Lactylation-driven NSUN2-mediated RNA m5C modification promotes perineural invasion in pancreatic cancer

Abstract

Background: Perineural invasion (PNI) is a key biological feature underpinning the high malignancy and poor prognosis of pancreatic ductal adenocarcinoma (PDAC). Lysine lactylation (Kla), a metabolite-stress-induced post-translational modification, plays crucial regulatory roles in diverse biological processes. The RNA methyltransferase NSUN2 is essential for cancer invasion and metastasis. However, the mechanisms by which NSUN2 contributes to lactylation-driven PNI in PDAC remain to be elucidated. Methods: We assessed tumor lactate / pan-lactylation, NSUN2 lactylation, and PNI in human PDAC cohorts with survival follow-up. Functional studies used PDAC cell lines for migration/invasion assays, dorsal-root-ganglion (DRG) co-culture, and neurite-outgrowth assays under lactate or enzymatic perturbations. Mechanistic interrogation combined NSUN2 knockout, CRISPR knock-in mutants at K692 (K692R/E), co-immunoprecipitation, RIP-seq, MeRIP-qPCR, and actinomycin-D chase to test mRNA binding, m5C modification, and stability of CDCP1/STC1. In vivo validation employed a sciatic nerve invasion model and a KPC genetically engineered mouse model to assess tumor-nerve infiltration and disease progression. Results: Lactylated NSUN2 is markedly upregulated in mice and human PDAC with more severe PNI, and is significantly associated with poorer prognosis. Functionally, inhibiting lactylation or blocking NSUN2 markedly attenuated tumor-nerve interactions and neural invasion. Mechanistically, lactate accumulation leads to the lactylation of NSUN2 at lysine 692 (K692), subsequently inhibiting its ubiquitination and degradation. lactylation of NSUN2 mediated m5C modification on CDCP1 and STC1 mRNA, enhanced their mRNA stability. Conclusions: This study identifies lactate-driven NSUN2 K692 lactylation as a key driver of perineural invasion in PDAC. We define a lactate-NSUN2-m5C-CDCP1/STC1 axis that links metabolic stress-induced lysine lactylation to mRNA methylation-dependent stabilization of pro-invasive transcripts, highlighting actionable therapeutic targets to restrain neural invasion and improve patient outcomes.

Keywords: NSUN2; lactylation modification; m5C; pancreatic ductal adenocarcinoma (PDAC); perineural invasion.

Figure 1

Elevated lactate and lactylation levels are associated with poor prognosis and PNI in PDAC. (A) Representative mIF staining for Pan-lac, Tuj1 and CK-19 expression in PDAC tissues grouped by PNI severity. Pan-lac reports lactylation; Tuj1 labels neurons; CK-19 identifies epithelial cancer cells. Scale bars, 200 μm. (B) Quantification of tissue lactylation levels in PDAC cases with or without severe PNI. (C-D) Kaplan-Meier survival curves showing OS (C) and DFS (D) in the PDAC cohort stratified by Pan-lac level (high vs low). (E) Immunoblot analysis of lactylation modification levels in PANC-1 cells following treatment with varying concentrations of L-lactate. (F) Immunofluorescence detection of Pan-lac in tumor cells treated as indicated. Scale bar, 20 μm. (G) PANC-1 cells subjected to LDHA/LDHB knockdown and L-lactate exposure were analyzed by Western blot to determine lactylation levels. (H-M) PANC-1 cells expressing siNC, LDHA siRNA, LDHB siRNA, or expressing both LDHA and LDHB siRNAs, followed by ± L-lactate treatment for subsequent experiments. (H) Schematic diagram of tumor cells in a Transwell assay. (I) PANC-1 Transwell migration/invasion: image panels (left) and measurements (right). Scale bar, 200 μm. (n = 3; one-way ANOVA; mean ± SD). (J) Schematic representation of the neuronal-tumor Transwell co-culture system. (K) Quantitative analysis of neurite outgrowth under the indicated conditions in the Transwell co-culture model. Phase-contrast images were obtained at 6 h intervals. (n = 3; two-way ANOVA; mean ± SD). (L) Schematic representation of the direct co-culture model between tumor cells and dorsal root ganglia (DRG). (M) (top) Representative fields from DRG co-cultures with tumor cells. (bottom) Summary statistics for tumor neurite invasion toward DRG. The black dashed line on the left indicates the growth boundary of the DRG, which on the right marks the growth boundary of PANC-1 cells. Scale bar, 500 μm. (n = 3; one-way ANOVA test; mean ± SD). ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05.

Figure 2

Lactate upregulates m5C and NSUN2 levels to promote PNI in PDAC. (A) Dot blot analysis showing m5C modification levels in total RNA isolated from PANC-1 cells treated with or without L-lactate. Methylene blue staining (below) indicates RNA loading, whereas the dot blot signal intensity (above) reflects the global m5C content. (B) Quantitative RT-PCR analysis of NSUN1, NSUN2, NSUN3, NSUN5, NSUN6, NSUN7 and DNMT2 mRNA expression levels in PANC-1 cells followed by ± L-lactate treatment. (C) Immunoblot analysis of NSUN1, NSUN2, NSUN3, NSUN5, NSUN6, NSUN7 and DNMT2 protein expression levels in PANC-1 cells under the same treatment conditions. (D) Representative mIF staining for NSUN2, Tuj1 and CK-19 grouped by PNI severity in PDAC tissues. Tuj1 marks neuronal cells, and CK-19 identifies tumor epithelial cells. Scale bars, 200 μm. (E) Correlation between NSUN2 expression levels and lactylation modification in 24 cases of PDAC tissues. Pearson correlation analysis was applied to estimate the coefficient (R) and significance level (P). (F-G) Kaplan-Meier survival curves showing OS (F) and DFS (G) in the PDAC cohort stratified by NSUN2 level (high vs low). (H-J) Functional tests were performed on PANC-1 cells transduced with either control vector or NSUN2-OE plasmid. (H) Transwell-based migration/invasion: image panels (left) and measurements (right). Scale bar, 200 μm. (n = 3; unpaired t test; mean ± SD). (I) Neurite extension in the Transwell co-culture was quantified under the specified conditions; phase-contrast images were acquired at 6-h intervals. (n = 3; paired t test; mean ± SD). (J) (left) Representative fields from DRG co-cultures with tumor cells. (right) Summary statistics for tumor neurite invasion toward DRG. Scale bar, 500 μm. (n = 3; unpaired t test; mean ± SD). (K-M) NSUN2-knockout PANC-1 cells were generated using the CRISPR/Cas9 system. Wild-type or NSUN2-knockout PANC-1 cells grouped by L-lactate treatment (with / without). (K) Transwell-based migration/invasion: image panels (left) and measurements (right). Scale bar, 200 μm. (n = 3; one-way ANOVA test; mean ± SD). (L) Neurite extension in the Transwell co-culture was quantified under the specified conditions; phase-contrast images were acquired at 6-h intervals. (n = 3; two-way ANOVA test; mean ± SD). (M) (left) Representative fields from DRG co-cultures with tumor cells. (right) Summary statistics for tumor neurite invasion toward DRG. Scale bar, 500 μm. (n = 3; one-way ANOVA test; mean ± SD). ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05.

Figure 3

Lactylation of NSUN2 inhibits its ubiquitination and degradation. (A) Immunoblot analysis of NSUN2 in tumor cells exposed to ddH2O or L-lactate treatment, followed by treatment with cycloheximide (CHX) for specified time points. (B) Immunoblot analysis of NSUN2 expression in PANC-1 cells carrying siNC or LDHA/LDHB siRNAs, followed by treatment with CHX for specified time points. (C) Immunoblot analysis of NSUN2 in tumor cells exposed to MG132 treatment, followed by treatment with CHX for specified time points. (D) Immunoblot analysis of NSUN2 in tumor cells exposed to NH₄Cl or 3-MA treatment, followed by treatment with CHX for specified time points. (E) Following plasmid transfection, PANC-1 cells received MG132/L-lactate treatment for the specified intervals. Anti-FLAG immunoprecipitation of SDS-treated whole-cell lysates was performed, and NSUN2 ubiquitination was evaluated by immunoblotting. The line labeled 'NSUN2 (Ub1)' indicates mono-ubiquitinated NSUN2, while the bracket labeled 'NSUN2 (Ubn)' marks high molecular weight polyubiquitinated NSUN2 species. (F) PANC-1 cells expressing indicated plasmids and LDHA/LDHB siRNAs were treated with MG132. Anti-FLAG immunoprecipitation of SDS-treated whole-cell lysates was performed, and NSUN2 ubiquitination was evaluated by immunoblotting. (G-H) Lactylation of NSUN2 in 293T cells (G) or PANC-1 cells (H) was confirmed using an immunoprecipitation (IP) assay. 293T cells or PANC-1 cells expressing FLAG-tagged NSUN2 and were exposed to ddH2O or L-lactate treatment, with concurrent treatment with MG132. After lysis, SDS-treated extracts were subjected to anti-FLAG IP, followed by immunoblot detection. (I) Lactylation modification levels of NSUN2 in PANC-1 cells following treatment with varying concentrations of L-lactate was confirmed using an immunoprecipitation (IP) assay. (J-M) Confirmation of NSUN2 lactylation in PANC-1 cells expressing specific siRNAs by immunoprecipitation (IP). PANC-1 cells expressing FLAG-tagged NSUN2 were treated with either ddH2O or L-lactate, followed by treatment with MG132. The cells were lysed for SDS pre-treated immunoprecipitation with anti-FLAG antibody followed by western blotting. (J) Lactylation of NSUN2 in tumor cells expressing siNC or LDHA siRNA was confirmed through immunoprecipitation (IP) assay. (K) NSUN2 lactylation in PANC-1 cells expressing siNC or LDHB siRNA was confirmed through immunoprecipitation (IP) assay. (L) NSUN2 lactylation in PANC-1 cells expressing siNC, LDHA siRNA, LDHB siRNA, or both LDHA and LDHB siRNAs by immunoprecipitation (IP). (M) An immunoprecipitation blot showing the interaction levels between NSUN2 and STUB1 in PANC-1 cells transfected with siNC, LDHA siRNA, LDHB siRNA, or both LDHA and LDHB siRNAs and treated with ddH2O or L-lactate, using NSUN2 antibody in vitro. (N) GST pull-down assay showing the interaction between STUB1 and NSUN2. (O) (Left) Immunofluorescence detection of NSUN2 and STUB1 in tumor cells treated as indicated. (Right) Colocalization analysis of NSUN2 and STUB1 in the nuclei of PANC-1 cells under the same treatment conditions. The intensity profiles corresponding to the regions of interest (ROI, indicated by yellow lines) are shown. Red curve: NSUN2 relative fluorescence intensity; green curve: STUB1 relative fluorescence intensity. The overlap of the red and green curves indicates the colocalization of NSUN2 and STUB1. Higher peak intensities correspond to greater levels of fluorescence expression. Scale bar, 20 μm. (P-Q) Manders' overlap coefficient (tM1) (P) and Pearson's correlation coefficient (Q) analysis of NSUN2-STUB1 colocalization in tumor cells treated as indicated. (n = 15; unpaired t test; mean ± SD). ****P < 0.0001.

Figure 4

NSUN2 is lactylated at K692. (A) Representative Coomassie Brilliant Blue staining image of NSUN2 protein from treated tumor cells. The boxed area marks NSUN2-specific bands analyzed by LC-MS. (B) Potential modifications of lysine (K) residues in NSUN2 were assessed by mass spectrometry. Putative lactylation sites (upper panel) and acetylation sites (lower panel) are depicted. (C) (Left) PANC-1 cells were transfected with indicated NSUN2 site mutations, followed by immunoprecipitation and and Western blotting to visualize NSUN2 lactylation with pan-lactyl and FLAG antibodies. (Right) Quantification of NSUN2 lactylation intensity normalized to Flag-NSUN2 (n = 3; mean ± SD; one-way ANOVA test). (D) Mass spectra for NSUN2 peptides lactylated at K692. (E-G) NSUN2-K692 mutant PANC-1 cells were generated using the CRISPR/Cas9 approach, followed by treatment with or without L-lactate and were used for the subsequent experiments. Residue K692 was replaced by arginine (K692R) or glutamine (K692E). (E) Lysates from treated PANC-1 cells were SDS-pretreated, immunoprecipitated with anti-FLAG, and probed for NSUN2 ubiquitination by Western blot. The line labeled 'NSUN2 (Ub1)' indicates mono-ubiquitinated NSUN2, while the bracket labeled 'NSUN2 (Ubn)' marks high molecular weight polyubiquitinated NSUN2 species. (F) (Left) Whole-cell lysates made from PANC-1 cells under the indicated treatments were used for immunoblot to assess NSUN2 protein expression. (Right) Quantification of NSUN2 intensity normalized to β-actin (n = 3; mean ± SD; one-way ANOVA test). (G) (Left) Co-IP immunoblot illustrating the binding of STUB1 to NSUN2 in PANC-1 cells with various NSUN2-K692 mutations. (Right) Quantification of STUB1 relative intensity (n = 3; mean ± SD; one-way ANOVA test). (H) In vitro binding of STUB1 to various domains of NSUN2. The in vitro translated STUB1 protein was incubated with GST, GST-NSUN2 (1-250), GST-NSUN2 (251-500), or GST-NSUN2 (501-767) in GST pull-down assays. The glutathione eluates and 15% input materials were analyzed by SDS-PAGE followed by immunoblotting with anti-STUB1 antibody. (I) Molecular docking analysis of NSUN2-STUB1 interaction before and after NSUN2 K692 lactylation. (J-L) NSUN2-K692 mutant PANC-1 cells were generated using the CRISPR/Cas9 approach, followed by treatment with or without L-lactate and were used for the subsequent experiments. (J) Transwell-based migration/invasion: image panels (left) and measurements (right). Scale bar, 200 μm. (n = 3; mean ± SD; one-way ANOVA test). (K) Quantitative evaluation of neurite outgrowth under the indicated conditions in the Transwell co-culture model. Phase contrast images were obtained at 6 h intervals. (n = 3; mean ± SD; two-way ANOVA test). (L) Schematic representation of the direct co-culture model between tumor cells and DRG. Scale bar, 500 μm. (n = 3; mean ± SD; one-way ANOVA test). ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05.

Figure 5

NSUN2 regulates CDCP1 and STC1 expression in an m5C-dependent manner. (A) Heatmap of mRNA changes in PANC-1 cells exposed to ddH2O or L-lactate treatment. (B-C) Pathway enrichment mapping (KEGG) in PANC-1 cells; dot size = pathway gene number, dot color = adjusted significance. (B) Pathway enrichment mapping (KEGG) of transcriptome sequencing between the control and lactate-treated groups. (C) Pathway enrichment mapping (KEGG) of NSUN2 RIP-seq between the control and lactate-treated groups. (D) NSUN2 binding Motif analysis of peak sequences using HOMER (v4.11.1) and visualization of nucleotide preferences. (E) NSUN2 RIP-seq pie charts for both the ddH2O treatment (left) and L-lactate groups (right) depicting the distribution of NSUN2 RIP-binding peaks. (F) Bioinformatic prediction pinpointed CDCP1 and STC1 as NSUN2-regulated genes. (G) MeRIP-qPCR confirmed m5C enrichment of CDCP1 and STC1 transcripts in PANC-1 cells ± L-lactate. (H-I) PANC-1 cells expressing shNC or NSUN2-specific shRNA, exposed to ddH2O or L-lactate treatment, were used for Actinomycin D assay to assess the half-life of CDCP1 (H) and STC1 (I) mRNA. (J-K) PANC-1 cells expressing shNC or NSUN2-specific shRNA, followed by specific treatment. (J) (Above) Immunoblot analysis of CDCP1 and STC1 in PANC-1 cell lysates prepared under the respective treatment conditions. (Below) Quantification of CDCP1 and STC1 intensity normalized to β-actin (n = 3; mean ± SD; one-way ANOVA test). (K) Total RNA isolated from treated PANC-1 cells was used for qRT-PCR analysis to assess the CDCP1 and STC1 mRNA expression. (L-M) PANC-1 cells expressing siNC, CDCP1 siRNA (L) or STC1 siRNA (M), followed by treatment with or without L-lactate, were used for subsequent experiments. Transwell-based migration/invasion: image panels (top) and measurements (bottom). Scale bar, 200 μm. (n = 3; one-way ANOVA test; mean ± SD). (N) The interaction of NSUN2 and the indicated mRNAs in PANC-1 cells exposed to ddH2O or L-lactate treatment was demonstrated by CLIP. (O) IGV tracks for CDCP1 and STC1 from RIP-seq analysis, peaks are marked with red triangles. (P) WT and potential m⁵C modification motif mutation CDCP1/STC1 sequences were presented. Red fonts represented the mutant bases. (Q-T) WT-CDCP1, Mut-CDCP1, WT-STC1 or Mut-STC1 was overexpressed in CDCP1/STC1 knockdown PANC-1 cells. (Q-R) MeRIP-qPCR analysis of CDCP1 (Q) and STC1 (R) mRNAs in PANC-1 cells under specific treatment conditions. (S-T) PANC-1 cells under specific treatment were used for Actinomycin D assay to assess the half-life of CDCP1 (S) and STC1 (T) mRNA. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05.

Figure 6

Impact of NSUN2 Lactylation on PNI in PDAC in vivo. (A) Schematic diagram of the sciatic nerve invasion model in nude mice. NSUN2-knockout PANC-1 cells were generated using the CRISPR/Cas9 system. Wild-type or NSUN2-knockout PANC-1 cells were treated with either L-lactate or PBS, followed by inoculation into the sciatic nerve of nude mice. PBS alone served as the control. (n=5 per group). (B-C) Sciatic nerve function scores (B) and sciatic nerve indexes (C) of mice treated as indicated. (n = 5; mean ± SD; Kruskal-Wallis test with Dunn's multiple comparisons test). (D) Representative gross morphology, intraoperative views, H&E and mIF images from the sciatic nerve invasion model. mIF demonstrates the expression of Pan-lac, NSUN2, CDCP1, and STC1 at the tumor-nerve interface. Dashed lines mark the tumor margins. Tuj1 serves as a neuronal marker, whereas CK-19 identifies tumor epithelial cells. Scale bar for H&E, 1000 μm; scale bar for mIF, 200 μm. (E-H) KPC mice (LSL-KRASG12D/+; LSL-TP53R172H/+; PDX-1-CRE+/+) were divided into six groups, each receiving orthotopic injections into the pancreas. Two groups were injected with adeno-associated virus (AAV) packaging short hairpin RNA negative control (shNC), while the other two groups were injected with AAV packaging shNSUN2. One week later, control and experimental mice received sodium lactate (1 g/kg) or PBS intraperitoneally, respectively. n = 15. (E) H&E and mIF images demonstrate the PNI features in KPC mice and expression of Pan-lac, NSUN2, CDCP1, and STC1. Tuj1 is used as a neuronal marker, while CK-19 denotes tumor epithelium. Scale bars: H&E and mIF, 200 μm. (F) Quantitative assessment of nerve density (above) and nerve count (below) in KPC mice. (n = 15; Chi-square test). (G) PNI frequency in KPC mice. (n = 15; Fisher's exact test). (H) Severity score of KPC mice. (n = 15; mean ± SD; one-way ANOVA test). ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05.

Figure 7

Association of NSUN2/CDCP1/STC1 with PNI and Unfavorable Prognosis in PDAC. (A) H&E and IHC images showing Pan-lac, NSUN2, CDCP1, STC1 expression in PDAC specimens (n = 142). Scale bars, 200 μm. (B) H-score of NSUN2 expression in PDAC patients with or without severe PNI. (C) ROC curve analysis based on NSUN2 H-scores to further evaluate its predictive ability for PNI risk. (D) Proportional distribution of Pan-lac, NSUN2, CDCP1, and STC1 expression in PDAC samples grouped by PNI severity (with or without severe PNI). (E) Lactylation of NSUN2 in PNI(-) or PNI(+) pancreatic cancer tissues (left) and quantification of NSUN2 lactylation intensity normalized to NSUN2 (n = 3; unpaired t test; mean ± SD). (F-I) Kaplan-Meier survival curves showing OS and DFS in the PDAC cohort stratified by CDCP1 (F-G) and STC1 (H-I) level (high vs low). (J) Diagram summarizing the proposed mechanism through which a lactate-enriched tumor microenvironment facilitates pancreatic cancer PNI by upregulating tumor cells NSUN2 lactylation and transcriptionally activating the m5C modification of neural invasion—related genes.