Rosmarinic acid prevents fibrillization and diminishes vibrational modes associated to β sheet in tau protein linked to Alzheimer’s disease

Abstract Alzheimer’s disease is a common tauopathy where fibril formation and aggregates are the hallmark of the disease. Efforts targeting amyloid-β plaques have succeeded to remove plaques but failed in clinical trials to improve cognition; thus, the current therapeutic strategy is at preventing tau aggregation. Here, we demonstrated that four phenolic diterpenoids and rosmarinic acid inhibit fibrillization. Since, rosmarinic acid was the most active compound, we observe morphological changes in atomic force microscopy images after treatment. Hence, rosmarinic acid leads to a decrease in amide regions I and III, indicating that rosmarinic acid prevents β-sheet assembly. Molecular docking study inside the steric zipper model of the hexapeptide 306VQIVYK311 involved in fibrillization and β sheet formation, suggests that rosmarinic acid binds to the steric zipper with similar chemical interactions with respect to those observed for orange G, a known pharmacofore for amyloid.


Introduction
Alzheimer's disease (AD), the most common form of dementia, is characterized by the deposition of two main protein aggregates in the brain: senile plaques consisting of amyloid-b (Ab) and neurofibrillary tangles (NFTs) composed mainly of the microtubule-associated protein tau and found as intraneuronal inclusion 1 . Tau protein also plays an important role in other tauopathies, such as progressive nuclear palsy (PSP), corticobasal degeneration (CBD), and Pick's disease. In these tauopathies, tau is located in neurons and glia 2,3 . Physiologically, tau regulates microtubule stability and axonal transport 4 , but in patients with AD it becomes increasingly phosphorylated. Hyperphosphorylated tau no longer binds to microtubules and accumulates in the soma and dendrites of neurons, forming intracellular deposits called paired helical filaments (PHFs) that are the main fibril component of NFTs 5,6 . Tau pathology occurs concomitantly with axonal transport deficits and neuronal degeneration 7 .
On the other hand, human tau is a natively unfolded protein whose structure has two fibril-forming motifs, 275 VQIINK 280 and 306 VQIVYK 311 . These motifs are within the microtubule-binding domain (4R) of tau (aa 244-373), and are frequently used to resemble tau aggregation in vitro 8,9 . Crystallization of the fibrilforming motifs shows that they are both necessary and sufficient for tau fibrillization. In particular, the glutamine residue in the 306 VQIVYK 311 hexapeptide is essential for tau fibrillization 8,10 . The hexapeptide 275 VQIINK 280 is of equal importance for tau fibril formation; unfortunately no crystal structure is available 11 . The protein secondary structure with the highest tendency to form fibrils are b sheets that can associate adjacent strands running either parallel or antiparallel 8 . The 306 VQIVYK 311 hexapeptide forms parallel b sheets that are linked through interdigitating non-polar side chains 12 . The correlation of pathological phosphorylation of tau triggers conformational changes that make it prone to form dimers or higher order oligomers 13,14 . Interestingly, improvement in the cognition of a transgenic mouse model displaying both NFTs and senile plaques required reduction of soluble tau but not of Ab plaques 15 . Over-expression of tau in neurons has deleterious effects, suggesting that intracellular tau aggregates are cytotoxic 16 .
The major focus of past drug developments for AD treatment has rested on Ab, although a significant number of candidate drugs have failed to slow down disease progression. For example, immunization against Ab was effective in reducing amyloid plaque load, but had little effect on cognitive functions in patients 17,18 . Moreover, a recent study shows that gamma-secretase inhibition failed in a clinical phase III trial 19 . Thus, it seems timely to consider alternate drug discovery strategies for AD, based on approaches such as misfolded tau reducing agent 20 . The development of small-molecules that inhibit aggregation of tau appears as a valid alternative strategy for developing new treatments for AD and other tauopathies 21 . It has been proposed that naturally occurring phytochemicals have the potential to prevent AD based on their neuroprotective, anti-amyloidogenic, anti-oxidative and antiinflammatory properties 22 . For instance, natural polyphenols have reported anti-aggregating capacity to prevent amyloid formation, [23][24][25] and a standardized turmeric extract reduced both amyloid-b and phosphorylated tau levels in a transgenic mouse model 26 . Likewise, novel compounds have been found with the ability to prevent tau aggregation, such as cinnamon extract, and fulvic acid among others 27,28 .
Rosmarinus officinalis L. belonging to Lamiaceae family has been used in medicine as anti-inflammatory and antimicrobial agent 29,30 . Since drugs targeting Ab have recently failed in clinical trials 31,32 , it seems timely to search for other naturally occurring polyphenols with relevant activities, such as tau inhibitors. Considering that tau pathology involves not only AD but several tauopathies, we investigated the effect of four phenolic diterpenes and one caffeoyl derivative, isolated from Rosmarinus officinalis L., over heparin-induced tau aggregation in vitro by using thioflavin T assay (ThT), atomic force microscopy (AFM), and Raman spectroscopy. In addition, we provide structure-based models of complex between the fibril-forming motif 306 VQIVYK 311 of tau and caffeoyl derivative to explain the mechanism of its tau anti-aggregation capacity.

Chemistry
Silica gel (Kieselgel 60, Merck 0.063-0.200 mm) and Sephadex LH-20 were used in column chromatography (CC). Technical solvents used in extraction were previously distilled and dried according to standard procedures. The extraction and purification of 1,2,3,4, and 5 were carried out essentially as described [33][34][35][36] . The purity of all compounds was estimated by using a HPLC-PDA (! 98%) (Agilent LC-1200). Tau 4R (four microtubule binding domain) recombinant protein was expressed in E coli. The expressed protein was purified by HPLC using a ProPac IMAC-10 Column. Aggregation and fluorescence assays were performed in a Biotek H1 monochromator-based multi-mode microplate reader. Samples were prepared on highly ordered pirolytic graphite (HOPG) as substrate, and atomic force images were obtained using a Nanoscope III equipment in tapping mode. Raman spectra and maps were obtained through a Confocal Raman Microscopy Alpha 300 (WITec GmbH) .

General experimental procedures of extraction and isolation
Rosmarinus officinalis L. was collected in Talca, VII Regi on, Chile. Samples from aerial parts were dried at room temperature in darkness. R. officinalis L. was identified by Prof. O. Garcia, and a voucher specimen (N RO-2015/1215) is kept at the Herbarium of Extreme Natural Product Laboratory of University of Chile.

Protein expression and purification of tau 4R
The fragment 4R (htau 244-372 ) was cloned into pET-28a vector (Novagen) to produce a His-tagged protein which was expressed in Escherichia coli strain BL21 (DE3) as described 28 . LB medium containing kanamycin (30 lg/ml) was inoculated with a stationary overnight culture. The culture was grown at 37 C to OD 600 of 0.5-0.6, and protein expression was induced by addition of 1 mM isopropyl b-D-1-thiogalactopyranoside (IPTG) for 4 h. The cells were pelleted and then resuspended in 50 ml of buffer A (50 mM sodium phosphate buffer pH 7.4, 500 mM NaCl, and 20 mM imidazole) supplemented with EDTA free anti protease tablet (Roche). The mixture was sonicated for 5 min, centrifuged at 10000 g for 30 min and then filtered (0.22 lm membrane filter, Millipore). The sample was loaded onto ProPac IMAC-10 Column (4 Â 250 mm Thermo dionex) with buffer A and buffer B (50 mM sodium phosphate buffer pH 7.4, 500 mM NaCl, 200 mM imidazole).
After loading, one-step wash with 30 mmol Imidazole (7.7% B) was done before elution at 200 mmol imidazole. Protein purity was verified on Coomassie Brilliant Blue-stained SDS-polyacrylamide gel. The concentration of purified 4R was determined using the extinction coefficient at 280 nm (1520 M À1 cm À1 ). The concentrated protein was stored at -80 C until use.

Tau fibrillization in vitro
The aggregation of 4R was induced with 40 lM of protein plus 10 lM of heparin in 100 mM of sodium acetate pH 7.0 for 24 h incubation time, with continuous shaking. Standard solution of 5 at 10 mM to 1 mM (DMSO) was prepared fresh prior to aggregation experiments. We used a black flat bottom 96 well plate (ThermoFisher scientific).

Atomic force microscopy images of tau protein
Each sample was prepared by pipetting 30 lL (40 lM of tau plus 10 lM of heparin) in the presence or absence of 5 (10 lM and 1 lM respectively) onto highly orientated pyrolytic graphite (HOPG). The samples were incubated at room temperature for 30 min and then carefully washed. Atomic force microscopy imaging was performed in tapping mode using a Nanoscope III (Veeco, CA) and Olympus AC240TS cantilevers exhibiting spring constants of 2 N/m at resonance frequencies of 55 kHz. To achieve minimal imaging forces between AFM stylus and sample, the drive amplitude was set between 0.5 and 1.0 V, and the amplitude set point was adjusted manually to compensate for the thermal drift in order to maintain good tracking of the sample.

ThT binding fluorescence assay of tau protein
The ThT fluorescence assay adopted here was modified from the reported by Pickhardt et al. 37 and Crowe et al. 38 . Briefly, to examine the inhibition of tau aggregation, the total volume of the reaction mixture was 100 lL, which included 40 lM of protein, 10 lM heparin in 100 mM sodium acetate, pH 7.0 plus 5 at different concentrations. Notice that for inhibition all components are included in the solution at the same time. After 24 h incubation time at 37 C, addition of 100 lL at a 50 lM solution of ThT was made and incubation continued for 1 h at room temperature prior to fluorescence reading. Then, the fluorescence was measured in a BioTek Synergy H1 multimode-multiple reader with an excitation wavelength at 440 nm and emission wavelength of 485 nm in a 96-well plate.
Raman spectra and maps of tau b sheets content The Raman spectra of 4R were recorded as described with modifications by using an Confocal Raman Microscope Alpha 300 (WITec GmbH) equipped with a solid state Compass Sapphire laser (Coherent Inc.) operating at 532 nm and a Back Illuminated CCD camera DV401A (Andor Technology Ltd.) cooled at À60 C 39 . A single mode 3.5/125 lm optical fiber was used (OZ Optics Ltd.) to connect the laser to a Zeiss (100X/0.9 NA) objective. The laser power was set between 10 to 20 mW. In order to generate scan images, we set 20 points per line over a surface of 5 lm (width) and 5 lm (height). Scan speed was set to 100 s/line. Integration time was set to 5 s. Tau protein sample (40 lM plus 10 lM of heparin) over glass (microscope slides) was dried and the spectra collected on several spots over the sample. The sample spectra were standardized from glass (spectra subtracted) 39 .

Molecular docking of 306 VQIVYK 311 fragment of tau and 5
Docking was performed by using the Glide method 40,41 . A grid box of 15 Å Â 10 Å Â 10 Å covered the whole cavity in the model. Docking parameters were used as in previous work 39 , and Glide extra-precision (XP) modes were explored during the search. The docking hierarchy begins with the systematic conformational expansion of the ligand followed by placement in the receptor site. Then minimization of the ligands (R and S enantiomers) in    the receptor field were carried out using the OPLS-AA force field with a distance-dependent dielectric of 2.0 42 . Afterwards, the lowest energy poses were subjected to a Monte Carlo procedure that samples the nearby torsional minima. The best pose for a given ligand was determined by the Emodel score, while different compounds were ranked using GlideScore 43 . The docking poses were analyzed by examining their relative total energy score. The more energetically favorable conformations were selected as best poses.

Statistical analysis
Statistical analysis was developed using Graphpad 6. Data were analyzed as Mean ± SE using Paired t-test and Anova. Significance was determined as p < 0.05.

Result and discussion
Phenolic diterpenes and caffeoyl derivative inhibit fibrillization of tau monitored by ThT fluorescence assay Interestingly, some phenolic diterpenoids such as carnosic acid has been reported as a potent activator of a nuclear factor erythroid 2 related transcription factor (Nrf-2), a major regulator of redox homeostasis in mammals, whose activity relies on a cathecol moiety 44,45. Also phenolic diterpenoid carnosic acid has shown effects on Ab peptides production in a cells model 46,47 . Another phenolic diterpenoid, carnosol, has been reported as an anti-inflammatory and anti carcinogenic agent, and blocked inactivation of muscarinic acetylcholine receptor [48][49][50] .
Moreover, rosmarinic acid isolated from the spice sage (Salvia officinalis) protects PC12 cells from neurotoxicity induced by amyloid-b 51 and has protective effects on transgenic mice carrying a double mutation linked to an amyloid precursor protein 52 . However, until now neither phenolic diterpenoids, nor 5 have been tested against tau aggregation. Thus, we isolated 1, 2, 3, 4, and 5 from Rosmarinus officinalis L., in order to demonstrate whether they may act as tau inhibitors by avoiding fibril formation in vitro and subsequent b sheet formation. The chemical structure of phenolic diterpenoids is shown in Figure 1. The 1 H-NMR spectrum of 1, 2, 3, 4, and 5 is consistent with the literature [32][33][34][35] . Tau protein largely exists as an intrinsically disordered protein 45 , however as long as motifs 275 VQIINK 280 and 306 VQIVYK 311 interact, tau is able to form paired helical filaments with cross b-sheets 8 . In addition, the benzothiazole dye, ThT, has been used as a fluorescence marker for amyloid fibrils; 53 giving a reliable signal to determinate fibril content in vitro. Therefore, we address the question of whether both phenolic diterpenoids and 5 are able to stop aggregation of tau by carrying out ThT fluorescence assay as an indicator of fluorescence intensity associated to aggregation. Thus, we demonstrate that all compounds are able to inhibit tau aggregation at 50 lM (Figure 2(A)). Considering the inhibition percentage, all naturally occurring compounds inhibit over 50%, and the most active compound was 5 (Table 1). Since 5 was the most effective compound, supplemental assays are performed in order to provide understanding of how 5 interacts with tau. First, we induced tau fibrillization and then we added 5 at different concentrations, after incubation we measured fluorescence intensity. Thus, we found that 5 inhibits tau in concentrations ranging from 10 lM to 100 lM, in dose-response manner (Figure 3(A)). In addition, we calculated IC 50 at 7.7 lM of 5 based on a nonlinear regression method (Figure 3(B)).
It is important to note that nitrocatechol derivatives such as tolcapone and entacapone exhibit anti-aggregation properties against both Ab and a-synuclein and tau hexpeptide [54][55] .
Considering that 5 was the most active compound and it has two cathecol moieties, we hypothesize that differences observed can be evidence of a main role of cathecol moiety.

Compound 5 diminishes fibril formation and reduces band assignment modes associated to b-sheets
The hallmark of Alzheimer s disease and tauopathies is the formation of paired helical filaments and subsequent fusion into neurofibrillary tangles (NFTs) whose core is mainly composed of b-sheets structure 56 . Moreover, kinetic studies of fibril formation by using ultraviolet resonance Raman (UVRR) spectroscopy have revealed different stages in fibrillation where at early stages it is possible to find fibrillary aggregates with b sheet and disordered content. As the process advances, the b sheet content also increases 57 . Therefore, there is a correlation between Raman spectroscopy, band assignment, and secondary structure 58 . The sensitive Amide I region is associated with C¼O stretch and is able to form a hydrogen bond with an N-H group of either inter-or intrabackbone chain giving a good signature of a b-sheet between wavenumbers 1674-1670 cm À1 . Moreover, Amide III region is associated to C-N stretching and N-H bending giving a b-sheet signature in the range of wavenumber 1242-1227 cm À1 . Hence, Raman spectroscopy provides a useful method to monitor b sheet formation 59 . Considering that 5 was the most active naturally occurring compound that we isolated from R. officinalis, we elucidate whether there are both morphological and signature changes in secondary structure once 5 interacts with tau protein. With this in mind, we induce tau aggregation by adding heparin, which promotes paired helical fibril formation. Atomic force microscopy images (Figure 4) show several twisted fibrils and oligomers at two distinct magnifications. Aggregation control experiments are presented in Figure 4(A) and in Figure 4(D) with higher magnification. Interestingly, once we challenge tau aggregation with 5 at 1 lM, a few straight-thin filaments and several oligomers are seen on the HOPG surface (Figure 4(B,E)). However, at 10 lM of 5 most fibrils and oligomers are diminished indicating that 5 is able to prevent fibril progression (Figure 4(C,F)). Raman's spectra of tau aggregation (24 h) and Raman vibrational bands are both shown in Figure 5(A), and the vibrational modes are listed in Table 2. Figure 5(A) shows band assignments associated to Amide I and Amide III that are linked to b sheets. In addition, we found band assignment associated to C-H deformation at 1451 cm À1 ( Figure  5(A)). However, the spectra of tau aggregation in the presence of 5 (10 lmol) ( Figure 5(B)), clearly shows a band assignment reduced in both Amide I and III region, respectively, and a signal corresponding to disordered structure at 1260 cm À1 38,57-60 is also detected, indicating that 5 prevents fibrillization and diminished b-sheet content. Considering that tau is a natively unfolded protein, 55 this suggests that some tau protein at least remains in the disordered state ( Figure 5(B)), differently from tau in absence of 5, where we found a massive peak at 1670 cm À1 indicative of b sheet antiparallel structure. Interestingly, inhibitors with a nitrocathecol moiety are able to prevent aggregation and b-sheets assembly as demonstrated by circular dichroism 55 .
Raman maps based on the integral of different Raman bands give the relative intensities of defined peaks in the spectral region. Thus, we analyzed peaks associated to b sheets (Amide I and III), and C-H deformation. As we compared aggregation control, with samples treated with 5, there are differences between peak intensities, including Amide I (1670 cm À1 ) ( Figure 6(B,D)), and Amide III (1240 cm À1 ) ( Figure 6(C,E), respectively), indicating that relative intensities associated to b sheets are also diminished. However, once we compare C-H deformation (1451 cm À1 ) in both controls and treated sample, significant differences were not found ( Figure  6(A,D)), indicating that sample amounts remain similar in both (control and treated) once we scan over the glass surface to determinate Raman spectra.

Molecular modeling of 5 binding to tau's 306 VQIVYK 311 hexapeptide
Tau is considered a natural unstructured protein 16 , although there is limited information on its exact structure. However, the  Orange-G is bound between steric zippers of VQIVYK with its long axis parallel to the fiber axis and forms salt links between its sulfonic acid groups and lysine ammonium ions from repeating strands in VQIVYK fibers. In addition, the aromatic rings of orange-G are packed against apolar side chains of Val309, and establish polar interactions with glutamine and lysine side-chains at the edges of the steric zipper. Recent works have used the information of the VQIVYK-orange-G crystals to propose interactions between ligands that prevent tau aggregation and the hexapeptide VQIVYK using docking method 55,61 .
Compound 5 has an elongated shape, aromatic and polar groups, and a negatively charged group, which match with the chemical features orange-G. With this in mind, we propose that rosmarinic acid inhibits tau aggregation by establishing chemical interactions with the fibril-forming hexapeptide VQIVYK.
To obtain a better understanding of a possible interaction of 5 with the 306 VQIVYK 311 motif in 4R oligomers, we model 5 binding (R and S enantiomers were considered) to b-sheets of poly-306 VQIVYK 311 hexamers. X-ray crystallography revealed that fiber-like complexes of 306 VQIVYK 311 consist of pairs of b-sheets, allowing small molecules to bind between sheets 9 . In particular, negatively charged molecules bound specifically to 306 VQIVYK 311 lysine side chains of adjacent sheets. Rosmarinic acid docking was done inside the cylindrical cavity formed by paired 306 VQIVYK 311 b-sheets (Figures 7(A,B)).
Our docking models depict that both 5 enantiomers are placed inside the cavity without any steric clashes (Figure 7(A,B)). According to our results, 5 and 306 VQIVYK 311 , chemical interactions are comparable to those observed for orange G 9 . Both rosmarinic acid enantiomers have the aromatic rings packed against apolar side chains of Val309 and form a hydrogen bond (HB) network with glutamine and lysine side chain groups at both sides of the steric zipper. The most important interaction is the salt link between the carboxylate group of rosmarinic acid and two Lys311 ammonium ions from 306 VQIVYK 311 fibers.
Taken together, our docking models predict a possible interaction of 5 with 306 VQIVYK 311 b-sheets with chemical interactions comparable to the ones reported for orange G. Therefore, we suggest that anti-aggregation effects of 5 are due to interactions with 306 VQIVYK 311 , in concordance with other reports that suggested the same mechanism for other known drugs.

Conclusions
In the present study, we isolated five compounds 1-5 from R. officinalis L. All of them demonstrated to inhibit tau aggregation over 50% at 50 lM. Moreover, we found that 5 was the most active compound. In addition, we found that both morphology and content of b sheet is affected once we challenged tau with 5. Docking experiment revealed that 5 interact with 306 VQIVYK 311 mainly through salt bridges. Since there is only palliative treatment against Alzheimer's disease and tauopathies, we consider important to search for promising naturally compounds as potential drug pharmacophore with catechol moities.

Funding
This work was supported from INACH RT 13-13 and Fondecyt Regular No 1150745 to Alberto Cornejo and Carlos Areche. Francisco Melo, Felipe Aguilar and Leonardo Caballero kindly acknowledge Anillo ACT1412 Conicyt-Chile. George Perry acknowledges The Semmes Foundation and a grant from the National Institute on Minority Health and Health Disparities (G12MD00759) from the National Institutes of Health.