In this amyloid-negative amnestic MCI cohort study, three main findings emerged in terms of conversion to dementia. First, the FBB SUVRs of the converter group were higher than those of the non-converter group in the temporal, parietal, and posterior cingulate cortices including the FBB composite score. Second, the SUVR of the parietal cortices predicted dementia conversion with the highest accuracy in this amnestic MCI with visually negative FBB PET scan compared to other VOIs. In particular, the combination of parietal, precuneus, and FBB composite score SUVRs predicted dementia conversion with the highest accuracy among our model comparisons. Third, although regional amyloid deposition is known to poorly correlate with neurological symptoms, we found that RCFT recognition test scores and regional amyloid burden of bilateral frontal cortices were negatively correlated in the converter group [
38,
39]. Taken together, this suggests the potential value of using subthreshold levels of amyloid deposition as a marker for predicting conversion to dementia from amnestic MCI, even in patients with visually amyloid-negative MCI on PET.
We found that the global deposition of amyloid was higher in the converter group than in the non-converter group. In particular, we confirmed that higher regional SUVRs were observed in the converter group than in the non-converter group, which is similar to the tau deposition of AD continuum corresponding to Braak stages III–IV [
40,
41]. Taken together, our findings suggest that subthreshold amyloid deposition plays a role in the conversion to dementia in two ways. First, subthreshold levels of amyloid deposition could induce pathological tau deposition in the brain. In fact, a previous study found that a low level or subthreshold level of Aβ was associated with deposition and spreading of pathologic tau protein even in a cognitively normal patient group [
4,
5]. In particular, the subsequent tau deposition in Braak stage I–II regions was best predicted by the baseline Aβ burden [
42‐
44]. Therefore, similar to the AD continuum, subthreshold Aβ deposition-induced pathologic tau deposition could be a causative factor of cognitive decline and conversion to dementia in these subpopulations [
5]. In a previous study, we found out that the hippocampus volume of the converter group was reduced compared to the non-converter group [
20]. In this study, the relationship between SUVR value and hippocampus volume was not found, suggesting the possibility of underlying tau pathology [
45]. However, it is difficult to figure out the rationale for amyloid deposition and dementia conversion in terms of ATN classification in this population, except for the possibility of T marker directly impacting on neurodegeneration (N) in the absence of an overt A marker. A subtle increase in the subthreshold amyloid levels might affect the pathologic tau protein aggregation. Second, subthreshold levels of Aβ deposition could lower the threshold for developing symptoms of combined pathological factors [
46]. Notably, as shown by Additional file
1, no statistically meaningful associations were observed between cortical thickness and SUVRs in either the converter or non-converter groups (Fig. S
2). This suggests that there might be other pathological mechanisms beside amyloid that lead to neurodegeneration. For example, several studies have been conducted in vascular cognitive impairment (VCI) patients investigating the association between comorbid amyloid pathology and vascular burden, and shown that combined vascular and amyloid pathology revealed poor cognitive function in each case [
47]. Furthermore, although still controversial, previous studies on VCI patients have found that patients with amyloid pathology had a poorer prognosis than those without amyloid pathology [
14,
48,
49]. Thus, the small amount of Aβ deposition present in these patients could lower the threshold for the emergence of cognitive impairment caused by other pathology. In fact, other studies have reported that TDP-43 and hippocampal sclerosis are often associated in AD patients [
50]. In particular, TDP-43 pathology is known to interact with amyloid deposition as well as pathological tau aggregation [
51]. From this perspective, it is possible that TDP-43 pathology or hippocampal sclerosis might be related to amyloid pathology at the subthreshold levels, but future studies are needed to confirm this.
The second major finding of our study was that the regional amyloid burden in the parietal and precuneus predicted conversion to dementia with the highest accuracy. These results corroborate the previous findings showing that there is a relationship between amyloid accumulation and cognitive changes in amyloid-negative subjects using the ADNI data [
7,
12]. Our results of analyzing ADNI dataset also revealed that the amyloid burden in the precuneus could predict cognitive decline in amyloid-negative MCI. Furthermore, even subthreshold levels of regional amyloid deposition could be prognostic factors in these patient groups. Other previous studies on the AD continuum have also revealed that the regional amyloid deposition in the parietal cortices and precuneus was important not only for cognitive decline, but also for conversion to dementia [
52‐
54]. Notably, the effect of amyloid deposition has been shown to be the most robust in the precuneus and posterior cingulate cortices, key areas constituting the default mode network (DMN), in the AD continuum [
55]. Interestingly, given our findings that the subthreshold levels of amyloid deposition in parietal cortices and precuneus are important for conversion to dementia, it is possible that these regions are responsible for cognitive decline. That is, these DMN-related regions may be dysregulated via amyloid deposition even at the subthreshold levels observed in our patient group [
19,
56].
The third finding of our study was that performance on the visual memory retrieval task was correlated with the regional amyloid deposition in the bilateral frontal cortices, including the middle frontal cortex, only in the converter group [
57]. Notably, there was no significant correlation between cognitive function and regional amyloid deposition in the non-converter group. In general, cognitive function along the AD continuum has been known to be more associated with the distribution of tau proteins than amyloid deposition [
58‐
60]. In fact, it is questionable that frontal amyloid deposition alone causes pertinent cognitive deficit [
61,
62]. Therefore, other pathologies such as pathological tau proteins or TDP-43 might be involved [
16,
63]. Nevertheless, we did not observe a correlation between amyloid burden and quantitative structural brain MRI analysis. Future studies are warranted to further assess the effects of subthreshold levels of amyloid burden on cognitive function and brain structure.
Limitations
There are some limitations in our study. First, the presence of other pathological proteins such as tau, TDP-43, and α-synuclein was not evaluated in this patient group [
64]. Given the pathological heterogeneity and uncertainty of amyloid-negative amnestic MCI patients, postmortem pathological examination along with the active use of various biomarkers should be considered to elucidate the underlying pathophysiology of this condition. Other than VCI, few models of combined pathology interactions have been studied and presented yet. Also, because our dataset excluded patients with vascular pathology, there is a limitation in explaining our findings compared to VCI. Therefore, further research is warranted on the interaction of combined pathologies within our dataset. Second, we cannot completely exclude the possibility of a conversion to Alzheimer’s dementia from amyloid-negative MCI [
12]. In other words, cumulating amyloid deposition might have gone above the threshold levels at some point in the course of time. Therefore, a follow-up PET scan should be performed to verify this. Finally, the follow-up neuropsychological assessments (SNSB) were not performed in all patients, which made it difficult to evaluate the changes in each cognitive domain. To overcome this, three experienced neurologists evaluated the patients according to the aforementioned criteria. However, due to the inherent problems with a retrospective study, it seems that selection bias could not be completely excluded from the process. The higher conversion to dementia rate of our patient group (36.4%) than other previous studies could also be due to selection bias. Future prospective studies, including longitudinal follow-up of structural and functional imaging with post-mortem pathological confirmations, are needed to better address the clinical progression of amyloid-negative MCI. Despite these limitations, our study assessed the effect of amyloid deposition on conversion to dementia in seemingly amyloid-negative MCI by excluding patients with an ischemic lesion and investigated the relationship between levels of regional amyloid uptake across each cognitive domain function by administering detailed neuropsychological tests.