Cx43 is mainly expressed in astrocytes in mouse retina
We first examined the expression and distribution of Cx43 in mouse retina. Figure 1 shows that Cx43 was mainly distributed in the retinal nerve fiber layer (NFL). Double immunostaining further showed that Cx43 was not co-labeled with brain-specific homeobox/POU domain protein 3A (Brn3a), a marker of RGCs (Fig. 1A, a1-a3), but was predominantly co-labeled with glial fibrillary acidic protein (GFAP), a marker of astrocytes (Fig. 1B, b1-b3). A small amount of Cx43 was expressed in the end-feet of Müller cells, as shown the co-localization of Cx43 and glutamine synthase (GS, a marker of Müller cells) in the NFL (Fig. 1C, c1-c3), and in microglia labeled by ionized calcium binding adapter molecule 1 (Iba-1) in the inner plexiform layer (IPL) (Fig. 1D, d1-d3).
Up-regulation of Cx43 phosphorylation in COH retinas
Dynamic changes in the protein levels of total and phosphorylated Cx43 in COH retinas were then examined. The COH mouse model was successfully produced, with the IOPs in the operated eyes ranging from 14.9 ± 0.3 mmHg at G4d (n = 41, all P < 0.001) to 17.7 ± 0.4 mmHg at G4w (n = 28, all P < 0.001), which was significantly higher than that of 0d or the corresponding un-operated eyes (9.93 ± 0.02, n = 141), and the sham-operated eyes (9.91 ± 0.02, n = 150) (Fig. 2A).
The phosphorylation of Cx43 at Ser373 site (p-Cx43Ser373) and at Ser368 site (p-Cx43Ser368) are two important modulated ways, which may affect Cx43 hemichannel functions [25–29]. In our COH retinas, total protein levels of Cx43 were significantly decreased at G4d and G1w (n = 6, P < 0.01 and 0.001 vs. control (Ctr), respectively), then returned to the control levels at G2w and G3w, and decreased again at G4w (n = 6, P < 0.01 vs. Ctr) (Fig. 2B, C). The expression of p-Cx43Ser373 was significantly decreased at G1w (n = 6, P < 0.05 vs. Ctr) and then quickly returned to the control levels (Fig. 2B, D), while the levels of p-Cx43Ser368 did not show significant change as compared with Ctr at G4d and G1w, followed by significant decrease from G2w to G4w (n = 6, P < 0.05 and 0.01 vs. Ctr) (Fig. 2B, E). These changes resulted in a significant increase of the ratio of p-Cx43Ser373/Cx43 (red dash line) at G4w (P < 0.001) (Fig. 2D) and an up-regulation of the ratio of p-Cx43Ser368/Cx43 (red dash line) at G4d and G1w (P < 0.01 and 0.001, respectively) (Fig. 2E). Since Cx43 proteins only expressed on cell membrane could form functional hemichannels, we further detected the Cx43 protein levels in the membrane component. As shown in Fig. 2F-I, changes of Cx43, p-Cx43Ser373, p-Cx43Ser368, ratios of p-Cx43Ser373/Cx43 and p-Cx43Ser368/Cx43 in membrane proteins of COH retinas were basically consistent with those in total retinal proteins, suggesting that the functions of Cx43 hemichannels were modulated after IOP elevation.
Dynamic changes of Cx43 expression caused by astrocyte plasticity in the ONH of COH mice
In the ONH, astrocytes form functional syncytia through the gap junctions constructed by Cx43, thereby communicating with RGCs and maintaining the ion and metabolic homeostasis of RGCs [10]. To test whether astrocytes may be responded to the changes of Cx43 after IOP elevation, we examined the activation of astrocytes in COH retinas by c-Fos immunostaining, an immediate early gene. As shown in Fig. 3, weak c-Fos positive fluorescent signals were detected in the ONH of sham-operated retinas (control, Ctr) (Fig. 3A, a1), and more c-Fos positive signals appeared at G4d in COH mice and then the number of positive signals was progressively increased until G4w (Fig. 3Bb1, 3Cc1, 3Dd1, 3Ee1, and 3Ff1). The c-Fos positive signals were co-located with GFAP (Fig. 3Bb3, 3Cc3, 3Dd3, 3Ee3, and 3Ff3). In addition, the cells in the inner and outer nuclear layers (INL and ONL) began to be activated at G1w, and the c-Fos positive signals in RGCs were seen from G2w (Fig. 3D-3F, arrows). These results suggest the astrocyte activation was prior to neurons in COH retinas. The increased protein levels of GFAP indicate that macroglial cells were activated after IOP elevation in COH retinas (n = 6, P < 0.05∼0.001 vs. Ctr) (Fig. 3G, H).
Then, we detected the astrocyte plasticity in the ONH after IOP elevation. Growth-associated protein 43 (Gap43), an activity-dependent plasticity protein [30], is implicated in axonal plasticity and regeneration [31]. We first identified the glial cell types that express Gap43 in the ONH. Our results showed that Gap43 was mainly expressed in astrocytes labeled by GFAP (Fig. 4A), scarce in Müller cells labeled by GS (Fig. 4B), less in oligodendrocytes labeled by O4 (Fig. 4C) and microglia labeled by Iba-1 (Fig. 4D). We then detected the expression of Cx43 in the Gap43 positive cells by double immunostaining (Fig. 4E). Although the number of Gap43 labeled cells kept unchanged during the whole period of IOP elevation (G4d-G4w) (n = 5 ~ 6, P > 0.05 vs. Ctr) as compared with the controls (Fig. 4E, F), the fluorescent intensity of Gap43 was increased in COH retinas (Fig. 4E, e8-e12). The number of Gap43 and Cx43 double labeled cells was significantly decreased from G4d to G2w in COH retinas (n = 5 ~ 7, P < 0.05 and 0.01 vs. Ctr), and then returned to the control levels (Fig. 4E, G). Similarly, Western blotting experiments showed that the protein levels of Cx43 in the ONH were decreased from G4d to G2w (n = 5, P < 0.01 vs. Ctr) (Fig. 4H, I). These results suggest that the plasticity of astrocytes in the ONH may cause the reduced expression of Cx43 during IOP elevation.
Rac1 regulates astrocyte response in COH retinas
The regulation of cytoskeleton and differentiation is involved in the plasticity of astrocytes. Previous studies have shown that Rac1, a GTPase of the Rho family, was specialized in the regulation of actin cytoskeleton dynamics [32, 33], which was involved in the pathogenesis of glaucoma. We examined whether Rac1 may be involved in the regulation of astrocyte activation in COH retina. As shown in Fig. 5A and 5B, the ratio of active Rac1 to Rac1 in COH retina was significantly increased at G4d and G3w (n = 6, P < 0.01 and 0.001 vs. Ctr, respectively), similar to our previous report [16]. Additionally, the expression of GFAP was increased in COH retina (normal saline (NS) G1w) (n = 6, P < 0.001 vs. Ctr), which could be largely reversed by NSC23766 (NSC) administration, an inhibitor of Rac1 (n = 6, P < 0.001 vs. Ctr and NS G1w) (Fig. 5C, D). IOP elevation in COH retina induced morphological changes of astrocyte from the smaller cell bodies with slender processes in control condition to the larger cell bodies with more branches and hypertrophic protrusion. These morphological changes of astrocytes could be partially reversed by the Rac1 inhibitor NSC23766 (n = 5, P < 0.001 vs. Ctr, P < 0.01 vs. NS G1w group) (Fig. 5E, F). Furthermore, inhibition of Rac1 by NSC23766 significantly increased the EtBr uptake of astrocytes in COH retinas (n = 6, P < 0.01 vs. NS G1w group), suggesting that Rac1 regulates the permeation of hemichannels in astrocytes. The increased permeation mediated by Rac1 inhibition was blocked by the Cx43 inhibitor Gap26 (n = 6, P < 0.05 vs. NSC G1w group), further demonstrating that it was the hemichannels mediated the EtBr uptake in astrocytes (Fig. 5G, H).
The above results show that changes in expression of Rac1 in COH retinas are opposite to those of Cx43. It is possible that Rac1 may regulate the expression of Cx43. To test this possibility, the Rac1 inhibitor NSC23766 was intravitreally injected 2 days prior to the operation of COH model. As shown in Fig. 6A and 6B, the total protein levels of Cx43 were significantly increased in COH retinas when Rac1 was inhibited (n = 6, P < 0.001 vs. NS G1w). At the same time, the expression of p-Cx43Ser373 was increased (Fig. 6A and 6C), while the expression of p-Cx43Ser368 was decreased (n = 5, P all < 0.05 vs. NS G1w) (Fig. 6A and 6D). Furthermore, the protein levels of Cx43 in membrane component were also significantly increased (n = 7, P < 0.001 vs. NS G1w). PAK1 is a downstream molecule of Rac1. Pre-injection of the Rac1 inhibitor NSC23766 significantly reduced the protein levels of active Rac1 and p-PAK1 in COH retinas, which resulted in decreased ratios of active Rac1/Rac1 and p-PAK1/PAK1 (n = 6, P < 0.01 and 0.05, respectively) (Fig. S1, A-C). In addition, pre-injection of the PAK1 inhibitor IPA-3 (4 mg/kg, i.p.) significantly reduced the protein levels of PAK1 and p-PAK1 in COH retinas (n = 6, P all < 0.001 vs. NS G1w) (Fig. S1, D-G). Moreover, although pre-injection of IPA-3 did not affect the protein levels of Cx43, the protein levels of p-Cx43Ser373 were up-regulated (n = 5, P < 0.001 vs. NS G1w) and the protein levels of p-Cx43Ser368 were down-regulated (n = 6, P < 0.01 vs. NS G1w), similar to that of the Rac1 inhibition (Fig. 6G-J). Co-IP experiments further revealed that there were interactions between Cx43 and active Rac1 or p-PAK1 (Fig. 6K, L). These results suggest that Rac1/PAK1 signaling pathway may directly regulate the phosphorylation of Cx43.
Rac1-mediated ATP release from astrocytes
In COH retinas, extracellular ATP concentrations were progressively increased from G2w to G4w (n = 5∼7, P < 0.05∼0.001 vs. Ctr) (Fig. 7A). However, when the ecto-ATPase inhibitor ARL67156 was added to the tissue lysate to prevent ATP degradation, a significant increase of ATP concentration was also observed at G4d and G1w (n = 6, P all < 0.01 vs. Ctr) (Fig. 7B), suggestive of a quick degradation of ATP at early stage of IOP elevation. Furthermore, inhibiting the activation of Rac1 by NSC23766 promoted the release of ATP at G1w and G2w (n = 6, P all < 0.01 vs. NS groups) (Fig. 7C, D), while inhibition of Cx43 by Gap26 or Gap19 significantly reduced the ATP levels (n = 6, P < 0.05 and 0.01, respectively) (Fig. 7C, D), suggesting that Rac1 regulates ATP release in COH retinas through affecting Cx43 hemichannels. Cx43 is expressed on astrocytes, Müller cells and microglia. In order to determine whether inhibition of Rac1 increase ATP release from microglia, a single dose of clodronate liposomes (Clo-lip) was intraperitoneally injected one day before the COH operation to remove microglia. Clo-lip treatment could remove about 66.1 ± 2.9 % of microglia in retina (Fig. S2). Figure 7E shows that removing retinal microglia did not affect the increased ATP levels caused by the Rac1 inhibitor NSC23766 in COH retinas, indicating that microglia was not the source of Rac1-mediated ATP release (Fig. 7E). In order to identify whether ATP may be released from astrocytes, the AAV2-EFS-DIO-ATP1.0 probe was injected into the retinas of GFAP-Cre mice to make the ATP probe express specifically in astrocytes (Fig. 7F). Three weeks after the AAV injection, NSC23766 was intravitreally injected before the COH operation. The fluorescent signals of GFP were detected in the living retina at G1w. We observed that the Rac1 inhibitor NSC23766 significantly increased the fluorescent signal density (n = 45 cells of 3 retinas, P < 0.001), as compared with the controls (NS) (n = 43 cells of 3 retinas) (Fig. 7G, H). These results indicate that GFAP-positive astrocytes contribute to the increased extracellular ATP concentrations.
We then explored whether conditional knockout of Rac1 in astrocytes may affect Cx43-mediated ATP release. Rac1 knockout in astrocytes was constructed by intravitreally injecting GFAP-Cre-AAV into Rac1flox/flox mice. The efficiency of Rac1 knockout in astrocytes was shown in Fig. S3. Compared with GFP-AAV-injected group, the number of GFAP and Rac1 co-localized positive signals was reduced in GFAP-Cre-AAV injected group (Fig. S3). The protein levels of Cx43 were increased in COH retinas at G1w in the Rac1 conditional knockout mice (n = 6, P < 0.01 vs. GFP-AAV group) (Fig. 7I, J), and the ATP concentrations were also increased in COH retinas of the Rac1 conditional knockout mice (Fig. 7K).
Rac1-mediated ATP release increases the survival of RGCs in COH retinas
What’s the role of the Rac1-mediated ATP release from astrocytes on RGC survival in glaucoma? We assayed the survival of RGCs by Brn3a immunostaining in whole flat-mounted retinas. Figure 8A shows representative images obtained from central and peripheral areas of retinas at the same angle (Fig. 8B). The average number of RGCs in both the central and peripheral areas of COH retinas at G1w was significantly increased in the Rac1 conditional knockout mice (n = 5, P < 0.001 and 0.01 vs. GFP AAV group, respectively), which could be blocked by Gap26 administration (n = 6, P < 0.001 and 0.05 vs. GFAP Cre AAV group, respectively) (Fig. 8C). These results suggest that conditional knockout of Rac1 in astrocytes may increase RGC survival through modulating Cx43.
We then explored how conditional knockout of Rac1 in astrocytes protects RGCs in COH retina. As shown in Fig. 8D, adenosine A1 receptor (A1R) and A3 receptor (A3R) were expressed in Brn3a-positive RGCs. IOP elevation did not change the expression of A1R (d2) and A3R (d5). Similar results were observed when Rac1 was inhibited by the NSC23766 injection in COH retinas (Fig. 8D, d3). However, injection of NSC23766 remarkably augmented the expression of A3R in Brn3a-positive RGCs in COH retinas (Fig. 8D, d6), which was further confirmed in conditional knockout of Rac1 in astrocytes (Fig. 8E). Furthermore, in the Rac1 conditional knockout mice, intravitreal injection of CGS 15943, an adenosine receptor antagonist, partially attenuated the increase in the number of Brn3a positive RGCs in both central and peripheral areas of COH retinas (n = 6, P < 0.01 and 0.001 vs. GFAP AAV Cre groups, respectively) (Fig. 8F-H). All these results suggest that Rac1-mediated ATP release increases the survival of RGCs through activating adenosine receptors.