NSUN2-Mediated Splice Site m5C Methylation Modication Promotes the Progression of Colon Cancer Through Modulating the Discrepant Cleavage of tRNA-Arg

Background: Hypoxia is a key driving factor for the tumour microenvironment restructuring, which leading to the variation of gene expression proling in cancer cells. Increasing evidence reveals the initial action of hypoxia in the epitranscriptomics including RNA methylation. The role of tRNA-derived fragments (tRFs) in regulating tumour metastasis potential has attracted attention. Methods: The expression of tRFs in colon cancer cells under hypoxia were evaluated based on full-transcript sequencing and bioinformatics analysis and their effects on colon cancer metastasis were detected by transwell assays. The role of C34 was veried by introducing mutation and articial m5C modication. The effects of NSUN2 on the biological characteristics of colon cancer cells were investigated on the basis of gain-of-function and loss-of function analyses. Lung metastasis model further uncovered the roles of NSUN2 and key tRF in tumour progression. Assays of RNA immunoprecipitation-qPCR (RIP-qPCR) were performed to identify that NSUN2 is a key methyltransferase for cysteine modication at C34 of tRNA-Arg. Results: The present study veried the up-expression of tRF (tRF-20-MEJB5Y13) and down-expression of tRFs (tRF-20-M0NK5Y93 and tRF-21-3OPP6N7KE) in colon cancer cells under hypoxia, and all of them were derived from different tRNA-Arg. Contradictory effects of these three tRNA-Arg-derived tRFs on metastatic potential of colon cancer were demonstrated in this study. The sequence differences and the key nucleotide bases of tRNA with the methylation modication potential among the source tRNAs were analysed. Notably, our data identied C34 of tRNA-Arg as a key site that may play an important role in hypoxia-mediated tRNA-Arg discrepant cleavage. We further investigated that NSUN2 mediated specic site methylation of tRNA-Arg at C34, thereby protecting tRNA from cleavage by endonuclease and subsequently promoting the mechanism involved. Research on tRFs which derived from tRNA-Arg under hypoxia revealed signicant differences in their expression at the molecular level, revealed the site differences among these three kinds of tRNA C34 and veried that the expression of NSUN2 was up-regulated in colon cancer cells. Signicantly, NSUN2-mediated splice site m5C methylation led to different cleavage outcomes of tRFs by acting on C34, playing a key role in promoting colon cancer metastasis. Our study focuses on changes in the tumour epigenetics transcriptomics with respect to tumour metastasis potential, and provided a deeper understanding of tumour metastasis in the molecular regulatory mechanism of initiating.

of gene expression, including RNA export, ribosome assembly, translation, and RNA stability. m5C modi cations can occur in the acceptor stem (C72), variable loop region (C47-C50), anticodon stem (C38) and anticodon loop (C34) of transfer ribonucleic acids (tRNAs) [4]. m5C in RNA molecules is catalysed by members of the NOL1/NOP2/SUN domain (NSUN) protein family, which includes 7 members of the highly conserved human NSUN family (NSUN1-7) and the DNA methyltransferase (DNMT) homologue DNMT2 [5]. NSUN2-mediated methylation of tRNA sites occurs speci cally in the anticodon or variable loop (VL), thereby protecting tRNA from endonucleic cleavage [6,7]. This protection is important to prevent the accumulation of tRNA-derived fragments (tRFs), which inhibit protein synthesis. For instance, loss-of-function mutations in the human NSUN2 and NSUN3 genes lead to neurodevelopmental and mitochondrial diseases, respectively [8][9][10][11]. According to bioinformatics analysis, NSUN2, as a major m5C methyltransferase, was reported to promote cell proliferation, migration and invasion of breast cancer [12] and gallbladder cancer [13] and was relevant to the poor prognosis of patients with head and neck squamous cell carcinoma [14]. Immunohistochemical (IHC) analysis showed that the protein expression of NSUN2 was elevated in all types of cancer, including oesophageal, gastric, liver, cervical, prostate, kidney, bladder, thyroid, and breast cancers [15]. In addition to their original description of NSUN2 as an RNA guanine-7 methyltransferase (RNMT), Frye and Watt observed high expression of NSUN2 in stage III/IV colon cancer [15], which was subsequently con rmed by other researchers [16]. Intriguingly, a circular RNA (circRNA) derived from the NSUN2 coding sequence called circNSUN2 (hsa_circ_0007380) has recently been identi ed as being frequently upregulated in patients with colorectal cancer (CRC) and predicts poor patient survival [17]. However, the expression level of NSUN2 in colon cancer cells and its functional role in tumorigenesis and progression remain unknown.
Recent studies have revealed important biological roles of tRNA beyond gene translation. Sohailtavazoie and Hanigoodarzi, used genome sequencing technology to measure the number of tRNAs in different types of cells and con rmed that tRNAs encoding speci c amino acids can promote tumour metastasis [18]. Although there are many ways to encode the same amino acid codon, it is found that only one codon of arginine corresponds to tRNA (tRNA ArgCCG, where CCG is the antisense codon of CGG) and one codon of glutamic acid tRNA (tRNA GluUUC) has the characteristics of signi cantly affecting tumor metastasis. By up-regulating or decreasing the expression of these two tRNAs, the metastatic potential also changes. The two codons of synonymous mutations (encoding the same amino acid) lost their transfer characteristics. It is worthy of interest that Arginine has an inseparable relationship with tumour genesis. Arginine has been identi ed as the only physiological precursor of NO [19]. One of the most important roles of arginine is that it participates in the synthesis of polyamines by transferring from the NO synthesis pathway. Polyamines are known to promote tumor growth, invasion and metastasis [20]. Moreover, under pathophysiological stresses such as hypoxia, some speci c types of mature tRNA or precursor tRNA are speci cally cleaved by endonucleases such as Dicer [21], Elac2 [22] and Ang [23] to produce tRFs or tRNA halves (tiRNAs), which belong to the short noncoding RNA family. Abnormal levels of tRFs have been observed in a number of human diseases, including cancer. Our study showed that hypoxia-induced changes in tRF levels may be closely related to the progression of colon cancer metastasis. There were signi cant differences in the expression of different tRFs under hypoxic conditions. Among the differentially expressed tRFs, two signi cantly downregulated tRFs (tRF-20-M0NK5Y93, tRF-21-3OPP6N7KE) and one signi cantly upregulated tRF (tRF-20-MEJB5Y13) showed the most obvious expression differences, which were consistent with the trend of sequencing results (see the basis of previous work) [24,25]. All three types of tRFs are derived from tRNAs corresponding to different codons for arginine (Arg). To distinguish these tRFs in this study, we named them tRF-ArgCCG down-1, tRF-ArgCCU down-2, and tRF-ArgUCG up-1 (the database of named sources and their sequences shown in  supplementary Table S1). Previous studies have con rmed that methyltransferase-mediated tRNA methylation regulates stress-induced tRNA shearing [26]. We speculated that there is some regulatory mechanism related to hypoxia-induced tRNA cleavage, and the levels of tRFs may be closely related to the speci c tRNA structure and speci c sites of RNA methylation. In this study, we identi ed C34 of tRNA as a key site that may play an important role in NSUN2 methyltransferase activity.
In mammals, the precise regulation of tRF cleavage in colon cancer under hypoxic conditions is still largely unknown. In this study, we identi ed an association between tRNA methylation at speci c sites under hypoxic conditions (with discrepant cleavage of tRNA-Arg) and the development of conlon cancer, and explored the mechanism of NSUN2 in this process. Our ndings reveal an unprecedented mechanism by which RNA with m5C regulates tRNA discrepant cleavage in conlon cancer and provide potential therapeutic strategies for colon cancer.

RNA sequencing
Total RNA was extracted with TRIzol (Invitrogen, Carlsbad, CA, USA). RNA integrity was analysed with an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). A Qubit RNA detection kit and a Qubit Fluorometer (Invitrogen, Carlsbad, USA) were used to measure the RNA concentration. The samples used in subsequent experiments met the following requirements: RNA integrity number (RIN) of tRF-ArgCCU down, 2.0; and 28S:18S ratio, 1.5. A sequencing library was generated and sequenced by Boao Biotechnology (Beijing, China), with 5 µg of RNA per sample. Ribosomal RNA (rRNA) and linear RNA were removed from total RNA using a Ribose Zero Magnetic Kit (Epicenter Technologies, Madison, WI, USA) and RNase R (Epicenter Technologies, Madison, WI, USA), respectively. The NEBNext Ultra RNA Library Preparation Kit from Illumina (Nebraska, USA) was used to construct a sequencing library according to the manufacturer's protocol. In NEBNext rst-strand synthesis reaction buffer (5x), RNA is fragmented into fragments of approximately 300 base pairs (bp) in length. First-strand cDNA was synthesized by adding reverse transcriptase and random hexamer primers to the RNA fragments, and second-strand cDNA was synthesized with 10x dUTP Mix in second strand synthesis reaction buffer. The end of the cDNA fragment undergoes an end repair process, which involves adding a single base and then connecting the Illumina sequencing adapter, and the cDNAs were digested with USER enzyme (NEB, USA) to construct a strand-speci c library. The generated libraries were then puri ed and subjected to PCR ampli cation before they were certi ed on an Agilent 2100 Bioanalyzer system and quanti ed using a KAPA Library Quanti cation Kit.

Database analysis
High and low expression levels of NSUN2 were de ned as greater and less than the median of the values of transcripts, respectively. NSUN2 expression in colon cancer and other cancers was analysed using the Gene Expression Pro ling Interactive Analysis (GEPIA) (http://gepia.cancer-pku.cn) and Human Cancer Metastasis Database (HCMDB) (http://hcmdb.i-sanger.com) online databases.
Stably infected Colon cancer cells were selected for further experiments via culture in puromycin (5 µg/ml) for 1-2 weeks.

Cell migration and invasion assay
The migration and invasion of Colon cancer cells were evaluated by Transwell assays, which were conducted with CytoSelect™ 24-well Cell Invasion assay kits. Brie y, polycarbonate lters (8-µm pore size, Corning) with or without 50% Matrigel (BD Bioscience, Bedford, MA) coating in the upper chamber were inserted into the wells of a 24-well plate for assessment of migration and invasion, respectively. Note that Matrigel was not required for the migration assay. Then, after cell transfection for 48 h, cell suspensions were seeded into the upper chamber, while 500 µl of complete Dulbecco's modi ed Eagle's medium was placed into the lower chamber. After 24 h of culture at 37°C, the cells were xed in methanol for 20 minutes, washed with phosphate-buffered saline (PBS) 3 times, and stained with 2% crystal violet solution. Images of migrated and invaded cells in each well were captured, and the number of cells in 3-5 randomly selected elds under a microscope was counted. Each assay was conducted in triplicate.

RIP-qPCR
The RIP assay was performed as described previously [28] with a Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, USA). Brie y, RKO and SW480 cells were lysed with RIP lysis buffer before A/G magnetic beads were used to immunoprecipitate antibodies that targeted RNA-binding proteins (RBPs). The magnetic bead binding complex was xed with a magnetic frame, and unbound material was washed away. RNA was extracted with TRIzol (Invitrogen) and subjected to qRT-PCR analysis.

RNA isolation and quantitative real-time PCR (qRT-PCR)
Total RNA was prepared from cell lines and tissues using TRIzol reagent (Thermo Fisher Scienti c) according to the manufacturer's instructions. Then, total RNA was treated with TURBO DNase (AM2239; Invitrogen) for 30 minutes for further puri cation following the manufacturer's instructions. The RNA in each sample was washed with ethanol before its concentration and purity were assessed using a NanoDrop ND2000 (Thermo Scienti c Inc., Waltham, MA) following the kit's instructions. Reverse transcription was performed using a cDNA Synthesis Kit (TaKaRa, Beijing, China).
qRT-PCR for each sample comprised 1 µl of cDNA, 3.5 µl of RNase-free water, 0.5 µl of a predesigned Nsun2 probe (Hs00214829_m1; Applied Biosystems) and 5 µl of TaqMan Fast Universal PCR Master Mix (4366073). qRT-PCR and data acquisition were performed with a Real-Time PCR System (Applied Biosystems).
The tRFs were reverse transcribed using Moloney murine leukaemia virus (M-MLV) reverse transcriptase (Promega, Madison, WI, USA). The forward and reverse primers of tRFs were designed and synthesized (Supplementary Table S2). qRT-PCR ampli cations were performed using SYBR Green qPCR Master Mix (ABI, 4367659) on an ABI 7300 real-time PCR detection system (the main primers used in this study are shown in Table S1).

Western blotting assay
Proteins were collected from cell lines following the manufacturer's protocol (KeyGEN BioTech). The protein concentration of each sample was assessed using the Pierce BCA Protein Assay kit (23225; Thermo Fisher) according to the manufacturer's instructions and a spectrophotometer with SoftMax software. Total protein from lysates was separated using sodium dodecyl sulfate-polyacrylamide gel vertical electrophoresis and then transferred onto polyvinylidene uoride membranes (PVDF, Millipore, USA). The membranes were immersed with speci c primary antibodies and then probed with secondary antibodies after they were washed three times with 0.1% Tris-HCl with Tween-20 (TBST). The primary antibodies used in this study were as follows: anti-MMP-2 (Abcam, ab92536, 1:1000), anti-MMP-7 ( (1:1). Eight weeks after injection, mice were killed, and metastatic lung tumours were analysed. The largest tumor was taken from the mouse lung and its volume was measured. The tumour volume was calculated using the formula V = 1/2 × larger diameter × (smaller diameter) 2 . Lung metastasis models, and tail vein injection models were performed. Metastases were detected using the IVIS Lumina II system (Caliper Life Sciences, Hopkinton, MA) for 10 minutes to observe GFP uorescence in the lung.

Statistics
Each assay was independently performed at least 3 times. Statistical analyses were performed using GraphPad Prism 8.0.3 software. Student's t-test (two tailed) and ANOVA were used to detect differences between two groups or among more than two groups, respectively. The Western blot results were quanti ed by ImageJ software (National Institutes of Health). All data are shown as the means ± standard deviation (SD).

Identi cation of signi cantly changed tRFs and their roles in Colon cancer metastasis under hypoxia
In this study, we observed that the expression of 18 tRFs differed signi cantly in colon cancer cells under hypoxic conditions (Fig. 1a-c). Among them, three tRFs showed the most signi cant difference, with two showing decreased expression decreased and one showing increased expression. Interestingly, all three were derived from tRNAs corresponding to different codons encoding Arg; the tRFs were labelled as tRF-ArgCCG down-1, tRF-ArgCCU down-2 and tRF-ArgUCG up-1. Therefore, tRF mimics and tRF inhibitors were transfected into RKO, SW480 and HCT116 cells. The transfection e ciency was detected by PCR assay, and then the the effect of key tRFs on DNA synthesis activities were detected by ow cytometry and the analysis showed no remarkable differences among three groups (Additional le 1: Figure S1g, h). Migration abilities were observed by transwell assays. The transwell assays results showed that downregulation of tRF-ArgCCG down-1 and tRF-ArgCCU down-2 could be associated with advanced progression of colon cancer cells, whereas upregulation of tRF-ArgUCG up-1 could drives colon cancer cells metastasis in vitro (Fig. 1d-k). **P < 0.01. The assays were repeated more than three times.
tRNA Arg C34 methylation exerted critical role in cleavage and colon cancer metastasis By comparing the sequences of the three tRFs, we found base differences at C34. Notably, C34 happens to be the rst base of the antisense codon and is a potential site for m5C modi cation. More importantly, the location of the differential tRF on the corresponding tRNA and C34 are as follows. C34 of tRF-ArgUCG up-1 corresponding to tRNA-Arg is uracil (U) rather than C, while the same position on tRF-ArgCCG down-1 and tRF-ArgCCU down-2 corresponding to tRNA-Arg is C (Fig. 2a). To verify that C34 is indeed the tRNA methylation modi cation site and is the root cause of the differential expression of tRF-ArgCCG down-1, tRF-ArgUCG up-1 and tRF-ArgCCU down-2 under hypoxic conditions, we mutated the C at position 34 to U in tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2 and the U at position 34 to C in tRNA-ArgUCG up-1. These constructs were transfected into cells, and the levels of tRF-ArgCCG down-1 and tRF-ArgCCU down-2 increased while those of tRF-ArgUCG up-1 decreased (Fig. 2b-d). Subsequently, we arti cially methylated position 34 of tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2, and the methylated e ciency was detected by mass spectrum (Additional le 2: Figure S2a, b) and the qPCR assays results showed that the expression of tRF-ArgCCG down-1 and tRF-ArgCCU down-2 decreased (Fig. 2e-g).
To determine the key role of C34 demethylation of tRNA-Arg, we investigated C34-methylated colon cancer cells. Wild-type and C34-methylated tRF-ArgCCU down-1 and tRF-ArgCCU down-2 constructs and the blank control were transfected into RKO, SW480, and HCT116 cells. The effects of the wild-type, C34 methylation of tRNA ArgCCU down-1 and tRF-ArgCCU down-2 on the proliferation of RKO, SW480 and HCT116 cells were examined by MTT assay. The results showed that tRNA ArgCCU down-1 and tRF-ArgCCU down-2 could inhibit colon cancer cells proliferation. However, after tRNA Arg C34 methylation, the inhibition lessened (Additional le 2: Figure S2c, d, e). Interestingly, the results showed that compared to the NC groups, the tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2 groups could obviously inhibit colon cancer cell invasion and migration. However, transfection with mutant tRNA-ArgCCG down-1 or tRNA-ArgCCU down-2 reduced the inhibition of colon cancer cell invasion and migration compared to that of cells without m5C methylation at C34 (Fig. 2h-s). *P < 0.05, **P < 0.01. ***P < 0.001. The assays were repeated more than three times.

NSUN2 overexpression distinguished colon cancer with predictive values
Sequencing results showed that there were 16 kinds of m5C methylation-related enzymes and others associated with RNA cleavage, which expressed signi cantly different compared to normoxia group (Fig.  3a, b). qRT-PCR results and Western Blot were performed to verify the sequencing results and all showed that the expression of NSUN2 is signi cantly increased under hypoxic conditions (Fig. 3c, d, e). Comparisons between metastatic and nonmetastatic foci in the database were performed to screen out differences in NSUN2 expression and the potential correlation between NSUN2 and colon cancer. NSUN2 expression in primary colon cancer tumours is obviously higher than that in the primary normal tissue (Fig. 3f, g). Notably, NSUN2 expression in primary tumours that metastasized was much higher than that in tumours without metastasis (Fig. 3h). These data indicate that NSUN2 may contribute to the progression of colon cancer. We also analysed the expression of other NSUN methyltransferases (NSUN 3-7) but failed to obtain positive results (Fig. 3i-r). This also veri ed the conclusion we made in our previous work that NSUN2 maight play a key role in modulating progression of colon cancer. Thus, we mainly focused on NSUN2 in this study. *P < 0.05. The assays were repeated more than three times. NSUN2 served as a promoter of colon cancer RKO and SW480 cells were treated with a short hairpin RNA speci c for NSUN2, resulting in approximately 90% reduction of its level when compared to that of PLK0.1 group (Figure Sa, b). Further, after the NSUN2 overexpression plasmid was transfected into colon cancer cells, the expression of NSUN2 mRNA and protein levels in the cells were signi cantly increased ( Figure S3c, d). PCR assay results showed that upon NSUN2 overexpression, the levels of tRF-ArgCCG down-1 and tRF-ArgCCU down-2 decreased while those of tRF-ArgUCG up-1 remained basically unchanged (Fig. 4a-f). In contrast, after NSUN2 knockdown, the expression of tRF-ArgCCG down-1 and tRF-ArgCCU down-2 increased, while the expression of tRF-ArgUCG up-1 remained basically unchanged ( Fig. 4i-n). Interestingly, transwell assays analysis showed that overexpression of NSUN2 could increase the invasion and migration ability, while down-regulation of NSUN2 could inhibit the invasion and migration ability of colon cancer cells (Fig. 4g, h, o, p). NSUN2 overexpression led to a decreased cell proliferation rate in the NSUN2 overexpression group compared to that of the pcDNA 3.1/vector group ( Figure S3e, f). A ow cytometry assay further demonstrated G2/M arrest were observed after NSUN2 overexpression ( Figure S3g, h). ns > 0.05, *P < 0.05, **P < 0.01. The assays were repeated more than three times.

NSUN2 exerts effects on the EMT and metastasis-related markers of colon cancer
Based on the fact that the epithelial-mesenchymal transition (EMT) is one of the dominating procedure and mechanisms for tumor metastasis, we further evaluated the effects of NSUN2 on EMT-related markers to further verify the ability of NSUN2 to promote the invasion and migration of colon cancer cells. Western blot analysis showed that NSUN2 overexpression could increase the expression of mesenchymal markers (such as N-cadherin, and Vimentin), the EMT-related transcription factors (ZEB1, Slug, Snail), invasion-related enzymes (such as MMP2, MMP3, MMP7, MMP9) and decrease the expression of epithelial markers (E-cadherin) (Fig. 5a, b, n, o). QPCR assays were performed to verify the changed mRNA expression of EMT-related markers after NSUN2 overexpressed plasmids transfection in RKO and SW480 cells (Fig. 5c-m, 5p-z). These results indicated that NSUN2 could in uence EMT process of colon cancer cells to modulate colon cancer progression. *P < 0.05, **P < 0.01. ***P < 0.001. The assays were repeated more than three times.
The interaction between NSUN2 and methylated C34 of tRNA-Arg determined tRNA-Arg cleavage RIP-qPCR was applied to detect the changes in the binding between NSUN2 and tRNA-ArgCCU down-1, tRNA-ArgCCU down-2 after using NSUN2 with ag. As expected, when compared to the input group, NSUN2 overexpression markedly increased the binding between NSUN2 and both tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2. Thus, the analysis of qPCR results showed obvious elevated expression of tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2 after NSUN2 pulldown (Fig. 5a, b). Furthermore, we used an m5C antibody to pull down tRNA-Arg with m5C methylation. The RIP-qPCR results showed that when NSUN2 was overexpressed, tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2 were enriched upon pulldown with an m5C-speci c antibody.
To further elucidate whether position 34 of tRNA-Arg determines its enzymatic cleavage and whether NSUN2 is a key methyltransferase for cysteine modi cation at position 34 of tRNA, we transfected tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2 with mutations at position 34 into colon cancer cells and detected the levels of tRF-ArgCCG down-1 and tRF-ArgCCU down-2, respectively, after overexpression of NSUN2. Analysis revealed that consistent with our previous results, the levels of tRF-Arg CCG down-1 and tRF-ArgCCU down-2 in cells transfected with mutant tRNA were markedly lower than those in cells transfected with wild-type tRNAs after over-expression of NSUN2. In cells transfected with wild-type tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2, overexpression of NSUN2 led to reductions in the levels of tRF-ArgCCG down-1 and tRF-ArgCCU down-2. However, in cells transfected with position 34 mutants of tRNA-ArgCCG down-1 and tRNA-ArgCCU down-2, NSUN2 overexpression hardly played a role in lowering the levels of tRF-ArgCCG down-1 and tRF-ArgCCU down-2 (Fig. 6f, h, j, l). Thus, the results demonstrate that NSUN2 is a key methyltransferase that functions by methylating cysteine at position 34 of tRNA, which, under hypoxic conditions, protects tRNA from cleavage. **P < 0.01. ***P < 0.001. The assays were repeated more than three times.

NSUN2 promotes colon cancer metastasis partly via tRNA-ArgCCG down-1 in vitro and in vivo
To further verify the relationship between the m5C methyltransferase activity of NSUN2 and tRNA-ArgCCG down-1, we used lentivirus to stably overexpress NSUN2 and tRNA-ArgCCG down-1 either separately or simultaneously in cells. It was found that overexpression of tRNA-ArgCCG down-1 can partially eliminate the effect of NSUN2 on suppressing tRNA-ArgCCG down-1 expression (Fig. 7a, c). More interestingly, tRNA-ArgCCG down-1 had a negative feedback effect on NSUN2 expression (Fig. 7b, d). To further illuminate the in uence of both NSUN2 and tRNA ArgCCG down-1 overexpression on the migratory and invasive activity of colon cancer cells, we performed transwell assays and found that up-regulation of tRNA ArgCCG down-1 could signi cantly break role of promoting metastasis of NSUN2 in vitro (Fig. 7e, f).
To investigate the effect of tRNA-ArgCCG down-1 on tumour formation and metastasis in vivo both independently and concomitantly with NSUN2 overexpression, CT-26 cells were implanted subcutaneously into the right ank of BALBC white mice as an orthotopic PC model or intravenously into the tail vein to observe lung metastasis. The BALBC white mice were divided into four groups, pcDNA 3.1/ NC mimics, NSUN2+/ NC mimics, pcDNA 3.1/ tRNA ArgCCG down-1 mimics, NSUN2+/ tRNA ArgCCG down-1 mimics, respectively. Furthermore, we developed a lung metastasis model (via tail vein injection) to highlight the effect of NSUN2 and tRNA-ArgCCG down-1 and used an in vivo optical imaging system (IVIS) for small mice. The most typical picture is showed as followed. BALBC white mice implanted with NSUN2+/NC CT-26 cells exhibited signi cantly higher GFP activity in the lung than those implanted with NC cells. In contrast, the GFP signal of mice bearing pcDNA 3.1/tRNA-ArgCCG down-1 mimic RKO cells was much lower than that of mice bearing NC cells. Notably, the GFP signal of mice in the NSUN2+/tRF-ArgCCG down-1 mimics group was similar to that of the NC group (Fig. 7g), which indicated that tRF-Arg CCG down-1 could inhibit tumour metastasis in vivo and weaken the role of NSUN2 promoting metastasis. Then we removed the lung tissues of the mice, stained it and took pictures, and got similar results (Fig. 7h). The largest tumor was taken from the mouse lung and its volume was measured. The tumour volume of each group is V1 = 0.259 mm 2 , V2 = 0.668 mm 2 , V3 = 0.118 mm 2 , V4 = 0.263 mm 2 , respectively. *P < 0.05, **P < 0.01. ***P < 0.001. The assays were repeated more than three times.

Schematic diagram of overall study
Under hypoxic stress conditions, tRNA is cleaved into tRFs, some of which show increased levels (tRF-Arg UCG up-1) and others that show decreased levels (tRF-Arg CCG down-1, tRF-Arg CCU down-2). The methylation of C34 on the source tRNA point and the level of methylation were analysed. Furthermore, the role of the key m5C methyltransferase NSUN2, which regulates the delicate balance of endonucleasemediated tRNA cleavage, in the changes in tRF levels was clari ed, and the effects of selective tRNA cleavage on metastatic potential were veri ed (Fig. 8).
Discussion tRFs can regulate a variety of biological processes [29]. Aberrant expression of tRFs have been observed in a variety of human diseases, including cancer, neurodegenerative diseases, and acquired metabolic diseases [29][30][31]. In this study, we found that the antisense codons of tRNA-Arg encoding different codons of Arg protected tRNA-Arg via m5C methylation at their respective bases at C34, thus reducing the sensitivity to of these tRNAs to enzyme digestion under stress conditions. Here, we observed a trend of increased NSUN2 expression in colon cancer cells under hypoxic conditions and determined that the upregulation of NSUN2 expression was signi cantly associated with tumour metastasis. We demonstrated that NSUN2 is the key methyltransferase responsible for the m5C modi cation at C34 of tRNA-Arg in colon cancer and is a prognostic factor for colon cancer. In addition, we found that NSUN2 promoted colon cancer migration and invasion in vitro and in vivo. Consistent with this, clinical data showed that there were higher levels of NSUN2 in tumour tissues than in adjacent normal tissues and that these increases were associated with poorer outcomes in patients.
M5C is an abundant RNA modi cation that is present in a wide variety of RNA classes, including cytoplasmic and mitochondrial ribosomal RNAs (rRNAs) and tRNAs, as well as messenger RNAs (mRNAs), enhancer RNAs (eRNAs) and a number of noncoding RNAs. The m5C modi cation is involved in many mammalian processes, such as RNA cleavage, protein translation, and stem cell renewal [32]. In eukaryotes, C5-methylation of Cs in RNA is catalysed by enzymes of the NOL1/NOP2/SUN domain (NSUN) family and by DNMT2. The rst methylated substrate of NSUN2 was identi ed as tRNA [4].
Previous studies identi ed three direct tRNA targets of human NSUN2: tRNA-Leu (C34), tRNA-Asp (C48, 49) and tRNA-Gly (C48, 49, 50) [33][34][35]. Brzezicha et al. showed that NSUN2 mediated methylation in an intron-dependent manner at C34 of pre-tRNA-Leu in HeLa cells [34]. Here, upon comparing the sequences of the three tRFs with changes in their levels, we discovered different bases at C34. Notably, position 34 is the rst base of the antisense codon, which is a potential site for m5C modi cation. More importantly, position 34 of tRF-ArgUCG up-1 corresponding to tRNA-Arg is U rather than C, while same position of tRNA ArgCCG down-1 and tRF-ArgCCU down-2 corresponding to tRNA-Arg is C. The results suggest that the tRNA-Arg antisense codons encoding Arg protect C34 of tRNA-Arg via NSUN2-mediated m5C methylation in colon cancer cells.
NSUN2 is involved in a variety of biological pathways [15], and overexpression of NSUN2 has been found in a variety of human cancer types [16]. In recent years, many studies have revealed the abnormal expression of NSUN2 in cancer and the role it plays in the pathogenesis and development of cancer [36,37]. Tatsuka et al. demonstrated that NSUN2 expression was abnormally upregulated in breast cancer [16]. A previous study also found that DNA hypomethylation is associated with increased NSUN2 expression in breast cancer. Overexpression of NSUN2 promotes cell proliferation, migration and invasion, while knocking out NSUN2 inhibits these processes in vivo and in vitro (overexpression of NSUN2 due to DNA hypomethylation is associated with metastatic progression in human breast cancer).
In skin cancer, when the expression levels of NSUN2 were compared in normal skin with early and advanced skin cancers, loss of NSUN2 expression was associated with increased malignancy of the tumour and with an increase in the number of tumour-initiating cells [38].
High expression of NSUN2 was observed in patients with stage III/IV colon cancer [15], and this observation was subsequently con rmed by other studies [16]. However, few studies have focused on the role and underlying mechanism of NSUN2 in the development of colon cancer. Interestingly, a circRNA derived from the NSUN2 coding sequence named circNSUN2 (hsa_circ_0007380) was recently found to be upregulated in patients with colon cancer and predicted a poor prognosis [17].In models of patientderived xenotransplantation (PDX), circNSUN2 appears to promote liver metastasis when overexpressed.
In addition, the circNSUN2/IGF2BP2 complex can lead to the promotion of EMT in colon cancer cells when bound to HMGA2 mRNA [17]. In the present study, we found that NSUN2 expression was upregulated in colon cancer, which could result in aberrant m5C modi cation at a speci c site (C34) of tRNA-Arg. Furthermore, increased NSUN2 expression was signi cantly associated with colon cancer progression in vivo or in vitro and indicated a poor prognosis.
Fundamentally, we observed that there were signi cant differences in the expression of tRFs. Among them, the expression of three tRFs showed the most signi cant differences, with two tRFs showing decreased expression and one showing increased expression, but all three were derived from tRNAs corresponding to codons that encoding Arg. This suggests that tRNAs encoding different codons of the same amino acid (e.g, Arg) are differently cleavage and produce different tRF subsets. Through the indepth study, we nally con rmed that NSUN2-mediated splice site m5C methylation by acting on C34, leading to discrepant cleavage outcomes of tRFs. Our previous study con rmed that tRNA-ArgCCG down-1 signi cantly affects the EMT process of colon cancer cells by regulating the expression of Claudin-1 [24]. In this study, the differential expression of these tRF subsets showing contradictory effects on the metastatic potential of colon cancer. In the present study, we con rmed that NSUN2 was the key methyltransferase in the methylation of the C at C34 in tRNA and this modi cation could reduce the cleavage e ciency of enzymes and protect tRNA from cleavage under hypoxic stress conditions. In the present study, we rst showed that NSUN2 expression was signi cantly upregulated in colon cancer cells under hypoxic conditions. Moreover, NSUN2 can promote the metastasis of colon cancer cells in vitro and in vivo. Further studies revealed that speci c tRFs are potential downstream molecules of NSUN2 regulation in colon cancer. In addition, NSUN2 can promote the metastasis of colon cancer cells in an m5C-dependent manner. This study suggests that NSUN2-mediated m5C methylation of tRFs is a novel mechanism involved in the development of colon cancer.

Conclusions
The present study on the methylation of noncoding RNA and its effect on abberant transcriptome levels further elaborates the selectivity of m5C tRNA modi cations on regulating the shearing process and nely adjust the mechanism involved. Research on tRFs which derived from tRNA-Arg under hypoxia revealed signi cant differences in their expression at the molecular level, revealed the site differences among these three kinds of tRNA C34 and veri ed that the expression of NSUN2 was up-regulated in colon cancer cells. Signi cantly, NSUN2-mediated splice site m5C methylation led to different cleavage outcomes of tRFs by acting on C34, playing a key role in promoting colon cancer metastasis. Our study focuses on changes in the tumour epigenetics transcriptomics with respect to tumour metastasis potential, and provided a deeper understanding of tumour metastasis in the molecular regulatory mechanism of initiating.

Abbreviations
EMT epithelial-to-mesenchymal CRC colorectal cancer

Consent for publication
Not applicable.

Availability of data and material
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. wrote the draft. YC, JS, ZT, and JW revised the manuscript. All authors have read and approved the nal manuscript. The red shades and blue shades represent high expression and low expression, respectively. b Signi cantly discrepant expressed tRFs were found by high-throughput sequencing analysis. c RT-PCR analysis was used to verify the expression of 18 kinds of tRFs in cell lines. Actin was the internal control.
d, e, h, i Transwell migration and invasion assays were performed to detect the motility of RKO and SW480 cells transfected with tR down-1 inhibitor, tRF-ArgCCU down-2 inhibitor. Columns are the average of three independent experiments. f, g, j, k tRF-ArgUCG up-1 overexpression led to increased migration and invasion of RKO and SW480 cells.

Figure 2
Key point of tRNA Arg plays a key role in tRNA leavage and tRF-mediated metastasis. a The location of the differential tRF on the corresponding tRNA and the position 34 base. The underlined is the antisense codon, and the red mark is the position of tRF. b, c, d qPCR assays identify the tRF-ArgCCG down-1, tRF-ArgUCG up-1 and tRF-ArgCCU down-2 expression of RKO, SW480 and HCT116 cells transfected with wildtype and mutant tRNA-5/6/7 24h. e, f, g qPCR to examine the expression level of tRF ArgCCG down-1 and tRF-ArgCCU down-2 in RKO/SW480/HCT116 cells transfected with WT and added methylated m5C tRNA ArgCCU down-1 and tRNA ArgCCU down-2. h-s compared to NC groups, tRNA ArgCCG down-1 and tRF-ArgCCU down-2 groups could remarkably inhibit colon cancer cells invasion and migration. However, when compared to that of C34 without m5C methylation, tRF-ArgCCG down-1 and tRF-ArgCCU down-2 produced from C34 methylated tRNA ArgCCG down-1 and tRNA ArgCCU down-2 weakened the ability of colon cancer cells invasion and migration. Figure 3 NSUN2 overexpression characterized colon cancer with predictive values. a Heatmap for differentially expressed m5C RNA methylated enzymes and RNase in colon cancer tissues under hypoxia compared with that in normoxia conditions. b Further analysis of heatmap. c Relative RNA expression of NSUN2 in RKO and SW480 cells under hypoxia analyzed by RT-qPCR. d, e Relative protein expression of NSUN2 in RKO and SW480 cells under hypoxia analyzed by western blot. f, g, h NSUN2 upregulation is associated with tumour in primary focus as well as advanced progression in colon cancer. i-r other NSUN methyltransferases (such as NSUN 3-7) expressed no signi cantly difference between normal tissue and primary tumour focus, nor primary tumour focus and metastasis focus. NSUN2 played a role as a promoter of colon cancer. a-f PCR assays results showed that after NSUN2 knockdown, the expression of tRNA ArgCCG down-1 and tRF-ArgCCU down-2 increased, while the expression of tRF ArgUCG up-1 remained basically unchanged. g Transwell assays analysis showed that down-regulation of NSUN2 could decrease the invasion and migration ability. i-n After NSUN2 overexpression, the expression of tRNA ArgCCG down-1 and tRF-ArgCCU down-2 decreased, while the expression of tRF ArgUCG up-1 remained basically unchanged. g, h, o, p overexpression and downregulated expression of NSUN2 could regulate the invasion and migration ability of colon cancer cells.  could weaken NSUN2 promoting metastasis ability and damage tRF-ArgCCG down-1 inhibiting metastasis ability by reducing its production. g. The BALBC white mice implanted with NSUN2+/ NC mimics CT-26 cells exhibited signi cantly higher GFP activity in lung, while the GFP signal of mice bearing pcDNA 3.1/ tRNA ArgCCG down-1 mimics RKO cells were much lower than that of mice in the NC group. Signi cantly, the GFP signal of mice in the NSUN2+/ tRF-Arg CCG down-1 mimics group was similar to the NC group. h General view lung tissue of each mice.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.