Co-loading of Levodopa and Curcumin Using Brain-targeted Protocells for Improving the Ecacy of Parkinson's Disease

Parkinson's disease (PD), one of the most common movement and neurodegenerative disorders, is challenging to treat. Levodopa is a common clinical drug for controlling the symptoms of PD, but it only replenishes the missing dopamine, can’t protect dopaminergic neurons. While curcumin as a neuroprotective agent has been reported for treatment of PD. Herein, we present a novel organic-inorganic composite nanoparticle with brain targeting (lf-protocells) for co-delivery of levodopa and curcumin, and demonstrate its attractive use as a biocompatible platform for PD treatment. The nanoparticle system is comprised of a lactoferrin (lf) modied lipid bilayer (LB) containing curcumin as its outer membrane and mesoporous silica nanoparticles (MSNs) containing levodopa as its supporting inner core. Our studies illustrate that the lf-protocells have a spherical morphology, and can be used to co-load levodopa and curcumin eciently; the combination of curcumin and levodopa alleviates the apoptosis of PD cells, decreases the expression of a-synuclein and increase the expression of tyrosine hydroxylase in SH-SY5Y cells as compared to single drug; the binary-drug loaded lf-protocells ameliorate oxidative stress and mitochondrial dysfunction as compared to combination of free drugs; lf-protocells improve signicantly the distribution in brain compared with unmodied protocells; binary-drug loaded lf-protocells have better performance of motor function in mouse than unmodied protocells and combination of free drugs. In conclusion, binary-drug loaded lf-protocells show better therapeutic ecacy in both cell model and mouse model of PD than combination of free drugs and lower toxicity than bare MSNs. These results suggest that lf-protocells can be used as a promising drug delivery platform for targeted therapy against PD and other diseases of the central nervous system.


Introduction
Parkinson's disease (PD) is the second most common chronic progressive neurodegenerative disorder after Alzheimer's disease (AD) as it affects almost 1-2% of population aged over 65 in the world [1] . The typical pathological hallmarks of PD include the loss of dopaminergic neurons in the substantia nigra pars compacta (SNc), the decrease of tyrosine hydroxylase (TH), and the appearance of Lewy bodies formed by the accumulation of misfolded α-synuclein (α-syn) protein [2] . TH catalyzes the conversion of L-tyrosine to dihydroxyphenylalanine, which is the initial and rate-limiting step in the biosynthesis of dopamine [3] .
Although the pathogenesis of PD is still unclear [4] , most reports show that it is associated with oxidative stress, neuro in ammatory reaction, mitochondrial dysfunction, cell apoptosis and so on [5] . Several lines of evidence have implicated that mitochondrial dysfunction is a key element in the pathogenesis of PD [6,7] . While, oxidative stress, as a consequence of mitochondrial dysfunction, is increased in the brain tissue of patients with PD [8] . Cell apoptosis comprises two main pathways: the extrinsic pathway that does not involve mitochondria, and the mitochondrial apoptotic pathway which is involved in the pathogenesis of PD [9] . Following the inhibition of mitochondrial electron transport chain (ETC) complex I, and production of reactive oxygen species (ROS), mitochondrial apoptotic pathways can be activated [9] . Furthermore, products of PD genes are linked to apoptosis. Overexpression of wild type or mutant asynuclein in cultured neurons also causes apoptosis [10] . A vicious cycle in which α-syn aggregation and mitochondrial dysfunction exacerbate each other could explain why these cellular changes are observed together in degenerating neurons in PD [2] .
Studies have linked dysfunction of ETC complex I to PD, and many of the genes whose mutations cause familial PD are involved in mitochondrial function and homeostasis [11] . As neurotoxins, both rotenone and MPTP are the mitochondrial complex I inhibitors, which inhibit complex I and consequently increase ROS production [12] . These neurotoxins also cause neuronal death and almost completely reproduce the typical pathological and clinical features of PD, including the loss of dopaminergic neurons, the accumulation of α-syn and the mitochondrial dysfunction [12] .
So far, levodopa is considered as the most effective medication for the symptomatic treatment of PD [13] .
It works by being converted into dopamine and compensating for the dopamine de ciency in the brain [13] . However, the e cacy of levodopa in the late stage of PD is signi cantly reduced due to its metabolism, subsequent low bioavailability and irregular uctuations of its blood concentration [14] .
Extensive metabolism of levodopa, principally through decarboxylation, O-methylation, transamination and oxidation, also generate side effects such as nausea, vomiting and cardiac arrhythmias [15] . The short half-life of levodopa is a major factor responsible for uctuating plasma levels and clinical response oscillations [16] . In general, > 95% of levodopa is metabolized in the gastrointestinal tract, liver and plasma, whereas only 1% of the ingested dose of levodopa penetrates the central nervous system (CNS) for the treatment of PD [17] .
Various therapeutic strategies have been developed to extend the bioavailability of levodopa, mainly through sustained release systems, but also through the combined therapies with inhibitors of decarboxylase and O-methyltransferase [18] . However, the presence of oxidative processes in the disease has opened up new research that has led to new strategies. It is noted that several drug candidates such as EPI-743 [19] , Baicalein [20] and NTCELL [21] currently used in clinical trials exert their therapeutic effects on PD by antioxidation, protecting mitochondria or acting on glial cells. It suggests that intervention of oxidative stress, improvement of mitochondrial dysfunction, resistance to neuroin ammation, protection of dopamine and other nerve cells are potential therapeutic strategies for PD [21] .
As a natural polyphenol extracted from the dry rhizome of zingiberaceae plants [22] , curcumin has numerous pharmacological properties such as anti-in ammation, antioxidation, anti-tumor and anti-nerve degeneration [23] . It reduces the high level of oxidative stress and neurotoxicity mainly through antioxidant effect and chelating metal ions, and thus achieves the protection of dopaminergic neurons and improves the level of dopamine [24] . Previous studies suggested that curcumin could be applied in the treatment of neurodegenerative diseases such as PD and AD [25] . Hence, if combined with levodopa, curcumin may lead to enhanced therapeutic effects of PD on the basis of levodopa supplementing neurotransmitters, since it works through antioxidation and repairing injured neurons. In other words, the combination of these two different durgs is expected to enhance the therapeutic effect of PD. So far, no relevant pharmacological study has been reported.
However, the physicochemical properties of hydrophilic levodopa and hydrophobic curcumin are quite different, which make their co-delivery confront with great challenges. So it is necessary to construct an eligible drug delivery system that e ciently co-load these two drugs. Protocells, consisting of a mesoporous silica nanoparticles (MSNs) core coated by a layer of lipid bilayer (LB), are a class of highly scalable nanocarriers that have caused great concern in drug delivery applications [26] . Inner sponge-like inorganic MSNs with favorable surface area, high porosity and adsorption performance, extraordinary drug loading capacity, are excellent carriers of hydrophilic drugs such as levodopa. The outer membrane LB can not only contain hydrophobic drugs like curcumin but also effectively seal the surface pores of the inner MSNs to prevent drug leakage, thus achieving the co-delivery of the different drugs. In addition, these drugs can be delivered simultaneously without affecting each other.
In our study, protocells delivery system co-loaded with levodopa and curcumin were constructed, brain targeting ligand-lactoferrin (Lf) was then modi ed on the LB surface to enhance cellular uptake and brain targeting [27] . The blood-brain barrier (BBB) formed by brain vessel endothelial cells linked together with tight junctions, presents a major obstacle to the delivery of drugs passing from the bloodstream to the CNS [28] . Lf receptor existing on the BBB mediated endocytosis is among the most e cient cellular uptake pathway, so the delivery system modi ed with Lf ligand may be more e cient under certain pathological conditions of the high expression of Lf receptor, such as PD [29] and AD [30] .
The physicochemical properties, e cacy in vitro, mechanism of action and preliminary pharmacodynamics in vivo of the drug delivery system named as binary-drug loaded lf-protocells (L/C-Lf-Pro) were investigated. With the modi cation of Lf ligand, the obstacle presented by the BBB are expected to be tackled. Moreover, L/C-Lf-Pro could offer a bene cial strategy in mitigating the adverse reaction and improving the therapeutic effect of levodopa.

Characterization of sp-MSNs and MSNs
The DLS test indicated that the prepared MSNs had outer diameter about 90 nm, similar to the sp-MSNs without TMB modi cation (Tab S1). TEM images showed that the MSNs retained high monodispersity and spherical morphology, and meanwhile possessed larger porosity compared with the sp-MSNs without TMB modi cation (Fig. S1). According to nitrogen adsorption data, with the enlarging pores modi cation, the BET surface area, pore volume, and pore size were changed from 743.78 m 2 /g, 1.01 cm 3 /g, and 5.44 nm to 572.99 m 2 /g, 1.15 cm 3 /g, and 8.00 nm respectively (Fig. S2, Tab. S2).

Characterization of lf-protocells and drug loading
As for 1 H-NMR spectra analysis, the solvent peak of D 2 O at δ 4.71 ppm was used as the reference (Fig.   1B). The 1 H-NMR spectrum of DSPE-PEG 2000 -MAL clearly showed characteristic peak of MAL group at δ 6.84 ppm (Fig. 1B, top), while the disappearance of MAL peak in the 1 H-NMR spectrum of DSPE-PEG 2000 -Lf (Fig. 1B, bottom) matched well with the fact that the MAL group had been reacted with the thiol group of Lf-SH. In a word, the 1 H-NMR spectrum demonstrated the formation of DSPE-PEG 2000 -Lf.
The DLS test indicated that lf-protocells were larger than MSNs, up to 177 nm in diameter. The presence of the coating lipid bilayer was con rmed by TEM. As can be seen, the lf-protocells appeared to be surrounded by a ring, and the appearances of the pores were relatively vague compared with TEM image of MSNs shown in Fig. S1. The co-loading of levodopa and curcumin had little in uence on the morphology of lf-protocells (Fig. 1C).
And the results of DLC and EE of various lf-protocells were shown in Tab. 1. The L-Lf-Pro yielded a levodopa DLC of 20.89 wt % (drug: MSNs) and a levodopa EE of 9.87 wt %, which were quite similar to that of binary-drug loaded lf-protocells (DLC of 21.44 wt % and EE of 10.72 wt % for levodopa). The C-Lf-Pro yielded a curcumin DLC of 1.73 wt % (drug: MSNs) and a curcumin EE of 86.67 wt %, which were also quite similar to that of binary-drug loaded lf-protocells (DLC of 1.89 wt % and EE of 88.31 wt % for curcumin). The co-loading of levodopa and curcumin showed a similar DLC and EE value to that of single-drug loaded lf-protocells, which was attributable to the possible deep immersion of levodopa in the MSNs core and the immersion of curcumin in the lipid bilayer.
In addition, the in vitro release pro les of levodopa and curcumin from binary-drug loaded lf-protocells in neutral or acidic medium were measured using HPLC (Fig. S3). Neutral medium of pH 7.4 simulated body uid such as blood, and acidic medium of pH 5.0 simulated endosomal/lysosomal environment. The results suggested that no levodopa and only 59.2% curcumin was released from lf-protocells for 48 h in pH 7.4 solution at 37 °C. When the pH is decreased to acidic condition (pH 5.0), lf-protocells released 100% of their curcumin and 72.6% of their levodopa within 48 hours. Therefore, we speculated that the release of levodopa was hindered by the inclusion of lipid bilayer outside the inner core, which would greatly reduce its leakage in the circulatory system. However, the drug release from lf-protocells in acidic conditions is initiated by LB destabilization.

Biocompatibility and cytotoxicity of MSNs and lf-protocells
Hemolysis and cell viability assays were performed to investigate biocompatibility and cytotoxicity of MSNs (as the inner core of lf-protocells) and lf-protocells.
The hemocompatibility of MSNs, protocells and lf-protocells was evaluated by hemolysis assay at the concentrations ranging from 50 to 1000 μg/mL. The hemolytic activity of MSNs, protocells and lfprotocell on RBCs was observed using digital photography and calculated by the formula of hemolysis percentage as described in part 4.9. Visual inspection of the hemolysis experiments showed a clear color differential between samples with varying degrees of hemolysis. Uncoated MSNs showed high hemolytic activity, roughly 40 % of the positive hemolysis control of RBCs when the concentration exceeded 400 μg/mL (Fig. S4). In contrast, with the LB coating, protocells showed the hemolytic activity below 5 % of the positive hemolysis control of RBCs, even at concentrations greater than 800 µg/mL (Fig. S4). In addition, the hemolysis ratio of lf-protocells with a concentration of 1000 µg/mL was still less than 5 % compared with the positive control group. Studies on other particle systems have determined that increasing the positive charge density of the particle leads to a decrease in hemolysis. When acidic silanols on MSNs have been masked by the lipid bilayer, no hemolysis is observed [31] . These results demonstrated that the LB coating on the surface of MSNs could reduce dramatically the hemolysis percentage, and improve markedly the biocompatibility of protocells; and modi cation of Lf ligand would not increase hemolytic activity.
The in vitro cell viability was performed by MTT assay to evaluate the toxicity of the MSNs and lfprotocells on SH-SY5Y cells. As shown in Fig. S5, over 90 % cell viability was achieved after incubating SH-SY5Y cells with 200 µg/mL of MSNs and lf-protocells respectively for 24 h, while over 85 % cell viability was achieved after incubating SH-SY5Y cells with 500 µg/mL of particles for 24 h.

Cytotoxicity of curcumin, levodopa and rotenone alone on SH-SY5Y cells
The cytotoxicity of curcumin alone, levodopa alone and rotenone alone to SH-SY5Y cells was evaluated.  had obvious toxic effect on cellular morphology. As previously reported [32] , a signi cant reduction of rotenone-induced cell viability could be observed. Our result of cytotoxicity of rotenone revealed that 1-20 μM of rotenone caused approximately 50 % of cell death after 24 h treatment (Fig. 2B). Moreover, high concentration up to 4 μM of rotenone also induced signi cant changes in cell morphology ( Fig. 2A).
To further investigate the effects of rotenone inductionred4 on SH-SY5Y cells, western blot analysis on several important apoptosis mediated signaling pathway proteins was performed. The expression levels of pro-apoptotic factors poly ADP-ribose polymerase (PARP), Caspase-3, Cleaved Caspase-3, Bax and anti-apoptotic factors Bcl-2 were detected in 1-20 μM Rot-induced SH-SY5Y cells after 24 h cultivation. Compared with blank group, rotenone signi cantly increased PARP, Cleaved Caspase-3 and Bax levels and cut down Caspase-3 and Bcl-2 levels, which indicated promotion effect on apoptosis in a dosedependent manner (Fig. S6A).
Since the typical pathological changes of PD include the deformation of dopaminergic neurons, the aggregation of intracellular α-syn protein and the decrease of TH expression, we measured the expression of α-syn and TH in Rot-induced SH-SY5Y cells. The results of western blot on TH and α-syn, shown in Fig.  S6B, demonstrated that rotenone, compared with blank group, signi cantly decreased the level of TH and increased the level of α-syn in a dose-dependent manner.

Protection of curcumin and levodopa against rotenone-induced cytotoxicity
The cytoprotective effect of curcumin or levodopa against rotenone-induced SH-SY5Y cytotoxicity was determined by pre-treating cells with different concentrations of curcumin or levodopa for 60 min prior to rotenone exposure. Pretreatment with curcumin alone at 1, 2, 4 μM for 60 min prior to rotenone exposure signi cantly increased the cell viability to 79.2 ± 4.4, 82.6 ± 5.6, 76.9 ± 4.4% of control, respectively ( The levels of ROS were determined by the intensity of DCF uorescence within cells. As shown in Fig. 3A, exposure of rotenone (4 μM) led to a dramatic increase of DCF uorescence intensity in SH-SY5Y cells compared with that in control group. After the pretreatment of single drug alone, neither curcumin nor levodopa, could decrease the high level of intracellular ROS induced by rotenone. However, co-treatment of curcumin and levodopa (free L/C), as well as treatment of L/C-Lf-Pro, signi cantly decreased rotenoneinduced ROS production (P < 0.05, P < 0.001, respectively). Interestingly, treatment with L/C-Lf-Pro markedly scavenged rotenone-induced ROS compared with free L/C group (P < 0.05), L group (P < 0.001), and C group (P < 0.001), respectively. In brief, our study showed that L/C-lf-protocells exerted signi cant effect on reduction of intracellular ROS level in SH-SY5Y cells activated by rotenone.
Superoxide dismutase (SOD), as an anti-oxidative defense enzyme, plays a crucial role in maintaining cellular redox homeostasis [33] . Thus, we determined the effect of lf-protocells on the enzyme activities of SOD. As shown in Fig. 3B, rotenone (4 μM) signi cantly decreased the SOD activity in SH-SY5Y cells (P < 0.01). Oxidative stress induced by rotenone was con rmed by the reduced expression of SOD. While curcumin & levodopa co-treatment in the form of free drugs (free L/C) and lf-protocells (L/C-Lf-Pro) signi cantly improved the enzyme activities of SOD reduced by rotenone (P < 0.01), and both of them were superior to single free drug treatment (P < 0.05). In addition, pretreatment with L/C-Lf-Pro increased the activity of SOD to 93.5 %, the highest among all groups, even though there was no statistical difference compared with free L/C group. In short, our study indicated that L/C-Lf-Pro could e ciently recover the SOD activity impaired by rotenone.

Lf-protocells improved rotenone-induced GSH depletion.
GSH is a tripeptide nonprotein anti-oxidant and redox regulator in brain in which GSH depletion probably occurs via oxidative damages caused by increased ROS with signi cant mitochondrial damage [34] .
Based on GSH neutralizing free radicals and reactive oxygen compounds in cells, the assessment of intracellular GSH content should be thought about in further exploration. Since L/C-lf-protocells exhibited a profound effect on scavenging Rot-induced intracellular ROS and ameliorating SOD activity, we speculated that L/C-lf-protocells could also improve Rot-induced GSH depletion in SH-SY5Y cells.
As shown in Fig. 3C, the content of total GSH decreased remarkably after incubation with rotenone (4 μM) compared with control group, whereas the total GSH content increased after the co-treatment of curcumin and levodopa (free L/C), as well as treatment of L/C-Lf-Pro. Noticeably, pretreatment with L/C-Lf-Pro attenuated Rot-induced total GSH depletion, the best among all groups, even though there was no statistical difference compared with free L/C group. These observations implied that L/C-Lf-Pro might provide protection against Rot-induced oxidative damage in SH-SY5Y cells via regulating GSH antioxidant system to cope with oxidative stress.
The occurrence and development of PD are closely related to oxidative stress, which are involved in the process of neuronal deformation in PD. The anti-oxidative therapeutic e cacy of binary-drug loaded lfprotocells were apparently con rmed by SOD and GSH experiment in vitro. The results of SOD and GSH were in conformity with the ROS assay. These ndings suggested that levodopa and curcumin co-loaded lf-protocells, a novel nanoparticle platform comprised of a Lf-modi ed LB containing curcumin as its outer membrane and MSNs containing levodopa as its supporting inner core, compared with free levodopa and curcumin, could effectively enhance the level of total GSH and SOD in damaged cells, reduce the oxidative damage, and thus protect cells against oxidative stress, improve the anti-oxidative therapeutic effect for PD.

Lf-protocells suppressed rotenone-induced reduction of the MMP
As the most important organelle, mitochondria plays a key role not only in supplying metabolic energy to cells in the form of ATP via directly participating in a number of metabolic reaction, but also in regulating the signal transmission during the apoptosis of cells [35] . The burst of ROS and depletion of intracellular SOD and total GSH can induce the drop of MMP. The collapse of MMP, in turn, can also stimulate ROS generating excessively and the level of SOD and total GSH declining [33] . Since L/C-lf-protocells exhibited a profound effect in ameliorating oxidative stress in SH-SY5Y cells induced by rotenone, we postulated that L/C-lf-protocells could also ameliorate Rot-induced oxidative damage to membrane and mitochondrial dysfunction.
As described in part 4.12, the Rh123 uorescence intensity was used to evaluate the MMP changes in cells. As shown in Fig. 4A, the Rh123 uorescence intensity of rotenone group was about half of that of control group, a remarkable decrease of uorescence intensity was observed after treatment with rotenone (P < 0.001). Pretreatment with free drugs (including L group, C group and free L/C group), attenuated the Rot-induced mitochondrial dysfunction as re ected by increasing the uorescence intensity of rotenone group (P < 0.05). In addition, pretreatment with L/C-Lf-Pro increased the uorescence intensity signi cantly compared with that of free L/C group (P < 0.05).
Furtherly, the change of Rh123 uorescence intensity can be displayed by a visual way with the help of uorescent confocal microscope (Fig. 4B). The green light intensity characterizing mitochondrial function weakened owing to the addition of rotenone, followed by becoming stronger after treatment with free L/C or L/C-Lf-Pro, which were in accordance with the former results.
Since MMP reduction could induce apoptotic signalling downstream involving cytochrome C release from mitochondria and subsequent caspase activation, and could also induce pro in ammatory signaling during apoptosis [36] , L/C-Lf-Pro which suppressed the reduction of MMP might reduce apoptosis. As a consequence, L/C-Lf-Pro were expected to provide protection against Rot-induced oxidative damage and mitochondrial dysfunction in SH-SY5Y cells.

Lf-protocells inhibited rotenone-induced neuronal apoptosis.
To investigate the detailed mechanisms, western blot analysis on several important apoptosis mediated signaling pathway proteins was performed. Anti-apoptotic protein Bcl-2 and pro-apoptotic protein Bax play key roles in the process of apoptosis [37] . The Bcl-2 protein could block apoptotic-like forms of death [38] . While protein Bax is the key component of apoptosis caused by mitochondrial stress and increase of membrane permeability after apoptosis stimulation [37] . In addition, once the mitochondrial membrane permeability suffers from damage, Bax may be released from the intermembrane space to initiate caspase activation in the cytosol [39] . Protein Caspase-3 has a vital role to play executor of apoptosis and could crack PARP [40] . The proenzyme form of Caspase-3 is hydrolyzed and activated to a Cleaved Caspase-3 (17 kd subunit + 19 kd subunit) [41] . Moreover, PARP is one of the main shearing targets of Caspase-3 in vivo and could be used as a marker of apoptosis [40] .
Our results, as shown in Fig. 5A, indicated that Bcl-2 and Caspase-3 were up-regulated in both free L/C group and L/C-Lf-Pro group compared with Rot group and single drug groups. Similarly, the expression of Bax, Cleaved Caspase-3 and PARP were down-regulated in both free L/C group and L/C-Lf-Pro group compared with Rot group, and the degree of decrease was markedly enhanced in binary-drug groups compared with single-drug groups. Based on the role of apoptosis mediated signaling pathway proteins as described above, these results suggested that combination of levodopa and curcumin might partially reverse the apoptosis of cells by reducing mitochondrial damage.
Since the typical pathological changes of PD include deformation of dopaminergic neurons, aggregation of intracellular α-syn protein and decrease of TH expression, we measured the expression of these PD related proteins using certain experiments.
The results of western blot on TH and α-syn showed that rotenone could signi cantly decrease the level of TH and increase the level of α-syn compared with Control group, while levodopa and curcumin reversed this trend (Fig. 5B). The addition of both free L/C and L/C-Lf-Pro increased TH level and decreased α-syn level. Interestingly, the performance of binary-drug group was better than single-drug group. Since TH is the key enzyme in dopamine synthesis, co-delivery of levodopa and curcumin is expected to promote the dopamine synthesis by increasing TH expression. Moreover, the addition of levodopa as a dopaminergic neurotransmitter supplement would further increase dopamine levels. Therefore, the results demonstrated that levodopa and curcumin could protect dopaminergic neurons through decreasing the aggregation of α-syn protein and increasing the activity of TH.
Overall, these studies implied that L/C-Lf-Pro pretreatment may alleviate the apoptosis of PD cells and protect dopamine neurons through the following mechanisms: reducing the accumulation of unfolded or misfolded α-syn proteins, enhancing the level of TH and transforming more levodopa into dopamine for supplementing the loss of dopamine in the brain, reducing the mitochondrial damage and dysfunction, down-regulating the expression of Bax protein and up-regulating the expression of Bcl-2, inhibiting the activation of Caspase-3 and thus exerting an anti-apoptotic ability, showing a protective effect in the degenerative changes of dopamine neurons nally.

Cellular Uptake
The bEnd.3 cells are the brain microvascular endothelial cells in the BBB. The BBB blocks passage of most drugs. In our study, Lf-modi ed protocells were expected to allow more drugs to pass through BBB by the speci c receptor-mediated transport. RBITC was labeled on the inner core of protocells by chemical bond, indicating the location of protocells. Calcein was loaded in protocells by the adsorption of MSNs, indicating the location of loading drugs. The cellular uptake of protocells and their loading drugs was investigated, through labeling with RBITC and Calcein respectively.
The results of cellular uptake were shown in Fig. S7. Only the blue nucleus stained with DAPI were visible in the protocells group, and no other uorescence was seen. While the lf-protocells group showed distinct red and green uorescence besides blue. The experimental results indicated that RBITC/Calceinprotocells were hardly internalized by bEnd.3 cells, while the modi cation of Lf ligand was bene cial to increase the cellular uptake of protocells and their loading drugs.

Biodistribution
Biodistribution of carrier after intraperitoneal injection of Cy5-labeled lf-protocells, unmodi ed protocells and bare MSNs (denoted as Cy5-lf-protocells, Cy5-protocells, Cy5-MSNs, respectively) were qualitatively assessed by luminescence imaging in vivo (Fig. 6A). The results indicated that the injection site (above the right leg of mice) showed strong uorescence intensity in all experimental groups after injection for 30 min. After 60 min, the uorescence intensity of lf-protocells group enhanced signi cantly in the brain. Interestingly, the brain uorescence intensity of lf-protocells was further increased, which was much higher than that of bare MSNs and unmodi ed protocells groups for 120 min. The uorescence of bare MSNs and unmodi ed protocells groups concentrated mainly in abdominal cavity, with little or no enrichment in the brain during the whole time of the test.
According to these results, lf-protocells showed speci c brain-targeting properties in mice, and may be distributed in the brain at a high concentration due to the modi cation of active targeting Lf ligand. While either bare MSNs or unmodi ed protocells had no brain targeting. Noteworthily, Lf is a type of ligand with brain-targeting potential, meanwhile, previous studies have demonstrated that the expression of Lf receptor was increased in the brain of PD patients [42] . Hence, the high expression of Lf receptor in cerebral microvascular endothelial cells can be utilized to achieve internalization through blood-brain barrier (BBB) by receptor-mediated transport pathway, so as to promote the brain targeting ability of lfprotocells and make them more enriched in the brain of PD patients [43] . A signi cant challenge of nanoparticles for targeting speci c disease sites is the fact that most nanoparticles become trapped in the liver [44] . Recent study showed that different shaped MSNs were mainly trapped in reticuloendothelial system of the liver, spleen, and lung, accounting for over 80% of the injected dose at 2 h after administration [45] . Obviously, with the coating of Lf modi ed membrane, lf-protocells partly avoided being trapped by the reticuloendothelial system and more enriched into the brain. Animal models of Parkinson's disease (PD) have been widely used in the past four decades to investigate the pathogenesis and pathophysiology of this neurodegenerative disorder. Classical models based on the systemic or local administration of neurotoxins, like MPTP, are able to replicate most of the pathological and phenotypic features of PD in mammals [46] . Exposure to MPTP induces a PD-like syndrome such as motor retardation in mice [47] . In the brain, MPTP is converted to MPP + , which is selectively transported into dopaminergic neuron axon terminals, causing oxidative stress, mitochondrial dysfunction, and cell death [48] .
In our study, the motor function of PD mice induced by MPTP was assessed using the open-eld test. The results indicated that motor function of PD mice was signi cantly attenuated from day 1 to day 5 in the MPTP group, compared with the control group (P < 0.001). Neither levodopa nor levodopa/curcumin did affect the improvement of motor function; however, L/C-Pro and L/C-Lf-Pro signi cantly rescued the reduction of motor function induced by MPTP (P < 0.05, P < 0.01, respectively) (Fig. 6B), and the improvement effect of L/C-Lf-Pro was superior to that of L/C-Pro (P < 0.05).
These results suggested that levodopa and curcumin co-loaded lf-protocells had better therapeutic effect on PD-related motor de cits than other groups. One of the possible reasons is that lf-protocells can package drugs into nanoparticles for elevating the bioavailability of drugs. Especially, lf-protocells have an active brain-targeting distribution, which can deliver more drugs into the brain and exploit the advantages to treat PD. But it is important to note that the behavioral judgement of autonomic activity in mice includes a variety of evaluation systems, one of which is selected in our experiment. Meanwhile, behavioral evaluation is only a part of PD e cacy at the animal level. Therefore, the PD therapy evaluation of lf-protocells needs to be further completed at the animal level in our study.

Discussion And Conclusion
In summary, this work prepared and characterized organic-inorganic composite nanoparticles (namely lfprotocells) which were formed by coating MSNs with lf-modi ed lipid bilayer for levodopa/curcumin codelivery to brain, evaluated therapeutic effects of the binary-drug loaded lf-protocells on PD both in vitro and in vivo, and explored the possible related mechanism (Fig. 7).
Firstly, we successfully synthesized a single carrier named lf-protocells to co-deliver levodopa and curcumin at an optimized drug ratio passing through the BBB for the treatment of PD. Various physicochemical characteristics of MSNs and lf-protocells have been elucidated. Protocells, which are formed by the encapsulation of MSNs cores within supported LB, have the advantages of both liposomes and MSNs: drugs with different properties, such as levodopa and curcumin, could be loaded in different positions of the carrier; the adhesion energy between the MSNs and LB suppresses large-scale membrane bilayer uctuations responsible for liposome instability and leakage, while the LB serves to retain watersoluble cargos within the MSNs [49] . Further safety evaluation in our study indicated that lf-protocells possessed high biocompatibility and low cytotoxicity. The biodegradation behaviors of MSNs have been concerned recently. Researchers [50] proved that MSNs exhibited the three-stage degradation behavior and nally maintained the slow degradation on day-scale. For the brain-targeted application Jun [51] indicated that the concentration of Si in the brain was decreased after injection for up to 7 days, by which time the level had decreased by nearly 92.4%. Afterwards, MSNs can be excreted mainly through renal routes containing either nanoparticles or degraded products [50] .
Several different types of natural products that have been identi ed as potential antioxidants compounds seem to be promising tools with therapeutic potential in oxidative stress-mediated diseases including PD [52] . As a mitochondrial protective antioxidant, curcumin was typically effective for PD treatment [53] . In our study, when combined with levodopa, curcumin enhanced the protective effect on the rotenoneinduced SH-SY5Y cell model of PD. Loaded in the inner core of brain-targeted lf-protocells, levodopa avoided releasing in the neutral medium of pH 7.4 simulated body uid until it was transported into the endosome with acidic condition. Furthermore, with the modi cation of Lf ligand, more uptake of lfprotocells allowed more dose of levodopa to enter the brain, thereby increasing the bioavailability of levodopa. It was worth noting that levodopa not only acted as a dopamine precursor, but also showed a certain degree of anti-oxidative activity. It was reported that levodopa oxidation products prevented H 2 O 2induced oxidative damage to cellular DNA in cultured tissue cells [54] .
Compared to that observed following treatment with free levodopa and curcumin, binary-drug loaded lfprotocells (L/C-Lf-Pro) could increase the distribution of drugs in the brain, thereby exhibit better effects on reducing the level of ROS, increasing the level of SOD and total GSH, ameliorating MMP reduction and decreasing the damage of oxidative stress to PD cells. Furthermore, L/C-Lf-Pro treatment not only reduced α-syn accumulation which could cause the formation of Lewy bodies, but also elevated the activity of TH which is the rate limiting enzyme for the synthesis of dopamine. Presumably, L/C-Lf-Pro not only convert more levodopa into dopamine to make up for the loss in the brain, but also inhibit MMP reduction, turn off apoptosis signaling pathway, alleviate dopaminergic neuron damage and restore cell survival rate, subsequently ameliorate the motor function de cits, since the loss of dopaminergic neurons leading to striatal dopamine depletion is the core mechanism underlying the cardinal motor features of PD [2] .
In vivo, the intraperitoneal injection of L/C-Lf-Pro resulted in improvement of motor function de cits, which was evidently superior to those of any other groups. However, more experiments are required to testify the current point of view. For example, an in vivo study using MPTP-induced animal models is necessary to be further completed.
Overall, L/C-Lf-Pro not only replenished the loss of dopamine in the brain, but also played the neuroprotective role of antioxidant and repairing injured neurons. The synergy of two different mechanisms can yield a better e cacy for PD, which has been successfully con rmed in vitro and in vivo. Therefore, L/C-Lf-Pro have the potential to provide a new approach for the development of drug delivery system with high-e ciency and important application prospect for curing PD. As far as we know, it is the rst time that combination therapy of levodopa & curcumin loaded lf-protocells has been applied in PD.
Our work also offered a basis for lf-protocells as a novel nanocarrier to the CNS. The most intriguing thing is the potential application of this approach to other CNS diseases where future lf-protocells could be extended to nanoparticle delivery system for a broad range of drugs, genes or cell-speci c targeting.

Synthesis of small pore MSN (sp-MSNs) and MSNs
Various mesoporous silica nanoparticles (MSNs) were prepared by sol-gel phase transition method according to the literature [55] with little modi cation. In brief, 0.437 g of CTAB and 0.472 g of Brij58 (as the structure-directing agents) were dissolved in 100 mL of 0.1 M phosphate buffer solution (pH 7.4). With vigorous stirring for 20 min at 60 ℃, 2.14 mL of TEOS (as the silica source) was added to the solution at a rate of approximately 30 drops per minute, and the stirring process at 60 ℃ was continued for another 8 h. In order to enlarge the pore size to increase adsorption, 2 mL of TMB was added after the addition of TEOS for 40 min [56] . Then, the stirring process at 60 ℃ was continued for another 8 h. To distinguish, we called the product synthesized without adding TMB as sp-MSNs, while synthesized with adding TMB as MSNs. The resulting white precipitate of sp-MSNs or MSNs was puri ed to remove remaining surfactant by centrifugation and washing with ethanol and water three times each. To remove the organic template, the sediment was re uxed for 24 h at 78 ℃ in acidic ethanolic solution (ethanol: hydrochloric acid (HCl) = 60:1, v/v). And the nal white precipitate was puri ed by centrifugation and washing with ethanol for three times.

Characterization of sp-MSNs and MSNs
For measuring particle size distribution such as mean particle size as well as PDI, sp-MSNs and MSNs were diluted to an appropriate volume with distilled water and measured by the dynamic light scattering (DLS) method (Malvern Instruments, UK) at room temperature. The nitrogen adsorption/desorption analysis was performed using an adsorption analyzer (ASAP 2460, Micromeritics, USA). According to the adsorption data, the Brunauer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) models were used to calculate the speci c surface areas and the pore size of sp-MSNs and MSNs. Morphology of sp-MSNs and MSNs were observed via transmission electron microscopy (TEM, JEM-100CX II, JEOL, Japan).

Preparation of lf-protocells and drug loading
Refer to Huan Meng's synthetic method of protocells [57] , the coating lipid bilayer consisted of a DPPC/Chol/DSPE-PEG 2000 mixture, the molar ratio of which was 75:20:5; these three excipients were dissolved in chloroform and placed into a round-bottom ask, the resulting lipid solution was dried to a lm by a rotary vacuum evaporator (RV10 digital, IKA, Germany) for 15 min at room temperature. Following the addition of 2 mL MSNs suspension to the as-prepared coating lipid lm at a ratio of 1:1 (w/w dry weight), probe sonication was used for 15 min with 15/15 s on/off working cycle at a power output of 50 W. Since the resulting suspension contained protocells, liposomes and others, the protocells were separated by centrifugation at 12000 rpm for 10 min, followed by washing three times in saline.
Lf modi ed protocells (denoted as lf-protocells) were prepared according to a previously described method with slight modi cations [27] . Lf was dissolved in borate-EDTA buffer (pH 8.0) containing Traut's reagent for 2 h at 4℃. The obtained Lf-SH was mixed with DSPE-PEG 2000 -MAL in PBS (pH 7.0) for 24 h under constant shaking in the dark at 4℃. Thereafter, the product was dialyzed and freeze-dried to acquire the conjugate DSPE-PEG 2000 -Lf in powder form. The coating lipid bilayer consisted of a DPPC/Chol/DSPE-PEG 2000 /DSPE-PEG 2000 -Lf mixture, the molar ratio of which was 75:20:4:1. Finally, lfprotocells were obtained according to the above preparation method of protocells.
To obtain levodopa-loaded MSNs, levodopa was dissolved in HCl (pH 1.0) with a concentration of 10 mg/mL, then 10 mg of MSNs were soaked in 2 mL of the levodopa solution for 48 h under light-sealed condition at room temperature.
To obtain curcumin-loaded lipid lm, 200 μg curcumin was added to the mixed lipid solution as previously described followed by rotary evaporation drying to form the lipid lm containing curcumin.
To achieve co-loading of levodopa and curcumin, the levodopa-loaded MSNs suspension was added to the lipid lm containing curcumin, followed by rehydration, sonication, centrifugation, and washing, similar to preparation method of protocells described above. The Fig. 1A depicted the preparation procedure for lf-protocells as a nanocarrier for co-loading of levodopa and curcumin.
For measuring particle size distribution such as mean particle size and PDI of various lf-protocells (unloaded and drug-loaded), nanoparticles were diluted to an appropriate volume with distilled water and measured by the DLS method (Malvern Instruments, UK) at room temperature. Unloaded and drug-loaded lf-protocells were characterized for morphology using a TEM (JEM-100CX II, JEOL, Japan). A minimum of three images for each sample was captured.

Encapsulation e ciency and drug loading content of lf-protocells
The encapsulation e ciency (EE) and drug loading content (DLC) of drug-loaded lf-protocells were determined by a subtraction method as the following equations: The concentration of levodopa was determined by high-performance liquid chromatography (HPLC) with the mobile phase composed of tetrahydrofuran and 0.1 % TFA (3:97, v/v), and UV detection operated at 280 nm; while the concentration of curcumin was determined by HPLC with the mobile phase composed of methanol and 4 % acetic acid (75:25, v/v), and UV detection operated at 430 nm.

Hemolysis assay of MSNs and lf-protocells
Hemolysis assays were used to assess the safety of MSNs and lf-protocells for in vivo applications. After collection of beagle dog blood samples (provided by our cooperative laboratory), red blood cells (RBCs) were collected by centrifugation at 1000 rpm for 10 min. After discarding the supernatant and washing three times with sterile isotonic saline, the RBCs suspension was diluted to a concentration of 2 % (v:v).
Subsequently, 0.4 mL of 2 % RBCs suspensions were mixed with 0.8 mL of MSNs (or lf-protocells) suspensions in saline at concentrations of 50, 100, 200, 400, 600, 800, and 1000 μg/mL, respectively. The hemolysis of RBCs in saline and 5 % Triton X-100 served as negative and positive control, respectively. The mixtures were incubated at 37 ℃ for 1 h and then centrifuged at 1000 rpm for 3 min. The supernatants were measured at 540 nm using a microplate reader (BioTek Synergy H1, USA). The following formula was used to calculate the hemolysis percentage: 4.10 In Vitro cell viability by MTT assay Cellular toxicity of MSNs and lf-protocells was determined by MTT assay. SH-SY5Y cells were seeded in 96-well plates (1 × 10 4 per well) and incubated for 24 h. Subsequently, cells were exposed to various concentrations of MSNs or lf-protocells for an additional period of 24 h, cell viability was then determined using a colorimetric assay with MTT.
To determine the toxicity of rotenone, SH-SY5Y cells were incubated with different concentrations of rotenone (0.5, 1, 2, 4, 10 and 20 μM) for 24 h, cell viability was then determined by MTT assay.
To determine the neuroprotective effects of free drugs (levodopa and curcumin) against rotenone-induced cytotoxicity, SH-SY5Y cells were cultured in 96-well plates (1 × 10 4 per well) for 24 h as above, and were pretreated with free drugs (the concentrations of curcumin were 1, 2, 4 µM and the concentrations of levodopa were 20, 40 µM). Then these cells were treated with 4 μM rotenone for 24 h, cell viability was determined by MTT assay.
For the MTT assay, each well was treated with 100 µl of MTT-labeling reagent (0.5 mg/mL), and the plate was incubated for an additional 3 h. The resulting formazan crystals were dissolved with 150 µl of dimethyl sulfoxide, and MTT reductions were detected at 595 nm by the microplate reader (BioTek Synergy H1, USA).

Measurement of intracellular ROS
In order to determine the cellular ROS level, Reactive Oxygen Species Assay Kit was used, in which 2', 7'dichloro uorescein diacetate (DCFH), a speci c ROS uorescence probe, can be oxidized into uorescent DCF in the presence of ROS, so the level of ROS can be measured by the intensity of DCF [58] . SH-SY5Y cells (about 1 × 10 5 cells/well in 6-well plates) were cultured for 24 h. After treatment as described in part 4.2 above (group 1-5&8) for 24h, cells were washed twice with PBS and then incubated with 4 μM DCFH-DA in the dark for 20 min. Cells were then washed twice with PBS and harvested in trypsinization. All samples were measured via FACS calibur ow cytometer (BD, USA). Fluorescent measurements were done with excitation and emission wavelengths set at 488 nm and 525 nm, respectively. The experiment was repeated in triplicate.

Measurement of intracellular superoxide dismutase (SOD) activity
The SOD activity was quanti ed by Total Superoxide Dismutase Assay Kit with WST-8. SH-SY5Y cells were seeded in a 6-well plate at 1 × 10 5 cells/well and treated for 24h as described in part 4.2 above (group 1-5&8). Cells were harvested and washed twice with PBS. Then, the SOD activity of all samples was determined according to the Kit manufacturer's instructions, and the absorbance value was read at 450 nm by the microplate reader (BioTek Synergy H1, USA). The protein content was determined using the BCA protein assay. Each experiment was performed in triplicate.

Determination of cellular total glutathione (GSH) level
The total GSH level was quanti ed by GSH and GSSG Assay Kit. SH-SY5Y cells were seeded in a 6-well plate at 1 × 10 5 cells/well and treated as described in part 4.2 above (group 1-5&8). After 24 h incubation, cells were collected and washed twice with PBS. Then each sample was determined according to the Kit manufacturer's instructions. Then, the absorbance value was measured at 412 nm by the microplate reader (BioTek Synergy H1, USA). Each experiment was performed in triplicate.

Measurement of mitochondrial membrane potential (MMP)
To monitor the MMP changes of cultured cells, the mitochondrial speci c uorescent dye Rh123 was used. SH-SY5Y cells were seeded in 6-well plates for the treatment listed in part 4.2 above (group 1-5&8).
After 24 h incubation, the medium of each sample was removed, then 1 mL FBS-free medium with 1 μM Rh123 was added, the incubation was continued for 30 min in the dark at 37 °C. Data were analyzed using a ow cytometer (BD Biosciences, USA).
To visualize the mitochondrial function of cultured cells, the medium was removed after 24 h incubation.
Each sample was incubated with 1 μM Rh123 for 30 min in the dark at 37 °C. Then cells were washed twice with PBS and photographed immediately using an Ultra High Resolution Microscope (TCS SP8 STED, Leica, Germany). The results were expressed as mean Rh123 uorescence intensity. The experiment was repeated in triplicate.

Western blot
SH-SY5Y cells treated as list in part 4.2 were lysed in RIPA lysis buffer supplemented with protease inhibitors. After quanti cation by BCA protein assay, samples were mixed with loading buffer (containing DTT) at a ratio of 4:1, and then boiled for 10 min. The resulting proteins were separated with SDS-PAGE and transferred onto a PVDF membrane. Then the membrane was sealed up in TBST with 5 % skim milk at room temperate for 2 h, and subsequently cultured with primary antibodies including anti-TH, anti-αsynuclein, anti-Bcl-2, anti-Bax, anti-Caspase-3, anti-PARP, anti-β-actin and anti-GAPDH at 4 °C overnight.
After TBST-washing in triplicate, the membrane was incubated with secondary antibodies HRPconjugated A nipure Goat Anti-Mouse IgG(H+L) and Anti-Rabbit IgG(H+L) for 1 h at room temperature.
Protein bands were visualized using enhanced chemiluminescence. The results were analyzed by Image J. GAPDH and β-actin were included as internal controls. The experiment was repeated in triplicate.
RBITC-conjugated MSNs were prepared as follow: 8.6 mg RBITC and 200 μL APTES were dissolved in 2mL absolute ethanol. The solution reacted for 24h in the dark under stirring. CTAB, Brij58 and TEOS were added and reacted as described in part 4.4. After that, 200 μL of the above reaction solution was added. The following procedures were the same as the synthetic method of MSNs. For Calcein loading, Calcein was soaked following a similar approach of levodopa-loaded MSNs, as described above.
RBITC/Calcein-protocells and RBITC/Calcein-lf-protocells were prepared refer to synthetic method of protocells as described above. The authors con rm that the following animal experiments were carried out in accordance with the ARRIVE guidelines, and all methods were carried out in accordance with relevant guidelines and regulations.

Biodistribution
To investigate the biodistribution behavior of lf-protocells in mouse, uorescence-labeled lf-protocells denoted as Cy5-lf-protocells were prepared using Cy5-conjugated MSNs as the inner core.
To evaluate the exploratory and locomotor activities by Open-Field Test, each mouse was placed into the center of a square open-eld box (24 cm * 24 cm * 10 cm, divided into sixteen squares with grids, made of transparent PVC) and acclimated for 10 min. The motor function of each mouse was evaluated by counting the number of line crossings during 5 min in the square open-eld box, persist for ve days [47] .

Data processing
All data were presented as mean ± SD (Standard Deviation) of triplicates measurement. All statistical analysis was done in Graph Pad Prism software. Statistical differences between groups were evaluated using a two-tailed Student's t test and differences with a P < 0.05 were considered signi cant.