Extracellular Vesicles as Drug Delivery System for Treatment of Neurodegenerative Disorders: Optimization of the Cell Source

Abstract

Extracellular vesicles (EVs) represent a next generation drug delivery system that combines nanoparticle size with extraordinary ability to cross biological barriers, reduced immunogenicity, and low offsite toxicity profiles. A successful application of this natural way of delivering biological compounds requires deep understanding EVs intrinsic properties inherited from their parent cells. Herein, we evaluated EVs released by cells of different origin, with respect to drug delivery to the brain for treatment of neurodegenerative disorders. The morphology, size, and zeta potential of EVs secreted by primary macrophages (mEVs), neurons (nEVs), and astrocytes (aEVs) were examined by nanoparticle NTA, DLS, cryoTEM, and AFM. Spherical nanoparticles with average size 110–130 nm and zeta potential around −20 mV were identified for all EVs types. mEVs showed the highest levels of tetraspanins and integrins compared to nEVs and aEVs, suggesting superior adhesion and targeting to the inflamed tissues by mEVs. Strikingly, aEVs were preferentially taken up by neuronal cells in vitro, followed by mEVs and nEVs. Nevertheless, the brain accumulation levels of mEVs in a transgenic mouse model of Parkinson’s disease were significantly higher than those of nEVs or aEVs. Therefore, mEVs were suggested as the most promising nanocarrier system for drug delivery to the brain.

Graphical Abstract

6. Short Summary

In this work, we aimed to develop a novel platform for brain delivery of therapeutics based on extracellular vesicles (EVs) for enabling a broader array for treatment of neurodegenerative disorders. EVs may inherit at some extent properties of their parent cells. Therefore, it is imperative to understand, what kind of producer cells should be used for EVs manufacture, and specifically, for delivery of therapeutics to the brain. Our data indicate that EVs released by inflammatory-response cells, specifically, macrophages, are the most promising candidates for delivery of therapeutics to the inflamed brain.

1. Introduction

Extracellular vesicles (EVs) have entered the field of drug delivery as biological alternatives to synthetic nanocarriers, liposomes and polymeric nanoparticles. These natural vehicles are being extensively investigated in therapeutic settings to treat various diseases, including cancer [1–10], neurological disorders (Alzheimer’s and Parkinson’s diseases, stroke) [8, 11], infectious diseases (meningitis, Human Immunodeficiency Virus (HIV), and HIV-related dementia) [12–16], joint diseases (inflammatory arthritis) [17], as well as autoimmune [18] and cardiovascular diseases (atherosclerosis and heart attack) [19–24]. EVs are known to be released by the most types of cells; their functions vary from waste disposal to transport of nucleic acids, lipids, and proteins to neighboring cells and distant organs [25]. Regarding using these vesicles for drug delivery, EVs are commonly considered as a combination of two types of vesicles, exosomes and microvesicles that display size and compositional heterogeneity dependent on their subcellular origin. Exosomes (30 nm −120 nm) are smaller in size and generated in multivesicular bodies. Microvesicles (50 nm – 500 nm) are larger and generated by outward budding of the plasma membrane [26]. Due to their overlapping sizes, surface markers, and lipid content, these two types of vesicles generally used together for drug delivery as a heterogeneous population.

Recently, our laboratories developed novel drug delivery systems using EVs for the transport of different therapeutic molecules, including the small molecule anticancer agents paclitaxel and doxorubicin for treatment of pulmonary metastatic cancer [27, 28] and triple-negative breast cancer (TNBC) [29]. We also utilized EV-based formulations of therapeutic enzymes, catalase and tripeptidyl peptidase-1 (TPP1), for treatment of Parkinson’s disease (PD) [30] and Batten CLN2 disease [31, 32] respectively. Moreover, EVs loaded with brain derived neurotrophic factor (BDNF) were engineered for treatment of neuroinflammation and neurodegeneration [33]. In these publications, we reported that using EVs as nanocarriers for therapeutic agents improved the pharmacokinetic profile of incorporated therapeutics and facilitated drug transport across biological barriers upon systemic administration resulting in significant therapeutic effects in different animal models. Herein, we focused on the further development of EV-based formulations for neurodegenerative disorders through optimization of the EVs source.

As demonstrated earlier by us and other investigators, EVs can cross the BBB, and deliver their therapeutic cargo to the CNS [30, 31, 34, 35]. To accomplish this transfer, EVs bind to target endothelial cellular membranes through receptor-mediated interactions and enter the cell by membrane fusion, and endocytosis, followed by entering brain parenchyma. Therefore, the size, charge, and morphology of EVs may affect their ability to accumulate and deliver therapeutic cargo to target cells. Next, it was reported that EVs contain biomolecules that are reflective of the cell origin [36, 37]. As such, different expression levels of adhesion proteins on their surface may also play a crucial role in their potency as drug delivery vehicles. For example, EVs homing towards inflamed tissues can be enforced by LFA1/ICAM1 interactions as indicated in our earlier reports [28, 38]. Finally, EVs may inherit their own intrinsic biological activity that can, in turn, reinforce therapeutic efficacy of EV-based formulations, enhancing transport and accumulation of EVs-incorporated therapeutics at the disease site. Thus, considering the cells’ phenotype, and the main characteristics of EVs released by these parent cells is essential to shed light on applicability of these vesicles for drug delivery.

In order to further advance the field based on evidence in the literature and our experience to date, we theorized that appropriate EV-based drug formulations could be best designed by optimizing the source of the EVs, i.e. parent cells. The present study serves to optimize this novel drug delivery system for treatment of neurodegenerative disorders by selecting the source of EVs and characterizing them extensively. Of note, EVs ability to target specific type of cells are known to reflects the mechanisms that govern intercellular communications, specifically between cells of neurovascular unit in the brain. Thus, cross-talk between microglia, neurons, and astrocytes by means of EVs are well-documented within the nervous system [39]. Therefore, we choose three types of origin cells that may be beneficial for CNS drug delivery: primary macrophages, neurons, and astrocytes, and evaluated them as a source for mEVs, nEVs, and aEVs, respectively. A possible impact of the cell type on EVs features that are relevant for drug delivery to the brain, i.e. size, zeta potential (ZP), and morphology was investigated. Furthermore, the expression of EV-specific proteins, tetraspanins and integrins, on their surface that play a crucial role in cell adhesion and targeting to the inflamed endothelial cellular membranes through ICAM and VCAM interactions was examined. This was done using western blot and targeted quantitative proteomics, a technique that has not been applied previously to measure integrins. Next, we compared accumulation of different types of EVs in neuronal cells in vitro. Finally, we examined biodistribution of systemically administered EVs from different cell origins in transgenic Parkin Q311X(A) mice. For the first time, we report superior (about two-fold) brain accumulation of mEVs compared to nEVs and aEVs in PD mice. Thus, the exploration of mechanisms involved in targeted CNS transport of EV-based drug formulations is crucial for their therapeutic application and translation the field of precision medicine.

2. RESULTS and Discussion

2.1. Isolation and Characterization EVs of Different Origin

To evaluate, whether cell source can determine specific properties important for EVs ability to deliver therapeutics to the brain, three parent cell types, primary macrophages, neurons, and astrocytes, were chosen as the most promising candidates in respect to the design of EV-based drug delivery system for CNS transport. The obtained EVs isolated from conditioned culture media of macrophages (mEVs), neurons (nEVs), and astrocytes (aEVs), were characterized in accordance with the recommendations made by the International Society of Extracellular Vesicles [26]. First, the purity of EVs was assessed by ZetaView QUATT Nanoparticle Tracking Analysis using Microscope PMX-420. Measurements were performed with EVs labeled with lipophilic fluorescent dye, CMDR (cell mask deep red) that is known to incorporate into lipid biolayers of cellular/EVs membranes. As a negative control, the medium only was also subjected to the same purification steps and labeling procedures. According to the obtained data, EVs samples were highly pure with number of vesicles close to 100% for macrophage- and astrocyte-derived EVs, of and 84% for neuron-derived EVs (Supplementary Table S1). Next, the size, zeta potential (ZP), and morphology of different origin EVs were studied by NTA, DLS, cryoTEM and AFM, respectively (Figure 1). NTA studies, showed average size 110–130 nm for EVs released by all three types of parent cells (Figure 1A) that were not statistically different between them. Of note, all samples showed a broad distribution of the zeta potential (Supplementary Figure S1), however, nEVs showed slightly more narrow distribution compared to mEVs and aEVs. This suggests that both mechanisms of cell entering known for EVs, i.e. endocytosis and fusion with plasma membrane, are equally available for all three types of nanocarriers. According to DLS and ZetaView (ZW) analysis, an average ZP for mEVs and nEVs around −20 mV. In particular, aEVs showed slightly less negative charge around −10 mV by ZW, or −15 mV by DLS. Finally, AFM studies (Figure 1 B–D) and CryoTEM studies (Figure 1 E–G) confirmed spherical morphology of all types EVs. Overall, the obtained data suggest that, at least from the standpoint of size, charge, and morphology, all three types of EVs nanocarriers should be equally considered for the drug delivery to the brain.

Figure 1. Characterization of EVs by NTA, DLS, ZW, AFM and CryoTEM.

Size and zeta potential (ZP) of EVs released by cells of different origin were obtained by NTA, DLS, and ZW (A). An average diameter was around 120–150 nm. To study morphology by AFM, EVs were deposited on positively charged mica (APS) for 2 min, washed with deionized water and dried under nitrogen flow. AFM and CryoTEM images revealed spherical particles with size about 60–100 nm for EVs released by macrophages (B, E), neurons (C, F), and astrocytes (D, G), respectively. N = 20. The bar: 200 nm (AFM), and 100 nm (CryoTEM).

It is known that EVs contain a specific composition of membrane proteins, tetraspanins and integrins, in a functionally active form that can regulate binding affinity to ligands on the recipient cell membranes or the extracellular matrix and facilitate their homing and accumulation in target tissues [40]. Therefore, we investigated whether cellular origin affects the amount of the selected membrane proteins on EVs. First, the levels of EV-specific markers, HSP90, TSG101, Integrin α, and CD9, in different types of EVs were examined by Western blot using Wes^™ technique (Figure 2). As seen in the Figure 2A, mEVs (1) showed abundance of tetraspanins and integrins, especially, Integrin α, CD11b, as well as TSG101 and HSP90 proteins. The expected protein sizes are listed in Figure 2B. Accordingly, nEVs (2) have considerably lower levels for almost all examined protein markers, except for HSP90, which is as much abundant in nEVs as in mEVs. The quantification of the expression levels by Compass SW software confirmed these results (Figure 2B). Furthermore, western blot showed a lack of HSP90 and CD11b in aEVs (3), and much lower expression levels of TSG101, and Integrin α compared to mEVs. Nevertheless, the very similar levels of tetraspanin CD9 were detected across all types of EVs. Overall, this suggests that mEVs may have advantage as drug delivery vehicles compared to nEVs and aEVs due to the greater expression of adhesive and targeting membrane proteins. Low, if any expression of EV-specific proteins, were identified in parent cells, except CD11b in macrophages (Supplementary Figure S2).

Figure 2. Expression levels of EV-specific tetraspanins and integrins in EVs released by different cell types.

mEVs or nEVs, or aEVs were isolated from conditioned media of parent cells and subjected to western blot analysis. Representative images obtained by Wes (A) indicate upregulated levels of tetraspanins and integrins in mEVs (1) for most examined proteins compared to nEVs (2) and aEVs (3). Quantification of the expression levels by Compass SW software (B) confirmed this result.

Next, we carried out a practical examination of proteins displayed on membranes of EVs from different cell type origin using a targeted quantitative proteomics analysis. This is a powerful approach used to understand global proteomic dynamics in a cell, tissue, or organism. In this work, we apply this approach to assess the protein profile in EVs of different origin. Specifically, we focused on the relative expression levels of surface adhesive proteins that facilitate EVs binding to cellular membranes, i.e. tetraspanins (CD81, CD63, and CD9), ALIX, Integrin β−1 (CD29), and Tsg101, as well as proteins that accomplish targeting to the inflamed endothelium (Integrins α and β, CD49, CD11, or LFA1, and CD18). Noteworthy, TSG101 and ALIX are known to mostly be present in the EVs lumen. The list of selected peptides from identified proteins of interest and their function is shown on Supplementary Table S2. To assess the expression levels, a sample of 20 μg total protein from each type of EVs was digested and examined by nanoLC-MS/MS with multiple reaction monitoring (MRM) acquisition. The MRM is the collision induced dissociation, fragmentation that occurs on the second quadrupole of the triple quad MS. Since LC-MS/MS is a physical method, independent of the use of antibodies, with MRM monitoring, we were able to detect and measure proteotypic peptides for all ten integrins and tetraspanins that were evaluated.

As revealed by the targeted proteomic analysis, mEVs showed significantly higher content of peptides from integrin and tetraspanin proteins that are known to facilitate adhesion and accumulation of EVs nanocarriers in recipient cells, compared to nEVs and aEVs (Figure 3A). Specifically, in most cases, peptides of TSG101, CD81, CD63, CD9, ALIX, and Integrin β in macrophage-derived EVs showed higher amounts compared to neuron- and astrocyte-derived nanocarriers. Furthermore, the comparative assessment of peptides of the proteins that target inflamed tissues (Integrin α4, Integrin αL, Integrin αM, and Integrin β) were investigated (Figure 3B). This is of importance for the therapy of different neurodegenerative disorders, including PD, as it has been demonstrated that chronic brain inflammation is present in PD patients [41]. The striking contrast in the amount of expressed integrins in mEVs compared to the other EVs types was revealed. In almost all cases, integrins in mEVs were present in significantly greater amount than in nEVs or aEVs (Figure 3B). This confirms western blot data suggesting that mEVs would target inflamed brain tissues and accomplish specific delivery of incorporated therapeutics to the PD brain.

Figure 3. Quantification of Different Integrins in EVs by NanoLC-MS/MS Targeted Proteomic Analysis.

EVs samples from primary macrophages (black bars), neurons (white bars), and astrocytes (grey bars) were digested (n = 2) with trypsin and examined by nano-liquid chromatography linked tandem MS (nanoLC–MS/MS) with multiple reaction monitoring (MRM) acquisition. Expression levels of specific proteins that facilitate adhesion (A) and targeting to the inflamed tissues (B) were assessed. The sample of 20 μg total protein was used, and 0.2 μg was injected. Macrophage-derived EVs showed significantly higher levels most of integrins and tetraspanins compared to neuron- and astrocyte-derived EVs. Values are means ± SD, * p < 0.05, ** p < 0.005.

2.2. Accumulation of EVs from Different Cell Origin in Target Neuronal Cells in vitro

The ability of nanocarriers to deliver the drug payload into target cells is crucial for the therapeutic efficiency of EVs formulations. Herein, we examined whether cell origin can affect their transport into target cells in vitro. EVs derived from macrophages, neurons, and astrocytes were labeled with lipophilic fluorescent dye, DIL, and then incubated with Cath.a neurons for different time points (Figure 4). Remarkably, astrocyte derived EVs showed the greatest accumulation in Cath.a neurons, compared to macrophage- and neuron- derived EVs at all time points examined. We hypothesized that this preferential accumulation of aEVs by neuronal cells reflects the mechanisms that govern intercellular communications between astrocytes and neurons in the brain. Furthermore, mEVs uptake by neuronal cells was significantly greater than nEVs. We suggested that the abundance of integrins and tetraspanins (Figure 3A) in mEVs resulted in more efficient adhesion and internalization of these carriers in target neuronal cells along with achieving the saturation at earlier time points. Of note, in vitro accumulation studies were carried out in the absence of inflammatory conditions, therefore the role of specific proteins that facilitate targeting to the inflamed tissues should be diminished in these settings.

Figure 4. Accumulation of EVs from different origin in Cath.a cells in vitro.

EVs were isolated from conditioned media of parent cells (macrophages, neurons, or astrocytes), and labeled with a fluorescent dye, DIL. Cath.a neuronal cells were incubated with DIL-EVs (1×10⁹ particles/well) for different time points, then washed with ice-cold PBS, and supplemented with Triton X100. Fluorescent levels were measured and adjusted for the number of cells. Significantly greater accumulation of aEVs was recorded in neuronal cells compared to mEVs and nEVs. N = 6, *p < 0.005, **p < 0.0005.

Biodistribution of EVs from different cell origin in Parkin Q311(X)A mice by bioluminescence imaging (IVIS)

To study the ability of EVs to deliver incorporated therapeutics to the brain in vivo, we examined brain accumulation of macrophage- neuron- and astrocyte-derived EVs in PD mouse model, Parkin Q311(X)A mice, upon systemic administration by IVIS (Figure 5). To visualize EVs nanocarriers, their lipid membranes were labeled with DIR, and confirmed equal labeling efficiencies between the three EVs preparations. Considerable levels of DIR-EVs in the brain were recorded at 4h – 480h time frame for all three types of EVs (Figure 5A). The quantification of fluorescent levels of DIR-EVs signals in the brain area of PD mice was assessed by Aura software (Figure 5B). As expected, signals of DIR-EVs released by macrophages (DIR-mEVs) in the PD mouse brain were significantly greater (p < 0.05) than those released by neurons or astrocytes (DIR-n-EVs and DIR-aEVs) throughout the entire observation period. Of note, kinetics of brain accumulation profiles for nEVS and aEVs were almost identical and considerably lower than mEVs (Figure 5B). At the endpoint (20 days), mice were sacrificed, perfused to eliminate blood content, and the main organs were imaged by IVIS (Figure 5C). The quantification of fluorescence levels at necropsy suggested that the fluorescence signal of DIR-mEVs in the brain was significantly higher even 20 days after administration than for DIR-nEVs or DIR-aEVs (Figure 5D, insert). Next, as expected, all types EVs were mostly accumulated in liver and spleen; although, aEVs and nEVs were accumulated at greater extent in these peripheral organs than mEVs (Figure 5D). The supine images of PD mice injected with DIR-EVs of different origin confirmed higher fluorescent signals in main excretion organs, liver, spleen, and kidney for aEVs and nEVs compared to mEVs at all time points (Supplementary Figure S3). These results suggest that systemically administered EVs released by macrophages accumulate in the inflamed brain of transgenic Parkin Q311(X)A mice at greater levels than neuron- or astrocyte-derived EVs.

Figure 5. Biodistribution of DIR-EVs released by different cells in Parkin Q311(X)A mice by IVIS.

PD mice (12 months old) were injected with DIR-EVs released by primary macrophages or neurons or astrocytes through i.v. route (5×10¹¹ particles/200 μL), and imaged up to 480 h. Brain representative images show prolonged DIR signal accumulation for all types of EVs, with higher levels for mEVs (A). At the endpoint (480 h), mice were sacrificed, perfused, and main organs (i.e., liver (1), lungs (2), spleen (3), kidney (4), and brain (5)) were imaged by IVIS (C), and quantification of DIR‐EVs signals in the organs of PD mice was assessed by Aura software. Quantification of DIR-EVs distribution in the brain at various points (B), and at necropsy at 480h (D) shows the highest brain signals in mice injected with mEVs compared to those injected with nEVs or aEVs. The highest DIR-EVs signals were recorded in the liver and spleen in mice injected with nEVs (D). The amount of mEVs carriers in the brain of PD mice were significantly greater compared to those injected with nEVs, or aEVs (D, insert). *p < 0.05, n = 5.

3. Conclusions

Due to their unique characteristics, EVs are increasingly being evaluated as drug delivery vehicles. Although in some cases, EVs can function as moieties that eliminate unwanted proteins from a parent cell, the most vital role is their ability to reach distal organs and tissues [42]. Despite numerous investigations, the EVs biology is still in its infancy with rapidly growing interest. EVs are phospholipid bilayer-enclosed vesicles, thus their membranotropic nature suggests that these drug carriers will be able efficiently interact with the target cells and deliver their therapeutic payload. One of the major obstacles of conventional synthetic nanocarriers is their inability to efficiently cross biological barriers, and specifically the blood brain barrier. In contrast, several reports published by us and others indicate that EVs have the ability to cross the BBB and deliver therapeutics to disease sites [30–32, 43–45]. Herein, we explored the possibility of using EVs of different cell origin as drug delivery vehicles for the treatment of neurodegenerative disorders, in particular for PD. We selected three types of EVs producer cells, primary macrophages, neurons, and astrocytes, collected EVs from the conditioned media, and examined their ability to reach inflamed brain tissues in transgenic model of PD, Parkin Q311(X)A mice. Macrophage derived EVs were chosen due to their ability to target inflamed tissues as was reported earlier [30–32, 38]. Neuron- and astrocyte-derived EVs were selected assuming that they may target neural cells due to a well-documented intercellular communication between astrocytes and neurons in the brain [46–48]. Of note, primary mouse cells were utilized in these studies to eliminate impact of the host immune system.

Regarding their ability to deliver therapeutics, EVs are known to enter the target cells through endocytosis, by the cell through endocytosis (clathrin- and caveolae-dependent, and -independent mechanisms), as well as lipid-raft-mediated internalization [49]. In addition, EVs can be internalized through phagocytosis, and micropinocytosis in phagocytes. Therefore, the size, surface charge and moiety can influence the interaction of these nanocarriers with target cells and blood components thus improving cellular uptake and penetration. Based on the NTA, DLS, and microscopy studies (AFM and CryoTEM), all three types examined EVs have relatively similar size, surface charge, and morphology, suggesting that at least from this standpoint, they can be internalized by all mentioned above mechanisms equally.

Next, it was reported that EVs attach to recipient cell surface by receptor-ligand interactions [50]. Thus, adhesion molecules, such as integrins and ICAMs are involved in the binding and internalization of EVs to the target cells [51–53]. Furthermore, specific molecules expressed on the EVs surface can facilitate targeting these nanocarriers to the inflamed tissues and increase their accumulation at the disease site. In particular, we reported that EVs uptake by activated endothelial cells in the tissues with inflammation is mediated by ICAM-1/LFA1 interactions [38]. Therefore, in order to take advantage of these attributes, we investigated expression of specific adhesion molecules on EVs of different origin using western blot and nanoLC-MS/MS targeted quantitative proteomics. In particular, different tetraspanins and integrins are known to be involved in EVs binding to the target cells and this may affect their internalization and the delivery of incorporated drugs. A quantitative analysis of western blot images indicated that mEVs showed significantly higher levels of adhesive proteins, compared to nEVs and aEVs. The nanoLC-MS/MS targeted proteomics data confirmed these results, and because of the high specificity and sensitivity of this method, it was possible to measure relative amounts of proteotypic peptides in all ten proteins studied, which was not feasible with western blot due to the limitation of specific antibodies. Indeed, finding a method for identification a “right” peptide is crucial for targeted proteomics approach. Unfortunately, at this point, we cannot provide a solution for this problem. Nevertheless, the expression of all peptides in mEVs assessed by targeted proteomics was greater than those in aEVs and nEVs. This ability to multiplex and universally measure the relative or absolute expression of many proteins allows targeted quantitative proteomics to be easily extended to additional proteins or to humans and other species, thus enabling further characterization of EVs. Of note, this is the first demonstration of the use of targeted quantitative proteomics to measure relative amounts of integrins in EVs, and this novel application should find expanded use for characterization of EVs, as it has been extensively adopted to measure enzymes and transporters involved in drug metabolism [54, 55]. In most cases, peptides associated with adhesive proteins in mEVs showed significantly greater amounts than those in EVs released by neurons or astrocytes. Moreover, according to the targeted proteomics analysis, integrins that accomplish targeting to inflamed tissues were also enriched in mEVs, while being mostly absent in nEVs and aEVs. Thus, the complementary results of both assays suggest that macrophage derived EVs should have a superior ability to be taken up by target cells and inflamed tissues delivering therapeutics to the disease site. In summary, the facilitated drug delivery by mEVs to the inflamed brain could be accomplished via both processes, specific interaction with receptors overexpressed on the inflamed endothelium, and facilitated adhesion to the membranes of target cells of neurovascular unit.

Next, the ability to accumulate in the target cells was first examined in in vitro studies with Cath.A neurons. Interesting, neurons showed preferential accumulation of astrocyte derived EVs, compared to mEVs and nEVs. We speculate that this effect may reflect active intercellular talk between neurons and astrocytes in the brain. Indeed, the underlying mechanisms of endogenous intracellular trafficking of aEVs and accumulation in recipient neuronal cells remain to be elucidated. Next, mEVs were taken by neuronal cells in significantly greater amounts compared to nEVs. This effect may be attributed to the high expression of integrins and tetraspanins in mEVs that provide efficient adhesion to target cells membranes and internalization in neuronal cells. Noteworthy, in vitro studies were conducted under non-inflamed conditions, therefore, macrophage released EVs would not be actively targeted to these cells. Thus, we reported earlier that upregulation of ICAM-1 receptor, a common process in inflammation, promoted macrophage derived EVs accumulation in brain endothelial cells in vitro [38].

Finally, accumulation of EVs of different origin was examined in PD mice by IVIS. Previously we used PD mouse models with acute inflammation and neurodegeneration induced by lipopolysaccharide (LPS), 6-hydroxydopamine (6-OHDA), or 1-methyl-4-phenyl tetrahydropyridine (MPTP) [30, 32, 56, 57]. However, the limitation of these toxin-induced PD models is that most of them resemble PD at late stages, whereas to model disease at early staged the genetic animal models are more appropriate. In this work, we used Parkin-Q311(X)A mice with mild brain inflammation that represents Turkish early-onset PD directed to DA neurons of the Substantia Nigra pars compacta (SNpc) and ventral tegmentum area (VTA) [58]. This widely used animal model exhibits several key features of the disease in humans such as a slow progressive course of phenotypic manifestations and neurodegeneration, adult-onset degeneration of nigrostriatal dopamine circuitry, motor deficits, and an altered response to L-DOPA treatment dependent on disease stage.

As expected, mEVs were found at significantly greater levels in the brain compared to other EVs. Interesting, kinetic profiles of brain accumulation for nEVs and aEVs were identical at almost all time points, except very late stages, where aEVs performed slightly better than nEVs. This suggests that even mild brain inflammation in Parkin Q311(X)A mice can play a crucial role in targeting and accumulation of mEVs. This is consistent with our earlier findings that neuroinflammation increased brain influx rate and brain accumulation of macrophage derived EVs in mice [38]. Accordingly, more nEVs and aEVs were found in peripheral organs, especially in liver. Nevertheless, aEVs were collected in the brain at later time point (20 days after administration) at higher amounts than nEVs, that were almost the same levels than mEVs. We speculate that mechanisms that underlying crosstalk between astrocytes and neurons might play a crucial role in this prolonged accumulation of astrocyte derived EVs in the brain. To this regard, we reported earlier that macrophage derived EVs were mostly co-localized with neurons, microglia and partially with endothelial cells [30].

Overall, the biology of EVs should be considered in every aspect of the design of this bio-inspired drug delivery system. Based on our investigations, macrophage derived EVs showed superior ability for drug delivery to the inflamed brain upon systemic administration. Indeed, our in vitro studies indicate that astrocyte derived EVs are taken up by neurons in greater quantities compared to macrophage derived EVs. However, the ability to penetrate across biological barriers and target inflamed tissues make macrophage derived EVs more suitable for drug delivery in these conditions. In addition, it is important to keep in mind that macrophage EVs producer cells may be differentiated into pro-inflammatory (M1) or anti-inflammatory (M2) subtypes. The EVs released from them may inherit at least partially their characteristics, although this is a complex phenomenon that needs further investigations. Thus, mEVs possess features that qualify them as a potential avenue for therapy and as a drug delivery system.

4. Experimental Section

Reagents

Lipophilic fluorescent dyes, 1,1’-dioctadecyl-3,3,3’,3’-tetramethylindo-carbocyanine iodide (DIR) and 1,1’-dioctadecyl-3,3,3’,3’-tetramethylindocarbocyanine (DIL), were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Murine macrophage colony-stimulating factor (MCSF) was purchased from Peprotech Inc. (Rocky Hill, NJ, USA). A human liver microsome (HLM) quality control (QC) (Gentest^™ mixed gender pool of 50) was purchased from Corning (Corning, NY, USA). Ultrapure water was obtained from a Picopure® 2 system (Hydro Service and Supplies, Inc. Durham, NC, USA). Trypsin Gold mass spectrometry grade (100 μg; item # V5280) was purchased from Promega (Madison, WI, USA). Cell culture medium and fetal bovine serum (FBS) were purchased from Gibco Life Technologies, (Grand Island, NY, USA). The solid phase extraction (SPE) cartridges for sample clean up were Strata^™-X 33u Polymeric Reversed Phase (10 mg/mL) (part no. 8BS100AAK) purchased from Phenomenex (Torrance, CA, USA). Ammonium bicarbonate, dithiothreitol, β-casein from bovine milk, sodium deoxycholate, iodoacetamide and acetic, formic and trifluoroacetic acids were purchased from Sigma-Aldrich (St. Louis, MO, USA). Acetonitrile (HPLC grade), methanol and flat cap 0.2 mL PCR tubes (catalog # 14230227) were purchased from Fisher Scientific (Pittsburg, PA, USA). All other chemicals were reagent grade.

Cells

Bone marrow-derived cells extracted from murine femurs (wild type littermates of Parkin Q311X(A) female mice, 2 mo. old) were seeded in T75 flask, and cultured for 10 days in the media supplemented with 1000 U/ml macrophage colony-stimulating factor (MCSF) to obtain primary bone-marrow macrophages (BMM) [59]. A few non-differentiated cells that do not attach to the flask was removed by washing. The purity of macrophages culture was determined by flow cytometry using FACSCalibur (BD Biosciences, San Jose, CA), and 95% of cells were found to be CD11b+. Mouse primary cultured neurons were isolated from mouse pups brain as described in [60]. Immortalized primary astrocytes from C57BL/6 mice were cultured in RPMI 1640 media (Sigma-Aldrich) supplemented with 10 % fetal bovine serum (FBS), and 1 % (v/v) of both penicillin and streptomycin. The cells were grown in an incubator with optimal culture conditions of 37 °C and 5% CO₂, and the medium was routinely replaced every 2–3 days.

For in vitro accumulation studies, mouse catecholaminergic CATH.a neurons were purchased from American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in RPMI-1640 medium supplemented with 8% normal horse serum (NHS), 4% FBS, and 1% penicillin-streptomycin. Cultures were maintained in a humidified incubator at 37°C and 5% CO2. CATH.a neurons were differentiated by adding 1 mM of N6,2’-O-dibutyryladenosine 3’,5’-cyclic monophosphate sodium salt (dbcAMP, Sigma-Aldrich) to the culture media every other day for 6–8 days.

EVs Isolation

Conditioned media from parental cells grown on 75T flasks (20 × 10⁶ cells/flask) was collected, and EVs were isolated using gradient centrifugation [52]. In brief, the culture supernatants were cleared of cell debris and large vesicles by sequential centrifugation at 300g for 10 min, 1000g for 20 min, and 10,000g for 30 min, followed by filtration using 0.2 μm syringe filters. Then, the cleared sample was spun at 100,000g for one hour to pellet the EVs, and supernatant was removed. The pelleted EVs (10¹¹ – 10¹² particles/flask) were washed twice with phosphate buffer solution (PBS). To avoid contamination by the FBS-derived EVs, FBS was spun at 100,000g for 12h to deplete EVs before the experiment. The conditioned media was retrieved after 72h. The recovery of EVs for each cell type was estimated by measuring the protein concentration using the Bradford assay and by Nanoparticle Tracking Analysis (NTA). The EVs yield for each cell type was as following: 1.6 × 10⁴ particles/cell for neurons; 1.7 × 10⁴ particles/cell for astrocytes, and 1.7 × 10⁴ particles/cell for macrophages. The calculated ratio of 2.0 × 10¹⁰ particles/μg protein indicated near 0% protein contamination for macrophage- and astrocyte-derived EVs, and 1.86 × 10¹⁰ particles/μg protein (7% contamination) for neuron-derived EVs. EVs suspension was stored in aliquots at −80° C before use.

Characterization of EVs by Nanoparticle Tracking Analysis (NTA), Dynamic Light Scattering (DLS), Cryogenic Transmission Electron Microscopy (cryoTEM), and Atomic Force Microscopy (AFM)

Molecular and structural characterization of EVs from different cell origin was performed according to the updated guidelines of the International Society for Extracellular Vesicles (MISEV2018) [26]. First, to confirm the purity of isolated EVs, we determined the number of particles by using lipophilic fluorescent membrane labeling dye CMDR^™ (cell mask deep red, ThermoFisher Scientific). This measurement provided the information about the vesicles vs. protein aggregates and non-membrane particles. For this purpose, stock EVs solutions were diluted with PBS 1:10 v/v ration and incubated with 1:100 diluted CMDR solution for 1 h at room temperature. Resulting mixtures were analyzed on ZetaView QUATT Nanoparticle Tracking Microscope PMX-420 (Particle Metrix, Germany) in scatter and fluorescent mode to compare particle counts. Measurements were performed at 11 positions and using the following settings: maximum area 1000, minimum area 5, minimum brightness 20, shatter 100, sensitivity 83 for scatter mode and 93 for fluorescent mode. Conditions for membrane labeling are summarized in Supplementary Table S1.

Next, the size, distribution, and number of particles for EVs released by primary macrophages, neurons, or astrocytes were also examined by NTA. For this purpose, EVs were prepared at concentration 0.01 mg/mL, and evaluated using NanoSight 500, Version 2.2 (Wiltshire, United Kingdom). The Zeta potential (ZP) was measured by DLS using the ZetaPlus’ Zeta Potential Analyzer (Brookhaven Instruments, Santa Barbara, CA, USA) equipped with a 35mW solid state laser (658 nm laser) as described in [61, 62]. ZetaView QUATT Nanoparticle Tracking Microscope PMX-420 (Particle Metrix, Germany) were used to trace ZP of individual particles by tracking an electrophoretic movement after electric field application. EVs were diluted to concentration about 1×10⁸ – 5×10⁸ with pre-filtered PBS. Measurements were performed at 11 positions and using the following settings: maximum area 1000, minimum area 5, minimum brightness 20. Next, the morphology of EVs was investigated by AFM. A drop of isolated EVs in 50 mM phosphate buffer, pH 7.4 at concentration 1×10⁹ particles/mL was placed on a glass slide and dried under an argon flow. The AFM imaging was operated as described earlier [63]. For CryoTem studies, prepared using Leica Cryo-plunger and Cu mesh grids with four pre-blot applications, 4 sec each. EVs from different cells origin were placed on grids and imaged using Talos Arctica TEM operating at 200 keV. Images were collected using automated data collection mode and following settings: 45,000X magnification, 3.284 nm pixel size, 5 microns defocus with total dose 30 e/Ų.

Characterization of EVs by Western Blot

The levels of proteins constitutively expressed in EVs from different types parent cells were identified by Western blot analysis, using Wes^™ Simple Western Blot device (ProteinSimple, San Jose, CA, USA). EVs were lysed with 1x RIPA buffer for 30 minutes at room temperature and 200 or 40 mg/mL of protein was denatured and loaded in Wes^™ multi-well plates following manufacturer’s instructions. The expression levels of EV-specific proteins were also examined in parent cells in control experiment. Protein concentrations were determined using BCA kit (Pierce Biotechnology, Rockford, IL, USA). For analysis of CD9, sample’s lysates de-glycosylated using PNGase F PRIME (Bulldog Bio, NZPP050) under non-denaturing conditions at the ration 1:9 v:v for 1 h prior to denaturalization. The protein bands were detected with primary antibodies described in Supplementary Table S3, and secondary Goat Anti-Rabbit HRP Conjugate (ready-to-use reagent, ProteinSimple, San Jose, CA, USA). The protein concentration for all samples were kept the same, 200 μg/well. A quantitative analysis of obtained images was carried out using Compass SW software.

Characterization of EVs by Label Free Targeted Quantitative Proteomics

Samples of EVs released by different types of parental cells were digested with trypsin, extracted with SPE and prepared for nanoLC-MS/MS analysis as described previously [64] with minor modification. For digestion of the EVs, samples (20 μg per digestion) were evaporated to dryness in a ThermoSavant SpeedVac, and to each tube was added 50 mM ammonium bicarbonate (100 μL), 40 mM dithiothreitol (10 μL), 0.5 mg/mL β-casein solution (10 μL) (an indicator that digestion had occurred and a marker of chromatography retention time) and 13.3 μL of 10% sodium deoxycholate (to aid with solubilization and denaturation). The samples were placed in an Isotemp Thermal Mixer (Fisher Scientific) and denatured for 40 min at 60 °C shaking at 500rpm. After cooling, 10 μL of 135 mM iodoacetamide was added and the samples were incubated at room temperature in the dark for 30 min. Ten microliters of a solution containing an Na⁺/K⁺-ATPase (membrane marker) stable isotope labeled (SIL) peptide standard (purchased from JPT Peptide Technologies, Berlin, Germany) in ~20/80 acetonitrile/50 mM ammonium bicarbonate was then added to each sample. Ten microliters of a 0.1 μg/μL solution (in 50 mM acetic acid) of trypsin was then added to give a 1/20 (w/w) trypsin/protein ratio. Samples were vortexed and digested for 20 h at 37 °C shaking at 300 rpm in the Isotemp Thermal Mixer. After digestion, 10 % TFA solution was added, such that the volume added was 10 % of the total volume of the digestion reaction, to stop the reaction. A deoxycholate precipitate formed. Samples were vortexed and centrifuged at 13.4K × g for 5 min to pellet the precipitate. The supernatant was transferred to fresh tubes (Eppendorf Protein LoBind, 0.5 mL) before cleaning up with SPE. The SPE cartridges (10 mg/mL; polymeric reversed phase) were conditioned with 250 μL methanol followed by 250 μL purified water. Samples was added and cartridges were washed with 150 μL of water. The sample was eluted with 60% acetonitrile/40% formic acid (0.1% solution) into Eppendorf 0.5 mL LoBind tubes. The eluate was evaporated and reconstituted in 50 μL of modified mobile phase A solution (water/acetonitrile/formic acid 98/2/0.1, i.e. 2 % acetonitrile). The reconstituted sample was then centrifuged at 13.4K × g for 5 min and the supernatant transferred to deactivated vial inserts (part # WAT094171DV; Waters, Milford, MA) for nanoLCMS/MS analysis.

NanoLC-MS/MS Analysis was performed on a nanoAcquity UPLC® (Waters) coupled to a SCIEX QTRAP 5500 (Framingham, MA) hybrid mass spectrometer equipped with NanoSpray® III source. The QTRAP 5500 was controlled with Analyst 1.5 software (SCIEX) and the nanoAcquity with UPLC® Console. Mobile phase A was 1 % acetonitrile in 0.1 % formic acid. Mobile phase B was acetonitrile (100 %). The injection volume was 0.2 μL (0.08 μg of protein or 0.4 % of each 20 μg sample) which was loaded onto a Symmetry® C18 2G-V/M trap column, 180 μm internal diameter × 20 mm length, 5 μm particle size (Waters part # 186006527). Trapping flow was 15 μL of mobile phase A for 1 min. After elution from the trap column peptides were separated at a flow rate of 1.3 μL/min on a BEH130 C18 column, 150 μm internal diameter × 100 mm, 1.7 μm particle size (Waters part # 186003550). The chromatographic gradient was 100 % of mobile phase A at start, decreasing to 58 % at 24 min, 5% at 24.5 min for 3 min and 100 % at 28 min for 7 min. Total run time was therefore 35 min.

Multiple reaction monitoring (MRM) analysis on the mass spectrometer was conducted in the positive mode with the ion spray voltage set at 4000 V. The GS1 pressure was set to 15 psi, the interface heater temperature to 150 °C and the collision gas flow to Medium. The nanospray was produced using an uncoated PicoTip^™ emitter (20 μm inner diameter, 10 μm tip inner diameter) by New Objective (Woburn, MA). Following selection of approximately twelve murine tetraspanin and integrin proteins, a previously used in silico approach [64–66] was employed to prepare a list of peptides. The in silico approach included NCBI Blast searching for whether peptides were proteotypic and, except in the case of one integrin (Supplementary Table S4), the use of curator-evaluated (Reviewed) protein sequences from the Universal Protein Resource (UniProt). Other criteria for peptide selection included the absence of methionine (M) and peptides that were of suitable hydrophobicity for trapping on the LC during injection to enable separation by the LC gradient. Where appropriate, peptides containing cysteine (C) were also avoided. Peptides were selected, where possible, to enable application to both human and murine proteins. Approximately two to four peptides were chosen per protein and exosome sample digests were tested for the presence of these peptides. MRM transitions (~30 to 50) for the peptides were predicted by Skyline 1.3 software from the University of Washington. The presence of multiple MRMs and the use of SSRCalc (University of Manitoba, Canada) retention predictions both aided in the identification of peptide peaks in chromatograms. When peptides were identified, where possible, collision energies for up to the five highest responding MRMs were optimized by a stepwise process in order that the highest possible signal would be obtained for each peptide. The three highest responding MRMs were then used, summing the areas, in the label free quantitative assessment. If a peptide could not be identified during the method development, it was discarded from the study. The sequences of the peptides that could be identified, and were used in the final quantitative assessment, their corresponding protein UniProt accession numbers and the three highest responding MRMs for each peptide are shown in Supplementary Table S4. Following optimization, the test samples were analyzed as a single batch to minimize variation for the method, which was both label and standard free. When more than one peptide was measurable for a protein, the peptide with the strongest response was employed for comparative assessment of protein expression across sources of EVs. Two represented MRM chromatograms for one of the most abundant integrins, TSG-101, with a strongly responding peptide, as well as a lower abundant integrin, CD63 are shown on Supplementary Figure S4. Finally, representative extracted ion chromatograms for proteotypic tryptic peptides, CD63.TATILDK and tsg101.DLKPVLDSYVFNDGSSR are shown on Supplementary Figures S5, and S6, respectively.

Accumulation of EVs in Neuronal Cells in vitro

EVs of different cell origin (230 μg total protein/mL) were stained with lipophilic dye, DIL (2 μmol). DIL-stained EVs were purified from non-incorporated label gel-filtration chromatography with Sepharose 6 BCL (Sigma-Aldrich). Cath.A neuronal cells were seeded into 96-well plate (50,000 cell/well), cultured for three days, and then incubated with DIL-labeled EVs for 4h and 24h. The concentration of EVs was selected in the preliminary studies with the escalation dose of EVs (Supplementary Figure S7). The cellular uptake of macrophage derived EVs displayed saturation at high EVs concentrations therefore, therefore, we selected a low EVs concentration (1×10⁹ particles/well) for further experiments. Following the incubation, the cells were washed three times with ice-cold PBS, and solubilized in Triton X100 (10%). Fluorescence in each sample was measured by Shimadzu RF5000 fluorescent spectrophotometer (λ_ex = 540 nm, λ_em = 565 nm). The amount of EVs accumulated in neuronal cells was normalized for the total protein content and expressed as a number of EVs per mg of the protein as means ± S.E.M. (n = 6). Cells incubated with the free dye were used as negative control. All EVs were prepared at the same level of fluorescence, and a separate calibration curve was used for each formulation. Calibration curves for each type of EVs are shown on Supplementary Figure S8.

Animals

Two breeding pairs of Parkin Q311X(A) mice were purchased from the Jackson Laboratory (Bar Harbor, ME, USA) twelve weeks of age, and were treated in accordance to the Principles of Animal Care outlined by National Institutes of Health and approved by the Institutional Animal Care and Use Committee of the University of North Carolina at Chapel Hill. Several cohorts of transgenic mice as well as their wild type controls were bred in house. The genotyping of pups was carried by PCR analysis of Parkin Q311X(A) gene was performed for parents, a negative control wild mouse, and several mice generations according to manufacturers protocol as described earlier [67]. All experiments were performed in 12 mo. old Parkin Q311X(A) mice on a C57BL/6 genetic background. Mice were housed in a temperature and humidity-controlled facility on a 12 h light/dark cycle and food and water were provided ad libitum.

EVs Biodistribution Studies in PD Mouse Model by Bioimaging and Infrared Spectroscopy (IVIS)

To track EVs of different origin in PD mice, nanocarriers were labeled with DIR, a lipophilic near-infrared fluorescent cyanine dye (emission peak of 790), according to manufacturer’s protocol. The fluorescence spectrum of this dye allows efficient penetration through the bones and tissues; therefore, DIR is the most appropriate for the imaging in living animals. Briefly, DIR stock solution in ethanol was added to EVs suspension (final concentration 2 μM) and incubated for 20 min at 37 °C. Then, DIR-EV_S were span down at 100,000×g to separate from non-incorporated dye, and further purified on Nap10 column. To reduce fluorescence quenching by fur and autofluorescence from solid diet, Parkin Q311X(A) mice (12 mo old) were shaved and kept on a liquid diet for 48 h prior to the imaging studies. Then, PD mice (N = 5) were injected with DIR-EVs through intravenous (i.v.) route (6 × 10¹¹ particles/200 μL/mouse) intra-tail vein bolus injection, with the use of a restraining device. The dose was optimized in order to inject maximum amount of EVs that would not cause possible occlusion in the blood stream due to the EVs aggregation. For the background autofluorescence assessment, all animals were imaged before the injections in the IVIS 200 Series imaging system (Caliper, Xenogen Co., Life Sciences), and autofluorescence levels were accounted in further investigations. At the end point, animals were sacrificed and perfused as described earlier [68]; main organs were removed, washed, post-fixed in 10% phosphate-buffered paraformaldehyde, and evaluated by Aura software (Spectral Instrument Imaging, Tucson, AZ, USA).

Statistical Analysis

For all experiments, data are presented as the mean ± SEM. Tests for significant differences between the groups in experiments regarding characterization of EVs of different origin were performed using a one-way ANOVA with multiple comparisons (Fisher’s pairwise comparisons) using GraphPad Prism 5.0 (GraphPad software, San Diego, CA, USA). For targeted quantitative proteomics data, the MRM peak area data was processed with MultiQuant 2.0.2 software (SCIEX). Peak areas for the three highest responding MRMs for each peptide were summed and responses between samples were compared for each peptide. The presence of multiple peptides for a protein provided extra confidence that the protein/tetraspanin/integrin was present. SIL or label free standards would be needed to compare abundance of proteins within a sample, notwithstanding the large differences in peptide signal sometimes seen between proteins. T-tests were used to determine whether peptide abundance differences between samples were significant (p < 0.05). For in vivo experiments, data are presented as the mean ± SEM. Tests for significant differences between the groups were performed using a t-test or one-way ANOVA with multiple comparisons (Fisher’s pairwise comparisons) using GraphPad Prism 5.0 (GraphPad software, San Diego, CA, USA) and Microsoft Excel (Microsoft Corp., 2016). If the differences between the groups were significant at the 0.05 level then pairwise tests were conducted using a Bonferroni correction for multiple comparisons. For each outcome measure in each experiment involving groups of N = 5 mice, sample means were tabulated along with their corresponding standard errors: the estimated mean ± one SEM is an approximate 66% confidence interval. The reported p-values have not been adjusted for multiple comparisons. After the main analyses were completed, sensitivity analyses were performed to guide our level of trust in the main results by examining their robustness/fragility. Examples of these auxiliary computations include analysis of residuals, and examination of the impacts of influential observations, questionable data values, and alternative transformations of measurement scales.