Metformin Targets the Metabolic Achilles Heel of Human Pancreatic Cancer Stem Cells

13 Pages·2013·3.02 MB·English

Metformin Targets the Metabolic Achilles Heel of Human
Pancreatic Cancer Stem Cells
Enza Lonardo1., Michele Cioffi1., Patricia Sancho1, Yolanda Sanchez-Ripoll1, Sara Maria Trabulo1,
Jorge Dorado1, Anamaria Balic1, Manuel Hidalgo2, Christopher Heeschen1*
1StemCells&CancerGroup,MolecularPathologyProgramme,SpanishNationalCancerResearchCentre(CNIO),Madrid,Spain,2GastrointestinalCancerClinicalResearch
Unit,ClinicalResearchProgramme(CNIO),SpanishNationalCancerResearchCentre,Madrid,Spain
Abstract
Pancreaticductaladenocarcinomascontainasubsetofexclusivelytumorigeniccancerstemcells(CSCs),whicharecapable
ofrepopulatingtheentireheterogeneouscancercellpopulationsandarehighlyresistanttostandardchemotherapy.Here
wedemonstratethatmetforminselectivelyablatedpancreaticCSCsasevidencedbydiminishedexpressionofpluripotency-
associated genes and CSC-associated surface markers. Subsequently, the ability of metformin-treated CSCs to clonally
expand in vitro was irreversibly abrogated by inducing apoptosis. In contrast, non-CSCs preferentially responded by cell
cyclearrest,butwerenoteliminatedbymetformintreatment.Mechanistically,metforminincreasedreactiveoxygenspecies
production inCSCand reduced theirmitochondrial transmembranepotential.Thesubsequent induction oflethalenergy
crisis in CSCs was independent of AMPK/mTOR. Finally, in primary cancer tissue xenograft models metformin effectively
reducedtumorburdenandpreventeddiseaseprogression;ifcombinedwithastroma-targetingsmoothenedinhibitorfor
enhancedtissuepenetration, while gemcitabineactually appeareddispensable.
Citation:LonardoE,CioffiM,SanchoP,Sanchez-RipollY,TrabuloSM,etal.(2013)MetforminTargetstheMetabolicAchillesHeelofHumanPancreaticCancer
StemCells.PLoSONE8(10):e76518.doi:10.1371/journal.pone.0076518
Editor:AnitaB.Hjelmeland,ClevelandClinic,UnitedStatesofAmerica
ReceivedMay3,2013;AcceptedSeptember1,2013;PublishedOctober18,2013
Copyright:(cid:2)2013Lonardoetal.Thisisanopen-accessarticledistributedunderthetermsoftheCreativeCommonsAttributionLicense,whichpermits
unrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalauthorandsourcearecredited.
Funding:TheresearchwassupportedbytheERCAdvancedInvestigatorGrant(Pa-CSC233460),EuropeanCommunity’sSeventhFrameworkProgramme(FP7/
2007–2013)undergrantagreementnu256974,theSubdireccio´nGeneraldeEvaluacio´nyFomentodelaInvestigacio´n,FondodeInvestigacio´nSanitaria(PS09/
02129),andtheProgramaNacionaldeInternacionalizacio´ndelaI+D,Subprogramma:FCCI2009(PLE2009–0105;bothMinisteriodeCienciaeInnovacio´n,Spain).
E.L.issupportedbytheRochePostdoctoralFellowshipProgram.M.C.issupportedbytheLaCaixaPredoctoralFellowshipProgram.Thefundershadnorolein
studydesign,datacollectionandanalysis,decisiontopublish,orpreparationofthemanuscript.
CompetingInterests:Theauthorshavedeclaredthatnocompetinginterestsexist.
*E-mail:[email protected]
.Theseauthorscontributedequallytothiswork.
Introduction metforminonAMPK/mTORsignalingresultsinreducedprotein
synthesis and cell proliferation [11,12]. Moreover, in established
Pancreatic ductal adenocarcinoma (PDAC) remains one of the PDAC cell lines metformin is also capable of inhibiting PDAC
most devastating cancers, and is the fourth leading cause of [13].Intriguingly,anotherrecentstudysuggestedthatCSCscould
cancer-relateddeathsinindustrialcountrieswitha5-yearsurvival betargetedbymetforminviare-expressionofmiRNAsimplicated
rate of less than 5% [1]. Many risk factors including smoking, indifferentiation,althoughthesedataarebasedonnon-validated
alcohol consumption, and chronic pancreatitis have been recog- cancer cellline-derived CSCs[14].
nized as potential risk factors for the development of PDAC [2].
Unlike the majority of differentiated cells within the tumor,
Epidemiologic studies also suggest that diabetes mellitus, partic-
CSCs have been shown to be highly resistant to chemotherapy
ularly type 2, is associated with enhanced risk for PDAC [3,4]. [15].Therefore,drugsthatselectivelytargetCSCsmayrepresent
Therefore,investigatorshaveembarkedonfindingaputativelink a more effective approach to overcome resistance and/or
between the use of anti-diabetic drugs and a reduced risk for the treatment relapse in PDAC. Here, we now provide compelling
development and/or progression of PDAC. Strikingly, in a evidencethatCSCsderivedfromadiversesetofprimaryhuman
retrospectiveanalysis,oraladministrationofmetformininpatients PDACs are highly vulnerable to metabolic reprogramming by
with diabetes mellitus type II was found to be associated with metformin resulting in long-term survival of preclinical mouse
reducedriskfordevelopingPDAC[5]aswellasbetteroutcomein models.
patients with establishedPDAC [6].
The primary systemic effect of metformin (Met) represents a
Results
decreaseinbloodglucoselevelsviareducedhepaticgluconeogen-
esis and increased glucose uptake in peripheral tissues [7]. We have previously shown that primary pancreatic CSCs can
Mechanistically, metformin indirectly activates AMP-activated be enriched in vitro as anchorage-independent three-dimensional
protein kinase (AMPK) signaling [8] and subsequently inhibits spheres,whichareenrichedforcellswithstemcell-likeproperties
mTOR activity, which is frequently increased in cancer cells [9] [15].AtotalnumberofninehumanPDACxenograftswereused
including pancreatic cancer stem cells (CSCs) as a highly with A6L, 163, 185, 215, 247, 253, and 286 being described
tumorigenic subpopulation [10]. This inhibitory effect of earlier [16,17] as well as 354 and JH029, which were obtained
PLOSONE | www.plosone.org 1 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
PLOSONE | www.plosone.org 2 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
Figure 1. Metformin targets pancreatic cancer stem cells. (A) Primary PDAC cells, but not normal pancreas cells express organic cation
transporter1,2,and3(n=3).(B)DefinitionofthetherapeuticrangeformetformininprimaryPDACcells.Numberofcellsgrowninthepresenceof
theindicatedconcentrationsofmetforminfor24h(n=6).(C)qPCRanalysisofCSCs-associatedgenesinspherestreatedwith3mMofmetforminfor
7days.Dataarenormalizedtothehousekeepinggeneandarepresentedasfoldchangeingeneexpressionrelativetocontrolcells(n=6).(D)
Representative Western blot illustrating reduced Oct4 protein expression in response to metformin treatment (n=3). (E) Representative flow
cytometry analysis for CSCs markers in spheres treated for 7 days with 3mM of metformin as compared to untreated spheres (upper panel).
SummaryofdataforPDAC-185,A6L,215,253,and354isshown(lowerpanel;n=6).
doi:10.1371/journal.pone.0076518.g001
Figure2.Metformindiminishesinvitroandinvivotumorigenicity.(A)Metformindecreasesthesizeofspheres.Representativeimagesof
spheresobtainedaftertreatmentwithmetforminfor7days.Quantificationofspheresize(n$6).(B)Sphereformationcapacityinthepresenceor
absenceofmetforminfor7days(n$6).(C)Self-renewalcapacityofcancerstemcellsisolatedfromtumorsrespondingpoorlyintermsoffirstpassage
sphere forming capacity. Cells were continuously passage as secondary and tertiary spheres treated with metformin or vehicle only during first
generationsphereformation(n=6).(D)ColonyformationforPDAC-185,A6L,215,253,and354evidencedby0.05%crystalvioletafter21days(n=3).
(E)Rareficationofinvivotumorigeniccancerstemcellsinspherestreatedwithmetforminascomparedtovehicle.(F)Invasionofsphere-derivedcells
after24hoftreatmentwithmetforminorcontrol(n=3).
doi:10.1371/journal.pone.0076518.g002
PLOSONE | www.plosone.org 3 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
Figure3.Metforminspecificallyeliminatescancerstemcells.(A)Numberofcellsgrowninthepresenceoftheindicatedconcentrationsof
metforminfor24h.(n=3).(B)QuantificationforKi67andDAPIinadherentcellsafter7doftreatmentwithmetforminorcontrol(n=3).(C)qPCR
PLOSONE | www.plosone.org 4 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
analysisforCyclinD1inadherentandsphere-derivedcellsafter7doftreatmentwithmetforminorcontrol.Dataarenormalizedtothehousekeeping
geneandarepresentedasfoldchangeingeneexpressionrelativetountreatedcells(n=6).(D)CellcycleanalysisdeterminedbyPropidiumIodide
staininginadherentcellsandspheresafter7doftreatmentwithmetforminorcontrol(n=3).(E)Cytometryanalysisofapoptoticcellsbydouble
stainingforAnnexinV/DAPIaftertreatmentwithmetforminorcontrolforadherentversussphere-derivedcells(n=3).
doi:10.1371/journal.pone.0076518.g003
usingthesamemethodology.Importantly,forinvitroexperiments proliferation (Fig. 3B), which was confirmed by reduced
all cells were freshly isolated from early passage xenografts and expression of cyclinD1 at mRNA (Fig. 3C) and protein level
cultured in low passages as adherent cells or anchorage- (Fig. S2B). Next, we analyzed the cell-cycle profile of adherent
independent spheres. Spheres are enriched in CD133+ cells cells versus CSCs by flow cytometry. Interestingly, the data
(Fig. S1A) and several other markers that have been associated showed that the functional effect of metformin on cell cycle
with a CSCphenotype ascompared toadherent cells [18]. progression was much more marginal for sphere-derived cells
suggesting a distinct mechanism responsible for their subsequent
Metformin decreases the expression of CSCs markers loss during extended treatment (Fig. 3D). Indeed, while metfor-
First, we established the expression of the organic cation mindidnotsignificantly altertherateofapoptoticadherentcells
transporter 1, 2, and 3 (OCT1-3) in our primary PDAC cells as evidenced by AnnexinV/Dapi staining (1.2360.37 fold
(Fig.1A),whicharecrucialforthecellular uptakeofmetformin. change), it resulted in a strong induction of apoptosis in sphere-
Metforminwasusedatconcentrationsthatarenotdirectlytoxicto derived cells (2.9561.10 fold change; P,0.01) (Fig. 3E). These
primary PDAC cells and non-transformed pancreatic cells (PSC, data are in line with a preferential elimination of CSCs by
pancreaticstellatecells;HDPE,humanductalpancreaticepithelial metformin,whereasitseffectsondifferentiatedprogenyaremostly
cells) (Fig. 1B), which are significantly lower as compared to related toinhibition ofproliferation.
concentrationsusedinpreviousstudieswithPDACcelllines(10–
30mM)[14].Next,wefoundsignificantchangesinmRNAlevels Mechanism of action
of CSCs genes (CD133, Alk4, Nodal, Activin and Smad2) and MetforminuniformlyreducedATPlevelbothinadherentcells
pluripotency-associated genes (Nanog, Oct4 and Sox2) following andspheresacrossthepanelofinvestigatedtumorsincludingtardy
treatment with metformin (Fig. 1C; utilized primers are listed in responders (Fig. 4A). Consistently, metformin induced a time-
Table S1), which were also confirmed at the protein level dependentincreaseofpAMPKinCSCs(Fig.4B,upperpanel)
(Fig. 1D). Strikingly, metformin appeared to preferentially with no apparent difference to non-CSC (data not shown).
eliminate CSCs as CSC-marker positive cells CD133, CD44, Moreover, we observed a similar net decrease in p70S6K,
CXCR4andSSEA-1declined,whiletheepithelialdifferentiation suggestinganefficientblockadeofthemTORpathway(Fig.4B,
maker EpCAMincreased (Fig.1E). lower panel). As the mTOR pathway is a central regulator of
autophagy, we analyzed whether metformin might induce
Metformin selectively diminishes in vitro and in vivo characteristic hallmarks of autophagy in CSCs versus non-CSCs,
tumorigenicity but our results do not support the notion that the selective
We next examined the functional effects of metformin on the elimination of CSCs by metformin can be rationalized by
self-renewalcapacityofCSCs.Thesphereformationassayshowed alterations in autophagy (Fig. S3A/B). To provide further
a strong decrease in the size of formed spheres by metformin evidence that AMPK/mTOR pathway is not mediating the
(Fig.2A&S1B)viainhibitionoftheexpansionoftheprogeniesof deleterious effects of metformin on CSCs, we investigated if the
CSCs,whichrepresentthebulkofthecellsoftheformedspheres. effects ofmetforminaremimickedbythedirect AMPKactivator
Even more importantly, we found that metformin significantly A769662 and/or the mTOR inhibitor rapamycin (Rapa).
decreased the number of actually formed spheres with the same Strikingly, neither sphere formation (Fig. 4C) nor colony
efficiency at 3 and 10mM (Fig. 2B). In order to examine the formation (Fig. 4D) was significantly inhibited by the two
long-term effect of metformin treatment on the self-renewal inhibitors, essentially ruling out AMPK/mTOR as the driving
capacityofcellsthatdidnotshowsignificantdifferenceduringfirst mechanism formetformin intargeting pancreaticCSCs.
passage sphere formation (Panc-185 and Panc-215), secondary Next,wehypothesizedthattheparticularsensitivityofCSCsto
and tertiary spheres were initiated without further metformin metformin might be attributable to anti-oxidant properties to
treatment.Formationofspheresinthesecondandthirdpassages metformin [19]. Indeed, metformin treatment significantly
wasdrasticallyhamperedsuggestingthatmetformintreatmenthad increased mitochondrial production of reactive oxygen species
irreversibly eliminated the majority of CSCs by inhibition or (ROS) in CSCs derived from all the used tumors, but these
abrogation of their self-renewal capacity (Fig. 2C). Consistently, changes were more pronounced in rapid responders (PDAC-253
pre-treatment with metformin resulted in the formation of fewer and354)ascomparedtothetardyrespondersPDAC-215and286
andsmallercoloniesascompared tocontrol(Fig.2D).Thegold (where metformin effects are not detectable in first generation
standard for CSC activity represents in vivo tumorigenicity. The sphere formation) (Fig. 4E). Interestingly, consistent data were
frequencyoftumorigeniccellsderivedfromPDAC-354,215,and obtained for the mitochondrial transmembrane potential, which
A6Lsphereswasregularlyveryhighwithvaluesbelow1:500and was most prominently reduced in rapid responders (Fig. 4F).
became markedly rarified by metformin treatment (Fig. 2E). Using the mitochondrial probe TMRE, we demonstrate that in
Finally,wefoundthatmetforminpretreatmentofpancreaticCSCs thesetumorsmetformindecreasedtheirmitochondrialtransmem-
subsequently reduced their migratory (Fig. S2A) and invasive branepotential(usedasageneralmarkerofmitochondrialtoxicity
capacity (Fig.2F). andinductionofapoptosis),whileitremainedunchangedoronly
slightly reduced in tardy responders. This direct effect on
Metformin specifically eliminates pancreatic CSCs mitochondriamayalsoexplainthedifferentialsensitivitybetween
Populationdoublingofadherentcellswasmarkedlyreducedby non-CSCsandCSCsasthelatterappeartobemoredependenton
metformin with strong intertumoral variation (Fig. 3A). Consis- theirmitochondriaforgeneratingenergy,whilenon-CSCsmainly
tently,Ki67expressionwasdiminishedsuggestinginhibitionofcell rely mitochondria-independent sources such as aerobic glycolysis
PLOSONE | www.plosone.org 5 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
Figure4.Mechanismofaction.(A)CellularATPlevelsinadherentcellsandspheresafter7doftreatmentwithmetforminorcontrol(n=3).(B)
Upperpanel:WesternblotanalysisofpAMPK,AMPK,andGAPDHinspherestreatedwithmetforminorcontrol.Lowerpanel:Westernblotanalysisof
pAMPK,pp70S6K,andGAPDHinadherentcellsandspherestreatedwithmetforminorcontrolfor7days(n=3).(C)Overalleffectofmetformin
PLOSONE | www.plosone.org 6 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
(3mM),inhibitionofmTOR(rapamycin;100ng/ml),anddirectionactivationofAMPK(A769662;10mM)onsphereformationand(n=6).(D)colony
formationforPDAC-215,253,and354cells(n=3).(E)Totalandmitochondrial(Mito)ROSproductionafter8hoursofcontrolormetformintreatment.
(F)Mitochondrialtransmembranepotentialafter8hoursofcontrolormetformintreatment.
doi:10.1371/journal.pone.0076518.g004
(Warburg effect) [20]. In addition, we also confirmed that the pancreatic CSC pool after termination of metformin treatment.
effectofmetforminonmitochondriawasindependentofAMPK/ Consistently,transplantationofthepretreatedcellsintoimmuno-
mTOR as neither the direct AMP activator A769662 nor the compromised mice revealed that the tumorigenicity of the cells
mTORinhibitorrapamycinwerecapableofmimickingtheeffects was indeed drastically reduced by short-term exposure to
of metformin (Fig.S4A/B). metformin. Most importantly, however, in vivo treatment of
established primary human cancers strongly responded to
Metformin stalls PDAC progression in vivo metformin withdisease regressionand subsequent stable disease.
First,westudiedtheeffectsofmetformininvivousingabrief7- TheidentifiedmechanismofactionformetformininCSCshas
day metformin treatment of PDAC-185. During the subsequent remainedlargelyunknown.Mostinvestigationsinbulktumorcells
follow-upof40daysweobservedasignificantreductionintumor support a simplified working model in which metformin exerts
growthandacompleteremissionin3out8tumorsformetformin anti-tumoral effects by indirectly activating AMPK followed by
treated-miceasopposedtonoremissioninthecontrolgroup(Fig. subsequent suppression of mTOR [8]. Directly targeting mTOR
S5). Consistent data were obtained for other PDACs with with rapalogs has also been considered as a target for anticancer
metforminmonotherapybeinghighlyeffectiveininducingdisease drugdevelopment.Whilethiswaseffectiveinpatientswithseveral
regression(Fig.5A–F).Severalimportantobservationsweremade cancers [24], preclinical studies in PDAC did not suggest a
during these studies. First, no gross toxicity was observed during particular high response rate (23%) [23]. Importantly, one of the
prolonged metformin treatment as body weight remained insensitive tumors in that study was PDAC-215, which we were
unchanged (Fig. 5A, right panel). Secondly, tumors slowly, able to also test in the present study using metformin. CSCs
but regularly relapsed after withdrawal of metformin on day 100 derived from this tumor showed a robust inhibition of secondary
(Fig.5A,leftpanel).Thirdly,whiletheadditionofgemcitabine sphereformationinvitroandefficienttumorregressioninvivo.On
invitroshowedanadditiveeffectontheCSCpopulation(Fig.5C), the other hand, direct activation of mTOR by A769662 or
no additive effect could be noted in vivo for metformin plus inhibition of mTOR by rapamycin did not mimic the strong
gemcitabine (Fig.5D). effectsthatweachievedwithmetforminsuggestingthattreatment
These data prompted us to hypothesize that gemcitabine was effects of metformin in pancreatic CSCs are independent of
not appropriately delivered to the PDAC tissue [21] and/or the alterations of theAMPK/mTOR axis.
stroma protected the CSCs from the deleterious effects of Human PDAC cells are known for their inherent tolerance to
metformin by promoting their stemness and subsequently resis- nutrition starvation, which enables them to survive under a
tance [22]. Therefore, we added the smoothened inhibitor SIBI- hypovascular (austerity) tumor microenvironment. It is an
C1asastroma/stellatecells-targetingagenttothecombinationof emerging paradigm that normal stem cells are characterized by
gemcitabineplusmetformin.Sincewehavepreviouslyshownthat predominant aerobic glycolysis and suppressed mitochondrial
the combination of gemcitabine plus smoothened inhibitor does oxidationwithsubsequentlylowermitochondrialATPproduction
not prevent relapse [10] and gemcitabine currently represents and ROS release [25]. Our data are in line with the notion that
standardtreatmentforPDAC,weonlytestedthiscombination.In pancreatic CSCs actually bear a highly mitochondrial-dependent
additiontoenhancingtissuedeliveryofgemcitabine,[21]thetriple metabolicprofile,whichisinstrikingcontrasttonormalstemcells,
combination also resulted in a doubling of tissue metformin but also distinguishes them from the bulk cancer cells. It has
concentration (23.160.82 versus 10.861.9mg/g). Subsequently previously been shown that metformin is slowly accumulated
reproducible disease regression was observed (Fig. 6A–C) and, 1000-fold within mitochondria and directly inhibits Complex 1
even more importantly, this combination was also effective in (NADHdehydrogenase),whichtranslocatesfourprotonsfromthe
tumors previously shown to be resistant to mTOR inhibition mitochondrialmatrixtotheintermembranespace,thusproducing
(Fig.6D)[23]. a proton gradient. The electron transport chain and oxidative
phosphorylation are coupled by this proton gradient across the
innermitochondrialmembrane,andtheirinhibitionappearstobe
Discussion
particularly lethal for CSCs [26]. Therefore, drugs such as
Here we demonstrate that the heterogeneous populations of metformin that target the oxidative mitochondrial metabolism
cancer cells harbored in primary human PDAC tissues differ in represent powerful therapeutic toolsfor attacking theCSCpool.
theirresponsetometformindependingontheirlevelofstemness. TheabilityofmetformintoselectivelyeliminateCSCsisinstark
While the bulk of more differentiated cancer cells reacted to contrast to the effects of gemcitabine, a chemotherapeutic drug
treatment with metformin with cell cycle arrest, a subset of cells thatkillsbulkcancercells,butnotcancerstemcells[15].Basedon
with distinct stemness features, namely CSCs actually underwent their distinct properties, metformin should synergize with
rapid apoptotic death due to energy crisis. Although the most gemcitabine to reduce both non-CSCs and CSCs in the mixed
apparenteffectformetformintreatmentinprimarysphereswasa population.Indeed,itwasonlyrecentlyshowninfourgenetically
marked reduction in sphere size consistent with a reduced differenttypesofbreastcancercelllinesthatmetforminselectively
proliferation rate of the more differentiated cells harbored in kills CSCs and that the combination of metformin and the
these spheres, subsequent serial passaging of the spheres clearly standard chemotherapeutic agent doxorubicin killed both CSCs
revealed theiralmost complete andirreversible lossof clonogenic and non-CSCs in vitro as well as reduced tumor mass and
propagationability,despitethefactthatmetformintreatmentwas prolongedremissionmoreeffectivelyinvivothaneitherdrugalone
stopped already after the first passage. These data demonstrate [27].WhilewecouldconfirmthishypothesisforPDACinvitro,our
thatmetforminvirtuallyexhaustedtheCSCfraction,butarealso in vivo treatment studies suggest that gemcitabine is actually
consistent with the notion that non-CSC do not replenish the dispensable as long as metformin is continued throughout the
PLOSONE | www.plosone.org 7 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
PLOSONE | www.plosone.org 8 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
Figure5.MetforminstallsPDACprogressioninvivo.(A)Leftpanel:PDAC-215tissuewasimplantedandtreatmentwasallocatedafterinitial
tumortakewasverified.Miceweretreatedwitheithergemcitabine(Gem)ormetformin(Met).Metforminwasdiscontinuedond100totestthe
potentialoftheremaininglesionstoinitiatediseaserelapse.Afterdocumenteddiseaserelapse,metformintreatmentwasre-administered.Right
panel:Bodyweightduringtreatment.(B)Leftpanel:Histologicalanalysis:Hematoxylin&Eosin(H&E),CyclinD1,andCaspase3.Rightpanel:Content
forCD44+cells.(C)Effectsofinvitrotreatmentasindicatedonsphereformationcapacity.(D)PDAC-A6Ltissueimplantedinmiceandtreatedas
indicated.(E)ContentforCD44+cells(upperpanel)andsphereformationcapacity(lowerpanel).(F)Histologicalanalysis:H&Eandcytokeratin19
(CK19;n=6).
doi:10.1371/journal.pone.0076518.g005
study. While these data certainly warrant further validation it Metformin bears an exceptional safety profile as only hepato-
appears reasonable to speculate that metformin monotherapy cytes,butnotothernon-transformed(stem)cellsexpressOCTfor
couldrepresentanoveltreatmentoptionforpatientsthataretoo efficient cellular uptake of metformin. Several clinical trials
fragile to tolerate the side effects of the chemotherapeutic agent (NCT01210911, NCT01167738, and NCT01488552) are cur-
gemcitabine [28]. rently testing metformin in locally advanced and metastatic
Unfortunately,however,alltestedtumorseventuallyprogressed PDAC. This will hopefully provide a definitive assessment of the
under metformin therapy. It remains to be determined if CSCs clinicaleffectsofmetformininPDAC.Itwouldbeofparticularly
exposedtolong-termmetformintreatmentswitchtheirmetabolic interest whether patients will also ultimately progress during
phenotype rendering them resistant to metformin’s effects on metformin treatment and whether this could be assessed and/or
mitochondria (acquired resistance) or whether a preexisting predicted by the analysis of circulating cancer (stem) cells. Their
subpopulation of CSCs is actually resistant to metformin and prospective isolation during relapse may provide important
eventually becomes the dominating clone (inherited resistance). mechanistic insightsinto thecauseof relapse.
This issue should be addressed in future studies by comparing
CSCs isolated from tumors that regressed under metformin Materials and Methods
treatment and CSCs isolated from tumors that eventually
progressed underthesameconditions.Extensivecharacterization Tumor samples
ofresistantclonesshouldprovidecrucialinsightsastowhetherthis After written informed consent had been obtained from the
is actually preventable or second line treatment options could be patients, excess tissues from resected pancreatic carcinomas was
offered. xenografted at Johns Hopkins Medical Institutions (Ethics board
Importantly, however, the addition of a stroma targeting approval: JHMIRB: 05-04-14-02 ‘‘A Feasibility Study for Indi-
smoothened inhibitor appears mandatory to achieve stronger vidualized Treatment of Patients with Advanced Pancreatic
and longer lasting response rates. The most likely explanation Cancer’’) and Hospital de Madrid - Centro Integral Oncolo´gico
could be two fold. One contributing factor to the failure of Clara Campal (FHM.06.10 ‘‘Establishment of bank for tumors
systemic therapies may be the abundant tumor stromal content andhealthytissueinpatientswithcancer’’),respectively,underthe
that is the characteristic of PDAC. The stroma represents the indicated Institutional Review Board-approved protocols [30].
majorityofthetumormass,andconsistsofadynamicassortment Briefly, excess tumor tissues not needed for clinical diagnosis
of extracellular matrix components and non-neoplastic cells during routine Whipple resections performed by surgeons that
including fibroblastic, vascular, and immune cells. Recent work were not involved in the present study were subsequently
hasrevealedthatthestromasupportsCSCs,[18,22,29]promotes implanted into immunocompromised mice. All patient informa-
invasiveness and metastasis as well as simultaneously serves as a
tionwasmadeanonymousbyremovalofanyinformation,which
physical barrier to drug delivery [21]. Therefore, hedgehog
identifies,orcouldleadtotheidentificationofthepatient.Noneof
pathway inhibition, in addition to providing better access for
thepatients hadundergoneneoadjuvant radiation or chemother-
systemicallyadministereddrug,mayalsoabrogatetheCSCniche
apy prior toresectionof thetumor.
rendering them more susceptible to the treatment effects of
metforminand/orgemcitabine.Futuremoreextensivestudieswill
Primary human PDAC cells
have to show if the development of resistance can be efficiently
Forinvitrostudies,tissuefragmentswereminced,enzymatically
prevented bythistreatment strategy.
digested with collagenase (Stem Cell Technologies, Vancouver,
Itmaybearguedthatthemetforminconcentrationsusedinour
BritishColombia)for90minat37uC[10]andaftercentrifugation
invitrostudies,albeitalreadysignificantlylowerthanthoseutilized
for5minat1,200rpmthepelletswereresuspendedandcultured
in previous studies, are still non-achievable in vivo and therefore
in RPMI, 10% FBS and 50units/ml penicillin/streptomycin. For
irrelevant from a mechanistic point of view. However, it is
someexperiments,thehumanPDACcelllineL3.6plwasusedas
important to note that metformin accumulates in tissues and
previously described[15].
particularly in mitochondria at concentrations several-fold higher
than those measured in blood [26]. This is particularly true for
Cancer stem cell-enriching culture
cellsthatareequippedwiththerespectivetransportersrelevantfor
metforminuptake,namelyOCT1-3.Asweshowhere,PDACcells PDAC spheres were generated and expanded in DMEM-F12
expressallthreeisoforms,butshowparticularlyhighexpressionof (Invitrogen,Karlsruhe,Germany)supplementedwithB-27(Gibco,
OCT1. While it is difficult to assess the true intracellular Karlsruhe,Germany)andbFGF(PeproTechEC,London,United
metformin concentrations in PDAC cells due to the massive Kingdom).103cells/mlwereseededinultra-lowattachmentplates
stroma content and necrotic areas in harvested tumors, our (Corning B.V., Schiphol-Rijk, The Netherlands) as described
analysis of the metformin concentration revealed higher metfor- previously.[31] After 7 d of incubation, spheres were typically
minconcentrationsinthePDACtissue(10.8mg/goftissue)than .75mm large with ,97% CD133high. For serial passaging, 7-
in the liver (8.9mg/g) and even more so if metformin was day-old spheres were harvested using 40mm cell strainers,
administeredincombinationwiththesmoothenedinhibitorSIBI- dissociated to single cells with trypsin, and then re-grown for
C1 (23.1mg/g). Therefore, our mechanistic in vitro studies are 7 d. Cultures were kept no longer than 4 weeks after recovery
indeed highlyrelevant forthe invivosetting. fromfrozen stocks (passage3–4).
PLOSONE | www.plosone.org 9 October2013 | Volume 8 | Issue 10 | e76518
MetforminTargetsPancreaticCancerStemCells
Figure 6. Additional targeting of the stroma prevents tumor relapse. (A) PDAC-185 tissue implanted in mice and treated as indicated
includingthesmoothenedinhibitorSIBI-C1.(B)ContentforEpCAM+andCD133+cells,respectively,(upperpanel)andsphereformationcapacity
(lowerpanel).(C)Histologicalanalysis:Hematoxylin&Eosin(H&E)andcytokeratin19(CK19).(D)mTORinhibitorresistantPDAC-253tissueimplanted
inmiceandtreatedasindicated(n=6).
doi:10.1371/journal.pone.0076518.g006
PLOSONE | www.plosone.org 10 October2013 | Volume 8 | Issue 10 | e76518