Serum miR-200c-3p and miR-6134 as diagnostic biomarkers for epithelial ovarian cancer.

Using microarray analysis, we identified four candidate miRNAs that met the selection criteria. Subsequent real-time RT-PCR validation confirmed that miR-200c-3p and miR-6134 were significantly upregulated in EOC serum compared to controls, with fold changes of 3.2 and 2.4, respectively (Table  2 ). Table 2 MiRNA microarray analysis and real-time RT-PCR for screening miRNAs Microarray analysis Real-time RT-PCR Global normalization EOC/Normal EOC/normal Mann-Whitney U Normal EOC Fold change Fold change p -value hsa-miR-16-5p 7.1 35.5 5 1.7 0.45 hsa-miR-23b-3p 14 35.6 2.6 1.5 0.059 hsa-miR-200c-3p 10 20.9 2.1 3.2 5.2.E-07* hsa-miR-6134 13.8 34.1 2.5 2.4 4.9.E-04* EOC , epithelial ovarian cancer; miRNA , microRNA; Real-time RT-PCR , realtime reverse transcription PCR; * p <0.05; statistically significant MiRNA microarray analysis and real-time RT-PCR for screening EOC , epithelial ovarian cancer; miRNA , microRNA; Real-time RT-PCR , realtime reverse transcription PCR; * p <0.05; statistically significant We explored associations between expression levels of the two miRNAs, CA125, HE4 and IL-6 with disease severity, histological classification, and recurrence (Fig.  2 ). In an analysis based on clinical staging classifications, the expression level of miR-200c-3p showed a significant increase between benign tumors and EOC. The expression level of miR-6134 significantly increased with severity of disease, when comparing normal control and benign tumors to stage I–II and more advanced stages. No significant difference was found between normal and benign tumors in CA125, HE4 and in IL-6. The expression of miR-6134, CA125 and HE4 showed significant differences for all histological classifications compared to the normal control and benign tumor groups. MiR-200c-3p were also significantly different with all histological classifications when compared with benign tumors. In contrast, IL-6 showed no significant difference for the mucinous carcinoma group. When investigated for the presence of recurrence, miR-200c-3p, CA125, and IL-6 showed significantly higher values in recurrent cases; however, miR-6134 and HE4 did not show a significant difference. Fig. 2 The expression levels of two miRNAs, CA125, HE4 and IL-6. a Staging classification; b Histology; and c Recurrence. Expression levels, as determined by real-time RT-PCR, were corrected for the mean expression level of the normal group and are shown on the y-axis. The median value in each group was depicted in each box-plot. a * p  < 0.05 vs. normal, † p  < 0.05 vs. benign tumors, ‡ p  < 0.05 vs. Stages I–II. b * p  < 0.05 vs. normal, † p  < 0.05 vs. benign tumors. c * p  < 0.05 vs. no recurrence The expression levels of two miRNAs, CA125, HE4 and IL-6. a Staging classification; b Histology; and c Recurrence. Expression levels, as determined by real-time RT-PCR, were corrected for the mean expression level of the normal group and are shown on the y-axis. The median value in each group was depicted in each box-plot. a * p  < 0.05 vs. normal, † p  < 0.05 vs. benign tumors, ‡ p  < 0.05 vs. Stages I–II. b * p  < 0.05 vs. normal, † p  < 0.05 vs. benign tumors. c * p  < 0.05 vs. no recurrence To evaluate the diagnostic performance of serum biomarkers for EOC, we compared the levels of miR-6134, miR-200c-3p, CA125, HE4, and IL-6 between two groups: normal controls vs. EOC, and benign ovarian tumors vs. EOC. Receiver operating characteristic (ROC) curve analyses were performed and AUC was calculated to assess each marker’s ability to distinguish between groups (Fig.  3 a, b). In the comparison between normal controls and EOC, HE4 demonstrated the highest AUC (0.929), followed by CA125 (0.870) and miR-6134 (0.818). In contrast, in the comparison between benign tumors and EOC, miR-6134 exhibited the highest AUC (0.933), followed by HE4 (0.882) and miR-200c-3p (0.848). While HE4 showed high AUC values, its sensitivity was limited (0.568) for both comparisons when using age-based cutoff values (70 pmol/L for women under 50 and 140 pmol/L for those 50 and older). To identify optimal biomarker combinations, the Akaike Information Criterion (AIC) was applied (Tables S1 and S2). For the comparison of normal vs. EOC, the best model (Combination 1) included miR-6134, CA125, and HE4, yielding the lowest AIC and an AUC of 0.961 (95% CI: 0.931–0.991). For the comparison of benign tumors vs. EOC, the optimal combination (Combination 2) consisted of miR-6134, CA125, HE4, and IL-6, with an AUC of 0.968 (95% CI: 0.942–0.995). Overall, both Combination 1 and Combination 2 outperformed individual markers in terms of AUC. Notably, although Combination 2 demonstrated the highest AUC, its 95% confidence interval overlapped only with that of miR-6134. This suggests that miR-6134 may offer comparable diagnostic performance to the combination panel. However, as its AUC point estimate was lower than that of Combination 2, the combination should still be regarded as the more accurate model. Further studies using formal statistical methods (e.g., DeLong’s test) are required to determine whether this difference is statistically significant. Fig. 3 Diagnostic value of individual and combined miRNAs, CA125, HE4 and IL-6 as markers in serum. ROC analyses were used for the discrimination of EOC. a ROC curves of Normal vs. EOC and benign tumors vs. EOC. b The performance of markers to detect EOC. AUC: area under the curve, NLR: negative likelihood ratio, NPV: negative predictive value, PLR: positive likelihood ratio. PPV: positive predictive value. The cutoff point of miRNAs and IL-6 was determined by the Youden index. Combination 1: Combination of miR-6134, CA125 and HE4; Combination 2: Combination of miR-6134, CA125, HE4 and IL-6 the best AUC combination determined by AIC (Table S1) Diagnostic value of individual and combined miRNAs, CA125, HE4 and IL-6 as markers in serum. ROC analyses were used for the discrimination of EOC. a ROC curves of Normal vs. EOC and benign tumors vs. EOC. b The performance of markers to detect EOC. AUC: area under the curve, NLR: negative likelihood ratio, NPV: negative predictive value, PLR: positive likelihood ratio. PPV: positive predictive value. The cutoff point of miRNAs and IL-6 was determined by the Youden index. Combination 1: Combination of miR-6134, CA125 and HE4; Combination 2: Combination of miR-6134, CA125, HE4 and IL-6 the best AUC combination determined by AIC (Table S1) We evaluated the ability of candidate miRNAs to distinguish benign tumors from EOC in patients stratified by CA125 levels using a cutoff value (Fig.  4 a). Among cases with CA125 below the cutoff of 35 U/mL, miR-6134 demonstrated the highest discriminatory ability (AUC = 0.927), followed by miR-200c-3p (AUC = 0.843). The combined performance of miR-200c-3p and − 6134, designated as combination 3, when CA125 < 35 U/mL revealed AUC = 0.935 and for CA125 ≥ 35 U/mL it was 0.934 (Fig.  4 b). Even in cases where CA125 was below the cutoff, the combination of miR-200c-3p and miR-6134 discriminated EOC from benign tumors. Therefore, the AUC indicated that combined miR-200c-3p and − 6134 were accurate markers when CA125 was < 35 U/mL or ≥ 35 U/mL. Fig. 4 ROC curves for benign tumors vs. EOC with CA125 values using a threshold of 35 units/mL. a ROC curves of benign tumors vs. EOC for CA125 <35 U/mL and CA125 ≧ 35 U/mL. b The performance of miRNAs to detect EOC. Combination 3: Combination of miR-200c-3p and − 6134 ROC curves for benign tumors vs. EOC with CA125 values using a threshold of 35 units/mL. a ROC curves of benign tumors vs. EOC for CA125 <35 U/mL and CA125 ≧ 35 U/mL. b The performance of miRNAs to detect EOC. Combination 3: Combination of miR-200c-3p and − 6134 Both miR-200c-3p and − 6134 showed no significant difference in expression levels between benign tumors and EOC in frozen tissues (Fig.  5 a, b). A significant difference was not also observed between stages I–II and more advanced stages, nor between recurrence and non-recurrence groups (data not shown). No correlation was observed between serum and the tissue in which miRNAs were expressed. Despite the expression of miR-200c-3p and − 6134 being examined in several cultured cell lines, including ovarian cancer, expression levels varied widely between cell types, and no trend in expression by cancer type was observed (Fig. S1 ). Overall, the expression levels of miR200c-3p and − 6134 in frozen tissues and various cultured cells behaved differently from those in serum. Fig. 5 The expression levels of miR-200c-3p and − 6134 in frozen tissues. Box plots show the relative expression levels of ( a ) miR-200c-3p and ( b ) miR-6134, normalized to the mean expression level in benign tumors, which was set to 1 The expression levels of miR-200c-3p and − 6134 in frozen tissues. Box plots show the relative expression levels of ( a ) miR-200c-3p and ( b ) miR-6134, normalized to the mean expression level in benign tumors, which was set to 1 We examined the localization of expression of miRNA signals in surgical specimens by ISH (Fig.  6 ). Of 81 EOC specimens, miR-6134 was positive in the cytoplasm of 15 cases (69%) of clear cell carcinoma and of 18 cases (69%) of serous carcinoma, respectively. We attempted to detect miR-200c-3p using the same specimens but did not obtain a signal. Namely, miR-6134 was detected in a majority of EOC surgical specimens by ISH and miR-200c-3p was uninformative. Fig. 6 The cytoplasm of cancer cells was stained with an miR-6134 probe by ISH at ×20 objective. To confirm expression in tissues of cases in which miR-6134 is highly expressed in serum, ISH with a specific miR-6134 targeting nucleic acid probe was performed on formalin-fixed paraffin-embedded surgical specimens. Intense miR-6134 ISH signal was detected in the cytoplasmic region, predominantly in epithelial cancer cell. a Clear cell carcinoma with a negative signal. b Clear cell carcinoma with a positive signal. c High grade serous carcinoma with a negative signal. d High grade serous carcinoma with a positive signal. EOC, epithelial ovarian cancer; ISH, in situ hybridization; miRNA, microRNA The cytoplasm of cancer cells was stained with an miR-6134 probe by ISH at ×20 objective. To confirm expression in tissues of cases in which miR-6134 is highly expressed in serum, ISH with a specific miR-6134 targeting nucleic acid probe was performed on formalin-fixed paraffin-embedded surgical specimens. Intense miR-6134 ISH signal was detected in the cytoplasmic region, predominantly in epithelial cancer cell. a Clear cell carcinoma with a negative signal. b Clear cell carcinoma with a positive signal. c High grade serous carcinoma with a negative signal. d High grade serous carcinoma with a positive signal. EOC, epithelial ovarian cancer; ISH, in situ hybridization; miRNA, microRNA

Serum specimens were obtained from patients, aged 19–85 (median 47.5) years, who attended the outpatient clinic at Fujita Health University Hospital, Aichi prefecture, Japan, for routine gynecological examinations from August 2014 to December 2017. All sera were taken before treatments such as surgery or chemotherapy commenced. As a reference for noncancerous control (normal), serum was collected from patients with infertility issues, but not those with other diseases. Serum specimens were obtained from patients with epithelial ovarian cancer (EOC; N  = 64, median age: 60 years) and non-cancerous controls (normal; N  = 40, median age: 38 years) for screening using miRNA microarray analysis and real-time reverse transcription (RT)-PCR (Fig.  1 ). Since obtaining biopsies from healthy volunteers raised ethical concerns, we recruited 40 healthy women undergoing infertility treatment as controls, defining them as “normal”. Next, serum specimens from 250 patients were examined for expression of the selected candidate miRNAs, CA125, HE4 and IL-6. This N  = 250 population included N  = 104 used to select candidate miRNAs with complete normal ( N  = 40) overlap. The sera of the 95 patients with EOC (median age: 57 years) were classified as histologic subtypes: clear cell ( N  = 27), serous carcinoma ( N  = 31), mucinous carcinoma ( N  = 11), endometrioid carcinoma ( N  = 16) and other subtypes ( N  = 10). The staging classification for EOC were Stages I and II ( N  = 57) and a more severe condition ( N  = 38). With a median follow-up of 54 months (range 1–90 months) across studies, the recurrence rate was 36.8% and the median time to recurrence was 13 months (range 1–63 months). The number of serum samples taken from. patients with a benign ovarian tumors (benign tumors) was 115 (median age: 44 years) as shown in Table  1 . Fresh surgical specimens derived from patients with EOC ( N  = 61) and benign tumors ( N  = 10) were promptly immersed in RNAlater (Thermo Fisher Scientific, Waltham, MA, USA) in the operating room after tumor excision, and then frozen for preservation. Fig. 1 Details of the study design Details of the study design From a patient cohort of 250 individuals, 104 specimens were used for extracting candidate miRNAs. The patients from whom the 71 frozen tissue specimens were derived were part of the population of 250 individuals; specimens were collected in the operating room. CA125, cancer antigen 125; EOC, epithelial ovarian cancer; HE4, human epididymis protein 4; IL-6, interleukin 6; miRNA, microRNA; RT-PCR, reverse transcription PCR. Table 1 Characteristics of the patient population ( N  = 250) Noncancerous control (Normal) N  = 40 Age, median (range) 38 (26–44) Benign ovarian tumors N  = 115 Age, median (range) 44 (19–85) Histology, N (%) Teratoma 32 (27.8%) Mucinous cystadenoma 27 (23.5%) Endometriosis 27 (23.5%) Serous cystadenoma 14 (12.2%) Fibroma/Thecoma 10 (8.7%) Struma ovarii 2 (1.7%) Other subtypes 3 (2.6%) CA125, N (%) < 35 U/mL 77 (67.0%) ≥ 35 U/mL 38 (33.0%) Epithelial ovarian cancer (EOC) N  = 95 Age, years, median (range) 57 (26–85) Follow-up time: months, median (range) 54 (1–90) FIGO stage, N (%) Stages I–II 57 (60%) Stages III–IV 38 (40%) Histology, N (%) Clear cell carcinoma 27 (28.4%) Endometrioid carcinoma 16 (16.8%) Serous carcinoma 31 (32.6%) Mucinous carcinoma 11 (11.6%) Other subtypes 10 (10.5%) CA125, N (%) < 35 U/mL 23 (24.2%) ≥ 35 U/mL 72 (75.8%) Prognosis Non-recurrence Recurrence N (%) 60 (63.2%) 35 (36.8%) Age, years, median (range) 55 (26–78) 61 (36–85) Follow-up time: months, median (range) 52 (23–90) 13 (1–67) Survival status Dead Alive N (%) 15 (15.8%) 80 (84.2%) CA125 , cancer antigen125; FIGO , International Federation of Gynecology and Obstetrics. Characteristics of the patient population ( N  = 250) CA125 , cancer antigen125; FIGO , International Federation of Gynecology and Obstetrics. For the microarray analysis, equal volumes of serum from all 64 EOC cases and all 40 control cases were pooled separately to generate composite samples for each group. Total RNA was extracted from 300 µL of serum using 3D-Gene RNA extraction reagent (Toray, Kamakura, Japan) according to the manufacturer’s instructions. The RNA integrity number (RIN) values were 2.70 for the EOC pool and 3.20 for the normal pool. RNA concentrations were 318 pg/µL and 210 pg/µL, respectively. The extracted total RNA was checked by Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and labeled with a 3D-Gene miRNA labeling kit (Toray). Half volumes of labeled RNAs were hybridized onto 3D-Gene Human miRNA Oligo chips (Toray). The annotation and oligonucleotide sequences of the probes were confirmed by miRBase miRNA database Release 21 (miRbase; https://www.mirbase.org ). After stringent washing, fluorescent signals were scanned with a 3D-Gene Scanner (Toray) and analyzed using 3D-Gene Extraction software (Toray). Raw data from each spot was normalized by substitution with a mean intensity of the background signal determined by all blank spot signal intensities and 95% confidence intervals (CI). Measurements of spots with signal intensities greater than 2 standard deviations of the background signal intensity were considered to be valid. The relative expression level of a given miRNA was calculated by comparing the signal intensities of the valid spots throughout the microarray experiments. The normalized data were globally normalized per array, such that the median of the signal intensity was adjusted to 25. The value of each gene was normalized by a method such that the median of the ovarian cancer to control ratio was equalized to one. All labeling, hybridization, washing, and scanning procedures were carried out in a single batch using the same lot of reagents and equipment. Therefore, batch effect correction was not required in this analysis. A total of 2,565 human miRNAs were covered by the 3D-Gene Human miRNA Oligo Chip based on miRBase Release 21, of which 1,693 miRNAs were detected above the signal threshold in the pooled serum samples. For initial candidate selection, we applied the following criteria: signal intensity > 10 and a fold change (EOC vs. normal) > 2. Based on these thresholds, four miRNAs (miR-16-5p, miR-23b-3p, miR-200c-3p, and miR-6134) were identified as candidates for further validation. To normalize serum RNA extraction efficiency, a synthesized cel-miR-39 was used as a synthetic spike-in control RNA oligonucleotide (Thermo Fisher Scientific) as it has no mammalian homolog. Ten microlitres of 0.1 fmol/µL cel-miR-39 was added to each 200 µL of serum specimen. Total RNA, extracted from 200 µL of serum using an miRNeasy Mini Kit (QIAGEN GmbH, Hilden, Germany) yielded a final elution volume of 30 µL. For cDNA synthesis from 2 µL of total RNA, a TaqMan Advanced MicroRNA cDNA Synthesis Kit (Thermo Fisher Scientific) was used according to the manufacturer’s instructions. Quantitative real-time RT-PCR was performed on 5.0 µL of a diluted (1:10) RT cDNA template using a 2X Fast Advanced Master Mix and 20X TaqMan Advanced miRNA Assays (both Thermo Fisher Scientific) on a QuantStudio7 (Thermo Fisher Scientific). TaqMan Advanced MicroRNA Assay probes were as follows: Homo sapiens (hsa)-miR-16-5p (Assay ID: 477860_mir); hsa-miR-200c-3p (Assay ID: 478351_mir); hsa-miR-23b-3p (Assay ID: 483150_mir); hsa-miR-6134 (Assay ID: 480842_mir); and Caenorhabditis elegans cel-miR-39-3p (Assay ID: 478293_mir). Relative fold changes were determined from Ct values using the 2 − ΔΔCt method. Data were normalized to cel-miR-39-3p to account for possible differences in the amount of starting RNA. To examine the local expression of miR-200c-3p and − 6134, which exhibited high expression in serum of EOC, frozen tissues were collected during surgery from 61 patients with EOC from whom serum and 10 patients with a benign tumors were obtained. One hundred milligrams of frozen tissue was thoroughly homogenized in 1,000 µL of QIAzol Lysis Reagent using TaKaRa Biomasher Standard (Takara, Shiga, Japan), followed by total RNA extraction using an miRNeasy Mini Kit. MicroRNA was reverse transcribed using a TaqMan miRNA Reverse Transcription Kit and then quantified by quantitative real-time PCR according to the manufacturer’s instructions, employing the following TaqMan MicroRNA Assays: hsa-miR-203-3p (002300), hsa-miR-6134 (476896_mat), RNU48 (001006), and RNU44 (001094) (Thermo Fisher Scientific). MicroRNA levels were normalized against a combination of the average signal of both RNU48 and RNU44, and presented as − ΔCt values. Serum CA125 and HE4 levels were measured by a chemiluminescent enzyme immunoassay (SRL, Tokyo, Japan). The cutoff value for CA125 was 35 U/mL, and for HE4, it was 70 pmol/L for premenopausal women and 140 pmol/L for postmenopausal women. However, since the study population did not ascertain menopausal age, considering that the average menopausal age among Japanese women is 50 years, we defined those below 50 years of age as 70 pmol/L and those 50 years or older as 140 pmol/L. The concentration of serum IL-6 were determined by enzyme-linked immunosorbent assay (ELISA) using ELISA MAX™ Deluxe sets for human IL-6 (BioLegend, Inc., San Diego, CA, USA), following protocols provided by the manufacturer. The cellular localization of miRNA was examined in 81 surgical formalin-fixed paraffin-embedded specimens of patients with EOC. Double (3′and 5′) digoxigenin-labeled miRCURY LNA detection probes (Exiqon, Vedbaek, Denmark) were used for visualization of the following miRNAs: hsa-miR-200c-3p (probe sequence: TCCATCATTACCCGGCAGTAT, Tm = 84 °C), hsa-miR-6134 (probe sequence: TCTACATCCTACCACCTCA, Tm = 84 °C), and U6 (CACGAATTTGCGTGTCATCCTT, Tm = 84 °C) as a positive control. For in situ hybridization (ISH), 3-µm–thick sections of formalin-fixed paraffin-embedded tissues were mounted on Superfrost glass slides and deparaffinized in xylene baths, followed by serial dilutions of ethanol and phosphate buffered saline (PBS). The slides were then immersed in 0.3% H 2 O 2 for 10 min at room temperature, washed twice with PBS, digested with 20 µg/mL proteinase K (Exiqon) at 37 °C for 15 min, and washed twice with PBS. The slides were pre-hybridized at 53 °C for 30 min in hybridization buffer (Exiqon) and then hybridized at 53 °C for 1 h with 40 nM probes for miR-6134 or 20 nM probes for U6 in hybridization buffer. After stringent saline sodium citrate washes, the slides were blocked with Protein Block Serum-free (Agilent Technologies, Santa Clara, CA, USA) and incubated with anti-digoxigenin–horse-radish peroxidase (POD), Fab fragments from sheep (Roche, Mannheim, Germany) diluted to 1:100 at 37 °C for 1 h. POD signals were visualized using a Liquid DAB + Substrate Chromogen System (Agilent Technologies), and slides were stained with hematoxylin for nuclear staining. The slides were then dehydrated and mounted with a coverslip. A microscope (Power BX-51; Olympus, Tokyo, Japan) was used for observation. All statistical analyses were performed using SPSS for Windows (ver. 22.0.0.0; IBM Corp, Armonk, NY, USA). Data were analyzed by two-tailed Mann–Whitney U tests. We defined p  < 0.05 as significant. We used a conventional receiver operating characteristic curve (ROC) with a Youden index to analyze miRNA and IL-6 levels to determine the cutoff points that yielded the highest combined sensitivity and specificity with respect to distinguishing patients with cancer from those with a normal histology. The interpretation of area under the curve (AUC) was as follows: 1.0, perfect match; 1.0–0.9, high accuracy; 0.9–0.7, moderate accuracy; 0.7–0.5, low accuracy; and 0.5, no better than chance. We used Akaike information criterion (AIC) to compare the goodness-of-fit of each model of combined miRNAs, and unified different markers to determine the predictive probability through logistic regression and then constructed ROC curves according to probability.

Globally, ovarian cancer is the most lethal gynecological cancer in women. In 2022, approximately 324,603 new ovarian cancer cases were diagnosed, accounting for 1.6% of all cancer cases, with an estimated 206,956 deaths [ 1 ]. The most common subtype is epithelial ovarian cancer (EOC). Prognosis is highly stage-dependent: the 5-year survival rate is 92% for stage I but drops to around 30% in advanced stages (II–IV) [ 2 ]. CA125 remains the most widely used serum tumor marker for EOC; however, its sensitivity in early-stage disease is only about 50%, and it can be elevated in benign conditions such as endometriosis or pelvic inflammatory disease [ 3 , 4 ]. These limitations have prompted investigation into alternative biomarkers, such as interleukin-6 (IL-6) [ 5 – 7 ] and human epididymis protein 4 (HE4) [ 8 – 13 ], but their overall diagnostic accuracy remains suboptimal, particularly in differentiating EOC from benign tumors. MicroRNAs (miRNAs) are non-coding RNAs 19–25 nucleotides in length that regulate gene expression by binding to the 3′ untranslated region of target mRNAs [ 14 ]. Approximately two-thirds of human genes are believed to be regulated by miRNAs, and currently, about 2,600 human miRNAs have been cataloged in the miRBase database [ 15 ]. Some act as oncogenes or tumor suppressors [ 14 ]. miRNAs are not only present in tissues but also circulate in blood, where they are bound to proteins or encapsulated in extracellular vesicles, rendering them highly stable and resistant to degradation [ 16 ]. These properties make circulating miRNAs attractive candidates for non-invasive cancer biomarkers. In ovarian cancer tissues, various miRNAs have been found to be differentially expressed compared to normal tissues. However, previous studies have reported inconsistent results regarding their diagnostic utility, likely due to differences in study design or sample types used [ 17 , 18 ]. Given these inconsistencies, further validation using matched serum and tissue samples is needed. Therefore, in this study, we aimed to identify promising serum biomarkers—including miRNAs, IL-6, and conventional markers such as CA125 and HE4—that can distinguish EOC from especially benign ovarian tumors [ 19 , 20 ]. We also examined the correspondence between serum and tumor tissue miRNA expression using matched surgical specimens.

Previous studies have reported that serum CA125 has a sensitivity of 0.80 and a specificity of 0.75 for the diagnosis of ovarian cancer [ 21 ]. In our study, CA125 demonstrated a sensitivity of 0.758 and a specificity of 0.875, indicating a diagnostic performance comparable to previous reports. These findings support the consistency of our results with existing knowledge and provide a reliable foundation for the subsequent analysis of additional biomarkers. Although the CA125 level was found to be elevated in the serum of patients with EOC, this marker has a low sensitivity in the early stages of EOC [ 3 ]. Elevated CA125 levels have also been reported in other physiological or pathological conditions such as menstruation, pregnancy, endometriosis, and inflammatory diseases of the peritoneum [ 22 – 24 ]. Therefore, the development of more accurate biomarkers is desired for patients with EOC. HE4 has been recognized as one of the most promising tests for early detection of ovarian cancer [ 8 – 10 ]. When combined with CA125, its sensitivity and specificity for detecting malignancies in adnexal masses improve significantly compared to CA125 alone [ 11 – 13 ]. However, despite its clinical utility, HE4 levels can be influenced by various benign and malignant diseases, and differences in cutoff values due to racial and patient background variations have prevented it from becoming a definitive diagnostic tool [ 10 , 25 – 30 ]. In our study, HE4 demonstrated high diagnostic performance for distinguishing between normal controls and EOC, showing the highest AUC (0.929) among the markers analyzed. To address the clinical challenge of distinguishing benign tumors from EOC, we further evaluated the diagnostic utility of candidate miRNAs. Notably, miR-6134 showed the highest AUC (0.933), followed by HE4 (AUC = 0.882) and miR-200c-3p (AUC = 0.848). Importantly, in this comparison, HE4 exhibited a particularly low sensitivity of 0.568. This relatively low sensitivity underscores the limitation of HE4 as a standalone diagnostic marker, aligning with previous reports that highlight the challenges associated with defining an optimal cutoff value across different populations [ 10 , 25 – 30 ]. This finding is especially relevant given that clinical differentiation between benign and malignant tumors is a critical challenge in early ovarian cancer detection. While HE4 maintains high specificity and AUC, its limited sensitivity emphasizes the need for supplementary biomarkers such as miR-6134 to enhance detection accuracy. Next, we assessed the diagnostic performance of various biomarker combinations by calculating their respective AUCs. Our analysis revealed that the combination of miR-6134, CA125, and HE4 (Combination 1) demonstrated the best discrimination between normal and EOC cases (AUC = 0.961, 95% CI: 0.931–0.991), while the combination of miR-6134, CA125, HE4, and IL-6 (Combination 2) achieved the highest AUC (0.968, 95% CI: 0.942–0.995) for differentiating benign tumors from EOC. To facilitate reproducibility, all numeric values used in these combined analyses, including miR-200c-3p, miR-6134, CA125, HE4, and IL-6, are summarized in Supplementary Table S3 . Among individual biomarkers, miR-6134 demonstrated the highest AUC (0.933) for differentiating benign tumors from EOC, and also maintained relatively high performance in the normal vs. EOC comparison (AUC = 0.818). Notably, the 95% confidence interval of Combination 2 overlapped only with that of miR-6134, suggesting comparable diagnostic performance. Given the simplicity and potential cost-effectiveness of single-marker testing, miR-6134 may be a particularly attractive candidate for further development. Nonetheless, due to software limitations, we were unable to perform formal statistical testing such as DeLong’s test. Instead, we compared 95% confidence intervals (CIs), which cannot confirm statistical significance. Future studies using appropriate statistical methods (e.g., DeLong’s test or bootstrapping in R) and larger cohorts are warranted to determine the incremental value of combined biomarker models and validate the diagnostic utility of miR-6134. Furthermore, in cases where CA125 was < 35 U/mL—a level typically considered negative—miR-6134 demonstrated a high sensitivity (0.826) and specificity (0.974), with an overall accuracy of 0.94. This suggests its potential utility as a supplementary marker in cases where CA125 fails to detect malignancy. Notably, combining miR-6134 with miR-200c-3p further enhanced diagnostic performance in this subgroup. This finding highlights the potential benefit of miR-based combinations, especially in cases where traditional markers like CA125 are insufficient. In the context of biomarker discovery, previous studies have shown that the miR-200 family is significantly detected in ovarian cancer tissues compared to normal tissues [ 31 , 32 ]. In our study, we quantified the expression levels of serum miRNAs identified by our original microarray screening and confirmed that two miRNAs, miR-200c-3p and miR-6134, were significantly upregulated as shown by real-time RT-PCR. Of these, the biological behavior and cellular role of miR-6134 have not been well described to date. Also consistent with the results of previous studies [ 33 – 35 ], we found that expression levels of miR-200c-3p, CA125, and IL-6 were significantly higher in cases showing recurrence of disease compared to those without any recurrence. In contrast, expression levels of miR-6134 did not correlate with recurrence status. In our study, we investigated the correlation between pre-intervention blood sampling results and subsequent recurrence events. We were particularly interested in elucidating whether conducting longitudinal blood sampling after an intervention could predict future recurrences; this is the intended subject of future research. To confirm the validity of miRNAs as serum tumor markers, we investigated their expression in surgically resected specimens using two methods: real-time RT-PCR on frozen specimens and ISH on paraffin-embedded tissues. The latter method allowed for assessment of local expression at the cellular level within the tumor, although its detection sensitivity was expected to be low. Indeed, miR-6134 was detected in 58–69% of clear cell and serous carcinoma specimens, confirming its expression in tumor cells. Conversely, we were unable to detect miR-200c-3p expression by ISH, possibly due to poor compatibility between the probe and tissues. While RT-PCR allows for direct comparison between serum and tumor expression, it analyzes only a portion of the tumor and may be affected by tumor heterogeneity. Our results revealed no correlation between expression levels in tumors and serum, and tissue-based analysis did not yield clinically useful information. While miRNAs are expressed in the tumor, they may behave differently from their circulating counterparts. One possible hypothesis for the lack of correlation between serum and tissue miRNA levels is the differential release and stabilization of miRNAs in circulation. Circulating miRNAs may originate not only from tumor cells but also from the surrounding microenvironment or immune cells in response to tumor presence [ 36 ]. Furthermore, miRNAs released into the bloodstream are encapsulated in exosomes or bound to proteins, which may selectively reflect specific biological processes distinct from overall tissue expression [ 37 ]. These mechanisms could contribute to the observed discrepancies and suggest that serum miRNA profiles provide complementary, rather than redundant, information to tissue-based analyses. These differences in miRNA release and stability between serum and tissue may explain the inconsistencies seen in previous studies using different sample types, such as plasma, whole blood, or tumor tissues [ 17 , 18 ]. In our study, miR-6134 was detected by ISH in approximately 69% of clear cell and serous carcinoma cases, whereas tissue qRT-PCR did not demonstrate significant upregulation compared with benign controls. This discrepancy highlights the methodological limitations of comparing ISH on FFPE sections with qRT-PCR on frozen tissues. Moreover, since ISH was performed only on malignant specimens, the absence of data from normal and benign tissues restricts the interpretation of its clinical significance. Importantly, the purpose of ISH in this study was to verify the intracellular localization of miR-6134 rather than to evaluate its diagnostic accuracy. Future studies incorporating normal and benign tissues into ISH analysis will be essential to better define the tissue-level relevance of this marker. This study has some strengths. First, while most miRNA studies have focused on tissue-based analysis, we conducted microarray-based screening using serum from our institution and identified miR-200c-3p and miR-6134 through validation with real-time RT-PCR. Second, the inclusion of samples from normal controls, patients with benign tumors, and those with EOC allowed us to construct comprehensive ROC curves. Third, we were able to examine miRNA expression in both serum and matched tumor tissues from the same patients. However, several limitations should be acknowledged. First, this study used overlapping samples in both the screening and validation phases due to the limited availability of pre-treatment clinical specimens from patients with histologically confirmed EOC and benign tumors. Specifically, the microarray screening utilized pooled RNA samples from 64 EOC and 40 normal controls, and these were partially included in the 250-case validation cohort. This overlap may introduce a degree of overfitting; however, our primary objective was to identify promising miRNA candidates, not to establish a definitive predictive model. To minimize potential bias, we employed a multi-step validation strategy including RT-PCR quantification, ROC analysis, and AIC-based model selection. Nonetheless, further validation in independent, multi-institutional cohorts is essential to confirm the robustness and clinical applicability of our findings [ 38 , 39 ]. Additionally, HE4 was analyzed using age 50 as a surrogate for menopausal status, which may not be accurate in all cases. Finally, our results suggest two potential clinical applications: screening in the general population and differentiation between benign and malignant tumors detected by imaging, both of which will require large-scale, multicenter evaluation.

We identified miR-6134 as a novel biomarker for EOC and demonstrated its diagnostic potential. The combination of miR-200c-3p and miR-6134, improved the accuracy of EOC detection, particularly in cases with CA125 < 35 U/mL. Our findings highlight the importance of using multiple biomarkers together to enhance diagnostic precision. Further validation in larger, multi-institutional cohorts is warranted to confirm the clinical utility of these biomarkers and support their potential application in EOC screening and differential diagnosis.

Supplementary Material 1: Table S1: Analysis of Akaike Information Criterion with CA125 and IL-6 in the serum as screening markers and their variable combinations in patients with EOC. Supplementary Material 1: Table S1: Analysis of Akaike Information Criterion with CA125 and IL-6 in the serum as screening markers and their variable combinations in patients with EOC. Supplementary Material 2: Table S2: Analysis of Akaike Information Criterion for serum miRNAs and IL-6 as their variable combinations in patients between benign tumor vs. EOC with CA125 values using a threshold of 35 units/mL. Supplementary Material 2: Table S2: Analysis of Akaike Information Criterion for serum miRNAs and IL-6 as their variable combinations in patients between benign tumor vs. EOC with CA125 values using a threshold of 35 units/mL. Supplementary Material 3: Table S3: Real-timeRT-PCR data for miR-200c-3p and miR-6134, and the ELISA data for IL-6, CA125, and HE4. Supplementary Material 3: Table S3: Real-timeRT-PCR data for miR-200c-3p and miR-6134, and the ELISA data for IL-6, CA125, and HE4. Supplementary Material 4: Fig. S1: miRNA expression in cultured cells. Supplementary Material 4: Fig. S1: miRNA expression in cultured cells.