We present the first study that directly compared BF volumes of HC, SCD, MCI and DAD patients in a large multi-center dataset. We found a significant effect of diagnosis on BF total and subnuclei volumes, which was most pronounced in Ch4p. In patients with clinically manifest DAD, almost all BF subnuclei were affected by volume reduction, whereas in MCI patients, volume reduction was yet confined to Ch4al and Ch4p. Moreover, Ch4p volume correlated significantly with cognitive performance measures (MEM and LANG) in participants with MCI. In the subgroup with available CSF amyloid markers, an interaction between diagnosis (SCD vs. HC) and CSF amyloid status was observed in the cognitively normal subgroups, indicating Ch4p volume reductions only in amyloid-positive SCD (but not amyloid-positive HC), which also showed linear positive correlations with Aβ42/40 ratios in the SCD subgroup, but not in the controls.
Our findings are consistent with the assumption that incipient BF atrophy already begins in individuals with preclinical AD pathology. BF atrophy most likely starts in Ch4p and subsequently spreads across the entire brain with progressing disease. Notably, the fact amyloid-positive HC subjects did not show comparable levels of BF atrophy, suggests that they present an earlier AD disease stage (NIA-AA stage 1) than the amyloid-positive individuals with SCD, whose cognitive complaints may already reflect subtle cognitive deterioration (NIA-AA stage 2), which may in turn be related to an early cholinergic deficit.
In accordance with previous in vivo MRI morphometric studies [9, 41], we found significant volume reductions of the BF and in most of its subregions in DAD patients. This is in line with previous post mortem studies in DAD, which have demonstrated that severe neurofibrillary degeneration and neuronal loss in the cholinergic BF, leading to reduced cholinergic innervation of the cerebral cortex [6, 7]. In addition, BF volume loss was most pronounced in the Ch4p subregion. This converges with histopathological data from DAD patients showing that, while the entire BF is affected by cell loss, neurodegeneration is most severe in Ch4p [42]. This observation mirrors the functional neuroanatomy of cholinergic modulation of cortical activity in AD, which is particularly affected in the superior temporal gyrus and temporal pole, key targets of the cholinergic projections from Ch4p [1, 43]. Unlike other BF subnuclei, Ch4p and Ch4al exhibited volume loss already in MCI patients, confirming earlier in vivo MRI studies [8], and providing evidence that the cholinergic BF subregions are not uniformly affected during earlier stages of AD progression. Instead, the posterior part of the NBM, Ch4p, appears to be most vulnerable to AD pathology. While this interpretation remains tentative owing to the cross-sectional nature of our data, this would be consistent with a recent study [15] which found longitudinal NBM, but not entorhinal atrophy progression over a two-year period already in healthy individuals with preclinical amyloid pathology, suggesting very early BF involvement in AD pathogenesis.
Our findings are compatible with the latter view by showing that Ch4p volume reductions are present in amyloid-positive SCD, while no atrophic changes were detectable in the entorhinal (and also hippocampal) areas. Neither SCD on its own, nor CSF amyloid pathology in clinically normal individuals were associated with NBM atrophy, while their combination was. While the lack of BF atrophy in healthy controls with amyloid pathology appears to contradict the abovementioned study [15], it has to be noted that their observations were based on longitudinal data over two years which may be more sensitive to detect degenerative processes in this early disease stage. Our notion that the simultaneous presence of amyloid pathology plus SCD symptoms reflects an advanced stage of preclinical AD is compatible with Vogel, Varga Dolezalova [44] who observed that cortical Aβ accumulation as assessed with PiB positron emission tomography (PET) and clinical SCD predicted cognitive decline in a complementary manner [45]. This is also in line with an increased hippocampal atrophy and lower cognitive performance in SCD with positive APOE status [46], even though APOE constitutes a risk factor rather than a biomarker for amyloid pathology. Taken together with findings that Ch4p volume loss precedes entorhinal atrophy and mediates longitudinal memory loss [15], one might conclude from our findings that SCD in amyloid-positive individuals constitutes a very early clinical manifestation of an AD-related cholinergic neuronal loss that cannot be fully compensated anymore, thereby leading to first symptoms of incipient cognitive decline.
This view is also supported by our correlation analyses showing that Ch4p volume decreases with increasing amyloid pathology as indexed by CSF Aβ ratios in SCD subsample. The relationship between amyloid pathology and BF atrophy is also supported by the literature. Cortical amyloid accumulation correlates with cholinergic system atrophy, as shown in animal models [3, 10] and post mortem studies in humans [12, 13]. The strong association between Ch4p atrophy and Aβ42/40 ratio in individuals with SCD may reflect downstream neurotoxic effects of amyloid aggregates and vulnerability of Ch4p cholinergic cells to amyloid-related neurodegeneration, which may trigger incipient functional impairments in innervated brain regions. This is compatible with previous findings showing that Ch4p volume reductions in SCD correlated with reduced glucose metabolism in the precuneus [23]. Reduced precuneus glucose metabolism has previously been suggested to constitute a highly sensitive biomarker for AD [47, 48], and in MCI, it is associated with a conversion rate toward DAD of more than 90% within two years [49, 50]. The AD-related early atrophy of Ch4p might also explain why individuals with both SCD and biomarker evidence for AD are at increased risk of future cognitive decline and progression to MCI and ultimately DAD [51-55], as in these individuals, Ch4p degradation has already started.
Interestingly, exploratory analyses for MCI patients did not reveal a significant interaction with CSF amyloid status, neither for the volume of Ch4p, nor other BF subnuclei. Even though surprising on a first view, this missing association was also reported by other groups [56]. They might suggest that at the earlier SCD stadium the amyloid deposition is the driving force, whereas at a later stage, the atrophy is more likely driven by other factors than amyloid. However, a more detailed analysis including tau and other biomarkers would be needed to further prove this hypothesis. Moreover, the reduced sample size and the cross-sectional nature of the imaging data may have precluded the detection of more subtle effects: Actually, a recent longitudinal study contrasting a large sample of amyloid-positive MCI individuals with amyloid-negative healthy controls found not only higher volume losses in the NBM (Ch4), but also in the medial septal nucleus/diagonal band of Broca of the MCI patients [57].
The lack of associations between BF volumes and cognitive domain scores in the clinically unimpaired HC and SCD groups is consistent with an earlier study in healthy elderly that found no correlations with specific neuropsychological measures, except for a positive association with a global intelligence score [58, 59] . Meanwhile, the linear relationships between volume loss of Ch4p and decline in memory and language performance within the MCI sample is partially consistent with a previous study that observed significant correlations of total BF volume with memory and executive test scores, respectively [59]. This suggests that the loss of cholinergic innervation due to NBM neurodegeneration may contribute to the decline of cognitive functioning during this disease stage, although we must note the exploratory nature of these finding due to the uncorrected significance threshold. The lack of complementary correlations in DAD patients seems inconsistent with a recent study with DAD patients under cholinesterase inhibitor treatment which observed significant associations between total BF volume and global cognition, as well as memory and executive functions [60]. Meanwhile, the current sample was much smaller, limiting the statistical power of the present analyses.
One potential methodological limitation of the present study is related to the limited spatial resolution of our MRI-based basal forebrain volumetry, considering the small size of the BF nuclei compared with the spatial resolution, image contrast, and the atlas used, which is based on post mortem data from one single subject. It should be noted, though, that the locations of the BF nuclei derived from this atlas are in good agreement with a cytoarchitectonical probability map derived from ten subjects [8, 9, 61]. Furthermore, even though the MR method is relatively coarse, spatial specificity is still high enough to assess basal forebrain anatomy sufficiently. Another limitation is that volumetric approaches can, unlike histopathological methods, not distinguish between cell shrinkage, cell death, glial or neural loss or reduction of extracellular space, Therefore, interpretations in terms of ‘atrophy’ or ‘volume loss’ remain debatable. Nevertheless, considering previous histopathological findings in the context of our MRI-based findings, neurodegeneration in the BF is the most likely interpretation of our results. The fact that volume changes were observed in the BF regions of interest, but not in proximal control regions (caudate, BA25), also argues against global effects, and for spatial specificity of our findings. Moreover, the limited CSF sample size precluded further subdivision according to both amyloid and tau pathology in this interim analysis. We note that exploratory analyses according to tau pathology (Supplement Table S6/Table S7) showed no significant effects. Dual stratification will probably become possible with future extensions of the dataset. Finally, as longitudinal MRI and cognitive data were not yet available for analysis, our cross-sectional approach could not track BF volume changes over time, which limits sensitivity to detect atrophic processes.
Despite these methodological caveats, important strengths of this study are the longitudinal multicentric study design, the relatively large group size, the availability of biomarkers and a broad neuropsychological test battery that was specifically designed to characterize SCD as broadly as possible. While longitudinal follow-up data collection is still evolving, it will be important to investigate the longitudinal trajectory of BF changes, its relation to evolving alterations of other brain structures like the entorhinal cortex and hippocampus as well as to cognitive decline, and to assess its predictive value for the conversion of SCD into MCI and ultimately DAD.