Genotyping and Plasma/Cerebrospinal Fluid Profiling of a Cohort of Frontotemporal Dementia–Amyotrophic Lateral Sclerosis Patients

Frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS) are part of the same pathophysiological spectrum and have common genetic and cerebrospinal fluid (CSF) biomarkers. Our aim here was to identify causative gene variants in a cohort of Greek patients with FTD, ALS and FTD-ALS, to measure levels of CSF biomarkers and to investigate genotype-phenotype/CSF biomarker associations. In this cohort of 130 patients (56 FTD, 58 ALS and 16 FTD-ALS), we performed C9orf72 hexanucleotide repeat expansion analysis, whole exome sequencing and measurement of “classical” (Aβ42, total tau and phospho-tau) and novel (TDP-43) CSF biomarkers and plasma progranulin. Through these analyses, we identified 14 patients with C9orf72 repeat expansion and 11 patients with causative variants in other genes (three in TARDBP, three in GRN, three in VCP, one in FUS, one in SOD1). In ALS patients, we found that levels of phospho-tau were lower in C9orf72 repeat expansion and MAPT c.855C>T (p.Asp285Asp) carriers compared to non-carriers. Additionally, carriers of rare C9orf72 and APP variants had lower levels of total tau and Aβ42, respectively. Plasma progranulin levels were decreased in patients carrying GRN pathogenic variants. This study expands the genotypic and phenotypic spectrum of FTD/ALS and offers insights in possible genotypic/CSF biomarker associations.


Introduction
It has been widely recognized that Frontotemporal Dementia (FTD) and Amyotrophic Lateral Sclerosis (ALS) occupy the two extremes of the same pathophysiological spectrum, sharing several histological features and genetic causes [1]. FTD is a highly inherited disorder, with 30-50% of patients reporting family history of a similar phenotype [2]. For

Blood Collection and DNA Extraction
For the genetic analyses, whole peripheral blood from participants was collected in ethylenediaminetetraacetic acid (EDTA) tubes, which were stored at −20 • C until DNA extraction and were only thawed once, just before the procedure. For the extraction of the genomic DNA from 400 µL of whole peripheral blood, the QIAamp DNA Blood Mini kit (Qiagen, CA, USA) was used. DNA concentration and purity were assessed spectrophotometrically at 260 and 280 nm. Participant anonymity was ensured through the designation of unique code identifiers for each DNA sample.

C9orf72 Repeat Expansions
Most of the patients (n = 94) were initially checked for C9orf72 hexanucleotide repeat expansions at the Neurogenetics Laboratory of Eginition Hospital according to a two-step protocol. This protocol included sizing PCR (amplification of the region that contains the hexanucleotide repeat with primers flanking this region), agarose electrophoresis and fragment analysis to identify samples with possible presence of the C9orf72 repeat expansion. This was followed by repeat primed PCR, to verify and separate samples with the C9orf72 repeat expansion, as previously described [21,22]. Samples were then electrophorized on an ABI 310 Capillary Analyzer (Applied Biosystems, Foster City, CA, USA) and analyzed on GeneScan v3.7 (2001, Applied Biosystems, Foster City, CA, USA). In addition, 21 patients were initially analyzed for the presence of C9orf72 repeat expansion by repeat primed PCR amplification and STR (short tandem repeats) PCR analysis at the Diagnostic Service Facility, Laboratory of Neurogenetics, University of Antwerp, Belgium.

Whole Exome Sequencing
In most patients that did not carry a C9orf72 hexanucleotide repeat expansion, Whole Exome Sequencing (WES) was performed ( Figure 1) according to the following procedures: (1) At Minotech Genomics Facility, Institute of Molecular Biology and Biotechnology (IMBB-FORTH, Crete) with the use of the Illumina NextSeq500 platform (n = 60). In detail, sequencing of 2 × 75 bp DNA fragments with at least 50× coverage, targeting regions of 45.3 Mb size, was performed. Libraries were prepared with the TruSeq ® Rapid Exome Library prep kit (Illumina, San Diego, CA, USA). Bioinformatics processing of the data derived from mapping on the hg19 reference genome and quality control of the results (e.g., number of readings and coverage quality) were performed by the BaseSpace ® software (Illumina, San Diego, CA, USA). Finally, genetic variation was identified with the VariantStudio ® software after comparison with the hg19 reference genome and drawing information from genetic databases, e.g., Human Genome Mutation Database (HGMD), ClinVar ® (National Center for Biotechnology Information, Bethesda MD, USA) (Annotation Excel file). (2) At Macrogen (Seoul, Korea), using the Illumina HiSeq4000 platform (n = 24). In specific, 2 × 100 bp DNA fragments were sequenced with an aim of at least 50x coverage. For the construction of genomic libraries, the Agilent Sure-Select Human All Exon V5 (not including UTRs) Target Enrichment System was used. (3) At Otogenetics (GA, USA), using the Illumina HiSeq2500 platform (n = 11). In detail, sequencing of 2 × 100 bp DNA fragments was performed aiming at coverage of at least 50× and targeting a region of 45.

Gene variant Identification and Verification
We specifically searched for variants in the most recognized FTD and ALS related genes, namely C9orf72, GRN, MAPT, TARDBP, FUS, CHMP2B, SQSTM1, UBQLN2, VCP,  OPTN, TBK1, SOD1 and CHCHD10, as well as the 3 genes associated with autosomal dominant AD (APP, PSEN1, PSEN2). Single nucleotide polymorphism (SNP) variants reported here were verified by Sanger sequencing to exclude false positive results.

Measurements of CSF and Plasma Biomarkers
The CSF levels of Aβ42, τT and τP-181 were measured in duplicate by commercially available ELISA kits (Innotest β-amyloid 1-42, hTau antigen and phospho-tau 181; Fujirebio, Gent, Belgium) according to the manufacturer's instructions. We chose to measure τP-181, as in neurodegenerative disorder research and clinical practice, tau phosphorylated at threonine 181 (τP-181) is the most commonly measured form of phosphorylated tau as a biomarker in the CSF [24] and, recently, in plasma [25]. Additionally, to better evaluate the τP-181 levels, the ratio of τP-181 to τT (τP-181/τT) was calculated [26]. TDP-43 was measured in triplicate by double-sandwich enzyme-linked immunosorbent assay (ELISA) using a commercial kit (Human TAR DNA-binding protein 43 ELISA kit; Cusabio Biotech, China). All determinations were performed using a four-parameter logistic curve and blindly to

Gene Variant Identification and Verification
We specifically searched for variants in the most recognized FTD and ALS related genes, namely C9orf72, GRN, MAPT, TARDBP, FUS, CHMP2B, SQSTM1, UBQLN2, VCP, OPTN, TBK1, SOD1 and CHCHD10, as well as the 3 genes associated with autosomal dominant AD (APP, PSEN1, PSEN2). Single nucleotide polymorphism (SNP) variants reported here were verified by Sanger sequencing to exclude false positive results.

Measurements of CSF and Plasma Biomarkers
The CSF levels of Aβ 42 , τ T and τ P-181 were measured in duplicate by commercially available ELISA kits (Innotest β-amyloid 1-42, hTau antigen and phospho-tau 181; Fujirebio, Gent, Belgium) according to the manufacturer's instructions. We chose to measure τ P-181 , as in neurodegenerative disorder research and clinical practice, tau phosphorylated at threonine 181 (τ P-181 ) is the most commonly measured form of phosphorylated tau as a biomarker in the CSF [24] and, recently, in plasma [25]. Additionally, to better evaluate the τ P-181 levels, the ratio of τ P-181 to τ T (τ P-181 /τ T ) was calculated [26]. TDP-43 was measured in triplicate by double-sandwich enzyme-linked immunosorbent assay (ELISA) using a commercial kit (Human TAR DNA-binding protein 43 ELISA kit; Cusabio Biotech, Wuhan, China). All determinations were performed using a four-parameter logistic curve and blindly to the clinical diagnosis. Cut-off levels were calculated by ROC (receiver operating characteristic) analysis with the optimal combination of sensitivity and specificity, as previously described [15,16].
Blood samples were collected between 08:00-10:00 a.m. (morning samples), transferred to EDTA-containing tubes and refrigerated until centrifugation (within 3 h) for the plasma isolation. Plasma isolated from these samples was subsequently kept in deep freeze (−80 • C). Plasma progranulin levels were measured using the Human Progranulin ELISA Kit (Adipogen Life Sciences, Liestal, Switzerland) [27].

Statistical Analysis
All numerical data were tested for normality and homogeneity of variances by the Shapiro-Wilk's and Brown-Forsyth tests, respectively. When appropriate, differences among groups were tested by one-way analysis of variance (one-way ANOVA) or two-way analysis of covariance (two-way ANCOVA), followed by Bonferroni correction for multiple comparisons. When deviations from normality and/or heterogeneity of variances were noted, Kruskal-Wallis test was performed, followed by Dunn's post-hoc test. Categorical data were compared among groups by the χ 2 -test.

Demographic Data
Demographic data are summarized in Table 1. The three patient groups (FTD, ALS, FTD-ALS) had comparable age and sex. Due to violations regarding normality and homogeneity of variances, disease duration was initially compared among groups with Kruskal-Wallis test that showed a significant difference (p < 0.0001). Additionally, Dunn's post-hoc test revealed significantly longer disease duration for FTD as compared to ALS patients (p < 0.0001).

Family History
Among the 130 patients within the FTD-ALS spectrum, 28.6% (16/56) of the FTD patients had at least one first degree relative affected with phenotype suggesting either FTD or ALS. For the ALS and FTD-ALS groups, 15.5% (9/58) and 25.0% (4/16), respectively, had positive family history for possible FTD or ALS, with at least one first degree relative affected (Table 1).  Table 2.

Other Causative Variants
From the analysis of the FTD/ALS genes in the WES derived data, causative variants were identified in the TARDBP, GRN, VCP, SOD1 and FUS genes in 11 patients (Table 3). Specifically, in three patients (two with pure ALS and one with FTD-ALS) we found causative variants in the TARDBP gene, namely the p.Met337Val (c.1009A>G), the p.Asn352Ser (c.1055A>G) and the p.Ile383Val (c.1147A>G) variant.  In three patients with FTD, all with the PPA phenotype, we found three causative variants in the GRN gene (Table 3). The two patients with splice site GRN variants (c.463-2A>G and c.934-1G>A) presented with an svPPA and nfaPPA phenotype, respectively. Both had increased CSF τ T and TDP-43 levels and τ P-181 and Aβ 42 levels within the reference range for healthy controls. Another FTD-PPA patient was detected with the p.Cys482Tyr GRN variant. She had increased CSF levels of τ T and τ P-181 , while the CSF levels of TDP-43 and Aβ 42 were within the reference range for healthy controls.
Additionally, in three patients with FTD and Inclusion Body Myopathy (IBM), all from Crete, we found two causative variants in the VCP gene (Table 3). A 63-year-old male patient with the pathogenic p.Arg159His (c.476G>A) heterozygous missense VCP gene variant was identified with the characteristic clinical picture of IBM and FTD [23]. The same VCP variant (p.Arg159His) was found in a 68-year-old patient with progressive muscle weakness and atrophies in all four extremities and the trunk and behavioral disturbances. The symptoms of the patient had started at the age of 58 years. Another patient was found to have myopathy, persistently increased Creatine PhosphoKinase (CPK), language deficits and Paget's disease of bone. He carried the p.Arg155His (c.464G>A) VCP change (Table 3).
Further, we found ALS causative variants in the FUS and the SOD1 genes (Table 3). Specifically, we detected the p.Gly506Val (c.1517G>T) FUS gene variant in a 82-yearold male with a predominately lower motor neuron involvement and no family history. Moreover, we found the p.Ser106Leu (c.317C>T) SOD1 gene variant in a 37-year old patient with personal and family history of ALS.

Association of Variants in FTD -ALS Genes with CSF Biomarkers
As a next step in our analyses, we proceeded in measuring CSF biomarker levels in this patient cohort and associated these levels with the results of C9orf72 repeat expansion determination and WES based genotyping. Our efforts focused on typical neurodegenerative disease-associated CSF biomarkers (TDP-43, Aβ 42 , τ P-181 and τ T ) and on genes that either have as their protein product the above biomarkers (TARDBP, APP, MAPT) or are commonly associated with FTD/ALS (C9orf72). Of these four genes, no genotype/CSF biomarker associations were possible for the TARDBP gene since very few patients with variants in this gene were found.

Association of C9orf72 Variants with CSF Biomarkers
Regarding the C9orf72 gene, in the cohort analyzed, besides the pathogenic hexanucleotide repeat expansion (see above), we found five rare intronic variants (c.444+31T>G, c.600+86A>T, c.666-120C>T, c.855+52A>G, c.856-37G>T) and the p.Asn207Ser (p.N207S) exonic variant. These C9orf72 variants differ in their location and nature (repeat expansion, intronic and exonic) and thus possibly have differential effects on gene expression and function. For this reason, these three types of genetic variants were analyzed separately. Thus, for this part of our study, we compared Aβ 42 , τ T , τ p-181 and TDP-43 CSF levels and the τ P-181 /τ T ratio in FTD and/or ALS patients carrying the hexanucleotide C9orf72 repeat expansion, rare intronic C9orf72 variants or the p.Asn207Ser (p.N207S) exonic C9orf72 variant to patients without these variants. Analyses were performed separately for FTD patients, ALS patients ± FTD and all patients combined.
These analyses revealed that in patients with ALS (with or without concurrent FTD), the CSF levels of τ P-181 were lower in C9orf72 repeat expansion carriers compared to noncarriers, whereas levels of TDP-43, Aβ 42 and τ T and the τ P-181 /τ T ratio did not differ between the two groups ( Figure 2). Additionally, carriers of rare intronic variants in C9orf72 had lower CSF levels of τ T (Figure 3). In contrast, this difference in τ T levels was not found for the more common C9orf72 p.Asn207Ser (p.N207S) variant (Figure 3). Results on C9orf72-based biomarker level comparisons that did not reveal statistically significant results or results not approaching statistical significance are not shown.

Association of MAPT and APP Gene Variants with CSF Biomarkers
When we compared levels of CSF biomarkers between groups of patients carrying different MAPT and APP gene variants, we found that the c.855C>T (p.Asp285Asp/p.D285D) synonymous MAPT variant carriers had significantly lower τp-

Association of MAPT and APP Gene Variants with CSF Biomarkers
When we compared levels of CSF biomarkers between groups of patients carrying different MAPT and APP gene variants, we found that the c.855C>T (p.Asp285Asp/p.D285D) synonymous MAPT variant carriers had significantly lower τ p-181 (Figure 4). There was no difference found for the other three biomarkers tested (TDP-43, Aβ 42 and τ T ), when comparing carriers of the p.Asp285Asp (p.D285D) change with non-carriers (data not shown). Additionally, all other MAPT variants similarly tested did not show an effect on biomarker levels (data not shown). Additionally, as shown in Figure 4, carriers of rare APP variants (p.Ala11Ala, p.Val118Ile, c.-44C>T/c.-49+321C>T, c.697+50G>A and c.697+11888delT) had lower Aβ 42 levels, whereas this was not found for the more common c.1000-31T>C intronic APP variant. Finally, other biomarkers (τ T , τ p-181 , TDP-43) analyzed relative to APP variants did not yield significant results (data not shown). groups ( Figure 5). For these analyses, Spearman ρ values were 0.74 (p < 0.0001) for FTD patients, 0.78 (p < 0.0001) for ALS patients and 0.62 (p = 0.031) for FTD/ALS patients. Correlation between other CSF biomarker measurements pairs (τP-181 vs. Aβ42, Αβ42 vs. TDP-43 and TDP-43 vs. τP-181) did not yield statistically significant results (data not shown).

Association of Pathogenic Variants in the GRN Gene with Plasma Progranulin Levels
To show that our findings for CSF biomarkers (i.e., that genetic variants affect CSF biomarker levels) are valid also for plasma biomarkers, we performed plasma progranulin quantification in the three patients with the GRN variants, c.463-2A>G, c.934-1G>A and p.Cys482Tyr (c.1445G>A), that were considered causative based on strong in silico predictions. Patients carrying the pathogenic GRN variants had more than two-fold lower plasma progranulin levels compared to FTD patients harboring the C9orf72 repeat expansion and to other FTD patients ( Figure 6). These findings verified the pathogenicity of the three novel GRN gene variants.

Correlation of CSF Biomarkers Values
As a next step in our analyses, we performed correlation analyses of CSF biomarker values (Aβ 42 , τ T , τ p-181 , TDP-43) in pairs across different patient groups. Results of these analyses showed that only τ T and τ p-181 CSF values were correlated, across all three patient groups ( Figure 5). For these analyses, Spearman ρ values were 0.74 (p < 0.0001) for FTD patients, 0.78 (p < 0.0001) for ALS patients and 0.62 (p = 0.031) for FTD/ALS patients. Correlation between other CSF biomarker measurements pairs (τ P-181 vs. Aβ 42 , Aβ 42 vs. TDP-43 and TDP-43 vs. τ P-181 ) did not yield statistically significant results (data not shown).

Association of Pathogenic Variants in the GRN Gene with Plasma Progranulin Levels
To show that our findings for CSF biomarkers (i.e., that genetic variants affect CSF biomarker levels) are valid also for plasma biomarkers, we performed plasma progranulin quantification in the three patients with the GRN variants, c.463-2A>G, c.934-1G>A and p.Cys482Tyr (c.1445G>A), that were considered causative based on strong in silico predictions. Patients carrying the pathogenic GRN variants had more than two-fold lower plasma progranulin levels compared to FTD patients harboring the C9orf72 repeat expansion and to other FTD patients ( Figure 6). These findings verified the pathogenicity of the three novel GRN gene variants.

Discussion
Our study presents a clinical series of Greek FTD/ALS spectrum patients by innovatively integrating phenotypic, genotypic and biomarker data. Causative genetic variants (either SNPs or C9orf72 hexanucleotide repeat expansions) were identified in 25 cases (19.2% of the total patient cohort). The frequency of patients carrying causative FTD/ALS gene variants in our series is higher compared to other similar series from Greece, most of them having evaluated either FTD or ALS cohorts separately [28][29][30]. This increased frequency probably reflects a higher proportion of FTD/ALS patients in our study, as well as the exclusion of patients with probable AD pathology through CSF biomarker measurements. Additionally, we have observed several unique associations between genotype and CSF/plasma biomarker levels, with our study being one of the first studies to adopt such an approach.
The most common genetic change detected in the cohort analyzed was, as expected, the C9orf72 repeat expansion (12.2%; 14 out of 115 patients tested). C9orf72 repeat expansions have been reported to 4-40% of patients with ALS [31]. In previously reported Greek ALS cohorts, 5-10% of cases carried the abnormal C9orf72 expansion [28,32], in accordance with our findings. Regarding the FTD group, we found a rate of C9orf72 repeat expansions (10.4%), similar to one of the studies from Greece [29], but almost two-fold higher compared to another Greek study [30]. The latter can be explained by the biomarker supported clinical diagnosis of FTD patients in our study, avoiding the contamination of this group with other pathologies presenting as a frontal-behavioral or language syndrome.
Additionally, using a WES based approach, we identified causative gene variants in 11 more patients: three variants in each of the TARDBP and GRN genes, two variants in the VCP gene (in three patients) and one variant in each of the SOD1 and FUS genes (Table 3). Variants in these genes are commonly described in FTD/ALS cohorts, including the most recent studies [33].
In detail, we detected three patients (two women with ALS and one man with FTD-ALS) with TARDBP pathogenic variants (p.Met337Val, p.Asn352Ser and p.Ile383Val). This is of special importance as TARDBP encodes TDP-43, a protein which is increasingly implicated in neurodegenerative processes and offers an attractive therapeutic target [34]. Pathogenic TARDBP gene variants are considered the cause of 1-7% of familial ALS cases and of several sporadic ALS cases [35,36], but they are less frequently reported in FTD [3,37]. Our 60-year old patient with the p.Ile383Val TARDBP gene variant presented with an FTD-ALS phenotype. This specific variant has been repeatedly described as pathogenic across the FTD/ALS spectrum [33,[38][39][40][41]. Additionally, three FTD patients of Greek origin harboring the p.Ile383Val TARDBP gene variant have been recently reported [30]. These data are a call for a systematic study of this variant's geographic distribution and significance. Furthermore, the c.1009A>G (p.Met337Val) TARDBP variant found in one of our patients with familial ALS has been repeatedly reported to segregate with ALS [42][43][44][45][46], including bulbar onset ALS [47]. The p.Asn352Ser TARDBP variant that we found in another ALS patient has also been often described in the literature [48][49][50] and functionally characterized [51,52].
In three of our patients, we found causative variants in the GRN gene (c.463-2A>G, c.934-1G>A and p.Cys482Tyr), all novel (not previously reported) but with strong evidence in favor of their pathogenicity. Variants in the GRN gene have been emerging as a frequent cause of FTD, with marked phenotypic heterogeneity [53][54][55] but most frequently associated with TDP-43 histopathology and PPA phenotype [56][57][58]. Our data add further support to the association of GRN variants with the PPA phenotype. The two splice site variants (c.463-2A>G, c.934-1G>A) described here have strong in silico evidence that they disrupt splicing, as indicated by their MaxEntScan and CADD scores (Table 3). Additionally, in the GRN gene, this type of variant (splice-site variants) is frequently described as causing FTD. In another FTD-PPA patient, we detected the p.Cys482Tyr GRN variant, which is essentially absent from public variant databases such as gnomAD and has strong computational evidence (CADD score = 32.0) in favor of its pathogenicity (Table 3). All three patients in the cohort we analyzed harboring causative GRN variants showed decreased plasma progranulin levels ( Figure 6). This is additional evidence of the pathogenicity of these variants, as it has been shown for other GRN variants [27,59].
Three male patients, all originating from Crete and all presenting with the IBM/FTD phenotype, were found to harbor heterozygous pathogenic VCP variants. These VCP variants were the p.Arg155His (c.464G>A) and, in two apparently unrelated patients, the p.Arg159His (c.476G>A) variant. Pathogenic VCP gene variants are associated with the characteristic clinical picture of IBM, FTD and ALS, often with Paget's disease of bone. This syndrome can be partially or fully developed depending on the respective causative VCP variant and other factors [60,61]. The p.Arg155His VCP variant has been repeatedly shown to co-segregate with the disease phenotype in multiple affected family members [61][62][63]. This variant is in a mutational hot spot and there are in vitro and in vivo functional studies and strong computational evidence (CADD = 24.6) supportive of its damaging effect. This p.Arg155His VCP missense change occurs at an amino acid residue where different missense changes (p.Arg155Pro, p.Arg155Leu, p.Arg155Cys, p.Arg155Ser) are known to be pathogenic [63]. Regarding the p.Arg159His variant, there are several reports describing patients with this variant in the literature [63], including the clinical description of one of our patients [23]. This variant is not present in population databases such as gnomAD, it has strong in silico evidence of pathogenicity (CADD = 23.2) and affects a conserved arginine residue (Table 3). This amino acid change has been shown to disrupt the function and structure of VCP and can lead to protein aggregates [64][65][66][67].
In addition, we detected the p.Gly506Val (c.1517G>T) FUS gene variant in a patient with ALS. The FUS protein is a widely expressed DNA/RNA-binding protein with functional homology to TDP-43 [68]. It is involved in transcriptional and translational regulation and in mRNA splicing and transport [69]. FUS gene pathogenic variants cause 5% of familial and 1% of sporadic ALS, with predominant lower motor involvement [68,70,71]. The p.Gly506Val FUS gene variant is not present in public databases, such as gnomAD, and there is computational evidence in support of its pathogenicity (CADD = 23.7). Additionally, there are reports on a different amino acid substitution (as p.Gly507Asp) at the same amino acid residue causing ALS [68,70]. Finally, in a 37-year old male with family history of ALS, we identified the p.Ser106Leu SOD1 variant. This variant has already been described in young ALS patients with rather slow disease progression, as was the case in our patient [72][73][74].
There have been several studies on the CSF levels of TDP-43, Aβ 42 , τ T , τ P-181 in patients with FTD and ALS [15,75], but very few have focused on the association of these levels with the genotype of these patients. Here we found several interesting genotype/CSF biomarker associations. Specifically, we found that levels of τ p-181 were lower in C9orf72 repeat expansion and MAPT c.855C>T (p.Asp285Asp) carriers compared to non-carriers. Additionally, carriers of rare C9orf72 and APP variants had lower levels of τ T and Aβ 42 , respectively.
Regarding C9orf72, it has been reported that 25% of patients with C9orf72 repeat expansion show low Aβ 42 levels in the CSF [76], but the effect of this expansion on tau and phospho-tau levels is not clear [77]. Our finding that C9orf72 changes (either the pathogenic repeat expansion or the rare intronic variants) can affect levels of τ P-181 and τ T , indicates that there is an association between C9orf72 and tau species that needs further assessment.
The effect of the MAPT c.855C>T (p.Asp285Asp) change on levels of τ p-181 in the FTD/ALS cohort we studied suggests that this variant could be an indirect modifier of tau phosphorylation (e.g., by affecting mRNA processing) and thus of pathophysiological significance. This is of interest, since the same effect, increased levels of τ p-181 , has been found for several MAPT variants in patients with AD [78]. Our findings extend the observations about this effect on another neurodegenerative disorder beyond AD.
Finally, the effect of rare APP variants on Aβ 42 levels suggests these variants are possibly important for APP processing by secretases, as it has already been shown for known pathogenic and non-pathogenic APP variants in AD and other neurodegenerative disorders [79][80][81].

Strengths and Limitations
The main strength of our work relates to the extent and depth of the genotypic and phenotypic analysis, that included both WES-based genotyping and CSF/plasma biomarker measurements.
One limitation to our work is the rather small cohort size; however, this is counterbalanced by the depth of the genotypic and phenotypic characterization. In addition to the plasma and CSF biomarkers used in our study, there is increasing interest in the possible role of other emerging biomarkers in ALS and FTD, including neurofilaments both in CSF [82,83] and plasma [84,85]. This is of special importance for patients with the FTD/ALS phenotype, given that there is evidence that elevated Neurofilament Heavy chain (NfH) levels could antedate the onset of ALS in FTD patients [83]. Our study was not designed to test the levels of these biomarkers in the plasma and CSF; however, future analyses are panned to include these important biomarkers.

Conclusions
In conclusion, we found causative genetic variants (in the C9orf72, TARDBP, GRN, VCP, FUS and SOD1 genes) with a high frequency (19.2%) in a cohort of Greek FTD/ALS spectrum patients. In addition, we observed several potentially important associations between the plasma/CSF biomarker levels and the genotype of several genes. Specifically, we found lower levels of τ p-181 in C9orf72 repeat expansion and MAPT c.855C>T (p.Asp285Asp) carriers compared to non-carriers. Additionally, carriers of rare C9orf72 and APP variants had lower levels of τ T and Aβ 42 , respectively. Finally, plasma progranulin levels were decreased in patients with the newly described pathogenic GRN variants.
Our results further expand the genotypic/phenotypic spectrum of FTD/ALS and provide insights in the interaction of the genome with plasma and CSF biomarkers. Taken together, our findings call for an integrated individualized approach in interpreting plasma and CSF biomarker levels, as these can be influenced by the genotype of the patient under study.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy restrictions.