Abstract
Cerebrospinal fluid (CSF) biomarkers are currently included in the diagnostic criteria for Alzheimer’s disease (AD), in particular, decreased concentrations of amyloid-β peptide 1–42 (Aβ42) in the CSF, coupled with increased levels of tau and phosphorylated tau proteins, are supportive of AD diagnosis. To date, the quantification of Aβ42 levels with antibody-dependent immunoassay shows a marked variability among different laboratories and is also affected by different pre-analytical factors, suggesting that part of Aβ42 peptides might be aggregated and thus undetected by antibodies. To bypass an antibody-dependent measurement, we determined the Aβ40 and Aβ42 levels by immunoblot. We analyzed CSF samples from 35 patients with clinical diagnosis of probable AD and from 15 age-matched normal controls; CSF Aβ levels were determined by two different ELISA kits and by immunoblot analysis. Aβ40 levels measured by ELISA were comparable to those obtained by immunoblot, whereas CSF concentrations of Aβ42 measured by ELISA were significantly lower compared to values obtained by immunoblot quantification. Biochemical analysis, following 1D- and 2D-PAGE analysis, showed that the qualitative composition of Aβ peptides in the CSF is similar in AD and controls but different from that of AD brain tissues. Moreover, sedimentation velocity in sucrose gradient of CSF and brain homogenate from AD demonstrated that Aβ42 in CSF is different from Aβ42 in brain in terms of solubility and aggregation state.
INTRODUCTION
In Alzheimer’s disease (AD), the cerebrospinal fluid (CSF) levels of amyloid-β peptide 1–42 (Aβ42), tau protein, and phosphorylated forms of tau (ptau) have diagnostic and prognostic values. Dubois et al. included CSF biomarkers such as Aβ42, total tau, and ptau in the diagnostic criteria for AD, highlighting that the reduction of Aβ42 and the increase of tau and ptau are supportive markers for AD diagnosis [1, 2]. In addition, the levels of these biomarkers are reported to be helpful in predicting the progression of mild cognitive impairment to AD, but also in differentiating AD from other dementias with 95% sensitivity and 87% specificity [2].
Aβ peptides show a high propensity to misfolding and aggregation following a seeding-nucleation mechanism, leading to protofibrils and fibrils through the formation of several intermediates, including soluble oligomers [3].
The propensity of specific Aβ peptides to aggregate is related to their primary structure, Aβ42 is more prone to aggregation than Aβ40, given the presence of two hydrophobic residues at C-terminus. Due to these biochemical properties, Aβ42 is mainly found in brain tissue under amyloid plaques, while Aβ40 is more soluble and is deposited in the cerebral vessel walls, inducing amyloid angiopathy [3].
Aβ monomers self-associate under oligomers, which are neurotoxic and ultimately polymeric structures [4]. However, their effect on tau hyperphosphorylation remains unknown [5].
Although the levels of Aβ peptides in the CSF have a high diagnostic value, their determination did not always report univocal results across different laboratories [2]. Significant differences of Aβ42 levels between AD patients and controls were reported and standard values and cut-off points differed among studies [2]. This variability might be related to Aβ conformations and/or to the affinity of antibodies for Aβ binding in different commercially available ELISAs.
Moreover, it is unclear whether Aβ ELISAs recognize only monomeric forms or also oligomers/multimers, except for ELISAs in which monoclonal antibodies against oligomers conformational epitopes had been specifically developed [6–8].
In this study, we analyzed Aβ levels in the CSF samples from subjects with a diagnosis of AD and age-matched controls, by commercially available ELISA kits and by immunoblot.
In addition, we showed that Aβ peptides composition and biochemical properties in CSF and brain tissue from AD and controls are significantly different.
MATERIALS AND METHODS
Patients and samples
This study was approved by the ethical committee of the Azienda Ospedaliera Universitaria Integrata of Verona, Verona, Italy. Written informed consent for participation in research was obtained in accordance with the Declaration of Helsinki (1964–2008) [9] and the Additional Protocol on the Convention of Human Rights and Biomedicine Concerning Biomedical Research (2005) [10].
Patients underwent lumbar puncture into L3-L4 subarachnoid space using a non-traumatic 20 gauge needle. CSF samples were collected in polypropylene tubes, centrifuged at 1,000 g for 10 min at 4°C, aliquoted and stored at –80°C. CSF samples were obtained from 35 patients diagnosed as probable AD, based on established criteria [11], and 15 age-matched normal control patients.
Frontal cortex tissue samples from five patients with definite AD diagnosis were also analyzed.
CSF analysis of Aβ40 and Aβ42 by ELISA
Aβ1–40 ELISA was performed by Human β Amyloid (1–40) ELISA Kit by Wako (Wako Pure Chemical Industries, Ltd. Osaka, Japan) based on monoclonal antibody BAN50 (directed to 1–10 Aβ residues) which specifically detects the N-terminus of Aβ and the HRP conjugated antibody BA27 directed to the C-terminus of Aβ40 (epitope not defined ending at aa 40) as detecting antibody and Innotest β-AMYLOID (1–40) by Fujirebio (Fujirebio Europe, Gent, Belgium) based on a mouse monoclonal 2G3, recognizing 35–40 Aβ sequence, as capturing antibody and the biotinylated antibody 3D6 (directed to 1–6 Aβ residues) as detecting antibody. CSF Aβ42 was measured by the Human β Amyloid (1–42) ELISA Kit from Wako (Wako Pure Chemical Industries, Ltd. Osaka, Japan) and the Innotest β-AMYLOID(1–42) from Fujirebio (Fujirebio Europe, Gent, Belgium). Wako assay uses BAN50 (directed to 1–10 Aβ residues) as capturing antibody, and the HRP conjugated antibody BC05 (directed to 35–43 Aβ residues) as detecting antibody. In the Fujirebio kit the capturing antibody is monoclonal 21F12 (directed to 34–42 Aβ residues) and Aβ42 is detected by the biotinylated antibody 3D6 (directed to 1–6 Aβ residues).
Determination of tau and ptau in CSF samples
Total tau and ptau were measured by Innotest hTAU Ag and Innotest Phospho-Tau (181P) ELISA kits from Fujirebio (Fujirebio Europe, Gent, Belgium) according to the manufacturer’s instructions.
Sample preparation and Aβ SDS-PAGE
CSF and brain samples were prepared as described [12]. Electrophoretic separation of Aβ peptide was performed by the urea based SDS-PAGE according to the protocol of Bibl et al. [12].
Serial dilutions from 4 to 64 pg of synthetic Aβ40 and Aβ42 peptides (Bachem, Bubendorf, Switzerland) were run together with CSF samples. The correct loading of Aβ peptides was shown by the linear increase in the optical densities. Due to the higher levels of Aβ40 in the CSF, CSF samples were normalized by loading neat CSF samples for Aβ42, while for Aβ40, CSF samples were 4 to 10 folds diluted. Each CSF sample was analyzed in duplicate and Aβ quantification was determined by densitometry.
Aβ 2D-PAGE immunoblot
For isoelectric focusing, 7-cm long precast gels with a linear immobilized pH range of 3–10 were used (Bio-Rad). Dry gels were swollen for 16 h in 125 μl of DeStreak Rehydration Solution (GE Healthcare) 0.5% IPG buffer (GE Healthcare) containing 35 μl of CSF or 10 μl of 10% brain homogenate. Isoelectric focusing was carried out at 20°C for 45 min with 500 V, 45 min with 1000 V, and 3 h with 8000 V in a cooled horizontal electrophoresis unit (IPGphor, GE Healthcare).
Immobilized pH gradient strips were then equilibrated for 30 min in 6 M urea, glycerol 20% (w/v), SDS 2.0% (w/v), bistris 0.36 M, bicine 0.16 M, DTT 1.0% (w/v) and loaded onto a urea gel and run for 2 h at room temperature. Immunoblot analysis was performed as described [12].
Aβ peptides were detected by 6E10 monoclonal antibody (Covance, Princeton USA) which recognizes amino acids sequence 3–8 of the Aβ peptide.
Sedimentation velocity in sucrose gradient
180 μl of CSF or 10% brain homogenate were incubated with 20 μl of 20% Sarkosyl in TBS for 30 min in ice and loaded atop a 10–60% step (10-15-20-25-30-60) sucrose gradient in TBS 1% Sarkosyl and centrifuged at 200,000 g for 1 h at 4°C. 11 fractions of 180 μl were collected from the top of the tube and prepared for immunoblot analysis.
Statistical analysis
Aβ42, Aβ40, tau, and ptau data are presented as mean±standard deviation of the mean and values range. To evaluate differences among the three methods, ELISA assays and immunoblot analysis, data were analyzed by one-way analysis of variance (ANOVA) and Tukey’s post-hoc analysis. Statistical differences between the results obtained in AD and controls using the same measurement method, were analyzed by unpaired t-test. A “p” value of <0.01 was accepted as significant.
All the analyses were performed using SPSS statistical software.
RESULTS
Determination of biomarker levels in CSF samples from AD patients and controls
Values of CSF Aβ42, total tau, and ptau, measured in all AD patients and controls by commercial Fujirebio ELISA kits, are reported in Table 1.
CSF biomarkers determination in Pathological (P) and control (C) patients
In CSF samples from AD patients the mean levels of Aβ42 were 296±132, 712±485 pg/ml for tau, and 72±38 pg/ml for ptau, while in controls mean levels of Aβ42 were 693±195 pg/ml, 251±39 pg/ml for tau and 28±12 pg/ml for ptau. In five CSF samples an overlap of Aβ42 levels was observed between AD and controls. Therefore, we also determined the Aβ42/ptau ratio, to more clearly distinguish controls from AD [2], as reported.
Mean Aβ/ptau ratio was 4.1±2.3 in CSF samples from AD, while control samples showed mean Aβ42/ptau ratio of 27.3±9.9. Among all analyzed samples, the highest value of Aβ/ptau ratio was 9.1 for AD while all control CSF samples were above Aβ42/ptau ratio of 14. Although in 8 CSF samples from AD patients Aβ/ptau were above the reported diagnostic cut off of 7, we did not observe any overlapping between AD and controls in Aβ/ptau ratio [2].
Determination by ELISA and immunoblot analysis of CSF Aβ42 and Aβ40 levels
To compare the sensitivity between commercially available ELISA kits, we tested the same CSF samples for Aβ42 by Fujirebio ELISA and by Wako ELISA kit, which uses well characterized and fully specific antibodies [14].
Wako ELISA and Fujirebio ELISA kits showed mean Aβ42 levels of 221±134 (range 29–597) pg/ml and 415±239 (range 39–1023) pg/ml respectively (Tables 1 and 2).
Mean of total Aβ42 values determination by immuno-blot method and commercially available ELISAs Fujirebio and Wako
ap Value between immunoblot analysis and Fujirebio ELISA is <0.01.
b p Value between immunoblot analysis and WAKO ELISA is <0.01.
c p Value between WAKO ELISA and Fujirebio ELISA is not significative.
To overcome the possibility that Aβ peptides are present in the CSF under different states of aggregation, a factor influencing the detection of Aβ, we determined the levels of Aβ peptides in CSF following samples denaturation and immunoblot analysis. The relative concentration of Aβ was determined by loading different amounts of Aβ peptides (Fig. 1).
Immunoblot with anti-Aβ antibody 6E10 of some CSF samples and serial dilution of recombinant Aβ42 (A) and Aβ40 (B) for western blot quantification.
Following solubilization of CSF samples with SDS and 8M urea, immunoblot analysis revealed Aβ42 mean levels of 1286±1043 (range 162–4294) pg/ml (Table 2). ANOVA analysis showed significantly higher levels by immunoblot than those determined by both ELISA assays, while mean levels of Aβ42 measured by WAKO and Fujirebio ELISAs did not differed (Table 2).
Conversely, the mean levels of Aβ40 determined by immunoblot were not significantly different from those obtained by both ELISA assays (Table 3). Aβ40 levels were similar in AD and controls by ELISAs and immunoblot (Table 4) and Aβ42/Aβ40 ratio discriminated significantly between AD and controls, regardless of the three methods used (data not shown).
Mean of total Aβ40 values determination by immunoblot method and commercially available ELISAs Fujirebio and Wako
a p Value between immunoblot analysis and Fujirebio ELISA is not significative.
b p Value between immunoblot analysis and WAKO ELISA is not significative.
c p Value between WAKO ELISA and Fujirebio ELISA is not significative.
Mean Aβ42 and Aβ40 values determination by immunoblot method and two commercially available ELISAs (Fujirebio and Wako) for each pathological (AD) and control group
Diagnostic accuracy of ELISA and immunoblot determination of Aβ42 in CSF samples
Aβ42 levels determined by ELISAs and immunoblot, were significantly decreased in AD compared to controls (Table 4).
The relative increase of Aβ42 levels determined by immunoblot analysis in both AD and controls compared to the two ELISA methods was comparable.
2D-PAGE immunoblot analysis of Aβ peptides in CSF and brain tissue
Since different Aβ peptides might share the same orthogonal electrophoretic mobility, we resolved CSF samples by 2D-PAGE analysis to confirm the reliability of the immunoblot quantification.
Semi-quantitative densitometric analysis of Aβ peptides spots in AD (Fig. 2A) and control CSF samples (Fig. 2B) within the same gel showed that Aβ40 band is composed by a prominent spot and by a second more acidic spot migrating at the same molecular weight (1–40* in Fig. 2). This minor spot accounts for less than 10% of Aβ40 band while Aβ42 is totally composed by a single spot indicating that the band observed by 1D-PAGE corresponds to Aβ42 only (Fig. 2C, D).
Representative 2D anti-Aβ immunoblot and relative densitometric analysis of control (A and C) and AD (B and D) CSF samples.
Then we compared qualitatively and quantitatively Aβ peptides in CSF and brain samples by 2D-PAGE and immunoblot, using a 2D map of known synthetic Aβ peptides as a reference map (Fig. 3A).
Representative 2D anti-Aβ immunoblot of recombinant Aβ peptides (A), CSF (B) and AD brain (C).
In 2D-PAGE separations of CSF samples, Aβ peptides 1–37, 1–38, 1–40, and 1–42 were detected at the expected theoretical pI (5.4) and molecular masses. Two additional spots with an acidic and a basic shift of about 0.5 pH unit were also detected (Fig. 3B).
In contrast, 2D-PAGE of brain tissue showed the presence of Aβ1 - 42, Aβ2 - 42, Aβ3 - 42, and a more basic spot at the same relative mobility of Aβ3 - 42 (Fig. 3C).
Size of Aβ aggregates in AD and controls CSF samples and in the brain tissue of AD subjects
We determined the size of aggregates of Aβ peptides in CSF of AD patients and controls as well as their extent of aggregation. Thus, we performed a velocity sedimentation through sucrose step gradients on CSF and brain tissue samples (Fig. 4). As expected, the amount of Aβ42 peptides was reduced in CSF samples from AD patients compared to controls, albeit the sedimentation pattern of Aβ peptides was comparable. Aβ peptides were detected mainly in fractions 1 to 4 (soluble small monomers/oligomers), while a weak signal was detected also in fraction 11 (Fig. 4A, B). In contrast, in AD brain tissue preparations Aβ peptides were characterized by insoluble aggregates of large sizes (Fig. 4C) and by Aβ peptides qualitatively and quantitatively different from those of CSF (compare Fig. 4A-C).
Representative immunoblot of velocity sedimentation through sucrose step gradients of control (A) and AD (B) CSFs and AD brain (C) samples. Numbers 1 to 11 indicate the fraction number collected from the top to the bottom.
DISCUSSION
In this work, we showed that two commercially available ELISAs provided non-univocal Aβ42 CSF levels as a likely effect of different Aβ-capturing antibodies, with the Fujirebio Aβ42 ELISA providing higher values than the WAKO ELISA kit [13]. This observation should be considered in light of a shared inter-laboratory consensus regarding the application of common methodologies in assessing Aβ42 CSF levels for both diagnostic purposes and selection of AD patients for clinical trials [15]. Mechanisms leading to decreased Aβ concentration in CSF are still unclear [16], and a number of hypotheses have been put forward, including accumulation of the peptide in the Aβ amyloid plaques [17], decrease in the Aβ42 generation [18], increased degradation in the brain [19], or increased clearance from neural tissues across the blood–brain barrier [20].
The most attractive, but not fully proven, hypothesis is that the decrease of Aβ42 levels might be linked to an epitope masking when using antibody-based detection methods [21]. We showed a significant difference in Aβ42 levels obtained by immunoblot or by ELISA, confirming that the epitope of Aβ42 peptide might be not accessible to capturing or detecting antibodies in native conditions and that solubilization of CSF samples in SDS and high concentration of urea overcomes the bias of immunological detection methods [12]. In particular, the amount of Aβ42, but not of Aβ40, detected by immunoblot is significantly higher compare to ELISAs indicating that, in the CSF, Aβ42 could be assembled into oligomers or aggregated with other interacting proteins [6].
Several studies indicated a matrix interference as the prevalent cause for inaccurate determination of Aβ42 concentration, and this is in keeping with our present results, showing considerably higher Aβ42 levels after sample treatment under denaturing conditions. However, denaturation did not affect differences in Aβ42 levels between AD and control groups, which were observed with all the three methods of measurement.
Effects of pre-analytical denaturation on Aβ42 measurement were also reported by Slemmon et al., who showed that, after denaturation and reverse-phase HPLC, considerable higher amounts of Aβ42 were revealed in CSF than in non-denatured CSF. However, after denaturation, Aβ42 concentrations in the AD group overlapped with those of the non-demented group, losing on one hand the differentiation between the disease and non-disease groups, but on the other hand indicated that Aβ is under different aggregation [21]. In contrast, Yang et al., investigating for the presence of aggregated Aβ42 by specific ELISA, showed similar CSF levels of soluble oligomeric Aβ42 between AD and controls [22].
The CSF level of Aβ40 is considered to most closely reflect the total Aβ load in the brain and CSF Aβ42/Aβ40 ratio is reported to improve differentiation of AD patients from other dementias and controls [23]. As expected, we did not find differences in Aβ40 levels between AD and control samples by all the three methods, confirming a selective decrease of Aβ42 in AD.
2D-PAGE immunoblot analysis showed that Aβ42 consists of a single spot, while the 1D-PAGE band identified as Aβ40 is resolved in a major spot and minor spot, accounting for 10%, a finding consistent with the slightly higher Aβ40 values obtained by 1D-PAGE immunoblot quantitation as compared to ELISAs.
Moreover 2D-immunoblot analysis showed that only Aβ42 is found both in brain tissue and CSF, whereas other Aβ peptides, Aβ37, Aβ38, and Aβ40, are selectively detected in CSF. On the contrary, Aβ3 - 42, which likely triggers tau hyperphosphorylation [24], is expressed in brain tissue but not in CSF, suggesting that these differences could reflect a selective release of Aβ peptides from neural tissues into the CSF.
In addition, the size of Aβ peptides aggregates in CSF and brain tissue from AD showed that CSF is composed by small soluble aggregates while Aβ peptides extracted from the brain tissue are exclusively composed by insoluble and aggregated species. These findings prove that CSF Aβ42 somehow differs from Aβ42 deposited in brain tissue. Conversely, sedimentation pattern analysis did not show qualitative differences in physico-chemical properties of CSF Aβ42 and other Aβ peptides between AD and controls.
Since immunoblot and ELISA are antibody-based methods, we showed that these immunological methods cannot discriminate between pathologic and normal Aβ42 conformers because the pathologic Aβ42 peptides level might probably be below their detection limit.
However, promising results in detecting Aβ misfolded oligomers have been obtained in CSF samples by highly sensitive Aβ Protein Misfolding Cyclic Amplification assay which clearly distinguishes AD patients from controls based on the extent of Aβ oligomers amplification [25].
