SNHG7 interacts with PCBP2 to promote CDKN2A expression and modulate cuproptosis in colorectal cancer.

Introduction
Colorectal cancer (CRC), a major cause of mortality, is the third most prevalent cancer and the second most frequent cause of cancer deaths worldwide[1,2]. Approximately 20 % of CRC patients were diagnosed in the advanced stage. Worth still, the prognosis of advanced CRC remains a challenge, with a 5-year survival rate of less than 20 %. In recent years, pathological and molecular tumor testing have gained the potential to improve prognosis and select appropriate therapy[3].
The major forms of regulated tumor cell death extensively studied to date include apoptosis, pyroptosis, ferroptosis, and necroptosis[4,5]. Recently, emerging evidence suggests that the accumulation of copper ionophores can induce cell death by binding multiple, structurally distinct small molecules in hundreds of cell lines. Copper ionophore directly binds to the fatty acylated components in the tricarboxylic acid (TCA) cycle, resulting in abnormal aggregation of lipoylated proteins and loss of Fe-S cluster-containing proteins, which leads to acute proteotoxic stress and eventually cell death[6,7]. The cyclin dependent kinase inhibitor 2A (CDKN2A) gene encodes the tumor suppressor proteins p16INK4a and p14ARF, which are critical regulators of cell cycle arrest and senescence[8,9]. Paradoxically, elevated CDKN2A expression is associated with poor prognosis across multiple cancer types, often serving as a biomarker for aggressive tumor behavior and resistance to certain therapies[10]​. In CRC, CDKN2A is often upregulated, correlating with advanced tumor stages, enhanced immune infiltration, and poor patient outcomes, making it a valuable prognostic biomarker[9,11]​. Furthermore, the CDKN2A gene has been implicated in cuproptosis by influencing mitochondrial respiration and protein lipoylation pathways[12]. CDKN2A promotes resistance to cuproptosis by modulating glycolysis and copper metabolism[13]. Nevertheless, the molecular mechanisms underlying CDKN2A dysregulation in CRC remain largely elusive.
Long non-coding RNAs (lncRNAs) are non-coding transcripts longer than 200 nucleotides in length that lack a significant protein-coding capacity[14,15]. LncRNAs are important versatile molecules engaged in a variety of tumorigenic processes by interacting with DNA, RNA, or proteins[16]. LncRNAs act as archetypes of decoys, signals, guides, and scaffolds to execute molecular functions[17]. The small nucleolar RNA host gene 7 (SNHG7) functions as an oncogenic lncRNA across various cancers, including CRC, by promoting cell proliferation, migration, and invasion while inhibiting apoptosis[18]. SNHG7 is significantly overexpressed in CRC and acts as an oncogenic lncRNA, promoting tumor progression through multiple mechanisms[19]​. It promotes CRC progression and anlotinib resistance by sponging miR-181a-5p to upregulate GATA6, contributing to poor patient survival[20]. Additionally, SNHG7 is implicated in cuproptosis resistance by modulating glycolysis and copper homeostasis via the SNHG7/miR-133b/CDKN2A axis, supporting a malignant phenotype and tumor microenvironment[13]. As a key component of the competing endogenous RNA network, SNHG7 is significantly upregulated in CRC tissues and linked to enhanced proliferation, apoptosis resistance, and metastasis[19]. It is further identified as a crucial disulfidptosis-related lncRNA within novel prognostic models, where its aberrant expression correlates with immune escape, poor responsiveness to immunotherapy, and high chemosensitivity[21,22]. Collectively, these findings highlight SNHG7 as a promising biomarker and therapeutic target in CRC management.
Poly(rC)-binding protein 2 (PCBP2), a member of the RNA-binding poly(C)-binding protein family, regulates tumor progression by modulating transcriptional and post-transcriptional processes, including pre-mRNA splicing, mRNA stabilization, and translational control[23]. PCBP2 plays a crucial role in CRC, functioning as a key regulator in tumor progression, therapy response, and prognosis. It enhances the mRNA stability of UKLF by binding to its 3′-UTR, promoting tumor growth and progression via the UKLF/SLC39A4 pathway, thereby presenting a potential target for molecular therapy in CRC[24]. Additionally, PCBP2 emerges as a predictive biomarker for favorable outcomes in locally advanced rectal cancer patients undergoing neoadjuvant chemoradiotherapy, indicating its utility in personalized treatment strategies[25]. Furthermore, PCBP2 is implicated in stress granule-associated oncogenic pathways, including TNF-α, PI3K-AKT-mTOR, and WNT-β-catenin signaling, highlighting its involvement in cellular stress responses and its potential as a prognostic marker across cancers[26].
The long noncoding RNA SNHG7 is closely linked to the CDK family, modulating cell cycle progression and proliferation by regulating the expression of key CDK-related genes[27], especially CDKN2A13. LncRNA associates with PCBP2 to enhance its stability and PCBP2 in turn binds to CDKN2A mRNA influencing its regulatory processes[28]. In this study, we identified a novel lncRNA, SNHG7, which is highly expressed in colorectal cancer. SNHG7 directly interacts with PCBP2 to enhance CDKN2A expression while activating the cuproptosis pathway and promoting glycolysis.
Cuproptosis is a recently characterized form of regulated cell death. Analysis of CRC databases identified the long non-coding RNA SNHG7 as an independent poor prognostic factor in CRC. Subsequent bioinformatic analyses suggested that SNHG7 may inhibit cuproptosis by regulating the cuproptosis suppressor CDKN2A, thereby reducing cell death in colorectal cancer cells. To experimentally validate these observations, we conducted both in vitro and in vivo studies. First, the predicted interaction between SNHG7 and PCBP2 was confirmed by RNA immunoprecipitation (RIP) assays. Consistent with this finding, database analyses revealed positive correlations between SNHG7 and PCBP2 expression, as well as between PCBP2 and CDKN2A. In addition, we confirmed that the cuproptosis inducer elesclomol effectively triggered cuproptosis in cultured cells. For in vitro functional assays, we selected colorectal cancer cell lines based on their endogenous SNHG7 levels: SW480 and RKO cells (with relatively high expression) were used for siRNA-mediated knockdown experiments, whereas HCT116 cells and a separate set of RKO cells were used for SNHG7 overexpression experiments. Western blotting and quantitative PCR analyses demonstrated that SNHG7 overexpression or silencing led to corresponding upregulation or downregulation of its interacting partner PCBP2 and the downstream effector CDKN2A. In vivo, HCT116 cells stably overexpressing SNHG7 were subcutaneously injected into nude mice to generate xenograft tumor models. Four experimental groups were established: vector control, SNHG7 overexpression, vector plus elesclomol, and SNHG7 overexpression plus elesclomol. Tumor growth monitoring revealed that SNHG7 overexpression significantly promoted tumor progression, an effect that was markedly reversed by elesclomol treatment. Final tumor volume measurements, together with qPCR, Western blotting, and histopathological analyses, suggested that SNHG7 reduces the cellular response to ESCu through the PCBP2/CDKN2A signaling axis, consistent with a cuproptosis-related phenotype. Together, our findings suggest that SNHG7 may serve as an important regulator linking glycolytic metabolism with cuproptosis-related pathways, highlighting its potential as a therapeutic target in CRC.

Materials and methods

Data acquisition and quality control
COAD, READ, STAD, LUAD and KIRC patients' data were downloaded from the portal.gdc.cancer.gov. The patients without survival information or the survival time of less than months were excluded. If the patients have more than one sample, the first sample was included. Finally, 461 CRC patients were included in our study. Finally, the 461 patients were divided into training group (n = 231) and test group (n=230) randomly by the “caret” package. The training group was used to build the lasso-cox risk score model, then the testing group, entire CRC group and cohorts of STAD, LUAD and KIRC were used to verify the effectiveness and the rationality of this risk score model.
The TCGA cohort (n = 461) was divided into training and testing sets by stratified random sampling on the event indicator using caret: createDataPartition with a fixed random seed (set.seed (123456787)) at a 1:1 ratio (50 %/50 %). LASSO Cox regression was performed in the training set using glmnet, with the penalty parameter λ selected by k-fold cross-validation (cv.glmnet, 10-fold by default). The final signature was constructed from features with non-zero coefficients at λ_min. Hazard ratios (HRs) were estimated using Cox proportional hazards models and are reported with 95 % confidence intervals.

Screening of the cuproptosis-related lncRNA
Signature of Cuproptosis-related mRNAs was obtained from the article: “Copper induces cell death by targeting lipoylated TCA cycle proteins[7]”, and the ten marker genes were included in the Fig3A of this article. Then, some genes were eliminated from the TCGA expression data if the expression of the gene was zero in over 25 percent of total samples. The “edgeR” package was used to do the differential gene expression analysis. Then the differential lncRNA was used to do the Pearson’s correlation analysis with the Cuproptosis-related mRNAs, and the LncRNAs with P value<0.001 and R2 >0.4 were included in the Cuproptosis-related lncRNAs.

Establishment of cuproptosis-related lncRNA prognostic model
The training group was used to build the prognostic model using the lasso-cox regression by the “glmnet” package. Then the best value of lambda was chosen to establish the model when the mean square error of the model is the minimum. And the risk score was calculated by the “predict” function. Moreover, the multivariate Cox model was established to assess the hazard ratios (HR) of each resulting lncRNAs by the “survival” package. Subsequently, the best cut-off point was calculated by the “surv_cutpoint” function in the “survminer” package. So the patients were divided into "high risk" or "low risk" groups according to their risk scores. And the Kaplan Meier (KM) survival and receiver operating characteristic (ROC) curve analysis were performed by the “survival” package using the overall survival (OS) time. The testing group and entire group were performed the same analysis as the training group to evaluate this prognostic model.

Clinical information and single gene prognostic analysis of the cuproptosis-related lncRNA signature
The single gene KM survival analysis was performed with the resulting CRLs. And the patients were divided into subgroups in terms of age (>65 and ≤65 years), gender (male and female), American Joint Committee on Cancer (AJCC) stage (III, IV), T stage (T3, T4), N stage (N0, N1, N2), M stage (M0, M1) and the microsatellite stability (MSS). Then the univariate and multivariate cox analyses were performed. Survival analysis and ROC curves, the area under the curve (AUC), and decision curve analysis (DCA) were used to assess the reliability of this model and validate whether these factors were independent prognostic factors.

Establishment of prognostic nomogram model
The nomogram was used to predict the 3- and 5-year survival rate of these patients based on the multivariate cox model. The Nomogram calibration curve was used to estimate the performance of the model.

Establishment of cuproptosis-related lncRNA-mRNA Co-expression network
The Pearson’s correlation analysis was performed between the resulting Cuproptosis-Related lncRNA and mRNA. The mRNAs with P value<0.001 and R2 >0 .4 were chosen to be considered to be genes regulating related pathways.

Gene set enrichment and drug sensitivity analysis
Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed to explore the biological function of mRNAs by the “clusterProfiler” package. Gene set enrichment analysis (GSEA) was used to analyze the activated or repressed pathway by differential genes of high- and low-risk groups. The data from genomics of drug sensitivity in cancer (GDSC) were used to be the training cohort and the expression data of CRC patients from TCGA were used to be the testing cohort. The IC50 of 198 drugs were predicted by the oncoPredict package[29].

Immune cell infiltration analysis
To explore the tumor purity, the stromal and immune scores were calculated by using the ESTIMATE algorithm. Immune cell meta genes were obtained from the “Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade[30]”. Calculation of the abundance of immune cell infiltration in high- and low-risk groups was performed based on these meta genes by the “GSVA” package.

Function exploration by scRNA sequencing analysis
The scRNA data (GSM4904234, GSM4904236, GSM4904239, GSM4904245) were obtained from the Gene Expression Omnibus (GEO) database. 9978 single cells were included to perform the quality control. Cells that have transcriptomes with more than 200 and fewer than 6000 expressed genes and mitochondrial genes occupying fewer than 5 % were preserved. Finally, 6086 cells were included in the study to do the dimension reduction and clustering by the “Seurat” package. T-distributed stochastic neighbor embedding (t-SNE) analysis was used to visualize the results in two dimensions. The “FindALLMarkers” function was used to identify marker genes of each cluster.

Statistical analysis
All statistical analyses and visualization were conducted with the R software (version 4.0.3). The GraphPad software (Graphpad Prism 10.0, Graphpad Software) and Microsoft Excel were used to compare the baseline clinical data for all patients with the Chi-square test. Statistical significance was assessed using Student's t-test (two-tailed), one-way ANOVA, or two-way ANOVA. Results are presented as the means ± SD, and p<0.05 was considered statistically significant. Statistical significance was indicated by *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. All independent experiments were conducted in triplicate or more.

Cell culture and treatments
Human colon cancer cell lines: NCM460, FHC, RKO, HCT116, SW480, CaCO2 and LOVO purchased from the Cell Bank of Type Culture Collection (China Academy of Sciences, Shanghai, China). These cell lines were cultured in high-glucose Dulbecco's Modified Eagle Medium (DMEM; GIBCO, Invitrogen, Shanghai, China) supplemented with 10 % fetal bovine serum (FBS; GIBCO, Invitrogen), 100 U/mL penicillin, and 100 µg/mL streptomycin. Cell cultures were maintained in a humidified incubator at 37°C under 5 % CO₂ and 95 % air. Cu(II)-Elesclomol (ESCu) (E4801, Selleck) is a Cu2+ complex of Elesclomol. A 48h pulse treatment with 10nM ESCu was performed to promote the occurrence of cuproptosis in SW480, LOVO, HCT116 and RKO cells and cells were lysed and collected after 48h. ESCu was dissolved in dimethyl sulfoxide (purchased from Sigma-Aldrich, Shanghai, China). Ferrostatin-1 (HY-100579, MCE) is an effective and selective inhibitor of ferroptosis. Belnacasan (HY-13205, MCE) is an effective and selective inhibitor of IL-converting enzyme (ICE)/caspase-1. All cell lines were verified to be free of mycoplasma contamination.

Total RNA isolation and qRT-PCR validation of CRC Cell lines
Total RNA was extracted from CRC cell lines and subcutaneous tumors in nude mice (purchased from Shanghai South Mode Co., Ltd, Shanghai, China) using the MolPure® Cell/Tissue Total RNA Kit (Yeasen, Shanghai, China). Complementary DNA (cDNA) was produced by RNA using the qPCR First Strand cDNA Synthesis Ready-to-Use Premix Kit (Yeasen, Shanghai, China). The specific primers used in this study were designed and synthesized by Sangon Biotech (Shanghai, China) and are detailed in Supplementary Table S1. According to the instructions, the following procedures were operated: First, enzymes are activated at 50°C for 2 mins and then 95°C for 2 mins. Second, 40 cycles of denaturation at 95°C for 15s. Then, annealing at 58°C for 20s. Finally, extension at 72°C for 30s. The relative expression levels of interested genes were calculated using the 2-ΔΔCT method. Beta-actin was used as the control. All the qPCR reactions were performed at least three times.

Lactate production assay and cell viability assay
The siRNAs of SNHG7 and control group were purchased from the GenePharma (Shanghai, China). SNHG7 overexpression and control plasmids and lentivirus were purchased from Guangzhou Youming Biotechnology Co (Guangzhou, China). The sequences of siRNA were listed in supplementary table S1. They were transfected into SW480 and RKO cells according to the manufacturer's instructions by jetPRIME (Polyplus-transfection, France). Overexpression plasmids and lentivirus were transfected into SW480 and RKO cells using jetPRIME. The Lactate production assay was performed following the manufacturer's protocols (Jiancheng Bioengineering, NanJing). The cells were seeded in 96-wells plates in 5000 cells per well. The CCK8 reagents (Dojindo, Japan) were added to perform the cell viability assay at the concentration of 10 % and after incubation for 2h, the absorbance of each well was measured at 450 nm with the microplate reader.

RNA immunoprecipitation (RIP) and RNA pull-down
RIP assays were conducted using the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, USA) following the manufacturer's protocol. SW480 cells were lysed on ice using RIP lysis buffer, and PCBP2 was added to the lysate along with resuspended beads, followed by incubation at room temperature. The RIP lysate was then combined with bead-antibody complexes in RIP immunoprecipitation buffer and incubated overnight at 4°C. Bead-bound immunoprecipitates were eluted using elution buffer for 30 minutes, and RNA was extracted using a phenol/chloroform/isoamyl alcohol mixture. The purified RNA was subsequently reverse-transcribed into cDNA, and qRT-PCR was performed for further analysis.

Western blotting
Indicated cells and tumor tissues were lysed using RIPA lysis buffer supplemented with protease inhibitors (Fode, China). The lysates were separated by 10 % SDS-PAGE and transferred to PVDF membranes, followed by immunoblotting with specific primary antibodies. The primary antibodies used included PCBP2 (15070-1-AP, Proteintech; 1:1000) and CDKN2A (AF5484, Affinity; 1:1000), with incubation at 4°C overnight. Following primary antibody incubation, the PVDF membrane was incubated with the corresponding HRP-conjugated secondary antibody (HRP-Goat Anti-Rabbit or HRP-Goat Anti-Mouse, diluted in TBST) for 1 hour at room temperature on a shaker. Protein bands were visualized using an enhanced chemiluminescence detection system (Bio-Rad, USA) and ECL reagent (Yeasen, China).

Immunohistochemical (IHC)
IHC staining was performed on paraffin-embedded tissue sections. Following deparaffinization, rehydration, and antigen retrieval, the sections were incubated overnight at 4°C with primary antibodies against CDKN2A (AF5484, Affinity; 1:50) and an anti-Ki67 antibody (AF0198, Affinity; 1:500). After extensive washing, sections were treated with an HRP-conjugated secondary antibody for 60 minutes, counterstained with hematoxylin, and visualized using a 3DHISTECH CaseViewer microscope (Pannoramic SCAN, Hungary). Quantification of CDKN2A and Ki67 expression was conducted by measuring the average optical density (AOD), calculated as AOD = integrated optical density (IOD) / area.

Tumor xenograft assay
Male BALB/c-nu mice (6-8 weeks old; body weight 20-22 g) were obtained from Shanghai South Mode Co., Ltd. (Shanghai, China). The mice were acclimatized under specific pathogen-free (SPF) conditions (humidity 50-55 %, 12 h light/dark cycle, temperature 24-25°C) for one week prior to the initiation of experiments. All animal experiments were conducted in strict accordance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health, USA). The experimental protocols were reviewed and approved by the Animal Care and Use Committee of Guangzhou Medical University, China.
Nude mice were randomly divided into four groups: (1) control group, (2) overexpression group, (3) control + ESCu group (1 mg/kg), and (4) overexpression + ESCu group, with 5-10 mice per group. Mice in the control and control + ESCu groups were subcutaneously injected with the HCT116 colorectal cancer cell line (5 × 10⁶ cells), while those in the overexpression and overexpression + ESCu groups received HCT116 cells overexpressing SNHG7. After a stabilization period of 5 days, the control and overexpression groups were administered intraperitoneal injections of corn oil every other day, whereas the control + ESCu and overexpression + ESCu groups received corn oil containing ESCu. Tumor length and width were measured daily for a total of 11 injections. At the end of the experiment, the mice were euthanized, and subcutaneous tumor samples were collected for further analysis.

Result

SNHG7 is highly expressed in CRC
The overall experimental design is briefly summarized as follows (Fig. 1. A). Data for patients with COAD and READ were retrieved from TCGA database and processed according to the criteria described in the Methods section. Following data filtering, the 461 patients were divided into training group (n=231) and testing group (n=230) randomly to build the Cuproptosis-related lncRNA prognostic signature. The baseline clinical data were shown in Supplementary Table S2, the result of Chi-square test indicated there was no significant difference between the train and test group. Through the data quality control, 5393 lncRNAs were reserved to perform the Pearson’s correlation analysis with the Cuproptosis-related mRNAs, and the 1710 LncRNAs with P value<0.001 and R2 >0.4 were included in the Cuproptosis-related lncRNA. Then those CRLs with zero expression in a quarter of the samples were excluded, and the resulting lncRNAs (n=131) were used to perform the log-rank and Cox survival analysis (Supplementary Table S3 and S4). The hub CRLs (n=11) were obtained by taking the intersection of the log-rank and Cox survival analysis results with p-value less than 0.05. The key CRLs were used to construct the clinical prediction model based on lasso and cox regression (family=cox). Finally, seven hub lncRNAs (AC010973.2, SNHG7, ELFN1-AS1, AC233992.3, MELTF-AS1, MNX1-AS1, AC106045.1) were identified (Fig. 1. B-C). Subsequently the best cut-off point of training, testing, and entire groups were calculated by the “surv_cutpoint” function in the survminer package. Patients over the cut-off point were classified as high-risk and those below were classified as low-risk (Fig. 2. D-F).
Sankey diagram was used to show co-expression network between cuproptosis-related mRNAs and CRLs (Fig. 1. D). GLS was co-expressed with AC010973.2, AC233992.3, MELTF-AS1, and SNHG7. MTF1 was co-expressed with MNX1-AS1. PDHA1 was co-expressed with AC010973.2, MELTF-AS1, SNHG7, AC106045.1, and ELFN1-AS1. PDHB was co-expressed with AC233992.3. Analysis of gene expression revealed that the seven hubs CRLs were significantly up-regulated in the tumor tissue (Fig. 1. E).
The expression levels of the seven CRLs are shown in the boxplot (Fig. 1. F). The seven hubs CRLs were highly expressed in the high-risk group. The single lncRNA KM survival analysis showed that SNHG7 implied a poor prognosis with high expression (Fig. 1. G).

SNHG7 signature offers strong prognostic power and clinical stability in CRC
With the increased risk score, the survival time was reduced and mortality was elevated (Fig. 2. A-C). Survival analysis indicated that Overall Survival was significantly shorter in the high-risk group in the training, test, and entire groups (p < 0.05; Fig. 2. G-I). The AUCs of Time-dependent ROC curves at 1, 3, and 5 years in the training group were 0.71, 0.73, and 0.75, respectively (Fig. 2. J). Then the testing and entire group were used to validate the accuracy of the prognostic model. The AUCs of the ROC in the two groups were acceptable for the lncRNA prognostic signature. All the AUC at 1, 3, and 5 years in the test and entire groups were over 0.7 (Fig. 2. K-L). The above results showed that the CRLs prognostic signature has good performance.
Furthermore, the gender, age, AJCC stage, T, N, M, and risk score were used to establish univariate and multivariate cox model. The results suggested the risk score is an independent risk factor in CRC, which indicated that the higher the risk score, the worse the prognosis (Fig. 3. A-B). In addition, compared with other clinicopathological features, ROC analysis was used to confirm that the risk score has higher accuracy (AUC=0.792) to predict the prognosis of patients (Fig. 3. C). The DCA constructed for this cox model demonstrated that the risk score has the best performance in predicting the OS rates in CRC cohorts (Fig. 3. D). We then investigated the ability of the signature for predicting OS in other cancer types (STAD and KIRC), it also exhibited good predictive performance (Fig. 3. G), which means that this signature may have wide applicability.
The nomogram was established by integrating the risk score, T, N, M stage, age, and gender. The parameter selection is based on multivariate cox regression. Every factor in the nomogram was adding the points based on its contribution to the OS (Fig. 3. E). And the calibration curve was used to demonstrate the reliability of the nomogram (Fig. 3. F). The result indicated that SNHG7 prognostic signature has good stability for clinical use.

Function and drug sensitivity analysis of SNHG7
The co-expression network between the cuproptosis-related lncRNAs and mRNAs in CRC was established and the target mRNAs (n=1500) with P value<0.001 and R2 >0.4 were chosen (Supplementary Table S5). The KEGG function enrichment analysis showed that these target mRNA may be involved in many cancer-related signal pathways and immune regulation processes such as CRC, VEGF signaling pathway, PD-L1 expression and PD-1 checkpoint pathway in cancer, necroptosis, mTOR signaling pathway, mitophagy, EGFR tyrosine kinase inhibitor resistance, and autophagy (Supplementary Fig. 3. A). GO analysis showed that it was involved in many biological processes such as oxidative phosphorylation, mitochondrial ATP synthesis coupled electron transport, regulation of autophagy, response to endoplasmic reticulum stress, and endoplasmic reticulum associated degradation (ERAD) pathway (Supplementary Fig. 3. B).
The data from genomics of drug sensitivity in cancer (GDSC) were used to predict the IC50 of the CRC patients from TCGA as the training cohort (Supplementary Table S6). The top five drugs that have the up-regulated or down-regulated IC50 with the lowest p-value in the high-risk group were screened out. The results showed that LY2109761_1852, AZD3759_1915, Cediranib_1922, PF.4708671_1129, and AZD2014_1441 have higher IC50, which indicates that the patients in the high-risk group are resistant to these drugs; OSI.027_1594, Acetalax_1804, Dihydrorotenone_1827, SB216763_1025, and PF.4708671_1129 have lower IC50, which indicate that the patients in the high-risk group are more sensitive to these drugs (Supplementary Fig. 3. C-D).
To further explore the function of CRLs in the single cell level, the scRNA sequencing analysis was performed using the datasets from GEO[31]. After the initial clustering by Seurat, we specifically selected the tumor epithelial cells (n=1704) for secondary clustering (Fig. 4. A-B). The “FindALLMarkers” function was used to identify marker genes of each cluster (Supplementary Table S7). Only the SNHG7 was identified in CRLs to be the marker gene of cluster 3,10,11 (Fig. 4. C). The GSEA enrichment showed glycolysis were activated in these clusters (Fig. 4. D). The KEGG analysis indicated these clusters may be involved in many cancer-related pathways such as the TCA cycle, mitophagy, ferroptosis, pyruvate metabolism and platinum drug resistance (Fig. 4. E).

The exploration for the role of SNHG7 in cuproptosis
qRT-PCR was performed to further determine the potential of as SNHG7 prognostic biomarkers in CRC. The results showed that SNHG7 was significantly increased in CRC cell lines (Fig. 5. A-B). We first examined the transfection efficiency in CRC cells using qPCR in SW480 and RKO cells with SNHG7-siRNA and HCT116 and RKO with SNHG7-overexpression (Fig. 5. C-D). To explore the change of the SNHG7 under cuproptosis condition, we pulsed the CRC cells with 100nM ESCu for 48h. Then, we found SNHG7 was up-regulated under the treatment of ESCu (Fig. 5. E).
To validate the biological function of SNHG7 in cuproptosis, we performed the lactate production assay, then we found that knockdown of SNHG7 will markedly decrease the level of glycolysis, while overexpressing SNHG7 markedly enhanced glycolysis. (Fig. 5. F-G). Furthermore, we can find the viability of SNHG7 in SNHG7-siRNA group were markedly dcreased compared with the control groups after treatment with ESCu whereas overexpression of SNHG7 notably increased cell survival (Fig. 5. I-L). The results indicated that SNHG7 may inhibit the cuproptosis by upregulation of level of the glycolysis.

SNHG7 interacts with PCBP2 to promote CDKN2A expression and modulate cuproptosis in CRC
Having established that SNHG7 modulates cuproptosis in CRC cells, we next sought to elucidate the underlying molecular mechanism. Previous studies have shown that several genes, including FDX1, CDKN2A, LIAS, LIPT1, DLD, and DLAT, are closely associated with cuproptosis. Our experiments demonstrated that SNHG7 overexpression downregulated FDX1, PDHB, DLD, and DLAT, whereas SNHG7 silencing induced the opposite effect (Fig.S M-V). These findings indirectly confirmed that the observed process was cuproptosis. Using the GEPIA database, we analyzed the relationship between SNHG7 and these genes in CRC patients and identified a positive correlation between SNHG7, PCBP2 and CDKN2A (Fig. 5. M-O). Additionally, using the catRAPID database, we found that SNHG7 has a strong potential to associate with the PCBP2 protein, prompting us to investigate PCBP2′s role in this process. Through RIP assays, we demonstrated that PCBP2 not only associates with SNHG7 but also interacts with CDKN2A (Fig. 6. A-D). When RIP was performed using an alternative exon of CDKN2A alongside SNHG7, no significant enrichment of the target genes was detected by gel electrophoresis, suggesting that a specific binding sequence is required for the RIP assay (Fig. 6C-D).
Subsequently, we conducted qPCR (Fig. 6. I-J) and western blot experiments (Fig. 6. E-H), revealing that silencing SNHG7 in CRC cell lines significantly decreased PCBP2 and CDKN2A expression, whereas overexpressing SNHG7 markedly elevated PCBP2 and CDKN2A levels (Fig. 6. K-L), confirming this correlation. Then, we also found PCBP2 and CDKN2A was up-regulated under the treatment of ESCu (Fig. 6. M-N). In addition, PCBP2 knockdown by siRNA in SNHG7-overexpressing cells significantly decreased cell viability relative to SNHG7 overexpression alone, with no significant difference compared with the control group (Fig. 5. H). To further determine whether CDKN2A mediates the function of SNHG7, we performed a rescue experiment by knocking down CDKN2A in SNHG7-overexpressing cells (Supplementary Fig 5. B). Previous studies have reported that PCBP2 can interact with lncRNAs to regulate cellular functions and gene expression. Consistent with these observations, our results suggest that SNHG7 associates with PCBP2 to facilitate CDKN2A expression, potentially influencing cuproptosis-related cellular responses.

SNHG7-PCBP2-CDKN2A promotes CRC initiation and progression
To further validate the tumor-promoting function of SNHG7 and its role in cuproptosis regulation, we established subcutaneous xenograft models in nude mice using HCT116 cells with either normal SNHG7 expression or stable SNHG7 overexpression. In each group, half of the mice were randomly selected for intraperitoneal injection of ESCu every other day. SNHG7 overexpression in CRC significantly enhanced tumor initiation, growth, and size in the mice. In contrast, treatment with ESCu significantly reduced tumor initiation, growth, and size in both groups compared to controls (Fig. 7. A-C).
Subsequent qPCR and western blot experiments showed that SNHG7 overexpression increased PCBP2 and CDKN2A expression, while ESCu treatment significantly reduced PCBP2 and CDKN2A expression in both groups (Fig. 7. D-G). Immunohistochemical staining further confirmed corresponding changes in PCBP2 and CDKN2A expression (Fig. 7. I-M). These findings indicate that SNHG7 associates with PCBP2, promoting CDKN2A expression and glycolysis, thereby modulating cuproptosis. Collectively, these results indicate that targeting CRC-related lncRNAs such as SNHG7 and their downstream signaling pathways may serve as a promising therapeutic strategy. Fig. 8

Discussion
Effective treatment of CRC remains a significant clinical challenge and a deeper understanding of its pathogenesis is essential for developing effective treatment strategies. Recently a novel form of cell death known as cuproptosis was identified which occurs through the direct binding of copper to lipoylated components of the tricarboxylic acid cycle[7]. LncRNAs are identified as a group of non-protein-coding transcripts with a length longer than 200 nucleotides, existing widely in nucleus and cytoplasm[17,28]. And lncRNAs have been elaborated to play vital functional roles in regulating gene expression at different molecular levels in colorectal cancer[18]. SNHG7, a long noncoding RNA frequently upregulated in various human cancers, promotes malignant behaviors such as tumor cell proliferation, migration, invasion, and chemoresistance through its interactions with miRNAs, transcription factors, and signaling pathways[32]. Experimental studies demonstrate SNHG7′s oncogenic role, with overexpression accelerating tumor growth and metastasis in vitro and in vivo, while knockdown reverses these effects and enhances cancer cell sensitivity to therapies[18]. Clinically, high SNHG7 expression correlates with worse prognosis, advanced disease stages, and poor survival outcomes, establishing it as a promising diagnostic, prognostic, and therapeutic target in cancer research[32]. SNHG7 is associated with poor prognosis in CRC patients, correlating with advanced tumor stage, increased lymphatic and distant metastases, and reduced overall survival​. In this study, we identified a significant elevation of SNHG7 expression in CRC. Mechanistically, SNHG7 was found to directly bind to PCBP2, facilitating the upregulation of CDKN2A, which enhanced cell proliferation and inhibited cuproptosis. Furthermore, our findings underscore the critical biological role of SNHG7 in CRC. These results highlight the potential of SNHG7 as a key driver of CRC progression and as a promising therapeutic target for its treatment.
In our study, we used the lasso and cox regression to bulid a cuproptosis-related lncRNA signature in CRC for the first time. The result showed that the risk score levels were negatively correlated with favorable outcomes. To fully understand the clinical potential of the risk model, we also established the nomogram with the risk score, age, gender, and T, N, M stage. Other cancer cohorts (STAD, LUAD and KIRC) from TCGA were also used to validate efficiency of this model. All of the results revealed the excellent predictive performance of the model. As a result, our model had a powerful role to predict the prognosis of CRC patients. After the co-expression analysis, SNHG7 was chosen to be the risk factors in this model. As is well known that, overexpression of SNHG7 positively correlated with cisplatin resistance in gastric cancer cells that elevated SNHG7 level is found in cisplatin resistant cells[33]. This result is also consistent with our scRNA functional analysis. Furthermore, elevated SNHG7 is associated with unfavorable overall survival, tumor progression, lymph node metastasis and distance metastasis in various carcinomas, and may be served as a promising biomarker to guide therapy[32].
Emerging evidence has linked copper homeostasis to key cellular processes in cancer. For instance, copper exposure has been shown to induce liver mitophagy via the PINK1/Parkin pathway, and this mitophagic response may attenuate copper-induced mitochondrial apoptosis [34]. The mitochondrial oxidative phosphorylation system is necessary for cellular metabolism, and the cuproptosis is dependent on mitochondrial respiration[7,35]. Hypoxia, Nutrient restriction, and Intracellular accumulation of reactive oxygen species will lead to the response to the ER stress in cancer cells. Besides, a large number of studies have revealed that ER stress in tumor cells can promote tumor progression by changing the function of immune cells in the tumor microenvironment or protect tumor cells from ER stress-induced apoptosis in ERAD pathway which assist tumor to a more aggressive, more proliferative phenotype[[36], [37], [38], [39]]. The KEGG and GO functional enrichment analysis indicated that SNHG7 was involved in the mitophagy, mitochondrial ATP synthesis coupled electron transport and response to endoplasmic reticulum stress. We also analyzed the CRLs in single-cell level, 1704 tumor epithelial cells were used to identify the clusters with high expression of SNHG7. By enrichment analyses, SNHG7 was found to be highly correlated with many cancer-related pathways such as the mitophagy, glycolysis, platinum drug resistance and TCA cycle which is an essential process for the cuproptosis[12]. In this study, we also found the expression of SNHG7 was significantly changed under the treatment of ESCu. The results of lactate production and cell viability assay were consistent with our enrichment analysis, SNHG7 may inhibit the cuproptosis by upregulation of level of the glycolysis. Therefore, identifying lncRNAs associated with cuproptosis in CRC may provide guidance for predicting prognosis of CRC patients and evaluating immune cell infiltration levels.
PCBP2 is a versatile RNA-binding protein that participates in various cellular processes, including mRNA stability, translation regulation, and stress response modulation[24,40]. PCBP2 is frequently overexpressed in CRC and plays a pivotal role in promoting tumor growth, metastasis, and resistance to therapies by stabilizing oncogenic mRNAs and enhancing their translation[41]​. PCBP2 regulates key oncogenic pathways in CRC by binding to the untranslated regions of target mRNAs, such as UKLF, thereby increasing their stability and expression, which facilitates cancer cell proliferation and invasion[24]​. PCBP2 acts as a molecular chaperone for intracellular iron, influencing metabolic processes crucial for CRC progression, including energy production and oxidative stress responses[25]​. This study found that elevated SNHG7 levels were associated with increased PCBP2 expression, contributing to a poorer prognosis in CRC.
CDKN2A is a critical regulator of cell cycle arrest and senescence, playing a pivotal role in inhibiting cell proliferation[42,43]. Aberrant CDKN2A expression disrupts cell cycle control, impairs tumor suppressor mechanisms, and is associated with tumor resistance and immune escape[44]. Moreover, CDKN2A is recognized as a cuproptosis-related gene[12]. In this study, we demonstrated that CDKN2A is regulated by SNHG7, which suppresses cuproptosis in colon cancer, thereby contributing to poorer tumor prognosis. Previous studies have suggested a functional link between CDKN2A and metabolic regulation. In addition, a report has indicated that CDKN2A may participate in the regulation of glycolytic metabolism [45]. Although the precise mechanism remains incompletely understood, our data showing increased lactate production following SNHG7 overexpression support the possibility that the SNHG7/PCBP2/CDKN2A axis contributes to metabolic reprogramming in colorectal cancer cells.
The regulatory mechanisms of RNA-binding proteins are highly intricate, encompassing not only the maintenance of mRNA stability but also the modulation of RNA processing and translation. Further investigation into the specific role of PCBP2 in RNA processing and translation is essential to gain a deeper understanding of the interplay between PCBP2, SNHG7, and CDKN2A. Future studies should continue to explore these regulatory pathways to uncover their broader biological and therapeutic implications.
We have read the other article about the CRLs in colon cancer that has been published carefully. CDKN2A mediates cuproptosis resistance in CRC by regulating copper homeostasis, glycolysis, and metabolic reprogramming[13]​​. Our work thoroughly elucidates the molecular mechanism by which SNHG7 associates with PCBP2 to upregulate CDKN2A expression, thereby promoting glycolysis and resistance to cuproptosis. This detailed analysis bridges critical gaps in understanding the interplay between lncRNAs, metabolic reprogramming, and copper metabolism in cancer. The rescue experiment further supports the functional involvement of CDKN2A in mediating SNHG7-driven phenotypes. These efforts were complemented by robust experimental methodologies, including qPCR, Western blotting, RIP, in vitro cell assays, and in vivo tumor models, ensuring the thorough validation of our findings. However, our study mainly relied on ESCu-induced phenotypes and metabolic readouts to evaluate cuproptosis-related responses. Although these observations are consistent with previously reported cuproptosis characteristics, additional copper-dependence validation experiments (such as copper chelation rescue assays) would further strengthen the mechanistic specificity. Future studies will address this limitation.
By integrating these elements, our study not only advances current knowledge but also sets a benchmark for future research on lncRNA-driven regulatory networks in CRC. Importantly, we translate our findings into actionable insights, identifying CRLs, particularly SNHG7, as promising therapeutic targets for CRC. This alignment of scientific discovery with clinical application underscores the translational potential of our work.

Limitations of the study
Despite the comprehensive experimental validation presented in this study, several limitations should be acknowledged. First, while we have demonstrated a direct interaction between SNHG7 and PCBP2, the specific binding sites and structural motifs mediating this interaction remain to be characterized. Second, similar to our findings, most studies have only demonstrated that CDKN2A inhibits cuproptosis. For instance, Baowen Xu et al. reported that CDKN2A mediates the resistance of fibrosis-promoting M2 macrophages to cuproptosis, thereby contributing to the progression of pulmonary fibrosis[46]. However, the precise molecular mechanism by which CDKN2A regulates cuproptosis remains unknown. This still requires further research.

Data and code availability statement
All of the raw data and original code could be made available upon request. Publicly available datasets were analyzed in this study. This data can be found here: https://portal.gdc.cancer.gov/ and https://www.ncbi.nlm.nih.gov/geo/.

Informed consent statement
Collection and usage of clinical and pathological specimens were approved by The Third Affiliated Hospital of Guangzhou Medical University’s Ethical Committee Board (Approval number: EN-2024300). All patients were informed of sample collection and usage. All animal experiments were approved by the Experimental Animal Ethics Committee of Guangzhou Medical University (Approval number: GY2024-592).

Glossaries
AJCC, American Joint Committee on Cancer; AOD, average optical density; AUC, area under the curve; CDKN2A, cyclin dependent kinase inhibitor 2A; cDNA, complementary DNA; CRC, colorectal cancer; DCA, decision curve analysis; ERAD, endoplasmic reticulum associated degradation; ESCu, Cu(II)-Elesclomol; GDSC, genomics of drug sensitivity in cancer; GEO, Gene Expression Omnibus; GO, Gene Ontology; GSEA, Gene set enrichment analysis; HR, hazard ratios; KEGG, Kyoto Encyclopedia of Genes and Genomes; IHC, Immunohistochemical; KM, Kaplan Meier; lncRNAs, long non-coding RNAs; MSS, microsatellite stability; OS, overall survival; PCBP2, poly(rC)-binding protein 2; RIP, RNA Immunoprecipitation; ROC, receiver operating characteristic; SNHG7, small nucleolar RNA host gene 7; TCA, tricarboxylic acid; t-SNE, T-distributed stochastic neighbor embedding.

Funds
This work was supported by the 10.13039/501100001809National Natural Science Foundation of China (82370529, 82303391), the Guangdong Basic and Applied Basic Research Fund (2022A1515111193), the Guangzhou Science and Technology Plan Project (2023A04J0581, 202201020142, and 2024B03J0466), and Guangzhou Municipal Science and Technology Program key projects (2023C-TS34).

CRediT authorship contribution statement
Qiusan Chen: Writing – review & editing, Writing – original draft, Software, Formal analysis. Qiuyu Song: Writing – review & editing, Formal analysis. Haonan Zhang: Writing – review & editing, Writing – original draft, Software, Project administration, Formal analysis, Data curation. Zhaoxian Lin: Writing – review & editing, Validation. Yulu Shi: Writing – review & editing, Formal analysis. Yixia Chai: Writing – review & editing, Data curation. Xianmei Fang: Writing – review & editing, Writing – original draft. Na Li: Writing – review & editing, Data curation. Yifeng Zheng: Writing – review & editing, Data curation. Yi Yang: Writing – review & editing, Validation. XueYing Wu: Writing – review & editing, Formal analysis. Shanping Wang: Writing – review & editing, Supervision, Project administration. Chengcheng He: Supervision, Project administration, Funding acquisition. Mingsong Li: Writing – review & editing, Supervision, Funding acquisition.

Declaration of competing interest
The authors declare that they have no competing financial or other interests to disclose.