SOD-1 Variants in Amyotrophic Lateral Sclerosis: Systematic Re-Evaluation According to ACMG-AMP Guidelines

Amyotrophic lateral sclerosis (ALS) is the most common type of motor neuron disease whose causes are unclear. The first ALS gene associated with the autosomal dominant form of the disease was SOD1. This gene has a high rate of rare variants, and an appropriate classification is essential for a correct ALS diagnosis. In this study, we re-evaluated the classification of all previously reported SOD1 variants (n = 202) from ALSoD, project MinE, and in-house databases by applying the ACMG-AMP criteria to ALS. New bioinformatics analysis, frequency rating, and a thorough search for functional studies were performed. We also proposed adjusting criteria strength describing how to apply them to SOD1 variants. Most of the previously reported variants have been reclassified as likely pathogenic and pathogenic based on the modified weight of the PS3 criterion, highlighting how in vivo or in vitro functional studies are determining their interpretation and classification. Furthermore, this study reveals the concordance and discordance of annotations between open databases, indicating the need for expert review to adapt the study of variants to a specific disease. Indeed, in complex diseases, such as ALS, the oligogenic inheritance, the presence of genes that act as risk factors and the reduced penetration must be considered. Overall, the diagnosis of ALS remains clinical, and improving variant classification could support genetic data as diagnostic criteria.


Introduction
Neurodegenerative diseases are a heterogeneous set of diseases of the central nervous system, characterized by the progressive degeneration of neurons' structure and function. ALS, also known as Charcot disease or Lou Gehrig's disease, is the most common motor neuron disease with a global annual incidence ranging from 1 to 3 cases per 100,000 population [1]. It is characterized by the progressive neuromuscular junctions dismantling and the motor neurons degeneration in the motor cortex, brain stem, and spinal cord. Loss of motor neurons leads to progressive paralysis and death from respiratory failure within 3-5 years of the onset of disease [2]. To date, despite ALS being known to be a multifactorial disease associated with genetic predisposition, environmental factors, and lifestyle, its causes are unclear. In 10% of cases, ALS is caused by a genetic mutation that can characterize several generations of the same family (familial ALS (fALS)) or the disease arises suddenly and without any known familial link (sporadic ALS (sALS)) [3]. The main genes directly associated with the onset of the disease and which account for almost 50% of fALS and 6% of sALS cases in the world are SOD1, TARDBP, FUS, and C9orf72 [4].
The first ALS gene associated with the adult-onset autosomal dominant form of the disease was mapped to the chromosome 21q22.1-22.2 and encodes for the cytoplasmic Cu/Zn superoxide dismutase (SOD1). Mutations in this gene were first described and linked to the disease in 1993 [5], and their frequencies range from 1% to 2% and from 13% to 20% in sALS and fALS patients, respectively [6]. SOD1 gene encodes a homodimer enzyme to the scoring rules expressed in Supplementary Figure S1, to choose a classification from the 5-level system.

Frequency Analysis of Variants
For the evaluation of the minor allelic frequency (MAF), gnomAD v2.1 (gnomad. broadinstitute.org) was used. According to the ACMG-AMP guidelines, if a variant exists in a large general population or a control cohort, it can be considered a criterion 2 moderate pathogen (PM2). Recently, Oza et al. [20] recommended downgrading the weight of the PM2 criterion to a supporting weight in guidance from the Sequence Variant Interpretation (SVI) working group (https://clinicalgenome.org/site/assets/files/5182/pm2_-_svi_recommendation_-_approved_sept2020.pdf, 27 January 2022). In our context, SOD1 variants completely absent from all databases or with a MAF < 0.001 met the moderate pathogen criterion 2 (PM2) [21,22].

Computational Tools Analysis
The pathogenicity of the variants was also evaluated using multiple in silico genomic tools, including SIFT [23], Polyphen-2 [24], MutationTaster [25], MutationAssessor [26], FATHMM [27], PROVEAN [28], and CADD [29]. Splicing prediction analysis was performed using SpiP and SpliceAI tools [30]. The convergence should be considered as supporting evidence of pathogenicity or benignity. An adverse consequence predicted with more than six tools is considered to be evidence to support pathogenicity (PP3), while at least three total and concordant predictors are required to support a benign state (BP4). However, if the in silico predictions disagree, this test should not be used for the classification of a variant.

Analysis of Functional Studies
Functional studies conducted in vitro or in vivo may show a detrimental effect on protein function, supporting the pathogenicity of the tested variant (PS3), or no effect, thus supporting a benign impact of the variant (BS3). As recommended by Brnich et al., several factors must be considered for the assignment of this criterion [13]. It is necessary to delineate the genetic mechanism of the disease to choose the appropriate functional analysis, as well as the experimental model to be used on which to perform functional tests. Assays performed must be coded in biological replicates, and an adequate number of control cases should be considered. Finally, the test result must be in line with the pathogenetic mechanism of the disease [13].
ALS is a complex biological system, and the precise mechanisms underlying selective cell death in the disease are unknown. Current understanding indicates that there may be a complex interplay between genetic factors, oxidative stress, excitotoxicity, protein aggregation, impaired axonal transport, and mitochondrial damage. Over the years, several in vitro and in vivo experimental models have been exploited to study the molecular processes underlying motor neuron degradation and disease progression, providing essential small pieces to the ALS puzzle [31,32]. However, only an integrated and complimentary use of different experimental models could improve our understanding of the pathogenic mechanism of the disease.
According to Brnich's recommendations and experimental evidence on SOD1 variants, we assigned three different weights to the PS3 criterion: PS3_Strong, PS3_Moderate, and PS3_Supporting. PS3_Strong was assigned to variants in which in vivo functional assay results showed toxic SOD1 function and disruption of mitochondrial dynamics, protein folding, axonal transport, and cellular metabolism. The PS3_Moderate and PS3_Supporting criteria were assigned to variants investigated by functional in vitro studies, following the flowchart shown in Supplementary Figure S2.

Critical Functional Domain Position
Pathogenic SOD1 variants associated with ALS are spread over the entire length of the gene. Hotspot mutations are characterized by the presence of many missense variants, in-Genes 2022, 13, 537 4 of 28 frame, and nonsynonymous variants, most of which have a pathogenicity classification. The universal protein source, UniProt (https://www.uniprot.org/, 29 January 2022), and bioinformatics tools can help identify mutations in established hotspots or functional domains. The PM1 rule considers reported variants within the domain and is assigned with the ratio pathogenic to total nonpathogenic variants is greater than 1.5, with at least 5 pathogenic variants within the domain. We established that the PM1 criterion can be upgraded to strong (PM1_Strong) when at least 14 pathogenic variants are reported in the functional domain; the moderate weight (PM1_Moderate) of this criterion was assigned in the presence of at least 7 pathogenic missense variants. Finally, we considered PM1_Supporting when 5-6 pathogenic mutations are reported in the domain (Supplementary Table S2). The distribution of SOD1 variants is shown in Figure 1A.

Evaluation of the PVS1 Criterion
A strong weight in the classification of ALS genes is represented by the PVS1 criterion. This is attributed to a "Null variant in a gene where loss of function (LOF) is a known mechanism of disease". Gain of function and several mechanisms are confirmed to be involved in ALS [33][34][35]. However, much evidence about SOD1 loss of function (LOF) suggests it may play a role in ALS pathogenesis, probably increasing the susceptibility to neurodegeneration. Overall, the major effect of SOD1 mutations in ALS is linked to the protein aggregation but these may also lead to loss of function of SOD1 activity or loss of the nuclear function, where SOD1 acts as a transcription factor [6].
According to the recommendations provided by the ClinGen SVI working group [10] and taking into account Guissart's considerations [36], we adapted ACMG-AMP guidelines for SOD1 by evaluating the right strength related to the PVS1 criterion and distinguishing the variants with PVS1_Very strong and PVS1_Strong. In particular, we assigned PVS1_Very strong criterion when variants do not undergo the nonsense-mediated mRNA decay (NMD) and PVS1_Strong when variants undergo NMD. A representation of the SOD1 gene showing the localization of premature termination codons (PTCs) that cause ALS and NMD escaping is reported in Figure 1B.

Comparison between Databases
To assess the concordance between annotations generated by ClinVar and LOVD databases, we compared a list of 202 variants of the SOD1 gene across several online resources. The databases investigated generate an automated interpretation of the variants. Specifically, Leiden Open Variation Database (LOVD-https://www.lovd.nl/, 5 January 2022) is a software database for collecting and visualizing variants in the DNA sequence [37]. ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/, 7 January 2022) is an archive that aggregates information on genomic variation and the various correlations with human health [38].

Statistical Analysis
The chi-square concordance test was performed for discordance-concordance and variant classification across the various databases, using the SPSS 20.0 test (SPSS, Chicago, IL, USA). Pathogenic/likely pathogenic SOD1 variants were compared using Fisher's test. A p-value < 0.05 was considered statistically significant.

Contribution of Each Criterion to the Final Classification
The contribution of each criterion to the final classification is shown in Figure  PM2 criterion was assigned to all 202 variants analyzed in this study. Pathogenic va of the SOD1 gene are often identified in a single patient or an isolated family. Th dence underlines the rarity of the variants and therefore allows the PM2 criterion quire a moderate level of evidence of pathogenicity.

Contribution of Each Criterion to the Final Classification
The contribution of each criterion to the final classification is shown in Figure 3. The PM2 criterion was assigned to all 202 variants analyzed in this study. Pathogenic variants of the SOD1 gene are often identified in a single patient or an isolated family. This evidence underlines the rarity of the variants and therefore allows the PM2 criterion to acquire a moderate level of evidence of pathogenicity.
The frameshift variants c.272_274dupACA and c.56_58delTCA cause a change in the length of the protein for which the PM4 criterion can be applied. Combined with the PM and PP criteria, these two variants were classified as likely pathogenic.

Concordance-Discordance between Open Databases and Reclassification Based on ACMG-AMP Guidelines
LOVD contains a total of 47 variants, of which only 18 are classified as pathogenic, likely pathogenic, or VUS. The comparison of these variants (18/202, 9%) revealed a discordant classification (p-value < 0.001) in two substitutions. The variant c.59A>G, defined as likely benign in LOVD databases, was reclassified as VUS, while the c.272A>C was reclassified from VUS to pathogenic (Supplementary Table S5).
In the present study, the PM5 criterion assigned to a "novel missense change at an aminoacidic residue where a different missense change determined to be pathogenic has been seen before" [9], showed a moderate impact in 114 variants (56.4%). Of these, 91 (79.8%) were likely pathogenic, while the remaining 23 (20%) were pathogenic.
According to the ACMG-AMP guidelines, the PP5 criterion was assigned to 90 variants (90/202; 44.6%) for which different studies were available in the literature. The presence of the PP5 criterion allowed the classification of 65 (72.2%) likely pathogenic, 23 (25.5%) pathogenic, and of the remaining 2 (2.2%) VUS variants. The splice region c.240-7T>G and the intron c.358-11A>G variants were characterized as VUS, with the PM2 and PP5 criteria, since no data from the literature were available.
Considering previously published functional studies for the evaluation of the variant effect and the impact of the mutation on the clinical phenotype, the PS3 criterion was assigned to 32 (32/202; 15.8%) variants (Supplementary Table S3). PS3_Strong was assigned to 10 (31.3%) variants classified as pathogenic. These pathogenic variants were characterized by the attendance of multiple functional studies performed in various experimental models (Supplementary Table S3) [39][40][41][42][43][44][45][46][47][48][49]. The p.Gly94Ala mutation, known as G93A, was used to create the first ALS-associated SOD1 transgenic mouse. This model, which expresses large amounts of mutant SOD1, develops adult-onset neurodegeneration of spinal motor neurons and progressive motor deficits leading to paralysis, and it is now widely used both for pathogenic molecular mechanisms and for preclinical studies useful for the specific therapy of affected patients [31,[39][40][41]. This mutation is currently the one studied in the largest number of animal models including C. Elegans, Saccharomyces Cerevisiae, Swine, and Danio rerio (Zebrafish) models. In the same way, experimental models of the transgenic mouse G93A and H46R (p.His47arg) show degeneration of the upper and lower motoneurons associated with a gravity directly proportional to the altered protein expression [31,39,42]. In addition, in vivo experiments show that SOD1-D91A (p.Asp91Ala), the most common mutation affecting SOD1 linked to ALS, is less toxic than other tested mutants, while homozygous mice develop fatal motor neuron disease similar to that observed in homozygous-D91A human ALS patients [41,43].
A Zebrafish model with G94R (p.Gly94Arg) variant has been developed, and this shows a mutated expression of SOD1 at a physiological level, which allows targeted pre-clinical studies [31,49]. For the variants G86R (p.Gly86Arg) [31,[45][46][47] and G38R (p.Gly38Arg) [31,47,48], there are several functional studies carried out on transgenic mouse models in which there are high expression levels of the mutated gene and differences in the age of onset and progression of the disease as well as lifespan. For the G86R variant, studies have also been carried out using the C. Elegans nematode model. In particular, the presence of this missense variant produces an abnormal function of neurons ad a cellular dysfunction at the muscular level. The same variant was tested on Drosophila melanogaster models showing a progressive deterioration of motor functions with a specific clinical phenotype. The same experimental model was also used to study the effect of variants G38R and I114T (p.Ile114thr), determining a decrease in the lifespan of the midge accompanied by abnormalities in movement (Supplementary Table S3).
Another criterion with a strong impact on the classification, the PS1, was assigned to only eight SOD1 variants (4%). Of these, seven (87.5%) were classified as pathogenic and one (12.5%), the c.355G>T, was classified as likely pathogenic.
The frameshift variants c.272_274dupACA and c.56_58delTCA cause a change in the length of the protein for which the PM4 criterion can be applied. Combined with the PM and PP criteria, these two variants were classified as likely pathogenic.

Concordance-Discordance between Open Databases and Reclassification Based on ACMG-AMP Guidelines
LOVD contains a total of 47 variants, of which only 18 are classified as pathogenic, likely pathogenic, or VUS. The comparison of these variants (18/202, 9%) revealed a discordant classification (p-value < 0.001) in two substitutions. The variant c.59A>G, defined as likely benign in LOVD databases, was reclassified as VUS, while the c.272A>C was reclassified from VUS to pathogenic (Supplementary Table S5).

Discussion
In this study, for the first time, ALS-associated SOD1 variants were systematically re-evaluated to perform clinical interpretation according to the ACMG-AMP guidelines. We also provided an updated assessment of variant concordance-discordance between bioinformatics tools used to evaluate variants' roles and the criteria application provided by the above-mentioned guidelines adapted for SOD1.
Previous research has reported that only 34% of variants have a concordant classification according to the ACMG-AMP guidelines [64]. The high percentage of discordance between classifications can be explained by how individual laboratories interpret and analyze variants data, based on published levels of evidence and the misuse of ACMG-AMP guidelines.
Various scientific evidence supports the idea that ALS is caused by a combination of genetic and environmental factors [65,66]. To understand and quantify the value of the contribution of genetics in the pathogenesis of the disease several studies have been conducted [46]. Although a large number of genes have been related to ALS, other variants are to be discovered, which could play adjunct roles in increasing risk or susceptibility to the disease. Specific mutations not identifiable by GWAS or massive sequencing methods, such as repeated expansions [67] or variants in noncoding regions, could belong to this category [68,69].
Human SOD1 is widely known to be an important protein in the antioxidant system. The main hallmark of fALS is the presence of SOD1 aggregates in the spinal cord of patients [70], and the chemically modified form of the protein in the spinal cord of sporadic ALS patients has also been found [71]. SOD1 mutations were the first to be directly associated with the disease pathogenesis and the classification of variants in such a complex pathology remains a challenge. The heterogeneity of clinical data and the difference in approaches used for the interpretation of variants are the main challenges in variant classification [72].
SOD1 gene has a high rate of rare variants, and an appropriate classification is essential for a correct ALS diagnosis. Furthermore, one of the greatest challenges in interpreting pre-existing assessments is the heterogeneity in the approaches used for variant analysis. In this study, we applied a standardized in silico analysis and assessed the frequency of the variants, as well as considering the presence of functional and segregation studies. Results showed that 136 (67.3%) variants met the criteria of likely pathogenic, 53 (26.3%) were classified as pathogenic (Figure 4), and the remaining variants (n = 13; 6.4%) were categorized as VUS.  Table S5). Finally, 111 terms (55%), searched for rs ID, did not match in ClinVar.

Discussion
In this study, for the first time, ALS-associated SOD1 variants were systematically reevaluated to perform clinical interpretation according to the ACMG-AMP guidelines. We also provided an updated assessment of variant concordance-discordance between bioinformatics tools used to evaluate variants' roles and the criteria application provided by the above-mentioned guidelines adapted for SOD1.
Previous research has reported that only 34% of variants have a concordant classification according to the ACMG-AMP guidelines [64]. The high percentage of discordance between classifications can be explained by how individual laboratories interpret and analyze variants data, based on published levels of evidence and the misuse of ACMG-AMP guidelines.
Various scientific evidence supports the idea that ALS is caused by a combination of genetic and environmental factors [65,66]. To understand and quantify the value of the contribution of genetics in the pathogenesis of the disease several studies have been conducted [46]. Although a large number of genes have been related to ALS, other variants are to be discovered, which could play adjunct roles in increasing risk or susceptibility to the disease. Specific mutations not identifiable by GWAS or massive sequencing methods, such as repeated expansions [67] or variants in noncoding regions, could belong to this category [68,69].
Human SOD1 is widely known to be an important protein in the antioxidant system. The main hallmark of fALS is the presence of SOD1 aggregates in the spinal cord of patients [70], and the chemically modified form of the protein in the spinal cord of sporadic ALS patients has also been found [71]. SOD1 mutations were the first to be directly associated with the disease pathogenesis and the classification of variants in such a complex pathology remains a challenge. The heterogeneity of clinical data and the difference in approaches used for the interpretation of variants are the main challenges in variant classification [72].
SOD1 gene has a high rate of rare variants, and an appropriate classification is essential for a correct ALS diagnosis. Furthermore, one of the greatest challenges in interpreting pre-existing assessments is the heterogeneity in the approaches used for variant analysis. In this study, we applied a standardized in silico analysis and assessed the frequency of the variants, as well as considering the presence of functional and segregation studies. Results showed that 136 (67.3%) variants met the criteria of likely pathogenic, 53 (26.3%) were classified as pathogenic (Figure 4), and the remaining variants (n = 13; 6.4%) were categorized as VUS.  In our study, the PS3 criterion proved to be fundamental for the pathogenicity classification of 32 variants, representing 62.5% of total pathogenic terms. However, the near absence of functional and segregation data for a large number of terms remains an important limiting factor, considering that functional evidence is the main scoring driver for pathogenic/likely pathogenic for missense variants. Compared with all the variants analyzed, it is clear that there are few functional studies, conducted in vitro or in vivo, which show a detrimental effect on the protein. Furthermore, the PS3 criterion is not automatically assigned in the various databases, as functional evidence must be searched by the geneticist consulting PubMed and other available sites (Mastermind, etc.). After the manual revision, the strength of the criterion can also be changed according to the relevance of the evidence.
The attribution of the PM1 criterion was based on scientific evidence showing the absence of specific mutational hotspot mutations in SOD1 linked to ALS, indicating the whole protein involved [73]. In SOD1, several variants have been described to cluster in all five exons of the gene identified as hotspots ( Figure 1A). Rather than defining clear function domains, these regions are defined by an increase in pathogenic variations identified in ALS patients. For this reason, we considered three different levels of strength of the criterion: PM1_Strong (≥14 variants), PM1_Moderate (at least 7 variants), and PM1_Supporting (5-6 variants).
All of the variants present in this study with a MAF value less than 0.001 and absents from all databases were considered to have a moderate pathogenicity level. The definition of a MAF threshold for ALS, in which significant confounding factors coexist (late-onset of the disease, age-dependent, variant-related reduced penetrance, and oligogenic inheritance), cannot ignore many factors difficult to evaluate, such as the inheritance model (monoallelic for most SOD1-ALS variants), the prevalence of the disease, and the penetrance. Failing consensus on allele frequency thresholds, the PM2 criterion was assigned to these cases.
For most of the missense variants involved in our study and classified as pathogenic or likely pathogenic, the PP5 criterion was considered supporting evidence with a very strong value, as the variants are present in several studies in the literature. However, this criterion must be used with caution as "the evidence is not available to the laboratory to perform an independent evaluation" [11].
A strong impact on the evaluation of a mutation is expressed by the PS1 criterion, which is attributed when the same amino acid variation of a well-known causative variant occurs. The variant could act directly through the specific change of DNA or could modify the protein sequence through various mechanisms, such as an alteration in splicing. To support the pathogenicity of genetic variants, it is necessary to have well-established and reproducible functional studies.
The PVS1 criterion is considered the greatest impact, characterized by identifying a variant that causes a loss-of-function (LoF) allele. Although the gain of function is causative for amyotrophic lateral sclerosis, SOD1 loss of function may play a role in ALS, contributing to the onset and progression of the disease [73,74]. Many nonsense or frameshift mutations may lead to a LOF of SOD1 by affecting its structures and/or interactions pattern [6]. mRNA containing premature termination codons are known to be rapidly degraded by nonsense-mediated mRNA decay and premature termination codons that escape NMD in SOD1, causing the production of truncated SOD1 protein, which may have high toxicity [36] ( Figure 1B).
An element of increasing importance is constituted by databases that report identified variants with their associated phenotypes and the interpretation of their functional role. In ALS, this concept is even more important since the vast majority of the variants identified are new and/or private, and for this reason, it is very important to share the knowledge obtained in dedicated databases. Another aspect to note is the presence/absence of SOD1 variants in these different databases. This study shows that many variants of this gene are absent in online archives used and, therefore, stresses that it is necessary to revise these data, so they are also open source, allowing the scientific community to search for updated information of genetic mutations, which are extremely important to the disease. A comparison of the annotations between open databases and our reclassification based on ACMG-AMP guidelines with the criteria adapted for SOD1 revealed a different rate of discordance. In particular, comparing ClinVar, we found discordance between pathogenic/likely pathogenic variants and VUS in 9.4% of terms. A case in point is the most common SOD1 mutation, p.Asp91Ala, which is still reported in ClinVar with conflicting interpretations of pathogenicity despite extensive evidence demonstrating pathogenicity according to the ACMG-AMP guidelines. In this public repository, many variants are associated with classification according to the ACMG-AMP guidelines. This archive accepts data derived from different working groups, the relevant literature, and documents from other web resources. To ensure the quality of the declarations, a review phase is carried out that involves evaluation by groups of experts. The variant data are then updated based on the new information presented, and therefore, an adequate time is required for the process. [39].
Based on the data obtained, it is evident that the mere application of the guidelines for interpreting genetic variants developed and refined so far cannot be considered sufficient in ALS to address various problems related to the implementation of new genetic analysis technologies in a diagnostic and clinical context. A manual reinterpretation of variants is essential to adapt the study to a particular disease, considering the specific characteristics of the gene and allowing for a consistent application. In this way, the different ACMG-AMP criteria acquire a different weight based on the type of pathology, scientific evidence, and knowledge of the variant itself.

Conclusions
In this study, we re-evaluated the classification of all previously reported SOD1 variants (n = 202) in ALSoD, Project MinE, and in-house databases by applying the ACMG-AMP criteria to ALS. Our results underline the importance of classifying variants involved in ALS and SOD1 variants with caution.
Specific guidelines for ALS would allow a more adequate classification of the large number of variants of uncertain significance, which have been constantly discovered. This can be accomplished by carrying out genetic research on large cohorts of patients and controls and by sharing the data obtained with the scientific community. Furthermore, it is crucial to develop and implement new analysis strategies that could facilitate the discovery and the validation of genetic factors involved in the pathogenesis of this disease that is still unknown. To date, the diagnosis of ALS remains, based on clinical, not genetic criteria. The improvement of the existing variant classification systems available could provide a better basis for using genetic findings as diagnostic criteria.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/genes13030537/s1. Table S1: The ACMG-AMP criteria and their relevance in ALS, Table S2: Assessment of the PM1 criterion in SOD1 variants, Table S3: List of in vivo and in vitro functional studies considered for the PS3 criterion attribution, Table S4: Characteristics of the SOD1 variants relevant to the PVS1 criterion, Table S5: List of SOD1 variants and concordancediscordance between open databases and reclassification, Figure S1: ACMG-AMP guidelines for combining criteria to classify sequencing variants, Figure S2: Decision tree for functional evidence PS3.
Author Contributions: Conceptualization, P.R. and F.L.C.; methodology, P.R.; formal analysis, P.R.; resources, P.R. and B.P.; data curation, P.R.; writing-original draft preparation, P.R.; writing-review and editing, P.R. and F.L.C.; supervision, F.L.C. All authors have read and agreed to the published version of the manuscript. Informed Consent Statement: This study did not involve humans.