At present, the pathogenesis of Amyotrophic Lateral Sclerosis (ALS) is not fully understood. Patients may wait as long as one year for a definitive diagnosis, which is still based on clinical criteria. In this regards, the identification of genetic hallmarks would greatly improve the diagnostic path, in particular for sporadic ALS forms. Short tandem repeats (STR) expansions have recently been found in patients with a clinical diagnosis of ALS as potentially causative of the disease and therefore as possible clinical biomarkers. Most of the previous studies identified STR expansions using Short Read Sequencing (SRS). Thanks to technology improvement, Long Read Sequencing (LRS) have recently become more accessible.In this paper, we compared SRS and LRS on a cohort of sALS patients and showed that SRS might fail in identifying pathological repeats as well as misidentify existing pathological repeats.Thus, we believe that our findings will be relevant to the broad readership of your journal, and be of inspiration for future studies that, by including large dataset and the proper sequencing technique, will help elucidating the ALS molecular mechanisms and consequently identify potential therapeutic targets.

Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disorder presenting progressive weakness of the bulbar and extremity muscles, leading to a wide-ranging clinical phenotype like Bulbar, Pseudobulbar and Limb ALS, and Limb and Mill’s variant [1–3]. Symptoms onset occurs between 58 and 63 years with a prevalence, among the European population, of 2.2 per 100,000 individuals and a higher incidence in males than in females [4, 5]. Cognitive impairment occurs in up to 50% of cases, with 15% of patients diagnosed with frontotemporal dementia [4]. Approximately 85% of ALS cases are sporadic (sALS) whereas the remaining 10%–15% are familial (fALS) [4, 6], both showing similar clinical presentations [5, 7]. ALS etiology is the result of numerous factors including genetic susceptibility, age-related cellular damage, and environmental exposures among which gender, geographical region, smoking, sportive activities and lead exposure [8, 9].

At present, approximately 30 genes have been associated to ALS [10–12]^. Four of them: C9ORF72, SOD1, TARDBP and FUS, account for approximately 60%–70% of familial ALS cases and 6%–10% of sporadic ALS cases, listed in order of decreasing frequency [13]. While known ALS disease genes account for a minority of sporadic cases, recent research highlights the potential role of noncoding structural variants and gene copy number variations in sALS susceptibility and phenotype modification [13]. Interestingly, the pathogenic form of the chromosome 9 open reading frame 72 (C9ORF72) gene is a G4C2 hexanucleotide repeat expansion (HRE) in the intron 1 between the non-coding first exons 1a and 1b [14]. C9ORF72 is currently the only short tandem repeat (STR) expansion proven to cause ALS and frontotemporal spectrum disorder (FTD), however, different expanded STRs distinctive of other neurodegenerative disease like ATXN1 (spinal cerebellar ataxia type 1 (SCA1)), ATXN2 (SCA2), ATXN8 (SCA8) and HTT (Huntington’s disease) [15] have been found in clinically diagnosed ALS patients and FTD cases [16, 17]. Among them, the CAG trinucleotide expansions in ATXN2 have been identified as risk factors for ALS [4, 18]. Interestingly, in addition to the more than 30 genes associated to fALS, heritability studies suggest a 60% genetic component for sALS as well [19] suggesting that sALS is triggered by a complex genetic variation far from being understood.

In this context, STR expansions could represent a valuable starting point to further investigate and clarify the genetic predisposition of sporadic cases as well. Proving the association between STR expansions and sALS will also have beneficial effects on the diagnosis of the disease, which might take up to 1 year, and is still based on clinical, electrophysiological, and radiological investigations while genetic variants or other biomarkers tests are rarely taken into consideration [20].

Next generation sequencing (NGS), like Illumina short-read sequencing techniques (SRS), and the development of various computational methods set the scene for genome-wide STR detection [21]. It has been shown that, due to technical limitations, SRS methods often lack sensitivity and specificity for detecting a significant proportion of structural variants (SVs) and tandem repeats [22]. These limitations can be addressed by long-read sequencing (LRS) which, unlike SRS, can directly sequence long repeat regions without the need for fragmentation. For example, SRS generates reads of 100–150 base pairs, which is much smaller than the thousands of base pairs typical of pathogenic STR expansions. Consequently, SRS is limited in detecting pathogenic STR expansions, but can still identify smaller STRs using specialized tools like GangSTR or ExpansionHunter. In contrast LRS may generate reads up to two megabases (Mb) in length allowing for more efficient detection of larger STR expansions [23]. In this context it is important to note that most reads from LRS platforms (e.g., PacificBiosciences (PacBio), Oxford Nanopore) are typically shorter ranging from 10 to 100 kb for PacBio and 20–200 kb for ONT) [24].

Long read NGS instruments have been on the market for the past decade. Initially, the lower yield, higher error rate, and higher costs of the instruments, have kept them from being more widely adopted. More recently, PacBio (PacBio) and Oxford Nanopore Technologies (ONT) have both been working successfully to make LRS more accessible. These technologies, with the aid of several available computational tools, such as “ExpansionHunter, STRetch, and Tandem Repeat Finder” [21], use information from flanking sequences to provide better alignment for pathogenic STRs. Being a relatively new technique, only a few studies have applied LRS to characterize disease-associated STRs [25–28]. In contrast, most of the previous studies on the characterization of ALS-associated pathogenic STRs were performed using SRS [16, 29, 30].

In our study, we used SRS to sequence the DNA from 47 sALS patients to investigate pathological STR expansion. LRS was then used to reanalyze samples from two patients, as the identification of pathogenic STR expansions raised ambiguities in the interpretation of their STR lengths.

As mentioned above, LRS offers distinct advantages over SRS, including the ability to directly sequence long repeat regions and accurately determine STR sizes, which is crucial for precise quantification of STR expansions.

Twelve, potentially pathogenic, STR loci located in the following genes: C9orf72, ATXN2, ATXN1, ATXN7, FMR1, DM1-AS, PPP2R2B, ATXN8OS, HTT, CACNA1A, ATXN3, and TBP, were genotyped in all the 47 sALS patients. Out of 564 genotypes, 308 (54.6%) passed quality filtering. Twenty-six genotypes (4.6%) failed level 1 general filters, 25 of these had only spanning and/or bounding reads, and 1 had low read depth. The remaining 230 genotypes (40.8%) failed the more stringent level 2 filter, which enforces a minimum call quality threshold to ensure precise repeat length estimation.

In each individual, on average, 56.4% (0%–92%) of the genotypes at the 12 loci passed filtering. No significant correlation was observed between the average sequencing depth and the number of loci genotyped in each individual (cor = 0.12, p = 0.42). Post filtering, the percentage of genotypes available at each loci ranged from 0% to 83% and there was a significant negative correlation with the number of genotypes called and the size of the repeat in the reference genome (cor = −0.72, p = 0.007). The only locus in which no genotypes passed the quality filtering for the STR located in the TBP gene which was the largest in the reference genome (114bp). More than half of the failed calls were due to the absence of reads fully enclosing the repeat, indicating that longer-read sequencing covering the entire repeat and its flanking regions is necessary for accurate genotyping. This underscores a key limitation of short-read sequencing when genotyping larger genomic repeats. The tandem repeat data are summarized in Table 2.

Out of 47 patients, 46 showed tandem repeat lengths within the normal range (Table 2). Only one patient (patient 21) showed pathogenic STR expansions in the HTT (40 CAG length), ATXN2 (36 CAG), and CACNA1A (46 CAG) genes, and intermediate length STR in the ATXN3 (50 CAG) and DM1-antisense RNA (39 CTG) genes. Patient 21 was further investigated by WGS LRS. The sequencing reads from patient 21 were visually inspected in the bam file using Integrative Genomic Viewer (IGV) to determine the repeat lengths, the LRS did not support the STR calls from the SRS data (Table 3).

When analyzing the C9orf72 intron 1, located between the non-coding first exons 1a and 1b, for potentially pathogenic hexanucleotide repeat expansion (HRE), the SRS did not reveal any allele neither in the intermediate nor in pathological length range. However, inspection of the calls made by GangSTR for the C9orf72 repeats, prior to filtering, suggested that two samples (patients 28 and 29) might carry pathogenic expansions at this locus. Such calls were filtered out due to low quality.

Moreover, the number of soft clipped reads observed over the C9orf72 loci (Figures 1a,b) also suggests the presence of STR expansions for both patient 28 and 29.

LRS was then performed for patient 28 only because the DNA available for patient 29 was insufficient for long-read library construction (unfortunately patient 29 died before obtaining the SRS data). LRS for patient 28 (Figure 1c) showed two reads over the C9orf72 hexanucleotide repeat containing large expansions: the first one containing an additional 4628 bp (771.3 repeats) and the second one containing an additional 6285 bp (1047.5 repeats), confirming the presence of the pathogenic C9orf72 STR expansion in the genome of patient 28 (Table 3). From a pathological standpoint, patient 28 exhibited, in addition to motor neuron disease symptoms, a dementia syndrome suggestive of frontotemporal dementia. The disease progression was very rapid. In contrast, patient 29 followed a typical ALS course, beginning with bulbar paralysis followed by the emergence of additional typical ALS symptoms.

ALS is a neurodegenerative disorder presenting phenotypic and genetic heterogeneity [37] with a multifaceted molecular basis difficult to characterize. The identification of reliable biomarkers could positively impact the understanding of the underlying disease mechanisms and, consequently, patient diagnosis and management. In this context, Feldman et al. [4] recently proposed to replace the categorization of sALS and fALS cases with a new binomial: genetically vs. non-genetically confirmed forms, respectively, underlying the importance of genetic testing in disease characterization. Despite the approximately 30 genes already associated to ALS [10–12] detecting, genetically, sALS forms is challenging because they display a clear genetic background in a minority of patients only [38]. Furthermore, the challenge increases when STR expansions are involved as predisposing mutations. For instance, no consensus on a specific disease-related threshold for various polyglutamine-associated disorders has been reached, since healthy individuals may also carry expansions in the pathological range [39]. Although C9orf72 expansions have been extensively associated with ALS/FTD, a disease-causing cut-off for the hexanucleotide repeats is still questioned [40]. Both healthy and affected individuals show repeats in the intermediate range (20-30 hexanucleotides), confirming that the pathological role of intermediate STR expansions is far from being understood [4, 41–45]. Furthermore, with the exception of C9orf72, the abundance of STR expansions in ALS patients compared to healthy control subjects is often narrow [30, 46]. For these reasons, the presence of STR expansions must be evaluated with a robust and reliable sequencing technique, displaying both high diagnostic sensitivity and specificity. Our study showed that even very long STR expansion might not be properly identified by SRS and, on the other end, SRS can falsely identify STR expansion not present in the subject genome. This could result in false negatives, delaying ALS diagnosis and hindering timely clinical management, access to disease-modifying therapies, multidisciplinary care, and clinical trial participation [47]. Although less common, false positive could lead to unnecessary and potentially anxiety-inducing tests.

Thus, LRS must be the preferred choice genotyping a patient DNA in search of pathological STR expansions.

From an epidemiological point of view, despite the limitation due to the relatively small patients cohort and the recruitment from a single Neurology Department, our study seems consistent with previous ones showing a 5–10% of sALS patients carrying C9orf72 STR expansion [16]. Because patient 29 showed SRS characteristics similar to patient 28, we might speculate that both carried pathological STR expansion. However, because no DNA was available for LRS, our hypothesis remains a conjecture. Thus, 4–5% of our sALS cohort showed the presence of expanded C9orf72 hexanucleotide repeats.

Our study highlighted the benefits of LRS for accurate characterization of large tandem repeats: SRS identified multiple REs in a patient which were not confirmed by long read sequencing. Conversely, in another patient, unfiltered calls from GangSTR (that did not pass the quality filtering) as well as manual inspection of the bam files suggested the presence of expanded alleles in C9orf72 which were further confirmed by LRS. This could lead to ALS misdiagnosis, resulting in either false negatives or false positives, along with the various problems associated with these types of medical errors. These point out SRS limitations in evaluating broader repeat sequences and large genomic rearrangements (32) and recommend the use of LRS to flank ALS clinical diagnosis [48].