UBE2M as a bridge spanning neddylation and cell cycle regulation in colorectal adenocarcinoma

TL;DR

This study shows that UBE2M links the neddylation process to cell cycle control in colorectal cancer, and targeting it with micafungin may help treat the disease.

Contribution

UBE2M is identified as a novel bridge between neddylation and cell cycle regulation in colorectal cancer.

Findings

UBE2M promotes CRC progression by neddylating USP39 and modulating PABPC1.

Micafungin inhibits UBE2M and suppresses CRC progression in mice.

Neddylation is associated with G2M phase progression in colorectal cancer.

Abstract

Protein neddylation is a post-translational modification process that modifies the functional state of proteins by conjugating NEDD8, a ubiquitin-like polypeptide, to the lysine residues of substrates. In various cancers, neddylation is upregulated and implicated in cancer progression via modulating cell cycle-related proteins. However, in colorectal cancer (CRC), the relationship between neddylation and the cell cycle remains incompletely understood. Here, by leveraging single-cell and bulk transcriptome data, we demonstrated that neddylation is associated with G2M phase progression in CRC. Through bioinformatic analysis, we identified ubiquitin conjugating enzyme E2 M (UBE2M) as a molecular bridge spanning neddylation and the cell cycle in CRC. To elucidate how UBE2M promotes CRC progression, we conducted in vivo and in vitro experiments to confirm the role of UBE2M in neddylating USP39, which in turn modulates the deubiquitination of PABPC1, enhances the translation efficiency of CCNB1 and propels the cell cycle progression of CRC. Regarding clinical application, we identified micafungin as an inhibitor of UBE2M capable of suppressing the regulatory axis and, consequently, hindering CRC progression. Therefore, this study underscores the potential role of UBE2M in bridging neddylation with the cell cycle and holds promise for advancing targeted therapies in CRC treatment.

Colorectal cancer is a major health concern worldwide, being the third most common and second deadliest cancer. This study explores a process called neddylation, which modifies proteins to control their function and stability, in colorectal cancer. The researchers focused on a protein called UBE2M, which is involved in neddylation and may play a role in cancer progression. The team used advanced techniques to analyze cancer cells from patients, identifying that UBE2M is overexpressed in colorectal cancer tissues. They found that UBE2M helps cancer cells grow by stabilizing another protein, PABPC1, which is crucial for cell cycle progression. By inhibiting UBE2M with a drug called micafungin, they were able to slow down tumor growth in mice. The study concludes that targeting UBE2M could be a promising strategy for treating colorectal cancer.

This summary was initially drafted using artificial intelligence, then revised and fact-checked by the author.

Figures

a). A total of 49,155 epithelial cells were extracted, and CNV levels between normal and CRC tissues were compared (Supplementary Fig. 1b). Using the infercnvpy package, epithelial cells were partitioned into 40 CNV-based subclusters, with their distributions and CNVs shown in Supplementary Fig. 1c–f. Using a threshold of 0.01 to classify cellular malignancy, we excluded normal epithelial cells (Leiden clusters 0, 1 and 25), ultimately identifying 37,537 malignant epithelial cells (Fig. [1d](#Fig1) and Supplementary Fig. 1g). Then we found that malignant epithelial cells exhibit higher neddyla

h) and WGCNA for bulk RNA-seq (Supplementary Fig. 2i) to select gene modules associated with both neddylation and cell cycle subpopulations. First, through the analysis of scRNA-seq data, we identified 36 distinct gene modules (Supplementary Fig. 2g) and subsequently investigated their associations with various cNMF clusters and neddylation scores. Our screening revealed that only seven modules (M6, M7, M9, M14, M21, M29 and M31) demonstrated significant co-association with both cNMF cluster 4 (average expression > 1.5 and percent expressed ≥ 50%) and high neddylation score (average expression

h). To validate the reliability of this axis, we additionally conducted a series of rescue experiments to confirm its role in promoting the progression of CRC. In a subcutaneous tumor-bearing mouse model, we found that tumor suppression resulting from UBE2M inhibition could be reversed by overexpressing PABPC1 (Supplementary Fig. 8e–g), and IHC analysis together with hematoxylin and eosin (H&E) staining confirmed the validity of this regulatory axis (Supplementary Fig. 8h). Furthermore, in cellular systems, we discovered that the inhibition of proliferation of CRC cells owing to UBE2M knockdow

d–f). Furthermore, we discovered that in UBE2M knockdown SW480 and DLD-1 cells, PABPC1 was rescued by the proteasome inhibitor MG132 (Fig. [5p,q](#Fig5)). This suggests that the UBE2M regulation of PABPC1 might be mediated by the ubiquitin–proteasome system. By using a series of ubiquitin mutants in which only one Lys residue was preserved while the others were substituted with Arg residues, we then corroborated this hypothesis in HEK293T cells, showing that UBE2M can prevent the degradation of PABPC1 by inhibiting K48-linked polyubiquitination (Fig. [5r](#Fig5)). Consistently, IP assays targe

b). CHX chase assays further demonstrated that USP39 knockdown accelerated PABPC1 degradation, while its overexpression prolonged the protein half-life of PABPC1 (Fig. [6l,m](#Fig6) and Supplementary Fig. 5c,d), collectively suggesting that USP39-mediated PABPC1 stabilization occurs through the ubiquitin–proteasome system. To further validate that PABPC1 serves as a substrate for USP39, we performed both in vivo and in vitro deubiquitination assays. In cellular systems, USP39 knockdown enhanced the ubiquitination levels of PABPC1 (Supplementary Fig. 5e). Furthermore, exogenous expression of US

h). Moreover, previous studies have demonstrated that PABPC1, by binding to eIF4G, can enhance the translation efficiency of CCNB1^[42](#CR42)^. Given that CCNB1 is a key regulatory protein of the G2M phase, and UBE2M is known to regulate the G2M phase, we hypothesized the existence of a UBE2M–USP39–PABPC1–CCNB1 axis that modulates the G2M phase in CRC. Our RIP assays confirmed that PABPC1 protein indeed binds to CCNB1 mRNA (Supplementary Fig. 6i). We further confirmed that UBE2M is incapable of modulating the mRNA level of CCNB1 (Supplementary Fig. 6c), which indicated that the regulation occ

a). We then confirmed that micafungin diminished the levels of CCNB1 mRNA bound by an equivalent amount of PABPC1 (Fig. [8a](#Fig8)). Subsequently, we validated on the WB level that micafungin could attenuate the expression of PABPC1 and CCNB1, whereas the expression level of UBE2M was not affected, which was consistent with previous findings^[47](#CR47)^ (Fig. [8b](#Fig8)). Further mechanistic analysis demonstrated that micafungin pharmacologically suppressed USP39 neddylation and augments ubiquitin-mediated degradation of PABPC1 (Supplementary Fig. 8b,c). Using micafungin at 43 μM, we observ

Full text

Introduction

Colorectal cancer (CRC) ranks as the third most common and second most lethal malignancy worldwide^1^. Despite advancements in multimodal therapies, including surgery, chemotherapy and immunotherapy, treatment resistance and recurrence remain major challenges, underscoring the urgent need to identify novel therapeutic targets and elucidate underlying mechanisms^2^.

Dysregulation of the cell cycle is a hallmark of cancer, enabling uncontrolled proliferation^3^. While the roles of cyclins, CDKs and their inhibitors are well documented^4,5^, the post-translational regulatory landscape, particularly neddylation, in CRC cell cycle control is less understood. Neddylation, analogous to ubiquitination, is a pivotal post-translational modification that conjugates Nedd8 to target proteins, thereby modulating their function, stability and localization^6–8^. This reaction is catalyzed by the coordinated actions of E1, E2 and E3 enzymes^9^.

In recent years, neddylation has garnered substantial attention in oncology. A growing body of research indicates that neddylation plays a important role in the initiation, progression and invasion of various cancers^10–13^. Concurrently, neddylation is intricately linked to the cell cycle, including regulation of the G1/S transition and induction of G2 phase arrest^14–16^. In CRC, neddylation influences immune response^17^, therapy sensitivity^18,19^, apoptosis^20^ and some cancer-associated signaling pathways^21,22^, yet its direct link to cell cycle regulation—especially the G2/M phase—remains elusive. Thus, our aim was to define the role of neddylation in regulating the cell cycle in CRC and to identify subsequent mechanistic drivers and therapeutic vulnerabilities. Furthermore, resistance to the NEDD8-activating enzyme inhibitor MLN4924 highlights the necessity to identify alternative targets within the neddylation pathway^23^.

Ubiquitin conjugating enzyme E2 M (UBE2M), a core neddylation E2 enzyme, works in concert with E1 and E3 enzymes to facilitate the transfer of Nedd8 to target proteins, thereby being involved in the neddylation process of a multitude of proteins^9^. While existing research hints at aberrant UBE2M expression in certain cancers and its correlation with tumor progression^12,24,25^, the study of its role in CRC is still in its nascent stages. Its overexpression in CRC tissues suggests a potential oncogenic role^26^, but its specific functions, targets and mechanisms in CRC cell cycle progression are unknown. On the basis of these gaps in knowledge, we hypothesized that UBE2M acts as a molecular bridge connecting neddylation to G2/M phase progression in CRC by stabilizing a key cell cycle regulator through a novel post-translational mechanism and that targeting this axis could yield therapeutic benefits.

Here, we have elucidated the link between neddylation and the cell cycle in CRC at both the cellular and whole-organism levels. Moreover, UBE2M, serving as a molecular bridge between neddylation and cell cycle regulation, has been demonstrated through both in vivo and in vitro experiments to be involved in CRC progression and to predict poor prognosis. Mechanistically, we discovered that UBE2M can promote the neddylation of USP39, leading to the deubiquitination of PABPC1 and, ultimately, enhancing the transcriptional efficiency of CCNB1 to advance cell cycle progression. In addition, we have identified micafungin, an inhibitor of UBE2M, as a potential therapeutic agent for clinical translation of CRC treatment.

Materials and methods

Data acquisition and processing

The single-cell RNA sequencing (scRNA-seq) data were sourced from five published CRC cohorts (CRC-SG1, CRC-SG2, SMC, KUL3 and KUL5)^27,28^, which contained 189 samples from 63 patients. We directly obtained the combined count expression matrices that had been quality-controlled according to the original publication through Synapse (syn26844071). Then, we used the Scanpy package to process the data. Following data normalization and selection of the top 2000 highly variable genes, we scaled the data and performed principal component analysis for dimensionality reduction, followed by batch effect correction using Harmonypy (v0.0.9). Cell clusters were identified through Leiden algorithm-based clustering (resolution range 0.2–1.2), and the global cellular distribution was visualized using the Uniform Manifold Approximation and Projection (UMAP) space. After processing 373,058 captured cells, we directly adopted the cell annotation strategy outlined in the original publication^27^ for annotation purposes. Subsequently, 49,155 epithelial cells were isolated for analysis. To enhance the multidimensionality of our validation process, we integrated a total of 1,829 CRC samples for comprehensive bulk-sequencing analysis (TCGA-COAD/READ, GSE39582, GSE38832, GSE17538 and GSE14333). The raw count data obtained from TCGA database were converted to transcripts per million (TPM) for normalization. The raw GEO datasets were processed using the ‘affy’ R package, with background correction and normalization implemented through the robust multiarray average algorithm. We annotated probes with gene symbols using the ‘hgu133plus2.db’ R package, as the GEO datasets above were all generated on the Affymetrix GPL570 platform. We excluded normal colorectal mucosal samples and retained only CRC tissues. The ‘Combat’ method was used to eliminate batch effects. Corresponding clinical and molecular characteristics of each patient are presented in Supplementary Tables 1 and 2. In addition, spatial transcriptomic (ST) data were procured from GSE226997 to enrich our data.

InferCNV analysis

In our study, we harnessed inferCNVpy (v0.4.2) to ascertain the copy number variation (CNV) score for each cell, with normal tissue cells serving as our comparative benchmark (Supplementary Fig. 1b). We analyzed the distribution of inferCNV clusters within both tumor and normal tissues and correlated these with their respective CNV averages (Supplementary Fig. 1c–f). Using a stringent cutoff value of 0.01, we could precisely discriminate between malignant and nonmalignant cells (Supplementary Fig. 1g). The CNV scores for each cell are presented in Supplementary Table 3.

cNMF and AUCell analysis

We leveraged the cNMF package (v1.5.4) to delineate gene expression signatures and classify cellular subtypes within malignant cell populations^29^. Through a systematic evaluation across k = 5–10 based on the stability and error metrics of cNMF, we identified k = 7 as the optimal number of components, for the point preceding the most pronounced drop in the stability curve (Supplementary Fig. 1i). Furthermore, a density threshold of 0.15 was established to demarcate distinct clusters (Supplementary Fig. 1j). The defined genes for the seven cNMF-derived clusters are presented in Supplementary Table 4. To characterize the neddylation phenotype, we used the AUCell scoring system (v1.20.1) to attribute neddylation-associated traits to individual cell clusters, thereby uncovering disparities within the cellular landscape.

Enrichment analysis

We performed various enrichment analyses, including Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), single-sample Gene Set Enrichment Analysis (ssGSEA) and Gene Set Variation Analysis (GSVA). GO and KEGG analyses were applied for group-wise comparisons, whereas ssGSEA and GSVA were used for individual-level assessments. All gene sets used in these analyses were derived from the Molecular Signatures Database (MSigDB)^30^.

Pseudotime analysis

To describe the developmental trajectories of the seven cNMF-identified malignant epithelial clusters, we first applied the CytoTRACE2 package (v1.0.0) to identify the origin cluster^31^. Subsequently, we used the PAGA method (scanpy v1.10.2) to demonstrate the interconnectivity and trajectories among these clusters. Detailed data are presented in Supplementary Table 5.

ST analysis

To bolster the robustness of our findings, ST data were acquired from GSE226997 containing four CRC samples. Raw sequencing data in FASTQ format were processed using the Space Ranger pipeline (version 2.0.0) to generate spatially resolved gene expression matrices, following alignment to the GRCh38 reference genome with default parameters. We used Seurat (v4.4.0) to conduct SCTransform normalization, integration and dimensionality reduction of the ST samples. Next, we applied the COSG package (v0.9.0) to identify the top 500 marker genes of the target cell clusters^32^. By quantifying and mapping these markers with the AddModuleScore function into the ST data, we could directly compare gene expression patterns against regions associated with neddylation phenotypes, providing a clearer understanding of the underlying molecular relationships.

hdWGCNA and WGCNA

To uncover gene modules correlated with phenotypic traits, we applied hdWGCNA (v0.3.03) to scRNA-seq data and WGCNA (v1.71) to bulk RNA-seq data^33,34^. For both analytical approaches, we chose a soft-thresholding power of 12 (Supplementary Fig. 2j). We selected gene modules related to neddylation and cell cycle for further target gene screening. Furthermore, within the bulk RNA-seq data, we identified genes with a correlation coefficient exceeding 0.7 with the cyan module and greater than 0.25 with cell cycle-associated modules, thereby refining our gene selection for downstream analyses (Supplementary Fig. 2k). Details are presented in Supplementary Tables 9–11.

Clinical specimens and cell culture

Surgically obtained CRC and adjacent normal mucosal tissues were collected from patients at the First Affiliated Hospital of Nanjing Medical University between 2017 and 2020. A total of 80 paired specimens, sourced from patients with CRC, were prepared into formalin-fixed paraffin-embedded tissue microarrays (TMA) conducted by Servicebio. This study was approved by the human ethics committee. The cell lines NCM460, LoVo, RKO, DLD-1, HT-29, HCT116, SW480, SW620 and HEK293T, sourced from the Cell Bank of Type Culture Collection at the Chinese Academy of Sciences, were propagated in their specific mediums enriched with 10% fetal bovine serum from Nanjing Ozfan Biotechnology. These cultures were maintained at an optimal temperature of 37 °C within a 5% CO_2_ incubator environment. All cell lines were authenticated by short tandem repeat (STR) profiling and confirmed to be free of mycoplasma contamination.

Cell transfection

We used lentiviral-mediated RNA interference and overexpression strategies for the genes UBE2M, USP39 and PABPC1, with short hairpin RNAs (shRNAs) and full-length target sequences obtained from Tsingke. Furthermore, we acquired plasmids for UBE2M tagged with Flag, PABPC1 tagged with Myc, USP39 tagged with His and Ub tagged with HA from GenePharma. All wild type and mutant plasmids for USP39, PABPC1 and Ub were also purchased from GenePharma. Lipofectamine 3000 (Invitrogen) was used for transient transfection of these plasmids into cells. Transfection efficiency was validated by quantitative real-time reverse transcription polymerase chain reaction (qRT–PCR) and western blot (WB) analysis. The shRNA sequences are presented in Supplementary Table 6.

RNA extraction and qRT–PCR

Total RNA was extracted from the samples using TRIzol reagent (Invitrogen). The extracted RNA was subsequently reverse transcribed into complementary DNA with the HiScript RT Mix (Vazyme). qRT–PCR was performed using a SYBR Premix Ex Taq Kit (TaKaRa Biotechnology) to amplify and detect gene expression levels. Primers used are presented in Supplementary Table 7.

WB

Proteins were extracted from cells and tissues using RIPA lysis buffer (Beyotime), and protein concentrations were determined through a bicinchoninic acid assay. We subjected the proteins to SDS–PAGE with Bio-Rad and then transferred them to Millipore polyvinylidene fluoride membranes. After a 30-min blocking period with QuickBlock (Beyotime), the membranes were incubated with primary antibodies at 4 °C overnight. The next day, the membranes were incubated with secondary antibodies for 2 h at room temperature, after which the blots were detected. β-actin served as an internal control for normalization. The results were quantified using ImageJ software (v1.53e). The ubiquitination assay was performed as previously described^35^. MG132, a proteasome inhibitor, and cycloheximide (CHX), an inhibitor of protein synthesis, were acquired from Selleckchem and Cell Signaling Technology, respectively. These inhibitors were incorporated into the culture medium at concentrations of 10 μM for MG132 and 100 μg/ml for CHX. All antibodies are presented in Supplementary Table 8. The original WBs are provided in the Supplementary Information.

Cell proliferation

Cells were collected with trypsin when they reached logarithmic growth phase. A subset of cells was seeded into a 96-well plate at 2000 cells per well and incubated for 1–4 days, and the optical density (OD) value (450 nm) was measured using the CCK-8 Cell Counting Kit (Beyotime). For the colony formation assay, cells were plated at 500 per well in 6-well plates and cultured for 10–14 days with medium changes every 3 days. The colonies were fixed with 4% paraformaldehyde, stained with crystal violet and photographed. The remaining cells were seeded at 10,000 per well in a 96-well plate and then treated with EdU working solution for 2 h before fixation and permeabilization. After rinsing with PBS, Azide 555 and Hoechst 33342 were applied for imaging. The ImageJ software was used to analyze the colony formation assay and EdU results.

Cell cycle and apoptosis

Flow cytometry was used to assess cell cycle progression and apoptosis in CRC cells. For cell cycle analysis, different groups of cells were fixed, stained with propidium iodide and analyzed using flow cytometry. To assess apoptosis, cells were stained with Annexin V-FITC/propidium iodide and examined using flow cytometry. Specific procedures were performed as previously described^36^.

Immunoprecipitation, RNA immunoprecipitation assay and MS

Cultured cells were lysed in NP-40 lysis buffer (Beyotime) with 30-min ice incubation. After centrifugation at 12,000g (4 °C, 15 min), the clarified supernatants were immunoprecipitated by rotary incubation with specific primary antibodies at 4 °C overnight. Subsequently, Protein A/G Plus-Agarose beads (ThermoFisher Scientific) were introduced to the antigen–antibody complexes under continuous rotation for 12 h at 4 °C. The bead-bound complexes were then pelleted and stringently washed three times with chilled wash buffer (1 ml, 20 min per cycle). After buffer removal, the captured proteins were eluted in SDS–PAGE sample buffer (Beyotime) through 5-min boiling at 100 °C. Then, the supernatant was collected for WB and mass spectrometry (MS) analyses. The MS analysis was performed on an Orbitrap Exploris 480 and identified with Mascot, with a 1% false discovery rate (FDR), by BGI Genomics. We performed RNA immunoprecipitation (RIP) according to the Magna RIP Kit protocol (Millipore). We mixed 5 μg Myc and IgG antibodies with 50 μl magnetic beads and incubated them with the cell lysates. Purified RNA extracted from the immunoprecipitated complex was subjected to reverse transcription and PCR amplification, and the final PCR products were subjected to agarose gel electrophoresis.

In vitro ubiquitination assay

The reconstituted reaction contained recombinant E1, E2, Flag–E4B (E3), Myc–PABPC1 (substrate) and ubiquitin in ATP-supplemented buffer (Enzo Life Sciences) with increasing amounts of purified Flag–UBE2M. After incubation at 37 °C for 1 h, reactions were analyzed using SDS–PAGE and immunoblotting with an anti-Myc antibody to detect polyubiquitinated species.

In vivo and in vitro deubiquitination assays

To perform the in vivo deubiquitination assay, HEK293T cells were transfected with plasmids encoding HA–Ub, Myc–PABPC1, His–USP39 or mutant His–USP39–ΔC306A for 48 h. The cells were then treated with 10 µM MG132 for 6 h, washed with prechilled PBS and lysed using RIPA buffer. HA-ubiquitinated PABPC1 was captured using an anti-Myc antibody, and its ubiquitination level was detected by WB analysis using an anti-HA antibody. In CRC cells, the HA-Ub plasmid was cotransfected into USP39 knockdown SW480 cells. HA-ubiquitinated PABPC1 was isolated with an anti-PABPC1 antibody, and its ubiquitination level was assessed using WB with an anti-HA antibody.

For the in vitro deubiquitination assay, we purified ubiquitinated PABPC1 from HEK293T cells transfected with Myc–PABPC1 and HA-Ub using affinity purification with anti-Myc agarose beads (ThermoFisher Scientific), followed by elution with Myc peptide. His-tagged USP39-WT, catalytically inactive ΔC306A or USP39-K6R were expressed in HEK293T cells, lysed, purified via Ni-NTA beads (ThermoFisher Scientific), washed with imidazole gradients and eluted with 250 mM imidazole. Neddylated USP39 was purified from HEK293T cells cotransfected with His-USP39 and NEDD8 plasmids. The ubiquitinated PABPC1 was then incubated with purified USP39-WT or USP39-ΔC306A in the reaction buffer (50 mM Tris-HCl, 50 mM NaCl, 10 mM dithiothreitol, 1 mM EDTA, 5% glycerol, pH 8.0) for the indicated times at 37 °C. The reactions were terminated by boiling in 1 × SDS sample buffer and analyzed by immunoblotting with anti-HA antibodies. Similarly, ubiquitinated PABPC1 was incubated with purified USP39-WT, USP39-K6R or USP39-NEDD8 for 1 h, followed by immunoblotting with an anti-HA antibody.

Cell synchronization

DLD-1 cells were cultured as previously described. To analyze cell cycle-dependent protein expression, cells were synchronized using a double thymidine block method. In brief, when cells reached approximately 50% confluency, they were treated with 2 mM thymidine (Sigma-Aldrich) for 16 h. The thymidine-containing medium was then removed, and cells were washed with PBS before being incubated in fresh complete medium for 8 h. Subsequently, a second thymidine block (2 mM) was performed for 16 h to achieve a highly synchronized population at the G1/S boundary. Finally, cells were released into fresh complete medium, and this time point was designated as 0 h. Whole-cell lysates were collected at sequential time points (0, 2, 4, 6, 8, 10 and 12 h) after release for subsequent analysis.

Immunofluorescence staining

The cells were fixed with paraformaldehyde, permeabilized with Triton X-100, blocked with bovine serum albumin and then incubated with primary antibody overnight at 4 °C, followed by secondary antibody at room temperature. 4,6-Diamidino-2-phenylindole was used to stain the nuclei. The specific procedure has been described before^35^. Subsequently, observations were recorded using a Leica SP5 confocal microscopy system (Leica Microsystems).

Animal model

We purchased 4-week-old BALB/c nude mice for subcutaneous tumor implantation. One million cells each in 100 μl were injected into each mouse’s flank. Concurrently, commencing on day 10, we administered daily intravenous injections of micafungin (MedChemExpress) dissolved in dimethylsulfoxide (DMSO) to an additional group of mice at a dose of 30 mg/kg. The control group received injections of DMSO alone. The tumor volume was monitored and measured weekly. A total of 28 days after injection, the tumors were excised, weighed and preserved for subsequent analysis. The study was approved by the Animal Care and Use Committee of Nanjing Medical University (approval no. IACUC-2411010).

Statistical analysis

All experimental data were analyzed using GraphPad Prism 9.2.0, presented as the mean ± standard deviation (s.d.). All experiments were performed at least in triplicate. All bioinformatics analyses were performed utilizing R (v4.2.0) and Python (v3.10). All other packages or functions that were not explicitly specified were used with their default parameters. The statistical significance of differences between groups was determined using a two-tailed Student’s t-test for pairwise comparisons or one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test for multiple group comparisons. The FDR was used to control for multiple comparisons in the statistical analysis. Correlations between variables were assessed using Spearman’s correlation coefficient. Kaplan–Meier analysis was performed to evaluate the overall survival rates. The chi-squared test was also conducted to explore the relationship between the expression of UBE2M and the clinicopathological features of patients with CRC. Statistical significance was set at P < 0.05.

Results

Neddylation is correlated to the cell cycle process in CRC at the scRNA-seq level

Initially, we outlined the workflow of this study in a schematic (Fig. 1a). On the basis of the scRNA-seq data integrated from 189 samples, a comprehensive single-cell landscape encompassing 373,058 cells was constructed (Fig. 1b). We downloaded the set of neddylation-related genes from the MSigDB database (REACTOME_NEDDYLATION). We then used AUCell analysis to quantify the neddylation pathway and mapped them into cells, observing that neddylation was significantly elevated in endothelial cells, epithelial cells and fibroblasts, with epithelial cells demonstrating the highest neddylation activity (P < 0.001). To refine the research focus, we prioritized investigating the interplay between neddylation and epithelial cells (Fig. 1c and Supplementary Fig. 1a). A total of 49,155 epithelial cells were extracted, and CNV levels between normal and CRC tissues were compared (Supplementary Fig. 1b). Using the infercnvpy package, epithelial cells were partitioned into 40 CNV-based subclusters, with their distributions and CNVs shown in Supplementary Fig. 1c–f. Using a threshold of 0.01 to classify cellular malignancy, we excluded normal epithelial cells (Leiden clusters 0, 1 and 25), ultimately identifying 37,537 malignant epithelial cells (Fig. 1d and Supplementary Fig. 1g). Then we found that malignant epithelial cells exhibit higher neddylation levels than normal cells (P < 2.22 × 10^−16^) (Fig. 1e). Consequently, our study concentrates on the neddylation status within the malignant epithelial cells. However, acknowledging the potential heterogeneity among malignant epithelial cells^37^, we used the cNMF algorithm to stratify these cells into seven subclusters (Fig. 1f,g), which were annotated with marker genes and enriched pathways (Supplementary Table 4), including Oncogen-drive, Mixed, p-EMT, Cyclin, Stress, Metabolism and Secretory (Fig. 1h,i). We then applied neddylation scores to these malignant epithelial subclusters and found the highest expression in cNMF cluster 4, the Cyclin subcluster, which was particularly enriched in cell cycle, G2M checkpoint and meiotic cell cycle phase transition (Fig. 1j,k). The high G2/M-phase fraction in cNMF cluster 4 further confirmed its identity as cell cycle-enriched subpopulation (Supplementary Fig. 1h). Meanwhile, to determine the origin of these cells, we used CytoTrace2 to infer their developmental trajectories (Supplementary Fig. 1k,l), revealing that cNMF cluster 4 occupied an earlier state in the trajectory (Fig. 1l). The observed co-occurrence of early differentiation signatures and enriched cell cycle pathways in cNMF cluster 4 aligns with the biological paradigm where primitive cells maintain their undifferentiated states through heightened proliferative activity—a mechanistic strategy to sustain tissue regeneration potential while delaying terminal differentiation^38,39^. In addition, we used the PAGA package to illustrate the correlations between the subclusters and the distribution of neddylation, which showed that cNMF cluster 4 was related to the other clusters and was enriched in the neddylation pathway (Fig. 1m,n). These results confirm the high enrichment of neddylation in malignant CRC epithelial cells and its close association with the cell cycle pathway.Fig. 1. Neddylation is correlated with the cell cycle in scRNA-seq data of CRC.a The flowchart of the entire study. b UMAP of all 373,058 cells in 189 samples, annotated with cell types sourced from the original literature. c Distribution of the neddylation AUCell score among all the cells shown in the UMAP. d Post-selection by the infercnvpy method; the inferred CNVs of the malignant and normal cells are presented as chromosome heat maps. The mutations, indicated in red for gain-of-function and blue for loss-of-function, were inferred on the basis of the averaged gene expression levels. e The neddylation score between malignant and normal epithelial cells using the AUCell method. The power analysis yielded an effect size of 0.682 with a 95% confidence interval (CI) of 0.648 to 0.716. f The clustering heat map illustrating the seven cNMF clusters and their respective annotations in identified malignant epithelial cells. g The distribution of the seven cNMF clusters. h,i The bubble plot and heat map showcase the top three marker genes for each of the seven cNMF clusters, accompanied by their respective GSVA annotations. The intensity of the red coloration signifies higher expression levels. The top section of the GSVA heat map indicates the proportion of cells in each cNMF cluster relative to the total cell population. j,k The distribution and comparison of the neddylation score in the seven cNMF clusters. Effect sizes for cNMF cluster 4 versus other clusters were: versus 1: 0.672 (95% CI, 0.635 to 0.710); versus 2: 0.844 (0.804 to 0.884); versus 3: 0.900 (0.855 to 0.945); versus 5: 1.403 (1.353 to 1.453); versus 6: 1.224 (1.172 to 1.277); versus 7: 0.606 (0.427 to 0.782). l The CytoTRACE2 scores of the seven cNMF clusters. Effect sizes for cNMF cluster 4 versus other clusters were: versus 1: 0.693 (95% CI, 0.655 to 0.730); versus 2: −0.005 (−0.044 to 0.034); versus 3: 2.554 (2.496 to 2.612); versus 5: 2.090 (2.034 to 2.145); versus 6: 2.013 (1.954 to 2.072); versus 7: 1.680 (1.500 to 1.861). m,n This illustration shows the pseudotime trajectory of the seven cNMF clusters, capturing their developmental progression from inception to the final stage, annotated with corresponding neddylation scores. o Malignant epithelial cells were stratified into high and low neddylation groups on the basis of neddylation scores. KEGG pathway enrichment analysis was performed for each group, with statistical significance determined using the hypergeometric test and FDR adjustment via the Benjamini–Hochberg method. Error bars represent the mean ± s.d.

Neddylation may affect the G2/M phase, with UBE2M being a potential key regulator

To rule out the possibility of cNMF classification bias, we also divided all malignant epithelial cells into high and low neddylation groups and performed enrichment analyses with differentially expressed genes, uncovering high neddylation group cells involved in the cell cycle, proteasome, RNA degradation and CRC pathways (Fig. 1o and Supplementary Fig. 2a–c). Subsequently, we observed a markedly higher proportion of cells with high neddylation levels during the G2M phase (P = 2.90 × 10^−196^) (Fig. 2a). To further examine the relationship between neddylation and the G2M phase as well as cNMF cluster 4, we used the CellCycleScoring function and cNMF package to quantify the score of the G2M phase and cNMF cluster 4. We then performed a correlation analysis between these two scores and the neddylation score. A significant correlation between the neddylation score and G2M phase score/Cyclin cluster was observed (G2M phase: r = 0.18, P = 1 × 10^−16^; Cyclin: r = 0.37, P = 1 × 10^−16^) (Fig. 2b and Supplementary Fig. 2d). This suggests that the impact of neddylation on the cell cycle of CRC may be focused on the G2M phase. To enhance the reliability of our findings, ST data from four CRC cases were additionally used to complement our existing results. Dimensionality reduction and clustering were performed on the ST data, identifying 12 subpopulations (Supplementary Fig. 2e). Neddylation displayed distinct expression patterns across different cellular subpopulations, with considerable variation in neddylation levels detected among the four patients with CRC (Supplementary Fig. 2f). We integrated tumor epithelium markers, cNMF cluster 4 signatures and neddylation-related genes into the ST data using the AddModuleScore algorithm. Spatial mapping demonstrated predominant neddylation enrichment within malignant epithelial compartments, with partial colocalization observed between cNMF cluster 4 and neddylation hotspots (Fig. 2c). This pattern was not universally observed—patient P1 displayed near-complete absence of the cNMF cluster 4 subpopulation, highlighting interpatient heterogeneity in the distribution of this neddylation-associated epithelial state. Collectively, our integrative analysis of scRNA-seq and ST data uncovered a potential link between neddylation and cell cycle progression across CRC specimens.Fig. 2. Neddylation correlates with the G2M phase, and UBE2M serves as a molecular link between neddylation and cell cycle control.a Proportions of cell cycle stages in high and low neddylation group of malignant epithelial cell clusters. b Linear relationship between neddylation score and quantified G2M phase score. The effect size was 0.182 (95% CI: 0.172 to 0.192). c Mapping of neddylation levels and cell cycle-related cluster in STs. The darker the red color, the higher the corresponding score. d A heat map showing the distribution of biological pathways and clinical features in patients with high- and low-neddylation CRC, as quantified by ssGSEA. e GSVA analysis of patients with high- and low-neddylation CRC. The FDR value was used for statistical analysis. f Distribution of G2M phase-related proteins and NEDD8 expression in patients with high- and low-neddylation CRC. Power analyses were as follows: CCNB1: effect size, 0.673, 95% CI, 0.578 to 0.767; CCNB2: effect size, 0.620, 95% CI, 0.527 to 0.714; CDK1: effect size, 0.927, 95% CI, 0.831 to 1.024; NEDD8: effect size, 0.370, 95% CI, 0.278 to 0.463. g Using hdWGCNA to scRNA-seq data to identify gene modules associated with both neddylation and the cell cycle. h Application of WGCNA to bulk-sequencing data to identify gene modules related to both neddylation and the cell cycle. i Identification of the target molecule by intersecting neddylation-related genes with modules selected by hdWGCNA and WGCNA. j PPI analysis of the intersected 15 genes. k The correlation of UBE2M and GPS1 with proteins associated with the G2M phase and NEDD8. l The selection process of the target gene. Error bars represent the mean ± s.d. ^^P < 0.05, ^^P < 0.01, ^^P < 0.001.

To validate our findings at the bulk RNA-seq level, we integrated TPM data from a total of 1,829 clinical patients. A heat map generated using the ssGSEA algorithm highlighted differences in clinical, molecular and biological pathway profiles between the high and low neddylation groups in patients with CRC, with the cell cycle clearly upregulated in the high neddylation group (Fig. 2d). GSVA analysis also confirmed that the pathway differences between the high and low neddylation groups included the G2M phase (Fig. 2e). In addition, we selected key molecules of the G2M checkpoint (CCNB1, CCNB2 and CDK1) and found that they were upregulated in the high neddylation group (P < 0.001). NEDD8, which is highly expressed in the high neddylation group, served as a validation of neddylation levels (P < 0.001) (Fig. 2f).

To elucidate the link between the neddylation pathway and the cell cycle, we used hdWGCNA for scRNA-seq (Supplementary Fig. 2h) and WGCNA for bulk RNA-seq (Supplementary Fig. 2i) to select gene modules associated with both neddylation and cell cycle subpopulations. First, through the analysis of scRNA-seq data, we identified 36 distinct gene modules (Supplementary Fig. 2g) and subsequently investigated their associations with various cNMF clusters and neddylation scores. Our screening revealed that only seven modules (M6, M7, M9, M14, M21, M29 and M31) demonstrated significant co-association with both cNMF cluster 4 (average expression > 1.5 and percent expressed ≥ 50%) and high neddylation score (average expression > 0.4 and percent expressed ≥ 40%) (Fig. 2g). Therefore, we extracted 947 genes from these modules. Subsequently, we performed analogous analyses of bulk RNA-seq data by correlating the eight identified gene modules with both the neddylation and cell cycle scores. Notably, only the MEcyan module showed the strongest association with these two traits (cor > 0.5, P < 0.001) (Fig. 2h). As a result, we selected 627 genes from the MEcyan module that showed the highest correlation with cell cycle pathways. By intersecting these two gene lists with a list of neddylation-related genes, we ultimately obtained 15 overlapping molecules. (Fig. 2i and Supplementary Table 11). Using protein–protein interaction (PPI) analysis, we selected molecules related to neddylation, ultimately identifying UBE2M and GPS1 (Fig. 2j). Then, correlation analysis suggested that UBE2M may exhibit a relatively stronger connection with G2M phase checkpoint regulation compared with GPS1 (Fig. 2k). Thus, our findings suggest that in CRC, UBE2M could act as a potential regulatory interface connecting the neddylation processes with cell cycle progression (Fig. 2l).

UBE2M is significantly upregulated in CRC and closely associated with the cell cycle

Previous bioinformatic results indicated that neddylation levels were upregulated in CRC tissues. To validate this conclusion, we conducted WB analysis of paired tumor–normal tissue samples from patients with CRC. Our findings revealed significantly increased levels of neddylated cullin family proteins, confirming the enhanced neddylation activity in tumor tissues (Supplementary Fig. 3a). We then focused on our target gene UBE2M, a biomarker of neddylation. We systematically analyzed its expression profile and explored potential downstream regulatory pathways. Our findings from the Tumor Immune Estimation Resource (TIMER) database, The Human Protein Atlas (HPA) and pan-cancer of TCGA bulk RNA-seq data revealed that UBE2M is among the highly expressed genes across various cancers, with notable overexpression in CRC (Fig. 3a and Supplementary Fig. 3b,c). Survival analysis on the KMplot website^40^ suggested that patients with higher UBE2M expression had poorer overall survival (P = 0.021), implying that UBE2M could be an oncogene (Fig. 3b). To validate our findings, we used clinical specimens from our center for further analysis. qRT–PCR analysis of 32 pairs of tumor and normal tissues confirmed that UBE2M messenger RNA levels were significantly elevated in tumors compared with those in normal tissues (Fig. 3c). WB analysis of 12 pairs of tumor and normal tissues further corroborated these results (Fig. 3d). To reinforce these findings, we used a TMA with 80 pairs of samples for validation, which consistently showed higher UBE2M protein expression in tumor tissues (Fig. 3e,f). These results indicate that UBE2M is overexpressed in tumors and associated with poor prognosis; however, the precise mechanisms of its regulation require further investigation.Fig. 3UBE2M is overexpressed in CRC, regulating the cell cycle and promoting tumor progression in vivo.a UBE2M mRNA levels in pan-cancer of TCGA bulk RNA-seq data. b Survival analysis between high and low UBE2M group on the KMplot website. c RT–qPCR analysis of 32 pairs of tumor and normal tissues from our center. d WB analysis of 12 pairs of tumor and normal tissues. e Representative IHC for UBE2M in tumor and adjacent normal tissues (n = 80). Scale bars, 100 μm. f AOD of UBE2M in tumor and normal tissues. g,h Downstream pathway prediction of UBE2M in GSCA. The FDR value was used for statistical analysis. i Downstream pathway prediction of UBE2M in our integrated bulk RNA-seq data. The FDR value was used for statistical analysis. j UBE2M mRNA levels across various cell lines. k UBE2M protein levels across various cell lines. l RNA-seq analysis was conducted on SW620 cells with overexpressed UBE2M and empty vector controls, followed by KEGG analysis. m Representative images of subcutaneous xenograft tumors derived from nude mice injected with DLD-1 cells with UBE2M knockdown and overexpression. n Analysis of tumor mass and weight across different groups. o H&E staining and IHC for UBE2M, Ki-67, PABPC1 and CCNB1 in the UBE2M knockdown and overexpression tumors. Error bars represent the mean (n = 3) ± s.d. ^^P < 0.05, ^^P < 0.01, ^^P < 0.001, ^***^P < 0.0001.

We predicted pathways potentially related to UBE2M using the Gene Set Cancer Analysis (GSCA) database^41^, identifying the cell cycle and apoptosis pathways as possible targets (Fig. 3g). In addition, in CRC, there were discernible differences in the cell cycle and apoptosis pathway scores between the high and low UBE2M expression groups (Fig. 3h and Supplementary Fig. 3d,e). Applying the high and low expression groups of UBE2M to our integrated bulk RNA-seq data further supported the involvement of UBE2M in cell cycle regulation (Fig. 3i and Supplementary Table 12). Building on this finding, we aimed to validate our hypothesis through a series of experiments. First, we initiated by confirming UBE2M expression in various colon cancer cell lines and normal intestinal epithelial cells. The results showed the highest expression in SW480 and DLD-1 cell lines, whereas there was no significant difference in SW620 cells (Fig. 3j,k). To delineate the regulatory effects of UBE2M overexpression, we performed RNA-seq analysis in SW620 cells transfected with either the UBE2M-expressing plasmid or empty vector control. The sequencing process was entrusted to Personalbio. Ultimately, KEGG enrichment analysis using the above data confirmed that UBE2M might be involved in the regulation of the cell cycle (Fig. 3l and Supplementary Table 13).

UBE2M promotes the malignant progression of CRC both in vivo and in vitro

Given these observations, we hypothesized that UBE2M may serve as a potential oncogenic driver in CRC. To validate this hypothesis, we conducted systematic functional experiments to characterize its role in CRC progression. We established loss-of-function models through UBE2M knockdown in SW480 and DLD-1 cells, complemented by gain-of-function models via UBE2M overexpression in SW620 cells. The efficiency of these genetic manipulations was rigorously validated through qRT–PCR and WB analyses at both transcriptional and translational levels (Supplementary Fig. 3f,g).

In vivo, we conducted subcutaneous tumorigenesis assays in nude mice, which revealed that mice injected with DLD-1 cells harboring UBE2M knockdown developed smaller and lighter tumors, whereas those injected with SW620 cells overexpressing UBE2M exhibited a greater tumor burden (Fig. 3m,n). Furthermore, immunohistochemical (IHC) analysis of these tumors revealed differential proliferation patterns: UBE2M-knockdown cohorts exhibited attenuation of Ki-67 staining intensity, while UBE2M-overexpressing groups showed enhanced Ki-67 nuclear positivity compared with control groups (Fig. 3o and Supplementary Fig. 3h).

On the basis of our hypothesis that UBE2M might participate in the regulation of the cell cycle and apoptosis, we proceeded with a suite of functional cellular assays. CCK-8, colony formation and EdU assays were performed to assess the effect of UBE2M on cell proliferation. The results indicated that UBE2M knockdown in SW480 and DLD-1 cells led to reduced CCK-8 growth (Fig. 4a), decreased colony numbers (Fig. 4c) and a lower count of EdU-positive cells (Fig. 4e). Conversely, overexpression of UBE2M in SW620 cells yielded opposite results (Fig. 4b,d,f). This confirmed that UBE2M positively regulates the proliferative capacity of CRC cells. Moreover, we observed a higher rate of apoptosis in UBE2M knockdown SW480 and DLD-1 cells than in UBE2M overexpressing SW620 cells, further validating our conclusions (Fig. 4g,h). In addition, cell cycle analysis revealed that knockdown of UBE2M in SW480 and DLD-1 cells mostly arrested them in the G2/M phase (Fig. 4i), whereas overexpression of UBE2M accelerated cell cycle progression in SW620 cells (Fig. 4j). Collectively, these results align with our initial hypothesis that UBE2M can stimulate cell cycle advancement, suppress apoptosis and thus promote cellular proliferation.Fig. 4UBE2M promotes CRC cells’ proliferation, inhibits apoptosis, and facilitates cell cycle progression in vitro.a CCK-8 analysis of UBE2M knockdown in SW480 and DLD-1 cells. b CCK-8 analysis of UBE2M overexpression in SW620 cells. c Representative images of colony formation assay in SW480 and DLD-1 cells transfected with Sh-NC and Sh-UBE2M. d Representative images of colony formation assay in SW620 cells transfected with empty vector and OE-UBE2M. e Representative images of EdU-positive cell counts in SW480 and DLD-1 cells transfected with Sh-NC and Sh-UBE2M. f Representative images of EdU-positive cell counts in SW620 cells transfected with empty vector and OE-UBE2M. g Analysis of apoptosis conducted in the above two cell lines with UBE2M knockdown utilizing flow cytometry. h Analysis of apoptosis conducted in UBE2M-overexpressed SW620 cells utilizing flow cytometry. i Cell cycle analysis of UBE2M-knockdown SW480 and DLD-1 cells with flow cytometry. j Cell cycle analysis of UBE2M-overexpressed SW620 cells with flow cytometry. Error bars represent the mean (n = 3) ± s.d. ^^P < 0.05, ^^P < 0.01, ^^P < 0.001.

UBE2M exerts an enhancing effect on PABPC1 at protein level

These findings suggest that UBE2M may play a regulatory role in the G2M phase; however, the underlying mechanisms remain unclear. In this segment of our research, we embarked on a preliminary exploration of the downstream molecular targets of UBE2M. Given the primary function of UBE2M as a neddylation enzyme, we conducted immunoprecipitation (IP) assays with UBE2M and subjected the associated proteins to silver staining and MS analysis (Fig. 5a and Supplementary Fig. 4a). By integrating proteomic data from TCGA-CRC, we identified three potential targets (Fig. 5b and Supplementary Table 14). Given that a previous study has reported the involvement of PABPC1 in the regulation of the G2M phase^42^, we hypothesized that PABPC1 might be a primary target of UBE2M and presented its MS profile (Fig. 5c). To validate the interaction between UBE2M and PABPC1, we first performed immunofluorescence staining to assess their subcellular colocalization. The results demonstrated that endogenous UBE2M and PABPC1 exhibit colocalization in the cytoplasmic compartment, suggesting a potential physical interaction between these two proteins (Fig. 5d). To further substantiate this finding, we conducted co-IP assays to investigate their interaction at the protein level. The results confirmed an endogenous interaction between UBE2M and PABPC1 in SW480 and DLD-1 cells (Fig. 5e,f). In addition, to rule out cell line-specific artifacts, we established an exogenous expression system in HEK293T cells, where both proteins were cotransfected. Consistent with the endogenous results, co-IP assays confirmed the interaction between UBE2M and PABPC1 in this heterologous system (Fig. 5g,h). shRNA-mediated UBE2M depletion resulted in no significant change in PABPC1 mRNA abundance (qRT–PCR, P > 0.05), indicating that this regulation may occur at the protein level (Supplementary Fig. 4b). Knockdown of UBE2M in SW480 and DLD-1 cells led to decreased PABPC1 protein expression (Fig. 5i,j), whereas UBE2M overexpression in SW620 cells resulted in increased PABPC1 (Fig. 5k). To substantiate these findings, we overexpressed UBE2M in SW620 cells in a dose-dependent manner and observed a corresponding increase in PABPC1 expression (Fig. 5l).Fig. 5UBE2M interacts with PABPC1 and inhibits its ubiquitination.a The top 10 proteins identified to interact with UBE2M through IP–MS analysis. b The intersection between the top 10 IP–MS-identified proteins and UBE2M-associated proteins from the LinkedOmics database. c Through IP and liquid chromatography–MS analysis, specific peptides of PABPC1 were detected. d Immunofluorescence staining confirming the colocalization of UBE2M with PABPC. e,f UBE2M and PABPC1 interactions in SW480 and DLD-1 cells confirmed by co-IP at the endogenous level. g,h UBE2M and PABPC1 interactions in HEK293T cells confirmed by co-IP at the exogenous level. i,j Monitoring alterations in PABPC1 upon UBE2M knockdown in SW480 and DLD-1 cells. k Observation of alterations in PABPC1 upon UBE2M overexpression in SW620. l Observing changes in PABPC1 with dose-dependent overexpression of Flag–UBE2M in SW620 cells. m–o The impact of UBE2M knockdown (SW480, DLD-1) or overexpression (SW620) on PABPC1 protein stability was assessed by treating cells with CHX (100 μg/ml) and monitoring PABPC1 levels at 0, 2, 4 and 8 h after treatment. p,q The effect of UBE2M knockdown on PABPC1 in SW480 and DLD-1 cells treated with MG132 (10 μM) for 6 h. r Evaluation of PABPC1 ubiquitilation linkage after cotransfecting HEK293T cells with Myc–PABPC1, Flag–UBE2M and various HA-Ub plasmids, specifically wild type, Lys11-only, Lys48-only and Lys63-only forms. s Observation of the ubiquitination levels of Myc–PABPC1 in DLD-1 cells with UBE2M knockdown. t Comparison of the neddylation levels of PABPC1 between UBE2M knockdown and control groups after the addition of the neddylation inhibitor MLN4924 (100 nM) in DLD-1.

Therefore, understanding the mechanism by which UBE2M positively regulates PABPC1 at the protein level is an issue that warrants further investigation. We hypothesized that UBE2M might either influence PABPC1 through the ubiquitin–proteasome pathway or directly neddylate PABPC1. To test our hypothesis, we performed a CHX (100 μg/ml) chase assay and monitored PABPC1 degradation over a time course (0, 2, 4 and 8 h). This analysis revealed that UBE2M knockdown in SW480 and DLD-1 cells promoted PABPC1 degradation, whereas UBE2M overexpression in SW620 cells slowed its degradation (Fig. 5m–o and Supplementary Fig. 4d–f). Furthermore, we discovered that in UBE2M knockdown SW480 and DLD-1 cells, PABPC1 was rescued by the proteasome inhibitor MG132 (Fig. 5p,q). This suggests that the UBE2M regulation of PABPC1 might be mediated by the ubiquitin–proteasome system. By using a series of ubiquitin mutants in which only one Lys residue was preserved while the others were substituted with Arg residues, we then corroborated this hypothesis in HEK293T cells, showing that UBE2M can prevent the degradation of PABPC1 by inhibiting K48-linked polyubiquitination (Fig. 5r). Consistently, IP assays targeting PABPC1 revealed that whereas knockdown restored the abundance of ubiquitinated PABPC1 (Fig. 5s), overexpression of UBE2M suppressed the ubiquitinated PABPC1 (Supplementary Fig. 4c). Given the low likelihood of UBE2M (an E2 enzyme) directly inhibiting PABPC1 ubiquitination, we hypothesized an indirect mechanism or potential neddylation by UBE2M that stabilizes PABPC1. To test this, we first confirmed E4B, which was reported to degrade PABPC1^43^, as a relevant E3 ligase for PABPC1 in our CRC model (Supplementary Fig. 4g,h). Subsequently, an in vitro ubiquitination assay demonstrated that increasing amounts of purified UBE2M did not alter the ubiquitination status of PABPC1 (Supplementary Fig. 4i). We further investigated the potential of UBE2M to modulate the neddylation of PABPC1. Although NEDD8 was found to be associated with PABPC1, and this association could be abolished by the neddylation inhibitor MLN4924, the neddylation level of PABPC1 showed no significant alteration upon UBE2M knockdown when compared with the negative control group. (Fig. 5t). These findings suggest that UBE2M is not directly responsible for neddylation of PABPC1; instead, it appears to modulate PABPC1 levels by indirectly inhibiting its ubiquitination.

The elevation of PABPC1 correlated with USP39-mediated deubiquitination, which is modulated by UBE2M

In light of the above findings, we propose that UBE2M does not directly modulate PABPC1; instead, it exerts regulatory effects through certain intermediaries or pathways. On the basis of the intersection of our IP–MS data and proteins correlated with PABPC1 in TCGA-CRC, we identified seven candidate proteins (Fig. 6a and Supplementary Table 14). Given that the elevation of PABPC1 is dependent on the inhibition of its ubiquitin-dependent degradation, and USP39 is a deubiquitinating enzyme^44^, we considered it plausible that USP39 participates in the deubiquitination of PABPC1, with its MS profile shown in Supplementary Fig. 5a. First, we used immunofluorescence staining to confirm the intracellular colocalization of USP39 and PABPC1 at the endogenous expression level (Fig. 6b). Subsequently, molecular docking with PyMOL (v2.5.4) revealed a potential binding between the two proteins (Fig. 6c). To substantiate this, we validated the interaction between USP39 and PABPC1 at both endogenous and exogenous levels through IP and WB analyses (Fig. 6d,e). In addition, we constructed truncated USP39 and PABPC1 mutants to confirm their interactions (Fig. 6f,h). We found that the M1 region of USP39 and the M1 region of PABPC1 were capable of binding (Fig. 6g,i), which is in line with the molecular docking results.Fig. 6USP39 facilitates the deubiquitination of PABPC1, a process regulated by UBE2M.a The intersection of the top 10 IP–MS-identified proteins and PABPC1-associated proteins from the LinkedOmics database. b Immunofluorescence staining confirming the colocalization of USP39 with PABPC1. c Molecular docking of USP39 and PABPC1 using PyMOL. d,e Co-IP of USP39–PABPC1 in SW480 and DLD-1 cells indicates the endogenous combination, with similar results observed in exogenous conditions in HEK293T cells. f,g Design of Myc–PABPC1 truncation mutants and co-IP validation of their binding to His-USP39 in HEK293T. h,i Design of His-USP39 truncation mutants and co-IP validation of their binding to His-USP39 in HEK293T. j Transfecting HEK293T with a positive gradient of wild type USP39 and its deubiquitination-inactive mutant ΔC306A to assess their effects on PABPC1. k The effect of USP39 knockdown on PABPC1 in SW480 cells treated with MG132 (10 μM) for 6 h. l,m The impact of USP39 knockdown in SW480 cells and overexpression in SW620 cells on PABPC1 treated with CHX. n In vivo ubiquitination assay in HEK293T cells demonstrating the effects of USP39-WT and its catalytically inactive mutant ΔC306A on PABPC1 ubiquitination. o After transfecting SW480 cells with K48R-mutated and wild type Ub plasmids, the effects on PABPC1 with and without USP39 knockdown were compared. p HEK293T cells were transfected with different site-mutated Myc–PABPC1 variants, His-USP39 and HA-Ub to elucidate the ubiquitination sites on PABPC1. q After transfecting HEK293T cells with K361R-mutated and wild type Ub plasmids, the effects on PABPC1 with and without USP39 knockdown were compared. r In HEK293T cells, His-USP39 was cotransfected with a gradient of Flag–UBE2M overexpression to observe the impact on PABPC1 expression levels. s Combined depletion of USP39 and UBE2M through shRNA-mediated cosilencing in SW480 cells demonstrated synergistic control over PABPC1 protein levels. t Reconstitution assay in HEK293T cells transfected with Flag–UBE2M, His-USP39 (wild type/catalytically inactive C306A), Myc–PABPC1 and HA-K48-linked ubiquitin demonstrates: (1) potentiation of the deubiquitinating enzyme activity of USP39 toward PABPC1 by UBE2M and (2) UBE2M-mediated PABPC1 regulation requiring functional USP39.

To investigate whether USP39 acts as a deubiquitinase for PABPC1, we systematically validated its enzymatic activity through multiple experimental approaches. First, USP39 overexpression elevated PABPC1 protein levels, whereas the catalytically inactive ΔC306A USP39 mutant^43^ failed to exert this stabilization effect (Fig. 6j) in HEK293T cells. Conversely, USP39 depletion in SW480 and DLD-1 cells reduced PABPC1 abundance, an effect abolished by proteasomal inhibition with MG132 (Fig. 6k and Supplementary Fig. 5b). CHX chase assays further demonstrated that USP39 knockdown accelerated PABPC1 degradation, while its overexpression prolonged the protein half-life of PABPC1 (Fig. 6l,m and Supplementary Fig. 5c,d), collectively suggesting that USP39-mediated PABPC1 stabilization occurs through the ubiquitin–proteasome system. To further validate that PABPC1 serves as a substrate for USP39, we performed both in vivo and in vitro deubiquitination assays. In cellular systems, USP39 knockdown enhanced the ubiquitination levels of PABPC1 (Supplementary Fig. 5e). Furthermore, exogenous expression of USP39-WT, but not the catalytically inactive USP39–ΔC306A mutant, reduced the ubiquitination of PABPC1 (Fig. 6n). The in vitro deubiquitination assay demonstrated that purified USP39-WT gradually decreased PABPC1 ubiquitination over time, whereas USP39–ΔC306A failed to exhibit this deubiquitination activity (Supplementary Fig. 5f). In addition, USP39 knockdown-induced proteasomal degradation of PABPC1 was reversed by the expression of the K48R ubiquitin mutant, demonstrating that USP39 counteracts K48-linked polyubiquitination to stabilize PABPC1 (Fig. 6o). These results confirmed that PABPC1 is regulated by the deubiquitinating activity of USP39. Using the GPS-Uber website (https://gpsuber.biocuckoo.cn), we predicted the top three ubiquitination sites of PABPC1 (Supplementary Fig. 5g) and performed mutation site validation. Subsequently, it was identified that the deubiquitination of PABPC1 by USP39 potentially targeted the K361 site. (Fig. 6p). Similarly, USP39 knockdown did not affect the deubiquitination of the K361R mutant Myc–PABPC1 (Fig. 6q).

The regulatory role of UBE2M on the USP39–PABPC1 axis remains to be elucidated. Compared with the overexpression of USP39 alone, the co-overexpression of UBE2M and USP39 in HEK293T cells synergistically enhanced the protein abundance of PABPC1. Importantly, this augmentation exhibited a dose-dependent relationship with UBE2M protein levels (Fig. 6r). In parallel, dual genetic ablation of UBE2M and USP39 elicited maximal suppression of PABPC1 expression, achieving significantly lower levels than those observed under single USP39 knockdown conditions (Fig. 6s). To determine whether the UBE2M-mediated regulation of PABPC1 requires USP39 activity, we performed rescue experiments in HEK293T cells by transfecting either wild type USP39 or its catalytically inactive mutant. As expected, UBE2M specifically enhanced USP39-WT-dependent deubiquitination of PABPC1, whereas this regulatory effect was completely abolished in cells expressing the USP39–ΔC306A mutant (Fig. 6t).

UBE2M mediates the neddylation of USP39, thereby enhancing its deubiquitination of PABPC1

We confirmed that UBE2M modulates PABPC1 expression in a USP39-dependent manner; however, the detailed mechanisms of this dependency are not well understood. Knockdown of UBE2M revealed no decrease in USP39 expression (Fig. 7a). Thus, we explored whether UBE2M could facilitate the binding between USP39 and PABPC1 and discovered that UBE2M indeed promotes their binding at both endogenous and exogenous levels (Fig. 7b,c). However, the specific mechanisms involved require further elucidation. Given the role of UBE2M as an E2 enzyme in neddylation, we questioned whether it might neddylate USP39, thereby enhancing its binding with PABPC1. We first verified that the knockdown and overexpression of UBE2M did not globally alter the expression of other neddylation-related proteins (Supplementary Fig. 6a). Our findings in DLD-1 and HEK293T cell lines confirmed that USP39 is neddylated, which is inhibited by MLN4924 (Selleckchem) (Fig. 7d,e). Subsequently, we observed that UBE2M knockdown led to reduced neddylated USP39, while overexpression had the opposite effect (Fig. 7f,g). In addition, knockdown of NEDD8 also diminished USP39 neddylation (Fig. 7h). Collectively, these results confirmed that UBE2M is capable of neddylating USP39.Fig. 7UBE2M neddylates USP39, thereby regulating the deubiquitination of PABPC1, leading to an increase in CCNB1.a Effects of UBE2M knockdown on USP39 in DLD-1. b The effect of UBE2M knockdown on the interaction between USP39 and PABPC1 in DLD-1 cells. c Transfection of Flag–UBE2M, His–USP39 and Myc–PABPC1 into HEK293T cells to observe whether UBE2M can promote the exogenous interaction between USP39 and PABPC1. d We incubated DLD-1 cells with MLN4924 and performed IP experiments to observe the inhibition of USP39 neddylation. e We transfected His–USP39 and NEDD8 into HEK293T cells, along with DMSO and MLN4924, and conducted IP experiments to observe whether the neddylation of USP39 was decreased by MLN4924. f Validation of the effect of UBE2M knockdown in DLD-1 cells on USP39 neddylation. g In HEK293T cells, exogenous Flag–UBE2M was transfected to evaluate the neddylation status of USP39. h Knockdown of NEDD8 in DLD-1 cells to observe the extent of USP39 neddylation. i We transfected the previously designed USP39 truncation constructs into HEK293T cells to identify which segment is involved in the neddylation process. j We transfected a series of His–USP39 lysine site mutants into HEK293T cells to determine which site is involved in neddylation regulation. k We transfected wild type and K6R mutant His–USP39 into HEK293T cells to compare alterations in the expression of PABPC1. l We transfected wild type and K6R mutant His–USP39 into HEK293T cells to compare the effects on the binding affinity between USP39 and PABPC1. m Comparison of the effects of transfecting USP39-K6R and USP39-WT plasmids on the deubiquitination of PABPC1 in HEK293T cells. n Evaluation of the impact of UBE2M overexpression on PABPC1 expression in SW620 cells transfected with either USP39-K6R or USP39-WT plasmids.

N-terminal domain mapping via truncation mutagenesis, combined with anti-NEDD8 immunoblotting, localized the neddylation-responsive region to the first 100 amino acids of USP39 (Fig. 7i). Systematic screening of lysine residue within this domain via site-directed mutagenesis (K6R/K16R/K29R/K51R/K73R/K91R) revealed K6 as the primary neddylation acceptor site, with a weakened NEDD8 signal observed exclusively in the K6R mutant (Fig. 7j). To verify that USP39-K6R lost its ability to deubiquitinate PABPC1, we conducted further protein functional validation in HEK293T cells. Compared with wild type USP39, the K6 mutant USP39 failed to increase PABPC1 expression (Fig. 7k) and showed reduced binding to PABPC1 (Fig. 7l) as well as increased ubiquitination of PABPC1 (Fig. 7m). In addition, we conducted an in vitro deubiquitination assay to verify that neddylated USP39 can enhance the deubiquitinating activity toward PABPC1. Nevertheless, this phenomenon was reversed by the K6R mutant of USP39 (Supplementary Fig. 6b). Furthermore, we found that the regulation of the deubiquitination of PABPC1 by UBE2M is dependent on USP39, and this function is negated by the K6 mutation in USP39 (Fig. 7n and Supplementary Fig. 6g), thereby reinforcing the notion that UBE2M enhances PABPC1 deubiquitination by neddylating USP39. Additional rescue experiments assessing cell proliferation and cell cycle phenotypes confirmed that the USP39-K6R mutant failed to promote CRC progression (Supplementary Fig. 7). Given the role of the UBE2M–USP39–PABPC1 axis in cell cycle regulation, we investigated whether its components exhibit cell cycle-dependent expression. We synchronized CRC cells at the G1/S phase using a double thymidine block and monitored protein levels at subsequent timepoints. As shown in Supplementary Fig. 6d, UBE2M and PABPC1 displayed synchronous fluctuations: their expression was low at G1/S, increased at early G2 and peaked during G2/M transition. Consistently, PABPC1 protein stability was also the highest in G2/M (Supplementary Fig. 6e). By contrast, USP39 protein abundance remained constant throughout the cycle. As UBE2M does not regulate USP39 expression but promotes its neddylation, we examined neddylation levels across phases and found that USP39 neddylation, unlike its expression, was markedly elevated in G2 and peaked at G2/M (Supplementary Fig. 6f).

Building on previous studies that confirmed that PABPC1 binds eIF4G to facilitate the formation of the translation initiation complex and boost mRNA translation efficiency^45,46^, we reaffirmed that PABPC1 binds eIF4G and is positively regulated by the UBE2M–USP39 axis, whereas this regulation was abolished by USP39-K6R (Supplementary Fig. 6h). Moreover, previous studies have demonstrated that PABPC1, by binding to eIF4G, can enhance the translation efficiency of CCNB1^42^. Given that CCNB1 is a key regulatory protein of the G2M phase, and UBE2M is known to regulate the G2M phase, we hypothesized the existence of a UBE2M–USP39–PABPC1–CCNB1 axis that modulates the G2M phase in CRC. Our RIP assays confirmed that PABPC1 protein indeed binds to CCNB1 mRNA (Supplementary Fig. 6i). We further confirmed that UBE2M is incapable of modulating the mRNA level of CCNB1 (Supplementary Fig. 6c), which indicated that the regulation occurs at the protein level. Further validation revealed that UBE2M promoted CCNB1 protein expression through the USP39–PABPC1 axis, whereas USP39-K6R transfection in SW620 cells abolished this regulatory effect. (Supplementary Fig. 6j). This revealed that the neddylation of USP39 by UBE2M enhanced PABPC1 deubiquitination and stability, leading to increased CCNB1 protein.

Employing micafungin to target UBE2M elicits promising therapeutic potential

In pursuit of better clinical translation, we noted recent literature reporting on micafungin, an inhibitor of UBE2M^47^. This study revealed that micafungin could impede UBE2M-mediated neddylation. Intriguingly, it also had the effect of stabilizing UBE2M. This prompted us to investigate the potential efficacy of this inhibitor in CRC treatment. To ascertain the functional impact of micafungin on CRC cells, we first determined the half-maximal inhibitory concentration (IC_50_) of this UBE2M inhibitor in DLD-1 cells to be 42.83 μM (Supplementary Fig. 8a). We then confirmed that micafungin diminished the levels of CCNB1 mRNA bound by an equivalent amount of PABPC1 (Fig. 8a). Subsequently, we validated on the WB level that micafungin could attenuate the expression of PABPC1 and CCNB1, whereas the expression level of UBE2M was not affected, which was consistent with previous findings^47^ (Fig. 8b). Further mechanistic analysis demonstrated that micafungin pharmacologically suppressed USP39 neddylation and augments ubiquitin-mediated degradation of PABPC1 (Supplementary Fig. 8b,c). Using micafungin at 43 μM, we observed its capacity to suppress cell proliferation (Fig. 8c–e), induce apoptosis (Fig. 8f) and cause cell cycle arrest at the G2M phase (Fig. 8g). In the xenograft model, intravenous administration of micafungin, commencing on day 10 at a daily dose of 30 mg/kg (Supplementary Fig. 8d), resulted in a significant reduction in tumor volume and mass compared with the control group receiving DMSO injections (Fig. 8h). These findings position UBE2M as a valuable prognostic indicator, and its inhibitor micafungin as a potential therapeutic agent for CRC.Fig. 8. Micafungin-induced UBE2M suppression attenuated the progression of CRC.a Comparison of the extent to which PABPC1 binds to CCNB1 mRNA under the influence of DMSO and micafungin (43 μM). b Evaluation of how UBE2M regulates the PABPC1–CCNB1 interaction under the influence of DMSO and micafungin (43 μM). c–e Evaluation of the capacity of micafungin (43 μM) to inhibit cell proliferation in vitro through CCK-8, colony formation and EdU-positive cell count assays in DLD-1 cells. f Evaluation of the ability of micafungin (43 μM) to inhibit cell apoptosis through apoptosis assays in DLD-1 cells. g Evaluation of the ability of micafungin (43 μM) to inhibit cell cycle progression through cell cycle analysis in DLD-1 cells. h In vivo, starting on day 10 after injecting DLD-1 cells into nude mice, we administered intravenous injections of DMSO and micafungin (30 mg/kg) until tumor excision and subsequently calculated the tumor volume and weight of the treatment groups. i A schematic diagram of the entire study’s findings on phenotype and mechanisms. Error bars represent the mean (n = 3) ± s.d. ^^P < 0.05, ^^P < 0.01, ^^P < 0.00.

UBE2M-mediated regulatory axis as a prognostic biomarker and therapeutic vulnerability in CRC

In the above mouse xenograft tumor experiment, we also found that both knockdown and overexpression of UBE2M could lead to coordinated changes in the expression levels of PABPC1 and CCNB1, which serves as compelling evidence (Fig. 3o and Supplementary Fig. 3h). To validate the reliability of this axis, we additionally conducted a series of rescue experiments to confirm its role in promoting the progression of CRC. In a subcutaneous tumor-bearing mouse model, we found that tumor suppression resulting from UBE2M inhibition could be reversed by overexpressing PABPC1 (Supplementary Fig. 8e–g), and IHC analysis together with hematoxylin and eosin (H&E) staining confirmed the validity of this regulatory axis (Supplementary Fig. 8h). Furthermore, in cellular systems, we discovered that the inhibition of proliferation of CRC cells owing to UBE2M knockdown could be rescued by OE-PABPC1, while the proliferation-promoting effect of UBE2M overexpression was suppressed by Sh-PABPC1 (Supplementary Fig. 9a–c). Similarly, apoptosis and cell cycle arrest induced by UBE2M knockdown were reversed by OE-PABPC1, and vice versa (Supplementary Fig. 9d,e). These data collectively demonstrate that the pro-tumorigenic activity of UBE2M in CRC progression is mechanistically linked to its regulation of PABPC1 stability and function. Consistent with the previous results, the expressions of PABPC1 and CCNB1 were higher in tumor tissues (Supplementary Fig. 10a). In addition, Supplementary Fig. 10b,c shows our TMA profiles for UBE2M and PABPC1. PABPC1 exhibited higher average OD (AOD) in tumor tissues, which was consistent with our previous conclusion (Supplementary Fig. 10d). After stratifying UBE2M on the basis of IHC scoring, we found that the high-expression group correlated with T stage and tumor size (Supplementary Table 15). We also observed that patients with higher UBE2M expression levels tended to exhibit relatively elevated PABPC1 levels (Supplementary Fig. 10e). Furthermore, analysis of our TMA cohort demonstrated a robust association between elevated levels of UBE2M and PABPC1 and poorer patient prognosis, as indicated by both Cox regression and Kaplan–Meier survival analysis (Supplementary Fig. 10f and Supplementary Table 15). Ultimately, we developed a comprehensive mechanistic diagram to delineate the molecular pathway through which UBE2M promotes CRC progression (Fig. 8i).

Discussion

Post-translational modifications of proteins are essential for the regulation of cellular physiology^48^, with neddylation emerging as a significant area of focus in oncological research^49^. Neddylation modulates the functionality, stability, subcellular localization and interaction networks of the target proteins^50,51^. In the realm of cell cycle regulation, a multitude of studies have demonstrated the involvement of various post-translational modifications in ensuring precise and orderly progression through cell cycle phases^52^. However, the detailed mechanisms by which neddylation influences cell cycle regulation in CRC cells have not yet been fully elucidated. Considering that disruption of cell cycle control is a critical hallmark of cancer development^53^, delving into the association between neddylation and the cell cycle in CRC is of paramount importance. Such insights are not only vital for unraveling the pathogenesis of CRC but also hold the potential to uncover key leads for the development of innovative diagnostic and therapeutic approaches.

In this study, we used a combination of single-cell transcriptomics and bulk transcriptomics to systematically investigate the relationship between neddylation and the cell cycle in CRC. At the single-cell level, we identified a robust correlation between neddylation and a tumor cluster associated with the cell cycle. At the organismal level, we found neddylation to be correlated with the G2M phase. These insights enhance our understanding of CRC pathogenesis at the cell cycle level and underscore the significance of neddylation as a novel regulatory node in cell cycle control. Utilizing powerful bioinformatics tools such as hdWGCNA, WGCNA and PPI, we successfully identified UBE2M, a neddylation marker, as a bridge connecting neddylation and the cell cycle. This not only enriches the repository of molecular markers for CRC but also provides potential biomarkers for precise diagnosis and prognosis assessment.

Most importantly, we experimentally investigated the molecular mechanisms by which UBE2M advances CRC progression. We confirmed elevated UBE2M expression in CRC via IHC, qRT–PCR and WB analysis, and found that high expression is associated with poor clinical outcomes, consistent with prior studies^6,26^. Previous research has established that UBE2M knockdown induces G2M arrest in lung and esophageal cancers through the inhibition of cullin neddylation and upregulation of CRL substrates such as p21, p27 and Wee1^14,15^. In liver cancer, UBE2M stabilizes β-catenin, thereby promoting Cyclin D1 and advancing the G1/S transition in CRC^16^. In breast cancer, UBE2M drives progression and therapy resistance by forming a positive feedback loop with the estrogen receptor^12^. Despite its conserved protumor role in multiple cancers, UBE2M may use tissue-specific mechanisms, a possibility that has not been systematically examined in CRC. Therefore, we conducted mRNA sequencing on cells overexpressing UBE2M compared with control vector-transfected cells, uncovering the downstream effects that are intimately associated with the cell cycle. Subsequent in vivo and in vitro validations have corroborated the capacity of UBE2M to act as a stimulant for the proliferation of CRC, an inhibitor of apoptosis, and, most importantly, a facilitator for progression through the G2M phase. Building on this, we used a series of IP–MS and ubiquitination-related WB experiments to demonstrate that UBE2M can suppress ubiquitination of PABPC1, which has been reported to participate in G2M phase progression^42^. However, UBE2M, a neddylation enzyme, does not directly modulate PABPC1 ubiquitination levels. Inspired by the IP–MS results, we proposed USP39 as a potential linker between UBE2M and PABPC1. Although USP39 is recognized as a spliceosome component, its deubiquitinating function has increasingly garnered attention^44,54^. Our WB analysis confirmed the ability of USP39 to deubiquitinate PABPC1, a process augmented by UBE2M. However, the precise mechanisms underlying this augmentation remain unclear. Given that UBE2M functions as a neddylation E2 enzyme, we hypothesized that UBE2M could mediate the neddylation of USP39, thereby promoting its binding to PABPC1, and this was experimentally verified. Subsequently, we also demonstrated that PABPC1 could enhance CCNB1 expression in CRC, and this process was under the regulatory influence of UBE2M. Therefore, dissection of the UBE2M–USP39–PABPC1–CCNB1 axis offers a novel perspective for understanding the malignant progression of CRC and provides a solid theoretical foundation for the development of therapeutic strategies targeting neddylation and the cell cycle. From the standpoint of clinical translation, we discovered that micafungin could inhibit the neddylation ability of UBE2M in CRC. These research findings are expected to pioneer new avenues for the diagnosis, personalized treatment and drug development of CRC, bringing new hope for improving the prognosis of patients with CRC.

Compared with previous studies on neddylation, this study distinguishes itself by unveiling the intricate connection between neddylation and the advancement of the G2/M phase in CRC at both the single-cell and bulk transcriptome levels. We delineated a new signaling cascade from UBE2M to the regulation of the cell cycle in CRC. This work not only remedies the previous lack of research on the interplay between neddylation and cell cycle dynamics in CRC but also offers a robust framework for synthesizing findings on neddylation across various cancer types.

This study has certain limitations. Although scRNA-seq and bulk RNA-seq provide valuable information, we require extensive proteomics data for validation. In addition, the limited sample size may introduce bias; thus, expanding the sample size and including a more diverse patient cohort could enhance the precision of the study. Regarding signaling pathways, while we identified the role of the UBE2M–USP39–PABPC1–CCNB1 axis in CRC progression, the complexity of cell cycle regulation suggests that additional neddylation-related pathways may interact with or parallel this axis. Elucidating the full interactome through TurboID, affinity purification–MS and systematic perturbation constitutes a critical and logical next step. For instance, our sequencing results further revealed that there may be a ribosome regulatory pathway downstream of UBE2M. Moreover, among the top 10 proteins enriched by MS, RPL11, RPL14 and RPL19 were all directly associated with ribosomes. This phenomenon implies that UBE2M may regulate the cell cycle through a broad spectrum of molecular pathways. Thus, future work should leverage systems biology to decipher intricate neddylation networks in CRC cell cycle control. The intricacies of the tumor microenvironment are difficult to replicate in vitro. Therefore, the development of more advanced models, such as three-dimensional tumor tissues or humanized animal models, is warranted to validate our findings. In summary, this study uncovered a new link between neddylation and the cell cycle in CRC, solidifying the role of UBE2M as a connector and elucidating the UBE2M–USP39–PABPC1 regulatory mechanisms of the G2M phase, offering novel avenues for CRC therapeutics.

Supplementary information

Supplementary Information Supplementary Tables