Introduction
Cerebral amyloid angiopathy (CAA) is mainly characterized by pathological deposition of Amyloid-β (Aβ) peptides in the walls of cortical and leptomeningeal vessels. CAA may lead to intracerebral hemorrhages such as microbleeds (CMB), hematomas (ICH), focal subarachnoid hemorrhage, and cerebral superficial siderosis (CSS). In cerebral autopsy series, Alzheimer disease (AD) is frequently associated with some degree of CAA [
1]. AD is neuropathologically defined by an abnormal aggregation of Aβ peptides in the brain parenchyma, along with neurofibrillary tangles composed of intra-neuronal abnormally hyperphosphorylated Tau proteins. Aβ peptides thus play a central role in both CAA and AD [
1]. Despite the existence of other peptides putatively causing CAA in rare genetic forms, Aβ aggregation is the major peptide responsible for CAA in elderly people and most CAA patients.
Among AD patients, 4 to 10% exhibit the first symptoms before the age of 65 defining early-onset AD (EOAD) [
2‐
4]. In this population, a monogenic cause is identified in ~2 to 77%, depending on ages at onset and family history [
5‐
8]. Such pathogenic variants in either the amyloid-β protein precursor (
APP), presenilin 1 (
PSEN1), or presenilin 2 (
PSEN2) genes or duplications of the
APP locus are autosomal dominantly inherited. After their discovery in 2006 [
9],
APP locus duplications were described in autosomal dominant EOAD series with various frequencies, 8% (95% CI, 2.6–17.1) in France [
9] and 2.7% (95% CI, 0.32–9.3) in the Netherlands [
10]. This copy number variant (CNV), located on chromosome 21, may encompass the
APP gene with or without surrounding genes. However, no correlation has been identified between the phenotype of
APP duplication carriers and the size of the duplicated locus or the genes included but the small number of reported families precluded any definite conclusion [
11].
APP is clearly the main gene as its duplication is sufficient to cause EOAD and/or CAA through overexpression. mRNA levels are increased ~1.5 times in carriers [
9,
12] with widespread Aβ deposits in the brain parenchyma and vessels [
13]. Recently, we reported a family with an
APP triplication (4 copies of
APP) and a phenotype similar to
APP duplications [
14].
Some degree of inter and intrafamilial phenotype diversity has been reported with
APP duplication but from a small number of families [
11,
15]. Most carriers presented AD-related cognitive decline and around 30% with CAA-related lobar spontaneous ICH upon presentation. Moreover,
APP duplication carriers were more likely to present seizures [
16]. Neuropsychiatric disorders with hallucinations related to a pathologically proven Lewy body (LB) disease have also been reported [
15]. Given the rarity of
APP duplications, little is known of the radiological pattern. Large discrepancies have been reported, from normal neuroimaging to severe CAA with multiple CMB or large inflammatory-related CAA [
17]. The proportion of
APP duplication carriers exhibiting CAA according to Boston-imaging criteria [
18] (except the age criterion) remains unknown and studies on AD cerebrospinal fluid (CSF) biomarkers are still required.
Here, we analyzed the clinical, radiological, and neuropathological features of 43 patients carrying 24 distinct APP duplications and compared their MRI features to 40 APP-negative CAA controls.
Material and methods
This retrospective study analyzed the phenotypic data of APP duplication carriers detected in France in the CNRMAJ center (Rouen University Hospital) from a nation-wide recruitment since 2006. Patients or legal representatives provided informed written consent for genetic analyses, in a medical and research setting (RBM 02-59, EudraCT 2009-010884-18) or GMAJ, NCT01622894, respectively approved by Paris Ile de France II and CPP Nord Ouest I ethics committees. This study was also conducted in accordance with the Declaration of Helsinki.
Inclusion and genetic analyses
The CNRMAJ laboratory of Rouen has a nation-wide recruitment of blood samples from patients with EOAD and/or with early-onset CAA (onset before 66 years). DNA isolated from whole blood samples of EOAD ± CAA were screened by Sanger or exome sequencing for exons 16-17 of
APP,
PSEN1, and
PSEN2 coding exons and for
APP duplication by quantitative multiplex PCR of short fluorescent fragments (QMPSF), as previously described [
19], and CAA patients (without AD phenotype) were screened for
APP pathogenic variants and duplications. Genes surrounding
APP were assessed by additional QMPSF or digital droplet PCR [
20] experiments in order to assess the size of each duplication and its gene content. All patients were genotyped for
APOE by Sanger sequencing. Patients were recruited regardless of the presence of a positive family history.
We included all patients exhibiting APP locus duplication including probands and their relatives. Before diagnosis, patients had neurological examination, neuropsychological assessment, and some had cerebral MRI and CSF AD biomarker analysis. Clinical data were retrieved from patients’ medical charts provided by each referring clinician.
The MRI patterns of these patients were compared to a control group of 40
APP-negative CAA controls (by sequencing exons 16 and 17 and by QMPSF) referred either to CNRMAJ, Rouen or to the Department of Genetics, Lariboisière Hospital, Paris, and fulfilling radiological criteria for probable CAA [
18].
MRI analysis
All cerebral 1.5 or 3-T MRI containing magnetic-susceptibility sequences were assessed. These blood-sensitive sequences, used to evaluate intracranial bleeds, were T2 gradient echo (T2 GRE) sequences, T2*, susceptibility weighted imaging (SWI), or a T2*-weighted angiography (SWAN), depending on the MRI machine and protocol used. The diagnosis of probable CAA was performed according to revised Boston diagnostic criteria, except for the age criterion (> 55 years) [
18]. In order to assess a consensus, two clinicians (LG and DW) independently rated hemorrhagic features in each brain region of interest. All hemorrhagic lesions were analyzed independently by both experts, by visually counting each lesion. The presence and number of CMBs (< 10 mm diameter) and ICH (≥ 10 mm) was evaluated [
21], and CSS was classified as focal (≤ 3 sulci) or disseminated (> 4 sulci) after recording the number of sulci involved [
22]. Hippocampal atrophy was scored according to Scheltens scale on 3D acquisition or 2D coronal T1-weighted-sequences [
23]. Perivascular white matter lesions were scored according to Fazekas scale on T2 or FLAIR-weighted sequences [
24]. Lesions were also classified according to location: lobar pre- or post-rolandic, deep, or in posterior fossa. In case of repeated MRI, only the latest images were considered for rating.
CSF analysis
CSF was obtained by lumbar puncture (LP). All centers used a common 10-ml polypropylene tube to collect CSF (catalog number 62.610.201; Sarstedt, Nümbrecht, Germany). All samples were aliquoted after centrifugation for 10 min at + 4 °C in polypropylene Eppendorf tubes and then frozen at – 80 °C within 1 h. Aβ42, Aβ40, Tau, and p-Tau protein measurements were taken using an enzyme-linked immunosorbent assay (ELISA) (Fujirebio-Europe, Ghent, Belgium) according to the manufacturer’s instructions. Analysis was performed in duplicate and a coefficient of variation (CV) less than 15% was considered as acceptable. In this case, the mean of the two measured values was taken as final result. All sites belong to the same national ePLM network which was created to enhance harmonization of procedures regarding CSF AD biomarkers [
25]. As preanalytical and analytical procedures might still have impact on the quantification, we set the normal thresholds for all CSF biomarkers following local laboratory normal ranges.
Neuropathological examination
Nine brain autopsies were available (BES_262, EXT_298, EXT_773, EXT_144, EXT_019, two patients from ROU_037 and two patients from EXT_028). The brains were fixed in a 10% formaldehyde solution buffer for at least 3 months. Seven-micrometer sections were cut from paraffin-embedded blocks of frontal, temporal including hippocampus, occipital lobes, and cerebellar hemispheres and brainstem. Sections were stained with hematoxylin–eosin, periodic acid Schiff, Orcein, Luxol-Phloxine. Routine immunohistochemical study was performed using antibodies directed against alpha-synuclein (diluted 1/200) (Zymed, Clinisciences, Montrouge, France), PHF tau (AT8, 1/20) (Innogenetics, Gent, Belgium), glial fibrillary acidic protein (GFAP, 1/300), PrP (1/50), ubiquitin (1/100), and the macrophagic marker CD68 (1/300) (Dakopatts, Trappes, France). Vascular and intraparenchymatous amyloid deposits were characterized using β-amyloid protein antibody (diluted 1/100) (Dakopatts, Trappes, France).
Statistical analyses
Results are expressed as mean ± standard deviation (SD) unless otherwise specified. Fisher exact tests and Welch two-sample tests were used to compare radiological characteristics between APP duplication carriers and the comparison CAA group. Point estimates of odds-ratios and mean differences between the two groups were accompanied by corresponding 95% confidence intervals, as provided by the R fisher.test and t.test functions, without continuity correction. Logistic regression was used with adjustment for time from symptoms onset to MRI. Exception was made for disseminated sulci. Because none of APP duplication carriers displayed disseminated loci, logistic regression could not be used. Adjusted OR and p-value were computed using Firth logistic regression with the logistf R package using default penalization parameters We analyzed radiological findings using the R statistical software version 3.6.2.
Data availability statement
De-identified database and statistical analysis plan will be shared upon reasonable request for 2 years after publication.
Discussion
In this study, we analyzed the clinical, radiological, and neuropathological features of 43 patients from 24 European families harboring an APP duplication, a rare cause of autosomal dominant AD and/or CAA, and compared their MRI features to those of 40 APP-negative CAA controls.
The wide range of different duplications shown here with distinct breakpoints (Fig.
3) and the diverse ethnicities of
APP carriers reported in literature, as Japanese cases for instance [
29], are not suggestive of a founder effect. Interestingly, one of our patients (EXT_773) harbored a de novo duplication. Given the reports of different-sized duplications and the identification of the first
APP triplication, the
APP locus appears to be a hotspot region for recombination, likely related to different regions with short tandem repeats [
14]. In our study, 41 out of 43 patients were symptomatic patients and symptoms occurred between 42 and 63 years. Overall, 90.2% of symptomatic patients had major neurocognitive impairment, with a clinical diagnosis of amnestic AD and prominent episodic memory impairment in 41.4%, atypical presentation with prominent behavioral impairment in 9.7%, and severe dementia with quick bedridden state in 21.9% of patients. According to pedigrees, isolated cognitive decline occurred in 13/24 (54.2%)
APP duplication families, with no reported ICH, whereas mixed presentation (cognitive decline in some affected relatives and ICH in others) occurred in 10 (41.6%) families. Mean age at onset of cognitive decline ranged from 42 to 60, highlighting the early onset and severity of cognitive impairment in patients harboring
APP duplications and more than 90% of the cohort showed cognitive decline before 59.
Based on 16
APP duplication carriers, our group previously described seizures occurring in 31% of cases [
16], with higher seizure risk compared to
PSEN1 or
APP point mutation carriers in EOAD. An epileptiform activity consecutive to Aβ overproduction which modulates presynaptic and postsynaptic transmission in mice models was suggested, in addition to the effect of brain hemorrhages [
30]. By adding 27 new carriers, our series further underlines the frequency of seizures in half of
APP duplication carriers, independently of symptomatic ICH occurrence (in only 29.2%) or CSS on MRI (present in only 14.2%). Consequently, seizure occurrence in EOAD or CAA, whatever the clinical presentation, or in familial dementia, could be an argument for
APP duplication screening.
Symptomatic ICH related to CAA occurred in 29.2% of our series, with a range at onset from 42 to 63 years, in line with the literature with a global rate of 30% when all reported
APP duplications were gathered. In other autosomal dominant hereditary CAA such as Dutch type (HCHWA-D), Iowa or Italian
APP point mutations, ICH occurred at similar ages (between 40 and 65 years) but with a higher prevalence of 75% in Iowa and Italian mutations [
31] and up to 100% in HCHWA-D with fatal outcome in 2/3 cases after the first event [
32]. Overall, 14/19 (73.6%)
APP duplication carriers with available MRI fulfilled the radiological criteria for CAA, while only 5 displayed no hemorrhagic features. We identified several MRI features in those 14 carriers: scarcity of CSS, less extended white matter lesions, and high number of CMBs with prominent posterior location. Nevertheless, those results were not significant anymore after adjustment for time from onset to MRI except for the disseminated CSS (OR = 0.06 [0.00–0.47], adjusted
p value = 0.0038). One obvious explanation may be the early age of onset of
APP duplication carriers, but this may also suggest a correlation between
APP duplications and radiological profile. The lack of power due to relatively small number of patients may have prevented those results to be significant in multivariate analysis. It has been shown that the
APOE genotype has an impact on the MRI characteristics of CAA patients with a correlation between
APOE2 and CSS and
APOE4 and higher number of CMBs [
33,
34]. The underlying mechanisms of CSS and CMBs are now considered to be different, if not opposite [
35]. In
APP duplication carriers, the pathogenic mechanism is the overproduction of Aβ [
12]. This mechanism could lead to a suggestive radiological pattern including CMBs but no disseminated CSS and less white matter lesions. On the opposite, altered perivascular clearance of the peptide Aβ is often described in late-onset and sporadic cases of CAA. Indeed, a higher number of CMBs is known to be associated with severe amyloid load [
36], which is consistent with the more severe amyloid deposits found in
APP duplication compared to other
APP point mutations [
13]. On the other side, interestingly, CSS has been specifically linked to clearance defect in CAA [
37].
Given the wide clinical and radiological heterogeneity, we sought a correlation between phenotype and duplication size. This question was initially raised with DS. Indeed, in line with the common sharing of an extra copy of the
APP gene, located on chromosome 21 and its consequent overexpression [
12], EOAD is highly penetrant in DS [
38]. However, ICH is a strikingly rare feature in DS, despite the deposition of large amounts of amyloid plaques in the brain parenchyma and vessels. A recent histological study comparing 34 DS to 4
APP duplication carriers revealed prominent diffuse Aβ plaques throughout the cerebral cortex in DS, associated with CAA confined to leptomeningeal vessels conversely to
APP duplication phenotypes which exhibited capillary and arterial intraparenchymatous CAA with fewer Aβ plaques [
13]. This suggests that one or several duplicated genes in DS may provide partial protection against the pro-hemorrhagic effects of
APP duplication [
39] similarly to
BACE2 involvement in age at onset of dementia in DS [
40]. Hence, we investigated whether any of the surrounding genes could account for the hemorrhagic features in
APP duplications but found no correlation (Fig.
3). Moreover, the high heterogeneity even between the same family, carrying the same duplication, is a strong argument for assuming that this variability cannot be explained at the level of the duplication itself. The pro-hemorrhagic effects of
APP duplications may more likely be modulated by more distant genes on chromosome 21 or other factors. The
APOE genotype should also be considered as all patients carrying an
APOE2 allele present hemorrhagic features on MRI in line with previous data obtained from sporadic CAA patients [
41].
Overall, numerous fibrillary tangles and senile plaques, observed in the hippocampal formation and the isocortex, consistent with a definite diagnosis of AD were observed in all patients but one (presenting recurrent ICH). CAA was diffuse and severe with thickening of the leptomeningeal vessels, as well as superficial and deep intraparenchymatous small arteries, capillaries and venules, and small infarcts often found near amyloid vascular deposits. Interestingly, LBs in the substantia nigra and locus coeruleus but also cortical structures were found in 2/9 (22.2%) carriers, one of them already reported [
10]. The association of LB-pathology with AD-type pathology is however well recognized in autosomal dominant AD due to
APP,
PSEN1, or
PSEN2 mutations but also DS [
42]. The present work underlines the need for
APP duplication screening in families with a mixed phenotype of EOAD and LB dementia.
Finally, some clinical and radiological features of CAA due to
APP duplications could be helpful to study sporadic CAA. HCHWA-D, due to a fully penetrant
APP mutation, was previously considered to explore new MRI biomarkers in sporadic CAA diagnosis [
43,
44]. More specifically, one additional hemorrhagic lesion (i.e., CSS) was added to the modified Boston criteria in 2010 [
45]. Additional radiological features are still a matter of debate in a new revised version of the Boston criteria, as the possibility to consider hemorrhages in deep brain territories when associated with other lobar locations, or CSS severity [
45]. In HCHWA-D, no deep CMBs have been reported. Nevertheless, we found 2 deep CMBs in
APP duplication carriers, and more importantly one autopsy described vascular amyloid deposits in deep grey vascular structures, in the absence of lipohyalinosis. Taken together, these results support a revision of the radiological criteria integrating lobar/deep ratio of CMBs. Regarding CSF profile, all but one
APP duplication carrier showed decreased Aβ42 levels, underlying the potential diagnostic value of this biomarker, as previously suggested from a sporadic CAA cohort by our group [
46].
In conclusion, phenotypes associated with APP duplications are characterized by EOAD and/or CAA with overall symptomatic ICH representing 30%. More than 10% of carriers showed an atypical presentation such as isolated behavioral disorders or hallucinations suggestive of Lewy body disease and frequent early seizures. Subsequent APP overproduction leads to high amyloid burden, notably within cerebral vessels as demonstrated by the neuropathological data, the high number of CMBs, and the possible occurrence of CAA-related inflammation. Overall, we suggest APP duplication screening in all patients with CAA or AD onset before or equal to age 65 or early-onset family history.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.