Effect of Phosphorylation and O-GlcNAcylation on Proline-Rich Domains of Tau

Abstract

Microtubule associated protein Tau (MAPT) is a phospho-protein within neurons of the brain. Aggregation of tau is the leading cause of tauopathies such as Alzheimer’s disease. Tau undergoes several post-translational modifications of which phosphorylation and O-GlcNAcylation are key chemical modifications. Tau aggregates into paired helical filaments and neurofibrillary tangles upon hyperphosphorylation whereas O-GlcNAcylation stabilizes the soluble form of Tau. How specific phosphorylation and/or O-GlcNAcylation events influence Tau conformations remains largely unknown due to the disordered nature of Tau. In this study, we have investigated the phosphorylation and O-GlcNAcylation induced conformational effects on a Tau segment (Tau_225–246) from the proline-rich domain (P2), by performing metadynamics simulations. We studied two different phosphorylation patterns; Tau_225–246 phosphorylated at T231 and S235, and Tau_225–246 phosphorylated at T231, S235, S237, and S238. We also study O-GlcNAcylation at T231 and S235. We find that phosphorylation leads to the formation of strong salt-bridge contacts with adjacent lysine and arginine residues, which disrupts the native β-sheet structure observed in Tau_225–246. We also observe the formation of a transient α-helix (²³⁸SAKSRLQ²⁴⁴) when Tau_225–246 is phosphorylated at four sites. In contrast, O-GlcNAcylation showed only modest structural effect and resembles the native form of the peptide. Our studies suggest the opposing structural effects of both PTMs and the importance of salt-bridges in governing the conformational preferences upon phosphorylation, highlighting the role of proximal Arginine and Lysine upon hyper-phosphorylation.

Graphical Abstract

Introduction

Protein posttranslational modification (PTM) is among the key phenomenon in regulating protein activity as well as its diversity in structure and function. To date, phosphorylation is the highly explored PTM as it can be detected in vivo and in vitro effortlessly.¹ Introducing a charged, anionic/dianionic tetrahedral phosphate group on Ser/Thr/Tyr side chain induces altered conformations in protein microenvironments, primarily due to the formation of salt bridges.² Another PTM analogous to protein phosphorylation is O-Glycosylation, which also targets the alcoholic side chains of Ser/Thr/Tyr. Amongst various O-glycosylation’s, the addition of β-linked N-acetylglucosamine (O-GlcNAc) is the only PTM which is observed within the nucleo-cytoplasmic compartments of all metazoans. O-GlcNAcylation and phosphorylation are involved in a complex interplay involving cell-signaling pathways, protein transcription and cytoskeletal regulatory proteins activity.^3,4 Protein phosphatase (an enzyme that removes O-phosphate) is found in a dynamic complex with O-GlcNAc transferase (an enzyme that attaches O-GlcNAc), indicating that, in many cases, this enzyme complex both removes O-phosphate and concomitantly attaches O-GlcNAc to serine and threonine.⁴ Many cellular functions are phosphoregulated. O-GlcNAcylation at these sites may interfere with critical control mechanisms due to reciprocal relationship with phosphorylation. Both modifications have been observed in a reciprocal relationship for tumor-associated SV-40 Large T antigen.^5–7 Several experimental studies have shown opposing structural and functional effects as well as cross-regulation of phosphorylation and O-GlcNAcylation.^6,8–10 Hence it is important to understand phosphorylation and O-GlcNAcylation induced structural effects on biologically crucial pathways such as cell signaling and etiology of the disease.

Tau (tubulin-associated unit), identified in 1975, is a family of microtubule (MT)-associated proteins which are expressed predominantly within neurites and axons in the adult brain.^11,12 The largest tau isoform (htau40) in the brain contains 441 amino acids comprising of two N-terminal inserts, proline-rich domain and four repeat domains that bind to MT (Figure 1).¹³ Tau interact with microtubules and stabilize the cytoskeleton of neurons.¹⁴ The short sequence ²²⁵KVAVVR²³⁰ localized to the second proline-rich region (P2) and the two hexapeptide motifs (²⁷⁵VQIINK²⁸⁰ and ³⁰⁶VQIVYK³¹¹) in repeats R2 and R3 shows the strongest interaction with MTs.^15–17 It has been found that the basic residues in these regions K225, R230, K274, and K281 are critical for MT binding and polymerization.^18,19

Figure 1.

Pictorial representation of tau protein, isoform htau40 (441 residues)⁴⁵. N1 and N2 are the two 29-residue inserts near the N-terminus; P1:I151-Y197 and P2: S198-L243 are proline-rich domain; R1-R4: Q244-N368, are the four repeats in the C-terminal half; R’ is 5^th repeat domain K369-S400 and C-terminal tail G401-L441. Definition of domains⁴¹: Projection domain (MI-Y197, does not bind to microtubules by itself), Assembly domain (S198-L441, binds to microtubules).

The neurofibrillary tangles (NFT) found in Alzheimer’s disease (AD) brains are composed primarily of the protein tau in a hyper-phosphorylated state, which disrupts microtubule assembly. Phosphorylation is the most common tau post-translational modification.²⁰ Tau protein consists of approximately 85 known phosphorylation sites among which there are 45 serine, 35 threonine, and only 5 tyrosine sites.^21–24 It is well documented that an increase in tau phosphorylation reduces its affinity for microtubules, which results in neuronal cytoskeleton destabilization depending on the number and location of phosphorylation sites.^25–28 Tau phosphorylation at key sites involving T231, S235, and S262 have a major role in dissociation of tau from microtubules. It was found that Tau phosphorylated at T231, S235 and S262 reduced their binding to microtubules by a factor of 26%, 9%, and 33% respectively when compared to native un-phosphorylated Tau.²⁹ Amniai et al. showed that phosphorylation of both Ser202/Thr205 and Thr231/Ser235 in the proline-rich domain of tau do not allow tubulin to polymerize into MT.³⁰ Phosphorylation at Thr231 and S235 is established as a valid biochemical marker for AD diagnosis.^31–33 Hyperphosphorylation of tau in AD is a direct result of it’s reduced O-GlcNAcylation.^34,35 AD is characterized by an inverse relation between tau phosphorylation and O-GlcNAcylation ^36–38 and in vitro, the de-glycosylation of tau tangles converts them into bundles of straight filaments and restores their accessibility to microtubules.³⁴ These studies suggest that O-glycosylation protects tau from hyper-phosphorylation and consequently is expected to prevent NFT formation.

Structurally Tau has been classified to belong to the family of intrinsically disordered proteins (IDP). The full-length crystal structure of Tau has not been solved. Solution studies of Tau and small fragments of Tau using predominantly NMR techniques have observed transient secondary structure elements, like α-helix and PPII, in specific regions of the sequence. Schwalbe et. al. performed NMR studies to gauge the structural impact of phosphorylation on the proline-rich region P2 (Tau_225–246) of tau which contains the T231 PTM site, which has been implicated in the binding of Tau with MT. As per the “jaws” microtubule-binding model proposed by Mandelkow et. al. and global hairpin structural model of tau, structural changes in the proline-rich regions could have a major effect on microtubule binding affinity.^31,39–41 Tau_225–246 also corresponds to a known recognition sequence of the WW domain of Pin1, which binds tau protein in a PPII conformation upon phosphorylation.⁴² These results suggest that hyperphosphorylation of proline-rich sequences in tau may lead to its global reorganization.

With this in mind, in this study, we investigate the effect of phosphorylation and O-GlcNAcylation on the proline-rich P2 domain (Tau_225–246) of Tau. Tau_225–246 is a 22-residue fragment of tau protein from the P2 (198–243) proline-rich domain just before the R1-R4 repeat units (Figure 1). In addition to T231, Tau_225–246 contains three additional PTM sites S235, S237, and S238. This allows us to study the influence of multiple PTMs on the overall structure of Tau_225–246. Additionally the availability of structural information from NMR experiments for native Tau_225–246, Tau_225–246 phosphorylated at two-sites (T231 and S235) and Tau_225–246 phosphorylated at four-sites (T231, S235, S237, and S238) allows us to compare the simulations directly to the experimental data. Owing to the disordered nature of the Tau peptides we perform molecular dynamics (MD) simulations using enhanced sampling techniques to improve the conformational sampling efficiency. We also explore the efficiency of the Amber ff99SBws⁴³ and CHARMM36m⁴⁴ force fields in describing the conformational dynamics of the IDPs and the influence of PTMs on the structure of the disordered peptides used in the study.

Methods

Systems

We study five systems as described in Table 1: (i) Tau_ff99SBws - Unmodified Tau_225–246 simulated with Amber ff99SBws force field (ii) Tau_C36m - Unmodified Tau_225–246 simulated with CHARMM36m force field (iii) Tau_2p - phosphorylated Tau_225–246, phosphorylated at residues T231 and S235 (iv) Tau_4p - phosphorylated Tau_225–246, phosphorylated at residues, T231, S235, S237 and S238 and (v) Tau_O-GlcNAc - O-GlcNAcylated Tau_225–246, O-GlcNAcylated at residues T231 and S235. All the phosphorylated and O-GlcNAcylated systems were studied using the CHARMM36m force field as described in detail in the simulation protocol section. Hereafter unmodified tau_225–246 simulated with Amber ff99SBws and CHARMM36m force fields will be referred to as tau_ff99SBws and tau_C36m respectively. Phosphorylated tau will be referred to as tau_2p (2 sites phosphorylated) and tau_4p (4 sites phosphorylated), and O-GlcNAcylated tau will be referred to as tau_O-GlcNAc. The choice of the systems was based on experimental work by Brister et al⁶. and Schwalbe et al.¹⁶ and the availability of NMR data for direct comparison with the simulation studies. It must also be noted that phosphorylation at residues T231 and S235 is observed in normal as well as AD brain, whereas S237 and S238 are found phosphorylated only in the AD brain.²⁰

Table 1:

Tau_225–246 peptide ensembles simulated.

System	PTM sites	Force Field	Water model	No of water molecules
Tauff99SBws	Unmodified	Amber ff99SBws	TIP4P/2005	5350
TauC36m	Unmodified	CHARMM36m	TIP3P	5212
Tau2p	T231, S235	CHARMM36m	TIP3P	5216
Tau4p	T231, S235, S237, S238	CHARMM36m	TIP3P	5213
TauO-GlcNAc	T231, S235	CHARMM36m	TIP3P	5159

Simulation protocol

For this study, we used Amber ff99SBws (an adaptation of Amber ff99SB to use TIP4P/2005 with scaled water interactions) and the CHARMM36m force field designed especially for simulating intrinsically disordered proteins. We perform our simulations for the unmodified peptide by employing both the force fields; Amber ff99SBws and CHARMM36m along with the TIP4P/2005⁴⁶ and TIP3P⁴⁷ water model respectively, whereas for phosphorylated and O-GlcNAcylated peptides only the CHARMM36m force field was used for the simulations. The CHARMM protein and carbohydrate force fields were used to account for the phosphate and carbohydrate PTMs for the simulation of modified peptides.^44,48 The starting geometry was set up by generating a random coil configuration of Tau_225–246 peptide. The dihedral values for the initial random coil are given in Table S1 of supplementary information file. The system is solvated in a truncated octahedron box with 6 nm spaced face and NaCl ions are added to neutralize the system charges. This was followed by 5000 steps of steepest descent to minimize the system followed by 100 ps of NVT and NPT equilibration at stochastic dynamics. Sixteen replicas are generated within a temperature range of 300 to 518.4 K. To adjust replica temperatures between 300 K and the highest temperature of interest geometric spacing is used to obtain uniform acceptance probability (~20%) between all adjacent replica pairs. All the replicas in each simulation are initiated from extended coil structures (Table S1). Each temperature replica was equilibrated for 200 ps to equilibrate the potential energy of the replicas. The potential energy was biased using 500 kJ/mol Gaussian widths, 1.2 kJ/mol initial Gaussian heights, with Gaussian potentials added every 2000 steps, with a bias factor of 36 to conduct simulations in the well-tempered ensemble (WTE). All metadynamics parameters are implemented in the plugin for free energy calculation, PLUMED v2.4.0.⁴⁹ The system dynamics were propagated using leapfrog stochastic dynamics with a 1ps friction coefficient. The replica exchange was attempted every 500 steps. A 12 Å cutoff was used for Lennard-Jones interactions.⁵⁰ Electrostatic interactions are calculated using the Particle-Mesh Ewald (PME) method with 1.0 nm real space cutoff.⁵¹ Each temperature replica was simulated for 200 ns NVT production simulations performed with the GROMACS⁵² version 2016.4 simulation package for a total simulation time of 3.2 μs, using a time step of 2 fs.

Only the 300 K temperature replica was used for analysis. The convergence of the simulations was checked by measuring the radius of gyration (Rg) and end-to-end (E2E) distance of the peptide. Based on the convergence analysis the last 100 ns of the production simulations were used for analyzing the results. Structural RMSD clustering analysis is performed based on the gromos method using a distance cutoff of 0.35 nm.

Results

For the present study, we simulate a 22 residue long fragment from the proline-rich domain of tau protein, Tau_225−246. The effect of phosphorylation on this fragment was also experimentally studied using NMR spectroscopy by Schwalbe et al.¹⁶ Here, we analyzed equilibrium ensembles of Tau_225−246 in its unmodified as well as post-translationally modified (phosphorylated and O-GlcNAcylated) forms.

Radius of Gyration and End-to-End Distances

To gain insight into the characteristics of the chain dimensions we calculate the distribution of the radius of gyration (R_g) for all systems as shown in Figure 2. For the unmodified peptides both the force fields predict similar behavior, albeit with significant differences. For Tau_C36m we observe a bimodal distribution with a peak around ~ 1 nm and ~ 1.5 nm respectively, while for Tau_ff99SBws we observe a prominent peak in the distribution around 1.35 nm which is intermediate to the peaks observed for Tau_C36m. For Tau_ff99SBws we also observe a slight shoulder around 1 nm. Upon examining the structures corresponding to these peaks we find that the lower peak corresponds to a loop conformation while the higher peak corresponds to an extended conformation. However, we note that the relative population distributions around these two peaks are reversed, with Amber ff99SBws favoring extended structures and CHARMM36m favoring the loop conformations. Upon phosphorylation at T231 and S235 (Tau_2p), we observe that the R_g peak at 1.5nm is diminished with the systems favoring a loop conformation. However upon additional phosphorylation at S237 and S238 (Tau_4p) we find a reappearance of the peak at 1.5 nm. The peak at 1 nm observed for the unmodified system is shifted to ~ 0.8 nm indicative of a further collapse in the geometry. Upon O-GlcNAcylation (Tau_O-GlcNAc) we observe a shift towards random coil structures with the distribution overlapping the Rg distribution of the unmodified peptide (Tau_C36m).

Figure 2.

Normalized radius of gyration (R_g) and end-to-end (E2E) distance distributions of Tau_ff99SBws, Tau_C36m, Tau_2p, Tau_4p, and Tau_O-GlcNAc. Peptide conformations corresponding to the peaks observed in the R_g distributions are also presented.

End-to-end (E2E) distance distributions, which provide insight into extended versus collapsed conformations, are also shown in Figure 2. The distributions follow a similar pattern as observed for R_g. For the simulation with the Amber ff99SBws force field, we observe a peak in the distribution around 3.5 nm, while for CHARMM36m we observe a peak around 1.5 nm followed by a broad distribution from 2nm - 6nm. Upon phosphorylation, we observe shifts towards collapsed conformations, with prominent peaks around 1.75 nm for both Tau_2p and Tau_4p. Upon O-GlcNAcylation we observe a broad distribution profile from 1nm – 6nm indicative of native-like random coil structures.

Secondary Structure Analysis

Gaining insights into the secondary structure preferences of the Tau_225–246 region are important to understand the binding of Tau with MT and the influence of PTMs. NMR, as well as circular dichroism (CD) studies, show that Tau_225–246 remains mostly disordered in solution. Some evidence of α-helix propensity has been observed in residues A239-R242 based on NMR chemical shift analysis. For this region, it has been observed that the fraction of α-helical content increases from 5% to 15% upon combined phosphorylation at T231 and S235 with additional phosphorylation at S237 and S238 resulting in a further increase in the helical content in the region. In contrast with PTM influenced changes in the downstream region the head region of Tau_225–246 peptide (²²⁵KVAVVR²³⁰) remains largely unaffected by phosphorylation. This observation is of interest as ²²⁵KVAVVR²³⁰ has been implicated in MT binding.

To gain insight into the secondary structure preferences of Tau_225–246 we performed secondary structure analysis to measure populations of common secondary structure elements using the DSSP algorithm.⁵³ The algorithm classifies the secondary structure of peptide based on hydrogen bonding interactions between backbone carbonyl and NH groups. Helices are characterized by (i, i+3), (i, i+4), and (i, i+5) hydrogen bond contacts as 3₁₀-helix, α-helix, and π-helix respectively. The average fraction of secondary structures for each system is presented in Figure 3. The coil state is the dominant conformation for every peptide ensemble, with population in the range of (61% −73%). This is followed by the bend conformations (14% - 24%). This observation is in line with the NMR and CD studies which have observed Tau_225–246 to remain predominantly disordered. Among the ordered structure, we find a slight preference for helical structures, α-helix (6%) and 3₁₀-helix (4%), exhibited by the Amber ff99SBws force field for the native peptide. While the CHARMM36m force field favors the β-sheet structure (6%) for the native peptide. This highlights slight differences arising due to the underlying force field. Upon PTM we observe slight shifts in the populations, but no drastic overall changes.

Figure 3.

DSSP based fraction of secondary structure for Tau_ff99SBws, Tau_C36m, Tau_2p, Tau_4p and Tau_OGlcNAc.

To further elaborate the picture, we also show per-residue secondary structure fractions in Figure S1 of the supporting information (SI) file. The residue-specific analysis allows us to identify regions that may adopt secondary structure elements. For the native peptide, we observe that the random coil state remains the most dominant conformation for both the Amber ff99SBws and CHARMM36m force fields, however, the tail regions of the peptide Tau_235–245 shows a preference for α-helical conformation in the Amber ff99SBws simulations. For the CHARMM36m simulations, we observe a slight preference for β-sheet geometries in the head (Tau_225–232) and tail (Tau_240–245) regions. This is also reflected in the representative structures presented in Figure 2 corresponding to the Rg peaks (b) and (c) for Tau_ff99SBws and Tau_C36m simulations respectively. Upon PTM we observe a reduction in the preference of β-sheet structure for Tau_2p and Tau_4p, and a reappearance of the preference for Tau_O-GlcNAc. These conformational preferences are consistent with the R_g analysis, with Tau_O-GlcNAc exhibiting a similar distribution to the unmodified system Tau_C36m. We also observe that Tau_4p exhibits a slight preference for α-helical conformation in the tail region Tau_236–245.

The availability of experimental NMR coupling constants and chemical shifts for the native wild peptide (Tau_wt) and the phosphorylated peptides, Tau_2P and Tau_4P allows us to analyze residue-specific structural features and compare the same with experimental observations.

¹J_CαHα coupling constants

Deviations in ¹J_CαHα coupling constants values from random coil values can provide useful insights into structural preferences. It has been found that residues in α-helical conformations exhibit positive deviations in the order of 4–5 Hz. The availability of experimental ¹J_CαHα coupling constants for Tau_225–246, Tau_2P, and Tau_4P allows us to calculate the deviations in the coupling constants (Δ¹J_CαHα) between the phosphorylated systems and the native peptide. The Δ¹J_CαHα values provide insight into the structural changes upon PTM. Significant Δ¹J_CαHα values available for the head region (Tau_225–231) and tail region (Tau_237–245) from the experimental study by Schwalbe et al.¹⁶ are also presented in Figures 4a and 4b respectively. ¹J_CαHα coupling constants can also be calculated from ϕ and ψ dihedral data obtained from MD simulations using Equation 1 proposed by Vuister et. al.⁵⁴

$${}^{1}J{C}_{\alpha }{H}_{\alpha }=A+B.\sin (\psi +138{)}^o+C.\cos 2(\psi {)}^o+D.cos2(\varphi +30{)}^0$$ (1)

Figure 4:

Residue specific differences (Δ¹J_CαHα) between ¹JC_αH_α couplings of Tau_2p/Tau_C36m, Tau_4p/Tau_C36m, and Tau_O-GlcNAc/Tau_C36m obtained from MD simulations. Δ¹J_CαHα values calculated from the experimental study by Schwalbe et al.¹⁶ for Tau_2p and Tau_4p are also presented for comparison. Δ¹J_CαHα values are presented for the (a) head region Tau_225–231 and (b) tail region Tau_237–245 of Tau_225–246.

Where, A = 140.3, B = 1.4, C = −4.1, D = 2.0

The Δ¹J_CαHα values evaluated from MD simulations are also presented in Figure 4.

The experimental Δ¹J_CαHα profiles reveal that upon phosphorylation at T231 and S235 (Tau_2P) the major changes with significant positive deviations (~1Hz) are observed for T231, S237 and K240 with minor positive deviations (0.5Hz) being observed for S238 and L243. These positive deviations are indicative of α-helical secondary structure preferences. The CHARMM C36m MD simulations capture the trend for the shift in the secondary structure preference at T231, S238, and L243, while we see no shift from the native conformations at S237 and K240. For Tau_4P, which is phosphorylated at T231, S235, S237, and S238 we observe shifts towards α-helical conformations at T231 and nearly all the tail residues from S238-Q244. This is also captured in the simulations with all the residues in the tail region from S238-Q244 exhibiting positive Δ¹J_CαHα deviation. The only exception to the trend being the large negative deviation observed for L243. In figure 4 we also present Δ¹J_CαHα values from the native peptide upon O-GlcNAcylation obtained from MD simulation. We do not observe any significant deviations in the head region of the peptide. In the tail region, we observe significant positive deviations at A239, L234, and T245. These residues do not exhibit any significant changes for phosphorylation, clearly indicating differences upon phosphorylation versus O-GlcNAcylation.

Chemical shift analysis

From all the MD simulations we also calculated the C_α, C_β, H_α secondary chemical shift values using the SPARTA+⁵⁵ program. In Figure 5 we present the Δδ¹³Cα−Δδ¹³Cβ chemical shift data, which allows us to comment on regions of contiguous secondary structure. The Δδ¹³Cα and Δδ¹³Cβ values are calculated relative to random coil values extracted from Tamiola et al.⁵⁶ Positive deviations from random coil values are indicative of α-helical propensity while negative values are indicative of β-sheets. From the experimental data we observe that there is a tendency for the head region of the peptide to adopt a β-sheet structure (Tau_225–230), while the tail region adopts the α-helical conformation especially upon phosphorylation (Tau_236–242). We also evaluate the secondary structure propensity score (SSP) using the chemical shift data. The same is compared with the experimental data and plotted in Figure 5. On comparing the Tau_C36m, Tau_2P and Tau_4P SSP scores we observe a shift towards α-helical preferences in the tail region, however, the absolute values are lower in comparison with the experimental data. Both the experimental and simulation data reveal that PTMs do not affect the ²²⁵KVAVVR²³⁰ region of the Tau peptide, which has been implicated in MT binding.

Figure 5:

Left Panel: Chemical shift calculated from simulated ensembles for unmodified Tau (ff99SBws and CHARMM36m) and Tau_2p compared with experimental data (from Schwalbe et al.¹⁶). Right Panel: Secondary structure propensity (SSP) score for unmodified Tau (ff99SBws and CHARMM36m) and Tau_2p and Tau_4p calculated using the SSP program (Marsh et al.⁵⁷) compared with the SSP score obtained from experimental data (from Schwalbe et al.¹⁶).

NOE Analysis

Nuclear Overhauser Effect (NOE) provides distance information from NMR experiments. NMR studies report an increase in the medium-range NOEs especially around residues near the phosphorylation sites. In Figure 6 we present the mean proton-proton distances between the various amino acids from MD simulations. The proton-proton distances are calculated as an average d_HH = 〈r⁻⁶〉^−1/6. The distances are classified as strong (0 Å < d_HH < 2.7 Å), medium (2.7 Å < d_HH < 3.5 Å) and weak (3.5 Å < d_HH < 5.0 Å). Only the shortest d_HH distance between the sets of amino acids is used to classify the interactions presented in the figure. From the Tau_ff99SBws simulations for the native peptide, we observe d_HH i, i+2 and i, i+3 interactions between the amino acids in the tail region of the Tau peptide favoring an α-helical conformation. This pattern is not observed in the Tau_C36m simulations of the native peptide. For the simulations with the CHARMM C36m force fields, we observe weak interactions between the head and the tail regions of the peptide, indicated by the cross-peaks. These interactions are intensified upon phosphorylation and correspond to a bent/hairpin conformation for the peptide. Upon O-GlcNAcylation we observe an interaction pattern similar to the native peptide. As observed in the NMR experiments the interactions between the amino acids far removed from each other are pronounced for the residues near the phosphorylation sites. The NOE analysis reveals that there is are strong interactions between the head and the tail regions of Tau_225–246 upon both phosphorylation and O-GlcNAcylation. However, the interaction pattern differs in the two PTMs. To gain further insight into these effects we investigate the system for residue-specific interactions, especially the formation of salt bridges and H-bonding interactions.

Figure 6:

Mean proton-proton distances between the various amino acids from MD simulations. The proton-proton distances are calculated as an average d_HH = 〈r⁻⁶〉^−1/6. The distances are classified as strong (0 Å < d_HH < 2.7 Å), medium (2.7 Å < d_HH < 3.5 Å) and weak (3.5 Å < d_HH < 5.0 Å).

Salt bridges

Electrostatic interactions are known to play a crucial role in protein stability. Interactions between the charged phosphate group of phosphoserine/phosphothreonine and neighboring basic amino acids (Arginine (R), Lysine (K) and Histidine (H)) residues can enhance the stability of the protein. An example of such stabilization was reported by Errington et. al.⁵⁸ who observed enhanced stabilization of α-helices by inducing a secondary salt-bridge sidechain interaction between phosphoserine and lysine placed in a i, i+4 arrangement. The tau peptide segment used in this study contains five basic epitopes (K225, R230, K234, K240, and R242) within the peptide. Site-directed mutagenesis study by Goode et al.¹⁹ indicated that K224, K225, and R230 are important for MT-binding and assembly. We investigate the ability of these basic residues to form salt bridges in Tau_2p and Tau_4p.

In Figure 7(a) and 7(b), we present the probability distribution of the possible salt-bridge interactions formed in both Tau_2p and Tau_4p respectively. For Tau_2p we observe that both pT231 and pS235 interact with R230, K234, K240, and R242, forming stable salt-bridges. pS235 exhibits a preference for forming a very stable salt bridge with R242, which is present in the tail region of the peptide sequence. The formation of these stable salt bridges explains the off-diagonal interactions observed in the NOE NMR experiments. In Figure S2 of the SI file, we present the simulation time-series of the possible salt bridge interactions. We observe a concomitant formation of all the observed salt bridges.

Figure 7:

Normalized probability distribution of (a) salt-bridge interaction for phosphorylated T231 and S235 (Tau_2p) and (b) T231, S235, S237 and S238 (Tau_4p) with basic residues: K225, R230, K234, K240 and R242. (c) and (d) representative structures of the salt-bridges for Tau_2p and Tau_4p respectively, only representative portion of the peptide have been shown for clarity.

Contrary to Tau_2p for Tau_4p, we observe the formation of selective salt bridges. pT231 and pS235 selectively interact with R230 and K234, respectively. While both pS237 and pS238 interact favorably with R242. In addition to the strong salt bridge with R242, we also observe interactions between pS237 and R230, K234 and K240, while pS238 interacts with R230 and K240. In Figure S2 of the SI, we present the simulation time series for these salt bridge interactions. We observe differences between the interaction patterns between Tau_2p for Tau_4p. The salt bridges in Tau_4p appear to be localized in comparison to Tau_2p. In Figure 7 (c) and (d) we present the representative structures of the salt-bridges. We observe that the unique turn/bend conformation of the Tau_225–246 peptide is brought about by the formation of the phosphate-lysine/arginine salt bridges. This highlights the structural influence of phosphorylation on Tau.

H-bonding

In the Tau_C36m and Tau_O-GlcNAc simulations we observe β-strand conformations in the head (²²⁶VAVVR²³⁰) and tail regions (²⁴¹SRLQT²⁴⁵) of the Tau_225–246 peptide segment. These β-strands arrange themselves into an anti-parallel β-sheet conformation via the formation of H-bonds between the backbone amide atoms of following pairs of amino acid pairs, A227-Q244, A227-L243, V229-S241, and R230-R242. The H-bond distributions corresponding to these pairs of H-bonds are presented in Figure 8a. For both the native (Tau_C36m) and the O-GlcNAcylated (Tau_O-GlcNAc) peptides the most significant H-bond pair is observed between R230 and R242. We note that both these arginine residues form salt bridges with the phosphate group in Tau_2p and Tau_4p. Thus, it is observed that the formation of the bend/loop conformation in the native and O-GlcNAcylated peptides is controlled by the formation of an anti-parallel β-sheet while the same in Tau_2p and Tau_4p is controlled by the formation of salt bridges. The involvement of R230 and R242 in the formation of salt bridges in Tau_2p and Tau_4p disrupts the formation of β-strands in these regions. We did not observe the formation of β-strands in Tau_ff99SBws simulations highlighting subtle differences between the Amber ff99SBws⁴³ and CHARMM36m⁴⁴ force fields. The observance of β-sheet structures especially in the N-terminal ²²⁵KVAVVRT²³¹ motif is in line with the NMR studies which report an 18% probability of this motif adopting a β-sheet structure.¹⁷ In addition to the H-bonds involved in the formation of the anti-parallel β-sheet, we also observe the formation of a significant H-bond between the free hydroxyl (-CH₂OH) group at C6 in GlcNAc with the backbone carbonyl oxygen of the adjacent amino acid (Figure 8b). O-GlcNAc at T231 interacts with the carbonyl oxygen of both P232 and R230, while the O-GlcNAc at S235 interacts only with the carbonyl oxygen of K234. These H-bonds control the orientation of the carbohydrate (GlcNAc) relative to the peptide backbone. In Figure S3 of the SI, we present the simulation time series corresponding to these H-bonding interactions. Moreover for Tau_4p we also found presence of transient helix in the tail regions of the peptide (235–240). The presence of the transient helix is a direct result of formation of backbone carbonyl-amide i-i+4 hydrogen-bonds between S235 (O)…A239 (HN), S237 (O)…S241 (HN) and S238 (O)…R242 (HN) (Figure S4 of SI).

Figure 8:

(a) Main-chain carbonyl and amide hydrogen bond contact responsible for antiparallel β-sheet formation. (b) H-bond between the free hydroxyl (-CH₂OH) group at C6 in GlcNAc with the backbone carbonyl oxygen of the adjacent amino acid. Representative structures have been presented to illustrate the H-bonding interactions.

Discussion

In this study, we have explored the conformational space of the Tau_225–246 segment belonging to the proline-rich domain (P2) of the tau protein and its post-translationally modified forms using enhanced conformational sampling WTE metadynamics simulations. The proline-rich domain of tau serves as a linker between the N-terminal sequence and the tubulin-binding domains.⁶ Structurally the proline-rich domain has largely been characterized as being in a random coil state. Our simulations show that although the coiled state remains the most dominant conformation for the native peptide, we do observe a force field dependent influence on the short segments of the peptide. We observed a preference for α-helical conformation for the tail regions Tau_235–245 in the Amber ff99SBws simulations, while for the CHARMM36m force field simulation we observed a preference for β-strand formation for both the head (Tau_225–232) and tail (Tau_240–245) regions of the peptide. The analysis of the radius of gyration (Rg) and end-to-end distances (Ree) reveal a preference for extended conformations for the native and O-GlcNAcylated peptides while phosphorylation induces collapsed bent conformations.

The secondary structure analysis revealed that PTMs (phosphorylation or O-GlcNAcylation) do not influence the conformation of the ²²⁵KVAVVR²³⁰ region, with no appreciable change in chemical shifts being observed in the simulations which agrees with the NMR derived chemical shifts data. However, closer inspection of the atomic level interactions reveals crucial binding interactions involving R230 in both phosphorylation and O-GlcNAcylation. Upon phosphorylation, R230 forms a stable salt bridge interaction with pT231 and pS235 in Tau_2p, while in Tau_4p the interaction is more selective to pT231. We note that this salt bridge does not influence the conformation of the preceding ²²⁵KVAVV²²⁹ region. Both these observations are in agreement with the NMR experiments by Schwalbe et. al. which alluded to the importance of the R230 forming a stable salt bridge with phosphorylated T231 in controlling the binding of Tau_225–246 segment to MTs.¹⁶ However, upon O-GlcNAcylation R230 forms a intramolecular H-bond with R242 which results in the formation of an anti-parallel β-sheet between ²²⁶VAVVR²³⁰ and ²⁴¹SRLQT²⁴⁵. Experimental studies on native Tau have found that ²²⁵KVAVVRT²³¹ and ²⁴³LQTA²⁴⁶ show a strong affinity to bind to MTs. This observation has provided credence to the “Jaws” model of MT binding, wherein the flanking proline-rich regions in Tau are also involved in binding with MT in addition to the MT binding domain.

Thus we observe that our results indicate the unique “on-off” behavior of R230 and R242 in controlling the conformation of these short peptide sequences. In the native peptide as well as upon O-GlcNAcylation R230 and R242 are responsible for the stabilization of the antiparallel β-sheet structure, however, upon phosphorylation (Tau_2p), both these Arginines are involved in salt bridges with pT231 and pS235 that results in a conformational change. Upon additional phosphorylation (Tau_4p) at S237 and S238 we observe a change in the interaction pattern with the formation of selective salt-bridges. We note that while T231 and S235 phosphorylation sites have been established as a valid biochemical marker for AD diagnosis, the same is not true for S237 and S238. It could be that only selective sites are targeted for phosphorylation. One of the unique features of both T231 and S235 is the presence of a preceding basic amino acid R230 and K234 and being followed by a proline P232 and P236, ie ²³⁰RTP²³² and ²³⁴KSP²³⁶. The R230-T231 bond is also targeted by calpain and caspase family of proteases which results in 20–22 kDa Tau fragments which have been observed in the patients with AD, highlighting the importance of this region of the Tau peptide and T231. ^59,60

Our detailed investigation into the influence of phosphorylation on the conformational preferences of Tau could also provide insights for the development of phospho-specific antibodies. Shih et. al. reported the development of one of the first phospho-specific antibody which could detect the pathological form of Tau in patients with AD. ⁶¹ The study specifically targeted the pT231 and pS235 phosphorylation sites. To gain insight into the specificity they also reported the crystal structure of a Fab fragment co-crystallized with ²²⁴KKVAVVR(pT²³¹)PPK²³⁴ at a resolution of 1.9 Å. In the crystal structure, six residues ²²⁴KKVAVVR(pT²³¹) were found to directly interact with the Fab fragment, with the remainder of the structure adopting a random coil conformation. A significant salt bridge was observed between the Arginine side-chain of Fab (R53) with the pT231 phosphate. Additionally, a water-mediated interaction was observed between pT231 and the side chain of R230. These observations highlight the importance of Phosphate-Lysine interactions in antibody recognition.

Conclusions

In conclusion, we find that both phosphorylation and O-GlcNAcylation influence the secondary structure of the Tau peptide segment. However, different effects govern the conformational preferences. The conformational transition upon phosphorylation is a direct influence of the formation of strong salt bridges with nearby lysine and arginine. For O-GlcNAcylation we did not find a direct influence on the conformational preference, however, the formation of strong H-bonds between the free –CH₂OH group and backbone amide carbonyl’s does not rule out the perturbative influence of O-GlcNAcylation. It is evident the phosphorylation induces a strong structural transition with Tau_225–246 favoring a bent conformation due to the formation of strong salt-bridges. These results are also in agreement with the NMR studies by Brister et. al. who observed opposing structural and functional effects for various short segments of the Tau protein.⁶ In terms of composition, the Tau peptide contains 3% Arginine sites and 10% Lysine sites, which can interact with phosphorylated threonine/serine. Additional simulation studies on longer constructs of the Tau peptide could provide significant information on the conformational preferences of native Tau and Tau fragments observed in Tauopathies.