Aggregation Kinetics and Filament Structure of a Tau Fragment Are Influenced by the Sulfation Pattern of the Cofactor Heparin

A pathological signature of Alzheimer’s disease (AD) is the formation of neurofibrillary tangles comprising filamentous aggregates of the microtubule associated protein tau. Tau self-assembly is accelerated by polyanions including heparin, an analogue of heparan sulfate. Tau filaments colocalize with heparan sulfate proteoglycans (HSPGs) in vivo, and HSPGs may also assist the transcellular propagation of tau aggregates. Here, we investigate the role of the sulfate moieties of heparin in the aggregation of a recombinant tau fragment Δtau187, comprising residues 255–441 of the C-terminal microtubule-binding domain. The effects that the selective removal of the N-, 2-O-, and 6-O-sulfate groups from heparin have on the kinetics of tau aggregation, aggregate morphology, and protein structure and dynamics were examined. Aggregation kinetics monitored by thioflavin T (ThT) fluorescence revealed that aggregation is considerably slower in the presence of 2-O-desulfated heparin than with N- or 6-O-desulfated heparin. Transmission electron microscopy revealed that tau filaments induced by 2-O-desulfated heparin were more slender than filaments formed in the presence of intact heparin or 6-O-desulfated heparin. The 2-O-desulfated heparin-induced filaments had more extensive regions of flexibility than the other filaments, according to circular dichroism and solid-state NMR spectroscopy. These results indicate that the sulfation pattern of heparin regulates tau aggregation, not purely though electrostatic forces but also through conformational perturbations of heparin when the 2-O-sulfate is removed. These findings may have implications for the progression of AD, as the sulfation pattern of GAGs is known to change during the aging process, which is the main risk factor for the disease.


Introduction
The assembly of microtubule-associated protein tau (MAPT, or tau; UniProtKB P10636) into filamentous aggregates is a pathological hallmark of neurogenerative diseases including Alzheimer's disease (AD) and Pick's disease, collectively known as tauopathies. 1,2 Tau is a water-soluble, intrinsically disordered protein under normal physiological conditions and its function is to assist the assembly and stabilisation of microtubules and other neuronal cytoskeletal elements. Six isoforms of tau occur in the adult human brain; 3 the largest (441-aa) isoform consists of two N-terminal inserts (N1, N2) and four repeat units (R1-R4) in the microtubule-binding C-terminal region. The shorter isoforms lack one or both of the N1 and N2 units and/or the R2 unit.
Neurofibrillary tangles of abnormally phosphorylated, aggregation-prone tau occur in the brains of patients affected by AD, and these consist predominantly of insoluble paired helical filaments (PHF) and straight filaments (SF). 4,5 The filaments consist of an ordered -sheet amyloid core flanked by a fuzzy, unstructured coat of around 200 residues that is invisible to transmission electron microscopy (TEM). 6 Recent visualisation by electron cryo-microscopy, supported by earlier solid-state NMR studies, 7,8 have revealed that the nanoscale polymorphism of tau PHF and SF originates from disease-specific molecular conformations of tau within the fibres, 5,9 raising the possibility that neuropathological strains may propagate in a prion-like mechanism. The molecular processes that lead to the formation and propagation of these strains in vivo are far from clear and our understanding has until recently depended upon mechanistic studies in vitro.
Studies of the aggregation mechanisms of intact or truncated tau proteins in vitro require the addition of arachidonic acid or polyanionic cofactors such as RNA and, most commonly, heparin to initiate aggregation. 10,11 Recent results from tissuederived filaments pose the question of whether the heparin-induced tau aggregation is pathologically relevant. 12,13 Heparin-induced filaments are heterogeneous, 13 possibly reflecting the heterogeneity and polydispersity of the heparin cofactor, and the protein molecular conformation differs somewhat from the structures associated with AD or Pick's. It is worth noting, however, that there is insufficient information about the structural variability of filaments obtained from different AD or Pick's brains, given the limited current data and challenges involved. Further, heparin is closely related to heparan sulfate, which was found early on to accumulate with neurofibrillary tangles in the AD brain 14 and, when conjugated to proteoglycans, assist the cellular uptake and possible propagation of tau via a prion-like seeding mechanism 15 . Scheme 1. (A) Sequence of the four microtubule binding repeats (R1-R4) and strand regions of the heparin-induced 2N4R tau polymorphs (orange, snake; blue, twisted; green, jagged) identified by cryo-EM. (B) Generic structure of the heparin disaccharide unit and list of the desulfated heparin derivatives prepared.
It has often been assumed that heparin-tau interactions are driven largely by non-specific ionic interactions between the protein and the sulfate and carboxylate groups of heparin, 16

Tau expression
The tau construct comprising residues 255-441 of human 4R tau (cDNA clone htau46) with the aggregation impeding N terminus removed, leaving the 2 nd and 3 rd repeat microtubule binding (MTB) units, including the highly amyloidogenic sequences VQIINK and VQIVYK, respectively. The protein was expressed and purified as previously described. 20

Transmission Electron Microscopy
Tau and heparin (20 and 5 μM, respectively) were incubated with Tris (30 mM), DTT (1 mM) at pH 7.5 at 37 °C for 24 hours. A 10 μL suspension was spotted onto carbon coated formar grids. After 5 minutes the excess liquid was removed via blotting. For negative staining, 10 μL of 2 % phosphotungstic acid was spotted onto the loaded grids, and left for 3 minutes before blotting the excess. Grids were viewed on a Jeol JEM-1010 electron microscope and images captured that were representative of the entire grid. Fibril widths were measured using ImageJ software and the averages from 50 measurements was calculated. The simulated 13 C-13 C spectrum (Supporting Information, Figure S1) was calculated from chemical shifts predicted from the cryo-EM structure of heparin induced 2N4R tau snake filaments, 12 using a C program written specifically for this purpose.

Desulfation of heparin influences tau187 aggregation kinetics
Incubation of tau187 ( Figure 1D). In summary, the overall rate of heparin-induced tau187 aggregation to completion is impeded to a greater extent by removing the 2-O-sulfates of heparin than it is by removing the 6-O-or N-sulfates. Although the data for tau187 is normalized to the final fluorescence emission that was measured, the curve is clearly continuing in an upward trajectory, indicating that aggregation had not reached its conclusion even after 12 h, and t0.5 is likely to be considerably longer than 6 h. The different effect of the selectively desulfated heparin derivatives argues against the tauheparin interaction being mediated only by non-specific charge interactions. Other chemical and/or conformational properties must be considered.

Concentration dependence of desulfated heparin-induced tau187 aggregation
To further investigate the effect of the sulfate groups of heparin on tau187 aggregation, the kinetics were followed (by ThT) at different concentrations of LMWH or  Table 1). The retarding effect of heparin at higher concentrations is known and has been attributed to a screening effect of high heparin concentrations because of increased ionic strength. 30 It has also been proposed that high heparin concentrations promote the formation of dead-end, off-pathway tau-heparin complexes. 31 A further explanation, which will be explored in the next section, is that higher heparin concentrations increase the number of 1:1 tau:heparin complexes that are below the critical mass necessary to nucleate filament growth.  Table 1. Means and error bars are shown for triplicate measurements.
When tau187 is incubated with 6-O-desulfated or N-acetylated LMWH, a 2-3-fold overall decrease in the rate of aggregation is seen relative to unmodified LMWH, with t0.5 being 0.9 -1.7 h (Figure 2, top right and bottom left; Table 1). Aggregation again follows a sigmoidal profile and the reduced rate appears to originate chiefly from a reduction in kn  Table   1. These rate constants describe a cofactor-independent aggregation mechanism represented by the equations, 27, 28 Here, M(t) and P(t) are the filament mass and number, respectively, at time t and m(t) is the monomer concentration. The term ( ) represents the formation of primary nuclei from monomers (m) with reaction order nc (typically assigned a value of 2). The term 2 ( ) 2 ( ) describes secondary nucleation in a reaction of order n2 (typically given a value of 2 or 3) with respect to the mass of monomer. The term Table 1. Summary of the apparent rate constants for tau aggregation in the presence of unmodified heparin and desulfated derivatives, obtained from the lines of best fit to the data in Figure 2. The calculations were also based on fixed values for the fibril dissociation rate constant (km = 1.0 M -1 s -1 ) and the reaction orders of primary nucleation (nc = 2) and secondary nucleation (n2 = 2). 27,28 The reaction orders do not necessarily correspond to the size of the nuclei, but heparin has been shown to interact with two tau molecules forming a dimer that nucleates fibril growth. 31 Dashes indicate where constants could not be calculated because of poor fits to the experimental data. Errors in t0.5 are given in parentheses.

[m]T and [h]T are the total concentrations of free protein monomer and free
heparin and Kd is the dissociation constant. The equilibrium is perturbed as mh recruits further monomers to form a nucleating species. Aggregation of the 4-repeat domain tau construct in the presence of heparin has been shown to be nucleation dependent, with a single heparin molecule binding tau forming an aggregationpromoting dimer, which then serves as a building block for further fibril growth. 31 We therefore assume that the smallest primary nucleating species N consists of two tau molecules bound to one heparin molecule, m2h. where nc = 1. Equations [3] and [4] thus take into account the role of heparin in the initiation of tau aggregation. Equation [1] becomes These equations likely oversimplify the tau-heparin interaction, but serve as a working hypothesis until the details of the model are further refined with experimental input. The simulated curves share similarities with the experimental data in Figure 2, the simulations do not replicate exactly the behaviour of the ThT curves and so quantitative analysis using global fitting of equations [4] and [5] to the curves was not attempted. However, qualitative inspection of numerically-simulated curves based on these adaptations suggest that t0.5 for aggregation and filament mass are both sensitive to heparin concentration and to Kd (Figure 3).
An equilibrium association constant exceeding 10 6 M -1 (i.e. Kd < 1 M) for tau and heparin has been reported. 30 According to Eq.  Table 1), although the latter curve is consistent with much slower aggregation than represented by the simulation. Hence, removal of the 2-O-sulfate group does not appear to reduce the affinity of heparin for tau187 according to these curves. The slower rate of tau aggregation in the presence of LMW-2OH, compared to the rates in the presence of the other heparin derivatives is reflected in the lower rate constants kn, k+ and k2 calculated by curve fitting (Figure 2 and Table 1), which indicates that 2-O-sulfate group is critical for tau primary and secondary nucleation and filament elongation. although there are also some important differences that indicate that our model is incomplete.

Heparin 2-O-desulfation, but not 6-O-desulfation, affects the structure and morphology of tau187 aggregates
We next used circular dichroism (CD) spectroscopy to follow the structural transformation accompanying tau187 aggregation in the presence of HMWH, HMW-6OH and HMW-2OH. CD has indicated that tau aggregation accompanies a transition from an unfolded structure to a partially folded structure with approximately ~36 % sheet. 32 Here, we obtained spectra from freshly prepared tau187-heparin solutions and again after incubation at 37°C for 8 h. In the absence of heparin, the spectrum changes  Filaments formed in the presence of HMW-2OH ( Figure 4C, right) are also sparsely distributed and display a sinusoidal curvature pattern. These filaments are noticeably more slender than the filaments formed with unmodified heparin, with mean width of 8.5 (± 2.6) nm ( Figure 4D, right). The width of these filaments is more typical of the single filaments that are the minor population of tau aggregates associated with AD. 33 INEPT-based methods that correlate nuclear spins through J-couplings, while the ordered core is observed selectively using Hartmann-Hahn cross-polarization combined with nuclear dipolar recoupling methods. 32,35 Here, two-dimensional 13 C-13 C dipolar correlation NMR spectra were obtained from fibrils formed by tau187 (20 μM) in the presence of 5 μM HMWH or HMW-2OH.
By using a short dipolar mixing time (10 ms), the spectra report on the motionallyrestrained residues forming the core of the fibrils. The more flexible outer residues give rise to weaker cross-peaks or are not observed. The peaks are quite broad, which probably reflects the known structural heterogeneity of tau filaments formed in the presence of heparin. 13 No attempt was made to sequentially assign the spectra owing to severe crowding and overlap, as the tau construct we studied is much larger than the constructs Figure 5. Solid-state magic-angle spinning 13 C NMR spectra of uniformly 13  previously studied by SSNMR. Sequential assignment and structural determination could in principle be achieved with a combination of 2D and 3D NMR experiments and selective isotope labelling, but this was beyond the scope of the present work. The experimental spectrum of tau filaments formed with HMWH is overlaid with a simulated spectrum (Supporting Information, Figure S1) generated from 13 C chemical shifts predicted from the cryo-EM structure of heparin-induced filaments (Table S1), 12 which are predominantly -sheet/hairpin folds. Good agreement between the simulated and experimental spectra suggests that the observed signals originate from the -sheet core of the fibrils. A striking reduction in cross-peak intensities, particularly for cross-peaks representing longer-range couplings, is seen in the spectrum of the fibrils induced by HMW-2OH, as compared to the spectrum of fibrils induced by HMWH ( Figure 5, A and B, top panels). The reduction in intensity is consistent with a higher degree of overall flexibility within the fibrils induced by HMW-2OH. In contrast to the 13 C-13 C spectra, 1 H- 13 C INEPT spectra of the exhibit many more peaks than are seen than in the spectrum of HMWH-induced filaments ( Figure 5, A

Discussion
Heparin is a convenient experimental aid to induce the rapid formation of tau filaments/fibrils in vitro for mechanistic and structural investigations. It is a close analog of the GAG heparan sulfate, and heparan sulfate proteoglycans (HSPGs) such as agrin are commonly associated with A plaques and NFTs in AD. The sulfation patterns of GAGs are known to change with ageing, 36, 37 the principal riskfactor for AD, and so there is biological incentive to investigate how GAG sulfation affects tau aggregation, using heparin as a representative GAG.
Experimental evidence gained over the last decade suggests that the progression of neurodegeneration in AD is driven by the transcellular propagation of tau aggregates, which seed protein aggregation in the recipient cells in a prion-like manner. 38,39 This process may not necessarily be limited to tau aggregates, as monomeric tau can also be internalised and seed the aggregation of endogenous tau. 40 The mechanism of tau release from neurons into the extracellular space is not currently known, but the cellular uptake of tau fibrils can occur via binding to HSPGs. 15 It is conceivable that HSPGs also promote the fibrillisation of tau monomers and oligomers in the extracellular space, just as heparin (and HS) do in vitro, in addition to facilitating the cellular uptake of fibrils.
Here we report that selectively altering the sulfation pattern of heparin affects the kinetics of heparin-induced tau187 aggregation and the mass of filaments formed, and can also modify the molecular structure of the filaments. The rate of aggregation is slower in the presence of all the desulfated heparins than in the presence of native heparin, which may be attributed in part to the lower negative We propose a basic mechanism for heparin-induced tau aggregation, which goes some way toward explaining the aggregation kinetics in the presence of the different heparin derivatives, but significant differences between the experimental ThT curves and calculated curves indicate that the model is incomplete. A more robust analysis may need to take into account the formation of oligomers, polymerization via multiple pathways and complex secondary nucleation processes involving protein fibrillar and heparin surfaces. As it stands, the basic model is consistent with the 6-O-desulfated and, to a lesser extent, N-desulfated heparin having lower affinity for tau than intact heparin. This interpretation agrees with previous work: using a heparin-immobilized chip, surface plasmon resonance was used to show that N-desulfation and 2-O-desulfation had no effect on heparin binding to a tau construct, whereas 6-O-desulfation severely reduced binding. 19 In addition, cellular studies have shown that heparin interacts with heparan sulfate binding sites in tau and prevents binding to cell-surface HSPGs, 15 but removal of the 6-O-sulfates from heparin abolishes the inhibitory effect whereas 2-O-desulfated heparin remains strongly inhibitory. 17 Further studies concur that 6-O-sulfation is critical for tau-heparan sulfate interactions and that this modification regulates uptake in human cell lines and mouse brain slice culture. 18 Interestingly, the 6-OS and NS glucosamine sulfates, but not the 2-OS iduronate sulfate, of heparin is required for binding to A fibrils, 41 suggesting a common function for this moiety in amyloid binding.
The most interesting finding here is that 2-O-desulfation of heparin results in a markedly greater reduction of the tau187 aggregation rate than does desulfation at the other two sites, and templates tau assembly into a different filament structure and morphology. These differences may be rationalised by considering the effects of removing sulfate groups on the conformation of heparin. In addition to losing a charge in the form of a sulfate group, the conformation of heparin derivatives is sensitive to the substitution pattern in several ways. The iduronate residue is known to be in an equilibrium of 1 C 4 and 2 S0 chair and skew boat forms in heparin (67:33 1 C 4 : 2 S0 ) 42  with relatively little perturbation of the overall conformation. 47 We propose that the unique conformational perturbations resulting from 2-O-desulfation must therefore modify the tau-heparin interaction in such a way that the conformation of tau is less amenable to self-assembly. One explanation, which is supported by the NMR data, is that the core -sheet region of tau is reduced and the flanking fuzzy coat region extends further into the core. Why a different filament structure and morphology occurs in the presence of 2-O-desulfated heparin can only be speculated upon at this stage. Alterations in the global and local conformation of heparin after 2-Odesulfation may reduce or otherwise change the charged surface that interacts with tau monomers. A distinct primary nucleating species may be formed that directs propagation along a structurally and kinetically altered pathway involving interactions between fewer core residues.
There is ample evidence that heparin sulfates accelerate the aggregation of monomeric tau in vitro. 48 Further work will be necessary to understand fully the nature of the tau species that interact with HSPGs during transcellular propagation, and to ascertain whether GAGs play a role in filament formation. It should be noted, that protein aggregates in the Alzheimer's brain principally contains phosphorylated tau (P-tau), rather than normal tau, and as such the higher negative charge density would likely repel polyanionic GAGs. The interplay between tau, P-tau and GAGs must therefore be considered in future investigations. The key message of this work is that to understand the role of GAG-protein interactions in the formation of amyloid fibrils, it is necessary to look beyond electrostatics and to consider the chemical and conformational effects conferred on GAGs by their sulfation patterns.