Efficacy of Rho-kinase Inhibition in Promoting Cell Survival and Reducing Reactive Gliosis in the Rodent Retina

Abstract

purpose. To analyze the outcomes of Rho-kinase (ROCK) inhibition on retinal cell survival and glial reactivity under adverse conditions.

methods. Organotypic cultures of mouse retinas were incubated with the specific ROCK-inhibitor H-1152P for 24 to 48 hours under serum deprivation. Cell damage was determined by ethidium homodimer-1 uptake and caspase-3 cleavage. Immunohistochemistry and Western blot were performed to detect reactive gliosis and to confirm the specificity of H-1152P. The cytokine profile of the culture medium was analyzed using a membrane-based array. H-1152P was administered intravitreally into rats before optic nerve crush (ONC) and the extent of apoptosis and reactive gliosis was determined after 7 days.

results. Cell damage in cultured retinas was significantly reduced in response to 1 μM H-1152P, particularly in the ganglion cell layer. This was associated with a decrease in the levels of glial fibrillary acidic protein (GFAP) isoforms and the number of reactive astrocytes, Müller cells, and microglia. The release of proinflammatory cytokines including TNF-alpha, interferon-gamma, and IL-6 was also reduced, which likely contributed to the significantly lower toxicity of the conditioned media collected from retinas incubated with H-1152P. H-1152P (1 μM) suppressed the ROCK-dependent phosphorylation of adducin without a considerable interference with the protein kinase A/C-mediated phosphorylation events, indicating the specificity of the inhibitor for ROCK. H-1152P also resulted in a significant decrease in the extent of apoptosis and reactive gliosis after ONC.

conclusions. These results demonstrate the neuroprotective effect of H-1152P-mediated ROCK-inhibition on retinal cells under stress, which may rely partly on the attenuation of glial cell reactivity.

Like elsewhere in the central nervous system (CNS), the transition of quiescent glial cells to the activated phenotype represents one of the earliest and most remarkable responses of the retina to various stress factors such as mechanical or ischemic injury. ^{1 2} The glial reactivity represents an effort to maintain homeostasis and normal neuronal function under stress. However, the persistence of reactive gliosis can also participate in the course of neuronal loss and is associated with diverse pathologic events in the retina such as age related macular degeneration, retinal detachment, diabetic retinopathy, and glaucoma. ^{3 4 5 6} Reactive glial cells can release immunomodulatory cytokines and reactive oxygen species, which diffuse into the adjacent regions and activate an apoptotic death program, accounting for the loss of neurons that might have survived the primary damage. ^{7 8 9} The expression of repulsive extracellular matrix molecules by reactive astrocytes also interferes with the regeneration of injured retinal ganglion cell (RGC) axons. ¹⁰ The extent of glial cell reactivity therefore appears to be a critical determinant in the transformation of a protective attempt to a severe response, facilitating the spread of apoptotic cell death and neurodegeneration even after the removal of the primary cause of damage. 

The repulsive guidance molecules expressed by the glial scar converge on the activation of the small intracellular protein RhoA from the Rho family of GTPases and its downstream effector Rho-kinase (ROCK). ROCK is a ubiquitously expressed serine-threonine kinase that mainly regulates the organization of the actin cytoskeleton and the associated dynamic events such as contraction, adhesion, motility, cell-cycle progression, and gene expression in numerous cell types. ^{11 12} The activation of ROCK either by RhoA or directly by the caspase-dependent cleavage of its autoinhibitory domain also regulates the morphologic events characterizing the execution phase of apoptosis ¹³ and possibly the intracellular signaling involved in the initiating stages in a cell type-dependent manner. ¹⁴ Consistently, there is increasing evidence demonstrating the protective effects of RhoA/ROCK-inhibition on adult retinas. For instance, the intraocular injection of the RhoA antagonist C3 is reported to increase both axonal regeneration and RGC survival after optic nerve axotomy in rats. ¹⁵ Moreover, the inhibition of ROCK was shown to decrease the extent of N-methyl-d-aspartic acid (NMDA)-induced neurotoxicity in rat retinas. ¹⁶ The treatment with the ROCK-inhibitor Y-27,632 also promoted the viability of primary RGCs, the RGC-5 cell line, ¹⁷ and the RGCs of rats with transient retinal ischemia. ¹⁸ Yet, despite these promising findings on neuronal regeneration and survival, the outcomes of RhoA-ROCK inhibition on glial cell reactivity in the retina under such injury paradigms still remain to be elucidated. 

Considering the interdependence between neurons and glial cells under both normal and pathologic conditions, the use of a model where the cell-cell interactions are preserved at a higher degree is believed to provide more insight into the outcomes of ROCK inhibition on the response of the retinal neurons and glia to stress. The organotypic culture of whole or partial retina serves as a convenient in vitro model for this purpose, since the cells can be maintained with the multi-layered arrangement, components of the extracellular matrix, and presumably the cell-cell associations being essentially preserved for several days to weeks. ¹⁹ The long-term maintenance of mouse retinas in serum-free culture was also reported to be possible, allowing the investigation of molecules synthesized by the retina and studying drug effects without the interference of unknown serum factors. ²⁰ In this study, the role of ROCK in retinal cell fate was therefore investigated by using H-1152P, the most specific of the commercially available ROCK-inhibitors, ¹² initially on organotypic cultures of mouse retinas under serum deprivation. The axotomy of the ganglion cells in this model deprives the cells of their efferent synapses and trophic support. The incubation in a minimal medium without serum supplement in turn accelerates the retrograde death of the RGCs. We sought to study the effect of H-1152P treatment on retinal cell survival under these severe conditions by histologic evaluations with particular emphasis on the ganglion cell layer (GCL). Furthermore, we characterized the outcomes of H-1152P administration on glial cell reactivity, cytokine release, and the impact of this process on the toxicity of the retinal conditioned medium. The specificity of H-1152P was also analyzed to ensure that the observed effects reflected the outcomes of ROCK-inhibition. Lastly, we studied the influence of H-1152P on the extent of apoptosis in the GCL and glial cell reactivity after optic nerve crush (ONC) in rats to gain insight into the in vivo events in the rodent retina associated with ROCK inhibition. 

Materials and Methods

Animals

All experiments were performed in accordance with the guidelines of the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Male NMRI mice (8 to 10 weeks old; Harlan-Winkelmann, Borchen, Germany) were used for the preparation of organotypic retinal cultures, and adult female Brown Norway rats (150 to 200 g body weight; Charles River, Sulzfeld, Germany) were used in ONC experiments. Animals were maintained in a 12-hour (h) light–dark cycle with food and water ad libitum. 

Organotypic Retinal Cultures

Male NMRI mice were killed in a CO₂ chamber. The eyes were enucleated under a laminar flow hood and collected into sterile 0.1% glucose-PBS (w/v). The cornea was cut using a scalpel and the lens and the vitreous body were removed. To facilitate the detachment of the retinas, the eyecup was sheared off with fine forceps. Retinas dissociated from the pigment epithelium were mounted onto cellulose nitrate filters pre-soaked in PBS (0.45 μm pore size, cut into approximately 1 × 1 cm pieces; Sartorius, Göttingen, Germany) with the GCL exposed. The retinal flatmounts were incubated in 1 mL of Dulbecco’s modified Eagle’s medium (DMEM)/F-12 (Gibco, Invitrogen, Carlsbad, CA) with or without H-1152P (Calbiochem, Merck, Darmstadt, Germany) for 24 or 48 h at 37°C and with 5% CO₂. 

For analyzing the toxicity of the retinal conditioned medium, flatmounted retinas were incubated for 24 h as described above. At the end of this period, the culture medium was collected and used immediately for the incubation of freshly mounted retinas for 24 h, with the addition of H-1152P when indicated. 

Assessment of Cell Death via Ethidium Homodimer (EthD)-1 Staining

At the end of the culture period, the retinas were incubated in 4 μM EthD-1 (Molecular Probes, Eugene, OR) in 0.1% glucose-PBS for 30 minutes, fixed with 4% paraformaldehyde (PFA)-PBS for 30 minutes, counterstained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI) to detect the nuclear morphology, and analyzed by fluorescence microscopy (Carl Zeiss Meditech GmbH, Göttingen, Germany) using analysis software (Soft Imaging System, Münster, Germany). Quantification of the stained cells was performed in 8 to 10 areas of 0.16 mm² per retina. 

Immunohistochemistry on Retinal Cryosections

Flatmounted mouse retinas and the rat eyes that underwent ONC were fixed in 4% PFA for 2 h, incubated in 4% sucrose-PBS (w/v) overnight at 4°C, kept in 20% sucrose (w/v) –5% glycerol (v/v) for 2 days at 4°C, and embedded in optimal cutting temperature (OCT) compound (Sakura Finetec, Torrance, CA). Transverse cryosections at 14 μm were fixed in ice-cold acetone for 10 minutes, air-dried, and washed with PBS. The sections were blocked in 3% BSA-PBS with 0.3% Triton X-100 (BSA-PBST) for 30 minutes at room temperature (RT) and incubated with the primary antibodies against cleaved caspase-3 (rabbit monoclonal, 1:100 in BSA-PBST; Cell Signaling Technology, MA), GFAP (rabbit polyclonal, 1:400; DAKO, Glostrup, Denmark) or CD11b (rat polyclonal, 1:100; Serotec, Oxford, UK for the mouse retinas, and mouse monoclonal, 1:100; GeneTex, San Antonio, TX for the rat retinas) overnight at 4°C in a humidified chamber. The sections incubated with the blocking buffer alone served as negative controls. After three PBS-washes of 5 minutes, the sections were incubated with biotin conjugated rabbit anti-rat (1:200; DAKO) secondary antibodies followed by Cy3-conjugated streptavidin (1:500; Jackson ImmunoResearch, West Grove, PA) or with Cy3-conjugated goat anti-rabbit antibodies (1:400; Jackson ImmunoResearch) or with Alexa 488-conjugated goat anti-mouse antibodies (1:400; Molecular Probes) for 1 h at RT. The nuclei were then counterstained with DAPI and the samples were analyzed by fluorescence microscopy. 

Immunohistochemistry on Flatmounted Retinas

For the immunohistochemical detection of GFAP and CD11b expression on flatmounted retinas, the protocol of Wang et al. ²¹ was performed with slight modifications. Briefly, the fixed retinas were washed for 1 h in PBS containing 0.2% Triton X-100, blocked in 3% BSA-PBS overnight at 4°C, and incubated with the antibodies against CD11b or GFAP diluted in 3% BSA-PBS for 48 h at 4°C. The negative controls were incubated with 3% BSA-PBS alone. Retinas were washed in PBS three times for 1 h each and incubated with the secondary antibodies indicated above overnight at 4°C. The nuclei were counterstained with DAPI and the retinas were analyzed by fluorescence microscopy. The quantification of the CD11b or GFAP positive cells was performed in four to five areas of 0.64 mm² on the flatmounted retinas (Openlab software; Improvision, Tübingen, Germany). 

Immunoblotting

Free-floating retinas (n = 8 for each treatment group) were homogenized in liquid nitrogen and lysed in 300 μL of ice-cold tissue lysis buffer (50 mM Tris-HCl [pH 7.4], 2 mM MgCl₂, 1% NP-40, 10% glycerol, 100 mM NaCl), and 1% protease inhibitor cocktail (Sigma-Aldrich, St. Louis, MO) added just before use. The lysates were cleared by centrifugation at 12,000g for 20 minutes at 4°C. The protein concentration of the supernatant was determined using the BCA assay (Pierce, Rockford, IL) according to the manufacturer’s instructions. Protein (50 μg) was run in 5% to 12% denaturing gels and transferred onto nitrocellulose membranes. Ponceau S staining (Sigma-Aldrich) was carried out to verify equal protein loading. Immunoblotting and signal detection by ECL were performed as described previously ²² by incubating the membranes with the primary antibodies against adducin (1:1000 in 5% milk powder/0.1% Tween-20/TBS; Abcam, Cambridge, UK), GFAP (1:2000; DAKO), phospho(S726)-adducin (1:500; Abcam), or phospho(T445)-adducin (1:200; kindly provided by Kozo Kaibuchi, Nagoya, Japan) overnight at 4°C followed by biotin conjugated secondary donkey anti-rabbit antibodies (1:1000; Jackson ImmunoResearch) and horseradish peroxidase-conjugated streptavidin (1:1000; Dianova) for 1 h at RT. The signal intensity was quantified (Quantity One software version 4.6.2; Bio-Rad, Munich, Germany). 

Membrane-Based Mouse Cytokine Array

The cytokine profile of the retinal conditioned medium was analyzed (RayBio Mouse Cytokine Antibody Array I; RayBiotech Inc., Norcross, GA) according to the manufacturer’s instructions, but with slight modifications. This array consists of membranes coupled with antibodies against 22 cytokines in duplicates as well as biotin-conjugated IgGs producing positive signals allowing the comparison of the relative expression levels among different membranes. The membranes were initially kept in the blocking buffer overnight at 4°C and incubated with the freshly collected conditioned media of retinas for 2 h at RT. After washing with two changes of the wash buffers supplied, the membranes were incubated with biotin-labeled anti-cytokine antibodies first at 4°C overnight, then at RT for an additional 2 h-period. This was followed by extensive washes and a 2 h-incubation with horseradish peroxidase conjugated streptavidin. Signal detection was performed by ECL as instructed and the spot intensity was measured (ImageQuant software; Molecular Dynamics, Sunnyvale, CA). 

Intravitreal Injections and Optic Nerve Crush

Female Brown Norway rats were anesthetized with an intraperitoneal injection of chloral hydrate (6 mL/kg body weight of a 7% solution). The eyes (n = 4 for 10 μM H-1152P, n = 6 for each of the other groups) were injected intravitreally with 2 μL of H-1152P (dissolved in PBS at 1, 10, and 100 μM) or PBS alone using a heat-pulled glass capillary connected to a microsyringe (Drummond Scientific, Broomall, PA) avoiding injury to the lens. Optic nerve crush was performed as follows. ²³ Briefly, the conjunctiva of the injected eyes was incised and the optic nerve was exposed by blunt dissection. A cross-action calibrated crush forceps was placed approximately 2 mm behind the globe and the nerve was partially crushed for 15 seconds. The rats were killed 7 days after ONC in a CO₂ chamber and immediately perfused by the transcardial injection of 250 mL PBS and 200 mL 4% PFA in PBS (pH 7.4). The eyes were then removed and processed for immunohistochemistry as described. 

Statistical Analysis

The differences between the groups were analyzed by one-way ANOVA and the comparison between two treatment groups was performed using the two-tailed, non-paired t-test assuming equal variance. P values < 0.05 were considered as significant. Data are presented as mean ± SE. 

Results

Optimal Concentration of H-1152P

To determine the optimal concentration of H-1152P for promoting cell survival, the organotypic cultures of mouse retinas were incubated with different concentrations of H-1152P (0 to 100 μM) for 24 h under serum deprivation. The damaged cells in the GCL were detected by staining with EthD-1, a fluorescent nucleic acid dye that cannot penetrate intact cell membranes. The ratio of the EthD-1 stained cells to the amount of total (DAPI-stained) cells was then calculated to determine the extent of cell damage. 

After 24-h serum deprivation, cells strongly positive for EthD-1 constituted 57.2% ± 9.2% of the total cell population in the GCL of the untreated retinas (n = 12 independent experiments). The majority of the cells with strong EthD-1 uptake also possessed condensed nuclei characteristic of apoptotic cells. Though the treatment with H-1152P at concentrations ranging from 1 to 100 nM slightly lowered the levels of EthD-1 uptake, as demonstrated by the decrease in the intensity of staining, no considerable difference was observed in the percentage of damaged cells (n = 4 experiments for 1 and 10 nM, n = 5 for 100 nM). However, the ratio of (EthD-1)+ cells was significantly reduced to 33.6% ± 8.2% in retinas incubated with 1 μM of H-1152P (n = 9, P < 0.000007). The addition of 20 μM H-1152P also resulted in a significant though more moderate protective effect (45% ± 10.9%, n = 8, P < 0.02), whereas increasing the concentration further to 100 μM led to a mild and insignificant decrease in cell damage (51.8% ± 8.2%, n = 4; Figs. 1A 1B ). 

In retinal damage, the activation of caspase-3 by proteolytic cleavage is an early intracellular event triggered by various apoptotic signals such as FasL, TNF-alpha, cytochrome-C, and elevated intracellular calcium. ²⁴ Based on this, immunostainings for cleaved caspase-3 were performed on retinal transverse cryosections to detect the extent of apoptosis in all the retinal layers. In untreated retinas (n = 4 independent experiments), the highest degree of immunoreactivity was observed in the GCL, with 72.7% ± 8.8% of the cells being strongly positive for cleaved caspase-3. However, the incubation with H-1152P (0.01 to 100 μM, n = 4 for each group) resulted in a significant reduction in the extent of apoptosis in the GCL at all the concentrations tested. Since the anti-apoptotic activity of the inhibitor was most significant and reproducible when administered at 1 μM (48.3% ± 2.7%, P < 0.002), we concluded that the optimal concentration of H-1152P under these culture conditions is 1 μM (Fig. 1C) . 

To determine the long-term effect of H-1152P at its optimal concentration, flatmounted retinas were incubated with or without 1 μM H-1152P for 48 h (n = 2 per group) and cell damage in the GCL was analyzed by EthD-1 and DAPI stainings. Serum deprivation for 48 h severely impaired cell survival in untreated retinas, as demonstrated by the increase in the number of (EthD-1)+ cells. This, together with the decrease in cell density as seen in the DAPI-staining, accounted for the rise in cell damage to 71.5% ± 6.1%. However, the incubation with 1 μM H-1152P significantly reduced this ratio to 40.8% ± 6.4% (P < 0.04), demonstrating the long-term protective effect of this treatment (Figs. 1D 1E) . Yet, owing to the difficulty to handle and process the untreated retinas for further immunohistochemical analysis after 48 h, we concentrated on the cellular changes occurring after 24 h. 

Decrease in the Reactivity of Retinal Glial Cells in Response to H-1152P

Considering the decisive role of the extent of reactive gliosis in the pathologic course of neuronal damage, we performed an immunohistochemical analysis of the GFAP on retinal cultures to detect the extent of astrocyte activation on the nerve fiber layer (NFL) after 24-h serum deprivation. Retinas fixed immediately after mounting served as controls reflecting the basal level of astrocytes at the starting conditions (T0). 

GFAP immunostaining at T0 was mainly localized to the spots injured during mounting, and appeared very weak in undamaged areas revealing a moderate population of astrocytes. However, a prominent network of reactive astrocytes strongly positive for GFAP was detected especially around the blood vessels of untreated retinas after 24-h serum deprivation. The astrocyte network was present in the retinas treated with 1 μM H-1152P, as well. However, the number of reactive astrocytes and the intensity of the GFAP-staining were considerably reduced close to the basal levels, demonstrating the impairment of astrocyte reactivity in response to H-1152P (Figs. 2A 2B) . 

Immunostainings of retinal transverse sections performed to detect the response of not only the astrocytes but also the Müller cells revealed only a few GFAP+ cells on the NFL and almost no reactive Müller cells at T0. However, the 24-h incubation in serum-free medium led to a dramatic upregulation of GFAP, as demonstrated by the strong staining on the NFL and the profound increase in GFAP+ Müller glia in untreated retinas. Yet, the coincubation with 1 μM H-1152P considerably reduced the number of both the activated Müller glia and the intensity of staining in the NFL (Fig. 2C) . 

Upregulation of GFAP in untreated retinas after 24-h serum deprivation and the H-1152P associated decrease in the levels of this protein were also confirmed by the immunoblot analysis of the corresponding retinal lysates. GFAP is a 49 kDa protein with several splice variants, which might be phosphorylated and glycosylated to varying degrees, resulting in several bands on Western blot analysis. ²⁵ Our results demonstrated the significant upregulation of all the detectable GFAP-isoforms in untreated retinas after 24 h in comparison to T0. Among these isoforms, a band of approximately 37 kDa, which could hardly be detected at T0, was of particular interest because it constituted the main portion of the GFAP positive bands in untreated retinas after 24 h. In contrast, treatment with 1 μM H-1152P led to a significant reduction of 32% to 47% in the intensity of all the detectable isoforms (P < 0.04, Figs. 2D 2E ). 

To characterize the response of the retinal microglia, we performed the immunohistochemical detection of CD11b, an integrin receptor upregulated in activated microglia, ²⁶ on freshly prepared retinal wholemounts as well as on retinas incubated for 24 h. In normal retinas, ramified microglia with long, thin, and bifurcated processes could be detected on the GCL. However, the number of (CD11b)+ cells and to some extent the intensity of the staining increased considerably after 24-h serum deprivation. The majority of the CD11b-positive cells in the control group also acquired an ameboid shape with retracted protrusions characteristic of reactive microglia. ²⁷ In contrast, the microglia in retinas incubated with 1 μM H-1152P possessed longer processes resembling the ramified morphology of quiescent microglia (Fig. 3) , demonstrating that the treatment with H-1152P also reduced the activation of the retinal microglia in vitro. 

Changes in the Cytokine Profile in Response to the ROCK-Inhibitor

Reactive glial cells are known to produce growth factors and immunomodulatory cytokines, the balance of which is critical for neuron survival. To determine whether the H-1152P-dependent decrease in glial cell reactivity also altered the release of these factors, we analyzed the cytokine profile of the retinal culture media using a membrane-based cytokine array, which allowed for the simultaneous detection of 22 mouse cytokines. Incubation of the retinas in the presence of 1 μM H-1152P led to a significant reduction in the amount of 11 of the cytokines that could be detected in the culture medium using this array. Among these, the cytokines interleukin (IL)-3, IL-6, and tumor necrosis factor (TNF)-alpha were reduced by 30% in a H-1152P dependent way whereas the levels of GM-CSF, IL-5, IL-9, IL-13, interferon-gamma, stem cell factor, soluble TNF receptor (STNFR)1, and thrombopoietin underwent an approximately 18% decrease (P < 0.05; Fig. 4 ; Table 1 ). 

H-1152P-Dependent Decrease in the Toxicity of Retinal Conditioned Media

Having observed an H-1152P–dependent reduction in the release of various proinflammatory cytokines, we next wanted to determine whether these secreted factors exerted toxicity on retinal cells. With this purpose in mind, whole-mounted retinas were incubated for 24 h with or without 1 μM H-1152P. The conditioned media were then collected and used immediately for the incubation of freshly prepared retinas. The percentage of damaged cells was then assessed by EthD-1 and DAPI stainings. 

The 24-h incubation of freshly prepared retinas (n = 3) with the conditioned medium from untreated retinas gave rise to a severe cell damage in the ganglion cell layer (71.6% ± 10.4%), which was significantly higher than that detected in retinas incubated in fresh culture medium (57.3% ± 9.2%, P < 0.05). When fresh H-1152P was added into this untreated retinal conditioned medium, a lower degree of cell damage was detected (57.7% ± 9.1%, n = 3). However, the conditioned medium of the retinas incubated with 1 μM of H-1152P resulted in a significantly lower cell damage (42.7% ± 4.4%, n = 3, P < 0.02; Fig. 5 ). These findings underline the possibility that H-1152P exerts its protective effect on retinal cells mainly by suppressing the release of certain toxic factors, and probably to a lesser extent by directly interfering with these molecules. 

Specificity of H-1152P Action in Mouse Retinal Cells

H-1152P suppresses the activity of ROCK mainly by occupying its ATP-binding domain and thus depriving this kinase of its source for the phosphorylation of its substrates. However, H-1152P can also inactivate protein kinase A (PKA) and protein kinase C (PKC) at approximately 600- to 1000-fold higher concentrations, owing to the homology between the ATP-binding domains of these kinases. ²⁸ To ascertain that the effects observed with 1 μM H-1152P reflect the outcome of ROCK-inhibition with minimal interference with the PKA/PKC pathways in the retina, we determined the levels of adducin and the extent of its phosphorylation at different residues by immunoblotting. Adducin is a cytoskeletal protein which can be phosphorylated by ROCK at the threonine (T)445 and T480 residues in the neck domain. In contrast, the serine (S)726, S713, and S662 residues in the MARCKS domain of alpha, beta, and gamma isoforms of adducin, respectively, are common phosphorylation sites for both PKA and PKC. ²⁹  

The total levels of adducin detected by using polyclonal antibodies recognizing the alpha (∼110 kDa) and gamma (∼83 kDa) but not the beta (97 kDa) isoforms were not found to be significantly altered in either treatment group. Additional bands of smaller molecular weights, which probably represent the degradation products of adducin, were also detected in both samples at a similar intensity, suggesting that the H-1152P treatment induces no significant difference in the total level of this protein after 24-h serum deprivation (Fig. 6A) . 

The PKA/PKC-dependent phosphorylation of adducin was analyzed using antibodies that recognize the phospho-S726/S713/S662 residues in the MARCKS domain of the three adducin isoforms. The signal intensity of these isoforms were also detected at similar levels in both treatment groups, demonstrating that the incubation with 1 μM H-1152P did not significantly interfere with the PKA/PKC-dependent phosphorylation of adducin (Figs. 6B 6D) . 

The ROCK-dependent phosphorylation of adducin was analyzed using a polyclonal antibody that was reported to recognize only the α-adducin isoform phosphorylated at T445 in the neck domain and the degradation products thereof. ³⁰ In untreated retinas a weak band of 110 kDa, presumably corresponding to this isoform, was indeed detectable. In addition, a stronger band of approximately 74 kDa, which might represent a caspase-3-dependent cleavage product of adducin ³¹ as well as several weaker bands of intermediate and smaller size, were also present. All these bands were found to be expressed at approximately 50% lower levels in retinas treated with 1 μM H-1152P (Figs. 6C 6E) . This decrease in the ROCK-dependent phosphorylation of adducin in H-1152P treated retinas without an accompanying decrease in the PKA/PKC-dependent phosphorylation or the total levels of adducin isoforms suggests that H-1152P administered at 1 μM specifically inhibits the ROCK activity without interfering with the PKA/PKC pathways in mouse retinas. 

H-1152P Reduces Apoptosis and Glial Cell Reactivity after ONC

The neuroprotective effect of H-1152P-mediated ROCK-inhibition on cultured retinas prompted us to analyze the outcomes of this treatment on retinal cells experiencing stress in vivo. For this purpose, rats were intravitreally injected with H-1152P (1 to 100 μM) or the vehicle alone (PBS) and subjected to ONC. The extent of apoptosis in the GCL and the degree of reactive gliosis was determined seven days after injury. 

The injury to the optic nerve resulted in a high level of apoptosis in rats that received PBS (n = 6 eyes), with 61.1% ± 9.2% of the cells in the GCL being positive for cleaved caspase-3. However, the injection of H-1152P significantly reduced the percentage of apoptotic cells at all the concentrations tested (n = 4 eyes for 10 nM, n = 6 eyes for 1 and 100 μM each), with 1 μM again exerting the most significant protective effect (39.1% ± 4.5% apoptotic cells, P < 0.0005; Figs. 7A 7B ). 

The high degree of apoptosis in rats that received PBS was accompanied by a notable glial cell reactivity, demonstrated by the strong GFAP expression in the NFL and in Müller cells as well as the abundance of CD11b positive microglia in the NFL and inner retina. However, similar to our in vitro observations, the injection of 1 μM H-1152P led to a considerable decrease in the reactivity of particularly the Müller cells and microglia after ONC (Figs. 7C 7D) . These findings therefore provide the first in vivo evidence for the protective effects of H-1152P treatment on the retina and demonstrate the response of the glial cells to this treatment. 

Discussion

The inhibition of ROCK-signaling emerges as an effective therapeutic approach for promoting the survival of retinal neurons and yielded promising results particularly on the RGCs in the past few years. However, the outcomes of ROCK inhibition on retinal cell fate in conjunction with glial cell reactivity are still unknown. In this study, we therefore analyzed the influence of this signaling molecule on retinal cell survival and reactive gliosis by using the specific ROCK-inhibitor H-1152P in both mouse organotypic retinal cultures under serum deprivation and after ONC. 

After 24-h serum deprivation, the extent of cell damage in the GCL of untreated retinas was found to be 57% and 73%, as determined by the EthD-1 stainings and cleaved-caspase three immunostainings, respectively (Figs. 1A 1B 1C) . The discrepancy in these results might have arisen due to the different phases of apoptosis the cells in a given region possibly go through. Since the membrane permeability is not dramatically altered in the early phases of apoptosis, ³² cells at this stage are likely to have little or no EthD-1 uptake even though there is detectable caspase-3 cleavage, accounting for the relatively low value for the extent of cell damage determined by the former method. Nevertheless, the administration of H-1152P resulted in a significant reduction in the amount of damaged cells detected by both methods under these stringent in vitro conditions, most effectively and reproducibly when administered at 1 μM. The lower concentrations of H-1152P (1 to 100 nM) did not promote a considerable recovery in terms of EthD-1 exclusion, though there was a significant reduction in caspase-3 cleavage with 10 nM of this inhibitor (cleaved caspase-3 immunostaining was not performed on retinas treated with 1 and 100 nM). This effect of 10 nM H-1152P was, however, prone to variability as reflected by the SEM (Fig. 1C) . Increasing the concentration further to 20 μM and 100 μM also resulted in a significant anti-apoptotic effect, but this effect became gradually less compared to 1 μM, possibly due to the interference of the inhibitor with other signaling cascades at higher concentrations, which we will refer to in more detail in the subsequent sections. 

The protective effect of 1 μM H-1152P, particularly on the axotomized ganglion cells and presumably the displaced amacrine cells in vitro, was associated with a marked reduction in the reactivity of astrocytes, Müller cells, and microglia. The H-1152P dependent decrease in reactive gliosis might have arisen in consequence to the general improvement in retinal cell survival. However, this ROCK-inhibitor might also have directly interfered with certain events in the glial cells, such as the regulation of the actin cytoskeleton. The RhoA-ROCK signaling is mainly involved in the regulation of stress fiber formation, which provides the cells with contractile strength. Consistently, the inhibition of this pathway interferes with cellular contraction and favors the formation of protrusions in diverse cell types. ^{11 12} Since the retraction of cellular processes and the acquisition of an ameboid morphology is the major characteristic feature of microglia reactivity, the H-1152P dependent reduction in this process might have arisen due to the impairment of contractility in these cells. 

The possible effects of H-1152P on the cytoskeleton of astrocytes and Müller cells are less apparent since the stellate shape of the astrocytes in the NFL appear conserved and the Müller cell morphology does not show distinct alterations when analyzed at this level. However, there is a marked reduction in the levels of the GFAP protein in these cells in response to H-1152P. The immunoblot analysis suggests that the GFAP upregulation in untreated retinas is confined particularly to the lower molecular weight isoforms (Figs. 2D 2E) , which are presumably generated by the phosphorylation of GFAP at the head domain and the subsequent cleavage of the full length protein. ³³ Such isoform specific differences have indeed been related to the motor neuron degeneration in mice ³³ and hyperphosphorylation of GFAP was observed in rats after transient cerebral ischemia ³⁴ and in Alzheimer’s disease patients. ²⁵ The H-1152P-dependent decrease in the reactivity of astrocytes and Müller cells might therefore be related to an interference with the post-translational modifications of GFAP, such as its phosphorylation at the head domain. This possibility, which remains to be investigated, gains further support from a recent study demonstrating the higher affinity of certain commercially available GFAP-antibodies to the phosphorylated form of GFAP. ³⁵  

Interestingly, the inhibition of ROCK in adult rats and cultured astrocytes was reported to be promoting astrocyte process outgrowth and stellation, favoring a reactive phenotype. ^{36 37 38} Though these findings may sound contradictory to our results demonstrating the anti-gliotic effect of ROCK-inhibition, it should be noted that most of these studies were conducted using Y-27,632 at micromolar concentrations. Y-27,632 is another synthetic ROCK-inhibitor, which can potently inactivate ROCK with a Ki value of 140 nM in cell-free assays. However, it is most commonly used at a concentration of 10 μM in vitro, ^{11 39} at which it can also inhibit protein kinase n (PRK2), a Rho/Rac-GTPase effector involved in the regulation of the actin cytoskeleton, with equal potency. ³⁹ The administration of Y-27,632 at higher concentrations (e.g., 50 μM), which may be attained by in vivo delivery, can also lead to the inhibition of other kinases including CamKII, PKC, cAMP-dependent kinase, and myosin light chain kinase (MLCK). ³⁹ Therefore, although the application of Y-27,632 proved to be an efficient means of modulating astrocyte morphology and yielded interesting results, it is likely that the effects of Y-27,632 treatment on astrocytes do not solely reflect the outcomes of ROCK-inhibition. 

The incubation of the retinal cultures with 1 μM H-1152P further resulted in a reduction in the release of various cytokines (Fig. 4 , Table 1 ), possibly due to the decrease in glial cell reactivity. Among these cytokines, TNF-alpha is a major proinflammatory molecule that is upregulated after ischemic or excitotoxic injuries in the CNS and a key inducer of RGC death. ^{40 41} TNF-alpha and IFN-gamma also activate the inducible nitric oxide (NO) synthase in glial cells, causing massive NO production which destroys the neighboring cells. Furthermore, IFN-gamma directs the synthesis of various cytokines. ^{42 43} Likewise, GM-CSF regulates microglial function and the cytokines IL-3 and IL-13 are involved in proinflammatory cascades. ^{43 44 45} H-1152P treatment also resulted in a decrease in the levels of the soluble TNF-alpha receptor, which promotes TNF-alpha production, ⁴⁶ and thrombopoietin, a hematopoietic growth factor with proapoptotic functions in the CNS. ⁴⁷ Interestingly, the pro-inflammatory cytokine IL-6, which exerts conflicting effects on RGC survival, ^{48 49 50} and the anti-inflammatory cytokines IL-5 and IL-13 ^{43 51} also underwent a significant reduction in response to H-1152P. Though it is complicated to draw a conclusion regarding the net outcome of this cytokine profile, it should be noted that the balance between the levels of various immunomodulators, rather than the individual molecules themselves is considered as a critical factor regulating the conversion of an initially neuroprotective immune response to a neurodegenerative reaction. ^{9 51}  

This view was further supported by the findings demonstrating the H-1152P-dependent decrease in the toxicity of the retinal conditioned media. The 24-h incubation of freshly prepared retinas with the conditioned medium from untreated retinas gave rise to a significant increase in the extent of cell damage from 57% to 72% (P < 0.05, Fig. 5 ). This effect could not be solely attributed to the immunomodulators released from the untreated retinas, since the 24-h incubation period was expected to result in the uptake of nutrients and the release of metabolic wastes, both of which decreasing the quality of the medium. However, the culture medium of the retinas incubated with H-1152P resulted in a significantly lower damage of approximately 43%, suggesting the presence of other toxic factors apart from the quality of the medium, which could be altered in the presence of H-1152P. Here, the inhibitor left in the conditioned medium might also have exerted direct neuroprotective effects, the extent of which remained unclear. Yet, since the addition of fresh H-1152P into the conditioned medium from untreated retinas appeared less effective in promoting recovery, H-1152P is likely to be exerting its protective effect mainly by reducing the amount of certain secreted molecules, and probably to a less extent by directly inhibiting their toxic activity. 

H-1152P is a very potent ROCK inhibitor that recently gained attention due to its beneficial effects on the relaxation of trabecular meshwork and ciliary muscle cells and the suppression of the wound healing activities of Tenon’s capsule fibroblasts. ^{22 52} This molecule binds the ATP-binding domain of ROCK with a Ki of 1.6 nM and shows considerably less affinity for PKA (Ki = 630nm), PKC (Ki = 9.270 μM), and MLCK (Ki = 10.1 μM) in cell-free assays. ²⁸ However, higher concentrations of H-1152P (0.1 to 10 μM) were required for the suppression of ROCK in various cell-based assays ^{22 28} and in our retinal cultures, possibly due to the competition with intracellular ATP present at the micromolar range. ¹¹ This, in turn, increases the risk of unintentionally targeting other proteins, with PKA and PKC being the most likely candidates owing to the relatively lower Ki values of H-1152P for these kinases. To elucidate this possibility, we analyzed the levels of adducin, a cytoskeletal protein that can be phosphorylated by ROCK, PKA, and PKC. Since the treatment with 1 μM H-1152P suppressed the phosphorylation of adducin at the domain recognized by ROCK but not PKA or PKC (Fig. 6) , we concluded that the effects observed reflected the outcomes of ROCK-inhibition. 

To analyze the outcomes of H-1152P treatment in retinas confronting stress in vivo, we performed ONC in rats, which is a reproducible, widely used, and well-characterized model of optic nerve trauma. ^{23 53 54} The intravitreal injection of H-1152P (1 to 100 μM) before ONC significantly reduced apoptosis in the GCL at all the concentrations tested, with 1 μM exerting the most protective effect. The neuroprotective activity of H-1152P at 1 μM achieved with a single application was also associated with a significant reduction in reactive gliosis. Considering that the vitreal volume of the rats used in this study is approximately 20 μL, the single injection of 2 μL of 1 μM H-1152P into the vitreous would theoretically result in an intravitreal concentration of approximately 91 nM. At this low concentration the inhibitor is unlikely to bind other kinases, which indicates the specificity and potency of H-1152P-mediated ROCK inhibition in vivo. 

Altogether, these data demonstrate the neuroprotective effect of H-1152P on retinal cells, particularly in the GCL, which was associated with a decrease in the reactivity of astrocytes, Müller cells, and microglia both in retinas cultured under serum deprivation and after ONC. Furthermore, the amount of immunomodulatory cytokines released from the cultured retinas underwent a considerable reduction, which likely affected the toxicity of the conditioned medium. These beneficial outcomes of H-1152P treatment might have arisen from the possible protective effects of ROCK-inhibition directly on the retinal neurons, which would then have led to a decrease in glial reactivity, or conversely, by primarily the suppression of excessive reactive gliosis which would in turn protect the remaining neurons from secondary injury. Alternatively, H-1152P might have targeted both the retinal neurons and the glial cells simultaneously and resulted in an additive outcome. Though it remains to be determined which of these possibilities underlie our findings, the safety of H-1152P and its beneficial effects on the retina under such adverse conditions are nevertheless noteworthy, considering the current interest in the application of this inhibitor at the anterior chamber for modulating the activities of different cells implicated in the course of glaucoma. These findings therefore not only provide more insight into the role of ROCK in retinal cells encountering stress but also highlight the efficacy of H-1152P treatment to promote retinal cell survival and attenuate reactive gliosis under unfavorable conditions. 

 Supported by Herbert Funke-Stiftung (AT) and Dr. Wolfbauer-Stiftung.

 Submitted for publication March 4, 2008; revised July 11 and August 1, 2008; accepted November 3, 2008.

 Disclosure: A. Tura, None; F. Schuettauf, None; P.P. Monnier, None; K.U. Bartz-Schmidt, None; S. Henke-Fahle, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

The optimal dose of H-1152P for promoting cell survival in retinas cultured under serum deprivation. (A) EthD-1 staining of the flatmounted retinas demonstrating the damaged cells in the GCL after 24 h. DAPI counterstaining of the nuclei was performed to determine the total number of cells in the fields analyzed. Note the condensed appearance of the nuclei suggestive of apoptosis in the majority of the cells exhibiting very strong EthD-1 staining. Bar = 25 μm. (B) Percentage of damaged cells in the GCL after 24 h. *P < 0.02; ***P < 0.000007 compared to control (n = 12 retinas for control (0 μM), n = 4 for 0.001, 0.01, and 100 μM, n = 5 for 0.1 μM, n = 9 for 1 μM, n = 8 for 20 μM H-1152P). (C) The anti-apoptotic effect of 1 μM H-1152P in the retina detected by cleaved caspase-3 immunostaining on retinal transverse sections. The percentage of cleaved caspase-3 positive cells in the GCL was quantified in 4 to 12 sections per retina (n = 4 retinas per group) and from approximately 4 fields of 0.14 mm² per section. *P < 0.02, **P < 0.004, ***P < 0.002 compared to control. (D) The long-term protective effect of 1 μM H-1152P in the GCL demonstrated by the considerably lower levels of EthD-1 uptake after 48 h-serum deprivation. (E) Percentage of cell damage after 48 h (n = 2 retinas per group, *P < 0.04).

Figure 2.

 

The reduction in astrocyte and Müller glia reactivity in response to 1 μM H-1152P. (A) GFAP immunostaining of the retinal wholemounts demonstrating the upregulation of GFAP+ astrocytes especially around retinal blood vessels (v) after 24 h in untreated controls (C), and the H-1152P (H) dependent reduction in both the number of reactive astrocytes and the intensity of staining close to the levels at time point 0 (T0). Bar = 100 μm. (B) Quantification of the GFAP+ astrocytes on the flatmounted retinas (n = 5 for T0 and C, n = 6 for H). (C) GFAP immunostaining of retinal transverse sections, showing GFAP upregulation not only in the nerve fiber layer (NFL), but also in Müller glia (arrows) in untreated controls, and the impairment of this process in response to H-1152P. (D) Representative immunoblot demonstrating the upregulation of all the detectable GFAP isoforms in untreated retinas (C) after 24 h-serum deprivation compared to the initial levels (T0), and the decrease in the intensity of these bands in response to 1 μM H-1152P (H). M, Molecular weight marker in kDa. Ponceau S staining was performed to verify equal protein loading. The arrow denotes the 49 kDa isoform whereas the block-arrow indicates the approximately 37 kDa band which was strongly upregulated in untreated retinas after 24 h. (E) Densitometric analysis of the intensity of the GFAP isoforms, normalized to the levels of total protein in each lane and expressed as fold increase with respect to the intensity at T0. The mean values were calculated from four independently performed blots. The H-1152P-associated decrease in the intensity compared to the control was significant for all the isoforms (P < 0.05).

Figure 3.

 

Representative images of the CD11b immunostaining on retinal flatmounts demonstrating (A) the basal levels of microglia in the NFL at time point 0 (T0) and (B) the activation of microglia after 24 h-serum deprivation (control). Note the retraction of the processes and the acquisition of an ameboid shape in the majority of the (CD11b)+ cells in this group. (C) Microglia in retinas treated with 1 μM H-1152P exhibited a more ramified morphology similar to the quiescent microglia detected at T0. Bar, (A–C) 25 μm. (D) Quantification of the (CD11b)+ cells (n = 4 retinas for T0, n = 5 retinas for the untreated and H-1152P treated groups).

Figure 4.

 

The H-1152P dependent decrease in cytokine release determined using a membrane-based cytokine array. Images demonstrate the cytokines (in duplicates) that were detected at significantly lower levels in the conditioned medium of retinas incubated with 1 μM H-1152P (+) compared with untreated controls (−). The positive controls on each membrane are shown to allow for the comparison of the spot intensity on different membranes. The mean intensity of the cytokine levels in response to H-1152P expressed as the percentage of the levels in untreated groups are denoted in Table 1 .

Table 1.

 

Levels of the Cytokines Released into the Culture Medium from Retinas Incubated with 1 μM H-1152P

Cytokine	% of Control (Mean ± SEM)
GCSF	92.24 ± 26.33
GMCSF	81.11 ± 9.42, ***
IL-2	80.03 ± 25.24
IL-3	66.67 ± 25.27, **
IL-4	97.89 ± 27.94
IL-5	81.22 ± 10.02, **
IL-6	68.89 ± 16.1, ***
IL-9	84.12 ± 16.31*
IL-10	86 ± 21.6
IL-12p40p70	83.85 ± 30.75
IL-12p70	86.24 ± 18.8
IL-13	74.04 ± 20.19*
IL-17	85.77 ± 18.57
IFN-γ	85.61 ± 7.71*
MCP-1	95.6 ± 5.03
MCP-5	94.69 ± 15.04
RANTES	90.89 ± 16.15
SCF	83.9 ± 12.33, **
sTNFR1	85.68 ± 13.23*
TNF-α	70.64 ± 18.89*
Thrombopoietin	83.74 ± 13.84*
VEGF	82.21 ± 20.22

 The intensity level of each cytokine was normalized to the levels of the positive controls (see Fig. 4 ). The mean values are calculated from 6 independent experiments (

*   P < 0.04,

**   P < 0.01,

***   P < 0.001). GCSF, granulocyte-colony stimulating factor; GMCSF, granulocyte-macrophage CSF; IFN-γ, interferon-gamma; IL, interleukin; MCP, monocyte chemoattractant protein; RANTES, regulated upon activation normal T-cell expressed and secreted; sTNFR1, soluble TNF-alpha receptor 1; TNF-α, tumor necrosis factor-alpha; VEGF, vascular endothelial growth factor.

Figure 5.

 

H-1152P associated decrease in the toxicity of retinal conditioned medium. Control denotes the extent of cell damage in untreated retinas incubated with serum-free DMEM/F-12 for 24 hours (see Fig. 1B ). Incubating freshly prepared retinal wholemounts with the conditioned medium (CM) of untreated retinas gave rise to a significant increase in the percentage of damaged cells. The addition of 1 μM H-1152P into this conditioned medium (CM+H) resulted in a lower level of damage. The extent of cell damage underwent a further reduction when the conditioned medium of retinas incubated with 1 μM H-1152P (CM(H)) for 24 hours was applied.

Figure 6.

 

H-1152P at 1 μM inhibits ROCK without interfering with PKA/PKC in mouse retina. Representative Western blot analysis demonstrating (A) the total levels of adducin isoforms; (B) the extent of the PKA/PKC-dependent phosphorylation of adducin at S726 in the MARCKS domain (p-(S726)-adducin); and (C) the levels of ROCK-dependent phosphorylation at T445 in the neck domain (p-(T445)-adducin) in retinas incubated with (+) or without (−) 1 μM H-1152P for 24 h. The arrow indicates the α-isoform of adducing, whereas the arrowhead denotes an approximately 74 kDa fragment that might represent a cleavage product of caspase-3. Since the putative 74 kDa product contains the neck domain but not the MARCKS domain, ³¹ it was not detected in the p-(S726)-adducin blots. The densitometric analysis of α-adducin phosphorylated at (D) S726 in the MARCKS domain or at (E) T445 in the neck domain confirms the distinct reduction in ROCK activity without a considerable change in PKA/PKC activity. The mean intensity values were calculated from three independently performed blots and normalized to the total protein levels (determined by the Ponceau S staining) in the corresponding lane. *P < 0.05, **P < 0.00002.

Figure 7.

 

H-1152P protects the cells in the GCL from apoptosis and attenuates glial cell reactivity after ONC. Adult rats were intravitreally injected with H-1152P or PBS (n = 4 eyes for 10 μM H-1152P, n = 6 eyes for each of the remaining groups) before ONC. The extent of apoptosis in the GCL was examined 7 days after injury via cleaved caspase-3 immunostainings (A, B), which demonstrate a high level of apoptosis in the GCL of rats injected with PBS and a significant decrease in cell damage in response to H-1152P particularly when administered at 1 μM. The quantification of the (cleaved caspase-3)+ cells in the GCL was performed on 4 to 8 sections that were interspaced by intervals of minimum 70 μm and from approximately 4 fields of 0.14 mm² per section. *P < 0.05, **P < 0.01, ***P < 0.0005 compared to control. (C) Representative images of the GFAP immunostainings. Note the high degree of reactivity in the NFL and in Müller cells (arrowheads) in the eyes injected with PBS and the significant reduction in this process in response to H-1152P. (D) CD11b immunostainings demonstrating the abundance of microglia (arrows) in the NFL, IPL, and INL in rats that received PBS injection before ONC. The number of the (CD11b)+ cells also underwent a considerable decrease in response to H-1152P.

The authors thank Michal Fiedorowicz for assisting with the optic nerve crush experiments and Kozo Kaibuchi for kindly providing us with the phospho-(T445) adducin antibody.