Consequences of blunting the mevalonate pathway in cancer identified by a pluri-omics approach

Abstract

We have previously shown that the combination of statins and taxanes was a powerful trigger of HGT-1 human gastric cancer cells' apoptosis¹. Importantly, several genes involved in the "Central carbon metabolism pathway in cancer", as reported in the Kyoto Encyclopedia of Genes and Genomes, were either up- (ACLY, ERBB2, GCK, MYC, PGM, PKFB2, SLC1A5, SLC7A5, SLC16A3,) or down- (IDH, MDH1, OGDH, P53, PDK) regulated in response to the drug association. In the present study, we conducted non-targeted metabolomics and lipidomics analyses by complementary methods and cross-platform initiatives, namely mass spectrometry (GC-MS, LC-MS) and nuclear magnetic resonance (NMR), to analyze the changes resulting from these treatments. We identified several altered biochemical pathways involved in the anabolism and disposition of amino acids, sugars, and lipids. Using the Cytoscape environment with, as an input, the identified biochemical marker changes, we distinguished the functional links between pathways. Finally, looking at the overlap between metabolomics/lipidomics and transcriptome changes, we identified correlations between gene expression modifications and changes in metabolites/lipids. Among the metabolites commonly detected by all types of platforms, glutamine was the most induced (6-7-fold), pointing to an important metabolic adaptation of cancer cells. Taken together, our results demonstrated that combining robust biochemical and molecular approaches was efficient to identify both altered metabolic pathways and overlapping gene expression alterations in human gastric cancer cells engaging into apoptosis following blunting the cholesterol synthesis pathway.

Fig. 1

Representative score plots of the Principal Component Analysis (PCA) from metabolomics and lipidomics analyses. Score plots of the metabolomics MS data acquired in the negative mode on the LC-MSn°2 platform (A), the LC-MSn°1 platform in the positive mode (B), the GC-MSn°1 platform (C), and the NMRn°1 platform (D). Experimental conditions: control (open squares), docetaxel (plain triangles), lovastatin (open circles), and lovastatin + docetaxel (plain circles). Cells were treated for 36 h. The two dimensions of the analysis and their contribution to the variance are indicated. The data are representative of results from all platforms

Fig. 2

Inter-platform correlation network. The network was established using the RV scores between all the data matrices generated by the 5 analytical platforms. The thickness of the links reflects the concordance level in metabolite identification between platforms. The LC-MSn°1 was used in the positive (LCMS1Pos) or negative (LCMS1Neg) modes; the LC-MSn°2 was used in the metabolomics positive (LCMS2MetaboPos) or negative (LCMS2MetaboNeg) modes or in the lipidomics positive (LCMS2LipidoPos) or negative (LCMS2LipidoNeg) modes

Fig. 3

Score plot of the “meta” PCA prepared from the calculated Dim1 coordinate from primary PCA analyses resulting from each data matrix. Experimental conditions: control (open squares), docetaxel (plain triangles), lovastatin (open circles) and (lovastatin + docetaxel) (plain circles). Cells were treated for 36 h. The two dimensions of the analysis and their contribution to the variance are indicated

Fig. 4

Changes in lipid content. Sphingomyelins (SM), ceramide (Cer 16:0), triglyceride (TG), and phosphatidylcholine (PC) fold-changes in cells treated with (lovastatin + docetaxel) vs. control. (**p-value<0.01). Fold changes (− or +) of the various lipid classes were represented as histogram bars. Cells were treated for 36 h

Fig. 5

Venn diagram of the specific and shared metabolites (M) and lipids (L). Metabolites detected by each technology and lipids detected by the LC-MS technology in the (lovastatin + docetaxel, 36 h) condition vs. control (fold change >1.5, p-value<0.05). The six metabolites common to LCMS2 and NMR were: glutamate, glutamine, myo-inositol, creatine, lactic acid, and fumarate

Fig. 6

Combined network analysis of metabolites and lipids. Data from both the metabolomics and the lipidomics analyses were combined. Samples used were from (lovastatin + docetaxel)-treated cells for 36 h. The shapes and colors were as follows: ovals for metabolites and rectangles for lipids; from yellow to dark orange, more and more up-regulated, from light blue to dark blue, more and more down-regulated

Fig. 7

PCA analysis of the kinetic effects of the (lovastatin + docetaxel) combination. The cells were treated with both drugs for the indicated times (H) and cell extracts were prepared for GC-MS analysis. Seventeen-hours control (open squares), 24 h control (gray squares), 36 h control (black squares), 17 h treament (empty circles), 24 h treatment (gray circles), and 36 h treatment (black circles). Three biological replicates of each treatment condition were performed

Fig. 8

Central carbon metabolism in cancer. Metabolite and gene expression changes in response to (lovastatin + docetaxel) treatment for 36 h. The metabolomics and lipidomics signaling pathways were overlaid with the alterations in gene expression following (lovastatin + docetaxel) treatment compared to control cells after 36 h. The color code was as follows: blue boxes and blue circles referred to down-regulated genes and down-regulated metabolites, respectively. Red boxes and red circles referred to up-regulated genes and up-regulated metabolites, respectively