Extracellular vesicles produced by NFAT3-expressing cells hinder tumor growth and metastatic dissemination

Metastases are the main cause of cancer-induced deaths worldwide. To block tissue invasion, development of extracellular vesicles (EVs) as therapeutic carriers, appears as an exciting challenge. To this aim, we took advantage of the anti-invasive function of NFAT3 transcription factor we identified previously in breast cancer and addressed the opportunity to transfer this inhibitory function by EVs. We show here that EVs produced by poorly invasive NFAT3-expressing breast cancer cell lines are competent to block in vitro invasion of aggressive cancer cells from different origins and, in cooperation with macrophages, inhibit cell proliferation and induce apoptosis. Moreover, this inhibitory effect can be improved by overexpression of NFAT3 in the EVs-producing cells. These results were extended in a mouse breast cancer model, with clear impact of inhibitory EVs on tumor growth and metastases spreading. This work identifies EVs produced by NFAT3-expressing breast cancer cells as an anti-tumoral tool to tackle cancer development and metastases dissemination.

Thus, in the present study, we evaluate the use of EVs as endogenous mediators to convey NFAT3 inhibitory properties and target cancer cells both in vitro and in vivo. Indeed, we show that EVs produced by low invasive luminal breast cancer cells are fully competent to inhibit both cell invasion in vitro of cancer cells from different origins and in vivo metastases formation in a mice model of breast cancer. Furthermore, besides blocking metastases arising, we demonstrate in vitro that these EVs are strong inhibitors of tumor growth in cooperation with macrophages. Strikingly, these EVs inhibitory effects rely on the expression of NFAT3 by EVs-producing cells, yet without any detectable transfer of NFAT3 to the recipient cells. To note, increase of NFAT3 expression in the EVs-producing cells appeared to be sufficient to significantly enhance EVs inhibitory function both in vitro and in vivo. Together, these results provide fundamental basis to develop and potentiate new EVs-derived therapeutic tools to treat aggressive cancers.

Results
EVs produced by poorly aggressive luminal breast cancer cells have anti-invasive properties in vitro on different cancer cell types. Having shown that NFAT3, significantly more expressed in luminal breast cancer, inhibits breast cancer cell invasion 9 , we evaluate here the possibility that EVs produced by luminal breast cancer cells might be competent to transfer this inhibitory capacity by NFAT3 to triple negative breast cancer cells lines. To this end EVs were isolated from conditioned medium of different cell lines, purified by the classical ultracentrifugation method and characterized by specific EV markers CD63, CD81 and Calnexin (Fig. S1). The size and concentration of MDA-MB-231 and T-47D EVs were determined by NTA (Nanoparticle Tracking Analysis) allowing to estimate the amount of EVs per producing cells (Fig. S1A).
To study their potential effect on the invasive capacity of triple negative breast cancer cell lines, we first treated the triple negative MDA-MB-231 breast cancer cells with EVs produced by luminal T-47D breast cancer cells. As controls, we tested on the same cell line the effect of EVs produced by MDA-MB-231 or by normal human fibroblasts originated from two different healthy donors (FHN21, FHN32) (Fig. 1A). Among the different EVs produced, only those originated from T-47D cells were reproducibly efficient in inhibiting MDA-MB-231 cell invasion compared to the EVs from other sources (Fig. 1A). Conversely, EVs produced by highly invasive MDA-MB-231 cells were able to significantly enhance T-47D cell invasion (Fig. 1B) as previously reported in vivo by Zomer et al. 5 . To confirm the inhibitory effect of the T-47D -produced EVs, we showed that these EVs were reproducibly able to inhibit the invasion of SUMP-159-PT, another triple negative breast cancer line (right panel, Fig. 1C). Moreover, this anti-invasive effect was not specific to the EVs produced by T-47D cells since EVs produced by MCF7, another luminal breast cancer cell line, were shown to be as efficient in blocking cell invasion of both MDA-MB-231 and SUM-159-PT triple negative breast cancer cell lines (left panel, Fig. 1C). These results demonstrate for the first time the specificity and capacity of EVs produced by luminal breast cancer cells to inhibit the invasion of highly aggressive triple negative breast cancer counterparts.
Next, we explored whether this inhibitory effect of T-47D-produced EVs could have a broader spectrum by examining their effect on the cell invasion of other aggressive cancer cell lines from melanoma (WM.266.4), glioblastoma (U87MG) or pancreatic origin (BXPC3). Indeed, results presented in Fig. 1D unveil that the anti-invasive function of EVs produced by T-47D cells was not restricted to highly invasive breast cancer cells but could be extended to other highly aggressive cancer cells from various origins. To note, EVs had no impact on cell proliferation and apoptosis (Fig. S1B). Altogether, these results pinpoint that EVs produced by luminal breast cancer cell lines are fully proficient in impairing highly aggressive cancer cells invasive capacity in vitro.
The anti-invasive capacity of EVs produced by T-47D is linked to the expression of NFAT3 in T-47D producing cells. To decipher the mechanisms responsible for the anti-invasive effect of the EVs generated by the luminal breast cancer cells, we took advantage of our previous work showing that luminal breast cancer cells specifically express NFAT3, an anti-invasive isotype of the NFAT transcription factors family 9 . We hypothesized that the endogenous NFAT3 expressed in T-47D cells might confer its inhibitory function to the derived-EVs. To validate this hypothesis, we generated stable T-47D cell lines expressing lentiviral constructs encoding either a shRNA control (T-47D shCtl) or two different shRNAs targeting endogenous NFAT3 (T-47D shNFAT3-3, T-47D shNFAT3-4) that reduced up to 60% of NFAT3 expression compared to the shCtl expressing cells ( Fig. 2A, left). As previously shown 9 , we confirmed that reducing endogenous NFAT3 expression in T-47D cells increased by two times their invasive capacity ( Fig. 2A, right). We produced EVs from these different T-47D clones and verified their homogeneity in size by NTA (Fig. S3). We did not observe any effect of NFAT3 downregulation on EVs production efficiency (Fig. S3A). In addition, and conversely to what we observed for T-47D-produced EVs, T-47D shNFAT3-3 and T-47D shNFAT3-4 produced EVs were ineffective to inhibit cell invasion neither of triple negative breast cancer cell lines (Fig. 2B), nor glioblastoma or pancreatic cell lines (Fig. 2C). These results demonstrate that EVs produced by poorly invasive luminal breast cancer cells are fully competent at inhibiting aggressive cancer cells invasion and rely on NFAT3 expression in the EVs producing cells.

NFAT3-expressing luminal breast cancer cells produced EVs are fully competent at inhibiting in vivo breast tumor growth and metastases burden. Since cell invasion is critical for metastases
spreading in vivo, we sought to examine whether these anti-invasive EVs might also be competent in preventing metastases burden in a murine breast cancer model. To this end we xenografted MDA-MB-231-luciferase positive cells (D3H2LN-LUC) in mice fat pad. We generated 4 groups of mice that were subjected to a weekly intra-tumor injection, starting 1 week after cells implantation, with either PBS as a control or the different EVs originated from T-47D shCtl cells or T-47D shNFAT3-3/shNFAT-4 cells. Tumor growth was evaluated once a week for each group by caliper measurements and after one month, metastases apparition was quantified once a week by luminescence with an IVIS in vivo imaging system.
During the first 30 days tumor growth was equivalent in all the 4 mice groups (Fig. 2D, Tumor volume, day 0-28). Strikingly, from day 42, only the group treated with the EVs produced by T-47D shCtl cells showed a and left untreated or were treated the following day with 3 × 10 8 pp/mL EVs isolated from by WT T-47D; from WT MDA-MB-231 or from 2 different female primary human dermal fibroblasts (FHN21, FHN32) and subjected to in vitro invasion assay for 6 h. Data from one representative experiment of two independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; **p < 0.005). (B) Poorly invasive luminal breast cancer cells T-47D were serum starved for 24 h and left untreated or were treated the following day with 3 × 10 8 pp/mL EVs produced by WT MDA-MB-231 subjected to in vitro invasion assay for 24 h. Data from one representative experiment of two independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; ***p < 0.001, compared to the untreated cells). (C) Highly invasive triple negative breast cancer cells MDA-MB-231 (left panel) and SUM-159-PT (right panel) were serum starved for 24 h and left untreated or treated the following day with 3 × 10 8 pp/mL EVs produced by WT T-47D or by WT MCF7 and subjected to an in vitro invasion assay for 6 h. Data from one representative experiment of two independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; **p < 0.005, compared to the untreated cells). (D) Highly invasive melanoma (WM.266.4), glioblastoma (U87MG) and pancreatic cancer cells (BXPC3) were serum starved for 24 h and left untreated or were treated the following day with 3 × 10 8 pp/mL EVs produced by WT T-47D and subjected to in vitro invasion assay for 6 h. Data from one representative experiment of two independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; **p < 0.005, compared to the untreated cells). For all data, the invasion index is calculated as a proportion of the number of invasive cells in treated wells compared to the number of invasive cells in the control well (−) arbitrarily set to 1. Whole cell lysates were revealed by an anti-NFAT3 (α-NFAT3) and normalization was done by revelation with an anti-actin (α-actin) on the same blot after cutting the membrane. Right panel: T-47D shCtrl, shNFAT3-3 or shNFAT3-4 were serum starved for 24 h and the following day subjected to an in vitro invasion assay for 24 h. Data from one representative experiment of three independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; **p < 0.005, compared to the untreated cells). (B) Highly invasive MDA-MB-231 or SUM-159-PT cells were serum starved and pre-treated or not with 3 × 10 8 pp/mL EVs produced by WT T-47D (T-47D) or T-47D shCtrl, T-47D shNFAT3-3, T-47D shNFAT3-4 and tested for their invasion for 6 h. Data from one representative experiment of three independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; **p < 0.005, compared to the untreated cells). (C) Glioblastoma U87MG cells or pancreatic cancer cells BXPC3 were treated as described above and tested for their invasion for 6 h. Data from one representative experiment of three independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; **p < 0.005, ***p < 0.001, compared to the untreated cells). For all data, the invasion index is calculated as a proportion of the number of invasive cells in Scientific RepoRtS | (2020) 10:8964 | https://doi.org/10.1038/s41598-020-65844-x www.nature.com/scientificreports www.nature.com/scientificreports/ This suggest that other mechanisms are engaged upon EVs treatment to favor the inhibition of tumor growth in physiological mode.
Moreover, we completed this experimental set by analyzing metastasis burden, one month after cancer cells injection. As shown in Fig. 2D (right panel, Metastases), the group of mice injected with the EVs produced by T-47D shCtl cells present a significative inhibition of metastases burden, from day 42 until the end of the experiment, to reach 99,5% of inhibition compared to the group of mice injected with either PBS or EVs produced by T-47D shNFAT3-3/shNFAT3-4. Representative images of metastases burden are shown for day 59 in Fig. S3B.
As a complement of our in vitro results, these later data show that EVs produced by NFAT3-expressing low invasive breast cancer cells can inhibit metastases spreading but also tumor growth of an aggressive triple negative breast cancer cell line.

EVs from NFAT3-expressing cells require macrophages to inhibit cell growth and to induce apoptosis in breast cancer cells.
Even though we were unable to detect anti-proliferative or pro-apotoptic effects of the T-47D-produced EVs in vitro, we reproducibly observed an inhibitory effect of these EVs on tumor growth in vivo. As for the anti-invasive effects observed, these inhibitory effects on the tumor growth were linked to NFAT3 expression in the EVs-producing cells. Therefore, we questioned the 2D cell culture system previously used and set up a protocol for breast cancer cell spheroids culture (see Mat & Meth section). Cells were seeded and cultured for 3 days in low attachment plates embedded in 3% Matrigel and spheroids size was monitored using the INCUCYTE device. Medium containing the apoptosis indicator (fluorescent caspase 3/7 substrate) was added to cells and supplemented with or without the T-47D-produced EVs. Again, we could not identify any anti-proliferative or pro-apoptotic effects of the EVs in these conditions (Fig. 3A).
To further explain the difference between our in vitro and in vivo results, we chose to monitor cell microenvironment. Indeed, it has been shown in vivo, that the tumoral microenvironment constituted by many cells, such as cancer associated fibroblasts and macrophages, can have a direct impact on tumor growth 14 . In our experimental conditions, this cell contingent might play crucial role and be necessary for the EVs-induced-inhibition of the tumor growth. Therefore, we evaluated by immunofluorescence the presence of macrophages in the tumor we generated by xenotransplantation of the MDA-MB-231 cells in the fat pad of the athymic nude mice. Human tumor cells were detected based on human cytokeratins expression and mouse macrophages owing to the detection of the F4/80 expression marker. As shown in Fig. 3B, mouse macrophages (anti-F4/80) were found to infiltrate the engrafted tumor (anti-Pan human cytokeratins). No increase of macrophages recruitment within the tumor was observed for mice treated with the inhibitory EVs compared to the untreated mice.
Thus, in order to mimic the tumoral context and to identify the putative role of these infiltrating macrophages in the tumor growth inhibition, we designed co-culture conditions and set up protocol to generate hetero-spheroids containing 2/3 of breast cancer cells (MDA-MB-231 or SUM-159-PT) and 1/3 of the murine macrophage cell line RAW 264.7 embedded in 3% Matrigel. Three days later, medium containing the apoptosis indicator (fluorescent caspase 3/7 substrate), supplemented either with or without the indicated EVs was added. In this cellular system, EVs had no effect on macrophages cell growth while apoptosis was undetectable in this cell line (Fig. S2). While inefficient in our previous settings, (Fig. 3A), in this 3D system complemented with mouse macrophages, EVs T-47D shCtl were fully competent to reduce hetero-spheroids size and to increase apoptosis compared to the untreated cells ( Fig. 3C; -and EVs T-47D shCtl). Moreover, the impact on hetero-spheroids growth and apoptosis induction was reversed upon NFAT3 downregulation in T-47D producing cells ( Fig. 3C; EVs T-47D shNFAT3/4). These results obtained for both triple negative breast cancer cell lines tested correlate with the in vivo effect of EVs on tumor growth and underline the cooperation of EVs with tumoral macrophages to hinder breast cancer cell growth.

Overexpressing NFAT3 in T-47D EVs-producing cells enhances the anti-invasive function in vitro.
As EVs therapeutic use is explored for anti-tumor therapies 1,2 , increment of their intrinsic anti-tumoral effects described here would be interesting. As illustrated by the present work, NFAT3 could be a pertinent candidate since inhibition of cell invasion in vitro absolutely required NFAT3 expression in EVs producing luminal breast cancer cells (Fig. 2). Then, we asked whether overexpression of wild-type NFAT3 or a constitutively active form of NFAT3 previously described 9 in these cells would be a way to potentiate the inhibitory capacity of derived EVs. To this end, we generated by lentiviral transduction stable T-47D cell lines expressing either the control vector with a Tomato tag (T-47D toCtl), the wild-type NFAT3 (T-47D toNFAT3) or the active N-terminal deletion mutant of NFAT3 fused to the Tomato tag (toΔNFAT3). After validation by Western Blot (Fig. 4A, lower panel), we confirmed that, compared to empty vector, wild-type NFAT3 transduction (toNFAT3) inhibits treated conditions compared to the number of invasive cells in the control condition (−) arbitrarily set to 1. (D) 5 × 10 5 MDA-MB-231 cells (D3H2LN-LUC) were injected into the left Fat Pad of each 6-weeks-old female mouse Athymic Nude mice. Tumor volume: Tumor growth is presented as the mean tumor volume (cm 3 ) ±SEM, from mice injected weekly intra tumor with PBS or with EVs T-47D shCtl (5 × 10 9 pp) or EVs T-47D shNFAT3-3, EVs T-47D shNFAT3-4, 7 days after cell xenotransplantation, 8 mice per groups. Metastases: Beginning day 35, weekly, bioluminescent images were acquired on the IVIS system to quantify the mean of photons flux produced by the metastatic cells. Metastases quantification is presented as the mean photon flux produced by the metastatic cells. Data from one representative experiment of two independent experiments is shown, all data are shown as mean ± SEM (n = 8 technical replicates;, *p < 0.05 **p < 0.005 and ***p < 0.001, compared to the PBS treated mice group). (2020) 10:8964 | https://doi.org/10.1038/s41598-020-65844-x www.nature.com/scientificreports www.nature.com/scientificreports/  (Fig. 4A, upper panel). These results are in coherence with previous data obtained after transient transfections 9 . We then isolated EVs from T-47D toCtl, toNFAT3 or toΔNFAT3, validate their size and concentration by NTA (Nanoparticle Tracking Analysis) along with the amount of EVs produced per cell (Fig. S4). We did not observe any impact of NFAT3 overexpression on EVs production efficiency (Fig. S4). To test their effects on the invasive capacity, cell growth and apoptosis of the different aggressive breast cancer cell lines used above. Compared to the effect of EVs T-47D toCtl cells (T-47D toCtl), EVs produced by T-47D toNFAT3 slightly but significantly increase their anti-invasive effect, with a much stronger effect with the EVs produced by T-47D toΔNFAT3 in MDA-MB-231 and SUM-159-PT breast cancer cell lines (Fig. 4B, upper panels). Interestingly, these results were reproducible in BXPC3 pancreatic cancer and U87MG glioblastoma cell lines (Fig. 4B, lower panels). These results show the central role of NFAT3 in the EVs-producing cells.
Therefore, we wondered if these results were specific for T-47D-producing cells and decided to test NFAT3 constructs overexpression in the non-tumoral cell line HEK293T. First, we generated EVs from wild-type HEK293T and tested their effect on MDA-MB-231 cell invasion in comparison with the EVs produced in wilt-type T-47D. Strikingly results presented in Fig. S6A showed that EVs produced in HEK293T cells were as efficient to inhibit MDA-MB-231 cell invasion as the EVs generated by the T-47D cells. Therefore, as shown if Fig. S6B, we evaluated the potential expression of endogenous NFAT3 in HEK293T cells and validated that indeed NFAT3 was specifically expressed in these cells as shown by a NFAT3 siRNA. We then generated stable clones of HEK293T overexpressing the different NFAT3 constructs described above (Fig. S6C). The size and concentration of HEK293T EVs were determined by NTA along with the amount of EVs produced per cell (Fig. S5). We did not observe any effect of NFAT3 overexpression on EVs production efficiency (Fig. S5).
As for the T-47D cell line, overexpression of the NFAT3 constructs in HEK293T cells generated EVs with an enhance anti-invasive capacity underlining once again the central role of NFAT3 expression in EVs-producing cells independently of the cell type (Fig. 6D). Thus, overexpression of the NFAT3 constructs could be a general feature for inducing production of EVs with an enhanced anti-invasive capacity. Confirmed in two different human cell lines, these data emphasize the role of NFAT3 expression in EVs-producing cells for efficient cell invasion inhibition.
We then tested the effect of T-47D toNFAT3/toΔNFAT3 on the in vitro 3D cancer cells/macrophages co-culture system described above. EVs generated from both T-47D toNFAT3/toΔNFAT3 cells exert a strongest effect on hetero-spheroids growth inhibition as on apoptosis induction compared to the EVs originated from toCtl cells (Fig. 4C). To note, no transfer of overexpressed NFAT3 protein nor RNA could be detected in recipient cells (data not shown). Altogether, these results confirm the central role of NFAT3 expression in EVs-producing cells for consecutive signalization and global anti-proliferative and anti-metastatic effects and highlight the feasibility of increasing these inhibitory capacities by overexpressing NFAT3 in the EVs-producing cells.

NFAT3 overexpression in EVs-T-47D producing cells increases anti-tumor properties and inhibit tumor growth on a pre-established tumor model in vivo. As ectopic expression of NFAT3 in
EVs-producing cells was suitable to enhance EVs anti-invasive, anti-proliferative and pro-apoptotic effects in vitro, we evaluated in vivo the anti-tumoral effects of modified EVs. To this end, we established 3 groups of mice that were subjected to weekly intra-tumor injection, starting 1 week after cells implantation, with either PBS as a control or the different EVs originated from T-47D toCtl cells or T-47D toΔNFAT3 cells. Tumor growth was evaluated twice a week for each group by caliper measurements and, after one month, metastases apparition was quantified twice a week by luminescence with an IVIS in vivo imaging system. Beginning on day 24, mice injected with the EVs produced by T-47D toΔNFAT3 or by T-47D toCtl showed a similar significant reduction of tumor growth rate, compared to the mice group injected with PBS, lasting for 10 days (Fig. 5A, Tumor volume, day 24-34).
Remarkably, from day 38 to the end of the experiment, if injection of T-47D toCtl-produced EVs exerted by itself an anti-tumor growth effect (41% tumor volume reduction) compared to PBS-treated group, the curtail of the tumor growth was significantly stronger for the group of mice injected with the EVs toΔNFAT3 cells (61% tumor volume reduction) (Fig. 5A, Tumor volume, day 34-62). Nonetheless, despite of the clear inhibition of tumor growth mediated by T-47D toΔNFAT3 produced EVs, these vesicles were not further efficient in preventing metastases apparition, compared to the group of mice injected with the EVs T-47D toCtl (Fig. 5A,  Metastases). Indeed, tested EVs T-47D toΔNFAT3 and EVs T-47D toCtl induced a similar reduction (95%) of detected metastases, compared to mice injected with PBS (Fig. 5A, Metastases), illustrating the absence of correlation between the tumor size and the consecutive extend of metastatic dissemination.
To get closer to clinical situations, we further evaluated EVs potential effectiveness on a pre-established tumor in vivo. To this end, D3H2LN-LUC cells were xenografted in the fat pad of 2 mice groups, with a measurement of the tumor growth twice a week during 1 month without any EVs injection. After this first month, once the tumor had grown, one group of mice was intra-tumorally injected, once a week, with PBS and the second group with (C) Hetero-spheroids containing 33% of the murine macrophage cell line RAW 264.7 with either MDA-MB-231 or SUM-159PT were plated in medium with 3% Matrigel in 96 wells ultra low attachment plates for three days to allow the formation of the hetero-spheroids formed. Then medium was added to each condition containing the apoptosis indicator (fluorescent caspase 3/7 substrate) supplemented with medium containing or not the different EVs (3 × 10 8 pp/mL) as indicated in the figure. The size of the spheroids and appearance of green fluorescence (apoptosis) were recorded every 2 h on an INCUCYTE apparatus for 4 days. Data are represented as the AUC of the spheroid size (upper panels) and apoptosis (lower panels) over 96 h. Data from one representative experiment of three independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; ***p < 0.001, compared to the untreated cells). www.nature.com/scientificreports www.nature.com/scientificreports/ EVs originated from T-47D toΔNFAT3 cells. Tumor growth was evaluated 2 times/week by caliper measurements and metastases apparition was quantified 2 times/week by luminescence with an IVIS in vivo imaging system. Strikingly, the results presented in Fig. 5B show that as soon as 2 days after the first EVs injection, the tumor growth was significantly reduced compared to the PBS-treated control group. This slow-down pursued until the experiment ends, to reach a 46% of inhibition (Fig. 5B, Tumor volume), validating the beneficial effect of T-47D toΔNFAT3-EVs to prevent tumor growth. On the opposite, no statistical beneficial effect of delayed EVs injection was observed on metastases apparition (Fig. 5B, Metastases). Globally, these results clearly demonstrate that the inhibitory effect of EVs on tumor growth can be technically improved by overexpressing ΔNFAT3 in T-47D producing cells and was efficient on a pre-established tumor, raising interesting concern in the cancer fight field.

Discussion
Identifying effective treatments of highly metastatic cancer (i.e. triple negative breast cancer (TNBC) and pancreatic cancer) is the urgent challenge of the next decade. Indeed, TNBC remain the breast cancer subgroup with the least benefit from targeted therapies 15 . Therefore, the early-stage TNBC patients are mainly treated with combinations of taxane and anthracycline chemotherapy with severe side effects and unfortunately a frequent recurrence of the metastases in the first 3 years after surgery 16,17 . These therapeutic situations underline the urgent necessity of finding innovative therapeutic approaches to treat these aggressive cancers. In this context, EVs-based therapy shows an exponential interest with in vitro and in vivo studies showing promising results 18 .
In our present study, we have shown for the first time that EVs originated from NFAT3-expressing poorly aggressive luminal breast cancer cells are competent to alone inhibit triple breast cancer cell lines invasion and, in cooperation with macrophages, spheroids growth in vitro. We verified that these EVs inhibitory capacities were not restricted to the EVs produced by the T-47D cell line as another luminal breast cancer cell line (MCF-7) was able to generate inhibitory EVs too. Moreover, we also established that the inhibitory functions of these EVs were not restricted to the MDA-MB-231 breast cancer cell as another triple negative breast cancer cell line (SUM-159PT) melanoma, pancreatic and glioblastoma cancer cell lines were also similarly affected by incubation with these EVs. In addition, our study extends these results in vivo by demonstrating that these early injection of EVs are efficient to inhibit metastases arising and tumor growth in an in vivo breast cancer mice model.
The tumor microenvironment is determinant in tumoral progression and among the numerous types of cells presents in this surrounding tissue, macrophages play a central role in tumoral evolution by interacting either directly or indirectly with the tumor cells 14 . We show here that infiltrated macrophages were present within the tumor in our breast cancer mice model and demonstrate, in vitro, that the presence of macrophages was compulsory for the EVs to inhibit breast cancer tumor cells growth associated with an increase of apoptosis. These data could be explained by two not exclusive hypotheses, either (i) by an EVs-based sensitization of breast cancer cells to intra-tumoral macrophages killing, or (ii) by a functional interaction between EVs and macrophages to enhance their intrinsic killing capacity. To note compared to data previously reported 19 ,We did not reveal in our work any specific M1 or M2 polarization of the macrophages as it has been previously reported 19 .
Mechanistically, we demonstrate the requirement of NFAT3 expression in the EVs-producing cells to generate efficient anti-tumoral EVs. These results could suggest that either NFAT3 protein or NFAT3 mRNA or NFAT3-regulated targets (proteins, RNA, miRNA, for example) need to be transferred by the EVs to the recipient cells to induce anti-tumoral effects. Nonetheless, we were so far unable to detect either the NFAT3 protein nor the NFAT3 mRNA in target cancer cells or EVs, arguing more in favor of NFAT3-regulated targets that might be involved in the anti-invasive/anti-metastatic and anti-proliferative effects of EVs produced by luminal breast cancer cells. These NFAT3-regulated targets have now to be identified and characterized.
Furthermore, we show that these EVs have the capacity to inhibit in vivo a pre-established tumor closest to the clinical situation of patients with measurable lesions at the diagnostic. However, in this context, no detectable effects of the EVs on metastases arising were detected. This last observation raises the question of the timing of metastases formation and suggest that the metastatic dissemination might occur early in the tumor's development, as seen in the patients, before EVs injection as it has been reported by other studies 20,21 . Moreover, these data suggest that rather than eliminating pre-exiting metastases EVs treatment might prevent tumor cells from leaving the primary tumor to colonize other organs.
The anti-tumoral EVs we describe in this study may represent a potential new therapeutic tool to impede breast cancer progression. Indeed, these EVs show an inhibitory effect not only on breast cancer cells but also on other cancer cells, from melanoma, pancreas and glioblastoma origin, emphasizing their pervasive interest for large therapeutic applications. Indeed, we have shown that overexpressing an active form of NFAT3 in EVs-producing cells was a meaningful approach to boost their inhibitory capacities. Moreover, the possibility of using a non-tumorigenic human cell line as the HEK293T to generate enhanced inhibitory EVs might open avenues for potential future developments of production of therapeutic EVs according to GMP (Good Manufacturing Practice). Further studies on the ΔNFAT3-EVs mechanisms of action and ways to enhance their potency are still needed, in particular to understand the molecular determinants present within the EVs and the mechanistic consequences within the recipients' cancer cells and the microenvironment cells. Associated with spheroids. Medium was then added to each condition containing the apoptosis indicator (fluorescent caspase 3/7 substrate) with medium containing or not the different EVs (3 × 10 8 pp/mL). The size of the spheroids and green fluorescence (apoptosis) was recorded every 2 h on an INCUCYTE apparatus for 4 days. Data are represented as the AUC of the spheroid size (upper panels) and apoptosis (lower panels) over 96 h. Data from one representative experiment of three independent experiments is shown, all data are shown as mean ± SEM (n = 3 technical replicates; ***p < 0.001, compared to the untreated cells).  EVs T-47D toCtl (5 × 10 9 pp) or EVs T-47D ΔNFAT3, 7 days after cell xenotransplantation, 10 mice per group. Metastases: Beginning day 35, weekly, bioluminescent images (where the primary tumor was shield with a black tissue) were acquired on the IVIS system to quantify the mean photon flux produced by the metastatic cells. Metastases quantification is presented as the mean photon flux produced by the metastatic cells. Data from one representative experiment of two independent experiments is shown, all data are shown as mean ± SEM (n = 10 technical replicates;, *p < 0.05 **p < 0.005 and ***p < 0.001 compared to PBS-injected mice; (**), (***) compared to EVs-T-47D toCtl-injected mice). (B). 10 5 MDA-MB-231 cells (D3H2LN-LUC) were injected into the left Fat Pad of each 6-weeks-old female Athymic Nude mouse. Tumor volume: Tumor growth is presented as the mean tumor volume (cm3) ±SEM, from mice injected weekly intra-tumor with PBS or with 5 × 10 9 pp EVs T-47D ΔNFAT3, 38 days after cell xenotransplantation, 10 mice in each group. Metastases: Beginning at day 45, weekly, bioluminescent images (where the primary tumor was shield with a black tissue) were acquired on the IVIS system to quantify the mean photon flux produced by the metastatic cells. Metastases quantification is presented as the mean photon flux produced by the metastatic cells. Data from one representative experiment of three independent experiments is shown, all data are shown as mean ± SEM (n = 10 technical replicates; ***p < 0.001, compared to the PBS treated mice group). cells were cultured in high glucose (4,5 g/L) DMEM. All media were complemented with 100 U/mL penicillin and 100μg/mL streptomycin (Thermo Fisher Scientific), an anti-mycoplasma (normocin 100μg/mL, Invivogen) and 10% fetal bovine serum (FBS) (PAN BIOTECH); 10% newborn calf serum (PAN BIOTECH) was added to the medium of NIH3T3. All cell lines were routinely tested for the absence of mycoplasma contamination by PCR. The anti-mycoplasma drug (normocin) and the puromycin were omitted in all EVs productions and experiments. To produce T-47D shCtl, T-47D shNFAT3-3 and T-47D shNFAT3-4 stable cell lines, T-47D were stably infected with shRNAs cloned in the pGIPZ lentiviral vector obtained from Dharmacon (sequences: shNFAT3-3: CAATGAACACCACCTTGGA; shNFAT3-4: AGTCTCAGGGAACATCCGC). Lentiviral shRNAs were produced in HEK293T cells using the trans-lentiviral shRNA packaging kit with calcium phosphate (Dharmacon) following the manufacturer's instructions. Viral particles were collected and concentrated. T-47D cells were infected, selected in presence of 1.5μg/mL of puromycin, and sorted by flow cytometry on the basis of GFP expression. Effective down-regulation of the endogenous NFAT3 was validated by Western blot (anti-NFAT3: 1:1000, no. AB 2267268, Thermo Fisher Scientific; anti-NFAT3: 1:1000, no. AB 650208, Santa-Cruz Biotechnology).

Generation and characterization of stable cell lines.
To produce T-47D toCtl, T-47D toNFAT3 and T-47D toΔNFAT3 stable cell lines, T-47D were stably infected either by vontrol viral particles or expressing NFAT3 or ΔNFAT3 particles (from cDNA cloned in the pLVX lentiviral vector tagged with a td-Tomato epitote obtained from Clontech). All the cloned sequences were verified by sequencing. Lentiviral cDNAs were produced in HEK293T cells using the trans-lentiviral ORF packaging kit with calcium phosphate (Dharmacon) following the manufacturer's instructions. Viral particles were collected and concentrated. T-47D cells were infected, selected in presence of 1.5μg/mL of puromycin, and sorted by flow cytometry on the basis of the Tomato expression. Effective expressions of the cloned cDNAs were validated by Western blot (anti-Tomato: 1:1000, no. AB2722750, Sicgen). All Western blots were normalized by probing with an anti-actin (1:5000, no. AB10979409, Thermo Fisher Scientific).

EVs isolation and characterization.
Briefly, the different cell lines used to produce EVs were cultivated in absence of normocin and puromycin to reach 70% confluence, at which point their normal culture medium was changed to medium containing 10% FBS that had previously been depleted from calf serum EVs. To obtain appropriate cell medium without calf serum EVs, the appropriate cell medium was prepared with 20% FBS and centrifuged at 167,000 g for 18 h at 4 °C. After centrifugation the supernatant was collected, filtered through a 0.22μm membrane and kept at 4 °C until use. Cells were cultured in EV-free medium for 48 h, after which cell culture supernatant was collected for EVs isolation and a total cell lysate was obtained to compare total cell proteins composition to that of the EVs pellet (for EVs characterization purposes). EVs were isolated by differential centrifugation/ultracentrifugation techniques. First, the medium was centrifuged at 300 g for 10 minutes at 4 °C to remove cells. To remove dead cells, the supernatant was collected to be centrifuged at 2000 g for 20 minutes at 4 °C. In order to eliminate cell debris, the supernatant was collected to be centrifuged at 10,000 g for 30 minutes at 4 °C. The supernatant was collected to be ultracentrifuged in Ultra-Clear tubes using a 45Ti rotor (Beckman Coulter) at 120,000 g for 90 minutes at 4 °C in an Optima XPN-80 ultracentrifuge (Beckman Coulter). The supernatant was discarded and the remaining pellet, containing EVs and contaminating proteins, was washed twice in a large amount of cold PBS, and centrifuged again for 90 minutes at 120,000 g at 4 °C. This last ultracentrifugation was done a second time using fresh PBS to thoroughly eliminate the contaminating proteins. The pellet, containing the EVs fraction, was then resuspended in a minimal amount of cold PBS. The final EVs solution was aliquoted in siliconized polypropylene microcentrifuge tubes (Sigma, #T3281-500EA) and aliquots were kept at -80 °C until use. EVs were characterized by NTA (Nanoparticle Tracking Analysis) using the Nanosight (Malvern Instruments, Software version NTA 3.1 Build 3.1.46). PBS used for sample dilution was filtered and checked for vesicles absence prior to use. Calibration was performed using silica microspheres of 0.10μm sold at a known concentration of 4,897 × 10 13 pp/mL (particles/mL) (Biovalley, #24041). Each sample was analyzed twice (using two different dilutions and aliquots) and 5 videos of 60 seconds were made per analysis. Final sample concentration was calculated as a mean of the different concentrations measured per analysis and as a proportion of the difference of concentration of the silica microspheres given by the Nanosight and the known concentration informed by the manufacturer. Total protein lysates from EVs were compared to that of their cells of origin by Western blot. Samples were prepared in non-reducing sample buffer with 2% SDS to EVs samples. 20 μg of protein were loaded per lane and separated with a 12.5% SDS-acrylamide gel. The antibodies used were: rabbit anti-human calnexin (1:100, no. AB2243890, Santa Cruz); mouse anti-human CD63 (1:1000, no. AB396297, BD Biosciences) and mouse anti-human CD81 (1:1000, no. AB2275892, Santa Cruz Biotechnology).
Invasion assay. Cells were seeded in 12-wells plates and starved the following day by removing the FBS.
Twenty-four hours later, medium or EVs was added to the wells as described in the figures. Twenty-four after adding the EVs, cells were trypsinized, resuspended in cell medium containing 1% BSA, centrifuged and washed once in free medium. Cells were then resuspended in medium containing 0.1% BSA and seeded in transwell chambers with 8μm pores (BD Falcon, #353097) in a 24-well companion plate previously coated with Matrigel (Corning, #356234 0.5 μg/insert). Conditioned medium isolated from NIH3T3 cells was added to the wells under the chambers and used as attractant for cell invasion. Cells were allowed to invade for different periods of time: 6 h for MDA-MB-231, SUM159PT, BXPC3, U-87MG, WM.266.4 and 24 h for T-47D and MCF7 cell lines Then, non-invading cells were removed from the top of the transwell with a cotton swab, whereas invading cells were fixed in 100% ethanol for 10 minutes and stained with hematoxylin (#HHS16-500mL, Sigma Aldrich) for 25 minutes. Within each experiment, each condition was tested in technical triplicates and the number of cells in the transwell whole field was counted to evaluate the cells' ability to invade the Matrigel. Thus, the invasion index was