Non-endometrioid and high-grade endometrioid endometrial cancers show DNA fragmentation factor 40 (DFF40) and B-cell lymphoma 2 protein (BCL2) underexpression, which predicts disease-free and overall survival, but not DNA fragmentation factor 45 (DFF45) underexpression

This study was performed using archived paraffin-embedded tissue samples from patients who were diagnosed with and treated for ECs in the Gynecology and Oncology Clinic of Jagiellonian University, Krakow, Poland. Samples from 365 patients who were consecutively diagnosed with EC from January 2007 to December 2012 were initially considered eligible for inclusion in the study; however, the medical records for 24 of these women were found to have missing data, and the samples from an additional 17 patients underwent autolysis. Thus, samples from 342 patients were ultimately eligible for inclusion in the study. Each patient was preoperatively diagnosed with EC based on the results of endometrial evaluations performed via dilatation and curettage (D&C) or hysteroscopy with biopsy. Each patient also underwent pelvic ultrasound or pelvic magnetic resonance imaging, abdominal ultrasound, and chest X-ray or chest computed tomography preoperatively. Each patient subsequently underwent total hysterectomy with bilateral salpingo-oophorectomy with or without pelvic lymph node biopsy, and patients with an increased risk of disease recurrence and/or nodal metastases underwent radical surgery and/or pelvic/para-aortal lymph node dissection. The omentum was removed in patients with tumors with histology corresponding to serous carcinoma, clear cell carcinoma, or carcinosarcoma. Each postoperative specimen was re-evaluated by two experienced pathologists, who were blinded to each other’s interpretations and to previous pathology reports, to confirm the final diagnosis of EC. All cases were restaged using current FIGO 2010 criteria [3]. Moreover, histological typing, grading (G), and the presence of MELF pattern myometrial invasion in the patients’ paraffin-embedded tissue specimens were performed according to the classification of the World Health Organization (WHO) and criteria presented by Murray et al. [6, 24] Body mass index (BMI) was calculated by dividing body mass, as determined by preoperative measurements of standardized morning weights, by the square of the body height (kg/m²), as determined by preoperative measurements of height. The date of menopause onset was defined as the date of the final menstrual period, i.e., the date after which no menses were reported for a subsequent period of 12 months. Sub-cohorts comprising patients with G1-2 endometrial adenocarcinoma (Group 1) and patients with non-endometrioid EC and G3 endometrial adenocarcinoma (Group 2) were subsequently organized. The characteristics of these patients and their tumors are presented in Table 1.

Representative postoperative tissue blocks were identified for immunohistochemical analysis. Four specimens from four patients with stage 4 disease that were obtained by D&C were also included in the analysis. Immunohistochemical staining was performed with a rabbit polyclonal antibody against DFF40 (Abcam, Cambridge, UK) diluted 1:50, a monoclonal mouse anti-human antibody against BCL2 (Leica Biosystems Newcastle Ltd., Newcastle Upon Tyne, UK) diluted 1:200 and a rabbit polyclonal antibody against DFF45 (Abcam, Cambridge, UK) diluted 1:100 using a previously described protocol [19, 20].

Sections that were stained but incubated without a primary antibody were used as negative controls. Jurkat cells (for DFF45 and DFF40) and human follicular lymphoma cells (for BCL2) were used as positive controls.

To obtain comparable outcomes to our formerly published results, the following procedure, which was previously developed and validated, was employed: two certified histopathologists, who were blinded to the study data and to each other, calculated the staining scores for DFF40, BCL2 and DFF45 for each slide in five high-power fields (× 40) using a 0-to-3 scale (0, no staining; 1, weak staining; 2, moderate staining; and 3, strong staining) [19, 20]. The cell staining percentage scores for DFF45 and DFF40 were determined as follows: 0 = expression in up to 10% of cells; 1+ = expression in 10–50% of cells; 2+ = expression in 51–80% of cells; and 3+ = expression in over 80% of cells. The cell staining percentage scores for BCL2 were calculated using the scale defined by Yigit et al.: 0 = expression in up to 5% of cells; 1+ = expression in 5%-25% of cells; 2+ = expression in 26%-50% of cells; and 3+ expression in over 50% of cells [25]. The intensity score was multiplied by the cell staining percentage score to obtain the final immunoreactivity score for each protein, which ranged from 0 to 9. Final scores of 0-1 were indicative of negative protein expression, whereas scores of 2-4 were indicative of low protein expression, and scores of 6-9 were indicative of high protein expression. Discrepancies regarding scoring were noted in 1.75% of the cases. In these instances, the corresponding samples were re-evaluated 2 weeks after the primary evaluation to achieve consensus regarding their scores and to minimize the possibility that the results of analysis would be affected by recall bias.

The clinical characteristics of the study groups were compared using the parametric Student’s t-test and the non-parametric Mann–Whitney U test as appropriate. The data are presented as the mean ± standard deviation (SD) or as the median ± standard error of the mean (SEM). The chi-squared test was employed to evaluate differences in biomarker immunoexpression, with Yate’s correction, if required, based on the expected frequencies of variables. Cox’s proportional hazard model was used to determine the hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) for the survival predictors. The Cox–Mantel test was used to compare OS (defined as the period between the initial surgery and time of death) and DFS (defined as the period between the initial surgery and time of disease recurrence), and the relative risks (RRs) of the factors associated with a poor prognosis were calculated for cases in which DFF40, DFF45, or BCL2 expression was negative. To evaluate interobserver/intraobserver agreement for the immunohistochemistry scores, we calculated the intraclass correlation coefficient (ICC) and corresponding 95% CIs for multiple histopathological evaluations. For these evaluations, Research Randomizer (www.​randomizer.​org) was employed to select 50 samples in which DFF40, DFF45, and BCL2 expression levels were re-evaluated at 2 weeks after the primary evaluation to prevent recall bias from affecting the study results. The Guidelines for Reporting Reliability and Agreement in Studies were used to verify these results [26]. Calculations were performed using STATISTICA data analysis software, version 12.0 (StatSoft, Inc. 2014. STATISTICA data analysis software system, version 12. www.​statsoft.​pl), and MedCalc Statistical Software, version 16.2.1 (MedCalc Software by Ostend, Belgium). A p value of 0.05 was considered statistically significant.

A total of 342 patients were included in analysis. The median follow-up duration was 64 (IQR: 42.00; range: 11-106) months. Multivariate analysis showed that stage 3+ disease (HR: 4.811; CI: 2391-9.681; p < 0.001), G3 disease (HR: 3.040; CI: 1.474-6.2711; p = 0.003), lymph node metastases (LN+) (HR: 2.834; CI:1.447-3.730; p = 0.002), and LVSI+ (HR: 1.976; CI: 1.047-3.730; p = 0.036) were significant predictors of OS, while age, BMI, positive peritoneal washings, and adjuvant therapy were not independent predictors of OS. Multivariate analysis also showed that stage 3+ disease (HR: 3.354; CI: 1.643-6.846; p < 0.001), LN+ (HR: 3.561; CI: 1.831-6.924; p = 0.001), and LVSI+ (HR: 2.068; CI: 1.108-3.862; p = 0.023), but not G3, age, BMI, positive peritoneal washings, or adjuvant therapy, were independent predictors of DFS. The clinicopathological characteristics of the study cohort and its sub-groups are shown in Table 1.

DFF40 and DFF45 displayed predominately nuclear expression, while BCL2 displayed cytoplasmic expression in all histological types of EC analyzed herein (Fig. 1). No differences in DFF40, DFF45, or BCL2 expression were observed between G1 ECs and G2 ECs in Group A, and no differences in DFF40, DFF45, and BCL2 expression were observed between G3 ECs and non-endometrioid tumors in Group B (Table 2). However, DFF40 and BCL2 expression levels were significantly higher in Group A than in Group B (p < 0.001 and p < 0.001; respectively). DFF45 expression levels did not differ between Groups A and B (Table 2).

The absence of DFF40 expression significantly increased the RRs for disease recurrence, LN+, LVSI+, and invasion of more than 50% of the myometrium (MI > 50%) in the entire cohort and in Group B (with the exception of the RR for MI > 50% in Group B) (Table 3). Similarly, BCL2 deficiency increased the RRs for disease recurrence, LN+, LVSI+, and MI > 50% in the entire cohort and in Groups A (with the exception of the RR for disease recurrence in Group A) and B (with the exception of the RR for MI > 50% in Group B). The absence of DFF45 expression was not associated with increases in the risks of any of the above factors, either in the entire cohort or in Group A or B (Table 3).

DFF40-negative cases showed significantly shorter OS in the entire cohort and in Group A and displayed significantly shorter DFS in the entire cohort, but not in Group A or B, compared with DFF40-positive cases (Fig. 2). DFF45 expression did not affect OS or DFS in the entire cohort or in Group A; however, DFF45-negative patients in Group B exhibited shorter DFS, but not OS. BCL2-negative patients displayed significantly shorter OS in the entire cohort and in Groups A and B compared with BCL2-positive patients (Fig. 2). Similarly, BCL2-negative patients displayed significantly shorter DFS in the entire cohort and in Groups A and B compared with BCL2-positive patients (Fig. 2).

Univariate analysis revealed that decreased DFF40 and BCL2 expression (but not decreased DFF45 expression) increased the negative HRs for OS (HR: 2.757; CI: 1.644-4.624; p < 0.001 and HR: 6.277; CI: 3.522-11.189; p < 0.001; respectively) and DFS (HR: and DFS (HR: 2.937; CI: 1.680-5.134; p < 0.001 and HR: 6.979; CI: 3.654-13.331; p < 0.001; respectively) in the entire cohort. We subsequently performed multivariate analysis of the relationships among the clinical features and DFF40, DFF45, and BCL2 expression levels and OS and DFS in the entire cohort. The results showed that stage 3+ disease (HR: 3.605; CI: 1.763-7.372; p < 0.001), G3 disease (HR: 2.467; CI: 1.845-5.135; p = 0.016), LN+ (HR: 3.290; CI: 1.865-5.805; p < 0.001), and BCL2 negativity (HR: 3.105; CI: 1.654-5.827; p < 0.001), but not LVSI+, MI > 50%, age, positive peritoneal washings, adjuvant therapy, or DFF40 or DFF45 negativity, were significant predictors of OS. Additionally, stage 3+ (HR: 2.682; CI: 1.312-5.487; p = 0.007), LN+ (HR: 3.997; CI: 2.217-7205; p < 0.001), and BCL2 deficiency (HR: 2.302; CI: 2.302-8.777; p < 0.001) were significant predictors of DFS.

We noted very high intra-rater agreement with regard to the immunohistochemistry scores for DFF40, DFF45, and BCL2 expression. The ICCs for agreement between the immunohistochemistry scores for the indicated proteins, which were determined by the abovementioned pathologists, are listed below.

As full inter-rater consensus was achieved for the expression levels of the above proteins, we did not evaluate inter-observer reliability.