Differentially expressed key genes between drug-resistant and -sensitive BRCA patients treated with diverse kinds of chemotherapy drugs
In order to characterize the molecular features in BRCA patients with diverse drug responses, we identified drug-resistant and -sensitive differentially expressed lncRNAs, genes, and miRNAs. In Cytoxan, Adriamycin, and Tamoxifen, the quantity of differentially expressed lncRNAs, genes, and miRNAs was variable (Figure 1A). We discovered that most of these differentially expressed molecules were downregulated. Some of these lncRNAs, genes, and miRNAs were differentially expressed in multiple kinds of chemotherapy drugs (Figure 1B). We identified 2446, 579, and 110 common drug response-related lncRNAs, genes, and miRNAs, respectively, in three kinds of drugs. However, several of these common drug response-related lncRNAs, genes, and miRNAs showed different expression patterns (Figure 1C). For example, 9.73%, 8.63% and 12.73% of molecules showed opposite expression patterns among Cytoxan, Adriamycin, and Tamoxifen, respectively. We also focused on a number of key lncRNAs, which also displayed different dysregulated expression patterns (Figure 1D). LncRNA CTB-164N12.1 was upregulated in Cytoxan and Adriamycin but downregulated in Tamoxifen. Several other lncRNAs, including LINC01028, CTD-2128A3.2, and RP11-167H9.4,revealedsimilar dysregulated expression patterns. These results indicate that certain lncRNAs, genes, and miRNAs play essential roles in the drug responses of BRCA patients. Furthermore, several lncRNAs showed more complex regulatory patterns in BRCA drug responses.
Specific features of drug-resistant and -sensitive dysregulated ceRNA networks
An integrated computational pipeline was designed to identify drug-resistant and -sensitive dysregulated ceRNAs. We found 1574, 1630, and 2990 drug-resistant dysregulated ceRNAs in Cytoxan, Adriamycin and Tamoxifen, respectively. Furthermore, 48, 44, and 213 drug-sensitive dysregulated ceRNAs were identified in Cytoxan, Adriamycin and Tamoxifen, respectively (Figure 2A). In order to better describe drug response-related dysregulated ceRNAs, drug-resistant and -sensitive dysregulated ceRNA networks were constructed for the aforementioned three drugs. The drug-resistant dysregulated ceRNA network of Tamoxifen had most edges and nodes (Figure 2B). Three drug-resistant dysregulated ceRNA networks showed scale-free topological features (Figure 2C and D). Only a small number of nodes in these networks had great degrees. This finding indicates that drug response-related dysregulated ceRNA networks are informative biological networks. We also extracted some core modules from the ceRNA networks of Cytoxan, Adriamycin, and Tamoxifen (Figure 2E). The core module in the Adriamycin network had 9 nodes (1 lncRNA, 1 miRNA, and 7 genes) and 15 edges. IGF2BP2-AS1 is an lncRNA located in the antisense chain of coding gene IGF2BP2. A previous study reported that IGF2BP2-AS1 is associated with survival in lung squamous cell carcinoma . Accumulating evidence has also suggested a link between dysregulation of IGF2BP2 and cancer . IGF2BP2 has been hypothesized to serve as an N6-methyladenosine reader for the promotion of cancer stemness-like properties and pancreatic cancer pathogenesis .
Evaluation ofpatterns in drug-resistant and -sensitive dysregulated ceRNAs
We used expression correlations between any two molecules in each dysregulated ceRNA to evaluate the patterns of drug-resistant and -sensitive dysregulated ceRNA networks. We found that the co-expressed correlations between miRNA-lncRNA, miRNA-gene, and gene-lncRNA interactions in both drug-resistant and -sensitive dysregulated ceRNA networks showed similar distributions (Figure 3A). PCC differences between drug-resistant and -sensitive groups for any one kind of interaction (gene-lncRNA, gene-miRNA,or miRNA-lncRNA) were also analyzed. We discovered that most of these differences were concentrated between 0.2 and 0.6 in the three kinds of drugs (Figure 3B). In addition, we analyzed the number of dysregulated interactions. Among the three kinds of drug-resistant dysregulated ceRNAs, two dysregulated interactions were most commonly observed. For example, there were 211, 746 and 617 ceRNAs with one, two, and three dysregulated interactions, respectively, in Adriamycin-resistant ceRNAs. In drug-sensitive dysregulated ceRNAs, most ceRNAs only displayed one dysregulated interaction (Figure 3C, D). These results indicate that our method was accurate and reliable for identifying drug-resistant and -sensitive dysregulated ceRNAs.
Common and specific drug-resistant and -sensitive dysregulated ceRNAs in BRCA
We also evaluated common and specific features of drug-resistant and -sensitive dysregulated ceRNAs. More than 90% of drug response-related ceRNAs were only dysregulated in one kind of drug (Figure 4A). However, a small number of ceRNAs were dysregulated in two or all three kinds of drugs. For example, ceRNA UCP1/miR-130b-3p/RMST were dysregulated in all three drugs (Figure 4B), and ceRNA FOXO1/miR-15a-5p/CLRN1-AS1 were dysregulated in two kinds of drugs. Although some drug response-related ceRNAs were dysregulated in multiple kinds of drugs, their dysregulation patterns were diverse. For example, ceRNA UCA1/miR-16-5p/BACE1 were dysregulated in the three kinds of drugs and all displayed three dysregulated interactions (Figure 4C). More ceRNAs showed different dysregulation patterns than similar ones. Dysregulated interactions were diverse for ceRNA OTX2-AS1/miR-16-5p/TPPP3 in all three kinds of drugs. In addition, we analyzed the frequency of lncRNAs, genes, and miRNAs to identify key molecules (Figure 4D). For example, key lncRNAs like HCP5, FAM182A, and NPHP3-AS1 were discovered. Newly identifiedkey genes included PTEN, HMGA2, and CCND1. We also found that26.28%, 61.05%, and 27.51% of lncRNAs, miRNAs, and genes occurred in three, two, and one kind(s) of drugs, respectively (Figure 4E). Notably, lncRNA HCP5, which had the highest frequency, was only dysregulated in Tamoxifen (Figure 4F). A number of other lncRNAs with high frequency were dysregulated in multiple drugs.
Cancer hallmarks in drug-resistant and -sensitive-dysregulated ceRNAs
In order to depict the roles of drug-resistant and -sensitive dysregulated ceRNAs, we performed a functional analysis of lncRNAs. These lncRNAs were enriched in several essential GO terms, including regulation of biosynthetic processes, response to gonadotropin, polyphosphate metabolic processes, and regulation of inositol phosphate biosynthetic processes (Figure 5A). Some of these GO enrichment terms were related to the development and progression of BRCA. For example, gonadotropin is a glycoprotein hormone that regulates the gonadal development of vertebrates as well as promotes the production and secretion of sex hormones. A large number of studies have reported that BRCA is well known for being strongly influenced by female steroids [28-30]. Inositol phosphate is also thought to play an essential role in BRCA . In addition, we explored the relationships between drug-resistant and -sensitive dysregulated ceRNAs and cancer hallmarks. We found that some cancer hallmarks, such as tissue invasion and metastasis, self-sufficiency, insensitivity to antigrowth signals, and evasion of apoptosis, were associated with multiple genes in drug-resistant and -sensitive dysregulated ceRNAs (Figure 5B). Together, these results indicate that drug-resistant and -sensitive dysregulated ceRNAs have specific functions in BRCA.
Efficacy of drug-resistant and -sensitive dysregulated ceRNAs as prognostic biomarkers in BRCA
Finally, we utilized a comprehensive computational method to evaluate the prognosis of drug-resistant and -sensitive dysregulated ceRNAs. Each drug type displayed a different percentage of survival-related drug-resistant and -sensitive dysregulated ceRNAs (Figure 6A). For example, we observed that 86.36% of drug sensitive-related dysregulated ceRNAs in Cytoxan were related to survival (Figure 6B). The HR values for genes, lncRNAs, and miRNAs in a given dysregulated ceRNA were diverse (Figure 6C). These values represent relative protection and risk for molecule survival. We also evaluated the survival of some key drug-resistant and -sensitive dysregulated ceRNAs (Figure 6D). Dysregulated ceRNA PURA/LINC02120/miR-15a-5p was significantly associated with survival (P<0.001). Dysregulated ceRNA MTDH/LINC00092/miR-542-3p was also significantly associated with survival (P< 0.001). In addition, BRCA patients with high-risk scores showed poorer prognoses. All of these results suggest that drug-resistant and -sensitive dysregulated ceRNAs could serve as effective prognostic biomarkers in BRCA.