LKB1 loss links the serine metabolic network to DNA methylation and tumourigenesis [RNA-seq]. Mus musculus

Intermediary metabolism generates substrates for covalent modification of chromatin, enabling potential coupling of metabolic states and epigenetic control. Here, we identify such a network as a major component of oncogenic transformation downstream of the LKB1 tumour suppressor. LKB1 encodes a serine-threonine kinase that integrates nutrient availability, metabolism and growth, however the mechanisms for LKB1-dependent tumour suppression remain elusive. By developing genetically engineered mouse and primary epithelial cell models and employing transcriptional, proteomics, and metabolic analyses, we find that oncogenic cooperation between LKB1 loss and KRAS activation, alterations commonly coinciding in human cancer, is fueled by pronounced mTOR-dependent induction of the serine-glycine-one carbon network coupled to S-adenosylmethionine generation. In concert, DNA methyltransferases (DNMT1, DNMT3A) are upregulated, leading to elevation in genomic 5-methylcytosine levels, with particular enrichment at retrotransposon elements, associated with silencing of retrotransposon transcription. Correspondingly, LKB1 deficiency renders cells highly sensitive to inhibition of serine biosynthesis and DNA methylation in vitro and in vivo. Thus, we define a hypermetabolic state resulting from LKB1 loss and KRAS activation that fuels changes in the epigenetic landscape. This state is critically required for the tumourigenic program of LKB1-mutant cells, suggesting novel points of therapeutic intervention in defined patient subsets. Overall design: RNA-sequencing was performed using total RNA isolated from 2 independent murine pancreatic ductal cell lines from KRASG12D/+ (K) or KRASG12D/+;LKB1-/- (KL) genotypes or from the two independent KL lines transduced with full length LKB1 cDNA (rescue). RNAseq libraries were prepared using standard Illumina protocols and sequencing was performed in the Illumina HiSeq 2500 instrument. Data was processed using a standard RNA-seq pipeline that used Tophat2 to align the reads to mm9, and the Cufflinks suite8 to calculate differential expression.