Elevated Growth Differentiation Factor 15 Expression Predict Poor Prognosis in Epithelial

1

Elevated Growth Differentiation Factor 15 Expression Predict Poor Prognosis in Epithelial Ovarian Cancer Patients

Ying Zhang1,2*, Wei Hua3*, Li-chun Niu2, Shi-mei Li2, Ying-mei Wang3, Lei Shang4, Cun Zhang1, Wei-na Li1, Rui Wang4, Bi-liang Chen5, Xiao-yan Xin5, Ying-qi Zhang1#, Jian Wang5#

1 The State Key Laboratory of Cancer Biology, Biotechnology Center, School of Pharmacy, the Fourth Military Medical University, Xi’an, Shaanxi 710032, China;

2 Department of Gynecology and Obstetrics, the People's Liberation Army 323 Hospital, Xi’an, Shaanxi 710045, China;

3State Key Laboratory of Tumor Biology,DepartmentofPathology,XijingHospital, The Fourth Military Medical University, Xi’an, Shaanxi 710033, China;

4 Department of Health Service, School of Public Health, Fourth Military Medical University, Xi’an, Shaanxi 710033, China;

5Department of Gynecology and Obstetrics, Xijing Hospital, Fourth Military Medical University, Xi’an, Shaanxi 710033, China;

* Ying Zhangand Wei Hua contributed equally to this work.

#Correspondence to: Dr. Jian Wang, Department of Gynecology and Obstetrics, Xijing Hospital, Fourth Military Medical University, 169 Changle West Road, 710033 Xi’an, Shaanxi, China;

Phone: +86-29-84775387, Email address:

#Correspondence to: Ph.D. Ying-qi Zhang, Ph.D, The State Key Laboratory of Cancer Biology, Biotechnology Center, School of Pharmacy, the Fourth Military Medical University, 17 Changle West Road, 710032 Xi’an, Shaanxi, China

Phone: +86-29-84774773, Email:

Running title: Elevated GDF15 Expression Predict Poor Prognosis in EOCs

Word count: Abstract: 172; Text: 4040.

ABSTRACT

Objective: The purpose of this study was to determine the expression of growth differentiation factor 15(GDF15), and explore its clinical significance inepithelialovariancancer (EOC) patients.Methods: The expression of GDF15 in EOC tissues and serum sampleswas evaluated using immunohistochemistry and enzyme-linked immunosorbent assay (ELISA) respectively. The association of GDF15 expression with clinicopathologic parameters was analyzed.Survival time was assessed using the Kaplan–Meier technique and Cox regression model.Results: Both in EOC tissues and serum, high GDF15 levels were obviously related with advanced FIGO stage, lymph node metastasis,ascites, and chemoresistance.Kaplan–Meier analysis indicated thatEOC patients with high GDF15 expression showed poorer progression-free survival (PFS) and overall survival (OS). Multivariate analysis demonstrated that GDF15 expression was an independent predictor of PFS in EOC patients. Conclusion: Our study shows that elevated GDF15 expression was associated with poor prognosis in EOC patients. We suggest that GDF15 is a novel biomarker for the early detection of EOC, prediction of the response to chemotherapy, and screening for recurrence in EOC patients.

Keywords:growth differentiation factor 15, epithelial ovarian cancer, biomarker, chemotherapy, prognosis

INTRODUCTION

Epithelial ovarian cancer (EOC)accounts for more than 80% of all malignant cancers of the female reproductive system, and is the leading cause of death from gynecological malignancies[1]. Owing to the lack of effective screening methods and specific symptoms early in the disease, over 70% of patientsare diagnosed at an advanced stage,and thus, the prognosis of EOC patients remains poor,with a 5-year overallsurvival (OS) rate of less than 25%[2, 3]. Serum cancer antigen125 (CA125)levelsare widely used to distinguish malignant from benign pelvic masses, monitor the response to cytoreduction and chemotherapy, and screen for disease recurrence in EOC patients[4-7]. However, the sensitivity and specificity of this serum test are low[4, 5, 8]. Therefore, novel, clinically effective biomarkers that can sufficiently predict the prognosis of EOC patients and identify platinum-resistant EOCshold great promise to improve the therapeutic effectsin patients with ovarian cancer.

Growth differentiation factor 15 (GDF15) is a secreted protein of the transforming growth factor-β (TGF-β) superfamily. GDF15 plays multiple roles in various pathologies, including inflammation, cancer, cardiovascular diseases, and obesity[9-11]. GDF15 is weakly expressed in most tissues, but is highly overexpressedunder pathologic conditions such as injury, inflammation, and various cancers[10, 11].Although GDF15 has beenreported to have both tumorigenic and anti-tumorigenic activities,considerable evidence indicates that GDF15 plays animportant role in carcinogenesis-related activities, such as proliferation, migration, apoptosis, and angiogenesis[12-15]. Serum GDF15 levels are markedly increased in patients with pancreatic,prostate, and colorectal cancers[12, 13, 16, 17]. All the above evidence indicates that GDF15 may bea useful biomarker for solid tumor detection. However, the role of GDF15 in the development and progression of EOC is largely unknown. To determine whether GDF15 can serve as a powerful diagnostic and prognostic factor in EOC, we evaluated the expression of GDF15 in EOC tissues and analyzed its association with clinicopathologic data. In addition, enzyme-linked immunosorbent assay (ELISA) was used to determine serum GDF15 levels in EOC patients and healthy controls.

MATERIALS AND METHODS

Tumor tissues

Between January 2010 and December 2013, we enrolled 145 patientswho had been diagnosed with primary EOCat the Department of Obstetrics and Gynecology, Xijing Hospital, Xi’an, China, and had available untreated tissue specimens. The eligibilitycriteria for this study included the following: (1)histologically provenEOC; (2)availability of clinical data and resected tissue; (3) postoperative treatment with standard platinum-basedadjuvant chemotherapy (cisplatin/paclitaxel or cisplatin/cyclophosphamide/doxorubicin); and (4)follow-up from the time of surgical intervention to 2014.The exclusioncriteria for this study were as follows: (1) histologictypesother than EOC; (2) preoperative radiation or chemotherapy; and (3) patientswhose cause of death was unknown. The studywas approved by the ethicscommittee of Xijing Hospital, and the patients provided informed consent before their inclusion into the study. Informed consent for the use of the tumor specimens was obtained either from the patients or from their next of kin.All patients underwent follow-up gynecological examinations, transvaginal/abdominopelvicultrasonography, radiological investigations, and serum CA125 measurements. All patients were followed up from the time of surgical intervention to 2014.

Tissue specimens were harvested intraoperatively, formalin-fixed, and paraffin-embedded. Each EOC sample was cut into 4-μm sections.Onesection was stained with hematoxylin–eosin (H&E) and used for morphological diagnosis, while the others were used for immunohistochemical analysis. The EOC samples were histologically classified according to the International Federation of Gynecology and Obstetrics (FIGO) criteria[18]. The histologic type and grade were classified according to the World Health Organization criteria[19]. All histologic specimens were analyzed by a single operator.

Blood samples

Bloodsamples were obtained before surgery and any medical treatment from 120 EOC patientsand 40 healthy controls (age, 46.3 ± 12.4 years) with no history of ovarian pathology or other systemic disease. Blood samples from the patients and controls were drawn into serum tubes and centrifuged at 1000g for 10 min. All serum samples were stored at-80°C until use.

Immunohistochemistry

A standard streptavidin–biotin complex method was used. Negative control slides for GDF15 antibody were prepared using non-specific mouse IgG, and breast cancer specimens were used as positive controls. Tissue specimens were de-waxed with xylene and gradually hydrated. After being blockedwith endogenous peroxidase and 3% H2O2–methanol for 10 min, the slides were incubated overnight at 4°C with the primary mouse monoclonal anti-GDF15 antibodies (Novus Biologicals, Littleton, CO) at a dilution of 1:800. After being washed three times for 5 min each with phosphate-buffered saline, the sections were incubated with a secondary antibody (Pierce, Rockford, IL) for 30 min followed by incubation with the avidin–biotin complex for a further 30 min. 3-3ʹ-Diaminobenzidine tetrahydochloride was used as a chromogen. All sections were counterstained with Gill’s hematoxylin.A pathologist then reviewed the immunohistochemical preparations in parallel with their corresponding H&E-stainedslides to confirm the diagnosis.

Evaluation of immune staining

The tumor cores were evaluated by specialist pathologists and oncologists blinded to the clinicopathologic characteristics of the patients. The scale of staining was semiquantitatively evaluated according to the percentage of stained cells and the staining intensity as previously described[20]. A brown precipitate observed on tissue sections indicated positive immunoreactivity with the primary GDF15 antibody. Whole-field inspection of the core was included in the assessment, and the immune staining was scored as follows: (1) the proportion of malignant cells positively stained with the anti-GDF15 antibody was scored as 0 (0%–4%), 1 (5%–24%), 2 (25%–49%), 3 (50%–74%), or 4 (75%–100%);(2) the intensity of immunostaining was graded as 0 (negative), 1+ (weak), 2+(moderate), or 3+ (strong);and (3) the two scoreswere multipliedto obtain the final score. Final scores of 0–4 indicated low expression, scores of 5–8 indicated moderate expression, and scores of 9–12 indicated high expression.

ELISA

ELISAwas used to measure serum GDF15 levels, according to the manufacturer’s instructions(GDF15 ELISA Kit; USCNLIFE, Wukan, China). Serum CA125 levels were determined at the Xijing hospital laboratory by using an immunoenzymometric assay and an immunoelectrochemiluminescence detection technique with a CA125 II ECLIA (electrochemiluminescence immunoassay) kit and Roche/Hitachi Modular Analytics E170 (Roche Diagnostics GmbH, Mannheim, Germany).

Statistical analysis

Statistical analysis was performed using SPSS 17.0 for Windows (SPSS Inc., Chicago, IL).Associations between GDF15 expression in EOC tissues and clinicopathologic variables were assessed using the chi-square test. Serum GDF15 levels were expressed as mean ± SD. The distributions of GDF15 values in serum were asymmetric; therefore, nonparametric analyses (Mann–Whitney U test and Kruskal–Wallistest) were used to compare median values between groups. The Fisher exact test was used to compare the clinicopathologic characteristics according to the serum GDF15 level.In addition, a receiver operating characteristic (ROC) curve was employed to obtain the area under the curve (AUC), sensitivity, and specificity. The survival probabilities of patients according to the GDF15 expression level were described using Kaplan–Meier curves and compared using the log-rank test. Factors that showed significant prognostic value on univariate regression analysis were evaluated with multivariate Cox regression analysis. A value ofp0.05 was considered statistically significant.

RESULTS

Characteristics of EOC patients

A total 145 patients diagnosed with primary EOC between January 2010 and December 2013 were studied. The characteristics of all the patients are summarized in Table 1.The age range of the patients was from 35 to 83 years (mean: 51.92±15.95 years).

Table 1. Characteristics of patients with ovarian cancer

Characteristic / Number of patients (%)
Age (years)
60 / 119 (82.1)
≥60 / 26 (17.9)
Pathologic type
Serous cystadenocarcinoma / 111 (76.6)
Mucinous cystadenocarcinoma / 14 (9.7)
Endometrioid adenocarcinoma / 9 (6.2)
Undifferentiated carcinoma / 11 (7.6)
Histologic differentiation
Well / 20 (13.8)
Moderate / 11 (7.6)
Poor / 114 (78.6)
FIGO stage
I+II / 41 (28.3)
III+IV / 104 (71.7)
Ascites
Negative / 23 (15.9)
Positive / 122 (84.1)
Lymph node metastasis
Negative / 72 (49.7)
Positive / 73 (50.3)
Response to first-line chemotherapy
Sensitive / 111 (76.6)
Resistant / 34 (23.4)
Distant metastasis
Negative / 117 (80.7)
Positive / 28 (19.3)
Recurrence
Negative / 66 (45.5)
Positive / 79 (54.5)
CA125 expression (U/ml)
500 / 52 (35.9)
501–1000 / 43 (29.7)
≥1000 / 50 (34.5)

Increased GDF15 expression in EOC tissues

EOC tumor cells displayed cytoplasmic GDF15 staining (Fig.1). The positive expression rate of GDF15 was 82.39% in EOC tissues. Weak or negative staining was observed in 93 samples (64.1%), moderate staining was observed in 24 samples (16.6%), and high GDF15 protein expression was detected in 28 tumor samples (19.3%). In contrast, GDF15 expression was never detected in normal ovarian tissues. As shown in Table 2, the expression of GDF15 in EOC tissues was significantly higher in patients with advanced FIGO stages (III+IV) than in patients with early-stage tumors (I+II;p=0.026).Further analysis showed that the expression of GDF15 in EOC tissues was related with ascites (p=0.017) and lymph node metastasis (p=0.003). EOC patients whowere resistantto first-line chemotherapy more frequently showed higher GDF15 expression than those who were sensitive to first-line chemotherapy (p=0.030). However, GDF15 expression in EOC tissues did not differ with pathologic type and histologic differentiation (p0.05).

Table 2.GDF15 expression and clinicopathologic parameters

Characteristic / GDF15 expression / χ2 / p value
Low / Moderate / High
Age (years) / 7.012 / 0.031
60 / 82 / 18 / 19
≥60 / 11 / 6 / 9
Pathologic type / 4.062 / 0.694
Serous / 70 / 19 / 22
Mucinous / 11 / 1 / 2
Endometrioid / 4 / 3 / 2
Others / 8 / 1 / 2
Histologic differentiation
Well / 15 / 3 / 2 / 2.460 / 0.673
Moderate / 6 / 3 / 2
Poor / 72 / 18 / 24
FIGO stage
I+II / 33 / 5 / 3 / 7.296 / 0.026
III+IV / 60 / 19 / 25
Ascites
Negative / 19 / 4 / 0 / 8.165 / 0.017
Positive / 74 / 20 / 28
Lymph node metastasis
Negative / 51 / 15 / 6 / 11.686 / 0.003
Positive / 42 / 9 / 22
Response to first-line chemotherapy
Sensitive / 72 / 22 / 17 / 6.611 / 0.030
Resistant / 21 / 2 / 11

Figure 1. Immunohistochemical micrographs of GDF15 protein in different ovarian tissues (400×).

(A)Negative, (B) low, (C) moderate, and (D) high GDF15 expression in EOC tissues.

We next examined the relationship between GDF15 expression and prognostic outcomes in EOC patients. All 145 EOC patients with optimally debulked tumors and available outcome data were included in the survival analysis. The median follow-up duration was 30.24 months (range, 40 to 79.3 months). The detailed clinical information of the 145 EOC patients divided according to GDF15 expression level (low, moderate,or high)was reviewed to determine the prognostic implications of GDF15 expression. Analysis using the Kaplan–Meier method showed that patients with high GDF15 expression had significantly shorter OS than those with low or moderate GDF15 expression (19.13 months versus 58.62 months and 47.11 months, p=0.000; Fig. 2). Similarly, the median postoperative progression-free survival (PFS) time was lower in patients with highGDF15 expression (11.13 months) than in those with low or moderate GDF15 expression (33.62 months and 48.43 months, respectively, p=0.000). The Kaplan–Meier curves indicated that high GDF15 expression was significantly associated with an increased risk of death.

Figure 2. Kaplan–Meier curves of survival durations in EOC patients grouped according to GDF15 expression. Both progression-free survival (PFS) and overall survival (OS) were significantly shorter in patients with high GDF15 expression than in those with low or moderate GDF15 expression.

To identify factors that affected OS, the five clinical factors listed in Table 3and the GDF15 levelswere included in a multivariate Cox regression analysis. The analysis revealed that tumor stage (p=0.023), response to first-line chemotherapy (p=0.004), and GDF15 level (p=0.014) were significantly associated with survival among the EOC patients.

1

Table 3. Multivariate analyses of factors associated with overall survival

Risk factor / Overall survival / Progression-free survival
HR / 95% CI / p value / HR / 95% CI / p value
Age (years)
(<60 vs. ≥60) / 1.54 / 0.51–2.82 / 0.495 / 1.52 / 0.75–2.78 / 0.323
Grade
(Poor vs. others) / 1.39 / 0.66–2.82 / 0.243 / 1.14 / 0.63–2.04 / 0.449
FIGO stage
(I + II vs. III + IV) / 0.41 / 0.16–0.43 / 0.023 / 0.41 / 0.32–0.87 / 0.019
Lymph node metastasis
(Positive vs. negative) / 2.27 / 1.45–10.51 / 0.016 / 2.25 / 1.07–2.94 / 0.040
Response to first-line chemotherapy
(Sensitive vs. resistant) / 4.72 / 2.79–9.87 / 0.004 / 3.64 / 2.62–7.51 / 0.017
GDF15 level
(High vs. low) / 6.60 / 1.656–26.28 / 0.014 / 6.19 / 1.468–23.87 / 0.039

Multivariate analysis and Cox proportional hazards regression model were used. Variables were adopted because of their prognostic significance demonstrated on univariate analysis (p0.05).

HR, hazards ratio; CI, confidence interval; FIGO, International Federation of Gynecology and Obstetrics.

1

Serum GDF15 level and its correlation with clinicopathologic characteristics

A total 120 blood samples taken from EOC patients before they underwent surgery were compared with samples taken from 40 healthy controls. Themean serum GDF15 concentration was significantly higher in the patient group than in thehealthy controls (1302.12 ± 415.03 pg/ml vs. 418.17 ± 301.66 pg/ml, p=0.047; Fig.3).

Figure3. Box plot of serum GDF15 levels in EOC patients and healthy controls. The serum concentrations of GDF15 in the EOC patients ranged from 550.17 pg/ml to 2402.26 pg/ml (median, 1302.12 pg/ml), and were significantly higher than the GDF15 levels in the control cohort of 40 healthy female volunteers (72.6–1410.78 pg/ml; median, 618.17 pg/ml; p=0.047, t-test).

As shown in Fig.4, the area under the ROC (AUROC) was calculated based on the serum GDF15 levels in 120 EOC patients and 40 healthy controls. The ROC analyses revealed that the estimated AUROC of serum GDF15 was 0.894, which was not superior to that of serum CA125 (0.924). The cutoff serum GDF15 level as determined using the Youdenindex was 748 pg/ml. At this cutoff, GDF15 showed a similar specificity (83.3%) to that of CA125 (88.1%), and a higher sensitivity (75.5%)than that of CA125 (68.2%).The combined AUC value of GDF15 and CA125 was 0.944, suggesting that the combination had a better performance than that of CA125 alonein the detection of EOC.

Figure4. ROC curves of serum CA125, GDF15(A), and the combination of CA125 with GDF15 (B) among EOC patients and healthy controls. (A) The estimated area under the ROC curve of serum GDF15 was 0.894 (95% confidence interval [CI]:0.791–0.957,p0.001). (B) The AUC value of the combination of GDF15 with CA125 was0.944 (95% CI: 0.856– 0.986, p0.001).

To determine the impact of elevated serum GDF15 levels in EOC patients, we analyzed the relationship between serum GDF15 levels and clinicopathologic characteristics (Table 4). Serum GDF15 levels were not correlated with pathologic type and differentiation(p0.05),but were significantly correlated with FIGO stage.Median serum GDF15 levels were obviously higher in patients with advanced stagesof EOC (III+IV) than in patients with early-stage EOC (I+II; 563.49 pg/ml vs. 987.23pg/ml, p=0.004). Moreover, the median serum GDF15 levels were higher in patients with ascites than in patients without ascites (586.3 pg/ml vs. 898.9pg/ml, p=0.026). Similarly, GDF15 levels were higher in patients with lymph node metastasis than in those without lymph node metastasis (663.8 pg/ml vs. 1044.8 pg/ml, p=0.039). Additionally, the serum level of GDF15 was significantly higher in EOC patients who were resistant to first-line chemotherapy than those who were sensitive to first-line chemotherapy (692.0 pg/ml vs. 1096.6pg/ml, p=0.030).

1

Table 4. Serum GDF15 levels by clinicopathologic characteristics of the patients

Characteristic / GDF15 (pg/ml)
Median (range) / p value / χ2 value / CA125 (U/ml)
Median (range) / p value / χ2 value
Pathologic type
Serous / 748.8 (37.9–2704.0) / 0.128 / 5.682 / 764.3 (75.0–2897.0) / 0.120 / 2.419
Non-serous / 868.8 (287.9-1475.2) / 481.8(35.6-1329.3)
Differentiation
Well / 541.9 (37.9–1266.8) / 0.221 / 3.019 / 566.90 (37.0–1808.0) / 0.011 / 6.443
Moderate / 782.8 (308.7–1214.0) / 1038.27(22.6–2492.8)
Poor / 1006.8 (256.0–2704.0) / 1285.79(97.5–2897.0)
FIGO stage
I+II / 563.49 (37.9–1266.8) / 0.004 / 13.158 / 488.92 (22.6–2492.8) / 0.039 / 4.263
III+IV / 987.23 (289.9–2704.0) / 1318.84 (575.0–2897.0)
Ascites
Negative / 586.3 (37.9–1652.2) / 0.026 / 4.937 / 700.9 (75.0–2518.0) / 0.022 / 9.609
Positive / 998.9 (69.2–2704.0) / 1134.0 (118.1–2897.0)
Lymph node metastasis
Negative / 663.8(37.9–1402.2) / 0.039 / 4.263 / 818.6(75.0–2518.0) / 0.120 / 2.419
Positive / 1044.8 (69.2–2704.0) / 1317.9(127.7–2897.0)
Response to first-line chemotherapy
Sensitive / 692.0 (37.9–1173.2) / 0.037 / 2.245 / 795.1(75.0–2225.0) / 0.009 / 2.876
Resistant / 1096.6(256.7–2704.0) / 1546.9(171.0–2897.0)

1

DISCUSSION

EOC is the leading cause of death among gynecologic malignancies. In spite of recent advances, over three-quarters of EOC patients are diagnosed at an advanced stage because of the asymptomatic nature of the early stage of the disease and the rapid progression of chemoresistant disease. Itis worthnotingthatovarian cancer represents a very diverse group of tumors. The epithelial category, which accounts for 90% of all ovarian cancers, is classifiedinto the following subtypes: (1)serous (50%); (2) endometrioid (10%–25%); (3) mucinous (5%–10%); (4) clear cell (4%–5%); (5) undifferentiated (5%); and (6) transitional cell cancer (rare)[21]. Over the last three decades, CA125 has been used for distinguishing malignant from benign pelvic masses, detecting recurrent disease, monitoring response to treatment, and for early detection[4, 6, 8]. However, serum CA125 is not an ideal biomarker for EOC screening because of its low sensitivity and specificity. Høgdall et al.[22]found that the test for CA125 is positive in 85%–90%of serous tumors, 40%–65% of clear cell and endometrioid tumors, and only6%–12% of mucinous tumors.Furthermore, serum CA125 levels may be in the normal range in 50% of symptomatic stage I patients and in about 10%–20% of advanced-stage patients[23-26].The identification of valuable diagnostic and prognostic biomarkers to improve the outcomes of EOC patients remains a challenge.