br Discussion br Our analysis of
Our analysis of correlations among OCM gene expression in tumor cell lines showed that expression levels of many OCM genes were correlated with each other, suggesting their possible co-regulation (Table 2). We also observed multiple modest associations of OCM gene expression with tumor cell response to individual antitumor agents in the GDSC-CCLE dataset (Table 1). While some of these associations confirm
several previously published reports [12,82], given the com-paratively weak correlations with drug response (Pearson r and Spearman ρ between 0.3 and 0.4), they should be inter-preted with caution and require further validation and investi-gation of their underlying molecular mechanisms. When six of the top 20 associations were examined in the NCI-60 dataset, the direction of all correlations was consistent with that in the GDSC-CCLE dataset, with higher expression levels of OCM genes being associated with drug sensitivity (Supplementary Table 2). However, only three out of six of these associa-tions were statistically significant when using Pearson correla-tion analysis of the full NCI-60 dataset. The NNMT-dasatinib association was also significant when comparing the high-est and the lowest NNMT-expressing cell lines, in agreement with a previously published report . This association is of particular interest given the bimodality of NNMT expression (Fig. 2C).
The antitumor agents for which cell line sensitivity was modestly associated with elevated expression of multiple OCM genes may be of particular interest because this could suggest a possibility that they may be more dependent on OCM. Fig. 3 shows four genes, GART, TYMS, SHMT2, and MTR, for which higher expression levels were significantly as-sociated with an increased sensitivity to 37 agents. Products of these OCM genes have unique functions. SHMT2 partic-ipates in the mitochondrial conversion of serine to glycine, which is accompanied by the mitochondrial production of 5,10-methyleneTHF from THF [1,7,8,28]. TYMS catalyzes the conversion of 5,10-methylene-THF to DHF [8,85,95]. This reaction is coupled with DNA synthesis, and during the S phase TYMS is transferred from the cytosol to the nu-cleus to provide deoxythymidine triphosphate for DNA repli-cation [1,7,8]. GART catalyzes multiple important steps in purine biosynthesis while generating cytosolic THF from 10-formylTHF [20,28,96], whereas MTR participates in homocys-teine remethylation to methionine using 5-methylTHF [20,85]. While all of these 12-O-tetradecanoyl phorbol-13-acetate catalyze important reactions in the OCM pathway, their roles are diverse.
Twenty-two of the 37 agents shown in Fig. 3 were as-sociated with two or more of the four genes, GART, TYMS, SHMT2, and MTR. Many agents in Fig. 3 are kinase inhibitors, which is consistent with associations of multiple kinases and kinase ligands with OCM gene expression (Table 3), indicat-ing that the action of some kinases or their ligands targeted by these agents may be directly or indirectly associated with OCM reactions. Because higher expression levels of GART, TYMS, SHMT2, and MTR were associated with increased sensitivity to the agents listed in Fig. 3, this may suggest that cancer cells harboring higher levels of OCM reactions cat-alyzed by the products of these genes may be more sensitive to these agents. Among possible explanations for such asso-ciations could be either non-drug-specific sensitivity of cancer cells with higher levels of OCM reactions to drug treatment, or a possibility that molecular targets of such drugs might be associated with higher levels of expression of OCM-related genes in cancer cell proliferation and survival.
A broad explanation for the association of elevated OCM gene expression with sensitivity to various agents listed in Table 1 may involve an established association of increased levels of one-carbon metabolism reactions in tumor cells with higher rates of cancer cell proliferation, increased aerobic glycolysis, NADH production, and elevated levels of TYMS- 33
mediated pyrimidine metabolism . It may be possible to suggest that a combination of such processes in rapidly grow-ing cancer cells could lead to small increases in the overall sensitivity of cancer cells to treatment with a variety of agents, resulting in modest associations listed in Table 1. Additionally, as discussed above, multiple genes whose products catalyze OCM reactions and a number of kinase genes may have el-evated transcriptional levels in tumor cells [97,99], and the correlation of elevated rates of OCM reactions with higher kinase expression and associated drug sensitivity could in-directly result from progression of tumorigenesis. While such non-specific association of OCM genes with expression and activity of kinase targets appears plausible, there could also be a possibility of more direct connections between the ac-tivity or expression of individual kinases and the genes in-volved in folate metabolism and transport. For example, one study  noted increased ERBB2 protein expression in mammary tumors of rats who received folate supplementa-tion. In addition, some reports [1,102] suggested regulation of transcriptional levels of PHGDH via ERBB2 overexpression, as well as possible regulation of SLC19A1 activity by phos-phorylation, which could possibly indicate a regulatory effect of kinase activity on SCLC19A1; notably, activities of both PHGDH and SLC19A1 contribute molecular components to OCM reactions. Therefore, it may be possible that association of OCM gene expression with expression of multiple kinase genes could result from a combination of indirect and direct causes.