Versican in epithelial ovarian cancer: Relation to hyaluronan, clinicopathologic factors and prognosis

Authors


Abstract

Versican, a proteoglycan previously reported to increase in other malignant tumours, was studied immunohistochemically in 299 primary epithelial ovarian cancers, their 43 metastases and 6 normal ovaries to evaluate its prognostic value and relation to hyaluronan, another extracellular matrix molecule increased in cancer and a binding partner of versican. The stainings were scored according to the area percentage of strong versican signal of total peri- and intratumoural stroma as low (<15%) or high (≥15%). Epithelial staining of the tumours was scored as positive or negative. Low and high area percentage of strong stromal versican staining were observed in 133 and 166 carcinomas, respectively. A low area percentage of strong stromal versican staining correlated with mucinous histology (p = 0.019) and early International Federation of Gynecologists and Obstetritians (FIGO) stage (p < 0.0005), whereas a high percentage was associated with reduced 5-year survival rate of the patients (44% vs. 32%; p = 0.032). Versican was associated with the cancer cells in 151 tumours and correlated with clear cell histology (p < 0.0005), early FIGO stage (p = 0.049) and increased recurrence-free survival (63% vs. 47%; p = 0.032). However, in Cox's multivariate analyses with the conventional prognostic factors included, neither stromal nor cancer cell-associated versican reached a significant prognostic value. Versican is thus enriched in the malignant stroma surrounding and promoting the growth of ovarian cancer, probably acting with hyaluronan, and associates with unfavourable prognosis but does not constitute an independent indicator of patient survival. © 2003 Wiley-Liss, Inc.

Ovarian cancer is the sixth most common cancer and cause of death from cancer in women. Incidence rates are highest in developed countries, except in Japan.1 Ovarian cancer patients have a poor prognosis because the majority of the patients (approx. 70%) are in advanced stages (stages III and IV) at the time of diagnosis.2

Versican is a chondroitin sulfate proteoglycan, which belongs to the aggrecan gene family.3 At its carboxyl terminus, versican has EGF-like, lectin-like and complement regulatory protein-like domains, whereas the amino-terminal domain binds hyaluronan with high affinity.4, 5 In addition, in vitro studies have shown that versican binds to CD44.6 In normal adult human tissues, versican is found in connective tissues, most smooth muscle cells, veins and arteries, cartilage, nerves, glandular epithelia and skin.7 Elevated versican levels have been found in many malignant tumours.8, 9, 10, 11, 12, 13, 14, 15 In prostate cancer, increased concentration of stromal versican is an independent predictor of outcome for patients with moderately differentiated tumours.16 Likewise, peritumoural versican is a strong predictor of relapse-free survival in breast cancer.17

We have shown previously that a high level of stromal hyaluronan is an independent prognostic factor of poor survival in epithelial ovarian cancer patients.18 The fact that versican binds hyaluronan and that they are highly expressed in breast14 and prostate cancer16 prompted us to study the expression of versican and its possible prognostic role in a large group of epithelial ovarian cancer patients.

MATERIAL AND METHODS

Patients

We analysed 299 specimens from women treated and diagnosed for epithelial ovarian malignancy at Kuopio University Hospital and Jyväskylä Central Hospital, Finland, from 1976 to 1992 and subsequently followed up until September 1996 (Table I). Patients who were treated before operation or patients who were not operated on were not included, nor were patients who died because of any postoperative complications (deaths during 1 month after operation). Forty-three of the patients also had metastatic lesions. These as well as 6 normal ovarian samples were included in analysis.

Table I. CLINICOPATHOLOGICAL CHARACTERISTICS OF THE PATIENTS (n = 299)
Variablen%
Histologic grade  
 14013
 210134
 315853
FIGO stage  
 I8328
 II4615
 III13846
 IV3211
Histologic type  
 Serous10736
 Mucinous299
 Endometrioid8027
 Clear cell3211
 Other epithelial5117
Chemotherapy response  
 Partial response3815
 Complete response13555
 Stable disease198
 Progressive disease5522
Primary residual tumour  
 None12044
 ≤2 cm5018
 >2 cm10338
Strong stromal versican  
 <15%13345
 ≥15%16655
Cancer cell-associated versican  
 Negative14850
 Positive15150
Stromal hyaluronan  
 Low (<35%)8729
 Moderate (35–75%)11539
 High (>75%)9732
Cancer cell-associated hyaluronan  
 Negative4615
 Positive25385

Staging of tumours was based on standards of the International Federation of Gynecologists and Obstetricians (FIGO). In addition to operative therapy, 9 (3%) patients received postoperative radiotherapy, 218 (73%) patients received postoperative chemotherapy and 36 (12%) patients received both of these adjuvant therapies. Until September 1996, disease recurrence was observed in 70 patients (23%), no recurrence in 93 patients (31%) and in 136 patients (46%) the tumour remained or was progressing. The age range of the patients at the time of the diagnosis was 17.7–84.9 years, and the median age at diagnosis was 61.7 years. The median follow-up time for all patients (n = 299) was 27 months (range, 1–237 months) and for patients still alive (n = 99) 104 months (range, 21–237 months).

Histology

Histologic typing and grading were reevaluated according to the World Health Organization (WHO) classification as shown previously.18

Immunohistochemistry

Five-micron-thick paraffin-embedded tissue sections of all tumours were stained immunohistochemically. Deparaffinized and rehydrated sections were heated in a microwave oven in citrate buffer (pH 6.0) for 3 × 5 min, incubated in the citrate buffer for 18 min and washed with PBS for 2 × 5 min. Endogenous peroxidase activity was blocked by 5% hydrogen peroxide for 5 min and then the sections were washed with PBS for 2 × 5 min. Nonspecific binding was blocked with 1.5% normal horse serum in PBS for 25 min at room temperature. The primary antibody (mouse monoclonal anti-human versican, clone 2B1; Seikagaku Corporation, Tokyo, Japan), which recognizes all forms of versican,8 was diluted with 1% BSA in PBS to 1:1,000 and incubated on the slides overnight at 4°C. The negative control was incubated with 1% BSA in PBS instead of the primary antibody. The slides were washed with PBS for 2 × 5 min and incubated with the biotinylated secondary antibody (anti-mouse IgG; ABC Vectastain Elite kit, Vector Laboratories, Burlingame, CA) for 35 min at room temperature. After this, slides were washed with PBS for 2 × 5 min, incubated for 45 min in preformed avidin-biotinylated peroxidase complex (ABC Vectastain Elite kit, Vector Laboratories) and washed twice for 5 min with PBS. The colour was developed with diaminobenzidine tetrahydrochloride (DAB) substrate (Sigma, St. Louis, MO). The slides were counterstained with Mayer's haematoxylin, washed, dehydrated, cleared and mounted with DePex (BDH, Poole, UK). Skin samples showing strong staining in the dermis were used as positive controls. The negative control (primary antibody omitted) in each staining batch did not show any positivity. The staining procedure of hyaluronan has been described previously.19

Evaluation of the stainings

All specimens were analysed by 3 observers (K.V., S.S., V-M.K.) for versican and by 2 observers (M.A., V-M.K.) for hyaluronan, unaware of the clinical outcome. First, the whole tissue sample was thoroughly examined to check that there was no nonspecific staining. After that, the intensity of the total peri- and intratumoural stromal versican in the whole section was graded into 3 categories: weak = 1, moderate = 2 and strong = 3. Both high-power and low-power fields were used for the evaluation of staining. If the staining pattern in the stroma around and inside the malignant areas was similar to that in the normal stroma, the staining intensity was considered weak. The strong intensity corresponded to that seen in strong staining dermal areas of skin samples used as standards. Moderate intensity was defined as a staining intensity between weak and strong. Indeed, disagreement in the assessment of stainings was found in <10% of the slides examined, and consensus was reached on further review. For statistical analyses, the samples were divided into 2 groups according to the median area percentage of stroma with the strongest (category 3) versican intensity of the total peri- and intratumoural stroma. The median value (15%) was chosen because it does not introduce a bias that could arise by the use of a minimum p-value approach.20 The percentage of versican-positive tumour cells of all carcinoma cells in the section was also estimated, but for statistical analyses, the tumours were grouped into 1 of the 2 categories: versican negative or positive. A tumour was considered positive if any tumour cell-associated versican staining was observed. Hyaluronan intensity in stroma was categorized as: weak = 1, moderate = 2, strong = 3. Grading of the strong stromal hyaluronan staining percentage of total peri- and intratumoural stromal area was done using the 33rd and 66th percentiles in a frequency distribution: low (<35%), moderate (35–75%), high (>75%) as described earlier.18

Statistical analyses

The statistical analyses were carried out by using the SPSS software for Windows Release 10.0 (SPSS, Chicago, IL). Spearman correlation coefficients and Wilcoxon tests were used to examine the relationships between continuous variables. χ2 test was used in analyzing frequency tables. Univariate survival analyses were based on the Kaplan-Meier method,21 and Cox's proportional hazards model in a forward stepwise manner with the log-likelihood ratio significance test22 was used in multivariate survival analyses. Overall survival was defined as the time between the date of surgery and the date of death due to ovarian cancer. Recurrence-free survival was defined by the time interval between the date of surgery and the date of recurrence. Probability values <0.05 were regarded as significant.

RESULTS

Versican staining in stroma

In normal ovaries (n = 6), staining was observed in the vessel walls and more frequently in the loose than in tight connective tissue. The staining of the normal ovarian stroma was generally of weak intensity, whereas in the ovarian tumours the high intensity of versican staining was relatively frequent (Fig. 1a,b). A histogram of the area percentage of strong versican positivity in stroma is shown in Figure 2a. The sample material was divided into 2 categories, one with low percentage of versican expression (<15%) and another with high percentage of total area with strong versican staining (≥15%). In the primary tumours, low and high percentage of strong stromal versican staining were detected in 133 (44.5%) and 166 (55.5%) cases, respectively. In the metastases, high percentage of strong stromal versican were observed in 22 (51.2%) cases. No significant difference was found in strong stromal staining between primary tumours and metastases (p = 0.281).

Figure 1.

(a) Normal ovarian tissue sample with weak intensity of versican staining. Scale bar = 50 μm. (b) Serous ovarian cancer showing strong stromal versican positivity (asterisk). Scale bar = 100 μm; insert 50 μm. (c) Clear cell ovarian cancer showing cancer cell-associated versican positivity (arrow). Scale bar = 100 μm; insert 100 μm. (d) Serous ovarian cancer showing strong stromal versican positivity (asterisk) and strong hyaluronan positivity (e) (asterisk) in stroma. Scale bar = 50 μm.

Figure 2.

Distribution of versican staining in different ovarian cancer samples. The columns represent the numbers of cases within the percentage range indicated in the figure. The staining for stroma (a) and cancer cells (b) was analysed separately.

Epithelial versican staining

The epithelium of normal ovaries (n = 6) was invariably versican negative (Fig. 1a). In contrast, half of the ovarian tumours (151/299) showed cancer cell-associated versican staining (Fig. 1c). However, the overall percentage of versican-positive cells was <5% in 77.3% of the samples (Fig. 2b). In most cases (138/151), versican staining was either in the cytoplasm or on the plasma membrane; however, a nuclear localization was present in 13 cases. Of the metastases available for study, 56% (24 of 43) were positive for epithelial versican. The frequency of cancer cell-associated versican staining in the primary tumours did not differ from that in the metastases (p = 0.128).

Hyaluronan staining

The percentage of strong hyaluronan staining was higher in the intra- and peritumoural stroma than in the normal ovarian stroma. In the carcinomas, high, moderate and low percentage of strong stromal hyaluronan were observed in 98 (32%), 116 (37%) and 95 (31%) cases, respectively. Sixteen percent (49 of 309) of the tumours did not contain hyaluronan-positive cancer cells. In 73% (227 of 309) of the cases, the fraction of hyaluronan-positive cancer cells was ≤10%.

Correlation of versican with hyaluronan expression

A significant positive correlation was found between the high percentage of strong stromal versican and strong stromal hyaluronan in the primary tumours (r = 0.413; p < 0.0005) (Fig. 1d–e). Cancer cell-associated versican positivity did not correlate with strong stromal hyaluronan staining (p = 0.052) or cancer cell-associated hyaluronan positivity (p = 0.224) in the primary tumours. The percentage of strong versican expression was lower than that of hyaluronan both in tumour stroma (z = −14.319; p < 0.0005) and epithelium (z = −8.668; p < 0.0005). In the metastatic lesions (n = 43), the high percentage of strong stromal versican staining correlated with strong stromal hyaluronan staining (r = 0.420; p = 0.005). Neither stromal nor epithelial versican were related to cancer cell-associated hyaluronan positivity in the metastases.

Versican and clinicopathologic characteristics

A high percentage of strong stromal versican was associated with advanced FIGO stage (χ2 = 14.5; p < 0.0005), a large (>2 cm) primary residual tumour (χ2 =9.7; p = 0.002) and serous histologic type (χ2 = 11.8; p = 0.019). Versican positivity in tumour cells was correlated with clear cell histologic type (χ2 = 29.3; p < 0.0005), the absence of primary residual tumour (χ2 = 8.9; p = 0.006) and early FIGO stage (χ2 = 7.9; p = 0.015) (Table II).

Table II. DISTRIBUTION (n/%) OF CLINICOPATHOLOGIC VARIABLES WITHIN DIFFERENT EPITHELIAL AND STROMAL VERSICAN CATEGORIES
VariableEpithelial versican n (%)Strong stromal versican n (%)
Neg.Pos.<15%≥15%
FIGO stage    
 I33 (40)50 (60)50 (60)33 (40)
 II19 (41)27 (59)20 (44)26 (56)
 III79 (57)59 (43)55 (40)83 (60)
 IV17 (53)15 (47)p = 0.015 8 (25)24 (75) p < 0.0005
Histologic type    
 Serous69 (65)38 (35)39 (36)68 (64)
 Mucinous11 (38)18 (62)18 (62)11 (38)
 Endometrioid37 (46)43 (54)31 (39)49 (61)
 Clear cell4 (13)28 (87)15 (47)17 (53)
 Other epithelial27 (53)24 (47) p < 0.0005 30 (59)21 (41) p = 0.019
Histologic grade    
 118 (45)22 (55)17 (43)23 (57)
 244 (44)57 (56)47 (47)54 (53)
 386 (54)72 (46) p = 0.120 69 (44)89 (56) p = 0.936
Primary residual tumour    
 None52 (43)68 (57)66 (55)54 (45)
 ≤2 cm22 (44)28 (56)25 (50)25 (50)
 >2cm64 (62)39 (38) p = 0.006 35 (34)68 (66) p = 0.002

Survival

The 5-year prognosis became significantly worse with increasing strong stromal versican for overall survival (OS) but not for recurrence-free survival (RFS). The 5-year prognosis of RFS was better when tumour epithelium was versican positive compared to negative epithelium (63% vs. 47%; p = 0.032) (Table III; Fig. 3). Other factors in univariate analysis significantly decreasing OS were advanced FIGO stage (p < 0.00005), poor differentiation (p < 0.00005), primary residual tumour >2 cm (p < 0.00005) and older than median age at diagnosis (p = 0.018). Factors predicting a short RFS were poor differentiation (p = 0.042), serous histologic type (p = 0.03), advanced FIGO stage (p = 0.0001) and primary residual tumour >2 cm (p < 0.00005). Combined expression of versican both in cancer cells and stroma did not significantly affect prognosis.

Table III. THE PROGNOSTIC SIGNIFICANCE OF VERSICAN IN KAPLAN-MEIER ANALYSIS
 n5-year prognosis (%)P
Overall survival(OS)
 Versican (epithelial)   
  Positive148420.3602
  Negative151330.3602
 Versican (strong stromal)   
  <15%129440.032
  ≥15%160320.032
Recurrence–free survival (RFS)   
 Versican (epithelial)   
  Positive84630.032
  Negative74470.032
 Versican (strong stromal)   
  <15%79570.455
  ≥15%79530.455
Figure 3.

(a) Kaplan-Meier curves for overall survival of epithelial ovarian cancer patients according to the percentage of low (<15%) and high (≥15%) strong stromal versican staining (log-rank analysis). (b) Kaplan-Meier recurrence-free survival curves of patients with cancer cell-associated versican-positive and -negative tumours (log-rank analysis).

Cox's multivariate analyses included the histologic type and grade, FIGO stage, primary residual tumour, age at diagnosis, adjuvant chemotherapy, as well as the percentage of stromal and cancer cell-associated versican. Neither stromal nor cancer cell-associated versican had statistically significant prognostic value in OS or in RFS, in contrast to the clinicopathologic factors shown in Table IV.

Table IV. THE INDEPENDENT PROGNOSTIC FACTORS IN COX'S MULTIVARIATE ANALYSIS FOR OVERALL SURVIVAL (n = 261) AND RECURRENCE-FREE SURVIVAL (n = 149)
FactorCategoryRR95% CIP
  1. a, Reference category; RR, relative risk; 95% CI, 95% confidence interval.

Overall survival    
Adjuvant chemotherapyPlatinum-containing therapya 0.012
 Nonplatinum therapy1.4821.015–2.1640.042
 None2.2981.256–4.2040.007
FIGO stageI–II vs. III–IV2.2671.343–3.8270.002
Primary residual tumourNegative vs. positive4.1652.360–7.352<0.0005
Histologic gradeGrade 1–2 vs. 31.5381.092–2.1670.014
Recurrence-free survival    
Primary residual tumourNegative vs. positive3.6252.175–6.044<0.0005
Histologic typeSerous vs. others0.5090.306–0.8450.009

DISCUSSION

The expression of versican was increased in the stroma of epithelial ovarian cancers, a finding completely in line with several previous reports indicating that tumours contain higher amounts of versican than the corresponding normal adult tissues.9–12, 14–16

In a normal ovary, versican is mainly expressed in the granulosa cells of growing follicles, where it serves as an important matrix constituent, together with hyaluronan, contributing to the preovulatory expansion of cumulus oophorus.23 In general, versican expression is induced upon tissue activation, whether it be due to follicle growth,23 inflammation,24 wound healing25 or development of an atherosclerotic lesion.26 The tissue remodelling and neovascularization that take place in wound healing have several similarities to those around an expanding tumour.27 The expression of versican occurs often in tissues and situations known to cause an upregulation of hyaluronan, cancers as one example. The high correlation between the contents of versican and hyaluronan in the present patient material fits well with this notion and suggests that together these molecules organize the extracellular matrix supportive for cancer spreading.

In the ovarian cancers, versican was mainly present in the peritumoural stroma, but there were also samples with some versican-positive tumour cells. Earlier reports on breast and prostate cancers have also suggested mainly stromal localization,14, 16, 17 whereas melanomas also show cell-associated versican.12 Whether the stromal or cancer cells act as the source of versican in the tumours remains to be settled, but it has been demonstrated in vitro that many malignant cells can synthesize versican,12, 28–30 and also that malignant cells can stimulate versican synthesis in stromal fibroblasts.17, 31 The cell-associated versican signal may either result from enhanced uptake32 or visualization of molecules under biosynthesis or secretion. The abundancy of cell-associated versican in early stage cancer as well as in cases without residual tumour probably explains the apparently positive contribution of cell-associated versican to patient survival.

Versican is expressed as several different splice variants, some of which show even opposite biologic influences on cell behaviour,33, 34, 35 a fact that may explain some of the differences in the prognostic importance of versican between ovarian cancer and other malignancies.16, 17 At least 3 different antibodies, presumably against different epitopes, have been used in the reports on versican expression in cancer.12, 14, 16, 17 This may also contribute to the variation in the results between the research groups.

Versican can modulate the invasion of malignant cells by reduced cell adhesion,36 increased proliferation37, 38 and enhanced migration,36, 39 in part perhaps by acting on EGF receptors through its EGF-like motifs.37, 40 The influences of hyaluronan on cell behaviour are very similar to those of versican, but the mode of action of the 2 molecules, whether additive or interdependent, is not known at the moment. The localization in the proliferating and invasive areas of the interstitial and perivascular cancer tissues41 suggests an important role for versican.

However, despite its close association with the appearance and spreading of cancer in the present, relatively large patient material, versican did not reach the status of an independent prognostic factor for survival when conventional prognostic factors were included in the Cox's multivariate analysis. This was in contrast to prostate cancer, in which an increased concentration of stromal versican emerged an independent predictor of outcome for patients with moderately differentiated tumours.16 Furthermore, peritumoural versican is a strong predictor of relapse-free survival in breast cancer.17 These data suggest that there is a difference between the strength of versican as a prognostic factor in ovarian vs. breast and prostate carcinoma, although the confirmation of this must await application of identical analysis techniques to the different carcinomas.

To conclude, during the progress of the epithelial ovarian cells to malignant tumours, versican increases in the adjacent stroma and in rare cases also in the tumour epithelium. The abundancy of stromal versican in ovarian cancer correlates with a bleak disease outcome, but does not provide a prognostic factor as strong as hyaluronan, its binding partner in the stromal extracellular matrix.18

Acknowledgements

The skillful technical assistance of Ms. H. Kemiläinen, Mr. O. Horto and Ms. R. Törrönen is gratefully acknowledged. We also thank Mr. A. Pelttari for photographic advice. Our study was supported by Finnish Cancer Foundation (V-M.K. and R.T.) and EVO funds of Kuopio University Hospital (V-M.K.) and Academy of Finland (M.T.).

Ancillary