Expression and clinical role of long non-coding RNA

YGYNO-977016; No. of pages: 8; 4C: Gynecologic Oncology xxx (2018) xxx–xxx

Contents lists available at ScienceDirect

Gynecologic Oncology journal homepage: www.elsevier.com/locate/ygyno

Expression and clinical role of long non-coding RNA in high-grade serous carcinoma Natalie Filippov-Levy a, Hallel Cohen-Schussheim a, Claes G. Tropé b, Thea E. Hetland Falkenthal c, Yoav Smith d, Ben Davidson b,e,⁎, Reuven Reich a,⁎⁎,1 a

Institute of Drug Research, School of Pharmacy, Faculty of Medicine, The Hebrew University of Jerusalem, Jerusalem 91120, Israel University of Oslo, Faculty of Medicine, Institute of Clinical Medicine, N-0316 Oslo, Norway Department of Oncology, Oslo University Hospital, Norwegian Radium Hospital, N-0310 Oslo, Norway d Genomic Data Analysis Unit, Hadassah Medical School, The Hebrew University of Jerusalem, Jerusalem 91120, Israel e Department of Pathology, Oslo University Hospital, Norwegian Radium Hospital, N-0310 Oslo, Norway b c

H I G H L I G H T S • HGSC lncRNA profiles differ from that of normal ovarian tissue. • LncRNAs are differentially expressed at different anatomic sites in HGSC. • ESRG and Link-A in HGSC effusions are candidate prognosticators of better survival.

a r t i c l e

i n f o

Article history: Received 9 September 2017 Received in revised form 3 January 2018 Accepted 3 January 2018 Available online xxxx Keywords: High-grade serous carcinoma Effusion Long non-coding RNA Array Survival

a b s t r a c t Objective. To profile long non-coding RNA (lncRNA) expression at the various anatomic sites of high-grades serous carcinoma (HGSC) and in effusion-derived exosomes. Methods. LncRNA profiling was performed on 60 HGSC specimens, including 10 ovarian tumors, 10 solid metastases and 10 malignant effusions, as well as exosomes from 30 effusion supernatants. Anatomic site-related expression of ESRG, Link-A, GAS5, MEG3, GATS, PVT1 H19, Linc-RoR, HOTAIR and MALAT1 was validated by quantitative PCR and assessed for clinical relevance in a series of 77 HGSC effusions, 40 ovarian carcinomas, 21 solid metastases and 42 supernatant exosomes. Results. Significantly different (p b 0.05) expression of 241, 406 and 3634 lncRNAs was found in comparative analysis of the ovarian tumors to solid metastases, effusions and exosomes, respectively. Cut-off at two-fold change in lncRNA expression identified 54 lncRNAs present at the 3 anatomic sites and in exosomes. Validation analysis showed significantly different expression of 5 of 10 lncRNAs in the 4 specimen groups (ESRG, Link-A, MEG3, GATS and PVT1, all p b 0.001). Higher ESRG levels in HGSC effusions were associated with longer overall survival in the entire effusion cohort (p = 0.023) and in patients with pre-chemotherapy effusions tapped at diagnosis (p = 0.048). Higher Link-A levels were associated with better overall (p = 0.015) and progression-free (p = 0.023) survival for patients with post-chemotherapy effusions. Link-A was an independent prognostic marker in Cox multivariate analysis in the latter group (p = 0.045). Conclusions. We present the first evidence of differential LncRNA expression as function of anatomic site in HGSC. LncRNA levels in HGSC effusions are candidate prognostic markers. © 2018 Elsevier Inc. All rights reserved.

1. Introduction ⁎ Correspondence to: B. Davidson, Department of Pathology, Norwegian Radium Hospital, Oslo University Hospital, Montebello, N-0310 Oslo, Norway. ⁎⁎ Corresponding author. E-mail addresses: [email protected] (B. Davidson), [email protected] (R. Reich). 1 R.R. is affiliated with the David R. Bloom Center for Pharmacy and the Adolf and Klara Brettler Center for Research in Molecular Pharmacology and Therapeutics at The Hebrew University of Jerusalem, Israel.

The sequencing of the human genome revealed that protein-coding genes constitute only ~2% of the whole genome. The non-protein transcriptome is generally referred to as non-coding RNAs (ncRNAs). Some of these genes, e.g. housekeeping ncRNAs, i.e. rRNA, are long-known, while others, such as regulatory ncRNAs, were recently revealed. ncRNAs are grouped based on transcript size. Those smaller than

https://doi.org/10.1016/j.ygyno.2018.01.004 0090-8258/© 2018 Elsevier Inc. All rights reserved.

Please cite this article as: N. Filippov-Levy, et al., Expression and clinical role of long non-coding RNA in high-grade serous carcinoma, Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.01.004

2

N. Filippov-Levy et al. / Gynecologic Oncology xxx (2018) xxx–xxx

200 bp are termed small ncRNAs and include microRNAs (miRNAs), endogenous siRNAs (endo-siRNAs), and PIWI-associated small RNAs (piRNAs). ncRNAs N 200 bp are referred to as long non-coding RNAs (lncRNAs), and can be further classified by function, i.e. enhancerassociated RNAs (eRNAs), promoter-associated transcripts (PATs) and circular RNAs (circRNAs) [1]. Analysis of the cellular transcriptome discovered that 10,000–32,000 transcripts are lncRNAs, ~ 11,000 are pseudogenes, ~ 9000 are small RNAs and only ~ 21,000 are protein-coding mRNAs [2]. These findings led to the understanding of the importance of ncRNAs, generating intensive study of this field. Studies on lncRNAs have revealed potential roles in regulating cellular activity: 1. Targeting proteins to specific genomic loci - it has been suggested that the lncRNA recruits PcG proteins (responsible for gene silencing) to specific locations [3]; 2. Recruiting histone-modifying complexes in cis - the X-inactivation specific transcript (Xist) is transcribed from only one X chromosome and causes transcriptional silencing; 3. Organizing chromatin domains, loops, chromosomes and nuclear structures [4]; 4. Function as precursors for small RNAs - multiple small RNAs are obtained from one lncRNA, and each mature transcript can be localized to a different place and have its one unique function; 5. Association with proteins - lncRNAs can be a part of a protein complex, functioning as coactivator, and can modulate protein cell localization; 6. Association with RNA - lncRNAs can modify transcript processing of pre-mRNA splicing and can also bind miRNA and titrate their availability to bind mRNA, thus acting like a sponge. Furthermore, it has been shown that lncRNA expression is cell type-dependent, and changes significantly in disease conditions [3]. Exosomes are 30–100 nm endosomal-derived vesicles secreted from various, though not all cells types. Exosome cargo consists of mRNA, miRNA, lncRNA, transcription factors, proteins and lipids. They are involved in cell-cell communication and cellular housekeeping. There is a difference between exosomes secreted from normal cells from those of cancer cells, both in quantity and in content [5]. A previous study from our lab showed that there is a difference between exosome content at different anatomic sites of ovarian carcinoma [6]. In the present study, we analyzed the expression of lncRNAs in highgrades serous carcinoma (HGSC) tumor cells at different anatomic sites, as well as exosomes isolated from effusion supernatants. We additionally assessed the clinical relevance of lncRNAs in HGSC effusions. 2. Materials and methods 2.1. Patients and specimens Specimens were submitted for routine diagnostic purposes to the Department of Pathology at the Norwegian Radium Hospital during the period of 1998 to 2008. HGSC specimens and clinical data were obtained from the Department of Gynecologic Oncology, Norwegian Radium Hospital. As the fallopian tubes have not been adequately assessed in this cohort, tumors in the ovary are specified as such without reference to primary site. The diagnosis of HGSC was made based on the combination of morphology (obvious nuclear atypia and the presence of multiple mitoses) and the presence of aberrant (diffusely positive or entirely negative) p53 immunostaining. The discovery cohort consisted of 60 HGSC specimens, including 10 ovarian tumors, 10 solid metastases and 10 malignant effusions, as well as exosomes from 30 effusion supernatants. The validation cohort analyzed using quantitative real-time reversetranscription polymerase chain reaction (qRT-PCR) consisted of 77 HGSC effusions (64 peritoneal, 13 pleural) from 77 patients. Additionally, 40 solid ovarian carcinoma specimens, 21 solid metastases, the majority omental, and 42 effusion supernatants were analyzed. Tumor specimens from the different anatomic sites were not patient-matched. Clinicopathologic data for the discovery (array) and validation cohorts are presented in Tables 1-A, 1-B.

Table 1-A Clinicopathologic parameters of the discovery cohort. Parameter

Effusions (n = 10)

Ovarian tumors (n = 10)

Solid metastases (n = 10)

Age (mean)

51–76 years (65)

52–81 years (64)

45–83 years (67)

FIGO stage I II III IV NA

0 0 5 5 0

2 0 6 2 0

0 0 7 2 1

Residual diseasea ≤1 cm N1 cm NA

4 4 2

8 1 1

4 5 1

216–7244 (681)b

60–3741 (730)

178–2478 (404)c

8 1 0 0 1d

6 0 0 0 4e

CA 125 at diagnosis (range; median)

Chemoresponse after primary treatment CR 5 PR 4 SD 1 PD 0 NA 0

Abbreviations: NA = not available; CR = complete response; PR = partial response; SD = stable disease; PD = progressive disease. a Among 7 patients with effusions who received surgery as upfront treatment, 3 were debulked to ≤1 cm, 4 to N1 cm. One of the remaining patients received neoadjuvant chemotherapy. Data were unavailable for 2 patients, including one who only received chemotherapy. b Available for 8 patients. c Available for 9 patients. d FIGO stage I patient who did not receive chemotherapy. e Disease response after chemotherapy could not be evaluated because of absent data, normalized CA 125 after primary surgery or missing CA 125 information and no residual tumor.

Effusions were centrifuged immediately after tapping, and cell pellets were frozen at − 70 °C in equal amounts of RPMI 1640 medium (GIBCO-Invitrogen, Carlsbad, CA) containing 50% fetal calf serum (PAA Laboratories GmbH, Pasching, Austria) and 20% dimethylsulfoxide (Merck KGaA, Darmstadt, Germany). Supernatants were frozen at −70 °C without any treatment. Effusions were diagnosed by an experienced cytopathologist (BD) based on morphology and immunohistochemistry performed on cell blocks prepared using the Thrombin clot method. Frozen sections from all solid tumors were reviewed by an experienced gynecopathologist (BD), and only specimens with tumor cell population N50% and minimal or no necrosis were included in this study. Informed consent was obtained according to national and institutional guidelines. Study approval was given by the Regional Committee for Medical Research Ethics in Norway. Signing of informed consent was waived by the Committee for patients diagnosed before 2007 as the majority of patients are deceased. 2.2. RNA and exosome extraction Solid samples were thawed and homogenized using mixed 1 mm and 2 mm zirconium oxide beads in the Bullet Blender (Next Advance, Averill Park, NY). Effusions were thawed and centrifuged and cells were transferred to a new tube before homogenization. Total RNA was isolated using the Tri-Reagent protocol (Sigma-Aldrich, St. Louis, MO). RNA quantity and quality were measured by NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA). One microgram of total RNA was transformed to cDNA using qScript cDNA synthesis kit (Quanta Biosciences, Gaithersburg, MD) according to the manufacturer's protocol. Exosomes extraction from effusion supernatants was performed using the ExoQuick kit (SBI, Mountain View, CA) followed by RNA

Please cite this article as: N. Filippov-Levy, et al., Expression and clinical role of long non-coding RNA in high-grade serous carcinoma, Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.01.004

N. Filippov-Levy et al. / Gynecologic Oncology xxx (2018) xxx–xxx

2.5. qRT-PCR and PCR

Table 1-B Clinicopathologic parameters of the validation cohort. Parameter

Effusions (n = 77)

Ovarian tumors (n = 40)

Solid metastases (n = 21)

Age (mean)

38–81 years (61)

31–86 years (60)

47–86 years (67)

FIGO stage I II III IV NA

0 2 44 31 0

1 7 26 4 2

0 0 18 3 0

Residual diseasea ≤1 cm N1 cm NA

37 30 10

36 2 2

9 6 6

11–43,800 (991)b

8–8230 (509)c

195–28,000 (1025)d

34 2 0 1 3

7 2 2 6 4

CA 125 at diagnosis (range; median)

Chemoresponse after primary treatment CR 28 PR 26 SD 6 PD 9 8 NAe

3

Abbreviations: NA = not available; CR = complete response; PR = partial response; SD = stable disease; PD = progressive disease. a Among 47 patients with effusions who received surgery as upfront treatment, 22 were debulked to ≤1 cm, 23 to N1 cm. Data were unavailable for 2 patients. b Available for 58 patients. c Available for 38 patients. d Available for 20 patients. e Disease response after chemotherapy could not be evaluated because of absent data, normalized CA 125 after primary surgery or missing CA 125 information and no residual tumor.

extraction with miRvana miRNA Isolation Kit (Ambion, Inc., Austin, TX). RNA quantity and quality were measured by NanoDrop 2000 (Thermo Fisher Scientific).

qRT-PCR was performed on the cDNA using the KAPA SYBERFAST Universal qPCR kit (Kapa Biosystems, Wilmington, MA) with the CFX Connect Real-Time system (Bio-Rad Laboratories, Hercules, CA). The final concentrations of template and primers were determined individually for every assay, calibrated based on a standard curve. Primer sequences are described in Table 2. Analysis was performed with BioRad CFX manager software. The fold change of genes of interest was calculated relative to RPLP0 or GAPDH and expressed as 2−ΔCT. PCR was performed with specific primers using KAPA2G Fast ReadyMix PCR Kit (Kapa Biosystems). The products were then loaded on agarose gel 1.5% (for product N120 bp) or 2% (for product b 120 bp). 2.6. Statistical analysis Statistical analysis was performed applying the SPSS-PC package (Version 24, Chicago IL). Probability of b0.05 was considered statistically significant. Comparative analysis of lncRNA expression in effusions, ovarian tumors, solid metastases and exosomes was performed using the Kruskal-Wallis H test. Analysis of the association between lncRNA expression levels in HGSC effusions and clinicopathologic parameters was executed using the Mann-Whitney U test. For this analysis, as well as for survival analysis, clinicopathologic parameters were grouped as follows: age: ≤60 vs. N60 years; effusion site: peritoneal vs. pleural; FIGO stage: III vs. IV; chemotherapy status: pre- vs. postchemotherapy specimens; residual disease (RD): ≤1 cm vs. N 1 cm; response to chemotherapy: complete response vs. partial response/stable disease/progressive disease. Progression-free survival (PFS) and overall survival (OS) were calculated from the date of the last chemotherapy treatment/diagnosis to the date of recurrence/death or last follow-up, respectively. Univariate survival analyses of PFS and OS were executed using the Kaplan-Meier method and log-rank test. Platinum resistance was defined as PFS ≤ 6 months according to guidelines published by the Gynecologic Oncology Group and progressive disease or recurrence was evaluated by RECIST criteria. Multivariate survival analysis was performed using the Cox regression model (Enter function).

2.3. Microarray analysis

3. Results

A custom-made SurePrint G3 Custom GE 8x60K lncRNA expression microarray with one-color microarray-based gene expression analysis (low input quick amp labeling) protocol from Agilent Technologies (Santa Clara, CA) was used. A total of 8 chambers on the microarray slide were used to represent the 3 anatomic sites and the exosomes in duplicates. 2 μl RNA from 5 specimens from each anatomic site (ovary, solid metastasis and effusion) were mixed and applied on a chamber in duplicates, i.e. a total of 10 samples analyzed. For the exosomes, 8 μl RNA from 15 samples in duplicate were used, a total of 30 samples. To compare the expression of ncRNAs in our series to that of normal ovary, we used the published GSE38666 (https://www.ncbi.nlm.nih. gov/geo/query/acc.cgi?acc=GSE38666) array dataset, which compared 12 normal and the 18 serous samples.

3.1. Array analysis Among a total of 17,696 ncRNAs included in the array, approximately two-thirds were expressed in HGSC (Supplementary Tables 1–4). Comparative array analysis of HGSC at different anatomic sites showed significant differences in the expression of lncRNAs between HGSC specimens resected from the ovary compared to extra-ovarian sites (Table 3). These differences were confirmed in PCA analysis comparing Table 2 Primer sequences. Name RPLP0

2.4. Bioinformatics Array analysis was done using PARTEK GENOMIC SUITE 6.6 gene expression package. PCA, HEATMAP and VENN Diagrams were generated after using flooring and log2 transformation of the 8 arrays. To enable comparison between the GSR38666 data set and our microarray, we normalized both using Z-core. The list of differentially expressed genes in the comparative analyses was generated using an absolute ratio N 2, with the FDR correction available only in the comparative analysis of exosomes vs. the ovarian tumors.

Forward primer

CCAACTACTTCCTTAAGATCATCC AACTA GAPDH GACAGTCAGCCGCATCTTCT H19 GCACCTTGGACATCTGGAGT HOTAIR GGTAGAAAAAGCAACCACGAAGC Linc-RoR TCTCTCACCAGCCACCTCAA ESRG ACAGCCTTGTACCCTGGTCT MEG3 AGCTGTTGAGCCTTCAGTGT GAS5 GTCTTGCCTCACCCAAGCTA PVT1 TGGTTTTAAGGGAGGCTGTG Link-A AACCAGTCACCCAACCAGAG GATS CAAACATGAGGGAGGGTTGGA MALAT1 GGATCCTAGACCAGCATGCC

Reverse primer ACATGCGGATCTGCTGCA TTAAAAGCAGCCCTGGTGAC TTCTTTCCAGCCCTAGCTCA ACATAAACCTCTGTCTGTGAGTGCC CAGAGTGGCGATGTGTTTGG CAATGGTGCGAAGCTGTGTT TGTGCTTTGGAACCGCATCA TGGAGACACTGTTTTAATCTTCTTG AGTCGGGGTCTTACATTCCA CACAGGCCAGATGGAGTTTT CTCCCTGGCCCTCATTTGTC AAAGGTTACCATAAGTAAGTTCCA GAAAA

Please cite this article as: N. Filippov-Levy, et al., Expression and clinical role of long non-coding RNA in high-grade serous carcinoma, Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.01.004

4

N. Filippov-Levy et al. / Gynecologic Oncology xxx (2018) xxx–xxx

the 4 specimens groups (Fig. 1-A), as well as in a Venn diagram of the expressed transcripts in solid metastases, effusions and exosomes (Fig. 1-B). Anatomic site-related differences in the expression of 10 lncRNAs chosen for validation is shown in the heat map designated Fig. 1-C. Our validated data show significant anatomic site variation similar to the data obtained from the array analysis (Table 4-A). As we had no access to normal ovarian epithelial tissue for comparison analysis, we compared the expression profile in our samples of the 10 lncRNAs chosen for validation to a published array. Our validated data show significant anatomic site variation similar to the data obtained from the array analysis (Table 4-B). Fig. 1-D demonstrates the validity of our approach. The expression of 8 genes was compared in our array and the published one, showing relative similarity of the two. 3.2. Validation of anatomic site-related expression Only a small fraction of the lncRNAs has attributed function. In order to make our data more biologically and clinically meaningful, we chose from the list of significantly differentially expressed lncRNAs to validate the expression of lncRNA transcripts with a known attributed function or previously reported expression in cancer. The expression of 10 lncRNAs was analyzed in a validation series of 180 specimens, including 77 effusions, 40 ovarian HGSC specimens, 21 solid metastases and exosomes from 42 effusion supernatants. LincRoR was not found in the analyzed specimens. Significant differences in expression were observed for 5 of 9 remaining lncRNAs (ESRG, Link-A, MEG3, GATS and PVT1, all p b 0.001; Fig. 1-E). ESRG expression was highest in exosomes, followed by solid metastases and ovarian tumors, with lowest expression in effusions (mean rank = 137, 105, 85 and 64, respectively). Link-A expression was highest in solid metastases, followed by ovarian tumors and effusions, with lowest expression in exosomes (mean rank = 122, 109, 92 and 55, respectively). MEG3 expression was highest in solid metastases, followed by ovarian tumors and exosomes, with lowest expression in effusions (mean rank = 120, 104, 98 and 71, respectively). GATS expression was similar in solid metastases and ovarian tumors, intermediate in effusions and lowest in exosomes (mean rank = 127, 128, 88 and 40, respectively). The same distribution was observed for PVT1 (mean rank = 111, 110, 95 and 54, respectively). 3.3. Association with clinicopathologic parameters and survival The clinical relevance of the validated lncRNAs was analyzed in the effusion cohort, which included the largest number of patients. MEG3 levels were significantly higher in pre-chemotherapy effusions tapped at diagnosis (n = 40) compared to post-chemotherapy specimens (p = 37; p = 0.017). Link-A expression was higher in effusions from older (N 60 years) patients (p = 0.049). Higher GATS expression was seen in specimens from patients who were sub-optimally debulked (p = 0.046), whereas the opposite was true for H19 (p = 0.032). Higher H19 levels were measured in effusions from patients whose tumors showed primary resistance to chemotherapy (PFS ≤ 6 months). The follow-up period for the 77 patients with HGSC effusions ranged from 1 to 155 months (mean = 37 months, median = 26 months). PFS ranged from 0 to 81 months (mean = 9 months, median = 6 months).

At the last follow-up, 70 patients were dead of disease, 5 were alive with disease and 1 was with no evidence of disease. One patient was lost to follow-up. In univariate survival analysis, higher ESRG levels were significantly related to longer OS in the entire cohort (p = 0.023), as well as in analysis limited to patients with pre-chemotherapy primary diagnosis specimens (p = 0.048; Fig. 2-A, B). Higher Link-A levels in postchemotherapy specimens was significantly associated with longer OS (p = 0.015) and PFS (p = 0.023; Fig. 2-C, D). Among clinicopathologic parameters, patient age was significantly associated with survival in the entire series (p = 0.014), whereas FIGO stage and RD volume, the latter assessed only in patients with upfront surgery, were unrelated to OS (p N 0.3; Supplementary Fig. 1-A to C). The parameters entered into the Cox multivariate analysis included patient age and ESRG levels, as well as lncRNAs with p-value b 0.2 in univariate analysis (Link-A: p = 0.187; MEG3: p = 0.133). Age was the only parameter with independent prognostic role (p = 0.029; (p N 0.3; Supplementary Table 5-A). None of the clinicopathologic parameters was related to OS in analysis limited to patients with pre-chemotherapy primary diagnosis specimens. Cox multivariate analysis was consequently not performed for this group. For patients with post-chemotherapy effusions, age was significantly related to OS (p = 0.029), with an association of p b 0.2 for FIGO stage (p = 0.104), while RD volume was unrelated to OS (Supplementary Fig. 1-D to F). In Cox multivariate analysis including these 2 parameters, as well as Link-A and ESRG (p = 0.015 and p = 0.187 in univariate analysis, respectively), Link-A was the only independent prognostic marker in this group (p = 0.045; Supplementary Table 5-B). None of the clinicopathologic parameters was related to PFS in patients with post-chemotherapy primary diagnosis specimens. Cox multivariate analysis was consequently not performed. 4. Discussion The majority of ovarian cancer patients are diagnosed with advanced stage disease, which is associated with poor outcome. There is therefore an unmet need to develop novel biomarkers for early detection of this disease. lncRNAs, initially considered junk RNA, turned out to control cellular processes by acting as scaffolds, decoys and guides by their ability to interact with DNA, RNA and proteins. Thereby, lncRNAs regulate gene expression via epigenetic control, splicing, and protein stability. Numerous studies have shown specific lncRNA expression patterns related to various physiological and pathological conditions. Whereas several lncRNAs have been studied in ovarian cancer in vitro and in clinical specimens, no comprehensive analysis has been performed analyzing their expression at various anatomical sites to the best of our knowledge. Data are additionally limited with respect to the clinical outcome of lncRNAs in this cancer. In the current study, we identified over 15,000 transcripts that were expressed in HGSC out of a total of 17,696 present in the array. Comparison of HGSC specimens at different anatomic sites identified several hundred transcripts with site-specific expression. Notably, there were profound differences between the solid specimens (ovarian and peritoneal) and the lncRNA profile in the exosomal fraction isolated from the effusion fluid. Bioinformatics analyses estimate about 20,000 lncRNA transcripts in the human genome, of which only about 300 currently have attributed

Table 3 Summary of microarray results. Anatomic site

Total

Not expressed

Expressed

p b 0.05

Overexpressed

Underexpressed

Metastasis vs. ovary Effusion vs. ovary Exosomes vs. ovary

17,696 17,696 17,696

4483 3538 1052

13,213 14,158 16,644

241 406 3634

114 137 1676

127 269 1958

Please cite this article as: N. Filippov-Levy, et al., Expression and clinical role of long non-coding RNA in high-grade serous carcinoma, Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.01.004

N. Filippov-Levy et al. / Gynecologic Oncology xxx (2018) xxx–xxx

5

Fig. 1. Long non-coding RNA (lncRNA) array results. A. PCA diagram demonstrating the difference in lncRNA expression between HGSC localized to the ovary (red), solid metastasis (blue), effusion (green) and exosomes (purple). The exosome cargo is clearly distinct from the other groups. B. Venn diagram showing the overlap in lncRNA content, at 2-fold change, between solid metastases, effusions and exosomes. These 3 anatomic sites were compared to the ovarian tumors as reference. Thus, all lncRNAs overexpressed in solid metastases, effusions or exosomes are present in the primary tissue as well. C. Heat map of the 10 validated genes showing the differential expression of lncRNA in tumor cells at the various anatomic sites and in exosomes. D. Comparison of the expression of 8 lncRNAs between the present study and GSE38666. E. Validated lncRNAs with significantly different expression at different anatomic sites (see text; all p b 0.001). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

function. In order to focus on differences which may have biological relevance based on current data, we chose to validate the expression of lncRNA transcripts with a known attributed function or previously reported expression in cancer. Among the 10 validated transcripts, 5

showed significantly different site-related expression with a pattern similar to the array. ESRG (Embryonic Stem Cell Related) expression was higher in solid specimens compared to effusion-derived cells (p b 0.001) and its

Table 4-A lncRNA selected for validation at the three anatomic sites and exosomes. lncRNA

ESRG

LINK-A

GAS5

MEG3

PVT-1

GATS

Linc-RoR

H19

HOTAIR

MALAT1

Validation

+

+

+

+

+

+

−a

+b

+b

+b

a b

Present in the array by not in the validation set. Not validated in exosomes.

Please cite this article as: N. Filippov-Levy, et al., Expression and clinical role of long non-coding RNA in high-grade serous carcinoma, Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.01.004

6

N. Filippov-Levy et al. / Gynecologic Oncology xxx (2018) xxx–xxx

Table 4-B Comparison between cancer and normal ovarian tissuea. lncRNA

ESRG

LINK-A

GAS5

MEG3

PVT-1

H19

HOTAIR

MALAT1

Fold change (+ or −) in tumor compared to normal tissue p-Value

4.58756 0.00038

1.30498 0.506009

1.16606 0.451995

1.96073 0.008954

2.62271 5.85E−07

1.84923 0.168358

3.30473 0.00775497

−1.98557 2.50E−02

a Data was downloaded from published array No. GSE38666. The array included tissue from 18 patients with ovarian carcinoma and 12 samples of normal ovarian surface epithelium. Data regarding GATS and Linc-RoR was not present in the array.

expression in effusions was associated with better OS. The precise role of this gene and its clinical relevance are unclear at present. However, its expression has been reported in certain tumors (intracranial germinoma and embryonal carcinoma) and in human embryonic stem cells [7]. LINK-A (long intergenic non-coding RNA for kinase activation) expression was higher in solid specimens compared to effusion-derived cells and very low in exosomes. Link-A expression was associated with better OS and PFS in patients with post-chemotherapy effusions, independently so for the former. LINK-A was reported to be differentially expressed in comparative analysis of triple-negative breast cancer

A

and matched normal tissue and promotes tumor growth through activation of normoxic hypoxia-induced factor-α (HIF1α). In glioma cells, LINK-A mediates invasion and proliferation by regulating lactate dehydrogenase A (LDH-A) [8]. Moreover, Link-A was found to hyperactivate AKT through Phosphatidylinositol-3,4,5-trisphosphate (PIP3), leading to resistance to AKT inhibitors [9]. These studies, though focused on other cancers, do not concur with our observations and suggest that the biological and clinical role of LINK-A may vary among different cancers. MEG3 (maternally expressed gene 3) expression was 10-fold higher in solid lesions and exosomes compared to effusion-derived cells. MEG3 was also found to be higher in pre-compared to post-chemotherapy

B

p=0.023

C

p=0.048

D

p=0.015

p=0.023

Fig. 2. ESRG and Link-A expression in HGSC effusions is associated with longer overall survival. A. Kaplan-Meier survival curve showing the association between ESRG expression and overall survival (OS) in the entire effusion cohort (n = 77). Patients with effusions with high (above median) ESRG expression levels (n = 38; solid line) had mean OS of 50 months compared to 29 months for patients with effusions having low ESRG expression levels (n = 39, dashed line; p = 0.023). B. Kaplan-Meier survival curve showing the association between ESRG expression and OS for patients with pre-chemotherapy effusions (n = 40). Patients with effusions with high (above median) ESRG expression levels (n = 19; solid line) had mean OS of 61 months compared to 30 months for patients with effusions having low ESRG expression levels (n = 21, dashed line; p = 0.048). C. Kaplan-Meier survival curve showing the association between Link-A expression and OS for patients with post-chemotherapy effusions (n = 37). Patients with effusions with high (above median) Link-A expression levels (n = 16; solid line) had mean OS of 43 months compared to 25 months for patients with effusions having low Link-A expression levels (n = 21, dashed line; p = 0.015). D. Kaplan-Meier survival curve showing the association between Link-A expression and progression-free survival (PFS) for patients with post-chemotherapy effusions (n = 37). Patients with effusions with high (above median) Link-A expression levels (n = 16; solid line) had mean PFS of 9 months compared to 5 months for patients with effusions having low Link-A expression levels (n = 21, dashed line; p = 0.023).

Please cite this article as: N. Filippov-Levy, et al., Expression and clinical role of long non-coding RNA in high-grade serous carcinoma, Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.01.004

N. Filippov-Levy et al. / Gynecologic Oncology xxx (2018) xxx–xxx

effusions. MEG3 expression was previously reported to be decreased in ovarian cancer compared to normal tissue and its promoter highly methylated [10]. Upregulation of MEG3 in ovarian cancer cell lines induced cell cycle arrest and inhibited proliferation [11]. MEG3 was further reported to activate WT p53 and repress Notch pathway in tumor cells [12,13]. Repressed MEG3 expression was an unfavorable risk factor for survival in breast cancer [14]. MEG3 is additionally associated with cisplatin resistance [15]. The higher expression of MEG3 in solid specimens in the present study may suggest that the tumor-promoting effect of this molecule predominantly exists at these anatomic sites, whereas it may be replaced by other LncRNAs in HGSC effusions. PVT-1 had comparable expression in tumor cells at all anatomic sites in our study. However, it was absent from the exosomal fraction. It was overexpressed in tumor tissue compared to normal tissue. In ovarian cancer PVT-1 overexpression was shown to be associated with cisplatin resistance [16] and is a part of a lncRNA signature that can predict survival and relapse [17]. This transcript represents a long non-coding RNA locus that has been identified as a candidate oncogene. Increased copy number and overexpression of this gene are associated with many types of cancers, including breast and ovarian cancers [18]. Consistent with its association with various types of cancer, transcription of this gene is regulated by the tumor suppressor p53 through a canonical p53-binding site, and it has been implicated in regulating levels of the proto-oncogene MYC to promote tumorigenesis [19]. HOTAIR (HOX antisense intergenic RNA) was highly expressed in all tumor samples without anatomic site-specific differences. However, a 3-fold increase compared to normal tissue was observed. HOTAIR expression is elevated in ovarian cancer compare to normal tissue [20]. HOTAIR expression was associated with poor patient survival in several studies [21–23]. However, no association with chemoresponse or survival was observed in our cohort. This may reflect differences in the series studies in terms of cohort size, specimen type (solid specimens vs. effusions) and histology (different histotypes vs. only HGSC). In ovarian cancer HOTAIR was shown to be involved in various cell pathways, including Wnt/β-catenin and NF-κB signaling promoting resistance to treatment with cisplatin or carboplatin [21,24,25], regulation of proliferation and apoptosis [23], and promotion of metastasis by regulation of matrix metalloproteinases (MMPs) and epithelial-tomesenchymal transition (EMT) [22,26]. Additionally, HOTAIR is part of different competing endogenous RNA (ceRNA) networks such as MAPK, PI3K and Rab22 [27–29]. MALAT-1 was expressed at all sites without significant differences, but its expression in tumor tissue was significantly higher than in normal tissue. A previous study found that MALAT-1 expression in ovarian cancer is elevated and promotes cell growth and migration [30]. Metaanalysis of various cancers, including ovarian cancer, reported that MALAT-1 was a marker of poor prognosis in the latter and disease recurrence [31,32]. GATS (stromal antigen 3 opposite strand); https://www.ncbi.nlm. nih.gov/nuccore/253970399/) was overexpressed in solids specimens compared to effusions, and its levels were higher in effusions from patients with larger RD volume. To our knowledge, there are no studies on GATS in ovarian cancer. Linc-ROR (long intergenic non-protein coding RNA-regulator of reprogramming) was not detected in our validation samples. There are currently no publications in English reporting expression of this molecule in ovarian carcinoma. Linc-ROR was reported to be elevated in breast carcinoma compared to paired normal tissue and contributed to EMT, migration, and invasion [33]. Linc-ROR was similarly shown to be overexpressed in in situ (DCIS) and invasive breast carcinoma compared to normal tissue in another study focusing on triple-negative breast carcinoma. In the latter study, it regulated invasion through suppression of miR-145, with resulting increase in ADP-ribosylation factor 6 (Arf6) [34].

7

H-19 had comparable expression level at all tumor sites and its expression was unrelated to survival in our cohort. H19 was previously reported to be expressed in ovarian serous cystadenomas, borderline tumors and carcinomas, and absent from their mucinous counterparts [35]. Loss of heterozygosity at this gene, as well as IGF-II, encoding insulin growth factor 2 (IGF2), was subsequently reported in ovarian carcinoma [36]. H19 was suggested to be a tumor suppressor negatively regulating IGF-II. However such association was not found in ovarian carcinoma [37]. A recent study linked H-19 expression to chemoresistance in vitro and recurrence in clinical HGSC [38]. Metaanalysis identified H-19 as part of a 6-lncRNA signature predicting recurrence in ovarian cancer [32]. In conclusion, this study is the first to compare lncRNA profiles at different anatomic sites in HGSC. Our data document unique lncRNA profiles at the various anatomic sites affected by this cancer, particularly between effusions and solid lesions, as well as between normal ovaries and tumor tissue. The differences between effusions and solid specimens reinforce our previous observations that carcinoma cells in effusions have distinct molecular characteristics, which may be attributed to growth in an anchorage-independent manner. We further report on significant differences between lncRNA expression in exosomes compared to tumor cells, particularly from solid lesions, suggesting a modifying effect for these vesicles on the tumor microenvironment. Lnc levels were further related to clinicopathologic parameters, though some relatively weakly, suggesting that further research is necessary in order to validate these findings. The association between lncRNA expression and longer survival for patients with effusions suggests that these molecules may have tumor suppressor functions, as previously reported for H19. Supplementary data to this article can be found online at https://doi. org/10.1016/j.ygyno.2018.01.004. Financial acknowledgment This work was supported by The Inger and John Fredriksen Foundation for Ovarian Cancer Research. Conflict of interest statement The authors declare that they have no competing interests.

References [1] X.D. Fu, Non-coding RNA: a new frontier in regulatory biology, Nat. Sci. Rev. 1 (2014) 190–204. [2] D.L. Cheng, Y.Y. Xiang, L.J. Ji, Lu XJ, Competing endogenous RNA interplay in cancer: mechanism, methodology, and perspectives, Tumour Biol. 36 (2015) 479–488. [3] J.E. Wilusz, H. Sunwoo, D.L. Spector, Long noncoding RNAs: functional surprises from the RNA world, Genes Dev. 23 (2009) 1494–1504. [4] T.R. Cech, J.A. Steitz, The noncoding RNA revolution-trashing old rules to forge new ones, Cell 157 (2014) 77–94. [5] L. Milane, A. Singh, G. Mattheolabakis, M. Suresh, M.M. Amiji, Exosome mediated communication within the tumor microenvironment, J. Control. Release 219 (2015) 278–294. [6] O. Vaksman, H.T. Stavnes, J. Kaern, C.G. Trope, B. Davidson, R. Reich, miRNA profiling along tumour progression in ovarian carcinoma, J. Cell. Mol. Med. 15 (2011) 1593–1602. [7] A. Lin, C. Li, Z. Xing, Q. Hu, K. Liang, L. Han, et al., The LINK-A lncRNA activates normoxic HIF1α signalling in triple-negative breast cancer, Nat. Cell Biol. 18 (2016) 213–224. [8] D. Wu, B. Zhao, X. Cao, J. Wan, Long non-coding RNA LINK-A promotes glioma cell growth and invasion via lactate dehydrogenase A, Oncol. Rep. 38 (2017) 1525–1532. [9] A. Lin, Q. Hu, C. Li, Z. Xing, G. Ma, C. Wang, et al., The LINK-A lncRNA interacts with PtdIns(3,4,5)P3 to hyperactivate AKT and confer resistance to AKT inhibitors, Nat. Cell Biol. 19 (2017) 238–251. [10] X. Sheng, J. Li, L. Yang, Z. Chen, Q. Zhao, L. Tan, Y. Zhou, J. Li, Promoter hypermethylation influences the suppressive role of maternally expressed 3, a long non-coding RNA, in the development of epithelial ovarian cancer, Oncol. Rep. 32 (2014) 277–285. [11] Y.L. Xiu, K.X. Sun, X. Chen, S. Chen, Y. Zhao, Q.G. Guo, Z.H. Zong, Upregulation of the lncRNA Meg3 induces autophagy to inhibit tumorigenesis and progression of epithelial ovarian carcinoma by regulating activity of ATG3, Oncotarget 8 (2017) 31714–31725. [12] D. Lv, R. Sun, Q. Yu, X. Zhang, The long non-coding RNA maternally expressed gene 3 activates p53 and is downregulated in esophageal squamous cell cancer, Tumour Biol. (2016 Oct 24) (Epub ahead of print).

Please cite this article as: N. Filippov-Levy, et al., Expression and clinical role of long non-coding RNA in high-grade serous carcinoma, Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.01.004

8

N. Filippov-Levy et al. / Gynecologic Oncology xxx (2018) xxx–xxx

[13] Q. Guo, Z. Qian, D. Yan, L. Li, L. Huang, LncRNA-MEG3 inhibits cell proliferation of endometrial carcinoma by repressing Notch signaling, Biomed Pharmacother 82 (2016) 589–594. [14] J.J. Zhang, S.H. Guo, B.Q. Jia, Down-regulation of long non-coding RNA MEG3 serves as an unfavorable risk factor for survival of patients with breast cancer, Eur. Rev. Med. Pharmacol. Sci. 20 (2016) 5143–5147. [15] J. Liu, L. Wan, K. Lu, M. Sun, X. Pan, P. Zhang, B. Lu, G. Liu, Z. Wang, The long noncoding RNA MEG3 contributes to cisplatin resistance of human lung adenocarcinoma, PLoS One 10 (2015), e0114586. [16] E. Liu, Z. Liu, Y. Zhou, Carboplatin-docetaxel-induced activity against ovarian cancer is dependent on up-regulated lncRNA PVT1, Int. J. Clin. Exp. Pathol. 8 (2015) 3803–3810. [17] P. Martini, L. Paracchini, G. Caratti, M. Mello-Grand, R. Fruscio, L. Beltrame, et al., lncRNAs as novel indicators of patients' prognosis in stage I epithelial ovarian cancer: a retrospective and multicentric study, Clin. Cancer Res. 23 (2017) 2356–2366. [18] Y. Guan, W.L. Kuo, J.L. Stilwell, H. Takano, A.V. Lapuk, J. Fridlyand, et al., Amplification of PVT1 contributes to the pathophysiology of ovarian and breast cancer, Clin. Cancer Res. 13 (2007) 5745–5755. [19] M. Cui, L. You, X. Ren, W. Zhao, Q. Liao, Y. Zhao, Long non-coding RNA PVT1 and cancer, Biochem. Biophys. Res. Commun. 471 (2016) 10–14. [20] H. Wu, X. Shang, Y. Shi, Z. Yang, J. Zhao, M. Yang, et al., Genetic variants of lncRNA HOTAIR and risk of epithelial ovarian cancer among Chinese women, Oncotarget 7 (2016) 41047–41052. [21] A.E. Teschendorff, S.H. Lee, A. Jones, H. Fiegl, M. Kalwa, W. Wagner, et al., HOTAIR and its surrogate DNA methylation signature indicate carboplatin resistance in ovarian cancer, Genome Med. 7 (2015) 108. [22] J.J. Qiu, Y.Y. Lin, L.C. Ye, J.X. Ding, W.W. Feng, H.Y. Jin, et al., Overexpression of long non-coding RNA HOTAIR predicts poor patient prognosis and promotes tumor metastasis in epithelial ovarian cancer, Gynecol. Oncol. 134 (2014) 121–128. [23] J.J. Qiu, Y. Wang, J.X. Ding, H.Y. Jin, G. Yang, K.Q. Hua, The long non-coding RNA HOTAIR promotes the proliferation of serous ovarian cancer cells through the regulation of cell cycle arrest and apoptosis, Exp. Cell Res. 333 (2015) 238–248. [24] J. Li, S. Yang, N. Su, Y. Wang, J. Yu, H. Qiu, et al., Overexpression of long non-coding RNA HOTAIR leads to chemoresistance by activating the Wnt/β-catenin pathway in human ovarian cancer, Tumour Biol. 37 (2016) 2057–2065. [25] A.R. Özeş, D.F. Miller, O.N. Özeş, F. Fang, Y. Liu, D. Matei, et al., NF-κB-HOTAIR axis links DNA damage response, chemoresistance and cellular senescence in ovarian cancer, Oncogene 35 (2016) 5350–5361.

[26] J. Wang, D. Chen, X. He, Y. Zhang, F. Shi, D. Wu, et al., Downregulated lincRNA HOTAIR expression in ovarian cancer stem cells decreases its tumorgeniesis and metastasis by inhibiting epithelial-mesenchymal transition, Cancer Cell Int. 15 (2015) 24. [27] T. Yiwei, H. Hua, G. Hui, M. Mao, L. Xiang, HOTAIR interacting with MAPK1 regulates ovarian cancer skov3 cell proliferation, migration, and invasion, Med. Sci. Monit. 21 (2015) 1856–1863. [28] L. Dong, L. Hui, HOTAIR promotes proliferation, migration, and invasion of ovarian cancer SKOV3 cells through regulating PIK3R3, Med. Sci. Monit. 22 (2016) 325–331. [29] Z. Zhang, J. Cheng, Y. Wu, J. Qiu, Y. Sun, X. Tong, LncRNA HOTAIR controls the expression of Rab22a by sponging miR-373 in ovarian cancer, Mol. Med. Rep. 14 (2016) 2465–2472. [30] Y. Zhou, X. Xu, H. Lv, Q. Wen, J. Li, L. Tan, et al., The long noncoding RNA MALAT-1 is highly expressed in ovarian cancer and induces cell growth and migration, PLoS One 11 (2016), e0155250. [31] Y. Wang, D. Xue, Y. Li, X. Pan, X. Zhang, B. Kuang, et al., The long noncoding RNA MALAT-1 is a novel biomarker in various cancers: a meta-analysis based on the GEO database and literature, J. Cancer 7 (2016) 991–1001. [32] K. Yang, Y. Hou, A. Li, Z. Li, W. Wang, H. Xie, et al., Identification of a six-lncRNA signature associated with recurrence of ovarian cancer, Sci. Rep. 7 (2017) 752. [33] P. Hou, Y. Zhao, Z. Li, R. Yao, M. Ma, Y. Gao, et al., LincRNA-ROR induces epithelial-tomesenchymal transition and contributes to breast cancer tumorigenesis and metastasis, Cell Death Dis. 5 (2014), e1287. [34] G. Eades, B. Wolfson, Y. Zhang, Q. Li, Y. Yao, Q. Zhou, lincRNA-RoR and miR-145 regulate invasion in triple-negative breast cancer via targeting ARF6, Mol. Cancer Res. 13 (2015) 330–338. [35] V. Tanos, D. Prus, S. Ayesh, D. Weinstein, M.L. Tykocinski, N. De-Groot, et al., Expression of the imprinted H19 oncofetal RNA in epithelial ovarian cancer, Eur. J. Obstet. Gynecol. Reprod. Biol. 85 (1999) 7–11. [36] C.L. Chen, S.M. Ip, D. Cheng, L.C. Wong, H.Y. Ngan, Loss of imprinting of the IGF-II and H19 genes in epithelial ovarian cancer, Clin. Cancer Res. 6 (2000) 474–479. [37] S.K. Murphy, Z. Huang, Y. Wen, M.A. Spillman, R.S. Whitaker, L.R. Simel, et al., Frequent IGF2/H19 domain epigenetic alterations and elevated IGF2 expression in epithelial ovarian cancer, Mol. Cancer Res. 4 (2006) 283–292. [38] Z.G. Zheng, H. Xu, S.S. Suo, Xu XL, M.W. Ni, Gu LH, et al., The essential role of H19 contributing to cisplatin resistance by regulating glutathione metabolism in highgrade serous ovarian cancer, Sci. Rep. 6 (2016), 26093.

Please cite this article as: N. Filippov-Levy, et al., Expression and clinical role of long non-coding RNA in high-grade serous carcinoma, Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.01.004