Differential early subcortical involvement in genetic FTD within the GENFI cohort

Highlights • Progressive and differential atrophy patterns are seen at presymptomatic stages across genetic groups.• Very early presymptomatic brain changes are detectable only by looking at small regions.• C9orf72 expansion carriers show the earliest and most widespread changes (cortex, pulvinar, cerebellum).• MAPT mutation carriers show early differences in the dorsolateral temporal cortex, amygdala, and hippocampus.• Late presymptomatic changes occur in GRN mutation carriers in dorsolateral prefrontal cortex, insula, and presubiculum.


Introduction
Frontotemporal dementia (FTD) is a common cause of early onset dementia. In about a third of the cases it is associated with an autosomal dominant inherited mutation in one of three genes: microtubuleassociated protein tau (MAPT), progranulin (GRN), and chromosome 9 open reading frame 72 (C9orf72) (Warren et al., 2013). For each of these genetic groups, there is evidence of a differential pattern of cortical atrophy (Chen and Kantarci, 2020), with changes occurring presymptomatically, up to twenty years before estimated phenoconversion Cash et al., 2018). Whilst these studies have been highly informative in describing the presence of brain changes in presymptomatic stages of the disease, they have focused less on subcortical structures, and in particular, they have not investigated specific subregions within the deep grey matter. However, due to advanced imaging methods, it is now possible to measure these individual nuclei and subregions in vivo on structural magnetic resonance scans, with prior studies in small cohorts showing changes at the symptomatic stage of genetic FTD (Bocchetta et al., 2016(Bocchetta et al., , 2019, but without any previous investigation of the presymptomatic period. Using data from the Genetic FTD Initiative (GENFI) cohort, we therefore aimed to examine the specific pattern of subcortical changes (including specific subregions), to determine which areas were impaired across the different disease stages of genetic FTD.

Methods
At the time of the fifth data freeze in the GENFI 2 study (03/03/ 2015-31/05/2019), 850 participants had been recruited across 24 centres in the United Kingdom, Canada, Italy, the Netherlands, Sweden, Portugal, Germany, France, Spain, and Belgium, of whom 804 had a volumetric T1-weighted magnetic resonance image acquired on a 3 T scanner. Another 26 participants were excluded as the scans were of unsuitable quality due to motion or other imaging artefacts, pathology unlikely to be attributed to FTD, or as they were carriers of mutations in one of the rarer genetic causes of FTD. All the remaining 778 participants were known to be either a carrier of a pathogenic expansion in C9orf72 or of a pathogenic mutation in GRN or MAPT (n = 480), or were non-carrier first-degree relatives (n = 298), who therefore acted as controls within the study. All aspects of the study were approved by the local ethics committee for each of the GENFI sites, and written informed consent was obtained from all participants.
All participants underwent a standardized clinical assessment as described previously . This included the CDR® plus NACC FTLD (Miyagawa et al., 2020) which was used to group the mutation carriers into stages: those with a global score of 0 were considered as asymptomatic, those with a score of 0.5 considered as possibly or mildly symptomatic (i.e. prodromal), and those with a score ≥ 1 were considered as fully symptomatic or phenoconverted (Table 1).
Left and right volumes were summed, and total intracranial volume was computed with SPM12 v6470 (Statistical Parametric Mapping, Wellcome Trust Centre for Neuroimaging, London, UK) running under Matlab R2014b (Math Works, Natick, MA, USA) (Malone et al., 2015). All segmentations were visually checked for quality with only one subject excluded from the cerebellar analyses due to the presence of an arachnoid cyst. Statistical analyses were performed in SPSS software (SPSS Inc., Chicago, IL, USA) version 26, with a linear regression analysis within each genetic group adjusting for age, sex, and scanner type (as there were significant differences between groups for each of these, Table 1), as well as total intracranial volume, with correction for multiple comparisons using the Benjamini & Hochberg method (Benjamini and Hochberg, 1995) using p = 0.05 for false discovery rate. The correction was performed separately for the genetic groups (MAPT, GRN, C9orf72), while considering the number of comparisons within each of the main regions (cortical, cerebellum, brainstem, thalamus, hypothalamus, amygdala, hippocampus, and other subcortical structures).

Total brain and cortical volumes
The total brain volume was significantly smaller in all genetic groups with CDR ≥ 1 when compared to controls (8-10% volumetric difference, p < 0.0005). However, it was also significantly smaller in C9orf72 expansion carriers at CDR 0 and 0.5 (1-3%, p ≤ 0.004) (Supplementary Table 1, Supplementary Fig. 1).
C9orf72 expansion carriers with a CDR ≥ 1 showed significantly smaller volumes than controls in all cortical regions, with the largest differences in the anterior and posterior insula (24%) (Supplementary Table 1, Supplementary Fig. 2). These two regions, together with the DLPFC, motor, dorsolateral temporal, lateral parietal and occipital cortex, were also significantly smaller in the C9orf72 expansion carriers with CDR 0 and 0.5 (2-7%, p ≤ 0.006). The temporal pole was significantly smaller than controls in those scoring 0.5 (6%, p = 0.004), while the orbitofrontal, posterior cingulate, sensory and medial parietal cortex were significantly smaller in those scoring 0 (1-4%, p ≤ 0.028), but these differences did not reach statistical significance in those scoring 0.5 (Supplementary Table 1, Supplementary Fig. 2).

Basal forebrain
Changes in the basal forebrain were only seen in MAPT mutation carriers (Supplementary Table 1, Fig. 2A), and only at the CDR ≥ 1 stage (15% smaller than controls, p < 0.0005).

Amygdala
All amygdalar regions were significantly smaller than controls for all mutation carriers with CDR ≥ 1 (p < 0.0005), with the MAPT group showing the largest differences, particularly in the superficial and Table 1 Demographic and clinical characteristic of the cohort divided by genetic group and CDR®+NACC FTLD global scores. Abbreviations: N/A not applicable, FTD frontotemporal dementia, bvFTD behavioural variant FTD, PPA primary progressive aphasia, NOS not otherwise specified, CBS corticobasal syndrome, PSP progressive supranuclear palsy, AD Alzheimer's disease, ALS amyotrophic lateral sclerosis.  Figure 2B). All regions were significantly smaller in C9orf72 expansion carriers at CDR 0, and in MAPT mutation carriers with CDR 0.5, with smaller volumes at CDR 0 in the superficial and accessory basal nuclei (2-4%, p ≤ 0.034) (Supplementary Table 1, Figure 2B). GRN mutation carriers with CDR 0 or 0.5 did not show any significant differences from controls.

Hippocampus
All hippocampal regions were significantly smaller than controls for all mutation carriers with CDR ≥ 1 (p < 0.0005), with MAPT mutation carriers being the genetic group with the largest differences (all above 30%) (Supplementary Table 1, Figure 2C). Differences were also seen in all regions in MAPT mutation carriers with CDR 0.5 (6-11%, p ≤ 0.019), and in the subiculum, presubiculum and tail (3-4%, p ≤ 0.020) in MAPT mutation carriers with CDR 0. In C9orf72 expansion carriers with CDR 0 there were significantly smaller volumes than controls in all regions except the tail and the subiculum (2-3%, p ≤ 0.015). The presubiculum was the only region significantly smaller in GRN mutation carriers with CDR 0.5 (8%, p = 0.016) with no significant differences at CDR 0.

Thalamus
C9orf72 expansion carriers showed significantly smaller thalamic regions in all stages, with the only exception being the ventromedial nucleus (which only became significant at CDR stage ≥ 1) and the ventral posterolateral nucleus, which did not quite reach statistical Fig. 1 Fig. 2. A-G. Plots representing the means and standard error bars for regional brain volumes for each of the stages in C9orf72, MAPT and GRN mutation carriers. Volumes as expressed as % the mean volumes in controls (y axis). * indicates a significant difference from controls after correcting for multiple comparisons. Abbreviations: Amygdala: CAT corticoamygdaloid transition area; Thalamus: LD laterodorsal, VLa ventral lateral anterior, VLp ventral lateral posterior, VPL ventral posterolateral, LGN lateral geniculate nucleus, MGN medial geniculate nucleus; Hypothalamus: AI anterior inferior, AS anterior superior, I-TUB inferior tuberal, S-TUB superior tuberal; Brainstem: SCP superior cerebellar peduncle. significance at CDR 0.5. The most affected regions at CDR 0 were the LD (13%), LGN (10%) and pulvinar (9%) (p < 0.0005) (Supplementary Table 1, Figure 2D). MAPT mutation carriers with CDR ≥ 1 showed significantly smaller volumes in all regions except LGN, with the main differences located in the MD, midline and LD regions (22-26%, p < 0.0005) but no differences at earlier stages. GRN mutation carriers also only showed significantly smaller regions at CDR ≥ 1, with the main differences located in the MD and midline regions (31%, p < 0.0005), followed by the anteroventral and LD (21-25%, p < 0.0005). No differences were found in the LGN and MGN.

Hypothalamus
All regions were significantly smaller than controls for all mutation carriers with CDR ≥ 1 (p < 0.0005), except for the i-tub regions for C9orf72 and GRN mutation carriers. In the CDR ≥ 1 group MAPT mutation carriers had the smallest volumes, with differences above 29% in the posterior and anterior regions (Supplementary Table 1, Figure 2E). C9orf72 expansion carriers were the only ones showing early differences, with the CDR 0.5 group showing smaller volumes than controls in the anterior superior and s-tub regions (5-7%, p ≤ 0.017), and the CDR 0 group showing smaller volumes than controls in the s-tub region (1%, p = 0.008).

Brainstem
GRN mutation carriers with CDR ≥ 1 showed smaller volumes in the superior cerebellar peduncle (5%, p = 0.011), midbrain and pons (7-8%, p < 0.0005), while MAPT mutation carriers scoring ≥ 1 showed smaller volumes in the midbrain (9%, p < 0.0005) (Supplementary Table 1, Figure 2G). No difference was detected in those with CDR 0 or 0.5, or in C9orf72 expansion carriers at any stage. Fig. 3 summarizes the sequential pattern of neuroanatomical involvement for each of the genetic groups, by indicating at which stage each region resulted significantly smaller than controls.

Discussion
In this study we have defined the pattern of involvement in subcortical brain regions and specific nuclei in genetic frontotemporal dementia. We have identified gene-specific changes in asymptomatic and prodromal stages through to fully symptomatic stages in C9orf72, MAPT and GRN mutation carriers. By looking at specific regions, in a large cohort of mutation carriers, we were able to identify small changes that occurs very early on in all genetic groups, which might go undetected when looking at the whole brain or at large regions.
The first brain regions showing differences from controls in C9orf72 expansion carriers without any detectable clinical symptoms were the thalamic regions (the pulvinar, LD and LGN in particular), the putamen, the CA regions with the dentate gyrus and the presubiculum, all the amygdalar regions, the s-tub region in the hypothalamus, the lobule VIIa-Crus II and VIIb of the cerebellum as well as several cortical regions. By the time C9orf72 expansion carriers reach the symptomatic phase, nearly all the regions in the brain become affected, with the exception of the caudate, nucleus accumbens, basal forebrain, brainstem, and the anterior and inferior cerebellum and the i-tub region of the hypothalamus. These results are in line with previous studies showing widespread involvement of the brain in C9orf72-associated FTD, well beyond the classical frontal and temporal regions of FTD Cash et al., 2018;Bertrand et al., 2018;Lee et al., 2017).
Among the thalamic regions, the pulvinar and LGN were particularly affected in C9orf72 expansion carriers, which is in line with previous research in both symptomatic and presymptomatic carriers Lee et al., 2017) and with the pathological accumulation of TDP-43 and dipeptide repeat proteins in those regions (Vatsavayai et al., 2016;Yang et al., 2017) Atrophy in these regions is linked to hallucinations and other psychotic symptoms as well as the altered processing of pain, features seen more commonly in C9orf72 expansion carriers than in other forms of FTD Ducharme et al., 2017;Fletcher et al., 2015;Kertesz et al., 2013).
Interestingly, the regions affected early in the cerebellum (lobule VIIa-Crus II and VIIb) are connected via the dentate nuclei to the ventral anterior and ventrolateral nuclei of the thalamus and from here to the DLPFC to regulate cognitive functions, and in particular goal-directed complex behaviours (Makris et al., 2003;D'Angelo and Casali, 2013;Palesi et al., 2015). The cerebellum is also connected via the anterior and ventral lateral posterior thalamic regions to the basal ganglia and the parietal and motor cortex (Palesi et al., 2015); all regions affected early in C9orf72-associated FTD. Dipeptide repeat proteins are also typical and abundant in the cerebellar cortex (Mackenzie and Neumann, 2016).
The most affected regions within the hippocampus and amygdala are among the ones previously shown to be atrophic in symptomatic C9orf72 expansion carriers (Bocchetta et al., , 2019 and connected to the temporal and posterior cortex (de Flores et al., 2017). The CA regions are abundant of dipeptide repeat proteins, with or without TDP-43 deposition (Mann and Snowden, 2017).
C9orf72 expansion carriers at CDR 0 showed reduced volumes in the s-tub region of the hypothalamus and later on in the anterior and posterior hypothalamus, leaving the i-tub as the only spared region as previously reported . The s-tub region includes the dorso-medial nucleus and lateral hypothalamic area, which regulate appetite and contain neuropeptide-expressing neurons and neuropeptide receptors (Parker and Bloom, 2012). Similarly, volume loss in the posterior hypothalamus and the presence of TDP-43 pathology have been linked to the development of abnormal eating behaviours, typical symptoms of bvFTD Piguet et al., 2011).
In MAPT mutation carriers, the only regions affected at CDR 0 were the superficial and accessory basal regions of the amygdala, and the subiculum, presubiculum and hippocampal tail. Such early differences in the amygdala could not be detected when looking at its volume as a whole, which only became significantly affected at a later stage. MAPT mutation carriers at CDR 0.5 additionally showed smaller volumes in the dorsolateral temporal cortex and in all the other hippocampal and amygdalar regions. Overall, the more medial regions of the amygdala (particularly the superficial, accessory basal and basal and paralaminar) tend to be affected more than the lateral regions. They are connected to key limbic regions and likely related to the development of symptoms associated with abnormal reward and emotional processing. These results are in line with previous in vivo studies on symptomatic mutation carriers (Bocchetta et al., , 2019 and with pathological studies: tau deposition is extensively found in the hippocampus and other limbic structures in MAPT mutation carriers (Ghetti et al., 2015).
By the time MAPT mutation carriers are fully symptomatic, we find lower volumes in the other key regions of the limbic system, such as the insula, anterior cingulate, mediotemporal cortex, nucleus accumbens and basal forebrain. This latter structure, the basal forebrain, was only affected in the MAPT genetic group, as previously reported . Other regions affected in this group include the midbrain, which forms part of a network that regulates emotion perception with the thalamus and amygdala (Liddell et al., 2005). All regions in the hypothalamus were also affected, although mainly in the superior and posterior regions as previously reported in a smaller cohort . Interestingly, the posterior region includes the mammillary bodies, connected via the fornix to the amygdala and hippocampus. Among the thalamic regions, the MD was the most affected, as previously reported : this region is connected to brain regions within the limbic network and plays a role in emotional and behavioural regulation, as well as executive function. This sequence of regional involvement and the localisation in the temporal lobe is in line with what has been reported in other studies in MAPT mutation carriers Cash et al., 2018;Whitwell et al., 2012;Olney et al., 2020). The vermis of the cerebellum, another important part of the limbic system, was not affected, in contrast with what was found by another study (Bocchetta et al., 2016). This could be due to the different way of classifying the symptomatic mutation carriers, and the presence of different scanner types and sample characteristics. However, the fact that the cerebellum was overall not affected in MAPT-associated FTD is in line with other studies .
GRN mutation carriers only showed significant atrophy at the CDR 0.5 stagethis was mainly cortical, affecting the DLPFC, and anterior insula, but there was also subcortical involvement of the presubiculum, a hippocampal region connected to the basal ganglia, frontal and parietal cortex, areas which are typically atrophic in GRN, as found here at the symptomatic stages and in other studies Cash et al., 2018;Olney et al., 2020); and which typically show TDP-43 accumulation (Mann and Snowden, 2017).
Previously described as spared in GRN mutation carriers (Bocchetta et al., 2016); we found here that lobule VIIa-Crus II, VIIb and VIIIa are affected later in the disease. These regions are connected, via the thalamus, to the DLPFC and primary sensorimotor cortex (Makris et al., 2003). Within the brainstem, the midbrain, pons, and superior cerebellar peduncle were atrophic, as previously found (Rohrer et al., 2010). The role of the brainstem in FTD is not yet fully understood, but TDP-43 pathology has been found previously in several nuclei of the midbrain and pons (Grinberg et al., 2011). When symptoms were clearly present in GRN mutation carriers, all the hypothalamic regions were also smaller than in controls, with the exception of the i-tub (similarly to the C9orf72 group). This region includes the arcuate nucleus, an important target for metabolic and hormonal signals (Parker and Bloom, 2012). Interestingly, in a previous histological study (and consistent with our findings), TDP-43 inclusions were not found in this region, but were abundant in the anterior, superior and posterior region of the hypothalamus (Cykowski et al., 2016).
C9orf72 expansion carriers showed by far the earliest and most widespread changes in the brain, compared to MAPT and GRN mutation carriers. Even the total brain volume was lower in C9orf72 expansion carriers at CDR 0 whilst only being affected at the fully symptomatic stage in MAPT and GRN mutation carriers. This result was found previously  and could suggest that C9orf72-associated FTD might be associated with a long and slow process of neurodegeneration which could start many decades before the onset of clinical symptoms, as also suggested by Staffaroni et al . GRN mutation carriers instead might have a more rapid process which occurs later and closer to symptom onset (Jiskoot et al., 2019). Longitudinal studies, such as in Staffaroni et al  and Whitwell et al , looking at the atrophy rates in the different disease stages could potentially provide a definite answer to whether this is the case.
This study has some limitations. Some of the nuclei are very small and we grouped them into combined regions, or clusters of nuclei. In the future, this could be addressed by imaging at higher field strengths (e.g. 7 T), enabling higher spatial resolution. There were differences in age, sex and scanner type for some of the groups, which we have taken into account by including these variables as covariates, although this cannot completely exclude their impact. The CDR 0.5 is smaller than the other groups, and is likely to be heterogeneous including both people who are truly in a mild prodromal stage, and others that score 0.5 due to 'questionable' symptoms that might instead be related to affective symptoms during the at-risk period.
By looking at in vivo regional volumetry, we have shown here a differential pattern of subcortical changes across severity stages in C9orf72, MAPT and GRN mutation carriers. By looking at a wide range of specific brain regions, for the first time we were able to measure small changes that occur in localized regions in the early stages of genetic FTD. These results suggest that these changes may be used as markers of neurodegeneration in future trials even during preclinical and prodromal periods. Further longitudinal studies, including multimodal imaging looking at brain connectivity networks and including correlations with cognitive and other biomarkers, will be vital to investigate these results further.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.