Review: An update on clinical, genetic and pathological aspects of frontotemporal lobar degenerations

Introduction

Frontotemporal lobar degeneration (FTLD) is a pathological term used for the description of a clinically, pathologically and genetically heterogeneous group of disorders, in which relatively selective degeneration of the frontal and temporal lobes is a prominent and common feature 1. The clinical term frontotemporal dementia (FTD) is used for the description of a group of early onset dementias, which is the second most common dementing disorder among individuals under the age of 65 years 2. However, in 25% of the cases FTD presents in old age 3. The estimated prevalence of FTD is 15–22/100 000 and population studies indicate an equal gender distribution 3. Clinically FTD patients can present with one of three canonical clinical syndromes: behavioural variant FTD (bvFTD) and two language variants, semantic dementia and progressive nonfluent aphasia (PNFA) (see later). FTD can overlap with motor neuron disease/amyotrophic lateral sclerosis (MND/ALS) (FTD-MND), corticobasal syndrome (CBS) and progressive supranuclear palsy (PSP) syndrome 4. FTD is a highly heritable disorder with approximately 30–50% of cases reporting a positive family history 5. Mutations in three genes microtubule-associated protein tau (MAPT), progranulin (GRN) and chromosome 9 open reading frame 72 (C9orf72) genes being responsible for most of the familial cases, and about 10–20% of all cases with FTD 5, 6. The current neuropathological classification of FTLDs recognizes five major subgroups, three of which are characterized by specific proteinaceous inclusions: tau in FTLD-tau, 43 kDa transactive response DNA-binding protein (TDP-43) in FTLD-TDP and fused in sarcoma (FUS) in FTLD-FUS. The protein nature of the ubiquitin-positive inclusions has not been identified in the fourth group and is currently classified as FTLD-UPS, represented mostly by cases of affected individuals of a Danish pedigree due to mutation in the charged multivesicular body protein 2B (CHMP2B) gene while no inclusions are found in a small minority of the cases designated as FTLD-ni. In this review, we also provide a summary of the evolution of the pathological and clinical concepts of FTD, and discuss the genetics and the main neuropathological findings of the FTLD subgroups and summarize issues related to the pathogenesis of the relevant disorders. The most recent classification is also provided in this review.

Evolution of the pathological concept of FTLDs

The concept of circumscribed cerebral cortical atrophy is over 120 years old (Figure 1). Between 1892 and 1906, Arnold Pick, the eminent Czech-Austrian neuropsychiatrist and head of the Department of Psychiatry at the Charles University in Prague, in a series of pioneering papers described a number of patients with such type of cortical atrophy. Pick, trained with Meynert in Vienna and Westphal in Berlin, is one of the forefathers of modern cognitive neurology, whose scientific ideas were greatly influenced by the work of John Hughlings Jackson in London 7-9. The first of Pick's studies on circumscribed cerebral atrophy is his seminal paper, published in the Praguer Medizinische Wochenschrift in 1892, in which he described the case of August H, a 71-year-old male with progressive cognitive decline and prominent speech disorder consisting of poor understanding of speech and written commands, paraphasias and partial preservation of repetition, for which Pick used Wernicke's term, ‘transcortical sensory aphasia’ 10. Post mortem examination, performed by Pick's pathologist colleague Chiari, confirmed cerebral atrophy, which was more prominent in the left than in the right hemisphere and Pick postulated that the particularly severe atrophy of the left superior temporal gyrus (‘atrophia cerebri praecipue haemisphaerii sin. in regione gyri primi lobi sphenoidalis’) was responsible for his patient's aphasia 10. Pick's original paper provided no detailed microscopic description and the argyrophilic globular neuronal cytoplasmic inclusions, subsequently named as Pick bodies and are considered pathognomonic of Pick's disease (PiD) today, were described and illustrated by Alois Alzheimer in 1911 (Figure 2) 11. As a result of clinicopathological observations a new disease entity was gradually emerging, and in 1925 the medical eponym PiD was first introduced by A. Gans as ‘Ziekte van Pick’ in the Dutch 12, and a year later by Onari and Spatz, who were aware of Gans's work, as ‘Picksche Krankheit’ in the German medical literature 13.

In the following decades, the clinical and neuropathological heterogeneity of cases with frontotemporal lobar atrophy became apparent, which was also supported by findings indicating that argyrophilic Pick bodies were present only in about 20% of larger case series 14. The neuropathological classification of PiD, put forward by Constantinidis and his colleagues in 1974 15, was based on this notion of clinical and morphological heterogeneity of a disease for which the umbrella term PiD was still used. These authors distinguished three major categories of PiD on the basis of the distribution of the pathology and the presence or absence of swollen, achromatic neurons (Pick cells) and the argyrophilic Pick bodies. Cases with both Pick bodies and Pick cells were classified as group A, cases that only possessed swollen neurons belonged to group B, while both Pick cells and Pick bodies were absent in group C. The striking absence of either Pick bodies or Alzheimer-type pathological changes in the majority of cases with FTD was also emphasized in the subsequent pioneering studies of the research groups of Lund and Manchester Universities 16, 17. A pattern of pathological changes was also identified in such cases, which includes loss of neurons accompanied by gliosis and microvacuolation of the neuropil in superficial cortical laminae without amyloid plaques, neurofibrillary tangles or other ‘silver positive’ changes (for a review see 18). In keeping with the ‘nonspecific’ nature of such pathology, the term ‘dementia lacking distinctive histological features’ (DLDH) was also coined and subsequently widely used 19. During the 1990s, as a prelude to the molecular understanding of FTLDs, ubiquitin-positive inclusions were recognized in extramotor cerebral structures such as neurons of the dentate fascia or the frontal and temporal cortices in cases with MND/ALS with associated FTD 20, 21. The link between ubiquitin-immunoreactive extramotor neuronal cytoplasmic inclusions, initially described in MND/ALS, and FTLD was subsequently firmly established by the recognition that such inclusions also occur in cases with progressive FTD, but without clinical evidence of either upper or lower motor neuron involvement, which resulted in the introduction of the FTLD with ubiquitin-positive inclusions (FTLD-U) concept 22, subsequently underpinned by several clinicopathological studies 23-26. This knowledge gave the impetus for pathological studies to be performed, which demonstrated that ubiquitin-positive inclusions are characteristic perhaps in a majority of clinically well-documented FTD cases 23, 26 and that the neuropathological diagnosis of DLDH is only rarely justified 26-28. The following two decades witnessed several landmark discoveries. In the 1990s, the tau isoform composition of sporadic tauopathies, such as PiD, PSP and corticobasal degeneration (CBD) 29-31, which according to the currently used classification system are part of the FTLD-tau spectrum 32, was identified. Furthermore, in 1998, mutations in the microtubule-associated protein tau gene (MAPT) were reported, indicating that a significant proportion of hereditary FTDs, previously linked to chromosome 17 (17q21–22), is due to mutations in this gene 33, 34. Further landmark discoveries include the recognition that TDP-43 is the main component of the ubiquitin-positive inclusions in the majority of the FTLD-U cases and MND/ALS 35. Involvement of TDP-43 in both FTLD-TDP and MND/ALS also provided a firm molecular link between FTLD and MND/ALS underpinning the notion that these two large groups of neurodegenerative disorders represent two, often overlapping ends of a disease spectrum. This is also supported by clinical data indicating that about 30% of patients with FTD develop clinical signs of MND/ALS and that 50% of MND/ALS patients show some evidence of cognitive deficits 36. In addition to the MAPT gene, a number of further genes associated with different forms of familial FTD and/or MND/ALS was identified, including the progranulin (GRN) gene, whose mutations are responsible for Frontotemporal dementia and parkinsonism linked to chromosome 17 (FTDP-17). FTDP-17 in families without harbouring an MAPT mutation 37, 38 or the much rarer mutations in the valosin-containing protein (VCP) gene in familial FTD 39 also associated with Paget's disease and inclusion body myositis 40, TARDP and FUS genes in MND/ALS 41, 42, as well as the CHMP2B gene in familial FTD linked to chromosome 3 (FTD-3), described in a Danish pedigree 43. The discovery in 2011 that intronic hexanucleotide (GGGGCC) repeat expansion in the C9orf72 gene is a common genetic cause of ALS/MND and FTLD (C9FTD/ALS), is a recent addition to the list of genes whose mutations play an important role in FTLD 44, 45. This rapid increase in knowledge also allowed that a molecular classification of FTLDs has become available (Figures 1 and 3), which is now widely accepted and used 32, 46.

Clinical concept of FTD

After the publication of Pick's original paper in the late 19th century, further publications documented that patients with circumscribed cerebral atrophy had both personality change and language impairment consistent with the modern day split of FTD into behavioural and language variants. However, despite these descriptions the early 20th century saw limited reports of such cases in the Western world, although in Japan the disorder Gogi aphasia (a progressive language disorder) was described in the 1940's 47.

The first modern accounts of the progressive language disorders in the Western literature were by Warrington 48 and Mesulam 49. Warrington described the selective impairment of semantic memory in a group of patients, although she did not use the term semantic dementia, which was coined a number of years later 1, 50. Mesulam independently described a group of patients with progressive language problems initially calling them ‘slowly progressive aphasia without generalized dementia’ 49, and then later ‘primary progressive aphasia’ (PPA) 51, a term that has stuck to this present day. Around the same time period, in the late 1980s and early 1990s, a group of researchers came together to start to understand better the bvFTD. This led to the first International Conference of Frontal Lobe Degeneration of non-Alzheimer type in Lund, Sweden and subsequently a set of diagnostic criteria in 1994 52.

The late 1990s and early 2000s saw a flowering of FTD research. One of the keys to this was the development of criteria for both the behavioural and language variants by a group of researchers in the field in 1998 1. These criteria coined the term FTLD and defined a behavioural variant (called FTD) and two variants of progressive aphasia: PNFA and semantic dementia. Arguments over the nosology of these language variants has continued over the years and in 2004, a third variant was described by Gorno-Tempini et al., called logopenic aphasia (LPA) 53. More recently, this variant was incorporated into a new set of diagnostic criteria for the progressive language disorders (PPA), which were renamed, the nonfluent/agrammatic variant, the semantic variant and the logopenic variants of PPA 53. Around the same time, an international consortium came together to update the diagnostic criteria for the behavioural variant, now known as bvFTD 54.

bvFTD is characterized by a set of behavioural symptoms including disinhibition, apathy, abnormal appetite (commonly a sweet tooth), loss of empathy and obsessive-compulsive behaviour. However, patients also develop cognitive impairment, usually executive dysfunction initially, but other cognitive domains also become involved as the disease progresses. Patients with nonfluent variant of PPA develop agrammatism and/or apraxia of speech, while those with the semantic variant develop anomia and impaired single word comprehension, later developing nonverbal semantic impairment as well; the LPA variant is characterized by word-finding pauses and impaired working memory.

Neither of the current diagnostic criteria describes the overlap of FTD with MND/ALS or the atypical parkinsonian disorders, CBS and PSP. The overlap with these disorders has been increasingly recognized in recent years with a substantial proportion of bvFTD patients (and a smaller number of nonfluent aphasia cases) developing MND/ALS or parkinsonism at some point during the disease process.

Genetics of FTD

A family history of a similar disease is common in FTD, typically in a pattern that suggests dominant inheritance 5. In populations of European ancestry, there are three major FTD disease genes (MAPT, GRN and C9orf72) and many rarer disease genes. It is also recognized that mutations in genes more commonly associated with related neurodegenerative diseases such as Alzheimer's disease, MND/ALS, Parkinson's disease or prion disease can be found in patients with clinical diagnoses of FTD. The frequency of mutations in specific regions or clinical centres varies according to population factors such as founder effects, drift and migration; however, typically, a clinic will find high single digit percentages of all FTD are explained by each of the three major genes individually. A substantial proportion of familial FTD, depending on how this is defined, remains unexplained 5. A core clinical and pathological phenotype is recognized for each gene, although clinical distinction is imperfect, particularly in the early stages when gene testing is discussed with patients and their families. Gene panel technologies are now available to screen all known common and rare FTD genes simultaneously 55, 56.

In pedigrees with several affected individuals, the location of the responsible gene can be narrowed by genetic linkage using markers across the genome. The region including the MAPT gene on chromosome 17 was first linked with familial FTD in 1994 57, and mutations in this gene were first described in 1998 33. MAPT gene mutations are typically associated with bvFTD. Associated syndromes can include parkinsonism, aphasia, early amnesia and semantic problems. Over 50 different causal mutations are now known. They fall into two broad categories: missense mutations, which alter the amino acid code of the protein usually in or near to the fourth repeat region of the microtubule-binding domains (see later), and splice site mutations, which alter the inclusion of exon 10 58. Common genetic variation near to the MAPT gene is a risk factor in PSP and Parkinson's disease 59, but not FTD.

Mutations in the GRN gene, coincidentally near to MAPT on chromosome 17, were identified as causal of FTD in 2006 37, 38. The protein product, progranulin, is a secreted glycoprotein, cleaved into granulin peptides and found in the brain and serum with roles in inflammatory diseases, diabetes and obesity 60. Complete loss of progranulin protein is associated with neuronal ceroid lipofuscinosis 61. Like MAPT mutations, they are most commonly associated with the bvFTD; however, nonfluent aphasia, CBS and other Parkinsonian syndromes have been reported. Over 150 variants of GRN have been reported in patients 62. The large majority of mutations are premature stop codons, splice site, exon/gene deletions or frameshift mutations, which result in a hemi-loss of functional protein. Serum levels of progranulin are reduced by ~50% in mutation carriers. Most of the missense changes in GRN, aside from those in or near to the transcription initiation codon are classified as variants of unknown significance. Variants of TMEM106b modify the clinical phenotype of GRN mutation associated FTD 63.

The C9orf72 gene on chromosome 9 has a hexanucleotide repeat region located in either in the promoter or intron 1 of the gene (depending on the transcript variant). When massively expanded, the mutation causes FTD, MND/ALS or the combined syndrome 44, 45. Particularly high frequencies of the C9orf72 expansion mutation are found in Finland because of a founder effect 64. Although the repeat region is variable in length in the healthy population (up to around 30 repeats), repeat expansions typically seen in patients are far more massively expanded, typically >400 repeats 44, 65, 66. We still do not have certainty about the smallest hexanucleotide expansion that confers risk. The core clinical phenotype is the combination of FTD and MND/ALS; however, the expansion mutation is commonly associated with pure FTD, pure MND/ALS and may be reported in patients labelled with AD, movement disorders, ataxias and nonspecific neuropsychiatric/neurodegenerative syndromes (for a review see 67). No other mutation types have been reported in the gene and the function of the C9orf72 protein is not well understood.

Mutations in several other genes have been uncommonly associated with FTD. These might best be classified as rare gene mutations, which typically cause FTD (CHMP2B and VCP, the latter in association with inclusion body myopathy and/or Paget's disease of bone), or more common gene mutations associated with neurodegeneration, but rarely the FTD phenotype (Alzheimer's disease genes: PSEN1, APP; prion disease gene: PRNP; ALS genes: FUS, TARDBP; others: CSF1R).

Neuropathology of FTLD – an overview of current classification

As a result of recent advances in the understanding of the molecular mechanisms associated with FTLDs, this heterogeneous group of diseases are now subdivided pathologically on the basis of the deposited abnormal intracellular protein aggregates 32, 46, 68 (Figure 3). In approximately 40% of all FTLD cases the microtubule-associated protein tau forms inclusions in neurons, and in some diseases, in both neurons and glial cells (both astrocytes and oligodendrocytes) (FTLD-tau). With the exception of a small minority of the cases, the tau-negative cases have neuronal inclusions, which were originally identified by their immunoreactivity for ubiquitin (FTLD-U) 22, 27, 69. The majority of the FTLD-U cases are associated with accumulation of TDP-43 35, and the term FTLD-TDP is assigned to this group 32, 46. In a smaller group of the FTLD-U, the inclusions are positive for the FUS protein (FTLD-FUS) 32, 70, 71, and such cases are responsible for the majority of tau and TDP-43-negative FTLDs representing about 5–10% of ubiquitin-positive FTLDs. FTLD-FUS includes three diseases, neurofilament inclusion body disease (NIFID), atypical FTLD-U (aFTLD-U) and basophilic inclusions body disease (BIBD) 70-73. Two neuropathological subtypes still remain elusive; one primarily representing familial FTLD with CHMP2B mutation (FTD-3) and with ubiquitin-positive inclusions, which are negative for tau, TDP-43 or FUS and is now termed FTLD-UPS, and a second group, termed FTLD-ni, which contains rare cases with no discernible pathological inclusions 32.

FTLD-tau

The natively unfolded tau protein with a suggested role in microtubule assembly and stabilization as well as regulation of axonal transport, is the most commonly misfolded protein forming intracellular inclusions in human neurodegenerative diseases 74. Tauopathies are responsible for a major group of FTLDs, designated as FTLD-tau 32, which represents a clinically, biochemically, genetically and pathologically heterogeneous group of diseases with the overarching feature that all forms possess abundant neuronal, and in some variants, both neuronal and glial filamentous inclusions. These are composed of abnormally hyperphosphorylated form of tau in the absence of amyloid-β deposits 74, 75; hence, they are ‘primary’ tauopathies 76. In addition to tau phosphorylation, which is an early event of tau aggregation, a number of other post-translational modifications of tau are also known. These include ubiquitination, acetylation, glycation, glycosylation with O-linked N-acetylglucosamine (O-GlcNAcylation), nitration, sumoylation, and truncation, which are also thought to be associated with the aggregation process 58. It is widely accepted that the inclusions contribute to disease pathogenesis in tauopathies as they occur in specific brain regions whose functions are altered by the tau pathology 77. In this regard, formation of tau aggregates and filaments, which possess a cross-β structure characteristic of amyloid fibrils, is a highly relevant process 58, which is also supported by experimental data indicating that tau aggregates can induce neurotoxicity 78 and are associated with nerve cell dysfunction 79.

The current molecular classification of FTLD-tau is based upon genetics and the biochemical composition of the inclusions. FTLD-tau includes patients with MAPT gene mutations (FTD and parkinsonism linked to chromosome 17 or FTDP-17 MAPT), PiD, PSP, corticobasal degeneration (CBD), argyrophilic grain disease (AGD), globular glial tauopathy (GGT) and NFT-dementia, which was recently redefined (somewhat controversially 80) and designated as primary age-related tauopathy or PART (Figures 3 and 4) 32, 81-84. It should be noted that the clinical presentation is not that of FTD in the majority of AGD and PART cases. Although both PSP and CBD are clinically often referred to as atypical Parkinsonian disorders, they are now recognized as disorders of both movement and cognition 85-89. It is well documented that both CBD and PSP patients may present with bvFTD 88-92 or language impairment (PSP-FTD variants) 89, 92-94. It is of note that the ‘cortical’ PSP variants, which, in addition to PSP-FTD also include PSP-CBS (PSP with CBS), are associated with increased neocortical tau load due to a shift of the tau burden from the basal ganglia towards the cerebral cortices 95-97. An intriguing aspect of several protein folding neurodegenerative disorders is clinical disease progression, which is thought to be due to a stereotypic spatial, time dependent expansion of the underlying pathology. Data from Alzheimer's disease indicate that clinical progression is dependent on such an anatomical progression of the neurofibrillary tau pathology, described by the Braak and Braak staging system, which is widely used in neuropathological practice 98. Recent experimental data indicate that tau can show prion-like propagation and spread, which is consistent with pathological disease progression in human disease 74. However, apart from AGD 99, staging of tau pathology has not been established in other forms of FTLD-tau such as PiD, CBD and PSP despite significant advancement in recognizing disease subtypes.

FTLD-tau can be divided on the basis of the predominant tau species making up the inclusions. The six tau isoforms, which are expressed in the adult human brain, are produced by alternative mRNA splicing from the MAPT gene 100, 101. The isoforms differ from one another by whether they possess 0, 1 or 2 29-amino-acid-long N-terminal inserts, encoded by exons 2 and 3 and by the presence or absence of a fourth 31-amino-acid-long repeat in the microtubule-binding domain of tau, encoded by exon 10. Two major groups of the six tau isoforms, each group containing three isoforms, can be differentiated on the basis of whether the repeat encoded by exon 10 is present or absent. Isoforms with four-repeat sequences are designated as 4-repeat-tau (4R-tau) while those without this sequence are called 3-repeat tau (3R-tau) 100. In PiD 30 and some forms of FTDP-17 MAPT 102 3R-tau is the predominant protein species in the tau filaments making up the inclusions, whereas 4R-tau assembles into filaments in CBD, PSP, GGT, AGD and some FTDP-17 MAPT variants 102-105. In a third group of diseases, which includes NFT-dementia and some FTDP-17 MAPT variants, both 3R-tau and 4R-tau are present in the inclusions (for a recent review see 102). The differences in the tau isoform composition of the tau filaments are also reflected by characteristic differences in the ultrastructural morphologies of the filaments 106.

For diagnosis post mortem microscopic investigation of the brain, including demonstration of the characteristic neuronal and if present glial inclusions, by tau, and if necessary, supplemented by 3R-tau and 4R-tau immunohistochemistry (Figure 4) together with documentation of the anatomical distribution of the tau lesions, is required. Microscopic studies may need to be supplemented with genetic and biochemical investigations in some cases, especially if a familial FTLD-tau variant is suspected. Although there is significant overlap between different forms of FTLD-tau, the presence of characteristic morphological features associated with the different tauopathies allows a specific diagnosis to be made in the majority of the cases. The presence of Pick bodies in PiD or other relevant neuronal and thread pathology in combination with characteristic glial inclusions, such as astrocytic plaques in CBD, tufted astrocytes in PSP, ramified astrocytes in PiD or numerous globular oligodendroglial and astrocytic inclusions in GGT are important helpers of the morphological diagnosis (Figure 4) (for a review see 107). The FTDP-17 MAPT variants may show similarities to PiD, PSP, CBD, AGD or GGT and variability within the same family is also known 58. For correlations between clinical presentation, genetics and pathology see Table 1.

FTLD-TDP

The discovery demonstrating the TDP-43 protein as the main component of the ubiquitin-positive inclusions in the majority of FTLD-U, designated as FTLD-TDP, and also in MND/ALS (ALS-TDP) resulted not only in a better understanding of the pathogenesis of these diseases, but this has also provided further support for a molecular classification of FTLDs 35. TDP-43 is a highly conserved and ubiquitously expressed multifunctional, heterogeneous nuclear ribonucleoprotein (hnRNP), encoded by the TARDBP gene located on chromosome 1. TDP-43 possesses nuclear localization and nuclear export signals and it shuttles between the cell nucleus and cytoplasm. TDP-43 also has an N-terminal domain, two RNA-recognition motifs (RRMs) involved in RNA and DNA binding and its glycine-rich C-terminal region contains most of the mutations causing familial MND/ALS and rarely familial FTD 108, 109. The C-terminal region of TDP-43 contains a prion-like protease resistant domain 110. TDP-43 protein expression is tightly controlled by autoregulatory mechanisms and both over and under expression results in impaired neuronal function 111. Cellular functions of TDP-43 include regulation of RNA splicing, translation, miRNA processes and mRNA transport and stability 112 with more than 6000 RNA targets. The aggregation of TDP-43 is associated with several post-translational modifications including phosphorylation of serine residues, ubiquitination, oxidation, lysine acetylation and C-terminal cleavage. Pathological examination, utilizing antibodies specific for phosphorylated TDP-43 (pTDP-43) confirms that the cellular TDP-43 aggregates contain phosphorylated epitopes 113. In disease TDP-43 aggregates can accumulate in both the cytoplasm and nuclei of affected neurons and glia, which results in cellular dysfunction. There is now evidence to suggest that TDP-43 in inclusions, the majority of which have been shown to show amyloid features 114, has cellular prion-like properties, which could have relevance for the pathomechanism of FTLD-TDP 115. The disease-associated TDP-43 making up the inclusions, is thought to exercise its deleterious effect via toxic gain of function due to overexpression or mutant forms, but given the number of its functions it is also likely that a loss of function effect also has a role in the pathogenesis of FTLD-TDP. FTLD-TDP has recently been reported to be associated with chronic traumatic encephalopathy 116. TDP43-positive inclusions are also found in around 90% of patients with hippocampal sclerosis, which is also a feature in the majority of FTLD-TDP suggesting a special relationship between these two pathologies 117, 118.

FTLD-TDP subtypes

Subtypes of FTLD-U were originally established, using findings of ubiquitin immunohistochemistry, and two major classification systems were in existence 69, 119. One of the strengths of these classifications is that they have been validated by clinical, genetic and pathological correlations. In order to eliminate competing classifications a single harmonized system was proposed in 2011, which has gained wide acceptance 120. Accordingly, the four different subtypes of FTLD-TDP, which are recognized, are based on the predominant lesion type(s) and the distribution of the pathological inclusions. The TDP-43 immunoreactive inclusion types considered include neuronal cytoplasmic inclusions (NCIs), neuronal intranuclear inclusions (NIIs), oligodendroglial inclusions and dystrophic neurites (DNs). Neurons containing TDP-43-positive inclusions lose their normal nuclear TDP-43 staining that can be observed in neurons without an inclusion, which in every day neuropathological practice is a helper of identifying affected neurons (Figure 5). FTLD-TDP type A is characterized by numerous NCIs, DNs and variable numbers of NIIs and the TDP-43 immunoreactive lesions are most numerous in layer 2 of affected cortices. FTLD-TDP type B is characterized by numerous NCIs in both the superficial and deeper cortical layers, but this subtype can exhibit the occasional DNs. FTLD-TDP type C is associated with abundant long DNs often with a corkscrew appearance throughout the cortical layers. Some TDP-43 type C cases are associated with corticospinal tract degeneration 121. FTLD-TDP type D is characterized by numerous NIIs, DNs and infrequent NCIs (Figure 5) 122, 123. It remains unclear whether differences in underlying pathophysiology or selective neuronal vulnerability determine the distinction between the FTLD-TDP subtypes via a single pathogenic mechanism or whether there are multiple mechanisms involved. For correlations between clinical presentation, genetics and pathology see Table 1.

FTLD-TDP in familial FTD

TDP-43 pathology is characteristic not only in sporadic, but also in some of the familial forms. In cases with mutations in the GRN gene FTLD-TDP type A is the characteristic morphology, while FTLD-TDP type D is restricted to cases with mutations in the VCP gene. The majority of the FTLD cases with C9orf72 repeat expansion mutations show FTLD-TDP type B pathology, but a proportion of them have changes consistent with type A and rare cases with type C pathology have also been reported 121, 124, 125. Families with pathogenic mutations in both the GRN and the C9orf72 genes have been documented 126, 127.

The characteristic aspect of the pathology of C9FTLD/ALS cases is the presence of star-like p62/sequestosome-1-positive, but TDP-43-negative NCIs in the hippocampus (granule cell layer and CA4 hippocampal subregion) and cerebellar cortex (cerebellar granule cells) (Figure 6) 44, 125, 128, 129, which are also positive for ubiquitin and the ubiquitin-binding protein ubiquilin-2. The high frequency of C9orf72 repeat expansion in both FTD and ALS has generated great interest in the underlying mechanisms in these diseases, of which several nonmutually exclusive possibilities exist (for a review, see 130). One of the mechanisms is RNA-related toxicity, based on evidence from microsatellite expansion diseases with large repeat expansions. In these disorders, the transcribed repeat RNA aggregates in the nucleus in discrete structures termed RNA foci, which sequester select RNA-binding proteins resulting in loss of their function, ultimately leading to disease. RNA foci have also been identified in C9FTD/ALS neurons 44 and, as the GGGGCC repeat expansions are bidirectionally transcribed, the RNA foci are formed of both sense and antisense transcripts, typically found in separate cells, but they can also co-localize in the same nucleus 131, 132. The possibility between high RNA foci burden and early age of disease onset has been raised 131, 133. Although the precise function of the C9orf72 protein is not known, it has been shown to be structurally related to a class of GDP/GTP exchange factors that activate Rab-GTPases, suggesting that it may have a role in vesicular trafficking 134. The possibility that toxicity is, at least partly due to a decrease/loss of the function of the C9orf72 protein is considered and such a hypothesis is supported by findings of decreased levels of GGGGCC repeat-containing transcripts in C9FTD brains 44, 135. A third mechanistic option is that both the sense and antisense transcripts of the GGGGCC expansion repeats are translated via the mechanism of RAN (repeat-associated non-ATG) translation into C9RANT (RAN-translated) protein aggregates (antisense: Pro-Arg, Pro-Ala, Pro-Gly; and sense: Gly-Ala, Gly-Arg, Gly-Pro), which form the dipeptide repeat proteins found in the p62-positive inclusions and thought to contribute to disease pathogenesis (Figure 6) 132, 136, 137. The topographical distribution of these dipeptide repeat proteins is similar regardless of the clinical phenotype 138, 139. RAN translation is ‘unconventional’ mode of translation with unknown mechanism, which occurs across expanded repeat tracts in the absence of an initiating codon 140. This mechanism has been shown in several microsatellite expansion disorders such as myotonic dystrophy, type 1 (DM1), spinocerebellar ataxia, type 8 (SCA8) and fragile X-associated tremor/ataxia syndrome (FXTAS) and data suggest that RAN translation may be a significant mechanism in C9FTD/ALS 130, 132, 141. Experimental data indicate that both arginine-rich proteins and repeat RNA contribute to C9orf72-mediated neurotoxicity 131.

FTLD-FUS

Neuropathological features of FTLD-FUS

The discovery that FUS mutations are the cause of a subgroup of familial MND/ALS gave the impetus for investigating the role of the FUS protein in sporadic, tau and TDP-43-negative, ubiquitin-positive FTLD-U cases. These studies showed that, indeed, the FUS protein is a major component of the pathological lesions in such cases comprising about 5–10% of all FTLD-U 70, 71, 73. This finding gave the basis for the introduction of a third smaller FTLD group, appropriately termed FTLD-FUS, which at present includes three conditions: NIFID 71, 156, aFTLD-U 70 and BIBD 73.

All three diseases classified under the umbrella term FTLD-FUS contain FUS-positive inclusions and show morphological similarities, but the differences in the inclusions types and their locations in the different brain regions is sufficient for a specific diagnosis to be made 157, 158. There are considerably more FUS-positive inclusions in NIFID compared with aFTLD-U in many anatomical regions including the cerebral cortex, medial temporal lobe structures, such as subiculum, entorhinal cortex and fusiform gyrus (but not in the granule cell layer of the dentate fascia), subcortical and brainstem nuclei 158. It is of note that a greater cell loss has been found in these anatomical regions in aFTLD-U, which may account for the lower numbers of inclusions in these areas. Hippocampal sclerosis has been described to be a prominent feature of aFTLD-U, but only occasionally seen in NIFID 72, 157-159. Different inclusion types have been documented in all three subtypes of FTLD-FUS 70, 71, 157, 158, 160. In NIFID, the morphology of the FUS immunoreactive NCIs is heterogeneous varying from small bean-shaped to larger annular shapes (Figure 7) 158, whereas NII were only occasionally found in NIFID cases. α-Internexin-positive inclusions are a prominent feature of all NIFID cases, although far less abundant than FUS-positive inclusions, whereas these are absent in both aFTLD-U and BIBD 157. The FUS-immunoreactive NCIs in aFTLD-U are often compact, round or bean-shaped and vermiform NII, described previously 159, and found throughout the neocortex, granule cells of the dentate gyrus and striatum 158. The third FTLD-FUS subtype, BIBD shows basophilic NCIs on the haematoxylin and eosin-stained histological sections of cerebral cortices and FUS immunohistochemistry highlights widespread NCIs seen not only in cerebral cortices, but also in the basal ganglia and brainstem 73, 157. In all three FTLD-FUS subtypes, the cerebellar cortex remain unaffected. FUS-positive neuronal intranuclear inclusions have also been described in polyQ inclusions in Huntington's disease, spinocerebellar ataxia types 1, 2, 3 and dentatorubro-pallidoluysian atrophy 161, 162.

The FUS-positive inclusion of NIFID, aFTLD-U and BIBD consistently contain the TATA-binding protein-associated factor 15 protein (TAF15) and variably contain the Ewing's sarcoma protein (EWS), which are other members of the FET family of proteins 163, 164. TRN1, responsible for the transport of FUS from the cytoplasm to the cell nucleus, has also been shown to be present in the NCIs and NIIs in all FTLD-FUS subgroups suggesting a role in the pathogenesis of these diseases 165. However, TAF15 and EWS and TRN1 are not present in the inclusions in familial MND/ALS with mutations of the FUS gene 163, 166. For correlations between clinical presentation, genetics and pathology see Table 1.

FTLD-UPS

Mutations of the CHMP2B gene, located on chromosome 3, are a rare cause of hereditary FTLD. The first mutation, affecting the splice acceptor site of the sixth (last) exon, was described in affected members of a Danish kindred with FTD-3 followed by the discovery of a Q165X mutation in a Belgian FTD family. Both mutations appear to have a common mechanism; deletion of the C-terminus of the CHMP2B protein (for a review, see 167). The CHMP2B protein is a component of the ESCRT-III complex (endosomal sorting complex required for transport-III), which has a role in protein degradation pathways.

Brains of affected individuals show marked frontotemporal atrophy, but the parietal lobe can also be affected. There is cortical microvacuolation of the neuropil accompanied by nerve cell loss and astrogliosis. Remaining cortical neurons show enlarged vacuoles, which are thought to represent aberrant large, late endosomes 167. Ubiquitin and p62-positive NCIs, which are tau, TDP-43 and FUS-negative, have been identified in the granule cells of the hippocampal dentate gyrus and in neurons of the frontal cortex 167, 168.

Conclusions and future directions

The last 10 years witnessed the discovery of two major FTD genes, the GRN and C9orf72 genes and proteins such as TDP-43 and FUS together with the recognition of the role of other FET proteins and TRN1 in the FUS inclusions in FTLD-FUS. Recognition of C9orf72 as a major gene in both FTD and MND/ALS trigged major research into the disease mechanisms of C9FTD-MND/ALS, which has provided significant results in a relatively short period of time. This remarkable increase in knowledge has resulted in a better understanding of the pathogenesis of several FTLD subgroups, firmly established a molecular link between FTD and MND/ALS, and also facilitated the introduction of a molecular classification, which has been widely accepted and followed in everyday diagnostic practice of neuropathologists. As in other neurodegenerative diseases, the concept of cell-to-cell propagation of disease-associated proteins underlying disease spread and progression has been studied in an FTLD-TDP subgroup with bvFTD, although this is still awaited in all forms of FTLD-TDP, tauopathies such as PSP and CBD, and in the different disease entities of FTLD-FUS.

Despite the advances several fundamentally important questions related to pathogenesis remain unanswered and we only enlist here a few. Is there a common pathogenic mechanism determining selective neuronal vulnerability and linking all forms of FTLDs or, as it seems more likely, multiple mechanisms exist? What are the downstream mechanisms driving the TDP-43 pathology to different cell types and/or different neuronal compartments in the four pathological subtypes of FTLD-TDP? Detailed pathological investigation of larger cohorts with c9orf72 repeat expansions is also required. The landmark findings of the past 20 years together with future discoveries, one may trust, will facilitate the translation of this knowledge into disease biomarkers, allowing precise clinical diagnosis and ultimately leading to the establishment of effective disease modifying therapies.

Author contributions

TL, JDR, SM and TR undertook a literature review and drafted the initial paper and prepared the figures. All authors were involved in editing the paper. TR did the final editing.