TDP-43/FUS in Motor Neuron Disease: Complexity and Challenges

Abstract

Amyotrophic lateral sclerosis (ALS), a common motor neuron disease affecting two per 100,000 people worldwide, encompasses at least five distinct pathological subtypes, including, ALS-SOD1, ALS-C9orf72, ALS-TDP-43, ALS-FUS and Guam-ALS. The etiology of a major subset of ALS involves toxicity of the TAR DNA-binding protein-43 (TDP-43). A second RNA/DNA binding protein, fused in sarcoma/translocated in liposarcoma (FUS/TLS) has been subsequently associated with about 1% of ALS patients. While mutations in TDP-43 and FUS have been linked to ALS, the key contributing molecular mechanism(s) leading to cell death are still unclear. One unique feature of TDP-43 and FUS pathogenesis in ALS is their nuclear clearance and simultaneous cytoplasmic aggregation in affected motor neurons. Since the discoveries in the last decade implicating TDP-43 and FUS toxicity in ALS, a majority of studies have focused on their cytoplasmic aggregation and disruption of their RNA-binding functions. However, TDP-43 and FUS also bind to DNA, although the significance of their DNA binding in disease-affected neurons has been less investigated. A recent observation of accumulated genomic damage in TDP-43 and FUS-linked ALS and association of FUS with neuronal DNA damage repair pathways indicate a possible role of deregulated DNA binding function of TDP-43 and FUS in ALS. In this review, we discuss the different ALS disease subtypes, crosstalk of etiopathologies in disease progression, available animal models and their limitations, and recent advances in understanding the specific involvement of RNA/DNA binding proteins, TDP-43 and FUS, in motor neuron diseases.

1. Introduction

Amyotrophic lateral sclerosis (ALS), a fatal neuromuscular disease characterized by degeneration of upper and lower motor neurons of spinal cord, was first described by the French neurologist Jean-Martin Charcot in 1869, as clinical and pathological symptoms of muscular atrophy and hardening of lateral spinal cord (Chio et al, 2009b; Leblond et al, 2014). Progressive bulbar, limb, thoracic, and abdominal muscle atrophy in ALS causes death within 3-5 years after onset of symptoms in most ALS patients, normally due to respiratory failure (Chio et al, 2009a). Clinically, ALS may impact bulbar or spinal innervated muscles at onset, based on the relative loss of upper and/or lower spinal neurons, which is symptomatically consistent with bulbar or limb motor defects. While spinal neurons are predominantly affected, loss of neurons may also be observed in the brainstem and motor cortex (Al-Chalabi et al, 2012).

ALS is the most common degenerative disorder of motor neurons in adults, with incidence rates of 2-5 per 100,000 individuals worldwide. Men are 1.3 times more likely to develop ALS than women (Chio et al, 2013; Leblond et al, 2014). Despite its relatively low incidence compared to other neurodegenerative diseases like Alzheimer’s Disease (AD) (Hebert et al, 2013); or Parkinson’s disease (PD) (de Lau & Breteler, 2006); the devastating physiological effects and rapid lethality are the prominent features of ALS. Most ALS cases are sporadic, while about 8-10% are inherited (Figure 1).

Figure 1. Illustration of the molecular and pathological features of sporadic and familial ALS.

The sporadic disease subtypes account for ~90 % of ALS cases and which could be classified based on inclusions and the protein within. In the recent years, mutations and defects in several new ALS causing genes have been implicated in distinct subgroups of ALS patients (FALS). However, proteins encoded by these genes are also found in protein inclusions/aggregates in sporadic patients (SALS). There is a distinct pattern of co-localization or overlap of pathology among the ALS-linked protein inclusions, which are indicated. *NR: not reported.

Many ALS patients (~36-51%) also exhibit cognitive impairment, with about 20% developing frontotemporal lobar degeneration (FTLD) (Broustal et al, 2010). The reverse is also seen, with some patients with FTLD developing ALS as well (Lipton et al, 2004). This has led to the theory that ALS and FTLD are part of a clinical spectrum of disease. FTLD accounts for 10-15% of dementias, making it the second most common type of dementia for people under the age of 65, after AD. Pathologically, FTLD includes multiple subtypes, including FTLD-tau, FTLD-DPR (dipeptide repeat proteins), FTLD-UPS (ubiquitin-proteasome system), FTLD-FUS (fused in sarcoma/translocated in liposarcoma), and FTLD-TDP-43 (Tar DNA binding protein 43), although the distinction among the subtypes is not very clear. Most patients with ALS-FTLD have TDP-43 pathology, including FTLD-DPR, raising the question about the relative toxicity of DPR compared to TDP-43 pathology. Early-onset ALS-FTLD patients typically present cognitive changes, followed by muscle weakness (primarily upper body weakness), behavioral changes (such as euphoria, indifference, personality changes, and language impairments), paucity of speech, impaired comprehension, and even mutism. However, unlike AD, overall memory is relatively preserved in most patients with ALS or ALS-FTLD (Mitsuyama, 1993). The overlapping pathogenesis of ALS and FTLD suggests that these motor neuron diseases and cognitive deficits could have a common molecular basis, which is discussed later in this article.

1.1. Etiopathology and disease subtypes

The etiopathogenesis underlying degeneration of motor neurons in ALS is complex. Endogenous factors like accumulation of pro-oxidant metals and free radicals, glutamate excitoxicity, protein aggregation, mitochondrial dysfunction, and deregulation of RNA processing have been associated with dysfunctional motor neurons (Ferraiuolo et al, 2011). Various environmental factors including cigarette smoking, occupational exposure to electromagnetic radiation, heavy metals, pesticides, diesel exhaust and head traumas may increase the risk of developing ALS (Garruto et al, 1985b; Mitra et al, 2014b; Perl et al, 1982; Schmidt et al, 2010; Weisskopf et al, 2009; Yanagihara et al, 1984) and even in familial cases, an interplay of genetic and environmental factors has been linked to increased susceptibility. The first evidence of environmental triggers of sporadic ALS (SALS) was established with the indigenous populations of Guam in the Mariana Islands and in the Kii peninsula of Japan where ALS occurred with unusual incidence among these populations. Guam-ALS patients presented higher iron levels and lower zinc levels in brain and chronic nutritional deficiencies of calcium that could lead to higher intestinal absorption of toxic metals and co-deposition of calcium, aluminum and silicon in neurons (Garruto, 1991; Garruto et al, 1985a; Yasui et al, 1993) Furthermore, cyanobacteria-derived neurotoxin, beta-Methylamino-L-alanine (BMAA) is a candidate neurotoxin, like metals in the Guam ALS and Parkinsonism-dementia complex (ALS/PD), particularly prevalent in indigenous populations of Guam and Rota who consumed cyanobacteria-infested cycads (Cox & Sacks, 2002; Garruto et al, 1980; Garruto et al, 1981; Hoffman et al, 1977; Plato et al, 2003). BMAA binds to N-methyl-D-aspartate (NMDA) and (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) (AMPA) receptors, producing a molecule that resembles glutamate and activates glutamate receptor 5, inducing oxidative stress (OS). BMAA also inhibits the cysteine/glutamate antiporter system X_c⁻, leading to glutathione depletion, which further enhances OS. Moreover, BMAA may be incorporated into proteins, causing their misfolding or aggregation (Arif et al, 2014; Rao et al, 2006; Weiss et al, 1989). However, it is unclear, which of these cytotoxic properties of BMAA may be critical for Guam ALS. In addition, the Kii peninsula population was later linked to some genetic contribution/ founder effect (Arif et al, 2014).

Despite the range of pathological and clinical features among various neurodegenerative disorders, the identification of misfolded protein-rich inclusion(s) has become one of the molecular hallmarks of neurodegeneration. ALS is no exception, as its pathology involves accumulation of proteinaceous aggregates including diverse proteins like redox regulator Cu-Zn superoxide dismutase 1 (SOD1), and RNA/DNA binding proteins TDP-43 and FUS/TLS. Furthermore, the genes encoding these proteins SOD1, FUS, TARDBP and C9orf72, which encodes a yet-to-be characterized protein, have been implicated in 4 major ALS subtypes (Figure 1).

TARDBP

In 2006, the gene coding for TDP-43, an RNA/DNA binding protein, was implicated in ALS as the major component of ubiquitinated inclusions (Arai et al, 2006; Neumann et al, 2006). TDP-43 is involved in RNA processing, including splicing, transcription, and transport. The primary histopathological feature in a major subset of ALS cases is the inclusion of TDP-43 in the cytoplasm of upper and lower motor neurons and in other regions of the central nervous system (CNS), including the frontal and temporal cortex. As previously mentioned, there is evidence of an overlapping histopathology between C9orf72 genetic lesions and TDP-43 inclusions, but the mechanism is not clearly characterized (Freibaum et al, 2015). Recent studies identified co-localization of TDP-43 and RNA-binding motif-45 (RBM45) inclusions and their RNA-dependent association in motor neurons of some cases of ALS (Collins et al, 2012; Mladinic et al, 2010). Like TDP-43, RBM-45 is mainly nuclear, but migrates to cytoplasm and co-localizes in SGs to associate with Kelch-like ECH-Associated protein 1, a component of anti-oxidant machinery (Bakkar et al, 2015). Furthermore, TDP-43 co-localizes in cytoplasmic inclusions with Poly-A binding protein - 1 (PABP-1); a stress granules marker (McGurk et al, 2014).

TDP-43 is ubiquitously expressed in most tissue and cell types, including neurons and glia in the CNS. Studies of ALS, FTD, and ALS-FTD have shown that TDP-43 is cleaved, hyperphosphorylated, ubiquitinated or mis-localized in cytoplasm in the form of insoluble inclusions (Arai et al, 2006; Feiguin et al, 2009; Kwong et al, 2007; Neumann et al, 2006). ALS and ALS/PDC patients in Guam also present TDP-43 inclusions as a secondary pathology (Maekawa et al, 2009). ALS-parkinsonism disease subtypes show TDP-43 pathology in motor neurons, hippocampus, amygdala, globus pallidus, caudate, and putamen (McCluskey et al, 2009). Mutations in TARDBP gene, mostly within the glycine-rich C-terminal domain (CTD), are associated with 1-3% of cases of ALS with familial history (Daoud et al, 2009; Valdmanis et al, 2009).

FUS

The FUS/TLS protein was linked to ALS in 2009, as a component of inclusions found in ALS patients (Kwiatkowski et al, 2009; Vance et al, 2009). Although TDP-43 and FUS exhibit similar structures, functions, and pathobiology in ALS, TDP-43 pathology is notably absent in ALS-FUS cases (Kwiatkowski et al, 2009; Ticozzi et al, 2009). Like TDP-43, FUS/TLS protein is sequestered in the cytosol of ALS-affected motor neurons. Mutations in FUS cause severe loss of motor neurons in the spinal cord, moderate loss of upper motor neurons, and are associated with juvenile ALS (Kwiatkowski et al, 2009; Ticozzi et al, 2009). Patients with ALS-FUS develop distinct phenotypic patterns: early onset, with aggressive clinical progression, or late onset, with slower disease progression (Ravits et al, 2013). Pathological analysis has indicated that truncating mutations in the FUS gene can generate more aggressive phenotypes than missense mutations (Waibel et al, 2013). Interestingly, ALS-FUS appears to be clinico-pathologically distinct from FTLD-FUS. FTLD-FUS is not usually associated with mutations and the inclusions contain not only FUS, but also other FET (FUS, EWS and TAF15) proteins (Mackenzie et al, 2011). ALS-FUS present inclusions that co-localizes with stress granule marker, PABP-1 (Gal et al, 2011) and Ataxin-2 (ATXN2) (Elden et al, 2010). In addition, an unusual case has been recently reported with mutations in both FUS and TARDBP genes (King et al, 2015). This particular case exhibited a FUS P525L mutation as well as the truncating TARDBP Y374X mutation, showing moderate FUS pathology and no TDP-43 pathology, with extensive granular and p62-positive, TDP-43-negative inclusions in the spinal cord and motor/neocortex (King et al, 2015).

Other major ALS subtypes and their distinct and overlapping pathology with ALS-TDP or ALS-FUS SOD1

The primary role of antioxidant metalloenzyme, SOD1, in healthy cells is to protect cells from OS by neutralizing the toxicity of superoxide radicals to dioxygen and hydrogen peroxide molecules (Fridovich, 1978). SOD1 was the first protein to be implicated in ALS in 1993 (Rosen et al, 1993). Subsequently, more than 160 mutations in the SOD1 gene are found in ~20% of familial ALS (FALS) cases and ~2% of SALS cases (Acevedo-Arozena et al, 2011; Andersen & Al-Chalabi, 2011; Pasinelli & Brown, 2006; Saccon et al, 2013). Besides the straightforward involvement of redox imbalances, the molecular mechanism of SOD1-mediated ALS is largely unknown. The disease-linked mutations of SOD1 do not usually impair its function, while misfolding associated with mutation of proteins appears to gain toxic properties leading to neurodegeneration. Misfolded mutant SOD1 proteins accumulate in motor neurons and glial cells of the spinal cord, which mediates excitoxicity, mitochondrial dysfunction, axonopathy, and endosomal trafficking that have been implicated in ALS-SOD1 (Bosco et al, 2010). Patients with ALS-SOD1 have neuronal loss in the anterior horn region of the spinal cord and positive p62 skein-like inclusions that are negatively immunoreactive for TDP-43. P62 is a protein degradation marker related specifically to the selective autophagy system; p62-positive inclusions are widely found in TDP-43-negative-star-shaped-inclusions found in ALS cases. Furthermore, ALS-SOD1 is less frequently associated with ALS-FTD (Tan et al, 2007).

C9orf72

In 1991, a multigenerational family with a history of ALS, FTLD, and ALS-FTLD was described which was later attributed to a mutation consisting of a hexanucleotide expansion repeat (GGGGCC) in the intron between non-coding exons 1a and 1b of the gene C9orf72, which codes for a protein with unknown function (Gunnarsson et al, 1991). The pathological repeat expansion may extend to hundreds of repeats in individuals harboring the genetic lesion, presenting unusual pathological features including RNA foci and aggregates of dipeptide repeats produced from repeat associated non-ATG (RAN) translation of the repeat expansion RNA, as well as TDP-43 positive aggregates. C9orf72 is the predominant ALS gene, associated with about half of all ALS cases, including ~10% of sporadic and nearly 40% of FALS and FTD cases. Moreover, ALS-C9orf72 involves motor neuronal loss in the anterior horn, which normally overlaps with TDP-43-positive inclusions. Strikingly, these patients develop numerous p62-positive, TDP-43-negative inclusions in the dentate gyrus, neocortex, and cerebellum (Al-Sarraj et al, 2011; Boxer et al, 2011; DeJesus-Hernandez et al, 2011). Furthermore, patients with the C9orf72 hexanucleotide expansion also have RBM45 inclusions and PABP-1 (Collins et al, 2012; McGurk et al, 2014). C9orf72 mutations appear to promote mis-localization of TDP-43 to the cytoplasm, in addition to causing malformed RNA molecules (Zhang et al, 2015). C9orf72 repeat expansion partly contributes to the high prevalence of ALS cases in Kii peninsula of Japan, where 20% of ALS patients carry the hexanucleotide repeat expansion (Ishiura et al, 2012).

Optoneurin (OPTN)

OPTN was firstly linked to ALS in 2010 (Maruyama et al, 2010) and its mutations that have been identified in ALS patients, including truncation and missense mutations are thought to have a loss-of-function pathology. OPTN is localized in cytoplasmic inclusions with ubiquitin and TDP-43 in ALS affected motor neurons (Hortobagyi et al, 2011). There are conflicting studies about OPTN co-localization with SOD1 and FUS (Deng et al, 2011a; Hortobagyi et al, 2011; Keller et al, 2012). OPTN inclusions are present in several other neurodegenerative diseases (Osawa et al, 2011).

Valosin-containing protein (VCP)

In 2010, mutations in VCP gene, coding for AAA+ATPase ubiquitn-dependant segregase, were linked to 1-2% of familial cases of ALS (Johnson et al, 2010). Recent studies demonstrated that mutations in VCP cause mitochondrial dysfunction that leads to reduced ATP production. The discovery of mutations in VCP gene led to the model of multisystem proteinopathies, where multiple tissues are affected (Watts et al, 2004). Although, VCP inclusions have been reported in ALS cases, and there are no reports on its co-localization with FUS nor TDP-43 inclusions in ALS cases, TDP-43 was observed to be co-localized with VCP in the cytoplasmic inclusions of FTLD patients, and its translocation between nucleus and cytoplasm is altered by mutations in VCP gene(Gitcho et al, 2009).

Ubiquilin 2 (UBQLN2)

UBQLN2 was linked to ALS in 2011, it was shown that mutations in UBQLN2 gene caused a dominantly male-to-male inherited form of ALS, with or without dementia (Deng et al, 2011b). Given that UBQLN2 regulates degradation of ubiquitinated proteins, and dysregulation in the ubiquitin-proteasome system (UPS) has been linked to ALS; the mutations in UBQLN2 causing ALS was believed to through impaired protein degradation pathways. Besides, ALS patients with mutations in UBQLN2, present inclusions and co-localized with ubiquitin, p62, TDP-43, FUS, OPTN but there are no reports on co-localization with SOD1 aggregates (Deng et al, 2011b; Williams et al, 2012).

ATXN2

the association of extended polyQ repeats in ATXN2 gene with ALS was observed in 2010; Normally ATXN2 carries 21 to 22 polyQ repeats, whereas in ALS cases ATXN2 polyQ repeats extends to 27 to 33 (Elden et al, 2010). ATXN2 functions in mRNA polyadenylation, miRNA synthesis and stress granules formation. ATXN2 cytoplasmic inclusions have been observed in spinal cord tissue of ALS patients. Furthermore, ATXN2 and TDP-43 inclusions co-localized in FTLD patients, while ATXN2 and FUS co-localization was observed in ALS patients (Elden et al, 2010; Farg et al, 2013). In ALS patients with ATXN2 polyQ repeats, ATXN2 and PABP-1 have been observed to colocalize in inclusions (McGurk et al, 2014). ATXN2 polyQ repeats enhanced stress-induced caspase-3 activation and TDP-43 cleavage as well as its phosphorylation (Elden et al, 2010; Farg et al, 2013).

HnRNPA2B1 and hnRNPA1

Hetereogeneous ribonucleoproteins harbor prion-like domains (PLD’s) are essential to assembly into self-seeding ribonucleoprotein granules or fibrils. Mutations in hnRNPA1 and hnRNPA2B1 has been observed in cases of ALS with increased tendency of self-seeding fibrils formation and stress-granule formation (Kim et al, 2013). It is well known that hnRNPA1 and hnRNPA2B1 interact with TDP-43 via its C-terminal domain, co-localization of these proteins with TDP-43 in cytoplasmic inclusions has not been observed in ALS cases (Honda et al, 2015).

NEK1

In a recent exome-wide study, the mutation of a gene that encodes the serine/threonine kinase NIMA (never in mitosis gene-A)-related kinase, NEK1, was found in~ 3% of ALS cases in European and European-American families (Kenna et al, 2016). It is worth to note that NEK1 is an established DNA damage response factor,(Chen et al, 2011)further studies may needed to focus on NEK1 mutation mediated DNA damage repair deficiency in in ALS pathology.

In summary, although the ALS group of motor neuron diseases exhibit common phenotypes, the molecular mechanisms underlying motor neuron dysfunction appear to be distinctive among disease subtypes. This review will focus primarily on the deregulation of RNA/DNA metabolism involved in ALS-TDP-43 and ALS-FUS, as well as emerging research in our understanding of complexity associated with loss of functions versus acquired toxicity of TDP-43 and FUS.

2. TDP-43 and FUS: A perspective into neurobiology and nucleic acid binding

2.1. TDP-43

The 414 amino acid-containing TDP-43, belonging to the heterogeneous nuclear ribonucleoprotein (hnRNP) family, is encoded by the TARDBP gene and is highly conserved among human, mouse, D. melanogaster, and C. elegans genomes (Wang et al, 2004). Domain analysis revealed that TDP-43 consists of two RNA recognition motifs, RRM1 and RRM2, and a disordered glycine-rich CTD (Figure 2). Structurally, TDP-43 closely resembles other hnRNP family proteins: hnRNP A1 and hnRNP A2/B1 (Dreyfuss et al, 1993). Functionally, TDP-43 appears to be a multi-tasking protein and is essential for cell survival, due to its involvement in transcriptional repression, pre-mRNA maturation and alternative splicing, mRNA transportation, microRNA biogenesis, interaction with noncoding RNA, autoregulation, and translational regulation of a number of key proteins (Figure 3). The critical importance of TDP-43 in many cellular functions both during development and in adults is underscored by the embryonic lethal phenotype of homozygous knock-out mice and the fact that its postnatal knock-out, through conditional gene inactivation, causes rapid lethality (Shan et al, 2010). Originally discovered as a transcriptional repressor binding to TAR DNA of the human immunodeficiency virus type 1 (HIV-1) (Ou et al, 1995), TDP-43 was subsequently found to repress the transcription of mouse SP-10 gene (Abhyankar et al, 2007) and human cyclin-dependent kinase 6 (Cdk6) (Ayala et al, 2008a) by binding to the regulatory elements of the respective promoters. In this context, the association of TDP-43 with chromatin, possibly through its RRM2 motif, both in the human brain and in cultured cells, is highly significant (Ayala et al, 2008b; Casafont et al, 2009; Thorpe et al, 2008).

Figure 2.

Schematic of TDP-43 and FUS protein structure. TDP-43 and FUS, both have a Prion-like domain, nuclear localization signal, nuclear export signal and RNA recognition motif (RRM). FUS has an additional RRM as well as a Zinc finger domain. Major familial mutations are indicated. In contrast to FUS, disease-linked TDP-43 mutations are clustered in the Glycine-rich C-terminal domain; whereas FUS disease-linked mutations are mainly clustered in the nuclear localization signal domain.

Figure 3.

TDP-43/FUS in healthy neurons bind to thousands of cellular RNA’s. They shuttle between the nucleus and cytoplasm, and play roles in miRNA biogenesis, pre-mRNA splicing, mRNA stability and transport. ALS affected motor neurons present altered cytoplasmic localization and nuclear clearance of the TDP-43 and FUS, together with deregulation in their posttranslational modification states, impacting their normal functions.

TDP-43 interacts with both UG-rich and non-UG-rich RNA sequences using its RRM domains. TDP-43 binds both single-stranded (ssDNA) and double-stranded DNA (dsDNA), with higher affinity to ssDNA at the TG repeat sequences. RRM1 may be involved in its interaction with DNA, since this region co-crystallizes with TG-rich ssDNA, forming thermostable dimeric assemblies (Kuo et al, 2014). RRMs contain two highly conserved sequences, namely RNP1 and RNP2. Residues Phe194 (conserved in RNP1) and Phe231 (conserved in RNP2) are involved in the interaction of TDP-43 with different UG- or TG-rich nucleic acid sequences (Buratti & Baralle, 2001). Mutational modification of Phe147 and Phe149 of RNP1 has been found to be sufficient to abolish the splicing regulatory functions of TDP-43 (Buratti & Baralle, 2001; Buratti et al, 2001; D’Ambrogio et al, 2009). The RRM2 domain is structurally distinct from the RRM1 domain. Interestingly, the RRM1 motif prefers to bind long (UG)₆ repeats, whereas RRM2 has a stronger binding affinity for short (UG)₃ repeats, suggesting that both the RRM domains are essential for normal cellular functions of TDP-43 (Kuo et al, 2009). TDP-43 also binds (UG)_nUA(UG)_m and poly-pyrimidine tract-containing RNA/DNA sequences (Sephton et al, 2011; Xiao et al, 2011). TDP-43 forms a homodimer with a domain arrangement similar to hnRNPA1, where the two RRM2 domains form a highly thermostable β4 strand. Thus, it is possible that in addition to its prion-like glycine-rich domain (277-414 amino acids), the RRM2 domain may also contribute to TDP-43 proteinopathy (Kuo et al, 2014).

Through its RNA-splicing functions, TDP-43 controls the expression levels and splice variants of many targets. TDP-43 autoregulates its own protein levels, as well as regulates the expression levels of other RNA-binding protein genes, including SRSF1, polypyrimidine tract-binding (PTB), and hnRNP L. (Buratti & Baralle, 2011). A UV-cross-linking immunoprecipitation-sequencing (CLIP-seq) study in mouse brain tissue showed that TDP-43 interacts with over 6,300 RNAs. The depletion of TDP-43 resulted in altered splicing of over 900 mRNAs (Polymenidou et al, 2011). The amount of TDP-43 protein in healthy cells is tightly maintained, possibly due to nonsense-mediated decay of mRNA (Lejeune & Maquat, 2005), and is likely to be critical for cellular homeostasis. TDP-43 CTD region 321-366 aa binds to the conserved 3′UTR proximal polyA₁ and distal polyA₄ sequences in its own TARDBP mRNA to generate its two major splice variants of 2.8kb (using intron 7 only) and 4.2kb (using both intron 6 and 7) in a ratio of ~1:3 (Ayala et al, 2011; D’Ambrogio et al, 2009). Overexpression of TDP-43 causes asymmetric interactions between TDP-43 and its nascent RNA to stall RNA polymerase II (RNA Pol II), leading to transcript degradation and maintaining the autoregulation of TDP-43 protein level (Avendano-Vazquez et al, 2012). Interestingly, TDP-35, a second splice variant of TDP-43, has been found to be expressed in the brains of ALS patients, through the use of the downstream start codon ATG^Met85 (Xiao et al, 2015). Expression of TDP-35 in primary neurons causes cytoplasmic aggregation and neuronal death (Xiao et al, 2015).

Recent studies found that TDP-43, but not FUS, is a component of nuclear Drosha complex, which is involved in microRNA regulation (Kawahara & Mieda-Sato, 2012). Cytoplasmic TDP-43 associates with the Dicer complex that contains transactivation-responsive RNA binding protein (TRBP), which recruits Argonaute 2 (Ago2) for efficient processing of pre-miRNAs (Chendrimada et al, 2005; Gregory et al, 2005). These studies indicate that TDP-43 can also affect gene regulation in complex with microRNA (miRNA). For example, TDP-43 forms a complex with miR-NID1, generated from intron 5 of NRXN1 suppressing the expression of NRXN1, which is crucial for synaptic vesicle release and maintenance of synapse ultrastructure (Kang et al, 2015).

Long non-coding RNAs (lncRNAs) confer another level of gene regulation by binding to complementary mRNA sequences during post-transcriptional processing (Beltran et al, 2008). It was recently discovered that growth-arrested DNA damage-inducible gene 7 (gadd7), a 754 nucleotide (nt)-long lncRNA induced by both growth arrest and DNA damage signaling, regulates Cdk6 expression by specifically binding to and dissociating TDP-43 from Cdk6 mRNA and directing it for decay (Liu et al, 2012). This implicates the direct interaction of TDP-43 with lncRNA and its involvement in cell cycle regulation. In FTLD-TDP-43 brains, TDP-43 showed highly increased binding to metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) and nuclear enriched abundant transcript 1 (NEAT1) non-coding RNAs. In disease conditions, the binding affinity of TDP-43 increases for deep intronic regions, downstream of silenced exons (Tollervey et al, 2011).

Global attenuation of protein synthesis and induction of enhanced protein quality control are the most important and immediate cellular responses following stress. Activated protein quality control machinery removes the damaged proteins and simultaneously activates molecular chaperones in order to improve protein folding in the endoplasmic reticulum lumen (Holcik & Sonenberg, 2005; Ron & Walter, 2007). In response to stress, TDP-43 translocates from the nucleus to cytoplasm, where it binds 14-3-3 proteins. This relieves the negative inhibition of FOXO transcription factors, allowing their translocation back to the nucleus to regulate expression of genes involved in the stress response (Salih & Brunet, 2008; Zhang et al, 2014). Upon exposure to oxidative stress, cells form cytoplasmic granular particles (≤5 μm) known as stress granules (SGs). SGs are RNA-containing cytoplasmic particles composed by RNA and RNA-binding proteins involved in RNA metabolism and translation. SGs have been linked to several neurodegenerative diseases, including ALS and FTLD (Volkening et al, 2009). Although the precise components of SGs, which appear to be dynamic, are not completely understood, some of the key components that have been identified include TDP-43, T-cell intracellular antigen 1 (TIA-1) and RasGAP SH3-domain binding protein 1 (G3BP). SG formation is a rapid and transient mechanism, which starts immediately after stress exposure and dissembles within a couple of hours (Gilks et al, 2004; Kedersha & Anderson, 2002; McDonald et al, 2011; Tourriere et al, 2003). Efficient SG kinetics and dynamics are critical to neuronal cell survival to respond to both acute and chronic OS, where TDP-43 plays a major role (Aulas et al, 2012).

2.2. FUS/TLS

Following the discovery of TDP-43, mutations in a second RNA/DNA-binding protein FUS/TLS was also implicated in ALS and FTLD. Located on chromosome 16 (Aman et al, 1996), the FUS gene encodes a 526 amino acid protein that belongs to the FET/TET family. Proteins in this family are defined by the presence of an N-terminal transcription activation domain SYGQ-rich region, a C2/C2 zinc finger (ZnF) motif, and one or more arginine-glycine-glycine (RGG)-repeat sequences (Morohoshi et al, 1998). The N-terminal of FUS contains a SYGQ-rich region, followed by a RGG box (RGG1), an RRM motif, another RGG box (RGG2), a ZnF motif and an additional RGG box (RGG3). The C-terminus contains a nonclassical nuclear localization signal (NLS) with conserved proline and tyrosine residues (PY-NLS) (Burd & Dreyfuss, 1994; Dormann & Haass, 2013; Iko et al, 2004; Lanson & Pandey, 2012; Lee et al, 2006; Morohoshi et al, 1998; Prasad et al, 1994; Zinszner et al, 1994)(Figure 2). In addition, bioinformatics analysis identified two PLDs in FUS, localized in its N-terminal (residues: 1-239) and RGG2 box (residues 391-405), respectively (Cushman et al, 2010; Gitler & Shorter, 2011). PLDs are regions enriched in polar aminoacids commonly found in RNA-binding proteins; it is hypothesized that PLD drives protein aggregation in neurons. PLDs in RNA binding proteins are essential for the protein to adopt a functionally aggregated state into ribonucleoprotein granules; but the relation between physiological function and disease is not well understood (King et al, 2012).

FUS is able to bind with RNA, ssDNA, and potentially with dsDNA (Tan & Manley, 2009). The RGG-ZnF-RGG domain has been suggested as the major RNA-binding sequence, with a preference for GGUG motifs.

While FUS is ubiquitously expressed in both the nucleus and cytoplasm in many cell types, it is predominantly nuclear in glial cells and neurons (Andersson et al, 2008). FUS was originally identified as an oncogene fused with transcription repressor C/EBP homologous protein 10 (CHOP) in human malignant myxoid liposarcomas (Rabbitts et al, 1993). FUS was also identified as an activator of ETS-related gene (ERG) in acute myeloid leukemia (Ichikawa et al, 1994; Panagopoulos et al, 1994) and in Ewing’s sarcoma tumors (Shing et al, 2003) by chromosomal translocation via its N-terminus, which was later shown to possess potent transcriptional activity (Prasad et al, 1994; Zinszner et al, 1994). Subsequent studies have illuminated the transcriptional regulatory role of FUS in global or specialized components of transcriptional machinery. For example, under specific conditions, FUS may have a role in activating transcription of certain nuclear hormone receptor by interacting with their DNA-binding domain (Fay et al, 1998; Tan & Manley, 2009). Recently, FUS was also shown to affect the transcription of RNA Pol III (Tan & Manley, 2010). Additionally, FUS affects gene expression by acting as co-regulator of several transcription factors, including nuclear hormone receptors (Powers et al, 1998), Spi-1/PU.1 (Hallier et al, 1998), NF-κB (Uranishi et al, 2001), and RUNX transcription factors (Li et al, 2010). The involvement of FUS in RNA splicing by binding splicing regulator or pre-mRNA was widely investigated as well (Sama et al, 2014). FUS was identified as component of the hnRNP complex assembled on the Adeno pre-mRNA (Cinzia Calvio, 1995, RNA, 1), and subsequently as a partner of SR family splicing factors, including SRm160, SRp75, and PTB (Calvio et al, 1995). A recent study has shown that FUS is able to directly bind to thousands of pre-mRNAs, with a preference for long introns compared to exons and 3′UTRs. In addition to transcription and splicing, FUS also plays a role as an mRNA transporter between the nucleus and cytoplasm. FUS displays a bidirectional transport in neuronal dendrites to shuttle mRNA into dendritic spines, a critical step for neuronal maturation (Fujii et al, 2005).

Furthermore, due to its ssDNA/dsDNA binding properties, FUS was presumed to be involved in maintaining genomic fidelity and the DNA damage response (DDR), which was subsequently demonstrated (Wang et al, 2013b). FUS was found to be a downstream factor of ataxia-telangiectasia mutated (ATM) and Poly(ADP-ribose) Polymerase (PARP) 1, and an interacting partner of histone deacetylase 1 (HDAC1) in DDR (Baechtold et al, 1999; Gardiner et al, 2008; Mastrocola et al, 2013; Rulten et al, 2014; Wang et al, 2013b). The involvement of FUS in DDR is discussed in detail in later section of this article.

2.3. Crosstalk of TDP-43 and FUS

Both TDP-43 and FUS are involved in multiple RNA metabolic processes (Figure 3). Growing evidence suggests a functional overlap of TDP-43 and FUS (Honda et al, 2013). Tibbetts and colleagues found that TDP-43 interacts with FUS physically in vivo and in vitro, and the interaction relies on the C-terminal Gly-rich and RRM2 domains of TDP-43 (Kim et al, 2010). The interaction is required for coordinately regulating mRNA expression of their common target, histone deacetylase 6 (HDAC6) (Kim et al, 2010). Interestingly, a more recent study by Cleveland and colleagues showed enhanced interaction of ALS-associated TDP-43 mutants Q331K and M337V with FUS in HeLa cells, which was speculated to perturb normal function of FUS (Ling et al, 2010). In animal models, the TDP-43 knockdown phenotype was rescued by overexpression of WT FUS in zebrafish, which suggests a potential backup role for TDP-43 (Kabashi et al, 2011a). In D. melanogaster FUS was shown to function downstream of TDP-43 in neurons, a role required to maintain normal locomotion and regular life spans (Wang et al, 2011). Furthermore, TDP-43 or FUS-silenced primary cortical neurons obtained from mouse fetal brains exhibited significantly overlapping transcriptome profiles. Specifically, around 25% of genes with changed expression levels and around 10% of genes with altered splicing overlapped, which suggests a substantial collaboration of TDP-43 and FUS in mRNA maturation and/or transportation (Honda et al, 2013).

Potential back-up functions of TDP-43/FUS for other RNA/DNA binding proteins like SMN was also evident in several studies. In the nuclei and nuclear gemini of coiled bodies in healthy neurons, FUS together with TDP-43 and SMN have been suggested to cross-talk for spliceosome maintenance (Belly et al, 2005; Wang et al, 2008).

There is evidence showing that transport ribonucleoprotein particles (tRNP) are affected in ALS-FUS as well as in ALS-TDP-43 (Sephton & Yu, 2015). The function of TDP-43 and FUS at synapses is not known, but emerging evidence suggests that these proteins play roles in synapse integrity (Hebron et al, 2013). Furthermore, formation of granules containing TDP-43, FUS and tRNP increases upon stimuli, and genetic deletion results in altered dendritic spines and branches in cultured primary motor neurons (Sephton & Yu, 2015). Overall, although the physiological or pathological significance of the interaction of TDP-43 and FUS remains largely unknown, further investigations to understand the impact of deregulation may shed light on the common molecular mechanisms of ALS-TDP-43 and ALS-FUS.

3. Neurotoxicity of TDP-43/ FUS

3.1. Cytoplasmic aggregation and functional/nuclei-specific loss of TDP-43/FUS

Accumulation of polyubiquitinated, misfolded protein aggregates is a shared feature in most neurodegenerative disorders, such as PD, which presents α-synuclein aggregates, and AD, in which the proteins amyloid beta peptide and MAP Tau form aggregates. TDP-43 was first described in 2006 as a major component of insoluble cytoplasmic aggregates found in spinal cord and brain samples from patients with ALS and FTLD (Neumann et al, 2006). Cytosolic TDP-43 aggregates were also identified in muscle tissue of patients with inclusion body myositis and Alexander disease (Salajegheh et al, 2009; Walker et al, 2014).

In ALS, TDP-43 is fragmented at its CTD into ~35 to ~25 kDa polypeptides that form cytoplasmic insoluble inclusions. Although the molecular triggers driving TDP-43 aggregation in spinal motor neurons is not clearly understood, in vitro studies have implicated several critical factors that can induce TDP-43 cleavage and aggregation (Li et al, 2015). TDP-43 can also form dimers that are enriched in cytosolic fractions. The expression of C-terminal fragments (CTFs) in cultured cells generally reproduces the aggregation pathology found in patients. Both the CTD and N-terminal domain (NTD) of TDP-43 can form aggregates as long as they contain the C-terminal end of the RRM2 domain, a β-sheet-rich region capable of forming stable dimers (Yang et al, 2010). Moreover, aggregation of fragments can seed full-length TDP-43 aggregation, reducing nuclear TDP-43. The crystal structure of C-terminal RRM2 revealed that it folds into a structure composed of two α-helices and five β-sheets. Cell culture studies have shown that RRM2 is required for aggregation, as cleaving at these sites removes the β-strand and α-helix of RMM2, preventing native folding (Zhang et al, 2009). This observation raises the question of whether or not RRM2 plays a role in TDP-43 aggregation via misfolding. Researchers studying the domain assembly of TDP-43 found that TDP-43 without the C-terminal tail is capable of forming homodimers via its NTD. It has also been shown that truncated RRM2 in the glycine-rich region forms fibrils in vitro, similar to those found in disease models, suggesting that RRM2 plays a role in the formation of cytoplasmic inclusions of TDP-43 (Zhang et al, 2009). Inclusions within brain samples of ALS patients labeled with antibodies recognized the C-terminus of TDP-43, but not the N-terminus, suggesting that inclusions are predominantly comprised of CTFs. Inclusions in spinal cord samples of ALS patients stained positive for both C-terminal and N-terminal fragments, suggesting that inclusions contain full-length TDP-43 (Zhang et al, 2009). Furthermore, TDP-43 dimer complex was identified in human brain tissues, suggesting that dimerization is a feature of TDP-43 proteinopathies (Shiina et al, 2010).

Reactive oxygen species (ROS) and increased OS markers have been reported in patients with ALS, which are critical factors linked to TDP-43 aggregation (LoGerfo et al, 2014; Nagase et al, 2015; Niedzielska et al, 2015). A marker of OS, 4-hydroxynonenal (HNE), is found to be elevated in the spinal cord, motor cortex, cerebrospinal fluid, and serum of SALS cases (Pedersen et al, 1998; Simpson et al, 2004). HNE induces aggregation and mis-localization of TDP-43 in cultured cells (Kabuta et al, 2015). Cysteine residues are strongly involved in TDP-43 insolubility induced by HNE. Furthermore, it has been reported that cysteine residue-mediated oxidation and disulfide crosslinking alters TDP-43 solubility and impairs nuclear function. This reversible alteration of TDP-43 solubility occurs via formation of disulfide bonds through the highly conserved cysteine residues in TDP-43. These TDP-43-disulfide crosslinks are detected in control brain samples, suggesting that disulfide crosslinking occurs in normal healthy brains in response to OS (Cohen et al, 2012). In addition, OS also promotes TDP-43 acetylation, which consequently drives TDP-43 aggregation (Cohen et al, 2015). Acetylated TDP-43 (Lys-145) is found in ALS spinal cord samples (Cohen et al, 2015). Altered metal homeostasis has also been linked to TDP-43-associated neurodegeneration. Zinc, but not copper and iron salts, promotes TDP-43 nuclear depletion and formation of cytoplasmic inclusions in cultured neurons (Zinszner et al, 1994), suggesting that TDP-43 nuclear depletion is a specific consequence of increased Zn levels (Caragounis et al, 2010).

Chronic immune activation is a common feature in motor neuron diseases. It has been reported that TDP-43 mRNA and protein levels are increased in the spinal cord of ALS patients (Swarup et al, 2011b), as well as elevated levels of activated transcription factor NF-κB, which is involved in a large number of normal cellular processes, including immune and inflammatory responses. It was found that TDP-43 and NF-κB p65 interact in glial cells and neuronal cells of ALS patients, as well as in transgenic mice overexpressing human TDP-43 (Philips & Robberecht, 2011; Swarup et al, 2011b). NF-κB is a key component of the immune response and has an emerging role in ALS and other neurological disorders. It has been reported that in cell culture, exposure of microglia and astrocytes to LPS induces cytoplasmic redistribution of TDP-43 (Correia et al, 2015). Additionally, NF-κB activation increases cytoplasmic levels of TDP-43, suggesting that chronic brain inflammation can drive cytoplasmic aggregation of TDP-43 (Correia et al, 2015).

TDP-43 aggregation has been widely related to TDP-43 toxicity. However, in vitro reduction of TDP-43 aggregation by small molecules/natural compounds effectively reduced TDP-43 aggregates in ALS cell models and yeast models, but did not reduce or prevent cell death (Liu et al, 2013). This suggests that while TDP-43 aggregation is a cytotoxic effect of TDP-43 proteinopathies, its prevention alone is not sufficient to rescue neurons from degeneration and death. Jackson et al., recently showed, that expression of upframeshift protein 1 (UPF1) could rescue motor functions in ALS-like paralysis rat model based on TDP-43 overexpression. Moreover, expression of UPF1 in yeast and neuronal cell culture reduces TDP-43 cytotoxicity (Jackson et al, 2015).

Since the first report that linked a number of mutations of FUS with patients with FALS, who also developed cytoplasmic inclusions in spinal motor neurons, a pathology similar to TDP-43 (Kwiatkowski et al, 2009; Vance et al, 2009), subsequent studies identified FUS-positive cytoplasmic inclusions not only in FALS but also in a small number of SALS patients (Deng et al, 2010). Cytosolic FUS is recruited into reversible SGs (Dormann & Haass, 2011; Li et al, 2013). Furthermore, SG markers have been observed in the inclusions from patients with FUS linked ALS or FTLD (Deng et al, 2014a). In addition, ALS associated FUS mutants show increased association with SGs compared to WT FUS, mutated FUS is able to bind and sequester WT FUS into SGs, which suggests a direct pathological role of SGs in ALS. Some researchers believe that although FUS accumulation in SGs is a reversible process in healthy neurons, under chronic stress, it could lead to pathological aggregation of FUS in SGs (Ling et al, 2013). A second school of thought suggests that recruitment of FUS into SGs prevents irreversible aggregation of FUS mis-localized to the cytoplasm (Shelkovnikova et al, 2013b). Recent study in a mouse model with post-natal elimination of FUS, shows that FUS has no effect in survival of motor neurons or function, suggesting that FUS-dependent neurodegeneration is likely to be due to a gain of toxicity (Sharma et al, 2016).

3.2. Neurotoxicity of TDP-43/FUS mutations

Since the discovery of TDP-43 as the major protein found in ALS cytoplasmic aggregates, several genetic screens have been conducted to identify TARDBP gene mutations (http://alsod.iop.kcl.ac.uk/Overview/gene.aspx?gene_id=TARDBP). Three separate research groups sequenced the entire coding region of the TARDBP gene and identified mutations associated with ALS (Gitcho et al, 2008; Kabashi et al, 2008; Sreedharan et al, 2008), which were mostly substitutions in serine and threonine residues. Cell and biochemical studies of these mutations provided evidence linking TDP-43 abnormalities and neurodegeneration. In an extensive survey of patients diagnosed with ALS, (80 FALS cases and 120 SALS cases), eight heterozygous mutations were identified, with the A315T mutation found to be the most common (Kabashi et al, 2008). The M337V mutation was identified in a British family, and two other missense variations G294A and Q331K, were reported by the same investigators (Sreedharan et al, 2008).

Since then, 47 missense mutations have been identified. M337V is one of the most common mutations in patients who presented with upper-limb early onset motor neuron disease. Recent studies reported increased levels of full-length and truncated TDP-43 in differentiated neurons derived from induced pluripotent stem cells (iPSC) from ALS patient lymphoblasts carrying the M337V mutation (Rutherford et al, 2008). The same group reported additional mutations, including two previously unknown mutations (N345K, I383V). Human lymphoblastoid cells expressing N345K, I383, and M337V presented with increased caspase-cleaved ~25 kDa fragments in cytoplasmic aggregates, suggesting a novel toxic gain-of-function through protein-protein interactions or accumulation of TDP-43 fragments leading to apoptosis (Gendron et al, 2013). Other studies identified several other mutations (G348C, R361S, N390D, N390S) that also presented increased insolublê25 kDa TDP-43 fragments (Daoud et al, 2009; Del Bo et al, 2009). TDP-43 spontaneously forms aggregates; mutations within the CTD (Q331K, M337V, Q343R, N345K, R361S, N390D) increase the number of aggregates, promote toxicity in vivo, and accelerate aggregation of recombinant TDP-43 in vitro. In addition, a mutation causative of ALS and FTLD (A328T) has also been found in patients with PD, as well as in patients with FTD with Parkinsonism (Borghero et al, 2011; Cannas et al, 2013).

D. melanogaster lacking TBPH (the TDP-43 ortholog) die as pupae, and rare survivors present with synaptic dysfunction of motor neurons, reduced lifespan, and progressive motor neuron loss. These features can be rescued by expression of human TDP-43. In contrast, introducing M337V or A315T mutants fails to rescue motor neuron degeneration. (Feiguin et al, 2009; Fiesel et al, 2010; Kabashi et al, 2011b).

Studies have shown that overexpression of TDP-43 results in reduced mitochondrial density in neurites of primary motor neurons (Cozzolino & Carri, 2012). This condition is exacerbated by mutant TDP-43 Q331K/M337V overexpression. In contrast, suppression of TDP-43 results in increased mitochondrial density (Wang et al, 2013a). Also, TDP-43 co-localizes with mitochondria in motor neurons, and this co-localization is enhanced by mutant expression, suggesting that TDP-43 has a role in regulating mitochondria dynamics. TDP-43 depletion results in increased abnormal neurites and decreased cell viability. TDP-43 mutants A315T/M337V/Q331K mis-localize to the cytosol and show abnormal neurites. Cytosolic expression of TDP-43 with mutated NLS variant also showed abnormal neurite morphology and reduced cell viability (Gitcho et al, 2008; Han et al, 2013; Wang et al, 2013a).

TDP-43 mutations affect the dynamics of SG assembly by increasing the propensity for SG assembly in the presence of sodium arsenite, a potent OS inducer. This also correlates with mutant TDP-43’s decreased nuclear localization in response to sodium arsenite (McDonald et al, 2011). Mutations disrupt cytoplasmic SGs that contain translationally silenced RNA transported to target sites, enabling the cell to carry out protein synthesis. TDP-43 mutants may also be incorporated in the early stage of SG formation, resulting in larger and deregulated granules (Dewey et al, 2011; McDonald et al, 2011).

The pathogenic mechanisms of TARDBP mutations are still unclear; mutations most likely affect normal functions of TDP-43 by gain-of-toxicity or loss-of-function, mediated by enhanced aggregation and nuclear depletion. Alterations in TDP-43 function have deleterious effects, including impaired RNA metabolism and generation of toxic byproducts. The involvement of TDP-43 pathology in ALS vs. FTLD is an interesting topic, which is still not completely understood. While the location/distribution of TDP-43 aggregates differ between ALS (spinal cord) and FTLD (more widespread in the brain including frontal and temporal lobes), the familial mutations in TARDBP are unique to ALS and are not found in other neurodegenerative disorders including FTLD (Blokhuis et al, 2013; Brouwers et al, 2010; Rutherford et al, 2008; Van Deerlin et al, 2008). However, some of the common TARDBP mutations, presumably involving vulnerable residues for somatic mutations were reported in a small number of sporadic FTLD cases (Benajiba et al, 2009; Borghero et al, 2011; Borroni et al, 2009; Chiang et al, 2012; Chio et al, 2010; Corrado et al, 2009; Kovacs et al, 2009). Similar scenario may occur in other neurodegenerative diseases involving TDP-43 pathology such as inclusion body myositis and Alexander disease. Further investigations are required to understand how these sporadic diseases with TDP-43 pathology manifest in presence of somatic TARDBP mutations.

Numerous mutations of FUS have also been reported since 2009, and to date, over 50 mutations/deletions have been identified to account for ~4% FALS and rare SALS cases. Some mutations are associated with early-onset disease; for example, ALS patients exhibiting the P252L mutation are often a very young age at onset (mean: 23.7 years old) and present rapid disease progression (patients die within 12 months) (Chio et al, 2009b; Lagier-Tourenne & Cleveland, 2009). Compared to ALS, only a few FUS mutations were discovered in association with FTLD (Broustal et al, 2010). The R521H mutation was identified in a patient initially diagnosed with behavioral disorders, but rapidly developed ALS (Broustal et al, 2010). Meanwhile, R521C, G156E mutations and G174-G175 deletion were identified in a FTLD patient who was concurrently diagnosed with motor neuron disease (Blair et al, 2010; Ticozzi et al, 2009; Yan et al, 2010). M254V, P106L mutation and Gly174-Gly175 deletion were discovered in patients with FTLD (Huey et al, 2012). These cases suggest that ALS and FTLD share similar pathology and genetics.

Unlike TDP-43, in which the majority of mutations cluster in the C-terminal prion-like region, over 50% of ALS-related FUS mutations have been discovered in the C-terminal NLS domain, implying a possible involvement of its nuclear import defects. Dormann et al., showed that the last 13 amino acids (514-526) within the NLS are necessary, but not sufficient for nuclear import of FUS. In this study, FALS-associated point mutations occurred in the C-terminal region at evolutionarily conserved residues including R521 (G/H/C), R522 (G), R524 (S), and P525 (L). Immunostaining of HeLa cells expressing the above mutants showed that the P525L mutant induces dramatic nuclear import defect, whereas R522G showed a moderate phenotype. Interestingly, although the ALS-linked R521 mutation is in a highly conserved amino acid among vertebrates, is the most frequently mutated residue in FUS, the R521G/H/C mutant shows only mild cytoplasmic localization, which suggests that the mutation of R521 may result in pathological dysfunctions of FUS beyond cytosolic accumulation. The cytoplasmic accumulation of FUS P525L was subsequently observed in neurons from rat hippocampus and frontal cortex, as well as in zebrafish eggs in the same study (Dormann et al, 2010); these results were later confirmed by other groups (Niu et al, 2012; Zhou et al, 2013). As mentioned above, FUS is recruited into SGs. The FUS R495X truncation mutant (a deletion of the last 32 C-terminal amino acids of FUS), which exists in early-onset ALS, was identified by Hayward and colleagues (Bosco et al, 2010). Also, an experimental mutant of G515X (a deletion of the terminal 12 C-terminal amino acids of FUS, which includes the main cluster of ALS-linked mutations) co-localizes with SGs in HEK-293 cells extensively (in ~70-80% cells) within 4-7 min of exposure to arsenite. However, the recruitment of the R521G mutant into SGs shows significant delay and lower intensity. The same phenomenon was subsequently observed in zebrafish embryos in response to heat shock (Bosco et al, 2010). Other ALS-FUS mutants were also reported to be associated with SGs; for example, R521C co-localized with SGs in zebrafish motor neurons under the treatment of heat shock and arsenite (Acosta et al, 2014), and R514S or P525L mutants formed SGs in HeLa cells. Also, double mutations of R514S and P525L, or triple mutations of R514S, R521C, and P525L, additively enhanced SG formation (Ito et al, 2011). A detailed analysis was conducted by Suzuki and colleagues to identify functional domains that contribute to SG formation. In this study, a series of FUS deletion constructs of conserved regions, including N-terminus, SYQG-rich domain deletion (ΔNT), Gly-rich region deletion (ΔGRR), NES deletion (ΔNES), RNA recognition motif deletion (ΔRRM), first RGG box deletion (ΔRGG1), ZnF motif deletion (ΔZnF), secondary RGG box deletion (ΔRGG2) and C-terminal deletion (ΔCT), as well as WT FUS were developed and overexpressed in HeLa cells. This experiment revealed that ΔCT formed SGs extensively, whereas ΔRGG2 formed few SGs, indicating that the C-terminus plays a critical role in SG formation (Ito et al, 2011). The assembly of SGs is an emergency response for cells to survive by limiting the translation of non-essential mRNAs and focusing on producing essential proteins under stress. The observation that TDP-43 and FUS accumulate in SGs in vitro and in vivo supports its pathological significance. Several models, including a gain-of-toxicity model, loss-of-function in SGs model, and nuclear loss-of-function model, have been proposed to explain the connection between TDP-43, FUS and SGs in ALS/FTLD pathogenesis, which need to be elucidated by further investigation (Li et al, 2013).

The cytoplasmic retention of ALS-associated mutant FUS and its role in deficient RNA processing were confirmed by multiple studies. Hicks and colleagues have shown that WT FUS, but not ALS-linked FUS mutants R521G, R522G, and Δ515 (deletion of the last 12 residues in the C-terminus), are able to directly modulate the alternative splicing of exon 7 to autoregulate protein levels, and the ability to regulate splicing is tightly correlated with nuclear retention (Zhou et al, 2013). Another study showed that HEK-293T cells overexpressing FUS with the R521G or R522G mutation displays a globally-affected transcription pattern, which resembles the effect of overexpression of WT FUS but the effects of siRNAs targeting FUS (van Blitterswijk et al, 2013). Intranuclear aggregation of FUS induced by the pathogenic mutation G156E was also proposed as an alternative pathogenic mechanism of FALS. The FUS G156E mutation occurs in an evolutionarily conserved region in mammals, reptiles, and zebrafish. Clinically, a patient with the G156E mutation presented with both upper and lower motor neuron signs concurrently with dementia, and developed FTD in his fourth decade (Ticozzi et al, 2009). Nomura et al. showed that the G156E mutant increases propensity of FUS for aggregation in vitro and in vivo, which supports the model of ‘seeding reaction’, in which the aggregated protein fibrils serve as a structural template facilitating the fibrillation of unaggregated protein. Consistently, FUS G156E caused a rapid progression of ALS after its onset (Nomura et al, 2014).

3.3. Deregulation of post-translational modifications

3.3.1. Abnormal phosphorylation

Hyperphosphorylated TDP-43 has been consistently identified in the cytoplasmic inclusions of spinal motor neurons in both ALS patients and model systems. Serines 409 and 410 are the most consistently phosphorylated sites in disease, although S379, S403 and S404 are also observed in disease pathology (Hasegawa et al, 2008; Neumann et al, 2009). The functional role of TDP-43 phosphorylation is not clear, and a correlation between protein insolubility and TDP-43 phosphorylation has been reported (Neumann et al, 2006). Phosphorylated TDP-43 exhibits a longer half-life than the non-phosphorylated form. This may be due to changes in UPS-mediated degradation or increased protein stability. Mutating serine residues 409 and 410 of TDP-43 to non-phosphorylatable alanine decreased TDP-43 neurotoxicity in C. elegans models (Liachko et al, 2010; Liachko et al, 2013). There are a number of kinases identified that control TDP-43 phosphorylation, including CK1/2, CDC7, and TTBK1/2 (Hasegawa et al, 2008). TDP-43 can be directly phosphorylated by CK1/2, and expression of hyperactive CK1 promotes the accumulation of cytoplasmic insoluble phosphorylated TDP-43 (Nonaka et al, 2016). The kinases CDC7 and TTBK1/2 have also been shown to directly phosphorylate TDP-43 in vitro. Pathologically, CDC7 and phospho-TDP-43 co-localize in disease-affected neurons (Liachko et al, 2013). Furthermore, a small molecule inhibitor of CDC7 reduced phosphorylation and neurodegeneration in C. elegans model and also limited TDP-43 phosphorylation in NSC-43 cells (Liachko et al, 2013). TTBK1 and TTBK2 protein levels are elevated in motor cortex of FTLD, and the kinases co-localize with TDP-43 inclusions in ALS spinal cord (Liachko et al, 2014). TDP-43 phosphorylation is not required for its cleavage, aggregation or toxicity in cell culture systems.

Phosphorylation of FUS was recently identified. Serine 42 was identified as the phosphorylated serine in WT, but not in CHOP-fused FUS by ATM, in response to ionizing radiation (IR)-induced genomic double-strand breaks (DSBs) (Gardiner et al, 2008). In another study, DNA-dependent protein kinase (DNA-PK) mediated multiple phosphorylation on the N-terminus of FUS, and this phosphorylation facilitated its cytoplasmic accumulation upon cellular DNA damage (Deng et al, 2014b). The association of FUS phosphorylation with pathogenesis in neurodegeneration was explored by Nukina and colleagues, who found that overexpression of mouse phosphomimetic mutation of serine 505 (513 in humans) predicted potential phosphorylation sites for kinases such as CK2, DNA-PK and GSK3, S505D. This phosphorylation can enhance nuclear clearance of ALS-linked mutants including G499D, H509P, R510K, R513G, and R516S in both N2a and COS-7 cell lines, as well as impair the splicing regulatory activity of FUS mutants H509P and R510K (Kino et al, 2011). These data suggest that neurodegeneration initiated by FUS mutants could be exacerbated by phosphorylation.

3.3.2. Ubiquitination

The role of ubiquitination of TDP-43 in the pathogenesis of disease is not clearly understood. Cells dispose of soluble TDP-43 proteins via the UPS, while aggregated TDP-43 appears to be predominantly cleared through autophagy (Scotter et al, 2014). Inhibition of UPS in cell culture increases cytoplasmic accumulation of TDP-43 and formation of intracellular aggregates. Disruption of the UPS contributes to increased levels of ubiquitinated TDP-43 in ALS (Zhang et al, 2010). Ubiquitination of TDP-43 may be a late event in disease pathogenesis because most TDP-43 inclusions are not ubiquitinated in early disease stages (van Eersel et al, 2011). Autophagosome-mediated degradation is also involved in TDP-43 protein turnover. TDP-43 was proposed to interact with UBQLN2, a protein that binds ubiquitinated proteins and transports them to the proteasome for degradation. UBQLN2 also has roles in macroautophagy and chaperone-mediated autophagy. Interestingly, mutations in UBQLN2 have been found in some families with FALS (Deng et al, 2011b). UBQLN2 expression in cell culture and D. melanogaster models promotes TDP-43 autophagosomal degradation (Filimonenko et al, 2007). Recently, the E3 ubiquitin ligase Parkin was proposed to ubiquitinate TDP-43, regulating its subcellular transport (Hebron et al, 2013). In D. melanogaster models, the ubiquitin-conjugating enzyme UBE2E3 promotes ubiquitination of TDP-43; in contrast, ubiquitin isopeptidase Y (UBPY) decreased TDP-43 ubiquitination (Hans et al, 2014). Furthermore, knockdown of UBPY promotes formation of insoluble TDP-43 aggregates and induced neurotoxicity in D. melanogaster (Hans et al, 2014).

While ubiquitination and hyperphosphorylation of TDP-43 are etiologically linked to motor neuron diseases, unlike TDP-43, FUS co-localization with ubiquitin has been observed (Deng et al, 2010) in post-mortem ALS brains and spinal cord tissue. However concluding evidence of FUS ubiquitination needs to be further investigated.

3.3.3. Methylation

Arginine methylation is frequently observed in RNA-binding proteins, a process in which nuclear importation is tightly regulated (Dormann & Haass, 2013). Although it has not been reported that TDP-43 is methylated, previous work has shown that FUS is highly methylated at most of its 37 arginine residues (Du et al, 2011; Rappsilber et al, 2003), and the protein arginine transferases (PRMT) 1 and 8 have been demonstrated to interact and catalyze arginine methylation of FUS. PRMT was found to methylate both wild-type (WT) and FUS FALS mutants at a comparable overall level, even when the mutation occurred at arginine residues (Jackel et al, 2015). Methylation enhances the toxicity of ALS-associated FUS mutants by regulating their subcellular localization as well as the formation of SGs (Dormann et al, 2012; Du et al, 2011; Scaramuzzino et al, 2013; Tradewell et al, 2012), although it is controversial whether or not methyltransferase activity is involved in the incorporation of FUS into SGs (Baron et al, 2013; Sama et al, 2013). Insights into how arginine methylation affects the subcellular localization of FUS have been given by Hass and colleagues (Dormann et al, 2012), who found that PRMT1-mediated arginine methylation within the RGG3 domain (but not PY-NLS) is necessary and sufficient to restore nuclear import of mutant FUS upon inhibition of methyltransferase activity by methylation inhibitor AdOx. In addition, the unmethylated RGG3 domain tightly binds with Transportin (TRN) to compensate the weaker association between TRN and PY-NLS with mutation P525L, whereas methylation in the RGG3 domain impairs its binding with TRN significantly in FUS peptides containing P525L mutation, but only slightly in WT peptide. This binding affinity difference could play a role in PRMT-dependent methylation in ALS-FUS. Interestingly, by using an antibody specific for FUS methylation, researchers in this study also revealed that FUS is highly methylated in cytoplasmic inclusions in ALS-FUS patients, but not in FTLD-FUS patients, indicating that arginine methylation may have distinct implications in the pathogenesis of the two neurodegenerative disorders (Dormann et al, 2012). Tibshirani et al., showed that nucleo-cytoplasmic shuttling of FUS is able to affect the distribution of PRMT1 in both cultured motor neurons and transgenic mice, and in turn regulates its nuclear substrates (Tibshirani et al, 2015). In this study, PRMT1 was found to distribute with overexpressed human WT FUS in nuclei or redistribute to the cytoplasm with ectopically expressed ALS-linked FUS mutants including R521H, P525L, and R495X in neurons, in both physiological and stress conditions. Furthermore, the depletion of PRMT1 in nuclei induced by the cytoplasmic inclusion of FUS mutant R521H reduced asymmetric arginine dimethylation of histone 4 (H4), a key substrate of PRMT1, and the methylation at arginine 3, which is required for the acetylation of histone 3 (H3) at lysine residues 9 and 14 (H3K9/K14ac). Accordingly, a decreased acetylation of H3K9/K14 was also demonstrated. Additionally, decreased H3K9/K14ac was observed in either WT FUS or R521H FUS-depleted nuclei (Tibshirani et al, 2015).

3.4. Protein truncation/degradation

TDP-43 harbors caspase 3 cleavage sites, and when cleaved, TDP-43 forms ~25 kDa and ~35 kDa fragments. Researchers found that in vitro incubation of TDP-43 with caspase 3 and caspase 7 produces fragments of ~42, ~35 and ~25 kDa (Zhang et al, 2007). Fragments of ~37 kDa phosphorylated at S409 and S410 have been observed, and are presumed to be the phosphorylated forms of 35kDa fragment (Zhang et al, 2007). Research suggests that once the CTFs are generated, they are phosphorylated and ubiquitinated. The accumulation of TDP-43 CTFs may lead to neuronal dysfunction. Intermediate length poly-glutamine expansions in ATXN2 are a risk factor for ALS (Elden et al, 2010). These expansions may activate caspase 3 promoting TDP-43 cleavage and aggregate formation. Hart and colleagues show that the intermediate-length ATXN2 polyQ expansions enhance stress-induced C-terminal cleavage of TDP-43 by the activation of stress-dependent multiple caspases, including caspase 3. Accumulation of caspase 3 in motor neurons may represent a risk factor for ALS cases that harbor Ataxin2 polyQ expansions (Hart & Gitler, 2012). N-terminal sequencing of brain TDP-43 CTFs revealed that the cleavage begins at site Arg 208 or Asp 219. TDP-43 has an NLS motif that lies within the NTD. Cleavage of TDP-43 removes this NLS sequence, but not the nuclear export signal (NES) sequence, promoting cytoplasmic accumulation of TDP-43. These cytoplasmic CTFs have been proposed to serve as primary seeds for the aggregation of TDP-43 into inclusions (Hart & Gitler, 2012). Unlike TDP-43, FUS fragments have not been identified in motor neuron models.

4. TDP-43/FUS animal models: Lessons learned and challenges

Expression of wild-type or FALS mutant TDP-43 or deletion of the endogenous TDP-43 homolog in animal models can cause motor dysfunction and neuronal death, modeling ALS and FTLD. Several loss-of-function and transgenic overexpression rodent models have been generated. TDP-43 homozygous knockout mice die between embryonic day 3.5 and 8.5; even though RNA splicing regulation function carried by TDP-43 is important, the exact cause of death is not known. (Kraemer et al, 2010; Sephton et al, 2010; Wu et al, 2010). Heterozygous mice with one functional copy of TARDBP are viable, although they develop mild motor dysfunctions and moderate pathology with age (Kraemer et al, 2010). To overcome lethality of loss of TDP-43 during development, TDP-43 was deleted specifically in motor neurons and spinal cord in post-natal mice using the Cre-lox system (Wu et al, 2012). This mice line presents muscle atrophy and motor neuron loss (Iguchi et al, 2013). A transgenic mouse line that expressed an ectopic microRNA to reduce endogenous TDP-43 ubiquitously presented muscle weakness, paralysis and neurodegeneration of cortical and spinal neurons, resulting in early lethality (Yang et al, 2014).

The transgenic rodent models vary depending on the selection of the promoter driving gene expression. Investigators have overexpressed TDP-43 and ALS-linked TDP-43 mutants using neuronal-specific promoters, inducible promoters, or ubiquitous promoters (Stallings et al, 2010; Wegorzewska et al, 2009; Xu et al, 2010; Xu et al, 2011). WT TDP-43 overexpression in mice presents accumulation of polyubiquitinated TDP-43, TDP-43 fragmentation, astrogliosis, microgliosis, axonal degeneration, neuronal loss, motor function impairments, and shortened lifespan (Sasaki et al, 2015). TDP-43 overexpressing mice also display behavioral deficits and motor neuronal loss (Tsai et al, 2010). Mice expressing human TDP-43 driven by the neuron specific Thy1 promoter (TAR4 mice) present nuclear ubiquitinated and hyperphosphorylated cytoplasmic inclusions in cortical neurons. These mice also accumulate ~25kDa and ~35kDa C-terminal fragments of TDP-43 that are characteristically found in ALS brain samples. However, the TAR4 mice do not exhibit limb paralysis (Wils et al, 2010). Mice expressing human TDP-43 driven by prion protein promoter (Prp) have decreased mouse TDP-43 mRNA levels, reactive gliosis, degenerating neurites and neurons in spinal cord. This mouse line also accumulates ~25kDa CTFs and ubiquitinated cytoplasmic inclusions. This mouse model presents motor deficits at 21 days and lethality around 2 months of age. Human TDP-43 expression affects mitochondrial distribution and integrity in mouse spinal cord neurons (Xu et al, 2010). Transgenic mice expressing human TDP-43 driven by bacterial artificial chromosome (BAC) have cognitive and motor deficits with ~25kDa and ~35kDa CTFs, reactive gliosis, and neuroinflammation (Swarup et al, 2011a; Xu et al, 2011).

Transgenic mouse models with inducible overexpression of human TDP-43 with defective NLS in forebrain, driven by a Dox-suppressible CamK2a promoter showed neuronal loss; however, phosphorylated and ubiquitinated TDP-43 pathological aggregates were less observed (Igaz et al, 2011). Interestingly, cytoplasmic expression of human TDP-43 resulted in an abnormal decrease of nuclear mouse TDP-43 and altered gene expression (Igaz et al, 2011). This mouse model did not develop ALS-like phenotype possibly due to the lack of expression in spinal cord. To overcome this, a second generation of Dox suppressible transgenic with defective NLS TDP-43 expressing line driven by neurofilament heavy chain promoter to express in brain and spinal cord was generated (Walker et al, 2015). This mouse line develops ALS-like phenotype with accumulation of phosphorylated cytoplasmic TDP-43 in brain and spinal cord with significant motor neuron loss and progressive motor impairment, eventually leading to death. Although, expression of human TDP-43 reduces endogenous mouse TDP-43 in neuronal nuclei, suppressing human TDP-43 expression after disease onset reduces the phosphorylated TDP-43 pathology, simultaneously increases nuclear mouse TDP-43 to rescue motor impairment and extension in lifespan (Walker et al, 2015).

Interestingly, while cytoplasmic mis-localization and fragmentation of TDP-43 is routinely observed in transgenic mouse models, inclusions are rarely observed. TDP-43 overexpression in D. melanogaster also results in loss of motor function and decreased dendrites and synapses (Ayala et al, 2005), while depletion of the TDP-43 ortholog results in reduced lifespan and locomotor defects due to alterations in dendrite branching and synapses (Feiguin et al, 2009). Zebrafish overexpressing TDP-43 exhibit abnormal swimming behavior and defective neuronal axon formation (Ayala et al, 2005). The fact that TDP-43 mutant expression is more toxic than WT TDP-43 is consistent with other animal models (yeast, chicken embryos, D. melanogaster, and mammalian cells). In yeast models, TDP-43 mutation accelerates aggregation of TDP-43. TDP-43 mutant expression alters SG formation, leading to increased formation of inclusions leading to toxic gain-of-function and cell death (Ash et al, 2010; Hanson et al, 2010; Kabashi et al, 2011b; Lagier-Tourenne & Cleveland, 2009; Polymenidou et al, 2011).

The first TDP-43 mouse model expressing ALS-associated TDP-43 mutant A315T driven by prion promoter presented ubiquitination with positive cortical neurons with loss of nuclear TDP-43 (Wegorzewska et al, 2009). These mice also have abnormal neuritis, decreased number of neurons in cortical layer V, and 20% motor neuron loss in ventral horn (Wegorzewska et al, 2009). In addition, TDP-43 A315T mice also develop apparent gastrointestinal dysfunction likely due to neurodegeneration of the myenteric plexus of the colon (Esmaeili et al, 2013; Guo et al, 2012). Transgenic TDP-43 M337V expression caused protein fragmentation and increased cytoplasmic expression compared to WT TDP-43 mice. Both WT and mutant TDP-43 proteins are neurotoxic upon overexpression, but mutant TDP-43 requires less overexpression than WT TDP-43 to induce neurotoxicity. TDP-43 mutant overexpression results in reactive gliosis, axonal and myelin degeneration, gait abnormalities, and early lethality (Arnold et al, 2013; Janssens et al, 2013; Swarup et al, 2011a; Xu et al, 2011).

A rat model expressing ubiquitous WT human TDP-43 displayed fragmentation, phosphorylation, and aggregation of TDP-43 and developed progressive degeneration of motor neurons without paralysis (Zhou et al, 2010). In comparison, transgenic rat models expressing M337V mutations become paralyzed and die as early as 29 days. Surprisingly, in rats expressing the M337V mutation, motor function can be rescued when overexpression is turned off, suggesting that therapeutic interventions targeting TDP-43 may be effective after disease onset (Zhou et al, 2010).

In summary, rodent models show TDP-43 toxicity is dose-dependent, few of the models have robust cytoplasmic aggregation, and TDP-43-toxicity appears to be critical in promoting disease. The major drawback of the available rodent models is that none of these models mimic the multi-faceted ALS disease phenotype and pathology, making it challenging to completely understand and address future therapeutic approaches. Most models do not demonstrate the key ALS hallmark of TDP-43 nuclear depletion and cytoplasmic sequestration and aggregation at the same time. In the case of overexpressing transgenic models, it is possible that TDP-43 autoregulation due to the presence of endogenous mouse TDP-43 could be a confounding factor. As knock-out of endogenous TDP-43 is lethal at the embryonic stage; conditional knockout model in which TDP-43 is targeted in a tissue-specific manner have been successful in recapitulating adult onset of SALS to an extent. A recently generated transgenic ΔNLS-TDP-43 mouse model did develop an ALS-like phenotype (Chiang et al, 2010).

Like ALS-TDP-43, several FUS-deficient and overexpression ALS-FUS animal models have been generated, which develop ALS-like phonotypes to varying extent. Although FUS knockout (FUS^−/−) mice with different genetic backgrounds have distinct features, inbred FUS^−/− mice showed defective B-lymphocyte development and activation, as well as perinatal death. Outbred FUS^−/− mouse displayed defects in spermatogenesis and increased radio-sensitivity but were able to reach adulthood. Both mice lines have high genomic instability (Hicks et al, 2000; Kuroda et al, 2000). The morphological observation in neurons from embryos of inbred FUS^−/− mouse was also reported, which provided the initial insight into the potential function of FUS in the CNS (Tolino et al, 2012). In this study, FUS^−/− primary hippocampal neurons showed irregularly branched dendrites, and numerous long and thin processes, like immature axons, extended from cell bodies but had only a single axon. Moreover, FUS-deficient neurons isolated from mice displayed significantly reduced dendritic spine density, but increased occurrence of filopodia-like spines, with morphology similar to thin and long cytoplasmic protrusions (Fujii et al, 2005). Studies in D. melanogaster and zebrafish demonstrated that loss of FUS directly leads to neuronal cell death (Kabashi et al, 2011a; Wang et al, 2011). Together, these studies suggested a critical role of FUS in neuronal maturation.

A number of transgenic FUS models have been established by independent research teams (Robinson et al, 2015; Shelkovnikova et al, 2013a). Human FUS mutants, such as R495X, H517Q, R521G, and R521C, were ectopically expressed into D. melanogaster, zebrafish or mouse models (Lanson et al, 2011). D. melanogaster expressing human FUS mutants (R518K, R521C and R521H) developed severe neurodegeneration characterized by disorganized ommatidia and loss of mechano-sensory bristles in eyes. In zebrafish, human WT FUS, FUS H517Q and FUS R521G exhibited a predominantly nuclear localization in the spinal cord while FUS R495X and G515X truncation mutants showed cytoplasmic accumulation, and all the mutants formed SGs in response to heat shock stress. FUS lacking NLS (FUS 1-359) transgenic mouse model showed several key features of human ALS including the loss of spinal motor neurons and peripheral nerve fibers or lower motor neuron populations in the brainstem. Mice expressing FUS that lack RRM domain and carry a R522G mutation showed significant neuronal proteinopathy but no apparent neurodegeneration in brain or brainstem region. However, both mouse models displayed shorter lifespans (Robinson et al, 2015; Shelkovnikova et al, 2013a).

5. New Perspectives: Involvement of Genome Damage and Repair Defects in ALS pathology

5.1. DNA damage in ALS-affected motor neurons

To date, growing evidence supports the notion that damaged DNA accumulates in neurons as aging progresses, and neurons are particularly susceptible to accumulate damaged DNA due to their lack of self-renewal and replication. Furthermore, the brain is more susceptible to DNA damage due to high neuronal activity and higher mitochondrial respiration, a process that produces ROS that can cause the formation of DNA lesions, including single-strand breaks (SSBs) and DSBs (Acosta et al, 2014; Hegde et al, 2012; Mitra et al, 2014a). A recent study suggested that normal neuronal activity may produce endogenous DSBs, the most lethal form of DNA damage, by the activation of the NMDA glutamate receptor involved in neuroplasticity (Suberbielle et al, 2013). Increased levels of γ-H2AX, a well-known DNA DSB marker (Wang et al, 2014), was observed in postmortem brain sections of FALS patients, carrying FUS R521C or P525L mutations, in comparison to age-matched controls. This suggests that accumulated DSBs due to the loss of DNA repair function of FUS in ALS (Wang et al, 2013b), and confirmed by a transgenic mouse model harboring the FUS mutation presented elevated levels of several DNA damage markers, including γH2AX, phosphorylated p53, and activating transcription factor 3 (ATF3) in the CNS (Qiu et al, 2014).

5.2. DNA repair defects and abnormal damage response: Involvement of FUS

Defective DNA repair and its possible role in ALS was hypothesized as early as 1982 when Bradley et al., proposed that abnormal DNA in ALS may arise from a deficiency of an isozyme of one of the DNA repair enzymes (Bradley & Krasin, 1982). Currently, there is emerging evidence suggesting that defective DNA repair is present in numerous neurological disorders, raising interest in studying the role of accumulation of DNA damage in these disorders including ALS (Madabhushi et al, 2015). Oxidative damage to DNA and p53 activation have been observed in motor neurons in ALS cases (Martin et al, 2005). Abnormal activities of DNA repair components including AP endonuclease 1, DNA glycosylase OGG1, mitochondrial DNA polymerase γ and PARP1 were observed in ALS patients or mouse model, which strongly supports significantly impaired oxidative DNA damage repair in ALS. Apurinic/apyrimidinic endonuclease (APE/Ref-1) protein levels are increased in ALS cases (Coppede, 2011; Martin et al, 2007; Shaikh & Martin, 2002).

The involvement of FUS in DDR already has been indicated, in which FUS was shown to bind with and promote annealing of complementary ssDNA and D-loop formation in super helical dsDNA, an essential step for homologous recombination (HR)-mediated DNA DSB repair, while the fusion form of FUS-CHOP is unable to promote DNA pairing (Baechtold et al, 1999). Several recent studies have provided insights into the molecular mechanisms underlying how FUS is involved in DDR. Independent investigations by two research groups demonstrated that FUS is recruited to and co-localizes with γ-H2AX at DNA damage sites (induced by UVA laser and micro irradiation) in a PARP 1-dependent manner in human A549 and U2OS cell lines (Mastrocola et al, 2013; Rulten et al, 2014). Additionally, the RGG2 domain has been shown to be sufficient for the recruitment of FUS to DNA damage sites, which is significantly intensified by PLD (Mastrocola et al, 2013). Gardiner and colleagues showed that FUS, but not the FUS-CHOP fused form, is phosphorylated at Ser42 by ATM in HEK-293 cells in response to irradiation-induced DNA damage, whereas Deng et al. showed that FUS is phosphorylated by DNA-PK (Deng et al, 2014b), which indicates FUS as a novel component of the PIKK family-mediated DDR signaling pathway. Furthermore, Wang et al., showed that FUS interacts with HDAC1 in primary mouse cortical neurons, and the interaction is required for an optimal HR and non-homologous end joining (NHEJ)-mediated DSB repair. Further, increased DNA damage levels were observed in NeuN-positive neurons from FALS patients harboring R521C or P525L FUS mutations, and FALS-associated mutations FUS-R244C, R514S, or R521C overexpressed in U2OS cells showed defective HR and NHEJ-mediated DSB repair (Wang et al, 2013b). Although the phenotypes of DNA repair deficiency have been linked to the loss of FUS, the molecular mechanism of how FUS participates in DNA damage recognition and/or repair is largely unclear. The contribution and mechanisms of FUS-related unrepaired DNA damage in the initiation and progression of degenerative diseases still needs to be established.

Persistent accumulation of unrepaired genome damage due to increased damage induction coupled with their defective repair in neurons, could lead to arrest of transcription and damage-induced neuroinflammation (Shiwaku & Okazawa, 2015). Consistently, recent studies have reported accumulation of R-loop structures (presumably formed by damage induced inhibition of transcription) (Elden et al, 2010; Salvi & Mekhail, 2015), and activation of inflammatory signaling in ALS affected brain tissue (Kawamata et al, 1992; Mantovani et al, 2009; Swarup et al, 2011b; Turner et al, 2004). Moreover, transcriptional stalling together with DNA repair inhibition is a recipe for exacerbating genome instability and enhancing hypermethylation of gene promoters leading to gene silencing (Schmitz et al, 2009). Together, these recent developments open up a new avenue of research on the implications of defective DNA repair and DDR in neuronal death in ALS, and their potential as targets for therapeutic approaches.

6. Concluding remarks

The groundbreaking discoveries in 2006 implicating aggregation/fibrillation of the RNA/DNA-binding protein TDP-43 and FUS in ALS and FTLD (Kwiatkowski et al, 2009; Neumann et al, 2006; Vance et al, 2009) triggered a flurry of research activities that led to the discovery of TDP-43/FUS mutations in hereditary ALS patients and the widespread presence of TDP-43 or FUS pathology in about 40% of other neurodegenerative diseases, including, AD and PD (Arai et al, 2006; Geser et al, 2008; Hasegawa et al, 2007; Nakashima-Yasuda et al, 2007). However, how TDP-43/FUS pathology triggers neuronal apoptosis still remains unclear.

Whether neurotoxicity of RNA/DNA binding proteins in ALS is due to its ‘gain-of-toxicity’ or ‘loss-of-function’ is a key question. The ‘gain-of-toxicity’ hypothesis had initially received much attention, primarily based on the striking neurodegenerative phenotypes in multiple TDP-43/FUS-overexpression/aggregation models (Ayala et al, 2011). Yet, recent studies demonstrating widespread nuclear clearance of TDP-43 in ALS-affected neurons and significant loss of functional TDP-43/FUS pool due to its aggregation (Polymenidou et al, 2011) and strong ALS phenotype in TDP-43-deficient (partial) mice (Chiang et al, 2010; Walker et al, 2015), warrant detailed investigation of the ‘loss-of-function’ model.

Thus, despite significant increase in our understanding of the pathological and biochemical changes in ALS and other neurodegenerative disorders, there is no current cure; available treatments only temporarily slow the disease progression, but do not prevent neuronal death. This underscores the necessity of an overarching approach to unravel the fundamental mechanisms of disease initiation and progression in order to design more effective ways to prevent the onset of ALS, delay its progression, and finally develop improved treatment protocols for ALS patients.

Highlights.

• Amyotrophic lateral sclerosis (ALS) is a group of motor neuron diseases involving a dozen distinct and overlapping protein inclusions.

• The complex neuropathology of RNA/DNA binding proteins TDP-43 and FUS in motor neuron disease is critically discussed.

• The loss of function due to nuclear clearance vs. gain of aggregating protein toxicity of TDP-43/FUS is comprehensively assessed.

• New avenues of research involving the role of genome damage and repair defects in FUS/TDP-43-associated ALS.

• Role of disease-linked TDP-43/FUS mutations in familial and sporadic ALS and other motor neuron diseases.

• Lessons learned from TDP-43/FUS animal models.

• Rationale and need for an overarching approach to unravel the fundamental mechanisms based intervention strategies.

Abbreviation List

AD

Alzheimer’s Disease

Ago2

Argonaute 2

ALS

Amyotrophic lateral sclerosis

ALS/PD

Amyotrophic lateral sclerosis and Parkinsonism-dementia complex

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

ATF3

activating transcription factor 3

ATM

ataxia-telangiectasia mutated

BAC

bacterial artificial chromosome

BMAA

beta-Methylamino-L-alanine

C/EBP

homologous protein 10

Cdk6

cyclin-dependent kinase 6

CLIP-seq

UV-cross-linking immunoprecipitation-sequencing

CNS

central nervous system

CTD

C-terminal domain

CTFs

C-terminal fragments

DDR

DNA damage response

DNA-PK

DNA-dependent protein kinase

DSBs

doublé-strand breaks

dsDNA

double-stranded DNA

ERG

ETS-related gene

FALS

familial Amyotrophic lateral sclerosis

FET

FUS, EWS and TAF15

FTLD

frontotemporal lobar degeneration

FUS/TLS

fused in sarcoma/translocated in liposarcoma

G3BP

RasGAP SH3-domain binding protein 1

gadd7

growth-arrested DNA damage-inducible gene 7

H3

histone 3

H4

histone 4

HDAC1

histone deacetylase 1

HDAC6

histone deacetylase 6

HIV-1

human immunodeficiency virus type 1

HNE

4-hydroxynonenal

hnRNP

heterogeneous nuclear ribonucleoprotein

HR

homologous recombination

iPSC

induced pluripotent stem cells

IR

ionizing radiation

LPS

lipopolysaccharides

MALAT1

metastasis-associated lung adenocarcinoma transcript 1

miRNA

microRNA

NEAT1

nuclear enriched abundant transcript 1

NES

nuclear export signal

NHEJ

non-homologous end joining

NLS

nuclear localization signal

NMDA

N-methyl-D-aspartate

NTD

N-terminal domain

OS

OXIDATIVE STRESS

PARP

Poly(ADP-ribose Polymerase

PD

Parkinson’s Disease

PLDs

Prion-like domains

PRMT

protein arginine transferases

Prp

Prion protein promoter

PTB

polypyrimidine tract-binding

PY-NLS

proline-tyrosine nuclear localization signal

RAN

repeat associated non-ATG

RBM45

RNA-binding motif 45

RGG

Arginine-glycince-glycine

RNA Pol II

RNA polymerase II

ROS

reactive oxygen species

RRM

RNA recognition motif

SALS

sporadic Amyotrophic lateral sclerosis

SGs

stress granules

SOD1

superoxide dismutase 1

SSBs

single-strand breaks

ssDNA

single-stranded DNA

TDP-43

TAR DNA-binding protein-43

TIA-1

T-cell intracellular antigen 1

TRBP

transactivation-responsive RNA binding protein

tRNP

transport ribonucleoprotein particles

UBPY

ubiquitin isopeptidase Y

UBQLN2

ubiquilin 2

UPF1

upframeshift protein 1

UPS

ubiquitin-proteasome system

WT

wild type

ZnF

Zinc finger