Background
Mild cognitive impairment (MCI) is a heterogeneous syndrome that can be due to Alzheimer’s disease (AD) and non-AD pathologies [
1]. The presence of an early and significant objective deficit of episodic memory is considered the main criterion supporting the diagnosis of typical AD and the best cognitive predictor of the development of AD dementia [
2,
3]. Though the amnestic syndrome of hippocampal type is the most typical presentation in prodromal AD, impairments in delayed recall tasks may be present in individuals with non-AD disorders, such as the behavioural variant of frontotemporal dementia (bvFTD) [
4], argyrophilic grain disease (AGD) [
5,
6] and the recently identified suspected non-AD pathology (SNAP) [
7‐
9].
Subjects with selective long-term memory impairment and a relatively stable or very slowly progressing (up to decades) condition have been reported [
10,
11]. This condition has been considered up to now as a phenotypical expression of a focal medial temporal lobe dysfunction possibly due to AD pathology [
10‐
12]. In particular, patients with AD with a limbic predominant deposition of tau protein present with a specific phenotype (e.g., old age at onset, predominant amnestic syndrome) and selective medial temporal lobe atrophy visualised by magnetic resonance imaging (MRI) compared with the other AD variants (i.e., typical AD and hippocampus-sparing AD) [
13].
Recently, patients with MCI with such a clinical phenotype and no clear in vivo evidence of AD pathology were reported in the literature as part of the heterogeneous clinical group labelled SNAP [
8,
9,
14‐
17]. This is a biomarker-based concept applied to individuals with evidence of neurodegeneration in the absence of cerebral amyloid load [
9]. This definition reflects the notion that pathologies other than AD may underlie neurodegenerative changes, as revealed by cerebrospinal fluid (CSF) and fluorodeoxyglucose-positron emission tomography (FDG-PET) biomarkers, in subgroups of patients clinically presenting with an AD-like phenotype.
In the present study, we assessed FDG-PET brain hypometabolism and cerebral amyloid load by PET imaging as well as CSF β-amyloid (1–42) protein (Aβ42) and tau values in a sample of subjects with MCI with predominant episodic memory impairment and a very slow rate of progression. The aim was to provide in vivo evidence to improve disease characterisation in this MCI subgroup.
Discussion
Our sample of subjects with aMCI with prevalent or exclusive long-term memory deficits and a slow cognitive progression did not show the FDG-PET hypometabolism pattern typical of AD. They had reduced glucose metabolism in the medial temporal lobe structures with no amyloid load visualised by PET imaging in these structures (
see Figs.
1 and
2).
In addition to the consistent pattern of reduced metabolism in the hippocampal structures, hypometabolism in the frontomedial cortex, insula and anterior superior temporal cortex was present in some subjects. A few subjects showed reduction of glucose metabolism in the posterior cingulate cortex, which can be interpreted as a functional disconnection effect due to the severe involvement of the hippocampal structures. As reported in the literature, even in the absence of grey matter loss or amyloid toxicity in the posterior cingulate cortex, grey matter loss in the medial temporal lobe structures is sufficient to cause remote metabolic effects in connected regions [
36].
This FDG-PET hypometabolic pattern involving limbic structures is similar to the pattern reported in some patients with bvFTD [
37,
38]. The subjects in our series, however, presented with a clinical phenotype of aMCI, and the additional mild behavioural changes found in many subjects did not fulfil the clinical criteria for bvFTD [
39] at either onset or follow-up.
As for the amyloid load visualised by [
18F]florbetaben-PET or CSF, we found evidence of β-amyloidosis in some subjects according to the amyloid-PET SUVr (i.e., 8 of 16 subjects showed intermediate or high SUVr values) and/or CSF Aβ
42 cut-off values [
19] (i.e., 10 of 20 showed low CSF Aβ
42 values). Notably, the cognitive profile and FDG-PET features of subjects with evidence of in vivo β-amyloidosis did not show differences from patients with MCI without amyloid load (
see Fig.
2 and Table
3). In addition, they did not progress to dementia during long-term follow-up (i.e., 58.3 ± 10.1 months). The clinical value of β-amyloidosis as a marker of prodromal AD in these subjects is poor, however. It must be taken into consideration that a consistent percentage of cognitively normal subjects (i.e., 16–30% [
27,
32]), as well as non-AD neurodegenerative patients [
27], show incidental amyloid load with mean neocortical SUVr values above the proposed cut-off score for prodromal AD [
40]. This is particularly true in elderly populations such as the one included in our study [
32,
40]. Thus, the β-amyloidosis seen in these subjects does not per se provide conclusive evidence for an “MCI due to AD” condition.
A pure or mixed limbic-predominant AD variant [
12] could be the possible underlying neuropathological substrate in those subjects presenting with low cortical amyloid uptake. However, it is unlikely in patients with no evidence of brain amyloidosis. The possibility of other pathological substrates needs to be considered in the presence of long-lasting aMCI. These include AGD, a late-onset tauopathy typically characterised by tau lesions spreading throughout the limbic system [
5,
6,
41]. AGD usually affects elderly subjects (mean age of onset 75–80 years), without sex differences, and is clinically characterised either by a severe amnestic syndrome with relative sparing of other cognitive functions [
5,
6] or by a clinical neuropsychological phenotype overlapping with that of typical AD dementia.
Mixed neuropathological conditions have been suggested as a possible underlying substrate in patients with MCI with a stable clinical picture. Multiple co-morbid neuropathologies were reported previously in a large neuropathological study of individuals (
n = 1337) followed longitudinally from normal or MCI status to death, showing that less than one-fourth of patients with MCI had “pure” AD at autopsy, whereas more than half of the stable patients with MCI had mixed AD pathology changes [
42]. Brain arteriolosclerosis was also reported as a possible neuropathological substrate in this MCI case series [
42].
No subject in our MCI sample progressed to dementia during the follow-up, notably not even the two patients who showed higher SUVr values. This finding also supports a non-AD condition overall in this subgroup and once more suggests a possible combined AGD and additional amyloid pathological substrate in those individuals with slightly elevated SUVr values. Some subjects with AGD indeed showed an associated AD-type pathology [
6]. AGD and AD pathologies may act as additive factors, because subjects with combined AGD with mild to moderate AD-type pathologies were more frequently associated with dementia progression than with pure AGD pathology [
5,
43].
We found mild behavioural disturbances in the majority of our subjects with aMCI. Although the presence of emotional blunting could suggest a diagnosis of a bvFTD condition, no other behavioural changes typical of bvFTD (e.g., disinhibition; impulsiveness; perseverative, stereotypic or compulsive/ritualistic behaviour; hyperorality; and dietary changes) were present. Abnormal behaviours such as those observed in our sample represent common initial symptoms in pathology-confirmed AGD [
44].
Although asymmetry of brain volume in medial temporal regions has been described in study of patients with advanced AGD [
45,
46], this anatomostructural aspect was not reported by the expert radiologists who evaluated conventional MRI scans in our series. This work was indeed performed in a clinical setting, and advanced MRI measures were not available. Future studies are needed to highlight the role of subtle and specific anatomical changes in the differential diagnosis of this MCI group.
Other possible pathological substrates also need to be considered, such as primary age-related tauopathy (PART) [
47] and TDP-43 pathology [
48]. PART is a tauopathy in the absence of β-amyloidosis in which medial temporal lobe structures are primarily involved and abnormally elevated CSF tau protein levels are often observed [
47]. Age-related deposition of TDP-43 protein is a neuropathological condition recently described in elderly people with impairments in episodic memory and no pathological diagnosis of AD or frontotemporal lobar degeneration [
48]. According to Nag et al. [
48], there is an association between this proteinopathy and age-related hippocampal sclerosis (i.e., reduction of the volume of CA1 and subiculum). Hippocampal sclerosis has been suggested by Bien et al. [
49] as the main cause of memory loss in patients with MCI with predominant episodic memory impairment, slow progression of neuropsychological deficits and preservation of other cognitive functions. The large majority of elderly people with hippocampal sclerosis included in the Nag et al. cohort [
48] had TDP-43 pathology.
Both PART and age-related TDP-43 proteinopathy can represent possible pathological substrates of the heterogeneous clinical spectrum called SNAP [
9]. Reports on subjects with SNAP MCI [
7‐
9,
14‐
17], as well subjects with PART [
47] and age-related TDP-43 [
48], indeed show some parallelism with the clinical features of our aMCI sample: advanced age, very mild cognitive symptoms, neurodegeneration without amyloidosis, and low risk of clinical neuropsychological decline. Only clinical and neuroimaging studies with neuropathological verification will provide conclusive results about the array of pathologies responsible for the clinical picture reported in such subjects with long-lasting stable amnestic MCI.