Axon Regeneration: A Subcellular Extension in Multiple Dimensions

Abstract

Axons are a unique cellular structure that allows for the communication between neurons. Axon damage compromises neuronal communications and often leads to functional deficits. Thus, developing strategies that promote effective axon regeneration for functional restoration is highly desirable. One fruitful approach is to dissect the regenerative mechanisms used by some types of neurons in both mammalian and nonmammalian systems that exhibit spontaneous regenerative capacity. Additionally, numerous efforts have been devoted to deciphering the barriers that prevent successful axon regeneration in the most regeneration-refractory system—the adult mammalian central nervous system. As a result, several regeneration-promoting strategies have been developed, but significant limitations remain. This review is aimed to summarize historic progression and current understanding of this exciting yet incomplete endeavor.

Different from other cell types, the functional identity of a neuron in all organisms is largely defined by its synaptic connections (i.e., the inputs it receives on the cell body and dendrites and the outputs its axon innervates). Neuronal injury or disease can disrupt this critical neural circuitry via damage to both dendrites and axons and eventually cell death. Magnifying the problem, neurons are postmitotic and thus lose the ability to proliferate. Fortunately, many injured neurons can survive and are sometimes able to regenerate their axons and dendrites, as seen in evolutionarily older organisms, the peripheral nervous system (PNS), and even the immature mammalian central nervous system (CNS). However, such regenerative events are largely inefficient in the adult mammalian CNS, often resulting in permanent functional deficits after injury and disease. Thus, there has been a long-term interest to develop strategies that promote neuronal regeneration and functional restoration.

At the cellular level, neuronal regeneration could occur at the axonal or dendritic compartments. Perhaps due to their importance for neuronal function and distinctive structure (i.e., impressive lengths) that contributes to high injury vulnerability, most efforts have been devoted to studying axon rather than dendrite regeneration. Depending on the regrowth initiation site (the axon shaft, branch, or terminals), the generic term “axon regeneration” is sometimes used for different morphological forms, including regenerative regrowth from injured sites or the distal uninjured segment, and compensatory sprouting from spared axons. Inefficient axonal regeneration in the mammalian adult CNS is attributed to a variety of cell intrinsic and extrinsic factors, as this review will discuss.

MILESTONES IN AXON REGENERATION

Understanding the processes and limitations of axon regeneration has been a key research topic for many decades. In the early twentieth century, Santiago Ramón y Cajal made a keystone contribution to the field by identifying axons as the basic unit of neuronal connectivity. He noted structural changes within axons and the surrounding environment occurred following injury. Since then, science has made tremendous progress in understanding the different factors necessary for, or limiting, the induction of axon regeneration. The current section aims to review milestones in mammalian axon regeneration research that have shaped the field and laid a critical foundation for future studies (Fig. 1).

Figure 1.

Major milestones of axon regeneration research. In 1907, Tello (pupil of Ramón y Cajal) produced the first evidence that damaged central nervous system (CNS) axons can regenerate if placed near the permissive peripheral nervous system (PNS) environment. Tello anastomosed an optic nerve stump (1) to a sciatic nerve graft (2) and observed a scar (3) as well as purported RGC axons crossing the site of anastomosis (4) regenerating into the sciatic nerve (7). (5) Indicates a vein, (6) indicates neurilemma of the sciatic nerve (adapted from Tello 1907). Tello's work was complemented by Ramón y Cajal's histological descriptions of the degeneration and regeneration of the CNS, published in 1913 and translated in 1928 (adapted from Ramón y Cajal 1928). In 1981, David and Aguayo confirmed Tello's findings that CNS neurons can regenerate in the permissive PNS. (1) Diagram of the dorsal aspect of a rat spinal cord, with a peripheral nerve “bridge” between the medulla and thoracic spinal cord. Cross sections of the medulla and spinal cord show where the bridge was directly attached. (2) The origin of axons innervating the graft were determined with horseradish peroxidase (HRP) retrograde labeling and found to originate in the brainstem and spinal cord (figure adapted from David and Aguayo 1981). In 1984, Richardson and Issa used peripheral nerve grafts to investigate the neuron intrinsic mechanisms of CNS injury and regeneration; long spinal axons of dorsal root ganglia (DRG) were 100 times more likely to regenerate into peripheral nerve grafts in the dorsal column (following bilateral dorsal column lesions) if their peripheral axons were also cut (adapted from Richardson and Issa 1984). In 1999, Neumann and Woolf showed that altering intrinsic growth capacity of DRGs with a similar preconditioning lesion can promote regeneration of central axons in CNS tissue. Animals with a peripheral axotomy 1 to 2 weeks prior to (but not concomitant with) a dorsal column lesion show growth into and above the lesion, and that DRG explants show increased neurite outgrowth following preconditioning lesion (image shows explant results; adapted from Neumann and Woolf 1999). Focusing on neuron extrinsic factors inhibiting regeneration, in 1988 Caroni and Schwab identified myelin-derived membrane proteins with nonpermissive substrate properties; a monoclonal antibody, IN-1, neutralizes the nonpermissive substrate properties (adapted from Caroni and Schwab 1988a). A landmark study by Park et al. in 2008 highlighted the utility of altering the intrinsic state of neurons to promote axon regeneration–PTEN deletion (conditional knockout [CKO]) was shown to promote exceptionally robust axon regeneration in an optic nerve crush model (scale bar, 100 µm) (adapted from Park et al. 2008). Figure created with BioRender.com.

Pre-1985: Peripheral Nerve Grafts Enable Central Nervous Axon Regeneration in Mammals

Based on morphological observations, Ramón y Cajal concluded that the CNS does not have the capacity for regeneration following injury and disease (Ramón y Cajal 1928). This stands in stark contrast to the PNS, which is intrinsically more regenerative than the CNS. An obvious question is what differences account for this dichotomy between the CNS and PNS. Ramón y Cajal hypothesized that CNS axons can regenerate if placed near the growth-permissive environment of the PNS. His pupil, Jorge Francisco Tello, tested this by anastomosing a sciatic nerve graft (i.e., PNS) to a cut optic nerve (i.e., CNS), and observed axons growing from the optic nerve into the sciatic nerve graft (Tello 1907). Tello suspected these axons were from regenerating retinal ganglion cells (RGCs).

It was not until the early 1980s that Ramón y Cajal and Tello's hypothesis was proven by Albert Aguayo and his colleagues (David and Aguayo 1981). In a landmark study, the authors used autologous sciatic nerve grafts to serve as a bridge between the medulla and spinal cord. Later, the authors examined the extent of CNS axon ingrowth into the PNS grafts via horseradish peroxidase (HRP)-mediated tracing. This provided definitive proof that the axons in the grafts were of CNS origin. Despite a small number of regenerating axons, Aguayo's work served to reinvigorate modern CNS research.

1985–1995: The Injury Environment Dictates Capacity for Axon Regeneration in Mammals

Inspired by Aguayo's results, the field began to search for factors within the CNS injury environment responsible for inhibiting regeneration. The hypothesis was that if these inhibitory factors are identified and removed, then CNS axon regeneration would be possible. The advent of neuronal culture techniques in the 1980s enabled the dissection of putative inhibitory molecules in vitro via reductionist approaches. In 1988, studies from Caroni and Schwab implicated CNS myelin-associated inhibitors (MAIs) as major contributors, in particular NogoA, to the inhibitory CNS environment (Caroni and Schwab 1988a,b). Since these early studies, numerous additional inhibitory factors have been identified (Wang et al. 2002; Filbin 2003).

However, removing these MAIs by genetic or pharmacological means only modestly improved axon regrowth in vivo (Zheng et al. 2003; Schwab and Strittmatter 2014), suggesting the existence of other inhibitors independent of oligodendrocytes or myelin in vivo. Indeed, components of the fibrotic/glial scar, which forms at the lesion site, appear to be another major source of inhibitory activity. Like in other parts of the body, the immediate injury response in the CNS initially serves to isolate the injury and protect the surrounding spared tissue, but later evolves into a barrier for axon regrowth (O'Shea et al. 2017; Tran et al. 2018). Deciphering the molecular nature of the inhibitory activity associated with such scar structures has been a major focus of the field for the past decades. Jerry Silver observed that growing axons cease around the region enriched in chondroitin sulfate proteoglycans (CSPGs), and proposed that those, and perhaps other extracellular matrix (ECM) molecules, are key factors for limiting axon regeneration (McKeon et al. 1995). Supporting this idea, application of chondroitinase ABC, an enzyme that degrades CSPGs, is able to promote axon regrowth in vivo to a limited extent (Bradbury et al. 2002; Lang et al. 2015). Thus, these studies provide convincing evidence for environmental inhibition, but also point to the insufficiency of removing these in promoting robust axon regrowth.

1990s: Switching of Neuronal State Impacts Axon Regeneration in Mammals

In parallel to studying environmental factors, several observations pointed to the importance of neuronal intrinsic mechanisms. For example, pioneering work by Marie Filbin suggested that different cAMP levels in PNS and CNS neurons contribute to the dichotomous regenerative responses (Filbin 2003). From the perspective of neuronal development, axon growth stops upon the completion of development, even for PNS neurons. So, the question became, how can they reenter a growth-competent state following injury? Initial insights came from a series of studies on “conditioning lesion models” first described by Richardson and Issa (1984), and later extensively characterized by Neumann and Woolf (1999). Dorsal root ganglion (DRG) neurons have a unique morphology, with their soma residing within the DRG and a single stem axon bifurcating into a central and peripheral branch: the peripheral branch (i.e., PNS) innervates peripheral targets, and the central branch (i.e., CNS) relays somatosensory information to the brain. A lesion to the peripheral branch will repair itself but a lesion to the central branch will not. Remarkably, an initial injury to the peripheral branch allows the central branch to regenerate when it is injured days later (Richardson and Issa 1984; Neumann and Woolf 1999). This phenomenon is known as the “conditioning effect,” as the peripheral injury “conditions” the initially dormant DRG neuron to reenter a growth-competent state by up-regulating a genetic program for regeneration. This internal change in cell state is sufficient for the central branch to overcome the “inhibitory” CNS injury environment. Although a set of regeneration-associated genes that include transcription factors (TFs) (e.g., ATF3, Sox11, STAT3) and growth-associated proteins (e.g., GAP43, SPRR1a, and cytoskeletal proteins), have been described as molecular players, it remains unclear which are the drivers or passengers of the conditioning lesion effects.

2000s: Altering the Intrinsic State of Neurons Can Promote Axon Regeneration

Despite studies in PNS neurons, it was not clear whether an intrinsic growth program plays a role in axon regeneration of adult CNS neurons. Drawing inspiration from evolutionarily conserved pathways that regulate cell growth, Park et al. (2008) hypothesized that mechanisms preventing overgrowth upon the completion of development are likely responsible for the reduced regenerative ability of adult CNS axons. In the early 2000s, the development of different transgenic mouse lines and the advance of AAV-based manipulation methods made it possible to search for neuronal inhibitors of axon regeneration in vivo. Screening specific tumor suppressor genes in adult RGCs identified that the gene encoding phosphatase and tensin homolog (PTEN) prevents apoptosis and promotes exceptionally robust axon regeneration following optic nerve crush (ONC). PTEN is a known inhibitor of the cell growth promoting PI3K/mTOR pathway. Follow-up studies showed that PTEN deletion is also effective in promoting regeneration in more complex cell types such as corticospinal neurons (Liu et al. 2010), and is effective even 1 year after injury (Du et al. 2015). These studies raised the possibility of manipulating intrinsic programs as an effective means of promoting CNS axon regeneration.

Intriguingly, manipulating PTEN/mTOR leads to the regeneration of only subsets of injured axons (Duan et al. 2015). Further studies revealed that even in a seemingly homogenous population of neurons, such as RGCs, different subtypes differ in intrinsic regenerative ability and response to injury and manipulations (Duan et al. 2015; Norsworthy et al. 2017; Tran et al. 2019). Importantly, the characterization of molecular properties of these different types of neurons (e.g., via modern single-cell technologies) offers a new perspective to dissect the mechanisms of the selective axon regenerative ability.

MODEL SYSTEMS FOR AXON REGENERATION

Mechanistically, axon regeneration is a multifaceted cell biological process, making it difficult to study. Experimental models to study the neuronal response to an injury in vitro or in vivo, can either focus only on one aspect to dissect the cellular and molecular events of axon growth or aim at more closely resembling the complex in vivo conditions. In the following section, we will introduce a limited number of these model systems and touch on their advantages to study axon regeneration.

Invertebrates and Zebrafish

Equipped with ample genetic and experimental tools (such as in vivo imaging and large-scale screening), invertebrates, in particular Caenorhabditis elegans and Drosophila, are the leading genetic model systems used to decipher axon regeneration (Fig. 2A; Fang et al. 2012; He and Jin 2016; Brace and DiAntonio 2017; Rasmussen and Sagasti 2017). For example, C. elegans has a simple nervous system containing only 302 neurons and a transparent cuticle, facilitating the study of axon injury of individual neurons labeled with a fluorescent marker and monitoring their regenerative response in vivo. In 2004, Yanik et al. presented the first study using a femtosecond laser surgery to axotomize a single axon without perturbing nearby processes in the living worm (Yanik et al. 2004). As a result of their spontaneous regenerative capacity and short reproductive cycle, the C. elegans model system is particularly powerful to identify molecules, such as dual leucine zipper kinase ([DLK]; discussed later), associated with axon regeneration. As vertebrates, zebrafish offer the advantage of sharing more similarities with mammals while providing an intrinsically pro-regenerative system to study axon regeneration (Becker and Becker 2014). For example, the growth-permissive bridge formed by astrocytes and other cell types following CNS injury in zebrafish offers a useful model to dissect the mechanisms of inhibitory scar formation in mammals versus the permissive cell bridge in zebrafish (Mokalled et al. 2016).

Figure 2.

Models to study axon regeneration. (A) Invertebrates (e.g., Caenorhabditis elegans and Drosophila melanogaster) and zebrafish are well established models to study axon regeneration. They have short reproductive cycles to quickly screen gene function. The transparent cuticle and wings, of C. elegans and Drosophila can be used for laser-induced axon regeneration studies in vivo. The growth permissive bridge at the injury site distinguishes the zebrafish as a central nervous system (CNS) regeneration-competent vertebrate. (B) Other vertebrates or mammals like the mouse, Mus musculus, have a regeneration competent peripheral nervous system (PNS), but incompetent CNS. In the mammalian system, (1) the dorsal root ganglion (DRG) injury is used to model regeneration-competent PNS injury, or (2) to study the “preconditioning” effect for central branch injury. (3) Sciatic nerve injury (SNI) is another PNS injury model that keeps the neural tube intact to facilitate regeneration. In the optic nerve crush (ONC), viral vectors are used to manipulate retinal ganglion cell (RGC) gene expression and/or anterograde axon labeling (4) to assess axon regeneration. Generally, this is measured from the injury site (5) and divided into short (>2 mm) and long (<2 mm) distance regeneration. Spinal cord injury (SCI) models used are (6) the contusion model to mimic clinical relevance, or more consistent models like (7) lateral or (8) full transection to study specific aspects of axon regeneration and the environmental contribution. Figure created with BioRender.com.

Mammals

In Vitro Injury Models

Upon isolation, most PNS neurons from all developmental stages and early developing CNS neurons can survive in culture, permitting the dissection of cellular and molecular mechanisms of axon growth. These neurons can be cultured on different substrates (e.g., MAIs and CSPGs), thereby offering unique models to probe into cellular mechanisms of axon regeneration (Witte et al. 2008; Al-Ali et al. 2017; Rodemer et al. 2020). Furthermore, with advanced technologies, these culture systems provide powerful systems to address questions about subcellular events of axon growth, such as structural organization of the cytoskeleton (Xu et al. 2013) and the source driving axon extension (Santos et al. 2020).

In Vivo PNS Injury Models

In comparison to in vitro models, in vivo models have additional advantages to assess (1) complex interactions between injured axons and its surroundings, and (2) functional/behavioral outcomes. Here we summarize the most used surgical paradigms (Fig. 2B). For PNS, sciatic nerve crush (SNC) is a well characterized model to study axon regeneration (Bauder and Ferguson 2012). Typically, a crush lesion transects axons of the nerve and the distal stumps undergo Wallerian degeneration. However, usually the Schwann cell basal laminae and endoneurial tube are left intact, providing a growth permissive environment for axon regeneration from the proximal stump into the tube to allow re-innervation of target cells (Namgung 2014; Gordon 2020). In addition to analyzing neuronal mechanisms, this model has been a powerful system to probe the neuron-immune interaction and their impact on axon regeneration (Kalinski et al. 2020).

In Vivo CNS Injury Models

As adult mammalian CNS neurons often do not survive in culture, in vitro systems have clear limitations. Despite being prone to variations and technical artifacts (Di Maio et al. 2011; Tuszynski and Steward 2012; Fischer et al. 2017), several in vivo models have been instrumental in studying axon regeneration.

Optic Nerve Crush (ONC). As a part of the CNS, the optic nerve consists of the axons of all RGCs, relaying visual information to the brain. The noninvasive surgical access to the optic nerve allows for the induction of a highly reproducible injury with the complete transection of all RGC axons. Because of the unique anatomical structure, this model represents a pure white matter injury model after which nearly no axons spontaneously regenerate. The RGCs are easily accessible to test different manipulations for their effects on neuronal survival and axon regeneration. For example, introduction of zymosan, an inflammation-inducing agent, was shown to trigger axon regeneration (Leon et al. 2000; Fischer et al. 2001). Furthermore, the intravitreal injection of AAVs to transduce and/or manipulate genes expressed by the RGCs has identified many factors as positive or negative regulators of axon regeneration (Park et al. 2008; Sun et al. 2011; Norsworthy et al. 2017), some of which have also shown their pro-regenerative effect in other CNS areas (Liu et al. 2010, 2017).

Spinal Cord Injury (SCI). Conserved injury responses and regenerative outcomes are shared after SCI in humans and other mammals. Thus, SCI models are used to analyze (1) axon regeneration, (2) the correlation and causality of an individual axon tract with behavioral performance, and (3) the clinical relevance of experimental treatments via assessing functional recovery. Different types of SCI models have been devised. As distinguishing regenerating axons from spared axons could be difficult, several consistent injury models with relatively clean background have been designed (Tuszynski and Steward 2012). For example, complete and lateral hemisection models involve precise transection of the entire or half spinal cord and are better for assessing the effects of specific manipulations on axon regeneration. On the other hand, as many human SCIs involve incomplete contusive injury, several versions of contusion injury have been designed to inflict a transient and acute injury to the spinal cord (Gruner 1992; Stokes 1992). Despite having more variable lesions, these models are widely used as preclinical experimental models. In addition, all of these traumatic injuries trigger complex injury responses in the environment, thus being useful to study the evolving interactions among different cell types (O'Shea et al. 2017).

FROM INJURY TO REGENERATION: A FEW DECISION-MAKING PROCESSES

Upon an injury, neurons have a series of decisions to make, starting with whether and how the axon will seal its injured terminal and communicate local injury information to the soma? Followed by, will the injured neuron survive, subsequently regenerate, and to what extent? Last, how, why, and at what point does regeneration stop? Importantly, such decision-making processes depend not only on axons, but also on their cell bodies and local environment. Using the various models described above, numerous observations have shed light onto these questions and this section will give an overview on the current understanding of different decision-making events in axons, cell bodies, and their environment following injury.

Axonal Decisions

Axonal Decision 1—Whether to Form a Growth Cone or Retraction Bulb

Following injury and Ca²⁺-mediated membrane resealing (Fig. 3(1)), the injured axons need to reorganize the membrane and cytoskeleton structure. During development, axon growth is guided by a growth cone at the axonal terminal (Lowery and Van Vactor 2009). The cytoskeleton associated with the growth cone and axon shaft provide the driving force for extension. A plausible hypothesis is that a growth cone-like structure forms in the tip of injured axons, which supports axonal stability and extension of regeneration-competent neurons (Blanquie and Bradke 2018). In contrast, regeneration-incompetent axons attempt to sprout following an injury but ultimately fail and form a “retraction bulb” with disorganized microtubule networks (Ertürk et al. 2007). Microtubule stabilizers/destabilizers have been shown to significantly alter regeneration phenotypes. For example, addition of pharmacological destabilizers to in vitro cultures of injured PNS neurons transforms their growth cones into growth-incompetent retraction bulbs (Derry et al. 1997; Khan and Ludueña 2003), whereas pharmacologic stabilization of microtubules reduces degeneration in CNS neurons in vitro (Hellal et al. 2011; Ruschel et al. 2015).

Figure 3.

Soma, axonal, and environmental events following axon injury. (1) The initial response to injury of any axon is an immediate influx of calcium into the tip of the severed axon, which triggers activation of calpains and subsequent Ca²⁺-dependent membrane fusion to reseal the ruptured end of the axon. (2) Calcium also leads to activation of local protein synthesis (e.g., of STAT3 and importin-β, which are retrogradely transported to initiate regenerative programs at the soma) and protein activity such as DLK-1. (3) Additionally, an injury-induced back-propagating calcium wave induces changes in gene expression, such as the (4) epigenetic changes seen in regenerative peripheral nervous system (PNS) neurons (e.g., dorsal root ganglion [DRG]) via activation of protein kinase Cµ (PKCµ), resulting in nuclear export of HDAC5 (histone deacetylase 5). HDAC5's absence in the nucleus enables a transition from repressive DNA histone methylation to permissive acetylation within the nucleus, which (5) allows transcription factors such as ATF3 to bind to accessible sites. (6) In contrast, nonregenerative axons of the adult central nervous system (CNS) do not induce similar epigenetic changes so that DNA remains in a methylated, repressive state. (7) The postinjury microenvironment plays a critical role in dictating regeneration competence. Activated microglia clear apoptotic cells and debris from the damaged extracellular matrix, while releasing inflammatory mediators to recruit leukocytes from the periphery. Neutrophils secrete damaging proteolytic enzymes. Recruited macrophages to the injury site become indistinguishable from activated microglia, performing phagocytosis and proinflammatory roles. NG2-expressing oligodendrocyte precursor cells (OPCs) also contribute to inhibition of axon regeneration. Finally, astrocytes contribute to the formation of the glial scar in the CNS but not PNS to form a chemical and physical barrier to axon regeneration. (PIC) Peripheral immune cell, (CK) cytokine, (Ast.) astrocyte, (PGs) proteoglycans, (GPs) glycoproteins, (GIMs) growth inhibitory molecules, (FC) foam cell, (MΦ) macrophage, (µG) microglia, (ROS) reactive oxygen species, (PMN) polymorphonuclear cell (neutrophil). Figure created with BioRender.com.

Axonal Decision 2—How to Inform the Cell Body

Early studies showed that some forms of axon regrowth can occur in the absence of cell bodies (Harris et al. 1987), indicating an axon-intrinsic injury response. However, for most types of sustained axon regrowth, the coordinated action between axonal compartments and their cell bodies is required. Thus, an important question is what and how injury signals are generated and transported to the soma.

Several pathways have been implicated, with the evolutionary conserved DLK/MAPK pathway having been studied most extensively (Fig. 3(2)). Increased activity of DLK-1 in response to axon injury was first discovered in C. elegans (Hammarlund et al. 2009). Axon-localized DLK, and perhaps its homolog LZK (leucine zipper–bearing kinase), acts as an injury sensor activated by calcium elevation and cytoskeleton disruption (Hammarlund et al. 2009; Ghosh-Roy et al. 2010; Hao et al. 2016). Both DLK and LZK belong to the evolutionarily conserved family of mitogen-activated protein kinase kinase kinases (MAPKKKs). Their activation leads to the activation of JNK and p38 MAP kinases, which are then retrogradely transported to the soma in a JIP3-dependent manner. As a result, the injury signals generated in injured axons are broadcasted in the entire neuron. A striking demonstration of the importance of the DLK-mediated injury signaling is that the genetic deletion of DLK prevents the axon regeneration induced by PTEN inhibition (Watkins et al. 2013). However, DLK and LZK were also identified as proapoptotic molecules in RGCs (Welsbie et al. 2013), consistent with the notion that uncontrollable activation of this stress pathway may result in neuronal death.

In addition to calcium influx at the severed axon tip, elevated intra-axonal calcium may arise from L-type voltage-gated calcium channels and the endoplasmic reticulum (ER), both engendering changes in signaling and gene expression (Fig. 3(3)). For example, protein kinase Cµ (PKCµ)-mediated export of HDAC5 induces epigenetic changes in DRG neurons that allow expression of immediate early genes (e.g., ATF3, Fig. 3(4–6); Cho et al. 2013). In addition, many other molecules have also been implicated as putative retrograde injury signals (Terenzio et al. 2017).

Figure 4.

The PI3K/AKT/mTOR pathway. The PI3K/AKT/mTOR pathway is activated through the binding of growth factors to the receptor tyrosine kinase (RTK) complex, which in turn activates PI3K. The activated form of PI3K converts PIP₂ to PIP₃, which then mediates the phosphorylation of AKT. AKT acts on a wide range of substrates involved in the regulation of important cellular function. This includes blocking negative regulators of mTOR, such as TSC1 or TSC2. Active mTOR is associated with high-protein synthesis, cell proliferation, and cell growth. Phosphatase and tensin homolog (PTEN) acts as a tumor suppressor that negatively regulates this pathway by removing the 3-phosphate from PIP₃. Loss of PTEN leads to overactivation of the mTOR pathway and has been shown to promote exceptionally robust axon regeneration following central nervous system (CNS) injury (Park et al. 2008; Liu et al. 2010). This effect can be abolished with the application of rapamycin, a known inhibitor of mTOR, indicating that the pro-regenerative effects of this treatment are mTOR-dependent (Park et al. 2008). mTOR-independent signaling mechanisms, however, appear to also be important for axon regeneration. This is underscored by the finding that genetic deletion of other negative regulators of mTOR, such as TSC2, only partially mimic the effects of PTEN deletion in the CNS (Park et al. 2008); therefore, PTEN deletion likely works via additional effectors (such as glycogen synthase kinase 3 [GSK3]) to promote axon regeneration. (RTK) Receptor tyrosine kinase, (PI3K) phosphoinositide 3-kinase, (PIP₂) phosphatidylinositol 4,5-bisphosphate, (PIP₃) phosphatidylinositol 3,4,5-triphosphate, (TSC) tuberous sclerosis complex, (mTOR) mammalian target of rapamycin. Figure created with BioRender.com.

Cell Body Decisions

Cell Body Decision 1—Apoptosis or Survival

Upon transmission of injury signals, the soma response results in one of three different outcomes: (1) neuronal death, (2) neuronal survival with, or (3) without axon regeneration. Many injured descending projection neurons survive but fail to regenerate after SCI, in contrast to injured PNS neurons, which survive and regenerate. After an ONC close to the proximal axonal segment, the majority of injured RGCs undergo cell death (Hu et al. 2012; Tran et al. 2019). However, neuronal survival increases if the axotomy occurs in the distal axonal segments within the optic tract. There are at least two explanations for such length-dependent death: the distal injury may spare other axonal branches (thus preserving some target-derived trophic signals) and/or induce less injury signals compared to the proximal lesion. However, it is unknown why the activation of these injury signals fails to induce neuronal death in other types of neurons such as peripheral DRG neurons.

Cell Body Decision 2—Conversion to a Growth-Competent or -Incompetent State

To support axon extension during development, different building blocks need to be synthesized and transported to axon terminals. Thus, proper coordination of signaling, metabolism, and gene expression is required. However, in later developmental stages, neurons need to switch to a “synapse-growth mode,” devoting metabolic and growth resources to synapse development and function. In RGCs, this switch occurs in the early postnatal period (Goldberg et al. 2002). For adult axon regeneration, it is necessary to reactivate the axon growth program in injured neurons. Understanding mechanisms that control axon growth programs has been an active frontier. Here we have listed a few examples.

Transcriptional Regulation. Neuronal injury induces extensive changes in gene expression programs. Of these, studies suggest correlative regeneration-associated genes (RAGs) and TFs that are causal determinants in axon regeneration. For example, in comparing axon growth-competent and -incompetent RGCs, Goldberg and his colleagues identified several members of the Krüppel-like factor (KLF) TF family as crucial gene regulators (Moore et al. 2009). In particular, KLF4 and KLF9 deletion or KLF6 and KLF7 overexpression can promote regeneration following injury during adulthood (Moore et al. 2009; Blackmore et al. 2012; Apara et al. 2017).

Newly developed RNA sequencing technology permits unbiased genome-wide comparison of gene expression in neurons under different conditions to predict putative master regulators. For example, profiling injured DRG neurons led to the identification of several putative master regulators including ATF3, SOX1, and Jun (Chandran et al. 2016; Renthal et al. 2020). Interestingly, injury-induced ATF3 expression returns to its resting level after their regenerating axons are reconnected with their targets (Renthal et al. 2020). Whereas ATF3 deletion prevents large-scale gene expression alterations in injured DRG neurons (Renthal et al. 2020), constitutive expression of ATF3 enhances regeneration in the peripheral branch of DRG neurons (Seijffers et al. 2007). STAT3 is another important pro-regenerative TF that is activated following peripheral, but not central, injury (Schwaiger et al. 2000). However, overexpression of STAT3 in DRG neurons can promote new growth and collateral sprouting of central axons (Bareyre et al. 2011). Importantly, as a key cytokine signaling pathway, the STAT3 pathway can also evoke regenerative responses of CNS neurons (Smith 2009; Sun et al. 2011; Jin et al. 2015).

In addition, Norsworthy et al. (2017), hypothesized that developmental axon growth programs might be encoded by certain TFs whose expression is diminished in the adult and showed SOX11 promotes axon regeneration after ONC. Independent studies showed that SOX11 expression also promotes the regrowth of corticospinal tract axons (Wang et al. 2015). Intriguingly, while SOX11 overexpression promotes regeneration of some RGC subtypes, it has been shown to cause cell death of other RGC populations (Norsworthy et al. 2017). Thus, there might be a functional cross talk between the programs for injury-induced neuronal death and axon regeneration.

Epigenetic Regulation. TFs impact the transcription of their target genes by binding to specific DNA sequence motifs across the genome. However, the chromatin accessibility around these binding sites is tightly regulated under conditions such as injury (Gaub et al. 2010; Finelli et al. 2013; Puttagunta et al. 2014; Palmisano et al. 2019). By using the assay of transposase-accessible chromatin using sequencing (ATAC-seq) to quantify genome-wide accessible chromatin (Buenrostro et al. 2013, 2015; Corces et al. 2017), many gene expression changes in injured and regenerating DRG neurons could be attributed to epigenetic regulation (Palmisano et al. 2019). This suggests an “epigenetic barrier” in which TF-binding sites may simply not be accessible in regeneration-incompetent neurons. Intriguingly, this “epigenetic barrier” to axon regeneration appears to preferentially silence many genes associated with axon growth in adulthood (Venkatesh et al. 2018).

DNA methylation alterations might be another avenue of epigenetic regulation (Fig. 3(4–6)). Age-related DNA methylation patterns can serve as a proxy for an epigenetic clock (Horvath 2013). Strikingly, this epigenetic clock can be reset by overexpressing Yamanaka TFs (OCT4, SOX2, KLF4 [OSK], and c-MYC) in somatic cells, which results in the loss of cellular identity and transforming them into pluripotent stem cells (Takahashi and Yamanaka 2006). Similarly, overexpression of c-MYC (Belin et al. 2015) or OSK (Lu et al. 2020) promotes axon regeneration in injured RGCs. Understanding why and how these epigenetic barriers form, and how to manipulate them, will be crucial to future efforts that seek to use pro-regenerative TFs with naturally inaccessible promoter or enhancer sites in mature CNS neurons.

Signaling Cascades. Several signaling cascades are activated in response to injury. To define key regulators, at least two approaches have been informative: (1) identifying the pathways required for axon growth in regeneration-competent neurons, and (2) defining the manipulations that enable axon regeneration in regeneration-incompetent neurons. As mentioned above, the DLK/MAPK pathway only functions in regeneration-competent neurons. In contrast, two other pathways have been documented to regulate axon regeneration in both conditions.

Phosphoinositide 3-Kinases (PI3Ks)/Mammalian Target of Rapamycin (mTOR). The PI3K/mTOR pathway is a central regulator of metabolism and cell growth (Saxton and Sabatini 2017), and has been implicated as an evolutionarily conserved regulator of axon regeneration. PTEN is a negative regulator of this pathway (Fig. 4). As aforementioned, PTEN deletion or inhibition is sufficient to promote regeneration of regeneration-incompetent neurons. Inhibiting mTOR abolishes the regeneration induced by PTEN deletion, suggesting the requirement of mTOR in this regenerative outcome. However, activating mTOR only, by genetic deletion of TSC2, led to less axon regeneration (Park et al. 2008), indicating that additional effectors, such as GSK3, might mediate the regeneration induced by PTEN inhibition. Additionally, mTOR is endogenously activated in regeneration-competent peripheral axons to enhance axon outgrowth following peripheral nerve injury (Abe et al. 2010). Thus, leveraging PI3K/mTOR regulated metabolism to develop manipulations that elevate PI3K/mTOR activity could be a useful approach to promote axon regeneration.

In fact, PI3K/mTOR is a major signaling pathway that mediates the neurotrophic effects of growth factors such as brain-derived neurotrophic factor (BDNF) or neurotrophin-3. However, despite the established roles of these growth factors in regulating neuronal survival and axon growth during development, direct administration of these factors has limited effects on promoting axon regeneration in adults (Giehl and Tetzlaff 1996; Lu et al. 2001; Hollis et al. 2009). This suggests that, in comparison to immature neurons, mature corticospinal neurons have reduced responsiveness to growth factors. Recent studies suggest that osteopontin, an integrin-binding protein, could sensitize neuronal responsiveness to insulin-like growth factor 1 (IGF1) and promote IGF1-induced axon regeneration, perhaps by facilitating IGF1-triggered IGF1 receptor clustering and activation (Duan et al. 2015; Bei et al. 2016; Liu et al. 2017). Thus, it would be interesting to develop other methods of sensitizing neuronal responses to neurotrophic factors for promoting axon regeneration.

Jak/Stat. The second pathway central to axon regeneration is JAK/STAT. Cytokines such as ciliary neurotrophic factor (CNTF) or interleukin 6 (IL-6), released by nonneuronal cells in response to injury, bind to their specific receptors to initiate signal transduction to the intracellular compartment by their shared receptor component gp130. Together this results in the activation of JAK kinases and STAT TFs (Heinrich et al. 1998). Early studies documented activation of this pathway in injured DRG neurons (Cafferty et al. 2004). The causal relationship of this pathway to axon regeneration was supported by the observation that deletion of SOCS3, a negative regulator of the JAK/STAT pathway, following ONC promotes axon regeneration. This process is further enhanced by additional treatment with CNTF (Sun et al. 2011). Moreover, the JAK/STAT pathway is thought to mediate, at least partially, the effects of inflammation-triggered axon regeneration (Leibinger et al. 2013). Manipulating this pathway to promote axon regeneration has been an active area of neural repair research (Leibinger et al. 2021).

Metabolic Switch. During early development, growing neurons require anabolic metabolism, using ATPs for synthesizing building blocks for neuronal growth. Upon completion of neural development, mature neurons convert to homeostatic metabolism. To support sustained axon regeneration, it is conceivable that injured neurons may need to reestablish anabolic metabolism. In fact, some regeneration-promoting molecules, such as mTOR and c-MYC, are known master regulators of anabolic metabolism (Stine et al. 2015; Saxton and Sabatini 2017). In light of active research of metabolic regulation in nonneuronal cells, it will be interesting to identify other novel avenues of manipulating cellular metabolism for axon regeneration.

The Environmental Decision: Whether or Not to Form an Inhibitory Scar

The initial traumatic injury immediately disrupts the axon, surrounding glial cells, and blood vessels (Fig. 3(7)). Depending on species, location, and developmental stage, subsequent injury responses lead to varied outcomes. In nonmammalian vertebrates, such as newt and zebrafish, a permissive bridge forms after SCI allowing regenerating axons to cross the lesion (Zukor et al. 2013; Mokalled et al. 2016; Cigliola et al. 2020). In mammals, a glial or fibrotic scar develops after a lesion to the CNS, but not the PNS. Understanding the mechanisms mediating such distinct responses has been intensely investigated for many years but many mysteries still remain (Sofroniew 2018; Tran et al. 2018).

In all types of injury, the initial insult leads to breakdown of the blood–brain barrier (BBB) and the infiltration of blood-derived immune cells. Simultaneously, nonneuronal residential cells, in particular astrocytes and microglia, are activated and participate in the formation of glial and fibrotic scar in the adult CNS. Such fibrotic scars appear to be built from stromal cells, but the precise origin is still under debate. While some studies suggest that the fibrotic scar is mainly from pericytes associated with blood vessels (Göritz et al. 2011; Dias et al. 2018), others support the predominant contribution of proliferative CNS fibroblasts (Dorrier et al. 2021).

Another major focus is whether inflammation is beneficial or detrimental. Depletion of macrophages has been shown to improve regenerative outcomes following SCI (Popovich et al. 1999), suggesting that macrophages contribute to the inhibitory environment of the CNS. Although inflammation following CNS injury certainly has deleterious effects, the immune response has also been reported to have beneficial effects on axon regeneration (Benowitz and Popovich 2011; Schwartz and Raposo 2014). A recent study shows the multifaceted ways in which microglia act in the process of scar-free wound healing after SCI in neonatal mice (Li et al. 2020). After neonatal and adult injury, microglia become activated; however, while their activation is sustained in adults, after neonatal injury microglia reestablish homeostasis within a few days. These transiently activated microglia express not only fibronectin and other ECM molecules to form bridges reconnecting the spinal cord but also proteinase inhibitors to facilitate inflammation resolution (Li et al. 2020). Importantly the down-regulation of these molecules in reestablished microglia prevents the formation of a fibrotic scar after neonatal injury. Similarly, transient induction of connective tissue growth factor a (CTGFa) is important for initial bridging events after SCI in zebrafish, but its persistent expression might contribute to the scar formation in the adult lesions (Mokalled et al. 2016). Thus, such tight temporal control of microglial responses might be crucial for efficient wound healing. Modulating microglia and perhaps other inflammatory cells represent a potential strategy to temper neuroinflammation and enhance wound healing and axon regeneration.

Astrocytes are resident cells unique to CNS. As an essential part of the BBB, one of their functions is the communication between the blood and the brain. Upon injury, astrocytes undergo rapid reactive gliosis with massive cellular and molecular changes. This drastic change allows astrocytes to isolate the injury from surrounding neural tissue and restore the BBB (Burda and Sofroniew 2014). However, later, these reactive astrocytes may become detrimental to regeneration. Early studies emphasized the inhibitory nature of these cells, largely due to their expression of CSPGs and other types of growth-inhibiting ECMs (Silver and Miller 2004). As a component of the glial scar, some reactive astrocytes are tightly interwoven with other nonneuronal cells (fibroblasts or pericytes) within the gradually accumulating and disorganized ECM. This alters the molecular and physical properties of local tissues and causes the formation of physical and biochemical barriers to axon growth (Hara et al. 2017). On the other hand, mounting evidence suggests that some reactive astrocytes may be supportive for axon regeneration via the secretion of growth factors (Anderson et al. 2016, 2018) and the formation of elongated cell processes that act as a physical scaffold for such axons (Wanner et al. 2013). It is interesting to note that most regenerating axons induced by PTEN deletion cross the lesion in association with astrocyte bridges (Zukor et al. 2013), reminiscent of the observations that immature astrocytes are permissive for axon growth (White and Jakeman 2008; Filous et al. 2010). Thus, like microglia, astrocytes might exhibit different functional states (i.e., naive, reactive, and scar-forming), thereby dynamically impacting axon regeneration and tissue reconstitution after injury. Similar to wound-healing responses in nonneural tissues, reactive astrocytes may undergo epithelial-to-mesenchymal transitions (Vivinetto et al. 2020). Further understanding cellular and molecular mechanisms that control the switch between these different states might offer new insights into developing methods that encourage a more permissive environment within the lesion.

Final Decision: To Regenerate or Not

After successful growth cone formation, there are at least two requirements for a regeneration-competent axon to sustain extension: (1) supply and transport of cellular building blocks, and (2) their assembly into new axonal segments. Accordingly, proper cell body responses, axonal transport, and membrane insertion are required.

Building Blocks and Energy for Regeneration: Axonal Transport and Translation

Unlike developing neurons, which have relatively short axons in embryos, axotomy in the adult often occurs in the distant, distal axonal segments. The required building blocks and energy necessary for regeneration must therefore be supplied to newly forming growth cones via local recycling (e.g., of actin and microtubules), local protein synthesis, and axonal transport. Considering that the diffusion of proteins requires ∼1 week to travel a length of 1 mm (Futerman and Banker 1996), axons largely rely on active transport along microtubule “highways” to both receive retrograde signals and send building blocks to an axon tip. These “cargoes” move along axons, using kinesins for anterograde transport and dyneins for retrograde transport.

Like proteins, diffusion of ATP from the soma to the regenerating tip does not happen at a rate sufficient to support the energy requirements of new outgrowth. Because most of a cell's energy is generated by mitochondria, it is essential that mitochondria are located near the areas of high energy demand. Microtubule-based transport of mitochondria to a growing axon occurs via a conserved motor/adaptor complex consisting of the mitochondrial RhoGTPases Miro1/2 and TRAK1/2 (Schwarz 2013). Increasing mitochondrial motility may promote axon regeneration (Cartoni et al. 2016; Zhou et al. 2016). Whereas such molecular machinery and regulation have been extensively characterized (Hirokawa and Takemura 2005), it is unknown whether it is possible to speed up such systems for more efficient axon regeneration. Local protein synthesis via axonal translation appears to be an evolved solution (Sahoo et al. 2018), but it remains unknown whether nonregenerative adult mammalian axons undergo axonal translation.

Membrane Insertion

Once the building blocks are in axons, they need to be properly assembled for axon extension. It is hypothesized that materials are added to the tip of extending axons during development via plasmalemmal precursor vesicles. However, little is known about membrane transport and insertion during stretch growth in postnatal development or postinjury axon regeneration (Pfenninger 2009). Economically speaking, adding new membranes to the proximal end of networked axons, followed by lateral diffusion, reduces the need for long-distance transport. On the other hand, membranes could be inserted at multiple sites along the entire length of an axon. The smooth ER (SER) is distributed throughout axons as a continuous tubular network, raising the possibility that phospholipids may be synthesized locally in the axonal SER (Renvoisé and Blackstone 2010). Intriguingly, overexpression of protrudin, an integral ER membrane protein, promotes axon regeneration through the mobilization of endosomes and ER into the distal part of injured axons (Petrova et al. 2020). Thus, it will be interesting to develop other manipulations that could increase the axonal supply of growth molecules and organelles to enhance axon regeneration.

PERSPECTIVE

Despite tremendous progress made in the past decades, there is still no effective treatment of promoting axon regeneration in clinical settings. This is due to a few issues. First, the adult mammalian CNS exhibits limited axonal regeneration, both in length and number, even when using the best combinations of treatments. It is still a mystery why some, but not other, types of neurons remain refractory to regenerate under these manipulations. The application of new technologies like single-cell -omics is promising to tease out the underlying mechanisms (Belin et al. 2015; Bray et al. 2019; Tran et al. 2019; Poplawski et al. 2020). Additionally, we do not understand why mammalian axon regeneration is so slow. This might be the result of certain molecular and cellular limitations of adult neurons. During development, axon growth has two stages: relatively short distance growth cone-guided extension in early embryogenesis, followed by “stretch growth” exerted by increased distance between the neuronal soma and tethered axon terminus during later development (Smith 2009). However, axon regeneration is completely growth cone dependent. It is unclear whether the cellular and molecular machinery of adult neurons can support the lengths and speed of axon extension needed for regeneration. Considering growth cone–free stretch growth, it would be interesting to explore the possibilities of speeding up axon regeneration with mechanistic-based manipulations (de Vincentiis et al. 2020; Franze 2020). Second, regenerating axons may not automatically translate into functional recovery. At least in injured optic nerves, most, if not all, regenerated axons are not myelinated, representing an additional hurdle for functional recovery (Bei et al. 2016; Wang et al. 2020). Further, regenerating axons may not be able to integrate into the functional circuits in injured spinal cord, because of injury-induced, circuit-level alterations such as excitation–inhibition imbalance (Edgerton et al. 2001; Courtine et al. 2009; Rossignol and Frigon 2011). In this regard, different manipulations have been developed to adjust the functional states of injuried spinal cords by electrical stimulation (Harkema et al. 2011; Gerasimenko et al. 2015; Gill et al. 2020) and other forms of neuromodulation (Chen et al. 2018; Brommer et al. 2021). Thus, ultimately combining these complementary strategies to deal with both hardware (anatomical connections) and software (neuromodulation) issues might be able to achieve effective functional restoration after injury and diseases.