Recent advances in cellular biosensor technology to investigate tau oligomerization

Abstract Tau is a microtubule binding protein which plays an important role in physiological functions but it is also involved in the pathogenesis of Alzheimer's disease and related tauopathies. While insoluble and β‐sheet containing tau neurofibrillary tangles have been the histopathological hallmark of these diseases, recent studies suggest that soluble tau oligomers, which are formed prior to fibrils, are the primary toxic species. Substantial efforts have been made to generate tau oligomers using purified recombinant protein strategies to study oligomer conformations as well as their toxicity. However, no specific toxic tau species has been identified to date, potentially due to the lack of cellular environment. Hence, there is a need for cell‐based models for direct monitoring of tau oligomerization and aggregation. This review will summarize the recent advances in the cellular biosensor technology, with a focus on fluorescence resonance energy transfer, bimolecular fluorescence complementation, and split luciferase complementation approaches, to monitor formation of tau oligomers and aggregates in living cells. We will discuss the applications of the cellular biosensors in examining the heterogeneous tau conformational ensembles and factors affecting tau self‐assembly, as well as detecting cell‐to‐cell propagation of tau pathology. We will also compare the advantages and limitations of each type of tau biosensors, and highlight their translational applications in biomarker development and therapeutic discovery.

microtubule stability by binding multiple different molecules. 5,6 Under pathological conditions such as abnormal posttranslational modifications, pathogenic mutations or hyperphosphorylation, tau misfolds with conformational changes, accumulates in the cytosol and initiates the fibrillogenesis cascade. 4 The aggregation pathway initiates with the spontaneous formation of tau oligomers from monomers and subsequently nucleates into paired helical filaments, and eventually intracellular NFTs (Figure 1). While the large insoluble NFTs have been the histopathological hallmark of AD and tauopathies, the soluble tau oligomers that are formed prior to fibril formation has been proposed to be the principal toxic species in recent studies. 7,8 These toxic tau oligomers promote cellular cytotoxicity [9][10][11] and induce cognitive deficits and neurodegeneration in animal models. [12][13][14][15] As a result, there is a shift in the therapeutic paradigm to inhibit or disrupt the formation of toxic tau oligomers, rather than the large insoluble fibrillar aggregates. [16][17][18][19] Tau oligomers exist as a heterogeneous ensemble of distinct assemblies with molecular diversity including both fibrillization competent and resistant, toxic, and nontoxic species. [20][21][22] Toxic tau oligomer contributes to tau pathology by initiating tau aggregation, inducing toxicity and propagating tau species through cell-to-cell spreading ( Figure 1). 23 Toxicity arises from cellular dysfunctions such as apoptosis induced by activated caspase and mitochondrial impairments, which impede synaptic energy production and lead to neuronal death. 24,25 The cell-to-cell spreading phenomenon is characterized by the ability of cells to secrete and uptake tau in the naked form 26,27 or through other vesicles such as exosomes. 28 In terms of tau aggregation, it is important to note that wild-type (WT) full-length tau is resistant to fibrillization and forms mostly soluble oligomers. [29][30][31] On the other hand, tau mutants and truncated forms of tau are more prone to oligomerization, with a higher tendency of forming insoluble tau aggregates or fibrils. 32 The most promising approach to target toxic tau oligomers will be to take advantage of the available knowledge on the conformational ensembles or structures of these species. 33,34 A substantial number of studies have made use of purified recombinant tau oligomers that are assembled in vitro to explore their biophysical properties (e.g., conformational changes and protein-protein interactions [PPIs]) and toxicity. 9,11,[35][36][37][38][39][40][41][42][43][44] While several studies have shown that WT fulllength tau forms dimers and trimers spontaneously due to disulfide bond formation, 35,36,38,[42][43][44] others have generated self-assembled tau oligomers in the presence of aggregation-prone mutations or truncations, or with the help of aggregation inducers such as heparin or tau seeds. 9,11,37,[39][40][41] However, it is often difficult to control the extent of aggregation to obtain toxic oligomers with these inducers as fibrils or a mixture of oligomers and fibrils may be formed, which will interfere with the investigations of tau oligomers. In addition, established protocols to generate tau oligomers and aggregates from purified proteins have been shown to produce different tau assemblies depending on aggregation conditions, without specific toxic tau species being identified. 45 Furthermore, it is important to note that although some of these purified tau oligomers are capable of inducing toxicity, they lack numerous chaperone proteins present in cells, and hence do not recapitulate tau oligomerization in the cellular environment. 46 Several cell-based studies have illustrated that tau oligomerization and accumulation in cells result in neurotoxicity and cell death. 25,[47][48][49] Cellular assays are also responsive to indirect pathways and various posttranslational modifications such as phosphorylation and methylation, 6 which play key roles in determining the formation of toxic tau oligomers. 50 Therefore, the biophysical and biochemical characterization of soluble tau oligomeric species in the cellular context is necessary and will provide insights to the disease mechanism. 51,52 However, these approaches are far less adopted, in part because of the lack of tools available for such studies. In this review, we will present the use of different biophysical strategies  (Table 1). We will also discuss the applications of these cellular biosensors to study cell-to-cell propagation of tau pathology and their potential as translational tools for biomarker development and therapeutic discovery. This information will provide insights to the understanding of the heterogeneity of tau oligomers and their role as molecular targets for therapeutic development of AD and related tauopathies.

| FRET-BASED BIOSENSORS
FRET is a process by which energy is transferred from a donor fluorophore to an acceptor fluorophore when an FRET pair, such as green and red (GFP/RFP) or cyan and yellow (CFP/YFP) fluorescent proteins, is in close proximity. An FRET pair is individually fused to interacting proteins and exhibits FRET when they are brought to a distance of less than 10 nm apart. 53,54 The different forms of cellular tau FRET biosensors engineered (Table 1) and the different methods of FRET measurements will be discussed below.

| Acceptor photobleaching-based FRET microscopy
The FRET technique to investigate tau-tau intermolecular interaction was initially introduced by expressing CFP or YFP tagged WT 0N4R or caspase-cleaved (ΔAA421-441) tau proteins at either N-or Cterminal in living human embryonic kidney 293 (HEK293) cells. 55 Through the acceptor photobleaching method, similar levels of FRET signals are detected in both WT CFP-tau/YFP-tau and CFP-tau/tau-YFP expressing cells in the absence of aggregation inducers (Figure 3 (a)). This indicates spontaneous oligomerization of WT tau, independent of the positions of fluorophore tagging, as an initial step in the aggregation cascade. The similar FRET observed from differential fluorophore tagging suggests that tau is adopting different conformations in the formation of oligomers and the presence of conformational ensembles in tau oligomerization. In the same study, a higher FRET signal is detected in the presence of glycogen synthase kinase 3 beta (GSK3β) co-expression with WT tau, illustrating a higher aggregation propensity of hyperphosphorylated tau which is confirmed by biochemical assays. Interestingly, a similar level of higher FRET is observed in caspase-cleaved tau which is more resistant to GSK3βmediated phosphorylation but known to form insoluble thioflavin-S (ThS) positive inclusions in cells (Figure 3(a)). 56 Indeed, the analysis of cell lysates shows that WT tau forms sarkosyl soluble fractions in native cellular environment, while the caspase-cleaved tau proteins are sarkosyl insoluble (Figure 3(b)), indicating that FRET originates from respective tau species. This is consistent with previous studies showing that truncated tau aggregates more rapidly than WT tau and has been reported to promote fibril formation, especially under aberrant phosphorylation. 57 It is worth emphasizing that the similar FRET levels obtained with different species and treatment conditions suggests the need to stringently resolve the exact species that corresponds to the observed FRET in order to accurately interpret the signals.

| Lifetime-based FRET measurement
Recently, we engineered two cellular WT 2N4R tau FRET biosensors to monitor intramolecular (GFP-tau-RFP) (Figure 2(a)) and intermolecular (tau-GFP/tau-RFP) (Figure 2(b)) interactions. We show efficient FRET signals through lifetime measurements in HEK293 cells expressing the biosensors, 58 recapitulating the global folding of the tau proteins in the intramolecular system 59,60 and their spontaneous oligomerization in the intermolecular system ( Figure 3(c)). We note that the FRET efficiency reflects the average of the ensemble intermolecular and intramolecular proximity between tau molecules which potentially contains multiple different conformations and interaction states. More importantly, there is no puncta formation in cells expressing the WT tau biosensors and they are not ThS positive, hence confirming that these are soluble tau oligomers and do not contain β-sheet species (Figure 3(d)). 58 This is also consistent with previous studies showing that WT tau is resistant to fibrillization and forms mainly tau oligomers. 29 Furthermore, we also expressed fluorophoretagged tau with P301L mutation (tauP301L-GFP/tauP301L-RFP) in both HEK293 and SH-SY5Y cells and show that higher FRET signal is observed in tauP301L FRET biosensor than the WT tau biosensor, indicating the higher propensity of oligomerization for tauP301L. 58 Importantly, we have also observed an increase in FRET with treatment of recombinant tau proteins to the biosensor cells, suggesting the possibility of using our biosensors in seeding experiments. As overexpression of fluorophore-tagged tau in HEK293 cells does not F I G U R E 1 Tau fibrillogenesis cascade and cell-to-cell propagation of tau pathology in Alzheimer's disease. Misfolded tau species is capable of forming both nontoxic and toxic soluble tau oligomers spontaneously. The tau oligomers can proceed to form paired helical filaments (PHFs) and neurofibrillary tangles (NFTs) which are large insoluble aggregates with β-sheet conformations. The fibrillar species can be secreted by host cells and transmitted to recipient cells which is capable of inducing further seeded oligomerization or aggregation, leading to cell-to-cell propagation of tau pathology. Schematics are created with BioRender.com illustrate toxicity, we overexpressed unlabeled WT tau or tauP301L in SH-SY5Y cells to study the effect of tau oligomers in cell cytotoxicity. 58 Overexpression of WT tau does not induce significant cell cytotoxicity but overexpression of tauP301L in SH-SY5Y cells induces significant toxicity, consistent with other observations. [61][62][63][64] Interestingly, it has been shown that overexpression of tauP301L does not induce fibril formation in SH-SY5Y cells, 65 suggesting that tauP301Linduced toxicity is due to toxic oligomers. The unlabeled tauRD aggregates are also shown to be released from one cell population and taken up by recipient tau biosensor cells to induce aggregation as illustrated through an increase in FRET. This indicates seeding 67,68 and the propagation of seeds between cells. 66,69 The use of the tauRD FRET biosensor has also been extended to the study of the initiation of pathological aggregation beginning with conversion of inert tau monomer to a seedcompetent form based on increasing accessibility of the hexapeptide motifs (VQIINK/VQIVYK) that promote aggregation. 21 This is consistent to a report that pathogenic tau mutations, alternative splicing and proline isomerization are all capable of destabilizing the local structure proximal to the hexapeptide motifs and triggering spontaneous aggregation as illustrated by an increase in FRET of the tauRD biosensor. 70 More recently, the tauRD FRET biosensor is applied to study the seeding capability of tau oligomers. Purified oligomeric assemblies containing 3-mer, $10-mer, and $20-mer as well as fibrils have been shown to increase FRET of the tauRD biosensor. 71 Treatment of heparin or a heparinoid compound (SN7-13) inhibits seeding by both oligomeric and fibrillar species and reduces FRET of the biosensor. 71 The development of these cellular FRET biosensor technologies provide the platforms to study the aggregation cascade of tau proteins, including monomers (intramolecular doubly labeled tau or F I G U R E 2 Schematic representation of tau biosensors based on fluorescence resonance energy transfer (FRET), bimolecular fluorescence complementation (BiFC), or split luciferase complementation (SLC) for monitoring intramolecular and intermolecular tau interactions in living cells. (a) Tau intramolecular FRET biosensor where FRET is observed when there is global folding of wild-type (WT) monomeric tau. 58 (b) Tau intermolecular FRET biosensors where FRET is observed when tau oligomers or aggregates are formed. WT tau is used for the formation of nonβ-sheet soluble tau oligomers and tau repeat domain (tauRD) with P301S mutation or truncated tau is used for the formation of β-sheet insoluble tau aggregates. 55,58,66 (c) BiFC tau fluorescence turn-off biosensor where fluorescence is absent when there is tau oligomerization or aggregation. 78,79 (d) BiFC/SLC tau fluorescence/luminescence turn-on biosensor where fluorescence or luminescence is present when tau oligomers or aggregates are formed. 80,87,89 Tau oligomer is drawn as a dimer for illustration but it can be any species more than a dimer (≥2-mers)

| Split-GFP complementation (fluorescence turn-off biosensor)
To create a tau fluorescence turn-off biosensor, Johnson group adopted a split-GFP complementation approach by fusing WT 0N4R tau protein directly to a small nonfluorescent GFP fragment (GFP11), and co-expressing in HEK cells with a large nonfluorescent GFP fragment (GFP1-10). 78,79 When tau exists as a monomer or low degree aggregate, the complementary large GFP fragment is able to access the small GFP fragment fused to tau, assembling the fluorescently active GFP. The reconstitution of active GFP is prohibited primarily due to steric hindrance when tau aggregates, leading to a decrease in GFP fluorescence intensity in cells. 78 Besides WT tau, the fluorescence turn-off approach was also applied to truncated tau (ΔAA421-441) and tau-2EC with two Ser-to-Glu mutations at Ser396/S404 to mimic phosphorylation of these sites. The results show that truncated tau and tau-2EC proteins are accumulating in cells to a greater extent with reduced GFP fluorescence as compared to WT tau in the absence of phosphorylation inducers. 78 The presence of active GSK3β significantly increases the GFP intensity of WT, which strikingly corresponds directly to the increased expression of WT tau as illustrated through immuno- blots. An increase in the fluorescence intensity of the turn-off biosensor should correspond to less aggregation with more monomers, but this is complicated with an increase in tau expression. Therefore, although active GSK3β is used, it is unclear from the fluorescence turn-off biosensor whether extensive phosphorylation of WT tau leads to an enhanced oligomerization or aggregation as these should have led to a decrease in GFP intensity. Conversely, no change in GFP intensity is observed for truncated tau regardless of phosphorylation and no change in tau expression is observed. 78 This is a conflicting observation from the FRET study where the truncated tau shows higher FRET signal with active GSK3β expression, which we will expect a decrease in GFP intensity in the split-GFP tau biosensor to indicate increasing aggregation; however, this is not observed in this study. Interestingly, tau-2EC, which is pseudophosphorylated at S396/S404 and is not efficiently further phosphorylated by active GSK3β, exhibited significantly decreased GFP intensity in the presence of active GSK3β, indicating that there is increased aggregation. 78 This is suggested to be due to the formation of sarkosyl insoluble aggregates and enhanced phosphorylation at other residues besides S396/S404. We should also note that this method is only feasible when there is an initial self-assembly between the two parts of the split-GFP to form monomer signals of GFP fluorescence, but the intrinsic affinity between these two GFP parts may affect the aggregation propensity of the protein of interest.

| Split-Venus BiFC (fluorescence turn-on biosensor)
In  Figure 4(a)) which follows a time-course-dependent manner (Figure 4 (b)). 80  To bring this important BiFC tool in vivo to directly monitor tau self-assembly in the mouse brain, a novel tau transgenic mouse expressing tauP301L-BiFC fluorescence turn-on biosensor was recently generated (Figure 4(c)). 87 Initially, there is a significant increase in BiFC fluorescence in the mouse brain in the first

| SLC-BASED BIOSENSORS
To improve on the BiFC assay, tau SLC assays are developed. SLC assays possess enormous dynamic range with superior detection sensitivity and more dynamic reversibility of the complementation. 88 Examples of SLC assays are detailed below with different types of split luciferase used in each assay (Table 1).

| Gaussia-Luc SLC
To generate an SLC assay, the Hyman group used WT 2N4R tau to fuse to either the N-or the C-terminal region of Gaussia luciferase (gLuc), forming a split-gLuc. 89 The formation of tau oligomers enables the complementation of two separate parts of split-gLuc, namely tau-L1 (residues 1-92) and tau-L2 (residues 93-163), which then reconstitutes gLuc activity. Robust gLuc activity is observed in HEK293 cells with co-expression of tau-L1 and tau-L2, as well as in the conditioned medium after 40 h of transfection ( Figure 5(a)), without observed toxicity from the tau split-gLuc expression. 89 The control, L1 and L2 only without fusing to tau, does not result in luciferase activity, indicating specific tau-tau self-association and oligomerization in the SLC assay. Linear correlation between split-gLuc activity and tau concentration in conditioned medium indicates an excellent assay sensitivity of 7.5 pg/ml tau-L1/L2 which is equivalent to 0.16 nM full-length tau as characterized by human total tau ELISA ( Figure 5(b)). More importantly, these tau species are not thioflavin-T (ThT) positive, indicating that tau-L1/L2 activity originates mostly from dimers and oligomers but not insoluble tau fibrillar species with β-sheet. 89 Furthermore, the tau-L1/L2 complementation is enhanced with addition of aggregation inducers, including heparin, preformed fibrils, 90

| NanoLuc SLC
Using a similar approach, a NanoLuc SLC assay is generated with Nand C-terminal of WT 2N4R tau fused to two different NanoLuc luciferase subunits (an 18 kDa polypeptide (large Bit) and a 1.3 kDa peptide (small Bit)) respectively. 92 NanoLuc activity is observed in human liver HuH-7 cells transfected with the two tau-NanoLuc fusion constructs for 24 h, illustrating the reconstitution of luciferase activity based on tau protein self-association and oligomerization. 92 The NanoLuc activity is reduced with treatment of protein kinase inhibitors, suggesting that the tau oligomerization is initiated by selfassociation of phosphorylated tau monomers and that the kinase inhibitors are either directly inhibiting tau phosphorylation or indirectly affecting alternative mechanisms. 92 LO

| Click beetle green-Luc SLC
In another study to investigate tau propagation and spontaneous internalization for cell-to-cell spreading of tau pathology, a tau SLC assay based on click beetle green luciferase (cbgLuc) is developed. 93 The N-and C-terminal halves of the cbgLuc are fused to the C-terminal of the tauRD containing P301S mutation, and the tau split-cbgLuc illustrates a basal cbgLuc activity upon spontaneous tau dimerization and oligomerization. 94,95 Interestingly, the luciferase signal is increased with treatment of exogenous tau oligomers (n ≥ 3) and fibrils, suggesting the minimal tau assembly capable of spontaneous cell uptake and seeding ( Figure 5(d)). 94

| TRANSLATIONAL APPLICATIONS OF CELL-BASED TAU BIOSENSORS
The cellular tau biosensors are continuously being used in biomarker development as well as therapeutic discovery such as identification of small molecules or antibodies that disrupt toxic tau interactions or alter tau conformational ensembles. In this section, we will discuss the sensitivity of each type of tau biosensor in detecting seeding activity and examples of tau biosensors used as translational tools in highthroughput screening (HTS) drug discovery.

| Detection of seeding activity as a biomarker for AD pathology
High seeding activity in a biological sample can be a new biomarker for a subset of subjects that are more likely to develop symptoms of pathological tau aggregation. The cellular tau biosensors have been widely used to characterize the seeding capability of in vitro protein samples and biological samples from both mice and human.

| TauRD FRET biosensor
The tauRD FRET biosensor is highly sensitive and specific to the detection of isolated tau oligomeric species and fibrillar aggregates from human patients and AD transgenic mouse brains that are seedcompetent. [96][97][98] Specifically, tauRD FRET biosensor has been used to test the tau seeding capacity of the soluble high-molecular-weight brain fraction containing mostly seed-competent oligomeric species from AD patients. 97,98 The FRET assay readily discriminates diseased and aged control brains, 97 and further illustrates heterogeneous seeding properties across the patients as characterized by different extents of FRET increase. 98 In another study, brain-derived tau oligo-

| TauRDΔK SLC biosensor
To examine whether tau containing exosomes derived from the cerebrospinal fluid (CSF) of AD patients are seed-competent, tauRDΔK SLC biosensor was generated with two halves of luciferase fusing to tauRD containing ΔK280 mutation and expressed in N2a cells. 28 Tau containing exosomes dramatically increase the luciferase bioluminescence signal by about 50% for AD patients and 40% for healthy controls, corresponding to an increase in tauRDΔK aggregation. 28 While exosomes from AD patients induced slightly higher aggregation than the controls, it is important to note that the difference is not statistically significant. 28 This suggests that exosomes from both AD and control CSF may contain different species of tau oligomers which are capable of triggering tau aggregation to a different extent, recapitulating the heterogeneity in tau oligomers. 98 By comparing the efficacy of these three types of biosensors in detecting seeding activity, tau seeds activate all of them with a halfmaximal effective concentration (EC 50 ) between 0.03 and 0.1 μM, although they have different dynamic ranges of detection ( Figure 6).
TauRDΔK SLC biosensor has the greatest dynamic range with an eightfold increase in signal between nontreated and treatment of the highest concentration of tauRD seeds (Figure 6(a)). 94 This is followed by a fourfold signal increase in tauRD FRET biosensor with treatment of tauRD seeds ( Figure 6(b) 66 and a twofold signal increase in tau-BiFC biosensor with treatment of K18-P301L tau (Figure 6(c)). 99 However, we should note that the seeds used in these studies are different, which might contribute to the difference in the sensitivity or fold change observed.

| HTS drug discovery
Fluorescence and bioluminescence biosensors constitute attractive and powerful tools for drug discovery campaigns, from HTS assays, to optimization of lead compounds, and preclinical evaluation of candidate drugs. In the context of tau oligomerization and aggregation, tau biosensors have been used in HTS assays to identify small molecules that interfere with tau-tau interactions, conformational changes of tau and other biological processes that alter the propensity of tau aggregation such as tau phosphorylation.

| WT 2N4R tau FRET biosensor
We have developed an HTS platform to discover small molecules that target tau oligomerization by combining fluorescent tau In a separate study, the above-mentioned WT 2N4R tau intermolecular FRET biosensor was adopted and expressed in SH-SY5Y cells to perform HTS on ChemBridge DIVERSet library of 10,000 compounds to identify 195 small molecules that inhibit tau oligomerization. 108 This small subset of compounds was then filtered by • SLC assays generate low background signals because they do not require external light. Oxidation of chemical substrate by the luciferase emits detectable light known as bioluminescence • SLC assays possess a binary on-or-off characteristic that increases the detection sensitivity • SLC assays can detect both association and dissociation of protein pairs • Kinetic measurements of PPI are permitted • Large and linear dynamic range of seven to eight orders of magnitude • Relatively rapid turnover of the enzyme allows for detection of high luciferase activity • Requirement for exogenous substrates to generate bioluminescence and potential false-negative results in the presence of a luciferase inhibitor • Unlike stable fluorescent signals, the bioluminescence signals in SLC assays gradually decrease because the luciferaseluciferin reaction is a chemical reaction • There is a time delay of 4-6 h from stimulus to response to allow transcription to occur • SLC assay is not suitable for monitoring PPI over a long time (>5 h) • Unable to measure physical distances between a protein pair as the minimum physical distance required to obtain an interaction signal in SLC is unknown Abbreviations: BiFC, bimolecular fluorescence complementation; FP, fluorescent protein; FRET, fluorescence resonance energy transfer; GSK3β, glycogen synthase kinase 3 beta; PPIs, protein-protein interactions; SLC, split luciferase complementation.

| WT 2N4R tau-BiFC biosensor
The Venus-based tau-BiFC biosensor was used to screen a FDAapproved drug library containing 1018 compounds, and identified levosimendan as a potent small molecule inhibitor of tau oligomerization (Figure 7(c)). 111 Levosimendan directly inhibits disulfide-linked tau oligomerization by covalently binding to tau cysteines as characterized by 14 C-isotope labeling of the small molecule. In addition, levosimendan prevents the formation of oligomers from monomers and is also capable of disassembling preformed tau oligomers or aggregates into monomers (Figure 7(d)), thus rescuing neurons from tauinduced toxicity. 111 The study also made a comparison between levosimendan and the well-known inhibitors of tau aggregation, methy- promising approaches to probe first-hand molecular insights of tau aggregation cascade and cell-to-cell spreading of tau pathology. [112][113][114] However, they differ in terms of the type of interactions they can detect, sensitivity and signal-to-noise ratio, the possibility of false-positive or false-negative signals, capability of spatiotemporal monitoring of interactions, and the instrumentation needed. 77,115,116 We have summarized the advantages and limitations of each technique (Table 2). In general, expressing these tau biosensors in living cells have the key advantage of preserving the native surroundings, such as the presence of molecular chaperones or other molecular constituents, in which the interaction takes place and is monitored. It has been shown that fluorophore tagging could prevent tau toxicity and nontagged tau should be used for toxicity studies. This also suggests that a common limitation of these biosensors is that they are not capable of monitoring tau-induced toxicity in cells by themselves, although treatment of exogenous tau oligomers or aggregates to the biosensors has been shown to induce cell cytotoxicity. 99 In the technical aspects, all three types of biosensors can be used for timecourse analysis and HTS assays, and they are sensitive to changes in local environment.

| Bimolecular fluorescence complementation
The BiFC approach allows for tagging of smaller fluorophores, a low background fluorescence as noise and signal detection even at low level of protein expression. 117,118 The multicolor BiFC assay allows concurrent imaging of several protein complexes within the same cell such as tautau interactions and interactions between tau and other proteins which might change tau aggregation propensity. 117 However, the process of fluorophore fusion is not reversible. Once a transient complex between the protein of interest is formed, the reconstituted fluorescence protein remains relatively stable, hence reducing its capability in continuously monitoring the tau aggregation and disaggregation process. 118 In addition, the split fluorescent protein might autonomously and spontaneously assemble, which will alter the binding properties of interacting tau and increase the false-positive signals. 115,117

| Split luciferase complementation
A key advantage of SLC assays is the low background signals as they do not require external light for detection and allows a binary on-oroff characteristic that increases the detection sensitivity. 119,120 It also has a large and linear dynamic range with seven to eight orders of magnitude. 121 However, the presence of a luciferase inhibitor may generate false-negative results. 122 In addition, it is unable to measure distance between proteins and is not suitable for monitoring protein interactions over a long time as the bioluminescence signals gradually decrease over time. 119

| Fluorescence resonance energy transfer
FRET is powerful in determining spatiotemporal dynamics and reversibility of PPIs instantaneously at real time. 54,118 Besides measuring direct tau-tau intermolecular interactions, it also allows measurements of tau intramolecular folding or conformational change. 58 However, it requires large fluorophore fusion which may cause steric hindrance. 123 In addition, it has disadvantages of irreversible photobleaching, being less sensitive due to small fractions of interacting proteins combine with each other at any time, and insensitive to distance outside the dynamic range of the detection of the FRET pair. 124,125 Despite these limitations, important recent advances in lifetime-based FRET measurements allow for data fitting of the acquired fluorescent waveforms with sufficient precision to analyze two or more samples having different lifetimes, and resolving multiple components with high accuracy with respect to both lifetime and mole fraction. 126,127 The potential of resolving the different populations of protein species and the distance distributions of PPIs through model fitting of the time-resolved FRET waveforms containing high-content information 127 can be in theory applied to the investigation of tau oligomers and aggregates. However, a major limitation of tau oligomer associated time-resolved FRET data lies in the lack of information such as the number of interacting tau monomers, the stoichiometry of the tau oligomer or well-defined structural states required to constrain the model in order to identify specific toxic species. This suggests the need to perform additional complementary biophysical and spectroscopic characterizations such as analytical ultracentrifugation, nuclear magnetic resonance, electron paramagnetic resonance, surface plasmon resonance, or Raman spectroscopy to obtain oligomer size and binding affinity, although these techniques are mostly conducted with purified proteins in current studies and may not be directly correlated to cellular observations.

| CONCLUSION AND FUTURE PERSPECTIVES
The advancements in the cellular biosensor technology allow for the studies of molecular mechanisms involved in tau oligomerization and aggregation, as well as in cell-to-cell spreading of tau pathology. The tau biosensors can further be used to study the effect of different tau isoforms, mutations, and posttranslational modifications on the formation of toxic tau oligomers. It might be important to perform cross-validation experiments with different biosensors to ensure an observation or finding is consistent and reproducible with different tools. It is also imperative to compare biophysical observations from cellular models to tau oligomers extracted from AD mouse models or patients to elucidate the structural states of the true toxic tau oligomeric species. The use of FRET, BiFC and SLC assays can also be used to study cross reactivity between tau and other intrinsically disordered proteins such as β-amyloid 128 and α-synuclein. 129 To improve on the physiological relevance, the tau biosensors can be expressed in human induced pluripotent stem cells. Furthermore, future drug discovery campaigns using tau biosensors should include screening of CNS-focused compound libraries to ensure tau-targeted small molecules have a high probability of crossing the blood-brain barrier for more effective treatment of AD and related tauopathies.

ACKNOWLEDGMENTS
The author would like to thank Dr Jialiu Zeng for proofreading the manuscript. This study is supported by a NUS Development Grant to C. H. L. from the National University of Singapore.