Amyotrophic lateral sclerosis (ALS), also known as Lou Gehrig’s disease, is a devastating neurodegenerative disorder defined by the progressive death of both upper and lower motor neurons in the brain and spinal cord. The initial clinical presentation of ALS is variable among patients. Many individuals present with spinal-onset disease, which manifests as muscle weakness in the limbs. Others present with bulbar-onset disease, characterized by dysarthria (speaking difficulty) and dysphagia (swallowing difficulty). In a small fraction of cases (3%–5%), the disease begins with a respiratory onset, marked by orthopnea or dyspnea, often with only mild or absent signs in the spinal or bulbar regions (Chiò et al., 2011; Hardiman et al., 2017).

Despite this initial heterogeneity, the disease typically spreads to adjacent body regions as it progresses. This leads to widespread muscle atrophy, spasticity, quadriplegia, and the eventual inability to walk, use the hands and arms, or swallow. Neuropsychiatric conditions, such as cognitive or behavioral impairments, may also develop, though less consistently. Ultimately, the progressive weakness of the respiratory muscles, resulting in a loss of the ability to cough and breathe without assistance, is what makes the condition life-shortening (Fig. 1A). The typical survival time is 2–5 years following the onset of symptoms, although a small number of patients live for more than 10 years after their diagnosis (Swinnen and Robberecht, 2014).

ALS is a rare disease with a global incidence of approximately 2 per 100,000 person-years and a prevalence of 4–8 per 100,000 people, though varying with regions (Longinetti and Fang, 2019; Xu et al., 2020; Brown et al., 2021). A characteristic pathophysiological feature of ALS is a proteinopathy involving the TAR DNA-binding protein 43 (TDP-43). As motor neurons undergo injury, the TDP-43 protein is lost from the nucleus and forms aggregates in the cytoplasm, which appear as compact or skein-like structures (Neumann et al., 2006).

ALS demonstrates remarkable genetic heterogeneity that ultimately converges on shared pathophysiological pathways, providing both challenges and opportunities for therapeutic development. Currently, ALS is categorized as either familial or sporadic. Approximately 10% of cases are considered familial ALS (fALS), with inheritance patterns that can be autosomal dominant, autosomal recessive, or X-linked. To date, over 30 ALS-associated genes have been identified, accounting for more than 70% of fALS cases. The remaining majority of cases are classified as sporadic (sALS). However, many of the pathogenic mutations first identified in fALS have also been found in a subset of sALS patients, about 15% (Goutman et al., 2022; Akçimen et al., 2023).

Pathogenic variants in four major ALS-associated genes, including Superoxide Dismutase 1 (SOD1), TAR DNA-binding protein (TARDBP), fused in sarcoma (FUS), and chromosome 9 open-reading-frame 72 (C9orf72), which are responsible for approximately 60% of familial cases and about 10% of sporadic cases (Akçimen et al., 2023).

SOD1 was the first gene identified as a genetic cause of fALS (Rosen et al., 1993). It encodes the superoxide dismutase 1 enzyme, which is critical for protecting cells from damage by superoxide radicals. Over 200 mutations in SOD1 have been reported, accounting for about 12% of familial and 1%–2% of sporadic ALS in European populations, with most being missense mutations (Abel et al., 2012). The mutant SOD1 protein is believed to acquire a toxic gain-of-function, causing neuronal death through mechanisms including excitotoxicity, oxidative stress, mitochondrial dysfunction, disrupted axonal transport, and non-cell-autonomous toxicity from neuroglia (Hayashi et al., 2016).

The hexanucleotide (GGGGCC) repeat expansion in C9orf72 represents the most common genetic cause of both ALS and frontotemporal dementia (FTD), accounting for approximately 40% of familial and 7% of sporadic cases (Majounie et al., 2012). While healthy alleles have up to 30 repeats, disease-associated alleles can have between 700 and 1600 repeats (DeJesus-Hernandez et al., 2011; Renton et al., 2011). The proposed mechanisms of neurodegeneration include gain-of-function toxicity from the formation of nuclear RNA foci, repeat-associated non-AUG (RAN) translation producing toxic dipeptide repeats (DPRs), and a loss-of-function effect due to C9orf72 haploinsufficiency (Ash et al., 2013; Balendra and Isaacs, 2018).

Mutations in TARDBP and FUS, encoding the RNA-binding proteins TDP-43 and FUS, respectively, each account for approximately 4% of familial and 1% of sporadic cases (Neumann et al., 2006; Vance et al., 2009; Akçimen et al., 2023). FUS mutations are notable for their association with early disease onset, with more than 60% of FUS mutation carriers presenting first symptoms before age 45 years, and juvenile cases frequently occurring with disease onset in the early twenties (Assoni et al., 2022). Most pathogenic FUS mutations cluster in the C-terminal nuclear localization sequence, disrupting normal nuclear import and leading to cytoplasmic accumulation of the protein (Moens et al., 2025). Pathogenic mutations in both genes are associated with a similar mislocalization of the respective proteins from the nucleus to the cytoplasm and the formation of cytoplasmic inclusions, highlighting the importance of these cellular processes and impaired protein homeostasis in ALS pathogenesis (Lattante et al., 2013).

Other ALS genes, though individually rare, have illuminated additional pathogenic pathways (Nijs and Van Damme, 2024). For example, TBK1 mutations affect autophagy regulation (Freischmidt et al., 2015), NEK1 variants, together with its binding partner C21orf2, are associated with DNA damage response (Kenna et al., 2016; Gregorczyk et al., 2023), KIF5A mutations impair axonal transport (Nicolas et al., 2018; Baron et al., 2022), and ATXN2 intermediate repeat expansions modify TDP-43 toxicity and influence disease risk (Elden et al., 2010; Glass et al., 2022; Vieira de Sá et al., 2024). Recently, techniques such as Genome-Wide Association Studies (GWAS) and Whole Exome Sequencing (WES) have identified 12 novel genes associated with ALS. These genes are primarily involved in cellular functions like protein homeostasis, DNA repair, RNA metabolism, vesicle transport, and mitochondrial function (Wang et al., 2023).

Despite genetic heterogeneity, ALS-associated mutations converge on several interconnected pathophysiological pathways, providing a framework for understanding both disease biology and therapeutic targets (Fig. 1B). The precise mechanisms that drive neurodegeneration in ALS remain incompletely understood. While comprehensive mechanistic reviews exist elsewhere, we provide a focused overview of key pathways most relevant to therapeutic development, discussed in subsequent sections.

Protein homeostasis dysfunction represents a central feature of ALS pathology. The accumulation of ubiquitinated TDP-43 inclusions in motor neurons occurs in over 97% of cases (Neumann et al., 2006). This pathological mislocalization of TDP-43 from nucleus to cytoplasm creates a dual pathogenic mechanism, causing both loss of normal nuclear RNA processing functions and toxic cytoplasmic gain-of-function effects, including altered RNA splicing, increased cellular stress responses, and DNA damage (Mitra et al., 2019; Suk and Rousseaux, 2020; Dyer et al., 2021; Benson et al., 2021).

Similarly, C9orf72 mutations disrupt protein homeostasis through complementary mechanisms, with loss of normal C9orf72 function impairing autophagy initiation while the pathogenic DPRs produced from the expanded repeat undergo liquid-liquid phase separation, creating additional toxic species (Webster et al., 2016; Boeynaems et al., 2017). Mutations in SOD1 and FUS also produce characteristic protein aggregates (Watanabe et al., 2001; Dormann et al., 2010), collectively suggesting that the failure to maintain proper protein folding and clearance represents a fundamental pathogenic mechanism across ALS subtypes.

Oxidative stress and mitochondrial dysfunction form a self-amplifying cycle in ALS. Mutant SOD1 loses its superoxide scavenging function and aggregates within mitochondria, impairing electron transport chain complexes (Tsang et al., 2014; Xu et al., 2022). TDP-43 pathology also contributes to this process, as cytoplasmic TDP-43 aggregates sequester nuclear-encoded mitochondrial proteins, creating a cascade where protein mislocalization leads to mitochondrial dysfunction and increased oxidative stress (Colombrita et al., 2009; Zuo et al., 2021). Additionally, C9orf72 DPRs have been shown to directly bind and impair mitochondrial complex V (Choi et al., 2019), demonstrating how multiple genetic forms of ALS converge on mitochondrial pathways. These mechanistic insights have directly informed therapeutic development, establishing the rationale for targeting oxidative stress and mitochondrial dysfunction with drugs like edaravone.

Neuroinflammation in ALS represents a complex and dynamic process involving multiple cell types with potentially opposing roles depending on disease stage (Liu and Wang, 2017). The immune response is coordinated primarily by resident CNS cells, where microglia undergo a phenotypic transition from initially neuroprotective M2 states toward proinflammatory M1 activation as disease progresses (Liao et al., 2012; Roberts et al., 2013; Hooten et al., 2015), suggesting that timing may be critical for anti-inflammatory interventions. Concurrently, astrocytes contribute to motor neuron toxicity through multiple mechanisms, most notably by promoting glutamate excitotoxicity, which has provided the mechanistic basis for riluzole therapy (Rothstein et al., 1996). The pathogenic role of astrocytes is further supported by studies showing that astrocytes derived from ALS patient fibroblasts are directly toxic to motor neurons in co-culture models (Meyer et al., 2014), while genetic studies demonstrate that removing mutant proteins specifically from astrocytes can slow disease progression in animal models (Wang et al., 2011).

Stress granule dysfunction represents a central convergent pathophysiology in ALS linking multiple genetic and molecular mechanisms. Stress granules are protective cytoplasmic assemblies that normally form during cellular stress through liquid-liquid phase separation, but in ALS they become pathological sites of protein aggregation. The oxidative stress and mitochondrial dysfunction discussed above create chronic cellular stress conditions that promote persistent stress granule formation (Ratti et al., 2020; Dudman and Qi, 2020). Meanwhile, the protein homeostasis dysfunction manifests directly in this pathway, as mutations in RNA-binding proteins, including TDP-43, FUS, and ATXN2 disrupt normal stress granule dynamics, causing them to mature into solid pathological aggregates rather than resolving after stress relief (Becker et al., 2017; Besnard-Guérin, 2020; Mariani et al., 2024; Di Timoteo et al., 2024). This creates a pathological cycle where stress granules become nucleation sites for the very protein aggregates that characterize ALS pathology, while C9orf72 DPRs further disrupt both stress granule assembly and clearance mechanisms (Boeynaems et al., 2017). Thus, stress granule dysregulation exemplifies how multiple pathogenic mechanisms converge into common pathological endpoints in ALS.

Mouse models have provided essential tools for ALS research, with each capturing only selected aspects of disease biology (Lescouzères and Patten 2024; Zhou et al., 2024). The wobbler mouse, first described in 1956 (Falconer, 1956), was the earliest proposed model of motor neuron disease, developing progressive upper and lower motor neuron degeneration. Following SOD1 discovery, the SOD1-G93A transgenic mouse became the most widely used model, but it lacks TDP-43 pathology and shows rapid progression unlike human disease (Gurney et al., 1994). TDP-43 models have proven challenging, with different expression approaches yielding varying outcomes regarding motor phenotypes and inclusion formation (Swarup et al., 2011; Lescouzères and Patten, 2024). FUS models require gain-of-function mutations to recapitulate disease features (Sharma et al., 2016), while C9orf72 models often produce inflammatory rather than motor neuron phenotypes, though some approaches can replicate key pathological features (Chew et al., 2015; Herranz-Martin et al., 2017). This diversity reflects the complexity of modeling human ALS genetics and highlights ongoing challenges in developing clinically relevant preclinical tools.

While ALS pathophysiology remains mysterious and highly heterogeneous, this complexity fundamentally informs therapeutic development challenges. Different mechanisms likely play varying roles in individual patients, explaining both the limited efficacy of broad-spectrum approaches and the promise of precision medicine strategies. This review’s primary focus is critically evaluating the therapeutic landscape rather than comprehensively reviewing pathophysiology. We synthesize evidence from pivotal randomized controlled trials (RCTs) and real-world evidence (RWE), examining how different therapeutic strategies have succeeded or failed in targeting these underlying pathways. By analyzing currently approved medications and therapies with definitive clinical results, we emphasize the divergent outcomes of gene-targeted versus broad-spectrum approaches and identify lessons that might reshape future ALS drug development strategies.