Tumor microRNA expression patterns associated with resistance to platinum based chemotherapy and survival in ovarian cancer patients

Gynecologic Oncology 114 (2009) 253–259

Contents lists available at ScienceDirect

Gynecologic Oncology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / y g y n o

Tumor microRNA expression patterns associated with resistance to platinum based chemotherapy and survival in ovarian cancer patients Ram Eitan a,b,⁎,1, Michal Kushnir c,1, Gila Lithwick-Yanai c,1, Miriam Ben David c, Moshe Hoshen c, Marek Glezerman a,b, Moshe Hod a,b, Gad Sabah a,b, Shai Rosenwald c, Hanoch Levavi a,b a b c

The Helen Schneider Hospital for Women, Rabin Medical Center, Petah Tikva, Israel Sackler School of Medicine, Tel Aviv University, Israel Rosetta Genomics, Rehovot, Israel

a r t i c l e

i n f o

Article history: Received 26 January 2009 Available online 14 May 2009 Keywords: MicroRNA Ovarian cancer Platinum resistance

a b s t r a c t Background. Ovarian cancer, the leading cause of gynecologic cancer deaths, is usually diagnosed in advanced stages. Prognosis relates to stage at diagnosis and sensitivity to platinum based chemotherapy. We aimed to assess the expression of microRNAs in ovarian tumors and identify microRNA expression patterns that are associated with outcome, response to chemotherapy and survival. Methods. Patients, who were surgically treated for ovarian cancer between January 2000 and December 2004 were identified. Patient charts were reviewed for clinicopathologic information, follow-up and survival. Total RNA was extracted from tumor samples and microRNA expression levels were measured by microarrays. Expression levels were compared between groups of samples and statistically analyzed. Results. Fifty-seven patients were identified to fit study criteria. Of them, 19 patients had stage I disease at diagnosis, and 38 patients, stage III. All patients received platinum based chemotherapy as first line treatment. 18 microRNAs were differentially expressed (p b 0.05) between stage I and stage III disease. Seven microRNAs were found to be significantly differentially expressed in tumors from platinum-sensitive vs. platinum-resistant patients (p b 0.05). Five microRNAs were associated with significant differences (p b 0.05) in survival or recurrence-free survival. High expression of hsa-mir-27a identified a sub-group of patients with very poor prognosis. Conclusions. We have found an array of tumor specific markers that are associated with response to platinum based first line chemotherapy. Expression of some of these miRNAs also correlated closely with prognosis. This approach can potentially be used to tailor chemotherapy and further management to specific patient needs. © 2009 Elsevier Inc. All rights reserved.

Introduction Epithelial ovarian cancer (EOC) is the fifth leading cause of cancerrelated deaths in women in the United States and the leading cause of gynecologic cancer-related deaths [1]. Annually, there are more than 22,000 new cases of ovarian cancer in the United States and over 16,000 deaths. Despite efforts to develop an effective ovarian cancer screening method, most patients still present with advanced (stages III–IV) disease [2]. Survival of patients diagnosed with ovarian cancer is known to closely correlate with stage at diagnosis. In the setting of primary advanced disease, the two most important prognostic factors for patients with advanced ovarian carcinoma are the amount of residual disease left after surgery and the response to platinum based chemotherapy [2,3]. A complete clinical response can

⁎ Corresponding author. The Helen Schneider Hospital for Women, Rabin Medical Center, Petah Tikva, Israel. E-mail addresses: [email protected], [email protected] (R. Eitan). 1 RE, MK and GL-Y contributed equally to this work. 0090-8258/$ – see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.ygyno.2009.04.024

be achieved in approximately 80%–90% of patients with early-stage disease and in 50% of patients with advanced-stage disease. Despite achieving clinical remission after completion of initial treatment, most patients with advanced epithelial ovarian cancer will ultimately develop recurrent disease [4]. Patients who have a prolonged disease-free-interval after first line platinum based chemotherapy, can be re-treated with platinum and are more likely to respond well to second line therapy. This group of patients has an improved prognosis with a prolonged disease-free interval and longer overall survival. Patients who have progressive disease during platinum treatment or who suffer first recurrent disease within a short period of time are termed platinum-resistant. These patients are given alternative chemotherapy regimens which offer relatively small total response rates reaching 20–30% at most and usually have a poorer prognosis. It is clinically important to identify biomarkers that may assist to detect and predict which patients with ovarian cancer will respond to platinum based chemotherapy and which patients will remain refractory to this standard treatment. Specific data may assist in tailoring

254

R. Eitan et al. / Gynecologic Oncology 114 (2009) 253–259

treatment to each patient's specific clinical situation during the initial management of their disease and also offer the opportunity for better counseling regarding prognosis. Comparison of the patterns of gene expressions in ovarian cancer and normal ovarian tissue using cDNA microarrays revealed several genes that are differentially expressed in ovarian cancer [5–7]. Others have identified patterns of gene expression that predict response to chemotherapeutic agents, and prognosis [8–10]. MicroRNAs (miRNAs), a group of 22-oligonucleotide RNA molecules are non-coding forms of RNAs involved in post-transcriptional gene regulation. As modulators of protein expression they may also operate as oncogenes and tumor suppressors [11]. These small RNAs act by binding to a complementary site in the 3′UTR of specific mRNA molecules, leading to either suppression of translation or cleavage of the mRNA. MiRNAs have been shown to be conserved through evolution, and are believed to represent 1–5% of the total genes in some species [12]. Studies have demonstrated an important role for miRNAs in various developmental, differentiation and maturation processes [13,14], and more recently, miRNAs has been suggested to be involved in pathological conditions, amongst them, cancer [15– 18]. Understanding the regulatory role of miRNA may lead to better understanding of the molecular events involved in different biological processes, and can lead to the development of diagnostic tools and a novel class of drug targets for therapeutic interventions. Using high-resolution array-based comparative genomic hybridization (CGH) researchers have demonstrated significant alterations in the copy number of genomic loci that contain miRNA genes in several cancers, including ovarian cancer [19].

In this study, we have assessed the expression pattern of miRNA in ovarian tumors. We determine that certain microRNA expression patterns predict outcome, response to chemotherapy and survival. Materials and methods Patients and samples Patients, who were surgically treated for ovarian cancer at the Rabin Medical Center between January, 2000 and December, 2004 were identified. All pathology slides were re-evaluated by an expert pathologist. Tumor histology was established and the diagnosis of EOC was confirmed. Only serous papillary and endometrioid histology were included in the study. Patients found to have a synchronous endometrial malignancy were excluded. For each patient, a formalinfixed, paraffin-embedded (FFPE) tumor sample was obtained and tumor cell content was evaluated by a pathologist. Only tumor samples with a minimum of 50% tumor tissue content were included. Patient charts were reviewed for clinicopathologic information — demographics, surgical procedure and findings, pathology, chemotherapy regimens and response, follow-up and survival. Optimal surgical cytoreduction was defined during the study period as the largest residual tumor diameter of 1 cm. Patients with progressive disease during first line platinum based chemotherapy or those who suffered recurrent disease within 6 months of completing first line therapy were termed platinum-resistant. Patients with no recurrence or with recurrences beyond 6 months were termed platinumsensitive. Survival time was calculated as the time from the end of

Fig. 1. (a) MicroRNAs differentially expressed in stage 3 ovarian cancers that are resistant (n = 12) or sensitive (n = 25) to platinum based chemotherapy. Seven microRNAs had pvalue (Mann–Whitney) smaller than 0.05, but did not pass a false discovery rate (FDR) of 0.1. Three of these microRNAs (p-value ≤ 0.011) passed at FDR = 0.3. Fold-change is the ratio of the median signals in the two groups. (b and c) Differential expression of microRNAs in stage III ovarian cancers that are resistant or sensitive to platinum based treatment. Expression scale (y-axis) shows the logarithm (base 2) of the normalized fluorescence signal by microarray (Methods). Boxplots show the median (horizontal line), 25 to 75 percentile (box) and extent of data (“whiskers”) for resistant (n = 12) and sensitive (n = 25) patients, for hsa-miR-27a (b, p-value = 0.0019, median expression 1.7-fold higher in resistant) and hsa-miR-378 (c, p-value = 0.0055, median expression 1.8-fold higher in sensitive).

R. Eitan et al. / Gynecologic Oncology 114 (2009) 253–259

treatment to the last follow-up date or death. Recurrence time was calculated as the time from the end of treatment to the time of detected recurrence/progression. The study was approved by the institutional review board of the Rabin Medical Center.

255

expression analysis (Fig. 1a) and in the progression-free survival analysis (Table 3), only genes that were targeted by at least one of these two microRNAs were listed. Results

RNA extraction Total RNA was extracted as described before [20]. Briefly, FFPE tissues were deparaffinized with xylene, washed in ethanol, and digested with proteinase K. The RNA was extracted with acid phenol:chloroform followed by ethanol precipitation and DNAse digestion. MicroRNA microarray Custom microRNA microarrays were prepared as described previously [20]. Briefly, 900 DNA oligonucleotide probes representing microRNAs were spotted in triplicate on coated microarray slides (Nexterion® Slide E, Schott, Mainz, Germany). 3–5 μg of total RNA were labeled by ligation of an RNA-linker, p-rCrU-Cy/dye (Dharmacon, Lafayette, CO; Cy3 or Cy5) to the 3′ end. Slides were incubated with the labeled RNA for 12–16 h at 42 °C and then washed twice. Arrays were scanned at a resolution of 10 μm, and images were analyzed using SpotReader software (Niles Scientific, Portola Valley, CA). Microarray spots were combined and signals normalized as described previously [20]. Previous studies have confirmed the validity of this measurement platform through comparison to a qRT–PCR platform [20].

Fifty-seven patients were identified to fit study criteria. Nineteen patients had stage I disease at diagnosis, 38 patients stage III at diagnosis. Due to small numbers, stage II and stage IV patients were excluded from the study. Table 1 lists clinical parameters for the study cohort, divided into stage I patients, and stage III patients who were either platinum-resistant or platinum-sensitive (see Methods and below). One patient was censored after 161 days (due to death of other causes) and is not included in Table 1. Median age of the study cohort was 58 years. Of the stage III patients, 18 had optimal surgical cytoreduction and 15 were left with sub-optimal residual disease at the end of surgery. Thirty five patients were diagnosed with serous adenocarcinoma and 22 with endometrioid histology. Most of the 19 patients with stage I disease were staged according to FIGO guidelines (Table 1). All of them had bilateral salpingooophorectomy (BSO), cytology washings and omentectomy per-

Table 1 Characteristics of stage I and III patients and tumors in the cohort used. Stage I

Stage III sensitive

Stage III resistant

19

25

12

Median Range

55 40–81

59 42–88

66 41–76

Stage 3a/b Stage 3c

NA NA

5 20

0 12

Grade 0–2 Grade 3

13 6

8 14

4 7

n Age

Data analysis

0.8699

Substage

Expression levels between groups of samples were compared using the Mann–Whitney non-parametric test. Only microRNAs which had a median signal higher than signal background levels (normalized fluorescence signal of ∼ 300) in at least one of the two groups were tested. Corrections for multiple comparisons were performed using the Benjamini–Hochberg “false discovery rate” (FDR) method [21]. Survival time course was studied using the Kaplan–Meier method, and groups were compared using the logrank test. Stability of microRNAs in survival analysis was assessed by repeated (100 times) random resampling (bootstrap) from the original dataset (maintaining group sizes). Multivariate analysis of microRNA expression (for hsa-miR-27a), grade, age, optimal cytoreduction, and histological type was performed using Cox regression. The values of these features were combined in order to predict progression times in stage III patients. Histological type was encoded such that endometrioid carcinoma samples were given a value of one, while serous carcinomas were given a value of zero. Similarly, a value of one or zero was assigned for samples with or without optimal cytoreduction, respectively.

0.19

Grade

1

Treatment

0.18 Cyclophosphamide with cisplatin Platinum as a single agent Paclitaxel with carboplatin

1

1

0

14

2

4

4

22

8

Serous Endometrioid

10 9

13 12

12 0

Optimal Suboptimal

18 0

16 6

2 8

Yes No

19 0

24 0

11 1

Yes No

10 9

16 9

5 7

Yes No

19 0

23 1

11 1

Yes No

12 7

10 14

1 11

Yes No

14 5

22 2

6 6

Histology

0.0033

Cytoreduction

Omentectomy

0.0084 0.33

Appendectomy

0.29

BSO

MicroRNA target prediction Targets were selected (Supplementary Table S1) from the intersection of the target prediction results by Targetscan [22] and Miranda [23]. Only targets with a Targetscan score lower than 0, and a Miranda score ≥ 150 were used. In order to retrieve only the most relevant targets, we listed only genes targeted by at least three microRNAs that we found to be associated with poor prognosis. This list included the microRNAs that were over-expressed in platinum-resistant stage III patients compared to platinum-sensitive stage III patients (Fig. 1a, including hsa-miR-27a, hsa-miR-23a, hsa-miR-30c, hsa-let-7g, and hsa-miR-199a-3p) and the microRNAs that were associated with significantly poorer recurrence-free survival (Table 3, including hsamiR-27a, hsa-miR-23a, and hsa-miR-21). In addition, since hsa-miR27a and hsa-miR-23a were significant in both the differential

p-value

1

LNS

0.0595

TAH

0.0091

For treatment type, Fisher's exact test was used to compare the treatment of “paclitaxel with carboplatin” to the other treatment types grouped together. For other parameters, p-values were calculated comparing the stage III platinum-sensitive patients to the stage III platinum-resistant patients. p-value for age was calculated using a two-sample unpaired t-test. For the other parameters, two-sided Fisher's exact test was used. Numbers for several characteristics that do not add up to the total number of patients in the relevant category indicate missing clinical data. One stage III patient was censored (died of other causes after 161 days) and was not included in this table. BSO indicates bilateral salpingo-oophorectomy; LNS indicates lymph-node sampling; TAH indicates total abdominal hysterectomy.

256

R. Eitan et al. / Gynecologic Oncology 114 (2009) 253–259

formed; 12 (63%) had lymph-node sampling (LNS); 10 (52%) appendectomy; and 14 (73%) total abdominal hysterectomy (TAH). All patients received platinum based chemotherapy as first line treatment. 21 patients received platinum as a single agent, 34 received paclitaxel with carboplatin, and 2 patients received cyclophosphamide with cisplatin. MicroRNA expression patterns correlate with stage of disease Time to progression and survival were clearly linked to stage in our study cohort of patients (Supplementary Fig. S1). We compared microRNA expression between stage I (n = 19) and stage III (n = 38) cases. 18 microRNAs (Table 2) were differentially expressed with p b 0.05 (Mann–Whitney test), including for example hsa-miR449b (Supplementary Fig. S2a, p = 0.048), but only hsa-miR-200a (Supplementary Fig. S2b, p = 0.00047) was significant when allowing a false discovery rate (FDR) of 10% [21]. Both of these microRNAs were more highly expressed in stage I ovarian cancers compared to stage III cases (Supplementary Fig. S2 and Table 2). MicroRNA expression patterns in patients with advanced ovarian carcinoma We studied the relation between microRNA expression and disease progression. Since patient prognosis (Supplementary Fig. S1) and disease characteristics vary for different stages of the disease (Table 1), we focused on the larger, higher risk group of patients in stage III. 25 patients achieved a complete response with no recurrence within 6 months of the end of treatment, and were termed platinumsensitive. Twelve patients had rapid progression of the disease (partial response or recurrence within 6 months of the end of treatment) and were termed platinum-resistant. The patient censored before 6 months was not included in this analysis. We initially examined tumor microRNA expression patterns in these two groups. Seven microRNAs (Fig. 1a) were found to be significantly differentially expressed in tumors from platinumsensitive vs. platinum-resistant patients (p b 0.05, 3 passed FDR = 0.3), including hsa-miR-27a (Fig. 1b, p = 0.0019), and hsa-

Table 2 MicroRNAs differentially expressed in stage I (n = 19) compared to stage III (n = 38) ovarian cancers. MicroRNAs over-expressed in stage III ovarian cancers MiR name

p-value

Hsa-miR-423-3p Hsa-miR-130a Hsa-miR-146b-5p Hsa-miR-193a-3p Hsa-miR-193a-5p Hsa-miR-491-5p Hsa-miR-23b Hsa-miR-125a-3p Hsa-miR-125a-5p Hsa-miR-451

0.0024 0.0033 0.0037 0.0056 0.013 0.028 0.028 0.030 0.034 0.035

Fold-change 1.33 1.86 2.27 1.42 1.60 1.40 1.10 1.27 1.25 1.93

MicroRNAs over-expressed in stage I ovarian cancers MiR name

p-value

Fold-change

Hsa-miR-200a Hsa-miR-200b Hsa-miR-34a Hsa-miR-513a-5p Hsa-miR-509-3p Hsa-miR-509-3-5p Hsa-miR-574-5p Hsa-miR-449b

0.00047 0.0043 0.0066 0.0068 0.0074 0.017 0.045 0.048

2.10 1.63 1.69 5.32 10.3 4.01 1.24 4.61

18 microRNAs had p-value (Mann–Whitney) smaller than 0.05. Ten microRNAs (with p-value ≤ 0.013) passed at a false discovery rate (FDR) of 0.2, only hsa-miR-200a passed at FDR = 0.1. Fold-change is the ratio of the median signals in the two groups.

Table 3 MicroRNAs which are associated with significant changes in survival or recurrence-free survival of stage III ovarian cancers. MiR name

p-value, time to progression

p-value, survival

Higher expression associated with

Hsa-miR-23a Hsa-miR-27a Hsa-miR-449b Hsa-miR-21 Hsa-miR-24-2⁎

0.0049 0.0176 0.1 0.0493 0.225

0.0025 0.0215 0.0379 0.222 0.0493

Poorer prognosis Poorer prognosis Better prognosis Poorer prognosis Poorer prognosis

p-values were calculated by the logrank test, comparing the survival or the time to progression of the two groups with high (n = 13) or low (n = 13) expression level of each microRNA (Fig. 2).

miR-378 (Fig. 1c, p = 0.0055) and hsa-miR-23a (p = 0.011). The differential expression of these three microRNAs between sensitive and resistant tumors was also observed in the subset of stage III patients treated by the combined paclitaxel with carboplatin treatment (n = 30), and in the subset of stage III serous tumors (n = 25). With the exception of hsa-let-7g, which was over-expressed in serous papillary tumors (p = 0.013), no difference was found between tumors of serous and endometrioid histologies with similar response to chemotherapy in the expression of relevant microRNAs (microRNAs that are listed in Table 2, Fig. 1 and Table 3). We next compared the prognosis of patients in groups stratified according the expression levels of individual microRNAs. For each microRNA, we divided the samples into tertiles according to high, intermediate or low expression level of the microRNA. We plotted the survival and the recurrence-free survival time curves for the three groups, and compared the survival and time to progression between the two groups with high or low microRNA expression levels. Five microRNAs were associated with significant differences (logrank p b 0.05) in survival or recurrence-free survival (Table 3). Hsa-miR23a and hsa-miR-27a were associated with significant differences in both survival and recurrence-free survival (Fig. 2). The association of these microRNAs with survival and chemoresistance was stable, survival differences maintaining significance (p b 0.05) in at least one third of random resampling (bootstrap) runs (Methods). Hsa-mir-27a expression was found to identify a sub-group of these patients with very poor prognosis (Fig. 3) that had progressive disease during first line chemotherapy and extremely short progression-free survival. In order to assess the relative contribution of various parameters on progression times in stage III, the Cox proportional hazards model was used. The parameters used were hsa-miR-27a expression (b = 0.99, p = 0.02), grade (b = − 0.15, p = 0.76), age (b = 0.03, p = 0.13), optimal cytoreduction status (b = − 1.1, p = 0.05) and histological type (b = 0.37, p = 0.51). The results thus indicated that grade, age and histological type do not contribute to progression times within stage III beyond the effect of hsa-miR-27a expression and optimal debulking status. In order to further examine the connection between hsa-miR-27a expression and optimal cytoreduction status, the association of this microRNA with disease progression was analyzed separately for stage III patients with or without optimal cytoreduction. For patients with optimal cytoreduction, hsa-miR-27a was not a good predictor of progression times (logrank p-value of 0.62, comparing the upper and lower tertiles). However, interestingly, for patients without optimal debulking, hsa-miR-27a was a significant predictor of progression times (logrank p-value of 0.046, comparing the upper and lower tertiles). Discussion It is well known that stage at diagnosis is one of the most important prognostic factors in ovarian cancer. Only 25% of patients are diagnosed with stage I disease confined to the ovary but these patients have excellent survival, reaching 85% at 5 years [4]. Some

R. Eitan et al. / Gynecologic Oncology 114 (2009) 253–259

257

Fig. 2. Kaplan–Meier curves showing recurrence-free survival (a, c) and disease-specific survival (b, d) for groups of patients with stage III disease, stratified by expression levels of hsa-miR-23a (a, b) and hsa-miR-27a (c, d). The sample set was divided into three groups with high expression levels (dashed-dotted line, n = 13), intermediate expression levels (dashed line, n = 12), and low expression levels (solid line, n = 13) of hsa-miR-23a (a, p-value = 0.0049, b, p-value = 0.0025) or hsa-miR-27a (c, p-value = 0.018, d, p-value 0.022). p-values are calculated by logrank test comparing the low and high expression groups (Table 3). Censoring events are marked by gray vertical lines.

investigators believe that stage I ovarian carcinoma is a completely separate disease from cases diagnosed with advanced disease [24], although this has not yet met with general acceptance. We have found that microRNA expression differs in tumors taken from patients with stage I ovarian carcinoma in comparison to their advanced (stage III) counterparts. Most of our patients were thoroughly staged. Identifying appropriately staged early ovarian cancer is always an issue, and the fact that not all patients were staged according to strict guidelines is a possible weakness of all retrospective data collection including ours. Strengthening the data is the stage I survival curve (Supplementary Figure S1) which is in accordance with expected stage I patient survival. Several microRNAs were significantly differentially expressed between the stage I and stage III ovarian cancers (Table 2). Of particular interest are hsa-miR-200a, hsa-miR-34a, and hsa-miR-449b, which were down-regulated in the advanced (stage III) tumors. We studied the relation of microRNA expression to the prognosis of ovarian cancer patients. To avoid confounding effects of stage, we performed this analysis in the group of 38 stage III ovarian cancer patients, and performed two types of analyses, which identified several microRNAs. Hsa-miR-378 was found to have significantly higher expression levels in the groups of patients that were sensitive as compared to resistant to treatment by platinum based chemotherapy (Fig. 1). Expression of hsa-miR-449b divided the patients into groups with significantly different disease-specific survival times. Patients with higher expression of hsa-miR-449b were found to have an improved overall survival (Table 3). Hsa-miR-23a and hsa-miR-27a were found to be significantly associated with outcome by both methods of analysis. High levels of these microRNAs were associated in both cases with a poorer prognosis.

Hsa-miR-449b bears a high similarity in sequence to the hsamiR-34 family. In particular, residues 2-8 (5′ end) of hsa-miR-449b are identical to those of hsa-miR-34a. Residues 2–8 of microRNAs, also referred to as the “seed” sequence, are the most strongly conserved sequences in microRNAs and microRNA families and are considered the most important for determination of mRNA targets of microRNAs [22]. Thus, similarity in the seed sequence may suggest

Fig. 3. Kaplan–Meier curves showing recurrence-free survival for groups of patients of stage III disease, stratified by expression levels of hsa-miR-27a. Exceptionally high expression level of hsa-miR-27a identifies a sub-group of patients (n = 5) with very poor prognosis. The samples with the highest expression level of hsa-miR-27a (normalized fluorescence signal N9500, see Methods) had a very low time to progression and were resistant to platinum based treatment, and most (4 out of 5) had an incomplete response. Samples with lower expression of hsa-miR-27a had a median survival time of 20.6 months.

258

R. Eitan et al. / Gynecologic Oncology 114 (2009) 253–259

similar activity in a cancer cell. In this study, we found hsa-miR-34a to be down-regulated in advanced (stage III) tumors. Hsa-miR-34a is known to be a downstream target that is positive regulated by p53 [25,26], and its down-regulation in advanced cancers may thus be related to p53 abnormalities. Hsa-miR-449b was similarly downregulated in advanced tumors while its high expression was associated with a better response to platinum based chemotherapy among stage III cases. The identity of seed sequence suggests that hsa-miR449b may have similar functions or gene targets to hsa-miR-34a, which is involved in p53-mediated apoptosis and can reduce cell proliferation [25,26]. It has been postulated that decreased expression of miR-34a may contribute to tumorigenesis by attenuating p53dependent apoptosis. If this holds true for the case of ovarian cancer and for hsa-miR-449b is yet to be determined. Yang and coworkers recently studied expression of microRNA in ovarian cancers of different stages [27]. Using a set of ovarian tumors of mixed histologies, they found high expression of hsa-miR-200a associated with higher stage ovarian cancers. Nam and coworkers recently described a correlation between tumor expression of microRNAs and cumulative survival in ovarian carcinoma tumor samples [28]. In their data, high expression of hsa-miR-200a was associated with tumors from patients with poorer survival. In contrast to their findings, in our data set, significantly higher expression of hsamiR-200a was found in early-stage disease, correlating with improved survival. In the study by Nam and coworkers, no data is provided regarding stage at diagnosis and its correlation to micrRNA expression. The study by Yang and coworkers included significant numbers of mucinous and clear cell cystadenocarcinomas, histologies not represented in our study. This might explain the discrepancy between the studies. Yang and coworkers found high hsa-miR199a-3p (previously named hsa-miR-199a⁎) expression associated with higher stage tumors. In our dataset, we found this microRNA more highly expressed in platinum-resistant tumors. In the dataset of Nam and coworkers, that included 20 cases of various stages, hsa-miR-27a microRNA was marginally associated with poor prognosis but did not reach significance. In our set of 38 cases of stage III ovarian cancers, hsa-miR-27a was significantly associated with poor survival and poor response to platinum based chemotherapy. This effect was maintained also in multivariate analysis that takes into account additional clinical factors including grade, age, histological type and optimal debulking. As seen in Fig. 3, exceptionally high expression of hsa-mir-27a was associated with extremely poor survival. Current methods of bioinformatic prediction of microRNA targets provide large numbers of potential targets, many of which are probably false positive results [29]. To narrow down the list of potential genes of interest (Methods) we used the intersection of two target prediction programs: Targetscan [22] and Miranda [23]. Targetscan bases predictions on the match of the microRNA seed with the mRNA 3′UTR and some additional sequence-based parameters [22], while Miranda bases predictions mainly on the alignment of the whole mature miR to the target in addition to the thermodynamic stability [23]. We looked for genes that are targeted by at least three of the “poor prognosis” microRNAs (microRNAs in Fig. 1a and Table 3) and are also targeted by hsa-miR-23a and hsa-miR-27a (see Methods). These genes (Supplementary Table S1) included EIF4EBP2, which is a target of hsa-miR-21, hsa-miR-23a, and less significantly of hsa-let-7g. In addition, the gene EIF4E3, is a predicted target of hsa-miR-23a and, less significantly, of hsa-miR-27a. These genes, together with EIF4G2 (a predicted target of hsa-let-7g), take part in the eIF4F complex [30] which was found to be associated with good prognosis in ovarian cancer [31]. Thus, high levels of hsa-miR-21, hsa-miR-23a, hsa-miR27a and hsa-let-7g may contribute to poor prognosis or chemotherapy resistance through their effect on this complex. A recent study showed that expression of miR-27a was up-regulated in multidrug resistant (MDR) ovarian cancer cell line compared to its parental (A2780) cell line [32]. Also found was that treatment with miR-27a antagomirs

decreased the expression of P-glycoprotein and MDR1 mRNA, and enhanced sensitivity due to the intracellular accumulation of cytotoxic drugs, suggesting an alternative mechanism for the effect of hsa-miR27a on chemoresistance. Many of our ovarian cancer patients who respond completely to first line chemotherapy and are with no evidence of disease at the end of treatment are unfortunately diagnosed with recurrent disease during follow-up. There are currently few prognostic markers used to predict response, survival and overall prognosis. In this study, we have found an array of microRNA markers that are associated with response to platinum based first line chemotherapy. This approach can potentially be used to tailor chemotherapy to specific patient needs, to help in the selection of the most suitable treatment for those at high risk for recurrence and to better counsel patients on prognosis and the strategies planned to better their outcome. These microRNAs present potential candidates for the development of future therapeutic agents. Conflict of interest statement Ram Eitan — no conflict of interest. Michal Kushnir — full time employee of Rosetta genomics receiving salary and holding company equity. Gila Lithwick-Yanai— full time employee of Rosetta genomics receiving salary and holding company equity. Miriam Ben David— full time employee of Rosetta genomics receiving salary and holding company equity. Moshe Hoshen— full time employee of Rosetta genomics receiving salary and holding company equity. Marek Glezerman— no conflict of interest. Moshe Hod — chairman of Rosetta Genomics medical board. Gad Sabah— no conflict of interest. Shai Rosenwald— full time employee of Rosetta genomics receiving salary and holding company equity. Hanoch Levavi— no conflict of interest.

Acknowledgment We thank Dr. Tamara Drozd from the Rabin Medical Center for her assistance. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.ygyno.2009.04.024. References [1] Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. Cancer statistics, 2007. CA Cancer J Clin 2007;57:43–66. [2] Winter III WE, Maxwell GL, Tian C, et al. Prognostic factors for stage III epithelial ovarian cancer: a Gynecologic Oncology Group Study. J Clin Oncol 2007;25: 3621–7. [3] Bristow RE, Tomacruz RS, Armstrong DK, Timble EL, Montz FJ. Survival effect of maximal cytoreductive surgery for advanced ovarian carcinoma during the platinum era: a meta-analysis. J Clin Oncol 2002;20:1248–59. [4] Ozols RF, Rubin SC, Thomas GM, Robboy SJ. Epithelial ovarian cancer. In: Hoskins WJ, Perez CA, Young RC, Barakat R, Markman M, Randall M, editors. Principles and practice of gynecologic oncology. 4th ed. Philadelphia, PA: Lippincott-Raven Publishers; 2005. p. 895–987. [5] Collins Y, Tan DF, Pejovic T, et al. Identification of differentially expressed genes in clinically distinct groups of serous ovarian carcinomas using cDNA microarray. Int J Mol Med 2004;14:43–53. [6] Zhang X, Feng J, Cheng Y, et al. Characterization of differentially expressed genes in ovarian cancer by cDNA microarrays. Int J Gynecol Cancer 2005;15:50–7. [7] Lee BC, Cha K, Avraham S, Avraham HK. Microarray analysis of differentially expressed genes associated with human ovarian cancer. Int J Oncol 2004;24: 847–51. [8] Sakamoto M, Kondo A, Kawasaki K, et al. Analysis of gene expression profiles associated with cisplatin resistance in human ovarian cancer cell lines and tissues using cDNA microarray. Hum Cell 2001;14:305–15. [9] Selvanayagam ZE, Cheung TH, Wei N, et al. Prediction of chemotherapeutic response in ovarian cancer with DNA microarray expression profiling. Cancer Genet Cytogenet 2004;154:63–6. [10] Goto T, Takano M, Sakamoto M, et al. Gene expression profiles with cDNA microarray reveal RhoGDI as a predictive marker for paclitaxel resistance in ovarian cancers. Oncol Rep 2006;15:1265–71. [11] Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 2004;116:281–97.

R. Eitan et al. / Gynecologic Oncology 114 (2009) 253–259 [12] Chen CZ. MicroRNAs as oncogenes and tumor suppressors. N Engl J Med 2005;353: 1768–71. [13] Lau NC, Lim LP, Weinstein EG, Bartel DP. An abundant class of tiny RNAs with probable regulatory roles in Caenorhabditis elegans. Science 2001;294:858–62. [14] Palatnik JF, Allen E, Wu X, et al. Control of leaf morphogenesis by microRNAs. Nature 2003;425:257–63. [15] Chen CZ, Li L, Lodish HF, Bartel DP. MicroRNAs modulate hematopoietic lineage differentiation. Science 2004;303:83–6. [16] Eder M, Scherr M. MicroRNA and lung cancer. N Engl J Med 2005;352:2446–8. [17] Yanaihara N, Caplen N, Bowman E, et al. Unique microRNA molecular profiles in lung cancer diagnosis and prognosis. Cancer Cell 2006;9:189–98. [18] Volinia S, Calin GA, Liu CG, et al. A microRNA expression signature of human solid tumors defines cancer gene targets. Proc Natl Acad Sci U S A 2006;103:2257–61. [19] Zhang L, Huang J, et al. MicroRNAs exhibit high frequency genomic alterations in human cancer. Proc Natl Acad Sci U S A 2006;103:9136–41. [20] Rosenfeld N, Aharonov R, Meiri E, et al. MicroRNAs accurately identify cancer tissue origin. Nat Biotechnol 2008;26(4):462–9. [21] Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Roy Statist Soc Ser B 1995;57:289–300. [22] Grimson A, Farh KK, Johnston WK, Garrett-Engele P, Lim LP, Bartel DP. MicroRNA targeting specificity in mammals: determinants beyond seed pairing. Mol Cell 2007;27:91–105. [23] John B, Enright AJ, Aravin A, Tuschl T, Sander C, Marks DS. Human microRNA targets. PLoS Biol 2005;3(7):e264.

259

[24] Kurman RJ, Visvanathan K, Roden R, et al. Early detection and treatment of ovarian cancer: shifting from early stage to minimal volume of disease based on a new model of carcinogenesis. Am J Obstet Gynecol 2008;198(4): 351–6. [25] He L, He X, et al. A microRNA component of the p53 tumour suppressor network. Nature 2007;447:1130–4. [26] Raver-Shapira N, Marciano E, Meiri E, et al. Transcriptional activation of miR-34a contributes to p53-mediated apoptosis. Mol Cell 2007;26:731–43. [27] Yang H, Kong W, He L, et al. MicroRNA expression profiling in human ovarian cancer: miR-214 induces cell survival and cisplatin resistance by targeting PTEN. Cancer Res 2008;68:425–33. [28] Nam EJ, Yoon H, Kim SW, et al. MicroRNA expression profiles in serous ovarian carcinoma. Clin Cancer Res 2008;14(9):2690–5. [29] Sethupathy P, Megraw M, Hatzigeorgiou AG. A guide through present computational approaches for the identification of mammalian microRNA targets. Nat Methods 2006;3(11):881–6. [30] Sonenberg N. eIF4E, the mRNA cap-binding protein: from basic discovery to translational research. Biochem Cell Biol 2008;86(2):178–83. [31] Armengol G, Rojo F, Castellví J, et al. 4E-binding protein 1: a key molecular “funnel factor” in human cancer with clinical implications. Cancer Res 2007;67(16): 7551–5. [32] Zhu H, Wu H, Liu X, Evans B, Medina DJ, Liu CG, et al. Role of microRNA miR-27a and miR-451 in the regulation of MDR1/P-glycoprotein expression in human cancer cells. Biochem Pharmacol 2008;76(5):582–8.