Neuronal ER Stress in Axon Injury and Neurodegeneration

Abstract

Injuries to CNS axons result not only in Wallerian degeneration of the axon distal to the injury, but also in death or atrophy of the axotomized neurons, depending on injury location and neuron type. No method of permanently avoiding these changes has been found, despite extensive knowledge concerning mechanisms of secondary neuronal injury. The autonomous endoplasmic reticulum (ER) stress pathway in neurons has recently been implicated in retrograde neuronal degeneration. In addition to the emerging role of ER morphology in axon maintenance, we propose that ER stress is a common neuronal response to disturbances in axon integrity and a general mechanism for neurodegeneration. Thus manipulation of the ER stress pathway could have important therapeutic implications for neuroprotection.

Introduction

Injuries of central nervous system (CNS) axons often result in permanent loss of vital functions due to axon degeneration and retrograde neuronal cell atrophy or even death. Preventing neurodegeneration is, therefore, critical for minimizing the severe consequences of CNS injuries and preserving neuronal function. Although neuroprotectants have long been sought, none has been found either for acute neural injuries, such as stroke, traumatic brain injury (TBI) and spinal cord injury (SCI), or for chronic neurodegenerative diseases such as Alzheimer's disease (AD), Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), multiple sclerosis (MS) and glaucoma^{1, 2}. The significant unmet clinical need for neuroprotectants is due to the lack of understanding of the key upstream signals that trigger the apoptotic cascade in injured neurons. Deciphering the mechanisms responsible for the retrograde death of axotomized and chronically degenerating neurons would allow us to identify molecular targets for the development of innovative and efficient neuroprotective treatments. Moreover, axonal degeneration is now understood to be an active process with a complex metabolic basis. This understanding opens the possibility of rescuing axons that have been injured either by trauma or diseases. The present review summarizes evidence that a key element in the response of neurons to injury of their axons is activation of neuronal endoplasmic reticulum (ER) stress. ER stress is a complex cascade of reactions that are normally activated when the ER, the organelle responsible for protein synthesis and proper folding, is overwhelmed by unfolded and misfolded proteins, a process that is called the unfolded protein response (UPR). We also consider the possibility that ER stress is initiated within the axon itself and thus could provide a target for axon protection after mechanical or metabolic insults.

Mechanisms of axotomy-induced neurodegeneration

Several hypotheses have been suggested to account for neurodegeneration after axon injury. Proposed mechanisms include: deprivation of retrogradely transported, target-derived neurotrophins; toxic influx of calcium ions through damaged axon membranes; and loss of synaptic connectivity and neuronal activity necessary for survival³. Excitotoxicity, oxidative stress and dysfunctional neuron-glia interactions may also contribute to neuronal cell death⁴. The available evidence, however, only partially supports any of these hypotheses. Because axon injury often is the initial pathology in both acute and chronic neurodegenerative diseases, and because axon degeneration often precedes neuronal cell body loss^5-8, understanding the detrimental signals induced by axotomy is essential for effective neuroprotection.

Responses to axotomy differ among neuronal types. For example, cortico-spinal neurons, may atrophy but the vast majority survive after transection in the spinal cord^{9, 10}, whereas most retinal ganglion cells (RGCs) die after optic nerve injury, even despite temporary rescue by delivery of exogenous trophic factors¹¹. Several features of RGCs make them a particularly useful system for investigating the mechanisms responsible for neuronal death after axotomy. The optic nerve consists of unidirectionally projecting axons sent exclusively from RGCs. The unequivocal separation of optic nerve from RGC perikarya greatly simplifies interpretation of the specific responses of the isolated neuronal cell body to injury of their axons. Interestingly, the severity and time course of RGC death are directly correlated with the distance between the axon lesion and the neuronal perikaryon: the farther the lesion from the cell body, the fewer and more slowly the RGCs die^{12, 13}. This correlation may explain why sectioning axons of the cortico-spinal tract (CST) in the spinal cord (far away from neuronal soma) does not induce significant short-latency cortical motoneuron death. Traditionally, this greater vulnerability of neurons to proximal axotomy has been attributed to dependence for survival on target-derived trophic factors provided by sustaining collaterals located along the axon. However, since there are no collateral branches along the course of the optic nerve, deprivation of target-derived neurotrophins cannot explain the increased vulnerability of RGCs to injuries close to their perikarya. An alternative hypothesis is that apoptosis is induced by signals derived from the tips of injured axons and transported retrogradely to the RGC perikaryon. The report that, after mild optic nerve crush, recovery of axonal transport induces additional RGC death, is consistent with the notion that an axon-derived death signal may be transferred retrogradely to the soma¹⁴. The report of reticulospinal neuron death in lampreys with a delay of 12 weeks or more after SCI¹⁵ raises the possibility that a very delayed form of corticospinal neuron death also might occur after SCI.

Recent studies from our laboratories have revealed that both acute and chronic optic nerve injuries induce ER stress in RCGs¹⁶ and that manipulating ER stress molecules exerts striking RGC-protection effects¹⁶. These findings support the emerging theme that axon injuries induce neuronal ER stress and that unresolved ER stress initiates neuron death. Consistent with this idea, spinal motoneurons showed signs of ER stress after axotomy^{17, 18} or ischemic injury^{19, 20}. ER stress markers also have been detected in cortex after stroke^{21, 22} and in motoneurons in a mouse model of ALS²³. Moreover, modulation of ER stress protected cortical neurons in models of ischemia and TBI²⁴, decreased spinal motoneuron death^{17, 23} and improved functional recovery after SCI^25-28. ER stress has been reviewed extensively in relation to chronic neurodegenerative diseases^29-33. It has, for example, been suggested to be a critical mechanism for neurodegeneration in several hereditary motor neuron diseases, in which ER stress is precipitated by accumulation of the mutant ER-resident protein seipin within moroneurons³⁴. Most aggregated proteins that are associated with chronic neurodegenerative diseases, however, are not confined to the ER, and whether ER stress is the cause or the consequence of the abnormal protein aggregation in the CNS remains unknown. ER stress has been documented in oligodendrocytes in demyelinating diseases³⁵, thus it may contribute to neurodegeneration indirectly. Recent studies have demonstrated that oligodendrocytes play a fundamental metabolic role in axon survival³⁶, and implicated oligodendrocyte dysfunction in the pathogenesis of ALS³⁷.

The present review focuses on the axon injury-induced neuronal ER stress and explores its critical role in neurodegeneration. Axon damage is an early manifestation of many neurodegenerative diseases and neuronal cell body loss can be secondary to primary axon damage^{1, 2, 38}. Thus the ER stress detected in neurodegenerative diseases may be due to axon pathology rather than protein aggregation, and unresolved ER stress may be responsible for progressive neuronal death. Studies of experimental glaucoma have also shown that ER stress molecules are increased in glaucomatous RGCs^{16, 39, 40} and that appropriate manipulations of ER stress molecules can prevent their degeneration¹⁶. We will focus on the discrete, sequential responses of neurons as they progress from neuronal subcompartmental damage (axon injury) to cell body apoptosis. We consider the cellular and molecular evidence that the axonal ER network senses harmful signals from axon injuries, and propose that neuronal ER stress is the common neuronal response to disturbances in axon integrity and that it is a general underlying mechanism for neurodegeneration. We also consider the potential therapeutic strategies targeting neuronal ER stress. Finally, we highlight the important questions that remain to be answered and conclude with a working model of neuronal ER stress initiation and propagation.

ER stress and the contradictory effects of UPR pathways on cell fate (Fig. 1)

Figure 1. The three UPR signaling pathways elicit both anti- and pro-apoptotic effects.

The release of BiP inhibition due to accumulation of unfolded or misfolded proteins activates three trans-ER membrane molecules, ATF6, IRE1α and PERK. 1) Upon ER stress, ATF6 is transported to the Golgi apparatus, where its cytosolic domain fragment (ATF6f) is released by enzymatic processing. ATF6f is a potent transcription factor and activates XBP-1 and other genes involved in recovery of ER function and cell survival. 2) IRE1α processes XBP-1 mRNA to generate the active transcription factor, spliced XBP-1 (XBP-1s). XBP-1s induces transcription of BiP and other genes involved in ER-associated degradation (ERAD), protein folding, lipogenesis, biogenesis and autophagy, and ultimate increase in ER mass and enhanced cell survival. Independent of its transcriptional function, XBP-1 directly interacts with and inhibits FOXO1 to block FOXO1-induced apoptosis. However, IRE1α also activates ASK1 and its downstream pro-apoptotic JNK. 3) PERK activation leads to phosphorylation of eIF2α. Phosphorylated eIF2α inhibits global protein synthesis but allows translation of ATF4, a transcription factor that induces CHOP expression, which leads to apoptosis. In addition, CHOP exerts feedback inhibition on phosphorylated eIF2α by inducing the expression of GADD34, which facilitates dephosphorylation of eIF2α. ATF4 and CHOP act together to promote apoptosis by increasing protein synthesis and inducing oxidative stress. 4) The release of calcium from the ER lumen through the calcium channels IP3R and RyRs could be either a cause or a consequence of ER stress. Elevation of the intracellular calcium level is a well-known mediator of apoptosis. Cytoplasmic Ca²⁺ triggers apoptosis through mitochondria and by activating CaMKII, which activates the apoptosis-inducing JNK.

ER is the cellular factory for biosynthesis of proteins, lipids and sterols. It exerts essential quality control functions in protein folding, maturation, trafficking and degradation. Moreover, ER is the major store of free calcium and contributes to critical cellular events by regulating intracellular calcium levels. The vital role of ER in the quality control of newly synthesized proteins makes it sensitive to the accumulation of unfolded or misfolded proteins. Elevations in the levels of these abnormal proteins beyond the handling capacity of the ER cause ER stress. This is revealed by the activation of characteristic signal transduction pathways that in aggregate are called the UPR⁴¹. The evolutionarily conserved UPR either counteracts the stress and enables the cell to survive, or introduces additional effects that lead to cell death. Thus the final outcome of UPR depends on the characteristics and levels of the downstream signaling molecules expressed, the dynamics of expression and the vulnerability of individual cell types: the cell may either achieve a new homeostasis of biosynthesis and intracellular calcium levels or initiate its own elimination.

The diverse stimuli that can trigger ER stress include hypoxia, nutrient deprivation, disruption of calcium homeostasis, high-fat diet and protein aggregates. All of these stimuli disturb ER functions and cause the accumulation of unfolded or misfolded proteins in the ER⁴¹. The ER senses the stress and initiates the canonical UPR through three ER-resident stress-sensing proteins: protein kinase RNA-like ER kinase (PERK), inositol-requiring protein-1 (IRE1α) and activating transcription factor-6 (ATF6) (Fig.1). The ER chaperone immunoglobulin-binding protein (BiP) binds to the three sensor molecules, which prevents their activation under homeostatic conditions. Increased unfolded or misfolded proteins recruit more BiP, which removes the inhibition and allows the dimerization and autophosphorylation of PERK and IRE1α, and translocation of ATF6 from ER to Golgi, thus leading to the activation of the UPR. ATF6 is cleaved sequentially by site-1 protease (S1P) and site-2 protease (S2P) in Golgi to generate an active transcription factor ATF6 fragment (ATF6f) ⁴¹. ATF6f is generally considered cytoprotective, although its downstream effectors have not been completely identified^{42, 43}. We will focus on two other UPR signaling pathways, which employ distinct downstream effectors, either complementing or opposing each other, to promote cell survival or death. These pathways may also converge at certain integration nodal points, however, to determine the final fate of the stressed cell.

The anti- and pro-apoptotic effects of the PERK-eIF2α-ATF4-CHOP pathway

This branch of UPR promotes both cell survival and cell death. PERK is a Ser/Thr protein kinase that phosphorylates eukaryotic translation initiation factor 2α (eIF2α), upon which cap-dependent mRNA translation is inhibited. Therefore PERK-phosphorylated eIF2α lowers ER stress by reducing the protein workload of ER. Translation attenuation by the PERK- eIF2α pathway is required for pancreatic β cell survival through translation attenuation⁴⁴; this process also enables other cell types to adapt to ER stress³⁰. A small molecule, Salubrinal, inhibits the dephosphorylation of eIF2α, thus attenuating protein synthesis and promoting motor neuron survival in mouse models of chronic degenerative illnesses^{23, 45}. Suppression of protein translation by phosphorylated eIF2α is not always protective, however, since it may block the expression of cell survival proteins⁴⁶.

The phosphorylated eIF2α allows selective translation of ATF4 through the internal ribosomal entry site (IRES)⁴¹. ATF4 induces the pro-apoptotic C/EBP homologous protein (CHOP)⁴¹. CHOP is believed to function as the major mediator of ER stress for apoptosis by regulating B-cell leukemia-2 (Bcl-2) family molecules, the principal intracellular regulators of cell survival and death. CHOP down-regulates anti-apoptotic Bcl-2⁴⁷ and up-regulates pro-apoptotic BH-3-only molecules Bim⁴⁸ and PUMA⁴⁹. CHOP also promotes oxidative stress through activation of ER oxidase 1α (ERO1α)^{50, 51}. ERO1α may also stimulate calcium release from ER through activation of calcium channel inositol (1,4,5)-trisphosphate receptors (IP3R)⁵². CHOP-induced calcium release from ER further activates calcium/calmodulin-dependent protein kinase II (CaMKII), which in turn induces death molecule Fas through activation of c-Jun N-terminal kinase (JNK)⁵³. JNK can also be activated by the IRE1α pathway (see details below), and thus may act as an integration point for both of these pathways. In addition, CHOP exerts feedback inhibition on phosphorylated eIF2α by inducing the expression of growth arrest and DNA damage-inducible protein-34 (GADD34), which facilitates dephosphorylation of eIF2α⁴¹. ATF4 and CHOP have recently been shown to act synergistically to promote protein synthesis and induce oxidative stress, which results in apoptosis⁵⁴. Importantly, superoxide induced by axon injury plays a critical role in RGC death^{55, 56}. Since ER stress is metabolically convergent with oxidative stress, it suggests that oxidative stress mediates ER stress –induced apoptosis³⁰.

The anti-apoptotic IRE1α-XBP-1 pathway vs the pro-apoptotic IRE1α-RIDD and IRE1α-JNK pathways

IRE1α is a bi-functional enzyme that contains both a Ser/Thr kinase domain and an endoribonuclease (RNase) domain. BiP release and subsequent occupation with misfolded protein allow the oligomerization and activation of IRE1α⁵⁷. IRE1α RNase activity cleaves XBP-1 mRNA unconventionally to generate the spliced form of XBP-1 (XBP-1s). XBP-1s is a potent transcription factor that induces various genes to restore ER homeostasis and promote cell survival by increasing ER biogenesis as well as promoting protein degradation through ER-associated degradation complexes (ERAD)⁵⁸. In addition to its transcriptional function, XBP-1 directly interacts with Forkhead box O1 (FOXO1) to assist its degradation through the proteasome system⁵⁹, which blocks FOXO-dependent apoptosis.

On the other hand, IRE1α also leads to cell death. Regulated IRE1-dependent decay (RIDD) is an XBP-1-independent signaling event downstream of IRE1α RNase activity. IRE1α degrades a spectrum of ER-located mRNAs by RIDD, which may contribute to apoptosis due to lost translation of cell survival genes^{60, 61}. IRE1α binds to an adaptor molecule, TNF-receptor-associated factor 2 (TRAF2) and activates apoptosis signal-regulating kinase 1 (ASK1), which in turn activates JNK⁶². JNK mediates both extrinsic death receptor-initiated and intrinsic mitochondria-originated apoptotic signaling pathways⁶³. ER-localized pro-apoptotic Bim and PUMA selectively activate the IRE1α-TRAF2-JNK pathway, but not XBP-1 splicing⁶⁴, indicating the causative link between the TRAF2-JNK branch of the IRE1α pathway and apoptosis. However, it is not known what determines whether the XBP1 splicing or JNK activation predominates, and therefore whether the IRE1α enhances cell survival or apoptosis.

How is ER stress initiated in axotomized neurons?

Our results indicate that the UPR is an intrinsic neuronal reaction to disrupted axonal integrity and that it could be a general upstream mechanism for neurodegeneration. It is, however, unlikely due to upregulation of unfolded or misfolded proteins in the ER. ER stress in the perikaryon might be activated indirectly by a retrograde injury signal. Alternatively, ER stress might be initiated locally in the axon and subsequently translocated to the cell body. At least three forms of retrograde signaling could be triggered by axon injury. These include rapid propagation of action potentials, interruption of target-derived neurotrophins, and slower dynein/microtubule-dependent transportation of locally synthesized signaling molecules^{65, 66}. The action potential triggers calcium influx both in the axon and the soma, which has been implicated in RGC death after optic nerve injury⁶⁷. The critical role of calcium in ER function, signaling and apoptosis³² makes it the leading candidate as the axonal damage signal that induces ER stress and apoptosis (Fig. 2). Smooth ER (sER) is the major intracellular calcium store and maintains a much higher calcium concentration (10-100uM) than cytoplasm (100-300nM) due to the activity of SERCA (smooth ER Ca²⁺ ATPase) in the ER membrane⁶⁸. Thapsigargin, a SERCA inhibitor, is the most commonly used experimental ER stress inducer, because of the sensitivity of ER to calcium homeostasis. Increase in axoplasmic calcium is an early neuronal response to axon injury and plays key roles in the cytoskeletal breakdown of Wallerian degeneration^{38, 69} and in neural excitotoxicity and apoptosis⁶⁸. The release of Ca²⁺ from ER through ryanodine receptors (RyRs) and IP₃Rs contributes significantly to the intra-axonal Ca²⁺ influx^{38, 68}, as does extracellular Ca²⁺ influx through ion-channels on axon membrane. The wave of intra-axonal Ca²⁺ elevation propagates retrogradely to the cell body, where it may disturb the calcium homeostasis and induce soma ER stress. Moreover, there is intimate physical interaction and signaling crosstalk between ER and mitochondria^69-71. Uptake by mitochondria of the Ca²⁺ released from ER results in opening of the mitochondrial permeability transmembrane pore (mPTP), energetic failure, outer membrane fragmentation, release of cytochrome C and finally apoptosis. Interestingly, mutations of presenilin 1 and 2 (PS1,2), which are located primarily in ER and involved in calcium release from ER, are linked with early onset of familial AD, indicating the possibility that the malfunction of the ER contributes to pathogenesis of AD⁷². In addition to calcium, an ever-increasing number of molecules have been identified as the retrograde injury signals from the axon to the cell body in peripheral neurons. These signals are mediated by the nuclear transport factors importin, RanGTPase and vimentin and the dynein complex, including members of the mitogen-activated protein kinase family (ERK, JNK and DLK1)^{65, 66} and transcription factors such as STAT3⁷³. It remains to be determined whether CNS neurons employ similar molecules and mechanism as retrograde injury signals. It will also be critical to determine whether and how these signaling molecules induce ER stress in the cell body.

Figure 2. The initiation and propagation of neuronal ER stress after axotomy.

The neuronal ER is a continuous membrane system that is comprised of the nuclear envelope, the sheet-like rER decorated with ribosomes in perikaryon, and the tubular sER, which is distributed throughout the axon as a continuous tubular network. ER stress may be induced in the cell body by retrograde signals including: 1) Injury-associated signaling such as Ca²⁺, ERK, JNK, DLK1 and STAT3. 2) ER stress molecules, such as XBP-1s, phosphorylated eIF2α and BiP, which are induced locally in axon. 3) Morphological changes in the axonal tubular sER.

Ca²⁺ is pumped into the ER lumen by the SERCA and released through RyRs and IP₃Rs to contribute to the increase of intra-axonal Ca²⁺ level early after axon injury. The wave of intraaxonal Ca²⁺ increase could propagate retrogradely to disturbant the calcium homeostasis and induce ER stress in the cell body. The uptake of the Ca²⁺ released from ER by mitochondria results in opening of mPTP, energetic failure, ROS production, outer membrane fragmentation, release of cytochrome C and finally apoptosis. Ca²⁺ activated calpain breaks down axonal cytoskeleton in Wallerian degeneration, which may subsequently change the shape and location of tubular sER. The morphology of the axonal sER is critical for axon maintenance, and relies on the interaction between microtubules and the ER-shaping proteins spastin, atlastin-1, REEP1 and reticulons. Alterations of these proteins cause axon abnormalities during development and disease, which may be due to impairing ER morphology and distribution. Axon injury disrupts the ER network which may cause ER stress directly. In addition, morphological changes of the axonal tubular sER itself or other locally activated ER stress signals may propagate along the ER endomembrane system to the cell body, where ER stress is activated.

It is possible that axon injury induces ER stress in the cell body and axon independently and nearly simultaneously. If so, then the earlier degeneration of the axon might be due to its greater sensitivity to ER stress. There are few studies of axonal ER stress, which might also be induced locally by calcium influx into the axon. It has been shown that XBP-1 mRNA splicing occurs in neurites and that the spliced form of XBP-1 protein is transported back to the neuronal soma after BDNF stimulation⁷⁴. LPA treatment also induces axonal eIF2α phosphorylation and BiP expression locally⁷⁵. These findings suggest that the UPR pathways can be activated in axons. If so, then this would raise the intriguing question of how ER stress is translocated retrogradely to the neuronal soma. Possible mechanisms include the retrograde transport of activated UPR molecules to the perikaryon, where they could induce cell death (Fig. 2). However, the unique distribution and morphology of the neuronal ER network may represent an even more intriguing mechanism for ER stress initiation and propagation as discussed below.

The neuronal ER network is a continuous membrane system that comprises the nuclear envelope; sheet-like rough ER (rER) decorated with polyribosomes, which is present predominantly in neuronal perikarya and proximal dendrites; and tubular sER, which is distributed throughout the axons and distal dendrites as a continuous tubular network^76-78. It is the major component of the cellular endomembrane system, which physically interacts with, and is functionally coupled with, other cellular organelles. Due to its function in biosynthesis, ER may support the massive membrane expansion that occurs during axon and dendrite formation, extension and maintenance⁷⁹; recent evidence demonstrates that the preservation of ER morphology in axons is important for normal neuronal development and maintenance^{77, 80}. The shape and proper position of the tubular sER network in axons is coordinated by interaction between spastin and other ER-shaping proteins, including atlastin-1, receptor expression enhancing protein 1 (REEP1) and reticulon-1^{77, 80}. Autosomal dominant mutations in these genes are responsible for the most common forms of hereditary spastic paraplegias (HSPs), a class of neurological disorders caused by degeneration of upper motoneurons and their axons in the CST⁸⁰. Remarkably, alteration of ER shape and distribution in axons appears to be a novel pathologic mechanism underlying both HSPs and ALS^{77, 80}. This association receives support from recent discoveries of the mutations of another ER-shaping protein reticulon-2 in a less common form of HSP⁸¹ and the ER-resident VAPB protein (vesicle-associated membrane protein-associated protein B) in a familial form of ALS⁸². However, the mechanisms by which changes in ER shape lead to neurodegeneration are unknown. One possibility is that damage of the tubular ER in the axon causes dysfunction of ER, which in turn triggers ER stress that can move along the endomembrane system of ER back to the cell body (Fig. 2). If so, then ER stress may also be involved in axon degeneration.

That axon degeneration is an actively regulated process is suggested by a mouse mutant in which Wallerian degeneration is delayed⁸³. The chimeric mutant protein slow Wallerian degeneration protein (Wld^s) contains the full-length NAD⁺-synthetic enzyme nicotinamide mononucleotide adenylyltransferase 1 (Nmnat1) at its C-terminal region, while the N-terminal region is required for translocation of Wld^s from the nucleus to the axon. Both of these features are necessary for axon protection⁸⁴, indicating that axonal NAD⁺ production is critical for axon survival. Another NAD⁺-synthetic enzyme, Nmnat2, is constantly transported to the axon to maintain axonal integrity⁸⁵. Nmnat2 is highly enriched in CNS neurons and is associated primarily with Golgi and ER, whereas Nmnat3 is located in mitochondria^{86, 87}. Since general elevation of cytosolic NAD⁺ level fails to protect injured axons, it has been suggested that the regulation of NAD⁺ levels in axonal ER and mitochondria is critical for axon survival, presumably due to its role in regulating cytosolic levels of calcium and reactive oxygen species (ROS)⁶⁹. Importantly, the elevation of calcium and ROS levels induces the mPTP, which links mitochondria dysfunction to axon degeneration^{88, 89}. Due to the close relationship between ER and mitochondria and their co-regulation of calcium and ROS^{30, 89}, it suggests that ER stress also plays a role in axon degeneration. Thus it would be very interesting to determine the relationship between the activities of Nmnats/NAD⁺ levels and ER function in axons after injury, which could indicate the molecular mechanism for axon degeneration.

Therapeutic strategies targeting ER stress for neuroprotection

If UPR activation is designed to help cells to cope with ER stress, why does it seem only to cause neuron death after axon injury? One possibility is preponderant pro-apoptotic UPR activation after axonal insults. Studies of cultured non-neuronal cells, for example, led to the proposal that the duration of IRE1/XBP-1 activation correlates with its protective effects⁹⁰. Consistent with this idea, we found that axotomy triggers differential activation of diverse UPR pathways in RGCs: whereas CHOP is robustly and persistently activated, XBP-1 is activated only transiently and modestly¹⁶. Differences in these responses might account for the prominent pro-apoptotic effect of ER stress in injured neurons. In addition, the unique polarization characteristics of the axonal compartments in neurons might favor retrograde transportation of specific signaling molecules which may also contribute to the differential activation of UPR.

Considerable experimental evidence shows the therapeutic potential of targeting ER stress in CNS injuries and suggests that better understanding of the responsible mechanisms is likely to generate even more effective treatments. For example, inhibition of ER stress is associated with neuroprotection in models of both ischemia and TBI²⁴. In addition, modulation of ER stress molecules BiP, CHOP, XBP-1 and ATF4 decreased motoneuron death¹⁷ and improved functional recovery after SCI^25-28. Our own work with axotomized RGCs has shown that manipulating CHOP and XBP-1 in opposite directions, by deleting CHOP or by overexpressing XBP-1s, exerts striking RGC-protection effects¹⁶. Since XBP-1 regulates numerous genes that are involved in multiple cellular functions, it is important to define the downstream mechanism by which XBP-1 contributes to neuroprotection. In a Drosophila model of AD, XBP-1 exerts a neuroprotective effect that is due to downregulation of RyR3, with consequent inhibition of calcium release from ER⁹¹. XBP-1 also could be deleterious to motoneurons, however, since neural specific deletion of XBP-1 delays the onset of ALS in SOD1 transgenic mice, potentially through upregulation of autophagy and clearance of SOD1 aggregates⁹². It remains to be determined why XBP-1 behaves differently in diverse neuronal populations. CHOP deficiency also protects neurons in several pathological conditions^29-33. We have found that the combination of increasing XBP-1 and deleting CHOP together exerts a significantly more potent neuroprotective effect on injured RGCs than targeting either pathway alone, suggesting the potential benefits of treatments that can act synergistically by targeting multiple UPR pathways¹⁶.

JNK, a toxic signal that is induced by axon injury^{93, 94}, is another promising molecular target for promoting neuroprotection, although JNK inhibition failed to slow PD in a clinical trial⁹⁵. JNK deficiency rescues RGCs after optic nerve crush⁹⁶, and ASK1, the upstream kinase of JNK, plays a key role in SOD1 induced ER stress and motoneuron loss⁹⁷. In addition, JNK inhibition modestly delays axonal Wallerian degeneration⁹³. JNK could be the mediator of apoptosis downstream of both PERK-CHOP-Calcium -CaMKII and IRE1α-TRAF2-ASK1 pathways and a nodal point that integrates two pathways that lead to neuron death.

Some obvious strategies based on recent advances in our understanding of ER stress await more extensive trials in vivo. For example, IRE1α activation can lead to cell death or survival, depending on the balance between its kinase and RNase activities. Its RNase activity is primarily directed to XBP-1 splicing and RIDD. IRE1α kinase activity is involved in JNK activation. It is now possible to block IRE1α kinase activity, thus inhibiting JNK activation and at the same time to fine tune IRE1α kinase activity to bias its RNase activity towards XBP-1 splicing but not RIDD^{60, 98}. It would be important to determine whether applying these strategies to modulate ER stress in vivo will enhance neuron survival.

Concluding remarks

Axon injury-induced ER stress reveals a new link between axon pathology and retrograde neurodegeneration. Because these sequential events may be a general phenomenon in both acute neural trauma and chronic neurodegenerative diseases, modulation of neuronal ER stress represents a novel and promising strategy for neuroprotection. The development of neuroprotective agents has proven to be challenging and complex. A more thorough understanding of how ER stress is activated in neurons, and how UPR pathways crosstalk with each other to determine cell fate, is a prerequisite for advances in the treatment of neurodegeneration. The correlative evidence described in this review suggests that local ER stress at the site of injury may elicit a retrograde signal that can induce cell death. What that signal is and how the transmission might function are unanswered questions. Since the ER network is continuous from nucleus to the most distal portions of the axon, it is possible that some signals of the UPR are transported within the ER membrane system. Finally, compelling evidence supports that neuron cell body death and axon degeneration are autonomous effects of injury and independent of each other. It is obvious, however, that these two sequential events must be linked in some fashion. Might ER stress be the common upstream mechanism responsible for neuron and axon degeneration in neurological diseases? More importantly, the complex and often contradictory characteristics of the diverse ER stress signaling molecules will demand systematic analysis of their definitive roles simultaneously in a broad spectrum of diseases to develop optimal combinatory neuroprotective strategies.