With the aim of identifying a single miRNA or a miRNA profile with diagnostic ability in sALS, we collected all the papers listed on PubMed, searching for “diagnosis and sporadic amyotrophic lateral sclerosis and miRNA” (Figure 2). The query revealed the publication of 17 articles (20 November 2021); from all these articles all those regarding genetic forms of fALS (21); those describing genetic animal models (22, 23); those not specifically related to sALS (24), all the reviews on diagnostic miRNAs in genetic and sporadic forms of ALS (16) and those non-accessible (25, 26) were excluded.

The methods used in all these articles to obtain sALS-associated miRNA profiles include microarray analyses, next generation sequencing (NGS), but also the analysis of a single miRNA by RT-qPCR in different populations, comparing healthy subjects or patients with other neurodegenerative pathologies to sALS patients. miRNAs have been isolated from either spinal cord tissues or from blood, serum or plasma samples.

In order to assess the diagnostic role of literature-identified miRNAs, we implemented a Naive Bayes classifier (27), using an R Bioconductor package, e1071 (28). Our classifier was applied on an ALS dataset downloaded from the Gene Expression Omnibus (GEO), GSE52917 (29). It contains miRNA expression levels extracted from human serum and we selected the expression levels of 18 sALS patients and 17 controls. The aim of Naïve Bayes classifier was to develop: (i) a predictive model which is able to learn how to distinguish sALS patients and controls in the training set, and (ii) use the generated predictive model to classify new samples in sALS or control in the testing set.

First, we normalized the expression levels of miRNAs performing the Min-Max Normalization method (30). Then, we divided the original data into a training (70%) and testing set (30%). The training set is represented by a matrix of samples belonging to sALS patients and controls (known classes) and selected miRNA expression levels for each sample. The testing set is composed of a matrix of samples of unknown classes with miRNA expression levels.

In order to verify the efficiency of our classification, we evaluated three parameters: accuracy, sensitivity, specificity in the testing dataset. Accuracy is defined as the number of correctly classified samples divided by the total number of classified samples. Sensitivity and specificity indicate the ability of the classifier to correctly classify the samples with and without a disease, respectively.

Microarray analysis of spinal cord samples of five sporadic ALS and five control patients with diseases other than neurodegenerative disorders revealed upregulation in 37 miRNAs, and the downregulation in 35 miRNAs (31). Further validation revealed a significant downregulation of miR-139-5p and a significant upregulation of miR-5572 in sALS. The prediction by TargetScan analysis of miR-5572 targets and the in vitro validation revealed that zinc transporters could be controlled by this miRNA. Zinc could be important as a cofactor for several biological activities.

The transcriptome deep sequencing of spinal cord ventral horns of post-mortem sALS human donors showed 1160 deregulated mRNAs and 18 miRNAs. The snapshot offered by transcriptomic profile revealed the downregulation of genes involved in relevant neurological functions (impulse transmission, synaptic transmission, calcium ion transport or neurotransmitter secretion) with a specific reduction of SNAP25 and STX1B gene expression, neuronal t-SNARE involved in vesicle trafficking and calcium dynamics. miRNA seq analysis revealed miR-124-3p, -127-3p, -133a, -136-3p, -182-5p, -183,5p, -218-5p, -219-5p, -409-3p, -410, -478b, -485-5p, -577, -873-5p, -889, to be downregulated, while miR-148a-3p, -155-5p and -221-3p to be upregulated in these tissues. Several of the decreased miRNAs are involved in TGFβ protein signaling, and in the control of phosphatidylinositol-3 kinase (PI3K)/AKT pro-survival pathway (32). Alteration in the level of miR-219-5p, found in human spinal cord by RNA-seq by D’Erchia et al. (32), was also observed in SOD1^G93A mice spinal cords possibly associated to gliosis (33).

In the first paper (34), the authors analyzed the expression of a single specific miRNA, miR-338-3p, in multiple samples, including blood leukocytes, serum, CSF, and spinal cord of sALS (n = 72 patients) compared to healthy controls (n = 62). In all these samples, this miRNA is overexpressed. The main potential target could be SLC1A2 (transporter that clears glutamate) and host gene apoptosis-associated tyrosine kinase (AATK), which is overexpressed during apoptosis of myeloid precursor cells by IL3 deprivation. The last process would link miR-338 to apoptosis of mature neurons observed in sALS.

The microarray analyses of 18 sALS sera compared to 16 matched healthy control sera revealed the downregulation of 11 miRNAs (35). Only two miRNAs were validated, miR-1234-3p and miR-1825. Of these two miRNAs, miR-1234 is able to distinguish sALS from fALS. Regarding their regulatory function, the functional network “Neurological Disease, Skeletal and Muscular Disorders, Cell-To-Cell Signaling and Interaction” seems to be the target of these two miRNAs, with NHPH3 and NLE1 being the two most significantly predicted common targets of the two miRNAs.

With a microarray-based approach, in Takahashi et al. (36) the authors characterized plasma miRNA profiles of 16 sALS patients and 10 healthy controls. The majority of miRNA were downregulated, and in the validation phase only miR-663-b, miR-4258, and miR-4649-5p were upregulated in sALS, with hsa-miR-663b and hsa-miR-4649-5p showing a negative correlation with disease duration from onset to end point (for details see Table 1). Regarding the pathways regulated by these miRNAs, also these authors highlighted that ALS disease may involve alteration in the synthesis and maturing processing of miRNAs, neuronal development and neurogenesis, transmission of nerve impulse (miR-663b), and cell proliferation, mitosis and protein folding (let-7f-5p). Some of these miRNAs (miR-4299 and miR-4649-5p) could be involved in cell adhesion, angiogenesis and axonal guidance, all processes involved in the sALS onset.

In the comparison of microarray profile of leukocytes from 5 Chinese sALS patients and 5 Chinese healthy controls, Chen et al. (37) found miR-34a, miR-100, miR-193b, miR-4485, significantly higher in sALS compared to healthy subjects, while miR-124, miR-183, miR-451, miR-3690, miR-3935, miR-4538, and miR-4701, were downregulated in the same comparison. The significant differences in the expression levels of four miRNAs (hsa-miR-183, hsa-miR-193b, hsa-miR-451, and hsa-miR-3935) were confirmed in RT-qPCR. The downregulation of miR-124 and the increase of miR-34a observed in microarray of sALS blood leukocytes (37) was also reported by Zhou et al. (33) in SOD^G93A animal model. After ROC curve analysis the authors claimed that the levels of miR-183, miR-193b, miR-451, and miR-3935 could be proposed as useful biomarkers for the diagnosis of sALS. Gene Ontology and KEGG database revealed the involvement of these miRNAs in the control of PI3K/AKT signaling pathway, MTOR signaling pathway, regulation of actin cytoskeleton, axon guidance, MAPK signaling, glioma and gap junction.

In Waller et al. (38), the authors used miRNA profile to identify a diagnostic signature in sera of sALS (n = 27) compared to that of healthy subjects or to patients with disease not related to ALS (n = 25). Fourteen significant diagnostic miRNAs in the comparison between sALS and healthy subject were found (for the details see Table 1). The validation phase on a separate case-control cohort confirmed that miR-143-3p, miR-206, (upregulated) and miR-374b-5p (downregulated) are differentially expressed between sALS and healthy subjects. In SOD1^G93A transgenic mouse model miR-206 expression, analyzed in murine spinal cord by RT-qPCR, was increased at the pre-onset of the disease (39) and this upregulation is similar to that found in sALS serum samples by Waller et al. (38). The overexpression of miR-143-3p increased during the development of the pathology, as well as the decrease of miR-374b-5p.

In the next generation sequencing (NGS) analysis performed on blood and neuromuscular junctions of 45 sALS versus 25 healthy subjects (40), the authors found 20 upregulated miRNAs and 22 downregulated miRNAs in blood, and 284 upregulated and 222 downregulated miRNAs in neuromuscular junctions of ALS compared to healthy subjects (Table 1). The validation analysis revealed that miR-223-3p, miR-326 and miR-338-3p were elevated in ALS neuromuscular junction compared to blood samples in 97% of the patients. Target prediction analyses revealed that these miRNAs regulate the brain-derived neurotrophic factor (BDNF) implicated in metabolic syndrome and neurodegenerative disease, but also the pathway of HIF1, including HIF1α, VEGFA, EGLN3, TFRC, and IGF. Some other targets are involved in neuronal compartment (ACTB, KALRN, MATN2, and RHOB), differentiation, morphogenesis and development.

miR-218 and miR-133a, decreased in blood and neuromuscular junctions of sALS patients (32, 38, 40) and miR-138 in spinal cords of sALS subject (32, 40), were found on the contrary upregulated in SOD1^G93A transgenic mouse model (41), possibly suggesting that these miRNAs in sALS control different genes and pathways from familial ALS.

miR-125b, downregulated in neuromuscular junction of sALS patiens (40), was found upregulated in spinal cord of SOD1^G93A mouse model (42). miR-9, downregulated in neuromuscular junction of sALS patiens (40), miR-124, downregulated in blood leukocytes (37) and in spinal cord (32) of sALS were also decrease in SOD1^G93A whole spinal cord (43).

Blood extracellular vesicles (EV) miRNAs from five sALS and age-matched healthy controls were isolated. The microarray analysis revealed 13 upregulated and 17 downregulated miRNAs (44). Among the possible biological processes regulated by these miRNAs, there are synaptic vesicle docking and exocytosis, positive regulation of neurotransmitter transport and secretion and synaptic vesicle cycle, suggesting a main role of synaptic vesicle-mediated transport in ALS development.

In the book chapter (45), the authors described all the possible genetic and epigenetic mechanisms that could generate altered expression of miRNAs found in fALS. The authors’ hypothesis is that the altered miRNA profile is a consequence of a deregulation in RNA transcription regulatory mechanisms, implying that miRNAs per se are not involved in the onset of the fALS pathology. In particular, they highlighted the role of TDP-43 protein as a modulator of AGO2. The lack of TDP-43 observed in fALS subjects could affect miRNA biogenesis, decreasing the expression of several miRNAs (miR-132-3p/5p, miR-143-5p/3p, miR-574-3p). Also, the loss of Dicer has been described in spinal cord degeneration, suggesting that the biogenesis and the processing of miRNA is critical in ALS development. The authors described a good correlation in the miRNA profiles between animal models and human samples. For instance, miR-155 has been described upregulated both in fALS in human and in SOD1^G93A animal model, perhaps for its regulatory role on TGFβf1, on the stimulation of proinflammatory macrophagic response and on increased secretion of cytokines. Similarly, miR-29 is highly expressed in humans and in the SOD1 mouse model. In human samples, as well as in mouse models, there is an increased activation of microglia, possibly supported by overexpression of miR-22, miR-125b, miR-146b, miR-155, and miR-365, being miR-125b involved in TGFβ regulation and miR-365 in IL6 production in animal models. Nothing is reported on miRNA alterations in the sporadic form of ALS. That is why this book chapter has not been included in our analytical review.

A remarkable event in ALS is also the mitochondrial dysfunction (46) due to the influence of miRNA activity on mitochondrial gene expression regulation. Indeed, mitochondrial miRNAs (mito-miRs) are a class of miRNAs that regulate mitochondrial gene expression and function, such as oxidative stress, cell metabolism, chemoresistance, and apoptosis. It could be useful, considering the role of mitochondria in sALS development by identifying those miRNAs whose loss or increase could impact on mitochondrial functions. miR-22 and miR-26b, altered in human blood samples of sALS (46), were found altered also in the spinal cord and brainstem of SOD1^G93A transgenic mice (33). In Catanesi et al. (46) miR-27a/b and miR-335-5p upregulation could modulate the mitochondrial dynamics and activates the pathway of caspase 3/7 in sALS. Similarly, the silencing of miR-151b is involved in the overexpression of PINK1, UCP2, and PKM, genes of neurodegeneration.

The analysis of sALS-associated miRNAs described in the literature is summarized in Table 1.

The analysis of the 10 articles in the literature on sALS identified 162 single miRNAs with potential diagnostic properties, coming from different profiles and tissues (spinal cord, CSF, blood, serum, plasma) although an agreement among publications on a single miRNA or a miRNA-panel is still lacking. For this reason, we proceeded with the bioinformatics analysis.

In order to assess the diagnostic role of 162 miRNAs presented in the first column of Table 1, we implemented a Naive Bayes classifier (27), using an R Bioconductor package, e1071 (28). The classification was performed on GSE52917, but as GSE52917 does not contain all 162 miRNAs the classification was performed considering the expression levels of 101 out of 162 miRNAs. The search criteria to identify GEO datasets for our analysis were the key words “sporadic Amyotrophic Lateral Sclerosis AND mirna AND serum” in Homo Sapiens (Access: 23 March 2022). GSE52917 was the only dataset with a suitable number of samples for each class (sALS and control).

The classification was applied considering: (i) the expression levels of all 101 miRNAs together, (ii) single miRNAs, and (iii) pairwise combinations of miRNAs.

We found an accuracy of 0.58, a sensitivity of 0.66, and a specificity of 0.50 when the classifier was performed on all 101 miRNAs together.

We improved the performance when the classifier considered some single miRNAs (Table 2). We obtained a better classification for miR-139-5p, miR-193b, miR-193b-1, miR-338-5p, miR-455-3p, miR-3911, miR-3911-1, miR-4687-5p, miR-4745-5p, and miR-4763-3p (cutoff of sensitivity > 60% and specificity > 60%). Specifically, miR-338-5p, miR-4745-5p, and miR-4763-3p achieved better performance.

Subsequently, we selected those miRNAs of Table 2 and we calculated Accuracy, Sensitivity and Specificity of classification using every possible pairwise combination of miRNAs. As shown, the best classification performance was achieved by miR-4745-5p (Accuracy = 0.83, Sensitivity = 0.83, Specificity = 0.83), although other miRNAs, such as miR-338-5p or miR-4763-3p, reached good classification performances. The Table 3 shows the pairwise miRNAs that obtained the better performance based on the condition of sensitivity and specificity > 70%. The analysis suggests the combination of miR-193b and miR-4745-5p expression could become a potential diagnostic tool, with an accuracy of 0.91, sensitivity of 0.83 and specificity of 1.

In summary, the classification analysis on single miRNA suggested 10 miRNAs (miR-139-5p, miR-193b-1, miR-193b, miR-338-5p, miR-455-3p, miR-3911, miR-3911-1, miR-4687-5p, miR-4745-5p, and miR-4763-3p) as potential diagnostic biomarkers of sALS. In addition, we also suggested the combined use of miR-193b and miR-4745-5p expression level evaluation as a sALS diagnostic tool.