New Mouse Retinal Stroke Model Reveals Direction-Selective Circuit Damage Linked to Permanent Optokinetic Response Loss

Abstract

Purpose.: Ischemic insults give rise to severe visual deficits after blood vessel occlusion. In this study we investigated the effects of retinal stroke on the direction-selective circuit of the inner retina in a new adult mouse model.

Methods.: The inner retinal blood flow was interrupted for 60 minutes by ligating the ophthalmic arteries and veins in the optic nerve sheath. The optokinetic response (OKR) was measured to assess ischemia/reperfusion-mediated functional deficits and structural changes were studied by immunohistochemistry.

Results.: Ischemia/reperfusion induced reactive gliosis and degeneration of the inner retina. The OKR was almost completely abolished from 7 days after reperfusion, whereas approximately 40% of retinal ganglion cells were still alive. Ischemia led to severe degeneration of the processes of starburst amacrine cells (SAC), which cell bodies are in the ganglion cell layer (ON SACs), and to a lesser extent of the dendrites of SACs, which cell bodies are in the inner nuclear layer (OFF SACs). In addition, the elimination of retinal ganglion cells, direction-selective ganglion cells, and ON SACs was much greater at 10 days and 21 days than that of OFF SACs. After reperfusion, P-Stat3 was transiently activated in ganglion cells, whereas P-Erk1/2 signal was specifically detected in Müller glia.

Conclusions.: These results show a pronounced destruction of the ON direction-selective circuit in the inner retina that correlated with the irreversible loss of the OKR early after ischemia/reperfusion.

Introduction

Inner retinal ischemic insults cause severe visual impairments after blood vessel occlusion. Central retinal artery occlusion corresponds to a situation of complete retinal vessel occlusion of variable durations that can trigger sudden and profound vision loss. ^1,2 A clot formation in the carotid artery that hampers the regular blood flow to the retina is responsible for central retinal artery occlusion and has thus been considered as an ocular “stroke.” ^2,3  

In retinal stroke, blood supply interruption leads to irreversible vision deficits by killing neurons in the inner retinal layers, essentially composed of amacrine cells and retinal ganglion cells (RGCs). Several animal models have been developed to investigate the cellular and molecular mechanisms involved in ischemia-induced retinal defects (for reviews, see Refs. 4, 5). One of the most popular methods used in rodents consists of elevating the intraocular pressure (IOP). ^4,6 This IOP can be increased by implanting a catheter in the anterior chamber of the eye, connected to an elevated reservoir of saline. However, several drawbacks limit the use of this procedure in particular for the study of central retinal artery occlusion: the pressure elevation thus obtained not only induces ischemia in the inner retina but also interrupts the choroid and the uveal circulations, ^4,5 and the IOP increase can exert mechanical lesions that add to ischemic damage. As an alternative, the model of photothrombosis has been applied in different central nervous system (CNS) regions, including the retina. ⁷ However, this method produces free radicals resulting in microthrombi in retinal and choroidal vessels and leads to permanent and severe ischemia. ^4,5,8 The administration of the vasoconstrictor endothelin-1 also was used as an ischemic model, ⁹ but can generate variable ischemic responses. ⁴  

In the present study, we sought to characterize a new ocular stroke model mimicking the damage caused by central retinal artery occlusion in the adult mouse retina, the rodent species that is the most used in experimental ophthalmology. The ligature of the ophthalmic vessels appears to be the most appropriate technique to induce selective ischemic damage in the inner retina, similarly to what has been done in rats. ^10,11 In the present study, our results show that the optokinetic response (OKR), a visual test controlled by direction-selective circuits in the inner retina, was permanently abolished at 1 week post injury while approximately 40% of RGCs were still alive. Importantly, the dendritic layer of starburst amacrine cells (SACs), which cell bodies are in the ganglion cell layer (ON SACs) was more affected than that containing SAC processes which cells bodies are in the inner nuclear layer (INL) and correspond to OFF SACs. ^12–15 Moreover, we observed massive RGC and ON amacrine cell death in the innermost layer of the retina, whereas OFF amacrine cells showed a limited reduction in number in the INL. 

Material and Methods

Animals and Retinal Surgeries

All procedures were conducted in accordance with the regulations of the Veterinary Office of the Canton of Zurich and with the ARVO statement for the use of Animals in Ophthalmic and Vision Research. Two- to 3-month-old male C57/BL6 mice (Charles River, Sulzfeld, Germany) were used in this study for retinal injuries and for tissue analysis. 

The surgical procedure to perform mouse retinal ischemia-reperfusion (I/R) was adapted from Lafuente and colleagues ¹⁰ (Fig. 1A). Briefly, under general isoflurane anesthesia (2.5%) the shaved skin of the mouse skull was opened horizontally. In the periphery of the eye, the superior rectus muscle was attached with a 6.0 suture to pull forward the eye and to get access to the optic nerve. The optic nerve dural sheath was cut longitudinally and a 9.0 suture was inserted between the dural sheath and the optic nerve. A knot was tied around the sheath to constrict the central ophthalmic artery and vein. The interruption of the inner retinal blood flow was monitored by doing a fundus examination under an operating microscope with a glass coverslip applied onto the cornea. Care was taken not to damage the optic nerve. After 60 minutes, the suture was carefully removed to allow reperfusion and the blood circulation was again verified by fundus examination. Animals received analgesic (buprenorphin, 0.01–0.05 mg/kg body weight) subcutaneously to relieve pain. Mice showing no visible blood circulation or an incomplete recovery after 10 minutes were excluded from the analysis. During the whole procedure, mice were placed on a heating pad to maintain constant their body temperature, a parameter that was reported to strongly influence RGC death. ¹⁶  

Figure 1

 

Ischemia/reperfusion induces gliosis and inner retinal damage. (A) Experimental procedure for retinal I/R in mice. (B) Retinal blood vessels were labeled with GS-IB4 and astrocytes were examined by immunohistochemistry for GFAP on retinal flat-mounts in intact and 21-day ischemic retinae. Top- or side-view projections from 3D reconstructions of confocal pictures showed that I/R induced inner retinal layer collapse leading to a close proximity of the three vascular plexus (red; GS-IB4) and activated gliosis in astrocytes and in Müller cell radial processes (green; GFAP, arrowheads). Confocal pictures were taken in the temporal quadrant. Scale bars: (B) top view = 40 μm; side view = 50 μm.

For comparison, another group of mice was subjected to optic nerve crush injury. The optic nerve was exposed according to the same experimental procedure (n = 3–4 mice/group). A 9.0 suture was used to tie a knot at approximately 0.5 mm from the optic nerve head to completely constrict the nerve for 20 seconds by paying attention not to damage the ophthalmic artery running in the optic nerve sheath. The suture was then carefully released as previously described. ^17,18 Animals received analgesic (buprenorphin, 0.01–0.05 mg/kg body weight) subcutaneously to relieve pain. 

Virtual Optomotor System and Measurement of the OKR

The virtual optomotor apparatus (OptoMotry; Cerebral Mechanics, Middlesex, UK) described by Prusky et al. ¹⁹ was used to evaluate the spatial frequency threshold of freely moving intact and ischemic mice (Fig. 2A; n = 6 mice/group) at different time points from reperfusion. Briefly, individual mice were placed on a platform in a middle of an arena surrounded by four computer screens. Moving gratings of variable spatial frequencies were presented on the screens in the clockwise and counterclockwise direction, eliciting an independent temporal-to-nasal stimulation of the left and right eye respectively. Full-contrast gratings with increasing spatial frequencies were presented to the mouse to determine the spatial sensitivity threshold. Spatial frequencies ranged between 0.042 cycle/degree (c/d) to the maximal frequency that the animal was able to discriminate (∼0.4 c/d) and that corresponded to the baseline values before injury. 

Figure 2

 

Retinal ischemia leads to permanent and complete OKR abolition. (A) Schematic representation of the virtual optomotor apparatus used to evaluate the OKR spatial frequency threshold. (B) The spatial frequency threshold was measured in intact and ischemic eyes of the same mice (n = 6 mice/group) at different time points after reperfusion. One week after injury, the OKR was almost completely abolished in ischemic eyes but did not change in intact eyes. At 2 and 3 weeks post ischemia, no OKR could be recorded in ischemic eyes compared with intact contralateral eyes.

Retinal Immunohistochemistry

The mice were intracardially perfused with 4% paraformaldehyde (PFA). After removal of the cornea and the lens, eyecups were prepared for cryosections (14-μm-thick) or retinae were rapidly flat-mounted. After an additional postfixation in 4% PFA for 1 hour, eyecups were cryoprotected in 30% sucrose. Retinal flat-mounts were permeabilized in ice-cold methanol for 20 minutes at −20°C before immunohistochemistry. Neuronal survival was examined at 10 and 21 days after I/R and at 5 and 14 days after optic nerve crush. Primary antibodies (see Table) were diluted in the blocking solution composed of PBS containing 0.3% Triton-X-100, 5% bovine serum albumin (BSA), and 0.05% sodium azide. After extensive washing, flat-mounts or frozen sections were incubated with corresponding secondary antibodies at room temperature (RT) and mounted with Mowiol anti-fading medium [10% Mowiol 4–88 (wt/vol); Calbiochem, Cambridge, UK, in 100 mM Tris, pH 8.5, 25% glycerol (wt/vol), and 0.1% DABCO (1,4-diazabicyclo[2.2.2]octane)]. Stained sections were analyzed with a Leica SPE-II confocal microscope (Leica, Wetzlar, Germany) with a ×40 (NA 1.25) oil immersion objective. Image stacks were acquired with a step size of 0.5 μm and a resolution of 1024 × 1024 pixels (0.27 μm/pixel). Blood vessel and astrocyte reconstructions were obtained from image stacks that were exported to the Imaris Software (Bitplane, Zürich, Switzerland) to create three-dimensional (3D) projections. Snapshots of the top- and side-view projections were captured in the orthogonal mode. 

Table

 

Antibodies Used for Immunofluorescence (IF) and Western Blotting (WB)

Name	Type	Dilution IF	Dilution WB	Source	Catalog Number
Anti β3-tubulin	rAb	(C)1:500; (F)1:200		Abcam (Cambridge, UK)	ab18207
GS-IB4 Alexa-594	lectin	(C)1:50; (F)1:50		Invitrogen	I21413
Anti-GFAP	mAb	(C)1:500; (F)1:500		Sigma (Buchs, Switzerland)	G-3893
Anti-CART	rAb	(C)1:2000; (F)1:1000		Phoenix Pharmaceuticals (Karlsruhe, Germany)	H-003-62
Anti-ChAT	gAb	(C)1:200; (F)1:100		Millipore (Zug, Switzerland)	AB144P
Anti β3-tubulin	mAb	(C)1:1000; (F)1:500		Promega (Dubendorf, Switzerland)	G712A
Anti-Brn3a	mAb	(C)1:200		Santa Cruz (Heidelberg, Germany)	sc-8429
Anti-Brn3b	gAb	(C)1:200		Santa Cruz	sc-6026
Anti-melanopsin	rAb	(C)1:2500		Advanced Targeting Systems (San Diego, California, USA)	AB-N38
Anti-TH	mAb	(C)1:200		Millipore	MAB318
Anti-calretinin	mAb	(C)1:1000		Swant (Marly, Switzerland)	6B3
Anti-NF	mAb	(C)1:100; (F)1:100		Dako (Baar, Switzerland)	M0762
Anti-GS	mAb	(C)1:500		Millipore	MAB302
Anti-P-Erk1/2	rAb	(C)1:100	1:1000	Cell Signaling (Leiden, The Netherlands)	4370
Anti-Erk1/2	rAb		1:1000	Cell Signaling	4695
Anti-P-Stat3	rAb	(C)1:100	1:500	Cell Signaling	9131
Anti-Stat3	rAb		1:1000	Cell Signaling	9132
Anti-P-Akt	rAb		1:1000	Cell Signaling	9275
Anti-Akt	rAb		1:1000	Cell Signaling	9272
Anti-CNTF	rAb		1:2000	Abcam	ab46172
Anti-GAPDH	mAb		1:20000	Abcam	ab8245
Anti-pan NaCh	mAb	(C)1:50-1:100		Sigma	A3562
Anti-paranodin	rAb	(C)1:500		Generous gift from J.-A. Girault, PhD, Paris, France	Not applicable

 (C), dilution used for retinal cryosections; (F), dilution used for retinal flat-mounts; gAb, antibody raised in goat; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; GS, glutamine synthetase; mAb, antibody raised in mice; NaCh, voltage gated sodium channel; NF, human neurofilament protein; rAb, antibody raised in rabbit.

Neuronal Survival and Cell Density

To evaluate neuronal survival, β3-tubulin–stained RGCs were imaged in the four quadrants of the retina with a Leica SPE-II confocal microscope at ×40 (NA 1.25), using a step size of 0.5 μm and a resolution of 1024 × 1024 pixels (0.27 μm/pixel). The number of RGC cell bodies was quantified in eight regions of 62,500 μm² at 1.0 mm and 1.5 mm from the optic disk, distributed in the four quadrants. The density of RGCs per square millimeter was calculated in the whole retina or in each quadrant. The same counting method was applied to determine SAC and cocaine- and amphetamine-regulated transcript (CART)-labeled direction-selective retinal ganglion cell (DSGC) densities. 

Paranodin/Sodium Channel Labeling

For paranodin/Caspr and sodium channel (NaCh) double staining in the optic nerve, fresh-frozen sections (20-μm thick) were cut immediately after mouse decapitation and placed at −80°C (n = 3 mice/group). Slides were fixed in ice-cold absolute methanol for 20 minutes at −20°C and washed in PBS. After blocking, mouse anti-pan NaCh and rabbit anti-paranodin (see Table) were applied on slices and left overnight at 4°C. To detect NaCh staining, a horse anti-mouse biotin-coupled antibody (1:100; Vector Laboratories, Peterborough, UK) was added for 2 hours at room temperature and a third incubation was performed with streptavidin linked to Alexa 488 (1:400; Invitrogen, Zug, Switzerland). A goat anti-rabbit antibody (1:1000; Invitrogen) was used to detect paranodin. Confocal pictures were taken with a Leica SPE-II microscope in the rostral part of the optic nerve. 

Western Blot Analysis

At 1 and 5 days after I/R, three mice per group were killed by cervical dislocation and retinae were quickly dissected in an Eppendorf tube and snap frozen in liquid nitrogen. Protein lysate and dosage was performed as previously described. ²⁰ Retinal proteins (20 μg/well) were resolved by electrophoresis on a 4% to 12% gradient polyacrylamide gel and transferred to nitrocellulose membranes. Membranes were preincubated in a blocking solution of 5% BSA dissolved in TBST (Tris-base 0.1 M, 0.2% Tween-20, pH 7.4) for 1 hour at room temperature, incubated with primary antibodies (see Table) overnight at 4°C and after washing, with a horseradish peroxidase–conjugated anti-mouse or anti-rabbit antibody (1:15,000; Pierce Biotechnology, Wohlen, Switzerland). Protein bands were detected by adding SuperSignal West Pico Chemiluminescent Substrate (Pierce Biotechnology) and after exposure of the blot in a Stella detector (Raytest, Zurich, Switzerland). Densitometric quantifications were done using the ImageJ software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA). 

Data Analysis

Bar graphs present mean ± SEM. Statistical analyses were performed by using a one-way ANOVA followed by the Tukey post hoc test (GraphPad Software, Prism 5; San Diego, California, USA). 

Results

Ischemia-Reperfusion Selectively Induces Inner Retinal Damage and Gliosis in Adult Mice

Ischemia-reperfusion in the inner retina of adult mice was carried out by ligating the optic nerve sheath for 60 minutes and by paying attention not to injure the optic nerve (Fig. 1A). The blood flow interruption and the reperfusion of the inner retina were controlled by fundus examination. Blood vessels were labeled with Griffonia simplicifolia isolectin B4 (GS-IB4) coupled to Alexa 594 while the glia-specific marker glial fibrillary acidic protein (GFAP) was detected by immunohistochemistry so as to investigate the possible effects of I/R on the vasculature and on gliosis in retinal flat-mounts (Fig. 1B). Compared with the intact control retina in which the three vascular plexus appeared clearly separated, 21-day ischemic retinae showed collapsed inner retinal layers, resulting in closer blood vessel plexus proximity (Fig. 1B, side view). This decrease in the thickness of the inner retina was probably due to I/R-mediated RGC and amacrine cell death. Blood vessels appeared intact, except for a few degenerated, acellular capillaries that also could be observed (Fig. 1B, arrow in the top-view projection). ²¹ In the astrocyte network covering blood vessels and RGC axons, the GFAP immunofluorescent signal was dramatically increased after ischemia (Fig. 1B, right column, top view). In addition, 3D reconstructions of the inner retina allowed observation of the strong upregulation of GFAP in Müller cell processes, suggesting that astrocytes and Müller cell gliosis persist a long time after retinal reperfusion (see arrowheads, Fig. 1B, side view). 

Ischemia-Reperfusion Leads to Complete Abolition of the OKR

A longitudinal follow-up of visual function changes was performed using the virtual optomotor system developed by Prusky and colleagues ¹⁹ (Fig. 2A). The rotation of the virtual cylinder elicited mouse head tracking in the same direction by reflexive neck movements defined as the OKR. The OKR threshold was determined by increasing the spatial frequency of gratings in intact animals and mice subjected to I/R. One week after ischemia, the OKR was almost completely abolished in the left ischemic eye compared with the right intact eye which threshold sensitivity remained unchanged (Fig. 2B). At 2 and 3 weeks after ischemia, no OKR could be recorded from the left ischemic eye compared with the normal contralateral eye response. These data suggest that 60 minutes of ischemia is sufficient to cause rapid and permanent loss of OKR. 

Ischemia-Reperfusion Alters Dendritic Connectivity in the Retina

The OKR depends on the activation of DSGCs that are sensitive to object motion. ²² The firing of DSGCs in response to specific direction stimulus depends on asymmetric GABAergic inputs from cholinergic SACs. ^23,24 Moreover, selective ablation of ON and OFF SACs has been shown to abolish the directional selectivity of ganglion cell responses and disrupted the OKR. ²⁵ ON and OFF SACs form synaptic contacts with DSGCs in the sublamina 4 (S4) and S2 of the inner plexiform layer (IPL), respectively. To determine the effects of I/R on the SAC-DSGC circuit connectivity, SACs and ON-OFF DSGCs were labeled by immunohistochemistry for acetylcholine transferase (ChAT) and CART, respectively, on cryosections (Fig. 3A). The cell bodies of OFF SACs were detected in the INL while ON SACs were localized in the ganglion cell layer (GCL) (arrowheads) ²³ (Fig. 3A). The two separate laminae containing OFF and ON SAC dendritic processes appeared in S2 and S4 layers of the IPL, as expected. In the intact retina, CART-positive cells colocalized with β3-tubulin, a specific marker for the whole RGC population (Fig. 3A, white stars). Previous studies reported that approximately 15% of RGCs expressed CART in the mouse retina, but some CART-positive non-SACs also were detected in the INL. ²⁶ The DSGC dendrites cofasciculated with the SAC extensions in S2 and S4 sublaminae of the IPL (Fig. 3A). ^22,23 The number of ChAT-positive SACs was markedly decreased in the GCL 5 days after I/R (arrowheads) and the ChAT-immunopositive ON-dendritic layer was strikingly weaker in the IPL (white arrow) than that of the OFF stratum (S2) (Fig. 3A). At the same time point, the CART staining intensity and the number of CART-positive cells were decreased (stars). The dendritic stratification of CART-positive cells appeared diffuse compared with the control situation. 

Figure 3

 

Ischemia/reperfusion induced severe dendritic damage in the inner retinal layers. (A) Identification of DSGCs and SACs on retinal cryosections in intact retinae and in ischemic retinae at 5 days after reperfusion. Blue: CART. Red: ChAT. Green: β3-tubulin. Five days after reperfusion, ischemic retinae showed a reduction in the number of CART-labeled DSGCs (stars) and fewer ChAT-positive SACs, mainly in the GCL (arrowheads). A weaker staining of the ON layer could also be noticed (arrows). (1, 2) Magnified pictures from (A) (×2.5). (B) Detection of RGC subtypes on retinal cryosections in intact and ischemic retinae at 5 days after reperfusion with specific markers: Brn3a (green), Brn3b (red), melanopsin (blue). (C) Calretinin-positive AII amacrine cell neurites (green) were damaged at 5 days post ischemia (arrows). (D) No obvious difference in the S1 dendritic lamina could be noticed after TH labeling for dopaminergic amacrine cells in I/R retinae. (E) TH-positive dopaminergic amacrine cells colocalized with CART and were preserved after retinal stroke. ONH, optic nerve head; OPL, outer plexiform layer; FL, fiber layer. Scale bars: 100 μm (A); 50 μm (B–E).

RGC-specific markers, such as Brn3a, Brn3b, and melanopsin, were detected by immunohistochemistry in the intact retina and at 5 days after I/R (Fig. 3B). The intensity of Brn3a (stars) and Brn3b (arrowheads) immunostainings was markedly decreased after ischemia, whereas the signal for melanopsin remained high, as in intact retinae (Fig. 3B). RGC protein expression changes suggest a differential susceptibility of RGC subpopulations to ischemic insults. 

The dendritic processes of the calretinin-positive AII amacrine cells (Fig. 3C) and of the tyrosine hydroxylase (TH)-positive dopaminergic amacrine cells (Fig. 3D) were also examined by immunohistochemistry. AII amacrine cell neurites were partially damaged; two of the three sublaminae subsisted 5 days after I/R, probably as a result of the synapse elimination in S3 apposed to ON SAC terminals in S4. In addition, no obvious difference could be noticed after I/R in the S1 sublamina of the IPL containing dopaminergic amacrine cell terminals (Fig. 3D). TH-positive cells were not lost and kept expressing CART after I/R (Fig. 3E). Together, these data suggest that I/R provokes severe but also differential dendritic loss of amacrine cells, appearing more pronounced in S3 and S4 containing DSGC-ON SAC synaptic connections. 

Ischemia-Reperfusion Disturbs Nodes of Ranvier Protein Expression in the Optic Nerve

Hypoperfusion has previously been shown to alter the distribution of nodal and paranodal proteins in different myelinated CNS regions, including the optic tract. ²⁷ The nodes of Ranvier are crucial for the rapid conduction of action potentials in long nerves of the CNS, such as the optic nerve. ^28,29 Typically, immunohistochemical stainings showed triplet pattern at the nodes of Ranvier composed of a very high density of voltage-gated sodium channels flanked by paranodin/Caspr clusters in paranodal segments. ^30,31 We then wondered if I/R affected the expression of paranodin/Caspr and voltage-gated sodium channels in optic nodes of Ranvier, a mechanism that may impair action potential conduction and could contribute to the abolition of the OKR that we observed (Fig. 4A). The labeling of the paranodin/sodium channel revealed a pronounced decrease of paranodin/Caspr and voltage-gated sodium channel expression 10 days after I/R. Interestingly, the integrity of some optic nerve and retinal axons seemed preserved after ischemic injury (Fig. 4B). Therefore, the molecular disorganization of ion channels at the nodes of Ranvier may be due to cell death, protein downregulation, or cluster disassembly that could participate in the quick and durable OKR deficits due to retinal stroke. 

Figure 4

 

Ischemia/reperfusion led to the molecular disruption of optic nerve nodes of Ranvier. (A) Immunofluorescent analysis of optic nerve nodes of Ranvier revealed a marked decrease in paranodin (red)/sodium channel (green) expression at 10 days post ischemia. (1′, 2′) Magnified pictures from (A) (×5). (B) The labeling of optic nerve axons (left) and of RGC axons in the retina (right) with an antibody directed against neurofilament (NF) proteins showed that some axons were preserved after I/R. The intensity of the staining was stronger after retinal injury. Scale bars: 50 μm (A). Magnification (A) = 10 μm; (B) optic nerve = 50 μm; retina = 100 μm.

Cocaine and Amphetamine Regulated Transcript-Labeled DSGCs and ON SACs Are Highly Sensitive to I/R

To evaluate the loss of RGCs, DSGCs, and ON and OFF SACs, immunohistochemistry was performed on retinal flat-mounts in intact mice and 10 and 21 days after I/R (Figs. 5A, 5B). In this aim, the same specific markers as in Figure 3 were used to identify the different retinal cell populations (Fig. 5A). The density of β3-tubulin–labeled RGCs was decreased to 41% and 26% at 10 and 21 days after I/R, respectively (Fig. 5B). The population of CART-labeled DSGCs was even more severely diminished with only 21% and 14% of remaining cells, respectively, at 10 days and 21 days. Strikingly, the density of ON SACs was more affected than the density of OFF SACs at the two time points, suggesting that these latter cells may be more resistant to ischemic insults. Quantitatively, 32% of ON SAC density survived at 10 days after I/R, whereas 76% of OFF SACs were still alive. Similarly, at 21 days after I/R, the density of ON SACs was diminished to 23%, whereas the OFF SAC density was only reduced to 52%. The density of each cell population also was evaluated in each individual retinal quadrant (Fig. 5C). RGCs tended to be better preserved in the temporal quadrant, although DSGCs and ON or OFF SACs seemed to be similarly affected in the four retinal quadrants (Fig. 5C). Overall, these data show that CART-expressing DSGCs and ON SACs are more vulnerable to I/R than the rest of RGCs and OFF SACs. 

Figure 5

 

Ischemia/reperfusion causes severe degeneration of cocaine and amphetamine-regulated transcript-positive direction-selective ganglion cells and of ON SACs. (A) Representative immunohistochemical stainings from intact and ischemic retinae (10 days after reperfusion) showing the distribution of CART-positive DSGCs (blue), ChAT-positive ON and OFF SACs (red), and β3-tubulin–labeled RGCs (green) on retinal flat-mounts. (A, B) Note the reduction in densities of RGCs, CART-labeled DSGCs, and ChAT-positive ON SACs (in the GCL) at 10 and 21 days after reperfusion. In contrast, ChAT-positive OFF SACs (in the INL) were more resistant to ischemic insult. The number written in each bar of the graph indicates the number of mice used for this experiment. (C) Quantification of cell densities in the four different retinal quadrants (Inf, inferior; Nas, nasal; Sup, superior; Temp, temporal). Statistical analysis was performed with a one-way ANOVA followed by a post hoc Tukey test comparing ischemic groups with the intact group (***P < 0.001). Scale bar: (A) = 100 μm.

Optic Nerve Crush-Mediated RGC Apoptosis Does Not Affect SAC Survival

We hypothesized that the expression of CART by RGCs may be required to maintain SACs alive; the downregulation of CART expression that we observed after I/R may lead to preferential ON SAC elimination in the GCL, whereas OFF SAC survival may be maintained by the presence of CART-expressing cells in the INL. Indeed, recombinant CART peptide administration in the brain has been found to prevent neuronal cell death after stroke. ³² To address the possible trophic dependence between ON SACs and CART-positive RGCs, we examined the survival of ON and OFF SACs after the targeted optic nerve crush–induced RGC cell death. The densities of β3-tubulin–expressing RGCs, CART-stained DSGCs, and ChAT-positive SACs were analyzed 5 days after optic nerve crush, right before RGC death, and 14 days after injury when most of the RGCs have disappeared (Fig. 6A). The density of surviving RGCs did not change at 5 days but was diminished to 22% at 14 days after optic nerve crush compared with intact retinae (Figs. 6B, 6C), consistent with previous studies in rodents. ^17,33,34 Interestingly, the density of CART-expressing DSGC was drastically lowered as early as 5 days post lesion (to 33%) and was even more reduced at 14 days (to 5%). In stark contrast, the number of ON and OFF SACs stayed unchanged at 5 and 10 days post injury (Figs. 6B, 6C). These data demonstrate that the massive and fast loss of CART-expressing RGCs can occur without affecting SAC survival. 

Figure 6

 

Targeted elimination of CART-positive DSGCs does not affect ON and OFF SAC survival. To evaluate the impact of their cell death on SAC survival, RGCs were selectively killed by optic nerve crush injury. (A) Representative immunohistochemical stainings in intact and axotomized (5 days) retinae. The same markers as in Figure 4 were used. (A, B) Optic nerve crush lesion induced massive degeneration of RGCs and CART-labeled DSGCs, whereas the density of ON and OFF SACs did not change. The number written in each bar of the graph indicates the number of mice used for this quantification. (C) Quantification of cell densities in the four different retinal quadrants. Statistical analysis was performed with a one-way ANOVA followed by a post hoc Tukey test comparing animals from the “ONC” group with the intact group (***P < 0.001). ONC, optic nerve crush. Scale bar: 100 μm (A).

Intracellular Signaling Pathways Activated by I/R in Mouse Retinal Cells

To clarify the molecular mechanisms activated in ischemic retinal cells, we analyzed the levels of phosphorylated signal transducer and activator of transcription (Stat)3, extracellular signal-regulated kinase (Erk)1/2, protein kinase B (Akt), and those of ciliary neurotrophic factor (CNTF) (Table) in retinal lysates by Western blotting 1 and 5 days after reperfusion. One day after ischemic injury, P-Stat3 was transiently upregulated in retinal lysates, whereas P-Akt levels stayed unchanged (Fig. 7A). Interestingly, the level of P-Erk1 was also increased at 1 day post ischemia (∼3-fold), whereas the expression of P-Erk2 did not vary either at 1 or 5 days. After 5 days, both P-Stat3 and P-Erk1 levels were decreased relative to 1 day but stayed significantly higher than in control mice. At the same time point, however, P-Akt level was significantly increased. The level of inflammatory cytokine CNTF was significantly upregulated at 5 days compared with uninjured retinal samples (∼4-fold). The P-Stat3 and P-Erk1/2 expression changes were localized by immunohistochemistry on retinal sections (Figs. 7B–D). After 1 day of reperfusion, P-Stat3 upregulation could be detected in β3-tubulin–positive RGCs (Fig. 7B). At 1 and 5 days, a robust enhancement of the P-Erk1/2 fluorescent signal was detected in the INL and in radial cells processes (Figs. 7C, 7D). P-Erk1/2 did not colocalize with the β3-tubulin RGC marker (Fig. 7C) but was expressed in Müller cell radial processes (arrows) and endfeet (arrowheads) stained for glutamine synthetase (Fig. 7D). Our data suggest that transient Stat3 signaling activation in RGCs may not be sufficient to protect these cells from cell death, ³⁵ whereas P-Erk1 activation in Müller glial may regulate gliosis after I/R. ³⁶  

Figure 7

 

Intracellular signaling pathways activated after retinal I/R. (A) The activation of the Jak/Stat3, Erk1/2, and PI3K/Akt signaling cascades was monitored by Western blot analysis, 1 and 5 days after I/R (n = 3 mice/group). The levels of P-Stat3 and P-Erk1 proteins were significantly increased 1 day after I/R. P-Akt and CNTF were upregulated at 5 days post ischemia. Ischemia/reperfusion injury also influenced the levels of total Stat3 and total Akt. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as a loading control. Statistical analysis was performed with a one-way ANOVA followed by a post hoc Tukey test (*P < 0.5; **P < 0.01; ***P < 0.001). Right column: values were obtained by densitometry and normalized to total, unphosphorylated proteins. Values for intact mice were set to 1. (B–D) Cellular localization of P-Stat3 and P-Erk1/2 in control retinae and in retinae after ischemic injury. (B) P-Stat3 upregulation was detected in RGCs labeled with β3-tubulin antibody at 1 day. (C) Fluorescent microscopy revealed that P-Erk1/2 signal was not colocalized with β3-tubulin–positive neuronal cells 1 day post injury. (Inset) Magnified pictures (×3). (D) In contrast, P-Erk1/2 signal was found in the radial processes (arrows) and the endfeet (arrowheads) of Müller cells identified with the specific glutamine synthetase (GS) marker. ONL, outer nuclear layer. Scale bars: 50 μm (B), 100 μm (C, D).

Discussion

In the current study, we reported that retinal ischemia led to virtually complete OKR abolition at 1 week after reperfusion, whereas approximately 40% of RGCs survived to the injury. Our data also revealed that the DSGC-ON SAC circuit was severely damaged in the mouse inner retina; the loss of CART-labeled RGCs and ON SACs was stronger than that of OFF SACs, and synaptic contacts in sublaminae S3-S4 presented a greater susceptibility to I/R-induced damage than the OFF SAC dendritic layer, near the INL. 

By ligating the arteries and veins embedded in the optic nerve sheath and without damaging the optic nerve itself, we wished to generate a pure I/R model in mice similar to the ischemic conditions encountered in central retinal artery occlusion. This is in contrast with the pressure elevation model induced by anterior chamber cannulation where the IOP increase might be accompanied with mechanical axonal damage. ³⁷ Using the same procedure as us in rats, Lafuente and coworkers ¹⁰ reported that after a 60-minute ischemia, approximately 40% of RGCs survived at 14 days after reperfusion, a value that is in the same range as what we found here after 10 days. By selectively ligating the mouse optic nerve sheath, we obtained very different effects on the survival of RGCs and amacrine cells when compared with optic nerve crush (Figs. 5, 6), suggesting that I/R on one side and optic nerve trauma on the other side activate distinct molecular mechanisms of cell death. 

The loss of OKR that we reported after 1 week of reperfusion did not correlate with the rate of RGC survival. This observation is consistent with another study that showed that IOP increase-mediated transient ischemia caused an approximately 50% decrease in OKR, whereas only approximately 16% of these cells in the GCL died 1 day after reperfusion. ³⁸ Other complementary functional tests could be used to assess deficits in other retinal cell populations, including other RGC types that survived after ischemia. For example, previous studies reported that the amplitudes of the oscillatory potentials of the ERG, for which amacrine cells are one of the main players, ^39,40 varied along with the duration of ocular ischemia. ^41,42 Cells from the inner retina (i.e., bipolar cells and Müller glial cells), are the main contributors of the ERG b-wave, suggesting that the measurement of the b-wave amplitude also could be an indicator of the physiological state of the inner retina after I/R. ^43–45 Alterations of the a- and b-wave amplitudes of the ERG have been shown in a model of transient (90-minute) ligature of ophthalmic vessels in rats. ⁴⁶ Moreover, visual evoked potential recordings in the superior colliculus or in the visual cortex could be useful to evaluate the decrease of retinal inputs to visual centers after injury. ⁴⁷ Indeed, surviving RGCs may not be able to transmit action potentials along their axons in the optic nerve and optic tract in which the anterograde and retrograde transports were shown to be affected after I/R. ^48,49 Further investigation will be needed to clarify these aspects in our ischemic mouse model. 

Early alterations in the motion-sensitive circuit of the retina are likely to underlie the rapid and dramatic loss of OKR that we observed after 1 week. Normal OKR requires the integrity of the connectivity between SACs and DSGCs. ²⁵ We found that terminal destructions between DSGCs and SACs preferentially occurred in the ON sublamina (S4) and was also obvious for CART RGCs in the plexiform layer. Rapid contact loss may also involve RGC dendritic arbor shrinkage, as shown in another model of ischemia in mice ⁵⁰ and in glaucoma models ^51–53 to which a decrease in spatial and temporal RGC response was associated. ⁵⁴ However, in contrast with what we observed in ischemic retinae, glaucomatous mice displayed important reduction in ChAT-positive cells in the INL as well as in the GCL. ^55,56 Recently, Rodger et al. ⁵⁷ observed that brain-derived neurotrophic factor (BDNF) delivery by gene therapy enhanced the dendritic field size in axotomized RGCs. It would then be interesting to see if, in our model, RGC dendritic length could be increased by BDNF so as to restore the OKR and motion perception. To our knowledge, our study is the first to report specific impairments of the DSGC-ON SAC system after ischemic injury. 

We found that ON SACs and RGCs were more susceptible to ischemic insults than OFF SACs. A possible molecular mechanism underlying the preferential RGC and ON SAC death may depend on the release of toxic cytokines, such as TNF-α and the robust activation of its cell death receptor, TNF-R1, in the GCL. After transient IOP increase-induced I/R, TNF-α and TNF-R1 were strongly upregulated in the RGC layer, whereas cells in the INL showed only TNF-R1 increase. ⁵⁸ It is therefore plausible that the elevation of TNF-α drives massive cell elimination in the GCL and is less efficient at killing OFF SACs in the more distal INL cells. In addition, the cell death encountered in glaucoma may occur in a TNF-α/TNF-R1–dependent manner ^59,60 and may involve the secretion of TNF-α by microglial cells and Müller cells. ^61–64 Whether more reactive oxygen species are produced in the GCL than in the INL where OFF SACs lie is not known. This question deserves further investigations to clarify the mechanisms of cell death in retinal stroke. 

Our data revealed a striking loss of CART, Brn3a, and Brn3b signals after ischemia. At this stage, it is not clear whether the loss of RGCs expressing these markers was due to cell death or to protein downregulation. For example, expression of the RGC-specific transcription factor Brn3a decreases in response to axonal lesion while axotomized RGCs are still alive. ³⁴ However, Brn3a appeared to be more sustained in surviving RGCs after optic nerve injury than Brn3b. ⁶⁵ This is in contrast with β3-tubulin, which is more stable and is considered as a reliable marker to quantify surviving RGCs. ⁶⁶ In addition to possible protein expression changes that may differ between Brn3a and Brn3b for instance, the susceptibility of RGCs to ischemic damage may vary between subpopulations; for example, it has been shown that photosensitive RGCs expressing melanopsin better survived to ischemic lesion or axotomy than the rest of RGCs. ^67,68  

A possible trophic relationship between CART-labeled RGCs and ON SACs has been addressed using the optic nerve crush paradigm, an injury type that results in selective RGC apoptosis. Based on our results, the expression of CART does not seem to be necessary to sustain ON SAC survival. In other brain regions, the injection of recombinant CART peptide (55-102) improved neuronal survival in rat stroke by upregulating BDNF expression. ³² The activation of endogenous survival mechanisms, including the transient upregulation of BDNF, ^69,70 failed to block rapid deleterious effects of ischemia. Similarly, the transient Stat3 phosphorylation observed in some RGC bodies was not protective after ischemia. However, a more sustained Stat3 overexpression could protect RGCs from cell death after ischemia. ³⁵ In addition, our results revealed a selective phosphorylation of Erk1 compared with Erk2, suggesting that P-Erk1 but not P-Erk2 was increased in Müller glia. The upregulation of Erk1 in Müller cells may be due to BDNF synthesis ⁶⁹ or to glutamate receptor activation. ^71,72 Further investigations would be needed to clarify the exact role of Erk1 activation in Müller glial cells in which it could be associated with the activation of transcription factors, ^71,73 promote gliosis, ³⁶ or stimulate the expression of cell-cycle genes, ⁷⁴ for example. 

In summary, our data revealed stronger cell damage in the ON pathway of the inner retina direction-selective circuit correlating with complete OKR loss in a new mouse stroke model. The important cellular damage in the ON SAC/DSGC layer inflicted by ischemia compared with the moderate cell alteration in the OFF SAC layer may be due to variable cytotoxic mechanism activation in the inner retina. 

Acknowledgments

The authors thank Franziska Christ for her technical help and Olivier Raineteau, PhD, for allowing us to use his SPE-II confocal microscope. 

Supported by the Swiss National Science Foundation Grants 31003A-149315-1 and 3100A0-1222527-2 (MES), the Forschungskredit of the University of Zürich Grant no. K-41210-01-01, and the Velux Stiftung (project no. 817) (VP). 

Disclosure: S. Joly, None; A. Guzik-Kornacka, None; M.E. Schwab, None; V. Pernet, None