The MiR-135b–BMAL1–YY1 loop disturbs pancreatic clockwork to promote tumourigenesis and chemoresistance

Subsequently, RT-PCR and western blotting revealed that miR-135b mimics markedly reduced the mRNA and protein levels of BMAL1, whereas miR-135b inhibitors increased BMAL1 expression in PC cells (Fig. 2e). The expression levels of miR-135b and BMAL1 in cohort 1 were negatively correlated (55 cases, Pearson r = –0.481, P < 0.001; Fig. 2f), and this result was further validated by bioinformatics analysis using high throughput RNA-sequencing data from The Cancer Genome Atlas (TCGA) PC cohort (178 cases, Pearson r = –0.251, P < 0.001; Fig. 2g, Supplementary Table S5). As the BMAL1 gene oscillates rhythmically in vivo, we explored whether miR-135b also harbours self-sustained oscillation by assessing its temporal pattern in HPDE6c7 and Panc-1 cells at 4-h intervals over a 24-h period. As shown in Fig. 2h, in both the synchronised normal and malignant pancreatic cells, the relative abundance of miR-135b was marked by rhythmic variations (P < 0.05) that were roughly in anti-phase with those of BMAL1. Collectively, these findings confirmed that miR-135b is a BMAL1-targeting miRNA in PC.

As the circadian clock targets CCGs to regulate cellular processes, we next investigated the influence of miR-135b on cancer-related CCGs^12,26. Here, we showed that miR-135b overexpression in MIA PaCa-2 cells significantly altered the expression of clock-controlled cell cycle checkpoints (Supplementary Fig. S3a), the clock-controlled DNA repair system (Supplementary Fig. S3b) and clock-controlled apoptosis (Supplementary Fig. S3c), leading to enhanced cell cycle progression and the inhibition of cell apoptosis; however, the restoration of BMAL1 expression largely reversed these changes by rescuing the local circadian gating control.

Then, we dissected the mechanism of GEM resistance. Flow cytometry analysis revealed that miR-135b significantly reduced the fraction of apoptotic MIA PaCa-2 cells during GEM treatment, whereas silencing miR-135b in Panc-1 cells led to a sharp increase in programmed cell death. Specifically, the modulation of BMAL1 expression was found to be effective against this miR-135b-mediated anti-apoptotic activity (Fig. 6g–i). We further measured the expression of apoptotic biochemical markers. After a 72 h-exposure to GEM, cleaved caspase-3 and PARP were significantly increased in BMAL1-upregulated MIA PaCa-2 cells and in miR-135b-downregulated Panc-1 cells, compared with their respective controls, which presented with mainly full-length and inactivated caspase-3 and PARP proteins (Fig. 6j). Overall, our findings demonstrated that dysregulation of the miR-135b–BMAL1 axis confers GEM resistance to PC cells.

On the other hand, genomic analysis identified two potential YY1-binding motifs inside the hsa-miR-135b promoter region (Fig. 7d). Therefore, we designed corresponding primer sets covering the two sites and performed a chromatin immunoprecipitation (ChIP) assay in PC cells with or without YY1 overexpression. As a result, YY1 protein was recruited to both binding sites, with the majority of YY1 bound to site 2 (Fig. 7e). Then, PGL3 vectors containing the wild-type or mutant putative binding sites of the miR-135b promoter were constructed and transfected into MIA PaCa-2 and Panc-1 cells. We found that ectopic YY1 expression significantly increased the luciferase activity of the wild-type miR-135b promoter; however, when mutations were present in binding site 1 or site 2, this enhancement was abrogated (Fig. 7f). These data indicated that YY1 activated miR-135b by directly binding to its promoter. Also, we found that the level of miR-135b was significantly upregulated by YY1 overexpression and was downregulated by YY1 silencing (Fig. 7g). In PC samples from two independent cohorts of patients, similar results were obtained showing that YY1 was positively correlated with miR-135b expression (cohort 1: 55 cases, Pearson r = 0.502, P < 0.001; Fig. 7h; TCGA cohort: 178 cases, Pearson r = 0.173, P = 0.021; Fig. 7i). Taken together, these findings suggested that YY1 transcriptionally activates miR-135b and forms a feedback loop with miR-135b–BMAL1 signalling.

Next, we explored the relevance of miR-135b–BMAL1–YY1 expression and the tumour response to chemotherapy. Thirty-seven eligible patients (cohort 2) that had been diagnosed with advanced PC and that had received GEM or GEM-based regimens as first-line chemotherapy were enroled for investigation (Supplementary Table S9). The OS and progression-free survival (PFS) curves showed that high miR-135b and low BMAL1 expression were significantly correlated with poor survival outcomes (Fig. 8d). Moreover, GEM-based treatment resulted in an objective response rate (ORR) of 16.2% and a disease control rate (DCR) of 62.2%; high miR-135b/low BMAL1/high YY1 expression facilitated GEM resistance, whereas those patients with low miR-135b/high BMAL1/low YY1-expressing tumours had relatively higher chemosensitivity (Supplementary Table S10). Thus, on the basis of the above findings, we speculated that identifying miR-135b–BMAL1–YY1 signalling activity may be useful for predicting GEM responsiveness.

Four human PC cell lines (MIA PaCa-2, AsPC-1, SW1990 and Panc-1) and HEK 293T cells were obtained from the American Type Culture Collection (ATCC). The immortalised human pancreatic duct epithelial cell line HPDE6c7 was acquired from Kyushu University, Japan. Cells were authenticated by short tandem repeat profiling and were cultured according to the manufacturer’s protocols. The cell lines were carefully checked for morphological consistency and for mycoplasma contamination using the Cycleave PCR Mycoplasma Detection Kit (TaKaRa). Prior to the experiments, the cells were serum-shocked with 50% horse serum (Gibco) for 2 h to achieve cellular synchronisation⁴⁷. Cell transfection was performed with Lipofectamine 2000 reagent (Invitrogen) in Opti-MEM (Gibco), according to the manufacturer’s instructions. Oligonucleotides used in this study were all purchased from GenePharma (Shanghai, China).

Luciferase constructs were generated by ligating oligonucleotides containing the wild-type or mutated putative target site of the BMAL1 3′-UTR into the pmirGLO vector (Promega) downstream of the luciferase gene; and the hsa-miR-135b promoter sequence with wild-type or mutated putative YY1-binding sites was amplified from human genomic DNA and cloned into the pGL3 vector (Promega). To generate miR-135b expression or knockdown (Anti-miR-135b) vectors, oligonucleotides encoding precursors or inhibitors of miR-135b were synthesised and subcloned into the BamHI and XhoI restrictive sites of pPG/miR/EGFP/blasticidin plasmid or pGCMV/EGFP/miR/blasticidin plasmid (GenePharma), and then verified by DNA sequencing. Empty vectors were used as controls. The expression sequence of miR-135b was as follows: 5′-TATGGCTTTTCATTCCTATGTGA-3′. The silencing sequence of miR-135b was as follows: 5′-TCACATAGGAATGAAAAGCCATA-3′. Stably transfected cells were selected by blasticidin (Invitrogen), and the transfection efficiencies were examined by PCR. Lentiviral vectors encoding human BMAL1 (Lv-BMAL1) or YY1 or expressing short hairpin RNA against BMAL1 (shBMAL1) or YY1 and the pcDNA 3.1 empty vectors (pcDNA) and plasmids carrying scrambled shRNA (Scramble) were all constructed as previously described^19,29,48.

Cells seeded in 96-well plates with a confluence of approximately 80% were transfected with luciferase reporter plasmids, pRL-TK-Renilla luciferase vectors (Promega), and the indicated RNA or YY1 expression constructs by Lipofectamine 2000 (Invitrogen). Forty-eight hours after transfection, firefly and Renilla luciferase activities were measured using a Dual-Luciferase Reporter Assay Kit (Promega).

Total RNA was extracted from cells or tissue specimens using TRIzol reagent (Invitrogen) and was then reverse-transcribed using a Transcriptor First Strand cDNA Synthesis Kit (Roche) for mRNAs or a One-Step Hairpin-it miRNAs qRT-PCR Quantification Kit (GenePharma) for miRNAs. The resulting cDNAs were used as templates for quantitative real-time PCR amplification. Relative RNA expression was evaluated using the comparative CT method and was normalised to U6 snRNA or GAPDH. Supplementary Table S11 shows the primer sequences used in this study.

Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer and proteins were then extracted and quantified using the bicinchoninic acid assay (BCA) kit (Beyotime Biotechnology, Jiangsu, China). Whole-cell lysates were separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) with 10% gels and were transferred to polyvinylidene difluoride membranes (Millipore). Membranes were incubated overnight with specific antibodies. Signals were visualised using ECL detection reagent (Beyotime). All antibodies were obtained from Abcam and Santa Cruz. β-actin was used as the internal control.

Cell proliferation was monitored using the Cell Counting Kit-8 (CCK-8) (Dojindo) according to the manufacturer’s instructions, and cell numbers were calculated based on relative absorbance at 450 nm. Colony formation assays were performed to assess long-term cell proliferation in vitro¹⁹.

To determine cell viability, cells were seeded in 96-well plates at an initial density of 3 × 10³ cells per well and incubated with gemcitabine (GEM) (Sigma-Aldrich, cat. no. 1288463) at various concentrations for 72 h. Then, 10 μL of CCK-8 solution was added into each well and incubated for 2 h before measurement. The IC₅₀ value refers to the GEM concentration that produced 50% of the maximum cell death.

In vitro wound-healing assays were performed to determine migration ability. Cells were cultured on 6-well plates forming a single cell layer, and then a straight scratch was made with a sterile pipette tip in the middle of the cell layer. Twenty-four hours later, cells that had migrated into the wound line were observed and the percent wound closure was calculated for five randomly chosen fields. Cell invasive ability was assessed by transwell assay in accordance with our previous report¹⁹.

Cells seeded in 6-well culture plates were subjected to different doses of GEM treatment for 72 h. Then, cells were collected, washed with PBS, and stained with Annexin V-FITC/ PI (KeyGEN Biotech) in the dark for 15 min before assessment by flow cytometry (FACS Calibur, BD Biosciences).

Twenty-four male Balb/c nude mice (4–5-weeks old, 18–20 g) were purchased from the Laboratory Animal Center, Chinese Academy of Sciences (Shanghai, China). Mice were maintained under a 12/12 h light/dark cycles with lights turning on at 8:00 am and off at 8:00 pm. All animal-related procedures were performed according to the ethical guidelines of the Institutional Animal Care and Use Committee of Shanghai Jiao Tong University. Mice were randomly divided into eight groups (n = 3 per group). Equal amounts of the indicated cells (2.5 × 10⁶) were subcutaneously implanted into the right back-side of each mouse. GEM treatment began after the tumours reached a volume of ~100 mm³. GEM was intraperitoneally injected at a dose of 100 mg/kg at 9:00 am every 3 days for a total of six times. Tumour volume was monitored every 4 days. Mice were then killed, and the xenografts were excised and weighed.

ISH was performed on paraffin-embedded samples as previously described⁴⁹. Sensitivity enhanced ISH kits (MK10301, Boster Biological Technology, Wuhan, China), and LNA-modified and DIG-labelled miR-135b probes (miRCURY LNA Detection probe, Exiqon) were used. The 5ʹ–3ʹ sequence was TCACATAGGAATGAAAAGCCATA, with digoxin labels at the 5ʹ and 3ʹ ends. Hybridisation, washing and scanning were carried out according to the manufacturer’s instructions. The intensity of miR-135b staining was scored according to the following standards: 0–1 (no staining), 1–2 (weak staining), 2–3 (medium staining) and 3–4 (strong staining). The percentage of miR-135b-expressing cells in three high-power fields of each individual sample was analysed. The expression scores were calculated as the intensity score × the percentage of positive cells. Individual samples were evaluated by two pathologists. The samples with expression scores greater than or equal to 2 were defined as high expression, and those with scores <2 had low expression. IHC analysis and the evaluation criteria were conducted according to previous methods¹⁹.

ChIP analysis was performed as described in our former report¹⁹. Briefly, MIA PaCa-2 and Panc-1 cells infected with YY1 expression vectors or the negative controls were cross-linked using 1% formaldehyde and lysed in SDS lysis buffer. Cell lysates were sonicated and then mixed with ChIP dilution buffer and protein A-Agarose/salmon sperm DNA (Millipore). After centrifugation, the supernatants were collected and divided into two parts: one part was for YY1 antibody detection (ab12132, Abcam), and the other part was for IgG reactivity. The next day, immune complexes were precipitated and rinsed. Immunoprecipitates were pelleted by centrifugation and incubated at 65 °C to reverse the protein-DNA cross-links. RNase A and proteinase K were then added, in turn, to recycle the DNA fragments. Purified DNA samples were dissolved in ddH₂O and subjected to PCR analysis.

A total of 92 patients with pathologically diagnosed PC and who had undergone operations, laparoscopic biopsies, or EUS-FNAs at Shanghai General Hospital and Zhejiang Province People’s Hospital were enroled in this study. The patients were divided into two independent cohorts. Cohort 1 included 55 pairs of PC and control normal pancreatic (NC) tissues collected between 2014 and 2016, with both fresh and paraffin-embedded specimens; cohort 2 included 37 paraffin-embedded PC samples, which were obtained between 2009 and 2014 with full follow-up data. None of these patients had previously received anticancer therapy. The study protocol was approved by the ethics committee of Shanghai General Hospital and the ethics committee of Zhejiang Province People’s Hospital. All human materials were obtained with informed consent, and all research was carried out in accordance with the provisions of the Helsinki Declaration of 1975.

Follow-up and prognostic studies were conducted in cohort 2 PC patients who had been diagnosed with unresectable advanced PC and had received GEM or GEM-based regimens as first-line chemotherapy. A total of 37 patients were eligible for inclusion, and detailed information is provided in Supplementary Table. The criteria for enrolment were as follows: histologically confirmed advanced PC (AJCC Stage III/IV); no previous chemotherapy or other anti-tumour therapies; life expectancy ≥2 months; GEM-based regimens for at least two cycles; ECOG performance status <2; and adequate haematologic, hepatic, and renal function. Chemotherapy outcomes were evaluated by computed tomography or magnetic resonance imaging every two cycles. The objective tumour response for target lesions was classified as complete response (CR), partial response (PR), stable disease (SD) or progressive disease (PD), according to the Response Evaluation Criteria in Solid Tumour (RECIST) version 1.1. The endpoints included ORR, DCR, OS and PFS. The OS and PFS times were calculated from the date that first-line chemotherapy treatment was started to the date of mortality and disease progression, respectively. S9

Three independent public data sets of human genome arrays for PC and normal pancreatic tissues were downloaded from GEO (http://www.ncbi.nlm.nih.gov/geo/↗). The GSE19650↗ data set consisted of 15 PC and 7 NC, GSE32676↗ consisted of 25 PC and 7 NC, and GSE16515↗ consisted of 36 PC and 16 NC. The GEO-developed R-based interactive tool GEO₂R (http://www.ncbi.nlm.nih.gov/geo/info/geo2r/↗) was used to compare expression profiles between samples. The expression values of specific genes were extracted from the original submitter-supplied files. The data were normalised to the average expression of the NC samples in the same probeset. TCGA RNA-seq data and corresponding clinical data from 178 PC patients were downloaded from the TCGA database (https://tcga-data.nci.nih.gov/tcga/↗) following approval of this project by the consortium. GSEA was performed to gain insight into the biological pathways involved in PC pathogenesis through the miR-135b–BMAL1 axis. Gene sets with FDRs of 0.25, a well-established cutoff value for the identification of biologically relevant genes, were considered to be enriched between the classes being compared. The gene sets collection (C2.CP: KEGG v5.2) from the Molecular Signatures Database-MsigDB (http://www.broad.mit.edu/gsea/msigdb/index.jsp↗) were used for enrichment analysis. Moreover, the profiling data from the GSE19650↗ cohort were uploaded (http://www.ingenuity.com↗) and used to conduct an ingenuity pathway analysis (IPA) to map the lists of BMAL1-associated differentially expressed genes between PC and normal pancreatic tissue. The ingenuity ‘core analysis’ typically generates canonical pathways and functional molecular networks pertaining to the data set in question based on the Ingenuity Pathway Knowledge Base (IPKB), which is derived from known functions and interactions of genes published in the literature.

Statistical analyses were performed using the programme R (www.r-project.org↗) and SPSS version 19.0 software. Data are presented as the means ± SEM of at least three biological repeats. Comparisons between groups were analysed with Student’s t-test, one-way ANOVA, and χ² tests. A P-value < 0.05 was considered to be statistically significant. Pearson Correlation analysis was used to determine the relationship between the expression levels of miR-135b, BMAL1 and YY1. OS and DFS curves were plotted according to the Kaplan–Meier method, and the log-rank test was used for comparison. Prognostic factors and survival data were further evaluated by the multivariate Cox regression analysis. Hazard ratios and 95% confidence intervals were derived from Cox’s proportional hazards model. The rhythms of clock-related genes and miR-135b were analysed using the JTK_CYCLE nonparametric algorithm, which is a highly efficient tool for characterising circadian oscillations and cycling variables²⁵.

Supplementary Information