in vivo fate mapping and expression analysis reveals ... · reveals molecular hallmarks of...

Cell Stem Cell

Resource

In Vivo Fate Mapping and Expression AnalysisReveals Molecular Hallmarksof Prospectively Isolated Adult Neural Stem CellsRuth Beckervordersandforth,1,10 Pratibha Tripathi,1,10 Jovica Ninkovic,1,4,10 Efil Bayam,1 Alexandra Lepier,4

Barbara Stempfhuber,1 Frank Kirchhoff,5,6 Johannes Hirrlinger,7 Anja Haslinger,2 D. Chichung Lie,2 Johannes Beckers,3,8

Bradley Yoder,9 Martin Irmler,3 and Magdalena Gotz1,4,*1Institute for Stem Cell Research2Research Group Adult Neural Stem Cells and Neurogenesis, Institute of Developmental Genetics3Institute of Experimental Genetics

Helmholtz Centre Munich German Research Centre for Environmental Health (GmbH), Ingolstadter Landstr.1,

85764 Neuherberg/Munich, Germany4Physiological Genomics, Medical Faculty, University of Munich, Schillerstr. 46, 80633 Munich, Germany5Department of Neurogenetics, Max Planck Institute of Experimental Medicine, 37075 Gottingen, Germany6DFG Research Centre for Molecular Physiology of the Brain, 37075 Gottingen, Germany7Carl-Ludwig-Institute for Physiology and N05 Neural Plasticity, Faculty of Medicine, University of Leipzig,Liebigstr. 21, 04103 Leipzig, Germany8Technical University Munich, Center of Life and Food Sciences Weihenstephan, 85354 Freising, Germany9University of Birmingham at Alabama, MCLM688, 1918 University Blvd, Birmingham, AL 35294, USA10These authors contributed equally to this work*Correspondence: [email protected]

DOI 10.1016/j.stem.2010.11.017

SUMMARY

Until now, limitations in theability toenrichadultNSCs(aNSCs) have hamperedmeaningful analysis of thesecells at the transcriptome level. Here we show viaa split-Cre technology that coincident activity of thehGFAP and prominin1 promoters is a hallmark ofaNSCs in vivo. Sorting of cells from the adult mousesubependymal zone (SEZ) based on their expressionof GFAP and prominin1 isolates all self-renewing,multipotent stem cells at high purity. Comparison ofthe transcriptome of these purified aNSCs to paren-chymal nonneurogenicastrocytesandotherSEZcellsreveals aNSC hallmarks, including neuronal lineagepriming and the importance of cilia- and Ca-depen-dent signaling pathways. Inducible deletion of theciliary protein IFT88 in aNSCs validates the role ofciliary function in aNSCs. Our work reveals candidatemolecular regulators for unique features of aNSCsand facilitates future selective analysis of aNSCs inother functional contexts, such as aging and injury.

INTRODUCTION

Thediscoveryofadult neurogenesisandneural stemcells (aNSCs)

has opened a novel field of research aiming to utilize these cells as

sources for repair. However, progress in the field has been

hampered by the limited knowledge about the molecular mecha-

nisms governing the unique properties of aNSCs as the source

of neurogenesis in the adult mammalian brain. Global molecular

analysis of this important cell type was not yet possible because

aNSCs could not be prospectively isolated. So far, aNSCs are en-

744 Cell Stem Cell 7, 744–758, December 3, 2010 ª2010 Elsevier Inc

riched as neurospheres in vitro (Reynolds et al., 1992; Richards

et al., 1992). However, neurosphere cells expanded in high

concentrations of EGF and FGF2 differ profoundly from their

in vivo counterparts, in regard to their fast proliferation (aNSCs

in vivo proliferate slowly), their predominant generation of glial

cells (aNSCs in vivo generate mostly neurons), and their expres-

sion of key fate determinants (Gabay et al., 2003; Hack et al.,

2005). Togain access to acutely isolated aNSCs, various attempts

have been used to enrich this population by fluorescence-acti-

vated cell sorting (FACS), but purification was either below 35%

(Basak and Taylor, 2007; Capela and Temple, 2002; Ciccolini

et al., 2005; Corti et al., 2007; Kawaguchi et al., 2001; Pastrana

et al., 2009) or the isolated fraction contained only a proportion

of the stem cells (Rietze et al., 2001). To overcome these limita-

tions, we aimed to improve the prospective isolation of aNSCs

by utilizing the knowledge about their glial identity and ciliated

nature (Chojnacki et al., 2009; Mirzadeh et al., 2008).

Cells expressing GFAP and GLAST are the source of adult

neurogenesis; they are able to self-renew in vivo (Ninkovic

et al., 2007) and give rise to multipotent neurospheres in vitro

(Garcia et al., 2004). Further analysis revealed that these cells

possess a radial glial-like morphology, are partially embedded

within the ependymal layer, and possess a cilium (Mirzadeh

et al., 2008), a morphology reminiscent of tanycytes (Chojnacki

et al., 2009). Conversely, multiciliated ependymal cells have

been found to be largely quiescent and rarely form self-renewing

neurospheres (Capela and Temple, 2002; Carlen et al., 2009;

Coskun et al., 2008). However, the delineation between radial

glia, tanycytes, and other ependymal cells is rather difficult

(Chojnacki et al., 2009) because it relies on a few markers that

are in fact nonexclusive, such as the Ca-binding protein

S100b, which is often used to identify ependymal cells (Coskun

et al., 2008) but is also present in astrocytes (see Figure S1

available online; Buffo et al., 2008). For the same reason,

.

mailto:[email protected]

http://dx.doi.org/10.1016/j.stem.2010.11.017

hGFAP-GFP βcateninmerge

ac.tubulin prominin1/ac.tubulin merge prominin1prominin1/hGFAP-GFP

C C´ C´´ C´´´ C´´´´

prominin1hGFAP-GFP/prominin1merge

LV

A A´ A´´

B B´ B´´

Figure 1. Immunostainings of SEZ of Adult hGFAP-GFP Transgenic Mice

Confocal images taken at the ventricular surface of whole mounts (B,C) and of sagital sections (A) of adult SEZ.

(A–A00) hGFAP-GFP+/prominin1+ cell (circle) intercalating between the ependymal cells; arrow points at prominin1+ cilia. Cells expressing only hGFAP-GFP are

located beneath the ependymal layer (arrowhead).

(B–B00) b-catenin staining (red) shows pinwheel organization with a GFP+ cell with a small apical surface surrounded by GFP-negative ependymal cells (see also

Movie S1). DAPI is shown in blue; for movie of the confocal stack see also Movie S1.

(C–C00 00) GFP+ cell (small arrow) with a single short cilium (stained for acetylated tubulin [ac. tubulin]) shows prominin1 immunoreactivity at its tip (large arrow).

For additional analysis of SEZ cell populations, see also Figure S1. LV, lateral ventricle; scale bars represent 20 mm.

Cell Stem Cell

Transcriptome of Prospectively Isolated aNSCs

discrimination between radial glia-like stem cells and other niche

astrocytes was so far not possible. To overcome the limitation of

single markers to define a distinct neural cell type, we followed

a new dual-labeling strategy, which allowed us to reliably

discriminate between aNSCs, niche astrocytes, and ependymal

cells and thereby purify these cells for transcriptome analysis.

RESULTS AND DISCUSSION

hGFAP-GFP+/Prominin1+ Cells Fulfill Stem Cell CriteriaIn VivoWe hypothesized that aNSCs may be identified and sorted by

GFP driven by the GFAP promoter (hGFAP-GFP mice) (Nolte

Cel

et al., 2001) and the prominin1 protein (CD133 in human) that

is located on microvilli and cilia of radial glial cells during devel-

opment (Pinto et al., 2008; Weigmann et al., 1997). In adult

hGFAP-GFP mouse brains, all GFP-labeled cells in the dien-

cephalon are colocalizing with astroglial markers like GFAP

and S100b (Figures S1A and S1F). In the SEZ, most of the

GFP+ cells were located beneath the layer of ependymal cells

and were immunoreactive for GFAP (Figures S1A–S1A00, arrow).

We also noticed that some weakly GFP+ cells were neuroblasts

as revealed by double labeling with doublecortin (DCX) (about

20%; see Figures S1C–S1C00 0). Analyzing for prominin1 immuno-

reactivity, we found that most of the GFP-expressing cells were

not immunoreactive for prominin1 (Figures 1A–1A00, arrowheads;

l Stem Cell 7, 744–758, December 3, 2010 ª2010 Elsevier Inc. 745

hGFA

P-G

FPGFP BrdUmerge merge prominin1

A B B´´ B´´´B´

linker iCre aa 1-59

NCre -fusion

GCN4-ORF polyAhGFAP linker iCre aa 1-59

NCre -fusion

GCN4-ORF polyAlinker iCre aa 1-59

NCre -fusion

GCN4 -ORF polyAhGFAP

linker iCre aa 60-343

CCre -fusion

GCN4 -ORF polyAP2 linker iCre aa 60-343

CCre -fusion

GCN4 -ORF polyAP2

hGFAP-NCre containing lentivirus

P2-CCre containing lentivirus

C

SE

Z C

AG

CAT

GFP

GFP D D´

LV LV

DCX/GFP

SE

Z C

AG

CAT

GFP

DCX/GFP/DAPI DCXGFP

E´´E

LV

E´ E´´´

RM

S C

AG

CAT

GFP

DCXGFP

F F´

DCX/GFP/DAPI

F´´

3 month after injection

3 month after injection

10 days after injection

DCX+ 5% 30% 75%

SEZ/RMS

0% 10% nd

70% 20% 3%

10% 15% nd

15% 25% 12%

nd nd 0%

55%

nd

OB

0%

nd

0%

45%

Marker

Survival 3 days

SEZ/RMS

10 days

SEZ/RMS

3 months

DCX+ and BrdU+

GFAP+

BrdU+ (1 hrs pulse)

GFP+only

NeuN+

I J

OB

CA

G C

AT G

FP

DCX/GFP/DAPI GFPG G´ G´´ G´´´

OB

CA

G C

AT G

FP

DCX/GFP/DAPI GFP

H H´ H´´

DCX

Cell Stem Cell


746 Cell Stem Cell 7, 744–758, December 3, 2010 ª2010 Elsevier Inc.

Cell Stem Cell


hereinafter called hGFAP-GFP+only cells). Interestingly, we

observed some GFP-labeled processes intercalating into the

ependymal layer (arrow in Figure 1A0) in a characteristic

pinwheel-like organization (Figures 1B–1B00; Movie S1). These

cells never expressed DCX and the apical processes often

bore a single acetylated tubulin-positive cilium showing promi-

nin1 immunoreactivity especially at its tip (Figures 1C–1C00 0, largearrow), whereas the basal processes were often in contact with

blood vessels. Notably, these cells expressing both GFP and

prominin1 (referred to as hGFAP-GFP+/prominin1+) were few in

number, whereas the majority of prominin1-immunopositive

cells lining the ventricle were multiciliated and GFP negative

(Figures 1A and 1C), thereby resembling ependymal cells

(Coskun et al., 2008; Mirzadeh et al., 2008). Moreover, these

prominin1+/GFP� cells (hereinafter called prominin1+only cells)

expressed commonly used ependymal markers like S100b and

Wdr16. However, these proteins were also expressed in other

SEZ cell types as well as in parenchymal astrocytes (Figures

S1D–S1F), illustrating the difficulty of distinguishing ependymal

cells, aNSCs, and astrocytes by immunohistochemical

‘‘markers.’’ Consistent with this, we also observed weak expres-

sion of GFAP in multiciliated prominin1+only ependymal cells

(Figures S1B–S1B00; see also Platel et al., 2009). These results

suggest that cells with radial glia characteristics integrating

partially into the ependymal layer can be detected by combined

expression of hGFAP-GFP and prominin1.

To further examine whether the double-positive cells possess

the hallmarks of aNSCs, we first examined whether they are slow

dividing and hence retain S-phase DNA label (such as BrdU) as

previously described for aNSCs (Doetsch et al., 1999). Among

the BrdU-label-retaining cells (BrdU for 2 weeks, followed by

a 3 week BrdU-free ‘‘chase’’ period, n = 112), 70% ± 2% were

hGFAP-GFP+/prominin1+ and these cells also exhibited the

characteristic radial glia morphology integrated into the ependy-

mal layer (Figures 2A–2B00 0, arrowheads indicate radial glial

process and arrows prominin1+ cilia), demonstrating that this

double-positive population is indeed slow dividing in vivo.

Functionally, a key characteristic of aNSCs is their neurogenic

nature. To examine whether the hGFAP-GFP+/prominin1+ cells

contribute to adult neurogenesis, we took advantage of

a recently developed fate mapping technique to follow exclu-

sively the progeny of cells coexpressing two markers (‘‘split-

Cre’’) (Hirrlinger et al., 2009). This has been achieved by splitting

the Cre-recombinase into two fragments and fusing them to the

dimerizing GCN4-coiled coil domain yielding the N-terminal

Figure 2. hGFAP-GFP+/Prominin1+ Cells Are Label Retaining and Neur

(A–B00 0) Confocal images of a SEZ sagital section from adult hGFAP-GFP mice th

a GFP+ cell with prominin1+ cilia (red; arrows) that has retained BrdU (white; circle

the adjacent pictures.

(C) Schematic drawing of the split-Cre lentiviral constructs showing the N-termina

the C-terminal Cre fragment under control of the human P2 promoter (prominin1

(D–F) Confocal images taken 10 days after injection of the lentiviral vectors into th

DCX negative and had radial glia morphology (D, D0), or were DCX+ neuroblasts (E

double-positive cells in the RMS are depicted by red arrowheads. For specificity

(G and H) 3 months after injection of the lentiviral vectors, there are GFP+ cells in

(G–G00 0) and a GFP+ neuron integrated into the superficial layer (H–H00).(I) Schematic indicating the localization of GFP+ cells of three injected animals 3

(J) Table shows marker gene expression of the GFP+ cells in the SEZ, RMS, and

Scale bars represent 20 mm (A, B); 50 mm (D–F); and 100 mm (G and H). LV, later

Cel

NCre and the C-terminal CCre fragment (Figure 2C). Because

two different promoters control the expression of the two Cre

fragments, only cells coexpressing the two markers have func-

tional Cre-recombinase, and will therefore undergo the genomic

recombination and permanently express the reporter gene (Hirr-

linger et al., 2009). We adapted this system to label GFP+/

prominin1+ cells by cloning CCre behind the P2 element of the

human prominin1 promoter (hP2-CCre) (Coskun et al., 2008)

and NCre behind the human GFAP promoter (hGFAP-NCre)

into a FUGW lentiviral backbone (Lois et al., 2002) to achieve

expression of functional CCre::NCre dimers only in cells ex-

pressing from both promoters hGFAP and P2.

We first tested the lack of recombination mediated by each of

the Cre fragments alone and observed no GFP reporter+ cells

(Nakamura et al., 2006) in neurospheres transduced with either

construct (Figures S2B, S2C, S2E, and S2F), whereas successful

recombination and many GFP reporter+ cells were detected

upon cotransduction with both N- and C-terminal fragments

driven by the hGFAP or P2 promoter in neurosphere cultures

(Figures S2D and S2G). To examine these cells and their progeny

in vivo, we injected lentiviruses expressing the two Cre frag-

ments into the SEZ of reporter mice. Recombined GFP+ cells

were observed already 3 dpi (days postinjection) in the SEZ (Fig-

ure 2J), but only when both viral vectors were coinjected. At 3

dpi, the majority of reporter+ cells (96%) were restricted to the

SEZ and expressed GFAP (70%). Interestingly, some reporter-

positive SEZ cells had radial glia morphology, consistent with

previous suggestions of aNSCs having long radial processes

(Figure 2D; Chojnacki et al., 2009; Mirzadeh et al., 2008).

Although we observed more transit-amplifying cells (TAPs)

than neuroblasts at short times after injection, the proportion of

labeled neuroblasts increased in the SEZ (from 5% at 3 dpi to

30% at 10 dpi) (Figures 2E–2E00 0) and extended into the rostral

migratory stream (RMS) (Figures 2F–2F00), indicating that

hGFAP-GFP+/prominin1+ cells indeed contribute to neurogene-

sis in vivo.

To test the labeled cells for the ability to self-renew and

generate neurons over a long period of time, we assessed the

identity of labeled cells 3 months after injection of the split-Cre

viral vectors. A small population of reporter-positive cells ex-

pressing GFAP (3%) was still present in the SEZ, indicating that

they did not transform into a more differentiated cell type. Most

importantly, these fate-mapped cells still gave rise to new neuro-

blasts, as shownby the fact that themajority (about 75%) ofGFP+

cells in the SEZ and rostral migratory stream (RMS) expressed

ogenic In Vivo

at obtained BrdU as described in Experimental Procedures. Example depicts

s). Dashed squares always delineate the field of higher magnifications shown in

l Cre recombinase fragment under the control of the hGFAP promoter (top) and

).

e SEZ of CAG CAT GFP reporter mice show GFP reporter+ SEZ cells that were

–E00 0). The latter had also migrated into the RMS (F–F00). Examples of DCX/GFP

of split-Cre approach see Figure S2.

the olfactory bulb (OB). Example of a DCX+ neuroblast at the entry of the OB

months after injection; one dot represents one cell.

OB 3 days, 10 days, and 3 month after injection; nd, not determined.

al ventricle.


Cell Stem Cell


doublecortin (DCX) and migrated into the olfactory bulb (OB).

These data suggest that there was ongoing neurogenesis from

the labeled cells even 3 months after injection. By then many

GFP reporter-positive cells were also found in the OB (73% of

all reporter-positive cells) and most of them were mature

(NeuN-positive, 45%) or immature (DCX-positive, 55%) neurons

(Figures 2G–2H00). Notably, labeled cells populated both glomer-

ular and granular cell layers (Figure 2I), suggesting that hGFAP

and prominin1 promoter activities do not mark a specific,

lineage-restricted population of neural stem cells, but rather

a heterogeneous population capable of generating different

subpopulations of the OB interneurons. Notably, the long-term

neurogenesis of these cells also rules out any reaction to the

injection of the virus that has long ceased by this time. Taken

together, the fate mapping analysis reveals the stem cell hall-

marks of the GFAP- and prominin1-coexpressing cells and also

provides an ideal tool that can be combined with any LoxP-

flanked genes for conditional gene targeting selectively in the

stem cell lineage with minimal effects on other niche cells.

hGFAP-GFP+/Prominin1+ SEZ Cells Comprise All CellsForming Self-Renewing, Multipotent NeurospheresGiven that hGFAP-GFP+/prominin1+ cells exhibit features previ-

ously ascribed to aNSCs in vivo, we then proceeded to isolate

them by FACS and examined whether this fraction would

generate multipotent, self-renewing neurospheres. Toward this

end, the SEZ tissue along the lateral wall of the lateral ventricle

(Figure 3A) was dissected from 8-week-old hGFAP-GFP mice,

dissociated into single cells, and stained with prominin1 anti-

bodies coupled to PE. FACS analysis (Figures 3B–3E) was per-

formed by setting the gates with negative control cells, either

from wild-type mice not expressing GFP or stained with isotype

controls (Figure 3B). Among live SEZ cells (dead cells were

excluded by PI staining) from hGFAP-GFP mice, 20% were

GFP+ and 9% were prominin1 labeled (Figure 3C; for further

details see Figure S3). Consistent with the observations in whole

mounts and sections, we observed a small population of hGFAP-

GFP+/prominin1+ cells (2.5%; Figure 3C). Importantly, when

cells were dissociated from the brain parenchyma at some

distance from any neurogenic site, such as the diencephalon

(Figure 3A), no hGFAP-GFP+/prominin1+ cells could be de-

tected, suggesting that this population is indeed different from

the parenchymal astrocyte population (Figures 3D and 3E).

Next we sorted the above described populations by FACS at

the single cell mode with <2000 events/s. Resorting of the sorted

cells showed 93% purity (Figure S3J), and immunocytochemical

analysis of cells plated immediately after FACS confirmed their

expected identities (Figure S4). For example, proteins known

to be enriched in astrocytes such as GFAP, GLAST, and GLT-1

were detected in most of the hGFAP-GFP+ cells sorted from

the diencephalon (96%) or the SEZ (85%–90%) as well as in

most of the hGFAP-GFP+/prominin1+ stem cell population

(78%). Conversely, the latter as well as the astrocytes from the

diencephalon did not contain any neuroblasts (DCX+ cells) nor

transient amplifying cells (TAPs) detected by Mash1 immuno-

staining and were hence pure glial cells. In contrast, the

hGFAP-GFP+only cells sorted from the SEZ also contained

some neuroblasts and TAPs (16% and 3%, respectively; see

Figure S4U). Most of the cells sorted for their prominin1 positivity


(but not for GFP) contained proteins expressed in ependymal

cells, such as S100b as well as the cilia protein acetylated tubulin

(89% and 90%, respectively; data not shown). However,

astrocytes also show partially weaker immunoreactivity for

these proteins, although at much lower levels (33% of hGFAP-

GFP+only cells from the SEZ; 82% of hGFAP-GFP+ cells from

the diencephalon; see Figure S1). This further supports the coex-

pression of many of the so-called marker proteins in both

astrocyte and ependymal cell populations. Importantly, the

prominin1-only cells were also purely ependymoglial cells and

did not contain any neuroblasts or TAPs.

To examine which of these populations would give rise to self-

renewing, multipotent neurospheres, single cells of each sorted

cell population were plated in single wells to ensure clonality of

the spheres (Buffo et al., 2008) and the number of neurospheres

(defined as cell clusters exceeding 50 mm in diameter, i.e., con-

taining at least 10–15 cells) was quantified 7–10 days after

plating. Sorted cells negative for the markers (hGFAP-GFP,

prominin1) did not generate any neurospheres (Figure 3F). Cells

positive only for prominin1 generated very few neurospheres

(9%; Figure 3F), consistent with previous work (Capela and

Temple, 2002; Coskun et al., 2008). Remarkably, 72% of the

hGFAP-GFP+/prominin1+ cells formed neurospheres, consistent

with the aNSC hallmarks of the double-positive cells in vivo.

Some neurospheres were also generated by hGFAP-GFP+only

cells (Figure 3F).

Because neurosphere formation by itself can not be used as

reliable indication of multipotency and self-renewal (Buffo et al.,

2008), we first examined multipotency by culturing the primary

neurospheres adherent and in the absence of mitogens

(5–7 days), followed by staining for the neuron-specific protein

bIII-tubulin orO4andGFAPdetectingglia.Notably, neurospheres

of cells sorted for prominin1+only generated exclusively glial cells

and hence were not multipotent (data not shown). Conversely,

most neurospheres (79%, n = 28) derived from hGFAP-GFP+/

prominin1+ cells gave rise to all three cell types (astrocytes,

oligodendrocytes, and neurons), indicating their multipotent

nature (Figures 3G–3I). Neurospheres derived from the hGFAP-

GFP+only cell fraction also gave rise to all three cell types (48%,

n = 35).

To examine self-renewal, primary neurospheres were individ-

ually transferred into separate wells, dissociated into single cells,

and cultured for 7 days. Strikingly, none of the neurospheres

formed by cells sorted for prominin1+only or hGFAP-GFP+only

could be passaged and formed secondary neurospheres (Fig-

ure 3F; n = 60 and 100). Thus, none of these populations

contained any self-renewing stem cells. Conversely, the neuro-

spheres sorted for both GFP+ and prominin1+ were self-renew-

ing (59%, n = 90; Figure 3F) and could further self-renew for

more than six passages (data not shown). Thus, our dual-

labeling protocol combining GFP expression from the human

GFAP promoter and prominin1 expression resulted in an enrich-

ment of multipotent, self-renewing neural stem cells above 70%,

and additionally this fraction also comprised all SEZ stem cells,

as none of the other cell populations formed self-renewing, mul-

tipotent neurospheres. This, together with the in vivo data,

further corroborates the stem cell hallmarks of the hGFAP-

GFP/prominin1 double-positive cells and their considerable

enrichment by FACS. Notably, the hGFAP-GFP+only SEZ cells

.

I

die

ncep

halo

n (n

on-n

euro

geni

c)

βIII tubulin GFAP O4

Diencephalon

SEZ

G

F

hGFAP-GFP prominin1-PE

hGFAP-GFP hGFAP-GFP prominin1-PE

H

0

20

40

60

80

hGFAP-GFP+prominin1+

hGFAP-GFP+only

prominin1+only

all negative

% o

f sin

gle

plat

ed c

ells

fo

rmin

g ne

uros

pher

es Secondary neurospheresPrimary neurospheres

prominin1+only

6.5%

hGFAP-GFP+

prominin1+

2.5%

hGFAP-GFP+only

17.1%

AC

hGFAP-GFP+ 11%

D E

WT Isotype control-PE

SE

Z (n

euro

geni

c)

Bprominin1+only

0%

hGFAP-GFP+

prominin1+

0%

hGFAP-GFP+only

0%

prominin1+only

0.2%

hGFAP-GFP+

prominin1+

0%

hGFAP-GFP+only

11%

Figure 3. FACS Analysis, Sorting, and Neurosphere-Forming Potential of the Sorted Cells

(A) Micrograph of a sagital section of an adult mouse brain with white boxes indicating the regions dissected for FACS.

(B–E) Dot plots depict GFP-negative cells and isotype control for PE in (B) and the proportion of the respective positive cells in (C)–(E). For further details on FACS

parameters Figure S3 and for molecular characterization of FACS-sorted cells, see Figure S4.

(F) Histogram depicting the frequency of neurosphere formation by single cells isolated by FACS 7–10 days after culturing (cells analyzed for primary neuro-

spheres: GFP+/prominin1+, n = 299; GFP+only, n = 213; prominin1+only, n = 120; cells negative for all markers, n = 430; cells analyzed for secondary neuro-

spheres: GFP+/prominin1+, n = 90; for the other populations see Results; data from three independent experiments. Error bars represent standard deviation.

(G–I) Micrographs depicting examples of bIII tubulin+ neurons (G), GFAP+ astrocytes (H), and O4+ oligodendrocyte progenitors (I) differentiating from neuro-

spheres generated by hGFAP+/prominin1+ cells showing their multipotent nature (multipotent: 79% [n = 28] of neurospheres generated from hGFAP+/prominin1+

cells; 48% [n = 35] of neurospheres from SEZ cells positive only for GFP).

Scale bars represent 20 mm.

Cell Stem Cell


lacking prominin1 were composed predominantly of astrocytes

(85%–90%) that lacked neurosphere-forming capacity, suggest-

ing that this astrocyte population in the SEZ is functionally

different and rather corresponds to niche astrocytes. Thus our

approach allowed us to separate aNSCs from other astrocytes

in and outside of the SEZ.

Transcriptome Analysis of aNSCs, Astrocytes,and Ependymal CellsThese highly enriched aNSCs (hGFAP-GFP+/prominin1+ cells)

were then examined by whole-genome Affymetrix MOE430

2.0 arrays. Their transcriptome was compared to the transcrip-

tome of parenchymal astrocytes that are not neurogenic

(hGFAP-GFP+ cells from the diencephalon), as well as to other

Cel

cell types in the SEZ, such as the multiciliated ependymal cells

(SEZ prominin1+only) and the SEZ hGFAP-GFP+only cells

comprising niche astrocytes and some of the stem cell progeny.

Hierarchical clustering of the expression data from all biological

replicates grouped the samples according to their cellular identity,

indicating high reproducibility of the sortings and the RNA amplifi-

cation protocol (Figure 4A). Interestingly, rather than the two

astroglial populations clustering with each other (the aNSCs and

the diencephalic astrocytes), the prominin1+only ependymal cells

were found to cluster closer to the nonneurogenic, diencephalic

astrocytes than the aNSCs and the SEZ hGFAP-GFP+only cells.

This is consistent with the fact that the ependymal cells and

diencephalic astrocytes are postmitotic differentiated glia cells,

whereas the aNSCs and hGFAP-GFP+only cells contain


SEZ diencephalon SEZ diencephalon

diencephalon

Ifitm

3

Rarres2

1110017D

15R

IK

microarray RT-PCR

35302520151050-5

hGFAP-GFP+/prominin1+

vs. SEZ hGFAP-GFP+only

hGFAP-GFP+/prominin1+ vs.

dienchGFAP-GFP+

aNSCs vs. prominin1+only

dienc vs.prominin1+only

020406080

100120

Sox11

microarrayRT-PCR

Fold

cha

nge

Sox11 IS

H

H

Fold

cha

nge

N

I

K L M

EDC

A B

F

G

hGFA

P-G

FP/S

ox11

/DC

X

J

Igsf1

Fold

cha

nge

020

4060

80

Thbs4 Th

bs4

ISH

Subependymal Zone

Subependymal Zone Subependymal Zone

hGFA

P-G

FP/T

hbs4

/DA

PI

diencephalon

O P QSubependymal Zone

R SmicroarrayRT-PCR

Cre

b

Cyc

linG

2

Cdk

4

Cyc

linD

2

Cen

trin2

FoxJ

1

Bm

p6

Cal

bind

in

FoxG

1Fe

ch

Mia

1

Rar

res2

RIK

K20

FoxG

1

hGFA

P-G

FP 1

hGFA

P-G

FP 2

hGFA

P-G

FP+ on

ly 1

hGFA

P-G

FP+ on

ly 2

hGFA

P-G

FP 3

hGFA

P-G

FP 4

prom

inin

1+ only

1

prom

inin

1+ only

2

hGFA

P-G

FP+

prom

inin

1+ 1hG

FAP

-GFP

+

prom

inin

1+ 2hG

FAP

-GFP

+

prom

inin

1+ 3hG

FAP

-GFP

+

prom

inin

1+ 4

Subependymal zonediencephalon

hGFAP-GFP+/prominin1+

vs. dienc hGFAP-GFP+

hGFAP-GFP+/prominin1+ vs. dienc hGFAP-GFP+

Figure 4. Confirmation of Microarray Analysis by RT-PCR, ISH, and Immunohistochemistry

(A) Hierarchical clustering of normalized expression data groups the samples in four distinct clusters representing diencephalic (dienc) astrocytes (green),

SEZ cells positive for only prominin1 (red), SEZ cells expressing hGFAPeGFP and prominin1 (aNSCs, violet), and SEZ cells positive for only hGFAP-GFP

(blue).

(B, K, O) Histograms depicting linear fold change of gene expression between the respective cell types indicated in the panels as measured by quantitative

RT-PCR (red bars) or by Affymetrix array analysis (black bars). Error bars represent standard deviation.

Cell Stem Cell



Cell Stem Cell


proliferating stem and progenitor cells (for comparison to embry-

onic radial glia see Figure S4).

To define genes enriched in aNSCs compared to other cell

types, we applied a stringent filter consisting of statistical signi-

ficance (FDR < 10%), at least 2-fold higher expression compared

to other cell populations, and an average expression level >50 in

aNSCs. The resulting sets of aNSC-enriched genes comprised

285 genes (295 probe sets, Table S6) compared to all other

cell populations (diencephalic astrocytes, SEZ prominin1+only,

SEZ hGFAP-GFP+only). The same approach was used to define

ependymal-enriched genes (Table S2) and astrocyte-enriched

genes (Table S3). For the pair-wise comparison of aNSCs versus

diencephalic astrocytes, we used the same filters as described

above, except FDR < 1%. This resulted in 392 genes (474 probe

sets, Table S5) higher expressed in aNSCs in comparison to

diencephalic astrocytes.

These cell type-specific transcriptomes can now be further

used to characterize different types of glial cells (see also Lovatt

et al., 2007; Tables S2 and S3). Indeed, previously used

‘‘markers’’ could not discriminate between astrocytes, aNSCs,

and ependymal cells in our analyses. Consistent with immuno-

stainings and previous functional analyses (Platel et al., 2010)

indicating a strong overlap of glial proteins in astrocytes, aNSCs,

and ependymal cells, we also observed many of the known so-

called astrocyte-specific genes being expressed in aNSCs as

well as in ependymal cells (Table S1). In most cases (64%),

parenchymal protoplasmic astrocytes expressed the highest

levels of these genes, such as GLAST, GLT-1, Aquaporin 4,

FGFR3, or the recently suggested new astrocyte marker Aldh1l1

(data not shown), but these proteins or reporter constructs were

also detectable in the other populations, such as the SEZ niche

astrocytes, the stem cells, or the multiciliated ependymal cells

(seealsoPlatel et al., 2010, for functional relevance). The situation

for identifying ependymal cells is evenworse because noneof the

previously used ependymal cell ‘‘markers’’ was specific or even

expressed at higher levels in the ependymal cells compared to

the other cell types. For example, the widely used ependymal

marker S100b was expressed at much higher levels in the

parenchymal astrocytes (see Figure S1) and c-Jun and FoxJ1

were expressed at higher levels in aNSC compared to the epen-

dymal cells (Table S1; see also Jacquet et al., 2009). However,

our transcriptome analysis now identifies sets of marker genes

that may be better to delineate these different cell types.

aNSC-Enriched Genes Are Specifically Expressedin the SEZNext, we validated the differences in gene expression that

emerged from the array analysis (according to the criteria estab-

lished in previous work [Pinto et al., 2008]; see also Supple-

mental Information) by RT-PCR, in situ hybridization (ISH), and

immunohistochemistry. From 47 genes selected for RT-PCR,

we confirmed 38 as differentially expressed (Figure 4B and

(C–J, L, M, P, Q) In situ hybridizations (ISH) of mRNAs as indicated in the pane

Thbs4 in the SEZ (arrows) and virtual absence of signal in the diencephalon (C

confirmation by Allen Brain Atlas, see Figures S5 and S6.

(N, R, S) Fluorescence micrographs depicting Sox11- and Thbs4-immunopositi

that Thbs4 is a secreted protein.

All scale bars represent 120 mm except (N) and (S) (50 mm).

Cel

data not shown). ISH further confirmed that genes selected as

being enriched in the aNSCs in comparison to parenchymal

astrocytes were indeed expressed in the SEZ and absent in

the diencephalon (Figures 4C–4F, 4L, 4M, 4P, and 4Q). For

example, Ifitm3, Rarres2,111001D15RIK, Sox11, and Thbs4

mRNAs as well as proteins were enriched in the SEZ and virtually

absent from the diencephalon astrocytes as demonstrated by

RT-PCR (Figures 4B, 4K, and 4O and data not shown), ISH

(Figures 4C–4J, 4L, 4M, 4P, and 4Q), and immunohistochemistry

(Figures 4N, 4R, and 4S). Further confirmation was obtained by

expression patterns from the Allen Brain Atlas (Figures S5 and

S6). Notably, genes with a higher expression in aNSCs

compared to diencephalic astrocytes showed two expression

patterns. Most of the genes showed specific expression in the

SEZ, such as the transcription factors Cbx1, Nfib, or the planar

polarity gene Celsr1 (Figure S5), whereas a few genes were

also expressed in basal ganglia (such as Id4 and Foxp1).

To further corroborate our microarray results, we also in-

spected genes enriched in diencephalic astrocytes as compared

to aNSCs, for example Igsf1 andDNER, a Notch ligand known to

promote maturation of Bergmann glia through activation of

Notch (Eiraku et al., 2005). For these genes we observed expres-

sion only in the diencephalon, but not in the SEZ (ISH see Figures

4I and 4J; Allen Brain Atlas see Figure S5). Thus, the gene

expression differences determined by our microarray analysis

were confirmed by independent approaches.

Neurogenic Fate Determinants Are Enriched in aNSCsCompared to Parenchymal AstrocytesThe gene expression differences observed in hGFAP-GFP- and

prominin1-coexpressing aNSCs in comparison to other cell

types now allow us to gain novel insights into the molecular

program regulating the remarkable features of these cells, one

of thembeing their capacity to generate neuronswhile other cells

in surrounding regions fail to generate neurons. To this end, we

compared the expression profile of SEZ aNSCs with the astro-

cytes residing outside the neurogenic niche and analyzed the

gene ontology (GO) terms associated with the 392 genes en-

riched in aNSCs compared to diencephalic astrocytes (Figures

5A and 5B). Besides higher expression of genes known to be ex-

pressed in undifferentiated progenitors or stem cells, such as

Nestin, Vimentin, andMusashi, we observed a significant enrich-

ment of the GO term ‘‘cell cycle and proliferation’’ (59 genes).

Indeed, aNSCs proliferate, whereas parenchymal astrocytes

are quiescent (see e.g., Buffo et al., 2008). Interestingly, another

significantly enriched GO category was ‘‘nervous system devel-

opment and differentiation’’ (30 genes), indicating that aNSCs

already upregulated mRNAs related to neurogenesis. In line

with this, we observed the presence of strikingly many neuro-

genic fate determinants associated with the GO term ‘‘regulation

of transcription’’ (19 genes), such as the subgroup C of Sox tran-

scription factors (TFs), Sox4 and Sox11 (Bergsland et al., 2006),

ls. Note the high expression of Ifitm3, 1110017D15RIK, Rarres2, Sox11, and

–Q), whereas the opposite pattern was observed for Igsf1 (I, J). For further

ve cells in the SEZ of hGFAP-GFP mice (arrows indicate Thbs4+ cells). Note


unknown gene function (84)

SE

Z N

CS

1

SE

Z N

SC

2

SE

Z N

SC

3

SE

Z N

SC

4

die

nc

G

FP

1

die

nc

G

FP

2

die

nc

G

FP

3

die

nc

G

FP

4

4 1 1 -4

SE

Z N

CS

1

SE

Z N

SC

2

SE

Z N

SC

3

SE

Z N

SC

4

die

nc

GF

P 1

die

nc

GF

P 2

SE

Z P

ro

m1 1

SE

Z P

ro

m1 2

die

nc

GF

P 3

die

nc

GF

P 4

SE

Z G

FP

1

SE

Z G

FP

2

A B

DC

ESEZ hGFAP-GFP+/prominin1+ dienc hGFAP-GFP SEZ hGFAP-GFP+only SEZ prominin1+only

2000

1500

1000

500

0

Dsg2

Calb1

Tgm2

Spag1

7Myo

7a

Crocc

CiliumCalcium ion binding

Epha5

Wnt4

Bmp6

Signal transduction

Fhod

3

Anxa2

Cytoskeleton

exp

resi

on v

alue

s

Calcium ion binding (17)

Cilium (11)

Cell adhesion (9)

Extracellular matrix (9)

Cytoskeleton organization (8)Metabolic processes (9)Signaling (5)

Cell cycle (3)

Anchoring junction (3)

Others (127)

Calcium

Cilium

Cell adhesionCytoskeleton organization

Signaling

Pkp2

Lgals

3

Cell adhesion

Unknown gene function (81)

Cell cycle (59)

Nervous system development (30)

Regulation of transcription (19)

Microtubule-based process (13)

Chromatin assembly/disassembly (8)

Calmodulin binding (6)Heterotrimeric G-protein complex (5)

Proximal/distal pattern formation (4)

Negative regulation of glial differentiation (3)

Other function (199)

Cell cycle

Nervous sytem development

Regulation of transcription

aNSCs vs. diencephalic astrocytes

aNSCs vs. all other populations

4 1 1 -4

Figure 5. aNSC-Enriched Gene Expression

(AandC)Heatmaprepresentationofexpressionvaluesofgenesenriched inaNSCs incomparison todiencephalicastrocytes (A)andtoall otherpopulations (diencephalic

astrocytes, prominin1+only, hGFAP-GFP+only) (C). Red/blue (A) and red/green (C) indicates higher/lower expression values as indicated in the scale bar (log2 scale).

(B and D) Pie charts depicting significantly enriched GO terms (p < 0.05; except for the categories ‘‘unknown’’ and ‘‘other gene function’’) associated with

aNSC-enriched genes in comparison to diencephalic astrocytes (B) and to all other populations (D). In brackets, the numbers of genes associated with the indi-

cated GO term are provided.

(E) Microarray expression data of selected aNSC-enriched genes grouped according to selected associated GO terms.

Cell Stem Cell



Cell Stem Cell


as well asMeis2 or Ascl1 (Parras et al., 2004) that are reported to

regulate GABAergic neurogenesis. Notably, although expression

levels of these TFs were about 103 lower in diencephalic astro-

cytes and ependymal cells, they were even higher in the hGFAP-

GFP+only cell population comprising the neuroblast progeny of

the stem cells. Indeed, at protein level these TFs (see e.g., for

Sox11 in Figure 4N) as well as doublecortin and synuclein were

detectable only in neuroblasts, suggesting that aNSCs already

upregulate the mRNA but do not yet express the protein at

detectable levels. The same expression pattern was observed

for other transcription factors and known neurogenic regulators

that have failed the stringent criteria of differential expression, for

example, the neurogenic TFs Pax6 and Dlx (Brill et al., 2008;

Hack et al., 2004, 2005) or the chromatin remodeling factor Mll

that is involved in the upregulation of the Dlx family of transcrip-

tion factors (Lim et al., 2009). They were all expressed about

2–103 higher in aNSCs compared to diencephalic astrocytes

but are even higher in the neuroblasts-containing SEZ GFP+only

cell population (Table S4), an expression profile that is also

shared by other TFs that have so far not yet been implicated in

neurogenesis. These are providing candidates for neurogenic

fate determinants in aNSCs, such as the orphan nuclear receptor

(Nr2f6), also named Ear2, previously shown to be involved in

nucleus coeruleus development (Warnecke et al., 2005), or the

transcriptional regulators Tox3 and Whsc1. These also

comprised epigenetic regulators, e.g., Uhrf1 (also Np95), amulti-

domain protein that connects DNA methylation and histone

modification (Rottach et al., 2010), and Hells/Lsh, a member of

the SNF2 chromatin remodeling family involved in the control

of DNA methylation patterns during embryonic development

(Sun et al., 2004). These proteins are predicted to be involved

in regulating neurogenesis from aNSCs because their expres-

sion levels were higher than in nonneurogenic glia but further

rose in the SEZ cells comprising neuroblasts. Understanding

the neurogenic regulators in aNSCs is of particular importance

for the attempts to elicit neurogenesis also outside these niches

in the adult brain parenchyma for repair.

Thus, one of the transcriptional hallmarks of aNSCs is the

increased expression level of neurogenic regulators, suggesting

that aNSCs in the SEZ are already ‘‘primed’’ for neurogenesis

by upregulating neurogenic factors atmRNA level. This transcrip-

tional bias toward neurogenesiswasmissedbefore because cells

cultured as neurospheres were examined. However, culturing

these cells with EGF and FGF2 promotes expression of glial

transcription factors, such as Olig2 (Gabay et al., 2003; Hack

et al., 2004), and only a few neurons are generated from cells

cultured as neurospheres, whereas aNSCs in vivo generate

mostly neurons. The profound difference in gene expression

between aNSCs isolated acutely by FACS and those cultured as

neurospheres is further evident from our transcriptome analysis,

in which we found little to no overlapwith the published transcrip-

tomes of neurosphere cells (Ivanova et al., 2002). These differ-

ences are probably due to (1) the gliogenic environment resulting

from growth factors in the medium; (2) one population being

highly enriched (hGFAP-GFP+/prominin1+ aNSCs), whereas neu-

rospheresarequiteheterogeneous; and (3)geneexpressionalter-

ation resulting from long culture conditions. This highlights again

the importanceof theexpressionprofileof acutely isolatedaNSCs

to unravel the molecular characteristics of aNSCs.

Cel

aNSCs Have a Unique Neural Stem Cell SignatureAlthough the comparisonbetweenSEZcells andparenchymal glia

revealedexciting insights intoneurogenic regulators, to identify the

unique molecular hallmarks of aNSCs, it is important to exclude

genes that are also expressed at high levels in the other SEZ

populations. Therefore, the expression profile of the aNSCs

(hGFAP-GFP+/prominin1+ cells) was depleted for genes ex-

pressed in all other sorted cell populations (prominin1+only

ependymal cells and hGFAP-GFP+only cells comprising mostly

astrocytes and someTAPs and neuroblasts). This filtering resulted

in a set of 285 genes enriched in aNSCs (Figure 5C; Table S6),

which were further confirmed by RT-PCR (see e.g., FoxJ1,

Bmp6, Mia in Figure 4B) as well as by the Allen Brain Atlas (83%

were SEZ enriched, n = 118; Figure S6 and data not shown).

This subtraction approach resulted in the elimination of genes

expressed broadly in the basal ganglia and SEZ (compare Figures

S5andS6) andexcluded, e.g.,Mll. This genewas identifiedbyLim

et al. (2006) as expressed in GFAP+ cells of the SEZ, but it is

expressed at even higher levels in the progeny of these cells, the

neuroblasts. This demonstrated that without comparison to the

other SEZpopulations, genesmay bemistaken for being enriched

in stemcellswhile they are expressedat evenhigher levels in other

niche cells. This approachmight also avoid anymisinterpretations

resulting from the 30% nonneurosphere-forming cells comprised

in the sorted hGFAP-GFP and prominin1 double-positive cells.

Although the cells in this population not generating neurospheres

may be quiescent stem cells, it is also possible that they do not

have stem cell characteristics. In the latter case they may be

similar to one or several of the other cell types, such that the

subtraction of genes expressed in other cell types will eliminate

them from the aNSC-enriched gene set. Indeed, the genes ex-

pressed at significantly higher levels only in aNSCs (Table S6)

were depleted of the neurogenic factors. Analysis of two of these

aNSC-enriched genes (Epha5, Lgals3) on protein level revealed

a rather selective signal in a subset of hGFAP-GFP+ radial astro-

cytes lining the lateral ventricle (Figures 6A–6B00 and 6C–6C00 0).Accordingly, the enriched GO terms associated with the

aNSC-enriched gene sets were quite different compared to the

previous analysis (aNSCs versus diencephalic astrocytes), as

reflected in the shift from neurogenesis and transcriptional regula-

tion to cilia function and signal transduction (Ca signaling, G

protein coupled receptors, Wnt, BMP, and TGF-b signaling).

It was rather surprising that the most enriched GO terms of the

aNSCs were associated with cilia function and Ca signaling,

because one would have expected that these pathways are

shared with ependymal cells (cilia) and parenchymal astrocytes

(Ca signaling). Instead, these categories became noticeable

only in the analysis of aNSC-enriched genes after depletion of

the genes expressed in other astrocytes (including niche astro-

cytes) and ependymal cells, indicating therefore elevated

expression levels of a specific subset of genes involved in Ca

signaling and cilia function in aNSCs. Ca signaling has been

described to regulate stem cell quiescence and proliferation in

other cell types (Horsley et al., 2008; Shepherd et al., 2007;

Tao et al., 2008). In contrast to Ca affecting NFATc-mediated

transcription in skin, aNSCs instead expressed high levels of

Aebp1 (Table S6), a transcriptional regulator interacting with

Ca/Calmodulin (Lyons et al., 2006). Also, cell adhesion and cyto-

skeletal components enriched in aNSCs are subject to Ca


merge Galectin3/hGFAP-GFP Galectin3 hGFAP-GFPC´ C´´ C´´´C

D E GF

hGFAP-GFPmerge Epha5 hGFAP-GFPmerge Epha5A LV

CP

A´ A´´ B B´ B´´

GlastCreERT2/+; CAG CAT GFP; IFT88flox/wt

GlastCreERT2/+; CAG CAT GFP; IFT88flox/delta

10

15

20

180

185

5

0

cells

per

SE

Z

*

*

GFP+ BrdU+ GFP+

BrdU+

GlastCreERT2/+; CAG CAT GFP; IFT88flox/wt

GFP/BrdU BrdU

GlastCreERT2/+; CAG CAT GFP; IFT88flox/delta

GFP/BrdU BrdU

SEZ hGFAP-GFP+/prominin1+

SEZ hGFAP-GFP+

only

SEZ prominin1+

only

300

200

100

0

exp

resi

on v

alue

s in

the

SE

Z

IFT88

Figure 6. Expression and Functional Analysis of aNSC-Enriched Genes In Vivo

(A and B) Fluorescence micrographs depicting Epha5-immunopositive cells in the SEZ of hGFAP-GFP mice.

(A–A00) Arrow indicates a GFP+ cell with radial glial morphology expressing Epha5.

(B–B00) Protoplasmic GFP+ astrocyte in the SEZ negative for Epha5 (arrowhead).

(C–C00 0) Lgals3-immunopositive cells in the SEZ of hGFAP-GFP mice. Arrows indicate GFP+ cells close to the ventricle colabeled with Galectin3; arrowheads

depicting a GFP+ niche astrocyte being negative for Galectin3.

(D) Microarray data showing that IFT88 is highly enriched in the aNSCs in comparison to the other SEZ populations.

(E–G) Ablation of cilia in aNSCs by deleting IFT88 with a GLAST::CreERT2 driver line (see also Figure S7).

(E) BrdU label-retaining experiments (as described in Experimental Procedures) revealed a significant decrease in the number of slowly proliferating cells in mice

carrying the IFT88floxed/D allele (n = 4) compared to IFT88floxed/wt animals (n = 3), although the induction rate is the same as shown by the GFP reporter.

(F and G) Fluorescence micrographs depicting BrdU label-retaining cells and CAG CAT GFP reporter construct to monitor the recombined cells in the SEZ of

IFT88floxed/wt animals (F) and IFT88floxed/D mice. Error bars represent standard deviation (E).

Scale bars represent 20 mm (A–C), 100 mm (F–G); CP, choroid plexus; LV, lateral ventricle.

Cell Stem Cell


signaling, such as Desmoglein 2, a junctional protein (Chitaev

and Troyanovsky, 1997), and the cytoskeletal proteins Fibulin 7

and Arvef. Desmoglein 2, amember of the larger cadherin family,


forms desmosomes that are usually connected with Plakophilin,

linking them to the intracellular cytoskeleton. Indeed, Plakophi-

lin2 was also enriched in the aNSC-enriched transcriptome,

.

Ependymal cells

Radial glia/astrocytes

Ventricle-contact aNSC

neuroblast

lateral ventricle

cilia

TAP

blood vessel

niche astrocyte

hGFAP-GFP prominin1

Figure 7. Schematic Drawing Summarizing Cell Types Examined in Whole Mounts/Sections and Isolated by FACS from the Adult SEZ

hGFAP-GFP+/prominin1+ radial glia like cells are connected to the ventricle via a small apical process and a primary cilia and to the blood vessels on the basal

side. Most of the hGFAP-GFP+ cells have no access to the ventricle and display astroglial features. Some neuroblasts also remain a weak GFP signal (indicated in

light green). Prominin1 labels multiciliated ependymal cells.

Cell Stem Cell


suggesting a rather specific molecular composition of the junc-

tions established by the aNSCs with notable differences to those

connecting radial glial cells (own unpublished observations).

Electron microscopic analysis has revealed different junctions

between aNSCs within the pinwheel core and between aNSCs

and neighboring ependymal cells (Mirzadeh et al., 2008), and

our microarray data now suggest specific molecular hallmarks

of aNSC junctions. These findings prompt the concept that Ca-

mediated signals may regulate delamination of the aNSCs and

thereby be involved in their activation or instructing quiescence.

Indeed, among the astrocytes in the SEZ, only those integrated

in the ependymal layer act as active stem cells (Pastrana et al.,

2009; and our analysis here), suggesting that regulation of this

integration by influencing the junctional connectivity may decide

whether aNSCs leave their active state and enter quiescence or

differentiate. Thus, Ca-mediated regulation of the junctional

adhesion may be involved in maintaining the stem cell pool. In

addition to these junctional Ca-regulated proteins, we also found

other cell surface molecules with selectively high expression in

the aNSCs, such as Galectins, that are linked to Ca signaling.

In neutrophils Galectin clustering causes an increase in cytosolic

Ca signals, probably by Ca release from internal stores (Karma-

kar et al., 2005). Besides the previously described expression of

Galectin1 (Lgals1) in SEZ astrocytes (Sakaguchi et al., 2007), we

also discovered Lgals3 as one of the most differentially ex-

pressed genes in the aNSCs. Taken together, this analysis

suggests novel candidates as molecular regulators of aNSC

hallmarks, possibly converging on Ca as a central signaling

mediator. Moreover, it provides a platform for comparison to

other stem cell signatures, prompting the suggestion that Ca

signaling may emerge as a pathway involved in regulating stem

cell features in several organs.

A particularly intriguing finding is the intersection between Ca

signaling and the second enriched GO term, cilia function.

Sensory cilia also influence Ca signals in cells, for example in

kidney cells where cilia deflection causes a rise in intracellular

Ca. Primary cilia are crucial not only for Ca signaling but also

for many other signaling pathways, and many receptors as well

as signal transduction components are concentrated in the

primary cilium (as reviewed in Eggenschwiler and Anderson,

Cel

2007). Moreover, in the developing dentate gyrus, the second

major neurogenic niche, cilia are required for the formation of

aNSCs and regulate neurogenesis by mediating Shh signaling

(Breunig et al., 2008; Han et al., 2008). Because these results

have been obtained by deletion of cilia during development,

the effects of cilia deletion specifically in aNSCs of the SEZ are

not yet known.

To address this issue, we deleted IFT88, a member of the in-

traflagellar transport group of cilia proteins, whose expression

was highly enriched in the aNSCs in comparison to all other

SEZ populations (Figure 6D). Its expression in diencephalic

astrocytes was also lower, though not as significant as required

for our statistical analysis. To inducibly delete IFT88 in adult

neural stem cells, we used the GLAST::CreERT2 mouse line

(Ninkovic et al., 2007). These mice were crossed further to

GFP reporter mice (Nakamura et al., 2006) to monitor the re-

combined cells. Because the genetic deletion of a mutated

allele of IFT88 (Ift88floxed/delta) (Haycraft et al., 2007) results in

cilia loss, this approach allowed us to selectively delete cilia in

aNSCs 4 weeks after tamoxifen application (Figure S7). To

examine the effect of cilia loss in aNSCs, we performed BrdU

label-retaining experiments as described above. Notably, the

number of BrdU-retaining, slowly dividing aNSCs was signifi-

cantly reduced in the SEZ of mice carrying the IFT88floxed/D

allele in comparison to IFT88floxed/wt animals (Figures 6E–6G),

demonstrating that disruption of cilia function results in pertur-

bation of aNSC function. These data support our transcriptome

analysis because the deletion of an aNSC-enriched factor

impairs aNSC function and shows the importance of cilia also

in aNSCs.

Taken together, our dual-labeling technique allowed reliable

discrimination between aNSCs, niche astrocytes, and ependy-

mal cells (Figure 7) and thereby the purification of these cells

for transcriptome analysis. Comparing the expression profiles

of aNSCs to other SEZ cell populations as well as to non-

neurogenic astrocytes from the brain parenchyma revealed

new insights into the molecular characteristics of aNSCs,

such as their priming toward the neuronal lineage as well as

the importance of specific cilia- and Ca-dependent signaling

pathways.


Cell Stem Cell


EXPERIMENTAL PROCEDURES

Animals

Human GFAP-GFP (Nolte et al., 2001), CAG CAT GFP reporter (Nakamura

et al., 2006), GLAST::CreERT2 (Mori et al., 2006), IFT88floxed/D, IFT88floxed/wt

(Haycraft et al., 2007), and C57BL6/J mice used here were kept under stan-

dard conditions. Experimental procedures were performed in accordance

with German and European Union guidelines and were approved by the insti-

tutional animal care committee and the Government of Upper Bavaria under

license numbers 211-2531-23/04 and 55.2-1-54-2531-144/07. Stereotactic

injections were performed as previously described (Brill et al., 2008) with 1 ml

of virus at �0.75 anterior/posterior, 1.2 medial/lateral, and �1.9 to �1.6

dorsal/ventral from Dura relative to Bregma in mm.

Immunohistochemistry

Whole mounts were prepared as described (Mirzadeh et al., 2008) and brain

sections were cut on the vibratome (80–100 mm) or on the cryostat (30 mm).

The tissue was directly processed for immunostaining (see Brill et al., 2008)

with the primary and secondary antibodies detailed in Supplemental

Information.

Constructs for the Split-Cre Fate Mapping

Human P2 promoter was PCR amplified from human genomic DNA with

50-GGTCCAATCAGAGTGCGT-30 and 50-CCCTTAGCTCGCCAGA-30 as

primers and subcloned in the pCMV (Stratagene, USA) vector containing

CCre. The hGFAP-NCre fragment was cloned into the pCMV as previously

described (Hirrlinger et al., 2009). Fusion products were then subcloned into

FUGW lentiviral backbone and lentiviral particles were produced as described

previously (Lois et al., 2002).

Preparation of Cells for FACS Sorting

Adult SEZ and diencephalon from hGFAP-GFP mice (n = 10 in each group)

were dissected, dissociated as described previously (Buffo et al., 2008). and

resuspended in staining solution (DMEM/10% FCS/0.02% NaN3) containing

primary antibody (PE-conjugated CD133 [1:100, BD Bioscience]) and incu-

bated for 20 min. After washing in FCS-containing medium, cells were resus-

pended in medium containing PI (1:1000, 10 min) and analyzed or sorted at

a FACS Aria (BD). Debris and aggregated cells were gated out by forward,

sideward scatter and PI staining. Gating was done with isotype control as

detailed in Figure S3. Because FACSAria collects four different fractions, the

SEZ populations of hGFAP-GFP+only, hGFAP-GFP+/prominin1+, and

prominin1+only were collected simultaneously. For further details please see

Supplemental Information. For neurosphere cultures, cells were plated as

described (Buffo et al., 2008) and further detailed in the Supplemental Informa-

tion. For immunostaining, FACS-sorted cells were plated in 24-well plate for

1 hr, fixed in 2% PFA for 15 min at room temperature, and stained after

washing in PBS with the primary and secondary antibodies detailed in Supple-

mental Information.

RNA Isolation, Microarray Analysis, and qRT-PCR

After sorting, the cells were centrifuged for 20 min at 10,000 rpm and lysed in

100 ml of lysis buffer (QIAGEN) containing N & P carriers (ExpressArt). Total

RNA was isolated with the RNeasy MICRO kit (QIAGEN) according to the

manufacturer’s instructions. Quality and concentration of total RNA was

examined with the Agilent Bioanalyser. Microarray analysis was performed

with Affymetrix MOE430 2.0 arrays as per manufacturer’s instructions (see

Supplemental Information).

For RT-PCR, cDNA was synthesized with SuperScript II (Invitrogen) as per

manufacturer’s instructions. qPCR was performed on an Opticon (Bio-rad)

with SYBRGreen Imastermix (Bio-rad) and expression levels were normalized

to GAPDH.

In Situ Hybridization

Digoxigenin-labeled RNA probes were synthesized by in vitro transcription

with the DIG labeling mix and T3, T7, or SP6 polymerase (Roche). In situ

hybridizations were performed on 80–100 mm thick vibratome sections via

standard protocol (Brill et al., 2008).


ACCESSION NUMBERS

The microarray data are available in the Gene Expression Omnibus (GEO)

database (http://www.ncbi.nlm.nih.gov/gds) under the accession number

GSE18765.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures,

seven figures, six tables, and one movie and can be found with this article on-

line at doi:10.1016/j.stem.2010.11.017.

ACKNOWLEDGMENTS

We particularly would like to thank Timothy McClintock and Frank Zaucke for

probes and antibodies; Jochen Graw for human DNA; and Andrea Steiner-

Mezzadri, Christiane Lach, and Timucin Ozturk for excellent technical help.

We are also particularly grateful to Armen Saghatelyan, Benedikt Berninger,

Emily Violette, Filippo Calzolari, Pia Johansson, Troy Ghashghaei, and Yves

Barde for excellent comments on the manuscript. This work is supported by

the DFG including the SFB596, 870, the Hans and Ilse Breuer Price, BMBF,

EUTRACC, HELMA, and CoReNe as well as the UAB Hepato/Renal Fibro-

cystic Diseases Core Center (UAB HRFDCC) DK074038.

Received: November 6, 2009

Revised: July 8, 2010

Accepted: November 1, 2010

Published: December 2, 2010

REFERENCES

Basak, O., and Taylor, V. (2007). Identification of self-replicating multipotent

progenitors in the embryonic nervous system by high Notch activity and

Hes5 expression. Eur. J. Neurosci. 25, 1006–1022.

Bergsland, M., Werme, M., Malewicz, M., Perlmann, T., and Muhr, J. (2006).

The establishment of neuronal properties is controlled by Sox4 and Sox11.

Genes Dev. 20, 3475–3486.

Breunig, J.J., Sarkisian, M.R., Arellano, J.I., Morozov, Y.M., Ayoub, A.E.,

Sojitra, S., Wang, B., Flavell, R.A., Rakic, P., and Town, T. (2008). Primary cilia

regulate hippocampal neurogenesis by mediating sonic hedgehog signaling.

Proc. Natl. Acad. Sci. USA 105, 13127–13132.

Brill, M.S., Snapyan, M., Wohlfrom, H., Ninkovic, J., Jawerka, M., Mastick,

G.S., Ashery-Padan, R., Saghatelyan, A., Berninger, B., and Gotz, M. (2008).

A dlx2- and pax6-dependent transcriptional code for periglomerular neuron

specification in the adult olfactory bulb. J. Neurosci. 28, 6439–6452.

Buffo, A., Rite, I., Tripathi, P., Lepier, A., Colak, D., Horn, A.P., Mori, T., and

Gotz, M. (2008). Origin and progeny of reactive gliosis: A source of multipotent

cells in the injured brain. Proc. Natl. Acad. Sci. USA 105, 3581–3586.

Capela, A., and Temple, S. (2002). LeX/ssea-1 is expressed by adult mouse

CNS stem cells, identifying them as nonependymal. Neuron 35, 865–875.

Carlen, M., Meletis, K., Goritz, C., Darsalia, V., Evergren, E., Tanigaki, K.,

Amendola, M., Barnabe-Heider, F., Yeung, M.S., Naldini, L., et al. (2009).

Forebrain ependymal cells are Notch-dependent and generate neuroblasts

and astrocytes after stroke. Nat. Neurosci. 12, 259–267.

Chitaev, N.A., and Troyanovsky, S.M. (1997). Direct Ca2+-dependent hetero-

philic interaction between desmosomal cadherins, desmoglein and desmocol-

lin, contributes to cell-cell adhesion. J. Cell Biol. 138, 193–201.

Chojnacki, A.K., Mak, G.K., and Weiss, S. (2009). Identity crisis for adult peri-

ventricular neural stem cells: Subventricular zone astrocytes, ependymal cells

or both? Nat. Rev. Neurosci. 10, 153–163.

Ciccolini, F., Mandl, C., Holzl-Wenig, G., Kehlenbach, A., and Hellwig, A.

(2005). Prospective isolation of late development multipotent precursors

whose migration is promoted by EGFR. Dev. Biol. 284, 112–125.

Corti, S., Nizzardo, M., Nardini, M., Donadoni, C., Locatelli, F., Papadimitriou,

D., Salani, S., Del Bo, R., Ghezzi, S., Strazzer, S., et al. (2007). Isolation and

.

http://www.ncbi.nlm.nih.gov/gds

http://dx.doi.org/doi:10.1016/j.stem.2010.11.017

Cell Stem Cell


characterization of murine neural stem/progenitor cells based on Prominin-1

expression. Exp. Neurol. 205, 547–562.

Coskun, V., Wu, H., Blanchi, B., Tsao, S., Kim, K., Zhao, J., Biancotti, J.C.,

Hutnick, L., Krueger, R.C., Jr., Fan, G., et al. (2008). CD133+ neural stem cells

in the ependyma of mammalian postnatal forebrain. Proc. Natl. Acad. Sci. USA

105, 1026–1031.

Doetsch, F., Caille, I., Lim, D.A., Garcıa-Verdugo, J.M., and Alvarez-Buylla, A.

(1999). Subventricular zone astrocytes are neural stem cells in the adult

mammalian brain. Cell 97, 703–716.

Eggenschwiler, J.T., and Anderson, K.V. (2007). Cilia and developmental

signaling. Annu. Rev. Cell Dev. Biol. 23, 345–373.

Eiraku, M., Tohgo, A., Ono, K., Kaneko, M., Fujishima, K., Hirano, T., and

Kengaku, M. (2005). DNER acts as a neuron-specific Notch ligand during

Bergmann glial development. Nat. Neurosci. 8, 873–880.

Gabay, L., Lowell, S., Rubin, L.L., and Anderson, D.J. (2003). Deregulation of

dorsoventral patterning by FGF confers trilineage differentiation capacity on

CNS stem cells in vitro. Neuron 40, 485–499.

Garcia, A.D., Doan, N.B., Imura, T., Bush, T.G., and Sofroniew, M.V. (2004).

GFAP-expressing progenitors are the principal source of constitutive neuro-

genesis in adult mouse forebrain. Nat. Neurosci. 7, 1233–1241.

Hack, M.A., Sugimori, M., Lundberg, C., Nakafuku, M., and Gotz, M. (2004).

Regionalization and fate specification in neurospheres: The role of Olig2 and

Pax6. Mol. Cell. Neurosci. 25, 664–678.

Hack, M.A., Saghatelyan, A., de Chevigny, A., Pfeifer, A., Ashery-Padan, R.,

Lledo, P.M., and Gotz, M. (2005). Neuronal fate determinants of adult olfactory

bulb neurogenesis. Nat. Neurosci. 8, 865–872.

Han, Y.G., Spassky, N., Romaguera-Ros, M., Garcia-Verdugo, J.M., Aguilar,

A., Schneider-Maunoury, S., and Alvarez-Buylla, A. (2008). Hedgehog

signaling and primary cilia are required for the formation of adult neural stem

cells. Nat. Neurosci. 11, 277–284.

Haycraft, C.J., Zhang, Q., Song, B., Jackson, W.S., Detloff, P.J., Serra, R., and

Yoder, B.K. (2007). Intraflagellar transport is essential for endochondral bone

formation. Development 134, 307–316.

Hirrlinger, J., Scheller, A., Hirrlinger, P.G., Kellert, B., Tang, W., Wehr, M.C.,

Goebbels, S., Reichenbach, A., Sprengel, R., Rossner, M.J., and Kirchhoff,

F. (2009). Split-cre complementation indicates coincident activity of different

genes in vivo. PLoS ONE 4, e4286.

Horsley, V., Aliprantis, A.O., Polak, L., Glimcher, L.H., and Fuchs, E. (2008).

NFATc1 balances quiescence and proliferation of skin stem cells. Cell 132,

299–310.

Ivanova, N.B., Dimos, J.T., Schaniel, C., Hackney, J.A., Moore, K.A., and

Lemischka, I.R. (2002). A stem cell molecular signature. Science 298, 601–604.

Jacquet, B.V., Salinas-Mondragon, R., Liang, H., Therit, B., Buie, J.D.,

Dykstra, M., Campbell, K., Ostrowski, L.E., Brody, S.L., and Ghashghaei,

H.T. (2009). FoxJ1-dependent gene expression is required for differentiation

of radial glia into ependymal cells and a subset of astrocytes in the postnatal

brain. Development 136, 4021–4031.

Karmakar, S., Cummings, R.D., and McEver, R.P. (2005). Contributions of

Ca2+ to galectin-1-induced exposure of phosphatidylserine on activated

neutrophils. J. Biol. Chem. 280, 28623–28631.

Kawaguchi, A., Miyata, T., Sawamoto, K., Takashita, N., Murayama, A.,

Akamatsu, W., Ogawa, M., Okabe, M., Tano, Y., Goldman, S.A., and Okano,

H. (2001). Nestin-EGFP transgenic mice: Visualization of the self-renewal

and multipotency of CNS stem cells. Mol. Cell. Neurosci. 17, 259–273.

Lim, D.A., Suarez-Farinas, M., Naef, F., Hacker, C.R., Menn, B., Takebayashi,

H., Magnasco, M., Patil, N., and Alvarez-Buylla, A. (2006). In vivo transcrip-

tional profile analysis reveals RNA splicing and chromatin remodeling as prom-

inent processes for adult neurogenesis. Mol. Cell. Neurosci. 31, 131–148.

Lim, D.A., Huang, Y.C., Swigut, T., Mirick, A.L., Garcia-Verdugo, J.M.,

Wysocka, J., Ernst, P., and Alvarez-Buylla, A. (2009). Chromatin remodelling

factor Mll1 is essential for neurogenesis from postnatal neural stem cells.

Nature 458, 529–533.

Cel

Lois, C., Hong, E.J., Pease, S., Brown, E.J., and Baltimore, D. (2002). Germline

transmission and tissue-specific expression of transgenes delivered by lentivi-

ral vectors. Science 295, 868–872.

Lovatt, D., Sonnewald, U., Waagepetersen, H.S., Schousboe, A., He, W., Lin,

J.H., Han, X., Takano, T., Wang, S., Sim, F.J., et al. (2007). The transcriptome

and metabolic gene signature of protoplasmic astrocytes in the adult murine

cortex. J. Neurosci. 27, 12255–12266.

Lyons, P.J., Mattatall, N.R., and Ro, H.S. (2006). Modeling and functional anal-

ysis of AEBP1, a transcriptional repressor. Proteins 63, 1069–1083.

Mirzadeh, Z., Merkle, F.T., Soriano-Navarro, M., Garcia-Verdugo, J.M., and

Alvarez-Buylla, A. (2008). Neural stem cells confer unique pinwheel architec-

ture to the ventricular surface in neurogenic regions of the adult brain. Cell

Stem Cell 3, 265–278.

Mori, T., Tanaka, K., Buffo, A., Wurst, W., Kuhn, R., and Gotz, M. (2006).

Inducible gene deletion in astroglia and radial glia—A valuable tool for func-

tional and lineage analysis. Glia 54, 21–34.

Nakamura, T., Kurokawa, K., Kiyokawa, E., and Matsuda, M. (2006). Analysis

of the spatiotemporal activation of rho GTPases using Raichu probes.

Methods Enzymol. 406, 315–332.

Ninkovic, J., Mori, T., and Gotz, M. (2007). Distinct modes of neuron addition in

adult mouse neurogenesis. J. Neurosci. 27, 10906–10911.

Nolte, C., Matyash, M., Pivneva, T., Schipke, C.G., Ohlemeyer, C., Hanisch,

U.K., Kirchhoff, F., and Kettenmann, H. (2001). GFAP promoter-controlled

EGFP-expressing transgenic mice: a tool to visualize astrocytes and astroglio-

sis in living brain tissue. Glia 33, 72–86.

Parras, C.M., Galli, R., Britz, O., Soares, S., Galichet, C., Battiste, J., Johnson,

J.E., Nakafuku, M., Vescovi, A., and Guillemot, F. (2004). Mash1 specifies

neurons and oligodendrocytes in the postnatal brain. EMBO J. 23, 4495–4505.

Pastrana, E., Cheng, L.C., and Doetsch, F. (2009). Simultaneous prospective

purification of adult subventricular zone neural stem cells and their progeny.

Proc. Natl. Acad. Sci. USA 106, 6387–6392.

Pinto, L., Mader, M.T., Irmler, M., Gentilini, M., Santoni, F., Drechsel, D., Blum,

R., Stahl, R., Bulfone, A., Malatesta, P., et al. (2008). Prospective isolation of

functionally distinct radial glial subtypes—Lineage and transcriptome analysis.

Mol. Cell. Neurosci. 38, 15–42.

Platel, J.C., Gordon, V., Heintz, T., and Bordey, A. (2009). GFAP-GFP neural

progenitors are antigenically homogeneous and anchored in their enclosed

mosaic niche. Glia 57, 66–78.

Platel, J.C., Dave, K.A., Gordon, V., Lacar, B., Rubio, M.E., and Bordey, A.

(2010). NMDA receptors activated by subventricular zone astrocytic glutamate

are critical for neuroblast survival prior to entering a synaptic network. Neuron

65, 859–872.

Reynolds, B.A., Tetzlaff, W., and Weiss, S. (1992). A multipotent EGF-respon-

sive striatal embryonic progenitor cell produces neurons and astrocytes.

J. Neurosci. 12, 4565–4574.

Richards, L.J., Kilpatrick, T.J., and Bartlett, P.F. (1992). De novo generation of

neuronal cells from the adult mouse brain. Proc. Natl. Acad. Sci. USA 89,

8591–8595.

Rietze, R.L., Valcanis, H., Brooker, G.F., Thomas, T., Voss, A.K., and Bartlett,

P.F. (2001). Purification of a pluripotent neural stem cell from the adult mouse

brain. Nature 412, 736–739.

Rottach, A., Frauer, C., Pichler, G., Bonapace, I.M., Spada, F., and Leonhardt,

H. (2010). The multi-domain protein Np95 connects DNA methylation and

histone modification. Nucleic Acids Res. 38, 1796–1804.

Sakaguchi, M., Imaizumi, Y., and Okano, H. (2007). Expression and function of

galectin-1 in adult neural stem cells. Cell. Mol. Life Sci. 64, 1254–1258.

Shepherd, M.C., Duffy, S.M., Harris, T., Cruse, G., Schuliga, M., Brightling,

C.E., Neylon, C.B., Bradding, P., and Stewart, A.G. (2007). KCa3.1 Ca2+ acti-

vated K+ channels regulate human airway smooth muscle proliferation. Am. J.

Respir. Cell Mol. Biol. 37, 525–531.

Sun, L.Q., Lee, D.W., Zhang, Q., Xiao, W., Raabe, E.H., Meeker, A., Miao, D.,

Huso, D.L., and Arceci, R.J. (2004). Growth retardation and premature aging

phenotypes in mice with disruption of the SNF2-like gene, PASG. Genes

Dev. 18, 1035–1046.


Cell Stem Cell


Tao, R., Lau, C.P., Tse, H.F., and Li, G.R. (2008). Regulation of cell proliferation

by intermediate-conductance Ca2+-activated potassium and volume-sensi-

tive chloride channels in mouse mesenchymal stem cells. Am. J. Physiol.

Cell Physiol. 295, C1409–C1416.

Warnecke, M., Oster, H., Revelli, J.P., Alvarez-Bolado, G., and Eichele, G.

(2005). Abnormal development of the locus coeruleus in Ear2(Nr2f6)-deficient


mice impairs the functionality of the forebrain clock and affects nociception.

Genes Dev. 19, 614–625.

Weigmann, A., Corbeil, D., Hellwig, A., and Huttner, W.B. (1997). Prominin,

a novel microvilli-specific polytopic membrane protein of the apical surface

of epithelial cells, is targeted to plasmalemmal protrusions of non-epithelial

cells. Proc. Natl. Acad. Sci. USA 94, 12425–12430.

.

in vivo fate mapping and expression analysis reveals ... · reveals molecular hallmarks of...

Documents