A microRNA meta-signature for pancreatic ductal adenocarcinoma

Key Paper Evaluation

Expert Review of Molecular Diagnostics Downloaded from informahealthcare.com by 92.234.37.213 on 02/28/14 For personal use only.

A microRNA meta-signature for pancreatic ductal adenocarcinoma Expert Rev. Mol. Diagn. Early online, 1–5 (2014)

Adam E Frampton*1, Elisa Giovannetti2, Nigel B Jamieson3, Jonathan Krell4, Tamara MH Gall1, Justin Stebbing4, Long R Jiao1 and Leandro Castellano4 1 Department of Surgery and Cancer, HPB Surgical Unit, Imperial College, Hammersmith Hospital campus, Du Cane Road, London, W12 0HS, UK 2 Department of Medical Oncology, VU University Medical Center, De Boelelaan 1117, 1081 HV, Amsterdam, The Netherlands 3 Academic Unit of Surgery, Faculty of Medicine, Glasgow Royal Infirmary, University of Glasgow, Alexandra Parade, Glasgow, G31 2ER, UK 4 Department of Surgery and Cancer, Division of Oncology, Imperial Centre for Translational and Experimental Medicine (ICTEM), Imperial College, Hammersmith Hospital campus, Du Cane Road, London, W12 0NN, UK *Author for correspondence: [email protected]

informahealthcare.com

Evaluation of: Ma MZ, Kong X, Weng MZ et al. Candidate microRNA biomarkers of pancreatic ductal adenocarcinoma: meta-analysis, experimental validation and clinical significance. J. Exp. Clin. Cancer Res. 32(1), 71 (2013). Due to its aggressive and late presentation, there is an urgent need for novel and reliable biomarkers for the diagnosis and prognostication of pancreatic ductal adenocarcinoma (PDAC). MiRNAs have been extensively profiled in PDAC tissues, biopsies, blood samples and other biofluids and their expression levels compared to normal and chronic pancreatitis (CP) specimens in order to identify the most relevant candidates. Consolidation of these activities has not been attempted until now. The evaluated meta-review by Ma et al. helps to define the use of miRNAs as biomarkers for detecting this tumor-type and predicting survival outcomes in PDAC. Based on frequency and consistency between microarray studies, they identified a miRNA meta-signature for recognising PDAC: upregulation of miR-21, 23a, 31, 100, 143, 155, and 221; with downregulation of miR-148a, 217 and 375. Furthermore, they validated high miR-21, high miR-31 and low miR-375 tumoural expression as independently prognostic for poor overall-survival (OS; n = 70). KEYWORDS: biomarkers • diagnostic • meta-signature • miRNA • pancreatic ductal adenocarcinoma • prognostic

The advent of high-throughput technologies for molecular profiling in cancer medicine has resulted in a wealth of information, but teasing out the most important and clinically useful biomarkers is often extremely difficult. Such is the case with miRNA expression data from PDAC profiling, although it is clear from more focused studies that a few select miRNAs are very influential during pancreatic tumorigenesis, such as oncomiR-21. Determining a common miRNA ‘signature’ for PDAC from all these studies has still not been achieved. To this end, Ma et al. [1] have performed a robust meta-review of the profiled PDAC miRNome and validated a 10-miRNA signature as a diagnostic tool. In addition, they demonstrated the prognostic value of three miRNAs, independent of other clinicopathologic variables in PDAC. This meta-signature will hopefully provide further insights into the processes underlying PDAC tumorigenesis, thereby supporting the development of clinical biomarkers and the identification of novel potential targets for therapeutic intervention.

10.1586/14737159.2014.893192

Summary of methods & results Methods Meta-profiling of the PDAC miRNome

An extensive literature search for microarray studies comparing PDAC to non-neoplastic pancreatic tissues was performed using Scopus, GEO, PubMed and ArrayExpress (up to May 2013). Other inclusion criteria were sufficient sample size, the use of miRNA microarray platforms, English language, available lists of miRNAs and their fold-changes. Following study selection, two meta-review approaches were employed. First, a vote-counting strategy was used to assess which assessed the number of studies that consistently identify a dysregulated gene with subsequent gene ranking in respect to total sample size and average foldchange. Of note, total sample size is considered more important than average fold-change. Second, the robust rank aggregation method was used which ‘detects genes that are ranked consistently better than expected and assigns a significance score for each gene’ [2]. These


2014 Informa UK Ltd

ISSN 1473-7159

1

Expert Review of Molecular Diagnostics Downloaded from informahealthcare.com by 92.234.37.213 on 02/28/14 For personal use only.

Key Paper Evaluation

Frampton, Giovannetti, Jamieson et al.

techniques allowed the creation of a miRNA meta-signature for PDAC. A meta-signature is the intersection of multiple differential expression profiles, whereby there are significantly more genes/miRNAs intersecting a given number of signatures than would be expected by chance [3]. This technique has recently been used to determine a miRNA meta-signature for lung cancer [4].

adherens junction; TGF-b signaling and cadherin signaling), axonal guidance, Wnt signaling, EGF receptor signaling and the RAS pathway. Interestingly, the 10 most strongly enriched GO processes also included multicellular organismal development and nervous system development, which have been less commonly associated with PDAC.

miRNA target gene enrichment analyses

miR-21, miR-31 & miR-375 expression have prognostic significance in PDAC

The gene targets for the miRNAs in the meta-signature were predicted using various algorithms. Targets scoring less than -1.25 on miRANDA/miSVR were selected for further analysis. Enrichment analyses for these genes were then created using Kyoto Encyclopedia of Genes and Genomes (KEGG), Panther and Gene Ontology (GO) tools in order to highlight the important biological processes and pathways. PDAC tissue samples

In order to validate the findings from the meta-review, a cohort of PDAC samples was collected and miRNA levels were measured. Between September 2010 and August 2011, 78 PDAC and adjacent normal pancreas specimens were obtained at surgery and macrodissected before being flash-frozen and stored. RNA was isolated and real-time reverse-transcription quantitative PCR (RT-qPCR) was used to quantify miRNA expression using specifically designed primers. For survival analyses, miRNA expression was dichotomized into high versus low based on the mean levels for each miRNA. Results Meta-review of miRNA profiling reveals a biologically relevant meta-signature for PDAC diagnosis

Eleven studies met the inclusion criteria. There were 538 PDAC versus 206 non-neoplastic samples in the pooled analysis, with a median of 47 patients per study. The results from the vote-counting strategy were comparable to those from the robust rank aggregation method. The latter method identified a statistically significant miRNA meta-signature for PDAC, including seven upregulated (miR-21, -23a, -31, -100, -143, -155 and -221) and three downregulated miRNAs (miR-148a, -217 and -375). The most frequently and consistently dysregulated miRNAs were miR-155 (eight studies), miR-21, miR-100 and miR-221 (each in seven studies). This diagnostic meta-signature was successfully validated by RT-qPCR using 78 PDAC samples (stages I–IIA, 41%; stages IIB–IV, 59%) and adjacent normal tissues. However, it is unclear whether the p-values were adjusted for multiple testing (i.e., 10 miRNAs were assessed). Furthermore, the clinical risk cut-offs for these miRNAs are unknown, as well as the sensitivity and specificity of the meta-signature itself for detecting PDAC. As such, it would be interesting to validate this meta-signature in an independent cohort to assess its diagnostic performance. The meta-signature was then found to be associated with extremely relevant GO processes and KEGG/Panther pathways known to be crucial to pancreatic tumorigenesis including cell– cell adhesion (also focal adhesion, regulation of actin cytoskeleton, doi: 10.1586/14737159.2014.893192

Next, a multivariate Cox proportional-hazards regression model was created using seven co-variates that were associated with poor OS at univariate analyses (i.e., pathological tumor (T) stage; resection margin status; vascular invasion and miR-21, miR-31, miR-155 and miR-375 expression). For this analysis, 70 patients were included, while the eight patients with metastatic stage IV disease were excluded. The multivariate model revealed that high miR-21 (HR: 2.599; 95% CI: 1.151–5.867; p = 0.021), high miR-31 (HR: 2.637; 95% CI: 1.298–6.635; p = 0.048) and low miR-375 (HR: 2.451; 95% CI: 1.429–5.135; p = 0.034) were associated with poor OS independent of other variables. However, follow-up for these 70 patients is not described, and neither is the subsequent administration of adjuvant chemotherapeutic and/or radiation therapy. We are told that the last patient sample collected was in August 2011 and as the paper was submitted in May 2013, we therefore assume that there was at least 21 months follow-up, which should be sufficient. Furthermore, the number of deceased at follow-up is not given. In a multivariate analysis, the ‘rule of tens’ states there should be a minimum of 10 outcome events per candidate predictor variable. Therefore, as seven co-variates were analyzed, all 70 patients would have to have succumbed during the follow-up in order to prevent overfitting. Expert commentary & five-year view

The purpose of meta-profiling is to discover whether a selected set of differential expression signatures share a significant intersection of genes or miRNAs (a meta-signature), and therefore, biological relatedness [3]. By using this approach, Ma et al. have succeeded in combining miRNA lists from several independent profiling studies and provided a pooled conclusion of miRNA expression in PDAC. The existence of a PDAC miRNA metasignature suggests that these miRNAs reflect the essential transcriptional and post-transcriptional features of this tumor type. It even highlighted regulation of relatively novel pathways, including axonal guidance, which was only recently implicated in PDAC by means of large-scale genomic sequencing efforts [5]. However, the usefulness of this meta-signature requires validation in more than just one independent cohort of patients. In addition, these patients and those in the profiling studies had varying stages of disease and resectability; therefore it would be wise to focus new molecular profiling experiments on the resectable (stage I–IIB) versus non-resectable (stages III–IV), as well as carefully evaluating survivorship. Indeed, our groups have recently produced two studies that have attempted to concentrate on these aspects. Giovannetti et al. [6] have profiled Expert Rev. Mol. Diagn.

Expert Review of Molecular Diagnostics Downloaded from informahealthcare.com by 92.234.37.213 on 02/28/14 For personal use only.

miRNA meta-signature for pancreatic cancer

19 stage-pT3N1 PDACs homogeneously treated with adjuvant gemcitabine that were also carefully selected according to their outcome (OS 30 months, n = 6; i.e., short/long OS). In this study, patients with long-OS had higher expression of miR-211, -1207-3p, -326 and lower expression of miR-4321. MiR-211 expression was then validated as the most predictive biomarker for treatment outcome in this subgroup of PDAC patients, and high levels were associated with better OS and DFS independent of other clinicopathologic variables. Unsurprisingly, none of the 10 miRNAs from the meta-signature discovered by Ma et al. [1] emerged among the most discriminatory miRNAs in this microarray. This is not only due to the focus on a particular subgroup of patients, but also to the laser-capture microdissection (LCM) of tissue samples that was performed prior to profiling. In our recent study [7], we concentrated on resectable PDAC patients (stage IIB) and performed an integrated molecular profiling that analyzed both miRNA and mRNA expression. This was done on macrodissected fresh tissues (n = 18), and then miRNA results were confirmed experimentally and ex vivo on LCM tissues. We found that compared to normal tissues, there were 20 differentially expressed miRNAs, and of those, three miRNAs (miR-21, -23a and -27a) had significantly enriched frequency of seed sequences within the differentially expressed genes in PDAC. Our integrated analysis therefore allowed us to prioritize dysregulated miRNAs, with the most biological relevance and influence on the PDAC transcriptome, for further investigation. We subsequently found that inhibition of this triple miRNA combination was able to reduce PDAC proliferation in vitro and in vivo greater than silencing oncomiR-21 alone. Furthermore, we demonstrated that high expression of the triple combination (miR-21/23a/27a) was independently prognostic for poor OS in resectable PDAC patients (n = 91). Thus, high expressors of this triple miRNA combination (miR-21/23a/27a) may benefit from anti-miRNA therapy, although the best way to deliver such a treatment and potential off-target effects are unknown. Considering the miRNAs within the meta-signature identified by Ma et al. [1], there are many important candidates. As in our study [7], and several previously, oncomiR-21 is known to be involved in PDAC tumorigenesis, invasion, metastasis and chemoresistance [8–10]. Recent analysis of The Cancer Genome Atlas pan-cancer miRNA-sequencing data has revealed that miR-21 is also the most highly expressed miRNA in many cancers [11]. Interestingly, Kadera et al. [12] have shown that miR-21 expression is increased within the stromal compartment, which is known to be responsible for the aggressive nature of PDACs. They conclude that ‘anti-miR-21 may represent a novel therapeutic strategy for dual targeting of both tumor and stroma in PDAC’ [12]. These findings bring into question whether molecular profiling studies should use LCM to dissect out the tumor cell compartment or not? Any biopsy or fine-needle aspiration of a pancreatic tumor will not sample just the cancerous cells, but also some of the desmoplastic stroma. Therefore, it seems reasonable that profiling is based informahealthcare.com

Key Paper Evaluation

on ‘bulk’ tumor, as biomarkers discovered in this way could be translated more easily into clinical practice. On the other hand, if we want to identify molecular aberrations in vivo and then work backward to reveal tumorigenic mechanisms and possible targets that could be ‘drugable’, then it makes sense to work on cell-specific genomic changes. In both our integrated molecular profiling [7], and the current study [1], tumors were macrodissected and there were miRNAs consistently identified as dysregulated in PDAC versus normal (i.e., upregulation of miR-21, -23a, -100, and -143; downregulation of miR-217). The other differences between these profiling results are likely to be due to the use of different microarray platforms, and this was one of the key reasons that Ma et al. [1] state for aggregating selected independent miRNA lists in order to reach a pooled conclusion. More importantly, different stages of disease were included in each microarray. Nevertheless, both studies found miR-23a upregulated in PDAC, along with its family member miR-27a, as well as miR-21 [1,7] MiR-27a is also significantly upregulated in PDAC in several other studies, including nonmicroarray ones [7,13–17]. Indeed, members of the miR-23a/ 24-2/27a cluster are known to be upregulated in many other malignancies, including leukemias, breast, gastric, lung, hepatocellular, bile duct and colorectal cancers [18–21]. While the context-dependent functions of this cluster in tumorigenesis are still under investigation, recent studies have elucidated several, mainly metastatic mechanisms for these miRNAs. Zheng et al. [21] have demonstrated in gastrointestinal cancers that FasL signaling activates ERK/MAPK, which inhibits E-cadherin expression, through induction of miR-23a, which is regulated by the NFAT4 and AP-1 complex. Loss of E-cadherin is the hallmark of the epithelial-to-mesenchymal transition and in PDAC is known to be associated with enhanced invasion and migration in vitro [22] and also with higher tumor grade, increased chemoresistance and worse prognosis in [23]. Acunzo et al. [20] have shown that HGF binds MET and induces miR-23a/24-2/27a cluster expression through transcription factor ELK1 in nonsmall-cell lung cancer. Moreover, miR-27a can then downregulate MET and EGFR, by either directly or indirectly targeting Sprouty2 (SPRY2). Of note, Li et al. [19] showed that EGF induces the expression of c-MYC, which promotes miR-23a/ 24-2/27a cluster expression and subsequently decreases SPRY2, thus promoting breast cancer cell migration and invasion and subsequent hepatic metastases. As a recent study also detected SPRY2 downregulation in PDAC, this highlights a potential role for this miRNA cluster in pancreatic tumorigenesis. Indeed, our integrated analysis [7] indentified several tumor suppressor genes targeted by miR-23a and -27a in PDAC [24]. The 10-miRNA meta-signature was able to distinguish PDAC from normal pancreas. However, the more clinically relevant problems are discerning PDAC from CP, or stratifying PDAC precursor lesions. Indeed, Bloomston et al. [25] found that there were 15 upregulated and 8 downregulated miRNAs in PDAC versus CP, including overexpression of miR-21 and -221 and decreased levels of miR-148a. Notably, these three miRNAs are components of the meta-signature. Therefore, it doi: 10.1586/14737159.2014.893192

Expert Review of Molecular Diagnostics Downloaded from informahealthcare.com by 92.234.37.213 on 02/28/14 For personal use only.

Key Paper Evaluation

Frampton, Giovannetti, Jamieson et al.

would be important to assess whether the meta-signature could effectively separate these two pathologies. In addition, several studies have now shown that miRNAs can differentiate between low and high malignant potential pancreatic cystic tumors [26]. Notably, two of the miRNAs in the meta-signature, miR-21 and -221, can predict malignant progression of pancreatic intraductal mucinous neoplasms (IPMN) [27]. However, it is unlikely that the entire meta-signature would be specific for this form of pancreatic cancer (i.e., carcinoma-ex-IPMN) or premalignant cystic lesions. In conclusion, based upon previous expression profiling studies, a miRNA meta-signature for PDAC appears to exist. While the mechanism of many of these miRNAs has been investigated, several remain to be elucidated. Future large-scale genomic profiling studies will expand our understanding of the genetic heterogeneity that underlies PDAC and detail the interactive role of miRNAs in tumorigenesis. MiRNAs hold potential as robust biomarkers to enhance the detection and

management of early PDAC through assessment in biofluids, including blood [28] and bile, and endoscopic ultrasound-guided fine-needle aspiration samples [27]. Currently, it remains uncertain if miRNAs can be used to reliably help detect PDAC preoperatively, identify high-risk premalignant neoplasms and/or stratify patients for various anti-cancer treatments. The next step will be to use these diagnostic and prognostic markers within prospective trials to ascertain if they can personalize outcome prediction and ultimately aid clinical decision-making. Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. No writing assistance was utilized in the production of this manuscript.

Key issues • This is the first study to consolidate miRNA expression profiling efforts in pancreatic ductal adenocarcinoma (PDAC) tissues. • A 10-miRNA meta-signature for PDAC diagnosis has been defined (upregulated: miR-21, -23a, -31, -100, -143, -155 and -221; downregulated: miR-148a, -217 and -375). • MiR-21, -31 and -375 are each independently prognostic in PDAC, although their combination was not assessed.

References 1.

2.

3.

4.

5.

6.

Ma M-Z, Kong X, Weng M-Z, et al. Candidate microRNA biomarkers of pancreatic ductal adenocarcinoma: metaanalysis, experimental validation and clinical significance. J Exp Clin Cancer Res 2013; 32(1):71

7.

8.

Kolde R, Laur S, Adler P, Vilo J. Robust rank aggregation for gene list integration and meta-analysis. Bioinformatics 2012; 28(4):573-80 Rhodes DR, Yu J, Shanker K, et al. Large-scale meta-analysis of cancer microarray data identifies common transcriptional profiles of neoplastic transformation and progression. Proc Natl Acad Sci USA 2004;101(25): 9309-14 Vo˜sa U, Vooder T, Kolde R, et al. Meta-analysis of microRNA expression in lung cancer. Int J Cancer 2013;132(12): 2884-93 Biankin AV, Waddell N, Kassahn KS, et al. Pancreatic cancer genomes reveal aberrations in axon guidance pathway genes. Nature 2012;491(7424):399-405 Giovannetti E, van der Velde A, Funel N, et al. High-throughput microRNA (miRNAs) arrays unravel the prognostic role of mir-211 in pancreatic cancer. PLoS One 2012;7(11):e49145

doi: 10.1586/14737159.2014.893192

9.

10.

11.

12.

Frampton AE, Castellano L, Colombo T, et al. MicroRNAs cooperatively inhibit a network of tumor suppressor genes to promote pancreatic tumor growth and progression. Gastroenterology 2014;146(1): 268-77; e218 Giovannetti E, Funel N, Peters GJ, et al. MicroRNA-21 in pancreatic cancer: correlation with clinical outcome and pharmacologic aspects underlying its role in the modulation of gemcitabine activity. Cancer Res 2010;70(11):4528-38 Hwang JH, Voortman J, Giovannetti E, et al. Identification of microRNA-21 as a biomarker for chemoresistance and clinical outcome following adjuvant therapy in resectable pancreatic cancer. PLoS One 2010;5(5):e10630 Moriyama T, Ohuchida K, Mizumoto K, et al. MicroRNA-21 modulates biological functions of pancreatic cancer cells including their proliferation, invasion, and chemoresistance. Mol Cancer Ther 2009; 8(5):1067-74 Hamilton MP, Rajapakshe K, Hartig SM, et al. Identification of a pan-cancer oncogenic microRNA superfamily anchored by a central core seed motif. Nat Commun 2013;4:2730 Kadera BE, Li L, Toste PA, et al. MicroRNA-21 in pancreatic ductal

adenocarcinoma tumor-associated fibroblasts promotes metastasis. PLoS One 2013;8(8): e71978 13.

Szafranska AE, Davison TS, John J, et al. MicroRNA expression alterations are linked to tumorigenesis and non-neoplastic processes in pancreatic ductal adenocarcinoma. Oncogene 2007;26(30): 4442-52

14.

Yihui M, Shuangni Y, Wugan Z, et al. miR-27a regulates the growth, colony formation and migration of pancreatic cancer cells by targeting Sprouty2. Cancer Lett 2010;298(2):150-8

15.

Zhang Y, Li M, Wang H, et al. Profiling of 95 microRNAs in pancreatic cancer cell lines and surgical specimens by real-time PCR analysis. World J Surg 2009;33(4): 698-709

16.

Jamieson NB, Morran DC, Morton JP, et al. MicroRNA molecular profiles associated with diagnosis, clinicopathologic criteria, and overall survival in patients with resectable pancreatic ductal adenocarcinoma. Clin Cancer Res 2012;18(2):534-45

17.

Ohuchida K, Mizumoto K, Kayashima T, et al. MicroRNA expression as a predictive marker for gemcitabine response after surgical resection of pancreatic cancer. Ann Surg Oncol 2011;18(8):2381-7

Expert Rev. Mol. Diagn.

miRNA meta-signature for pancreatic cancer

18.

19.

Expert Review of Molecular Diagnostics Downloaded from informahealthcare.com by 92.234.37.213 on 02/28/14 For personal use only.

20.

21.

Chhabra R, Adlakha YK, Hariharan M, et al. Upregulation of miR-23a-27a-242 cluster induces caspase-dependent and -independent apoptosis in human embryonic kidney cells. PLoS One 2009; 4(6):e5848

22.

Li X, Liu X, Xu W, et al. c-MYC-regulated miR-23a/24-2/27a cluster promotes mammary carcinoma cell invasion and hepatic metastasis by targeting Sprouty2. J Biol Chem 2013;288(25):18121-33

23.

Acunzo M, Romano G, Palmieri D, et al. Cross-talk between MET and EGFR in non-small cell lung cancer involves miR-27a and Sprouty2. Proc Natl Acad Sci USA 2013;110(21):8573-8 Zheng H, Li W, Wang Y, et al. miR-23a inhibits E-cadherin expression and is regulated by AP-1 and NFAT4 complex

informahealthcare.com

from normal pancreas and chronic pancreatitis. JAMA 2007;297(17): 1901-8

during Fas-induced EMT in gastrointestinal cancer. Carcinogenesis 2014;35(1):173-83

24.

25.

von Burstin J, Eser S, Paul MC, et al. E-Cadherin regulates metastasis of pancreatic cancer in vivo and is suppressed by a SNAIL/HDAC1/HDAC2 repressor complex. Gastroenterology 2009;137(1): 361-71; e365 Gall TM, Frampton AE. Gene of the month: E-cadherin (CDH1). J Clin Pathol 2013;66(11):928-32 Ma Y, Yu S, Zhao W, et al. miR-27a regulates the growth, colony formation and migration of pancreatic cancer cells by targeting Sprouty2. Cancer Lett 2010; 298(2):150-8 Bloomston M, Frankel WL, Petrocca F, et al. MicroRNA expression patterns to differentiate pancreatic adenocarcinoma

Key Paper Evaluation

26.

Frampton AE, Gall TM, Giovannetti E, et al. Distinct miRNA profiles are associated with malignant transformation of pancreatic cystic tumors revealing potential biomarkers for clinical use. Expert Rev Mol Diagn 2013;13(4):325-9

27.

Farrell JJ, Toste P, Wu N, et al. Endoscopically acquired pancreatic cyst fluid microRNA 21 and 221 are associated with invasive cancer. Am J Gastroenterol 2013; 108(8):1352-9

28.

Schultz NA, Dehlendorff C, Jensen BV, et al. MicroRNA biomarkers in whole blood for detection of pancreatic cancer. JAMA 2014;311(4):392-404

doi: 10.1586/14737159.2014.893192