Neuroimaging features of C9ORF72 expansion

Hexanucleotide expansion intronic to chromosome 9 open reading frame 72 (C9ORF72) has recently been identified as the most common genetic cause of both familial and sporadic amyotrophic lateral sclerosis and of frontotemporal dementia with or without concomitant motor neuron disease. Given the common frequency of this genetic aberration, clinicians seek to identify neuroimaging hallmarks characteristic of C9ORF72-associated disease, both to provide a better understanding of the underlying degenerative patterns associated with this mutation and to enable better identification of patients for genetic screening and diagnosis. A survey of the literature describing C9ORF72 neuroimaging thus far suggests that patients with this mutation may demonstrate symmetric frontal and temporal lobe, insular, and posterior cortical atrophy, although temporal involvement may be less than that seen in other mutations. Some studies have also suggested cerebellar and thalamic involvement in C9ORF72-associated disease. Diffuse cortical atrophy that includes anterior as well as posterior structures and subcortical involvement thus may represent unique features of C9ORF72.


C9ORF72 neuroimaging features
Most neuroimaging studies on C9ORF72 thus far have examined atrophy patterns by using T1-weighted magnetic resonance imaging (MRI) in symptomatic patients with the behavioral variant of FTD (bvFTD) or in combined cohorts of all mutation carriers representing clinical diagnoses of bvFTD, ALS, FTD-MND, and/or PPA. In these contexts, C9ORF72 expansion has been asso ciated primarily with relatively symmetrical (bilateral) atrophy most prominent in the frontal and temporal lobes and the insula (Table 1), which are all regions previously implicated in FTD. Whereas one group found a predominance of temporal atrophy [9], other studies have found less temporal lobe involvement in C9ORF72 mutation carriers (C9 + ) as compared with patients with other FTD mutations [11] and sporadic disease [12,13]. Th e group with predominant temporal atrophy was composed of a signifi cantly larger proportion of PPAdiagnosed patients than other studies, which may explain this diff erence in fi ndings and highlights the heterogeneity in C9 + -associated diagnoses. C9ORF72 mutation carriers may also harbor several subtle neuroimaging features that are uncommon in sporadic bvFTD and that distinguish it from other muta tions (described in detail in the following section). Most notably, C9 + patients appear to have more parietal and occipital cortical involvement [8,9,[11][12][13][14], so that cortical atrophy often appears quite diff use (Table 1). Th is diff use atrophy pattern is much less common in other genetic and sporadic forms of FTD and may be a hallmark of C9ORF72 expansion. Interestingly, some C9 + patients have been described as clinically aff ected yet demonstrate no visually detectable brain

Abstract
Hexanucleotide expansion intronic to chromosome 9 open reading frame 72 (C9ORF72) has recently been identifi ed as the most common genetic cause of both familial and sporadic amyotrophic lateral sclerosis and of frontotemporal dementia with or without concomitant motor neuron disease. Given the common frequency of this genetic aberration, clinicians seek to identify neuroimaging hallmarks characteristic of C9ORF72-associated disease, both to provide a better understanding of the underlying degenerative patterns associated with this mutation and to enable better identifi cation of patients for genetic screening and diagnosis. A survey of the literature describing C9ORF72 neuroimaging thus far suggests that patients with this mutation may demonstrate symmetric frontal and temporal lobe, insular, and posterior cortical atrophy, although temporal involvement may be less than that seen in other mutations. Some studies have also suggested cerebellar and thalamic involvement in C9ORF72associated disease. Diff use cortical atrophy that includes anterior as well as posterior structures and subcortical involvement thus may represent unique features of C9ORF72.
atrophy [8,9,14]; a recent report described two such patients with slowly pro gressive bvFTD (bvFTD-SP) charac terized by a long disease course and nonprogressive brain atrophy, and both of them were C9 + [14].
In addition to fi nding diff use cortical atrophy, some groups have found involvement of the cerebellum [9- 11,13] or bilateral thalamus [10,13,14] (or both) in C9 + , which may further distinguish C9 + neuroimaging patterns from those of other mutations and suggests that subcortical changes may contribute to symptoms [10,13]. Cerebellar fi ndings have been further substantiated pathologically by the presence of ubiquitin/p62-positive, TDP-43-negative neuronal cytoplasmic inclusions in this area in subsets of C9 + FTD/ALS cases included in the aforementioned neuroimaging studies [8-10,12] as well as Whitwell et al. [11] (2012) n = 19 C9 + patients showed diff use atrophy involving the frontal, temporal, parietal, and occipital lobes.
C9 + patients showed more atrophy in parietal, occipital, lateral, frontal, and cerebellar regions compared with sporadic FTD. Sporadic FTD showed more atrophy in the medial frontal lobe.
C9 + patients showed more atrophy in parietal, occipital, lateral, frontal, and cerebellar regions compared with MAPT. MAPT showed more atrophy in the anterior temporal regions.
C9 + patients without MND showed more parietal and thalamic atrophy than sporadic bvFTD. Sporadic FTD showed more medial frontal atrophy. C9 + patients with MND showed more dorsal frontal, posterior, and cerebellar atrophy compared with sporadic FTD-MND. others [15][16][17][18] and may be a unique pathologic fi nding associated with C9ORF72 expansion [15]. Little assessment of C9ORF72-associated features has yet been performed using neuroimaging modalities beyond T1-weighted MRI. Findings from multimodal imaging, such as positron emission tomography (PET) or single-photon emission computed tomography (SPECT), are most often concordant with structural imaging fi ndings but can sometimes reveal additional information not readily apparent from MRI alone. In C9 + patients, in whom marked heterogeneity in presentation may complicate diagnosis, complementary fi ndings from other imaging modalities may be particularly informative for confi rming clinical and structural imaging fi ndings. For example, Floris and colleagues [19] presented a case report of a patient with C9ORF72 expansion associated with FTD-parkinsonism-upper motor neuron disease. In addition to presenting with prominent behavioral disturbances consistent with bvFTD, the patient presented with marked visuospatial impairment and hallucinations, which, along with parkinsonism, suggested a diff erential diagnosis of dementia with Lewy bodies. Th is patient demonstrated bilateral frontotemporal and posterior atrophy on structural MRI. Perfusion SPECT demonstrated reduced uptake bilaterally in both frontotemporal and parietal regions, consistent with structural imaging fi ndings and the behavioral and visuospatial impairments. Genetic screening confi rmed pathogenic expansion of C9ORF72 and a diagnosis of bvFTD [19]. Similarly, Khan and colleagues [14] described one C9 + bvFTD-SP patient who had no atrophy upon visual inspection of MRI and had a consistent FDG-PET scan showing results within normal limits. Th ese fi ndings, though concordant across imaging modalities, were interesting given the patient's clinical presentation of bvFTD and suggest disruption of behavior in the context of limited visible changes on both structural and metabolic imaging.
Even within a survey of six more 'typical' C9 + patients, Boeve and colleagues [8] described a variety of fi ndings from SPECT and PET imaging. In fi ve carriers, SPECT demonstrated signifi cant hypoperfusion in the anterior and middle cingulate gyri in comparison with controls. In one of these patients, there was no visual atrophy in structural MRI at the time of SPECT imaging, and frontal cortical atrophy was more evident only on structural MRI two years later [8], suggesting that early changes in hypometabolism preceded cortical atrophy. In four of fi ve patients who underwent FDG-PET imaging, anterior cingulate demonstrated signifi cant hypometabolism; posterior cingulate metabolism was normal in all four patients, and frontal cortical hypo metabolism ranged from mild to severe [8]. Interestingly, the fi fth patient demonstrated hypometabolism in parietal/precuneus regions with relative sparing of frontal cortical regions, a pattern more consistent with Alzheimer's disease than FTD/ALS. Th ese fi ndings, though not directly comparable, indicate that multimodal imaging sometimes may reveal abnor malities not detectable with structural MRI alone. Th is is particularly true when characterizing atypical clinical presentations and also establishes a breadth of imaging variability across C9 + patients. Th is high level of hetero geneity across individuals refl ects the multitude of clinical phenotypes associated with C9 + and may distin guish C9ORF72 expansion from other FTDassociated gene mutations, which are described in the next section.

C9ORF72 versus GRN and MAPT neuroimaging characteristics
Previous studies have assessed the neuroimaging characteristics unique to other FTD-spectrum gene mutations, most notably in granulin (GRN), which encodes the protein progranulin and results in TDP-43 pathology, and in MAPT, the gene coding for the tau protein, which characterizes the other major type of FTD pathology. MAPT mutation carriers are generally charac terized by relatively symmetrical atrophy predominantly in the anterior and medial temporal lobes, orbitofrontal cortex, and fornix [7,20]. Whitwell and colleagues [20] also found cerebellar involvement in MAPT mutation carriers. In contrast, GRN mutations are often associated with strongly asymmetric atrophy, aff ecting either hemisphere and involving the inferior frontal, temporal, and parietal lobes, with additional white matter involvement [7,20]. Examples of typical atrophy patterns in MAPT, GRN, and C9ORF72 mutation carriers are shown in Figure 1.
Whitwell and colleagues [11] performed modeling to identify neuroimaging predictors of C9ORF72 expansion as compared with GRN and MAPT and found that smaller left sensorimotor cortices, right occipital lobe and left cerebellum, and larger left inferior temporal lobe all independently contributed to a prediction of C9 + status. Using 14 of 39 total regions of interest, they could correctly classify C9ORF72 mutation carriers with 74% accuracy and achieved 93% classifi cation accuracy with 26 variables [11]. Although neuroimaging hallmarks are more likely to be used in conjunction with, rather than in place of, molecular genetic and family history information [21], the ability to accurately identify C9 + patients from other FTD mutation carriers further supports the hypothesis that diff erent genetic lesions result in diff erent patterns of brain atrophy. Identifi cation of gene-specifi c neuroimaging hallmarks may provide insight into the underlying pattern and type of pathology, which could be important information in the advent of pathology type-specifi c therapeutic interventions or for use as a biomarker in clinical trials. Previous studies have suggested that FTLD-TDP-43 type may be consistent with specifi c patterns of structural atrophy [22,23]. Both TDP-43 (harmonized [24]) type A and B pathologies have been associated with C9 + FTD/ALS [8-10, [16][17][18] and are broadly consistent with neuroimaging hallmarks for each: type A (Mackenzie type 1, Sampathu type 3) was associated with more dorsal frontotemporal, inferior parietal, striatal, and thalamic atrophy, and type B (Mackenzie type 3, Sampathu type 2) was asso ciated with relatively symmetrical posterior frontal, medial temporal, prefrontal, orbitofrontal, and insular cortex atrophy [22,23]. Th e association of neuroimaging fi ndings and pathology in less common clinical presenta tions of C9 + cases, however, remains to be determined.
Th e role of genotype in changes in brain structure over time provides another insight into early hallmarks of disease and its underlying phenomenology. Although the sample size was small, Mahoney and colleagues [10] found that annualized rates of brain atrophy were greatest in GRN (n = 4) carriers, followed by carriers of C9ORF72 (n = 5) and MAPT (n = 6), although mean atrophy rates did not diff er signifi cantly between groups [25]. Previously, Whitwell and colleagues [26] demonstrated similar fi ndings whereby GRN muta tion carriers had higher annual rates of whole-brain atrophy in comparison with MAPT mutation carriers. Boeve and colleagues [8] reported on at least 2 years of longitudinal follow-up in eight C9 + patients, who showed progression of atrophy specifi cally in frontal lobes and ventricular enlargement. Previously, GRN mutations were shown to exhibit asymmetric volume loss mainly in inferior frontal, superior temporal, and inferior parietal lobes, precuneus, and cingulate cortex over time [25]. In MAPT mutation carriers, longitudinal volume loss is symmetric and involves anteromedial temporal lobes, orbitofrontal cortex, and white matter tracts, including corpus callosum [25].
Of note, Khan and colleagues [14] described two patients with bvFTD-SP who demonstrated no signifi cant brain atrophy over the course of 3 years and 8 years, respectively; the seeming lack of progressive atrophy contributed some uncertainty to the patients' initial bvFTD diagnosis. A small minority of patients with mutation in C9ORF72 have been described with similar absences of visible brain atrophy in MRI [8,9], and this stands in stark contrast to the often insidious progression of volume loss displayed with the other main gene mutation causing TDP-43 patho logy, GRN. Th e possibility that some patients harbor ing C9ORF72 repeat expansion may not show progressive brain atrophy thus seems quite unusual in comparison with other mutations, and it will be impor tant to investigate potential factors that modify rates of progression, including the number of hexanucleotide repeats, as the technology becomes available.
Broadly, MAPT atrophy is symmetric and more ventral whereas GRN atrophy is asymmetric and more dorsal [25], and this diff erence may refl ect degeneration in diff erent functional networks that may be selectively vulnerable to FTLD [27]. Although initial lesions (genetic, developmental, environmental, and so on) may diff erentially determine which brain region is fi rst aff ected, studies in FTD mutation carriers as well as in other diseases strongly suggest that once a degenera tive process is set in motion, that process will continue in a circumscribed pattern [25] that may be determined by connections refl ecting intrinsic organization of func tion al brain systems [27]. Zhou and colleagues [28] suggest that neurodegenerative processes may begin within a single 'epicenter' and spread through a specifi c network of functional paths. In the case of mutation carriers, genetically mediated vulnerability may enhance this spread. Two key questions thus remain: how does C9ORF72 expan sion alter vulnerability, and upon what path does it tread?
Findings from published neuroimaging studies thus far suggest that subcortical structures, including the thalamus and cerebellum, may be uniquely aff ected by mutation in C9ORF72. As reviewed by Schmahmann and colleagues [29], these regions are both interconnected with the entire cerebral cortex. If these subcortical regions serve as the epicenter for C9ORF72 pathology, then their diff use cortical connectivity may be a clue to the mechanisms leading to the diff use cortical involvement that seems characteristic of this mutation.
Subcortical involvement may also explain the development of symptoms even in patients who have relatively little cortical atrophy. In a meticulous review, Schmahmann and Pandya [30] describe disconnection syndromes that can occur in the context of specifi c subcortical involvement of the basal ganglia, thalamus, and cerebellum. Th e authors propose that neural architecture determines function, that specifi c connections between subcortical nodes defi ne behavior, and that fi ber tracts linking cerebral cortical regions to one another enable the coordination necessary for complex behaviors [30]. In this framework, changes in the connectivity of these subcortical structures with the cortex could account for some of the behavioral, executive function, and motor symptoms that have recently been associated with C9ORF72 FTD/ALS, even in patients who show relatively little cortical atrophy. Th is hypothesis will have to be addressed in future studies integrating structural and functional imaging methodology and linking them to symptoms [13,31].

Conclusions
Mutation carriers with FTD demonstrate prominent sym metric atrophy in frontal and temporal lobes and insula. Main hallmarks specifi c to C9ORF72 may include relatively diff use changes that involve posterior as well as anterior cortical regions and bilateral thalamic and cere bellar atrophy. Involvement of temporal lobes does not appear to be as prominent in C9 + patients with bvFTD but is still seen in those with PPA. Occipital and cere bellar atrophy and relative sparing of temporal lobes may be distinguishing features of C9ORF72 relative to GRN-or MAPT-associated disease. Finally, in some expansion carriers, atrophy may not be apparent upon visual inspec tion of MRI despite clear clinical symptoms, and this may refl ect the eff ects of disease on subcortical structures.
Although signifi cant eff orts have been undertaken to characterize C9ORF72 expansion carriers, much of this work has been performed on incomplete archival data or samples of convenience, resulting in diverse datasets and study designs that can be diffi cult to compare directly, particularly in small cohorts. Future investigations of neuroimaging characteristics in C9ORF72 expansion carriers will require careful clinical characterization and study designs that are properly controlled with respect to diagnosis and image ascertainment. Comparisons of C9 + FTD with sporadic disease may further complement work comparing C9ORF72 expansion with other FTDcausing mutations. Furthermore, very few studies have examined the imaging features of C9ORF72 in modalities other than T1-weighted structural MRI. While neuroimaging for diagnostic assessment of mutation carriers will most likely complement-rather than replace-molecular genetic characterization, these signatures may serve a signifi cant role in earlier identifi cation and diagnosis of patients with mild behavioral syndromes or family histories of unknown etiology or both. In addition, neuroimaging may serve a critical role in the prediction of symptoms and in the assessment of drug eff ects during treatment trials, particularly during preclinical stages of disease. In the future, functional assessment of connectivity and subcortical network mapping may shed light on the mechanistic underpinnings of C9ORF72 pathogenicity, particularly in the context of visually undetectable brain atrophy, and may complement structural imaging in diagnosis and longitudinal assessment.