Transcript
PERSPECTIVES ceased at this time.Over the past 40 years,however,it has become clear that the adultbrain also retains stem cells that produce neu-rons and glia throughout an animal’s life.This slow realization has challenged formerpreconceptions about brain development,and has provided an opportunity to exploreexperimentally the identity ofneural stemcells and the mechanisms by which they gen-erate differentiated progeny.In particular,itoffers a unique opportunity to identify celltypes that function as adult stem cells andtheir lineage relationship to embryonic stemcells.The understanding and manipulation of neural stem cells in the embryo and the adultwill have profound therapeutic implications,and will yield insight into how the brain isbuilt and maintained throughout the life of an animal.Fundamental knowledge about stem cellsin the blood,skin and other tissues has comeprimarily from studies in adults,in whichlong-term experimental intervention andtransplantation are possible 1 .Similarly,exper-imental manipulations ofgerminal centres inthe adult brain are beginning to reveal sur-prising insights.For example,stem cells in theadult vertebrate brain express molecularmarkers and ultrastructural characteristics of mature glia 2–5 .Given the prevailing view thatglia represent a developmental endpoint,thischallenges old interpretations regarding thesrcin ofneuronal and glial cells.Anotherseries ofrecent observations indicates thatradial glia,long considered to be precursorsofastrocytes but not neurons,have propertiesofembryonic stem cells 6–8 .This observationraises questions about the lineage relationshipbetween stem cells that maintain adult brainand embryonic stem cells involved in de novo organ assembly.Here we review the glial characteristics of stem cells in the adult brain and suggest thepossibility that glia-like cells might be stemcells at earlier developmental times,thus pro-viding a hypothetical continuum between theglia-like stem cells ofthe adult brain and theneuroepithelial cells in the embryo. Heritage of interpretation, not facts Misconceptions about the srcin ofneuronsand glia have plagued our understanding of brain ontogeny since the early days ofdevel-opmental neuroscience.Soon after thedescription ofglia as supporting cells ofthenervous system,many researchers subscribedto Virchow’s assumption that glia in the cen-tral nervous system (CNS) had a mesenchy-mal origin (reviewed in REF.9 ).This inferencewas refuted in the latter part ofthe nineteenthcentury,when histological techniques beganrevealing the structure ofthe early neuralepithelium.Wilhelm His 10 examined thisregion and discovered that glial cells srcinat-ed within the CNS primordium.He incor-rectly concluded,however,that neurons andglia were produced from two separate popu-lations ofprogenitor cells (FIG.1a) .Roundedmitotic cells near the neural tube lumen werereported by His to be neuronal precursorsand the so-called ‘spongioblasts’(elongatedcells that probably correspond morphologi-cally to what we now call radial glia) werethought to produce glia.From this work emerged the dominant hypothesis that in theearly neural tube,predetermined precursorsproduced either neurons or glia.Schaper 11 and later Sauer 12 showed that the elongatedcells and the rounded mitotic cells ofthe earlyneuroepithelium were indeed the same cells atdifferent stages ofthe cell cycle.Nevertheless,the idea ofseparate srcins for neurons andglia remained heavily entrenched in theneurosciences.In reality,the complexity of For many years, it was assumed that neuronsand glia in the central nervous system wereproduced from two distinct precursor poolsthat diverged early during embryonicdevelopment. This theory was partiallybasedon the idea that neurogenesis andgliogenesis occurred during different periodsof development, and that neurogenesisceased perinatally. However, there is nowabundant evidence that neural stem cellspersist in the adult brain and support ongoingneurogenesis in restricted regions of thecentral nervous system. Surprisingly, thesestem cells have the characteristics of fullydifferentiated glia. Neuroepithelial stem cellsin the embryonic neural tube do not showglial characteristics, raising questions aboutthe putative lineage from embryonic to adultstem cells. In the developing brain, radial gliahave long been known to produce corticalastrocytes, but recent data indicate thatradial glia might also divide asymmetrically toproduce cortical neurons. Here we reviewthese new developments and propose thatthe stem cells in the central nervous systemare contained within the neuroepithelial → radial glia → astrocyte lineage. Most adult tissues retain a reservoir ofself-renewing,multipotent stem cells that cangenerate differentiated tissue components.Until recently,the brain was thought to repre-sent an exception to this general rule.Fordecades,neurobiologists subscribed to theidea that neural stem cells were depleted inthe perinatal brain and that neurogenesis NATURE REVIEWS | NEUROSCIENCE VOLUME 2 | APRIL 2001 | 287 A unified hypothesis on the lineageofneural stem cells Arturo Alvarez-Buylla,José Manuel García-Verdugo and Anthony D.Tramontin OPINION ©2001 Macmillan Magazines Ltd 288 | APRIL 2001 | VOLUME 2 www.nature.com/reviews/neuro PERSPECTIVES to indicate that primary neural precursors inthe adult brain might have characteristics of mature glial cells.We argue that there is a con-tinuum in the neural stem-cell lineage fromthe embryo to the adult (FIG.1b) . Germinal centres persist in the adult Ongoing neurogenesis in the adult conflictswith the classical view that neurogenic ger-minal centres vanish sometime after birth.Instead,it suggests that,somewhere in thebrain,a continuum ofgerminal activitymust persist.The subgranular layer (SGL) of the hippocampal dentate gyrus 39–41 and thesubventricular zone (SVZ) ofthe lateral ven-tricle 42 are the only two neurogenic regionsthat have been described in the adult mam-malian brain.Proliferating cells in the SGLgive rise to young neurons that migrate ashort distance and differentiate into hippo-campal granule neurons 41,43 .It has been sug-gested that small dark cells in this region areneuronal progenitors 39 ,but scant data existregarding the SGL cellular composition,orthe identity ofthe SGL stem cells.Germinalactivity in the adult rodent SVZ 44 ,and thatofother vertebrates 45–50 ,is more extensivethan that observed in the SGL,spanningmuch ofthe lateral wall ofthe lateral ventri-cle.Proliferation ofSVZ cells continuesthroughout life 51,52 .In neonatal 53 and adultrodents 54 ,cells born in the SVZ migrate tothe olfactory bulb,where they differentiateinto interneurons.Unlike the SGL,the cellular compositionand architecture ofthe SVZ has been de-scribed 55 (FIG.2) .SVZ neuroblasts (type A cells)migrate in homotypic chains 56,57 through anetwork ofinterconnecting pathways dis-tributed throughout the wall ofthe lateralventricle 44 .These chains are ensheathed bythe processes ofslowly proliferating type Bcells that have properties ofastrocytes 56 .Scattered along the chains oftype A cells areclusters ofrapidly dividing immature type Ccells (a transit amplifying progenitor),whichare often interposed between type B and typeA cells 55 .Many ofthe chains oftype A cellsmerge in the anterior and dorsal SVZ,form-ing a restricted path called the rostral migra-tory stream (RMS).The RMS carries morethan 30,000 neuroblasts per day into theolfactory bulb,where a fraction ofthese dif-ferentiate into granule and periglomerularneurons 54 .A layer ofmulticiliated ependymalcells lies adjacent to the SVZ and contacts theventricle.Cell–cell interactions within theSVZ (FIG.2) and with the ependymal layer areprobably crucial for the maintenance ofthelocal microenvironment,or neurogenicniche 58 .was usually associated with the generation of glia rather than neurons.The concept that no new neurons wereformed in the adult brain began to change inthe 1960s with the pioneering work ofJosephAltman,who suggested that new nerve cellswere produced in restricted regions oftheadult brain 24–27 .Even though Altman’s sugges-tions were based on careful light microscopicanalysis of[ 3 H]-thymidine-labelled cells,theneuronal identity ofthe new cells was notdemonstrated using ultrastructural or electro-physiological evidence.Thus,the idea that theadult brain was incapable ofneurogenesis pre-vailed for two more decades until Nottebohmand colleagues conclusively demonstratedtelencephalic neuronal replacement in theadult avian brain 28,29 .Adult neurogenesis hasalso been shown in other taxa and is nowaccepted as a common feature ofvertebratebrains 30 .It is important,however,to clarify atthe outset that although new nerve cells areincorporated into certain adult brain regions,there are other regions where this does notseem to take place.Furthermore,in all adultbrain regions where neuronal recruitment hasbeen demonstrated,only a subset ofthe neu-ronal types are involved 31–37 .So,many types of neuron do not seem to be replaced through-out the life ofthe organism.Evolutionary con-straints have been suggested to limit neuronalreplacement in the adult brain 38 .Nevertheless,brain regions do exist where new neurons arecontinually generated.This raises questions asto the identity ofthe neural stem cells withinadult germinal centres and the relationship of these cells with those in the developing brain.As we discuss below,new results are beginningthe early vertebrate neuroepithelium and fetalneural germinal zones makes it very difficult,ifnot impossible,to accurately trace thesehypothetical lineages.Work in invertebrates,where the fate of individual precursor cells can be followedmore precisely,has provided examples inwhich a neuron and a glial cell might be pro-duced from a final single division in alineage 13 .This concept,however,has not beenfully incorporated in the vertebrate field.Some retroviral lineage studies in mammals,primarily in neocortex and during late histo-genic stages,have suggested the presence of glial-restricted and neuronal-restricted prog-enitors 14,15 .Others,however,have suggestedthat ventricular-zone (VZ) cells divide to pro-duce both neurons and glia 16–22 .Interpretationofthese experiments is difficult because virallyencoded gene inactivation and tangentialmigration can prevent the construction of precise lineage trees.Associated with the idea ofseparate srcinsfor neurons and glia grew another dogma,perhaps one ofthe most publicized views of brain development:the generation ofnewnerve cells occurs only during embryonic andperinatal development.This view was bol-stered by the scarcity ofmitoses in the adultCNS,a phenomenon that has never beendemonstrated in differentiated neurons.Furthermore,brain germinal centres aremarkedly modified around birth and manyresearchers thought that the ventricular zonewas converted into the postnatal ependymallayer 23 .It was widely believed that these germi-nal centres became vestigial or quiescent soonafter birth.Any postnatal proliferative activity RG NeuronsRG NeuronsAstrocyte NeuronsRG NPRG NPNPNeuronsNeuronsNeuronsAstrocyteNECommittedneuronalprecursorCommittedglialprecursorNeuronsGlia a b NeuronsNE NEAstrocyteAstrocyte NPNeurons Figure 1 | Classical and proposed stem-cell lineages.a | Historically, it was believed that neuroepithelialcells (NE) in the early neural tube produced two separate pools of committed glial and neuronal progenitors. b | Recent data support a different model in which NE cells either produce or transform into radial glia(RG). Radial glia can divide asymmetrically to produce neurons, glia and perhaps other cell types not shownhere. Neuronal production by radial glia might occur directly (left panel), indirectly via transit amplifyingneuronal progenitors (NP) (right panel) or by some combination of these two schemes. In mammals, radialcells disappear perinatally, when they are thought to transform into astrocytes. In the adult subventricularzone, these astrocytes continue to produce neurons, glia and perhaps other cell types. In several non-mammalian species, radial glia persist into adulthood and continue to produce new neurons. Not includedin this figure is the possibility that NE cells, radial glia or astrocytes might, at times, divide symmetrically toincrease their populations. ©2001 Macmillan Magazines Ltd PERSPECTIVES SVZ astrocytes show several characteristicsthat are similar to neuroepithelial stem cells inthe embryonic neural tube.Both cell typesexpressnestin 55 ,an intermediate filament pro-tein found in neuroepithelial stem cells 76 ,andboth cell types contact the ventricle lumen.Ependymal cells line the luminal surface oftheventricle and are typically described as sepa-rating the SVZ from the cerebrospinal fluid.However,on closer examination using elec-tron microscopy,the ependymal layer does notappear entirely contiguous.In normal mice,asmall number ofastrocytes make direct con-tact with the ventricle by extending a thin cel-lular process between ependymal cells 73 .These astrocytes extend a single cilium intothe cerebral spinal fluid,similar to the ciliumfound on neuroepithelial cells 73,77–79 .Unlikethe multiple long ependymal cilia,this singlecilium is short and contains a 9+0 micro-tubule arrangement.The formation ofthiscilium on SVZ astrocytes and their transientcontact with the ventricle suggest that inter-action with neighbouring ependymal cells orthe cerebrospinal fluid might be involved inthe regulation ofstem-cell proliferation anddifferentiation in the SVZ.The identification ofastrocytes as neuralstem cells is surprising given the classical per-ception that these cells are end points in theglial lineage.Astrocytes encompass a diversegroup ofcells,and it remains to be determinedwhich ofthese can function as stem cells.Nevertheless,SVZ type B cells have many char-acteristics ofastrocytes,including elaborateprocesses that compartmentalize the SVZ 55 .These cells also have a network ofintermedi-ate filaments that contain GFAP,considered bymany to be a definitive hallmark ofa termi-nally differentiated astrocyte.The high degreeofstructural and biochemical specializationsuggests that SVZ astrocytes in the adult brain Astrocytes as stem cells Ifgerminal centres persist in the postnatal andadult brain,which cells within these regionsare the primary precursors or stem cells? Self-renewing cells from the adult SVZ have beenpropagated in vitro in both adherent and non-adherent cultures by using high concentra-tions ofepidermal growth factor (EGF),basicfibroblast growth factor (bFGF) or both 59–62 .On removal ofgrowth factors,the progeny of these cells can differentiate into neurons,astro-cytes and oligodendrocytes 63 .These in vitro progenitors have been considered to be neuralstem cells 64–66 .Although these studies revealcell types that can self-renew and become mul-tipotent in response to growth-factor sig-nalling and culture conditions,it is not knownwhether these cells correspond to the primaryprecursors ofneurons and glia in vivo .On the discovery ofadult SVZ cells thatcould behave like stem cells in culture,it waswidely expected that immature,undifferenti-ated stem cells would be encountered in vivo .This expectation was based in part on the mis-leading idea that stem cells should lack mark-ers or structural features generally attributedto more mature cells.In several mature tissues,it is clear that stem cells express characteristicsofdifferentiated cells 1,67–69 .Recently identified neural stem cells in theSVZ ofthe adult brain possess characteristicspreviously attributed to fully differentiatedglia.Vitally labelled SVZ astrocytes canrespond to EGF signalling in vitro to produceprogeny that can be passaged and differentiat-ed into neurons and glia 2 .SVZ astrocytes canalso behave as neural stem cells when culturedin calfserum,EGF and bFGF 5 .Consistent withthese observations,oligodendrocyte precur-sors isolated from the rat optic nerve can beinduced to differentiate into type-2 astrocytes in vitro .These astrocytes,in turn,whenexposed to FGF,generate cells with neuralstem-cell characteristics 70 .It has been also sug-gested that multiciliated ependymal cells canfunction as stem cells under specific cultureconditions 71 .However,in other studies usingdifferent culture conditions,ependymal cellsdid not behave as stem cells 2,5,72 .Furthermore,there is no convincing evidence that multi-cilated ependymal cells can divide in vivo 2 .Fora further discussion on the controversy overependymal cells as stem cells,see REF.4 .The above observations show that astro-cytes can behave as neural stem cells in vitro ,but in vivo studies are required to determinewhether astrocytes are stem cells in the normalbrain.Doetsch et al. 73 infused the antimitoticdrug cytosine- β - D -arabinofuranoside (Ara-C)into the SVZ for six days to eliminate all rapid-ly dividing type A (neuroblasts) and type Ccells.This treatment spares some SVZ astro-cytes,probably because they proliferate at amuch slower rate 73 .Twelve hours after Ara-Cremoval,SVZ astrocytes began incorporatingBrdU (bromodeoxyuridine).Ependymal cellswere also spared by the antimitotic treatment,but none ofthese incorporated the mitoticmarker in this study.Dividing SVZ astrocytesgave rise to type C cells,which in turn generat-ed neuroblasts.Within 14 days,the entire SVZwas regenerated.Using a method developed to specificallylabel the progeny ofastrocytes in vivo 74 ,Doetsch et al. 2 also showed that SVZ astro-cytes are neuronal precursors in the SVZunder non-regenerating conditions.Doetsch et al. injected an avian leukosis retrovirus(RCAS) containing the alkaline phosphatase(AP) gene into the SVZ oftransgenic mice.These mice were designed to express the RCASreceptor under the control ofglial fibrillaryacidic protein (GFAP);thus viral infection wasdirected at astrocytic B cells.Three and a half days after RCAS injection,AP-positive cellswere found in the SVZ and en route to theolfactory bulb.By 14 days post-injection,many AP-positive interneurons had integratedinto the olfactory bulb.Some SVZ astrocytesalso remained labelled at this time,butlabelled ependymal cells were never observed.The presence ofAP-positive astrocytes in theSVZ 14 days after infection indicates that thesecells might self-renew in vivo .This question,however,merits further investigation.Anothercrucial goal for future studies will be to deter-mine whether SVZ astrocytes can produce celltypes other than olfactory-bulb neurons,suchas large pyramidal neurons or oligodendro-cytes.Recent work indicates that oligoden-drocytes might be produced in the adultSVZ 75 ,but the identity ofthe oligodendrocyteprogenitor in this region is unclear. NATURE REVIEWS | NEUROSCIENCE VOLUME 2 | APRIL 2001 | 289 AECB a CCBCACCAAACA AB c BBBCACCAAAC b A AA AA ABBBCCB Figure 2 | Architecture and putative lineages in the adult subventricular zone.a | The adultsubventricular zone (SVZ) consists of neuroblasts organized as chains (type A cells, red) ensheathed bythe processes of slowly dividing SVZ astrocytes (type B cells, blue). Adjacent to these chains are transitamplifiers (type C cells, green). Ependymal cells (type E cells, pink) line the ventricle and are sometimesdisplaced by type B cells that interdigitate between the ependyma to contact the ventricle. b , c | Hypothetical lineage trees illustrate astrocytic type B stem-cell division. These cells might divideasymmetrically ( b ) or symmetrically ( c ) to produce neurons. The number of cell divisions of any SVZ celltype are unknown. Repeated divisions of type C cells might greatly amplify the number of neuroblastsproduced. Further amplification occurs by the division of these neuroblasts. ©2001 Macmillan Magazines Ltd 290 | APRIL 2001 | VOLUME 2 www.nature.com/reviews/neuro PERSPECTIVES and rat neocortex can produce neurons andglial cells in vitro .Interestingly,the progenyproduced by cultured radial cells depended onthe time at which the cells were collected andplated.When radial glia were isolated fromE14–E16 embryos (E,embryonic day),that is,during peak times ofcortical neurogenesis,they produced primarily neuronal colonieswith few glial and mixed colonies.When radi-al glia were isolated and cultured at E18,thatis,when cortical neurogenesis is nearly com-plete,very few neuronal clones were observedand most colonies were glial or mixed.It willbe interesting to test whether this apparentswitch from neuronal to glial production alsooccurs in radial cells ofthe lateral wall ofthelateral ventricle,where neurons continue to beproduced throughout postnatal life.To investigate the proliferative potential of cortical radial glia in vivo ,Noctor et al. 8 inject-eda retrovirus that expressed green fluores-cent protein (GFP) into the ventricle of E15–E16 rat embryos.Twenty-four hoursafter injection,cells with a radial glial mor-phology were infected by the virus andexpressed high levels ofGFP.These cells wereimmunopositive for vimentin,consistent witha radial glial identity.Three days after injec-tion,radially arrayed clones ofbetween 2 and17 (mean = 5)cells were observed.Each clonecontained several cells that were immunopos-itive for neuronal tubulin and usually onlyone vimentin-positive radial cell.Noctor et al. then used time-lapse videomicrography tofollow the development ofthese clones,andfound that infected cortical radial glia dividedasymmetrically to produce neuroblasts thatmigrated into the cortex,typically along theradial process ofthe same radial cell thatproduced them (FIG.3) .Electrophysiologicalrecordings from clonally related cells wereused to verify the identity ofradial glia andtheir neuronal progeny.Radial cells had lowinput resistances and no voltage-dependentconductances,consistent with a precursor celltype.Migratory neuroblasts had higher inputresistances and voltage-dependent conduc-tances,consistent with an immature neuronalidentity.Taken together,these data provide aclear demonstration that radial glia can pro-duce cortical neurons.It will be important totest the range ofcell types that can be derivedfrom radial cells in the developing brain andto determine ifradial glia in other brainregions also behave as neuronal stem cells.Together,the above results indicate thatradial glial cells are not merely glial progeni-tors,but probably correspond to the primaryprecursors for neurons and glia in the VZ.Assuch,these radial cells could become thefocus ofintense investigation as they mightcontact the ventricle lumen and extend a sin-gle short cilium into the cerebrospinal fluid( REF.92 and A.D.T.,J.M.G.-V.and A.A.-B.,unpublished observations).This cilium has a9+0 microtubule organization and appearsidentical to the cilia found on neuroepithelialcells 77–79 and adult SVZ type B cells 73 .As mentioned above,most radial glia in themammalian brain disappear within days toweeks after birth.In many non-mammalianspecies,however,radial glia persist into adultlife 84,93–98 .Neurogenesis continues into adult-hood in many,ifnot all ofthese species.Radialglia have been studied closely in the avianbrain,where they continue to divide through-out life.Radial-glia division correlates spatiallyand temporally with the appearance ofnewneurons in the avian forebrain,leading to theproposition that these radial cells are neuronalprecursors 99 .Radial-glia somata are found inthe avian VZ,where they contact the ventriclelumen and express the same short ciliumfound in mammalian stem cells.Furthermore,radial glia in the adult avian brain undergointerkinetic nuclear migration 92 .Finally,retro-viral lineage studies ofthe adult avian telen-cephalon have described clones containingradial glia and multiple young neurons 100 .Inretroviral lineage studies ofthe developingchick optic tectum,most clones contain severalneurons,but only one,or rarely,two radial glialcells 17 .Although the cell initially infected by theretrovirus cannot be defined,such a clonalcomposition is consistent with an asymmetri-cally dividing radial cell giving rise to onedaughter cell that remains as a radial glial celland one that generates neurons.More recently the potential ofradial gliahas been tested in vitro .Malatesta et al. 7 reported that radial glia isolated from mouseare not simply stem cells,but have other rolesin the organization and maintenance ofthegerminal centre.This raises questions aboutthe molecular and cellular mechanisms bywhich these cells generate their progeny.Doastrocytes dedifferentiate and lose their pro-cesses,relinquishing their supportive influ-ence on neighbouring cells,or can astrocyticstem cells divide,while maintaining theirstructural functions? This in turn raises ques-tions about what we call a differentiated cell. Radial glia as stem cells Radial glia arise during early VZ development.These cells have their soma in the VZ,andpossess a long process that extends towardsthe pial surface 80–82 .These processes are com-monly thought to serve as guides on whichneuroblasts migrate to reach their final desti-nation 83 .Sometime after neuronal productionceases,radial glia in the mammalian brainretract from the ventricular and pial surfacesand differentiate into astrocytes 81,84–87 .As thename implies,it has been classically thoughtthat radial cells are glia and function as glialprogenitors.Radial glia share multiple features withneuroepithelial cells in the early embryo.Inthe primitive neural tube,neuroepithelialcells contact both the ventricular and pial sur-faces.As development proceeds,the wall of the neural tube thickens and neuroepithelialcells might transform into elongated radialglia by maintaining their ventricular and pialcontacts.These two cell types express theintermediate filament nestin 76,88 ,which hasalso been described in cultured neural stemcells 63 .Radial glia undergo mitosis and inter-kinetic nuclear migration in a fashion verysimilar to neuroepithelial stem cells 89–91 ,and A VZ Ba Bb IZ/ SVZIZ 5 mm 5 mm VZCPIZ t = 0 220 420 495 535 585 705 845 122010 mm t = 0 730 t =18901650 Figure 3 | Time-lapse videomicroscopy of radial glial-cell division.A | A single radial cell is shown 24hours after infection with a retrovirus expressing green fluorescent protein ( t = 0 minutes). The successivepanels show the same cell descending to the ventricular surface, dividing, and then translocating to thetop of the ventricular zone (VZ). The daughter cell can be seen migrating along the radial fibre towards theintermediate zone (IZ). Ba | Another radial glial cell at 24 hours after infection ( t = 0 minutes), showing itsfibre contact with both pial and ventricular surfaces. Bb | After division, its daughter (arrows) migratedalong the radial fibre into the IZ and was confirmed to be positive for the neuron-specific marker TUJ1.(CP, cortical plate.) (Adapted with permission from REF.8 © (2001) Macmillan Magazines Ltd.) ©2001 Macmillan Magazines Ltd PERSPECTIVES predicted fate ofradial glia.Interestingly,Gaiano et al. also reported AP-labelled astro-cytes in the SVZ ofthe lateral ventricular wall.This observation lends support to the ideathat radial glia in the lateral wall ofthe lateralventricle might transform into SVZ astro-cytes.Further experiments are required tofully examine this possibility. Concluding remarks In contrast to the classical model that suggest-ed that neurons and glia are derived from twoseparate ‘branches’ofa lineage tree (FIG.5a) ,we hypothesize that neural stem cells are con-tained within a continuum that forms the‘trunk’ofa lineage tree (FIG.5b) .From thistrunk,the different committed progenitorsor terminally differentiated cells emerge.Depending on the time ofdevelopment,cellswithin this trunk have neuroepithelial,radialglial or astrocytic characteristics.As illustratedabove,these cells have different morphologyand can express different markers.However,they also share some common characteristics,such as the single cilium or the expression of nestin.As neural stem cells go from neuro-epithelial cell → radial glia → astrocyte,theymight continue to self-renew and generatemultiple cell types,but these different stemcells might not be precisely equivalent.Thehold much ofthe programme on which thebrain is assembled.Radial glia seem to have animportant structural role in the embryo,pro-viding spatial alignment between germinalcentres close to the brain ventricles and themore superficial brain regions where neuronstake up residence (the ‘glial coordinate system’ofNieuwenhuys 101 ).This radially organizedalignment system might be particularlyimportant for the development and evolutionofthe mammalian neocortex 102 .However,theidea that radial glia are just a framework oftheembryonic brain must change,as these cellsalso seem to be primary precursors for theimmense numbers ofneurons that use thisscaffolding for their migration and subse-quent organization.This new realization hasimportant implications for the understandingofthe molecular and cellular mechanisms of brain development.For instance,a youngneuron that migrates radially might need todo so on the radial fibre ofits progenitor.So,quite unexpectedly,the cell type in theembryonic brain with apparently the mostcomplex and ‘differentiated’features (longradial processes,elaborate cytoskeleton,com-plex endings in the pial surface,and contactswith blood vessels) turns out to be a primaryprecursor.Should we continue calling thesecells glia,or should the name glia be reservedfor the mature non-neuronal cells ofthe ner-vous system 99 ? The lineage of stem cells Classically,embryonic neuroepithelial cellswere believed to produce at least two separatepools oflineage-restricted progenitors,oneneuronal and the other glial.Neuronal pro-genitors were thought to disappear as thebrain matured and germinal centres becamequiescent or vestigial.The more recent identi-fication ofmultipotent precursors and ongo-ing neurogenesis in the adult brain suggests,instead,that a continuum exists betweenembryonic and adult neural stem cells.Thestudies described here indicate that SVZastrocytes in the postnatal brain are neuralstem cells.The origin ofSVZ astrocytes isunknown,but it seems likely that,as in thecortex,radial glia produce these cells.We haveproposed that perhaps radial glia are elongat-ed neuroepithelial cells,and continue tobehave as multipotent self-renewing progeni-tors in the embryonic and perinatal brain.Wehypothesize that neural stem cells are con-tained within the developmental lineage thatproceeds from neuroepithelial cells,throughradial glia,to astrocytes (FIG.4) .Perhaps the best illustration ofthis pro-posed lineage for stem cells is the recent studyofGaiano and colleagues 6 .This group usedretroviral transfection to introduce constitu-tively active Notch1 into mouse telencephalicprogenitors at E9.5.Notch is believed tomaintain cells in a proliferative progenitorstate and inhibit neuronal differentiation 103 .Surprisingly,activated Notch promoted radialglia identity in infected cells,consistent withthe idea that radial glia are produced fromneuroepithelial cells and that they possessstem-cell potential.Most ofthe research onradial glia has been done in the developingcortex,where these cells ultimately transforminto astrocytes.It seems likely that radial gliain other parts ofthe brain share the same ori-gin and fate.This is an especially provocativehypothesis in the lateral wall ofthe lateralventricle — the future site ofadult SVZ neu-rogenesis and one region where astrocyteshave been shown to function as neural stemcells.Data from Gaiano et al. 6 indicate thatthis might be the case.The virus they used totransfect activated Notch1 into telencephalicprogenitors also carried the alkaline phos-phatase (AP) reporter gene so that infectedcells could be identified histochemically.Asdiscussed above,activated Notch promotedradial glial differentiation in the embryonicVZ.When infected cells were visualized inanimals on postnatal day 42,AP was expressedprimarily by astrocytes,consistent with the NATURE REVIEWS | NEUROSCIENCE VOLUME 2 | APRIL 2001 | 291 Neuroepithelium Embryonic histogenesis Adult SVZRadialgliaAstrocyteNeuronalprogenitorNeuroblastNeuroepithelialcell?? Figure 4 | Unified hypothesis for neural stem-cell development. Left panel: Neural stem cells (purple)in the early neuroepithelium extend from the ventricle to the pia. Middle panel: Like neuroepithelial stemcells, many radial glia (purple) also contact the ventricular and pial surfaces. Radial glia might be neuralstem cells, perhaps an elongated form of the neuroepithelial cell. Radial glia are known to dividesymmetrically (not shown) or asymmetrically (arrows) to produce neurons (red) that migrate into the cortexalong the fibre of their progenitor. Radial glia might produce neurons directly or indirectly through transitamplifying cell types (green). Right panel: Radial glia transform into cortical astrocytes later indevelopment. Cells derived from radial glia might come to reside in the adult subventricular zone (SVZ)(blue). Like radial glia and neuroepithelial cells, some SVZ astrocytes contact the ventricle and extend ashort 9+0 cilium into the cerebrospinal fluid. These astrocyte-like cells behave as stem cells in that theyself-renew and produce neurons (arrows), possibly through intermediate cell types (green). ©2001 Macmillan Magazines Ltd
