Oncogene Regulated Release of Extracellular Vesicles

Oncogenes alter EV abundance and size

We sought to understand how different oncogenes alter the number and contents of EVs released from transformed cells. MCF10A cells are derived from non-tumorigenic breast tissue but are susceptible to transformation by a wide range of oncogenes, allowing us to study diverse oncogene signaling pathways (Debnath et al., 2002; Martins et al., 2015). We generated a panel of isogenic MCF10A cell lines engineered to express a variety of the most commonly mutated, amplified or overexpressed oncogenes in breast and other types of cancers (Martins et al., 2015). In order to directly compare the influence of specific oncogenic signals on EV properties we used a panel of 10 cell lines in which individual oncogenes were stably expressed (Martins et al., 2015) and compared these to cells transduced with empty vectors (i.e., Puro/Blast) (Figure 1A).

EVs were isolated using a differential ultracentrifugation method as depicted in Figure 1B (Thery et al., 2006) and processed to determine their number, size, protein and RNA content. Prior to EV collection, each cell line was cultured for 48 hours in defined medium (KSR) that does not contain exogenous EVs, normally abundant in serum-containing media. This allowed us to ensure that EVs were derived from each of the isogenic MCF10A lines. To assess the number and size of EVs produced we utilized two complementary assays. First, we performed nanoparticle tracking analysis (NTA) using conditioned media from 10 different isogenic MCF10A lines. To focus mainly on small EVs (sEVs) conditioned media was pelleted at 1000g and 2000g spins to remove cell debris, apoptotic bodies and filtered through a 0.2μ filter to remove large vesicles (lEVs) (Figure 1C–D). We found that 8 out of 9 oncogene-expressing cell lines significantly increased the number of particles released compared to control (Puro/Blast) cells (Figure 1C). Amongst the oncogenes tested, MYC and AURKB-overexpressing MCF10A cells release the greatest number of EVs compared to control cells, also significantly higher than any of the other oncogene-expressing cell lines (Figure 1C). Furthermore, the size distribution of the EVs confirmed that the majority of released vesicles are sEVs <200nm in size (Figure 1D) consistent with known size distribution for exosomes (Colombo et al., 2014). MYC and AURKB-derived EVs were smaller with a modal particle size of 116.9 nm and 119.1 nm, respectively (Figure S1A). Thus, we observed that most oncogenes induced significantly increased EV release, though MYC and AURKB vesicles were far more numerous on a per cell basis but tended to be smaller (Figure 1C–D).

NTA does not permit direct visualization of individual vesicle morphology. We therefore used transmission electron microscopy (EM) to evaluate qualitative differences and confirm size changes via direct visualization of sEVs collected from the 100K fraction (Figure 1B) of a subset of cell lines: MYC, AURKB and HRAS(G12V) oncogene expressing and control MCF10A cells. We observed the typical cup-shaped morphology previously described for sEVs (Figure 1E) (Sahoo et al., 2011) in the 100K pellets from all of the cell lines. To elucidate the size differences of sEVs we measured the diameter of each sEV by random sampling from micrographs. The mean size of Puro/Blast sEVs is 120nm whereas the mean size of sEVs from AURKB and MYC overexpressing cells are 99nm and 96nm respectively. HRAS(G12V) sEVs were of similar size and appearance to Puro/Blast sEVs with a mean size of 122 nm. MYC and AURKB sEVs showed smaller size variance (sd ± 29, ±30nm, respectively) compared to control and HRAS(G12V) sEVs which demonstrated a wider size distribution (sd ± 52, ±42nm, respectively) (Figure 1F). These data indicate that different oncogenes alter the size but not the morphology of released sEVs. EM images also revealed the presence of some non-vesicular (NV) particles (exomere and protein aggregates) in the 100K pellet (Figure 1E), although prior fractionation studies have revealed that the majority of exomeres are retained in the supernatant of 100,000g spin (Zhang et al., 2019).

Since oncogenes induced differences in both the size and number of secreted sEVs, we asked if the total biomass of released sEVs was altered across the panel of oncogenes. Using the number of sEVs released and their size distribution from NTA, we determined the volume of cellular biomass released as sEVs and normalized to total cell number using a spherical model. We found that MYC and AURKB expressing cells released ~19.5- and ~ 14.6-fold more biomass compared to control MCF10A cells respectively (Figure 1G). We next measured cell volume to estimate percentage of cellular biomass released by EVs during a 48hr period of culture. We measured cell sizes using Countess automated cell counter and calculated cell volume using a spherical model. Control and HRAS(G12V) overexpressing MCF10A cells released less than 1% of their cellular volume whereas MYC and AURKB expressing MCF10As released 7.7% and 4.1% of their total cellular volume, respectively, via sEVs (Figure S1B). Remarkably, the volume of cells expressing either oncogene is within the same order of magnitude as control cells. Thus, MYC and AURKB overexpressing cells can significantly alter their biomass flux via sEV release compared to control MCF10A cells to maintain cellular homeostasis.

The protein content of EVs is regulated by the driver oncogene

We postulated that specific oncogenes may alter not only the size and number of sEVs released, but also their contents. To elucidate the heterogeneity of sEVs released from different oncogenes, we performed proteomic analysis of equal amounts of the 100K pellets from five different MCF10A cell lines including, control, AURKB, CCND1, HRAS(G12V), and MYC. We identified 2608 proteins with 2+ peptides using DIA profiling (Table S1). ANOVA analysis identified 319 proteins that were significantly different between the analyzed 100K fractions. PCA analysis of significantly different proteins showed that MYC and AURKB extracellular protein fractions cluster together, as do Puro/Blast and CCND1 fractions, whereas HRAS(G12V) fractions were distinct from the other groups (Figure 2A). Across all the samples we observed some shared proteins, as well as oncogene-specific proteins enriched in the 100K fractions (Figure 2B). We next sought to identify enriched biological pathways within these oncogenic EVs. KEGG pathway analysis of enriched proteins from MYC 100K fractions identified RNA transport and spliceosome function (Figure 2C), pathways known to be upregulated by MYC in tumor cells (Hsu et al., 2015; Kress et al., 2015). The HRAS(G12V) 100K fraction was enriched with ribosome and many metabolic pathway proteins (Figure 2C). Proteins involved in metabolism and glycolysis were also identified in exosomes from mutant KRAS(G13D) colorectal cancer cells in previous studies (Beckler et al., 2013; Zhang et al., 2018b). Our observations suggest that the enrichment of metabolic pathways could be a common RAS-induced EV phenotype.

We next examined how protein markers associated with EVs differ among oncogenic EVs (Figure 2D and 2E). We included the 10K fraction in our immunoblot analysis since some of the EV markers tested are enriched in large EV (lEV) fractions. Actinin-4 is an actin-binding cytoskeleton protein identified as a lEV marker (Jeppesen et al., 2019; Kowal et al., 2016). Actinin-4 is highly abundant in HRAS(G12V)-derived lEVs (10K) (Figure 2E). Similarly, annexin-A1, which is characterized as specific lEV marker distinct from exosomes (Jeppesen et al., 2019), is enriched in the 10K fraction compared to the 100K fractions for most oncogenes, yet highly abundant in HRAS 100K fraction (Figure 2D and 2E). These results suggest that HRAS(G12V) EVs appears to be enriched with vesicles originating from plasma membranes.

Tetraspanins (i.e., CD9, CD63 and CD81) are enriched in classical exosomes but can be found in other types of EVs as well (Crescitelli et al., 2013; Kowal et al., 2016). All three tetraspanins are very abundant in AURKB and MYC fractions suggesting classical exosomes are enriched in these oncogenic fractions (Figure 2D and 2E). Alix, one of the most abundant exosomal protein, is especially highly enriched in AURKB 100K fractions (Figure 2D and 2E). Annexin-V, an apoptotic marker protein commonly identified in purified sEV fractions but absent from classical exosomes (Jeppesen et al., 2019), is significantly enriched only in the 10K and 100K fractions from HRAS(G12V) pellets (Figure 2D and 2E).

To confirm the purity of isolated EVs, we tested both the 10K and 100K fractions with known cellular markers. We tested ER marker Grp94, mitochondrial marker Cytochrome C, and a pan-histone antibody to see if our fractions were contaminated with cellular debris or apoptotic bodies. The 100K fractions were devoid of these cellular markers in all oncogenes and Grp94 and Cytochrome C was barely detected in the 10K fraction of CCND1 (Figure S2B). Moreover, we tested our fractions with markers of NV particles (exomeres) since EM images revealed some NV particles. Ago2, MVP and HSPA13 are shown to be markers of NV fractions (Jeppesen et al., 2019; Zhang et al., 2018a, 2019). HSPA13 is not detected by proteomics or western blot analysis in our 100K fractions. Low level of Ago2 is present in proteomic and western blot analysis and enriched in AURKB 100K fractions (Figure 2D and 2E). Interestingly, MVP is highly abundant in both 10K and 100K fractions and especially enriched in CCND1 preps (Figure 2D and 2E). Although MVP is suggested as NV marker (Jeppesen et al., 2019), it was found to be enriched in the 10K fraction in this study as well as in prior study (Kowal et al., 2016). Finally, we performed density gradient ultracentrifugation of the MYC 100K fractions and found that many exosome markers reside in the lighter density fraction (sEV) while markers of NV fractions such as MVP and PKM1 (glycolytic enzyme) were in the higher density fraction (NV fraction) (Figure S2D). These observations suggest that the 100K crude pellets of oncogenes contain some NV particles, but their abundance is also altered by different oncogenes. Based on these observations the role of oncogenes in the release of NV particles and their functional roles await further investigation.

Oncogenes alter secreted miRNA composition

We next examined if secreted miRNA content depends on the oncogenic driver. We isolated equal amounts of RNA from the 100K fractions or the corresponding cells that released them. We performed qPCR analysis using a panel of 384 miRNA primers. We identified 95 miRNAs (about 1/4 of the panel) that were differentially expressed in oncogene-expressing cells compared to control cells (Figure 3A, Table S2). 72 of these miRNAs were also enriched in oncogene-derived 100K fractions compared to control 100K fractions, however there were an additional 74 miRNAs that were differentially sorted into oncogene-derived 100K fraction despite not being dysregulated in cells (Figure 3A). We performed two types of analyses on these data: 1. We asked if each oncogene caused either preferential sorting into, or exclusion from, the 100K fraction by examining the concentration of each miRNA in the 100K fraction and comparing that to the concentration of miRNAs within the cell of origin. 2. We asked if the miRNAs within the 100K fraction from each cell type differed from that of the control 100K fraction. Our goals were to discover what miRNAs the cells may be preferentially releasing, possibly to maintain homeostasis or to influence cells in their environment. We also sought to understand what miRNA handling and sorting processes may be employed when each oncogene is driving cellular phenotypes.

First, we determined which miRNAs are enriched in the 100K fraction versus their cell of origin (Figure 3B). Although the released versus cellular miRNA profiles look different for each oncogene, CCND1-overexpressing cells seem most similar to control cells in the hierarchical clustering (Figure 3B). Certain miRNAs (e.g., hsa-miR-18–3p, hsa-miR-425–3p, hsa-miR-663a) were not only enriched in several oncogene-derived fractions but also in control fractions suggesting their release is independent of oncogenic driver. On the other hand, there are certain miRNAs that are preferentially released into the 100K fraction in an oncogene-specific manner (Figure 3B).

Second, we compared miRNAs differentially released from oncogene-derived 100K fraction compared to control 100K fraction (Figure 3C, Table S3). There were 16 deregulated miRNAs common to all oncogenes either released or retained compared to control fraction (Figure 3C). AURKB and MYC derived fractions shared the highest number of deregulated miRNAs, released or retained, suggesting a common propensity to release these same miRNAs.

Third, we identified miRNAs that are uniquely over or underrepresented in 100K fractions from each oncogene-expressing cell line. CCND1 derived 100K fraction has the lowest number of differentially sorted miRNAs (3), whereas 100K fractions from AURKB, HRAS and MYC oncogene have several unique miRNAs (i.e., 24, 14, 16 respectively) (Figure 3C).

The 100K fraction-enriched miRNAs from this dataset can potentially be explored as oncogene specific markers. We selected miRNAs that are enriched in at least one of the oncogene derived fractions and presented them in Figure 3D to visualize unique or shared miRNAs (Figure 3D). miRNA-298 was the only miRNA that was upregulated in all the oncogene-derived fractions compared to control MCF10A EV fractions. There were several miRNAs that are unique to MYC, AURKB and HRAS fractions (Figure 3D). For example, miR-95 is significantly enriched in HRAS(G12V) 100K fraction and miR-34a-5p and miR-195–5p, which are well-known tumor suppressors, is highly abundant only in the MYC 100K fraction. Our data strongly suggests that oncogenes bias which miRNAs are released and identified in the 100K fraction. This could result from oncogene specific expression of the secretion machinery, localization of the miRNA handling apparatus or miRNA binding proteins.

Thus, we asked whether the RNA motifs found within the differentially sorted miRNAs could identify oncogene-specific miRNA binding proteins. We examined known miRNA motifs shown to control sorting of miRNAs into EVs and asked if these motifs are enriched in miRNAs that are sorted into 100K fraction in an oncogene specific manner. The motif hEXO “GGYU” was found to be enriched in miRNAs sorted into hepatocyte EVs by the RNA binding protein SYNCRIP (HNRPQ) (Santangelo et al., 2016) (hereinafter hEXO SYNCRIP). Likewise, two different miRNA motifs were reported to influence sorting of miRNAs into T cell-derived EVs (Villarroya-Beltri et al., 2013). The motif “GGAG” has been shown to bind to hnRNPA2B1 and is proposed to sort miRNAs bearing this motif into EVs (hereafter hnRNPA2B1 EXO). Finally, the motif “UGCA” was found to be enriched in miRNAs retained in cells (hereafter CL).

We first asked if any of these motifs were found in miRNAs either preferentially released into the 100K fraction or preferentially retained in cells in common between all oncogene expressing cell lines. All three motifs were detected in half of these miRNAs (Figure 3E). The hEXO SYNCRIP binding motif was equally represented (25%) in both miRNAs enriched in EVs and those retained in cells (Figure 3E). Interestingly, hnRNPA2B1 EXO motif was observed in 16% of the miRNAs released into the 100K fractions whereas it was detected in only 4.3% of miRNAs retained in cells. In contrast, the CL motif was abundant in miRNAs retained in cells (17%) while only 6% of the miRNAs enriched in the 100K fraction contain the CL motif (Figure s3E, Table S4). Although the hEXO SYNCRIP motif was detected in more miRNAs, hnRNPA2B1 EXO motif seems to be more EV specific in MCF10A oncogene expressing cells. It is also noteworthy that half of the miRNAs either differentially secreted into the 100K fraction or retained in cells do not contain any of the previously reported sorting motifs. This suggests that miRNA secretion or cellular retention may occur via RNA binding proteins and/or mechanisms not previously described.

To further investigate oncogene-specific release, we analyzed miRNAs from MYC and HRAS derived fractions and examined the prevalence of previously identified sorting motifs. We categorized miRNAs that are common to both oncogenes, enriched in one of them compared to other, and not enriched in either (Figure 3F). miRNAs that are enriched in the 100K fraction common to both MYC and HRAS had equal representation (~25%) for both the hEXO SYNRICP and hnRNPA2B1 EXO motifs (Figure 3F). However, the hEXO SYNCRIP motif was more abundant (46%) in miRNAs that are enriched in HRAS fractions compared to MYC fractions (Figure 3F, Table S4). Interestingly, miRNAs not enriched in HRAS or MYC fractions (i.e., enriched in AURKB and/or CCDN1 fractions) are two times more likely to have the hnRNPA2B1 EXO motif (22%) compared to hEXO SYNCRIP motif (11%). Hence, miRNAs released into HRAS derived fractions appear to contain the hEXO SYNCRIP motif more frequently than other oncogenes tested. In aggregate, mechanisms of miRNA release may be differentially employed by each cancer driver.

We also sought to determine if miRNAs enriched in the 100K fractions are associated with oncogenic or tumor suppressive functions (Table S5). For example, of the 6 miRNAs that were preferentially secreted into the 100K fractions from MYC overexpressing cells as compared to cells expressing the other oncogenes, all six have been categorized as targeting tumor suppressor or pro-growth processes ((Wang et al., 2010); Table S5). This suggests that MYC high cells eliminate growth-suppressive miRNAs via preferential EV and/or exomere release. In contrast, many of the miRNAs preferentially released from HRAS cells have been reported to have pro-tumorigenic functions (Table S5), perhaps stimulating the growth of cells within the tumor microenvironment.

Sphingolipid metabolism is altered for multiple oncogenes transformed cells but is especially critical in MYC overexpressing cells.

Due to their cone-shape structure, ceramide lipids induce negative curvature triggering inward budding of endosomal membranes (Trajkovic et al., 2008). We postulated that changes in lipid metabolism might therefore contribute to alterations in EV release by distinct oncogenes. We performed metabolomic analysis to determine how different oncogenes alter steady state abundance of lipid species in cells. We discovered significant changes in different lipids from MCF10A cells transformed by CCND1, AURKB, MYC or HRAS(G12V) relative to control MCF10A cells (Figure 4A; Table S6). Many different lipid species were highly enriched in MYC- and HRAS-overexpressing cells compared to CCND1 and AURKB overexpressing MCF10A cells (Figure 4A). Lipids with large head groups such as lysophosphatidylcholine (LPC) or phosphatidylinositol (PI) phosphates support positive curvature observed during plasma membrane budding, while phosphatidylethanolamine (PE), phosphatidic acid (PA) and diacylglycerol (DAG) which have smaller head groups, favor negative curvature (McMahon and Boucrot, 2015; Zimmerberg and Kozlov, 2006). MYC- and HRAS-overexpressing MCF10A cells were enriched with both types of lipids compared to control cells (Figure 4A). Nevertheless, ceramide lipids which induce negative curvature to generate exosomes were significantly enriched only in MYC-overexpressing MCF10A cells (Figure 4B). Ceramide can be generated via three major metabolic pathways as depicted in Figure 4C.

To evaluate lipid metabolic pathways that could contribute to EV release, we performed RNA sequencing of AURKB, MYC or HRAS(G12V) overexpressing cells which we compared to control MCF10A cells. We sought to identify deregulated enzymes in the ceramide pathway. We found that many genes involved in ceramide production were dysregulated in all three oncogenic lines. However, N-SMases, SMPD3 and SMPD4, were significantly upregulated only in MYC overexpressing cells which have the greatest abundance of ceramide lipids (Figure 4D).

These data suggest that MYC-overexpressing cells may use ceramide generation through sphingomyelin hydrolysis to produce high levels of exosomes. We tested the neutral sphingomyelinase inhibitor (GW4869) on a subset of oncogenic lines and measured EV release through NTA. We observed a significant decrease in EV release from MYC and AURKB overexpressing cells, but not in control cells, or those expressing HRAS or CCND1 oncogenes (Figure 4E). The MYC overexpressing cells were most responsive to GW4869 treatment, suggesting that ceramide-dependent EV production is especially important in MYC high cells.

We next asked if N-SMase inhibitor treatment can alter EV protein composition from cells transformed with different oncogenes (Figure 4F). As expected, majority of EV markers exhibited significant decrease in abundance following GW4869 treatment in MYC and AURKB cells (Figure 4F, and S3). Alix was one of the most significantly downregulated protein in both AURKB and MYC 100K fraction upon GW4869 treatment, supporting the role of N-SMases in the production of syndecan-syntenin-Alix containing exosomes (Baietti et al., 2012). However, we did not see a consistent decrease in CD63 levels as observed in Syndecan-Syntenin-Alix containing exosomes(Baietti et al., 2012), suggesting that MYC and AURKB cells might produce other types of CD63-containing EVs from an alternative pathway upon N-SMase inhibition (Figure 4F, and S3).

Although HRAS(G12V) EV numbers were not changed upon GW4869 treatment (Figure 4E), we observed changes in the subtypes of released EVs in both fractions (Figure 4F, and S3). Clathrin and actinin-4 abundance were significantly decreased while Cav2- and Annexin-V-containing EVs increased in HRAS(G12V) fractions (Figure 4F, and S3). Moreover, significant changes in many markers in 10K fraction, especially lEV marker actinin-4, strengthens the observation that ceramide might also play a role in budding from the plasma membranes (Menck et al., 2017).

ESCRT pathway component TSG101 increases EV release from AURKB and MYC overexpressing cells

The endosomal sorting complex required for transport (ESCRT) is a major regulator of MVB formation and is therefore also critical for EV release. ESCRT-0 complex recognizes and clusters ubiquitinated proteins, while ESCRT-I and -II complexes deform endosomal membranes to form buds where cargo is sorted. Finally, ESCRT-III together with accessory proteins cleave the buds to form intraluminal vesicles (ILVs) (Colombo et al., 2014; Henne et al., 2011; Wollert and Hurley, 2010) (model shown in Figure 5A). Gene expression of multiple components of the ESCRT pathway are deregulated in the context of different oncogenes compared to control cells (Figure 5B).

We sought to determine if specific oncogene expressing cells are reliant on ESCRT components for EV release. We selected two ESCRT components previously identified to be important for EV release, TSG101 and ALIX (also called PDCP6IP) (Baietti et al., 2012; Colombo et al., 2013), and depleted their expression via siRNAs. We then performed NTA on EVs from these depleted cells (Figure 5C). We did not observe significant changes in the number of released EVs following TSG101 knockdown from control and HRAS-overexpressing MCF10A cells. However, TSG101 knockdown significantly decreased the number of EVs from AURKB and MYC overexpressing cells, and this decrease was most pronounced in AURKB overexpressing cells (Figure 5C). ALIX depletion was not associated with diminished EV release in any of the cell lines. Thus, TSG101, but not ALIX, expression is important for mediating the number of EVs released from MYC and AURKB expressing cells.

We next explored if EV subtypes changed following depletion of TSG101 or ALIX. As expected, most classical exosome markers decreased in the MYC and AURKB 100K fractions upon TSG101 knockdown (Figure 5D, 5E). Although TSG101 knockdown did not change the EV numbers from HRAS overexpressing cells (Figure 5C), the decrease in exosomal markers CD81, Alix, and syntenin-1 was compensated via increase of EV subtypes containing clathrin- and Cav2-enriched EVs (Figure 5D, 5E). Likewise, TSG101 depletion increased the Cav2- and clathrin-containing EVs in MYC fractions.

ALIX depletion did not alter the number of released EVs in any three oncogenic lines tested (Figure 5C). Still, as expected, the release of syntenin-1-containing EVs significantly decreased in all oncogenes (Baietti et al., 2012) (Figure 5D, 5E). Upon ALIX depletion, a significant increase in the subtype of CD81- and TSG101-containing EVs resulted in compensation for MYC and AURKB 100K fraction (Figure 5D, 5E). Like TSG101 depletion, ALIX depletion in HRAS cells increased the subtypes of Cav2, actinin-4, and clathrin-containing EVs (Figure 5D, 5E).

These results demonstrate that upon TSG101 or ALIX depletion, oncogene-expressing MCF10A cells can induce EV release from alternative pathways (e.g.; either originating from MVBs or plasma membranes). This compensation seems to be oncogene specific. Amongst the oncogenes tested, AURKB-overexpressing cells appear to rely on ESCRT-dependent pathways more than other oncogene-overexpressing cells.

Decreased lysosomal genes and activity is associated with increased EV release

Recent studies have found that increased EV secretion can occur under conditions in which lysosome function is impaired (Adams et al., 2021; Latifkar et al., 2019; van de Vlekkert et al., 2019). Therefore, we explored if lysosomal function might be inhibited in MYC and AURKB expressing cells, the two oncogenes that account for the greatest number of released EV. Indeed, KEGG pathway analysis of RNA expression revealed that lysosomal mediated degradation is one of the most downregulated pathways for MYC and AURKB overexpressing cells, but not HRAS high cells (Figure 6A). Expression of multiple lysosome-associated enzymes are downregulated in MYC and AURKB overexpressing MCF10A cells compared to control cells (Figure 6B). The MiT/TFE family of transcriptional factors, which include TFEB, TFE3 and MITF, are key regulators of lysosome biogenesis and function (Palmieri et al., 2011; Perera and Zoncu, 2016; Sardiello et al., 2009). MiT/TFE factors are upregulated in pancreatic cancers and are required for maintaining high autophagy and lysosome activity that fuel the bioenergetic needs of tumors (Perera et al., 2015). Consistent with downregulation of a lysosome associated gene signature in MYC and AURKB cells, we find that TFE3 and TFEB are also significantly downregulated in these cells relative to control (Puro/Blast) and HRAS cells (Figure 6C). In accordance with TFEB/TFE3 downregulation, many TFEB regulated autophagy and lysosomal genes are also downregulated in AURKB and MYC cells (Figure 6D). In addition, Lysotracker dye staining, which measures lysosome pH showed a significant decrease in MYC cells, consistent with decreased lysosomal activity (Figure 6F). As maintenance of acidic luminal pH is required for lysosomal degradation, we next determined the expression and activity of lysosome hydrolases. We found that several hydrolases including Cathepsin B (CTSB), Cathepsin D (CTSD), and Acid alpha-glycosidase (GAA) and the Mannose 6-phosphate receptor (M6PR) - a transmembrane protein essential for transport of hydrolases to the lysosome – showed significantly reduced expression in MYC and AURKB cells relative to control cells and some were increased in HRAS cells (Figure 6E). Accordingly, CTSB activity was also reduced in MYC and AURKB cells (Figure 6G), but increased in HRAS cells. Together, these data indicate that MYC and AURKB expression leads to a profound decrease in lysosome activity which, in accordance with recent studies (Adams et al., 2021; Latifkar et al., 2019; van de Vlekkert et al., 2019), favors increased EV release.

MYC overexpression increases EV production but can be attenuated by MEK/ERK signaling

In the isogenic MCF10A system MYC overexpression resulted in the greatest increase in EVs released compared to parental cells (Figure 1). MYC regulates a wide variety of processes, including but not limited to, proliferation, apoptosis and metabolism (Camarda et al., 2017; Gabay et al., 2014; Meyer and Penn, 2008) and is overexpressed in many of the most aggressive types of human cancers, such as lymphoma, receptor triple negative breast cancer, neuroblastoma and liver cancer (Horiuchi et al., 2014; Lin et al., 2012). We sought to determine if MYC overexpression affects EV production from other cell types. First, we asked if turning MYC off alters EV production. EC4 cells are derived from a conditional MYC transgenic mouse model of liver cancer in which MYC transgene expression is rapidly inhibited by treatment with doxycycline (Anderton et al., 2017; Cao et al., 2011). When MYC expression was turned off for 2 days, EV production decreased 5-fold (Figure 7A). Thus, acutely inhibiting MYC expression can rapidly reprogram cells to decrease EV release. Second, we utilized human retinal pigment epithelial (RPE) cells that constitutively overexpress human MYC (RPE-MYC) or a control plasmid (RPE-NEO) (Goga et al., 2007). MYC overexpression significantly increased EV release from RPE cells as well (~ 2-fold) (Figure 7B).

We next sought to understand how EV release is altered in human tumor cells in which MYC is known to be one of the common oncogenic drivers. First, we measured EV release from three different acute myeloid leukemia (AML) lines which have variable MYC expression (Figure 7C) and MYC copy number (Brondfield et al., 2015). We observed a strong correlation between the level of MYC overexpression and EV release in AML lines (Figure 7C, F). Next, we analyzed EV release in neuroblastoma cancer lines in which MYCN amplification is a common characteristic of aggressive neuroblastomas with the poor patient outcome (Bosse and Maris, 2016). We picked five MYCN amplified neuroblastoma lines with a range of MYCN overexpression and observed a similar trend where MYCN high cells released higher numbers of EVs (Figure 7D and 7F).

Last, we explored triple negative breast cancer cell lines, a tumor type in which MYC is frequently overexpressed (Horiuchi et al., 2012), but in which additional oncogene pathways are deregulated. For example, the SUM159 line has mutations in HRAS and PIK3CA whereas the MDA-MB-231 line has mutations in BRAF and KRAS (Figure S4A) (Cerami et al., 2012; Gao et al., 2013). When we analyzed EV release from these five different MYC-high TNBC lines, we did not observe a correlation between MYC abundance and EV release (Figure 7E and 7F). We sought to determine if interactions with other oncogenic pathways in TNBC cells might account for this discrepancy in EV release. MDA-MB-231 cells produced the lowest number of EVs among the TNBC lines. We noticed that multiple genes involved in RAS/MEK/MAPK pathways were mutated in these cells (e.g., KRAS, BRAF, NF1) (Figure S4A). Hence, we asked if activated signaling via this pathway might diminish EV release despite the abundance of MYC expression. To test this hypothesis, we checked phosphorylated ERK, which is a downstream signal in the activated RAS/MEK/MAPK pathway (Figure 7G) and observed a significant inverse correlation between activated p-ERK levels and EV release in TNBC cells (Figure 7H). To determine if inhibition of MEK, which blocks ERK phosphorylation and activation, could elicit increased EV release in these cells, we treated MDA-MB-231 cells with the MEK inhibitor trametinib. Trametinib treatment at doses sufficient to inhibit ERK phosphorylation, resulted in a significant increase in EV release from both MDA-MB-231 and BT549 cells, without inducing an appreciable change in cell viability (Figure 7I–J and S4B). Hence, while MYC overexpression induces a cellular program that increase EV release, it can be attenuated by other signaling pathways that alters cellular homeostasis.

Limitations of the Study

We acknowledge that there is an interplay between multiple oncogene pathways when they are activated together in some TNBC cells; activation of these oncogenic pathways together affects the abundance of EV release. Furthermore, we found differential oncogene-dependent miRNA release, however other types of small RNAs, not explored in this study, may also be identified as oncogene-associated RNA secretome markers. Finally, our biomass calculations were based on a spherical vesicle volume model using NTA. Further biophysical characterization of released vesicles and particles could be performed by AFM or CryoEM to assess biomass loss more precisely.

RESOURCE AVAILIBILITY

EXPERIMENTAL MODEL AND SUBJECT DETAILS

METHOD DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were performed using GraphPad Prism 9 software. Groups were compared using unpaired or paired Student’s t-test. Pearson correlations were calculated using Prism 9. One way ANOVA followed by multiple test correction (control FDR with standard Benjamini-Hochberg procedure) was performed using Scaffold DIA software for proteomics data. R package was used to create heatmaps. Details regarding the statistical analysis of the data are provided in the corresponding figure legends and/or the Methods sections above. p values of less than 0.05 are considered to be significant.