Distinct Neuroanatomical Correlates of Neuropsychiatric Symptoms in the Three Main Forms of Genetic Frontotemporal Dementia in the GENFI Cohort

INTRODUCTION

Frontotemporal dementia (FTD) is the second most common cause of young-onset dementia. It is a highly heritable disease with an autosomal dominant mode of inheritance being found in around a third of cases [1, 2]. Mutations in three main genes have been identified in familial FTD: microtubule associated protein tau (MAPT), progranulin (GRN), and chromosome 9 open reading frame 72 (C9orf72) [3]. Among neurodegenerative dementias, FTD often mimics primary psychiatric conditions. Indeed, patients with FTD are more likely to receive a prior inaccurate psychiatric diagnosis [4].

The most common subtype, namely behavioral variant FTD (bvFTD), is characterized by apathy, loss of empathy, perseverative and compulsive behavior, disinhibition, and hyperorality [5]. These symptoms often lead to bvFTD being misdiagnosed with either late onset depression or obsessive-compulsive and impulse-control disorders. Other neuropsychiatric symptoms (NPS) such as delusions or hallucinations are frequently present in FTD with a third of patients having psychotic symptoms in neuropathologically confirmed cases [6]. Although FTD may mimic late-onset schizophrenia or bipolar spectrum disorders in some cases, psychosis and mood disorders are not considered clinical hallmarks of bvFTD [7] and are not included within current diagnostic criteria.

Recent genetic advances in the understanding of familial forms of FTD have rekindled interest in the borderlands between FTD and psychiatric disorders, by revealing a high frequency of NPS in genetic FTD, particularly among C9orf72 and GRN mutation carriers [8–10]. NPS sometimes emerge many years prior to dementia onset and appear to have different phenomenology across FTD-related mutations [10, 11]. Cross sectional results from the Genetic FTD Initiative (GENFI) have previously shown evidence for neuroanatomical changes in presymptomatic FTD mutation carriers [12, 13]. The insula has been shown to be the earliest affected region in genetic FTD, a key hub in the salience network which has been considered as an influential model in several psychiatric disorders such as schizophrenia [14]. To our knowledge, no study has been carried out to date, which assesses the correlation between NPS and neurostructural changes across the three main genes associated with FTD. Characterizing the neural correlates across familial FTD would provide important insight into the distinctive structures involved in the neuropsychiatric phenotype in FTD and therefore enable further understanding of the pathophysiological substrate and neurobiological basis of NPS. The pattern of brain atrophy could ultimately guide targeted genetic testing particularly in cases of atypical and overlapping FTD/neuropsychiatric presentations. In this study, we applied voxel-based morphometry (VBM) to a large international cohort of presymptomatic and symptomatic carriers of FTD-related mutations. We hypothesized that neuroanatomical correlates of NPS would be distinct between the different forms of genetic FTD and would bring evidence of overlapping brain structural changes with primary psychiatric disorders.

MATERIALS AND METHODS

RESULTS

Among the 167 mutation carriers, 45% had a mutation in the GRN gene, 36% in the C9orf72, and 19% in the MAPT. Participants consisted of 122 presymptomatic (63 GRN, 37 C9orf72, and 22 MAPT) and 45 symptomatic (12 GRN, 23 C9orf72, and 10 MAPT) carriers (see Table 1). Individuals with MAPT mutations were significantly younger (mean age 44.3 years versus 52.1 and 51.3 years in C9orf72 and GRN, respectively; p = 0.013). According to the Frontotemporal Dementia Rating Scale (FRS) [19], in the symptomatic group: three participants (1 GRN, 1 C9orf72, and 1 MAPT) were mildly impaired, three (1 GRN, 1 C9orf72, and 1 MAPT) were moderately impaired, 18 (3 GRN, 12 C9orf72, and 3 MAPT) were severely impaired, 15 (5 GRN, 7 C9orf72, and 3 MAPT) were very severely impaired and 5 (2 GRN, 1 C9orf72, and 2 MAPT) were profoundly impaired; FRS was not available for one patient. There was no significant correlation between mean NPS score and disease severity in symptomatic carriers (rS = 0.21, p = 0.15). There were no significant differences between the three genetic groups in terms of severity scores nor frequency of hallucinations (visual, auditory and tactile), delusions, depression and anxiety (see Table 2). Distribution of NPS among presymptomatic and symptomatic carriers is displayed in Supplementary Table 1.

Statistical parametric maps showing correlations between each NPS score and GM density in the three genetic groups are reported in Figs. 1 to 6. For visual hallucinations (Fig. 1), in the GRN group, we found significant negative correlations between the clinical score and GM in a large network of areas including mainly the parietal lobe (left superior parietal lobule and left angular gyrus), inferior occipital gyrus, left thalamus and left middle frontal gyrus (these clusters remained significant after FWE correction for multiple comparisons p < 0.05), with further areas in the cerebellum and temporal lobe (Supplementary Table 2). In the C9orf72 group, we identified a significant negative correlation with GM in the parietal lobe (left superior parietal lobule and left angular gyrus) and left occipital pole (Supplementary Table 3). In the MAPT group, visual hallucinations score correlated with GM in the precentral gyrus, left lingual gyrus, right fusiform gyrus, and left angular gyrus (Supplementary Table 4). In the GRN group, auditory hallucinations (Fig. 2) were associated with the left anterior insula, left thalamus, frontal lobe, temporal lobe, and the cerebellum when correcting for multiple comparisons (Supplementary Table 5). In the C9orf72 group, a significant negative correlation was found in the left middle frontal gyrus (Supplementary Table 6). In the MAPT group, the auditory hallucinations score was related to GM atrophy in the right angular gyrus (Supplementary Table 7). Areas related to tactile hallucinations (Fig. 3) in the GRN group (Supplementary Table 8) were similar to those seen in auditory hallucinations. In the C9orf72 group, we found a significant negative correlation in the inferior frontal gyrus and posterior cingulate gyrus (Supplementary Table 9). Significant areas were seen in the MAPT group within the frontal lobe and left lingual gyrus (Supplementary Table 10).

Fig.1

VBM analysis showing areas of significant correlation between the presence and severity of visual hallucinations and GM density across the FTD genetic groups. Statistical parametric maps were thresholded at p < 0.001 uncorrected and rendered on a study-specific T1-weighted MRI template in MNI space. Analyses were adjusted for age, gender, total intracranial volume, and scanner type. The color bar indicates the Z-scores.

Fig.2

VBM analysis showing areas of significant correlation between the presence and severity of auditory hallucinations and GM density across the FTD genetic groups. Statistical parametric maps were thresholded at p < 0.001 uncorrected and rendered on a study-specific T1-weighted MRI template in MNI space. Analyses were adjusted for age, gender, total intracranial volume, and scanner type. The color bar indicates the Z-scores.

Fig.3

VBM analysis showing areas of significant correlation between the presence and severity of tactile hallucinations and GM density across the FTD genetic groups. Statistical parametric maps were thresholded at p < 0.001 uncorrected and rendered on a study-specific T1-weighted MRI template in MNI space. Analyses were adjusted for age, gender, total intracranial volume, and scanner type. The color bar indicates the Z-scores.

Fig.4

VBM analysis showing areas of significant correlation between the presence and severity of delusions and GM density across the FTD genetic groups. Statistical parametric maps were thresholded at p < 0.001 uncorrected and rendered on a study-specific T1-weighted MRI template in MNI space. Analyses were adjusted for age, gender, total intracranial volume, and scanner type. The color bar indicates the Z-scores.

Fig.5

VBM analysis showing areas of significant correlation between the presence and severity of depression and GM density across the FTD genetic groups. Statistical parametric maps were thresholded at p < 0.001 uncorrected and rendered on a study-specific T1-weighted MRI template in MNI space. Analyses were adjusted for age, gender, total intracranial volume, and scanner type. The color bar indicates the Z-scores.

Fig.6

VBM analysis showing areas of significant correlation between the presence and severity of anxiety and GM density across the FTD genetic groups. Statistical parametric maps were thresholded at p < 0.001 uncorrected and rendered on a study-specific T1-weighted MRI template in MNI space. Analyses were adjusted for age, gender, total intracranial volume, and scanner type. The color bar indicates the Z-scores.

As displayed in Fig. 4, GRN mutation carriers showed a significant association between the delusions score and GM atrophy mainly in the left thalamus and left anterior insula when correcting for multiple comparisons (Supplementary Table 11). In the C9orf72 group, we found a significant correlation between delusions and GM in the left middle and superior frontal gyri (Supplementary Table 12). In the MAPT group, the most significant areas for delusions were in the temporal lobe (left middle temporal gyrus and left hippocampus), frontal lobe, parietal lobe, left subcallosal area, left anterior insula, and fusiform gyrus when correcting for multiple comparisons (Supplementary Table 13).

For the depression score (Fig. 5), we found significant correlations within the GRN group, with GM in the frontal lobe, right subcallosal area, bilateral anterior insula, left middle cingulate, right cerebellum, temporal lobe, and right middle occipital gyrus (these clusters survived after FDR correction) (Supplementary Table 14). In C9orf72 expansion carriers, the depression score was correlated with GM atrophy in the parietal lobe (left precuneus and superior parietal lobule) and left superior frontal gyrus (Supplementary Table 15). In the MAPT group, we identified significant negative correlations with GM in the left postcentral gyrus, temporal lobe, right fusiform gyrus, right lingual gyrus, superior motor cortex, and left cerebellum (Supplementary Table 16).

Finally, for the anxiety score (Fig. 6), GRN mutation carriers showed significant correlations with GM atrophy in areas including the frontal lobe, left anterior insula, left precuneus, left posterior cingulate, temporal lobe, left caudate, left superior occipital gyrus, and bilateral cerebellum when correcting for multiple comparisons (Supplementary Table 17). In the C9orf72 group, we found significant correlations in the inferior frontal gyrus, left cerebellum, left fusiform gyrus and right superior temporal gyrus (Supplementary Table 18). In the MAPT group, the main significant areas were in the temporo-parietal and frontal lobes as well as bilateral cuneus (Supplementary Table 19).

Table 3 summarizes the regional grey matter correlates of NPS from the VBM analysis.

DISCUSSION

Our findings highlight the involvement of widespread cortical and subcortical brain structures, as part of large-scale networks, in the occurrence of NPS in FTD mutation carriers. These results suggest a strong association of psychotic symptoms with GM atrophy in the anterior insula in GRN mutation carriers and, to a lesser extent, in MAPT carriers. Moreover, the GRN group showed particular involvement of the left thalamus, left middle frontal gyrus, and cerebellum in the emergence of psychosis. In the C9orf72 group, psychotic symptoms were mainly associated with changes in the frontal cortex, whereas the MAPT group involved the temporal lobe and angular gyri. The presence of mood disorders was mostly associated with GM atrophy in frontal-insular cortex, in addition to precuneus and posterior cingulate, in the GRN group. Parietal and frontal cortices as well as cerebellum alterations correlated with mood disorders in C9orf72 carriers, whereas in MAPT carriers, depression and anxiety were mainly associated with the temporo-parietal cortex.

The involvement of the insula in the emergence of psychotic symptoms in GRN mutations carriers supports the aberrant salience hypothesis previously reported in mental disorders [20, 21]. The insula is critical for the regulation of a number of affective and cognitive functions, including body-self consciousness and subjective emotional experience. It is thought to integrate multisensory inputs and coordinate the allocation of the brain’s attentional resources to the most relevant stimuli [22]. The anterior part of the insula, which is especially correlated to NPS in our study, was shown to be particularly specialized for interoception, subjective feelings, and emotional awareness [23]. Psychotic symptoms including delusions and all sensory modalities of hallucinations were also significantly associated with greater GM atrophy in the left thalamus within the GRN group. The disturbances of thalamo-cortical networks were suggested to contribute to the emergence of psychotic symptoms in schizophrenia [24].

The involvement of large-scale networks in the emergence of NPS in FTD is also reflected by the significant association of NPS with cerebellar alteration in the GRN group. A growing body of evidence arising from functional magnetic resonance imaging (fMRI) implicates cortico-cerebellar connectivity loops in the pathophysiology of primary psychiatric conditions [25]. Schmahmann et al. considered the psychosis spectrum as an integral part of the cerebellar cognitive affective syndrome and linked it to damage of the fastigial nucleus and the vermis, defined as “the limbic cerebellum” [26, 27]. Our results showed evidence of frontal involvement in delusions among C9orf72 expansion carriers which is inconsistent with a previous study suggesting an association between delusions and GM atrophy in the precuneus [28]. Limited data are available on the specific content of delusional thoughts in C9orf72 carriers. Authors have reported odd somatic complaints, somatoform delusions and body image distortion [11, 29]. This observation is supported by the evidence of altered body scheme perception and processing among patients with C9orf72 expansion [30]. Interestingly, the superior and middle frontal gyri along with inferior frontal and posterior cingulate gyri, which correlated respectively with delusions and tactile hallucinations in our C9orf72 group, have previously been considered as part of the neural correlates of somatoform disorders [31]. The MAPT group was mainly distinguished by the involvement of the left temporal lobe in delusional symptoms, in line with previous reports suggesting a predominant temporal alteration in MAPT mutation carriers [12, 32]. This is also consistent with previous studies that reported an association between GM atrophy in temporo-limbic areas and vulnerability to psychosis [33].

As expected, visual hallucinations were mainly characterized by the involvement of the visual cortex including the occipito-parietal and occipito-temporal regions across the three genetic mutations, consistent with previous studies which demonstrated structural and functional disconnectivity within visual processing areas in hallucinatory experiences [34]. The correlation of visual hallucinations with GM in the left angular gyrus was shared by all genetic groups. The left angular gyrus alteration has previously been considered as a structural substrate in schizophrenia [35], but its role in the emergence of visual hallucinations remains unexplored. Based on our results, we might speculate that the angular gyrus, as part of the heteromodal parietal association cortex, plays presumably an important role in the psychopathology of visual hallucinations in genetic FTD. The GRN group was characterized by a significant correlation between thalamic atrophy, particularly posteriorly (including the pulvinar) and visual hallucinations. Vulnerability to visual hallucinations as a result of thalamic dysfunction has been reported in neurodegenerative diseases such as Lewy body dementia (LBD) [36]. Interestingly visual hallucinations have previously been reported as an important distinctive feature in GRN carriers which could sometimes even present with a LBD-like phenotype [10]. Atrophy of the frontal lobe has also been found to correlate with visual hallucinations in both GRN and MAPT groups. Several studies have shown visual hallucinations to be linked to frontal structures, as reflected in previous correlational analyses performed in LBD and Parkinson’s disease [37].

Mood disorders in the GRN group were mainly related to a fronto-insular-cerebellar alteration. We also noted particular involvement of the posterior portions of the default-mode network (DMN) (precuneus and posterior cingulate cortex) in anxiety and of left hippocampus in both anxiety and depression. Left precuneus atrophy has also been found to be correlated to depression in C9orf72 expansion carriers. This study supports previous findings of the role of DMN alterations in depression and anxiety disorders [38–40]. The MAPT group was characterized by the involvement of right temporo-occipital areas in depression and left temporo-parietal regions in anxiety. This is in line with the prominence of mood disorders in neurodegenerative disease presenting with a major alteration of posterior cortical areas as illustrated by posterior cortical atrophy [41, 42]. Disparate findings for depression could be reconciled through its conceptualization as a constellation of symptoms related to several psychopathological mechanisms (self-related mental representation, referential judgments, negative self-focused thoughts…), which can be underpinned by different structural and neurobiological dysfunctions [43, 44]. Though the structural models underlying the psychopathology of depression can be distinct, they may involve overlapping functional networks and neurobiological systems including serotoninergic, noradrenergic and dopaminergic tracts.

Anxiety was previously reported as a prominent feature in C9orf72 expansion carriers [45]. It is interesting to note that, despite a recognized pattern of cerebellar involvement in the C9orf72 expansion [46], its atrophy has proved to be only related to anxiety in our study, but not to psychosis as previously reported [47]. Variability in findings between studies may be due to several factors, including sample size, different neuroimaging techniques, and medication. Overall, we found the most significant results in the GRN group.

The lack of specific identification of each symptom’s nature through the use of the NPS scale instead of a full psychiatric questionnaire is a limitation. Psychotic symptoms are a multidimensional phenomenon, which is difficult to deconstruct through a clinical scale, therefore caution should be taken in interpreting the results. For instance, odd somatic complaints in C9orf72 expansion carriers may be mistaken for anxiety disorders. Further studies should be performed based on detailed psychiatric assessment, including expanded items of valid psychiatric rating scales, to improve the detection of subtle NPS and better refine their phenomenology. Connectivity-based neuroanatomical methods will be required in order to provide a better understanding of the underlying structural substrates and networks dynamics of psychotic symptoms.

Despite these limitations, to our knowledge, this is the first study to investigate the neuroanatomical correlates of a wide range of NPS in the three main mutations responsible for genetic FTD. Moreover, our findings, taking into account the presymptomatic phase of the disease, support that NPS may be considered as prodromal hallmarks of early neurodegeneration, within the scope of mild behavioral impairment (MBI). The concept of MBI was developed as a potential pre-dementia stage. New diagnostic criteria and a standardized scale to measure MBI have recently emerged [48, 49]. Affective dysregulation and abnormal perception/thought content were stated as MBI subcategories. A previous study showed that MBI patients had a high conversion rate to dementia (71.5%), mainly evolvement to FTD [50]. Therefore, the study of NPS offers the opportunity of early diagnosis and could allow measurement of future disease-modifying preventive strategies in at-risk individuals.