Adult spinal cord progenitor cells are repelledby netrin-1 in the embryonic and injured adultspinal cordAudrey Petit*, Drew L. Sellers*, Daniel J. Liebl†, Marc Tessier-Lavigne‡, Timothy E. Kennedy§, and Philip J. Horner*¶

*Department of Neurosurgery, University of Washington, Seattle, WA 98195; §Centre for Neuronal Survival, Montreal Neurological Institute, McGillUniversity, Montreal, QC, Canada H3A 2B4; †The Miami Project to Cure Paralysis and Department of Neurosurgery, University of Miami School of Medicine,Miami, FL 33101; and ‡Genentech, San Francisco, CA 95688

Edited by Corey S. Goodman, Renovis, South San Francisco, CA, and approved September 11, 2007 (received for review May 2, 2007)

Adult neural progenitor cells (aNPCs) exhibit limited migration in vivowith the exception of the rostral migratory stream and injury-inducedmovement. Surprisingly little is known regarding those signals reg-ulating attraction or inhibition of the aNPC. These studies demon-strate that aNPCs respond principally to a repulsive cue expressed atthe embryonic floor plate (FP) and also the injured adult CNS. Adultspinal cord progenitor cells (aSCPs) were seeded onto organotypicslice preparations of the intact embryonic or injured adult spinal cord.Cell migration assays combined with genetic and molecular pertur-bation of FP-derived migration cues or aSCP receptors establishnetrin-1 (Ntn-1) but not Slit-2, Shh, or Ephrin-B3 as the primaryFP-derived repellant. When slices were prepared from injured spinalcord, aSCP migration away from the injury core was Ntn-1-depen-dent. These studies establish Ntn-1 as a critical regulator of aSCPmigration in the intact and injured CNS.

adult stem cells � floor plate � migration � guidance

Neural stem cell (NSC) migration in the mature nervous systemis very limited unless triggered by injury or cell misregulation,

as in cancer (1–3). These data suggest the existence of potentinhibitors or the absence of attractants leading to restricted cellmovement. Regulated migration is a potent mechanism for creatingNSC compartments and specialization. Migration inhibitors haveparticular significance for the adult mammalian nervous system,which has limited self-repair. Lesion resolution is typified bypersistent cystic voids that never regenerate neural cells (4). Incontrast, embryonic brain and many systems of non-mammals showlimited scarring and impressive tissue reconstitution (5, 6). Duringdevelopment, neural migration, patterning, and axonal guidanceare directed by a spatiotemporal integration of chemoattractive andchemorepulsive signals (7, 8). Many of these guidance molecules areexpressed in the mature CNS and may play key roles in the controlof adult stem cell migration or patterning in the healthy andpathological CNS (9–13). There has been a limited analysis of NSCmigration in the adult rodent rostral migratory stream from thesubventricular zone to the olfactory bulb (14–17) and migrationfrom the subventricular zone or subependymal region after head orspinal cord (SC) injury, respectively (18, 19). The migratory cuesthat guide these processes are largely unknown, but candidates suchas the CXCR4 chemokine receptor and its ligand, stromal cell-derived factor (SDF-1alpha), have emerged. These factors arethought to be mediators of inflammation-induced stem cell homing(20). But, to our knowledge, the developmentally potent migratorycues such as slit, netrin, and integrins have yet to be studied in eitherof these systems.

We cocultured adult SC progenitor cell (aSCPs) with embryonicorganotypic slices of SC to screen for adult NSC migration cues. Wedemonstrate that aSCPs migrate away from the floor plate (FP).When we perturbed Slit-2, Shh, Ephrin-B3, and Ntn-1 function,only down-regulation of Ntn-1 removed the repellent effect of theFP. We then tested the role of Ntn-1 in a slice culture model of SCinjury where aSCPs are normally repelled from the injury core. Our

results establish Ntn-1 as a key repellent cue for aSCPs that maysignificantly restrict lesion remodelling in the injured adult CNS.

ResultsaSCPs Migrate Away from the FP. The aSCP is distributed throughoutthe SC and has the unique capacity of self-renewal and multipo-tency (21). In addition, aSCP are one of the first cell populations torespond to CNS insult and generate astrocytes and oligodendro-cytes, important for tissue homeostasis (22). NSCs were isolatedfrom whole SC preparations of adult transgenic and reporter mice.They were expanded in serum-free media containing the mitogensFGF and EGF, and multipotency was periodically confirmed[supporting information (SI) Experimental Procedures].

To explore the potential cues involved in the control of adultprogenitor migration, we cocultured GFP-expressing aSCPs onembryonic day 12 (E12) organotypic SC slices. An aggregate ofaSCPs was placed at the midline of the slice, at the FP that expressesmany guidance molecules (Fig. 1A). By 2 days in vitro (DIV), apattern of cell migration was apparent that became distinct by 5DIV. The aSCPs aligned at the ventral neuroepithelium (VN) withno aSCPs persisting at the FP (Fig. 1 A–D). The aSCPs establishedthis migratory pattern in 97% of the E12SC slice–aSCP cocultures.

To better understand this pattern of migration, we repeated theE12 SC slice–aSCP coculture, but with the roof plate, anothermajor source of guidance molecules, as a midline structure. TheaSCPs did not migrate, but stayed at the roof plate even after 5 DIV(Fig. 1E). These results suggest that the migration pattern requiresthe FP.

Embryonic-derived SC progenitor cells (E12) also were capableof proliferation, self-renewal, and multipotency in vitro (SI Fig. 5)and resettled similarly to the aSCP, indicating a potentially commonmechanism of migration.

To rule out selective cell proliferation within the VN or increasedcell death (apoptosis) in FP development, we performed an analysisof bromodeoxyuridine incorporation and apoptosis detection byimmunostaining. These studies demonstrated equal rates of deathand proliferation across the open-book preparation. In addition,aSCP placed at the FP remained undifferentiated with the excep-tion of cells at the VN, which exhibited a predominantly astroglial

Author contributions: A.P. and P.J.H. designed research; A.P. performed research; A.P.,D.L.S., D.J.L., M.T.-L., and T.E.K. contributed new reagents/analytic tools; A.P. and P.J.H.analyzed data; and A.P., T.E.K., and P.J.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviations: aSCP, adult SC progenitor cell; CM, conditioned media; DIV, days in vitro; En,embryonic day n; FP, floor plate; HEK, human embryonic kidney cell; KO, knockout; Ntn-1,netrin-1; NSC, neural stem cell; SC, spinal cord; VN, ventral neuroepithelium.

¶To whom correspondence should be addressed. E-mail: phorner@u.washington.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0703240104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0703240104 PNAS � November 6, 2007 � vol. 104 � no. 45 � 17837–17842

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

Mar

ch 1

0, 2

020

phenotype. These data indicate that cell migration was the likelymechanism behind the distribution of progenitors (SI Fig. 6).

aSCPs Migrate in Response to Specific Signals Expressed by the FP.Because aSCPs aligned on each side of the FP and stopped at theVN (Fig. 1 A–D), two alternative mechanisms were possible:repulsion by the FP or attraction by the VN. To differentiaterepulsion from attraction, we created three different E12SCslice–aSCP coculture conditions. In the first experiment, weplaced an FP explant ectopically on the slice (Hoechst; SI Fig. 7A and B). The aSCPs migrated to alternate sides of the endog-enous FP and reproduced this pattern in register with the ectopicFP (SI Fig. 7 C and D). In the second experiment, we alsodissected away one part of the FP from the SC slice (SI Fig. 7E),and aSCPs established the migration pattern, but only on the halfof the slice with the preserved FP (SI Fig. 7F). For the thirdexperiment, we dissected one part of the SC slice (SI Fig. 7H) androtated it by 180° to place the dorsal neuroepithelium adjacentto the FP (SI Fig. 7G). The aSCPs migrated to either side of theFP and did not show any preference for the VN compared withthe dorsal neuroepithelium (SI Fig. 7 I and J). The aSCPs arespecifically repulsed by the FP, and they have the ability torespond to guidance molecules expressed during neural tubedevelopment.

Slit-2, Ephrin-B3, Shh, and Ntn-1 as Potential Migration Cues. Toidentify the guidance molecule(s) responsible for the migrationpattern, we screened for guidance receptors expressed by aSCPsby using RT-PCR for Slit-2, Ephrins, and Ntn-1 receptors. aSCPsexpress two of the three Slit receptors, Robo-1 and -2, but not -3(SI Fig. 8A). At E12, Ephrin-B3 is highly expressed at the FP (SIFig. 8 K and K�) (23), and aSCPs express many Ephrin-B

receptors: Eph-B1, -B2, -B3, -B4, and -B6 (SI Fig. 8J). aSCP alsoexpresses the Ntn receptors of the DCC family, DCC andneogenin, and the UNC5 family, UNC5A, UNC5B, and UNC5C(Fig. 2A).

Because aSCPs express receptors for both Ntn-1 and Slits, andbecause these ligands are all highly expressed at the FP, we usedfour methods to determine the responsiveness of aSCPs to thesemolecules: (i) a coculture of aSCPs on an isolated substrate ofhuman embryonic kidney cells (HEKs) (cell patch); (ii) a cocultureof aggregates of aSCPs and HEKs in a three-dimensional collagenmatrix; (iii) a transwell migration assay with conditioned media(CM); and (iv) a stripe migration assay with recombinant Ntn-1.

Ntn-1 Is a Chemorepellent for aSCPs. We used a transwell assay inwhich CM from HEK, HEK-Slit2, HEK-Ntn1, and recombinantNtn-1 were placed in the lower chamber, and aSCPs were placedin the upper chamber (Fig. 2G). The presence of Ntn-1 but notSlit2 in the lower chamber almost completely blocked aSCPmigration into the lower chamber and onto the underside of thefilter. Only 6–7% of the cells crossed the filter with HEK-Ntn1or recombinant Ntn-1 at 100 ng/ml and 200 ng/ml (Fig. 2H)compared with 38–39% for the HEK and HEK-Slit2.

We used the collagen-embedded explant cocultures assay (24) todetermine if Slit-2 and Ntn-1 might be responsible for FP-mediatedrepulsion. Aggregates of aSCPs were cocultured with aggregates ofHEK, HEK-Slit2, or HEK-Ntn1 cells. When the aSCPs were incontact with the HEK and the HEK-Slit2 aggregates, 50–60%migrated into the aggregates (SI Fig. 8 D and E). In contrast, �5%of aSCPs migrate into HEK-Ntn1 aggregates (Fig. 2 J). We also usedan assay in which the substrate for aSCP growth was a patch ofHEK, HEK-Slit2, or HEK-Ntn1 cells (Fig. 2B). aSCPs grew ran-domly over the coverslips when the patches of cells were HEK orHEK-Slit2 (SI Fig. 8 B and C and Fig. 2 C and D). In contrast, aSCPsdeveloped a distinctive distribution when in contact with a patch ofHEK-Ntn1 cells (Fig. 2 E and F). The aSCPs were clearly repelledby the Ntn-1-expressing cells and migrated outside the patch toform a clear border between the patch and the laminin-coatedcoverslip.

To directly test the repellent activity of Ntn-1, we used a stripeassay (Fig. 2K) to prepare a substrate of alternating bands ofrecombinant Ntn-1 and laminin (Fig. 2 K and L). The aSCPswere clearly repelled by Ntn-1, as indicated by the near-totalabsence of cells attaching to stripes of Ntn-1. These resultssupport the conclusion that aSCPs are not repelled in responseto Slit-2 but only in response to Ntn-1. In both slice and stripeassays, Ntn-1 repulsion appears to depend on DCC (deleted incolorectal cancer) function (SI Fig. 9 A and B).

The Repellent Action of Ntn-1 Is Not Mediated by a Proliferation orSurvival Mechanism. aSCPs were cultured with CM from HEK,HEK-Slit2, HEK-Ntn1, and recombinant Ntn-1 for 3 days. The cellswere pulsed with BrdU and fixed after 6 h, 12 h, 1 DIV, 2 DIV, or3 DIV. BrdU-positive cells were observed as soon as 6 h after thepulse with HEK-Slit2 and at 12 h for HEK. The number of positivecells remained increased and was similar for both conditions until3 DIV (SI Fig. 10). In contrast, very few BrdU-positive cells wereobserved with HEK-Ntn1 and recombinant Ntn-1. After TUNELstaining, very few labeled cells were observed in any of theconditions. Ntn-1 does not increase aSCP death or proliferation andfurther supports the central concept of Ntn-1 mediating a migratoryresponse of the aSCPs.

Slit-2, Ephrins-B, and Shh Are Not Implicated in the FP MigrationPattern. Although these results support the conclusion that Ntn-1functions as an effective FP-derived repellent, they do not establishwhether Ntn-1 is sufficient to cause repulsion from the FP in theorganotypic slice model. Therefore, we investigated the potentialcontributions of Slit-2, Ephrin-B3, Shh, and Ntn-1 to the repulsive

Fig. 1. aSCPs migrate in response to embryonic molecular cues expressed in theVN. (A) The neural tube was cut along the dorsal midline, opened, and sectionedtoobtainorganotypic slices. (B)aSCPswerecoculturedwithE12slicepreparationsof SC. An aggregate of aSCPs was placed at the midline, over the FP. (C and D)After 2 DIV, aSCPs migrated away from the FP and accumulated on each side ofthe FP. After 5 DIV, aSCPs established a clear and reproducible pattern of migra-tion. (D) A transverse z-stack reconstruction shows the absence of aSCPs at the FPand that cells penetrate the slice. (E) To have the roof plate (RP) centered, theneural tube was cut along the ventral midline and opened. The aSCPs did notmigrate away from the midline. (Scale bars, 100 �m.)

17838 � www.pnas.org�cgi�doi�10.1073�pnas.0703240104 Petit et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

0, 2

020

activity of the FP and to the induction of the migration pattern byrepeating the E12SC slice–aSCP coculture with methods to perturbtheir function.

We first evaluated the possible involvement of Slit-2 byinhibiting its extracellular action. We added CM containingRobo/Fc with the extracellular portion of Robo fused to an Fcdomain of the human Ig to the E12SC slice–aSCP coculture.Robo/Fc binds to extracellular Slits and inhibits the function ofthose expressed in the SC slice (25). The addition of Robo-FcCM did not affect frequency of pattern formation (SI Fig. 8 F–I).

Next, we tested the role of Ephrin-B3 by preparing SC organo-typic slices from Ephrin-B3 KO embryos. Down-regulation ofephrin-B3 in the SC slice did not affect the aSCP migration pattern.The FP pattern was still observed for 77% of Ephrin-B3�/�, 83% ofEphrin-B3�/�, and 87% of Ephrin-B3�/� SC slices (SI Fig. 8 L–Nand Q). We also blocked the ability of the aSCPs to respond toEphrin-B family by isolating a fresh population of aSCPs fromEphB1/B3 double knockout (KO) mice (SI Fig. 11 D–F). The FPpattern was observed for 90% of EphB1/B3�/�-derived aSCPs anddid not significantly differ from the probability of pattern formationobserved in wild-type aSCPs (97%) (SI Fig. 8 O–Q).

To determine whether Shh is required for FP repulsion, wepretreated aSCPs with cyclopamine, a plant alkaloid that selectivelyinhibits Shh signaling (26). We also preincubated the SC slices withShh-N antibody against the biologically active amino-terminalfragment of Shh (27). After these two treatments, aSCPs stillmigrated away from the FP (in 88–96% of the slices; SI Fig. 8 R–V).These experiments suggest that Slit-2, Ephrin-B3, and Shh have alimited role in the elaboration of the FP migration pattern.

Down-Regulation of Ntn-1 Disrupts the Pattern of Migration. Todetermine whether Ntn-1 signaling is required for the repellentactivity of the FP and the establishment of the migration pattern,we repeated the coculture assay under conditions of decreasedNtn-1 expression. First, when cocultured with Ntn-1�/� SC orga-notypic slices, aSCPs stayed at the FP and did not migrate selec-tively to the VN (Fig. 3 A–D). Ninety-four percent of the wild-typeSC slices resulted in the typical FP migration pattern, whereas only15% of the Ntn-1�/� SC and 63% of the Ntn-1�/� exhibited thepattern.

Because Ntn-1 is absent throughout development of the Ntn-1KO, it could, during development, conceivably influence the ex-pression of other guidance cues downstream. We sought to controlfor the effects of altered development of the Ntn-1�/� mouse bycreating a model to selectively down-regulate Ntn-1 expression in awild-type organotypic slice by using an Ntn-1 siRNA (Fig. 3 E–Kand Tables 1 and 2). We found that siRNA against Ntn-1 success-fully and selectively down-regulated Ntn-1 gene expression inHEK-Ntn1 cells compared with nontreated (32.6%), GAPDHsiRNA-treated (27.9%), and negative siRNA-treated (27.9%) cells(Fig. 3 E and Table 1). GAPDH siRNA (positive control) and thescrambled siRNA as negative control did not decrease Ntn-1expression.

We next examined if siRNA (Ntn-1) application on E12SC slicesblocked aSCP migration or prevented the development of the FPmigration pattern. After 3 days of siRNA treatment, the FP nolonger promoted migration, and only 28% of the slices exhibited theFP pattern compared with 97% of controls (Fig. 3 F–J). After Ntn-1siRNA treatment, there is a corresponding decrease in �-gal

Ntn-1. (I and J) HEK (I) and HEK-Ntn1 (J) aggregates cocultured with aggregatesof aSCPs in collagen matrix. After 3 DIV, the aSCPs migrated into the aggregatesof HEK, but they did not migrate in the HEK-Ntn1 aggregate even when in closecontactwith it. (K)Thestripeassay. (L)aSCPsculturedonasubstrateofalternatingstripes of Ntn-1 (lower images illustrate the position of Ntn-1/blue bead stripes)andlaminin.aSCPsgrewonlaminin interstripesbutdidnotgrowonNtn-1stripes.(Scale bars, 100 �m.)

Fig. 2. aSCPs are repulsed by Ntn-1. (A) RT-PCR amplification of RNA isolatedfrom aSCPs in proliferation conditions. aSCPs expressed the five receptors forNtn-1: DCC, neogenin, UNC5A, UNC5B, and UNC5C (lane 1, ladder; lane 2,positive control with RNA isolated from E10.5 mouse brain; lane 3, RNA isolatefrom aSCP; lane 4, negative control containing no cDNA; lane 5, negativecontrol, cDNA treated with DNase). (B) The migration assay. The substrate foraSCPs is a patch of HEK cells on a laminin-coated coverslip. (C–F) aSCPs werecultured on HEK (C and D) or HEK-Ntn-1 (E and F) patch substrates. The aSCPsgrew randomly on control HEK cells and laminin whereas they migrated awayfrom the patch of HEK-Ntn-1 cells and aligned at the border between HEK-Ntn-1 and laminin. (G) Transwell insert migration assay. The aSCPs are placedin the upper chamber, and the CM or recombinant Ntn-1 is in the lowerchamber. (H) Percentage of cells crossing the membrane insert. aSCPs mi-grated similarly into the HEK CM, HEK-Slit2 CM, and control Neurobasal.Significantly fewer cells migrated into the HEK-Ntn1 CM and 100–200 ng/ml

Petit et al. PNAS � November 6, 2007 � vol. 104 � no. 45 � 17839

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

Mar

ch 1

0, 2

020

expression at the FP of slices prepared from Ntn-1�/� slices (Fig. 3J)compared with the nontreated slices (Fig. 3I). siRNA against Ntn-1applied to the SC slices down-regulated Ntn-1 gene expressioncompared with nontreated (32%), GAPDH siRNA-treated(41.8%), and neg siRNA-treated (30.8%) slices. The expression ofEphrin-B3 or GAPDH was not significantly affected (Fig. 3 K and

Table 2). As a control for the effects of perturbations on the extentof migration from the FP, we measured the most lateral spread ofaSCPs on slices and found no differences (control: 158 � 13 �m;Net-1 KO: 171 � 10 �m; Net-1 siRNA: 191 � 15 �m). These dataindicate that among the guidance molecules that were tested, onlyNtn-1 plays a major repellent role in the migration pattern ofaSCPs.

Ntn-1 Is Sufficient and Necessary to Repel aSCPs from an Injury Core.These data show that Ntn-1 is up-regulated at the injury site afteran SC lesion. The lesion core is a zone that becomes devoid ofneural elements over time. Because we demonstrated that aSCPsare repelled by Ntn-1, it could participate in the repulsion of aSCPsfrom the lesion site after an SC injury. We developed an in vitromodel to study the repulsion of aSCPs observed in vivo (Fig. 4A).Organotypic SC slices, prepared from an adult mouse after an SChemisection, were cocultured with aSCPs. After a hemisection,Ntn-1 expression is increased in dorsal white matter compared withother regions (Fig. 4 G and H). In the uninjured preparation, aSCPsremained within the dorsal columns where they were originallyplaced (Fig. 4B; 606 � 46 �m from the central canal). However, 3days postinjury, aSCPs migrated away from the injury core that wasdelineated by a gliotic ring and proteoglycan deposition (Fig. 4C;285 � 50 �m; P � 0.0001; n � 39). The model reproduces keyelements of the repulsion seen after hemisection injury and allowsus to interrogate the molecular signals responsible for repulsionfrom the lesion core. To determine if Ntn-1 could be totally orpartially implicated in the repulsive activity at the injury site, wetreated the injured SC slices with Ntn-1 siRNA for 3 DIV beforeplacing the cells at the dorsal columns. In two replicate experiments,we found that Ntn-1 siRNA treatment significantly down-regulatedNtn-1RNA (Fig. 4F). Importantly, Ntn-1 down-regulation (Fig. 4E;613 � 85 �m) or blocking of DCC function (SI Fig. 9C) preventedthe repulsion of aSCPs from the lesion core, whereas digestion ofcondroitin sulfate proteoglycans had no effect (P � 0.0001; n � 18;

Table 1. Comparison of Comparison of Ntn-1 expression aftersiRNA treatment on HEK-Ntn1 cells

Geneexpression Treatment

Treatment

Ntn-1 siRNA,%

GAPDH siRNA,%

neg siRNA,%

Ntn-1 No treatment 32.6 7.3 19.2Ntn-1 Ntn-1 siRNA 27.9 27.9GAPDH No treatment 19.2 30.6 7.9

The relative expression of Ntn-1 and GAPDH was quantified from theintensity of bands in Fig. 3E. Bold underlined text indicates a significantdecrease of Nnt-1 expression after Ntn-1 siRNA treatment.

Table 2. Comparison of Ntn-1 expression after siRNA treatmenton E12 SC slices

Geneexpression Treatment

Treatment

Ntn-1siRNA, %

GAPDHsiRNA, %

negsiRNA, %

Ntn-1 No treatment 32 16.8 2.7Ephrin-B3 10 10.8 10.7Ntn-1 Ntn-1 siRNA 41.8 30.8Ephrin-B3 18.8 0.7GAPDH for Ntn-1 No treatment 2.1 23.8 1.2GAPDH for

Ephrin-B35.5 13.3 5.2

The relative expression of Ntn-1 and GAPDH was quantified from theintensity of bands in Fig. 3K. Bold underlined text indicates a significantdecrease of Nnt-1 expression after Ntn-1 siRNA treatment.

Fig. 3. aSCPs are repulsed by the FP in response to Ntn-1. (A–L) Ntn-1function-blocking experiment. aSCPs were cultured on organotypic SC slicesisolated from Ntn-1 KO embryos (A–C). After 5 DIV, the aSCPs showed a weakpattern of migration for 63% of Ntn-1�/� slices (B) and only for 15% of theNtn-1�/� slices (*, P � 0.005 and P � 0.0001, respectively). (E–H) The functionof Ntn-1 was also blocked by using Ntn-1 siRNA. (E) We first tested theefficiency by applying 30 nM Ntn-1 siRNA with siPORT amine transfectionagent to HEK-Ntn1 cells for 3 DIV, silencer-positive control GAPDH siRNA, andsilencer-negative control siRNAs were used. HEK-Ntn1 cells had a significantdecrease in Nnt-1 expression. (F) Experimental design. (G and H) After 3 daysof Ntn-1 siRNA treatment, only 28% of the slices still presented a migrationpattern (*, P � 0.0001). (I and J) After Ntn-1 siRNA treatment of SC slicesprepared from Ntn-1�/� embryos, the �-gal expression at the FP decreasedcompared with the nontreated slices. (K) To control the efficiency of the siRNAtreatment, total RNA was isolated from E12SC slices, and expression of Ntn-1,GAPDH, and Ephrin-B3 was evaluated by RT-PCR. (D and H) The percentage ofslices that exhibit the FP pattern. These data show that a decrease in Ntn-1expression almost completely blocked the migration of aSCPs and the estab-lishment of the FP pattern. (Scale bars, 100 �m.)

17840 � www.pnas.org�cgi�doi�10.1073�pnas.0703240104 Petit et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

0, 2

020

Fig. 4D). These data demonstrate that (i) organotypic slice prep-arations of the lesion core lead to repulsion of aSCPs and (ii)down-regulation of Ntn-1 is sufficient to allow aSCPs to survive andmigrate within the lesion zone.

DiscussionThe experiments outlined here lead to a better understanding of thecues that affect the motility of adult SC stem cells in the intact andinjured CNS. We demonstrated that aSCPs migrate and form an FPpattern of resettlement. Our observations confirm that Ntn-1 alonemediates the observed aSCP repulsion and the establishment of theFP migration pattern. The aSCP expresses DCC, neogenin, UNC5A,UNC5B, and UNC5C. Manipulation of Ntn-1 in a variety of in vitroparadigms demonstrates that Ntn-1 produces a clear repulsiveeffect. Importantly, we also demonstrate that Ntn-1 significantlyparticipates in the repulsion of aSCPs from an SC injury site in anin vitro model of SC injury. These data establish Ntn-1 as a key aSCPrepulsive cue and illuminate the clear dominance of this cuecompared with either eph/ephrin family members, Slit-2, Shh, orinjury-related inhibitors like condroitin sulfate proteoglycans.Hence, Ntn-1 is likely a key mediator influencing aSCP migrationin the intact CNS but also may influence the mode of progenitormigration in settings of injury and degeneration.

aSCPs Migrate in Response to a Guidance Cue. Although studies havedocumented that directed progenitor migration occurs in the adultbrain and SC, the underlying molecular cues of these processes arestill largely unknown (14–19, 28–30). Our data indicate that Ntn-1is a major mediator of adult NSC repulsion. Netrins remainconstitutively expressed in the adult vertebrate CNS, but theirnormal or postinjury roles are unknown (9, 11, 12, 31, 32). Studiesin the lamprey demonstrate a continued role of adult-derived Ntn-1as an axonal repellent (33) and DCC may function as a negativeregulator of cell motility. Finally, the relative level of adult UNC-5protein increases while DCC expression decreases compared withthe embryonic SC, potentially indicating a dominant repulsive rolefor Ntn-1 in the adult (9, 34). In a complex CNS injury environment,many factors other than Ntn-1 expression alone could contribute tothe repulsion of aSCPs. For example, it is suggested that Ntn-1 cancollaborate with other molecules such as Shh (35, 36) and Slit (25)to guide axons, and that extracellular factors such as laminin-1 canaffect Ntn-1 responsiveness through effects on the intracellular levelof cAMP (37). The expression of receptors by the aSCP can alsoaffect their response to Ntn-1. By itself, DCC mediates attraction,but DCC and UNC5 receptor families can form a complex thatmediates repulsion (38). The relative amount of cell-surface DCCor UNC-5 protein may determine the type of cell response:attraction or repulsion (9). For example, after optic nerve injury, adown-regulation of Ntn receptors modifies the sensitivity of axonsto Ntn (12, 39). Future work is needed to actively track andmanipulate migration cues and their cognate receptors in an in vivosetting. In addition, the molecular pathway responsible for Ntn-1signaling in aSCP needs to be determined to guide molecularapproaches in vivo.

aSCPs Are Repulsed by Ntn-1 Through a Short-Range Response. Therehas been a short-range function attributed to Ntn-1 (40, 41), and ithas been shown that axon outgrowth-promoting activity and sub-cellular distribution of Ntn-1 is mainly associated with the mem-brane fraction (11, 42). Here, aSCPs seemed to be repulsed onlywhen in close proximity to a source of Ntn-1, so Ntn-1 may act tocreate tight regional borders in the intact CNS and also in an injurysetting. The pattern of migration we observed indicates that thecells stop directly after exiting the FP region where Ntn-1 is highlyexpressed. However, Ntn-1 is also expressed at a lower level in theventral two-thirds in the mouse neural tube (43). Hence, the highconcentration of Ntn-1 could be a critical factor. The collagencoculture assay also suggests that aSCPs need to be in closeproximity with the Ntn-1 source to activate repulsion. In our Ntn-1down-regulation experiments, the concentration of Ntn-1 needed tobe high to maintain repellent activity.

Ntn-1 Is Sufficient to Repel aSCPs from the Central Lesion Core. Theseexperiments demonstrate that Ntn-1 is sufficient and necessary torepel aSCPs from an acute SC lesion. Ntn-1 expression is up-regulated at the lesion core after an SC hemisection (44) and asciatic nerve lesion (10). Because of its expression by oligodendro-cytes, Ntn-1 may be an inhibitor of axon regeneration after SCinjury (11, 45). Our in vitro model of SC injury shows thatdown-regulation of Ntn-1 is sufficient to block the repulsive capac-ity of the injury site. Other repulsive molecules could also partic-ipate in this repulsive activity (46). In a complex in vivo environ-ment, these other candidates will have to be considered.Nonetheless, our data and the existing literature place Ntn-1 at thesite of an early SC lesion and establish Ntn-1’s role in setting lesionboundaries assumed by the aSCPs after a traumatic injury.

In conclusion, the identification of an endogenous factor thatregulates aSCP repulsion from the necrotic/cystic zone after an SCinjury will provide an important tool toward endogenous cellularregeneration and lesion repair. Future research directions shouldfocus on the broad applicability of our observations to establishwhether Ntn-1 is a repellent cue for all adult CNS-derived progen-

Fig. 4. In an in vitro SC injury model, application of Ntn-1 siRNA decreased therepulsion of aSCPs from the injury site. (A) Experimental design. (B–E) CoronaladSC slice cocultured with aSCPs. (B) In the control adSC slices with no injury, theaSCPs placed at the dorsal columns did not migrate outside this original position.(C) After hemisection, the aSCPs migrated away from the injury site, delimited byglial fibrillary acidic protein/condroitin sulfate proteoglycans (GFAP/CSPG) reac-tive area. (D) After treatment with chondroitinase ABC, the aSCPs still migratedaway from the injury site. (E) After treatment with 30 nM Ntn-1 siRNA, the aSCPsstayed at the injury site, similar to control. (F) To determine the efficiency of thesiRNAtreatment, totalRNAwas isolatedfromadSCslices,andexpressionofNtn-1was evaluated by RT-PCR. Nnt-1 expression decreased significantly. 1, no le-sion/no siRNA; 2, no lesion/Ntn1 siRNA; 3, lesion/no siRNA; 4, lesion/Ntn1 siRNA.(G and H) Ntn-1 expression is up-regulated at the dorsal columns after hemisec-tion. No expression was observed in other regions of the SC slice or uninjuredslices. (Scale bars, 100 �m.)

Petit et al. PNAS � November 6, 2007 � vol. 104 � no. 45 � 17841

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

Mar

ch 1

0, 2

020

itors and whether Ntn-1 expression is typical of other CNS lesionsand mammalian species. Understanding the pathways involved inNtn-1-mediated NSC guidance is likely to provide insight into thefunction of other guidance molecules expressed in the adult CNSand lead to strategies for therapeutically directing NSC migrationwithin the damaged brain.

Materials and MethodsSee SI Experimental Procedures for further details.

Isolating and Coculture of aSCPs on Embryonic SC Slices. Embryonic(E12) SC progenitor cells and aSCPs (3 mo) were isolated fromtransgenic mice expressing GFP under the �-actin promoter and anEph B1/B3 double KO. Organotypic slices were prepared fromneural tubes from C57/BL-6, Ntn-1, and ephrin-B3 (E11.5–12.5)embryonic KO mice.

Migration Assays. Collagen gel cocultures. Cell aggregates were pre-pared as previously descried (47). aSCP aggregates were positioned0–200 �m apart from HEK-Ntn1 or HEK-Slit2 cell aggregates for3 DIV (25).Stripe assay. Recombinant mouse Ntn-1 (200 ng/ml; R&D Systems,Minneapolis, MN) was mixed with blue fluorescent polymer mi-crospheres (Duke Scientific, Palo Alto, CA). A coverslip precoatedwith mouse laminin (10 �g/ml; Invitrogen, Carlsbad, CA) was putonto a silicone matrix with 90-�m-wide parallel channels (MaxPlank Institute, Tubingen, Germany), and the Ntn-1 was injected inthe channels (48). aSCPs were plated onto the stripes substrate for3 DIV.Cell patch. Cell suspensions of HEK, HEK-Slit2, and HEK-Ntn1were plated on laminin-coated coverslips. After 1 h, aSCPs werecultured on this patch substrate of adherent HEK cells and lamininfor 3 DIV.Transwell migration assay. aSCP suspensions were plated at a densityof 4 � 105 cells per ml on noncoated or laminin-coated polycar-bonate transwell culture inserts (Corning, Acton, MA). Cell mi-gration was evaluated for six conditions, including CM from HEK,HEK-Slit2, HEK-Ntn1, and recombinant Ntn-1 at 50, 100, and 200ng/ml. At 24 or 48 h, cells were scraped from the upper and lowerfilter and counted.Function-blocking assay. SC organotypic slices were dissected fromEphrin-B3 (49, 50) or an Ntn-1 KO mouse (51). The Ntn-1 mutant

allele is severely hypomorphic but is not a complete null. Hetero-zygote and homozygote slices were compared with the wild typefrom the same litter.

To inhibit Ntn-1 function, we used small interfering RNAs(siRNA) to down-regulate Ntn-1 expression. We also inhibited theNtn-1 by blocking the function of the DCC receptor with amonoclonal DCC antibody [OP45SP1 DCC (Ab-1) EMD Bio-sciences/Calbiochem, San Diego, CA]. To inhibit Slit-2 function, weadded Robo1/Fc-CM (2:1, 1:1) (52). We controlled the activity ofSlit-2 and Robo1/Fc-CM with a motor neuron repulsion assay (25).To block the SHH signaling, we used either SHH-N (5E1 IgG;Developmental Studies Hybridoma Bank, University of Iowa, IowaCity, IA) or inhibited the function of Smoothened (SMO; part ofSHH-receptor complex) with cyclopamine; tomatidine is a negativecontrol (5, 10, and 20 �M; Toronto Research Chemicals, Toronto,ON, Canada).

SC Injury Model. Female C57B/6 mice (3 mo) were subjected to a(T8) dorsal hemisection as previously described (22). One hourafter injury, the mouse was killed, and coronal slices were preparedfrom the lesion core and noninjured animals and cultured for 6DIV. siRNA and transfection agent (Ambion, Foster City, CA) wasapplied at 0 h, 1 DIV, and 2 DIV after the hemisection; and after3 DIV, aSCP aggregates were placed at the dorsal columns,representing the acute injury site. The slices were also treated withchondroitinase ABC (53).

Analysis and Image Processing. The presence of an FP pattern wasdefined as the accumulation of the majority (�90%) of aSCPs oneach side of the FP. Statistical analysis was performed by a Fisher’sexact test with a confidence interval set at 95% and a t test(GraphPad Prism; GraphPad, San Diego, CA). In the in vitro injurymodel, we evaluated cell migration by determining the distance ofthe aSCP GFP intensity center from the central canal (Metamorph6.2). All experiments were repeated at least three times.

We thank Dr. Jane Wu (Northwestern University, Chicago) for providingHEK-Slit2 cells; Dr. Robert Rostomily (University of Washington) forcyclopamine; and Loan Nguyen and Don Maris for technical assistance.A.P. is a Fonds de la Recherche en Sante du Quebec fellow. This work wassupported by grants from the Craig H. Nielsen Foundation and NationalInstitutes of Health Grant R01 N546724 (to P.J.H.). P.J.H. is a member ofthe University of Washington Institute for Stem Cell and RegenerativeMedicine and The Center on Human Development and Disability.

1. Hagg T (2005) Trends Neurosci 28:589–595.2. Dirks PB (2001) J Neurooncol 53:203–212.3. Picard-Riera N, Nait-Oumesmar B, Baron-Van Evercooren A (2004) J Neurosci Res 76:223–231.4. Silver J, Miller JH (2004) Nat Rev Neurosci 5:146–156.5. Carlson BM (2005) Anat Rec B New Anat 287:4–13.6. Ferretti P, Zhang F, O’Neill P (2003) Dev Dyn 226:245–256.7. Tessier-Lavigne M, Goodman CS (1996) Science 274:1123–1133.8. Yamada T, Pfaff SL, Edlund T, Jessell TM (1993) Cell 73:673–686.9. Manitt C, Kennedy TE (2002) Prog Brain Res 137:425–442.

10. Madison RD, Zomorodi A, Robinson GA (2000) Exp Neurol 161:563–570.11. Manitt C, Colicos MA, Thompson KM, Rousselle E, Peterson AC, Kennedy TE (2001)

J Neurosci 21:3911–3922.12. Ellezam B, Selles-Navarro I, Manitt C, Kennedy TE, McKerracher L (2001) Exp Neurol

168:105–115.13. Goldshmit Y, McLenachan S, Turnley A (2006) Brain Res Brain Res Rev 52:327–345.14. Morshead CM, Craig CG, van der Kooy D (1998) Development (Cambridge, UK) 125:2251–2261.15. Lois C, Alvarez-Buylla A (1994) Science 264:1145–1148.16. Lois C, Garcia-Verdugo JM, Alvarez-Buylla A (1996) Science 271:978–981.17. Doetsch F, Garcia-Verdugo JM, Alvarez-Buylla A (1997) J Neurosci 17:5046–5061.18. Lin RC, Matesic DF, Marvin M, McKay RD, Brustle O (1995) Neurobiol Dis 2:79–85.19. Frisen J, Johansson CB, Torok C, Risling M, Lendahl U (1995) J Cell Biol 131:453–464.20. Imitola J, Raddassi K, Park KI, Mueller FJ, Nieto M, Teng YD, Frenkel D, Li J, Sidman

RL, Walsh CA, et al. (2004) Proc Natl Acad Sci USA 101:18117–18122.21. Shihabuddin LS, Horner PJ, Ray J, Gage FH (2000) J Neurosci 20:8727–8735.22. Horky LL, Galimi F, Gage FH, Horner PJ (2006) J Comp Neurol 498:525–538.23. Liebl DJ, Morris CJ, Henkemeyer M, Parada LF (2003) J Neurosci Res 71:7–22.24. Tessier-Lavigne M, Placzek M, Lumsden AG, Dodd J, Jessell TM (1988) Nature 336:775–778.25. Brose K, Bland KS, Wang KH, Arnott D, Henzel W, Goodman CS, Tessier-Lavigne M, Kidd

T (1999) Cell 96:795–806.26. Incardona JP, Gaffield W, Kapur RP, Roelink H (1998) Development (Cambridge, UK)

125:3553–3562.27. Ericson J, Morton S, Kawakami A, Roelink H, Jessell TM (1996) Cell 87:661–673.28. Nguyen-Ba-Charvet KT, Picard-Riera N, Tessier-Lavigne M, Baron-Van Evercooren A,

Sotelo C, Chedotal A (2004) J Neurosci 24:1497–1506.

29. Gotts JE, Chesselet MF (2005) J Neurosci Res 80:160–171.30. Jin K, Sun Y, Xie L, Peel A, Mao XO, Batteur S, Greenberg DA (2003) Mol Cell Neurosci

24:171–189.31. Livesey FJ, Hunt SP (1997) Mol Cell Neurosci 8:417–429.32. Deiner MS, Kennedy TE, Fazeli A, Serafini T, Tessier-Lavigne M, Sretavan DW (1997)

Neuron 19:575–589.33. Cohen AH, Mackler SA, Selzer ME (1988) Trends Neurosci 11:227–231.34. Manitt C, Thompson KM, Kennedy TE (2004) J Neurosci Res 77:690–700.35. Charron F, Stein E, Jeong J, McMahon AP, Tessier-Lavigne M (2003) Cell 113:11–23.36. Salinas PC (2003) Trends Neurosci 26:641–643.37. Shewan D, Dwivedy A, Anderson R, Holt CE (2002) Nat Neurosci 5:955–962.38. Hong K, Hinck L, Nishiyama M, Poo MM, Tessier-Lavigne M, Stein E (1999) Cell 97:927–941.39. Petrausch B, Jung M, Leppert CA, Stuermer CA (2000) Mol Cell Neurosci 16:350–364.40. Mitchell KJ, Doyle JL, Serafini T, Kennedy TE, Tessier-Lavigne M, Goodman CS, Dickson

BJ (1996) Neuron 17:203–215.41. Winberg ML, Mitchell KJ, Goodman CS (1998) Cell 93:581–591.42. Serafini T, Kennedy TE, Galko MJ, Mirzayan C, Jessell TM, Tessier-Lavigne M (1994) Cell

78:409–424.43. Kennedy TE, Wang H, Marshall W, Tessier-Lavigne M (2006) J Neurosci 26:8866–8874.44. Wehrle R, Camand E, Chedotal A, Sotelo C, Dusart I (2005) Eur J Neurosci 22:2134–2144.45. Manitt C, Wang D, Kennedy TE, Howland DR (2006) J Neurosci Res 84:1808–1820.46. Niclou SP, Ehlert EM, Verhaagen J (2006) J Neurotrauma 23:409–421.47. Kennedy TE, Serafini T, de la Torre JR, Tessier-Lavigne M (1994) Cell 78:425–435.48. Walter J, Kern-Veits B, Huf J, Stolze B, Bonhoeffer F (1987) Development (Cambridge, UK)

101:685–696.49. Henkemeyer M, Orioli D, Henderson JT, Saxton TM, Roder J, Pawson T, Klein R (1996)

Cell 86:35–46.50. Williams SE, Mann F, Erskine L, Sakurai T, Wei S, Rossi DJ, Gale NW, Holt CE, Mason

CA, Henkemeyer M (2003) Neuron 39:919–935.51. Serafini T, Colamarino SA, Leonardo ED, Wang H, Beddington R, Skarnes WC, Tessier-

Lavigne M (1996) Cell 87:1001–1014.52. Wu W, Wong K, Chen J-h, Dupuis S, Wu JY, Rao Y (1999) Nature 400:331–336.53. Asanbaeva A, Masuda K, Thonar EJ, Klisch SM, Sah RL (2007) Arthritis Rheum 56:188–

198.

17842 � www.pnas.org�cgi�doi�10.1073�pnas.0703240104 Petit et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

0, 2

020