A novel autophagy‐related lncRNA prognostic risk model for breast cancer

Long non‐coding RNAs (lncRNAs) are well known as crucial regulators to breast cancer development and are implicated in controlling autophagy. LncRNAs are also emerging as valuable prognostic factors for breast cancer patients. It is critical to identify autophagy‐related lncRNAs with prognostic value in breast cancer. In this study, we identified autophagy‐related lncRNAs in breast cancer by constructing a co‐expression network of autophagy‐related mRNAs‐lncRNAs from The Cancer Genome Atlas (TCGA). We evaluated the prognostic value of these autophagy‐related lncRNAs by univariate and multivariate Cox proportional hazards analyses and eventually obtained a prognostic risk model consisting of 11 autophagy‐related lncRNAs (U62317.4, LINC01016, LINC02166, C6orf99, LINC00992, BAIAP2‐DT, AC245297.3, AC090912.1, Z68871.1, LINC00578 and LINC01871). The risk model was further validated as a novel independent prognostic factor for breast cancer patients based on the calculated risk score by Kaplan‐Meier analysis, univariate and multivariate Cox regression analyses and time‐dependent receiver operating characteristic (ROC) curve analysis. Moreover, based on the risk model, the low‐risk and high‐risk groups displayed different autophagy and oncogenic statues by principal component analysis (PCA) and Gene Set Enrichment Analysis (GSEA) functional annotation. Taken together, these findings suggested that the risk model of the 11 autophagy‐related lncRNAs has significant prognostic value for breast cancer and might be autophagy‐related therapeutic targets in clinical practice.

together, these findings suggested that the risk model of the 11 autophagy-related lncRNAs has significant prognostic value for breast cancer and might be autophagyrelated therapeutic targets in clinical practice.

K E Y W O R D S
autophagy, breast cancer, long non-coding RNAs (lncRNAs), prognosis, risk model a crucial role in many physiological processes and various pathological events, including stress and starvation adaptation, metabolism, inflammation, neurodegenerative disorders and cancer. [4][5][6][7] Over the past few years, an increasing number of studies have indicated that autophagy participates in the development and progression of breast cancer. 8,9 Therefore, identifying key regulators of autophagy is of great importance for both theoretical basis and clinical practice.
Long non-coding RNAs (lncRNAs) are a series of transcript RNAs longer than 200 nucleotides without the capacity of protein-coding. 10 LncRNAs are considered as one of the most sensitive and specific cancer biomarkers, which participate in the development and progression of various cancers at different levels including epigenetic, transcriptional and post-transcriptional regulation. [11][12][13][14] Moreover, accumulating studies have suggested that lncRNAs promote cancer progression and predict worse prognosis in numerous cancers via regulating autophagy. [15][16][17] Therefore, it is valuable to identify key lncRNAs closely related to autophagy and prognosis in breast cancer.
In the present study, we analysed a data set of lncRNA expression in breast cancers from The Cancer Genome Atlas (TCGA) and screened out autophagy-related lncRNAs with prognostic value.
We identified an eleven autophagy-related lncRNA signature with the potential to predict the survival prognosis of breast cancer patients.

| Patient data sets
Breast cancer patients with clinical information and pathology records were obtained from the TCGA (https://cance rgeno me.nih.

| Identification of autophagy-related lncRNAs in breast cancer
A total of 395 autophagy-related encoding genes (mRNAs) were extracted from the Molecular Signatures Database of Gene Set Enrichment Analysis (GSEA: M27935, M6328 and M10281). Finally, 912 autophagy-related lncRNAs were identified by constructing autophagy-related mRNA-lncRNA co-expression network according to the criteria of |Correlation Coefficient| > 0.4 and P < .001 by Pearson correlation analysis using the Limma R package. 18

| Identification of autophagy-related lncRNA prognostic signatures for breast cancer
To identify autophagy-related lncRNAs associated with survival, we performed univariate Cox proportional hazards analysis according to the criteria of P < .01. Subsequently, multivariate Cox analysis was conducted to construct the optimal prognostic risk model based on the Akaike information criterion (AIC = 1444.62), using the survival R package. Based on the following formula, the risk score for each patient was calculated. coef (lncRNAn) was defined as the coefficient of lncRNAs correlated with survival. expr (lncRNAn) was defined as the expression of lncRNAs.
Based on the median risk score, breast cancer patients in the TCGA were divided into a high-risk group and a low-risk group.
Kaplan-Meier survival analysis was performed to estimate the survival difference between the two groups by using the survival and survminer R packages.

| Independent prognostic analysis and ROC curve plotting
To assess the relationship of survival prognosis with clinicopathological factors and risk score, we performed univariate and multivariate Cox regression analyses using the Survival R package.
Time-dependent receiver operating characteristic (ROC) curves were drew to estimate the predictive accuracy for survival time by different clinical pathological factors and risk score using the survivalROC R package.

| Statistical analysis
All statistical analyses were performed using R software (version 3.6.2). A co-expression network of the 11 autophagy-related lncR-NAs-mRNAs with prognostic value was established and visualized using Cytoscape and Sankey diagram. The correlation between 11 autophagy-related lncRNA expressions and clinicopathological factors was analysed by ggpubr R package. Principal component analysis (PCA) was performed for effective dimension reduction, pattern recognition and exploratory visualization of high-dimensional data of the whole-genome, 395 autophagy-related encoding genes and the risk model of the 11 autophagy-related lncRNA expression profiles, respectively. 19,20 Gene Set Enrichment Analysis (GSEA) was used for functional annotation. GSEA (https://www.gsea-msigdb. org/gsea/index.jsp) is a powerful analytical approach for interpreting genome-wide expression profiles. 21 GSEA focuses on gene sets rather than just high scoring genes, which can detect biological processes such as several cancer-related pathways, metabolic pathways, transcriptional programmes and stress responses. GSEA tends to be more reproducible and more interpretable to analyse molecular profiling data. Two-tailed P < .05 was considered statistically significant.

| Evaluation of the risk model of 11 autophagyrelated lncRNAs as an independent prognostic factor for breast cancer patients
To evaluate whether the risk model of the above 11 autophagy-related lncRNAs is an independent prognostic factor for breast cancer, univariate and multivariate Cox regression analyses were conducted. The hazard ratio (HR) of the risk score and 95% CI were 1.507 and 1.308-   Figure 3C), indicating that the prognostic risk model of the 11 autophagy-related lncRNAs for breast cancer is considerably reliable. Taken together, these all indicated that F I G U R E 1 Identification of autophagy-related lncRNAs with significant prognostic value in breast cancer. A, The forest showed the HR (95% CI) and p-value of selected lncRNAs by univariate Cox proportional hazards analysis. B and C, A co-expression network of the 11 autophagy-related lncRNAs-mRNAs with prognostic value was constructed and visualized using Cytoscape and Sankey diagram

| Correlation of the expression of the 11 autophagy-related lncRNAs with clinicopathological factors
To further assess whether the 11 autophagy-related lncRNAs par-

| Different autophagy statuses in the lowrisk and high-risk groups
PCA was performed to compare the difference between low-risk and high-risk groups based on the risk model of the 11 autophagy- of the 11 autophagy-related genes were enrichment in autophagy processes and oncogenic signatures ( Figure 6).
Above these, it all indicated that the low-risk and high-risk groups showed different autophagy and oncogenic statuses.

| D ISCUSS I ON
In the field of clinical treatment, although the overall survival of breast cancer patients has gained great improvements, metastasis and recurrence of breast cancer have constantly grown, which were the major causes of breast cancer mortality. Considerable evidence has revealed that the role of autophagy in the development of cancer is a doubleedged sword. That is, autophagy may serve as either a pro-survival or pro-death mechanism under different circumstances. 22,23 Consistent with the theory, autophagy functioned dual roles in breast cancer.
Autophagy induced the recurrence of metastatic breast cancer by elevating the survival of dormant breast cancer cells. 8 Cytostatic autophagy suppressed the proliferation and metastasis of triple-negative breast cancer cells. 9 Furthermore, numerous researches indicated the crucial role of lncRNAs in autophagy-inducing progression or inhibition of various cancers, such as hepatocellular carcinoma, lung cancer and breast cancer. 24-26 Based on the above, it led us to find potential specific lncRNAs associated with autophagy and the survival prognosis. In this study, we identified the risk model of the 11 autophagyrelated lncRNAs as an independent prognostic factor for breast cancer.
So far, among these 11 autophagy-related lncRNAs, only LINC01016 and LINC00578 have been studied in breast cancer or other cancers.
LINC01016-miR-302a-3p/miR-3130-3p/NFYA/SATB1 axis plays a crucial role in the occurrence of endometrial cancer. 27 In addition, LINC01016 is a direct target of ERα, associated with survival prognosis of breast cancer. 28 It has been reported that LINC00578 is associated with OS in pancreatic cancer and lung adenocarcinoma. 29,30 In accordance with our results, LINC01016 was low-risk autophagy-related  To date, the focus of precision genomic medicine is to find out accurate specific predictive factors for survival prognosis from large medical data sets with clinical outcomes. 35 Thus, there have been some researches aiming to explore autophagy-related prognostic factors using bioinformatics analysis in recent years. Over the past year, three distinct prognostic risk models of autophagy-related encoding genes in breast cancer have been established based on TCGA database by using different screening criteria and statistics methods. [36][37][38] Meanwhile, due to the crucial function of lncRNAs in autophagy, autophagy-related lncRNAs in cancer also arouse more attention. [15][16][17] Recently, the prognostic risk models of autophagy-related lncRNAs were constructed in several cancers including bladder urothelial carcinoma and glioma. 39,40 However, the role of autophagy-related lncRNAs in prognosis of breast cancer remains unclear.
We therefore conducted this research and found out a novel eleven autophagy-related lncRNA prognostic risk model, which may assist clinicians in making individual effective therapeutic decisions.
However, our study has some limitations. First, we applied traditional statistical analyses to build and evaluate the prognostic risk model of 11 autophagy-related lncRNAs. Although the methods have been utilized and validated in many researches, it is critical to improve our further studies with more advanced methodologies and technologies in the future. To further verify our bioinformatics prediction results, in-depth studies on the 11 autophagy-related lncRNAs, including functional experiments and molecular mechanisms, are needed.

| CON CLUS ION
In conclusion, we identified a novel autophagy-related prognos- In the future, with prospective validation, the 11 autophagy-related lncRNAs signature may improve predictive accuracy and guide individualized therapy for breast cancer patients.

ACK N OWLED G EM ENTS
The data of this study were downloaded from The Cancer Genome Atlas (TCGA), and we gratefully acknowledge the patients and operations. This study was supported by grants from the National Natural Science Foundation of China (No. 81773163, 81602564).

CO N FLI C T S O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data utilized in this study are included in this article, and all data supporting the findings of this study are available on reasonable request from the corresponding author.