Article Interkinetic Nuclear Migration Is Centrosome Independent and Ensures Apical Cell Division to Maintain Tissue Integrity Graphical Abstract Authors PaulinaJ.Strzyz,HyunO.Lee,..., LouisC.Leung,CarenNorden Correspondence [email protected] In Brief Strzyzetal.showthat,inzebrafish, pseudostratifiedretinalneuroepithelia apicalnuclearmigrationpriortomitosisis ahighlyreproduciblephenomenon.It doesnotdependoncentrosomenumber, location,integrity,orpositionofmitotic entry.Thisensuresthatproliferativecells robustlydivideapicallyandsafeguards tissuearchitectureandmaturation. Highlights d ApicalIKNMishighlyrobustandoccursindependentlyofthe centrosome d CDK1activityisnecessaryandsufficienttodriveapicalIKNM d Nonapicaldivisionsperturbintegrityofthepseudostratified neuroepithelium d Apicallocalizationofdivisionssafeguardstissuearchitecture andmaturation Strzyzetal.,2015,DevelopmentalCell32,203–219 January26,2015ª2015ElsevierInc. http://dx.doi.org/10.1016/j.devcel.2014.12.001 Developmental Cell Article Interkinetic Nuclear Migration Is Centrosome Independent and Ensures Apical Cell Division to Maintain Tissue Integrity PaulinaJ.Strzyz,1HyunO.Lee,1JaydeepSidhaye,1IsabellP.Weber,1LouisC.Leung,1andCarenNorden1,* 1MaxPlanckInstituteofMolecularCellBiologyandGenetics,Pfotenhauerstraße108,01307Dresden,Germany *Correspondence:[email protected] http://dx.doi.org/10.1016/j.devcel.2014.12.001 SUMMARY the millimeter scale in the primate neocortex (Rakic, 1972). Nucleiofthesecellslocateallalongtheapicobasalaxisresult- Pseudostratified epithelia are widespread during inginthestratifiedappearanceofthetissue(LeeandNorden, animal development and feature elongated cells 2013).Intheproliferativestate,whenPSEexpandandcellsdo whose nuclei adopt various positions along the api- not yet differentiate, divisions take place at apical positions cobasal cell axis. Before mitosis, nuclei migrate with cleavage planes perpendicular to the apical surface (Das towardtheapicalsurface,andsubsequentdivisions et al., 2003; Morin et al., 2007; Nakajima et al., 2013; Weber etal.,2014;Xieetal.,2013).Itisproposedthatsuchperpendic- occur apically. So far, the exact purpose of this nu- ular division angles are crucial for the maintenance of tissue clear migration remained elusive. One hypothesis integrityinthePSEofDrosophilaimaginaldisk(Nakajimaetal., was that apical migration ensures that nuclei and 2013) and chick neural tube NE (Morin et al., 2007). In rodent centrosomesmeetformitosis.Weheredemonstrate NE,includingretinaandhindbrain,however,anincreaseinnon- thatinzebrafishneuroepitheliaapicalnuclearmigra- perpendicular divisions has been observed upon the onset of tionoccursindependentlyofcentrosomepositionor neurogenesis(CayouetteandRaff,2003;Kosodoetal.,2004). integrity. It is a highly reproducible phenomenon ApicalmitosesobservedinallPSEresultfromapicalnuclear linkedtothecellcycleviaCDK1activity.Wepropose migration in G2 (Kosodo et al., 2011; Leung et al., 2011). that the robustness of bringing nuclei apically for Followingapicaldivision,nucleiaredistributedmorebasallyuntil mitosis ensures that cells are capable of reintegrat- thecompletionofSphase,afterwhichtheyreturntotheapical ingintotheepitheliumafterdivision.Nonapicaldivi- sideanddivideagain.Thisbidirectionalmovementofnucleiis knownasinterkinetic nuclearmigration(IKNM).Since itsinitial sions lead to cell delamination and formation of cell observation in 1935(Sauer,1935), IKNMhas beenextensively clusters that subsequently interfere with neuronal studied in many model organisms and tissues (Grosse et al., layering. Therefore, positioning divisions apically in 2011; Meyer et al., 2011; Norden et al., 2009; Rujano et al., pseudostratified neuroepithelia could serve to safe- 2013;Tsaietal.,2010).Thecytoskeletalmachineriesthatdrive guardepithelialintegrityandenableproperprolifera- apicalIKNMvarydependingontissuetype.Inshorterepithelia tionandmaturation. (i.e.,NEofthezebrafishorDrosophilaPSE),forcesformoving nucleiaregeneratedbyactomyosincontractions(Leungetal., 2011; Meyer et al., 2011; Norden et al., 2009; Rujano et al., INTRODUCTION 2013). Inthe more elongated mammalian neocortex, however, mainly microtubules (MTs) and their motors are involved (Hu Pseudostratified epithelia (PSE) are found during the develop- etal.,2013;Kosodoetal.,2011;Tsaietal.,2010).Collectively, mentofmanyorganisms.Theygiverisetovarioustissuesina thesestudiesprovidedsubstantialmechanisticinsightsintonu- widerangeofanimals,includinginvertebrateslikeNematostella clear kinetics and the cytoskeletal machineries driving IKNM. andDrosophila,aswellasvertebratesincludingzebrafish,chick, Whatissofarlessexploredthoughisthebasicfunctionofapical rodents,andhumans(Grosseetal.,2011;Kosodoetal.,2011; IKNM.Wefurtherlackanexplanationwhydivisionsoccurexclu- Meyeretal.,2011;Nordenetal.,2009;Rujanoetal.,2013;Spear sively at the apical surface. One study suggests that apical andErickson, 2012).Pseudostratified neuroepithelia (NE)have mitosisisrelatedtotheexistenceofamitoticzoneattheapical gatheredspecialattention,astheygeneratemanypartsofthe endfoot (Hu et al., 2013). According to another hypothesis, nervoussystem. apicalIKNM ismainlya result of theapical position ofcentro- CellswithinPSEareattachedtobasalandapicallaminaeand somesduringinterphase,astheseorganellesareimportantfor arehighlyelongated.Theircentrosomesarelocalizedtotheapi- spindleformationbutalsoserveasbasalbodiesforprimarycilia cal surface during the whole cell cycle (Miyata, 2008; Norden (Miyata,2008;TavernaandHuttner,2010).Additionally,thepo- et al., 2009; Tsai et al., 2010). Cell lengths range from 40– sitionofthenucleusbeforeapicalIKNMhasbeenlinkedtothe 50 mm in NE of zebrafish or the imaginal disk of Drosophila likelihoodofneurogenesis:Twostudiesprovidedevidencethat (Meyer et al., 2011; Norden et al., 2009) to 250 mm in rodent the basal depth of nuclei before apical IKNM can be linked to neocortex (Hu et al., 2013; Kosodo et al., 2011) and up to the probability of a neurogenic division in correlation with an DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. 203 apicobasalNotch/Deltagradient(BayeandLink,2007;DelBene Movie S1 (available online), in the zebrafish retinal NE, the etal.,2008).Sofar,however,allthesestudiesdealtwiththepro- centrosomebecomesmotileonlyshortlybeforethenucleusrea- cessofapicalIKNMitselfbutdidnotinvestigatetheimpactof ches the apical endfoot. On average, the centrosome travels apicaldivisionsontissuedevelopmentandmaturation. 6.7mmtowardthenucleus(Figure1E,n=51cells,fourembryos), Herewesetouttoexploretheimportanceofapicaldivisions similartoresultsintherodentneocortex(Huetal.,2013).This for tissue maturation in proliferating pseudostratified zebrafish meansthatcentrosomesmaintaintheirapicallocalizationduring retinalNE.Weshowthatincontrasttorecentfindingsinrodent thewholecellcycleandthatthemigrationofnucleiresultsinnu- neocorticalPSE(Huetal.,2013)mitoticentryisnotrestrictedto cleus-centrosomeassociation. theapicalendfoot.Wefurtherdemonstratethattheapicallocal- IfindeedthemainreasonfortheoccurrenceofapicalIKNM izationofthecentrosomeisnotaprerequisiteforapicalnuclear wastoensurethatthecentrosomeandthenucleusmeetbefore migration, as apical IKNM still occurs when centrosomes and division, apical IKNM should not take place when the nucleus nucleimeetnonapicallyorincaseswhencentrosomeintegrity and centrosome associate nonapically. To test this idea, we isperturbed.Interestingly,apicalmigrationeventakesplaceaf- introduced additional nonapical ‘‘centrosomes,’’ by interfering ternonapicalmitoticentry.Thisfindingimpliesthatonceapical with the centriole duplication pathway. To allow for temporal IKNMistriggered,a‘‘pointofnoreturn’’independentofcentro- control we used heat shock (HS)-inducible constructs (Clark somepositionispassed.Wefindthatthisupstreamtriggeristhe etal.,2012;Nordenetal.,2009).Wecoexpressedanoverstabi- activation of CDK1, which is necessary and sufficient for the lized(OS)versionofPlk4—thekeyregulatorofthecentrosome onsetofapicalIKNM.Wealsorevealthatlocalizingallmitoses duplicationpathway(Hollandetal.,2010)andadominant-nega- anddivisionsapicallyisofgeneralimportanceforproperprolif- tive(DN)formofCep152,whichperturbscentriolarrecruitment eration and integrity of pseudostratified NE, as induction of of Plk4 (Cizmecioglu et al., 2010; Coelho et al., 2013) (Fig- nonapicaldivisionsperturbsretinaldevelopment.Wetherefore ureS1A).Incellsexpressingbothconstructsfusedtothefluo- suggest that apical migration of nuclei and subsequent apical rescentproteinmKate2,weobservedtheformationofnonapical divisionsaremechanismswhoserobustnesssafeguardstissue centrosome-like structure appearing as mKate2 positive foci integrityandthusrepresentsimportantfirststepstoorchestrate thatwereg-tubulinpositive(Figure2A).Thesefocirecruitedcen- tissuematuration. trin as well as centrosome targeted GFP-PACT (Figures S1B andS1C).Additionally,liveimagingofEB3-GFPdemonstrated RESULTS that they can act as microtubule organizing centers (MTOCs) as nonapical MT nucleation was observed (Figures 2B and MitoticEntryCanOccurNonapicallyinZebrafish S1D; Movie S2). Together, this shows that these foci indeed RetinalNE functionallyresemblecentrosomes. WeusedthedevelopingzebrafishretinalNEtostudytheoccur- We next investigated whether the interaction of nuclei with renceandreproducibilityofapicalIKNMandapicalmitosesin suchnonapicalcentrosomestriggersnonapicalmitosisandsub- PSE. We first asked whether in zebrafish NE mitotic entry is sequently impairs apical IKNM. Intriguingly, this was not the restrictedtotheapicalendfootaspreviouslysuggestedforthe case. While we observed MT nucleation emanating from the rodentneocortex(Huetal.,2013).Ifthiswerethecase,apical centrosome-likestructures(Figure2B),nucleistilldisplayedapi- IKNMwouldbeindispensableformitoticentryandsubsequent calmovementwithsimilarkineticsasinthecontrolsituation(Fig- divisions.Asactomyosincontractionsarethemainforcegener- ures 3C and 3D; Movie S2). Furthermore, all MTOCs of a cell atorsduringapicalIKNMinzebrafishNE,weblockedactomy- could often be observed to cluster at the apical surface (Fig- osin contractility using the myosin II inhibitor blebbistatin ure S1D). pH3staining confirmedthatmitotic figures localized (Nordenetal.,2009).Notably,followingdrugtreatment,rounded apically (n = 28 of 29, seven embryos) (Figure 3A). Altogether cellsthatassociatedwithcentrosomeswereobservedatnonap- this demonstrates that enabling the mere interaction between icallocations(Figure1A).Thesecellsstainedpositivelyforthe thenucleusandanMTOCatnonapicalpositionsdoesnotper- mitoticmarkerphospho-Histone3(pH3),andapicalaswellas turbapicalIKNM. basalprocessesofcellscouldbeobserved,arguingagainsta Intheconditioninwhichweinducednonapicalcentrosomes, puredelaminationphenomenon(Figures1Band1C). however, the apically localized centrosome was still present. WeconcludethatinzebrafishNEmitoticentryisnotrestricted Therefore, we could not exclude the possibility that the apical to the apical zone. This raises the question why nevertheless position of this centrosome prompted nuclear migration. To nucleimigrateapicallybeforemitosis. investigate this option, we aimed to induce nonapical centro- some-nuclear association in cells that featured normal centro- ApicalIKNMStillOccurswhenNucleusand some number. We noticed that interference with N-cadherin CentrosomesMeetNonapically via a DN construct (Wong et al., 2012), as well as colcemide Wenextaimedtotestthepreviouslyraisedhypothesisthatapi- treatment can lead to nonapical association of centrosomes calIKNMisaresultoftheapicallocationofcentrosomesinPSE andnuclei(Figures2C,2D,andS2A–S2C).Therefore,weused (Miyata,2008;TavernaandHuttner,2010)andmainlyservesto these approaches to investigate whether nonapical centro- bring nuclei and centrosomes into close proximity before divi- some-nucleus association influences apical IKNM. In both as- sion.Wefirstanalyzedthenuclearandcentrosomal dynamics says, cells kept their apical and basal attachments arguing during the cell cycle using proliferating cell nuclear antigen against mere cell delamination (Figures S2A, S2B, and S2D). (PCNA),whichunambiguouslylabelsallcellcyclephases(Leung Interestingly, in the DN-N-Cadherin as well as the colcemide et al., 2011). As seen in the typical example in Figure 1D and condition nuclei that got in contact with centrosomes at basal 204 DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. A B C E D Figure1. MitoticEntryIsNotRestrictedtotheApicalSurfaceinZebrafishRetinalNE;CentrosomesAreMaintainedApicallythroughoutthe CellCycle (A)ConfocalscanofacellexpressingRas-mKate2(green)andcentrin-GFP(magenta)inanembryotreatedwith200mMblebbistatin1.5hrbeforeimaging.The cellshowsnonapicalmitoticrounding(filledarrow)andisassociatedwithcentrosomes(openarrows). (B)ConfocalscanofacellexpressingRas-GFP(green)stainedforpH3(magenta)inanembryotreatedwith200mMblebbistatin1.5hrbeforefixing.Thecell featuresnonapicalpH3signal(filledarrow)whilemaintainingapicalandbasalattachments(openarrows). (C)Confocalscansoftheretinaeofanembryotreatedwith200mMblebbistatin1.5hrbeforefixing(lower)andacontrolembryo(upper)stainedforpH3(magenta). Blebbistatintreatedcellsentermitosisnonapically(lower,arrows). (D)Time-lapseofthedynamicsofthenucleus-centrosomepairwithrespecttocellcycleprogression.PCNA-RFPlabelsnucleiandmarksthecellcyclestage (gray).Ras-mKate2(gray)labelscellmembranes.Centrin-GFP(red)labelscentrosomes.Onenucleusislabeledwithayellowdot.Thearrowhighlightsthe positionofthecentrosome.Timeisinhr:min.TheframesarefromMovieS1. (E)Centrosomepositionpriortocentrosomesplitting(left)andthemeanvalueofthepositionofcentrosomesatmitosis(middle)withrespecttonuclearlength (right),n=51cells,4embryos. Scalebarsrepresent10mm.Thedottedlinerepresentstheapicalsurface,andthesolidlinerepresentsthebasalside. positions entered mitosis basally, as demonstrated by the mitoticentry,apicalIKNMstilloccurred.Interestingly,whencen- appearance of cell rounding, chromosome condensation, and trosomes were mispositioned in the DN-N-Cadherin or colce- nuclearenvelopebreakdown(NEB)(Figures2Cand2D;Movie mide condition, apical IKNM typically started shortly before or S3).Remarkably,however,despite thepossibilityof nonapical evenaftermitoticentry(Figures2C,2D,3E,and3F).Incontrast, DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. 205 A B C D Figure2. ApicalIKNMPersistsinCellsinwhichCentrosomesandNucleusMeetNonapicallyEvenfollowingNonapicalMitoticEntry (A)ConfocalscanofaHS-DN-Cep152-mKate2/HS-OS-Plk4-mKate2expressingembryo.PositivecellsfeaturecytosolicmKate2signalaswellasfociofsignal alongtheapicobasalaxis(magenta).Fociarepositiveforg-tubulinstaining(green).NonapicalfocipositiveformKate2andimmunopositiveforg-tubulinare markedwithyellowarrows.HSwasperformed8hrpriortofixing.SeealsoFigureS1. (legendcontinuedonnextpage) 206 DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. incontrolcells,mitoticentryonlytookplaceoncetheapicalside ibility is achieved independently of centrosomes. As IKNM is was reached (Figure 3C). pH3 staining of embryos expressing tightly linked to the cell cycle, we aimed to elucidate the up- DN-N-Cadherinshowedthatmitoticfigureswerefoundatapical streamcellcyclecomponentsresponsibleforIKNMrobustness positions(n=36of36,13embryos)(Figure3B).Thisunderlines andreproducibility.ItisknownthatapicalIKNMinvariablyoc- thatevenincasesinwhichcentrosomeandnucleicanmeetnon- cursaftercompletionofSphase,whichisinG2(Kosodoetal., apically, apical IKNM and apical division still take place. 2011;Leungetal.,2011).Additionally,itisknownthatcellcycle TogetherthissuggeststhatinitiationofapicalIKNMoccursinde- progression is tightly coordinated by the activity of cyclin- pendentlyofapicalcentrosomeposition.Additionally,itimplies dependentkinases(CDKs)andtheirassociatedcyclins(Fisher that the apical position of the centrosome cannot be the sole etal.,2012).EntryintoG2isregulatedbytheactivityofCDK1 reasonfornucleitomigrateapicallypriortocelldivision. in complex with cyclins A or B (Pines and Rieder, 2001). We therefore hypothesized that CDK1 is the cell cycle-associated CentrosomeIntegrityIsDispensableforApicalIKNM molecule that controls the timing of apical IKNM. This predic- SofarwehavedemonstratedthatapicalIKNMisstillreproduc- tionwassupportedbyresultsfromourpreviousworkshowing ibly initiated when centrosomes and nuclei associate nonapi- thatcellcyclearrestinducedbyCDK1inhibitionleadstostalled cally.Anindicationthatcentrosomesmightnotbeessentialfor IKNM (Leung et al., 2011). To reproduce these findings, we apicalnuclearmigrationingeneralcamefromapreviousstudy used the CDK1 inhibitor RO3306 (Vassilev, 2006). Indeed, in Drosophila. This study showed that flies without centrioles upon CDK1 inhibition, cells completed S phase, as indicated developrelativelynormally(Bastoetal.,2006)despitefeaturing by the disappearance of nuclear PCNA foci, but stalled in G2 epithelia that display IKNM (Meyer et al., 2011; Rujano et al., without initiating apical IKNM (Figures 4C, 4E, and 4F; Movie 2013). We therefore tested whether centrosomal integrity is a S5). This is in contrast to controls in which apical IKNM prerequisitefornuclearapicalmigrationinthezebrafishNE.To occurred upon the disappearance of nuclear foci (Figures 1D achievethis,weperformedlaserablationofcentrosomesinsin- and 4B; Movie S5). This suggested that CDK1 activity is glecellsusingcentrinlabelingasareadout.Nucleardynamics required to induce apical nuclear movements. Next, we were followed after ablation, and cells were grouped into two explored whether CDK1 activation is also sufficient to trigger categories:(1)cellsinwhichsomecentrinsignalremainedbut apicalIKNM.Tothisend,wemadeuseofthefactthattheac- was much weaker and/or more diffuse; these cells most likely tivity of CDK1 is blocked during S phase due to an inhibitory featuredonlyremnantsofcentrosomalmaterial(ninecells,eight phosphorylation by Wee1 kinase (Tang et al., 1993). Inhibiting nucleiperformedapicalmigration)and(2)cellsinwhichnocen- Wee1activityhasbeenshowntoinduceprecociousCDK1ac- trinsignalwasobservedafterablation(fivecells,fivenucleiper- tivity (McGowan and Russell, 1995). To test whether inducing formed apical migration). This shows that the vast majority of earlyCDK1activationprematurelytriggersapicalIKNM,wein- nucleifollowingcentrosomeablationinitiatedmovementtoward hibited Wee1 using PD0166285 (Leijen et al., 2010). Remark- the apical side. Some cells subsequently did not progress ably, Wee1 inhibition consistently caused nuclear migration to throughmitosisandremainedroundedatapicalpositions(Fig- occur already during S phase (Figures 4D–4F; Movie S5), ure4A;MovieS4),whereasothersmanagedtocompleteapical showing that indeed CDK1 activation alone is sufficient to division(MovieS4).Collectively,thisarguesthateventhepres- trigger apical nuclear movement independent of cell cycle enceofanintactcentrosomeisnotessentialforapicalmigration phase. ofnuclei. Together,ourresultsthereforestronglyarguethatCDK1ac- Takentogether,inallexperimentalsettingsinwhichweinter- tivity is necessary and sufficient to initiate apical IKNM. They fered withcentrosome position orcentrosome integrity, apical also offer an explanation as to why apical IKNM occurs even IKNMstilloccurred.ThisstronglyarguesthatapicalIKNMinze- when nuclei and centrosomes meet nonapically or when brafishNEisaparticularlyrobustphenomenonthatstilloccursin centrosome integrity is impaired. Once cells finish S phase, challengedconditionsindependentlyofapicalcentrosomes. CDK1 activation most likely leads to a ‘‘point of no return’’ for apical movement. Apical IKNM subsequently occurs ApicalIKNMIsTriggeredbyCDK1Activity evenincasesinwhichnucleialreadyassociatedwithcentro- OurexperimentsshowthatapicalIKNMisahighlyrobustphe- somes and initiated NEB or chromosome condensation nomenon. Next we wanted to understand how this reproduc- nonapically. (B)Time-lapseofacellexpressingHS-DN-Cep152-mKate2,HS-OS-Plk4-mKate2(green),andthedynamicMTsmarkerHS-EB3-GFP(gray).(Left)Cytosolic mKate2signalandnonapicalfoci(cyanoutlinedarrows)inthecellofinterestareshown.TheremainingpanelsshowdistributionofdynamicMTs.Clearfociof nonapicalMTnucleationcanbeobserved(yellowarrows).Intheinsets,magnifiedregionsofnonapicalMTnucleationsitesareshowninfirelookuptable(in panels4and7,1.53magnification,remainingpanels33magnification).Thereddotmarksthepositionofthenucleus.HSwasperformed12.5hrpriortotime- lapse.Timeisinhr:min.TheframesarefromMovieS2. (C)Time-lapseofacellexpressingDN-N-Cadherin.HS-H2B-RFPlabelsnuclei/chromatin(magentaintheupperandgrayinthelower)andcentrin-GFP-RNA labelscentrosomes(green,upperpanelonly).Thecentrosome(yellowarrow)associateswiththenucleus(cyan,arrow)inanonapicalposition(upper).Thecell entersmitosisnonapicallyasvisualizedbychromosomecondensation(lower,00:40,cyanarrow).ApicalIKNMandapicaldivisionoccur.HSwaspreformed 17.5hrpriortotime-lapse.Timeisinhr:min.TheframesarefromMovieS2.SeealsoFigureS2. (D)Time-lapseofacellinanembryotreatedwith100mMcolcemide.ThecellexpressesRas-mKate2(gray)andcentrin-GFP(green).Inthecellofinterestthesplit centrosome(yellowarrows)travelsbasallyandassociateswiththenucleus(yellowdot)nonapically.Afternonapicalcellrounding(01:50),thecellperformsapical IKNM.Time-lapsewasstarted5hrafterdrugaddition.Timeisinhr:min.TheframesarefromMovieS3. Scalebarsrepresent10mm.Thedottedlinerepresentstheapicalsurface. DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. 207 A B C D E F Figure3. ApicalIKNMPersistsDespiteNonapicalMitoticEntry (AandB)pH3stainings(green)ofembryosexpressingHS-DN-Cep152-mKate2/HS-OS-Plk4-mKate2(A)orDN-N-Cadherin(B)(bothmagenta). (C–F)Tracksofnuclearpositionaroundmitosisin(C)controlcells,(D)cellsfeaturingnonapicalcentrosomes,(E)cellsexpressingDN-N-Cadherin,(F)cellstreated with100mMcolcemide.Timeshownintimepoints(tp).1tp=5min.Nuclearpositionwastracked10tp(=50min)priortoand10tpaftermitoticentryoruntil division.Nuclearpositionhasbeenmeasuredfromthebaseofthenucleustothebasallaminaandwasnormalizedwithrespecttocelllength.Eachtrack representsasinglenucleus;tpofmitoticentryismarkedwithblueline(alwaysattp=11). Scalebarsrepresent10mm. ApicalNonperpendicularDivisionsDoNotPerturb integrityofDrosophilaandchickPSE(Morinetal.,2007;Naka- RetinalTissueArchitectureandMaturation jima et al., 2013). Therefore, one assumption was that apical Atthispoint,weunveiledthatapicalIKNMisaparticularlyrepro- IKNM is a prerequisite for successful tissue development of ducibleeventthatdependsonCDK1activityandtherebyoccurs PSEbyenablingthecontrol ofcleavage planesatdivision.To eveninchallengedconditions.Wenextaimedtofindoutwhether investigatewhetherthisisthecase,weexaminedhowinterfer- the reproducibility of bringing all nuclei apically for division is encewithapicalcleavageplanesimpactstheintegrityofretinal importantfortissuedevelopmentperse.Onepossibleexplana- NE.Weusedpreviouslypublishedmorpholinos(MOs)targeting tionisthatonlyatapicallocationscancleavageplanesbepre- aPKCl/zasthesehavebeenshowntoinducenonperpendicular cisely controlled leading to perpendicular divisions. In support divisions(Cuietal.,2007).aPKCl/zMOswereinjectedmosai- ofthisidea,ithasbeenpreviouslysuggestedthatperpendicular callyinto8-to32-cellstageembryostogetherwithfluorescently cleavage planes are implicated in the maintenance of tissue tagged RNAs to label affected cells (Norden et al., 2009). We 208 DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. observedthatdivisionstookplaceapicallyat34hourspostferti- aPKCwaslinkedtoaCAAXdomain,thustargetingtheproteinto lization(hpf)(Figure5A).Thisisincontrasttoapreviousreport themembrane(Ossipovaetal.,2007).Inductionofthisconstruct usingthesameMOcombination(Cuietal.,2007).Themostlikely led to a gross alteration of aPKC distribution. The proteinwas reasonsforthisdiscrepancyaredifferencesineffectiveMOcon- observed to localize to cell membranes apically as well as all centrationorthefactthatweinjectedMOsatthe8-to32-cell alongtheapicobasalcellaxis(Figures6Band6C).Becauseof stage to avoid early morphological defects as opposed to the the involvement of aPKC in epithelial cell polarity, we investi- 1-cellstageinthepreviousstudy.Inaccordancewiththework gated whether polarity components still localized correctly to byCuietal.(2007),however,theclearbiasforperpendicularapi- the apical membrane in cells expressing aPKC-CAAX. To this calcleavageswaslostinaPKCl/zmorphants(Figures5Band end, we first analyzed Par3 localization in these cells. We 5C).Toensurethatmorphantcellsdidnotdisplaymarkedpolar- coinjected a HS-inducible Par3-GFP together with HS-aPKC- ity defects, we performed staining against the tight junction CAAX-mKate2ormembraneboundHS-Ras-mKate2asacon- component ZO-1and confirmedthat no differences to control trol.Inbothcases,someresidualnonapicalsignalofPar3was epitheliawerefound(Figure5D).Weperformedliveimagingto observed(FiguresS3AandS3B).Thisisexplainedbythefact monitorcentrosomeandPar3behaviorandobservedthatafter thatHSinductionmostlikelyleadstosomeproteinoverexpres- nonperpendiculardivisionsalsothemorebasaldaughterquickly sion. Nevertheless, the bulk of Par3 signal was localized to a repositioned its centrosome as well as Par3 toward the apical distinct apical domain both in aPKC-CAAX and control cells side (Figures 5E and 5F; Movie S6). This confirmed that cells (FiguresS3AandS3B).Additionally,westainedaPKC-CAAX-ex- wereabletore-establishpolarityandreintegrateintothetissue pressingcellsagainstthetightjunctionmarkerZO-1.ZO-1signal byrecreatingtheirbipolarmorphology.Subsequently,weinves- wasfoundapicallyandcolocalizedwithapicalPar3(FigureS3C). tigatedwhetherperturbingapicalcleavageplanesleadstode- We next tested whether aPKC distribution all along the cells’ fects in tissue maturation. To address this issue, we injected apicobasal axis interfered with actomyosin organization. We aPKCl/z MOs into Tg(Ath5:GAP-GFP) embryos. Ath5-pro- visualizedF-actinbymosaicallyexpressingtheCalponinhomol- moter-driven membrane-GFP labels the first-born retinal neu- ogydomainofUtrophin,Utr-GFP,whichbindsF-actinwithout rons—theretinalganglioncells(RGCs).WeexaminedtheRGC stabilizingit(Burkeletal.,2007).Inthecontrolsituation,F-actin layer to deduce whether early neuronal layer formation was was organized as filaments along the apicobasal axis of cells impaired. This was not the case, as RGC layer formation was (Figure6A).Thisisinaccordancewithourearlierobservations similar to control embryos in aPKCl/z morphants (Figure 5G). (Leungetal.,2011;Nordenetal.,2009).Interestingly,whenwe Additionally,morphantcellscontributedtotheRGClayer(Fig- expressedaPKC-CAAX,actinfilamentorganizationwasdisrup- ure 5G). This argues that nonperpendicular apical divisions in ted,andthesignalwasmorediffusethaninthecontrolcondition thezebrafishretinalNEdonotmarkedlyperturbtissuearchitec- (Figure 6B). This argues that aPKC-CAAX interferes with actin tureandmaturation.Cellsarisingfromsuchdivisionscanreinte- organizationandmighttherebyrepresentavalidtooltoimpair grateintothedevelopingepitheliumanddifferentiate.Therefore, apicalIKNMandinducenonapicaldivisions.Wenextassessed controlofcleavageplanepositioningalonedoesnotexplainthe the cell division positions in aPKC-CAAX-expressing cells. robustnessandreproducibilityofapicalnuclearmigrationbefore Alreadyat30hpf/6hrpostHS(hphs),aboutathirdofalldividing divisioninPSE. aPKC-CAAX-positive cells (27%, n = 26 cells, 11 embryos) dividedatectopicbasallocations.Asexpected,thesenonapical InterferencewithActinDistributionLeadstoImpaired divisionsfeaturedrandomizedcleavageplanes(Figure6D). ApicalIKNMandNonapicallyDividingCells TheseresultsshowthatinaPKC-CAAX-expressing cellsthe As we observed that perpendicular cleavage planes are not interferencewithactinorganizationcanindeedleadtoimpaired absolutelymandatoryforthemaintenanceofPSEarchitecture, apicalIKNM,whichinturnresultsinasignificantamountofnon- wespeculatedthattheapicalpositionofdivisionsitselfprovides apicaldivisions. relevant advantages for PSE development. To test this idea, wewantedtoexplorehownonapicaldivisionsinfluencetissue TheOffspringofNonapically,Nonperpendicularly developmentandmaturation.Toachievethis,weneededacon- DividingCellsPerturbsRetinalTissueMaturation ditionthatinterferedwithapicalmotionofnucleiwithoutblocking WenexttestedtheimpactofnonapicaldivisionsofaPKC-CAAX- cytokinesis (as direct blocking of the actomyosin machinery expressingcellsontissuedevelopmentandretinalmaturation. would)orcellcyclecontinuation(asthiswouldnotallowcelldi- Interestingly, before and during division, cells showed bipolar vision).Weandothersrecentlyshowedthatactomyosinaccu- morphology,arguingagainstdelamination(Figures6C,6E,6F, mulationsbasaltothenucleuscanbelinkedtotheoccurrence andS3D).Followingdivision,theapicaldaughterwasstillinte- ofapicalIKNM(Leungetal.,2011;Rujanoetal.,2013).Actomy- gratedintheepithelium,whilethebasaldaughterlosttheapical osindistributioninepithelialcellshasbeenpreviouslyshownto attachmentanddisplayedprotrusiveactivity(Figure6C;Movie beregulated byaPKC in variouscontexts(Even-Faitelson and S7).Thisimpliedthatfollowingnonapicaldivisiontheabilityof Ravid,2006;Kishikawaetal.,2008;Uberalletal.,1999).Interest- themorebasaldaughtertoreintegrateintotheepitheliumwas ingly, in the elongated cells of the pseudostratified epithelium, impaired.Tosubstantiatethisfinding,wemonitoredcentrosome aPKCisconfinedtotheapicalsideanddoesnotoccuratbaso- behavior and Par3 distribution in nonapically dividing aPKC- laterallocations(Clarketal.,2012).Therefore,wehypothesized CAAX-positive cells. We demonstrated that while the apical that altering aPKC distribution in these cells could influence daughtercouldrepositionthecentrosometotheapicalsurface actomyosinorganizationandmightthusperturbnuclearmigra- following division, this process was inefficient in the basal tion.Totestthisidea,weusedaHS-inducibleconstructinwhich daughter(Figures6EandS3D;MovieS7).Similarly,theapical DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. 209 A B C D E F (legendonnextpage) 210 DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. daughterinheritedapicalPar3signal,maintainedduringdivision, arguing that they could not reintegrate into the epithelium, as whilethebasaldaughtercreatedanectopicPar3domain(Fig- seen in the aPKC-CAAX condition. We next tested whether ure6F;MovieS7).Insum,thesedatashowthatapicalprocess also in this condition RGC layer formation can be disturbed. regrowthandre-establishmentofradialmorphologyarenoteffi- Like in the aPKC-CAAX condition, we used the Tg(Ath5:GAP- cientinmorebasaldaughtersofnonapical,nonperpendiculardi- GFP)lineasareadout.Herewehadtofixtheembryoswithout visions,andconsequently,suchcellscannotrobustlyreintegrate knowinginwhichonestheaccesstotheapicalsurfacehadbeen intotheepithelium. sufficiently blocked earlier in development. Nevertheless, we Subsequently,wetestedtheeffectofnonapicaldivisionslater found different embryos that showed basal cluster formation. indevelopment.Notably,at24–30hphs,aPKC-CAAX-express- LikeintheaPKC-CAAXcondition,holesintheAth5layerwere ingcellswereobservedatectopicbasalpositions,formingclus- observed that contained pH3-positive cells. Additionally, as ters of rounded cells. These cells continued proliferating, as seen previously, these proliferating cell clusters disturbed demonstrated by pH3 staining (Figure 7A). We next explored neuronallayerformation(Figures7E,S4B,andS4C). howsuchbasalcellclusterformationaffectstissuematuration Together,theseresultsstronglyarguethattheoccurrenceof and architecture. To this end, we injected aPKC-CAAX into nonapicaldivisionscanbelinkedtoperturbationsofPSEinteg- Tg(Ath5:GAP-RFP) embryos. Remarkably, Ath5 expression rityleadingtodefectsintissuematuration.Therefore,wesug- was not observed in aPKC-CAAX-positive cell clusters, while gest that the remarkable reproducibility of apical IKNM and neighboringcontrolcellsdifferentiatednormallyintoRGCs(Fig- subsequentapicaldivisionsservestosafeguardtissueintegrity ure 7B). These RGCs had to arrange around the aPKC-CAXX andmaturationinproliferativezebrafishNE. positiveclusters,resultinginholesintheRGClayer(Figure7B, lower).Thisdemonstratesthatbasalclusterformationofcycling DISCUSSION aPKC-CAAX-expressing cells negatively impacts retinal devel- opmentandtissueformation. In this study, we investigated the purpose and significance of To ensure that what we observed for basal divisions in the localizingdivisionstotheapicalsurfaceinzebrafishNE.Were- aPKC-CAAX condition was a general phenomenon, we aimed vealedthat(1)mitoticentryisnotrestrictedtoapicalpositions, to induce such divisions in otherwise nonperturbed cells. To (2) apical IKNM depends on CDK1 and occurs independently thisend,wemadeuseoftheknownfactthatinhibitionofPlk1 ofcentrosomepositionorintegrity,(3)nonperpendicularapical leadstoamitoticarrest(Le´na´rtetal.,2007).Therebyanaccumu- divisions are tolerated and do not cause tissue perturbations, lationofroundedcellsthatarestalledinmitosisformsasteric (4)afternonapical,nonperpendiculardivisionscausedbyinter- hindrance at the apical surface (Weber et al., 2014). After this ferencewithactomyosindistributionorsterichindranceatapical boundary has been formed, cells undergoing apical IKNM locations,cellsfailtoefficientlyreintegrateintothetissue;they cannotreachtheapicalsurfaceleadingtononapicalmitoticen- instead establish proliferative basal cell clusters that perturb tryofthesecells(Weberetal.,2014).ToinhibitPlk1activity,we overall tissue maturation. Figure 8 presents a schematic sum- hereusedageneticapproachandexpressedaDNconstructun- maryofthesefindings. derHSpromoter(Smitsetal.,2000).DN-Plk1wasexpressedina subsetofcells(markedbyadditionalexpressionofH2B-RFP), ApicalMigrationOccursIndependentlyofCentrosome leadingtotheformationofroundedapicalcellsstalledinmitosis PositionandIntegrity,aswellasMitoticEntry thatblockedtheapicalsurface(Figure7C).Wechoseembryosin It was recently suggested that in the highly elongated cells of whichwedetectedsuchblockingoftheapicalsurfaceforliveim- therodentneocorticalPSE,mitoticentryisrestrictedtotheapi- aging. In all embryos imaged, we observed that beneath the calterminus(Huetal.,2013).InthezebrafishretinalNE,how- blockedapicallayer,unperturbedcontrolcellslabeledbyH2B- ever, mitotic entry occurred at nonapical positions when the GFPandras-GFPmRNAinjectionindeedperformednonapical IKNMmachinerywasinhibitedorwhencentrosomesandnuclei divisionsshown by pH3staining and live imaging (Figures7C, met basally. Our observation is in accordance with a recent 7D,andS4A;MovieS8).Interestingly,earlyaftertheformation study reporting that nonapical mitotic entry can take place oftheapicalhindrance,divisionsoccurredclosetothebarrier, also in the PSE of the Drosophila imaginal disk (Liang et al., and cells still featured an apical process during division (Fig- 2014).Interestingly,wedocumentedthatapicalmigrationper- ure7D,upper).Later,however,theoffspringofthesecellsalso sisted despite mitotic entry at basal locations and even upon divided at very basal positions (Figures 7D, lower, and S4A), perturbing centrosome integrity. We further show that CDK1 Figure4. ApicalIKNMDoesNotDependonCentrosomeIntegrity;CDK1ActivityIsNecessaryandSufficientforApicalIKNM (A)Time-lapseimagingofacellfollowinglaserablationoftheapicalcentrosome.(Left-mostpanel)Cellpriortoablation.ThenucleusislabeledwithPCNA-GFP (green).ThemembraneisvisualizedwithRas-GFP(green)andthecentrosomewithcentrin-mKate2(magenta,whitearrow).Afterablation(00:00),nocentrin signalcanbedetectedintheapicalprocess(whitearrow).Theyellowboxinpanel00:00showsa23magnifiedregionoftheapicalendfoot.Thefollowingpanels shownuclearbehaviorfollowingablationmarkedbyayellowarrow.Timeisinhr:min.TheframesarefromMovieS4.SeealsoFigureS2. (B–D)Time-lapseofcellsexpressingPCNA-GFP(gray)upontreatmentwith(B)DMSOonly,(C)CDK1inhibitor(RO3306),(D)Wee1inhibitor(PD166285).Red arrowsmarkcellsinS-phase(PCNAdots),andgreenarrowsmarkcellsinG2(disappearanceofPCNAdots).Timeisinhr:min.TheframesarefromMovieS5. (E)TracksofnuclearmovementsinthedifferentdrugconditionsstartingattheonsetofapicalIKNM(PD0166285),theentryintoG2phase(RO3306),orboth (control). (F)Averagespeedofapicalnuclearmigrationindifferentconditions.Speedwascalculatedastotaldistancetraveledbynucleiovertime(mm/min).Errorbars representSEM. Scalebarsrepresent10mm.Apicalsurfaceisrepresentedbyadottedlinein(A)andasolidlinein(B)–(D). DevelopmentalCell32,203–219,January26,2015ª2015ElsevierInc. 211