Frontotemporal dementia (FTD) is the second most common cause of early-onset dementia after Alzheimer’s Disease (AD); also, it is associated with progressive degeneration of the frontal and temporal lobes (Bang et al., 2015). Behavioral variant FTD (bvFTD) is the most common subtype of FTD and is characterized by personality change and social dysfunction (Rascovsky et al., 2011). The severity of behavioral manifestations of bvFTD is commonly evaluated using the frontal behavioral inventory (FBI), which is a caregiver-based questionnaire consisting of 12 positive and 12 negative behavioral symptoms (Kertesz et al., 1997, 2000; Milan et al., 2008).

The concept of “reserve” refers to individual differences in the capacity to withstand the pathological burden of neurodegenerative diseases (Katzman et al., 1988). Compared with people with a low reserve, people with a higher reserve can cope with a greater pathological burden before the onset of symptoms; they may show a less severe clinical presentation under similar pathological burdens. However, if the pathological burden increases, participants with a high reserve show a faster decline in the clinical presentation than those with a low reserve (Stern, 2012; Barulli and Stern, 2013). Cognitive reserve (CR) is a widely accepted concept in AD (Stern, 2012; Lee et al., 2019). Moreover, the concept of motor reserve (MR) has recently emerged in Parkinson’s Disease (PD) (Chung et al., 2020a,b). Previous research has suggested that CR is an environmental factor that contributes to heterogeneous cognitive function mediated by the neuroanatomical structure, metabolism, or cerebral blood flow; the precise mechanism remains unclear (Placek et al., 2016; Maiovis et al., 2018; Beyer et al., 2021).

Using FBI scores and statistical modeling, Borroni et al. (2012) identified four behavioral subgroups in participants with bvFTD, namely “disinhibited,” “apathic,” “language,” and “aggressive.” Furthermore, Premi et al. (2013) proposed the behavioral reserve (BR) hypothesis in FTD with cerebral single-photon emission computed tomography (SPECT); this is similar to the concepts of CR in AD and of MR in PD. These studies revealed that educational attainment was the only measure associated with a disinhibited phenotype. These studies also observed greater hypoperfusion in the right inferior frontal gyrus, left medial frontal gyrus, and right caudate in those with a higher education level than those with a lower education level.

The degree of behavioral manifestation varies in participants with bvFTD, even with similar degrees of cortical atrophy or white matter (WM) destruction. In this study, we have proposed a concept of BR to understand these individual behavioral differences. We have suggested a novel model to conceptualize BR based on the original definition of reserve, using T1 imaging and fractional anisotropy (FA) as surrogates of the neuropathological burden of gray matter (GM) and WM, respectively. Subsequently, we have tested whether BR can explain the heterogeneous clinical presentation of bvFTD.

Neural compensation, one of the neural reserve mechanisms, is a brain network newly developed to compensate for the disruption of the pre-existing brain network due to the disease pathology. In this respect, we also investigated the neural correlates and the associated networks of BR by estimating the BR of each participant with bvFTD based on their FBI score and neuroimaging abnormalities. Then, we identified the BR-associated functional brain networks using network-based statistics (NBSs) analysis.

In this study, we have proposed a concept of BR and investigated the neural correlates of BR in bvFTD, based on the definition of reserve (i.e., the discrepancy between the pathological burden and clinical manifestation) (Stern, 2012; Barulli and Stern, 2013). We hypothesized that the presence of BR explains the differences in the behavioral performances, despite similar pathological burdens, among participants with bvFTD. First, the nBR was defined, calculated, and validated. We then identified an interaction between the nBR marker and the neuroimaging abnormalities and its effect on behavioral performance. Finally, we identified the functional brain network associated with the nBR marker.

We wish to explain the individual variability of behavioral problems in bvFTD through the BR hypothesis. BR was considered to represent the difference between the estimated and observed behavioral performances. Our linear model calculated the estimated behavioral score from neuroimaging abnormalities. We used cortical thicknesses and FA values from DTI images to reflect the neuropathological burden of the GM and WM, respectively. Considering that AD-specific ROIs were more highly correlated with education (which is a proxy of CR) than with the global area (van Loenhoud et al., 2017), we additionally applied the neuropathological burden in the frontal, temporal, and insular lobes (these are representative pathological areas in bvFTD) (Du et al., 2007; Pan et al., 2012). Our experiments showed a higher correlation with these areas than with the global area. In the WM, the sagittal stratum (including the inferior longitudinal fasciculus and the inferior fronto-occipital fasciculus), cingulum (cingulate gyrus and hippocampus), fornix, stria terminalis, superior longitudinal fasciculus, superior fronto-occipital fasciculus, and uncinate fasciculus were selected as tracts of interest (Zhang et al., 2009).

The FBI scores and neuroimaging abnormalities were analyzed to create the model. The FBI score for negative symptoms was significantly correlated with the cortical thicknesses and FA values of the global and fronto-temporo-insular areas; however, the FBI score for positive symptoms was not. The cortical thickness and FA map-related FBI score for positive symptoms were considered to have no statistically significant areas because the difference in the scores between the participants was small. This can be attributed to the narrow distribution of the FBI scores for positive symptoms. Although the positive symptoms were more diagnostic for bvFTD, the negative symptoms were the most prominent symptoms at all stages (Benussi et al., 2021). According to the FBI manual, scores 25–30, 30–40, and >40 indicate mild, moderate, and severe disease states, respectively. The mean FBI score of our participants was 27.8; therefore, our data tended to represent participants with a mild disease status. In addition, the prevalence of genetic factors in bvFTD and AD is 30 and 5%, respectively (Rohrer et al., 2009). Because of the high proportion of genetic components, the environmental factors that are presumably core factors for BR may be relatively small.

The apathetic and language phenotypes (groups comprising negative items) also showed no significant correlations with neuroimaging abnormality. The apathetic phenotype included only two FBI items and excluded the major negative symptoms, such as indifference, disorganization, inattention, and personal neglect. The results for the language phenotype are thought to be because of the few language symptoms among the patients with FTD in this study, as only patients considered to have behavioral variants were selected and those considered to have semantic variants or progressive nonfluent aphasia were excluded. No significant correlations were observed between the nBR marker and the education level and occupation complexity, which were previously reported as proxies of other reserves in AD, FTD, and PD. In previous studies, only in cases with the disinhibited phenotype of bvFTD (which mostly consisted of the FBI score for positive symptoms), correlations between the education level and hypoperfusion in the frontotemporal area were observed on SPECT. However, apathic and language phenotypes, which were a part of the negative symptoms in the FBI, showed no correlated areas (Borroni et al., 2012; Maiovis et al., 2018). Our results were consistent with those of previous studies, suggesting that the education level may not be associated with BR.

Our major findings supported the BR hypothesis and demonstrated nBR marker-associated neuroimaging correlates. The high nBR group showed a higher behavioral score with lower neuroimaging abnormalities and a rapid decline of the behavioral score as reflective of cortical atrophy; this progressed similarly to CR in AD and MR in PD. Although our study is not a longitudinal study, the effect of the interaction between the nBR marker and neuroimaging abnormalities on the behavioral score was compatible with the traditional reserve concept (Stern, 2012; Barulli and Stern, 2013; Lee et al., 2019; Chung et al., 2020b). It has been reported that apathy is associated with the right dorsolateral prefrontal cortex, anterior cingulate, and putamen in FTD (Zamboni et al., 2008); attention is associated with the ventral prefrontal cortex in the attention-deficit/hyperactivity disorder (Durston et al., 2006); insights are associated with the orbitofrontal cortex and the frontal pole in the AD and FTD (Hornberger et al., 2014); and semantic performance is associated with the anterior temporal cortex in FTD (Williams et al., 2005).

The nBR marker was associated with the left frontal lobe, left precuneus, left cingulate, left middle and inferior temporal, left banks superior temporal, left supramarginal, and left inferior parietal cortical thicknesses. These areas overlapped with the previously reported areas associated with negative symptoms and further correlated with the default mode network (DMN). The DMN is divided into the ventral and dorsal medial prefrontal cortices (mPFCs), posterior cingulate, and precuneus (Raichle, 2015). The ventral mPFC is related to a personality change, emotional response, and mood control (Damasio, 1996; Bechara et al., 1997; Simpson et al., 2001). The dorsal mPFC is related to the regulation of emotional behavior and judgment of another person’s emotional state (Ochsner et al., 2004a,b).

Given that the negative symptoms of the FBI score are apathy, emotional flatness, inflexibility, personal neglect, and loss of insight, the results of these previous studies are in line with our results. The behavioral scores in those with lesser neuroimaging abnormalities were similar between the FA maps and cortical thickness analysis. Furthermore, as the FA value decreased, the corresponding decline in the behavioral score did not differ between the high and low BR groups. In FTD, structural connectivity is known to degrade prior to cortical atrophy (Gordon et al., 2016). Thus, nBR is more closely related to the GM. In addition, the BR, similar to other reserves, can be understood by two mechanisms: neural reserve and neural compensation (Stern, 2012). Neural reserve is a pre-existing brain network with a greater capacity to cope with a neuropathological burden; on the other hand, neural compensation is a brain network newly developed to compensate for the disruption of the pre-existing brain network due to the disease pathology. However, it remains unclear whether increased brain networks related to nBR in bvFTD are mainly associated with the capacity of neural reserve or neural compensation. From this hypothesis’s perspective, it can be interpreted that the GM of the high nBR group tolerated neurodegeneration better than the GM of the low nBR group (neural reserve mechanism). It can also be interpreted that compared to in the low nBR group, the WM in the high nBR group is better adapted to neurodegeneration by developing the compensation neural network identified in NBS analysis (neural compensation mechanism).

In addition, we also identified an nBR-related brain network (high nBR > low nBR) composed of the anterior cingulate gyrus, paracingulate gyri, left frontal operculum cortex, left occipital pole, right pallidum, right insular cortex, left parahippocampus, and right inferior frontal gyrus (pars opercularis). The anterior cingulate gyrus and the anterior insular cortices are a part of the salience network, and the paracingulate gyri, frontal operculum cortex, and pars opercularis of the inferior frontal gyrus are adjacent to the anterior cingulate and anterior insula. In bvFTD, salience network disruption is correlated with the clinical severity (Zhou et al., 2010; Day et al., 2013). Furthermore, the paracingulate gyri and the frontal pole were correlated with the recall performance in bvFTD. The functional connectivity between the occipital pole and the parahippocampus may be related to language ability in the FBI’s negative items.

Our study has several limitations. First, bvFTD has several genetic causes. Genetic perspectives include C9orf72, MAPR, and GRN genes, while proteinopathy includes the tau and TDP-43 proteins. However, we used a clinical diagnosis instead of a pathological or genetic-based diagnosis. Second, the neuropathological burden was calculated indirectly from MRI and did not include cellular or molecular data. Nevertheless, we used GM and WM to reflect the neuropathological burden as accurately as possible. Third, the estimated nBR using the residual approach could vary depending on which variables were used (Pettigrew and Soldan, 2019). Although we considered all available variables (age, sex, cortical thickness, and the mean FA values) and outcome measures (FBI score) in the model, we need to validate the nBR using another biomarker or outcome measure in the future. Finally, we could only suggest the effect of the interaction between nBR markers and neuroimaging abnormalities on the behavioral score of negative symptoms in this cross-sectional study. Hence, a longitudinal study would help confirm whether greater nBR would delay the decline of behavioral score or disease progression.

In conclusion, we propose a novel concept of BR in bvFTD, which is associated with the individual’s capacity against its neuropathological burden, especially for negative symptoms. The nBR marker-related GM areas and the functional brain networks were also found centered at the frontotemporal areas. Participants with a greater nBR marker would show lesser clinical manifestations of the same neuropathological burden. These findings can be used to predict the clinical progression of each individual with bvFTD, thus enabling physicians to provide appropriate interventions when available.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Institutional Review Board of the Samsung Medical Center. The patients/participants provided their written informed consent to participate in this study.

SHK performed the statistical analyses. SHK, JHP, and YJ performed the study design. BHL and SWS performed the data collection. SHK, BHL, SWS, and YJ performed the drafting of the manuscript. SHK, YJK, and PL performed the imaging data processing and analysis. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.