Biomanufacturing and lipidomics analysis of extracellular vesicles secreted by human blood vessel organoids in a vertical wheel bioreactor

Abstract

Background

Extracellular vesicles (EVs) derived from human organoids are phospholipid bilayer-bound nanoparticles that carry therapeutic cargo. However, the low yield of EVs remains a critical bottleneck for clinical translation. Vertical-Wheel bioreactors (VWBRs), with unique design features, facilitate the scalable production of EVs secreted by human blood vessel organoids (BVOs) under controlled shear stress, using aggregate- and microcarrier-based culture systems.

Methods

Human induced pluripotent stem cell-derived BVOs cultured as aggregates or on Synthemax II microcarriers within VWBRs (40 and 80 rpm) were compared to static controls. The organoids were characterized by metabolite profiling, flow cytometry, and gene expression of EV biogenesis markers. EVs were characterized by nanoparticle tracking analysis, electron microscopy, and Western blotting. Lipidomics provided insights into EV lipid composition, while functional assays assessed the impact of EVs in a D-galactose-induced senescence model.

Results

VWBR cultures showed more aerobic metabolism and higher expression of EV biogenesis genes compared to the static control. EVs from different conditions were comparable in size, but the yields were significantly higher for microcarrier and dynamic cultures than static aggregates. Lipidomic profiling revealed minimal variation (< 0.36%) in total lipid content; however, distinct differences were identified in lipid chain lengths and saturation levels, affecting key pathways such as sphingolipid and neurotrophin signaling. Human BVO EVs demonstrated the abilities of reducing oxidative stress and increasing cell proliferation in vitro.

Conclusions

Human BVOs differentiated in VWBRs (in particular 40 rpm) produce 2–3 fold higher yield of EVs (per mL) than static control. The bio manufactured EVs from VWBRs have exosomal characteristics and therapeutic cargo, showing functional properties in in vitro assays. This innovative approach establishes VWBRs as a scalable platform for producing functional EVs with defined lipid profiles and therapeutic potential, paving the way for future in vivo studies.

Introduction

Human induced pluripotent stem cells (iPSCs) have revolutionized biomedical research, emerging as a versatile platform for drug discovery, disease modeling, and therapeutic evaluation [1,2,3,4]. With their capacity to differentiate into virtually any cell type, iPSCs have been directed in vivo to generate all major brain cell types, including neurons, neural progenitor cells, microglia, oligodendrocytes, pericytes, and vascular endothelial cells. To further bridge the gap between in vitro studies and the complexity of human biology, three-dimensional (3D) organoids derived from iPSCs have become indispensable. These 3D structures partially recapitulate and mimic the intricate cellular architecture and microenvironment in vivo, surpassing the limitations of traditional 2D cultures. By capturing the dynamic interactions and genetic intricacies of the tissues, iPSC-derived organoids have uncovered groundbreaking insights into the molecular and genetic underpinnings of complex neurological disorders, opening new avenues for understanding and treating these conditions [5,6,7].

Organoid generation is commonly achieved using ultra-low attachment plates, where aggregation occurs due to enhanced cell-cell adhesion in suspension, usually under static conditions. However, static cultures face significant challenges, primarily due to limitations in nutrient diffusion, which make them less ideal to mimic in vivo organ growth. The absence of microcirculation in static cultures leads to insufficient nutrient and oxygen delivery, as well as the accumulation of metabolic waste [8,9,10]. To address these limitations, dynamic 3D culture systems utilizing bioreactors have gained increasing attention. Among the most widely used bioreactors are rotating wall and stirred tank (spinner flask) reactors [11,12,13,14,15]. However, these systems have inherent limitations. Spinner flasks, for example, induce high shear stress due to horizontal rotation, which can create regions of varying shear rates within the reactor, leading to variability in culture conditions [13, 14]. Additionally, a rotating wall bioreactor has been shown to require high speeds for sufficient agitation which can result in high shear rates, and the scalability is limited [11, 12].

Vertical Wheel Bioreactors (VWBRs) have been recently established for stem cell culture that addresses the limitations of commonly used bioreactors [16]. Its novel vertically rotating wheel allows the propeller to account for more than 80% volume of the U-shaped bottom ensuring homogenous shear rates within the system generated from both radial and axial flow [17, 18]. The unique design of the VWBR impeller allows for efficient mixing at very low shear rates (10-fold lower than traditional stirred tank bioreactors) and have been shown to be applicable to scale up to 500 L [17, 18]. The dynamic system provided by the VWBR has been shown to enhance cell production and reduce manufacturing costs when expanding human iPSCs as aggregates [19,20,21]. Along with human iPSC expansion, the VWBR has also been utilized with other cell types for expansion, such as human mesenchymal stem cells (MSCs) and natural killer cells [22, 23]. Differentiation of human iPSCs into pancreatic islets and cerebellar organoids has also been successfully done within the dynamic microenvironment provided by the VWBR and have shown faster differentiation commitment while maintaining high cell quality [24, 25].

Extracellular vesicles (EVs) are nanoscale, phospholipid-bound particles (30–1000 nm) secreted by cells to mediate intercellular communication through the delivery of bioactive cargo, including proteins, nucleic acids, lipids, and growth factors [26]. A key subpopulation of EVs, exosomes (30–200 nm), are formed via endosomal sorting complex required for transport (ESCRT)-dependent and -independent pathways [27]. Due to their role in cellular communication, EVs exhibit tissue-specific homing capabilities, making them promising candidates for targeted therapeutic drug delivery, including in challenging sites such as the brain. The therapeutic efficacy of EVs depends significantly on their parent cell source and administered dose [28]. Notably, EVs derived from iPSCs demonstrate anti-senescent, anti-inflammatory, and anti-apoptotic properties due to their unique cargo composition [29,30,31,32,33]. Similarly, EVs obtained from 3D organoids may carry cargo that more accurately mirrors the physiological and structural complexity of human tissues in vivo [34, 35]. However, a major challenge in leveraging EVs for the clinical treatment of brain-related diseases lies in scaling up EV production to meet therapeutic demands. A comprehensive report of 50 preclinical animal studies highlighted a median EV dose of 2.75 mg of EV protein per kg of body weight per administration. Meeting these clinical dose requirements necessitates further research to increase EV generation, enrichment, and establish novel scalable production methods [36].

Our previous studies evaluated the influence of agitation speed on EV generation from human MSCs grown on microcarriers [22], undifferentiated human iPSCs grown on microcarriers and as aggregates [37], human iPSC-derived forebrain spheroids [16], and retinal organoids [38] in VWBRs. This study seeks to address the following key questions: How does VWBR affect iPSC differentiation to blood vessel organoids (iBVOs)? How does VWBR affect iBVO EV production and lipid composition? And what are the characteristics and therapeutic effects of iBVO EVs? While our previous studies have performed the EV cargo analysis by proteomics and microRNA-sequencing, lipid profiling remains unexplored. To the best of our knowledge, this is the first comprehensive study of the lipid profiles of EVs secreted by 3D human iBVOs cultured in a dynamic VWBR system. By providing novel insights into the lipid composition of iBVO EVs and evaluating their therapeutic potential, this research advances the development of dynamic bioreactor systems for scalable organoid-derived EV production as a cell-free therapeutic platform for treating neurological disorders.

Materials and methods

Human iPSC culture

Human iPSCs (Pluristyx) were thawed from the vial of the Matched Research Grade human iPSC Working Cell Bank (WCB) generated under current good manufacturing practice (cGMP) from Lonza (Walkersville, MD, USA). Human iPSCs were reprogrammed from human CD34 + umbilical cord blood cells using a non-integrating episomal vector reprogramming method (Certificate of Analysis and Lutheran Hospital Institutional Review Board Approval available upon request). Human iPSCs cryogenically preserved in CryoStor CS10 (StemCell Technologies Inc., Vancouver, Canada) were thawed and plated into growth factor reduced Matrigel (BD Biosciences, Franklin Lakes, NJ, USA)-coated Corning CellBIND T-25 cm² and T-75 cm² Cell Culture Flasks (Corning Inc., Corning, NY, USA). Cells were seeded in the presence of Y27632 (10 µM, Sigma-Aldrich, St. Louis, MO, USA) at a density of 15,000 cells/cm² for 3 days as P + 1, and 5,000 cells/cm² for 4 days as P + 2, in mTeSR media (StemCell Technologies Inc.) and cultured under standard condition (5% CO₂, 37 °C) with daily media change. Cells were passaged after incubation for 7 min with Accutase (StemCell Technologies Inc.). Cells were cultured through 2 passages and then harvested for inoculation of the VerticalWheel bioreactors and 3D static control.

iBVO differentiation from human iPSCs

Human iPSCs were harvested for seeding in respective culture conditions. There are five culture conditions: aggregates formed in the VWBR without microcarriers (Ag) at two respective RPMs 40 RPM and 80 RPM (Ag 40 and Ag 80), cells attached to microcarriers that grew in VWBR (MC) at two respective RPMs (MC 40 and MC 80), and aggregates formed in static ultralow attachment 6-well plates without microcarriers (6 W). From Day 0 (D0) to D1, the cells were maintained in mTeSR and after D1 cells were cultured in DMEM/F12 plus 2% B27. Media were changed on D1, D3, D6, D8, and D10 and the growth factors were added according to the protocol (Fig. 1). Cell growth and differentiation was evaluated and D12 spent media were collected for EV isolation.

Fig. 1

Schematics of experimental design. (A) Human iPSC seeding conditions; (B) Differentiation timeline; (C) Cell characterizations; and (D) Extracellular vesicle characterizations. Ag: aggregates grown in the VWBR. MC: cells attached to Matrigel coated microcarriers in the VWBR. 6-well: aggregates grown as static culture in 6-well plates. 40: 40 rpm; 80: 80 rpm. Created by BioRender

Microcarrier-based differentiation

Corning^® Low Concentration Synthemax^® II Microcarriers (360 cm²/gram, Corning Inc.) were added to the Vertical-Wheel bioreactors (PBS Biotech Inc., Camarillo, CA, USA) at a concentration of 20 g/L of working volume. For 100 mL (0.1 L), 2 gram was used, and total surface area was 720 cm². The bioreactor was inoculated with iPSCs at a seeding density of 8 × 10⁴ cells/mL, i.e., 1.12 × 10⁴ cells/cm². The vessels were filled up to the seeding volume with mTeSR media (60 mL for the 0.1 L vessels) in the presence of Y27632 (10 µM). Vessels were maintained at 30 RPM for 24 h for cell attachment and then raised to 40 or 80 RPM. On day one, 70% of the media were exchanged to DMEM/F12 (Gibco Inc., Billings, MT, USA) containing B27 (Gibco Inc.). Media exchange was performed on D3, D6, D8, D10, and the cells were harvested on D12. On D1, D3, and D6, Wnt activator CHIR99021 (6 µM, Sigma-Aldrich, St. Louis, MO, USA) was added into the media; and on D8 and D10, fibroblast growth factor (FGF)2 (20 ng/mL, PeproTech, Inc., Cranbury, NJ, USA) was added.

Aggregate-based differentiation

The vessels were filled up with 100 mL medium and maintained at 40 RPM or 80 RPM. For the aggregates in ultra-low attachment 6-well plates (Corning Inc.), each well was filled with 3 mL media. Cells equivalent to microcarrier-based bioreactors (8 × 10⁴ cells/mL) were seeded. The feeding timeline was the same as microcarrier-based differentiation. Three independent runs were performed for microcarrier and aggregate conditions.

Cell number and metabolite measurements

For each bioreactor, 3 mL of samples were taken out for cell count and metabolite analysis on day 3, 6, 8, 10, and 12. One mL of samples was removed for microscopic observation. The samples were spun at 500 g for 5 min. The supernatant was then removed and frozen for metabolite analysis. The spent media were analyzed with a BioProfile Flex2 (Nova Biomedical, Waltham, MA, USA) analyzer for metabolite concentration. The cells were re-suspended and dissociated by Accutase. The cell suspension was loaded into a Via1-Cassette (ChemoMetec, Bohemia, NY, USA) and put into a NucleoCounter^® NC-200 (ChemoMetec, Bohemia, NY, USA) that determined cell concentration. Each measurement was performed with three replicates.

Cell harvest, media collection, and EV isolation by ExtraPEG ultracentrifugation

For bioreactors, microcarriers and aggregates in the media were collected while maintaining 48 RPM agitation. For static cultures, the aggregates were collected in media. After centrifugation at 500 g for 5 min, the culture supernatants were stored at -80 °C for EV isolation. The pellets were washed with phosphate buffered saline (PBS) and then dissociated in Accutase. The cells were suspended in FreSR-S cryopreservation medium (StemCell Technologies Inc.) and then stored at -80^oC for further analysis. For EV isolation, ultracentrifugation with ExtraPEG method was used. Conditioned media were sequentially spun (500 g for 5 min, 2000 g for 10 min, 10,000 g for 30 min) to remove cell debris, apoptotic bodies, large vesicles, etc. Polyethylene glycol (PEG)-6000 was added to the supernatant to a final ratio of 8% PEG and 0.5 M NaCl and stored for 24 h at 4 °C in order to enrich EVs. The solution was spun at 3,200 g for one hour and the supernatant was discarded. The remaining pellet was suspended in 1 mL PBS and ultracentrifuged at 127,000 g for 70 min at 4 °C to remove residue PEG. The EV pellet was then suspended in PBS using a benchtop shaker at 1500 rpm for 5 min. EVs were diluted to 10⁸–10⁹ particles per mL in EV-free PBS for nanoparticle tracking analysis.

Nanoparticle tracking analysis (NTA)

NTA was performed on the isolated EV samples in triplicate to determine size distribution and particle concentration, on a NanoSight LM10-HS instrument (Malvern Instruments, Malvern, UK) configured with a blue (488 nm) laser and sCMOS camera. For each replicate, three videos of 60 s were acquired with camera shutter speed fixed at 30.00 ms. To ensure accurate and consistent detection of small particles, the camera level was set to 13, and the detection threshold was maintained at three. The laser chamber was cleaned thoroughly with particle-free water between each sample reading. The collected videos were analyzed using NTA software (version 3.4 Build 3.4.003) to obtain the mode and mean size distribution, as well as the concentration of particles per mL of solution. Compared to the mean size, the mode size is usually a more accurate representation because the vesicle aggregates may affect the mean size.

Immunocytochemistry

The cell samples from different culture conditions were thawed and replated onto Matrigel-coated 24-well tissue culture plate and allowed to grow for 12 days. The culture medium was aspirated, and the cells were washed with PBS. Then 4% paraformaldehyde (PFA) was added for 20–30 min at room temperature (RT) for cell fixation. For permeabilization required to detect intracellular markers, the fixed cells were washed with PBS and 100% ethanol (EtOH) was added and incubated for 5 min at RT. The EtOH was then aspirated out and the cells were washed with PBS. For nonspecific binding blocking, the PBS was aspirated, and 5% fetal bovine serum (FBS) in PBS (blocking buffer) was added for 30 min at RT or one hour at 4 °C. Then the blocking buffer was aspirated and the cells were incubated in primary antibody (Supplementary Table S1) diluted in 1% FBS in PBS for 60 min at RT. The samples were washed and then incubated in secondary antibody diluted in 1% FBS in PBS for 60 min at RT in the dark. The cells were washed and then stained with Hoechst 33342 for cell nuclei. The samples were imaged via an epifluorescence microscope with Zeiss Axio Observer 7 with Colibri7 light source and Hammamatsu Flash 4.0 sCMOS camera using a 20 × 0.8NA (Zeiss).

Flow cytometry

All cells from each culture condition were dissociated by Accutase, and the single cells in suspension were fixed with 4% PFA for 15 min The samples were resuspended in PBS at 10⁷ cells per mL. Nine volumes of cold 100% methanol were then added, and the cells were incubated on ice for 15–30 min for permeabilization. The samples were spun at 300 g for 5 min, resuspended in blocking buffer, and incubated at 4 °C for 15 min. Primary antibodies (Supplementary Table S1) prepared in a staining buffer (SB, 1% FBS in PBS) were then added to the cells and incubated at 4 °C for one hour. After incubation and washing, the secondary antibody was added and incubated at 4 °C for 30–60 min in the dark. After incubation, the cells were washed and resuspended in 150 µL of SB. The samples were acquired with a BD FACSCanto™ II flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA) and analyzed against isotype controls using FlowJo software.

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)

Total RNA was isolated using TRIzol following the vendor’s instructions (Invitrogen). Reverse transcription was carried out using 2 µg of total RNA, anchored oligo-dT primers (Operon) and Superscript III (Invitrogen). Primers for specific target genes (Supplementary Table S2) were designed using the software Oligo Explorer 1.2 (Genelink). For normalization of expression levels, β-actin was used as an endogenous control. Using SYBR1 Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA), real-time PCR reactions were performed on an ABI7500 instrument (Applied Biosystems). The amplification reactions were performed as follows: 2 min at 50^oC, 10 min at 95^oC, and 40 cycles of 95^oC for 15 s, 55^oC for 30 s, and 68^oC for 30 s following with a melt curve analysis. The Ct values of the target genes were first normalized to the Ct values of the endogenous control β-actin. The corrected Ct values were then compared to the experimental control. Fold changes in gene expression were calculated using the comparative Ct method: $\:{2}^{{-(\varDelta\:C}_{t\:treatment}-{\varDelta\:C}_{t\:control})}$ to obtain the relative expression levels.

Western blot assay

EV samples were lysed in a radio-immunoprecipitation assay buffer (150 mM sodium chloride, 1.0% Triton X-100, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 50 mM Tris, pH 8) with addition of Halt protease inhibitor cocktail (Fisher Scientific, Hampton, NH, USA). The supernatant was collected, and protein lysate concentration was determined and normalized to the lowest sample concentration. Proteins were separated by 12% Bis-Tris-SDS gels and transferred onto a nitrocellulose membrane (Bio-rad, Hercules, CA, USA) for blocking with 5% nonfat dry milk (w/v) in Tris-buffered saline (10 mM Tris-HCl, pH 7.5, and 150 mM NaCl) with 0.1% Tween 20 (v/v) (TBST) and incubated overnight with primary antibody (Supplementary Table S1) at 4 °C. Next, the membranes were washed and incubated with IR secondary antibodies (LI-COR, Lincoln, NE, USA) for three hours at room temperature. Membranes were washed three times in TBST and imaged using the LI-COR Odyssey (LI-COR) system.

Lipidomics analysis

Quantitative Lipidomics was performed by Metware Biotechnology Inc. The samples were taken out from the − 80∘C refrigerator and thawed on ice. 1 mL of the extraction solvent (MTBE: MeOH = 3:1, v/v) containing internal standard mixture was added to the samples. After whirling the mixture for 15 min, 200 µL of ultrapure water was added. The samples were vortexed for 1 min and centrifuged at 12,000 rpm for 10 min. 500 µL of the upper organic layer was collected and evaporated using a vacuum concentrator. The dry extract was dissolved in 200 µL reconstituted solution (ACN: IPA = 1:1, v/v) prior to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. The data acquisition instruments consisted of Ultra Performance Liquid Chromatography (UPLC) (Nexera LC-40) and tandem mass spectrometry (MS/MS) (Triple Quad 6500). With in-house database, lipids were annotated based on their retention time and ion-pair information from Multiple Reaction Monitoring (MRM) mode. In MRM mode, the first quadrupole screens the precursor ions for target substance and excluded ions of other molecular weights. After ionization induced by the impact chamber, the precursor ions were fragmented, and a characteristic fragment ion was selected through the third quadrupole to exclude the interference of other non-target ions.

In vitro functional assays using a D-galactose induced senescence model

Human blood vessel organoids (iBVO) were derived from human iPSCs. iBVO from 6-well culture condition were dissociated into single cells by accutase and replated in DMEM/F12 + 2% B-27. To develop an in vitro senescence model, iBVO were incubated with 150 mM of D-galactose (Sigma) dissolved in PBS. After 48 h, the iBVO were treated with EVs from the five culture conditions (6-well, Ag 80, Ag 40, MC 80, and MC 40) at a dosage of: 2 × 10⁶ particles/cell to compare with PBS vehicle control. After four days, the iBVO were used for various biochemical assays.

Biochemical assays

Reactive oxygen species (ROS) assay

The cells were washed with PBS and treated with 25 µM carboxy-H₂DCFDA (Invitrogen). After 30 min incubation at 37 °C (protected from light), the cells were washed using PBS three times, and resuspended in PBS. The fluorescence was measured by a microplate reader (BioRad Laboratories). The images were captured using a Zeiss epifluorescence microscope.

CellTrace cell proliferation assay

The cells were washed with PBS and the assay was performed as described in the manufacturer’s protocol (Invitrogen). The CellTrace DMSO stock solution was diluted in PBS and added to the culture. The cells were incubated for 20 min at 37°C. Then the solution was removed and the cells were washed twice with culture medium. The cells were incubated with fresh culture medium for 10 more minutes. The fluorescence was measured by a microplate reader. The images were captured using a Zeiss epifluorescence microscope.

MitoSox green (MSG) mitochondrial activity assay

The cells were washed with PBS and the assay was performed as described in the manufacturer’s protocol (Invitrogen). A 1 mM stock solution of MSG was prepared by dissolving the MSG reagent in 10 µL of anhydrous N, N-Dimethylformamide (DMF). A working solution was prepared by adding 10 µL of 1 mM stock to 10 mL of PBS. The cells were incubated with 100 µL per 96 well plate of working solution and incubated for 30 min at 37 °C and 5% CO₂. The fluorescence was measured by a microplate reader. The images were captured using a Zeiss epifluorescence microscope.

Transmission electron microscopy (TEM)

TEM was performed to confirm the morphology of EVs as shown in our previous publication [39]. Briefly, EV isolates were resuspended in 50–100 µL of sterile filtered PBS. For each sample preparation, intact EVs (15 µL) were dropped onto Parafilm. A carbon coated 400 Hex Mesh Copper grid (Electron Microscopy Sciences, EMS) was positioned using forceps with coating side down on top of each drop for one hour. Grids were washed with sterile filtered PBS three times and then the EV samples were fixed for 10 min in 2% PFA (EMS, EM Grade). After washing, the grids were transferred on top of a 20 µL drop of 2.5% glutaraldehyde (EMS, EM Grade) and incubated for 10 min at room temperature. Grid samples were stained for 10 min with 2% uranyl acetate (EMS grade). Then the samples were embedded for 10 min with 0.13% methyl cellulose and 0.4% uranyl acetate. The coated side of the grids were left to dry before imaging on the Transmission Electron Microscope HT7800 (Hitachi, Japan).

Statistical analysis

Experimental results were expressed as means ± standard deviation (SD). Statistical comparisons were performed by one-way ANOVA and Tukey’s post hoc test for multiple comparisons, and significance was accepted at p < 0.05. For comparisons of two conditions, student’s t-test was performed for the statistical analysis.

Results

Differentiation of blood vessel organoids from human iPSCs in the VWBR

The experimental workflow is depicted in Fig. 1. Blood vessel organoids were differentiated from human iPSCs in VWBRs across three independent experimental runs, with data from Run 2 presented as representative results. Morphology images showed that the aggregates significantly increased in size during differentiation (from ~ 100 μm to ~ 1000 μm in diameter). By Day 12 (D12), the aggregates in the Ag 40 group exhibited larger perimeters and areas compared to those in the 6-well group (~ 3000 μm vs. ~2000 μm). In contrast, the aggregates in the Ag 80 group demonstrated perimeter sizes similar to the 6-well group (~ 2000 μm) (Fig. 2A, B and C). The smaller size for the Ag 80 group than the Ag 40 group is probably due to the influence of the shear stress (~ 0.4 vs. 0.2 dyn/cm² on average). Aggregate merging was observed for the 6-well group, whereas Ag 80 and Ag 40 groups maintained relatively uniform spherical shape (Fig. 2D). Microcarrier-based cultures displayed clumping during cell growth, with the MC 80 group forming larger and more numerous clumps compared to the MC 40 group. pH trends aligned with cell growth, decreasing from 7.5 to 7.0 as cell density increased (Fig. 2E). Cell viability at D12 ranged from 60 to 80%, with no statistically significant differences among the culture conditions, although the MC 80 condition exhibited slightly lower viability compared to aggregate-based conditions (static or dynamic) (Fig. 2F). Growth kinetics demonstrated that dynamic aggregation (Ag 40 and Ag 80) greatly promoted cell growth, reaching 3–4 × 10⁶ cells/mL and 37- to 50-fold cell number increase over 12 days, while microcarrier conditions sustained growth rate similar to static culture, reaching 1–2 × 10⁶ cells/mL and yielding 12–25 fold increase in cell number over 12 days (Fig. 2G).

Fig. 2

3D blood vessel organoid differentiation from human iPSCs on microcarriers and as aggregates in VWBRs. (A) Culture morphology over time. Scale bar of 4x: 500 μm. Scale bar of 10x: 200 μm. For the aggregates: (B) Perimeter size (µm). (C) Area size (µm²). (D) Roundness ratio. (E) pH kinetics. (F) Cell viability. (G) Cell number growth kinetics. (Run 2 data). * indicates p < 0.05. N.S.: not significant

Similar to Run 2, morphological analyses of Runs 1 and 3 demonstrated consistent aggregate growth across both static and dynamic culture conditions (Supplementary Figure S1A and S2A). In Run 1, aggregates in the Ag 80 condition exhibited larger areas and perimeters compared to static cultures. However, in Run 3, significant variability in aggregate size across conditions resulted in no statistical differences (Supplementary Figure S1B, 1 C and S2B, 2 C). For 6-well static cultures, the area, perimeter, and roundness ratio of aggregates were consistent across runs. Notably, the Ag 40 condition showed the greatest variability in area and perimeter sizes, with aggregates in Run 2 being larger than those in Runs 1 and 3, possibly due to the variations of aggregate sampling from VWBR or the sensitivity of the cells to the shear stress. Similar to Run 2, microcarrier cultures in Runs 1 and 3 formed clumps and clusters, with no significant differences in roundness ratios across all conditions or runs (Supplementary Figure S1D and S2D). Regardless of the run, pH values followed trends consistent with cell growth, decreasing with higher cell densities (Supplementary Figure S1E and S2E). Cell viability showed no significant differences among culture conditions, although Run 1 exhibited the lowest viability overall (Supplementary Figure S1F and S2F). In terms of cell number kinetics, Run 1 revealed that Ag 40 achieved the highest cell concentration (~ 1.6–2.0 × 10⁶ cells/mL), while Ag 80 and MC 80 were comparable to static controls (~ 1.2 × 10⁶ cells/mL). MC 40 showed the lowest cell concentration (Supplementary Figure S1G). In Run 3, by day 12, Ag 40 and Ag 80 maintained higher cell concentrations (~ 2 × 10⁶ cells/mL) than static controls (~ 1.2 × 10⁶ cells/mL), whereas MC conditions were comparable to or slightly below static controls (~ 1.0 × 10⁶ cells/mL) (Supplementary Figure S2G). Taken together, the aggregate culture of VWBR had similar or larger size (1.0-1.5 fold) of aggregates than static control and the cell numbers were 1.3-2.0 fold higher across all runs. The MC cultures did not show higher cell number than static control in general, possibly due to the influence of shear stress in VWBR.

Differentiation marker expression was evaluated on Day 24 cells following the thawing and replating of cryopreserved Day 12 cells. Immunocytochemistry analysis confirmed the presence of key iBVO markers across all culture conditions. These markers included CD31, an endothelial cell marker indicative of blood vessel lining, ZO-1, associated with tight junction maintenance, and VE-Cadherin, a later-stage endothelial maturation marker (Fig. 3A). Flow cytometry further quantified the expression of CD31 and ZO-1. Most conditions exhibited moderate levels of CD31 and ZO-1 expression (45.4–50.4%). Notably, the Ag 40 condition demonstrated the highest expression levels, with CD31 and ZO-1 reaching 85.5% and 81.3%, respectively (Fig. 3B). Gene expression analysis via RT-qPCR showed upregulation of ZO-1 in all conditions except MC 40, compared to the 6-well control (Supplementary Figure S3). VEGF expression remained stable across conditions, with a slight downregulation in Ag 80. CD105 showed slight downregulation in MC 80 and MC 40, while Ag 80 exhibited modest upregulation. Gene expression analysis for Ag 40 could not be performed due to insufficient cell samples. Taken together, these findings confirm that VWBR cultures supported the differentiation and expression of essential iBVO markers, highlighting the potential for generating blood vessel organoids.

Fig. 3

Endothelial and vascular differentiation. (A) Immunostaining of Endothelial and Vascular differentiation markers CD31, ZO-1 and VE-Cadherin. (B) Flow Cytometry analysis of differentiation markers CD31 and percentage differentiated of all culture conditions compared to an unstained control. (C) Flow Cytometry analysis of differentiation markers ZO-1 and percentage differentiated of all culture conditions compared to an unstained control. Scale bar: 100 μm

Metabolite analysis of VWBR based differentiation

Metabolite analysis was performed on the spent media of different culture conditions. The results showed that all culture conditions maintained normal K⁺, Ca²⁺, and Na⁺ concentration levels, but the 6-well conditions consistently had higher concentrations of K⁺, Ca²⁺, and Na⁺ (Fig. 4A). PO₂ (partial pressure of oxygen) levels for all culture conditions were high (ambient air pressures, ranging from ~ 170 mmHg to ~ 210 mmHg), but could be attributed to excess gas exchange due to exposure to free air prior to measurement [40]. The 6-well static control had higher values after day 5 than other VWBR conditions. For the PCO₂ levels, the values were lower than the optimal 35–45 mmHg range for typical cell culture, also possibly due to the exposure to free air prior to measurement (Fig. 4B) [40, 41]. The 6-well static control had lower values before day 12 than other VWBR conditions. Both aggregate conditions (Ag 80 and Ag 40) had the highest glucose consumption rate (e.g., lowest glucose concentration after day 8). MC 40 and 6-well group had medium glucose consumption, and MC 80 had the lowest glucose consumption, which is consistent with the pattern of the cell growth kinetics (Fig. 4C). For the lactate production, the 6-well plate static control had the highest values, reaching 2 g/L while the other VWBR conditions had similar lactate concentration lower than 1.5 g/L. When the mole ratio of lactate to glucose was calculated, all culture conditions showed values between 1.0 and 2.5 (Fig. 4C). The 6-well plate static control had the highest values of 2.0-2.5, indicating anaerobic metabolism. All conditions typically increased in ratio over time, indicating more anaerobic metabolism when cell number increased. Ag 40 condition showed the lowest ratio of 1.2–1.3 in general, indicating more aerobic metabolism. All other VWBR conditions had the values of 1.3–1.7. Similarly to glucose consumption, glutamic acid decreased over time for all culture conditions. Ammonium generation was consistent and similar for all culture conditions except the 6 well condition that experienced an increase in generation from D8 onward (Fig. 4D). Glutamine was not able to be measured due to the use of Glutamax. When glucose consumption was normalized to cell number, Ag 40 and Ag 80 had the lower values after day 8 compared to other conditions. Taken together, VWBR cultures, in particular the aggregate controls, exhibited different metabolic pathways compared to static 6-well plate control.

Fig. 4

Metabolite analysis of human iPSC-derived blood vessel organoids in VWBRs. (A) Changes in K⁺, Na⁺, and Ca²⁺ concentrations over time. (B) Changes in PO₂ consumption, and PCO₂ generation. (C) Changes in glucose consumption, lactate generation, and the ratio of lactate generation to glucose consumption. (D) Changes in glutamic acid generation, NH₄⁺ generation, and glucose consumption per million cells

Run 3 was more similar to Run 2 than Run 1 (there was one day power outrage around day 8 in Run 1). Ca²⁺, K⁺, and Na⁺ concentration levels were normal for all runs. The trend of the metabolites was similar with these ions, with the 6 well condition having the highest concentrations while the other VWBR conditions were similar to each other (Supplementary Figure S4 and S5). PO₂ levels for all culture conditions were high (ambient air pressures, ranging from ~ 170 mmHg to ~ 250 mmHg) but were consistent with Run 2. The same trend could be found in the PCO₂ levels which were lower than the optimal 35–45 mmHg range for typical cell culture but consistent to the values found in Run 2 (Supplementary Figure S4 and S5). For glucose metabolism, the lactate to glucose mole/mole ratio was the highest for 6-well plate static control with values of 2–3, indicating the anaerobic metabolism. All other VWBR conditions had values around 1.5 due to more aerobic metabolism. The glutamine metabolism showed the decreased glutamic acid and ammonium generation along the differentiation.

EV biogenesis, glycolysis, and autophagy analysis

Gene expression of cells at D12 was determined for all the conditions. For EV biogenesis in ESCRT-independent pathways, Ag40, MC80, and MC40 groups showed significantly higher expression of Rab7a, by 5-fold, 3-fold, and 2-fold respectively compared to the static control (Fig. 5A). For Rab27b, Ag 40 and Mc 40 conditions showed 2-fold increase compared to the control. Rab7a showed slightly increase for MC80 and MC40. For SMPD2, Ag40, MC80, and MC40 groups showed significantly higher expression, by 2.8-fold, 2-fold, and 2-fold respectively compared to the static control. Similarly, Ag40, MC80, and MC40 groups showed significantly higher expression than the static control by 3-fold for SMPD3. Ag80 condition had significantly lower expression than all the other conditions for all the markers. For ESCRT dependent pathway (Fig. 5B), only MC80 showed higher ALIX expression than the static control. For HRS, Ag40 and MC80 had 2.8-3.0 fold higher expression than the static control, while MC 40 had 1.5 fold higher expression. For STAM1, Ag40, MC80, and MC40 groups showed significantly higher expression than the static control by 4-fold. Similarly, Ag40, MC80 and MC40 had 3–4 fold higher STAM2 expression than the 6-well static group. Somehow, Ag 80 showed lower expression than all the other conditions for all the markers. This may be related to the mRNA sample preparation or the variations of the collected cell samples of Ag80 group due to aggregate heterogeneity. Taken together, Ag 40 and MC groups of VWBR cultures promoted the expression of EV biogenesis genes.

Fig. 5

EV biogenesis, glycolysis, and autophagy pathway analysis by RT-qPCR. Comparison of mRNA expression levels at D12. (A) ESCRT independent genes. (B) ESCRT dependent genes. (C) Glycolytic genes. (D) Autophagy genes. ESCRT: endosomal sorting complex required for transport. N = 3. * indicates p < 0.05

The expression of glycolytic activity genes was determined. Ag40, MC80 and MC40 had ~ 4-fold higher HK2 expression and 2-fold higher PDK1 than the 6-well static group (Fig. 5C). MC80 and MC40 had slightly higher (1.5-2 fold) LDHA expression and PKM2 was slightly higher for MC40 than the 6-well group. The Ag 80 condition showed low expression for all the markers again (Fig. 5C). For cellular autophagy, AMPK (1.5–2.2 fold), ATG5 (3–4 fold), and ATG16L1 (2–3 fold) were significantly higher for Ag40, MC80 and MC40 compared to the static control (Fig. 5D). Ag40 was slightly higher for BECN1, and MC80 was slightly higher for LAMP1, For TFEB, Ag40 was significantly higher (> 100-fold) and MC80 was 2-fold higher than static control. The upregulation of glycolytic genes and the genes responsible for regulation of autophagy was consistent with our previous study for hMSC EV production in VWBR [22].

EV isolation and characterization

EVs were isolated from D12 media of all culture conditions. The EV size and concentrations were measured by NTA, and the representative EV size distributions from Run 2 are shown in Fig. 6A. The mean sizes for all conditions were around 250–400 nm (Fig. 6B). The mode sizes were around 200 nm, with Ag40 and MC80 having lower value of around 180 nm (in exosome size range). The static 6-well culture condition showed a slightly larger mean size (~ 230 nm) as well as and the MC 40 condition (~ 270 nm). The mode size is a more accurate representation because aggregation may affect the mean size. EV yields were calculated by normalizing EV numbers to spent media and cell number (Fig. 6C). The 6-well condition had the lowest yield when normalized by volume of spent media. Ag 40, Ag 80, and MC40 had 2–3 fold higher EVs than the 6-well plate control. MC80 had ~ 5 fold higher EV numbers possibly due to easy EV release from cells and higher shear stress. When normalized per million cells, both microcarrier conditions exhibited significantly higher yields (4.7-fold of the 6-well for MC 80; 2-fold of the 6-well for MC 40). Static and dynamic aggregate conditions (6-Well, Ag 80 and Ag 40) showed similar yields, possibly due to EV retention inside the ECM network of aggregates. TEM images showed that these EVs presented doubled-layered cup-shaped morphology for all the conditions (Fig. 6D). Only the low RPM groups (the preferred conditions) and the static condition were used for confirmation of exosomal markers (Fig. 6E and Supplementary Figure S6). Calnexin (the negative exosome marker) was enriched in cell lysates, but not shown in the EVs. HSC70, TSG101, and CD81 are positive exosome markers that were shown in the EVs from all three culture conditions.

Similar to Run 2, the representative size distributions of Run 1 and Run 3 indicated that the majority of EVs was within the range of exosome size (~ 200 nm) (Supplementary Figure S7A, S7B and S8A, S8B). Similar mode sizes between culture conditions and between runs were observed. Yield calculation with EVs per mL spent media show that consistently, across all runs that the 6-well condition yielded the lowest number of EVs. The VWBR conditions showed 2–6 fold higher EVs per mL of spent media than that of the static 6-well culture (Supplementary Figure S7C and S8C). When yield was normalized by EVs per million cells, similar or higher yields were observed for Ag 40 and Ag 80, while the MC 80 and MC40 were always higher than the 6-well condition (Supplementary Figure S7C and S8C). When the results of all the three runs were combined, the EV mode size was similar for all the conditions (Fig. 6F). For EV number per mL spent media, which reflects the total EV production, the VWBR conditions secreted 2–3 fold higher EVs than static 6-well control. When normalized to cell number, which reflect EV release mechanism, the MC 80 and MC40 had 2–3 fold higher EVs than static 6-well control and slightly higher yields were observed for Ag 40 and Ag 80. Taken together, VWBRs significantly promoted total EV secretion.

Fig. 6

EV isolation and characterizations. (A) The representative particle size distribution measured by nanoparticle tracking analysis. (B) EV mean and mode sizes for run 2. (C) EV particle numbers normalized to media volume, (i) EVs per mL media; and to cell number, (ii) EVs per million cells for run 2. (D) TEM images, scale bar: 100 nm. (E) Western blot bands of exosomal markers. (F) Data combined from three runs: (i) EV size, (ii) EVs/mL media, and EVs/million cells. N = 3 independent runs, * indicates p < 0.05. N.S. not significant

Lipidomics analysis of EV lipid composition

Lipidomic analysis reveals the lipid composition of the EVs from 6-well, Ag 40, and MC 40 conditions and the results showed that lipid subclasses were all comparable, with content varying less than 0.3% between each culture condition (Fig. 7A). A radar plot further visualized this while showing that the lipid subclasses most prevalent in content are MG (monoglycerides), FFA (Free Fatty Acids), DG (Diglycerides), PC (phosphatidylcholine), PMeoH (phosphatidylmethanol) and, Cer-NS (ceramide-non-substituted lipids) (Fig. 7B). All samples analyzed had similar lipid subclass compositions and similar total lipid molecule content (Fig. 7C). Within certain subclasses there are statistically significant differences within different conditions, e.g., chain unsaturation. Chain unsaturation is the number of double bonds in the fatty acid chain. The abundance of lipid compounds with the same number of unsaturated bonds between sample groups can vary such as in the Cer-NDS group where Ag 40 had the most abundance of Cer-NDS with both zero and one double bond, MC 40 as the median, and 6-well with the least abundance (Fig. 7D). Cer-NDS can play a crucial role in various cellular processes, including cell signaling, apoptosis (cell death), and cell differentiation. Saturation can affect substrate specificity, membrane properties, cellular signaling and regulatory mechanisms [42,43,44,45]. Chain length is another property within subclasses that can have statistically significant differences between samples. Chain length can affect permeability of lipid membranes due to membrane thickness, and melting temperatures which also contribute to membrane fluidity [46, 47]. LPC-O is a lipid subclass that had statistically significant chain length differences between samples with the 6-well group being the highest, followed by MC 40, and Ag 40 as the lowest (Fig. 7E). This trend held true for both 16 and 18 length carbon chains of LPC-O but there was no statistical difference in abundance with the 20-carbon chain LPC-O. LPC-O plays roles in membrane structure and integrity, lipid metabolism and cell signaling. The chain length of LPC-O can affect their metabolism via reaction kinetics changing product formation, upregulating or downregulating cellular functions influenced by LPC-O such as proliferation, wound healing, and migration [48, 49]. Heatmap analysis reveals the differences among the three conditions, where each condition had its unique lipid profiles (Fig. 7F).

Fig. 7

Lipidomics analysis of lipid composition of blood vessel organoid-derived EVs. (A) Loop diagrams of lipid subclass composition. In the color codes that display the subclasses of lipids and their percentages of the overall composition, the three numbers correspond to 6 W, Ag 40, and MC 40 respectively. (B) Radar of changes in lipid subclass content. (C) Total Lipid Molecule Content of three culture conditions. (D) Abundance of lipid compounds with the same number of unsaturated bonds between sample groups. (E) Differences between lipids with different chain length. (F) Hierarchical Cluster heatmap of the relative quantification of each lipid

The differentially expressed lipids (DELs) were compared between Ag 40 vs. 6-well and shown in Volcano plot (Fig. 8A). There were 23 upregulated lipids (e.g., 2 carnitine lipids, 12 ceramide lipids, 1 free fatty acid lipid, 2 phosphatidylethanolamine lipids, 2 phosphatidylglycerol lipids, 1 phosphatidylinositol lipid, 1 phosphatidyl-myo-inositol ethyl ester lipid, 1 phosphatidylserine lipid, and 1 sphingomyelin lipid) and 24 downregulated lipids (e.g., 4 carnitine lipids, 3 ceramide lipids, 1 diacylglycerol lipid, 1 hexosylceramide lipid, 9 lysophosphatidylcholine lipids, 1 phosphatidylglycerol lipid, 3 sphingomyelin lipids and 2 triglyceride lipids) in Ag 40. The most upregulated lipids were in the ceramide subclass which regulate cell growth, differentiation, structural integrity and membrane fluidity, as well as apoptosis and inflammation regulation [50,51,52]. The most downregulated lipids were in the lysophosphatidylcholine subclass which can act as a pro-inflammatory mediator for neuroinflammation, promoting tumor progression and endothelial dysfunction [53, 54]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis showed the top 20 significantly different lipids enriched in the corresponding pathways, such as Sphingolipid metabolism and signaling pathway, AGE-RAGE signaling pathway, Necroptosis, and Neurotrophin signaling pathway (Fig. 8A).

The similar differential lipid comparison was performed between MC 40 vs. 6-well. There were 18 upregulated lipids (e.g., 2 carnitine lipids, 3 ceramide lipids, 4 hexosylceramide lipids, 2 phosphatidylcholine lipids, 1 phosphatidylglycerol lipid, 5 sphingomyelin lipids and 1 sphingolipid) and 14 downregulated lipids (e.g., 1 carnitine lipid, 2 ceramide lipids, 1 lysophosphatidylcholine lipid, 2 phosphatidylethanolamine lipids, 2 phosphatidylcholine lipids, 2 phosphatidylethanolamine lipids, 1 phosphatidylinositol lipid, 2 phosphatidylserine lipids, and 1 sulfated hexosyl ceramide lipid) in MC 40. The most upregulated lipids were in the sphingomyelin subclass which have been shown to be essential in preventing loss of synaptic plasticity, cell death and neurodegeneration [55]. Previous studies have shown that sphingomyelin plays key roles in the processing and aggregation of common neuroinflammation agent Aβ utilizing models to mimic exosomes, linking unregulated sphingomyelin to Alzheimer’s formation and proliferation [56,57,58]. There was no subclass of lipids that was significantly downregulated with respect to the other downregulated subclasses. The top 20 significantly different lipids enriched are related to the Sphingolipid metabolism and signaling pathway, Glycine, serine, and threonine metabolism, Linoleic acid metabolism etc. (Fig. 8B).

Fig. 8

Lipidomics analysis of lipid cargo of blood vessel organoid-derived EVs for pathway analysis. (A) KEEG enrichment diagram and volcano plot of differential lipids between the Ag 40 and 6 W culture conditions. (B) KEEG enrichment diagram and volcano plot of differential lipids between the MC 40 and 6 W culture conditions. (C) KEEG enrichment diagram and volcano plot of differential lipids between the MC 40 and Ag 40 culture conditions

The differential lipid comparison was also performed between MC 40 vs. Ag 40. There were 22 upregulated lipids (e.g., 3 carnitine lipids, 6 ceramide lipids, 2 diacylglycerol lipids, 1 hexosylceramide lipid, 4 lysophosphatidylcholine lipids, 1 phosphatidylcholine lipids, 1 phosphatidylethanolamine lipid and 4 sphingomyelinlipids) and 56 downregulated lipids (e.g., 15 ceramide lipids, 1 free fatty acid lipid, 1 hexosylceramide lipid, 1 lysophosphatidylcholine lipid, 2 phosphatidylethanolamine lipids, 1 lysophosphatidylglycerol lipid, 2 phosphatidic acid lipids, 3 phosphatidylcholine lipids, 21 phosphatidylethanolamine lipids, 3 phosphatidylglycerol lipids, 1 phosphatidylinositol lipid, 2 phosphatidylserine, 3 sulfated hexosyl ceramide lipids and 1 sphingomyelin lipid) in MC 40. The subclass with one of the most upregulated lipids was in the lysophosphatidylcholine subclass, which is the key to the integrity of cell membrane and has also been linked to neuronal damage, cell loss, and neurodegenerative diseases such as Alzheimer’s and Parkinson’s disease [59, 60]. The most downregulated lipids were in the phosphatidylethanolamine subclass which is involved in vesicle formation and neurotransmission [61]. The top 20 significantly different lipids enriched are related to the Ether lipid metabolism, Sphingolipid metabolism and signaling pathway, Metabolic pathways, Necroptosis, etc. (Fig. 8C). These findings indicate that there is difference related to Necroptosis and Metabolic pathways between bioreactor cultures and static control. Given the enhanced oxygen and nutrient delivery of dynamic cultures, the static control had higher level of Necroptosis.

Principal component analysis (PCA) plot showed the distinct clusters of Ag 40 EVs while 6 well and MC 40 EVs had wider spread clusters (Supplementary Figure S9A). The correlation network diagrams of DELs displayed the degree of connection of differential lipids between EVs. The dots in the figure represent the various differential lipids, and the size of the dot is related to the Degree of connection. The larger the dot, the greater the Degree of connection, i.e., the more dots (neighbors) connected to it. Red lines represent positive correlations, and blue lines represent negative correlations. Compared to the 6 well EVs, the Ag 40 EVs had more differential lipids, with significantly more negative correlations than that of the MC 40 EVs. Comparing MC 40 EVs to Ag 40 EVs yielded the largest Degree of connections with multiple positive and negative correlations (Supplementary Figure S9B). The differential lipid regulation network diagrams were built using KEGG pathways to create regulatory interaction networks. Red dots represent a metabolic pathway, yellow dots represent enzymes, green dots represent a metabolic pathway of background material, purple dot represents molecule modules, and blue dot represents a chemical interactions reaction (Supplementary Figure S9C). The significant DELs were classified based on pathway annotation and the number of lipids and the proportion of the total lipids are plotted to a bar plot (Supplementary Figure S10). The most differential KEEG pathways among EVs were all different but what stayed consistent between EVs is the high percentage of metabolism pathway lipids that make up total lipid count. Taken together, the differentially expressed lipids in the EVs of the three culture conditions are mainly related to Sphingolipid metabolism and signaling pathway and metabolism regulation.

In vitro functional analysis

An induced in vitro senescence model was used for evaluating EV therapeutic potentials. Human blood vessel organoids (iBVO) derived from iPSCs were incubated for 48 h with 150 mM of senescence inducing agent D-Galactose and then treated with EVs from different culture conditions at a dose of 2 × 10⁶ EVs/cell. ROS assay showed that D-galactose increased ROS when compared to no treatment, indicating successful senescence induction (Fig. 9A). The fluorescence was normalized to the average value of the cells in a senescent state. EV treatment from the MC 80 condition did not affect ROS production, but every other culture condition (6 well, Ag 80, Ag 40 and MC 40) significantly reduced ROS generation. For CellTrace proliferation assay, fluorescence was normalized to the inverse value of the average fluorescence of the cells in a senescent state (Fig. 9B). The results showed that D-galactose decreased cell proliferation when compared to no treatment, indicating successful senescence induction. EV treatment from the MC 80 condition provided no significant change in cell proliferation. Every other culture condition (6 well, Ag 80, Ag 40 and MC 40) significantly increased cell proliferation, indicating the therapeutic effect of the isolated iBVO EVs on senescence. The performance of MC 80 group may be due to the altered EV cargo or different optimal EV dose. Mitochondrial activity assay showed that D-galactose had no significant effect when compared to no treatment, indicating that D-galactose did not affect mitochondrial activity (Fig. 9C). EV treatment regardless of culture condition had no significant effect on mitochondrial activity.

Fig. 9

In vitro functional assays in an iBVO senescence model for EV therapeutic evaluation. (A) (i) ROS reduction relative to cells induced with D-galactose. (ii) Representative ROS images; (B) Relative cell proliferation compared to cells induced with D-galactose. (ii) Representative cell proliferation images; (C) (i) Relative mitochondrial activity compared to cells induced with D-galactose. (ii) Representative mitochondrial activity images. Scale bar: 100 μm. N = 3, *: p < 0.05, **: p < 0.01, ***: p < 0.001. N.S.: not significant

Discussion

VWBR resulted in different metabolism and gene expression compared to static culture

In this study, human blood vessel organoids were derived from iPSCs in dynamic microenvironments utilizing VWBRs. These organoids were derived as aggregates, at a high shear rate (Ag 80) (corresponding to ~ 0.4 dyn/cm² on average) or low shear rate (Ag 40) (corresponding to ~ 0.2 dyn/cm² on average), as well as on microcarriers at a high shear rate (MC 80) or low shear rate (MC 40) [22]. The dynamic VWBR cultures were compared to a static aggregate culture (6-well). All five culture conditions had the same initial seeding density of 8 × 10⁴ cells/mL, but over time the VWBR conditions had similar or better cell growth than the static control, which can be shown via cell counting measurements as well as pH changes over the differentiation period.

Metabolite concentrations, including glucose, lactate, and NH₄⁺ provide further evidence for increased cell growth of dynamic aggregates. The metabolite data of microcarrier culture are more similar to the dynamic aggregate cultures, which showed more aerobic metabolism (with the lactate to glucose ratio at 1.0-1.5) compared to the anaerobic metabolism (with the lactate to glucose ratio at around 2.0) in the static control. This similarity between dynamic aggregate and microcarrier cultures has been shown in our previous study [16]. Chang et al. demonstrated that human forebrain spheroids derived from iPSCs in VWBRs as aggregates or on microcarriers shared more similarities on a metabolic level indicated by the lactate to glucose ratio, being more indicative of aerobic metabolism compared to the static control. It has been shown that VWBRs display increased stem cell expansion in a 3D dynamic environment when compared to static 2D cultures [62, 63]. In dynamic environments, the core of the aggregates suffered less from the poor nutrient, waste and oxygen transfer occurred in static cultures, which reduced cell viability near the center of aggregates [64,65,66,67]. The static culture exhibited lower area and perimeter sizes than that of the Ag 40 condition. Regardless of shear rate however, the aggregate conditions displayed slightly lower cell-number-normalized glucose consumption. This is consistent with studies performed by Fogg et al. [64,65,66,67] and Chang et al. [16], which found that larger size aggregates of MSCs and iPSCs had lower active metabolisms. The different growth rate trends found between this study and Chang et al. could be linked to the differences in differentiation protocol and culture conditions of human iPSCs. Previous research showed that induction of iPSCs to different cell lineages can have an effect on cell growth rate [68, 69].

Regarding differentiation, flow cytometry reveals that most culture conditions had no effect on differentiation efficiency, however the Ag 40 condition nearly doubled detection of key blood vessel markers such as CD31 and ZO-1. Most studies of iPSCs in VWBR focused on expansion rather than differentiation, and those that do primarily focused on aggregates [24, 25]. The utilization of microcarriers in VWBRs in this study was shown not to impair differentiation potential, similar to high shear dynamic aggregate cultures and static control. At the molecular level, RT-qPCR analysis of cellular genes at D12 revealed similar or higher expression of glycolic pathway genes for Ag 40 and Ag 80 compared to the 6-well group, which was consistent with glucose metabolism determined by metabolite analysis. Consistent with our previous study of human MSC EV generation in VWBR, autophagy gene expression was upregulated in bioreactor culture compared to the static control, possibly due to the presence of shear stress [22]. During EV biogenesis, multivesicular bodies (MVBs) can fuse with lysosomes where the contents are degraded in autophagy, or fuse with the plasma membrane where the intraluminal vesicles (ILVs) are secreted from cells as exosomes [70]. For example, ATG5 can inhibit the vacuolar proton pump on MVBs/ILVs and prevent the content degradation, therefore redirecting MVB to the plasma membrane to be secreted as exosomes [71]. Correlatively, ESCRT dependent and independent EV biogenesis genes were upregulated in VWBR compared to the static control. This correlation may indicate that the EV biogenesis is related to cellular metabolism and autophagy.

Consistent with EV biogenesis markers, the secreted EVs had higher yields (2–4 fold higher EVs per mL spent media) in VWBRs across three runs than the static control. When normalized to cell number, EV release was much higher (2–3 fold) in MC conditions than the aggregate conditions. The secreted EVs could be entrapped in the extracellular matrix network inside the aggregates, while the cells attached to the surface of MCs and formed tissue with lower thickness in the MC group that can easily release the EVs. Similar results were observed for undifferentiated human iPSC aggregates vs. MC [37] and forebrain spheroids vs. MC [16] cultures in our previous studies.

Human blood vessel organoid EV cargo analysis by lipidomics

EVs from all the conditions had a mode size of ~ 200 nm, which means the dynamic environment did not significantly change the EV size. Both MC conditions showed the highest particle numbers per cell, 6-well condition showed the least yield, and the Ag conditions showed yields in the middle. When normalizing the EV number to the spent media volume, the VWBR conditions showed 2–3 fold higher EVs than the static control in the three runs. The pattern of EV secretion matched the expression of ESCRT independent and dependent EV biogenesis genes tested by RT-qPCR. This provides further evidence that VWBRs greatly improve EV production from iBVO. Microcarrier cultures generated a higher EV yield when normalized to cell number since the secreted EVs are readily released into the media, compared to EVs secreted from cells within organoids. This is consistent with our previous studies about iPSC-EVs from VWBRs as well as human forebrain spheroid-derived EVs from VWBRs, which had an increase in EV yield with microcarriers than aggregate cultures [16, 37]. In aggregate cultures, the cell-secreted EVs may be retained and sequestered by the extracellular matrix networks inside the organoids, with only the outer layer of the organoids exposed to shear stress. Shear stress has been proposed to cleave EVs once ligand binding creates membrane tethers [72, 73]. It has also been proposed that high intracellular levels of Ca²⁺ are related to EV production in MSCs [74, 75]. Consistently, this study exhibited lower Ca²⁺ concentrations in the dynamic culture media than the static control, indicative of higher calcium flux and intracellular levels, thus promoting EV generation [16]. Similar results were observed for K⁺ and Na⁺ concentrations, showing the higher ion flux and intracellular levels in dynamic culture than the static control.

The shear stress on the blood vessel organoids did not only have an effect on EV generation but may also affect the membrane composition and cargo packaging of the EVs. In this study, EVs from 6-well, Ag 40, and MC 40 groups underwent lipidomic analysis and exhibited a distinct distribution in PCA plot. The Ag 40 condition had the most distinct population while the MC 40 and 6-well populations had large variations and some overlap, indicating that the lipid composition of those two conditions may be more similar to each other than the Ag 40 condition. The heatmap compared the distinct lipids tested in all samples and the upregulated or downregulated lipids were the least for MC 40 versus 6-well control. Regardless of the distinct lipid compositions of the tested samples, the total lipid molecule content remained similar between the samples and lipid compositions varied between the samples by less than 1%. Compared to the control, Ag 40 and MC 40 had lower concentration of unsaturated lipids per subclass, therefore having higher rigidity, and tighter packing of cargo at the sacrifice of membrane fluidity. Compared to the control, Ag 40 and MC 40 also had longer chain lengths of lipids per subclass also increasing EV rigidity. This rigidity could have resulted in harder EV uptake and less of a therapeutic effect when compared to the 6-well control. KEEG enrichment shows sphingolipid metabolism and signaling pathways are the most prevalent mechanisms that differ from the 6-well control. These lipids are key in membrane structure and cellular recognition and provide further evidence for the differences in EV permeability yielded by different microenvironments.

Function of biomanufactured human blood vessel organoid-derived EVs

The therapeutic effect of human blood vessel organoid EVs was evaluated via an in vitro senescence model, in which 1 × 10⁴ cells/well of day 24 blood vessel organoids were incubated with 150 mM senescence-inducing D-Galactose for 48 h and then treated with EVs from different conditions at a dose of 2 × 10⁶ EVs/cell for 24 h. Common indicators of senescence include an increase in oxidative stress (i.e., ROS production), decrease in cellular proliferation, and decrease in mitochondrial activity. This may be due to ROS’s ability to modulate miRNA families such as miR-34a whose downregulation and upregulation have been linked to neurological diseases [76, 77]. All these indicators can be observed in cells in a senescent state and in other neurological diseases such as Parkinson’s disease and Alzheimer’s disease [78,79,80,81].

In this study, the increased ROS production was observed after 48 h D-Galactose induction, indicative of cellular senescence. EVs from most culture conditions significantly decreased ROS to levels lower than prior to senescence induction (6-well, Ag 40, Ag 80, and MC 40). The significant therapeutic effect could be attributed to the high dosage of EVs used, ensuring high cellular uptake. MC 80 EVs had no therapeutic effect on ROS despite the high EV dosage, indicating the cell culture condition had an adverse effect on the EVs. It is unknown whether the high shear rate had an effect on ESCRT dependent and independent packing of miRNAs or other relevant cargo. Ag 80, Ag 40, and MC 40 groups all yielded significant reduction in ROS at similar levels to the static 6-well culture. This suggests that human blood vessel organoid-derived EVs grown on microcarriers are more susceptible to changes in cargo packaging as shear rate is increased. This reduction of ROS is consistent with previous studies utilizing EVs derived from both human MSCs and iPSCs [82,83,84].

A CFSE cell-proliferation assay also confirmed the senescent state of the cells by the reduction in cell proliferation after D-galactose induction. Similar to the ROS assay, MC 80 EVs had no therapeutic effect, suggesting that any cargo packaging discrepancies that affected ROS reduction had an effect on cellular proliferation as well. Ag 80, Ag 40, and MC 40 all yielded a significant increase in cell proliferation higher than that of iBVOs prior to inducing senescence. This increase of cell proliferation due to EV treatment is consistent with previous studies [29, 85].

A mitoSox Green mitochondrial activity assay yielded no significance of EV therapeutics. Previous studies have shown that EVs can affect mitochondrial pathways. For example, Peruzzotti-Jametti et al. showed that neural stem cell-derived EVs transport functional cargo to restore mitochondrial function in target cells as shown by an enrichment of mitochondrial proteins in mtDNA-deficient L929 Rho cells [86]. The results in this study more likely provide evidence that D-galactose induction did not significantly affect the mitochondrial pathways as it did with ROS and cell proliferation. The exact mechanism of D-galactose induced senescence is not fully understood but common theories such as oxidative stress, inflammation and mitochondrial disfunction have all been proposed as mechanisms [87,88,89]. A study by Cao et al. showed that there are distinct differences in mitochondrial activity and glycolysis in D-galactose induced vs. naturally senescent astrocyte cells [87]. These observations indicate that a more complicated senescent model or alternative models (e.g., using Alzheimer’s disease patient-derived samples) may need to be tested [87].

To further understand the influence of cell culture condition on EV therapeutic effects to prevent senescence, further experiments may need to be performed to better understand the effects of shear rate and aggregation on EV cargo packing and composition. An in vitro dosage testing study is also needed to determine the minimal and optimal amount of EVs to yield significant changes in senescence indicators as well as estimate a rate of EV uptake over time. In vivo EV kinetics, clearance and site specificity studies may be needed down the line to determine clinical viability of the EVs produced from VWBRs. Further analysis to provide insights on the scale of production of human blood vessel organoid EVs in dynamic cultures is needed for potential pre-clinical and clinical applications.

Conclusion

This study demonstrated that VWBR conditions generated 2–3 fold higher EVs (per mL) of human blood vessel organoids than static control, possibly due to more aerobic metabolism and upregulation of EV biogenesis genes. Dynamic aggregate differentiation at low shear stress promoted marker expression of human blood vessel organoids. While microcarrier-based differentiation at low shear stress generated more EVs than the aggregates, the aggregate condition results in developmentally more mature population. EV lipidomics revealed that Ag 40 and MC 40 had lower concentration of unsaturated lipids per subclass, therefore having higher rigidity. Sphingolipid metabolism and signaling pathways are the most prevalent mechanisms in Ag 40 and Mc 40 that differ from the 6-well control. Ag-EVs and MC-EVs showed therapeutic effects on in vitro senescence model. To our knowledge, this is the first study to report a systematic and comprehensive analysis of EV lipidomics of human blood vessel organoids differentiated on microcarriers and as aggregates in a dynamic bioreactor (VWBRs). This study provides a scalable biomanufacturing platform with defined lipid compositions for human blood vessel organoid EV production for the applications in drug screening and cell-free therapy in clinical settings.

The integration of dynamic bioreactors with advanced lipidomic profiling in this study paves the way for refining EV-based therapies tailored for precision medicine. The scalable nature of this platform opens the door to producing EVs at clinical-grade quality, potentially transforming the landscape of regenerative medicine and drug delivery systems. With further optimization, this approach could lead to groundbreaking advancements in treating complex diseases such as neurodegenerative disorders and vascular pathologies, marking a significant leap toward personalized, cell-free therapeutic solutions.