Müller glia phagocytose dead photoreceptor cells in a mouse model of retinal degenerative disease

Abstract

Retinitis pigmentosa is a devastating, blinding disorder that affects 1 in 4000 people worldwide. During the progression of the disorder, phagocytic clearance of dead photoreceptor cell bodies has a protective role by preventing additional retinal damage from accumulation of cellular debris. However, the cells responsible for the clearance remain unidentified. Taking advantage of a mouse model of retinitis pigmentosa (Rho^P23H/P23H), we clarified the roles of Müller glia in the phagocytosis of rod photoreceptor cells. During the early stage of retinal degeneration, Müller glial cells participated in the phagocytosis of dying or dead rod photoreceptors throughout the outer nuclear layer. Nearly 50% of Müller glia engaged in phagocytosis. Among the Müller phagosomes, >90% matured into phagolysosomes. Those observations indicated that Müller glial cells are the primary contributor to phagocytosis. In contrast, macrophages migrate to the inner part of the outer nuclear layer during photoreceptor degeneration, participating in the phagocytosis of a limited population of dying or dead photoreceptor cells. In healthy retinas of wild-type mice, Müller glial cells phagocytosed cell bodies of dead rod photoreceptors albeit at a lower frequency. Taken together, the phagocytic function of Müller glia is responsible for retinal homeostasis and reorganization under normal and pathologic conditions.—Sakami, S., Imanishi, Y., Palczewski, K. Müller glia phagocytose dead photoreceptor cells in a mouse model of retinal degenerative disease.

Retinitis pigmentosa (RP) describes a group of highly heterogeneous blinding disorders that lead to severe visual impairment and blindness (1). Human patients and animal models of RP demonstrate the progressive death of rod photoreceptor cells (1, 2). In mammals, rod photoreceptor cells account for most of the photosensitive cells in the retina, constituting ∼95% of the photoreceptors in humans (3) and ∼97% in mice (4). The progressive death of rod photoreceptor cells in RP is eventually followed by the death of neighboring cone photoreceptor cells. To compensate for the gradual degeneration of these photoreceptor cells, a remodeling of retinal structures is necessary, but the identity of the cells facilitating that remodeling process has not been ascertained. An especially critical part of the process is the phagocytic clearance of dead neurons, which otherwise cause additional damage to the retina, likely through inflammatory processes. In understanding the etiology of retinal degenerative disorders, the responsible phagocytic cells need to be identified and characterized. Among the phagocytic cells, neuroglia are proposed to possess phagocytic functions in the retina (5, 6); however, it remains unclear which glial cell types are responsible for the rapid response to retinal damage by phagocytosing dead photoreceptors.

Müller glial cells are a retina-specific glial cell type. Individual Müller glial cells span almost the entire thickness of the retina, including the outer nuclear layer containing photoreceptor cell bodies (7). In contrast, microglia cells, another type of neuroglia, are excluded from the outer nuclear layer of healthy retinas (6). Previous transmission electron microscopy (EM) studies suggested the presence of cellular debris within Müller glial cells, which was considered to have resulted from phagocytic activity. According to those studies, Müller glia contained the pyknotic residues in fetal human retina (8), lipofuscin-like structures in ground squirrel retina (9), and electron-dense cellular debris in developing chicken and quail retinas (10, 11). Thin sheets of Müller glial cell membranes were also demonstrated to surround apoptotic cell bodies in the outer nuclear layer (ONL) in degenerative CBA/Ki mouse retinas (12). From these studies, however, it has been challenging to determine whether electron-dense debris was simply wrapped by the plasma membranes of Müller glial cells or internalized within Müller glial cells. The normally dense packing of Müller glial cells and photoreceptors prohibited such analyses. Moreover, without adequate controls in earlier EM studies, it was impossible to interpret whether electron-dense materials or apoptotic bodies, which were often described as phagosomes, were indeed the result of phagocytosis. Alternatively, the cellular debris could have resulted from autophagy as part of the self-renewal of Müller glial cells (13). Thus, it remains unclear whether Müller glial cells can phagocytose dead neurons during retinal degeneration.

Despite technical limitations, the potential phagocytic activity of Müller glia has been demonstrated during the past 80 yr. After subretinal or intravitreal injection of foreign materials, those materials were subsequently observed within Müller glial cells (14–19). Such studies indicated that Müller glia is capable of engulfing exogenous materials; nevertheless, those results did not provide insights into the in situ clearance of damaged retinal neurons. More-recent studies took advantage of a TUNEL assay and demonstrated that Müller glial cells contained TUNEL⁺ DNA fragments, which were believed to have originated from dead retinal neurons (20, 21). Confounding those studies, however, such DNA fragments could have originated from the Müller glial cells themselves (22). It remains unknown whether dead photoreceptors, generated by retinal degenerative disorders, are engulfed and digested by Müller glia through phagocytosis. Considering the importance of phagocytosis in human RP, the role of Müller glia needs to be established in a relevant mammalian model that is genetically tractable.

Visualization of Müller glial cell–mediated phagocytosis has been challenging because of the rapid clearance of cell bodies (23) and the low steady-state number of dead photoreceptor cells (24) at any given moment. Previously, we generated the P23H rhodopsin-mutation knock-in mouse, which recapitulated the genetic conditions of autosomal-dominant RP (25). In this study, we analyzed the phagocytic clearance of rod photoreceptor cells in P23H knock-in mice. The severity of the visual phenotype is dependent on the dosage of the Rho^P23H alleles; ∼50% of rod photoreceptor cells are lost in the heterozygote P23H (Rho^P23H/+) knock-in mouse retina, whereas ∼90% of rod photoreceptors are lost in the homozygote P23H (Rho^P23H/P23H) knock-in mouse retina by 1 mo of age (25). These observations indicated that phagocytic degradation was extremely active in this model of retinal degeneration. Despite rapid clearance, dead rod photoreceptor cell bodies were observed in the Rho^P23H/P23H mouse retina; therefore, this model presents an opportunity to visualize the phagocytic events of Müller glial cells, which were elusive in the past. To analyze the phagocytic events, we designed a method to study phagosomes and phagolysosomes in individual and dissociated Müller glia from Rho^P23H/P23H mice. These studies were complemented by contemporary EM and fluorescence microscopy of intact retinas, documenting the role of Müller glial cells in phagocytosing photoreceptor neurons as well as the supportive role of macrophages at an early stage of retinal degeneration. Thus, this study clarifies the important role of Müller glia in the initial remodeling process of the retina during the onset of degenerative events.

MATERIALS AND METHODS

Animals

Mouse rhodopsin P23H mutant knock-in mice (Rho^P23H/P23H) were backcrossed to the C57BL/6J background (000664; The Jackson Laboratory, Bar Harbor, ME, USA) >6 times (25). Wild-type (WT; Rho^+/+) C57BL/6J mice were used as controls. Mice were fed a standard diet (Prolab Isopro RMH 3000; LabDiet, St. Louis, MO, USA) and water, and reared under 12/12 h light/dark cycles. For euthanasia and eye tissue collection, mice were exposed to carbon dioxide and decapitated. All animals were raised and maintained at the Animal Resource Center of Case Western Reserve University (Cleveland, OH, USA). Animal procedures and experiments were approved by the Institutional Animal Care and Use Committee of Case Western Reserve University. Animal procedures also conformed to the recommendations of both the American Veterinary Medical Association (Schaumburg, IL, USA) Guidelines for the Euthanasia of Animals and the Association for Research in Vision and Ophthalmology (Rockville, MD, USA).

Microscopy

For ultrastructural EM, we imaged thin-plastic sections (∼80 nm) by transmission EM (1200 EX; JEOL, Tokyo, Japan) as previously described (25). For Brightfield light microscopy, thick-plastic sections (∼400 nm) were stained with 0.5% Toluidine Blue O dissolved in 0.1 M sodium phosphate buffer (pH 7.0). They were imaged with a DM 6000 B upright microscope (Leica Microsystems, Wetzlar, Germany) equipped with an HCX PL apochromat (APO) ×63 oil objective lens [numerical aperture (NA) = 1.40–0.60], a MicroPublisher 5.0 real-time viewing color charge-coupled device (CCD) camera (QImaging, Surrey, BC, Canada), and Image-Pro software (Media Cybernetics, Rockville, MD, USA).

For confocal imaging of flat-mount retinas and a subset of dissociated Müller glial cells, we used a TCS SP5 II confocal microscope (Leica Microsystems) equipped with a Chameleon Vision laser system (Coherent, Santa Clara, CA, USA), a TCS SP2 MP confocal microscope (Leica Microsystems) equipped with a Chameleon XR MP laser (Coherent), and a TCS SP8 confocal microscope (Leica Microsystems). The 2′-[4-ethoxyphenyl]-5-[4-methyl-1-piperazinyl]-2,5′-bi-1H-benzimidazole trihydrochloride trihydrate (Hoechst 33342) was excited with a 760-nm Chameleon infrared laser (Coherent) or a 405 nm laser. Cy3 was excited with a 543-, 552-, or 561-nm laser, Alexa Fluor 488 was excited with a 488-nm laser, and Alexa Fluor 647 was excited with a 633-nm laser. We used an HCX PL APO ×63 oil objective lens (NA = 1.40–0.60) with zoom 1 for analysis of retinal sections (Fig. 2 and Supplemental Fig. S1), zoom 1–2 for analysis of dissociated Müller glial cells (Figs. 4A, C, 5A, B, and 6C), and zoom 5 for analysis of the spherical processes of the macrophages in flat-mount retinas (Fig. 3C). For analysis of dissociated Müller glial cells, labeled with anti-RP1 (Supplemental Fig. S3), we used an HCX PL APO CS2 ×63 oil objective lens (NA = 1.40) with zoom 1. For the statistical analysis of retinal flat mounts (Fig. 3A, B, D, E), we used an HCX PL APO ×40 oil objective lens (NA = 1.3) with zoom 1.

Figure 2.

Phagocytic Müller glial cells are distinct from macrophages. Retinas prepared from PND 14 Rho^P23H/P23H and Rho^+/+ mice were labeled with antibodies detecting the Müller glial cell marker GS and the macrophage marker CD11b. Nuclei were labeled with Hoechst 33342 (Hoechst). A maximum projection image of the Rho^P23H/P23H mouse retina was generated from 47 confocal sections spanning 13.63 µm in the z dimension. A maximum projection image of the Rho^+/+ mouse retina was generated from 40 confocal sections spanning 11.56 µm in the z dimension. In Rho^P23H/P23H mouse retina, thin Müller glial cell profiles contain elongated processes surrounding nuclei (arrows in merged and anti-GS panels). Such immunostaining from anti-GS (green) did not overlap with the macrophage marker anti-CD11b (red). Macrophages either migrated (cross) or extended their processes (arrowheads) into the ONL. In Rho^+/+ mouse retinas, processes from each Müller glial cell appeared as a single, thin profile spanning the ONL. Macrophages were located exclusively at the OPL (asterisks). OS, outer segment of photoreceptor cells. Scale bar, 20 µm.

Figure 4.

Müller glial cells from Rho^P23H/P23H mouse retinas contain dead rod cell bodies. A–D) Fluorescence images of dissociated Müller glial cells from PND 14 Rho^P23H/P23H (A, B) and Rho^+/+ (C, D) mouse retinas. Müller glial cells retaining their intact outer processes (op), inner processes (ip), and end feet (ef) were analyzed. Müller glial cells were labeled with anti–glutamine synthetase (GS, green), and nuclei were counterstained with Hoechst 33342 (blue). A) A Müller glia containing a small pyknotic nucleus in its op. The small pyknotic nucleus (p*) and Müller glial cell nucleus (s) of the same image are shown in the inset. B) A Müller glial cell containing a rod photoreceptor cell body (arrow, r*) at the op. Insets demonstrate the nuclei of a Müller glial cell (s) and a rod photoreceptor cell (r*). C, D) Most Müller glial cells from Rho^+/+ mice did not phagocytose other cell somas, as shown by this representative Müller glial cell. Insets demonstrate the nuclei of the Müller glial cells (s). To unambiguously identify the op, F-actin–enriched microvilli, located at the apex of the outer processes (ap, arrowheads), were labeled with Alexa Fluor 488–conjugated phalloidin (red) in B and D. A, C) Maximum projections of confocal images are shown. B, D) Epifluorescence images are also shown. B, D) To show the details of the entire Müller glial cells, single cells were imaged at 2–3 different focal depths. E) Percentages of Müller glial cells containing dead cell bodies in their outer processes. Rho^+/+ (n = 4) and Rho^P23H/P23H (n = 3) mice were compared at PND 14. Scale bars, 20µm, 4 µm for insets. ***P < 0.001 by Student’s t test.

Figure 5.

Most phagosomes containing rod nuclei mature into phagolysosomes. A, B) Plasma membrane of dissociated and intact Müller glial cells from PND 14 Rho^P23H/P23H mice were labeled with antibodies against EAAT1 (Müller glial cell plasma membrane marker, green) and Lamp1 (red). Nuclei were counterstained with Hoechst 33342 (blue). A) Maximum projection images spanning 8.97 µm thickness are shown at low magnification (a–d) and higher magnification (a’–d’). B) Optical sections of phagolysosomes at different z heights are shown. C) Dissociated Müller glial cell is shown at low magnification (a–d) and high magnification (a′–d′). The cell was labeled with antibodies against GS (green) and Lamp1 (red). Nuclei were counterstained with Hoechst 33342 (blue). The apex of outer processes of Müller glial cells are often closely associated with condensed nuclei of rod photoreceptor cells (r) after mechanical dissection (A, C). op, outer process; ip, inner process; ef, endfeet; r*, spherical nuclei of rod cell bodies localized inside Müller glia; s, soma of Müller glial cell. Scale bars, 20 µm (Ad, Cd), 4 µm (Ad’, B, Cd’).

Figure 6.

Müller glial cells in Rho^+/+ mice phagocytose apoptotic cell bodies at low frequency. A, B) Müller glial cells dissociated from PND 11 Rho^+/+ retinas were labeled with antibodies against the Müller glial cytoplasmic marker GS (green). To visualize F-actin–rich apical microvilli, the cells were labeled with phalloidin (red). The nuclei were counterstained with Hoechst 33342 (blue). Although at low frequency, dissociated Müller glial cells from Rho^+/+ mice retinas contained dead cell bodies at their outer processes (A, p* indicated by arrow) and somas (B, p* indicated by arrows). In addition to engulfed cell bodies, multiple photoreceptor cell bodies also were attached to the outer processes (op) of Müller glial cells (B), but those cell bodies were not engulfed within the cytoplasm. C) A dissociated Müller glial cell from PND 11 Rho^+/+ mouse retina is shown at low magnification (a–e) and high magnification (a′–e′). The cell was labeled with antibodies against the Müller glia plasma membrane marker (EAAT1, green), lysosome marker (Lamp1, red), rod photoreceptor marker (Rhodopsin, magenta), and DNA dye Hoechst 33342 (blue). Maximum-projection images spanning 15.1 µm thickness are shown. A pyknotic cell body was engulfed at the op of a Müller glial cell (arrow, p*). In addition, the op interacted with rhodopsin containing cellular debris (arrowheads). a’–e’) Optical sections of an engulfed pyknotic cell body. A, B), The same Müller glial cells are shown at 2 different focal depths (a, b–e). s, soma of Müller glial cell; M, nucleus of Müller glial cell. Scale bar, 20 µm (Be, Ce) and 4 µm (Ce’).

Figure 3.

The morphology of macrophages in the ONL of Rho^P23H/P23H mouse retinas. A, B) The 3-dimensional projections were constructed from stacks of images covering the outer and middle part (outer/middle) of the ONL, inner part of the ONL, and the OPL. A) In Rho^+/+ mouse retina at PND 14, macrophages were not observed at the ONL and only observed at the OPL (asterisks). B) In Rho^P23H/P23H retinas at PND 14, most macrophages were located proximal to the OPL (asterisks). Macrophages formed spherically shaped processes within the ONL (arrows). Such processes were observed more frequently in the inner part, compared with the outer part, of the ONL. A, B) The 3-dimensional projection images were constructed from stacks of images covering 15 µm thickness. C) A high-magnification image of macrophage processes (red) in the ONL of Rho^P23H/P23H mouse retina at PND 14. These processes engulfed pyknotic (p*) or spherical (r*) rod nuclei labeled with Hoechst 33342 (blue). D) In Rho^P23H/P23H mouse retinas at PND 14, round-shaped processes of macrophages were observed at lower frequency within the outer part, compared with the inner part, of the ONL. E) Average numbers of macrophage somas/100 µm² of the ONL (n = 4 mice for Rho^+/+, n = 3 mice for Rho^P23H/P23H at PND 14 and PND 18. Macrophages were labeled with anti-CD11b. Scale bars, 20 µm (B) and 4 µm (C). ***P < 0.001 by Student’s t test.

Epifluorescence images of Müller glial cells were used for the statistical analysis of dissociated Müller glial cells. Those images were captured with a DMI 6000 B epifluorescence microscope (Leica Microsystems) equipped with an HCX PL APO ×63 oil objective lens (NA = 1.40–0.60) and a Retiga EXi CCD camera (QImaging), and Image-Pro software.

Immunohistochemistry of retinal sections

Mouse eye cups were fixed for 2.5 h at 4°C with gentle agitation in PBS (136 mM NaCl, 2 mM KCl, 8 mM Na₂HPO₄, 1 mM KH₂PO₄, pH 7.4) containing 5% (w/v) sucrose and 4% (w/v) paraformaldehyde. After fixation, eye cups were incubated sequentially in PBS containing 5, 10, 15, and 20% (w/v) sucrose for >30 min each at 4°C. Eye cups were then infiltrated with a 2:1 mixture of PBS containing 20% sucrose and OCT compound (Sakura Finetek, Torrance, CA, USA) at 4°C and then frozen with 2-methylbutane cooled by liquid N₂ (26). Frozen eye cups were stored at −80°C until use. Retinal sections were cut at a thickness of 7 or 14 µm with a CM1850 cryostat microtome (Leica Microsystems) and stored at −80°C until use. Retinal sections were air dried at 37°C and then rehydrated and blocked with PBS containing 5% (v/v) goat serum, 0.3% (v/v) Triton X-100, and 0.02% (w/v) sodium azide. After blocking, sections were incubated with primary antibodies diluted in PBS containing 0.3% Triton X-100 and 0.02% sodium azide overnight at room temperature with gentle agitation. Rabbit anti-glutamine synthetase IgG (6.5 mg/ml; MilliporeSigma, Burlington, MA, USA) was diluted at 1:1500–3000, and biotin-conjugated rat anti-CD11b IgG (0.5 mg/ml; BD Biosciences, San Jose, CA, USA) was diluted at 1:20–500. After incubation, sections were washed with PBS and then incubated with secondary antibodies and Hoechst 33342 diluted in PBS containing 0.3% Triton X-100 and 0.02% sodium azide for 2 h at room temperature with gentle agitation. Concentrations of secondary antibodies and dye were as follows; Alexa Fluor 488–conjugated goat anti-rabbit IgG (2 mg/ml; Thermo Fisher Scientific, Waltham, MA, USA) at 1:500 dilution, Cy3-conjugated goat anti-rat IgG (1.5 mg/ml; Jackson ImmunoResearch Laboratories, West Grove, PA, USA) at 1:330 dilution, and Hoechst 33342 (10 mg/ml; Thermo Fisher Scientific) at 1:813 dilution. After incubation, sections were washed with PBS and then mounted with Vectashield Mounting Medium (Vector Laboratories, Burlingame, CA, USA) and covered with No. 1.5 cover glass. To encompass the entire thickness of a sample, optical sections were collected every 0.3 µm with a Leica TCS SP5 II confocal microscope, and maximum projection images were analyzed (Fig. 2).

Immunohistochemistry of flat-mounted retinas

For labeling macrophages in flat-mounted retinas, we followed the protocol described by Noailles et al. (27), with modifications. To prepare eye cups, the cornea and lens were removed from the eyes in PBS. Eye cups then were flattened by making 4 incisions. To track dorsoventral and temporonasal directions, retinas were labeled with tissue-marking dye (Electron Microscopy Sciences, Hatfield, PA, USA) and then detached from the retinal pigment epithelium (RPE) in PBS. Detached retinas were transferred onto culture inserts with a pore size of 0.4 µm (PICM03050; MilliporeSigma) with their ganglion cell–side up. Retinas were fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (0.014 M NaH₂PO and 0.086 M Na₂HPO₄, pH 7.5) for 50 min at room temperature with gentle agitation. Retinas were then washed with PBS containing 5% sucrose and stored in PBS containing 20% sucrose at 4°C to increase tissue permeability toward antibodies. Retinas were processed and labeled as described by Noailles et al. (27), with the following dilutions of antibodies and Hoechst 33342; biotin-conjugated rat anti-CD11b IgG at 1:500 dilution, Cy3-conjugated goat anti-rat IgG at 1:1500 dilution, and Hoechst 33342 at 1:1500 dilution. Antibodies and dye were diluted in PBS containing 0.5% (v/v) Triton X-100 and 0.02% sodium azide. After labeling, retinas were washed with the same buffer. Subsequently, retinas were fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.5) for 5 min at room temperature and then washed with PBS. Retinas were mounted with Vectashield Mounting Medium and No. 1.5 cover glass. Images were captured with a Leica TCS SP5 II confocal microscope. To capture a large field, we used an HCX PL APO ×40 oil objective lens (NA = 1.3) with zoom 1 (Fig. 3A, B). To analyze details of the spherical processes, the ONL was imaged with an HCX PL APO ×63 oil objective lens (NA = 0.60–1.40) with zoom 5 (Fig. 3C).

Four mice were analyzed for each genotype (Rho^P23H/P23H and Rho^+/+); >3 regions (dorsal, ventral, nasal, and temporal)/retina were analyzed for each mouse. In Rho^P23H/P23H mice, the entire thickness of the ONL and outer plexiform layer (OPL) encompassed 3–4 stacks of images. Each stack covered a thickness of 15 µm, with optical sections collected every 0.3 µm. Spherical processes of macrophages were counted (Fig. 3D) using top views of 3-dimensional volume images generated from a stack of images (369 × 369 × 15 µm) using LAS X software (Leica Microsystems) (Fig. 3B). In Fig. 3C, a maximum projection image of macrophages was generated from a 28-µm stack of images using LAS X software and merged with an optical section of nuclei labeled with Hoechst 33342. For analyses of Rho^P23H/P23H mouse retinas in Fig. 3E, macrophages were counted with maximum projection images generated from stacks of images (369 × 369 × 15 µm). For analysis of Rho^+/+ mouse retinas, entire thicknesses of the ONL were visually inspected, and the absence of macrophages in the ONL was confirmed. Entire thicknesses of the OPLs were imaged as previously described.

Dissociation of Müller glial cells

For the dissociation of Müller glial cells from mouse retina, we modified the protocol designed for the rat retina (28). Retinas were collected from mouse eyes in ice-cold HBSS (Thermo Fisher Scientific). Retinas were then incubated with ∼3 U/ml of papain (MilliporeSigma) and 2.5 mM cysteine in modified Locke’s solution A [3.6 mM KCl, 112.5 mM NaCl, 20 mM NaHCO₃, 10 mM N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES), 0.02 mM disodium EDTA, 2.4 mM MgCl₂, 30 mM d-(+)-glucose, and 2.5 mM EGTA, pH 7.3] for 20 min at 37°C. All modified Locke’s solutions used for this procedure were equilibrated with 95% O₂ and 5% CO₂ gas. Papain was then inactivated by incubating the retinas with 10% goat serum in modified Locke’s solution B (modified Locke’s solution A without the EGTA and with 1.2 mM CaCl₂, pH 7.5). After the inactivation, DNase I (100 U/ml) and 0.1% (w/v) bovine serum albumin (BSA) were added to the same solution, and the incubation was continued at 37°C for 2 min. After washing, retinas were transferred into modified Locke’s solution B containing 2 mM glutamine (MilliporeSigma), 1× minimum essential medium amino acids (MilliporeSigma) and 1× minimum essential medium vitamins (MilliporeSigma). In that solution, Müller glial cells were mechanically dissociated by gently pipetting the retinas repetitively using a siliconized P1000 pipette tip (Thomas Scientific, Swedesboro, NJ, USA). After dissociation, cells were diluted with the same medium and transferred into glass-bottom dishes (MatTek, Ashland, MA, USA), which were pretreated with 0.1 mg/ml poly-d-lysine hydrobromide (MW > 300,000; MilliporeSigma) and 0.1 mg/ml Phaseolus vulgaris erythroagglutinin (Vector Laboratories). To allow Müller glial cells to descend and then attach to the glass, cells were incubated at 16°C for ∼1 h with 95% O₂ and 5% CO₂. Cells were then immediately fixed with 4% paraformaldehyde in modified Locke’s solution B for 5 min at room temperature and then washed with PBS containing 0.5% (w/v) BSA. Fixed cells were stored in PBS containing 0.5% BSA and 0.02% sodium azide (PBSB) at 4°C until use.

Immunohistochemistry of dissociated Müller glial cells

Dissociated Müller glial cells were blocked with ice-cold PBSB containing 5% (v/v) goat serum and 0.1% (w/v) saponin for 20 min and were then incubated with primary antibodies diluted with PBSB containing 0.1% saponin (PBSBS) for 1 h at 4°C. Primary antibodies were used at the following dilutions: rabbit anti–glutamine synthetase (anti-GS) IgG at 1:3000 and rat 1D4B anti-lysosome-associated membrane glycoprotein 1 (Lamp1) IgG2a supernatant (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, USA) at 1:50. After washing cells with ice-cold PBSBS 4 times for 5 min each, cells were incubated with secondary antibodies and Hoechst 33342 at 4°C for 1 h. Secondary antibodies and Hoechst 33342 were at the following dilutions: Alexa Fluor 488–conjugated goat anti-rabbit IgG at 1:500, Cy3-conjugated goat anti-rabbit IgG (1.5 mg/ml; Jackson ImmunoResearch Laboratories) at 1:330, Cy3-conjugated goat anti-rat IgG at 1:330, and Hoechst 33342 at 1:813. After washing the cells with PBSBS, they were incubated with 0.2 U/ml of Alexa Fluor 488–conjugated phalloidin (Thermo Fisher Scientific) in PBSB for 20 min at 4°C to label the apical microvilli of the Müller glia (29). For statistical analysis, Müller glial cells were visually inspected and imaged with a Leica DMI 6000 B microscope. To analyze whole structures of Müller glial cells, each Müller glial cell was imaged at different z heights. For the statistical assessment of phagocytic Müller glial cells in Fig. 4E, we analyzed only those Müller glial cells that retained their intact outer processes with phalloidin-positive apical microvilli (representative images in Fig. 4B, D) and determined whether those outer processes contained dead cell bodies. Müller glial cells dissociated from postnatal day (PND) 11 Rho^+/+ mice (n = 3 mice), PND 14 Rho^+/+ mice (n = 4 mice), and PND 14 Rho^P23H/P23H mice (n = 3 mice) were analyzed. For the statistical analysis of phagosomes and phagolysosomes in Rho^P23H/P23H mice (PND 14), we analyzed dead cell bodies located within the outer processes of dissociated Müller glial cells (n = 3 mice).

To obtain brighter fluorescence signals for detailed structural analysis with confocal microscopy (Figs. 4A, C, 5A, B, and 6C and Supplemental Fig. S3), cells were labeled as previously described, with the following modifications. Cells were incubated with primary antibodies for 3 h, washed with PBS containing 0.1% saponin and 0.02% sodium azide (PBSS), incubated with secondary antibodies for 2 h, and then washed with PBSS. Samples were fixed with 4% paraformaldehyde in modified Locke’s solution B at room temperature and then washed with PBS containing 0.02% sodium azide. Antibodies and Hoechst 33342 were diluted in PBSS and used at the following final dilutions: rabbit anti-GS IgG at 1:3000, rat 1D4B anti-Lamp1 IgG2a supernatant at 1:30, rabbit anti-EAAT1 IgG (Abcam, Cambridge, United Kingdom) at 1:200, mouse anti-rhodopsin (1D4 IgG, 5 mg/ml) at 1:5000, chicken anti-RP1 (EAP15) (30) at 1:1600, Hoechst 33342 at 1:542, Alexa Fluor 488–conjugated goat anti-rabbit IgG at 1:333, Alexa Fluor 647–conjugated goat anti-mouse IgG (2 mg/ml; Thermo Fisher Scientific) at 1:333, Cy3-conjugated goat anti-rat IgG at 1:220, and Cy3-conjugated goat anti-chicken IgY (1.5 mg/ml; Jackson ImmunoResearch Laboratories) at 1:220. As negative controls, cells were incubated with secondary antibodies only or incubated with normal IgGs. Maximum projection images prepared for this manuscript were generated from stacks of images collected with a Leica TCS SP2 MP, TCS SP5 II, or TCS SP8 confocal microscope. Thicknesses of Müller glial cells varied, and thus, thicknesses of image stacks were adjusted accordingly; the thicknesses were 8.4 µm in Fig. 4A; 13.96 µm in Fig. 4C; 8.9 µm in Fig. 5A; 15.1 µm in Fig. 6C; 10.45 µm in Supplemental Fig. S3A, top panel; 7.76 µm in Supplemental Fig. S3A, bottom panel; 11.34 µm in Supplemental Fig. S3B; and 12 µm in Supplemental Fig. S3C. To preserve the structures, Müller glial cells were dissociated with gentle mechanical stress. Thus, Müller glial cells were often weakly attached to rod photoreceptor cells that were not phagocytosed (Figs. 5 and 6B, C, and Supplemental Fig. S3A, B).

Image analysis

For EM images (Fig. 1B, C) and some of the fluorescence images presented in this article (Fig. 4A, inset, D and Figs. 6Ab, e, Ba, b, e, Ca’–e’, optical section images in Supplemental Fig. S3, and Hoechst images in Fig. 4C and Supplemental Fig. S3A–C), contrast and brightness were adjusted with Photoshop CC software (Adobe Systems, San Jose, CA, USA). Only linear adjustments were applied to the entire images. Several fluorescence microscopy images (Figs. 4B, D, 5C, and 6A, B) were subjected to deconvolution with Velocity software (PerkinElmer, Waltham, MA, USA) using point-spread functions, which were estimated based on the imaging conditions.

Figure 1.

Dead photoreceptor cell bodies are located within elongated cells. A) Light microscopy images of retinas derived from Rho^P23H/P23H, Rho^P23H/+ and Rho^+/+ mice at PND 14. Dead photoreceptor cell bodies (arrows) were located inside elongated structures (encircled by dashed red lines) in the Rho^P23H/P23H mouse retina. Dead cell bodies were rarely observed in Rho^P23H/+ or Rho^+/+ mouse retinas. Os, outer segment of the photoreceptor cell. B) Transmission EM images of retinas derived from Rho^P23H/P23H (a–d), Rho^P23H/+ (e), and Rho^+/+ (f) mice at PND 14. a–e) Dead cell bodies were spherical and electron dense. Those dead cells retained their nuclear envelopes (surrounded by blue lines) and were located inside phagocytes (surrounded by red lines) with dense glycogen granules (arrows) or sparse glycogen granules (arrowheads). f) In most Rho^+/+ mouse retinal sections, dead photoreceptor cell bodies were not observed in the ONL. Panels a and c are higher-magnification images of b and d, respectively. C) Transmission EM images of retinas derived from Rho^P23H/P23H mice at PND 14. a) Elongated disc membranes (asterisk), inner segment (white arrowhead), and ciliary protrusion (white arrows) were surrounded by RPE microvilli. b) Within the RPE cell, a membranous structure with the characteristics of a rod inner segment (arrowhead) was observed. The structure contained abundant and condensed mitochondria as observed in the inner segment ellipsoid. A ciliary protrusion (cp) connected to an inner segment is observed in the middle of the image. Scale bars, 20 µm (A), 4 µm (Bb–d, f, Cb), 500 nm (Ba, e, Ca).

Statistics

Data in the bar graphs are expressed as means ± sd. sd was calculated based on ≥3 replicates, as stated in the results or in the figure legends. P values were calculated by the Student’s t test (2-tailed). P values for the analyses were as follows: P = 4.66 × 10⁻¹² (Fig. 3D), P = 7.98 × 10⁻⁵ (Fig. 3E, left), P = 2.81 × 10⁻¹⁴ (Fig. 3E, right), P = 3.78 × 10⁻⁵ (Fig. 4E), and P = 0.0166 (Supplemental Fig. S1).

RESULTS

Initial retinal remodeling occurs through clearance of rod photoreceptor cells

The P23H mutation of the rhodopsin gene is the major cause of autosomal-dominant RP in humans. In P23H mutant rhodopsin knock-in homozygote mice (Rho^P23H/P23H), rod photoreceptor cells degenerate rapidly (Fig. 1A) (25, 31, 32). Despite the rapid degenerative process, dead rod photoreceptor cell bodies do not accumulate in the retina; rather, they are quickly cleared from the ONL (Fig. 1A) (25, 31, 32). Such rapid clearance of dead photoreceptor cells is a common feature and is the initial phase of remodeling, which occurs in photoreceptor-degenerative disorders, such as RP and Leber congenital amaurosis (23, 24).

To study the initial remodeling of the retina because of photoreceptor degeneration, we compared the retinal structures of Rho^P23H/P23H, Rho^P23H/+, and Rho^+/+ mice by light microscopy and EM (Fig. 1). In Rho^P23H/P23H mice, rod photoreceptor cells degenerate because of the effect of mutant P23H rhodopsin, which is prone to protein misfolding (32–35). The loss of rod photoreceptor cells is most significant between PND10 and PND15. The thickness of the ONL in Rho^P23H/P23H mice is indistinguishable from Rho^+/+ mice up to PND 10 but then is dramatically reduced to ∼50–60% of Rho^+/+ mice by PND 15 (32). The thickness of the ONL in Rho^P23H/P23H mice continues to diminish and is less than one-quarter of that of age-matched Rho^+/+ mice at PND 20 (32). Thus, the retinas in this study were analyzed at PND 14, when rod photoreceptors degenerate rapidly in Rho^P23H/P23H mice (25, 31). At that time, the number of condensed rod nuclei that had lost unique chromocenters (36) was significantly greater in Rho^P23H/P23H than it was in Rho^P23H/+ and Rho^+/+ retinas (Fig. 1A). These condensed nuclei were spherical (Fig. 1A, arrows) and were often surrounded by elongated structures spanning the ONL (Fig. 1A, dashed red line). In contrast, the somas of healthy rod cells were adjacent to each other and tightly packed. Two cell types are known to be elongated and span the entire ONL. One type is the Müller glia that radially transects the entire retina, whereas the other is the photoreceptor cell, which spans the ONL.

By transmission EM (Fig. 1B), we found that the elongated structures spanning the ONL were not associated with rod photoreceptors; rather, they were derived from cells engulfing dead photoreceptor cell bodies. A subpopulation of the dead cell bodies in the ONL was engulfed within the inflated cytosol of other cells from phagocytic events (Fig. 1Ba–e). The plasma membranes of dead cells (Fig. 1Ba, c, e; green lines) were distinct from those of the engulfing cells (Fig. 1Ba, c, e; red lines) and enclosed their own nuclei. Typically, those dead cells contained round-shaped nuclei, which had lost their original chromatin structure and had become pyknotic but were still surrounded by nuclear envelopes (Fig. 1Ba, c, e; blue lines). Sometimes, the dead cells retained organelles (Fig. 1Ba) within their cytoplasmic areas, demarcated by individual plasma membranes (Fig. 1Ba; green line). In contrast to the contents of the dead cells, the phagocytotic cells (Fig. 1Ba, c, e; arrows) contained expanded cytoplasmic areas along with a variety of organelles, including smooth endoplasmic reticulum, mitochondria, and small electron-dense aggregates with the appearance of glycogen granules (Fig. 1Ba, c, e). Müller glial cells are known to contain abundant glycogen granules that are not observed in photoreceptor cells (37) and are observed far less frequently in other phagocytic cells, such as macrophages (38, 39). Thus, the elongated cells with numerous glycogen granules (Fig. 1Ba, c, e; arrows) were likely Müller glial cells engaged in phagocytosis. We also observed elongated phagocytes with a smaller number of glycogen granules (Fig. 1Bc; arrowhead). That type of elongated cell was most likely a phagocytic macrophage (discussed in the next section). In Rho^P23H/+ mouse retinas, rod photoreceptor cells degenerate slowly, and thus, it was more challenging to find pyknotic cells (Fig. 1A). Although rare, we were able to observe pyknotic cells in Rho^P23H/+ retinas in the cytoplasmic areas of cells containing a variety of organelles and glycogen granules (Fig. 1Be). As a control, we also analyzed PND 14 Rho^+/+ mouse retinas by EM (Fig. 1Bf). No dead cell bodies were detected in the cytosol of other cells. Rather, photoreceptor cell bodies were densely packed, forming columns. Between the columns of rod cell bodies, Müller glial cells containing numerous glycogen granules passed through the ONL (Fig. 1Bf; red line). Thus, these studies provide ultrastructural evidence that Müller glial cells are the principal cells, whereas macrophages are the secondary cells, engulfing dead rod cells during the onset of RP.

In Rho^P23H/P23H mice, rod photoreceptor cells form ciliary protrusions but do not form outer segments efficiently (Fig. 1A, C) (31). Thus, rod inner segments are located adjacent to the RPE microvilli. We found disc membranes and ciliary protrusions that were surrounded by RPE microvilli (Fig. 1Ca; asterisk and arrows). We also observed inner segments surrounded (Fig. 1Ca; arrowhead) and internalized within the RPE (Fig. 1Cb; arrowhead). Taken together, these studies indicate that phagocytosis was induced by photoreceptor degeneration and that phagocytosis is an active event in Rho^P23H/P23H mice at PND 14.

Müller glia and macrophages have complementary roles in the phagocytosis of dead photoreceptors

To gain further evidence of phagocytosis by Müller glia, we conducted immunofluorescence studies. Retinal cryosections were labeled with an antibody that is specific for GS (Fig. 2, green) and is expressed in Müller glia, RPE cells, and astrocytes in the retina (40). Because RPE cells or astrocytes are absent from the photoreceptor layer and demonstrate distinct morphologies, Müller glial cells can be readily distinguished from those other cell types. Hoechst 33342, a fluorescent DNA dye (Fig. 2, blue), was used to morphologically identify pyknotic nuclei characteristic of end-stage apoptotic or necrotic cells. In Rho^P23H/P23H mouse retinas at PND 14, outer processes of Müller glial cells often formed spherical structures, surrounding photoreceptor cell bodies in the ONL (Fig. 2, anti-GS, arrows). In WT (Rho^+/+) mouse retinas at PND 14, most Müller glial cells appeared as single processes traversing the ONL and did not branch or surround photoreceptor cell bodies (Fig. 2, anti-GS). Those data suggest that, under pathologic conditions, Müller glial cells change their shape in the ONL and engulf dead photoreceptor cell bodies.

In the retina, microglia cells are considered the major phagocytic cell type (6). To understand their contribution to clearing dead photoreceptors, microglia and infiltrating macrophages were labeled with an antibody specific for CD11b (Fig. 2, red). CD11b⁺ microglia and other macrophages are collectively called macrophages in this study. In Rho^P23H/P23H mouse retinas, most macrophage somas resided in the OPL, where photoreceptor synapses are located (Fig. 2, anti-CD11b, asterisk). Their processes extend to the ONL and form round structures (Fig. 2, arrowheads). Some macrophages also migrated into the ONL (Fig. 2, cross). In control experiments, macrophages of Rho^+/+ mice were located solely in the OPL and did not invade the ONL (Fig. 2, asterisks). Thus, these experiments indicated that Müller glial cells and macrophages can phagocytose different subsets of dead rod photoreceptor cells, and such phagocytic activities are stimulated by photoreceptor degeneration induced by the P23H mutant rhodopsin. To understand the relative contributions of Müller glia and macrophages, round processes wrapping rod cell bodies were counted, and their average numbers were compared (Supplemental Fig. S1). Based on that analysis, round processes derived from Müller glia were about 4-fold more numerous than those derived from macrophages in Rho^P23H/P23H mice. Round processes were barely observed in the ONL of Rho^+/+ mice. Taken together, Müller glia has a major role in the phagocytosis of dead photoreceptors during the early stage of retinal degeneration (PND 14).

The phagocytic activity of macrophages is spatially limited at the initial stage of retinal degeneration

In the healthy retina, microglia cells are located in the plexiform layers, where synapses are formed and maintained (6). In pathologic retinas, microglia cells become activated and migrate or extend their processes into degenerating regions (Fig. 2). The processes of activated microglia extend in multiple directions, and their shapes are hard to visualize in vertical sections of the retina. To study the relationship between these cell processes and dead cell bodies, we analyzed the localization of macrophages, including microglia in retinal flat mounts (Fig. 3). In Rho^+/+ mouse retinas, macrophages labeled with anti-CD11b were not observed within the ONL (Fig. 3A). Within the OPL, macrophages extended their processes horizontally (Fig. 3A, asterisks). In Rho^P23H/P23H mouse retinas at PND 14, most macrophages were located within the OPL (Fig. 3B, right panel, asterisks). However, a subpopulation of macrophages migrated or extended their processes into the ONL (Fig. 3B, middle panel, asterisks and arrows). Those ONL-localized macrophages had spherical processes (Fig. 3B, arrows) and often contained multiple dead cell bodies (Fig. 3B, C). The sizes of the spherical processes were consistent with the wrapping of somas of dead photoreceptor cells. Detailed colocalization analysis confirmed that round-shaped processes of macrophages (red) contained either pyknotic nuclei (Fig. 3C, p*) or spherically shaped nuclei (Fig. 3C, r*) of rod photoreceptor cells. Those observations indicated that macrophages were actively engaged in the phagocytosis of dead photoreceptor cell bodies within the ONL.

Despite their involvement, the role of macrophages is rather limited at that early stage (PND 14) of retinal degeneration in Rho^P23H/P23H mice. Macrophages and their processes were more pronounced in the inner part of the ONL (Fig. 3B, D), suggesting that their phagocytosis is restricted to that region. In Rho^+/+ mouse retinas (PND 14, n = 4 mice), macrophages were not observed in the ONL but were observed in the OPL (Fig. 3A, E). In Rho^P23H/P23H mouse retinas, the number of macrophages increased ∼3-fold within the ONL (n = 3 mice, P < 0.0001) (Fig. 3E), and ∼1.5-fold (P < 0.01) within the OPL from PND 14 to 18. Those results suggested that the macrophages progressively invaded the ONL from PND 14 to 18, and invasion occurred from the OPL side.

Müller glia dissociated from Rho^P23H/P23H mouse retinas contain dead rod photoreceptor cell bodies at their outer processes

The same flat-mounted retinas described above for the analysis of macrophages were not appropriate for measuring Müller glia phagocytosis because of the dense packing of the Müller glia. To circumvent that issue, we designed a method to analyze isolated Müller glial cells (Supplemental Fig. S2). In dissociated cell preparations of Rho^P23H/P23H mouse retinas, nearly 50% of Müller glial cells contained pyknotic or spherically shaped cell bodies of dead rod photoreceptor cells at their outer processes (Fig. 4). Pyknotic and spherical rod nuclei (Fig. 4A, B, p* and r*) were clearly distinguishable from Müller glial nuclei, which were occupied with less-condensed euchromatin and had several small, condensed chromocenters (Fig. 4A, B, s). As a control, <1% of Müller glial cells dissociated from Rho^+/+ mouse retinas contained dead cell bodies of rods, either pyknotic or spherically shaped, at their outer processes (Fig. 4C–E). During phagosome maturation, lysosomes fuse with phagosomes and form phagolysosomes. The formation of phagolysosomes indicates active degradation of phagocytosed materials. Thus, to confirm the phagocytic activity of Müller glia, we next analyzed phagolysosome formation in Müller glial cells by labeling phagolysosomes with anti-Lamp1, a marker of lysosome-associated membrane glycoprotein 1 (Fig. 5). The plasma membrane and the cytoplasmic area of Müller glial cells were labeled by anti-EAAT1 (Fig. 5A, B) and anti-GS1 (Fig. 5C), respectively. Confocal microscopy demonstrated that pyknotic (Fig. 5A, p*) and spherical (Fig. 5A, r*) nuclei of rod cell bodies localized inside the plasma membrane of Müller glial cells (Fig. 5A, B, green). We further confirmed that spherical rod nuclei (Fig. 5C, r*) were within the cytoplasm of Müller glial cells (Fig. 5C, green). Moreover, those rod nuclei were wrapped by Lamp1⁺ phagolysosomal membranes (Fig. 5A–C, red). Those data strongly suggested that the dead cell bodies found in the ONL were internalized in Müller glial cells of the Rho^P23H/P23H retina. Within the cytoplasmic area of Müller glial cells, most (90.7 ± 1.8%, n = 3 animals) of the engulfed nuclei were observed in phagolysosomes, indicating phagosomes mature quickly into phagolysosomes. Cell bodies located inside these phagolysosomes were often pyknotic and weakly labeled with Hoechst 33342, suggesting that the rod-derived DNA was digested. Those data demonstrated that dead photoreceptor cell bodies were phagocytosed and degraded by Müller glial cells in the Rho^P23H/P23H mouse retina.

Those observations indicate that the lower portion of rod photoreceptor cells, containing the cell body, was phagocytosed by Müller glial cells of Rho^P23H/P23H mice. We further investigated whether sensory cilia, which are located at the upper portion of rod photoreceptor cells, are also phagocytosed by Müller glial cells. Both in Rho^+/+ and Rho^P23H/P23H mice, RP1 immunopositive sensory cilia were observed outside the Müller glia (Supplemental Fig. S3, red). However, we did not observe significant RP1 immunoreactivity within the Müller glia (Supplemental Fig. S3). Those results suggested that Müller glia were not the major contributor for clearing photoreceptor sensory cilia. Overall, our immunofluorescence results (Supplemental Fig. S3) are consistent with the EM analysis (Fig. 1C), indicating that the upper portions of rod photoreceptor cells, cilia and inner segments, are phagocytosed by the RPE cells.

Müller glia phagocytose dead cells in WT retina at low frequency

In addition to photoreceptor cell death because of RP-causative gene mutations, programmed cell death is a naturally occurring event during normal development as well as during the maintenance of the retina (41). In Rho^+/+ mice, developmental cell death continues to be observed in the ONL until ∼3 wk after birth (42, 43). Thus, we determined whether Müller glia could also contribute to phagocytosis in healthy retinas of Rho^+/+ mice. Müller glial cells dissociated from Rho^+/+ mouse retinas engulfed condensed nuclei and cell bodies of other cells (Fig. 6) albeit at low frequencies (1.39% at PND 11 and 0.26% at PND 14). Those Müller glial cells also interacted with materials derived from rod photoreceptor cells, including rod cell bodies (Fig. 6B) and rhodopsin-containing debris (Fig. 6C); however, those materials did not localize inside the Müller glial cells. The plasma membrane of Müller glial cells is known to contact the photoreceptor cell bodies and inner segments in healthy retinas (44, 45). Thus, our results confirm that the phagocytosed cell bodies observed in our studies are distinct from those cells attached as part of normal cell–cell contacts in the retina. We observed condensed nuclei both in the outer processes (Fig. 6A, p*) and somas (Fig. 6B, p*). In those cells, nuclei of Müller glial cells, which have less-condensed euchromatin structures (Fig. 6, M), were also detected. Thus, those additional condensed nuclei did not originate from the Müller glial cells themselves, but they were engulfed by Müller glial cells through phagocytosis. The outer processes and somas of Müller glial cells are located in the outer and inner nuclear layer, respectively (44). Therefore, these results suggest that Müller glial cells are involved in the phagocytosis of both photoreceptor cells and secondary neurons. As further evidence for the phagocytosis of rod photoreceptors, a subpopulation of the engulfed cell bodies (Fig. 6C, p*) weakly stained with anti-rhodopsin mAb 1D4 (Fig. 6C a–e, for low magnification and Fig. 6Ca’–e’, for high magnification). Our results are consistent with the previous observation (42) that programmed cell death is occurring at low frequency in WT retina during postnatal development. Because the invasion of macrophages is not observed in the ONL of healthy retinas (Figs. 2, 3), Müller glial cells likely have a primary role in the clearance of the dead cell bodies of photoreceptors and other cell types during the development and maintenance of the neural retina.

DISCUSSION

Our studies demonstrate that Müller glia phagocytose dead rod photoreceptor cells in mice harboring an RP-causative mutation. Müller glia are in proximity to all cell types in the retina (44) and thus can act as fast responders to ameliorate retinal damage, which otherwise can be exacerbated by the accumulation of dead cells. In RP, those dead cells are cleared, thus thinning of the ONL is observed because of the loss of photoreceptor cells, which is a prominent feature of this blinding disorder. Unlike Müller glia, macrophages are not usually found in the ONL and are thus unable to quickly respond to degenerative stress. At the early stage of retinal degeneration in PND 14 Rho^P23H/P23H mice, macrophages slowly migrated and populated primarily the inner part of the ONL. The contribution of macrophages to the phagocytic activity, therefore, is limited to the inner part of the ONL. Our observation is in contrast to the behavior of microglia and other professional phagocytes in the light-induced photoreceptor degeneration of the albino mouse model, in which microglia and other professional phagocytes quickly spread throughout the photoreceptor cell layers within 24 h of light exposure (46). Unlike the Rho^P23H/P23H model, rod degeneration occurs through excessive activation of the phototransduction cascade, and thus, the initial degenerative stress is localized to the outer segments and later spreads throughout the entire photoreceptor structure in the light-induced model. Thus, degenerating photoreceptors may attract professional phagocytes from both the vitreal and scleral sides, in addition to residential microglia of the retina. In the Rho^P23H/P23H mouse model, rod outer segments form poorly, and thus, cellular stresses are localized more prominently to the inner regions of rod photoreceptors, attracting and inducing migration of macrophages from the OPL. Müller glia, which reside in the ONL, do not require cell migration and appear to contribute to phagocytosis throughout the ONL. The outer processes of Müller glial cells, which span the entire ONL, are capable of quickly wrapping and phagocytosing dead photoreceptor cells, which is especially important during the initial stage of retinal degeneration. Müller glial cells also are the most abundant glial cell type in the retina (45). Based on our quantitative analysis, Müller glia are 4 times more active than macrophages are in engulfing dead cell bodies in Rho^P23H/P23H mice at PND 14. Nearly 50% of Müller glial cells are engaged in phagocytosis when retinal degeneration is underway, contributing significantly to the clearance of dead cell bodies.

Previous studies examined Müller glia phagocytosis of foreign materials, such as latex beads or carbon particles that were intravitreally injected (17, 18). Müller glia are also capable of phagocytosing biologic materials, such as melanin granules and erythrocytes, injected into the vitreous or subretinal space (47, 48). Nevertheless, the use of foreign substances may confound our understanding of the in vivo clearance of cellular debris. Unlike biologically relevant phagocytotic clearance, phagocytosed melanin granules and other foreign substances are not degraded and instead reside in Müller glia for an extended period (14, 15, 17, 18). Cultured Müller glial cells of gold fish retina are capable of phagocytosing latex beads ex vivo; however, those Müller glia are incapable of phagocytosing those materials in vivo (49). Other studies frequently employed a TUNEL assay to document the phagocytosis of dead neurons by Müller glial cells (5). In degenerating retinas of mammals and teleost fish (20, 21, 50), more Müller glial cells became TUNEL⁺ within their cytoplasmic regions. These results were interpreted as a sign of phagocytosis of dead cells by Müller glia, resulting in the transfer of TUNEL⁺ DNA fragments to the cytosol of Müller glial cells. Nevertheless, it is unclear how the phagocytosed materials, which are within endocytic compartments, would leak into the cytosol. Under similar experimental conditions, an inhibitor of phagocytosis, O-phospho-l-Ser, did not change the number of those TUNEL⁺ cells (21), suggesting that TUNEL⁺ DNA fragments were incorporated into Müller glial cells independent of phagocytic events. Recently, documented cytoplasmic-fusion mechanisms may also explain such transfer of DNA fragments, independent of phagocytosis (51). Moreover, past studies suggested that TUNEL⁺ DNA fragments might have originated from the Müller glial cells undergoing degeneration (22). In contrast to those indirect studies, we verified, by direct visualization, the internalization of dead cell bodies of rod photoreceptors as phagosomes and phagolysosomes in dissociated and intact Müller glia.

Previous studies also reported the possible contributions of Müller glial cells to the phagocytic activity in various vertebrate species, including human (8, 10, 12, 52), but those earlier studies failed to clearly demonstrate the phagocytosis of dead neurons. Earlier EM studies documented the presence of cellular debris within Müller glial cells (8, 10, 11). However, it was difficult to determine the origin of the debris. It was also challenging to determine whether the debris resulted from phagocytosis or self-lysis of Müller glial cells. Here, comparisons of Rho^P23H/P23H and Rho^+/+ mice allowed us to quantitatively demonstrate that dead rod cell bodies were internalized and phagocytosed by Müller glial cells. Rho^P23H/P23H mice are characterized by the rapid degeneration of rod photoreceptors. The nearly complete degeneration of rod cells occurs by PND 30 (25, 31). Because of the rapidity and high number of dying or dead cells at the early stages of degeneration, the Rho^P23H/P23H model allows the straightforward visualization of cells phagocytosed by Müller glia. Interpretations of the earlier EM studies were also challenging because of the proximity of Müller glial cells with densely packed photoreceptor cells. Therefore, it was difficult to ascertain whether Müller glial cells were internalizing or just wrapping other cells with their elaborate lamellar plasma membranes. To circumvent this issue, we took advantage of dissociated Müller glial cell preparations. We analyzed dissociated cells by immunocytochemistry to identify the Müller glia outer processes engaged in phagocytosis. The dead cell bodies localized within EAAT1⁺ plasma membranes and frequently colocalized with the lysosome marker Lamp1. Those results directly demonstrate that rod cell bodies are indeed internalized and degraded by phagolysosomes. Our study demonstrates that 90% of phagosomes containing dead cell bodies matured into phagolysosomes, suggesting Müller glial cells are capable of not only engulfing but also quickly sending the phagocytic contents to phagolysosomes for degradation.

In summary, Müller glia cells clear dead photoreceptor neurons in retinal degenerative disorders. We improved the methods to visualize phagocytic activities of individual and dissociated Müller glial cells. These improved methods will allow the visualization of Müller cell phagocytosis in other murine models of inherited blinding disorders that demonstrate a slower progression of photoreceptor degeneration (24). We also demonstrate that the effect of Müller glia phagocytosis is not limited to degenerating retinas but extends to healthy retinas as part of developmental cell death (42, 43) because dead cell bodies were found at low frequency within Müller glial cells isolated from Rho^+/+ retinas. These observations are consistent with the idea that Müller glial cells are involved in ontogenically programmed cell death (10, 11).

In the past 2 decades, Müller glial cells have gained attention in the field of regenerative medicine because they are capable of proliferating and transdifferentiating into retinal neurons after neuronal cell death (53, 54). An intriguing idea is that the stimulation of a phagocytic receptor is the triggering event for retinal regeneration by Müller glial cells. Supporting that notion, inhibition of the phagocytic receptor disrupts Müller glial proliferation, which is a prerequisite for their transdifferentiation (21). After the initiation of phagocytosis by Müller glial cells, communication and crosstalk between Müller glial cells and macrophages may also occur (55) and contribute to the coordination of their phagocytic activities, neuroprotection, and neuroregeneration. The present study, therefore, sets a foundation for future examination of the relationship between phagocytic signaling pathways among different glial types and retinal reorganization facilitated by Müller glia.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

APO

apochromat

BSA

bovine serum albumin

CD11b

integrin αM, macrophage-1 antigen α subunit

EAAT1

excitatory amino acid transporter 1

EM

electron microscopy

GS

glutamine synthetase

ip

inner process of Müller glial cell

IS

inner segment of photoreceptor cell

Lamp1

lysosome-associated membrane glycoprotein 1

NA

numerical aperture

ONL

outer nuclear layer

op

outer process of Müller glial cell

OPL

outer plexiform layer

PBSB

PBS containing 0.5% bovine serum albumin and 0.02% sodium azide

PBSBS

PBS containing 0.5% bovine serum albumin, 0.02% sodium azide, and 0.1% saponin

PBSS

PBS containing 0.1% saponin and 0.02% sodium azide

PND

postnatal day

RP

retinitis pigmentosa

RPE

retinal pigment epithelium/epithelial

WT

wild type

AUTHOR CONTRIBUTIONS

S. Sakami, Y. Imanishi, and K. Palczewski designed the research; S. Sakami performed the research; S. Sakami analyzed the data; and S. Sakami, Y. Imanishi, and K. Palczewski wrote the manuscript.