Suppression of cancer stemness p21-regulating mRNA and microRNA signatures in recurrent ovarian cancer patient samples.

0.95 MB·English

by  Gallagher, Michael F

Gallagheretal.JournalofOvarianResearch2012,5:2
http://www.ovarianresearch.com/content/5/1/2
RESEARCH Open Access
Suppression of cancer stemness p21-regulating
mRNA and microRNA signatures in recurrent
ovarian cancer patient samples
Michael F Gallagher1,2*, Cynthia CBB Heffron1,4, Alexandros Laios1,3, Sharon A O’Toole1,3, Brendan Ffrench1,2,
Paul C Smyth1, Richard J Flavin2, Salah A Elbaruni1,2, Cathy D Spillane1,2, Cara M Martin1,2, Orla M Sheils1,2 and
John J O’Leary1,2
Abstract
Background: Malignant ovarian disease is characterised by high rates of mortality due to high rates of recurrent
chemoresistant disease. Anecdotal evidence indicates this may be due to chemoresistant properties of cancer stem
cells (CSCs). However, our understanding of the role of CSCs in recurrent ovarian disease remains sparse. In this
study we used gene microarrays and meta-analysis of our previously published microRNA (miRNA) data to assess
the involvement of cancer stemness signatures in recurrent ovarian disease.
Methods: Microarray analysis was used to characterise early regulation events in an embryonal carcinoma (EC)
model of cancer stemness. This was then compared to our previously published microarray data from a study of
primary versus recurrent ovarian disease. In parallel, meta-analysis was used to identify cancer stemness miRNA
signatures in tumor patient samples.
Results: Microarray analysis demonstrated a 90% difference between gene expression events involved in early
regulation of differentiation in murine EC (mEC) and embryonic stem (mES) cells. This contrasts the known parallels
between mEC and mES cells in the undifferentiated and well-differentiated states. Genelist comparisons identified a
cancer stemness signature set of genes in primary versus recurrent data, a subset of which are known p53-p21
regulators. This signature is present in primary and recurrent or in primary alone but essentially never in recurrent
tumors specifically. Meta-analysis of miRNA expression showed a much stronger cancer stemness signature within
tumor samples. This miRNA signature again related to p53-p21 regulation and was expressed prominently in
recurrent tumors. Our data indicate that the regulation of p53-p21 in ovarian cancer involves, at least partially, a
cancer stemness component.
Conclusion: We present a p53-p21 cancer stemness signature model for ovarian cancer. We propose that this may,
at least partially, differentially regulate the p53-p21 mechanism in ovarian disease. Targeting CSCs within ovarian
cancer represents a potential therapeutic avenue.
Keywords: Ovarian cancer, chemoresistance, recurrent disease, cancer stem cell, p53, p21
Background ovariancancerdevelopssilently,withmostpatientssymp-
Ovarian cancer is the leading gynecological malignancy, tom-free,onlypresentingatanadvancedstage.Treatment
affecting more than 200,000 women per annum world- ofprimarydiseasegenerallyconsistsofsurgicalremovalof
wide [1,2]. This is largely due to high rates of chemore- themalignancyincombinationwithplatinum-basedtreat-
sistant recurrence associated with the disease. Primary ments.Inrecentyears,chemotherapeuticagentcarbopla-
tin has proved successful in eliminating primary
malignancy while reducing side effects for the patient
*Correspondence:[email protected]
1DepartmentofHistopathology,UniversityofDublin,TrinityCollege.Trinity [Reviewedin[3]]. Mechanistically,platinum-based drugs
CentreforHealthSciences,StJames’Hospital,Dublin8,Ireland bind nucleotides within the DNA backbone, causing
Fulllistofauthorinformationisavailableattheendofthearticle
©2012Gallagheretal;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative
CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and
reproductioninanymedium,providedtheoriginalworkisproperlycited.
Gallagheretal.JournalofOvarianResearch2012,5:2 Page2of11
http://www.ovarianresearch.com/content/5/1/2
cross-linking. In response, cells activate DNA-repair cells with pluripotent and nullipotent EC cells can
mechanismsthatultimatelyresultinapoptosis.Today,the establish mechanisms required for functional malignant
majorityofprimaryovarianmalignanciesaresuccessfully differentiation. The cells are so similar that EC cells are
treated, where upto 80% ofwomenwillrecover [2].The used as an easily cultured model of ES biology, reflect-
remaining 20% may be explained by late presentation of ing the difficulty of targeting CSCs without damaging
the disease by asymptomatic women. Alarmingly, up to non-malignant stem cell populations [16-18].
80%ofthesesurvivorswilldevelopchemoresistantterm- In this study we first used gene microarrays to assess
inalrecurrentdiseasewithintwoyears,whichisaccepted upstream regulation of differentiation in murine EC
as the main factor in fatality rates. We have previously (mEC) and mES cells. Our analysis describes aberrant
usedcomparativemicroarrayanalysistodemonstratethat regulation of differentiation in EC cells. Subsequently,
primaryandrecurrentdiseasehavesubstantiallydifferent we compared mEC genelists to our previously published
gene and microRNA (miRNA) expression profiles [4,5], primary versus recurrent tumor sample data [5]. We
whichwecontinueinthisstudy. described the presence of a cancer stemness p53-p21
Current treatment of recurrent disease, which is simi- regulatory mechanism in ovarian tumor samples. This
lar to treatment of primary disease, has proved ineffec- mechanism is employed by primary disease and sup-
tive. Thus, recurrent disease must be fully characterised pressed in recurrent disease. Subsequently, we con-
and novel therapeutic approaches developed. One such ducted a meta-analysis of our previously published
approach involves targeting cancer cells with stemness human EC (hEC) and tumor sample miRNA data [15,6].
properties. These cancer stem cells (CSCs) have been We report that cancer stemness signature miRNAs are
described in ovarian cancer [Reviewed in [6]] and have more relevant to ovarian cancer than cancer stemness
several properties with relevance to recurrent ovarian signature genes. We detail substantial recruitment of
cancer. CSCs are sufficient to regenerate malignancy in stemness signature miRNAs by recurrent disease. Thus
vivo via extensive self-renewal and differentiation. recurrent tumors suppress and activate stemness signa-
Tumor regeneration from CSCs is remarkably efficient, ture genes and miRNAs respectively. Our analysis indi-
where a single CSC is often sufficient to re-establish dis- cates that cancer stemness mechanisms are specifically
ease [7,8]. CSCs proliferate well in the hypoxic condi- and differentially regulated in primary and recurrent
tions found in the tumor microenvironment [9,10]. As ovarian malignancy, with obvious implications for
they differentiate, CSCs quickly develop neo-vasculature treatment.
to fuel further tumorigenesis. Perhaps the most alarming
aspect of CSCs is their uninhibited proliferation in the Methods
presence of chemotherapeutic agents. It is broadly Cell Culture
accepted that CSCs play a role in most, if not all, pri- Murine ES (ES-E14TG2a) and EC cells (pluripotent
mary malignancies. Theoretically, the persistence of a ‘SCC-PSA1’ and nullipotent ‘Nulli-SCC1’) were pur-
single CSC post-intervention could be sufficient to chased from ATCC, cultured on murine irradiated fibro-
explain chemoresistant recurrence. However, the role of blasts in DMEM supplemented with 10% foetal bovine
CSCs in recurrent ovarian disease is poorly understood. serum, 4 mM L-glutamine (Invitrogen) and 100 U/ml of
Ultimately we must develop methods of targeting speci- penicillin/streptomycin (Invitrogen Corporation, Carls-
fic CSC populations as part of a combined anti-cancer bad, CA, USA) and spontaneously differentiated via
strategy. removal of feeder layer. Human EC cells were retinoic
Many studies have demonstrated the presence of CSCs acid-differentiated as previously described [15].
in ovarian malignancy [6]. However, establishing ovarian
CSC models in culture has proved challenging. In this Tumor Samples
study we employed an embryonal carcinoma (EC) model Tumor sample data was previously published [5,6].
of cancer stemness. Originally derived from malignant Briefly, two cohorts of primary and recurrent samples
teratomas that can develop in the ovary, EC cells are the were assessed. Cohort 1 contained 5 primary and recur-
original and best characterised CSC model [11-14]. We rent serous papillary adenocarcinomas (Grade 3). Cohort
have previously shown high relevance between EC cells 2 contained 3 paired ovarian cancers from the same
and ovarian serous carcinoma patient samples at the patient but with different histologies: papillary serous,
miRNA level [15]. Pluripotent EC cells can differentiate mixed mullerian and clear cell carcinomas.
into cells representing all three germ layers and are con-
sidered the malignant equivalent of embryonic stem Microarray Analysis
(ES) cells [11-14]. Nullipotent EC cells can avoid differ- RNA was isolated using the RNeasy kit (Qiagen, West
entiation in vivo to generate poorly-differentiated, Sussex, UK) as per manufacturer’s protocol. Digoxi-
highly-malignant tumors [11-14]. Comparison of ES genin-UTP labelled cRNA was synthesized via the
Gallagheretal.JournalofOvarianResearch2012,5:2 Page3of11
http://www.ovarianresearch.com/content/5/1/2
Chemiluminescent RT-IVT Labelling Kit v2.0 (Life Microarray data was validated through qPCR analysis,
Technologies, Foster City, CA, USA) and hybridized to showing excellent correlation (Figure 1). An overview of
Mouse Genome Survey arrays (Life Technologies) as per the number of differentially expressed genes in pluripo-
manufacturers’ instructions. Data was filtered to a sig- tent (SCC-PSA1) and nullipotent (Nulli-SCC) mEC and
nal/noise ration threshold > 3 in at least one sample mES cells is shown in Table 1. At cut-offs of 0.05 (p-
®
using R and further analysed using Spotfire (Life Tech- Value) and ± 2.0 (fold change) SCC-PSA1 cells alter the
nologies). Genelists were generated using cut-offs of expression of 724 genes: 202 upregulated and 522 down-
0.05 (p-Value) and ± 2.0 (fold change). Functional rela- regulated at fold change levels between +18 and -18
tionships were analysed using DAVID [19,20]. Pathways (Additional File 1). Top ten SCC-PSA1 genes are char-
associations of predicted targets of miRNAs highlighted acterised by receptor activity and growth and differentia-
were generated using DIANA miRPath [21] using cut- tion/development roles (Table 2). Noteworthy events
offs of ≥ 2 genes per pathway and p-value ≤ 0.05. include upregulation of apoptosis (Bcl)-related gene
Bid3 and downregulation of Cav2 tumor suppressor [24]
qPCR Analysis and metastasis-linked Nupr1 [25]. Functional relation-
2 μg total RNA was used to synthesis cDNA using the ship analysis identified upregulation of developmental
High Capacity cDNA Archive Kit (Life Technologies) as pathways and downregulation of transcription regulation
per manufacturer’s instructions. Microarrays were vali- processes and Toll-Like Receptor (TLR), Interleukin-2
dated using 36 pre-designed TaqMan assays (Life Tech- and cancer pathways (Additional File 1).
nologies). Gene expression values were generated using Nulli-SCC cells responded to differentiation stimuli
the 2^-ddCt method [22]. microRNA was isolated using through the upregulation of 185 and downregulation of
the mirVANA kit (Ambion) and miRNA TaqMan qPCR 152 genes at levels from -6.3 to 14.0 fold (Additional
(Life Technologies) [23] analysis carried out as pre- File 2). Top ten genes included signal transducers and
viously described [5,6,15]. Data plotted represents the regulators of development/differentiation and malig-
mean value across a minimum of n = 3. Error bars nancy (Table 2). Notable genes include hypoxia and
represent standard error of the mean. tumor growth regulator Loxl2 [26] and tumor suppres-
sor Serpini2 [27]. Interestingly Ssa2 is downregulated, a
Results gene that is commonly expressed on the surface of
Microarray analysis of early mEC and mES differentiation apoptotic cells. Functional analysis identified upregula-
It is well established that ES and EC cells express similar tion of signal transduction regulators and downregula-
gene profiles in the undifferentiated and well-differen- tion of growth regulators (Additional File 2).
tiated (one week or later) states [16,11,17,18]. In con- Upstream differentiation of mES cells is characterized
trast, our understanding of the earlier, upstream by substantial levels of upregulation: 554 upregulated
regulation of differentiation is sparse. We hypothesized and 832 downregulated genes at levels of 232 to -68
that comparison of early differentiation of mES and fold (Additional File 3). Top ten genes are populated
mEC cells would identify cancer-specific differences in with receptors and developmental regulators (Table 2).
upstream regulation of stem cell differentiation. Addres- Tll1 is linked to cardiac development, the first organised
sing this we used microarray analysis to assay early system formed during embryogenesis. Notably, a key
(three day) differentiation of mES and mEC cells. RNAi gene, Eif2c4, is upregulated during differentiation,
Figure1Validation of mEC and mES microarraydata. Microarray data wasvalidatedthrough qPCR(TaqMan)analysisof agroupof 36
genes.Datapresentedrepresentsgeneexpressionchangeindifferentiatedcellscomparedtoundifferentiatedandshowsgoodcorrelationfor
mES(A)andPSA-SCC1(B)datasets.
Gallagheretal.JournalofOvarianResearch2012,5:2 Page4of11
http://www.ovarianresearch.com/content/5/1/2
Table 1Anoverview ofthe numbers andpercentage Table 2 Top ten genes differentially expressed (D/U) dur-
overlap ofdifferentially expressed genes(D/U) during ing early differentiation of mES and mEC cells.
early differentiation ofmES andmEC cells. (Continued)
CellType GeneNumber %Overlap Kcnk4 Kchannel,subfamilyK,member4 -4.6
Upreg Downreg mES SCC-PSA1 Nulli-SCC Cpne2 CopineII -4.6
mES 554 832 10 <1 Nulli-SCC
SCC-PSA1 202 522 33 <1 Upregulated
Nulli-SCC 185 152 <1 <1 Flt1 FMS-liketyrosinekinase1 5.3
Npy5r NeuropeptideYreceptorY5 5.0
Olfr786 Olfactoryreceptor786 4.4
perhaps reflective of involvement of the RISC complex
Fau FBR-MuSVubiquitouslyexpressed 4.1
[28]. Upregulated mES genes regulate development, sig-
Gm392 Genemodel392,(NCBI) 3.7
nalling and gene expression while downregulated genes
Gm449 Genemodel449,(NCBI) 3.2
regulate morphogenesis, particularly growth factor bind-
Loxl2 Lysyloxidase-like2 2.9
ing. Stemness-linked pathways such as Wnt-catenin and
Aoc3 Amineoxidase,coppercontaining3 2.8
Hedgehog signalling were upregulated while signalling
Serpini2 Serineproteinaseinhibitor1 2.8
pathways including TLR and TGF-ß were downregu-
Olfr870 Olfactoryreceptor870 2.7
lated (Additional File 3).
Downregulated
Ssa2 SjogrensyndromeantigenA2 -5.3
Aberrant upstream regulation of differentiation in mEC
4930486G11Rik RIKENcDNA4930486G11gene -4.6
cells
1700052K11Rik RIKENcDNA1700052K11gene -4.5
A comparison of mES and mEC early differentiation
Refbp2 RNAandexportfactorbindingprotein2 -4.4
genelists is summarised in Table 1 and detailed in addi-
2900011O08Rik RIKENcDNA2900011O08gene -3.7
tional files 1, 2 and 3. In contrast to documented undif-
Tmem62 Transmembraneprotein62 -3.7
ferentiated and well-differentiated comparisons, 90% of
Tirap TIRdomain-containingadaptorprotein -3.6
the mES genelist differed to the mEC genelist at this
Defb13 Defensinbeta13 -3.5
Es31 Esterase31 -3.4
Table 2Top ten genesdifferentially expressed (D/U) Nap1l5 Nucleosomeassemblyprotein1-like5 -3.3
during early differentiation ofmES andmEC cells. mES
GeneSymbol GeneName Fold Upregulated
Change H1foo H1histonefamily,memberO,oocyte- 64.2
SCC-PSA1 specific
Upregulated Mrgprh MAS-relatedGPR,memberH 60.8
Olfr1450 Olfactoryreceptor1450 9.0 Mak10 MAK10homolog 59.6
Fgf5 Fibroblastgrowthfactor5 8.9 V1rd11 Vomeronasal1receptor,D11 57.8
Dscam Downsyndromecelladhesionmolecule 7.7 B230317F23Rik RIKENcDNAB230317F23gene 41.1
Afp Alphafetoprotein 5.2 Gdpd3 Glycerophosphodi-phosphodiesterase3 39.2
Rbp4 Retinolbindingprotein4,plasma 5.0 Na Genemodel979,(NCBI) 32.3
Slc28a2 Solutecarrierfamily28,member2 4.5 Eif2c4 Euktranslationinitiationfactor2C,4 30.6
Bid3 BH3interactingdomain,apoptosis 4.3 Tll1 Tolloid-like 22.9
agonist Na SimilartoIggamma-2achainprecursor 20.1
Igfbp5 Insulin-likegrowthfactorbindingprotein 4.2 Downregulated
5
MP4 ProlinerichproteinMP4 -70
Irs4 Insulinreceptorsubstrate4 3.8
Pck1 Phosphoenolpyruvatecarboxykinase1 -49.3
Downregulated
Fgfrl1 Fibroblastgrowthfactorreceptor-like1 -38.3
Clec2d C-typelectindomainfamily2,memberd -11.3
Rpgrip1l Rpgrip1-like -20.6
Fbln5 Fibulin5 -6.5
Olfr508 Olfactoryreceptor508 -16.0
Cav2 Caveolin2 -5.6
Eif5a2 Euktranslationinitiationfactor5A2 -15.6
Irf5 Interferonregulatoryfactor5 -5.5
9130015A21Rik RIKENcDNA9130015A21gene -14.6
Lsp1 Lymphocytespecific1 -5.1
Tecta Tectorinalpha -11.0
Olfr787 Olfactoryreceptor787 -4.9
Fancc Fanconianemia,complementationgroup -10.7
Fxyd4 FXYD-containingiontransportregulator -4.8 C
4
Pax9 Pairedboxgene9 -9.9
Nupr1 Nuclearprotein1 -4.7
Gallagheretal.JournalofOvarianResearch2012,5:2 Page5of11
http://www.ovarianresearch.com/content/5/1/2
earlier time point (Table 1). Similarly, almost 70% of the Table 3Percentage gene expression ofmEC-specific
SCC-PSA1 genelist differed from the mES genelist genesexpressed in primary versusrecurrent tumor
(Table 1). Functional relationship analysis indicates that samples(Group A expressedsimilarly in primaryand
quite different mechanisms are activated during early recurrent samples).
differentiation of mEC and mES cells. This included %GeneExpression %GeneExpression
mES-specific upregulation of p53 signaling pathway GeneName Tumor SCC-PSA1 GeneName Tumor Nulli-SCC
genes (Additional File 3). There is very little overlap (P/R) (U/D) (P/R) (U/D)
between Nulli-SCC and the other cell types (Additional GroupA GroupA
Files 1, 2 and 3). Only four genes are upregulated by Dusp26 325.2 Egln3 259.9
SCC-PSA1 and downregulated by Nulli-SCC cells, while Hsf2 222.5 Ndufab1 245.4
only two are downregulated by both cell types. The Pdzk1 212.8 Gpr6 229.6
downregulation of symporters, signal-transducing mem- Sdsl 226.2 Ltbr 212.2
brane proteins, which are upregulated by pluripotent Ndufs6 -203.1 Golga5 -208.5
cells, may indicate a potential counteraction of differen- Sox4 -208.1 Slc15a1 -214.6
tiation. Upstream regulation of differentiation represents GroupB Gpatc3 -236.7
a substantial difference between these cell types, sup- Itgbl1 432.6 -396.1 Dgcr8 -323.3
porting our hypothesis. While similar genes maintain Kcnmb4 397.0 267.5 Tirap -364.3
the self-renewal state in each cell, different mechanisms Nkx3-1 375.2 260.0 GroupB
are employed to regulate the early events in Tmprss2 368.3 373.0 Cask 442.1 -220.5
differentiation. Cthrc1 334.6 298.7 Stau2 292.9 -201.6
Pdgfc 283.2 -392.8 Bnip3 283.4 217.4
A SCC-PSA1 p53 mechanism is expressed in primary and Plaur 259.9 -263.2 Pfkp 239.4 251.0
maintained in recurrent tumors Hoxb2 239.3 -375.5 Pak6 202.4 -213.8
We have previously published microarray analysis of pri- Col4a5 210.8 -320.7
mary versus recurrent tumor samples [5,6]. The study Gata2 -276.5 -243.0
contained two cohorts. Cohort 1 represents a group of
matched primary and recurrent tumors while cohort 2
and B) those expressed in primary but not recurrent
represents primary and recurrent tumor samples from
tumors (Table 3). Many of these genes have links to
the same patients. In this study, raw microarray data
stemness and malignancy. Tmprss2 is a transmembrane
from the primary versus recurrent study was reanalysed
signalling protein that is upregulated in prostate cancer
in an identical fashion to mES and mEC data described
[29]. Cthrc1 is a Smad2/3 (TGF-b signalling) inhibiting
above (Additional File 4). Primary versus recurrent dis-
Wnt signalling modulator that is differentially expressed
ease and mEC genelists were then compared. Genes
in invasive breast cancer and several solid tumors [30].
altered similarly in mEC and mES data were not consid-
Nkx3-1 is a metastatic marker transcription factor
ered to be cancer-specific and were removed from this
expressed in prostate cancer [31]. Pdgfc is a cisplatin-
analysis. Comparison of mEC and tumor data identified
associated growth factor [32]. Col4a5 is linked to several
16 SCC-PSA1 genes expressed in tumor samples
cancers while Plaur is a regulator of tissue reorganisa-
(Figure 2, Table 3). These genes group into those that
tion [33]. Ndufs6 is an oxidative phosphorylation
are A) expressed in both primary and recurrent tumors
Figure2ExpressionofSCC-PSA1stemnesssignaturemRNAsinrecurrenttumors.Sixteengeneswereidentifiedasdifferentiallyexpressed
inSCC-PSA1mECcellsandintumordata.Dataispresentedasthepercentagechangeingeneexpressionindifferentiatedcomparedto
undifferentiatedSCC-PSA1cells(blue)andinprimarytumorsamplescomparedtorecurrent(red).GeneExpressionvaluesaredetailedintable3
(Microarraydatap-Value≤0.05).
Gallagheretal.JournalofOvarianResearch2012,5:2 Page6of11
http://www.ovarianresearch.com/content/5/1/2
enzyme linked to cervical cancer [34]. Pdzk1 is linked to interactingkinasethatisarequiredforchemoresistancein
oestrogen-sensitivity in breast and ovarian cancers [35]. prostatecells[52].Collectively,anECcancerstemnesssig-
Sdsl is a cancer-specific metabolic enzyme [36]. Only natureexpressedintumorsamplesislinkedtomaintained
one gene, Gata2, an endodermal differentiation marker p53regulationandsuppressionofp53’smaintarget,p21,
[37] was upregulated by mEC cells and expressed higher inrecurrentdisease.
in primary tumors than in recurrent.
When scrutinised, we noted that several of the genes Recruitment of cancer stemness signature miRNAs during
highlighted above have been defined as p53 regulators recurrence
in various models, as now described. Dusp26 is a p53- Having identifiedgenelevel overlaps, we next conducted
inhibiting phosphatase that negatively regulates prolif- overlapmeta-analysisofourpreviouslypublishedmiRNA
eration of epithelial cells [38]. Stemness gene Sox4 is a data for primary and recurrent patient samples and
p16 and p53 regulator in cancer cells [39] while Hsf2 is humanEC(hEC)early(threeday)differentiation[5,6,15].
a regulator of p53 stability [40]. Hoxb2 has been linked The earlier study identified cancer stemness signature
to p205 regulation of p53 and is a well known regulator miRNAs:thosemiRNAsinvolvedinthedifferentiationof
of EC differentiation [41]. Collectively, our analysis indi- hEC cells. Specifically, our previous tumor study high-
cates that both primary and recurrent ovarian tumors lighted60miRNAs(52upand8downregulated)inrecur-
express this ‘p53-regulating stemness signature’. rentdisease[6].Ofthese,55miRNAs(92%)areexpressed
inhECcells(AdditionalFile5).21recurrentdisease-speci-
A NULLI-SCC p21 mechanism is suppressed by recurrent fic miRNAs are linked to differentiation of pluripotent
tumors NTera2 hEC cells (Figure 4). We have previously shown
Despitethereducedgenelistsize,14Nulli-SCCgeneswere thatnullipotent2102EphECcellsexpressalargenumber
expressedinA)bothprimaryandrecurrenttumorsorB) ofmiRNAsatsubstantiallyhigherlevelsthanNTera2cells
primary tumors only (Figure 3, Table 3). These genes [15].Herewereportthat26(43%)recurrentdisease-speci-
related to apoptosis/cellular proliferation, signaling and ficmiRNAsareexpressedathigherlevelsin2102Epcells
regulation. Dgcr8 (Pasha) is a key miRNA biosynthesis thaninNTera2(Figure5).Thus,developmentofrecurrent
gene[42],whileTirapisaregulatorofTLRsignaling[43]. tumorsinvolvesrecruitmentofcancerstemnesssignature
TNF-family related Ltbr and hypoxia-linked Egln3 are miRNAs. Specific examples include miR-9, which is the
apoptosisregulators[44,45].Gpr6isadevelopmentregu- mostdownregulatedmiRNAinrecurrenttumorsandis>
lator expressed inumbilical cord cells [46]. Ndufab1 is a 1000% higher expressed inundifferentiated 2102Ep cells
TGF-bsignalingrelatedNADPHenzyme[47].Slc15a1is compared to NTera2, and miR-206, which is in the top
involvedindrugabsorptioninthesmallintestineandhas tenmiRNAsupregulated byrecurrenttumorsanddown-
been linked to several cancers and metastasis [48]. regulatedduringNTera2differentiation.Molecularpath-
Coupledwiththisistherecurrentsuppressionofapopto- way relationships between predicted gene targets of the
sis regulatorsBnip3 and Stau2[49,50]. Notably, two p21 miRNAs highlighted were identified using DIANAmir-
regulators are expressed higher in primary tumors com- PATH(Additionalfile5).Whilelittlepathwayoverlapwas
paredtorecurrent:Caskmediatestheexpressionofp21to observedin gene array data, miRNA data showed strong
control cell proliferation [51] while Pak6 is a p21 pathway associations. Pathway analysis highlighted
Figure3ExpressionofNulli-SCCstemnesssignaturemRNAsinrecurrenttumors.Fourteengeneswereidentifiedasdifferentiallyexpressed
inNulli-SCCmECcellsandintumordata.Dataispresentedasthepercentagechangeingeneexpressionindifferentiation-stimulatedcompared
toundifferentiatedNulli-SCCcells(blue)andinprimarytumorsamplescomparedtorecurrent(red).GeneExpressionvaluesaredetailedintable
3(Microarraydatap-Value≤0.05).
Gallagheretal.JournalofOvarianResearch2012,5:2 Page7of11
http://www.ovarianresearch.com/content/5/1/2
Figure 4 Expression of NTera2 stemness signature miRNAs in recurrent tumors.Twenty one miRNAs wereidentified as differentially
expressedindifferentiatedNTera2hECcellsandintumordata.DataispresentedasthepercentagechangeinmiRNAexpressionin
differentiatedNTera2hECcells(blue)andinrecurrenttumorscomparedtoprimary(red).Valuesrepresentthemeanofatleastn=3anderror
barsthestandarderrorofthemean.
Figure 5 Expression of 2102Ep stemness signature miRNAs in recurrent tumors. Twenty six miRNAs were identified as differentially
expressedinundifferentiated2102EpcomparedtoundifferentiatedNTera2hECcellsandintumordata.Dataispresentedaslog (fold
10
change).miRNAspresentedshowedalteredexpressioninundifferentiated2102EpcellscomparedtoundifferentiatedNTera2(blue)andin
recurrenttumorscomparedtoprimary(red).Valuesrepresentthemeanofatleastn=3anderrorbarsthestandarderrorofthemean.
Gallagheretal.JournalofOvarianResearch2012,5:2 Page8of11
http://www.ovarianresearch.com/content/5/1/2
alteration of several cancer pathways (miRs-10b, -100, related to p53-p21 regulation. In parallel, recurrent
-106, -107, -128 and let-7g) as well as Wnt and TGF-b tumors recruit a population of miRNAs with close links
stemnesssignalingpathways(mirs-10b,-100,-106,-128a tothedevelopmentofhighlymalignant,poorly-differen-
and 137). Finally, we assessed the expression of p53-p21 tiatedtumorsfromnullipotenthECcells.
regulatingmiRNAsinthesedatasets.TwomiRNAs,miRs- Different genetic profiles are employed by primary and
106aandb,arevalidatedtargetsofp21[52]thatareupre- recurrent ovarian tumors [5,6]. In this study we demon-
gulated in recurrent disease and expressed in hEC cells. strate that malignant stem cell differentiation genes are
Notably, miR-106b expression in 2102Ep cells is double expressed in either primary tumors or both primary and
that ofNTera2 cells. Incontrast, miR-155, the only vali- recurrent tumors but essentially never in recurrent
dated p53-regulating miRNA, is unaltered in recurrent tumors-specifically. Some CSC mechanisms are similarly
tumors.Wenotethatthep53signalingpathwaywashigh- employed in primary and recurrent tumorigenesis. In
lightedforlet-7gandmiRs-106band-107inpathwayana- addition, an obvious implication of our study is that
lysis(Additionalfile5).Inoverview,wefindthatmiRNAs CSCs that survive chemotherapy to repopulate recurrent
linked to 2102Ep malignancy are highly relevant to pri- disease can do so using different mechanisms than those
maryandrecurrenttumors. employed in primary disease. Functional relationship
analysis indicated that these stemness signature genes
Discussion have a particular relevance to cellular proliferation and
AlthoughCSCsareobvioussuspectsinthedevelopment apoptosis. Several of the genes highlighted are known
ofrecurrentovarianmalignancy,arelationshiphasyetto ‘p53-p21 signaling regulators’. Mechanistically this
beestablishedordescribedindetail.Anecdotal evidence relates to regulation of p53-p21 processes, where p53
includes altered regulation of Notch3 in chemoresistant regulation is enhanced and p21 regulation no longer
ovariandisease andthe clearparallel betweenepithelial- required in recurrent tumors. This is supported by
mesenchymal transition (EMT) and CSC differentiation increased expression of p21 repressing miRNAs in
mechanisms[53,54].Inthisstudyweconductedmicroar- recurrent tumors and strong predicted targeting of p53
ray and meta-analysis of mRNA and miRNA expression signaling genes by tumor-specific miRNAs. Altered p53-
in primary and recurrent tumor samples and an EC p21 regulation is the primary mechanism through which
model of cancer stemness. Our analysis reiterates that cancers avoid apoptosis and stimulate cellular prolifera-
development of primary and recurrent ovarian disease tion. Predictably, we did not find loss of p53 or p21 in
involvesquitedifferentmechanisms:thousandsofgenes recurrent disease (data not shown). It appears that p53-
are differentially expressed. At the gene level, recurrent p21 regulation is required at both stages of ovarian
tumors appear to repress a cancer stemness signature malignancy. In Figure 6 we present a schematic to
Figure6Suppressionofcancerstemnessp53-p21regulationinrecurrenttumors.Primarydiseaseischaracterisedbytheexpressionof
p53-regulatingstemnesssignaturegenesSox4andSdsl,whichiscontinuedandenhancedwithDusp26expressionintherecurrency.In
contrast,p21regulatingstemnesssignaturegenesPak6andCaskareexpressedinprimarydiseaseandsuppressedinrecurrency.Thisprocessis
paralleledbyrecruitmentofstemnesssignaturemiRNAsbyrecurrentdisease.
Gallagheretal.JournalofOvarianResearch2012,5:2 Page9of11
http://www.ovarianresearch.com/content/5/1/2
illustrate the p53-p21 regulators highlighted in out Recurrenttumordevelopmentinvolvesthesuppression
study. We propose that these genes and miRNAs regu- of twice as many genes as are specifically activated
late p53-p21 signaling, at least partially, in primary and (Cohort1).Thisindicatesthatrecurrentmalignancydoes
recurrent disease (Figure 6). Indeed, this is likely to be a not require a substantial number of mechanisms
component of a larger mechanism. This p53-p21 regu- employed by primary tumors. Specifically, angiogenesis
lating component appears to play a role in primary and development genes are turned off by recurrent dis-
tumors that is not used during recurrence. We refer to ease as malignancy genes are turned on. The upregula-
this as a p53-p21 regulating mechanism within the can- tion of polycystic ovary-associated gene Fabp4 and
cer stemness signature (genes altered during differentia- ovariancancergenePrkcbp1maybeofparticularimpor-
tion of EC cells but not by ES cells). As a key tance. There was little overlap between genes altered in
tumorigenesis component, differential regulation of cohort 1 and cohort 2, which altered genes more asso-
stemness-linked p53-p21 mechanisms in primary and ciated with malignancy and less with differentiation.
recurrent disease is an important outcome of this study Functional relationship analysis revealed that recurrent
and will be the subject of ongoing analysis. disease no longer requires homeostasis or stimulus
It is well established that EC and ES cells are highly response processes while upregulating catalytic activity
similar in the undifferentiated and well-differentiated andproteinbindingprocess.Ingeneral,recurrentdisease
states [16-18]. This illustrates the significant challenges behavesmoreasadevelopingcancerratherthantheche-
to the concept of targeting CSCs in a manner that micalstressresponsesrequiredbyprimarydisease.
does not harm the non-malignant stem cell pool. In
this study we have identified upstream regulation of Conclusion
differentiation as a substantial difference between EC CSCstargetingisapotentialavenuethroughwhichtreat-
and ES cells, supporting our hypothesis. While down- mentofrecurrent,chemoresistantovariancancermaybe
regulated mEC and mES genes displayed similarity, improved.Thisiscomplicatedbythesimilaritiesbetween
upregulated SCC-PSA1 genes were almost 90% specific cancer and non-cancer stem cells and our poor under-
to malignancy. This supports a model where normal standingofrecurrentovariandisease.Wehaveidentified
and malignant stem cells employ similar mechanisms the early eventsof stem cell differentiationas a key area
to maintain the self-renewal state. The different phe- ofdifference betweencancer and non-cancerstemcells.
notypes developing from differentiation, therefore, are Furthermore, we have highlighted the association of a
related to activation of specific malignant or non- p53-p21relatedcancerstemnesssignaturewithinovarian
malignant genes. Both cell types alter genes related to disease. Our data suggests that a stem cell involved in
similar processes: receptor-mediated signalling of development of recurrent disease employs different
development/differentiation. Thus the differentiation of mechanismsoftumorigenesis.Ourstudysuggeststhatit
malignant and non-malignant cells is driven by a diver- may be possible to target early differentiation events in
gent group of genes. It is noteworthy that the primary- CSCs without damaging non-cancer stem cells, which
recurrent genetic switch contained an equally strong wouldhave broad implications for treatments. Our data
Nulli-SCC cell signature, despite the much reduced indicatesthatsuchtherapiesshouldbeindependentlytai-
genelist. Nulli-SCC cells avoid differentiation through loredforprimaryandrecurrentovariandisease.CSCtar-
maintained levels of gene and miRNA expression to geting during treatment of primary disease is likely to
generate highly malignant tumors [11]. While a small have anegativeimpact onrecurrent tumorigenesis.CSC
number of molecular events take place in these cells targeting in recurrent disease should be developed with
response to differentiation, these appear to have a par- considerationtoindependentmechanisms.Development
ticular relevance to the difference between primary and ofstrategiestoachievethiswillcontinueinourgroup.
recurrent disease. Stemness genes are never expressed
by recurrent disease only, suggesting a less stem-like Additional material
profile. These genes have a particular relevance to cel-
lular proliferation and apoptosis, including p53-p21 Additionalfile1:Microarrayandfunctionalrelationshipanalysisof
regulation. Of particular note is the downregulation in earlydifferentiationresponseofSCC-PSA1cells.Microarrayand
functionalrelationshipanalysisofearlydifferentiationresponseof
Nulli-SCC cells of TLR signaling adapter Tirap, a gene
SCC-PSA1cells.Thisfilecontainsgenelistsofgenesupregulatedand
that is constantly expressed in primary and recurrent downregulatedbySCC-PSA1cellsinresponsetodifferentiationstimulus.
disease. TLR signaling has received increased attention Thisfilealsocontainsresultsofanalysistoidentifyfunctional
relationshipsbetweenthesegenes.
in both cancer and stemness studies in recent years
Additionalfile2:Microarrayandfunctionalrelationshipanalysisof
[55]. In summary, recurrent disease appears to have
earlydifferentiationresponseofNULLI-SCCcells.Microarrayand
more correlation with nullipotent cells rather than EC functionalrelationshipanalysisofearlydifferentiationresponseof
cells with good differential potential. NULLI-SCCcells.Thisfilecontainsgenelistsofgenesupregulatedand
Gallagheretal.JournalofOvarianResearch2012,5:2 Page10of11
http://www.ovarianresearch.com/content/5/1/2
5. LaiosA,O’TooleS,FlavinR,MartinC,KellyL,RingM,FinnSP,BarrettC,
downregulatedbyNULLI-SCCcellsinresponsetodifferentiationstimulus. LodaM,GleesonN,D’ArcyT,McGuinnessE,SheilsO,SheppardB,O’
Thisfilealsocontainsresultsofanalysistoidentifyfunctional
LearyJ:PotentialroleofmiR-9andmiR-223inrecurrentovariancancer.
relationshipsbetweenthesegenes.
Moleccancer2008,7:35.
Additionalfile3:Microarrayandfunctionalrelationshipanalysisof 6. BerryNB,BapatSA:Ovariancancerplasticityandepigenomicsinthe
earlydifferentiationresponseofmEScells.Microarrayand acquisitionofastem-likephenotype.JOvRes2008,1:8.
functionalrelationshipanalysisofearlydifferentiationresponseof 7. KleinsmithLJ,PierceGB:Multipotentialityofsingleembryonalcarcinoma
mEScells.Thisfilecontainsgenelistsofgenesupregulatedand cells.CancerRes1964,24:1544-1552.
downregulatedbymEScellsinresponsetodifferentiationstimulus.This 8. Al-HajjM,WichaMS,Benito-HernandezA,MorrisonSJ,ClarkeMF:
filealsocontainsresultsofanalysistoidentifyfunctionalrelationships Prospectiveidentificationoftumorigenicbreastcancercells.PNAS,USA
betweenthesegenes. 2003,100:3983-3988.
Additionalfile4:Microarrayandfunctionalrelationshipanalysisof 9. SilvanU,Diez-TorreA,ArluzeaA,AndradeR,SilioM,ArechagaJ:Hypoxia
primaryversusrecurrenttumorsamples.Microarrayandfunctional andpluripotencyinembryonicandembryonalcarcinomastemcell
relationshipanalysisofprimaryversusrecurrenttumorsamples. biology.Differentiation2009,78:159-168.
Thisfilecontainsgenelistsofgenesupregulatedanddownregulatedin 10. KeithB,SimonSC:Hypoxia-induciblefactors,stemcellsandcancer.Cell
primaryversusrecurrenttumorsamples.Thisfilealsocontainsresultsof 2007,129:465-472.
analysistoidentifyfunctionalrelationshipsbetweenthesegenes. 11. AndrewsPW,MartinMM,BahramiAR,DamjaovI,GokhaleP,DraperJS:
Embryonicstem(ES)cellsandembryonalcarcinoma(EC)cells:opposite
Additionalfile5:ComparisonofmicroRNAexpressioninhECand
sidesofthesamecoin.BiochemicalSocietyTransactions2005,33(Pt
primaryversusrecurrenttumorsamples.ComparisonofmicroRNA
6):1526-30.
expressioninhECandprimaryversusrecurrenttumorsamples.This
12. AndrewsPW,FendersonB,HakomoriS:Humanembryonalcarcinoma
filedetailstherelativeexpressionpatternsandlevelsofmicroRNAsin
cellsandtheirdifferentiationinculture.Intjandrology1987,10(1):95-104.
hECcellsandprimaryversusrecurrenttumorsamplesandthepathway
13. AndrewsPW,DamjanovI,SimonD,BantingGS,CarlinC,DracopoliNC,
associationsofgenestargetedbythesemicroRNAs.
FoghJ:Pluripotentembryonalcarcinomaclonesderivedfromthe
humanteratocarcinomacelllineTera-2.Differentiationinvivoandin
vitro.LabInvest1984,50(2):147-162.
14. AndrewsPW:Fromteratocarcinomastoembryonicstemcells.PhilTrans
Abbreviations RSocLond,B2002,357:405-417.
CSC:cancerstemcell;EC:embryonalcarcinoma;h:human;mEC:murine 15. GallagherMF,FlavinRJ,ElbaruniSA,McInerneyJK,SmythPC,SalleyYM,
embryonalcarcinoma(cells);mES:murineembryonicstem(cells);miRNA: VenckenSF,O’TooleSA,LaiosA,LeeMY,DenningK,LiJ,AherneST,
microRNA;P/R:Primarycomparedtorecurrent;qPCR:quantitative LaoKQ,MartinCM,SheilsOM,O’LearyJJ:RegulationofmicroRNA
polymerasechainreaction;U/D:Undifferentiatedcomparedtodifferentiated biosynthesisandexpressionin2102Epembryonalcarcinomastemcells
ismirroredinovarianserousadenocarcinomapatients.Jovres2009,
Acknowledgements 2:19.
TheauthorswishtoacknowledgethesupportofCancerResearchIreland 16. JosephsonR,OrdingCJ,LiuY,ShinS,LakshmipathyU,ToumadjeA,LoveB,
andTheEmerCaseyFoundation. ChesnutJD,AndrewsPW,RaoMS,AuerbachJM:Qualificationof
embryonalcarcinoma2102Epasareferencelineforhumanembryonic
Authordetails stemcellresearch.Stemcells2007,25:427-446.
1DepartmentofHistopathology,UniversityofDublin,TrinityCollege.Trinity 17. SchwartzCM,SpivakCE,BakerSC,McDanielTK,LoringJF,NguyenC,
CentreforHealthSciences,StJames’Hospital,Dublin8,Ireland.2Molecular ChrestFJ,WerstoR,ArenasE,ZengX,FreedWJ,RaoMS:NTera2:amodel
PathologyResearchLaboratory,CoombeWomen’sandInfants’University systemtostudydopaminergicdifferentiationofhumanembryonicstem
Hospital,Dublin8,Ireland.3TheDepartmentofObstetricsandGynecology, cells.Stemcellsdev2005,14:517-534.
TrinityCentreforHealthSciences,StJames’Hospital,Dublin8,Ireland.4The 18. SkotheimRI,LindGE,MonniO,NeslandJM,AbelerVM,FossåSD,DualeN,
DepartmentofHistopathology,CorkUniversityHospital,Wilton,Cork,Ireland. BrunborgG,KallioniemiO,AndrewsPW,LotheRA:Differentiationof
humanembryonalcarcinomasinvitroandinvivorevealsexpression
Authors’contributions profilesrelevanttonormaldevelopment.Cancerres2005,65:5588-5560.
MFGdesignedandcarriedoutthecomparisonsandwasthemainauthorof 19. HuangDW,ShermanBT,LempickiRA:Systematicandintegrativeanalysis
themanuscript.CCBBH,MFGandPCScarriedoutmicroarrayexperiments oflargegenelistsusingDAVIDBioinformaticsResources.NatProtocols
andanalysis.BFconductedstatisticalanalysis.AL,SAO’T,RJF,SAE,CMM, 2009,4(1):44-57.
OMSwereinvolvedintheoriginalexperimentsandcontributedtodata 20. DennisGJr,ShermanBT,HosackDA,YangJ,GaoW,LaneHC,LempickiRA:
analysisandinterpretation.CDScontributedtomanuscriptdesign.JJO’L DAVID:DatabaseforAnnotation,Visualization,andIntegratedDiscovery.
oversawtheproject.Allauthorshavereadandapprovedthefinal GenomeBiol2003,4(5):P3.
manuscript. 21. PapadopoulosGL,AlexiouP,MaragkakisM,ReczkoM,HatzigeorgiouAG:
DIANA-mirPath:IntegratinghumanandmousemicroRNAsinpathways.
Competinginterests Bioinfor2009,10,(1093/bioinformatics/btp29).
Theauthorsdeclarethattheyhavenocompetinginterests. 22. LivakKJ,SchmittgenTD:Analysisofrelativegeneexpressiondatausing
real-timequantitativePCRandthe2-ddCtmethod.Methods2001,
Received:18October2011 Accepted:19January2012 25:402-8.
Published:19January2012 23. LaoK,XuNL,YeungV,ChenC,LivakKJ,StrausNA:Multi-plexingRT-PCR
forthedetectionofmultiplemiRNAspeciesinsmallsamples.BiocBioph
References ResComm2006,343:85-9.
1. JemalA,SiegelR,WardE,HaoY,XuJ,MurrayT,ThunMJ:Cancerstatistics, 24. HanSY,DruckT,HuebnerK:Candidatetumorsuppressorgenesat
2008.CA2008,58:71-96. FRAG7GarecoamplifiedwithMETanddonotsuppressmalignancyin
2. CannistraSA:Canceroftheovary.NEngljmed2004,351(24):2519-29. gastriccancer.Genomics2003,18(2):105-7.
3. MuggiaF:Platinumcompounds30yearsaftertheintroductionof 25. CanoCE,HamidiT,SandiMJ,IovannaJL:Nupr1:theswiss-knifeofcancer.
cisplatin:implicationsforthetreatmentofovariancancer.Gyneoncol Jcellphysiol2011,266(6):1439-43.
2009,112:275-81. 26. BoccaC,BozzoF,CannitoS,ParolaM,MigiliettaA:Celecoxibinactivates
4. LaiosA,O’TooleSA,FlavinR,MartinC,RingM,GleesonN,D’ArcyT, epithelial-mesenchymaltransitionstimulatedbyhypoxiaand/or
McGuinnessEP,SheilsO,SheppardBL,O’LearyJJ:Anintegrativemodel epidermalgrowthfactorincoloncancercells.Moleccarcinog2011,
forrecurrenceinovariancancer.Moleccancer2008,7:8. (ePub).