1.1. Mutations Causing Amyotrophic Lateral Sclerosis (ALS)

ALS is a heterogeneous neurodegenerative syndrome leading to muscle weakness and eventually atrophy. In Europe, the median incidence rate is 2.08 per 100,000 population (1.47–2.43), corresponding to 15,355 (10,852–17,938) cases [1]. In the United States, for 2018, the National ALS Registry found 21,665 persons meeting the case definition for a prevalence of 6.6 per 100,000 persons [2]. While 1%–10% of patients autoreport a familial predisposition for ALS (fALS), screening of ALS gene panels has revealed that at least 20%–36% of all patients irrespective of reported medical history carry a claimed pathogenic variant in one or more ALS genes. fALS is inherited in an autosomal dominant or recessive mode, frequently with reduced penetrance, and pleiotropy has been demonstrated for mutations in at least three genes [3]. In 1993, missense mutations in the small gene encoding the free radical scavenging enzyme Cu-Zn-SOD (SOD1) were linked to causing fALS in 13 of 18 studied pedigrees with high-penetrant ALS in multiple generations [4]. This sparked a global screening for mutations in SOD1 and since then, some 234 mutations have been claimed to cause ALS. Though SOD1 mutations are rare in some seemingly similar populations (e.g., prevalent in the United Kingdom, rare in Ireland) in many countries mutations have been found in 7%–23% of diagnosed fALS cases but also in 3%–7% of sALS cases [5–7].

SOD1, a homodimer containing copper and zinc ions, is expressed constitutively at a high level, with differences between tissues and cell types, and can be detected in most subcellular compartments including cytoplasm, in the mitochondrial intermembrane space, nucleus, and endoplasmic reticulum [8–11]. SOD1’s established role is to scavenge superoxide free radicals generated in the normal cell metabolism to hydrogen peroxide and water. Other reported functions include involvement in ferroptosis, as a transcription factor, a regulator of RNA stability, or a component of nutrient signaling [12–14]. SOD1 is ubiquitously and constitutively expressed in all cells in the human body and accounts for up to 1% of total cytoplasmic proteins in some areas of the CNS [15].

Although the SOD1 protein has been extensively studied, the pathomechanism by which SOD1 mutations cause ALS remains elusive. Originally, SOD1 mutations were proposed to cause ALS via a loss-of-function mechanism, but currently, the gain of function mechanism is favored. Several mechanisms have been proposed to cause the disease upon SOD1 inactivation: excitotoxicity, oxidative stress, ER stress, mitochondrial dysfunction, axonal transport disruption, prion-like propagation, and noncell autonomous toxicity of neuroglia [16]. Since 1993, over 200 different SOD1 mutations distributed over the entire gene have been described, mostly missense, with quite a few mutations affecting the SOD1 polypeptide length (e.g., nonsense, insertion, deletion, and splice site variants) [17]. Most SOD1 “structural” variants are not deemed pathogenic by the American College of Medical Genetics and Genomics (ACMG) standards (similarly to the missense ones); moreover, for many of them, loss-of-function ALS mechanisms have been proposed.

This study is aimed at analyzing the phenotypic characteristics of the ALS-causing SOD1 mutations that affect the polypeptide length with a particular stress on the ages at symptom onset and death.

2.1. Statistical Analyses

Associations between individual SOD1 mutations carried by ALS patients and their clinical phenotype were investigated using survival analysis methods. For most patients, we were only able to collect data on two clinically relevant endpoints defined to estimate disease progression: age at symptom onset and age on death (overall survival). Kaplan–Meier curves showing survival for patients stratified by specific SOD1 mutation or group of mutations (frameshift (FV) and premature stop but nonframeshift variants (nFVs)) were compared using the log-rank test (for two curves) or the chi-square test (for more than two curves). Associations of endpoints with mutation features were analyzed using a Cox proportional hazards regression model. The age of ALS onset was compared between mutations by the Mann–Whitney test. p < 0.05 was considered statistically significant. Statistica 13 was used for calculations.

3.1. Associations of Clinical Endpoints With Individual Mutations

Statistical analysis was performed on 29 SOD1 mutations (excluding two homozygous mutations causing juvenile ALS: p.Cys112Trpfs^∗11 and p. Val119del) found in 128 individuals from 44 families. Due to a lack of clinical data, it was not possible to stratify patients by clinical phenotype (e.g., site of onset or personality changes, psychiatric symptoms suggestive of FTD) or clinically relevant endpoints (e.g., wheelchair use or respiratory failure). However, survival analysis of age at onset, age at death, and overall disease duration from onset to death were informative. Only pooled family data were available for five mutations and 28 patients (p.Val30Aspfs^∗8 (n = 12), p.Val119_Val120insPheLeuGln (n = 2), p.Val119_Val120insPheLeuGln (n = 2), p.Gly28_Pro29del (n = 8), and p.Lys129Glyfs^∗6 (n = 4)).

When patients were considered individually, mean data from the relevant family was assigned to each member of the family. Generally, earlier age of onset turned out to be a significant, strong factor for earlier age of death (HR = 0.725 per year, 95% CI: 0.725–0.804, p = 3 × 10⁻²⁴). A significant difference (p < 10⁻⁷) was found between FVs and nFVs for all analyzed variables: protein length, first affected amino acid, number of affected SOD1 structures, as well as clinical data: age of onset and age of death (Table 1, Supporting Information 2). The year of publication was significantly different between the two groups (median year 2012 for FV and 2018 for nFV, p = 0.0051). FV significantly increased the risk that carriers will develop ALS (HR = 2.553, 95% CI: 1.722–3.786, p = 3.09 × 10⁻⁶). Because the FV and nFV groups were significantly different as regards most crucial variables, further statistical analyses were performed for each group separately.

In the nFV group, the p.Gly28_Pro29del mutation affecting the v-loop of SOD1 was the least severe (HR = 0.224, 95% CI: 0.069–0.727, p = 0.013) compared with all other nFVs as regards early ALS onset, and indeed the presence of homozygous patients further confirms this finding. On the other extreme was the p.Lys79_Arg80insSerIle mutant affecting the zinc-binding domain, found to be the strongest predictor of early ALS onset, but this conclusion should be considered with caution since only a single patient with this mutation has been reported. The distance in the protein sequence of the first affected amino acid from the N-terminus of SOD1 was positively associated with an earlier development of ALS at a borderline statistical significance (HR = 1.008 per one amino acid, 95% CI: 1.000–1.016, p = 0.051) and with shorter mean disease duration (HR = 1.008, 95% CI: 1.000–1.016, p = 0.04). The HR value of only 1.008 seems to be extremely close to 1, but it should be stressed that it is calculated per one amino acid, counting from the SOD1 N-terminus. Thus, a mutation further downstream in the protein sequence by, for example, 10 amino acids would increase the risk of earlier ALS onset and of shorter survival with ALS by a nontrivial 8%.

In the FV group, mutations affecting the electrostatic loop close to the C-terminus of SOD1 were associated with later ALS onset (HR = 0.393, 95% CI: 0.203–0.763, p = 0.0057) and were also associated with a longer survival from ALS onset to death (HR = 0.266, 95% CI: 0.131–0.540, p = 0.00024). Interestingly, variants close to either the N- (AA 1–90, log-rank = 2.189771, p = 0.02854) and C-terminus (AA > 142, log-rank = 2.92443, p = 0.00345) of SOD1 were risk factors of earlier ALS onset compared with the mutations affecting the middle part of the protein (AA 91–142, see Figure 1). This feature is easily explained in the case of the proximal mutations but not for distal ones: A frameshift mutation close to the N-terminus of the protein affects a larger section of the polypeptide sequence than a more distal mutation. One can only speculate that an SOD1 molecule with most of its amino acid sequence, and plausibly also tertiary structure, identical with a native one and a disturbed relatively small C-terminus part could be more “toxic” than the mutant disturbed in most of its C-terminus region.

The risk factor for early death age was FV, and in this group, the mutation in the electrostatic loop (affecting the C-terminus of SOD1) was a protective factor.

No difference in survival time from ALS diagnosis was associated with patients’ sex or family history of the disease (fALS vs. sALS).

Frameshift mutations occur in a finite number of protein sequence C-termini combinations. In our data set, 16 FVs occur in 11 different C-termini motifs, with HisThrGlyGlyPro (HTGGP) common for three mutations: p.(Val98Leufs^∗17), p.Ser108Leufs^∗15, p.Cys112Trpfs^∗11, and GlnArgTrpLys (QRWK) for four mutations: p. Leu127Glyfs^∗6, p.Leu127SerfsTer7^∗, p.Lys129Glyfs^∗6, and p.Asn132Glnfs^∗5. We did not observe any significant difference in clinical data (age of onset and age of death) compared to mutations that do not impact the protein sequence on C-terminus.

This study analyzes for the first time a comprehensive dataset of age at onset, age at death, and disease duration in ALS patients with nonmissense SOD1 mutations. Sex or familial disease history did not affect ALS onset, disease duration, or age of death. In contrast, different mutations were associated with different ages of symptom onset and death. We provide evidence that through adult life without symptom onset ranging from 21.4 to late 84 years (mean: 48.4), the shift toward younger ages of onset is evident in the penetrance in individuals with nonmissense SOD1 mutation compared to ALS patients without identified mutation in associated genes, and is comparable to ALS with missense SOD1 mutations.

Although SOD1 mutations are generally considered fully penetrant, we observed incomplete penetrance more often than in missense SOD1 mutations, which might be age related; however, this observation needs further investigation.

Guissart et al. [39] showed that all pathogenic premature termination codons (PTCs) described so far in SOD1 can theoretically escape nonsense-mediated mRNA decay (NMD), resulting in the production of a truncated protein, supporting the hypothesis that haploinsufficiency is not an underlying mechanism of SOD1 mutant-associated ALS. Those authors found that even when there is massive degradation of mRNA harboring a PTC in the NMD-triggering region, a small residual amount of mutant mRNA could still be detected. Here, we did not stratify the dataset according to the likelihood of NMD. Guissart and colleagues also suggested that PTCs found in NMD-triggering regions are not pathogenic, but such variants have in fact been identified in ALS patients (e.g., p.W33^∗). For all patients with nonmissense mutations for which SOD1 activity has been determined, its level was decreased, similarly to missense SOD1 mutations [40]. For at least two homozygous mutations (p.Val119del and p.Cys112Trpfs^∗11), a complete loss of SOD1 function was observed [11, 24, 30]. These findings suggest SOD1 haploinsufficiency as a pathological factor, at least coexisting with gain of function.

The relationship between gain and loss of function of SOD1 in the pathomechanism of ALS should be further investigated. First, the ALS-causing mutations that cause gain of function must also cause loss of function of SOD1. On the other hand, there are patients who have ALS due to mutations that cause only loss of SOD1 function. These seemingly contradictory mechanisms can coexist, either attenuating or enhancing each other depending on a time point in the biological process or environmental factors. Our results highlight the need for functional studies of mutations that result in truncated SOD1 protein.

Among the 38 mutations studied here, four were intron ones affecting splicing: c.240-7T>G (p.Lys79_Arg80insSerIle), c.358-304C>G (p.Val120_Gln154delinsGlnLeuLysLysLeuProLys), c.358-11A>G (p.Val120_Gln154delinsPhePheThrGly), and c.358-10T>G (p.Val119_Val120insPheLeuGln). All of them were associated with fALS and considered as pathogenic mainly because of the availability of extensive family trees with access to samples allowing the respective authors to prove cosegregation. Since most ALS patients are sporadic and most familial cases offer only a limited sample access, it seems plausible that numerous ALS cases of undetermined origins could be explained by such mutations, which are routinely omitted in most analyses as variants of uncertain significance. This implies the need to report and investigate all variants, including those in the noncoding regions.

Family history was reported for 35 out of the 38 mutations with clinical information provided. In the 23 mutations (66%), and 123 of 140 cases, 88% were described as causal for fALS, which is comparable to that reported for all SOD1 mutation carriers worldwide (84%) [41]. This supports the true pathogenicity of these mutations, but this shift toward familial cases may indicate that in sporadic cases SOD1 mutations could be underreported. It is likely that in the case of identified SOD1 variants, especially the sporadic ones, the same mechanism can be observed as in the case of published negative results. This is even more likely if one considers that the absence of a family history may be due to incomplete penetrance, inadequate recording, or small family size, which may mask familiarity. At the same time, the low penetrance of the mutation in our dataset is noted, for example, p.Gly28_Pro29del and p.Glu134del mutations were identified in both sporadic and familial cases. To understand the pathomechanism of SOD1 mutations, it is crucial to encourage both scientists and publishers to publish case reports that do not necessarily prove pathogenicity.

Individual SOD1 nonmissense mutations show differences between them in terms of age at symptom onset, age at death, or disease duration, which are also observed in missense SOD1 variants. The underlying mechanism remains unclear but consistent disease course in different families or environments (same mutations identified in different countries and continents) may lean towards the nature of the individual mutation itself.

SOD1 mutations have been annotated variably in the literature, complicating comparison of carriers of the same mutation across reports. Most of the mutations were published without counting the first methionine, but we also observed errors in the predicted protein consequence (e.g., the HGVS Nomenclature p.Leu107_Ser108del vs. the literature nomenclature S105deltaSL, the HGVS Nomenclature p. Gly130Trpfs^∗3 vs. the literature nomenclature p.Lys129fsTer5 or Lys128 fsX131). Our study organizes protein nomenclature according to the HGVS Nomenclature recommendation inferred from reported cDNA annotation or an attached electrophoretogram of Sanger sequencing.

Nonmissense variants account for 14% of the 165 SOD1 variants affecting the protein sequence reported in the gnomAD (April 17, 2024) (frameshift n = 11, inframe deletion n = 4, stop gained n = 5, start lost n = 1, stop lost n = 2, missens = 144), which is slightly less than among the variants identified in ALS patients (Supporting Information 4) [42]. Those nonmissense variants are rare (identified in 1 up to 11 individual subjects, the latter in the case of p.Ter155SerextTer6) and annotated as heterozygous. p.Trp33^∗, p.Asp97Metfs^∗8, p.Leu107_Ser108del, p.Asp126Thrfs^∗24, and p.Glu134del have also been reported in ALS patients. The presence or absence of a variant in the gnomAD does not indicate its pathogenicity, nor does it mean that variants will not be identified in ALS patients, as the gnomAD contains whole exome and whole-genome sequences aggregated from many large-scale sequencing projects. Well-established ALS pathogenic variants such as p.Ala5Val are also present in the gnomAD. Noteworthy, there are rare variants that most likely affect splicing in the gnomAD that have also been identified in ALS patients but (with a few exceptions) were not further studied in the context of ALS.