Abstract
The optic nerve conveys information about the outside world from the retina to multiple subcortical relay centers. Until recently, the optic nerve was widely believed to be incapable of re-growing if injured, with dire consequences for victims of traumatic, ischemic, or neurodegenerative diseases of this pathway. Over the past 10–20 years, research from our lab and others has made considerable progress in defining factors that normally suppress axon regeneration and the ability of retinal ganglion cells, the projection neurons of the retina, to survive after nerve injury. Here we describe research from our lab on the role of inflammation-derived growth factors, suppression of inter-cellular signals among diverse retinal cell types, and combinatorial therapies, along with related studies from other labs, that enable animals with optic nerve injury to regenerate damaged retinal axons back to the brain. These studies raise the possibility that vision might one day be restored to people with optic nerve damage.
Keywords
Introduction
Vision relies on high-fidelity signal conduction from the retina to subcortical target areas via the optic nerve, a white-matter tract containing bundles of axons that originate in retinal ganglion cells (RGCs). Like most neurons in the central nervous system (CNS), RGCs cannot regenerate damaged axons, and consequently, optic nerve damage resulting from incidents of head trauma, ischemia, or glaucoma generally lead to a permanent loss of vision. In addition to discovering potential treatments for optic neuropathies, the rodent optic nerve has been a valuable model for understanding and counteracting the general inability of mature CNS neurons to regenerate axons after injury (Li, Schlamp, & Nickells, 1999; McKinnon, Schlamp, & Nickells, 2009). Several strategies discovered in the optic nerve model have already proven to be effective in promoting axon regeneration in the injured spinal cord and may one day lead to therapies for multiple traumatic, ischemic or neurodegenerative conditions.
In this chapter, we summarize efforts by our group and others to regenerate the optic nerve, including the discovery of multiple cell-intrinsic and cell-extrinsic factors that promote or suppress RGC survival and axon regeneration, and conclude with studies pointing to the feasibility of regenerating injured axons all the way from the retina to appropriate areas of the brain.
Nerve regeneration in lower vertebrates
Unlike mammals, cold-blooded vertebrates such as fish and frogs can spontaneously regenerate their optic nerves and recover visually guided behaviors throughout life (Grant & Keating, 1992; Sperry, 1948). This phenomenon has been attributed to a less inhibitory environment within the nerve after transection (e.g., less scar formation at the injury site) (Becker & Becker, 2002; Lee-Liu, Mendez-Olivos, Munoz, & Larrain, 2017), de-differentiation of optic nerve oligodendrocytes (Ankerhold, Leppert, Bastmeyer, & Stuermer, 1998), expression of growth-promoting cues by retinal astrocytes (Hirsch, Cahill, & Stuermer, 1995), and a stronger cell-intrinsic capacity for growth (Abdesselem, Shypitsyna, Solis, Bodrikov, & Stuermer, 2009; Becker & Becker, 2014; Benowitz & Lewis, 1983; Benowitz, Shashoua, & Yoon, 1981; Benowitz, Yoon, & Lewis, 1983; Bodrikov et al., 2017; Elsaeidi, Bemben, Zhao, & Goldman, 2014; Kusik, Hammond, & Udvadia, 2010; Priscilla & Szaro, 2019; Skene & Willard, 1981b; Veldman, Bemben, & Goldman, 2010; Veldman, Bemben, Thompson, & Goldman, 2007). Moreover, unlike the near-complete loss seen after optic nerve injury in mammals, optic nerve injury leads to only modest RGC death in zebrafish and frog, (Bernhardt, Tongiorgi, Anzini, & Schachner, 1996; Liu, Yu, Deaton, & Szaro, 2012; Zou, Tian, Ge, & Hu, 2013). Cold-blooded vertebrates can even regenerate the retina following partial injury, as Müller cells, the radial glia of the retina, de-differentiate to produce progenitor cells that replace lost tissue (Fausett & Goldman, 2006). After ablation of the inner retina, Müller cells can give rise to new RGCs that extend axons down the optic nerve (Fimbel, Montgomery, Burket, & Hyde, 2007; Reier & Webster, 1974). Mammalian Müller glia initially exhibit a similar response to injury by upregulating markers associated with cell stress and becoming hypertrophied (Bringmann et al., 2009; Thomas, Ranski, Morgan, & Thummel, 2016); however, this process arrests and results in gliotic scarring, whereas zebrafish and Xenopus Müller glia, triggered by post-injury retinal signaling, dedifferentiate into a retinal progenitor-like state and go on to replace lost neurons (Goldman, 2014; Langhe et al., 2017; Nelson et al., 2013; Wan & Goldman, 2016; Wan, Ramachandran, & Goldman, 2012; Zhao et al., 2014). Insights from zebrafish have been successfully applied to stimulate regeneration in the mammalian retina, indicating that exogenous factors can override the loss of an intrinsic regenerative capacity in mammals (Jorstad et al., 2017; Ueki et al., 2015; Yao et al., 2018).
Mammalian RGC axon regeneration through a peripheral nerve graft
In his monumental study, The Degeneration and Regeneration of the Nervous System (1928), Cajal described experiments by his student Tello showing that some RGCs of the adult rabbit were able to regenerate their axons into a peripheral nerve graft attached to the cut end of the optic nerve. These observations provided the first evidence that, under certain experimental conditions, CNS neurons possess some regenerative ability (Ramon y Cajal, 1991), and planted the seeds for modern research into CNS regeneration (Fawcett, 2018). Many decades after Tello’s experiment, the PNS nerve graft strategy was revisited by Aguayo and colleagues, who observed that axons grew into a grafted sciatic nerve fragment following complete removal of a segment of the thoracic spinal cord (Richardson, McGuinness, & Aguayo, 1980), or a nerve fragment bridging the spinal cord and medulla (David & Aguayo, 1981), or, confirming Tello’s findings, a nerve graft connecting the retina to the brain (Aguayo et al., 1991). In the latter case, some RGC axons projected through the graft to form synapses in the superior colliculus (So & Aguayo, 1985; Vidal-Sanz, Bray, Villegas-Perez, Thanos, & Aguayo, 1987). However, a major limitation to the peripheral nerve graft strategy is that CNS axons fail to continue growing beyond the graft, limiting the potential for functional recovery. The use of combinatorial treatments has had some success in overcoming this problem (Drummond et al., 2014; Lee, Rosen, Weinstein, van Rooijen, & Noble, 2011; Theisen et al., 2017). Although the clinical future of this approach is yet uncertain, these studies had an immediate and enormous impact in the field.
Axon regeneration through the injured optic nerve
Although the peripheral nerve graft experiments showed that mature RGCs retain some capacity for axon growth, regeneration through the optic nerve itself was long considered to be impossible. Schwab and others showed that CNS myelin is considerably more inhibitory to axon growth than PNS myelin (Caroni, Savio, & Schwab, 1988; Schwab, 1990; Schwab & Caroni, 1988), and this difference was widely thought to be the primary cause of regenerative failure in the CNS. However, Berry’s group proposed that trophic factors produced by Schwann cells were the basis of RGC axon regeneration through peripheral nerve grafts, and went on to show that implanting an autologous fragment of peripheral nerve into the vitreous chamber of the rat eye induced extensive axon regeneration through the optic nerve itself (Berry, Carlile, & Hunter, 1996). These studies provided the first evidence that the optic nerve was not an insuperable barrier to regeneration provided that RGCs were adequately stimulated. One factor that was initially overlooked, however, was the possible contribution of the numerous immune cells seen in the implants (Berry et al., 1996).
Intraocular inflammation induces optic nerve regeneration
Our lab found that an unintentional injury to the lens induced considerable axon regeneration, and that this effect could be mimicked by inducing intraocular inflammation with zymosan (Leon, Yin, Nguyen, Irwin, & Benowitz, 2000; Yin et al., 2003). This regeneration was found to be associated with a change in RGCs’ intrinsic growth state, as evidenced by a massive up-regulation of SPRR1A, GAP-43, and other growth-associated proteins (GAPs, or regeneration-associated gene products, RAGs) in a pattern similar to that seen during peripheral nerve regeneration (Fischer, Petkova, Thanos, & Benowitz, 2004). Genetic deletion of two receptors that are expressed by inflammatory cells, Toll-like receptor 2 (TLR2) and dectin-1, eliminates the pro-regenerative effects of Zymosan despite not altering the general profile of infiltrative cells (Baldwin, Carbajal, Segal, & Giger, 2015). β-glucan is a component of Zymosan that stimulates cells of the innate immune system via dectin-1, and curdlan, a particulate form of β-glucan, mimics the effects of Zymosan on regeneration. These studies raise the question of whether the positive factors associated with inflammation can be identified to promote regeneration in a clinically useful way.
Oncomodulin and SDF1
Our earlier work showed that the carbohydrate mannose, which is abundant in the vitreous and cerebrospinal fluid, stimulates appreciable axon growth from goldfish RGCs and moderate outgrowth from cultured mature rat RGCs. These effects require elevation of cAMP, and are strongly augmented by a protein secreted by activated macrophages (Li et al., 2003; Yin et al., 2003). Using column chromatography, mass spectrometry and bioassays, we identified the 11 kDa Ca2+-binding protein Oncomodulin (Ocm) as a major growth-promoting factor associated with inflammation (Yin et al., 2006). Ocm is secreted by infiltrative neutrophils and macrophages and accumulates in the neural retina 12–24 h after inducing intraocular inflammation by binding to a high-affinity receptor on RGCs (Kd ∼ 28 nM) (Kurimoto et al., 2010, 2013; Yin et al., 2006). Elevation of cAMP alone induces only modest axon regeneration (Monsul et al., 2004), but is required for Ocm and other trophic factors to bind to their cognate receptors on RGCs (Kurimoto et al., 2010; Meyer-Franke et al., 1998; Shen, Wiemelt, McMorris, & Barres, 1999; Yin et al., 2006). Delivery of Ocm and a cAMP analog via slow-release polymer beads mimics the pro-regenerative effects of zymosan, whereas blocking the effects of Ocm with either a neutralizing antibody or a blocking peptide strongly suppresses the effects of zymosan (Fig. 1)(Kurimoto et al., 2013; Yin et al., 2006, 2009). Regeneration is also diminished by immune-depletion of neutrophils, implying that these first responders of the innate immune system mediate most of the effects of inflammation on optic nerve regeneration (Kurimoto et al., 2013). Ocm has also been reported to synergistically promote axon outgrowth in RGCs when combined with a small interfering RNA against the Nogo-66 receptor (Cui, Kang, Hu, Zhou, & Wang, 2014), and to contribute to the “conditioning lesion” effect in the peripheral nervous system, enabling injured dorsal root ganglion (DRG) neurons to extend axons following the accumulation of infiltrative cells into peripheral nerves and DRGs (Kwon et al., 2013).

Oncomodulin mediates much of the effect of inflammation on optic nerve regeneration. A, B. Cross-sections of mouse retina immunostained for oncomodulin (Ocm) in a normal adult mouse (A) and in a mouse receiving intraocular Zymosan 24 hr earlier (B). C, D. Cells extracted from the vitreal chamber 12 hr after Zymosan stained for Gr1, a neutrophil marker (C), Ocm (D), and DAPI (C). Inset: infiltrative cells with polymorphic nuclei. E-H. Sections through the mature rat optic nerve immunostained for GAP-43 2 weeks after optic nerve injury. Blank polymer beads induce a small amount of regeneration (E). Inclusion of Ocm plus CPT-cAMP enhances growth greatly (F). The effects of intraocular inflammation (by lens injury, LI) are blocked by P1, a peptide antagonist of Ocm (I), but not by P3, a control peptide (J) (Kurimoto et al., 2013; Yin et al., 2009; Yin et al., 2006).
A second growth factor associated with intraocular inflammation is the chemokine stromal cell-derived factor 1 (SDF1, CXCL12). SDF1 acts through the receptor CXCR4, which is expressed in neurons, inflammatory cells, and other cell types (Khan et al., 2018; Reaux-Le Goazigo, Van Steenwinckel, Rostene, & Melik Parsadaniantz, 2013; Wang et al., 2017), as well as through CXCR7 (Balabanian et al., 2005). SDF1 has a wide range of effects on CNS development and hematopoiesis (Belmadani et al., 2005; Nagasawa et al., 1996). It is highly expressed in infiltrative macrophages and acts synergistically with Ocm to induce optic nerve regeneration (Yin et al., 2018). Deletion of SDF1 in myeloid cells (using CXCL12fl/fl-LysMCre -/+ mice) or deletion of its receptor CXCR4 in RGCs (CXCR4fl/fl mice injected intraocularly with AAV2-Cre virus) diminishes inflammation-induced optic nerve regeneration by ∼ 1/3 and completely eliminates the benefits of inflammation on RGC survival (Yin et al., 2018). Blockade of both Ocm and SDF1 decreases inflammation-induced regeneration by 70–80% (Yin et al., 2018). In gain-of-function experiments, although SDF-1 has only modest effects on regeneration by itself (Heskamp et al., 2013; Yin et al., 2012), SDF-1 combined with Ocm and cAMP mimics most of the pro-regenerative effects of intraocular inflammation (Yin et al., 2018). The level of SDF1 associated with intraocular inflammation appears to be below optimal levels, as adding exogenous SDF1 to intraocular inflammation doubles the number of axons that regenerate the full length of the optic nerve and increases the number of axons that extend through the optic chiasm, optic tract, and into the dorsal lateral geniculate nucleus (LGN) (Yin et al., 2018). SDF1 exerts its effects by activating PI3K signaling, elevating intracellular cAMP, and antagonizing the repellant effects of slit/robo (Chalasani et al., 2007; Yin et al., 2018, 2012). Thus, Ocm and SDF1 are two of the major pro-regenerative constituents of intraocular inflammation and together may be useful in promoting optic nerve repair clinically.
Ciliary neurotrophic factor (CNTF) has also been proposed to mediate the effects of intraocular inflammation on axon regeneration (Muller, Hauk, & Fischer, 2007). Although CNTF and other cardiotrophin family chemokines become elevated in the eye after intraocular inflammation (Kurimoto et al., 2013; Leibinger et al., 2009), at physiological concentrations, recombinant CNTF (rCNTF) by itself has little axon-promoting effects on RGCs in cell culture (Cohen, Bray, & Aguayo, 1994; Yin et al., 2006, 2009) and weak or no effects on optic nerve regeneration in vivo (Leon et al., 2000; Lingor et al., 2008; Pernet & Di Polo, 2006; Qin, Zou, & Zhang, 2013; Smith et al., 2009; Weise et al., 2000). In the peripheral nerve grafting paradigm, although high concentrations of CNTF augment axon regeneration (Cui & Harvey, 2000), these effects may be largely due to the chemotactic effects of CNTF on macrophages (Blanco, Vega-Melendez, De La Rosa-Reyes, Del Cueto, & Blagburn, 2019; Cen et al., 2007). In contrast to recombinant CNTF, AAV2-mediated CNTF delivery induces considerable axon regeneration through the optic nerve, but this effect is associated with an infiltration of inflammatory cells in the eye that express Ocm, SDF1, and other trophic factors (Xie, Yin, Gilbert, & Benowitz, 2019). One reason for the low efficacy of CNTF alone is that SOCS3, a repressor of the Jak-STAT signaling pathway, increases postnatally and increases even further after optic nerve injury (Fischer, Petkova, et al., 2004; Park et al., 2009; Qin et al., 2013). Accordingly, deletion of SOCS3 amplifies the effects of rCNTF (Smith et al., 2009). In addition, mature RGCs do not express detectable CNTFRα, the specific receptor subunit for CNTF (Y. Yin, R. Kawaguchi, D. Geschwind, G. Coppola, L. Benowitz, unpublished), although elsewhere in the nervous system, this subunit can be released from one type of cell and become anchored to another cell type via a glycosylphosphatidylinositol linkage to form part of a tripartite receptor complex with LIFRβ and glycoprotein 130 (gp130) (Davis et al., 1993). CNTFRα is heavily expressed on astrocytes and inflammatory cells, and it will be important to determine the extent to which the effects of AAV2-CNTF on optic nerve regeneration are mediated by factors secreted by these cells. Nonetheless, CNTF and related trophic factors do appear to play an important role in this system, as double deletion of CNTF and leukemia inhibitory factor (LIF) accelerates RGC death after optic nerve injury and prevents regeneration (Leibinger et al., 2009).
Regarding other factors, fibroblast growth factor-2 (FGF2) stimulates some axon regeneration (Sapieha, Peltier, Rendahl, Manning, & Di Polo, 2003) and the neurotrophic factor BDNF increases RGC survival after optic nerve injury, albeit transiently, and paradoxically nullifies the effects of intraocular inflammation on optic nerve regeneration (Mansour-Robaey, Clarke, Wang, Bray, & Aguayo, 1994; Almasieh, Wilson, Morquette, Cueva Vargas, & Di Polo, 2012; Di Polo, Aigner, Dunn, Bray, & Aguayo, 1998; Pernet & Di Polo, 2006). The transient neuroprotection afforded by BDNF is partly due to the downregulation of BDNF’s cognate receptor, TrkB, in RGCs after injury (Cheng, Sapieha, Kittlerova, Hauswirth, & Di Polo, 2002; Cui, Tang, Hu, So, & Yip, 2002), and overexpression of TrkB leads to enhanced survival of RGCs in vitro and in vivo (Cheng et al., 2002; Hu, Cho, & Goldberg, 2010; Martin et al., 2003; Osborne et al., 2018). Insulin-like growth factor (IGF-1) combined with osteopontin enhances axon regeneration selectively in α-RGCs, which represent ∼ 6% of the total RGC population (Duan et al., 2015). GDNF and related receptors have a profound effect on RGC survival, too, and their effects on optic nerve regeneration is not yet clear (Koeberle & Ball, 1998, 2002).
During development, aging, and certain pathological conditions, cells can reduce their intrinsic responsiveness to particular extracellular signals (Goldstein, 2002) via let-7-mediated inhibition of the PI3K-mTOR signaling pathway (Zhu et al., 2011). The RNA-binding protein Lin28 is highly expressed in embryonic stem cells and is suppressed in most adult tissues (Moss & Tang, 2003; Yang & Moss, 2003). Overexpression of Lin28 eliminates the inhibitory effects of let-7, promotes an insulin-sensitized state, and enhances expression of IGF1 and the IGF1 receptor (Polesskaya et al., 2007; Zhu et al., 2011). Overexpression of Lin28 promotes repair of several tissues (Shyh-Chang et al., 2013), including sciatic nerve regeneration in mice (Wang et al., 2018), and virally-mediated overexpression of Lin28 in the retina induces sustained optic nerve regeneration without affecting RGC survival (Wang et al., 2018). Surprisingly, this latter effect is mediated by Lin28 acting in amacrine cells (AC), the inhibitory interneurons of the retina, reducing AC activity and improving RGCs’ responsiveness to IGF1 (via localization of IGF1 receptors to RGCs’ primary cilia) (Zhang et al., 2019). Blocking inhibitory neurotransmission likewise improves IGF1 responsiveness, suggesting that cell non-autonomous, circuit-based mechanisms normally suppress RGC survival and optic nerve regeneration (Zhang et al., 2019).
Cell-Intrinsic regulators of axon regeneration
RGCs exhibit robust axon growth during development but lose this capacity early in the postnatal period (Chen, Jhaveri, & Schneider, 1995). This effect is mediated by RGCs receiving synaptic inputs from amacrine cells (Goldberg, Klassen, Hua, & Barres, 2002), and is associated with major changes in their program of gene expression (Wang et al., 2007). Some of the most striking changes occur in transcription factors belonging to the Kruppel-like factor (Klf) family (Moore, Apara, & Goldberg, 2011; Darcie L. Moore, Blackmore, & Goldberg, 2009). Elevation of Klf-4 and -9 in RGCs suppresses axon outgrowth, whereas decreasing their expression in mature RGCs promotes optic nerve regeneration (Apara et al., 2017; Apara & Goldberg, 2014; Moore et al., 2009; Qin et al., 2013; Trakhtenberg et al., 2018). Klf-4 acts in part by suppressing signaling through the JAK-STAT pathway (Moore et al., 2009; Qin et al., 2013), whereas Klf-9 regulates expression of dual specific phosphatase-14 (DUSP14), a suppressor of signaling through the mitogen activated kinase (MAPK) pathway (Galvao et al., 2018). In contrast to Klf-4 and -9, Klf-6 and Klf-7 decrease in expression during RGC development (Moore et al., 2009) but increase after optic nerve injury in zebrafish, where they regulate expression of regeneration-associated genes including Tα1 Tubulin (Veldman et al., 2007). Klf-7 overexpression promotes axon sprouting and regeneration in the corticospinal tract of adult mice (Blackmore et al., 2012).
An important, developmentally-regulated suppressor of axon regeneration is PTEN (phosphatase and tensin homolog), a negative regulator of PI3 kinase and its downstream effectors (AKT, mTOR, S6 kinase) and other pathways (Huang et al., 2019). Conditional deletion of PTEN substantially improves RGC survival, albeit transiently, and enables α-ON RGCs to regenerate axons into the optic nerve (Duan et al., 2015; Park et al., 2008). PTEN deletion also promotes axon growth in adult corticospinal neurons (Liu et al., 2010). Combining PTEN deletion with other pro-regenerative treatments strongly augments optic nerve regeneration, as discussed later.
Other studies point to post-transcriptional regulatory mechanisms of axon regeneration (Liu et al., 2012; Weng et al., 2018). Nerve injury increases levels of N6- methyladenosine, the most abundant modification of mRNA (Weng et al., 2018). Deletion of the N6-methyltransferase complex component Mettl14 mediates this epitranscriptomic modification and attenuates RGC axon regeneration in mice. Another RNA-binding protein, heterogeneous nuclear ribonucleaprotein (hnRNP)K, is upregulated after optic nerve injury in Xenopus laevis, and hnRNP K knockdown suppresses protein but not mRNA expression of medium neurofilament NF-M and GAP-43 and blocks axon regeneration (Liu et al., 2012).
The role of local axonal protein translation has been studied extensively in the developing Xenopus optic nerve (Shigeoka et al., 2016), and similar regulatory mechanisms are likely to be involved in optic nerve regeneration. In the rat spinal cord, ascending axons regenerating into a peripheral nerve graft contain mRNA of proteins associated with axon growth, i.e. β-actin, GAP-43, Neuritin, Reg3a, Hamp, and importin β1 (Kalinski et al., 2015), while in an in vitro model of axonal regeneration, mTOR protein levels are upregulated in sensory axons via local translation, in turn regulating local translation of injury-associated importin β and STAT3 (Terenzio et al., 2018).
Extrinsic inhibitors to Axon regeneration
In addition to cell-intrinsic suppressors of axon growth, regenerating axons encounter multiple cell-extrinsic suppressors of growth in their microenvironment, including myelin-associated inhibitors (MAIs), inhibitory proteoglycans, repulsive developmental cues, and physical barriers.
MAIs include Nogo, myelin associated glycoprotein (MAG, Siglec-4), and oligodendrocyte myelin glycoprotein (OMgp)(Wang et al., 2002), all of which can limit axon growth in various models (Boghdadi, Teo, & Bourne, 2018; Caroni & Schwab, 1988; McKerracher et al., 1994; Mukhopadhyay, Doherty, Walsh, Crocker, & Filbin, 1994; Schnell & Schwab, 1990). In normal development, myelination generally follows axon growth and targeting. Accordingly, MAIs may normally suppress aberrant axon sprouting and synaptic changes in developed CNS circuits, thereby stabilizing and protecting established connections and functions. However, when axons encounter myelin after CNS injury or disease, these same factors limit regrowth and recovery.
MAIs bind to a neuronal receptor complex including splice variants of the Nogo receptor (NgR1, NgR3), plus LINGO-1, and either the p75 low affinity neurotrophin receptor or TROY (Boghdadi et al., 2018; Wang et al., 2002). The receptor PirB also responds to MAIs to suppress axon growth (Atwal et al., 2008), and other receptors exist for specific MAIs, such as NgR2 (Robak et al., 2009) and the receptors for gangliosides GD1 and GT1b and for MAG (Giger et al., 2008). MAI binding to various receptors leads to RhoA GTPase activation of ROCK (Rho-associated coiled-coil-containing protein) (Domeniconi et al., 2005), a point of convergence for multiple growth-inhibitory stimuli. ROCK activation leads to growth cone collapse and cytoskeletal rearrangements. RhoA is an attractive therapeutic target in patients with CNS injury, and this can be achieved with the bacterial enzyme C3 ribosyltransferase, which inactivates RhoA (Boato et al., 2010; Chan et al., 2005; Dergham et al., 2002; Fehlings et al., 2018; Fehlings et al., 2011; Fischer, Petkova, et al., 2004; Lehmann et al., 1999).
In the optic nerve, virally directed expression of dominant-negative NgR or of C3 ribosyltransferase alone does not induce appreciable axon regeneration after nerve injury (Fischer, He, & Benowitz, 2004; Fischer, Petkova, et al., 2004). These results are similar to those seen elsewhere in the nervous system, where genetic knockout of NogoA or of NgRs results in only modest enhancement of axon regrowth in the injured spinal cord (Fink, Strittmatter, & Cafferty, 2015; Kim, Li, GrandPre, Qiu, & Strittmatter, 2003; Schnell & Schwab, 1990; Simonen et al., 2003; Wang et al., 2011). This result in part can be explained by the redundant roles of individual MAIs and their receptors (Lee et al., 2010). However, an even more relevant factor is that mature CNS neurons have only a limited intrinsic capacity for growth. Accordingly, considerable axon regeneration can be achieved after optic nerve injury by combining intraocular inflammation with expression of dominant-negative NgR or C3 ribosyltransferase (to inactivate Rho), or triple deletion of all three isoforms of NgR (Dickendesher et al., 2012; Fischer, He, et al., 2004; Fischer, Petkova, et al., 2004). Similarly impressive regeneration of ascending sensory fibers into or through the spinal cord has been achieved by combining strategies that activate neurons’ intrinsic growth state while also counteracting environmental inhibition (Steinmetz et al., 2005; Wang et al., 2012). Other strategies that have been successful after SCI include use of an NgR1 decoy protein (Wang et al., 2011) or administration of an antibody to NogoA (Fouad, Klusman, & Schwab, 2004; Freund et al., 2007), which has produced promising results in a human clinical trial (Kucher et al., 2018; Cafferty, Duffy, Huebner, & Strittmatter, 2010).
Chondroitin sulfate proteoglycans (CSPGs) are extracellular matrix components produced by cells that accumulate in the scar that forms after CNS injury, and represent another barrier to axon regeneration (McKeon, Hoke, & Silver, 1995; Rudge & Silver, 1990; Tran, Warren, & Silver, 2018). Growth-inhibitory CSPGs act through RPTPσ, LAR, NgR1, and NgR3 to inhibit growth; deletion of RPTPσ, NgR1, -2 and -3 substantially augments inflammation-induced optic nerve regeneration (Dickendesher et al., 2012; Fry, Chagnon, Lopez-Vales, Tremblay, & David, 2010; Lang et al., 2015). The enzyme chondroitinase ABC degrades CSPGs and also enhances plasticity after CNS injury, and experiments in the rodent and primate spinal cord have shown moderate improvements with chondroitinase (Bradbury et al., 2002) (Rosenzweig et al., 2019), especially when combined with other treatments (Garcia-Alias & Fawcett, 2012). In the optic nerve, a genetically engineered chondroitinase strongly enhances the effects of intraocular inflammation on axon regeneration (Pearson, Mencio, Barber, Martin, & Geller, 2018). It should be noted, however, that the glial scar appears to have some beneficial roles, as its deletion can impair recovery after CNS injury (Anderson et al., 2016).
Regeneration in the mature CNS is also hindered by upregulation of molecules that otherwise repel axons during development such as wnt, netrin, ephrins, somaphorin, and slit (Goldberg et al., 2004; Hollis, 2016; Song et al., 1998; Tran et al., 2018).
RGC Survival
In mice and rats, RGCs begin to die within a few days after optic nerve injury and nearly all are lost within a month, largely via apoptosis (Berkelaar, Clarke, Wang, Bray, & Aguayo, 1994; Garcia-Valenzuela, Gorczyca, Darzynkiewicz, & Sharma, 1994; Misantone, Gershenbaum, & Murray, 1984; Quigley et al., 1995). Although axon regeneration and cell survival involve distinct mechanisms (Chierzi, Strettoi, Cenni, & Maffei, 1999; Goldberg, Espinosa, et al., 2002), prolonging RGC survival after optic nerve injury is nonetheless crucial to slow neurodegeneration and thereby extend the critical window for treatments that might stimulate regeneration. Over the past 30 years or so, multiple mechanisms have been found to contribute to RGC death, but for the most part, the relationships among these mechanisms are still poorly understood.
Zinc and NO signaling in RGC survival and axon regeneration
One of the earliest changes seen in the retina after optic nerve injury is an elevation of ionic (free, or mobile) zinc (Zn2+) (Li et al., 2017). Zinc is essential for the functioning of all cells, and is required, for example, for the proper folding and function of transcription factors and enzymes (Kochanczyk, Drozd, & Krezel, 2015). In addition, in many neurons, free Zn2+ accumulates in synaptic vesicles (Kimura & Kambe, 2016), where it is co-released with neurotransmitters, most prominently in cortical neurons where it modulates glutamatergic synapses and synaptic plasticity (Nakashima & Dyck, 2009; Pan et al., 2011; Sensi, Paoletti, Bush, & Sekler, 2009; Sensi et al., 2011). However, Zn2+ dyshomeostasis can be lethal (Aras & Aizenman, 2011; Sensi et al., 2009) and can impair mitochondrial function, stimulate proapoptotic factors, and lead to the production of reactive oxygen species (ROS) (Lieven, Hoegger, Schlieve, & Levin, 2006; Sensi, Yin, Carriedo, Rao, & Weiss, 1999) that can react with nitric oxide (NO) to produce peroxynitrite and activate MAPK-p38 (Bossy-Wetzel et al., 2004). MAPK/p38 signaling directly and indirectly activates voltage-gated potassium channels, which induce potassium efflux and cell death (Brown, 2010). Because elevation of intracellular Zn2+ beyond the nanomolar range is neurotoxic, an elaborate array of zinc transport proteins (ZnTs), import proteins (ZIPs) and buffers (e.g., metallothioneins) have evolved to tightly regulate cytosolic Zn2+ levels (McAllister & Dyck, 2017).
Zn2+ levels increase in the retina within an hour of optic nerve injury. Surprisingly, however, this elevation occurs in synaptic vesicles of amacrine cells, the inhibitory interneurons of the retina, and only appears in RGCs after 2-3 days (Li et al., 2017). Chelating Zn2+ increases RGC survival and axon regeneration (Fig. 2). These findings implicate Zn2+ elevation in retinal interneurons as a crucial step in a multi-cell signaling pathway that links optic nerve injury to RGC degeneration. The recent results of (Zhang et al., 2019) further establish the importance of amacrine cells in modulating the survival and regenerative capacity of RGCs after optic nerve injury.

Zn2+ elevation after optic nerve injury and effects of chelation on RGC survival and optic nerve regeneration. Top row: Normal retina shows little mobile zinc (Zn2+) (left); there is a dramatic increase in amacrine cell terminals in the inner plexiform layer (IPL, arrows) starting≤1 h after optic nerve injury and peaking at 1 day (right). Middle: Retinal whole-mounts immunostained for βIII tubulin. Two weeks after nerve injury, only 17% of RGCs survive; treatment with the chelator TPEN (far right) enables nearly half of RGCs to survive long-term. Bottom: Optic nerve two weeks after injury shows no axons beyond the injury site (asterisk) without treatment, but extensive regeneration with early Zn2+ chelation (GAP-43 ICC) (Y. Li et al., 2017).
Metallothionein-bound Zn2+ can be liberated by reactive oxygen species or peroxynitrite (ONOO)(Aravindakumar, Ceulemans, & De Ley, 1999; Hidalgo, Aschner, Zatta, & Vasak, 2001; Sensi et al., 1999; Spahl, Berendji-Grun, Suschek, Kolb-Bachofen, & Kroncke, 2003). Using a newly developed nitric oxide (NO) sensor, we found that optic nerve injury leads to a rapid, persistent upregulation of NO (Li et al., 2017) which depends on the neuron-specific form of nitric oxide synthase (NOS1, nNOS). To fully understand the crosstalk between NO and Zn2+ in the retina it will be necessary to determine whether blocking ROS generation decreases zinc elevation, attenuates NO production, and, ultimately whether this augments RGC regeneration and survival after optic nerve injury.
In neurons, NOS1 is commonly associated with NMDA receptors, and NOS1 activation requires calcium influx through the glutamate- and voltage-regulated ion channel (Christopherson, Hillier, Lim, & Bredt, 1999). Inhibiting NMDA receptors eliminates the Zn2+ signal in the IPL, suggesting that NO generation and Zn2+ elevation lie downstream of NMDA receptor activation (Li, Andereggen, Yuki, Rosenberg, & Benowitz, 2016). NOS1 is expressed exclusively in amacrine cells, pointing to the existence of an upstream signal linking optic nerve crush to NOS1 activation. The most probable NMDA activator is glutamate, which can enter the extracellular space through synaptic release or from a reversal of glutamate transporters (e.g., GLT-1, GLAST and EEAT5), which normally function to take up extrasynaptic glutamate (Danbolt, 2001). In the retina, GLT-1 is expressed in two isoforms, GLT-1a which is primarily expressed by glia, and GLT-1b which is primarily expressed in bipolar cells. Global inhibition of GLT-1 prevents Zn2+ accumulation in amacrine cell terminals, as does bipolar cell-specific knockout of GLT-1. These results point to a reversal of glutamate transport in bipolar cells and activation of NMDA receptors as the steps that precede Zn2+ accumulation (Hanovice, Li, Benowitz, & Rosenberg, 2018). These results also point to a previously unknown interneuronal signaling pathway that culminates in Zn2+ accumulation in the inner retina and suppression of RGC survival and regeneration after optic nerve injury. Our initial efforts in understanding this pathway are beginning to reveal the contours of this pathway, including the generation of NO by NMDAR-mediated NOS1 activity, driven by glutamate exported through the GLT-1b transporter in bipolar cells. The signals that link optic nerve injury to the inversion of glutamate transport through GLT-1b in bipolar cells remain to be identified.
Activation of dual leucine zipper kinase (DLK, MAP3K12) has also been identified as an early and critical determinant of RGC survival after optic nerve injury. DLK is a mitogen-activated kinase that positively regulates both the JNK and p38 MAPK signaling pathways (Tedeschi & Bradke, 2013). DLK signaling increases after optic nerve crush, increasing JNK1-3-dependent activation of the transcription factors c-JUN, activating transcription factor 2 (ATF2), myocyte-specific enhancer factor 2 (MEF2A), and SRY-box 11 (SOX11) to regulate RGC survival (Welsbie et al., 2013, 2017). DLK activity is enhanced by leucine zipper kinase (LZK) activity, such that double knockout of both kinases further protects RGCs (Welsbie et al., 2013, 2017). However, DLK signaling is also required for axon regeneration (Ghosh et al., 2011; Watkins et al., 2013), and axon outgrowth induced by Pten deletion is blocked by DLK deletion. It will be important to identify downstream targets of DLK that can be manipulated to improve RGC survival without impairing axon regeneration.
Microglia and A1 astrocytes
Microglia, the resident immune cells of the CNS, respond to injury by retracting their long, ramified processes and altering their cellular functions (Graeber, Li, & Rodriguez, 2011). Activated microglia are characteristic of many neurodegenerative diseases (Salter & Stevens, 2017), including glaucoma, and in an animal model of this disease, repressing microglial activation substantially improves RGC survival (Bosco et al., 2008, 2012; Williams et al., 2016, 2019). Our lab found that deleting TNF-α, or one of its receptors, or the complement receptor on microglia, strongly protects RGCs in a mouse model of ocular hypertension (OHT)-induced glaucoma (Nakazawa et al., 2006). In conformity with that finding, we found that Etanercept, a decoy receptor for TNF-α, prevents microglial activation, astrocyte polarization and attenuates RGCs loss in a rat OHT model (Roh et al., 2012).
Activation of A1 astrocytes after optic nerve injury promotes synapse degradation and increases expression of pro-apoptotic factors. After optic nerve injury, reactive astrocytes secrete pro-inflammatory agents, such as Lcn2, NO, and TNFα, which directly impact neuronal function (Bi et al., 2013; Lebrun-Julien et al., 2009; Sofroniew, 2014; Zamanian et al., 2012). Although reactive astrocytes can be beneficial by clearing cell debris and providing neurotrophic support to neurons, the pro-inflammatory “A1” phenotype exhibits highly neurotoxic functions (Liddelow et al., 2017; Zamanian et al., 2012). Liddelow, et al. have proposed that activated microglia directly initiate A1 astrocyte polarization by secretion of pro-inflammatory proteins, including tumor necrosis factor-α (TNFα), interleukin 1α (IL-1α), and the complement protein C1q (Liddelow et al., 2017). Inhibiting TNFα, IL-1α, and C1q after optic nerve injury either with blocking antibodies or by genetic deletion, was sufficient to block A1 astrocyte polarization even in the presence of activated microglia, and greatly increased RGC survival. These results suggest that axotomy-induced RGC death is driven largely by neurotoxic A1 astrocytes. However, microglia are also known to have beneficial roles after injury, including controlling debris clearing and synapse maintenance (Herzog et al., 2019; Norris et al., 2018). One study reported that ablation of microglia does not improve RGC survival after nerve injury (Hilla, Diekmann, & Fischer, 2017), raising the question of whether incomplete microglia ablation is insufficient to block A1 astrocyte polarization, or whether the loss of beneficial microglial functions is offset by the loss of harmful microglial functions. Therefore, further research into understanding how we can program microglia towards beneficial phenotypes and prevent neurotoxic signaling is needed to determine if we can harness these cells to improve RGC survival after injury.
Downstream pathways
Bax and activated caspase-3, two hallmarks of apoptosis, appear in RGCs approximately 2 days after optic nerve injury, and inhibiting either of these prolongs RGC survival, as does blocking caspase 2 or caspase 9 or overexpressing Bcl-2 (Cenni et al., 1996; Isenmann, Wahl, Krajewski, Reed, & Bahr, 1997; Kermer et al., 1999, 2000; Li, Schlamp, Poulsen, & Nickells, 2000; Vigneswara, Berry, Logan, & Ahmed, 2012). Immediately preceding the initiation of the apoptotic response is the degeneration of the nucleus, suggesting that axonal damage leads to active remodeling of RGC chromatin (Janssen, Mac Nair, Dietz, Schlamp, & Nickells, 2013). Histone proteins do in fact undergo a marked deacetylation at about 3 days post injury, which is associated with nuclear translocation of histone deacetylase 3 (HDAC3), and which renders chromatin less accessible for transcription. Accordingly, inhibition of HDAC activity increases the survival of RGCs after injury (Chindasub, Lindsey, Duong-Polk, Leung, & Weinreb, 2013; Gaub et al., 2011; Janssen et al., 2013; Koriyama, Sugitani, Ogai, & Kato, 2014; Pelzel, Schlamp, & Nickells, 2010; Schmitt, Pelzel, Schlamp, & Nickells, 2014; Zhang et al., 2012). RGC-specific knockdown of HDAC3 suppresses histone H4 deacetylation and reduces RGC degeneration, though without attenuating the diminished expression of RGC-specific genes (e.g. Brn3, Thy1, Sncg, Nrn1) (Schmitt et al., 2014). These results suggest the involvement of other HDACs, with the exact contributions of these and other mediators of cell death and cell survival remaining to be elucidated.
Combinatorial treatments for long-distance optic nerve regeneration
In the past few years, several treatments have enabled a modest number of RGCs to regenerate their axons the entire length of the optic nerve, through the optic chiasm, and into the brain, in some cases forming synapses in the appropriate subcortical visual target areas.
As noted earlier, intraocular inflammation enhances RGC survival and axon regeneration by virtue of elevating retinal expression of oncomodulin, SDF1, and perhaps other factors that are secreted by neutrophils and macrophages. Combining intraocular inflammation with elevation of cAMP and PTEN deletion (in RGCs and other cells infected with an AAV2 expressing an anti-pten shRNA) has strongly synergistic effects, increasing axon regeneration 10-fold compared to any of the treatments alone, and enabling some axons to reach the optic chiasm by 6 weeks (Kurimoto et al., 2010). At later timepoints, regenerating axons begin to enter the brain and reinnervate multiple subcortical visual nuclei, including the LGN, suprachiasmatic nucleus (SCN), olivary pretectal nucleus (OPN), superior colliculus (SC), and medial terminal nucleus (MTN), leading to a limited recovery of simple visual reflexes such as the optomotor response (Fig. 3) (de Lima et al., 2012). Regenerating axons become myelinated and recover the electrical domains at paranodal and nodal regions that are responsible for the propagation of action potentials. However, myelination of growing axons proceeds very slowly (Marin et al., 2016).

Example of combinatorial treatment leading to long-distance regeneration, target reinnervation, and partial recovery of visual responses. a. Longitudinal section through the mouse optic nerve shows axons regenerating the full length of the nerve (red) in a mature ptenflx/flx mouse 12 weeks after injury. pten was deleted in RGCs 2 weeks prior to nerve injury via intraocular AAV2-Cre. Mice received intraocular Zymosan (to induce inflammation) plus CPT-cAMP at the time of nerve injury and again at 3 and 6 weeks. Axons were labeled by intraocular injection of cholera toxin B fragment (CTB) 5 days before sacrifice. Asterisk shows injury site. b-d. Higher magnification of the area within the rectangle in a stained for CTB (b), the growth-associated protein GAP-43 (c) and both (d). e. Regenerating CTB-labeled axons in the lateral geniculate nucleus (LGN, red) express VGluT2 (green), a marker for RGC glutamatergic synaptic vesicles. f. Regenerating axons (red) lie within the dLGN (dotted line; tissue co-immunostained for NeuN). g,h. Evidence of remyelination. Note oligodendrocyte process (arrow) in early stages of ensheathing an axon in the optic nerve 12 weeks after nerve damage (g). i,j. Optomotor test. Mice were placed on an elevated platform surrounded by a display simulating circumferential rotation of a striped pattern (i). Ten weeks after nerve damage, mice that showed anatomical evidence of central reinnervation (blue) showed a stronger optomotor response (measured in cycles/degree) than controls receiving partial treatment (red, j). k. Circadian photoentrainment: average % of overall activity in one hour bins. Mice were maintained on a continuous cycle of lights on between 7:00 AM and 7:00 PM (yellow bars; blue bars denote lights off period). Ambient light was phase-shifted by 3 h on day 3. Although individual mice in all groups showed activity cycles of ∼ 24 h, the activity patterns of blind mice and mice with incomplete regeneration became asynchronous. In contrast, mice with successful regeneration showed synchronous activity that was phase-shifted compared to normal mice (de Lima et al., 2012).
In other studies, long-distance regeneration was achieved by co-deletion of PTEN and SOCS3 plus overexpression of intraocular CNTF, with some axons reaching the chiasm after 4 weeks (Sun et al., 2011). This effect was further enhanced by overexpression of c-Myc (Belin et al., 2015). When the optic nerve was injured close to the optic chiasm, co-deletion of PTEN and SOCS3 plus intraocular AAV2-CNTF led to reinnervation of the SCN within one month and the reappearance of physiological responses in SCN neurons within 2 months (Li et al., 2015). Another study used co-deletion of PTEN and SOCS3 together with virally mediated expression of Osteopontin and the trophic factors CNTF and IGF-1, with the lesion performed close to the SC. In this case, RGCs axons extended through the injury site and invaded the SC. Although the axons remained unmyelinated, signal conduction beyond the injury site was observed within 3 months after injury when a potassium channel blocker was applied (Bei et al., 2016).
Further studies showed that stimulating neuronal activity with visual stimulation or chemogenetics, combined with overexpression of cRheb1, a positive regulator of the mTOR signaling pathway, synergistically promoted optic nerve regeneration, with reinnervation of subcortical visual nuclei reported to occur as early as 3 weeks after injury (Lim et al., 2016). Overexpression of the anti-apoptotic protein Bcl-2 combined with virally-expressed CNTF led to a significant increase in RGC survival and regeneration, with a few axons reaching the chiasm 5 weeks after injury (Leaver, Cui, Bernard, & Harvey, 2006). Deletion of BAX, a pro-apoptotic member of the Bcl-2 family, enabled up to 80% of RGCs to survive 8 weeks after optic nerve injury and enabled RGCs to regenerate axons even when virally-mediated CNTF expression was delayed for several weeks. In this case, some axons reached central visual target areas 8 weeks later (Yungher, Ribeiro, & Park, 2017). It should be noted, however, that because BAX deletion occurred in all cells, and because virally-mediated CNTF expression promotes optic nerve regeneration largely by activating glia and/or inflammatory cells (L. Xie, H.-Y. Gilbert, Y. Yin, L. Benowitz, in preparation), the mechanisms underlying this phenomenon remain uncertain.
Over the past 2 - 3 decades, optic nerve regeneration has gone from being considered impossible to becoming a reality. Clearly, much more needs to be done to increase the number of axons that reach their appropriate destinations, determine whether regenerating axons form a topographically organized map of visual space in the LGN and SC, and assess visual acuity. Perhaps these advances will take the form of manipulating currently unidentified transcriptional and epigenetic regulators of the regenerative program, perhaps by identifying additional potent trophic agents for RGCs, perhaps by finding ways to alter the immune and glial response to injury, perhaps by re-programming the entire retina, or perhaps by something that is currently outside our conceptual models (unknown unknowns). In any event, the strides that have been made to date are considerable and point to the possibility that one day we may be able to restore vision after optic nerve injury in a clinically meaningful way.
Footnotes
Acknowledgments
We are grateful for the support of the Adelson Medical Research Foundation, NIH/NEI (R01EY027881, NEI R01EY024481, 2T32EY007145-19), and NINDS IDDRC HD018655.
