Exosomes as mediators of platinum resistance in ovarian cancer

Exosomes have been implicated in the cell-cell transfer of oncogenic proteins and genetic material. We speculated this may be one mechanism by which an intrinsically platinum-resistant population of epithelial ovarian cancer (EOC) cells imparts its influence on surrounding tumor cells. To explore this possibility we utilized a platinum-sensitive cell line, A2780 and exosomes derived from its resistant subclones, and an unselected, platinum-resistant EOC line, OVCAR10. A2780 cells demonstrate a ~2-fold increase in viability upon treatment with carboplatin when pre-exposed to exosomes from platinum-resistant cells as compared to controls. This coincided with increased epithelial to mesenchymal transition (EMT). DNA sequencing of EOC cell lines revealed previously unreported somatic mutations in the Mothers Against Decapentaplegic Homolog 4 (SMAD4) within platinum-resistant cells. A2780 cells engineered to exogenously express these SMAD4 mutations demonstrate up-regulation of EMT markers following carboplatin treatment, are more resistant to carboplatin, and release exosomes which impart a ~1.7-fold increase in resistance in naive A2780 recipient cells as compared to controls. These studies provide the first evidence that acquired SMAD4 mutations enhance the chemo-resistance profile of EOC and present a novel mechanism in which exchange of tumor-derived exosomes perpetuates an EMT phenotype, leading to the development of subpopulations of platinum-refractory cells.


INTRODUCTION
Ovarian cancer remains the leading cause of death of all gynecological malignancies due, in part, to a dearth of treatment options and the evolution of platinumresistant disease. The clinical introduction and use of platinum-based chemotherapeutics, specifically cisdiamminedichloroplatinum(II) (cisplatin), significantly improved overall survival (OS) by ~6 months in 29% of patients with ovarian cancer, leading the way to its adoption as the backbone of most chemotherapeutic regimens [1,2]. In the mid-1980's, a cisplatin analog, Carboplatin [cis-diammine(1,1-cyclobutanedicarboxylato) platinum(II)], with an improved toxicity profile and equivalent therapeutic efficacy, replaced cisplatin as standard of care [3][4][5]. The last major advance occurred in the early 1990's with the introduction of the mitotic inhibitor, paclitaxel, that further improved OS 3-15 months (depending on the study) when used in combination with platinum [6][7][8]. Despite these advances, tumor progression and recurrences still occur in approximately 80% of advanced stage ovarian cancer patients and ultimately result in mortality as a result of acquired resistance to platinum-based antineoplastic drugs. Platinum-resistance has been attributed to several factors including, but not limited to, increased glutathione synthesis, increased drug efflux, increased DNA damage repair, and/or increased ability to undergo epithelial to mesenchymal transition (EMT) [9][10][11][12][13][14][15].
Current interest in the EOC field is focused on the importance of intercellular cross talk mediated by soluble and insoluble factors between the EOC tumor www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 7), pp: 11917-11936

Research Paper
and stromal cells during development, progression, and evolution of drug-resistance [16][17][18]. The EOC tumor microenvironment includes recruited host cells (i.e., endothelial cells, fibroblasts, and macrophages) that communicate with tumor cells and often are reeducated to supply functions which enhance metastasis, vascularization, and immuno-evasion. For instance, EOC tumors have been shown to produce significant levels of interleukin (IL)-6 which trigger recruitment of monocytes from the peripheral blood, and via activation of STAT3, convert these cells into tumor supportive M2 tumor associated macrophages [19]. Work from our lab demonstrated fibroblasts from normal human ovaries secrete high levels of hepatocyte growth factor, (HGF) which binds to and activate c-MET mediated signaling on EOC tumor cells. This interaction leads to changes in biological processes including increases in tumor proliferation and metastasis [20]. The mechanisms of cellular communication in these types of experiments include secretion of soluble factors such as cytokines, mitogens, and growth factors; however, recently nanosized vesicles (30-150 nm) termed exosomes have been shown to be released by tumor cells and are emerging as a novel vehicle of cell-cell communication within the development and progression of EOC [21,22].
Exosomes are multifunctional vesicles that have essential roles in normal physiology [23] as well as pathological processes [24][25][26]. They are formed through the endosomal pathway when the membrane of the late endosome undergoes inward budding resulting in the formation of intraluminal vesicles. The late endosome, now termed a multivesicular body (MVB), escapes merging with the lysosomal pathway and instead travels back and fuses with the cell membrane, releasing its contents (exosomes) into the extracellular space [27]. Exosomes are informative molecules that carry cargo representative of their originating cell including nucleic acids, cytokines, membrane-bound receptors, and a wide assortment of other biologically active lipids and proteins [28,29]. This cargo remains functional upon entry or fusion with a recipient cell, thus exosomal transfer is now considered an important form of cell-cell communication in normal and pathological states such as cancer [30]. Exosomes can travel systemically throughout the body and are able to target cells at very distal locations [31]. They contribute to EOC development by inducing immune evasion, assisting in the establishment of secondary tumor niches, and serve as 'packaged delivery vehicles' for cross talk between tumor cells and the surrounding stroma [32,33]. More recently, exosomes have been shown to have a functional role in the development of chemotherapy resistance in breast, non-small cell lung, and prostate cancer; however, their role in platinumresistance in EOC is unknown [34][35][36]. While emphasis in the EOC field is focused on tumor-stroma communication during neoplastic advancement, we report here, the first evidence of the importance of the intricate exchange of exosome-mediated crosstalk within EOC tumors leading to chemotherapeutic resistance by way of activation of EMT. Based on these novel findings, we propose the release of exosomes is a mechanism by which neoplastic EOC cells 'educate' each other, thereby exasperating the development of platinum-resistant disease.

Exosomes from carboplatin-resistant cells modify the platinum sensitivity profile of recipient cells
Exosomes carry a range of nucleic acids and proteins and can have a significant impact on the phenotype of recipient cells as reported in melanoma, breast, nonsmall cell lung, and gastrointestinal stromal tumors (GIST) amongst others [35,[37][38][39][40]. For this phenotypic effect to occur, exosomes need to fuse with target cell membranes, either directly with the plasma membrane or with the endosomal membrane after endocytic uptake. To examine if exosomes isolated from platinumresistant clones are taken up by the platinum-sensitive parental, population we utilized the classic A2780 (IC 50 of 11 µM carboplatin) cell line and two independentlyderived carboplatin resistant clones, C30 (IC 50 of 325 µM carboplatin) and CP70 (IC 50 of 120 µM carboplatin) [41][42][43] (Supplementary Table 1). We first isolated extracellular vesicles from conditioned media of A2780, CP70, and C30 cells by ultracentrifugation [44,45]. Isolated vesicles displayed between ~80-150 nm in size as determined by Nanoparticle Tracking Analysis (NTA) and scanning electron microscopy and contained common exosomal markers such as ALIX, TSG101, CD60, and the absence of either β-actin or GRP78 suggesting a highly enriched exosome population (Supplementary Figure 1). To visualize uptake of exosomes by recipient cells, the exosomes were labeled with PKH67 green and the cells with PKH red fluorescent membrane dyes. Recipient A2780 cells were then exposed to exosomes (number of exosome equivalent to 1 µg of total exosomal protein/10,000 cells) derived from A2780 (autologous), C30, or CP70 cells for up to 24 hours. Regardless of the source, exosome uptake was rapid and uniform across all experimental groups ( Figure 1A, Supplementary Figures  2-4).
We have previously shown that the phenotypes of normal primary myometrial cells are dramatically altered towards the characteristics of donor tumor cells following uptake of tumor-derived exosomes from GIST cells and patient samples [38]. To investigate if a platinum-resistant phenotype could be transferred by this mechanism we treated platinum-naive A2780 cells with exosomes (number of exosome equivalent to 0.5 µg of total exosomal protein/5,000 cells) from A2780 (autologous), CP70, C30, www.impactjournals.com/oncotarget Overlay of A2780 cells (Red) exposed to exosomes (Green) derived from A2780, CP70, or C30 cell lines for 24 hours. DAPI (Blue) is included. B. Illustration of experimental protocol. C. Viability of A2780 cells pre-treated with PBS (Black), or exosomes derived from A2780 (Red), CP70 (Green), or C30 (Dark Green) cell lines followed by exposure to carboplatin. D. Caspase 3/7 activity as measured using Caspase Glo ® luminescent reporter system in A2780 cells exposed to the same conditions as C. E-F. Colony formation assay and corresponding analysis of 200 A2780 cells pre-treated for 24 hours with exosomes (as above) and then exposed to constant levels of 0, 1, or 5 µM carboplatin. All values are normalized to vehicle. Fold changes are relative to control (PBS -no exosome group) *P<0.05, **P<0.01. www.impactjournals.com/oncotarget or vehicle (PBS) for 24 hours prior to treatment with 20 µM carboplatin for 48 hours ( Figure 1B). We observed a near 2-fold (P<0.05) and 1.5-fold increase in viability of A2780 cells exposed to CP70-derived and C30-derived exosomes, respectively, as compared to autologous exosomes or PBS ( Figure 1C). Parallel analysis of caspase 3/7 activity by way of a fluorescent reporter assay revealed an average 20% decrease in cleavage of caspases 3 and 7 indicating reduced apoptosis in cells treated with C30-or CP70-derived exosomes as compared to controls ( Figure  1D). Finally, a colony formation assay of 200 A2780 cells pre-treated with exosomes for 24 hours prior to exposure to either 1 or 5 µM carboplatin revealed an ~15% increases in the number of colonies when pretreated with exosomes derived from C30 and CP70 cell lines as compared with controls ( Figure 1E-1F).
To confirm this observation is not limited to a select lineage of cells, we investigated the effects of exosomes derived from the unrelated, platinum-resistant OVCAR10 cell line (IC 50 of 200 µM carboplatin) on A2780 cells and the effects of C30-and OVCAR10-derived exosomes on the A1847 (IC 50 of 67 µM carboplatin) and OVCAR5 (IC 50 of 44 µM carboplatin) ovarian cancer cell lines. OVCAR10 was previously derived from a high-dose carboplatin and cisplatin refractory advanced ovarian tumor [11], while A1847 and OVCAR5 were derived from untreated advanced ovarian tumors [41,46]. Exosomes were isolated and characterized as before (Supplementary Figure 1 and data not shown). Similar to previous observations, A2780 cells treated with OVCAR10-derived exosomes exhibited up to a 1.7-fold (P<0.05) increase in viability as well as a 50% decrease in caspase 3/7 activity following treatment with 80 µM carboplatin for 48 hours (Figure 2A-2B). Likewise, colony formation assays of 100 A2780 cells pre-treated with exosomes for 24 hours demonstrated up to a 20% increase in colony formation upon treatment with 1 µM carboplatin ( Figure 2C-2E). We also observed an increase in colony size in A2780 cells treated with exosomes derived from C30 and OVCAR10 upon exposure to 1 µM carboplatin (Supplementary Figure 5). A1847 cells exhibited up to a 1.7-fold (P<0.05) increase in viability and up to a 75% decrease (P<0.005) in caspase 3/7 activity upon exposure to 50 µM carboplatin for 48 hours only in cells treated with exosomes derived from CP70, C30, or OVCAR10 cell lines as compared to autologous exosome controls ( Figure 3A-3B). In addition, A1847 cells pre-treated with exosomes derived from CP70, C30, or OVCAR10 derived exosomes, resulted in a near 3-fold increase in colony formation upon exposure to 5 µM carboplatin as compared to controls ( Figure 3C-3E). Similar results were seen in OVCAR5 cells, which demonstrated up to a 2-fold increase in viability (P<0.05) and 50% decrease in caspase 3/7 activity (P<0.005) upon exposure to 50 µM carboplatin in cells pre-treated with exosomes derived from C30 and OVCAR10 cell lines as compared to autologous exosome controls ( Figure 4A-4B). We also observed up to a 3-fold increase in colony formation in cells treated with exosomes derived from CP70, C30, or OVCAR10 cell lines as compared to autologous exosome controls ( Figure 4C-4E). Taken together our data demonstrate the development of platinum resistant disease may, in part, be mediated by cell-cell communication via exosomes.

Exosomes induce a protracted phenotypic change
Recent studies by our lab and others have shown the effects of tumor-derived exosomes on recipient neighboring and distal cells can be both transient as well as prolonged and, in some cases, even exert a permanent phenotypic shift [24,38,47]. Therefore, we next asked if the acquired platinum-sensitivity profile was durable. We repeated the studies as indicated above and utilized an additional highly platinum-resistant cell line C200 (IC 50 >500 µM), which was derived from C30 by continuous exposure to 200 µM cisplatin [41]. A2780 cells were exposed to autologous, CP70-, C30-, or C200-derived exosomes (1.0 µg exosome/10,000 cells plated), for 24 hours after which the media was replaced and cells were cultured for up to 14 days in standard conditions. Two-weeks after exosome exposure the recipient A2780 cells exposed to C200 or CP70-derived exosomes maintained a more resistant (~1.2-fold increase in viability, P<0.05) phenotype as compared to controls. This increase in resistance was consistent across 20, 40, and 80 µM concentrations of carboplatin ( Figure 5).

Exosomes derived from platinum-resistant cells trigger EMT in platinum-sensitive A2780 cells
Two recent studies have presented evidence which challenges the significance of EMT in cancer metastasis but suggest that inhibition of EMT suppresses chemoresistance in lung and pancreatic carcinomas [48,49]. Continuing with this line of thought, there is increasing evidence that EOC cells utilize EMT as a mechanism of escape from the deleterious effects of platinum-based chemotherapy, as evident from the observed up-regulation of mesenchymal markers upon treatment with platinumbased compounds [13,15,50]. Likewise, exosomes have been shown to carry cargo which triggers EMTlike changes in breast, colorectal, and urinary cancers [51][52][53]. Given this information we questioned the effects of exogenous exosome exposure on changes in EMT characteristics. We utilized A2780 and their platinumresistant derivatives CP70, C30, and C200 cells, which, as we have previously published, all express a basal mesenchymal morphology [14,54]. A2780 were treated with exosomes as above, and evaluated for changes in expression of mesenchymal markers, including mRNA and miRNA expression as well as changes in morphology.
For mRNA analysis we chose a panel of previously published genes to characterize changes in EMT [14]. A2780 cells exposed to exosomes from A2780 platinumresistant subclones (i.e., CP70, C30, or C200) for 24 hours demonstrated at least a 50% reduction (P<0.05) in the expression of the epithelial markers dystroglycan, a cell-surface laminin receptor that links the cytoskeleton to the extracellular matrix in a variety of epithelial tissues [55,56], E-cadherin, a cell-cell adhesion molecule and is down-regulated during EMT in numerous cancers [57,58], and to a lesser extent EpCAM, the epithelial cell adhesion molecule expressed in epithelia and epithelialderived neoplasms [59,60]. Complimentary to these results we observed significant (P<0.05) increases in the mesenchymal markers palladin, which plays a role in modulating the actin cytoskeleton and is ubiquitously expressed in cells of mesenchymal origin [61] and TWIST, an embryonic transcription factor which is frequently reactivated in cancer and correlate with poor prognosis. Additionally, we observed up to a 70-fold increase (P<0.05) in vimentin, a type III intermediate filament protein expressed in mesenchymal cells [62] in A2780 cells treated with C200-derived exosomes ( Figure 6A). We also observed that A2780 cells treated with exosomes from platinum-resistant cells displayed a 2-to 5-fold (P<0.05) decrease in KLF4, a transcription factor that regulates genes essential to EMT including E-cadherin and vimentin and has been shown to regulate EMT in multiple types of cancers [63][64][65][66] (Figure 6A).
To further examine markers associated with EMT we examined MicroRNA-21 (miR-21) expression. miR-21 is perhaps, one of the most well studied miRs in regards to

Figure 2: A2780 cell platinum sensitivity is influenced by other platinum-resistant EOC cells. A-B. Viability (A) and
Caspase 3/7 activity (B) cell lines exposed to exosomes derived from autologous or OVCAR10 cell lines and treated with carboplatin for 48 hours. All values are normalized to vehicle. Fold-change is relative to control (autologous treated exo group) C-F. Colony formation assay (C) raw data (D) and corresponding analysis (F) of 100 A2780 cells pre-treated for 24 hours with exosomes (autologous, CP70, C30, and OVCAR10) and then exposed to constant levels of 0, 1, or 5 µM carboplatin. All values are normalized to viability. Fold-change is calculated relative to control. *P<0.05. www.impactjournals.com/oncotarget EMT in cancer [67][68][69]. Specifically, in regards to EOC, up-regulation of miR-21 has been shown to correlate with decreased sensitivity to both paclitaxel and cisplatin and has been associated with EOC cell stemness [70][71][72]. We found that CP70, C30, and OVCAR10 cells have 18-, 5-, and 6-fold increases in mature miR-21 as compared to A2780 cell lines, respectively (Supplementary Figure 6). Importantly, A2780 cells treated with CP70-and OVCAR10-derived exosomes demonstrate a nearly 4-fold increase in miR-21 as compared to cells treated with PBS or autologous-derived exosomes ( Figure 6B). Furthermore, there were no significant changes in primary miR-21 (pri-miR-21) levels in A2780 cells treated with exosomes ( Figure 6B), suggesting either transfer or increased processing of miR-21 following uptake of exogenous exosomes.
We further observed that, while A2780 cells themselves have a partial mesenchymal phenotype, they displayed an increase in spindle-like morphology when exposed to C30-and C200-derived exosomes, and to a lesser-extent, CP70-derived exosomes ( Figure 6Carrows). Taken together these data demonstrate the transfer of a resistant phenotype may be driven, in part, through transfer or upregulation of factors which contribute to EMT.

Figure 3: A1847 cell platinum sensitivity is influenced by other platinum-resistant EOC cells. A-B. Viability (A) and
Caspase 3/7 activity (B) cell lines exposed to exosomes derived from autologous, CP70, C30 or OVCAR10 cell lines and treated with carboplatin for 48 hours. All values are normalized to vehicle. Fold-change is relative to control (autologous treated exo group) C-F. Colony formation assay (C) raw data (D) and corresponding analysis (F) of 100 A1847 cells pre-treated for 24 hours with exosomes (autologous, CP70, C30, and OVCAR10) and then exposed to constant levels of 0, 1, or 5 µM carboplatin. All values are normalized to viability. Foldchange is calculated relative to control. *P<0.05, **P<0.005. www.impactjournals.com/oncotarget

Alterations in SMAD4 are found in platinumresistant EOC cell lines
To identify potential alterations contributing to drug resistance in these models at the genomic level, Next Generation DNA sequencing was performed across 12 ovarian cell lines using a panel of 48 cancerassociated genes using the TruSeq© Amplicon Cancer Panel (Illumina) (Supplementary Table 2). Through this analysis we identified previously unreported somatic mutations in the MH2 domain of SMAD4, which were exclusive to the highly platinum resistant cell lines (C30, CP70, and OVCAR10) (Figure 7 and Supplementary Figure 7A). Specifically, all three cell lines, C30, CP70, and OVCAR10, contain a common SMAD4 S344I mutation (OVCAR10 was homozygous for the S344I mutation), while C30 and CP70 each harbor a second unique mutation (SMAD S411C and SMAD G508A , respectively) ( Figure  7A). The sequencing results were confirmed by sanger sequencing ( Figure 7B and data not shown). These initial findings suggest that changes in SMAD4 may contribute or correspond with a loss of platinum sensitivity and warrant further investigation.
We next used in silico analysis to determine the potential functional impact of these SMAD4 mutations [73]. The SMAD4 Q388R variant, a missense change discovered in OVCAR4, a cell line derived from a patient's tumor that was refractory to cisplatin, but with a low level of in vitro platinum-resistance [41,74], received a score of 0.54 and was predicted to have no effect on protein function. The SMAD4 S344I mutation, which is common to all 3 highly platinum resistant cell lines, was assigned a

Figure 4: OVCAR5 cell platinum sensitivity is influenced by other platinum-resistant EOC cells. A-B. Viability (A) and
Caspase 3/7 activity (B) cell lines exposed to exosomes derived from autologous, CP70, C30 or OVCAR10 cell lines and treated with carboplatin for 48 hours. All values are normalized to vehicle. Fold-change is relative to control (autologous treated exo group) C-F. Colony formation assay (C) raw data (D) and corresponding analysis (F) of 100 OVCAR5 cells pre-treated for 24 hours with exosomes (autologous, CP70, C30, and OVCAR10) and then exposed to constant levels of 0, 1, or 5 µM carboplatin. All values are normalized to viability. Foldchange is calculated relative to control. *P<0.05, **P<0.005. value of 2.765 and predicted to have a considerable effect on function. The second acquired mutations identified in CP70 (SMAD4 G508A ) and C30 (SMAD4 S411C ) were both predicted to have a high impact on SMAD4 function (ranked at 3.365 and 3.395, respectively) (Supplementary Figure 7B).

Dysregulation of TGF-β/SMAD signaling is found in clinical EOC samples
SMAD4 is a key component of TGF-β/SMAD signaling. It binds to receptor activated SMADs (SMAD2 or SMAD3), forming a heterodimer which allows for translocation into the nucleus, where it binds additional co-factors and initiates transcription of target genes [75,76]. To determine how the TGF-β/SMAD signaling pathway might be altered in clinical EOC cases, we used in silico data mining approaches and The Cancer Genome Atlas (TCGA) database. We identified multiple genetic aberrations within one or more key components of the TGF-β/SMAD signaling pathway. Most notably, 21% (61 of 316) of high-grade serous EOC samples have deletions and/or down-regulation of SMAD4 which significantly (p<0.001) correlates with deletions and/or down-regulation of SMAD2 (deleted or down-regulated in ~28% (81 of 316) of the cases ( Figure 8A). In addition, enrichment analysis of samples with available miRNA expression data (n=518) revealed specific miRNAs that are differentially expressed in cases with loss of SMAD4 as compared to normal controls. These include miRNAs known to impact or be regulated by TGF-β/SMAD signaling, such as miR-142 [77], miR-146A [78,79], and miR-29B [80]. Importantly, miR-21 was significantly (p=0.00417) overexpressed in cases with a loss of SMAD4 expression (n=61) (Figure 8B-8arrow). Additionally, out of 316 cases with complete mRNA data, we found no significant change in overall survival (OS) in cases harboring dysregulation in SMAD4 (n=61) or SMAD2 (n=81), indicating that loss of SMAD4 and/or SMAD2 is not a significant predictor of outcome in primary EOC ( Figure 8C-8D). However, patients exhibiting a loss of SMAD3 expression (n=16), albeit small in number, have a significant increase in OS (~44 versus 65 months, p=0.0386), suggesting that activation of the SMAD3 pathway could be important in EOC development ( Figure  8E). Continuing with this analysis, SMAD2 mRNA levels are down-regulated in 32% of TCGA platinum resistant EOC tumors (N=197) as compared to 24% in platinum sensitive (N=90) and SMAD3 tends to be down-regulated in the platinum sensitive cohort as compared to the resistant (7% and 1% respectively) further highlighting the complexity and potential importance of SMAD  Figure 8). In ovarian cancer, SMAD3 acts alone or in conjunction with SMAD4 to regulate transcription of EMT target genes [81]. Importantly, SMAD4, SMAD3, and other key components of the TGF-β/SMAD signaling pathway (i.e., TGFβr1, TGFβr2, and SMAD2) are found within exosomes isolated from tumor cell lines and from human plasma samples ( Supplementary Figures 1 & 9).

Mutant SMAD4 cell lines and exosomes elicit a resistant phenotype in recipient cells
To provide direct evidence that mutations in SMAD4 contribute towards a loss of platinum-sensitivity, we exogenously overexpressed SMAD4 S344I , SMAD4 S411C , or SMAD4 WT in A2780 cell to generate A2780 S344I , A2780 S411C , and A2780 WT cell lines, respectively. Of these cell lines, A2780 S344I demonstrated the highest level of platinum resistance, e.g., 2-fold (P<0.05) increase in viability at both 20 and 40 µM carboplatin ( Figure 9A). A2780 S344I cells also exhibited an increase in mesenchymal markers N-cadherin and ZEB1 as well as a decrease in the mediator of apoptosis, PDCD4 (Programmer of Cell Death 4) ( Figure 9B), which is known to regulate EMT [82]. Likewise, we observed a more mesenchymal phenotype upon exposure of A2780 S344I cells to 20 and 40 µM carboplatin as compared to A2780 WT ( Figure 9C), even in the background of the endogenous SMAD4.
Taken together, these data provide the first evidence that mutations, specifically within SMAD4, enhance the native chemo-resistance profile of EOC. Specifically, cells harboring these mutations have survival advantages by way of increased EMT in response to carboplatin. In addition, these cells generate and secrete exosomes capable of modifying the platinum-sensitivity profile of surrounding neoplastic cells which enhances the development of platinum-resistant disease.

DISCUSSION
Because the ovarian tumor mass is composed of a heterogeneous population of cells with a high degree of individual morphologies and genomic instability, the development of platinum-based chemotherapy resistance is multifactorial and frequently due to a variety of changes in specific proteins, genes, or in gene regulation that are advantageous to chemotherapy resistance [10,[83][84][85][86][87][88][89][90]. It is commonly believed that chemotherapeutic treatment used during cancer treatment can cause or select for intrinsically resistant populations of cells which possess one or more of these advantages. For example, Stronah and colleagues have shown that, in response to cisplatin, DNA-dependent protein kinase (DNA-PK) selectively phosphorylates AKT-S473 in the nucleus of platinum-resistant, but not sensitive cells, leading to the clinical development of platinum-resistant disease [91]. Whether by selection of resistant clones or infliction of secondary aberrations, some tumor cells survive primary rounds of therapy, and eventually give rise to new cells which form more therapy resistant tumors. Here we have investigated how drug resistant populations of cells perpetuate the development of platinum-resistant disease in ovarian cancer via exosomal mitigated cell-cell communication. It is well understood that intracellular transport of genetic material, including miRNAs and mRNAs as well as biologically active proteins from one cell to another, occurs via microvesicles [36,[92][93][94][95][96]. We and others have shown that extracellular vesicles may be involved in transformation of surrounding cells via transport of miRNAs and oncogenic proteins [26,38,40,97,98]. Our current findings suggest that, as the ovarian tumor develops and evolves to escape therapeutic attack, exosomes may be a previously unappreciated factor, which contribute to the advancement of drug resistant disease.
We present evidence demonstrating that exosomes derived from platinum-resistant EOC cells can transfer a portion of their chemo-resistant phenotype to platinumsensitive cells and this increase in resistance corresponds with increased EMT and mutations in SMAD4. SMAD4 has been shown to be necessary for the transcription of EMT related genes through SMAD4/SMAD3 signaling [81] and EMT has been well established as a mediator of drug resistance in ovarian cancer [15,[99][100][101]. Our focused analysis of TCGA ovarian cancer data did not identify mutations in SMAD4; however, we did identify significant dysregulation of key TGF-β/SMAD genes, despite the fact that the vast majority of samples are from primary tumors. The lack of mutations within SMAD4 is not surprising: however, given that we hypothesize that somatic mutations in SMAD4 are either selected for or cause by platinum-based chemotherapy. Here we provide Figure 10: Exosomes from Mutant SMAD4 Cell Lines Transfer Platinum Resistance. A. Viability of A2780 cells exposed to PBS or exosomes derived from A2780 WT , A2780 S411C , and A2780 S344I cell lines for 24 hours prior to carboplatin treatment (10 µM for 72 hours). Viability is normalized to vehicle within each exosome group. B. Dose-response curves of A2780 cells pre-exposed to exosomes derived from A2780 WT or A2780 S344I cell lines for 24 hours followed by carboplatin treatment (0, 10, 20, 40, 80, and 160 µM) for 72 hours. *P<0.05. the first evidence that mutations within the SMAD4 gene directly affect the platinum-sensitivity of EOC cells. In addition, we are the first to report that cells harboring these mutations produce exosomes capable of transferring this resistant profile.
TGF-β/SMAD signaling components are found within and on exosomes including SMAD2, SMAD4, TGFβrI and TGFβrII ( Supplementary Figures 1 and 9); however, a momentary transfer of protein alone does not provide sufficient evidence to explain a persistent phenotypic switch. Evidence suggests that the transfer of exosomal mRNAs between cells may lead to more prolonged phenotypic changes [102,103]. We have identified very low levels of the SMAD4 mRNA within a portion of exosomal samples (Crow and Godwin, unpublished data); however, we were unable to sequence this transcript, suggesting that the whole SMAD4 mRNA transcript may not be present. Given this information, and the fact that SMAD4 is a transcription factor, we speculate that the exosomal contents responsible for the observed morphological changes may be products of aberrant TGF-β/SMAD signaling.
Current evidence suggests that miR-21 may both be a product of and mediator of TGF-β signaling in many pathological conditions [104][105][106]. We identified a 3-to 4-fold increase in miR-21 expression in platinum-sensitive cells treated with platinum-resistant exosomes and in the parental platinum-resistant cells. Up-regulation of miR-21 in multiple types of cancers, including ovarian, positively correlates with increased EMT [107][108][109]. Therefore, a complementary mechanism to alter platinum sensitivity is exosome-induced up-regulation of primary miR-21 (pri-miR-21) in recipient cells. Of relevance, recent studies by Davis and colleagues demonstrate that SMAD proteins can bind with the Drosha complex to increase processing of pri-miR-21 to pre-miR-21 [110]. Indeed, we observed increased miR-21 in A2780 cells treated with platinumresistant exosomes but no significant changes in pri-miR21, suggesting increased miRNA processing may be an important factor in EOC. Although exciting, further research is warranted to further define the exosomal cargo that is directly responsible for influencing platinum sensitivity.
In summary, we propose a new mechanism by which tumor cell-cell cross talk may actively enhance chemotherapy resistance throughout treatment. It is important to note that the experiments conducted here are limited to a single exosome treatment, and do not mimic the potentially constant exposure to exosomes which likely occurs in vivo. This is especially relevant in EOC, where neoplastic cells have enhanced exosomal output as compared to normal epithelial cells [111]. Taken together, this work underscores the importance of cell-cell communication in the advancement of platinum resistant disease and provides novel insight as to how exosomal transfer further enhances EMT in response to frontline chemotherapy.

Cells and culture conditions
We utilized human ovarian cancer cells A2780, C30, CP70, C200, OVCAR5, A1847, and OVCAR10 [11,43,112]. Ovarian cancer cell lines A2780, A1847, and OVCAR5 were authenticated by using multiplex short tandem repeat (STR) testing and compared to historical reference DNA preserved in the lab for these cell lines. Testing was performed by the Clinical Molecular Oncology Laboratory at KUMC, a CLIA/CAP-accredited molecular diagnostics laboratory using the Promega PowerPlex 16 System used for human identity testing run on an Applied Biosystems instrument. All cell lines were cultured in RPMI-1640 (Gibco, Thermo Fisher) media supplemented with 10% (v/v) exosome-depleted FBS, 2 mm L-glutamine, 0.2 units/mL human insulin, and 100 units mL penicillin-streptomycin at 37°C with 5% CO 2 . Exosome-depleted FBS was obtained by centrifuging FBS for 18 hours at 100,000 x g followed by filtration through a 0.22 µm filter.

Exosome isolation
Cell lines were cultured to 70-80% confluency and conditioned media was collected after 24-48 hours, spun for 10 minutes at 2000 x g, and pooled together. Exosomes were isolated by differential centrifugation as previously reported [23]. Briefly media was spun for 45 minutes at 10,000 x g to pellet large vesicles and twice at 100,000x g to pellet and wash exosomes. Exosome pellets were resuspended in 50-100 µL of cold PBS and stored at -80°C.

Nanoparticle tracking analysis
Purified exosomes were resuspended in 100 µL of 0.22 µm filtered PBS and analyzed using a NanoSight LM10 instrument (NanoSight, Salisbury, United Kingdom). Analysis was performed by applying a monochromatic 404 nm laser to diluted exosomal preparation and measuring the Brownian movements of each particle. The Nanoparticle Tracking Analysis software version 2.3 was used to analyze 60 second videos of data collection to give mean, median, and mode of vesicle size and concentration.

Electron microscopy
Exosomes were purified as above and fixed using 2% glutaraldehyde in 0.1 M sodium cacodylate buffer at 4°C overnight. The pellet was rinsed in 0.15 M sodium cacodylate buffer (pH 7.4), followed by a post fixation in 1% osmium tetroxide containing 0.1% potassium ferricynide buffered in 0.1 M cacodylate buffer for 1 hour. Exosomes were dehydrated through a series of ethanol washes followed by a propylene oxide bath for 10 minutes. Prepared exosome pellets were embedded in half propylene oxide/half embed 812 resin and cured in a 60°C oven overnight. 80 nm sections were cut using a Leica UC7 ultramicrotome and were picked up on copper thin bar 300-mesh grids and contrasted with 4% uranyl acetate and Sato's lead stain. Samples were examined using a transmission electron microscope JEOL JEM-1400 TEM at 80 KV. Images were captured using a digital camera.

Exosome uptake and fluorescent microscopy
Exosomes were labeled with PKH67 Green Fluorescent Cell Linker (Sigma Aldrich, St. Louis, MO) in a modified protocol as described previously [38]. A2780 cells were labeled with PKH26 Red Fluorescent Cell Linker (Sigma-Aldrich) according to manufacturer's instructions. Adhered A2780 cells were exposed to exosomes at a concentration of 1 µg/10,000 cells plated in time points of 0.5, 1, 2, and 24 hours. Media was removed upon completion of all time points and cells were gently washed once with PBS and fixed with 4% paraformaldehyde. After washing with PBS, cover slides were attached using a mounting medium containing DAPI. 3-6 images were taken of each time point and overlaid using MetaMorph Software (Molecular Devices, SunnyVale, CA).

Cell viability assays
Cells were treated with carboplatin (SelleckChem, Houston, TX) diluted from a 20 mM stock in PBS or PBS alone as a vehicle control. Cell viability was assessed using Cell Titer Blue (ThermoFisher), as previously published [113]. Fluorescence was read using the Tecan Plate reader by 560/590 excitation/emission spectra. All assays were conducted in technical triplicates and replicated at least 2 times. Caspase 3/7 activity was conducted using Caspase-Glo ® 3/7 (Promega), according to manufacturer's instructions. Statistical significance between experimental and control groups was determined using Student's T-test. Values of < 0.05 were considered significant.

Colony formation
EOC cells were grown in 10 cm dishes and treated with 10 µg equivalence of exosomes per 100,000 cells for 24 hours. EOC cells were then trypsonized and counted. 100 or 200 cells were plated/well in a 6 well dish and allowed to adhere overnight. The next day either 1 or 5 µM carboplatin was added (PBS was used as a vehicle control). The media + carboplatin was replaced every 7 days. The assay was stopped when a majority of wells grew colonies observable to the naked eye. Media was removed and the cells were washed with PBS prior to fixation with a 3:1 combination of methanol and acetic acid. Colonies were stained with 0.1% crystal violet in methanol. Images of the plates were taken using the AlphaImager™ system and colonies were counted and measured using Adobe™ PhotoShop software.

SDS page and western blot analysis
Exosome samples and cell lysates (prepared in RIPA buffer) were separated by adding 40 µg protein on 7%, 10%, or 4-20% Mini-PROTEAN ® TGX ™ Precast Gels, (BioRad, Hercules, CA) and transferred to a supported nitrocellulose membrane (BioRad). The membranes were blocked with 5% non-fat milk for one hour. Membranes were incubated with primary antibodies (Supplementary Table 3) overnight and washed thrice for 10 minutes before addition of HRP conjugated anti-rabbit or anti-mouse secondary antibody (BioRad) for 1 hour. Membranes were then washed and treated with ECL Western Blotting Substrate (Fisher Scientific) according to manufacturer's instructions.

Real time PCR
Cells were harvested at 80% confluency and RNA was extracted using Trizol and Direct-sol™ RNA miniprep system (Zymo Research, Irvine CA). Reverse transcriptase was performed using the GoScript™ Reverse Transcriptase System (Promega, Madison, WI) according to manufacturer's instructions. Real time PCR was conducted using Sso Advanced™ real-time PCR Master Mix (BioRad). Primers for EMT-related genes were taken from Yew et al [14]. For miR-21 identification, TaqMan Probes (Invitrogen) and RT-PCR primers specific to Mir21 (Invitrogen) were used according to manufacturer's instructions. Primers for primary miR-21 were designed as previously published [110].

Next generation sequencing DNA extraction
EOC cells at 80% confluency were trypsonized, collected, and washed 2x with cold PBS. DNA isolation was accomplished using the DNEasy blood and tissue kit (Qiagen, Hilden Germany) according to manufacturer's instructions.

Library preparation
The TruSeq Amplicon Cancer Panel (Illumina) was used according to manufacturer's instructions. Amplicon libraries were generated by hybridizing pairs of oligonucleotides specific to targeted regions to each DNA sample. Unbound oligonucleotides were removed, and DNA polymerase and ligase were used to connect bound oligonucleotides by extension and ligation. Primers containing index sequences for multiplexing and adapter sequences for cluster generation were used for PCR amplification. Libraries were quantified using the KAPA Library Quantification Kit (KAPA Biosystems, Inc.) specific to Illumina platforms and optimized for the Roche LightCycler 480 ii. Sample libraries were normalized and equal volumes pooled in the final multiplexed sequencing libraries.

Sequencing and data analysis
Pooled libraries were sequenced on a MiSeq instrument using a 2 x 150 paired-end format using the Custom Amplicon workflow. Base calls were generated on-instrument with the Real Time Analysis (RTA) software (Illumina). Reads were aligned to the Homo Sapiens -UCSC (h19) genome assembly and the Somatic Variant Caller (Illumina) was used for identification of variants. The Illumina Variant Studio software was used to annotate all detected variants. All alternate variant calls were required to have Q scores of at least Q30 and occur at a frequency of ≥ 15%.

Mutant plasmid generation and transfection
The SMAD4 plasmid (pcDNA FLAG-SMAD4M; Joan Massague, Addgene) was altered using the Quick Change II Site-Directed Mutagenesis Kit (Agilent, Santa Clara, CA) according to manufacturer's instructions. Primers were designed using the Quick Change Primer Design software (Agilent). All sequence validation was done using Sanger sequencing by GeneWiz, INC. Transfection of plasmids into A2780 cells was accomplished using Lipofectamine® 2000 (Invitrogen) according to manufacturer's instructions. Transfected cells were maintained in media containing 500 µg/mL G418 (Corning).

Statistical analysis
Statistical analysis was performed using the two tailed Student's t-test on both excel and Graph Pad Prism Programs and one-way ANOVA analysis on Graph Pad Prism.