Comparative proteomic analysis of three major extracellular vesicle classes secreted from human primary and metastatic colorectal cancer cells: Exosomes, microparticles, and shed midbody remnants

Cell‐derived extracellular vesicles (EVs) are evolutionary‐conserved secretory organelles that, based on their molecular composition, are important intercellular signaling regulators. At least three classes of circulating EVs are known based on mechanism of biogenesis: exosomes (sEVs/Exos), microparticles (lEVs/MPs), and shed midbody remnants (lEVs/sMB‐Rs). sEVs/Exos are of endosomal pathway origin, microparticles (lEVs/MPs) from plasma membrane blebbing and shed midbody remnants (lEVs/sMB‐Rs) arise from symmetric cytokinetic abscission. Here, we isolate sEVs/Exos, lEVs/MPs, and lEVs/sMB‐Rs secreted from human isogenic primary (SW480) and metastatic (SW620) colorectal cancer (CRC) cell lines in milligram quantities for label‐free MS/MS‐based proteomic profiling. Purified EVs revealed selective composition packaging of exosomal protein markers in SW480/SW620‐sEVs/Exos, metabolic enzymes in SW480/SW620‐lEVs/MPs, while centralspindlin complex proteins, nucleoproteins, splicing factors, RNA granule proteins, translation‐initiation factors, and mitochondrial proteins selectively traffic to SW480/SW620‐ lEVs/sMB‐Rs. Collectively, we identify 39 human cancer‐associated genes in EVs; 17 associated with SW480‐EVs, 22 with SW620‐EVs. We highlight oncogenic receptors/transporters selectively enriched in sEVs/Exos (EGFR/FAS in SW480‐sEVs/Exos and MET, TGFBR2, ABCB1 in SW620‐sEVs/Exos). Interestingly, MDK, STAT1, and TGM2 are selectively enriched in SW480‐lEVs/sMB‐Rs, and ADAM15 to SW620‐lEVs/sMB‐Rs. Our study reveals sEVs/Exos, lEVs/MPs, and lEVs/sMB‐Rs have distinct protein signatures that open potential diagnostic avenues of distinct types of EVs for clinical utility.


INTRODUCTION
Secretion and exchange of extracellular vesicles (EVs) by most cell types is a central paradigm for intercellular communication [1][2][3].EVs play a critical role in normal physiology and pathophysiology such as cancer, neurodegenerative disorders, and infectious diseases.These secretory organelles are evolutionary conserved with the capacity to modulate recipient cell phenotype/function by horizontal transfer of intrinsic cargo constituents such as oncogenic proteins, cytokines, infectious proteins (amyloid-β proteins, prions, malarial proteins), RNA species (miRNAs, mRNAs, lncRNA), lipids, and metabolites [4].The utility of EV-based derived biomarkers have gained significant attention in recent years and there are several notable examples of the translatability and diagnostic potential of these markers, demonstrating promise for clinical utility [1,[5][6][7][8][9].
MPs are formed by outward blebbing of the cell plasma membrane [1].In contrast to exosomes, MPs are more ellipsoid in shape and heterogeneous with respect to size (50 to ∼2000 nm diameter) [1], density 1.08-1.14g/mL; to date, no stereotypic protein markers have been described for MPs.
While much is known about exosomes and MPs, little is known about recently reported shed midbody remnants (sMB-Rs) [14,24].
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.For decades the prevailing view was that cytokinetic abscission occurred asymmetrically and that the MB remnant was inherited by one daughter cell, where it persists as a midbody remnant (MB-R) and is subsequently engulfed by autophagy whereupon it is degraded intracellularly [24,25].Accumulating evidence now shows that MB-Rs can also be released into the extracellular space (shed MB-Rs,

Statement of significance
Extracellular vesicles (EVs), a class of secreted membraneencapsulated organelles that play a crucial role in intercellular communication, can be classified into two broad groups-small EVs and large EVs.Exosomes are subtype of small EVs (sEVs/Exos) and microparticles (MPs) are a subtype of large EVs (lEVs/MPs).Recently, our group discovered a novel class of EV termed shed midbody remnants (sMB-Rs), a subtype of large EVs (lEVs/sMB-Rs).Here, we report the large-scale purification and biochemical characterization of sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs released from the primary (SW480) and metastatic (SW620) human isogenic colorectal cancer (CRC) cell lines.We dissect the MS-based protein profiles of these three EV types to yield insights into metastatic factors and signaling molecules fundamental to tumor progression.We show that the protein signatures of the three EV types are molecularly distinct from one another.
In contrast to sEVs/Exos and lEVs/MPs, lEVs/sMB-Rs are selectively enriched with centraspindlin (a tetrameric complex of dimeric Kinesin Family Member 23 (KIF23/MKLP1 and dimeric RACGAP1), RNA granule-associated proteins, mitochondrial proteins, spliceosome complex proteins and a variety of RNA-binding proteins and ribonucleoproteins.
A salient finding of our study was the observation that sMB-Rs; diameter 200-600 nm, density: 1.22-1.30g/mL) following symmetric cytokinetic abscission and potentially be taken up by nonsister cells [26].An unanswered question in the EV field relates to our knowledge of what subset of cytoplasmic proteins selectively traffic to the different EV classes-are they distinct?-and,if so, how they might impact on EV functionality.As a first step toward addressing this question we have undertaken, a comprehensive comparative proteomic analysis of sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs secreted from human primary and metastatic colorectal cancer (CRC) cells using mass spectrometry.Using a combination of differential ultracentrifugation and isopycnic iodixanol density centrifugation [14,[27][28][29] sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs were purified from the culture media of isogenic CRC cells SW480 (surrogate of CRC adenocarcinoma) and SW620 cells (surrogate of lymph node-metastatic CRC cancer).We report here the identity of proteins selectively enriched in each of the three EV classes secreted from SW480 and SW620 cells, including oncoproteins.

Protein quantification and western blotting
Protein quantification of EV samples, and western blot analyses were determined as described [30].For Western blot analysis membranes were probed with primary antibodies (anti-mouse ALIX,
Unassigned precursor ion charge states and singly charged species were rejected, and peptide match disabled.The isolation window was set to 1.4 Th and selected precursors fragmented by high-energy collision dissociation (HCD) with normalized collision energies of 25 with a maximum ion injection time of 110 msec.Ion target values were set to 3e6 and 1e5 for survey and MS/MS scans, respectively.Dynamic exclusion was activated for 30 s. Data was acquired using Xcalibur software v4.0 (Thermo Fisher Scientific) Raw data were pre-processed as described [33] and processed using MaxQuant [34] (v1.6.0.1) with Andromeda (v1.5.6), using a Human-only (UniProt #133,798 entries) sequence database.Data were searched as described [35] with a parent tolerance of 10 ppm, fragment tolerance of 0.5 Da and minimum peptide length 6, with false discovery rate 1% at the peptide and protein levels, tryptic digestion with up to two missed cleavages, cysteine carbamidomethylation as fixed modification, and methionine oxidation and protein N-terminal acetylation as variable modifications, and data analyzed with label-free quantitation (LFQ).The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD041529.

Differential protein enrichment analysis
LFQ intensities of peptide ions identified in sEVs/Exos, lEVs/MPs, and lEVs/sMB-R preparations were statistically analyzed using the edgeR software package [36].Briefly, LFQ intensities of each protein were normalized based on "effective library size" [36] for each sample using a trimmed mean of M-values (TMM) [37].P-values were calculated using Benjamini-Hochberg method [38] and normalized LFQ intensities were presented as log2.
For CRC cancer progression-related protein identifications, detected proteins in this study were compared with the Cancer Gene Census from COSMIC database (GRCh38, COSMIC, version 97, https://cancer.sanger.ac.uk/census#cl_search).
For protein abundance analysis, SW480-combined EVs were compared with SW620-combined EVs.Highly-enriched proteins in each comparison were identified using the criteria-log2fold change <−1 or >1 with p-value < 0.05.[41] in R.
A full list of uniquely-identified proteins found in sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs secreted from SW480 and SW620 cells is given in Tables S2, S3, and S4, respectively.

Interrogation of highly-enriched proteins in SW480-/SW620-derived sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs
To gain insights into similarities and differences between the three EV classes, a deep interrogation of the protein profiles for SW480-/SW620-derived sEVs/Exos, lEVsMPs, and lEVs/sMB-Rs was undertaken to gain insights into similarities and differences between the three EV subtypes.For this comparative analysis, datasets for exosomes from SW480 and SW620 cells were combined and, likewise, datasets for lEVs/MPs and lEVs/sMB-Rs.Highly-enriched proteins for each combined vesicle subtype were identified using the criteria: log2fold change < −1.0 or >1.0 with a p-value < 0.05.A complete list of highly-enriched protein identifications in sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs is provided in Tables S5, S6, and S7, respectively.
The number of protein identifications highly-enriched in SW480-/SW620-sEVs/Exos (data combined) were compared to corresponding SW480-/SW620-derived lEVs/MPs (data combined) and the number of proteins enriched in SW480-/SW620-derived sEVs/Exos were compared to lEVs/sMB-Rs.The criteria for proteins that preferentially traffic into one or another EV class was based on those proteins not being evident, or of very low abundance, in one EV class, but not the other two EV classes.In Figure 3A, it can be seen that there are 95 proteins highly-enriched in sEVs/Exos when compared to lEVs/MPs, and 409 proteins highly-enriched in sEVs/Exos when sEVs/Exos are compared to lEVs/sMB-Rs.When these two datasets are compared (see Venn diagram) 80 proteins selectively traffic to sEVs/Exos.A list of the 80 selectively enriched sEVs/Exos protein identifications is given in Table S5.
In the next comparison, 108 proteins were found to be highlyenriched in lEVs/MPs compared to sEVs/Exos and 319 proteins were highly enriched in lEVs/MPs compared to lEVs/sMB-Rs.When these two data sets were compared (Venn diagram, Figure 3B) only 14 proteins were found to selectively traffic to lEVs/MPs.A list of the 14 proteins selectively enriched in MPs compared to sEVs/Exos and lEVs/sMB-Rs is given in Table S6.
In a third comparative analysis, 604 proteins were found to be highly-enriched in lEVs/sMB-Rs when lEVs/sMB-Rs were compared to sEVs/Exos, and 533 when lEVs/sMB-R proteins were compared to lEVs/MPs (Figure 3C).Of these, 492 proteins were found to selectively traffic to lEVs/sMB-Rs (a list of selectively-enriched proteins in lEVs/sMB-Rs is given in Table S7).

Normalized LFQ intensity (a)
EV type

Highly-enriched cancer associated cargo proteins in SW480-/SW620-EV classes that modulate CRC progression
It is well recognized that sEVs/Exos secreted from human metastatic CRC cells harbor metastatic factors and signaling pathway components that engage in crosstalk between tumor and stromal cells in the tumor microenvironment [31].To gain insights into how the proteome of individual EV classes might impact on CRC progression, a differential protein enrichment analysis between SW480-EVs (combined sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs) and SW620-EVs (combined) was performed.This analysis revealed 230 cancer-associated proteins that are highly enriched in combined SW480-EVs (such as CD44, STAT1, MDK, TGM2, EGFR, FAS, CLDN7) and 264 cancer-associated proteins in SW620-EVs such as RICTOR, MACC1, PRKACA, TGFBR2, MET (Figure 4A).A complete list of highly-enriched protein identifications in this analysis is provided in Table S9).
Using human protein atlas cancer database resource of genes linked from COSMIC cancer database, we showed 14 commonly identified proteins in our proteomic data and COSMIC cancer database (Table S1).Of these 14 commonly identified proteins include KRAS, betacatenin (CTNNB1), Proto-oncogene SRC, and DNA mismatch repair proteins MSH2 and MSH6 (Table 2).Interestingly, the MSH proteins were detected only in lEVs/sMB-Rs (Table 2), while KRAS, CTNNB1, and SRC protein abundance is higher in sEVs/Exos and lEVs/MPs compared to lEVs/sMB-Rs (Table 2).
We next questioned whether the EV classes exhibit different functionalities in cancer progression.Proteomic analysis showed that many
In summary, proteomic analysis showed selective enrichment of cancer -associated proteins in different EV classes from primary CRC (SW480) and metastatic CRC (SW620) cell lines.

DISCUSSION
In this study, we performed a detailed comparative proteome analysis of three EV classes (sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs) secreted from isogenic human primary and metastatic CRC cells-SW480 cells (surrogate of adenocarcinoma) and SW620 (surrogate of metastatic colon cancer) [43].
sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs were enriched and purified in high yield from the same preparation of SW480/SW620 CM using a combination of differential centrifugation and isopycnic buoyant density (iodoxinol/OptiPrep) centrifugation (Figure 1A).
Interestingly, proteins purported to be involved in lEVs/MP biogenesis such as ARRDC1 [47], ARF1 [48], ARF6 [49], RHOA [50], and ANXA5 [11] are not only enriched in lEVs/MPs but also in sEVs/Exos derived from SW480 and SW620 in this current study.This paradox raises the specter of EV purity-for example, possible cross contamination with other EV types is highly probable, especially in the case of lEVs/MPs and sEVs/Exos which overlap in size distribution and have similar biophysical properties such as buoyant density.
The observed enrichment of RNA granule and mitochondrial proteins in lEVs/sMB-Rs is intriguing.While typical RNA granule and mitochondria isolation methods [51,52] are similar to the EV subtype isolation methods used in this study, TEM images of lEVs/sMB-Rs derived from both SW480 and SW620 cells (Figure 1F) did not show any evidence of intact mitochondria (see additional TEM images of lEVs/sMB-Rs in Figure S1).However, it is evident that mitochondrial bodies can be released from cells as mitovesicles and transferred to recipient cells [53,54].As these mitochondrial bodies are ∼100-200 nm (small than lEVs/MPs and lEVs/sMB-Rs) [54] and size ranges of EVs and mitovesicles overlap [55], it is possible that mitochondrial bodies/mitovesicles are co-isolated or incorporated in lEVs/MPs and lEVs/sMB-Rs.Biological experiments are needed to prove this concept.Interestingly, Skop and colleagues reported the identification of mitochondrial (26%), nuclear (16%), and ribosomal (13%) proteins in the midbody proteome [2] and RNA localization, nuclear transport, RNA splicing, mitochondrion organization rate as top fold enrichment GO annotation profiles of midbody proteome and interactome using PANTHER [56,57].Furthermore, 25/492 selectivelyenriched lEVs/sMB-R proteins identified (RACGAP1, KIF4A, KIF23, CEP55, PLK1, for example) co-identified with proteins listed in the MiCroKITS v.4.0 database [58] (a database of proteins temporally and spatially localized in distinct subcellular positions including midbody, centrosome, kinetochore, telomere, and mitotic spindle during cytokinesis (cell division/mitosis) (http://microkit.biocuckoo.org,Table S7).In another midbody remnant study enriched vesicular traffic transport and protein-translation related proteins were reported in the "Flemmingsome" (referred as "post-abscission midbody" or "midbody remnants") using STRING functional association network [59].
Cancer progression-associated proteins have been previously shown to selectively traffic to EVs [30,31,61].In our present study receptors such as CXCR4 and TGFBR2/AXL were uniquely detected in SW480-sEVs/Exos and SW620-sEVs/Exos, respectively.A possible reason for receptor enrichment in sEVs/Exos is that receptors such as EGFR, HER2, ERBB3, and ERBB4 bind to their cognate ligands and are then internalized into the intercellular early endosome which further develops to late endosome and multivesicular body (MVB), respectively [62].MVBs that contain receptor-intact ILVs (exosomes) can fuse either with lysosomes (leading to proteolytic degradation) or traffic to the plasma membrane whereupon sEVs/Exos are released into the extracellular milieu [63].
A salient finding was the detection of the DNA mismatch repair protein MSH2 and MSH6 only in sMB-Rs (Table 2).MSH2 and MSH6 form a dimeric complex [73] which is implicated in hereditary nonpolyposis CRC (HNPCC) [74].
In summary, the proteome of SW480-/SW620-lEVs/sMB-Rs is dis- DNA mismatch repair proteins (MSH2 and MSH6) that play a major role in CRC progression are uniquely detected in lEVs/sMB-Rs.The characterization of major intrinsic EV subtypes (sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs) in this study open avenues for exploring their clinical utility as protein biomarkers for diagnosis and screening of CRC.

F I G U R E 1 Figure 2
Figure 2 compares the protein profiles of sEVs/Exos, lEVs/MPs, and

F I G U R E 4
label free precursor intensity) normalized with protein length.* = Uniquely identified proteins.# = Commonly identified colorectal cancer-related proteins with COSMIC cancer database.-= Undetected in samples.Identification of cancer progression-related proteins and KEGG pathways in EVs derived from SW480 and SW620 cell lines.(A) Differential protein enrichment analysis of highly-enriched (log2 fold change > 1, p-value < 0.05) cancer-associated cargo proteins in SW480-EVs (230 proteins) and SW620-EVs (264 proteins).(B) KEGG pathway analysis (ranked by p-value) of highly-enriched cancer-associated proteins found in SW480-EVs and SW620-EVs.cancer-progression-associated proteins are highly or uniquely sorted in the separate EV classes.For instance, protein-related to genetic stability such as PARP1 was highly-enriched in SW480-lEVs/sMB-Rs and histone deacetylase (HDAC1) and its substrate MSH6 (DNA mismatch repair protein) were uniquely detected in SW480-lEVs/sMB-Rs (Table tinct from sEVs/Exos and lEVs/MPs proteomes.The lEVs/sMB-R proteome is high enriched with mitochondrial proteins (membrane proteins and enzymes), RNA granule proteins, splicing factors, ribonucleoproteins, histone subunits, translation initiation factors and integral components of midbodies.SW480/SW620 cell-derived sEVs/Exos are highly-enriched in tetraspanins/glycoproteins (TSPAN1, TSPAN14, CD63, CD81, CD82) and ESCRT components (TSG101, CHMP1A, CHMP4B).lEVs/MPs are highly-enriched in enzymes (DTYMK, IMPA1, and MRI) and membrane-associated proteins (SLC29A2, FGFR4).This study provides, for the first time, an in-depth comparative proteomic analysis of three EV classes (sEVs/Exos, lEVs/MPs, and lEVs/sMB-Rs) which were purified simultaneously from two CRC cell types (SW480 and SW620 cells).This comparative proteome study paves the way to advancing the characterization of EV classes and in doing so may impact on our understanding of intercellular communication.