Astrocytes communicate with each other through gap junctions. Astrocytes have considerable plasticity and play important roles in balancing the normal retinal, ONH, and brain activity by maintaining homeostasis, regulating neuronal activity, and participating in immune responses (Table 1, Figure 2).

Studies using SD rats with chronically elevated IOP induced by episcleral vein cauterization (EVC) found that retinal astrocytes lost GFAP expression 3 days after EVC, however their density increased steadily from 2 weeks to 6 months after EVC (Kanamori et al., 2005). Studies using SD rats or Swiss mice with laser-induced ocular hypertension (OHT) for 2–3 weeks found that astrocytes-occupied retinal area was reduced (Ramírez et al., 2010; Gallego et al., 2012; Table 2, Figure 3).

A study of GFAP-labeled astrocytes in human POAG retinas (donor age, 87.1 ± 6.9 years) found retinal astrocytes are spatial-dependent and heterogeneous. While the peripapillary regions (2 mm from the optic disc) had increased cell density and GFAP expression of astrocytes along large vessels, the mid-peripheral (6–8 mm from the optic disc) and peripheral retinas had reduced cell density and GFAP expression in most astrocytes, increased GFAP expression only found in some individual astrocytic bundles (Wang et al., 2002). Based on the available clinical records, there was no clear relationship between the severity of signs of glaucomatous damage and the changes in astrocytes (Wang et al., 2002). Another study of Cx43/GFAP-labeled astrocytes in human POAG retinas (donor age: 70 and 86 years) found increased connexin43 immunoreactivity in the peripapillary and mid-peripheral retina in association with glial activation (Kerr et al., 2011).

Similar spatial-dependent heterogeneity was found in DBA/2J mouse model of glaucoma, which suggested that astrocyte reactivity (increased density, soma, and GFAP expression) occur in micro-domains in the retina (Formichella et al., 2014). The mean density of astrocytes did not change, but astrocytes are re-distributed to produce lower or higher astrocyte density areas. Surprisingly, most astrocytes had decreased GFAP expression, while only a tiny group of astrocytes exhibited high levels of GFAP expression.

These results indicated that retinal astrocyte activation in glaucoma is heterogeneous, stage-dependent, and spatially dependent. In the early stage of OHT/glaucoma, or the peripheral retina of late-stage glaucoma, cell density and GFAP expression of astrocytes are reduced (Table 2, Figure 3). Astrocyte activation seems only to occur in some retinal areas of later-stage glaucoma. Astrocyte decline in the retina of glaucoma animals may contribute to RGCs loss (Formichella et al., 2014).

The responses of ONH astrocytes to glaucoma are also temporally or stage-dependent (Table 3, Figure 4). A moderate and transient elevation of IOP (30 mmHg for 1 h) by anterior chamber cannulation connected to a saline reservoir induced reactive morphological remodeling of mouse ONH astrocytes that peaked at 3 days, including hypertrophy, process retraction, and simplification of cellular shapes (Sun et al., 2013). There were no significant changes in the gene expression profile, no ectopic cell division, no cell death, and no damage to the optic axons. Interestingly, the morphological remodeling was reversible and recovered from 7 days to 6 weeks post-IOP normalization (Sun et al., 2013). A short-term acute elevation IOP (60 mmHg for 8 h) by anterior chamber cannulation in male Brown Norway rats induced transient and reversible ONH astrocyte actin bundle reorientation without axon damage (Tehrani et al., 2016).

Persistent OHT for 3–10 days can induce morphological remodeling (process retraction) and cell division of ONH astrocytes (Figure 4). Notably, based on these observations, it was proposed that the damage to the RGC axons is not mechanical but is a consequence of localized loss of metabolic support from the ONH astrocytes (Dai et al., 2012). ONH astrocytes in mice with microbeads induced-OHT for 3 days had overall thinner processes, a reduced area fraction of GFAP expression, and an increased number of nuclear clusters (Ling et al., 2020). The junctions between astrocytes and their basement membranes were disrupted, resulting in the reorientation of cell processes, new collagen formation, and cell proliferation after 1 week of OHT in mice (Quillen et al., 2020). Studies using female Albino Swiss rats with anterior chamber magnetic microspheres induced OHT found that withdrawal of astrocytic processes in the dorsal side of the ONH occurs at 7 days of OHT induction, coinciding with the site of initial axon loss. There is no indication of distortion or compression of the axons (Dai et al., 2012). Studies using female Wistar rats with laser induced OHT found that vimentin⁺ astrocytes increased in the post-laminar ONH 10 days after laser treatment, indicating that the astrocyte activation is an early event (Son et al., 2010).

Persistent OHT for 3–14 days can induce morphological remodeling (process retraction and reorientation, loss of gap junction), reduced GFAP expression, and cell division of ONH astrocytes (Figure 4). Studies using male Brown Norway rats with episcleral vein injection of hypertonic saline found that ONH astrocytes were stressed 3–14 days after OHT induction, as evidenced by reduced expression of Cx43 and GFAP and increased expression of PCNA (Johnson et al., 2000). About 2% of ONH cells were divided in this rat model. Rat ONH of ocular hypertension for 5 weeks had more Sox2+ astrocytes than controls, and co-staining of Ki67 and Sox2 proved that 80% (anterior ONH) and 66% (transition zone of ONH) of mitotic cells were astrocytes (Lozano et al., 2019). Reduced gap junction (Cx43 expression), cell division, and decreased GFAP expression are early features of ONH astrocytes, suggesting that ONH astrocytic cytoskeletal reorganization was the early event associated with OHT (Johnson et al., 2000), including the reorientation of ONH astrocyte processes from transverse to longitudinal, as indicated by phalloidin-labeling (Tehrani et al., 2014). GFAP expression increased in ONH astrocytes only at a late stage or with long-term (about 1 month) OHT induction in rats (Johnson et al., 2000).

DBA/2J mouse models of glaucoma are widely used to study ONH astrocytes (Figure 4). ONH astrocytes can phagocytose large axonal evulsions. Some ONH astrocytes at the myelination transition zone (MTZ) express the phagocytosis-related galectin-3 (Gal3). These cells up-regulated Gal3 expression in a γ-synuclein-dependent way in DBA/2J mouse, suggesting that failure to clear axon-derived debris at the MTZ, such as γ-synuclein, may be related to axon loss (Nguyen et al., 2011).

DBA/2J mice also show ONH astrocyte remodeling, including process retraction and sprouting. GFP-labeled astrocytes in DBA/2J mice had thickened and simplified processes. They became smaller and retracted processes, thus having a reduced spatial coverage (Lye-Barthel et al., 2013). DBA/2J mice had diminished ramification and reorientation of astrocyte processes in the early phase of the disease (Cooper et al., 2016, 2018). There was localized sprouting of new processes in the early stages (6 months) of the disease before detectable RGC death. These new processes had no GFAP protein and grew into the axon bundle (Lye-Barthel et al., 2013). In microbeads-induced OHT mice, GFP-labeled MTZ astrocytes extend new processes and invade ONH axon bundles. These longitudinal processes are likely a common feature of glaucomatous ONH astrocytes, as most processes are transverse in naïve ONH astrocytes (Wang et al., 2017). These new longitudinal processes contain sparse intermediate filaments and no subcellular organelles, but some apparently degenerating mitochondria may be of axonal origin (Wang et al., 2017).

In late-stage glaucoma patients, the ONH astrocytes are activated and characterized by morphologic changes (enlarged soma and thicker cellular processes in the prelaminar region, and round-shape cell body and loss of processes in the LC) and increased GFAP expression (Varela and Hernandez, 1997; Hernandez et al., 2000; Wang et al., 2002). The gap junction protein Cx43 also increased in LC and the retina of post-mortem human POAG eyes (Kerr et al., 2011; Figure 4). In a monkey model of glaucoma, ONH astrocytes demonstrate rounding and migration from the core of the cribriform plates as early as 4 weeks after the elevation of IOP (Hernandez and Pena, 1997). Consistent with changes in human glaucoma, ONH astrocytes were also activated in old DBA/2J mice, with increased expression of GFAP and Cx43 (Son et al., 2010; Cooper et al., 2018). The activated astrocytes can form glial scars to cover lesions of axonal degeneration (Bosco et al., 2016).

Most studies regarding astrocytes in the visual brain were performed in the monkey glaucoma model. In monkeys with laser-induced OHT for 2–15 weeks, GFAP expression increased in the lateral geniculate nucleus (LGN), especially in the layers receiving a neuronal input from the high IOP eye (Sasaoka et al., 2008). GFAP expression increased in the LGN and the visual cortex (V1), accompanied by loss of neuronal metabolic activity as assessed with cytochrome oxidase (CO) histochemistry (Lam et al., 2009). Reactive astrogliosis occurred in the magnocellular and parvocellular LGN layers of monkeys with unilateral glaucoma. A quantitative study indicated a linear relationship between increased GFAP expression and ONH axon loss (Dai Y. et al., 2012). LGN astrocyte activation in the monkey glaucoma model can also be detected by PET imaging with [¹¹C] PK11195, a PET ligand for peripheral-type benzodiazepine receptor (PBR). These data suggested that activated glial markers such as PBR in the LGN may be useful noninvasive biomarkers for diagnosing glaucoma (Shimazawa et al., 2012).

In the superior colliculi (SCs) of DBA/2J mouse, loss of anterograde transport of RGC axon was accompanied by hypertrophy of brain-derived neurotrophic factor (BDNF)-expressing astrocytes, which might represent an intrinsic mechanism to mitigate the effects of RGC axon damage (Crish et al., 2013). In male Long-Evans rats with unilateral OHT induced by EVC, pro-inflammatory markers (such as IL-1β and TNFα) and astrocyte activation were identified in hypertensive and normotensive RGC projection sites in the SCs. These results suggested a complicated role of the SCs in the propagation of neuroinflammatory events induced by unilateral OHT (Sapienza et al., 2016). In a ferret model of OHT, loss of neurons and activation of astrocytes occurred in the LGN layers receiving projections from OHT eyes especially in the C layer (Fujishiro et al., 2020).

In summary, the response of astrocytes to glaucoma is quite heterogeneous and is stage, region, and spatially dependent. In the early stage of glaucoma, cell density and GFAP expression of retinal astrocytes are reduced, and the morphological remodeling of ONH astrocytes is reversible. Reduced Cx43 and GFAP expression and cell division are early features of ONH astrocytes. Longitudinal processes with degenerating mitochondria are a common feature of glaucomatous ONH astrocytes. Withdrawal of ONH astrocytic processes may result in localized loss of metabolic support from the ONH astrocytes to RGC axons. The astrocytes in the visual brain also respond to glaucoma insult. Understanding the molecular mechanism of these responses is necessary to develop an effective therapy for glaucoma in the future.

The ONH astrocytes exhibit increased complement C3 expression in DBA/2J mice, without detectable RGC dysfunction and axon degeneration (Harder et al., 2017). Knocking out the C3 gene significantly increased the number of DBA/2J eyes with nerve damage and RGC loss after IOP elevation. The C3-dependent astrocytic response involves EGFR signaling. EGFR inhibitor AG1478 substantially increased the number of eyes with moderate or severe glaucoma. Thus, in the mouse model of glaucoma, ONH astrocytes produce C3 and activate EGFR signaling to support RGC survival (Harder et al., 2017; Figure 2). This result is consistent with observations that patients with primary angle-closure glaucoma (PACG) or POAG have lower plasma C3 levels, which were negatively associated with the disease severity (Li et al., 2017, 2018).

However, inhibition of C3 activation in the retina of DBA/2J mice, by intravitreal injections of AAV2.CR2-Crry, can suppress C3d deposition in RGCs, which results in significant long-term neuroprotection of ONs and the retina. Thus, it is possible that there are compensatory changes following C3 knockout that do not occur with local inhibition of C3 activation (Bosco et al., 2018).

Similarly, the C3 cleavage product C3a can recruit microglia and infiltrating monocytes that express C3a receptor-1 (C3ar1), but C3ar1 is a detrimental neuroinflammatory factor in DBA/2J glaucoma, as C3ar1 knockout lowered the risk for RGC degeneration (Harder et al., 2020). Indeed, C3a/C3 ratio increases in aqueous humor (AH) and serum in progressive POAG patients but not in stable POAG patients (Hubens et al., 2021). In contrast to C3 in astrocytes, C1q upregulation in microglial cells is also an early response to IOP elevation but is detrimental to RGCs as described below (Howell et al., 2011; Guttenplan et al., 2020).

In the CNS, A1 neurotoxic astrocytes are activated by the NF-κB signaling pathway (Lian et al., 2015). In human donor eyes with glaucoma (Yang et al., 2011) and glaucoma rat model (Tezel et al., 2012), astrocyte-specific nuclear factor-kappaB (NF-κB) is critical for the transcriptional regulation of neuroinflammation (Figure 2). NF-κB represents a family of transcription factors, including NF-κB1, NF-κB2, RelA (also named p65), RelB, and c-Rel. The NF-κB proteins are usually sequestered in the cytoplasm by IκB kinases. Thus, the transcriptional activity of NF-κB depends on the degradation of IκB via a process that requires specific kinases (IκK). The IκK-subunit β (IκKβ) is the central activating kinase involved in IκB degradation. GFAP-Cre/ERT2 induced astrocyte-specific knockout of IκKβ or p65 gene reduced pro-inflammatory cytokines (such as TNFα, IFNγ, IL1, and IL2) in the retina and optic nerve of OHT mice. Inhibition of astroglial NF-κB signaling reduced RGC death and improved the retinal function of OHT mice. Interestingly, IκKβ or p65 deletion in astrocytes also reduced pro-inflammatory cytokine production in microglia (Yang et al., 2020).

In the CNS, activated microglia-derived cytokines, including IL-1α, TNFα, and C1q, activate A1 neurotoxic astrocytes (Liddelow et al., 2017). DBA/2J mice with C1qa gene mutation were protected from RGC degeneration (Howell et al., 2011). Triple knockout IL-1α, TNFα, and C1q genes reduced reactive astrogliosis and protected RGCs (especially SMI-32⁺ α-RGCs) from microbeads-induced chronic OHT (Guttenplan et al., 2020; Sterling et al., 2020). Interestingly, subcutaneous injections of NLY01, a long-acting GLP-1R (glucagon-like peptide 1 receptor) agonist, can reduce microglia production of C1q, TNF-α, and IL-1α and RGC loss in this OHT model (Sterling et al., 2020). Ceramides can induce RGC cell death by inducing TNF-α secretion from optic nerve head astrocytes (Fan et al., 2021).

Because the C1qa gene mutation and triple knockout are all germline and not astrocyte-specific, if these protective effects are solely related to the reduced reactive astrogliosis is unknown. Germline knockout of these cytokine genes can cause changes in other cell types, for example, RGCs, Müller glia, and microglia, which may also contribute to these observed protective effects in OHT mice.

Caspase-8 is the initiator caspase of the TNFR-mediated extrinsic apoptosis. It inhibits necroptosis mediated by RIPK3 and MLKL (Fritsch et al., 2019), and promotes NF-κB-mediated cell survival and inflammation (Su et al., 2005). Caspase-8 is detected in the apoptotic RGCs in human and animal models of glaucoma (Yang et al., 2011; Chi et al., 2014). GFAP-Cre/ERT2 induced astrocyte-specific knockout of caspase-8 suppressed neurodegenerative inflammation and protected RGC structure and function in microbeads-induced OHT eyes (Yang et al., 2021; Figure 2).

The astrocytes exhibit increased STAT3 expression in human glaucoma (Yang et al., 2011) and animal glaucoma models (Johnson et al., 2011; Zhang et al., 2013; Lozano et al., 2019). STAT3 is a member of the Jak-STAT signaling pathway and is activated by phosphorylation through cytokines and growth factors. STAT3 is crucial for A2 reactive astrogliosis (Herrmann et al., 2008; Zhang et al., 2013) and mediates the neuroprotective functions of reactive astrocytes (Justicia et al., 2000; Yamauchi et al., 2006; Zhang et al., 2013; Ben Haim et al., 2015; Sun et al., 2017). Indeed, in transient and chronic OHT mice, GFAP-Cre induced astrocyte-specific knockout of STAT3 impaired astrogliosis (loss of the large distinct glial tubes and the overall honeycomb arrangement, hypertrophy of astrocyte processes) in the ONH, and exacerbated RGC injury (losing 20%–30% more RGCs compared with control) and reduction in the positive scotopic threshold response (Sun et al., 2017; Figure 2). Similar results are also observed using the retrobulbar injection of STAT3 inhibitor to rats with transient OHT (Wong et al., 2015). Thus, early reactive astrocyte remodeling in glaucoma is adaptive and beneficial and is essential for the survival of RGCs.

Astrocytic processes are organized in parallel and share cytoplasmic information via communication through gap junctions (GJs) comprised primarily of connexin 43 (Cx43). Gap junctions also allow functional coordination among the members of the neurovascular unit, such as glial cells, endothelial cells, pericytes, and neurons. Astrocytes create and store glycogen as a safeguard against stress. They provide nutritional supplements and energy resources for surrounding neurons (Vecino et al., 2016). This metabolic collaboration between astrocytes and neurons is important to tissue survival during injury. Cx43 expression decreases during the first 2 weeks after OHT induction (Johnson et al., 2000), but increases at the late stage of human glaucoma (Kerr et al., 2011), and DBA/2J mouse model of glaucoma (Son et al., 2010; Cooper et al., 2018). There is evidence that elevated IOP activates C3 and EGFR signaling, which is beneficial to RGCs at the early stage of the disease (Harder et al., 2017). However, EGFR signaling can phosphorylate Cx43 and shut down gap junctions, thus disrupting the intercellular communication between astrocytes (Malone et al., 2007). So, both early and late stages of glaucoma may suffer from the reduced function of the gap junction.

GFAP-Cre-ERT2 induced astrocyte-specific knockout of Cx43 gene (Cx43 CKO) in the retina, optic nerve, and superior colliculi (SCs). These Cx43 CKO mice proved that astrocytes in the OHT-stressed ONH can receive glycogen from other astrocytes in the un-stressed contralateral ONH via gap junction. Thus, Cx43 CKO eyes lost visual function much faster than controls upon unilateral IOP elevation. These results demonstrate that glucose and its metabolites are redistributed from healthy to stressed ONH through an astrocyte Cx43-mediated network within the optic projection (Cooper et al., 2020). However, GFAP-Cre induced Cx43 knockout indicated that gap junctional coupling between retinal astrocytes exacerbates RGCs loss in ischemia-reperfusion retinal injury (Toychiev et al., 2021). In addition, in contrast to these beneficial effects of astrocyte gap junction in glaucoma, neuronal gap junctions are detrimental to RGCs in glaucoma, as pharmacological blockade of gap junctions or knock out connexin 36 (Cx36) subunits, which are highly expressed in retinal neurons, protected RGCs and optic nerve axons in a mouse model of glaucoma (Akopian et al., 2017). The possible reason is that gap junctions can form conduits through which toxic molecules from dying cells pass to and harm coupled neighbor neurons.

In animal models of glaucoma, the mitochondria of many cell types (such as astrocytes, Müller glia, and RGCs) become dysfunctional, resulting in energetic deficits, increased ROS, and calcium imbalance. Mitophagy can remove damaged mitochondria and is the major mechanism for mitochondrial quality control. Impaired mitophagy can cause neurodegenerative diseases such as Parkinson’s disease (Geisler et al., 2010; Vives-Bauza et al., 2010). Mitochondrial uncoupling proteins (Ucp) can uncouple the electron transport chain from ATP synthase activity and decrease membrane potential (Ψm). The mitochondrial uncoupling protein 2 gene (Ucp2) is expressed in the retina, can decrease electron transport chain efficiency and ROS level, and increase energy expenditure (Arsenijevic et al., 2000). GFAP-Cre-ERT2 induced astrocyte-specific knockout of Ucp2 gene, increased levels of mitophagy in the retina, decreased oxidative protein modification, and reduced RGC death in microbeads-induced OHT mice. Ucp2 deletion facilitates increased mitochondrial function by improving quality control (Hass and Barnstable, 2019; Figure 2). These results are consistent with findings that overexpression of the E3 ubiquitin ligase parkin, or optic atrophy type 1 (OPA1) can restore dysfunctional mitophagy in OHT rats and increase RGC survival (Dai et al., 2018; Hu et al., 2018). Astrocytes can internalize damaged RGC mitochondria (Davis et al., 2014). Ucp2 knockout, which enhances glial mitophagy, may also enhance trans-cellular degradation of damaged RGC mitochondria or other components to protect against mitochondrial damage in glaucoma (Hass and Barnstable, 2019).

In summary, these studies suggested that astrocytes have both beneficial and detrimental functions to RGCs in glaucoma. They can regulate neuroinflammation, mitochondrial dysfunction, and energy redistribution in glaucoma (Table 4).