Pancreatic cancer is the eighth most common cancer and has an overall 5-year survival rate lower than 10%. Because of their ability to regulate gene expression, microRNAs can act as oncogenes or tumor-suppressor genes and so have garnered interest as possible prognostic and therapeutic markers during the last decade. However, the prognostic value of microRNA expression in pancreatic cancer has not been thoroughly investigated. We measured the levels of miR-155, miR-203, miR-210, miR-216, miR-217 and miR-222 by quantitative RT-PCR in a cohort of 56 microdissected pancreatic ductal adenocarcinomas (PDAC). These microRNAs were chosen as they had previously been shown to be differentially expressed in pancreatic tumors compared to normal tissues. The possible association of microRNA expression and patients' survival was examined using multivariate Cox's regression hazard analyses. Interestingly, significant correlations between elevated microRNA expression and overall survival were observed for miR-155 (RR = 2.50; p = 0.005), miR-203 (RR = 2.21; p = 0.017), miR-210 (RR = 2.48; p = 0.005) and miR-222 (RR = 2.05; p = 0.035). Furthermore, tumors from patients demonstrating elevated expression levels of all 4 microRNAs possessed a 6.2-fold increased risk of tumor-related death compared to patients whose tumors showed a lower expression of these microRNAs. This study provides the first evidence for an oncogenic activity of miR-155, miR-203, miR-210 and miR-222 in the development of pancreatic cancer as has been reported for other tumor types. Furthermore, the putative target genes for these microRNAs suggest a complex signaling network that can affect PDAC tumorigenesis and tumor progression.
Because of its poor prognosis, pancreatic cancer is one of the leading causes of cancer-related death, despite its relative low incidence with 9 cases per 100,000 people.1 The poor prognosis of patients is a result of the late clinical presentation and the high metastatic potential. Three-fourths of pancreatic carcinomas are ductal adenocarcinomas (PDAC).2
There are molecular prognostic markers known for pancreatic cancers. For example, higher levels of protein expression of VEGF or EGFR have been shown to associate with poorer survival rates.3 Besides non-physiological KRAS activation, which is mutated in about 95% of all pancreatic carcinomas,4 inactivation of known tumor-suppressor genes such as p53 or SMAD4 (deleted in 75% or 55% of all PDACs, respectively) have been demonstrated.5, 6
MicroRNAs are small, non-coding RNAs of endogenous origin, which mainly function as negative regulators of gene expression. The association of altered microRNA expression with cancerogenesis as well as tumor progression is well established.7–9 There is a growing number of microRNAs, which are classified as oncogenes or tumor-suppressor genes.10 For instance, miR-17–92-cluster has gained interest by being regulated via c-myc and its ability to accelerate tumor formation.11, 12 Also, let-7 expression was described to correlate with a poor prognosis when being underexpressed in adenocarcinomas of the lung.13 Furthermore, the miR-15–16 cluster targeting BCL-2 was shown to be down-regulated in chronic lymphocytic leukemia (CLL) and to contribute to tumor progression.14 Altogether, misregulated expression of microRNAs seems to have a major impact on tumor progression and prognosis in many malignancies.
Interestingly, recent studies have demonstrated that the expression of a subset of microRNAs varies in tumors compared to normal tissues and tumors derived from other patients15–18 summarized in Supporting Table 1. However, reports about a correlation of expression levels of these microRNAs to the patient's outcome are limited. Specifically, the elevated expression of miR-196a-2 and miR-21 has been shown to significantly correlate with a poor outcome for PDAC patients.18, 19 The aim of this study is to analyze the expression levels of the subset of microRNAs, previously shown to be differentially expressed in PDAC and explore a potential effect on PDAC survival. The microRNAs analyzed were miR-155, miR-203, miR-210 and miR-222 that were recently described by Szafranska et al. as differentially expressed in PDAC compared to normal pancreatic tissue, as well as the pancreas-specific microRNAs miR-216 and miR-217.17
Material and Methods
The analyzed cohort of patients underwent primary surgery for PDAC in the years 2001–2005 in the Dept. of Surgery 1 at the University of Ulm, Germany. PDAC tissue of 56 patients (22 females and 34 males, age-range 34–80 years; mean age: 61.7 years) was snap-frozen directly after surgery and used for microdissection and subsequent RNA isolation. All patients with tumors giving intact RNA were included into the study. The mean observation time was 15.99 (range: 1–61) months, the median survival rate 15.26 (range: 1–52) months. All patients gave written informed consent and approval of the local ethics committee was obtained.
Microdissection and RNA isolation
For the exclusion of stromal tissue and enrichment of neoplastic cells, fresh frozen tissue samples from the 56 PDAC patients were microdissected on cresyl violet stained sections according to the method described by Burgemeister et al.20 Subsequent RNA isolation was performed using the Innuprep RNA mini-kit (AJ Innuscreen GmbH, Berlin, Germany) and the integrity of the isolated total RNA was confirmed with the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA).
cDNA synthesis and measurement of microRNA expression
Ten nanogram of total RNA of each sample were used for cDNA synthesis of the selected microRNAs miR-155, miR-203, miR-210, miR-216, miR-217 and miR-222 using stem-loop-primers (single kits by Applied Biosystems, Darmstadt, Germany). Reverse transcription reactions were carried out with the TaqMan Reverse Transcription Kit according to manufacturer's protocol (Applied Biosystems, Darmstadt, Germany). For each of the selected microRNAs, real-time-PCR measurements with specific Taqman-primers were performed in duplicate measurements on the Rotorgene 3000 Real-time PCR System (LTF, Wasserburg, Germany), determining a mean CT-value for each sample. CT values of the different samples were compared with the ΔΔCT method described by Livak and Schmittgen.21 Expression of 18S rRNA was homogenous in all samples and served as internal reference.
An EFNA3-3′UTR sequence carrying the binding region of miR-210 was inserted in a psiCheck2-Vector (Promega, Madison, WI) harboring Renilla Luciferase sequence and Firefly Luciferase sequence. The construct was transfected in HEK293 cells via lipotransfection (Interferin, Polyplus Transfection, New York, NY). After cotransfection of control microRNA or miR-210-mimic, respectively, Renilla Luciferase activity was measured using the Dual-Glo Kit (Promega, Madison, WI). Firefly Luciferase activity served as an internal control.
After cultivation under hypoxic or normoxic conditions for 24 hr, cells were harvested and protein isolated using RIPA buffer. Cell lysates were loaded on an SDS gel and the protein was transferred to a PVDF membrane. Unspecific protein binding was blocked by incubation in 5% non-fat dry milk for an hour. Protein expression was analyzed with primary antibodies against EFNA3 (1:500 from R&D, Wiesbaden, Germany) and then incubated with secondary antibodies (anti-mouse; DakoCytomation, Glostrup, Denmark). After washing, protein bands were visualized with an ECL kit (Amersham, Freiburg, Germany).
Cell colony formation assay
Cells of the non-invasive mamma carcinoma cell line MCF-7 were treated with miR-210 antagomir (2′-O-methylribonucleotid-inhibitor against miR-210; Ambion, Austin, TX) and cultivated under hypoxic and normoxic conditions for 24 hr, respectively. Inhibitors were transfected in cells at a concentration of 100 nM using Oligofectamine Reagent (Invitrogen, Karlsruhe, Germany). A negative inhibitor (Ambion) was transfected into a parallel cell sample cultivated under the same conditions and served as a negative control. Twenty-four hours after transfection and hypoxic treatment, cells were seeded in 25 cm2 flasks in defined cell aliquots and have been allowed to grow for at least 10 days. Then the cell colonies were fixed with 3% formalin and stained with Giemsa stain (Sigma, St. Louis). Cell colony forming assays were performed in triplicates for each cell line.
Cut-off values for statistical analyses were set according to the median expression levels of the specific microRNAs. Multivariate survival analysis according to Cox's proportional hazards regression model (adjusted for tumor stage, sex and type of tumor resection) were performed with SPSS 12.0 software (SPSS, Chicago). p-Values < 0.05 were considered significant.
Expression levels of 6 microRNAs were measured in 56 PDAC samples by qRT-PCR and compared to 18S rRNA expression according to the ΔΔCT method.21 The median expression was 0.23 (range: 0.006–122.9) for miR-155, 0.155 (range: 0.009–53.5) for miR-203, 0.15 (range: 0.001–39.9) for miR-210, 0.31 (range: 0.034–121.07) for miR-222, 0.25 (range: 0.003–213.3) for miR-216 and 0.56 (range: 0.022–9539.12) for miR-217.
In bivariate correlations we found a highly significant correlation between the expression of miR-155, miR-210 and miR-222 (p = 2.8 × 10−20 for miR-155 with miR-210, p = 6.2 × 10−8 for miR-210 with miR-222 and p = 5.7 × 10−10 for miR-155 with miR-222, respectively). MiR-203 expression was also significantly correlated to the expression of these 3 microRNA (p = 3.3 × 10−13 for miR-203 with miR-210; p = 1.6 × 10−13 for miR-203 with miR-155; p = 2.5 × 10−16 for miR-203 with miR-222).
Four out of 6 analyzed microRNAs, which are known to be tumor-associated and which had been previously shown to be expressed at higher levels in PDACs compared to normal tissues (miR-155, miR-203, miR-210 and miR-222) showed a significant correlation with the prognosis of PDAC patients. In multivariate Cox's regression hazard analysis, a higher expression of miR-155 resulted in a 2.50-fold increased risk of cancer-related death (p = 0.005). Furthermore, an elevated expression of miR-203, miR-210 and miR-222 was significantly correlated with a worse outcome of PDAC patients (RR = 2.21; p = 0.017; RR = 2.48; p = 0.005 and RR = 2.05; p = 0.035, respectively, Table 1). On the other hand, a lower expression of miR-217 correlated with a worse outcome of PDAC patients (RR = 2.33; p = 0.027), while miR-216 expression showed no influence on patients' survival (RR = 1.20, p = 0.576). Subsequently, we excluded patients with advanced disease (UICC stage IVb—concomitance of distant metastases) and/or local R2-resection since it can be assumed that prognosis of these patients is determined by occurrence of relapses or metastasis rather than other biological characteristics as expression levels of microRNAs. In multivariate Cox's regression hazard analyses, the relative risk of tumor-related death of patients with higher miR-203, miR-210 or miR-222 expression was increased (RR = 2.29; p = 0.023; RR = 2.50; p = 0.01 and RR = 2.20; p = 0.038, respectively), but not for patients with higher miR-155 expression (RR = 2.20; p = 0.029). In this analysis, neither miR-216 nor miR-217 expression were significantly correlated with the survival of PDAC patients (p = 0.141 and p = 0.066, respectively, Table 1). However, results of statistical analyses with all PDAC patients or at excluding patients with advanced disease did not differ significantly from each other, suggesting that expression of the 4 microRNAs (miR-155, miR-203, miR-210 and miR-222) is correlated with prognosis independent of the progression of the disease.
Table 1. Multivariate Cox's regression hazard analysis for correlation of microRNA expression with prognosis of patients with PDAC
To explore the effect of higher expression of all 4 significant microRNAs on PDAC patients' survival, we partitioned the data into 3 groups: 0 = low expression of all 4 microRNAs (n = 10); 1 = higher expression of 1, 2 or 3 microRNAs (n = 14), 2 = high expression of all 4 microRNAs (n = 15). A multivariate Cox's regression hazard analysis revealed a significant correlation (p = 0.034) between patients' outcome and the overexpression of all 4 microRNAs compared to the low expression of all 4 microRNAs in their PDACs. The observed risk of tumor-related death with 5.24-fold is moderate additive, i.e., 2.7-fold to 3.0-fold higher than the effect of any individual microRNA (Table 1, Fig. 1).
To study exemplary the function of microRNAs identified to have prognostiv relevance in PDACs; miR-210 was investigated in more detail. We performed a database search for putative target proteins using the search algorithms TargetScan 5.0, PicTar and miRanda (TargetScan: www.targetscan.org; PicTar: http://pictar.bio.nyu.edu/; miRanda: http://cbio.mskcc.org/cgi-bin/mirnaviewer/mirnaviewer.pl). A putative protein target identified by all of these databases is ephrin-A3 (EFNA3), a receptor-tyrosine-kinase ligand.
First, we analyzed the regulation of the EFNA3-3′UTR by the miR-210 in luciferase-assays. The EFNA3-3′UTR sequence carrying the binding position of miR-210 was inserted in a psiCheck2-Vector. Cotransfection of this vector with a miR-210-mimic or a control microRNA was performed. Transfection of miR-210-mimic reduced luciferase activitiy to 57.7% compared to the luciferase activity of cells with the vector control (p = 0.0093; 2-tailed student's t-test; Fig. 2).
Next we analyzed whether inhibition of miR-210 affects protein expression of EFNA3. We applied a miR-210 antagomir and transfected it into MCF-7 cells. Inhibition of miR-210 under hypoxic conditions resulted in an increase of EFNA3 protein to 86.1% compared to 46.2% in the untreated control (Fig. 3). However, EFNA3 protein did not increase significantly after inhibition of miR-210 in normoxic cells, suggesting another mechanism of translational control of EFNA3 mRNA under normoxia.
Functional effects of the inhibition of miR-210 on the clonogenic survival of MCF-7 cells were studied in colony formation assays. Cells were transfected with miR-210 antagomirs and cultivated under hypoxic or normoxic conditions for 24 hr. The inhibition of miR-210 resulted in a reduction of the clonogenic survival of MCF-7 cells to 59.6% under normoxia and to 23.8% under hypoxia, respectively, compared to the untreated control sample (Fig. 4). On the other hand, the transfection of the MCF-7 cells with a control inhibitor not targeting a specific microRNA did not result in a significant alteration of clonogenic survival.
Our results strongly indicate that patients whose PDACs express higher levels of any of the microRNAs, miR-155, miR-203, miR-210 and miR-222, but not miR-216 or miR-217, will have a poorer overall survival. Furthermore, the combined overexpression of all 4 microRNAs increases the relative risk of tumor-related death from a range between 2.20-fold and 2.50-fold to 5.24-fold. These data suggest an oncogenic function of these microRNAs in PDAC progression. Alterations in microRNA expression patterns in various pancreatic cancer entities are well studied. Lee et al. identified several microRNAs up-regulated or down-regulated in pancreatic carcinoma compared to normal pancreatic tissue. Among others, miR-221, miR-181a and -181c, miR-155, miR-100 and miR-21 were upregulated, whereas miR-375 was down-regulated.16 In another study, Roldo et al. investigated microRNA expression patterns in normal pancreatic tissue, endocrine pancreatic carcinoma and acinar carcinomas and compared them to each other.15 While differences between the expression patterns in endocrine versus acinar carcinomas were found, both entities of carcinomas were reported to be different compared to normal benign tissues by over-expressing miR-107, miR-103, miR-23a, miR-26b, miR-192 and miR-342.15 Intriguingly, in comparison to the results of Lee et al.,16 only miR-107 was also shown to be up-regulated in pancreatic endocrine carcinoma, while, in contrast, Roldo et al. described miR-155 as down-regulated in pancreatic endocrine tumors.15
The microRNAs analyzed in our study were selected according to the results of Szafranska et al., who compared the microRNA expression patterns of normal pancreatic tissue and PDAC samples, and found 84 microRNAs to be differentially expressed.17, 22 These included pancreas-specific microRNAs, such as miR-216 and miR-217, which were down-regulated in PDAC, but also miR-155 and miR-221, which were upregulated in these tumors. Szafranska et al. also observed miR-222 and miR-210 to be overexpressed, which have both been defined as oncogenic microRNAs.17 Furthermore, in a study by Bloomston and colleagues, miR-196a was strongly enriched in PDAC, and also miR-155, miR-210 and miR-221–222; miR-21, miR-100, miR-103, miR-107 and the miR-181 family members were found to be up-regulated in pancreatic tumors compared to normal pancreas tissue.18 MiR-196a was subsequently shown to be a significant prognostic marker.18 Recently, Zhang and coworkers identified also miR-196a and miR-221–222 but additionally the microRNAs miR-15b, miR-95, miR-186, miR-190 and miR-200b to be overexpressed in pancreatic cancer tissues and cell lines.23
It was interesting to note that the prognostic relevant microRNAs showed a strong correlation in their expressions to each other. This may point to a concerted transcriptional control of their genes and/or a stabilization of these microRNAs but this relationship needs further investigation.
In this study, miR-155 appeared to have a strong effect on survival of PDAC patients (RR = 2.50, p = 0.005), although the effect becomes less pronounced upon exclusion of cases with advanced disease (RR = 2.20, p = 0.029). miR-155 has been validated as a pancreatic biomarker, distinguishing normal pancreatic tissue from early pancreatic neoplasias.24 Furthermore, miR-155 has been found to be important in lymphogenesis and B-cell maturation, but it is also overexpressed in several types of B-cell lymphomas.25, 26 In addition, miR-155 and an orthologue of the Kaposi-Sarcoma-Virus negatively regulate BACH-1, a known transcriptional repressor, and LDOC1, a regulator of apoptosis, whose down-regulation contributes to tumorigenesis.27 TP53-induced nuclear protein 1 gene has been described to be down-regulated by miR-155, accelerating pancreatic tumor development.28
miR-210 was also found to effect survival of the PDAC patients (RR = 2.48, p = 0.005). miR-210 expression has previously been shown to be regulated by HIF-1α and to be up-regulated under hypoxic conditions.29 Its experimentally validated targets comprise of ephrin-A330; this study and neuronal pentraxin 1.31 Furthermore, inhibition of miR-210-mediated repression of ephrin-A3 has been suggested to restrict tubulogenesis and chemotaxis of endothelial cells,30 what seems to contribute to angiogenesis and accelerated tumor growth. Our results show the 3′UTR of EFNA3 is regulated by miR-210 and inhibition of miR-210 reduces EFNA3 protein levels as well as clonogenic cell survival of MCF-7 cells. We suggest that EFNA3 is a target protein of miR-210 and its expression also affects clonogenic cell survival. However, miR-210 also targets e2f transcription factor 3 (E2F3), a regulator of cell cycle progression and slows down tissue growths, but also selects more aggressive tumor cells, giving them a growth advantage.32
Importantly, miR-210 has also been described to be an independent negative prognostic factor for breast cancer patients.33
miR-203 (RR = 2.21, p = 0.017) was recently reported to be overexpressed in different tumors such as ovarian, bladder and colon.34–36 Similar to the data presented here in PDAC patients, in colon adenocarcinoma patients, the overexpression of miR-203 was shown to be significantly correlated with a poorer survival,36 suggesting an oncogenic function of miR-203. In contrast, miR-203 has been proposed to possess tumor-suppressor qualities in some hematopoetic malignancies.37, 38 Specifically, miR-203 was shown to inhibit the expression of ABL-1 and BCR-ABL-1, resulting in an inhibition of cell proliferation.
miR-222 (RR = 2.05, p = 0.035) has been described to be overexpressed in endothelial cells and to target c-kit in HUVECs and therefore to control the angiogenic properties of these cells.39, 40 Another validated target of miR-222 is p27 (Kip1), a cell cycle inhibitor, and its translational repression results in a growth inhibition in various cancer cell lines.41–43 However, upregulation of miR-222 alone is not sufficient for the occurrence of uncontrolled cell proliferation, but enables cells to enter the S-phase through avoiding the restriction point, which is controlled by its target proteins p27 (Kip1) and p57 (Kip2).44
In summary, miR-155, miR-203, miR-210 and miR-222 have been shown to play an important role not only in PDAC growth and development, but also in other malignancies. An in silico search for putative target proteins for miR-155 miR-203, miR-210, miR-222 results in a long list of proteins that can affect essential cellular functions as transcription, angiogenesis, cell cycle, epigenetics/chromatin remodeling, apoptosis and immune response (Supporting Table II; Fig. 5). Comparison of the putative mRNA targets for these 4 microRNA via TargetScan revealed, that at least miR-155 and miR-222 share several putative targets critical for the development and growth of tumors, among them MYBL1, TP53INP1, RAB1A, VEZF1, MAP3K10, SOX1 and NOVA1. miR-210 shares the brain-derived neurotrophic factor (BDNF) with miR-155 as a putative target and the zinc finger protein 148 (ZNF148) with miR-222. Furthermore, one of miR-155 putative targets is HIF-1α, which is the inductor of miR-210 under hypoxia.29 However, the overlapping putative targets of these 4 microRNAs are few. Therefore, it might be better to consider “target pathways” rather than target proteins. All prognostic relevant microRNA can affect several different pathways in tumor cells at the same time, what could explain the moderate additive effect of all 4 microRNAs on prognosis of PDAC patients (Figs. 1 and 5).
One potential mechanism for the decrease in survival seen in tumors that overexpress these microRNAs could lie in their noted effects on the immune system. Specifically, miR-155 and miR-222 target immune genes in different pathways,45 leading to chromatin remodeling and gene silencing as well as the initiation of immune-escape mechanisms. By modulating these pathways, miR-155 and miR-222 could contribute to the selection of more aggressive cancer cells and thus might be the interesting targets for the therapy of PDAC. Additionally, 3 targets were found to be regulated by one of these 2 microRNAs and to be differentially regulated in PDAC: BACH1 and HIF-1A (miR-155) and PPP3R1 (miR-222). These genes are upregulated in PDAC compared to normal tissue.46, 47 This finding is somewhat surprising since an upregulation of microRNAs is considered to repress transcription or translation of target genes. However, microRNAs can act differently, i.e., repress translation in cycling/proliferating cells but can activate translation in quiescent cells as possibly stem cells.48 Cancer stem cells decisive for tumorigenesis, tumor progression and finally patients' prognosis can arise from stem cells, still express markers of stem cells and have the ability of self-renewal.49 At the molecular level, the alterations of stem cell self-renewal pathways have been recognized as an essential step for cancer stem cell transformation. Recently, we could show that an elevated expression of the stem cell self-renewal gene Hiwi is associated with a poor prognosis in soft tissue sarcomas and in male PDAC patients.50, 51 It would be of interest to study if altered microRNA expression is related to occurrence of cancer stem cells and can help to identify these cells.
In conclusion, we have found that an elevated expression of miR-155, miR-203, miR-210 and miR-222 is correlated with a poor outcome of PDAC patients. Our study further confirms the oncogenic potential of these 4 microRNAs and demonstrates their prognostic value in PDAC.
The authors thank Dr. G. Bond for critical reading of the manuscript and helpful discussions. H.T. was supported by the Deutsche Krebshilfe and T.G. by a graduate research scholarship from the Land Sachsen Anhalt.