The aim of the present work was to improve our understanding of the structural and functional basis of the constellation of behavioural symptoms of FTD, by studying them with data-driven approaches, and relating them to two different imaging modalities (structural MRI and FDG-PET) in combination and non-linearly.
We found that the variability of behavioural and psychological symptoms in an FTD cohort was best captured by three components, which we labelled as (i) Apathy, (ii) Disinhibition versus depression/mutism, and (iii) Psychosis.
The fact that ratings for apathy and disinhibition from behavioural questionnaires (such as the FrSBe and the NPI) loaded on two different components contributes to the ongoing discussion on whether apathy and impulsivity represent opposite ends of a one-dimensional continuum or rather they tend to co-occur. Earlier clinical-anatomical studies aimed at capturing the variability of the constellation of behavioural and psychological symptoms in the FTD spectrum had identified two presentations with distinct neural correlates: one predominantly characterized by disinhibition and impulsivity, and the other predominantly characterized by apathy and inertia (Le Ber et al., 2006; Snowden et al., 2001; Zamboni et al., 2008). Whereas these early studies assumed that such “disinhibited” and “apathetic” profiles were the opposite ends of a behavioural continuum (i.e., a patient could have one or the other presentation), it has been now demonstrated that disinhibition and apathy usually co-occur in the same patient with FTD (Kok et al., 2021; Lansdall et al., 2017; Peters et al., 2006), and often coexist in cognitively healthy young individuals (Petitet et al., 2021). Our findings support the hypothesis that apathy and impulsivity may coexist to variable degrees but remain independent constructs with separate neuroanatomical correlates. They suggest that there may be patients who are both apathetic and disinhibited, as well as patients who are apathetic and depressed.
The second Component, labelled as Disinhibition versus depression/mutism, was the only component that loaded specific behavioural disturbances on both its negative and positive ends. More precisely, it contrasted ‘disinhibition’ with ‘depressive mood’, highlighting aspects of disinhibition related to mania and abnormally elevated, expansive mood. But it also contrasted ‘disinhibition’ with ‘expressive deficit’ and ‘speech articulation defect’, highlighting aspects of disinhibition related to the prepotent verbal response and excessive garrulous chatter that FTD patients may present. Thus, we may assume that Component 2 captured several aspects of the multifaceted phenomenon associated with the broad term ‘disinhibition’, including those reflecting enhanced impulsivity or hyperactivation of the processes that generate the impulse, as well as those related to the loss of the knowledge of social rules or impairments in the suppression of prepotent responses and resistance to distractor interference (Magrath Guimet et al., 2021; Migliaccio et al., 2020). Importantly, this component did not change when we excluded patients that had started with language disturbances from the PCA, suggesting that it was not simply driven by their aphasia but rather captured behavioural variability across the different presenting phenotypes.
The third Component, labelled as Psychosis, remained stable and distinct even when increasing the number of extracted components in the PCA. This is consistent with findings in several previous studies (Aalten et al., 2008) and with the hypothesis that psychotic symptoms identify a specific phenotype in dementia (Ballard et al., 2020; Murray et al., 2014).
The second aim of the present study was to examine how the identified components of behavioural variability relate to changes in brain structure (MRI) and metabolism (FDG-PET). We preliminarly studied each modality separately with regression models exploring linear correlations: the unimodal VBM results were consistent with previous studies that had performed VBM correlational analyses of single behavioural questionnaires (Rosen et al., 2005; Sheelakumari et al., 2020; Zamboni et al., 2008). Interestingly, there were no regions of significant correlation between grey matter volume and Psychosis (Component 3).
By using a newly developed fusion analysis we then studied, for the first time, how the identified components of behavioural variability relate to the two imaging modalities in conjunction, i.e., whether they are mainly associated with changes in structure (MRI), metabolism (FDG-PET), or both. In fact, it would be reasonable to think that some symptoms may mainly derive from alterations in the metabolism and not be associated with detectable atrophy, which takes longer to occur. Some other symptoms, instead, may be a direct consequence of the neurodegenerative process, which causes cell death and synapsis loss, seen as focal grey matter atrophy. In addition, our multimodal decomposition technique allowed us to also uncover nonlinear relationships, as depicted by the trajectory plots often showing relationships that were flat for some portion and then changed or were even U-shaped, whereas previous studies had mainly searched for linear relationships.
The fusion analysis of MRI and PET data showed that voxels in which grey matter volume and metabolism decreased with increasing values of Apathy were mainly located in the anterior insula and anterior cingulate cortex, regions known to be part of the salience network (SN), and with hypometabolism in the right prefrontal cortex. The SN is specifically thought to be involved in detecting and processing salient information (Seeley et al., 2007). Another trajectory showed that increasing values on the Apathy component were associated with decreasing volume in the right cingulate and bilateral putamen, paired with largely decreased metabolism in the right prefrontal cortex. These two trajectories of multimodal covariation seem to capture what has been indicated as the motivational and cognitive components of apathy, respectively (Ducharme et al., 2018). Interestingly, in both trajectories a decreasing volume for bilateral subcortical structures was associated with hypometabolism of the right prefrontal cortex.
The fusion analysis on the Disinhibition versus depression/mutism component identified a trajectory with large clusters of hypometabolism, more than for atrophy, associated with increasing depression, mutism, and stillness in the left prefrontal cortex and in the sensory-motor cortex bilaterally. These regions have been associated, respectively, with language production, motor control, and depression (Davis et al., 2010; Ray et al., 2021). Another trajectory of the same Component captured instead clusters of atrophy, which was predominant for this trajectory, associated with increasing disinhibition, which was localised in the anterior insula bilaterally and right anterior cingulate. In addition, hypometabolism also involved the temporal poles. According to one functional interpretation of frontal-subcortical circuits (Tekin & Cummings, 2002), temporo-limbic structures are part of the orbitofrontal circuit, whose dysfunction is characterized by disinhibition syndromes including irritability, impulsivity, and undue familiarity. This has been interpreted both as primarily frontal, i.e., due to the loss of inhibition by the frontal monitoring system on the limbic system responsible for instinctual behaviors (Cummings, 1995), but also as primarily subcortical, i.e., due to the impaired risk perception mechanisms (Ghika, 2000).
Lastly, the fusion analysis on the Psychosis component mainly showed results from PET rather than VBM, suggesting that the symptoms described by Component 3 have greater functional rather than structural substrates. Among the trajectories associated with increasing scores of ‘psychosis’, one showed small clusters of atrophy in the basal ganglia (striatum) from the VBM and larger clusters of hypometabolisms in the prefrontal cortex bilaterally from the PET. This component may capture the mesolimbic dopaminergic pathway, the dysfunction of which has been associated with positive symptoms in schizophrenia (McCutcheon et al., 2018; McCutcheon et al., 2019). Another multimodal trajectory showed that increasing scores of psychosis are also associated with increasing metabolism in visual and auditory cortices, in line with the hypothesis that psychotic productive symptoms derive from aberrant hyperfunctioning primary sensory areas (Alderson-Day et al., 2016; Zmigrod et al., 2016).