Main

Primary peritoneal carcinoma is an uncommon disease that occurs exclusively in women and diffusely involves both abdominal and pelvic peritoneum. Primary peritoneal carcinoma is similar to ovarian serous carcinoma histologically, however it is characterized by minimal surface ovarian involvement or minimal surface invasion of the ovarian cortex. The reported incidence of primary peritoneal carcinoma is 10–15% that of ovarian serous carcinoma.1, 2, 3 Although the recognition of primary peritoneal carcinoma is increasing its pathogenesis remains obscure. Chromosome 17 has been implicated as a potential location of genetic events important in the pathogenesis of both tumour types.4, 5, 6 High frequencies of loss of heterozygosity have been described at chromosomal band 17p13.1 (the genomic locus for p53 and microRNA cluster miR-195 and miR-497) in both tumour types.4 Primary peritoneal carcinoma exhibits distinct allelic loss patterns in comparison to ovarian serous carcinoma, implying that different sets of tumour-suppressor genes may be involved in the development of both tumour types.5

MicroRNAs are a class of small non-coding RNAs, approximately 22 nucleotides long that have been found to negatively regulate gene expression. They have been found to have roles in cell growth, differentiation, apoptosis and tumourigenesis.7, 8, 9, 10, 11, 12, 13, 14 Furthermore, unique microRNA expression profiles have been able to classify various cancers. Recently, microRNAs were implicated in the development of ovarian cancer: 39 microRNAs are differentially regulated between tumour and normal ovarian tissue.9

MicroRNA production involves cleavage of a long nascent transcript (primary microRNA) to form a 70–100 nucleotide hairpin precursor (precursor microRNA), which is in turn processed to form a microRNA duplex. One strand of this duplex is incorporated into the RISC complex where it will bind through partial sequence homology to the 3′ UTR of target mRNAs causing their translational repression.15, 16, 17 A variety of studies have linked aberrant microRNA expression to carcinogenesis where they can potentially act as both oncogenes and tumour-suppressor genes. Individual microRNAs such as miR-15 and miR-16 that target bcl-2 are frequently deleted and downregulated in CLL and thus can act as tumour-suppressor genes.7, 18

The purpose of this study was to examine the differential expression of a select group of microRNAs and their confirmed targets in primary peritoneal carcinoma relative to ovarian serous carcinoma. These microRNAs were chosen based on cluster site (17p13.1), their potential target pathways and on their relative fold changes as determined in a pilot study (unpublished observations, RJ Flavin). Furthermore, DNA copy number alterations at genomic loci for selected microRNAs have been described in ovarian serous carcinoma.19

Materials and methods

Ethical approval for the study was obtained from the St James's and Federation of Dublin Voluntary Hospitals ethics committee.

Case Selection and Tumour Sample Preparation

In total, 34 matched serous tumours (n=17 for primary peritoneal carcinoma and serous ovarian carcinoma) of advanced FIGO stage and grade were selected from archival formalin-fixed, paraffin-embedded tissue, between the years 1991 and 2006 from St James's Hospital, Dublin. All tumour samples were taken from the ovary. H&E slides of all tumours were reviewed by a histopathologist (RF) and original diagnoses confirmed. Primary peritoneal carcinoma was identified according to the Gynaecological Oncology Group criteria as previously described20 namely: (1) both ovaries were of normal size or enlarged by a benign process, (2)involvement of the extra-ovarian sites was greater than involvement on the surface of either ovary and (3) microscopically the ovarian component was either: non-existent, confined to the ovarian surface epithelium or involved the ovarian surface epithelium and underlying cortical stroma with tumour size not more than 5 × 5 mm.

Archival blocks were selected that contained over 90% tumour with contaminating stromal tissue estimated to be not more than 10%. Those tumours with >10% stromal contamination (n=10) were laser capture microdissected, using the PixCell 11â„¢ System (Arcturus Engineering Inc., CA, USA), to yield homogenous populations of malignant epithelial cells lining the papillae, as previously described.21

RNA was extracted from archival material using RecoverAll™ Total Nucleic Acid Isolation Kit (Ambion Ltd, Cambridgeshire, UK) following the manufacturer's protocol. RNA quantity was assessed using a Nanodrop® ND-1000 Spectrophotometer (Wilmington, USA). Table 1 lists the clinicopathological characteristics of the cases selected. The paraffin tissue microarray composed of 68 serous tumours arrayed in quadruplicate (n=30 for primary peritoneal carcinoma; n=38 for ovarian serous carcinoma).

Table 1 Clinicopathological characteristics of tumour cohort
Full size table

Stem-Loop RT-PCR

Applied Biosystems TaqMan® microRNA assays are designed to detect and quantify mature microRNAs using looped-primer real time PCR. All assays were taken from an early access panel, which contained 330 individual assays for identified human microRNAs (AB). Primer sequences are available on demand. Assays included: miR-16, miR-195 and miR-497. The protocol involves three steps: reverse transcription, pre-PCR amplification and real-time PCR, and has been previously described.23 Each RT reaction contained 10 ng of total RNA, and assays were gene dose corrected using let-7 as endogenous control.

Immunohistochemical Stains and Statistical Analysis

Sections (4 μm) of the ovarian serous carcinoma tissue array were cut and mounted on glass slides. For antigen unmasking, heat-mediated antigen retrieval was performed on deparaffinized sections using Triology (p53, bcl-2) in citrate buffer (10 mmol/L sodium citrate buffer (pH 6.0)) before incubation with primary antibodies. p53 (Vision Biosystems, Newcastle, UK) and bcl-2 (Dakocytomation, Copenhagen, Denmark) protein levels were examined using mouse monoclonal antibodies at dilutions of 1:50 and 1:30, respectively. Antibody staining was performed using the Ventana Nexus automated system (Tucson, Arizona, USA) using the avidin–biotin procedure.

Stained TMA slides were graded jointly by two pathologists (RF, CB). A modified visual semiquantification method was used as previously described,24 using a two-score system for immunointensity (II) and immunopositivity (IP). II and IP scores were summated. The semiquantification for II was scored on a scale of: 0, negative; 1, weak; 2, moderate and 3, strong. The semiquantification for percentage of IP cells was scored on a scale of 1 (1–10%), 2 (11–40%), 3 (41–70%) and 4 (>70%). This produced immunoscore values ranging from 0 to +7. Scores from all cores from one case were averaged. A cutoff value of ⩾5 was used to determine immunohistochemical positivity.

MicroRNA Target Prediction

The analysis of microRNA clusters and predicted targets were determined using the miRGen algorithm (http://www.diana.pcbi.upenn.edu/miRGen). Predicted targets of miR-195 and miR-497 were determined by union of target programs: DIANA-microT, miRanda, miRBase, PicTar and TargetScans.

Gene Ontology Analysis

Gene ontology analysis was performed using an online database known as the Panther Classification System (http://www.pantherdb.org). A binomial statistical tool was employed to compare the gene list to a reference list (ie the complete human genome) to determine over- or underrepresentation of PANTHER classification categories.

Statistical Analysis

Real-time stem-loop RT-PCR data analysis was performed using Real-Time StatMiner™ software from Integromics™ (www.integromics.com). Fold changes were calculated on filtered and ΔΔCT method normalized data using the CT method. P-values were calculated using a t-test.

The similarity in associations between tumour histotype and clinicopathological characteristics were examined by means of Fisher's exact test (Analyse-It™ Software Ltd) for discontinuous variables and a t-test for parametric continuous variables. The association between histotype and protein expression levels (p53 and bcl-2) were examined by means of a Mann–Whitney test.

Kaplan–Meier estimates of survival time were compared using the logrank test (MedCalc™ Software Ltd). All tests were two-tailed, and the significance level was set at P<0.05.

Results

Real-Time Quantitative Stem-Loop RT-PCR Analysis of Differential MicroRNA in Primary Peritoneal Carcinoma Relative to Ovarian Serous Carcinoma

We used real-time stem-loop RT-PCR analysis to examine 17 matched cases for differential expression of miR-16, miR-195 and miR-497 in primary peritoneal carcinoma relative to ovarian serous carcinoma. All microRNAs were downregulated (Figure 1). miR-16 was downregulated by 4.42-fold (P=0.02) and miR-195 by 7.75-fold (P=0.006), whereas miR-497 was downregulated by 1149 fold (P=1.38E-11).

Figure 1
figure 1

Mean logarithmic fold change (RQ) of microRNAs in primary peritoneal carcinoma (PPC) relative to ovarian serous carcinoma (OSC). miR-16 (P=0.02); miR-195 (P=0.006); miR-497 (P=1.38E-11).

Full size image

Immunohistochemical Analysis of p53 Expression in Primary Peritoneal Carcinoma Relative to Ovarian Serous Carcinoma

Our next investigation was to examine the protein expression levels of p53 and to see if there was any difference in relative expression between primary peritoneal carcinoma relative to ovarian serous carcinoma. This was undertaken as p53 is found at the same genomic locus on chromosomal band 17p13.1 as the microRNA cluster miR-195 and miR-497.

Background normal ovarian surface epithelium was predominantly non-reactive for p53. p53 overexpression was detected in 63% (19/30) of primary peritoneal carcinoma vs 84% (32/38) of ovarian serous carcinomas (Figure 2). In primary peritoneal carcinoma, the mean p53 immunoreactivity was 4.6 and in ovarian serous carcinoma, the mean p53 immunoreactivity was 6.0. The difference in p53 immunoreactivity between both tumour types was statistically significant (P=0.04). The results of the immunohistochemical analysis of p53 expression are summarized in Table 2.

Figure 2
figure 2

Protein expression of p53 and bcl-2 in tissue microarray: No staining (‘0’) for p53 (a) and bcl-2 (b) in normal ovarian surface epithelium ( × 20). Strong (‘3’) staining for p53(c) and bcl-2 (d) in primary peritoneal carcinoma and ovarian serous carcinoma ( × 200).

Full size image
Table 2 Summary of protein expression of p53 and bcl-2 in primary peritoneal carcinoma and ovarian serous carcinoma
Full size table

Immunohistochemical Analysis of bcl-2 Expression in Primary Peritoneal Carcinoma Relative to Ovarian Serous Carcinoma

Our next investigation was to examine the protein expression levels of bcl-2 and to see if there was any difference in relative expression between primary peritoneal carcinoma and ovarian serous carcinoma. This was undertaken as bcl-2 is a validated target of miR-1625 and downregulation of this microRNA should be accompanied by an increased expression of its protein target.

Background normal ovarian surface epithelium was predominantly non-reactive for bcl-2. bcl-2 expression was detected in 20.69% (6/29) of primary peritoneal carcinoma vs 21.05% (8/38) of ovarian serous carcinomas (Figure 2). In primary peritoneal carcinoma, the mean bcl-2 immunoreactivity was 5.2 and in ovarian serous carcinoma, the mean bcl-2 immunoreactivity was 3.2. The difference in bcl-2 expression between both tumour types was statistically significant (P=0.02). The results of the immunohistochemical analysis of bcl-2 expression are summarized in Table 2.

Potential Targets of miR-195 and miR-497

Using the miRGen database, we looked for putative targets of miR-197 and miR-495 (no targets have been experimentally validated currently). Results were incorporated into PANTHER for gene ontology classification of the predicted mRNA targets. The cadherin, T-cell activation and WNT signalling pathways were significantly (P<0.01) overrepresented relative to a reference list. Potential targets include E-cadherin, α-catenin, cyclin D1 and cyclin D2. Table 3 lists the putative gene targets in significantly overrepresented pathways.

Table 3 Potential genes targeted by miR-195 and miR-497 in the cadherin, WNT signalling and T-cell activation pathways
Full size table

Discussion

In this study, we evaluated the expression of a select group of microRNAs and their target genes in primary peritoneal carcinoma relative to matched cases of ovarian serous carcinoma. Specifically, we show decreased p53 expression and downregulation of miR-195 and miR-497 from the microRNA cluster site at chromosome 17p13.1 in primary peritoneal carcinoma relative to ovarian serous carcinoma. Previous data have indicated that chromosome 17 may be a potential location of genetic events important in the pathogenesis of both tumour types.4, 5, 6 Throughout chromosome 17 frequent genetic deletions have been identified, suggesting that multiple tumour-suppressor genes may exist on this chromosome.26, 27, 28, 29 Our data therefore implicate miR-195 and miR-497 as potential tumour-suppressor genes.

Since its first description by Swerdlow in 1959, primary peritoneal carcinoma has been treated similarly to ovarian serous carcinoma by surgery and chemotherapy.30, 31 Despite histological and clinical similarities, some significant differences are seen in clinical, epidemiological and molecular characteristics of the two tumours, suggesting primary peritoneal carcinoma may be a distinct disease entity. Patients with primary peritoneal carcinoma tend to be older,32, 33 and when suboptimally debulked may have better survival rates than ovarian serous carcinoma patients.3, 34, 35 The prognosis of primary peritoneal carcinoma compared to ovarian serous carcinoma of the same stage is controversial with some studies reporting worse, same or better survival.32, 36, 37 A more recent matched case-comparison study demonstrated platinum resistance and impaired survival in patients with advanced primary peritoneal carcinomas.37 Of note, a recent epidemiological study found that parity and obesity are associated with an increased risk of peritoneal cancer, unlike ovarian serous cancer where there is an inverse correlation.38 The somewhat different patterns in terms of risk for primary peritoneal cancer suggests that primary peritoneal cancer may develop along different molecular pathways. Furthermore, differences in loss of heterozygosity,4, 5, 39 gene expression patterns40, 41 and cDNA copy number gains and loss have also been demonstrated between the two tumours possibly reflecting different carcinogeneic pathways.

Previous studies have suggested serous ovarian cancer may have a unifocal origin42, 43, 44, 45, 46 and that primary peritoneal cancer may have a multifocal origin.4, 40, 41, 47 Recent work proposes that epithelial ovarian cancer arises from three proposed origins, including the ovarian surface epithelium or mullerian inclusions, fallopian tube mucosa and mullerian epithelium elsewhere in the peritoneal cavity.2, 48, 49, 50, 51 Specifically, there is a growing body of evidence to suggest that both ovarian serous carcinoma and primary peritoneal carcinoma may arise from a precursor lesion in the distal tubal fimbria,49, 50, 51 and thus both tumour types may be genetically related. Importantly, our results do not preclude a common site of origin; they probably reflect temporal and spatial heterogenous microRNA expression in both tumour types, which maybe a consequence of biological properties such as preferential site of growth and spread.

Four recent studies reported aberrantly expressed microRNAs in human ovarian cancer.9, 14, 19, 52 Zhang et al19 described DNA copy number alterations at all genomic loci for our selected microRNAs in ovarian serous carcinoma. In the study by Iorio et al,9 miR-181a and miR-195 were downregulated in ovarian serous carcinoma, whereas Nam et al52 demonstrate differential expression of miR-16. Yang et al14 did not describe dysregulation of any of our selected microRNAs, however their group used microRNA microarrays (which are less sensitive and have a smaller dynamic range in comparison to PCR) to generate pilot data. Furthermore, a different RNA extraction protocol (Trizol Invitrogen) was used.

Interestingly, Iorio et al9 looked at differential expression of microRNAs in ovarian tumours with surface involvement and pelvic peritoneal involvement. miR-101, miR-182*, miR-22 and miR-133a were upregulated in those tumours with ovarian surface involvement, whereas miR-302c was upregulated in those tumours with pelvic peritoneal involvement. Though not stated in their study it would seem that ovarian surface and pelvic peritoneal involvement does not satisfy the Gynaecological Oncology Group diagnostic criteria for primary peritoneal carcinoma, rather reflecting pelvic and surface involvement by an ovarian serous carcinoma primary. Our data demonstrate decreased p53 expression and downregulation of a select group of microRNAs in tumours meeting Gynaecological Oncology Group criteria for primary peritoneal carcinoma relative to ovarian serous carcinoma. As the tissues studied were all taken from ovarian samples, it is possible that the differences observed are related to tumour progression (eg the peritoneal tumours began as p53 positive, miR-16, miR-195 and miR-497 retained tumours that then lost these features as they progressed and metastasized to the ovary. The ovarian tumours, as primary tumours, would have retained their p53 and microRNA profiles). Of note, dysregulation of miR-16, miR-195 and miR-497 have also been found in a range of other cancers, including chronic lymphocytic leukaemia, prostate, pancreas, cervical carcinomas together with sarcomas and benign uterine leiomyomata (Table 4).12, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63

Table 4 List of cancers associated with dysregulation of miR-16, miR-195 and miR-497
Full size table

Little is known about the mechanisms of microRNA deregulation in neoplastic tissue. MicroRNA genes are commonly located at minimal regions of amplification, loss of heterozygosity and breakpoint regions, suggesting that abnormal microRNA profiles can be caused by somatic gene mutation.19 Of note, previous studies have demonstrated high rates of loss of heterozygosity in both primary peritoneal carcinoma and ovarian serous carcinoma.4, 5, 39 A microRNA DNA copy number study of human ovarian cancer in combination with breast cancer and melanoma showed that a high proportion of genomic loci containing microRNA genes exhibit DNA copy number alterations.19 Furthermore, our group has demonstrated dysregulation of Dicer and Drosha proteins (both of which coordinate microRNA processing) in ovarian serous carcinoma.64

Recent research suggests that DNA methylation may also be involved in the regulation of microRNA expression in ovarian cancer, and the methylation can be reversed by DNA methyltransferase inhibitors.9, 65 Indeed, microRNA control and deregulation may be more complex as widespread microRNA repression by transcription factors such as c-myc (that targets miR-195/miR-497) have also been found to contribute to tumourigenesis.66

Our data demonstrate decreased p53 expression in primary peritoneal carcinoma relative to ovarian serous carcinoma. There was lower p53 overexpression in our cohort of primary peritoneal carcinomas vs ovarian serous carcinomas (63 vs 84%) with a lower mean p53 immunoreactivity score (4.6 vs 6.0; P=0.04). This difference may be due to distinct allelic loss patterns in both tumour types which has been demonstrated in previous studies.4, 5 The tumour-suppressor gene p53, which maps to chromosome 17p13.1, has a key role in cell-cycle regulation and apoptosis. Over 60% of ovarian carcinomas express mutations in the p53 gene.67, 68, 69 Previous studies70, 71, 72, 73 have reported both higher and lower p53 overexpression in ovarian serous carcinoma relative to primary peritoneal carcinomas, however these results were not statistically significant. True differences in p53 overexpression may however have been masked in these studies by both sample size and scoring methodology. The bcl-2 gene, an experimentally validated target of miR-16 in B-cell lymphomas,25 results in dysregulation of apoptosis. Expression of bcl-2 has been observed in 57% of ovarian cancer cases, and increased bcl-2 expression has been associated with improved patient survival.67, 69

Our study demonstrates both downregulation of miR-16 in primary peritoneal carcinoma relative to ovarian serous carcinoma and protein overexpression of bcl-2 in primary peritoneal carcinoma relative to ovarian serous carcinoma. In contrast to our study, previous groups have noted similarities in quantitative bcl-2 expression70 between primary peritoneal and ovarian serous carcinoma. However differences in qualitative expression were not examined specifically.

Potential overrepresented pathways targeted by mir-195 and miR-497 include the cadherin, WNT and T-cell activation pathways. Dysregulation of putative target genes within these lists have been described in ovarian cancer.74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88 E-cadherin and α-catenin, components of the cadherin pathway, have previously been found to be dysregulated in metatstatic ovarian carcinoma and serous effusions.79, 80, 81 Polymorphisms in cyclin D2 and protein kinase C isoform expression have been linked with prognosis in ovarian carcinoma patients.87, 88 AKT3, a component of the T-cell activation pathway, has been found to be highly expressed in a subset of ovarian serous carcinomas, and has been implicated as a key mediator of ovarian oncogenesis by the PI3K/AKT pathway.78 PIK3R1, part of the same family of lipid kinases has also been implicated as an oncogene in ovarian tumourigenesis.86 These putative targets indicate that miR-195 and miR-497 may have multiple biological roles in primary peritoneal carcinoma tumourigenesis.

In conclusion, we provide definitive evidence that there is downregulation of a select group of microRNAs in tumours meeting Gynaecological Oncology Group criteria for primary peritoneal carcinoma relative to ovarian serous carcinoma. Specifically, we show decreased p53 expression and downregulation of miR-195 and miR-497 from the microRNA cluster site at chromosome 17p13.1 in primary peritoneal carcinoma relative to ovarian serous carcinoma. Further investigation is needed to examine the functional role of miR-195 and miR-497 as potential tumour-suppressor genes in primary peritoneal tumour development.