Proteomic profiling in cerebral amyloid angiopathy reveals an overlap with CADASIL highlighting accumulation of HTRA1 and its substrates

Cerebral amyloid angiopathy (CAA) is an age-related condition and a major cause of intracerebral hemorrhage and cognitive decline that shows close links with Alzheimer's disease (AD). CAA is characterized by the aggregation of amyloid-β (Aβ) peptides and formation of Aβ deposits in the brain vasculature resulting in a disruption of the angioarchitecture. Capillaries are a critical site of Aβ pathology in CAA type 1 and become dysfunctional during disease progression. Here, applying an advanced protocol for the isolation of parenchymal microvessels from post-mortem brain tissue combined with liquid chromatography tandem mass spectrometry (LC–MS/MS), we determined the proteomes of CAA type 1 cases (n = 12) including a patient with hereditary cerebral hemorrhage with amyloidosis-Dutch type (HCHWA-D), and of AD cases without microvascular amyloid pathology (n = 13) in comparison to neurologically healthy controls (n = 12). ELISA measurements revealed microvascular Aβ1-40 levels to be exclusively enriched in CAA samples (mean: > 3000-fold compared to controls). The proteomic profile of CAA type 1 was characterized by massive enrichment of multiple predominantly secreted proteins and showed significant overlap with the recently reported brain microvascular proteome of patients with cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), a hereditary cerebral small vessel disease (SVD) characterized by the aggregation of the Notch3 extracellular domain. We found this overlap to be largely attributable to the accumulation of high-temperature requirement protein A1 (HTRA1), a serine protease with an established role in the brain vasculature, and several of its substrates. Notably, this signature was not present in AD cases. We further show that HTRA1 co-localizes with Aβ deposits in brain capillaries from CAA type 1 patients indicating a pathologic recruitment process. Together, these findings suggest a central role of HTRA1-dependent protein homeostasis in the CAA microvasculature and a molecular connection between multiple types of brain microvascular disease.


Introduction
Cerebrovascular pathologies are a major cause of stroke, cognitive decline, and dementia thus posing a significant burden for aging societies [17,27,59]. Disorders of the brain microvasculature, collectively termed cerebral small vessel diseases (SVDs), are particularly common in the aging brain and encompass a variety of sporadic and hereditary conditions affecting small and medium-sized vessels in the cerebral cortex, subcortical white matter, and deep white matter [48,68]. Among the most frequent pathologies is cerebral amyloid angiopathy (CAA), an important cause of intracerebral hemorrhage and cognitive decline [2,13]. Vascular amyloid is also seen in a substantial proportion of patients with Alzheimer's disease (AD) [20]. CAA is characterized by the misfolding and excessive vascular deposition of amyloid-β (Aβ) peptides, which are generated by multiple proteolytical processing of the β-amyloid precursor protein (APP) [60]. While mostly occurring as a sporadic condition, CAA can also develop from rare APP mutations such as the E693Q mutation, which causes hereditary cerebral hemorrhage with amyloidosis-Dutch type (HCHWA-D) [33].
Cerebrovascular amyloid deposits can be observed in both large (e.g. leptomeningeal) vessels and small parenchymal arterioles and capillaries, with their presence in capillaries determining the classification of CAA into two subtypes. Type 1 is defined by the presence of Aβ pathology in capillaries and may show additional Aβ deposition in non-capillary blood vessels. In contrast, Aβ pathology in type 2 is restricted to leptomeningeal and cortical arteries and arterioles [58]. Vascular Aβ accumulation is primarily considered a consequence of reduced clearance of parenchymal Aβ, for which different mechanisms are discussed depending on vessel type: while an impairment of transendothelial clearance is regarded as the predominant pathological process in capillaries, a reduction in perivascular Aβ clearance might prevail in larger blood vessels [2,69]. The capacity of these physiological clearance pathways is believed to decline with age and under pathological conditions resulting in incomplete Aβ elimination and its focal build-up at different locations within the vasculature [12,50]. Capillary CAA is frequent in AD [3], is associated with both microvascular occlusions and disturbances of cerebral blood flow [57], and might contribute to cognitive decline in AD [24]. An invariant histopathological feature in CAA patients is a disruption of the vascular architecture including mural cells loss [21,32,65,66] and alterations of the extracellular matrix (ECM) with thickening, splitting and duplication of the basement membrane [43].
The molecular mechanisms linking vascular Aβ deposition to vessel pathology and dysfunction remain largely elusive. However, recent proteomic studies have identified a variety of proteins and pathways with a possible role in CAA pathophysiology. Two studies isolated leptomeningeal and large cortical vessels from cryopreserved post-mortem tissue by manual dissection or laser-capture microdissection [28,38]. A third study used biopsy material from patients undergoing surgery for large lobar hemorrhages [18]. Collectively, these investigators found several proteins including apolipoprotein E (APOE), clusterin (CLU, also known as apolipoprotein J) and vitronectin (VTN) to be enriched in vessels from CAA patients and to co-localize with Aβ deposits evidencing co-aggregation or recruitment of these proteins into Aβ deposits. APOE and CLU were further shown to interact with Aβ and to influence its aggregation and clearance in an in vitro model of CAA that recapitulates the intramural periarterial drainage process [18]. In a fourth proteomic study, microdissected cortical tissue showing either parenchymal or vascular Aβ pathology from AD patients with or without capillary CAA was used and norrin and collagen VIα2 (COL6A2) identified as highly selective CAA markers [25].
Despite the abundancy and presumed functional relevance of vascular amyloid pathology in the brain [45], the specific molecular changes in small parenchymal vessels from CAA type 1 patients remain insufficiently characterized. Recent advances in the protocols for isolating brain microvessels from autopsy samples have facilitated their targeted biochemical analysis [6,42,72]. For instance, Bourassa and colleagues examined microvasculature-enriched fractions from human cerebral cortex by a combination of Western blot, enzyme-linked immunosorbent assay (ELISA), and immunofluorescence analyses and found CAA patients to exhibit alterations in the abundance of several endothelial markers and of proteins involved in Aβ production and clearance [6]. To our knowledge, however, there have been no focused efforts to characterize the full range of proteomic changes in parenchymal microvessels from CAA type 1 patients.
Applying untargeted proteomics to isolated brain microvessels, we recently assessed the microvascular proteome of CADASIL, a monogenic form of ischemic SVD that is caused by mutations in the NOTCH3 gene and characterized by vascular deposition of the Notch3 extracellular domain (Notch3 ECD ) [72]. This study demonstrated a distinct proteomic profile characterized by the accumulation of multiple proteins. Among the most strongly enriched proteins was the serine protease hightemperature requirement protein A1 (HTRA1). Moreover, many of the accumulating proteins were found to be HTRA1 substrates suggesting a loss-of-function signature. Notably, loss-of-function mutations in HTRA1 cause yet another form of hereditary SVD [23,63]. We further demonstrated co-localization of HTRA1 with Notch3 ECD deposits and an accumulation of various HTRA1 substrates in brain parenchymal microvessels from CADASIL patients. Collectively, these findings provided evidence for a loss of HTRA1 proteolytical function as a critical step in CADASIL pathogenesis.
The current study aimed to characterize the proteomic profile of brain parenchymal microvessels in CAA type 1 to identify key molecular targets implicated in disease pathogenesis. The analysis resulted in the detection of a distinct CAA type 1 profile encompassing multiple secreted proteins and ECM constituents. This profile, which was not observed in AD cases, showed a remarkable overlap with the brain microvascular proteome of CADASIL [72] highlighting the accumulation of HTRA1 and several of its substrates. We further identify serum amyloid P component (APCS) and PRSS23, another serine protease, as novel HTRA1 substrates. Our findings suggest a role of HTRA1 in CAA type 1 pathogenesis thus further highlighting shared mechanisms among distinct types of cerebral SVD.

Human brain tissue
Cryoconserved human brain autopsy samples (cortex and adjacent white matter from occipital or parietal lobe) from 12 neuropathologically confirmed CAA patients, 12 neurologically healthy control subjects and 13 neuropathologically confirmed AD patients were obtained from the Netherlands Brain Bank (Netherlands Institute for Neuroscience, Amsterdam). All CAA cases had been classified as capillary type (type 1) by routine neuropathological autopsy examination.

Microvessel isolation
Microvessels were isolated from 100 mg of frozen brain tissue as previously described [72]. In brief, brain tissue was minced and homogenized in cold minimum essential medium using a glass tissue grinder (Wheaton). After adding Ficoll to a final concentration of 15% and centrifugation at 6000 × g for 20 min at 4 °C, the resulting pellet was resuspended in 1% bovine serum albumin (BSA) diluted in PBS, transferred onto a nylon mesh (40 µm), and extensively washed with cold PBS. Microvessels were collected by flushing the inverted nylon mesh and centrifugation at 3000 × g for 5 min. Purity was checked by light microscopy. Throughout the procedure, plastic material was coated with 1% BSA in PBS.

Protein extraction and Aβ ELISA
Protein extraction of isolated microvessels was performed in a buffer containing 4% SDS, 100 mM Tris-HCl pH 7.6, 100 mM DTT. Samples were processed by Precellys tissue homogenizer (5 × 30 s, 10,000 rpm, 30 s pause), heated for 3 min at 95 °C and subsequently sonicated (5 × 30 s, amplitude 100%, duty cycle 50%) with intermediate cooling using the VialTweeter sonicator (Hielscher). Lysates were cleared by centrifugation at 16,000 × g for 15 min, supernatants collected, and protein concentration determined using the colorimetric 660-nm assay according to the manufacturer's instructions (Thermo Fisher Scientific). Aβ 1-40 and Aβ 1-42 species were quantified using the V-PLEX Plus Aβ Peptide Panel 1 (6E10) enzyme-linked immunosorbent assay (ELISA) Kit (Meso Scale Diagnostics) according to the manufacturer's instructions. Samples were diluted 1:100 and measured in triplicates. The resulting Aβ levels were normalized to total protein concentrations.

LC-MS/MS
A protein amount of 30 µg vessel lysate was subjected to proteolytic digestion by a modified single-pot solidphase-enhanced sample preparation (SP3) protocol [26]. Resulting peptides were desalted, dried by vacuum centrifugation and dissolved in 20 µL 0.1% formic acid. A micro-flow LC-MS/MS system composed of a modified Dionex UltiMate 3000 RSLCnano System coupled to a Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific) was used for all proteomic measurements in this study, as described in detail by Bian et al. [5]. Peptides (v/v) formic acid and 3% DMSO was applied as follows using a flow rate of 50 µL/min: 0 min, 0.5% B; 0.2 min, 2% B; 52.7 min, 24% B; 60.2 min, 35% B; 60.5 min, 90% B; 62.7 min, 0.5% B. Sample loading and column wash was performed at an increased flow rate of 100 µL/min. The column was heated to 55 °C. MS1 spectra were acquired at a resolution of 120,000, a scan range from 360 to 1400 m/z, a maximum injection time of 100 ms and an AGC target of 3E6. The Top 20 precursors were subjected to higher-energy c-trap dissociation with a normalized collision energy of 28%. A resolution of 15,000, an AGC target of 1E5, an isolation window of 1.6 m/z and a maximum injection time of 22 ms was applied. The dynamic exclusion was set to 30 s.

Proteomic data analysis
The data were analyzed with the Maxquant software [16] version 1.6.17.0 and searched against a reviewed canonical FASTA database of Homo sapiens (UniProt, download: June 26th 2021, 20,395 entries). To recalibrate the peptide masses within a window of 20 ppm, the option first search was used. Main search was performed for peptides and peptide fragments within a mass tolerance of 4.5 and 20 ppm respectively. N-terminal acetylation and oxidation of methionine were set as variable, carbamidomethylation of cysteine as static modification. The false discovery rate (FDR) was adjusted to less than 1% for both peptides and proteins. For label-free quantification (LFQ) of proteins, at least two ratio counts of unique peptides were required. To determine the significance of protein abundance changes between the different groups, LFQ intensities were log 2 -transformed and a two-sided Student's t-test was applied. Relative quantification and statistical analysis were performed for proteins identified in at least six samples of each group using the following significance threshold: p value < 0.05 and log 2 LFQ ratios > 1.0 and < − 1.0 (corresponding to 2.0-fold and 0.5fold changes). The R software packages GOChord and SuperExactTest (version 1.0.4) [67] were used for generating the circos diagram and for statistically assessing the proteomic profile overlaps.

Sample characterization
Cryopreserved post-mortem brain samples from patients with cerebral amyloid angiopathy (CAA) type 1 (n = 12, mean age 80.5 ± 9.8 years), neurologically healthy controls (n = 12, mean age 79.8 ± 8.0 years) and patients diagnosed with Alzheimer's disease (AD) (n = 13, mean age 77.6 ± 8.5 years) ( Table 1) were obtained from the Netherlands Brain Bank. To verify the presence or absence of Aβ pathology we performed immunofluorescence staining of tissue sections of selected cases along with confocal microscopy imaging (Fig. 1a). CAA samples showed extensive co-localization of Aβ immunoreactivity with parenchymal arterioles and capillaries demonstrating a high microvascular Aβ load whereas there was no microvascular Aβ immunoreactivity in the AD cases, which in turn were characterized by parenchymal plaque staining. Control samples were devoid of any Aβ immunoreactivity. Next, we validated our recently established brain microvessel isolation protocol [72] (see workflow in Fig. 1a) in one of the CAA cases. Confocal microscopy imaging of vessel preparations stained for the basement membrane marker collagen IV and for Aβ revealed the presence of intact capillaries and enabled visualization of vascular amyloid deposits in high resolution (Fig. 1b). Aβ immunoreactivity was mostly detectable as patches of variable size in confined vessel segments. To obtain a quantitative measure of the load of Aβ 1-40 and Aβ 1-42 peptides in our samples, we performed ELISA measurements on the same microvessel lysates prepared for liquid chromatography/tandem mass spectrometry (LC-MS/MS) analysis. Aβ 1-40 and Aβ 1-42 were massively enriched in CAA samples (mean: 3226-fold for Aβ 1-40 and 111-fold for Aβ 1-42 compared to controls) and showed the highest level in the HCHWA-D case (Fig. 1c, d and Additional file 1: Table 1). In contrast, AD cases exhibited isolated enrichment of Aβ 1-42 (mean: 11-fold compared to controls). The brain microvasculature of CAA type 1 displays a distinct proteomic profile with enrichment of multiple secreted proteins For the proteomic analysis, individual vessel preparations were subjected to LC/MS-MS followed by label-free quantification (LFQ) of protein abundances. As a first step, we performed a comparison of CAA samples with neurologically healthy controls. A total of 3752 proteins consistently identified by at least two unique peptides in at least six samples of each group were used for analysis ( Fig. 2a and Additional file 2: Table 2). We found 35 proteins to be significantly altered with an abundance ratio of > two-fold (log 2 ratio > 1) or < 0.5-fold (log 2 ratio < − 1). The majority of them (n = 32) were enriched, 10 of which showed a ratio of > four-fold (Fig. 2b) revealing a strong overall tendency towards protein accumulation. As inferred from their mean intensity-based absolute quantification (iBAQ) intensities, significantly altered proteins were distributed across the full abundancy range (Fig. 2c). However, most of the proteins that were strongly enriched in CAA demonstrated high abundance, including apolipoprotein E (APOE), clusterin (CLU) and pleiotrophin (PTN). In contrast, the three depleted proteins prenylcysteine oxidase 1-like (PCYOX1L), VGF nerve growth factor inducible protein (VGF) and ubiquitin specific peptidase 39 (USP39) showed an overall lower abundance. A subcellular localization analysis based on UniProt database information revealed strong overrepresentation of secreted proteins (23 out of 35 proteins, 66%) (Fig. 2d). While some of these proteins such as APOE, CLU, serum amyloid P component (SAP, encoded by the APCS gene), vitronectin (VTN) and tissue inhibitor of metalloproteinases 3 (TIMP3) have previously been reported to be enriched in vascular amyloid deposits [18,25,38,62], we also identified several novel proteins including serpine E2 (SERPINE2), olfactomedin-like protein 3 (OLFML3), the complement components C3, C1QB and C1QC, the extracellular matrix protein tenascin-C (TNC), the secreted glycoprotein slit guidance ligand 2 (SLIT2), PTN and the serine protease 23 (PRSS23) that have so far not been associated with CAA . Next, we compared the proteome of brain microvessels from AD cases versus controls. A total of 3780 consistently identified proteins were used for analysis (Fig. 3a). Overall, we identified 82 proteins with a significantly altered abundance ratio of > two-fold (log 2 ratio > 1) or < 0.5-fold (log 2 ratio < − 1) (Fig. 3b and Additional file 3: Table 3). As in the CAA profile, significantly altered proteins distributed across a broad abundancy range (Fig. 3c). However, the overall extent of protein accumulation was substantially lower, with TNC as the single protein exceeding an abundance ratio of 4. Also, the subcellular localization profile of the 82 significantly altered proteins differed substantially from the CAA type 1 profile, in that there was an overrepresentation of cytoplasmic proteins but no enrichment of secreted proteins (Fig. 3d). This is in accord with the absence of detectable vascular Aβ deposits in the AD samples and indicates a less pronounced disruption of protein homeostasis. As a third step, we compared the microvascular proteomic profiles of CAA and AD considering proteins that surpassed the significance and fold change threshold (p < 0.05 and log 2 fold change < − 1.0/ > 1.0) in either dataset. There was a significant overlap of 12 proteins, which further showed directional consistency, whereas 93 proteins did not overlap. Out of those, 23 and 70 proteins were specifically altered in CAA and AD, respectively (Fig. 4a). The comparison of individual LFQ ratios for CAA type 1 and AD (each compared to the control group) demonstrated a CAA type 1-specific cluster with high abundant changes whereas this was not observed for proteins constituting the AD-specific profile or for proteins that overlapped between the CAA type 1 and AD profiles (Fig. 4b). Subcellular localization analysis highlighted that the vast majority of proteins constituting the CAA-specific profile were annotated as secreted (21 out of 23 proteins, 91%) whereas proteins constituting the AD-specific profile or overlapping between the two profiles were distributed among all main subcellular localizations (Fig. 4c). Collectively, these findings define a distinct proteomic profile with enrichment of multiple secreted proteins in CAA type 1.

CAA type 1 and CADASIL microvessels show a similar profile of protein accumulation
The accumulation of secreted proteins in the CAA type 1 profile was reminiscent of our recently determined brain microvascular proteome of CADASIL [72]. We, therefore, investigated a possible relationship by comparing proteins quantified in both datasets and exhibiting significantly altered abundance ratios (CAA type 1: n = 38, CADASIL: n = 67). This comparison revealed a significant overlap of 12 proteins (p = 1.47 × 10 -13 ) representing almost one third of the CAA type 1 profile (Fig. 5a). Notably, there was no overlap between the AD and CADASIL profiles. The abundance changes of the proteins overlapping between CAA type 1 and CADASIL showed complete (100%) consistency in directionality, with all proteins enriched in both datasets (Fig. 5b). Moreover, there was a significant correlation (R = 0.58, p < 0.05) of the log 2 LFQ ratios of the 12 overlapping proteins. Furthermore, all the overlapping proteins belonged to the secretory category providing additional evidence for shared mechanisms between CAA type 1 and CADASIL possibly related to the presence of abnormally folded proteins (Aβ and Notch3 ECD , respectively) that characterize these conditions. Plotting the Aβ 1-40 levels (as determined by ELISA) against the iBAQ intensities of the 12 proteins shared by the CAA and CADASIL profiles revealed significant correlations for all but 2 proteins ( Fig. 5c and Additional file 4: Figure 1) indicating that their accumulation is indeed Aβ-dependent and not a general feature of disrupted vessel architecture. Of note, this included data from a patient with HCHWA-D, which were fully consistent with the results in sporadic CAA type 1 patients.

Accumulation of HTRA1 and its substrates in CAA type 1
Among the proteins shared by the CAA type 1 and CADASIL profiles was HTRA1, a serine protease genetically linked to hereditary SVD [23,63] and further found to co-localize with Notch3 deposits in the vasculature of CADASIL patients [72]. Hence, we performed immunofluorescence co-staining of HTRA1 and Aβ in isolated capillaries from a CAA type 1 sample, which revealed a near-complete co-localization (Fig. 6a). This confirms and extends previous findings from Hondius et al. [25] and suggested a recruitment of HTRA1 to pathological Aβ deposits including a possible sequestration process.
To investigate whether the overlap between CAA type 1 and CADASIL could be attributed to HTRA1 substrate enrichment and thus indicate a loss-of-function mechanism as proposed for CADASIL [72], we conducted a three-way comparison between the CAA type 1, the CADASIL and our published HTRA1 −/− mouse microvessel profile. Notably, proteins accumulating in HTRA1 −/− mice, can be considered putative HTRA1 substrates, as these mice do not develop protein deposits. Eight of the 12 proteins that were enriched in both the CAA type 1 and CADASIL profiles (OLFML3, APOE, VTN, TIMP3, SERPINE2, CLU, NDP and PRSS23) also showed increased abundancy in HTRA1 −/− microvessels (Fig. 6b, c) suggesting that these proteins indeed represent HTRA1 substrates. HTRA1 was exclusively found in the wild-type animals and thus did not contribute to the signature. The same applies for APCS, which was not identified in mice. Thus, only C3 and C1QC were consistently quantified but showed no significant changes in HTRA1 −/− mice. Importantly, this finding was fully confirmed in a recent independent study on HTRA1-deficient mice [34]. Also, seven of the eight proteins (OLFML3, APOE, VTN, TIMP3, SERPINE2, CLU and NDP) have previously been shown in proteolysis assays to be HTRA1 substrates [1,9,14,44,64,72] (Fig. 6c). To explore the candidacy of APCS and PRSS23 as potential substrates of HTRA1, we next performed cell-based cleavage assays: APCS was the most strongly enriched protein in the overlap signature; PRSS23 shares features with HTRA1 in that it represents a member of the family of chymotrypsin-type serine proteases and is primarily secreted by astrocytes [53,61]. We expressed epitope-tagged constructs in HEK293 cells, a commonly used approach for analyzing HTRA1 proteolytic activity in a cellular setting [4,72], co-incubated substrate-and protease-containing supernatants and investigated the resulting protein levels using Western Blotting (Fig. 6d,  e). While HTRA1 was secreted as a ~ 55-kDa protein (bands of lower molecular weight represent autoproteolysis products), APCS was detected as a single band of ~ 51 kDa (26-kDa full-length protein plus the 25-kDa epitope tag) and PRSS23 migrated as a 37-kDa band (most likely generated from the 45-kDa full-length protein by intracellular cleavage at a furin consensus site). Upon co-incubation with wild-type HTRA1, APCS and PRSS23 levels were strongly reduced close to detection limit, whereas this was not the case upon co-incubation with HTRA1 S328A , an artificial mutation which eliminates the catalytic serine residue resulting in a complete loss of proteolytic activity, or in the presence of a HTRA1specific inhibitor. We thus identified APCS and PRSS23 as novel HTRA1 substrates providing further support for a role of functional deficiency of HTRA1 in CAA type 1 pathogenesis.

Discussion
Applying untargeted proteomics to isolated parenchymal microvessels from post-mortem CAA type 1 brain samples, we identified a distinct profile that was characterized by an enrichment of multiple secreted proteins and substantially overlapped with the proteomic profile of CADASIL, a genetic form of cerebral SVD. This overlap, which was not observed in a comparison with the brain microvascular profile of AD patients lacking vascular amyloid pathology, can for the most part be attributed to the accumulation of HTRA1 and several of its substrates. HTRA1 is a protease implicated in the regulation of extracellular protein homeostasis and centrally involved in the pathogenesis of yet another form of hereditary cerebral SVD [23,63]. Collectively, these findings indicate a critical role of impaired HTRA1-mediated protein degradation in the CAA type 1 microvasculature and suggest a c log 10 iBAQ intensity CAA case number

Fig. 5
The CAA proteomic profile overlaps with the CADASIL proteomic profile. a Venn diagrams illustrating the overlap between proteins significantly altered in CAA (red) or AD patients (blue) with proteins significantly altered in CADASIL (teal) as reported in our earlier study [72]. A comparison between the CAA and CADASIL proteomic profiles results in an overlap of 12 proteins (p = 1.47 × 10 -13 ) (orange), whereas no overlap was observed between the AD and CADASIL proteomic profiles. For better comparability, the CADASIL dataset was analyzed using the same significance and fold change thresholds as applied to the CAA and AD samples (p < 0.05 and log 2 fold change < − 1.0/ > 1.0). b Scatter plot of the log 2 LFQ ratios in CAA versus CON against CADASIL versus CON. Significantly altered proteins are highlighted according to the colors in a, non-significantly altered proteins are shown in grey. Overlapping proteins are labeled with gene names, as is NOTCH3. The serine protease HTRA1 is highlighted in green. c Profile plot of Aβ 1-40 levels (determined by ELISA) (black) and iBAQ intensities for the overlapping proteins (orange) in each individual CAA patient. For formal assessment of statistical correlations see Additional file 1: Figure 1 shared mechanisms across different types of cerebral SVD. In many respects, the current study represents a major advancement over previous proteomic investigations on CAA [18,25,28,38]. Aside from focusing on isolated brain microvessels we determined the proteome of CAA patients with pronounced capillary Aβ pathology. Generating proteomic data on a large sample of patients and controls in unprecedented depth and using label-free quantification, we identified a distinct set of proteins with strongly increased abundance. This profile recapitulates the enrichment of proteins reported in earlier studies (APCS, APOE, CLU, HTRA1, NDP, TIMP3, VTN), while also revealing an accumulation of several proteins so far not associated with CAA including PTN, SLIT2, TNC, the complement components C3, C1QB and C1QC as well as OLFML3, PRSS23 and SERPINE2. For a considerable fraction of these proteins we provide evidence for a link to HTRA1 proteolytic function (see below), but other proteins including PTN, SLIT2, TNC and the complement components might independently contribute to pathogenesis via alternative mechanisms: PTN, a pericyte-secreted neurotrophic growth factor that is exclusively expressed in the brain, has been shown to protect neurons against ischemic and excitotoxic injury [46]. Its recruitment to microvascular Aβ deposits might result in functional inactivation of PTN and thus enhance the adverse effects of vascular dysfunction on neuronal integrity in CAA. The extracellular signaling factor SLIT2, has been implicated in angiogenesis and the regulation of vascular permeability [30,35] and might thus contribute to the loss of BBB integrity in CAA [36]. TNC, an ECM glycoprotein linked to inflammatory processes in the brain, has been shown to associate with cored Aβ plaques as well as vascular Aβ deposits in AD patients and cognitively normal elderly individuals showing plaque pathology [40] and might represent an immune response modulator during CAA pathogenesis [70]. Support for a role of inflammatory processes further comes from the increased abundance of C3, C4A, C1QB and C1QC, components of the classical and alternative complement pathways.  Fig. 6 Accumulation of HTRA1 and its substrates in the CAA proteomic profile. a Immunofluorescence staining of an isolated capillary from a CAA patient (visualized by collagen IV staining) demonstrates a near complete co-localization of Aβ and HTRA1 immunoreactivity. b Venn diagram illustrating the overlap between proteins significantly altered in CAA with proteins significantly altered in CADASIL and in HTRA1 deficient mice as reported in our earlier study [72]. Eight of the shared proteins in CAA and CADASIL were also enriched in HTRA1 deficient mice (red filling). c Table with LFQ ratio information in the CAA, CADASIL and HTRA1 deficient proteomic profiles for the overlapping proteins, as for APCS, C3, C1QC and HTRA1. The latter four proteins are shared between CAA and CADASIL but not part of the HTRA1 deficient profile as they were not identified (APCS), not significantly altered (C3 and C1QC) or exclusively detected under HTRA1 wild-type conditions (HTRA1). Similar proteomic changes in HTRA1 deficient mice were recently reported by Kato et. al [34]. For most proteins in vitro cleavage data have been reported earlier. d, e APCS and PRSS23 are HTRA1 substrates. Shown are immunoblots of conditioned supernatants from HEK293 cells expressing PRSS23 (d) (detected via Myc-tag) or APCS (e), co-incubated with supernatants containing wild-type or active-site mutant (S328A) HTRA1 (detected via V5-tag) in the presence or absence of a HTRA1-specific inhibitor (5 μM). Molecular weight marker bands in kDa are indicated. The band above APCS in the S328A condition represent an unspecific cross-reactivity of the antibody. HTRA1 bands of lower molecular weight represent autoproteolysis fragments We found the proteomic profile of CAA type 1 to show remarkable overlap with the profile of brain microvessels in CADASIL, another SVD caused by protein misfolding. Like CAA, CADASIL is characterized by vascular protein deposits, whose primary constituent, however, is Notch3 ECD , a receptor fragment generated at the surface of mural cells. Notch3 ECD deposits differ from Aβ deposits regarding structure [72] and ultrastructure [31]. As such, the enrichment of a shared set of proteins, most notably HTRA1, is noteworthy. HTRA1 is implicated in ATP-independent protein quality control across multiple species including removal of misfolded or mislocalized polypeptides [15]. The extensive co-localization of HTRA1 with Aβ deposits in isolated brain capillaries is in accord with this concept, although the mechanisms of HTRA1 recruitment remain to be identified. HTRA1 accumulation related to vascular Aβ pathology has previously been reported on a qualitative level in human brain vasculature obtained by microdissection [25] and in vessel-enriched fractions from Tg-SwDI mice [52] and TgDI rats [51]. Indeed, a recent review of available literature proposed there is an overlap between the profiles of CAA and CADASIL, which, however, was not formally assessed [71].
The majority of highly enriched proteins shared by our CAA type 1 and CADASIL profiles also accumulate in brain microvessels of Htra1 knockout mice [34,72] and thus represent putative HTRA1 substrates. Accordingly, HTRA1-mediated processing of a number of these proteins has been demonstrated by us and others in in-vitro cleavage assays [1,9,14,44,64,72]. We now expand on this list by demonstrating HTRA1-dependent processing of APCS and PRSS23 providing further support for an important role of HTRA1 activity in SVD pathogenesis. Several observations suggest that the accumulating proteins we found here to overlap between the CAA type 1 and CADASIL profiles and to possibly represent a HTRA1 loss-of-function signature contribute to the pathophysiology of CAA type 1 and are not a mere reflection of vessel degeneration. First, inactivating mutations in HTRA1 have previously been shown to cause degenerative changes in brain microvessels [23,47,54] thus placing loss of HTRA1 function upstream of vessel damage. Second, we did not find these proteins to accumulate in brain microvessels from patients with sporadic SVD, which likewise show degenerative changes (unpublished results). Third, some of the HTRA1 substrates including TIMP3 have previously been shown to contribute to SVD pathogenesis [10,11]. Still, the precise mechanisms linking HTRA1 to vascular degeneration in CAA type 1 remain unknown. Notably, a recent large-scale sequencing study in the general population has uncovered an association of rare loss-of-function HTRA1 variants with white matter lesion volume, a quantitative marker of cerebral SVD, indicating a potential role of HTRA1 also in sporadic SVD forms [37].
In light of our findings, we propose a refined model for the molecular function of HTRA1. We hypothesize that, under physiological conditions, HTRA1 degrades excess proteins thus safeguarding extracellular proteostasis. Support for this comes from studies demonstrating HTRA1-mediated degradation of oligomeric/fibrillar Aβ and tau species [19,49,56] and from previous findings showing an accumulation of multiple constituents of the vascular ECM in Htra1-deficient mice [34,72]. In CAA type 1 and CADASIL, HTRA1 is recruited into the vascular Aβ and Notch3 ECD deposits characterizing these vasculopathies. We propose, this occurs to remove misfolded and excess polypeptides. Progressive growth of the deposits may result in a sequestration and depletion of HTRA1 from the extracellular environment leading to functional inactivation, substrate accumulation and eventually disruption of proteostasis [22]. While direct evidence for this concept remains to be presented, we would argue that the accumulation of multiple HTRA1 substrates supports our hypothesis. Upon first sight, HTRA1-related genetic cerebral SVDs lacking distinct protein deposits seem to be at odds with this concept. Of note, however, brain vessels of these patients exhibit prominent abnormalities in angioarchitecture including intimal thickening and ECM expansion [23,47] consistent with altered proteostasis. An alternative explanation for the vascular structural alterations would be dysregulated TGFβ signaling [23,29,39]. However, we found constituents of the TGFβ pathway to be either not detected in a sufficient number of samples or abundance changes did not reach significance. A proteomic investigation of vessels from patients with HTRA1-related familial cerebral SVD was not feasible due to a lack of appropriate autopsy material.
While HTRA1 is present in human senile plaques [19], it was not enriched in brain microvessels from our AD patients. Accordingly, there was no accumulation of HTRA1 substrates that we found here to be shared between the CAA type 1 and CADASIL profile. Still, we found the CAA type 1 and AD profiles to show some overlap, evidencing shared molecular pathways independent of vascular Aβ deposition. Overall, this overlap was less pronounced in terms of the abundance ratios of individual proteins, which further belong to various cellular pathways. As such, the underlying mechanisms may be more complex.
Specific strengths of this study include the application of untargeted and quantitative proteomics to isolated microvessel preparations from a large sample