Small extracellular vesicle‐derived miR‐574‐5p regulates PGE2‐biosynthesis via TLR7/8 in lung cancer

Abstract Intercellular communication plays an essential role in lung cancer (LC). One of the major players in cell‐cell‐communication is small extracellular vesicles (sEV). SEV trigger various biological responses by transporting cellular cargo to target cells. One essential sEV component are microRNAs (miRs), whose transport has recently attracted increasing research interest. We report that prostaglandin E2 (PGE2), a key inflammatory lipid mediator, specifically induces the sorting of miR‐574‐5p in sEV of A549 and 2106T cells. We found that sEV‐derived miR‐574‐5p activates Toll‐like receptors (TLR) 7/8, thereby decreasing PGE2‐levels. In contrast, intracellular miR‐574‐5p induces PGE2‐biosynthesis. Consequently, the combination of intracellular and sEV‐derived miR‐574‐5p controls PGE2‐levels via a feedback loop. This was only observed in adeno‐ but not in squamous cell carcinoma, indicating a cell‐specific response to sEV‐derived miRs, which might be due to unique tetraspanin compositions. Hence, we describe a novel function of miR‐574‐5p unique to adenocarcinoma. Intracellular miR‐574‐5p induces PGE2 and thus the secretion of sEV‐derived miR‐574‐5p, which in turn decreases PGE2‐biosynthesis in recipient cells.


. Cell lines and cell culture conditions
To ensure reproducibility of the results, we adopted the main recommendations of the MISEV2018 guidelines (Thery et al., 2018) in the experimental setup. An overview of the applied major recommendations is given in Table S1.

. Lung tumour tissue
Lung tumour tissue samples were collected from 16 different patients with NSCLC (nine AC and seven SCC). The study protocol for tissue donation was approved by the ethics committee (Ethik Kommission am Fachbereich Humanmedizin der Justus-Liebig-Universität Giessen) of the University Hospital of Giessen and Marburg (Giessen, Germany) and is in accordance with national law and with Good Clinical Practice/International Conference on Harmonization guidelines. Written informed consent was obtained from each patient or the patient's next of kin (AZ 58/15) (25-27).

sEV purification
Directly after harvesting, cell culture supernatants were centrifuged at 2000×g and room temperature (RT) for 20 min. Afterward, 1 ml supernatant was centrifuged in an Optima™ XPN-80 ultracentrifuge (Beckman Coulter, Brea, USA) at 21,000×g and 4 • C for 1 h in a 1.5 ml polypropylene tube (Beckman Coulter, Brea, USA). The supernatant was transferred to a fresh tube and centrifuged at 100,000×g and 4 • C for 1 h. Finally, the supernatant was discarded and sEV pellets were resuspended in 1× phosphate buffered saline (PBS, Gibco, Carlsbad, USA). To ensure constant sEV concentrations between different experiments (Thery et al., 2018), the quantity of sEV was measured via UV-Vis spectroscopy at an absorption wavelength of 280 nm. Purified sEV were stored at 4 • C and used for experiments within 48 h.

sEV characterization
SEV populations were characterized (Thery et al., 2018) with the ExoView R100 platform (NanoView Biosciences, Boston, USA). Cell culture supernatants of A549 and 2106T cells were incubated and stained on tetraspanin chips (CD81, CD9, CD63, IgG control, NanoView Biosciences, Boston, USA) following the manufacturer's instructions. Briefly, unpurified cell culture supernatants were incubated on the chips for 18 h at RT. Chips were washed, blocked, and incubated with antibodies against CD9 (CF488A-labeled), CD63 (CF647-labeled), and CD81 (CF555-labeled) for 1 h at RT (all NanoView Biosciences, Boston, USA). Each antibody was previously diluted 1:600 in blocking solution. Finally, chips were washed, dried, and imaged with the ExoView R100 platform. The approximate particle count after ultracentrifugation was determined by interferometry (IM) vesicle sizing and single sEV cargo staining of Syntenin-1 using the ExoView R100 platform according to the manufacturer's instructions. IM measurement detected particles within a range of 50-200 nm. SEV numbers were determined with a Zetasizer Nano S (Malvern Panalytical, Malvern, UK). To this end, 1 ml of sEV-containing cell culture supernatant was measured in a 2.5 ml PMMA macro cuvette (Brand, Wertheim, GER). For the measurements, the refractive index of the material was set to 1.45 with an absorption of 0.001. The refractive index of the dispersant was set to 1.345 and the viscosity to 0.94 cP in order to resemble the characteristics of cell culture supernatant. Samples were measured at 21 • C for 20 s. For transmission electron microscopy (TEM), sEV from A549 and 2106T cells were purified and resuspended in 1× PBS. Then, 15 μl of purified sEV (protein concentration ∼0.025 mg/ml) were incubated on a formvar carbon coated nickel grid (Plano, Wetzlar, GER) for 10 min at RT. Samples were fixed with 2% formaldehyde (FA; Carl Roth, Karlsruhe, GER) for 10 min and then washed 3× with Milli-Q water (MQ) and post-stained with aqueous uranyl acetate (4%, w/v). SEV were imaged with a Zeiss EM109 electron microscope.

. Overexpression of miR--p in sEV
In order to overexpress miR-574-5p in sEV, the XMIRXpress Lenti system (System Biosciences, Palo Alto, USA) was used. A corresponding negative control was generated with the XMIRXP-NT system (System Biosciences, Palo Alto, USA). A549 cells were seeded at a density of 5 × 10 5 cells/well in a 6-well plate 24 h before transfection and were transfected with 2 μg of either miR-574-5p overexpression (oe) or negative control plasmid and Lipofectamine 2000 (Invitrogen, Karlsruhe, GER) according to the manufacturer's instructions. 2106T cells were seeded into 12-well plates with a density of 1 × 10 5 cells/well 24 h before transfection and transfected with polyethylenimine (PEI, Sigma-Aldrich, Darmstadt, Germany). In brief, 2 μg plasmid were mixed with 100 μl DMEM. 6 μl PEI (1 g/L, pH10) were mixed with 100 μl DMEM and vortexed well. Plasmid and PEI mixes were combined, vortexed, and incubated for 15 min. Afterward, the plasmid/PEI mixture was added to the cells dropwise. Both cell lines were allowed to produce miR-574-5p oe or negative control sEV for at least 18 h until supernatants were harvested. The overexpression was confirmed via real-time quantitative PCR (RT-qPCR) analysis.

. RNase and Triton X- treatment of sEV
According to the MISEV2018 guidelines (Thery et al., 2018), purified miR-574-5p sEV were either treated with 2.5 U/μl RNase I (New England Biolabs, Ipswich, USA) for 20 min at 37 • C or with 1% TritonX 100 (Carl Roth, Karlsruhe, GER) for 10 min at RT and then with 2.5 U/μl RNase I for 20 min at 37 • C. An untreated sample was used as a reference control. NEBuffer 3 (New England Biolabs, Ipswich, USA) was added to all samples. Total RNA was isolated using the miRNeasy Mini Kit (Qiagen, Hilden, GER). After RNA extraction, the amount of miR-574-5p was analysed via RT-qPCR.

. Western blotting
For Western blot analysis, cells were lysed with a tissue protein extraction reagent (T-PER, Thermo Fisher Scientific, Waltham, USA) for 30 min on ice. Protein concentrations were determined via Bradford assay (Bio-Rad Laboratories, Hercules, USA). Western blot analysis was performed with 20-30 μg protein as previously described . The membranes were incubated with specific primary antibodies against mPGES-

. PGE  enzyme-linked immunosorbent assay (ELISA)
For PGE 2 measurements, cells were first stimulated with sEV for 24 h and then treated with 10 μM arachidonic acid (AA, Sigma-Aldrich, Darmstadt, GER). 15 min after the addition of AA, supernatants were harvested and centrifuged at 17,000×g and 4 • C for 5 min to eliminate cell debris. Supernatants were transferred to fresh tubes and PGE 2 concentrations were measured using a PGE 2 ELISA (514010, Cayman Chemicals, Ann Arbor, USA) according to the manufacturer's instructions. All samples were measured in duplicates. The absorbance was determined with the Tecan Infinite M 2000 (Tecan Group, Männedorf, CH).

. Tetrazolium reduction assay
Cell proliferation and metabolism were assessed with a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazoliumbromide (MTT) (Carl Roth, Karlsruhe, GER) reduction assay. Cells were seeded at a density of 1 × 10 4 cells/well in a 96-well plate and treated with 2 μg/ml sEV (as described above) 24 h after seeding. After 24 h, cells were incubated with MTT (5 mg/ml in cell culture medium) for 3 h at 37 • C. The medium was removed and 100 μl DMSO/well were added. The quantity of formazan was measured with the Tecan Infinite M 2000 plate reader at an absorption wavelength of 570 nm. A reference measurement at 630 nm was performed to subtract nonspecific background.

. Statistics
Results are shown as the mean + standard error of the mean (SEM) of at least three independent experiments. Statistical analysis was carried out by Student's unpaired t-test (two-tailed) or one-way ANOVA with Tukey's multiple comparison test using Graph-Pad Prism 6.0. Experimental differences were considered as significant for P≤0.05 (indicated as * for P≤0.05, ** for P≤0.005, *** for P≤0.001, and **** for P≤0.0001, or § for P≤0.05, § § § for P≤0.001, and § § § § for P≤0.0001).

. MiR--p and mPGES- colocalize in human NSCLC tissue
To compare PGE 2 -biosynthesis in AC and SCC and to analyse if the miR-574-5p/CUGBP1 decoy mechanism is present in NSCLC tissue, we stained human lung tissue sections of AC and SCC patients for miR-574-5p and mPGES-1. To this end, miR-574-5p was visualized via ISH using complementary LNA-probes and mPGES-1 via IHC with a specific antibody. The use of serial sections allowed a comparison of all staining within the same tumour sample. In AC tissue sections, miR-574-p and mPGES-1 were localized within NSCLC cells. MiR-574-5p was detected in the nuclei, whereas mPGES-1 was present in the cytoplasm of cancer cells (Figure 1a, b). In SCC sections, miR-574-5p and mPGES-1 expression were much higher compared to AC. Both were present within the cancer cells. Like AC samples, miR-574-5p was located inside the nuclei and mPGES-1 in the cytoplasm. In addition, miR-574-5p and mPGES-1 were strongly expressed in inflamed/necrotic areas of the tissue (Figure 1c, d), which was not observed in AC samples ( Figure S1A, B). Staining controls did not show specific signals for miR-574-5p or mPGES-1 ( Figure  S1C). It is important to note that negative control ISH was performed on a tissue section from the same SCC tumour sample used for miR-574-5p ISH. Additionally, CUGBP1 was stained via IHC and detected in the nuclei of cancer cells in AC and SCC, although it was not present in inflamed/necrotic areas of SCC ( Figure S1D, E). CUGBP1 and mPGES-1 primary antibodies were of the same isotype. The comparison of both staining confirms that both antibodies specifically bound their target. Our results show that the interaction partners of the miR-574-5p/CUGBP1 decoy mechanism are present within the same cells in NSCLC tissue. Thus, they can theoretically interact with each other in NSCLC cells in order to modulate PGE 2 -biosynthesis.
F I G U R E  A549-and 2106T-derived sEV differ in their tetraspanin composition. (a, b) Unpurified A549-and 2106T-derived sEV were analysed using the ExoView R100 platform (NanoView Biosciences). SEV were captured at specific antibody-coated spots against CD81, CD63, and CD9 (bar graphs). Captured sEV were further analysed for tetraspanin colocalization with specific fluorescent antibodies (pie charts). (c, d) The size distribution of A549-and 2106T-derived sEV was analysed using the ExoView R100 platform. Data are presented as mean +SEM of two independent biological experiments with three technical replicates In order to further analyse PGE 2 -biosynthesis in both tumour types, we generated spheroid mono-and co-cultures of A549 (AC) and 2106T (SCC) cells with human lung fibroblasts ( Figure S2). We chose the 3D model system over conventional 2D cell culture since it more closely mimics in vivo tumour characteristics such as nutrient gradients and cellular interactions. Also, it was shown that spheroid cultures constitute a more physiological model to study PGE 2 regulation (Kock et al., 2020). In general, 2106T spheroids revealed higher protein and mRNA levels of mPGES-1 and COX-2 than A549 spheroids ( Figure S2A, B, S3A-D). While IL-1β stimulation was capable of inducing mPGES-1 only in A549 spheroids, COX-2 was induced in spheroids of both cell types. There were no differences in CUGBP1 or miR-574-5p levels between both cell types ( Figure S2C, D). But miR-574-5p levels were slightly decreased with IL-1β stimulation and further decreased upon co-cultivation with fibroblasts in all spheroids ( Figure  S2D). In 2106T spheroids, decreased miR-574-5p levels did not coincide with a lower expression of nervous system overexpressed protein 20 (NOXP20), which carries the precursor of miR-574-5p in its intron (Zhang et al., 2014) (Figure S3F). Since NOXP20 mRNA levels usually correlate with miR-574-5p levels, this suggests that other factors, like an increase in miR-574-5p secretion, may be responsible for the reduced levels in 2106T spheroids.
In summary, our results confirm that A549 and 2106T cells both express the interaction partners of the miR-574-5p/CUGBP1 decoy mechanism. However, differences in the expression profile and inducibility of PGE 2 -synthesizing proteins indicate a celltype-specific regulation of PGE 2 -biosynthesis.

. A-and T-derived sEV differ in their tetraspanin composition
In order to further characterize the observed differences between A549 and 2106T cells, we next focused on the role of sEV in cell-cell-communication, which is a crucial factor in the progression of NSCLC (Zheng et al., 2018). To this end, we analysed sEV of A549 and 2106T cells using the ExoView R100 platform. This system allows the characterization of unpurified sEV, which is a great advantage because manipulation of sEV populations during the purification process is avoided (Bachurski et al., 2019;Daaboul et al., 2016). In our analysis, we focused on the characterization of the tetraspanin composition. The ExoView R100 platform can identify colocalized tetraspanins on a single sEV level by measuring CD81, CD63, and CD9 in the same sample. We observed that A549-and 2106T-derived sEV strongly differed in their CD63 levels (Figure 2a, b). While most 2106T-derived sEV were bound to the CD63 spots (47% of total sEV), only 16% of total A549-derived sEV were bound to CD63 spots. In 2106Tderived sEV, the CD81 spots showed the lowest number with 20% of total sEV. In contrast, most A549-derived sEV were bound to the CD9 spot with 47% of total sEV. Next, we analysed the colocalization of CD81, CD63 and CD9 (Figure 2a, b, pie charts). CD81 and CD9 positive sEV showed very similar colocalization patterns in both cell types. In contrast, CD63 positive sEV revealed cell-specific differences. In 43% of A549-derived CD63 positive sEV, CD9 and CD63 colocalized. 14% showed colocalization of CD81 and CD63, while 12% of sEV were positive for all three tetraspanins. 31% of CD63 positive A549-derived sEV did not colocalize with other tetraspanins. In contrast, in 2106T-derived sEV, 80% were only positive for CD63. 13% of CD63 positive 2106T-derived sEV showed colocalization of CD9 and CD63, while 5% were positive for CD81 and CD63. Only 2% of all CD63 positive 2106T-derived sEV were positive for all three tetraspanins. These results show that A549-and 2106T-derived sEV greatly differ in their tetraspanin composition, which might influence the interaction between these sEV and possible target cells. Finally, we analysed the size distribution of A549-and 2106T-derived sEV (Figure 2c, d). Both sEV types had a small number of sEV with a diameter of 100-200 nm. Most sEV had a diameter of 50 nm, which also marks the detection limit of the ExoView R100 platform. Therefore, it is likely that A549 and 2106T cells also secrete sEV smaller than 50 nm, which were not detected with this approach. Taken together, our results indicate that A549-and 2106T-derived sEV did not differ in size distribution but showed a different tetraspanin composition.
The unique tetraspanin composition of A549-and 2106T-derived sEV constitutes another crucial difference between both cell types, which could have a great influence on the interaction with target cells.

. PGE  induces the secretion of miR--p in sEV
Tissue staining and spheroid cultures demonstrated that miR-574-5p and mPGES-1 are localized in NSCLC cells and that PGE 2biosynthesis is specifically regulated in A549 and 2106T cells. Moreover, we showed that A549-and 2106T-derived sEV had unique tetraspanin compositions. Thus, we aimed to further analyse the sEV secretion in both cell types. Since it is known that PGE 2 is overexpressed in many NSCLC cases, we stimulated A549 and 2106T cells with 5 nM PGE 2 to mimic the inflammatory environment of the tumour. After 2 and 8 h, we purified sEV from cell culture supernatants using differential ultracentrifugation and isolated the total RNA of sEV. MiR-574-5p, -21-5p, -486-5p, and -16-5p levels were quantified via RT-qPCR. These miRs are known to exert physiological functions in LC and are discussed as functional biomarkers (Foss et al., 2011;Han et al., 2020;Liu et al., 2017;Reis et al., 2020;Tian et al., 2019). We found that only miR-574-5p was significantly enriched in A549-and 2106Tderived sEV after PGE 2 stimulation (Figure 3a, b). A549-derived sEV showed a significant 3-fold increase in miR-574-5p 2 h after stimulation (Figure 3a). 8 h after PGE 2 treatment, A549 cells secreted sEV with reduced miR-574-5p content. Like A549 cells, 2106T cells also secreted miR-574-5p in sEV in response to PGE 2 . However, we observed the highest level of miR-574-5p after 8 h, not after 2 h as in A549 cells (Figure 3b). In addition, we measured miR-574-5p secretion in another NSCLC cell line in response to PGE 2 stimulation. The results showed that the H1650 cell line was also prone to miR-574-5p secretion 8 h after stimulation ( Figure S4A). Finally, the secretion of other miRs was not significantly affected by PGE 2 stimulation in 2106T and A549 cells. This is particularly interesting since the basal level of miR-574-5p is generally lower than that of the other miRs in both cell lines ( Figure S4B, C).
We next examined whether the increase in sEV-derived miR-574-5p was associated with a higher expression of miR-574-5p ( Figure S4D, E). We did not observe a significant increase in intracellular miR-574-5p after PGE 2 stimulation in A549 or 2106T cells. Thus, the increased secretion of miR-574-5p was not caused by higher intracellular expression levels of miR-574-5p.
To test whether the increase in sEV-derived miR-574-5p correlated with higher sEV numbers or higher miR-574-5p levels per vesicle, we measured sEV numbers using a Zetasizer Nano S ( Figure S4F, G). There were no significant changes in sEV numbers in response to PGE 2 treatment which suggests that miR content per sEV was increased.
Additionally, we blocked endogenous PGE 2 -biosynthesis with the mPGES-1 inhibitor CIII (Foss et al., 2011). Both cell lines showed a significant decrease in sEV-derived miR-574-5p upon mPGES-1 inhibition (Figure 3c). These results indicate that both external stimulation with PGE 2 and intracellular PGE 2 -synthesis regulate miR-574-5p secretion. To assess which PGE 2 receptor (EP receptor) induces miR-574-5p secretion after PGE 2 stimulation, we treated A549 and 2106T cells with EP receptor agonists. Stimulation with EP1/EP3 agonists resulted in increased secretion of miR-574-p in A549 and 2106T cells (Figure 3d), while EP2 and EP4 agonist stimulation did not (Figure 3e, f). Here, we observed different secretion time points compared to PGE 2 stimulation, which is probably linked to different binding affinities and component stabilities of the agonists and PGE 2 .
In sum, our results strongly indicate that PGE 2 rapidly and specifically induces the sorting of miR-574-5p into sEV of A549 and 2106T cells via EP1/3 receptors.

. Overexpression of miR--p in A-and T-derived sEV
To analyse the physiological function of sEV-derived miR-574-5p, we established an overexpression system for both cancer cell lines as previously established in HEK-293 cells (Hegewald et al., 2020). The system increases the loading of miR-574-5p into sEV (miR-574-5p oe sEV). Control experiments were performed with sEV loaded with a scrambled miR (ScrC sEV). We compared miR-574-5p levels of both sEV types via RT-qPCR and detected a ∼9-fold increase of miR-574-5p in A549derived oe sEV (Figure 4a). 2106T-derived oe sEV showed a ∼4-fold increase of miR-574-5p compared to ScrC sEV (Figure 4b). We purified sEV via differential ultracentrifugation and confirmed the successful isolation of sEV with TEM analysis ( Figure S5A). To determine the number of sEV isolated after ultracentrifugation, we performed a representative analysis of oe sEV of 2106T cells with ExoView R100 and determined an approximate particle count of 30,000 vesicles per ml of supernatant ( Figure S5B). To verify whether miR-574-5p was loaded into sEV and not attached to the outside, we performed an exemplary RNase protection experiment with A549-and 2106T-derived miR-574-5p oe sEV as we had done previously (Shurtleff et al., 2016). For this purpose, sEV were treated with RNase alone or in combination with a detergent, followed by RT-qPCR analysis ( Figure S5C, D). We showed that the majority of miR-574-5p was protected within sEV with engineered miR-574-5p level. The disruption of sEV followed by an RNase digest led to a significant decrease in miR-574-5p levels. Hence, miR-574-5p was protected in samples without detergent, indicating that most of the miRs were located within sEV.
Finally, we showed that miR-574-5p oe sEV were taken up by A549 and 2106T cells using live cell imaging (Figure 4b, d). Both cell types started to internalize sEV within a few minutes. While most A549-derived sEV were distributed within the whole cell, some 2106T-derived sEV seemed to accumulate at the cell membrane (Figure 4d, white arrows). To find out whether the accumulation of sEV at cell membranes was vesicle-or cell-specific, 2106T and A549 cells were also treated with sEV of the other cell line. When 2106T cells were incubated with A549-derived sEV, an increased accumulation of sEV at the cell membrane was observed 30 min after addition of the sEV ( Figure S6A). In contrast, A549 cells appeared to take up 2106T-derived sEV (S6B). This suggests a cell-specific interaction between cells and sEV.
Either way, we were able to show that both cell lines internalize engineered sEV, which enables a physiological interaction. To control for cytotoxic effects of miR-574-5p oe sEV, we performed a tetrazolium reduction assay ( Figure S7A, B). To this end, A549 and 2106T cells were stimulated with miR-574-5p oe and ScrC sEV. 24 h after sEV treatment, we did not observe any impact on cell viability or cell metabolism caused by treatment with miR-574-5p oe sEV. We thus conclude that we successfully established cell-specific miR-574-5p oe and control sEV to analyse the physiological role of sEV-derived miR-574-5p. These engineered sEV are internalized by both cell types and do not exert cytotoxic effects on cells.
After 24 h, we analysed mPGES-1 protein levels by Western blot analysis. In A549 cells, ScrC sEV led to a significant 1.6-fold increase in mPGES-1 protein expression compared to the untreated control (Figure 5a and Figure S7D). In contrast, stimulation with miR-574-5p oe sEV led to a slight decrease by 0.9-fold. Compared to ScrC sEV treated samples, miR-574-5p oe sEV significantly decreased mPGES-1 protein levels by 0.5-fold. Treatment with the TLR7/8 ligand resulted in significantly lower mPGES-1 levels of 0.7-fold compared to ScrC sEV-treated samples and the untreated control. Blocking TLR7/8 receptors significantly prevented the mPGES-1 decrease induced by miR-574-5p oe sEV or R848. Overall, these results indicate that activation of TLR7/8 via sEV-derived miR-574-5p regulates mPGES-1 in A549 cells. Moreover, regulation of mPGES-1 also affected PGE 2 -levels (Figure 5c). We observed significantly lower PGE 2 -levels in A549 cells treated with miR-574-5p oe sEV or R848. Administration of the TLR7/8 inhibitor also blocked this downregulation.
Because miR-574-5p is a major post-transcriptional regulator of PGE 2 -biosynthesis, we measured intracellular miR-574-5p levels after treatment with miR-574-5p oe sEV ( Figure S7C). As expected, we observed an increase in intracellular miR-574-5p in response to sEV uptake 1 h after treatment. Interestingly, intracellular miR-574-5p was significantly decreased 3 h after sEV treatment compared to the 1 h values and the untreated control. These results suggest that intracellular miR-574-5p levels are also regulated by sEV-derived miR-574-5p. As observed for mPGES-1 regulation, this may also be due to the activation of TLR7/8 in A549 cells.
In contrast to A549 cells, we did not observe significant effects in 2106T cells using the same experimental setup (Figure 5b, Fig. S7E). 2106T cells did not show regulation of mPGES-1 upon stimulation with either the TLR7/8 ligand R848 or the TLR7/8 antagonist ODN 2088 Control (ODN 2087). Thus, our results indicate that A549 and 2106T cells respond cellspecifically to sEV-derived miR-574-5p. While miR-574-5p oe sEV activated TLR7/8 signalling in A549 cells and thus regulated mPGES-1-dependent PGE 2 -biosynthesis, this effect was not observed in 2106T cells. Similar results were shown with the TLR7/8 ligand. Taken together, these results indicate that 2106T cells do not respond to TLR7/8 activation as A549 cells do.

 DISCUSSION
A better understanding of intercellular communication in the tumour microenvironment is a valuable basis for developing new cancer treatment approaches. Recent studies identified sEV as key players in processes that contribute to disease development (Mathieu et al., 2019). This is mainly mediated by sEV cargo. One major cargo is miRs, which are highly associated with the hallmarks of cancer (Bhome et al., 2018;Garzon et al., 2006). In addition, it is well known that tumour development can also be induced by inflammatory mediators (Mantovani et al., 2008). But the interaction between inflammatory signalling and sEVderived miRs in the tumour microenvironment has not yet been studied. A key inflammatory mediator in cancer development is PGE 2 , which promotes tumour progression, immune evasion, and angiogenesis (Wang & Dubois, 2010). In this study, we demonstrate that PGE 2 significantly induces sEV-miR-574-5p secretion in A549, 2106T and H1650 cells. This is of particular interest because intracellular miR-574-5p regulates mPGES-1-dependent PGE 2 -biosynthesis in lung AC (Emmerich et al., 2020;Saul et al., 2019). This increased secretion of sEV-derived miR-574-5p was not caused by higher intracellular miR-574-5p levels, nor by increased sEV numbers. These results demonstrate for the first time that PGE 2 specifically triggers the sorting of a miR, here miR-574-5p, into sEV and that this sorting is mediated by EP1 and EP3 receptors. Several studies show that increased miR-574-5p plasma levels of LC patients correlate with tumour development (Foss et al., 2011;Han et al., 2020;Peng et al., 2016). Particularly interesting is the fact that the level of circulating sEV-derived miR-574-5p is not only elevated in LC patients compared to healthy controls, but also significantly decreases after tumour resection (Han et al., 2020). Taken together with the fact that miR-574-5p specifically regulates PGE 2 -biosynthesis in NSCLC (Emmerich et al., 2020;Saul et al., 2019), miR-574-5p can be considered as a promising candidate for a minimally invasive biomarker for the stratification of NSCLC patients . We showed that the localization of miR-574-5p and mPGES-1 in NSCLC tumour tissues correlated and that AC and SCC differ in the expression of miR-574-5p and mPGES-1 in tumour tissues and 3D tumour models. Furthermore, PGE 2 -induced secretion of miR-574-5p was significantly reduced by a PGE 2 inhibitor. These results support the hypothesis that the level of extracellular miR-574-5p correlates with the level of PGE 2 in the tumour. This will not only enable an initial patient classification, but also the monitoring of treatment success. Hence, the suitability of miR-574-5p as biomarker for NSCLC should be further investigated in clinical studies.
Next, we analysed cell-specific sEV characteristics. Our results revealed a unique tetraspanin composition on the surface of sEV for A549 and 2106T cells. Other studies already showed that the tetraspanin composition on sEV probably influences the selection of target cells and the mechanism of sEV uptake (Horibe et al., 2018;Rana et al., 2012). Also, the expression of surface proteins and receptors on the cell membrane strongly influences sEV uptake. Since the composition of membrane proteins is unique to different cell types, cell-specific sEV uptake is likely (Jadli et al., 2020). Taken together, the membrane characteristics of sEV and recipient cells allow cell type-specific responses to sEV. We speculate that A549 and 2106T cells respond differently to sEV because some 2106T-derived sEV accumulated at the cell membrane. Thus, specific interactions with sEV and different uptake mechanisms can be expected.
Hence, we describe a novel function for miR-574-5p, which is dependent on the transfer via sEV. So far, it was not clear whether sEV only transfer miRs or also impact their physiological function. We describe that intracellular and sEV-derived miR-574-5p exerts opposing functions on the PGE 2 -biosynthesis. Depending on the uptake mechanism, an sEV-derived miR can be released at different locations within the cell (McKelvey et al., 2015). In A549 cells, sEV-derived miR-574-5p can interact with the endosomal TLR7/8 receptors and thus decrease PGE 2 -levels. Consequently, sEV transfer miR-574-5p to a novel subcellular location which is not reached by intracellular miR-574-5p. This demonstrates that sEV not only act as miR shuttles, but are also capable of changing their mode of action. Contrary to A549 cells, we did not observe the described mechanism in 2106T cells. In these cells, sEV-derived miR-574-5p did not affect PGE 2 -synthesis, although 2106T cells express TLR7/8 similarly to A549 cells ( Figure S7F, G). The lack of this regulatory mechanism could explain the higher mPGES-1 levels in 2106T spheroid cultures compared to A549 spheroids. The fact that sEV-derived miR-574-5p does not exert an autocrine function on 2106T PGE 2 -biosynthesis, supports our initial theory of cell-specific interactions with sEV. This is consistent with our live cell experiments which indicate a cell-typespecific sEV uptake and with other studies describing cell-type-specific uptake mechanisms for similar sEV (Horibe et al., 2018). Furthermore, it is conceivable that 2106T-derived sEV physiologically target other recipient cells rather than interacting with their donor cell. Specific targeting of recipient cells has been shown previously for other cell types (Denzer et al., 2000;Mallegol et al., 2007). Here, we show that A549-and 2106T-derived sEV interact differently with their donor cells and elicit different physiological responses when it comes to mPGES-1-dependent PGE 2 -biosynthesis. The observation that sEV-derived miR-574-5p decreases PGE 2 -levels in A549 cells stands in contrast to other studies with immune cells that showed increased PGE 2 -synthesis after TLR7/8 activation (Hattermann et al., 2007;Salvi et al., 2016). We attribute this discrepancy mainly to cell-specific effects. One possible downstream signalling pathway after TLR7/8 activation in A549 cells is nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB) (Dowling, 2018). Interestingly, NFκB-p65, a subunit of the NFκB transcription complex, significantly reduces intracellular miR-574-5p (Ku et al., 2017). NFκB-p65 can regulate transcription within 30 min after activation (Zambrano et al., 2014). This is in line with our results. We demonstrated that sEV-derived miR-574-5p reduces intracellular miR-574-5p expression. Therefore, we hypothesize that sEV-derived miR-574-5p acts as a TLR7/8 ligand and activates NFκB signalling which results in decreased intracellular miR-574-5p and PGE 2 -levels. This assumption supports the hypothesis that TLR7/8 activation has antitumoral effects (Michaelis et al., 2019;Schön & Schön, 2008).
In sum, we demonstrated that PGE 2 modulates the sorting of miR-574-5p into sEV of A549 and 2106T cells. Furthermore, we analysed the physiological function of sEV-derived miR-574-5p in recipient cells and showed that it decreases mPGES-1dependent PGE 2 -biosynthesis in A549, but not in 2106T cells. This cell-specific PGE 2 regulation is of great interest for the development of new therapeutic approaches in the treatment of NSCLC. The benefits of PGE 2 inhibitors for the treatment of LC are well known. Our results can contribute to the development of new subtype-specific therapies based on a more efficient use of PGE 2 inhibitors. However, the results of this study should be characterized in additional AC and SCC cell lines, primary cancer cells, and in vivo models. Future studies need to further investigate the relationship between sEV uptake and the physiological function of sEV in the tumour microenvironment to unravel the therapeutic potential of our findings. F I G U R E  Schematic diagram of the PGE 2 feedback loop regulation mediated by miR-574-5p and TLR7/8. The miR-574-5p/CUGBP1 decoy regulates PGE 2 -biosynthesis intracellularly. High intracellular miR-574-5p induces PGE 2 -biosynthesis. PGE 2 triggers the secretion of miR-574-5p in sEV. In recipient A549 cells, sEV-derived miR-574-5p activates TLR7/8 signalling, which leads to decreased miR-574-5p, mPGES-1 and PGE 2 -levels Finally, while intracellular miR-574-5p induces PGE 2 -biosynthesis via a decoy mechanism with CUGBP1, sEV-derived miR-574-5p exerts opposite functions on the regulation of PGE 2 . This novel mechanism is mediated by the transfer of miR-574-5p via sEV, which enables an interaction with endosomal TLR7/8. Hence, the interplay between intracellular and sEV-derived miR-574-5p controls PGE 2 -levels in A549 cells via a feedback loop ( Figure 6). This study demonstrates for the first time that the same miR can have different functions depending on whether it is expressed inside a cell or transferred to a cell via sEV.

A C K N O W L E D G E M E N T S
We are grateful to Prof. Dr. Beatrix Suess (TU Darmstadt, Germany) for proofreading our manuscript and scientific support. We are grateful to Prof. Dr. Dieter Steinhilber (Goethe University Frankfurt, Germany) for scientific support. We thank Prof. Dr. M. Cristina Cardoso (TU Darmstadt, Germany) for providing access to her microscope facilities and Prof. Dr. Oliver Clemens (University of Stuttgart, Germany) for providing access to the Zetasizer Nano S. We also thank Dr. Siavash Mansouri (MPI, Bad Nauheim) for technical and experimental support. In addition, we would like to thank Prof. Dr. Alexander Loewer (TU Darmstadt, Germany) for providing access to his microscope facility and Petra Snyder (TU Darmstadt, Germany) for technical support. Finally, we thank Kai Breitwieser for proofreading and experimental support, Leon Koch for experimental support and Maria Fauth for proofreading.
Open access funding enabled and organized by Projekt DEAL.

C O N F L I C T O F I N T E R E S T
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

AU T H O R C O N T R I B U T I O N S
Julia Donzelli and Eva Proestler performed the experiments and analysed the data. Julia Donzelli wrote the manuscript. Eva Proestler contributed to manuscript writing. Anna Riedel performed spheroid experiments and supported secretion experiments. Sheila Nevermann performed RNase protection assay experiments, supported secretion experiments and contributed to manuscript writing. Brigitte Hertel performed the TEM analysis. Rajkumar Savai, Andreas Guenther and Stefan Gattenlöhner provided human lung tissue samples. Rajkumar Savai and Karin Larsson supported the tissue staining experiments. Meike J. Saul conceived the study, designed, and supervised the overall project, and wrote the manuscript. All authors conducted the quality assurance of the paper and reviewed the manuscript.