NSUN5 is Upregulated and Positively Correlated with Translation in Human Cancers: A Bioinformatics-based Study

The role of RNA m 5 C (5-methylcytosine) and RNA m 5 C methyltransferases (RCMTs, including NSUN1, NSUN2, NSUN3, NSUN4, NSUN5, NSUN6, NSUN7 and TRDMT1) in human cancers remains largely unknown. In this study, GEPIA2 was used to compare the expression of RCMTs in human cancers and that in associated normal tissues, and to analyze the prognosis value of NSUN5 expression. UALCAN was used to compare the methylation level of NSUN5 promoter in human cancers and that in associated normal tissues. LinkedOmics was used perform BPs (biological processes), CCs (cellular components), MFs (molecular functions) and KEGG pathways analyses of NSUN5-correlated genes in each cancer one by one. We found that six RCMTs (NSUN1-NSUN5 and TRDMT1), especially NSUN5, were generally upregulated in human cancers, that the hypomethylation of NSUN5 promoter may be responsible for its upregulation, and that overexpressed NSUN5 predicted poorer prognosis and was positively correlated with translation in human cancers. The function of NSUN5 in human cancers and its mechanism need to be validated by biological experiments. (NOP2), NSUN2, NSUN3, NSUN4, NSUN5, NSUN6, NSUN7 and TRDMT1 (DNMT2). [5] Only a small number of studies has reported the role of m 5 C and RCMTs in human cancers. For example, m 5 C and RCMTs (NSUN1, NSUN3 and TRDMT1) mediate chromatin organization and 5-azacytidine response and resistance in leukemia. [6] m 5 C, induced by upregulated NSUN2, promotes pathogenesis of bladder cancer through stabilizing oncogene mRNAs. [7] m 5 C regulation and its functions in human cancers remains largely unknown and need further research. ribosomal RNA; SARC, Sarcoma; SKCM, Skin Cutaneous Melanoma; STAD, Stomach adenocarcinoma; TGCT, Testicular Germ Cell Tumors; THCA, Thyroid carcinoma; THYM, Thymoma; tRNA, transfer RNA; UCEC, Uterine Corpus Endometrial Carcinoma; UCS, Uterine Carcinosarcoma; UVM, Uveal Melanoma.


Introduction
According to MODOMICS (https://iimcb.genesilico.pl/modomics/), there are at least 171 types of RNA modi cations so far. [1] m 6 A (N 6 -methyladenosine) is the most pervasive RNA modi cation, and can affect growth and metastasis in multiple cancers (glioblastoma, lung cancer, hepatocellular carcinoma, et al) by in uencing RNA biogenesis, splicing, degradation or translation. [2] However, the role of other types of RNA modi cations in human cancers is poorly studied.
In recent years, m 5 C (5-methylcytosine) has been found to be widespread in human rRNA (ribosomal RNA), tRNA (transfer RNA) and mRNA (messenger RNA). [3,4] There are eight kinds of human RNA m 5 C methyltransferases (RCMTs): NSUN1 (NOP2), NSUN2, NSUN3, NSUN4, NSUN5, NSUN6, NSUN7 and TRDMT1 (DNMT2). [5] Only a small number of studies has reported the role of m 5 C and RCMTs in human cancers. For example, m 5 C and RCMTs (NSUN1, NSUN3 and TRDMT1) mediate chromatin organization and 5-azacytidine response and resistance in leukemia. [6] m 5 C, induced by upregulated NSUN2, promotes pathogenesis of bladder cancer through stabilizing oncogene mRNAs. [7] m 5 C regulation and its functions in human cancers remains largely unknown and need further research.
In this study, using public databases, we analyzed the expression, prognostic value and possible function of NSUN5 (the most obviously upregulated RCMT) in multiple human cancers.

Results
The expression of RCMTs, especially NSUN5, is generally upregulated in cancers GEPIA2 [8] (http://gepia2.cancer-pku.cn/#index) was used to explore the expression of RCMTs in cancers. As is shown in Table 1 and Fig. 1, the expression of RCMTs, except NSUN6 and NSUN7, is generally upregulated in cancers, compared to associated normal tissues respectively. In detail, NSUN1, NSUN2, NSUN3, NSUN4, NSUN5, NSUN6, NSUN7 and TRDMT1 were respectively upregulated in 23,19,18,18,25,7,12 and 13 types of cancers out of all the 31 types of cancers. Among them, the number of cancer types with upregulated NSUN5 was the most. (Table 1)   The hypomethylation of NSUN5 promoter may be responsible for its upregulation in cancers To explore why NSUN5 was generally downregulated in cancers, we compared the methylation level of NSUN5 promoter in cancers and associated normal tissues using UALCAN [9] (http://ualcan.path.uab.edu/). Among the 28 types of cancers with DNA methylation data, the methylation level of NSUN5 promoter was downregulated in BRCA, HNSC, KIRC, LIHC, PAAD, STAD, TGCT, THCA and UCEC. Importantly, in BRCA, HNSC, LIHC, PAAD, STAD, TGCT and UCEC, the methylation level of NSUN5 promoter was downregulated, while the expression of NSUN5 was upregulated. Notably, the methylation level of NSUN5 promoter had a trend of downregulation in BLCA, CESC, ESCA, KIRP, LUAD, LUSC, PRAD, SARC, SKCM and THYM, although it was not signi cant. It may be due to the small quantity of normal tissues. (Fig. 2) To sum up, the methylation level of NSUN5 promoter was generally downregulated in cancers, and may contribute to the upregulation of NSUN5, especially for BRCA, HNSC, LIHC, PAAD, STAD, TGCT and UCEC.

Higher expression of NSUN5 predicts poorer prognosis in multiple cancers
We then used GEPIA2 to explore whether the expression of NSUN5 in cancers was related with the prognosis of cancer patients. Among the 33 types of cancers (

Higher expression of NSUN5 was positively correlated with translation in cancers
To explore the function of NSUN5 in cancers, using LinkedOmics [10] (http://linkedomics.org/login.php), GO (gene ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway analyses of NSUN5-correlated genes was performed in each cancer one by one. For example, in ACC, the co-expression analysis of NSUN5 (mRNA sequencing data) was shown in Supplementary Table 1 and Fig. 5A; the top 50 most signi cantly (according to FDR (false discovery rate)) NSUN5-positively-correlated and NSUN5-negatively-correlated genes were listed in Fig. 5B; the BPs (biological processes), CCs (cellular components), MFs (molecular functions) and KEGG pathways in which NSUN5-positively-correlated genes (FDR ≤ 0.05 and normalized enrichment score > 0) and NSUN5-negatively-correlated (FDR ≤ 0.05 and normalized enrichment score < 0) genes were signi cantly enriched were shown in Fig. 5C. The analyses process of the other 31 types of cancers were similar to that of ACC (data not shown).
The BPs, CCs, MFs and KEGG pathways in which NSUN5-positively correlated genes and NSUN5-negatively-correlated genes were signi cantly enriched in all the 32 types of human cancers were listed in Supplementary Tables 2-9.
To found the common BPs, CCs, MFs or KEGG pathways in human cancers, the number of cancer types with a same BP, CC, MF or KEGG pathway was calculated. The BPs, CCs, MFs or KEGG pathways were ranked by the times they appeared in cancers. The top 10 BPs, CCs, MFs and KEGG pathways in which NSUN5-positively-correlated genes were signi cantly enriched were listed in Tables 2-5, respectively. The top 10 BPs, CCs, MFs and KEGG pathways in which NSUN5-negatively-correlated genes were signi cantly enriched were listed in Supplementary Tables 10-13, respectively.  BP analyses of NSUN5-correlated genes in cancers using Linkedomics. The BP in which NSUN5-positively-correlated genes were signi cantly enriched (FDR ≤ normalized enrichment score > 0) was denoted by "√", otherwise, it was denoted by "×". The BPs were ranked by the number of cancer types with "√" and only shown. BP, biological process.
A previous study showed that loss of the rRNA methyltransferase NSUN5 impairs global protein synthesis (translation) and normal growth. [11] Interestingly, in this study, we found that higher expression of NSUN5 in most cancers was positively correlated with translational elongation, rRNA metabolic process, ribonucleoprotein complex biogenesis, translational initiation, ( Table 2) ribosome, (Tables 3 and 5) rRNA binding and unfolded protein binding, (Table 4) which were all associated with protein synthesis. For example, in ACC, higher expression of NSUN5 was positively correlated with translational elongation, rRNA metabolic process, ribonucleoprotein complex biogenesis, translational initiation and ribosome. (Table 6) The corresponding gene (mRNA) symbols were listed in Table 6.  CC analyses of NSUN5-correlated genes in cancers using Linkedomics. The CC in which NSUN5-positively-correlated genes were signi cantly enriched (FDR ≤ normalized enrichment score > 0) was denoted by "√", otherwise, it was denoted by "×". The CCs were ranked by the number of cancer types with "√" and only were shown. CC, cellular component.  Number of  cancer  types with  "√"   30  29  28  28  27  27  27  24  24  24  24 MF analyses of NSUN5-correlated genes in cancers using Linkedomics. The MF in which NSUN5-positively-correlated genes were signi cantly enriched (FDR ≤ 0.05 and normalized enrichment score > 0) was denoted by "√", otherwise, it was denoted by "×". The MFs were ranked by the number of cancer types with "√" and only the top 10 MFs were shown. MF, molecular function. KEGG pathway analyses of NSUN5-correlated genes in cancers using Linkedomics. The pathway in which NSUN5-positively-correlated genes were signi cantly enriched (FDR ≤ 0.05 and normalized enrichment score > 0) was denoted by "√", otherwise, it was denoted by "×". The pathways were ranked by the number of cancer types with "√" and only the top 10 pathways were shown. KEGG, Kyoto Encyclopedia of Genes and Genomes. In summary, higher expression of NSUN5 was positively correlated with translation-related mRNA, BPs, CCs, MFs and KEGG pathways in human cancers.

Discussion
In the present study, by data mining, we found that out of all the eight kinds of RCMTs, six kinds of RCMTs (NSUN1, NSUN2, NSUN3, NSUN4, NSUN5 and TRDMT1) were generally upregulated in human cancers. It indicates that m 5 C and RCMTs may play important roles widely in human cancers. In fact, it has been reported that in leukemia, NSUN3 and TRDMT1 could bind hnRNPK, NSUN1 could bind BRD4, both leading to formation of 5-Azacitidine-sensitive chromatin structure. [6] While NSUN2 could drive bladder cancer progression by stabilizing the mRNA of HDGF through upregulating its m 5 C level. [7] However, the function of m 5 C and RCMTs and its mechanism in other human cancers remain largely unclear. Researches on these aspects should be carried on in the future.
It has been reported that NSUN5 is a rRNA methyltransferase, introducing m 5 C3782 into human 28S rRNA. [11] NSUN5 de ciency altered ribosome function, leading to impaired global protein synthesis and normal growth. [11] In this work, we revealed that NSUN5 was the most obviously upregulated RCMT, that higher expression of NSUN5 predicted poorer prognosis, and that NSUN5 expression was positively correlated with translation-related mRNAs, BPs, CCs, MFs and KEGG pathways in multiple human cancers. In addition, the hypomethylation of NSUN5 promoter may be responsible for its upregulation in cancers.
To sum up, we assumed that higher expression of NSUN5 could promote growth by inducing global translation through upregulating translation-related mRNAs in human cancers. The role of NSUN5 in human cancers should be con rmed by biological experiments and the mechanism how NSUN5 upregulates translation-related mRNAs needs further exploration.

Conclusions
Totally, bioinformatics analysis showed that NSUN5 was an oncogene, overexpressed NSUN5 predicted poorer prognosis and was positively correlated with translation in human cancers.
The "Single Gene Analysis-Survival Analysis" module in GEPIA2 was used to analyze the prognostic value of NSUN5 in human cancers. The median expression of NSUN5 was used as the cutoff, which distinguished High-NSUN5-Group from Low-NSUN5-Group.

UALCAN
The "TCGA analysis" module in UALCAN [9] (http://ualcan.path.uab.edu/) was used to compare the promoter methylation level of NSUN5 in human cancers and that in associated normal tissues. DNA methylation level was indicated by Beta value ranging from 0 (unmethylated) to 1 (fully methylated).

LinkedOmics
The "LinkFinder" module in LinkedOmics [10] (http://linkedomics.org/login.php) was used to nd NSUN5-correlated genes each cancer one by one. Both "Search Dataset" and "Target Dataset" were chosen as "RNAseq", and the statistical method was chosen as "Spearman Correlation test".