2.1 Patients and tissue samples

A total of 300 patients in the First Affiliated Hospital of Wenzhou Medical University (January 2014 and December 2016, follow-up to December 2022) and 1500 patients in the Second Affiliated Hospital of Wenzhou Medical University (January 2018 and April 2023, follow-up to June 2024), who underwent surgical resection for CRC, were included in this study as Cohort1 and Cohort2, respectively. All the CRC tissues and paired adjacent normal tissues (at least 10 cm away from the negative margin) were formalin fixed and paraffin embedded to construct CRC tissue microarrays, and ultimately confirmed by histopathologic analysis. Clinical survival characteristics were collected for each patient. After immunohistochemistry (IHC) staining, the spots with more than 50% tissue loss were excluded, and finally 266 CRC tissues and 26 adjacent normal tissues were left in the Cohort1 and 1293 CRC samples in the Cohort2. The study was approved by the Review Board of the First Affiliated Hospital of Wenzhou Medical University (KY2022-202) and Second Affiliated Hospital of Wenzhou Medical University (2021-K-42-01). Written informed consent was obtained from the patients before the study began.

2.2 RNA m5C dot blot assay

Total RNA was extracted from CRC cells, and then treated with deoxyribonuclease I (DNase) and quantified for the purpose of loading varying RNA amounts (400, 600, 800 ng) onto Hybond-N+ membranes (Beyotime). Post a rapid drying phase, the membranes underwent UV crosslinking at 254 nm for 60 s, followed by blocking with 5% milk for 1 h and subsequent overnight incubation at 4°C with an anti-m5C antibody and corresponding secondary antibody. Ultimately, the membranes were analyzed using enhanced chemiluminescence and visualized with the Bio-Rad Imaging System.

2.3 Xenograft and lung metastasis models

For xenograft studies, female BALB/c nude mice (4 weeks) were randomly assigned to one of two groups. CRC cells overexpressing NSUN2, NSUN2 knockout cells and control cells were harvested (2 × 10⁶) and subcutaneously implanted into the flank regions of female BALB/c nude mice (4 weeks). Subsequently, the volumes of developing tumours were estimated based on length and width measurements (length × width² × .5) obtained at 2-day intervals. When tumours had reached a predetermined size, the mice were euthanized, and tumour specimens were excised. Finally, xenografts samples were thereafter sectioned at 5 µm for subsequent immunohistochemical analysis.

For lung metastasis models, NSUN2 overexpressing CRC cells were transfected with lentivirus containing a luciferase reporter gene (OBiO Technology). To generate a lung metastasis model, mice were randomly assigned to one of two groups of nude mice, and we injected CRC cells (5 × 10⁵) and intraperitoneally administered d-luciferin (15 mg/kg, BioGold) into the tail veins. At 5-day intervals, the mice were placed under isoflurane-induced anaesthesia and imaged for luciferin emission using an In Vivo Imaging System (IVIS) Spectrum imaging system (PerkinElmer) to track cancer cell dissemination.

All animals used in this study were purchased from Hangzhou Medical College. All animal experiments were performed in compliance with the Wenzhou University's Policy on the Care and Use of Laboratory Animals (WZU-2024-074/075).

2.4 Lapatinib sensitivity assay

For lapatinib IC50 assays, HCT116 and DLD-1 cells were plated in 96-well plates (5000 cells/well) and transfected with either wild-type NSUN2 (NSUN2-WT) or double mutant NSUN2 (NSUN2-DM) plasmids. Following treatment with a gradient of lapatinib (MCE) concentrations, cells were incubated for 48 h before determining the IC50 using CCK-8 assay.

For in vitro assays, mice were subcutaneously injected with DLD-1 cells overexpressing NSUN2 (OE) or control cells (NC); two groups received daily oral gavage of lapatinib at a concentration of 50 mg/kg. Subcutaneous tumour volumes were subsequently recorded, and when tumours had reached a predetermined size, the mice were euthanized, followed by the excision and documentation of tumour tissues. The tissues were paraffin-embedded and sectioned to 5 µm for subsequent analysis.

2.5 Statistical analysis

PASW Statistics (version 18; IBM), GraphPad Prism 7.0 software, and R program (version 4.2.3) were used for Statistical analysis. Data are presented as the means ± SD or percentage. Statistical comparisons between the two groups were conducted utilizing a two-tailed Student's t-test or χ² test. Kaplan–Meier analysis was employed for survival analysis, and the log-rank test was utilized to compare survival curves. Multivariable Cox regression analysis was performed to exclude confounding factors affecting survival outcomes. The correlation between the expression levels of two genes was evaluated using Spearman correlation analysis. Note that *p < .05, **p < .01 and ***p < .001 were considered statistically significant.

3.1 NSUN2 is highly expressed in CRC and predicts poor prognosis

To evaluate the expression of NSUN2 in human tumour tissues, we noted that NSUN2 was highly expressed in the majority of solid tumours, including colon and rectal adenocarcinoma through TCGA pan-cancer analysis (Figure 1A,B). Next, Kaplan–Meier analysis of CRC patients from TCGA-COAD and GEO (GSE17536, GSE17537 and GSE29621) was employed to assess the role of NSUN2 in the prognosis of CRC patients. As shown in Figure 1C, high NSUN2 expression was correlated with an unfavourable prognosis in both overall survival (OS; p = .016) and disease-free survival (DFS; p = .024). Freshly CRC tissues were then collected and detected to reveal a notably increased expression of NSUN2 protein in CRC tumours compared with normal tissues (Figure 1D). To further investigate the patterns of NSUN2 expression and its correlation with clinical pathological characteristics in patients with CRC, we collected 266 CRC paraffin tissues from First Affiliated Hospital of Wenzhou Medical University (Cohort 1) and another 1293 CRC paraffin tissues from the Second Affiliated Hospital of Wenzhou Medical University (Cohort 2) to construct CRC tissue microarrays. IHC staining for NSUN2 in the CRC tissue microarrays revealed that NSUN2 protein levels were higher in CRC tumour tissues than in adjacent normal tissues (Figure 1E,F). We then divided the patients with CRC into two groups based on the NSUN2 IHC staining scores, that is, the NSUN2-high and NSUN2-low groups. Clinicopathological features analysis showed that NSUN2 high expression was closely associated with positive rate of Ki-67 (Table 1; p < .001). Kaplan–Meier survival analysis revealed that CRC patients in the NSUN2-high expression group had shorter OS (p = .022 in Cohort 1; p < .002 in Cohort 2). In addition, Kaplan–Meier analysis of DFS indicated that high NSUN2 expression was associated with recurrence (Figure 1G–I). Multivariate Cox regression analysis showed that elevated NSUN2 expression served as an independent prognostic indicator for OS (Table 2; Hazard Ratio (HR), 1.72; 95% CI, 1.35–2.19; p < .001). Collectively, these findings suggest that NSUN2 is highly expressed in CRC tissues and correlated with an unfavourable prognosis in CRC patients.

3.2 Overexpression of NSUN2 strengthens CRC cell proliferation and metastasis

To investigate the tumourigenic effects of NSUN2 on CRC cells, we first evaluated the expression characteristics of NSUN2 in CRC cell lines and normal cell lines (FHC). As shown in Figure S1A, NSUN2 was mainly expressed in nucleus and partially expressed in the cytoplasm. Western blot detection revealed that NSUN2 expression was relatively higher in DLD-1 cell lines and slightly lower in HCT116 cell lines (Figure S1B). Consequently, we selected these two cell lines for subsequent experiments, and constructed stable NSUN2-overexpressing HCT-116 and DLD-1 cell lines (Figure 2A). Overexpression efficiency was evaluated by western blotting (Figure 2A). We conducted CCK-8 and EdU assays to evaluate the proliferative effects of NSUN2 in two CRC cell lines with NSUN2 overexpression. As shown in Figure 2B–D, NSUN2 overexpression significantly promoted the proliferation of DLD-1 and HCT-116 cells. Colony formation assays were performed to examine the long-term effects of NUSN2 on CRC cell proliferation. We found that CRC cells overexpressing NSUN2 formed more colonies than the control CRC cells at 2 weeks (Figure 2E). Next, Transwell assays were carried out to evaluate the metastasis capabilities of CRC cells with and without NSUN2 overexpression. As shown in Figure 2E, the migration and invasion of CRC cells were markedly elevated in CRC cells overexpressing NSUN2 than in control CRC cells.

To evaluate the function of NSUN2 in CRC progression in vivo, we subcutaneously injected stable NSUN2-overexpressing HCT-116 cells into nude mice to establish mouse xenograft models (Figure 3A) and monitored the growth of the resulting tumours. As shown in Figure 3B–D, tumour growth in the NSUN2-overexpression group of mice was significantly greater than that in the control group. Moreover, IHC staining confirmed that overexpression of NSUN2 significantly upregulated Ki-67 expression in the xenograft tumour tissues (Figure 3E). The results of the xenograft assay demonstrated that overexpression of NSUN2 promoted CRC progression in vivo.

To monitor the effect of NSUN2 on metastasis of CRC cells in vivo, lung metastasis models were constructed by injecting NSUN2-overexpressing HCT-116 cells into the tail veins of nude mice. The NSUN2-overexpressing HCT-116 cells showed more obvious metastatic ability and formed more tumour foci than the control HCT-116 cells, as shown by the photon number and haematoxylin–eosin (H&E) staining (Figure 3F,G). These results confirmed that overexpression of NSUN2 promotes metastasis of CRC cells in vivo.

3.3 Knockout of NSUN2 inhibits CRC tumourigenesis and progression

To further confirm the biological function of NSUN2 in CRC, we generated NSUN2-knockout HCT-116 and DLD-1 cell lines using the CRISPR/Cas9 system. The knockout efficiency of NSUN2 was verified by western blotting (Figure 4A and Figure S1). CCK-8, EdU and colony formation assays were performed to assess the effect of NSUN2 knockout on proliferative ability, and Transwell assays were performed to assess the effect of NSUN2 knockout on metastatic ability of CRC cells. As shown in Figure 3B–E, the proliferation and metastatic abilities of HCT-116 and DLD-1 cells were significantly suppressed by knockdown of NSUN2. Taken together, these results show that NSUN2 promotes the proliferation and metastasis of CRC cells.

3.4 NSUN2 promotes CRC progression via both m5C-dependent and m5C-independent mechanisms

NSUN2 is a classic nucleolar 5-methylcytosine RNA methyltransferase, and its biological and oncogenic functions are generally believed to depend on its m5C catalytic activity, that is, generation of m5C at target RNAs. Previous studies have shown that both the release site (cysteine 271) and catalytic site (cysteine 321) of NSUN2 are key for its m5C methyltransferase activity. Therefore, we generated two enzymatic-dead NSUN2 mutants by introducing point mutations at cysteine 271 and 321 (Figure 5A). Strikingly, similar to our previous study on gastric cancer cells, we found that overexpression of both the wild-type and enzymatic-dead mutant NSUN2 in HCT116 cells enhanced their proliferation and metastatic abilities (Figure 5B,C).

Next, we constructed an NSUN2-DM to further confirm that the biological function of NSUN2 was not dependent on its m5C modification ability (Figure 5D). M5C dot blot assays were performed to investigate whether the level of m5C mRNA in CRC was affected by the expression of NSUN2-WT or NSUN2-DM. As shown in Figure 5E, m5C levels were significantly decreased when NSUN2 was knocked out, whereas m5C levels were restored when NSUN2-WT, but not NSUN2-DM, was overexpressed in both HCT116 and DLD-1 cells. However, CCK-8 and Transwell assays revealed that overexpression of both NSUN2-WT and NSUN2-DM rescued the proliferation and metastatic abilities of CRC cells after NSUN2 knockout (Figure 5F,G and Figure S2). Taken together, these data suggested that the non-m5C-dependent function of NSUN2 also plays an important role in the malignant progression of CRC.

3.5 The m5C-independent function of NSUN2 activates the ErbB-STAT3 signalling pathway through its interaction with CUL4B

To investigate the molecular mechanism underlying the m5C-independent promotion of CRC progression for NSUN2, we attempted to identify the signalling pathways downstream of NSUN2. Previous research have found that activation of the ErbB pathway is closely correlated to m5C levels and the expression levels of its regulators. Therefore, we explored whether an enzymatic-dead NSUN2 mutant could upregulate the ErbB signalling pathway. We noted substantial decreases in the expression levels of HER2, phosphorylated EGFR and phosphorylated STAT3 in NSUN2-knockout CRC cells, whereas NSUN2-WT overexpression reversed the reductions in these levels. As anticipated, complementation with m5C enzymatic-dead NSUN2-DM also restored these levels, although the effect was not as pronounced as that for NSUN2-WT (Figure 6A). These results indicate that NSUN2 activates the ErbB signalling pathway in both an m5C-dependent and m5C-independent manner.

In pursuit of understanding the mechanism underlying the m5C-independent regulation of the ErbB signalling pathway by NSUN2, we attempted to identify the proteins that interact with NSUN2. We found that CUL4B bound tightly to NSUN2 using co-IP (Figure 6B). Additionally, overexpression of CUL4B in CRC cells also immunoprecipitated NSUN2 (Figure 6B), and both NSUN2-WT and NSUN2-DM interacted with CUL4B. We used a molecular docking model to predict the binding affinity and interacting regions between NSUN2 and CUL4B (Figure 6C). The computational analysis indicated a stability of −239.9 kcal/mol for these two proteins. Subsequently, to validate the domain that interacts with CUL4B, we constructed a mutant NSUN2 with a 216–270 amino acid deletion. As shown in Figure 6D, deletion of the 216–270 amino acid segment resulted in the loss of interaction between NSUN2 and CUL4B, thereby confirming that this region of the NSUN2 protein is essential for its interaction with CUL4B. These results indicate that NSUN2 interacts with CUL4B independent of its m5C modification activity and that it may exert its biological functions through CUL4B.

A preceding investigation has delineated that HER2 is a downstream target of CUL4B in gastric cancer. Therefore, we speculated that NSUN2 might interact with CUL4B to regulate the expression of the HER2 signalling pathway. CUL4B, a constituent of the largest family of ubiquitin E3 ligases, is implicated in a myriad of biological processes. Previous studies have shown that CUL4B is primarily involved in the formation of the CRL4B complex, which recognizes and ubiquitinates downstream substrate proteins. Recent studies have shown that CUL4B directly modulates downstream genes expression at the transcriptional level. Therefore, we evaluated the biological effects exerted by CUL4B in CRC cells (Figure S3). CUL4B knockdown via siRNA markedly suppressed both the proliferation and metastatic capabilities of CRC cells. Next, we knocked down CUL4B in CRC cells while simultaneously overexpressing either NSUN2-WT or NSUN2-DM and evaluated the changes in the ErbB signalling pathway and malignant phenotype of CRC cells. The expression levels of HER2, phosphorylated EGFR and phosphorylated STAT3 decreased upon CUL4B knockdown, and only overexpression of NSUN2-WT restored the levels of these proteins. Overexpression of NSUN2-DM did not rescue the defects in the ErbB signalling pathway (Figure 6D). Similarly, overexpression of NSUN2-WT partially restored the metastatic ability of CUL4B-knockdown HCT116 and DLD-1 cells, whereas overexpression of NSUN2-DM did not (Figure 6E–G and Figure S3C). Collectively, these findings indicated that the m5C-independent functions of NSUN2 might depend on CUL4B expression.

3.6 NSUN2 enhances the therapeutic efficacy of lapatinib in CRC

The application of molecular targeted inhibitors has brought a revolution in cancer therapy. These agents exhibit higher selectivity and fewer side effects as they specifically target molecular markers on cancer cells, thereby inhibiting tumour growth and metastasis. Therefore, we aimed to identify inhibitors that could target CRC with high NSUN2 expression. Among these, lapatinib, a potent dual EGFR-HER2 inhibitor, attracted our attention. We found that overexpression of NSUN2-WT increased the sensitivity of CRC cells to lapatinib. Interestingly, a similar effect was observed in cells overexpressing the m5C enzymatic-dead NSUN2-DM (Figure 7A). To further validate the therapeutic effect of lapatinib on CRC with high NSUN2 expression, a xenograft model was established in nude mice through the subcutaneous implantation of NSUN2 overexpression DLD1 cells. We then treated the mice with lapatinib, (orally, at 50 mg/kg per day). Evaluation of subcutaneous tumour volume and their growth curves showed that although the differences between the NC- and NSUN2-overexpressing groups were not significant after lapatinib treatment, the size of NSUN2-overexpressing tumours was significantly reduced compared with the untreated NSUN2 group after lapatinib treatment (Figure 7B–D). H&E staining and assessment of Ki-67 levels using IHC staining showed the level of Ki-67 in the subcutaneous tumour tissue in each group (Figure 7E). Additionally, the expression correlations of NSUN2, CUL4B, EGFR and HER2 were examined via IHC in xenografts and CRC patient tissues. The results indicated a positive correlation in the expression levels of these proteins (Figure S4 and Figure 7F). Spearman's analysis revealed significant positive correlations between NSUN2 and CUL4B (R = .672; p < .001), EGFR (R = .477; p < .001) and HER2 (R = .503; p < .001) in CRC patients (Figure 7E). Collectively, these findings revealed that NSUN2 upregulates EGFR/HER2-STAT3 pathway and enhances the therapeutic efficacy of lapatinib in CRC.