Main

Ovarian cancer is responsible for more cancer deaths among women in the Western world than any other gynecologic malignancy.1 An initial surgical approach is essential for aggressive cytoreduction and proper staging of the disease, since minimal residual tumor after surgery is a major factor of better response to chemotherapy and survival.2 Intravenous combinated chemotherapy with taxol plus carboplatin is the current regimen of choice for the treatment of advanced ovarian cancer and is followed by a 50% complete pathologic remission rate.3

Resistance to chemotherapy is, however, a major concern. Indeed, although significant proportions of women respond to chemotherapy, the majority of responders (approximately 60–75%) eventually relapses and dies from recurrent disease while 20–30% of patients never experience a clinical remission.4 Chemotherapy resistance in ovarian cancer is broad and encompasses diverse, unrelated drugs, suggesting more than one mechanism of resistance. Until now, very few markers were found to predict tumor response to chemotherapy and prognosis in ovarian cancer.5, 6, 7 Recent advances in microarray technology led to identification of gene signatures that can help to improve diagnosis of ovarian cancer8 and in vitro drug resistance9, 10 but not clinical response to chemotherapy.11

We recently analyzed the gene expression patterns in advanced (FIGO stages III and IV) primary serous papillary carcinomas of the ovary displaying different response to first line chemotherapy in an attempt to identify specific molecular signatures associated with clinical response to chemotherapy.12, 13 Initially, the expression profiles of 15 chemoresistant serous papillary carcinomas (recurrence ≤6 months) and 10 chemosensitive serous papillary carcinomas (recurrence ≥30 months) tumors were independently analyzed which allowed the identification of 155 genes with different expression in the chemoresistant or the chemosensitive phenotype. The 155 genes differently expressed at a P-value cutoff of 0.01 were upregulated or downregulated at least 2-fold in chemoresistant tumors in comparison with chemosensitive tumors. Functional classes of these differently expressed genes mainly include metabolism (30%), cell growth and maintenance (18%), signal transduction (12%), immune response (12%), cell organization and biogenesis (11%), transport (9%) and apoptosis (3%); the remainder (5%) have unknown functions.

This experiment prompted us to test the hypothesis that the detection of corresponding markers at the protein level by immunohistochemistry may prove clinically applicable to the daily practice of pathology to predict the response to chemotherapy. We decided to analyze 10 markers for which commercial antibodies were available on an independent, uniform cohort of patients with serous papillary carcinomas.

Materials and methods

Patients included in this study were operated between January 1998 to December 2003 for an advanced ovarian cancer at the CHUQ-L'Hôtel-Dieu hospital in Quebec City, Canada. Inclusion criteria were: serous papillary carcinoma histology, FIGO Stages III or IV and chemotherapy received after the surgery. The grade was evaluated using criteria defined by Silverberg,14 which was in use in our institution during the accrual period. Clinical response to chemotherapy was evaluated using modified RECIST criteria.15 The follow-up was available until death or to the date the study was closed (31 July 2004). The study was approved by the Institutional Ethical Committee.

One representative block of each ovarian tumor was selected for the preparation of the tissue arrays. Three 0.6 mm cores of tumor were taken from each tumor block and placed, 0.4 mm apart, on a recipient paraffin block using a commercial tissue arrayer (Beecher Instruments, Sun Prairie, WI, USA). The cores were randomly placed on one of three recipient blocks to avoid immunohistochemistry evaluation biases. Four micron-thick sections were cut for the hematoxylin–eosin (H&E) staining and immunohistochemistry analyses.

The antibodies were selected based on the capacity of the corresponding gene to predict ovarian cancer prognosis in our previous micro-array study.13 A serious constraint was their commercial availability. The antibodies are presented in the Table 1. P53 and Ki-67 were included in our study because of their general interest in oncology. Immunohistochemistry staining was performed using the avidin-biotin complex method. Briefly, one representative 4 μm tissue section was cut from the tissue array blocks. Sections were deparaffinized and rehydrated in graded alcohols, then incubated with blocking serum for 20 min. The antibody dilutions, retrieval method, and incubation conditions are detailed in Table 2. Sections were incubated with a biotinylated secondary antibody (Dako, Carpinteria, CA, USA) and then exposed to a streptavidin complex (Dako, Carpinteria, CA, USA). Complete reaction was revealed by 3–3′ diaminobenzidine and the slide was counterstained with hematoxylin. Positive controls used in each case are described in Table 2. Negative controls consisted of tissue sections incubated with phosphate-buffered saline (0.16 M, pH ∼7.5) instead of the primary antibody. For three antibodies (CD36, FOSB and MMP1) the staining was weak and we used the catalyzed signal amplification system (Dako, Carpinteria, CA, USA).

Table 1 Antibodies used in the study
Full size table
Table 2 Dilution and technique used for each antibody
Full size table

Positive staining was defined when more than 10% of cells expressed the marker, except for Ki-67 for which 20% was defined as a threshold.16 The relationship between marker expression and patients' age, tumor grade, tumor size and the type of chemotherapy received was evaluated by the χ2 t-test. Cox regression analyses were performed to estimate the association between tumor expression and progression free survival. Progression free survival was defined as the time from surgery to the first observation of disease progression, recurrence or death. Multivariate analyses, taking into account standard or strongly associated prognostic variables, were performed to identify independent prognostic factors. A significant association was considered when P-value was below 0.05 and a trend for values between 0.05 and 0.1. Kaplan–Meier curves were done to show progression-free survival for each marker. The immunohistochemistry staining was analyzed independently by two pathologists (BT, IP) blinded to clinical data and progression.

Results

The Study Population

During the accrual period, we retrieved 235 consecutive cases operated in our hospital for an advanced serous papillary ovarian carcinoma, stage III and IV. Seventy-seven cases were excluded. Of them six refused chemotherapy, 10 died before the beginning of chemotherapy, clinical information were incomplete in 34 cases, in one case the stage was uncertain, eight patients had preoperative chemotherapy, one had a tumor of uncertain origin and 17 were still under chemotherapy at the last follow-up. A total of 158 cases responded to all inclusion criteria.

Table 3 shows the major clinical characteristics of the patients. The age ranged from 28 to 88 years (median: 61 years). Tumors were mainly grade 3 (67%) and stage III (78%). Seventeen patients had a second cancer, of them 10 were from the breast, three from the colon, one from the endometrium, two from the skin and there was one malignant lymphoma. A majority of patients (71%) received an intravenous combination of platinum and taxol, which was associated with a lower risk of progression compared to other combinations (Hazard ratio: 0.44 [0.26; 0.74]; P=0.002). The median baseline CA125 was 800 U/ml and a higher than average CA125 level was associated with increased risk of progression (Hazard ratio: 1.72 [0.99; 2.97]; P=0.05).

Table 3 Patients' characteristics
Full size table

Figure 1 shows the status of the patients at the end of the chemotherapy regimen and their evolution during the study. Ninety-seven patients had a complete response, of whom 77 underwent recurrences and 32 finally died of their disease. Twenty-nine patients had partial response or stable disease at the end of the chemotherapy, of whom 21 had progression and 12 died. Thirty-two patients had progression of their cancer under chemotherapy and 21 died. The median follow-up period of the cohort was 26.1 months. Fifty percent of the patients had a progression or a recurrence within the first 12 months of follow-up. At 5 years, only 13% of patients are recurrence-free and 43% are alive.

Figure 1
figure 1

Distribution of patients according to their response to chemotherapy and evolution.

Full size image

Figure 2 shows examples of immunostaining with the different antibodies used in this study. MMP1 gave a cytoplasmic granular staining in cancer cells while stromal cells were negative (Figure 2a); CD36 was expressed in the cytoplasm of cancer cells and stromal cells were negative (Figure 2b); HSP10 showed a cytoplasmic and granular staining of both cancer and stromal cells (Figure 2c); FOSB showed a granular cytoplasmic staining mostly limited to cancer cells (Figure 2d); GST had a diffuse cytoplasmic staining limited to cancer cells (Figure 2e); pan-cathepsin was present in both cancer and stromal cells (Figure 2f); prostaglandin D synthetase was present in less than 5% of cancer cells and was not retained for this study; RbAp46 gave a strong positive nuclear staining in basically all cases and was not further analyzed; Siva and MIP2 were completely negative despite repeated attempts with various retrieval systems, antibody concentration, incubating time, or signal enhancements systems and were not retained for the study; p53 (Figure 2g) and Ki-67 (Figure 2h) gave a nuclear staining.

Figure 2
figure 2figure 2

Marker expression by immunohistochemistry: (a) MMP1; (b) CD36; (c) HSP10; (d) FOSB; (e) GST; (f) pan-cathepsin; (g) p53; (h) Ki-67.

Full size image

Relation Between Markers and Risk Factors

No association was found between CD36, HSP10 and FOSB and any risk factor. MMP1 overexpression was associated with an older age (P=0.01). Overexpression of GST was positively associated with higher initial serum CA15 (P=0.02). Pan-cathepsin expression was associated with higher grade (P=0.02) and higher initial CA125 levels (P=0.01). p53 tended to be associated with an older age (P=0.06). High Ki-67 levels (>20%) was associated with higher grade (0=0.03). Furthermore, we found no linear correlation between gene expression obtained by microarray and protein expression obtained by immunohistochemistry (data not shown).

Relation Between Marker Expression and Progression-Free Survival

Table 4 shows the prevalence of expression of each marker along with bivariate and multivariate analyses to predict progression-free survival. Multivariate analyses taking into account standard or strongly associated prognostic variables (age, grade, stage, type of chemotherapy, initial CA125) were performed to identify independent prognostic factors. Multivariate analyses showed a significant association between HSP10 expression and a lower risk of progression after chemotherapy (HR: 0.6; CI: 0.42–0.87; P=0.007). A Kaplan–Meier curve for HSP10 and progression-free survival is depicted in the Figure 3. High proliferation rate (Ki67>20%) showed a tendency to predict a lower risk of progression post chemotherapy (HR: 0.72; CI: 0.51–1.03; P=0.07). A trend was found for MMP1 overexpression to predict a higher risk of progression (HR: 1.61; CI: 0.94–2.79; P=0.08).

Table 4 Cox regression analysis to predict progression-free survival
Full size table
Figure 3
figure 3

Kaplan–Meier curve for progression-free survival and HSP10.

Full size image

Discussion

Our study shows that HSP10 is the only significant factor of delayed progression in patients exposed to chemotherapy. These findings confirm those of our micro-array study12 and are also consistent with the biology of HSP10.

Indeed, heat-shock proteins (HSPs) are important molecules in oncology. Five types of primary HSPs are currently known, and they are designated according to their molecular weight (HSP27, 60, 70, 90 and 110).17 HSPs bind and stabilize proteins to prevent the creation of aggregates during protein synthesis, transmembrane transport or stress, such as high temperatures.18 There are some cochaperones of low molecular weight that often form a biologically active complex. The HSP10-HSP40 complex is known to facilitate interactions between primary HSP and the substrate.17 The HSP10-HSP60 complex is involved in apoptosis through caspase activation.19 Therefore, in cases with constitutively low HSP10 levels, chemoresistance, as we observed, may be explained by a lack of capacity to induce apoptosis. However, in another micro-array study, HSP10 mRNA was found to be overexpressed in breast cancer cell cultures exposed to oxyplatin and 5-FU, suggesting that HSP10 overexpression, rather than low expression, might be associated with chemoresistance.20 However, in such an instance, higher HSP10 levels might not be associated with chemoresistance but might rather be induced by exposure to chemotherapy agents.18, 19

In addition to the reaction to stress, HSPs play an important role in carcinogenesis. HSP60 and its cochaperone HSP10 are expressed early during the development of a malignant phenotype.21 In colon and uterine cervix cancers, HSP 10 and 60 expression levels are increased as cells progress from their normal state to dysplasia and cancer.19, 22 HSPs are not expressed only by cancer cells. Indeed, higher levels of HSPs 10 and 60 were found in the cytoplasm of lymphocytes in lymph nodes with metastatic colon cancer than in non-metastatic lymph nodes.23 The presence of HSP10 was also detected in the serum and ascites from ovarian cancer patients and it was found that HSP10 in the serum may play a role in T lymphocyte inhibition, allowing cancer cells to escape immunitary surveillance.24

However, the prognostic significance of HSPs is not clear. In head and neck cancer25 and in breast cancer,26 HSP27 was not found to predict survival but HSP27 predicted neck cancer failure after radiation therapy.27 A few studies investigated the role of HSPs in ovarian cancer. HSP27 was not found to predict response to chemotherapy5 but HSP 60 overexpression was associated with a poor prognosis.28

In our study, Ki-67 failed to significantly predict progression-free survival. However, high proliferation rate was associated with higher tumor grade and Ki-67 tended to predict a better survival without progression, suggesting that high proliferation is associated with chemosensitivity. This is consistent with data from the literature, high proliferation rate, as measured by Ki-67, being a marker of recurrence in soft tissue sarcoma,29 head and neck cancer30 and urothelial tumors.31 It also predicted response to neoadjuvant anthracycline-based chemotherapy in breast cancer32 and response to radiation therapy in head and neck cancer.16, 33

In advanced ovarian carcinoma, high Ki-67 index was predictive of recurrence34 and poor prognosis.35, 36 However, as in our study, low proliferation predicted poor response to chemotherapy in ovarian carcinomas.37

In our study, there was a tendency for MMP1 to predict a poorer progression-free survival. MMP1 overexpression by immunohistochemistry was associated with a poor prognosis in both colon38 and esophageal39 cancers. However, no such prognostic study has been reported in ovarian cancer. MMP1 expression was found to be negligible in benign cystadenomas but it was overexpressed in both tumor and stromal cells of serous papillary carcinomas.40 Interestingly, a 2G mutation on the MMP1 promoter was found to be associated with MMP1 overexpression.41 The authors suggest that such genetic abnormality may be associated with an increased risk of developing ovarian cancer, although others did not reach such a conclusion.42

The fact that, of the 12 genes selected by microarrays comparing chemosensitive and chemoresistant tumors, HSP10 was the only significant marker of progression-free survival by immunohistochemistry is intriguing. This may suggest that the 2-fold ratio between high and low expression by microarrays defined in our previous study,12 is not high enough to be clinically relevant by immunohistochemistry. Furthermore, we found no linear correlation between gene expression obtained by microarray and protein expression obtained by immunohistochemistry. Current literature also shows that the correlation between mRNA and protein levels is insufficient to predict protein expression levels from quantitative mRNA data. Indeed, for some genes, while mRNA levels are of the same value, the protein levels may vary by more than 20-fold and, conversely, proteins with similar levels may have respective mRNA transcript levels that vary by as much as 30-fold.43 Further studies should focus on the clinical and immunohistochemistry relevance of data obtained by microarrays.

We conclude that, despite a lack of direct relation between mRNA and proteins levels, gene expression analysis coupled with immunohistochemistry allowed us to identify high HSP10 and possibly, high proliferation and low MMP1 as potential markers of response to chemotherapy. Future studies should be aimed at developing a prognostic index combining the immunohistochemistry markers to predict the response to chemotherapy.