PDAC-derived exosomes enrich the microenvironment in MDSCs in a SMAD4-dependent manner through a new calcium related axis

Tumor genetics and escape from immune surveillance concur in the poor prognosis of PDAC. In this study an experimental model was set up to verify whether SMAD4, deleted in about 55% PDAC and associated with poor prognosis, is involved in determining immunosuppression through Exosomes (Exo). Potential mechanisms and mediators underlying SMAD4-dependent immunosuppression were evaluated by studying intracellular calcium (Fluo-4), Exo-miRNAs (microarray) and Exo-proteins (SILAC). Two PDAC cell lines expressing (BxPC3-SMAD4+) or not-expressing (BxPC3) SMAD4 were used to prepare Exo-enriched conditioned media, employed in experiments with blood donors PBMCs. Exo expanded myeloid derived suppressor cells (gMDSC and mMDSC, flow cytometry) and altered intracellular calcium fluxes in an SMAD4 dependent manner. BxPC3-SMAD4+, but mainly BxPC3 Exo, increased calcium fluxes of PBMCs (p = 0.007) and this increased intracellular calcium trafficking characterized mMDSCs. The analysis of de-regulated Exo-miRNAs and transfection experiments revealed hsa-miR-494-3p and has-miR-1260a as potential mediators of SMAD4-associated de-regulated calcium fluxes. Eleven main biological processes were identified by the analysis of SMAD4-associated de-regulated Exo-proteins, including translation, cell adhesion, cell signaling and glycolysis. A reverse Warburg effect was observed by treating PBMCs with PDAC-derived Exo: BxPC3 Exo induced a higher glucose consumption and lactate production than BxPC3-SMAD4+ Exo. Conclusion: PDAC-derived Exo from cells with, but mainly from those without SMAD4 expression, create an immunosuppressive myeloid cell background by increasing calcium fluxes and glycolysis through the transfer of SMAD4-related differentially expressed miRNAs and proteins.


INTRODUCTION
Pancreatic ductal adenocarcinoma (PDAC) is still one of the most lethal forms of cancer worldwide despite the accumulating knowledge gained on its biology and genetics [1][2][3]. Recent studies exploring the whole genome and exome sequences revealed that the genetic landscape of PDAC is much more complex than previously believed, but they also confirmed that activating mutations of KRAS and inactivating mutations in the tumor suppressor genes p53, CDKN2A and SMAD4 characterize this tumor type [4,5]. The loss of SMAD4, an almost unique event in gastrointestinal malignancies (e.g. colorectal adenocarcinoma), which occurs in about 50% of PDAC [2], appears to be of particular interest. PDAC patients carrying SMAD4 loss have a worse prognosis [6,7], and this event, in vitro, favors cell proliferation and migration, while antagonizing cell senescence [8]. It is widely believed that the molecular basis linking SMAD4 loss with increased tumor growth and aggressiveness depends on the absence of Smad4 protein, which mediates the effects of the transforming growth factor (TGF)-β and bone morphogenetic protein, which inhibit cell proliferation and migration, and trigger apoptosis [9]. Yet this underlying mechanism cannot alone explain all PDAC features associated with SMAD4 loss, as recently demonstrated by Whittle et al. [10], who found that heterozygous SMAD4 mutation is associated with increased cellular proliferation but attenuates the metastatic potential of PDAC, while the complete loss of SMAD4 restores metastatic competency by regulating the expression of the transcription factor Runx3. Moreover, SMAD4 loss might underlie a more aggressive tumor pattern not only because tumor cells acquire a pro-proliferative and pro-metastatic phenotype, but also because it might create a favorable soil for tumor growth and metastasis by conditioning the surrounding stroma [11,12].
The PDAC microenvironment, characterized by a dense desmoplastic reaction driven mainly by pancreatic stellate cells and pancreatic cancer associated fibroblasts, include immune cells, which have an imbalance toward an immunosuppressive and protumorigenic phenotype [13,14]. Immunosuppressive T reg lymphocytes, M2 polarized tumor associated macrophages (TAM) and myeloid derived suppressor cells (MDSCs) prevail over immune effector CD8 + T cells, dendritic cells (DCs) and M1 polarized TAM in the tumor microenvironment, in blood and in lymphoid organs in both PDAC animal models and humans [15][16][17][18][19][20][21][22][23]. The immune imbalance in PDAC is an evolving phenomenon, starting in the early phases of carcinogenesis (i.e. in PanINs), a specific chronology of distinct immune cells subsets derangement paralleling disease progression [24]. While lymphocyte lineage alterations with T reg accumulation appear to occur early, alterations in myeloid lineage with MDSCs accumulation appear to occur later in PDAC [20].
Overall, PDAC-associated immune cell alterations concur in masking tumor cells from immune detection, thus favoring tumor progression. This observation is borne out by clinical data demonstrating the negative prognostic value of immunosuppressive cellular accumulation, and experimental research is now being conducted to establish whether therapeutic benefit might be gained by targeting immunosuppressive cells [14,20].
Immune cell alterations at the tumor site, and in nearby and distant organs, appear to depend on tumorderived and immunomodulatory molecules, including GM-CSF, S100A8/A9 heterocomplex and chemokines, which target the Ras/MAPK, Jak/Stat, PI3K and TGFβ pathways [14,25]. Soluble mediators, however, are not the only means of 'communication' between tumor cells and the surrounding stromal cells, extracellular micro vesicles also being involved [26]. Exosomes (Exo), defined as micro vesicles with a diameter ranging from 30 to 150 nm, are released by cells and can transfer from one cell to another a rich cargo of molecular messengers, including miRNAs, proteins, lipids and DNA, which remains highly stable within the Exo shuttle [27]. The role of Exo has been highlighted in carcinogenesis, metastases, drug resistance and immunosuppression [28][29][30][31]. Among the mechanisms activated by cancer derived Exo in the specific immune cellular setting, Stat3 and NF-κB activation has been suggested to occur in a TLR-dependent fashion due to Hsp70, miR-203 or as yet unidentified proteins [32,33].
When investigating the immune cell response to cancer, the role of calcium signaling should be taken into account, since it plays a relevant role in the elaboration of the adaptive immune response thought to be crucial to cancer control [34,35]. It remains to be elucidated whether and, if so, how variations in the mutational landscape of PDAC, with a special focus on SMAD4 deletion, accelerates immunosuppression; this would lead to a greater understanding leading to new therapeutic targeting. In the present study we demonstrate that PDAC derived Exo imbalance immature myeloid cells subsets, enhancing MDSCs while lowering DCs, by altering intracellular calcium fluxes, these effects being magnified in cases of SMAD4 loss. The characterization of miRNAs, and the protein cargo of PDAC derived Exo, allowed us to discover SMAD4-associated de-regulated miRNAs and proteins, and to identify, among the more extensively de-regulated miRNA, hsa-miR-494-3p and hsa-miR-1260a, and among proteins, VDAC1 and spectrin beta, all involved in calcium balance. Finally, SMAD4 loss determined enrichment in the Exo protein cargo of glycolytic enzymes, these Exo being able to enhance myeloid cell glycolysis. www.impactjournals.com/oncotarget

PDAC-derived Exo change the balance between MDSCs and DCs
To ascertain whether pancreatic cancer cells modify the equilibrium between immunosuppressive and immune effector cells in an SMAD4-dependent manner and whether any imbalance is mediated by soluble factors or Exo, PBMCs were cultured for four days in complete, Exo enriched and Exo free non conditioned (NC) and conditioned media (CM) obtained from BxPC3 and BxPC3-SMAD4+ cells. T cells and immature myeloid cells subsets were analyzed by flow cytometry and data from eight donors were collected. Only complete CM obtained from BxPC3-SMAD4+ cells caused a mild but significant expansion of CD4 + CD25 + (Repeated measures ANOVA: F = 10.77, p = 0.003), while reducing CD8 + T cells (F = 8.348, p = 0.007) (Supplementary Figure 1).
Within the CD11b expressing immature myeloid cells, four subsets were identified on the basis of the expression of CD14 and HLA-DR markers: monocytes (CD14 + DR + ), DCs (CD14 -DR + ), granulocytic (gMDSCs, CD14 -DR -) and monocytic (mMDSCs, CD14 + DR -) MDSCs. Figure 1 shows a typical example of CD11b gating and the subsequent analysis of CD11b + cells on the basis of CD14 and HLA-DR expression (upper panels). Complete and Exo free CM obtained from both BxPC3 and BxPC3-SMAD4+ cells induced a significant increase of CD11b + immature myeloid cells (Repeated measures ANOVA: F = 11.190, p = 0.007 and F = 15.540, p = 0.001 respectively), while Exo enriched fraction did not (F = 0.102, p = 0.793) (Figure 1, bottom panels). Cellular apoptosis (Annexin V) was not affected by any BxPC3 and BxPC3-SMAD4+ CM (Supplementary Table 1). Figure 2 illustrates the distribution of the four myeloid cells subsets obtained in the above-described experimental conditions. Complete CM obtained from both BxPC3-SMAD4+ and BxPC3 cells caused a significant increase in the percentage of mMDSCs (Repeated measures ANOVA: F = 12.50, p = 0.006) while reducing both DCs (F = 12.82, p = 0.009) and gMDSCs (F = 10.590, p = 0.013). Exo enriched media enhanced the percentage of mMDSCs (F = 3.749, p = 0.059) but also gMDSCs (F = 6.027, p = 0.034) in a SMAD4 independent manner, while Exo free media reduced the DCs population in the presence of SMAD4 (F = 7.711, p = 0.014). The CD11b + CD14 -DR + subset was confirmed to represent the DCs subpopulation in a separate series of two independent flow cytometry experiments with the CD80 and CD86 markers, which expression was not altered by BxPC3 or BxPC3-SMAD4+ CM (Supplementary Figure 2). Macrophages, classified as CD11b + CD16 + cells, were mainly allocated within the monocytes subset and in part in the mMDSCs subset, and they were not affected by treatment with cancer conditioned complete, Exo enriched and Exo free media (Supplementary Table 2).
The data obtained indicate that PDAC cells induce the expansion of CD11b + immature myeloid cells in a SMAD4-independent manner through the release of soluble mediators. Within this expanded population of myeloid cells, PDAC through off balance between immature myeloid cells subsets, lowering DCs and gMDSCs, while enhancing mMDSCs, whether in the absence or presence of SMAD4 gene expression. Exo from BxPC3 appear to mediate the expansion of any immunosuppressive cell subtype.

PDAC lowers TNF release by PBMCs in a SMAD4-dependent manner
IL-1β, IL-4, IL-6, IL-10, TGF-β1 and TNF were assayed in PBMCs' supernatants before and after four days in NC and in CM. IL-1β, IL-4, IL-6, IL10 did not vary and levels were below the lower limit of detection in all conditions. The levels of TGF-β1 measured in PBMCs supernatants collected after four days of culture did not significantly vary between NC (5,18±0,46 μg/L, mean±SEM from 4 replicated experiments), BxPC3 CM (5,25±0,74 μg/L) and BxPC3-SMAD4+ CM (4,89±0,40 μg/L) (One-way Anova: F = 0.465, p = 0.556). With respect to basal levels, a significant increase of TNF was found in NC PBMCs after four days of culture (from 4.81±0.16 pg/mL to 265.4±94.9 pg/mL, mean±SEM from 19 donors; Student's t test for paired data: t = 2.755, p = 0.013). The levels of TNF measured in PBMCs supernatants after four days in BxPC3 and BxPC3-SMAD4+ CM were expressed as percentage with respect to the corresponding TNF levels of NC PBMCs considering any donor of this series of 19 donors. The results are shown in Supplementary Table 3, which also reports the results obtained in a subset of 12/19 donors, for which parallel experiments were performed using Exo enriched NC and CM. With respect to NC media, the release of TNF by PBMCs was reduced by BxPC3-SMAD4+ CM, but mainly by BxPC3 CM. Exo enriched media did not reproduce the inhibitory effect observed with the employment of complete media.

PDAC-derived Exo increase intracellular calcium fluxes in a SMAD4 dependent manner
With the aim to verify whether Ca 2+ ions were involved in the observed PDAC dependent changes in myeloid cell subsets and TNF secretion, [Ca 2+ ] i fluxes were studied in PBMCs treated in the same conditions described above and Figure 3 shows the results of representative experiments for any studied condition each reporting results from 8 cells. Minimal and irregular variations in www.impactjournals.com/oncotarget [Ca 2+ ] i were observed in NC complete and Exo-enriched media ( Figure 3, upper panels), while an increased frequency of [Ca 2+ ] i was observed in PBMCs cultured in BxPC3-SMAD4+, but mainly in BxPC3 complete CM, the results being reproduced by Exo-enriched CM. To compare the results, we classified variations in [Ca 2+ ] i as absent or present and, in this latter case, as regular peaks or irregular oscillations. In each experiment the number of cells belonging to the first (stable [Ca 2+ ] i ), second (peak [Ca 2+ ] i ) or third group (irregular [Ca 2+ ] i ) were counted. BxPC3-SMAD4+, but mainly BxPC3 complete CM caused a significant increase in the number of cells with a peak or irregular behavior of [Ca 2+ ] i (X 2 = 13.480, p = 0.009) and this finding was confirmed when cells were incubated with Exo-enriched CM (X 2 = 14.212, p = 0.007). In the bottom panel of Figure 3 the percentages of PBMCs showing an irregular or a peak [Ca 2+ ] i pattern after they have been incubated in complete or Exo enriched NC or pancreatic cancer CM are shown.

DCs do not pulse whereas regular intracellular calcium fluxes characterize mMDSCs
To verify whether any specific pattern of [Ca 2+ ] i was correlated with any specific myeloid cell subset, PBMCs from 3 donors were incubated with BxPC3-SMAD4+ CM for 96 hours to enrich the immunosuppressive populations before performing FACS sorting. Monocytes, DCs, gMDSCs and mMDSCs were collected, seeded, and cultured in NC media for 48 hours before [Ca 2+ ] i analysis. Figure 4 shows [Ca 2+ ] i in the four myeloid cellular subsets. Monocytes had both peaks and irregular [Ca 2+ ] i whereas DCs did not pulse. FACS sorted gMDSCs hardly adhered to coverslips glass slides and only two records were available, but both cells had a stable profile. Conversely mMDSCs were characterized by the presence of well defined regular peaks and by the absence of irregular [Ca 2+ ] i . FACS sorted gMDSCs and mMDSCs were stained with May-Grünwald-Giemsa ( Figure 4, bottom panels).

The SMAD4-dependent Exo miRNAs 494-3p and 1260a induce intracellular calcium fluxes in PBMCs
According to the microarray analysis, the heat map of the differentially expressed miRNAs was generated by unsupervised hierarchical clustering analysis ( Figure 5). The most deregulated 30 miRNAs, detected by highest absolute value of logarithmized estimated fold change, are reported in Supplementary Table 4. Microarray data were validated using real time PCR, analyzing two of the most de-regulated miRNAs, hsa-miR-494-3p and hsa-miR-1260a. The Ct of three experimental replicates, each run in duplicate, obtained in BxPC3 derived Exo ranged from 30.93 to 31.91 for hsa-miR-494-3p and from 27.81 to 27.85 for hsa-miR-1260a. After normalization to U6 small RNA Exo content, the relative expression of any miRNA in BxPC3-SMAD4+ derived Exo was referred to that of BxPC3. It was confirmed that hsa-miR-494-3p was under expressed (relative quantification = 0.8±0.0, mean±SD), while hsa-miR-1260a was over expressed (1.2±0.0) in BxPC3-SMAD4+ with respect to BxPC3 derived Exo. To ascertain whether the differential expression of these two miRNAs was correlated with variations in calcium fluxes of myeloid cells, transfection induced expression and inhibition of hsa-miR-494-3p and of hsa-miR-1260a in both BxPC3 and BxPC3-SMAD4+ cells were performed. PBMCs were maintained in complete CM obtained from non transfected and transfected or inhibited pancreatic cancer cells.

SMAD4 loss determines enrichment in Exo proteins of the glycolysis biological process
The differentially expressed proteins between BxPC3-SMAD4+ and BxPC3 derived Exo, evaluated by SILAC experiments, were defined as proteins with a ratio between heavy (BxPC3) and light (BxPC3-SMAD4+) isoforms below 0.667 and above 1.5, respectively. The identified proteins with a ratio below 0.667 were those under-expressed, while those with a ratio above 1.5 were over-expressed in BxPC3 with respect to BxPC3-SMAD4+ Exo. Supplementary Table 5 reports the Uniprot accession number, protein description, and heavy/light ratio of the SMAD4-related differentially expressed proteins in tumor-derived Exo. Using David software, a comprehensive analysis was made of differentially expressed proteins between BxPC3 and BxPC3-SMAD4+ derived Exo. The GO-terms for biological processes found to be significant after Benjamini and Hochberg adjustment for multiple comparisons were selected (Supplementary Figure 6). Eleven biological processes were identified, including translation, cell adhesion, cell signaling and glycolysis. Focusing on glycolysis, we found that BxPC3 cells had a higher degree of glucose consumption and of lactate production than BxPC3-SMAD4+ cells ( Figure  7, upper left panel). To verify whether the different expression of glycolytic enzymes between BxPC3 and BxPC3-SMAD4+ derived Exo had an impact on glucose metabolism of myeloid cells, glucose and lactate were assayed in the supernatants of PBMCs cultured for 2, 24, 48, 72 and 96 hours in Exo-enriched CM and NC media, the results from three independent experiments each made in duplicate being shown in the right panels of Figure 7. BxPC3, not BxPC3-SMAD4+, Exo enriched media induced significantly greater glucose consumption and lactate production with respect to Exo enriched NC media. To verify whether PDAC-derived Exo transfer the glycolytic enzymes into PBMCs, the activity of LDH, one of the most differentially expressed proteins between BxPC3 and BxPC3-SMAD4+ derived Exo, was assayed in PBMCs after they have been cultured for 24 hours in NC and in serially diluted BxPC3 Exo-enriched CM. LDH activity in PBMCs increased in the presence of Exo with respect to NC (

DISCUSSION
Alterations in genetic and epigenetic, transcription and signaling programs converge and concur in determining PDAC, a tumor expected to become the second most common cause of cancer-related mortality in 2030 in the absence of advances in therapy [36]. A relevant co-determinant in PDAC progression is the immunosuppressive ground associated with this tumor type [3]. In order to gain a deeper understanding of the complexity of immunoregulation in PDAC, in this study we provide evidence that SMAD4 loss imbalances immature myeloid cellular subsets in favor of MDSCs by altering intracellular calcium doing this through miRNAs and proteins Exo transfer from cancer cells to myeloid cells.
We first observed that, in the presence of PDAC CM, T cells subsets were unbalanced towards immunosuppression: CD8 + T immunoeffector cells were  reduced while CD4 + CD25 + cells underwent expansion. These "in vitro" results are is in agreement with the decreased CD8 + and increased CD4 + CD25 + Foxp3 + T cell stromal infiltration found in human tumors, and described to be associated with shorter survival [18,37]. Interestingly, these PDAC-driven effects on T cells were SMAD4-dependent, being observed when BxPC3 cells were forced to express SMAD4. The immunosuppressive T cell phenotype induced by PDAC CM might be mediated by soluble factors but also by factors carried in Exo, currently considered an important route for cell to cell communication [32,[38][39][40]. Both the soluble fraction and the Exo enriched fractions of PDAC CM failed to reproduce the effects on T cells evoked by complete media. This might alternatively or additionally depend on: 1. co-operation between soluble factors and Exo being required to alter T cells phenotypes; 2. magnitude of observed variations in T cells, subsets being too small to be reproduced by any single fraction. Moreover, PDACderived factors (e.g. GM-CSF) may directly trigger the variations observed in T cells phenotypes [20], but these variations are more likely to be secondarily induced by myeloid cells, a potential main target of PDAC-derived immunomodulatory factors [23,41]. In line with this hypothesis, PDAC CM significantly altered myeloid cell subsets in a SMAD4 independent manner, causing the reduction of DCs and of their main cytokine, TNF, in favor of the induction of mMDSCs. gMDSCs, unlike mMDSCs, were inhibited by PDAC CM and this fits well with the hypothesized equilibrium between these two immunosuppressive myeloid cell populations, i.e. the expansion of one set being correlated with the inhibition of the other set [20]. Differently form complete media, Exo were able to induce the expansion of both mMDSC and gMDSC subsets, being these effects SMAD4 -related.
Since myeloid cells with an immunosuppressive phenotype prevailed in PDAC CM, we verified whether immunosuppressive cytokines increased in this setting. IL-1β, IL-4, IL-6 and IL10 were below the respective limits of detection, while TGF-β1 increased in PBMCs culture media on the fourth day of culture (2 fold) [20,41]. However, no significant differences were found between NC and PDAC conditioned PBMCs. This finding might not support the hypothesis that TGF-β1 is associated with PDAC induced immunosuppression, but it might also be consequent to 1) the autocrine/ paracrine action of this cytokine, the amount produced being rapidly taken up by producing and/or adjacent cells thus limiting the cytokine quota released in culture media and 2) our results come from one particular cell line which definitively does not represent the whole complexity of PDAC immunosuppressive networtk. In this network macrophages should also be considered, because they might enhance tumor-initiating capacity and suppress CD8 + T cells when educated by tumor cells [19], although in our experimental model macrophages appeared unaffected. This finding per se does not contradict the established concept that in PDAC tissue TAM increase and concur in enhancing tumor progression and chemoresistance. Our experimental model allows to investigate the effects of tumor derived soluble products and Exo on peripheral blood mononuclear cells, the precursors of TAM, which complete differentiation requires a more intimate and complex contact of monocytes with tumor cells and other stromal cells [22]. PDAC CM and Exo in particular might induce the expansion of immunosuppressive myeloid cells by activating critical intracellular signaling pathways, mainly JAK/STAT kinase, TGF-β and RAGE [25]. Calcium calmodulin is a critical pathway involved in many cellular functions, including myeloid cell differentiation [35,42]. Regular [Ca 2+ ] i pulses are required to maintain normal myeloid cellular function, and any disruption of this mechanism might underlie alterations in phenotype and function such as those observed in the present study [43]. This was the premise of our subsequent experiments conducted to evaluate [Ca 2+ ] i fluxes in the above described experimental settings. PBMCs cultured in NC medium showed irregular variations in [Ca 2+ ] i resembling background noise. A higher frequency of [Ca 2+ ] i spikes were found in PBMCs cultured in BxPC3-SMAD4+, but mainly in BxPC3 complete CM, and these variation were reproduced by Exo enriched media, suggesting that in case of SMAD4 loss cancer derived Exo magnify their ability to induce an increased calcium trafficking in PBMCs. Ca 2+ responsible for this increased cellular activity might derive from the extra-cellular space and the intracellular stores [42]. We believe that both are involved, since after blocking membranal Ca 2+ channels with verapamil, an overall reduction in [Ca 2+ ] i fluxes was observed but the difference between control and PDAC treated cells was maintained (data not shown). Since the variations in [Ca 2+ ] i pattern mirrored those in myeloid cell phenotypes, we analyzed the [Ca 2+ ] i pattern of single cells populations after they have been FACS sorted, and demonstrated that DCs are static while mMDSCs present regular and well defined [Ca 2+ ] i peaks. Only few FACS sorted gMDSCs could be analyzed, since they did not adhere to the glass supports and this might be consequent to increased apoptosis of sorted cells in standard media for 48 hours, in agreement with Stromnes et al. [20]. These data support the notion that PDAC cells and, in particular, PDAC derived Exo reduce DCs and expand mMDSCs through [Ca 2+ ] i changes, this phenomenon being SMAD4-dependent. This is the first time that such a mechanism has been described in literature.
Among the potential causative agents responsible for this pathogenic mechanism, SMAD4-related differences in the amount of Exo release might be hypothesized, also considering that a dose-related effect of BxPC3 derived Exo in inducing gMDSCs was found (data not shown). However, the role of SMAD4-differentially expressed miRNAs and/or proteins carried in PDAC-derived Exo, which transfer their cargo from cancer cells to immune cells thus altering their behavior should also be taken into account. Therefore we screened the differentially expressed miRNAs and proteins in BxPC3 and BxPC3-SMAD4+ derived Exo by microarray analysis and SILAC respectively.
A number of miRNAs were found to be differentially expressed between Exo from the two cell lines studied. We selected two of the most SMAD4-associated de-regulated miRNAs, the hsa-miR-494-3p, which was previously demonstrated by Li et al. to be closely associated with SMAD4 expression [44], and hsa-miR-1260a, which association with SMAD4 is reported here for the first time. To ascertain whether these two miRNAs have any role in the calcium balance of myeloid cells, transfection and inhibition experiments were performed. In the absence of SMAD4 (BxPC3) the expression of both hsa-miR-494-3p and hsa-miR-1260a caused a progressive accumulation in intracellular calcium in PBMCs. The expression of hsa-miR-1260a augmented the percentage of PBMCs with regular calcium peaks and induced cells to form large clusters only when SMAD4 was also expressed (BxPC3-SMAD4+), suggesting that these hsa-miR-1260a-related effects require other cancer derived SMAD4-related co-factors, being SMAD4-related proteins potential candidates. The most de-regulated proteins between SMAD4 expressing and non expressing cellular derived Exo were mainly involved in translation, cell and matrix adhesion, integrin and Ras signaling, and this is in agreement with a more aggressive SMAD4-associated phenotype. Three proteins overexpressed in BxPC3 with respect to BxPC3-SMAD4+ derived Exo, appear of potential interest for further studies because of their known effects on immune cells and/or trans membrane ion fluxes: VDAC1, Versican and Spectrin beta [45,46]. A number of glycolytic enzymes were overexpressed in BxPC3 than in BxPC3-SMAD4+ Exo, and glycolysis was one of the core biological processes significantly altered in an SMAD4 associated manner and this allows us to add SMAD4 to the repertoire of cancer-associated genetic alterations, which include KRAS activating mutations, TP53 loss of function and MYC overexpression, which regulate the metabolic reprogramming of cancer cells, also known as the Warburg effect [47]. The SMAD4 related enrichment of glycolytic enzymes in cancer derived Exo might underlie the reverse Warburg effect, the metabolic reprogramming involving stromal cells, such as cancer associated fibroblasts, cancer stem cells, and immune cells. We demonstrated that PDAC derived Exo transfer to myeloid cells the glycolytic enzyme activity of LDH, this finding being correlated with a direct and SMAD4-associated impact of Exo on myeloid cell glycolysis, since BxPC3 derived Exo caused a greater glucose consumption and lactate production than BxPC3-SMAD4+ Exo. However Exo might impact on cellular glucose metabolism not only through the transfer of glycolytic enzymes, but also because they might interfere with gene transcription and translation. miRNAs are candidate molecules in this context, and based on this we verified whether one of the most highly expressed miRNA in BxPC3-SMAD4+ derived Exo, the hsa-miR-1260a, could be implicated in the regulation of glycolysis by transfection/inhibition experiments. The forced expression of this miRNA caused a reduced glucose consumption and lactate production in SMAD4-expressing cells suggesting that hsa-miR-1260a is an antagonist of the Warburg effect. The present study therefore demonstrates that metabolic reprogramming with increased glycolysis and conversion of glucose into lactate is SMAD4 and hsa-miR-1260a dependent and that tumor cells can transfer the propensity to elevated glycolysis to stromal cells via Exo. This finding opens new ways for understanding the relationship between cancer metabolism and immune suppression recently reported to be dependent on the metabolic competition between tumor and immune cells [48].
In conclusion, SMAD4 deletion might underlie a more aggressive PDAC phenotype through Exo transfer to immune cells of proteins and miRNAs that, by altering intracellular calcium and glycolysis, may cause the expansion of immunosuppressive myeloid cells.

Cell lines
The pancreatic cancer cell lines BxPC3, known to carry a homozygous deletion of the SMAD4/DPC4 gene, and BxPC3-SMAD4+, obtained from the BxPC3 cells stably transfected with the pBK-cytomegalovirus (CMV)-SMAD4/DPC4 expression vector, were used. The characterization of the cellular model including the validation of the transfection efficacy has been described by us elsewhere [49]. Culture details are reported in Supplementary Methods.

Exo enrichment for cellular experiments
Non conditioned media, BxPC3 and BxPC3-SMAD4+ conditioned media (CM), all containing 1% FCS, were enriched in Exo following the differential ultracentrifugation procedure detailed in Supplementary Methods.

Isolation of human peripheral blood mononuclear cells (PBMCs)
Human PBMCs were isolated from blood donors' buffy coats by differential density gradient centrifugation (Histopaque ® -1077, Sigma-Aldrich, Milano, Italy, F/H). After being washed twice with saline solution to remove contaminating platelets and centrifuged at 1,200 rpm for 10 minutes, PBMCs were treated with a hemolysis solution (NH 4 Cl, KHCO 3 , EDTA Na 4 ) for 10 minutes, centrifuged at 1, 200 rpm for 10 minutes and finally used for the experiments.

Experimental design
PBMCs were cultured in complete, Exo-free and Exo-enriched CM and non conditioned media. After 4 culture days, PBMCs were analyzed by flow cytometry and cytokines were measured in the supernatants. Intracellular calcium fluxes were analyzed in the same experimental conditions after 3 culture days by epifluorescence (Fluo4).

Flow cytometry analysis
Human PBMCs from 8 blood donor's buffy coats were seeded in a six well culture plate (6x10 6 in each well) and kept in culture for 96 hours with 3 mL of complete, Exo-free and Exo-enriched non conditioned, BXPC3 CM and BxPC3-SMAD4+ CM reconstituted with 10% FCS. After 96 hours, PBMCs were collected by scraping, and analyzed by flow cytometry as detailed in Supplementary Methods.

Cell sorting
20x10 6 PBMCs were cultured for 48h in 100 mm culture dishes in 10 ml of BxPC3-SMAD4+ CM. At least 20 culture dishes were prepared. After collection by scraping, 200x10 6 PBMCs were incubated in the dark for 30 minutes with the following monoclonal antibodies (Beckman Coulter, Miami, FL, USA): 110 µL HLA-DR-PC5, 80 µL CD14-PC7, 80 µL CD45-ECD and 80 µL CD11b PE. Afterwards, PBMCs were washed with PBS containing 2% FCS, centrifuged at 1200 rpm for 10 minutes and re-suspended with PBS containing 2% FCS to obtain 15x10 6 PBMCs per mL. The above-described four mononuclear cell subsets were FACS-sorted (BD FACSAria III, BD Biosciences, San Jose, CA, USA). Sorted cells were seeded on coverslips that had been inserted within each well of six well culture plates and cultured for 48 hours in 2 ml of complete control medium before [Ca 2+ ] i fluxes study. After [Ca 2+ ] i fluxes study, each coverslip was also stained in May-Grünwald-Giemsa. This experiment was repeated three times with buffy coats from three blood donors.

Intracellular calcium fluxes analysis
2x10 6 PBMC were seeded on coverslips that had been inserted in each well of six well culture plates and cultured for 72 hours in 2 ml of complete or Exo-enriched control, BxPC3 CM, BxPC3-SMAD4+ CM reconstituted at 10% FCS. Coverslips were then processed for the [Ca 2+ ] i fluxes study as previously described [49], using the intracellular calcium tracer Fluo-4AM (Invitrogen S.R.L.) at 5 μM. Three independent experiments, each made in triplicate, were performed.

Cytokines assay
After centrifugation at 1,200 rpm for 10 minutes of PBMCs supernatants after 96 hours culture in the abovedescribed conditions IL-1β, IL-4, IL-6, IL-10, TGF-β1 and TNF were analyzed by chemiluminescent immunometric assays (Immulite, Siemens Healthcare Diagnostic, UK) according to the manufacturer's specifications. For all the experimental conditions, at least three independent sets of experiments were performed.

Glucose, lactate and lactate dehydrogenase (LDH) assays
Glucose and lactate in culture supernatants were measured on the automated ARCHITECT System following the manufacturer's instructions (Abbott Laboratories, Abbott Park, IL, USA). LDH enzymatic activity was determined in PBMCs lysates by using the automated COBAS System according to the manufacturer's instructions (Roche Molecular Diagnostics, Pleasanton, CA, USA). www.impactjournals.com/oncotarget miRNAs microarray analysis The miRNAs expression profiling of BxPC3 and BxPC3-SMAD4+ CM Exo was performed in duplicate samples using the Agilent SurePrintG3 Human miRNA (8x60K) microarrays (custom design by Comprehensive Biomarker Center, CBC, recently re-named to Hummingbird Diagnostics GmbH, Heidelberg, Germany) as detailed in Supplementary Methods.

SILAC experiment
BxPC3 and BxPC3-SMAD4+ cell lines were cultured in RPMI 1640 MEDIA FOR SILAC with 10% dialyzed fetal bovine serum (FBS), with the addition of either with the non-labelled aminoacids Lysin and Arginine (light medium) or with the labelled 13 C 6 -Lysine and 13 C 6 15 N 4 -arginine (heavy medium) (Chemical Research 2000 srl, Rome, Italy). In a second experiment, the same cell lines were cultured by swapping media, BxPC3 being maintained in heavy medium and BxPC3-SMAD4+ maintained in light medium, thus creating two biological replicates of the same experiment. After eight days, cell media were changed with fresh media prepared as specified above but without the serum addition, in order to reduce any contaminant in the proteomic analyses, cells being cultured in this condition for 16 hours before media collection and exosomes enrichment made by ultracentrifugation (Beckman Coulter L-80XP Ultracentrifuge, Type 70 Ti rotor) as detailed in supplementary methods. The ratios for light/ heavy and heavy/light were calculated for each identified protein of the two experiments. Proteins were considered significantly altered if the average value, calculated for the two ratios, was above 1.5 or below 0.667

Statistical analysis
The statistical analysis of data was made by Repeated measures ANOVA, Tukey's multiple comparisons test, Student's t test for paired data and the chi square test using the GraphPad prism software (ver 6.04).