High glucocorticoid receptor expression predicts short progression-free survival in ovarian cancer

2.1. Patients

Tissue microarrays (TMAs) were constructed from 481 patients with ovarian cancer collected at the time of definitive debulking surgery at Oregon Health & Science University Hospital and the University of Southern California during a 15-year period from 1995 to 2010. An additional 4 patients had benign conditions. Specimen collection and retrospective review of patient medical records were performed under a protocol approved by the local Institutional Review Boards (IRB). The review included outpatient and inpatient treatment, including surgery and chemotherapy. Study outcomes included overall survival (OS) and time to recurrence or progression, each measured from the time of definitive surgery. The duration of OS was the interval between definitive surgery and death. The duration of progression-free survival (PFS) was the interval between definitive surgery and first recurrence or progression. Observation time was the interval between definitive surgery and last contact (death or last follow-up). Data were censored at 120 months for patients who did not reach a study endpoint of recurrence, progression, or death. Among the 341 patients who received primary cytoreductive surgery followed by adjuvant chemotherapy (Cohort 1) there were 243 progression events and 201 deaths with a median follow up time of 43.4 months (Interquartile range 26.3–71.2). An additional 144 cases included in the tissue microarrays (Cohort 2, OHSU & USC) consisted of an expanded complement of cases described in further detail below.

A separate group of 20 patients who received neoadjuvant chemotherapy was selected from the University of Chicago tissue bank (Cohort 3). Overall, a total of 130 cases received neoadjuvant chemotherapy. Of these, 35 cases had a biopsy specimen prior to receiving chemotherapy in addition to post-treatment surgical specimen at the time of interval cytoreductive surgery. Twenty cases were identified with adequate tumor cellularity in both specimens (pre- and post- chemotherapy) for IHC analysis.

2.2. Tissue microarray construction

TMAs were constructed as described previously by Kononen et al. [14]. Briefly, after carefully choosing the morphologically representative region from the hematoxylin and eosin (H&E) section, a one-millimeter core was punched from the individual paraffin-embedded block (donor block), and transferred to the receiver paraffin-embedded block (receiver block). To overcome tumor heterogeneity, core biopsies were performed from three different areas of each tumor. One section was stained with H&E to confirm the presence of the tumor by light microscopy. The tumor subtypes and grade were re-reviewed for confirmation by a single experienced pathologist (PMF). Histologic subtypes were based on the World Health Organization (WHO) guidelines. Histologic grading was based on the Silverberg grading system [15]. Staging was performed according to the 1988 FIGO classification [16].

2.3. Immunohistochemistry

GR IHC was performed as follows. Four-micron sections were cut from formalin-fixed, paraffin-embedded whole tissue sections or TMAs. Endogenous peroxidase was blocked with 0.3% hydrogen peroxide for 5 min. Antigen retrieval was carried out in high pH (pH 8) citrate buffer for 3 min in a steam-cooker. Then, sections were incubated for 1 h with GR antibody (rabbit monoclonal antibody, clone# D8H2 XP, Cell Signaling, Danvers, MA) at 1:500 dilution at room temperature. A subsequent reaction was performed with the biotin-free HRP enzyme labeled polymer of the Envision plus detection system (Dakocytomation, Carpinteria, CA). Diaminobenzidine complex was used as chromogen. Breast and ovarian cancer tissues were used as positive controls. In negative controls, normal goat serum was substituted for primary antibody resulting in a lack of detectable staining. Ten whole sections of ovarian cases that were included in the array were also stained for comparison and showed homogenous nuclear staining throughout the tumor indicating that the TMA cores were representative of the tissue. One pathologist (PMF) reviewed the IHC slides twice at an interval of two months. Staining intensity and percentage of positive tumor cells were recorded. The staining intensity was interpreted as none, weak, moderate, or strong (i.e. 0, 1+, 2+, 3+). Percentage of tumor staining was categorized as follows: 0 = no positive staining; 1 = 1–5%; 2 = 6–25%; 3 = 26–50%; 4 = 51–75%; 5 ≥ 76%. Final analyses focused on categorization of tumor expression of GR as low when tumor cells exhibited 0 or 1+ intensity staining or as high when tumor cells exhibited 2+ or 3+ in >1% of tumor cells.

2.4. Statistical analysis

Of the 485 specimens in the TMA, the primary analysis focused on 341 specimens from patients with invasive EOC who received primary debulking surgery followed by adjuvant chemotherapy and had complete clinical follow up data available, including residual disease status (Cohort 1, OHSU & USC). The remaining 144 cases included in the tissue microarrays (Cohort 2, OHSU & USC) consisted of 12 borderline tumors, 42 early stage (I and II) EOC (23 treated with surgery alone, 19 treated with surgery and adjuvant chemotherapy for whom follow up data was incomplete), 26 advanced stage (III and IV) EOC cancers treated with primary surgery and adjuvant chemotherapy for whom follow up data was incomplete, 7 carcinosarcoma cases, 53 advanced stage cases treated with neoadjuvant chemotherapy followed by interval debulking surgery and adjuvant chemotherapy, and 4 cases with benign disease. A schema of study cohorts is shown in Supplemental Fig. S1. All 485 specimens in Cohorts 1 and 2 were collected at the time of primary or interval debulking surgery at OHSU & USC, including the specimens from the 53 patients who received neoadjuvant chemotherapy, whose specimens were collected after exposure to chemotherapy. Finally, Cohort 3 consisted of samples from 20 patients from the University of Chicago tissue bank who received neoadjuvant chemotherapy and for whom pre- and post-chemotherapy specimens were available.

To evaluate the relationship between GR expression and clinical characteristics, the Pearson’s Chi-square method or Fisher’s Exact test was used to test for association with age group, histologic subtype, FIGO stage (early versus advanced) and presence of residual disease after debulking surgery. For our analysis, early stage was defined as FIGO stage I/II and advanced as FIGO stage III/IV. In our dataset, gross residual disease was defined as ≥1 cm remaining and no gross residual disease was defined as <1 cm disease remaining. A Mantel-Haenszel test of trend was used to test for correlation of GR expression with grade. PFS and OS were estimated using the Kaplan-Meier method, with the difference between groups compared by log-rank test. Cox proportional hazards model was also performed to assess the association of GR expression with PFS or OS, with multivariate analysis adjusted for age (in increments of 10 years), histological cell type, grade, stage (early versus advanced), and the presence of gross residual disease after surgical debulking. Pre- and post-chemotherapy cases (Cohort 3) were compared by Wilcoxon rank sum test using STATA. All other statistical analyses were performed by SAS version 9.4 (SAS Institute, Cary, NC). All p-values are two-sided and a p-value of ≤0.05 was considered statistically significant.

3.1. Patient characteristics

Table 1 shows clinical and pathologic tumor characteristics for Cohort 1. This cohort consisted of 341 patients who received upfront debulking surgery followed by adjuvant chemotherapy and for whom complete clinical and follow up data were available. The median age was 59 years (range 24–89 years). The majority of the tumors (n = 240 (70.9%)) were of serous histology; nine cases were low grade serous (Grade 1) and 231 were high grade serous (Grade 2 or Grade 3). The remaining cases included clear cell (n = 42 (12.3%)), endometrioid (n = 32 (9.4%)), mucinous (n = 17 (5.0%)), and other (n = 10 (2.9%)). Most tumors were high grade (Grade 2 or Grade 3, n = 311 (91.2%)) and advanced stage (FIGO III or IV, n = 245 (71.8%)).

3.2. GR expression varies with histologic subtype, grade and stage

High GR expression, defined as >1% of tumor cells with 2+ or 3+ intensity staining, was present in 133 of 341 patients (39%) of tumors in Cohort 1 (Table 1). Representative stained sections of the TMAs are illustrated in Fig. 1. There was no association between GR expression and patient age at diagnosis (p = 0.772). High GR expression was significantly associated with histologic subtype (p < 0.001), with the highest prevalence observed in the serous tumors (n = 115/240 (47.9%)) and the lowest prevalence observed in the endometrioid tumors (n = 2/32 (6.3%)). Clear cell and mucinous tumors had high GR expression in 11 of 42 (26.2%) and 2 of 17 (11.8%) cases, respectively. High GR expression also correlated positively with increasing grade (p = 0.001) and with advanced stage at diagnosis (early versus advanced, p = 0.037).

GR expression was analyzed in the remaining cases included in the TMAs (Cohort 2) in order to evaluate GR expression in premalignant cases and other histologies (carcinosarcomas), increase the number of early stage tumors, and evaluate tumors collected at the time of interval debulking surgery. These cases were excluded from the primary statistical analyses, which were performed exclusively in Cohort 1. Clinical characteristics for all 481 EOC patients in Cohorts 1 and 2 are shown in Supplemental Table S1. Carcinosarcomas and serous cases had the largest proportion of cases with high GR expression (n = 4/7 (57.1%), and n = 128/265 (48.3%), respectively, Fig. 2A). As was observed when just Cohort 1 cases were considered, high GR expression was rare in endometrioid cases (n = 5/47 (10.6%)). Few borderline tumors exhibited high GR expression (n = 2/12 (16.7%), Fig. 2B). Early stage tumors treated with surgery alone (earlyS) or surgery and adjuvant chemotherapy (earlySCT) had a similar proportion of cases with high GR expression (n = 7/23 (30.4%) and n = 37/116 (31.9%), respectively). Advanced stage tumors treated with neoadjuvant chemotherapy before interval debulking surgery had an increased rate of high GR expression compared with unmatched advanced stage cases treated with primary debulking surgery (62.3% versus 41.9%, p = 0.006, respectively). In the 4 benign cases, no GR staining was observed in the ovarian epithelial cells.

The prevalence of GR expression in neoadjuvant cases may be due to high GR expression pre-treatment; alternatively, chemotherapy may cause selection for tumor cells with high GR expression. To examine whether high GR expression in neoadjuvant cases is present pretreatment or whether chemotherapy selects for chemotherapy-resistant cells with high GR, GR expression was examined in an independent cohort of 20 patients for whom paired pre- and post-neoadjuvant chemotherapy specimens were available (Cohort 3, Supplemental Table S2). Of the 20 pairs, 18 had evaluable staining. GR expression was not statistically different between paired pre-treatment and post-treatment specimens (p = 0.583, Fig. 2C). High GR expression (2+ or 3+ staining) was present in 15 out of 18 (83%) specimen in both groups.

3.3. High GR expression is associated with decreased PFS

High GR expression predicted an increased risk of disease progression in Cohort 1 (hazard ratio (HR) = 1.66, 95% confidence interval (CI) = 1.29–2.14, Fig. 3A). Median PFS was 15-months shorter in patients whose tumors had high GR expression compared to those whose tumors had low GR expression (20.4 versus 36.0 months respectively, p < 0.001, Fig. 3A). There was no significant difference in OS by GR status (Fig. 3B). When adjusted for other clinical variables in multivariate analysis, high GR expression remained an independent predictor of PFS (HR = 1.41, 95% CI 1.08–1.84, p = 0.012, Table 2) but not of OS. Other clinical variables with independent prognostic value for PFS and OS, respectively, included age at diagnosis (p = 0.024 and p = 0.007), stage (p < 0.001 and p < 0.001), and the presence of residual disease after primary surgery (p < 0.001 and p < 0.001), consistent with prior studies [17].

The level of GR expression needed to have an impact on outcomes is not known. The relationship of PFS with staining intensity and percent staining as nominal variables was also examined (Supplemental Table S3). A strong correlation was observed between the magnitude of GR intensity and the percentage of tumor cell nuclei with positive GR staining (p < 0.001). Supplemental Fig. S2A illustrates the ranked relationship between the intensity of GR staining and PFS with the longest PFS observed in patients with 0 or 1+ intensity staining and the shortest PFS observed with 2+ or 3+ intensity staining (p < 0.001, Supplemental Fig. S2A). A statistical trend was also observed when PFS was categorized by the percentage groups from 0 to 5 (p = 0.022, Supplemental Fig. S2B). Patients with zero GR positive tumor cells had the best outcomes and patients with any percentage of GR positive tumor cells had worse outcomes. Because of the strong correlation of intensity with percentage staining, we chose to characterize our cases by intensity using a dichotomous variable for ease of interpretation.

3.4. High GR expression shows a trend toward decreased PFS across histologic subtypes, grades and stage

As expected, the majority of cases in the study were of high grade serous histology. However, when only advanced stage high grade serous cases were considered, there was only a marginal difference in PFS (HR = 1.34, 95% CI 1.00–1.78, p = 0.048, Fig. 3C) and no difference in OS (HR = 0.95, 95% CI 0.68–1.30, p = 0.740, Fig. 3D) between cancers with high and low GR expression.

Further subgroup analyses were then performed, starting initially with patients diagnosed with early stage disease. Fig. 3E shows that among the 96 cases of early stage disease in Cohort 1 those with high GR expression had decreased PFS and increased risk of disease progression compared to those with low GR expression (HR = 2.68, 95%CI 1.16–6.19, p = 0.016). To examine this further, the 42 early stage cases in Cohort 2 were evaluated separately. Fig. 3F shows that a strong correlation of high GR expression with poor PFS was again observed in this cohort (HR 8.35, 95% CI 0.93–74.79, p = 0.023), independently validating this finding. Clinical characteristics for early stage patients are shown in Supplemental Table S4.

High GR expression also correlated with decreased PFS in the subsets with advanced stage disease, grade 2, and no gross residual disease (Fig. 4A). A trend toward decreased PFS was observed for high GR expression across all histologic subtypes (Fig. 4A). The trend was also present for all grades, however confidence intervals for the subgroups are wide for many variables examined. While there is no association of high GR expression with OS for Cohort 1 as a whole, or in the early stage cases, Fig. 4B demonstrates that high GR expression is associated with a shorter OS for the subset of patients with no gross residual disease.

Supplemental Table S5 displays the results of univariate and multivariate modeling for PFS for high versus low GR expression for all patients in Cohort 1, and in the subsets with early stage, advanced stage, advanced stage high grade serous ovarian cancer, or no gross residual disease. These analyses were also performed for the relationship of PFS with GR expression categorized as 0, 1+, 2+ or 3+ intensity (Supplemental Table S5).