Old versus New Mechanisms in the Pathogenesis of ALS

Abstract

Amyotrophic Lateral Sclerosis (ALS) is recognized as a very complex disease. As we have learned in the past 20 years from studies in patients and in models based on the expression of mutant SOD1, ALS is not a purely motor neuron disease as previously thought. While undoubtedly motor neurons are lost in patients, a number of alterations in those cell‐types that interact functionally with motor neurons (astrocytes, microglia, muscle fibers, oligodendrocytes) take place even long before onset of symptoms. At the same time, disturbance of several, only partly inter‐related physiological functions play some role in the onset and progression of the disease. Traditionally, mitochondrial damage and oxidative stress, excitotoxicity, neuroinflammation, altered axonal transport, ER stress, protein aggregation and defective removal of toxic proteins have been considered as key factors in the pathogenesis of ALS, with the relatively recent addition of disturbances in RNA metabolism. This complexity makes the search for an effective treatment extremely difficult and prompts further studies to reveal other possible, previously unappreciated aspects of the pathogenesis of ALS. In this review, we focus on previous knowledge on ALS mechanisms as well as new facets emerging from studies on genetic ALS patients and models that may both provide precious information for a novel therapeutic approach.

Introduction

An unquestionable set of genetic evidence indicates that disturbed RNA metabolism is a key player in ALS pathogenesis 106. Indeed, more and more genes with an established or probable role in RNA processing have been discovered as to be causally linked to familial as well as to apparently sporadic forms of the disease. In particular, the identification of mutations in TARDBP and FUS, both genes encoding RNA binding proteins, initially proposed a pivotal role for aberrant RNA processing, a topic that was previously not fully appreciated as a possible mechanism of the disease pathogenesis. However, it was the more recent identification of the hexanucleotide repeat expansion in the C9orf72 gene as the most frequent genetic cause for ALS that truly and completely shifted the focus of ALS research and placed RNA mis‐processing in the spotlight. Mutations in several other genes encoding proteins involved in RNA metabolism have been described: although they occur only in a small fraction of fALS cases, at the same time they strengthen the relevance of this pathway in the pathogenesis of ALS.

Yet, ALS is far from being a pure RNA‐dysfunction disease. A wealth of experimental data has been accumulated over the last 20 years indicating that dysfunctions in a number of cellular functions, that is, mitochondrial metabolism and regulation of oxidative balance, modulation of neuronal excitability, control of the inflammatory response, axonal transport, protein folding and degradation, do have a role in disease pathogenesis 26. As new information is rapidly getting available, it is becoming clear that previous knowledge can be integrated into the new one, thus helping in shedding light into a real intricate disease.

TDP‐43 and FUS: RNA DYS‐Metabolism as a Novel Player in ALS

Mutations in the gene TARDBP (Trans active response DNA binding protein), encoding the 43 KDa protein TDP‐43, have been identified in 2008 and account for about 3% of fALS and about 1.5% of sALS cases 46, 65, 78, 126.

Shortly after this discovery, mutations in the gene encoding for the fused in sarcoma/translocated in liposarcoma (FUS/TLS) protein were identified in about 4% of familial ALS patients and rare sporadic cases 74, 78, 133. Together, these discoveries caused a great enthusiasm and opened a new perspective in the field. Indeed, FUS and TDP‐43 share structural and functional homology, being both DNA/RNA binding proteins belonging to the family of heterogeneous ribonucleoproteins (hnRNPs) 78. These are proteins that are able to shuttle between the nucleus and the cytoplasm where they play various functions in RNA‐related pathways, including transcription, pre‐mRNA splicing, mRNA transport and local translation 78. At steady state, FUS and TDP‐43 are localized predominantly in the nucleus, while ALS mutations cause a relocation of the proteins from nucleus to cytoplasm and their accumulation into stress granules 10, transient cytoplasmic assemblies where nonessential mRNAs are translationally repressed in response to stress 3. The pathological significance of this process is unclear, although it might imply a role of both FUS and TDP‐43 in stress‐granule mediated translational repression, a central process in the control of gene expression which is emerging as a crucial issue in the field of neurodegenerative diseases, including ALS 10. Yet, as it is believed that stress granules are permissive sites for protein aggregation, the localization of mutant TDP‐43 and FUS (and even of their wild‐type forms under stress conditions) into stress granules might represent the initial step that eventually generates TDP‐43 or FUS‐positive inclusions that typically mark affected tissues from patients (see below) 31, 84. Overall, these findings imply both a loss of the normal protein function in the nucleus and a gain of toxic function of the aggregated forms of TDP‐43 and FUS in the cytoplasm as potential pathogenic mechanisms. However, they do not provide straightforward clues about the pathological mechanisms whereby TARDBP and FUS mutations cause neurodegeneration. So far, the approach that has been most extensively used to answer this question consisted on high‐throughput screenings of the RNAs that are bound to FUS and TDP‐43. To date, several thousands of RNA species targeted by TDP‐43 107, 120, 131, 138 and FUS 56, 59, 76, 112 have been identified, but which of these interactions have physiological and/or pathological implications is still really unclear. It has been proposed that a set of common transcripts that are targeted by both proteins might be the most relevant for the disease, but again experimental evidence showing that this is the case are still lacking. Interestingly, most of them are pre‐mRNAs with very long introns (>100 kb) and the expression of these types of mRNAs, which typically encode proteins involved in neuronal functionality, is enriched in brain, thus providing a potential explanation for the selective neuronal vulnerability associated to FUS or TDP‐43 dysfunction 76.

Considering the thousands of RNA targets and the multiple roles in RNA processing that both FUS and TDP‐43 might have, it is not surprising that alterations in subcellular distribution and/or deposition in insoluble inclusions that characterize ALS patients have potential deleterious consequences for each step of RNA‐related cellular pathways (Figure 1A).

Figure 1.

Converging pathogenic mechanisms triggered by RNA‐related mutant genes. A. FUS and TDP‐43 have a central role in different steps of RNA related pathway. While they are mainly localized in the nucleus, where they act as transcriptional and splicing regulators, both FUS and TDP‐43 are also involved in cytoplasmic mRNA transport and local translation. ALS mutations alter their normal cellular distribution, relocating both proteins in cytoplasmic protein aggregates and/or stress granules. This causes both a loss of their normal, and mainly nuclear functions, and a gain of toxic function by protein aggregates in the cytoplasm. In particular, mutated FUS and TDP‐43 affect the proper splicing regulation of thousands of RNA targets. This might also be a result of sequestration of snRNPs and SMN splicing complexes in the cytosol, which in turn leads to their depletion from nuclear gems. Further, mutated FUS and TDP‐43 alter stress granule dynamics, thus impairing stress granules‐mediated translational repression in condition of stress. Local translation into distal neuronal processes and neuromuscular junctions (MNJ) might also be affected, as TDP‐43 plays a role in the axonal transport of defined target mRNAs for local translation. Finally, ALS‐associated mutation in FUS and TDP‐43 affect nuclear protein import, as in the case of C9orf72. B. The transcription of mutant C9orf72 gene lead to the expression of both sense and antisense C9orf72 RNA transcripts containing expanded G4C2 (and C4G2) repeat, which can affect normal cellular pathway in numerous manners. First, the C9orf72 RNA transcripts accumulate in the nucleus as RNA foci which interact and sequester different RNA binding proteins, leading to alterations in their normal cellular functions. Interaction between G4C2 repeat expansion and nucleolin causes nucleolar stress and defects in rRNA biogenesis. Similarly, the hexanucleotide repeat binds several splicing factors, such as hnRNP H, hnRNP A3, SRSF2, thus impairing splicing regulation of hundreds of genes. Second, expanded repeats impact on the control of nucleo‐cytoplasmic trafficking of both mRNAs and proteins by targeting mRNA export adaptors and the nuclear pore complex machinery. As a consequence, mRNA translation is affected. Protein translation might also be impaired by sequestration of translational regulators, such as initiation and elongation factors, which in turn might lead to impairment in stress granules‐mediated translation repression. Finally, bypassing normal surveillance mechanisms in an obscure way, expanded RNAs translocate from nucleus to cytoplasm, where they are translated through an unconventional process independent from the presence of an upstream ATG (RAN translation). The ribosomal reading of the sense and antisense transcripts lead to the expression of five different poly‐dipeptide repeat proteins (DPRs). DPRs accumulate in cytosolic inclusions that affect cell viability by unknown mechanism, but they also target nuclear RNA processing and protein transport leading to a detrimental loop of RNA dysfunction.

The majority of TDP‐43 and FUS binding sites in RNAs are introns, suggesting that a major function of these proteins is to control the splicing of hundreds of gene transcripts. Based on the possibility that ALS mutations might induce neurodegeneration as a consequence of both loss‐of‐function and gain‐of‐function splicing defects, different transcriptomic studies have been performed after protein knockdown or overexpression of wild‐type and mutant FUS and TDP‐43 4, 55, 76, 107, 110. Not surprisingly, an altered expression of thousands of aberrantly spliced mRNA molecules emerged in all of these studies, strengthening the link between the two proteins with splicing regulation. In addition to a direct alteration on this process, ALS‐related mutations may affect also the network of protein–protein interaction of FUS and TDP‐43 with other splicing factors. Indeed, FUS interacts with both RNA polymerase II and the small nuclear ribonucleoprotein complex snRNP U1, thus suggesting that it might have a role in coupling transcription to splicing 143. Moreover, FUS and TDP‐43 interact with the SMN complex, which is essential for spliceosome assembly 132, 140 and mutant FUS sequesters both SMN and snRNPs in protein aggregates in the cytoplasm, leading to alteration in the alternative splicing process 45, 142. Further, ALS mutations disrupt the interactions between TDP‐43 and different hnRNPs (A1, A2/B1, C1/C2 and A3) 12. Finally, both FUS and TDP‐43 are required for the formation of nuclear particles called gems, which are involved in the final stages of snRNP modification and assembly and are typically lost in spinal motor neurons from ALS patients 60, 132. Interestingly, mutant SOD1, a major determinant of familial ALS, impairs the recruitment of SMN into nuclear gems, which are significantly reduced in spinal motor neurons of mutant SOD1 mice 69. Considering that splicing alterations have been clearly reported in cell and mouse models of SOD1‐ALS 6, 81, one interesting possibility is that motor neurons are particularly vulnerable to splicing defects that might well represent a converging mechanism of different types of genetic ALS. This conclusion is further supported by the observation that mutations in SMN1 gene coding for SMN are responsible for spinal muscular atrophy (SMA), the most common form of juvenile motor neuron degeneration, and that SMN is a significant modifier of ALS pathogenesis in mouse models and in human patients 1.

TDP‐43 and FUS also play a critical role in the biogenesis of noncoding regulatory RNAs such as microRNA (miRNA) 43, interacting with the Drosha miRNA processing complex 51 and, at least for TDP‐43, with Dicer complexes 70. Moreover, TDP43 binds directly to some miRNAs and silencing of TDP43 significantly alters the abundance of several miRNAs in human cultured cells 13, 70. Importantly, the expression levels of these microRNAs were found up‐ or down‐regulated in some ALS patients, although the results were somehow inconsistent, which might depend on the different tissues analyzed 38.

FUS and TDP‐43 are key component of neuronal RNA transport granules, which contain mRNAs that must be transported from the cell body to axons and dendrites and locally translated, and that are, therefore, essential for the normal functionality of highly polarized cells as neurons in general, and motor neurons in particular 2, 39, 40, 86, 135. Interestingly, disease‐associated mutations in TARDBP impair this process 2, 86.

In addition to their function in RNA trafficking, cytoplasmic FUS and TDP‐43 might be involved in the regulation of mRNA translation. Although a direct role of these proteins in the control of translation machinery is still lacking, it is clear that sequestration of mRNA transcripts into stress granules in response to various cellular insults, such as oxidative, heat or endoplasmic reticulum stress, might impact on the overall control of stress granules‐mediated translational repression. As a matter of fact, stress granules dynamics is strongly affected by ALS mutations 84, and both genetic and pharmacological modulation of this process provide beneficial effects on animal models of the disease 32, 72.

C9orf72: Solid Foundations for RNA DYS‐Metabolism in ALS

A mutation in the C9orf72 gene is the most frequent cause of ALS, accounting for about 40% of familial and about 7% of sporadic ALS cases 89, and consists of an expansion of a GGGGCC (G4C2) hexanucleotide repeat between the noncoding exons 1a and 1b 28, 111. In healthy individuals, the hexanucleotide sequences have a median length of two repeats (range 0–20), whereas most of C9orf72‐associated ALS patients have several hundreds or even thousands of repeats. However, the minimal pathogenic repeat size remains undefined 113.

Although the mechanisms whereby the G4C2 repeat expansion leads to neurodegeneration are currently unknown, the nature of the mutation and its analogies with other noncoding repeat expansion disorders suggest two potential pathogenic mechanisms, including a loss of function of the protein encoded by C9orf72 gene, and a gain of toxic function of RNA transcripts containing the expanded repeat. While haploinsufficiency of C9orf72 protein, whose function is only scarcely characterized 35, 82, 145, does not seem to cause neurodegeneration 73, the simple overexpression of expanded G4C2 repeats causes reduced viability in cellular models and motor deficits in Drosophila melanogaster and mice 18, 37, 80, 139, 148. In these cases, cell toxicity could be mediated by the accumulation of pre‐mRNAs containing the expanded repeat and/or the formation of poly‐dipeptide proteins, which are unconventionally produced by the so called repeat‐associated non‐ATG (RAN) translation.

Intracellular accumulation of pre‐mRNAs containing repeat expansions as RNA foci is a clinical feature of C9orf72‐associated patients: both sense and antisense RNA foci have been indeed detected in cortical, hippocampal, cerebellar and spinal motor neurons of ALS patients 44, 77, 94. RNA foci may exert a deleterious gain‐of‐function effect through the sequestration of RNA binding proteins, whose decreased availability might lead to alterations, once again, in various steps of RNA processing. To date, different groups have reported several candidate binding partners, including splicing and translational factors, and identified the presence of various of these proteins in RNA foci 21, 33, 52, 80, 97, 115, 139. Although a pathogenic role of specific repeat binding proteins in ALS neurodegeneration has not yet been definitively proved, emerging data have started to shed light on the possible toxic mechanisms induced by expanded RNAs (Figure 1B). We have recently observed that the expression of a (G4C2)₃₁ repeat expansion in cultured cells induces stress granules‐associated translational repression, which is accompanied by a marked accumulation of poly(A) mRNAs in cell nuclei, thus suggesting that defective trafficking of mRNA, as a consequence of impaired nuclear mRNA export, might contribute to the pathogenesis of C9orf72 ALS 115. Nuclear retention of mRNAs was later confirmed by Freibaum et al in the nuclei of Drosophila cells expressing expanded G4C2 repeats and in induced pluripotent stem‐cell‐derived neurons from C9orf72‐associated patients 37. Importantly, at least 18 proteins involved in the nucleo‐cytoplasmic transport act as genetic modifiers (enhancers or suppressors) of eye degeneration induced by expanded G4C2 in Drosophila 37, further suggesting the involvement of this pathway in the pathogenesis of C9orf72 ALS. A similar conclusion was obtained by Zhang et al, who picked up the protein RanGAP1, a central regulator of protein trafficking between nucleus and cytoplasm, as a key target of the repeat expansion 148. In particular, the overexpression of RanGAP (orthologue of human RanGAP1) is a potent suppressor of eye degeneration and locomotor defects observed in Drosophila model expressing G4C2 repeats. RanGAP is able to bind the hexanucleotide repeat, can colocalize with repeat RNA foci and is mislocalized in repeat‐expressing flies, iPSC‐derived neurons and in C9orf72 ALS patient brain tissue. Altered localization of RanGAP is associated to a loss of function mechanism, as demonstrated by the impairment of import of nuclear proteins in Drosophila cells and in iPSC neurons. Interestingly, pharmacological targeting of nuclear export machinery rescues the neurodegenerative phenotype, suggesting a potential therapeutic approach 148.

A direct effect of repeat expansion on RNA processing has also emerged from gene expression profiling, which revealed a number of differentially expressed genes and exons in purified motor neurons and other tissues from C9orf72‐associated patients compared to normal controls 22, 109, 117. Notably, among the genes found aberrantly spliced, enrichment in genes encoding post‐transcriptional regulatory factors was observed, suggesting a potentially detrimental vicious cycle for proper RNA metabolism.

It is therefore clear from this short survey of the available experimental evidence, that RNA processing is a crucial issue in ALS pathogenesis, a concept that is further strengthened by a number of mutations in other genes encoding for proteins involved in RNA metabolism that were found to be associated, although rarely, with ALS.

Mutations in the gene encoding senataxin (SETX), a DNA/RNA helicase predicted to be involved in several steps of gene expression and the maintenance of genomic integrity, have been identified in rare cases of autosomal dominant juvenile ALS 9, 17. Mutations in the gene encoding angiogenin (ANG), a tRNA specific RNase involved in the transcription of ribosomal RNA, were found in adult onset forms of familial and sporadic ALS 50. Mutations in elongation protein 3 (ELP3), the histone H3/H4 acetyl transferase which is the catalytic subunit of the elongator complex, regulator of transcriptional elongation and post‐transcriptional processing of tRNA, were also associated to ALS 123. Moreover, mutations in genes encoding Ewing sarcoma breakpoint region 1 (EWSR1) and TATA box binding protein‐associated factor 15 (TAF15) also cause ALS 23, 24, 130. Both these proteins are RNA binding proteins structurally and functionally related to the other disease‐associated protein FUS, that form together the FET family, and play a regulatory role in transcription and alternative splicing. Additional genetic mutations were more recently identified in different hnRNPs, such as hnRNP A1, hnRNP A2/B1 and Matrin3 63, 71. Altogether, these findings reinforce the notion that RNA metabolism is a vulnerable point for motor neurons, although the pathogenic mechanisms triggered by mutations in all these genes remain to be defined.

Dipeptide Repeat Aggregation: A New Player in An Old Scenario

The discovery that expanded nucleotide repeats can be translated into polypeptides (RAN peptides) with mechanisms that bypass the conventional rules of mRNA translation, such as the presence of an ATG‐initiated open reading frame, not only represents one of the most exciting and important findings in basic research over the last years, but it also promises to deeply impact on our knowledge of the mechanisms underlying different disease of the nervous and muscular tissues. To date, spinocerebellar ataxia type 8 (SCA8), myotonic dystrophy type 1 (DM1) and Fragile‐X tremor ataxia syndrome (FXTAS) have been shown to be associated to this process 19. Most importantly, growing experimental evidence points to RAN peptides as major determinants of C9orf72 ALS pathogenesis.

Both sense and antisense C9orf72 transcripts containing G4C2 repeat expansion undergo RAN translation, resulting in the expression of five dipeptide repeat proteins (DPR): Glycine‐Proline (GP), Glycine‐Arginine (GR), Glycine‐Alanine (GA), Proline‐Alanine (PA) and Proline‐Arginine (PR) 5, 44, 96, 98, 150. Analysis of postmortem tissues from C9orf72‐associated patients has shown that these dipeptide repeat proteins form cytoplasmic p62 positive and TDP‐43 negative inclusions and, less frequently, intranuclear inclusions in neuronal cells, which might both contribute to the disease pathogenesis 88, 90. The translational products of C9orf72 repeat expansion seem to be sufficient per se to cause ALS neurodegeneration, as expression of synthetic constructs lacking G4C2 repeats and producing DPRs are toxic in cellular models, as well as in Drosophila 93, 95, 129, 137, 141, 147. However, these evidence are inconsistent with the observations that the amount of DPR inclusions does not seem to correlate with the clinical severity in ALS patients 88, and that the DPR protein aggregates are mainly distributed in brain regions such as cerebellum, hippocampus, neocortex, while are rare in the brainstem and in the spinal cord and, strikingly, almost absent in motor neurons 27, 48. Yet, the presence of soluble DPRs cannot be excluded in these tissues, where they might interfere with the normal cellular pathways independently from their accumulation into detectable inclusions.

How DPRs cause neurotoxicity is completely unknown. Interestingly, even DPR proteins seem to be capable of affecting RNA metabolism. Different groups have indeed shown that Arginine‐rich DPRs, in particular PR, translocate to the nucleus, where they localize and aggregate in the nucleoli, leading to nucleolar stress response and alteration in RNA biogenesis 75, 129, 137. Notably, nucleolar stress is also triggered by the interaction of expanded RNA transcript with nucleolin, an essential nucleolar protein 52. Moreover, several proteins involved in nucleocytoplasmic transport have been identified as potent modifiers of toxicity induced by PR dipeptides in Saccharomyces cerevisiae 64. Finally, among the recently identified GR and PR interacting proteins, many are RNA binding proteins and ribosomal proteins, further suggesting a crucial role of this pathway in the disease pathogenesis 129. Interestingly the analysis of poly‐GA coaggregating proteins revealed an interactome network enriched for proteins involved in the ubiquitin‐proteasome system regulation 93, thus pointing to protein aggregation as a major player in ALS, a concept that has been repeatedly invoked to explain the action of other aggregated proteins in ALS.

Protein Aggregation: A New Scenario for An Old Player

Protein aggregates in motor neurons are a pathological hallmark of ALS, which may be regarded as a classical proteinopathy. Motor neuron degeneration is often preceded by the formation of inclusions containing ALS associated proteins that are usually also ubiquitin‐positive. At variance with mutant SOD1, which is found aggregated exclusively in patients carrying SOD1 mutations, and mutant FUS, which is aggregated mainly in patients carrying FUS mutations and only in a few juvenile and sporadic forms of the disease 29, TDP‐43 is aggregated in most patients with sporadic and familial ALS including those carrying C9orf72 expansion 87 or other unrelated mutations (ie, OPTN, UBQLN2, VCP, ANG, ATXN1, PFN1) 29. A complex interplay between the aggregated ALS proteins and the mis‐functioning of specific mechanisms that are involved in their proper folding/disposal seems to be involved in ALS (Figure 2).

Figure 2.

Protein aggregation‐related toxicity in ALS. SOD1, TDP‐43 and FUS may share toxicity mediated by protein aggregates that form both as a consequence of gene mutations and/or as a result of oxidative or ER stress. In the case of FUS and TDP‐43, this process can be enhanced by their localization into stress granules. Correct protein handling by the Ubiquitin Proteasome System (UPS), as well as proper aggregate removal by the autophagic machinery can be hampered by the same aggregated proteins, but also by dysfunction/mutations in proteins that regulate these processes. As a result, aggregates accumulate over time and can propagate into neighboring cells following their extracellular release, which is due to discharge of seeds of aggregates by dying cells, or mediated by exosomes. Inclusions of Dipeptide Repeat Proteins (DPRs) that are formed by an aberrant translation of the C9orf72 expanded repeat, are thought to share the same toxic mechanisms as the other aggregation‐prone ALS factors.

The mechanisms of mutant SOD1 aggregation are quite well understood as it involves cysteine‐mediated polymerization and amyloid fibrils formation 15; intracellular accumulation of aggregates is favored by impairment of protein degradation systems in neurons 8 and, to a lesser extent, in muscle cells 42. It is also quite clear that these aggregates are linked to induction of mitochondrial damage, another recognized facet of ALS pathology 25, 36, 57. Conversely, the mechanisms ruling aggregation of wild‐type and mutant TDP‐43 are still largely debated, with evidence attributing a main role in the cysteine‐mediated aggregation process to the C‐terminal RNA‐recognition motifs (RRMs) of the protein, as well as to a glycine‐rich sequence containing a prion‐like domain 105, 122, 136 or to the N‐terminal domain 11, 114, 146.

Oxidative stress promotes TDP‐43 cysteine oxidation and disulfide crosslinking 20, 122, a fact that may explain aggregation of the wild‐type protein in most ALS patients. It also promotes TDP‐43 acetylation on two specific lysine residues (K145 and K192) within the RNA‐binding domains; in turn, acetylation may prompt aggregation and loss of function of the protein 20. A direct physical interaction between histone deacetylase 6 (HDAC6) and TDP‐43 exists in vivo 54 and, thus, it is also possible that under exposure to oxidative stimuli TDP‐43 aggregation prevents effective HDAC6 binding.

Oxidative stress also prompts hyperphosphorylation of aggregated TDP‐43 on C‐terminal serine residues 58 and this may contribute to TDP‐43 aggregation in patients 53, although other studies suggest that phosphorylation prevents rather than promotes TDP‐43 aggregation 83.

Aggregation of wild‐type and mutant FUS has also been investigated, although most studies were focused on FUS mis‐localization in specific protein aggregates such as stress granules (see above) 10 rather than in large unfunctional aggregates. FUS is also able to form RNA‐dependent aggregates which are distinct from stress granules 121, with both the low‐complexity domain and the R/G rich domain of FUS contributing to assembly of fibrous β zipper structures 118. As for SOD1 and TDP‐43, oxidative stress accelerates the formation of FUS granules 127. Most important, as other neurodegeneration‐linked proteins including TDP‐43, FUS contains disordered, prion‐like regions that render it very prone to aggregation. In particular, as demonstrated in a recent elegant study by Patel et al, FUS function requires its existence in the form of “liquid droplets” that convert with time to an aggregated state, and this transition is accelerated by ALS mutations 104.

Why motor neurons should be more sensitive to protein aggregates than other cell‐types is still not clear. However, as discussed in two recent reviews 42, 61 glia and muscle cells may be better equipped than neurons to handle misfolded proteins and counteract toxicity due to aggregates, possibly because they are able to better activate molecular chaperones and protein degradation systems including the immunoproteasome.

Correct protein folding and inhibition of protein aggregation is facilitated by a quality control system that involves a network of proteins including molecular chaperones and the ubiquitin proteasome system (UPS). A number of studies reported defects in this system in models for mutant SOD1 toxicity (reviewed in [8]) and in postmortem tissue from patients 66, and a recent study in two strains of G93A‐SOD1 mice correlates severity of phenotype in terms of early onset and fast progression with low levels of soluble chaperones and with malfunction of the proteasome degradation machinery 91. Thus, increasing the level of chaperones or even inducing autophagy may constitute an approach to removal of SOD1 aggregates as suggested from studies in models 67.

However, the relation between ALS pathogenesis and autophagy is still not clear. For instance, rapamycin, an MTOR‐dependent autophagic activator, accelerates disease progression in the SOD1(G93A) mouse model of ALS 144. Conversely, treatment with trehalose (an MTOR‐independent autophagic inducer) was reported to have a number of beneficial effects in these mice, including delay of onset, prolonged life span, reduction of motor neuron loss, decreased SOD1 aggregation and ubiquitinated protein accumulation in one study 149, while it became less effective at delaying further disease progression as the disease progressed and it failed to extend the survival of the same mice in another study 85. Interestingly, BECN1 levels are upregulated in mutant SOD1 mice but life span is increased in these mice if they are made haploinsufficient for Becn1 compared with littermate control animals. Furthermore, an altered equilibrium between monomeric and oligomeric forms of mutant SOD1 is observed in the spinal cord of these mice, suggesting an abnormal interaction of the mutant protein with the BECN1‐BCL2L1 complex 100.

TDP‐43 proteostasis is also maintained by the coordinated action of the UPS and autophagy, which may be particularly important for clearing TDP‐43 oligomers and aggregates as these are found in most ALS patients. Activators of the UPS or autophagy promote TDP‐43 clearance and/or mitigate toxicity in models overexpressing TDP‐43 119. Interestingly, autophagy activation with rapamycin reduces the accumulation of FUS‐positive SGs and also reduces neurite fragmentation and cell death in neurons expressing mutant FUS under oxidative stress 116.

In this context, it is worth recalling that mutations in Ubiquilin 2, a member of the ubiquilin family, which regulates the degradation of ubiquitinated proteins, are linked to abnormal protein aggregation and neurodegeneration in a small subset of familial ALS and ALS/FTLD patients and that Ubiquilin 2 dysfunction is found also in patients without UBQLN2 mutations 30. Mice expressing mutant forms of UBQLN2 variably develop a motor phenotype at 3–4 months accompanied by large neuronal cytoplasmic inclusions and ubiquilin‐2‐positive inclusions that colocalize with ubiquitin, p62/SQSTM, optineurin, and occasionally TDP‐43, but never with FUS 14.

Protein misfolding may also be a consequence of ER stress. Dysfunction in protein handling in the ER and the following stress are typically associated with neuronal damage in SOD1 models and in patients 68, 92, 128 including those carrying FUS mutation 34, 134, with a controversial role of protein disulphide isomerase (PDI), a disulphide bond‐modulating ER chaperone that also facilitates the ER‐associated degradation of misfolded proteins 62.

ER stress is activated by the expression of poly(GA) DPRs in cultured cells and primary neurons 147, although the relevance of this phenomenon in patients has not been documented. ER stress is seen also in skeletal muscle across the lifespan of G93A‐SOD1 mice; activation begins before onset of symptoms and increases with disease progression and most probably contributes to muscle atrophy and weakness in ALS 16. Interestingly, it has been recently reported that ALS‐typical mutant SOD1, TDP‐43 and FUS perturb protein transport in the early secretory pathway between ER and Golgi compartments by distinct mechanisms, but each process is dependent on Rab1, which colocalizes with the mutant proteins and is probably misfolded itself 125.

Prion Propagation: A New Mechanism for the Toxicity of ALS Protein Aggregates

Whether TDP‐43 and FUS are really toxic inside neurons, and by which mechanism beside protein and RNA sequestration (if any), is also still not entirely clear.

What is now emerging as an interesting facet of the toxicity of protein aggregates in ALS is the possibility that these aggregates do not act simply through a loss‐of‐function due to sequestration of the proteins (or RNAs?), but rather actively diffuse damage to neighbor cells via a prion‐like mechanism that might contribute to the noncell autonomous nature of the disease.

ALS neurodegeneration typically begins focally and then spreads in an orderly propagating process 108 that is reminiscent of the seeding and self‐propagation seen in prion disease 79. As mentioned above, both TDP‐43 and FUS contain prion‐like domains that render them quite prone to spontaneous aggregation. Thus, it is tempting to assume that these domains allow diffusion of toxic aggregates to neighboring cells as prions do. However, to consider ALS a prion‐like disease, several conditions must be fulfilled 124.

First, ALS misfolded proteins must have “seeding” properties and form self‐aggregates. This condition seems to be verified for SOD1 103, for TDP‐43 41 and FUS 101.

Second, protein aggregates must be able to spread and propagate to neighboring cells. This is the case, at least in cultured cells, for TDP‐43 102. Aggregated mutant SOD1 is able to enter cells efficiently by macropinocytosis and to nucleate aggregation of the cytosolic protein with a self‐perpetuating mechanism 99. This process does not require cell‐to‐cell contact, but obviously depends on the extracellular release of aggregates, which can be due to discharge of seeds by dying cells or mediated by exosomes 49. Mutant SOD1 is secreted by a mechanism involving exosomes in NSC34 cells 47 and astrocyte‐derived exosomes efficiently transfer mutant SOD1 to spinal neurons and induce selective motor neuron death 7, which is the third and final property to be possessed by a protein to be considered “prion‐like.” TDP‐43 is also secreted via exosomes at least in part and it is also able to induce neurotoxicity in vitro 102.

Whether FUS possesses all the required prion characteristics and whether C9orf72 dipeptide repeat proteins are capable of similar spreading mechanisms is currently not known.

Conclusions

Studies sprouting from the individuation of the main genetic causes of ALS have greatly increased our knowledge of the processes playing a role in the pathogenesis of this complex disorder. Most importantly, they have substantially enlarged and modified our view of the molecular factors that are relevant for the development of the disease. Although the real relevance of some of these factors is still to be conclusively ascertained, precious suggestions for therapy may hopefully soon derive and be translated into clinical trials.