Wallerian degeneration as a therapeutic target in traumatic brain injury

Abstract

Purpose of review

Diffuse or traumatic axonal injury is one of the principal pathologies encountered in traumatic brain injury (TBI) and the resulting axonal loss, disconnection, and brain atrophy contribute significantly to clinical morbidity and disability. The seminal discovery of the slow Wallerian degeneration (WD) mice (Wld^s) in which transected axons do not degenerate but survive and function independently for weeks has transformed concepts on axonal biology and raised hopes that axonopathies may be amenable to specific therapeutic interventions. Here we review mechanisms of axonal degeneration and also describe how these mechanisms may inform biological therapies of traumatic axonopathy in the context of TBI.

Recent findings

In the last decade, SARM1 and the DLK/LZK MAPK cascade have been established as the key drivers of WD, a complex program of axonal self-destruction which is activated by a wide range of injurious insults, including insults that otherwise leave axons structurally robust and potentially salvageable. Detailed studies on animal models and postmortem human brains indicate that this type of partial disruption is the main initial pathology in traumatic axonopathy. At the same time, the molecular dissection of WD has revealed that the decision that commits axons to degeneration is temporally separated from the time of injury, a window that allows potentially effective pharmacological interventions.

Summary

Molecular signals initiating and triggering WD appear to be playing an important role in traumatic axonopathy and recent advances in understanding their nature and significance is opening up new therapeutic opportunities for TBI.

Introduction: Essentials of axon biology

The extensive network of nerves traversing the human body evident to anyone who has performed gross dissections underlines the important fact that nerve cells are, in essence, their axons: the majority of the cytoplasm of projection neurons is in their axonal processes. Two characteristic examples are the human sensory or motor neuron where a 50–100 micron-diameter soma is attached to an axon that runs for one meter or more, and the human basal forebrain cholinergic neuron with a 30-micron soma and a cumulative axon length that may reach an estimated 100 meters with all its branches [1]. The substantial axonal outlay of the nerve cell made axotomy, a non-lethal perturbation away from the nucleus, an extremely popular method for studying neuronal responses to injury [2, 3]. Of course, the axon is much more than a cytoplasmic process, however massive. The remarkable geometry of the projection neuron introduces a number of biological challenges, not the least of which is the need to ship organelles and key metabolites or neurotransmission-related molecules to distal axons and terminals for sustenance and signaling. The axon is also a conduit for target-derived trophic signals that are retrogradely transported to secure and maintain neuronal survival and phenotype[4–6]. The interdependence between soma and the axon/terminal is best illustrated in the pathological patterns seen after axotomy. The soma undergoes more or less stereotypical changes including dissolution of the Nissl substance, reduction in the expression of neurotransmitter-related enzymes, aberrant phosphorylation of neurofilaments [7] and, in some cases of proximal axotomy, cell death [8]. Meanwhile, the distal axon undergoes an orderly type of fragmentation into ovoid formations via both autonomous and glial-derived mechanisms that are best described in the PNS and are known as Wallerian degeneration (WD) [9–11].

The previous classical concepts underlying the dependency of the soma on the axon and vice versa were cast aside in the early 90s with the discovery of Ola mice characterized by a remarkably slow WD [12, 13]. In this phenotype, separated axons can survive for weeks while maintaining the ability to transmit action potentials [14]. This phenomenon is the result of Wld^s, a protein that is encoded by a spontaneous gain-of-function mutation and is the result of an in-frame fusion of the complete sequence of nicotinamide mononucleotide adenylyltransferase 1 (Nmnat1, a NAD+ synthesizing enzyme) at the C-terminus, a 70 amino-acid sequence from the N-terminus of Ube4b (an E4-type ubiquitin ligase) at the N-terminus, and a linker between the two comprised of a unique 18 amino acid sequence from a read-through of Nmnat1 5′ UTR [15, 16]. NMNAT activity in Wld^s is thought to substitute for the axonal isoform, NMNAT2, which is depleted in the axon after transection [17, 18]. The suppression of WD by a particular genetic arrangement led to the surprising realization that axonal degeneration is not simply a passive response to deprivation of key support from the cell body (Fig.1). The notion of an autonomous mechanism of axonal degeneration was further reinforced by the fact that Wld^S appears to have no effect on neuronal cell body in classical models of apoptosis, i.e. NGF-deprivation of sympathetic neurons or the neonatal motor neuron axotomy model [19, 20]. Conversely, well-established and effective anti-apoptotic strategies such as Bak or Bax deletion and Bcl-2 overexpression that protect the neuronal cell body do not confer axonal protection [21].

Figure 1.

Evolution of concepts on axonal biology over time. Top, the axon as a (massive) extension of the neuronal cytoplasm. Middle, the axon as a metabolic and trophic conduit. Bottom, the axon (and perhaps the terminal) as independent biological domains. Red profiles represent glial support.

The current state of cell biology of WD

Subsequent to the discovery of the Wld^s gene and its protective role in axotomized peripheral axons, more models of axonal degeneration were found to be responsive to the mutation or to its biochemical surrogate, i.e. excess NMNAT in the axon, leading to a molecular signature that is progressively viewed as pathognomonic of WD [10, 15]. In addition to the well-established protection of axons in peripheral nerve injury, the expression of Wld^s or NMNATs appears to preserve axons in taxol neuropathy[22], glaucoma[23, 24], excitotoxic injury[25], ischemia[24, 26], and the 6-hydroxydopamine model of Parkinson’s disease[27, 28]. These findings, on top of similar observations made in simpler injury models in the fly[29, 30] helped shape the prevailing notion that a) WD is a genetically instructed, evolutionarily conserved program of axonal demise akin to apoptosis for the cell body, and b) that this program may be activated in diverse pathways of injury including conditions that do not immediately destroy the structural continuity of the axon, yet trigger axonal degeneration. As we will discuss in the next section, axonopathy in traumatic brain injury (TBI) belongs to this set of conditions with partial insults but is featured by clearly defined time windows that offer therapeutic opportunities.

The molecular landscape of WD has been reshaped in recent years to include signals that clarify and extent observations made in Wld^s-based models. It is now generally accepted that a required instructive signal is the activation of Sterile alpha and TIR motif containing 1 (SARM1) that degrades NAD⁺, a promiscuous redox cofactor essential for axonal maintenance that is generated de novo from tryptophan or salvaged from the pool of its metabolites via the action of NMNAT [31–34]. The exact mechanism and timing of SARM1 activation is not well understood. A key initiating event may be NMNAT2 degradation via the activation of the stress MAPK pathway [35] or an atypical SCF E3 ligase complex [35, 36], and this is perfectly consistent with prior work on the protective role of Wld^s. However, the depletion of NMNAT2 does not only lead to loss of its product, NAD+ [17], but also to a buildup of its substrate NMN which may play a role in SARM1 activation [37–39]. The latter might explain why deletion of SARM1 is more potent in protecting axotomized axons than Wld^s [40]. The digestion and depletion of axonal NAD⁺ and ATP [41] may suffice to cause axonal degeneration while leaving the soma intact [42], although some of the enzymatic products of SARM1 activation such as ADPR and cADPR might also play a role in degeneration [31, 43]. Similar to the Wld^s mutation, SARM1 deletion or interference is protective not only in peripheral nerve axotomy but also in models of peripheral neuropathy [42, 44–46], and in more generic types of injury such as mitochondrial dysfunction/oxidative stress [47] and excitotoxicity [48](Fig. 2). Although NMNAT proteins can also function as chaperons in a fashion unrelated to their enzymatic activity and, as such, they may protect against proteotoxic insults [49], this interesting mechanism is not well understood in the context of injury.

Figure 2.

Key pathways involved in axonal degeneration after injury. This is a simplified diagram of molecular processes associated with Wallerian degeneration. Yellow: initiating events related to passive or active reduction in NAD+ levels in axons, via depletion or degradation. Red: “execution” events dependent on activation of SARM1. Blue: auxiliary or modifying signals related to the activation of the MAPK pathway, which also signals retrograde perikaryal responses. There may be mutual reinforcement of signals between SARM1 and MAPK cascades.

Besides NMNAT and SARM1, the stress MAPK cascade, which has an established role in retrograde somatic responses to injury, may also participate in WD (Fig 2). A key player here is dual leucine zipper kinase (DLK, MAP3K12) along with the highly related leucine zipper kinase (LZK, MAP3K13). DLK is a ubiquitous and highly conserved retrograde injury signal [50, 51] and remains elevated in the nerve cell due to feedback phosphorylation and suppression of ubiquitination by its downstream product JNK [52]. DLK, MKK4/7 and JNKs combine with the scaffolding protein JIP3 in signaling complexes that are retrogradely transported in vesicles [53, 54]. Via this retrograde mechanism DLK appears to regulate axotomy- or NGF deprivation-induced neuronal death [55–58]. Dlk knockout alone can rescue axotomized neurons [57, 58], but the combined knockout with Lzk may be substantially more effective [57]. Although the retrograde effects of MAP3K on the cell body are clearly established, less is known about the effects of these kinases on axons. One axonal effect may be secondary to transcriptional processes via signals such as the upregulation of PUMA which triggers an anterograde axonal degeneration program [59]. A more direct mechanism, however, is JNK1-mediated phosphorylation of SARM1 which reinforces its NADase activity[60], as well as DLK/LZK-mediated degradation of the palmitoylated NMNAT2 and of the axon-protective molecule SCG10 [35](Fig. 2). Of note, injury-induced DLK/LZK activation may also serve to promote axonal regeneration [50, 51], a condition that stresses the complexity of the role of stress MAPK signaling and the need to be cautious with respect to potential molecular targeting for therapeutic purposes in the future.

The key role of the axon and its pathology in TBI

Traumatic brain injury is a calamity with 2–3 million new emergency room visits a year in the United States alone [61]. Main neuropathologies associated with TBI are focal contusions [62] and diffuse axonal injury (DAI). Diffuse axonal injury is a primary axonal pathology initiated by inertial loading of axons, usually as a result of rotational acceleration of the head. Although it is classically encountered in motor vehicle crashes and high-impact falls, it may occur across TBI causes and degrees of severity including single and repeat concussions and blast injuries [63–69]. In severe cases, DAI leads to multifocal axonal disruption associated with micro-hemorrhages or separation of gray with white matter in the form of gliding contusions, and secondary disconnection and brain atrophy (Fig. 3)[70, 71]. Because of the preferential involvement of associative and commissural tracts DAI can cause pervasive cognitive, behavioral, and emotional disorders or complex motor and sensory deficits. Gyrencephalic brains including the human brain may be especially prone to this type of pathology, in part because of the sheer bulk of white matter but perhaps also the three-dimensional organization of intertwined white matter tracts [63–65, 67, 68](Fig. 4). However, the elemental pathology of DAI, i.e. primary traumatic axonal injury (TAI), can be largely reproduced in the lissencephalic rodent brain [66, 72–74].

Figure 3.

MRI images showing progressive degenerative changes in the brain of this 57 year-old man 20 years after motor vehicle crash with diffuse axonal injury (A-B, arrows), including gliding contusions (A). Note the marked atrophy of the corpus callosum, resulting from massive axonal degeneration (C, arrows).

Figure 4.

Compared to gray matter, the white matter preferentially increases in the mammalian brain with increasing complexity, primarily in associative and commissural tracts (A). For example, only 8% of the mouse brain is occupied by white matter whereas the corresponding figure for the human brain is 50%. The relative increase (Δ) in white compared to gray matter is expressed by the simplified equation ΔW=ΔG^4/3. The real exponent may be slightly smaller than 4/3 (Discussed by Zhang K & Sejnowski TJ (2000). A universal scaling law between gray matter and white matter of cerebral cortex. Proc Natl Acad Sci U S A. 2000 May 9;97(10):5621–6). What is more, white matter in complex brains is organized in tracts intertwined with each other (B), such that biomechanical acceleration forces may become significantly enhanced at the points of crossing. Images in (A) taken from neurosciencelibrary.org of the University of Wisconsin and Michigan State Comparative Mammalian Brain Collections, and the National Museum of Health and Medicine. Image in (B) is courtesy of Dr. E. Fieremans from Validation methods for diffusion weighted magnetic resonance imaging in brain white matter. Department of Electronics and Information Systems, Ghent University. Ghent, Belgium. 2008.

Traumatic axonal injury is a primary axonal insult [72, 75–77] and is featured by structural compromise of the axon and metabolic /neurochemical impairments such as inappropriate ionic flux and depletion of energy stores [78, 79]. Early structural events may include fracture of microtubules or mechanoporation of the axolemma eventually resulting in excess calcium influx/propagation and the activation of cysteine proteases that degrade the cytoskeleton [80–83]. Some of these phenomena are the direct effect of the biomechanical force to which the axon submits more or less passively, yet other events involve active signaling through complex molecular programs, including signals initiating and triggering WD (see below). Select processes (e.g. cytoskeletal changes, transportation deficits and the progressive formation of axonal swellings/spheroids) resemble changes encountered in axotomy [84, 85] or models of demyelination [86–88]. Both myelinated and unmyelinated axons are involved in TAI and, in the case of myelinated axons, pathology seems to begin at the node of Ranvier [89–91]. With notable exceptions [89–91], the focus of research in TAI has been the axons themselves. The role of the myelin sheath is poorly understood, but this is the case of many diseases of the CNS [9].

Wallerian degeneration in traumatic axonopathy

As alluded in the previous section, TAI ay lead to progressive degenerative phenomena over the course of hours or days, a process that we have termed traumatic axonopathy [74, 92]. Proceeding via a number of pathobiological processes reviewed in the previous section, traumatic axonopathy is featured by characteristic dystrophic changes such as axonal varicosities and spheroids. These hallmark pathologies develop within minutes and can last for hours and days, especially in larger gyrencephalic brains [72, 93]. In many cases, these dystrophic changes progress to secondary axotomy with the formation of end bulbs. Based on the PNS axotomy paradigm, one would be tempted to think that this the point of trigger of WD in the detached distal axon segment. However, recent work from our group suggests that even axonal varicosities and spheroids are suppressed by deleting SARM1 [74], a result suggesting that these “earlier” lesions may be manifestations of WD in the course of traumatic axonopathy. Although, in our recent work, this was demonstrated with the combination of genetic methods and high-resolution neuroanatomy[74], there is some prior evidence that spheroids and other degenerative phenomena occurring even acutely after injury may share mechanisms with or represent manifestations of WD [85, 94].

High-resolution anatomical tools and careful neuropathological analysis offer new opportunities for examining the problem of WD in the course of traumatic axonopathy. For example, in two well-characterized CNS tracts that undergo primary TAI after diffuse TBI (impact acceleration), there are well-formed injury fronts of dystrophic axons and spheroids. In the case of the mouse optic nerve such an axonal disruption front is between the orbital apex and the optic chiasm [92]. In the case of the corticospinal tract (CST), this front is located at the great crossings of the medulla, i.e. the crossing of CST with reticulospinal tract and the decussation of the CST itself. Although axons in these experimentally reproducible fronts and axon segments distal to them show classical spheroid swellings and end-bulbs (Fig.5A), there are also ovoid formations with prominent dysmyelination typical of WD more evident on semithin sections or at the ultrastructural level [74, 92] (Fig 5B–C). In the case of CST, there is also distal axonal fragmentation and denervation of targets in the spinal cord (Ziogas and Koliatsos, personal observations).

Figure 5.

Characteristic lesions encountered in the corticospinal (A,C) and visual (B) systems after impact acceleration TBI. (A) is an image from a CLARITY preparation through the brain stem of a YFP-H mouse. (B) is an image of a hematoxylin & eosin-stained semithin preparation through the distal optic nerve. (C) is an electron microscopic image through the corticospinal tract. Although most lesions in (A) are spheroid swellings and end bulbs, most lesions in (B) have a myelin ovoid configuration and so do lesions in (C). In (C), a1 is a normal axon, a2 is an abnormal axon and arrows indicate a myelin septum.

Therapeutic opportunities

In contrast to the all-or-none nature of experimental axotomy, the biomechanical disruption of the axon in TAI leads to varying degrees of damage, from complete transection and rapid degeneration to no injury at all. Perturbed axons in the middle may either degenerate or heal over a period of time [75, 93, 95] (Fig. 6). As we have shown with impact acceleration injury in the visual system, although a majority of RGCs are injured based on the presence of DLK-dependent induction of c-JUN in the nucleus, only a third of injured retinal ganglion cells and their axons eventually degenerate[96]. In the optic nerve stretch model, degeneration of retinal ganglion cells occurs much later than would be typical for primary transection [97], an observation indicating the presence of a period of time during which the injured axon is going through a molecular decision process. In other models of TBI, late axonal degeneration seems to occur at 4–12 hours, and for reasons that are not yet understood, is further delayed in humans [72, 98].

Figure 6.

The variable/selective nature of TAI. In contrast to conventional axotomy where all axons are immediately transected, the biomechanical forces of TBI have variable effects on axons, from no effect (1) to transection and death (3), with various types and degrees of injury (2–2´) in the middle. Injured axons may proceed to death (as in 3) or repair (back to 1). The degenerative course may be prepared and triggered by molecular signals associated with WD. Many injured axons may be salvageable if we can turn off “suicide” WD signals and allow repair to take its course.

The delay in the commitment of the injured axon to WD allows a window of opportunity for therapeutic interventions and there is emerging support of this idea from pharmacological studies in vitro. For example, cleavage of a protease-sensitized SARM1 construct up to 2 hours following axotomy fully suppresses axonal degeneration [41]. Although there are still no available SARM1 inhibitors, it has been suggested that the enzyme can be indirectly inhibited by suppressing the action of NAMPT and hence the built up of NMN [37–39]. Delayed treatment of in vitro transected axons with the NAMPT inhibitor FK866 up to 4h hours post-injury also suppresses WD [99]. Similar observations have been made by targeting the MAPK cascade which is activated within minutes after injury [54]. Pretreatment with DLK/LZK inhibitors, seems to delay WD in cultured primary neurons for at least 30h following axotomy, an effect that could be explained by an increase in the axonal levels of NMNAT2 and prevention of its depletion [35]. Others have shown that whereas pre-treatment with JNK inhibitors does not delay WD [55], treatment at the time of injury or within an hour is able to suppress it [54, 55]. We have recently confirmed some of these observations using microfluidic axotomy platforms [100].

Based on this discussion, we propose that partially damaged, salvageable axons that represent the majority of the injured axons in human and animal TBI may be restored to normal functionality if key WD signals such as SARM1 and select MAPKs are promptly suppressed. In the absence of an ongoing degenerative drive, such timely anti-WD interventions should have long-term benefits for axons and neural systems. This prospect has important therapeutic implications because of the availability of small molecules that can inhibit MAPKs and, indirectly, SARM1. Some of these compounds have been already tested or are FDA-approved for other indications and seem to be effective in models of TAI in vitro and in vivo [74, 100].

Conclusions

In summary, WD represents a complex genetic program of axonal breakdown which, when inhibited, allows transected axons to survive independently and transmit action potentials for weeks after injury. Although in the case of axonal transection models targeting WD and preserving “zombie” axons would be of no translational benefit, it is also true that WD-like processes are present in more complex diseases and may contribute to the eventual demise of axons that are partially injured and are, therefore, salvageable. This type of incomplete injury is typical of traumatic axonopathy and therefore it is an attractive therapeutic target. The same concept may apply to types of axonal injury other than mechanical trauma, for example axonopathies encountered in hypoxia-ischemia and other white matter disorders.

Key Summary Points.

• Wallerian Degeneration (WD) is an axonal program of self-destruction that can be activated by a variety of injurious stimuli, including traumatic axonal injury (TAI), a very common pathology associated with rotational acceleration in the course of TBI.

• WD is driven by conserved pathways involving SARM1 with a possible participation of the DLK/LZK MAPK cascade, both activated by injury and resulting in axon dissolution with hours or days.

• The molecular decision that commits axons to WD is temporally separated from the time of injury, thereby allowing a therapeutic window for therapeutic intervention

• Pharmacological inhibition of WD may protect partially injured, salvageable axons in the context of TAI and potentially prevent disconnection and brain atrophy.