Introduction

Hepatocellular carcinoma (HCC), the predominant form of liver cancer, ranks as the sixth most common cancer and the second highest cause of cancer-related fatalities. Its 5-year survival rate is a dismal 18% [1]. Despite considerable advancements in both clinical and experimental cancer treatments, the overall prognosis for HCC patients remains grim due to high rates of tumor recurrence and metastasis post-surgery. Prior studies indicate that the progression of HCC involves a complicated mix of factors and genetic alterations. Thus, deciphering the molecular basis of HCC recurrence and metastasis is vital for enhancing prognosis and treatment approaches [2].

Anoikis, a specific form of programmed cell death, is initiated when cells detach from the surrounding extracellular matrix (ECM) [3]. Evading this process can aid in the proliferation of cancer cells. Anchored growth is crucial in tumorigenesis, particularly in relation to cancer cell metastasis and dissemination [4]. The exact mechanism that enables tumor cells to resist Anoikis remains unclear. Current research suggests that aspects such as epithelial-mesenchymal transition (EMT), integrins, intracellular adhesion complexes, intercellular cadherin and immunoglobulin superfamily, growth factor signaling, and cell cycle progression may all contribute to resistance [5]. Anoikis resistance is also associated with drug resistance and immune cell infiltration in HCC; for example, hepatoma-derived growth factor-related protein-3 promotes anchor-dependent growth and drug resistance of HCC cells [6]. Chen’s research has identified Anoikis-apoptosis-related subgroups demonstrating that BAK1, SPP1, BSG, PBK, and DAP3 are highly correlated with immune infiltration [7].

Long-chain noncoding RNA (lncRNA) is an RNA molecule that does not code for proteins and is longer than 200 nucleotides. A growing body of literature suggests that lncRNA plays a significant role in the development of liver cancer, including processes such as cell proliferation, apoptosis, metastasis, differentiation, and chemoresistance. LncRNA operates its biological functions through various mechanisms, including genomic imprinting, chromatin remodeling, miRNA-lncRNA, and lncRNA–protein interactions [8]. Certain lncRNAs have been identified as regulators of Anoikis or anchorage-independent growth processes. For example, Liu et al. showed that the LncRNA-FOXD2-AS1/miR7/TERT pathway enhances survival, anchorage-independent growth, and desiccation in thyroid cancer cells [9]. Sun et al. found that lncRNA-LEF1-AS1/miR-30-5p/SOX9 facilitates anoikis-related drug resistance and desiccation in rectal cancer cells [10]. At present, research on the role of lncRNAs in fostering anoikis resistance in HCC is scant, necessitating further studies to elucidate this relationship.

Results Identification of differentially expressed anoikis-related genes and functional enrichment analysis

The flowchart depicting the methodology of our study is illustrated in Fig. 1. We compiled a list of 36 genes associated with anoikis from both the Gene Ontology (GO) database and the MSigDB database (Supplementary File 1, Supplemental Digital Content 1, https://links.lww.com/ACD/A520). To compare the expression levels of these 36 genes between HCC tumor tissues and adjacent tissues, we used gene expression data obtained from 365 HCC tumor tissues and 50 adjacent tissues. Based on the criteria of |fold-change| > 1 and P < 0.05, we identified 21 differentially expressed genes (DEGs) (Fig. 2a and b), all of which were upregulated in HCC tumor tissues. Subsequently, functional enrichment analysis was performed to identify GO and Kyoto Encyclopedia of Genes and Genome (KEGG) terms associated with the 21 DEGs. The following GO projects showed significant enrichment: anoikis (P = 4.94E−57), regulation of anoikis (P = 1.67E−34), negative regulation of anoikis (P = 1.68E−23), intrinsic apoptotic signaling pathway (P = 2.17E−08), and negative regulation of Wnt signaling pathway (P = 1.13E−06) (Fig. 2c). KEGG results indicated that the DEGs were mainly enriched in infection-related biological pathways including hepatitis B (Fig. 2d).

Fig. 1:

Flowchart of the study.

Fig. 2:

Analysis of functional enrichment for genes with differential expression related to anoikis. (a) A volcano plot illustrating 21 genes with differential expression. (b) Heat map displaying the variance in expression of anoikis-associated genes in HCC and surrounding tissues (blue denotes low expression level; red signifies high expression level). (c) Enriched GO terms for genes exhibiting differential expression linked to anoikis. (d) KEGG pathways that are enriched for genes showing differential expression connected to anoikis. GO, gene ontology; HCC, hepatocellular carcinoma; KEGG, Kyoto Encyclopedia of Genes and Genome.

Construction of anoikis-related anoikis-related lncRNA signature

To identify lncRNA linked with anoikis, we conducted Pearson correlation analysis on 1202 lncRNA with |Cor| > 0.4 and P < 0.001 (Supplementary File 2, Supplemental Digital Content 2, https://links.lww.com/ACD/A521). This led to the discovery of 880 differentially expressed lncRNAs (logFC > |1|), as shown in Fig. 3a and Supplementary File 3, Supplemental Digital Content 3, https://links.lww.com/ACD/A522. From this pool, we selected 104 at-risk lncRNA based on the results of univariate Cox regression analysis (hazard ratio > 1, P < 0.001) (Supplementary File 4 Supplemental Digital Content 4, https://links.lww.com/ACD/A523). The HCC samples were randomly divided into training (70%, n = 256) and test (30%, n = 119) cohorts to construct a risk model. We performed LASSO Cox regression analysis to obtain a risk model that included six variables when the partial likelihood deviation reached the minimum (Fig. 3b and c), and stepwise multivariate Cox regression analyses were used to confirm the prognostic significance of five genes: AL031985.3, AC026412.3, DDX11-AS1, MKLN1-AS, and TMCC1-AS1 (Table 1). We found that the expression levels of these lncRNA were significantly associated with poor prognosis in Kaplan–Meier survival analysis (P < 0.05) (Fig. 3d), and we calculated a risk score formula. We also established co-expression networks using the five independent lncRNA and anoikis-related genes, revealing the risk association of all five lncRNA (Fig.3e and f). Finally, we divided patients into high- and low-risk groups according to the median score in the training, test, and entire cohorts (Supplementary File 5, Supplemental Digital Content 5, https://links.lww.com/ACD/A524, Supplementary File 6, Supplemental Digital Content 6, https://links.lww.com/ACD/A525, and Supplementary File 7, Supplemental Digital Content 7, https://links.lww.com/ACD/A526).

Table 1 - Multivariate Cox regression analysis of five prognostic lncRNAs lncRNAs Coefficient Hazard ratio AL031985.3 0.1672 1.1820 DDX11-AS1 0.4675 1.5960 AC026412.3 0.5321 1.7024 MKLN1-AS 0.3445 1.4113 TMCC1-AS1 0.2528 1.2877

lncRNAs, long noncoding RNAs.

Fig. 3:

Construction of prognostic risk signature for patients with HCC based on anoikis-related lncRNAs in the training set. (a) The volcano plot of anoikis-related differentially expressed lncRNAs. (b) Distribution plot of the partial likelihood deviation of the LASSO regression. Six variables were retained when the partial likelihood deviation reached the minimum (log lambda = −2.4). (c) Distribution plot of the LASSO coefficient. (d) The expression of the five lncRNAs and poor prognosis for the high expression group. (e) Sankey diagram of the relationship between lncRNAs and mRNAs. *P < 0.05. (f) The study established a prognostic co-expression network to depict the correlation between the most important five pyroptosis-associated lncRNAs and mRNAs. HCC, hepatocellular carcinoma; lncRNAs, long noncoding RNAs.

Evaluation of anoikis-related lncRNA signature

The performance of the model was evaluated by using the receiver operating characteristic (ROC) curve over time. The areas under the curve (AUCs) for 1-, 3-, and 5-year survival in the overall cohort were 0.771, 0.685, and 0.692, respectively. These results were consistent with the AUCs obtained for 1-, 3-, and 5-year survival in the training cohort, which were 0.795, 0.694, and 0.714, respectively. Similarly, the AUCs for 1-, 3-, and 5-year survival in the test cohort were 0.706, 0.677, and 0.710, respectively (see Fig. 4a). In the training cohort, patients with higher expression levels of the five ARlncRNAs were classified as high-risk, and their overall survival time decreased as their risk scores increased. The high-risk group had a significantly shorter overall survival time than the low-risk group (Fig. 4b). The same risk score formula was applied to the entire cohort testing cohort (Fig. 4c) and the training cohort (Fig. 4d), and the results were consistent with those obtained in the testing cohort (Fig. 4e).

Fig. 4:

Evaluation and validation of anoikis-related lncRNA signature for overall survival in patients with HCC in three datasets. Risk scores and expression profiles of five lncRNA signatures in the high- and low-risk groups showed in the training cohort. (a and b) ROC analyses and Kaplan–Meier survival in the entire cohort (left), training cohort (middle), and testing cohort (right), respectively. Risk scores and expression profiles of five lncRNA signatures in the high- and low-risk groups showed in the entire cohort (c), training cohort (d), and testing cohort (e). HCC, hepatocellular carcinoma; lncRNAs, long noncoding RNAs; ROC, receiver operating characteristic.

Correlation between anoikis-related lncRNA signature and clinical pathological features of hepatocellular carcinoma

We conducted both univariate and multivariate Cox regression analyses to determine whether ARlncSig was an independent prognostic factor in HCC, while taking into account variables such as age, gender, grade, stage, and tumor-lymph node metastasis (TNM) staging across the entire cohort. The results of the univariate Cox regression analysis showed that both ARlncSig and tumor stage were significantly associated with the overall survival of HCC patients (Fig. 5a). Moreover, the multivariate Cox regression analysis revealed that both ARlncSig and tumor stage were independent prognostic factors for the overall survival of HCC patients (Fig. 5b). Additionally, we compared the ROC curve, AUC at 1 year, and the predictive ability of risk scores with age, gender, grade, stage, T, N, and M. Our findings indicated that the risk scores had better predictive ability than other factors, with an AUC of 0.765 (Fig. 5c). Furthermore, we examined the association of ARlncSig with clinical pathology across the cohort by stratifying cohorts according to age, gender, grade, stage, T, N, and M. The heat map results demonstrated high expression of ARlncSig in the high-risk patient group, and the prognostic significance of risk score in HCC patients was linked to the clinical pathological characteristics of grade (P < 0.001) and stage (P < 0.05) (Fig. 5d).

Fig. 5:

Evaluation of the prognostic anoikis-related lncRNAs signature. (a) Forest plot for univariate Cox regression analysis. (b) Forest plot for multivariate Cox regression analysis. (c) The ROC curve of the risk score and clinicopathological variables. Heat map to show the connection between clinical pathology, clusters, and risk scores. (d) The results indicated that the expression of the five lncRNAs in the high-risk group was significantly higher than that in the low-risk group, and this trend was more obvious in the high-risk group. lncRNAs, long noncoding RNAs; N, lymph node; M, metastasis; ROC, receiver operating characteristic; T, tumor.

Construction and evaluation of the prognostic nomogram

To further investigate the prognostic role of ARlncSig in HCC patients, we conducted single and multivariate Cox regression analyses and explored its association with various clinical factors, including age, gender, grading, staging, T, N, and M. Subsequently, we developed nomograms that integrated risk-scoring models along with other clinical pathological characteristics. These were based on data derived from the TCGA-HCC cohort and were designed to forecast survival rates at 1, 3, and 5 years. Our results showed that the nomograms outperformed single factors in predicting overall survival, as evidenced by the calibration curves closely matching the standard curves(Fig. 6a and b).

Fig. 6:

Construction and evaluation of the nomogram for clinicopathological characteristics and risk signature. (a) Nomogram combining ARlncSig and clinicopathological characteristics for predicting prognosis of HCC patients in the entire cohort. (b) Calibration curve analysis of the nomogram for the probability of overall survival at 1, 3, and 5 years. (c) Cumulative risk curve. (d) Decision curve analysis (DCA) of the predictions. ARlncSig, anoikis-related lncRNA signature; HCC, hepatocellular carcinoma; N, lymph node; M, metastasis; T, tumor. ***P < 0.001.

Moreover, the cumulative risk curve revealed that high-risk patients had higher ARlncSig expression levels compared to low-risk patients at the same survival time (Fig. 6c). Moreover, decision curve analysis revealed that the predictive models, when combined, offered superior clinical utility compared to any individual variable (Fig. 6d). Overall, our findings suggest that ARlncSig may serve as an effective prognostic tool for managing HCC patients.

Gene set enrichment analysis

To investigate the potential molecular mechanisms underlying the association between ARlncSig and HCC, we employed gene set enrichment analysis (GSEA) to perform biological functional annotation. Our analysis revealed that the high-risk group was primarily enriched in ECM receptor interaction, cell cycle, and neuroactive ligand receptor interaction pathways, while the low-risk group was mainly enriched in metabolic pathways such as fatty acid metabolism, primary bile acid biosynthesis, glycine, serine, threonine metabolism, tryptophan metabolism, beta-alanine metabolism, and butanoate metabolism (see Fig. 7a). These results suggest that ARlncSig may be involved in the regulation of various cellular processes that are critical for HCC progression, including ECM interactions, cell proliferation, and neural signaling. Moreover, the distinct metabolic profiles observed between the high- and low-risk groups may reflect differences in the energy demands and nutrient utilization strategies of HCC cells with varying ARlncSig expression levels. Overall, our findings provide new insights into the potential functional roles of ARlncSig in HCC pathogenesis and suggest that it could be a promising prognostic biomarker and therapeutic target for this deadly disease.

Fig. 7:

Functional GSEA analysis of ARlncSig. (a) Immune-related GO terms significantly enriched in high-risk patients. (b) Representative KEGG pathways significantly enriched in high-risk patients. ARlncSig, anoikis-related lncRNA signature; GO, gene ontology; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genome.

The widely acknowledged role of cell cycle and extracellular mechanisms in conferring resistance to anoikis in tumor cells is significant [11]. Recent research indicates that neural receptors, such as the TrkB receptor involved in nutrition, may play a crucial role in the development of resistance to anoikis. Additionally, abnormal metabolism of various substances, including glucose, amino acids, fatty acids, and nucleotides, has been found to be associated with this resistance [12]. Our GO analysis has highlighted a cluster of pathways related to immune functions in the high-risk group, particularly those associated with the biological activities of immunoglobulins. These pathways include the immunoglobulin complex, immunoglobulin receptor binding, circulating immunoglobulin complex, humoral immune response mediated by circulating immunoglobulin, B cell receptor signaling pathway, antigen binding, and phagocytosis recognition (Fig. 7b).

The continuous process of cellular transformation involves modifying the cellular mechanisms responsible for sensing and responding to stimuli from various cell surface molecules, including integrins, cadherins, and the immunoglobulin family of cell adhesion molecules located at adhesion complexes. These alterations enable cells to develop resistance against anoikis, a form of programmed cell death triggered by detachment from the ECM. By evading anoikis, these transformed cells gain the ability to migrate to different organs, ultimately leading to the formation of metastases [5]. The results obtained through GSEA suggest that the ARlncSig marker plays a significant role in immune regulation and may be useful in guiding immunotherapy for patients with HCC. These findings highlight the potential of using the risk score associated with ARlncSig as a biomarker for identifying patients who may benefit from specific immunotherapeutic interventions.

Immune analysis of risk scoring

Anoikis is a type of programmed cell death that occurs when cells lose their anchoring to the ECM. In the case of cancer, the ECM derived from tumors can affect immune cells and result in ineffective immunotherapy and tumor immunosuppression [13]. To understand the relationship between ARlncSig and tumor immune cell infiltration, we analyzed the correlation between the risk score and immune cells. Our analysis revealed that there were significant differences in the levels of immune cell infiltration between the high and low subgroups (Fig. 8a). Furthermore, we investigated the levels of multiple factors related to immune cell infiltration, including antigen presenting cell co-stimulation, C-C chemokine receptor, checkpoint, cytotoxic activity, major histocompatibility complex (MHC) class I, parainflammation, type I interferon (IFN) response, and type II IFN response, between the high and low subgroups. Our findings showed that there were significant differences in the levels of these factors between the two subgroups (Fig. 8b).

Fig. 8:

Immune microenvironment analysis in different risk groups. (a) Heat map of all significantly differential immune responses between high- and low-risk groups based on TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ, MCPCOUNTER, XCELL, and EPIC algorithms. Immune cell scores. Immune function scores (b), immune checkpoints (c), and m6A methylation (d) in high- and low-risk groups based on ssGSEA algorithm. ssGSEA, single-sample gene set enrichment analysis.

It appears that the study focused on exploring the relationship between risk scores and immune checkpoints in the clinical treatment of HCC. The researchers found that certain immune checkpoints, namely BTNL2, CD40, and TNFRSF25, were expressed differently in high- and low-risk groups (Fig. 8c). In addition, the study also looked at the role of N6-methyladenosine (m6A) in regulating immune cells through various mechanisms in HCC [14]. The researchers found that m6A regulators, including METTL3, METTL14, WTAP, KIAA1429, RBM15, ZC3H13, YTHDC1, YTHDC2, YTHDF1, YTHDF2, HNRNPC, FTO, and ALKBH5, had significant differences between the high- and low-risk groups (Fig. 8d). These results suggest that ARlncSig may play a critical role in regulating the immune response to tumors and may be useful in predicting the efficacy of immunotherapy for patients with cancer.

Drug reaction characteristics of anoikis-related lncRNA signature

The study examined the role of ECM in inducing resistance to cell death and cell adhesion-mediated resistance, and the efficacy of chemotherapeutic agents, including immunotherapy, in relation to the risk score developed by the researchers. The findings revealed that the high-risk group was more sensitive to cisplatin, oxaliplatin, sorafenib, and gemcitabine, but less sensitive to 5-fluorouracil and paclitaxel than the low-risk group (Fig. 9a–f). Cisplatin, doxorubicin, and sorafenib are the first-line drugs recommended for treating HCC in China’s standard for diagnosis and treatment of primary liver cancer (2022 edition). Medications such as paclitaxel and 5-fluorouracil have been shown to possess adjuvant therapeutic properties for HCC. The predictive model identified in this study could be a potential predictor of chemosensitivity.

Fig. 9:

Drug effectiveness of different risk groups and expression of ARLs in HCC patients: (a) 5-fluorouracil, (b) cisplatin, (c) oxaliplatin, (d) sorafenib, (e) paclitaxel, and (f) gemcitabine. ARLs, anoikis-related LncRNA; HCC, hepatocellular carcinoma.

Validation of the anoikis-related long noncoding RNAs in hepatocellular carcinoma tissues and cell lines

To validate the expression of seven anoikis-related LncRNA (ARLs) in their signature, the researchers collected eight pairs of liver cancer tissues and adjacent tissues from the Affiliated Tumor Hospital of Nantong University. After performing RT-PCR on the extracted total RNA, they found that three out of the five high-risk ARLs (AL031985.3, AC026412.3, DDX11-AS1, MKLN1-AS, and TMCC1-AS1) had higher expression levels in liver cancer tissues than in adjacent tissues (Fig. 10a).Due to the insufficient sample size, further research is needed to investigate the differences in the expression of the five high-risk ARLs between HCC tissues and adjacent non-tumor tissues. The RT-PCR results showed that, except for AC026412.3, DDX11-AS1, and TMCC1-AS1 which showed no significant difference in expression between some HCC cell lines and the normal cell line L-02, the expression levels of the five high-risk ARLs (AL031985.3, AC026412.3, DDX11-AS1, MKLN1-AS, TMCC1-AS1) were generally higher in the HCC cell lines HepG2 and Huh7 (P < 0.05), while their expression was often the lowest in normal liver cells L-02 (Fig. 10b).

Fig. 10:

The expression of five lncRNAs in HCC patients and different drug-sensitive cell lines. (a) Relative expression of five ARLs in HCC patients. (b) Expression of MKLN1-AS, TMCC1-AS1, DDX11-AS1, AL031985.3, and AC026412.3 in cell lines (ns, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; Student’s t-test). ARLs, anoikis-related LncRNA; HCC, hepatocellular carcinoma; lncRNAs, long noncoding RNAs.

Discussion

Liver cancer is a prevalent malignancy and ranks third in cancer-related deaths. The survival rate of patients greatly depends on distant metastasis after treatment [15], with the majority of cancer-related deaths occurring due to it [16]. The complex mechanism of distant metastasis involves various factors, one of which is resistance to anoikis [17]. Thus, it is crucial to identify relevant biomarkers that predict patient prognosis. Previous studies already developed anoikis-related gene prognosis models for different cancers, yet none have explored lncRNA prognosis models for liver cancer [18–20]. In this study, a predictive signature consisting of five anoikis-related lncRNAs was constructed based on various downstream targets, and its specificity and sensitivity were validated in three cohorts.

Several studies have developed candidate lncRNA models for apoptosis-related pathways of HCC. DDX11-AS1 is used in prognosis models related to ferroptosis, cuproptosis, and cancer cell stemness [21–23]. AL031985.3 is associated with liver cancer ferroptosis, cuproptosis, pyroptosis, and autophagy-related prognosis models [24–26], while AC026412.3 is included in the model predicting copper mortality and focal mortality in HCC [27,28]. MKLN1-AS and TMCC1-AS1 are also part of liver cancer-related models such as pyroptosis, ferroptosis, autophagy, necroptosis, and autophagy, respectively [25,27,29–31]. Overall, the identified lncRNAs may be closely related to liver cancer apoptosis and offer new avenues for targeted therapy.

Anoikis is a critical apoptotic mechanism that prevents tumor metastasis by inhibiting cancer cells from escaping the natural ECM and metastasizing to distant sites [12]. Tumor cells that evade anoikis use various mechanisms, including adaptations to transcription signaling pathways, oxidative stress responses, crosstalk with cell cycle and growth proteins, EMT-related mechanisms, and changes in glucose, fatty acid, amino acid, and nucleotide metabolism [32]. Moreover, research suggests that members of the neurotrophic factor family and their receptors, like BDNF and TrkB, are linked to anoikis resistance in various cancer types, such as endometrial carcinoma, cervical carcinoma, and renal cell carcinoma [33–35]. In addition, GSEA analysis implies that the development of anoikis resistance allows malignant cells not only to survive but also to secure their position, thus extending their life cycle within the biological system. Anoikis resistance has been pinpointed as a necessary condition for both intrahepatic and extrahepatic metastasis of HCC [36].

Autophagy, ferroptosis, pyroptosis, and necroptosis are additional apoptotic mechanisms that are linked to tumor immunity, but the relationship between immunity and anoikis remains unclear [37,38]. Our study results indicate that the risk score we constructed is associated with immune cells such as CD4+ T cells, CD8+ T cells, natural killer cells, monocytes, and macrophages, which promote hepatitis to liver cancer [39]. Immunological functions such as Cytolytic_activity, MHC_class_I, Type_I_IFN_Response, and Type_II_IFN_Response also differ between the high- and low-risk groups. Effector T lymphocytes enhance the cytolytic activity of HCC and offer a promising new treatment option [40]. Low expression of MHC class I is linked to high postoperative recurrence in liver cancer [41]. Furthermore, immune checkpoints such as PD-1, PD-L1, TME, CD27, CD28, CD40, and CD48 are highly expressed in patients with HCC, and our results show high expression of immune checkpoints such as CD27, CD28, CD40, and CD48 [42]. Therefore, our risk model can provide clinicians with a reference for developing tailored immunotherapy for their patients.

Despite our promising results, our study has limitations. We primarily based our analysis on bioinformatics results and lack systematic experimental verification of lncRNA functions. We also require further research to predict and explain the relationship between immune infiltration results and risk scoring models for specific clinical treatment. In addition, the number of samples included in our experiment is too small.

Conclusion

The signature identified in this study shows promise for evaluating both the prognosis and microenvironment of patients with HCC. It could also serve as a new standard for selecting optimal treatment strategies for HCC patients.

Materials and methods Data collection

We downloaded RNA sequencing (RNA-seq) data and clinical information on hepatocellular carcinoma (LIHC) from TCGA website (https://gdc.cancer.gov/). To correct for batch effects from non-biological bias, we used the ‘combat’ algorithm from the SVA R package. We obtained GTF files from Ensembl to annotate and identify lncRNAs. After removing duplicate patients, those without complete follow-up information, and those with zero survival time, we included 50 normal samples and 365 tumor samples in our analysis.

Acquisition of anoikis-related genes

To identify genes associated with anoikis, we obtained a total of 36 protein-encoding genes from the MSigDB (http://www.gsea-msigdb.org/gsea/msigdb) and GO databases (http://geneontology.org/) using the keyword ‘anoikis’. We utilized the limma R package [43] to analyze the differential expression of these genes between normal and tumor samples. Our screening criteria included P < 0.05 and |log|FC| > 1, resulting in the identification of 22 DEGs that are related to anoikis.

Acquisition of anoikis-related lncRNA

Our study involved an examination of 1203 distinct lncRNAs that demonstrated co-expression with the 36 anoikis-related genes. We subjected all lncRNAs to a Pearson correlation analysis, setting a threshold of |Cor| > 0.4 and P < 0.001. Subsequently, we employed the limma R package for assessing the differential expression of these lncRNAs between normal tissue and tumor samples. The screening criteria were set as P < 0.05 and logFC > |1|. This led us to identify 881 differentially expressed lncRNAs that have potential links with anoikis.

Random grouping of data

We used the ‘Caret’ package in R version 6.0.88 to randomly divide 365 HCC patients into two sets: a training set, which included 70% of the patients, and a test set, comprising the remaining 30%. Both these sets contained patients with comprehensive prognostic information and survival times that were not zero (Table 2). Univariate Cox proportional hazards regression was performed on the batch-adjusted cohort with a significance level of P < 0.05 to identify lncRNAs associated with survival. The training cohort underwent the minimum absolute contraction and selection operator (LASSO) algorithm and multi-factor Cox regression model, which were subjected to 10-fold cross-validation and 1000 cycles of 1000 random stimuli to prevent over-fitting [44,45]. The R-package ‘glmnet’ identified genetic markers containing biomarkers that help prognosis and calculated the risk score for each sample in the entire cohort using the formula: Risk score = (lncRNA1 coefficient × lncRNA1 expression) + (lncRNA2 coefficient × lncRNA2 expression) + ... + (lncRNAn coefficient × lncRNAn expression). Patients in TCGA were divided into high- and low-risk groups based on the median risk score in the training and test sets, respectively. Kaplan–Meier curves were plotted to compare overall survival between the risk groups, and the patient’s risk curve was drawn. The ROC package was used to draw the ROC curve and calculate the AUC to evaluate model accuracy. Univariate and multivariate Cox regression analyses were conducted on clinical information of all HCC patients to evaluate whether the risk-scoring model showed good predictive ability independent of other clinicopathological characteristics such as age, gender, grade, and stage. Nomograms that included relevant clinical parameters (sex, grade, age, stage), TNM stage, and risk score were constructed using the ‘rms’ and ‘regplot packages’ to evaluate patient survival at 1, 3, and 5 years. The model’s superiority was further assessed through correction curves, cumulative risk curves, and clinical decision curves.

Table 2 - The information of clinicopathological characteristics of 374 hepatocellular carcinoma (HCC) patients Clinical characteristics Total (377) % Age at diagosis (years) 53 (16–90) Futime: follow-up time (m) 27.06 (0.03–122.5) Status Alive/death 130/235 35.6/64.4 Gender Female/male 119/246 32.6/67.4 Stage I/II/III/IV/NA 170/84/83/4/24 46.6/23.0/22.7/1.1/6.6 Grade G1/G2/G3/G4/NA 55/175/118/12/5 15.1/47.9/32.3/3.3/1.4 T-classification T1/T2/T3/T4/TX/NA 180/91/78/13/1/2 49.3/24.9/21.4/3.6/0.3/0.5 M-classification M0/M1/MX 263/3/99 72.1/0.8/27.1 N-classification N0/N1/NX/NA 248/4/112/1 67.9/1.1/30.7/0.3

Data expressed as mean (min–max).

NA, not applicable.

Functional enrichment analysis

GSEA was conducted using the ‘clusterProfiler’ software package in R to identify potential pathways that are enriched or functionally associated with anoikis marker-related lncRNA markers between high- and low-risk groups [46]. The samples from TCGA were categorized into high- and low-risk groups, determined by an optimal threshold. The GSEA analysis was performed with a P-value cutoff of 1, minGSSIZE of 15, and maxGSSIZE of 300, using the KEGG dataset c2.cp.kegg.symbols.gmt and the GO dataset c5.go.symbols.gm. A P-value less than 0.05 was considered statistically significant.

RNA extraction and quantitative real-time fluorescent quantitative PCR

RNA extraction was performed on HCC cell lines or clinical samples using TRIzol reagent (Takara BioInc, Shiga Prefecture, Japan) according to the manufacturer’s protocol. The cDNA was extracted using a reverse transcription system reagent (Vazyme, #R233-01). Quantitative real-time PCR was carried out on the StepOnePlus system (Applied Biosystems, Shanghai, China) using SYBR qPCR Master Mix (Vazyme, #Q511-02) and 10 μM primers. Relative expression values were normalized to the reference gene (GADPH). The primer pairs used in this study are provided in Table 3.

Table 3 - PCR primer sequences Gene Primers AL831985.3 F: 5′-AGTTCTGAGGTGGAGAGGC-3′
R: 5′-GGAAGAAGGAGAGCGTGAG-3′ AC026412.3 F: 5′-GAGGACTCTCCATTTGTGGT-3′
R: 5′-TGGGGTTTGTTAGGTTTGGT-3′ MKLN1-AS F:5′-CTTTGGGACAGAATGGTGGA-3′
R: 5′-CAGGCAGAAGGACAGAGAGG-3′ TMCC1-AS1 F: 5′-GCTGTGAAGAACGGTGTTCA-3′
R: 5′-GGCACGTTTCAGAGGAAGAG-3′ DDX11-AS1 F: 5′-GTCTGACCCAGATGGAGAAG-3′
R: 5′-GCCAACCTTGTGAACCTGAA-3′
Tumor immune cell infiltration

The immune infiltration levels of the high- and low-risk groups were estimated using seven algorithms: TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ, MCPCOUNTER, XCELL, and EPIC [47]. In order to assess the immune function differences between the high- and low-risk groups based on ARlncSig, we utilized GSVA, GSEABase, limma software packages, and single-sample gene set enrichment analysis algorithms. Following this, we conducted an analysis of the variations in gene expression levels for 47 immune checkpoints and 12 m6A factors across both groups.

Drug sensitivity analysis

The semi-maximal inhibitory concentration values, also known as IC50, for various chemotherapeutic drugs were sourced from the Genomics of Drug Sensitivity in Cancer (SC) database, which can be accessed at https://www.cancerrxgene.org/ [48]. Utilizing the ARlncSig model, a total of 198 pharmaceutical agents underwent susceptibility testing. This was conducted between high and low subgroups using the oncoPredict and limma software packages to analyze the data.

Cell lines and culture conditions

The cell lines were obtained from the National Certified Cell Culture Collection Center in Shanghai, China. Huh7 and HepG2 cells were cultured in DMEM medium (Gibco, Fengxian, Shanghai) with 10% fetal bovine serum and 1% penicillin-streptomycin. L-02 cells were cultured in RPMI medium (Gibco) with 10% fetal bovine serum and 1% penicillin-streptomycin.

Statistical analysis

The statistical analysis was carried out utilizing RStudio and the appropriate software packages. The Kaplan–Meier method and log-rank test were employed to analyze overall survival. The prognostic significance of lncRNA in patients with HCC was assessed by means of univariate and multivariate Cox regression analyses. Pearson correlation analysis was utilized for correlation analysis. A P-value less than 0.05 was considered statistically significant.

Acknowledgements