Transcriptomic analysis and fusion gene identifications of midbody remnants released from colorectal cancer cells reveals they are molecularly distinct from exosomes and microparticles

Previously, we reported that human primary (SW480) and metastatic (SW620) colorectal (CRC) cells release three classes of membrane‐encapsulated extracellular vesicles (EVs); midbody remnants (MBRs), exosomes (Exos), and microparticles (MPs). We reported that MBRs were molecularly distinct at the protein level. To gain further biochemical insights into MBRs, Exos, and MPs and their emerging role in CRC, we performed, and report here, for the first time, a comprehensive transcriptome and long noncoding RNA sequencing analysis and fusion gene identification of these three EV classes using the next‐generation RNA sequencing technique. Differential transcript expression analysis revealed that MBRs have a distinct transcriptomic profile compared to Exos and MPs with a high enrichment of mitochondrial transcripts lncRNA/pseudogene transcripts that are predicted to bind to ribonucleoprotein complexes, spliceosome, and RNA/stress granule proteins. A salient finding from this study is a high enrichment of several fusion genes in MBRs compared to Exos, MPs, and cell lysates from their parental cells such as MSH2 (gene encoded DNA mismatch repair protein MSH2). This suggests potential EV‐liquid biopsy targets for cancer detection. Importantly, the expression of cancer progression‐related transcripts found in EV classes derived from SW480 (EGFR) and SW620 (MET and MACCA1) cell lines reflects their parental cell types. Our study is the report of RNA and fusion gene compositions within MBRs (including Exos and MPs) that could have an impact on EV functionality in cancer progression and detection using EV‐based RNA/ fusion gene candidates for cancer biomarkers.


INTRODUCTION
Extracellular vesicles (EVs) are a class of secreted membraneencapsulated organelles that play a pivotal role in cell-cell communication in normal and pathophysiological processes [1,2].These secretory organelles are evolutionary conserved and contain proteins, lipids, metabolites, genetic material such as DNA fragments and a broad spectrum of RNA species including protein-coding (mRNA), non-protein coding RNAs, micro-RNAs, and pseudogene transcripts [1,3].Because RNA species are diverse, they can regulate cellular activities at many levels including RNA metabolism, RNA transcription, and protein translation [4].EV-derived RNAs have been shown to horizontally transfer to recipient cells to mediate biochemical and phenotypic changes in recipient cells [5][6][7][8].For these reasons, including their resistance to degradation [9], EV-encapsulated RNAs have become attractive candidates for potential body-fluid-based diagnostic disease biomarkers [10].Importantly, RNA profiles of circulating EVs in human patients have been shown to reflect (mirror) the RNA profile of disease biopsy material [11,12].
Over the past two decades much effort in the EV field has been directed towards understanding what subset of cytoplasmic proteins selectively traffic to different EV populations, especially exosomes, in the context of different biological systems, and how EV cargo impacts on recipient cell functionality [17][18][19].This information has been catalogued in several publicly available databases [20][21][22].Recently, we reported a comprehensive comparative label-free MS/MS-based protein analysis of three distinct EV populations -Exos, MPs, and MBRsreleased from human primary and metastatic colorectal cancer (CRC) cells using a combination of differential ultracentrifugation and isopycnic iodixanol density centrifugation [23].The three EV populations were sequentially purified in milligram amounts from the culture media of isogenic CRC cells SW480 (surrogate of CRC adenocarcinoma) and SW620 cells (surrogate of lymph node-metastatic CRC cancer) [23].
We showed that Exos, MPs, and MBRs have distinct protein signatures and highlighted the diagnostic potential of these distinct EVs types, especially MBRs, for clinical utility [23].Because it is generally accepted in the disease biomarker field that multi-marker analysis increases the reliability and sensitivity of a disease diagnostic [24], we undertook, a comparative transcript profiling analysis of the same Exos, MPs, and MBRs sample preparations used for our protein profiling [23].We report here the identity of RNA transcripts (protein-coding, lncRNA, pseudogene transcripts) and fusion genes selectively enriched in each of the three EV classes secreted from isogenic SW480 and SW620 human colorectal cancer cell lines.

Large-scale purification of extracellular vesicles
Extracellular vesicle subtypes (Exos, MPs, and MBRs) for this transcriptome study were taken from same EV preparation used for our previously published EV proteomic study [23].This approach was designed to allow us to define proteogenomic relationships for these three EV subtypes.Briefly, SW480 and SW620 cells (3 × 10 7 cells) were transferred to CeLLine AD-1000 Bioreactor classic flasks (Integra Biosciences) and grown in continuous culture with 1% Insulin-Transferrin-Selenium (ITS) (Gibco) and 1% Penicilin/Streptomycin (Gibco) [23,25].
Cell culture media (CM, 30 mL was harvested each day, 6 days for each biological replicate).Cell culture medium was immediately centrifuged at low speed to remove floating cells, cell debris (500 × g, 10 min; and 2000 × g, 10 min), and the supernatant stored at −20 • C prior to EV subtype purification.For each biological replicate (a total of three biological replicates), large-scale purification of Exos, MPs, and MBRs were performed in tandem from 180 mL CM, as described elsewhere [23].Briefly, SW480 and SW620 culture media were sequentially centrifuged at 500 × g for 5 min, 2000 × g for 10 min, and 10,000 × g for 30 min at 4 • C. The 10K pellet was subjected to isopycnic iodixanol (OptiPrep) gradient centrifugation centrifuged at 100,000 × g for 18 h at 4 • C to separate MPs (low buoyant density, fraction #7, 1.10 g/mL) from MBRs (high buoyant density, fraction #9, 1.14-1.15g/mL).MPs and MBRs were recovered from fractions #7 and #9, respectively, by centrifugation at 100,000 × g for 18 h at 4 • C, and the pellets resuspended in PBS (500 µL) and then re-centrifuged at 10,000 × g, 4 • C for 30 min.The pellets were resuspended in PBS (150 µL) for RNA isolation and sequencing.The 10K supernatant was centrifuged at 100,000 × g, 4 • C for 1 h to harvest crude Exos.The crude Exos pellet was resuspended in 500 µL PBS and subjected to OptiPrep buoyant density gradient centrifugation as described above and purified Exos harvested (fraction #7 at a buoyant density of 1.10 g/mL).The Exo pellets were resuspended in PBS (150 µL) for RNA isolation and sequencing.Yields of EVs from 180 mL CM, measured by protein concentration, were in the range 868-987, 749-827, and 463-660 µg for Exos, MPs and MBRs, respectively [23].

Total RNA isolation and quality control
Total RNA from SW480-/ SW620 EV samples and parental cell lysates was extracted using 1 mL TRIzol reagent (Invitrogen) from each sample by incubation for 5 min at 25

cDNA library construction and strand-specific transcriptome sequencing
RNA libraries of EV subpopulations were constructed using Illumina, HiSeq 3000/4000 PE Cluster Kit and sequenced by HiSeq4000 system (Illumina), according to the manufacturer's protocol.Briefly, rRNA in total RNA sample was depleted by Ribo-Zero depletion kit (Illumina).
Total RNA was fragmented and transcribed to be single strand DNA.
The second-strand cDNA was then synthesized using dUTP, dATP, dGTP, and dCTP as nucleotides.Adenine nucleotide was conjugated with first-and second-strand cDNA at 3′ end using PCR technique.
Adaptors were then conjugated at both 3′ and 5′ ends of cDNA, and cDNA was enriched by PCR technique with dTTP, dATP, dGTP, and dCTP as substrates.cDNA libraries were sequenced using the Illumina HiSeq4000 system and RNA sequence of all samples was outputted as raw FASTQ files (paired-end reads).Three biological replicates were performed for each sample.[31]).All raw RNA sequencing data can be found at NCBI database repository: GSE244455.
Previously annotated fusion genes were identified with The Cancer Genome Atlas (TCGA) RNA-Seq analysis [33].
Briefly, transcript expression was normalized using DESeq method

Transcripts uniquely found in MBRs:
To identify unique transcripts for each EV class, we focused on transcripts that were uniquely detected in both SW480-and SW620-derived EV subtypes: 509 and 894 unique transcripts were found in SW480-/SW620-MBRs, respectively (Figure 1B A comprehensive list of unique transcripts found in SW480/ SW620-MBRs, including those unique to one or the other, is listed in Table S3).S3 for full list).
Principal component analysis (PCA) demonstrated closer correlation of transcript cargo for Exos and MPs when compared to MBRs from the same parental cell type.(Figure 1D).Inspection of Figure 1E shows that the most abundant transcript biotypes in all six EV samples are protein-coding, and pseudogene transcripts (a list of transcript biotype classifications is given in Table S4).S7 and S8).Interestingly, GO terms such as cytoplasmic stress granule and transcription factor complex were enriched in MBRs compared to Exos (Figure S1A).

LncRNA and pseudogene transcripts and their association with RNA-binding proteins in MBRs
RBPs listed in the ENCORI-RBP target database (Figure S2B, Table S9).Next, we functionally annotated the 27 identified RBP partners using a g:GOst functional profiling from gProfiler [36].GO analysis revealed that the 27 RBPs involve ribonucleoprotein complex/ spliceosomal complex (HNRNP families, U2AF2), ribonucleoprotein granule (ELAVL1, FUS, IGF2BP families) and nuclear speckles (ALYREF, METTL3) (Table S10).Interestingly, 26 of the 27 RBPs were found in the same EV sample preparations analyzed by GelLC-MS/MS at the proteome level [23]; and highly-enriched lncRNA and pseudogene transcripts are predicted to bind with the RBPs shown in Table 1.
Tumor suppressor gene (MSH2) [45] was found fused to several genes (fusion genes) in all EV classes.MSH2 fusion genes such as PLAGL1-MSH2, METRNL-MSH2, and HNRPLL-MSH2 are among the top 10 most abundant fusion genes observed in all EV classes in this study, but, surprisingly, not detectable in corresponding cell lysates (Figure 4B and 4C).MSH2 fusion gene transcripts are most abundant in MBRs (relative to Exos and MPs) (Figure 4D).

Highly-enriched cancer-associated transcripts in SW480-/SW620-EV classes
EVs secreted from cancer cells mediate the tumor microenvironment to initiate and maintain hallmarks of cancer such as sustaining cell division and growth, evading immune cells, resisting cell death, reprograming neighboring cells, and acquiring genome stability [47].
Because EV classes in this study were secreted from human CRC SW480 cells (surrogate adenocarcinoma) and SW620 cells (surrogate lymph-node metastasis) we asked whether transcript profiling of SW480/ SW620 cell EVs might yield insights into cancer progression.

DISCUSSION
This study provides an in-depth transcriptomic analysis of EV classes (Exos, MPs, and MBRs) derived from the isogenic human colorectal cancer SW480/SW620 cell line model (SW480 cells, surrogate of adenocarcinoma, and SW620 cells, surrogate of CRC metastasis) [48].Previously, we examined the Exos miRNA signatures from the isogenic human colorectal cancer cell lines SW480 (adenocarcinoma) and SW620 (lymph-node metastatic cancer) to gain insights into role of miRNAs in colon cancer progression [49].In one of our previous papers we reported comparative RNA profiling of miRNAs, lncRNAs, and transcriptomic profiling of classic mRNA, pseudogenes and fusion genes in Exos and MPs secreted from the human colorectal cancer LIM1863 [11,50]; this study indicated that a variety of RNAs are sorted in EV classes.
During late stage cytokinesis, midbody remnants (MBRs), membrane-encapsulated organelles, are disassembled and released into the extracellular space [51,52], suggesting that they can be detected from biofluids or culture media.Previously, we demonstrated that MBRs can be isolated and purified from culture media of colorectal cancer cell lines using a combination of ultracentrifugation and iodixanol density separation [15] and have distinct proteome profiles from Exos and MPs [23].Previous studies have demonstrated that MBRs contain RNAs that are involved in localized translation and abscission [53,54].In our present study, we questioned whether MBRs transcriptome profiles are also distinct from Exos and MPs.To address the question, we isolated total RNA from MBRs, MPs, and Exos from SW480 and SW620 cell lines.For the first time, we sequenced the total RNA and performed differential transcript expression analysis between MBRs compared to other two EV classes.
Importantly, "cytoplasmic stress granule" and "transcription factor complex" GO terms are significantly enriched in MBRs compared to Exos (Figure S1A).Enrichment of mitochondrial and ribonucleoprotein transcripts in MBRs (relative to Exos and MPs) and high abundance of mitochondrial/RNA-binding proteins was also detected in MBRs from previous proteomic studies [59,60], including our proteomic study [23], and poses the question whether mitochondria and ribonucleoprotein complexes might have co-purified along with MBRs.Previous studies, reported elsewhere, have shown that mitochondria and ribonucleoprotein complexes/RNA granules can be released into the extracellular space either in a free-form [61,62] or embedded within EVs [63,64].Mitochondria are known to sediment at 7000-12,000 × g (10 min) [65,66] and 18,000 × g (15 min) for RNA granules [67].Interestingly, MBRs not only contain ribonucleoprotein-complex transcripts but also their ribonucleoprotein complexes protein partners such as splicing factors, and translation initiation factors [23].This observation suggests a tight relationship of RNA and their cognate protein complexes in MBRs.Following this observation, we next questioned whether enriched transcripts in MBRs might bind to RBPs.Target RBPs for the highly-enriched transcripts in MBRs were identified ( 27RBPs in total) in the ENCORI database [37].Functional annotation of these 27 RBPs suggested GO terms such as "ribonucleoprotein complex"/ "spliceosomal complex" (HNRNP families, U2AF2), "ribonucleoprotein granule" (ELAVL1, FUS, IGF2BP families) and "nuclear speckles" (ALYREF, METTL3) (Table S10).Intriguingly, 26/27 putative RBPs (Table 1) are found in our proteome analysis [23].For example, NEAT1-202 which is associated with RNA retention and RNA granules [68] is predicted to bind to RNA/stress granule protein markers (FUS, TARDBP, IGF2BP2, TAF15, ELAVL1) and ribonucleoproteins (HNRNP families) (Figure 3B) which is consistent with Chen, et al. [41].The co-existence of these RBPs and their cognate transcripts suggests they might occur as pre-formed complexes in MBRs.It is interesting to speculate that pre-formed RNA/protein complexes in MBRs might be implicated in RNA sorting [69,70].Whilst this concept has been proposed as an RNA-sorting mechanism in Exos [69,70], it has not been previously reported for larger EVs such as MBRs.A recent midbody study shows MBR (MBRs) are an active, large translating extracellular vesicle with RNA cargo with high abundance of ribonucleoproteins [51] that could determine cell fate and proliferation [16,71,72].Both our proteomic study of MBRs [23] and our transcriptomic study, presented here, also suggest that MBRs are involved in RNA translation process because: (1) proteins related to translation initiation process, spliceosome and ribonucleoproteins are highly abundant in MBRs [23], (2) protein-coding transcripts that are related to cell fate and cell proliferation such as TGFB1 [73], SOX12 [74] are highly enriched in MBRs.
EV-containing fusion genes are gaining much attention as potential targets in biomarker discovery [75].In our present study, 770 fusion genes were detected across all samples (cell lysate, Exos, MPs, and MBRs).To our knowledge this is the first report in the literature demonstrating the presence of fusion genes in MBRs and at much higher levels compared to Exos, MPs and cell lysates (Figure 4A).A salient finding was the detection of the tumor suppression gene MSH2 linked to many other genes.The fusion gene analysis identified 33 novel MSH2 fusion genes in EV subtypes Exos, MPs, MBRs, but not in SW480/SW620 cell lysates (Figure 4D).MSH2 is a key mammalian DNA mismatch repair (MMR) gene [76], and mutations or deficiencies in mammalian MSH2 gene result in microsatellite instability (MSI + ) [77].SW480 and SW620 cells have been characterized as microsatellite stable (MSS) [78], and it is interesting that most MSH2 fusion genes were not detected in SW480-and SW620-cell lysate, presumably, due to their low levels apart from DNAJC1-MSH2 which was detected in one biological replicate of SW480 cell lysate, Table S11).It is not clear why (or how) low-abundance MSH-2 fusion genes, presumably at almost undetectable levels in the cytoplasm, selectively traffic to EVs where they are found at abundant levels.If MSH-2 fusion genes are considered to be toxic to genome stability, their selective enrichment in EVs might offer a mechanism for their removal from the cell.Clearly, this hypothesis warrants further experimentation.
In this study we found that our transcript data (for example, EGFR) correlates with enriched levels of EGFR protein found in SW480-Exos, SW480-MPs, and SW480-MBRs [23].SW620-EVs were enriched in lncRNA (MALAT1-202 [85]), transcription factors (SOX2-201 [86]) and signaling transcript molecules such as metastasis-associated in colon cancer 1/ MET transcriptional regulator (MACC1-202 [87]).This finding is consistent with our observation that the MACC1 protein is also enriched in SW620-Exos, -MPs, and -MBRs [23] compared to the same EV classes derived from SW480.Reactome pathway analysis demonstrated that SW620-EVs are enriched in pathways related to translation processes and signaling pathways -for example, "Oncogenic MAPK pathway" (Figure 5C) which is consistent with previously reported transcriptome analyses of primary colorectal adenocarcinomas and metastatic colorectal cancer [88] where metastatic cancers have been shown to display aberrant RNA translation processes [89,90].
In contrast to other published EV studies [91][92][93][94][95][96], which show MPs co-purifying with MBRs (and Exos), in our present study we used a sequential centrifugation isolation strategy that resulted in highlyenriched MBR, Exos, and to a lesser extent MP, sample preparations.
Our previous proteome comparative analysis of MPs from the human isogenic colorectal cancer cell lines SW480 and SW620 did not reveal any statistically-significant proteins [23], relative to MBRs and Exos, that could be used as potential markers for this class of EVs.Interestingly, with our present Exos/MP/MBR proteogenomic comparison strategy, while we observed previously reported MP markers (RAC-GAP1 [96], KIF23 [96], ANXA1 [95], ANXA5 [93,94], EMMPRIN [92], ARF6 [91]), they were not significantly-enriched in MPs when compared to MBRs (centralspindlin complex markers, RACGAP1, KIF23) or Exos (ANXA1, ANXA5, EMMPRIN, ARF6).Because our study did not reveal distinct proteomic and transcriptomic profiles of MPs (relative to MBRs and Exos), it raises the question as to whether MPs are indeed a distinct class of EVs, or possibly a purification artifact (commixture of MBRs and Exos?).Intriguingly, MPs might be a minor subset of MBRs, arising from high-speed centrifugation.Further experiments are warranted to address this polemic.

CONCLUSION
The transcriptome of MBRs is dissimilar to Exos and MPs with high enrichment of mitochondrial transcripts, and lncRNA/pseudogene

[ 34 ]
. To identify significantly enriched transcripts in each EV subtype, we combined normalized transcript expression from the same EV subtype (from 2 cell lines) as one and we performed Wald tests in three specific EV class comparisons: (1) Exos versus MPs, (2) Exos versus MBRs, and (3) MPs versus MBRs.For cancer-progression-related transcript identification, combined normalized transcript expression from EV classes derived from SW480 cells were compared with combined normalized transcript expression from EV classes derived from SW620 cells.Log 2 fold change and pvalue were used to identified highly-enriched transcripts (Log 2 fold change ← 1 or > 1, p-value < 0.05).

1
Transcriptomic analysis pipeline and transcript profiling of Exos, MPs, and MBRs derived from SW480 and SW620 cells.(A) RNA sequencing clean reads from Exos, MPs, and MBRs were aligned against Human GRCh38 index and gene annotation (Human ENSEMBL GRCh38, v.96) using HISAT2 alignment software which identified 16,347 transcripts (FPKM > 1.5) across all samples (Exos, MPs, and MBRs).Differential transcript expression was performed using DESeq2 package with cutoff Log2 fold change > 1 or ← 1 and p value < 0.05 for highly-enriched transcript in each comparison (see section 2.6).ENSEMBL human gene annotation (GTF) and featureCounts were used to determine read counts and annotate each transcript into four subclasses: protein-coding, long non-coding, pseudogene and other (e.g., short non-coding).Fusion genes were identified using ChimeraScan v.0.4.5.and those with total fragment > 10 were considered for further analysis (FPKM = Fragments Per Kilobase of transcript per Million mapped reads).(B) A three-way Venn diagram of transcripts identified in SW480-Exos, -MPs, and -MBRs (C) A three-way Venn diagram of transcripts identified in SW620-Exos, -MPs, and -MBRs.(D) Principal component analysis (PCA) of total transcriptome of Exos, MPs, and MBRs derived from SW480 and SW620 cell lines.(E) Number of protein-coding, long non-coding, pseudogene, and other transcripts identified in Exos, MPs, and MBRs derived from SW480 and SW620 (FPKM > 1.5 in three biological replicates).

F I G U R E 2
Differential transcript expression analysis of MBRs compared to Exos and MPs derived from SW480 and SW620 cells (combined).(A) A two-way Venn diagram of highly enriched transcripts in MBRs compared to Exos and MPs.(B) Differential transcript expression analysis revealed 2276 transcripts highly-enriched in MBRs compared with Exos (log2 fold change > 1, p value < 0.05) and (C) 1541 transcripts highly-enriched in MBRs compared with MPs (log2 foldchange > 1, p value < 0.05).(G) Heat map illustration of highly-enriched transcripts in SW480-/SW620-derived Exos, MPs, and MBRs (scale shown is normalized counts subtracted by mean and divided by standard deviation).

F I G U R E 3
LncRNA and pseudogene transcripts and their association with RNA-binding proteins (RBPs) in Exos, MPs, and MBRs derived from SW480 and SW620 cells (combined).(A) Heat map illustration of highly-enriched lncRNA and pseudogene transcripts in Exos, MPs, and MBRs (scale shown is normalized counts subtracted by mean and divided by standard deviation).(B) Binary interaction networks of RBPs (from our previously publish proteomic study[1]) and three highly-enriched lncRNA transcripts identified in MBRs (NEAT1-202, KCNQ1OT1-201, and GABPB1-AS1-202).F I G U R E 4 Identification of fusion genes in Exos, MPs, MBRs, and cell lysate derived from SW480 and SW620 cells.(A) Enrichment level of fusion genes of Exos, MPs, MBRs, and cell lysate derived from SW480 and SW620.(B) A four-way Venn diagram of fusion genes identified in SW480-Exos, MPs, -MBRs, and -cell lysate (top 10 fusion genes shown in boxes, fusion genes highlighted in red are MSH2 fusion genes).(C) A four-way Venn diagram of fusion genes identified in SW620-Exos, -MPs, -MBRs, and -cell lysate (top 10 fusion genes shown in boxes, fusion genes highlighted in red are MSH2 fusion genes).(D) Heat map illustration of MSH2 (tumor suppressor) fusion genes in SW480-/SW620-derived Exos, MPs, MBRs, and cell lysate (scale shown is average total fragment of each fusion gene in log2).

F I G U R E 5
Identification of cancer progression-related transcripts and Reactome pathway analysis in EVs derived from SW480 and SW620 cells.(A) Differential transcript expression analysis of highly-enriched (log2 fold change ← 1, p value < 0.05) cancer-associated transcripts in SW480-EVs (1759 transcripts) and SW620-EVs (3923 transcripts).(B) Heat map illustration of cancer-associated transcripts in SW480-/SW620-derived Exos, MPs and MBRs (scale shown is normalized counts subtracted by mean and divided by standard deviation).(C) Reactome pathway analysis of highly-enriched transcripts in SW480-EVs (n = 1759) and SW620-EVs (n = 3923) (p value < 0.05).
transcripts that are predicted to bind to ribonucleoprotein complexes, spliceosome and RNA/stress granule proteins.The transcriptome of MBRs is high enriched with a variety of fusion genes, some of which have not been, hitherto, described in the literature (CDK6-ATIP1B2, ADD3-PARN).Exos contain transcripts related to release and biogenesis of Exos such as BICD2-201 and CAV2-201.Unlike the other two EV classes, MPs do not contain a distinct transcript signature.Importantly, the expression of cancer progression-related transcripts found in EV classes derived from SW480 and SW620 cell lines reflects their parental cell types and positively correlate with protein abundance levels[23].This study provides, for the first time, a comprehensive transcriptomic analysis of three different EV classes (MBRs, Exos, MPs) released from the human CRC cells SW480 and SW620.In summary, our findings provide a better understanding of RNA and fusion gene compositions in EV classes and how they might impact on EV functionality in cancer progression.Additionally, our study identifies several EV-based RNA/ fusion gene as potential candidates as colon cancer biomarkers/diagnostics.

Exos/MPs a Exos/MBRs b MPs/Exos c MPs/MBRs d MBRs/Exos e MBRs/MPs f RNA-binding proteins* lncRNA Pseudogene lncRNA Pseudogene lncRNA Pseudogene lncRNA Pseudogene lncRNA Pseudogene lncRNA Pseudogene
[23]27 RBPs co-identified in proteomic data[23]and ENCORI-RBP target database.a Highly-enriched transcripts in Exos compared to MPs. b Highly-enriched transcripts in Exos compared to MBRs.c Highly-enriched transcripts in MPs compared to Exos.d Highly-enriched transcripts in MPs compared to MBRs. e Highly-enriched transcripts in MBRs compared to Exos.f Highly-enriched transcripts in MBRs compared to MPs.S3A).The novel fusion genes CDK6-ATIP1B2 and ADD3-PARN are uniquely detected in SW480-EVs (but not in SW480 cell lysate) with selective enrichment in SW480-MBRs (Figure