Hypoxia-related lncRNA Signature to Predict the Survival of Patients with Clear Cell Renal Cell Carcinoma

Background: Renal cell carcinoma is the most common aggressive tumor of the genitourinary system. The main pathological subtype is clear cell renal cell carcinoma (ccRCC), and its treatment options are very limited. Therefore, identifying specic markers of renal clear cell carcinoma is of great signicance for diagnosis and prognosis. Methods: From the TCGA database, we obtained information on 611 patients with renal clear cell carcinoma to analyze the relationship between hypoxia-related lncRNAs and overall survival. According to the coexpression of hypoxia genes and lncRNAs, genes related to hypoxia were identied. Difference analysis and Cox regression analysis were applied to assess survival-related risk factors. According to the median risk score of hypoxia-related genes, patients were divided into high-risk and low-risk groups. According to these gene characteristics and clinical parameters, a nomogram map was built, and GSEA was used for gene function annotation. RT-qRCR, Western Blot and Flow Cytometry were used to determine the role of SNHG19 in RCC cells. Results: By analyzing the coexpression of hypoxia genes and lncRNAs, 310 hypoxia-related genes were obtained. Six sHRlncRs were signicantly correlated with the clinical outcomes of patients with ccRCC. Four sHRlncRs (AC011445.2, PTOV1-AS2, AP004609.3, and SNHG19) with the highest prognostic values were included in the group to construct the HRRS model. The high-risk group had a shorter OS than the low-risk group. HR-lncRNAs were considered to be an independent prognostic factor and associated with OS. The high- and low-risk groups showed different pathways in GSEA. Experiments showed that SNHG19 plays essential roles in autophagy and apoptosis of RCC cells. Conclusion: Our research shows that we established and veried a hypoxia-related lncRNA model that accurately correlates with ccRCC patients. This study also provides novel insights into hypoxia-based mechanisms and provides new biomarkers for the poor prognosis of ccRCC patients. A graphic nomogram based on the HR-lncRNA signature with clinicopathologic features was developed using R software (version 4.0.3) to predict the probable 3-year and 5-year survival of ccRCC patients. The time-varying ROC curve and calibration chart were used to verify the prediction accuracy of the nomogram. The clinical results predicted by the nomogram are displayed on the x-axis and y-axis, in which the 45-degree dashed line represents the ideal prediction.


Introduction
Renal cell carcinoma (RCC) is the most common genitourinary cancer in adults, and 270,000 new cases occur every year worldwide and account for 2-3% of all malignant tumors (1). According to different molecular genetic features, RCC was divided into different histopathologic types. Accounting for 70 to 80% of RCC, clear cell renal cell carcinoma (ccRCC) constitutes the main histopathological subtype (2). Patients with ccRCC are usually diagnosed as advanced and have a poor prognosis because of its insidious onset and partly due to the lack of typical symptoms, effective diagnosis and treatment methods (3). ccRCC is not sensitive to chemotherapy and radiotherapy and shows an inherited predisposition to metastasize and in ltrate (4). However, molecular targeting therapy is effective for advanced or metastatic ccRCC. The biological behavior of ccRCC is dynamic and complex (5).
Consequently, it is essential to search for a robust prognostic predictor with increased resistance to metastatic ccRCC (6).
Tumor biomarkers are substances or processes that can be used for prognosis, risk evaluation and therapeutic responses. Long noncoding RNAs (lncRNAs) are seen as new prognostic biomarkers for many cancers, such as hepatocellular carcinoma (7), gastric cancer(8), lung cancer (9) and pancreatic cancer (10). In recent years, lncRNArelated genes have attracted more attention due to their prediction accuracy. Hypoxia is an important feature of the tumor microenvironment, is associated with poor prognosis, and promotes tumor cell invasion, proliferation, angiogenesis, metastasis and treatment resistance (11). Some studies have reported that under hypoxic conditions, lncRNAs promote oral squamous cell carcinoma invasiveness via lncRNA HAS2-AS1 (12). In pancreatic cancer, lncRNAs promote cancer metastasis via epithelial-mesenchymal transition (13). Based on the important roles of lncRNAs in tumors, many studies have focused on potentially predicting progression and prognosis. lncR-ZNF180-2 is signi cantly expressed in advanced RCC (14). To our knowledge, although many studies have found that lncRNAs are differentially expressed in many cancers, the role of lncRNAs in predicting ccRCC remains unclear. Therefore, identifying promising prognostic markers based on hypoxia-related lncRNA expression is eagerly anticipated.
This research was designed to provide further insight into the clinical potency prognosis estimation based on hypoxia-related lncRNAs in ccRCC. HR-lncR genes with different expression levels were analyzed, and a risk score model was established to predict the poor prognosis of ccRCC patients. Then, we further calculated correlations between them and OS. Additionally, we constructed a nomogram to assess the clinical signi cance and validated the model. We also identi ed underlying biological processes and molecular mechanisms through GSEA. Finally, we built a more personalized accuracy prediction model for ccRCC, which provides a novel perspective. We also found that SNHG19 played an important role in the regulation of cell growth.

Acquisition of HR-lncRs
We downloaded the transcriptome data and clinical data from ccRCC patients in the TCGA database. The patients with OS<30 days were excluded. In total, 611 ccRCC samples were analyzed, including 539 ccRCC samples and 72 normal tissues. Transcriptome data were converted from Ensembl IDs to gene names and were divided into lncRNAs and mRNAs. We also obtained 310 hypoxia-related genes (M641, M5466, M11033 and M10508) from MSigDB.
Clinicopathologic characteristics of the high-and low-risk groups We established a risk evaluation model using the top 4 hypoxia-related lncRNAs (AC011445.2, PTOV1-AS2, AP004609.3, and SNHG19) among the 6 hypoxia-related lncRNAs. Based on the intermediate risk score, the ccRCC samples were divided into high-and low-risk groups (Fig. 1C). Patients with higher risk scores had a higher mortality rate than patients with lower risk scores (Fig. 1D). As the risk score increased, the expression levels of SNHG19, PTOV1-AS2 and AC011445.2 were elevated, while that of AP004609.3 was decreased (Fig. 1E). Patients with lower risk scores had longer survival times (Fig. 2).

Correlation between HRRS and clinicopathologic indicators
A strong association was found between the sHRlncRs and clinicopathological features of ccRCC, such as age, sex, stage, T stage, N stage and M stage. We found that the risk score differed by stage, T stage, N stage and M stage but not by age and sex (Fig. 3).
Furthermore, we conducted a strati cation analysis to investigate the prognostic value of hypoxia-related lncRNAs.
The patients were grouped according to age (≤ 65 and > 65), sex (female and male), tumor grade (grades 1-2 and grades 3-4), stage (stages I-II and stages III-), T stage (T1/T2 and T3/T4), N stage (N0 and N1/N2/N3) and M stage (M0 and M1). As shown in Fig. 4A-G, univariate analysis showed that age, grade, stage, T stage, N stage and M stage were signi cantly associated with OS (P < 0.05). These results suggest that the prognostic signature can precisely identify the prognosis of patients relative to other clinicopathological characteristics.
The HR-lncRNA signature is an independent prognostic factor Univariate and multivariate regression analyses were used to determine the risk factors for assessment. The univariate analysis revealed that the overall survival rate was obviously related to age, stage, grade, T stage, N stage, M stage, and HRRS (P<0.05). However, multivariate analysis showed an increase for only the HRRS, and age was signi cantly related to the overall survival rate ( Table 1)  Evaluation of the prognostic prediction nomogram including the HR-lncRNA prognostic signature risk score Nomograms are tools commonly used by clinicians to accurately predict survival time for a patient by calculating the nomogram score based on the points assigned for each prognostic factor included in the nomogram (15). To accurately estimate the 3-year and 5-year survival probabilities, we established a nomogram by risk score calculated from the HR-lncRNA prognostic signature and other clinicopathological factors (age, sex, grade, stage and N stage) and III-IV) to verify the predictive performance of the risk scores. The subgroup analysis results indicated that the overall survival rate (OS) of the low-risk group was signi cantly higher than that of the high-risk group (Figs. 6).
However, there were no signi cant differences in the N1-3 groups (p = 0.247), which may be due to the small sample size.

Gene set enrichment analysis
We used GSEA to study the potential molecular mechanisms of lncRNA signaling related to hypoxia in the progression of ccRCC. The cancer hallmark results showed base excision repair, homologous recombination, glycerophospholipid metabolism and cytosolic DNA sensing pathways (Fig. 7A).

Discussion
RCC is one of the most common types of urological tumors, with a high recurrence rate and a mortality rate of more than 40%(16, 17), especially when metastasis occurs. Almost 30% of patients undergoing radical or partial nephrectomy experience recurrence or metastasis, which also means a poor prognosis(18). Although innovative and multimodal treatment strategies, including immunotherapy and targeted therapy, have provided many ccRCC patients with more novel options and have prolonged their survival time, the remaining patients have not yet received satisfactory treatment (2,19,20).
Hypoxia is a microenvironmental feature of many tumors and is related to poor prognosis and treatment failure.
Therefore, nding robust prognostic biomarkers remains an urgent challenge. With the rapid development of bioinformatics technology, new prognostic markers in ccRCC can be identi ed. Because of their improvement in predictive accuracy in comparison to standard benchmarks, researchers have recently paid more attention to lncRNA-based benchmarks (21)(22)(23).
lncRNAs are a newly discovered type of noncoding RNA molecule involved in regulating tumor cell development, differentiation, proliferation and apoptosis (24). Hence, they are potential biomarkers that can predict tumor risk and survival outcomes. However, to date, there are no prognostic biomarkers based on hypoxia-related lncRNA expression pro les in ccRCC patients. Nomograms can be used for personalized risk assessment according to the characteristics of patients and are widely used to evaluate tumor prognosis (15). In our study, we used the clinical features of the hypoxia-related lncRNA signature to construct a nomogram. Moreover, compared with the traditional TNM staging system, our nomogram was more accurate and has higher clinical value.
We observed that our hypoxia-related lncRNA signature was involved in cell base excision repair, homologous Although our research shows that hypoxia-related lncRNA signatures are stable, there remain some limitations. First, our ndings must be further validated in other independent cohorts to verify the accuracy of the hypoxia-related lncRNA prognostic signature. Second, our study is based on a single cohort study of 611 patients from an international database and has been veri ed internally. Moreover, there is little research on our studied hypoxia-related lncRNAs. The mechanism of action of lncRNAs related to hypoxia in ccRCC needs further experimental veri cation.
In summary, we constructed a prognostic marker containing 4 lncRNAs associated with hypoxia to predict the OS of ccRCC and validated the marker. We constructed a prognostic signature comprising 4 hypoxia-related lncRNAs to predict the OS of ccRCC and veri ed this signature. Furthermore, the characteristics of the 4 lncRNAs related to hypoxia that we established can predict the clinical value of this predictive nomogram model. It can better predict the prognosis of ccRCC patients than the traditional TNM staging system. We hope that this signature will provide new references for the current prediction of ccRCC prognosis and provide new ideas for hypoxia-related research and treatment strategies.

Patient data acquisition
We  Supplementary Table S1.

Hypoxia-related lncRNAs
We downloaded the hypoxia-related gene set from the Molecular Signatures Database v4.2 (Cellular response to hypoxia, M5466; Genes known to be induced by hypoxia M5466, M11033 and M10508; https://www.gseamsigdb.org/gsea/msigdb/index.jsp) and obtained 310 hypoxia-related genes. Coexpression analysis was applied to analyze the association between the hypoxia score and lncRNA expression in patients with ccRCC. HR-lncRs were identi ed by the criteria of a P value < 0.001 and |cor| > 0.8.

Survival-related HR-lncRs
We used univariate Cox analysis to identify hypoxia-related lncRNAs whose expression levels were signi cantly correlated (P < 0.05) with the OS of patients with ccRCC. We also used the hazard ratio (HR) to classify sHRlncRs into deleterious and protective groups. Subsequently, the candidate sHRlncRs were selected for the following study.

Establishment of the hypoxia-related risk score model (HRRS)
We used multivariate analysis to verify the reliability of the sHRlncRs and used the integrated sHRlncRs as an independent prognostic indicator to develop the HRRS (p<0.05, Table 2). We used the differential expression of sHRlncRs to perform a risk score model, which used the median risk score as the cutoff point. The ccRCC patients were divided into high-and low-risk groups.      Survival curve of ccRCC patients. Kaplan-Meier survival curve analysis showed that the OS of ccRCC patients in the high-risk group was signi cantly shorter than that of ccRCC patients in the low-risk group.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.