Cerebral amyloid angiopathy (CAA) is characterized by leptomeningeal and cortical beta-amyloid (Aβ) accumulation in small vessels (Vinters, 1987). CAA particularly affects the aged population, and it is associated with Alzheimer’s disease (AD; Ellis et al., 1996; Keage et al., 2009; Arvanitakis et al., 2011). Previous studies have indicated that CAA is characterized by vascular occlusion and local ischemia (Olichney et al., 1995; Cadavid et al., 2000; Thal et al., 2009). Besides, the local fragility of the vessels can result in singular or chronic intracerebral or subarachnoid bleeding (Vinters, 1987; Itoh et al., 1993; Qureshi et al., 2001, 2009; Nakata-Kudo et al., 2006; Charidimou et al., 2012). The typical CAA finding of linear cortical superficial siderosis (cSS) winds around the gyri and sulci of the brain and the blood component hemosiderin, in particular, can be measured in gradient-echo-T2*-weighted-imaging (GRE-T2*-WI), a sequence of MRI (Linn et al., 2008, 2010; Kumar, 2010). Recent studies pointed out a diagnostic (Linn et al., 2010; Charidimou et al., 2013a) and prognostic (Charidimou et al., 2013b; Linn et al., 2013; Wollenweber et al., 2019) value of cSS determination in patients with CAA. However, the detailed mechanisms and associations between Aβ aggregation and cSS are still unclear. The aforementioned findings suggest that there is a regional relationship between Aβ accumulation and cSS, and we aimed to test this hypothesis by in vivo imaging. In this regard, fibrillar Aβ is now sufficiently detectable by FDA-approved PET tracers in living patients (Clark et al., 2011; Curtis et al., 2015; Sabri et al., 2015), and modern data analysis pipelines allow to investigate the PET and MRI data in matched spatial orientation (Wollenweber et al., 2014). We hypothesized that regionally pronounced Aβ accumulation leads to a colocalized increase of cSS due to higher vessel vulnerability in these regions. In turn, this could facilitate improved judgment of Aβ-PET positivity by the specific analysis of cSS-positive brain regions. However, in a recent single-case analysis from our group (Brendel et al., 2020b), a patient with probable CAA did not show a regional increase of the Aβ-PET signal in brain areas affected by cSS. Thus, we aimed to explore this phenomenon in a larger cohort of ten patients with CAA who underwent GRE-T2*-WI MRI and ¹⁸F-florbetaben Aβ-PET. A detailed segmentation approach was used to investigate the regional agreement between fibrillar Aβ accumulation and cSS.

The patients presented themselves to the respective clinics with focal neurological symptoms (past intracranial hemorrhage n = 3, memory complaints n = 4, past ischemic stroke or transient ischemic attack n = 2, or presumed focal seizures n = 1). Of the four women (66.2 ± 14.4 years) and six men (78.0 ± 4.9 years), three patients were diagnosed with a possible CAA and seven patients with a probable CAA, according to the modified Boston criteria (Linn et al., 2010). HCs were not significantly different in age (patients: 73.3 ± 10.9 years; HC: 65.9 ± 5.7 years; p = 0.073) and sex (patients: M/F = 6/4; HC: M/F = 8/6; χ² = 0.020; p = 0.889). Details of the cohort are provided in Table 1.

According to the multiregion classifier (Figure 2), nine patients (9/10 = 90%) showed at least one, and all patients showed an average of 37.0 ± 12.6 Aβ-PET-positive brain regions. Four patients had a focal cSS (≤3 sulci) with a mean of 2.0 ± 1.4 cSS⊕ regions, and the other six patients had a disseminated cSS (>3 sulci) with a mean of 27.0 ± 14.8 cSS⊕ regions (Linn et al., 2010). The patient without any Aβ-positive region was rated with a disseminated cSS on MRI with 13 cSS-positive regions. From a total of 170 cSS⊕ regions in all 10 patients, 40 were observed unilateral, whereas the remaining 130 were observed bilateral (65 in each hemisphere). The frequency of cSS- and Aβ-positive patients (as shown in Figure 2) for each region is plotted in Figure 3A, and Spearman’s Rho analysis showed a slightly significant positive correlation (r_s = 0.277, p = 0.032). However, this association was driven by temporal regions characterized by both limited cSS and Aβ-PET positivity (Figure 3B). As shown in Table 2, concurrent cSS- and Aβ-positive atlas regions (94/600 = 15.7%) did not occur more frequently than statistical random distribution. There were no significant differences in patient characteristics among patients with high and low cSS- and Aβ-positive region overlap (Table 1). Transverse brain sections of one CAA patient (female, 75 years) are illustrated in Figure 4.

To substantiate the descriptive results, a detailed quantitative approach was pursued to investigate the regional relationship between cSS and Aβ deposition in more detail. In this study, we tested the SUVr differences between the VOI groups defined earlier. All SUVr MVs and SDs are provided in Table 3A, whereas all p-values of the two-tailed pairwise Student’s t-tests are provided in Table 3B. The results of all composite cerebral lobes (Figure 5A) and the frontal lobe (Figure 5B) are illustrated separately in a boxplot. When analyzing all composite cerebral lobes (n = 151 regions), the group of cSS⊕1 regions revealed significantly lower Aβ SUVr when compared to cSS⊕0 regions (cSS⊕1: 1.462 ± 0.293; cSS⊕0: 1.538 ± 0.265; p < 0.0001), indicating that the tracer uptake was higher in the surrounding brain region of cSS when compared to the direct localization of cSS. This was also reflected by lower SUVr of cSS⊕1 regions when compared to combined cSS⊕ regions (cSS⊕1: 1.462 ± 0.293; cSS⊕: 1.517 ± 0.254; p < 0.001). Thus, we questioned if Aβ SUVr in the vicinity of cSS-positive regions indicates the hypothesized Aβ accumulation due to higher vessel vulnerability. However, we did not find an elevated Aβ SUVr in cSS⊕0 regions when compared to contralateral regions, if they were cSS-negative (= contralateral-to-cSS⊕) (n = 35; cSS⊕0: 1.568 ± 0.300; contralateral-to-cSS⊕: 1.515 ± 0.260; p = 0.603).

We asked if significant effects could be masked by a combination of regions deriving from different lobes. However, the separate analysis of the frontal lobe (n = 78) still revealed lower Aβ SUVr in cSS⊕1 than in cSS⊕0 regions (cSS⊕1: 1.466 ± 0.287; cSS⊕0: 1.557 ± 0.258; p < 0.001), and these Aβ SUVr were lower than in combined cSS⊕ regions (cSS⊕1: 1.466 ± 0.287; cSS⊕: 1.525 ± 0.246; p < 0.05). We did not find an elevated Aβ SUVr in cSS⊕0 frontal regions when compared to contralateral-to-cSS⊕ regions (n = 18; cSS⊕0: 1.673 ± 0.221; contralateral-to-cSS⊕: 1.594 ± 0.236; p = 0.226). A stand-alone analysis of the temporal lobe, parietal lobe, and posterior cingulate gyrus (PCG) displayed no significant differences between any VOI groups after FDR and Bonferroni correction.

Pathogenetic considerations suggest that Aβ accumulation and cSS could have a regional association, and the aim of this study was to evaluate this potential linkage in a cohort of ten CAA patients with cSS pathology and Aβ-PET imaging. We observed that regional Aβ-PET positivity correlated with regional cSS positivity in 60 predefined brain regions. However, concomitant Aβ-PET- and cSS-positive regions did not occur more frequently than expected from a statistical random distribution. Using detailed quantitative segmentation approaches, we explored that brain regions affected by cSS showed decreased Aβ-PET SUVr compared to brain areas in the vicinity to cSS. However, regions in the vicinity to cSS did not show altered Aβ-PET SUVr compared to cSS-negative contralateral brain regions.

For a proper interpretation of our results, the small sample size needs to be considered in detail and as a limitation. One out of the ten patients did not show any Aβ-positive brain region (Figure 2). We used a clear Aβ-negative control cohort with a priori exclusion of outliers in the single regions to ensure sensitivity for Aβ-PET positivity in the single regions of patients with CAA. Thus, we assumed that this single case did not have significant Aβ deposition in the brain. However, all patients were diagnosed with possible or probable CAA according to the modified Boston criteria (Linn et al., 2010), and a positivity rate of 90% fits to previous studies reporting Aβ positivity in patients with probable CAA between 60 and 100% (Jang et al., 2019). The prevalence of CAA increases with age and is associated with dementia syndromes such as AD but often remains asymptomatic until death (Ellis et al., 1996; Keage et al., 2009; Arvanitakis et al., 2011). The age distribution of our study cohort was consistent with the typical age of patients with CAA (Arvanitakis et al., 2011) and was slightly higher than our control group. Only one patient had a CAA atypical age of only 45 years, so a hereditary cause could be considered (O’donnell et al., 2000; Knudsen et al., 2001; Linn et al., 2010). However, this case was not the one without Aβ-positive regions.

As apparent by the analysis of the multiregion classifier (Figure 2), Aβ and cSS affect all cerebral brain lobes. The temporal lobe was less affected by Aβ deposition, which is consistent with the previous literature discussing the occipital lobe as the main CAA localization (Vinters and Gilbert, 1983; Vinters, 1987; Attems et al., 2007, 2011). The cSS of our sample was represented strongest in the parietal lobe and least represented in the temporal lobe. Larger studies refer to the occipital (Shams et al., 2016) or frontal lobe (Zonneveld et al., 2014) as the most cSS-affected lobe, and the temporal lobe, as the less cSS-affected lobe (Zonneveld et al., 2014). Accordingly, previous studies vary as to the main localization of Aβ and cSS, but our sample was in line with the literature. We calculated Spearman’s correlation between the frequency of Aβ-positive patients and the frequency of cSS-positive patients in all 60 atlas regions, showing a slightly positive trend of colocalization (Figure 3). In this study, the temporal lobe showed regions with both low Aβ and low cSS, among them hippocampus and amygdala. Thus, the borderline positive correlation was likely driven by the low positivity rates of both biomarkers in the temporal lobe. This was confirmed by an analysis excluding the temporal lobe, which did not reveal significant colocalization. For an in-depth analysis, we investigated Aβ-PET positivity and cSS in all 600 regions to determine whether Aβ- and cSS positivity are colocalized (Table 2). Since 333 of the 600 regions were Aβ-positive (333/600 = 55.5%) and 170 were cSS-positive (170/600 = 28.3%), we expected 94.4 regions to be both cSS- and Aβ-positive in a statistical random distribution. With 94 out of 600 regions comprising combined cSS- and Aβ positivity, we did not find evidence for a non-random distribution. Similarly, the visual inspection of single cases (Figure 4) did not indicate that increased Aβ deposits are present in direct or immediate vicinity to cSS deposition. This cohort-based finding confirmed the previous case report of one patient from this study (Brendel et al., 2020b).

The atlas-based analysis provided a robust descriptive analysis of cSS and Aβ pathology distribution, but we also applied a methodologically more sophisticated approach to increase sensitivity for colocalization by segmenting the atlas regions. Therefore, we subsegmented all cSS-positive atlas regions (cSS⊕) into a part with only cSS (cSS⊕1) and a part surrounding the cSS but excluding direct cSS (cSS⊕0). The respective contralateral atlas regions were used as a reference group once they were cSS-negative (contralateral-to-cSS⊕). When analyzing all composite brain lobes (Figure 5A) and the frontal lobe separately (Figure 5B), less Aβ-PET signal was observed in regions with only cSS compared to composed atlas regions and compared to cSS surroundings. These group differences were proved to be significant despite rigorous correction for multiple testing. Among other factors, areas with cSS deposits may experience remodeling and structural changes in the brain parenchyma and cortical layers. The structural remodeling also alters the vascular supply and blood flow in the tissue; vessels may recede or be remodeled. This implies that the reduced vascular supply or the reduced cerebral blood flow could underlie the observed SUVr that decreases, since this semiquantitative measure of PET does not account for the alterations of perfusion changes (Berckel et al., 2013). Due to the accumulation of toxic species, not only changes in vascular structures but also neuronal loss and brain atrophy are expected to occur in the vicinity to cSS. In this regard, we did not account for partial-volume-effects (PVE) due to atrophy since our segmentation approach did not fit into available PVE correction models (Brendel et al., 2015). Hence, signal loss due to atrophy could be another confounder of the observed decreased SUVr. Then, the altered plaque morphology in terms of reduced Aβ plaque density could also lead to a reduced signal in PET (Biechele et al., 2021). In this regard, we noted that diffuse and vascular Aβ occurring at cSS sites provides fewer tracer binding sites when compared to cored Aβ plaques which could, in turn, hamper the detection by PET. Further studies with larger cohorts, a longitudinal study design, and complementary immunohistochemical examinations would be desirable to validate our results. Longitudinal studies will be of particular interest to better understand the mechanisms linking the occurrence of amyloid-related imaging abnormalities (ARIA) occurring both spontaneously in CAA-related inflammation with high levels of anti-amyloid autoantibodies (Antolini et al., 2021) and after exposure with monoclonal antibodies against plaque-deposited Aβ of AD immunotherapy (Chantran et al., 2019).

Our results are consistent with recent studies of methodologically different approaches that do not identify Aβ as a key mediator of neurodegeneration after stroke (Hagberg et al., 2020) or do not detect increased Aβ levels in PET in patients with pathologically confirmed CAA burden at the lobar level (McCarter et al., 2021). However, a large-scale population-based study identified a positive association between lobar cerebral microbleeds and Aβ load as assessed by PET, whereas there was no association between deep cerebral microbleeds and Aβ accumulation (Graff-Radford et al., 2019). Furthermore, a recent neuropathological investigation identified associations between advanced CAA in leptomeningeal vessels and cSS (Charidimou et al., 2020). Thus, vascular vulnerability, in general, has associations with Aβ accumulation, but cSS, in particular, does not further increase the regional probability of an elevated Aβ load. The clinical goal of this study was to determine whether there is a diagnostic benefit to look specifically at cSS-affected brain regions when classifying a patient as visually Aβ-positive or -negative. The results of our cohort study suggest that sensitivity for Aβ-PET positivity cannot be increased by considering cSS information from MRI.

The colocalization of Aβ and cSS was not detectable with commonly used PET and MRI methodology in vivo in a small sample of patients with CAA. Therefore, the specific visual judgment of Aβ-PET positivity in cSS-positive regions is probably not a promising approach for the classification of Aβ-PET in patients with CAA.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Ethics Committee of the Medical Faculty of the Ludwig Maximilian University of Munich. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

AF contributed to writing, data analyses, and statistical analyses. SS, HW, and JL contributed to MRI data analyses. DJ, KB, FW, and CC contributed to patient recruitment, patient evaluation, and data analyses. AF, AR, PB, and MB contributed to PET scans and PET data analyses. FW, CC, and MB contributed to study design and conception. AF, CC, and MB contributed to writing, drafting, and manuscript conception. All authors added significant scientific input and intellectual content to the manuscript.

MB received speaker honoraria from GE Healthcare, Roche and LMI and is an advisor of LMI. AR and PB received speaker honoraria from GE Healthcare. JL received speaker honoraria from Bayer Vital GmbH and is an advisory board member of Mediaire GmbH and Biogen GmbH. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.