Production cell analysis and compound‐based boosting of small extracellular vesicle secretion using a generic and scalable production platform

Extracellular vesicles (EVs) are a novel format of advanced therapeutical medicinal products (ATMPs). They can act regenerative or immune‐modulatory as cell therapy substitutes or as a platform for designer exosomes. The biotechnological production of therapeutic EVs is still very much uncharted territory so standardized host cells, production setups, and isolation methods are not yet implemented. In this work, we present a tangential flow filtration (TFF) and fast‐performance liquid chromatography (FPLC)‐based size exclusion chromatography (SEC) purification setup that is compatible for industry applications. Moreover, we evaluated a series of potential host cell lines regarding their EV productivity, characteristics, and biological functionality. It was found that telomerase‐immortalized Wharton's jelly mesenchymal stromal cells (WJ‐MSC/TERT273) secrete high amounts of EVs per cell with regenerative capabilities. On the other hand, Cevec's amniocyte producer cells® (CAP®) and human embryonic kidney (HEK293) suspension cells are suitable platforms for designer EVs with high yields. Finally, we aimed to boost the EV secretion of HEK293 cells via chemical adjuvants and verified four compounds that heighten cellular EV secretion in a presumably cAMP‐dependent manner. A combination of fenoterol, iodoacetamide, and dinitrophenol increased the EV yield in HEK293 cells threefold and cellular secretion rate fivefold.

Although other extracellular vesicle (EV) subtypes like ectosomes and apoptotic vesicles are, due to their origin, biologically distinct from exosomes it is not possible to strictly distinguish them by size and there is no single, ubiquitous surface marker to vouch for the respective EV type. Therefore, we follow the Minimal Information for Studies of Extracellular Vesicles (MISEV) guidelines from 2018 (Théry et al., 2018) and more precisely call our 30-150 nm samples small EVs (sEVs), not exosomes.
Cancer cells are known to secrete high levels of EVs facilitating immune evasion and metastatic niche formation (Mashouri et al., 2019).
Professionally antigen-presenting cells' exosomes bear the major histocompatibility complex (MHC)-I and -II on their surface and can mediate an immune response (Raposo et al., 1996), making those EVs a promising tool for, for example, cancer immunotherapy (P. Yang et al., 2021). Mesenchymal stromal cell (MSC) EVs are especially interesting (Shekari et al., 2021). Known to be potent in tissue regeneration processes, it has been shown that MSC exosomes facilitate the same effects even without the host cells (Phinney & Pittenger, 2017).
On the other hand, exosomes can serve as a platform for designer EVs (Jafari et al., 2020). Modifications of therapeutic cargo and tropism-determining molecules allow for applications ranging from gene therapy (Duan et al., 2021) and vaccination (Sabanovic et al., 2021) to tailored treatments of diseases like, for example, Parkinson's disease (Kojima et al., 2018).
For industrial production of EVs, product definition and low yields remain major challenges. Also, there is no gold-standard isolation method and most established protocols lack the scalability for preparative purposes beyond lab scale. Thus, we aimed for a scalable and reproducible method that fits with the requirements of pharmaceutical industry.
As indicated above, the host cell is defining for their respective sEV characteristics but also secretion rate. Depending on the application, different cells could be relevant as platforms for therapeutic sEV production. To shed light on their sEVs characteristics and production capabilities we investigated several cell types in this study. We chose WJ-MSC/TERT273 cells because of their therapeutic potential and HEK293 and CAP ® suspension cells as generic platform for production of designer sEVs. Suspension HEK293 cells (HEK293sus) were compared to adherent growing HEK293 (HEK293adh) cells to evaluate the impact of different cultivation systems. Furthermore, SKOV3 cancer cells were included as a possible positive control for a high secretion rate of sEVs and CHO-K1 sEVs were characterized, too, due to their importance as a host cell in biotechnological production processes.
Finally, we addressed the issue of low sEV yields by adding a series of compounds to cultivation media. Most biomedical literature aims for a reduction of exosome secretion in tumors. In a large study trying to identify inhibitors of exosome secretion in cancer, Datta et al. (2018) found a number of activators among 4580 tested substances in pancreatic cancer cells. Some of these compounds could later be confirmed as boosters for exosome secretion also in MSCs . Ludwig, Yerneni, et al. (2020) worked with prostate cancer cells and found that an inhibitive treatment with iodacetamide (IAA) and 2,4-dinitrophenole (DNP) resulted in elevated 2′,3'-cAMP and higher exosomal protein levels, presumably due to adenosine receptor signaling. Based on these studies, we tested 11 compounds in total for their potential to boost small sEV yields in a biotechnological production setting with HEK293 suspension cells.
Cell density and viability were determined via trypan blue staining with a CEDEX analyzer (Roche).

| sEV purification
Cell culture supernatant was centrifuged 5 min at 500g and 4°C, followed by 30 min at 2000g and 4°C to and subsequently filtered (0.22 μm) and concentrated up to 200-fold via tangential flow filtration in a Sartocon ® Slice cartridge (Sartorius) with a 300 kDa cut-off filter (Sartorius). The concentrate was loaded on an ÄKTApurifier FPLC system (Cytiva) equipped with a 250 ×10 mm Sepharose CL-2B (Cytiva) SEC column. A description of the FPLC method is attached in Supporting Information: Table 1. For sEV quantification from precleared cell culture supernatant, the first two centrifugation steps at 500 and 2000g were followed by a 45 min centrifugation at 10,000g and 4°C.

| Flow cytometric exosome marker analysis
EV surface proteins were analyzed with MACSPlex Exosome kit, human (Miltenyi Biotec). 1E7 particles were diluted in 120 μl MACSPlex buffer in 1.5 ml tubes. After addition of 15 μl MACSPlex Exosome Capture Beads, the suspension was incubated overnight at room temperature and 450 rpm in the dark. Beads were washed with 500 μl MACSPlex Buffer 5 min at 3000g and 15 μl detection antibody cocktail (anti-CD9, anti-CD63, anti-CD81) was added and incubated 1 h at 450 rpm in the dark. The samples were then washed twice with 500 μl MACSPlex Buffer 5 min at 3000g, before measurement in an MACSQuant ® Analyzer 10 flow cytometer (Miltenyi Biotec).

| Transmission electron microscopy
For negative staining the carbon-coated transmission electron microscopy (TEM) grids were glow discharged to make the surface hydrophilic. Five microliters of the sample solution was pipetted onto the grid and incubated for 10 min. Then the grid was floated on three droplets of water and put on a droplet of 2% uranyl acetate in water for 10 min and then floated two times on fresh droplets with 2% of uranyl acetate. The samples were imaged in a JEM-1400 (JEOL) transmission electron microscope.   Over the course of the purification steps, samples were taken for nanoparticle tracking analysis (NTA). The concentration of particles in the supernatant and centrifuged samples was between 4.4E9 and 3.7E9 ml −1 and increased to 4E11 ml −1 after filtration and approx.

| Wound-healing assay
100-fold concentration via TFF (Figure 2c). During SEC purification, most detectable particles (1.4E11 ml −1 ), eluted in the first peak, followed by 9E9 ml −1 in the intermediate fraction and 3.2E9 ml −1 in the second peak. This indicates a majority of sEVs elutes in the first peak after SEC.
F I G U R E 1 Overview of the applied purification workflow using centrifugation, filtration, concentration, and size exclusion chromatography

| Characterization of different host cells for sEV secretion rates and functionality
As the host cell defines sEV characteristics, different cells could be of relevance as platform production systems depending on the respective application. Therefore, we selected different cell types, analyzed production capabilities and their sEV's characteristics and functionality.
Of course, a higher absolute sEV yield was obtained in suspension cell systems due to three-dimensional growth and higher cell densities, but cell-specific sEV secretion shows WJ-MSC/TERT273 and SKOV3

| Increasing sEV secretion
After establishing a reliable sEV production platform, we addressed the problem of relatively poor yields during current production processes, which is a major challenge for the applicability of sEVs What remains to be optimized is the approximately 80% loss of sEVs after purification when compared to sEV levels in the supernatant. It is difficult to classify the efficiency of our method in contrast to others because these calculations seem to be rarely made.
The few examples published showed similar losses (approx. 80%) with an ultracentrifugation/SEC setup (Simon et al., 2020) and even higher losses (approx. 90%) with an affinity purification method (Hung et al., 2018). Presumably, a large proportion of the losses is due to adsorption of vesicles to tube or other plastic surfaces (Evtushenko et al., 2020). The use of specialized plasticware or buffers that prevent adsorption (Görgens et al., 2022) might therefore be a good approach to tackle this issue in the future.
Evaluating potential sEV production host cells we analyzed appearance, size, secretion, and surface markers as well as wound healing and angiogenetic effects. Appearance-wise, sEV samples derived from suspension and adherent culture systems differed in TEM. Presumably, the prevalence of smear in the sEV samples from adherent cells was derived from the exosome-depleted serum that was supplemented to their cultivation media. The suspension cell samples, on the other hand, lacked these artifacts as these cells were cultivated in chemically defined medium. Thus, media complexity should be considered as a potentially disruptive factor for sEV isolation and product purity. Although chemically defined medium is currently preferred in most cases, some cells, especially stem cells and primary cells, will be very difficult, if not impossible, to adapt.
The investigated potential production cell lines in this work had very different origins and possible applications. HEK293 cells are an established system for therapeutics production, especially viral vectors, and may take over the current role of CHO cells in the future (Pulix et al., 2021). While the differences observed between the tested adherent and suspension, HEK293 cells were mainly caused by the cultivation system and respective cell growth, sEV secretion rates per cell and sEV characteristics were rather similar between cultivation systems. This shows that in studies on sEV productivity it can be difficult to ensure comparability due to growth behavior, | 995 studies utilizing HEK293 sEVs for ATMP development, for example, for delivery of therapeutic mRNAs (Kojima et al., 2018;Z. Yang et al., 2020). Considering the downsides of serum for sEV purification, the much higher cell densities and, thus, yields the use of a suspension setup is preferable.
While CAP ® cells performed very similar to CHO cells regarding sEV functionality, their human origin makes them feasible for sEV production and easier to analyze. From all tested suspension cells, they displayed the highest sEV secretion per cell and sEV yield overall. In a previous study, CAP ® cell-derived engineered sEVs could be shown to function as carrier for apoptotic miRNAs to target ovarian cancer cells (Zeh et al., 2019). Their relatively high sEV secretion rate, non-tumorigenic origin, and approved functionality for exosomal miRNA delivery makes CAP ® cells a good platform for sEV-based therapeutics.
Among the tested cell types, MSCs are the most promising source of therapeutic sEVs by nature. Their sEVs have been shown to facilitate many of the regenerative effects of MSC cell therapies without some of the downsides coming with cells (Phinney & Pittenger, 2017). In a case study from 2014, MSC exosomes were successful in treating graft-versus-host disease (Kordelas et al., 2014).
Since then, there have been lots of works investigating the therapeutic potential of MSC sEVs (Ha et al., 2020;Harrell et al., 2019). Our data underline the wound healing and pro-angiogenetic potential even of immortalized WJ-MSC/TERT273 sEVs. Among the investigated sEVs, the WJ-MSC/TERT273-derived ones displayed the highest ratio of CD29 on their surface. CD29, that is, Integrin beta-1, conveys celladhesion and -motion and is also a marker of MSCs (Rahimzadeh et al., 2016) that might be an important factor in the observed woundhealing effect (Grose et al., 2002;Raghavan et al., 2000). Furthermore,

(b)
F I G U R E 6 (a) Scheme of hypothetical mechanism behind increased sEV secretion. IAA and DNP inhibit central energy metabolism, leading to cAMP accumulation. Forskolin activates adenylate cyclase, while fenoterol is an agonist of beta-2-adrenergic receptors that can activate adenylate cyclase via activated G-protein, thus also increasing cAMP levels. This activates protein kinase A, which acts on calcium channels to increase activated Ca 2+ levels in the cell, facilitating vesicle release. (b) Relative cAMP and activated Ca 2+ levels in cells treated with FT, FK, IAA/DNP, or FT + IAA/DNP. Significance calculation via unpaired one-way ANOVA. Significance niveous indicated versus untreated control. DNP, 2,4-dinitrophenole; FT, fenoterol; FK, forskolin; IAA, iodacetamide; sEV, small extracellular vesicle.
the cell-specific secretion of sEVs was the second highest among the tested cells. However, a very slow growth rate outweighed this and led to a relatively low yield of sEVs. In the future, innovative cultivation systems allowing high cell densities while retaining stem cell characteristics will likely be the solution to increase yields for MSC sEVs.
Since SKOV3 cells were derived from a human ovarian adenocarcinoma (Dini et al., 1980), they were included as a control and representative of a tumor cell line, which have repeatedly been reported to secret high amounts of exosomes to facilitate angiogenesis, induce premetastatic niche formation, invasion as well as drugresistance and immune-evasion (Mashouri et al., 2019;Wortzel et al., 2019). Our results confirmed SKOV3 cancer cells as a positive control for high secretion, displaying the highest sEV secretion by far with around 2500 sEVs per cell. Angiogenesis, and, to a smaller degree, wound healing was also mediated by SKOV3 sEVs presumably by some of the surface proteins found, namely CD133, CD326, and CD44. CD133, also referred to as prominin-1, and CD326, the epithelial cell adhesion/activating molecule (EpCAM), are both cancer stem cell markers (Kim et al., 2016), with EpCAM being one of the first tumor-associated antigens discovered acting in a tumor growthpromoting manner (Huang et al., 2018). CD44 is also known as homing cell adhesion molecule (HCAM) and participates in a plethora of cellular functions like cell migration and development but is, depending on its many isoforms and posttranslational modifications, also involved in tumor development, especially epithelial-mesenchymal transition (H. Xu, Yang, et al., 2020). However, given the abundance of tumorgrowth-inducing elements like EpCAM on their surface and the role of cancer sEVs in vivo, this of course disqualifies them as a host cell for therapeutic EV production.
Although CHO cells are the workhorse in biotechnological production and are used as host cells for the majority of recombinant therapeutic proteins (Walsh, 2018), they are not suitable for therapeutic EV production due to immunological concerns.
Despite that, CHO-K1 cells secrete a considerable amount of sEVs (190 per cell) which is interesting as these were shown to transport regulatory RNA (Keysberg et al., 2021), inhibit apoptosis (Han & Rhee, 2018) and increase cell growth (Takagi et al., 2021 They achieved an up to threefold increase with a combined treatment of N-methyldopamine and norepinephrine. Some compounds, for example, FT, that did have an effect in prostate cancer cells did not in MSCs, demonstrating the necessity to verify compound effects for each cell type. In our effort to boost sEV secretion in HEK293sus cells, we found that a treatment with 100 μM FT and 5 μM IAA/DNP led to a significant improvement of sEV secretion, resulting in an approximately threefold increased sEV yield and fivefold increased cellspecific secretion. Treatment had only minor effects on cell growth and viability. High viability, in particular, is paramount for quantification of exosomal sEV secretion, because apoptotic vesicles can be in many cases hard to distinguish and overlap with exosomal vesicles causing artificially high sEV counts (Théry et al., 2018). The relevance of apoptosis, already at 80% viability, was shown in a previous study of us (Keysberg et al., 2021).
The increase was transferable from a 1 ml well plate scale to a 50 ml Erlenmeyer shaker and even 1.2 l bioreactor cultivation with subsequent SEC purification. However, the effect was lower in the bioreactor runs because they were one-to-one transfers of the smallscale setting, adding compounds at the same cell concentrations and using the same transfection-optimized HEK TF medium that lacks some anti-aggregation factors. When moving to a biphasic fed-batch, growing the cells to high densities, and then adding the compounds while retaining a high viability, better results might be achievable.
Some studies claimed that exosome release in general seems to be adenosine receptor-cyclic adenosine monophosphate (cAMP)/protein kinase A (PKA) signaling and Ca 2+ dependent (Islam et al., 2008;Ludwig, Azambuja, et al., 2020;Messenger et al., 2018;Savina et al., 2003). All our effective compounds are known to indirectly affect cAMP signaling, so this could be the link to their shared impact in this context. FT is a beta-2 adrenergic agonist (Jozwiak et al., 2011) so it could raise intracellular cAMP levels by adenylyl cyclase activation (Billington et al., 2017). FK, too, is an activator of adenylyl cyclase that is often used to study the role of cAMP (Alasbahi & Melzig, 2012). IAA is a known inhibitor of glycolysis (Schmidt & Dringen, 2009) and DNP acts as an inhibitor of succinate dehydrogenase and, hence, oxidative phosphorylation (Susheela & Ramasarma, 1971). Consequently, combined treatment with IAA and DNP can result in a nontoxic accumulation of cAMP and increase of sEV secretion as has been shown earlier (Ludwig, Yerneni, et al., 2020

DATA AVAILABILITY STATEMENT
Additional data and workflow information are provided in the supplementary material. The other data that support the findings of this study are available on request from the corresponding author.