Development and assessment of sensitive immuno-PCR assays for the quantification of cerebrospinal fluid three- and four-repeat tau isoforms in tauopathies
Abstract
Characteristic tau isoform composition of the insoluble fibrillar tau inclusions define tauopathies, including Alzheimer's disease (AD), progressive supranuclear palsy (PSP) and frontotemporal dementia with parkinsonism linked to chromosome 17/frontotemporal lobar degeneration-tau (FTDP-17/FTLD-tau). Exon 10 splicing mutations in the tau gene, MAPT, in familial FTDP-17 cause elevation of tau isoforms with four microtubule-binding repeat domains (4R-tau) compared to those with three repeats (3R-tau). On the basis of two well-characterised monoclonal antibodies against 3R- and 4R-tau, we developed novel, sensitive immuno-PCR assays for measuring the trace amounts of these isoforms in CSF. This was with the aim of assessing if CSF tau isoform changes reflect the pathological changes in tau isoform homeostasis in the degenerative brain and if these would be relevant for differential clinical diagnosis. Initial analysis of clinical CSF samples of PSP (n = 46), corticobasal syndrome (CBS; n = 22), AD (n = 11), Parkinson's disease with dementia (PDD; n = 16) and 35 controls revealed selective decreases of immunoreactive 4R-tau in CSF of PSP and AD patients compared with controls, and lower 4R-tau levels in AD compared with PDD. These decreases could be related to the disease-specific conformational masking of the RD4-binding epitope because of abnormal folding and/or aggregation of the 4R-tau isoforms in tauopathies or increased sequestration of the 4R-tau isoforms in brain tau pathology.
Abbreviations used
-
- 3R-tau
-
- three-repeat tau isoform
-
- 4R-tau
-
- four-repeat tau isoform
-
- AD
-
- Alzheimer's disease
-
- CBD
-
- corticobasal degeneration
-
- CBS
-
- corticobasal syndrome
-
- FTDP-17
-
- frontotemporal dementia with parkinsonism linked to chromosome 17
-
- iPCR
-
- immuno-linked quantitative polymerase chain reaction
-
- MAPT
-
- Tau gene
-
- MTBR
-
- microtubule-binding repeat region
-
- NPH
-
- normal pressure hydrocephalus
-
- PDD
-
- Parkinson's disease with dementia
-
- PD
-
- Parkinson's disease
-
- PSP
-
- progressive supranuclear palsy
Insoluble aggregates of abnormally hyperphosphorylated microtubule-binding protein, tau, form the inclusions characterising the broad class of progressive neurodegenerative disorders called tauopathies. Alternative splicing of exons 2, 3 and 10 of the tau gene (MAPT) results in six brain isoforms of tau (Goedert et al. 1989). Alternative splicing of exon 10 codes leads to tau isoforms with three- (3R-tau) or four (4R-tau) microtubule-binding repeat domains (MTBR) with only 3R-tau in embryonic brain and comparable levels of 3R- and 4R-tau in normal adult brain (Goedert et al. 1989). However, it is now clear that imbalances in 3R- and 4R-tau homeostasis are important in disease pathogenesis. This was first suggested by disease-specific isoform profiles of the pathological tau aggregates. For example, in progressive supranuclear palsy (PSP), cortical basal degeneration (CBD), argyrophilic grain disease and familial FTDP-17 cases with exon 10 splicing mutations, the neurofibrillar tangles and glial inclusions are predominantly 4R-tau (Togo et al. 2002; de Silva et al. 2003), whereas Pick body pathology in Pick's disease is predominantly 3R-tau (de Silva et al. 2003) and neurofibrillary tangles in AD contain both 3R- and 4R-tau isoforms. This is now supported by genetics where a large proportion of mutations causing inherited familial FTDP-17t affect exon 10 splicing regulatory motifs or destabilise a predicted pre-mRNA stem-loop structure at the splice donor site causing increased splicing of exon 10 with the general result of elevated levels of 4R-tau isoforms (Hutton et al. 1998; Spillantini et al. 1998). This is therefore central and sufficient for neurodegeneration. Furthermore, although PSP and CBD are largely sporadic, without MAPT mutations, the common MAPT H1 haplotype is strongly associated with increased risk of these disorders (for review see (Vandrovcova et al. 2010)). H1c, a subhaplotype of H1 drives this association (Pittman et al. 2005) and in brain, the H1c allele has higher levels of MAPT transcription and splicing of exon 10, with the effect of an increase in 4R-tau isoforms (Myers et al. 2007). This could in part lead to the characteristic 4R-tau dominant pathology in PSP and CBD and, combined with the neuropathological observations, these genetic findings reiterate the importance of disturbed 3R-/4R-tau isoform homeostasis, particularly, elevated 4R-tau, in the aetiology of some of the tauopathies. Interestingly, it is now clear that the MAPT H1 haplotype together with the α-synuclein gene (SNCA) are most consistently the strongest risk factors for sporadic PD (Lill et al. 2012). This is surprising as PD is not a tauopathy and would suggest that changes in tau homeostasis could contribute to neurodegenerative pathogenesis, but without the necessary formation of overt tau neurofibrillary pathology, perhaps implicating toxic oligomeric intermediates (Patterson et al. 2011).
The past decade has seen sustained efforts to discover clinically relevant diagnostic biomarkers for the early diagnosis of neurodegenerative disorders with minimal invasiveness to the patient. To date, no single biomarker consistently achieves diagnostic accuracy. CSF tau is elevated in AD, but it is a combination of increased tau and decreased Aβ that gives the best diagnostic accuracy for AD and this combination can identify prodromal AD and also differentiate AD from other types of dementia (for review, see Blennow et al. 2010). However, the other tauopathies do not have a reliable diagnostic test. Many tauopathies are often clinically misdiagnosed because of their overlapping clinical features. For example, overall diagnostic accuracy for PSP is 70–75% (Hughes et al. 2002) and misdiagnoses include idiopathic PD, CBD and FTDP-17 (Josephs and Dickson 2003). Corticobasal syndrome (CBS), previously defined as CBD, has no definitive clinical predictor of pathological outcome with its association with a more varied spectrum of pathologies, including classical CBD (a 4R-tauopathy with presence of extensive neuronal and glial tau inclusions), PSP and AD-type pathology (Ling et al. 2010). An objective laboratory test based on CSF would therefore be extremely useful to help neurologists with the early diagnosis of these conditions and monitoring disease progression/recovery in patients in response to drug intervention or novel treatments. To date, CSF tests for total tau (t-tau) and phosphorylated tau (p-tau) in PSP and CBS have not provided consistent disease-specific changes (Arai et al. 1997; Urakami et al. 2001).
In this work, we describe the successful development and validation of immuno-PCR (iPCR) assays for measuring the trace levels (< 100pg/mL) of CSF 3R- and 4R-tau and preliminary assessment of their diagnostic utility in tauopathies. Immuno-PCR is a highly sensitive detection method first described by Sano and colleagues (Sano et al. 1992), whereby chromogenic enzyme substrates are replaced by a DNA molecule conjugated to the detection antibody, enabling 100- to 10,000-fold improvements in detection compared to standard ELISAs. These assays were based on 3R- and 4R-tau sandwich ELISAs (Luk et al. 2009, 2010) that utilised the well-characterised 3R- and 4R-tau-specific monoclonal RD3 and RD4 antibodies (de Silva et al. 2003).
Materials and methods
Subjects and clinical protocol
CSF samples obtained from four European centres are summarised in Table S1 (see supplementary material). Cohort A, was from Hospital Clínic, Barcelona, Catalonia, Spain. Control subjects were healthy individuals (normal neurological examination with no parkinsonism and with MMSE (Folstein et al. 1975) > 28/30) who were admitted for knee replacement surgery and agreed to donate CSF when receiving the lumbar tap for intradural anaesthesia, with CSF being obtained ahead of the intradural infusion of the anaesthetic drug. In all five cases that underwent autopsy, the time from lumbar puncture to death ranged from 2 to 3 years. Cohort B, was from the Paracelsus-Elena Klinik, Germany. Three controls were normal pressure hydrocephalus (NPH). Noting that many NPH cases have underlying PD or AD, the NPH controls we used were investigated carefully and did not have PD by Queen Square Brain Bank Criteria, brainstem ultrasound, levodopa test and AD by neuropsychological testing and they were idiopathic by MRI criteria and improved with spinal tap test. Cohorts C and D were, respectively, from the Sahlgrenska University Hospital, University of Gothenburg, Sweden and the VU University Medical Center, Amsterdam, the Netherlands. Controls were without any known neurodegenerative and psychiatric disease.
PSP patients recruited in all three cohorts were diagnosed according to the NINDS-SPSP criteria for clinically probable PSP(Litvan et al. 1996). In the absence of autopsy-validated clinical diagnostic criteria, we used generally accepted diagnostic criteria for inclusion of CBS cases(Litvan et al. 2000), whereas CBD was defined post-mortem using pathological criteria (Dickson et al. 2002). PD was diagnosed according to the Queen Square Brain Bank clinical diagnostic criteria (Hughes et al. 1992). PDD was defined as DSM-IV-R criteria for dementia (American Psychiatric Association, 2000) and the recently proposed Movement Disorders Society diagnostic criteria for PDD (Emre et al. 2007). The clinical diagnosis of dementia was additionally supported by an MMSE score < 25, in the absence of alternative explanations such as psychomotor slowing, motor and speech disorders or severe depression. All PDD patients were examined by neurologists and either a psychologist and/or psychiatrist for the differential diagnosis of depression.
The study was approved by the Joint Research Ethics Committee of the UCL Institute of Neurology, the National Hospital for Neurology and Neurosurgery, London, and the local ethics committees at the participating centres. All controls and patients gave informed written consent after full explanation of all the procedures.
Lumbar puncture and CSF processing
CSF collection from patients and controls was carried out in all centres using the same standard protocol. Lumbar puncture was performed in the morning and CSF collected into polypropylene tubes, immediately centrifuged and stored in aliquots in polypropylene tubes at −80°C until analysis.
Development of immuno-PCR for 3R- and 4R- tau
The immuno-PCR (iPCR) was developed based on the sandwich configuration of the 3R- and 4R-tau ELISAs (Luk et al. 2009, 2010). For specific recognition and capture of the isoforms, we used the well-characterised RD3 and RD4 monoclonal antibodies (de Silva et al. 2003) and for detection, we used a sheep pan-tau antibody-DNA conjugate (Luk et al. 2009) for quantitation by real-time quantitative PCR. The latter was designed to recognise regions flanking and as close as possible to the MTBR (Luk et al. 2009) to maximise the detection of both full-length and proteolytic fragments of tau that could still contain the MTBR.
Buffers and reagents were obtained from Chimera Biotec, Dortmund, Germany. Affinity-purified sheep anti-tau (Luk et al. 2009) was conjugated to DNA at Chimera Biotec using their proprietary Imperacer technology as described previously (Adler et al. 2005). Washing steps were carried out using Tecan Hydroflex washer (Tecan Ltd, Theale, UK). First, monoclonal capture antibodies RD3 and RD4 (30 μL) were coated at various concentrations (1μg/mL, 4μg/mL, 5μg/mL, 10 μg/mL and 20 μg/mL) overnight at 4°C on Topyield modules (Nunc). Strips were washed with wash buffer A (Chimera) and blocked for 5 min at 22 ± 3°C with 240 μL of direct block buffer (Chimera). Thirty microlitres of recombinant 3R- and 4R- tau (rPeptide, Bogart, GA, USA; 10 ng/mL, 100 pg/mL, 10 pg/ mL and 0 pg/ mL) diluted in sample dilution buffer were added onto the Topyield strips in duplicates. The strips were incubated on a rotary shaker for 1 h at 22 ± 3°C and washed three times with buffer B followed by three times with buffer A (Chimera). After washing, sheep anti-tau-DNA conjugate was diluted in conjugation dilution buffer at various concentrations (1 in 30, 1 in 300 and 1 in 1000). Thirty microlitres of the diluted sheep anti-tau DNA conjugate were sequentially added onto the washed strips in duplicates and incubated for 30 min at 22 ± 3°C on a rotary shaker. The strips were washed twice with buffer B followed by three times wash with buffer B. After washing, PCR-Mastermix (Chimera) containing the primers for amplification of the DNA tag (30 μL) were added into each strip well. DNA amplification was carried out by real-time quantitative PCR using the Stratagene MX3000P real-time quantitative PCR system. Cycling conditions were: Initial step of 95°C for 4 min followed by 50 cycles of 50°C for 30 s, 72°C for 30 s and 95°C for 12 s. Fluorescence of incorporated label was read at the end of each 72°C step. The resulting Ct values were inversely proportional to tau concentrations, where negative controls (NC) had the highest value. Delta Ct values were calculated by subtracting Ct values from the total number of cycles carried out in the real-time PCR. This allows direct comparison between iPCR and ELISA as concentrations are proportional to the resulting ELISA signal.
To allow direct comparison between the iPCRs and ELISAs, the Ct values from the real-time PCR were converted to delta Ct by subtracting from the total number of cycles carried out in the experiment (50-Ct). To optimise the iPCR assays, various concentrations of RD3 and RD4 antibodies were used to coat the Topyield strips. The delta Ct values of 10 pg/mL, 100 pg/mL and 10 ng/mL of recombinant tau were normalised against the delta Ct value of 0 pg/mL. Supplementary Figures S1a and S1b show different coating concentrations of RD3 and RD4 against normalised delta Ct values of 3R- and 4R- recombinant tau respectively. Coating concentrations of 4 μg/mL of RD3 and 5 μg/mL of RD4 were shown to be optimal (Figure S1a and S1b). These coating concentrations produced high normalised delta Ct values of 0.26 ± 0.40 and 0.29 ± 0.12 for the lowest tau concentration tested (10 pg/mL) for 3R- and 4R- tau iPCRs, as well as greater differences in normalised delta Ct values between tau at 10 pg/mL and 100 pg/mL, potentially resulting in a steeper standard curve, hence, higher assay sensitivity. The optimal sheep anti-tau-DNA dilutions were determined as 1 in 300 for the 3R- tau assay (Supplementary Figure S1c) and 1 in 1000 for the 4R-tau assay (Supplementary Figure S1d). These dilutions also produced greater differences in normalised delta Ct readings between 10 pg/mL and 100 pg/mL compared with other sheep-anti-tau DNA dilutions used.
Optimisation of CSF incubation conditions
The optimal RD3 and RD4 concentrations (4 μg/mL and 5 μg/mL respectively) were coated on Topyield strips overnight at 4°C. The strips were washed and blocked as described previously and processed under sequential incubation (addition of tau followed by anti-tau DNA conjugate) and the combined incubation conditions (addition of both tau and anti-tau DNA simultaneously) respectively. For the sequential incubation experiments, 30 μL of recombinant 3R-tau standards (10 ng/mL, 100 pg/mL, 10 pg/mL and 0 pg/mL) diluted in artificial CSF (150 mM NaCl, 2 mM CaCl2, 1.2 mM MgCl2, 0.5 mM KH2P04, 1.5 mM K2HPO4 and 10 mM glucose, pH 7.3) at a ratio of 1 : 1, 1 : 3 and 1 : 10 were added in duplicates onto the washed strips. Artificial CSF is formulated to closely mirror the electrolyte composition and osmolality of endogenous CSF and, being more physiologically compatible, is commonly used as a sterile diluent or flushing fluid for delivery of molecules to the brain. Following incubation for 1 h at 22 ± 3°C and washing, 30 μL of optimal sheep anti-tau DNA concentration was added onto the strips and incubated for 30 min at 22 ± 3°C. The strips were then washed and PCR master mix was added and the plate was processed as previously described. For the combined incubation experiments, recombinant 3R- tau standards (10 ng/mL, 100 pg/mL, 10 pg/mL and 0 pg/mL) in artificial CSF diluted in sample dilution buffer at a ratio of 1 : 1, 1 : 3 and 1 : 10, together with an equal volume of optimal sheep anti-tau DNA was prepared in Protein LoBind Eppendorf tubes. Thirty microlitres of the combined incubation mixture was added onto the coated strips in duplicates and incubated overnight at 4°C. The strips were washed and processed as described previously and fluorescence was detected by real-time PCR.
To optimise the iPCR assays for CSF, artificial CSF diluted in sample dilution buffer at 1 : 1, 1 : 3 and 1 : 10 were compared. The 3R-tau assays diluted at 1 : 1 of artificial CSF gave the highest normalised delta Ct values at 1.69 ± 0.37, 3.83 ± 0.04 and 7.21 ± 0.26 for the three 3R-tau concentration tested (10pg/mL, 100pg/mL and 10ng/mL) (Supplementary Figure S1e), resulting in high assay sensitivity. Supplementary Figure S1f shows the 4R-tau assays carried out with the combined incubation step outperformed the sequential incubation format, resulting in higher normalised delta Ct values for 10pg/mL, 100pg/mL and 10ng/mL of 4R tau.
Parallelism
The parallelism between calibrant buffer and artificial CSFs as well as individual or pooled CSFs against artificial CSFs were studied by spiking these matrices with 3R- and 4R- tau at 100 000, 10 000, 100 and 10 pg/mL. The parallel relationships for buffer and artificial CSF for the 4R- and 3R- tau iPCR are shown in Figures S2a and S2b, respectively. Figure S2c also shows the parallelism between the 4R- tau spiked in artificial CSF, an individual CSF and a pooled CSF sample. These parallel relationships suggested the absence of endogenous binding between CSF tau isoforms and other CSF components.
CSF analysis by immuno-PCR
Topyield modules (Nunc) were coated with 30 μL of RD3 and RD4 antibodies (4 μg/mL and 5 μg/mL respectively) overnight at 4°C before the strips were washed and blocked as previously described. After washing, 40 μL of CSF samples or recombinant 3R- tau standards (10 ng/mL, 1 ng/mL, 100 pg/ mL, 10 pg/mL and 1 pg/mL and 0 pg/mL) diluted in artificial CSF were added into an equal volume of sample dilution buffer in Protein Lobind Eppendorf tubes. Eighty microlitres of 1 : 300 sheep anti-tau DNA conjugate was added into the diluted CSF or 3R-tau standard series. For the 4R-tau assay, 125 μL of CSF samples or recombinant 4R-tau in the same standard series in artificial CSF was also added into an equal volume of sample dilution buffer (125 μL) and 250 μL of sheep anti-tau DNA conjugate (1 in 1000). Thirty microlitres of 3R- and 180 μL of 4R-tau standard mixture along with sample mixtures were added into the blocked and washed Topyield strips coated with RD3 or RD4 antibodies. The reaction mixture was incubated overnight at 4°C. The strips were washed, processed and quantified by real-time PCR as previously described.
As a result of differences in harvest and storage inherent in CSF collections from multiple centres as well as variability between assays, case-control cohorts from each contributing centre were assayed and analysed in single batches, i.e. all samples from each centre were assayed together on the same day. This is important as we see the clear centre effects when looking at the 3R- and 4R-tau data from different centres.
Additional CSF analyses
Total tau (t-tau) and phospho-tau (p-tau) levels were measured for cases from cohort A and D by Innotest hTAU and Innotest phospho-tauP181 ELISA kits (Innogenetics, Gent, Belgium) (Blennow et al. 1995; Vanmechelen et al. 2000) and we deduced the ratio of 3R-tau:t-tau, 4R-tau:t-tau, 3R-tau:p-tau and 4R-tau:p-tau of controls, CBD, PSP, PDD and AD from these cohorts.
Statistical analysis
GraphPad Prism software (GraphPad Software Inc, La Jolla, CA, USA) was used for statistical analysis. Non-parametric Kruskal–Wallis test followed by a post hoc Dunn's multiple comparison were used to analyse samples which did not pass the normality test to compare CSF 3R- and 4R- tau concentrations among healthy controls and disease groups. With data represented in Fig. 5, when comparing only PSP versus controls, we used an unpaired t-test with Welch's correction.
Results
Performance of immuno-PCRs for 3R- and 4R-tau
Figure 1 shows the significantly enhanced dynamic ranges (10–10,000 pg/mL) and minimal detection limits (10pg/mL) for both the 3R- and 4R-tau iPCRs compared to the corresponding ELISAs. Parallelism between calibrant buffer and artificial CSFs as well as individual or pooled CSFs against artificial CSFs by spiking these matrices with 3R- and 4R-tau suggested the absence of endogenous binding or interference between CSF tau isoforms and other CSF components (Supplementary Figure S2). With percentage recovery and interassay variations, we also demonstrated reproducibility and acceptable recoveries of the assays (Table S2; see supplementary material).
Analysis of tau isoforms in the clinical cohorts separately
There were no statistical age differences between each diagnosis within each cohort (Table S1; see supplementary material).
We analysed all cohorts separately to assess independent performance. No statistical differences in CSF 3R-tau levels were observed between the different subgroups in all four cohorts (Fig. 2).
Figure 3 shows levels of 4R-tau in the four cohorts. In cohort A, we showed that 4R-tau levels in PSP and PDD are significantly lower than the control group (p < 0.05). We saw similar trends in cohorts B, C and D. Levels of 4R-tau in cohort B were in the following order: Controls > CBS > PSP > PDD and, in cohorts C and D, mean 4R-tau concentrations in controls were highest compared to CBS, PSP and AD. These trends of lowered 4R-tau in PSP were similar to the results obtained from cohort A, although not statistically significant. In cohort D, 4R-tau levels in the AD cases were significantly less (p < 0.05) than in the control group.
Analysis of tau isoforms in clinical cohorts combined
There were no statistical differences in the ages of all patients combined between all diagnoses (Table S1; see supplementary material).
When levels of tau isoforms from the four separate measurements were combined, there were no significant differences in CSF 3R-tau between different subgroups (Fig. 4a). On the other hand, combined analysis showed 4R-tau levels (Fig. 4b) were significantly lower in PSP (p < 0.01) and AD (p < 0.001) compared with controls. The level of 4R- tau in CSF of AD was also significantly lower than that of PDD (p < 0.01).
Because of the possibility of intercentre and interassay effects resulting from differences in pre-analytical factors pertaining to each contributing centre, and the sensitivity of the iPCR assays, we randomly selected 17 controls and 22 PSP samples from all four cohorts and analysed them on the same plate. The results confirmed a statistically significant decrease in CSF 4R-tau in PSP compared with controls (p < 0.05) using an unpaired t-test with Welch correction (Fig. 5).
Ratio of tau isoforms to total tau and phosphorylated tau (Cohorts A & D)
We measured the total tau (t-tau) and phospho-tau (p-tau) levels from cohorts A and D and deduced the 3R-tau: t-tau, 4R-tau: t-tau, 3R-tau: p-tau and 4R-tau: p-tau ratios. Mean 3R-tau: t-tau ratios (median in brackets) were between 0.015 (0.014) ± SEM 0.0028 and 0.094 (0.084) ± SEM 0.022 (Fig. 6a) and, mean 3R-tau: p-tau ratios were between 0.11 (0.09) ± 0.02 and 0.86 (0.62) ± SEM 0.20 (Fig. 6b). The values for both 3R-tau: t-tau ratio and 3R-tau: p-tau ratio were significantly lower (p < 0.05 and p < 0.001 respectively) in AD compared with PDD.
The mean ratios of 4R-tau: t-tau (Fig. 6c) were as follows (median in brackets):
Controls: 0.166 (0.09) ± SEM 0.044, CBS: 0.037 (0.011) ± SEM 0.012, PSP: 0.040 (0.021) ± SEM 0.010, PDD: 0.067 (0.068) ± SEM 0.016 and AD: 0.0058 (0.0017) ± SEM 0.0029.
Mean ratios of 4R-tau: p-tau (Fig. 6d) were as follows:
Controls: 1.204 (0.814) ± SEM 0.289, CBS: 0.168 (0.092) ± SEM 0.044, PSP: 0.24 (0.115) ± SEM 0.06, PDD: 1.65 (1.49) ± SEM 0.38 and AD: 0.046 (0.01) ± SEM 0.023.
The 4R-tau: t-tau ratio was significantly reduced in PSP (p < 0.05) and AD (p < 0.001) compared with controls (Fig. 6c). Similarly, significantly reduced 4R-tau: p-tau ratio is observed in AD (p < 0.001) compared with controls (Fig. 6d) and the 4R-tau: p-tau ratio is significantly higher in PDD when compared with CBS and PSP (p < 0.01) and AD (p < 0.001) (Fig. 6d).
Moreover, a significant decrease in 4R-tau: t-tau ratio (p < 0.05) was also observed in AD compared with PDD.
Autopsy-confirmed cases (Cohort A)
Five cases in Cohort A came to post-mortem pathological confirmation during the course of this work. For example, one patient who presented in clinic with Richardson's syndrome (PSP-RS; Williams et al. 2005), was pathologically confirmed as PSP. This case had the third lowest level of 4R-tau in all of Cohort A (13.9 pg/mL) with only a clinical PDD (13.8 pg/ml) and another clinical PSP case (12.1 pg/mL) having lower 4R-tau levels. The levels in these three cases are lower than both the mean (43.2 pg/mL) and the lowest value (20.6 pg/mL) of the controls in Cohort A.
Three clinical CBS cases that were pathologically confirmed as CBD had higher 4R-tau levels (20.6, 30.4 and 32.2 pg/mL) than the PSP case, but still below the mean of the controls. Lastly, a pathologically confirmed PDD case also had low 4R-tau levels (28.1 pg/mL), but it is to be noted that this case had remarkable concomitant AD pathology (Braak stage III-C fulfilling the CERAD criteria for definite AD pathology).
Discussion
Total tau (t-tau) levels in CSF range from 243 ± 127pg/mL in 51–70-year olds, to 341 ± 171 pg/mL in those over 71 years (Sjögren et al. 2001). With detection thresholds of 10 pg/mL, the immuno-PCR (iPCR) platforms we describe here are therefore sensitive enough for CSF tau measurements. The parallel relationships of the iPCRs in both calibrant buffer and 1 : 1 diluted CSF suggest the absence of matrix effects because of interference or binding with CSF components. We also showed that the iPCR assays are reproducible with acceptable recoveries, further supporting the specificity of the assays.
Brain protein changes in the CSF could be surrogates for pathological changes taking place in the brain. For example, the increase in tau tangle formation and tau load in AD brains is consistently mirrored by an increase in t-tau and p-tau in CSF (for review, see Hampel et al. 2010). However, studies of PSP and CBS CSFs have not shown any consistent changes in CSF t-tau and p-tau compared to controls (Arai et al. 1997; Urakami et al. 2001). With the characteristic tau isoform content of the fibrillar inclusions in tauopathies, we surmised instead that tau isoforms in the brain could be reflected in the CSF and we expected specific increases of CSF 4R-tau in the 4R-tauopathies, PSP and CBS/CBD. Instead, we observe reduced 4R-tau not only in PSP and CBS but also AD (mixed 3R-/4R-tauopathy) and PDD. Although PD is not traditionally described as a tauopathy, we have shown that increased AD-like pathology (β-amyloid plaques and tangle tau) increases risk of dementia in PD (PDD), suggesting that pathological tau changes could contribute to the dementia in these cases (Compta et al. 2011). Furthermore, CBS is now recognised to present with a varied spectrum of pathologies apart from the classical CBD-type (4R-tau) pathology. In one study of 21 cases with a clinical diagnosis of CBS, only five had CBD pathology, giving a positive predictive value of 23.8%; six others had PSP pathology, five had AD and the remaining five had other non-tau pathologies (Ling et al. 2010).
The reductions of 4R-tau in cases compared with controls are statistically significant within individual cohorts and with combined data from all cohorts, as well as when PSP and control samples from all cohorts were combined in a single assay to test for interassay and intercentre variability. Furthermore, despite their limitation to a subset of the cases studied, the 4R-tau: t-tau and 4R-tau: p-tau ratios, on one hand, and the CSF values from autopsy-confirmed cases, on the other, are in keeping with the findings from the assessment of all the cohorts. Although measures for individual cases preclude formal statistical comparison and should therefore be interpreted with caution, the autopsy-confirmed cases were of interest as they increased our confidence in correct clinical diagnosis and show us where the individual tau isoform values fall within the range of values for cases and controls.
A similar inverse relationship, i.e. lowered CSF levels of the pathological substrate, has been observed with the Aβ1–42 peptide in AD patients (Strozyk et al. 2003; Tapiola et al. 2009) and α-synuclein in PD patients (Tokuda et al. 2006). Lower CSF Aβ1–42 levels correlates with increased brain amyloid load as revealed with the in vivo amyloid PET imaging tracer 11C-PIB (Pittsburgh Compound B), suggesting that brain amyloid deposition depletes the CSF of Aβ1–42 (Fagan et al. 2006; Forsberg et al. 2008). It is therefore possible that the lowered 4R-tau in CSF similarly reflects selective accumulation of this more fibrillogenic isoform in diseased brain, and shared pathogenic processes in PSP, AD and PDD.
However, specific reduction in 4R-tau could also be due to loss of the RD4 epitope through disease-specific proteolysis of tau proteins and/or conformational epitope masking and further occlusion by tau aggregation in disease. Carboxy-terminal truncation of tau has been suggested to be a primary event that facilitates conversion to pathological conformation of tau (Horowitz et al. 2006; Wray et al. 2008; Borroni et al. 2009) (for review, see (Kovacech and Novak 2010)). The disease-specific proteolytic events could be accompanied by conformational changes such as adoption of a ‘paperclip’ configuration because of folding over of the amino- and carboxy-termini and interaction with the exon 10-encoded repeat of the MTBR (Jeganathan et al. 2006).
Slemmon and colleagues showed that interactions with other CSF components in part cause the perceived reduced levels of CSF Aβ1–42 in AD (Slemmon et al. 2012). Although our iPCR assays are free of matrix effects, it is possible other CSF components contribute to the reduced 4R-tau signal. In fact, work by Ono and colleagues showed that CSF from AD and Lewy body dementia cases specifically enhanced fibrillisation of recombinant α-synuclein (Ono et al. 2007), suggesting that CSF from neurodegenerative proteopathies contain pathological seeds for the further conformational conversion and aggregation of normal protein. Further study of this may contribute to our understanding of the possible role of CSF environment on the aggregation and masking of 4R-tau.
The relationship between brain- and CSF tau could be further complicated by selective release of tau variants (isoforms as well as proteolytic fragments) from the brain into the CSF. It is clear that there is a complex set of processes involving tau isoform imbalances, proteolytic processing, selective aggregation, variable tau pathology and brain–CSF exchange and these assays could help us investigate this.
We are conscious of the potential shortcoming of this project as it involved a large multicentre collection of CSF samples. It is now clear that with the trace levels of CSF protein coupled with the very sensitive ELISA assays, studies are compromised by pre-analytical factors involving the entire process from clinical diagnosis and CSF collection to storage and the ensuing protein degradation (Teunissen et al. 2009). To account for inter-centre differences, we therefore carried out separate analyses of data for each cohort as well as for the combined cohorts and noted that the selective depletion of 4R-tau is consistent in all collections.
To assess the utility of these iPCR in clinical diagnosis, we are now carrying out further studies with standardised operating protocols across centres, larger cohorts, longitudinal design and the use of additional or alternative biomarkers (ratios with the other available CSF tau-related markers or PET imaging probes for brain tau deposition that might emerge in the future). If reductions in 4R-tau reflect tau aggregation, the 4R-tau iPCR assay may be used to monitor disease progression in tauopathies, and its combination with t-tau/p-tau and Aβ1-42 might assist the differential diagnosis between different tauopathies. In general, longitudinal changes in tau isoforms could track disease progression and be of value in clinical trials of anti-4R-tau deposition agents, as well as their add-on value as part of the diagnostic work up in cases with an uncertain clinical diagnosis (unclassifiable parkinsonism).
Acknowledgements
The authors thank the patients and their families who contributed to this study – without their cooperation, this project would not have been possible. We thank Sven Schulz and Mark Spengler at Chimera Biotech GmbH for training with the use of the Imperacer system. This work was funded by the UK Medical Research Council (RdS; G0501560), Progressive Supranuclear Palsy (Europe) Association, Cure PSP+, The Brain Research Trust (Peacock Trust), the Wolfson Foundation and the Reta Lila Weston Trust for Medical Research.
RdS has licensed RD3 and RD4 antibodies to Upstate/Millipore. The authors declare no other potential conflicts of interest.
Author contributions
Conception, design and planning of project: CL, YC, RdS, AJL.
Assay development: CL, YC, RdS, GH.
Sample collection and manuscript preparation: All authors.
