Background
The amyloid-β peptide (Aβ) is a key pathological effector of Alzheimer’s disease (AD) [
1]. Aβ is a short polypeptide generated by processing of a larger type I transmembrane spanning glycoprotein, the amyloid precursor protein (APP), through the successive action of proteolytic enzymes called β-secretase and γ-secretase [
2,
3]. APP can undergo alternative proteolytic processing [
4]; indeed in the main pathway APP is cleavage by α-secretase within the Aβ domain, precluding Aβ formation [
5]. Several members of membrane-bound disintegrin metalloproteinase (ADAM) family have been proposed as α-secretases, mainly ADAM10, ADAM17 (TACE), and ADAM9 [
6], but other ADAM family members, such as ADAM8, may also cleave APP [
7]. However, convincing evidence, particularly data from in vivo studies [
8,
9], indicates that ADAM10 is the enzyme acting as the main physiologically relevant α-secretase [
10].
The major neuronal β-secretase, the beta-site APP cleaving enzyme 1 (BACE1; [
11] is present in CSF [
12] in a soluble and truncated form, and increased β-secretase activity and BACE1 protein levels have been investigated as biomarkers for AD [
13‐
16]. The presence in CSF of γ-secretase components, and particularly components of the catalytic subunit presenilin-1, have also been assessed recently as AD biomarkers [
17,
18]. However, to our knowledge, only ADAM17/TACE activity has been assessed in both CSF [
19] and plasma [
20,
21]; while the potential of ADAM10 as an alternative AD biomarker has so far only been investigated in platelets [
22,
23] and other blood cells [
24]. ADAM proteases, similar to BACE1, are type I transmembrane proteins, but also include secreted isoforms [
6]. Indeed, ADAM10 and ADAM17 have been shown to be secreted outside cells in exosomes [
25]. Recently, an in-depth analysis of the human CSF endopeptidome enabled identification of several ADAM10 peptides [
26].
In this study, we investigated the occurrence of ADAM10 in human CSF and whether altered levels of this protein occur in AD. We have characterized the full-length and truncated forms of ADAM10 in CSF, as well as immature forms of the protein that need to be taken into consideration for the design of an appropriate strategy for development of further assay approaches. We report that the full-length and truncated forms of ADAM10, but not the immature forms, decrease in AD CSF compared to control CSF.
Methods
Patients
CSF samples were obtained from the Clinical Neurochemistry Laboratory (Mölndal, Sweden) from patients who sought medical advice because of cognitive impairment. In total, 27 patients with AD (7 men and 20 women, mean age 71 ± 1 years) and 26 age-matched non-AD controls (NADC; 18 men and 8 women, mean age 70 ± 2 years) were included. Patients were designated as AD or NADC according to CSF biomarker levels using cutoffs that are > 90% specific for AD: total tau (T-tau) > 400 ng/L, P-tau > 60 ng/L and Aβ42 < 550 ng/L [
27]. For details about classical CSF biomarker levels see Table
1. All AD patients fulfilled the 2011 NIA-AA criteria for dementia [
28]. No other clinical data were available for the subjects. The CSF samples used for the present study were de-identified leftover aliquots from clinical routine analyses, following a procedure approved by the Ethics Committee at University of Gothenburg. This study was also approved by the Ethics Committee at the Miguel Hernandez University.
Table 1
Demographic data and classic CSF biomarker levels
NADC | 70 ± 2 [55–88] | n = 26 (8F/18M) | 773 ± 29 [1010–561] | 238 ± 13 [138–365] | 36 ± 2 [21–51] |
AD | 71 ± 1 [55–86] | n = 27 (20F/7M) | 414 ± 15* [544–251] | 689 ± 48* [1420–443] | 88 ± 5* [164–61] |
Cell cultures
For obtaining conditioned cell-culture medium CHO cells (450,000 cells/well) were grown in six-well plates for 48 h in Dulbecco’s modified Eagle’s medium (DMEM) plus GlutaMAX™ (Gibco® Life Technologies, Paisley, UK) supplemented with 5, 1 or 0.5% fetal bovine serum (FBS; Gibco) and 100 μg/mL penicillin/streptomycin (Gibco). After 48 h, the cell medium was recollected, centrifuged for 15 min at 1500×g at 4 °C, and frozen for future analysis.
Western blotting and biochemical measurements
Samples of CSF (30 μL) and cell medium (20 μL) were denatured at 98 °C for 5 min and resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions. Following electrophoresis, proteins were blotted onto nitrocellulose membranes (Bio-Rad Laboratories GmbH, Munich, Germany). Bands of ADAM10 immunoreactivity were detected using an antibody specific for the mid-region of ADAM10 (hereafter referred to as the ectodomain antibody; rabbit polyclonal; OAGA02442, Aviva Systems Biology, San Diego, USA), an anti-N-terminal (rabbit polyclonal, ab39153, Abcam, Cambridge, UK), and anti-C-terminal ADAM10 antibody (rabbit monoclonal; ab124695, Abcam). Blots were then probed with the appropriate conjugated secondary antibodies, and imaged on an Odyssey Clx Infrared Imaging System (LI-COR Biosciences, Lincoln, NE, USA). Band intensities were analyzed using LI-COR software (Image Studio Lite). All samples were analyzed at least in duplicate. Ponceau staining served to monitor potential loading inaccuracies in individual blots. Immunoreactive ADAM10 signal for each band was normalized to the immunoreactivity of the corresponding band from a control CSF sample (aliquots from the same sample), resolved in all the blots. For the estimation of the (50 + 55 kDa)/80 kDa ratio for each sample (see the “
Results” section), the unprocessed immunoreactivity for each band was considered.
Serum and CSF albumin concentrations were measured by immunonephelometry on a Beckman Image Immunochemistry system (Beckman Instruments, Beckman Coulter). The CSF/serum albumin ratio was calculated as CSF albumin (mg/L)/serum albumin (g/L). CSF albumin levels were also estimated by western blotting using a rabbit polyclonal antibody (SAB2100098, Sigma-Aldrich, Saint Louis, USA).
Sucrose density gradient ultracentrifugation
ADAM10 complexes were fractioned by ultracentrifugation at 250,000×g on a continuous sucrose density gradient (5–20%) for 4 h at 4 °C in a Beckman TLS 55 rotor. CSF aliquots (65 μL) were carefully loaded onto the top of the gradient containing 2 mL of 0.15 M NaCl, 50 mM MgCl2, and 0.5% Brij 97 in 50 mM Tris-HCl (pH 7.4). After centrifugation, ~ 14 fractions were collected gently from the top of the tubes. Enzyme markers of known sedimentation coefficient, β-galactosidase, catalase, and alkaline phosphatase were used in the gradients to determine the approximate sedimentation coefficients.
Measurement of T-tau, P-tau, and Aβ42 by ELISA
Total tau (T-tau), phosphorylated tau (P-tau), and Aβ1–42 (Aβ42) concentrations in CSF were measured using INNOTEST ELISA methods (Fujirebio Europe, Gent, Belgium).
Statistical analysis
All the data were analyzed using SigmaStat (Version 3.5; Systac Software Inc.) using a Student’s t test (two-tailed) or a Mann-Whitney U test for single pairwise comparisons, and determining the exact p values. The results are presented as means ± SEM, and the correlation between variables was assessed by linear regression analyses.
Discussion
There is a need to identify additional CSF biomarkers of AD. The knowledge that APP metabolism and Aβ production and aggregation are key steps in AD pathogenesis makes proteins involved in the pathological processing of APP, including secretases such as ADAM10, reasonable candidates for analysis in CSF. However, since secretases are transmembrane proteins, their assessment in CSF was not considered until recent years.
Previous studies have revealed that, in addition to proteins, CSF contains many endogenous peptides [
38,
39], including ADAM10 peptides [
26]. In this study, we demonstrate the presence in human CSF of the mature and immature full-length ADAM10 protein, as well as a membrane cleaved large fragment (sADAM10). As sADAM10 can be released by proteolytic processing from the membrane [
29], this suggests the potential for truncated isoforms to be present in CSF. Indeed, recent reports indicate the possibility that ADAM10 levels can even be measured in human serum by an enzyme-linked immunosorbent assay (ELISA; [
40,
41].
In our previous study [
26] using LC-MS analysis, we identified several short peptide fragments of ADAM10 in human CSF, matching sequences located at the N-terminus of the protein as well peptide fragments located close to the transmembrane domain of the protein. In this study, several different molecular mass bands of ADAM10 were detected by western blot analysis using N- and C-terminal anti-ADAM10 antibodies. Thus, in addition to a sADAM10 isoform attributed to the immunoreactive band of ~ 50 kDa molecular mass, other ADAM10 species retaining the intracellular C-terminal domain are present in the CSF. Moreover, as some of the sequences identified by LC-MS analysis were homologous to the N-terminal prodomain, this indicated that, unexpectedly, immature proADAM10 also reached the CSF. Thus, other full-length isoforms of the protein co-exist in the CSF with sADAM10. The presence of proADAM10, together with ADAM10f, has been described at the cell surface [
10].
The mechanisms by which these membrane-resident ADAM10 species reach the CSF are unknown, but neuronal death may be a contributing factor. Interestingly, proADAM10 and ADAM10f were also detected in culture media from CHO cells. Abundant ADAM10 has been found in exosomes of bovine endometrial stromal cells cultured at hypoxic conditions [
42]. Thus, an exosomal contribution of ADAM10-CSF cannot be discounted. Interestingly, ADAM10 is also enriched in synaptic vesicles [
43], being one of many synaptic proteins identified and measured in CSF [
44]. In this context, we and others have reported evidence of the presence in the CSF of “unprocessed” forms of several transmembrane proteins, such as BACE1 [
34], APP [
35,
45], and the multi-pass presenilin-1 (PS1) [
18,
34]. Thus, the existence of a membrane-resident protein in CSF is not an unusual finding [
46]. Recently, we also characterized in CSF the existence of C-terminal fragments of APP, which include the transmembrane domain [
47].
The occurrence in CSF of proteins which still maintain their transmembrane and intracellular domains is also relevant for the development of strategies for their quantitative estimation. ADAM10, similar to many other transmembrane proteins exists as a dimer in the brain [
33]. Both the transmembrane [
48] and cytoplasmic [
33] domains can participate in dimerization of ADAM10, a feature that may be is an inherent property of ADAM metalloproteinases. In the present study, we demonstrated by gradient centrifugation that sADAM10 and ADAM10f are present in the CSF as large complexes. Further studies will be necessary to clarify the biochemical properties of these homomeric complexes, but our preliminary analysis indicates that the species in NADC CSF are similar, if not identical to the species in the AD CSF. We have previously demonstrated the occurrence of APP heteromers in CSF, comprising both sAPPα/sAPPβ and also soluble full-length APP, and we have shown that these heteromers affect the determination of sAPP by ELISA [
35]. Given that the distinct ADAM10 species also form complexes, the development of an accurate ELISA protocol for the estimation of CSF-ADAM10 levels may require more knowledge about the potential variable stoichiometry and stability of these complexes. In fact, our early attempts to assess ADAM10-CSF levels by ELISA have resulted in poorly reproducible data (~ 60% intra-assay variability in CSF samples; ELISA kit from MyBioSource, Inc. San Diego, CA, USA). A previous study also reported difficulties in assessing ADAM10 in CSF, discarding their presence in the fluid [
49]. In this study, to circumvent this issue, we analyzed ADAM10-CSF levels by SDS-PAGE.
Our determination of the different species of ADAM10 in CSF by western blotting indicated that, in AD cases, there is a decrease in sADAM10 and ADAM10f, but not in the immature forms. Since amyloidogenic processing of APP is expected to be altered in the Alzheimer brain, parallel changes in the levels of α-secretase and β-secretase might be expected. However, it is still unclear if α-secretase and β-secretase are inversely correlated during pathological progression, as the proteolytic products sAPPα and sAPPβ displayed similar trends in the CSF [
45]. Data on ADAM10 in human brain are scarce, but the majority of the data indicate an overall decrease in ADAM10 mRNA, protein, and/or activity in the brain of AD patients compared to age-matched controls [
50]. However, at least in platelets, the decrease of ADAM10 protein in AD patients is not caused by a reduction in ADAM10 mRNA [
51]. Thus, the regulation of expression and activity of ADAM10 may be complex, being regulated by several pathways, epigenetically, and at translational and post-translational levels [
50], and affected by normal aging [
30]. In this context, it may be important to evaluate α-secretase activity in CSF. Enzymatic activity assays in CSF are usually based on the use of specific substrates (synthetic peptides) favorable for the assessment of a concrete activity, but as mentioned previously, other enzymes, in addition to ADAM10, display α-secretase-like activity. Indeed, elevated activity levels for ADAM17/TACE have been found in both CSF [
19] and plasma [
20,
21] from subjects with AD. It appears important to decipher the physiopathological significance of the differential regulation in AD for ADAM10 and ADAM17. Anyhow, in the view of the apparent increase in ADAM17 levels and decrease in ADAM10 protein levels in AD CSF, it is questionable whether an enzymatic activity assay for ADAM10 in CSF, based in the use of synthetic peptides (which can be cleaved by multiple proteases), should be used to measure changes in the CSF α-secretase activity of AD patients. The general requirements for secretase cleavage are not strict, and we cannot exclude the possibility that other CSF enzymes that may cleave the synthetic peptides also being detected. Moreover, emerging evidence indicates that the plasma membrane with its unique dynamic properties may additionally play an important role in controlling sheddase function, as physicochemical properties of the lipid bilayer that govern the action of ADAM-proteases [
52]. Accordingly, determination of enzymatic activities does not appear to be the most adequate and sensitive molecular tool to evaluate ADAM10, and other secretases, as a potential CSF biomarkers. Therefore, only ELISA assays based on pan-specific antibodies for concrete ADAM10 species, and including pre-treatment methods designed to disaggregate complexes, may be a reliable approach to assess ADAM10 protein levels and enzymatic activity.
Despite the limited precision of western blotting for quantitative analysis, we consider that mature forms of ADAM10 in CSF constitute potential new biomarkers of AD. The western blotting technique is also an important limitation regarding the number of cases to analyze. A corroborative study in a large and independent cohort should be necessary to assess the potential of this new biomarker for AD diagnosis. New studies should also address analysis of aging in normal conditions, adequate balance of gender in the pathological group, and more importantly, the specificity of the changes for AD confronting other dementias. In this study, we defined neurochemically AD and NADC subjects based on the levels of classical AD biomarkers. The NADC subjects included in this study also exhibited cognitive impairment; therefore, the control group is not totally comprised of healthy individuals and probably includes several non-AD conditions. Despite the inherent difficulty of the clinical diagnosis to truly assess the potential of new diagnostic biomarkers, tentative studies including other dementias clinically diagnosed will be of particular relevance, especially with longitudinal follow-up and establishment of definitive diagnosis. The involvement of ADAM10 in other pathological processes such as traumatic brain injury, inflammation, brain tumors, stroke, and psychiatric diseases [
48,
53] also deserves the analysis of CSF-ADAM10 levels.