Discovery of a diagnostic biomarker for colon cancer through proteomic profiling of small extracellular vesicles

HT-29 (ATCC® Number HTB-38™), HCT-116 (ATCC® Number CCL-247™) CC epithelial cells and CRL-1541 (ATCC® CRL-1541™) colon normal fibroblasts were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% EV-depleted fetal bovine serum (FBS) and 1% antibiotic/antimycotic solution. Cells were obtained from the American Type Culture Collection (Manassas, VA, USA) and tested for mycoplasma contamination by PCR.

HT-29 and HCT-116 cells (50,000 cells/well) were seeded into 24-well plates in DMEM supplemented with 10% EV-depleted FBS and cultured to 90% confluence. The medium was replaced with serum-free medium after 0, 12, 24, and 48 h, and 0.5 mg/mL MTT (3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide) solution was added to each well and incubated for 3 h at 37 °C. Five hundred microliters of 100% isopropanol was added, 200 μL of this mixture was transferred to 96-well plates, and absorbance was measured at 495 nm.

Small-EVs were purified by differential centrifugation as described previously [23, 24]. Briefly, supernatants from CC cells were subjected to differential centrifugation at 300, 1500, 2500, and 100,000×g. Small-EVs (100,000×g pellet) were resuspended in phosphate-buffered saline (PBS) for further analysis. As an alternative method for purification of small-EVs from plasma, the plasma was mixed with ExoQuick (System Biosciences, Mountain View, CA, USA). An optimized ExoQuick method was used as described in our previous study [25]. In the optimized ExoQuick method, after incubation at 4 °C for 30 min, the samples were centrifuged at 1500×g for 30 min, and the pellet was washed three times with PBS. These extra steps were required to reduce contamination with plasma proteins such as albumin. Small-EV proteins were quantified using the Pierce BCA Protein Assay kit (Rockford, IL, USA) after treatment with RIPA buffer (Cell Signaling Technology, Danvers, MA, USA).

Purified small-EVs were deposited onto pure carbon-coated EM grids. After staining with 2% uranyl acetate, the grids were dried at 25 °C and visualized at 6000× and 10,000× magnification using the Hitachi H-7000 transmission electron microscope (Tokyo, Japan).

The number of small-EVs was measured by NTA as described in our previous study [26]. Suspensions containing small-EVs from plasma or cell culture medium were analyzed using the Nano-Sight LM10 instrument (NanoSight, Wiltshire, UK). For this analysis, a monochromatic laser beam (405 nm) was applied to a dilute suspension of small-EVs. A video of 30- to 60-s duration was recorded at a rate of 30 frames/s, and small-EV movement was analyzed using NTA software version 2.2 (NanoSight). NTA post-acquisition settings were optimized and remained constant between samples, and each video was analyzed to estimate the concentration.

Protein concentrations of small-EVs in HT-29, HCT-116, and CRL-1541 cells were determined using the BCA assay after cell lysis. Digestion was performed as described previously [24]. Briefly, 50 μg protein was resuspended in 100 mM triethylammonium bicarbonate (TEABC; pH 8) containing 6 M urea, 5 mM EDTA, and 2% SDS. The proteins were chemically denatured by adding 10 mM dithiothreitol and incubating at 60 °C for 20 min. Next, proteins were alkylated by adding 50 mM iodoacetamide at room temperature and incubating for 20 min. The protein solution was mixed with acrylamide/bisacrylamide solution (30% v/v, 29:1) containing 10% (w/v) ammonium persulfate and tetramethylethylenediamine. The gel was cut into small pieces and washed three times with 50 mM TEABC containing 50% acetonitrile (ACN). Proteolytic digestion was performed using trypsin (protein:trypsin = 50:1 w/w) in 50 mM TEABC and incubating overnight at 37 °C. The digested peptides were extracted from the gel by exchanging with two buffers: 0.1% formic acid (FA) in 50 mM TEABC and 0.1% FA in 50% ACN. The eluents were concentrated using a Speedvac and desalted using an HLB cartridge (Waters, Milford, MA, USA).

Small-EVs were analyzed by LC–MS/MS as described previously [27]. One microgram of extracted peptide was analyzed by nano-ultra-high-performance LC (UPLC) (Waters) and tandem mass spectrometry using a Q-Tof Premier (Waters). Peptides were injected into a 2 cm × 180 μm trap column and resolved in a 25 cm × 75 μm nanoACQUITY C18 column (Waters) using the LC system. Mobile phase A was composed of water containing 0.1% FA and mobile B phase was 0.1% FA in ACN. The peptides were resolved using a gradient of 3–45% mobile phase B over 135 min at a flow rate of 300 nL/min. All samples were analyzed in triplicate. The method included a full sequential MS scan (m/z 150–1600, 0.6 s) and five MS/MS scans (m/z 100–1990, 0.6 s/scan) for the five most intense ions present in the full-scan mass spectrum.

For protein identification, MS raw data were converted into peak lists using MASCOT Distiller version 2.1 (Matrix Science, London, UK) with default settings. All MS/MS raw data were analyzed using MASCOT version 2.2.1 (Matrix Science) [27]. Mascot was used to search the Swiss-Prot database (release 2018_07) with human taxonomy. The database search against Mascot was performed with a fragment ion mass tolerance of 0.5 Da and parent ion tolerance of 0.2 Da. Two missed cleavages were allowed for trypsin digestion. Carbamidomethylation of cysteine residues and oxidation of methionine residues were considered as variable modifications. To evaluate the false discovery rate (FDR) of protein identification, this analysis was repeated using identical search parameters and validation criteria against a randomized decoy database created by Mascot. Peptide identities were assigned if their Mascot ion scores corresponded to p < 0.05. Proteins with more than 2 peptides were identified with confidence. FDRs with Mascot protein scores > 34 (p < 0.05) ranged below 2%.

Proteins were resolved by SDS-PAGE, transferred onto nitrocellulose membranes, probed with each primary antibody, and incubated with horseradish peroxidase-conjugated secondary antibody. The blots were visualized with enhanced chemiluminescence detection reagents (Thermo Fisher Scientific, Waltham, MA, USA) and quantified using enhanced chemiluminescence hyperfilm (AGFA, Morstel, Belgium). The following primary antibodies were used against CD63 (ab68418, 1:1000), HSP70 (ab2787, 1:1000), CD9 (ab92726, 1:1000), Alix (ab56932, 1:1000), LGALS3BP (ab123921, 1:1000), SLC1A5 (ab58690, 1:1000), CLDN7 (ab53044, 1:1000), RAI3 (ab155557, 1:1000; all from Abcam, Cambridge, UK), and TSPAN1 (H00010103, 1:1000; Abnova, Taipei City, Taiwan).

The method for detecting small-EV proteins has been described previously [21]. Anti-CD63 antibody (1:250)-coated 96-well plates were blocked with PBS containing 0.05% Tween 20 (PBS-T) for 3 h at room temperature. Serial dilutions of mouse serum were prepared in PBS-T supplemented with 10% horse serum. The diluted solution was added to the plates in duplicates and incubated for 2 h at room temperature. After washing, the samples in the plates were reacted with the monoclonal anti-TSPAN1 antibody pre-incubated with a horseradish peroxidase-linked secondary antibody for 30 min and developed using 3,3′,5,5′-tetramethylbenzidine-containing peroxide. The reaction was stopped, and optical density values were measured at 450 nm using an automated iMark (Bio-Rad, Hercules, CA, USA).

Data are presented as the mean ± SD, as indicated in each graph. The Student’s t-test was used to evaluate the differences between means for normally distributed immunoblotting data. The significance of the ELISA results was analyzed using the Mann-Whitney test. All p-values were two-tailed, and p-values less than 0.05 were considered statistically significant. Receiver operating characteristic (ROC) curves were generated based on logistic regression from the ELISA data. Statistical analysis was performed using SPSS 22.0 for Windows (SPSS, Inc., Chicago, IL, USA). The area under ROC curve (AUC) calculation provided a convenient number. AUC < 0.5 indicated that the test showed no difference between the two groups, while 0.5 < AUC < 1.0 indicated that the test yielded perfect differentiation between groups at *p < 0.05, **p < 0.01, ***p < 0.001, or ****p < 0.0001.

To identify potential biomarkers for diagnosing colon cancer, HT-29 and HCT-116 human colon cancer cell lines were used in this study. A schematic workflow is shown in Fig. 1.

We analyzed cell viability at different incubation times under starvation conditions to optimize the incubation time and observed reduced cell viability at more than 24 h incubation (Fig. 2a). Isolated small-EVs were analyzed by western blotting, TEM, and NanoSight. Immunoblotting revealed that the small-EV fraction was highly enriched in small-EV markers such as CD63 and Alix (Fig. 2b). It also showed rounded morphology (Fig. 2c) and a size distribution (Fig. 2d) consistent with typical small-EVs.

We performed proteomic analysis of small-EVs from CC cell lines by MS to identify small-EV proteins. CC cell-derived small-EVs were analyzed in two biological and three analytical replicates using nano-UPLC–MS/MS. Small-EV proteins were searched in the Swiss-Prot database with human taxonomy using the MASCOT software. From among high-confidence peptide sequences with error rates < 5%, we identified 316 and 324 proteins in HCT-116- and HT-29-derived small-EVs, respectively. This analysis showed that proteins with a peptide hit score (PHS) > 1 were identified with high confidence by multiple peptides and overlapping gene names were removed. A total of 464 proteins were identified in CC cell-derived small-EVs with high confidence (PHS > 1) (Fig. 3a; see Additional file 1: Table S1). Among these, 176 proteins were common between the two cell lines. We compared the identified proteins with the data in ExoCarta and observed that 90% of the identified small-EV proteins were present in this database (Fig. 3a). For bioinformatics analysis, including prediction of the function, biological process, cellular components, localization, and pathways of the exosomal proteins, we used the Protein Analysis Through Evolutionary and Relationship (PANTHER) software. The identified proteins were associated with at least one annotation term. All searched proteins were categorized as having binding (36.3%), catalytic (32.2%), or structural (13.6%) activities (Fig. 3b). The distribution of biological processes was as follows: cellular (29.1%), metabolic (19%), and cellular component organization or biogenesis (12.2%) (Fig. 3c). As shown in cellular components, 35.9% proteins were predicted to be localized in the cytoplasm, 24.5% in organelles, and 17.5% in macromolecular complexes (Fig. 3d). The distribution by protein class was as follows: enzyme modulator (15.9%), nucleic acid-binding (13.2%), and cytoskeletal protein (12.1%) (Fig. 3e). Signaling pathway analysis of the identified proteins confirmed their associations with a variety of pathways (Fig. 3f).

From the identified proteins, we selected candidate proteins to identify CC-specific proteins. The filtration step to select the candidate proteins is shown in Fig. 1. A total of 464 small-EV proteins were identified in the HCT-116 and HT-29 cell-derived small-EVs by mass spectrometry. Among these, 415 proteins known to be present in EVs were selected. Proteins detected in HCT-116- and HT-29-derived small-EVs, but not in small-EVs from CRL-1541 or normal colon cells (see Additional file 2: Table S2), were selected (352). Next, we excluded housekeeping proteins such as actin, myosin, and ribosomal proteins (231). Extracellular matrix proteins and outer membrane proteins were selected to detect small-EVs in the plasma without a purification step (114). Finally, we selected the following five proteins based on previous studies: TSPAN1, galectin-3-binding protein (LGALS3BP), neutral amino acid transporter B(0) (SLC1A5), claudin 7 (CLDN7), and retinoic acid-induced protein 3 (GPRC5A).

We performed immunoblotting analysis to detect candidate proteins in CC cells and confirm whether these proteins observed in the proteomic analysis were truly expressed in the small-EVs isolated from CC cells. CRL-1541 cells were analyzed and compared with CC cells. HSP70 was used as a known small-EV marker and β-actin was used as a cytosol marker for cell lysates. All candidate proteins were distinctly detected in CC cell-derived small-EVs (Fig. 4). Among these, expression of three proteins (SLC1A5, LGALS3BP, and TSPAN1) was increased in small-EVs isolated from CC cell lines.

To verify the levels of selected proteins in clinical samples, we used human plasma from HCs and patients with colon cancer. Detailed data related to clinical sample characteristics are provided in Table 1. After isolating small-EVs from the plasma, the selected proteins in small-EVs isolated from HC (n = 3) and CC patients (n = 9) (Fig. 5a) were verified by immunoblotting. Relative expression of the candidate biomarker proteins is shown in Fig. 5b in HC vs CC patients. TSPAN1 expression (HC:CC = 1:1.5, p = 0.013) was significantly higher in small-EVs isolated from the plasma of CC patients than in those of HC patients; however, LGALS3BP (HC:CC = 1:2.3, p > 0.05), SLC1A5 (HC:CC = 1:0.6, p > 0.05), CLDN7 (HC:CC = 1:0.4, p > 0.05), and GPRC5A (HC:CC = 1:0.4, p > 0.05) expression was not significantly different between the two groups (Fig. 5b).

We validated TSPAN1 in small-EVs from HCs (n = 30) and CC patients (n = 37) by ELISA. TSPAN1-positive small-EVs in plasma were captured using a coated anti-CD63 antibody and detected using an anti-TSPAN1 antibody (Fig. 6a). Our previous studies showed that this method was effective for detecting small-EVs [31, 32]. These results indicated that the TSPAN1 level was significantly higher in CC patients than in HCs (Fig. 6b). Additionally, ROC curves were generated using ELISA results to describe diagnostic performance (Fig. 6c), and AUC was determined for TSPAN1. Fig. 6 shows that the AUC of TSPAN1 for differentiating between CC patients and HCs was 0.828 (95% CI = 0.73–0.926), with a sensitivity of 75.7% and specificity of 66.7%.