We first asked whether Hsp27 is associated with pathological pTau aggregates in AD brains. We obtained human frontal cortex tissue slices from two cases of AD patients and two cases of age-matched healthy controls (Table 1). Immunofluorescence staining using antibodies anti-pTau^Ser262 and anti-Hsp27 detected the presence of pTau in two AD cases, but absent in two healthy controls (Figure 1). Notably, Hsp27 was detected at a high level in AD brain slices, but at a much lower or undetectable level in healthy control brain slices (Figure 1), consistent with previous findings of elevated Hsp27 levels in AD brains (Björkdahl et al., 2008; Renkawek et al., 1994). More importantly, Hsp27 was found to co-localize with the pTau aggregates in both AD cases (Figure 1), suggesting the pathological relevance between Hsp27 and pTau pathology in AD. Note that not all pTau aggregates contain Hsp27 and vice versa, especially in case 1, which may reflect the heterogeneity of the protein aggregates.

To investigate the influence of Hsp27 on the abnormal aggregation of pTau and its neuropathology in vivo, we used a Drosophila tauopathy model where human pathogenic mutant Tau (Tau^R406W), associated with FTD with parkinsonism linked to chromosome 17 (FTDP-17), is expressed in the nervous system by a pan-neuronal driver elav^C155-GAL4. Our previous work has demonstrated that the expression of human Tau^R406W in the Drosophila nervous system leads to age-dependent neurodegeneration recapitulating some of the salient features of tauopathy in FTDP-17 (Ali et al., 2012; Ma et al., 2020). To quantitatively assess the pTau species in the brain, we carried out western blot analysis using a total Tau antibody 5A6 (Johnson et al., 1997), a pTau^Ser262 specific antibody (Iijima et al., 2010), and a hyper-phosphorylated Tau antibody AT8 that recognizes hyper-phosphorylation at Ser202 and Thr205 sites (Goedert et al., 1995). Hsp27 overexpression significantly reduced hyper-phosphorylated Tau levels 2 and 10 days after eclosion (DAE) (Figure 2A and B, Figure 2—source data 1; Figure 2—source data 2).

We next examined the morphology of the fly brain and the accumulation of hyper-phosphorylated Tau by immunofluorescence staining. We used the AT8 antibody to label hyper-phosphorylated Tau because phosphorylation at Ser202 and Thr205 occurs after the phosphorylation of S262 and is associated with advanced pathology (Wesseling et al., 2020). Consistent with previous findings, brains with neuronal expression of Tau^R406W exhibited an accumulation of filamentous pTau and a reduction of brain neuropil size indicative of neurodegeneration (Figure 2C–F). Interestingly, we found that neuronal expression of Tau^R406W led to a significant upregulation of endogenous Hsp27, which was highly enriched in the synaptic areas (Figure 2C and D), suggesting that Hsp27 might be a part of the neuronal stress response network that is upregulated upon proteotoxic stress. Importantly, overexpression of Hsp27 restored the size of brain neuropil and suppressed the accumulation of filamentous pTau (Figure 2C–F), suggesting that Hsp27 protects against mutant Tau^R406W induced neurodegeneration.

Synaptic loss is a hallmark of human tauopathy (McGowan et al., 2006). To evaluate the synaptic integrity, we took advantage of the Drosophila visual system with highly organized paralleled photoreceptor structures (Fischbach and Dittrich, 1989). We have previously shown that Tau^R406W expression resulted in synaptic aggregation of hyper-phosphorylated Tau as well as a significant reduction of Bruchpilot (Brp), an active zone associated-cytoskeletal matrix protein, at lamina cartridge, suggesting a severe loss of the active zone structures in the presynaptic terminals (Ma et al., 2020). As shown in Figure 2G–I, we expressed Tau^R406W with either GFP (control) or Hsp27 in the photoreceptors using GMR-GAL4 and found that overexpression of Hsp27 led to a remarkable reduction of synaptic pTau and enhanced BRP localization, indicating a restoration of synaptic integrity.

Taken together, we showed that Hsp27 protects against synaptic dysfunction in a Drosophila tauopathy model by reducing pTau aggregation and ameliorating pTau-induced synaptic degeneration.

We next sought to examine the binding of Hsp27 with pTau and its effect on pTau amyloid aggregation in vitro. We purified the longest Tau isoform – Tau40 that contains both the projection region and four microtubule-binding repeat domains (R1–R4), as well as K19, a truncated Tau construct that contains R1, R3, and R4 (Figure 3A). Then phosphorylated Tau, including pTau40 and pK19, were prepared by phosphorylating purified Tau40 and K19 using kinase MARK2 which was previously identified to be important in mediating pTau pathology in AD (Ando et al., 2016; Drewes, 2004; Gu et al., 2013). Two dimensional (2D) ¹H-¹⁵N HSQC spectra enable us to monitor each of the MARK2 phosphorylation sites in both pTau40 and pK19. Consistent with previous report (Schwalbe et al., 2013), eight MARK2 phosphorylation sites including pS262, pS293, pS305, pS324, pS352, pS356, pS413, and pS416 were identified on the HSQC spectrum of pTau40 (Figure 3—figure supplement 1A). Four phosphorylation sites including pS262, pS324, pS352, and pS356 were identified on the HSQC spectrum of pK19 (Figure 3—figure supplement 1B). Most of these phosphorylation sites are within the fibril forming core region of Tau, and two of them including pS262 and pS356 have been widely reported to be essential for the detachment of pTau from MT and cause increased toxicity in animal models (Hanger et al., 1998; Hanger et al., 2009; Ali et al., 2012).

We used BioLayer Interferometry (BLI) assay to measure the binding affinity of Hsp27 with pTau and unphosphorylated Tau, respectively. pTau and Tau were immobilized on biosensor tips, and the association and dissociation curves were measured in the presence and absence of various concentrations of Hsp27 to calculate the equilibrium dissociation constant (K_D) between Hsp27 and pTau/Tau. Notably, the K_D value of Hsp27/pTau40 is ~1.58 µM (Figure 3B; Figure 3—source data 1). The K_D value for Hsp27/pK19 is ~1.98 µM (Figure 3C; Figure 3—source data 1) which is similar to that of Hsp27/pTau40. In addition, both the association (k_on) and dissociation (k_off) rates of Hsp27 to pK19 are at the same order of magnitude of Hsp27 to pTau40. These results suggest that pK19 serves as the key region for binding of Hsp27 to pTau40. In sharp comparison, the binding of both Hsp27/Tau40 and Hsp27/K19 were much weaker which cannot be detectable under our experimental condition (Figure 3B and C; Figure 3—source data 1). These results demonstrate that hyper-phosphorylation significantly enhances the binding of Hsp27 to both pTau40 and pK19.

We further assessed the influence of Hsp27 on the amyloid aggregation of pTau40 and pK19 by thioflavin T (ThT) fluorescence assay and negative-staining electron microscopy (EM). The results showed that Hsp27 effectively inhibited the amyloid aggregation of both pTau40 and pK19 in a dose-dependent manner (Figure 3D and E; Figure 3—source data 1). Together, our data show that Hsp27 can specifically bind to hyper-phosphorylated Tau including pTau40 and pK19, and efficiently prevent their amyloid aggregation.

To investigate the structural basis of the interaction between Hsp27 and pTau, we performed solution NMR spectroscopy, and titrated Hsp27 to ¹⁵N-labeled pTau40 and ¹⁵N-labeled pK19, respectively. Strikingly, the 2D ¹H-¹⁵N HSQC spectra showed obvious chemical shift changes (CSDs) of all the eight phosphorylated Ser (pSer) residues of pTau40 and four pSer residues of pK19 upon titration of Hsp27 (Figure 4A and B). The highly overlapped signals in the pTau40 HSQC spectrum exclude us to obtain a full assignment of pTau40. Alternatively, we accomplished the backbone assignment of pK19, which enables us to map the binding interface of pK19 with Hsp27. As shown in Figure 3C, the regions containing the four pSer residues of pK19 harbor prominent CSDs. Among them, the two pSer residues in R4 showed the largest CSDs above 0.06 ppm (Figure 4C; Figure 4—source data 1). In addition, titration of Hsp27 to pK19 induced dramatically intensity drops of most residues of pK19 (Figure 4C; Figure 4—source data 1), implying the direct binding of Hsp27 to pK19 which increases the molecular weight upon pK19-Hsp27 complex formation, and thus increases line width of the cross peaks. By contrast, addition of the same amount of Hsp27 to unphosphorylated Tau including Tau40 and K19 only results in slightly CSD and intensity perturbations (Figure 4—figure supplement 1; Figure 4—figure supplement 1—source data 1). Together, our data suggest that phosphorylation of Tau promotes the interaction between Hsp27 and pTau by forming the specific interaction between Hsp27 and the phosphorylated residues of Tau.

Hsp27 composes of a central α-crystallin domain (ACD) flanked by a flexible NTD and a flexible C-terminal domain (CTD). To further dissect the role of different domains of Hsp27 in binding pK19, four different truncations of Hsp27 were prepared (Figure 5A). Their binding affinities to pK19 were determined by BLI, respectively. Notably, the isolated NTD binds to pK19 with a K_D value of ~0.57 µM which is similar to the binding affinity of the full-length Hsp27 to pK19 (Figure 5B; Figure 5—source data 1). ΔCTD which contains both NTD and ACD display a similar binding affinity to pK19 (0.67±0.01 µM; Figure 5B; Figure 5—source data 1). In contrast, both ACD and ΔNTD which do not contain NTD show no detectable binding to pK19 at the same condition (Figure 5C and Figure 5—figure supplement 1; Figure 5—figure supplement 1—source data 1). Together, these data suggest that Hsp27 NTD plays an essential role in mediating the binding of Hsp27 to pK19.

Next, we performed cross-linking mass spectrometry (CL-MS) to map the interacting regions of Hsp27 and pK19. pK19 was used since it contains less amino acids but keeps the key region of pTau40 for binding Hsp27 to simplify the data analysis. NTD of Hsp27 contains 0 lysine residue. Whereas, negatively charged residues including glutamine acid (E) and aspartic acid (D) evenly distribute within the three different domains of Hsp27 (Figure 5D). Thus, we performed a two-step coupling procedure using cross-linkers 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) and sulfo-N-hydroxysulfosuccinimide (sulfo-NHS) to cross-link Hsp27 and pK19. EDC causes direct conjugation of carboxylates (-COOH) to primary amines (-NH₂) of target molecules, and sulfo-NHS is introduced for stabilization. We identified two pairs of cross-linked segments between Hsp27 and pK19 with a confidence score of <10^–6 by mass spectrometry, including E3_Hsp27-K331_pK19 and D30_Hsp27-K311_pK19 (Figure 5D and E and Table 2). Notably, both cross-linked segments are formed by residues located at R3 of pK19 and NTD of Hsp27 (Figure 5D), which is consistent with our NMR titration that residues at R3 show slightly stronger intensity drop upon addition of Hsp27, and the domain truncation mapping results. Together, our data demonstrate that NTD is essential to mediate pK19 binding.

To further pinpoint the binding regions of pK19 by NTD and other domains of Hsp27, we titrated full-length and the four truncations of Hsp27 to ¹⁵N-labeled pK19 and collected a series of HSQC spectra. Titration of both NTD and ΔCTD induce significant CSDs mainly located on the four pSer sites of pK19 (Figure 6A and B; Figure 6—source data 1), which recapitulates that titrated by full-length Hsp27. In contrast, titration of ACD or ΔNTD caused much smaller CSDs on the four pSer sites of pK19. Moreover, addition of NTD and ΔCTD induces overall intensity drops of pK19 signals (~I/I₀<0.8), while titration of ACD and ΔNTD only cause slightly intensity drops. These results demonstrate that Hsp27 mainly uses its NTD to directly interact with the pSer residues of pK19 for pTau binding, which is consistent with the BLI results.

Notably, we found that despite that titration of ACD does not induce CSDs of the pSer sites, it causes relative small but detectable CSDs (>0.01 ppm) and intensity drops (I/I₀<0.7) of the highly amyloidogenic region - ³⁰⁶VQIVYK³¹¹ of pK19, especially residues V309, Y310, and K311 (Figure 6A and B; Figure 6—source data 1). This demonstrates that ACD is capable of directly binding to pK19 via its VQIVYK region. Of note, the cross peaks of V306, Q307, and I308 severely overlapped with a few residues of pK19 which excludes us to obtain the accurate CSD and intensity changes upon titration. Together, our results demonstrate that different domains of Hsp27 display distinct binding preferences to different regions of pK19. NTD exhibits a high binding affinity to the multiple pSer sites of pK19, while ACD weakly binds the amyloidogenic VQIVYK segment of pK19.

Finally, we examined the role of different domains of Hsp27 in preventing amyloid aggregation of pK19 and pTau40. As shown in Figure 7A and B (Figure 7—source data 1), NTD-containing Hsp27 variants including full-length Hsp27, ΔCTD, and NTD exhibits potent inhibitory activity in preventing both pTau40 and pK19 fibrillation. This suggests that strong binding of NTD to the phosphorylated sites of pTau is important in preventing pTau fibrillation. Interestingly, ACD and ΔNTD display weakened but not negligible inhibitory activity of pTau, especially for pK19. This suggests that binding of ACD to the VQIVYK segment of pTau may result in moderate inhibition of pTau fibrillation. Together, our data show that the NTD is predominately employed to bind pTau to prevent its fibrillation, while ACD also contributes chaperone activity to a less degree by binding to the highly aggregation-prone VQIVYK region of pTau.

Human brain samples were originally from the tissue bank of the Center for Neurodegenerative Disease Research (CNDR) at the University of Pennsylvania. 10% neutral buffered formalin (NBF)-fixed, paraffin-embedded human frontal cortex tissues from two AD patients and two age-matched normal controls were sectioned into 6 μm slices, and immunostained as described before (He et al., 2018). Briefly, after deparaffinization and rehydration, the brain slices were incubated with anti-human Hsp27 (1:200, Cat. # ab2790) and anti-pTau^Ser262 (1:200, Cat. # ab131354) antibodies overnight at 4 °C followed by a 2 hr incubation of Alexa Fluor-conjugated secondary antibodies (Thermo Fisher Scientific), and then mounted with DAPI-containing Fluoromount-G (SouthernBiotech, USA).

The following fly strains were used: UAS-Hsp27 (generated in this study); UAS-Tau^R406W obtained from Dr. Mel Feany (Wittmann et al., 2001); yw, UAS-GFP, elav-GAL4, GMR-GAL4 obtained from Bloomington Drosophila Stock Center. Flies were reared on cornnmeal-molasses-yeast medium at 22°C, 65% humidity, with 12 hr light/12 hr dark cycles.

For protein extraction in flies, 10 heads of each group were homogenized in radioimmunoprecipitation assay (RIPA) buffer (Sigma-Aldrich, R0278). Samples were mixed with Laemmli sample buffer containing 2% SDS, 10% glycerol, 62.5 mM Tris-HCl (pH 6.8), 0.001% bromophenol blue, and 5% β-mercaptoethanol, and heated at 95 °C for 10 min. Proteins were separated by SDS-polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane. After blocking at room temperature for 1.5 hr, the membrane was incubated with primary antibody at 4 °C overnight, followed by secondary antibody for 1.5 hr at room temperature. Imaging was performed on an Odyssey Infrared Imaging system (LI-COR Biosciences) and analyzed using Image Studio (ver 4.0). Primary antibody dilutions were used as follows: anti-pTau^{Ser202/Thr205} (AT8, 1:500; ThermoScientific, Rockford, IL, USA), anti-pTau^Ser262 (1:1000; ThermoScientific), anti-β-actin (1:5000, Sigma-Aldrich, St Louis, MO, USA), anti-total-Tau (5A6, 1:250, Developmental Studies Hybridoma Bank, Iowa City, IA, USA).

Fly brains were dissected in phosphate-buffered saline (PBS, pH 7.4) and fixed with 4% formaldehyde at room temperature for 10 min. After 10 min washing in PBTx (PBS containing 0.4% v/v Triton X-100) three times, brains were incubated with primary antibodies diluted in PBTx with 5% normal goat serum at 4 °C overnight. Fly brains were incubated with conjugated secondary antibodies at room temperature for 1 hr, followed by 4’,6- diamidino-2-phenylindole (DAPI, 1:300, Invitrogen, Carlsbad, CA, USA) staining for 15 min. Brains were mounted on slides with VECTASHIELD Antifade Mounting Medium (Vector Laboratories Inc, Burlingame, CA, USA). After that, brains were imaged using an Olympus IX81 confocal microscope coupled with a 60×oil immersion objective lens with a scan speed of 8.0 µs per pixel and spatial resolution of 1024×1024 pixels. Images were processed using FluoView 10-ASW (Olympus) and Adobe Photoshop CS6 and quantified using Fiji/Image J (ver 1.52). Primary antibody dilutions were used as follows: anti-pTau^{Ser202/Thr205} (AT8, 1:250), anti-BRP (nc82, 1:250, Developmental Studies Hybridoma Bank), anti-Hsp27 (1:250, Abcam, Cambridge, MA, USA). The experimenter was blinded to the genotype during confocal scans. The sample size was not predetermined by statistical calculations, but the number of flies and independent biological replicates were large enough to achieve sufficient statistical power, according to our previous publications (Ma et al., 2020; Zhu et al., 2019). All the data were included in the final analysis.

All plasmids used in this study are available upon request. Genes of full-length human Hsp27 (UniProt accession number P04792), and the truncations of Hsp27 including NTD (residues of 1–84), ΔCTD (residues of 1–176), ΔNTD (residues of 85–205), and ACD (residues of 85–175) were inserted into a pET-28a vector with an N-terminal His₆-tag and a following tobacco etch virus (TEV) protease cleavage site. Sequencing of all constructs was verified by GENEWIZ company (Suzhou, China).

All proteins were expressed in Escherichia coli BL21(DE3) cells, grown to an OD₆₀₀ of 0.8 and induced with 0.3 mM IPTG overnight at 16 °C. Hsp27 and its variants were purified with HisTrapFF column (GE Healthcare) with the Tris buffer (50 mM Tris-HCl, 100 mM NaCl, a gradient of 0~500 mM imidazole, pH 8.0). The N-terminal His₆-tag was removed using TEV protease in the buffer of 50 mM Tris-HCl, 100 mM NaCl, pH 8.0. The cleaved proteins were immediately loaded onto the size-exclusion chromatography column Superdex 75 26/60 (GE Healthcare) with a PBS buffer of 50 mM sodium phosphate, 50 mM NaCl at pH 7.0.

Human Tau40 and K19 were over-expressed and purified as previously described (Barghorn et al., 2005). Briefly, Tau/K19 was purified by a HighTrap HP SP (5 ml) column (GE Healthcare), and followed by a Superdex 75 gel filtration column (GE Healthcare).

For ¹⁵N-labeled proteins, protein expression was the same as that for unlabeled proteins except that the cells were grown in M9 minimal medium with ¹⁵NH₄Cl (1 g l⁻¹). Purification of ¹⁵N-labeled proteins was the same as that of the unlabeled proteins.

The purity of proteins was assessed by SDS-PAGE. Protein concentration was determined by BCA assay (Thermo Fisher).

Phosphorylation of Tau40/K19 by MARK2 kinase was carried out following a method described previously (Ma et al., 2020). Briefly, Tau40/K19 was incubated with cat MARK2-T208E at a molar ratio of 10:1 in a buffer of 50 mM Hepes, pH 8.0, 150 mM KCl, 10 mM MgCl₂, 5 mM ethylene glycol tetraacetic acid (EGTA), 1 mM PMSF, 1 mM dithio-threitol (DTT), 2 mM ATP (Sigma), and protease inhibitor cocktail (Roche) at 30 °C overnight. pTau40/pK19 was further purified by HPLC (Agilent) to remove kinase, and lyophilized. The sites of phosphorylation were quantified using 2D ¹H-¹⁵N HSQC spectrum according to previous publications (Schwalbe et al., 2013).

All NMR experiments were performed at 298 K in the NMR buffer of 50 mM sodium phosphate, 50 mM NaCl, and 10% (v/v) D₂O at pH 7.0. 3D HNCA and HNCACB experiments were collected on an Agilent 600 MHz spectrometer for backbone assignment of K19, while 3D HNCA, HNCOCA, CBCACONH, and HNCACB experiments were collected for backbone assignment of pK19. NMR titrations were performed on a Bruker 900 M or Agilent 800 MHz spectrometer. Each sample (500 µl) was made of 50 µM ¹⁵N-pTau40/pK19/Tau40/K19, in the absence or presence of Hsp27 and its variants at the indicated concentrations. Intensity changes were calculated by I/I₀. And chemical shift deviations (CSDs, Δδ) were calculated using equation,Δδ=Δδ1H2+0.0289Δδ15N2

Where Δδ1H and Δδ15N are the chemical shift differences of amide proton and amide nitrogen between free and bound state of the protein, respectively. All NMR spectra were processed using NMRPipe (Delaglio et al., 1995), and analyzed using Sparky (Lee et al., 2015) and NMRView (Johnson, 2004).

The binding kinetics of the Tau proteins to Hsp27 WT and variants were measured by BLI on an ForteBio Octet RED96 system (Pall ForteBio LLC). Experiments were performed at room temperature using the assay buffer of 50 mM sodium phosphate, 50 mM NaCl, pH 7.0. Tau proteins including pTau40, Tau40, pK19, and K19 were firstly biotinylated by incubating 0.5–1 mg/ml proteins with biotin at a molar ratio of protein: biotin of 2:3 at 4 °C for 30 min, then the excess biotins were removed by desalting column (Zeba Spin Desalting Columns, Thermo). Then biotinylated Tau proteins were immobilized onto streptavidin biosensors (ForteBio) individually, and incubated with varying concentrations of Hsp27 WT and variants as indicated in the figure. The kinetic experiments were performed following the protocol previously (Zhang et al., 2021), in which an auto-inhibition step was used to eliminate the non-specifically binding of Hsp27 and its variants to biosensors. The resulting curves were corrected using the blank reference and analyzed by the ForteBio Data Analysis software 9.0.

ThT kinetics of pTau/pK19 in the absence and presence of Hsp27 and its variants were recorded using a Varioskan Flash Spectral Scanning Multimode Reader (Thermo Fisher Scientific) with sealed 384-microwell plates (Greiner Bio-One). The assay buffer is 30 mM PBS, 2 mM MgCl2, 1 mM DTT, 0.05% NaN₃, pH 7.4. 0.5% (v/v) of fibril seeds (the seeds were prepared by sonicating fibrils for 15 s) were added to promote the fibril formation of pTau/pK19. ThT fluorescence with a final ThT concentration of 30 µM in each sample was measured in triplicates with shaking at 600 rpm at 37 °C with excitation at 440 nm and emission at 485 nm.

Five μl of samples were applied to fresh glow-discharged 300-mesh copper carbon grids and stained with 3% v/v uranyl acetate. Specimens were examined by using Tecnai G2 Spirit TEM operated at an accelerating voltage of 120 kV. Images were recorded using a 4K × 4K charge-coupled device camera (BM-Eagle, FEI Tecnai).

Cross-linking experiments were modified from the protocol previously (Gong et al., 2015). 0.4 mg EDC (final concentration of ~2 mM) and 1.1 mg Sulfo-NHS (final concentration of ~5 mM) were added to 1 ml of Hsp27 solution and reacted for 15 mins at room temperature. Then 1.4 µl 2-mercaptoethanol (final concentration of 20 mM) was added to quench the activity of EDC. Then pK19 was added to the reaction system at an equal molar ratio with Hsp27, and the proteins were allowed to react for 2 hr at room temperature. Then Tris was added to the reaction system with a final concentration of 50 mM to quench the reaction. Cross-linking products were analyzed by SDS-PAGE to assess the cross-linking efficiency. Before MS analysis, proteins were precipitated with acetone; the pellet was resuspended in 8 M urea, 100 mM Tris (pH 8.5) and digested with trypsin at room temperature overnight. The resulting peptides were analyzed by online nanoflow liquid chromatography tandem mass spectrometry (LC−MS/MS). And the mass spectrometry data were analyzed by pLink (Yang et al., 2012).