To investigate the colocalization of hTau40 and hEFhd2 in aggregate structures, immunogold labeling and electron microscopy was used. Briefly, a parafilm platform was prepared in a humidifying chamber for all incubation steps. Twenty microliters of each sample were fixed with 2% glutaraldehyde (EMS, cat#16120) for 10 min. Then, a 300-mesh carbon-coated nickel grid (EMS, cat#FCF300NI) was placed on a 5 μL drop of each fixed sample spotted on the parafilm for 1 min. Then each grid was rinsed in one 10 μL drop of sterile water that was then wicked away using Whitman filter paper (Capillary Blotting and Wicking applications, cat#GB003). This step was repeated with a 20 μL drop of blocking solution (5% normal goat serum +0.1% bovine serum albumin in TBS). Next, the grids were placed over a 20 μL drop blocking solution for 30 min blocking at room temperature. After blocking, each grid was incubated with a 20 μL drop of primary antibodies for 1.5 h at room temperature. Primary antibody solution was a mixture of Tau 13 1:2500 (Biolegend, cat# 835201 and EFhd2 1:10, rabbit (Prosci, cat# 5657) diluted in TBS/ 5% normal goat serum. After incubation, grids were rinsed with sterile filtered TBS three times 1 min each. The grids were then incubated with a 20 μL drop of secondary antibodies mixture for 1 h at room temperature. Secondary antibodies were 15 mm gold-conjugated goat anti-rabbit (EMS, cat# 25112) and 6 nm gold-conjugated goat anti-mouse (EMS, cat# 25124) diluted in TBS/ 5% NGS 1:20. Subsequently, grids were washed with TBS six times 1 min each. Lastly, the grids were rinsed with a 10 μL drop of water followed by another rinse with a 10 μL drop of VitroEase (2% methylamine vanadate, Thermo Scientific, cat# A51037)). The last step was staining the grid on a 10  μL drop of VitroEase for 2 min. Grids were stored in grid boxes to fully dry before taking the micrographs.

TEM was used to visualize the morphological changes of tau aggregates induced by hEFhd2 (Cox et al., 2016). At the end of the polymerization reaction, unless the samples were processed for immunogold, all samples were processed for TEM using the same procedure. Beforehand, a parafilm platform was prepared on which grid handling took place. We used 300 mesh carbon-coated copper grids (EMS, cat# FCF300-CU). First, 20 μL of each sample were fixed with 2% glutaraldehyde for 10 min. Then, a 5 μL drop of each sample was spotted on the parafilm. The grids were placed on the top of sample drops for 1 min. The grids were rinsed by picking up a 10 μL drop of sterile water and wicking it away using Whatman filter paper. The final step was incubating the grids over a 10 μL drop of 2% uranyl acetate (EMS, cat#22400) for 1 min. The grids were allowed to fully dry in a closed grid box before taking the micrographs using a JEOL JEM-1400 Plus electron microscope at 80 kV and 5,000X and 15,000X magnification (25,000X and 40,000X for immunogold staining). Images were captured with an AMT XR81 digital camera and AMT software version 602.6 (Advanced Microscopy Techniques).

As described in Combs et al. (2017), a nondenaturing sandwich ELISA assay (sELISA) is instrumental to quantify tau oligomeric modifications in disease brains and recombinant protein (Combs et al., 2017). Herein, a slightly modified version of the assay was used to assess EFhd2-induced aggregates formed with hTau40^m and hTau40^ARA. Unless stated otherwise, all steps were undertaken at room temperature with shaking at 200 rpm. Washing and blocking were performed using 200 μL/well. All other steps performed using 50 μL/well. The capture antibodies used were Tau13, TOC1, or Alz50. Tau13 (Biolegend, cat# 835201) is a pan-tau monoclonal mouse IgG1antibody that reacts with monomeric and aggregated tau. TOC1 is a monoclonal mouse IgM antibody that was developed against tau dimers. It is a conformation-dependent antibody whose epitope is presumably revealed with dimerization and oligomerization (RRID#: AB_2832939; Kanaan Lab) (Patterson et al., 2011; Ward et al., 2013). Alz50 is another tau conformation-specific antibody. It is monoclonal mouse IgM that recognizes discontinuous epitope in misfolded tau 2–10 aa and 312–342 aa (RRID#: AB_2313937; Davies lab) (Wolozin et al., 1986; Ksiezak-Reding et al., 1988, 1995; Goedert et al., 1991; Carmel et al., 1996). At the end of overnight incubation of polymerization reaction, samples were initially diluted in 1X PBS to 40 nM (Tau13), 20 nM (TOC1, in presence of ARA), 40 nM (TOC1, in the absence of ARA; and Alz50). High binding 96-well plates (Corning, cat# 3590) were coated with TOC1, Tau13, or Alz50 diluted to 2 ng/μL in 1X PBS and incubated overnight at 4°C. Additional wells were coated with only 1X PBS and were used as a negative antibody control. Sample wells were strictly washed twice with ELISA wash buffer (100 mM borate acid, 25 mM sodium borate, 75 mM NaCl, 0.25 mM thimerosal, 0.4% (w/v) bovine serum albumin, 0.05% (v/v) Tween-20) and then blocked for 1 h with 5% non-fat dry milk (NFDM) prepared in ELISA wash buffer. Then, sample wells were carefully washed twice with ELISA wash buffer followed by adding the diluted samples for 1.5 h. Sample wells were washed 4 times with ELISA wash buffer and incubated with the detection antibodies; rabbit polyclonal pan-tau R1 (RRID#: AB_2832929; Kanaan lab) at 1:10 k for 1.5 h. Afterwards, sample wells were carefully washed 4 times and incubated with goat-anti-rabbit-HRP at 1:5000 (Vector Labs, cat# PI-1000) for 1 h. Wells were washed 4 times before developing with 3,3′,5,5′-tetramethylbenzidine (TMB, Sigma, cat# T8665) 8 min (Tau13, Alz50 and TOC1). Reactions were stopped using 3.5% sulfuric acid. Absorbance readings were measured at 450 nm on a SpectraMax Plus 384 microplate reader (Molecular Devices). Absorbance values of no capture negative control were first subtracted for sample values; then, the background-corrected values were converted to percent light absorbed using the equation % A = 1 − 10 − A ∗ 100 , where A is equal to absorbance at 450 nm.

All data were analyzed using GraphPad Prism v9.5 (San Diego, www.graphpad.com, RRID:SCR_002798). Before running statistical analysis, outliers were detected and removed from the data. ROUT method with false discovery rate of 1% was used for detection of outliers. Normal distribution was tested using the Shapiro–Wilk test. For comparisons between two groups, unpaired T-test and Mann–Whitney test were used for Gaussian and non-Gaussian distribution samples, respectively. For multiple groups, Kruskal-Wallis test (followed by Dunn’s multiple comparison post hoc test) was used for non-Gaussian samples. p-Values were calculated with a 95% confidence internal, if nothing is mentioned, it is nonsignificant: otherwise, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Data presented were shown as mean ± SEM.

Previously, our research established the connection between EFhd2 and pathological tau in both a transgenic tau model and postmortem brain tissues of tauopathies (Vega et al., 2008; Ferrer-Acosta et al., 2013b). Furthermore, we demonstrated that EFhd2 influences the conformation of tau by increasing its β-sheet structure (Vega et al., 2018). In addition, EFhd2 has been shown to affect tau’s liquid–liquid phase separation by promoting the formation of solid-like structures in vitro (Vega et al., 2019). However, whether EFhd2 can directly drive further aggregation of either monomeric or filamentous tau has yet to be investigated.

In our experimental approach, we utilized arachidonic acid (ARA)-induced tau fibrillization as an in vitro model of tau filaments and oligomers (Combs et al., 2017). The polymerization reaction proceeded for 16–18 h at room temperature using either a single recombinant protein or by co-incubating equimolar concentrations of a protein mixture. Subsequently, each reaction was fixed and placed on carbon-coated copper grids for visualization via transmission electron microscopy (TEM). For simplicity, hTau40^m refers to full length human tau hTau40 incubated overnight at room temperature without inducer, while hTau40^ARA refers to hTau40 incubated overnight with ARA. Additionally, hTau40^m/hEFhd2 and hTau40^ARA/hEFhd2 refer to the co-incubation of recombinant human EFhd2 with hTau40 in the absence and presence of ARA, respectively.

To establish a reference to which we could compare the co-incubation of EFhd2 and hTau40, each of the two proteins was first incubated separately both in the absence and presence of ARA. In line with previous research findings, monomeric hTau40 (hTau40^m) incubated overnight at room temperature without ARA exhibited no detectable filaments or aggregates (Figure 1A) (Kuret et al., 2005; Patterson et al., 2011; Cox et al., 2016). In contrast, Figure 1B shows that overnight incubation of hTau40 in the presence of ARA induced the formation of oligomers (asterisk), as well as short (inverted triangle) and long (open arrowhead) filaments (hTau40^ARA). Our prior studies established that EFhd2 self-oligomerizes without a nucleation factor or external inducer (Ferrer-Acosta et al., 2013a). Consistently, EFhd2 incubated overnight at room temperature formed short filamentous structures (Figure 1C). However, the addition of ARA reduced EFhd2 filament formation, as seen in Figure 1D. These findings align with our previously reported observations regarding the impact of heparin on EFhd2 self-oligomerization (Ferrer-Acosta et al., 2013b).

EFhd2 co-incubation with hTau40^m and hTau40^ARA led to the formation of distinct protein aggregates (Figure 2). EFhd2 induced the formation of amorphous aggregates when added to hTau40^m. Those aggregates were not observed in any of the recombinant proteins alone (Figure 2A, arrows; compared with Figures 1A,C). In addition, short filaments were noticed surrounding the larger protein aggregates (Figure 2A, arrowheads). These short filaments could represent EFhd2 self-oligomerization, as observed in Figure 1C. In contrast, when hEFhd2 was added to hTau40 in the presence of ARA (hTau40^ARA/hEFhd2), we detected aggregates that are different from those observed in hTau40^m/hEFhd2 (Figure 2A vs. Figure 2B). The observed aggregates in hTau40^ARA/hEFhd2 seem to be formed through coalescing filaments together in a way that forms proteins tangled or intertwined together (arrows in Figure 2B). The protein aggregates observed in hTau40^ARA/hEFhd2 are clearly different from hTau40^ARA filaments (Figure 1B).

Quantitatively, hTau40^ARA/hEFhd2 and hTau40^m/hEFhd2 aggregates are significantly different in area (Figure 2E). As described in the methods section, we quantified the average area for EFhd2 filaments (Figures 1C, 2E, dotted line) and hTau40^ARA filaments (Figures 1B, 2E, solid line) and used them as a baseline above which we calculated the area of the protein aggregates observed in hTau40^m/hEFhd2 and hTau40^ARA/hEFhd2, respectively. We also subtracted the detected electron dense speckles that represent artifacts of the staining process. The distribution of hTau40^ARA/hEFhd2 aggregate areas highlights the formation of larger aggregates than the average area of long tau filaments in hTau40^ARA (Figure 2E, dashed line), indicating that the overnight co-incubation of hEFhd2 and hTau40 in the presence of ARA led to the formation of larger distinct protein aggregates. In addition, the area of hTau40^m/hEFhd2 aggregates is larger than the area of the EFhd2 filaments (Figure 2E, dotted line). These results indicate that incubation of EFhd2 with either hTau40^m or hTau40^ARA induces the formation of protein aggregates with different structural characteristics.

EFhd2 protein bears a net negative charge. Hence, it was plausible to ascribe the observed protein structures formed by incubating hEFhd2 with hTau40 in the absence or presence of ARA to mere electrostatic interaction between the two proteins. Alternatively, EFhd2 in equimolar concentration might create a crowded environment that could exert changes on tau folding and promote the formation of the observed aggregates. To address those two explanations, we conducted a control experiment using GST protein. We chose GST because it shares some physicochemical properties with EFhd2 (i.e., molecular weight and isoelectric point). Importantly, GST has not been shown to be associated with tau. Thus, we co-incubated recombinant GST with hTau40 in the absence or presence of ARA (Figures 2C,D). GST neither promoted changes in hTau40^m nor did it induce similar structures like those formed with hEFhd2 (Figure 2A vs. Figure 2C). Likewise, adding GST to hTau40 in the presence of ARA did not lead to the formation of large protein aggregates as observed when hEFhd2 (Figure 2B vs. Figure 2D). These results strongly support the argument that the protein aggregates observed during the co-incubation of hEFhd2 with hTau40, with or without ARA, can be clearly attributed to the protein–protein interaction between EFhd2 and tau.

To determine the colocalization of both hEFhd2 and hTau40 within the observed aggregates, immunogold electron microscopy was employed. The detection of hTau40 and hEFhd2 was carried out using Tau13 and anti-EFhd2 antibodies, respectively, immediately after the overnight polymerization reaction. Control experiments were initially conducted to demonstrate the specificity of primary antibodies (Supplementary Figures S1, A–C) and secondary antibodies (Supplementary Figure S1D) used in the immunogold labeling. Figure 3A illustrates that both hTau40 (small particles, arrowheads) and hEFhd2 (large particles, arrows) colocalize on the same amorphous protein aggregated structure. The electron density of amorphous protein aggregates can be challenging to focus when using immunogold labeling in electron microscopy. The lack of defined structural features and the uneven distribution of electron-dense material within amorphous aggregates affects the clarity of the imaging, including the precise localization of immunogold labels. Therefore, focusing on amorphous aggregates is more challenging compared to well-defined structures with clear boundaries. Consistently, colocalization of hTau40 (small particles, arrowheads) and hEFhd2 (large particles, arrows) was evident in hTau40^ARA/hEFhd2 (Figure 3B). Interestingly, hEFhd2 was detected in the protein dense area where hTau40 filaments are coalescing into the aggregates (Figure 3B). These results endorse that the observed aggregates comprise hEFhd2 and hTau40.

The aforementioned results along with previous studies confirm the direct association between hEFhd2 and hTau40. Nonetheless, to rule out that antibody binding to the protein aggregates are technical artifacts (e.g., due to sample fixation with glutaraldehyde), we opted to use nondenaturing sELISA as an additional approach to verify the association of hEFhd2 and hTau40 on the same structures. The assays were conducted using Tau13 as capture antibody and anti-EFhd2 as detection antibody. Control samples were hTau40 in the absence (hTau40^m) or presence (hTau40^ARA) of ARA. As demonstrated in Figure 3C, EFhd2 signal was not detected in these samples. In contrast, EFhd2 signal was detected in both hTau40^m/hEFhd2 and hTau40^ARA/hEFhd2 samples. The data collectively attest to the protein–protein interaction between hTau40 and hEFhd2.

To further investigate the extent of hEFhd2-hTau40 aggregation, we explored whether the formation of these aggregates relies on the concentration of hEFhd2. When equimolar hEFhd2 and hTau40^m were used, the expected aggregates formed (compare Figures 2A, 4A). Reducing hEFhd2 concentration by half to 1 μM showed minimal differences in the aggregate structure of hTau40^m/hEFhd2 (Figure 4B). However, quantitative analysis revealed that 1 μM hEFhd2 led to the formation of aggregates with significantly smaller areas compared to 2 μM hEFhd2 (Figure 4D). Importantly, it should be noted that, as described earlier (Figure 2E), hTau40^m/hEFhd2 aggregates had a larger area than the area of hEFhd2 self-oligomeric filaments baseline (Figure 4D, dotted line). Conversely, with 0.5 μM hEFhd2, no aggregates were detected above the baseline (Figure 4C). These results indicate that the formation of hTau40^m/hEFhd2 aggregates is indeed hEFhd2 concentration dependent.

We also examined the effect that different hEFhd2 concentrations have on the formation of aggregates when incubated with hTau40^ARA (hTau40 in the presence of ARA). EM micrographs in Figures 5A,B illustrate unnoticeable structural differences in hTau40^ARA/hEFhd2 when 2 and 1 μM hEFhd2 concentrations were used. In contrast, detection of the hTau40^ARA/hEFhd2 entangled protein aggregates was remarkably diminished when 0.5 μM hEFhd2 was used and short filaments predominated in the fields (Figure 5C). Moreover, statistical analysis presented in Figure 5D showed a significant difference in aggregate area when either 2 μM or 1 μM hEFhd2 were used in comparison to 0.5 μM hEFhd2. No significant difference was detected in the aggregates area between 2 μM and 1 μM hEFhd2. These results imply that the extent of hEFhd2-hTau aggregation is directly correlated with varying EFhd2 concentrations.

The co-incubation of hEFhd2, hTau40, and ARA (a robust inducer of tau fibrillization) elicited the entanglement of hEFhd2 with hTau40 filaments, suggesting that hEFhd2 did not interfere with ARA-induced hTau40 filament formation (Figure 2B). Figure 6A summarizes the experimental paradigm followed to examine how the addition of hEFhd2 to hTau40 prior to or after ARA impacts formation of protein aggregates. In Figure 6C, hTau filaments (hTau40^f) were generated by initially incubating hTau40 with ARA for 24 h, followed by incubation with hEFhd2 for additional 16 h (hTau40^f/hEFhd2). Conversely, in Figure 6D, we incubated hEFhd2 and hTau40 (hTau40^m/hEFhd2) for 24 h, followed by ARA for additional 16 h (hTau40^m/hEFhd2/ARA). TEM was used to validate the formation of hTau40 filaments after 24 and 40 h incubation. The micrographs show that hTau40^f at 24 h (before adding hEFhd2) and at 40 h (the total experimental duration) are virtually the same as hTau40^ARA formed overnight (Supplementary Figures S2, A–C). Similarly, we did not notice a difference between hTau40^m/hEFhd2 at 24 or 40 h and hTau40^m/hEFhd2 after overnight reaction (Supplementary Figures S2, D,E). Figure 6B shows the consistent formation of hTau40^ARA/hEFhd2 aggregates as a reference.

Introducing hEFhd2 after ARA-induced tau filament formation (hTau40^f/hEFhd2) comparatively reduced the formation of intertwined filamentous structures (Figure 6B vs. Figure 6C). Additionally, isolated filaments were visible alongside the intertwined filamentous structures (Figure 6C). The quantitative analysis showed a significant decrease in the area of hTau40^f/hEFhd2 entangled filamentous structures in comparison to the area of structures that emerged when all three components were co-incubated simultaneously hTau40^ARA/hEFhd2 (Figure 6E). It is important to note that despite the differences in area, the observed intertwined filamentous structures seem morphologically similar (Figures 6B,C).

These results indicate that hEFhd2 may be more effective at entangling hTau40 filaments during their formation rather than after they are fully formed. In other words, hEFhd2 may not interfere with the initial formation of hTau40 filaments; rather, it promotes their entanglement as they are being generated. Along this line of thinking, pre-incubation of hTau40 and hEFhd2 (hTau40^m/hEFhd2) did not interfere with ARA-induced hTau40 filament formation (Figure 6D). Notably, filamentous structures radiate from the amorphous protein aggregates when hTau40 and hEFhd2 were incubated before the addition of ARA (Figure 6D). The observed structures are generally similar to those shown in Figures 6B,C. Moreover, the area of aggregates formed when ARA was added after the incubation of hEFhd2 and hTau40 (hTau40^m/hEFhd2/ARA) showed no significant difference from those observed when all three components were co-incubated (hTau40^ARA/hEFhd2) (Figure 6E). Taken together, these results signal that hEFhd2 does not interfere with the ARA-induced formation of hTau40 filaments. We can infer that hEFhd2 could be more effective at entangling tau filaments during ARA-induced tau polymerization than after long hTau40 filaments have been fully formed.

Tau conformational changes are among the pivotal pathological events associated with neuronal toxicity (Brunden et al., 2008; Spires-Jones et al., 2009; Lasagna-Reeves et al., 2012; Cowan and Mudher, 2013; Chung et al., 2021). Hence, several tools have been geared toward characterizing tau conformation associated with its aggregated forms to understand their spatial and temporal evolution during tauopathies. Tau oligomeric complex 1 (TOC1) antibody is a conformation-dependent antibody that recognizes tau oligomers (Patterson et al., 2011; Ward et al., 2013). TOC1 epitope is exposed upon oligomerization and presumably is masked with further tau fibrillization. Sandwich ELISA (sELISA) is used to quantify and characterize TOC1 signal in tauopathies and in vitro hTau fibrils (Tiernan et al., 2016; Combs et al., 2017). Thus, sELISA was conducted here to determine whether hEFhd2-hTau aggregates have conformational changes detected by TOC1.

hTau40^m/hEFhd2 and hTau40^ARA/hEFhd2 aggregates were formed as described earlier and validated using TEM. The protein aggregates were then subjected to sELISA. As illustrated in Figure 7A, sELISA using Tau13 as capture antibody and R1 (pan-tau) as detection antibody was conducted to detect total hTau40 (monomeric and aggregated) levels. The results point to nearly comparable tau levels in hTau40^m/hEFhd2 and hTau40^m. In contrast, a significantly lower hTau40 level is detected in hTau40^ARA/hEFhd2 compared to hTau40^ARA.

We assessed TOC1 reactivity of aggregated hTau40 in the presence of hEFhd2. As expected, TOC1 showed no affinity for hTau40^m (monomeric tau without ARA), which served as a negative control (Figure 7B). Consistent with previous research (Tiernan et al., 2016; Combs et al., 2017), TOC1 successfully captured hTau40^ARA, confirming the formation of ARA-induced hTau40 oligomers and the associated conformational changes (Figure 7B). Interestingly, hTau40^m/hEFhd2 (hEFhd2 and hTau40 in the absence of ARA) was not captured with TOC1 (Figure 7B). This result indicates that the formation of hTau40^m/hEFhd2 aggregates does not involve the conformation change detected by TOC1. Alternatively, amorphous protein aggregate might mask the conformational epitope. Conversely, TOC1 reactivity in hTau40^ARA/hEFhd2 (hEFhd2 and hTau40 in the presence of ARA) was significantly higher compared to hTau40^ARA (Figure 7B). These results indicate that TOC1 captures hTau40^ARA/hEFhd2 aggregates more effectively than hTau40^ARA. That could imply that ARA-induced hTau filament formation in the presence of hEFhd2 leads to greater exposure of the TOC1 specific conformational epitope in comparison to hTau40^ARA. It is important to note here that statistical analysis was not conducted to compare between samples in the absence of ARA to samples in the presence of ARA (See Material and Methods).

We also tested if hEFhd2 induces a tau conformational change that could be detected by Alz50. Alz50 antibody recognizes a discontinuous epitope that involves tau’s N-terminus and microtubule-binding repeat domains, which is formed due to a conformational change associated with tau oligomer formation (Wolozin et al., 1986; Ksiezak-Reding et al., 1988; Goedert et al., 1991; Ksiezak-Reding et al., 1995; Carmel et al., 1996). Following the same methodology, we conducted sELISA to analyze Alz50 reactivity. Figure 7C demonstrates that Alz50 captured hTau40^ARA. Alz50 signal was also detected with hTau40^m, suggesting that during the incubation time some monomeric hTau40 adapted the conformational change detected by Alz50. The incubation of hEFhd2 with hTau40 in the absence (hTau40^m/hEFhd2) or presence (hTau40^ARA/hEFhd2) of ARA significantly reduced Alz50 reactivity compared to hTau40^m and hTau40^ARA, respectively. Reduced Alz50 signal in hTau40^ARA/hEFhd2 could be due to the formation of a different tau conformation that masks the Alz50-epitope.