Myoferlin Depletion in Breast Cancer Cells Promotes Mesenchymal to Epithelial Shape Change ...

12 Pages·2012·0.85 MB·English

Myoferlin Depletion in Breast Cancer Cells Promotes
Mesenchymal to Epithelial Shape Change and Stalls
Invasion
Ruth Li1, William E. Ackerman, IV1, Cosmin Mihai4, Leonithas I. Volakis2, Samir Ghadiali2,3,
Douglas A. Kniss1,2*
1LaboratoryofPerinatalResearch,DivisionofMaternal-FetalMedicine,DepartmentofObstetricsandGynecology,TheOhioStateUniversity,Columbus,Ohio,United
StatesofAmerica,2DepartmentofBiomedicalEngineering,TheOhioStateUniversity,Columbus,Ohio,UnitedStatesofAmerica,3DivisionofPulmonary,Allergy,Critical
CareandSleepMedicine,DepartmentofInternalMedicine,TheOhioStateUniversity,Columbus,Ohio,UnitedStatesofAmerica,4PacificNorthwestNationalLaboratory,
Richland,Washington,UnitedStatesofAmerica
Abstract
Myoferlin (MYOF) is a mammalian ferlin protein with homology to ancestral Fer-1, a nematode protein that regulates
spermaticmembranefusion,whichunderliestheamoeboid-likemovementsofitssperm.Studiesinmuscleandendothelial
cellshavereportedontheroleofmyoferlininmembranerepair,endocytosis,myoblastfusion,andtheproperexpressionof
variousplasmamembranereceptors.Inthisstudy,usinganinvitrohumanbreastcancercellmodel,wedemonstratethat
myoferlin is abundantly expressed in invasive breast tumor cells. Depletion of MYOF using lentiviral-driven shRNA
expression revealed that MDA-MB-231 cells reverted to an epithelial morphology, suggesting at least some features of
mesenchymal to epithelial transition (MET). These observations were confirmed by the down-regulation of some
mesenchymalcellmarkers(e.g.,fibronectinandvimentin)andcoordinateup-regulationoftheE-cadherinepithelialmarker.
CellinvasionassaysusingBoydenchambersshowedthatlossofMYOFledtoasignificantdiminutionininvasionthrough
MatrigelortypeIcollagen,whilecellmigrationwasunaffected.PCRarrayandscreeningofserum-freeculturesupernatants
fromshRNAMYOFtransducedMDA-MB-231cellsindicatedasignificantreductioninthesteady-statelevelsofseveralmatrix
metalloproteinases.ThesedatawhenconsideredintotosuggestanovelroleofMYOFinbreasttumorcellinvasionanda
potential reversion toan epithelialphenotype uponloss ofMYOF.
Citation:LiR,AckermanWEIV,MihaiC,VolakisLI,GhadialiS,etal.(2012)MyoferlinDepletioninBreastCancerCellsPromotesMesenchymaltoEpithelialShape
ChangeandStallsInvasion.PLoSONE7(6):e39766.doi:10.1371/journal.pone.0039766
Editor:WanjinHong,InstituteofMolecularandCellBiology,Singapore
ReceivedApril23,2012;AcceptedMay30,2012;PublishedJune27,2012
Copyright:(cid:2)2012Lietal.Thisisanopen-accessarticledistributedunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsunrestricted
use,distribution,andreproductioninanymedium,providedtheoriginalauthorandsourcearecredited.
Funding:ThisworkwassupportedbytheOhioStateUniversityPerinatalResearchandDevelopmentFund(DAK),NIHK08HD49628(WEA),andtheNational
ScienceFoundationNanoscienceScienceandEngineeringCentergrant#EEC-0425626(RL)andChemical,Bioengineering,Environmental,andTransportSystems
(CBET)#1134201(LV,DAK,SG).Thefundershadnoroleinstudydesign,datacollectionandanalysis,decisiontopublish,orpreparationofthemanuscript.
CompetingInterests:Theauthorshavedeclaredthatnocompetinginterestsexist.
*E-mail:[email protected]
Introduction protease-independent, and collective migration) [7,8]. Consider-
ableprogresshasbeenmadeinidentifyingthepotentialmolecular
Breastcanceristhesecondleadingcauseofcancermortalityin components that mediate cell migration, and there is growing
women [1], with the majority of the deaths due to metastatic
evidence that intracellular vesicle trafficking of key proteins is
ratherthanlocalizeddisease[2].Unrestrainedcelldivisionelicited
crucial for efficient migration [9]. In tumor cell migration, the
by somatic and/or germline mutations in several oncogenes and
recycling of focal adhesion proteins (i.e., integrin receptor
tumor suppressor genes such as TP53 and Rb, and resistance to
recycling) through endocytosis/exocytosis is now thought to
programmed cell death are hallmarks of tumorigenesis [3,4]. For
contribute to the maintenance of polarized movement [10,11].
tumor cells to efficiently metastasize, they often undergo a
Moreover,endocytosis/exocytictraffickinghasbeenimplicatedin
pernicioustransformationcharacterizedbydramaticallyincreased
thedeliveryofproteolyticenzymes-includingmatrixmetallopro-
migrationandinvasivecapacity.Specifically,thespreadofcancer
teinases(MMPs)-toinvadopodia,specializedprotrusionsutilized
from a localized, self-contained tumor through tissue stroma and
by cells fordegradation of ECM [12,13].
into distant organs requires that cells achieve atypical, robust
Ferlin proteins, an evolutionarily ancient family of large
motility and the capacity to aggressively degrade extracellular
integral membrane proteins [14], have been implicated in
matrix (ECM), enabling them to invade surrounding tissues and
vesicle trafficking in a variety of physiological settings. All
vessels of the blood and lymphatic systems, and subsequently
mammalian ferlins derive their names based on homology to
establish nascent, secondarytumors [5,6].
the C. elegans protein FER-1 (FERtilization defective-1). In
The enhanced migration and invasive capacity of metastatic
roundworms, fer-1 is required for the fusion of specialized
tumor cells are the subject of intense investigation, and it is now
vesicles (membranous organelles) with the plasma membrane at
appreciated that several forms of cancer cell motility exist (i.e.,
the leading edge of cell migration in spermatozoa; in the
single-cell, mesenchymal and amoeboid, protease-dependent,
PLoSONE | www.plosone.org 1 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
absence of functional FER-1 protein, the normal amoeboid Results
locomotion of sperm is impaired, and infertility results [15].
Human patients harboring FER1L1 (FER-1-like 1, dysferlin) Myoferlin in Breast Cancer Cells Correlates with Cell
mutations manifest one of two autosomal recessive forms of Invasiveness
muscular dystrophy - limb girdle muscular dystrophy type 2B TocorrelatetheexpressionofMYOFwiththeinvasivecapacity
and Myoshi myopathy [16], due to the inability of skeletal ofbreastcancercells,weanalyzedfivemammarycelllines:MCF-
muscle fiber sarcolemma to repair damaged muscle cells during 10A,MCF-7,T47D,BT549,andMBA-MB-231.TheMCF-10A
the normal course of biomechanical wear and tear [17–19]. line was chosen as a model for non-cancerous, non-invasive
Mutations in FER1L2 (otoferlin) result in non-syndromic mammary epithelial [32]. MCF-7 and T47D cells model non2/
deafness (DFNB9), due to the failure of synaptic vesicles to low-invasivebreastcancers,whileBT549andMDA-MB-231have
fuse and exocytose their cargo at the presynaptic plasma high invasivecapacity [33].
membrane [20,21]. FER1L3 (myoferlin, MYOF), historically Examination of MYOF mRNA expression in each of the cell
considered to be a muscle-specific protein, has not yet been linesindicatedthatMYOFexpressionwas2.44-foldhigherinthe
directly associated with a distinct mammalian disorder, but invasive lines (BT549 and MDA-MB-231) relative to non2/low-
recent studies have indicated that its deletion results in impaired invasive cells (MCF-10A, MCF-7,and T47D)(Figure 1A andB).
mouse myoblast fusion into mature skeletal myotubes [22]. Incongruence,higherMYOFproteinexpressionwasobservedby
Furthermore, MYOF has been shown to mediate caveolae- immunoblotting in the highly-invasive breast cancer cell lines
dependent endocytosis in human endothelial cells [23]. relative to those with low invasive potential (Figure 1C). To
expand upon this, we examined the transcriptional profiling data
RecentreportsindicatethatMYOFiscriticallyinvolvedinthe
availableon51establishedbreastcancerlinesgeneratedbyNeve
function and/or stability of plasma membrane receptor tyrosine
and colleagues (http://cancer.lbl.gov/breastcancer/data) [34].
kinases (RTKs). Bernatchez etal. demonstrated that ablation of
Those 14cell lines which clustered into the invasive, basal B-like
MYOF in vascular endothelial cells resulted in instability and
subsetexhibitedstatisticallysignificantlygreaterMYOFexpression
rapid degradation of the vascular endothelial growth factor
compared to the 25cell lines of the luminal phenotype as
receptor 2 (VEGFR-2) [24]. Further research demonstrated that
determined by two MYOF probesets (201798_s_at and
theexpressionofanotherangiogenictyrosinekinasereceptor,Tie-
211864_s_at, Figure S1).
2, is also attenuated upon MYOF depletion in endothelial cells
Buildingontheseobservations,weminedpublishedmicroarray
[25].Finally,Demonbreunet al.reportedthatknockoutofMYOF
data available through the ArrayExpress database (www.ebi.ac.
in mouse muscle resulted in diminished insulin-growth factor-1
uk/arrayexpress) [35]. Our query returned microarray data
receptor (IGFR) response and accumulation of the receptors in
(accession: E-TABM-276) from Cheng et al. [36] consisting of a
vesicles targeted fordegradation [26].
panel of 23 microdissected primary invasive ductal carcinoma
The studies reported above when considered in toto suggest
tissues,28adjacentstromaltissues,and10healthycontrols.When
that MYOF may be a regulator of vesicle fusion events that
MYOF expression was categorized by disease state, we observed
deliver essential cargos (e.g., growth factor receptors, cell
significantly higher MYOF mRNA expression in tissues from
adhesion molecules, and other cell-surface proteins) to and/or
patients with invasive ductal carcinoma compared with samples
from the plasma membrane. Inasmuch as tumor cell metastatic
from healthy individuals (Figure S2). Collectively, these data
dissemination involves the convergence of receptors, signal
suggest a potential correlation between MYOF over-expression
transduction pathways, adhesion and matrix proteolytic mole-
andinvasive breast carcinoma.
cules at the leading edge of a tumor cell, we proposed that
MYOF (and perhaps other ferlin proteins) may be involved in
Myoferlin Knockdown Promotes a Mesenchymal to
one or more steps in tumor progression and/or spread.
Epithelial Change
Microarray and proteomic studies in the cancer literature have
Given a possible association between MYOF expression and
reported MYOF expression in breast cancer specimens and
invasive potential, we examined whether MYOF contributed to
relevant cell lines [27–29]. For example, in one microarray
breast cancer cell invasion. We selected the MDA-MB-231 cells
study, MYOF was 1 of 39 genes found to be over-represented
for these experiments since they exhibited high levels of MYOF
in breast carcinoma [28]. These observations prompted us to
protein expression in our survey, and were previously shown to
hypothesize that MYOF may be an important protein in breast
have the greatest invasive capacity compared to a large panel of
cancer cells for their mobilization during cellular migration
breast cancer cell lines [34]. Stable lines of MYOF-deficient
and/or invasion.
MDA-MB-231 cells (231MYOF-KD) were generated using lentivi-
We recently reported on the mathematical modeling of the
rus-based delivery of short hairpin ribonucleic acids (shRNAs)
role of MYOF in breast cancer cell invasion [30]. In the targetinghumanMYOF(FigureS3).Controlcelllines(231LTV-ctrl)
current manuscript, we present novel data confirming the
were generated in parallel using a non-human gene targeting
computational modeling and show, in addition, that MYOF-
construct.SelectiveknockdownofMYOFproteinexpressionwas
deficient MDA-MB-231 human breast cancer cells exhibit a
confirmedbyimmunoblotting(Figure1D).Ofthethreeconstructs
more epithelial shape compared to the mesenchymal morphol-
initiallyscreened(#10628,10630,and1522),#10628yieldedthe
ogy of wild-type MDA-MB-231 cells, suggesting that some
most efficient knockdown (mean 94% compared with controls,
features of a mesenchymal-to-epithelial transition (MET) phe-
n=4)and wasused insubsequent experiments.
notype [31] may exist in the absence of functional MYOF
An unexpected, but interesting shape change appeared in the
protein. Matrigel and collagen I based invasion assays demon- 231MYOF-KD cells, wherein they exhibited a more epithelial-like
strated that depletion of MYOF diminished the invasive ability
(cobblestone) morphology which diverged from the fibroblastic,
of cancer cells. In congruence with the mathematical modeling spindleshapeofthewildtypeand231LTV-ctrlcells(Figures2and
results, we report that the decreased invasive capacity is due, at 3).The231MYOF-KDmorphologywasstable,andwasmaintained
least in part, to the down-regulation of MMP expression in after 10 continual passages in culture (data not shown). By
MYOF-deficient cells. immunofluorescent microscopy, we noted that 231MYOF-KD cells
PLoSONE | www.plosone.org 2 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
Figure1.Myoferlinexpressioninbreastcancer.(A)QuantitativeRT-PCRresultsofMYOFmRNAlevelsinapanelofbreastcancercelllines,
normalizedto18SandcomparedtothenonmalignantbreastepithelialcelllineMCF-10A(mean6s.d.,n=3percelltype).Nostatisticalsignificance
wasdetectedincomparingpairsofcelllines(Kruskal-WalliswithDunn’smultiplecomparisontest).(B)RepresentationofthedatainsubpanelA
grouped by known invasive capacity of the cells (2 tailed, p=0.003, Mann Whitney). (C) Immunoblotting for MYOF in breast cancer cell lines.
DensitometrymeasurementswasdonebynormalizingthedensityofMYOFtorespectiveGAPDHoractinstaining,andfurthernormalizedtothe
nonmalignantmammaryepithelialcelllineMCF-10A(n=4,nostatisticalsignificancedetectedbyKruskal-WalliswithDunn’smultiplecomparison
test).Animmunoblotimageisshowninthesubpanel.(D)Myoferlinknockdowninbreastcancercells.ImmunoblotshowingtheexpressionofMYOF
in MDA-MB-231 wild type (WT), lentiviral transduction control (LTV-ctrl), and MYOF knockdown cell lysates using three Sigma-Aldrich MISSIONH
shRNAconstructs(#1522,10628,and10630).
doi:10.1371/journal.pone.0039766.g001
formedmorecellclustersrelativetocontrolcells.Whenexamined TheMET-likemorphologicalchangespromptedustoexamine
in greater detail using atomic force and scanning electron whether the expression of common epithelial-mesenchymal
microscopy, the morphological divergence between 231MYOF-KD transition(EMT)markers(e.g.vimentin,fibronection,N-cadherin,
and control cells was very prominent (Figure 3). At the cellular and E-cadherin) was affected by MYOF depletion. By immuno-
level, 231LTV-ctrl cells often appeared elongated with tapering blotting,231MYOF-KDcellsexpressedhigherlevelsoftheepithelial
cytoplasmic poles, having one or more small lamellipodia and marked E-cadherin, and lower levels of mesenchymal markers,
relatively short filopodia, and actin filaments oriented along the fibronectin and vimentin, than 231LTV-ctrl cells (Figure 4). MDA-
long axis of the cell (Figure 3). In contrast, 231MYOF-KD cells MB-231 cells do not natively express N-cadherin, so it was not
exhibited a flattened, polygonal shape with broad lamellipodia, surprising that N-cadherin was undetectable in 231LTV-ctrl and
andelongatedfilopodiathattendedtoorienttowardneighboring 231MYOF-KD cells (data not shown). The results of this screen of
cells. Collectively, these results suggested a potential reversion of EMT markers suggest that the morphologic shift associated with
epithelial-mesenchymal transition, i.e., mesenchymal-epithelial MYOF-depletion may be correlated with MET changes at the
transition (MET), at themorphologylevel. molecular level. In a survey using a qRT-PCR-based array of
PLoSONE | www.plosone.org 3 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
Figure 2. Morphology change following myoferlin depletion in MDA-MB-231 cells. Immunofluorescence micrographs showing the
morphologyofMDA-MB-231wildtype(WT),lentiviraltransductioncontrol(LTV-ctrl),andmyoferlinknockdown(MYOF-KD)stablecelllinesinculture.
NotethemoreepithelialmorphologyofMYOF-KDcellscomparedtothemoremesenchymalappearanceoftheWTandLTV-ctrlcells.Bar=50mm.
doi:10.1371/journal.pone.0039766.g002
EMT markers, we found that the mRNA levels appeared to Myoferlin Depletion Attenuates Cell Invasiveness
correlatewiththeWesternresultsinthatfibronectionmRNAwas MYOF deficiency was not associated with a change in
down-regulated 3.46-fold while that of E-cadherin was up- chemotaxis (Figure 5A) in Boyden chamber assays when 10%
regulated 2.58-fold in 231MYOF-KD cells compared with controls, fetal bovine serum was used as the chemoattractant. However,
althoughnosignificantchangewasfoundforvimentin(TableS1). 231MYOF-KD cells exhibited a significant ,38% reduction in
invasiveness into Matrigel relative to 231LTV-ctrl cells (Figure 5B).
TheseresultssuggestthattargeteddisruptionofMYOFexpression
Myoferlin Depletion Does Not Alter Cell Proliferation or
selectivelyattenuatesinvasivenessthroughreconstitutedbasement
Cell Cycle Distribution membrane matrix,with littleeffect onchemotactic migration.
We next subjected 231LTV-ctrl and 231MYOF-KD cells to
proliferation assays. First, to assess general proliferative status, Myoferlin Depletion Alters MMP Expression
MTS assays were conducted on wild-type, 231LTV-ctrl, and The reduced invasive capacity of 231MYOF-KD cells led us to
231MYOF-KD cells. We observed no change in the growth rates study whether proteins associated with tumor-microenvironment
of the three cell types (Figure S4A), indicating that MYOF does interactions may be altered following MYOF depletion. Specifi-
not profoundly alter tumor cell growth. To verify this result, we cally,giventhatMYOFhasbeenimplicatedinexocytosis[37],we
then carried outcellcycle analysison the three celltypes. Again, hypothesized that its depletion might decrease the secretion of
we detected no change in cell cycle activity in MYOF-depleted matrix metalloproteinases (MMPs) and/or tissue inhibitors of
cells compared towild-type or231LTV-ctrl cells (Figure S4B). MMPs (TIMPs), as these are vital endopeptidases for degrading
PLoSONE | www.plosone.org 4 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
Figure3.MorphologychangefollowingmyoferlindepletioninMDA-MB-231cells.Atomicforceandscanningelectronmicroscopyimages
showingthespindle,elongatedshapeoflentiviralcontrol(LTV-ctrl)cellsandthemoreflatandcircularmorphologyofmyoferlindepleted(MYOF-KD)
cells.AFMimagingshowspronouncedactinstressfibers(blackarrows)orientedalongthelongaxisbeingevidentinthecontrolbutnotintheMYOF
depletedcells.Cytoplasmicpoles,lamellipodiaandfilopodiaareobservableintheSEMimages.Whitearrowheadsindicatetheleadingedgeofcells.
doi:10.1371/journal.pone.0039766.g003
ECM constituents [38]. To that end, we collected 24hour confirmedthediminution(,97%)ofpro-MMP1secretionintothe
conditioned, serum-free supernatants from 231LTV-ctrl and extracellular milieu by solution-phase ELISA of 231LTV-ctrl and
231MYOF-KD cells and performed an array-based multiplex 231MYOF-KD cells (Figure 6B). To confirm the functional
sandwich ELISA to screen the expression of several of these association between MYOF knockdown and MMP1 depletion,
endopeptidases and their regulators (Figure S5). We observed a weinvestigatedtheinvasivecapacityof231MYOF-KDcellsthrough
reductioninthesecretionofMMP-1,-2,-3,and-8from231MYOF- acollagenImatrix,sincefibrillarcollagenisakeysubstrateforthe
KD cells (MMP-2, -3, and -8 were undetectable in 231MYOF-KD MMP1enzyme[39]andMatrigellackscollagenI[40].Insupport
cells). MMP-13 levels were not drastically affected, while in of these previous observations, 231MYOF-KD cells exhibited a
contrast, levels of MMP-9 and -10 were elevated in 231MYOF-KD ,62%reductionininvasionthroughcollagenImatrixcompared
cells. In addition, the 231MYOF-KD cells had reduced levels of with 231LTV-ctrl cells (Figure 6C).
TIMP-4,butnonotablechangesinTIMP-1and-2releaseswere We next tested whether MMP1 expression was diminished at
evident. Of the detectable MMPs, the most pronounced change the transcriptional level using qRT-PCR. Surprisingly, MMP1
was the decrease in secreted MMP1. However, given the small mRNAlevelswerereducedoveranaverage18.8-foldin231MYOF-
samplesizeofthisscreeningexperiment(n=1),statisticalanalysis KDcellswhencomparedwiththe231LTV-ctrlcells(FigureS6).This
couldnotbeperformedtovalidatethesignificanceoftheobserved finding of MMP1 mRNA down-regulation in 231MYOF-KD cells
fold changes. Therefore, to validate the change in MMP1, wasstriking,astherehavebeennoreportsfromotherlaboratories
multiplebiologicalreplicatesofconcentrated24hoursupernatant suggesting that MYOF may contribute to gene expression
from 231LTV-ctrl and 231MYOF-KD cells were subjected to regulation. Since the diminution of MYOF in the MDA-MB-
immunoblotting (Figure 6A), and secreted MMP1 was verified to 231 cells was mediated by shRNA (which mimics the function of
bedecreased in231MYOF-KDcells comparedtocontrols.Wealso microRNA molecules [41]), we recognized the potential that
PLoSONE | www.plosone.org 5 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
Figure 4. Expression of EMT markers following myoferlin depletion in MDA-MB-231 cells. (A) Representative immunoblots (n=3) of
selectEMTmarkersinlentiviralcontrol(LTV-ctrl)andmyoferlindepleted(MYOF-KD)cellsserum-starvedfor24h.(B)Graphsillustratingthesemi-
quantitativeevaluationoftheexpressionofEMTmarkersbydensitometryanalysisofblots(n=3).Densitywaspresentedinthegraphas‘‘relative
density(%)’’withdensityoflentiviralcontrolsnormalizedto100%,andstatisticaltestingwasdonewiththe1-samplet-test.
doi:10.1371/journal.pone.0039766.g004
MMP1mRNAcouldbedown-regulatedasanoff-targeteffectof the #10628 construct did not show MMP1 to be a potential
the RNAi targeting construct. We therefore screened additional target.Construct#10628isfromaclonecatalogedbytheRNAi
MYOF-targeting shRNA constructs. Although two constructs Consortium (http://www.broadinstitute.org), and a search
(Figure S3) yielding partial MYOF knockdown (mean 64–66% through the consortium database did not show the MYOF
knockdown compared with controls, n=4) were not associated shRNA construct as a non-targeting clone with matching
with reductions in MMP1 (data not shown), an additional transcripts in the MMP1 gene. Finally, the construct sequence
construct (#320398) that targets the coding sequence of MYOF was placed into the miRBase [43] to search against known
(FigureS3)resultedinameanof91%MYOFknockdown(n=3) miRNAs.Therewerenomatchesofthesequenceagainstmature
and was also associated with significant reductions in MMP1 miRNAs, and all matches to stem-loop sequences using the
release (Figure S7AandB). SSEARCH method were from non-human species (silk worm,
In addition, we performed an extensive in silico analysis to mosquito, and platypus) and were not significant (E-value
examinepotentialoff-targeteffectsofconstruct#10628.Asearch ..0.01).
throughthehumannucleotidedatabaseforsequencessimilarityto We did additional PCR screening using a commercial
the21nucleotidesequenceof#10628didnotrevealMMP1asa extracellular matrix and adhesion molecules array to profile the
transcript with significant alignment (BLASTN version 2.2.24+) expression of MMP genes (Table S2). The screen validated the
[42]. A search through the microRNA database TargetScan MMP1PCRresultsshowingdown-regulationofMMP1mRNAin
Custom (http://www.targetscan.org/, release5.1,accessed Octo- 231MYOF-KDcells.Specifically,ofthefourMMPs(MMP1,10,11,
ber10,2010)usingthe2–8mersequenceof‘‘CTCTACT’’from 14) expressed at reasonable levels in the MDA-MB-231 cells (i.e.
PLoSONE | www.plosone.org 6 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
Figure5.MyoferlindepletionreducesinvasivebutnotmigratorycapacityofMDA-MB-231scontrol(LTV-ctrl)andMYOFdepleted
cells (MYOF-KD). (A) Boyden chamber migration assay of MDA-MB-231 cells moving across 8mm porous membranes towards a 10% serum
gradientfor24h(mean6s.d.,n=3,unpaired2-tailedt-test).(B)24hBoydenchamberinvasionresultsofMDA-MB-231cellsacrossa100%Matrigel
coated8mmporousmembranetowardsa10%serumchemoattractant(mean6s.d.,n=3,unpaired2-tailedt-test).
doi:10.1371/journal.pone.0039766.g005
Cp value ,30), only MMP1 was found to be significantly MYOF depletion leads to fundamental changes in the mode of
regulated($62fold-change)atthetranscriptionlevelbyMYOF- breast tumor cell motility (Volakis et al., in preparation, J
depletion. Biomechanics), suggesting that MYOF loss also contributes to
disturbances in cell migration on two-dimensional surfaces (work
Discussion inprogress).
Tumor cell invasion into distant tissues to establish secondary
We provide in silico and experimental evidence that MYOF malignanciesisahallmarkofcancermetastasisandcancer-related
plays a critical and previously unrecognized role in breast tumor death[4].Inmanycases,enhancedtumorcellmigrationaswellas
cell invasion. Mining of transcriptome and proteome databases localanddistantinvasionareprecededbyanEMTthatconverts
from breast cancer cell lines revealed higher levels of MYOF sessile epithelial cells into actively migratory and invasive cells
mRNA and protein in the more invasive lines compared with [46,47].Inorderforefficientmigrationandinvasiontooccur,in
poorly invasive breast tumor cells or normal mammary epithelial many instances cancer cells must secrete several MMPs for
cells (i.e., MCF-10A). Published microarray datasets using breast basement membrane and stromal ECM degradation [48]. We
cancercelllines[34]alsosupportgreaterMYOFexpressioninthe probedtheinvasivecapacityofbreasttumorcells(MDA-MB-231)
invasive basal B group compared with the luminal subtype. In a followingshRNAdirecteddepletionofMYOF(shRNAMYOF),and
separate study, MYOF was found to be 1 of 39 genes selectively noted a profound deficit in the ability of cells to invade through
over-expressedinbreastcarcinomausing20humanbreastcancer Matrigel (basement membrane mimic) [49] and collagen type I
cases[28].ThesedatasuggestedtousthatMYOFisapotentially (stromal ECM mimic) [50]. These observations indicated that
important proteinin breast cancerbiology. MYOF disruption may perturb one or more MMPs that are
The premise that MYOF contributes to cancer invasion was essentialforeffectivecellinvasionthroughtheECMorcollagenI
based on several reports implicating a role for ferlin proteins in [48]. Indeed, screens of several MMPs using qRT-PCR and
intracellular vesicle trafficking, including cell motility. For proteinimmunoassayarrayssuggestedthatMMPs-1,3,8,12,13,
example, Ward and colleagues demonstrated that temperature- 14, and16were down-regulated in MDA-MB-231cells inwhich
sensitive mutations in the ancestral fer-1 gene in C. elegans led to MYOF was knocked down by RNAi (Figure S5 and Table S2).
fertility impairments yielding immotile spermatozoa secondary to Interestingly, MMP-9 appeared to be up-regulated under these
defectsinvesiclerecyclingat theleadingedgeofthecell[15,44]. same conditions, indicating the MYOF depletion results in
Mechanistically,mutationsinC2domainsoffer-1alteredcalcium selective changesinMMPs.
sensitivityandsubsequentlyimpededcalcium-dependentfusionof To evaluate the effect of abolishing MYOF on MMP
intracellular vesicles to the plasma membrane that drives the expression,weprobedMMP-1further,becauseourdatarevealed
amoeboid movementof theroundworm spermcells [15]. Vesicle a nearly complete loss of MMP-1 when shRNAMYOF was
trafficking has been implicated in several important steps during expressed in MDA-MB-231 cells. Quantitative RT-PCR and
cell motility [9]. Based on these observations, coupled with the WesternblottingshowedessentiallynodetectableMMP-1mRNA
observationthatMYOFexpressionwasincreasedinbreastcancer or protein in 231MYOF-KD cells. Moreover, we detected virtually
specimensandbreastcancercelllineswithhighinvasivepotential, no pro-MMP-1 reactive protein by ELISA. Thus, it appears that
wehypothesizedthatMYOFdepletionwouldimpairthemotility MYOF may play a role in MMP gene expression in addition to
ofbreastcancercells.Andthisappearedtobethecase.End-point other mechanisms it may influence, such as MMP secretion via
Boydenchamberassaysaccessingcellmotilityinitiallydetectedno vesicle trafficking. We are currently exploring the mechanistic
overall change in migration during a 24hour assay period. This basis for this apparent alteration in MMP gene expression in
was somewhat surprising in light of the work of Achanzar etal. human breast tumor cells.
[45] demonstrating that mutation of fer-1 resulted in profound To authenticate that our shRNA results were not due to off-
defectsinamoeboidmigration ofC.elegansspermcells.However, targeteffectsontheMMPmRNAs,weconductedanextensivein
recentworkfromourlabusinglive-cellimagingclearlyshowsthat silico analysis using several software packages available in the
PLoSONE | www.plosone.org 7 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
evaluatetheinvasivecapacityof231LTV-ctrland231MYOF-KDcells(n=3,
mean6s.d.,Kruskal-Wallistest/Dunn’smultiplecomparisonanalysis).
doi:10.1371/journal.pone.0039766.g006
public domain and verified that none of the shRNAMYOF
constructs predicted the targeting of any nucleic acid sequences
within the human MMP-1 gene. In addition, we found MMP1
expression to be also down-regulated when MYOF expression
was knocked down using an alternative shRNA construct with
similar knock down efficiency.
Another unanticipated finding in the current studies was the
partial reversion of the mesenchymal-like shape in wild-type and
controlMDA-MB-231cellstoamoreepithelial-likephenotypein
231MYOF-KD cells. We verified this potential MET event using
immunofluorescence, scanning, and atomic force microscopy. In
addition, we also demonstrated the likelihood of an MET in the
setting of depleted MYOF using immunoblotting that revealed
down-regulationoffibronectinandvimentinandup-regulationof
E-cadherin.
Inarecentpaperfromourgroup,MYOFdepletionwasshown
to lead to the down-regulation of the phosphorylation of several
receptortyrosinekinases,includingEphB4,FGFR2,Hck,IGF-IR,
JAK2, TXK, VEGFR2 [30]. This observation is in agreement
with the work of Demonbreun etal. demonstrating that loss of
MYOFledtoimpairedinsulin-likegrowthfactorIreceptor(IGF-
IR)signalinginamurinemodel[51].Inseparatereports,MYOF
was shown to be required for Tie-2 (angiopoietin-1 receptor)
activation [25] and vascular endothelial growth factor receptor 2
(VEGFR-2) expression in vascular endothelial cells [24]. More-
over, Sharma etal., showed that disruption of dysferlin, a close
molecular cousin of MYOF, in endothelial cells led to poly-
ubiquitinationandproteasomaldegradationofplateletendothelial
cellular adhesion molecule-1 (PECAM-1/CD31) signaling[52].
These data, whenconsidered collectively and in the context of
ourbreasttumorcellinvestigations,implicateMYOFasacritical
participant in key processes within the intracellular vesicle
trafficking mechanism including cell motility, MMP expression/
secretion,andRTKactivation.Ourworkisthefirstexamination
oftheroleofMYOFincancercellbiologyandmayshedlighton
essential features of vesicle shuttling of important cargos within
tumor cells that participate incancer progression andmetastasis.
Materials and Methods
Cell Culture Methods and Cell Line Authentication
Human MCF-10A (generously provided by Dr. Joan Brugge,
Boston,MA,[32,53]),MCF-7(HTB-22, ATCC, Bethesda,MD),
MDA-MB-231(HTB-26,ATCC),T47D(HTB-133,ATCC),and
BT549 (generous gift from Dr. Lisa Yee, Columbus, OH; HTB-
122, ATCC) cells were used in this study. MCF-10A cells were
maintained in Dulbecco’s Modified Eagle Medium Ham’s F12
(DMEM/F12,Gibco,Carlsbad,CA)supplementedwith5%horse
serum (Gibco), 20ng/ml hEGF (PeproTech, Rocky Hill, NJ),
0.5 mg/ml hydrocortisone (Sigma-Aldrich, St. Louis, MO),
100ng/mlcholeratoxin(Sigma-Aldrich),10 mg/mlbovineinsulin
Figure 6. Myoferlin depletion attenuates MMP1 expression
(Sigma-Aldrich),and1%penicillin/streptomycin(Gibco).MCF-7
and collagen I invasion capacity of MDA-MB-231 cells. (A)
Representative(n=3)immunoblottingresultsofsecretedMMP1in24h andMDA-MB-231cellsweremaintainedinDMEMwith4.5 g/L
serumstarvedsupernatantofMDA-MB-231lentiviralcontrol(LTV-ctrl) D-glucose supplemented with 10% fetal bovine serum (FBS,
andmyoferlinknockdown(MYOF-KD)cells.Recombinanthumanmatrix Gibco).T47DandBT549cellswereculturedinRPMI-1640with
metalloproteinase-1 was used as a standard. Verification of myoferlin 10%FBS.
knockdownwasdoneinthecorrespondingcelllysateswithGAPDHasa
loadingcontrol.(B)Secretedpro-MMP1wasevaluatedin231LTV-ctrland The cell lines used in this study were authenticated by DNA
231MYOF-KD cells (n=2) using ELISA (mean 6 s.d., Kruskal-Wallis test/ profiling using short tandem repeat (STR) analysis on a
Dunn’smultiplecomparisonanalysis).(C)ResultsfromBoydenchamber PowerPlex 1.2 System (Promega, Madison, WI) at John Hopkins
invasion assays using a coating of 3mg/ml of rat tail collagen I to University (Fragment Analysis Facility, Baltimore, MD). Fur-
PLoSONE | www.plosone.org 8 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
thermore, to rule out cell line contamination as the cause for Atomic Force (AFM) and Scanning Electron Microscopy
the morphology changes seen in MDA-MB-231 myoferlin- (SEM)
deficient cells (231MYOF-KD), the identity of the 231MYOF-KD
MDA-MB-231 lentiviral control and myoferlin-deficient cells
cells as MDA-MB-231s was also authenticated through short were seeded in 60mm cell culture dishes (5–86105 cells), and
tandem repeat profiling. allowedtogrowto80–90%confluenceforAFManalysis.Priorto
AFManalysis,culturemediawasexchangedforCO independent
2
Lentiviral shRNA Transduction media (Invitrogen) supplemented with 1% FBS, 1% Penicillin/
Recombinant lentiviral particles containing non-target control Streptomycin and 4 mM L-Glutamine, for optimal cell survival
shRNA and human myoferlin targeted shRNA during the atomic force measurements. A Bioscope II instrument
(TRCN0000010628, TRCN0000010630, TRCN0000001522, (Veeco,Plainview,NY)mountedonthestageofanAxiovert200
and TRCN0000320398, Figure S3) in the pLKO.1 vector were inverted optical microscope (Zeiss, Oberkochen, Germany) was
purchased from Sigma-Aldrich (MISSIONH). The myoferlin used in contact mode for live cell imaging. Silicon nitride (SiN)
constructs are referred to as #10628, #10630, #1522 and triangular cantilevers (Veeco) 200mM in length with a tip angel
#320398, respectively, in the body of the manuscript. For h=35uandanominalspringconstantkof0.01 N/mwereusedto
lentiviral transduction, cells were seeded in 24-well culture plates imagethecells.Imageacquisitionwasdoneatascanningrateof
and incubated overnight at 37uC in 5% CO in a humidified 0.5 Hzwith5126512linesresolutionforrecordingtheheightand
2
atmosphere. Media was replaced with media containing 8mg/ml thedeflectionchannels.Allimageswererecordedwithascansize
hexadimethrine bromide (Sigma-Aldrich) and lentiviral particles area of 70–80mm2 covering ,5 cells, with 4–5 areas completed
wereaddedtosubconfluentatamultiplicityofinfection(MOI)of percelltype.Theexperimentswererepeatedthreetimesandthe
at least 1. After overnight incubation (37uC, 5% CO ), virus- datawereanalyzedusingversion7.30oftheNanoScopesoftware
2
containing supernatant was replaced with complete media and (Veeco).
incubated overnight. Transduced cells were selected in media For SEM, cells cultured on thermoplastic coverslips (Therma-
containing an appropriate puromycin concentration as predeter- noxTM,Nunc,Waltham,MA)werefixedwith3%glutaraldehyde,
mined bya puromycin killcurve. post-fixedwith1%OsO4anddehydratedthroughanethanoland
hexamethyldisilazane series. Samples were mounted onto SEM
studsandpreparedforSEMusingaPelcoModel3sputtercoater
RNA Extraction and Quantitative Real Time-Polymerase
with gold-palladium (0.07 mbar, 17mA, 110s) and imaged with
Chain Reaction (qRT-PCR)
an FEI Nova NanoSEM microscope (FEI Company, Hillsboro,
Total RNA was extracted using TRIzolH reagent (Invitrogen, OR).
Carlsbad,CA)accordingtothemanufacturer’sprotocoluptothe
chloroform extraction and centrifugation step. The resulting Protein Extraction and Immunoblotting
aqueous phase was mixed with an equal volume of 70% ethanol
CellswererinsedwithPBSandlysedfor30minincoldRIPA
andappliedtoanRNeasyminicolumn(QIAGEN,Valencia,CA)
lysisbuffer supplemented withprotease inhibitor cocktail (Sigma-
and processed according to the manufacturer’s protocol with on
Aldrich)and1mg/mlofpepstatinA(USB,SantaClara,CA).Cell
column DNAse digestion. Total RNA was quantified by UV
lysates were centrifuged for 10min at 17,0006g and 4uC.
absorbance at 260nm and 280nm on a NanoDrop 2000
Supernatantaliquotswereassayedforproteinconcentrationusing
(Thermo Scientific, Waltham, MA), and 1–2mg total RNA from
the BCA Protein Assay (Pierce, Rockford, IL) with bovine serum
each sample were reverse-transcribed to cDNA using oligo-dT18
albuminasa standard.Totalproteinfromeachlysate(30 mg per
and random hexamer primers with a Transcriptor First Strand
lane) were resolved by SDS-PAGE and transferred to nitrocellu-
cDNA Synthesis Kit (Roche Applied Science, Indianapolis, IN).
losemembranes.Non-specificbindingwasblockedbyincubation
Quantitative RT-PCR was performed using an equal amount of
inTris-bufferedsaline(pH8)containing0.1%Tween-20and5%
cDNA per sample on a LightCyclerH 480II System (Roche
non-fat dry milk. Membranes were probed with primary
Applied Science) using primer-probe sets specific to MYOF, antibodies against: b-actin (Santa Cruz Biotechnology, Santa
MMP1, RLPO and 18S rRNA (Applied Biosystems, Carlsbad,
Cruz,CA),GAPDH(Chemicon,Billerica,MA),myoferlin(Sigma-
CA) with the LightCyclerH 480 Probes Master Mix, and results
Aldrich), E-cadherin(EP700Y, Abcam, Cambridge, MA), vimen-
analyzed withtheLightCyclerH480Software.
tin (Sigma-Aldrich V6630), or MMP1 (R&D Systems, Minneap-
olis,MN)dilutedinblockingsolution.Afterwashing,membranes
Immunofluorescence were exposed to horseradish peroxidase-conjugated secondary
MDA-MB-231 cells cultured on glass coverslips and serum- antibodies and immune complexes were revealed with Super-
starved for 24h were fixed for 1 h in 4% (w/v) paraformalde- SignalWestFemtochemiluminescentsubstrate(Pierce),visualized
hyde/PBS, permeabilized for 15min with 0.2% (v/v) Triton X– usingtheVersaDocImagingSystemandanalyzedwithQuantity
100/PBS, and blocked for 1 h in 5% normal goat serum/1% Oneanalysissoftware (Bio-Rad, Hercules, CA).
nonfat dry milk/PBS. Samples were incubated with antibodies
directed against myoferlin (1:400, Sigma) overnight at 4uC. After Supernatant Collection and Immunoblotting
washing, cells were exposed to fluorochrome-conjugated second- Cellconditionedmediawerecollectedafterthecellswereplated
aryantibodies(MolecularProbes).Afterfurtherstringentwashing atequaldensitiesin100mmdishes,thenserumstarvedin4 mlof
in PBS, actin cytoskeleton was stained with AlexaFluor 488- basal media for 24h. The conditioned media were concentrated
phalloidin(MolecularProbes)for20min.Nucleiwerestainedand 50-foldwithAmiconUltra-410-kDacentrifugalfilters(Millipore,
cellsmountedwithProLongHGoldAntifadeReagentwithDAPI Billerica, MA), and equal volumes of concentrated media were
(Invitrogen)andvisualizedwithconfocallaserscanningmicrosco- reconstituted in sample buffer and resolved by SDS-PAGE.
py (Olympus FV 1000 Spectral Confocal system; Olympus HumanrecombinantMMP1(rhMMP-1)fromconditionedmedia
America Inc.,Center Valley, PA). of rhMMP-1 expressing NS0 mouse myeloma cells (WBC024;
R&DSystems)wasusedasapositivecontrol.Gelswereprocessed
PLoSONE | www.plosone.org 9 June2012 | Volume 7 | Issue 6 | e39766
MyoferlininBreastCancerEMT&Invasion
perimmunoblottingproceduresasdetailedintheprevioussection, conducted in the same manner but with the addition of 20 ml of
using anti-human mouse monoclonal antibodies against MMP-1 growth factor reduced Matrigel (BD) or 3 mg/ml collagen I (rat
(MAB901; R&D Systems) and fibronectin (610077; BD Trans- tail, Gibco) to the top of the insert and allowed to gel for 1h at
duction Laboratories, San Jose,CA). 37uC,5%CO .
2
MMP Expression Array and MMP1 ELISA Quantitative Reverse Transcription PCR Array
ECMproteins(MMP1,-2,-3,-8,-9,-10,and-13andTIMP1,- RT2ProfilerTMPCRarray(SABiosciences,Frederick,MD)was
2, -3, and -4) in conditioned, serum starved supernatants of performed for EMT-associated genes (cat #PAHS-090) and
lentiviral-transduction control (MDA-MB-231LTV-ctrl) and MDA- human extracellular matrix and adhesion molecules (cat #
MB-231MYOF-KD cells were determined by an antibody array PAHS-013) following manufacturer’s protocol. Briefly, total
(RayBioH human matrix metalloproteinase antibody array 1, RNA was extracted as described in the qRT-PCR section from
RayBiotech, Inc., Norcross, GA) following the manufacturer’s lentiviralcontrolandmyoferlin-deficientMDA-MB-231cellsthat
protocol. Each array membranes contained duplicates for each had been in culture for 10consecutive passages. Complementary
ECMproteinalongwithpositiveandnegativecontrols.Theblots DNA was generated from 1mg of total RNA per cell type using
were imaged and quantified by densitometry (Quantity One randomhexamersandoligo-dT18primersaspartoftheRT2First
software, Bio-Rad), and values normalized to total protein of the Strand Kit (SABiosciences). Equal amounts of dilute cDNA was
corresponding celllysates. mixedwithLightCyclerH480SYBRGreenIMastermix(Roche)
Secretedpro-MMP-1inculturedmediaofcellswasdetectedby andaliquotedtoeachwellofthePCRarrayplatecontainingpre-
quantitative ELISA (DMP100; R&D Systems). Conditioned filledgene-specificprimersets,andPCRwasperformedaccording
cultured media were collected by plating equal densities of each to manufacturer’s instructions for the Roche LightCyclerH 480.
cell type in 100mm culture dishes followed by serum-starvation TheLightCyclerH480Software(RocheAppliedScience)wasused
for24h.Unconcentrated,conditionedmediasampleswererunin to calculate the threshold cycle (crossing point, Cp) values for all
duplicateintheELISAandprocessedaccordingtomanufacturer’s the transcripts in the array. The Cp values were exported into a
instructions. Colorimetric results were read at 450nm with a spreadsheet-based PCR array data analysis template (SABios-
wavelength correction of 570nm in a MRX Microplate Reader ciences) to calculate fold changes in gene expression using the
(Dynex Technologies, Chantilly, VA). MMP-1 concentrations DDCtmethod.
weredeterminedbyinterpolationusinghumanrecombinantpro-
MMP-1asastandard.Incaseswherethesecretedpro-MMP1was Statistical Analysis
undetectable, it was assigned the assay’s lowest limit of detection
Graphical representation and statistical analysis of data were
(0.021 ng/ml).
done using Prism version 5 (GraphPad). Error bars represent
standard deviationinallcases,unlessdifferentlynoted.Statistical
MTS Proliferation Assay and Cell Cycle Analysis
significance was determined by appropriate statistical tests
Cells wereplatedin96-wellplatesatadensityof2500and/or followingnormalitytestingofthedata,usingtheKS,D’Agostino
5000 cells per well and assayed at 24, 48, and 72h after plating & Pearson omnibus and/or Shapiro-Wilk normality tests. Com-
using the CellTiter 96 AQueous Non-Radioactive Cell Prolifer- parisonsbetween2groupswithGaussiandistributionwerecarried
ation Assay (Promega), according to the manufacturer’s instruc- out with the 2-tailed student t-test for unpaired samples, while
tions. those among 3 or more groups were carried out with one-way
For flow cytometric analysis of MYOF depleted breast cancer ANOVA followed by Tukey’s Multiple Comparison Test. Data
cells,cellswerefixedin70%ethanolandstainedinPBScontaining exhibiting a non-Gaussian distribution were analyzed with either
0.1% Triton X-100, 200mg/mL of RNase, and 50mg/mL theMannWhitneystatisticaltest(2groups)ortheKruskal-Wallis
propidiumiodide(Sigma-Aldrich).DNAcontentwasmeasuredon withDunn’smultiplecomparisonpost-test(3ormoregroups).For
aFACSCaliburorLSRIIflowcytometer(Becton-Dickinson)),and analysis, results were regarded as significant if p values were less
data were acquired and analyzed using flow cytometer software than 0.05.
(BD).Eachanalyzedsamplecontainedatleast16106cells.
Supporting Information
Migration and Invasion Assays
Cells were washed with PBS, trypsinized, pelleted and Figure S1 Analysis of the myoferlin gene expression in
resuspended in serum-free DMEM. Cells were then seeded onto breastcancercells.Graphicalrepresentation(mean6s.d.)of
Boyden Chamber inserts (8 mm pores; Millipore) at a density of expression data from two microarray probe sets for myoferlin
7.56104cellsin24-wellplateswith10%FBScontainingDMEM reported in the study by Neve and colleagues [34], showing a
in the bottom chamber as a chemo-attractant. In parallel, input higher expression (Kruskal-Wallis test/Dunn’s multiple compari-
controls were seeded with the same number of cells in the same sonanalysis)intheBasalBcellscomparedwiththeluminalcells.
volumesofserum-freeand10%FBSmediawithoutinserts.Cells (TIF)
were incubated (37uC, 5% CO ) for 24 hours. The inserts were
2 FigureS2 GraphicalrepresentationfromArrayExpress
processedbyrinsinginPBSandfixationwith3.7%formaldehyde
gene expression atlas (accession # E-TABM-276) of
containing0.05%crystalvioletfor10min.Afterrepeatedwashes
MYOF mRNA expression level in breast tissue samples
with PBS and distilled water the chambers were air-dried.
from healthy patients and patients with invasive ductal
Migrated cells on the bottom of the inserts were collected with
carcinoma.
cottonswabs andplaced intoEppendorf tubes.Crystal violet dye
(TIF)
was extracted from the Q-tips and input control wells with 80%
methanol for 30min, and quantified at 570nm. Percentage Figure S3 Myoferlin lentiviral constructs. Target and
migrationwascalculatedusingtheratioofthemigratedcellsover sequence information of lentiviral constructs used to generate
the total cells (input control) to determine the percentage of cell myoferlin-deficientcelllines.Boldlettersindicatecodingsequence.
migration. For invasion, the Boyden Chamber assay was (TIF)
PLoSONE | www.plosone.org 10 June2012 | Volume 7 | Issue 6 | e39766