Scalable Generation of Nanovesicles from Human-Induced Pluripotent Stem Cells for Cardiac Repair

Extracellular vesicles (EVs) from stem cells have shown significant therapeutic potential to repair injured cardiac tissues and regulate pathological fibrosis. However, scalable generation of stem cells and derived EVs for clinical utility remains a huge technical challenge. Here, we report a rapid size-based extrusion strategy to generate EV-like membranous nanovesicles (NVs) from easily sourced human iPSCs in large quantities (yield 900× natural EVs). NVs isolated using density-gradient separation (buoyant density 1.13 g/mL) are spherical in shape and morphologically intact and readily internalised by human cardiomyocytes, primary cardiac fibroblasts, and endothelial cells. NVs captured the dynamic proteome of parental cells and include pluripotency markers (LIN28A, OCT4) and regulators of cardiac repair processes, including tissue repair (GJA1, HSP20/27/70, HMGB1), wound healing (FLNA, MYH9, ACTC1, ILK), stress response/translation initiation (eIF2S1/S2/S3/B4), hypoxia response (HMOX2, HSP90, GNB1), and extracellular matrix organization (ITGA6, MFGE8, ITGB1). Functionally, NVs significantly promoted tubule formation of endothelial cells (angiogenesis) (p < 0.05) and survival of cardiomyocytes exposed to low oxygen conditions (hypoxia) (p < 0.0001), as well as attenuated TGF-β mediated activation of cardiac fibroblasts (p < 0.0001). Quantitative proteome profiling of target cell proteome following NV treatments revealed upregulation of angiogenic proteins (MFGE8, MYH10, VDAC2) in endothelial cells and pro-survival proteins (CNN2, THBS1, IGF2R) in cardiomyocytes. In contrast, NVs attenuated TGF-β-driven extracellular matrix remodelling capacity in cardiac fibroblasts (ACTN1, COL1A1/2/4A2/12A1, ITGA1/11, THBS1). This study presents a scalable approach to generating functional NVs for cardiac repair.


Introduction
The need for effective strategies to promote cardiac tissue protection and subsequent regeneration in the heart following a major cardiac event (i.e., myocardial injury, ischaemia) remains unmet [1][2][3][4][5][6]. Given the limited regenerative capacity of the human heart, stem cellbased therapies have emerged as a promising strategy for cardiac repair and restoring heart function [7,8]. However, clinical trials using cell-based therapy report a low engraftment rate and are associated with the risk of arrhythmia [9] and teratomas [8]. Importantly, the poor engraftment of stem cells post-transplantation suggests that secreted factors potentially mediated the observed cardiac repair [10,11].
However, scalable generation of EVs (particularly from stem cells) remains an ongoing technical challenge that has limited their clinical utility [29][30][31]. Thus, in recent years, technologies to generate particles that mimic EVs (nanovesicles, NVs) directly from stem cells have garnered therapeutic interest for their high yield and reparative function [31][32][33][34][35][36][37]. For example, NVs have been generated in high quantity from mesenchymal stem cells (MSCs) with demonstrated myocardial protective effects in in vitro (pro-angiogenic, prosurvival) and in vivo (scar size reduction, preservation of cardiac function) settings [38]. However, MSCs are inherently difficult to isolate and maintain in large quantity [39,40]. This limitation can be addressed by employing pluripotent stem cells such as iPSCs that are easy to generate autologously from individual somatic cells (easily sourced from skin, hair, peripheral blood, and bodily fluids such as urine) [41][42][43], have unlimited proliferative capacity and can be maintained in culture long-term [44][45][46][47]. Therefore, iPSCs are proposed as a preferred alternative source for scalable generation of NVs. However, whether iPSC derived NVs have cardiac repair function remain unknown. Here, we reported a reproducible strategy to generate functional NVs from different human iPSCs efficiently and demonstrated their therapeutic potential as a functional surrogate for natural EVs for tissue repair.

Generation and Characterization of Nanovesicles from iPSCs
We next generated NVs by serially extruding iPSCs through three filter membranes of decreasing pore size; this process termed serial extrusion [55] (Figure 2A). The extruded material was then subjected to density gradient separation (DGS), a gold standard for EV purification, whereby majority of proteins were enriched in fraction 5 with floatation density of~1.13 g/mL, which is characteristic floatation density of natural EVs [56]. Protein distribution across density fractions resulted in 350 µg yield (~66% of total protein signal for both iPSC models) ( Figure 2B). Further single particle tracking analysis highlighted striking enrichment/abundance of particles in fraction 5 (~1.6 × 10 15 particles/mL) (Supplementary Figure S4). To ascertain that this fraction contains NVs, we employed cryo-electron microscopy which revealed abundant membranous particles that are spherical in shape and morphologically intact, reminiscent of natural EV morphology ( Figure 2C). We refer to these particles as CERA-NVs or CL2-NVs depending on their parental cell origin. Both cryo-electron microscopy and single nanoparticle tracking analysis revealed these NVs were~110 nm in mean size (ranging~50-300 nm) ( Figure 2D,E). Single nanoparticle tracking analysis also showed that NVs diameter remained constant even after 10 freeze/thaw cycles (F/T, Supplementary Figure S2).
We typically obtained~98 µg of NVs from a 6-well plate iPSC culture at~80% confluency (~1.7 µg of protein/cm 2 ). We next compared the yield of NVs with natural EVs (purified using DGS as previously described [56]) released by the same number of both iPSCs. We found that NV yields were a striking 900-fold higher when compared to natural EV yield ( Figure 2F). Thus, our data show we can generate NVs from iPSCs in large quantities, with consistent short processing time typically requiring less than 4 h. Int. J. Mol. Sci. 2022, 23, x FOR PEER REVIEW 6 of 28

Uptake of iPSC NVs by Different Cardiac Cells
Similar to natural EVs, engineered NVs should also possess the ability to interact with target cells, and potentially be internalized, to deliver cargo and execute functional response. Our NV proteome included proteins such as integrins (ITGA6/B1) and tetraspanins (ADAM10 and CD9), that are present on EV surface membrane [70] and mediate EV uptake [71]. We thus assessed whether NVs can be taken up by two cardiac cell types: cardiomyocytes (human iPSC-derived cardiomyocytes, CMs) and cardiac fibroblasts (primary human cardiac fibroblasts, hCFs). We labelled NVs with lipophilic tracer DiI and subjected labelled NVs to cushion-based centrifugation to separate unbound dye as previously described [56]. Fluorescence microscopy revealed that NVs are readily taken up by recipient cells within 2 h ( Figure 3A,B). Labelled NVs appeared as punctate structures reminiscent of natural EVs that are taken up by target cells ( Figure 3A,B, inset). Confocal microscopy analysis of recipient cells along z-axis revealed that NVs are internalised (Supplementary Figure S6).

Uptake of iPSC NVs by Different Cardiac Cells
Similar to natural EVs, engineered NVs should also possess the ability to interact with target cells, and potentially be internalized, to deliver cargo and execute functional response. Our NV proteome included proteins such as integrins (ITGA6/B1) and tetraspanins (ADAM10 and CD9), that are present on EV surface membrane [70] and mediate EV uptake [71]. We thus assessed whether NVs can be taken up by two cardiac cell types: cardiomyocytes (human iPSC-derived cardiomyocytes, CMs) and cardiac fibroblasts (primary human cardiac fibroblasts, hCFs). We labelled NVs with lipophilic tracer DiI and subjected labelled NVs to cushion-based centrifugation to separate unbound dye as previously described [56]. Fluorescence microscopy revealed that NVs are readily taken up by recipient cells within 2 h ( Figure 3A,B). Labelled NVs appeared as punctate structures reminiscent of natural EVs that are taken up by target cells ( Figure 3A,B, inset). Confocal microscopy analysis of recipient cells along z-axis revealed that NVs are internalised (Supplementary Figure S6).

iPSC NVs Functionally Regulates Cell Survival, Angiogenesis, and Fibroblast Activation In Vitro
Next, given their proteome composition we investigated whether following uptake, these iPSC-derived NVs modulate function of target cells in context of cardiac repair. One of the most prevalent cardiovascular diseases includes ischaemic heart disease, in which the heart is starved of oxygen and nutrients due to an impaired blood supply [72]. Clinical

iPSC NVs Functionally Regulates Cell Survival, Angiogenesis, and Fibroblast Activation In Vitro
Next, given their proteome composition we investigated whether following uptake, these iPSC-derived NVs modulate function of target cells in context of cardiac repair. One of the most prevalent cardiovascular diseases includes ischaemic heart disease, in which the heart is starved of oxygen and nutrients due to an impaired blood supply [72]. Clinical intervention entails reperfusion to restore blood flow. However, reperfusion can paradoxically exacerbate cellular dysfunction and cell death, in a time-dependent manner [73]. Effective therapeutic strategy limiting this damage (termed ischaemia-reperfusion injury) should thus focus on maintaining CM survival and tissue re-vascularisation [6].
We first assessed whether NVs could promote survival of reoxygenated CMs exposed to low oxygen culture condition (4 h of hypoxia) ( Figure 4A). A single dose of NV treatment significantly (p < 0.0001) reduced CM cell death when compared to vehicle treatment ( Figure 4B,C). NVs from both iPSCs displayed similar degree of CM protection, with percentage of cell viability comparable to CMs cultured in normoxia condition. Similarly, NV-treated HUVECs displayed significantly (p < 0.05) enhanced organisation into tube-like structures on a Matrigel matrix, when compared to vehicle treatment ( Figure 4D-F). Cardiac fibrosis is characterized by elevated deposition of extracellular matrix proteins in the cardiac interstitium and contributes to systolic and diastolic dysfunction [74,75]. TGF-β is induced as an upstream regulator of cardiac fibrosis, promoting fibroblast activation (via upregulated expression of alpha (α)-smooth muscle actin (α-SMA)) that assemble into contractile fibres and increases ECM deposition [76]. We show that NVs from both iPSCs significantly (p < 0.0001) attenuated TGF-β mediated expression of α-SMA in human cardiac fibroblasts (hCF) ( Figure 4G-I).
Collectively, our data show that NVs from both iPSCs protect cardiomyocytes and promote endothelial cell tubule formation against simulated ischaemia-reperfusion injury, and attenuate fibroblast activation induced by TGF-β.

NVs Reprogram Target Cell Proteome to Support Cardiac Reparative and Protective Phenotype
We next investigated whether the function of NVs can be further supported by changes in proteome of cells treated with NVs in Figure 4. MS quantified similar number of proteins in target cells following treatment with NVs or vehicle control. In all three functional assays, we found that NV-treated target cell proteomes displayed high correlation coefficient (>0.80) and clustered together, suggesting similar reprogramming capacity of NVs ( Figure 5A). This further supports comparable level of NV-mediated CM survival, endothelial cell tube formation, and TGF-β-mediated α-SMA expression in hCF. We highlight MS-based validation of NV-attenuated protein expression of α-SMA in TGF-β-treated hCF ( Figure 4I, shown by Western blotting) ( Figure 5B).
At a proteome level NV-treated HUVECs, compared to control treatment, showed significantly higher abundance of 33 proteins that included key angiogenesis regulators such as MFGE8 [86][87][88][89], MYH10 [90] as well as the cardiac function regulator VDAC2 [91] ( Figures 5C and S7B). In NV-treated HUVECs, Gene Ontology analysis (Reactome) of these clusters identified a significant enrichment of processes and pathways associated with translation (p = 1.78 × 10 −6 ), and glucose metabolism (p = 6.39 × 10 −4 ) [92,93] (Figure 5D, Supplementary Tables S6 and S8). (HUVEC) tube formation, and TGF-β-mediated human primary cardiac fibroblast (hCF) activation. Protein identification, LFQ (normalised) intensity, and Pearson correlation for cells in response to NV treatment, and vehicle control for each assay (n = 4). (B) MS-based quantitation (LFQ intensity) of ACTC1 (actin alpha cardiac muscle 1) or GAPDH expression in hCFs in response to TGF-β (vehicle, UT) and following NV treatment (CERA-or CL2-NVs). Data represented as mean ± s.e.m. *** p < 0.0005, ns, non-significant. (C) Differential protein abundance (fold change, of LFQ intensity, log 2 ) of selected protein markers for each assay following NV treatment (CERA or CL2), relative to vehicle controls (yellow) (p < 0.05). Proteins selected include pro-survival, pro-angiogenic, and anti-fibrotic associated markers identified in study. (D) Gene Ontology enrichment analysis (ranked, p < 0.05) (Reactome and GO cellular component) of proteins identified differentially enriched for each assay. Hierarchical cluster analysis of NV treatments and vehicle controls for each assay were determined (Supplementary Figure S7, ANOVA, p < 0.05, fold change (FC): 1.5), to reveal distinct clusters of differential protein expression for each assay. Differential protein subsets for each assay (Supplementary Figure S7) were then mapped using Reactome and GO cellular component and plotted as adj. p-val for each category/assay/treatment.
Therefore, we highlight the capacity of single dose treatment of NVs derived from iPSCs to reprogram target cell proteome to support cardiac reparative and protective phenotype.

Discussion
The therapeutic potential of extracellular vesicles (EVs) lies within their multimodal capacity to protect and repair damaged tissues [69,[99][100][101][102][103][104][105][106] but challenges in their large-scale production are a bottleneck to their clinical utility [29,30]. In this study, we report a rapid size-based extrusion strategy to generate large quantities of EV-like membrane-limited nanovesicles (NVs) from human iPSCs of therapeutic potential. While EV isolation takes approximately 22 h from conditioned media collection, and EV purification [107,108], NVs can be obtained from the same number of donor cells within 4 h. Importantly, following density purification, the fraction containing the highest protein and particle yield resulted in a 900-fold higher protein amount of NVs compared to purified EVs. We also showed that the diameter/size of NVs is not affected by consecutive freeze-thaw cycles. Collectively, our study highlights scalability, ease-of-use, high-throughput and shorter production time, selective use of donor cells and economical-all important logistical considerations for EV-based therapies [30,41].
Additional considerations for therapeutic utility include NVs ability to influence target cell function, carry bioactive cargo of interest, and induce a desired functional response [109]. Functionally, we show that these NVs can be taken up by various cell types (cardiomyocytes, cardiac fibroblasts, and endothelial cells). We chose two iPSC cells as a source of NVs because they contain proteins that influence cardiomyocyte survival, regulate fibroblast activation, and promote angiogenesis, several of which were packaged into NVs during extrusion. These include proteins implicated in wound healing (ACTN1, FLNA, MYH9, ACTC1, ILK), stress response/translation initiation (eIF2S1/S2/S3/B4), hypoxia response (HMOX2, HSP90, GNB1), and extracellular matrix organization (ITGA6, MFGE8, ITGB1). Consistent to functional pathway enrichment in NV cargo, NVs also elicit a heterogeneous response in target cells along with their proteome reprogramming supporting the functional response (Figures 3-5). Contrastingly, components not enriched in NVs compared to parental cells included nucleic acid/protein binding proteins (spliceosome/nuclear proteins) consistent with a previous report; why there is this difference in the NV composition warrants further investigation [55].
In fact, several of these proteins such as HSP20/27/70/90, GJA1, ITGA6/B1, HMGB1, and ILK that have been shown to be packaged in natural EVs from stem cells and dictate reparative function (angiogenesis, proliferation, migration, immune modulation, and cell survival) [1][2][3][4][5]68,69]. Naturally secreted EVs versus NVs potentially have different molecular composition; EV biogenesis includes active cargo sorting mechanisms [71,110,111] whereas NV generation involves random cargo sampling from parental cells [55]. Despite these potential cargo differences, our iPSC NVs were able to recapitulate cardiac repair functions attributed to naturally produced EVs from stem cells in the field [12][13][14][15][16]112,113]. Our functional data showed that after hypoxia, NVs upregulate cell survival proteins (STAT4 [77], CNN2 [78], THBS1 [79,80], RHEB [81], IGF2R [82]) and promote cardiomyocyte survival. In the same context, NVs also supported endothelial tube formation and elevated expression of proteins associated with angiogenesis (MFGE8 [86][87][88][89], MYH10 [90]). Furthermore, NVs attenuated TGF-β mediated activation of human cardiac fibroblasts and its consequential expression of pro-fibrotic proteins (ACTN1 [94], CCN2 [76], COL1A1/2/4A2/ 12A1 [74,75], ITGA1/11 [95], THBS1 79,97,98]). Altogether, these results suggest that NVs regulate distinct and key processes of tissue repair that could be applied to the heart in the context of hypoxia/reoxygenation. Current treatment following MI is the restoration of blood flow in the heart (reperfusion). If reperfusion occurs promptly (less than 2 h) after the infarct there is very limited damage [114]. Often, patients are intervened more than 2 h after the MI, which attributes increased tissue damage and heart failure [115]. Following reperfusion, treatments only prevent further ischemic events and have not shown any reparative effects to the infarct zone [116]. Indeed, preservation of tissue and cellular function is imperative to maintain functionality of the heart, with multiple signalling pathways involved that influence survival through stress fibre formation and contractile proteins, anti-apoptotic and cardiomyocyte survival, and promotion of blood vessel growth and development [6]. NV treatment during reperfusion represents a potential combination therapy to induce repair by promoting angiogenesis, cardiomyocyte survival and attenuation of extracellular matrix deposition. Such emerging strategies using NVs could further impact cardiac senescence and aging, delivering stemness factors (including PIM1 [117] or nucleostemin/GLN3 [118]) to influence impairment of myocardial biology and tissue repair following injury. Whether iPSC NVs could deliver such regulatory cargo (GLN3 [118], interestingly present in both iPSC source NVs) that may impact expression levels of markers of stemness and multipotency for reparative function remains to be determined.
Knowledge of the functional cargo that therapeutic EVs deliver could be exploited to engineer loaded NVs via robust incorporation (either during extrusion, i.e., active loading or via sonication/incubation, i.e., passive loading) [119]. Moreover, our extrusion process is also amenable to introducing targeting moieties either as lipid-conjugated or lipid binding protein-conjugated targeting peptides, and hence warrants further investigation [120][121][122][123][124][125]. Some of the key advantages of NVs include their low immunogenicity, intrinsic cell targeting properties, and enhanced stability in circulation which make them attractive in targeted drug therapy [30,126]. Future studies investigating their administration pathways, pharmaco-kinetics/-dynamics, dosing, biodistribution, and clearance, and therapeutic efficacy in pre-clinical models is a prerequisite [30,109,126,127].
Similar to EV-based therapeutics, the choice of donor cells for NV generation is critical. Indeed, EVs from different cell origins contain diverse cargo [110,111,128] (including surface membrane composition [70]) and exert distinct functions [129,130]. Although MSCs are extensively utilised, their isolation remains difficult, time consuming and/or invasive (liposuction, surgical resection, umbilical cord, or bone marrow aspiration) [131][132][133]. In contrast, iPSCs, which can be generated from skin (fibroblasts and keratinocytes) [134,135], extraembryonic tissues [136], peripheral blood [137], cord blood [136], and urine [42,43], have shown similar reparative effects [14,15]. Moreover, iPSCs maintain pluripotency even after >50 passages [44][45][46][47] important for scalability. In this study, as a proof of principle we show that NVs generated from two types of iPSCs (lentiviral [49] and episomal [50]) also display reparative function. We show that although highly similar in their cargo and target cell effect, differences in cell and NV proteome from different iPSC donors may contribute to such observations in cell uptake/internalisation and tissue reparative function. Whether this further impacts the membrane composition of NVs to alter capacity of cell targeting remains to be determined.

Human Induced Pluripotent Stem Cells (hiPSCs)
Human iPS-Foreskin-2 (CL2) cell line [49], kindly provided by James A. Thomson (University of Wisconsin), and CERA007c6 (CERA) [48] iPSC line were maintained on vitronectin-coated plates in TeSR-E8 medium (Stem Cell Technologies, VA, USA) according to the manufacturer's protocol. Briefly, cells were cultured until confluent (7 days), enzyme free reagent (ReLeSR, Stem Cell Technologies) was used to detach cells and cell aggregates were re-seeded into new vitronectin pre-coated plates in a wash-free manner. Media was replaced every 2-3 days. Brightfield images were obtained using an inverted microscope (Olympus IX71, Tokyo, Japan) at 40× magnification.

Cell Proliferation Assay
Cell proliferation was performed by immunostaining with Ki67 antibody. Briefly, cells cultured on coverslips were fixed in 10% neutral buffered formalin and permeabilised with 0.2% triton X-100. After blocking with serum-free blocking solution (Thermo Fisher Scientific) for 10 min, cells were incubated with primary antibody against Ki67 (0.29 µg/mL, rabbit monoclonal IgG; Abcam) antibody and counterstained with 4 6-diamidino-2-phenylindole (DAPI, 1 µg/mL; Thermo Fisher Scientific) for nuclear staining. Images were taken using BX-61 Olympus fluorescence microscope (Tokyo, Japan). The number of proliferative cells (Ki67 positive) was counted from three random fields and expressed as a percentage over total number of cells (DAPI positive).

iPSC Nanovesicle Generation and Purification
Generation of nanovesicles (NVs) was performed as described [140] with modifications. Briefly, adherent iPSCs (CL2 and CERA) were individually harvested (following PBS wash) using a solution of EDTA (10 mM) (3 × 3 min round incubation) and cell suspension spun at 500× g for 5 min. The pellet was resuspended in PBS and the cell suspension sequentially extruded through 10, 5, and 1 µm polycarbonate membranes (19 mm; Advanti Polar lipids, 610010) (13 times across each filter, Whatman). The extruded NVs were subsequently purified using 10% OptiPrep™ (Stemcell Technologies) density cushion (step gradient formed by overlaying extruded sample on 10% and 50% iodixanol) and centrifuged at 100,000× g for 2 h at 4 • C. Seven equal fractions were obtained, diluted in PBS (to 1.5 mL), centrifuged at 100,000× g for 2 h at 4 • C (TLA-55 rotor; Optima MAX-MP ultracentrifuge) and resuspended in PBS and stored at −80 • C until further use. The yield (protein) and density of each fraction was determined as described [56].

Cryo-Electron Microscopy
Cryo-electron microscopy (Tecnai G2 F30) on NVs was performed as described [108,141]. Briefly, NVs from each iPSC model (~1 µg protein, non-frozen samples prepared within 2 days of analysis) were transferred onto glow-discharged C-flat holey carbon grids (ProSciTech Pty Ltd., Kirwan, Australia). Excess liquid was blotted and grids were plungefrozen in liquid ethane. Grids were mounted in a Gatan cryoholder (Gatan, Inc., Warrendale, PA, USA) in liquid nitrogen. Images were acquired at 300 kV using a Tecnai G2 F30 (FEI, Eidhoven, The Netherlands) in low dose mode. Size distribution of particles was calculated for the 12 fields of view (~300 different vesicles for each preparation).

Nanoparticle Tracking Analysis
Particle concentration and size was determined using nanoparticle tracking analysis (NTA, ZetaView, Particle Metrix, PMX-120; 405 nm laser diode) for all density fractions (volume normalized) [108]. Samples were prepared in 1 mL of PBS (14190-144, Thermo Fisher Scientific) and particle diluent analyzed in experimental triplicate; 11 positions were captured with the following parameters: camera sensitivity: 80, min area: 5, max area: 1000, brightness: 30, min trace length: 15, temperature: 25 • C. Calibration beads (Nano FCM, S16M-Exo) were used for instrument setup. Capture was performed at medium video setting, corresponding to 30 frames per position. ZetaView software 8.5.10 was used to analyse acquired data. For freeze-thaw analyses, NVs were placed on dry-ice, thawed and this cycle repeated for 1, 5, and 10× before NTA analysis.

NV Recipient Cell Uptake
Human CMs and CFs were seeded and cultured on gelatin coated glass bottom chamber µ-slides (8 well, Ibidi) to~80% confluency. The medium was supplemented with DiI-labelled NVs (6 µg, concentration of 30 µg/mL) or PBS vehicle control at 37 • C for 2 h to allow uptake. For NV staining, NV was incubated with fluorescent dye DiI (Vybrant™ DiI Cell-Labeling Solution, 1:200 dilution, Invitrogen, Waltham, MA, USA, V22885) at 1 µM concentration for 15 min at 37 • C as described [142]. Briefly, labelled NV and PBS (DiI) were centrifuged at 20,000× g for 50 min on a 10% OptiPrep™ cushion, washed with PBS and resuspended in 50 µL of PBS. Following uptake, cells were washed twice in PBS and fixed using 4% formaldehyde for 5 min. Nuclei were stained for 7 min with Hoechst 33342 stain (Thermo Fisher Scientific) (10 µg/mL) prior to imaging by Nikon A1R confocal microscope equipped with resonant scanner, using a 20× WI (1.2 NA); (Nikon, Tokyo, Japan). Images were sequentially acquired. The XY image resolution was 1024 × 1024 at 0.033 FPS, 4× averaging, 2.4 dwell time. 3D images were taken by Z-stack of approximately 15 µm, 25 steps, at a resolution of 1024 × 1024, 8× averaging 2.4 dwell time. NS studio was used to render images.

In Vitro Model of Hypoxia/Reoxygenation
Human cardiomyocytes (CM) and human endothelial cells (HUVECs) at~80% confluency were placed inside a hypoxia incubator chamber, and purged with nitrogen gas for 5 min The cell chamber was placed inside an incubator at 37 • C for either 4 h (for CMs) or 24 h (for HUVECs). After the hypoxia period, cell conditioned medium was replaced with respective culture media containing NVs (30 µg/mL) or PBS and cultured in 37 • C humidified 5% CO 2 incubator for 24 h (for CMs) or 6 h (for HUVECs) to simulate reoxygenation.

iPSCs and Cardiomyocyte Survival Assay
For iPSC characterisation and CM survival following hypoxia/reoxygenation, cell viability was determined by 3 µg/mL propidium iodide (PI, Thermo Fisher Scientific, Waltham, MA, USA) and 5 µg/mL Hoechst 33258 (Sigma-Aldrich, St. Louis, MO, USA) dual staining (30 min). Images were captured at 200× magnification with an inverted microscope (Olympus IX71). The number of dead cells (PI positive) were counted from three random fields for each replicate for a total of 4 replicates each and expressed as a percentage of total cells (Hoechst 33258 positive).

Tube Formation Assay
Following 24 h of hypoxia, HUVECs were plated either into 96-well µ-plates (ibidi) for microscopy or flat-bottom 96-well plates for MS. Cells were treated with either NVs (single dose, 30 µg/mL) or vehicle control (PBS), and incubated in normoxia conditions (37 • C, 5% CO 2 , humidified incubator) for 6 h. Cells were washed with PBS and bright field images obtained using an inverted microscope (Inverted Confocal Nikon A1r Plus NIR) at 200x magnification. Manual identification of tubes was performed followed by ImageJ quantitation. Results are shown as average number of tubules per condition.

Fibroblast TGF-β-Mediated Activation Assay
hCF cells were seeded at confluency in 24-well plate pre-coated with 1% gelatin and incubated for 24 h at 37 • C, 5% CO 2 to allow the cells to attach. Cells were serum-starved for 24 h (to obtain basal levels of α-SMA; activation) prior to treatment with 5 ng/mL of TGF-β (or non-treatment control) for 24 h (conferring hCF activation/α-SMA expression), followed by treatment with NVs (30 µg/mL; 9 µg) or vehicle control (PBS) for 24 h at 37 • C, 5% CO 2 . Medium was removed, followed by 3 washes with PBS, and cells were immediately lysed in-well with cell lysis buffer (1% SDS, 50 mM HEPES pH 8) on ice for 5 min, heat treated at 95 • C for 5 min prior to protein quantification (microBCA™ Protein Assay Kit, Thermo Scientific, 23235).

Proteomics: Data Processing and Informatics/Visualisation
Identification and quantification of peptides was performed using MaxQuant (v1.6.14.0) with its built-in search engine Andromeda [146], as described [12,143,147]. Human-only (UniProt #78,120 entries) sequence database (March 2021) with a contaminants database was employed. N-terminal acetylation and methionine oxidations were set as variable modifications. False discovery rate (FDR) was 0.01 for protein and peptide levels. Enzyme specificity was set as C-terminal to arginine and lysine using trypsin protease, and a maximum of two missed cleavages allowed. Peptides were identified with an initial precursor mass deviation of up to 7 ppm and a fragment mass deviation of 20 ppm. Protein identification required at least one unique or razor peptide per protein group. Contaminants, and reverse identification were excluded from further data analysis. 'Match between run algorithm' in MaxQuant [148] and label-free protein quantitation (maxLFQ) was performed. All proteins and peptides matching to the reversed database were filtered out.
Perseus [149] and R studio were used to analyse the proteomic data and generate plots. G:Profiler, Reactome, STRING, and Cytoscape (v3.9.1) were used for enrichment analysis. Protein lists for samples were generated in Perseus (v1.6.14.0) [150]. For cell, NV, and NV-treated cell proteomes, proteins were identified at least once in two biological replicates per group. Protein intensities (maxLFQ) were log 2 transformed. Statistical analysis and plots were generated using Perseus, R studio and GraphPad Prism. Principal component analysis, Pearson correlation matrix, and hierarchical clustering was performed in R studio using Euclidian distance and average linkage clustering, with missing values imputed at z-score 0 for heatmap generation only. R was used for data visualisation (ggplot2, ggpubr packages). In all instances significance was p < 0.05 unless otherwise indicated.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/ijms232214334/s1. Author Contributions: J.L. contributed to project coordination, experimental design, NV generation/characterisation, uptake and functional assays, proteomic samples preparation, data generation, and data analysis, bioinformatics analyses, preparation of figures/manuscript drafting, wrote the manuscript and preparation of figures. A.R. contributed to project coordination, experimental design, bioinformatics analyses, analysed functional/proteome data, provided intellectual input, and preparation of figures/manuscript. J.G.L. contributed to iPSC culture/maintenance, cardiomyocyte differentiation, and functional assays. H.F. contributed to proteomic sample preparation and mass spectrometry data generation. B.C. contributed to functional assays. S.Y.L. contributed to iPSC culture/maintenance, cardiomyocyte assay, and reagents for functional assays. D.W.G. contributed to project development and coordination, experimental design, data analysis, wrote the manuscript and preparation of figures, funding, and provided intellectual input. All authors discussed the results and commented on the manuscript. All authors have read and agreed to the published version of the manuscript. Data Availability Statement: Data generated or analysed during this study are included in this published article (and its Supplementary Materials files) or available from Data Repositories. Raw proteome data (cell, NV, and NV-treated cell proteomes) and parameter/search information is available from the ProteomeXchange Consortium via the PRIDE partner repository (#PXD036654, http://www.proteomexchange.org/; accessed 12 September 2022). Functional enrichment annotations were retrieved using g:Profiler (https://biit.cs.ut.ee/gprofiler/; accessed 12 September 2022). Hierarchical clustering was performed in R and Perseus using Euclidian distance and average linkage clustering, with missing values imputed at z-score 0. R was also used for data visualisation.

Conflicts of Interest:
The authors declare no conflict of interest.