Functionalization with TAT-Peptide Enhances Blood-Brain Barrier Crossing In vitro of Nanoliposomes Carrying a Curcumin-Derivative to Bind Amyloid-Β Peptide

Alzheimer’s Disease (AD) is the most common neurodegenerative disease in the elderly population [1-3]. The mechanisms underlying AD are not yet completely clear but genetic, pathological and biochemical clues suggest that the progressive production and subsequent accumulation of the amyloid-β (Aβ) peptides play a central role. Under abnormal conditions, the accumulation of Aβ progressively forms oligomeric, multimeric and fibrillar aggregates and culminates with the formation of extracellular plaques, one of the morphological hallmarks of the disease [4-6].


Introduction
Alzheimer's Disease (AD) is the most common neurodegenerative disease in the elderly population [1][2][3]. The mechanisms underlying AD are not yet completely clear but genetic, pathological and biochemical clues suggest that the progressive production and subsequent accumulation of the amyloid-β (Aβ) peptides play a central role. Under abnormal conditions, the accumulation of Aβ progressively forms oligomeric, multimeric and fibrillar aggregates and culminates with the formation of extracellular plaques, one of the morphological hallmarks of the disease [4][5][6].
In the search of new strategies for AD treatment, molecules able to stabilize the soluble Aβ conformation, to destabilize the altered amyloidogenic conformer, and to prevent its aggregation could be effective inhibitors of amyloid plaque formation and very potent drug candidates [7][8][9][10]. Among them, curcumin (1, 1-diferuloylmethane) is a natural molecule with a wide variety of biological effects [11,12]. Curcumin is an active principle of the perennial herb Curcuma longa (commonly known as turmeric) which is used in the Indian traditional diet and as herbal medicine [13]. Recent studies indicate a role for curcumin as a potential anti-amyloid agent in vitro [14][15][16] and in vivo [17,18]. However, curcumin is unstable in aqueous media [19] and presents low bioavailability following delivery through oral or parenteral route [20]. The attachment of curcumin on the surface of Nanoparticles (NP) might increase the drug bioavailability [21]. Moreover, NP offer an attractive tool in driving drugs to the brain, due to their high potential for surface functionalization with ligands able to promote the crossing of the Blood-Brain Barrier (BBB) [22][23][24], a tightly packed layer of endothelial cells surrounding the brain TAT functionalization was assessed by surface plasmon resonance technique. The BBB crossing was investigated by measuring the NL permeability across a transwell BBB cellular model, made by human cerebral microvascular (hCMEC/D3) cultured cells, following [ 3 H]sphingomyelin added to nanoliposome as tracer by means of liquid scintillation counting and curcumin-derivative by mass spectrometry.

TAT-peptide synthesis and characterization
The sequence corresponding to residues 48-57 of human TAT protein was synthesized on an automated Applied Biosystems synthesizer model 433A (Applied Biosystems, Foster City, California) at 0.1 mM scale with NOVASYN-TGA resin, using Fmoc-protected L-amino acids. The peptides were bearing at the C-terminal a tryptophan residue for monitoring the peptide by fluorescence and ended with cysteine residue for covalent coupling with mal-PEG-DSPE. Amino acids were activated by reaction with O-(Benzotriazol-1-yl)-N,N,N',N'-tetramethyluronium tetrafluoroborate and N,Ndiisopropylethylamine. A capping step with acetic anhydride after the last coupling cycle of each amino acid was included. Peptides were cleaved from the resin with trifluoroacetic acid/water/3, 6-dioxa-1, 8-octanedithiol (90:5:5 vol/vol/vol, 180 min at 25°C), precipitated, and washed with diethyl ether. Crude peptide was then purified by reverse-phase high-performance liquid chromatography on a semipreparative C18 column (Symmetry 300; Waters Corporation, Milford, Massachusetts), and peaks collected were characterized by matrix assisted laser desorption/ionization mass spectrometry (MALDI-TOF). The sequence was GRKKRRQRRRPPQGWC (2065.47 g/mol) and purity was higher than 95%.

Preparation of Curc-decorated nanoliposomes (Curc-NL) by 'click chemistry'
Click reaction between azido-PEG-lipid containing NL and Curc was performed using the Huisgen 1, 3-dipolar cycloaddition of azides and terminal alkynes, with some modification from [30]. Briefly, an aqueous solution of CuSO 4 (8 mM) and sodium ascorbate (145 mM) were added to a solution of Curc (100 mM in DMSO). The obtained mixture was added to NL and the reaction was gently stirred for 6 h, pH 6.5, at 25°C. The resulting mixture was purified by ultrafiltration with Amicon Ultra-15 devices and by gel filtration through a Sepharose 4B-CL column (25×1 cm). NL elution was assessed by DLS (Dynamic Light Scattering). The presence of Curc linked to lipid (Curc-PEG-lipid) in NL after purification was verified by mass spectrometry (see below). Phospholipid recovery was determined as described above [39].

Functionalization of Curc-NL with TAT-peptide (TAT-Curc-NL)
TAT-peptide was added to maleimide containing Curc-NL in PBS to give a final peptide-to-maleimide molar ratio of 1.2:1. The mixture was incubated overnight at 25°C. Peptide-bound NL was separated from the unbound peptide using a PD-10 column. The yield of coupling and the amount of coupled peptide was assessed by tryptophan fluorescence intensity measurements, as reported [40].

Characterization of NL
All NL preparations were characterized in terms of size, ζ-potential, polydispersity index and stability by DLS, as described [38].

HPLC-MS/MS method for Curc-PEG-lipid quantification
The HPLC system consisted of an Alliance separation module 2695 (Waters, Milliford, MA, USA) and a chromatographic X Terra C8 column 3.5 µm, 150×2.0 mm (Waters, Milliford, MA, USA) coupled with a security guard C8 cartridges 2.1×10 mm (Waters, Milliford, MA, USA), held at 30°C. The HPLC system was coupled with a Micromass Quattro Micro triple-quadrupole mass spectrometer (MS, Waters, Milliford, MA, USA), controlled by Mass-Lynx software 4.0 and equipped with an electrospray ionization interface using argon as collision gas. MS analyses were done using positive ionization and multiple reactions monitoring mode, measuring the fragmentation products of the deprotonated pseudo-molecular ions of Curc-PEGlipid. The choice of fragmentation products and the optimization of collision-induced dissociation energies and other instrumental parameters were done in continuous-flow mode, using Curc-PEG-lipid solutions (50% methanol containing 0.05% formic acid) as standard. Samples were analyzed with the ion spray needle operating at 4.0 kV, cone voltage at 43 V, collision energy at 30 eV, source and desolvation temperatures at 100 and 300°C, respectively. The m/Z of the bicharged Curc-PEG-lipid ion was 823.0 and the principal ion transition 823.0>377.3 were selected for quantification. The Mobile Phases (MP) were composed by: 0.05% formic acid in acetonitrile/water, 80:20 (MP-A), and 0.05% formic acid in acetonitrile (MP-B). The HPLC system was set up to operate at a flow rate of 0.2 mL/min (from 0 to 100% of solvent B in 10 min, hold at 100% for 10 min and re-equilibration for 6 min at 0% of solvent B). The retention time for Curc-PEG-lipid was 11.9 min.

Interaction of NL with Aβ investigated by Surface Plasmon Resonance (SPR)
SPR experiments were conducted by using a SensiQ semi-automatic SPR machine (ICx Technologies) with two parallel flow channels; one was used to immobilize Aβ fibrils while the other was used as "reference" (empty surface). Aβ aggregates were prepared as already described [38] and immobilized on a COOH5 sensor chip (ICx Technologies) by amine coupling chemistry using the same procedure described by Le Droumaguet et al. [41]. The final immobilization level was ~5000 Response Unit (RU, 1 RU = 1 pg/mm²). The empty "reference" surface was prepared following the same immobilization procedure, without the addition of the peptide. Sensorgrams were obtained via subsequent injections of not-functionalized NL and TAT-Curc-NL flowed at 30 µL/min for 5 min, at three different concentrations (2, 6, 12 µM, of total lipids) over the immobilized ligand or the reference surface, at the same time. The non-specific binding to the reference channel was automatically subtracted from the total signal. The binding profiles were processed by Qdat Software (ICx Technologies) and a "global" fit of the entire series of curves, obtained according to the pseudofirst order 1:1 Langmuir interaction model and according to the same model incorporating mass transport limitation, was used to calculate association (K a ) and dissociation (K d ) rate constant and the corresponding KD. [25][26][27][28][29][30][31][32][33][34][35] were seeded at a concentration of 27000 cells/cm 2 and grown in tissue culture flasks coated with 0.1 mg/ mL rat tail collagen type 1, in EBM-2 medium supplemented with 5% fetal bovine serum, 1% Penicillin-Streptomycin, 1.4 µM hydrocortisone, 5 µg/mL ascorbic acid, 1/100 chemically defined lipid concentrate 10 mM HEPES and 1 ng/mL basic fibroblast growth factor. The cells were cultured at 37°C, 5% CO 2 /saturated humidity. Cells culture medium was changed every 2-3 days.

Assessment of NL cytotoxicity on hCMEC/D3 cells
hCMEC/D3 cells (30000 cells/cm 2 ) were grown on 12-well plates coated with type I collagen. Medium was replaced and NL (100 µM) suspended in cell culture medium were incubated at 37°C with the cells for up to 48 h, corresponding to 2-fold the cell doubling time. After treatment, the cell viability was assessed by MTT (3-(4, 5-dimetiltiazol-2-yl)-2, 5-diphenyltetrazolium bromide) assay, as described [40]. Each sample was analyzed at least in triplicate.

NL uptake studies carried out by confocal-laser-scanning microscopy (CLSM)
CLSM was employed in order to study the cellular uptake of fluorescently labelled NL. Pictures were taken using an LSM710 inverted CLSM equipped with a Plan-Neofluar 63×/1.4 oil objective (Carl Zeiss, Oberkochen, Germany). Excitation was performed using two V/VISlaser diode 25 mV (405-488) and Ar-laser (540 nm) at 10% intensity. The pinhole was set to 1 Airy. Image acquisition was done sequentially to minimize cross-talk between the fluorophores. 30000/cm 2 cells were cultured for 2 days on rat type I collagen-coated cover slips (diameter 25 mm) positioned in culture dishes, and incubated at 37°C with Curc-NL or TAT-Curc-NL containing BODIPY-Sm suspended in serum free medium for 3 h, rinsed 3 times with PBS and fixed with a 10% formalin solution. NL design has been optimized to achieve recognition by the targeted cells, and approaches such as the one here presented can be used to determine whether targeting is indeed achieved. All the results presented from now on were obtained in serum free medium. Before moving towards in vivo studies, we will verify whether NL targeting is preserved in more realistic biological environments. After 3 washes with PBS, cells were permeabilized with 0.2% Triton-X100 in PBS for 15 min, then rinsed twice and incubated with a solution of 1% phalloydin (actin filaments staining) in PBS for 1 h, then with 20 µM DAPI (nuclear staining) in PBS for 10 min and finally with LAMP-1 (1:200) (late endosomes and early lysosomes staining) or EAA1 (1:200) (early endosomes staining), for 4 h, at 25°C. After 3 washes in PBS, the samples were mounted using polyvinyl alcohol mounting medium.

NL uptake and permeability across hCMEC/D3 cell monolayer
To quantify the association of NL with hCMEC/D3 cells and the permeability across a monolayer made with the same cells, two NL components, [ 3 H]-Sm and Curc-PEG-lipid, were followed by liquid scintillation counting and MS, respectively.
For uptake experiments, NL containing [ 3 H]-Sm as tracer were suspended in serum free culture medium at 100 μM of total lipids concentration, added to hCMEC/D3 cells (30000 cells/cm 2 ) grown in 12-well plates, and incubated for 3 h at 37°C. The medium was withdrawn and cells were washed with 1 mL culture medium containing 0.5% serum, 1 mL PBS and with 400 µL trypsin 0.1% for 3 min, followed by PBS washing. Finally, the cells were detached with 400 μL trypsin 0.25%/EDTA. The suspensions were mixed with 5.0 mL Ultima Gold scintillation cocktail and counted with a Tri-Carb 2200 CA Liquid Scintillation Analyzer (Packard) for [ 3 H]-Sm quantification. To quantify Curc-PEG-lipid by MS, samples were extracted with 4 volumes of ethyl-acetate; after shaking for 15 min, samples were centrifuged for 2 min at 1000 rpm, the solvent layer recovered and dried under nitrogen at 25°C. The residues were then dissolved in 100 µL MP-A and 20 µL were injected into the HPLC system at 10°C. A calibration curve was prepared containing known amounts of Curc-PEG-lipid (linearity range from 0.5 to 300 ng), and then extracted using the same procedure. Peak area of Curc-PEG-lipid was used for quantification by MS.
For permeability experiments across a cell monolayer, hCMEC were seeded in a 12-well transwell inserts coated with type I collagen. 0.5 mL of cell suspensions containing 6×10 4 cells were added to the upper (donor) chamber which was inserted into the lower (acceptor) chamber containing 1 mL of the culture medium. A tight cell monolayer was usually formed 8-10 days after seeding, judged by three criteria: (1) the cells formed a confluent monolayer without visible spaces between cells under a light microscope; (2) the height of the culture medium in the upper chamber had to be at least 2 mm higher than that in the lower chamber for at least 24 h; and (3) a constant TEER (Trans Endothelial Electrical Resistance) value, measured using an EVOM Endohm chamber (World Precision Instruments, Sarasota, FL) was obtained. Wells were used when TEER value was higher than 40 Ω•cm 2 . Transendothelial permeability coefficient (PE) was calculated as previously described [40]. All the permeability experiments were performed in serum free culture medium at 37°C, adding 100 μM (as lipids) NL incorporating about 20000 dpm/ml of [ 3 H]-Sm into the upper chamber. After adding NL to the upper compartment, aliquots were taken from the lower compartment at different times (0-60-180 min) for liquid scintillation counting and for Curc-PEG-lipid quantification, as above described. In parallel experiments, the efflux of the hydrophilic marker [ 14 C]-sucrose (200 µM) added in the upper chamber was measured to evaluate the paracellular permeability. To monitor cell monolayer integrity during permeability studies, we performed co-incubation with unlabelled NL and [ 14 C]-sucrose. At 0, 60, 180 min [ 14 C]-sucrose radioactivity was counted and PE was calculated as described [40].

Statistical analysis
Each experiment was performed at least in triplicate. The differences were evaluated for statistical significance using Student's t-test.

Time-dependent stability of Curc
Fluorescence spectra of Curc and native curcumin were recorded after incubation for different times. The results showed that the maximum fluorescence intensity of native curcumin was reduced of more than 50% in the first 2 h of incubation, compared with a reduction of only 10% for Curc. Moreover, Curc fluorescence was also detectable after longer incubation times (Figure 1).

Curc-NL and TAT-Curc-NL preparation and characterization
Curc was coupled to NL through a 'click reaction' between its triple bond and the azido function of azido-PEG-lipid, present in NL as 10 molar % of total lipids. Theoretical structure of Curc-NL is depicted in figure 2. The yield of the 'click reaction' ranged between 75 and 95%, as determined by quantifying Curc-PEG-lipid amount in Curc-NL dispersions by mass spectrometry ( Figure 3A). Curc-NL preparations resulted monodispersed, with a mean diameter of 179.7 ± 3.5 nm and negatively charged, as reported in table 1.
TAT-peptide was coupled to Curc-NL through reaction between its cysteine thiol group and maleimide function onto the distal end of PEG chains of mal-PEG-DSPE lipid. After purification of TAT-Curc-NL from the unbound peptide by gel filtration, fluorescence emission spectrum was recorded to verify the presence of TAT-peptide onto NL. The tryptophan fluorescence spectrum of TAT-Curc-NL displayed a blue shift to 350 nm in the emission maximum, with respect to 354 nm of TAT-peptide alone ( Figure 3B), suggesting that the coupling of TAT with the mal-PEG-DSPE lipid occurred, as previously reported for other peptides [42]. The yield of coupling was between 60 and 70%. After the coupling with TAT-peptide, TAT-Curc-NL preparations are still monodispersed, with slight modifications in size and charge ( Table  1). The size of all NL preparations remained unchanged for at least 48 h (data not shown).

SPR analysis of NL binding to Aβ peptide
SPR was exploited to investigate the association and dissociation of NL to Aβ fibrils ( Figure 4). TAT-Curc-NL, as well as not functionalized NL, was injected at concentration of 2, 6, 12 µM total lipids over parallel flow channels of the same sensor chip with immobilized Aβ fibrils or nothing (reference surface). The non-specific binding to the reference channel was automatically subtracted from the total signal. Only negligible binding was detected for not functionalized NL ( Figure  4A); on the contrary, TAT-Curc-NL interacted with immobilized Aβ as indicated by the marked increase of RU signal versus time ( Figure 4B). The "global" fit of the entire series of curves was obtained according to the pseudo-first order 1:1 Langmuir interaction model. Fitting was significantly better by incorporating mass transport limitation, and are shown as grey lines in figure 4B. The KD value of TAT-Curc-NL towards Aβ calculated on total lipids concentration was in the low nanomolar range: KD=48 nM.

Assessment of NL cytotoxicity on hCMEC/D3 cells
NL toxicity was tested using MTT assay. All the NL preparations tested were non-toxic, since the cell viability of hCMEC/D3 cells

Cellular uptake of NL studied by CLSM
Cellular uptake by hCMEC/D3 cells of Curc-NL and TAT-Curc-NL fluorescently-labelled with BODIPY-Sm was qualitatively evaluated by CLSM ( Figure 5). In the case of Curc-NL very low amounts of fluorescence associated to cells were detected after 3 h incubation. On the contrary, the coupling of NL with TAT-peptide increased the amount of fluorescence associated to hCMEC/D3 cells. The images showed the presence of hot green spots below the plasma membrane and near the perinuclear recycling region.
Successively, we performed a staining of late-endosomes and earlylysosomes by means of LAMP-1 and early endosomes by means of EAA1. Our results show that Curc-NL and TAT-Curc-NL did not colocalize with early endosomes or late-endosomes and early-lysosomes, though their fluorescence was present in the same optical section ( Figure 5), thus indicating their cellular uptake.

Cellular uptake and permeability investigated by radiochemical technique and mass spectrometry
We quantitatively evaluated the uptake of Curc-NL and TAT-Curc-NL by hCMEC/D3 cells using [ 3 H]-Sm-labelling. After 3 h incubation with radiolabelled NL, and mild trypsin treatment to remove membrane surface-adsorbed radioactivity, cells were detached with trypsin/EDTA solution and the radioactivity stably associated with cells was counted.  NL were functionalized with Curc and TAT-peptide and purified as described in the text. A) Representative mass spectrometry spectrum of Curc-PEG-lipid, the product of the reaction between Curc and azido-PEG-lipid containing NL; B) Representative fluorescence spectra of TAT-peptide before (red) and after (black) conjugation to maleimide containing NL (λ ex 280 nm). Curc=curcumin-derivative; NL=nanoliposomes.
Furthermore, the uptake of NL has been evaluated by quantification of Curc-PEG-lipid by HPLC-MS/MS. When hCMECD/3 cells were incubated with Curc-NL or TAT-Curc-NL, Curc stably associated with cells was 0.37 ± 0.04% or 1.18 ± 0,16% (p<0.05) of the administered dose, respectively. These results proved that the cellular uptake of Curc-NL strongly increased after NL functionalization with TATpeptide. The ζ-potential of the here prepared NL was slightly negative, nevertheless the TAT sequences added to NL surface has been found to drive a direct interaction of TAT-NL with hCMEC/D3 cells to sustain their greater intracellular uptake.

Discussion
With the number of people with dementia increasing rapidly worldwide, more than 36 million people with the disease today and more than 115 million predicted for the year 2050, care for people with dementia will put an unprecedented burden on health and social systems. Therefore, the search for effective therapies and early diagnosis, still lacking today, is imperative. So far, several therapeutic strategies have been proposed. Since extracellular aggregates (plaques) of the cytotoxic Aβ peptide in the AD brain are hallmarks of the disease, one therapeutic strategy considers Aβ as target [43].
This study was aimed at the preparation and characterization of NL double-functionalized to bind Alzheimer's disease Aβ peptide and to cross the BBB.
To bind Aβ peptide, NL functionalized with a curcuminderivative (Curc) with high affinity for Aβ [30] were utilized. Curc has a good stability, as proved within the present investigation, much improved compared to curcumin itself, probably due to the absence of the chemically labile β-diketone moiety [31]. The coupling of Curc molecules to the surface of NL by 'click chemistry' proceeded easily and almost quantitatively at 25°C in our experimental conditions. Among different nanoparticles suitable for click chemistry [44], or already utilized for coupling with Curc [30,41], we decided to use nanoliposomes, due to their known non-toxic and non-immunogenic features, fully biodegradable and structurally versatile nature [45].
In the present investigation we utilized NL composed of sphingomyelin and cholesterol; different from the formulation already reported [30]. This is a formulation repeatedly utilized in vivo for therapeutic purposes, displaying good circulation times in blood, biocompatibility and resistance to hydrolysis [46].
Starting from this point, we further functionalized Curc-NL with the aim to confer them the ability to cross the BBB. Development of strategies to deliver drugs to the central nervous system is of high importance, because many drug candidates are not able to permeate the BBB. Among all the different ligands proposed for BBB passage (as endogenous and foreign proteins, antibodies, and peptides) [24], here we employed a modified cell penetrating TAT-peptide. TAT protein from HIV-1 or certain small regions of such protein were shown to enter cells when added to the surrounding media [47], and to carry heterologous proteins across the BBB [32]. TAT-peptide used in the present investigation was modified by adding at the C-terminal a tryptophan residue for spectrofluorimetric quantitative detection and a cysteine residue for the binding to maleimide function of mal-PEG-DSPE lipid incorporated into NL during preparation. Before testing the ability of TAT-Curc-NL to cross the BBB in vitro, we worried whether the functionalization with TAT affected or not the reported ability of Curc-NL [30]  of NL towards the peptide, assessed by SPR, confirmed that doublefunctionalized NL retained the ability to bind amyloid β-peptide.
Successively, we observed that the functionalization of Curc-NL with TAT strongly enhances NL uptake by human brain capillary endothelial cells (hCMEC/D3) in comparison to Curc-NL.
Thus, TAT-Curc-NL increased availability at the neurovascular junction of the cerebral microvasculature forming the BBB will be particularly useful to treat neurological disorders such as AD and cerebrovascular amyloidosis.
Interestingly, our experiments showed that TAT-Curc-NL were uptaken and clustered in the peri-nuclear region, with reduced cytoplasmic distribution, and that their uptake did not configure an active endocytic mechanism. The mechanism involved in TAT traslocation through cellular membranes is not completely understood, but our observations suggested that the exclusion from endosomal/ lysosomal pathway could facilitate transcellular passage, as already reported [48]. Moreover, it should be underlined that bypassing the endocytic route would be very beneficial when using NP as drug carriers because the putative drug could be degraded within the endosomal/ lysosomal compartment [49].
The ability of TAT-Curc-NL to cross the BBB in vitro was assessed on a transwell system made of cultured hCMEC/D3 cells, displaying characteristics and functionality mimicking the basic features of the BBB. The PE of NL, measured by following two of its components, Sm (adding [ 3 H]-Sm as a tracer) and Curc (followed by MS technique) was much higher for TAT-Curc-NL with respect to Curc-NL, proving the effectiveness of double-functionalized NL to flux across the cellular monolayer. Interestingly, the comparable PE values of [ 3 H]-Sm and Curc-PEG-lipid, suggested that TAT-Curc-NL might be still intact after BBB passage.

Conclusion
Our results show that TAT-Curc-NL are able to cross an in vitro BBB model and display an high affinity for Aβ peptide. Thus, TAT-Curc-NL might be considered as promising tool for implementing innovative strategies for drug and contrast agent delivery to AD brain. Further studies aiming at investigate in vivo the pharmacokinetic and biodistribution of TAT-Curc-NL both in healthy and in AD transgenic mice are required.