br Nicotinamide N methyltransferase the
Nicotinamide N-methyltransferase, the product of NNMT, is an important drug target. It is involved in the conver-sion of nicotinamide to 1-methylnicotinamide that uses S-adenosylmethionine as a methyl donor, in methylation of pyridines and their structural analogs, and in biotransforma-tion of drug agents [74–77]. NNMT plays important roles in oncogenesis and cancer progression and is overexpressed in a variety of cancers, resulting in aberrant methylation of cancer-related proteins, and its phosphorylation is inhibited by an ALK inhibitor, RO5424802 [75,78,79]. Downregulation of NNMT was observed to induce apoptosis in breast cancer cell lines . In our study, higher NNMT gene expression lev-els were associated with sensitivity to dasatinib (r = −0.340) and to an Lck/Src inhibitor, WH-4-023 (r = −0.315). In con-trast, they were also weakly (r between 0.302 and 0.325) as-sociated with chemoresistance to six additional agents from several CytochalasinD (Table 1). Among these agents, association of NNMT expression with chemoresistance to methotrexate, vorinostat, WZ3105, and XMD13-2 also satisfied the criteria of ρ > 0.3 and FDR adjusted p < 0.05 in Spearman correla-tion analysis (Table 1), demonstrating the robustness of this association even though the NNMT gene expression was bi-modally distributed. Some studies have suggested that genes with bimodal transcriptional patterns may serve as diagnos-tic and prognostic cancer biomarkers [80,81]. Based on the biological importance of NNMT in cancer [72,75,78,79], bi-modality of its gene expression, and the statistical association of NNMT gene expression with cellular response to multiple drugs, this gene and its product may be of potential clinical interest in cancer treatment.
Association of elevated baseline NNMT expression with dasatinib sensitivity was further validated in the NCI-60 dataset, in which we observed a statistically significant differ-ence in drug sensitivity values between the cell lines with the highest and the lowest NNMT expression, with a very large ef-fect size (Cohen’s d = −2.179, FDR adjusted p = 0.0219; Sup-plementary Table 2). Pearson correlation analysis of the entire NCI-60 dataset also showed a statistically significant correla-tion between NNMT expression and sensitivity to dasatinib (r = −0.364, FDR adjusted p = 0.0156; Supplementary Table
Fig. 2 Histograms and density functions showing the distributions of expression values of selected folate metabolism-related genes among 635 cell lines with available CCLE gene expression data that had matching GDSC drug response data. (A) SLC46A1; (B) NNMT; (C) GART; (D) TYMS; (E) SHMT2; (F) MTR. Horizontal scale represents log2-transformed gene expression values. The left vertical scale represents cell line counts, whereas the right vertical scale represents density values.
2). Our results were in agreement with a prior report by Huang et al.  who found an association of NNMT expression with dasatinib sensitivity; in contrast, Chen et al.  reported that NNMT expression was a biomarker of resistance to dasa-tinib. In agreement with our findings of divergent directions of association with sensitivity to treatment, NNMT expression has been reported to be associated with resistance to radi-ation and to several cancer drugs, but with sensitivity to a sulphonamide cryptopleurine analogue .
Previous reports had suggested that BHMT, which encodes betaine-homocysteine methyltransferase, has a protective role against cancer, and its deletion in mouse models promoted fatty liver disease and hepatocellular carcinoma [84,85]. Higher levels of BHMT expression were associated with sensitivity to kinase inhibitors erlotinib and lapatinib (r = −0.359 and −0.387, respectively; Table 1). Our examination of these associations in the NCI-60 dataset (Supplementary Table 2) showed a partial validation of the BHMT-erlotinib association, which was significant when using Pearson correlation analysis in the full NCI-60 dataset (r = −0.349, FDR adjusted p = 0.0156), but the difference in erlotinib sensitivity between the highest and the lowest BHMT-expressing cell lines was not statistically significant. No significant association of BHMT expression with log(GI50) of lapatinib was observed in the NCI-60 dataset (Supplementary Table 2).
The increased expression of MAT2B, which encodes the beta subunit of the L-methionine S-adenosyltransferase II that participates in the biosynthesis of S-adenosylmethionine (SAM) and contributes to growth of colon and liver cancer cells , was associated in the GDSC-CCLE dataset with sensitivity to an ALK inhibitor crizotinib (r = −0.345), a kinase inhibitor sunitinib (r = −0.306), and a gamma-secretase in-hibitor Z-LLNle-CHO (r = −0.334; Table 1); however, in the NCI-60 dataset the association of MAT2B expression with crizotinib sensitivity was not significant (Supplementary Ta-ble 2). In the GDSC–CCLE dataset, higher expression lev-els of ALDH2, the product of which affects folate levels in vivo , were associated with resistance to Z-LLNle-CHO (r = 0.319). Elevated expression of MTHFD2, the product which is involved in the mitochondrial OCM pathway, was weakly (r = −0.308) associated with sensitivity to a small molecule IKKε inhibitor, KIN001-135. This gene is overex-pressed in many cancer categories and is required for sur-vival of malignant cells . Elevated expression levels of MTHFD2 had been associated with shorter survival rate of patients with colorectal cancer and lung adenocarcinoma .