The colorectal cell lines HCT116, DLD1, SW480 and HT29, were purchased from ATCC. Cells were cultured in a humidified atmosphere of 95% air, 5% CO₂, using recommended medium and 10% FBS. Formalin-fixed paraffin-embedded (FFPE) colorectal cancer samples were collected from Zhongshan Hospital, Shanghai, China. The study protocol was reviewed and approved by the ethics committee of Zhongshan Hospital. miRIDIAN shMIMIC lentiviral microRNAs (OpenBiosystems) were used to express miRNAs in colorectal cancer cells. For miRNA activity assay, pmirGLO Dual-Luciferase miRNA target expression vector (Promega) was used.

Small RNA was extracted from FFPE tumor samples using High Pure miRNA isolation kit (Roche). RNA quality was assessed by absorbance spectrometry and electrophoresis on NanoDrop and Bioanalyzer 2100 instruments (Agilent). Agilent human genome microRNA microarray was used for miRNA profiling. Raw mean signal, total probe intensities and total gene intensities were uploaded into GeneSpring GX 10.0 software. Differences were considered statistically significant when the Benjamini-Hochberg false discovery rate-corrected p<0.05.

For microRNA, we used stem-loop quantitative RT-PCR. Briefly, 25 ng of total RNA was reverse transcribed using the miRCURY First-strand cDNA kit and the miRCURY microRNA primer sets. QPCR was performed with the Sequence Detection System 7900HT (Applied Biosystems) using the miRCURY LNA™ SYBR^® Green Master mix. The comparative Ct (ΔΔCt) method was used to determine the expression level of miRNA, and 5S RNA and U6B as endogenous controls (16–18).

Glycolysis rates were measured by following the conversion of 5-³H-glucose to ³H₂O as described previously (19). Briefly, cells were washed once in PBS before incubation in Krebs buffer without glucose for 30 min at 37°C. The Krebs buffer was then replaced with Krebs buffer containing 10 mM glucose spiked with 10 μCi of 5-³H-glucose. After 1 h, triplicate samples of media were transferred to PCR tubes containing 0.2 N HCl and the amount of ³H₂O generated which was determined by diffusion. Lactate was measured with a colorimetric kit according to the manufacturer’s instructions (Biovision). Lactate production was normalized to the number of cells. Cellular oxygen consumption rates were measured using Biological oxygen consumption analysis system (Teket).

All statistical analysis was carried out using SPSS version 18.0 statistical software (SPSS). Two-tailed test and p-values <0.05 for significance were used.

Survival in patients with stage III colorectal cancer is totally different after standard treatment. Therefore, this cohort is quite suitable for prognostic analysis. miRNA expression profiles were evaluated in 12 subjects with stage III colorectal cancer. Thirty-two miRNAs showed >1.5-fold (p<0.05) difference. Compared with patients with <5 years survival, 21 miRNAs were up-regulated and 11 miRNAs were down-regulated in patients with ≥5-year survival. Expression map through unsupervised hierarchical clustering showed a clear separation of the two groups (Fig. 1A). To pursue underlying mechanisms, miRNA targets were analyzed by multiple databases (miRanda, TargetScanHuman and PicTar). Of 21 protective miRNAs, miR-124, miR-137 and miR-340 targeted the alternative splicing proteins of PKM: PTB1/hnRNAPA1/hnRNAPA2. This group of proteins has recently been found to regulate the Warburg effect, which is critical for cancer growth (20). To better understand the role of miRNA in cancer metabolism, miR-124, miR-137 and miR-340 were chosen for further study.

Poor prognosis is related with multiple aspects of colorectal cancer, including proliferation, survival and invasion. miRNA target prediction suggested that miR-124, miR-137 and miR-340 were probably involved in cancer metabolic reprogramming. The main aim of metabolic reprogramming is to sustain rapid cancer growth. Then we asked whether growth of colorectal cancer cells was regulated by these miRNAs. HCT116 cells were infected with lentivirus containing the precursor of miRNA or non-silencing control sequence (NS) (Fig. 1B). MiR-145, as a positive control, was known to have potent antiproliferative activity in colorectal cancer (21–23). miR-124, miR-137 or miR-145 inhibited the growth of HCT116 cells. Growth inhibition was more significant in cells treated with three miRNAs simultaneously (Fig. 1C). Consistent results were obtained in other colorectal cancer cell lines (Fig. 1E). To further corroborate the anti-proliferative role of miR-124, miR-137 and miR-340 we used the miRNA specific inhibitor LNA (locked nucleic acid), to reverse miRNA-mediated growth inhibition. After 2 weeks of puromycin selection, miRNA stably expressed HCT116 cells were obtained. Then these cells were transfected with specific LNAs, respectively. LNA rescued cells from miRNA-induced growth inhibition (Fig. 1D). Therefore, miR-124, miR-137 and miR-340 had important roles of regulating colorectal cancer cell growth.

miRNA target prediction indicated that PTB1, hnRNAPA1 and hnRNAPA2 contained binding sites of these miRNAs. The protein level of PTB1 was significantly reduced in HCT116 cells transfected by miR-124, miR-137 or miR-340. miR-124 or miR-340 repressed expression of hnRNAP2 protein, and hnRNAP1 was targeted by miR-137 (Fig. 2A). Consistent results were also found in mRNA level of target genes (Fig. 2B). To verify these transcripts targeted by miRNAs, 3′UTR luciferase reporter plasmids were constructed. Mutation sites were selected according to their evolutionary conservation. Constructs with intact miRNA binding sites of PTB1 were expressed at significantly lower levels of luciferase activity in the presence of these miRNAs. miR-124 and miR-340 repressed luciferase activity of wild type reporter of hnRNAP2. Only miR-137 repressed luciferase activity of wild-type reporter of hnRNAP1 (Fig. 2C). Moreover, expression of PTB1, hnRNAP1 or hnRNAP2 compromised the growth inhibition induced by these miRNAs (Fig. 2D). These data indicated that the PKM alternative splicing proteins were required for miRNA-mediated growth inhibition of colorectal cancer cells.

The two isoforms, PKM1 and PKM2, were produced from mutually exclusive alternative splicing of PKM gene. This process is mainly regulated by PTB1, hnRNAPA1 and hnRNAPA2 (15). Cancer cells were characterized by switching PKM1 to PKM2 expression. PKM1 or PKM2 was assayed by specific exon-overlapping primers, respectively. miR-124, miR-137 and miR-340 induced an increase of PKM1 mRNA, and concurrent decrease of PKM2 mRNA (Fig. 3A). The ratio of PKM1/PKM2 was significantly increased in cells treated by three miRNAs simultaneously (Fig. 3B). Consistent results were obtained by analyzing their protein levels (Fig. 3C). Next we asked whether the switch from PKM2 to PKM1 was essential for miRNA-mediated growth regulation. Knockdown of PKM2 by siRNA in HCT116 cells resulted in growth inhibition, which was consistent with published reports (10,14). In the absence of PKM2, these miRNAs had no effect on cell growth (Fig. 3D). These data indicated that miR-124, miR-137 and miR-340 inhibited colorectal cancer growth through regulating PKM alternative splicing.

PKM2, as a key enzyme for glycolysis, is selectively expressed in cancer cells and is critical for cell proliferation (9,24). Cellular glycolytic rate was measured by conversion of 5-³H-glucose to ³H₂O (14,25,26). miR-124, miR-137 and miR-340 showed a decrease of the glycolytic rate (Fig. 4A). Switching to PKM1 expression resulted in cells preferentially metabolizing glucose by oxidative phosphorylation rather than glycolysis. Oxygen consumption was increased in the presence of miR-124, miR-137 and miR-340 (Fig. 4B). Consistently, these miRNAs promoted cells to convert less pyruvate to lactate (Fig. 4C). Pyruvate was also decreased in these miRNA-treated cells (Fig. 4D) due to higher activity of PKM1 than PKM2 (12). On the other hand, recovering PKM2 expression compromised the miRNA-mediated glycolytic reprogramming (Fig. 4A-D). In lower oxygen condition, energy generation is strongly dependent on glycolysis. Compared with normoxia, miRNAs induced more apparent growth inhibition in hypoxia (Fig. 4E). Taken together, these data suggested that glucose metabolism of colorectal cells were regulated by miR-124, miR-137 and miR-340.