Phenotypic Heterogeneity of Monogenic Frontotemporal Dementia

Frontotemporal dementia (FTD) is a genetically and pathologically heterogeneous disorder characterized by personality changes, language impairment, and deficits of executive functions associated with frontal and temporal lobe degeneration. Different phenotypes have been defined on the basis of presenting clinical symptoms, i.e., the behavioral variant of FTD, the agrammatic variant of primary progressive aphasia, and the semantic variant of PPA. Some patients have an associated movement disorder, either parkinsonism, as in progressive supranuclear palsy and corticobasal syndrome, or motor neuron disease (FTD–MND). A family history of dementia is found in 40% of cases of FTD and about 10% have a clear autosomal-dominant inheritance. Genetic studies have identified several genes associated with monogenic FTD: microtubule-associated protein tau, progranulin, TAR DNA-binding protein 43, valosin-containing protein, charged multivesicular body protein 2B, fused in sarcoma, and the hexanucleotide repeat expansion in intron 1 of the chromosome 9 open reading frame 72. Patients often present with an extensive phenotypic variability, even among different members of the same kindred carrying an identical disease mutation. The objective of the present work is to review and evaluate available literature data in order to highlight recent advances in clinical, biological, and neuroimaging features of monogenic frontotemporal lobar degeneration and try to identify different mechanisms underlying the extreme phenotypic heterogeneity that characterizes this disease.

introduction Frontotemporal dementia (FTD) is the second most common cause of dementia in the presenile age group (<65 years of age), and accounts for 5-15% of all cases of dementia, with a prevalence of 10-15 per 100,000 subjects in the age group of 45-65 years (Bird et al., 2003;Feldman et al., 2003;Borroni et al., 2015b).
A family history of dementia is found in 25-50% of cases of FTD and about 10% have a clear autosomal-dominant inheritance (Seelaar et al., 2008;Rohrer et al., 2009). Genetic studies have identified several genes associated with monogenic FTD: in 1998, mutations in the microtubule-associated protein tau (MAPT) gene on chromosome 17 were identified in families with FTD and parkinsonism (Hutton et al., 1998;Poorkaj et al., 1998;Spillantini et al., 1998). In spite of this finding, there have been numerous reports of families with linkage to chromosome 17q21 but without evidence of MAPT pathogenic mutations (Froelich et al., 1997;Kertesz et al., 2000;Rosso et al., 2001;Rademakers et al., 2002).
In 2006, the second FTD-related gene on chromosome 17 was discovered in the progranulin (GRN) gene Cruts et al., 2006), accounting for even more cases than mutations in MAPT explaining all the 17q21-linked autosomal-dominant FTD cases not accounted for by MAPT mutations (www.molgen. ua.ac.be/ADMutations/).
Eventually, a genetic linkage to a locus on chromosome 9p was found in many families with FTD and motor neuron disease (Morita et al., 2006;Vance et al., 2006;Luty et al., 2008;Le Ber et al., 2009;Boxer et al., 2011); only recently in 2011, two separate groups identified the defect as being an expanded hexanucleotide repeat in a non-coding region of the chromosome 9 open reading frame 72 (C9orf72) gene Renton et al., 2011).
Clinical FTD is associated with different types of underlying neuropathology, and the term frontotemporal lobar degeneration (FTLD), characterized by the relative selective degeneration of the frontal and temporal lobes, is used to describe the pathological hallmarks of the disease. Abnormal intracellular accumulation of insoluble proteins have been identified with immunohistochemical techniques and, according to the predominant neuronal inclusions, FTLD cases have been categorized into several types (Mackenzie et al., 2008. At first, neuronal and glial inclusions of hyperphosphorylated tau protein were identified in FTLDtau, which represent nearly 30-40% of FTLD cases, and are found in patients carrying MAPT gene mutations, as well as in Pick's disease, PSP, and corticobasal degeneration (CBD) (Josephs et al., 2006).
Only briefly after the discovery of the GRN gene, TAR DNAbinding protein 43 (TDP-43) inclusions were identified as the predominant hallmark of the originally identified ubiquitin positive inclusions, which were termed FTLD-U , and subsequently renamed FTLD-TDP (Arai et al., 2006;Neumann et al., 2006). TDP-43 inclusions have also been identified in the majority of amyotrophic lateral sclerosis (ALS) and FTD-MND cases, in particular in patients with mutations of the C9orf72 gene. Four FTLD-TDP subtypes (A-D) have been proposed based on the location of these inclusions .
Finally in 2009, the ~10% of FTLD cases negative for tau or TDP-43 pathology were characterized by inclusions immunoreactive for FUS, which were then labeled as FTLD-FUS (Munoz et al., 2009;Neumann et al., 2009b;Rademakers et al., 2009). Furthermore, the inclusions positive for FUS were found to also stain for other members of the FET family of DNA/RNA-binding proteins, comprising the TATA-binding protein-associated factor 15 (TAF-15) and Ewing sarcoma protein (EWS), distinguishing FTLD with FUS pathology from ALS with FUS mutations .
Patients with monogenic FTD often present with an extensive phenotypic variability, even among different members of the same kindred carrying an identical disease mutation.
The objective of the present work is to review and evaluate available literature data in order to highlight recent advances in clinical, biological, and neuroimaging features of monogenic FTLD and try to identify the hallmarks of different monogenic FTDs.

MAPT Mutations
The MAPT gene on chromosome 17, encoding the tau protein, was the first gene identified in families with FTD and parkinsonism (Hutton et al., 1998;Poorkaj et al., 1998;Spillantini et al., 1998). Tau is a natively unfolded protein that is abundantly expressed in the central nervous system where it regulates microtubule assembly and stabilization, being involved in signal transduction and axonal transport (Goedert et al., 1989b). Forty-four pathogenic mutations in the MAPT gene have been identified so far (Cruts et al., 2012), causing the accumulation of hyperphosphorylated tau protein in neurons and glial cells (Hutton et al., 1998).
Alternative mRNA splicing of exons 2, 3, and 10 of the MAPT gene leads to the assembly of six tau isoforms that are expressed in the adult human brain (Neve et al., 1986;Goedert et al., 1989a;Andreadis et al., 1992). The tau isoforms differ from each other by the presence or absence of a 29-aminoacid or 58-aminoacid insert in the amino-terminal half and by the inclusion or not of a 31-aminoacid repeat encoded by exon 10 of MAPT in the carboxy-terminal half of the protein. The inclusion or exclusion of exon 10 results in the production of three isoforms with four repeats each (4R) or three isoforms with three repeats each (3R), respectively (Spillantini and Goedert, 2013). These repeats comprise the Phenotypes in FTD Frontiers in Aging Neuroscience | www.frontiersin.org microtubule-binding domains of tau, with 4R tau having greater predisposition for microtubule assembly than 3R tau (Goedert and Jakes, 1990).
Mutations can be differentiated into missense mutations in exons 9-13, influencing the normal function of the tau protein in microtubule stabilization, and intronic and some coding mutations modifying the splicing of exon 10 at the mRNA level, resulting in a disproportionate ratio of 3R/4R tau isoforms (Goedert et al., 1989a).
Heterogeneous clinical phenotypes with high penetrance (van Herpen et al., 2003) have been observed in patients with mutations in the MAPT gene, even in families bearing the same pathogenic mutation Janssen et al., 2002;Boeve et al., 2005).
Disease onset is also highly variable, ranging between 45 and 65 years, with a mean age of onset of 55 years, nearly 10 years earlier than in patients with GRN mutations (Van Swieten and Spillantini, 2007). The disease may however occasionally present between 20 and 30 years (Van Swieten and Spillantini, 2007;Momeni et al., 2010) or after 70 years in a few mutations (Hayashi et al., 2002). The mean disease duration is 9 years, ranging from 5 to 20 years (Seelaar et al., 2011).
The clinical phenotype of patients with MAPT mutations can be divided into a dementia-dominant type and a parkinsonismdominant type. The most common presentation, as with GRN mutations, is bvFTD, characterized by behavioral inhibition, impaired social behavior, and obsessive-compulsive disorder (Van Swieten and Spillantini, 2007). As with GRN mutations, patients may subsequently present parkinsonism, which rarely is the exclusive clinical presentation, and more often develops in association with bvFTD (Van Swieten and Spillantini, 2007;Rohrer et al., 2011b). However, some patients have a primary parkinsonian syndrome as in CBS or, more rarely, as in PSP (Dickson et al., 2007;Van Swieten and Spillantini, 2007). Parkinsonian features consist of symmetrical bradykinesia, postural instability and rigidity affecting axial and appendicular musculature, absence of resting tremor, and poor or no responsiveness to levodopa (Ghetti et al., 2015). In patients carrying the N279K mutation, parkinsonism is an early feature and asymmetric resting tremor and levodopa responsiveness have been observed (Tsuboi et al., 2002). Numerous other motor signs and symptoms have been observed, comprising supranuclear gaze palsy, upper and lower motor neuron dysfunction, dystonia, myoclonus, postural and action tremor, apraxia of eyelid opening and closing, dysphagia, and dysarthria (Ghetti et al., 2015).
Patients may also develop language impairment, which usually is not a presenting feature (Seelaar et al., 2008;Rohrer et al., 2010b); indeed the P301L mutation was recently observed in a family with impaired single word comprehension suggestive of verbal semantic impairment (Ishizuka et al., 2011), and the V363I and G304S mutations have been reported in cases with avPPA (Munoz et al., 2007;Villa et al., 2011).
As with GRN mutations, there is no clear correlation between the type of mutation and the clinical presentation; however, anecdotal observations suggest that exonic mutations that do not affect the splicing of exon 10 are usually associated with a dementiapredominant phenotype, and that intronic and exonic mutations that affect exon 10 splicing, leading to an overproduction of 4R tau, tend to be associated with a parkinsonism-predominant phenotype (Ghetti et al., 2015).
In some cases, episodic memory impairment has been observed in patients with MAPT mutations, in particular in the R406W mutation, mimicking Alzheimer's disease (AD) (Tolboom et al., 2010), or in the recently described duplication of the MAPT gene, where the predominant memory impairment was preceded by clinical symptoms characteristic of bvFTD (Rovelet-Lecrux et al., 2010).
Neuroimaging studies have observed heterogeneous patterns of cortical atrophy, but with predominant involvement of the frontal and temporal lobes, with occasional involvement of the parietal lobes Janssen et al., 2002;Boeve et al., 2005;Ghetti et al., 2008;Whitwell et al., 2009a,b). The atrophy is particularly prominent in the anteromedial temporal lobes Whitwell et al., 2009a,c;Rohrer et al., 2010b), with some studies showing that anteromedial atrophy is a feature of early stages of disease (Boeve et al., 2005). In the majority of individuals, patterns of cortical atrophy are relatively symmetrical Ghetti et al., 2008;Rohrer et al., 2011a); however, asymmetry has been observed in some cases (Boeve et al., 2005).
Atrophy of the basal ganglia and of the brainstem have been observed in studies on individual patients, possibly correlating with the parkinsonian features frequently associated with the disease (Arima et al., 2000;Soliveri et al., 2003;Slowinski et al., 2009).
In spite of the fact that temporal lobe atrophy is the most typical finding in patients with MAPT mutations, the pattern of cortical atrophy can differ depending on the specific mutation. Indeed MAPT mutations affecting the splicing of exon 10 are mainly associated with medial temporal atrophy, while mutations in the coding region are primarily associated with lateral temporal lobe atrophy (Whitwell et al., 2009a;Ghetti et al., 2015). These findings, if confirmed in other cohorts of subjects, could help identifying the underlying MAPT mutation.
As in structural imaging, functional imaging with F-18 fluorodeoxyglucose positron emission tomography (FDG-PET) often shows reduced frontal and/or temporal hypometabolism, which resembles the patterns observed in sporadic FTD (Frank et al., 2007).
As in GRN mutations, structural imaging changes can be identified years before expected onset of symptoms in asymptomatic adults at risk of genetic FTD. Rohrer et al. observed differences, between MAPT mutation carriers and non-carriers, in the hippocampus and amygdala 15 years before expected onset, followed by the temporal lobe 10 years before expected onset, and in the insula 5 years before expected onset (Rohrer et al., 2015b).
Functional connectivity in the default mode network (DMN) is altered in MAPT subjects before the occurrence of both atrophy and clinical symptoms, suggesting that changes in functional connectivity are early features of the disease. Asymptomatic MAPT subjects show reduced connectivity in the DMN predominantly between precuneus and lateral temporal lobe, but also medial prefrontal cortex, compared to age-matched controls (Whitwell et al., 2011a). Significantly decreased fractional anisotropy (FA) and increased radial diffusivity in bilateral uncinate fasciculi and reduced FA in the forceps minor have also been observed in asymptomatic MAPT carriers (Dopper et al., 2014).
A preliminary study with the F-18-T807 PET ligand for tau in a patient carrying a MAPT mutation has demonstrated a classic frontotemporal distribution characteristic of inherited tauopathies (Ghetti et al., 2015). These ligands, which characterize tau pathology in vivo (Fodero-Tavoletti et al., 2011;Xia et al., 2013), will undoubtedly improve our future understanding of the pathogenesis of disorders characterized by tau deposition (Villemagne et al., 2015).

GRN Mutations
Since 2006, 69 distinct pathogenic mutations have been identified in the GRN gene, accounting for up to 20% of familial and 5% of sporadic FTD cases (Cruts et al., 2012). Progranulin, which is expressed in many cell types, including neurons and microglia, is a secreted growth factor known for its role in biological processes, including cellular and tissue development, inflammation, wound healing, and cancer, and for its neurotrophic properties. It is proteolytically processed into peptides called granulins, the function of which, in the nervous system, is still largely speculative (Bateman and Bennett, 2009).
Almost all pathogenic mutations cause disease by haploinsufficiency Cruts et al., 2006), reducing the levels of progranulin in cerebrospinal fluid, plasma, and serum in symptomatic and asymptomatic GRN mutation carriers (Ghidoni et al., 2008;Finch et al., 2009;Sleegers et al., 2009). Indeed, the measurement of plasmatic progranulin levels is now largely used to screen patients for GRN sequencing (Schofield et al., 2010;Almeida et al., 2013).
A number of factors that influence progranulin expression have been identified in patients carrying GRN mutations. Genetic variants in TMEM106B (Van Deerlin et al., 2010), a gene encoding for the still uncharacterized transmembrane protein 106B, seem to delay the onset of disease in GRN mutation carriers, perhaps increasing the levels of progranulin (Cruchaga et al., 2011;Finch et al., 2011;van der Zee et al., 2011). Granulin expression is also regulated by microRNAs, including miR-29b and miR-107 (Jiao et al., 2010;Wang et al., 2010), and by sortilin (SORT1), a receptor for neurotrophic factors, which mediates progranulin endocytosis and regulates the plasmatic levels of progranulin (Carrasquillo et al., 2010;Hu et al., 2010).
The clinical presentation in patients with GRN mutations is highly variable, and phenotypic heterogeneity can be observed in the same family carrying an identical pathogenic mutation Gabryelewicz et al., 2010;Yu et al., 2010;Chen-Plotkin et al., 2011), with an age-dependent and incomplete penetrance (Cruts and Van Broeckhoven, 2008). Age at symptom onset varies significantly, ranging from 35 to 89 years, with an average of 65 years, and a penetrance of 50% by the age of 60 and 90% by the age of 70 years van Swieten and Heutink, 2008;.
Within families, there is considerable variation in the age of onset, with a difference of up to 20 years between consecutive generations (Seelaar et al., 2008;van Swieten and Heutink, 2008). Neither age of disease onset nor the duration of the disease seems to have any correlation with the type or location of the GRN mutation.
The initially reported and most common clinical diagnosis in patients carrying GRN mutations is bvFTD, with significant apathy and social withdrawal Cruts et al., 2006). In addition, subsequent studies identified, at onset, a clinical presentation characterized by aphasic disorders consistent with avPPA Mesulam et al., 2007;Le Ber et al., 2008), or CBS (Masellis et al., 2006). The aphasic disorder associated with GRN mutations seems nevertheless to be distinct from the avPPA with apraxia of speech observed occasionally in patients with FTLD-tau (Rohrer et al., 2010a,c), or from the logopenic variant of PPA (lvPPA), commonly associated with an atypical presentation of AD pathology (Gorno-Tempini et al., 2011;Rohrer et al., 2012).
Extrapyramidal features have been reported in about 40% of patients with GRN mutations and include CBS with limb apraxia, asymmetrical parkinsonism, and dystonia, although they are rarely a primary symptom and usually emerge after the onset of behavioral or language symptoms (Yu et al., 2010), and only occasionally respond to levodopa (Carecchio et al., 2009;Di Fabio et al., 2010).
In almost 25% of patients hallucinations and delusions can be predominant, frequently causing a misdiagnosis of dementia with Lewy bodies (DLB) (Le Ber et al., 2008;Kelley et al., 2010;Yu et al., 2010).
Furthermore, an AD-like phenotype can be observed in patients with episodic memory deficit, which occurs in 10-30% of cases, or with parietal deficits, such as dyscalculia, visuospatial symptoms, and limb apraxia. This phenotype can be occasionally associated with the c.154delA mutation in GRN (Kelley et al., 2010) or with the genetic variant rs5848, where the minor T allele, located in the 3′-untranslated region of GRN, was found to increase the binding of miR-659 to GRN (Brouwers et al., 2008;Rademakers et al., 2008;Lee et al., 2011;Sheng et al., 2014), consequently reducing serum progranulin levels (Hsiung et al., 2011).
A recent meta-analysis studied the association between the GRN polymorphism rs5848 and risk of neurodegenerative diseases, including FTLD, AD, Parkinson's disease (PD), and ALS, concluding that rs5848 was associated with an increased risk of AD and PD in the homozygous and recessive models but not with FTLD (Chen et al., 2015).
Since GRN mutations are associated with FTD-TDP pathology, which is frequently observed also in ALS and FTD-ALS, it prompted the search for GRN mutations in patients with associated motor neuron disease. One study identified a missense mutation in a patient with FTD-ALS (p.S120Y) and in a single case of limb onset sporadic ALS (p.T182M) , while a later study identified clinical features compatible with MND in 5.4% of patients carrying a GRN mutation; in this cohort, concomitant motor neuron disease was much more common in patients with FTD-TDP pathology not bearing a GRN mutation (26.3%) (Chen-Plotkin et al., 2011).
Recently, it has been shown that a complete GRN deficiency due to a homozygous GRN loss-of-function mutation causes the rare lysosomal storage disease neuronal ceroid lipofuscinosis (Smith et al., 2012). These findings suggest that lysosomal storage disorders and GRN-associated FTLD may share common features (Götzl et al., 2014).
Neuroimaging studies have highlighted that patients carrying a GRN mutation are more likely to show a pronounced asymmetrical atrophy (Huey et al., 2006;Bronner et al., 2007;Beck et al., 2008;Pickering-Brown et al., 2008) affecting either the left or right hemispheres, involving primarily the inferior frontal (Bronner et al., 2007;Beck et al., 2008;Ghetti et al., 2008), temporal and inferior parietal lobes (Mesulam et al., 2007;Pickering-Brown et al., 2008), as well as long intrahemispheric association white matter (WM) tracts (Borroni et al., 2008;Rohrer and Warren, 2011). A prominent feature, which is unusual in FTLD, is the predominance of parietal involvement (Whitwell et al., 2007;Beck et al., 2008;Le Ber et al., 2008), recently confirmed also in presymptomatic mutation carriers (Rohrer et al., , 2015bCruchaga et al., 2009). These findings have been also confirmed by voxel-based morphometry (VBM) studies, which detected a widespread and severe pattern of gray matter loss predominantly affecting the frontal, temporal, and parietal lobes (Whitwell et al., 2009b(Whitwell et al., , 2012. Along with cortical involvement, atrophy of deep subcortical structures has been identified in patients carrying GRN mutations, in particular in the caudate nucleus and in the thalamus bilaterally, possibly correlating with the parkinsonian features that are frequently associated since the early stages of the disease (Premi et al., 2014b).
Subcortical WM involvement, ranging from T2-weighted MRI hyperintensities (Nuytemans et al., 2007;Le Ber et al., 2008;Kelley et al., 2009) to widespread WM lesions may be linked to progranulin pathological processes in a subset of GRN mutation carriers .
Annualized rates of whole brain atrophy, and consequently of ventricular expansion, are greatest in GRN, followed by sporadic FTD, C9ORF72, and lastly MAPT (Whitwell et al., 2015), in which similar rates of hippocampal atrophy have been observed (Whitwell et al., 2011b).
Using single photon emission computed tomography (SPECT), hypoperfusion of the hippocampus and that of the cingulate gyrus have been observed, denoting the association of these regions with the pathogenesis of GRN mutations (Le Ber et al., 2008).
Recent studies have highlighted that structural imaging and cognitive changes can be identified years before expected onset of symptoms in asymptomatic adults at risk of genetic FTD. In the study by Rohrer et al., in the GRN group, differences were observed between carriers and non-carriers in the insula at 15 years before expected onset, then in the temporal and parietal lobes at 10 years before expected onset, with the earliest subcortical area affected being the striatum at 5 years before expected onset. GRN mutation carriers also showed significantly greater asymmetry than non-carriers at 5 years before expected onset (Rohrer et al., 2015b).
In functional connectivity studies, Premi et al. observed in asymptomatic carriers of the GRN Thr272fs mutation, decreased brain connectivity within the left frontoparietal network and increased connectivity in the executive network compared with healthy controls (Premi et al., 2013b;. Furthermore, the TMEM106B at-risk polymorphism (T/T) was associated with decreased connectivity within the ventral salience network (i.e., middle frontal gyrus) and the left frontoparietal network (i.e., left precuneus) (Premi et al., 2013a). Further studies have confirmed the patterns of brain atrophy and WM tract abnormalities in frontal-parietal circuits and within the uncinate fasciculus, supporting the hypothesis that WM damage is an early feature in the FTD disease process and can be detected at least a decade before the estimated symptom onset in asymptomatic mutation carriers (Dopper et al., 2014;Pievani et al., 2014).

C9orf72 Mutations
The strong association between FTD and ALS has become increasingly apparent due to a rapidly expanding body of clinical, epidemiological, genetic, and molecular evidence (Gros-Louis et al., 2006) and, in 2006, a locus was mapped on chromosome 9p21 in FTD-ALS families (Vance et al., 2006). Until 2011, despite numerous efforts, the disease mutation remained elusive Morita et al., 2006;Valdmanis et al., 2007;Luty et al., 2008;Le Ber et al., 2009;Boxer et al., 2011;Pearson et al., 2011). Eventually, DeJesus-Hernandez and colleagues observed the non-Mendelian inheritance of a GGGGCC hexanucleotide repeat, located in a non-coding region of the C9orf72 gene, in the large VSM-20 family (Vancouver, San Francisco and Mayo family 20) affected by FTD-ALS, with the intuition that the repeat expansion, which was identified in the parents but not in the affected children by using fluorescent PCR, was too large to be amplified by the PCR method and could only be identified by repeated-primed PCR assay and Southern blot analysis . With the use of next-generation sequencing, another group independently recognized the repeat expansion in a Welsh family affected by FTD-ALS (Renton et al., 2011).
Until recently, C9orf72 was a completely uncharacterized protein with unknown function. The pathogenesis underlying the C9orf72 mutation still remains unclear, but likely includes haploinsufficiency through loss of gene expression (Gijselinck et al., 2012;Cruts et al., 2013;, gainof-function mechanisms with secondary RNA toxicity caused by sequestration of RNA-binding proteins Lagier-Tourenne et al., 2013), or both (Rohrer et al., 2015a).
Two different protein isoforms of the C9orf72 gene are predicted to be generated from a total of three different transcripts variants, depending on the localization of the GGGGCC repeat, either in the promoter or in the intron 1 of the gene . All three different transcripts have been found to be reduced in blood (Mori et al., 2013a;Davidson et al., 2014), cortex, cerebellum, and spinal cord (Belzil et al., 2013;Donnelly et al., 2013). This reduced expression possibly reflects, to a limited extent, the hypermethylation of a CpG island upstream of the repeat (Xi et al., 2013;Belzil et al., 2014;Liu et al., 2014), and the methylation of lysine residues on H3 and H4 histones, which bind the GGGGCC repeats and suppress transcription (Xi et al., 2013;Davidson et al., 2014;Russ et al., 2014), acting as a possible innate protective mechanism McMillan et al., 2015;Xi et al., 2015).
As previously mentioned, the neuropathological hallmarks, associated with C9orf72 mutations, are the TDP-43 immunoreactive inclusions, i.e., subtypes A and B (Mahoney et al., 2012;Lashley et al., 2014), which have been observed in many cerebral structures, including the extramotor cerebral cortex, hippocampus, basal ganglia, substantia nigra, and lower motor neurons of the brainstem and spinal cord . In addition to the TDP-43 pathology, a distinctive and singular feature is the presence of abundant TDP-43 negative, p62-positive inclusions, with a unique star-like morphology, in the cerebellum, hippocampus, and frontal cortex (Al-Sarraj et al., 2011;Simon-Sanchez et al., 2012;Snowden et al., 2012;Troakes et al., 2012;Mori et al., 2013b).
Recently, Mori and colleagues tried to characterize these inclusions, speculating that they may consist of dipeptide repeat (DPR) proteins which are the result of the bidirectional translation of the non-coding GGGGCC expansion repeat through a mechanism of non-ATG-initiated translation, termed RAN translation (Mori et al., 2013c). Indeed, the researchers demonstrated that the GGGGCC expansion is bidirectionally translated into five distinct DPR proteins (Mori et al., 2013c), which have been observed in high load in neocortical regions, hippocampus, cerebellum and, to some extent, in motor neurons (Mackenzie et al., 2013). Moreover, the recent observation, in a Belgian C9orf72 carrier, of neuropathological aggregates of a DPR core surrounded by TDP-43 inclusions, suggested that DPR aggregation due to RAN translation could be the initial pathological event that possibly precedes and potentially triggers TDP-43 accumulation (Mackenzie et al., 2013;Mori et al., 2013c;van der Zee and Van Broeckhoven, 2014).
These findings have led to the suggestion that C9orf72 expansion cases should be reclassified, from a neuropathological prospective, as FTLD-DPR, since identified cases can fit into one of two FTLD-TDP subtypes (A or B), or might even lack TDP-43 pathology (Snowden et al., 2012;Mori et al., 2013c).
Despite these new findings implicating both loss-of-function and gain-of-function mechanism with new insights into the roles of sense (GGGGCC) and antisense (GGCCCC) repeat RNA and DPR proteins, the relative contribution to these mechanisms and the molecular pathways that lead to neurodegeneration are yet to be elucidated (Rohrer et al., 2015a).
The number of repeat units of the C9orf72 expansion in affected patients can range from at least several hundred to thousands of copies, whereas unaffected healthy controls harbor 2-24 repeat units Beck et al., 2013;van Blitterswijk et al., 2013). So far, there is no clear evidence for intergenerational anticipation, although some studies have identified earlier disease onset in subsequent generations (Stewart et al., 2012;Benussi et al., 2014).
Whilst it is now rather clear that expansions of several hundred copies are definitely pathogenic, it is not known which is the smallest hexanucleotide expansion that can confer a risk of developing the disease. Indeed, in vitro studies have demonstrated an inverse correlation between repeat size and gene expression starting at just 9 repeats, with a reduction of up to 50% in expression for 24 expansion repeats (van der . Furthermore, as in other non-coding repeat expansion disorders, evidence for somatic instability of the C9orf72 repeat has been observed (DeJesus-Hernandez et al., 2011), resulting in a range of mutation sizes in particular tissues with variation between tissues in the same individual. This possibly suggest that intermediate repeats detected in blood-sample-derived DNA might be associated with massive expansions in the brain (Rohrer et al., 2015a), making it even harder to determine correlations between genetics, phenotype, and neuropathology (Lavedan et al., 1993;Matsuura et al., 2004). Indeed, a recent study tried to estimate, with Southern blotting techniques and densitometry, the repeat size of the most abundant GGGGCC expansion species in patients with ALS and FTD-ALS. Comparing different tissues, substantial variation in repeat sizes between samples from the cerebellum, frontal cortex, and blood have been identified. In particular, longer repeat sizes in the cerebellum seem to be associated with a survival disadvantage, indicating that expansion size does affect disease severity .
As with GRN mutations, TMEM106B has been identified as a genetic modifier also in C9orf72 cases, demonstrating a significantly later age at death and age at onset for the rs1990622 major allele (T) carriers (Gallagher et al., 2014).
Since the discovery of the C9orf72 mutation, testing of patients' cohorts worldwide has found an average frequency of the pathogenic expansion in 37% of familial ALS, 6% of sporadic ALS, 21% of familial FTD, and 6% of sporadic FTD. Epidemiological studies have all established that the C9orf72 mutation is the major genetic cause of familial ALS (more frequent than superoxidose dismutase 1 mutations), and comparable in frequency to GRN mutations in familial FTD cases Renton et al., 2011;Chiò et al., 2012;Hsiung et al., 2012;Majounie et al., 2012;Cruts et al., 2013;.
Age at disease onset is widely variable, ranging from 27 to 83 years (mean 50 years), with about 50% patients affected by age 60 and with a penetrance reaching 90% at age 70. Disease duration is also variable, ranging from 1 to 22 years, with the development of ALS as a strong negative prognostic factor Chiò et al., 2012;Cooper-Knock et al., 2012;Gijselinck et al., 2012;Hsiung et al., 2012;Stewart et al., 2012).
Neuroimaging studies using VBM analysis have observed extensive atrophy of the frontal, insular, temporal, and parietal cortical and subcortical regions, including prominent atrophy of the thalamus and cerebellum (Mahoney et al., 2012;Whitwell et al., 2012;Irwin et al., 2013).
Concordant with gray matter atrophy, WM atrophy of long intrahemispheric, commissural, and corticospinal tracts has been observed with diffusion tractography (Mahoney et al., 2012).
As with structural neuroimaging, functional neuroimaging using FDG-PET and SPECT has shown frontal, temporal, and parietal cortical and subcortical alterations Adeli et al., 2014;Suárez-Calvet et al., 2014). Seed-based intrinsic connectivity analyses in C9orf72 carriers versus non-carriers showed connectivity reductions in the salience and sensorimotor networks, with severity of behavioral symptom correlating with diminished salience network connectivity and increased DMN connectivity (Lee et al., 2014).
As in GRN and MAPT mutations, presymptomatic imaging signatures can be identified in asymptomatic adults at risk of genetic FTD. Rohrer et al. observed differences, between C9orf72 mutation carriers and non-carriers, in subcortical areas, including the thalamus, the insula, and posterior cortical areas 25 years before expected onset, followed by the frontal and temporal lobes 20 years before expected onset, and in the cerebellum 10 years before expected onset (Rohrer et al., 2015b).

VCP Mutations
Missense mutations in the VCP gene, located on chromosome 9p13.3, were originally reported in 2004 in North American families with a specific autosomal dominant multisystem syndrome called inclusion body myopathy with Paget disease of bone and/ or FTD (IBMPFD) (Watts et al., 2004), with neuropathology consistent with type D TDP-43 pathology .
Valosin-containing protein, also known as p97, is a member of the ATPases associated with various cellular activities (AAA) family of proteins and is involved in multiple cellular processes, including protein degradation via both proteasome and autophagy pathways, in membrane fusion, transcriptional activation, apoptosis, and molecular chaperoning (Nalbandian et al., 2011;Yamanaka et al., 2012). The role of mutant VCP in the pathogenesis of IBMPFD or FTD-ALS remains elusive.
So far, 19 pathogenic mutations in VCP have been described (Cruts et al., 2012). The frequency of VCP mutations in FTD is about 1.6%, making it a rare but significant cause of FTD . Nearly 80% of VCP mutation carriers have a positive family history with a wide variability in phenotypic presentation, even within the same family.
The disease usually begins at the age of 40 with musculoskeletal symptoms not dissimilar to limb-girdle muscular dystrophy, affecting more than 80% patients. Moreover, nearly 60% patients develop adult-onset Paget disease of bone (Watts et al., 2004). FTD usually develops later on in the disease course (age 55-60 years), affecting about 30% of patients, with a classical bvFTD phenotype (Kovach et al., 2001;Guyant-Marechal et al., 2006;Kimonis et al., 2008); occasional psychiatric disorders have been reported (Jacquin et al., 2013). However, although rare, VCP mutations have been found in patients with a dementia-only phenotype (van der Zee et al., 2009). Early language deficits and semantic impairment have been rarely described (Kim et al., 2011).
VCP mutations have been identified also in patients with an MND/ALS phenotype, occasionally associated with FTD (Johnson et al., 2010;Miller et al., 2012;Hirano et al., 2015); these findings are supported by the pyramidal tract dysfunction which has been observed in some cases (Kumar et al., 2010).
The term multisystem proteinopathy has been recently proposed to replace the term IBMPFD to describe this phenotype (Benatar et al., 2013).

TARDBP Mutations
Transactive response DNA-binding protein 43 (TDP-43) is a ubiquitously expressed and highly conserved ribonucleoprotein, encoded by the TARDBP gene on chromosome 1, with a putative role in the regulation of transcriptional activity, of messenger RNA splicing, exon skipping, and microRNA biogenesis of several genes. It has been proposed that, in neurodegenerative disorders, TDP-43 can be redistributed from the nucleus to the cytoplasm and sequestered into inclusions that are mainly composed of abnormally phosphorylated and C-terminally truncated TDP-43 fragments (Buratti and Baralle, 2008;Polymenidou et al., 2011;Tollervey et al., 2011). The identification of mutations within the TARDBP gene as causative for both familial and sporadic ALS provided additional proofs for the pathogenic role of TDP-43 in neurodegeneration (Kabashi et al., 2008;Rutherford et al., 2008;Sreedharan et al., 2008;Van Deerlin et al., 2008).
TDP-43 inclusions, as previously outlined, have been observed also in non-TARDBP-associated genetic forms of FTLD, including GRN, VCP, and C9orf72 mutations, suggesting that dysregulation of TDP-43 might be a pivotal downstream pathogenic mechanism involved in neurodegeneration in each of these disorders .

CHMP2B Mutations
The CHMP2B gene, located on chromosome 3p11.2, encodes the charged multivesicular body protein 2B, which has been identified in 2005 in a large Danish family with an autosomal form of FTLD (Skibinski et al., 2005), first described in 1987 (Gydesen et al., 1987). The average age of onset in this family was 55 years, ranging from 46 to 65 years. Since 2005, researchers identified CHMP2B missense variants in FTD and FTD-ALS cases, representing less than 1% of familial forms of FTLD, although the pathogenicity of these variants remain to be elucidated (Parkinson et al., 2006;Cox et al., 2010).
The phenotypic expression of the originally described family comprises bvFTD that may be associated with late parkinsonism, dystonia, myoclonus, and pyramidal symptoms (Skibinski et al., 2005).

FUS Mutations
The genetic heterogeneity of FTD is further emphasized by the occurrence of rare mutations in the FUS gene, located on chromosome 16, which was first identified as a fusion oncogene in human liposarcomas (Crozat et al., 1993;Rabbitts et al., 1993).
FUS, as TAF-15 and EWS, belongs to the FET family of DNA/ RNA-binding proteins, which are highly conserved, ubiquitously expressed, multifunctional binding proteins (Tan and Manley, 2009) that can bind to a large number of overlapping RNA targets (Hoell et al., 2011). Although the precise physiological function of FUS has not been fully elucidated, growing evidence suggests that the protein is involved in various cellular processes, including cell proliferation, DNA repair, transcription regulation, and multiple levels of RNA and microRNA processing (Lagier-Tourenne et al., 2010;Deng et al., 2014).
FUS mutations in patients with clinical FTLD are rare, and only four FUS mutations have so far been identified in patients with FTD-MND/ALS or in their families (Ticozzi et al., 2009;Blair et al., 2010;Broustal et al., 2010;Yan et al., 2010). Moreover, two cases with a pure FTD phenotype without motor neuron involvement Phenotypes in FTD Frontiers in Aging Neuroscience | www.frontiersin.org have also been reported (Van Langenhove et al., 2010;Huey et al., 2012). Nevertheless, no autopsy data are yet available to confirm FUS pathology in any patient with FTLD and FUS mutations.
In contrast to ALS with FUS pathology, which is mostly caused by FUS mutations, FTLD-FUS tends to be sporadic, with none yet associated with a mutation in the FUS gene .
Different pathogenic mechanisms in ALS-FUS and FTLD-FUS also arise from a recent study in which, in ALS-FUS, no co-accumulation of other FET-proteins into FUS-positive inclusions was found, in remarkable contrast to FTLD-FUS, in which TAF-15 and EWS were found to co-accumulate in FUS-positive cytoplasmic inclusions .
Structural neuroimaging studies show predominant volume loss in the anterior temporal lobe, amygdala, anterior hippocampus, anterior insula, and anterior cingulate cortex (Lee et al., 2012). Bilateral anterior striatal atrophy is particularly prominent in the head of the caudate nucleus, which appears to be significantly greater in FTLD-FUS compared with FTLD-TDP and FTLD-tau, suggesting that severe caudate atrophy may be a useful clinical feature to predict FTLD-FUS pathology Lee et al., 2012).
Homozygous loss-of-function mutations in TREM2 on chromosome 6p21, encoding the triggering receptor expressed on myeloid cells 2 protein, have been initially associated with an autosomal recessive form of early onset dementia presenting with bone cysts and consequent fractures called polycystic lipomembranous osteodysplasia with sclerosing leukoencephalopathy, also known as Nasu-Hakola disease (Paloneva et al., 2000(Paloneva et al., , 2002. In 2013, two studies identified a rare heterozygous variant (R47H) on the second exon of TREM2, showing an approximately three-to fivefold increased risk in late onset AD (Guerreiro et al., 2013c;Jonsson et al., 2013). Furthermore, homozygous and compound heterozygous TREM2 exon 2 mutations were identified in rare families presenting with autosomal recessive bvFTD without any clinical or radiological evidence of bone involvement (Giraldo et al., 2013;Guerreiro, 2013;Guerreiro et al., 2013a;Lattante et al., 2013;Rayaprolu et al., 2013;Le Ber et al., 2014). Patients typically have very young age at onset (20-50 years) and, in addition to behavioral changes, they present with early parietal and hippocampal deficits, extrapyramidal features, and seizures. Neuroimaging frequently shows frontotemporal atrophy and extensive WM abnormalities (Guerreiro, 2013;Le Ber et al., 2014).
Recently, Borroni and colleagues evaluated the frequency of heterozygous TREM2 mutations as risk factor in FTD patients, confirming that the Q33X, R47H, and T66M mutations were significantly associated with the disease and absent in healthy subjects , in line with further reports (Cuyvers et al., 2013).

Conclusion
The striking clinical, genetic, and pathological heterogeneity are undisputable features of FTLD. The clinical phenotypes range from social to language impairment, variably associated with MND/ALS or extrapyramidal symptoms (see Table 1). In the past 9 years, considerable progress has been made toward unraveling the genetic causes of FTLD, with apparently all common FTD-related genes discovered, comprising C9orf72, MAPT, GRN, TARDBP, VCP, CHMP2B, and FUS, currently identifying 50-60% of the familial forms of FTD, although rare mutations in other genes could explain a small number of the remaining families .
Neuropathology is variably associated with two predominant hallmarks, namely FTLD-tau (found in MAPT mutations) and FTLD-TDP (found in GRN, C9orf72, TARDPB, and VCP mutations), although it is occasionally associated with FTLD-FUS (found in FUS mutations only in MND/ALS), FTLD-U (found in CHMP2B mutations) and the newly identified FTLD-DPR in C9orf72 mutations.
Why different genes linked to different molecular pathways are responsible for a neurodegenerative process involving mainly the frontotemporal lobes is still a matter for debate, and the shared mechanisms that lead to predominant frontal and temporal degeneration remain to be identified.
As previously outlined, frequently there is poor correspondence between the type of mutation or neuropathological feature and the clinical phenotype; indeed, neuropathology is  predicable only in known genetic defects and few clinical or neuroanatomical features have a specific molecular association (Rohrer et al., 2011a). Both clinically and anatomically, different genetic mutations show substantial overlap between diseases. However, certain distinctive clinical and neuroimaging features emerge and might help in driving neurology practice in molecular diagnostics. Patients with C9orf72 repeat expansion usually have an early disease onset and a family history of FTD (mainly bvFTD with disinhibition or svPPA), which is variably associated with MND/ALS. MAPT mutation carriers often present with a very early age of onset bvFTD, associated with parkinsonism, which is a common feature later in the disease course in all types of genetic FTLD. GRN mutations carriers frequently present with bvFTD, avPPA, and concomitant CBS at onset. TARDBP sequencing should be performed in familial or sporadic C9orf72 negative FTD-MND/ALS cases. The association of inclusion body myopathy and Paget's disease are indicative of VCP mutations (Rohrer and Warren, 2011;Borroni and Padovani, 2013;. In addition to these distinct clinical features, from a neuroimaging perspective, strikingly asymmetric interhemispheric atrophy with frontoparietal involvement is indicative of GRN mutations, while relative symmetric atrophy with particular involvement of the anterior temporal lobes is indicative of MAPT mutations (Rohrer et al., 2010b;Whitwell et al., 2012). On the other hand, C9orf72 mutation carriers tend to have a more generalized and symmetrical atrophy (Rohrer et al., 2010b;Whitwell et al., 2012;) (see Table 2).
Numerous attempts have been made to identify an in vivo biomarker that can accurately predict the underlying neuropathology and consequently point out which gene should be analyzed, but a clear-cut association between clinical or neuroradiological features and characteristic histological inclusions has not been yet established (Kim et al., 2007;Josephs et al., 2008;McMillan et al., 2013). Plasma progranulin levels have been shown to be reliable predictors of GRN mutations, and should be performed before GRN sequencing (Finch et al., 2009;Ghidoni et al., 2012). Furthermore, the recently identified cerebrospinal phospho-tau181 to total-tau ratio appears to predict FTLD-TDP pathology with high accuracy, thus leading to the search of genes associated with TDP-43 neuropathology (Hu et al., 2013;Grossman et al., 2014;Borroni et al., 2015a).
Considering the vast clinical, genetic, and pathological heterogeneity associated with this disease, the development of accurate in vivo biomarkers is mandatory for guiding genetic screening, for defining distinctive therapeutic approaches and helping in patients' selection in future clinical trials.