Bone morphogenetic protein 7 regulates reactive gliosis in retinal astrocytes and Müller glia

Tissue processing and fluorescent immunohistochemistry

Eyes from C57BL/6J mice (Jackson Laboratories, Bar Harbor, ME) were dissected from the heads of euthanized animals, washed in PBS (40 mM K₂HPO₄, 123.2 mM NaCl, 8.4 mM NaH₂PO₄; Fisher Scientific, Pittsburgh, PA), and fixed in 4% paraformaldehyde for immunohistochemistry. The eyes were then incubated in an ascending series of sucrose (5%, 10%, 15%, and 20%) made in 0.1M phosphate buffer, pH 7.4 overnight. The tissues were frozen in a 3:1 20% sucrose-in phosphate buffer and optimum cutting temperature (OCT) solution. Thick sections (10 μm) were cut using a Leica CM3050 S cryostat and placed on a Superfrost Plus slide (Fisher Scientific) treated with Vectabond (Vector Labs, Burlingame, CA), and were stored at −80 °C until used for immunohistochemistry. Antibodies used for immunohistochemistry are listed in Table 1. For immunohistochemistry, sections were fixed with 4% paraformaldehyde for 30 min. Sections were then washed in 1X PBS, permeabilized in methanol, and subjected to antigen retrieval by placing the sections in 1% sodium dodecyl sulfate (SDS, Fisher Scientific) in 0.01M PBS for 5 min. To aid in autofluorescence reduction, sections were treated with 1% sodium borohydride in PBS (Fisher Scientific) for 2 min at room temperature, then rinsed with PBS. Tissue was blocked by incubating with 10% serum in 1X PBS containing 0.25% Triton X-100 (Bio-Rad, Hercules, CA) at room temperature for 1 h and incubated with the primary antibody, diluted in 0.025% TritonX-100 PBS with 2% blocking serum, overnight at 4 °C. The following day, the slides were incubated in Dylight conjugated secondary antibody (Jackson Immunoresearch, West Grove, PA) at 1:800 diluted with 1X PBS, for 1 h at room temperature, then washed twice with 1X PBS for 5 min each rinse, and mounted with ProLong Gold with 4',6-diamidino-2-phenylindole (DAPI, Invitrogen, Grand Island, NY) or counterstained with Hoechst solution (diluted 1:500 in 1X PBS; Sigma-Aldrich, St. Louis, MO) and mounted with Aqua Poly/Mount (Vector Labs). For labeling of mouse tissue slides with GS, blocking and overnight incubation with primary antibody was performed as specified by the Vector mouse-on-mouse immunodetection kit (Vector Labs). For immunolabeling with NCAN, SMAD, BMPR1A, BMPR1B, and TAK1, biotin streptavidin amplification was performed. Following overnight incubation with the primary antibody, the sections were first incubated with biotinylated antibody (1:1,000, Vector Labs) for 1 h and then streptavidin-conjugated Dylight (1:33, Vector Labs) for 1 h at room temperature. For colabels of phospho-SMAD1/5/8 and phospho-TAK1 with sex determining region Y box 2 (SOX2), sections were subjected to heat antigen retrieval following methanol incubation. Briefly, the sections were incubated in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween-20, pH 6.0) at 65 °C for 1 h, allowed to cool for 20 min, rinsed 3× in water and 1X in PBS, and blocked with 10% horse serum in 1X PBS containing 0.25% Triton X-100 for 1 h at room temperature. Following overnight incubation with the primary antibodies, the slides were incubated with biotinylated horse anti-rabbit (1:300, Vector Labs) for 1 h and then Alexa Fluor anti-goat (1:500, Invitrogen) and horseradish peroxidase (HRP, 1:500, Biolegend) for 1 h. The slides were then washed 2× in 0.1M Tris-HCl, 0.15M NaCl, 0.05% Tween-20 (TNT) buffer and incubated with fluorescein plus amplification reagent diluted 1:300 in 1X tyramide signal amplification (TSA) amplification diluent (Perkin Elmer, Waltham, MA) for 5 min. The slides were washed 2× in TNT buffer, 2X in PBS, counterstained with Hoechst solution, and mounted with Aqua polymount (Vector Labs). Slides were viewed under an Olympus Fluoview FV 1000 confocal microscope.

Isolated retinas were processed and used for reverse transcriptase quantitative PCR (RT-qPCR) and western blot analysis. Briefly, animals were euthanized by CO₂ exposure. Eyes were then dissected using sterile forceps and the retina isolated. The retinas were homogenized and processed for RNA using the RNeasy Mini Kit as per the manufacturer’s protocol (Qiagen, CA). Briefly, isolated retinal tissues or cells were homogenized in buffer RLT. Seventy percent ethanol was added and the solution centrifuged in an RNeasy spin column. The flow through was discarded, and 700 µl of buffer RW1 was added and centrifuged. Following this step, 500 µl of buffer RPE was added to the spin column and centrifuged. This step was repeated once, following which the RNA was eluted with 40 µl of RNase free water. RNA was quantified using Nanodrop 2000c (Thermoscientific, Rockford, IL). For protein isolation, retinas isolated were homogenized and incubated with radioimmunoprecipitation assay (RIPA) lysis buffer (0.1% SDS, 50 mM Tris-HCl pH 7.5, 1% Triton-X100) supplemented with 0.01 mM phenyl methyl sulfonyl fluoride (PMSF; Roche Diagnostics, Indianapolis, IN) and 1:25 protease inhibitor cocktail (RPI Corp., Mount Prospect, IL) for 2 h at 4 °C. The tube was then centrifuged at 13,680 × g for 20 min at 4 °C and the supernatant aspirated into a new tube.

Cell isolation, cytospinning, and immunocytochemistry

Retinal cells were isolated as previously described [33]. Briefly, mouse retinas were prepared in sterile 0.9% saline. Approximately 200 mg of retinal tissue was incubated with saline containing 0.4 mg/ml papain for 30 min at 37 °C. The tissues were triturated with a siliconized pipette and dissociated single cells were centrifuged at 500 × g for 10 min. The cells were fixed in 4% paraformaldehyde for 10 min, centrifuged, and washed with saline. The cells were then placed onto a Superfrost Plus slide (Fisher Scientific) treated with Vectabond (Vector Labs) using a Cytospin4 (Thermoscientific) and stored at −80 °C until used for immunocytochemistry. For immunocytochemistry, the cells were permeabilized in methanol, incubated with 1% SDS in 0.01M PBS, and washed in 1X PBS. The slides were then incubated with 1% sodium borohydride, blocked using the mouse on mouse kit (Vector Labs), and colabeled with GS and the BMP receptor antibodies (BMPR1A, BMPRIB, and activin-like kinase receptor 2 [ALK2]). Following overnight incubation with the primary antibodies, the slides were incubated with the appropriate secondary antibodies with biotin streptavidin amplification used for the receptor antibodies. The slides were then washed, counterstained with Hoechst solution, and mounted with Aqua polymount.

Cell culture

The mouse retinal astrocytes were isolated as previously stated [34,35]. Cells were cultured in DMEM (Catalogue # D1152, Sigma-Aldrich, St. Louis, MO) containing EC growth supplement (Sigma-Aldrich), 1% penicillin/streptomycin (Sigma-Aldrich), 1 mM sodium pyruvate (Life Technologies, Grand Island, NY), 20 mM HEPES (Sigma-Aldrich), 2 mM Glutamine (Life Technologies), 1X nonessential amino acids (Sigma-Aldrich), heparin (Sigma-Aldrich), 10% fetal bovine serum, and 44 U/ml of murine recombinant interferon γ (R&D Systems, Minneapolis, MN). The cells were grown on Cellbind™ dishes (Fisher Scientific) in an incubator with 5% CO₂ at 33 °C and passaged every 3–4 days using trypsin-EDTA (Sigma-Aldrich). The human MIO-M1 Müller glial cell line was obtained from Dr. G.A. Limb at the Institute of Ophthalmology, University College London [36]. MIO-M1 cells were grown in DMEM containing 10% fetal bovine serum and 200 mM glutamine (Life Technologies). The cells were grown in sterilized tissue culture dishes (BD Falcon, Corning, NJ) in an incubator at 5% CO₂ and 37 °C and passaged every 6–7 days using trypsin-EDTA and the medium was changed every 3–4 days.

Treatment of retinal astrocytes and MIO-M1 cultures in vitro with sodium peroxynitrite (Cayman Chemicals, Ann Arbor, MI) was performed as previously stated [37]. Briefly, confluent astrocyte cell cultures were washed 3× with PBS supplemented with 0.8 mM MgCl₂, 1 mM CaCl₂, and 5 mM glucose. They were then incubated in 1 ml of 1X PBS (50 mM Na₂HPO₄, 90 mM NaCl, 5 mM KCl, 0.8 mM MgCl₂, 1 mM CaCl₂, and 5 mM glucose, pH 7.4), followed by three additions of sodium peroxynitrite at a concentration of 0.15 mM. The first bolus of sodium peroxynitrite was added to one edge of the dish and the buffer was swirled to allow mixing of the peroxynitrite throughout the dish. This step was repeated twice while changing the edge at which the addition was made and then incubating for 5 min. The buffer was then removed, replaced with the respective growth medium, and placed in the appropriate incubator. The cells were then processed after 32 h for protein or RNA. Some cultures were treated with recombinant BMP7 and BMP 4 (R&D Systems) reconstituted in 0.4% HCl-PBS. Dishes were treated with either vehicle or 100 ng/ml of BMP7 or BMP4 for 24 or 36 h. Cells were then processed for either RNA or protein for analysis via RT-qPCR or western blotting, respectively.

Bone morphogenetic protein 7 injections in vivo

The mice were injected intravitreally with BMP7 (1 μl of 20 ng/μl). Intravitreal injections were performed with a pump microinjection apparatus (Harvard Apparatus) and pulled glass micropipettes. Each micropipette was calibrated to deliver 1 μl of vehicle or vehicle containing BMP7. The mice were anesthetized under a dissecting microscope; the sharpened tip of the micropipette was passed through the sclera, just behind the limbus, into the vitreous cavity. Once the micropipette was in place, the vehicle or BMP7 was injected using a PL-1000 picospritzer (Harvard Apparatus, Holliston, MA). D3 or D7 post-injection, the animals were sacrificed and retinas were dissected for RNA and protein preparations. Each experiment was performed in triplicate.

Western blot analysis

Antibodies used for western blot analyses are listed in Table 1. The total protein was quantified using a Pierce BCA Protein Assay Kit (Thermoscientific) following the manufacturer’s protocol. Briefly, 5 μl of sample was mixed with 495 μl of working reagent (prepared by mixing Pierce BCA Reagent A and B in 50: 1 ratio) and 500 μl of water. This solution was incubated at 65 °C for 30 min, followed by colorimetric analysis using a NanoDrop 2000c spectrophotometer (Thermoscientific). Fifty μg of total protein mixed with the loading dye at a 3:1 ratio was then loaded and run on a 4%–20% SDS polyacrylamide gel (Expedeon, San Diego, CA) at 125 V for 1 h. Proteins were transferred to a polyvinylidene fluoride (PVDF) membrane (Bio-Rad) via a wet transfer and subjected to immunoblotting. Prior to incubation with the antibody, the membrane was blocked using a 5% milk solution in Tris-buffered saline–Tween (TBST; composition – 20 mM Tris base, 137 mM sodium chloride, 1 M HCl, 0.1% Tween-20, at pH 7.6) for 1 h. The blots were then incubated with the primary antibody diluted in TBST at 4 °C overnight. The blots were washed twice with TBST and then incubated with a peroxidase-conjugated secondary antibody (Thermoscientific) diluted to 1:5,000 in TBST for 1 h in the dark at room temperature. The blots were incubated with either Pierce ECL Western Blotting Substrate (Thermoscientific) or SuperSignal West Femto Chemiluminescent Substrate (Thermoscientific), and the bands visualized on X-ray films (Thermoscientific). β-TUBULIN was used as a loading control and densitometry of the blots was performed using Image J software.

Real-time quantitative PCR

Prior to cDNA synthesis, RNA samples were run on a 1% agarose gel to confirm the overall quality of the total RNA. The cDNA was synthesized from 1 µg of total RNA with an iScript cDNA Synthesis Kit (Bio-Rad) according to the manufacturer’s protocol and diluted to 1:20 before adding to qPCR wells. Real-time quantitative PCR (RT-qPCR) was performed using the 7300 RT detection system (Applied Biosystems, Carlsbad, CA) with the Power SYBR green PCR master mix (Invitrogen). The primer pairs used have been listed in Table 2 (mouse) and Table 3 (human). The total volume for each reaction was 20 µl using the diluted cDNA, corresponding to 5 ng of initial total RNA and 0.4 mM of each primer. The cycler conditions used were as follows: initial denaturation at 95 °C for 10 min, 45 cycles of denaturation at 95 °C for 10 s, annealing at 60 °C for 20 s and extension at 72 °C for 20 s, followed by a final extension at 72 °C for 5 min. Efficiency of the primer sets was determined by the standard curve method, where efficiency E=((10(−1/Ct2 – Ct1)) – 1) × 100. A no template control and an internal control using primers specific for β-2 microglobulin (B2M) were used for each run. The amplified samples were run on a 2% agarose gel to confirm that the amplification was of the right size. The change in the gene expression levels was done using the 2 ^–ΔΔCt method, where “Ct” is the crossing threshold value.

Statistical analysis

Statistical analysis of RT-qPCR data was performed by unpaired t test between the control and treated groups. Statistical analysis of densitometry results was performed using the Student t test. All analyses were performed using SPSS software (IBM) and Excel 2010 (Microsoft).

Müller glia express bone morphogenetic protein type IA, IB, and activin like kinase receptor 2 receptors

BMPs bind preferentially to three type I receptors; receptor IA, IB, or ALK2. Double-label immunofluorescence was used to determine whether any of the type I BMP receptors was expressed by Müller cells or retinal astrocytes. Sections through adult murine retina were colabeled with antibodies specific for GS, which labels Müller glial cells and retinal astrocytes, and BMPR1A, BMPR1B, or ALK2 (Figure 1). All three receptors showed similar labeling patterns in the retina with Müller glial cell processes labeled in the outer nuclear layer (ONL), outer plexiform layer (OPL), and GCL (Figure 1A-C). Photoreceptor outer segments (OS) and ganglion cell bodies were also clearly labeled for all three receptors. To confirm the label of Müller glial cells and their processes by antibodies to type I receptors, enzymatically dissociated cells were colabeled with antibodies to GS and BMPRIA, BMPR1B, or ALK2 (Figure 1D). Immunolabeled Müller glia were readily apparent in dissociated samples, as they had retained their shape throughout the processing. All three receptors appeared to colabel GS (+) cells, indicating that the cells are responsive to BMPs in vivo.

Bone morphogenetic protein 7 signaling components in the retinal glia

The BMP signaling pathway members present in the Müller glial and retinal astrocytes were also investigated to determine which pathways might be activated in the presence of BMPs. To investigate the canonical BMP pathway, sections were colabeled with antibodies against Sox2, which labels Müller glial cells, retinal astrocytes, and cholinergic amacrine cells in the mature murine retina, and SMAD1. SMAD1 was localized throughout cells of the inner nuclear and GCLs, and was colocalized with SOX2 (+) cells (Figure 2A). To determine whether the BMP-kinase pathway could be activated in response to BMPs, sections through the adult retina were colabeled with antibodies to GS and TAK1 (Figure 2B). While TAK1 was prominently expressed in the retina, most notably in a subpopulation of inner nuclear layer (INL) cells and the GCL, little to no co-expression could be detected in GS (+) cells (Figure 2B). Last, potential involvement of the FRAP-STAT pathway in Müller glial BMP signaling was investigated by colabeling sections with GS and FKBP12 (Figure 2C). FKBP12 label was noted in the inner nuclear, ganglion cell, and inner plexiform layers (IPLs). No apparent colabel was detected of FKBP12 and GS (Figure 2C).

Bone morphogenetic protein 7 can trigger changes in retinal astrocytes and MIO-M1 Müller cells resembling mild reactive gliosis

The role of BMP7 in reactive gliosis was first tested by the addition of the factor to an in vitro system using isolated retinal astrocytes or a Müller glial cell line, MIO-M1 [34-36]. Previously characterized retinal astrocytes have been shown to express paired box 2 (PAX2), NG2, and GFAP [34,35]. To use these cells, their ability to exhibit changes associated with reactive gliosis first had to be determined. The cells were briefly treated with vehicle or sodium peroxynitrite, a strong oxidizing agent that is released by injured neurons and has been shown to trigger reactive gliosis [38]. Following treatment, levels of mRNAs known to be regulated during reactive gliosis were examined 16 (not shown) and 32 h after treatment using RT-qPCR (Figure 3). In the analysis of the RT-qPCR levels throughout this study, two thresholds had to be met in order for the change to be considered convincing: 1) The comparative change with controls had to be statistically significant, and 2) an average increase in levels had to reach 1.5-fold or above that of controls, or similarly, a decrease had to reach 0.5-fold or below. No statistically significant changes in expression were noted between vehicle and peroxynitrite 16 h following treatment (not shown). In contrast, there were statistically significant increases that reached the 1.5-fold threshold in the level of many mRNAs within the reactive gliosis panel in peroxynitrite-treated cells in comparison to vehicle-treated ones, including Pax2, lipocalin 2 (Lcn2), Mmp9, Pcan, S100β, toll-like receptor 4 (Tlr4), tissue inhibitor of matrix metalloproteinase 2 (Timp2), and Gfap (Figure 3A). There was also a statistically significant decrease in the glutamate aspartate transporter (Glast), which also reached the level of 0.5 and below threshold. Levels of protein were investigated by western blotting for three proteins, namely GFAP, thioredoxin-interacting protein (TXNIP), and PAX2 (Figure 3B). While there was a modest increase in GFAP and PAX2 protein in comparison to control vehicle-treated cells, there was no increase in levels of TXNIP 36 h following treatment (Figure 3B).

To test whether BMP7 treatment could also trigger signs of reactive gliosis in vitro, retinal astrocytes were treated with either vehicle or BMP7 for 24 or 36 h. Following 24 h of vehicle or BMP7 treatment, retinal astrocytes treated with BMP7 showed a statistically significant increase in transcript levels above the 1.5-fold threshold for Ncan, Pcan, acid-sensing ion channel 1 (Asci1), Glast, Txnip, Pax2, and Gfap (Figure 3A). There was also a decrease in levels of Mmp14 in BMP-treated versus vehicle-treated retinal astrocytes that met the 0.5 threshold and was a statistically significant change in comparison to vehicle-treated cells (Figure 3A). Western blotting was used to assess the levels of a subset of proteins in vehicle- and BMP-treated cells. A statistically significant increase in protein levels of GFAP and TXNIP protein was noted for BMP7-treated cells compared to vehicle-treated cells at 24 h after treatment (Figure 3B). However, there was no detectable differences in the levels of TXNIP or PAX2 at 24 h (Figure 3B). Cells treated with BMP7 for 36 h showed changes in levels of a slightly different cohort of mRNAs; Mmp11, Timp2, Ncan, and Mmp9 were modestly increased, while Asic1, inwardly rectifying potassium channel 2.1 (Kir2.1), and Pax2 were decreased slightly (Figure 3A). By western blotting, small but statistically insignificant increases were observed in GFAP, TXNIP, and PAX2 (Figure 3B).

Similar to the retinal astrocyte cultures, treatment of the Müller glial cell line MIO-M1 with peroxynitrite also led to increases in some of the markers in the reactive gliosis panel (Figure 4A). Western blots showed a modest increase in protein levels of GFAP and S100β; however, no change was noted in PAX2 expression (Figure 4B). Treatment of MIO-M1 cells with BMP7 yielded similar, but slightly different changes in mRNA levels following treatment for 24 or 36 h. At 24 h, Gfap and Pax2 mRNA levels were increased in comparison to control vehicle-treated cells above the 1.5-fold threshold, while there was a decrease in Txnip, Gal3, Lcn2, and Mmp9 levels below the 0.5-fold threshold (Figure 4A). Western blotting showed statistically significant changes in GFAP, S100β, and TXNIP protein levels in comparison to controls; however, increases in PAX2 were not statistically significant (Figure 4B). Treatment of cells with BMP7 for 36 h yielded increases in Gfap and Tlr4 mRNA levels in comparison to control cultures and a decrease in secreted phosphoprotein 1 (Spp1; Figure 4A). By western blot, densitometric analysis showed a statistically significant decrease in PAX2 levels and no significant changes in GFAP, S100β, or TXNIP levels in comparison to vehicle-treated cells (Figure 4B).

To determine whether treatment of cells with another BMP family member, BMP4, led to similar changes in reactive gliosis markers, retinal astrocytes and MIO-M1 cells were treated with BMP4 for 24 or 36 h, and changes in mRNA levels were assessed by RT-qPCR. Overall, there were fewer statistically significant changes in BMP4-treated cells at or over the level of a 1.5-fold increase or a 0.5-fold decrease in comparison to vehicle-treated cells in both retinal astrocytes and MIO-M1 cells (Figure 5). In retinal astrocytes, Txnip and GS were increased more than 1.5-fold over control and the difference was statistically significant. Only Mmp9 showed a decrease in comparison to control at 24 h following treatment. In comparison, at the 36 h time point, only the decrease in Mmp9 was statistically significant and reached the 0.5-fold level. In the MIO-M1 cells, only Asic1 and Spp1 were upregulated at least 1.5-fold over levels of control mRNA at 24 h following addition. Further, at the 36 h time point, a statistically significant increase at the 1.5-fold level in endothelin receptor type B (Ednrb) and Asic1 were the only increases noted at 36 h following addition of BMP4 (Figure 5).

Intravitreal injection of bone morphogenetic protein 7 induces reactive gliosis

To further explore the role of BMP7 in reactive gliosis, wild-type adult mouse eyes underwent intravitreal injections of vehicle or BMP7. To determine the BMP signaling cascade in BMP7-mediated gliosis, the vehicle- and BMP7-injected retinas were colabeled with SOX2 and phospho-SMAD 1/5/8 (Figure 6) or phospho-TAK1 (Figure 7). While the vehicle-injected retina showed little or no phospho-SMAD and SOX2 colabeling, the BMP7-injected retinas showed an increased colabel for phospho-SMAD 1/5/8 and the Müller glial marker SOX2 (Figure 6). Colabeling was also performed for SOX2 and phospho-TAK1 to determine the activation of the BMP-MAPK noncanonical signaling cascade. The vehicle-injected retina showed phospho-TAK1 label primarily in the GCL (Figure 7), while the BMP7-injected retina showed an increase in the phospho-TAK1 label in the INL and an increase in colabeling with SOX2 (Figure 7). Further, retinas at 3 and 7 days after injection were examined using RT-qPCR, immunofluorescence, and western blot analysis. Three days following injection, the only statistically significant change that reached the threshold was a decrease in Pax2 and Gal3 mRNA levels (Figure 8A). At the 7 day time point, however, statistically significant increases in mRNA levels above the threshold were observed for Txnip, Glast, Mmp9, Lcn2, GS, Kir2.1, and Mmp14. There were also several decreases noted at 7 days after injection, specifically in Tlr4, Pax2, and Egfr (Figure 8A). Immunohistochemistry on sections through the injected retinas at 3 days following injection revealed no readily apparent changes in GFAP, S100β, or NEUROCAN in BMP7-injected retinas as compared to vehicle-injected ones. In contrast, there was a marked increase in GFAP, S100β, and NEUROCAN in BMP7-injected retinas when compared with their control counterparts (Figure 8B). In addition, Nomarski optics of the injected retina showed some thickened Müller glial processes in the IPL, reaffirming the claim that BMP7 induces hypertrophy of at least a subpopulation of Müller cells at 7 days after injection (Figure 9). By western blot, small but significant increases were observed in S100β and TXNIP at 3 days following injection in BMP7-injected retinas, and while there were increases in GFAP, S100β, and TXNIP at 7 days after injection in BMP-injected retinas, none of the changes were statistically significant (Figure 8C).

Regulation of glutamine synthetase during reactive gliosis

Reactive gliosis is a common and a complex response observed following neural injury [39]. One response typical of Müller cells undergoing gliosis is a change in levels of glutamine synthesis, the enzyme critical for glutamate-glutamine cycling in the nervous system. Under normal circumstances within the retina, GS prevents the excitotoxic effects of glutamate and is necessary for ammonium detoxification [40]. In the retina, GS is widely expressed in the astrocytes and Müller glial cells [41]. Levels of GS are exquisitely regulated by injury in an injury- or disease-state manner. For instance, expression of GS is downregulated in disease states in which photoreceptors (the major source of glutamate in the retina) degenerate [42,43]. However, in disease states where the presence of GS is necessary to detoxify the retina when high levels of ammonia, glutamate, or glutamate agonists are present (such as hepatic retinopathy or following kainic acid injection), expression levels of the enzyme increase [44,45]. Finally, levels of GS can remain stable in some injury and disease states, including diabetic retinopathy and following nerve crush [46,47]. In the studies presented here, GS levels increased following addition of BMP7 in vitro and in vivo, suggesting that the BMP7 addition mimics situations where there is an increase in glutamate or NH4+ concentrations in the retina. Further studies will be necessary to determine whether differential regulation of BMP7 occurs in a disease-dependent manner.

Bone morphogenetic protein 7 triggers reactive gliosis via the SMAD and the transforming growth factor-β activated kinase pathway

Reactive gliosis encompasses a wide range of responses, ranging from hypertrophy to regulation of transporter channels and enzymes, as well as cell migration, proliferation, and dedifferentiation [20]. The triggers for these different molecular and functional responses vary with the injury or disease state. Many growth factors have been implicated in regulating one or more of the changes associated with gliosis. For instance, epidermal growth factor has been associated with Müller glial proliferation, FGF2 with the scarring response, and CNTF with GFAP upregulation [17,23]. Further, inflammation is also a commonly observed response during gliosis. Factors such as TNF-α and CNTF have been known to play a role in inflammatory response of reactive gliosis, while TNF-α and some of the interleukins are also known to induce proliferative reactive gliosis [48-50].

BMPs have also been associated with various changes that occur during reactive gliosis; however, many of the studies to date have focused on the role of the BMPs following injury to the spinal cord [28,29,51]. Initial studies indicated that BMP signaling increased at the site of spinal cord lesions, and treatment of astrocytes in vitro with BMPs 4 or 7 increased expression of CSPGs typically associated with glial scars [52]. Further, inhibition of BMPs following spinal cord injury leads to a decrease in CSPGs and an increase in the ability of axons to regenerate over time [28,51]. Similarly, knockout studies for two of the type 1 BMP receptors have shown that BMPR1A is critical in the hypertrophy response that occurs early in reactive gliosis and BMPR1B is critical for some aspects of the scarring response. Little is known about the role of BMPs in injury or disease within the retina or optic nerve. Detailed descriptions of the expression patterns of the BMPs, BMP receptors, and several BMP-binding proteins have been described in the developing and adult optic nerve, the optic nerve head, and the trabecular meshwork [26,53,54]. An increase in BMP signaling within the trabecular meshwork has been implicated in blocking TGF-β-driven changes in gene expression [55-57]. Finally, BMP stimulation downstream of EGF signaling has recently been implicated in Müller glial cell proliferation [58].

In the studies detailed here, we showed that the Müller glial cells express BMPR1A, BMPR1B, and ALK2 receptors. While it is probable that astrocytes are also expressing the receptors, due to the intensity of the label in the ganglion cell and inner limiting membrane area, we were unable to determine this for certain. BMP7 preferentially binds the ALK2 receptor and transduces the signal via SMAD 1/5/8 [59]. To examine BMP signaling in the retinas, we first analyzed the expression and glial localization of the canonical and noncanonical BMP signaling components, namely SMAD, TAK and FRAP. While colabeling showed that SMAD and TAK were localized with Müller glia, with SMAD to a greater extent than TAK, FRAP showed no localization (Figure 2). Analysis of the activated forms of SMAD (phospho-SMAD) and TAK (phospho-TAK) revealed that there is activation of both in BMP7-injected retinas (Figure 6 and Figure 7). Apart from the glial cells, several other cells also showed positive labels for both phospho-SMAD and phospho-TAK. This upregulated BMP signaling in both macroglial and other cells of the retina suggests that BMP7 can regulate gliosis either by directly acting on the retinal glia or indirectly by regulating other retinal cells.

In comparison to BMP7, treatments with BMP4 did not appear to induce gliosis in retinal astrocytes or Müller cells in vitro. While BMP4 has been shown to have a neuroprotective effect when injected before a toxic insult, BMP7 was not shown to have the same protective effect [60]. The differential response of retinal neurons and macroglia to BMP4 and 7 is indicative of the involvement of different signaling pathways, but further study is necessary to elucidate the mechanism of this differential response.

Differences in response patterns

Apart from growth factors, various molecular factors such as adenosine tri-phosphate (ATP), nitric oxide (NO), and so on can also trigger reactive gliosis [20]. Although roles of specific factors have been implicated in certain aspects of gliosis such as GFAP, vimentin, and neurofilament expression [14,60], there does not seem to be a clear correlation between the various gliosis markers and its regulators. GFAP is used as the standard indicator of gliosis. However, it and other markers such as GS, MMPs, and others have been shown to be regulated differentially depending on the stress to which they are subjected [42,47,61,62]. In this study, RT-qPCR analysis of RNA from BMP7-injected retinal tissue compared to the RNA prepared from treated retinal astrocytes and Müller glial cells showed a greater deviation from their respective control treatments (Figure 3, Figure 4, and Figure 8). This could be attributed to the fact that the cells in vitro represented a pure population of cells, while the retinal tissue represented a mixed population of macroglia, microglia, neurons, and components of the vasculature. We also observed differences in the profile of regulated genes between the sodium peroxynitrite and BMP7-treated cells. This observation is consistent with the idea that various factors are responsible for driving specific aspects of the gliosis response [13,14,21].

In addition to differential expression between peroxynitrite and BMP7 treatments in vitro, as well as in vivo versus in vitro responses, we also noted that there were substantial differences between the regulation of various markers at the mRNA and protein level. This asymmetry has also been noted by other investigators who have compared mRNA and protein levels, indicating that there may be multiple levels of regulation involved in gliosis [63-67]. Although mRNA levels showed increased regulation both in vivo and in vitro, a similar change was not detected at the protein level. mRNA translation and protein levels are regulated in the cell by several systems, namely regulation of mRNA localization and translational repression mediated by RNA binding proteins, regulation via miRNAs, the ubiquitin proteasome system, and changes in protein stability [68,69]. Here, we did not observe a linear correlation between the mRNA and protein levels in the gliosis models analyzed. We thus hypothesize that protein is regulated at the mRNA translation level, mediated either by RNA binding proteins or by miRNAs. Furthermore, protein stability may also be regulated in BMP7-mediated gliosis, leading to differences in mRNA and protein levels. mRNA binding protein such as cytoplasmic polyadenylation element binding protein (CPEB) has been previously shown to upregulated in gliosis [70]. One of the downstream targets of the TAK-mediated BMP cascade is p38-MAPK [59,71]. p38-MAPK can regulate the phosphorylation, and thereby, the activation of CPEB [70]. miRNAs are small, 22–25 nucleotide long, noncoding RNAs that play an important role in cell fate and development. miRNAs have been found that are specific to the nervous system, as well as to the retina, and BMPs have also been shown to regulate miRNA levels [72-75]. Further, regulation of miRNAs has also been observed in astrocytes following injury to the spinal cord [76]. The miRNAs bind to their target RNA and can either direct it to degradation or repress translation. ceRNA interaction mediated by miRNA, which can target more than mRNA, has been previously reported in diabetic retinopathy in RPE cells [77]. Similar interactions may also be prevalent in the glial cells leading to translational regulation.

In this study, we observed a more severe response in gliosis in vivo as compared to in vitro. Immunohistochemistry for BMP receptors in the mouse retina showed a large population of neuronal cells other than Müller glia and retinal astrocytes to express them. Thus, binding of BMP7 to receptors in these cells and to Müller glia could amplify the gliosis signal, leading to a comparatively severe response in the mouse retina.

Extracellular matrix and reactive gliosis

One of the responses seen during reactive gliosis is ECM modifications. These changes in the matrix are primarily due to the regulation of MMPs, which remove ECM molecules, and the regeneration-inhibiting chondroitin sulfate, ultimately leading to the formation of a glial scar [20,78]. The formation of the scar serves to protect the injured site while also preventing axonal regrowth [25].We observe here an increase in RNA levels of ECM molecules such as Ncan, Pcan, and Mmps, both in vivo and in vitro (Figure 3, Figure 4, and Figure 8). Furthermore, immunohistochemistry of the BMP7-injected mouse retina also showed an increase in localization of Ncan in the INLs of the retina (Figure 8). Other members of the TGF-β family have been previously shown to increase proteoglycan production in reactive glial cells [16]. Here, we observed an increase in Ncan at the 7 day stage when compared to the earlier 3 day time point, a result similar to a previous study [16]. Further study is necessary to determine whether the effects of BMPs on ECM production are direct or indirect.