MicroRNA Expression Abnormalities in PancreaticEndocrine and Acinar Tumors Are Associated WithDistinctive Pathologic Features and Clinical BehaviorClaudia Roldo, Edoardo Missiaglia, John P. Hagan, Massimo Falconi, Paola Capelli, Samantha Bersani,George Adrian Calin, Stefano Volinia, Chang-Gong Liu, Aldo Scarpa, and Carlo M. Croce
A B S T R A C T
PurposeWe investigated the global microRNA expression patterns in normal pancreas, pancreatic endocrinetumors and acinar carcinomas to evaluate their involvement in transformation and malignant progres-sion of these tumor types. MicroRNAs are small noncoding RNAs that regulate gene expression bytargeting specific mRNAs for degradation or translation inhibition. Recent evidence indicates thatmicroRNAs can contribute to tumor development and progression and may have diagnostic andprognostic value in several human malignancies.
Materials and MethodsUsing a custom microarray, we studied the global microRNA expression in 12 nontumor pancreasand 44 pancreatic primary tumors, including 12 insulinomas, 28 nonfunctioning endocrine tumors,and four acinar carcinomas.
ResultsOur data showed that a common pattern of microRNA expression distinguishes any tumor typefrom normal pancreas, suggesting that this set of microRNAs might be involved in pancreatictumorigenesis; the expression of miR-103 and miR-107, associated with lack of expression ofmiR-155, discriminates tumors from normal; a set of 10 microRNAs distinguishes endocrine fromacinar tumors and is possibly associated with either normal endocrine differentiation or endocrinetumorigenesis; miR-204 is primarily expressed in insulinomas and correlates with immunohisto-chemical expression of insulin; and the overexpression of miR-21 is strongly associated with botha high Ki67 proliferation index and presence of liver metastasis.
ConclusionThese results suggest that alteration in microRNA expression is related to endocrine and acinarneoplastic transformation and progression of malignancy, and might prove useful in distinguishingtumors with different clinical behavior.
Pancreatic endocrine tumors (PETs) may occursporadically or as part of multiple endocrine neopla-sia type 1 syndrome.1 These neoplasms are clinicallyclassified as functioning (F-PET) or nonfunctioning(NF-PET), according to the presence of symptomsdue to hormone hypersecretion. F-PETs are mainlyrepresented by insulinomas. At diagnosis, metastaticdisease is observed in only 10% of insulinomas butin up to 60% of NF-PETs, and most PET-relateddeaths are caused by liver metastasis.1 The malig-nant potential among PETs varies greatly and can-not be predicted on the basis of histologicappearance. In fact, the vast majority of PETs arewell-differentiated endocrine tumors (WDETs) and
are defined as carcinoma (WDEC) only when inva-sion or metastases are identified.1
Pancreatic acinar cell carcinoma (PACC) is anextremely rare tumor type distinct from ductal ade-nocarcinoma and PET, although some overlap withPET is testified by both the expression of neuroen-docrine markers in one third of the cases and theexistence of mixed acinar-endocrine carcinomas.2
PACC is always malignant with a median survival of18 months, which lies between that of pancreaticductal adenocarcinoma and endocrine neoplasms(6 months and 40 months, respectively3).
Little is known about the molecular pathogenesisof PETs.4 Inactivation of multiple endocrine neoplasiatype 1 gene is the most frequent genetic event iden-tified in sporadic PET,5 whereas mutations in genes
From the Departments of MolecularVirology, Immunology and MedicalGenetics and Comprehensive CancerCenter, Ohio State University, Colum-bus, OH; Departments of Pathologyand Surgical and GastroenterologicalSciences, Universita di Verona, Verona;and the Department of Morphology andEmbryology, University of Ferrara,Ferrara, Italy.
Submitted December 28, 2005; acceptedJune 29, 2006; published online ahead ofprint at www.jco.org on September 11,2006.
Supported by Program Project Grants No.P01CA76259 and P01CA81534 from theNational Cancer Institute (C.M.C.), by aKimmel Scholar award (G.A.C.), Associa-zione Italiana Ricerca Cancro (AIRC; A.S.),Milan, Italy; Fondazione Cassa di Rispar-mio di Verona (Bando 2005), Italy; Minis-teri Universita e Salute, Rome, Italy;European Community Grant No.PL018771; and Fondazione GiorgioZanotto, Verona, Italy.
C.R. and E.M. contributed equally tothis work. A.S. and C.M.C. contributedequally to this work.
Terms in blue are defined in the glossary,found at the end of this article and onlineat www.jco.org.
Authors’ disclosures of potential con-flicts of interest and author contribu-tions are found at the end of thisarticle.
Address reprint requests in the UnitedStates to Carlo M. Croce, MD, OhioState University, Comprehensive CancerCenter, Wiseman Hall Room 385K, 410W 12th Ave, Columbus, OH; e-mail:[email protected]; and in Europeto Aldo Scarpa, MD, Verona University,strada Le grazie 8, 37134 Verona, Italy;e-mail: [email protected].
typically involved in pancreatic adenocarcinoma are uncommon.5
Even less is known regarding the molecular anomalies of PACC.6 Nogene expression profile data are available for PACC, and our under-standing of gene expression changes that occur in PET is still at aninitial phase.7
MicroRNAs are small (20 to 24 nucleotides) noncoding RNAgene products that serve critical roles in cell proliferation, apoptosisand developmental timing by negatively regulating the stability or trans-lational efficiency of their target mRNAs.8 Currently, 462 unique maturehuman microRNAs are known (http://microrna.sanger.ac.uk). Aberrantexpression of microRNAs has been linked to cancers,9,10 and diagnostic/prognostic characteristics of specific cancer types can be distinguishedbased on their microRNA profiles.11-19 Functional studies also havelinked aberrant microRNA expression to carcinogenesis.20-23
We investigated the global microRNA expression patterns innormal pancreas, primary pancreatic endocrine tumors and acinarcarcinomas to evaluate their involvement in transformation and ma-lignant progression of these tumor types.
MATERIALS AND METHODS
Patient Data, Neoplastic-Cell Enrichment, and
RNA Extraction
All samples were from frozen primary tumors collected at the Depart-ment of Pathology, Verona University, Verona, Italy. All tumors were spo-radic, as assessed by personal and family histories. PET were diagnosed andclassified according to WHO criteria.1 They included 28 nonfunctional and 12functional tumors. The 28 NF-PET included 11 WDETs and 17 WDECs. The12 F-PET were insulinomas, comprising 11 WDETs and one WDEC. Diagno-sis of PACC was confirmed by immunohistochemical expression of lipase,amylase, and trypsin. As a control, normal pancreas was taken in 12 corre-sponding patient specimens.
A neoplastic cellularity higher than 90% was obtained by cryostat enrich-ment. RNA was extracted with Trizol (Invitrogen, Carlsbad, CA) from 1020-�m thick cryostat sections, checking the cell composition of the sampleevery five sections. RNA integrity was confirmed using the Agilent 2100 Bio-analyzer (Agilent Technologies, Palo Alto, CA).
MicroRNA Microarray Hybridization and Quantification
MicroRNA labeling and hybridization were performed as described pre-viously,24 using 5 �g of total RNA. Our microRNA microarray (Ohio StateUniversity Comprehensive Cancer Center, version 2.0) contains probes for460 mature microRNAs spotted in quadruplicate (235 homo sapiens, 222 musmusculus, and three Arabidopsis thaliana) with annotated active sites. Often,more than one probe set exists for a given mature microRNA. Additionally,there are quadruplicate probes corresponding to most precursor microRNA.Hybridization signals were detected with Streptavidin-Alexa647 conjugate andscanned images (Axon 4000B) were quantified using the Genepix 6.0 software(Axon Instruments, Sunnyvale, CA).
Computational Analysis of microRNA Microarray Data
Most of the analysis and graphics were generated using R software ver-sion 2.0.1 and Bioconductor version 1.625 packages. Sequentially, the blankand probe controls spots were removed from the data set of the 56 microRNAmicroarrays, and the local background was then subtracted from the mediansignal. Next, the data were normalized using a variance-stabilizing transforma-tion stratified, within each array, by grid using the vsn package.26 The dataresulting from this transformation can be considered as equivalent to logarith-mic transformed data.26 Subsequently, the genefilter package was used toremove all the spots whose intensities were lower than 99th percentile of theintensities of blank spots in all the arrays. To select the differentially expressedgenes between pairs of relevant sample categories, the data obtained were thenanalyzed by direct two-class unpaired comparison using the samr package. To
increase stringency, we considered only the microRNA probes that had at leastthree significantly differentially expressed spots out of the four replicas con-tained in the array. The fold change reported in the tables is the median valueof these spot replicas.
Hierarchical cluster analysis was performed using the aggregate val-ues of replicate spots obtained applying Tukey’s median polish algorithm.The analysis was done using the 200 probes containing the mature mi-croRNA sequences with the highest interquartile range. The distance metricsused to cluster samples and genes were Pearson correlation and Euclideandistance, respectively. The agglomerative method was the complete-linkage.The output was visualized using Maple Tree (version 0.2.3.2, http://mapletree.sourceforge.net/). All data were submitted using MIAMExpress to the ArrayExpress database (http://www.ebi.ac.uk/arrayexpress/).
Northern Blotting
Five micrograms of total RNAs prepared from samples derived from thesame patients but different from those used in the microarray experiment wererun on 15% Criterion polyacrylamide gel electrophoresis/Urea gels (Bio-Rad,Hercules, CA), transferred onto Hybond-N� membranes (Amersham,Piscataway, NJ) and hybridized overnight with 32P end-labeled probes at37°C in ULTRAhyb-Oligo hybridization buffer (Ambion, Austin, TX). Theprobes were antisense oligonucleotides relative to the mature microRNAsand to 5S RNA as normalizer. Membranes were washed at 37°C twice for 30minutes each with 2XSSC/0.5% sodium dodecyl sulfate (SDS), analyzedusing a Typhoon 9410 phoshorimager (Amersham) and quantified usingImageQuant-TL (Amersham). Blots were stripped by boiling in 0.1%aqueous SDS for 5 minutes and were reprobed several times.
RESULTS
MicroRNA expression profiles were determined for 12 nontumorpancreas and 44 primary pancreatic tumors, including 40 PETs andfour PACCs, using a custom microarray platform that was provento give robust results, as validated by several studies.12,17,19,24 Theunsupervised hierarchical clustering, using the 200 most variablemicroRNAs, showed a common microRNA expression patterndistinguishing PET and PACC from normal pancreas (Fig 1). No-tably, PACCs fell into a unique cluster that was part of the wider clusterincluding all PETs, whereas there was no distinctive pattern betweeninsulinomas and NF-PET.
We then searched for microRNAs showing differential expres-sion between classes of samples. Class comparison analysis showed thedifferential expression of several microRNAs between normal tissueand either PACC or PET, while a smaller number of microRNAs werefound to be differentially expressed between PET and PACC (Fig 2). Indetail, class comparison analysis identified 87 up- and eight down-regulated microRNAs in PET versus normal pancreas, while PACChad 30 microRNAs upregulated and seven downregulated when com-pared with the healthy tissue. Only 10 microRNAs were differentiallyexpressed between PET and PACC, and four were unique to WDECwith respect to PACC.
Common microRNA Expression Pattern Distinguishes
Pancreatic Endocrine and Acinar Tumors From
Normal Pancreas
The vast majority of the differentially expressed microRNAsfound in PACC versus healthy tissue were also found in PET versushealthy tissue. In fact, 28 of the 30 (93%) microRNAs overexpressed inPACC were also found upregulated in PET. Similarly, five (71%)of seven underexpressed microRNAs were downregulated in bothtumor types (Fig 2). This overlap, together with the fact that only a
limited set of microRNAs were differentially expressed between PETand PACC or among PET subtypes, is suggestive of a pattern ofmicroRNA expression common to acinar and insular derived tumors.
Among the upregulated microRNAs in PET that are alsocommon to PACC, seven were validated by Northern analysis.MiR-103 was the best discriminator for all pairwise comparisons ofnormal pancreas, acinar carcinomas, and endocrine tumors (Fig3). The expression of miR-107 paralleled that of its highly homologousmiR-103, and the significant overexpression in tumors versus normalof miR-23a, miR-26b, miR-192, and miR-342 was also confirmed(online only Fig 1).
Among the downregulated microRNAs in PET, Northern blotof miR-155 showed the lack of detectable expression in both PETand PACC (Fig 3). Although miR-155 was not among the top listeddownregulated genes in PACC, its low expression in this tumor typewas also detected by microarray, as shown in the box-and-whiskersplot of Figure 2.
Limited Set of microRNA Distinguishes Pancreatic
Endocrine From Acinar Tumors
The direct comparison of PET and PACC showed only 10 up-regulated microRNAs, all of which were also overexpressed in PET
Fig 1. Unsupervised hierarchical clustering of 12 normal pancreas and 44 pancreatic tumors, including 22 well-differentiated pancreatic endocrine tumors (WDETs), 18well-differentiated pancreatic endocrine carcinomas (WDECs) and four pancreatic acinar cell carcinomas (PACCs). WDET included 11 insulinomas (INS) and 11 NF-PET; WDECincluded 1 INS and 17 NF-PET. (A, B) Some microRNAs upregulated in PET versus normal tissue (in red); (C) two microRNAs upregulated in INS versus NF-PET (in blue); (D)some microRNAs downregulated (in green).
MicroRNA in Pancreatic Endocrine and Acinar Tumors
www.jco.org 4679Downloaded from jco.ascopubs.org on August 16, 2016. For personal use only. No other uses without permission.
versus normal tissue. In contrast, no microRNA was found to bespecifically up- or downregulated in PACC.
Overexpression of miR-204 Is Specific to Insulinomas
and Correlates With Immunohistochemical Expression
of Insulin
The comparison of insulinomas with NF-PET identified onlythree microRNAs that were significantly overexpressed in insulino-mas, including miR-204, its homologous miR-211, and miR-203. No-tably, the immunohistochemical expression of insulin correlated withmiR-204 expression more strongly than with insulin mRNA expres-sion (Fig 4). In fact, logistic regression analysis, based on negative orpositive immunohistochemical staining, showed that the protein ex-pression was predicted by both insulin mRNA and miR-204 expres-sion (P � .001); however, in a multivariate model only miR-204expression retained statistical significance (P � .001).
Because miR-375 was suggested to be specifically expressed inmouse pancreatic islets and to function as a negative regulator ofinsulin exocytosis,27 we investigated its expression in a panel ofnontumor human adult tissues and our samples by Northern blot.MiR-375 was detected in only healthy pancreas (online only Fig 2),and its expression level was higher in tumors versus nontumorpancreas, but showed no difference between insulinomas and non-functioning tumors.
Expression of miR-21 Is Strongly Associated With the
Proliferation Index and Presence of Liver Metastasis
The evaluation of expression profiles to identify microRNAsdiscriminating PETs based on either metastatic status or proliferationindex identified only miR-21 as significant (Fig 5). This is not surpris-ing, because these two tumor characteristics are interconnected. Infact, all metastatic PETs had a proliferation index more than 2%,
Fig 2. Venn diagram illustrating the rela-tionships between sets of microRNAsfound differentially expressed using directtwo class unpaired comparisons. Circlesinclude the total number of differentiallyexpressed microRNAs in the pairwisecomparison indicated. The intersection ar-eas identify the number of differentially ex-pressed microRNAs in common betweeneach comparison. Normal, nontumor; PET,pancreatic endocrine tumor; PACC, pancre-atic acinar cell carcinoma.
Fig 3. The overexpression of miR-103 andlack of expression of miR-155 is peculiar topancreatic insular and acinar tumors. Box-and-whiskers plots showing the expressionlevels of (A) miR-103 and (B) miR-155 mea-sured by microarray analysis in 12 nontumorpancreas (normal) and 44 pancreatic tumors,including 22 well-differentiated pancreaticendocrine tumors (WDETs), 18 well-differentiated pancreatic endocrine carcino-mas (WDECs) and four pancreatic acinar cellcarcinomas (ACCs). Median intensity is high-lighted by bold lines. (C) Northern blot anal-ysis parallels the microarray expression data.
whereas no tumor with a lower proliferation score was metastatic.Furthermore, miR-21 also distinguished between NF-PETs orWDECs with high (Ki67 � 2%) and low (Ki67 � 2%) proliferationindex. Another interesting observation is that miR-21 was also over-expressed in PACCs versus nontumor pancreas.
Identification of Putative mRNA Targets for
Differentially Expressed microRNAs
Three programs (miRanda, TargetScan, PicTar, respectivelyavailableathttp://www.microrna.org/mammalian/index.html,http://genes.mit.edu/targetscan/, http://pictar.bio.nyu.edu/) were used toidentify predicted targets of selected microRNAs, namely miR-103/miR-107, miR-155, miR-204/miR-211, and miR-21. To increase thestringency of the analysis, we considered only target genes that werefound from all three algorithms. Because the same tumor samples andfive nontumor pancreas analyzed for microRNA expression have alsobeen evaluated for gene expression profiles with a custom oligonucle-otide microarray (Missiaglia et al, manuscript in preparation), weassessed the status of predicted mRNA targets in PET and normaltissue as well as among PET with different clinicopathologic charac-teristics. A two-sample t-test analysis identified several putative targetgenes that were either up- or downregulated, namely 28 up- and sevendownregulated for miR-103/107, two up- and two downregulated foreither miR-155 or miR-204/211, and one up- and one downregulatedfor miR-21. Notably, the mRNA expression of PDCD4 gene, a putativetarget of miR-21, was downregulated in liver metastatic PET and in
tumors with high proliferation index, showing an inverse correlationwith the expression of miR-21 (Fig 6).
DISCUSSION
The results of our survey of microRNA expression profiles innormal pancreas, pancreatic endocrine tumors and acinar carcino-mas may be summarized as follows: a common microRNA expres-sion profile distinguishes both endocrine and acinar tumors fromnormal pancreas; the expression of miR-103 and miR-107 associ-ated with lack of expression of miR-155 discriminates tumors fromnormal; a limited set of microRNAs is specific to endocrine tumorsand is possibly associated with the endocrine differentiation or tumor-igenesis; miR-204 expression occurs primarily in insulinomas andcorrelates with immunohistochemical expression of insulin; and ex-pression of miR-21 is strongly associated with proliferation index andliver metastasis.
Unsupervised hierarchical clustering of the expression profilesshowed that both tumor types were separated from normal pancreas.Although PACCs fell into a unique cluster, this was part of the widercluster including all PETs. Although we identified many more differ-entially expressed microRNAs in PET versus healthy than betweenPACC versus healthy, the vast majority of differentially expressedmicroRNAs in PACC were similarly altered in PET. It is worth noting
Fig 4. miR-204 overexpression is specificto insulinomas and correlates with immu-nohistochemical expression of insulin. (A)Box-and-whiskers plot showing the ex-pression level of miR-204 measured bymicroarray analysis in 12 nontumor pan-creas (normal), 12 insulinomas (INSs), 28nonfunctioning pancreatic endocrine tu-mors (NF-PETs) and four pancreatic acinarcell carcinomas (ACCs). Median intensity ishighlighted by bold lines. (B) Strong corre-lation observed between miR-204 expres-sion and insulin staining assessed byimmunohistochemistry. Significant correla-tion was found also between miR-204 ex-pression and insulin mRNA expression(Pearson correlation � 0.42; 95% CI, 0.12to 0.65; P � .008). (C, D) Northern blotanalysis confirms the microarray expres-sion data.
MicroRNA in Pancreatic Endocrine and Acinar Tumors
www.jco.org 4681Downloaded from jco.ascopubs.org on August 16, 2016. For personal use only. No other uses without permission.
that healthy pancreas is largely formed by acini and therefore repre-sents the ideal normal counterpart for the analysis of acinar cellcarcinomas, whereas pancreatic islet cells would represent thehealthy counterpart for pancreatic endocrine tumors. Unfortu-nately, we had no preparations of these cells available. Nonetheless,the finding of a largely concordant pattern of differentially ex-pressed microRNAs between acinar and insular tumors, including28 upregulated and five downregulated genes, suggests that this setcommon to both tumor types might be related to pancreatic neo-plastic transformation. Providing additional support for this asser-tion, several microRNAs differentially expressed in both our tumortypes have been found differentially expressed also in breast,17
colon11 and B-cell leukemia.12 In addition, at least 20 of the differ-entially expressed microRNAs in our tumors have been identifiedas having either growth related or apoptotic effects in the lungA549 or cervical HeLa carcinoma cell lines.28
Furthermore, both PACC and PET showed the coordinate over-expression of miR-17, miR-20 and miR-92-1, which are contained in apolycistronic cluster. This miR-17-92 cluster has been described to actas an oncogene in association with MYC gene.22 Notably, overexpres-sion of MYC has been reported in PET and also in hyperplastic islets,suggesting its involvement in the early phases of insular tumorigene-sis.29 In addition, induction of MYC in islet or acinar cells of mouse invitro or in vivo models produces endocrine tumors30,31 or mixedacinar/ductal adenocarcinomas,32 respectively, while suppression ofMYC-induced apoptosis leads to islet cell carcinoma.33
The expression of the two highly homologous miR-103 and miR-107 together with the lack of expression of miR-155 was distinctive oftumors versus nontumor tissue. Interestingly, miR-103/107 have beenfound to be overexpressed in several tumor types.34 The finding thatmiR-155 was expressed in healthy pancreas and lacking in both PETand PACC is rather interesting as overexpression of miR-155 has beenobserved in lymphomas16,35,36 and breast cancer,17 finding that haveled to the speculation that miR-155 may be an oncogenic mi-croRNA.10 This may not be unexpected, as microRNAs expressed inadults are tissue specific37 and the consequences of microRNA misex-pression is highly dependent on the cell-specific expression pattern ofmRNAs that are microRNA regulated.28
Ten microRNAs were peculiarly overexpressed in PET and dif-ferentiated this tumor from both PACC and normal pancreas. Theseincluded miR-99a, 99b, 100, 125a, 125b-1, 125b-2, 129-2, 130a, 132,and 342. These microRNAs may be characteristic of either normalpancreatic endocrine differentiation or endocrine tumorigenesis.Conversely, no microRNA was found to be specifically up- or down-regulated in PACC, though the limited number of PACC samples mayhave affected the power of the analysis.
Although the microRNA profiles were almost indistinguishablebetween insulinomas and nonfunctioning endocrine tumors, theoverexpression of the two closely related miR-204 and miR-211 wasrestricted to insulinomas. Notably, miR-204 expression correlatedwith the immunohistochemical expression of insulin. In this respect,miR-375 has been recently reported to be specifically expressed in
Fig 5. Expression of miR-21 is stronglyassociated with presence of liver metastasisand tumoral proliferation index. The box-and-whiskers plots show the different expres-sion level of miR-21, measured bymicroarray analysis, (A) between pancreaticendocrine tumors with (Meta�) or without(Meta�) liver metastasis, and (B) betweentumors with a proliferation index morethan 2% (high) or � 2 (low), as measured byKi67 immunohistochemistry. (C) Northernblot analysis confirms microarray expressiondata. WDEC, well-differentiated pancreaticendocrine carcinoma; WDET, well-differentiated pancreatic endocrine tumor.
mouse pancreatic islets and to function as a negative regulator ofinsulin exocytosis.27 Our data showed that this microRNA is expressedin human normal pancreas as well as in acinar and endocrine tumors.However, no difference was found in its expression level betweeninsulinomas and nonfunctioning endocrine tumors.
We also determined whether microRNA expression was corre-lated with the clinical characteristics of PETs. Our results showed thatmiR-21 overexpression is associated with both enhanced Ki67 prolif-eration index and liver metastasis. MiR-21 overexpression has been
observed in several cancers, including glioblastoma, breast, lung andcolon cancers.9,10,15,17 A cancer-related function of miR-21 is alsosupported by knockdown experiments in glioblastoma cells showingthat this microRNA has an antiapoptotic function.20 In this respect,the programmed cell death 4 (PDCD4) gene, putatively targeted bymiR-21, was found significantly downregulated in our metastatic andhigh proliferative PET, and showed an inverse correlation with theexpression of miR-21. This gene has been reported to act as a tumorsuppressor implied in cellular invasion and metastasis.38 Further-more, PDCD4 expression is lost in progressed carcinomas of lung,breast, colon, and prostate,39 and a tumor suppressor role for PDCD4has been also reported in neuroendocrine tumor cells.40
MicroRNAs exert their biologic effects by targeting specificmRNAs for degradation or translational inhibition. To get insightsinto the biologic implications of the most interesting microRNAsshowing altered expression in pancreatic tumors (ie, miR-103/miR-107, miR-155, miR-204/miR-211, and miR-21), we searched predictedtargets that were in common among those identified by three differentalgorithms, discussed in Results. Then, to evaluate whether there was acorrelation between the expression of microRNAs and that of theirpredicted targets, we took advantage of the microarry expression pro-files of the same tumor and normal samples (Missiaglia et al, manu-script in preparation). Among the selected targets that were containedin our microarray, we found several up- and downregulated genes.Interestingly, the predicted target genes of miR-103/107 were overex-pressed more frequently than expected. This finding parallels that ofBabak et al, 37 who reported a low correlation between microRNAexpression and their predicted mRNA targets in a set of 17 differentmouse tissues. This supports the currently favored model that mostmicroRNAs act more likely through translational inhibition withoutmRNA degradation.41
In conclusion, our study suggests that alteration in microRNAexpression is related to endocrine and acinar neoplastic transforma-tion and progression of malignancy.
REFERENCES
1. Kloppel G, Perren A, Heitz PU: The gastroen-teropancreatic neuroendocrine cell system and itstumors: The WHO classification. Ann N Y Acad Sci1014:13-27, 2004
2. Ohike N, Kosmahl M, Kloppel G: Mixed acinar-endocrine carcinoma of the pancreas: A clinicopath-ological study and comparison with acinar-cellcarcinoma. Virchows Arch 445:231-235, 2004
3. Holen KD, Klimstra DS, Hummer A, et al:Clinical characteristics and outcomes from an institu-tional series of acinar cell carcinoma of the pancreasand related tumors. J Clin Oncol 20:4673-4678, 2002
4. Zikusoka MN, Kidd M, Eick G, et al: Themolecular genetics of gastroenteropancreatic neu-roendocrine tumors. Cancer 104:2292-2309, 2005
5. Perren A, Komminoth P, Heitz PU: Moleculargenetics of gastroenteropancreatic endocrine tu-mors. Ann N Y Acad Sci 1014:199-208, 2004
6. Abraham SC, Wu TT, Hruban RH, et al: Geneticand immunohistochemical analysis of pancreatic aci-nar cell carcinoma: Frequent allelic loss on chromo-some 11p and alterations in the APC/beta-cateninpathway. Am J Pathol 160:953-962, 2002
7. Hansel DE, Rahman A, House M, et al: Metproto-oncogene and insulin-like growth factor bindingprotein 3 overexpression correlates with metastaticability in well-differentiated pancreatic endocrine neo-plasms. Clin Cancer Res 10:6152-6158, 2004
8. Ambros V: The functions of animal microRNAs.Nature 431:350-355, 2004
9. Caldas C, Brenton JD: Sizing up miRNAs ascancer genes. Nat Med 11:712-714, 2005
11. Michael MZ, O’Connor SM, van Holst PellekaanNG, et al: Reduced accumulation of specific microRNAsin colorectal neoplasia. Mol Cancer Res 1:882-891, 2003
12. Calin GA, Liu CG, Sevignani C, et al: MicroRNAprofiling reveals distinct signatures in B cell chroniclymphocytic leukemias. Proc Natl Acad Sci U S A101:11755-11760, 2004
13. Ota A, Tagawa H, Karnan S, et al: Identifica-tion and characterization of a novel gene, C13orf25,as a target for 13q31-q32 amplification in malignantlymphoma. Cancer Res 64:3087-3095, 2004
14. Takamizawa J, Konishi H, Yanagisawa K, et al:Reduced expression of the let-7 microRNAs in humanlung cancers in association with shortened postoperativesurvival. Cancer Res 64:3753-3756, 2004
15. Ciafre SA, Galardi S, Mangiola A, et al: Exten-sive modulation of a set of microRNAs in primaryglioblastoma. Biochem Biophys Res Commun 334:1351-1358, 2005
16. Eis PS, Tam W, Sun L, et al: Accumulation ofmiR-155 and BIC RNA in human B cell lymphomas.Proc Natl Acad Sci U S A 102:3627-3632, 2005
17. Iorio MV, Ferracin M, Liu CG, et al: MicroRNAgene expression deregulation in human breast can-cer. Cancer Res 65:7065-7070, 2005
18. Lu J, Getz G, Miska EA, et al: MicroRNAexpression profiles classify human cancers. Nature435:834-838, 2005
19. Calin G, Ferracin M, Cimmino A, et al: Aunique microRNA signature associated with prog-nostic factors and disease progression in B cellchronic lymphocytic leukemia. N Engl J Med 353:1793-1801, 2005
20. Chan JA, Krichevsky AM, Kosik KS:MicroRNA-21 is an antiapoptotic factor in humanglioblastoma cells. Cancer Res 65:6029-6033,2005
21. Cimmino A, Calin G, Fabbri M, et al: MiR-15 andmiR-16 induce apoptosis by targeting BCL2. Proc NatlAcad Sci U S A 102:13944-13949, 2005 [erratum inProc Natl Acad Sci U S A 103:2464, 2006]
Fig 6. Inverse correlation between the expression of miR-21 and its putativemRNA target PDCD4. The plot shows the expression of miR-21 and PDCD4mRNA in normal pancreas (*), metastatic (Œ), and nonmetastatic (‚) PET. Arobust locally weighted regression function has been used to fit a line among datapoints. Pearson correlation � �0.49 (95% CI, �0.68 to 0.23; P � .001).
MicroRNA in Pancreatic Endocrine and Acinar Tumors
www.jco.org 4683Downloaded from jco.ascopubs.org on August 16, 2016. For personal use only. No other uses without permission.
22. He L, Thomson JM, Hemann MT, et al: AmicroRNA polycistron as a potential human onco-gene. Nature 435:828-833, 2005
23. Johnson SM, Grosshans H, Shingara J, et al:RAS is regulated by the let-7 microRNA family. Cell120:635-647, 2005
24. Liu CG, Calin GA, Meloon B, et al: An oligo-nucleotide microchip for genome-wide microRNAprofiling in human and mouse tissues. Proc NatlAcad Sci U S A 101:9740-9744, 2004
25. Gentleman RC, Carey VJ, Bates DM, et al:Bioconductor: Open software development for com-putational biology and bioinformatics. Genome Biol5:R80, 2004
26. Huber W, von Heydebreck A, Sultmann H, et al:Variance stabilization applied to microarray data calibra-tion and to the quantification of differential expression.Bioinformatics 18:S96-S104, 2002 (suppl 1)
28. Cheng AM, Byrom MW, Shelton J, et al:Antisense inhibition of human miRNAs and indica-tions for an involvement of miRNA in cell growth
and apoptosis. Nucleic Acids Res 33:1290-1297,2005
29. Pavelic K, Hrascan R, Kapitanovic S, et al:Molecular genetics of malignant insulinoma. Anti-cancer Res 16:1707-1717, 1996
30. Katic M, Hadzija M, Wrischer M, et al: An invitro model of the early genetic events in multistagecarcinogenesis of malignant insulinoma. Carcino-genesis 20:1521-1527, 1999
31. Lewis BC, Klimstra DS, Varmus HE: Thec-myc and PyMT oncogenes induce different tumortypes in a somatic mouse model for pancreaticcancer. Genes Dev 17:3127-3138, 2003
32. Sandgren EP, Quaife CJ, Paulovich AG, et al:Pancreatic tumor pathogenesis reflects the caus-ative genetic lesion. Proc Natl Acad Sci U S A88:93-97, 1991
33. Pelengaris S, Khan M, Evan GI: Suppressionof Myc-induced apoptosis in beta cells exposesmultiple oncogenic properties of Myc and triggerscarcinogenic progression. Cell 109:321-334, 2002
34. Volinia S, Calin GA, Liu CG, et al: A microRNAexpression signature of human solid tumors definescancer gene targets. Proc Natl Acad Sci U S A103:2257-2261, 2006
35. Metzler M, Wilda M, Busch K, et al: Highexpression of precursor microRNA-155/BIC RNA inchildren with Burkitt lymphoma. Genes Chromo-somes Cancer 39:167-169, 2004
36. Kluiver J, Poppema S, de Jong D, et al: BIC andmiR-155 are highly expressed in Hodgkin, primarymediastinal and diffuse large B cell lymphomas.J Pathol 207:243-249, 2005
37. Babak T, Zhang W, Morris Q, et al: ProbingmicroRNAs with microarrays: Tissue specificity andfunctional inference. RNA 10:1813-1819, 2004
38. Jansen AP, Camalier CE, Stark C, et al: Char-acterization of programmed cell death 4 in multiplehuman cancers reveals a novel enhancer of drugsensitivity. Mol Cancer Ther 3:103-110, 2004
39. Goke R, Barth P, Schmidt A, et al: Pro-grammed cell death protein 4 suppresses CDK1/cdc2 via induction of p21(Waf1/Cip1). Am J PhysiolCell Physiol 287:C1541-C1546, 2004
40. Goke R, Gregel C, Goke A, et al: Programmedcell death protein 4 (PDCD4) acts as a tumor sup-pressor in neuroendocrine tumor cells. Ann N YAcad Sci 1014:220-221, 2004
We thank Stefano Barbi, PhD, for critically reviewing the bioinformatic sections.
Appendix
The Appendix is included in the full-text version of this article, available online at www.jco.org. It is not included in the PDF version(via Adobe® Reader®).
Authors’ Disclosures of Potential Conflicts of InterestThe authors indicated no potential conflicts of interest.
Author Contributions
Conception and design: Aldo Scarpa, Carlo M. CroceFinancial support: Aldo Scarpa, Carlo M. CroceProvision of study materials or patients: Massimo Falconi, Paola Capelli, Samantha Bersani, Aldo ScarpaCollection and assembly of data: Claudia Roldo, Massimo Falconi, Paola Capelli, Aldo ScarpaData analysis and interpretation: Claudia Roldo, Edoardo Missiaglia, John P. Hagan, Massimo Falconi, Paola Capelli, George Adrian Calin,
Stefano Volinia, Chang-gong Liu, Aldo Scarpa, Carlo M. CroceManuscript writing: Claudia Roldo, Edoardo Missiaglia, George Adrian Calin, Aldo ScarpaFinal approval of manuscript: Carlo M. Croce
GLOSSARY
MicroRNA profiles: The study of the global expression ofmicroRNAs in tissues.
MicroRNAs: Endogenous noncoding RNAs approximately 22 nucleo-tides long that regulate gene silencing by post-transcriptional mechanismssuch as cleavage or translational repression.
