Diagnostic and prognostic values of circulating growth differentiation factor-15 and osteopontin in uterine sarcoma

Abstract

Uterine sarcoma is an aggressive malignancy that is difficult to distinguish from benign leiomyomas pre-operatively. This study aimed to identify circulating biomarkers to improve differential diagnosis and prognostication. The serum and tissue levels of growth differentiation factor-15 (GDF15), progranulin (PGN), and osteopontin (OPN) were measured in patients with uterine sarcoma (n = 38) and leiomyoma (n = 67). The levels were correlated with diagnosis and patient survival. The serum and tissue levels of GDF15 and OPN were significantly higher in uterine sarcomas than in leiomyomas (p < 0.001). High GDF15 levels were associated with significantly poorer progression-free survival (PFS) (p < 0.001). In a multivariate analysis including established markers cancer antigen 125 (CA125) and lactate dehydrogenase (LDH), only GDF15 remained an independent predictor of PFS [hazard ratio (HR) 3.01, 95% confidence interval (CI) 1.04–8.67, p = 0.042]. The prognostic power of GDF15 was confirmed by an analysis that excluded carcinosarcomas. GDF15 and OPN are promising biomarkers for pre-operatively differentiation of uterine sarcoma from leiomyoma. Furthermore, GDF15 is a strong, independent prognostic factor for PFS in patients with uterine sarcoma and has the potential to improve diagnostic accuracy and clinical management.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-24481-y.

Introduction

Uterine sarcomas are aggressive gynecological tumors with poor prognosis. While they rarely account for 1–2% of malignant uterine tumors, their incidence tends to increase¹. These include leiomyosarcoma (LMS), high-grade or low-grade endometrial stromal sarcoma (HG- or LG-ESS), and undifferentiated uterine sarcoma (USS). Carcinosarcoma (CS) is classified as a metaplastic uterine carcinoma according to the 2009 International Federation of Gynecology and Obstetrics (FIGO) classification², although many studies have included CS as a uterine sarcoma.

HG-ESS and USS often occur after menopause, whereas LG-ESS occurs frequently before menopause and LMS occurs frequently in 50–55-year-old women. On the other hand, uterine leiomyomas are very common benign uterine tumors that occur in about one-third of women over the age of 30 years, whereas a high incidence with a diagnosis of surgically treated leiomyoma was reported in women aged 40–59 years with more than 80%³, suggesting that a reliable preoperative diagnostic tool to differentiate uterine sarcoma from leiomyoma is needed. However, it is generally difficult to distinguish uterine sarcomas from leiomyomas using only conventional blood parameters and/or imaging analyses⁴. Moreover, diagnostic and prognostic biomarkers for uterine sarcomas have not yet been established⁵. Therefore, the development of a diagnostic and prognostic method that utilizes reliable and easily available biomarkers in daily clinical practice to distinguish leiomyomas and predict the prognosis of patients with uterine sarcoma is needed.

Circulating proteins derived from blood samples are easily available, non-invasive, and relatively inexpensive to measure compared to tissue samples, imaging data, or other circulating biomarkers⁶, suggesting that circulating proteins might be a possible diagnostic or prognostic biomarker in uterine sarcoma. A review of 37 studies on circulating protein biomarkers in soft tissue sarcoma and 11 studies on uterine sarcoma reported cancer antigen 125 (CA125), lactate dehydrogenase isozyme (LDH), and growth differentiation factor-15 (GDF15) as possible diagnostic biomarkers to differentiate from leiomyoma⁷. However, further studies are needed to determine the diagnostic and prognostic utility of circulating protein biomarkers in uterine sarcomas.

We previously identified candidate circulating protein biomarkers for diagnosing uterine sarcoma by exploring public gene expression databases, including the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA). We identified candidate biomarkers of GDF15, progranulin (PGN), and osteopontin (OPN), that exhibited significantly different expression levels of the corresponding protein-coding genes in uterine sarcoma and leiomyoma. We also conducted a preliminary study using enzyme-linked immunosorbent assay (ELISA) to measure GDF15, PGN, and OPN levels in the blood samples of patients with uterine sarcomas and leiomyomas. We have shown that the serum concentrations of GDF15, PGN, and OPN are significantly higher in patients with uterine sarcomas than in those with leiomyomas, suggesting that these bioactive factors could be diagnostic biomarkers for differentiating between uterine sarcomas and leiomyomas, leading to possible prognostic biomarkers⁸.

Based on these in vitro findings, we examined in this current study the diagnostic impact, clinical relevance, and prognostic impact of GDF15, PGN, and OPN in a validation cohort of patients with uterine sarcoma. Hence, this study aimed to identify circulating biomarkers for improved differential diagnosis and prognostication and assist in the development of targeted treatment options for uterine sarcomas.

Methods

Patients and treatment

This study included 38 patients with uterine sarcoma who were treated between 2015 and 2019 at the Department of Obstetrics and Gynecology, University of Fukui, and the Department of Gynecology, National Cancer Center Hospital. This study also included 67 patients with uterine leiomyoma treated between 2015 and 2019 at the Department of Obstetrics and Gynecology, University of Fukui, which was pre-operatively suspected as uterine sarcoma according to the following imaging findings: (a) ultrasound exhibiting an enlarged tumor and the characteristic “mosaic pattern” on ultrasonic power Doppler images and (b) Magnetic Resonance Imaging (MRI) exhibiting enlarged uterine tumor of mesenchymal origin showing heterogeneous high-signal intensity on T2-weighted images and/or the characteristic “enhancement” on contrast-enhanced MRI⁹. Serum and formalin-fixed paraffin-embedded tissue samples were obtained from all patients treated at the University of Fukui and were analyzed retrospectively. Serum samples were also provided by the National Cancer Center Biobank, Japan, and analyzed retrospectively. Clinical and pathological factors were evaluated by reviewing medical charts and pathology records. Patients with histologically confirmed uterine leiomyoma and uterine sarcoma were included, and a definitive histopathological diagnosis was made by certified pathologists based on the World Health Organization (WHO) classification. The patients’ treatment included a combination of debulking surgery and adjuvant chemotherapy according to the clinical guidelines of the Japan Society of Gynecologic Oncology. Patients were followed up for at least 24 months after their first visit or until death. The study protocol was approved by the Institutional Review Board of the University of Fukui (Institutional Review Board (IRB) number: 20150028).

Serum samples and enzyme-linked immunosorbent assay

The ELISA kits for GDF15, PGN, and OPN were purchased from R&D Systems (Minneapolis, MN, USA). Blood samples were collected within two weeks before debulking surgery or tissue biopsy and centrifuged at 3000 rpm for 10 min at 4 °C to separate the serum. The serum concentrations of bioactive factors were measured according to the manufacturer’s instructions. Serum samples were diluted 1:10 for GDF15 in Colibrator Diluent RD5-20, 1:10 for PGN in Colibrator Diluent RD6-23, and 1:10 for OPN in Colibrator Diluent RD5-24 (R&D Systems (Minneapolis, MN, USA)). All analyses were performed in duplicate. Optical densities (ODs) were determined using a microtiter plate reader at 450 nm and compared to a standard curve. The blank was subtracted from the duplicate readings for each standard and sample, and the concentrations were reported in ng/mL.

Tissue samples and immunohistochemistry

Formalin-fixed, paraffin-embedded, 2.5 μm sections were obtained from the samples collected. Immunohistochemistry (IHC) staining was performed using the avidin–biotin–peroxidase complex method as previously described¹⁰. The antibodies used for IHC staining are listed in Supplementary Table 1. The intensity and distribution of GDF15, PGN, and OPN immunohistochemical staining were evaluated using a semi-quantitative method (immunoreactive score (IRS)), as described previously¹⁰. IRS was calculated as follows: IRS = signal intensity (SI) × percentage of positive cells (PP), where SI is the optical stain intensity graded as 0 = no, 1 = weak, 2 = moderate, and 3 = strong staining, and PP is the degree of positively stained cells defined as 0 = no staining, 1 ≤ 10%, 2 = 11–50%, 3 = 51–80%, and 4 ≥ 81%). Immunohistochemical staining was performed by 2 independent observers. Immunohistochemical staining was performed by 2 independent observers. Both observers were blinded to histopathological findings, clinical data, and other experimental results. Each dataset was reviewed as the consensus decision of both observers after a minimum interval of 4 weeks to avoid any decision threshold bias due to reading-order effects.

Statistical analysis

The outcome measures were progression-free survival (PFS) and overall survival (OS). Both PFS and OS were assessed from the date of debulking surgery or tissue biopsy, if surgery was not performed. Tumor progression was confirmed by either tissue biopsy or serial imaging, which showed evidence of disease progression. The Mann–Whitney U test was used to analyze the differentiation between uterine leiomyoma and sarcoma by the value of each serum marker or IRS of each protein, and the relationships between clinical characteristics and the value of serum markers. To complement the validation of the results caused by the small and imbalanced samples, a Monte Carlo simulation (N = 10,000 permutations) was employed. Correlative studies using Spearman’s rank correlation were used to determine the correlation between clinical characteristics and the value of serum markers or the value of each serum marker and the IRS of each protein. Receiver operating characteristic (ROC) curve analysis was performed to determine the optimal cutoff values for discrimination with high accuracy based on the area under the curve (AUC) for each serum marker. Precision-recall (PR) curves were constructed for each biomarker by ranking samples according to their predicted scores in descending order and calculating precision and recall at each threshold. The area under the precision-recall curve (AUPRC) was computed using the trapezoidal rule. Bootstrap resampling (N = 1000 iterations) was employed to calculate the 95% confidence intervals (CI) for the AUPRC values. PR curves are particularly appropriate for imbalanced datasets because they focus on the performance of the positive (sarcoma) class and are not influenced by a large number of true negatives. The Kaplan–Meier curve was used to assess the relationship between serum markers and PFS and OS, and the log-rank test was used to calculate statistical significance. Cox proportional hazards regression modeling was used for univariate and multivariate analyses. Significance was defined as p < 0.05 (2-sided testing). All statistical analyses were performed using SPSS version 24 (IBM Corp., Armonk, NY, USA)¹¹.

Results

Patient characteristics

The clinical information of the 67 patients with uterine myoma and 38 patients with uterine sarcoma is summarized in Table 1. The histopathological subtypes were leiomyoma (n = 67), carcinosarcoma (n = 24), leiomyosarcoma (n = 7), undifferentiated uterine sarcoma (n = 4), and low-grade endometrial stromal sarcoma (n = 3). Among uterine sarcomas, 16 patients (42.1%) had FIGO stage I or II, whereas 22 patients (57.9%) had FIGO stage III or IV. The median follow-up period was 37.8 months (range, 0.3–76.0 months). 27 patients (71.1%) showed tumor progression during the follow-up period and 18 patients (47.4%) died. Because the 2009 FIGO classification provides a new staging system, classifying carcinosarcoma as metaplastic uterine carcinoma, we also summarized 14 patients with uterine sarcoma, excluding carcinosarcoma, in Supplementary Table 2. Eight patients (57.1%) had FIGO stage I or II disease, whereas six patients (42.9%) had FIGO stage III or IV disease. The median follow-up period was 27.4 months (range, 0.3–76.0 months). Eleven patients (78.6%) had tumor progression during the follow-up period, and 7 patients (50.0%) died.

Table 1.

Patient and tumor characteristics.

Characteristic	n	%
Uterine myoma
Total number of patients	67	100
Histology
Leiomyoma	67	100
Uterine sarcoma
Total number of patients	38	100
Histology
Carcinosarcoma	24	63.2
Leiomyosarcoma	7	18.4
Undifferentiated uterine sarcoma	4	10.5
Low grade endometrial stromal sarcoma	3	7.9
FIGO stage
I, II	16	42.1
III, IV	22	57.9
Tumor progression
Absent	11	28.9
Present	27	71.1
Death
Absent	20	52.6
Present	18	47.4

FIGO: International Federation of Gynecology and Obstetrics.

Differential diagnosis between uterine myoma and sarcoma according to various clinical factors

Uterine myomas and sarcomas were correlated with various clinical factors to determine their differential diagnostic impact. The mean age in uterine sarcoma (62.47 years, 45–92) was significantly larger than that of uterine myoma (44.51 years, 24–60) (p < 0.001). There were no significant differences in the mean values of LDH (p = 0.342), CA125 (p = 0.317), and PGN (p = 0.821) between the uterine myomas and sarcomas. The mean value of GDF15 in uterine sarcoma (1.42 ng/ml, 0.26–5.17) was significantly larger than that of uterine myoma (0.44 ng/ml, 0.08–1.38) (p < 0.001). The mean value of OPN in uterine sarcoma (22.04 ng/ml, 3.16–82.24) was also significantly larger than that in uterine myoma (7.58 ng/ml, 2.65–21.74) (p < 0.001) (Table 2). Using Fisher’s test to determine optimal cutoffs, ROC and PR curve analyses were used to test and compare the performance of the serum markers under investigation. With a cut-off of 0.45, GDF15 could significantly differentiate between uterine myoma and sarcoma (AUC 0.883, 95% CI 0.812–0.953). With a cut-off of 10.61, OPN could significantly differentiate between uterine myoma and sarcoma (AUC 0.881, 95% CI 0.810–0.951). No significant differences in LDH, CA125, or PGN levels were observed. (Fig. 1A). GDF15 and OPN also demonstrated superior diagnostic performance with AUPRC values of 0.888 and 0.870, respectively, representing 2.1-fold and 2.0-fold improvements over the baseline random classifier (prevalence = 0.414). (Fig. 1B).

Table 2.

Differential diagnosis between uterine myoma and sarcoma according to various clinical factors.

	Uterine myoma	Uterine sarcoma
Number of patients	67	38
Age (years)
Means (range)	44.51 (24–60)	62.47 (45–92)	p < 0.001*
LDH (U/L)
Means (range)	217.84 (140.00–713.00)	285.71 (127.00–1185.00)	p = 0.346
CA125 (U/mL)
Means (range)	44.85 (8.20–492.60)	62.29 (8.00–364.00)	p = 0.319
GDF15 (ng/mL)
Means (range)	0.44 (0.08–1.38)	1.42 (0.26–5.17)	p < 0.001*
PGN (ng/mL)
Means (range)	49.52 (29.47–76.75)	59.75 (31.92–362.53)	p = 0.824
OPN (ng/mL)
Means (range)	7.58 (2.65–21.74)	22.04 (3.16–82.24)	p < 0.001*

LDH: lactate dehydrogenase isozyme; CA125: cancer antigen 125; GDF15: growth differentiation factor-15; PGN: progranulin; OPN: osteopontin.

Fig. 1.

Receiver operating characteristic curves (A a–e) and Precision-Recall curves (B a-e) analyses for differentiating diagnosis between uterine myoma and sarcoma according to serum markers of LDH, CA125, GDF15, PGN, and OPN. (A) a) AUC is 0.556 (p = 0.342, 95% CI 0.433–0.679), with 229.50 determined as the optimal cut-off for LDH. b) AUC is 0.563 (p = 0.317, 95% CI 0.438–0.689), with 52.10 determined as the optimal cut-off for CA125. c) AUC is 0.883 (p < 0.001, 95% CI 0.812–0.953), with 0.45 determined as the optimal cut-off for GDF15. d) AUC is 0.513 (p = 0.821, 95% CI 0.396–0.631) with 73.56 determined as the optimal cut-off for PGN. e) AUC is 0.881 (p < 0.001, 95% CI 0.810–0.951), with 10.61 determined as the optimal cut-off for OPN. (B) The red dashed line represents the baseline prevalence (random classifier performance, 41.4%). (a) AUPRC is 0.590 (95% CI: 0.437–0.741) for LDH. (b) AUPRC is 0.502 (95% CI 0.360–0.687) for CA125. (c) AUPRC is 0.888 (95% CI 0.793–0.948) for GDF15. (d) AUPRC is 0.531 (95% CI 0.369–0.675) for PGN. (e) AUPRC is 0.870 (95% CI 0.768–0.944) for OPN.

Uterine myomas and sarcomas, excluding carcinosarcomas, also correlated. The mean age in uterine sarcoma (61.71 years, 45–92) was significantly larger than that of uterine myoma (44.51 years, 24–60) (p < 0.001). Regarding serum markers, no significant differences in the mean values of CA125 (p = 0.175) and PGN (p = 0.144) were observed between the uterine myomas and sarcomas. The mean LDH value in uterine sarcoma (423.57 U/L, 174.00–1185.0) was significantly larger than that in uterine myoma (217.84 U/L, 140.00–713.00) (p = 0.002). The mean value of GDF15 in uterine sarcoma (2.02 ng/ml, 0.26–5.17) was significantly larger than that in uterine myoma (0.44 ng/ml, 0.08–1.38) (p < 0.001). The mean value of OPN in uterine sarcoma (28.82 ng/ml, 3.16–82.24) was also significantly larger than that in uterine myoma (7.58 ng/ml, 2.65–21.74) (p < 0.001) (Supplementary Table 3). In ROC curve analysis, with a cut-off of 230.50, LDH could significantly differentiate between uterine myoma and sarcoma (AUC 0.770, 95% CI 0.614–0.927). With a cut-off of 0.89, GDF15 could significantly differentiate between uterine myoma and sarcoma (AUC 0.891, 95% CI 0.771–1.000). With a cut-off of 11.10, OPN could significantly differentiate between uterine myoma and sarcoma (AUC 0.861, 95% CI 0.722–1.000). No significant differences in CA125 and PGN levels were observed. (Supplementary Fig. 1A). In the PR curve analysis, GDF15 and OPN also demonstrated superior diagnostic performance, with AUPRC values of 0.892 and 0.813, respectively, representing 5.2-fold and 4.7-fold improvements over the baseline random classifier (prevalence = 0.173). (Supplementary Fig. 1B).

Value of LDH, CA125, GDF15, PGN, and OPN in relation to clinical factors of patients with uterine sarcoma

Various clinical factors of uterine sarcomas were correlated with LDH, CA125, GDF15, PGN, and OPN levels. Significant correlations were identified between high LDH levels and carcinosarcoma (p = 0.001). Significant correlations were also identified between high CA125 levels and FIGO stages III-IV (p < 0.001). No significant correlations were observed between GDF15, PGN, or OPN and clinical parameters (Supplementary Table 4). Uterine sarcomas, excluding carcinosarcomas, also correlated. Significant correlations were identified between high LDH and CA125 levels and FIGO stages III-IV (p = 0.029 and p = 0.005, respectively). No significant correlations were observed between GDF15, PGN, or OPN levels and the clinical parameters (Supplementary Table 5). The age of the patients and GDF15 or OPN levels were analyzed using Spearman correlative studies to confirm the correlations between them. The correlation between age and GDF15 was y = 57.83 + 3.26x, R² = 0.104, and was not significant (p = 0.181) in the uterine sarcoma cohort, whereas the correlation between them was y = 41.23 + 7.5x, R² = 0.061, and was significant (p = 0.010*). The correlation between age and OPN was y = 61.99 + 0.02x, R² = 9.507E − 4, and was not significant (p = 0.406) in the uterine sarcoma cohort. The correlation between them was y = 46.75 + 0.3x, R² = 0.098, and was not significant (p = 0.098) (Supplementary Fig. 2).

Clinical performances of LDH, CA125, GDF15, PGN, and OPN in patients with uterine sarcoma

Using Fisher’s test to determine optimal cutoffs, ROC curve analysis was used to test and compare the performance of the serum markers under investigation. LDH could significantly detect tumor progression with a cut-off of 225.50 (AUC 0.731, 95% CI 0.574–0.887) (p = 0.027). CA125 also significantly detected tumor progression with a cut-off of 18.20 (AUC 0.817, 95% CI 0.659–0.974) (p = 0.005). No significant differences were observed in GDF15 (p = 0.296), PGN (p = 0.311), and OPN (p = 0.079) (Supplementary Fig. 3A). In terms of OS, CA125 was significantly detected with a cut-off of 29.00 (AUC 0.722, 95% CI 0.551–0.893) (p = 0.023). No significant differences in LDH (p = 0.057), GDF15 (p = 0.108), PGN (p = 0.726), or OPN (p = 0.193) were observed (Supplementary Fig. 4A).

The same analysis was used for uterine sarcomas, excluding carcinosarcomas. LDH could significantly detect tumor progression with a cut-off of 202.50 (AUC 1.000, 95% CI 1.000–1.000) (p = 0.010). GDF15 could significantly detect tumor progression with a cut-off of 1.09 (AUC 0.939, 95% CI 0.801–1.000) (p = 0.024). PGN could significantly detect tumor progression with a cut-off of 45.86 (AUC 0.909, 95% CI 0.739–1.000) (p = 0.036). OPN was significantly associated with tumor progression (AUC 0.939, 95% CI 0.801–1.000) (p = 0.024). No significant differences were observed in the CA125 levels (p = 0.052) (Supplementary Fig. 3B). In terms of OS, GDF15 was significantly detected with a cut-off of 1.39 (AUC 0.857, 95% CI 0.658–1.000) (p = 0.025). No significant differences in LDH (p = 0.180), CA125 (p = 0.225), PGN (p = 0.749), or OPN (p = 0.064) were observed (Supplementary Fig. 4B).

Prognostic effect of LDH, CA125, GDF15, PGN, and OPN in patients with uterine sarcoma

Kaplan–Meier survival curves showed that patients with high levels of LDH, CA125, and GDF15 had significantly poor PFS (p = 0.006, p = 0.007, and p < 0.001, respectively) than patients with low levels, but not PGN and OPN (p = 0.143 and p = 0.078, respectively) (Fig. 2A). Moreover, patients with high levels of LDH, CA125, and GDF15 showed significantly poorer OS (p = 0.001, p = 0.011, and p = 0.032, respectively) than those with low value, but not PGN, and OPN (p = 0.119 and p = 0.079, respectively) (Supplementary Fig. 5A). The same analysis was used for uterine sarcomas, excluding carcinosarcomas. Kaplan–Meier survival curves showed that patients with high LDH, CA125, GDF15, PGN, and OPN levels showed significantly poor PFS (p = 0.005, p = 0.029, p = 0.006, p = 0.011, and p = 0.016, respectively) than patients with low value (Fig. 2B). Moreover, patients with high levels of LDH, CA125, GDF15, and OPN showed significantly poorer OS (p = 0.033, p = 0.019, p = 0.014, and p = 0.013, respectively) than those with low levels, but not PGN (p = 0.064) (Supplementary Fig. 5B).

Fig. 2.

Kaplan–Meier survival curves for progression-free survival rates among patients with (A a-e) or without (B a-e) carcinosarcoma according to the serum markers of LDH, CA125, GDF15, PGN, and OPN. (A) (a) PFS in patients with high value of LDH (≥ 225.50, solid line) and low LDH levels (< 225.50, dotted line). Patients with high LDH levels showed poorer PFS (p = 0.006) than patients with low LDH levels. (b) PFS in patients with high value of CA125 (≥ 18.20, solid line) and low value of CA125 (< 18.20, dotted line). Patients with high CA125 levels showed poorer PFS (p = 0.007) than patients with low CA125 levels. (c) PFS in patients with high value of GDF15 (≥ 2.22, solid line) and low value of GDF15 (< 2.22, dotted line). Patients with high GDF15 levels showed poorer PFS (p < 0.001) than patients with low GDF15 levels. (d) PFS in patients with high PGN levels (≥ 52.85, solid line) and low PGN levels (< 52.85, dotted line). No difference in PFS was apparent according to PGN value (p = 0.143). (e) PFS in patients with high (≥ 15.38, solid line) and low (< 15.38, dotted line) OPN values. Patients with high OPN levels showed poorer PFS (p = 0.078) than patients with low OPN levels. (B) (a) PFS in patients with high LDH levels (≥ 202.50, solid line) and low LDH levels (< 202.50, dotted line). Patients with high LDH levels showed poorer PFS (p = 0.005) than patients with low LDH levels. (b) PFS in patients with high value of CA125 (≥ 24.85, solid line) and low value of CA125 (< 24.85, dotted line). Patients with high CA125 levels showed poorer PFS (p = 0.029) than patients with low CA125 levels. (c) PFS in patients with high value of GDF15 (≥ 1.09, solid line) and low value of GDF15 (< 1.09, dotted line). Patients with high GDF15 levels showed poorer PFS (p = 0.006) than patients with low GDF15 levels. (d) PFS in patients with high PGN levels (≥ 45.86, solid line) and low PGN levels (< 45.86, dotted line). Patients with high PGN levels showed poorer PFS (p = 0.011) than patients with low PGN levels. (e) PFS in patients with high (≥ 14.24, solid line) and low (< 14.24, dotted line) OPN. Patients with high OPN levels showed poorer PFS (p = 0.016) than patients with low OPN levels.

Univariate analysis showed that high levels of LDH, CA125, and GDF15 were significantly associated with poor PFS (p = 0.008, 0.011, and < 0.001, respectively). FIGO stages III-IV were also significantly associated with a poor PFS (p = 0.014). In the multivariate analysis, GDF15 was significantly associated with poor PFS (p = 0.042) and was an independent prognostic factor for PFS (Table 3). Univariate analysis showed that high LDH and CA125 levels were significantly associated with a poor OS (p = 0.003 and 0.017, respectively). FIGO stages III-IV were also significantly associated with a poor OS (p = 0.032). In multivariate analysis, LDH was significantly associated with poor OS (p = 0.007) and was an independent prognostic factor for OS (Supplementary Table 6). The same analysis was used for uterine sarcomas, excluding carcinosarcomas. Univariate analysis showed that high levels of CA125, GDF15, PGN, and OPN were significantly associated with poor PFS (p = 0.046, 0.024, 0.033, and 0.030, respectively). FIGO stages III–IV were also significantly associated with a poor PFS (p = 0.036). In multivariate analysis, GDF15 was significantly associated with poor PFS (p = 0.045) and was an independent prognostic factor for PFS (Table 3). Univariate analysis showed that high levels of CA125, and GDF15 were significantly associated with poor OS (p = 0.036 and 0.030, respectively). In multivariate analysis, no independent prognostic factors for OS were identified (Supplementary Table 7).

Table 3.

Prognostic factors for progression-free survival of uterine sarcoma with or without carcinosarcoma of uterine sarcoma selected by Cox’s uni- and multivariate analysis.

Uterine sarcoma with carcinosarcoma					Uterine sarcoma without carcinosarcoma
Variables	Univariate analysis		Multivariate analysis		Variables	Univariate analysis		Multivariate analysis
	Hazard ratio (95% CI)	p	Hazard ratio (95% CI)	p		Hazard ratio (95% CI)	p	Hazard ratio (95% CI)	p
Age (≥ 55)	1.35 (0.59–3.10)	0.473			Age (≥ 55)	2.37 (0.62–9.04)	0.206
FIGO stage (III-IV)	2.89 (1.25–6.70)	0.014*	1.72 (0.63–4.69)	0.288	FIGO stage (III-IV)	4.03 (1.09–14.87)	0.036*	0.62 (0.06–6.05)	0.683
Histology (CS)	1.27 (0.59–2.75)	0.540
LDH (≥ 225.0)	2.82 (1.31–6.08)	0.008*	1.71 (0.67–4.35)	0.258	LDH (≥ 202.5)	60.70 (0.34–11,007.99)	0.122
CA125 (≥ 18.20)	3.56 (1.33–9.51)	0.011*	1.73 (0.51–5.95)	0.382	CA125 (≥ 24.85)	5.06 (1.03–24.88)	0.046*	0.27 (0.01–9.22)	0.468
GDF15 (≥ 2.22)	6.48 (2.48–16.94)	< 0.001*	3.01 (1.04–8.67)	0.042*	GDF15 (≥ 1.09)	11.39 (1.38–93.90)	0.024*	32.92 (1.08–1004.14)	0.045*
PGN (≥ 52.85)	1.75 (0.82–3.76)	0.148			PGN (≥ 45.86)	9.77 (1.21–79.25)	0.033*	4.82 (0.41–56.69)	0.211
OPN (≥ 15.38)	2.04 (0.91–4.56)	0.084			OPN (≥ 14.24)	5.64 (1.18–26.89)	0.030*	3.59 (0.52–24.70)	0.194

FIGO: International Federation of Gynecology and Obstetrics; LDH: lactate dehydrogenase isozyme; CA125: cancer-antigen 125; GDF15: growth differentiation factor-15; PGN: progranulin; OPN: osteopontin; CI: confidence interval.

Differential diagnosis between uterine myoma and sarcoma according to immunoreactive score of each protein

To determine the differential diagnostic impact of GDF15, PGN, and OPN in the tumor tissues of uterine myomas and sarcomas, their IRSs were calculated after IHC staining. Their expression was mainly observed in sarcomatous lesions compared to healthy uterine smooth muscle tissues. The mean IRS of GDF15 in uterine sarcoma (3.13 (range, 1–6)) was significantly higher than that in uterine myoma (0.12 (range, 0–1)) (p < 0.001). The mean IRS of PGN in uterine sarcoma (8.13 (range, 1–12)) was significantly higher than that in uterine myoma (0.65 (range, 0–3)) (p < 0.001). The mean IRS of OPN in uterine sarcoma (10.13 (range, 4–12)) was significantly higher than that in uterine myoma (1.65 (range, 0–4)) (p < 0.001) (Supplementary Table 8). Representative images of hematoxylin and eosin (HE) and IHC staining are shown in Fig. 3A. To evaluate the correlation between expression in tumor tissue and serum, the IRSs in IHC and the values in ELISA assays for each protein obtained from the same cohort were analyzed using Spearman correlative studies. The correlation for GDF15 was y = 0.34 + 0.74x, R² = 0.374, and was significant (p = 0.002). The correlation for PGN was y = 0.85 + 0.03x, R² = 0.222, and not significantly correlated (p = 0.107). The correlation for OPN was y = 1.86 + 0.15x, R² = 0.544, and was significant (p = 0.002) (Fig. 3B).

Fig. 3.

(A) Representative uterine leiomyoma (upper) and sarcoma (bottom) showing immunostaining for HE (a), GDF15 (b), PGN (c) and OPN (d) (magnification, × 200). Scale bar is 50 μm. (B) Correlation between immunoreactive score and ELISA assays for GDF15 (a), PGN (b), and OPN (c). a) The correlation for GDF15 was y = 0.34 + 0.74x, R² = 0.374 (p = 0.002). b) The correlation for PGN was y = 0.85 + 0.03x, R² = 0.222 (p = 0.107). c) The correlation for OPN was y = 1.86 + 0.15x, R² = 0.544 (p = 0.002).

Discussion

In the present study, we demonstrated that GDF15 and OPN are promising biomarkers for the pre-operatively differentiation of uterine sarcomas from leiomyomas, as evidenced by their elevated serum levels and high tissue expression. Furthermore, GDF15 is a strong independent prognostic factor for PFS in patients with uterine sarcoma. Combined with our previous preliminary studies, GDF15 could serve as a potential target protein for therapeutic purposes and a possible diagnostic and prognostic biomarker.

Uterine sarcomas are aggressive tumors with poor prognosis, and their rarity and atypical appearance often cause diagnostic difficulties. Minimally invasive surgery, such as laparoscopic or robot-assisted surgery, is standard for benign leiomyomas but can be catastrophic if an unsuspected uterine sarcoma is present, as the use of morcellation devices can lead to widespread dissemination of malignant cells⁴. Therefore, effective preoperative triage using blood parameters and/or imaging analyses is critical. While markers such as LDH and CA125 have been reported to be potentially useful, their diagnostic value is often limited, although it may be improved in combination with other markers or imaging^7,12. Advanced imaging modalities such as MRI and positron emission tomography (PET) have also been explored. MRI can offer high accuracy for detecting uterine sarcomas, but differentiation from atypical leiomyomas remains a persistent challenge¹³. Similarly, ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET can be confounded by high uptake in benign leiomyomas, influenced by factors such as menstrual cycle¹⁴. Novel radiotracers, such as 16α-[¹⁸F]-fluoro-17β-estradiol (¹⁸F-FES), are in development but are not yet used in routine clinical practice¹⁵. This highlights the pressing need for reliable and readily available biomarkers.

Our study focused on GDF15, PGN, and OPN, which have all been reported to be intricately involved in cancer progression, including the promotion of cell proliferation, migration, invasion, and metastasis^16–21. In the current study, serum GDF15 and OPN levels were significantly higher in uterine sarcomas than in myomas. This is the first report of OPN as a differential diagnostic biomarker for this malignancy, and is consistent with a previous study on GDF15²². When we excluded carcinosarcoma cases, LDH also became a significant differentiator, which aligns with previous research on leiomyosarcoma¹².

An important finding was that GDF15 and OPN levels did not correlate with any clinicopathological status, including FIGO stage or histology. This suggests that they could be valuable differential diagnostic biomarkers, even in the early stages of the disease, which is often the most challenging clinical scenario. Moreover, GDF15 and OPN levels were also not correlated with age, although they can be one of the most upregulated proteins with age and have a protective role against inflammatory damage and other stresses^23,24, suggesting that both GDF15 and OPN have potential utility as diagnostic biomarkers independent of the influence of aging. In contrast, the utility of CA125 may be limited by physiological variations, such as the menstrual cycle in premenopausal women, which could explain the lack of a significant difference in our cohort²⁵.

In terms of patient outcome, our multivariate analysis revealed that GDF15 was an independent prognostic factor for PFS in uterine sarcoma, both with and without carcinosarcoma. In contrast, the LDH level was an independent prognostic factor for OS in the full sarcoma cohort. The prognostic significance of PFS is particularly noteworthy as it provides insight into the time frame for tumor recurrence and the development of treatment resistance. A reliable biomarker for PFS can be more useful than one for OS in guiding treatment choices in daily clinical practice and developing novel therapeutic agents^26,27. Our findings on GDF15’s prognostic role are partly consistent with those of Trovik et al.²², although they also reported correlations between GDF15 and factors like advanced stage and tumor diameter, which we did not observe. This discrepancy could be due to differences in cohort composition or the evolution of treatments over the last decade, which may have improved the outcomes for patients with advanced disease. The inclusion of carcinosarcoma in our study is a notable distinction, and the prognostic power of GDF15 in this mixed cohort suggests that it may be secreted by both the carcinomatous and sarcomatous components of these tumors.

At the tissue level, all three proteins, GDF15, PGN, and OPN, were significantly overexpressed in sarcomas compared to myomas, suggesting their potential as tissue-based diagnostic markers. However, only the IRSs of GDF15 and OPN were significantly correlated with the serum levels. This finding is consistent with their superior performance as serum-based diagnostic markers in our study. Our results are consistent with those of Matsumura et al., who suggested that PGN expression in tumor tissues, but not necessarily in serum, might be a potential diagnostic and prognostic biomarker²⁸. Our study is the first to report the potential value of OPN as a diagnostic biomarker for uterine sarcoma in both serum and tissue. Although the prognostic value of GDF15 in uterine sarcoma tissue has not been previously reported, its association with poor survival in other cancers, such as gastric and colorectal cancers, suggests that this is a promising avenue for future investigation^29,30.

The biological role of GDF15 in cancer progression is well documented. It has been implicated in suppressing apoptosis, promoting epithelial-to-mesenchymal transition, and driving chemoresistance, angiogenesis, and metastasis through various signaling pathways³¹. It has also been reported that GDF15 produced by tumor-associated macrophages can promote programmed death ligand 1 (PD-L1) expression, facilitating immune evasion³². These mechanistic actions provide a strong biological rationale for GDF15 as a key player in sarcoma progression and as a valid therapeutic target.

Although the findings of this study are promising, we acknowledge some limitations. First, it was retrospective, although we used a validation cohort. Second, owing to the rarity of the disease, the sample size was small, particularly for the subgroup of uterine sarcomas excluding carcinosarcoma, which limits the statistical power of these specific analyses. Therefore, we analyzed the diagnostic accuracy using a resampling method to complement the validation of the results caused by the small and imbalanced sample and explained that the small and imbalanced samples minimally affected the results.

Conclusion

To our knowledge, this is the first study to provide new insights into the clinical diagnostic impact of GDF15 and OPN in patients with uterine sarcoma. GDF15 has also emerged as an independent prognostic factor for PFS. These findings serve as the first report to explain the role of GDF15 and OPN in a comprehensive patient cohort using ELISA and IHC, suggesting that they provide important insights into the development of novel diagnostic and prognostic markers, and targeted therapies for patients with uterine sarcoma.

Abbreviations

LMS

Leiomyosarcoma

HG- or LG-ESS

High-grade or low-grade endometrial stromal sarcoma

USS

Undifferentiated uterine sarcoma

CS

Carcinosarcoma

FIGO

International Federation of Gynecology and Obstetrics

CA125

Cancer-antigen 125

LDH

Lactate dehydrogenase isozyme

GDF15

Growth differentiation factor-15

GEO

Gene Expression Omnibus

TCGA

The Cancer Genome Atlas

PGN

Progranulin

OPN

Osteopontin

ELISA

Enzyme-linked immunosorbent assay

MRI

Magnetic resonance imaging

WHO

World Health Organization

IRB

Institutional Review Board

OD

Optical densities

IHC

Immunohistochemistry

IRS

Immunoreactive score

SI

Signal intensity

PP

Percentage of positive cells

PFS

Progression-free survival

OS

Overall survival

ROC

Receiver operating characteristic

AUC

Area under the curve

PR

Precision-Recall

AUPRC

Area under the precision-recall curve

CI

Confidence interval

HE

Hematoxylin and eosin

PET

Positron emission tomography

¹⁸F-FDG

¹⁸F-fluorodeoxyglucose

¹⁸F-FES

16α-[¹⁸F]-fluoro-17β-estradiol

PD-L1

Programmed death ligand 1

Author contributions

HT, TM, and MO designed experiments. HT, TM, and MO were involved in the acquisition, analysis, and interpretation of data. HT wrote the manuscript. MU, TK, and YY proofread and revised the manuscript. YY supervised study. All the authors have read and approved the final manuscript.

Funding

This study was supported by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (C-23K08862) (HT). The funders had no role in the design of the study; collection, analysis, and interpretation of data; or in writing the manuscript.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study protocol was approved by the Institutional Review Board of the University of Fukui Hospital (IRB Number: 20150028). First, it was a retrospective study. The patients provided written informed consent to participate in the study and anonymous clinical data were used. Patients provided written consent for the use of residual samples and were offered an opt-out option as disclosed on the website (https://www.u-fukui.ac.jp/cont_about/disclosure/privacy/).Registry and Registration No is N/A. Animal Studies is N/A. All methods were performed in accordance with approved guidelines and regulations.