Overcoming drug resistance is one of the biggest challenges in cancer chemotherapy. In this study, we examine whether targeting the long noncoding RNA taurine upregulated gene 1 (TUG1) could be an effective therapeutic approach to overcome drug resistance in pancreatic ductal adenocarcinoma (PDAC). TUG1 was expressed at significantly higher levels across 197 PDAC tissues compared with normal pancreatic tissues. Overall survival of patients with PDAC who had undergone 5-FU–based chemotherapy was shorter in high TUG1 group than in low TUG1 group. Mechanistically, TUG1 antagonized miR-376b-3p and upregulated dihydropyrimidine dehydrogenase (DPD). TUG1 depletion induced susceptibility to 5-FU in BxPC-3 and PK-9 pancreatic cell lines. Consistently, the cellular concentration of 5-FU was significantly higher under TUG1-depleted conditions. In PDAC xenograft models, intravenous treatment with a cancer-specific drug delivery system (TUG1-DDS) and 5-FU significantly suppressed PDAC tumor growth compared with 5-FU treatment alone. This novel approach using TUG1-DDS in combination with 5-FU may serve as an effective therapeutic option to attenuate DPD activity and meet appropriate 5-FU dosage requirements in targeted PDAC cells, which can reduce the systemic adverse effects of chemotherapy.
Targeting TUG1 coupled with a cancer-specific drug delivery system effectively modulates 5-FU catabolism in TUG1-overexpressing PDAC cells, thus contributing to a new combinatorial strategy for cancer treatment.
Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive human malignancies, in which overall median survival is less than 5 months and the 5-year survival rate is around 7% (1). The poor prognosis is due to several complex factors, including the lack of efficient biomarkers for diagnosis and the lack of potent therapeutic strategies against advanced PDACs (2). Particularly, in case of treatment failure, many of the patients with PDAC inevitably suffer from resistance to chemotherapy (3). The standard chemotherapy for advanced PDACs uses 5-fluorouracil (5-FU) or gemcitabine, that is, FOLFIRINOX (FFX) regimen (5-FU, leucovorin, irinotecan, and oxaliplatin) or GnP regimen (gemcitabine combined with paclitaxel albumin-bound nanoparticles; ref. 4). Recent comprehensive studies on advanced PDACs have shown the overall survival (OS) to be better among patients who received first-line FFX than in those receiving GnP (4, 5). However, the OS of patients receiving FFX is still less than a year (4). Therefore, development of a novel strategy to strengthen the effect of chemotherapy and/or overcome the chemoresistance is imperative.
Acquisition of resistance to chemotherapeutic drugs is caused by multiple mechanisms involving drug metabolism (6). For example, after administration of 5-FU, only a small fraction is converted into cytotoxic metabolites intracellularly, while 80% of 5-FU is catabolized to the non-cytotoxic 5-fluoro-5,6-dihydrouracil (DHFU) by the enzymatic activity of dihydropyrimidine dehydrogenase (DPD), which is encoded by the dihydropyrimidine dehydrogenase (DPYD) gene (7). There is a consistent relationship between the enzymatic activity of DPD and pharmacologic activity of 5-FU (8). Therefore, aberrant upregulation of DPD enzyme might be one of the key determinants of 5-FU chemoresistance in PDACs, as reported in different types of cancers (9, 10).
Long noncoding RNA (lncRNA) is a versatile and dynamic regulator of gene expression via interaction with DNA, RNA, and proteins (11, 12). Such interactions cooperatively promote cancer progression (13, 14). Recently, several lncRNAs have been identified as new regulators of chemotherapy resistance in various cancers (6, 15). For example, LINC00261 is involved in 5-FU sensitivity via regulation of DPYD promoter methylation in esophageal cancer (16). Therefore, targeting lncRNAs has recently been proposed as a novel cancer therapeutic strategy. We previously reported marked anti-tumor effects of depleting taurine upregulated gene 1 (TUG1) in gliomas (17). However, to date, no practical approach to modulate lncRNAs for overcoming chemoresistance or enhancing chemotherapeutic efficiency has been reported in cancer treatment.
In this study, we identified a new lncRNA, TUG1-based mechanism for regulation of 5-FU metabolism in PDACs. We provide strong evidence that the targeting of TUG1 by antisense oligonucleotides (ASO) coupled with our new cancer-specific drug delivery system (DDS; ref. 17), which can reduce or avoid the systemic adverse effects, is a novel potent therapeutic option for patients with PDAC, especially in combination with 5-FU–based chemotherapy.
Materials and Methods
Human PDAC tissues
PDAC (n = 49) and adjacent normal pancreatic tissues (n = 15) were obtained from 49 patients, who had undergone surgical resection at Nagoya City University Hospital, Nagoya, Japan (Japanese cohort). Patients with PDAC were dichotomized into a high TUG1 group and low TUG1 group according to median TUG1 expression as assessed by the number of spots in the RNA-FISH analysis. Of these patients, 12 patients had received either adjuvant 5-FU chemotherapy [tegafur/gimeracil/oteracil potassium (S-1); n = 6, tegafur/uracil (UFT); n = 4] or gemcitabine monotherapy (n = 2). PDAC (n = 148) and adjacent normal pancreatic tissues (n = 111) were also obtained from patients who had undergone surgical resection at Pancreas Center, The First Affiliated Hospital with Nanjing Medical University, Nanjing, China (Chinese cohort). We defined the group with TUG1 expression >2.0-fold the median level of adjacent normal tissues as high TUG1 group, while the others were defined as low TUG1 group. All patients received adjuvant 5-FU–based chemotherapy; S-1 (n = 122), S-1+albumin–bound paclitaxel particles (n = 1), S-1+oxaliplatin (n = 2), S-1 + gemcitabine (n = 20), FFX (n = 1), and capecitabine (n = 2). Kaplan–Meier survival curves for overall survival based on TUG1 expression (high group, n = 56; low group, n = 17 in the Chinese cohort) were generated for patients with PDAC with N1 and N2 whose survival data were fully available. Samples were collected after obtaining appropriate Institutional Review Board approval and written informed consent from the patients (approval number 60–19–0115 and 2017–0408–2 for Japanese and Chinese cohorts, respectively). Diagnosis of pancreatic cancer was determined histologically by experienced pathologists, based on the WHO classification of tumors of the digestive system. Disease-free survival (DFS) was defined as the term from the start of chemotherapy up to recurrence.
Cell lines and transfection treatment
BxPC-3 (ATCC), CFPAC-1 (ATCC), MIAPaCa-2 (ATCC), PK-9 (ATCC), PK-8 (RIKEN Cell Bank), and KLM-1 (RIKEN Cell Bank) cell lines were maintained in RPMI1640 medium (FUJIFILM Wako Pure Chemical) containing 5% FBS (Thermo Fisher Scientific) and 1% antibiotic–antimycotic (FUJIFILM Wako Pure Chemical). Panc-1 (ATCC) cell lines was maintained in DMEM (FUJIFILM Wako Pure Chemical) containing 5% FBS and 1% antibiotic–antimycotic. All of these cell lines were analyzed within 6 months after thawing. These cell lines were authenticated by short tandem repeat profiling by the JCRB Cell Bank and tested negative for Mycoplasma on thawing. The characteristics of these cell lines are shown in Supplementary Table S1 (18–20). Cells were transfected with 50 nmol/L ASO, 50 nmol/L siRNA, or 30 nmol/L miRNA precursor using Lipofectamine 3000 (Thermo Fisher Scientific), according to the manufacturer's instructions. The ASO, siRNA, and miRNA precursor used are listed in Supplementary Table S1.
RNA extraction and quantitative reverse transcription-PCR
RNA extraction and qRT-PCR were performed as described previously (14). Expression levels of target genes were normalized to that of GAPDH or 18S rRNA. Expression levels of target miRNAs were normalized to that of RNU6B. Oligonucleotide primers for TaqMan PCR assays and primer sets for SYBR Green assays are shown in Supplementary Table S1.
Drug sensitivity assay and synergism analysis
Cells were seeded onto 96-well plates at 1 × 104 cells/well and treated with 5-FU (titrated doses of 1 pmol/L, 1 nmol/L, 10 nmol/L, 1 μmol/L, 10 μmol/L, 100 μmol/L, 1 mmol/L, 10 mmol/L, 50 mmol/L or 10 mol/L; Sigma Aldrich), gemcitabine (1 pmol/L, 10 pmol/L, 100 pmol/L, 1 nmol/L, 10 nmol/L, 100 nmol/L, 1 μmol/L, 50 μmol/L, or 10 mol/L; FUJIFILM Wako Pure Chemical), or Gimeracil (DPD inhibitor, 1 μmol/L; Santa Cruz Biotechnology), for 48 hours. Cell viability was assessed using Cell Count Reagent SF (Nacalai Tesque). The assay was conducted in triplicate. Drug sensitivity was determined by IC50 value (half maximal inhibitory concentration) using GraphPad Prism 5 (GraphPad Software). Combination index values were calculated using CompuSyn software (ComboSyn; ref. 21).
RNA and miRNA expression analysis via microarray technology
RNA and miRNA microarrays were performed as described previously (17). The arrays were scanned using an Agilent Microarray Scanner (G2565BA, Agilent Technologies). The scanned images were analyzed using the Feature Extraction software, version 12.0 (Agilent Technologies) with background correction. Data analysis was performed with GeneSpring GX, version 7.3.1 (Agilent Technologies). Expression data were centered on a median using the GeneSpring normalization option. Experiments were performed in duplicate throughout the analysis.
Cells were fixed with freshly made 4% paraformaldehyde and blocked in PBS containing 2% BSA (01281–84, Nacalai Tesque) for 30 minutes at room temperature. Samples were immunostained with anti-DPD antibody (PA5–22302, 1:500, Thermo Fisher Scientific) for 24 hours at 4°C. Primary antibody–antigen complexes were visualized using anti-rabbit Alexa Fluor 488 secondary antibody (A11070, 1:500, Thermo Fisher Scientific) for 1 hour at room temperature. Nuclei were counterstained using 4′, 6-diamidino-2-phenylindole (DAPI, #4083, 1:1,000, Cell Signaling Technology). Images were obtained using a Leica DMI6000B microscope (Leica Microsystems). The integrated density of the DPD signal was measured using ImageJ software (22) using region-of-interest polygon selections. The total integrated density of each field was calculated. For hematoxylin–eosin staining, PDACs were fixed in 4% paraformaldehyde for 24 hours and washed in PBS. Fixed tissues were embedded in paraffin, sectioned, and stained with hematoxylin-eosin.
RNA-FISH was performed using ViewRNA ISH Cell Assay Kit and ViewRNA ISH Tissue Assay Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. The ViewRNA probe sets used were human TUG1 (#257362510, Thermo Fisher Scientific) and miR-376b-3p (#222432815, Thermo Fisher Scientific). Images were obtained with a Leica DMI6000 B microscope. TUG1 spots were counted by ImageJ software using polygonal region-of-interest. More than 100 nuclei were counted per field.
RNA-binding protein immunoprecipitation assay
RNA immunoprecipitation assay was performed using RiboCluster Profiler RIP-Assay Kit (MBL International), according to the manufacturer's instructions. RNA–protein complexes were immunoprecipitated with anti-AGO2 antibody (RN005M, MBL International), anti-IgG antibody being used as a negative control (PM035, MBL International). Immunoprecipitated RNA was analyzed by qRT-PCR.
Western blot analysis
Western blotting was performed as described previously (14). Fifty micrograms of each protein were taken for this analysis. The primary antibodies used were: rabbit polyclonal anti-DPD (PA5–22302, 1:1,000, Thermo Fisher Scientific), rabbit polyclonal anti-MIB1 (11893–1-AP, 1:1000, ProteinTech), mouse monoclonal anti-ENT1 (sc-377283, 1:1,000, Santa Cruz Biotechnology), mouse monoclonal anti-MRP5 (sc-376965, 1:1,000, Santa Cruz Biotechnology), rabbit polyclonal anti-TYMP (ab226917, 1:1000, Abcam, Cambridge, UK), rabbit monoclonal anti-TYMS (ab108995, 1:500, Abcam), and mouse monoclonal anti-β-actin (#3700, 1:1,000, Cell Signaling Technology). HRP-linked anti-mouse IgG (7076S, 1:1,000, Cell Signaling Technology) and HRP-linked anti-rabbit IgG (7074S, 1:1,000, Cell Signaling Technology) were used as secondary antibodies. The band intensity was quantified by ImageJ software.
Construction of the DPD and TUG1 expression vector
cDNA was prepared from total RNA extracted from BxPC-3, using the PrimeScript RT Master Mix (Takara). DPYD gene (XM_005270562.3) was amplified from the cDNA by PCR using KOD-Plus-Neo (TOYOBO). The primer sequences are shown in Supplementary Table S1. The TUG1 (NR_002323.2) expression vector was constructed by fusion of TUG1 exon 1, 2, and 3 fragments, as described previously (17). The DNA fragment was then cloned into pcDNA3.4 vector (Thermo Fisher Scientific). The DPYD and TUG1 sequence in the vector construct was validated by conventional sequencing analysis. BxPC-3, PK-9, and Panc-1 cells were seeded into 96-well plates at 1 × 104 cells/well and treated with 5-FU at different concentrations. The cells were then transfected using the TUG1 expression vector (10 ng) or the DPD expression vector (100 ng) by Lipofectamine 3000 for 48 hours. Cell viability was assessed using Cell Count Reagent SF. The assay was conducted in triplicate.
Measurement of compounds
After 48 hours of 5-FU treatment, we collected BxPC-3 and PK-9 cells and the culture supernatant. LC/MS-MS was used to measure 5-FU and its metabolites in BxPC-3, PK-9 cells, and the culture supernatant (23). Samples were processed by Acquity UPLC HSS T3 1.8 μm column (2.1 × 150 mm; Waters) in 0.1% formic acid as the mobile phase, and subjected to ultra-performance liquid chromatography-tandem mass spectrometry (ACQUITY UPLC-MS/MS; Waters). The multiple reaction-monitoring transitions were recorded [5-FU: m/z 131 > 114, 5-fluoro-dihydrouracil (DHFU): m/z 133 > 88, 2-fluoro-3-ureidoprppionate (FUPA): m/z 151 > 108, and thymine-d4: m/z 131 > 114]. The chromatographic data were processed using the MassLynx software (Waters).
Flow cytometry analysis
Cells were seeded into 6-well plates at 1 × 105 cells/well and transfected with ASO and siRNA for 72 hours. Cells were then collected and apoptosis assessed using Annexin V-FITC Apoptosis Detection Kits according to the manufacturer's instructions (Nacalai Tesque). The samples were analyzed on a Gallios flow cytometer (Beckman Coulter). Data were analyzed using FlowJo software 10.6.1 (BD Biosciences).
Xenograft mouse model of PDAC and treatment
Animal protocols were approved by the Animal Care and Use Committee of Nagoya University Graduate School of Medicine (approval number 20271). BxPC-3 cells (2 × 106/mouse) were injected subcutaneously into 6-week-old female NOD/SCID mice (Japan SLC). A cyclic Arg-Gly-Asp (cRGD) peptide–conjugated polymeric micelle was used as the DDS of ASOs in vivo (17, 24, 25). Two weeks after the injection, CTRL-DDS (1 mg/kg per day) or TUG1-DDS (1 mg/kg per day) were intravenously injected every 3 days for 24 days; 5-FU (20 mg/kg per day) was intraperitoneally injected every 3 days for 24 days (n = 3 each for 4 experimental groups: CTRL-DDS, TUG1-DDS, 5-FU and CTRL-DDS, and 5-FU and TUG1-DDS). The accumulation of DDS in tumor tissue was confirmed by an in vivo spectral imaging system (IVIS; IVIS Lumina II, Xenogen). Tumor volumes were calculated using the formula (W × W × L)/2 (W, width and L, length). PK-9 cells (3 × 106/mouse) were inoculated into the pancreatic tail of 6-week-old female NOD/SCID mice to establish an orthotopic xenograft model. Two weeks thereafter, CTRL-DDS (1 mg/kg per day) or TUG1-DDS (1 mg/kg per day) was intravenously injected every 3 days for 21 days; 5-FU (20 mg/kg per day) was intraperitoneally injected every 3 days for 21 days (CTRL-DDS; n = 3, TUG1-DDS; n = 5, 5-FU and CTRL-DDS; n = 4, 5-FU and TUG1-DDS; n = 5). After treatment, pancreatic tissue was harvested and evaluated for tumor volume and weight. Tumor volumes were calculated using the formula (W × W × L)/2.
Tissue for the patient-derived xenograft (PDX) was derived from a moderately differentiated PDAC tumor (KRAS mutated; Gly12Arg) obtained from a patient who underwent surgical resection at Cancer Institute, Japanese Foundation for Cancer Research. The tissue was collected after obtaining the appropriate institutional review board approval (approval number 2012–1001) and written informed consent of the patient. The tumor tissue was intradermally implanted into a 5-week-old female NSG mouse (Charles River Laboratories Japan). Two weeks after the injection, the PDX mice were treated every 3 days for 24 days (n = 3 each for 4 experimental groups: CTRL-DDS, TUG1-DDS, 5-FU, and 5-FU and TUG1-DDS). Tumor volumes were calculated using the formula (W × W × L)/2. Cy5-labeled TUG1-DDS (10 μg/mouse per day) was intravenously injected into PDX mouse models for 2 days. Tumor cells and blood vessels were stained by γ-glutamyl-trans-hydroxymethyl rhodamine green (gGlu-HMRG, MERCK) or Dextran-TRITC. Accumulation of DDS in the tumor and stroma was confirmed by in vivo confocal laser scanning microscopy (iCLSM; Nikon) after 2 and 6 hours of injection. Cy5 fluorescence intensity was obtained from at least three views in each tumor and stroma. Fluorescence intensities were measured using NIS Elements (Nikon).
Data are presented as the mean ± SEM. P values of statistical significance are indicated as *, P < 0.05; **, P < 0.01; ***, P < 0.001. Differences in the quantified data between groups were compared using one-way ANOVA, followed by a Bonferroni post hoc test, or t test. Two-way ANOVA was used for multiple comparisons in the anti-tumor activity assay. Fisher exact test was used to assess differences in clinical features of PDACs. All reported P values were two-sided. Statistical analyses were performed using GraphPad Prism 5 software and Excel 2016.
The microarray data have been deposited in the Genomic Expression Archive (GEA) under accession numbers E-GEAD-367, E-GEAD-368, E-GEAD-369, and E-GEAD-371. The PDAC data were derived from The Cancer Genome Atlas (TCGA; https://portal.gdc.cancer.gov/) and Genotype Tissue Expression (GTEx; https://gtexportal.org/home/) through Gene Expression Profiling Interactive Analysis (GEPIA; http://gepia.cancer-pku.cn/; ref. 26).
TUG1 overexpression is related to poor prognosis in patients with PDAC who underwent 5-FU–based chemotherapy
RNA-FISH analysis in 49 PDACs revealed the mean number of TUG1 spots to be > 2.0-fold upregulated in PDACs compared to that in adjacent normal pancreatic tissues (n = 15; P < 0.001, Fig. 1A and B, Japanese cohort). The 49 PDAC patients were dichotomized into a high TUG1 group (n = 25) and a low TUG1 group (n = 24) according to whether the number of TUG1 spots counted was above or below the median (Supplementary Tables S2 and S3). Overexpression of TUG1 was observed in PDAC cells, but not in stromal cells. The expression levels of TUG1 did not significantly correlate with tumor differentiation, stage, tumor extension, and spread to nearby lymph nodes of PDACs (Supplementary Table S2). Interestingly, low TUG1 patients tended to show longer DFS than high TUG1 patients after adjuvant 5-FU (S-1, UFT) or gemcitabine chemotherapy (median DFS, 499 days and 254.5 days, respectively, P = 0.06; Fig. 1C; Supplementary Table S3).
To further investigate whether TUG1 expression correlated with the effectiveness of chemotherapy in clinical PDACs, we examined another independent cohort of 148 PDACs, who underwent adjuvant 5-FU–based chemotherapy, along with 111 adjacent normal pancreatic tissues (Chinese cohort). qPCR analysis showed 77% of PDACs expressed > 2.0-fold higher TUG1 than the adjacent normal tissues (Fig. 1D). These 148 patients with PDAC were also divided into a high TUG1 group (n = 115) and a low TUG1 group (n = 33). Levels of TUG1 expression were not apparently different across each PDAC characteristic, consistent with the Japanese cohort (Supplementary Table S2). However, among the patients with advanced PDAC (i.e., lymph node metastasis-positive), OS in high TUG1 group was significantly shorter than that in low TUG1 group (log rank, P < 0.05; HR, 2.53; 95% confidence interval (95% CI), 1.05–6.10; Fig. 1E; Supplementary Table S3).
Analysis of PDACs (n = 179) from the public databases TCGA and GTEx showed the expression of TUG1 in PDACs is significantly upregulated compared with that in normal pancreatic tissues (P < 0.05, Supplementary Fig. S1A). The higher expression of TUG1 in PDACs was correlated with poor prognosis (log-rank test, P < 0.05; HR, 1.6; 95% CI, 1.01–2.69, Supplementary Fig. S1B).
TUG1 inhibition improved 5-FU sensitivity
Differential outcomes in PDACs, between high and low TUG1 groups, after chemotherapy, inspired us to examine the effects of TUG1 on chemosensitivity. Significant positive correlations between the expression levels of TUG1 and the IC50 of 5-FU (r = 0.88, P < 0.01), but not of gemcitabine (r = -0.10, P = 0.81), were observed in seven pancreatic cancer cell lines whether or not they harbored KRAS mutations, and independent of differentiation status (Fig. 2A and B; Supplementary Table S1). Depletion of TUG1 by specific antisense oligonucleotides (TUG1-ASO), which had been validated previously (Supplementary Fig. S2; ref. 17), dramatically sensitized the resistant cell lines (BxPC-3 and PK-9) to 5-FU (IC50, 1,185 μmol/L to 3.2 μmol/L, and 26.3 μmol/L to 3.8 μmol/L, respectively), whereas it minimally affected Panc-1, which was already 5-FU sensitive and expressed a relatively lower level of TUG1 (IC50, 7.9 μmol/L to 4.5 μmol/L; Fig. 2A and C). Depletion of TUG1 also moderately sensitized PDAC cells to gemcitabine, as seen in BxPC-3, PK-9, and Panc-1 cells, regardless of TUG1 expression (IC50, 219 nmol/L to 66 nmol/L, 49.6 nmol/L to 22.1 nmol/L, and 334 nmol/L to 155 nmol/L, respectively; Fig. 2D). Combination index (21) revealed TUG1 depletion synergistically enhanced the effects of both chemotherapeutic drugs, especially 5-FU, in BxPC-3 and PK-9 (combination index, 0.0002 and 0.007, respectively; Fig. 2E).
TUG1 repressed the expression of miR-376b-3p
To explore the mechanism by which TUG1 affected chemosensitivity, especially to 5-FU, we examined the miRNA-sponge effect of TUG1 using miRNA microarray analysis (13, 17). Forty-three miRNAs were upregulated by > 2.0-fold after TUG1 depletion by TUG1-ASO in BxPC-3 in vitro (TUG1-depleted BxPC-3). Furthermore, we had subcutaneously inoculated BxPC-3 cells into mice, and treated them intravenously with TUG1-ASO coupled with a cancer-specific DDS: an Arg-Gly-Asp (cRGD) peptide-conjugated polymeric micelle (TUG1-DDS; n = 4; 1 mg/kg per day, every 3 days for 24 days; refs. 17, 24, 27). Twelve miRNAs were upregulated by > 2.0-fold in BxPC-3 tumors in vivo by TUG1-DDS treatment. Finally, six miRNAs were found to be commonly upregulated both in vitro and in vivo (Fig. 3A; Supplementary Table S4).
Among the six miRNAs, miR-376b-3p and miR-376a-3p, both having the same seed sequence, were the first and second most upregulated miRNAs (Supplementary Table S4). Upregulation of miR-376b-3p by TUG1 depletion was further verified by qPCR and RNA-FISH analyses (Fig. 3B).
Correlation coefficient analysis using TCGA data revealed a negative relationship between TUG1 and miR-376b-3p expression in PDACs (r = −0.209, P < 0.01; Fig. 3C). PDACs with low miR-376b-3p levels showed worse prognosis than those with high miR-376b-3p levels (log-rank, P < 0.05; HR, 0.45; 95% CI, 0.21–0.98; Fig. 3D). The expression of miR-376a-3p also showed a negative correlation with that of TUG1 in TCGA data (r = −0.15, P < 0.05). Low expression of miR-376a-3p is associated with worse prognosis in PDACs than those with high miR-376a-3p levels (log-rank, P < 0.13; HR, 0.64; 95% CI, 0.35–1.14; Supplementary Fig. S3A and S3B).
miR-376b-3p is one of the 268 miRNAs that interact with TUG1, as revealed by a large-scale cross-linking immunoprecipitation (CLIP) coupled with sequencing analysis (starBase v2.0; ref. 28). The seed sequence of the miR-376b-3p-binding region was found in TUG1 (Fig. 3E). RNA immunoprecipitation assay, using antibodies against argonaute 2 protein (AGO2), revealed the interaction of AGO2 with both TUG1 and miR-376b-3p in BxPC-3 and Panc-1 (Fig. 3F).
TUG1 increased DPD expression by repressing miR-376b-3p expression
To identify target genes regulated by TUG1 via miR-376b-3p, gene expression microarray analysis was conducted using RNA isolated from the aforementioned TUG1-depleted BxPC-3 cells. We found 214 genes to be commonly downregulated by > 2.0-fold by TUG1 depletion, both in vitro and in vivo (Fig. 4A). Among the 214 genes, 34 were potential target genes of miR-376b-3p as per in silico analysis by miRWalk 2.0 (Fig. 4A; ref. 29). Among the 34 candidate genes, the expression levels of 16 were substantially higher in PDACs than in normal pancreatic tissues, and were positively correlated with TUG1 expression in TCGA and GTEx database (Supplementary Table S5; Supplementary Fig. S4A). Among the 16 genes, the most significantly repressed gene by TUG1 depletion was DPYD (Supplementary Table S5). PDACs with high DPYD expression showed significantly worse prognosis (TCGA, log rank, P < 0.01; HR, 2.2; 95% CI, 1.35–3.68; Supplementary Fig. S4B). DPYD expression was positively correlated with TUG1 expression (r = 0.24, P < 0.01), and inversely correlated with miR-376b-3p expression (r = -0.208, P < 0.01; Fig. 4B and C). Consistent with the TCGA database, significant positive correlations between the level of expression of TUG1 and DPYD were observed in the seven pancreatic cancer cell lines tested (r = 0.75, P < 0.05; Supplementary Fig. S4C).
Depletion of TUG1 and overexpression of miR-376b-3p significantly repressed both DPYD mRNA and DPD protein expression in both BxPC-3 and PK-9 cells, which are resistant to 5-FU and show high TUG1 and DPYD expression (Fig. 4D–F). Neither DPYD mRNA nor DPD protein expression was detected in Panc-1 (Fig. 4D and F). Notably, overexpression of miR-376b-3p sensitized BxPC-3 and PK-9 to 5-FU (Fig. 4G).
IHC analysis demonstrated that the expression of DPD increased in PDACs compared with that in adjacent normal pancreatic tissues (Fig. 4H). The number of TUG1 RNA-FISH spots and fluorescence intensity of DPD were significantly correlated with each other (r = 0.56, P < 0.05; Fig. 4I; Supplementary Table S6).
PDACs acquired 5-FU resistance via TUG1-miR-376b-3p-DPD axis
Depletion of DPD efficiently reduced the IC50 of 5-FU in BxPC-3 and PK-9, but not in Panc-1 (Fig. 5A; Supplementary Fig. S5A). In addition, the reduced IC50 of 5-FU due to TUG1 depletion was partially rescued by the overexpression of DPD in BxPC-3 and PK-9 (Fig. 5B; Supplementary Fig. S5B). Notably, overexpression of TUG1 decreased 5-FU sensitivity (IC50, 0.82 μmol/L to 2.5 μmol/L) via downregulation of miR-376b-3p together with an increased expression of DPD in Panc-1 cells (Supplementary Fig. S5C–S5F). In contrast, neither depletion of TUG1 nor overexpression of miR-376b-3p changed the mRNA and protein levels of equilibrative nucleoside transporter 1 (ENT1), multidrug resistance-associated protein 5 (MRP5), thymidine phosphorylase (TYMP) or thymidylate synthase (TYMS), all of which are involved in 5-FU metabolism and confer 5-FU resistance (Supplementary Fig. S6A and S6B; ref. 30).
DPD catabolizes 5-FU to an intermediate metabolite, DHFU, which is consequently degraded into FUPA and α-fluoro-β-alanine (FBAL) through a degradation cascade (Fig. 5C; refs. 31, 32). We directly measured the concentration of 5-FU and its metabolites in TUG1-depleted BxPC-3 and PK-9 cells using LC/MS-MS. After 48 hours of TUG1 depletion, the level of intracellular 5-FU became higher than that in control (Fig. 5D). DHFU concentration was below detectable levels in either condition. While intracellular concentration of FUPA was below detectable levels in either condition, it was significantly lower in the TUG1-depleted samples of cell culture supernatants, compared with that in control, indicating the delayed conversion of 5-FU after TUG1 depletion (Fig. 5D). We then co-treated BxPC-3 and PK-9 with 5-FU and DPD inhibitor (33). Interestingly, treatment with 1 μmol/L DPD inhibitor partially enhanced the efficacy of 5-FU, though not as effectively as TUG1 depletion did (Fig. 5E).
Combination therapy with 5-FU and TUG1-DDS suppressed tumor growth
We examined the effect of combination therapy with 5-FU and TUG1-ASO in PDAC xenograft mouse model in vivo. Mice bearing BxPC-3 xenograft tumors were intravenously treated with TUG1-DDS and/or 5-FU every 3 days for 24 days (Fig. 6A). IVIS Imaging confirmed the accumulation of TUG1-DDS labeled with Alexa-647 in the tumor (Fig. 6B). TUG1-DDS significantly suppressed tumor growth compared with CTRL-DDS and 5-FU (Fig. 6C). Remarkably, combination therapy using 5-FU and TUG1-DDS further suppressed the tumor growth than TUG1-DDS alone (Fig. 6C). Reduced expression of TUG1 and DPD, along with an increased expression of miR-376b-3p, were observed in the TUG1-DDS–treated tumors (Fig. 6D–F). We further examined the effect of combination therapy using orthotopic xenograft mouse models with PK-9 cells. Consistent with results from the subcutaneous xenograft mouse model, combination therapy significantly suppressed tumor growth compared with 5-FU or TUG1-DDS alone (Supplementary Fig. S7A–S7F). No apparent adverse effects were observed in any animals.
TUG1-DDS was also administered to the clinically relevant PDX mouse model, which was derived from a moderately differentiated PDAC tumor (KRAS mutated; Gly12Arg) obtained from a patient. Using iCLSM imaging system, we found TUG1-DDS to accumulate in tumor cells, although some of the TUG1-DDS molecules were trapped by stromal cells after 2 hours of intravenous injection (Fig. 6G). Combination therapy resulted in significantly greater suppression of tumor growth and decrease of TUG1 expression than CTRL-DDS, TUG1-DDS, or 5-FU alone (Fig. 6H and I). Taken together, our data indicated that DDS effectively and specifically delivered TUG1-ASO to PDAC cells, despite the cells being surrounded by thick stromal cell layers.
In this study, we demonstrated TUG1-miR376b-3p-DPD axis to be frequently dysregulated in PDACs, causing acquisition of 5-FU resistance via enhancement of 5-FU catabolism. Intravenous treatment with 5-FU plus TUG1-ASO, coupled with cancer-specific DDS, efficiently repressed PDAC growth in vivo. Over the past decades, studies have focused on the control of both toxicity and efficacy of 5-FU due to the dysregulation of drug-metabolizing enzymes (34). 5-FU is first converted to 5-fluorodeoxyuridine (FUdR) by TYMP. FUdR is subsequently metabolized into 5-fluorodeoxyuridine monophosphate (FdUMP), which inhibits TYMS activity, resulting in cytotoxicity (31). Thus, cellular levels of FdUMP determine TYMS activity, both of which affect 5-FU efficacy (35). Toxicity of 5-FU can be also caused by deleterious polymorphisms in the DPYD gene, which result in the accumulation of 5-FU, whereas an increased expression of DPD in tumors usually results in resistance to 5-FU (34, 36, 37). To increase the efficacy of 5-FU, S-1 was designed to continuously release 5-FU in combination with potent DPD modulator components (38, 39). However, although S-1 is effective for some patients with PDAC, its effect is limited (40). Our approach using TUG1-DDS as an effective modulator to decrease the catabolic activity of DPD and meet appropriate 5-FU dosage requirements in targeted PDAC cells may provide a new layer of combination strategy for effective cancer treatment.
Previous studies revealed that overexpression of TUG1 in gastrointestinal cancers with poor prognosis and chemoresistance (41) and that this also plays an important role in the progression of PDACs (42, 43). In this study, we add a novel role of TUG1 in the acquisition of 5-FU resistance in PDACs. Functional experiments showed TUG1 depletion enhanced sensitivity to 5-FU, via increased cellular concentration of miR-376b-3p, which in turn efficiently targets DPYD, in the 5-FU–resistant PDAC cells. Previous studies have shown miR-27a, miR-27b, miR-134, and miR-582–5p, which are identified as DPYD-targeting miRNAs (44), to be overexpressed while the protein level of DPD is reduced, in a pancreatic cancer cell line MIAPaca-2. In contrast, miR-27b and miR-134 levels were significantly lower in lung cancers (44). In colon cancer cells, miR-494 targets DPYD by binding to the 3′ untranslated region of DPYD (45). miR-302b enhances the sensitivity of hepatocellular carcinoma cells to 5-FU by targeting Mcl-1 and DPYD (46). These studies collectively indicated DPD protein expression to be regulated by different miRNAs in a tissue-specific manner. Our microarray analysis did not detect these miRNAs as a posttranscriptional regulator of DPYD.
Notably, expression levels of enzymes and transporters such as ENT1, MRP5, TYMP, and TYMS, which had been previously reported to determine the antitumor activity of 5-FU (30, 35, 47, 48), were not changed by TUG1 depletion, indicating that TUG1 regulates 5-FU sensitivity via the miR376b-3p-DPD axis in PDACs.
Interestingly, TUG1 depletion caused synergism with Gem, although not as efficiently as with 5-FU. We identified mind bomb 1 (MIB1), a multi-domain RING-type E3 ligase that sustains Notch activity (49, 50), as another profoundly activated target gene of the TUG1-miR376b-3p axis in both in vitro and in vivo experiments (Supplementary Table S5; Supplementary Fig. S8A–S8J). Activated Notch signal promotes tumor growth and causes resistance to gemcitabine by alteration of the apoptotic pathway in PDACs (51, 52). Our functional experiments showed that both TUG1 depletion and miR-376b-3p precursor significantly suppressed cell proliferation by decreasing MIB1 expression, thereby suggesting TUG1 depletion to possibly enhance the efficacy of gemcitabine by reactivating miR-376b-3p against MIB1 (Supplementary Fig. S8A–S8J). Furthermore, these additional effects of the TUG1-miR376b-3p axis may explain why treatment with 5-FU along with TUG1 depletion reduced PDAC cell viability more efficiently than 5-FU plus DPD inhibitor.
Pancreatic cancer is characterized by stromal layers surrounding tumor cells, which is associated with drug resistance, impeding drug delivery by physically blocking the cytotoxic chemotherapeutics (53–55). To reduce TUG1 expression in a mouse xenograft model, we used cRGD ligand-conjugated polyion complex micelles, which have a well-defined size of 35 nm in diameter (24, 25). The micelles acquire the ability to remain in the bloodstream due to their polyethylene glycol coating. RGD peptides conjugated on the micellar surface are ligand molecules for targeting αvβ3 and αvβ5 integrins, which are frequently overexpressed in pancreatic cancer cells (27, 56). These targetable polymeric micelles retained ASO accumulation within PDAC tumors. Consequently, high cellular uptake of micelles via endocytosis and transcytosis-mediated penetration was observed, along with enhanced gene-silencing activity (17, 24). Indeed, TUG1-DDS successfully accumulated in tumors and exerted significant antitumor activity in PDAC PDX tumors, which were surrounded by a thick stroma. Although further investigation is required, TUG1-DDS is a powerful strategy for targeting PDACs via facilitated ASO delivery even beyond the stroma.
In conclusion, our data provided a strong and novel rationale for targeting TUG1 as a practical therapeutic approach to treat PDACs. Depletion of TUG1 acted as a potential modulator to decrease the catabolic activity of DPD, specifically in PDAC cells. Treatment with TUG1-DDS, which can reduce or avoid the systemic adverse effects, combined with 5-FU–based chemotherapy, might be a promising therapeutic approach against PDACs, particularly those with resistance to 5-FU.
No disclosures were reported.
Y. Tasaki: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. M. Suzuki: Conceptualization, software, formal analysis, supervision, visualization, writing–original draft. K. Katsushima: Conceptualization, formal analysis, funding acquisition, investigation. K. Shinjo: Resources, supervision, validation, visualization. K. Iijima: Supervision, validation, methodology. Y. Murofushi: Data curation, formal analysis, investigation, visualization, methodology. A. Naiki-Ito: Resources, data curation, methodology. K. Hayashi: Resources, data curation, methodology. C. Qiu: Resources, data curation, formal analysis, validation. A. Takahashi: Resources, data curation, software, formal analysis, funding acquisition, investigation, visualization, methodology. Y. Tanaka: Resources, data curation, software, visualization, methodology. T. Kawaguchi: Resources, data curation, software, visualization, methodology. M. Sugawara: Resources, data curation, software, visualization, methodology. T. Kataoka: Data curation, formal analysis, investigation, methodology. M. Naito: Resources, data curation, visualization. K. Miyata: Resources, data curation, formal analysis, visualization. K. Kataoka: Conceptualization, resources, data curation. T. Noda: Resources, data curation, investigation, visualization, methodology. W. Gao: Resources, data curation, software, formal analysis, supervision, investigation. H. Kataoka: Resources, data curation, investigation, methodology. S. Takahashi: Resources, data curation, validation, methodology. K. Kimura: Resources, data curation, validation. Y. Kondo: Conceptualization, data curation, supervision, funding acquisition, validation, investigation, writing–original draft, project administration, writing–review and editing.
This study was performed as a research program of P-CREATE, Japan Agency for Medical Research and Development (19cm0106108h0004, Y. Kondo; 19cm0106202s0404, M. Suzuki; 17cm0106202h0002, A. Takahashi), the Grant-in-Aid for Scientific Research, the Japan Society for the Promotion of Science (17H03582, 20H03511, Y. Kondo), JST/PRESTO grant (JPMJPR17H7, A. Takahashi), and of the Research Grant of the Princess Takamatsu Cancer Research Fund (15-24712, Y. Kondo).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.