Emerging drug targets in amyotrophic lateral sclerosis

Abstract

Introduction: Amyotrophic lateral sclerosis (ALS) is a progressive and devastating neurodegenerative disease resulting from injury and death of upper and lower motor neurons. Symptoms initially include muscle weakness and twitching and subsequently progress to muscle atrophy, complete loss of limb use, respiratory difficulties and ultimately death. Multiple biological mechanisms have been implicated in ALS and a complex etiology has been described. As a result, drug discovery researchers have few validated targets to pursue and patients have few therapeutic options.

Areas covered: Identification of new drug targets in ALS can be facilitated by a detailed understanding of the processes and genes that contribute to pathogenesis. Accordingly, this review summarizes current hypotheses regarding underlying mechanisms for motor neuron susceptibility in ALS. An overview of emerging and tractable drug targets that could result in therapeutic breakthroughs is provided.

Expert opinion: Despite the immense progress that has been made in understanding ALS over the last decade, riluzole remains the only approved drug to treat ALS. Combining structure-guided drug design applied to validated and pharmaceutically tractable targets with disease-relevant phenotypic screens will allow for the identification of novel drug targets and potentially breakthrough therapeutics for ALS.

Keywords::

• amyotrophic lateral sclerosis
• motor neuron
• phenotypic screening
• structure-guided drug design

1. Introduction

1.1 Motor neuron susceptibility in amyotrophic lateral sclerosis

Amyotrophic lateral sclerosis (ALS), also known as Lou Gehrig's disease, is a progressive and devastating neurodegenerative disease resulting from injury and death of upper and lower motor neurons. This degeneration of motor neurons causes weakness and atrophy of limb musculature progressing to a complete loss of arm and leg function, speech and swallowing difficulties and ultimately death, usually from respiratory failure. The prognosis is bleak for ALS patients whose average survival after first symptom onset is 3 years [1].

Multiple underlying factors responsible for the selective motor neuron degeneration in ALS have been described. Motor neurons are large cells and have extensive dendritic and axonal processes. While the cell body of an alpha motor neuron is 50 – 70 μm in diameter, the axon length can oftentimes be greater than 1 m. Motor neurons conduct action potentials to muscle fibers where larger axon diameters increase the rate of neuronal conduction velocity. Motor neuron neuronal filaments (NFs) function to increase axonal diameter in myelinated axons in a process known as radial axonal growth [2]. A hallmark of both sporadic and familial ALS (FALS) is the presence of cytoplasmic inclusions, many of which contain phosphorylated NFs [3,4].

The metabolic demands in maintaining an axon of 1 m length or more requires a high level of mitochondrial activity. This high metabolic demand is consistent with the observation that motor neurons are particularly susceptible to hypoxia. For example, deletion of the hypoxia responsive element in the vascular endothelial growth factor (VEGF) promoter caused a selective degeneration of motor neurons and concomitant ALS-like phenotype in mice [5]. Motor neurons are involved in neurotransmission to skeletal muscle involving both glutamatergic and acetylcholine neurotransmitter systems. Thus, motor neurons are susceptible to oxidative stress. Elevated protein carbonyl levels, increased 3-nitrotyrosine levels and high 4-hydroxynonenal levels (an indicator of lipid peroxidation) are observed in ALS patients [6-8]. Despite high metabolic activity and exposure to oxidative stress, motor neurons do not have high levels of glutathione compared to hepatocytes or astrocytes [9] and thus appear particularly vulnerable to oxidative stress [10]. High levels of oxidative stress can predispose proteins to unfolding and aggregation, yet in motor neurons there is a high threshold for induction of the stress response and expression of heat shock proteins (Hsps), which was associated with a failure to activate HSF1 [11]. Similarly, challenge of spinal motor neurons with ZnCl₂, paraquat or toxic glutamate concentrations did not lead to induction of metallothioneins (zinc-binding proteins with antioxidant properties) where, in contrast, other neuronal populations did induce metallothioneins [12]. The attenuated Hsp response of motor neurons could represent an opportunity for therapeutic intervention as it has been shown that treatment with arimoclomol, a coinducer of Hsps, delays disease progression in ALS mice [13]. Arimoclomol is now in the clinic for ALS patients [14]. It is not clear why a neuronal type that is exposed to high levels of stress has difficulty mounting a heat shock or other protective responses to injury. Nonetheless, the high threshold necessary to express Hsps may contribute, to some extent, to the vulnerability of motor neurons in ALS and may represent an opportunity for therapeutic intervention.

Lower motor neurons receive extensive glutamatergic input from upper motor neurons and excitatory interneurons. Motor neurons express alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA), kainate and N-methyl-D-aspartate (NMDA) glutamate receptors (GluRs) where glutamate is the neurotransmitter used for fast synaptic transmission in the CNS. AMPA receptors are highly expressed on motor neurons and consist of four homo- or hetero-oligomeric receptor subunits (GluR1, GluR2, GluR3, and GluR4). It has been shown pharmacologically that there is a significant pool of AMPA receptors expressed in motor neurons that are permeable to calcium [15-17]. Glutamate released into the synaptic cleft stimulates AMPA receptors where signaling is terminated through uptake of glutamate by the excitatory amino acid transporter (EAAT2). It has been shown that there is a decrease in EAAT2 levels in ALS patients [18,19], which may lead to overstimulation of AMPA receptors and excessive calcium influx, initiating a number of detrimental processes and excitotoxicity in motor neurons [20].

Natural cellular processes exist to counteract excessive calcium influx from overstimulation of AMPA receptors. Cytoplasmic calcium can be transported into the mitochondria by the mitochondrial uniporter. Calcium is then slowly released through the mitochondrial Na/Ca exchanger and then ultimately leaves the cytoplasm through a number of mechanisms including the plasma membrane Na/Ca exchanger [21,22] and sarcoendoplasmic reticulum calcium transport ATPase (SERCA) [23]. It has been shown that mitochondria from spinal cord have decreased calcium-buffering capacity compared to other brain mitochondria [24] which could be a result of a lower density of mitochondria in motor neurons compared to other neurons.

The endoplasmic reticulum (ER) is also an important and high capacity storage site for calcium. Calcium plays an important role in protein folding and processing within the ER [25,26]. Elevated intracellular calcium will cause calcium-induced calcium release through the ER ryanodine and triphosphoinositol (IP3) receptors and potentially deplete ER calcium stores [21,27]. There is an important crosstalk between ER and mitochondrial calcium stores as ER calcium released into the cytosol can be uptaken by the mitochondria and then ultimately reuptaken into the ER by SERCA pumps [23,27]. This calcium signaling between compartments could become disturbed in ALS where it is tempting to speculate, that the depletion of ER calcium stores that could happen with glutamate-receptor overstimulation will lead to activation of the unfolded protein response (UPR) [27]. Although this seems to be a valid hypothesis, it has yet to be unequivocally proven [27].

Intracellular calcium can also be buffered by calcium-binding proteins (CaBPs). Of note, motor neuron populations, such as the oculomotor and abducens nuclei, which express CaBPs such as parvalbumin, calbindin, calretinin, and neurocalcin are spared in ALS patients, while other motor neurons that do not express these proteins are not [28-31]. Expression of CaBPs in cultured motor neurons are also protective against mutant superoxide dismutase (SOD), excessive glutamate and oxidative stress [32], which is consistent with the clinical observation that neurons expressing CaBPs are spared in ALS.

The ER orchestrates a complex symphony of protein distribution for axonal transport, exocytosis, and for transport to Golgi, mitochondria and nuclear compartments of the cell. In the context of a motor neuron, with the demands of maintaining large cells with long axons and extensive dendritic morphology, it is not surprising that these cells could be susceptible to ER stress. It has been noted that ER stress, as indicated by activation of the UPR, occurs in ALS [21,33]. The UPR inhibits protein synthesis and results in the expression of ER resident chaperones to assist in protein folding. Quantitative Western blotting of human spinal cord tissue from sporadic ALS (SALS) patients showed that three UPR sensors, inositol-requiring enzyme1 (IRE1), activating transcription factor 6 (ATF6) and protein kinase RNA-like ER kinase (PERK) were upregulated from fourfold to sixfold compared to control patients [34]. Also in this study, the UPR chaperones protein disulfide isomerase (PDI) and PDI-related protein Erp57 were upregulated in ALS patients. These results were consistent with an earlier study showing increased expression of PDI and eukaryotic initiation factor 2-alpha (eIF2α) in the spinal cord of ALS patients compared to control patients [35]. Similarly, in 60-day-old mutant superoxide dismutase 1 (SOD1) mice, before ALS symptoms emerged, UPR sensor proteins IRE1, ATF6 and PERK were elevated and a high level of expression of PDI was observed compared to WT mice [21,34].

In summary, motor neuron susceptibility in ALS may have many contributing factors. Motor neurons are large in size and have long axons. The metabolic requirements to sustain these cells are inherently large. Motor neurons express high levels of calcium permeable AMPA receptors and by nature of their function, are often exposed to elevated levels of calcium, particularly during intense stimulation. Motor neurons may have a limited ability to induce Hsps, reduced glutathione levels, and many motor neurons do not express CaBPs which reduce calcium-buffering capacity. In light of the enhanced susceptibility of motor neurons to stress, it is perhaps no surprise that there are more known mutated genes causing FALS (Table 1) then there are for more common neurodegenerative diseases such as Alzheimer's disease (AD) or Parkinson's disease (PD).

Table 1. Genes associated with FALS.

ALS subtype	Gene	Chromosome region	% of FALS cases	Alternative phenotype	Function
ALS1	SOD1	21q22.1	20	None	Oxidative stress
ALS2	Alsin	2q33	Rare	Spastic paraplegia	Vesicle trafficking
ALS3	Unknown	18q21	Rare	Unknown	Unknown
ALS4	SETX	9q34	Rare	AOA2	RNA processing
ALS5	Unknown	15q15-21.1	Rare	Unknown	Unknown
ALS6	FUS/TLS	16q12	4	FTD	RNA processing
ALS7	Unknown	20p13	Rare	Unknown	Unknown
ALS8	VAPB	20q13.33	Rare	SMD	Vesicle trafficking
ALS9	ANG	14q11	Rare	FTD	RNA processing
ALS10	TDP-43	1p36.22	5	FTD/FTLD-U	RNA processing
ALS11	FIG4	6q21	Rare	Unknown	Vesicle trafficking
ALS12	OPTN	10p15-p14	Rare	Unknown	Vesicle trafficking
ALS13	ATXN2	12q24.12	Rare	Unknown	RNA processing
ALS14	VCP	9p13.3	Rare	IBMPFD	Vesicle trafficking/UPS
ALS15	UBQLN2	Xp11.21	Rare	Unknown	UPS/Autophagy
Undesignated	DAO	12q24	Rare	Unknown	Glutamate transmission
Undesignated	DCTN1	2p13	Rare	FTD	Vesicle trafficking
Undesignated	C9Orf72	9p21	23 – 40	FTD	RNA processing?
Undesignated	PFN1	17p13.3	Rare	Unknown	Actin dynamics

ALS: Amyotrophic lateral sclerosis; ANG: Angiogenin; AOA2: Ataxia with oculomotor apraxia 2; ATXN2: Ataxin-2; DAO: D-amino oxidase; DCTN1: Dynactin 1; FIG4: Factor induced gene 4; FTD: Frontotemporal dementia; FUS/TLS: Fused in sarcoma, translocated in liposarcoma; IBMPFD: Inclusion body myopathy, Paget's disease and frontotemporal dementia; OPTN: Optineurin; PFN1: Profillin 1; SETX: Sentaxin; SMD: Spinal muscular dystrophy; SOD1: Superoxide dismutase; TDP-43: TAR DNA binding protein; UBQLN2: Ubiquilin-2; VAPB: Vesicle-associated membrane protein-associated protein B; VCP: Valosin containing protein.

1.2 Current and future therapy for ALS patients

Riluzole, a potent anticonvulsant drug [36], has been approved to treat ALS since 1995, yet despite its approval for this use, the efficacy of riluzole is modest. Riluzole will extend the lifespan of ALS patients by a few months on average [37]. Riluzole may have several different mechanisms of action; however, the biological relevance of these for its effects in ALS is not completely understood. Riluzole noncompetitively blocks excitatory amino acid receptors [38], inhibits glutamic acid release and inactivates voltage-dependent sodium channels [39]. Some of the electrophysiological effects of riluzole can be blocked by pertussis toxin suggesting that it may interact with a G-protein driven mechanism and pathway [40].

There are a number of therapeutics that are currently in clinical trials for ALS and these have been reviewed elsewhere [37,41-43] and therefore will not be discussed in this present report. Additional development candidates not reviewed elsewhere include GSK1223249, an anti-Nogo-A antibody [44]. Nogo-A is a high molecular weight membrane protein of spinal cord myelin. Nogo-A functions to inhibit neurite outgrowth and can trigger growth cone collapse [45,46] which suggests that Nogo-A is an important factor involved in limiting neurite regeneration and repair of injured axons. In preclinical studies, Anti-Nogo-A antibody treatment promotes recovery of manual dexterity after cervical lesion in primates [47]. Genetic ablation of Nogo-A in SOD mice resulted in a moderate but significant increase in lifespan, increased the number of motor neurons, and eliminated ubiquitin inclusions (a marker of cell stress) [48]. Nogo-A expression has also been shown to be a prognostic marker for ALS early in the course of the disease [49]. A Phase I trial of GSK1223249 in ALS patients has been completed where we await safety and pharmacokinetic (PK) results of this clinical study.

Tirasemtiv, formerly known as CK-2017357, is an imidazo-pyrazine with a molecular mass of 230 Da. Tirasemtiv is a fast skeletal muscle troponin activator which increases fast skeletal muscle sensitivity to calcium, resulting in enhanced skeletal muscle force and slowing of time to muscle fatigue [50]. Tirasemtiv demonstrated potentially clinically relevant pharmacodynamic effects in a completed Phase IIa evidence of effect clinical trial in ALS patients [51]. At 6 h after dosing, patients demonstrated a positive change in overall status and decreased muscle fatigability. Data also demonstrated a statistically significant increase in the maximum volume of air patients could inhale and exhale [51]. A Phase IIb clinical trial for Tirasemtiv is expected in the near future as this looks to be a promising, albeit palliative approach to treating ALS.

Histone deacetylase (HDAC) inhibitors have been discussed as potential therapeutics for neurodegenerative diseases for some time where their limitation lies in their lack of selectivity and concomitant toxicity [52]. Inhibition of histone deacetylation can modulate gene expression and protein synthesis [53] and thus presents a biological basis for potential treatment of ALS. HDAC inhibitors have also been shown to be efficacious in ALS mouse models in a number of studies [54-58]. A Phase II clinical trial of sodium phenylbutyrate in patients with ALS demonstrated safety and tolerability at high doses, as well as a trend toward an increase in histone acetylation in patients' blood samples [59]. As this clinical trial was not powered to observe a potential clinical benefit, it remains to be seen whether phenylbutyrate will show efficacy in ALS patients.

2. Genes associated with FALS

Approximately 5 – 10% of ALS patients have a family history and/or the presence of a known genetic mutation and thus has familial ALS (FALS), while the remainder of patients are characterized with SALS [60]. Symptoms first appear for patients with FALS in the fourth to fifth decade of life, while features of the condition for SALS usually appear later. FALS and SALS are indistinguishable based on clinical symptoms. Genes associated with FALS can be grouped into five categories: i) those that effect RNA processing, ii) those that effect vesicle trafficking, iii) those that affect oxidative stress, iv) those that effect the ubiquitin proteasome system (UPS) and autophagy, and v) those that currently have an unknown function (Table 1).

2.1 Genes associated with RNA processing

TAR DNA-binding protein of 43 kDa (TDP-43) is a major component of ubiquitinated protein aggregates found in the CNS of ALS patients and those suffering from frontotemporal lobar degeneration with ubiquitinated inclusions (FTLD-U) [61,62]. To date, 44 different mutations of this gene have been identified that account for close to 5% of all FALS patients [53]. TDP-43 binds to an estimated one-third of all mouse and human brain RNAs [63,64]. TDP-43 has a profound effect on mRNA splicing and mature mRNA levels as determined from TDP-43 knockdown studies [63,64]. TDP-43 participates in the processing of many mRNAs that are involved in neuronal function, such as neurexins, the NMDA receptor and NF-L [63]. Fused in sarcoma or translocated in liposarcoma (FUS/TLS) is another gene involved in RNA processing, which has structural and functional similarities to TDP-43, where a total of 46 mutations have been identified accounting for ∼5% of all FALS patients [65]. In ALS, wild type and mutant TDP-43, as well as FUS accumulate in the cytoplasm and lose their nuclear localization [65]. Interestingly, TDP-43 and FUS mislocalization and aggregation is observed in other diseases such as FTLD-U, AD and Huntington's disease (HD) [65]. One unanswered question is why are TDP-43 and FUS so susceptible to aggregation? They are certainly large, multi-domain proteins that have poly-functional roles in mRNA processing. It is possible that posttranslational modifications such as phosphorylation and proteolysis may enhance the susceptibility of these proteins to aggregation [65]. Interestingly, one ALS-associated mutation in TDP-43 enhances its stability and promotes a complex with FUS, suggesting that this mutation in TDP-43 may perturb FUS function [66]. This result leaves the door open for a potential toxic gain of function activity for some TDP-43 mutations.

Other proteins that are genetically linked to ALS that are involved in RNA processing include Senataxin (SETX) which is a DNA/RNA helicase protein that affects gene transcription [67]. It has been demonstrated that depletion of SETX in HeLa cells causes an increase in readthrough RNA and Pol II density downstream of the polyA site, suggesting an involvement in transcriptional termination [68]. Mutations in SETX account for <1% of FALS cases. Angiogenin (ANG) is a secreted RNase that accounts for <1% of FALS cases. ANG has been shown to induce RNA cleavage in astrocytes where FALS associated mutations in ANG abrogate its RNase activity [69].

2.2 SOD1

Cu/Zn SOD was the first gene to be linked to ALS. SOD is responsible for the catabolism of superoxide radicals to hydrogen peroxide and oxygen. SOD1 is a 153 amino acid, functional homodimer that binds to Cu and Zn. SOD is ubiquitously expressed and comprises ∼ 1% of all cytoplasmic proteins [70]. Mutations that occur in the SOD gene account for ∼ 20% of all FALS cases [71]. The concept that mutant SOD1 causes aberrant oxyradical reactions, explaining its toxic gain of function, is controversial and has been reviewed elsewhere [71,72]. However, of note, it has also been suggested that the oxidative damage caused by mutant SOD1 may be attributed to Cu binding outside of the active site [73].

2.3 Genes involved in vesicle trafficking and other genes of interest

The size of a motor neuron and length of its axon can cause increased metabolic burden. Thus, it is not surprising that mutations in genes involved in vesicle trafficking may be linked to ALS. One such gene is Alsin (ALS2), which functions as a Rab5 guanine nucleotide exchange factor [74]. Mice deficient in ALS2 exhibit age-dependent neurological and motor neuron deficits that are associated with altered endosome trafficking [75]. For example, in wild type cortical neurons that were stimulated with brain-derived neurotrophic factor (BDNF), we observed a typical accumulation of Trk-B in the perinuclear area, which is indicative of retrograde transport. In contrast, in neurons derived from Als2^−/− mice, this was not observed [75]. Vesicle-associated membrane protein-associated protein B (VAPB), factor-induced gene 4 (FIG4), and Optineurin are other FALS genes that are also involved in vesicle trafficking [76-78] and have been similarly linked to FALS (Table 1).

Mutations in the C9Orf72 gene have been shown to cause up to 40% of all FALS cases and are also associated with FTLD-U [79]. Although the function of the C9Orf72 gene is not known, the presence of RNA nuclear foci in C9Orf72 FALS patient brain suggests that alternative mRNA splicing is dysregulated [80]. Most recently, mutations within the profilin 1 (PFN1) gene have been shown to cause FALS [81]. PFN1 is an actin-binding protein that plays an important role in regulating actin dynamics. PFN1 enhances the rate of actin polymerization by binding to actin-ADP to catalyze nucleotide exchange facilitating formation of actin-ATP. PFN1 bound actin-ATP is a substrate of Formin which adds actin-ATP to the barbed ends of nascent actin fibers, with a subsequent release of PFN1 [82]. From a functional perspective, a mutation in PFN1 that blocked actin-binding activity was shown to inhibit neurite outgrowth in a neuroblastoma cell line [83]. In a follow-up study to determine the prevalence of PFN1 gene mutations in FALS, this gene was sequenced in a cohort of 94 FALS patients of European ancestry. No PFN1 mutations were identified in this cohort suggesting that PFN1 mutations are a rare cause of FALS [84]. Despite that mutations in PFN1 may account for a low frequency of FALS patients; PFN1 gene mutations and their linkage to FALS have demonstrated a novel mechanism that can cause the disease.

3. New targets for ALS

3.1 EphA4

Ephrin type-A receptor 4 (EphA4) is a receptor tyrosine kinase that consists of an extracellular ligand-binding domain, a transmembrane region, and an intracellular kinase domain that can accommodate multiple docking sites for downstream signaling [85]. Upon ligand binding, the EphA4 receptor becomes activated and initiates signaling in the forward direction. However, ephrins, the ligands of the Eph receptor, are attached to the cell membrane through a glycosylphosphatidylinositol link which enables signal transduction in the reverse direction [86]. This bidirectional signaling is a key feature of Ephrin biology.

EphA4 is enriched at excitatory synapses in motor neurons and in the hippocampus [87,88]. Genetic knockdown of EphA4 in the mouse suggests that ephrin-A:EphA forward signaling is required for selection of lateral lower motor column neurons toward the dorsal limb, thus contributing to the topographic organization of motor neuron projections [87,89]. It has also been shown that the EphA4 receptor tyrosine kinase regulates spine morphology where its activation reduces spine length and density in hippocampal slices [90]. Subsequently EphA4 forward signaling was shown to decrease phosphorylation of focal adhesion kinase (FAK), and proline-rich tyrposine kinase 2 (Pyk2), thus regulating dendritic spine remodeling by affecting β1 integrin signaling pathways [91]. In the hippocampus, EphA4 receptor reverse signaling may overactivate its ligand glial ephrinA3 and inhibit glutamate uptake by the glutamate transporter [88]. Recently, in a genetic screen using a SOD1 zebrafish model, the most protective knockdown identified was of the gene Rtk2, which has 67% identity to human EphA4 [92]. Further validation of the importance of this target were the results in SOD1^G93A mice where a 50% knockdown of EphA4 expression significantly enhanced motor function (rotarod performance) and increased survival time [92]. In patients with ALS, EphA4 expression inversely correlated with disease onset and survival where loss of function and EphA4 mutations are associated with longer survival [92]. These findings suggest that EphA4 represents a compelling biological target for ALS.

From a drug discovery perspective, EphA4 is a tractable drug target. Compounds of the pyrrolyl benzoic acid scaffold have been identified that block the interaction of ephrins with the EphA4 receptor [93]. In addition, small molecules have been identified that bind to the ATP pocket and block EphA4 kinase activity with high affinity [94]. A crystal structure of the EphA4 tyrosine kinase domain in the apo- and dasatinib-bound state has also been determined which will facilitate structure-guided drug design [95]. Key insights from the crystal structure suggest a path forward for kinase receptor selectivity between EphA4 and other Ephrin receptors [95]. There are a number of commercially available EphA4 biochemical kinase assays that are amenable to high throughput screening (HTS) which should facilitate the identification of leads for drug discovery efforts. Further, a number of antibody reagents are readily available suggesting that the development of cell-based characterization assays will not represent an obstacle for drug discovery. In addition to the potential role for EphA4 as a drug target for ALS, it has been implicated to play a role in gastric cancer [96] and blocking EphA4 is effective in preclinical models of spinal cord injury [97]. At present, however, it does not appear that EphA4 has been intensely investigated in other neurodegenerative indications.

3.2 Nuclear erythroid 2-related factor 2

Nuclear erythroid 2-related factor 2 (Nrf2) is a transcription factor, which under non-stressed resting conditions, is normally located in the cytoplasm bound to Kelch-like ECH-associated protein 1 (Keap1). Under conditions of oxidative stress, oxidative modification of Keap1 leads to the dissociation of Nrf2 from Keap1. Nrf2 can then translocate to the nucleus, where Nrf2 binds the antioxidant response element (ARE) and facilitate the expression of over 250 cytoprotective genes including heme oxygenase-1 (HO-1), glutathione, thioredoxin (TRX) and Hsps. The Nrf2/Keap1/ARE system is a major cellular defense system for cells undergoing oxidative stress [98]. As described above, oxidative stress has been implicated to play a role in ALS [99] and the Nrf2/Keap1 system was shown to be impaired in motor neurons in a SOD1 mouse model [100]. Interestingly, selective overexpression of Nrf2, in astrocytes, using double transgenic SOD1^G93A/GFAP-Nrf2 mice, extended survival by over 20 days compared to SOD1^G93A mice [101]. This protective effect was mediated in part by an increase of astrocyte-secreted glutathione [101]. Triterpenoid small molecule activators of the Nrf-2/Keap1 system have extended survival in SOD1^G93A mice [102]. Bardoxolone Methyl (formerly RTA402), a triterpenoid has completed a Phase II clinical trial for stage 4 chronic kidney disease (CKD) [103] and is currently in a Phase III trial for this indication. As there is good rationale that activation of Nrf2/Keap1 system would be helpful to ALS patients, clinical testing using a CNS permeable Nrf2 activator could be considered.

3.3 DJ-1

DJ-1 is a 20 kDa protein that is ubiquitously expressed and widely conserved throughout species [104]. DJ-1 was identified as an oncogene and subsequently linked to familial PD [105]. DJ-1 is a multifunctional oxidative stress response protein that confers protection to cells undergoing oxidative and or mitochondrial stress [106,107]. Multiple functions for DJ-1 have been described. For example, DJ-1 has been shown to negatively regulate the apoptotic signaling kinase1 (ASK1) [108,109]. DJ-1 has also shown to suppress the proapoptotic PTEN phosphatase and thus is an enhancer of the Akt survival pathway [109,110]. DJ-1 is a redox sensitive protein where it has been shown that oxidation of cysteine 106 on the DJ-1 protein is required for its protective function [106,107]. Interestingly, excessive oxidation of this cysteine to the sulfonic acid renders the protein inactive [106]. DJ-1 is also linked to ALS where DJ-1 mutations in one Italian family caused Parkinsonism-Dementia ALS (PDALS) complex [111]. PDALS complex has been particularly well characterized in patients in the Kii peninsula of Japan and in Guam [112] where environmental as well as genetic factors may play a role in the development of disease. Thus, ALS and PD pathogenesis, may in some cases share a common etiology. At the molecular level, DJ-1 forms complexes with mutant SOD1 and ameliorates its toxicity [113]. As oxidative stress may play a role in many neurodegenerative diseases, it is not surprising that a protein like DJ-1 could be implicated in both ALS and PD. Although DJ-1 is protective in cellular and animal models of oxidative stress, it has not proved simple to assay for small molecule binders in a high-throughput biochemical format. One virtual screen identified potential binders of DJ-1 whose binding properties were confirmed using a quart crystal microbalance [114]. These compounds exhibited specificity for DJ-1 and prevented oxidative stress-induced cell death in SH-SY5Y and in primary ventral mesencepahlon cells [114]. In addition, these DJ-1-binding compounds were active in blocking dopaminergic cell death and blocked movement abnormalities in a 6-hydroxydopamine injected PD rat model [114]. It was reported that the protective effects of these compounds were mediated through binding to an oxidized form of DJ-1. It remains of interest to test these compounds in cellular and animal models of ALS.

DJ-1 levels are reported to be lower in DJ-1 familial PD patients and, as such, increasing levels of DJ-1 could be a therapeutic strategy for familial and sporadic PD and ALS [115,116]. Thus, compounds that bind and stabilize DJ-1 may be useful therapeutics. Binding assays such as those that use surface plasmon resonance (SPR) could also be used to identify small molecules that interact with DJ-1 in a medium throughput manner. A structure for DJ-1 has been solved by X-ray crystallography which may further facilitate a drug discovery program for this target [117]. Compounds of interest could be run through cellular oxidative stress protection assays in cells expressing mutant or wild type DJ-1. Although there are many inherent risks with this approach, DJ-1 is a validated target for PD and may be of relevance for ALS.

3.4 Apoptotic signaling kinase1

Apoptotic signaling kinase1 (ASK1) is a 1374 amino acid polypeptide consisting of a central serine/threonine kinase domain and coiled-coil domains at the N- and C-terminus [118,119]. ASK1 is a MAP3K that plays an essential role in cellular stress. In its resting state, ASK1 is a homodimer stabilized by its C-terminal coiled-coil domain. The N-terminal domain is bound by the redox regulatory protein, thioredoxin, which prevents its activation [120,121]. ASK1 is activated by ROS, ER stress, calcium ion influx and other stressful stimuli [122]. ASK1 is a regulator of the JNK and p38 MAP kinase cascades in stress signaling [118,122]. Oxidative stress leads to disulfide bride formation of thioredoxin and dissociation from ASK1, triggering a conformational change in the ASK1 dimer leading to phosphorylation and activation of its kinase activity [120,121]. DJ-1 can bind to ASK1 at its N-terminal domain and negatively regulate ASK1 phosphorylation activity (Figure 1). Further levels of oxidative stress lead to diminished DJ-1 levels, unregulated ASK1 activity, and subsequent cell death (Figure 1). ASK1 knockout is protective in ALS and PD mouse models [121,123]. Specifically, in SOD1^G93A/ASK1^−/− mice, mean survival times were significantly longer and neuronal cell death was ameliorated in the spinal cord compared to SOD1^G93A mice [123]. Furthermore, ASK1 has been reported to negatively regulate the proteasome, possibly by phosphorylating and inhibiting the ATPase activity of Rpt5 [124]. Thus, there is good biological rationale suggesting that well-tolerated ASK1 inhibitors may be useful for the treatment of ALS [125].

Figure 1. Hypothetical scheme for activation of ASK1 by oxidative stress. In resting conditions, ASK1 is bound to and negatively regulated by TRX. Under conditions of oxidative stress, TRX becomes oxidized and dissociates from the ASK1 signaling complex (signalosome). DJ-1 becomes oxidized under conditions of oxidative stress (*) and subsequently binds to ASK1, negatively regulating its activity. Continued oxidation of DJ-1 can lead to a conformational change (**) where it dissociates from the ASK1 signalosome, relieving the negative inhibition. ASK1 becomes phosphorylated and is now activated resulting in the activation of JNK and p38 pathways. This schematic connects ASK1, DJ-1 and p38 targets in a common pathway initiated by oxidative stress.

From a drug discovery perspective, ASK1 appears to be a tractable target. A crystal structure of ASK1 has been determined [119] which should facilitate structure-based drug discovery. ASK has three isoforms (ASK1-3) and preliminary analysis suggests that selectivity between ASK1 and ASK2 is possible, yet selectivity between ASK1 and ASK3 is unlikely [126]. It is not known whether selectivity between the three ASK isoforms is necessary for an appropriate tolerability/efficacy profile. Selectivity between ASK1 and the broader kinome also appears possible from a preliminary analysis [126]. ASK1^−/− mice are healthy and viable, yet studies have shown that ASK1 does play a role in innate immunity [127]. Thus, the benefit to cost ratio of an ASK1 inhibitor may override any underlying safety concerns, particularly in ALS patients, where there is a dire unmet medical need. Although there have been reports of ASK1 inhibitors in development for insulin resistance and cardiac indications, their use in neurodegeneration seems relatively unexplored [128].

3.5 HDAC6 stabilizers/inhibitors

Currently the nonselective HDAC inhibitor, phenylbutyrate, is undergoing clinical evaluation for the treatment of ALS [59]. HDAC6, a class II HDAC, represents a particularly intriguing target for ALS as there appears to be rationale for both a HDAC6 stabilizer/activator and a HDAC6 inhibitor for therapeutic application to ALS. HDAC6 has been shown to facilitate the formation of aggresomes, activate autophagy, and may be involved in the expression of Hsps [129,130]. Some of these neuroprotective functions may be independent of the deacetylating activity of HDAC6 [129-131]. TDP-43 and FUS may function in a complex to regulate HDAC6 mRNA expression where a loss of TDP-43 function may decrease levels of HDAC6 expression in motor neurons [132]. In this manner, a HDAC6 stabilizer, which could increase HDAC6 levels, would be helpful in ALS. Small molecule chaperones or stabilizers have been successfully pursued for some targets [133]. However, this approach is not universally applicable to all proteins.

There is also rationale for potential therapeutic role for a HDAC6 inhibitor in ALS. HDAC6 is the major α-tubulin deacetylating enzyme [134-136]. HDAC6 inhibition protects against oxidative stress-induced neurodegeneration and promotes neurite outgrowth in cortical neurons co-cultured with Chinese hamster ovary (CHO) cells expressing myelin-associated glycoprotein (MAG) [137]. In addition, HDAC6 inhibition promotes neurite outgrowth in dorsal root ganglion neurons exposed to MAG [137]. Furthermore, in a mouse model of mutant Hsp27-induced Charcot-Marie-Tooth (CMT) disease (a peripheral neuropathy), HDAC6 inhibitors restored axonal transport, and partially restored the CMT phenotype both at the behavioral and electrophysiological levels [138]. From a pharmacological perspective, HDAC6 represents a tractable target. The HDAC6 knockout mice are fully functional and viable, suggesting that the reduction of HDAC6 activity may be less likely to induce mechanism-based toxicity [139]. In addition, selective HDAC6 inhibitors have been synthesized [140,141], thereby obviating the potential for nonspecific toxicity that may be observed following treatment with CNS penetrant pan-HDAC inhibitors.

3.6 Nuclear factor-kappaB

Nuclear factor-kappaB (NF-κB) consists of a dimer of the p65 and p50 proteins. In the inactivated state, NF-κB is kept in the cytoplasm complexed with the inhibitory protein inhibitor of kappaB (IκBα). Extracellular stimuli such as TNF, reactive oxygen species and cytokines can activate many transmembrane receptors which can then activate IκB kinase which then phosphorylates the inhibitory protein IκBα. After phosphorylation, IκBα becomes ubiquitinated and then degraded. The NF-κB transcription factor then translocates to the nucleus to turn on protein expression [142]. Aberrant regulation of NF-κB has been linked to cancer, inflammatory disease, septic shock and autoimmune disease [142,143]. Recently, TDP-43 was shown to interact with the p65 subunit of NF-κB in mouse microglia cells after LPS simulation [144]. From a functional perspective, a gene reporter assay was used to show that TDP-43 acts as a coactivator of p65. Furthermore, this interaction was shown to occur in the context of human ALS patients where TDP-43 and p65 were coimmunoprecipitated from spinal cord extracts of ALS patients but not from spinal cord extracts of control patients. Consistent with an inflammatory component of ALS, it was also demonstrated that TDP-43 overexpression in glia or macrophages causes hyperactive inflammatory responses following a LPS challenge. Treatment of TDP-43 transgenic mice with Withaferin-A (WA), an NF-κB inhibitor, significantly improved motor function as demonstrated by enhanced rotorod performance and WA-treated mice had a 40% reduction in the number of partially denervated neuromuscular junctions [144]. Corroborating the results from this study, high levels of circulating TNFα have also been observed in other cohorts of ALS patients and in other animal models of ALS [145-148]. Using spinal cord organotypic cultures, it was shown that TNFα potentiates glutamate-induced motor neuron cell death which was blocked by inhibition of the NF-κB pathway [149]. Thus, chronic NF-κB activation may be a contributing factor to ALS disease and or disease progression and as such is a relevant therapeutic target. There are many approaches to successfully inhibit the NF-κB pathway with small molecules and these have been thoroughly reviewed elsewhere [150,151]. The advancement of these compounds toward clinical testing for ALS may offer hope to patients in the future.

3.7 P38α

The P38α is a member of the mitogen-activated protein kinase (MAPK) family of serine/threonine kinases. The p38α is widely expressed in endothelial, inflammatory cells, microglia and in the spinal cord [152]. The p38α is activated in response to stressful stimuli such as osmotic stress, oxidative stress, LPS, growth factors and inflammatory cytokines (TNFα, IL-1β, IL6) [152,153]. The p38 is activated in motor neurons and microglia of SOD1^G93A mice [154-157] and in ALS patients [157,158]. The p38α directly phosphorylates NFs and aberrant accumulation of phosphorylated NFs are hallmark pathological features of ALS [157]. Deficits in axonal transport are an early event observed in the disease course of SOD1 mutant transgenic mice [159]. Additional substrates for p38α include the heavy chain of the molecular motor kinesin [160]. In a series of studies, it was shown that one functional consequence of p38-mediated phosphorylation of kinesin was to inhibit fast axonal transport, an effect that was blocked by pharmacological p38 inhibition [160,161]. Of interest, p38 inhibition had a marked effect on motor neuron survival in the SOD1^G93A mouse model, yet only elicited a modest effect on survival [162]. It has been shown that riluzole protects against glutamate-induced deficits in NF axonal transport [163,] and it is therefore tempting to speculate that a combination of riluzole with a p38 inhibitor may work synergistically to reverse axonal transport deficits. There are a number of p38 inhibitors that have now entered into clinical testing for inflammatory indications and these have been reviewed extensively elsewhere [164,165]. It would be of interest to understand if a p38 inhibitor in combination with riluzole could provide significant benefit to ALS patients.

3.8 Activation of autophagy and the UPS for protein clearance

The UPS and autophagy/lysosome pathways are the two major cellular systems that degrade and clear proteins. In ALS, cytoplasmic inclusions consisting of hyperphosphorylated neurofilaments, TDP-43 and other proteins are a hallmark of the disease [61,62]. Clearance of these protein inclusions could be helpful in ameliorating symptoms of the disease as we do not know for certain whether inclusions are directly toxic themselves or represent an attempt to package aberrantly folded proteins that may not be cleared otherwise, into an innocuous compartmentalized form. Proteasomes predominantly degrade shorter half life nuclear and cytoplasmic proteins that need to be unfolded. In contrast, lysosomes degrade larger membrane proteins, oligomers and aggregates, and organelles in a process called autophagy. When proteins oligomerize, aggregate and form large inclusions, they become inaccessible to the proteasome. HD is a neurodegenerative disorder caused by a long CAG repeat in the Huntington (Htt) protein. Thus this mutant protein contains an abnormally long polyglutamine tract where after cleavage of the mutant protein, fragments containing the polyglutamine have a tendency to aggregate, forming inclusions. These inclusions, which are a hallmark of HD, can interfere with neuronal function, ultimately leading to cell death [166]. It was shown that inhibition of mTOR by rapamycin induces autophagy and reduces toxicity of mutant Htt in fly and mouse models of HD [167]. In addition, a screen for autophagy enhancers was performed which identified a number of small molecule modulators of autophagy [168]. It remains to be determined whether a similar approach will yield benefits in ALS. However, evidence of the relevance of enhancing autophagy as a potential therapeutic approach for ALS comes from a report by Hetz et al. [33]. In this study, they investigated the contribution of X-box binding protein-1 (XBP-1), a transcription factor involved in the UPR, for its contribution to FALS. They observed that knockdown of XBP1 in NSC34 cells transiently transfected with mutant SOD enhanced clearance of mutant SOD aggregates by macroautophagy [33]. This finding in a cellular system was validated in vivo where genetic knockdown of XBP1 in female mutant SOD mice prolonged survival by 10 days and increased autophagy and SOD1 degradation [33]. Thus, we look forward to the results of testing small molecule enhancers of autophagy in ALS animal models.

It has similarly been postulated that enhancing proteasomal clearance of proteins would have broad benefit across neurodegenerative diseases. Proteasomal degradation is a complex and highly regulated process. Proteins are brought to the proteasome after ubiquitination for degradation. Once at the proteasome, proteins must be deubiquitinated and unfolded before entry into the proteasome. The deubiquitinating function is performed by a family of proteins termed deubiquitinating enzymes (DUBs). There are three resident DUBs that are associated with the proteasome: Usp14, Uch37 and Rpn11. Usp14 has an ubiquitin chain trimming activity which in biochemical experiments slowed down proteasomal degradation of ubiquitinated protein substrates [169]. In addition, it was shown that small molecule inhibitors of Usp14 enhanced the clearance of TDP-43, Ataxin-3 and other proteins of relevance for neurodegenerative diseases in cellular assays [169]. It remains to be seen whether small molecule inhibitors of Usp14 will be efficacious in in vivo models of neurodegeneration.

3.9 Phenotypic screens aimed at identifying novel ALS targets

Despite some potentially promising clinical candidates currently under evaluation and some of the new promising drug targets listed above, a lack of a detailed understanding of the underlying mechanisms that cause ALS has hampered drug discovery. Cellular phenotypic screens can help facilitate the identification of new drug targets for ALS. Phenotypic screening is a type of screening used in drug discovery to identify small molecules that alter a cellular phenotype such as neurite outgrowth, expression of a particular protein or cell death. In a blinded phenotypic screen of 1040 FDA-approved drugs, it was discovered that many β-lactam antibiotics are potent stimulators of the glutamate transporter, EAAT2 (also called GLT1) expression [170]. Glutamate uptake by the transporter is one of the major modes of inactivation of glutamate signaling and thus increased expression of this transporter may be of benefit for ALS. Ceftriaxone, one of the β-lactam antibiotics identified in that study, elevated GLT1 expression threefold in cellular models and also raised GLT1 expression in rodents. In SOD1^G93A mice, administration of Ceftriaxone significantly spared motor neurons and had a modest effect on prolonging survival [170]. In a Phase II study for ALS, Ceftriaxone demonstrated good tolerability and is now in an ongoing Phase III study [171].

Similarly, cellular phenotypic screens can be used to identify compounds that enhance vesicle trafficking, RNA processing and clearance of aberrantly folded proteins. Not only can previously approved drugs be used in these screens, but screening could also be performed on larger and more diverse collections of compounds. One limitation of phenotypic screens is the lack of identification of the molecular drug target. This lack of understanding may make a traditional medicinal chemistry-based structure activity relationship (SAR) effort more difficult. However, novel chemical proteomics technologies have emerged that could help in target deconvolution following a phenotypic screen [172]. Stable isotope labeling for cell culture (SILAC) is one such technique based on mass spectrometry that allows for identification and quantification of proteins. Of relevance for identifying targets from phenotypic screens, using SILAC, cells can be metabolically labeled with different isotopes of arginine and lysine. The compound of interest can then be chemically tethered to sepharose beads and added to the cellular lysate. Specifically bound proteins can be identified, and their K _d values determined using relative quantitative mass spectrometry. Although careful controls must be considered with this technique, it has been successful in identifying novel binding targets of small molecule compounds.

One potential limitation in the development of new therapeutics targeting ALS is the observed discordance between interventions that demonstrate activity in ALS animal models and efficacy in the clinical setting. The reasons for this discrepancy are manyfold and have been highlighted elsewhere and are therefore not in the scope for this review [173]. As a potential mechanism for mitigating this issue, a drug repositioning approach has been suggested [173]. This approach would test as many FDA approved drugs as possible, in a small number of patients to observe if a marked clinical benefit was noted in individual patients. Endpoints in such a trial would include improvement of strength or cessation of disease progression. Using cellular phenotypic screens or directly testing a number of previously approved drugs in a smaller number of ALS patients may decrease the overall time to realize success in the search for new therapeutics to treat this disease.

4. Conclusions

Numerous mechanisms have been described that may contribute to the pathogenesis of ALS. The primary target of degeneration in ALS, the motor neuron, is large and has a high metabolic requirement. Motor neurons are exposed to a high level of oxidative stress as a result of primary signaling using excitatory neurotransmitters; yet these neurons appear particularly vulnerable in that additional protective mechanisms to counterbalance this stress appear to be lacking. Many motor neurons do not express high levels of calcium-buffering proteins which likely also contribute to their susceptibility in ALS.

Identification of mutations in TDP-43, FUS and most recently C9Orf72 that cause FALS have changed the way the field thinks about this disease from a mechanistic and clinical perspectives. There are currently 15 described ALS gene subtypes (Table 1) and many more unnamed genes that contribute to FALS. As this field of study progresses, it appears probable that many more genes will be categorized in the near future. Many of the ALS-associated genes can be broadly categorized into those that effect RNA processing, vesicular trafficking and oxidative stress. The recently discovered FALS gene PFN1 has illustrated a new mechanism of motor neuron degeneration. As additional genes are identified, these will further help to advance our understanding of ALS in the future.

As described above, there are a number of promising novel ALS drug targets on the horizon that may lead to breakthrough therapeutics. Genetic knockdown studies evaluating EphA4 and ASK1 kinases have significant biological effects in prolonging survival in the SOD animal models. Inhibition of HDAC6 has also shown efficacy in CMT animal models. New technology platforms such as relative quantitative mass spectrometry using SILAC will further help to deconvolute hits from phenotypic screens which should lead to additional ALS drug targets in the future, as well as provide further mechanistic understanding around this complex, multifactorial disease.

5. Expert opinion

There are many challenges facing researchers attempting to identify and develop new therapeutics for ALS. First, as demonstrated from ALS genetic studies, it is clearly a multifactorial disease with multiple potential underlying pathological mechanisms that can converge at the common endpoint of motor neuron degeneration. Despite the potential multifactorial nature of the pathogenesis of the disease, clinical studies typically utilize a heterogeneous group of ALS patients for clinical trials. Thus, it remains formally possible that compounds that have previously failed in clinical trials may exhibit efficacy in a small subset of ALS patients that are appropriately stratified. Further complicating the ability to translate nonclinical findings to effective treatments is the lack of universal concordance between results using current animal models and clinical outcomes. Given the large number of molecules that are reported to provide efficacy in nonclinical models, it may be argued that these models and/or the endpoints used for assessment of efficacy are set at a lower hurdle relative to clinical studies in patients. Another possibility that must be considered is whether the current clinical study paradigms are sufficiently sensitive and/or whether it is too late to therapeutically intervene with current treatment approaches once patients are symptomatic. Similar arguments have been forwarded in the treatment of other neurodegenerative disorders such as AD, where it has been suggested that earlier intervention may be needed to effectively impact the course of the disease. Treatment studies that allow for earlier intervention in the course of ALS, before unequivocal diagnosis, could result in an enhanced efficacy profile for the current standard of care, riluzole, as well as potentially allow compounds that have previously failed in clinical trials to demonstrate efficacy. In order to enable this approach, new biomarkers that can accurately predict disease progression need to be identified.

Despite the fact that genetics play a prominent role in ALS research, particularly with the discovery of mutations in TDP-43 and FUS that cause FALS and more recently with C9Orf72 and PFN1, these discoveries have not yet translated into new therapeutic approaches. New animal models such as the TDP-43 transgenic mouse should allow us to test new hypotheses and gain additional insight into mechanisms underlying ALS pathology. Such models may also prove to be more predictive for clinical efficacy. The zebrafish model of ALS allows for both small molecule and genetic screens directly in a medium throughput mode in vivo. Such a model system provides the opportunity to screen low-to-medium-sized compound or siRNA libraries in a reasonable time frame. Drug screening using ALS patient-specific induced pluripotent stem cells represents an alternate, albeit exciting approach that has tremendous potential from both a research and clinical perspective. Use of stem cells in this manner could allow for the high-throughput screening of large chemical libraries to identify compounds of relevance. Once a target is identified, the combination of traditional drug discovery approaches such as structure-guided drug design on compelling new targets like EphA4, ASK1, Nrf2 and HDAC6 with phenotypic screening approaches holds the promise to enhance the rate at which new drug targets for ALS are discovered and translated into clinical testing paradigms. Phenotypic screening historically has been less well utilized due to the fact that it is often difficult to deconvolute the molecular target for compounds of interest that emerge after a screening campaign. However, with new technologies such as relative quantitative mass spectrometry using SILAC, the ability to identify a compound's cellular-binding partners in a reasonable time frame is significantly enhanced. Phenotypic screens with endpoints focused on protein clearance, RNA processing, vesicle trafficking and motor neuron survival have the potential to result in multiple new targets for ALS.

Finally, the growing emphasis on drug repositioning approaches for the treatment of ALS represents a significant opportunity to expedite possible treatments to patients. Testing previously FDA-approved compounds in smaller numbers of ALS patients has the potential to lead to exciting new findings. Also, using combination therapy, taking advantage of pharmacological synergy, as is increasingly accepted as standard of care for cancer and hypertension, is an additional approach that should be considered. By combining the recent advances in genetics, chemical proteomics and informatics, we expect a number of new and exciting drug targets for this indication, and hope for ALS patients in the future.

Article highlights.

• Overview of potential mechanisms underlying motor neuron susceptibility in ALS is provided.

• Current genes linked to FALS including recently identified Profilin 1 and C9Orf72 are covered.

• In-depth discussion of the use of phenotypic screens to identify new small molecules that could be of benefit in ALS and new technology useful in deconvoluting hits from phenotypic screens.

• Discussion of the tractability, from a drug discovery perspective, of emerging targets for ALS such as EphA4, ASK1, DJ-1 and HDAC6.

• Insight into the challenges facing those developing therapeutics for ALS and how some of these challenges may be mitigated from a clinical and research perspective is provided.

Acknowledgments

The authors thank E Johnson and D Walker for helpful discussions and Z Ren and R Artis for the critical reading of this manuscript. This manuscript is dedicated to the memory of L Gehrig and H Katz, may we all follow our dreams.

Declaration of interest

M Bova and G Kinney are both employees of ELAN Pharmaceutical.

Notes

This box summarizes key points contained in the article.