Abstract
Background: Pancreatic neuroendocrine neoplasms (pNEN) are rare tumors with a poor prognosis. Although increasing data have accumulated on the molecular pathology of pNEN, very scarce data exist on microRNAs in pNEN and no data are published on microRNAs as potential biomarkers of pNEN in serum. This study aimed to identify microRNA signatures of pNEN in tissue and serum. Materials and Methods: We included tissue samples from 37 patients with pNEN, 9 patients with non-neoplastic pancreatic pathology, seven samples of micro-dissected pancreatic islets and serum samples of 27 patients with pNEN, as well as of 15 healthy volunteers. MicroRNA expression profiles were established using real-time quantitative Polymerase Chain reaction (PCR) for 754 microRNAs. Results: MicroRNA signatures differed between pNEN, pancreatic islets and total pancreas, with virtually no overlap between the groups of de-regulated microRNAs. Expression of miR-642 correlated with Ki67 (MiB1) score and miR-210 correlated with metastatic disease. When comparing microRNA levels in serum from patients with pNEN and healthy volunteers, 13 microRNAs were more abundant in the serum of patients. MiR-193b was also up-regulated in pNEN tissue when compared to pancreatic islets and remained significantly increased in serum even when corrected for multiple testing. Conclusion: Evaluation of microRNAs appears to be promising in the assessment of pNEN. In particular, miR-193b, which is also increased in serum, may be a potential new biomarker of pNEN.
The incidence of pancreatic neuroendocrine neoplasms (pNEN) is low (approximate annual incidence of <1 per 100,000) but has been reported to be rising over the last decades (1). Although most pNEN are low grade lesions referred to as neuroendocrine tumors instead of carcinomas according to the current WHO classification (2), the prognosis is still dismal, with reported 5-year survival rates of no more than 33% in those with non-functional tumors (3). The clinical presentation depends on the functional status of the tumor and is frequently dominated by paraneoplastic hormone secretion (e.g. insulin, glucagon and others). For non-functional tumors, the majority of patients report on only few clinical symptoms, caused e.g. by tumor obstruction. It is, thus, not surprising that metastatic disease is frequently already present at the time of diagnosis (4). To screen for early detection, as well as for monitoring disease activity following initial diagnosis and therapy, a number of tumor biomarkers have been established, such as chromogranin A, 5-hydroxyindolic acid and specific hormone products secreted by the tumor. However, the accuracy of these markers has been shown to be low (5).
Recent work on the molecular basis of pNEN defined alterations in signaling associated with angiogenesis, mammalian Target of Rapamycin (mTOR), and somatostatin receptor signalling (6). These insights have been instrumental for the development of new therapeutic strategies targeting these pathways (7). They have been complemented by a first report on the establishment of a gene expression-based classifier generated from peripheral blood samples from patients with gastroenteropancreatic neuroendocrine tumors (8).
MicroRNAs are small non-coding RNAs with the capability to regulate gene expression on the post-transcriptional level by binding to mRNA molecules. Binding of a microRNA to the seed sequence of the target mRNA can halt translation or degradation of the mRNA molecule. A single microRNA may target hundreds of mRNAs and a single mRNA may comprise of seed sequences for several microRNAs. By inhibiting the expression of oncogenes and tumor-suppressor genes, microRNAs can act as oncomiRs and tumor-suppressor miRs, respectively (9, 10). MicroRNA signatures have been discussed as valuable biomarkers in tumour typing and individual risk stratification. From the analytical point of view, microRNAs are stable in serum and plasma samples (11, 12). Moreover, microRNAs can be therapeutically targeted. Only few data have been published on microRNA signatures of gastroenteropancreatic neuroendocrine neoplasms limited to studies from Ruebel et al. (13), Roldo et al. (14) and Li et al. (15). Herein, we investigated the global microRNA expression of pancreatic neuroendocrine neoplasms in tumor specimen and serum in comparison to normal pancreatic tissue, isolated pancreatic islets and serum from healthy volunteers.
Materials and Methods
Patients and samples. The local Ethics Committee approved use of tissue and serum for this study (no. 13-093). Tissue samples were retrieved from the archive of the pathology department. All tumor cases (n=37) were evaluated by two pathologists (CK and CT) and grading and staging was adapted according to the WHO 2010 (2) and UICC TNM (16) classification. Data on gender, age, mitotic count, Ki67 index and TNM are summarized in Table I. Tumorous areas were dissected to achieve a tumour cell content >90%. Samples from non-neoplastic pancreas (n=9) were histologically examined before and after cutting sections for microRNA extraction to exclude significant inflammatory infiltration. These tissue samples contained predominantly exocrine parenchyma and were not enriched for endocrine cells. Pancreatic islets samples (n=7) were microdissected using PALM Micro Beam C microdissection system (P.A.L.M. Microlaser Technologie, Bernried, Germany). Serum samples of 15 healthy volunteers and 27 patients with pNEN were included in the study. Samples of 10 of the patients were collected before surgical or medical therapy was initiated.
Nucleic acid extraction. RNA from tissue samples, as well as from the pancreatic islets samples, was extracted using RecoverAll kit (Ambion, Austin, TX, USA) according to the manufacturer's recommendations. Serum samples were mixed with three volumes of QIAzol (Qiagen, Hilden, Germany) and subsequently RNA was isolated using the miRNeasy Mini Kit, as defined by the manufacturer (Qiagen).
Quantitative real-time polymerase chain reaction (qRT-PCR). MicroRNAs from TaqMan® Array MicroRNA A and B Cards were examined for all samples applying the Megaplex Pool protocol for microRNA expression analysis (Applied Biosystems, Foster City, CA, USA). Briefly, each reverse transcription reaction contained 350 ng RNA or 3 μl of total RNA from serum. In the case of the serum samples and microdissected islets, a pre-amplification reaction was performed after cDNA synthesis. TaqMan Universal PCR Master Mix was used for quantitative PCR on the TaqMan 7900HT Fast Real-Time PCR System (Applied Biosystems).
Data transformation and normalization. Data were available for 53 samples derived from three different pancreatic tissues, and for 42 samples derived from serum. Out of these samples, data of six tumour cases were available from both pNEN and serum. For each sample, the intensity of 754 human microRNAs was measured in terms of Ct values. The U6snRNA was assessed four times in parallel and used as endogenous control for each sample. The microRNA intensity values were obtained as a relative copy number using the Ct values of the measured microRNAs and Ct values of the U6snRNA for the tissue samples and plasma samples separately: In a first step, the derived Ct values were log2-transformed. Secondly, the mean of the transformed U6snRNA Ct values was subtracted from the transformed Ct values of the measured microRNAs per sample. Since the samples derived from the islets of Langerhans were pre-amplified, a median normalization was applied to the data to obtain comparable microRNA intensity levels across the different tissues. In a last step, normalized microRNA intensity values were multiplied by −1; thereby higher intensity values represent higher expression levels of the respective microRNAs.
Statistical analysis. For association analyses, we considered only those microRNAs that had at least 75% non-missing values in the respective groups, i.e. in 28 out of 37 patients with pNEN, six out of seven pancreatic islets samples, and seven out of nine samples from non-neoplastic pancreas (tissue), as well as in 21 out of 27 patients with pNEN, and 12 out of 15 samples of healthy volunteers (serum). Finally, the number of ‘analysis-ready’ microRNAs was: A) N=187 tissue samples from patients with pNEN vs. samples from non-neoplastic pancreas; B) N=109 patients with pNEN vs. pancreatic islets samples; and C) N=157 serum samples from patients with pNEN vs. healthy volunteers. For association analysis, linear regression was applied with microRNA intensity as dependent variable and group status as independent variable. Association results with a Bonferroni corrected p-value of less than 0.05 (corresponding to p<2.7×10−4, p<4.6×10−4, and p<3.1×10−4 for analyses A to C, respectively) were considered as statistical significant. Correlation of microRNA profiles between the tissue and serum samples was assessed by Pearson's correlation coefficient.
Results
Normal pancreatic parenchyma, pancreatic islets and pNEN differed with regard to their microRNA profiles. When we compared pNEN to exocrine pancreas, 16 microRNAs were found down-regulated and 6 up-regulated. In contrast, when pNEN samples were compared to pancreatic islets, 9 microRNAs were down-regulated and 18 up-regulated after correcting for multiple testing (Figure 1). There was virtually no overlap between the two groups of microRNAs. Only miR-19b, miR-146b and miR-720 were differentially expressed in pNEN compared to normal pancreatic parenchyma and to pancreatic islets; however, they differed when comparing the tissue origin, as the first two were overexpressed in pNEN when compared to islets but down-regulated when compared to normal pancreatic tissue and vice versa for miR-720. When correlating microRNA expression to the pT stage, metastatic spread, Ki67 (MiB1) score and mitotic count, only miR-642 correlated positively to the MiB1 score (p=4.0×10−6) and miR-210 to metastatic disease (p=7.4×10−5). MicroRNA expression levels did, however, not correlate to pT stage or mitotic count. Nor was there a significant correlation found between microRNA expression levels and age or gender. MicroRNA signatures in serum were very similar in patients with pNEN and healthy volunteers. However, miR-193b was significantly more abundant in serum of patients (p=2.02×10−5). We aimed to reduce heterogeneity by restricting our analysis to the 10 serum samples which were collected before any therapeutic intervention. The association p-value was still nominally significant for miR-193b (p=2.86×10−3) (Figure 2A). However, the effect sizes were almost unchanged and the p-value did not reach significance after Bonferroni correction, which was most probably a result of a reduced statistical power due to the substantially smaller sample size. Interestingly, miR-193b was also overexpressed in pNEN tissue compared to islets cells (p=0.039, Figure 2B), but for the 6 patients with both tissue and serum samples, there was much lower correlation between expression levels in serum and in tissue [mean correlation coefficient (r) of the six samples: r=0.528] compared to the mean correlation coefficient within the same tissue (r=0.851 and r=0.907 for pancreas and serum, respectively).
Summary of clinical and pathological data of patients with pancreatic neuroendocrine neoplasia.
Differentially expressed microRNAs between pancreatic neuroendocrine neoplasia (P), pancreatic tissue (C) and isolated islets (I). Red: expression levels greater than the median value; blue: expression levels less than the median value.
Discussion
Rather specific microRNA signatures characterize cells and tissues. Therefore, it is crucial to carefully select the non-neoplastic tissue for comparison to tumor data. As pNEN cells are transformed from neuroendocrine cells, we would speculate that the appropriate non-neoplastic tissue should not be based on total pancreatic tissue, which includes endocrine and exocrine tissue, but rather on endocrine cells derived from pancreatic islets. Thus, differentially expressed microRNAs in pNEN and normal pancreatic tissue may rather reflect the difference between endocrine and exocrine pancreatic cells than the differences between neoplastic and non-neoplastic endocrine cells.
Among the microRNAs that were up-regulated in pNEN in comparison to islet cells, miR-140 is of special interest, since its elevation has been described in serum of patients with various cancer types (17). Interestingly, it has been shown to act as a potential oncomiR in gliomas (18). On the other hand, increasing evidence is emerging that miR-140 may act as a tumor suppressor by targeting insulin-like growth factor receptor-1 in non-small cell lung cancer (19), transforming growth factor β receptor-1 and fibroblast growth factor-9 in hepatocellular carcinoma (20), and sex determining region Y box 2 (SOX2) in breast cancer (21). These diverse roles of miR-140 in cancer suppression or promotion are mirrored by comparable data on miR-150, which is described as tumor suppressor in T-cell/natural killer cell lymphomas (22) but acts as a potential oncomiR in extranodal marginal zone B-cell lymphomas (23, 24), indicating that the specific functions of microRNAs as tumour suppressor or oncomiRs also depends on the tissue of origin.
miR-193b levels (-delta log CT values) in serum of patients with pNEN and healthy volunteers (A) and in tissue samples of pNEN, pancreatic control tissue and islets (B). The bottom and top of the boxes are the lower and upper quartiles, respectively. The horizontal bar within the box is the median. The lower and upper ends of the whiskers represent the smallest and highest value within the 1.5 interquartile range of the lower and upper quartile, respectively. With respect to this convention, circles display outliers.
MiR-210 is associated with metastatic stage. Differential expression of miR-210 has been described for liver metastasis in colorectal cancer (25) and is known to be induced by hypoxia-inducible factor 1-alpha. The potential role of miR-210 in adaption to hypoxia-induced stress (26) might be of importance for the metastatic potential of pNEN. Obviously it is worthwhile to test for the prognostic value of miR-210 together with miR-642 in pNEN. However, follow-up and events were not sufficient for this analysis in our dataset.
Very few microRNAs were different in serum between patients with pNEN and healthy volunteers. Surprisingly, these were not the most de-regulated microRNAs in pNEN tissue samples, and the abundance of microRNAs in serum is not correlated to their level in tumor tissue. However, miR-193b was also more highly expressed in pNEN tissue than in islet cells (p=0.039). The mechanism of microRNA access to the blood is unclear. Several options have been discussed. They may be released to blood during tumor necrosis or apoptosis. However, necrosis is uncommon in G1 and G2 neuroendocrine tumours and this is most likely not an important source of circulating microRNAs in pNEN. MicroRNAs may also be secreted in an ATP-dependent manner as exosome-enclosed or argonaute-2/nucleophosmin1 (AGO2/NPM1)-associated microRNAs (12). Only in the case of necrosis or apoptosis would a correlation between circulating microRNA levels in blood and tissue levels be expected, but not necessarily in the case of active secretion, particularly as the stimuli causing microRNA secretion in pNEN are unknown. Furthermore, it can be speculated that the tumor is not the source of circulating miR-193b at all, but may directly or indirectly induce the secretion of miR-193b from other cells. Currently, we do not have convincing data to clarify the source of circulating miR-193b. MiR-193b has been ascribed a number of effects important for tumor biology. Walter and colleagues showed that miR-193b is overexpressed in high-grade prostatic carcinomas and in prostate cancer with perineural and lymphatic invasion and extracapsular extension (27). It further appears to modulate the TNF-related apoptosis-inducing ligand (TRAIL)-induced apoptosis pathway (28). Via cyclin-dependent kinase-2 regulation through direct targeting of inhibitor of growth family member 5, microRNA-193b may promote mesenchymal stem cell proliferation (29), but has also been described as silencing embryonic stem cell renewal (30).
In conclusion, we have shown that pNEN are characterized by a particular microRNA signature and that microRNA expression levels in the tumor correlate to those in metastases (miR-210) and to MiB1 score (miR-642). Therefore, miR-210 and miR-642 should be tested as prospective outcome parameters. Circulating miR-193b was found at higher levels in serum of patients with pNEN and may represent a very promising new biomarker for detecting pNEN but this certainly needs to be confirmed in larger series.
Acknowledgements
We thank Anette Aufseß for excellent technical assistance and Katrin Kalies and Jeanette Erdmann for granting access to the PALM and TaqMan, respectively. The study was supported by an unrestricted educational grant from IPSEN Pharma GmbH (Ettlingen), Germany.
Footnotes
-
Conflicts of Interests
CT and GB received research funding from IPSEN Pharma GmbH (Ettlingen), Germany and Novartis Pharma GmbH, Nürnberg, Germany; CS, NG, HW, CK, VB, ACF, TK, JKH, NB, and HL have nothing to disclose.
- Received January 31, 2014.
- Revision received March 11, 2014.
- Accepted March 13, 2014.
- Copyright© 2014 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
