Royal College of Surgeons in Ireland e-publications@RCSI Physiology and Medical Physics Articles Department of Physiology and Medical Physics 30-4-2015 Circulating miRNAs miR-34a and miR-150 associated with colorectal cancer progression. Sinéad T. Aherne Dublin City University Stephen F. Madden Dublin City University David J. Hughes Royal College of Surgeons in Ireland, [email protected] Barbara Pardini Human Genetics Foundation, Turin Alessio Naccarati Human Genetics Foundation, Turin See next page for additional authors Citation Aherne ST, Madden SF, Hughes DJ, Pardini B, Naccarati A, Levy M, Vodicka P, Neary P, Dowling P, Clynes M12. Circulating miRNAs miR-34a and miR-150 associated with colorectal cancer progression. BMC Cancer. 2015;15(1):329 This Article is brought to you for free and open access by the Department of Physiology and Medical Physics at e-publications@RCSI. It has been accepted for inclusion in Physiology and Medical Physics Articles by an authorized administrator of e-publications@RCSI. For more information, please [email protected]. Authors Sinéad T. Aherne, Stephen F. Madden, David J. Hughes, Barbara Pardini, Alessio Naccarati, Miroslav Levy, Pavel Vodicka, Paul Neary, Paul Dowling, and Martin Clynes This article is available at e-publications@RCSI:https://epubs.rcsi.ie/physiolart/58 — Use Licence — This work is licensed under aCreative Commons Attribution-Noncommercial-Share Alike 4.0 License. This article is available at e-publications@RCSI:https://epubs.rcsi.ie/physiolart/58 Aherneetal.BMCCancer (2015) 15:329 DOI10.1186/s12885-015-1327-5 RESEARCH ARTICLE Open Access Circulating miRNAs miR-34a and miR-150 associated with colorectal cancer progression Sinéad T Aherne1*, Stephen F Madden1, David J Hughes2, Barbara Pardini3, Alessio Naccarati3,5, Miroslav Levy4, Pavel Vodicka5, Paul Neary6, Paul Dowling1,7 and Martin Clynes1 Abstract Background: Screeningfortheearlydetectionofcolorectalcancerisimportanttoimprovepatientsurvival.Theaim ofthisstudywastoinvestigatethepotentialofcirculatingcell-freemiRNAsasbiomarkersofCRC,andtheirefficiency atdelineatingpatientswithpolypsandbenignadenomasfromnormalandcancerpatientgroups. Methods:Theexpressionof667miRNAswasassessedinadiscoverysetof48plasmasamplescomprisingnormal, polyp,adenoma,andearlyandadvancedcancersamples.ThreemiRNAs(miR-34a,miR-150,andmiR-923)werefurther examinedinavalidationcohortof97subjectsdividedintothesamefivegroups,andinanindependentpublicdataset of40CRCsamplesandpairednormaltissues. Results:HighlevelsofcirculatingmiR-34aandlowmiR-150levelsdistinguishedgroupsofpatientswithpolypsfrom thosewithadvancedcancer(AUC=0.904),andlowcirculatingmiR-150levelsseparatedpatientswithadenomasfrom thosewithadvancedcancer(AUC=0.875).Inaddition,thealteredexpressionofmiR-34aandmiR-150inanindependent publicdatasetoffortyCRCsamplesandpairednormaltissueswasconfirmed. Conclusion:WeidentifiedtwocirculatingmiRNAscapableofdistinguishingpatientgroupswithdifferentdiseasesof thecolonfromeachother,andpatientswithadvancedcancerfrombenigndiseasegroups. Keywords:Colorectalcancer,CirculatingmiRNAs,miR-34a,miR-150,miR-923 Background Screening at risk populations for CRC has significantly Colorectal cancer (CRC) poses a significant threat to the improved the outcome for patients, for instance diagno- health of global populations; it is the second most com- sis while the disease remains localised to the colon dra- monly diagnosed cancer in females and the third in matically improves patient survival, and removal of early males [1]. CRC develops in a progressive fashion during lesions such as adenomatous polyps may prevent disease which normal colon epithelial cells transform to form formation [4]. There are currently several potential benign growths such as polyps. These polyps may then screeningtestsavailabletodetectCRCincludingthefae- progress to benign adenomas, and ultimately to invasive cal occult blood test (FOBT), flexible sigmoidoscopy cancer lesions. The progression of the cancer has also (FS), optical colonoscopy (OC) and computed tomog- been associated with sequential genetic changes in genes raphy colonography (CTC). FOBT is a simple, cheap such as K-RAS, APC, DCC, and P53 [2]. However CRC and safe test that relies on the assumption that large ad- is a heterogeneous disease with various patient-related enomas and cancerous lesions may bleed, and that these confounding factorssuch asthe anatomic location ofthe blood products are detectable in the faecal matter of pa- tumour, race/ethnicity of the patient, and genetic and tients. Although cheap and non-invasive, this test is vul- dietary interactions influencing the development of the nerable to false positive and negative results due to disease[3]. incorrect sample storage, or confounding medical com- plaints such as haemorrhoids. The other examinations involve more costly and invasive procedures which al- *Correspondence:[email protected] though allow direct access to colorectal lesions also 1MolecularTherapeuticsforCancerIreland,NationalInstituteforCellular Biotechnology,DublinCityUniversity,Glasnevin,Dublin9,Ireland Fulllistofauthorinformationisavailableattheendofthearticle ©2015Aherneetal.;licenseeBioMedCentral.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative CommonsAttributionLicense(http://creativecommons.org/licenses/by/4.0),whichpermitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginalworkisproperlycredited.TheCreativeCommonsPublicDomain Dedicationwaiver(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthisarticle, unlessotherwisestated. Aherneetal.BMCCancer (2015) 15:329 Page2of13 suffer from low patient acceptance and procedural risks and cancer groups. To facilitate this we performed such asperforationofthe colon[4]. miRNAprofilingfor667miRNAsonadiscoverysetof48 The focus of the scientific community has thus shifted plasma samples comprising 8 normal, 8 polyp, 16 aden- to exploring the identification of non-invasive bio- omasamples,8earlystagecancersamples(stageI/II),and markers of disease from bio-fluids such as saliva, urine, 8advancedcancersamples(stageIII/IV).Threecandidate and blood. MicroRNAs (miRNAs) are nucleic acid miRNAs; miR-34a, miR-150, and miR-923 were then fur- markers that have been recently investigated in this con- ther examined in a validation cohort of 97 independent text. MiRNAs are short (20-22nt) non-coding RNAs that plasmasamplescomprising20normal,20polyp,20aden- negatively regulate gene expression through either oma samples, 23 early stage cancer samples, and 14 ad- mRNA degradation or translational repression [5]. vanced cancer samples. In addition, we confirmed the MiRNA expression has been shown to be altered in can- altered expression of two of the miRNAs in an independ- cerous tissue compared to normal tissue and different ent dataset of 40 CRC samples and their paired normal miRNAs have been attributed oncogenic and tumour tissues. We found circulating levels of miR-34a and miR- suppressor qualities [6]. In 2008, Chen et al. detected 150 to be capable of distinguishing patients groups with miRNAs in the serum and plasma blood components of benign and malignant diseases of the colon from each humans and other animals. This primary study illus- other, and sets of miRNAs that distinguish patients with trated that miRNAs remain stable in serum after being advanced cancer from benign disease groups. Specifically, subject to severe conditions such as extremely low or wefoundhighlevelsofcirculatingmiR-34aandlowmiR- high pH, 10 freeze-thaw cycles, extended storage, boil- 150 levels to distinguish patients with polyps from those ing, and RNase digestion [7]. In addition to their pres- with advanced cancer, and low circulating miR-150 levels ence in serum and plasma, miRNAs have also been to separate patients with adenomas from those with ad- detected in other body fluids such as urine, saliva, and vancedcancer. amniotic fluid making them ideal potential candidates as non-invasive biomarkers ofdisease[8]. Methods Expression levels of circulating miRNAs have shown Patientsselectionandsamplecollection some potential at distinguishing cancer patients and Cases with positive colonoscopy results for malignancy, healthy controls for prostate [9], ovarian [10], lung confirmed by histology as colon or rectal carcinomas, [11,12],andbreastcancers[13].Severalstudieshavealso were recruited between December 2007 and December investigated circulating miRNA levels for the detection 2010 at the Department of Surgery, Adelaide and Meath of CRC. Initial approaches analysed small numbers of Hospital and at theThomayer Hospital in Prague, Czech circulating miRNAs in CRC patient samples compared to Republic. Control subjects or subjects diagnosed with normal controls [14]. Other groups performed miRNA polyps or adenomatous polyps were selected during the profiling on pooled plasma samples and validated candi- same period from individuals undergoing colonoscopy date biomarkers on additional individual samples [15], for various gastrointestinal complaints (macroscopic andothersperformedprofilingonasmallnumberofCRC bleeding, positive faecal occult blood test or abdominal tissue/serum/plasma samples before validation in a larger pain of unknown origin). The participating subjects gave sample set [16]. These studies have produced conflicting written informed consent inaccordance with theDeclar- results[17]andsorecently,groupshavebeguntoperform ation of Helsinki at the precipitating site that was ap- profiling on larger sample sets and included plasma from proved by Tallaght Hospital/St. James’s Hospital Joint patients with adenomas in addition to CRC to improve Research Ethics Committee, The Adelaide and Meath thespecificityofdiseasedetection[18]. Hospital, Dublin, Incorporating The National Children’s In 2008, a guideline was released from the American Hospital, Tallaght, Dublin 24, Ireland and the Ethical Cancer Society which highlighted the importance for pa- Committee of the Institute of Experimental Medicine, tients to have access to screening tests that will facilitate Prague, Czech Republic. SeeTable 1 for clinical informa- cancer prevention through the early detection of cancer, tion onsamplesused. andthedetectionandremovalofpolyps[4].Acleardeficit Two separate patient cohorts were identified, a discov- in the search for circulating biomarkers for the early de- ery set (n=48) comprising 8 normal, 8 polyp, 16 aden- tection of CRC to date is the lack of adenomatous polyp oma samples, 8 early stage cancer samples (stage I/II), samples and the lack of separation of advanced and early and 8 advanced cancer samples (stage III/IV), and a val- stage cancers represented in studies [14-16,18-20]. The idation set (n=97) comprising 20 normal, 20 polyp, 20 aimofthisstudywasthereforetoinvestigatethepotential adenoma samples, 23 early stage cancer samples, and 14 of circulating cell-free miRNAs not only as biomarkers of advanced cancer samples. In addition, an independent CRC, but also their efficiency at delineating patients pre- public dataset [21] of quantitative real-time PCR (qRT- senting with polyps and benign adenomas from normal PCR) raw data was downloaded from the NCBI GEO Aherneetal.BMCCancer (2015) 15:329 Page3of13 Table1Clinicalinformationonthediscoveryand and quantitative PCR (qPCR) were performed on equal validationplasmasamplecohorts volumesofRNAfromeachsampleaccordingtothemanu- DiscoveryCohort facturer’s instructions using TaqMan® MicroRNA Reverse n(M/F) Age Transcription Kit (Cat no 4366596, Applied Biosystems) and Megaplex RT Primers to convert the miRNAs to Normal 8(4/4) 67±11 cDNA, TaqMan® PreAmp Master Mix (Cat no 4391128, Polyps 8(4/4) 65±7 Applied Biosystems) and Megaplex PreAmp Primers for a Adenoma 16(8/8) 56±6 preamplification step before real-time analysis. qPCR was EarlyStageCancer(StageI/II) 8(4/4) 65±10 performed using TaqMan® Universal Master Mix II, no AdvancedCancer(StageIII/IV) 8(4/4) 68±8 UNG (Cat no 4440048, Applied Biosystems) on the ValidationCohort 7900HTFastReal-TimePCRsystem(AppliedBiosystems). n(M/F) Age The Sequence Detector System software version 2.2.2 was Normal 20(12/8) 63±8 utilisedtogeneratestudyfilesusingafixedthresholdvalue of0.1forstatisticalanalysis(accessionno:GSE67075). Polyps 20(11/9) 57±7 Adenoma 20(12/8) 62±10 EarlyStageCancer(StageI/II) 23(10/13) 63±12 ValidationofmiRNAexpressionusingqRT-PCR AdvancedCancer(StageIII/IV) 14(9/5) 67±8 Individual TaqMan® miRNA assays were used for Mdenotesmale;Fdenotesfemale. miRNA quantification in the 97 plasma samples in the validation cohort. To improve reverse transcription effi- archive (accession no: GSE28364) which contains infor- ciency a miRNA multiplex RT primer pool was made mation on 40 CRC samples and their paired normal from the singleplex RT primers of the four miRNAs to tissues. be analysed; miR-34a, miR-150, miR-923, and miR-let7e Plasma samples were collected according to standard (this miRNA was used as the endogenous control as it phlebotomy procedures. 10 ml of blood sample was col- showed very little variation in the discovery cohort, ΔC t lected into EDTA plasma tubes and immediately placed SD=0.865). 100 μl of each 20X RT primer were added to in ice. Thetubes were centrifuged at 1000 x g for 10 mi- an RNase-free microfuge tube. The tube was dried in a nutes at 4°C. Plasma was denuded by pipette from the speed vacuum (MAXI dry plus, Medical Supply Company, cellular material, aliquoted into cryovial tubes, labelled Ireland)at50°Cfor1hour.Theprimerswerere-suspended and stored at -80°C until the time of analysis. The time in100μlofnuclease-freewaterand300μlof0.1XTEbuf- from sample procurement to storage at -80°C was less ferwasaddedtoyielda5XmultiplexRTprimerpool.The than 3 hours. Each plasma sample underwent no more TaqMan® MicroRNA Reverse Transcription Kit (Cat no than 3freeze/thawcycles priortoanalysis. 4366596, Applied Biosystems) was used to perform reverse transcriptionreactions.Eachreactioncontained1.8μlofRT RNAextraction buffer (10X), 0.18 μl of dNTPs (25 mM), 3.6 μl of miRNA TotalRNAwasisolatedfrom60μlofeachplasmasample multiplexRTprimerpool(5X),1.2μlofMultiscribeRTen- using the miRNeasy mini kit (Cat no 217004, Qiagen). zyme(50U/μl),5.22μlofnuclease-freewaterand6μlofex- The Qiagen supplementary protocol (Purification of total tractedtotalRNA.Thereactionswereincubatedat16°Cfor RNA, including small RNAs, from serum or plasma) was 30 minutes, 42°C for 30 minutes and 85°C for 5 minutes utilised with the following modifications: thawed plasma (G-STORM,GS1,SomertonBiotechnologyCentre,UK). sampleswerecentrifugedat1000xgfor5minutesat4°C Real-time PCR analysis was performed on 96 well toremoveexcessdebrisfromsamples,RNAwasextracted plates (Cat no 4346906, Applied Biosystems). Technical from the upper 50 μl of each sample. To elute the RNA, triplicate PCRs were performed for each sample, and no 50μlofnuclease-freewaterwasaddedtoeachspincolumn template controlsandapooledsample containing cDNA and incubated for 1 minute at room temperature before from all 97 samples were included on each plate to en- centrifuging into non-stick RNase-free microfuge tubes sure inter-plate reproducibility. Each reaction contained (CatnoAM12350,Ambion)toelutetheRNA. 1 μl of TaqMan miRNA assay (20X), 10 μl of TaqMan® Universal Master Mix II, no UNG (Cat no 4440048, MiRNAprofilingofplasmawithTaqMan®low-density Applied Biosystems), 7.67 μl of nuclease-free water, and arrays 1.33 μl of cDNA. The reactions were incubated at 95°C TaqMan® Array Human MicroRNA A and B Cards v2.0 for 10minutes,and40cycles of95°Cfor 15secondsand (Cat no 4400238, Applied Biosystems) were employed to 60°C for 15 seconds onthe 7900HTFastReal-Time PCR examine the expression of 667 miRNAs in 48 plasma system (Applied Biosystems). The Sequence Detector samples in the discovery cohort. Reverse transcription System software version 2.2.2 was utilised to generate Aherneetal.BMCCancer (2015) 15:329 Page4of13 study files using a fixed threshold value of 0.1 for statis- sum test was used to determine significance between the ticalanalysis. groups. Logistic regression (LR) and receiver operator charac- Statisticalanalysis teristic (ROC) curve analysis were performed on miR- In the discovery cohort (n=48), each miRNA was nor- 34a, miR-150 and miR-923 in the validation cohort. The malised by the ΔΔC method using the average within markers were combined using LR and the ROC curves t sample C value [22]. This technique involves the use of were used for interpretation of the models generated. t the mean expression value of all expressed microRNAs The area under the curve (AUC) from the ROC curve in a given sample as a normalisation factor for micro- for a given model was used to determine the probability RNA real-time quantitative PCR data. Thus the average of a correct prediction. The LR model for single miR- within sample C value for each card is calculated by NAs or combinations ofmiRNAs which gavethe highest t averaging all miRNA Ct values for each individual sam- AUC was considered the most discriminating model and ple. This was performed using the Bioconductor package therefore the best marker at distinguishing between the HTqPCR (www.bioconductor.org). The non-parametric groups of interest. All calculations were carried out in Kruskal-Wallis test was used to determine between the R statistical environment (http://cran.r-project.org/) group variations by rank as the data was not normally usingtheHTqPCR andstatspackages. distributed. A Wilcoxon rank sum test was subsequently used to perform pair-wise comparisons between the 5 Results groups for the significant miRNAs identified by the DifferentialexpressionofmiRNAsinthediscoverycohort Kruskal-Wallis test. This study examined the expression of 667 miRNAs in As an alternative to spiking un-related miRNA con- plasma samples of a discovery cohort of 48 patients with structs into our samples we utilised the miRNA profiling benign and malignant disease of the colon compared to data of the discovery cohort of samples to choose an ap- age and sex matched disease-free controls (Table 1). propriate endogenous control for use in the validation Statistical analysis revealed 73 miRNAs that have signifi- cohort. This involved analysing the expression of all 667 cantly different levels (p-value<0.05) in at least one of the miRNAs across all 48 samples in the discovery cohort disease groups (polyp n=8, adenoma n=16, early stage allowing us to choose one of the least variant miRNAs. cancer(stageI/II)n=8,andadvancedcancer(stageIII/IV) As MammU6 showed highly variant expression in the n=8) compared to the healthy controls n=8 (Table 2). discovery cohort, miR-let7e was chosen for useasan en- FortymiRNAsweresignificantlydifferentinadvancedcan- dogenous control for the validation set as it was one of cer, 22 in early stage cancer, 7 in adenoma, and 22 in the the least variant miRNAs in the discovery phase experi- polyp group compared to normal controls. Ten miRNAs ment (ΔC standard deviation of 0.86). When the let7e were significantly altered in both the early stage and ad- t Cs were examined across all samples in the validation vanced cancer groups compared to the normal controls; t cohort this miRNA proved an appropriate endogenous miR-923, miR-801, miR-144*, miR-135a*, miR-500, miR- control with a C standard deviation of 1.64. Statistically 497, miR-150, miR-30c and RNU48 showed lower levels t significant differences were determined using the non- while miR-532-3p was more abundant in the cancer parametric Wilcoxon rank sum test. The p-values for groups compared to the controls. MiR-34a was signifi- the validation set were adjusted using the Benjamini and cantly higher in early stage cancer compared to healthy Hochbergmethod[23]toaccountformultipletesting. controls. For consistency, the independent public dataset from miRNAs were prioritised for subsequent confirmation Reid et al. [21] (accession no: GSE28364) was normal- in the validation sample set if they showed consistently isedusingthesame approachusedtoanalyse thediscov- altered levels between the control group and each of the ery cohort qRT-PCR data. This independent study used disease groups, and if all or most of these changes were TaqMan® Array Human MicroRNA Cards v2.0 to ana- deemed to be statistically significant. MiR-34a was lyse miRNA expression in 40 CRC tumour samples and chosen for validation as it was increased in the plasma of their paired normal tissues. In order to mimic this struc- diseased patients compared to controls. MiR-150 and tureinour validation plasma sample cohort,wegrouped miR-923 were chosenfor validation as their plasma levels samples into ‘non-malignant’ and ‘malignant’groups. As progressivelydecreasedasthesample/diseasegroupspro- there were only two groups (normal versus cancer) in gresstowardmalignancy(Table2). this analysis, the Wilcoxon rank sum test was used to determine significantly differentially regulated miRNAs. AlteredlevelsofmiR-34a,miR-150,andmiR-923inthe For this analysis of the validation cohort, miR-34a, miR- validationcohort 150 and miR-923 were first normalised against the en- ThethreecandidatemiRNAs;miR-34a,miR-150,andmiR- dogenous control (miR-let7e) and the Wilcoxon rank 923wereanalysedinavalidationcohortof97independent Aherneetal.BMCCancer (2015) 15:329 Page5of13 Table2miRNAswithsignificantlydifferentlevelsindiseasegroupscomparedtonormalinthediscoverycohort MiRNAs NormalvsPolyp NormalvsAdenoma NormalvsEarlyCancer NormalvsLateCancer Log FC PValue Log FC PValue Log FC PValue Log FC PValue 2 2 2 2 hsa-let-7b −0.823 0.007 −0.618 0.038 −0.712 0.083 −0.298 0.328 hsa-let-7c 3.407 0.442 4.27 0.136 4.794 0.105 0.812 0.753 hsa-let-7g −0.749 0.195 −0.231 0.153 −1.177 0.010 −3.933 0.234 hsa-miR-135a 13.518 0.050 0.614 0.968 −5.76 0.171 3.738 0.655 hsa-miR-140-5p 0.672 0.038 0.122 0.834 0.558 0.083 −2.645 0.021 hsa-miR-146b-3p 11.496 0.281 −1.962 0.511 −1.377 0.904 10.34 0.043 hsa-miR-150 −0.523 0.161 −0.706 0.032 −1.491 0.015 −2.771 0.000 hsa-miR-15b −0.538 0.130 −0.362 0.238 −2.019 0.001 −0.746 0.105 hsa-miR-182 7.054 0.178 −2.384 0.864 −3.173 0.685 −11.924 0.032 hsa-miR-183 −0.192 0.599 −1.533 0.667 −3.09 0.792 −15.707 0.010 hsa-miR-190 −0.139 0.874 −9.03 0.043 −16.68 0.018 −7.941 0.075 hsa-miR-191 0.138 0.505 −0.049 0.383 −0.979 0.083 0.509 0.279 hsa-miR-192 3.363 0.721 2.913 0.061 3.349 0.505 −2.113 0.036 hsa-miR-193a-5p 4.76 0.105 2.668 0.417 3.744 0.798 −1.73 0.012 hsa-miR-194 2.184 0.105 2.992 0.452 3.008 0.234 −1.904 0.016 hsa-miR-199a-3p 1.037 1.000 −0.095 0.264 −0.217 0.234 −2.353 0.010 hsa-miR-19a 1.229 0.050 0.318 0.976 1.58 0.028 0.831 0.083 hsa-miR-19b 1.125 0.010 0.302 0.976 1.447 0.065 0.696 0.028 hsa-miR-204 12.069 0.005 6.087 0.096 6.565 0.751 3.338 0.790 hsa-miR-210 2.864 0.003 −3.207 0.927 1.987 0.028 −1.692 0.105 hsa-miR-21 −0.232 0.050 −0.239 0.238 0.561 0.130 −3.683 0.065 hsa-miR-219-1-3p −2.985 0.382 −1.557 0.610 3.221 0.488 8.939 0.142 hsa-miR-23a 3.751 0.574 −1.265 0.667 0.948 0.400 3.146 0.012 hsa-miR-25 0.717 0.721 0.052 0.120 0.853 0.574 −0.621 0.038 hsa-miR-30b 0.04 0.721 −0.117 0.452 −0.964 0.028 −0.88 0.130 hsa-miR-30c −0.017 0.878 −0.231 0.172 −0.845 0.038 −6.451 0.038 hsa-miR-323-3p −0.461 0.328 0.328 0.452 0.243 0.574 −2.93 0.130 hsa-miR-337-5p 6.448 0.359 4.019 0.391 3.964 0.547 12.303 0.007 hsa-miR-34a 18.817 0.003 12.09 0.119 15.428 0.018 11.626 0.055 hsa-miR-365 0.257 0.878 −1.16 0.061 −0.884 0.279 −5.763 0.010 hsa-miR-370 9.181 0.040 4.218 0.283 −2.578 0.790 6.114 0.065 hsa-miR-377 0 NA 4.206 0.221 3.063 0.382 14.295 0.013 hsa-miR-451 0.216 1.000 −0.341 0.120 −0.111 0.721 −1.955 0.007 hsa-miR-486-3p −0.225 0.382 0.062 0.569 0.176 0.574 −5.158 0.010 hsa-miR-486-5p −0.813 0.065 −0.241 0.787 −0.562 0.798 −2.187 0.003 hsa-miR-500 −5.223 0.018 −6.169 0.084 −11.283 0.007 −9.256 0.011 hsa-miR-503 −3.035 0.382 −1.631 0.610 −3.035 0.382 7.943 0.225 hsa-miR-532-3p 2.845 0.195 3.061 0.452 1.402 0.015 2.436 0.050 hsa-miR-532-5p 4.749 0.105 3.772 0.787 4.688 0.130 0.31 0.834 hsa-miR-542-3p 3.213 0.382 0 NA 0 NA 7.988 0.076 hsa-miR-548d-3p −3.186 0.382 −0.077 0.958 2.882 0.700 12.243 0.117 hsa-miR-548d-5p 9.445 0.076 1.349 0.536 2.736 0.382 14.573 0.013 hsa-miR-654-3p 10.564 0.012 4.161 0.153 3.815 0.250 11.441 0.007 Aherneetal.BMCCancer (2015) 15:329 Page6of13 Table2miRNAswithsignificantlydifferentlevelsindiseasegroupscomparedtonormalinthediscoverycohort (Continued) hsa-miR-660 0.544 0.234 0.228 0.106 0.447 0.721 −4.865 0.007 RNU48 −3.549 0.365 1.637 0.878 −12.871 0.064 −7.855 0.075 hsa-let-7a-3p 10.399 0.148 8.684 0.217 −6.667 0.171 10.717 0.118 hsa-miR-135-3p −0.835 0.195 −0.512 0.417 −1.688 0.002 −2.27 0.001 hsa-miR-136-3p 5.14 0.028 −0.617 0.357 −0.031 1.000 2.942 0.016 hsa-miR-138-1-3p −0.353 0.065 −0.038 0.697 −0.033 0.721 −1.37 0.038 hsa-miR-144-5p −1.373 0.065 −1.208 0.011 −3.164 0.001 −2.299 0.028 hsa-miR-151-3p 0.398 0.574 0.343 0.697 0.623 0.574 2.57 0.007 hsa-miR-16-1-3p 2.883 0.873 −7.605 0.117 7.397 0.290 4.03 0.424 hsa-miR-221-5p −6.94 0.171 1.828 0.732 −6.94 0.171 10.556 0.183 hsa-miR-222-5p 3.311 0.566 −3.168 0.479 −2.906 0.590 12.137 0.104 hsa-miR-25-5p 18.99 0.018 5.097 0.334 17.026 0.024 12.007 0.117 hsa-miR-30a-3p 0.171 0.382 −0.048 0.881 0.736 0.003 −3.516 0.279 hsa-miR-30e-3p −0.136 0.959 −0.8 0.038 −1.034 0.038 −1.083 0.574 hsa-miR-30e 0.798 0.003 −0.048 0.881 1.107 0.010 0.466 0.065 hsa-miR-497 5.305 0.004 −0.143 0.302 5.1 0.011 4.304 0.001 hsa-miR-509-3p −0.329 0.083 −0.06 0.653 0.825 0.234 −5.263 0.007 hsa-miR-559 −11.278 0.009 1.504 0.991 1.144 0.804 −7.323 0.007 hsa-miR-605 −3.301 0.164 4.458 0.233 −0.111 0.974 12.867 0.006 hsa-miR-609 −1.583 0.349 −0.832 0.613 −0.1 1.000 1.618 0.576 hsa-miR-610 −0.557 0.038 −0.463 0.291 0.493 0.382 −1.327 0.021 hsa-miR-632 −0.766 0.105 −0.495 0.350 0.176 0.721 −2.335 0.010 hsa-miR-645 −1.016 0.019 −0.628 0.135 −0.426 0.224 −5.643 0.001 hsa-miR-668 14.982 0.004 7.463 0.072 −2.096 0.549 11.805 0.017 hsa-miR-7 4.86 0.181 1.229 0.437 5.291 0.048 0.28 0.611 hsa-miR-768-3p 6.126 0.085 −7.483 0.206 −9.603 0.059 −9.625 0.059 hsa-miR-801 1.108 0.270 −0.261 0.787 −3.102 0.000 −6.163 0.008 hsa-miR-923 −1.635 0.000 −1.184 0.001 −2.833 0.000 −6.284 0.007 RNU24 11.491 0.000 9.167 0.000 1.901 0.291 3.178 0.089 RNU48 −3.073 0.074 0.595 0.649 −10.127 0.001 −9.95 0.000 miRNAshighlightedinboldwerechosenforvalidationinthevalidationcohortof97plasmasamples. plasma samples comprising 20 normal, 20 polyp, 20 aden- miR-150 distinguished the polyp and adenoma groups oma samples, 23 early stage cancer samples, and 14 ad- from the advanced cancer ((FC -2.31, p-value=0.007, vancedcancersamples(seeTable3). FC -2.55, p-value=0.001 respectively) (Figure 1B). Fol- Among the three miRNAs analysed in the validation co- lowingadjustmentformultipletesting,miR-923wasnot hort, only miR-34a distinguished the normal and precan- foundtohavesignificantlyalteredlevelsinthevalidation cerouslesiongroupsfromthediseasesamples(Figure1A). cohort(Figure1C). MiR-34a expression was significantly increased in the ad- enoma (FC 2.09, p-value=0.028) and early stage cancer ValidationofalteredmiR-34a,miR-150,andmiR-923 (FC 2.84, p-value=0.002) groups compared to healthy expressioninanindependentdatasetofmatchedcolon controls, and moderately in the advanced cancer group tumourandnormaltissues (FC 1.80, p-value=0.081). The levels of this miRNA In an effort to determine whether the levels of circulat- were also significantly higher in the adenoma (FC 2.71, ing cell free miRNAs that were observed in our study p-value=0.002), early stage cancer (FC 3.69, p-value= reflected the biology of the tumour, we investigated their 3.90e-5) and advanced cancer (FC 2.34, p-value=0.006) expression in a publically available qPCR dataset of CRC groups compared to the polyp samples. Alternatively, tumours and matched normal tissues [21]. All miRNAs Aherneetal.BMCCancer (2015) 15:329 Page7of13 Table3Foldchangesandassociatedp-valuesformiR-34a,miR-150,andmiR-923inthevalidationcohort miR-34a miR-150 miR-923 FC Pvalue FC Pvalue FC Pvalue NormalvsPolyp −1.3 0.286 1.52 0.218 1.24 0.537 NormalvsAdenoma 2.09 0.028 1.68 0.160 1.19 0.537 NormalvsEarlyStageCancer 2.84 0.002 −1.03 0.800 1.19 0.675 NormalvsAdvancedCancer 1.8 0.081 −1.52 0.282 −1.7 0.537 PolypvsAdenoma 2.71 0.002 1.11 0.735 −1.04 0.904 PolypvsEarlyStageCancer 3.69 0.000 −1.57 0.160 −1.05 0.779 PolypvsAdvancedCancer 2.34 0.006 −2.31 0.007 −2.11 0.113 AdenomavsEarlyStageCancer 1.36 0.381 −1.73 0.104 −1.01 0.867 AdenomavsAdvancedCancer −1.16 0.691 −2.55 0.001 −2.03 0.113 EarlyStageCancervsAdvancedCancer −1.58 0.169 −1.47 0.529 −2.02 0.113 Non-CancervsEarlyStageCancer 3.07 0.000 1.42 0.160 1.19 0.675 Non-CancervsCancer 2.62 0.001 1.29 0.013 −1.11 0.779 Normal&PolypvsAdenoma 1.9 0.002 1.98 0.282 1.08 0.537 Boldtextdenotessignificantcomparisons.Non-Cancercomprisesnormal,polyp&adenomagroups,Cancercomprisesbothearlystage&advancedcancergroups. found to be significantly differentially expressed in the DiagnosticpotentialofcirculatingmiR-34a,miR-150,and independent dataset are listed in Additional file 1. As miR-923forthedetectionofdiseaseofthecolon this independent dataset only contains information on Once the altered expression of miR-34a, miR-150, and cancer and adjacent normal samples, the normal, polyp miR-923 was confirmed in our 97 validation plasma and adenoma plasma samples in our validation cohort samples, and in the independent tissue samples, LR and were combined into a ‘non-malignant’ group and the ROC analyses were used to evaluate the potential of early stage and advanced cancer plasma samples were these miRNAs to distinguish between the disease and combined into a ‘cancer’ group to facilitate comparison control blood plasma samples. First LR was used to ofthemiRNA expression changes. identify the linear model with the best discriminatory Figure 2 illustrates that the altered levels of all three power between sample groups, and the quality of this circulating miRNAs observed in the plasma samples of model was depicted by the area under the curve (AUC) the validation cohort mirrors the expression changes in of the ROC curve. For individual miRNAs, AUCs ranged the CRC tissues in the independent tissue sample set. from 0.488 to 0.875 (see Additional file 2). In an effort Mir-34a was found to be significantly up-regulated in to identify the most powerful candidates as diagnostic both the validation and the independent sample sets, markers of disease, we focused on comparing sample 2.62 fold up-regulated (p-value=<0.001) and 1.71 fold groups that showed significant differential expression of up-regulated (p-value=8.50e-6) respectively (Figure 2A). morethanoneofthethreemiRNAs.Thuswehaveidenti- In further examining the expression of miR-34a in the fied sets of miRNAs that could distinguish advanced can- combined plasma sample groups in the validation co- cer from benign disease groups (Figure 3). Individually, hort, it was observed that it was also able to distinguish plasmalevels ofmiR-34a,miR-150,and miR-923discrim- the non-malignant group (normal, polyp & adenoma inate polyp samples from advanced cancer samples with groups) from the early stage cancer group (FC=3.07, AUC=0.796 (CI:0.646-0.947), 0.825 (CI:0.681-0.969), and p-value=<0.001), and the normal and polyp groups from 0.746 (CI:0.412-0.817) respectively. The discriminatory the benign adenoma group (FC=1.90, p-value=0.001). poweroftheanalysiswasimprovedbycombiningmarkers Additionally,miR-150 wassignificantlydown-regulated in miR-34a (p-value=0.016, CI:-2.911- -0.314) and miR-150 both the cancer plasma (FC=-1.29, p-value=0.013) and (p-value=0.031, CI:0.392-2.640) which increased the cancertissue(FC=-2.55,p-value=4.104e-08)samples.Al- AUC to 0.904. The Akaike’s Information Criterion (AIC) thoughnotsignificantly,miR-923wasalsodown-regulated was used as a measure of the quality of the model. The inboththecancerplasma(FC=-1.11,p-value=0.779)and modelwiththelowest AICwasconsideredthebestfiti.e. cancer tissue (FC=-2.26, p-value=0.353) samples. See the combination of miRNAs that produced the lowest Table 3 for all FCs and associated p-values for miR-34a, AIC. Where the addition of a miRNA did not lower the miR-150,andmiR-923inthevalidationcohort. AIC it was excluded as the simplest model that best