Large-scale bioreactor production of extracellular vesicles from mesenchymal stromal cells for treatment of acute radiation syndrome

Abstract

Background: Hematopoietic acute radiation syndrome (H-ARS) occurring after exposure to ionizing radiation damages bone marrow causing cytopenias, increasing susceptibility to infections and death. We and others have shown that cellular therapies like human mesenchymal stromal cells (MSCs), or monocytes/macrophages educated ex-vivo with extracellular vesicles (EVs) from MSCs were effective in a lethal H-ARS mouse model. However, given the complexity of generating cellular therapies and the potential risks of using allogeneic products, development of an "off-the-shelf" cell-free alternative like EVs may have utility in conditions like H-ARS that require rapid deployment of available therapeutics. The purpose of this study was to determine the feasibility of producing MSC-derived EVs at large scale using a bioreactor and assess critical quality control attributes like identity, sterility, and potency in educating monocytes and promoting survival in a lethal H-ARS mouse model.

Methods: EVs were isolated by ultracentrifugation from unprimed and lipopolysaccharide (LPS)-primed MSCs grown at large scale using a hollow fiber bioreactor and compared to a small scale system using flasks. The physical identity of EVs included a time course assessment of particle diameter, yield, protein content and surface marker profile by flow-cytometry. Comparison of the RNA cargo in EVs was determined by RNA-seq. Capacity of EVs to generate exosome educated monocytes (EEMos) was determined by qPCR and flow cytometry, and potency was assessed in vivo using a lethal ARS model with NSG mice.

Results: Physical identity of EVs at both scales were similar but yields by volume were up to 38-fold more using a large-scale bioreactor system. RNA-seq indicated that flask EVs showed upregulated let-7 family and miR-143 micro-RNAs. EEMos educated with LPS-EVs at each scale were similar, showing increased gene expression of IL-6, IDO, FGF-2, IL-7, IL-10, and IL-15 and immunophenotyping consistent with a PD-L1 ^high, CD16 ^low, and CD86 ^low cell surface expression. Treatment with LPS-EVs manufactured at both scales were effective in the ARS model, improving survival and clinical scores through improved hematopoietic recovery. EVs from unprimed MSCs were less effective than LPS-EVs, with flask EVs providing some improved survival while bioreactor EVs provide no survival benefit.

Conclusions: LPS-EVs as an effective treatment for H-ARS can be produced using a scale-up development manufacturing process, representing an attractive off-the-shelf, cell-free therapy.

Keywords: Acute radiation syndrome; Exosomes; Extracellular vesicles; Mesenchymal stromal cells; TLR4.

Fig. 1

Proposed GMP manufacturing platform for MSC-EV production. A Mesenchymal stromal cells (MSCs) isolated from bone marrow (BM) should be characterized and qualified before B making a master cell bank (MCB), C followed by an expansion to generate multiple working cell banks (WCB). D Early expansion in flasks (Passage P0–P1) is followed by E expansion in a closed system bioreactor (P2) in serum free media. F EVs may be isolated directly with differential ultracentrifugation steps or concentrated beforehand using tangential flow filtration (TFF). G Resuspension of the EV pellet followed by sterile filtration (0.22 u) possibly with endotoxin removing capability. H The final EV testing and monitored storage with a consistent quality control (QC) strategy is needed to fulfill regulatory requirements for product release for I in vivo or ex vivo clinical testing. Created with Biorender.com

Fig. 2

EV and LPS-EV particle size, yield, protein content and surface marker profile from flasks versus bioreactor. A Mean and mode (± SEM) of EV particle diameters from multiple flask production runs of EVs from conditioned media collected after 24-h (24H Flask) or after LPS stimulation (24H + LPS Flask) (N = 10 biological replicates) compared to EVs produced in multiple a hollow–fiber bioreactor collected after four 24-h cycles (24H-96H Bioreactor (N = 4 biological replicates) and after 24-h of LPS stimulation (24H + LPS Bioreactor). Overall, the mean and mode particle diameters of EVs or LPS-EVs between production methods were reproducible and not significantly different from each other. B Comparison of mean particle yields per 10 ⁵ cells (± SEM) from conditioned media of multiple flask runs (24H Flask and 24H + LPS Flask), (N = 10 biological replicates) bioreactor runs (24H-96H Bioreactor) (N = 4 biological replicates) and (24H + LPS Bioreactor) or with LPS stimulation and 24H + LPS Bioreactor). There was a significant (t-test) increase (p ≤ 0.05) in yield produced in the bioreactor runs for EVs (24H-96H Bioreactor) compared to the respective flask runs. C Mean protein content (± SEM) of flasks (N = 10 biological replicates) and bioreactor EVs (N = 4 biological replicates) or LPS-EVs based on mg protein / 10¹¹ EV particles. The EVs of the 24H-Flask production runs had significantly more protein/ 10¹¹ particles compared to the 24H-96H-Bioreactor (t-test ** p < 0.005). D Characterization of surface markers (mean (± SEM)) present on EVs (24H Flask) or LPS-EVs produced in flasks from multiple runs from MSC F1 and F2 MSC isolates (N = 2 biological replicates preformed in duplicate) and bioreactor MSC isolate (96H Bioreactor and 24H + LPS Bioreactor) from MSC isolate B as determined by MACSPlex flow cytometry. The EVs were stained with 37 different bead surface marker populations and compared by mean fluorescence intensity. The same set of surface markers were expressed in both EVs and LPS-EVs produced at both scales. However, when the expression levels in flask EVs and flask LPS-EVs were compared by Kruskal–Wallis with a Dunn post-test and several surface markers (CD146, CD29, CD44, MCSP, CD9 and CD49e) were found to be higher in the flask, *p ≤ 0.05, **p ≤ 0.005, ***p ≤ 0.0005

Fig. 3

Comparison of the mi-RNA and m-RNA cargo of EVs produced by flask or bioreactor process. RNA-seq was performed on EV produced in flasks from six different MSC isolates (N = 6 biological replicates, F_1 to F_6) and in the bioreactor EVs (96H) of one MSC isolate B done in triplicate (N = 3, technical replicates B_1 to B_3). A The heatmap profile after applying variance stabilizing transformation (VST) to reduce background of the 100 most abundant messenger-RNAs and B micro-RNA in the flask and bioreactor EVs. C Correlation between the RNA-seq data sets was performed using Pearson correlation analysis, where the correlation coefficient (R) value ranges from 1.0 to 0.0, implying complete to no correlation between data sets D Volcano plot of significantly upregulated and down-regulated mi-RNA and m-RNA in flask EVs compared to bioreactor EVs

Fig. 4

Gene expression of monocytes educated with EV or LPS-EV produced by flask or bioreactor process. Monocytes from 3 isolates were educated with flask (F_1) or bioreactor (B_1) produced EVs or LPS-EVs flask to generate flask EEMos, flask LPS-EEMos, bioreactor EEMos, or bioreactor LPS-EEMos After education, monocytes were collected, RNA isolated and analyzed by RT-PCR for gene expression (N = 3 to 6 biological replicates for each isolate). The fold-change of gene expression (± SEM) normalized to a GAPDH housekeeping gene and compared to untreated control monocytes A IL-6, B IL-8, IDO, FGF2 and C IL-7, IL-10, IL-12, and IL-15. Groups compared by Kruskal–Wallis with a Dunn’s post-test, *p < / = 0.05, ** p < / = 0.005, *** p < / = 0.0005 **** p < / = 0.0001

Fig. 5

Flow cytometric analysis of human monocytes educated with EVs or LPS-EVs produced by flasks or bioreactor. Monocytes from 3 isolates were educated with flask EVs (24H flask or 24H + LPS flask) from the F_1 MSC isolate, or bioreactor produced EVs (96H Bioreactor or 24H + LPS Bioreactor) from the B MSC isolate to generate flask EEMos, bioreactor EEMos, flask LPS-EEMos or bioreactor LPS-EEMos. Cells were analyzed by flow cytometry (N = 3 biological replicates). The percent (%) CD14⁺ cells for each marker (± SEM) is shown A CD86, HLA-DR, PD-L1 and CD163 B CD16, CD73 and CD206. Groups were compared by Kruskal–Wallis with a Dunn post-test *p < / = 0.05, ** p < / = 0.005, *** p < / = 0.0005 **** p < / = 0.0001 between groups is designated by bars as compared to control monocytes

Fig. 6

Treatment with EVs or LPS-EVs from flask or bioreactor in mice after lethal ARS. On day 0, NSG mice received 4 Gy of lethal radiation followed by an i.v. treatment 4 h later with vehicle control (PBS), or 5 × 10 ⁹ of flask-EVs, flask LPS-EVs (24H flask or 24H + LPS flask) from the F_1 MSC isolate, bioreactor-EVs, or bioreactor LPS-EVs (96H Bioreactor or 24H + LPS Bioreactor) from the B MSC isolate. Mice were followed for A overall survival, B clinical scores (percent weight loss, posture, activity, and fur texture) and C percent weight change. The final mean percent weight change and clinical score were carried over after death to allow for comparison by Kruskal–Wallis with a Dunn post-test between groups at a given time point. Results pooled from three separate experiments with 4 to 12 mice/group. *p < .05, ***p ≤ .005, ****p ≤ .0001