Size-dependent immunomodulation by MSC secretome: soluble factors target innate pathways, components larger than 100 kDa regulate T cell proliferation

Abstract

Background

Mesenchymal stromal cells exert their immunoregulatory effects through a complex secretome constituted of soluble factors and extracellular vesicles (EVs). While the immunomodulatory activity of the secretome has been demonstrated, the contribution of each fraction remains poorly defined. In particular, there is little knowledge about which bioactive molecules are responsible for the effect.

Methods

Human peripheral blood mononuclear cells (PBMCs) were treated with resiquimod in presence of clarified or concentrated secretome by tangential flow filtration (TFF), or fractions derived from ultracentrifugation. Supernatant was collected and used to treat THP-1 dual cells, a reporter cell line, to evaluate NF-κB and IRF pathway immunomodulation. T cell proliferation was measured via dye dilution and flow cytometry. Human PBMCs were treated with PHA/IL-2 in presence of clarified or concentrated secretome.

Results

Clarified secretome and the soluble factors fraction from ultracentrifugation inhibited NF-κB and IRF activation in a dose dependent manner. This effect is reduced in presence of concentrated secretome and lost when treated with pelleted EVs. Soluble factors below 5 kDa were responsible for this effect, which was partially mediated by Prostaglandin E2. However, T-cell proliferation was inhibited in the same dose-dependent manner by all TFF-concentrated secretome, regardless of cutoff size, while clarified samples had little effect.

Conclusions

These findings indicate a complex mechanism of action where soluble factors and components larger than 100 kDa modulate immunity but through different pathways. This mechanistic distinction highlights the importance of considering secretome composition when designing cell-free MSC-based therapies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13287-025-04670-2.

Introduction

Extracellular vesicles (EVs) are secreted by all type of cells, and carry a variety of molecules, such as proteins, nucleic acids and lipids from their parent cells, all enclosed by a lipidic bilayer. EVs are heterogeneous and differ in their biogenesis, which can be of endosomal origin or derived from the plasma membrane. They can also differ by their size, cargo composition, but also functional activity [1]. In the study of mesenchymal stromal cells (MSCs), MSC-derived EVs carry secreted molecules which mediate the immunosuppressive and regenerative activities of their parent cells. As such, they are considered a promising cell-free alternative for clinical applications and regenerative medicine [2]. However, MSC-EVs still suffer from a variety of bottlenecks which hinders the development of these vesicles as a clinically relevant alternative. The EV community has issued standard guidelines for the study of EVs, but the field still suffers from a lack of standardization in production, characterization, concentration/purification methods and functional assays [3].

Furthermore, the mechanism of action of MSC-EVs has shown to be complex and multiparametric. More specifically, a clear distinction can be made between two components of MSC-derived secretome: the free soluble factors and the EVs themselves [4]. Some of these soluble factors can also interact with the plasma membrane of EVs forming a “corona” [5], which complexifies even more the understanding of which fraction, if not both, holds the regenerative and immunomodulatory properties. The variability in EV processing methods, which affects the concentration levels of soluble factors and/or EVs each in different ways [6], makes any global understanding of the mechanism of action particularly challenging.

To further elucidate the mechanism of action, we investigated the immunomodulatory activity of the MSC-derived secretome and its tangential flow filtration (TFF)-concentrated fractions using different molecular weight cutoffs. To study the immune modulatory properties on innate immune response, a potency using peripheral blood mononuclear cells (PBMCs) with a THP-1 reporter cell line was developed. In combination, another assay was used, which assessed CD3⁺ cell proliferation. We evaluated the expression of classical EV markers and analyzed the global protein composition of the secretome using a multiplexed protein quantification assay. Our results demonstrate that the MSC-derived secretome exhibited stronger anti-inflammatory activity than its corresponding TFF-concentrated fractions, regardless of the cutoff size, suggesting that inhibition of innate pathways is likely mediated by soluble factors smaller than 5 kDa.

To investigate this further, we quantified key immunomodulatory soluble factors below the 5 kDa threshold such as prostaglandin E2 (PGE2) and kynurenine. Kynurenine is an immunomodulatory metabolite which is produced via indoleamine 2,3 dioxygenase-(IDO) activity. PGE2 is another anti-inflammatory mediator secreted by MSCs. We identified PGE2 as being at least partially responsible for the observed anti-inflammatory effects. However, these findings did not correlate with the anti-proliferative activity on CD3⁺ cells, indicating that the immunomodulatory effect is multifactorial and likely involves distinct molecular pathways.

Methods

Cells

Human adipose stromal cells (ASC) were purchased from Etablissement Francais du Sang (EFS) at P0. They were further amplified in aMEM (Gibco) + 5% platelet lysate (PAN Biotech) until a working cell bank was stored in liquid nitrogen at P3, which constituted the cells used for this study.

THP-1 dual were purchased from Invivogen. They were amplified until a working cell bank was realized and was stored in liquid nitrogen. Cells were cultivated in RPMI1640 + Glutamax (Gibco), with addition of 100 µg/mL of Normocin (Invivogen), 100 µg/mL of Zeocin (Invivogen), 10 µg/mL of Blasticidin (Invivogen), 100 U/mL of Penicilicin + Streptomycin (Gibco), 10% FBS (Biowest), 25 mM HEPES (Cytiva). Cells were passaged twice every week and were diluted in their media without centrifugation to a concentration between 250 000 and 500 000 cells/mL. Cells were used under passage 20, as recommended by supplier’s instructions. Cells were cultivated for 2 weeks before use after thawing.

Cell culture and EV production

ASCs were seeded at 2000 cells/cm² in 300 cm² flasks (TPP) or 875 cm² multiflasks (Falcon) depending on the number of cells seeded. Cells were cultivated for one week in aMEM (Gibco) + 5% platelet lysate (PAN Biotech) before passage, with a single intermediate media change. At confluence, cells were passed using TrypLE Express (Gibco) and seeded into a spinner-flask for repeated production or in flasks for further amplification. For secretome vs. post TFF studies, 100mL, 1 L spinner-flasks (BellCo) or 10 L single-use stirred tank bioreactors were used for 3D cell culture and for EV production. For ultracentrifugation studies, 100 mL spinner-flasks (BellCo) were used. For different cutoffs of TFF and PGE2 experiments, 10 L single-use stirred tank bioreactors were used. Cells were seeded on Cytodex1 microcarriers (Cytiva) to grow in a 3D setting. After seeding, complete media was added, and cycles of stirring and resting were set for homogeneous cell adhesion.

After cycles, spinner-flasks were set at 34 RPM, and bioreactors were set at 63 RPM. After cell expansion, secretome/EV production was started. Spinner-flasks were submitted to a shear stress equivalent to a mean kolmogorov length of 37 μm, while bioreactors were set at a mean kolmogorov length of 65 μm for four hours.

Cell culture follow-up was realized by volume sampling of the spinner-flask or bioreactor. Cell number and viability were obtained by staining cells on the surface of the microcarriers with NucBlue Live and NucGreen Dead (Invitrogen). A cell counting script on Python was used to accurately determine the number of cells at the surface of the microcarriers and cell viability. A minimum of ten photos were taken for every 3D cell culture follow-up.

Glucose and lactate were monitored throughout the 3D cell culture using glucose and lactate strips (NovaBiomedical). Media was changed when glucose dropped below 10 mg/dL.

Downstream process and storage

At the end of the production, the harvest from spinner-flasks containing all the released EV was clarified through a 0.45 μm filter (Thermo Scientific). For bioreactors, the harvest was clarified using a standard filter bag 20 μm (Entegris) to remove microcarriers and cells. Harvest was further clarified with a Sartoguard PES 0.2 μm filter. Post clarification (PC) samples were stored at -80 °C, waiting for further processing with TFF.

After PC, the media was processed and diafiltered by TFF using either a 5, 10, 30 or 100 kDa membrane. Post TFF (PT) samples were aliquoted in PBS (Gibco) for characterization or potency assays and were stored at -80 °C until further use.

For ultracentrifugation, PC samples were ultracentrifuged with a Beckman Coulter Optima MAX XP at 150 000 g for 2 h. Supernatant was collected and stored at -80 °C, and pellet was resuspended in 1 mL of PBS (Gibco) and stored at -80 °C until further use.

EV and secretome characterization

Protein concentrations were measured using a Qubit 3000 Fluorimeter and the Protein Assay kit (ThermoFisher).

Bulk-EV phenotyping was carried out using either the I/O EV MACSPLEX kit (Miltenyi) or the MSC MACSPLEX kit provided by Miltenyi as part of the beta testing of their product. Samples were processed for both techniques following manufacturer’s protocol for an overnight incubation between beads and EVs, in 1.5 mL tubes. Samples were analyzed by flow cytometry, gating on single beads, through measurement of median fluorescence intensity (MFI) of each specific bead population on a Attune NxT (AFC2). Data was analyzed using Flowjo software v10.8.1 and was represented for each marker as the MFI relative to the total expression of all the markers analyzed by the MACSPLEX kit.

PGE2 lipid levels were determined using Prostaglandin E₂ Express ELISA Kit (Cayman Chemical), following manufacturer’s instructions. For this dosage, EVs were submitted to bath sonication (3 × 10 min, separated by vortexing between each run) to induce EV lysis, following a previously described protocol by Fitzgerald et al [7].

Kynurenine levels were determined using Kynurenine to Tryptophan ratio ELISA pack (Immusmol), following manufacturer’s instructions. EVs were lyzed using RIPA 0.5X prior to dosage.

Bioactive proteins concentration in lyzed samples were measured using the ProCartaPlex Human Immune Monitoring 65-Plex Panel (Invitrogen) (G-CSF, GM-CSF, IFN-α, IFN-γ, IL-1α, IL-1β, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-12p70, IL-13, IL-15, IL-16, IL-17 A, IL-18, IL-20, IL-21, IL-22, IL-23, IL-27, IL-31, LIF, M-CSF, MIF, TNFα, TNFβ, TSLP, BLC, ENA-78, Eotaxin, Eotaxin-2, Eotaxin-3, Fractalkine, Gro-α, IP-10, I-TAC, MCP-1, MCP-2, MCP-3, MDC, MIG, MIP-1α, MIP-1β, MIP-3α, SDF-1α, FGF-2, HGF, MMP-1, NGFβ, SCF, VEGF-A, APRIL, BAFF, CD30, CD40L, IL-2R, TNF-RII, TRAIL, TWEAK). Samples were lyzed using 0.25X RIPA (Abcam) and 1X Halt’s Protease Inhibitor (ThermoFisher), with an overnight incubation at 4 °C. Analyte measurement was performed using a Bio-Plex 200 (Bio-Rad Laboratories). The standard curves were determined using a 5-parameter logistic regression model.

Functional assays

Blood from donors was purchased from EFS. Peripheral blood mononuclear cells (PBMCs) were isolated using Lymphoprep (Stemcell) by conventional Ficoll density gradient centrifugation. PBMCs were frozen in 90% FBS (Biowest) and 10% dimethylsulfoxide (DMSO, ThermoFisher) until further use. For CD3+ anti-proliferative activity assay, PBMCs were thawed in RPMI 1640 + Glutamax (Gibco) + 10% FBS and were rested for 24 h after thawing at 37 °C and 5% CO2. They were then marked with carboxyfluorescein succinimidyl ester (CFSE, ThermoFisher) at 0.5 M and seeded at 100 000 cells/well in a 96-well plate (ThermoFisher). 100 µL of each PT samples were added in a dilution cascade and 30 µL of medium + PHA/IL2 (10 µg/mL; 50IU) were added to promote T-cell proliferation. The plate was then incubated for 72 h at 37 °C and 5% CO2. After incubation, the cell supernatant was harvested and cells were labeled with a viability marker (Zombie Yellow, Biolegend) and an anti-CD3 antibody (Brilliant Violet 421, Biolegend) before analysis by cytometry on a Attune NxT, using Flowjo software and proliferation tool.

For PBMC NF-κB and IRF modulation, PBMCs were thawed in RPMI 1640 + Glutamax (Gibco) + 10% FBS and were rested for 1 h after thawing at 37 °C and 5% CO2. They were then treated with our samples, media or dexamethasone, and resiquimod (R848) was added to all wells expect media controls to induce inflammatory stimulation. All samples are evaluated as triplicates or quadruplicates. The plate is incubated for 24 h at 37 °C and is then centrifuged at 500 g for 5 min to pellet cells. Supernatant is collected and divided into three different plates. One plate containing 25 µL is used for cytotoxicity assay, using CytoTox 96^® Non-Radioactive Cytotoxicity Assay (Promega). Values with cell viability below 70% were discarded for all experiments, as immunomodulatory response could be due to cytotoxicity (except for Fig. 5B). Another plate, containing 25 µl as well, is frozen at -20 °C for 1 h to ensure all potentially collected PBMCs are dead and do not cause immunogenic reaction. After freezing, 160 µL of THP-1 dual in RPMI 1640 + Glutamax (Gibco) + 10% FBS are added, and the plate is incubated for 24 h at 37 °C. The third plate is stored at -20 °C as a supernatant stock until further use. After incubation, 20 µL of supernatant are transferred to one transparent 96 well plate (ThermoFisher) for NF-κB reading and one white nontransparent 96-w plate (ThermoFisher) for IRF reading. 90 µL of QuantiBlue solution (Invivogen) is added to the first plate and is incubated for 30 min at 37 °C or until the first well turns blue as per supplier’s instructions. After incubation, the plate is read for absorbance at 620 nm using a SpectraMax Id3 (Molecular Devices).

Fig. 5.

PGE2 might be responsible for the anti-inflammatory effect of MSC-derived secretome. (A) ELISA dosage of kynurenine shows depletion after TFF. (B) Similar to kynurenine, ELISA dosage of PGE2 shows decreased levels after TFF. (C) PGE2 inhibitors inhibit PC inhibitory properties on NF-κB. PC complex composition of signaling proteins might explain a lower anti-inflammatory activity, compared to PGE2 alone. PC and PT samples were added a concentration of 9e8 particles/mL. (D) Cell viability of PBMCs after potency assay, assessed through LDH measurement. Statistical analysis was carried out using GraphPad Prism 9, using unpaired parametrical t-test

For IRF reading, 50 µL of QuantiLuc (Invivogen) is added, and the plate is read immediately for luminescence at 100 ms of reading time.

For PGE2 inhibition studies, TG4-155 (TargetMol) and GW627368 (MedChemExpress) were added at a concentration of 5 µM.

Statistical analysis

Data was analyzed using GraphPad Prism 9.0. Parametric t-test were used for one-on-one comparisons, while one-way ANOVA was used to check the impact of one specific parameter.

Results

Validation of a PBMC screening assay for anti-inflammatory properties of EVs

To understand anti-inflammatory properties of EVs, we developed a potency assay for high throughput screening of secretome/EV samples (Fig. 1A). To validate the assay, the use of frozen PBMCs was compared to fresh ones, the repeatability and reproducibility of the assay was evaluated. Finally, PBMC cytokine levels in supernatants were compared to NF-κB and IRF activation readouts in THP-1 dual cells.

Fig. 1.

Validation of a PBMC screening assay for NF-κB pathway inhibitory properties of EVs. (A) Graphical protocol of the potency assay. (B) Fresh PBMCs were more responsive to EVs than frozen ones - each dot corresponds to the mean of the NF-κB pathway activation of a clarified or a concentrated sample relative to controls, assayed on at least 2 different donors. Repeatability was assayed as the %CV between at least 3 independent experiments, using the same donor. (C) Repeatability was evaluated on 3 different donors. Values over the bar correspond to %CV. (D) Reproducibility was evaluated as the %CV of at least 3 independent experiments with 3 different donors, highlighting different donor response. Values over the bar correspond to %CV. (E) Multiplexed dosage of cytokines showed levels of pro-inflammatory cytokines linked with NF-κB dosed in PBMC supernatant evolve in the same trend as THP-1 Dual activation. High dose (HD): 6e8 particles/mL; medium dose (MD): 3e8 particles/mL ; low dose (LD): 1.5e8 particles/mL. Statistical analysis was carried on GraphPad Prism 9.0, using a parametric t-test

Fresh PBMCs were more responsive to clarified and concentrated samples, compared to frozen cells, with a difference of 0.11 ± 0.09 for NF-κB normalized response. While inhibition of NF-κB was slightly lower on frozen cells, the trend was similar compared to activation on fresh cells. Thus, the use of frozen PBMCs was therefore accepted, while acknowledging that a secretome showing no apparent activity on frozen cells may still exhibit at least 10% inhibition of NF-κB (Fig. 1B).

Repeatability was assessed by calculating the coefficient of variation (%CV) from at least three independent experiments, each performed in triplicate, using PBMCs from the same donor.

This assessment was conducted on three different donors. The mean %CV for the high dose (HD) sample (34%) was higher to that of the R848 control (17%), indicating high intra-donor variability. Higher %CV may be caused by lower mean values, while standard deviation (SD) remains the same. Indeed, mean SD between control (0.187) and HD (0.167) were similar, highlighting low intra-donor variability. In contrast, the low dose (LD) sample showed equivalent %CV (32%), but higher SD (0.333), suggesting increased variability (Fig. 1C).

Reproducibility was evaluated using PBMCs from three different donors, each tested in at least three independent experiments in triplicate. While %CV (17%) and SD (0.18) for R848 stimulation alone remained stable with the addition of different donors, %CV increased to 45% and SD elevated to 0.29 for the HD sample, highlighting donor-to-donor variability (Fig. 1D).

Levels of pro-inflammatory cytokines TNF-α and IL-1β measured in PBMC supernatants were consistent with NF-κB activation measured in THP-1 dual cells. However, IL-6 levels—particularly for the PT sample—were lower compared to THP-1 dual readouts. Overall, the cytokine profiles associated with NF-κB activation paralleled the THP-1 dual results, supporting the validity of this cell line as a proxy for PBMC-based assays.

Finally, a similar analysis between fresh and frozen cells, repeatability, reproducibility, and correlation between PBMC cytokine levels and THP-1 dual cells activation was performed for the IRF pathway. Fresh PBMCs were also more responsive compared to frozen cells, with a difference of 0.20 ± 0.08 in IRF normalized response (Supplementary Fig. 1.A). For HD, mean %CV for repeatability assessment was 14%, which was equal to R848 mean %CV. (Supplementary Fig. 1.B). Addition of multiple donors for reproducibility increased %CV to 16% (Supplementary Fig. 1.C), showing that inter-donor variability is quite low for IRF pathway. Levels of IFN-α and IP-10 were similar to IRF activation in THP-1 Dual, except for HD_PT (Supplementary Fig. 1.D).

TFF-concentrated secretome has lower levels of bioactive proteins and reduced anti-inflammatory properties

To understand which fraction of the MSC-derived secretome holds the therapeutic effect, we compared clarified (PC) and both clarified and concentrated (PT) MSC-derived secretome (Fig. 2A). Size distribution measured by nanoparticle tracking analysis (NTA) showed no significant differences between PC and PT (Fig. 2B). Both PC and PT samples expressed classical EV and MSC markers. This confirms that concentration did not affect the EV fraction of the secretome. Increase of CD63 and CD73 expression relative to particle concentration after TFF might be associated to a reduction of non-EV particles. (Fig. 2C). Quantification of signaling proteins (pg of signaling protein/10^9 particles) in four PC/PT pairs revealed high variability among PC samples (Fig. 2D). Particles were concentrated using TFF by a 10.1 ± 1.7 fold, but dosed cytokines levels in pg of signaling protein/10^9 particles were reduced by a mean of 72.3 ± 26.6%. This suggests that most cytokines are present in the soluble phase and are effectively removed by TFF.

Fig. 2.

Clarified secretome has higher anti-inflammatory activity compared to its concentrated counterpart. (A) 19 samples of secretome were clarified and concentrated by TFF. (B) Size distribution is equal between PC and PT. (C) MACSPLEX analysis shows expression of both EV and MSC markers for PC and PT samples (n = 12). (D) Principal Component Analysis (PCA) of cytokine dosage relative to particles (pg/10^9 particles) shows decrease of nearly all signaling proteins (n = 4). Direction of arrows indicate higher concentration (pg/10^9 particles). (E) At equal particles/mL concentration, samples processed by TFF show reduced immunomodulatory activity on NF-κB pathway, (F) and higher cytotoxicity compared to PC (n = 19). Concentrations of samples are between 6e8 and 9e8 particles/mL. Statistical analysis was carried on GraphPad Prism 9.0, using a paired parametric t-test for E and F, and a two-way ANOVA multiple comparison for C

We also observed that the EV cargo content, dosed in PT samples, is quite stable across different batches. Dosed signaling proteins included various pro and anti-inflammatory cytokines, showing the complex composition of secretome (Fig. 2D). Using the previously described potency assay (Fig. 1), PC showed a stronger inhibition of NF-κB pathway compared to PT samples. PBMCs were treated at equivalent concentration of particles/mL (Fig. 2E). Similar results were observed on IRF (Supplementary Fig. 2.A). Additionally, PT samples exhibited higher cytotoxicity compared to PC, though statistical significance was not reached (Fig. 2F). While concentration with TFF increased particles/protein ratio (Supplementary Fig. 2.B), protein concentration in assay wells was not correlated with the difference in modulation between PC and PT, with no correlation between PT and protein concentration in wells (Supplementary Fig. 2.C). Overall, these results suggest that the anti-inflammatory activity of the MSC-derived secretome is mostly carried by soluble factors, removed from the sample by TFF. EVs showed no significant activity in our potency assay, except for a moderate impact on IL-6 secretion.

Ultracentrifuge supernatant and clarified secretome show NF-kB pathway inhibitory properties, in contrast with pelleted EVs

To confirm that soluble factors are responsible for the anti-inflammatory activity, three samples of clarified MSC-derived secretome (PC) were subjected to ultracentrifugation. This process separates extracellular vesicles (EVs) and larger particles into the pellet, while soluble factors remain in the supernatant (SN) (Fig. 3A). Although the particle concentration in the SN decreased, it remained detectable by NTA, indicating that not all EVs were efficiently pelleted. The size distribution of PC closely resembled that of isolated EVs (Fig. 3B). PC and SN exhibited comparable inhibition of the NF-κB pathway. In contrast, EVs alone showed a non-significant trend toward (p = 0.08 and 0.09) increased NF-κB activity (a slight pro-inflammatory signature) that warrants confirmation with greater statistical power. (Fig. 3C). Similar effects can be observed for IRF pathway (Supplementary Fig. 3). Combined with previous findings, these results confirm that soluble factors hold the anti-inflammatory activity, and not EVs.

Fig. 3.

NF-κB inhibitory properties are carried by soluble factors. (A) Graphical protocol. (B) Size distribution by NTA shows ultracentrifugation supernatant (SN) has smaller particles, as it is depleted from most vesicles. PC and pelleted EVs show no difference in size distribution (n = 3). (C) Potency assay shows that PC and SN show similar inhibition of NF-κB pathway, while EVs alone have a slight pro-inflammatory effect. Each dot corresponds to the mean of the three samples, for one donor. Two donors were used for this experiment (orange and purple). Medium dose (MD) and low dose (LD) for PC and EV correspond respectively to 4.5 and 2.25 particles/mL. For SN, because the concentration in particles could not reach previous levels, samples were assayed as undiluted for MD and diluted by 2 for LD. Statistical analysis was carried out using GraphPad Prism 9, using paired parametrical t-test

Soluble factors below 5 kDa are responsible for the immune modulatory effect

To understand which soluble factors were carrying the immunomodulatory effect on NF-kB, three different samples of PC were concentrated by TFF with different cutoff sizes: 5 kDa, 10 kDa, 30 kDa, and 100 kDa (Fig. 4A). In our TFF approach, membranes with a given nominal cutoff allow molecules smaller than this size (e.g., PGE₂ and kynurenine) to pass into the permeate, while larger molecules are retained. By comparing PC and PT fractions, we can assess whether low-molecular-weight bioactive factors contribute to the observed effects. As recommended by the International Society of Extracellular Vesicles (ISEV), all TFF-processed samples retained extracellular vesicles, confirmed by cryo-transmission electron microscopy (cryoTEM) (Fig. 4B), and expression of classical EV (red) and MSC (blue) markers (Fig. 4D). Differences in MFI between PT samples are probably associated with differential concentration of non vesicular particles detected by NTA. Size distribution by NTA were similar for all PT samples. 5 kDa and 10 kDa appear to have slightly bigger particles (Fig. 4C). As previously observed, PCs samples inhibited NF-κB pathway, confirming our previous findings. However, all samples concentrated by TFF lost their anti-inflammatory properties, regardless of TFF cutoff size. (Fig. 4E). Similar results were also observed for IRF pathway (Supplementary Fig. 4.A).

Fig. 4.

Soluble factors below 5 kDa modulate NF-κB pathway, but do not mediate T-cell anti-proliferative properties. (A) Three different samples were processed through TFF with different cutoff sizes. (B) CryoTEM imaging shows EVs in all PT samples. (C) Size distribution of samples by NTA was similar, with higher-sized particles in 5 kDa & 10 kDa samples. (D) Surface marker expression assessed by MACSPLEX show that all samples expressed classical EV and MSC markers. (E) PC retained immunomodulatory activity on NF-κB, which was lost after TFF, regardless of the cutoff (n = 3 donors). (F) However, all PT samples demonstrated CD3 + anti-proliferative properties, relative to stimulated controls (n = 2 donors). Statistical analysis was carried on GraphPad Prism 9.0, using two-way ANOVA multiple comparison for D

To further assess the immunoregulatory activity of PT, a second assay based on CD3⁺ T cell proliferation was performed. All PT samples induced a dose-dependent reduction of CD3⁺ T cell proliferation, regardless of the cutoff size used (Fig. 4E). In comparison, PC samples showed a pro-proliferative effect in this assay (Supplementary Fig. 4.B). These observations suggest a possible contribution of EVs to the anti-proliferative effect, in contrast to their minimal role in NF-κB or IRF modulation.

Overall, these results indicate that, in our potency assay, soluble factors smaller than 5 kDa are primarily responsible for the anti-inflammatory effect on cytokine secretion. However, these factors do not appear to mediate the anti-proliferative effect on CD3⁺ T cells, suggesting distinct mechanisms underlying these two activities.

PGE2 partially mediates MSC-secretome anti-inflammatory activity

Among the soluble factors below 5 kDa, two anti-inflammatory mediators were dosed to understand the impact of TFF on their levels. Kynurenine levels were depleted by TFF (Fig. 5A), supporting the selective loss of small soluble mediators through this concentration process. PGE2 levels were quantified in one batch across all PC and PT samples after TFF with various cutoff sizes. Similarly to kynurenine, PCs contained high concentrations of PGE2 (mean: 1904 pg/mL), which were markedly reduced following TFF (mean: 84.5 pg/mL), irrespective of the cutoff used (Fig. 5B). This is equivalent to a 96% loss of PGE2 after TFF. These data indicate that PGE2 is predominantly secreted as a soluble factor and is not associated with EVs.

Purified PGE2 demonstrated strong anti-inflammatory activity, at concentrations found in PC (1904 pg/mL – called 1X PGE2) and in PT (84.5 pg/mL – called 0X PGE2), even saturating the system (Supplementary Fig. 5). Although PC exhibited some anti-inflammatory activity, it was lower than that observed with purified PGE2. The addition of PGE2 inhibitors TG4-155 (TG) and GW627368X (GW) to PC abolished its anti-inflammatory effect, suggesting a central role for PGE2 in modulating inflammation in this fraction. In contrast, addition of the same inhibitors to PT had no effect or induced a pro-inflammatory response, possibly due to off-target effects, particularly of TG. Furthermore, supplementation of PC and PT samples with equivalent concentrations of purified PGE2 restored their anti-inflammatory activity (Fig. 5C), supporting its functional relevance. Similar observations were made regarding the IRF signaling pathway (Supplementary Fig. 6). PBMC viability was not statistically different across clarified and concentrated conditions, except for the addition of TG and GW with PC. (Fig. 5D).

Discussion

This study demonstrates the use of a validated immunomodulatory potency assay [8] to identify the bioactive fraction responsible for immunomodulatory effects on innate pathways within MSC-derived secretome. This, in turn, might shed some light on the mechanism of action of MSC-EV based products and supports progress toward clinical translation.

In potency assays, the use of frozen PBMCs offers practical advantages in terms of logistical ease and batch consistency. However, freezing can affect PBMCs gene expression [9], cytokine expression levels [10] and specific population viability. Notably, lymphocytes are mostly preserved, while monocytes and innate lymphoid cells are more sensitive to cryopreservation [11].

This did not appear to significantly impact on the overall results, as the overall trend remained consistent despite reduced responsiveness in frozen cells. In our assay, PBMCs were stimulated using the TLR7/8 agonist R848. TLR7 and TLR8 are expressed in plasmacytoid dendritic cells (pDCs) and myeloid dendritic cells, and both receptors are also found on monocytes [12]. Monocytes play a key role in innate immune responses by releasing pro-inflammatory cytokines [13]. Therefore, cytokine secretion in R848-activated PBMC supernatants is likely driven primarily by monocyte activation, which may in turn influence other immune cell subsets. Post-thaw cell viability exceeded 90% across all donors (data not shown), indicating that our freeze/thaw protocol preserved overall cell integrity. This is consistent with previous reports showing comparable viability levels, even after a month [11]. In our protocol, PBMCs were rested for 1 h following thawing. Although PBMC resting protocols are not yet standardized, a 1-hour rest has been reported to improve the detection of T cell markers compared to 0 or 24-hour rest periods [14]. Whether resting is beneficial or not depends on the immune cell population under investigation [15]. Additionally, factors such as cryopreservation duration and the composition of the freezing buffer also influence PBMC recovery and function post-thaw [16, 17]. Optimizing these parameters may improve the responsiveness of PBMCs to MSC-derived secretome. In summary, while frozen PBMCs may underestimate the full extent of immunomodulatory effects, it remains a practical and valid approach for a quick and reproducible assessment of the anti-inflammatory properties of MSC-derived secretomes.

We demonstrated that the clarified MSC-derived secretome exhibits stronger immunomodulatory activity on innate immune responses compared to its TFF-concentrated counterpart. Both samples were assayed at equivalent particles concentration, thus reinforcing that loss of immunomodulatory properties on innate pathways was due to soluble factors. This observation suggests that TFF remove key low-molecular-weight soluble factors responsible for the therapeutic effects of the secretome. In regard to secretome composition, PC samples showed high variability in levels of soluble factors, which was reduced after TFF. This is likely to be associated with process parameters such as cell population doubling level / passage [18]. The MSC-derived secretome concentrated via TFF exhibited no detectable effect in the PBMC-based inflammatory assay but induced a dose-dependent inhibition of CD3⁺ T cell proliferation, with no significant differences observed across the 5 to 100 kDa cutoff conditions. This suggests distinct mechanisms of action: while innate immune modulation appears to be primarily mediated by low molecular weight soluble factors, the inhibition of T cell proliferation may involve larger components—likely extracellular vesicles (EVs) or molecule complexes exceeding 100 kDa in size. Notably, in the case of anti-inflammatory molecules, TGF-β mediates immunomodulatory properties of MSCs [19]. TGF-β is often secreted as part of a large latent complex composed of three subunits, including the latent TGF-β binding protein (LTBP), which molecular weight is around 120 to 210 kDa [20]. Thus, secreted TGF-β might also be mediating the inhibiting effect on CD3 + proliferation. In contrast, the clarified secretome did not reduce CD3 + cell proliferation. These findings support the hypothesis of multiple, compartmentalized mechanisms of action within the MSC-derived secretome. They highlight the importance of dissecting the respective contributions of soluble factors and EVs to fully understand and harness their immunomodulatory and therapeutic potential [21]. Regarding the difference in immunomodulatory properties between soluble factors and EVs, Papait et al. showed contrasting results compared to our study. They demonstrated that both the total secretome and the EV-depleted fraction exhibited anti-proliferative effects on PBMCs stimulated with anti-CD3 antibodies, whereas isolated EVs had no significant effect [22]. Similarly, other studies have confirmed that the complete secretome exerts greater immunomodulatory activity than EVs alone [23, 24]. Conversely, some reports suggest that EVs may be more potent than conditioned media in certain contexts [25]. These discrepancies likely reflect differences in production protocols, concentration methods, and functional assay designs, making direct comparisons and definitive conclusions challenging [26]. Mitchell et al. demonstrated that secretome and EV both protected against senescence, but only EVs decreased inflammation levels. Furthermore, EVs and soluble factors modulated different parameters in an acute muscle injury model [27]. This potential dual mechanism of action of soluble factors/EVs indicates a need to concentrate secretome without any alteration. To our knowledge, no DSP method which fits this criteria, let alone scalable [6]. A concentration of secretome might also not be needed, as the concentration of our PC samples for different kDa TFF was 3.0 ± 0.84e9 particles/mL (n = 3). In comparison, clinical trials reported over the last years used a quantity between 1e8 and 1e10 particles per treatment [28]. Evaluating batch size, recovery of particles after storage and sterilizing filtration might allow the use of unconcentrated secretome for clinical trials.

Our findings indicate that the observed anti-inflammatory effects on innate immune response are primarily mediated by soluble components with a molecular weight below 5 kDa.

Many candidate molecules have been described in the literature as potential carriers of functionality of MSC-derived secretome or MSC-EVs. However, most of them, specifically proteins, such as IL-10 [29, 30], TGF-β [31], or TSG-6 [32] have sizes bigger than 5 kDa. Amongst molecules under 5 kDa, PGE2 is a lipid mediator involved in regulation of immunity. PGE2, secreted by MSCs, exerts its anti-inflammatory properties through binding with EP2 and EP4 receptors on immune cells, which in turn leads to increased cyclic adenosine monophosphate (cAMP) levels and activation of protein kinase A, mediating immunomodulatory activity [33]. Through this mechanism, PGE2 modulates T-cells, macrophages and DC immune activation [34]. In our study, PGE2 levels were high in PC (2168 ± 569 pg/mL, n = 7), and depleted to basal levels after TFF (30 kDa, 166 ± 71 pg/mL, n = 7). This depletion of PGE2 coupled with other bioactive proteins might explain the reduced immunomodulatory effect of PT samples on NF-kB. In comparison, PGE2 concentration in MSC conditioned media were of similar levels in other studies, but was shown to increase sharply after priming with IL-1β [35, 36]. As PGE2 could be a key soluble factor, priming with IL-1β could be an interesting follow-up in further studies.

We also showed that while PC carries high doses of PGE2, it does not have the same anti-inflammatory effect compared to pure PGE2 introduced at the same levels. Our results highlight that the MSC-derived secretome is a highly complex and dynamic mixture of soluble factors, encompassing both anti-inflammatory mediators such as PGE2, and pro-inflammatory cytokines. While PGE2 is known to exert potent immunosuppressive effects, particularly by downregulating innate immune activation, its action within the clarified secretome may be attenuated or counterbalanced by the presence of pro-inflammatory molecules. This intricate balance between opposing signaling pathways could explain the reduced anti-inflammatory activity observed in PC compared to PGE2 alone. The overall immunomodulatory outcome likely depends on the relative abundance, stability, and interplay of these molecules, emphasizing the need for deeper characterization of the secretome’s composition to fully understand its therapeutic potential.

High concentrations of PGE2 appeared to saturate the anti-inflammatory response in PBMCs, showing that a low concentration of PGE2 would be enough to trigger anti-inflammatory activity. Indeed, dose-response of PGE2 showed a diminution of 50% of NF-κB activity for concentrations as low as 51 pg/mL (Supplementary Fig. 5). This means levels of PGE2 in PT would still be able to trigger an anti-inflammatory response, which is probably downplayed by other signaling proteins still present after concentration.

Addition of PGE2 inhibitors decreased the anti-inflammatory effect in PC, but also in PT. This shows that PGE2 plays an important role in immunomodulation on PBMCs. However, we have not investigated whether this would be linked to the levels of PGE2 in PC and PT samples, to PGE2 secreted by PBMCs themselves, or both. Further studies should be carried to evaluate how PGE2 plays as a functional molecule on immune cells.

IDO is an enzyme that catalyzes the degradation of tryptophan into kynurenine and downstream metabolites [37]. IDO activity, along with kynurenine production, has been shown to promote the differentiation of regulatory T cells (Tregs) [38], reduce inflammatory responses in LPS-stimulated THP-1 cells [39], and induce a tolerogenic phenotype in pDCs, thereby reducing their antiviral capacity [40]. Results show that PC samples contain kynurenine, which is depleted after TFF, regardless of cutoff size (Fig. 5B). However, the levels on kynurenine observed in PC are very low compared to what can be observed physiologically, which are closer to the µM [41]. Thus, it is unlikely that kynurenine has any key role in the immunomodulatory differences between PC and PT, though we have not further investigated with an inhibitor of kynurenine.

Many other soluble factors under 5 kDa with anti-inflammatory activity linked with MSC therapeutic properties have been described in the literature. Amongst them, lactate [42, 43], bioactive lipids [44, 45], but also itaconate [46] could have key role in immunomodulatory activity on innate immune pathways observed from PC samples. Further studies could be focused on measuring levels of these bioactive molecules. The variety of molecules and the diversity of mechanism of action hints that not a single molecule can solely be responsible for the effect. The immunomodulatory properties of MSC-derived secretome are probably mediated by a variety of bioactive molecules, all interacting with immune cells.

In the case of TFF, to our knowledge, no studies have yet directly compared the differences in functional activity between clarified and concentrated secretome. The impact of DSP on the MSC-derived secretome underlines the importance of a standardization of optimal DSP processes. Numerous DSP methods have been described and reviewed, which all have different ways of impacting the levels of bioactive soluble factors [6]. For example, ultracentrifugation removes the functional corona layer of EVs, composed of soluble factors interacting with the plasma membrane of EVs [5]. This effect is not observed with TFF [5], hinting that our bioactive molecules are indeed soluble factors.

Further studies to understand how purification and concentration of our product may affect its functionality are needed. Pertaining to this, studies have shown the difference of protein content between secretome and EVs [47], or when comparing different TFF cutoffs [48]. Regarding the effect on PBMCs, we did not investigate whether MSC-derived secretome exert direct effects on all immune cell types or selectively target specific subsets, subsequently influencing other immune populations indirectly through cytokine release by the initially affected cells. Previous studies have shown that MSC-EVs enhanced Tregs production in an antigen presenting cell dependent manner [49]. Further studies may investigate how soluble factors and EVs selectively modulate specific subsets of PBMCs.

As shown by this study, a single potency assay is not enough to assess the complexity of MSC-derived secretome immunomodulatory activity [50]. A specific parameter evaluated through a potency assay might not fully correlate with therapeutic properties. For example, CD73 activity is uncorrelated with immunomodulatory capabilities [51]. Furthermore, the in vitro assays used in this study may not fully reflect the effects of EVs in an in vivo model. More studies should be carried to understand if any of our in vitro assays are predictive of some in vivo results. More specifically, studies have reported that some MSC-derived secretome concentrated using UC showed promising effects in chronic kidney disease [52], graft versus host disease (GVHD) [53]. Assessing if in vitro potency assays are predictive of in vivo effect is crucial and should be explored in further studies.

In conclusion, taken together, our findings underscore the critical role of soluble factors in modulating innate immune responses, with a comparatively limited contribution from EVs in this specific context. Among the soluble components, PGE2 emerges as a major mediator of anti-inflammatory activity, likely contributing substantially to the observed immunomodulatory effects on innate immune pathways. However, the complexity of the MSC-derived secretome suggests that this activity does not rely on a single molecule, but rather from the interplay of multiple signaling factors—both pro- and anti-inflammatory—in dynamic balance. Interestingly, while EVs showed minimal activity in assays measuring modulation of innate immunity, components larger than 100 kDa, potentially EVs, demonstrated dose-dependent effects on CD3⁺ T-cell proliferation, indicating distinct and possibly complementary mechanisms of action between the soluble and vesicular fractions. This functional compartmentalization suggests that both components may contribute to the therapeutic efficacy of MSC-derived products, but through different immunological pathways.

Future investigations should aim to further characterize the individual and synergistic roles of these secretome fractions, including identifying key bioactive molecules and elucidating their mechanisms of action. Such insights will be crucial for optimizing manufacturing processes, standardizing potency assays, and improving the safety and efficacy of MSC-derived therapies. Overall, this work represents an important step toward a deeper understanding of the mechanisms of action underlying MSC-derived secretome and extracellular vesicles, which is essential for their successful clinical translation from bench to bedside.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ASC

Adipose stromal cells

cryoTEM

Cryo-transmission electron microscopy

%CV

Coefficient of variation

EFS

Etablissement Français du sang

EV

Extracellular vesicles

GW

GW627368X

HD

High dose

IDO

Indoleamine 2, 3-dioxygenase

ISEV

International Society of Extracellular Vesicles

LD

Low dose

MD

Medium dose

MSC

Mesenchymal stromal cell

PBMCs

Peripheral blood mononuclear cells

PC

Post clarification

pDCs

Plasmacytoid dendritic cells

PGE2

Prostaglandin E2

PT

Post TFF

SD

Standard deviation

SN

Supernatant

TFF

Tangential flow filtration

TG

TG4-155

Tregs

Regulatory T cells

UC

Ultracentrifugation

Author contributions

CW, JV, TF and JPH conceived and designed the work. CW produced EVs in 100 and 1 L spinner-flasks. CW and SL concentrated secretome by TFF. CW concentrated secretome by UC. CW analyzed multiplexed-dosage, characterized EVs by MACSPLEX and NTA, performed the immunomodulatory assay on NF-κB and IRF, and PGE2 dosage. CW and IR evaluated T-cell proliferation. IR acquired data on kynurenine levels. CW analyzed the data and performed the statistical analyses. All authors contributed to the general input and discussion/interpretation of the results. CW wrote the manuscript and assembled the figures. CW, JV, TF, NB and JPH substantively revised the manuscript. All authors reviewed the manuscript and approved of its final version.

Funding

CW is a student from the FIRE PhD program funded by the Bettencourt Schueller foundation and the EURIP graduate program (ANR-17-EURE-0012).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval and consent to participate

Blood samples used in this study were obtained from residual cytapheresis material provided by the French Blood Establishment (EFS), under agreement 2023-2026-024 CCPSL EVERZOM. Donors provide general consent for the use of residual blood products in research at the time of donation. Samples were fully anonymized and are considered biological waste; therefore, no additional ethics approval or informed consent was required according to French regulations.

Consent for publication

Not applicable in this study.

Competing interests

CW, IR, SL, JV, TF are employed by Everzom.