Abstract Background Ubiquitination—a pivotal post-translational modification that orchestrates cellular homeostasis and oncogenic pathways—remains underexplored as a pancancer regulatory hub. Although ubiquitination dysregulation is linked to tumor progression, a comprehensive, multicancer framework integrating prognostic, molecular, and microenvironmental landscapes is lacking. Methods This study integrated data of 4,709 patients from 26 cohorts across five solid tumor types (lung cancer, esophageal cancer, cervical cancer, urothelial cancer, and melanoma) and mapped the molecular profiles to the interaction network. Cox regression and the Kaplan-Meier survival method were employed for prognostic analysis. Functional enrichment and protein–protein interaction analyses were performed to identify the key downstream pathways and genes. Findings were validated using independent patient cohorts, cell line models, and in vivo experiments. Results Key nodes and prognostic pathways within the ubiquitination-modification network were identified. A conserved ubiquitination-related prognostic signature (URPS) effectively stratified patients into high-risk and low-risk groups with distinct survival outcomes across all analyzed cancers. URPS may serve as a novel biomarker for predicting immunotherapy response, with the potential to identify patients who are more likely to benefit from immunotherapy in clinical settings. A comprehensive analysis of URPS-associated proteins revealed novel cancer-related interaction partners as potential drug targets. At the single-cell resolution, URPS enabled more precise classification of distinct cell types and was associated with macrophage infiltration within the tumor microenvironment. In vivo, in vitro, and patient cohort analyses, demonstrated that OTUB1-TRIM28 ubiquitination plays a crucial role in modulating MYC pathway and influencing patient prognosis. Conclusion We constructed a pancancer ubiquitination regulatory network and prognostic model, revealing important pathways, and offering insights into predicting patient prognosis and understanding biological mechanisms. Keywords: Histology, Immunotherapy, Lung Cancer, Esophageal Cancer, Kidney Cancer __________________________________________________________________ WHAT IS ALREADY KNOWN ON THIS TOPIC * Ubiquitination is a critical, reversible, and enzymatically regulated process essential for diverse cellular functions including proteolysis, metabolism, signaling, and cell cycle regulation. It plays a complex and crucial regulatory role in cancer development, progression, metabolic reprogramming, and impacts immunotherapy efficacy. WHAT THIS STUDY ADDS * This study demonstrates that the OTUB1-TRIM28 ubiquitination regulatory enzyme influences the histological fate of cancer cells by modulating MYC and its downstream, and altering oxidative stress, ultimately leading to immunotherapy resistance and poor prognosis in patients. Ubiquitination score positively correlates with squamous or neuroendocrine transdifferentiation in adenocarcinoma. We established a ubiquitination-related prognostic signature for predicting overall survival in pancancer patients receiving surgery or immunotherapy. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY * This study unveils a novel strategy for drug development targeting traditionally “undruggable” targets like MYC, whereby ubiquitination regulatory modifiers for such targets are screened through constructed pancancer ubiquitination regulatory network, providing new therapeutic alternatives for improving immunotherapy efficacy and patient prognosis. Background Ubiquitination is the second most common and critically important post-translational modification (PTM) of proteins following phosphorylation.[39]^1 This process has immense significance and plays a critical role in cellular processes, including metabolism, protein degradation, signal transduction, and regulation of the cell cycle and gene expression.[40]^2 Ubiquitination plays an essential and complex role in these cellular processes, intricately contributing to the maintenance of cellular function and proper function. The ubiquitin–proteasome system comprises ubiquitin and its degradation by the proteasome. Ubiquitination is responsible for 80–90% of cellular proteolysis,[41]^3 and is a reversible PTM. The ubiquitination process is regulated through a cascade of reactions mediated by ubiquitin-activating enzymes, ubiquitin-conjugating enzymes, and ubiquitin ligases.[42]^4 Ubiquitin or ubiquitin chains attached to substrate proteins can be removed by deubiquitinating enzymes. Ubiquitination plays a crucial regulatory role in the domain of tumor regulation and is strongly implicated in various aspects of cancer development and progression. Specifically, it regulates tumor metabolic reprogramming and is involved in various processes, including cell survival, proliferation, and differentiation.[43]^5 Additionally, it influences the protein levels of programmed cell death 1/programmed cell death ligand 1 in the tumor microenvironment (TME), thereby increasing the efficacy of immunotherapy.[44]^6 Squamous cell carcinoma (SQC), adenocarcinoma (ADC), and neuroendocrine carcinoma (NEC) are the three major histological subtypes of malignant tumors. Tumors from different organs with common embryonic cell origins, differentiation pathways, and histological characteristics may exhibit similar characteristics related to carcinogenic processes.[45]^7 For instance, both SQCs and ADCs, and in some cases NECs in some cases, may share specific molecular signatures or pathways involved in tumor development. Notably, tumors may simultaneously exhibit characteristics of two of these distinct histological subtypes. For example, a tumor may exhibit a mixture of SQC and ADC characteristics or a combination of ADC and NEC features.[46]^8 Moreover, tumors can also undergo transformations between subtypes,[47]^9 further complicating diagnosis and treatment. In cases where clinical presentations are concentrated, the relationship between the pathological characteristics and malignant biological behavior remains unclear. Patients with cervical ADCs have worse prognosis than those with SQCs, particularly following immunotherapy retreatment.[48]^10 With the extensive application of immunotherapy in the treatment of malignancies, melanoma and urothelial carcinoma are frequently referenced in immunotherapy research.[49]^11 In addition to assessing how different pathological types influence prognosis, researchers are investigating the potential mechanisms through which these tumor subtypes influence responses to immunotherapy. Additionally, cervical NECs may have a distinct prognosis and behaviors, showing a greater propensity for metastasis and stromal invasion. However, lung ADCs exhibit a more favorable prognosis than SQCs, whereas lung NECs typically exhibit the poorest prognosis.[50]^12 13 Transcriptome data from The Cancer Genome Atlas (TCGA) comprehensively depicts the molecular characteristics of various cancers.[51]^14 The pathological features of specific ADCs and SQCs are included in the datasets of lung, esophageal, and cervical cancers. These data enhance our understanding of tumor progression and patient outcomes at the cellular and biological level. TCGA-based studies have identified high-frequency mutations in genes related to the cell cycle, receptor tyrosine kinase signaling, squamous cell differentiation, and chromatin remodeling.[52]^15 Such insights facilitate cancer classification and subtyping, enabling the identification of patient subgroups with shared molecular profiles for personalized treatment. However, most studies rely on established pathological classifications rather than molecular profiling for histological tumor classification, which may lead to inaccurate findings and suboptimal therapeutic outcomes.[53]^16 The present study integrated public data from 23 distinct datasets across six cancer types, including bulk RNA sequencing (RNA-seq) and single-cell RNA-seq (scRNA-seq). Data were sourced from five databases, including TCGA and Gene Expression Omnibus (GEO). Using this comprehensive dataset, we constructed and verified a ubiquitination modification network and established a prognostic model using least absolute shrinkage and selection operator (LASSO) Cox regression. We then identified the shared characteristics of homologous tumor data from different sites. By integrating pathway enrichment analysis results, we determined that ubiquitination was upregulated in SQCs and NECs, followed by oxidative phosphorylation and the MYC pathway, as observed in both bulk and single-cell data. Additionally, we validated that the prognostic model can not only predict the overall survival (OS) in surgical patients but also holds distinct value in predicting immunotherapy efficacy. Methods Data collection and construction of the ubiquitination regulatory network Considering the differences in response to immunotherapy between ADC and SQC, the lung, esophageal, and cervical cancer datasets included at least five patients with tumors classified as the classic pathological subtypes SQC or ADC within TCGA database. Given the extensive application of immunotherapy, we downloaded RNA-seq data and clinicopathological features of patients from these three cancer types, as well as from the TCGA database for urothelial carcinoma and melanoma. The correlation coefficient matrix was standardized via significance screening (p value<0.05). Further details are provided in the [54]online supplemental methods. Prognostic analysis of biomarkers using Cox regression and the Kaplan-Meier method Ubiquitination scores were analyzed for prognosis, and the downstream genes were screened using the LASSO algorithm to further identify key genes associated with prognosis. Further details are provided in the [55]online supplemental methods. Functional enrichment analysis, gene set variation analysis, and protein–protein interaction analysis For gene set enrichment analysis (GSEA), the gene subset (h.all.v7.1.symbols.gmt) was downloaded from the Molecular Signatures Database. Protein–protein interactions (PPIs) were analyzed, and enrichment analyses were performed using the search tool for the Retrieval of Interacting Genes/Proteins database. Signaling pathway variation scores for each sample were calculated using the ‘GSVA’ (gene set variation analysis) R package.[56]^17 Validation of the prognostic and biological significance of the ubiquitination pair Transcriptomic data from non-small cell lung cancer, small cell lung cancer, and non-tumoral lung tissues were obtained from the GEO dataset [57]GSE30219, whereas the transcriptomic profiles of patients with lung cancer undergoing immunotherapy were obtained from the [58]GSE135222 and [59]GSE126044 datasets. Additional transcriptome sequencing data and clinicopathological characteristics of patients with urothelial carcinoma and melanoma receiving immunotherapy were obtained from the IMvigor210 cohort and [60]GSE91061 datasets, respectively. Integrative analysis of melanoma progression was performed using the [61]GSE15605 dataset, which contains matched samples of normal skin, primary melanoma, and metastatic melanoma, to delineate the origins of histopathological heterogeneity. Transcriptomic profiles with corresponding prognostic information for esophageal SQCs and ADCs were retrieved from the [62]GSE53625 and [63]GSE183924 datasets, respectively. Validation cohorts for cervical cancer were established using [64]GSE44001 and [65]GSE56303 datasets. [66]Table 1 presents the comprehensive sample characteristics, including histopathological subtypes, treatment regimens, and clinical endpoints. Further details are provided in the [67]online supplemental methods. Table 1. Basic information regarding the sample included in bulk analysis (N (%)). Characteristics TCGA [68]GSE30219 [69]GSE135222 [70]GSE126044 IMvigor210 [71]GSE91061 [72]GSE183924 [73]GSE53625 [74]GSE44001 [75]GSE56303 N=2,778 N=307 N=27 N=16 N=262 N=57 N=37 N=358 N=300 N=29 Histology  ADC 650 (23.4) 85 (27.7) NA 7 (43.8) 0 (0) 0 (0) 37 (100) 0 (0) 64 (21.3) 4 (13.8)  SQC 840 (30.2) 100 (32.6) NA 9 (56.2) 0 (0) 0 (0) 0 (0) 179 (50.0) 221 (73.7) 25 (86.2)  Normal 125 (4.5) 14 (4.5) NA 0 (0) 0 (0) 0 (0) 0 (0) 179 (50.0) 0 (0) 0 (0)  Others 1,163 (41.9) 108 (35.2) NA 0 (0) 262 (100) 57 (100) 0 (0) 0 (0) 15 (5.0) 0 (0) OS/PFS/DFS/RFS  Alive/stable 1,670 (62.9) 72 (27.9) 6 (22.2) 5 (31.2) 87 (33.2) 11 (19.3) 21 (56.8) 73 (40.8) 262 (87.3) NA  Death/progression 983 (37.1) 186 (72.1) 21 (77.8) 11 (68.8) 175 (66.8) 46 (80.7) 16 (43.2) 106 (59.2) 38 (12.7) NA  Median (IQR) 21.7 (12.5–44.6) 48.5 (14.0–91.0) 1.97 (1.18–6.32) 2.55 (0.95–8.85) 8.0 (3.1–17.9) 18.1 (7.6–28.1) 16.8 (11.6–24.3) 34.7 (13.0–60.4) 47.8 (30.5–66.9) NA  Immunotherapy No No Anti-PD-1 Nivolumab Atezolizumab Nivolumab Durvalumab No No No [76]Open in a new tab ADC, adenocarcinoma; anti-PD-1, antibody against the programmed cell death protein-1; DFS, disease-free survival; NA, not assessed; OS, overall survival; PFS, progression-free survival; RFS, relapse-free survival; SQC, squamous carcinoma; TCGA, The Cancer Genome Atlas. ScRNA-seq data were integrated from five anatomical systems. Pulmonary datasets, including normal epithelium, ADC, SQC, and NEC subtypes, were obtained from the Single-cell Lung Cancer Atlas (LuCA) and the Human Tumor Atlas Network. For esophageal malignancies, scRNA-seq profiles of ADCs, SQCs, and adjacent normal tissues were retrieved from the [77]GSE222078, [78]GSE160269, and [79]GSE196756 datasets. Single-cell transcriptomes of bladder cancer were obtained from the [80]GSE135337 dataset, which enabled the cellular-level analysis of urothelial carcinoma heterogeneity. Finally, the scRNA-seq data for cervical SQCs and ADCs were obtained from the [81]GSE197461 dataset. [82]Table 2 presents the detailed sample information for single-cell analysis. Table 2. Basic information regarding the sample included in single-cell analysis (N (%)). Characteristics LuCA HTAN [83]GSE197461 [84]GSE222078 [85]GSE160269 [86]GSE196756 [87]GSE135337 N=298 N=49 N=8 N=10 N=64 N=6 N=8 Samples  Adenocarcinoma 156 (52.3) 24 (49.0) 5 (62.5) 10 (100) 0 (0) 0 (0) 0  Squamous carcinoma 41 (13.8) 0 (0) 3 (37.5) 0 (0) 60 (93.8) 3 (50.0) 0  Control 86 (28.9) 4 (8.2) 0 (0) 0 (0) 4 (6.2) 3 (50.0) 1 (12.5)  Others 15 (5.0) 21 (42.8) 0 (0) 0 (0) 0 (0) 0 (0) 7 (87.5) Cell types  Epithelial 169,223 (18.8) 64,301 (43.7) 5,879 (21.2) 390 (0.8) 44,730 (21.4) 4,390 (11.6) 8,267 (74.0)  Immune 670,409 (74.6) 74,806 (50.8) 16,576 (59.9) 44,592 (85.1) 111,028 (53.2) 29,077 (77.0) 446 (4.0)  Stromal 58,790 (6.6) 8,030 (5.5) 5,234 (18.9) 7,405 (14.1) 52,901 (25.4) 4,310 (11.4) 2,453 (22.0) Organ Lung Lung Cervix Esophagus Esophagus Esophagus Bladder [88]Open in a new tab HTAN, Human Tumor Atlas Network; LuCA, Single-cell Lung Cancer Atlas. Cell culture and transfection Human lung cancer cell lines (H520, HCC95, H2286, and H1792), and an ESSC cell line (KYSE30) were cryopreserved in our department. Cells were cultured in RPMI 1640 (Gibco, USA) supplemented with 10% fetal bovine serum in an incubator at 37°C with 5% CO2. RNA isolation, reverse transcription, quantitative PCR, and mRNA sequencing Total RNA was isolated from the cells using an RNA-quick purification kit (RN001, ES Science). Subsequently complementary DNA was synthesized using the Prime Script RT reagent kit following the manufacturer’s instructions (RR036A, TaKaRa). Relative messenger RNA (mRNA) expression was determined via quantitative PCR using the SYBR Green Master Mix (11202, Yeasen). Further details are provided in the [89]online supplemental methods. Western blotting and coimmunoprecipitation For western blotting (WB), cell-derived protein samples were lysed in RIPA lysis buffer. For coimmunoprecipitation (Co-IP), protein extracts were prepared using IP lysis buffer (C1054, Applygen) supplemented with protease inhibitor cocktails (#7012, Cell Signaling Technology). The extracts were then incubated with specific antibodies (E5Q6W, Cell Signaling Technology) overnight at 4°C while they were on a rotating platform. Further details are provided in the [90]online supplemental methods. Liquid chromatography–tandem mass spectrometry analysis Proteomic data for H2286 cells were generated by PTM BIO (Hangzhou, China). Cells were washed three times with prechilled phosphate-buffered saline (PBS), collected in centrifuge tubes, and centrifuged at 400× g and 4°C for 5 min. After removing the supernatant, qualitative proteomic analyses were performed, including protein extraction, enzymatic digestion, liquid chromatography-mass spectrometry, tandem analysis, and bioinformatics analysis. Measurement of reactive oxygen species To determine the intracellular levels of reactive oxygen species (ROS), 1×10⁶ cells were harvested and then incubated with 10 µm DCFH-DA (Beyotime, Shanghai, China) at 37°C for 20 min. The cells were then washed three times with PBS and immediately analyzed via immunofluorescence. Mouse models Five-week-old female BALB/c-nu mice were used to establish a subcutaneous xenograft model. H2286 and H520 cells transfected with sh-OTUB1 or control vectors, as well as with OTUB1 or the negative control, were resuspended in PBS. The transfected H2286 cells were subcutaneously inoculated at a density of 5×10^6 cells per mouse in 100 µL of PBS, with four mice in each group. Tumors were measured every 3 days starting from day 10. Tumor volume (mm^3) was determined as follows: volume = (length × width^2)/2. This study was approved by the Institutional Animal Care and Use Committee of the Peking Union Medical College, Chinese Academy of Medical Sciences. Principal component analysis, uniform manifold approximation and projection, and statistical analysis The R software (v4.2.1) was used for all statistical analyses. Principal component analysis (PCA) was performed using the ‘factoextra’ and ‘FactoMineR’ R packages. The uniform manifold approximation and projection (UMAP) dimensionality reduction analysis of single-cell data was performed using the ‘Seurat’ R package. The results of dimensionality reduction and clustering were statistically analyzed using the ‘vegan’ package. Analysis of similarities was used to compare gene expression and cell ratios among the clusters. For the survival analysis, the p value was calculated using the log-rank test. Statistical significance was set at p<0.05. Results Identification of the key nodes of the ubiquitination-modified network and significant prognostic pathways Complete clinical and transcriptional information of 1,294 patients with lung cancer was obtained from TCGA lung cancer dataset and [91]GSE30219. In addition, 693 patients with cervical cancer were included from TCGA cervical cancer, [92]GSE44001, and [93]GSE56303 datasets, which were used to infer the histological characteristics of some patients. Overall, we identified 395 patients with esophageal ADCs and SQCs in TCGA esophageal cancer, [94]GSE53625, and [95]GSE183924 datasets. [96]Online supplemental figure S1 presents the detailed workflow of the study. We identified five key deubiquitinating enzymes with conserved functions and significant regulatory relationships across lung, esophageal, and cervical cancers, along with their associated ubiquitin-linked modules, following a step-by-step correlation analysis ([97]figure 1A). Certain deubiquitinating enzymes and ubiquitin ligases may contribute to stable expression of genes, whereas other combinations exert polarized amplification effects on the activation of biological pathways. Gene Ontology enrichment analysis indicated that OTUB1 and UCHL3 negatively modulated protein ubiquitination ([98]figure 1B). In contrast, USP12, USP34, and MYSM1 abnormally activated the ubiquitin ligase pathways ([99]figure 1C). We successfully established a ubiquitination regulatory network, and the detailed schematic is presented in [100]figure 1D. Cox regression analysis revealed a significant association between the paired scores of OTUB1, and UCHL3 and poor prognosis, indicating that these genes may activate crucial oncogenic pathways ([101]figure 1E). Consequently, we employed the LASSO algorithm to identify the key downstream genes within the ubiquitination network in TCGA lung cancer dataset, which included the largest number of patients ([102]figure 1E,F). PPI network analysis was used to examine the interactions among OTUB1, MYC, and TRIM28 ([103]figure 1G). GSEA further revealed that OTUB1 and TRIM28 influence gene ubiquitination and significantly activate the MYC pathway ([104]figure 1H). [105]Figure 1I presents the relationships between key nodes in the ubiquitination network, which are potential target genes, and biological pathways, including downstream oncogene activation and oxidative phosphorylation. Figure 1. Identification of the key nodes of the ubiquitination-modified network and significant prognostic pathways: The heatmap displays the potential correlation between deubiquitinating enzymes (DUBs) and ubiquitin-conjugating enzymes (E2) or ubiquitin ligases (E3) (A). The heatmap shows the impact of the OTUB1/UCHL3 and E2/3 pairs on the negative regulation of ubiquitination (B). The heatmap reveals the impact of DUB-E2/3 pairs on the ubiquitin ligase complex (C). The screening process of genes used to establish ubiquitination regulatory network is described (D). Cox regression analysis indicates that the OTUB1-TRIM28-UBE family score is related to poor prognosis (E). There is a relationship between MYC and the key nodes of ubiquitination modification networks (F). The dot plot of the hallmark pathway enrichment results shows the key pathways downstream of the ubiquitination network (G). The key molecules in the downstream pathway are shown in the circle plot (H). [106]Figure 1 [107]Open in a new tab Establishment and validation of prognostic models A Cox proportional hazards regression model was performed on the gene sets identified using LASSO in patients with lung cancer in TCGA to establish a ubiquitination-related prognostic signature (URPS). The areas under the time-dependent receiver operating characteristic curves (AUROCs) were calculated to assess the prognostic performance. In the training cohort, AUROC values for URPS at 1, 3, and 5 years were 0.644, 0.648, and 0.666, respectively ([108]figure 2A). Kaplan-Meier (K-M) survival analyses revealed significantly greater OS in the low-risk group than in the high-risk group across both training types ([109]figure 2B). [110]Figure 2C shows the distribution of patient survival and risk scores calculated using the URPS. [111]Figure 2D presents a heatmap illustrating the clinical outcomes, histological features, and intricate expression patterns of URPS. The UPRS was subsequently validated in independent patient cohorts with the same cancer. The model demonstrated a better predictive performance using the [112]GSE30219 dataset ([113]figure 2E). The results of the K-M survival analysis ([114]figure 2F) and patient distribution ([115]figure 2G) were consistent with those of the training set. Analysis of gene expression patterns revealed upregulation of USP12 and its associated genes in patients with good prognosis, whereas other URPS genes were significantly downregulated ([116]figure 2H). Considering its wider application, we included patients with lung cancer who received immunotherapy to explore the value of the URPS. Given the limited OS observed in patients receiving immunotherapy, we explored the applicability of URPS to progression-free survival, discovering its promising efficacy ([117]figure 2I). The survival analysis indicated that high-risk patients often experienced rapid disease progression during immunotherapy ([118]figure 2J). The original data and analysis scripts ([119]online supplemental file 1) for the lung cancer immunotherapy cohort are provided in the [120]online supplemental materials. The patient distribution ([121]figure 2K) and gene expression profiles ([122]figure 2L) showed consistent URPS patterns between patients treated with immunotherapy and those treated with surgery. Figure 2. Identification and establishment of a ubiquitination-related prognostic signature (URPS): The AUC values of the ROC curves predict the 1-year, 3-year, and 5-year overall survival (OS) of the signature in the TCGA-LUNG (A) and [123]GSE30219 (E) cohorts. Kaplan-Meier (K-M) survival curves for OS defined by the URPS are presented for high-risk and low-risk patients in TCGA-LUNG (B) and [124]GSE30219 (F) cohorts. Distributions of the risk score and survival status of patients with lung cancer in the TCGA-LUNG (C) and [125]GSE30219 (G) cohorts. The expression patterns of URPS and the clinical features of patients from the TCGA-LUNG (D) and [126]GSE30219 (H) cohorts are shown. The URPS predicts the AUC value of the ROC curve for 1-year progression-free survival (PFS) in patients with lung cancer receiving immunotherapy from the [127]GSE135222 and [128]GSE126044 datasets (I). K-M survival curves show the survival differences between high-risk and low-risk patients as defined by the URPS (J). The distribution of risk scores and survival status in patients with lung cancer receiving immunotherapy provides insights into the prognostic value of the URPS (K). The expression patterns of URPS and the patient cohort are displayed (L). ADC, adenocarcinoma; AUC, area under the curve; AUROC, areas under the receiver operating characteristic curve; FP, false positive; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; NE, neuroendocrine; ROC, receiver operating characteristic; SQC, squamous cell carcinoma; TCGA, The Cancer Genome Atlas; TP, ture positive. [129]Figure 2 [130]Open in a new tab Subsequently, we validated the URPS in four additional cancer types from the same database used to construct the ubiquitin regulatory networks. The URPS exhibited prognostic predictive power in patients with cervical cancer comparable to that in the training cohort ([131]online supplemental figure S2A), effectively identifying patients with poor prognosis. Furthermore, its expression pattern in patients with cervical cancer mirrored that observed in patients with lung cancer. Validation in an independent cohort of patients with cervical cancer who underwent radical resection further demonstrated the superior predictive performance of the URPS ([132]online supplemental figure S2B). To further confirm the robustness of the URPS, we analyzed additional external cohorts of patients with esophageal cancer. The URPS exhibited a strong discriminative ability in patients with esophageal cancer, exhibiting enhanced prognostic accuracy for all esophageal cancer cases compared with esophageal SQCs alone ([133]online supplemental figure S2C). Furthermore, it demonstrated excellent predictive power for long-term survival in patients with SQC ([134]online supplemental figure S2D). The URPS was also validated in patients with localized esophageal cancer receiving adjuvant immunotherapy ([135]online supplemental figure S2E). The model demonstrated outstanding predictive performance, with URPS-identified high-risk patients showing a significantly worse prognosis and frequent relapse within 2 years. Notably, URPS gene expression patterns differed between patients with esophageal ADC and SQC. Considering its potential relevance to immunotherapy, we evaluated the URPS in the IMvigor210 cohort. URPS effectively distinguished the immunotherapy-sensitive subgroups in this cohort ([136]online supplemental figure S2F). The URPS demonstrated exceptional discriminative performance in the surgical cohorts of patients with bladder cancer ([137]online supplemental figure F2G) and kidney papillary cell carcinoma ([138]online supplemental figure F2H), reliably identifying patients with unfavorable outcomes. Furthermore, in the homologous TCGA melanoma cohort, the URPS remained significantly associated with poor prognosis ([139]online supplemental figure F2I). Additionally, the URPS demonstrated comparable efficacy among patients with melanoma undergoing immunotherapy ([140]online supplemental figureF2J). URPS and prognosis based on histological differentiation The URPS was significantly associated with tumor histological outcomes, as illustrated in the heatmaps in [141]figure 2, [142]online supplemental figure S2. To explore the disparities in the ubiquitination patterns between tumors and adjacent tissues, as well as among tumor subtypes, PCA was first performed on 1,563 samples, including SQCs, ADCs, and normal tissues from TCGA, to reduce dimensionality. Significant heterogeneity was observed not only across tumors from different organs ([143]figure 3A), but also between tumors of different cellular origins ([144]figure 3B). When GSVA was used to quantify the ubiquitination-mediated biological functions, SQCs exhibited relatively high ubiquitination scores ([145]figure 3C), whereas normal tissues showed the lowest level of ubiquitination ([146]figure 3D). A similar trend was observed for oxidative phosphorylation ([147]figure 3E) and the MYC pathways ([148]figure 3F). Figure 3. Varied ubiquitination levels in different types of lung cancer affecting the MYC pathway: PCA for dimensionality reduction revealed that tumor samples from different sites (A) and different cell origins (B) were clustered in accordance with their species. The distribution of scores within the ubiquitin-mediated proteolysis pathway is presented within the samples (C). Boxplots display the differences in ubiquitination (D), oxidative phosphorylation (E), and MYC pathway (F) scores among adenocarcinoma, squamous cell carcinoma, and normal tissues. PCA dimensionality reduction in the [149]GSE30219 dataset shows samples of different cell origins (G). Boxplots present the variances in the scores of the MYC (left panel) and ubiquitination-mediated proteolysis (right panel) pathways among the tissues originating from different cell types (H). ADC, adenocarcinoma; CARCI, carcinoid; KEGG, Kyoto Encyclopedia of Genes and Genomes; NEC, neuroendocrine carcinoma; PCA, principal component analysis; SQC, squamous cell carcinoma. [150]Figure 3 [151]Open in a new tab We hypothesized that the ubiquitination modification network influences oxidative phosphorylation through the MYC pathway, thereby altering patient prognosis and immunotherapy response. Accordingly, this was validated using the exogenous lung cancer sequencing dataset [152]GSE30219, which includes a broader range of pathological types, including NECs such as small cell lung cancer. Compared with carcinoid tumors, small-cell and large-cell NECs exhibited higher levels of ubiquitination modifications ([153]figure 3G), indicating a close association with the degree of tumor malignancy rather than with histological features only. Notably, the differences in MYC pathway and ubiquitination scores were consistent or even more pronounced ([154]figure 3H). However, analyzing the biological functions in bulk data is insufficient to elucidate the mechanism of action at the cellular level. Subsequently, we analyzed the relationship between ubiquitination and tumor biology using single-cell data. We employed UMAP for dimensionality reduction. No significant batch effects were observed between the data, and cells with the same histological origin exhibited consistent profiles ([155]figure 4A). Ubiquitin-mediated proteolysis ([156]figure 4B) and MYC target ([157]figure 4C) pathway scores were significantly higher in neuroendocrine tumor cells. Considering the potential batch effects, this was further validated in the LuCA cohort ([158]figure 4D). In this cohort, SQCs exhibited higher levels of ubiquitination modifications ([159]figure 4E) and oxidative phosphorylation ([160]figure 4F) than ADC. When the ubiquitination modification network proposed in [161]figure 1A was applied at the single-cell level, the results remained consistent. Broadly similar nodes and inter-relationships were observed in the regulatory network ([162]figure 4G). Single-cell analysis revealed significantly higher URPS distribution in SQCs and relatively higher in NECs ([163]figure 4H). Correlation analysis indicated NME1 and TRIM28 as the genes most closely associated with the MYC and oxidative phosphorylation pathways ([164]figure 4I). Pancancer validation of URPS ubiquitination networks was conducted across single-cell datasets of esophageal, cervical, and bladder cancers, as well as the bulk datasets mentioned in [165]online supplemental figure S1. Consistent trends were observed in the integrated single-cell datasets of esophageal ADCs and SQCs ([166]online supplemental figure S3A,B). Both MYC pathway activity and URPS levels were higher in tumor tissues than in the normal tissues, with SQCs exhibiting higher levels than ADCs. Correlation analyses revealed strong associations between the NME1/OTUB1 and MYC pathways, whereas USP12, PTPN21, MAN1C1, and MBNL2 formed distinct interaction modules, consistent with bulk data findings in lung, esophageal, and cervical cancers ([167]online supplemental figure S3C,D). Data from TCGA and patients with immunotherapy-treated urothelial carcinoma revealed the conserved role of the TRIM28/OTUB1-NME1 axis in single-cell analyses ([168]online supplemental figure S3E–G). Finally, in melanoma progression models, both MYC activity and ubiquitination modification scores were correlated with tumorigenesis and metastasis, with NME1 playing a central functional role across cancer types ([169]online supplemental figure S3H). Figure 4. Similar ubiquitination regulation in different cell types at single-cell resolution: Potential differences between different cell types (left panel) rather than cohorts (right panel) shown by UMAP dimensionality reduction of lung cancer single-cell data (A). Boxplots display the differences in ubiquitination-mediated proteolysis (B) and MYC (C) pathway scores among different cell types in the combined data. The UMAP dimensionality reduction shows the distribution of specific cell types corresponding to cell classes (D). Boxplots illustrate the differences in ubiquitination-mediated proteolysis (E) and oxidative phosphorylation (F) pathway scores among different cell types in the LuCA cohort. The heatmap exhibits the potential correlation between DUB and E2 or E3 at the single-cell level (G). A boxplot is used to illustrate the differences in URPS among diverse cell types within the consolidated dataset (H). The correlation map displays the connections between genes associated with URPS and biological pathways (I). DUB, deubiquitinating enzyme; E2, ubiquitin-conjugating enzymes; E3, ubiquitin ligases; EMT, epithelial-mesenchymal transition; LUAD, lung adenocarcinoma; LuCA, Single-cell Lung Cancer Atlas; LUSC, lung squamous cell carcinoma; NE, neuroendocrine; NSCLC, non-small cell lung cancer; UMAP, uniform manifold approximation and projection; URPS, ubiquitination-related prognostic signature. **** represents P < 0.0001. [170]Figure 4 [171]Open in a new tab Impact of URPS on the TME In prior analyses, URPS was associated with tumor tissue types at both the bulk and single-cell levels. It also demonstrated prognostic value in patients with lung or esophageal cancers treated using immune checkpoint inhibitors (ICIs). We explored the relationships between URPS and non-tumor cells across diverse cancer types. Significant disparities in the cell composition patterns across tumor types were observed in the TCGA cohort ([172]figure 5A). A relatively elevated number of macrophages was observed in the SQCs ([173]figure 5B). Conversely, a disproportionately greater proportion of memory B cells was present in ADCs ([174]figure 5C). In addition to the ‘CIBERSORT’ (Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts) algorithm, we applied the TIMER method to estimate B-cell proportion, revealing a significant correlation with URPS ([175]figure 5D). Correlation analysis demonstrated that the MAN1C1 gene in URPS was considerably associated with B-cell markers ([176]figure 5E). Furthermore, we validated the relationships among URPS, tumor tissue classification, and immune cell ratio in the [177]GSE30219 cohort ([178]figure 5F). Compared with SQCs and ADCs, NECs exhibited a higher proportion of memory B cells ([179]figure 5G). Consistent with the findings of TCGA pancancer analysis, MAN1C1 and USP12 may be key contributors to this disparity ([180]figure 5H). However, in lung cancer alone, an inverse positive correlation was observed between URPS and memory B cell numbers ([181]figure 5I). This may be attributed to the different cancer types or the additional inclusion of NECs. In patients with lung cancer treated using ICIs, URPS levels significantly differed between responders and non-responders ([182]online supplemental figure S4A), potentially due to the reduced infiltration of macrophages ([183]online supplemental figure S4B,E) and CD8^+ T cells ([184]online supplemental figure S4C,F) in the TME, leading to suboptimal therapeutic efficacy. Correlation analyses revealed a potential regulatory mechanism analogous to USP12-mediated modulation ([185]online supplemental figure S4D,I,L,O). In patients with esophageal ADC receiving durvalumab, elevated URPS scores were observed in relapsed cases ([186]online supplemental figure S4G), accompanied by reduced M2 macrophage abundance ([187]online supplemental figure S4H). Conversely, in urothelial carcinoma immunotherapy cohorts, URPS levels ([188]online supplemental figure S4J) and M0 macrophage proportions ([189]online supplemental figure S4K) were significantly higher in refractory populations, with URPS potentially driving M0/M2 polarization. The melanoma immunotherapy dataset mirrored the patterns observed in esophageal cancer, where URPS was inversely correlated with treatment response ([190]online supplemental figure S4M) and macrophage infiltration ([191]online supplemental figure S4N). Figure 5. Potential associations between histological type, TME, and URPS: PCA dimension reduction using the proportion of immune cells reveals disparities in tumors of diverse histological classifications (A). The boxplot distinctly presents the variances in the proportions of macrophages (B) and memory B cells (C) among different tumors respectively. The scatter plot illustrates that the URPS is negatively correlated with the proportion of B cells, as calculated by TIMER (D). The correlation map of URPS, T cell, macrophage, B-cell proportions and B-cell marker genes uncovers their intimate connections (E). PCA exhibits immune cell composition variances in the [192]GSE30219 dataset (F). The boxplot delineates the discrepancies in memory B cells among ADCs, SQCs and NECs (G). The correlation atlas reveals URPS, the T/M/B-cell proportion and B-cell marker gene associations (H). The scatter plot illustrates a positive correlation between the proportion of memory B cells and URPS (I). ADC, adenocarcinoma; CARCI, carcinoid; CEAD, cervical adenocarcinoma; CESC, cervical squamous cell carcinoma; ESAD, esophageal adenocarcinoma; ESSC, esophageal squamous cell carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; NE, neuroendocrine; NEC, neuroendocrine carcinoma; PCA, principal component analysis; SQC, squamous cell carcinoma; TME, tumor microenvironment; URPS, ubiquitination-related prognostic signature. [193]Figure 5 [194]Open in a new tab Validation in cell lines and patient cohorts We performed Co-IP experiments using total cells extracted from the small-cell lung cancer cell line NCI-H2286, which highly expresses MYC, to validate the ubiquitination modification network and identify MYC-associated ubiquitination regulatory enzymes. A total of 1,157 potential interactors were identified in the Co-IP bound by MYC ([195]figure 6A). Notably, 45.6% (528/1,157) and 50.4% (583/1,157) of the MYC-interacting groups overlapped with those reported by the Qing laboratory[196]^18 and Beli laboratory,[197]^19 respectively. Among them, two URPS proteins encoded by genes TRIM28 and NME1, were found to bind directly to MYC in all three datasets and were associated with the MYC pathway ([198]figure 6B). Figure 6. Validation of the effect of OTUB1-TRIM28 on the MYC pathway for poor prognosis: Venn diagram shows the overlapping MYC-interacting partners among the current study and data from the laboratories of Beli and Qing (A). The circle plot presents the key binding proteins that influence downstream pathways (B). The scatter plot demonstrates a positive correlation between MYC and the ubiquitination pathway, particularly in the NCI-H2286 cell line (C). Co-Immunoprecipitation (Co-IP) was employed to detect the interactions between endogenous MYC and proteins such as OTUB1 and TRIM28 (D). Knockdown of OTUB1 has a significant effect on the MYC pathway as well as the ubiquitination network (E). After knockdown (left panel) and overexpression (right panel) of OTUB1, a CCK-8 assay was used to measure the proliferation of NCI-H2286 cells (F). Overexpression of OTUB1 promotes tumor colony formation (G). OTUB1 knockdown reduces MYC ubiquitination and binding protein levels in Co-IP (H). A xenograft tumor growth assay was conducted using NCI-H2286 cells after the knockdown of OTUB1 (I). ROS levels in NCI-H2286 cells overexpressing OTUB1 and TRIM28; light curve: negative control (J). In the CHCAMS esophageal cancer cohort, TRIM28 is linked to poor prognosis (right panel). Representative IHC staining images (left panel) show low (upper slide) and high (lower slide) expression of TRIM28 (K). In the CHCAMS lung cancer cohort, TRIM28 is correlated with shorter survival times (right panel). Representative images (left panel) show low (upper slide) and high (lower slide) expression of TRIM28 (L). FITC, Fluorescein 5-isothiocyanate; IHC, immunohistochemistry; NC, negative control; ns, not significant; ROS, reactive oxygen species. [199]Figure 6 [200]Open in a new tab OTUB1 and UCHL3 were also predicted to bind to MYC. Analysis of the ssGSEA data in the Cancer Cell Line Encyclopedia revealed that the ubiquitination-mediated proteolysis pathway was positively correlated with that of the MYC pathway, confirming a strong interaction between ubiquitination and the MYC pathway ([201]figure 6C). Notably, the H2286 small-cell lung cancer cell line exhibited the highest activities of both pathways. Therefore, we validated the interaction between MYC and components of the URPS in this cell line. In addition to TRIM28, which was previously inferred to bind MYC, Co-IP experiments revealed MYC interactions with OTUB1, UBE2T, and CHEK2 ([202]figure 6D). Subsequently, we verified the potential relevance of OTUB1 by knocking down OTUB1 and performing RNA-seq. The results revealed downregulation of NME1 expression ([203]figure 6E). WB experiment also revealed reduced MYC protein levels, indicating that OTUB1 may affect NME1 transcription through the MYC pathway. In the CCK-8 assay, OTUB1 knockdown inhibited H2286 cell proliferation, whereas its overexpression promoted proliferation ([204]figure 6F). Plate cloning experiments confirmed that OTUB1 overexpression promoted tumor colony formation ([205]figure 6G). Furthermore, Co-IP experiments demonstrated that OTUB1 knockdown reduced both MYC ubiquitination and the level of its interacting protein when normalized to consistent input samples ([206]figure 6H). H2286 cells overexpressing OTUB1 and TRIM28 exhibited significantly elevated levels of ROS and oxidative phosphorylation ([207]figure 6I). Subsequently, the oncogenic potential was validated in OTUB1/TRIM28-overexpressing esophageal carcinoma (KYSE30, ([208]online supplemental figure S5A), lung SQC (HCC95, [209]online supplemental figure S5B; H520, [210]online supplemental figure S5C), and lung ADC (H1792, ([211]online supplemental figure S5D) cell lines. In vivo xenograft models confirmed that OTUB1 knockdown suppressed lung cancer growth ([212]figure 6J), whereas its overexpression promoted tumorigenesis ([213]online supplemental figure S5E,F). In esophageal ([214]figure 6K), lung ([215]figure 6L), and bladder ([216]online supplemental figure S5G) cancer cohorts from CHCAMS, we validated the prognostic performance of TRIM28. Notably, high TRIM28 expression was associated with poor prognosis. Finally, RNA-seq of lung cancer cells overexpressing OTUB1 and TRIM28 revealed significant upregulation of the MYC and oxidative phosphorylation pathways ([217]online supplemental figure S5H). These regulatory relationships ([218]online supplemental figure S5I) were consistent with those predicted by the previously established ubiquitination regulatory network. Discussion This study involved a comprehensive exploration of ubiquitination in a pancancer setting and has several significant implications. Identification of key nodes and prognostic pathways within the ubiquitination modification network offers valuable insights into the molecular mechanisms underlying cancer progression. The establishment and validation of URPS in patients with different cancer types who underwent surgery and/or received immunotherapy highlight its potential as a tool for personalized medicine. Demonstrating the biological significance of URPS in both bulk and single-cell data highlights that, despite the complexity and heterogeneity of cancer, certain commonalities and stability occur across different levels. The OTUB1-TRIM28 axis is crucial for MYC ubiquitination and influences patient outcomes. Notably, although URPS-mediated regulation of the MYC pathway is a key determinant, it is not the sole factor influencing prognosis. Metabolic pathways, such as oxidative phosphorylation, also affect patient outcomes. Previous studies have identified oxidative phosphorylation as the primary source of endogenous ROS, which affects the prognosis by interfering with cellular autophagy.[219]^20 These findings indicate that it may serve as a specific target for therapeutic interventions. Future studies should explore the detailed mechanisms by which this ubiquitination process regulates the MYC pathway and investigate potential drug targets. Previous research has focused on the relationships between ubiquitinated substrates and enzymes. For example, the UbiBrowser database calculates potential correlations using deep learning algorithms based on a significant amount of known information regarding protein interactions, experimental data related to ubiquitination, and multiple protein biological signatures.[220]^21 The Integrated Annotations for Ubiquitin and Ubiquitin-Like Conjugation Database (IUUCD) reflects the degree of PPIs by integrating data such as mRNA expression profiles, PTMs, and mutant polymorphisms, along with information such as experimentally determined intensity or affinity of interactions.[221]^22 These resources are crucial in advancing our understanding of the complexity of the ubiquitination regulatory networks. However, our results are more reliable than those obtained from the UbiBrowser database. Our analysis builds on the foundational data provided by these databases and further strengthens the findings through experimental validation. Compared with the IUUCD, we quantitatively assessed the relationships between nodes in the ubiquitination regulatory network using absolute measures and elucidated the underlying downstream biological pathways, thereby enhancing the dimensionality of knowledge. A key strength of our study is the integration of multiple data sources and experimental approaches, which provides a robust framework for future cancer research. Nevertheless, this study has some limitations. First, this study primarily focused on a limited number of cancer types, and further validation across diverse cancer types is warranted. In addition, the biological functions of other components of the ubiquitination regulatory network must be further elucidated. In conclusion, this study identified critical hub nodes within the ubiquitination regulatory network and established a predictive model with strong prognostic value for patients undergoing surgery and immunotherapy. Supplementary material online supplemental file 1 [222]jitc-13-8-s001.R^ (750B, R) DOI: 10.1136/jitc-2025-012539 online supplemental file 2 [223]jitc-13-8-s002.xlsx^ (18KB, xlsx) DOI: 10.1136/jitc-2025-012539 online supplemental file 3 [224]jitc-13-8-s003.docx^ (4MB, docx) DOI: 10.1136/jitc-2025-012539 online supplemental file 4 [225]jitc-13-8-s004.docx^ (18.1KB, docx) DOI: 10.1136/jitc-2025-012539 Acknowledgements