Here, we show that pS214 and pS324 phosphorylation-dependent binding of 14-3-3 increases Tau solubility and decreases Tau aggregation in neurons. Molecularly, these effects are based on (i) 14-3-3 mediated “scavenging” of TaupS214/pS324 molecules from the MT surface based on Tau binding competition between MTs and 14-3-3, and (ii) the inhibition of TaupS214/pS324 condensation, potentially en route to the formation of pathological Tau species3,55, and amyloid-like aggregation by stoichiometric 14-3-3 binding. However, sub-stochiometric 14-3-3 concentrations promote Tau condensation and may therefore increase the risk for Tau aggregation. Together, our data elucidate the importance of 14-3-3 proteins in modulating Tau pathobiology, in which the formation of stoichiometric, soluble 14-3-3:TaupS214/pS324 complexes reduces the risk of unchaperoned phospho-Tau accumulation and spontaneous aggregation in the cytosol at the earliest step.
In the healthy brain, a robust occupation of soluble (not bound to MTs) phospho-Tau molecules by 14-3-3 and other chaperoning proteins could be essential for efficient prevention of Tau aggregation. In contrast, inhibition of chaperone binding could enable or enhance phospho-Tau aggregation.
To test whether the binding of 14-3-3 proteins can suppress neuronal Tau aggregation, we expressed pro-aggregant FTD-mutant human Tau (eGFP-Tau) in primary hippocampal mouse neurons, which leads to spontaneous eGFP-Tau aggregation into NFT-like, fibrillar aggregates in a subset of neurons (~2% NFTs in DMSO control; Fig. 1a, b). Co-immunoprecipitation from neuronal lysates confirmed the association of 14-3-3 with Tau (Supplemental Fig. S1a). Treatment with BV02, an inhibitor of 14-3-3 proteins binding to phosphorylated clients, increased the fraction of neurons with eGFP-Tau aggregates in a dose-dependent manner (~4% NFTs at 10 μM BV02, ~7% NFTs at 40 μM BV02). In contrast, increasing 14-3-3 binding by treatment with fusicoccin (50 μM) decreased neuronal Tau aggregation (~1% NFTs) and simultaneously increased the concentration (=fluorescence intensity) of soluble eGFP-Tau in the soma of neurons without aggregates (Fig. 1c). Together, these data suggested that 14-3-3 binding increases cytosolic Tau solubility, whereas reducing 14-3-3 binding promotes Tau aggregation in neurons.
To test whether 14-3-3 would colocalize with aggregated or soluble phospho-Tau, we performed immunostainings on eGFP-Tau expressing neurons. We found that, in contrast to previous reports from human brain immunostainings, 14-3-3 was mostly excluded from NFT-like eGFP-Tau aggregates in our neuron model (Fig. 1d), in which 14-3-3 showed a punctate (=granular) staining pattern in dendrites, axons, and somata (Fig. 1e), similar to previous reports from human brain without Tau NFTs.
Interestingly, Tau phosphorylated at PKA-target sites pS214 and pS324, two of the phosphorylation sites mainly relevant for 14-3-3:Tau binding, showed limited colocalization with MTs and a cytoplasmic, granular distribution, similar to 14-3-3 (Fig. 1f). In contrast, a Tau phospho-epitope previously not reported as relevant for 14-3-3:Tau binding (pS202/pT205, target sites of, e.g., Cdk5 and GSK3β) stained microtubules in naïve and eGFP-Tau overexpressing neurons (Fig. 1f). 14-3-3 granules seemed to arrange along Tau coated MTs (Fig. 1g). These data indicate that Tau binds MTs less efficiently than Tau, which could be related to its 14-3-3 binding.
Given the previously suggested link between 14-3-3, Tau and MT stability in axon development and regeneration, we established an in vitro model for the phosphorylation-dependent binding of Tau to 14-3-3, in order to probe the effect on Tau’s MT binding.
We generated phospho-null mutants and phospho-variants (by in vitro phosphorylation with purified kinases) of recombinant full-length human Tau (2N4R isoform; Fig. 2a) and determined their binding affinities to the 14-3-3ζ isoform, which is the most abundant isoform in the brain.
The PKA target sites, S214 and S324, were shown to be necessary for Tau binding to the 14-3-3σ isoform. Our phospho-null mutants therefore contained serine-to-alanine mutations at S214 and S324 — individually or together (Tau, Tau, and Tau = TauS2A). In vitro phosphorylation with PKA introduced phosphates at S214 and S324 in wild-type Tau (PKA-Tau) but not at respective mutated serine residues in Tau, Tau, and Tau, as confirmed by Western blot (Fig. 2b).
Next, we determined the binding strength between 14-3-3ζ and different PKA-Tau variants using thermal stability assays (measuring the temperature needed to denature 14-3-3:Tau complexes; Fig. 2c). 14-3-3ζ:PKA-Tau complexes demonstrated the highest thermal stability, followed by 14-3-3ζ:PKA-Tau and 14-3-3ζ:PKA-Tau. By contrast, 14-3-3ζ:PKA-Tau was no more stable than 14-3-3:non-phosphorylated Tau. Furthermore, phosphorylation of Tau by Cdk5, which phosphorylates S202/T205 (Supplemental Fig. S1b, c) but not S214/S324 (Fig. 2d), also did not stabilize 14-3-3ζ:Tau interactions, and in fact decreased stability (Fig. 2c). This confirmed previous observations that, among all PKA phosphorylation sites in Tau, phosphorylation of S214 and S324 is individually and synergistically important for 14-3-3ζ:Tau binding.
To further assess whether the differences in thermal stability arose from different 14-3-3ζ:Tau binding affinities, we performed Tau phospho-peptide (pS2) displacement assays. Here, we titrated full-length Tau variants into pS2:14-3-3ζ and measured the reduction in fluorescently-labeled pS2 anisotropy resulting from its displacement from the complex. The pS2 Tau phospho-peptide consisted of two short Tau sequences with phosphate groups at S214 and S324, connected by an unstructured linker (pS2: 38 aa; Tau-GGGSGGGSGGG-Tau; Fig. 2a and Supplemental Fig. S1d). pS2 showed efficient binding to 14-3-3ζ (K ~1 M), which also synergistically depended on phosphorylation at S214 and S324, as determined by fluorescent anisotropy and thermal stability assays, (Fig. 2e, f and Supplemental Fig. S1e). In the performed displacement assays (Fig. 2g and Supplemental Fig. S1f), pS2 was efficiently out-competed by full-length PKA-Tau, to a lesser degree by PKA-Tau and PKA-Tau, but not by PKA-Tau or non-phosphorylated Tau. Cdk5-Tau was also not able to displace pS2 from 14-3-3ζ.
We note that PKA-Tau contains multiple phosphorylated residues in addition to pS214 and pS324, as demonstrated in phos-tag gels by a pronounced upshift compared to non-phosphorylated Tau (Supplemental Fig. S1g). These additional phosphorylation sites should also be present in non-binding PKA-Tau. Cdk5-Tau also contained phosphorylation at multiple sites (upshift in phos-tag gel), but not at pS214 and pS324 (Fig. 2d), and did not bind 14-3-3ζ. Therefore, additional Tau phosphorylation sites other than pS214 and pS324 may have less impact on Tau’s binding to 14-3-3ζ.
Together these data confirm the previously reported necessity of Tau phospho-epitopes pS214 and pS324 for the binding to 14-3-3 and the relevance of PKA activity in this process, here for the 14-3-3ζ isoform (Fig. 2h). Phospho-epitopes pS214 and pS324 are located in regions relevant for Tau aggregation and microtubule binding; however, these specific phospho-sites have not been associated with Tau pathology in disease. We thus suggest that 14-3-3 proteins bind physiological soluble rather than pathological aggregated Tau molecules and thereby promote their solubility.
Non-phosphorylated Tau has a high affinity to MTs (Tau:MTs K ≈1 μM), but the binding affinity drops upon phosphorylation of serine and threonine residues in and around TauRD. For example, PKA-Tau binds MTs with K ≈10 μM. Since we found that 14-3-3ζ has a tenfold stronger affinity for PKA-Tau (K ≈1 μM; Fig. 2f), we hypothesized that 14-3-3 would compete for PKA-Tau binding with the MT surface. Sequestration of PKA-Tau from MTs by 14-3-3ζ could explain why Tau staining was absent from MTs in neurons (Fig. 1f).
To test this idea, we performed in vitro MTs pelleting assays, in which MTs were polymerized in the presence of PKA-Tau or unphosphorylated Tau and in the absence or presence of 14-3-3ζ. After centrifugation, the pellet (P) fraction contained MTs and MT-bound Tau, whereas the supernatant (S) contained unbound Tau and 14-3-3. Indeed, the fraction of unbound PKA-Tau (increased S/P ratio) was significantly higher in the presence, compared to the absence, of 14-3-3ζ (Fig. 3a, b and Supplemental Fig. S2a). The fraction of unbound non-phosphorylated Tau was unaffected by the presence of 14-3-3ζ, as expected in the absence of 14-3-3ζ:Tau binding. These results thus indeed suggest that 14-3-3 reduces MT binding for PKA-Tau but not Tau (Fig. 3c).
Intrigued by these findings, we aimed at visualizing the effect of 14-3-3ζ on the binding of PKA-Tau to MT in vitro by confocal microscopy using fluorescently-labeled Tau and 14-3-3ζ variants (1-2% labeled proteins). We first polymerized MTs from tubulin-Alexa594 in the presence of non-phosphorylated Tau, PKA-Tau, or PKA-Tau (1% fluorescently Dylight488-labeled proteins). Surprisingly, we observed colocalization of all three Tau variants with MTs (Fig. 3d-f), indicating that PKA-phosphorylation alone was not sufficient to detach Tau from MTs, despite the decrease in MT binding affinity for PKA-Tau (K ≈10 μM) compared to non-phosphorylated Tau (K ≈1 μM). In fact, the results from the MT pelleting assay also suggested that >50% (S/N ratio >1) of PKA-Tau was still bound to MTs (Fig. 3b).
When performing MT imaging in the presence of 14-3-3ζ, however, non-phosphorylated Tau (not binding 14-3-3ζ) still co-localized with MTs, whereas PKA-Tau (binding 14-3-3ζ) was now absent from MTs (Fig. 3g, h). These observations supported our hypothesis that 14-3-3ζ binding promotes the dissociation of PKA-Tau from MTs.
To prove that this was due to phospho-site specific binding of 14-3-3ζ, and not general electrostatic of unspecific interactions, we used a combination of non-binding mutants: PKA-Tau, which contained PKA-induced phosphorylation except at pS214 and pS324, and 14-3-3ζ, which contained an arginine-to-alanine mutation in its binding pocket and therefore binds phospho-Tau with lower affinity. In this non-binding combination, PKA-Tau remained on MTs (Fig. 3i), confirming that the binding of 14-3-3ζ was needed to sequester PKA-Tau from MTs. These findings may explain how overexpression of 14-3-3 can lead to a destabilization of the neuronal MT cytoskeleton, i.e., by excessive sequestration of phospho-Tau from MTs.
When imaging MTs, we noticed that all Tau variants formed condensates with free tubulin that attached to polymerized MTs (Fig. 3d-f), likely formed from excess, unbound Tau, which recruits free tubulin. We previously reported similar observations. When adding 14-3-3ζ to PKA-Tau:MT preparations, only a few and smaller PKA-Tau:14-3-3ζ condensates could be observed (Fig. 3g), some of which also contained tubulin. In contrast, adding 14-3-3ζ to either Tau:MT or PKA-Tau:MT preparations (both not binding 14-3-3ζ; Fig. 3f, h) promoted Tau:tubulin condensation and led to the formation of MT “junctions” with large condensates in their center. 14-3-3ζ co-partitioned into these condensates and did not coat outgrowing MTs, indicating that, when not binding Tau, 14-3-3ζ had a higher affinity to Tau condensates than to Tau bound to MTs or MT themselves. These observations were reminiscent of what we saw in neurons, where 14-3-3 granules partially aligned with MTs (Fig. 1g).
Together, the data suggest that in the absence of specific binding via Tau phospho-sites pS214 and pS324, 14-3-3 may promote Tau condensation without interfering with Tau’s MT binding. However, phosphorylation-dependent binding to 14-3-3 removes PKA-Tau from MTs via binding competition but may not promote its condensation. Notably, both Tau MT binding as well as pathological Tau aggregation have previously been reported to involve Tau condensation. Modulation of Tau condensation by 14-3-3 binding may therefore underlie the previously suggested involvement of 14-3-3 in both these processes.
To examine the mechanisms behind 14-3-3ζ modulation of Tau condensation, we tested its impact on the condensate formation of different phospho-Tau variants. Again, we phosphorylated Tau in vitro using different kinases that do (PKA) or do not (Cdk5 and Fyn) introduce phospho-groups at the crucial 14-3-3ζ binding sites S214 and S324. Increasing concentrations of 14-3-3ζ progressively inhibited PKA-Tau condensation (measured as condensate surface coverage). An equimolar concentration of 14-3-3ζ dimers (two molecules of 14-3-3ζ to one molecule of PKA-Tau fully abolished PKA-Tau condensation (Fig. 4a). Cdk5-Tau and Fyn-Tau condensation decreased by only 20% at these concentrations. This suggested that condensate inhibition was driven by specific binding of 14-3-3ζ to PKA-Tau.
To test whether 14-3-3ζ binding to specifically pS214 and pS324 was necessary to inhibit Tau condensation, we incubated recombinant PKA-Tau (binding 14-3-3ζ) or PKA-Tau (not binding 14-3-3ζ) with increasing concentrations of 14-3-3ζ (10 μM Tau variants; 2% DyLight488-labeled PKA-Tau or PKA-Tau; 2% DyLight650-labeled 14-3-3ζ) in condensation buffer (25 mM HEPES, pH 7.4, ~5 mM NaCl, 5% PEG). PKA-Tau and PKA-Tau each showed pronounced condensation in the absence of 14-3-3ζ (Fig. 4b, c), which confirmed previous observations. Increasing concentrations of 14-3-3ζ inhibited condensation of PKA-Tau but not PKA-Tau. 14-3-3ζ co-condensed with both Tau variants, but did not form condensates alone (Supplemental Fig. 2b). At equimolar concentrations of 14-3-3ζ (10 μM) and PKA-Tau (10 μM), when about half of PKA-Tau monomers are in complex with 14-3-3ζ dimers, no condensates were formed. In contrast, PKA-Tau condensation remained largely unaffected by 14-3-3ζ at this concentration ratio (Fig. 4d).
To further confirm the idea that Tau condensation can be tuned through 14-3-3 binding affinity, we modulated 14-3-3:PKA-Tau binding using 14-3-3ζ, which reduces but does not abolish PKA-Tau binding. Incubating PKA-Tau with increasing concentrations of 14-3-3ζ led to inhibition of condensation, however, to a lesser degree than observed for wild-type 14-3-3ζ (Fig. 4e, f). This further supported a reciprocal relation between Tau condensation and 14-3-3ζ:Tau binding affinity. Condensation of PKA-Tau was not affected by 14-3-3ζ. Tau condensation can thus be regulated via its 14-3-3 binding affinity.
In summary, these observations show that stoichiometric binding of 14-3-3ζ dimers to Tau monomers phosphorylated on S214 and S324 suppresses Tau condensation and therefore increases Tau solubility. This effect seemed to be largely independent of other Tau phosphorylation sites. Interestingly, suppression of biomolecular condensation by 14-3-3 binding has been suggested for other physiological and disease-associated 14-3-3 binding partners.
Co-condensation of 14-3-3ζ with phosphorylated Tau was previously reported, but most details of the molecular mechanisms and the connection to 14-3-3:Tau binding remain unclear. Our data showed that stoichiometric 14-3-3ζ binding suppresses PKA-Tau condensation in a concentration-dependent manner. PKA-Tau bound to 14-3-3 appears to lose its ability to participate in condensate formation, and increasing 14-3-3 concentrations may deplete the pool of PKA-Tau available for condensation. We therefore hypothesized that re-establishing the availability of PKA-Tau should bring back PKA-Tau condensation.
To test this idea, we titrated the pS2 Tau peptide into an equimolar solution of PKA-Tau and 14-3-3ζ (10 μM each) that did not show condensation (Fig. 4b, d). At 10 μM pS2, PKA-Tau condensation indeed started to occur (Fig. 5a), suggesting that pS2 was outcompeting PKA-Tau from 14-3-3ζ dimers at this concentration, thereby re-enabling condensation of now unbound PKA-Tau. Importantly, pS2 and 14-3-3ζ both co-enriched with PKA-Tau in condensates (Fig. 5b), indicating that 14-3-3ζ co-partitioned into PKA-Tau condensates while bound to pS2, similar to 14-3-3ζ:PKA-Tau complexes at lower 14-3-3ζ concentrations. Co-condensation of 14-3-3ζ with PKA-Tau, therefore, likely involved protein regions other than the Tau binding site of 14-3-3ζ.
The crystal structure of 14-3-3ζ dimers in complex with the pS2 peptide (Fig. 5c, d, Supplemental Fig. S3, and Table 1) revealed that pS2 bound in the standard 14-3-3ζ substrate binding groove established after 14-3-3 dimerization. pS2 residues pS214 and pS324 were binding to the two different 14-3-3ζ dimer subunits (Fig. 5e, f), analogous to previously suggested binding modes of individual Tau and Tau peptides to the 14-3-3σ isoform. The structure further showed many negatively charged areas on the 14-3-3ζ dimer surface (Fig. 5g). It is known from other studies that negatively charged biomolecules can co-condense with Tau, driven by multivalent electrostatic interactions. Thus, co-condensation of Tau with 14-3-3ζ – being overall acidic (pI of 14-3-3ζ = 4-5;) and having solvent-exposed negative charges on their surface (Fig. 5g) — could be driven by electrostatic interactions, e.g., with the positively charged TauRD. Indeed, recent NMR data showed that 14-3-3ζ, in addition to strong interactions with Tau phospho-sites pS214 and pS324, establishes many “weaker” interactions with the TauRD, which may be involved in co-condensation.
We determined whether electrostatic interactions contribute to Tau:14-3-3ζ and PKA-Tau:14-3-3ζ condensation using different buffer salt (NaCl) concentrations that screen electrostatic protein-protein interactions to different degrees. Tau:14-3-3ζ condensates became smaller with increasing NaCl concentration, but remained observable at physiological salt concentrations (100 mM NaCl) (Fig. 5h). Electrostatic interactions thus played a role in stabilizing Tau:14-3-3ζ condensates. Additionally, we found that 14-3-3ζ promoted Tau condensation at net charge-matched concentrations (net charge of negative charges in 14-3-3ζ molecules = net charge of positive charges in Tau; 1-5 μM 14-3-3ζ at 10 μM Tau) (Supplemental Fig. S4), which is typical for electrostatically driven Tau condensation.
For PKA-Tau:14-3-3ζ and PKA-Tau:14-3-3ζ, however, 100 mM NaCl fully inhibited condensation. For the PKA-Tau variants, having a lower net charge because of negative charges added by phosphate groups, co-condensation with 14-3-3 may be more sensitive to buffer ion strength because of overall weaker multivalent electrostatic interactions with 14-3-3ζ. Thus, for phospho-Tau variants binding 14-3-3, both weaker multivalent interactions and masking of Tau stretches relevant for Tau condensation most likely contribute to inhibition of co-condensation.
When monitored over time, Tau condensates exhibit a “maturation” process, during which Tau molecules lose their mobility in condensates and form species that seed Tau aggregation. Independent of condensation, Tau aggregation can also be induced by direct aggregation into amyloid-like fibrils. Binding of Tau by 14-3-3 seems to work efficiently against both Tau aggregation pathways, as demonstrated by the inhibition of PKA-Tau condensation in vitro (Fig. 4b) and Tau aggregation in neurons (Fig. 1a, b). However, the promotion of Tau condensation by 14-3-3 in non-binding conditions could induce Tau aggregation via condensate maturation. We investigated in vitro how 14-3-3 would modulate Tau amyloid aggregation and condensate maturation in phospho-site dependent binding and non-binding conditions.
First, to test whether Tau amyloid aggregation was modulated by 14-3-3ζ binding, we performed Thioflavin-T (ThioT) aggregation assays for Tau and PKA-Tau in the absence or presence of 14-3-3ζ. Here, we used the FTD-mutant pro-aggregant Tau to facilitate timely amyloid aggregation within 1-2 days. 14-3-3ζ inhibited PKA-Tau aggregation but promoted amyloid aggregation of non-phosphorylated Tau (Fig. 6a), similar to previous observations. This indicated that occupation of TauRD by 14-3-3ζ in PKA-Tau (binding conditions) can prevent aggregation into amyloid-like fibrils, whereas 14-3-3 may act as a polyanionic cofactor, triggering Tau aggregation in non-binding conditions.
Next, we investigated whether 14-3-3 interactions would alter Tau condensate maturation. The mobility of proteins within condensates is determined by the amount and strength of their interactions, and for Tau, typically decreases during condensate maturation. We used fluorescence recovery after photobleaching (FRAP) to evaluate the mobility (diffusion) and mobile fraction of Tau in and 14-3-3ζ in 14-3-3:Tau condensates. By bleaching small regions (<20% of the condensate area), we particularly probed these parameters inside condensates. Both unphosphorylated Tau and 14-3-3ζ were highly mobile in condensates (mobile fractions: Tau 61%, t = 4 s; 14-3-3ζ 85%, t = 7 s) (Fig. 6b, c). Similarly, phosphorylated non-binding mutant PKA-Tau demonstrated a comparable fraction of mobile PKA-Tau and a slightly reduced fraction of mobile 14-3-3ζ molecules (mobile fractions: PKA-Tau 64%, t = 7 s; 14-3-3ζ 72%, t = 13 s). By contrast, the mobile fractions in PKA-Tau:14-3-3ζ condensates was decreased (mobile fractions: PKA-Tau 50%, t = 6 s; 14-3-3ζ 50%, t = 17 s). This indicated that 14-3-3:Tau binding reduced the mobility of both binding partners in condensates. Notably, PKA-Tau recovered more slowly than non-phosphorylated Tau, suggesting that Tau phosphorylation (aside from S214 and S324) generally increased interactions in Tau condensates. These interactions were likely of an electrostatic nature, given that PKA-Tau condensates were more sensitive to NaCl levels in the buffer than non-phosphorylated Tau condensates (Fig. 5h).
Lastly, to assess whether 14-3-3 binding would alter condensate maturation, we compared the time-dependent loss of Tau molecular mobility inside Tau and PKA-Tau condensates (measured by FRAP) in the absence and presence of 14-3-3ζ. Between 2 and 6 h after condensate formation, Tau FRAP decreased similarly in all conditions tested (Fig. 6d), showing that neither the presence nor the binding of 14-3-3ζ changed Tau condensate maturation. This suggested that the progressive loss of Tau mobility in condensates depended on Tau:Tau and not Tau:14-3-3 interactions, and that Tau:14-3-3ζ condensates could still catalyze Tau aggregation.
