Response of Retinal Connexin43 to Optic Nerve Injury

Abstract

Purpose.: To characterize the spatial and temporal expression of Connexin43 (Cx43) after partial optic nerve transection and evaluate its relationship to retinal ganglion cell (RGC) loss and retinal glial response.

Methods.: Partial, unilateral, superior optic nerve transection was performed in 150 Wistar rats. The retinas were evaluated at 8 and 24 hours and 3, 7, 14, 28, and 56 days after injury. Immunohistochemical analysis identified changes in several markers including Cx43 immunoreactivity (ir), RGC counts (Brn3a), and retinal astrocytes (GFAP).

Results.: After injury, superior retinal Cx43-ir peaked at 3 days (192.1% of control; P = 0.0002) and 28 days (212.1% of control; P < 0.0001) and troughed at 14 days (73.8% of control; P = 0.0028) and 56 days (72.5% of control; P = 0.0232). Inferior retinal Cx43-ir was elevated at only 28 days (127.4% increase; P = 0.0481). Superior RGC loss began at 3 days (84.0% of control; P = 0.0454) and continued to decline by 56 days (18.8% of control; P < 0.0001). Inferior RGC loss began at 28 days (73.4% of control; P = 0.0021). An increase in GFAP-ir occurred in the superior retina from day 3 (153.7% of control; P = 0.0017) and from day 28 (186.7% of control; P = 0.0013) in the inferior retina, persisting in both the superior and inferior retina to 56 days (P = 0.0027).

Conclusions.: A biphasic upregulation of retinal Cx43 protein occurs in the superior retina with peaks at 3 and 28 days after injury, but at only 28 days in the inferior retina. There is an associated loss of RGCs and a retinal astrocytic inflammatory response.

Connexin43 (Cx43) is the most ubiquitously expressed gap junction protein in the central nervous system (CNS) and is predominantly found in brain astrocytes. ¹ Gap junctions are specialized cell-to-cell contacts that provide direct intercellular communication of molecules less than 1200 Da in size, which includes nutrients, metabolites, second messengers, cations, and anions. ² In the retina, Cx43 is principally found in macroglia, forming gap junctions between various combinations of astrocytes and Müller cells. ³  

There is accruing evidence that changes in the spatial and temporal expression of Cx43 occur in various models of CNS injury and that the severity of the injury governs the specific response of astrocytic Cx43. Generally speaking, mild to moderate injury leads to increased Cx43 immunoreactivity (ir) at the lesion site or in vulnerable CNS areas. ^4,5 Severe injury results in decreased Cx43-ir at the lesion site, which is surrounded by a zone of increased Cx43-ir in what could be described as the penumbra of the injury. ^{4,6 –14} These observations indicate that astrocytic gap junction communication is an important mediator after CNS injury, and the preponderance of present evidence suggests that preventing Cx43 upregulation that occurs after CNS injury increases neuronal survival. ^{15 –18} The most likely mechanisms by which Cx43 modulation leads to neuroprotection is by blocking Cx43 gap junctions, which would prevent the spread of cell death from areas of injured neurons to healthy bystanders, ¹⁹ or by gap junction hemichannel events. ²⁰  

Optic nerve injury models are useful in study of CNS disorders due to the accessibility of the optic nerve axons and their distant retinal ganglion cell (RGC) bodies in the retina. ²¹ After complete transection of the optic nerve in rats, there is an initial 1-week period of RGC body survival, followed by substantial RGC loss during the next 2 weeks. ²² RGC axons in the optic nerve distal to the injury site degenerate more rapidly. There is consistent evidence that RGCs die by apoptosis. ^23,24 A body of evidence now shows that the cellular presence of Cx43 and gap junction communication strongly influences apoptotic activity though there is debate as to whether it is facilitatory or inhibitory. ^20,25,26  

A model in which the superior optic nerve axons are transected was developed to exploit the topographic distribution of RGC axons, generating within the eye a superior zone in which there is direct RGC axonal injury and an opposite inferior hemifield that might serve as a zone without direct injury that is exposed to a toxic environment, similar to that of the penumbra zone in CNS lesions. ²⁷ This paradigm has demonstrated that secondary injury in the inferior retina exhibits slower and less significant RGC loss than in the primarily injured superior retina. ²⁸ This approach allows investigation of the relevance of spatial or temporal changes in Cx43 in the retina after partial optic nerve transection. The present research details the relationship of changes in Cx43 in both superior and inferior retina in this model, relating them to apoptotic RGC neuronal loss and the response of retinal glia. 

Materials and Methods

Animals

Wistar rats (250–350 g) were obtained from the Vernon Jansen Unit (VJU), University of Auckland. All animal procedures in this study were approved by the Animal Ethics Committee at the University of Auckland and were in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Rats were housed under standard conditions and fed food and water ad libitum. 

Partial Optic Nerve Transection Model

The technique of partial optic nerve transection was adapted from methods outlined by Levkovitch-Verbin et al. ²⁷ Briefly, unilateral partial optic nerve transection was performed in animals that were anesthetized with intraperitoneal ketamine (80 mg/kg) and xylazine (4 mg/kg) and topical benoxinate 0.4% eye drops. Intraocular disease was excluded by examination after instillation of mydriatic eye drops (tropicamide 1.0%). A superior conjunctival incision was made, and the eye was gently retracted outward with forceps, exposing the nerve behind the eye. The meninges overlying the optic nerve was carefully divided superiorly, exposing the optic nerve. A 200-μm incision was then made across the optic nerve with diamond step knife and 45° blade (Huco Vision, St. Blaise, Switzerland) at a point 0.5 to 1.0 mm behind the eye. The conjunctival incision was self-closing. Each surgically injured eye was inspected ophthalmoscopically to ensure patency of blood flow to the eye. The same procedure was performed for sham surgery, but the diamond step knife was sheathed and placed on the exposed optic nerve. 

Animals were euthanatized by carbon dioxide (CO₂) inhalation at 8 and 24 hours; 3, 7, and 14 days; and 1 and 2 months after partial optic nerve transection. A single conjunctival suture was placed superiorly for orientation. Enucleation with maintenance of a long optic nerve segment was carefully performed. 

Immunohistochemistry

Six animals per time point were used for immunohistochemical analysis. Samples were immediately postfixed after extraction by submersion in 1% paraformaldehyde for 30 minutes, followed by overnight cryoprotection with 15% sucrose. Tissue was then embedded in OCT compound and snap frozen with liquid nitrogen. Sagittal cryosections of 16-μm thickness were mounted on slides (Superfrost Plus; Menzel-Gläser, Braunschweig, Germany), dried for 30 minutes at room temperature, and stored at −20°C. Sections including the optic nerve head as a landmark were used for immunolabeling. 

The sections were further fixed in −20°C ethanol for 10 minutes if they were for Cx43 and astrocyte labeling; otherwise this step was omitted. After a thorough rinsing in phosphate-buffered solution (PBS), the sections were preincubated in the appropriate blocking buffer (10% normal goat serum and 0.1% Triton X-100 in PBS for all except RGC labeling in which 10% horse serum was substituted) for 1 hour at room temperature. 

Primary antibodies were applied overnight at 4°C for Cx43 (rabbit polyclonal anti-Cx43, 1:2000; Sigma-Aldrich, St. Louis, MO), retinal ganglion cells (anti-Brn-3a, 1:500, Santa Cruz Biotechnology Inc., Santa Cruz, CA), astrocytes (mouse monoclonal anti-glial fibrillary acidic protein [GFAP-Cy3 conjugate], 1:1000; Sigma-Aldrich), Müller cells (mouse monoclonal anti-glutamine synthetase [GS], 1:200; Abcam, Cambridge, UK), activated microglia (mouse monoclonal anti-OX-42 [CD11b], 1:100; Serotec, Oxford, UK), and blood vessel endothelial cells (mouse monoclonal anti-isolectin IB4; IL-B4)/Alexa 594 conjugate, 1:100; Invitrogen-Molecular Probes, Eugene, OR). 

After rinsing in PBS, appropriate secondary antibodies were applied for 2 hours at room temperature, including goat anti-rabbit Alexa 488 (1:1000; Invitrogen, Carlsbad, CA), goat anti-mouse Alexa488 (1:500; Invitrogen), goat anti-mouse Cy3 (1:500; Jackson ImmunoResearch Laboratories Inc.), and donkey anti-goat Cy3 (1:1000, Jackson ImmunoResearch Laboratories Inc.). 

After further rinsing in PBS, the sections were mounted in antifade reagent with DAPI (ProLong Gold; Molecular Probes) on coverslips, which also provided precise identification of retinal and nerve morphology. 

Imaging

Immunofluorescence was analyzed using a microscope (DMRA; Leica, Wetzlar, Germany) fitted with a digital camera (model DS-U1; Nikon, Tokyo, Japan) and a confocal laser scanning microscope (model FV1000l with FluoView software; ver. 1.7a; Olympus, Tokyo, Japan). Single-slice confocal images were captured at a speed of 4.0 μs/pixel, with a resolution of 1024 × 1024 pixels, using a Kalman average of 4 and on sequential mode if using multiple lasers. Gain and offset were standardized using control tissue for all analyses. Images of the retina containing the RGC layer and the inner nuclear (INL) and outer nuclear (ONL) cell layers, were captured at 60× magnification, except for RGC analysis (40×). Three retinal images were acquired both superior and inferior to the optic nerve head, separated by a distance of one eyepiece field (60× eyepiece field, 300 μm; 40× eyepiece field, 450 μm). 

Cx43 Immunohistochemical Analysis

Retinal Cx43 immunohistochemical analysis was performed qualitatively and quantitatively on three sections of each animal. For quantitative analysis, confocal images were exported as TIFF files into ImageJ software (version 1.41o; developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html). Image area was cropped to include the retinal glia only, and image type was converted to 8-bit black and white. A consistent threshold of 55 was applied to reduce background immunoreactivity, and an automated immunolabel spot count was then performed (Fig. 1). 

Figure 1.

 

Postimaging analysis for Cx43-ir (A–C) and RGC (D–F) quantification. (A) The original single-slice confocal image of Cx43-ir converted to (B) an 8-bit black and white image, with a final image (C) after a threshold of 55 was applied. (D) The original single-slice confocal image of Brn3a-labeled RGCs converted to (E) an 8-bit black-and-white image, with final image (F) after a despeckle and two erode processes.

RGC Immunohistochemical Analysis

Brn3a has been validated as a reliable, efficient marker to quantify RGCs in optic nerve-injured retinas. ²⁹ For quantification of RGCs, Brn3a TIFF files from two sections of each animal were converted to 8-bit black and white and a threshold of 33 was applied. A “despeckle” process followed by two “erode” processes were applied, to allow for easier discrimination of individual RGCs. A single, masked observer made manual cell counts on the final images (Fig. 1). 

Astrocyte Immunohistochemical Analysis

Retinal GFAP immunohistochemical analysis was performed qualitatively and quantitatively on the same sections used for Cx43 immunohistochemical analysis (three sections per animal). For quantitative analysis, GFAP TIFF files were converted to 8-bit black and white and a threshold of 37 was applied, and an automated area count of immunolabeled cells was performed. 

Inflammatory Response Immunohistochemical Analysis

Qualitative assessment of immunolabeled activated microglia (OX-42) and blood vessel endothelial cells (isolectin-IB4 [IL-IB4]) were performed on retinal sections. A basic quantitative analysis of OX-42 immunolabeling was also performed. Aggregated label, subjectively deemed to represent one cell, was manually counted in individual retinal layers. 

TUNEL Assay

Analysis of apoptotic cells by the terminal deoxynucleotidyl transferase deoxyuridine triphosphate (dUTP) nick-end labeling (TUNEL) method was performed with an in situ fluorescein apoptosis detection kit (ApopTag Plus; Chemicon International, Temecula, CA) the prescribed methodology. Briefly, sections were rinsed in PBS before and after being postfixed in a precooled ethanol/acetic acid (2:1) solution at −20°C for 5 minutes. An equilibration buffer was subsequently applied for a minimum of 10 seconds. After this, sections were treated with a solution containing terminal deoxynucleotidyl transferase and digoxigenin-conjugated nucleotides for 1 hour at 37°C. This step was necessary to attach the digoxigenin-dUTPs to the terminal end of nucleic acids in DNA fragments caused by the apoptotic process. Sections were then incubated in a stop/wash buffer for 10 minutes at room temperature to halt the enzymatic dUTP addition. After washing in PBS, a fluorescein conjugated anti-digoxigenin antibody was applied to the sections for 30 minutes at room temperature, followed by further PBS washing. The assay was completed by mounting the slides with a DAPI mounting medium on coverslips. 

Polymerase Chain Reaction of Cx43 mRNA and Western Blot Analysis

To determine the presence of Cx43 mRNA real-time quantitative polymerase chain reaction (RTqPCR) was performed. Three animals were selected from each of the control and 3-day posttransection groups. Extracted retinal samples were frozen immediately in liquid nitrogen. Total RNA was isolated from the experimental and sham retinal samples by the acid guanidinium thiocyanate-phenol-chloroform extraction method. The samples were homogenized with RNA extraction reagent (TRIzol; Invitrogen, Carlsbad, CA) and purified (PureLink RNA Micro Kit; Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. A spectrophotometer (Nanodrop ND-1000; Thermo-Scientific, Wilmington, DE) was used to assess RNA quantity and quality. One hundred nanograms of total RNA was reverse transcribed into cDNA (SuperScript VILO cDNA synthesis kit; Invitrogen) in a 20-μL reaction mixture under the following conditions: 25°C for 10 minutes; 42°C for 120 minutes; and 85°C for 5 minutes. The cDNA product was then amplified by PCR. PCR was performed on a real-time rotary analyzer (Rotor-Gene 6000; Corbett Life Science, Sydney, Australia). The reaction was initiated with a commercial system (FastStart Universal SYBR Green Master [Rox]; Roche, Basel, Switzerland) according to the manufacturer's specifications. The 50-μL reaction mixture consisted of 25 μL of green nucleic acid gel stain (FastStart SYBR Green [Rox]; Roche), 5 μL of cDNA, 0.5 μL of forward primer, 0.5 μL of reverse primer, and 19 μL of PCR-grade water. The primers used for connexin43 were 5′-GATTGAAGAGCACGGCAAGG-3′ (sense) and 5′-GTGTAGACCGCGCTCAAG-3′ (antisense). The following conditions were used for 40 cycles: denaturation at 95°C for 10 minutes, annealing at 95°C for 15 seconds, and extension at 60°C for 1 minute. 

The expression of connexin43 was normalized to β-actin, a housekeeping gene. The primers used for β-actin were 5′-GATTTGGCACCACACTTTCTACA-3′ (sense) and 5′-ACTTTGGTCATCTTTTCACGGTTGG-3′ (antisense). Each PCR reaction was repeated three times for every eye at all time points. In addition, negative controls were performed without reverse transcriptase. 

For Western blot analysis, eyes were enucleated from three control and three animals 3 days after transection, and the retina dissected and immersed in liquid nitrogen. The samples were homogenized in ice-cold phosphate-buffered saline containing 150 mM sucrose, 15 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (pH 7.9), 6 mM potassium chloride, 2 mM EDTA acid (pH 8.0), 1 mM EGTA (pH 8.0), and a protease inhibitor (Complete Protease Inhibitor Cocktail; Roche). After the addition of 1% Triton X-100 (Sigma-Aldrich), the homogenate was incubated for 1 hour on ice. The supernatant was collected after centrifugation at 10,000 rpm for 10 minutes and combined with a loading dye (125 mM Tris [pH 6.8]), 0.8% sodium dodecyl sulfate [SDS], 2% glycerol, and 0.1% B-mercaptoethanol). 

Equal amounts of protein were loaded onto a 10% SDS polyacrylamide gel and electrophoretically transferred to a nitrocellulose membrane. A molecular weight standard (BenchMark Protein Ladder; Invitrogen) was included. After incubation in blocking solution (5% nonfat milk powder and 0.1% Tween-20 in Tris-buffered saline [TBS]) for 1 hour at room temperature, the membrane was incubated sequentially with 1:6500 rabbit polyclonal anti-connexin43 antibody and 1:4000 rabbit anti-GAPDH antibody for 12 hours at 4°C. The membrane was then washed thoroughly with 0.1% Tween-20 in TBS and incubated with Alexa488-conjugated secondary antibodies at a dilution of 1:10,000 for 1 hour at room temperature. Blots were washed with 0.1% Tween-20 in TBS before detection with Western blot detection reagents (Amersham ECL Plus; GE Health care, Piscataway, NJ). Chemiluminescence was detected with an imaging system (LAS-3000; Fujifilm, Tokyo, Japan) and analyzed with ImageJ software. 

Statistics

Data were analyzed with used Student's t-test or one-way ANOVA, whichever was appropriate (Prism 5 for Windows; GraphPad, San Diego, CA). P < 0.05 was considered to be significant. 

Results

Biphasic Retinal Cx43 Response after Partial Optic Nerve Transection

Cx43-ir was primarily present in the retinal nerve fiber layer (RNFL) and ganglion cell layer (GCL), but was sparsely present in the superficial retinal layers (Fig. 2). In the RNFL and GCL, Cx43-ir was found to predominantly co-localize with GFAP-ir, representative of retinal astrocytes or Müller cell end feet (Figs. 3, 4). However, minimal co-localization of Cx43-ir was observed in the RNFL and GCL with GS-ir, a marker for Müller cells, and with IL-B4-ir, a marker for blood vessel endothelial cells (Fig. 3). In the more superficial retinal layers, Cx43-ir primarily co-localized with IL-B4-ir, but not with GS-ir or GFAP-ir, representative of Müller cell processes (Fig. 3). 

Figure 2.

 

Superimposed single-slice confocal images of retinal sections labeled for Cx43 (green) and DAPI (blue) showing a predominance of Cx43-ir in the RNFL and GCL, with minimum Cx43-ir in the superficial retinal layers. IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer. Scale bar, 50 μm.

Figure 3.

 

Superimposed single-slice confocal images of retinal sections labeled for Cx43 (green), DAPI (blue), and various markers (red). In the RNFL and GCL, (A) shows co-localization between Cx43-ir and GFAP-ir (red), (B) shows minimal co-localization between Cx43-ir and IL-B4-ir (red), and (C) shows minimal co-localization between Cx43-ir and GS-ir (red). In the superficial retinal layers, (D) shows no co-localization between Cx43-ir and GFAP-ir, (E) shows some co-localization between Cx43-ir and IL-B4-ir, and (F) shows no co-localization between Cx43-ir and GS-ir. All images are the same magnification. IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer. Scale bar, 30 μm.

Figure 4.

 

Superimposed single-slice confocal images of superior retinal sections labeled for Cx43 (green), GFAP (red) and DAPI (blue). Baseline Cx43-ir and GFAP-ir displayed in (A) control retina, is predominantly seen in RNFL and GCL. No significant change in Cx43-ir or GFAP-ir is seen at (B) 8 hours after injury. Increased Cx43-ir is seen at (C) 24 hours, (D) 3 days, (E) 7 days, and maximally at (G) 28 days after injury. Decreased Cx43-ir is seen at (F) 14 and (H) 56 days after injury. (D) to (H) show a progressive increase in GFAP-ir, reaching significance from (D) 3 days on. All images are the same magnification. IPL, inner plexiform layer; INL, inner nuclear layer. Scale bar, 50 μm.

After injury, differences in superior retinal Cx43-ir were observed in RNFL and GCL (Fig. 4), but not in the other retinal layers. Quantitative analysis revealed a biphasic response of Cx43-ir in the retina, peaking at days 3 and 28 after injury (Fig. 5). The changes in Cx43 spot counts for the total, superior, and inferior retina are outlined in Table 1. Superior retinal Cx43-ir was significantly increased compared with controls, as early as 24 hours after injury (141.7% of control, SD ± 53.1%; P = 0.0063) and continued to increase at 3 days (192.1% of control, SD ± 88.7%; P = 0.0002). By 7 days, it had dropped to 136.6% (SD ± 41.3%; P = 0.0043) of control, and at 14 days, the levels were significantly lower than the control (73.8% of control, SD ± 18.0%; P = 0.0028). A further maximum increase in Cx43-ir occurred at 28 days (212.1% of control, SD ± 90.1%; P < 0.0001), followed by another decrease at 56 days (72.5% of control, SD ± 37.6%; P = 0.0232). No significant change in inferior retinal Cx43-ir was noted except at 28 days after injury (127.4% increase cf. controls, SD ± 40.9%; P = 0.0481; Table 1). 

Figure 5.

 

(A) Quantification of Cx43-ir spot counts per retinal section. Superior (light gray bars) and inferior (dark gray bars) spot counts stacked to show total retinal counts. Compared with the control, a significant increase in Cx43-ir spot counts was observed in the superior retina at 24 hours and 3, 7, and 28 days after injury and in the inferior retina at 28 days. A significant decrease was seen in the superior retina at 14 and 56 days after injury when compared with control superior retina. (B) Quantification of RGCs per three superior retinal sections. Number of superior (light gray bars) and inferior (dark gray bars) RGCs stacked to show the total number of RGCs. Compared with the control, a significant loss of superior RGCs was observed from 7 days after injury. Inferior RGCs were significantly lower than the control at 28 days after injury. Note that the increase in superior Cx43-ir preceded superior RGC loss and an increase in inferior CX43-ir occurred in conjunction with inferior RGC loss. Results are expressed as the mean ± SEM. *Denotes statistical significance; P < 0.05.

Table 1.

 

Retinal Cx43 Spot, RGC, and GFAP Area Counts

Time Point	n	Cx43 Spot Count/Section		P vs. Controls	RGC Count/Section		P vs. Controls	GFAP Area Count/Section (× 10−6)		P vs. Controls
Control	18	Total	265.6 ± 98.3		Total	84.5 ± 9.7		Total	13.8 ± 3.9
		Sup.	143.4 ± 52.7	—	Sup.	42.6 ± 8.8	—	Sup.	7.0 ± 2.6	—
		Inf.	109.4 ± 74.7	—	Inf.	41.9 ± 5.0	—	Inf.	6.8 ± 2.6	—
Sham	7	Total	265.1 ± 96.8	0.99	Total	82.0 ± 11.2	0.63	Total	15.9 ± 4.5	0.28
		Sup.	145.4 ± 42.9	0.99	Sup.	41.3 ± 7.0	0.77	Sup.	8.6 ± 3.3	0.21
		Inf.	121.7 ± 54.2	0.70	Inf.	40.7 ± 6.4	0.68	Inf.	7.2 ± 2.6	0.72
8 h	18	Total	299.0 ± 116.4	0.36	Total	79.2 ± 13.5	0.29	Total	14.8 ± 4.6	0.50
		Sup.	183.2 ± 80.1	0.087	Sup.	39.3 ± 8.8	0.38	Sup.	8.0 ± 3.5	0.32
		Inf.	115.8 ± 74.4	0.80	Inf.	39.9 ± 10.1	0.56	Inf.	6.8 ± 2.8	0.95
24 h	18	Total	303.3 ± 93.2	0.25	Total	81.9 ± 12.7	0.58	Total	14.7 ± 4.1	0.52
		Sup.	206.1 ± 77.3	0.0063*	Sup.	37.3 ± 7.1	0.14	Sup.	8.4 ± 2.7	0.14
		Inf.	97.2 ± 48.6	0.57	Inf.	44.3 ± 8.1	0.40	Inf.	6.3 ± 2.7	0.59
3 d	18	Total	421.4 ± 170.3	0.0019*	Total	78.3 ± 13.3	0.21	Total	18.9 ± 5.2	0.0025*
		Sup.	279.4 ± 129.0	0.0002*	Sup.	35.8 ± 6.9	0.045*	Sup.	10.8 ± 3.9	0.0017*
		Inf.	142.0 ± 68.6	0.19	Inf.	42.6 ± 9.2	0.83	Inf.	8.1 ± 2.0	0.11
7 d	18	Total	330.6 ± 87.6	0.044*	Total	67.8 ± 10.4	0.0005*	Total	19.5 ± 4.7	0.0004*
		Sup.	198.7 ± 60.0	0.0043*	Sup.	28.5 ± 4.8	<0.0001*	Sup.	11.1 ± 3.2	0.0002*
		Inf.	131.8 ± 51.7	0.31	Inf.	39.3 ± 7.7	0.34	Inf.	8.4 ± 2.5	0.074
14 d	18	Total	185.0 ± 56.1	0.0048*	Total	58.3 ± 14.9	<0.0001*	Total	18.2 ± 6.7	0.024*
		Sup.	107.3 ± 26.2	0.0028*	Sup.	18.7 ± 6.7	<0.0001*	Sup.	8.8 ± 3.2	0.078
		Inf.	77.7 ± 47.1	0.14	Inf.	39.6 ± 10.4	0.50	Inf.	9.4 ± 5.2	0.071
28 d	18	Total	461.5 ± 150.8	<0.0001*	Total	39.8 ± 13.7	<0.0001*	Total	28.5 ± 8.3	<0.0001*
		Sup.	308.5 ± 131.0	<0.0001*	Sup.	9.1 ± 4.7	<0.0001*	Sup.	15.8 ± 5.2	<0.0001*
		Inf.	153.0 ± 49.1	0.048*	Inf.	30.8 ± 9.6	0.0021*	Inf.	12.7 ± 6.7	0.0013*
56 d	16	Total	224.6 ± 119.4	0.28	Total	33.0 ± 10.9	<0.0001*	Total	25.26 ± 14.1	0.0027*
		Sup.	105.5 ± 54.6	0.023*	Sup.	8.0 ± 3.7	<0.0001*	Sup.	12.6 ± 8.7	0.015*
		Inf.	119.1 ± 82.4	0.72	Inf.	25.0 ± 7.7	<0.0001*	Inf.	12.7 ± 6.8	0.008*

*  Statistically significant.

The presence of Cx43 mRNA and protein in control and injured retinal samples was confirmed by RT-qPCR and Western blot analysis, respectively. It was not possible to accurately quantify relative levels of Cx43 mRNA or protein by either method. Extracted retinal samples were shown through immunohistochemical techniques, to contain variable quantities of both neurosensory and pigmented retinal layers. The underlying retinal pigment epithelium, a structure expressing particularly high levels of connexin43, overwhelmed changes occurring in the inner retinal layers which were of a much smaller scale. The inability to separate retinal layers consistently and therefore distinguish the origin of the connexin43 mRNA or protein resulted in the decision not to use RTqPCR or Western blot analysis as a quantitative tool. 

Cx43 Upregulation Coincides with Significant RGC Loss

RGC loss was determined by quantitative analysis of Brn3a-labeled RGCs in two retinal sections per animal with six animals per time point (Fig. 6). The number of superior RGCs after injury was significantly reduced compared with the control number at 3 days (84.0% of control, SD ± 16.2%; P = 0.0454) and continued to decline at 7 (66.9% of control, SD ± 11.4%; P < 0.0001), 14 (43.8% of control, SD ±15.6%; P < 0.0001), 28 (21.3% of control, SD ± 11.1%; P < 0.0001), and 56 (18.8% of control, SD ± 8.8%; P < 0.0001) days. A significant decrease in the number of inferior RGCs was measured at 28 (73.4% of control, SD ± 22.8%; P = 0.0021) and 56 (59.6% of control, SD ± 18.4%; P < 0.0001) days but not earlier (Table 1, Fig. 5). 

Figure 6.

 

Superimposed single-slice confocal images of retinal sections labeled with Brn3a (red) and DAPI (blue). Baseline Brn3a-positive RGCs displayed in (A) control retina. Arrowhead: a displaced amacrine cell in the GCL. A significant loss in superior RGCs was seen at (B) 3 days through to (C) 56 days after injury. A significant loss in inferior RGCs was seen at (D) 28 days after injury. All images are the same magnification. IPL, inner plexiform layer. Scale bar, 50 μm.

The onset of superior RGC loss at 3 days appeared to coincide with the early peak in superior Cx43-ir, which was initially elevated from 24 hours through to 7 days after injury. Similarly, at 28 days, the significant loss of inferior RGCs compared with controls occurred in conjunction with a significant elevation of inferior retinal Cx43-ir. By 56 days, the total number of RGCs was not significantly different from that at 28 days (P = 0.2347), and total Cx43 had returned to levels not significantly different from the control. The TUNEL assay confirmed the presence of a low number of TUNEL-positive cells in the RGC layer of multiple retinal sections. 

Retinal Astrocyte Activation Occurs after Partial Transection

Evaluation of the retinal glial response after optic nerve injury was performed by analysis of GFAP-ir and GS-ir. After injury, a qualitative increase in superior retinal GFAP-ir was observed in the RNFL (Figs. 4, 7), but not the other retinal layers (Fig. 7). No significant qualitative change in GS-ir was observed in any of the retinal layers after injury (Fig. 7). 

Figure 7.

 

Superimposed single-slice confocal images of superior retinal sections labeled with DAPI (blue) and glial markers (red). (A) GFAP-ir leveled to highlight Müller cell processes extending into superficial retinal layers. With the same leveling parameters, minimal change was observed in the GFAP-positive Müller cell processes in (B) 28 days after injury. In contrast, although oversaturated, GFAP-ir was increased in the RNFL in (B) compared with that in (A). No significant change in GS-ir was seen throughout the retina in (C) control retina and (D) 28 days after injury. All images are same magnification. IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer. Scale bar, 50 μm.

Quantitative analysis of GFAP-ir in the same three retinal sections per animal used for Cx43-ir analysis confirmed a progressive increase in the area of GFAP-ir in the superior retina that became significant compared with the control at 3 (153.7% increase, SD ± 55.6%; P = 0.0017), 7 (158.2% increase, SD ± 46.1%; P = 0.0002), 28 (224.9% increase, SD ± 74.7%; P < 0.0001), and 56 (178.9% of control, SD ± 124.2%; P = 0.0149) days after injury. The area of GFAP-ir in the inferior retina was significantly higher than in the control at 28 (186.7% increase, SD ± 97.7%; P = 0.0013) and 56 (186.2% of control, SD ± 100.0%; P = 0.0018) days after injury (Table 1). The increased GFAP-ir observed after partial transection was similar in magnitude to the increase in Cx43-ir but persisted longer, remaining significantly elevated, even at 56 days (P = 0.0027). 

Retinal Inflammatory Response after Partial Transection

In addition to retinal astrocyte activation, OX-42-ir and IL-B4-ir were assessed after injury and compared to Cx43-ir. A maximum increase in OX-42-ir was observed 14 days after injury in multiple retinal layers both in the superior and inferior retina (Fig. 8). A basic quantitative analysis of OX-42-ir in the superior and inferior retina confirmed a consistent increase at 14 and 28 days after injury (Fig. 9). No colocalization was observed between OX-42-ir and Cx43-ir at high magnification (Fig. 8). Because OX-42 is located in the plasma membrane, Cx43 immunolabeling surrounding OX-42-positive cells in the GCL is likely to be expressed on other cell types. An analysis of retinal vasculature with isolectin-IB4 did not show any qualitative difference in distribution or caliber at any time point after injury. 

Figure 8.

 

Confocal images of retinal sections labeled for OX-42 (red) and DAPI (blue) in (A) to (C). (A) Control retina showed minimal OX-42-ir. A multilayered increase in OX-42-ir was observed at (B) 14 days after injury and (C) 28 days after injury. (D) A high-magnification, z-stacked image of an OX-42-positive cell (red) double labeled for Cx43 (green) showing no co-localization. IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer. Scale bar: (A–C) 30 μm; (D) 5 μm.

Figure 9.

 

A consistent increase in OX-42-positive cells was observed in the GCL, IPL, INL, and OPL at 14 and 28 days after injury, compared with the controls. Results are expressed as the mean ± SEM. IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer.

Discussion

The present study is the first to evaluate the spatial and temporal response of retinal Cx43 after an in vivo model of optic nerve injury. Optic nerve injury evoked a biphasic response of retinal Cx43 protein expression in the retinal area corresponding to axonal damage (superior), and a delayed elevation in the retinal area unaffected by the primary injury (inferior). In both areas, the peak in Cx43-ir coincided with significant RGC loss. There was also an associated inflammatory response including activation of astrocytes and microglia. 

More specifically, after partial transection of the superior optic nerve, the superior retina showed a significant increase in Cx43-ir as early as 24 hours after injury, peaking to almost 200% of control at day 3, and remaining elevated 7 days after injury. This increase in superior retinal Cx43-ir coincided with, or even preceded, significant superior RGC loss which began at 3 days after injury, with RGC counts declining to 66.9% of control at 7 days after injury. Superior retinal Cx43-ir dropped to subnormal levels at 14 days but increased again to maximum levels by 28 days after injury. By comparison, inferior retinal Cx43-ir significantly changed at only 28 days after injury, when it was were elevated to 127% of control levels. This was again paralleled by significant inferior RGC loss at 28 days, with inferior RGC counts down to 73% of control. 

RGC loss by Brn3a analysis of retinal sections compared relatively well with data from Levkovitch-Verbin et al., ²⁷ who used retrograde Fluorogold labeling of retinal wholemounts, bearing in mind that our incision was 50 μm deeper than theirs. With respect to superior RGC loss, our data showed less loss at 7 days (−33%) than their 8-day time point (−63%), but more loss at 28 days (−80%) than their 4 week time point (−46%). Of interest, both sets of data do not show significant differences in inferior RGC counts until 4 weeks (−27% cf. −39%). TUNEL analysis confirmed the presence of a low number of TUNEL-positive cells likely to represent RGC apoptosis. This is in keeping with data from Isenmann et al., ³⁰ who detected the occasional TUNEL-positive RGC from 2 to 24 hours after axotomy. Their data showed a peak of TUNEL-positive RGC apoptosis at 6 days (18 cells in six sections), which dropped sharply to occasional cells by 28 days. 

Attempts were made to quantify the total amount of retinal Cx43 mRNA using RTqPCR and Cx43 protein via Western blot, to further elucidate whether the changes in Cx43-ir seen were due to de novo synthesis or changes in the cellular distribution of Cx43 protein. While both techniques confirmed the presence of Cx43 mRNA and Cx43 protein in both control and injured tissue, quantification results were misleading secondary to contamination of the tissue specimens by retinal pigment epithelium, a structure whose abundance in Cx43 protein masked changes in the neurosensory retina. Current evidence points toward de novo synthesis as the most likely explanation. Cameron et al. ³¹ showed upregulation of Cx43 transcripts using a DNA microarray and real-time PCR analysis after mechanical injury to zebrafish retina. After focal cerebral ischemia, reactive astrocytes have been shown to envelope the lesion and upregulate Cx43 mRNA protein expression. ⁷ Complete transection of the adult rat spinal cord led to an upregulation of Cx43 mRNA and protein on GFAP-positive astrocytes within hours, remaining significantly elevated up to 28 days after the injury. ⁹ However, the possibility remains that the increase in Cx43-ir that we observed resulted from the cellular redistribution of astrocytic gap junction protein from a pool of Cx43 not normally detectable by immunohistochemistry, as proposed by Hossain et al. ⁴ after a global cerebral ischemia model in which Cx43-ir increased but Cx43 protein levels by Western blot did not. 

To identify the retinal cell types expressing Cx43, double-labeling immunohistochemical techniques were used. Cx43-ir predominantly co-localized with GFAP-ir in the RNFL and GCL, which could represent Cx43 expression by either retinal astrocytes or the end feet of Müller cells as has been previously studied by Zahs et al. ³ However, the minimal co-localization we observed between Cx43-ir and GS-ir, a Müller cell–specific marker, throughout the retina suggests that the former is most likely. This is in agreement with Ball et al., ³² who demonstrated Müller cell Cx43 staining in goldfish and mudpuppy retinas but not the rat retina and with the vast amount of literature studying astrocytic Cx43 (for review, see Chew et al. ³³ ). In the more superficial layers, Cx43-ir showed some co-localization with IL-B4-ir, a marker for blood endothelial cells, which did not change after transection injury. This is in contrast to the upregulation in Cx43 observed in small vessel walls after spinal cord injury ¹⁵ and skin injury. ^34,35 An increase in Cx43 gap junction communication has also been observed in the microglia of primary glial cell cultures ³⁶ and brain stab wounds, ³⁷ but no definite Cx43-ir co-localization with OX-42-ir was noted in this study. 

The inflammatory response observed after partial transection is also of interest. GFAP-positive glial activation was seen to predominantly occur in the retinal astrocytes of the RNFL and GCL in the superior retina until 28 days when the inferior also became significantly higher. Surprisingly, there was no increase in GFAP-ir or GS-ir in the other retinal layers, as has been described in rodent Müller cell fibers after optic nerve crush injury. ^38,39 Perhaps an explanation is the difference in injury model, as a crush would affect a greater volume of axons than a partial transection. Astrocyte activation paralleled Cx43 levels well except at 14 and 56 days after injury where superior Cx43-ir was decreased despite increased GFAP-ir. At these times, decreased gap junction communication was occurring despite a persisting astrocytosis. The microglial response to injury was more widespread with an increase in OX-42-ir observed throughout the layers of the superior and inferior retinas. This could represent either direct proliferation of the microglial population, recruitment and differentiation of circulating monocytes into microglia, or activation and hypertrophy of resting microglia. ⁴⁰  

The retinal Cx43 changes observed after distal optic nerve injury suggest that the RGC-glia interplay is highly significant in the retinal response to RGC axon injury. There is abundant evidence that RGC loss after injury to their axons at the optic nerve head or optic nerve is signaled from the injury site through retrograde axonal transport, leading to decreased neurotrophic support from the central target and/or the arrival at the RGC cell body of molecules signaling injury. ^41,42 Our data support the participation of astrocytic upregulation of Cx43 in the early phenomena of this injury process. The initial increase in the level of Cx43 in the superior retina that we detected coincided with, or even preceded, significant superior retinal RGC death, suggesting that the Cx43 changes are not a reactive or secondary event, but instead occurs early in the injury pathway. The retinal astrocytes may well be reacting to local signals and events generated by RGC body injury. We speculate that cell-to-cell interactions with the RGCs in the initial stages of injury lead to astrocytic and microglial activation. Retinal glia abundantly express Cx43 ³ and are known to play a role in the regulation of neurotransmitters, such as glutamate, growth factors, and inflammatory processes after damage. ^43,44 It is possible that the Cx43 localized in retinal macroglia in close proximity to the RGC cell layer may regulate excitotoxic RGC apoptosis. In brain slices subjected to acute injury, reactive gliosis and upregulation in astrocytic Cx43 have been observed. ^17,45,46 Lesion spread has been documented to occur in the slices via gap junction–mediated calcium signaling between astrocytes. ^{17,47 –50} Mechanical damage on the cut surface of the slices may induce hemichannel opening, allowing release of ATP into the extracellular space. ^{51 –53} Increase in ATP then activates G-protein-coupled receptor signaling cascade, which results in elevation of intracellular calcium levels, further release of ATP ⁵⁴ and glutamate. ⁵⁵ Toxic signals such as mobilized calcium and inositol triphosphate may also passively diffuse into cells coupled by gap junctions. ^56,57 This amplification of toxic signals causes irreversible cellular damage, extending the area of the lesion. This process may well occur in the retina, as well. A further possibility is that astrocytes are linked via their Cx43 gap junctions in a functional syncytium. It is conceivable that injury at the level of the optic nerve is signaled through this astrocytic network back into the retinal astrocytes. However, this is unlikely, given that the astrocytic Cx43 response is limited to the superior retina alone until later in the process. 

We used the partial transaction model to segregate and study the responses of primarily and secondarily injured RGC and the zones of retina in which they reside. As reported by Levkovitch-Verbin et al. ²⁷ and indirectly suggested by Yoles and Schwartz, ⁵⁸ the secondary degeneration process lags in time from primary RGC loss, both in the later and less severe RGC death in the inferior retina. The model has been used successfully by other groups to determine the effects of glatiramer acetate and the calcium channel blocker lomerizine on primary and secondary degeneration of the optic nerve and RGCs. ^28,59,60 Our work adds the important new finding that there was a parallel dissociation in time between Cx43 upregulation in upper and lower retina in our model, with Cx43 increase delayed in the zone of putatively secondary degeneration inferiorly, just as RGC cell body loss was delayed there. This result could be interpreted to mean that the interastrocytic communication via gap junctions, while potentially permitting broad geographic effects, is in fact limited to the zone of active injury. We cannot be certain at this time as to whether the Cx43 changes are beneficial or detrimental to RGC survival, although it is tempting to speculate the latter, given the above-mentioned literature and that the increase in superior retinal Cx43-ir precedes RGC loss. The spacing between later time points may well be masking the exact relationship between Cx43 increase and RGC loss at this time. Studies to inhibit or to stimulate Cx43 presence and function are needed to investigate these issues. 

The authors recognize that there are certain limitations to the injury model. Injury consistency was a problem because of the mobility of the optic nerve in the orbit. A shallow lesion may result in less changes in immunohistochemical markers and a lesion passing into the inferior optic nerve due to inadvertent rotation of the optic nerve would be even more detrimental by leading to inferior immunohistochemical changes unrelated to delayed death. However, no significant inferior RGC loss was seen at 7 days, which is known to be the time when almost 50% of axotomized RGCs die. Another limitation was the range of physiologic variability seen with various immunohistochemical markers such as Cx43, as seen with the high standard deviations even in the control retinas. To work around this issue, we used six animals per time point for immunohistochemistry and thoroughly examined three retinal sections per animal to create a large numerical database. 

In summary, Cx43 protein is shown to be upregulated in retinal zones corresponding to optic nerve injury beginning at 24 hours and peaking at 3 days, with a second peak throughout the retina at 28 days. This study is the first to demonstrate the biphasic response of Cx43 in the superior retina after superior optic nerve transection. The temporal and spatial relationship between Cx43 upregulation, astrocytic activation and RGC loss indicates that retinal glial gap junction communication may have a role in mediating RGC damage. Future studies will be directed at modulating Cx43 expression to determine whether it has an impact on RGC survival. 

Footnotes

 Supported by an Alcon Save Sight Society Grant.

Footnotes

 Disclosure: S.S.L. Chew, None; C.S. Johnson, None; C.R. Green, None; H.V. Danesh-Meyer, None