Snail, Slug, and Smad-interacting protein 1 as novel parameters of disease aggressiveness in metastatic ovarian and breast carcinoma

Authors

  • Sivan Elloul M.Sc.,

    1. Department of Pharmacology and Experimental Therapeutics, School of Pharmacy, Faculty of Medicine, The Hebrew University of Jerusalem, Jerusalem, Israel
    Search for more papers by this author
  • Mari Bukholt Elstrand M.D.,

    1. Department of Gynecologic Oncology, The Norwegian Radium Hospital, University of Oslo, Oslo, Norway
    Search for more papers by this author
  • Jahn M. Nesland M.D., Ph.D.,

    1. Department of Pathology, The Norwegian Radium Hospital, University of Oslo, Oslo, Norway
    Search for more papers by this author
  • Claes G. Tropé M.D., Ph.D.,

    1. Department of Gynecologic Oncology, The Norwegian Radium Hospital, University of Oslo, Oslo, Norway
    Search for more papers by this author
  • Gunnar Kvalheim M.D., Ph.D.,

    1. Department of Oncology, The Norwegian Radium Hospital, University of Oslo, Oslo, Norway
    Search for more papers by this author
  • Iris Goldberg Ph.D.,

    1. Department of Pathology, Sheba Medical Center, affiliated with the Sackler School of Medicine, Tel-Aviv University, Tel-Aviv, Israel
    Search for more papers by this author
  • Reuven Reich Ph.D.,

    1. Department of Pharmacology and Experimental Therapeutics, School of Pharmacy, Faculty of Medicine, The Hebrew University of Jerusalem, Jerusalem, Israel
    Search for more papers by this author
    • Reuven Reich, Ph.D. is affiliated with the David R. Bloom Center for Pharmacy at the Hebrew University.

  • Ben Davidson M.D., Ph.D.

    Corresponding author
    1. Department of Pathology, The Norwegian Radium Hospital, University of Oslo, Oslo, Norway
    • Department of Pathology, The Norwegian Radium Hospital, Montebello N-0310 Oslo, Norway
    Search for more papers by this author
    • Fax: (011) 47 22508554


  • The authors dedicate this study to the memory of Dr. Iris Goldberg, a close friend and a collaborator for many years.

  • Presented at the 10th International Congress of the Metastasis Research Society, Genoa, Italy, September 17–20, 2004.

Abstract

BACKGROUND

It was demonstrated previously that the Snail family of transcription factors and Smad-interacting protein 1 (Sip1) regulate E-cadherin and matrix metalloproteinase 2 (MMP-2) expression, cellular morphology, and invasion in carcinoma. For the current study, the authors analyzed the relation between the expression of Snail, Slug, and Sip1; the expression of MMP-2 and E-cadherin; and clinical parameters in patients with metastatic ovarian and breast carcinoma.

METHODS

One hundred one fresh-frozen, malignant effusions from patients who were diagnosed with gynecologic carcinomas (78 ovarian carcinomas and 23 breast carcinomas) were studied for mRNA expression of Snail, Slug, Sip1, MMP-2, and E-cadherin using reverse transcriptase-polymerase chain reaction analysis. Snail mRNA and E-cadherin protein expression levels also were studied in ovarian carcinoma effusions using in situ hybridization and immunocytochemistry. The results were analyzed for possible correlation with clinicopathologic parameters in both tumor types.

RESULTS

E-cadherin mRNA expression was lower in breast carcinoma (P = 0.001), whereas Snail expression was higher (P = 0.003). The Snail/E-cadherin ratio (P < 0.001) and the Sip1/E-cadherin ratio (P = 0.002) were higher in breast carcinomas. Sip1 mRNA expression (P < 0.001) and Slug mRNA expression (P < 0.001) were correlated with the expression of MMP-2 in ovarian carcinomas. The Sip1/E-cadherin ratio was higher in primary ovarian carcinomas at the time of diagnosis compared with postchemotherapy ovarian carcinoma effusions (P = 0.003), higher in Stage IV tumors compared with Stage III tumors (P = 0.049), and higher in pleural effusions compared with peritoneal effusions (P = 0.044). In a univariate survival analysis of patients with ovarian carcinoma, a high Sip1/E-cadherin ratio predicted poor overall survival (P = 0.018). High E-cadherin mRNA expression predicted better disease-free survival (P = 0.023), with a similar trend for a low Slug/E-cadherin ratio (P = 0.07). High Snail mRNA expression predicted shorter effusion-free survival (P = 0.008), disease-free survival (P = 0.03), and overall survival (P = 0.008) in patients with breast carcinoma.

CONCLUSIONS

Transcription factors that regulate E-cadherin were expressed differentially in metastatic ovarian and breast carcinoma. Snail may predict a poor outcome in patients who have breast carcinoma metastatic to effusions. E-cadherin expression generally was conserved in effusions from patients with ovarian carcinoma, but the subset of patients with postulated Sip1-induced repression of this adhesion molecule had a significantly worse outcome. This finding was in agreement with the stronger suppression of E-cadherin by Snail and Sip1 in breast carcinoma effusions, a clinical condition associated with extremely poor survival. Cancer 2005. © 2005 American Cancer Society.

Carcinoma is a disease of altered growth, differentiation, and tissue organization. These cellular activities are crucial for embryonic development and maintenance of proper structure and function of the mature organism. Disruption of these events, as evidenced in carcinoma, results in loss of tissue differentiation and promotes invasion and dissemination. The process of acquisition of the invasive phenotype by epithelial tumors can be regarded as a pathologic version of the epithelial to mesenchymal transition (EMT) of embryogenesis.

Local invasion and distant metastasis are the main determinants of carcinoma-related morbidity and mortality. Two critical cellular events involved in these processes are increased proteolytic activity and loss of adhesion.1 Matrix metalloproteinases (MMP), a family of zinc-dependent and calcium-dependent enzymes, are central mediators of the increased proteolysis during tumor progression through their ability to degrade basement membrane and extracellular matrix (ECM) components.2 Cadherins, a family of Ca2+-dependent integral membrane glycoproteins, are located at the cell-cell adherens junctions, where they mediate homophilic contact with neighboring cells.3–5 Cadherins interact through their carboxy-terminal intracytoplasmic domain with β-catenin (88 kilodaltons [kD]) and γ-catenin (80 kD). These, in turn, bind to α-catenin, which is a 102-kD protein that links actin molecules.6 It has been shown that E-cadherin, a member of the family expressed in epithelial cells, is an inhibitor of invasion.7, 8 Abnormal (absent, reduced, or localized to cell compartments other than the cell membrane) expression of E-cadherin and catenins has been reported in various human malignancies and has been associated with tumor progression.9, 10 It has been shown that inactivation or down-regulation of E-cadherin expression occurs through genetic (mutations) and epigenetic (CpG-promoter hypermethylation, transcriptional regulation, and posttranslational modification) mechanisms.11, 12

However, studies of gastric and breast carcinomas have shown that down-regulation of E-cadherin expression in primary tumors may be followed by elevated expression in solid metastases.13, 14 In agreement with those studies, we previously reported up-regulation of E-cadherin in ovarian carcinoma effusions and solid metastases compared with corresponding primary tumors.15 These results, in addition to the rare occurrence of E-cadherin mutations in ovarian carcinoma,16 suggest that epigenetic factors regulate E-cadherin expression in ovarian carcinomas.

The Snail superfamily of transcription factors consists of several proteins that have four to six zinc-finger domains and share high homology across species.17, 18 It has been shown that Snail, which is the first identified member of the family, along with additional family members that are expressed in Drosophila (Worniu, Escargot), mediates events in mesoderm, neuroectoderm, and other organ development in the embryo.17 Recent research has shown a role for this family in mediating EMT, migration, invasion, and survival, cellular events that play a role in the progression of epithelial malignancies.18 Several studies have shown that Snail family members, including Snail and Slug, are able to repress the transcription of E-cadherin, increase MMP expression and activity, and mediate EMT and invasion in human and nonhuman cell lines.19–28 Another zinc-finger protein, Smad-interacting protein 1 (Sip1),which is a member of the crystallin-enhancer binding factor 1 family,29 is a DNA-binding transcriptional repressor that has been implicated in similar cellular events.26, 30 Analysis of human carcinomas showed that the expression of Snail correlates inversely with that of E-cadherin in solid breast carcinoma31, 32 and that it is associated with higher histologic grade and lymph node metastasis.31 Snail and Sip1 showed inverse correlations with E-cadherin expression in gastric carcinomas of the diffuse and intestinal types, respectively.33

Breast carcinoma and ovarian carcinoma are major contributors to tumor-related morbidity and mortality in women.34 Both tumors metastasize to the serosal surfaces and are associated with the formation of malignant effusions, but this event is seen more commonly both in general and as a presenting symptom in ovarian carcinoma.35 Ovarian carcinoma involves both the pleural and peritoneal cavities, whereas breast carcinoma spreads predominantly to the pleural cavity, where it is the etiology of approximately 25% of malignant effusions.36–38 In addition, the survival of patients with ovarian carcinoma can exceed 5 years after an initial presentation with ascites39, 40 but is extremely poor once a malignant effusion is diagnosed in breast carcinoma, with a median survival of 5 months41 and 11 months42 reported in 2 series. The expression and clinical role of Snail family members in ovarian carcinoma has not been studied to date.

In the current study, our objectives were to analyze the possible mechanism of E-cadherin transcriptional regulation in ovarian and breast carcinoma effusions, define tumor type-related patterns, and characterize different biologic and prognostic groups within our ovarian and breast carcinoma cohort. For these purposes, we analyzed the expression of Snail, Slug, Sip1, E-cadherin, and MMP-2 in effusions from patients with metastatic ovarian and breast carcinoma. We demonstrate frequent expression of all three transcription factors in both tumor types, with significantly lower E-cadherin and higher Snail expression in breast carcinomas. Furthermore, we report that the Snail/E-cadherin and Sip1/E-cadherin ratios differ between ovarian and breast carcinomas and that the Sip1/E-cadherin ratio is higher in more clinically advanced ovarian carcinomas (International Federation of Gynecology and Obstetrics [FIGO] Stage IV and pleural effusions), with lower values after chemotherapy. Finally, we present data suggesting that Snail may predict a poor outcome in patients with breast carcinoma who have pleural metastasis and that the Sip1/E-cadherin ratio may be a novel predictor of poor survival in patients with ovarian carcinoma.

MATERIALS AND METHODS

Ovarian Carcinomas

All specimens and relevant clinical data were obtained from the Department of Gynecologic Oncology, Norwegian Radium Hospital (Table 1). The analyzed material consisted of 78 fresh, nonfixed, malignant peritoneal effusions (n = 54 specimens) and pleural effusions (n = 24 specimens) that were obtained from 63 patients who were diagnosed with epithelial (predominantly serous) ovarian carcinoma (70 effusions), from 1 patient with serous carcinoma of the fallopian tube (1 effusion), and from 6 patients who were diagnosed with primary peritoneal carcinoma (7 effusions; total = 70 patients). Due to their closely linked histogenesis and phenotype, all of these tumors are referred to as ovarian carcinomas in the following sections.

Table 1. Clinicopathologic Data from the Ovarian Carcinoma Cohort (n = 70 patients)
CharacteristicNo. of specimens
  • FIGO: International Federation of Gynecology and Obstetrics; NA: not available.

  • a

    Includes four patients with clear cell carcinomas.

  • b

    Includes specimens from inoperable patients (four patients) and patients who underwent surgery in hospitals in which tumor grade was not scored and the primary tumor grade could not be accessed (six patients).

  • c

    Four patients had inoperable disease, and four patients had no record.

  • d

    One patient who had inoperable disease underwent limited biopsy to establish a diagnosis of malignancy.

  • ePrior to sampling for 78 effusions.

Age (yrs) 
 Range35–79
 Mean60
FIGO stage 
 Stage II1
 Stage III35
 Stage IV34
Grade 
 Grade 16
 Grade 222
 Grade 332a
 NAb10
Residual disease 
 < 2 cm34
 ≥ 2 cm28
 NAc8
Histology 
 Serous57
 Mucinous1
 Clear cell4
 Mixed epithelial4
 Undifferentiated3
 NAd1
Chemotherapyc 
 No39
 Yes39

Breast Carcinomas

The material consisted of 23 effusions from 21 female patients with histologically verified breast carcinoma. Tumor type was available for 20 specimens. These consisted of 19 infiltrating duct carcinomas and 1 lobular carcinoma. Full clinicopathologic data were available for 16 patients with duct carcinoma (Table 2). For these 16 patients, data also included progression-free survival (PFS), the time from primary diagnosis to the appearance of pleural effusions (effusion-free period [EFS]), and overall survival (OS).

Table 2. Clinicopathologic Data from 16 patients with Breast Carcinomaa
ParameterNo. of specimens
  • TNM: tumor, lymph node, and metastasis.

  • a

    Full clinicopathologic data were unavailable for five patients who underwent surgery in community hospitals and were referred to The Norwegian Radium Hospital the time they had tumor progression to effusion.

  • b

    Chemotherapy was either combined cyclophosphamide, methotrexate, and 5-fluorouracil or combined 5-fluorouracil, epirubicin, and methotrexate.

  • c

    Includes one patient who had a concomitant recurrence in the liver and one patient who had a recurrence in the lung.

Age at diagnosis (yrs) 
 Range33–73
 Mean46
Stage (TNM classification) 
 Stage I (T1N0M0)2
 Stage II (T1N1M0, T2N1M0, T3N0M0)7
 Stage III (T2N2M0, T3N1M0, T4N2M0)6
 Stage IV (T4N2M1)1
Grade 
 Grade 212
 Grade 34
Receptor status 
 Positive9
 Negative7
Surgery 
 No2
 Yes (with adjuvant chemotherapy)3
 Yes (without adjuvant chemotherapy)11
Chemotherapyb 
 No5
 Yes11
Radiotherapy 
 No6
 Yes9
 Unknown1
Site of first recurrence 
 Pleura4
 Bonec6
 Axilla3
 Chest wall1
 Regional lymph nodes1
 Disseminated1

All effusion specimens were submitted for routine diagnostic purposes to the Section of Cytology, Department of Pathology, Norwegian Radium Hospital during the period between June 1998 and August 2002. Effusions were processed immediately after tapping. Each diagnosis was established by evaluation of smears and cell block sections from formalin fixed, paraffin embedded pellets and then were characterized further using immunocytochemistry with broad antibody panels against carcinoma, mesothelial and leukocyte epitopes, as detailed previously,43, 44 Informed consent was obtained according to national Norwegian and institutional guidelines.

Reverse-Transcriptase Polymerase Chain Reaction Analysis

Total RNA from all 101 effusions was extracted using a commercial kit (Tri-Reagent; Sigma Chemical Company, St. Louis, MO), and 0.5 μg of total RNA were reverse-transcribed using M-MLV Reverse Transcriptase (Promega, Madison, WI) with incubation for 2 hours at 37 °C, followed by 5 minutes at 95 °C, and diluted to 1:5 with RNase-free water. Reverse-transcriptase polymerase chain reaction (RT-PCR) analysis was performed on cyclic DNA (cDNA) samples with a DNA thermal cycler (Eppendorf Mastercycler gradient; Eppendorf, Hamburg, Germany) using primer sets for Snail, Slug, Sip1, E-cadherin, MMP-2, and 28S.23, 33, 45, 46 Primer sequences were as follows: Snail: sense, 5′-TGCGCGAATCGGCGACCC-3′; antisense, 5′-CCTAGAGAACCGCTTCCCGCAG-3′ (product size, ≈ 600 base pairs [bp]); E-cadherin: sense, 5′-TCCATTTCTTGGTCTACGCC-3′; antisense, 5′-CACCTTCAGCCAACCTGTTT-3′ (product size, 361 bp); Sip1: sense, 5′-GCGGCATATGGTGACACA-3′; antisense, 5′-TGCCACTAAACCCGTGTGTA-3′ (product size, 466 bp); Slug: sense, 5′-GCCTCCAAAAAGCCAAACTACTA-3′; antisense, 5′-GTGTGCTACACAGCAGCC-3′ (product size, 888 bp); MMP-2: sense, 5′-CACCTACACCAAGAACTTCC-3′; antisense, 5′-AACACAGCCTTCTCCTCCTG-3′ (product size, 327 bp); and 28S: sense, 5′-GTTCACCCACTAATAGGGAACGTGA-3′; antisense, 5′-GGATTCTGACTTAGAGGCGTTCAGT-3′ (product size, 212 bp).

The cycle parameters were as follows: Snail: denaturation at 94 °C for 1 minute, annealing at 61 °C for 1 minute, and extension at 72 °C for 2 minutes for 37 cycles; E-cadherin: denaturation at 94 °C for 30 seconds, annealing at 60 °C for 30 seconds, and extension at 72 °C for 1 minute for 36 cycles; Sip1: denaturation at 95 °C for 15 seconds, annealing at 53 °C for 30 seconds, and extension at 72 °C for 20 seconds for 36 cycles; Slug: denaturation at 95 °C for 1 minute, annealing at 58 °C for 1 minute, and extension at 72 °C for 1 minute for 40 cycles; MMP-2: denaturation at 94 °C for 30 seconds, annealing at 55 °C for 1 minute, and extension at 72 °C for 1.5 minutes for 33 cycles; and 28S: denaturation at 94 °C for 15 seconds, annealing at 63 °C for 20 seconds, and extension at 72 °C for 10 seconds for 16 cycles. Products were separated on 1.5% agarose gels, isolated using the Invisorb® Spin DNA extraction kit (Invitek GmbH, Berlin, Germany), and sequenced. The HT-1080 fibrosarcoma cell line was used as control in all Snail, Slug, Sip1, and MMP-2 reactions. An endometrial carcinoma biopsy was used as control for the E-cadherin RT-PCR reaction. Gels were photographed by the Kodak EDAS 290 system (Eastman Kodak, Rochester, NY). Densitometer analysis of films was performed using a computerized image-analysis program (NIH IMAGE 1.62, 1999 version).

Snail, Slug, Sip1, E-cadherin, and MMP-2 mRNA levels were established by calculating the target molecule/28S ratio (all samples were scored for band size compared with a control sample that was derived from the HT-1080 cell line). All measurements for clinical materials were carried out at the exponential phase of the reaction, as verified prior to sample analysis. This was verified for each gene separately. Results are shown as the average of two independent measurements of the RT-PCR reaction for each gene.

Immunocytochemistry

Sixty ovarian carcinoma effusions (45 peritoneal, 15 pleural) were analyzed for E-cadherin protein expression using the HECD-1 antibody (Zymed Laboratory, Inc., San Francisco, CA). Sections were stained using the improved biotin-streptavidin amplified detection system (BioGenex, San Ramon, CA). Negative controls underwent a similar staining procedure, with the exclusion of primary antibody application, or were stained with mouse myeloma protein of the same isotype as the primary antibody used. An E-cadherin-expressing endometrial adenocarcinoma specimen was used as positive control.

In Situ mRNA Hybridization

Specific antisense oligonucleotide DNA probe for the Snail mRNA transcript25 was obtained from Sigma Chemical Company. The specificity of the probe was verified using a sense probe (Sigma Chemical Company). A poly d(T)20 oligonucleotide probe (Research Genetics, Huntsville, AL) was used to verify the integrity and lack of degradation of mRNA in each sample. cDNA probes were hyperbiotinylated. The stock dilution was diluted with probe dilutent (Research Genetics) immediately before use.

Sixty-five ovarian carcinoma effusions (48 peritoneal and 17 pleural) were analyzed using in situ hybridization (ISH). Tissue sections (4 μ thick) of formalin fixed, paraffin embedded specimens were mounted on ProbeOn Plus slides (Fisher Scientific, Pittsburgh, PA). Sectioning was performed in RNase-free water. ISH was carried out using the microprobe manual staining system (Fisher Scientific).47 Hybridization of the probes was carried out as described previously.48 Positive enzymatic reaction in this assay stained dark blue. An infiltrating duct carcinoma of the breast shown to express Snail mRNA in pilot studies was used as positive control in each hybridization reaction. Controls for endogenous alkaline phosphatase included treatment of the sample in the absence of the probe and the use of chromogen alone.

Evaluation of ISH and Immunohistochemical Results

Staining in tumor cells was scored as negative, focal, or diffuse, corresponding to staining in 0%, 1–20%, and 21–100% of cells, respectively. At least 500 cells were scored when present. No specimen contained < 100 tumor cells. The pathologist who scored the slides (B.D.) was blinded to any clinical details related to the patients.

Statistical Analysis

Statistical analysis was performed applying the SPSS-PC package (version 10.1; SPSS Inc., Chicago, IL). A probability < 0.05 was considered statistically significant. Clinical and pathologic data were available for the majority of patients, as detailed in Tables 1 and 2. Survival data were available for all 70 patients with ovarian carcinoma and 16 patients with breast carcinoma. The mean follow-up was 31 months (range, 1–80 months), and 81 months (range, 6–234 months) for patients with ovarian carcinoma and breast carcinoma, respectively. A comparative analysis of ovarian and breast carcinoma effusions that included Snail, Slug, and Sip1 expression and the Snail/E-cadherin, Slug/E-cadherin, and Sip1/E-cadherin ratios were performed using the Mann–Whitney U test. The association between Snail, Sip1, and Slug mRNA expression (the median was used to establish low and high expression levels) and MMP-2 (as a continuous variable) were performed similarly using the Mann–Whitney U test. Analysis of the association between RT-PCR expression results and clinicopathologic parameters in the ovarian carcinoma cohort were undertaken using the Mann–Whitney U test (for two categories [e.g., effusion site]) or the Kruskal–Wallis H test (for greater than two categories [e.g., histologic grade]). Univariate survival analyses were executed using the Kaplan–Meier method and the log-rank test. For these analyses, expression was grouped as low or high based on median values. For patients with ovarian carcinoma who had more than one effusion, only expression in the first specimen was analyzed. The parameters analyzed were DFS and OS for the ovarian carcinoma cohort and EFS, DFS, and OS for the breast carcinoma cohort.

RESULTS

Ovarian and breast carcinoma cells in effusions express different levels of E-cadherin mRNA but have comparable MMP-2 mRNA expression: Our first objective was to analyze E-cadherin expression in the studied material. E-cadherin mRNA was found in 75 of 78 ovarian carcinoma effusions, with expression levels showing a 100-fold variation (range, 5–580% of control level) (Fig. 1). Immunohistochemistry in ovarian carcinoma effusions confirmed these findings, with protein expression in 59 of 60 specimens (21 specimens with expression in 1–20% of cells, 38 specimens with expression in 21–100% of cells) (Fig. 2A–D). All breast carcinoma effusions were positive for E-cadherin, with expression levels showing a 33-fold variation (range, 8–251% of control level) (Fig. 1); however, E-cadherin expression was significantly lower than in ovarian carcinoma effusions (mean rank, 32 vs. 56; P = 0.001, Mann–Whitney U test) (Fig. 3A). These results suggest a more preserved E-cadherin expression in ovarian carcinoma compared with breast carcinoma at this anatomic site.

Figure 1.

Reverse transcriptase-polymerase chain reaction analysis of Snail, Slug, Smad-interacting protein 1 (Sip1), E-cadherin (E-cad), and matrix metalloproteinase 2 (MMP-2) mRNA expression in eight ovarian carcinoma effusions (seven peritoneal and one pleural) and seven breast carcinoma effusions (all pleural). In ovarian carcinomas, an inverse ratio between mRNA expression of Snail, Slug, and Sip1 family members and the expression of E-cadherin can be observed in some specimens (e.g., E-cadherin and all 3 factors in Lane 7, Snail and E-cadherin in Lane 8). The association is less evident with respect to MMP-2. In breast carcinoma specimens, Slug expression is seen in only three of seven effusions, whereas Snail and Sip1 expression is strong in four of seven specimens and six of seven specimens, respectively. Strong E-cadherin mRNA expression is seen in only two specimens and appears to be related negatively to Slug (Lanes 3 and 6) and Sip1 (Lane 3) expression, rather than to Snail expression. Similar to ovarian carcinoma, the association between Snail transcription factors and MMP-2 is less evident. All results were calculated as a ratio to the 28S expression values. P: positive control; N: negative control.

Figure 2.

E-cadherin protein and Snail mRNA expression in ovarian carcinoma cells from effusions. (A–C) Immunocytochemistry for E-cadherin in three effusions. The majority of carcinoma cells show strong membrane expression, which is pronounced most in areas of adhesion to neighboring cells (arrows). (D) Markedly reduced expression of E-cadherin protein is observed in another effusion. The majority of cells are negative (arrow), with few showing focal expression. (E) Snail mRNA expression is seen in a primary breast infiltrating duct carcinoma that was used as positive control for in situ hybridization (ISH) analysis. All tumor cells are positive (arrow). (F) Hybridization with a d(T) probe confirms the presence of preserved mRNA in a peritoneal effusion (arrow). (G) A Snail-negative ovarian carcinoma effusion. Cells are counterstained with nuclear fast red. (H, I) Hematoxylin and eosin (H&E) staining (H) and ISH (I) for Snail in an ovarian carcinoma effusion. Cells show moderate hybridization intensity (arrow). (J, K) H&E staining (J) and ISH (K) for Snail in another ovarian carcinoma effusion. Cells show strong hybridization intensity (arrow; ISH: nitroblue tetrazolium/5-bromo-4-chloro-3-inodolyphosphate staining, counterstained with nuclear fast red).

Figure 3.

E-cadherin (e-cad), Snail, Slug, and Smad-interacting protein 1 (Sip1) mRNA expression in ovarian carcinoma effusions (n = 78 samples) and breast carcinoma effusions (n = 23 samples). Values on the Y-axis represent expression relative to control sample value. (A and B) Significant differences in the expression of E-cadherin and Snail in the two tumor types are shown. Ovarian carcinomas express significantly more E-cadherin (P = 0.001) and less Snail (P = 0.003). (C and D) The distribution of expression for Slug and Sip1 is shown. Differences for these two factors did not show a significant difference (all analyses were performed using the Mann–Whitney U test).

MMP-2 mRNA expression was found in 76 of 78 ovarian carcinoma effusions (range, 2–486% of control value) (Fig. 1) and in all breast carcinoma effusions (range, 12–313%) (Fig. 1). These levels did not differ statistically (P = 0.18, Mann–Whitney U test).

Snail, Slug, and Sip1 mRNA Is Expressed Frequently but Variably in Ovarian and Breast Carcinoma Effusions

In this analysis, we compared the expression levels of these three transcription factors in our ovarian and breast carcinoma specimens. RT-PCR showed Snail mRNA expression in 68 of 78 ovarian carcinomas (range, 1–1059% of control value) (Fig. 1) and in 22 of 23 breast carcinomas (range, 0–575% of control value) (Fig. 1), with significantly lower expression in ovarian carcinomas (mean rank, 46 vs. 67; P = 0.003, Mann–Whitney U test) (Fig. 3B). ISH showed Snail expression in 43 of 65 ovarian carcinoma effusions (30 specimens with expression in 1–20% of cells, 13 specimens with expression in 21–100% of cells) (Fig. 2E–K). Slides that were hybridized with a sense probe were negative. Slug mRNA was found in 56 of 78 ovarian carcinomas (range, 1–478% of control value) (Fig. 1) and in 13 of 23 breast carcinomas (range, 2–528% of control value) (Fig. 1) using RT-PCR. This difference was not significant (P = 0.2, Mann–Whitney U test) (Fig. 3C). Sip1 mRNA was expressed in 76 of 78 ovarian carcinomas (range, 5–1990% of control) (Fig. 1) and all breast carcinomas (range, 71–847% of control) (Fig. 1). The difference in expression was not significant (P = 0.5, Mann–Whitney U test) (Fig. 3D). These data suggest that expression of the transcription repressors Slug and Sip1 is comparable in metastatic breast and ovarian carcinomas, whereas Snail shows higher expression in breast carcinoma.

The More Pronounced Suppression Of E-Cadherin mRNA Expression in Breast Carcinoma Compared with Ovarian Carcinoma Cells in Effusions May Be Related to Sip 1 and Snail, but Not to Slug

It has been shown that Snail transcription factors and Sip1 repress E-cadherin expression. In the current analysis, we studied the relative expression of E-cadherin as a function of Snail family member expression by analyzing the ratio between their values. Statistical analysis showed significantly higher Sip1/E-cadherin ratio (mean rank, 66 vs. 45; P = 0.002, Mann–Whitney U test) and Snail/E-cadherin ratio (mean rank, 74 vs. 42; P < 0.001, Mann–Whitney U test) in breast carcinomas compared with ovarian carcinomas. The Slug/E-cadherin ratio was comparable in both tumors (P = 0.5, Mann–Whitney U test). These data suggest a greater degree of E-cadherin suppression by Sip1 and Snail in breast carcinoma effusions compared with effusions that originate from ovarian carcinoma. The data related to expression differences between ovarian carcinomas and breast carcinomas are summarized in Table 3.

Table 3. Expression Values and Ratios of the Analyzed Molecules in Ovarian and Breast Carcinoma Effusionsa
RatioOvarian carcinomaBreast carcinomaP value
  • Sip1: Smad-interacting protein 1.

  • a

    Mean rank values are shown (Mann–Whitney test).

Snail46670.003
Slug52430.2
Sip149540.5
E-cadherin56320.001
Matrix metalloproteinase 253430.2
Snail/E-cadherin4274< 0.001
Slug/E-cadherin50460.5
Sip1/E-cadherin45660.002

Slug and Sip1 Transcription Factor Expression Correlates with MMP-2 Expression in Ovarian Carcinoma

It has been shown that Snail transcription factors induce the transcription of MMP-2. Therefore, we analyzed the association between these molecules in metastatic ovarian and breast carcinomas. Higher expression (cut-off at the median value) of Sip1 mRNA (mean rank, 48 vs. 28; P < 0.001, Mann–Whitney U test) and Slug mRNA (mean rank, 50 vs. 29; P < 0.001, Mann–Whitney U test) was correlated with higher MMP-2 mRNA expression in ovarian carcinomas. This correlation was not found in breast carcinomas. These data suggest that Sip1 and Slug may be involved in transcriptional regulation of MMP-2 in ovarian carcinoma.

Sip1 Expression and Suppression of E-Cadherin Correlate with Chemotherapy Status and Clinicopathologic Parameters in Patients with Ovarian Carcinoma

We previously showed that expression levels of growth factors49 and signal transduction molecules50 are altered after chemotherapy in patients with ovarian carcinoma, possibly due to the selective death of tumor subpopulations or the specific effect of chemotherapy on the biochemistry of tumor cells. In the current analysis, we compared ovarian and breast carcinoma effusions that were obtained at primary diagnosis with postchemotherapy specimens. Expression in breast carcinoma effusions showed no association with chemotherapy status, receptor status, histologic grade, or disease stage (Mann–Whitney U test, data not shown). In ovarian carcinomas, Sip1 mRNA expression (mean rank, 46 vs. 33; P = 0.011, Mann–Whitney U test) and the Sip1/E-cadherin ratio (mean rank, 46 vs. 31; P = 0.003, Mann–Whitney U test) were significantly higher in effusions that were obtained prior to the administration of chemotherapy compared with disease recurrence specimens. The reduced expression of Sip1 mRNA and the Sip1/E-cadherin ratio correlated with treatment with both platinum agents (P = 0.018 and P = 0.008, respectively, Mann–Whitney U test) and paclitaxel (P = 0.013 and P = 0.009, respectively, Mann–Whitney U test). Expression of MMP-2, E-cadherin, Slug, and Sip1 and the Snail/E-cadherin and Slug/E-cadherin ratios were comparable (data not shown). Analysis of the relations between expression data and clinicopathologic parameters showed significantly lower E-cadherin expression (P = 0.023, Mann–Whitney U test), a higher Sip1/E-cadherin ratio (P = 0.049, Mann–Whitney U test) in FIGO Stage IV tumors compared with Stage III tumors, and a higher Sip1/E-cadherin ratio in pleural effusions compared with peritoneal effusions (P = 0.044, Mann–Whitney U test). These data suggest that Sip1 expression and the suppression of E-cadherin may be related to disease progression and that preferential targeting of Sip1-expressing tumor cell populations by chemotherapy may occur in effusions.

Snail mRNA Expression Level in Breast Carcinoma Effusions Predicts Poor DFS, EFS, and OS

We recently showed that breast carcinoma cells in effusions have a unique pattern of protein and mRNA expression compared with the corresponding solid lesions, and this pattern is relevant for predicting disease progression to effusion.51, 52 In the current study, we used recently updated clinicopathologic data to analyze the predictive role of Snail, Slug, Sip1, E-cadherin, and MMP-2 and their expression ratios in this cohort. In univariate survival analysis, high Snail mRNA expression predicted shorter DFS (P = 0.03) (Fig. 4A), EFS (P = 0.008) (Fig. 4B), and OS (P = 0.008) (Fig. 4C). Histologic grade was the sole clinicopathologic predictor in this patient group (P = 0.007 for EFS and P = 0.004 for OS; data not shown). These data suggest that Snail may be a novel prognostic marker in metastatic breast carcinoma.

Figure 4.

Snail mRNA expression in breast carcinoma cells in effusions predicts poor survival. (A) This Kaplan–Meier survival curve shows the correlation between Snail expression and disease-free survival (DFS). Patients who had specimens that showed lower Snail mRNA levels (n = 8 patients; solid line) had a mean DFS of 96 months (median, 34 months) compared with 22 months (median, 24 months) for patients who had specimens that showed high Snail expression (n = 8 patients; dashed line; P = 0.03). (B) A similar analysis was performed for effusion-free survival (EFS). Patients who had specimens that showed lower Snail expression (n = 8 patients; solid line) had a mean EFS of 127 months (median, 111 months) compared with 36 months (median, 26 months) for patients who had specimens that showed low Snail expression (n = 8 patients; dashed line; P = 0.008). (C) This Kaplan–Meier survival curve for overall survival (OS) shows that patients who had specimens that showed low Snail expression (n = 8 patients; solid line) had a mean OS of 130 months (median, 96 months) compared with 42 months (median, 31 months) for patients who had specimens that showed high Snail expression (n = 33 patients; dashed line, P = 0.008).

E-Cadherin and the Sip1/E-Cadherin Ratio Predict Survival in Ovarian Carcinoma

In the last analysis, we evaluated the potential value of the studied molecules in predicting survival in ovarian carcinoma cells in effusions. In a univariate survival analysis of patients with ovarian carcinoma, a high Sip1/E-cadherin (P = 0.018) ratio predicted poor survival (Fig. 5A). E-cadherin mRNA expression predicted better disease-free survival (DFS) (P = 0.023) (Fig. 5B), with a similar trend for a low Slug/E-cadherin ratio (P = 0.07) (Fig. 5C). These data suggest that the Sip1/E-cadherin ratio is a novel prognostic marker in patients with ovarian carcinoma.

Figure 5.

Expression of E-cadherin, Slug, and Smad-interacting protein 1 (Sip1) in ovarian carcinoma cells in effusions predicts survival. (A) This Kaplan–Meier survival curve shows the correlation between the Sip1/E-cadherin ratio and overall survival (OS). Patients who had specimens with a high Sip1/E-cadherin ratio (n = 34 patients; dashed line) had a mean OS of 30 months (median, 24 months) compared with 44 months (median, 40 months) for patients who had specimens with a low Sip1/E-cadherin ratio (n = 34 patients; solid line; P = 0.018). The 2 remaining patients in the cohort had E-cadherin-negative tumors and therefore were not analyzed for the Sip1/E-cadherin ratio. (B) This Kaplan–Meier survival curve shows the correlation between E-cadherin mRNA expression and disease-free survival (DFS). Patients who had specimens that showed higher E-cadherin expression (n = 39 patients; solid line) had a mean DFS of 17 months (median, 10 months) compared with 7 months (median, 3 months) for patients who had specimens that showed low E-cadherin expression (n = 31 patients; dashed line; P = 0.023). (C) This Kaplan–Meier survival curve shows the trend toward a correlation between the Slug/E-cadherin ratio and DFS. Patients who had specimens with a high Slug/E-cadherin ratio (n = 35 patients; dashed line) had a mean DFS of 10 months (median, 6 months) compared with 20 months (median, 10 months) for patients who had specimens with a low Slug1/E-cadherin ratio (n = 33 patients; solid line; P = 0.07). The 2 remaining patients in the cohort had E-cadherin-negative tumors and, thus, were not analyzed for the Sip1/E-cadherin ratio.

DISCUSSION

E-cadherin expression in carcinoma is variable and appears to depend on both tumor type and site. Analysis of E-cadherin expression in malignant effusions from patients with gastric and pancreatic carcinoma showed reduced expression of this adhesion molecule in the majority of specimens.53 In contrast, three recent studies showed frequent expression of this molecule in carcinoma cells of various origins (breast, lung, ovary, gastrointestinal tract) in effusions.15, 54, 55 It is interesting to note that an analysis of 4 breast carcinoma effusions showed staining in < 10% of cells in all specimens.54 In the current study, E-cadherin mRNA expression was found in all breast carcinoma effusions, as anticipated for tumors that originated predominantly from ductal carcinomas and in agreement with previous studies of solid breast carcinomas (for review, see Berx and van Roy56). However, E-cadherin mRNA levels were significantly lower in breast carcinoma effusions compared with ovarian carcinoma effusions. This finding is in agreement with the morphology of these two diseases in effusions, because dissociation into single cells is observed more frequently in breast carcinomas, many of which fail to show the classical “cell ball” morphology. Ovarian carcinomas, conversely, tend to have cohesive groups in all but the most poorly differentiated tumors. These differences in morphology and E-cadherin expression suggest differential regulation of E-cadherin expression by various transcription factors. In agreement with this hypothesis, we found significantly higher Snail mRNA expression in effusions from patients with breast carcinoma.

We hypothesized that the ratio between Snail transcription factor and E-cadherin expression may reflect the degree of repression better, because an inverse correlation between these values was shown previously in breast and gastric carcinomas.28, 31, 32 An analysis of these parameters showed significantly higher Sip1/E-cadherin and Snail/E-cadherin ratios in breast carcinoma effusions compared with their ovarian counterparts. These results suggest that Sip1 and Snail may be the transcription factors responsible for E-cadherin suppression in breast carcinoma, which also appears to be the case with gastric carcinomas.32

MMP-2 is a marker of poor prognosis in patients with solid ovarian carcinoma,48 and its expression is elevated in effusions from patients with ovarian carcinoma57 and breast carcinoma52 compared with primary tumors. It was shown recently that Snail and Sip1 induce the synthesis of several MMPs in vitro in hepatocellular carcinoma,26 and Snail had the same effect on MMP-2 in squamous carcinoma cell lines.27 The current data show a correlation between the expression of Slug and Sip1 and MMP-2 mRNA levels in ovarian carcinoma effusions and suggest a possible role for Snail transcription factors in MMP-2 regulation in effusions.

We found extreme differences in mRNA expression levels for all molecules analyzed in the current study using RT-PCR. In our view, these differences are due to the sensitivity of the densitometric analysis, which detects very low expression levels (i.e., < 10% of control) that would have been scored as negative by the naked eye. An additional factor contributing to these differences is the handling procedure for effusion material in our laboratory. Specimens arrive within minutes after tapping and are frozen immediately. This fact, added to the ability of cells to remain viable longer in effusions compared with biopsies after removal from the patient, contributes to the ability to detect very high mRNA expression levels in relevant specimens. Previous studies from our group documented that this also was true with respect to other sensitive epitopes, such as phosphorylation sites on activated proteins.50

The serosal cavities are unique anatomic sites that differ from solid organs in terms of the microenvironment and the growth/survival conditions of tumor cells. In recent years, along with others, we have shown that the clinical and prognostic role of many molecules is different in effusions compared with solid tumors (for review, see Davidson et al.58) and that the administration of chemotherapy may alter the expression of key molecules in ovarian carcinoma.48, 49 In agreement with the above-mentioned studies, we found that the Sip1/E-cadherin ratio and Sip1 mRNA expression were significantly lower in postchemotherapy specimens that were obtained at the time of disease recurrence. Both cisplatin, which induces the formation of DNA adducts, and microtubule-targeted antitumor agents, such as paclitaxel, exert their action through intracellular signaling pathways that are related to cell-cycle proteins and apoptosis/death signals.59, 60 Tumor cells develop resistance to these drugs by mechanisms that include activation of the prosurvival PI3K/Akt signaling pathway, inactivation of p53, and overexpression of HER2/neu and Bcl-2.59, 60 Although the current data do not prove any direct link between chemotherapy and the transcription factors studied, they may suggest that cells with higher expression of these molecules are targeted preferentially by chemotherapy, a finding that may be related to higher proliferation. The lower E-cadherin expression level and the higher Sip1/E-cadherin ratio found in FIGO Stage IV tumors, compared with Stage III tumors, and the higher Sip1/E-cadherin ratio in pleural effusions compared with peritoneal effusions, support a relation between Sip1 expression and suppression of E-cadherin and disease progression.

In the survival analysis, higher expression of E-cadherin in ovarian carcinoma effusions predicted better DFS, whereas Snail expression in breast carcinoma effusions predicted worse DFS, EFS, and OS. Both findings were in agreement with the aggressive behavior of epithelial cells undergoing EMT, a process characterized by the loss of E-cadherin and the acquisition of N-cadherin in which it has been shown that Snail plays a role (for review, see De Wever and Mareel61). In addition, a higher Sip1/E-cadherin ratio predicted poor survival in patients with ovarian carcinoma. These data are the first to document a correlation between survival and these transcription factors in human malignancies and are in agreement with the more aggressive clinical behavior reported in the form of lymph node metastasis related to Snail expression in breast carcinoma.28 Because the Sip1/E-cadherin ratio was significantly higher in breast carcinoma effusions compared with the entire ovarian carcinoma cohort, a clinical condition associated with extremely poor survival, we hypothesize that a high Sip1/E-cadherin ratio may characterize a subset of patients with ovarian carcinoma who have a similarly poor prognosis.

The results of the current study documented frequent expression of the Snail, Slug, and Sip1 transcription factors, with a possible role in regulating E-cadherin and MMP-2 expression in ovarian and breast carcinoma metastatic to effusions. Differences in the expression of these factors and their ability to repress E-cadherin may provide one molecular explanation for the different clinical significance of metastasis to effusions in these malignancies. Snail is a novel predictor of poor survival in patients with breast carcinoma, and there is a similar role for the Sip1/E-cadherin ratio in patients with ovarian carcinoma.

Acknowledgements

The authors gratefully acknowledge the competent technical help of Mrs. Ellen Hellesylt at the Department of Pathology, Norwegian Radium Hospital, in performing the immunohistochemical analysis.

Ancillary