Secreted midbody remnants are a class of extracellular vesicles molecularly distinct from exosomes and microparticles

During the final stages of cell division, newly-formed daughter cells remain connected by a thin intercellular bridge containing the midbody (MB), a microtubule-rich organelle responsible for cytokinetic abscission. Following cell division the MB is asymmetrically inherited by one daughter cell where it persists as a midbody remnant (MB-R). Accumulating evidence shows MB-Rs are secreted (sMB-Rs) into the extracellular medium and engulfed by neighbouring non-sister cells. While much is known about intracellular MB-Rs, sMB-Rs are poorly understood. Here, we report the large-scale purification and biochemical characterisation of sMB-Rs released from colon cancer cells, including profiling of their proteome using mass spectrometry. We show sMB-Rs are an abundant class of membrane-encapsulated extracellular vesicle (200-600 nm) enriched in core cytokinetic proteins and molecularly distinct from exosomes and microparticles. Functional dissection of sMB-Rs demonstrated that they are engulfed by, and accumulate in, quiescent fibroblasts where they promote cellular transformation and an invasive phenotype. Rai et al. characterise the properties of secreted midbody remnants, showing they are distinct from exosomes and microvesicles. The authors also find that these vesicles are engulfed by cells and promote anchorage independent growth and invasive phenotypes in NIH3T3 fibroblasts.

A t the terminal stage of cell division, the prospective daughter cells remain connected by a thin intercellular bridge containing the midbody (MB), a transient organelle responsible for mediating final abscission during cytokinesis [1][2][3] . The MB is then inherited by one of the newly-formed daughter cells where they perform non-mitotic functions [3][4][5][6][7][8] . While MB-Rs are degraded by autophagy 3 they can also accumulate intracellularly and influence cell fate 8 . An alternative fate of MB-Rs involves extracellular secretion as sMB-Rs [9][10][11] . Over the past 30 years the prevailing view has been that sMB-Rs are degraded extracellularly 11 . However, recent evidence shows that sMB-Rs can also be engulfed by non-sister cells 8,12 . Because MB-Rs can influence cell signalling and cell fate in daughter cells [8][9][10]13,14 , it has been speculated that they might perturb cell signalling at distal sites 15 , however, this question remains largely unexplored.
Previously, we observed that MKLP1 (also known as KIF23 or ZEN-4) 16 , a component of the centraspindlin complex, copurified with shed microvesicles (sMVs, also referred to as microparticles and ectosomes 17 ) isolated from the cell culture medium of the human colorectal cancer (CRC) cell line LIM1863 18 . Because centraspindlin is a core component of MBs 16 and MB-Rs 8 , we reasoned that MB-Rs might be released from cell lines into culture medium (CM) in sufficient quantities to permit their biochemical and functional characterisation. The focus of this paper is directed at the large-scale preparation of sMB-Rs that would allow us to undertake their biophysical and functional characterisation, to ask whether in vitro release of MKLP1 from LIM1863 CRC cells is a general phenomenon or a cancer-cell specific process, and whether MKLP1 is an indicative marker of sMB-Rs.

Results and discussion
Cancer cell-derived midbody remnants are shed into the extracellular space. To determine whether sMB-Rs are secreted into the extracellular space, we first investigated whether MKLP1 could be used as a reliable marker for MB/MB-R detection. For this purpose we used CRC cells (SW620) grown in 2D culture. Fluorescent microscopy revealed that the MKLP1 antibody readily stained punctate structures nested in the middle of the intercellular bridge between dividing cells (revealed using betatubulin antibody) typical of MBs 1-3 ( Supplementary Fig. 1). Next we examined whether MKLP1 stained for MB-Rs. During the final stages of cytokinesis the inner leaflet phosphatidylserine (PS) of the MB membrane flips to the outer leaflet 19,20 resulting in PS enrichment in the outer leaflet of the MB-R membrane. Importantly, PS on outer leaflet of sMB-Rs is required for their engulfment by cells 20 . In our study, fluorescent microscopy revealed that MBs localised between dividing SW620 cells did not stain with annexin V (Fig. 1a, top panel), whereas annexin V-and MKLP1-staining structures similar in size to MB-Rs associate with non-dividing cells. This observation is consistent with the presence of a MB-R inherited by one of the daughter cells post cytokinesis (Fig. 1a, middle panel). Moreover, co-staining of annexin V-and MKLP1 of MB-Rs was detected in the extracellular space (Fig. 1a, bottom panel) consistent with these particles being sMB-Rs. These anti-MKLP1-staining MB-R/sMB-R punctate structures were distinct from cellular debris as evidenced by lack of genomic DNA staining (Hoechst stain); their size range (~500 nm particle diameter) is consistent with previous report for sMB-Rs 9,10,19 . Collectively, these data show that MKLP1 can be used as a stereotypic marker for CRC cell MB/sMB-Rs. To provide additional proof that MB-Rs are secreted from cell lines into the extracellular space we generated fluorescently-labelled MB-Rs. For this purpose, we constructed SW620 cells stably-expressing plasma membrane-targeting GAP43 (1-20 a.a.) 21 fused to GFP (SW620-GAP-GFP cells) (Supplementary Fig. 2). In Fig. 1b it can be seen that fluorescently-labelled/ MKLP1-positive MB-Rs are shed extracellularly.
MB-Rs have been reported to accumulate in small subpopulations of cancer cells and associate with stemness 8 . It can be seen in Supplementary Fig. 3a, b that while only a small population of SW620 and SW480 cells (0.9-1.6%) grown in culture accumulate MB-Rs, a large pool of non-cell associated MB-Rs are observed, a finding consistent with shedding of MB-Rs into the extracellular medium. This observation was corroborated using an antibody to the centraspindlin component RACGAP1 ( Supplementary  Fig. 3a). To rule out the possibility that MB-R shedding is an artefact of 2-D cell culturing, we prepared 3-D cultures of SW620 and SW480 cells in Matrigel TM matrix (Fig. 1c) and tested for MB-R shedding. These data show that both SW620 and SW480 cells grown as spheroids also shed MB-Rs into their extracellular space (Fig. 1c, upper and middle panels).
Next, we set out to determine whether cancer cells shed MB-Rs in vivo. For this purpose, we established SW620-GAP-GFP cells as subcutaneous xenografts in mice. Immunohistochemical analysis of ectopic tumours from these mice using MKLP1 antibodies revealed GFP-tagged MB-Rs in the extracellular space (Supplementary Fig. 4). To determine whether sMB-Rs could be detected in human colon cancer tissues, we analysed MKLP1-or RAC-GAP1-antibody based immunohistochemical images of human colon cancer tissues publicly available from the Human Protein Atlas (http://www.proteinatlas.org/) (Fig. 1d, Supplementary Figs. 5 and 6). Strikingly, large pools of MKLP1-or RACGAP1staining punctate structures were detectable in the extracellular space (lumen) of CRC tissues in greater abundance compared to non-disease colon tissue ( Supplementary Fig. 7). Further, we also show that in stark contrast to non-polarised SW480-/ SW620spheroids that shed their sMB-Rs non-directionally (Fig. 1c), highly-polarised spheroids SW1222 cells (Fig. 1c, lower panel) or organoids cultured from mouse-derived intestinal crypts (Supplementary Fig. 8) shed their MB-Rs into the central lumen (stained by filamentous actin). These observations are consistent with the emerging role of MB-Rs in cell polarity 6,7 and boost the notion that cancer cells actively shed MB-Rs into the extracellular space.
sMB-Rs can be isolated from the culture medium of SW620 cells in high yield. As a first step towards purifying sMB-Rs from CRC SW620 cell line culture medium it was important to establish which of the two major EV classes sMB-Rs belong to -exosomes or shed microvesicles/microparticles 17 . sMB-Rs from CRC SW620 cell line culture medium it was important to establish which of the two major EV classes sMB-Rs belong to -exosomes or shed microvesicles/microparticles 17 . EV classes differ in size range (exosomes typically 30-200 nm and sMVs, 50-1300 nm) and mode of biosynthesis (exosomes being of endosomal origin and sMVs forming by plasma membrane budding) 17 . Using a differential centrifugation strategy 22 , we first separated sMVs (which pellet at 10,000 × g) from exosomes which pellet at 100,000 × g (Fig. 2a) and showed sMB-Rs co-pellet with crude sMVs, but not exosomes as evidenced by western blot analysis using RACGAP1 antibody ( Supplementary Fig. 9). Similar results were obtained using SW480 and LIM1863 cell lines ( Supplementary Fig. 9). We next used an orthogonal step, isopycnic (iodixanol-density) centrifugation 23,24 , to further fractionate sMVs based on buoyant density (Fig. 2b, c). Two well-separated sMV fractions with distinct buoyant densities were foundlow-density sMVs (sMV-LD fractions 7&8, 1.13-1.14 g ml −1 ) and high-density sMVs (sMV-HD fractions 9&10, 1.22-1.30 g ml −1 ) (Fig. 2a-c, Supplementary  Fig. 10). Centraspindlin markers MKLP1 and RACGAP1 identified the sMV-HD fraction as highly enriched in sMB-Rs; this fraction was subjected to further biochemical and functional characterisation.
Both sMV-HD and -LD fractions displayed low or nondetectable levels of ALIX, TSG101, CD81 and CD82 and CD63 using Western Blotting (Fig. 2d, Supplementary Figs. 11 and 12), however, both ALIX and TSG101 were detected in sMB-R proteome data set (Supplementary Data 1). The yield of sMB-Rs (i.e., sMV-HD fraction) from 450 ml of SW620 cells grown in continuous culture (10 days, 3 Bioreactor Cell Line™ flasks) was 1080 µg protein (~67.986 × 10 6 sMB-R particles). Cryo-electron microscopy revealed sMB-Rs (sMV-HD) with particle diameters in the range 200-600 nm partially overlapping with sMV-LD particles, but significantly larger than exosomes (30-200 nm) (Fig. 2e, f). This finding is comparable to particle size determinations obtained using nanoparticle-tracking analysis ( Supplementary Fig. 13) and sMB-Rs (~300 nm) based on conventional electron microscopy 9 . The presence of MKLP1-positive sMB-Rs in sMV-HD fractions, but not sMV-LD or exosome fractions, was further validated by aldehyde sulfate latex bead capture/fluorescence microscopy ( Fig. 2g). In RACGAP1 immunoprecipitation analysis of the sMV-HD fraction, MS-based proteomics identified MKLP1, an integral component of the MB centraspindlin complex of RACGAP1 and MKLP1/KIF23 16 and other midbody components known to interact with RACGAP1, including PLK1, RHOA, CIT, KIF14 and KIF1A (Supplementary Data 2, Supplementary Fig. 14). We further showed that MB-R shedding by cancer cells is a widespread phenomenon, being observed for primary (e.g., SW480, LIM1863) as well as metastatic CRC cell lines (COLO 205 and T84) and the breast cancer cell line MDA MB 231 (Supplementary Figs. 15 and 16). Moreover, these data indicate that sMB-Rs (sMVs-HD) represent a significant portion of total secreted EVs (exosomes and sMVs-LD fractions) depending on cancer cell type (Fig. 2h).
Protein profiling of shed midbody remnants. Next, we performed a comparative proteome analysis of SW620 cell-derived sMB-Rs (sMV-HD), sMV-LD and exosomes using a label-free quantitative mass spectrometry (MS) approach 18  b Immunofluorescence microscopy analysis of SW620-GAP-GFP cells using anti-MKLP1 antibodies. Nuclei (blue) were stained with Hoechst stain. Inset: higher magnification of GFP-tagged sMB-Rs in the extracellular space. Scale bar, 10 µm. c Immunofluorescence microscopy analysis of SW620, SW480 and SW1222 cells cultured in Matrigel TM matrix using anti-MKLP1 antibodies (green) and Alexa Fluor 594 Phalloidin (red) to stain actin. Nuclei (blue) were stained with Hoechst stain. White arrow heads point show sMB-Rs. Scale bar, 10 µm. d Immunohistochemistry analysis of normal human colon tissue and colon cancer tissue (adenocarcinoma) using anti-MKLP1 and anti-RACGAP1 antibodies. Red arrows indicate anti-MKLP1 or anti-RACGAP1 staining extracellular sMB-Rs. Images obtained from Human Protein Atlas (http://www.proteinatlas.org/) with permission.
proteins were identified in sMB-Rs, 2153 in sMVs-LD, and 1929 in exosomes (Supplementary Data 1) and 382, 144 and 236 proteins, respectively are uniquely identified (Venn diagram, Fig. 3a) indicating that these three vesicle types are molecularly distinct from one another. The relative abundance of proteins in each EV subtype, based on normalised spectral counts, is shown in the heatmap (Fig. 3b). Notably, proteins associated with cytokinesis such as microtubule-bundling proteins 25 , the centraspindlin complex 16 and chromosomal passenger complex 26 are selectively enriched in sMB-Rs (the sMV-HDs fraction), but not in sMV-LD and exosome fractions. These cytokinesis-signature proteins found exclusively in sMB-Rs boost our argument that sMB-Rs represent a new category of EV, hitherto undescribed in the EV literature.
To further address the functionality of sMB-R proteins, we conducted a gene-annotation enrichment and pathway analysis (DAVID 38 ) (version 6.8) using the Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) databases (Supplementary Data 5). This analysis revealed 207 proteins in b Heatmap illustration of proteins identified in sMB-R (sMV-HD), sMV-LD and Exos. Proteins present in higher abundance in sMB-R (red) as compared to sMV-LD and Exos include conserved cytokinetic proteins as well as additional cytokinetic proteins. *Proteins uniquely identified in sMB-Rs. **Proteins enriched (fold change >2) in sMB-R compared to sMV-LD and Exos. c STRING-based protein-protein interaction network analysis of 928 enriched proteins in sMB-Rs (sMV-HD) compared to sMV-LD and Exos. The interactions were "evidence"-based, with "experiments" as active interaction source and interaction threshold set at 0.900 (highest confidence). Disconnected nodes in the network are hidden. Proteins identified under biological processes or molecular processes (Gene Ontology) are indicated. Centralspindlin complex components (RACGAP1 and KIF23/MKLP1) are also indicated. Peptide spectral profiles are displayed on the right. f Immunofluorescence microscopy of SW620 cells using anti-MKLP1 and anti-KRAS G12V antibodies. Nuclei (blue) were stained with Hoechst stain. White arrows indicate the position of MB and MB-Rs. Inset represents higher magnification. Scale bar, 10 µm. g Western blot analysis of exosomes, crude 10,000 x g sMVs, and isopycnic (iodixanol-density) gradient centrifugation fractions of sMV-LD and -HD/ sMB-Rs using anti-KRAS G12V antibody. COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-021-01882-z ARTICLE COMMUNICATIONS BIOLOGY | (2021) 4:400 | https://doi.org/10.1038/s42003-021-01882-z | www.nature.com/commsbio sMB-Rs involved in regulation of "signal transduction"; amongst these 'MAPK signalling", "Ras signalling pathway" and "Pathways in cancer" are preeminent ( Supplementary Fig. 17).
Shed midbody remnants are taken up by fibroblasts. To address whether sMB-Rs from cancer cells can influence non-cancer cells, like exosomes and microparticles, we treated NIH3T3 fibroblasts with purified sMB-Rs for 2 h and then used MLKP1 and RAC-GAP1 antibodies to evaluate vesicle uptake (Fig. 4a). Compared to untreated fibroblasts, sMB-R-treated fibroblasts displayed an approximate 4-fold increase in uptake and accumulation of MKLP1/RACGAP1-positive sMB-Rs (in green, up to eight sMB-R green puncta per recipient cell) (Fig. 4a, b). Uptake of sMB-Rs was evident within 1 h (Supplementary Fig. 19). Confocal microscopy revealed that sMB-Rs were internalised by NIH3T3 fibroblasts (Fig. 4c) and can deliver their protein cargo (Fig. 4d).
We envision that several cargo proteins in sMB-Rs collectively signal in recipient cells. These include MEK1/2/3, IQGAP1, PAK1/2, RAC1, SEPTINN7, SPTAN1, TLN1, XPO1 and YWHAB, which are effectors of various signalling pathways. Additionally, in our proteomic data set we also detect FGF3 and its receptor FGFR4. Because sMB-Rs released into the extracellular space during symmetric abscission contain the membrane envelope that originates from the plasma membrane of the parent cell, it is conceivable that MB-Rs might deliver functional FGF3/ FGFR4 signalling complex to the recipient cells and activate its downstream signalling pathways 39 . It is well documented that soluble secreted signalling molecules bound to their cognate receptors can be loaded onto the EV-surface (for example, TFGβ-1 40 ). Non-specific binding of particles on EV surfaces has also been observed in physiological conditions such as binding of lipoprotein particles to blood EVs 41 . Whether specific or nonspecific binding of factors on vesicular surfaces is of physiological significance remains an open question. Thus, we anticipate that several players collectively signal upon sMB-R uptake in recipient cells. In addition, although sMB-Rs are internalised by fibroblasts, downstream signalling could be mediated by engagement of surface receptors 20 .
SW620 cell-derived sMB-Rs promote anchorage independent and invasive phenotype in fibroblasts. Because cancer EVs play an important role in cellular transformation, such as acquisition of invasive phenotype and anchorage independent cell growth capacity 42 , we reasoned that SW620-derived sMB-R uptake by NIH3T3 fibroblasts might promote cell invasion and anchorageindependent growth capacity. To test whether sMB-Rs promote cell invasion, NIH3T3 fibroblasts were incubated with sMB-Rs (30 µg ml −1 ) for 2 h, overlaid onto Matrigel TM matrix coated inserts on a Transwell invasion assay plate, and the number of invading cells quantified (Fig. 5a). In contrast to control (untreated) fibroblasts that failed to invade, sMB-R-treated fibroblasts displayed significantly higher invasive capacity (>14fold increase, p < 0.005, Fig. 5a); this invasive capacity was attenuated by pre-treatment of NIH3T3 fibroblasts with MEK inhibitor selumetinib (AZD6244). Next, to test whether sMB-Rs promote anchorage-independent growth, NIH3T3 fibroblasts were incubated with sMB-Rs (30 µg ml −1 ) for 2 h and grown in 0.6% soft-agar suspension in a soft agar colony formation assay (Fig. 5b). Compared to control fibroblasts treated with vehicle alone that failed to form colonies, sMB-R-treated fibroblasts formed significantly greater numbers of colonies (>14-fold increase, p < 0.005) on soft agar, which was attenuated by pretreatment of NIH3T3 fibroblasts with selumetinib. This fibroblast cell transforming capacity of sMB-Rs is comparable to that of exosomes and sMVs-LD ( Supplementary Fig. 20). Furthermore, we subjected purified sMB-Rs to size-based filtration (220 nm) to remove any remaining exosomes (Fig. 5c). The pro-invasive signalling capacity on fibroblasts by purified sMB-Rs was solely observed in the retentate that contains larger sMB-Rs versus the flowthrough, which contains residual exosomes and soluble secretome. We also show using 400 nm polystyrene latex beads that the acquired invasive capacity is not due to mere exposure to particles (Fig. 5d). Additionally, depletion of sMB-Rs using Annexin V resulted in~30% reduction in the invasive phenotype of NIH3T3 fibroblasts (Fig. 5d). Although Annexin V has been shown to promote invasion in cancer cells 43,44 , Peterman et al. 20 showed that PS on the outer leaflet of sMB-Rs is required for their engulfment by cells, thus any residual Annexin V carrying over to the invasion assay, albeit in very low amounts, would potentially reduce sMB-R-uptake by fibroblasts and thereby reduce their invasive capacity. The mechanism by which sMB-Rs exert long-term effect is currently not understood. Similar to our findings, one-time treatment of MDA-MB-231 cells with sMB-Rs from Hela cells resulted in enhanced soft agar colony forming capacity (as evident from larger colonies over 14 days) 20 . Further, Peterman et al. showed that MB-Rs taken up by recipient cells were still present after 48 h post-feeding, potentially allowing for continuous signalling through at least two different pathways: αVβ3-FAK-Src and EGF-EGFR. Similar observations has also been made in the EV field where a single stimulation of U373 cells with EGFRvIII-containing EV caused a 2-fold increase in anchorageindependent soft-agar colony formation of U373 cells (over a period of 3 weeks), whereas exposure to the equivalent amount of microvesicles devoid of EGFRvIII resulted in no significant increase in colony forming capacity 45 .
These findings suggest that upon uptake of sMB-Rs, initial signalling in recipient cells is sufficient to support or enhance a phenotype, in this case anchorage independent growth, for up to 10-14 days. Alternatively, MB-Rs may persist longer than 48 h to continually signal. In this regard, several studies have shown that MB-Rs persist in cells for extended periods. For example, accumulation of asymmetrically inherited MB remnants (up to 20 MB-Rs per cell) was shown to persist in stem cells and cancer cells 8 . Exogenously supplemented sMB-Rs have also been shown to persist for up to 48 h in cells following uptake 20 . While, phagocytosed particles are rapidly degraded within 2-5 h by fusing with lysosomes, internalised sMB-Rs can persist within actin-coated endosomes and thereby evade lysosomal degradation 8,20 . In our study, we show that sMB-Rs can persist in recipient cells for up to 48 h ( Supplementary  Fig. 21). However, it is conceivable that in vivo, continuous exposure to sMB-Rs might be required to support reprogrammed phenotype in recipient cells, as is the case of exosomes within the TME 45 . Although EVs have been previously shown to transfer mutated proteins that function in recipient cells, whether sMB-Rs deliver functional mutant KRAS warrants future investigation. On the other hand, sMB-Rs might be able to mediate epigenetic reprogramming of cells in a manner similar to exosomes 46 . In conclusion, characterising the biochemical properties of sMB-Rs is an indispensable first step towards understanding their underlying functionality. In this work we show that colon cancer-derived sMB-Rs can be isolated in high yield from the culture medium of SW620 cells grown in continuous culture tanks. Using a combination of differential centrifugation (10,000 × g) and OptiPrep™ (iodixanol) density gradient centrifugation~1 mg of highly-purified sMB-Rs (based on protein concentration) exhibiting a range of particle diameter (200-600 nm) and buoyant density (in the range 1.22-1.30 g ml −1 ) were obtained. GeLC-MS/MS analysis shows that sMB-Rs contain "cytokinesis signature proteins" (microtubule-bundling proteins, the centraspindlin complex (MKLP1/KIF23and RACGAP1) and chromosomal passenger complex proteins) not seen in exosomes and sMVs/microparticles. Functional studies show that sMB-Rs, like exosomes and low-density sMVs (buoyant density 1.08-1.14 g ml −1 ) can be taken up and accumulate in quiescent fibroblasts where they promote cellular transformation and a proinvasive phenotype 47,48 . Collectively, our findings show for the first time that sMB-Rs represent a third major class of EV molecularly distinct from exosomes and shed microvesicles/microparticles (Fig. 6). Our findings provide significant insights into sMB-R biology, suggesting this class of circulating EV may not merely constitute remnants of cytokinesis, but might also possess an unexpected role in cancer biology. Isolation of exosomes, sMV-LD and sMV-HD/sMB-Rs. Cells (SW620, SW480, MDA MB 231 and U87) were cultured in CELLine AD-1000 Bioreactor classic flasks (Integra Biosciences) 50 . SW620 or SW480 cells (3 × 10 7 ) in 15 ml of RPMI media (supplemented with EV-depleted 10% FBS and 1% Pen/Strep), or MDA MB 231 or U87 cells (3 × 10 7 ) in 15 ml of DMEM medium (supplemented with EVdepleted 10% FBS, 1% Pen/Strep) were added to the lower cell-cultivation chamber. The upper nutrient supply chamber contained 500 ml RPMI or DMEM (5% FBS, 1% Pen/Strep) that was replaced every 4 days. Cells in the lower cell cultivation chamber were allowed to adhere for 72 h at 37°C with 10% CO 2 . Thereafter, the lower chamber was gently washed with serum-free RMPI or DMEM medium. For SW620 or SW480 cells, 15 ml of RPMI media (supplemented with 0.5% (v/v) insulin transferrin selenium (Invitrogen) and 1% Pen/Strep) was added. For MDA MB 231 or U87 cells, 15 ml of DMEM medium (supplemented with EV-depleted 10% FBS, 1% Pen/Strep) were added to the lower cell-cultivation chamber. Thereafter, culture medium (CM) in the cell cultivation chamber was replaced each day. CM was sequentially centrifuged at 500 × g for 5 min (4°C) and 2000 × g for 10 min (4°C). The resultant supernatant was centrifuged at 10,000 × g (30 min) to pellet crude sMVs. The supernatant was then centrifuged at 100,000 × g (1 h) to pellet crude exosomes. Crude sMVs and crude exosomes were resuspended in 500 µl PBS and subjected to isopycnic (iodixanol-density) ultracentrifugation 24,51 . Briefly, a discontinuous gradient of OptiPrep TM (iodixanol solution) was prepared by layering 40% (3 ml), 20% (3 ml), 10% (3 ml) and 5% (2.5 ml) of iodixanol solution in a 14 × 89 mm polyallomer tube (Beckman Coulter). Dilutions of iodixanol solution were made in 0.25 M sucrose/10 mM Tris (pH 7.5) solution. Crude sMVs and crude exosomes (in 500 µl PBS) were overlaid and subjected to centrifugation at 100,000 × g for 18 h (4°C). Next, twelve 1-ml fractions were collected, diluted in PBS (2 ml) and centrifuged at 100,000 × g for 1 h. The supernatant was discarded and pellets were further washed in PBS (500 µl) with final resuspension made in 100 µl of PBS and stored at −80°C until further use. Human colon carcinoma LIM1863 cells were cultured in T175 flasks in RPMI medium supplemented with 0.5% (v/v) insulin transferrin selenium and 1% Pen/ Strep and CM harvested as previously described 18 and subjected to EV isolation strategy as described above.
For 3-D culture, 500 cells were mixed with 50 µl Growth Factor-Reduced Matrigel TM matrix (Corning) and overlaid onto a Nunc® Lab-Tek® Chamber Slide™ system. The matrix was allowed to polymerise at 37°C for 1 h and gently overlaid with growth medium. After 4-8 days, 3-D cultures were fixed with 2% aqueous formaldehyde and subjected to immunofluorescence assay.
Animal experiments were performed in accordance with La Trobe University Ethics committee guidelines. SW620 GAP GFP cells (1 × 10 6 cells/site) were subcutaneously injected into both inguinal regions of two NOD/SCID male mice to establish a total of four tumour xenografts. After 4 weeks, mice were killed, and tumours were excised, fixed in 4% aqueous formaldehyde, incubated in 20% sucrose solution for 48 h, embedded in optimum cutting temperature solution (Tissue-Tek®) and frozen (using isopentane). Sections (20 µm) were then subjected to immunofluorescence assay using mouse anti-MKLP1 antibody (1:100).
Isolation and culturing of intestinal crypts (as organoids) from small intestine or colon of C57BL/6 mice was performed using Gentle Cell Dissociation Reagent (STEMCELL TM Technologies) and IntestiCult™ Organoid Growth Medium (Mouse) (STEMCELL TM Technologies) as per manufacturer's instructions. Organoids were cultured in Growth Factor-Reduced Matrigel TM matrix for 7-10 days, fixed with 2% aqueous formaldehyde and subjected to immunofluorescence assay.
To quantify sMB-R particle number, 10 µg of sMB-Rs were subjected to immunofluorescence labelling as described above, using either mouse anti-MKLP1 (1:100, Santa Cruz Biotechnology) or mouse IgG isotype matched antibody (Abcam), and probed with Alexa Fluor 488-conjugated goat anti-mouse IgG (1:200). Labelled sMB-Rs were embedded in Matrigel TM , and MKLP1 + particles imaged using Zeiss AxioObserver Z1 microscope and numbers quantified using Image J software v1.49e. RACGAP1 immunoprecipitation assay. Dynabeads TM Protein G (Life Technologies) (10 µl) were conjugated with 1 µg RACGAP1 antibody (Santa Cruz) or mouse IgG isotype matched antibody (BD Biosciences) for 15 min at room temperature under continuous rotation. Antibody-bead conjugates were collected using a magnet. Next, sMB-Rs (200 µg) were solubilized in 0.5% TX-100-PBS (supplemented with Complete TM EDTA-free Protease Inhibitor Cocktail (Roche) and PhosSTOP TM Phosphatase inhibitor (Roche)) on ice for 30 min. Samples were centrifuged at 5000 × g for 1 min. The resultant supernatant was then incubated with antibody-conjugated Dynabeads TM Protein G for 2 h at 4°C. Beads were washed 3× with 0.2% TX-100-PBS and proteins eluted in SDS sample buffer and analysed by GeLC-MS/MS.
GeLC-MS/MS and data analysis. Proteomic experiments were performed in two independent biological replicates (with technical duplicates) using GeLC-MS/MS for each sample (sMB-Rs, exosomes, sMV-LD) as described previously 18 . Raw data was processed using Proteome Discoverer (v2.1, Thermo Fischer Scientific) and searched with Mascot (Matrix Science, London, UK; v2.5), Sequest (Thermo Fisher Scientific, San Jose, CA, v1.4.0.288), and X! Tandem (v2010.12.01.1) against the UniProt Human database comprising 71,785 entries. Data was searched with a parent tolerance of 10 ppm, fragment tolerance of 0.5 Da and minimum peptide length 7, with FDR 1% at the peptide and protein levels. Peptide spectral matches were validated using Percolator based on q-values at a 1% false discovery rate (FDR) 52 . Scaffold Q+/Q + S (Proteome Software Inc., Portland, OR, v4.8.7) was employed to validate MS/MS-based peptide and protein identifications from database searching 53 . Initial peptide identifications were accepted if they could be established at greater than 95% probability as specified by the Peptide Prophet algorithm. Protein identifications were accepted, if they reached greater than 99% probability and contained at least two identified unique peptides. Protein probabilities were assigned using Protein Prophet 54 . The relative abundance of a protein within a sample was determined using normalised spectral count (Nsc) 24 .
GO enrichment analysis of proteins was conducted using DAVID Bioinformatics Resources 6.8 (https://david.ncifcrf.gov/). KEGG pathway analysis was conducted as previously described 50 . Protein-protein interaction networks were generated using STRING (http://string-db.org) 55 . Heatmaps were generated using R-package software.
sMB-R uptake and KRAS G12V transfer assay. NIH3T3 fibroblasts were grown on Nunc® Lab-Tek® Chamber Slide™ system to 70% confluency. The medium was supplemented with sMB-R (5 µg), exosomes (5 µg) or PBS vehicle and cells further cultured at 37°C for 2 h to allow uptake. The ratio of sMB-R particles to cells are 50:1 (Fig. 4d). Cells were then subjected to immunofluorescence microscopy analysis using anti-MKLP1 or anti-KRAS G12V mutant-specific antibodies.
Soft-agar colony formation assay. NIH3T3 fibroblasts (20,000 cells) in 100 µl DMEM (1% Pen/Strep) were stimulated with 20 µg of SW620-sMB-R or 20 µl PBS vehicle for 2 h at 37°C. The ratio of sMB-R particles to cells are 250:1 (Fig. 5b). Where indicated, experiments were performed in the presence of 10 nM selumetinib or DMSO vehicle. Fibroblasts were then mixed with 300 µl 0.3% agarose (in DMEM with 10% FBS, 1% Pen/Strep) that was pre-warmed to 40°C in a water bath. The mixture was overlaid onto wells of a 24-well plate pre-coated with 300 µl 0.6% agarose (in DMEM with 10% FBS, 1% Pen/Strep). The mixture was allowed to solidify at 37°C for 15 min. The wells were then gently overlaid with 500 µl DMEM (5% FBS, 1% Pen/Strep) supplemented with 10 nM selumetinib (AZD6244) or DMSO vehicle and maintained at 37°C for 10 days. Culture medium was replaced every 2 days. Colonies were imaged using Zeiss AxioObserver Z1 microscope (Zeiss) under bright-field.
Transwell-Matrigel TM invasion assay. Transwell-Matrigel TM invasion assay was performed as previously described 47 . Briefly, Transwell inserts (8 µm pore size, Corning) were coated with 100 µl of 1 mg ml −1 growth factor reduced Matrigel TM and allowed to polymerise for 4 h at 37°C. NIH3T3 fibroblasts (50,000 cells) in DMEM (1% Pen/Strep) were incubated with either sMB-R (30 µg ml −1 ) or PBS alone for 2 h at 37°C. The ratio of sMB-R particles to cells are 75:1 (Fig. 5a, c, d). Where indicated, 400 nm Polystyrene latex beads (Thermo Fisher Scientific) were incubated with cells at 75:1. Cells were then carefully overlaid onto Matrigel TMcoated inserts. The inserts were placed into wells of 24-well plate companion plate (Corning) that contained DMEM (5% FCS, 1% Pen/Strep) supplemented with either sMB-R (30 µg ml −1 ) or PBS alone. Invasion chambers were incubated overnight (~16 h) at 37°C to facilitate invasion. Experiments were performed in the presence of 10 nM selumetinib or DMSO vehicle, as indicated. Inserts were washed, cells fixed (4% (v/v) formaldehyde, 5 min), and nuclei stained with Hoechst stain (10 µg ml −1 ) for 20 min. Non-invading cells were removed from the upper side of the inserts using cotton swab. Nuclei of fibroblasts that invaded to the lower side of the insert were imaged using Zeiss AxioObserver Z1 microscope. Centre of the membrane was imaged for each inset. Images were quantified using Image J software v1.49e.
Generation of SW620-GAP-GFP cells. pE-Growth-associated protein (GAP) (1-20 a.a., MLCCMRRTKQVEKNDEDQKI)-GFP plasmid was transfected using Lipofectamine TM 2000 (Invitrogen) into SW620 cells that were seeded to 70% confluency in 6-well plate. Briefly, 10 µg of plasmid was mixed with 10 µl of Lipofectamine TM 2000 in 500 µl RPMI medium at room temperature for 20 min. Cells were washed and overlaid with 1.5 mL RPMI (10% FBS). Plasmid-Lipofectamine TM mixture was then overlaid onto the cells. Cells were incubated at 37°C with 10% CO 2 . Cells with stable expression of GAP-GFP fusion proteins were selected following multiple rounds of single cell cloning into wells of 96-well plate. Expression of the GAP-GFP fusion protein in the expanded colonies was monitored using a Zeiss AxioObserver Z1 microscope and analysed by BD FACSCanto II HTS (BD Biosciences) using FlowJo software (TreeStar).
Statistics and reproducibility. Quantitative data represented as mean ± standard error of mean (s.e.m.). Statistical analyses were performed using GraphPad Prism software (one-way ANOVA (Turkey test)) with P < 0.05 considered as statistically significant. No method of randomisation was used. Investigators were not blinded to allocation during experiments or outcome assessment.
For proteomics analysis, experiments were performed in biological duplicate (with technical duplicates). Data were searched with a parent tolerance of 10 ppm, fragment tolerance of 0.5 Da and minimum peptide length 7, with FDR 1% at the peptide and protein levels. Peptide spectral matches were validated using Percolator based on q-values at a 1% false discovery rate (FDR) 52 . Scaffold (Proteome Software Inc., Portland, OR, v 4.3.4) was employed to validate MS/MS-based peptide and protein identifications from database searching 54 . Initial peptide identifications were accepted if they could be established at greater than 95% probability as specified by Peptide Prophet 53 . Protein identifications were accepted, if they reached greater than 99% probability and contained at least two identified unique peptides. Protein probabilities were assigned by Protein Prophet 54 .
Fluorescence microscopy analysis of MB-R shedding was performed ≥3× for SW620 cells and 2× for SW480 cells (Fig. 1a and Supplementary Fig. 3). Fluorescence microscopy analysis of MB-R shedding by SW620, SW480, LIM1215 (Fig. 1c) and mouse intestinal organoids ( Supplementary Fig. 8) was performed 3×. Shedding of MB-R in vivo was analysed in 4 SW620-GAP-GFP tumour xenografts (Fig. 1d). Cryo-EM imaging in Fig. 2e was performed ≥3× for Exos, 3× for sMV-LD and 2× for sMB-R with similar results. Numbers of EVs counted from Cryo-EM analysis in Fig. 2f were 60 for exos, 50 for sMV-LD and 30 for sMB-R. NTA analysis of three EV subtypes in Supplementary Fig. 13 was performed ≥3×. Immunofluorescence detection of sMB-R on aldehyde sulphate latex beads (Fig. 2g) was performed 2×. Co-IP experiment was performed 2×. Relative protein abundance of three EV subtype was performed in duplicate from five different cell lines (Fig. 2h).
Western blot-based validation of specific proteins for each EV-subtype was performed 2× for Fig. 2d, Supplementary Figs. 9, 12, and 15. Western blot analysis for Fig. S10 was performed 1×. Isolation of three EV subtypes from CM was performed ≥3× for SW620, SW480 and LIM1863 cells and 2× for MBA MB 231 and U87 cells. Western blot analysis of MB-R shedding into CM by multiple cell types in Fig. S16 was performed 1×. Western blot analysis of Supplementary Fig. 10 was performed 1×.
For Western blot, antibodies were validated as noted on manufacturer's website. For immuno-fluorescence, antibodies were validated as noted on manufacturer's website. Uncropped/entire Western blot images are provided in Supplementary  Fig. 22.
Reporting summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
Raw mass spectrometry data is deposited in the PeptideAtlas: #PASS01206 or can be accessed at http://www.peptideatlas.org/PASS/PASS01206 Source data for Figs. 2f, i, 4b, 5a-d, S7 and S20b have been provided as Supplementary Data 9.
Any remaining information can be obtained from the corresponding author upon reasonable request.