Background
Frontotemporal dementia (FTD) refers to a group of progressive neurodegenerative disorders associated with predominant frontal and/or temporal lobe atrophy [
1]. FTD is clinically heterogenous and encompasses three main phenotypes: a behavioal variant (bvFTD), semantic dementia (SD), and progressive nonfluent aphasia (PNFA). BvFTD is characterized by profound behavior and personality changes [
2,
3], with atrophy predominantly observed in the frontal and anterior temporal lobes, particularly affecting the insula, orbitofrontal, and medial prefrontal cortices [
4]. SD is marked by progressive non-modality specific loss of semantic knowledge [
5]. Brain imaging in SD consistently report marked asymmetric (usually more left lateralized) anterior temporal lobe atrophy [
6,
7]. PNFA typically presents with impaired expressive speech and language skills [
2,
5] and atrophy in the left inferior frontal gyrus and anterior insula [
8,
9].
In addition to cortical damage, FTD also involves changes in subcortical brain regions such as the amygdala. The amygdala is located in the medial temporal lobes and plays a central role in various cognitive and socio-emotional processes such as memory and learning [
10]. A plethora of cross-sectional studies have found significant amygdala damage in all clinical forms of FTD [
6,
11‐
16]. Yet, they have yielded inconsistent results, and only a few have directly compared amygdala changes across FTD phenotypes. For instance, while some studies have found bilateral and symmetrical amygdala atrophy in bvFTD [
13,
17,
18], others have not [
12,
19,
20]. In contrast, pronounced and asymmetrical amygdala atrophy (usually left greater than right) has been reported in SD, with variable involvement of the right amygdala [
6,
19,
21]. Furthermore, previous findings on amygdala atrophy in PNFA are limited and inconclusive, with studies reporting bilateral [
11] or left-confined [
19] atrophy compared to controls. These discrepancies may reflect variances in methodology (e.g., whole brain versus a-priori region of interest approach; manual versus automated segmentation), patient characteristics (e.g., disease duration and severity), and sample sizes.
The few existing longitudinal investigations of amygdala volumes in FTD have also yielded mixed results. A recent case–control study of bvFTD patients found no progressive atrophy over 12–16-months [
17], contrasting to Bejanin [
13], which detected greater volume loss in bvFTD compared to controls over a similar follow-up duration. Of note, several studies have used atlases that combined amygdala with the hippocampus [
22‐
25], making it difficult to determine to what extent each structure contributed to the observed decline. Research that directly compare progression trajectories of amygdala atrophy among FTD phenotypes is also notably lacking.
Crucially, the amygdala is a heterogenous structure comprising several nuclei with widespread interconnections with other cortical and subcortical brain regions [
26,
27]. These nuclei are traditionally grouped into three subregions based on their anatomical and functional characteristics [
26‐
28]. The basolateral (BLA) subregion receives substantial inputs from sensory cortices and is reciprocally connected with the frontal cortex, thalamus, hippocampus, and the occipital lobes [
26,
29,
30]. Functionally, the BLA subregion has been shown to be involved in decision-making, memory, and fear learning [
31]. The superficial (SUP) subregion plays a key role in processing olfactory and social information [
26], while the centromedial (CeM) subregion serves as the main output region that mediates motor behaviors such as flight reaction and startle responses, via its projections to the hypothalamus, basal forebrain, and the brainstem regions [
26,
32]. Advances in imaging techniques have made it possible to differentiate these nuclei on standard structural magnetic resonance imaging (MRI) in healthy and clinical populations in vivo [
33]. Recent research has identified specific amygdala nuclei volume reductions in the genetic forms of FTD [
34], the extent of such changes in the clinical forms of FTD, however, has not yet been explored.
Overall, previous research on amygdala atrophy in FTD has shown inconsistencies, with few studies directly comparing amygdala volumes across all FTD phenotypes. In addition, no previous research has examined whether amygdala subregions undergo distinct volumetric trajectories with disease progression. To address these gaps, this study aimed to investigate the profiles of amygdala atrophy in each clinical form of FTD, and track these changes at the subregional level over the disease course. Here, we applied a cutting-edge automated segmentation technique and accounted for intra-individual variability over time within a longitudinal cohort of well-characterized clinically diagnosed FTD patients, to establish the rates of atrophy of amygdala subregions in the main FTD syndromes. Considering the existing evidence, we hypothesized that all FTD variants would exhibit reduced amygdala volumes relative to healthy controls at baseline, with SD showing more severe amygdala atrophy compared to the other groups. We did not formulate any specific hypotheses of longitudinal trajectories of subregional change due to the scarcity of existing literature in this regard. Importantly, we also included typical Alzheimer’s disease (AD) as a disease control group of interest. Amygdala atrophy is frequently reported in AD, and some previous studies have shown comparable extent of amygdala atrophy and rates of annual progression between AD and bvFTD [
20,
24,
35]. Our study also aimed to uncover any potential differences in specific amygdala subregions between these diseases.
Discussion
Applying a novel automated brain segmentation protocol, our study demonstrates distinct amygdala subregional atrophy trajectories with disease progression in the canonical FTD syndromes. In particular, two findings enhance the existing body of knowledge on brain structural changes in FTD. First, our results revealed that the amygdala decline trajectories are driven by volumetric alterations in amygdala subregions specific to each FTD phenotype, with hemisphere-specific patterns. Second, we uncovered that amygdala atrophy presents early in the disease course of bvFTD and SD, whereas it occurs later in the disease course in PNFA. Importantly, the observed atrophy is mediated by changes in different subregions, underscoring the phenotypic heterogeneity across FTD variants. Overall, our systematic investigations provide a thorough understanding of the volumetric changes of amygdala in FTD as disease progresses and highlight differential involvement of amygdala subregions in the clinical forms of FTD. These findings are discussed in detail in the subsequent sections.
Consistent with our hypothesis, overall amygdala volume reduction was most severe in SD at baseline compared with all the other groups. SD also exhibited the most asymmetrical atrophy (left > right), which is a characteristic feature of the left-lateralized SD variant [
6]. The degree of asymmetry reduced with disease progression as demonstrated by the disproportionate volume reduction in the right amygdala over time. Our result is in keeping with previous research that documented a greater rate of degeneration in the right temporal lobe in left-lateralized SD patients [
46,
47]. We extend this finding by showing that the effect is driven specifically by significant progressive volume loss in the BLA and SUP subregions, and much less so in the CeM subregion. Indeed, the right CeM subregion appeared to be relatively spared over the disease course, whereas the left CeM subregion underwent significant atrophy at baseline but did not show prominent decline over time. Importantly, the BLA and SUP subregions have a similar cytoarchitecture to the cerebral and olfactory cortex. In contrast, the CeM subregion is thought to be an extension of the ventral striatum [
48]. These cytoarchitectural differences may potentially contribute to differential vulnerability of the amygdala subregions to progressive degeneration in SD. In addition, the observation that the BLA and SUP subregions are particularly affected in SD is consistent with their known involvement in emotion processing. The functional implications of this differential vulnerability remain to be fully explained.
In bvFTD, we found significant bilateral amygdala atrophy compared to the controls at baseline, albeit less pronounced than that observed in SD, aligning with some previous studies [
11,
17,
18]. We further revealed a symmetrical pattern of amygdala damage across all subregions, suggesting a uniformity in the extent of amygdala involvement in bvFTD. Despite an early and severe atrophy at baseline, bvFTD exhibited a less pronounced, and more symmetrical pattern of annual volume decline compared to SD and PNFA, highlighting distinct trajectories of amygdala atrophy progression across FTD phenotypes. It is plausible that more rapid pathological changes in the early phases of the disease are followed by a plateauing phase. Here, our study included scans up to 3.5 years after the baseline assessment, which amounts to half of the average disease duration in bvFTD generally reported. Longer follow-up periods and the use of non-linear modeling approaches will be needed to thoroughly test this hypothesis and provide a comprehensive account of amygdala atrophy trajectory in bvFTD.
In marked contrast to bvFTD and SD, the amygdala was preserved in PNFA at baseline. Over time, however, PNFA demonstrated widespread amygdala volume loss, notably in the right BLA and SUP subregions. Our findings highlight that amygdala volume reduction occurs later in the disease course in PNFA relative to bvFTD and SD. Importantly, while PNFA is described as primarily a language disorder with relative sparing of behavioral abnormalities [
5], several studies have reported the emergence of behavioral changes as early as 1-year post-baseline assessment affecting empathy [
49]. Others have also reported increased behavioral and emotional disturbances in this group 6 years after baseline assessment [
49‐
51]. Given that behavioral abnormalities are primarily linked with damage to neuroanatomical structures in the right hemisphere, it is plausible that global amygdala volume loss, notably the pronounced decline in the right hemisphere, underlies the emergence of emotional and behavioral disturbances observed in PNFA patients. As aforementioned, the BLA and SUP subregions are associated with roles in coordinating sensory input and processing social information, respectively [
26]. Our results point to a possible functional dissociation between the left and right BLA and SUP subregions, with a stronger implication of the right side in social cognition. Whether volume reduction in these specific subregions correlates with the emergence of behavioral symptoms remains to be investigated, which may shed additional light on the functional specificity of amygdala subregions in humans. Moreover, our findings may also partly explain the inconsistent finding of amygdala atrophy in PNFA, as the atrophy may be absent or too subtle to be detected by cross-sectional studies at the time of early presentation.
Notably, we identified a similar profile of amygdala change at baseline and over time in bvFTD and AD. This is in keeping with previous evidence [
24,
35], suggesting that alterations in total amygdala volume and/or its subregions do not discriminate between bvFTD and AD. It is well established that patients with bvFTD exhibit a higher prevalence and greater severity of behavioral disturbances compared to patients with AD [
52], and these disturbances are often associated with amygdala abnormalities. Some examples include emotion recognition deficits [
53,
54], reward deficits [
55], and changes in eating behaviors [
56,
57]. This seemingly surprising similarity in amygdala atrophy profile between bvFTD and AD may suggest a complex interplay between the amygdala and other brain regions, such as the hippocampus and orbitofrontal cortex, which collectively underpin various cognitive and socio-emotional functions. Several studies using diffusion weighted imaging techniques have reported more prominent and widespread white matter damage in bvFTD compared to AD, especially in the frontal and temporal regions [
20,
58]. It is plausible that severe disruption in the connectivity between amygdala and other brain regions in bvFTD underlies their profound behavioral symptoms, despite the similar degree of volume reductions within the amygdala with AD. Yet, to date, little is known about connectivity alterations of the amygdala subregions.
Some potential limitations of the study need to be noted. First, the automated segmentation protocol used in this study employs a probabilistic atlas constructed from high resolution ex vivo MRI data [
33]. It has been argued that FreeSurfer may tend to overestimate boundaries when processing low-quality images [
59]. We cannot fully exclude that the spatial resolution (1mm
3) of our T1 MR images may have had an impact on segmentation accuracy. This point, however, was mitigated by our careful manual quality control of all our T1 MR images that were checked for the presence of movement artefacts. Nevertheless, our volume measurements may need to be interpreted with caution, especially in the case of the CeM subregion, which comprises the smallest amygdala subregion and is associated with greater variability. Of relevance here, we quantified annualized volume decline by calculating the percentage of annual change in TIV-corrected amygdala volumes. Given that percentages provide a relative measure of change based on a baseline measurement, small changes in volume can appear significant in percentage terms if the baseline value was small, whereas large changes might appear less impactful when the baseline value was also large. Therefore, when interpreting the longitudinal decline rates, it is important to consider both the size of the subregions and the potential influence of percentage-based measurements on perceived changes. In addition, to minimize the risk of inaccurate boundary placement between nuclei, we clustered individual nuclei into three subregions according to their structural and functional characteristics. As such, changes in specific nuclei within these subregions cannot be discerned. Another possible limitation is our explicit decision to exclude patients with right-lateralized SD, which presents with extensive atrophy in the right temporal lobes, given their small number. Whether the severity and trajectory of amygdala changes in this group mirror those seen in the typical left-lateralized SD will be important to investigate, particularly in light of the renewed interest in this syndrome [
60,
61].
In summary, amygdala abnormalities are of clinical relevance in FTD, given its known role in cognitive and socio-emotional functions. Our study confirmed amygdala involvement in all FTD phenotypes, and for the first time, provided longitudinal data for volume changes in amygdala subregions with disease progression. Our findings suggest that amygdala subregions are differentially affected by FTD pathology, and have implications for differential diagnosis and monitoring of disease development and progression in FTD.