Semisynthetic Modification of Tau Protein with Di-Ubiquitin Chains for Aggregation Studies

Ubiquitin, a protein modifier that regulates diverse essential cellular processes, is also a component of the protein inclusions characteristic of many neurodegenerative disorders. In Alzheimer’s disease, the microtubule associated tau protein accumulates within damaged neurons in the form of cross-beta structured filaments. Both mono- and polyubiquitin were found linked to several lysine residues belonging to the region of tau protein that forms the structured core of the filaments. Thus, besides priming the substrate protein for proteasomal degradation, ubiquitin could also contribute to the assembly and stabilization of tau protein filaments. To advance our understanding of the impact of ubiquitination on tau protein aggregation and function, we applied disulfide-coupling chemistry to modify tau protein at position 353 with Lys48- or Lys63-linked di-ubiquitin, two representative polyubiquitin chains that differ in topology and structure. Aggregation kinetics experiments performed on these conjugates reveal that di-ubiquitination retards filament formation and perturbs the fibril elongation rate more than mono-ubiquitination. We further show that di-ubiquitination modulates tau-mediated microtubule assembly. The effects on tau protein aggregation and microtubule polymerization are essentially independent from polyubiquitin chain topology. Altogether, our findings provide novel insight into the consequences of ubiquitination on the functional activity and disease-related behavior of tau protein.


Introduction
Ubiquitination is a prevalent post-translational process that regulates essential eukaryotic cell functions such as protein homeostasis, signaling, and cellular localization [1][2][3]. It consists of the covalent addition of the small protein ubiquitin to the side chain of lysine residues in a target protein. The reaction is controlled by the coordinated action of a ubiquitin-activating enzyme (E1), a ubiquitin-conjugating enzyme (E2) and a ubiquitin ligase (E3), which catalyze the formation of an isopeptide bond between the carboxyl group of the C-terminal glycine of ubiquitin and the ε-amino group of the substrate's lysine [2][3][4][5]. Ubiquitination of a ubiquitin molecule through one (or more) of its seven lysine residues (Lys6, Lys11, Lys27, Lys29, Lys33, Lys48 and Lys63) or its amino terminus produces polymers named polyubiquitin chains [2,5]. Polyubiquitin chains have different biological functions depending on the type of linkage that connects the ubiquitin units. Lys48and Lys11-linked chains regulate proteasome-mediated protein degradation, Lys63-linked chains are involved in endocytosis, cell signaling, and cellular response to DNA damage [2,6], while the Indeed, we recently showed that CHIP (Carboxy terminus of Hsp70-interacting protein), an E3 ligase of tau protein [29], ubiquitinates the protein at more than ten sites [30]. To overcome the inherent limitation of the enzymatic approach and to obtain site-specific ubiquitination of target proteins, chemists have developed a vast array of semisynthetic methods that are based on the chemical conjugation of protein precursors [31]. Chemical ubiquitination strategies based on non-native isopeptide bond formation are often easier to implement and give high yields of the product conjugates. Among these, disulfide-coupling chemistry has proven to be highly versatile, efficient and robust, and has already been applied in many studies [30][31][32][33][34]. As a replacement for isopeptide linkage, this semisynthetic method generates a disulfide bond between a Cys residue placed in a specific position of the target protein and the C-terminal aminoethanethiol of a ubiquitin derivative obtained by intein processing. Recently, we described the production and characterization of tau protein mono-ubiquitinated at three different positions using disulfide-directed methodology [30]. With this approach, we obtained novel insight into the diverse effects that lysine mono-ubiquitination at different sites exerts on tau protein aggregation.
Here, we introduce a method based on directed disulfide bond formation in combination with enzymatic synthesis of ubiquitin chains to obtain controlled polyubiquitination of tau protein, thus adding a further level of complexity in the study of ubiquitinated tau protein. We produced and characterized the aggregation process of tau protein species homogenously modified with di-ubiquitin molecules at position 353, one of the ubiquitinated sites found in AD-brain filaments. The versatility of the method allowed us to incorporate two types of di-ubiquitin chains (Lys48-or Lys63-chains) characterized by different structures, and to explore the impact of these prototypical ubiquitin polymers on tau protein aggregation.

Results
In the present work, we optimized a semisynthetic strategy based on disulfide-coupling chemistry to covalently attach a di-ubiquitin molecule to the tau protein. Ubiquitin-substrate conjugates obtained using this method are functional mimics of the native ubiquitinated counterparts, with the synthetic linkage being only one atom longer than the native isopeptide bond [32,34]. We selected the construct tau4RD, which represents a short form of the protein, spanning residues Q244-E372 plus an initial Met ( Figure 1). Tau4RD has been widely used to study the aggregation behavior of tau protein as it includes the majority of the microtubule binding region and most of the residues involved in the assembly of the cross-β structure forming the core of the tau filaments [35]. To obtain site-specificity, we used a tau4RD mutant devoid of the endogenous cysteines and where a single cysteine was installed at the desired position. We chose to introduce the modification at position 353 because this site was found polyubiquitinated in PHF-tau [26,28], it is part of the fibril core found in different tauopathies [26], and mono-ubiquitination at this position had an observable influence on the aggregation kinetics without hindering the formation of short fibrils [30]. Thus, it was deemed suitable for evaluating the effect of an extended ubiquitin chain on tau filament assembly. are based on the chemical conjugation of protein precursors [31]. Chemical ubiquitination strategies based on non-native isopeptide bond formation are often easier to implement and give high yields of the product conjugates. Among these, disulfide-coupling chemistry has proven to be highly versatile, efficient and robust, and has already been applied in many studies [30][31][32][33][34]. As a replacement for isopeptide linkage, this semisynthetic method generates a disulfide bond between a Cys residue placed in a specific position of the target protein and the C-terminal aminoethanethiol of a ubiquitin derivative obtained by intein processing. Recently, we described the production and characterization of tau protein mono-ubiquitinated at three different positions using disulfide-directed methodology [30]. With this approach, we obtained novel insight into the diverse effects that lysine monoubiquitination at different sites exerts on tau protein aggregation.
Here, we introduce a method based on directed disulfide bond formation in combination with enzymatic synthesis of ubiquitin chains to obtain controlled polyubiquitination of tau protein, thus adding a further level of complexity in the study of ubiquitinated tau protein. We produced and characterized the aggregation process of tau protein species homogenously modified with diubiquitin molecules at position 353, one of the ubiquitinated sites found in AD-brain filaments. The versatility of the method allowed us to incorporate two types of di-ubiquitin chains (Lys48-or Lys63chains) characterized by different structures, and to explore the impact of these prototypical ubiquitin polymers on tau protein aggregation.

Results
In the present work, we optimized a semisynthetic strategy based on disulfide-coupling chemistry to covalently attach a di-ubiquitin molecule to the tau protein. Ubiquitin-substrate conjugates obtained using this method are functional mimics of the native ubiquitinated counterparts, with the synthetic linkage being only one atom longer than the native isopeptide bond [32,34]. We selected the construct tau4RD, which represents a short form of the protein, spanning residues Q244-E372 plus an initial Met (Figure 1). Tau4RD has been widely used to study the aggregation behavior of tau protein as it includes the majority of the microtubule binding region and most of the residues involved in the assembly of the cross-β structure forming the core of the tau filaments [35]. To obtain site-specificity, we used a tau4RD mutant devoid of the endogenous cysteines and where a single cysteine was installed at the desired position. We chose to introduce the modification at position 353 because this site was found polyubiquitinated in PHF-tau [26,28], it is part of the fibril core found in different tauopathies [26], and mono-ubiquitination at this position had an observable influence on the aggregation kinetics without hindering the formation of short fibrils [30]. Thus, it was deemed suitable for evaluating the effect of an extended ubiquitin chain on tau filament assembly. The method of preparation of di-ubiquitinated tau4RD is illustrated in Figure 2. First, using linkage-specific enzymes, we assembled di-ubiquitin molecules with Lys48-and Lys63-linkage ( Figure 2b, and 3a,b). We used the ubiquitin mutants Lys48Arg or Lys63Arg as distal units and Ub- The method of preparation of di-ubiquitinated tau4RD is illustrated in Figure 2. First, using linkage-specific enzymes, we assembled di-ubiquitin molecules with Lys48-and Lys63-linkage (Figures 2b and 3a,b). We used the ubiquitin mutants Lys48Arg or Lys63Arg as distal units and Ub-SH, a ubiquitin derivative bearing a C-terminal aminoethanethiol linker obtained from intein cleavage with cysteamine, as proximal unit (Figure 2a). Homogenous di-ubiquitin chains ending with a C-terminal thiol (Ub 2 (48/63)-SH) were then purified and allowed to react with 5,5 -Dithiobis(2-nitrobenzoic acid) (DTNB) to produce asymmetric activated disulfides (Figures 2c and 4a,b). Finally, the activated disulfides were incubated with tau4RD, modified with a unique cysteine at position 353, to produce the disulfide-linked conjugates Ub 2 (48)tau4RD(353) and Ub 2 (63)tau4RD(353) (Figure 2c). The protein conjugates were obtained at purity of >95% ( Figure 3c) and their identity was verified by mass spectrometry (Figure 4c,d). These samples were then used to perform aggregation experiments and their behavior was compared with that of mono-ubiquitinated Ub-tau4RD(353) and of the unconjugated cysteine-free protein (tau4RD∆C). SH, a ubiquitin derivative bearing a C-terminal aminoethanethiol linker obtained from intein cleavage with cysteamine, as proximal unit (Figure 2a). Homogenous di-ubiquitin chains ending with a C-terminal thiol (Ub2(48/63)-SH) were then purified and allowed to react with 5,5′-Dithiobis(2nitrobenzoic acid) (DTNB) to produce asymmetric activated disulfides ( Figure 2c, and 4a-b). Finally, the activated disulfides were incubated with tau4RD, modified with a unique cysteine at position 353, to produce the disulfide-linked conjugates Ub2(48)tau4RD(353) and Ub2(63)tau4RD(353) ( Figure  2c). The protein conjugates were obtained at purity of > 95% ( Figure 3c) and their identity was verified by mass spectrometry (Figure 4c,d). These samples were then used to perform aggregation experiments and their behavior was compared with that of mono-ubiquitinated Ub-tau4RD(353) and of the unconjugated cysteine-free protein (tau4RD∆C).    The kinetics of filament formation of the prepared proteins was followed by monitoring changes in Thioflavin T (ThT) fluorescence (Figure 5a,b). In this assay, the fluorescence emission of ThT increases upon its specific binding to the β-sheet rich structure characteristic of tau filaments. The sigmoidal profile of the kinetic curves reflects the cooperative nature of the nucleation-dependent aggregation process. The initial flat portion of the curve, corresponding to the lag phase, is followed by a steep transition (the growth or elongation phase) and a flat terminal part (plateau phase). As shown previously [30], aggregation of tau4RD∆C in the presence of heparin is very rapid, characterized by an early transition midpoint at ∼5 h, and a fast fibril growth, with an elongation time of 0.7 ± 0.2 h (Figure 5a,b, Table 1). The addition of unconjugated di-ubiquitin molecules (Ub2(48/63))   The kinetics of filament formation of the prepared proteins was followed by monitoring changes in Thioflavin T (ThT) fluorescence (Figure 5a,b). In this assay, the fluorescence emission of ThT increases upon its specific binding to the β-sheet rich structure characteristic of tau filaments. The sigmoidal profile of the kinetic curves reflects the cooperative nature of the nucleation-dependent aggregation process. The initial flat portion of the curve, corresponding to the lag phase, is followed by a steep transition (the growth or elongation phase) and a flat terminal part (plateau phase). As shown previously [30], aggregation of tau4RD∆C in the presence of heparin is very rapid, characterized by an early transition midpoint at ∼5 h, and a fast fibril growth, with an elongation time of 0.7 ± 0.2 h (Figure 5a,b, Table 1). The addition of unconjugated di-ubiquitin molecules (Ub2(48/63)) The kinetics of filament formation of the prepared proteins was followed by monitoring changes in Thioflavin T (ThT) fluorescence (Figure 5a,b). In this assay, the fluorescence emission of ThT increases upon its specific binding to the β-sheet rich structure characteristic of tau filaments. The sigmoidal profile of the kinetic curves reflects the cooperative nature of the nucleation-dependent aggregation process. The initial flat portion of the curve, corresponding to the lag phase, is followed by a steep transition (the growth or elongation phase) and a flat terminal part (plateau phase). As shown previously [30], aggregation of tau4RD∆C in the presence of heparin is very rapid, characterized by an early transition midpoint at~5 h, and a fast fibril growth, with an elongation time of 0.7 ± 0.2 h (Figure 5a,b, Table 1). The addition of unconjugated di-ubiquitin molecules (Ub 2 (48/63)) at an equimolar ratio with tau4RD∆C did not affect the aggregation curves (Figure 5a), excluding the possibility that significant tau-ubiquitin inter-molecular contacts interfered with filament formation. Moreover, in the experimental conditions used, neither Lys48-nor Lys63-linked di-ubiquitin formed fibrils by themselves ( Figure 5a), thereby allowing us to interpret the experimental data in terms of ubiquitination-induced perturbations of the aggregation of tau protein.
at an equimolar ratio with tau4RD∆C did not affect the aggregation curves (Figure 5a), excluding the possibility that significant tau-ubiquitin inter-molecular contacts interfered with filament formation. Moreover, in the experimental conditions used, neither Lys48-nor Lys63-linked di-ubiquitin formed fibrils by themselves (Figure 5a), thereby allowing us to interpret the experimental data in terms of ubiquitination-induced perturbations of the aggregation of tau protein. The sigmoidal aggregation profiles obtained with Ub2(48)tau4RD(353) and Ub2(63)tau4RD(353) indicate that di-ubiquitinated tau protein samples were capable of forming filamentous aggregates (Figure 5b). From visual inspection of the aggregation curves measured on all of the tau protein samples, we noted a significant difference in the maximum fluorescence at plateau for the monoubiquitinated tau protein in comparison with the other proteins. Because signal intensity is influenced by specific amyloid properties and the proteins under investigation have different structures and shapes, the reduction in ThT fluorescence observed for mono-ubiquitinated tau protein was likely due to a different affinity for ThT and not to the number of fibrils formed.
Based on a quantitative analysis of the aggregation kinetics, we found that filaments formation by di-ubiquitinated tau protein was significantly delayed compared to the unconjugated protein, as deduced by the longer lag phase and elongation time of the di-ubiquitinated proteins (Figure 5b, Table 1). Additionally, di-ubiquitination was found to specifically affect the transition midpoint and elongation time more than mono-ubiquitination. Indeed, Ub2(48)tau4RD(353) and Ub2(63)tau4RD(353) displayed a similar transition half time of ∼21 h, which was significantly shifted The sigmoidal aggregation profiles obtained with Ub 2 (48)tau4RD(353) and Ub 2 (63)tau4RD(353) indicate that di-ubiquitinated tau protein samples were capable of forming filamentous aggregates (Figure 5b). From visual inspection of the aggregation curves measured on all of the tau protein samples, we noted a significant difference in the maximum fluorescence at plateau for the mono-ubiquitinated tau protein in comparison with the other proteins. Because signal intensity is influenced by specific amyloid properties and the proteins under investigation have different structures and shapes, the reduction in ThT fluorescence observed for mono-ubiquitinated tau protein was likely due to a different affinity for ThT and not to the number of fibrils formed.
Based on a quantitative analysis of the aggregation kinetics, we found that filaments formation by di-ubiquitinated tau protein was significantly delayed compared to the unconjugated protein, as deduced by the longer lag phase and elongation time of the di-ubiquitinated proteins (Figure 5b, Table 1). Additionally, di-ubiquitination was found to specifically affect the transition midpoint and elongation time more than mono-ubiquitination. Indeed, Ub 2 (48)tau4RD(353) and Ub 2 (63)tau4RD(353) displayed a similar transition half time of~21 h, which was significantly shifted with respect to the value determined for Ub-tau4RD(353) (t 0.5~1 5 h), and all the values of modified tau protein were larger compared to that of tau4RD ∆C (t 0.5~5 h). Likewise, the elongation time showed an analogous trend for the investigated samples (Table 1). By comparison of the aggregation curves, it emerged that the kinetics of the two di-ubiquitinated proteins was similar.
Ub 2 (48)tau4RD(353) and, to a lesser extent Ub 2 (63)tau4RD(353), produced aggregates that reacted with A11 (Figure 5c), an antibody capable of recognizing prefibrillar oligomers of diverse proteins [36]. Thus, the ability of the tau protein component to form intermediate amyloidogenic species was maintained after modification, as shown previously for mono-ubiquitinated tau4RD [30]. The ability of both Ub 2 (48)tau4RD(353) and Ub 2 (63)tau4RD(353) to form mature fibrils was confirmed by the TEM images (Figure 6a,b), which showed the presence of well-formed twisted filaments. The morphological analysis revealed that filaments were characterized by a large width of 20 ± 2 nm and a narrow width of 15-16 nm, and a twist crossover repeat of 57-58 nm ( Table 2), indicating that the overall morphology of tau4RD∆C filaments was not heavily modified by di-or mono-ubiquitin conjugation (Figure 6a-d, Table 2).
Taken together, the obtained data clearly indicate that the incorporation of protein modifiers in the microtubule binding domain of tau protein at position 353 interferes with the aggregation mechanism but does not abrogate the formation of mature fibrils. The addition of one ubiquitin moiety to tau protein determines the inhibition of fibrils formation. The inhibitory effect is even stronger when a second ubiquitin moiety is attached to the proximal ubiquitin unit, as it resulted from the substantial increase of the duration of both midpoint transition and elongation time for the di-ubiquitinated species. However, despite their known structural differences, the topology of the investigated di-ubiquitin molecules (Lys48-or Lys63-linked) did not influence the process of fibrils formation. Thus, it appears that the inhibitory effect of di-ubiquitination at the 353 site is caused by the increased steric hindrance which impairs microscopic events that lead to fibrils formation. with respect to the value determined for Ub-tau4RD(353) (t0.5 ∼15 h), and all the values of modified tau protein were larger compared to that of tau4RD ΔC (t0.5 ∼5 h). Likewise, the elongation time showed an analogous trend for the investigated samples (Table 1). By comparison of the aggregation curves, it emerged that the kinetics of the two di-ubiquitinated proteins was similar. Ub2(48)tau4RD(353) and, to a lesser extent Ub2(63)tau4RD(353), produced aggregates that reacted with A11 (Figure 5c), an antibody capable of recognizing prefibrillar oligomers of diverse proteins [36]. Thus, the ability of the tau protein component to form intermediate amyloidogenic species was maintained after modification, as shown previously for mono-ubiquitinated tau4RD [30]. The ability of both Ub2(48)tau4RD(353) and Ub2(63)tau4RD(353) to form mature fibrils was confirmed by the TEM images (Figure 6a,b), which showed the presence of well-formed twisted filaments. The morphological analysis revealed that filaments were characterized by a large width of 20 ± 2 nm and a narrow width of 15-16 nm, and a twist crossover repeat of 57-58 nm ( Table 2), indicating that the overall morphology of tau4RDΔC filaments was not heavily modified by di-or mono-ubiquitin conjugation (Figure 6a-d, Table 2).
Taken together, the obtained data clearly indicate that the incorporation of protein modifiers in the microtubule binding domain of tau protein at position 353 interferes with the aggregation mechanism but does not abrogate the formation of mature fibrils. The addition of one ubiquitin moiety to tau protein determines the inhibition of fibrils formation. The inhibitory effect is even stronger when a second ubiquitin moiety is attached to the proximal ubiquitin unit, as it resulted from the substantial increase of the duration of both midpoint transition and elongation time for the di-ubiquitinated species. However, despite their known structural differences, the topology of the investigated di-ubiquitin molecules (Lys48-or Lys63-linked) did not influence the process of fibrils formation. Thus, it appears that the inhibitory effect of di-ubiquitination at the 353 site is caused by the increased steric hindrance which impairs microscopic events that lead to fibrils formation.  After having established how di-ubiquitination affects tau fibril formation, a process associated with disease, we set out to describe its consequence on tau protein functional activity. Specifically, we investigated the impact of Lys48-and Lys63-linked di-ubiquitin on tau-mediated tubulin polymerization. The assay was performed using Ub 2 (48)tau4RD(353) or Ub 2 (63)tau4RD(353), and in the presence of tau4RD∆C as a control (Figure 5d). Microtubule (MT) polymerization was monitored by following the increase in absorbance at 350 nm. The kinetics of MT assembly in the presence of tau4RD∆C resembled previous results reported on tau4RD [37]. The data acquired in the presence of Ub 2 (48)tau4RD(353) or Ub 2 (63)tau4RD(353) indicates that the presence of the di-ubiquitin chains at position 353 of tau protein, moderately but significantly inhibits MT polymerization. Indeed, after 300 min of incubation of tubulin with the conjugates, we observed~77% of MT formation (referred to 100% for tau4RD∆C). This effect was independent from the topology of the di-ubiquitin chain, as the observed curves were almost superimposable. 3.0 ± 0.9 Table 2. Morphological properties of the twisted filaments of tau protein samples obtained from the analysis of TEM images.

Discussion
The microtubule-associated protein tau plays a central role in the pathogenesis of AD. Besides being heavily phosphorylated [20], tau protein in PHFs from AD brains is found to be mono-or polyubiquitinated at multiple sites [26][27][28]. The observation that most of these sites belong to the microtubule binding region, a domain involved in the formation of the filaments core, suggests the attempt of the neurons to get rid of aggregated tau protein forms via the UPS. This hypothesis is supported by the fact that PHF-tau was found to be modified by Lys48-and Lys11-linked chains [28], both being signals for proteasomal degradation, and by mono-ubiquitin, recently recognized as an additional signal for proteasomal targeting [38]. These observations suggest a possible role of ubiquitination in the formation and clearance of pathological tau protein species.
In a bid to advance our understanding of the impact of ubiquitination on tau protein aggregation and function, we recently started developing methods to attain controlled ubiquitination of tau protein samples for molecular-level investigations. In a previous work, we produced mono-ubiquitinated tau4RD samples and found that the impact of the modification on tau protein fibrillogenesis was site-dependent [30]. While ubiquitination at Lys311, located within the PHF motif of the R3 domain, essentially abolished filament formation, modification at Lys254 and Lys353 changed the aggregation kinetics but did not arrest fibrillation completely. To follow up on this study, in the present work, we set out to investigate whether a minimal polyubiquitin chain, a di-ubiquitin, could affect the aggregation behavior of tau protein.
To obtain site-specific di-ubiquitination, we combined enzyme-mediated preparation of di-ubiquitin derivatives with disulfide-directed ligation to tau4RD, expanding our approach originally developed to produce mono-ubiquitination [30]. Based on the previous findings with mono-ubiquitination, di-ubiquitin was installed at position 353, a site that was deemed suitable for evaluating the effect of an extended ubiquitin chain on filament assembly. Because polyubiquitin chains exist in a variety of topologies and structures, here we chose to investigate the prototypical Lys48-and Lys63-linked chains as being representative of polymers that preferentially adopt compact or extended conformations, respectively [7]. Lys48-linked di-ubiquitin predominantly adopts a closed conformation in which the functional hydrophobic patches of mono-ubiquitin are sequestered at the ubiquitin/ubiquitin interface [9]. In contrast, Lys63-linked di-ubiquitin chains are extended and adopt an open conformation with no direct contact between the neighboring ubiquitin subunits, thereby exposing the hydrophobic patches of mono-ubiquitin [10]. Thus, we hypothesized that different architectures of polyubiquitin chains could differently affect substrate aggregation.
The performed aggregation experiments revealed that both di-ubiquitinated tau protein conjugates were less prone to form fibrils compared to the unconjugated protein (Figure 5b), although they retained the capability to form A11-positive prefibrillar oligomeric species (Figure 5c) and twisted filaments ( Figure 6). Moreover, the dimeric ubiquitin modifier increased significantly both the half-time of the transition and the fibril elongation time with respect to mono-ubiquitin, with the time constants following the order: di-ubiquitinated >> mono-ubiquitinated > unmodified ( Figure 5b, Table 1). However, the topological differences between the two di-ubiquitin chains (Lys48-or Lys63-linked) tested do not differentially interfere with the process of fibrils formation.
It is established that the duration of the lag phase depends on the rates of multiple parallel microscopic processes, such as the formation of primary nuclei and their amplification through elongation and secondary nucleation processes [39]. Because variations of the former process do not affect the growth phase, while changes in elongation and secondary nucleation modify both the lag phase and the growth phase, it seems possible that di-ubiquitination of tau protein could interfere with the growth and proliferation of primary nuclei rather than with their formation. This was more evident for di-ubiquitinated, rather than mono-ubiquitinated, tau protein. Since the inhibitory effect was dependent on the length of the ubiquitin chain, the impact on tau protein aggregation is likely be a result of a combination of structural motif at the tau-ubiquitin conjugating point and of the presence of the distal ubiquitin moiety which adds steric hindrance. Yet, we do not exclude the possibility that a further elongation of the polyubiquitin chain could elicit different results, particularly if considering that longer ubiquitin chains have lower thermodynamic stability and can form fibrils themselves under specific conditions [40].
To interpret our results in the wider context of post-translational modifications of tau protein, we reviewed recent studies which explored the impact of single or multiple phosphorylation of full-length tau protein or tau4RD [37,41]. Phosphorylation of both Ser262 and Ser356 was found to significantly alter the aggregation kinetics of tau4RD [37], and single-site phosphorylation affected tau protein aggregation in a sequence-specific manner [37,41]: phosphorylation of Ser305 had a significant impact on fibril formation; however modification at Ser356 did not significantly perturb the aggregation of tau protein.
Here we have shown that di-ubiquitination at 353, close to residue 356, allows the formation of fibrils, although with a reduced aggregation rate, in line with the observation that this site is excluded from the ordered core (residues 272-330) in heparin-induced tau filaments [42]. Additionally, recent cryo-EM studies show that tau filaments from brain tissues [25,26] are characterized by a long core that includes Lys353 and suggest that bulky modifiers, such as mono-or di-ubiquitin, in that position could affect the mechanism of formation of tau protein aggregates.
Tau protein is known to play a central role in the assembly and stabilization of microtubules [18]. Therefore, we explored whether di-ubiquitination could regulate tau-mediated tubulin polymerization. Based on our observations, it emerges that the presence of either Lys48-or Lys63-linked di-ubiquitin at position 353 moderately but significantly inhibits MT polymerization. This effect was found to be independent of the di-ubiquitin linkage topology (Figure 5d). Our findings are consistent with previous studies reporting that multiple mono-ubiquitination of tau protein reduces MT binding affinity [43], and phosphorylation at Ser356 (close to Lys353) slightly impairs MT assembly [37]. Taken together, the evidence on site-specific phosphorylated or on multi mono-ubiquitinated tau protein, and our results obtained with the di-ubiquitinated proteins indicate that post-translational modifications around position 353 do not crucially affect tau-assisted MT polymerization and that the functional activity of tau protein is highly dependent on the sites of modification.
In conclusion, we were able to obtain, for the first time, samples of tau protein conjugated to Lys48-or Lys63-linked di-ubiquitin at a specific site and to study the effect of these modifications on tau protein aggregation and function in comparison with the mono-ubiquitinated species. We demonstrated that the conjugation of tau protein to di-ubiquitin chains at the 353 position significantly delayed, but did not completely inhibit, tau protein aggregation, in analogy with mono-ubiquitination. Quantitative analysis of the aggregation kinetics revealed a more pronounced effect of di-ubiquitin compared to mono-ubiquitin during elongation of filaments, rather than during primary nucleation events. However, linkage topology had a minor effect on the measured kinetic parameters. Finally, we demonstrated that di-ubiquitination modulated the tau-mediated microtubule assembly, thus highlighting the potential role of ubiquitination in the regulation of tau protein function. For a more comprehensive overview, further efforts will have to be directed to produce tau protein modified with longer polyubiquitin chains linked in different positions.
We believe that our findings also provide crucial information in light of recent structural studies on tau filaments from degenerated brain tissues. Ubiquitin is proposed to play a structural role by mediating specific inter-protofilament packing and promoting the formation of fibril subtypes, specific of the different tauopathies [26]. Moreover, future investigations are required to elucidate the synergic effects played by different post-translational modifications, such as ubiquitination, phosphorylation and acetylation, in the transition of tau protein to toxic species, with the aim to understand their implications in the different pathologies.
The ubiquitin mutants K48R, K63R, D77 were produced with the same protocol used for wild-type ubiquitin that is described in [44].
The recombinant enzymes human His-tagged E1, GST-tagged E2-25K, yeast His-tagged Mms2, yeast GST-tagged Ubc13 were produced as previously described [45]. In the present work, the GST tag was removed from the E2-25K protein by incubating the clean fusion protein attached to the GSH-resin with thrombin.
To obtain a ubiquitin bearing an aminoethanethiol C-terminal group (Ub-SH) required for the disulfide-coupling reaction, we first produced a chimeric protein where ubiquitin was cloned to the N-terminal of the GyrA intein in a pET22 vector. In this way, the chimeric protein has a C-terminal His-tag. The Ub-intein-His protein was produced in BL21(DE3) cells at 37 • C overnight in auto-inducing medium and purified by immobilized nickel affinity chromatography (IMAC) according to standard protocols. Cleavage of Ub-SH was obtained by incubating the clean fusion protein in a buffer at pH 7.5 containing: Tris-HCl 20 mM, EDTA 1 mM, cysteamine 40 mM, and Tris(2-carboxyethyl)phosphine (TCEP) 3 mM for 48 h at 10 • C. Ub-SH was further purified by reverse -IMAC and a superdex-75 gel filtration column when required.
It is important to note the thiol containing proteins Ub 2 (48/63)-SH and tau4RD(353) were first incubated with a large excess of DTT that was then properly removed by size exclusion chromatography just before the disulfide coupling reaction.
The procedure to obtain mono-ubiquitinated tau protein at position 353 (Ub-tau4RD(353)) by a similar disulfide-coupling strategy is described in [30].

Thioflavin T Aggregation Assay
Thioflavin T aggregation assays were carried out in 96-well dark plates in a Tecan Infinite M200PRO Plex plate reader at 30 • C with cycles of 30 s of orbital shaking at 140 rpm and 10 min of rest throughout the incubation. ThT fluorescence measurements were taken every 11 min, using excitation wavelength of 450 nm and recording fluorescence emission at 480 nm.
Samples contained 0.01 mM proteins in 20 mM sodium phosphate buffer at pH 7.4 and 50 mM NaCl (with 0.02% NaN 3 and protease inhibitors with EDTA), incubated with equimolar amount of heparin and ThT. Samples containing tau protein variants were filtered through a 100 kDa cut-off filter (Sartorius) to remove pre-existing large oligomers. Each measurement was performed in three replicates. The aggregation curves were analyzed by fitting each individual experimental data set with the following sigmoidal function [48]: