Dense core vesicle markers in CSF and cortical tissues of patients with Alzheimer’s disease

Post-mortem human AD (n = 23) and non-AD (n = 22) brain samples (aged 49–86) were obtained from the Institute of Neuropathology Brain Bank IDIBELL Hospital Universitari de Bellvitge (Hospitalet de Llobregat, Spain) under an agreement with the local ethics committee; demographic data are presented in Table 1. Individuals were selected based on the post-mortem diagnosis of AD according to the ABC score (A, Amyloid phases Thal; B, Braak stages of neurofibrillary tangle pathology; C, CERAD stages) following the National Institute on Aging-Alzheimer's Association (NIA-AA) clinical research criteria [36]. The AD-diagnosed subjects corresponded to A3-A4, V-VI (B3) and C3, whereas the non-AD subjects were A0, 0-I/II (B0-BI) and C0.

Both AD patients and control subjects (healthy volunteers) were recruited from the SPIN cohort in the Memory Unit at the Hospital de la Santa Creu i Sant Pau [37]. This study was approved by the local ethics committee and was carried out in accordance with the Declaration of Helsinki. All patients (or their nearest relatives) and controls gave informed consent to participate in the study. Extensive clinical, neuropsychological, MRI and molecular examinations were performed in all subjects. CSF samples (n = 66) were collected by lumbar puncture between 9 am and noon. Centrifugation at 4 °C for 10 min at 2000 × g and storage of 500 μl aliquots in polypropylene tubes at − 80 °C were accomplished within 1 h after collection. All AD patients fulfilled clinical criteria for probable AD according to the revised NIA-AA criteria [38] and had a CSF biomarker profile consisting of decreased Aβ_1-42 plus high T-tau and P-tau levels (ELISA tests from Innogenetics, Ghent, Belgium), indicating high likelihood of being due to AD. The cut-off values we used to define our AD cohort in this study were 550 pg/ml for Aβ_1-42, 350 pg/ml for T-tau, and 61 pg/ml for P-tau [39]. The control group was defined according to the following criteria: objective cognitive performance within the normal range (performance within 1.5 standard deviation) on all tests from a specific test battery, clinical dementia rating scale score of 0, no significant psychiatric symptoms or previous neurological disease, and a non-pathological CSF biomarker profile. The average Mini-Mental State Examination (MMSE) score was 21.6 ± 4.4 for AD patients, whereas control subjects had a score of 28 or higher. AD and control groups were well matched for age at CSF collection. Demographic information is presented in Table 2.

Home-made and commercial antibodies against DCV proteins used in this study have been extensively validated for western blot and immunohistochemical methods in previous papers and recognize processed and precursor molecular forms. Polyclonal antibodies against PC2 (LS18 kindly provided by Dr I. Lindberg, University of Maryland) and SgII/Secretoneurin have been described elsewhere [40, 41]. Polyclonal antibodies against PC1/3 were purchased from Abcam (ab3532, Abcam Plc, Cambridge, UK) and Thermo (PA1-057, Thermo Fisher Scientific, Waltham, MA) and polyclonal antibodies against PC2 were from GeneTex (GTX23533, GeneTex Inc, Irvine, CA) [42‐44]. Polyclonal antibodies against SgIII (HPA006880, Prestige Antibodies® supported by The Human Protein Atlas) [29, 45] and monoclonal anti-β-actin peroxidase antibodies (AC-15) were purchased from Sigma (Sigma-Aldrich, Diesenhofen, Germany). Monoclonal and polyclonal antibodies against CPE were obtained from BD Biosciences (610758, BD Transduction Laboratories, San Jose, CA) and GeneTex (GTX23533, GeneTex Inc), respectively [45, 46]. Polyclonal antibodies against CysC (ABC20) [47] and monoclonal antibodies against GFAP (MAB3402) were obtained from Millipore Iberica (Madrid, Spain). Monoclonal and polyclonal antibodies against CHMP2B (MAB7509) and CK1δ (AF4568) were purchased from R&D Systems (Minneapolis, MN). Monoclonal antibodies against LAMP-1 (H4A3) and neurofilament light chain (2835) were obtained from Developmental Studies Hybridoma Bank at the University of Iowa (Iowa City, Iowa) and Cell Signaling (Leiden, The Netherlands), respectively. Antibodies against Aβ (M087201-2) and hyperphosphorylated tau (AT8) were from DAKO (Glostrup, Denmark) and Innogenetics (Gent, Belgium), respectively. A complete antibody list is provided in Additional file 1: Table S1.

Human brain samples were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, by immersion. After being cryoprotected in a 30% sucrose solution, the samples were frozen and sectioned with a cryostat. For peroxidase immunohistochemistry, histological sections were blocked for 30 min in phosphate buffered saline (PBS) supplemented with 10% methanol and 3% H₂O₂ and then washed in PBS. Pretreatment with formic acid was used to enhance labeling of plaques. To avoid nonspecific binding, brain sections were blocked in PBS containing 10% serum, 0.2% glycine, 0.1% Triton X-100, and 0.2% gelatin for 1 h at room temperature. Incubations with the primary antibodies were carried out overnight at 4 °C in PBS containing 1% fetal bovine serum, 0.1% Triton X-100 and 0.2% gelatin. Bound antibodies were detected using the avidin–biotin-peroxidase system (Vectastain ABC kit, Vector Laboratories Inc., Burlingame, CA). PBS containing 0.05% diaminobenzidine (DAB) and 0.01% H₂O₂ was used to visualize the peroxidase complex. Sections were mounted, dehydrated, and coverslipped in Eukitt® (Sigma-Aldrich, Diesenhofen, Germany). Semiquantitative analysis of CysC immunoreactivity on DAB-stained slices was carried out on the outer and inner layers of the AD and control parietal cortices (n = 3, each). Images at 20 × magnification (2 per section, 2 sections per sample) were randomly taken using identical acquisition microscope parameters. The signal intensity of 0.1 × 0.1 mm² fields was measured using the ImageJ® software (NIH, Bethesda, MD). Levels of CysC were initially quantified as arbitrary density units and subsequently AD values were expressed as a percent change from control measurement. Double-labeling fluorescent immunohistochemistry was carried out by incubation with different fluorochrome-conjugated secondary antibodies (Alexa Fluor 488 and Alexa Fluor 568; Molecular Probes, Eugene, OR), and cell nuclei were stained with Hoechst (Sigma-Aldrich, Diesenhofen, Germany). Endogenous autofluorescence was avoided using Sudan Black B (Sigma-Aldrich, Diesenhofen, Germany). Sections were mounted in Mowiol (Merk Chemicals Ltd., Nottingham, UK) and visualized with a Leica TCS SPE scanning confocal microscope. Colocalization analyses of DCV proteins with GVD markers were performed by ImageJ software, using the threshold-corrected Mander’s correlation coeficient, which ranges between 1 (high-colocalization) and 0 (low-colocalization). The specificity of the immunolabeling was checked omitting primary antibodies, using nonspecific IgG instead of them and using a previous incubation of the primary antibodies with an excess of antigen (Proteintech Group Inc., Deansgate, Manchester, UK). No immunostaining was observed in these conditions.

Equal volumes of lumbar CSF samples were collected in vials and brain tissues were homogenized in ice-cold lysis buffer containing 50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 5 mM MgCl₂,1 mM EGTA, 1% Triton X-100, and a protease inhibitor cocktail (Roche Diagnostics GmbH, Mannheim, Germany). The CSF samples and postnuclear brain lysates were mixed with 3 × sample buffer (188 mM Tris–HCl, 30% glycerol, 3% SDS, 0.01% Bromophenol Blue, and 15% β-mercaptoethanol), electrophoresed in 10%–12% SDS-acrylamide gel (Bio-Rad Laboratories, Hercules, CA), and then transferred to polyvinylidene difluoride immobilization membranes (Whatman® Schleicher & Schuell, Keene, NH). Tissue results were normalized for total protein content (data obtained from Ponceau staining scans). Membranes were blocked in 5% nonfat milk powder solution in Tris-buffered saline and Tween-20 (TBST; 140 mM NaCl, 10 mM Tris/HCl, pH 7.4 and 0.1% Tween-20) for 1 h at room temperature and then incubated with primary antibodies in blocking solution overnight at 4 °C. After numerous washes in blocking buffer, the membranes were incubated for 1 h with horseradish peroxidase-conjugated secondary antibodies (Bio-Rad Laboratories, Hercules, CA). Enhanced chemiluminescence reagents ECL™ (GE Healthcare, Buckinghamshire, UK) and X-ray films (AGFA) were used to visualize bound antibodies. Blot images were taken with a scanner and densitometric values were obtained using a Java-based image processing software (ImageJ® software). The amount of total proteins in brain homogenates (10–15 µg) and CSF volume (2–5 µl) and exposure times of the films were empirically determined to obtain the adequate linear range for quantitative analysis. For protein forms with different electrophoretic mobilities, a single average exposure of the film is shown in the figures, whereas quantitative analysis of each form was performed at different and suitable amounts of total protein/volume and exposure time of the films.

The SgII/secretoneurin radioimmunoassay was performed as described previously [41]. In brief, samples were denatured to avoid protein degradation and antiserum was added in the radioimmunoassay buffer and incubated for 24 h at 4 ºC. Then iodinated secretoneurin (200 cpm/μL) was added and samples were incubated for an additional 24 h at 4 ºC. The non-bound tracer was separated by dextran-coated charcoal absorption and pulled down by centrifugation. The iodinated tracer remaining in the supernatant fraction was quantified, with a detection limit of 1–2 fmol.

Quantitative data were statistically analyzed using GraphPad Prism 6.01® software (GraphPad Software, San Diego, CA) with t-test and Mann–Whitney test. All data are presented as the means ± SEM. Significance was set at P < 0.05. Correlations of the measured values were examined using the Bonferroni-corrected Spearman correlation coefficient. A significance cutoff of P ≤ 0.0019 based on Bonferroni correction of multiple comparisons and P ≤ 0.05 in non-corrected comparisons were applied.

To examine DCV changes in the cerebral cortex of AD patients, we performed Western blotting and immunostaining analyses in homogenates and histological sections from AD patients and age-matched controls. PC1/3, PC2, CPE, SgIII and SgII were abundantly found by immunoblotting in cortical samples (Additional file 1: Fig. S1). In addition to the unprocessed and mature forms, higher and lower electrophoretic mobility bands were detected with polyclonal antibodies against DCV proteins, likely corresponding to the described aggregated and cleaved species [30‐32]. A robust detection for all of these proteins was noted in CSF samples, essentially their precursor and mature forms. Volumes as small as 1 µL were enough to detect some of these proteins in the CSF (i.e. SgIII). As previously reported, only the monomeric form of the secretory protease inhibitor CysC was detected in the CSF (~ 14 kDa), whereas cortical tissues also displayed oligomeric/aggregated species [48, 49]. When performing analyses of similar CSF volumes, levels of membrane and cytosolic proteins such as synaptophysin (Syp) and β-actin (Additional file 1: Fig. S1), as well as the potential AD biomarker neurofilament light chain (data not shown), were under the detection limit. Thus, besides the abundance in brain tissues, DCV proteins are broadly detected in the CSF, feasibly corresponding with released species.

Next, we examined the levels of distinct molecular forms of secretory proteins in AD brains by immunoblotting (n = 7 per tissue and condition) (Fig. 1 and Additional file 1: Fig. S2). Comparing DCV protein amounts (normalized to membrane-transferred total protein) in hippocampus and parietal cortex of AD cases with controls, we only found changes for the 72-kDa precursor form of PC2 (proPC2). The proPC2 species was increased 1.7-fold in parietal cortex (P = 0.038) and 2.4-fold in hippocampus (P = 0.0006). Interestingly, the ~ 28 kDa form of CysC was significantly decreased in the parietal cortex of AD patients (35%, P = 0.038). This electrophoretic band probably corresponds to dimeric or aggregated forms of CysC, which have been associated with amyloid damage [48, 49]. Finally, a prominent reduction was evident in the content of the SV-specific integral protein Syp both in parietal cortex (30%, P = 0.042) and in hippocampus (50%, P = 0.017) of AD patients. These results show that the global levels of different DCV markers are mostly preserved in AD cortices, whereas ubiquitous SV proteins are dramatically reduced.

To examine the alterations of DCV proteins in situ, we performed peroxidase immunohistochemical and confocal immunofluorescence analyses in brain sections from control (n = 8) and AD cases (n = 9). In controls, PC1/3, PC2, CPE, SgIII, SgII and CysC were widely distributed throughout the cerebral cortex, consistent with previous studies [45, 50‐52] (Fig. 2). All of these proteins were abundant in neuronal populations of the neocortex and hippocampus (Fig. 2a), although a differential and robust immunolabeling for CPE was detected filling dendritic shafts (Fig. 2a, c). Immunostaining for SgIII, CPE, and CysC was also detected in GFAP⁺ astrocytes (Fig. 2b), whereas prohormone convertases were absent in glial cells. In neurons, subcellular structures labeled for secretory proteins included somatic and dendritic granules, varicose fibers, and terminal-like buttons, as we previously reported [45]. Granular compartments immunostained for typical DCV markers were negative for the lysosomal marker LAMP1 (not shown), whereas CysC immunofluorescence was largely associated with LAMP1⁺ structures (Fig. 2c). Immunostaining alterations for secretory proteins in AD cortices were mainly associated with senile plaques (Fig. 3a). All typical DCV components were aberrantly accumulated, to different degrees, in AT8⁺  dystrophic neurites surrounding amyloid plaques, in the parietal cortex and hippocampus (in each examined AD cases) (representative examples are shown in Fig. 3b, c). In some samples, intense immunolabeling for SgIII and CysC was detected in reactive astrocytes close to the amyloid deposits (data not shown), as described elsewhere [45, 51]. No major changes were detected for secretory proteins in non-plaque areas of the AD brains, including tangle-bearing neuronal somata identified by the AT8 antigen (excepting cells showing GVD, see below). Only a slight, but consistent, increase in CysC was found in pyramidal neurons located in outer and inner layers of the AD parietal cortex (Additional file 1: Fig. S3).

Unexpectedly, we found GVD-shaped structures immunolabelled for some DCV proteins in CA1 pyramidal somata (and to a lesser extent in CA2 and subiculum) in each of the four examined AD hippocampi (Fig. 4a). These granules had strong immunostaining for PC2 and moderate immunostaining for PC1/3, whereas only one AD case displayed scarce CPE⁺ large granules (Fig. 4a). An average of 241 ± 4 CA1 pyramidal neurons containing aberrant PC2⁺ granules per mm² was observed in AD (n = 4), whereas age-matched controls were virtually devoid of these (a few cells, and in only 1 out of 5 cases). Double labeling of PC1/3, PC2 or CPE and the established GVD markers CK1δ and CHMP2B substantiated the neuropathological identity of these large granules (Fig. 4b). Mander’s correlation coefficient for CK1δ overlapping with PC2 was 0.55 ± 0.13 and for PC2 overlapping with CK1δ was 0.62 ± 0.20 (n = 12), whereas Mander’s coefficient for CHMP2B overlapping with PC1/3 was 0.43 ± 0.25 and for PC1/3 overlapping with CHMP2B was 0.54 ± 0.28 (n = 10). These values indicate good colocalization between DCV proteins and GVD bodies in the AD hippocampus. Additionally, an atypical colocalization of PC2 and PC1/3 with the lysosomal marker LAMP1 was also detected in GVD bodies (Fig. 4b). Interestingly, the PC2- and PC1/3-immunolabelled GVD inclusions were associated with neurofibrillary tangle pathology in AD samples (Fig. 4c, d). Most pyramidal neurons displaying GVD bodies positive for PC2 (81%) and PC1/3 (77%) also contained hyperphosphorylated tau (AT8⁺). Although the GVD inclusions positive for DCV proteins were detected in neurons containing mature neurofibrillary tangles, the DCV-labeled aberrant granules were more often present in early-stages of neurofibrillary tangle formation (Fig. 4c, d).

Taken together, these results show that an aberrant accumulation of DCV components is associated with two neuropathological hallmarks of the AD cerebral cortex, plaque-related dystrophic neurites and GVD bodies.

Next, we investigated the levels of DCV proteins in 66 CSF samples from phenotypically well-characterized AD patients (n = 33) and cognitively normal controls (n = 33) with immunoblotting and radioimmunoassay methods. No differences in the CSF total protein content were found between AD and controls, either by Bradford assay or Coomassie staining of SDS-PAGE gels (Fig. 5a). However, levels of most molecular forms of DCV proteins were decreased in the CSF from AD patients compared with those from cognitively normal controls (Figs. 5b and 6). Representative immunoblots for different secretory proteins of the same AD and non-AD individuals are illustrated in Fig. 5b. The most prominent reduction in AD was found for unprocessed PC1/3 (− 42% ± 5.7%, P = 0.0003). Although levels of the mature form were apparently lower in AD, they did not reach statistical significance (P = 0.075) (Fig. 6). When we quantified the levels of mature and unprocessed forms of PC1/3 together, a significant decrease was detected in the AD group (− 24%, P = 0.03). Consistently, we also detected an apparent PC1/3 cleaved form displaying a lower electrophoretic mobility identified by the polyclonal antibodies used in this study. When analyzing this molecular species together with the other two bands, we found decreased levels of total PC1/3 in the CSF from AD patients compared to controls (P = 0.05). In addition, each of the two bands detected by PC2 antibodies that represent precursor and mature forms, was statistically decreased in AD samples (− 16%, P = 0.024 and − 12%, P = 0.011, respectively; Fig. 6), as well as in a pooled quantification (− 18%, P = 0.0009). The DCV-contained exopeptidase CPE was reduced by ~ 20% in the CSF of the patient cohort (P = 0.016). Regarding members of the granin family, the most abundant form of SgIII in the CSF, proSgIII, was decreased in AD (P = 0.047), whereas no significant differences were detected for the processed protein. An analysis of the two forms showed no differences between AD and non-AD individuals (P = 0.612). We also determined the CSF levels of SgII/secretoneurin with an radioimmunoassay assay. No statistical differences were found between AD and control groups. Finally, a reduction of about 33% was observed for the secretory protein CysC in the CSF of AD patients (P = 0.001).

To investigate the possible relationships of secretory protein forms with age, cognitive status (MMSE), and the core CSF biomarkers of AD (Aβ_1–42, P-tau and T-tau), Spearman correlation analysis was carried out in the AD cohort (33 individuals) (Fig. 7). No molecular forms of secretory proteins were significantly correlated with age or Aβ_1–42 (Fig. 7). In contrast, we detected strong correlations, positive between the mature PC2 form and MMSE score (P = 0.0001, r = 0.6449) and inverse between PC2 and T-tau (P = 0.0004, r =  − 0.5837), applying Bonferroni correction. Non-corrected analysis further evidenced an inverse correlation of P-tau with PC2 and PC1/3 and T-tau with PC1/3, and a direct correlation of MMSE with PC1/3. Additionally, corrected analysis evidenced positive correlations between levels of P-tau and CPE (P = 0.0013, r = 0.5377) and the precursor form of SgIII (P = 0.0019, r = 0.5216), whereas non-corrected analysis showed a direct correlation of levels of CPE, CysC and different forms of SgIII with P-tau and T-tau (Fig. 7).