Background
Ovarian cancer (OC) is a heterogeneous tumor with the highest mortality rate and worst prognosis among gynecological malignancies [
1]. There were 313,959 new cases and 207,252 deaths globally in 2020 [
2]. The onset of OC is hidden. Patients often have no obvious symptoms in the early stage, and are often diagnosed when symptoms such as abdominal distension, abdominal pain, and weight loss appear in the late stage. Despite advances in diagnostic techniques and therapeutic strategies in recent years, the long-term survival of OC patients remains unsatisfactory [
3]. Most patients with advanced-stage disease relapse and develop chemoresistance within a few years [
4,
5]. Therefore, there is an urgent need to identify potential tumor prognostic markers and new therapeutic targets to guide treatment decisions and improve prognosis.
Mitophagy, the selective engulfment of dysfunctional or redundant mitochondria by autophagosomes and subsequent degradation in lysosomes, has been established as a major mechanism of mitochondrial quality control [
6]. Abnormal mitophagy is associated with many diseases, including cardiovascular diseases [
7], kidney disease [
8], neurodegenerative disease [
9], and cancer [
10,
11]. The impact of mitophagy on cancer cells are multi-dimensional, and the specific role and mechanism of mitophagy in different cancers and at different cancer stages is still unclear [
11]. The regulation of this process is critical for maintaining cellular homeostasis and has been implicated in acquired drug resistance [
12]. In mammals, there are two main molecular regulatory mechanisms for mitophagy [
13]: the Parkin-dependent pathway, involving PINK1/Parkin-mediated mitophagy; and the Parkin-independent pathway, in which mitophagy is mainly mediated by receptor proteins, such as BNIP3L/NIX, BNIP3, and FUNDC1 [
11]. In addition, more and more new receptor molecules that can regulate mitophagy have been identified in recent years. A recent study showed that mitophagy is closely associated with cisplatin resistance in OC, and cisplatin resistance can be curtailed by blunting Bnip3-mediated mitophagy [
14].
Long non-coding RNAs (lncRNAs) can bind to DNA, RNA, and proteins, and thus participate in gene regulation at the transcriptional, post-transcriptional and epigenetic levels [
15]. LncRNAs are frequently dysregulated in cancer cells [
16]; therefore, they can be considered as therapeutic, diagnostic and prognostic factors for cancer [
15‐
17]. Thus, we explored the prognostic value of lncRNAs associated with mitophagy in OC and the possible regulatory mechanisms between lncRNAs and mitophagy-related genes. Our study may be valuable and meaningful for identifying potential prognostic markers and therapeutic targets in OC.
Methods
Data acquisition
RNA sequence, somatic mutation, and copy number variation (CNV) data, as well as clinicopathological information of 379 patients with OC, were downloaded from The Cancer Genome Atlas (TCGA) database. mRNA and lncRNA data were annotated using GTF files from Ensembl (
http://asia.ensembl.org). We used Perl to integrate and extract lncRNA expression and corresponding clinicopathological data, including the patient number, age, stage, grade, survival status, and survival time. As reported in a previous article, we also used the Reactome database to obtain data on three mitophagy-related signaling pathways: PINK1/Parkin-mediated mitophagy (R-HSA-5,205,685), receptor-mediated mitophagy (R-HSA-8,934,903), and mitophagy (R-HSA-5,205,647) [
18]. Based on the analysis of the combined gene set data, we identified 29 mitophagy-related genes (Additional file 1: Table
S1). The “maftools” package in R software was used to present the mutation data of mitophagy-related genes.
Identification of mitophagy-related lncRNAs
The correlation between lncRNAs and mitophagy-related genes extracted from the TCGA-OV dataset was calculated using the “corrplot” package, and mitophagy-related lncRNAs were screened out according to the p < 0.05 and |R| ≥ 0.4 screening criteria, using Pearson correlation analysis.
Construction of the prognostic signature
Univariate Cox regression analysis was used to identify the mitophagy-related lncRNAs associated with the prognosis of patients with a setting of p < 0.05. LASSO Cox regression analysis with ten-fold cross-validation and forward stepwise regression analysis were then used to conduct a prediction signature of mitophagy-related lncRNAs. At the same time, the risk score of each patient was calculated using the following formula: risk score = Σ (expression value of each lncRNA × corresponding coefficient). The median risk score was used to classify patients into high- and low-risk groups.
Validation of the prognostic signature
To validate this model, we performed a Kaplan Meier (KM) analysis to show the survival differences between the high- and low-risk groups and visualized the survival curves using the “survminer” and “survive” R packages. The 1-, 3-, and 5-year receiver operating characteristic (ROC) curves were drawn to evaluate prognostic prediction efficiency using the KM “survival ROC” R package. We performed univariate and multivariate Cox regression analyses to determine whether our risk model could independently predict prognosis in patients with OC. Furthermore, we divided the TCGA-OV dataset into two sets randomly, TCGA-training and -testing datasets, for internal validation of the model.
Tissue collection
Samples of 51 OC and 40 normal ovarian tissues were collected from the tissue specimen Bank of Shengjing Hospital of China Medical University between 2015 and 2019. All OC patients had not received chemotherapy, radiotherapy or other antitumor therapy before surgery. This study was approved by the Ethics Committee of Shengjing Hospital of China Medical University, and informed consent was obtained from all patients.
Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)
Total RNA was extracted using TRIzol reagent (Takara Bio, Kusatsu, Japan) and evaluated using the NanoDrop 2000 system (Thermo Fisher Scientific, Carlsbad, CA, USA) to determine RNA purity and concentration. RNA samples were reverse transcribed using TransScript Uni All-in-One First-Strand cDNA Synthesis SuperMix for qPCR (One-Step gDNA Removal; TransGen Biotech, Beijing, China). PerfectStart Green qPCR SuperMix (TransGen Biotech) was used for qPCR using an ABI 7500 Fast System. The primer sequences of LINC00174 is as follows: forward: GGCCCAACACTTCCCTCAAA, reverse: CAGGGAGAAACGACCTGGAG. We used β-actin as an internal reference and the 2−ΔΔCt method for gene expression analysis.
Cell culture and transfection
We purchased cells from the Chinese Academy of Sciences Cell Bank (Shanghai, China) and cultured them in RPMI 1640 medium (Seven, Beijing, China) with 10% fetal bovine serum (FBS; Procell, Wuhan, China) in a 5% CO2 atmosphere at 37 °C. Long intergenic non-protein coding RNA 00174 (LINC00174) shRNA plasmid was purchased from GeneChem (Shanghai, China). We used lipofectamine 3000 (ThermoFisher Scientific, Waltham, MA) for transfection according to the manufacture’s protocol.
Cell viability assay
The CCK-8 kit (GK10001, GLPBIO, Montclair, CA, USA) was used to assess the viability of cells. We inoculated the cell suspension in a 96-well plate (2 × 103 cells/well), and added CCK-8 solution (10 µL per well) every 24 h, then incubated for 2 h. We measured the absorbance at 450 nm by using a microplate reader.
We seeded 1000 cells per well of a 6-well plate. After 10 days in culture, we fixed and stained the cells with 4% paraformaldehyde (PFA) and 0.5% crystal violet, respectively; and then counted the colonies (> 50 cells).
Cell scratch assay
We seeded cells in 6-well plates and waited for cells to grow to 90% confluency. We gently draw a straight line in each well with a 200 µL pipette tip, then washed the well 3 times with phosphate-buffered saline (PBS), and imaged the scratches with a microscope (Nikon, Japan) under 10x objective lens. Cells were cultured for 24 h in FBS-free medium before images were captured again.
Transwell migration assay
Cell migration assay was performed using transwell chamber (8-µm pore size transwell filter) in a 24-well plate. We added 700 µL of medium containing 10% FBS to the lower chamber and 200 µL FBS-free medium (2 × 104 cells) to the upper chamber. After incubation for 24 h, we fixed the cells with 4% PFA and stained the cells with 0.5% crystal violet. The stained migrated cells on the membrane were photographed under 20x objective lens and manually counted.
Apoptosis assay with flow cytometry
After transfection for 48 h, the cells were collected and resuspended into a single cell suspension. Then we added Annexin V-FITC and PI staining solution (Vazyme, Nanjing, China) according to the manufacturer’s protocol. The cells were incubated in the dark for 10 min at room temperature and then analyzed by flow cytometry (Beckman Coulter, Brea, CA, USA).
Construction of a nomogram model
The nomogram, which combined clinicopathological information and risk score, was plotted using the “rms” R package and analyzed to improve clinical applicability. Calibration curves were constructed to evaluate the consistency between the predicted and actual survival rates. Decision curve analysis (DCA) was used to incorporate patients or decision-makers preferences for clinical utility [
19]. Meanwhile, area under the curve (AUC) values were calculated using the “survival,” “survminer,” and “time ROC” packages to compare the differential performance between the nomogram, risk score and clinicopathological information.
Drug sensitivity analysis
We used the R package “pRRophetic” to predict the drug response. We used Ridge’s regression to estimate the half-maximum inhibitory concentration (IC50) of each patient, and 10-fold cross-validation to estimate the accuracy of the prediction. The drug sensitivity analysis was based on the Genomics of Drug Sensitivity in Cancer (GDSC) database [
18].
GSVA analysis and construction of a ceRNA Network
Gene set variation analysis (GSVA) is an unsupervised and nonparametric method for assessing the enrichment of gene sets associated with mRNA expression data [
20]. We used the “GSVA” packages to assess the potential differences in biological functions between the high- and low-risk groups and constructed a ceRNA network based on selected mitophagy-related lncRNAs and corresponding mitophagy-related genes. The miRDB [
21] and miRWalk [
22] websites were used to identify microRNAs (miRNAs) interacting with selected mitophagy-related lncRNAs and mitophagy-related genes, respectively. The overlapping miRNAs were selected to construct the ceRNA network, which was visualized using Cytoscape software.
Statistical analysis
Statistical analyses were performed using R (v4.0.0) and GraphPad Prism 9 (La Jolla, CA, USA). Univariate and multivariate Cox regression analyses were used to evaluate the independence of the mitophagy-related lncRNA signature in OC. For comparisons between the two groups, the unpaired Student’s t-test was used for variables with normal distribution. The Mann–Whitney U test was used to analyze variables with non-normal distribution. Analysis of variance (ANOVA) or the Kruskal–Wallis test was used to compare three or more groups. Statistical significance was set at p < 0.05 unless otherwise specified. * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.
Discussion
OC is the deadliest gynecological cancer. The current first-line treatment of OC includes cytoreductive surgery and platinum–taxane chemotherapy [
23]. Following frontline treatment, tumor recurs in most patients with OC; the five-year survival rate is approximately 45% [
23]. Chemoresistance is a significant hindrance to therapeutic efficacy in patients with OC [
24]. Although angiogenesis inhibitor bevacizumab and poly (ADP-ribose) polymerase (PARP) inhibitors, such as olaparib and niraparib, have shown efficacy in prolonging progression-free survival (PFS) in recent years, they do not extend OS [
25,
26], indicating the need for more effective therapy.
Mitophagy, as a key mitochondrial quality control mechanism [
8], plays an important role in the process of carcinogenesis, including its progression and treatment; and serves as an important regulatory mechanism for maintaining intracellular and extracellular homeostasis [
11]. Mitophagy is considered a double-edged sword in cancer. On the one hand, it can reduce oxidative stress by clearing dysfunctional or redundant mitochondria, which may prevent carcinogenesis; on the other hand, it may protect tumor cells from apoptosis or necrosis by helping cancer cells survive under stress, thereby promoting cancer progression. Overall, the molecular mechanisms involved in the regulation of mitophagy are diverse and complex, and involve crosstalk [
11]. Recent studies have revealed that mitophagy plays an important role in OC, particularly in chemotherapy resistance [
14,
27]. Therefore, a better understanding of the role of mitophagy in OC development and chemoresistance may provide new prognostic markers and targets for the clinical treatment of OC.
In our study, we investigated the mutation status of mitophagy-related genes in samples from the TCGA-OV dataset and observed that 4.36% of the samples harbored gene mutations. Of all the mitophagy-related genes examined in the OC patient samples,
MFN1,
MFN2, and
UBC exhibited the highest mutation frequencies.
MFN1 and
MFN2 are GTPases essential for mitochondrial fusion [
28], and
UBC plays a key role in maintaining ubiquitin homeostasis [
29]. Mitochondrial fusion, division, and ubiquitin homeostasis play pivotal roles in mitophagy. In addition, we noted that alterations in CNV were common among mitophagy-related genes. These alterations may be the main factor responsible for the disturbed expression of some mitophagy-related genes, especially those encoding PINK1 and MFN1. KM analysis showed that 19 mitophagy-related genes were associated with the prognosis of patients with OC. For example,
ATG5,
FUNDC1, and
TOMM5 were determined to be prognostic protective factors, whereas
ATG12,
MFN1,
MFN2,
PINK1, and
PRKN were prognostic risk factors.
Recently, a close correlation between mitophagy-related genes and lncRNAs was reported [
30‐
33], and their interactions can regulate the expression of target genes and cellular biological functions. However, the roles of mitophagy-related lncRNAs in OC remain unclear and require further investigation. In this study, based on five mitophagy-related lncRNAs, a prognostic risk model was constructed. Furthermore, the results showed that the risk model is a reliable prognostic indicator of OC and that high risk scores are indicators of poor prognosis. Moreover, the risk model was correlated with clinicopathological features such as age, stage, and grade.
One of the five mitophagy-related lncRNAs, LINC00174, is an lncRNA whose high expression was associated with poor prognosis in our risk model. LINC00174 plays an oncogenic role in several cancers, including colorectal cancer [
34,
35], renal clear cell carcinoma [
36], hepatocellular carcinoma [
37], breast cancer [
38], glioma [
39‐
42], osteosarcoma [
43], thymic epithelial tumors [
44], and lung cancer [
45,
46]. Contrary to these conclusions, Cheng et al. suggested that LINC00174 acts as a tumor suppressor gene in non-small cell lung cancer, and overexpression of LINC00174 inhibits NSCLC cell migration and proliferation, and induces apoptosis [
47]. However, the role of LINC00174 in OC has not been reported; thus, it was selected for experimental verification. We observed that the expression of LINC00174 in OC tissue was higher than that in normal ovarian tissue. At the same time, we detected the expression of LINC00174 in three OC cell lines. In OVCAR3 cells, the expression of LINC00174 was significantly higher than that in the normal ovarian epithelial cell lines. In contrast, in the other two OC cell lines, there was no increase in its expression. Therefore, we knocked down LINC00174 in OVCAR3 cells to study its role in the development of OC and demonstrated that cell viability and proliferation of OVCAR3 cells were significantly inhibited, as well as cell migration ability, while cell apoptosis was promoted. These results prove that LINC00174 plays a cancer-promoting role in OC and may become a potential therapeutic target for patients with OC.
To further understand the clinical applicability of the risk score, a nomogram combining the risk score and clinicopathological information was constructed, which proved to be a feasible tool for predicting the survival probability of OC patients. The GDSC database provides access to explore association between risk score and clinical treatment [
18]. We noted that the sensitivity to 40 anti-tumor drugs was closely related to the risk score, providing a promising prospect for individualized treatment of OC patients in clinical practice. A higher risk score could predict decreased sensitivity to therapeutic drugs such as cisplatin and olaparib in OC patients, indicating that a higher risk score may be associated with cisplatin resistance. Therefore, the risk score is an independent prognostic tool as well as triggers further consideration of the relationship between mitophagy-related lncRNAs and OC therapeutics.
GSVA results showed that the high- and low-risk groups were enriched in different signaling pathways, further explaining the heterogeneity between the two groups and their potentially different mechanisms. The ceRNA mechanism is a widely reported method by which lncRNAs regulate mRNA expression. Therefore, we predicted miRNAs that may interact with mitophagy-related lncRNAs, and miRNAs that may interact with mitophagy-related genes using online websites and then constructed a ceRNA network. We speculated that mitophagy-related lncRNAs, such as LINC00174, may regulate the expression of downstream mitophagy-related genes through a ceRNA mechanism, thereby affecting the progression of OC. However, further studies are required to validate these findings.
Our study had some limitations. First, we did not validate our model using an external dataset. Second, we did not experimentally validate all five lncRNAs to clarify the practical application of the model in clinical practice, and we did not conduct relevant experiments on mitophagy. We also constructed a ceRNA network to speculate on the possible regulatory mechanism between mitophagy-related lncRNAs and mitophagy-related genes; however, we did not experimentally validate the ceRNA network. Nonetheless, the present model was validated using internal datasets; therefore, the results are still reliable and acceptable.
In summary, to the best of our knowledge, there are no previous reports on the use of mitophagy-related lncRNAs to predict the prognosis of patients with OC. We successfully constructed and verified a risk model based on five mitophagy-related lncRNAs for OC and identified a key lncRNA, LINC00174, that may contribute to OC development. The results of our study may contribute to further understanding of the role of mitophagy-related lncRNAs in OC progression as well as drug treatment responses, which highlights the potential of this model in prognosis prediction and targeted therapy of OC. Well-designed experiments are needed in the future to further verify the reliability of the model and the molecular mechanisms by which LINC00174 promotes OC cell proliferation and migration.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.