RETRACTED ARTICLE: Exosome-transmitted miR-128-3p increase chemosensitivity of oxaliplatin-resistant colorectal cancer

All cell lines were purchased from Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). Oxaliplatin (Sanofi-Synthelabo, NY, USA) was purchased from the Second Hospital of Shandong University. Oxaliplatin-resistant cell lines (HCT116OxR and HT29OxR) were established as previously described [2, 21, 22]. For exosome co-cultures, 1 μg/ml (as determined by BCA protein assay (Thermo Fisher Scientific, USA)) of exosomes were added to the culture medium of recipient cells (5 × 10⁶).

Tissue samples were collected prior to initiation of oxaliplatin therapy from surgical specimens or biopsies of advanced CRC patients and obtained informed consent from the Second Hospital of Shandong University (n = 86), Qilu Hospital of Shandong University (n = 67) and Shandong Provincial Traditional Chinese Medical Hospital (n = 20) between July 2008 and February 2018. This study was approved by the Ethics Committee of the Second Hospital of Shandong University. Patients in the oxaliplatin group were treated with at least six cycles of oxaliplatin, while patients in untreated group received no chemotherapy or stopped oxaliplatin therapy (< 21 days) due to adverse effects. Detailed clinical characteristics of patients in training phase are listed in Additional file 1: Table S1.

Tumor response to chemotherapy was assessed as a 3-dimensional volume reduction rate or tumor response rate (radiologic assessment), and evaluated as per the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines [23]. Patients with symptomatic deterioration, or appearance of new lesions, or radiologic assessment of ≥25% tumor regrowth in validation phase were allocated to the progressed disease (PD) group (n = 35) and remaining to non-PD group (n = 40). PFS was defined as the duration from tumor resection to PD. All patients were followed up in the clinic every 3 months (0–2 years), 6 months (2–4 years), and yearly until death or February 2018. Follow-up studies included computed tomography of the abdomen and postsurgical physical examination.

Exosomes were isolated from FHC cell conditioned medium by 0.22 μm filtration and ultracentrifugation as we previously described [24]. Transmission electron microscopy (JEM-1-11 microscope, Japan) was used to image exosomes at 100 keV, and quantified by Nanosight NS300 instrument (Malvern Instruments Ltd. UK) equipped with NTA 3.0 analytical software (Malvern Instruments Ltd. UK).

Immunohistochemistry was done as previously reported [22]. Positive cells were counted in five random fields per slide. Interpretation of staining intensity of Bmi1 or MRP5 was made independently by two specialists, as no staining = 0, weak staining = 1 (1–25%), moderate staining = 2 (26–50%), and strong staining = 3 (51–100%).

Cellular and exosomal RNAs were isolated using the miRNeasy Micro Kit (QIAGEN, Valencia, CA, USA). First-strand cDNA was synthesized with random primers using High Capacity cDNA Reverse Transcription Kit (Takara, Dalian, China). qPCR was performed using Power SYBR Green (Takara, Dalian, China) on a CFX96 Real-Time PCR Detection System (Bio-Rad, USA). Data was collected and normalized to U6 levels (for cellular miR-128-3p), GAPDH (for cellular Bmi1 and MRP5 mRNA) or miR-16 (for exosomal miR-128-3p) [25]. MicroRNA primers were synthesized by Ribobio (Guangzhou, China). mRNA primers are listed in Additional file 2: Table S3.

Total protein of cells or exosomes was extracted with RIPA buffer (Sigma-Aldrich) and quantified with the BCA assay (Pierce, Rockford, IL), and Western Blot preformed as previously described [21]. Protein band intensity was quantified by densitometry using Image Lab software (Bio-Rad, Hercules, CA, USA). All antibodies used are shown in Additional file 3: Table S2.

Lentiviral plasmids encoding miR-128-3p and negative control were designed and produced by HANBIO (Shanghai, China). HCT116OxR and HT29OxR cells were transfected with lentivirus (pHB-U6-MCS-CMV-ZsGreen-PGK-PURO) at a multiplicity of infection (MOI) of 20 and 15, respectively. The cells were then selected with 1 μg/ml puromycin for 3 days. pcDNA3.1 vector containing Bmi1-wt, Bmi1-mut, MRP5-wt or MRP5-mut and control were purchased from GENECHEM (Shanghai, China). miR-128-3p mimics, inhibitor and control were produced by GENECHEM (Shanghai, China). Plasmid, mimics, inhibitor and negative control were transfected using Lipofectamine2000 (Invitrogen, California, USA) according to the manufacturer’s instructions.

Cells viability was determined by Cell Counting Kit 8 (Dojindo, Japan) and measured at OD450 nm with the Thermo Scientific Multiskan FC (Thermo Fisher Scientific Corporation, USA).

Cell migration was measured with Transwell assays, briefly cancer cells (2 × 10⁴ cells/well) were divided into different groups and seeded in the upper chambers in serum-free media with or without the Matrigel membrane. Meanwhile the lower chambers were loaded with RPMI1640 containing 5% FBS. After incubation at 37 °C, 5% CO₂ for 24 h, the lower chamber was imaged using an inverted microscope. The upper chamber was cleaned with a cotton swab and the lower chamber was immersed and washed with PBS, fixed with 4% paraformaldehyde, stained with 0.1% crystal violet, washed three times with water, and imaged by Inversion Microscope (Zeiss, Germany).

Four-weeks old male BALB/C nude mice were purchased from Weitonglihua (Peking, China). HCT116OxR cells (5 × 10⁶ cells per mouse) were injected subcutaneously into the right flank of nude mice. Two weeks later, the nude mice generated tumors approximately 200 mm³ in size. Purified exosomes (5 μg) or PBS was then injected intratumorally twice weekly with or without oxaliplatin treatment (90 mg/m²). The tumor size and bodyweight were measured twice per week. After seven weeks, mice were sacrificed and tumor tissues were prepared for histological examination. Tumor volume (mm³) = 0.5 × width² × length. All animal work was performed according to the Health guidelines, and protocols were approved by the Institutional Animal Care and Use Committee of Shandong University.

PKH67 (Sigma-Aldrich, USA) (1 μM) was used to label exosomes according to manufacturer’s instructions. 24 h after PKH67-labeled exosomes were incubated with HCT116OxR, 4′,6-diamidino-2-phenylindole (DAPI) (Invitrogen, USA) was used for cell nuclear staining. The slides were fluorescently visualized with a laser scanning microscope Axio-Imager-LSM800 (ZEISS, Germany). Rhodamine-conjugated secondary antibody (Cell Signaling Technology, USA) for γ-H₂AX protein and DAPI for nuclear staining. The slides were visualized for immunofluorescence with a laser scanning microscope (Zeiss, Germany).

Gene Pulser X Electroporator (Bio-Rad, USA) was used to electroporate miR-128-3p into exosomes as previously described [22, 26, 27]. Briefly, 2 μg exosomes and 400 nmol miR-128-3p mimics were mixed in 400 μl of electroporation buffer at 4 °C. After electroporation at 350 V and 150 μF, the mixture was incubated at 37 °C for 30 mins to fully recover the membrane of exosomes.

Cells were co-transfected with Dual-Luciferase reporter system using pmiR-REPORT™ luciferase vectors containing wild-type or mutant 3`-UTR of Bmi1 and MRP5 and miR-128-3p mimics or miR-128-3p mimic-NC using Lipofectamine 2000 (Invitrogen, California, USA). Luciferase activity was measured by Dual-Luciferase Reporter Assay System (Beyotime Biotechnology, Shanghai, China) 48 h after transfection. Each group was run in triplicate in 24-well plates.

Intracellular Pt content was quantified as described previously [28]. Briefly, after treatment with oxaliplatin (30 μM), cells were lysed overnight with 0.1% triton-X-100 and 0.2% nitric acid. Cell lysates were subject to Atomic Absorption Spectrophotometry AA-6880 (SHIMADZU, Japan) and protein concentration was determined by BCA protein assay. DNA was isolated using DNeasy Blood & Tissue Kits (Qiagen, Valencia, CA) according to specifications and quantified with a NanoDrop spectrophotometer (Thermo Fisher Scientifc, USA). The same DNA hydrolysate was used for Pt measurements by Atomic Absorption Spectrophotometry AA-6880 (SHIMADZU, Japan).

To obtain oxaliplatin-resistant colorectal cancer cells, we treated HCT116 and HT29 (lowest IC50 of seven CRC cell lines to oxaliplatin, Additional file 4: Figure S1A) in vitro with escalating oxaliplatin concentrations and then grafted cells into nude mice and performed cycles of oxaliplatin treatment along with three passages in vivo. CRC cells from the third passage xenografts that acquired oxaliplatin resistance at over clinically relevant concentrations (2 μM) [2] were named HCT116OxR and HT29OxR (Fig. 1a). Compared to parental cells, HCT116OxR and HT29OxR cells responded poorly to oxaliplatin, as illustrated by an increased IC50 and decreased drug-induced apoptosis (Fig. 1b, c and Additional file 4: Figure S1B). Meanwhile, resistant cells had phenotypic changes including: loss of intercellular adhesion, spindle-cell morphology (loss of cell polarity), and increased pseudopodia formation (Fig. 1d). Furthermore, resistant cells exhibited higher migration and invasion than their parental cells (Fig. 1e) and enhanced motility as observed in wound healing assay (Fig. 1f). Western blot analysis of resistant cells revealed typical of changes in cells with chemotherapy-induced EMT including decreased E-cadherin protein expression but dramatically increased N-cadherin, vimentin and fibronectin (Fig. 1g). In addition, quantitative platinum (Pt) analysis following oxaliplatin treatment (30 μM) for 24 h showed total intracellular Pt and DNA-bound Pt was lower in resistant than parental cells (Fig. 1h and i). Immunofluorescence assay using nuclear foci of γ-H₂AX expression as an indicator of DNA double-strand breaks revealed that oxaliplatin induced more DNA damage in parental cells than oxaliplatin-resistant cells (Fig. 1j).

Using RT-qPCR, we found all seven CRC cell lines expressed lower miR-128-3p than normal intestinal epithelial FHC cells (Additional file 4: Figure S1C). Importantly, miR-128-3p expression levels were also markedly higher expressed in HCT116 and HT29 compared with other CRC cell lines (Additional file 4: Figure S1C). Additionally, miR-128-3p expression was significantly decreased in resistant cells compared to respective parental cells (Fig. 2a). To further elucidate the role of miR-128-3p in oxaliplatin resistance, we stably overexpressed miR-128-3p using a lentiviral vector expression system (lenti-miR-128-3p) in oxaliplatin-resistant CRC cells. As expected, resistant cells transfected with lenti-miR-128-3p showed miR-128-3p levels several orders of magnitude higher than lenti-negative control (lenti-NC) transfected cells (Fig. 2b and Additional file 4: Figure S1D). Enhanced miR-128-3p expression in resistant cells reduced IC50 and increased cell apoptosis following oxaliplatin treatment (Fig. 2c, d and Additional file 4: Figure S1E–G). miR-128-3p overexpression significantly upregulated E-cadherin expression and downregulated N-cadherin, vimentin, and fibronectin expression in resistant cells (Fig. 2e and Additional file 4: Figure S1H–J). Moreover, miR-128-3p inhibited the migration, invasiveness, and motility of resistant cell lines (Fig. 2f, g Additional file 5: Figure S2A, B). Importantly, we found miR-128-3p downregulates drug efflux in CRC cells. Quantitative Pt analysis showed that following a 24 h oxaliplatin incubation (30 μM), total intracellular Pt (Fig. 2h and Additional file 5: Figure S2C) and DNA-bound Pt (Fig. 2i and Additional file 5: Figure S2D) was markedly higher in lenti-miR-128-3p transfected cells than control cells, confirming that reduced cellular Pt content is an important mechanism of oxaliplatin resistance. After 24 h of oxaliplatin treatment, nuclear foci of γ-H₂AX levels in lenti-miR-128-3p cells were significantly increased, but remained low in lenti-NC cells (Fig. 2j and Additional file 5: Figure S2E). Finally, to determine whether miR-128-3p sensitizes CRC cells to chemotherapeutic agents in vivo, lenti-miR-128-3p transfected HCT116OxR cells were implanted subcutaneously into nude mice then treated with oxaliplatin. Our data indicated that miR-128-3p overexpression significantly decreased oxaliplatin resistance in HCT116OxR xenografts in vivo (Fig. 2k and l). These data support our in vitro findings, indicating that miR-128-3p ameliorates oxaliplatin-resistant CRC in vitro and in vivo. Collectively, these results demonstrate that decreased miR-128-3p expression is indispensable for oxaliplatin resistance in CRC cells.

As we observed miR-128-3p was substantially downregulated in both HCT116OxR and HT29OxR cells compared to parental cells, we felt it was important to establish clinical relevance in human CRC. Therefore, we analyzed miR-128-3p tissue levels in an independent large-scale sample set using RT-qPCR. Kruskal–Wallis test analysis indicated miR-128-3p expression levels in tissues were significantly lower in patients suffering from tumor relapse after oxaliplatin therapy (0.065; 0.033–0.088) compared with patients that responded well to neoadjuvant oxaliplatin therapy (0.169; 0.101–0.237) or who were therapy naive (0.204; 0.086–0.277) (all at P < 0.001). Compared with noncancerous tissues (1.014; 0.937–1.904), miR-128-3p expression levels were also markedly reduced in CRC tissues (P < 0.001, Fig. 3a and Additional file 1: Figure Table S1). We then explored the predictive value of miR-128-3p levels for oxaliplatin response in CRC patients. Expression levels of miR-128-3p were downregulated in patients with progressive disease (PD) during oxaliplatin therapy than those without PD (non-PD) (Fig. 3b). Receiver operating characteristic (ROC) curve analyses showed that miR-128-3p has a strong capability for discriminating non-PD patients from PD patients with an area under ROC curve (AUC) value of 0.868 (95% CI: 0.770–0.935, Fig. 3c). At an optimal cut-off value of 0.227, the sensitivity and specificity were 65.0 and 91.4%. Furthermore, the median miR-128-3p expression level (0.269) was used to categorize CRC patients into two groups; high-level (n = 61) and low-level (n = 61). Similar clinical characteristics between control and oxaliplatin treatment groups were observed before treatment. Associations between miR-128-3p and pathologies of CRC patients in validation phase are summarized in Additional file 2: Figure Table S3. Although Kaplan–Meier survival analysis indicated higher miR-128-3p expression was associated with longer progression-free survival (PFS) in CRC patients, there is limited benefit (Fig. 3d). Patients with low miR-128-3p expression had a poor PFS in oxaliplatin treatment (Fig. 3e). Conversely, the high miR-128-3p expression group exhibited a superior PFS after receiving oxaliplatin compared to the control group (Fig. 3f). Cox regression univariate and multivariate analysis revealed that oxaliplatin therapy correlated with improved PFS of CRC patients with high miR-128-3p expression (Table 1). Thus, miR-128-3p could serve as an independent predictor for oxaliplatin response in CRC patients.

Previous studies suggest miRNAs can be packaged into exosomes and functionally delivered to target cells to directly modulate target mRNAs [29]. Herein, we transfected FHC cells with lenti-miR-128-3p or lenti-NC. Subsequently, extracellular exosomes were isolated from the FHC supernatant after 48 h and identified with electron microscopy by their typical cup-shaped morphology (40 to 120 nm) and by exosomal markers (CD63 + CD9 + GM130-) (Fig. 4a and b). The miR-128-3p expression was markedly higher in lenti-miR-128-3p FHC cells (FHC-128) and their exosomes (128-exo) than in lenti-NC FHC (FHC-NC) and associated exosomes (NC-exo) (Additional file 5: Figure S2F and Fig. 4c). We then determined whether miR-128-3p was indeed present within exosomes. As expected, miR-128-3p expression in culture medium was unchanged upon RNase A treatment but significantly decreased when treated with RNase A and Triton X-100 simultaneously (Fig. 4d), suggesting that released miR-128-3p was protected by double-layer membrane instead of being directly released. When 128-exo (labeled with membrane phospholipid dye PKH67) were incubated with HCT116OxR cells, recipient cells exhibited high uptake efficiency, as measured by laser scanning confocal microscope (Fig. 4e). After 24 h` incubation, > 80% of recipient cells were positive for PKH67 fluorescence (Additional file 6: Figure S3A), suggesting that 128-exo are effectively internalized by HCT116OxR cells. 128-exo co-incubations increased miR-128-3p levels in resistant cells nearly 350-fold, while NC-exo did not affect miR-128-3p expression levels (Fig. 4f and Additional file 6: Figure S3B). Moreover, prolonged incubation caused a corresponding increase in miR-128-3p levels in HCT116OxR cells (Fig. 4g). Additionally, actinomycin D (RNA polymerase II inhibitor) did not significantly alter miR-128-3p levels, excluding the possibility of endogenous induction of miR-128-3p in recipient resistant cells (Fig. 4h). Taken together, these data demonstrate FHC-128 cells efficiently secrete exosomes containing miR-128-3p that can be directly transferred to HCT116OxR cells.

We next investigated whether exosome-transferred miR-128-3p could ameliorate chemosensitivity in resistant cells. In a cell viability and Annexin V/PI apoptosis assay, oxaliplatin-resistant cells incubated directly with 128-exo displayed elevated oxaliplatin sensitivity (Fig. 5a, b and Additional file 6: Figure S3C–E) which was abolished by transfecting recipient cells with a miR-128-3p inhibitor (anti-128). To exclude other factors in exosomes besides miR-128-3p in mediating oxaliplatin sensitivity, we electroporated miR-128-3p mimics directly into exosomes and it did not affect the physical properties of the exosomes (Additional file 6: Figure S3F). Indeed, HCT116OxR cells co-cultured with exosomes successfully up-took miR-128-3p mimics (Additional file 6: Figure S3G) and also exhibited increased oxaliplatin sensitivity (Additional file 6: Figure S3H). Importantly, incubated with 128-exo suppressed EMT, as characterized by upregulation of epithelial markers and downregulation of mesenchymal markers in resistant cells, while anti-128 blocked this progression (Fig. 5c and Additional file 7: Figure S4A–C). Moreover, 128-exo suppressed migration and invasion of resistant CRC cells across Transwell filters (Fig. 5d and Additional file 7: Figure S4D). And wound healing assays further showed that 128-exo suppressed motility in resistant cells (Fig. 5e and Additional file 7: Figure S4E). Both these affects were abolished with addition of anti-128. Using atomic absorption spectroscopy (AAS), we found that the total intracellular Pt and DNA-bound Pt levels in resistant cells was significantly increased when treated with 128-exo (Fig. 5f, g Additional file 7: Figure S4F, G). At 24 h` post oxaliplatin treatment, the nuclear foci of γ-H₂AX levels in the control group remained low but were significantly increased in the 128-exo co-culture group (Fig. 5h and Additional file 8: Figure S5A). Additionally, we found that the restored oxaliplatin sensitivity in recipient cells was sustained for at least 10 days following removal of 128-exo (Additional file 8: Figure S5B). Importantly, intra-tumor injection of miR-128-3p exosomes restored oxaliplatin response in resistant cells in vivo (Fig. 5i and j) and was accompanied by an increase in tumor miR-128-3p levels (Additional file 8: Figure S5C). Furthermore, miR-128-3p mimic-loaded exosomes were also effective. Therefore, these results confirm that miR-128-3p overexpression through 128-exo reverses oxaliplatin resistance in CRC cells in vitro and vivo.

We further sought to identify the mediators of miR-128-3p-driven oxaliplatin resistance. Previous reports suggest interactions of Bmi1/E-cadherin and ATP-dependent glutathione S-conjugate export pump (e.g. MRP5) are important regulators [30‐32]. To identify the potential pathway, we performed bioinformatic analysis with TargetScan and Miranda. This revealed the 3`-UTR of Bmi1 and MRP5 contain a predicted binding site for miR-128-3p (Fig. 6a). Furthermore, we found an inverse correlation between miR-128-3p levels and Bmi1 mRNA (r = − 0.445, P < 0.001) or MRP5 expression mRNA (r = − 0.538, P < 0.001) in CRC tissues (Fig. 6b and c). To verify whether Bmi1 and MRP5 are direct targets of miR-128-3p, Dual-Luciferase reporter system with pmiR-REPORT™ luciferase vectors containing wild-type or mutant 3`-UTR of Bmi1 and MRP5 was used. Co-transfection of miR-128-3p mimics significantly suppressed the luciferase activity of the reporter containing wild-type 3`-UTR, but not the mutant reporter (Additional file 9: Figure S6A and B). These data reveal that Bmi1 and MRP5 are direct functional targets of miR-128-3p. RT-qPCR and western blot assay indicated that the Bmi1 and MRP5 mRNA and protein levels were lower in miR-128-3p overexpressing cells compared to controls (Additional file 9: Figure S6C–E). To further validate these results, we employed a ‘rescue’ experiment by transfecting pcDNA3.1 vector carrying Bmi1 or MRP5 expression cassette with wild or mutated type seed sequences for miR-128-3p (Bmi-wt/Bmi1-mut or MRP5-wt/MRP5-mut) at its 3`-UTR into lenti-miR-128-3p transfected resistant cells. A cell viability and apoptosis assay indicated that only transfection of resistant cells with Bmi1-mut or MRP5-mut developed a resistant phenotype, while co-transfection with a Bmi1-wt or MRP5-wt did not and were silenced by miR-128-3p (Fig. 6d, e and Additional file 9: Figure S6F–H). Western blots assay demonstrated that transfecting resistant cells with Bmi1-mut permitted Bmi1, N-cadherin, vimentin, and fibronectin protein expression and suppressed E-cadherin protein expression, while transfection with Bmi1-wt was silenced by miR-128-3p and could not recover EMT (Fig. 6f and Additional file 10: Figure S7A). Moreover, ‘rescuing’ Bmi1-mut expression in the presence of miR-128-3p enhanced the invasion and migration of HCT116OxR cells (Fig. 6g, h and Additional file 10: Figure S7B, C). Collectively, these data suggest that miR-128-3p regulates chemotherapy-induced EMT of CRC cells by targeting Bmi1. We further sought to identify the drug efflux mechanism of miR-128-3p-driven oxaliplatin resistance. Western blotting revealed a significant increase in MRP5 protein in the MRP5-mut transfected cells, as MRP5 proteins in the mock, control groups or cells transfected with MRP5-wt remained unchanged (Fig. 6i and Additional file 10: Figure S7D). Importantly, transfected MRP5-mut significantly decreased the total intracellular Pt and DNA-bound Pt in lenti-miR-128-3p transfected HCT116OxR cells compared with the cells transfected with mock, control and MRP5-wt (Fig. 6j, k and Additional file 10: Figure S7E, F). The nuclear foci of γ-H₂AX expression levels revealed that oxaliplatin induced less DNA damage at 24 h in MRP5-mut group than other groups in lenti-miR-128-3p transfected resistant cell lines (Fig. 6l and Additional file 10: Figure S7G). In addition, treatment with 128-exo significantly decreased Bmi1 and MRP5 expression at protein levels in resistant cell lines, and miR-128-3p inhibitor in recipient cells disrupted this effect (Fig. 6m and Additional file 10: Figure S7H). Furthermore, forced miR-128-3p expression through intra-tumor injection of exosomes restored oxaliplatin response in resistant cells through blocked Bmi1 and MRP5 associated signaling in vivo (Fig. 6n). These findings indicate that 128-exo potentially facilitates oxaliplatin sensitivity in CRC cells by negatively regulating Bmi1 and MRP5 expression, two genes involved in oxaliplatin-induced EMT and drug efflux respectively.