Glycogen synthase kinase 3 signaling in neural regeneration in vivo

Abstract

Glycogen synthase kinase 3 (GSK3) signaling plays important and broad roles in regulating neural development in vitro and in vivo. Here, we reviewed recent findings of GSK3-regulated axon regeneration in vivo in both the peripheral and central nervous systems and discussed a few controversial findings in the field. Overall, current evidence indicates that GSK3β signaling serves as an important downstream mediator of the PI3K–AKT pathway to regulate axon regeneration in parallel with the mTORC1 pathway. Specifically, the mTORC1 pathway supports axon regeneration mainly through its role in regulating cap-dependent protein translation, whereas GSK3β signaling might be involved in regulating N⁶-methyladenosine mRNA methylation-mediated, cap-independent protein translation. In addition, GSK3 signaling also plays a key role in reshaping the neuronal transcriptomic landscape during neural regeneration. Finally, we proposed some research directions to further elucidate the molecular mechanisms underlying the regulatory function of GSK3 signaling and discover novel GSK3 signaling-related therapeutic targets. Together, we hope to provide an updated and insightful overview of how GSK3 signaling regulates neural regeneration in vivo.

Introduction

Glycogen synthase kinase 3 (GSK3), a serine/threonine kinase, was initially identified as a key regulatory enzyme in skeletal glucose metabolism (Hemmings and Cohen, 1983; Wang and Roach, 1993). Two GSK3 isoforms, GSK3α and GSK3β, share 85% overall amino acid sequence homology and show 95% identity in catalytic kinase domains (Woodgett, 1990). As research continues, ∼100 substrates of GSK3 have been explored (Beurel et al., 2015), and various signaling pathways involving GSK3 have been identified. Unlike most kinases, GSK3 has the sustained kinase activity and is inactivated by serine phosphorylation (Ser21 for GSK3α and Ser9 for GSK3β) (Dajani et al., 2001; Liang and Chuang, 2007). Homozygous knockout (KO) of GSK3β resulted in embryonic lethality (Hoeflich et al., 2000), whereas the GSK3β^+/− heterozygous mutant mimicked the behavioral and molecular effects of lithium (a GSK3 kinase inhibitor) (O’Brien et al., 2004) and suffered from impaired memory reconsolidation (Kimura et al., 2008), which highlights the crucial roles of GSK3β in nervous system development. Recent studies have explored the potential mechanisms through which GSK3 signaling coordinates neurogenesis and neuronal polarization during development, as previously summarized (Hur and Zhou, 2010). In mature neurons, GSK3β signaling has been shown to mediate phosphatidylinositol 3-kinase (PI3K)-dependent axon regeneration in the mammalian peripheral nervous system (PNS) after injury (Saijilafu et al., 2013b). Considering that the kinase activity of GSK3β increases in mature retinal neurons with little regenerative capacity (Guo et al., 2016), GSK3β signaling has also been revealed as an important mediator that regulates axon regeneration in the central nervous system (CNS).

Several in vitro studies have shown that GSK3β controls axon growth by regulating cytoskeletal assembly in growth cones (Zhou et al., 2004; Jiang et al., 2005). MAP1B, CRMP2, CLASPs, and APC act as key mediators of GSK3 signaling to coordinate microtubule states during axon growth (Goold et al., 1999; Zhou et al., 2004; Yoshimura et al., 2005; Zhou and Snider, 2005; Hur et al., 2011). Moreover, a substantial proportion of axon regeneration-related transcription factors, such as CREB, Smad1, p53, and c-Jun, were also found to be regulated by GSK3 signaling (Liu et al., 2012; Saijilafu et al., 2013b).

In addition to in vitro studies, great efforts have been devoted to enhancing axon regeneration in vivo in the past decade. Here, we reviewed recent findings of axon regeneration in vivo directly or indirectly regulated by GSK3 signaling, discussed the underlying mechanisms, and proposed our view on future research directions.

Roles of GSK3 signaling in the regulation of axon regeneration in vivo

Neurons in the CNS have little ability to support axon regeneration because of their poor intrinsic regenerative capacity and the inhibitory extrinsic environment. However, when the PNS is injured, axon regeneration can be strong (Mahar and Cavalli, 2018; Qian and Zhou, 2020; Varadarajan et al., 2022). Spinal cord injury (SCI) is a commonly used model for studying axon regeneration in the CNS. A study reported that lithium, a widely used GSK3 inhibitor, could stimulate the regrowth of corticospinal tracts (CSTs) rostral to a transection site after spinal cord unilateral transection injury in mice (Dill et al., 2008). Moreover, the application of SB415286, a more specific GSK3 inhibitor, induced the regeneration of CST axons upon dorsal transection and spinal cord contusion in rats. Importantly, locomotor recovery was enhanced in the GSK3 inhibitor-treated group, providing the first solid evidence that GSK3 inhibition can enhance CNS axon regeneration in vivo. Another widely used model for the study of CNS axon injury and repair is the optic nerve crush (ONC) model, in which the axons from retinal ganglion cells (RGCs) in the optic nerve are injured. Similar to the observations in the SCI model, GSK3β KO or knockdown (KD) mice showed moderately increased optic nerve regeneration compared with wild-type mice in the ONC model (Guo et al., 2016; Miao et al., 2016; Leibinger et al., 2017; Ahmed et al., 2019), providing clear genetic data that GSK3 inactivation promotes CNS axon regeneration in vivo. Lens injury-induced inflammatory stimulation and phosphatase and tensin homolog (PTEN) deletion are two well-known approaches for enhancing optic nerve regeneration (Leon et al., 2000; Müller et al., 2007; Park et al., 2008). Because GSK3 is inactivated by AKT-mediated serine phosphorylation, a mutant serine-to-alanine (S/A) knockin (KI) mouse line, GSK3(α/β)^S/A, was developed to create active mutant mice (McManus et al., 2005). Interestingly, such mice can survive into adulthood without obvious phenotypes in the nervous system, challenging the importance of serine phosphorylation in regulating GSK3 activity. Nevertheless, two studies from the same laboratory provided evidence that optic nerve regeneration induced by lens injury or PTEN deletion was reduced in GSK3(α/β)^S/A mice (Leibinger et al., 2017, 2019). Taken together, these studies consistently showed that GSK3 inhibition is an essential signaling event in promoting axon regeneration in the CNS.

The sciatic nerve contains mixed peripheral axons of sensory neurons in the dorsal root ganglion (DRG) and motor neurons in the spinal cord. Sciatic nerve injury (SNI) is the most studied model for investigating axon regeneration in the PNS. Using this model, we revealed that GSK3 acts downstream of PI3K signaling to regulate sensory axon regeneration (Saijilafu et al., 2013a). Furthermore, we established that the transcription factor Smad1, rather than β-catenin, is a downstream mediator of PI3K–GSK3 signaling (Saijilafu et al., 2013a). In another study, we showed that GSK3β is the target of microRNA-26a (miR-26a), which is necessary for sensory axon regeneration (Jiang et al., 2015). In support of this finding, GSK3β KD reversed sensory axon regeneration impairment induced by miR-26a inhibition (Jiang et al., 2015). Moreover, we demonstrated that CLASPs, microtubule plus-end-binding proteins activated by GSK3 inhibition, are necessary for sensory axon regeneration in vivo (Hur et al., 2011). Collectively, these studies provide consistent evidence that GSK3 inhibition is necessary and sufficient to enhance axon regeneration in vivo.

Inhibition of GSK3β but not GSK3α promotes axon regeneration in vivo

Mediated by the PI3K pathway, GSK3 is inactivated through distinct phosphorylation sites in two isoforms (Ser9 in GSK3β and Ser21 in GSK3α). Researchers have found that, despite highly similar structures, deletion of GSK3β alone induced significant axon regeneration, whereas deletion of GSK3α alone yielded little effect. Additionally, deletion of both GSK3α and GSK3β did not exert an additive effect compared with deletion of GSK3β alone (Guo et al., 2016; Miao et al., 2016; Leibinger et al., 2017). One likely explanation for the different roles of the two isoforms in controlling axon regeneration is their distinct substrates in neurons. CRMP2, a key microtubule-binding protein uniformly expressed in the nervous system, is a well-known substrate of GSK3 and a pivotal mediator that promotes axon growth and regeneration (Liz et al., 2014; Leibinger et al., 2019). Interestingly, an early study demonstrated that GSK3β KO abrogated the phosphorylation of Thr509 and Thr514 of CRMP2, yet normal phosphorylation still existed in the cortex lacking GSK3α (Soutar et al., 2010), thus providing one potential substrate accounting for the distinct roles of GSK3 isoforms in regulating axon regeneration in vivo. Future systematic screening of GSK3α/β substrates would provide more isoform-specific substrate candidates underlying different biological functions in the nervous system.

Roles of GSK3 serine phosphorylation in the regulation of axon regeneration

As mentioned above, GSK3(α/β)^S/A mice do not have overt neurodevelopmental phenotypes and can survive into adulthood, suggesting that GSK3 serine phosphorylation is not the only mechanism for regulating its kinase activity (Hur and Zhou, 2010). Interestingly, when using this mouse line to study axon regeneration, controversial results were obtained. Our early study demonstrated that sensory axon regeneration in mutant mice was comparable to that in wild-type mice, both in vitro and in vivo (Zhang et al., 2014). Using the same mouse line, a later study from Gobrecht et al. (2014) showed that sensory axon regeneration was significantly enhanced, suggesting that GSK3 activation is important in axon regeneration. The reasons for this discrepancy remain elusive. However, there were indeed a few different experimental details in the two aforementioned studies. For instance, in the in vitro experiment, we quantified the average length of the longest axon of each neuron, whereas in the latter study, the researchers quantified the total axon length of each neuron. For in vivo sensory axon regeneration experiments, we genetically labeled sensory neurons and their axons with enhanced green fluorescent protein and measured the average length of axons traced from the crush site to the distal axonal tips in the whole-mount sciatic nerves, whereas in the latter study, the researchers performed SCG10 (a regeneration-associated marker) immunostaining on sciatic nerve sections and quantified the number of labeled axons at several distances beyond the injury site.

Mechanistically, in our study, the protein levels of GSK3 in GSK3(α/β)^S/A KI mice remained the same as those in wild-type mice. Additionally, the levels of phospho-CRMP2 in DRG neurons of GSK3(α/β)^S/A KI mice were the same as those of wild-type mice, indicating that GSK3β kinase activity in GSK3(α/β)^S/A KI mice was not increased (Zhang et al., 2014). This result provides a potential explanation for the unchanged sensory axon regeneration phenotype in GSK3(α/β)^S/A KI mice. In support of this finding, overexpression of the GSK3β (S9A) mutant in the sensory neurons of GSK3(α/β)^S/A KI mice inhibited sensory axon regeneration, indicating that increasing the total GSK3 protein level and its kinase activity exerted a negative effect on axon regeneration. The explanation for the enhanced sensory axon regeneration in GSK3(α/β)^S/A KI mice observed in Gobrecht et al. (2014) was that higher GSK3 activity could maintain the level of dynamic microtubules in the growth cone via MAP1B phosphorylation, which is also necessary for efficient axon growth. However, studies from the same laboratory showed that enhanced optic nerve regeneration induced by lens injury or PTEN KO was significantly impaired in GSK3(α/β)^S/A KI mice, likely due to increased levels of phospho-CRMP2 (not able to bind to microtubules) (Leibinger et al., 2017, 2019). What are the potential explanations for the opposite effects between sensory axon and optic nerve regeneration observed in GSK3(α/β)^S/A KI mice? One potential reason is that an optimal level of GSK3 kinase activity is important for successful axon regeneration by balancing microtubule stability and dynamics in different neuronal types. Obviously, more studies from different researchers are required to answer these questions.

GSK3 signaling acts in parallel with mTORC1 signaling to regulate axon regeneration in vivo

How the molecular players of GSK3β signaling, either upstream or downstream, coordinate to regulate neural regeneration in vivo is summarized in Figure 1. Deletion of PTEN promotes robust axon regeneration after ONC (Park et al., 2008) or CST transection (Liu et al., 2010). Recent studies have shown that PI3K-mediated AKT activation acts downstream of PTEN deletion to enhance optic nerve regeneration (Guo et al., 2016; Miao et al., 2016; Huang et al., 2019). Specifically, PTEN deletion increases AKT phosphorylation and activation. Injection of AAV-caAKT, a constitutively active form of AKT (Kohn et al., 1996), into the vitreous body promoted both optic nerve regeneration and RGC survival after ONC (Guo et al., 2016). Among the three isoforms of AKT, AKT3 is the predominant isoform in the retina, and it was shown to promote much more robust axon regeneration and RGC survival than AKT1 (Miao et al., 2016). Conversely, AKT3 KO significantly reduced axon regeneration induced by PTEN deletion. mTORC1 is one of the most studied signaling mediators downstream of AKT, and the PI3K–AKT–mTORC1 pathway is a key axis for protein synthesis and cell growth (Manning and Cantley, 2007; Laplante and Sabatini, 2012). Rapamycin, an mTORC1 inhibitor, partially blocked axon regeneration induced by PTEN deletion (Park et al., 2008) or AKT overexpression (Guo et al., 2016). Moreover, deletion of essential regulators of mTORC1 signaling, such as regulatory-associated protein of mTOR (Raptor) or mTOR itself, significantly decreased but did not abolish AKT3-induced axon regeneration (Miao et al., 2016). Similarly, blocking the downstream of mTORC1 via overexpression of S6K1 dominant-negative mutant or 4E-BP1 constitutively active mutant also significantly reduced but did not abolish AKT3-induced regeneration (Miao et al., 2016). These results suggest that the mTORC1 signaling pathway only partially mediates PTEN/PI3K–AKT-induced axon regeneration, and mTORC1-independent pathways may also be involved.

Figure 1.

GSK3 signaling regulates axon regeneration in vivo by coordinating protein translation, gene transcription, and cytoskeletal organization. AKT, downstream of PTEN/PI3K signaling, regulates axon regeneration via two independent parallel pathways: mTORC1-mediated protein translation machinery and GSK3 signaling. During axon regeneration, GSK3 kinase activity is regulated by AKT-mediated phosphorylation, and the protein level of GSK3β in neurons can be regulated by miR-26a. Downstream of GSK3β, the transcription factor Smad1 plays an important role in the regulation of gene transcription necessary for axon regeneration. GSK3β inhibition can also activate eIF2Bε to enhance anisomycin-sensitive protein translation and induce axon regeneration. Finally, GSK3 signaling is a regulatory hub for many microtubule-binding proteins, such as CLASPs, CRMP2, APC, and MAP1B, which coordinately act to control axon extension at the growth cone. Future studies using newly developed techniques are required to discover novel regulators and substrates of GSK3β specifically involved in the regulation of axon regeneration. TFs, transcription factors.

GSK3 inactivation by phosphorylation acts downstream of AKT. Indeed, increased phosphorylation of GSK3β was observed in PTEN-deleted and active AKT-overexpressing RGCs (Guo et al., 2016). Functionally, overexpression of the GSK3β (S9A) active mutant significantly reduced AKT-enhanced axon regeneration (Guo et al., 2016; Miao et al., 2016), indicating that GSK3 inactivation is necessary for AKT-induced axon regeneration. In contrast, either GSK3β deletion or overexpression of a kinase-dead mutant GSK3β (K85A) in RGCs led to modest but significant optic nerve regeneration (Guo et al., 2016), providing direct evidence that GSK3 deletion or inactivation is sufficient to enhance axon regeneration. Importantly, deleting Raptor together with overexpressing the GSK3β (S9A) mutant almost completely abolished AKT-induced axon regeneration (Miao et al., 2016), further confirming that mTORC1 activation and GSK3β inactivation are two independent parallel pathways downstream of AKT that promote CNS axon regeneration. Interestingly, although rapamycin only partially inhibited AKT-induced axon regeneration, it almost completely blocked AKT-enhanced RGC survival (Guo et al., 2016), indicating that AKT promotes RGC survival mainly through the mTORC1 pathway. However, GSK3β did not regulate RGC survival (Guo et al., 2016). Taken together, these results demonstrate that inactivation of GSK3β, a direct downstream target of AKT, is both necessary and sufficient to enhance axon regeneration but not RGC survival.

How does AKT regulate GSK3β activity in neurons to control axon regeneration? In the traditional pathway, AKT is activated by PI3K signaling, as indicated by phosphorylation at Thr308/Ser473 (Thr305/Ser472 in AKT3) residues, subsequently phosphorylating and inactivating GSK3β. However, Miao et al. (2016) showed that AKT could regulate GSK3β activity in two opposite ways depending on the specific residue of AKT phosphorylation in RGCs. Overexpression of AKT3 (T305A), an AKT3 mutant that cannot be activated, in RGCs resulted in a slight increase in phospho-S6 levels but not in axon regeneration. Surprisingly, overexpression of the AKT3 (S472A) mutant induced even more axon regeneration than that of wild-type AKT3, suggesting that Ser472-phosphorylated AKT3 suppresses axon regeneration. These results revealed that AKT3 phosphorylation at Thr305 and Ser472 promoted and impaired axon regeneration, respectively, contradictory to the conventional mechanism. Several early studies have shown that AKT3 is phosphorylated at Ser472 by mTORC2 (Hresko and Mueckler, 2005; Sarbassov et al., 2005; Guertin et al., 2006). If so, mTORC2 may act to inhibit axon regeneration. However, deleting rapamycin-insensitive companion of mTOR (Rictor), a key component of mTORC2, did not induce axon regeneration, whereas combining Rictor deletion and AKT3 overexpression had an additive effect on axon regeneration (Miao et al., 2016). One potential explanation is that the phosphorylation of Ser472 on AKT3 by mTORC2 might inhibit axon regeneration by affecting GSK3β activity. The AKT3 (S472A) mutant markedly elevated the phosphorylation levels of GSK3β at Ser9 (Miao et al., 2016). Additional studies are necessary to validate these results and elucidate the underlying mechanisms.

In addition to the regulation of GSK3 kinase activity via phosphorylation, it has also been reported that the expression level of GSK3 contributes to the control of axon regeneration. In sensory neurons, GSK3β was identified as the target of miR-26a. As a result, the inhibition of endogenous miR-26a in sensory neurons markedly elevated the protein level of GSK3β, thereby leading to impaired sensory axon regeneration in vitro and in vivo (Jiang et al., 2015). Furthermore, GSK3β KD could rescue sensory axon regeneration impaired by miR-26a inhibition (Jiang et al., 2015).

GSK3 signaling regulates axon regeneration via mTORC1-independent protein translation

Unlike CNS neurons, sensory neurons in the PNS can robustly regenerate their peripheral axons. In our earlier study (Saijilafu et al., 2013a), SNI activated the PI3K–AKT pathway and inactivated GSK3β in sensory neurons. Functionally, the expression of a dominant-negative mutant of PI3K in sensory neurons via in vivo electroporation significantly blocked sensory axon regeneration in vivo. Downstream of PI3K–AKT, the mTORC1 inhibitor rapamycin could not block the regenerative axon growth of sensory neurons, whereas the general protein translation inhibitor cycloheximide completely blocked sensory axon regeneration, indicating the requirement of mTORC1-independent translation mechanisms. Similarly, AKT–mTORC1 and its downstream protein translation regulatory molecules, such as S6K1 and 4E-BP, were not involved in GSK3β deletion-induced optic nerve regeneration. However, another protein synthesis inhibitor, anisomycin, compromised optic nerve regeneration in GSK3β-deleted mice (Guo et al., 2016). Taken together, these results provide clear evidence that GSK3 signaling regulates axon regeneration in vivo via mTORC1-independent protein translational mechanisms (Figure 2).

Figure 2.

GSK3 inactivation promotes axon regeneration through mTORC1-independent protein translation. Optic nerve regeneration induced by PTEN deletion and PI3K–AKT activation can be partially blocked by rapamycin, a specific inhibitor of mTORC1 signaling, indicating the requirement of mTORC1-mediated, cap-dependent protein translation. However, optic nerve regeneration induced by GSK3 deletion can only be blocked by the general protein translation inhibitor anisomycin but not rapamycin, indicating the requirement of cap-independent protein translation. Similarly, spontaneous sensory axon regeneration mediated by PI3K–AKT–GSK3 signaling can be impaired by another general protein translation inhibitor, cycloheximide, but not rapamycin. Taken together, downstream of PI3K–AKT signaling, mTORC1 activation and GSK3 inactivation can lead to cap-dependent and cap-independent protein translation, respectively. Based on previous studies, GSK3 inactivation might lead to cap-independent translation via either eIF2B dephosphorylation and activation or mRNA methylation.

Eukaryotic mRNA has a monomethylated cap structure at the 5′ terminus, and protein translation usually involves multiple eukaryotic initiation factors (eIFs), methionyl initiator transfer RNA (Met-tRNAi), and guanosine triphosphate (GTP) to initiate 5′ cap-dependent translation (de la Parra et al., 2018; Figure 3). Specifically, the 40S ribosomal subunit interacts with eIF1/3/5, which then binds to the eIF2-GTP/Met-tRNAi complex to form the 43S preinitiation complex. The eIF4 complex, including eIF4A/E/G, recognizes the mRNA 5′ cap and functions with the 43S complex to search for the translation initiation codon AUG, which is promoted by eIF2-bound GTP hydrolysis (de la Parra et al., 2018). Therefore, eIF2B regulates a rate-limiting step in translation initiation (Campbell et al., 2005; Pavitt, 2005). During cap-dependent translation, eIF4E interacts with the mRNA cap, and its availability is negatively controlled by its binding protein 4E-BP (Gingras et al., 1999; Roux and Topisirovic, 2018). Moreover, 4E-BP phosphorylation by mTORC1 abolishes their interactions and activates eIF4E. Thus, in the presence of rapamycin to inhibit mTORC1, eIF4E is secluded by unphosphorylated 4E-BP, and cap-dependent protein translation is impaired (Roux and Topisirovic, 2018). If so, how does GSK3β inhibition promote axon regeneration in an mTORC1-independent manner? GSK3β can phosphorylate eIF2Bε at the Ser535 site and inhibit protein synthesis (Wang, 2001; Pap and Cooper, 2002). The phosphorylation level of eIF2Bε at Ser535 is higher in adult nonregenerative retinas, indicating reduced translation capacity (Guo et al., 2016). Thus, GSK3β deletion would enhance axon regeneration by reducing eIF2Bε phosphorylation and activating its translation activity. Indeed, overexpression of a dominant-negative mutant eIF2Bε (E572A) significantly eliminated the newly synthesized proteins and axon regeneration, both of which were induced by GSK3β deletion (Guo et al., 2016). Conversely, overexpression of a constitutively active mutant eIF2Bε (S535A) was sufficient to promote axon regeneration, which could be eliminated by anisomycin but not rapamycin (Guo et al., 2016). When cap-dependent translation is impaired, how eIF2Bε mediates protein translation downstream of GSK3β remains unclear. One potential mechanism is that GSK3β inactivation results in cap-independent protein translation that supports axon regeneration.

Figure 3.

mTORC1-regulated cap-dependent protein translation and mRNA methylation-mediated, cap-independent protein translation. In eukaryotes, mRNA has a monomethylated cap structure at the 5′ terminus that is necessary for cap-dependent protein translation with multiple eIFs. Specifically, the 40S ribosomal subunit interacts with eIF1/3/5 to form the 43S preinitiation complex. Meanwhile, the eIF4A/E/G complex recognizes the mRNA 5′ cap and works with the 43S complex to search for the initiation codon AUG. Finally, GTP-bound eIF2 binds to Met-tRNAi to initiate translation upon GTP hydrolysis. Such an eIF2-dependent process is the rate-limiting step that can be repressed by GSK3-mediated eIF2 phosphorylation. Importantly, the availability of eIF4E, which interacts with the 5′ cap, is negatively regulated by its binding protein 4E-BP. Therefore, in mTORC1-dependent protein translation, mTORC1 can phosphorylate 4E-BP and inhibit eIF4E, leading to translation initiation (upper). In the absence of mTORC1, eIF4E is secluded by unphosphorylated 4E-BP and cap-dependent translation cannot proceed, and thus cap-independent translation is activated to support GSK3 inactivation-induced axon regeneration (lower). One possibility is that, in the absence of GSK3, eIF2-GTP/Met-tRNAi could be activated to initiate translation without the 5′ cap. Studies have shown that m⁶A modification of the mRNA 5′UTR is a major approach that enables cap-independent protein translation. Thus, GSK3 inactivation may activate the mRNA methyltransferases Mettl3/14 and subsequent mRNA methylation, leading to cap-independent translation in the absence of the eIF4 complex.

During the past years, much effort was devoted to epigenetic modification of DNA and RNA, including N⁶-methyladenosine (m⁶A) modification of mRNA via methyltransferases Mettl3/14 (Zaccara et al., 2019; Boulias and Greer, 2022). Importantly, mRNA methylation is emerging as a major approach for enabling cap-independent protein translation (Meyer et al., 2015; Coots et al., 2017). Specifically, in the absence of the eIF4 cap-binding complex, ribosomal initiation complexes, including eIF1/2/3 and 40S, could be assembled successfully in m⁶A-containing mRNAs, and a single 5′UTR m⁶A is sufficient to induce translation by binding to eIF3. Moreover, more m⁶A sites in the 5′UTR were found in several types of stressed cells by analyzing transcriptome-wide m⁶A mapping datasets, suggesting that the 5′UTR plays a critical role in m⁶A-mediated, cap-independent translation in response to cellular stresses (Meyer et al., 2015; Coots et al., 2017). SNI inactivates GSK3, which mediates subsequent axon regeneration (Saijilafu et al., 2013a). In a recent study (Weng et al., 2018), immunostaining and genome-wide profiling revealed that SNI elevated m⁶A-tagged mRNAs in sensory neurons. Interestingly, many of these mRNAs were transcribed from regeneration-associated genes (RAGs) or genes encoding translational machinery proteins, suggesting the involvement of cap-independent translation. Functionally, deletion of the methyltransferase Mettl14 in sensory neurons drastically reduced the synthesis of new proteins induced by SNI and impaired axon regeneration. Similarly, Mettl14-mediated mRNA methylation was also necessary for PTEN deletion-induced optic nerve regeneration (Weng et al., 2018), in which GSK3 inactivation was also involved. These results provided strong evidence that both spontaneous sensory axon regeneration in the PNS and enhanced optic nerve regeneration in the CNS required mRNA methylation and cap-independent protein translation (Figure 3). Thus, GSK3 signaling-induced axon regeneration is likely regulated by mRNA methylation-mediated, cap-independent protein translation. To date, no direct evidence supports the role of GSK3 signaling in the regulation of mRNA methylation or cap-independent translation during axon regeneration. Future investigation is required to directly address this issue.

GSK3 signaling regulates neural regeneration by reshaping the transcriptomic landscape

GSK3β signaling could regulate gene transcription in addition to protein translation in sensory neurons to control peripheral nerve regeneration in vivo. Using two compartmental culture devices insulating sensory neuronal soma from their axons, we found that PI3K–GSK3 signaling functioned specifically in the neuronal soma but not in axons to support axon regeneration, suggesting its role in regulating transcription (Saijilafu et al., 2013a). In the canonical Wnt pathway, GSK3β phosphorylates β-catenin, leading to its degradation (Wu and Pan, 2010). Upon Wnt activation, GSK3β is sequestered away, which subsequently results in β-catenin stabilization to regulate gene transcription. Studies have provided evidence that the Wnt–β-catenin pathway promotes spinal cord regeneration in fish and optic nerve regeneration in both fish and mice (Garcia et al., 2018). However, the protein level of β-catenin in regenerating sensory neurons was not altered (Saijilafu et al., 2013a), and deletion of β-catenin did not influence sensory axon regeneration in vivo. Instead, the transcription factor Smad1 was identified as the downstream target of the PI3K–GSK3β pathway in sensory neurons to control axon regeneration. Smad1 KD in sensory neurons impaired sensory axon regeneration in vivo, whereas overexpression of an active mutant of Smad1 alone could not enhance axon regeneration, indicating that Smad1 must coordinate with other factors to promote sensory axon regeneration. Finelli et al. (2013) showed that peripheral nerve injury increased the level of histone acetylation, which is an important epigenetic regulatory factor for shaping the cellular transcriptomic landscape and has to coordinate with the transcription factor Smad1 to regulate the transcription of many RAGs. Therefore, GSK3-regulated Smad1 works with histone modification to regulate neuronal gene transcription.

In addition, recent progress in cellular reprogramming during neural regeneration has shed some light on novel mechanisms by which GSK3 signaling reshapes neuronal transcriptomics. As summarized in our recent review (Yang et al., 2022), after neural injury, mature sensory neurons in the PNS reprogram their transcriptomic landscape back to a state similar to that in younger neurons (Zhou et al., 2006; Chandran et al., 2016; Renthal et al., 2020). Similarly, mature CNS neurons could transiently reverse back to a younger cellular state at the transcriptional level after injury but fail to maintain it (Poplawski et al., 2020). Such rejuvenation of neurons leads to the restoration of the axon growth capacity of mature neurons, thus promoting axon regeneration. Interestingly, many genes identified to enhance CNS neural regeneration to date are factors known to be involved in reprogramming induced pluripotent stem cells (iPSCs), such as Lin28, Klf4, Oct4, Sox2/11, and c-Myc (Jing et al., 2012; Belin et al., 2015; Wang et al., 2015; Norsworthy et al., 2017; Galvao et al., 2018; Nathan et al., 2020). More importantly, the combined expression of three Yamanaka factors (Oct4/Sox2/Klf4) in RGCs enhanced optic nerve regeneration in both young adult and aged mice (Lu et al., 2020). Furthermore, the Jak–Stat3 pathway, which enhances CNS neural regeneration, has been shown to act downstream of leukemia inhibitory factor (LIF) to help keep embryonic stem cells (ESCs) in a pluripotent state (Niwa et al., 1998; Raz et al., 1999). Meanwhile, many studies in the iPSC field have shown that targeted somatic cells are rejuvenated after being reprogrammed into pluripotent stem cells (Ocampo et al., 2016; Cornacchia and Studer, 2017; Roux et al., 2022). Although neurons cannot be reprogrammed back to stem cells, reprogramming factors and signaling pathways involved in stem cell pluripotency could be key mechanisms underlying the reprogramming of the cellular state of neurons at the transcriptional level, supporting neural regeneration (Yang et al., 2022). To date, several pieces of evidence have shown that GSK3 signaling is involved in reprogramming the cellular transcriptomic landscape. First, it is well known from early studies that GSK3 signaling, downstream of canonical Wnt signaling, plays an essential role in maintaining the pluripotency of human or mouse ESCs. Sato et al. (2004) showed that LIF-induced Stat3 was insufficient to maintain the undifferentiated state of human ESCs, whereas GSK3 inhibition could sustain pluripotency by promoting the expression of Oct3/4 and Nanog. Second, several studies have shown that replacement of serum in the culture medium with both MEK and GSK3 inhibitors (2i) could better maintain ESCs in the naïve ground state or promote the reprogramming efficiency of iPSCs (Silva et al., 2008; Ying et al., 2008; Doble and Woodgett, 2009; Feng et al., 2009; Schlesinger and Meshorer, 2019). Importantly, the effect of 2i on maintaining pluripotency was achieved by modifying the cellular chromatin and epigenetic states, such as reducing global DNA methylation and local histone 3 lysine 27 trimethylation (H3K27me3) (Sim et al., 2017; Schlesinger and Meshorer, 2019). For instance, GSK3 inhibition decreased the expression of DNMT3 in ESCs, thereby downregulating DNA methylation without affecting 5hmC levels (Sim et al., 2017). In another study (Weng et al., 2017), removal of the intrinsic epigenetic barrier by active DNA demethylation could trigger robust axon regeneration in the adult mammalian nervous system. Mechanistically, PNS injury induces DNA demethylation and activates RAGs in a Tet3- and thymine DNA glycosylase-dependent manner, leading to functional axon regeneration in adult DRG neurons. Moreover, PTEN deletion-induced axon regeneration of mature RGCs was attenuated upon Tet1 KD. The histone methyltransferase Ezh2, a potential target of GSK3β (Ma et al., 2013), was also documented to rejuvenate mature CNS neurons at the transcriptomic level by targeting H3K27me3, thus favoring axon regeneration (Wang et al., 2022). However, to date, no direct evidence has shown that GSK3β acts upstream of these epigenetic modifiers to control axon regeneration. Clearly, further research is required to determine the underlying mechanisms. Third, Smad1, downstream of GSK3 signaling, has been shown to enhance reprogramming to pluripotency through its target gene, inhibitor of differentiation (Hayashi et al., 2016). Moreover, increased histone acetylation promotes pluripotency (Li et al., 2020, 2022). These results are consistent with the finding that increased histone acetylation works together with GSK3–Smad1 signaling to reprogram the transcriptomic landscape of mature sensory neurons for regeneration (Finelli et al., 2013).

In addition to iPSCs, in zebrafish after retinal injury, Müller glial cells can be reprogrammed at the transcriptional level to become pluripotent retinal progenitor cells, which then redifferentiate into different retinal cell types for functional repair (Lahne et al., 2020). In response to retinal injuries, Müller glial cells produce several growth factors and cytokines, such as Wnt, insulin, HB-EGF, TNFα, ADP, and CNTF, which contribute to their reprogramming and retinal regeneration in an autocrine or paracrine manner. Downstream of these factors, many well-known signaling pathways regulate retinal regeneration in zebrafish, including PI3K–AKT, GSK3β, ERK, and Jak (Goldman, 2014), all of which regulate axon regeneration (Qian and Zhou, 2020). Moreover, these signaling pathways act through multiple transcription factors and/or RNA-binding proteins to form regeneration-associated transcriptional cascades, such as β-catenin, Stat3, Ascl1, c-Myc, and Lin28, which also regulate axon regeneration (Goldman, 2014; Lahne et al., 2020; Qian and Zhou, 2020). In particular, the Wnt–GSK3β–β-catenin pathway is best studied, and pharmacological inhibition of GSK3β can initiate Müller glial cell reprogramming and progenitor formation in uninjured retinas (Ramachandran et al., 2011). Unlike that in zebrafish, retinal injury in mammals cannot induce Müller glial cell reprogramming to form progenitor cells (Garcia-Garcia et al., 2020; Lahne et al., 2020; Qian et al., 2021). Based on knowledge from zebrafish, forced activation of GSK3β–β-catenin signaling in mouse Müller glial cells could stimulate Müller glial cells to reprogram to progenitor cells and redifferentiate into rod photoreceptors (Yao et al., 2016, 2018). Importantly, such a strategy could not restore vision in a mouse model of congenital blindness via de novo genesis of rod cells. Similarly, GSK3β KO in mouse Müller glial cells was sufficient to promote their reprogramming without retinal injury via the Lin28/let-7 pathway (Yao et al., 2016). Collectively, these results suggest that GSK3β signaling could reprogram the cellular transcriptomic landscape to control neural regeneration.

Finally, downstream of GSK3 signaling, CRMP2, CLASPs, APC, Tau, and MAP1B are potential GSK3β substrates in regulating microtubule dynamics during axon regeneration at the growth cone, which have been summarized before (Zhou and Snider, 2005; Hur et al., 2012; Liu et al., 2012, 2020). Except for CRMP2 and CLASPs (Hur et al., 2011; Liz et al., 2014; Leibinger et al., 2019), whether APC, Tau, or MAP1B acts downstream of GSK3β signaling to regulate axon regeneration in vivo has not been directly tested.

Future directions

GSK3β signaling plays a pivotal role in the regulation of axon regeneration in vivo by orchestrating gene transcription and protein translation in the neuronal soma and cytoskeletal dynamics at the growth cone. Many studies have shown that several factors that enhance neural regeneration converge on mTORC1 signaling to regulate protein translation. As discussed above, mTORC1 regulates cap-dependent protein translation, whereas GSK3β controls neural regeneration and regeneration-associated protein translation via an mTORC1-independent mechanism. Because protein translation is necessary for GSK3β-mediated sensory axon and optic nerve regeneration, it is reasonable to hypothesize that GSK3β signaling is involved in controlling cap-independent translation via an mRNA methylation-mediated mechanism.

Given the broad regulatory function of GSK3β signaling, it may be another common mediator of neural regeneration. For instance, our recent studies have shown that Lin28, an RNA-binding protein and a well-known reprogramming factor, could enhance both optic nerve and CST axon regeneration; overexpression of Lin28 in neurons not only activated mTORC1 signaling but also inactivated GSK3β (Wang et al., 2018; Nathan et al., 2020). Furthermore, Lin28 has been shown to act downstream of the Wnt–GSK3β–β-catenin pathway to reprogram mouse Müller glial cells after retinal injury (Yao et al., 2016). Thus, exploring the roles of GSK3β signaling in all known pathways that govern neural regeneration would identify new upstream regulators of GSK3β (Figure 1). Moreover, recent rapid progress in multiomics-based sequencing and big data analysis provides useful tools for exploring signaling mechanisms at the transcriptomic, proteomic, and/or metabolomic levels in a spatiotemporal manner. Therefore, future investigations using such cutting-edge strategies will be of great importance for revealing the downstream targets of GSK3β signaling that contribute to neural regeneration.

Despite the substantial role of GSK3β in coordinating axon regeneration, GSK3β itself might not be an ideal target because of its broad substrates and wide participation in various physiological and pathological processes. Therefore, deciphering the detailed molecular networks that involve GSK3β signaling would be crucial for future studies. Emerging technologies, such as protein microarrays and proximity labeling (Han et al., 2018; Syu et al., 2020; Qin et al., 2021), could be used to identify novel and specific GSK3β-binding partners in specific cell types or microenvironments. These newly discovered binding proteins may include novel regulators of GSK3β activity or new GSK3β substrates specifically related to neural regeneration (Figure 1), which could be potential therapeutic targets for promoting neural regeneration in a more targeted manner.

Conflict of interest: none declared.