Transcriptome and long noncoding RNA sequencing of three extracellular vesicle subtypes released from the human colon cancer LIM1863 cell line.

Extracellular vesicles comprise three main classes – exosomes, shed microvesicles (sMVs or microparticles) and apoptotic bodies 1 12 . Previously we reported that sMVs and two distinct populations of exosomes are released from the highly polarised LIM1863 colon carcinoma cell-derived organoids 8 ; based on their protein profiles, the two exosome subtypes are consistent with one originating from the apical surface(EpCAM-Exos), the other (A33-Exos) from the basolateral surface 8 . Because the three EV types have distinct protein profiles, based on GeLC-MS/MS 8 , and miRNA profiles 11 , based on small RNA sequencing analysis, we surmised that cellular long RNA species (mRNA and lncRNA) might also be selectively enriched in these EVs. Exosomes and sMVs were purified using sequential immunocapture 8 and consisted of vesicles ranging in size from 50–120 nm for exosomes and 50–1500 nm for sMVs 8 10 . The integrity of these EV preparations was further assessed by transmission electron microscopy and western blot analysis for the presence of exosomal (CD63, CD81, CD82, Alix, Tsg101) and sMV (Kif23) markers 10 11 . Next, we prepared cDNA libraries for large RNAs from parental LIM1863 cells (whole cell lysates, CL) and LIM1863 cell-derived sMVs and A33-/EpCAM-Exos 11 . Transcriptome data for these 4 samples (EV samples were pooled from over 400 individual culture media collections) yielded 4.58 to 6.39 G raw data (50.8 to 71 million reads) and 3.75 to 4.7 G clean data (41.7 to 52.2 million reads). Clean reads were aligned to the human genome sequence (version GRCh37.74) using TopHat2 and gene expression was profiled using Cufflinks2 13 and matched human genome annotation provided by Ensembl ( http://www.ensemble.org ). Of the 50,101 mRNA/ncRNAs (>1 fragments per kilobase of transcript per million fragments mapped (FPKM) in at least one library) identified across all samples (40,271 for CL, 29,493 for sMVs, 19,789 for A33-Exos, and 25,865 for EpCAM-Exos), see Fig. 1A , 3,940 were significantly enriched in EVs relative to CL (1,717 sMVs, 2,543 A33-Exos, 2,565 EpCAM-Exos transcripts) using the following criteria: log 2 FC (fold change) >1, p-value 0.7 ( Supplementary Dataset ). To validate the RNA-Seq results, 7 genes (BCL7C, EEF1G, RAB13, RSP3, TPT1, SCARB1, and SCD) were chosen for quantitative real-time PCR (qRT-PCR) ( Fig. 1B ). The selected mRNAs were significantly enriched, positively or negatively, in at least one comparison group. Three genes (CKS1B, GAPDH, and MFL2) present in CL and all EVs were used as internal controls for normalisation of qRT-PCR data. The results showed that expression patterns for these genes were in excellent agreement with the RNA-Seq findings. The Ensembl Automatic Gene Annotation System 14 ( http://www.ensemble.org ) and GENCODE 15 were employed to annotate the 3,940 RNA transcripts enriched in LIM1863-derived EVs, relative to the parent cell. Of these, 2,389 are protein-coding mRNAs, 1,028 lncRNAs, 206 short noncoding RNAs (sncRNAs) and 317 are pseudogene-derived transcripts ( Fig. 1C ). Interestingly, Ensembl/GENCODE analysis revealed that the 2,389 protein-coding RNAs list contains 282 and 4 mRNAs predicted to be targets of nonsense-mediated decay (NMD) 16 and non-stop decay 17 , another mRNA surveillance pathway, respectively. The distribution of the 2,389 protein-coding mRNAs in the EV subtypes revealed 631 to be common to all EVs while 577, 467, and 127 are significantly enriched, relative to CL, in A33-Exos, EpCAM-Exos and sMVs, respectively ( Fig. 1D ). In the next phase of our annotation strategy we focussed on those protein-coding mRNAs in the 2,389 dataset which were common to both Ensembl 18 19 and UniProtKB/Swiss-Prot/TrEMBL 20 21 22 databases. This annotation process identified 1,937 transcripts encoding canonical proteins, 348 transcripts encoding protein isoforms (including splice-variant proteins), and an additional 119 transcripts encoding proteins annotated in the Ensembl, but not the UniProtKB database – we refer to the latter as ‘missing’ proteins (i.e., gene-encoded proteins where there is no confirmatory protein/peptide information). Further analysis of the 1,937 canonical protein dataset revealed 1,674 protein-coding transcripts, 259 NMD transcripts and 4 predicted to be the target of non-stop-decay ( Fig. 2 ). The distribution pattern of gene-encoded canonical proteins in the EV subtypes showed 446 encoded proteins common to the three EV subtypes, 383 selectively enriched in A33-Exos relative to CL, 337 in EpCAM-Exos and 89 selectively enriched in sMVs ( Fig. 2A ). To gain insight into the function of the 446 cellular canonical mRNAs that are selectively enriched in three EV subtypes we performed a Gene Ontology (GO) analysis. Interestingly, GO terms were related to ‘translation’ (GO:0006412), ‘ribosome biogenesis’ (GO:0042254), and ‘rRNA processing’ (GO:0006364) in the biological process category ( Fig. 2D ). These significantly enriched GO terms ( p  < 0.01) common to all LIM1863 EVs indicate the possibility of a hitherto, unrecognized role of EVs in cell-cell communication, especially in protein translation-related processes such as ‘ribosome biogenesis for energy metabolism’ and ‘cellular growth’ – processes which are of central importance and considered as hallmarks of cancer 23 24 . These GO terms were also prominent in selectively-enriched canonical mRNAs in A33-Exos and EpCAM-Exos ( Fig. 2D ); in contrast, for sMVs the GO terms ‘negative regulation of cellular protein metabolic process’ (GO:0032269) and ‘negative regulation of protein modification process’ (GO:0031400) are exclusively represented in the biological processing category ( Fig. 2D ). The most significant KEGG pathways for commonly enriched mRNAs and selectively enriched mRNAs in A33-Exos and EpCAM-Exos were ‘ribosome’ (hsa03010) and ‘ribosome’ (hsa03010)/’spliceosome’ (hsa03040), respectively ( Supplementary Table S1 ). These GO/KEGG pathway analysis findings imply that the sMVs, A33-Exos and EpCAM-Exos may have different functional roles in recipient cells. An inspection of the GO terms showed that the most prominent transcripts enriched in all EV subtypes, relative to parental LIM1863 cell lysates, were ribosomal protein mRNAs encoding 37 and 52 proteins in the small (40 S) and large (60 S) ribosome subunits ( Supplementary Dataset ). Other significantly enriched transcripts common to all EV subtypes were tumour protein, translationally-controlled 1 (TPT1), the EEFs (eukaryotic translational elongation factors EEF1G, EEF1B2, EEF1D and EEF2), FTL (ferritin light polypeptide), RAB13, the transcriptional elongation factors TCEB1/TCEB2, and 13 transcripts encoding cellular proteins associated with the mitochondrial complex. We next looked at prominent canonical transcripts differentially distributed in the individual EV subtypes compared to CL. There are three protein classes potentially transcribed by these mRNAs that stand out – eukaryotic translation initiation factors (EIFs), the heterogeneous nuclear riboproteins (HNRNPs), and mitochondrial ribosomal proteins (MTRPs). For example, A33-Exos contain 6 EIFs (EIF-3E, -3L, -4A1, -2AK1, -4E, and EIF5E), one HNRNP (HNRPM), and 16 MTRPs (MRPL-21, -55, -42, -30, -49, -48, -36, -13, -48, -37, -53, -1, MRPS-28, -14, -17, -35, -18C, and -18C) not seen in sMVs and EpCAM-Exos. In contrast, EpCAM-Exos there are 6 EIFs (EIF-3F, -3E, -3K, -3M, 4EBP1, and -1), 6 HNRNPs (HNRNP-A1, -A3, -C, -PH1, and -PL), and 8 MTRPs (MRPL-33, -22, -13, -51, -18, 53, -48, and MRPS36) exclusive to this EV. In the case of sMVs enrichment of 3 EIFs (EIF-3K, -3M and -2B4), two HNRNPs (HNRNPC and HNRNPH1) and three MTRPs (MRPL-52, -21 and -12) are selectively enriched in this EV subtype. For a summary of these data, see Supplementary Dataset . We next wanted to look at potential protein isoform sequences – i.e., protein products generated by alternative splicing, alternative promoter usage and alternative translation initiation 22 . A total of 336 mRNAs encoding isoform proteins were observed to be enriched (when compared to CL) in EVs released from LIM1863 cells – of which 74 are common to all EV subtypes and 21, 80, and 82 differentially distribute in sMVs, A33-Exos and EpCAM-Exos, respectively ( Fig. 2B , Supplementary Dataset ). TPT1 and a large group of ribosomal protein transcripts (RPL12, RPL13, RPL17, RPL18, RPL28, RPL31, RPLP0, RPLP1, RPS24 and RPS29) are common to all EV subtypes, while two IL32 isoform transcripts (ENST00000440815 and ENST00000530890) are specifically enriched in sMVs, three mRNAs encoding isoforms of small EDRK-rich factors (SERF1A, SERF1B and SERF2), and 4 splice variant transcripts for transmembrane proteins (TMEM126B, TMEM134, TMEM14B and TMEM54) preferentially distribute to EpCAM-Exos. Using the DAVID Bioinformatics Resource 25 26 51 (common), 17 (sMVs), 56 (A33-Exos), and 59 (EpCAM-Exos) of these enriched splice variant transcripts were found to be recognized in the GO biological process category ( Fig. 2E , Supplementary Table S2 ). Those splice variant transcripts common to all EV subtypes are mainly involved in the biological processes of ‘translational elongation’ (GO:0006414) and ‘translation’ (GO:0006412). Interestingly, splice variant mRNAs that selectively distribute to sMVs, A33-Exos and EpCAM-Exos are found in different GO terms – for example, sMVs (‘cellular macromolecular complex assembly’ (GO:0034622) and ‘subunit organization’ (GO:0034621)), EpCAM-Exos (‘macromolecule catabolic process’ (GO:0009057) and ‘protein modification by small protein conjugation’ (GO:0032446)); no significant biological process GO terms (p-value < 0.01) were observed for isoform mRNAs enriched in A33-Exos. These observations point to different functional roles of LIM1863 EV subtypes in recipient cells. Interestingly, of the 119 EV-enriched mRNAs observed in Ensembl but ‘missed’ in UniProtKB 27 are common to all the EV subtypes 9, 27 and 27 selectively distribute to sMVs, A33-Exos and EpCAM-Exos, respectively ( Fig. 2C , Supplementary Dataset ). Significant GO Biological Process annotations were found for transcripts common to all EV subtypes and those selectively enriched in EpCAM-Exos. For example, two transcripts (ENST00000439403 and ENST00000372099, GTF3A and GTF3C5) probably function in “rRNA transcription” (GO:0009303) and “transcription from RNA polymerase III promoter” (GO:0006383), while ENST00000444743 and ENST00000433710 (TIMM23 and MIPEP) are involved in ‘protein transport and targeting in mitochondria’. Next we asked how many of these 119 missed transcripts might originate from the same genes encoding canonical/isoform proteins. This analysis revealed 63/119 transcripts not originating from these genes ( Supplementary Table S3 ) – of these the most prominent transcript (ENST00000446260) encodes an uncharacterised protein from the chromosome1 open reading frame 122 (C1orf122 gene). Needless to say, further dissection of these 63 missing transcripts not seen in UniProtKB may accelerate the completion of the Human Proteome Project 27 28 29 . Increasing evidence indicates that missense mutations and unique cancer-derived fusion transcripts are potential sources of neoantigens, which could provide a valuable source of disease biomarker and novel therapeutic approaches 30 31 . Using TopHat2 we identified 268 novel alternative splicing events ( Supplementary Table S4 ) and 33 fusion genes were identified using Software ChimeraScan version 0.4.5 ( Supplementary Table S5 ). To increase the reliability of these searches we set stringent criteria – >100 fragments in the case of alternative splicing sites and >10 fragments for fusion gene connection coverage. Although none of the identified fusion genes have been reported in CRC, three have been reported in other cancers - SH3D19/LRBA in primary myelofibrosis 32 ; RIPK2/OSGIN2 in primary urethral clear-cell adenocarcinoma 33 ; and GOLT1A/KISS1 in bladder cancer 34 . Interestingly, most of the alternative spicing events occurred in ribosomal protein and TPT1 (TCTP) genes, whose normal transcripts are also highly expressed and enriched in the EVs (see above). While we do not see any splice variant forms of ribosomal or TCTP proteins in our datasets it is interesting to speculate that the corresponding transcripts we observe in EVs may be translated upon EV uptake in recipient cells; this, in turn, may contribute to onset of apoptosis and cancer, as described elsewhere 35 36 37 . Because the role of pseudogenes in physiology and disease is gaining importance 38 39 we next asked whether there are any pseudogene transcripts in LIM1863-derived EVs. Using the same criteria for protein-coding mRNAs, a total of 317 pseudogene transcripts were found to be enriched (>100 FPKM) in EVs released by LIM1863 cells, especially in A33-Exos ( Fig. 3A , Supplementary Dataset ). Common EV-enriched pseudogene transcripts include 105 processed and one unprocessed pseudogene transcripts. Of these, the 10 most highly-expressed pseudogene transcripts are RPL41P1, RPL39P3, EEF1A1P5, CTD-2031P19.4, CTB-63M22.1, RP11-742N3.1, RP11-122C9.1, RPL9P8, RPL9P9 and RP11-466H18.1. Eleven processed pseudogene transcripts preferentially distribute to sMVs (MRPS10P1, TECRP1, ATP5EP1, RP11-158M2.6, POLR2KP1, CTD-2218G20.1, CTD-3141N22.1, RP11-486A14.1, AL049542.1 , PSMC1P5 and AC012615.1 ), 113 in A33-Exos and 31 in EpCAM-Exos. Of note, 67 of the 113 pseudogene transcripts specifically enriched in A33-Exos were ribosomal protein pseudogenes, 4 derived from eukaryotic translation elongation factor 1 alpha 1 gene, and 3 were from ferritin, heavy polypeptide 1 gene. It is now apparent that the vast majority of the genome is transcribed as non-protein-coding RNA (ncRNA) 40 and that many of these ncRNAs are of crucial importance for normal development and physiology and for disease 41 42 . In this study, we identified 206 sncRNAs and 1,028 lncRNAs preferentially distributed to LIM1863 cell-derived EVs ( Fig. 1C ). In the case of sncRNAs we see 10 cellular sncRNAs enriched in all three EV subtypes and 18, 50 and 58 that selectively distribute to sMVs, A33-Exos and EpCAM-Exos, respectively. Detailed category information (miRNA, misc_RNA, rRNA, snRNA and snoRNA) for the EV-enriched sncRNAs is given in Fig. 3B and Supplementary Dataset . The miRNAs seen in this study were either primary miRNA transcripts (pri-miRNAs) or miRNA precursors (pre-miRNAs). Four EV-enriched pre-miRNAs (MIR3661, MIR941-1, MIR1282 and a novel miRNA AC009065 ) identified in sMVs were highly expressed (>100 FPKM), of them MIR3661 (ENST00000577394) was shared with EpCAM-Exos. In A33-Exos, we found two known pre-miRNAs (H19 and MIR1182) and five novel pre-miRNAs. Interestingly, H19 encodes miR-675 which regulates tumour suppressor RB in human colorectal cancer 43 . EpCAM-Exos was identified to contain MIR671 and another 8 novel miRNAs (for a detailed list of identified pri-miRNAs, see Supplementary Dataset ). Several miscellaneous snRNAs (misc_RNA) were enriched (>100 FPKM) in LIM1863-derived exosomes ( Supplementary Dataset ), especially 7SL and Y_RNA transcripts. Interestingly, 7SL is an RNA component of the SRP (signal recognition particle), which associates with the ribosome and targets nascent proteins to the endoplasmic reticulum for secretion or membrane insertion 44 45 . Y_RNAs are small noncoding RNAs, which are components of Ro60 ribonucleoprotein particle, a target of autoimmune antibodies in patients with systemic lupus erythematosus 46 . They are also necessary for DNA replication through interactions with chromatin and initiation proteins 47 48 . In our study while we do not see canonical Y_RNA products (e.g., Y-RNAs 1,3,4, and 5) we do observe homologs of Y-RNA that bare RNA sequence similarities 49 ; the function of this category of misc_RNAs it yet to be elucidated. Ribosomal ribonucleic acid (rRNA), the RNA components of the ribosome is essential for protein synthesis in all living organisms. It constitutes the predominant material within the ribosome, which is approximately 60% rRNA and 40% protein by weight. In this study we identified 5 and 33 rRNA transcripts enriched specifically in A33-Exos and EpCAM-Exos, respectively. The number of rRNA transcripts specifically enriched in EpCAM-Exos (12) is 6 fold greater than that in A33-Exos (2) ( Fig. 3C ). It is interesting that rRNAs enriched in the LIM1863-derived exosomes and sMVs are 5 S rRNAs. The snRNA (small nuclear ribonucleic acid) class of small RNA molecules found within the splicing speckles and Cajal bodies of the cell nucleus in eukaryotic cells 50 . Also referred to as spliceosome RNAs, the snRNAs are integral components of the spliceosome, a large ribonucleoprotein (RNP) made up of over 200 different proteins and five snRNAs - U1, U2, U4, U5 and U6 50 . snRNAs, the largest group of sncRNAs identified in this study are significantly enriched in LIM1863 released EVs; 35, 23 and 49 snRNA transcripts are enriched in sMVs, A33-Exos and EpCAM-Exos, respectively ( Table 1 ). Small nucleolar RNAs (snoRNAs) are a class of snRNAs responsible for guiding a series of site-specific post-translational modifications of rRNAs, tRNAs and snRNAs 51 52 . There are two main types of snoRNAs – H/ACA box snoRNAs (direct modification of nucleoside uridine to pseudouridine) and the C/D box snoRNAs which directs methylation of nucleosides 51 . In total we identified 40 EV-enriched snoRNAs, in which 8, 17 and 11 selectively distribute to sMVs, A33-Exos and EpCAM-Exos, respectively ( Table 2 ). In addition to the above, there is a further class of transcripts referred to as lncRNAs which are operationally defined as transcripts that are >200 nt in length and lack protein coding capability 53 . lncRNAs can be roughly classified as intergenic, intragenic/intronic, and antisense-based on their position relative to protein-coding genes 41 53 . In this study we identified 1,028 lncRNAs ( Fig. 1C ) comprising the following categories – 83 antisense lncRNAs, 554 processed transcripts, 303 retained-intron transcripts and 88 lincRNAs (long intergenic non-coding RNAs). Of these, 217 cellular lncRNAs were enriched in all three EV subtypes while 131 are enriched in both exosome types and 49, 369, and 149 preferentially distribute to sMVs, A33-Exos and EpCAM-Exos, respectively ( Fig. 1D ). Abundant antisense lncRNAs selectively enriched in both exosome subtypes include RP5-940J5.9, RP11-290D2.6, as well as 7 ZFAS1 (ZNFX1 antisense RNA 1) and 5 C17orf76-AS1 (FAM211A antisense RNA 1) isoforms. Interestingly, C17orf76-AS1 (LRRC75 antisense RNA1) is up-regulated in 5-fluorocil resistant CRC cell lines and is reported to regulate apoptosis 54 . We found several highly-enriched antisense lncRNAs (Log2FC > 2) common to all three EV subtypes (e.g., ZFAS1). Antisense lncRNA AC007193.8 uniquely distributed to sMVs, RUSC1-AS1, TM4SF1-AS1, DLGAP1-AS1 and DLGAP1-AS1to A33-Exos, while SETD5-AS1, DNAJC27-AS1 and TTC28-AS1 selectively distributed to EpCAM-Exos. Of these antisense lncRNA ZFAS1, which has been reported in breast cancer tissue 55 , liver cancer 56 and CRC 57 , is thought to function as an oncogene by destabilisation of p53 and interaction through the CDK1/cyclin B1 complex leading to cell cycle progression and inhibition of apoptosis 57 . In the case of lincRNAs, several EV-enriched lincRNAs attracted our attention because of their significance in cancer progression and diagnosis, such as small nucleolar RNA host genes (SNHG5, SNGH6 and SNHG8), growth arrest specific transcript 5 (GAS5), LINC00493, TP53 target 1 (TP53TG1), MIR4435-1 host gene (MIR4435-1HG) and H19. In the case of GAS5, its down-regulation is reported to be a poor prognosticator of cancers such as breast 58 , prostate 59 , gastric 60 , lung 61 , bladder 62 , colorectal 62 and cervical 62 . GAS5 is reported to act by enhancing G1 cell cycle by regulating cyclin-dependent kinase 6 (CDK6) 63 . For a list of enriched antisense lncRNAs and lincRNAs in LIM1863-derived EVs, see Supplementary Dataset . Because ncRNAs can function as molecular scaffolds to specify higher-order organisation in ribonucleoprotein (RNP) complexes and in chromatin states 64 65 we next asked whether EV cargo includes RBPs and if any of our identified RNA transcripts could potentially bind to these RBPs (i.e., cognate binding partners). Highly-purified LIM1863-derived EVs were prepared using a combination of differential centrifugation (for sMVs) and sequential immunoaffinity capture (for A33-Exos and EpCAM-Exos), as described by Tauro et al . 8 66 . GeLC-MS/MS revealed a total of 151 RBPs and 71 ribosomal proteins ( Supplementary Table S6 ), as defined 65 67 . For subsequent analysis the ribosomal proteins were not taken into consideration. It can be seen in the Venn diagram ( Fig. 3C ) that 12, 17, and 5 cellular RBPs selectively distribute to A33-Exos, EpCAM-Exos, and sMVs, respectively. Interestingly, of these 6 members of the heterogeneous nuclear RNA-binding proteins (hnRBPs) 68 , 16 eukaryotic initiation factor (eIF) family proteins 23 , 3 RNA helicases of DEAD box family (DDX) 69 and 8 splicing factors, some of which (e.g., hnRNPA2B1) have been implicated in the sorting of RNAs into EVs 70 . Next we used StarBase (v2.0) 71 , which comprises >6000 entries and 111 CLIP-Seq experiments, to identify those RNAs and RBPs from our LIM1863 EV datasets that could potentially interact. Table 3 reveals that 7 of the 151 RBPs seen in our MS-based studies could bind to many of the RNAs we obtained by RNA-Seq. We also interrogated our data for evidence of RNP complexes of snRNAs and RBPs associated with spliceosome subunits. Our findings ( Fig. 4 ) reveal evidence for the presence of two spliceosome complexes – the U1 and U2 subunits – that are critical for pre-mRNA processing 50 . To our knowledge this is the first report of EV-associated ribonucleoprotein particles. Finally, we asked whether any of the RNA species identified in LIM1863-derived EVs might be implicated in CRC or other cancer types. As shown in Table 4 we list several EV-enriched mRNAs seen in our study (e.g., TPT1, several ribosomal protein (RP) genes, EEF1A1, EEF1B2 and FTL, and lncRNAs, such as SNHG5, SNHG6, SNHG7, SNHG8, ZFAS1, H19 and LINC00116) have also been reported to be up-regulated in tumour tissues. It is interesting to note that GAS5, which has been reported to be down-regulated in CRC tumour tissue 72 as well as HCC 73 and pancreatic cancer 74 , is highly expressed and significantly enriched in all LIM1863-derived EVs. Since GAS5 overexpression has been implicated in cell growth arrest and apoptotic induction 58 , it is speculated that its release from the cells via EVs might provide a mechanism for lowering its cellular concentration, in a manner similar to PTEN in glioblastomas 75 . We used two CRC gene expression studies (SRP022054 76 and SRP029880 77 ) from the SRA database ( https://www.ncbi.nlm.nih.gov/Traces/sra/sra.cg ) - 4 and 18 pairs of tumour/normal tissues from CRC patients, respectively - to validate the expression levels of candidate mRNAs and lncRNAs found in our study. Heat maps for these two studies ( Supplementary Fig. S1 ) revealed several RNAs species from our LIM1863 studies that are up-regulated in tumour biopsies that warrant further investigation as potential blood-based CRC biomarkers.

All methods were carried out in accordance with the approved guidelines of La Trobe Institute for Molecular Science. LIM1863 cells 78 were initially cultured in 175 cm 2 flasks (Invitrogen, Carlsbad, CA) with RPMI-1640 supplemented by 5% foetal calf serum (FCS), 0.1% insulin-transferrin-selenium (ITS, Invitrogen), 100 U/ml penicillin and 100 μg/ml streptomycin at 37 °C and 5% CO 2 . Cells (~3 × 10 7 ) were transferred into the Cultivation chamber of a CELLine CL-1000 Bioreactor classic flask (Integra Biosciences) and cultured at 37 °C and 5% CO 2 atmosphere as previously described 11 . Culture medium was replaced twice a week and the cell suspension from the Cultivation chamber was harvested every 48 hr. LIM1863 cells were cultured in serum-free medium supplemented with insulin-transferrin-selenium for 24 h, according to Ji et al . 11 . The culture medium (CM) was collected and subjected to differential centrifugation at 4 °C, first at 480  g for 5 min followed by 2,000  g for 10 min to remove intact cells and cell debris and then a final centrifugation step (10,000  g , 30 min) to isolate shed microvesicles (sMVs) 7 79 . The resulting supernatant was centrifuged at 100,000  g to harvest crude exosomes which were then fractionated into two distinct exosome subpopulations (A33-Exos and EpCAM-Exos) by sequential immunocapture using Dynabeads™ (Invitrogen) loaded with anti-A33 monoclonal antibodies 5 in tandem with anti-EpCAM (CD326)-monoclonal antibody bound magnetic microbeads (Miltenyi Biotec), as described 8 . Protein quantification based on protein staining densitometry was performed by 1D-SDS-PAGE/SYPRO ® Ruby, as previously described 79 80 . A33-Exos and EpCAM-Exos subpopulations (1 μg/10 ml PBS) were applied for 2 min to 400 mesh copper grids coated with a thin layer of carbon. Imaging was performed using a JEOL JEM-2010 transmission electron microscope operated at 80 kV, as described 8 11 . sMV/exosome preparations (10 μg protein) were lysed in SDS sample buffer, resolved by SDS-PAGE, electrotransferred as previously described 11 79 . Membranes were probed with primary mouse anti-CD9 (1:1000, BD Biosciences), mouse anti-Alix (1:1000, Cell Signaling), and mouse anti-A33 (1 μg/ml, a kind gift from Dr. A. Scott, Ludwig, Austin Campus). Membranes were further incubated with secondary antibodies horse radish peroxidase (HRP)-conjugated anti-mouse IgG (1: 15,000, Sigma) and IRDye 800 goat anti-mouse IgG (1: 15000, Li-COR Biosciences). All antibody incubations were carried out using gentle orbital shaking at RT. Proteins were visualised by incubating membranes with Western HRP substrate (Merck-Millipore) followed by imaging with ChemiDoc MP System (Bio-Rad) or imaged directly with the Odyssey Infrared Imaging System, version 3.0 (LI-COR Biosciences, Nebraska USA). Total RNA was extracted using TRIzol ® Reagent (Life Technologies) according to the manufacturer’s protocol, as described previously 11 and the quality and quantity determined using an Agilent 2100 Bioanalyzer (Agilent Technologies). These RNA samples were subsequently used for cDNA library construction (LIM1863 whole cells (CL), A33-Exos, EpCAM-Exos, sMVs) and Illumina sequencing which was performed by Beijing Genomics Institute (BGI, Shenzhen, China). A total amount of 1 μg RNA per sample was used for cDNA library construction using the TruSeq™ RNA Library Preparation Kit v2 protocol (Illumina). Briefly, after DNase digestion and RNA purification, poly(A) mRNA was purified from total RNA using Dynal™ oligo(dT)-attached magnetic beads (Invitrogen-GIBCO). The mRNA was chemically cleaved into small fragments (~200 nt) by exposure to divalent cations under elevated temperature (94 °C, 8 min) in Elute/Prime/Fragment Mix buffer (Illumina). These cleaved RNA fragments were used to synthesize first strand cDNAs using random hexamer-primers and reverse transcriptase (SuperScript ® II Reverse Transcriptase, Invitrogen-GIBCO). Second-strand cDNA synthesis was performed using DNA Polymerase I (Invitrogen-GIBCO) and RNase H (Invitrogen-GIBCO). Double-stranded cDNA fragments were purified using AMPure XP beads (Beckman Coulter) and remaining overhangs converted into blunt ends via treatment with T4 DNA polymerase/Klenow DNA polymerase/T4 exonuclease/polymerase using End Repair Mix (Illumina). After adenylation of 3′ ends of DNA fragments using Klenow DNA polymerase the sample was purified using AMPure XP beads. After ligation of Illumina sequencing adaptors to the cDNAs, cDNA fragments were gel-purified using a 1.5% Tris-borate-EDTA polyacrylamide gel (Invitrogen), size-selected (250–350 bp) and amplified by PCR. Amplified cDNA libraries were quality controlled (Agilent 2100 Bioanalyzer and qRT-PCR). The final product should be a band at approximately 260 bp (for single-read libraries). qRT-PCR was used for quantification of sequencing adapter ligation and cDNA libraries were sequenced using an Illumina HiSeq2000 sequencer (typical read lengths 90 bp) and 200 bp paired-end reads were generated. Nucleotide sequences were represented based on an image format, where images generated by the sequencer were converted into nucleotide sequences using a base-calling pipeline (Illumina). Raw reads were saved as fastq formatted files. The raw reads were cleaned by removing adaptor sequences, low quality reads containing >50% bases with quality [QA] ≤15, and reads with >2% undefined nucleotides [N]. Clean reads can be publically accessed from the Sequence Read Archive (SRA) of NCBI using the accession number SRA180512. Clean reads were aligned to the human genome sequence and mRNA and ncRNA profiles for each library were determined using the programs TopHat2 and Cufflinks (v2.2.1) 13 81 , respectively. Briefly, human genome (GRCh37.74) and gene annotation data were downloaded from Ensembl database ( http://www.ensembl.org/index.html ) and TopHat2 was used to align the clean reads to the human genome by incorporating the Bowtie2 ( http://bowtie-bio.sourceforge.net/bowtie2/index.shtml ) algorithm. The processed alignment result file was used to profile mRNA and ncRNA expression using Cufflinks. Gene biotype annotation by Ensembl Automatic Gene Annotation System 14 and GENCODE 15 were used to distinguish mRNAs from ncRNAs (pseudogene transcripts, long noncoding RNAs and short noncoding RNAs). Paired alignment reads were used to calculate FPKM (fragments per kilobase of transcript per million mapped reads) values for mRNA/ncRNA transcript expression levels. FKPM values were calculated as follows: Here C, N and L represent the substitute reads with fragments, the total number of sequenced reads and the length of a particular mRNA/ncRNA, respectively. A cut-off value of >1 FPKM was used as a limit for mRNA/ncRNA detection across all samples. Canonical and variant human mRNAs were classified by linking UniProt and Ensembl annotated protein sequence databases. mRNAs and ncRNAs enriched in EV samples, relative to CL, were selected using the following criteria: FC (fold change) >2, p-value 0.7; where FC values represent RNA enrichment changes according to the formula: p-values were calculated using the PERL module COMPARISON based on Poisson distribution 82 , and probability values were obtained using the R package NOIseq 83 . TopHat2 81 and ChimeraScan 84 (v0.4.5) were used to identify alternative splicing and gene fusion events, according to the protocols, respectively. Gene ontology terms for mRNAs enriched in EVs and KEGG pathway analysis were determined using the DAVID Bioinformatics Resource v 6.7 ( http://david.abcc.ncifcrf.gov ) 85 . EV-enriched mRNAs and lncRNAs identified in this study (relative to CL) were analysed against gene expression data in matched tumour and normal tissues of colon cancer patients from SRA database ( https://www.ncbi.nlm.nih.gov/Traces/sra/sra.cgi? ). Two SRA studies with accession numbers of SRP022054 76 and SRP029880 77 used deep sequencing technology to identify gene expression levels in 4 and 23 pairs of normal/tumour tissues, respectively. Raw sequencing data was obtained and re-analysed based on the criteria established previously for identification of over-expressed mRNAs and lncRNAs. Total RNA (CL and EVs) were reverse transcribed into cDNA and synthesised with Reverse Transcription (RT) Master Mix (Applied Biosystems) according to the manufacturer’s instructions. Forward and reverse primers for seven selected mRNAs (SCARB1, SCD, TPT1, EEF1G, BCL7C, RPS3, and RAB13) and three internal controls (CKS1B, GAPDH and MLF2) were designed using OligoArchitec TM Online (Sigma-Aldrich, for primer sequences, see Supplementary Table 7 ) and synthesised by Integrated DNA Technology Inc. PCR experiments were optimised to determine the sample concentration, primer concentration and reaction temperature for each primer pair. SsoAdvanced TM Universal Supermixes (5 μl, Bio-Rad Laboratories Inc.), forward primer (10 μM, 0.5 μl), reverse primer (10 μM, 0.5 μl) and RNase-free H 2 O (2 μl) were added to 2 μl of cDNA sample (12 ng/μl) per reaction. Technical triplicate analyses for each mRNA were performed for each sample on CFX96 Touch TM Real-Time PCR Detection System (Bio-Rad Laboratories Inc.). The CFX Manager TM Software (v3.1) was used to analyse the mRNA expression levels and to calculate the expression changes between samples.

How to cite this article : Chen, M. et al . Transcriptome and long noncoding RNA sequencing of three extracellular vesicle subtypes released from the human colon cancer LIM1863 cell line. Sci. Rep. 6 , 38397; doi: 10.1038/srep38397 (2016). Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Secretion and reciprocal exchange of EVs between cells is emerging as a central paradigm in cancer biology, especially in the tumor microenvironment. EVs contain protein/RNA/DNA/lipid-laden cargoes which upon uptake by recipient cells play a critical physiological role in healthy and pathological conditions. Previously, we reported the isolation of three EV subtypes from the human CRC cell line LIM1863 –sMVs and two exosome populations (basolateral A33-Exos and apical EpCAM-Exos). Although profiling their protein 8 10 and miRNA 11 cargoes has shed much light on the nature of these vesicles, identifying their RNA cargo will further assist our understanding of how they modulate recipient cell behavior. In this study we identified a total of 2,389 cellular protein coding RNAs, 317 pseudogene transcripts, 1,028 long noncoding RNAs and 206 short noncoding RNAs that preferentially distributed to LIM1863 EVs. For each RNA category we identified a number of RNA transcripts commonly enriched in all EV subtypes, relative to CL, as well those specifically enriched in each EV subtype. TPT1 transcripts, mRNAs encoding ribosomal proteins, FTL, and lncRNAs ZFAS1 and SNHGs, which are commonly enriched in the three EV subtypes, are up-regulated in previously published CRC tumor tissues 76 77 when compared to matched normal colon tissue. Additionally, we observe several novel RNAs that warrant further analysis as potential CRC prognostic biomarker candidates and targets for clinical management. Among the protein coding RNAs, we found 446 mRNAs encoding canonical proteins including TPT1, ribosomal proteins, FTL and EEFs, 74 mRNAs encoding isoform proteins and 27 mRNAs without protein evidence in UniProtKB (i.e., ‘missing’ proteins). Novel alternative splicing and gene fusions we identified in LIM1863 cells and derived EVs warrant further study as possible neoantigen sources. Finally, we observed 151 RNA-binding proteins (RBPs) in LIM1863 EVs - 7 of which are reported to bind to RNA transcripts identified in our study by RNA-Seq. Remarkably, interrogation of the Spliceosome Database ( http://spliceosomedb.ucsc.edu/ ) identified the RBPs and cognate snRNAs for the ribonucleoprotein complexes U1 and U2. To our knowledge this is the first report of RNPs in EVs and raises the possibility that EVs may play a key role modulating mRNA splicing upon uptake in recipient cells.