showing that only the recoverin+ MGCs activate the GFP transgene (Supporting Information Fig. S6D–S6F″). In all cases, exposure to high-intensity light (470–490 nm) close to the rhodopsin absorption peak, for EGFP or Syto 13 excitation, is expected to fully bleach all regenerated rhodopsin.

Functional properties were, then, investigated by using the patch-clamp technique to define the electrophysiological properties of 12-day-old hMG-derived cells. Using perforated-patch recordings to prevent the loss of macromolecular components, similar membrane potential, electrical resistance, and capacitance were measured for hMG-derived cells (−52.5 + 3.2 mV; 2.2 ± 0.3 GΩ; 3.1 ± 0.2 pF) and adult mouse rods (−54.1 + 2.7 mV; 1.5 ± 0.5 GΩ; 4.2 ± 0.6 pF). Membrane capacitances, an index of cell membrane area, indicate that recordings were obtained from cells having small size, such as adult rods.

Interestingly, differentiated MGC-derived cells also express the transcripts for a set of genes coding for the potassium channel subunits that play key roles in visual signal processing by rods and also affect photoreceptor viability [30] (Fig. 6A). Among the expressed genes, KCNB1 codes for a retina-specific voltage-gated potassium channel that is first rapidly activated and then slowly inactivated by membrane depolarization, prompting us to investigate its functional expression by hMGC-derived cells. Indeed, hMGC-derived rod photoreceptors generate fast-activating outward currents that were then slowly inactivated during the 500 ms-long voltage steps (Fig. 6C). Current amplitudes and their kinetic properties were comparable between hMGC-derived rod photoreceptors and acutely dissociated adult mouse rods (Fig. 6E) and clearly different from the slowly activating and noninactivating outward currents generated by adult mouse and human MGCs in response to depolarizing voltage steps (Fig. 6G) and from mouse rod bipolar cells (not shown). We did not observe fast inward currents in response to membrane depolarization, which are typical of ganglion and amacrine cells.

Figure 6.

Functional properties of hMGC-derived rod photoreceptors. (A): Reverse transcriptase polymerase chain reaction analysis in hMGC differentiated progeny for HCN1, KCNB1, KCNV2, and KCNJ14 genes coding for the combination of potassium channel subunits expressed by adult rod photoreceptors. Positive and negative controls, using either hRET or only PA6 cells (−), respectively, were included. (B): The picture on the left shows a living hMGC-derived rod photoreceptor during recording with a patch pipette. The picture on the right shows the Syto-13 fluorescence emitted by a condensed nucleus (arrow) belonging to the recorded hMGC-derived rod photoreceptor. All recorded cells with small and condensed nuclei had membrane capacitance, an index of total membrane surface, lower than five pF and similar to adult mouse rods, providing an independent confirmation that nuclear morphology correlates with cell functionality. (C, E, G): Sweep plot currents evoked by either 500-millisecond-long depolarizing voltage steps ranging from −20 to 40 mV in 10-mV steps (C, E) or 2-second-long depolarizing voltage steps from −90 to +30 mV in 20-mV steps (G). The dashed traces plot the 0-current level. (D, F, H): Circles plot average current amplitudes ± SEM as a function of stimulating voltages (I/V curve) for hMGC-derived cells (n = 3) (D), adult mouse rods (n = 3) (F), and mouse MGCs (n = 3) (H). Slopes of I/V curves indicate similar high-membrane resistance (>1 GΩ) for adult mouse rods (F) and hMGC-derived cells (E). The I/V curves cross the voltage axis close to −50 mV in (E, F), indicating similar membrane potential in bright light for both adult rods and hMGC. On the other hand, mouse MGCs have a 10-fold lower membrane resistance and membrane potential close to −90 mV. Scale bar = 10 μm (B). Abbreviations: hMGC, human Müller glia cell; hMC, human Muller glia Culture; hRET, human retina.

Retinal Grafting of Differentiated Human MGCs

However, the in vitro culture conditions did not allow us to follow the subsequent maturation of the induced photoreceptors and their associated morphological changes. Thus, as proof of principle, we set out to transplant differentiated Müller cells into recipient retinas to verify their ability to complete photoreceptor differentiation and integrate into the host tissue. Proliferating hMGCs were transduced with a GFP-expressing lentivirus and both transduction competence and transgene expression were highly efficient.

GFP+ hMGCs were primed using PA6 coculture and bFGF withdrawal (4 days) and supplemented with 1 mM Tau. On priming, cells were transplanted into neonatal immunodeficient Rag2/g(c) double mutant retinas using standardized subretinal injections [31, 32]. Overall, six eyes were injected and the results were analyzed 4 weeks after transplantation. In all the injected eyes, we scored GFP+ cells scattered through the recipient retinas. In many cases, we found cells integrated into the appropriate layer (ONL, Outer Nuclear Layer) and displaying some morphological features of rods but lacking an evident outer segment (Fig. 7). They were immunoreactive for Crx, recoverin, phosducin, and GNAT1, and simultaneously were negative for adult Müller glia markers such as GS or GFAP (Fig. 7). These data indicate that hMGCs survive and integrate in the appropriate retinal layer in the grafted tissue and retain the capability to express rod-specific markers. However, more studies will be necessary to verify whether transplanted hMGCs will complete the photoreceptor differentiation program and therefore acquire functional properties integrated with the host retinal circuits.

Figure 7.

Subretinal human Müller glia cell (hMGC) transplantations in neonatal immunodeficient mice. GFP+(A, C, G, I, M, O) Tau-primed MGC-differentiated cells are detected along the entire subretinally injected retinas, and most of them exhibit loss of glial features and associated molecular markers. A subset of incorporated cells localizes and expresses outer nuclear layer-specific markers such as recoverin (D), Crx (K, Q), GNAT1 (J, P). Single-channel 11-μm-thick confocal images of Hoechst nuclear counter staining (B, H, N), GFP (C, I, O), recoverin (D) and GS (E) or GNAT1 (J, P) and Crx (K, Q) immunostainings were merged in Figures ([A] and [F]) or ([G], [L], [M], and [R]), respectively. Scale bar = 10 μm (A, G, H), 5 μm (B–F, H–L, F, G, I, J). Abbreviations: hMGC, human Muller glia cell; hMC, human Muller glia Culture; hRET, human retina.


In this study, we established the experimental conditions to isolate and expand a homogenous population of human and mouse adult MGCs. In this particular context, MGCs could self-renew for a relatively extensive period providing a number of cells sufficient for several applications. Importantly, the procedure supported MGC culture derivation from both routine surgical samples as well as postmortem retinal tissues. Thus, our results demonstrate for the first time that a large number of primary cultures can be derived starting from human retinal samples, a relatively easy source to be collected. During proliferation, MGCs exhibited the expression of many molecular markers normally associated with an activated state in vivo. Thus, it is likely that in vitro derivation procedures, including tissue disaggregation and culture media conditions, could trigger activation of MGCs bursting their proliferative activity and sustaining their differentiation potential. Indeed, it has been shown in vivo that mouse MGCs might generate neuronal amacrine cells following re-entry into the cell cycle on retinal damage [12]. However, only if potent neurogenic transcription factors are induced by viral transduction in vivo, MGCs acquire a broader neurogenic potential with a substantial higher efficiency [8]. However, in our in vitro conditions, both mouse and human MGCs displayed a significant wider differentiation potential producing various types of retinal neurons as well as rod photoreceptors. Thus, in vitro conditions seem to exalt MGC potential in a way that cannot be achieved in vivo. This phenomenon holds striking similarities with the behavior of brain reactive astrocytes. In fact, these glial cells undergo a process of reactive gliosis on brain injury, which lacks appreciable neurogenic potential in vivo. However, on isolation from the brain parenchima and in vitro culturing, reactive astrocytes exhibit a robust and long-lasting neurogenic differentiation potential [33]. Thus, both reactive glial cells in the brain and retina might be prevented to exert their full potential by inhibitory signals present in the adult tissue.

Despite both mouse and human MGCs exhibiting an intrinsic neurogenic activity when isolated and cultured in vitro, this potential was strongly enhanced by cocolturing with PA6 feeder layers together with molecule supplementation. PA6 are bone-marrow stromal cells originally employed to support in vitro growth and long-term expansion of highly purified hematopoietic stem cells. However, these cells have been widely employed to support neurogenic differentiation of mouse and human embryonic stem ES cells [22, 23]. In fact, PA6 feeder layers offer a rapid method to convert naïve pluripotent cells into a wide range of neuronal phenotypes. Similarly, we report here that PA6 cells sustain neuronal and rod photoreceptor conversion of adult MGCs, indicating a broader role of these cells in promoting neuronal lineage commitment.

Overall, our data remarkably expand the concept of adult MGC plasticity that up to now in rodents has been limited to the acquirement of some retinal interneuron phenotypes exclusively [13]. With respect to retinal degenerative diseases, in vitro generation of rod photoreceptors is particularly relevant, as about 95% of the photoreceptors in humans are of the rod type, which is predominantly lost in individuals with retinitis pigmentosa. Furthermore, both adult mouse and human MGC generate postmitotic photoreceptors, which display those features which have been closely associated with an efficient survival and functional integration after transplantation in vivo [34].

These results clearly suggest MGCs to be likely candidates for exploitation as cell replacement tools in retinal degenerative disorders. Indeed, there is an urgent need to find alternative restorative therapies, such as cell replacement, for those clinical conditions in which photoreceptors have already degenerated, and therefore, gene therapy procedures are no longer feasible. However, as yet there are no available adult somatic retinal cell types that can be isolated, expanded, and efficiently generate photoreceptors. For over a decade, efforts have been focused toward the characterization of a putative population of retinal stem cells derived from the ciliary margin [35, 36]. However, these cells exhibit reduced proliferation ability with only a minor fraction (below 10%) able to differentiate into photoreceptors. Moreover, latest studies provide evidence that these cells retain features of differentiated ciliary pigmented cells even after extensive in vitro culture and lack the potential to differentiate into mature photoreceptor cells [37, 38]. Alternatively, important advances were achieved using human ES or iPS (induced-Pluripotent Stem Cells) for the generation of photoreceptor precursors. In the last 3 years, a series of protocols have been established for this purpose [22, 39–41]. However, all of them contemplate an extremely long differentiation time window associated with a low efficiency in photoreceptor generation, which conflicts with a readily therapeutic application of these cells. Moreover, although iPS cells offer a system which overcome ethical and legal controversies associated with ES cells, both type of pluripotent cells are endowed with a tumorigenic potential whose adverse effects need to be carefully controlled before any possible therapeutic exploitation.


Thus, human Müller glia represents a new source of photoreceptor progenitors that might offer some advantages with respect to cells previously described. In fact, these results indicate that hMGCs represent a local and expandable source of retinal progenitors endowed with excellent potential for photoreceptor cell commitment. Furthermore, hMGCs rapidly differentiated toward rods within a timeframe fourfold to sixfold faster than conventional differentiating pluripotent stem cells. These dynamics further consolidate the evidence, which suggests a similarity between MGCs and retinal progenitors, at least from a developmental point of view. Finally, hMGCs can be easily isolated from adult retinal tissue and lack any tumorigenic potential as far as we have tested. These findings unveil new opportunities for the treatment of retinas that have been injured or affected by degenerative pathologies. Under these circumstances, hMGC could be isolated and cultured for autologous or allogenic cell transplantation purposes. Thus, hMGCs should be considered an attractive cellular source for future retinal regenerative therapies and their possible exploitation warrants future investigation.


We thank Y. Sasai, V. Arshavsky, and A. Swaroop for providing antibodies, T. Matsuda and A. Auricchio for demonstrating subretinal injections and T. Plati and A. Biffi for sharing their Rag2/gamma(c) double mutant mouse colony. M. Codenotti is acknowledged for providing biological samples. We are thankful to G. Cossu and M. Andreazzoli for critical reading of the article and A. Auricchio, E. Strettoi, and R. Galli for helpful discussions. We also thank M. John for the English language revision. This work was supported by grants from IAPB Italian Branch (International Agency for the Prevention of Blindness) and Telethon (GGP07181) to V.B.


The authors indicate no potential conflicts of interest.