Introduction
Intracellular aggregates of the microtubule-associated protein tau define Alzheimer’s disease (AD) and related neurodegenerative tauopathies. In AD, tau progressively accumulates in defined patterns that involve brain networks [
6]. This has been hypothesized to occur via formation of aggregate “seeds” in a single neuron or group of neurons that exit and gain entry to neighboring or synaptically connected cells. The seeds then serve as templates for amplification of specific pathological tau assemblies [
17,
18]. Our lab developed the first generation of tau “biosensor” cells to detect pathological seeds, here termed “v1.” These express the tau repeat domain (RD: aa244-368) containing a single disease-associated mutation (P301S) fused to cyan and yellow fluorescent proteins (CFP/YFP) [
21]. Exposure of biosensor cells to pathological forms of tau (monomer or larger assemblies) induces Tau RD-CFP/YFP aggregation that creates fluorescence resonance energy transfer (FRET) [
13,
21]. This may be detected by microscopy or flow cytometry [
15,
21]. Biosensor cells spontaneously take up free tau aggregates, or “naked” seeds [
20], but admixture with a cationic lipid-based transfection reagent such as Lipofectamine 2000 increases seeding by ~ 100–1000-fold [
21]. We have previously used the original biosensor cells to detect tau seeding activity in mouse and human brain prior to neurofibrillary tangle pathology. This indicates that seed formation is perhaps the earliest detectable tau-related pathological event [
16,
24], and thus could be an excellent disease biomarker. Reliable detection of seeding activity in peripheral fluids such as CSF or blood from living subjects has not yet been established, but could improve specific diagnosis of tauopathy.
Numerous studies have sought to amplify tau seeds from cerebrospinal fluid, either using biosensor cells [
37], in vivo propagation in a mouse model [
36], or in vitro amplification assays such as real-time quaking-induced conversion (RT-QuIC) [
30,
31]. Three groups have reported detection of seeding activity in antemortem CSF of living individuals [
31,
36,
37], but the sensitivity and specificity of these approaches are not established. Most mass spectrometry (MS) studies have observed that tau in the CSF contains the N-terminal and/or mid-region of tau, but not the RD [
2,
8,
33]. Tau RD forms the core of tau oligomers and fibrils and is required for seed formation [
9,
11]. However a recent study reports that AD CSF contains higher concentrations of tau regions within the RD than previously measured, and that levels of a fragment at the C-terminus of RD correlate well with disease progression [
23]. In this study we have created a highly sensitive second generation of biosensors and tested their ability to detect tau seeding in antemortem human AD CSF.
Materials and methods
Generation of biosensor cell lines (Tau RD(P301S)v2L and Tau RD(P301S)v2H)
We used the previously described lentiviral FM5-YFP plasmid containing the tau segment 246 to 378 with the P301S mutation [
32], replaced the human ubiquitin C (Ubc) promoter with a human cytomegalovirus (CMV) promoter, and replaced the YFP sequence with a cerulean 3 or mClover3 coding sequences. The sequence linking the tau segment and the coding sequence of the fluorophore (Cer or Clo) was optimized to the following sequence: GSAGSAAGSGEF [
40]. Low passage HEK293T cells (P5) were thawed and passaged with antibiotic free media twice before co-administration of lentivirus encoding tau RD(P301S)-Clo/Cer. After four passages, single cells were cell sorted by FACS based on low or high signal for both mCerulean3 and mClover3, termed version 2 low (v2L) or version 2 high (v2H). Monoclonal colonies were expanded and characterized as described. Second generation biosensors were used according to established protocols [
21].
Western Blot
We prepared cell lysates by resuspending frozen cell pellets (~ 1 million cells) in 100µL of 0.25% Triton X-100 with protease inhibitors and incubating for 30 min on ice, followed by centrifugation at 21,000×g for 15 min. We adjusted clarified supernatants to a concentration of 1 mg/mL as determined by Pierce 660 nm assay and SDS-PAGE was performed with 5 µg of total protein loaded onto a 4–20% BisTris gel. After transferring the protein to a PVDF membrane, we blocked it with 5% milk in 0.1% TBS-T for 1 hr at room temperature. To detect tau protein we used HJ10.3 (David Holtzman, Washington University, St. Louis, MO), a mouse monoclonal antibody that binds the RD of tau (amino acids 250–268), at a 1:10,000 dilution in blocking buffer for 4hrs at RT. To detect the fluorescent proteins fused to tau, we used the Rockland anti-GFP antibody (cat. 600-101-215), which binds all GFP variants used in this work, at a 1:10,000 dilution, in blocking buffer. After blotting with appropriate secondary antibodies and imaging, we stripped and reblotted the membranes for GAPDH with (6C5, Fisher cat.NC9537307) at a 1:5000 dilution in blocking buffer.
Recombinant tau fibrils
We synthesized wild-type full-length (2N4R) tau and purified it as previously described [
14]. We incubated 8 µM purified recombinant tau with 8 µM heparin and 10 mM DTT at 37C for 48 h in 10 mM HEPES, 100 mM NaCl, PH 7.4. We verified the quality of fibrils by transmission electron microscopy.
Human AD and mouse brain tissue
All mice were housed and cared for according to the UT Southwestern animal care and use guidelines and all applicable U.S. laws governing animal research. Mice: Wild type C57BL/6 (stock #00,064, Jackson Laboratory), Tau knockout (stock #007,251), and PS19 mice expressing human 1N4R tau with the P301S mutation under control of the mouse prion promoter (Prnp) [
43] (stock #008,169) mice, all 9 months old. We transcardially perfused all mice, then removed and immediately flash-froze the brains in N
2(l). We obtained frozen frontal cortex human brain tissue from 5 cases with a histopathological diagnosis of AD from the brain bank of the Alzheimer’s Disease Center UT Southwestern. We homogenized brain tissue in 10% w/vol of 1 × TBS with protease inhibitor cocktail (Roche) at 4 °C using a dounce homogenizer followed by intermittent probe sonication (Omni International) for 10 min. We centrifuged homogenates at 21,000×
g for 15 min at 4 °C to remove cellular debris and determined protein concentrations by BCA assay (Thermo Fisher).
Human CSF
We obtained human lumbar CSF from the UT Southwestern O’Donnell Brain Institute Biorepository, along with clinical data including age, sex, and CSF t-tau, p-tau, and Aβ42 levels as measured by ADmark clinical assay (Athena Diagnostics).
Immunoprecipitation from CSF
50 μl of Dynabeads Protein A (Thermo Fisher) were washed per the manufacturer protocol and incubated with 10 μg of polyclonal antibody (TauA, Diamond Lab, available upon request) targeting the first microtubule-binding repeat of tau for 1 h at room temperature. We then added washed beads to 1 or 5 ml of human CSF and incubated them with rotation overnight at 4 °C. We eluted captured proteins in low pH elution buffer (Pierce) and neutralized the buffer with 1:10 1 M Tris pH 8.5 with a final volume of 120 μl.
Seeding assays
Cells were passaged every 48-72 h and never allowed to reach > 80% confluence. We found that over-confluence and sparse passage conditions both reduced cell health and contributed to background false positive FRET and variability in FRET measurements. This was particularly true of the v2H cells. We plated v1 P301S [
21] or v2L or v2H HEK biosensor cells in 96-well plates at 20,000 cells per well 24 h before treatment. We allowed dilutions of recombinant tau fibrils or brain homogenates in Opti-MEM (Thermo Fisher), 30 μl total volume, or immunoprecipitation eluents 120 μl total volume, to come to room temperature. We mixed 1.5 μl of Lipofectamine 2000 transfection reagent (Invitrogen) with 28.5 μl of Opti-MEM for each sample and incubated it at room temperature for 5 min before mixing it with the sample. After incubating the mixtures at room temperature for 30 min, we divided them among 3 wells of a 96-well plate. After 48 h, we trypsinized and fixed cells in 2% PFA and suspended them in flow cytometry buffer (1 × HBSS, 1% FBS, 1 mM EDTA). We determined percent FRET positivity of each well by flow cytometry as previously described [
15].
Discussion
Tau assemblies that act as templates for their own amplification (seeds) may underlie progression of neurodegenerative tauopathies, and assays that measure the levels of these pathogenic forms thus have great utility. While highly sensitive and specific conformational antibodies would be ideal, amplification of tau seeds in purified systems [
27,
29‐
31] or in cultured “biosensor” cells has been the mainstay of sensitive and specific detection of pathological tau. Biosensor cells created by our group or others are now widely reported to detect pathological forms of tau [
4,
7,
10,
13,
15,
16,
21,
24,
25,
34,
35,
38,
42]. Reliable measurement of tau seeding activity in a peripheral fluid such as CSF could be very useful in detection of incipient tauopathy. Consequently, we have tried repeatedly and unsuccessfully to detect pathological tau in human CSF or blood. We reasoned that a more sensitive biosensor system might solve the problem. We optimized expression of tau RD(P301S)-Clo/Cer in HEK293T cells and created a biosensor with 100-fold improved sensitivity versus the original line [
21]. Interestingly, this increase was noted only with AD-derived seeds and not with synthetic heparin-treated fibrils (Fig.
2). This is unlikely to be due to differences in average fibril size as we have found no differences in the curves with synthetic fibrils sonicated for five minutes to decrease the fibril size (Additional file
3: Fig. 3). The reason for this difference remains to be determined, but it likely involves a preference of the P301S RD construct for the structure of the AD strain over those of synthetic fibrils, which are known to have distinct tertiary structure [
11,
44]. The v2H line should be especially useful to quantify tau seeds that are of low abundance. Given their linear dynamic range, the v2L line may be more useful to quantify seeding in samples with stronger signal. The utility of the original v1 biosensor assays to detect early evidence of tau pathology in brain tissue has already been demonstrated, and we anticipate that the v2H cell line will enhance detection of pathological tau beyond current capabilities.
Considering the high sensitivity of the v2H cells to detect pathological tau, we were surprised at our failure to detect seeding activity in CSF from antemortem AD subjects, as this has been reported previously by others [
31,
36,
37]. We can envision several reasons for this. First, there may be differences in the way samples were prepared and added to biosensor cells. Alternatively, though we have never observed cross-reactivity between tau and other amyloid proteins in the biosensor cells, other factors present in CSF samples could trigger tau aggregation, rendering false positive results.
Another possibility is that published RT-QuIC assays of tau seeding exceed the cell-based biosensor system in sensitivity. RT-QuIC assays have demonstrated tau seeding in post-mortem CSF, though with less sensitivity than in brain [
30,
31]. Post-mortem CSF may contain intracellular tau released after death, and thus pre-mortem CSF is a more accurate reflection of clinical utility. A 4R RT-QuIC assay sensitive for PSP and CBD seeds showed higher mean signal in groups of PSP and CBD pre-mortem CSF relative to a group of controls but not with sufficient sensitivity and specificity to apply clinically [
31]. Sensitivity of the RT-QuIC assays is demonstrated by consistent amplification from extremely dilute samples of brain homogenate. By combining IP with the v2H seeding assay we detected seeding from 31.6 pg of total brain protein diluted in 1 ml of CSF, representing a dilution factor of about 10
9. This is very similar to the published sensitivities of 3R and 4R tau RT-QuIC assays for seeding from AD, Pick, PSP, and CBD brain homogenates [
29‐
31].
Finally, it is possible that seed-competent conformations of tau are confined to the brain and do not enter the CSF. To have seeding activity, tau species must contain portions of the RD, which forms the core of all known tau fibrils [
9,
11]. Quantification of tau sequences by MS indicates that the proline-rich mid-region (aa197-243) of tau is most abundant in human CSF while sequences within the RD are much rarer [
2,
3]. Similar results have been obtained using IP-MS [
8], and plate-based immunoassays with antibody pairs targeting different regions [
12]. This pattern of fragments differs considerably from that found in the brain. Stable isotope labeling kinetic studies of tau metabolism and turnover in human neurons have found a regulated truncation and secretion of tau species containing only N-terminal and mid-regions [
33], which may explain the observed pattern of CSF tau fragments. While total tau levels in the CSF can rise due to passive release with neuronal death, such as in acute stroke [
19], elevated CSF tau in AD patients represents truncated, rather than full-length species [
3], indicating that it is likely driven by differences in processing and secretion. Seed-competent tau can be released into the extracellular fluid in cell culture models [
26,
41], but so far, except for one study [
37], it has been difficult to document seeding activity in the interstitial fluid, let alone the CSF.
While RD fragments probably constitute a minority of tau species in human CSF, several studies have detected appreciable concentrations, reporting levels in the low pg/ml [
2,
3,
8,
33]. Notably, Blennow et al
. recently published a novel Simoa assay of tau fragments ending at amino acid 368 (tau368), the C-terminus of RD. They found concentrations of tau368 in CSF from AD patients averaging about 20 pg/ml [
5]. Due to the highly fragmented nature of CSF tau, previous MS-based studies may have been limited by immunocapture with antibodies against N-terminal or mid-region epitopes, or with low affinity for RD fragments. Recently, Horie et al. have used an antibody-independent method of solid-phase protein extraction paired with MS to measure ng/ml concentrations of tryptic tau RD fragments in CSF. Importantly, these tryptic fragments do not reflect the native fragmentation pattern in CSF, and while three RD fragments were elevated in AD CSF relative to control, they differed in their concentration in CSF and correlation with disease stage. This suggests that the RD fragments in CSF may differ from seed-competent RD in the brain and might not contain the intact AD amyloid core region.
If even a tiny fraction of RD-containing tau species leaks from the brain into the CSF, the ultrasensitive assay should detect seeding in AD. Horie et al
. measured tau RD fragments in the CSF at 400 pg/ml. Thus it is interesting that The v2H biosensor detects seeding from as little as 31.6 pg of total protein from an AD brain, of which 0.89% is tau as measured by ELISA. This corresponds to a concentration of 281 fg/ml of AD-derived tau in the spiked CSF sample. Therefore we would expect to detect seeding if the RD tau fragments in AD CSF were at all representative of tau in the brain. The lack of detection suggests a mechanism acting specifically on seed-competent tau to sequester it in the brain. Seed-competent conformations of tau RD may be less likely to enter the CSF due to their propensity to aggregate, analogous to the decrease in CSF Aβ42 in AD [
22], or due to a binding affinity for HSPGs on the surface of cells [
20]. This may limit the usefulness of tau seeding as a biomarker. Regardless, the ability of the v2H cell line to quantify trace amounts of tau seed from biological specimens may prove useful in further studies of tauopathy.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.