P-values report on the significance of the difference between TRs and enzyme pdfs (KolmogorovSmirnov test). With Mutations and and and. Data are meanstandard deviation across three replicates. Comput. Genome-related datasets within the E. coli Genetic Stock Center To quantify the number of extracted cells, confluence was divided by the characteristic cell size (Supplementary Fig. enzymes directly or indirectly regulated by a TR. Genes. In agreement with the genetic basis of ccRCC, we identified VHL and HIF-1A among the top 1% of TRRUST-listed24 transcriptional regulators that potentially mediate metabolic rearrangement in ccRCC (Fig. The Cancer Genome Atlas Research Network. Metab. F0392) with a reduced content of low molecular weight compounds, to improve the accuracy of metabolite quantification. Subramanian, A. et al. 70-Gene signature as an aid to treatment decisions in early-stage breast cancer. Bioeng. Science 347, 779784 (2015). 1). 8, ncomms16018 (2017). Cellular Neuroscience, Neurodegeneration & Repair Program. 10, e1003572 (2014). 2a, Supplementary Fig. For 2181 putatively annotated ions exhibiting a significant linear dependency between extracted cell number and ion intensities (linear regression p-value3.4e7, Bonferroni-adjusted threshold), we report Z-score normalized relative abundances fitted over approximately 15 individual measurements per cell line (Fig. Regulation of cAMP-responsive element-binding protein-mediated transcription by the SNF2/SWI-related protein, SRCAP. Nature Communications (Nat Commun) Alam, M. T. et al. d Predicted network of modulatory interactions between TRs, metabolites and kinases. Article In the more specific cases described above, our analysis of HIF-1A and VHL activities across cell lines recapitulate the action of HIF-1A inhibitor vorinostat47 and HIF-1A inducer imiquimod48 (Fig. 31, 419425 (2013). Consistent with the hypothesized functional dependency between glucose and one-carbon metabolism, we found that among all drug classes, the strongest correlation with glycolytic flux was observed with cell line sensitivities to antifolate drugs and inhibitors of dihydrofolate reductase (DHFR)i.e., cell lines with lower glucose uptake are more sensitive to inhibitors of folate biosynthesis (Fig. A call to rethink some published values. Only pathways with a q-value0.01 are labeled. The CGSC Database of E. coli genetic information includes genotypes and reference information for the strains in the CGSC collection, the names, synonyms, properties, and map position for genes, gene product information, and information on specific mutations and references to primary literature. Han, H. et al. Altogether, our work also suggests that, while clearly far from typical in vivo conditions, in vitro cell line systems represent an invaluable discovery tool to investigate metabolic regulatory mechanisms that can still generalize to in vivo conditions and clinical settings. PLoS. TRs can directly regulate metabolic fluxes by modulating enzyme abundance, i.e. TRRUST database24 served as the source of TR-target gene interactions relevant in human, including 748 human TRs and 1975 non-TR gene targets. 2). As a proof of concept, we measured rates of glucose uptake and lactate secretion in each cell line as a proxy for glycolytic flux. that are increased with low glycolytic flux, were significantly (hypergeometric test, q-value0.001) enriched for one-carbon metabolism (Fig. All authors contributed to preparing the manuscript. Clin. Monroy, M. A. et al. 2) with PC1. However, often only few metabolic intermediates of metabolic pathways, typically the end product37, can allosterically regulate TRs. A complete description of the linear regression model used to quantify associations between variation in drug sensitivity and TR activity can be found in Supplementary Methods. 399, 11271139 (2011). Proc. The starting cell density for metabolomics experiments in 96-well plates (Nunc cat.no. Cell 159, 402414 (2014). ADS Immediately prior to MS analysis, the plates were thawed on ice, and the extracted cells were scraped off the bottom of each well using a multichannel pipet with wide-bore tips. 1, 20 (2013). no. 1d, Supplementary Fig. 2e). Fuhrer, T., Heer, D., Begemann, B. This procedure enables selecting only annotated ions exhibiting a linear dependency between measured intensities and extracted cell number (Fig. First, we quantified relative metabolite abundance per cell using a multiple linear regression scheme. Cascante, M. et al. Boxplot of Z-score normalized metabolite levels are grouped by the cell line tissue of origin. changing maximum flux capacity, or by indirectly affecting substrate availability of proximal metabolic reactions, which can in turn result in local changes of fluxes25,26. b Correlations between metabolite levels and TR activity profiles in relation to the distance between TRs and enzyme targets in the metabolic network21. 2 However, Rutter et al noticed that some patients developed CRC despite earlier . Barretina, J. et al. This methodology adopts a mathematical model to approximate TR-target regulatory interactions and integrates prior network information with the expression of target genes across multiple conditions to regress the activity of the respective TRs, delivering a relative measure of TR activity. Finally, pooled cell extracts for each experiment (pooled from five cell lines processed within the same experiment) as well as aliquots of cell-free extraction solvent were added to each measurement plate as control samples, and the plates were sealed and stored at 4C until injection. At each sampling time point, two replicate 96-well plates were processed (plates A and B). However, among KEGG pathways with known HIF-1A target genes, we found a significant enrichment of metabolic intermediates almost confined to TCA cycle (hypergeometric test, q-value<0.05), suggesting that changes in HIF-1A activity alone are sufficient to affect TCA cycle. Correlation analysis was used to find associations between TRs and metabolites. Here, we follow each cell line along the linear growth phase, sampling every 24h across 5 days. It is worth noting that the estimates we obtain with this approach are correct within an unknown scaling factor, and hence we determine a relative measure of activity for each of the 728 TRs across the NCI-60 cell lines (Supplementary Data2). The increase in abundance of TCA cycle intermediates upon HIF-1A knockdown is consistent with the previously reported regulatory role of HIF-1A as a repressor of oxidative metabolism40 via induction of pyruvate dehydrogenase kinase (PDK), and is in agreement with the enrichment analysis of functional associations in our TR-metabolite association network (Fig. Principal component analysis of relative metabolite abundance per cell revealed a strong trend across the 54 cell lines (PC1, 58.9% explained variance, Supplementary Fig. Other top-ranking TRs include YY1 that has been shown to interact with hypoxia-inducible factors42 (see also Supplementary Discussion and Supplementary Fig. After correcting for possible instrumental artifacts, we implemented a second step in order to derive comparative measurements of metabolite abundance for each cell line. Here, we established an in silico framework for generating hypotheses on regulatory interactions between TRs, metabolites and kinases (Fig. It is tempting to speculate that cancer cell line diversity in glucose uptake and lactate secretion observed in vitro potentially reflects regulatory programs acquired during an earlier adaptation to in vivo nutrient availability and stresses. However, it is important to emphasize that gene regulatory interactions between enzymes and TRs per se are not sufficient to functionally regulate the activity of metabolic pathways. Originally established by Liao et al.29, NCA provides a mathematical framework for reconstructing TR regulatory signals (TR activity) from gene expression profiles. A complete description of the analysis can be found in Supplementary Methods. Reinhold, W. C. et al. 7). Hence, because TR activity is governed by complex post-transcriptional and post-translational mechanisms, monitoring TR gene expression or protein levels is an inadequate proxy of their activity27,28 (Supplementary Fig. Genome-related datasets within the E. coli Genetic Stock Center Nat. Systematic identification of metabolites controlling gene - Nature Introduction. In order to find metabolites whose relative abundances correlate with TR activity, we calculated pairwise Spearman correlations between all 2181 annotated metabolites and 728 TRs across the 54 cell lines (Supplementary Data3). A. et al. Colored squares indicate the significance of the enrichment analysis. Box edges reflect 25th and 75th percentiles, center dots indicate medians, whiskers include extreme data points, and outliers are gray circles. Get what matters in cancer research, free to your inbox weekly. 5). Natl Acad. In total, we tested 6,753,600 models, and found 1888 interactions which significantly (FDR0.1%) improved the explained variance in the activity of 96 TRs out of the 100 TRs for which protein abundance data was available11. Cells were maintained in medium with dFBS throughout all experiments. Guo, T. et al. In the third and last step, we take into account systematic changes in metabolite abundances related to differences in cell size (i.e., cell volume) between the 54 cell lines to derive comparative estimates of intracellular metabolite concentration. By analyzing the coordinated changes in baseline transcriptome, proteome and metabolome with the aid of a gene regulatory network and model-based fitting analysis, we investigated the bi-directional exchange of signaling information between TRs and metabolic pathways. Zelezniak, A., Sheridan, S. & Patil, K. R. Contribution of network connectivity in determining the relationship between gene expression and metabolite concentration changes. The activated form of a transcriptional regulator, rather than its expression/protein level, regulates gene promoters, and a TRs activity is imprinted in the expression levels of its target genes. Journal peer review information: Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. CAS The procedure is described in details in the respective sections. For each ion, the values are specific of each cell line while the constant term in the model is fixed (i.e., expected ion signal at zero confluence that is independent of the cell line). Preprint at https://www.biorxiv.org/content/10.1101/268953v2 (2018). Genet. Huang, S.-W. et al. Sciacovelli, M. & Frezza, C. Oncometabolites: unconventional triggers of oncogenic signalling cascades. A next generation connectivity map: L1000 platform and the first 1,000,000 profiles. 8). To understand the interplay between transcriptional regulation and emerging metabolic phenotypes, we generated a genome-scale map of TR-metabolite associations. The black line represents average TR-metabolite distance (y-axis) at different levels of absolute TR-metabolite correlation (x-axis). 276, 4072140726 (2001). When testing these associations between groups of TRs regulating similar cellular processes and drugs with a shared mode of action (MoA), we found expected and potentially new functional associations (Fig. Hence, besides generating experimentally testable hypotheses on condition-specific regulatory roles of TRs in metabolism (Fig. 1a). To generate an empirical network of associations between TRs and metabolites, we systematically correlated the activity of 728 TRs with relative levels of individual metabolites across cell lines (Fig. Of note, we found that prior to normalization, the variance across three biological replicates at the same time-point was equally low in cell confluence (median: 7.4% CV) and raw ion intensities (median: 13%, Supplementary Fig. For example, hypoxia-inducible factor 1 alpha (HIF-1A)39 is reported to act on regulatory elements upstream of nearly 50 enzymes in several central metabolic pathways (Fig. A new software tool for the segmentation of bright-field microscopy images (Supplementary Note) allows automated cell number quantification. Nature 537, 544 (2016). A small-molecule inhibitor of the aberrant transcription factor CBF-SMMHC delays leukemia in mice. Here, we develop a unique experimental workflow for the parallel profiling of the relative abundance of more than 2000 putatively annotated metabolites in morphologically diverse adherent mammalian cells. The mission of the CGSC is to act as a central repository for genetically modified strains of E. coli created by the research community, so that they may be readily available to other researchers and educators. For visualization in Fig. Hence, similarly to gene expression signatures used clinically to guide treatment decisions49, metabolome-based signatures of TR deregulation could open new possibilities in aiding the selection of personalized therapeutic treatments. While the impact of phosphorylation has been systematically studied55,56, much less is known about the potential role of metabolites in the allosteric regulation of TR activity. The cancer cell line encyclopedia enables predictive modelling of anticancer drug sensitivity. After thawing, the 54 adherent cell lines were expanded in cell culture flasks (Nunc T75, Thermo Scientific) at 37C and 5% CO2 in RPMI-1640 (Biological Industries, cat.no. 40, D809D814 (2012). This fictitious TR mimics the general effect of proliferation rates on transcription. As a result, each TR is embedded in a sub-network of 50 TRs and their target genes from the full network. 6). 31, 357361 (2013). Correlation coefficients yielding a 99.9% confidence interval (0.1% FDR) were obtained from the pooled list of absolute correlation coefficients by finding the smallest correlation coefficient that exceeds the maximum value among 99.9% lowest correlation coefficients). Notably, TRs with the highest scores were enriched (permutation test, q-value0.05) for enzyme targets in ubiquinone biosynthesis, insulin signaling, TCA cycle and glycolysis/gluconeogenesis (Fig. 1b). mTOR Inhibitors induce cell-cycle arrest and inhibit tumor growth in epsteinbarr virusassociated T and natural killer cell lymphoma cells. In order to validate this association, we monitored dynamic intracellular metabolic changes upon HIF-1A mRNA degradation (i.e., siRNA knockdown) in IGROV1 ovarian cancer cells, exhibiting an average basal level of HIF-1A activity (Supplementary Fig. Cardoso, F. et al. After this step, we retained 2181ions with a regression p-value below a threshold value of 3.4e-7 (adjusted by the number of cell lines and ions) in at least one cell line, and that showed a significant dependency with the extracted cell number in more than 80% of cell lines (Supplementary Fig. Nat. Edge-size connecting the TR hub and metabolic pathways reflects the number of links found in either network. In the meantime, to ensure continued support, we are displaying the site without styles 2d and Supplementary Fig. 3c) involve the accumulation of metabolites in the oxidative branch of TCA cycle (Fig. This approach overcomes several of the major limitations in generating large-scale comparative metabolic profiles across cell lines from different tissue types or in different conditions, and was applied here to profile 54 adherent cell lines from the NCI-60 panel9. Article 2h). Anal. However, this methodology typically requires a large number of null tests to derive an accurate estimate of 0 (estimate of the proportion of true null p-values)65. The proportionality (i.e., parameter) between ion intensity and extracted cells depend on the intracellular concentration of the metabolite. 3a). & Zamboni, N. Quantification and mass isotopomer profiling of -keto acids in central carbon metabolism. 6 and Supplementary Discussion). Most interactions involve the combined action of a metabolite and a kinase (93% of interactions, Supplementary Data5). Liao, J. C. et al. Chen, J. C. et al. CAS Article N. Engl. 15, 157170 (2012). performed the metabolome experiments. and S.D. We thank Uwe Sauer and Nicola Zamboni for supporting this work and providing laboratory facilities, Dimitris Christodoulou, Maren Diether, Nicola Zamboni and Victor Chubukov for helpful feedback and discussions. We found that enzyme gene expression tends to more strongly correlate with levels of proximal metabolites (Fig. 7, 4962.e8 (2018). 2h) are associated to the regulation of key steps in glycolysis, such as the regulatory factor X3 (RFX3) regulating the glucokinase gene32, the nuclear factor erythroid 2-related factor 1 (NFE2L1) and its interacting partner MAFG, involved in the regulation of oxidative stress response and diverse glycolytic genes33. Nature 499, 4349 (2013). Cite this article. Free Radic. Evolutionary Medicine at Yale. The transcription factor Rfx3 regulates -cell differentiation, function, and glucokinase expression. Our approach consists of two main steps. The authors declare no competing interests. 1e) and metabolic differences across mouse tissues20 (Supplementary Fig. 4a, Supplementary Data3). Four key steps control glycolytic flux in mammalian cells. Here, we adopted sparseNCA implementation by Noor et al.67 (Matlab code downloaded from https://sites.google.com/site/aminanoor/softwares). Metabolite profiling identifies a key role for glycine in rapid cancer cell proliferation. Multiple cell . Semenza, G. L. HIF-1: upstream and downstream of cancer metabolism. B 64, 479498 (2002). Cell. By combining MS profiles throughout cell growth (Supplementary Fig. Biol. Science 336, 10401044 (2012). Cancer 15, 686694 (2015). KEGG pathway enrichment analysis showed that these putatively annotated metabolites were strongly overrepresented in fatty acid metabolism (Supplementary Fig. We found a global regulatory signature coordinating glucose- and one-carbon metabolism, suggesting that regulation of carbon metabolism in cancer may be more diverse and flexible than previously appreciated. Only ions annotated to KEGG identifiers are shown. Metabolic control analysis in drug discovery and disease. The herein-proposed workflow for large-scale metabolome profiling is directly applicable to the study of dynamic metabolic responses to external stimuli18, and can scale to larger cohorts that are now within reach of other molecular profiling platforms61. 1 and Supplementary Note). g Correlations between metabolite levels and gene expression in relation to their distance in the stoichiometric network of human metabolism21. Here, we used our framework to profile the intracellular metabolomes of 54 adherent cell lines from eight different tissue types in the NCI-60 cancer cell line panel. To test the extent to which our conclusions hold true beyond the specific TR-regulatory network and cell lines tested here, we expanded the TR-gene target network to include any enzyme whose transcript levels correlate with TR activity, i.e. designed and implemented the image analysis framework. Paglia, G. & Astarita, G. Metabolomics and lipidomics using traveling-wave ion mobility mass spectrometry. To this end, raw data were first corrected for instrument drift by normalizing for possible batch/plate effects. & Li, H. Regulatory architecture determines optimal regulation of gene expression in metabolic pathways. 2), which strongly correlates with cell line volumes derived from cell diameters measured in ref. Biotechnol. By submitting a comment you agree to abide by our Terms and Community Guidelines. 83, 70747080 (2011). Of note, even if the current knowledge of TR-target genes is incomplete, few gene targets can be sufficient to estimate TR relative activities using this approach. Dev. Metabolic alterations in lung cancer-associated fibroblasts correlated with increased glycolytic metabolism of the tumor. Moreover, out of 728 TRs in the TR-metabolite association network, HIF-1A was recovered within the top 2% of TRs with the strongest associations to metabolites affected by HIF-1A knockdown (Fig.