Identification of a Serum Three-MicroRNA Signature for Cervical Cancer Diagnosis
Cervical cancer (CC) remains a leading cause of cancer-related mortality among women globally. Current screening methods, including human papillomavirus (HPV) testing and cytology-based approaches, suffer from limitations such as high false-positive rates, delayed detection, and insufficient sensitivity. These challenges underscore the urgent need for non-invasive biomarkers with improved diagnostic accuracy. Circulating microRNAs (miRNAs), small non-coding RNAs stably present in serum or plasma, have emerged as promising candidates for cancer biomarkers due to their stability and disease-specific expression patterns. This study aimed to identify and validate a serum miRNA signature for the diagnosis of cervical cancer through a multi-stage experimental approach.
Experimental Design and Participant Recruitment
The study enrolled 108 CC patients and 108 normal control (NC) participants from the First Affiliated Hospital of Nanjing Medical University between 2016 and 2017. Inclusion criteria for CC patients were histologically confirmed cervical cancer, while exclusion criteria included prior treatments such as chemotherapy or radiotherapy, concurrent malignancies, or severe systemic diseases. Clinical characteristics, including Tumor Node Metastasis (TNM) staging and histological subtypes, were documented (Supplementary Tables 1 and 2). Ethical approval was obtained from the Institutional Review Board (No. 2016-SRFA-148).
Multi-Stage miRNA Screening and Validation
The investigation comprised four sequential phases: screening, training, testing, and external validation.
Screening Stage
Twenty CC and ten NC serum samples were pooled into two CC pools and one NC pool. miRNA expression profiling was performed using the Exiqon miRCURY-Ready-to-Use PCR-Human-panel-I + II-V1.M platform, which evaluates 174 serum/plasma miRNAs. Selection criteria included a cycle threshold (Ct) value 1.5 or <0.67 in CC pools compared to the NC pool. This process identified 29 candidate miRNAs (Supplementary Table 3). Additionally, four miRNAs (miR-196a-5p, miR-218, miR-21-5p, miR-20a-5p) previously implicated in CC were included for analysis.
Training and Testing Stages
Candidate miRNAs were further evaluated using quantitative reverse transcription-polymerase chain reaction (qRT-PCR) in the training cohort (30 CC vs. 30 NC) and testing cohort (60 CC vs. 60 NC). The Bulge-Loop™ miRNA qRT-PCR Primer Set and SYBR Premix Ex Taq II were employed for amplification. Normalization was performed using cel-miR-39 (exogenous control), miR-16-5p (endogenous serum control), and RNU6B (U6; tissue control). Relative miRNA expression was calculated via the 2^−ΔΔCt method.
Three miRNAs demonstrated consistent dysregulation across cohorts:
- miR-20a-5p and miR-122-5p were significantly upregulated in CC serum (P 1.5).
- miR-133a-3p was downregulated (P < 0.01; FC <0.67) (Figure 1, Supplementary Table 4).
Diagnostic Performance of the miRNA Panel
Receiver-operating characteristic (ROC) curve analysis evaluated the diagnostic potential of individual miRNAs and their combination. In the combined training and testing cohorts (90 CC vs. 90 NC):
- miR-122-5p yielded an area under the curve (AUC) of 0.672 (95% CI: 0.581–0.763; sensitivity = 67.5%, specificity = 65.4%).
- miR-20a-5p achieved an AUC of 0.681 (95% CI: 0.615–0.747; sensitivity = 62.0%, specificity = 79.5%).
- miR-133a-3p showed an AUC of 0.666 (95% CI: 0.620–0.712; sensitivity = 68.9%, specificity = 69.3%) (Supplementary Figure 2).
A logistic regression model combining the three miRNAs significantly improved diagnostic accuracy:
- Logit (P) = 0.541 + 0.396 × miR-20a-5p + 0.368 × miR-122-5p − 0.843 × miR-133a-3p.
- The panel achieved AUCs of 0.816 (training + testing; sensitivity = 70.5%, specificity = 85.6%), 0.833 (training cohort), 0.813 (testing cohort), and 0.808 in external validation (18 CC vs. 18 NC; sensitivity = 74.6%, specificity = 72.5%) (Supplementary Figure 3).
Subgroup Analyses and Clinical Relevance
Subgroup analyses assessed the miRNA panel’s performance across TNM stages and histological subtypes:
- Histological Subtypes: No significant differences in miRNA expression were observed between squamous cell carcinoma and adenocarcinoma (P >0.05).
- TNM Stages: miR-122-5p and miR-20a-5p were consistently upregulated in early (stage I/II) and advanced (stage III/IV) CC compared to NCs (P <0.05). The panel reliably discriminated CC patients at all stages from NCs, with AUCs ranging from 0.661 (stage IV) to 0.722 (stage III) (Supplementary Figure 4).
Tissue and Exosomal miRNA Expression
To explore the biological relevance of the identified miRNAs, their expression was analyzed in tissue (24 CC vs. 24 NC) and serum-derived exosome samples (24 CC vs. 24 NC):
- Tissue Samples: miR-20a-5p was significantly upregulated in CC tissues (P <0.05), suggesting a potential tumor-derived origin (Supplementary Figure 5).
- Exosomal Samples: miR-122-5p and miR-20a-5p were enriched in CC-derived exosomes (P <0.05), implicating exosome-mediated communication in CC pathogenesis (Supplementary Figure 6).
Functional Annotation of miRNAs
Bioinformatics analysis using DIANA-miRPath v3.0 revealed that miR-122-5p, miR-20a-5p, and miR-133a-3p are involved in cancer-related pathways, including:
- Cell cycle regulation
- p53 signaling pathway
- Pathways in cancer (Supplementary Table 5).
These findings align with the miRNAs’ roles in tumorigenesis and highlight their potential as therapeutic targets.
Conclusion
This study identified a serum three-miRNA signature (miR-20a-5p, miR-122-5p, miR-133a-3p) with robust diagnostic performance for cervical cancer. The panel demonstrated high specificity and sensitivity across multiple validation stages and TNM stages, outperforming individual miRNAs. While further large-scale studies and mechanistic investigations are needed, this signature holds promise as a non-invasive tool to enhance CC diagnosis and reduce reliance on conventional methods with inherent limitations.
doi:10.1097/CM9.0000000000001327
Was this helpful?
0 / 0