HnRNP proteins controlled by c-Myc deregulate pyruvate kinase mRNA splicing in cancer

Abstract

When oxygen is abundant, quiescent cells efficiently extract energy from glucose primarily by oxidative phosphorylation, whereas under the same conditions tumour cells consume glucose more avidly, converting it to lactate. This long-observed phenomenon is known as aerobic glycolysis, and is important for cell growth. Because aerobic glycolysis is only useful to growing cells, it is tightly regulated in a proliferation-linked manner. In mammals, this is partly achieved through control of pyruvate kinase isoform expression. The embryonic pyruvate kinase isoform, PKM2, is almost universally re-expressed in cancer, and promotes aerobic glycolysis, whereas the adult isoform, PKM1, promotes oxidative phosphorylation. These two isoforms result from mutually exclusive alternative splicing of the PKM pre-mRNA, reflecting inclusion of either exon 9 (PKM1) or exon 10 (PKM2). Here we show that three heterogeneous nuclear ribonucleoprotein (hnRNP) proteins, polypyrimidine tract binding protein (PTB, also known as hnRNPI), hnRNPA1 and hnRNPA2, bind repressively to sequences flanking exon 9, resulting in exon 10 inclusion. We also demonstrate that the oncogenic transcription factor c-Myc upregulates transcription of PTB, hnRNPA1 and hnRNPA2, ensuring a high PKM2/PKM1 ratio. Establishing a relevance to cancer, we show that human gliomas overexpress c-Myc, PTB, hnRNPA1 and hnRNPA2 in a manner that correlates with PKM2 expression. Our results thus define a pathway that regulates an alternative splicing event required for tumour cell proliferation.

Figure 1. hnRNP proteins bind specifically to sequences flanking E9

a, Schematic diagram of PKM splicing. b, Position of probes spanning the E9 or E10 5′ splice sites (top). After ultraviolet crosslinking, proteins were detected by autoradiography (bottom). Position of molecular mass standards in kDa is indicated at left. c, Affinity chromatography using EI9(50–68). Bound proteins were separated by SDS–PAGE and Coomassie stained. Bands excised for mass spectrometry are indicated. d, Sequence of EI9(50–68); the putative hnRNPA1/A2 binding site is indicated in bold italics (top). Ultraviolet crosslinking with wild-type RNA, or RNA with a mutation in the putative hnRNPA1/A2 binding site, is shown in the bottom panel. e, Position of I8 and I9 (top). Ultraviolet crosslinking using I8 or I9 substrates is shown in the bottom left panel. Ultraviolet crosslinking reactions were immunoprecipitated with either anti-PTB (BB7) or anti-HA antibodies (bottom right panel). f, Ultraviolet crosslinking with I8 and the mutant derivative I8mu, sequences indicated above. Putative PTB binding sites in I8 are underlined.

Figure 2. PTB, hnRNPA1 and hnRNPA2 are required for high PKM2/PKM1 mRNA ratios

a, Scheme for assaying PKM1/PKM2 ratios in human cells. b, Immunoblots showing protein levels after the indicated siRNA treatment. Protein bands were quantified after LI-COR Odyssey scanning and normalized to GAPDH. c, The indicated splicing factors were depleted by siRNA, followed by PKM splicing assay outlined in a. Products corresponding to M1 and M2 isoforms are indicated with arrows. The PKM1 percentage is indicated below. d, PKM1 and PKM2 levels assayed after the indicated siRNA treatment in 293 cells.

Figure 3. Expression of PTB, hnRNPA1, hnRNPA2 and c-Myc correlates with PKM2 expression in C2C12 cells and tumours

a, PKM splicing assay after the indicated number of days of C2C12 differentiation. b, Immunoblots for the indicated proteins were performed throughout differentiation, and normalized to GAPDH (day 0 5 1). c, RNA was extracted from brain tissue or tumour samples and assayed for PKM mRNA isoforms. d, Lysates were immunoblotted for PTB, hnRNPA1, hnRNPA2 or c-Myc and normalized to actin. Sample order is the same for RT–PCR and immunoblotting.

Figure 4. c-Myc upregulates PTB, hnRNPA1 and hnRNPA2 and alters PKM splicing

a, Immunoblotting using NIH-3T3 cells stably expressing control or c-Myc-targeting shRNAs. Signals were quantified and normalized to actin. b, RT–PCR using the same cell lines as in a, using B2m (β2-μglobulin) as a loading control. Real-time RT–PCR was performed separately to quantify the relative levels of Ptb, Hnrnpa1 and Hnrnpa2 mRNAs in control and c-Myc knockdown cells, using Rpl13a as a reference gene. Relative levels of each are shown below each panel, with s.d. indicated (n = 3). c, Pkm1/Pkm2 ratios in control and c-Myc knockdown cells determined as in Fig. 2a. d, A model for PKM splicing regulation. Top: in adult tissues, low expression of PTB, hnRNPA1 and hnRNPA2 allows for recognition of E9 by the splicing machinery and disrupts intronic structures favourable for E10 inclusion. Bottom: in embryonic and cancer cells, PTB, hnRNPA1 and hnRNPA2 are upregulated, bind to splicing signals flanking E9 and repress its inclusion. Binding of these proteins around E9 and possibly to other sites creates an intronic structure favourable to E10 inclusion.