Proteomics analysis identifies new markers associated with capillary cerebral amyloid angiopathy in Alzheimer’s disease

Case selection

Fast immunohistochemistry for LCM

Sections (10 μm) of fresh frozen occipital tissue were mounted on PEN-membrane slides (Leica), air-dried and fixed in 100% ethanol for 1 min. After air drying the tissue was wetted with sterile PBS. Anti-Aβ (clone IC16, detecting N-terminal part of Aβ [16]) was applied at a 1:100 dilution in sterile PBS (pH 7.5) and incubated for 20 min at RT. After washing 3 times for 30 s in sterile PBS, HRP labelled rabbit anti-mouse (DAKO) was applied at a 1:100 dilution in sterile PBS and incubated for 15 min at RT. Sections were briefly washed (3 × 30 s) and freshly prepared 3,3′ diaminobenzidine (DAB) solution was applied and left to incubate for 5 min to visualize antibody binding. Sections were thoroughly washed in ultra-pure H₂O and incubated with 1% (w/v) toluidine blue in ultrapure H₂O for 1 min as a counterstain. Sections were then washed in ultra-pure H₂O twice for 1 min and twice in 100% ethanol for 1 min and air dried.

Brain tissue preparation and laser capture microdissection (LCM)

Laser capture microdissection (LCM) was performed as described previously [17]. LCM was performed using a Leica AS LMD system (Leica). Cortical layers II to VI which were randomly selected from control tissue and selected based on the presence of severe Aβ pathology in the case of AD and CAA were collected in Eppendorf tubes containing 30 μl M-PER lysis buffer (Thermo Scientific) supplemented with reducing SDS sample buffer (Thermo Scientific). Between 10 and 20 tissue sections with a thickness of 10 μm were captured using LCM, yielding an equal volume each of 1.0 × 10⁹ μm³. Micro-dissected tissue was stored at − 80 °C until further use.

Protein separation by electrophoresis and in-gel digestion

Micro-dissected tissue lysates were incubated at 95 °C for 5 min to denature the proteins, followed by incubation with 50 mM iodoacetamide for 30 min at RT in the dark to alkylate the cysteine residues. To reduce protein complexity, samples were size separated on a NuPAGE® 4–12% Bis-Tris acrylamide gel using MOPS SDS running buffer (Invitrogen) according to the manufacturers’ protocol.

Gels were fixed in a solution containing 50% (v/v) ethanol and 3% (v/v) phosphoric acid in H₂O for 3 h at RT and stained with Colloidal Coomassie Blue (34% (v/v) methanol, 3% (v/v) phosphoric acid, 15% (w/v) ammonium Sulphate, and 0.1% (w/v) Coomassie brilliant blue G-250 (Thermo Scientific), overnight while shaking. Destaining was performed in ultra-pure water under gentle agitation for several hours to reduce background staining (Additional file 1: Figure S1 ). Each gel lane was sliced into 12 equal sized parts to reduce sample complexity during later mass spectrometry analysis and each part was cut into blocks of approximately 1 mm³ and collected in an Eppendorf tube. Gel fragments were destained in ultrapure water with 50 mM NH₄HCO₃ and 50% (v/v) acetonitrile overnight. Gel fragments were dehydrated using acetonitrile for 20 min and dried for 30 min using a SpeedVac. The gel parts were rehydrated in 70 μl of ultra-pure water containing 50 mM NH₄HCO₃ and 10 μg/ ml trypsin (sequence grade; Promega) and incubated overnight at 37 °C to facilitate digestion of the proteins. Peptides were extracted twice with a solution containing 0.1% (v/v) trifluoric acid and 50% (v/v) acetonitrile for 20 min. The samples were dried using a SpeedVac and stored at − 20 °C until further analysis.

Mass spectrometry analysis

The peptides of the individual sample fractions were dissolved in 15 μL of 0.1% (v/v) acetic acid of which 10 μL was loaded onto a nano-liquid chromatography (nano-LC) system (Eksigent). The peptides were separated using a capillary reversed phase C18 column that had been equilibrated with 0.1% (v/v) acetic acid at a flow rate of 400 nL/min. The peptides were eluted by increasing the acetonitrile concentration linearly from 5 to 40% in 80 min and to 90% in 10 min, using the same flow rate. Eluted peptides were transferred into the LTQ/Orbitrap MS (Thermo Scientific) by Electro Spray Ionisation (ESI). The Orbitrap was operated in the range of m/z 350–2000 at a full width at half maximum resolution of 30,000 after accumulation to 500,000 in the LTQ with one microscan. The five most abundant precursor ions were selected for fragmentation by collision-induced dissociation (CID) with an isolation width of 2 Da.

Protein inference and relative protein quantification

MaxQuant software was used for spectrum annotation, protein inference, and relative protein quantification [18]. Spectra were annotated against the Uniprot human reference proteome database (version 2016_04). Enzyme specificity was set to Trypsin/P, allowing at most two missed cleavages. Carbamido-methylation of cysteine was set as a fixed modification, and N-acetylation and methionine oxidation were set as variable modifications. Mass deviation tolerance was set to 20 ppm for monoisotopic precursor ions and 0.5 Da for MS/MS peaks. False-discovery rate cut-offs for peptide and protein identifications were set to 1% for both. The minimum peptide length was seven amino acids. Identified proteins that had the same set of peptides or a subset of peptides compared to another protein, were merged into one protein group. Peptides that were shared between different proteins were assigned to the protein with most peptide evidence (so-called ‘Razor’ peptides). Only protein groups with at least a single unique and a single Razor peptide were included. For relative protein quantification MaxQuant LFQ intensities based on at least a single shared peptide ratio were used [19].

Statistical analysis of differential protein expression

To identify proteins that differ in abundance between the different experimental groups an ANOVA (Kruskal–Wallis test) was performed using the Perseus software platform [20], adhering to a significance cut-off of p ≤ 0.05. The p values were not corrected for multiple testing to include more proteins and provide a broad impression of the differences in the proteome.

Conditions that were set for inclusion of CAA selective proteins comprise of three approaches (A, B and C) that are visualized in Fig. 2. Approach A: T-tests (two-sided, assuming unequal variances, performed using Excel (Microsoft)) were performed contrasting the three experimental groups. When there was a significant difference (p < 0.05) between both the control group versus CAA, and the AD group versus CAA, a protein was labelled as CAA specific. Approach B: If the number of quantitative values in the control group was zero or one while the AD and CAA groups both had two or more quantitative values, than a t-test was performed between the AD and the CAA group. When the AD group had zero or one quantitative values while the control and CAA groups both had two or more quantitative values a t-test was performed between the CAA and control group. Approach C: In the case of zero or single quantitative values in both the control and AD groups, proteins were included based exclusively on a minimum of four quantitative values in the CAA group. Also, we included proteins with zero or single quantitative values in the CAA group and four or more values in both the AD and control groups.

ANOVA (Kruskal–Wallis test) and posthoc Dunn’s multiple comparison tests on immunoblot data and immunohistochemical data was performed using Graphpad Prism (GraphPad Software).

Immunoblot analysis

Immunohistochemical analysis

Fresh frozen or paraffin embedded human occipital tissue was cut (5 μm). For frozen tissue the sections were placed on a SuperFrost Microscope Slide (VWR, PA, USA) and air-dried overnight at room temperature (RT). Prior to staining, the sections were fixed in 100% acetone for 10 min. For paraffin sections the paraffin was removed by washing in xylene. Next, the sections were washed in decreasing concentrations of ethanol (100%, 96% and 70% (v/v)). Endogenous peroxidase activity was quenched by incubating in methanol with 0.3% H₂O₂ for 30 min at RT. Next, antigen retrieval was performed by submerging the slides in citrate buffer (pH 6) and heating in an autoclave.

Primary antibodies were diluted in antibody diluent (VWR) and incubation was performed overnight at 4 °C. All primary antibodies and corresponding dilutions used are listed in Table 2. After incubation the sections were thoroughly washed in PBS (pH 7.4) for 30 min followed by incubation of an HRP-labelled secondary antibody, Envision (DAKO) for 30 min. Again, the sections were thoroughly washed in PBS (pH 7.4) for 30 min and then incubated with DAB to visualize antibody binding. Counterstaining of the nuclei was performed by incubation in hematoxylin for 3 min followed by extensive washing in running tab water for 5 min. Next, the slides were dehydrated by incubation in increasing concentrations of ethanol consisting of 70% (v/v), 96% (v/v) and 100% (v/v) ethanol. The slides were then incubated in xylene and mounted using Quick-D mounting medium. A negative control was made by omission of the primary antibody. Quantification of the staining was done using ImageJ using the threshold colour plugin.

Selection of cases, controls and analysis of brain tissue

Three groups with a total of 20 cases were assembled based on careful neuro-pathological inspection: 1) cognitively healthy control cases (n = 6) without any Aβ pathology or tau pathology, 2) AD cases with severe Aβ plaque pathology but no vascular deposits (no CAA) (n = 7) and 3) AD cases with severe nearly pure CAA type-1 pathology and a negligible amount of plaque pathology (n = 7). From here, these groups will be mentioned as “control”, “AD” and “CAA”, respectively. Inclusion of these cases was done based on histochemical analysis using Congo-red and additional IHC for Aβ on the occipital frozen tissue intended for LC-MS-MS analysis.

We focussed our analysis on the occipital lobe as this region is the most frequently and severely affected by CAA pathology. Tissue sections of human occipital lobe from all selected cases were mounted on PEN-foil slides and Aβ pathology was visualized using fast immunohistochemistry. Grey matter tissue was isolated using LCM. Tissue isolation from the AD cases and CAA cases was focused on occipital lobe grey matter areas with severe Aβ pathology, i.e. high plaque load or high CAA type-1 burden, respectively. This was done to selectively enrich the input material for the proteomics analysis for these types of Aβ pathology. For control cases occipital lobe grey matter areas from the same anatomical region were selected for isolation. LCM-collected tissue samples were lysed and proteins were separated using SDS-PAGE. Each PAGE sample lane was divided into 12 fractions and subjected to in-gel trypsin digestion (Fig. 1).

Protein quantification and global protein expression profiles

To identify and quantify proteins, liquid chromatography followed by mass spectrometry (LC-MS-MS) was performed on the 20 laser-dissected tissue samples. This allowed quantification of 2427 proteins in total and approximately 1500 proteins identified per individual case (Additional file 3: Figure S3), with a minimum of one tryptic peptide detected. All quantified proteins are listed in Additional file 4: Table S1.

To gain insight into the global similarities and differences between the three groups and the individual cases an ANOVA (Kruskall Wallis) was performed. This yielded 309 proteins that have a significant difference (p < 0.05) in abundance between any of the experimental groups. Using these proteins in an unsupervised clustering analysis, three different expression signatures were obtained. The protein expression signatures of the AD and CAA groups appeared largely similar, whereas both were different from the control group (Additional file 5: Figure S4A). Unsupervised clustering analysis of the individual cases using the 309 ANOVA-identified proteins, separated the controls from the disease cases (Additional file 5: Figure S4B). The CAA and AD cases were not separated on the basis of the full set of differentially expressed proteins indicating that overall their protein expression profile is largely similar. One CAA case (case #5) clustered with the control cases indicating that the protein expression profile of this sample is more similar to the control cases than to other CAA or AD cases. Visualizing the expression profile of case #5 next to the average expression profiles of the three groups confirmed the resemblance of case #5 to the control group, but also showed several proteins that are similar in expression to the AD and or CAA groups (Additional file 6: Figure S5A).

Identification of proteins selectively altered in CAA type-1

After calculating the multiple testing corrected false discovery rate (FDR) only CLU was considered significant. This is likely due to the relatively low sample size of this exploratory study and the high inter-individual variance that is inevitably associated with the use of human tissue. Given the explorative nature of this study we relaxed criteria and adhered to the uncorrected p-values for protein inclusion.

Importantly, using label-free mass spectrometry to identify and quantify proteins, the absence of data for a number of proteins is observed. Despite great improvements in the speed and sensitivity of MS analysers missing data is almost unavoidable. When quantitative data are absent in one group while being present in the other group(s), this likely indicates differences in abundance, which might represent interesting candidate marker proteins. Therefore, absence of data in one or more patient groups required 2 additional approaches to also consider these proteins in this study. An overview of the 3 complementing strategies for protein inclusion is shown in Fig. 2 and a complete description is present in the methods section. Note that any given protein is only considered using a single approach as these approaches are mutually exclusive.

Using approach B (Fig. 2), proteins with a significant difference were included, and APP, UBLCP1, SRI, NDP, PNP, C1orf123, DHX15, SYNPO, TPM1, CADPS2 and SERPINA3 (Fig. 3b and Table 3b) were identified as proteins selectively present in CAA type-1. Peptide data on APP indicates that quantification was based on two peptides in which the most abundantly detected peptide (LVFFAEDVGSNK) is part of Aβ.

Approach C (Fig. 2) resulted in the identification of HLA-DRA, HLA-DQA2, HTRA1, APCS, COL6A2, MOB2, POTEI, KIAA1468, TMF1 and SGIP1 (Fig. 3c and Table 3c) as CAA specific proteins.

Earlier, Case #5 was identified as having an expression profile resembling a control case. Case #5 was found positive for Alzheimer type 2 astrocytes, possibly related to high alcohol intake, and exhibited relatively low tau pathology. Otherwise, this case showed no pathological abnormalities when compared to the rest of the CAA type-1 group. However, the expression of several CAA selective markers that we identified was inspected for case #5. The levels of these markers correspond well with the other cases of the CAA group (Additional file 6: Figure S5B), indicating that these proteins are inseparably linked to the pathology of CAA type-1. In addition, although the number of cases is too small to do valid statistics, we observed no clear relation between gender and expression of the markers (Additional file 7: Figure S6).

To determine whether the above-described approaches were indeed appropriate in selecting CAA specific proteins, we performed additional immunoblotting and immunohistochemical (IHC) analysis.

Confirmation of MS data using immunoblotting and immunohistochemical analysis

Of the proteins described in Table 3 we selected APOE (approach A), NDP (approach B), HTRA1, APSC and COL6A2 (approach C), based on the fold change or specific expression in the CAA type-1 group compared to the AD and control groups, to confirm our mass spectrometry results. Immunoblotting was performed on whole tissue lysates of the same cases as used for the mass spectrometry analysis. When comparing the CAA group with the control group we found significant differences in NDP expression (Fig. 4). For APOE, APCS and COL6A2, the data showed the same trend of increased abundance in the CAA group as the proteomics data, but the differences did not reach significance. A likely explanation for this is the higher variation of expression of these proteins in the tissue used for immunoblotting, which in contrast to the mass spectrometry exploratory analysis, was not selectively enriched for pathological burden using LCM, and instead included white matter, leptomeningeal vessels and grey matter with a lower pathological burden. To unequivocally demonstrate CAA related expression, we turned to IHC analysis of these same proteins, which, in contrast to immunoblotting, allows region specific analysis similar to the LCM-LC-MS-MS analysis. For this a separate cohort was used consisting of cognitively healthy control cases (n = 8) without any Aβ or tau pathology, 2) AD cases with severe Aβ plaque pathology but no vascular deposits (no CAA) (n = 8) and 3) AD cases with severe CAA type-1 pathology (n = 6).

First Aβ pathology was visualized and its presence was confirmed in AD and CAA type-1 cases showing plaques and vascular Aβ pathology, respectively (Fig. 5b and c). Then, IHC analysis was performed to gain information on the localization of the selected proteins. IHC for NDP showed pronounced immunoreactivity in CAA type-1 cases that appeared associated to the vasculature. NDP immunostaining in CAA, appeared to be associated with both compact Aβ depositions as well as more diffuse staining in the parenchyma in cases that exhibit dyshoric Aβ deposits. Staining was more pronounced related to capillaries compared to larger vessels. The AD cases with plaques were nearly devoid of immunoreactivity, controls did not show any immunoreactivity for NDP (Fig. 5d-f). Different antibodies against NDP showed similar results (data not shown).

COL6A2 IHC showed some immunoreactivity in control and AD cases which was restricted to leptomeningeal vessels (Additional file 8: Figure S7) and a few large vessels in the brain tissue. In CAA type-1, immunoreactivity for COL6A2 was highly increased and includes brain capillaries and larger vessels. Immunoreactivity was mostly associated with the endothelium and/or the adventitia (Fig. 5g-i). Similar results were obtained using two different antibodies for COL6A2 (data not shown). HTRA1 IHC showed clear overlap with Aβ in both AD and CAA. Compact and diffuse staining was observed related to the vessels in the CAA cases and showed plaque pathology in the AD cases without CAA. Control cases were all negative for HTRA1 (Fig. 5j-l). IHC for APOE resulted in pronounced staining of the vasculature in CAA cases and appeared related to compact deposits as well as more diffuse dyshoric deposits. Also immunoreactivity of APOE was observed in the AD cases related to the Aβ plaques, although the staining was less intense than that related to the vascular amyloid in the CAA cases (Fig. 5p-r). APCS IHC illustrated the presence of this protein in relation with both diffuse and compact Aβ pathology in both the CAA and AD group. However, staining related to the plaque pathology was less intense than that related to the vascular Aβ pathology (Fig. 5m-o).

For quantification of the IHC, images were obtained at sites that, for the AD and CAA cases, had high Aβ pathological burden in nearby sections of the same tissue block. The percentage of positively stained pixels over a total of 5 images from each case was determined. Although this method is semi-quantitative, it allowed a region-specific analysis in line with the tissue obtained by laser capture dissection that was at the basis of the original mass spectrometry analysis. Using IHC quantification we observed strongly increased immunoreactivity in CAA type-1 cases for NDP and COL6A2, and moderate increases for APCS, HTRA1 and APOE, confirming the MS results (Fig. 6).

Specificity in relation to other small vessel diseases

To assess the specificity of Aβ, NDP, COL6A2, APCS and APOE in relation to other small vessel diseases we performed additional IHC on cases that present various types of vascular defects, i.e., cotton wool plaque pathology, prion CAA, cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), hypertension related small vessel disease and Cathepsin A-related arteriopathy with strokes and leukoencephalopathy (CARASAL). IHC was performed on sections that exhibited the relevant pathological characteristics of each disease including an additional CAA type 1 case. Immunoreactivity for Aβ, NDP, COL6A2, APOE and APCS was confirmed in the CAA type-1 case (Fig. 7a-e). Immunoreactivity for these proteins was also assessed and confirmed in case exhibiting hereditary cerebral haemorrhage with amyloidosis Dutch type (HCHWA-D), which is a heredity form of CAA-type-1 (Additional file 9: Figure S8).

In tissue with cotton wool plaques the pathology was confirmed using IHC for Aβ displaying pathology around capillaries and larger vessels with dyshoric changes extending deep into the parenchyma (Fig. 7f). Intense NDP immunoreactivity was observed related to capillaries and the dyshoric changes and to a lesser extend directly lining the larger vessels. COL6A2 showed intense immunoreactivity lining the capillaries and larger vessels (Fig. 7g). IHC for APOE and APCS presented a highly positive, staining that appeared similar to the Aβ staining.

The PrP-CAA tissue was confirmed negative for Aβ pathology (Fig. 7k) and showed positive for Prp (data not shown). NDP immunoreactivity was highly increased and localised to the affected vessels. NDP staining was more pronounced around affected capillaries than at affected larger vessels. COL6A2 immunoreactivity was clearly present around capillaries and larger affected vessels, of which the vascular pathology was clearly observed using haematoxylin. Although co-occurring in largely the same vessels COL6A2 did not generally co-localize with the deposits, but instead localized more internally as a component of the basal membrane. APOE and APCS were also highly present and co-localized with the Prp deposits.

In the white matter no immunoreactivity was seen for Aβ or any of the marker proteins (Fig. 7p-t) in the control case. The CADASIL case showed characteristic pathology in the white matter, thickened vessel walls, in the white matter as is common with this condition. No Aβ pathology was present (Fig. 7u) Mild immunoreactivity for NDP was present and of the assessed proteins COL62A staining was most pronounced. APOE and APCS also displayed mild immunoreactivity related to the affected vessels (Fig. 7v-y).

In hypertension related small vessel disease the presence of COLA6A2 was most prominent while immunoreactivity for NDP, APOE and APCS was low but present (Fig. 7z-ad). Interestingly IHC analysis of the CARASAL cases showed only the prominent presence of COL6A2 in affected vessels while NDP, APOE and APCS were absent.