All data (including age, gender, histologic grade, OS time, survival status and gene expression) were from the TCGA database. After data collation, 1096 breast cancer and 112 standard tissue samples were finally obtained (excluding data with survival time less than 30 days and unrecorded data collection).

The arginine methylation-related genes included in the study were obtained from the GeneCards database. First, we investigated the differentially expressed genes (DEGs, (|log2 FC|> 1). Next, Cox regression analysis was performed to identify key LncRNAs as prognostic features of BC.

We constructed a risk score (RS) based on 8 predictive features. r i s k   s c o r e = ∑ i = 1 n c o e f   A M T   L n c S i g i   X   E X P   A M T   L n c S i g i .

The “Coef AMT LncSigi” in the formula represents the value of the multifactorial regression coefficient, and the “EXP AMT LncSigi” indicates the expression of lncRNA.

Immune cell activity and pathways were calculated for each sample using single-sample GSEA (ssGSEA), and then the proportion of infiltrating immune cells in the tumor samples was assessed and immune-related drugs were evaluated and predicted using the “pRRophetic” R package.

We purchased two cell lines (MCF-7 and MDA-MB-231) from the Cell Bank of the Chinese Academy of Sciences. Cells with good growth conditions were taken and digested with trypsin. Then, the cells were inoculated in 6-well plates(1×10⁵) and placed in the 5% CO2 cell culture incubator for lentiviral transfection experiments until the cell confluence was 20~30%. Lentiviral expression vectors with Z68871.1 sequence were purchased from Shanghai Jikai Gene Medical Technology Co. According to the instructions of the reagent supplier, cells were infected with packaged lentivirus.

MCF-7, MDA-MB-231 experimental group and control group were digested with trypsin and resuspended as a single cell suspension, respectively; 1000 cells/well were put into the corresponding cells in 6-well Petri dishes and placed in the cell incubator for two weeks. The number of colonies was counted for analysis. Colony formation was used to assess the proliferative capacity of the cells. At the same time, in order to further understand the effect of z68871.1 on the radiosensitivity of breast cancer cells, MCF-7 cells were irradiated with doses of 0Gy, 3Gy and 6Gy, and the cell survival fraction was observed by colony formation assay.

Transwell chambers (Corning Costar, 8 μm) were primed with Matrigel matrix gel. The transfected control and experimental cells were resuspended with an FBS-free DMEM medium. After 24 hours of cultivation, we fixed the cells with 4% PFA, stained and counted them.

Cells in the logarithmic growth phase were taken to prepare a single-cell suspension. The cells were counted with a glass plate, and the cell concentration was 2 × 104/ml. Take the cell culture plate, and each group of cells is set at 2000/100 µ l/well, with 3 parallel samples in each group, and add the whole culture medium. Continue to cultivate in the incubator. CCK8 reagent (10 µL/well) was added to 96 well plates after 0, 24, 48, 72 and 96 hours, respectively. Cell-free medium wells can be used as blank control. Continue to incubate in the cell incubator for 1-4 hours. The microplate reader measured the absorbance value of 450nm. Draw the growth curve of each group of cells.

Five-week-old female nude mice (Guangdong Pharmachem Biotechnology Co., Ltd., Guangzhou, China) were divided into two groups. MCF-7 cells were divided into the NC and Z68871.1-ox groups, respectively. The number of cells injected subcutaneously into each mouse was 1×106, and weekly live imaging was performed to record the growth of tumors. This study was approved by the ethics committee of the Cancer Hospital Affiliated with Shandong First Medical University.

R software (https://www.r-project.org/) was used for all analyses. The results of clone formation and transwell invasion experiments were analyzed using ImageJ software, and GraphPad Prism 8.0 was used to analyze the statistical significance of the difference between the two groups. Correlation coefficients were examined using the Pearson correlation test. P < 0.05 was considered significant.

Based on the GeneCards database, 111 arginine methylation-associated genes were obtained, and the screening criterion was a correlation score of >10 for protein-coding genes ( Supplementary Data Sheet 1 ). Subsequently, 21 arginine methylation-associated DEGs were obtained by differential analysis of BC and normal breast tissues ( Figures 1A, B ). Enrichment analysis demonstrated that the above genes were enriched in cell growth and division, hormone and immune, cytokine receptor and tumor-related biological function signaling pathways ( Figures 1C, D ). Then, 1199 arginine methylation-associated lncRNAs ( Supplementary Data Sheet 2 ) were screened by arginine methylation-associated DEGs. Univariate Cox regression screened 48 prognostic-related lncRNAs ( Supplementary Material 3 ). Multivariate Cox regression identified 8 lncRNAs with the best predictive correlation: AL122010.1, OTUD6B-AS1, EGOT, AP005131.2, Z68871.1, AC024361.1, LINC00987, ST7-AS1 ( Table 1 ), and we further visualized the correlation between lncRNAs and mRNAs using Cytoscape ( Figures 1E, F ), showing that Z68871.1, AP005131.2 and AC024361.1 were most closely associated with mRNAs.

We used the RS = (0.542214503 × Z68871.1 expression) + (0.571403384 × OTUD6B-AS1 expression) - (0.8962291962× AC024361.1 expression) - (0.576665716 × LINC0098 expression) - (0.76381661 × AL122010.1 expression) - (0.717527719 × ST7-AS1 expression) - (0.497395506 × AP005131.2 expression) - (0.411082836 × EGO expression) to calculate the score of each patient. Patients in the model were categorized into two groups (high- and low-risk groups), and patient scores were positively related to prognosis ( Figures 2A, B ). lncRNA expression levels also showed significant differences ( Figure 3C ). The ROC curves for the model training set data were 0.75 (1 year), 0.762 (3 years), and 0.777 (5 years), and the K-M curves also demonstrated a poor prognosis for patients with high RS ( Figures 2D, E ).

Using the same threshold, the validation set was divided into two groups. Patient distribution was similar to the training set ( Figure 2F ), and high scores were positively associated with worse prognosis ( Figure 2G ). The validation set AUC value and lncRNA expression levels were similar to the training set ( Figures 2H, I ). K-M analysis proved that the model predicted well (p<0.05, Figure 2, J ).

The results of unifactorial and multifactorial analyses of clinical characteristics demonstrated that the model could predict patient prognosis independently of other clinicopathologic data ( Figures 3A, B ). Next, we constructed a nomogram diagram based on clinicopathologic data T, N, M grading, and risk score, which was used to calculate survival probabilities for clinical workup ( Figure 3C ). Upon further analysis, the risk model was significantly correlated with patient staging and prognosis ( Figures 3D, E ). The ROC curves suggest that predictive risk models perform best when compared to other clinical data ( Figure 3F ). Calibration curves show good agreement between model predictions and actual conditions ( Figures 3G–I ).

To further validate the accuracy of the risk model, we analyzed the survival curves of individual genes in the model ( Figures 4A–H ). We found that similar to the multifactorial results, high expression levels of AC024361.1, LINC00987, AL122010.1, EGOT, ST7-AS1, and AP005131.2 were correlated with a better prognosis; on the contrary, high expression levels of OTUD6B-AS1 and Z68871.1 were associated with a poor prognosis of patients.

We further investigated the biological functions of prognostic factors using GSEA and found that many immune-related pathways and cell metabolism-related pathways were associated with high-risk populations ( Figures 5A, B ). we further performed ssGSEA enrichment analysis and found that in the high-risk populations, most cells (DC cells, Th1 cells, NK cells, Tregs, etc.) were significantly elevated ( Figure 5C ). APC co-stimulation, CCR, T cell co-inhibition, and checkpoint were significantly observed in the high-risk populations ( Figure 5D ). Meanwhile, in order to better understand their correlation, we specifically quantified the relationship coefficients between immune-related pathways and the proportion of immune cells ( Figures 5E–G ). Given the importance of immune checkpoints for BC, we also analyzed 21 common immune checkpoints and found that PDCD1LG2, CD86, IDO1, CD276, TNFSF9, and NRP1 were higher in the high-risk populations ( Figure 5H ). Finally, relevant drug prediction analysis showed that the IC50 values of Etoposide, Rapamycin, Lenalidomide, and Cisplatin were higher in the high-risk populations ( Figure 5I ).

Among these eight screened genes, most of them, except Z68871.1 and AC024361.1, have sufficient basic experimental studies in cancer. According to our preliminary analysis, the closest association between Z68871.1 and mRNA was found, so it is reasonable to suspect that it plays a crucial role in BC. Therefore, in the following experiments, we investigated whether Z68871.1 plays a role in BC development. First, we overexpressed Z68871.1 in two BC cell lines, and in a cellular functionalization assay, a cell cloning assay showed that overexpression of Z68871.1 enhanced cell viability in BC cells ( Figures 6A, B ). Next, we also examined the invasive ability of cells and found that overexpression of Z68871.1 significantly enhanced their invasive ability ( Figures 6C, D ). Next, we used the CCK8 assay to detect the effect of z68871.1 on cell proliferation. The results showed that in the cell line, there was no significant change in each group at 24 hours. After 72 hours, the proliferation ability of overexpressing z68871.1 cells was stronger ( Figures 6E, F ). In order to further understand the effect of z68871.1 on the radiosensitivity of breast cancer cells and provide useful information for clinical treatment and prognosis, MCF-7 cells were irradiated with doses of 0Gy, 3Gy and 6Gy, respectively. The radiosensitivity of MCF-7 cells was evaluated by plate clone formation assay ( Figures 6G, H ). The survival rate of the nc group at 3Gy and 6Gy doses was lower than that of the ox group (P <0.05). The overexpression of z68871.1 attenuated the sensitivity of MCF-7 cells to radiation. Finally, to further test the functional effects of Z68871.1 on BC, we established an animal tumor model to explore the biological effects of Z68871.1. After the mouse tumor model was established, each mouse was subjected to in vivo imaging at intervals of 1 weekday to monitor the tumor growth process and plot the growth curve of intracranial meningiomas. The fluorescence values of the intracranial tumors in the two groups of nude mice did not show any significant difference in the first 2 weeks after cell injection, and the fluorescence values of the tumors in the two groups began to change on day 21 (P <0.05). Twenty-eight days later, the fluorescence values of tumors overexpressing Z68871.1 were remarkably higher than those of the control group ( Figures 6I, J P < 0.001). These suggest that overexpression of Z68871.1 enhances the proliferation and invasion of BC in a nude mouse tumor model.