Exploring dementia and neuronal ceroid lipofuscinosis genes in 100 FTD-like patients from 6 towns and rural villages on the Adriatic Sea cost of Apulia

Frontotemporal dementia (FTD) refers to a complex spectrum of clinically and genetically heterogeneous disorders. Although fully penetrant mutations in several genes have been identified and can explain the pathogenic mechanisms underlying a great portion of the Mendelian forms of the disease, still a significant number of families and sporadic cases remains genetically unsolved. We performed whole exome sequencing in 100 patients with a late-onset and heterogeneous FTD-like clinical phenotype from Apulia and screened mendelian dementia and neuronal ceroid lipofuscinosis genes. We identified a nonsense mutation in SORL1 VPS domain (p.R744X), in 2 siblings displaying AD with severe language problems and primary progressive aphasia and a near splice-site mutation in CLCN6 (p.S116P) segregating with an heterogeneous phenotype, ranging from behavioural FTD to FTD with memory onset and to the logopenic variant of primary progressive aphasia in one family. Moreover 2 sporadic cases with behavioural FTD carried heterozygous mutations in the CSF1R Tyrosin kinase flanking regions (p.E573K and p.R549H). By contrast, only a minority of patients carried pathogenic C9orf72 repeat expansions (1%) and likely moderately pathogenic variants in GRN (p.C105Y, p.C389fs and p.C139R) (3%). In concert with recent studies, our findings support a common pathogenic mechanisms between FTD and neuronal ceroid lipofuscinosis and suggests that neuronal ceroid lipofuscinosis genes should be investigated also in dementia patients with predominant frontal symptoms and language impairments.

recruited in 6 small and isolated towns and rural villages on the Adriatic Sea cost of Apulia, an Italian southern region characterized by a distinctive historic and geographic isolation over the centuries: Bari (320.000 inhabitants, 116 km 2 ), Andria (99.671 inhabitants, 407 km 2 ), Lecce (95.269 inhabitants, 238 km 2 ), Putignano (26.000 inhabitants, 99,11 km 2 ), Polignano a mare (17.925 inhabitants, 62 km 2 ), Tricase (17.421 inhabitants, 43,64 km 2 ) (Fig. 1). Given their isolation, these areas display a high level of consanguinity and parental isonomy 14 and the local population is highly inbred and enriched for rare causative alleles or highly penetrant risk factors of strong effect size. The patients were selected from the SLAP-DEM registry for rare neurodegenerative disorders in Puglia, south Italy. 75% displayed behavioural FTD (bvFTD), 21% primary progressive aphasia (PPA), and 1% patient showed FTD-ALS. Three patients were diagnosed with AD, and 2 of these were then diagnosed with FTD with memory onset and PPA during the disease progression. Average age at onset was 63 years (43-85y), 55% of the patients presented early-onset (< 65 years) and 24% very early-onset (≤ 55 years) ( Table 1). All patients were evaluated with a complete neuropsychological assessment, structural and functional neuroimaging, blood chemistry tests and electromyography if the neurological examination showed motor-neuronal signs.
This study and all experimental protocols were approved by the ethics committee on human research of the University Hospital and Polyclinic of Bari and the ethics committee of the Hospital of Lecce. Written informed consent was obtained from each subject enrolled in the study. DNA extraction. DNA was extracted from blood using the automated DNA extractor AutoGenFlex STAR (AutoGen, Holliston, MA, USA) according to the manufacturer's protocol.
C9orf72 repeat expansion study. A repeat-primed PCR was performed to screen for the presence of the GGG GCC hexanucleotide repeat expansion in C9orf72 as previously described 15 . Positive and negative controls  Variants were filtered for (1) heterozygous non synonymous, stop gain/loss, frameshift insertions/deletions and splice mutations that were (2) absent or very rare (minor allele frequency ≤ 0.001) in the public databases NHLBI ESP6500 (http:// evs. gs. washi ngton. edu/ EVS/ ) and ExAC03 (http:// exac. broad insti tute. org/) and (3) predicted pathogenic by at least one of the following in silico software algorithms: MetaSVM, MetaLR 23 and CADD Phred score ≥ 20 (University of Washington and HudsonAlpha Institute for Biotechnology, Huntsville, AL).

Sanger sequencing.
To verify that the variants reported in this study were not an artifact of the exome sequencing process, Sanger sequencing was performed using an ABI BigDye Terminator Cycle Sequencing Kit on an ABI 3730xl Sequencer. Sequence traces were analyzed using Sequencher (version 4.2; Gene Codes Corporation, Ann Arbor, MI, USA).
The pipeline of our study has been described in Fig. S1. All methods were carried out in accordance with relevant guidelines and regulations.

Results
We identified 17 rare coding variants in the selected genes. Most of them, 12/17 (70%), were singletons, 5 were novel variants. In our cohort TYROBP, UBQLN2, APP, PSEN1, NOTCH3, SNCA, GBA, CLN2, CLN3, CLN5 did not present any rare coding variant (  24 . One patient (RSA_bvFTD_50) referred as apparently sporadic, carried two different variants in GRN. One missense mutation (c.T415C, p.C139R) in exon 5 leading to a predicted partial loss of functional protein and suggested as pathogenic by in silico and in vitro studies 25 . This mutation has been associated with behavioral frontotemporal dementia, semantic dementia, Alzheimer's disease and corticobasal syndrome 26 . The second variant is a novel (e.g. not present in public databases) nucleotide deletion (c.1165delT, p.C389fs) in exon 10 predicted to give rise to a frameshift leading to the partial loss of function (Table 2). This patient presented with behavioral symptoms at age 63 (apathy, social retire and delusions). Four years later, he was completely socially inappropriate, unable to communicate and dependent in all daily activities with sphincter incontinence. www.nature.com/scientificreports/ Only one individual (1%) carried a pathologic C9orf72 hexanucleotide repeat expansion (37 repeats). The carrier was a male sporadic case and displayed bvFTD with non-fluent aphasia and a very early age at onset (44 years).
We report a rare and likely non-pathogenic variant in PSEN2 p.Q445E, mapping outside the alpha helix surface of the transmembrane domains (TMs), where all the pathogenic mutations have been reported (alphahelix rule) 27 .
Interestingly, we detected also 2 variants in CSF1R TK flanking regions (aa 538-581 and 911-972) (p.E573K and p.R549H). Although mutations in the TK domain (exons 12-22, aa 582-910) have been reported as pathogenic 12 , mutations in the TK flanking regions have been linked to AD 28 and particularly p.E573K is characterized by a significantly decreased autophosphorylation compared to the wild-type CSF1R and has been previously reported in a patient presenting ischemic embolic stroke without the classical HDLS clinical feature but periventricular white matter abnormalities, unrelated to the recent infarct 29 .
Moreover, we report one SORL1 mutation in the valosin-containing protein (VCP) domain (p.R744X) that was associated to AD with severe language impairment and PPA and was not detected in a member of the same family that had been initially diagnosed with AD and successively with FTD with memory onset (Table 3, Fig. 2A, B). SORL1 p.R744X was also found in an unaffected family member from the third generation (HIII2), aged 42 years, who should be considered at risk (average age at onset in Family H is 68 years).
The novel stop-gain mutation in SORL1 (p.R744X) clusters in a very well conserved domain across different species (Fig. 2C) and maps to exon 16, carrying several mutations that have been linked to familial and sporadic AD [30][31][32] . None of the sporadic FTD cases carried SORL1 variants in the VCP domain.
While a second heterozygous LoF mutation was identified within L2HGDH (p.R335X) in family H, this nonsense mutation did not segregate with disease. Finally, we report 4 mutations that cluster within genes highly expressed in the brain and already associated to developemental cognitive impairment and intellectual disability 33 (https:// www. omim. org/) (Table S3, Fig. 2E). Moreover, OXA1L has been linked to mitochondrial encephalopathy and AP5Z1 and SOX5 to hereditary  34 . Although these mutations do not meet all the filter criteria, given the critical role in CNS development, they may be disease modifiers. It is possible that these mutations (CLCN6 p.S116P, SORL1 p.R744X, L2HGDH p.R335X) lead to haploinsufficiency due to a nonsense-mediated decay (NMD) or either the generation of a truncated protein. Due to the lack of RNA samples available, it was not possible to perform a transcript analysis and demonstrate the absence of the mutant allele and therefore discriminate between the two mechanisms.
Family H. The clinical course of patients within Family H is characterized mainly by language impairment (HII2, HII4) and memory problems (HII2 and HII3) (Fig. 2A). The clinical diagnosis of affected family members includes probable AD, PPA and FTD with memory onset. The clinical features of the family members are summarized in Table 3.
The proband of the family died at 65 years of age and no samples were available for genetic evaluation. However, relatives described the patient as suffering from a dementing syndrome with behavioral and personality changes at the age of 60 years old.
HII2. At 62 years of age, the patient presented with memory impairment and spatiotemporal disorientation. Eight years after the onset of symptoms, she developed language problems that progressively worsened over four years with anomie, paraphasic errors and agrammatism progressing to mutism. At the age of 74 years, the patient was bed-ridden and completely dependent for all the daily activities. A neurological examination revealed spastic hypertony in all four limbs, increased and severe deep tendon reflex, mixed aphasia and global cognitive impairment. The clinical diagnosis was consistent with AD with severe language impairment. The patient deceased, aged 80 years old.
HII3. At 73 years of age, the patient developed short-term memory problems, depression and showed apathetic behavior. Three years later, aged 76 years, a neuropsychological examination revealed spatiotemporal disorientation. Long-term memory impairment and attention-execution deficits characterized the disease progression. An MRI scan, showed a marked anterior atrophy (Fig. 2D). The patient has been diagnosed with FTD with memory onset. www.nature.com/scientificreports/ HII4. At 69 years of age, the patient presented with language impairment (anomie and stutter). Over the next four years, language problems progressed to complete mutism with alexia and agraphia. A neurological examination revealed a complete motor aphasia without any remarkable language comprehension impairment. Her behavior was socially appropriate. An MRI scan, performed three years after the onset of symptoms, revealed predominant anterior atrophy (Fig. 2D). A SPECT scan showed left temporo-parietal hypoperfusion. The patient was diagnosed with PPA. Importantly, HII2 presented AD dementia and spastic paraplegia at the 4 limbs. Although this is a typical sign of patients with pathogenic mutations in PSEN1 35 , we have not detected any coding mutation in PSEN1 in this family. However, we report a rare heterozygous missense mutation in AP5Z1, a gene that have been associated to autosomal recessive spastic paraplegia type 48 (SPG48) 36 . Nevertheless, the MRI did not present any typical feature of hereditary spastic tetraparesis: no periventricular white matter hyperintensities or thin corpus callosum (Fig. 2D). However, we cannot exclude that this mutation may modify the disease phenotype.
Neuronal ceroid lipofuscinosis genes. We report a novel and likely pathogenic variant identified in CLCN6 (p.S116P) leading to a T to C transition in the last nucleotide of exon 5 (c.346 in coding DNA reference sequence NM_001286.2), at position − 1 of the exon 5 splice donor site (Fig. 3B). The same mutation may alternatively result in exon 5 skipping or act as a missense mutation (c.T346C, predicting a p.S116P substitution), that may modify the protein activity.
This heterozygous mutation (CLCN6, p.S116P) segregates with the FTD-like phenotype in Family E and has been found in all the three affected siblings of Family E (EII1, EII2, EII5). Furthermore, also an asymptomatic member in the third generation (EIII3) carried the CLCN6 p.S116P variant. EIII3, aged 52 years, was likely too young to manifest the phenotype (average age at onset for the affected members was 71.3 years) ( Table 4, Fig. 3A,  B). Importantly, CLCN6 p.S116P was the only novel putative loss of function mutation predicted as damaging by at least 2 out of 3 in silico prediction softwares (MUTATION TASTER, POLYPHEN2, SIFT), highly expressed in the brain and highly conserved (Grantham > 50, PhastCons > 0.4 and GERP > 4) ( Table S4, Fig. 3C), segregating with the disease phenotype in Family E, therefore this was the most likely mutation that could have explained the disease in this family.
Family E. Affected members of Family E display a heterogeneous clinical picture, ranging from probable AD, to bvFTD and PPA. The clinical features of the affected members are summarized in Table 4.
The proband of the family deceased at 65 years of age, due to pneumonia and could not be included in the genetic screening. However, relatives described the patient presenting with a dementing syndrome with cognitive and memory impairment.
EII1. At 71 years of age, the patient presented with deficits in short-term memory and attention. No behavioral changes or language impairment were reported. Mild personality fluctuations appeared only during the latter course of the disease. At 76 years of age, the patient was diagnosed with probable AD and later with FTD with memory onset. She was disoriented and died of pneumonia at 80 years of age. EII5. At the age of 75 years, the patient developed a change in personality with aggressive behavior. After three years, he displayed language problems worsening to include mutism and aphasia. After nine years of disease, the patient died due to a cardiac arrest. An MRI scan revealed diffuse cortical atrophy particularly marked in the anterior frontotemporal lobes.

Discussion
We carried out exome sequencing in 100 familial and apparently sporadic patients with FTD-like spectrum disorders and screened dementia and NCL genes. Among the dementia genes, we identified 3 likely pathogenic variants in GRN in 3 sporadic cases (p.C105Y, p.C389fs, p.C139R), one C9orf72 expansion in one sporadic case, 2 CSF1R mutations in the TK flanking regions and one loss of function mutation in SORL1 (p.R744X) in 2/3 affected members of Family H. Additionally, we detected a novel putative LoF mutation in a NCL gene, CLCN6 p.S116P, segregating with FTD with memory onset and PPA in Family E ( Table 2).
We recently reported a GRN novel splice site mutation, GRN c.709-2A > T, in a multigenerational family from the same geographic area 37 in Apulia and in this cohort identified only 3 moderately to frankly pathogenic mutations in 3 apparently sporadic bvFTD cases and showed that GRN mutations may account for only a minority of FTD cases (6.4%), in contrast to the high prevalence of GRN mutations that have been described in a cohort of the nearby Calabria region, where the overall contribution of GRN mutations was 53% (17/32) increasing to 71.4% in patients with family history of dementia (15/21) 38 . Analogously, the frequency of C9orf72 expansions (1%) is much lower than the ones reported in other European countries and Italy particularly (6%) 39 and this is likely not related to the North-South axis as the detected prevalence in Germany was 4.82% and, on the other hand, in Spain 25.49% 39 . This may further point to the isolation of these villages. www.nature.com/scientificreports/ Interestingly, we reported 2 mutations in CSF1R in the TK domain flanking regions (aa 538-581 and 911-972): p.E573K and p.R549H, detected in 2 apparently sporadic patients with late-onset bvFTD. Although mutations in the TK regions of CSF1R (exons 12-22, aa 582-910) are causative for hereditary diffuse leukoencephalopathy with spheroids (HDLS), which clinically manifests as early-onset bvFTD-like with additional Parkinsonism, extrapyramidal or pyramidal signs 40 , also mutation in the CSF1R TK flanking regions have been already associated to early onset PPA 28 and particularly p.E573K leads to a partial loss of the kinase activity and has been reported in a patient with ischemic embolic stroke without the typical clinical features of HDLS 29 , suggesting that missense mutations in the TK flanking regions leading to only a decreased TK activity may cause a milder phenotype compared to HDLS.
Among the dementia genes we detected a loss of function mutation in the VPS10 of SORL1 (Aa 124-757), p.R744X, in 2/3 affected members of Family H displaying late-onset AD with severe language impairment and PPA with pyramidal signs. This mutation was also found in an asymptomatic at risk member of the third generation (HIII2), aged 42 years (average age at onset in Family H is 68 years) and was not detected in another familial member, displaying FTD with memory onset, suggesting that SORL1 (p.R744X) may influence AD with language problems and PPA and that there may be additional genetic modifiers responsible for different phenotypic manifestations.
Importantly, SORL1 variants clustering in the VPS10 domain have been reported as pathogenic and to segregate within AD families 41 particularly with extrapyramidal signs like parkinsonism 42 and language impairment 43 and also to vascular dementia 44 and small vessel disease 45 . Therefore, our finding may support the role of SORL1 influencing motor function and language skills in dementing disorders.
Finally we report a near splice site mutation in CLCN6, p.S116P, segregating with an heterogeneous phenotype (bvFTD, FTD with memory onset and PPA) in Family E.
This mutation has been also reported in an asymptomatic member in the third generation (EIII3) that , aged 52 years, may manifest the phenotype later in life (average age at onset for the affected members was 71.3 years).
CLCN6 encodes for the protein CIC-6, a Clchannel protein that is almost exclusively expressed in neurons. It co-localizes with late endosomes and mediates the exchange of endosomal Clfor cytosolic H +46 . It is plausible that this putative loss of function mutation may lead to a less efficient late endosomal acidification, thus compromising the protein degradation and the autophagosomal pathway, which are pH dependent. Therefore, it may affect TDP-43 degradation, contributing to its cytoplasmatic deposition.
Importantly, in vivo studies with Clcn6 -/mice recapitulate some of the histological and clinical features of lateonset NCL, characterized by the accumulation of storage material (saposinB, lamp-1, cathepsin D and lysosomal acid phosphatase) in the lysosomal system, leading to mild cognitive impairment and behavioral abnormalities 46 . Remarkably, a growing number of studies has shown that NCL and FTD may share common pathogenic mechanisms. First, GRN heterozygous LoF mutations cause FTD whereas homozygous LoF mutations cause NCL 10,47 .
Second, heterozygous mutations in the Cathepsin F (CTSF) gene, that in homozygosity are causative for adult-onset NCL, have been recently reported in a patient with early-onset FTD and motor symptoms 9 .
Third, NCL is characterized by pathological alterations typical of FTD and vice versa: NCL presents a different degree of TDP-43 phosphorylation and GRN-associated FTD is characterized by the elevation of lysosomal proteins and accumulation of saposin B, subunit c of mitochondrial ATP synthase (SCMAS), ubiquitin and p62 protein 48 . Fourth, TMEM106B, VCP, CHMP2B and SORT1, harbor variants identified as disease causing or risk factors for FTD and seem to play a role in endosomal trafficking [49][50][51][52] . As with CLCN6, TMEM106B and CHMP2B co-localize to the late endosomes and appear to be involved in the endosome-lysosome fusion. This represents a critical step for the autophagosome-mediated degradation of proteins and may be involved in TDP-43 turnover. Moreover, CLCN6 has been associated to increased levels of N-terminal cleavage product of the B-type natriuretic peptide (NT-proBNP), a well-established biomarker for dementia 53,54 .
The strength of our study relies on the enormous advantage of performing a genetic analysis in a very inbred FTD cohort from geographically and historically isolated areas and therefore enriched for rare alleles with high penetrance and strong effect size. On the other hand, a limitation of our study is represented by the lack of multigenerational and expanded families to analyze the segregation of rare pathogenic alleles.
Our study includes SORL1 VPS mutations, CSF1R and CLCN6 in the genetic spectrum associated to dementing syndromes with frontal signs, memory deficits, language impairment and pyramidal signs and in concert with a growing body of evidence supports the potential shared pathogenic ground underpinning FTD-like disorders and adult-onset neuronal ceroid-lipofuscinosis.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).