Implication of neuropilin 2/semaphorin 3F in retinocollicular map formation

PATTERNS & PHENOTYPES

Implication of Neuropilin 2/Semaphorin 3F inRetinocollicular Map FormationT. Claudepierre,1† E. Koncina,2† F.W. Pfrieger,1 D. Bagnard,2 D. Aunis,2 and M. Reber2*

Neural representations of the environment within the brain take the form of topographic maps whoseformation relies on graded expression of axon guidance molecules. Retinocollicular map formation, fromretinal ganglion cells (RGCs) to the superior colliculus (SC) in the midbrain, is mainly driven by Ephreceptors and their ligands ephrins. However, other guidance molecules participate in the formation of thismap. Here we demonstrate that the receptor Neuropilin-2 is expressed in an increasing nasal–temporalgradient in RGCs, whereas one of its ligands, Semaphorin3F, but not other Sema3 molecules, presents agraded low-rostral to high-caudal expression in the SC when mapping is underway. Neuropilin-2 and itscoreceptor Plexin A1 are present on RGC growth cones. Collapse assays demonstrate that Semaphorin3Finduces significant growth cone collapse of temporal, but not nasal, RGCs expressing high levels ofNeuropilin-2. Our results suggest that Neuropilin-2/Semaphorin3F are new candidates involved inretinotopy formation within the SC. Developmental Dynamics 237:3394–3403, 2008. © 2008 Wiley-Liss, Inc.

Key words: axon guidance; neural map; semaphorin; neuropilin

Accepted 1 September 2008

INTRODUCTION

Organisms require precise represen-tations of their surroundings to re-spond appropriately to sensory stim-uli. These representations often takethe form of topographic maps, wherethe spatial relationship between a setof input neurons is conserved follow-ing their projections into the targetregions. One of the best studied repre-sentations is the retinocollicular map(or retinotectal map in nonmamma-lian vertebrates), where ganglion cellsin the retina (RGCs) project to one oftheir targets in the midbrain, the su-perior colliculus (SC). During the de-velopment of these projections, RGCaxons navigate along a stereotypic

pathway from the retina to the SCbetween embryonic day (E) 16 andpostnatal day (P) 0 then undergo to-pographic organization in the SC dur-ing the first postnatal week (Erskineand Herrera, 2007). On their way tothe SC, they respond to molecularcues along the optic pathway and tocues that elicit terminal branchingand topographic (retinotopic) mappingwithin the SC. The retinocollicularmap is defined by projections of thenasal–temporal and the dorsal–ven-tral axes in the retina, along the cau-dal–rostral and lateral–medial axes,respectively, in the SC. Sperry’s che-moaffinity hypothesis is the favoritemodel to explain how the topographic

maps are formed. It states that thespecificity of topographic projections isdetermined by interacting molecular“tags” that are distributed in comple-mentary gradients on both projectingaxons and their targets (Sperry,1963). Within the retinocollicular sys-tem, the receptor tyrosine kinasesEphs and their ligands the ephrinsare considered as one of those molec-ular tags (Flanagan and Vanderhae-ghen, 1998; McLaughlin and O’Leary,2005). Gradients of Eph receptors andephrin ligands drive the formation ofthe topographic retinocollicular map;however, it has been suggested thatother guidance molecules are involvedin this process (Feldheim et al., 2000,

1Department of Neurotransmission/Neuroendocrine Secretion, Inst. Cell. Integ. Neurosci. (INCI) UMR 7168/L2 CNRS/ULP, Centre deNeurochimie, Strasbourg, France2INSERM U.575 Physiopathologie du Systeme Nerveux, Strasbourg, FranceGrant sponsor: University Louis Pasteur; Grant number: BQR 2005-086, INSERM.†Drs. Claudepierre and Koncina contributed equally to this work.*Correspondence to: M. Reber, Molecular Neurobiology Laboratory, The Salk Institute, 10010 N. Torrey Pines Road, La Jolla,CA 92037. E-mail: [email protected]

DOI 10.1002/dvdy.21759Published online 15 October 2008 in Wiley InterScience (www.interscience.wiley.com).

DEVELOPMENTAL DYNAMICS 237:3394–3403, 2008

© 2008 Wiley-Liss, Inc.

2004; Brown et al., 2000; Hindges etal., 2002; McLaughlin et al., 2003b;Reber et al., 2004). For example, func-tional studies revealed that repulsiveguidance molecules (RGMs) and theirreceptor neogenin are involved inchick retinotectal map formation(Matsunaga et al., 2006) but no reti-nocollicular map defects were ob-served in RGMa mutant mice(Niederkofler et al., 2004). In addition,Wnt-Ryk signalling has been shown toparticipate in the development of theretinotectal map in both chick andmouse (Schmitt et al., 2006). Recently,Buhusi and collaborators demon-strated the involvement of the L1 ad-hesion molecule complexed to ankyrinin retinocollicular map formation (Bu-husi et al., 2008). Other guidance mol-ecule candidates are the semaphorins,particularly the semaphorin 3 class ofproteins. This family, composed ofseven members (Sema3A to Sema3G),are secreted molecules expressedthroughout the central nervous sys-tem (CNS) during development and inthe adult (Pasterkamp and Kolodkin,2003; Huber et al., 2003; de Winter etal., 2004). Sema3 signalling is medi-ated by heteromeric complexes of re-ceptors containing the binding-unitNeuropilin 1/2 (Npn-1/2) associatedwith transducing coreceptors such asmembers of the plexin-A subfamily,plexin A1/A2 (PlxA1/A2). The Sema3signaling system is involved in axonguidance in several CNS areas, in-cluding the visual system and recep-tor complexes triggering Sema3 sig-nalling are expressed in the retina ofpostnatal and adult mouse (Pas-terkamp and Kolodkin, 2003; Huberet al., 2003; de Winter et al., 2004; deWit and Verhaagen, 2003).

Retinocollicular map formation hasbeen mainly based on gradients of ret-inal receptors and collicular ligands(e.g., Eph receptors and ephrins li-gands). We tested whether Sema3 sig-naling could contribute to retinotopicmapping in the SC by analyzing theexpression of Sema3 in the SC and ofneuropilins/plexin As in the retina inpostnatal mice at P1 and P8. We de-tected a low-nasal to high-temporalgradient of Npn-2 in RGCs and ex-pression of Sema3 ligands in the SC ofnewborn mice. Interestingly, mRNAfor Sema3F, a high-affinity Npn-2 li-gand, displayed a low-rostral to high-

caudal expression in the SC at P1.Moreover, Sema3 expression becameundetectable at P8 when the map ismature. In addition, functional invitro assays demonstrated that tem-poral RGCs, exhibiting high Npn-2expression, were more sensitive toSema3F-induced repulsion and col-lapse than nasal RGCs. Taken to-gether our data suggest that retinalNpn-2 and collicular Sema3F are in-volved in retinocollicular map forma-tion.

RESULTS

Nasal–Temporal Gradient ofNeuropilin-2 in the RGCLayer

To explore the role of Sema3 signalingin retinocollicular map formation, weanalyzed by in situ hybridization theexpression patterns of neuropilins andplexin As in the mouse retina at P1and P8 when mapping is underway(Brown et al., 2000; Reber et al.,2004). Npn-2 was strongly expressedin the RGC layer at P1 (Fig. 1A). Highmagnification of the temporal and na-sal poles (insets in Fig. 1A) suggesteda graded expression. Analysis of the insitu hybridization signal confirmedthe presence of a Npn-2 gradient run-ning from low-nasal to high-temporal(Fig. 1E) but not along the dorsal-ven-tral axis (not shown). Npn-1 mRNAwas also detected in the RGC layer atP1, although expression levels werelow and appear uniform along boththe nasal–temporal (Fig. 1B) and dor-sal–ventral axis (not shown). TheplexinAs, PlxA1 and PlxA2, werestrongly and homogeneously ex-pressed in the RGC layer at P1 (Fig.1C,D). Semiquantitative reverse tran-scriptase-polymerase chain reaction(RT-PCR) on nasal versus temporalRGCs that were immunoisolated fromP1 mice and cultured for 2 days con-firmed the in situ hybridization data(Fig. 1E,F). Npn-2 showed a signifi-cantly sevenfold stronger expressionin the temporal pole compared to thenasal pole (relative expression in tem-poral RGCs, 0.48 � 0.06; and nasalRGCs, 0.07 � 0.05; P � 0.01 Wilcoxontest), whereas Npn-1, PlxA1, andPlxA2 exhibited no significant differ-ence between nasal and temporal

poles (Wilcoxon test) confirming theirhomogenous expression (Fig. 1F).

At P8, the graded Npn-2 expressionalong the nasal–temporal axis was nolonger detectable and replaced by auniform distribution (Fig. 2A and in-sets). Both in situ signal analysis andsemiquantitative RT-PCR confirmedthe uniform Npn-2 expression (Fig.2E,F). Similar to P1 retinae, Npn-1expression was low and uniform at P8(Fig. 2B,F). In addition, the expres-sion patterns of both PlxA1 and PlxA2did not change in RGCs (Fig. 2C,D,F)although strong PlxA1 expression wasdetected in the inner margin of theinner nuclear layer (INL; Fig. 2C, ar-rowhead in inset) in accordance withprevious data (de Winter et al., 2004).Expression levels of Npn-2, Npn-1,PlxA1, and PlxA2 were confirmed bysemiquantitative RT-PCR on nasalversus temporal RGCs immunoiso-lated from P8 mice and cultured for 2days. No statistical difference could berevealed for all the genes tested (Fig.2F, Wilcoxon test). The presence of agraded expression of Npn-2 in RGCsalong the nasal–temporal axis at P1suggests a role of this molecule in reti-nocollicular mapping along the pro-jecting rostral–caudal axis of the SC.We, therefore, analyzed the expres-sion of the Sema3 ligands in the SC.

Graded Expression ofSema3F in the SuperiorColliculus

In rodent midbrain, the SC presents alaminated structure composed of thestratum zonale (SZ, 20 �m), the stra-tum griseum superficiale (SGS, 200-400 �m) or superficial gray (SuG)layer and the stratum opticum (SO,100 �m). RGC axons enter the SCthrough the SO layer and then mi-grate upward to form synapses in theSGS/SuG layer (Huerta et al., 1983;Yamagata et al., 2006; May, 2006). Weanalyzed the expression of Sema3molecules in the SGS/SuG layer at P1and P8. In situ hybridization showedstrong expression of Sema3F in theSGS/SuG layer (dashed rectangle) atP1 (Fig. 3A) following a low-rostral tohigh-caudal gradient (Fig. 3B). In con-trast, Sema3A, 3C, 3D, and 3E dis-played weak and uniform signalsthroughout the SC (dashed rectanglein Fig. 3C–F). However, Sema3A, 3D,

NEUROPILIN 2/SEMAPHORIN 3F AND RETINOCOLLICULAR MAPS 3395

Fig. 1.

Fig. 2. Expression of the Sema3 receptor complexes in retina of P8mouse. A–D: Brightfield micrographs of retinal sections from P8 mice,with high magnification views of nasal and temporal poles (insets) afterin situ hybridization with probes for Npn-2 (A), Npn-1 (B), PlxA1 (C), andPlxA2 (D). GCL, ganglion cell layer; INL, inner nuclear layer. Arrowheadsindicate the inner margin of the INL in insets in C. E: Distribution of theNpn-2 signal along the nasal-temporal axis of the retina at P8.The linesrepresent counts from 3 different experiments. F: Mean transcript levelsof indicated genes normalized to Gapdh in nasal (N) and temporal (T)ganglion cells in the retina (RGCs) immunoisolated from P8 mice. Scalebars � 100 �m.

Fig. 3.

Fig. 1. Expression of the Sema3 receptor com-plexes in retina of postnatal day (P) 1 mouse.A–D: Brightfield micrographs of retinal sectionsfrom P1 mice, with high magnification views ofnasal and temporal poles (insets) after in situhybridization with probes for Npn-2 (A), Npn-1(B), PlxA1 (C), and PlxA2 (D). NBL, neuroblastlayer; GCL, ganglion cell layer; N, nasal; T, tem-poral. E: Distribution of the Npn-2 signal alongthe nasal–temporal axis of the retina at P1. Thelines represent counts from four different exper-iments. F: Mean transcript levels of indicatedgenes normalized to Gapdh in nasal (N) andtemporal (T) ganglion cells in the retina (RGCs)immunoisolated from P1 mice (n � 5, Wilcoxontest, **P � 0.01). Scale bars � 100 �m.

Fig. 3. Expression of Sema3 in superior collicu-lus (SC) of postnatal day (P) 1 mouse.A,C–F: Brightfield micrographs of SC sectionsfrom P1 mice along the rostral–caudal axis afterin situ hybridization with probes for Sema3F (A),Sema3A (C), Sema3C (D), Sema3D (E), andSema3E (F, arrows indicate rostral and caudalpoles; R, rostral pole; C, caudal pole; SC, su-perior colliculus; IC: inferior colliculus). Dashedlines indicate the SGS/SuG layer. B: Distribu-tion of the Sema3F signal in the SGS/SuG layeralong the rostral–caudal axis of the SC at P1.The lines represent counts from 3 different ex-periments. Scale bars � 1 mm.

3396 CLAUDEPIERRE ET AL.

3E, and 3F showed strong expressionin the inferior colliculus (IC) at P1,suggesting that, similar to ephrin-As,Sema3 molecules may form a repel-lent barrier preventing RGCs axonsfrom overshooting the SC during earlymap formation (Frisen et al., 1998).Sema3B was not detected at P1 (notshown). At P8, no signal was detectedfor any of the Sema3 molecules tested(not shown). The graded expression ofSema3F, unlike any other Sema3 mol-ecules, in the SGS/SuG where RGCaxons form synapses suggests a role ofSema3F in RGC axons targeting inthe SC.

Subcellular Localization ofSema3 Receptor Complexes

If Sema3F receptor complexes are in-volved in axon targeting, they must belocated on axons and/or growth cones.Therefore, we analyzed the distribu-tion of the main components trigger-ing Sema3F signaling, Npn-2 and itscoreceptor PlxA1 (Takahashi andStritmatter, 2001; Murakami et al.,2001; Sahay et al., 2003). Immunohis-tochemical analysis of their distribu-tion in RGCs in vivo is difficult due topossible colocalization in Muller glialcell endfeet, in astrocytes and in com-pacted axons in the optic nerve fiberlayer on top of the RGCs. Moreover, itis technically difficult to detect andlocalize receptor complexes on RGCaxons within the SC. Therefore, weperformed immunocytochemical stain-ing of Npn-2 and PlxA1 on culturednasal (Fig. 4A,C) and temporal (Fig.4B,D) RGCs purified by immunoisola-tion from P1 mice. Immunopanning ofRGCs has been used to purify postna-tal RGCs from rat (Barres et al., 1988;Meyer-Franke et al., 1995) and wasrecently adapted to mouse (Steinmetzet al., 2006; see Experimental Proce-dures section). This technique allowsthe visualization of individual RGCs,from the soma to the growth cones,that cannot be identified in otherpreparations such as retinal explants.Immunocytochemical staining ofRGCs isolated from P1 mice and cul-tured for 4 days, when a majority ofaxons (arrows in Fig. 4A–D) andgrowth cones (arrowheads in Fig.4A–D) can be clearly distinguished,shows a punctate signal for Npn-2 andPlxA1 (Fig. 4A–D). At high magnifica-

tion, Npn-2 and PlxA1 are clearly de-tected on both axons and growth cones(insets in Fig. 4B,D). Moreover, ourresults showed that temporal RGCsdisplay a stronger Npn-2 immunocy-tochemical signal when compared tonasal RGCs (Fig. 4A,B), confirmingthe in situ hybridization and RT-PCRdata. Overall, this punctuate stainingsuggests that the receptor complexesform clusters on axons/growth conemembranes, consistent with their rolein axon guidance.

Npn-2/Sema3F Binding inthe Retinocollicular System

We next determined whether Npn-2binds to Sema3 molecules in the SC insitu by receptor-affinity probe (RAP),which consists of a recombinant formof Npn-2 protein coupled to alkaline-phosphatase (Npn-2-AP; Huber et al.,2003). Sagittal sections of P1 SC incu-bated with Npn-2-AP show strong al-kaline phosphatase (AP) staining inthe SGS/SuG layers consistent withthe presence of Sema3F (Fig. 5A, ar-rowheads). Signal analyses reveal agraded Npn-2-AP staining, running

from low-rostral to high-caudal in ac-cordance with a Sema3F gradient(Fig. 5B). To confirm that the gradientof Sema3F is appropriately located inthe SC, we examined LacZ activity inP1 SC of Isl2-�LacZ mutant mice (Fig.5C). These mice express LacZ in 50%of the RGCs allowing visualization ofRGC axons innervating the SC (Tha-ler et al., 2004; Supp. Fig. 2 in Reberet al., 2004). Axons enter the SO (bluestripes Fig. 5C) and migrate upwardin the SGS/SuG where they form syn-apses (May, 2006). The location of theNpn-2 AP staining in serial sections(Fig. 5A) matches with the weak LacZstaining (Fig. 5C) in the SGS/SuGlayer, but not in the SO layer. Thisconfirms that the Sema3F protein gra-dient is indeed located in the SGS/SuG layer where RGCs form syn-apses. Control experiments (no Npn-2AP) show absence of specific AP stain-ing (Fig. 5D). Together these data sug-gest a role of Sema3F in rostral–cau-dal and/or lamina specific mapping inthe SC.

As a complementary approach, we de-termined whether RGCs bind Sema3Fduring retinocollicular map forma-

Fig. 4. Localization of the Sema3 receptor complex in cultured ganglion cells in the retina (RGCs)immunoisolated from nasal and temporal retinae. A–D: Fluorescence micrographs of nasal (A,C)and temporal (B,D) RGCs immunoisolated from P1 mice. RGCs were cultured for 2–4 days andimmunostained for Npn-2 (A,B) and PlxA1 (C,D). C,D: Higher magnification views (insets) of thegrowth cones. Scale bars � 50 �m in A–D; 10 �m in insets in B,C.

NEUROPILIN 2/SEMAPHORIN 3F AND RETINOCOLLICULAR MAPS 3397

tion. Using a recombinant-human-Sema3F-AP complex (Chen et al.,1997) as ligand-affinity probe, we ob-served strong binding of Sema3F inthe RGC layer at both P1 (Fig. 5E,F)and P8 (Fig. 5H,I). DAPI (4�,6-diami-dine-2-phenylidole-dihydrochloride)

staining of P8 retinal sections con-firms that Sema3F-AP labels theRGC layer (Fig. 5J). Control experi-ments using conditioned mediumfrom mock-transfected HEK cellsshow no AP staining (Fig. 5G). Al-though this technique enables visu-

alization of Sema3F binding toNpn-2, it masks the presence of theNpn-2 gradient due to the presenceof this receptor on RGC axons run-ning along the apical pole of the RGClayer. Furthermore, part of thestaining might be due to Sema3F-APbinding to other receptors (i.e., Npn-1).

Temporal, But Not Nasal,RGCs Are Collapsed bySema3F

Our expression data suggest that agradient of Sema3F/Npn-2 signalingis involved in retinocollicular map for-mation. If this is indeed the case, thenSema3F should affect temporal andnasal RGCs growth behavior differ-ently. We tested this hypothesis usinga collapse assay on cultures of immu-noisolated RGCs (Steinmetz et al.,2006) treated with Sema3F-contain-ing medium. Note that these cultureslack other retinal neurons or glialcells, enabling the study of the effectsof exogenous Sema3F without otherSema3 molecules, which could inter-fere with Npn-2 signaling (de Winteret al., 2004; and data not shown). Thiscannot be accomplished in retinal ex-plant cultures. To detect effects of theNpn-2 gradient, we isolated RGCsfrom the nasal and temporal side sep-arately. Presence of Sema3F in theconditioned medium was confirmed bymeasuring AP activity (see the Exper-imental Procedures section). After 2 to4 days in culture, P1 nasal and tem-poral RGCs were incubated for 1 hrwith control (mock-transfected HEKcells) or Sema3F-conditioned medium.These 2 to 4 days culture period isnecessary for the RGCs to regrow ax-ons and to form a growth cones. Time-lapse sequences show retraction andcollapse of RGC growth cones within18 min after Sema3F treatment (t �0) (Fig. 6A). Staining with fluoro-phore-conjugated phalloidin (phal-loidin-TRITC, Invitrogen) revealedgrowth cone morphology allowing usto score growth cones as collapsed ornoncollapsed. Criteria for collapsedgrowth cones were 2 or less filopodiaand the absence of lamellipodia (Fig.6B,C). Quantitative analyses of col-lapse in the presence of Sema3F con-taining medium normalized againstcollapse in supernatant from MOCK-

Fig. 5. In situ binding of Npn-2 ligand in the postnatal day (P) 1 superior colliculus (SC) and ofSema3F receptors in the P1 retina. A: Brightfield micrographs of sagittal sections of SC from P1mice after receptor affinity probe staining using Npn-2-alkaline phosphatase (AP). Arrowheadsindicate the Npn-2-AP staining intensity along the rostral–caudal axis. Lines mark the SO layer (R,rostral; C, caudal). B: Distribution of the Npn-2-AP staining intensity along the rostrocaudal axis ofthe SC in P1 mice. The lines represent counts from three different experiments. C: LacZ staining ofsagittal SC sections from P1 Isl2-�LacZ mice. Arrowheads indicate the ingrowing axons of ganglioncells in the retina (RGCs). Line separate the SZ, SGS/SuG, and SO layers. D: Control AP staining.E,F,H,I: Brightfield micrographs of retinal section showing ligand affinity probe staining usingSema3F-AP in P1 (E,F) and P8 (H,I) mice. G: Control AP staining. J: DAPI (4�,6-diamidine-2-phenylidole-dihydrochloride) staining of the section shown in I (NBL, neuroblast layer; IPL, innerplexiform layer; INL, inner nuclear layer; GCL, ganglion cell layer; ONF, optic nerve fiber; SZ,stratum zonale; SGS, stratum griseum superficiale; SO, stratum opticum). Scale bars � 1 mm inA,C,D, 100 �m in E,G,H, 15 �m in F,I,J.

3398 CLAUDEPIERRE ET AL.

transfected HEK cells supernatants(mean, 34 � 2.0%) reveal thatSema3F induces collapse in 4% ofnasal growth cones, whereas tempo-ral growth cones exhibit a threefoldhigher incidence of collapse at ap-proximately 12% (Fig. 6D). This dif-ference is statistically significant(Mann-Whitney test, Fig. 6D). Theoverall low collapse percentage isprobably not a consequence of a lowSema3F concentration as we used 3different batches of conditioned me-dium, each containing high levels ofSema3F-AP activity. The selectiveeffect of Sema3F on axons from tem-poral RGCs suggests that gradedNpn-2/Sema3F signaling is involvedin the proper targeting of RGCs ax-ons within the superior colliculus.

DISCUSSION

Our results indicate that bothSema3F and Npn-2 display graded ex-pression in the SC and the retina ofmice, respectively, when retinocollicu-lar targeting within the SC is under-way. In situ hybridization revealedgradients of Npn-2 in mouse RGCsalong the nasal–temporal retinal axis

and of Sema3F in the SGS/SuG layerof the SC along the rostral–caudalaxis. Npn-2 and its coreceptor PlxA1,are localized on RGC axons andgrowth cones. Receptor/ligand affinityprobe revealed selective binding ofNpn-2 and Sema3F to the SGS/SuGlayer in the SC and to the RGCs, re-spectively, in compliance with previ-ous data showing that Sema3F is ahigh affinity ligand for Npn-2 (Chen etal., 1997). Notably, our vitro studiesdemonstrate a stronger repulsive ef-fect of Sema3F on Npn-2-carrying ax-ons from temporal than from nasalRGCs. Altogether these data suggestthat Sema3F/Npn-2 may be involvedin retinocollicular map formationalong the rostral–caudal axis of theSC. The graded expression of bothNpn-2 in the retina and Sema3F inthe SC supports their role as posi-tional information molecules for topo-graphic map formation, in accordancewith Sperry’s chemoaffinity hypothe-sis (Sperry, 1963). Immunodetectionin vivo would not allow further gradi-ent identification at the protein levelas Npn-2 is localized on RGC axons.These axons are intermingled in thefiber layer and run toward the optic

disc in the center of the retina pre-venting any discrimination betweentemporal and nasal RGCs axons. Forthe same reason, indirect visualiza-tion of Npn-2 receptors in vivo usingSema3F-AP LAP did not reveal agraded labeling of the RGCs. How-ever, we showed in vitro that Npn-2protein is more strongly expressed intemporal compared with nasal RGCs.

Gradients of secreted molecules asindicated by in situ hybridization pro-vide a first estimate of the shape of thefunctional gradient. Further evidenceof the Sema3F protein gradient in theSC are provided by Npn-2-AP RAP ex-periments, which show a rostral–cau-dal gradient. This gradient was moreshallow compared to the gradient de-termined by in situ hybridization.This may be a consequence of baselinelevels of low affinity ligands likeSema3D competing with Sema3F andenhancing the overall Npn-2-AP sig-nal. Graded expression at the mRNAlevel is unusual for a secreted mole-cule, whose gradients are generallythought to be created by protein diffu-sion. In the present study, graded ex-pression at the transcript level may bedue to the large size of the SC (4–5mm at P1 along the rostral–caudalaxis), which may preclude coverage bydiffusion alone. Indeed, theoreticalstudies demonstrated that diffusionyields gradients that are inefficientwhen growth cones have to travel dis-tances exceeding 1 mm (Goodhill,1997; Goodhill and Baier, 1998). If so,the functional gradient of Sema3Fprotein, sensed by RGC growth cones,would be set at the mRNA level. Asimilar principle has been shown forthe membrane-bound Eph/ephrinmolecules whose functional gradientsoriginate at the transcriptional level(McLaughlin et al., 2003a).

Members of the Sema3 family,namely Sema3A and 3C, control lam-ina-specific connections of developingcortical neurons (Bagnard et al., 1998,2000; Polleux et al., 1998, 2000). TheSC is also a laminated structurewhere ingrowing RGC axons enterand connect to the superficial layersSGS/SuG following a lamina-specificarborization (Huerta et al., 1983;Yamagata et al., 2006). Diffusion ofSema3F proteins may create a shal-low dorsal-ventral gradient across theSGS/SuG laminae. The distance of mi-

Fig. 6. Selective effect of Sema3F on growth cones of temporal versus nasal ganglion cells in theretina (RGCs). A: Phase-contrast micrographs showing a time-lapse sequence of RGC growth coneretraction and collapse. At t � 0�, Sema3F was added to the medium. B,C: Fluorescence micro-graphs of RGCs stained with fluorescently labeled phalloidin indicating noncollapsed (B) andcollapsed (C) growth cones. D: Box-plot showing normalized percentage of collapsed growthcones from nasal and temporal RGCs in the presence of Sema3F (n � 6, Mann-Whitney test: **P �0.01). Scale bars � 5 �m.

NEUROPILIN 2/SEMAPHORIN 3F AND RETINOCOLLICULAR MAPS 3399

gration in this case ranges between200 and 400 �m, compatible with thetheoretical operating range of a diffus-ible gradient (less than 1 mm;Goodhill, 1997; Goodhill and Baier,1998; Yamagata et al., 2006). There-fore, complementary to the positionalinformation provided along the ros-tral–caudal axis, Sema3F may alsoparticipate in the control of the lami-na-specific arborization of RGCsand/or prevent RGC axons to enterinappropriate layers.

Our functional studies of primarycultures of immunoisolated RGCsshowed a threefold higher incidence ofgrowth cone collapse in temporal,compared to nasal RGCs in the pres-ence of Sema3F. The collapse-induc-ing effect of Sema3F is in accordancewith previous results demonstratingthat Sema3F triggers a repulsive ef-fect on Npn-2-carrying axons (Sahayet al., 2003; Atwal et al., 2003; Clou-tier et al., 2004; Pascual et al., 2005).We used cultures of immunopurifiedRGCs instead of retinal explant cul-tures as they allow an optimized de-tection of single axons and growthcones from isolated neurons and avoidany side effect of coexpressed Sema3molecules by other retinal cells thatmight interfere with exogenousSema3F (de Winter et al., 2004; andunpublished observation). The shortculture period (2 to 4 days) did proba-bly not affect the graded expression ofSema3 receptor complexes as indi-cated by RT-PCR data, but we cannotexclude that it diminished expressionlevels. The overall percentage of col-lapse is only approximately 12% fortemporal RGCs (compared with thecontrol collapse). One explanationwould be that EphA/ephrin-A signal-ing accounts for most of the map for-mation along the rostral–caudal axisof the SC (Reber et al., 2004; Lemkeand Reber, 2005). Indeed, we previ-ously demonstrate that EphA/eph-rin-A signalling accounts for 75% ofthe map along the rostral caudal axisof the SC (Reber et al., 2004). There-fore, other refinement mechanisms,such as correlated electrical activity(McLaughlin et al., 2003a; Shah andCrair, 2008) and axon targeting mole-cules must operate to provide the re-maining 25% of topographic precision.Our results suggest that Npn-2/Sema3F signaling may be part of

these refinement mechanisms by pro-viding a fraction of the remaining 25%of topographic precision, therefore, in-volving Npn-2 and Sema3F in retino-collicular map formation in mouse.

A role of Sema3 signaling in RGCmapping, in addition to Eph/ephrin, issuggested by other studies. Indeed, inXenopus, Sema3A shows a low-ante-rior to high-posterior gradient in thetectum (the nonmammalian SC) whenretinotectal mapping forms and RGCgrowth cones are responsive to tectalSema3A, exhibiting collapse, branch-ing and repulsive turning behavior. InXenopus, unlike rodents and chick,tectal cells proliferate throughout lifefollowing a rostral–caudal gradient ofdifferentiation; therefore, high-poste-rior expression of Sema3A seems toprevent RGC axons from projecting inthe immature tectum (Campbell et al.,2001). Numerous studies have shownsemaphorin expression in the Ze-brafish visual system (Yu et al., 2004;Liu et al., 2004; Sakai and Halloran,2006; Callander et al., 2007). Func-tional and expression analysis demon-strate that Sema3D and Npn-1/2 areexpressed in the Zebrafish tectum andretina, respectively. Sema3D guidesRGC axons in the contralateral optictract and participates to retinotectalmapping along the dorsoventral axisof the tectum presumably throughNpn-1 signaling (Liu et al., 2004). An-other recent study showed thatSema3Gb and Sema3Fb are expressedin the anterior tectum, whereasSema3Aa and 3E are found in the pos-terior tectum, reinforcing a putativerole of semaphorin signalling in theformation of Zebrafish retinotectalprojections by restricting RGC axonsto their target (Callander et al., 2007).In chick, Sema3D is expressed in ananterior–posterior increasing gradi-ent (Luo et al., 1995). Moreover,Sema3A and 3E show differential ex-pression in the chick tectum andSema3E collapses RGC axons,whereas Sema3A has no effect onRGCs collapse (Steinbach et al., 2002;Steffensky et al., 2006). Expressionstudies in rodents, which focusedmainly on semaphorin and neuropilinexpression in the retina during devel-opment and in the adult, showedhomogenous expression of Npn-2,Npn-1, Sema3A, 3B, 3C, and 3E (Gari-ano et al., 2006; de Winter et al.,

2004). Consistently, we showed a lowhomogenous Npn-1 expression inRGCs during the first postnatal week(de Winter et al., 2004). Unlike deWinter and collaborators, we showeda low-nasal to high-temporal Npn-2gradient in RGCs at P1 only. More-over, we demonstrated that Sema3F,a high-affinity ligand for Npn-2 (Chenet al., 1997) is expressed in a comple-mentary gradient (from low-rostral tohigh-caudal) in the SC at P1. Togetherwith our data, these results argue fora general role of Sema3 signaling inthe formation of the retinocollicular/retinotectal projections by restrictingRGCs axons to their target and/or giv-ing positional information to the pro-jecting RGC axons. Although gradedexpression of semaphorin 3 familymembers in the visual system seemsto be conserved through evolution,there may be species-specific differ-ences regarding ligand/receptor acti-vation. Indeed, Sema3A triggers re-pulsion of RGC axons in Xenopus,while it has no effect on rodent RGCs(Campbell et al., 2001; Goldberg et al.,2004). Similarly, Sema3E collapsesaxons of chick RGCs while Sema3Ahas no effect (Steinbach et al., 2002;Steffensky et al., 2006). In rodents,Sema3A, 3B and 3C have no effect onRGC axons (Goldberg et al., 2004),whereas we demonstrate that Sema3Frepels mouse RGCs. A role of Sema3F/Npn-2 in retinocollicular projectionformation would be confirmed by theanalysis of Sema3F and Npn-2 mousemutants (Giger et al., 2000; Sahay etal., 2003). However, the presence ofother Sema3 members in the SC andtheir redundancy may attenuate theretinocollicular phenotype in thosemutants.

EXPERIMENTALPROCEDURES

Tissue Processing/RGCCulture

C57/Bl6 mice were decapitated andretinae and/or colliculi were dissectedin phosphate buffered saline (PBS).For RGC culture and RT-PCR, retinaewere then cut in 3 equal parts com-prising the nasal, central and tempo-ral thirds. Culture of immunoisolatedmouse RGCs were prepared as de-scribed (Steinmetz et al., 2006).

3400 CLAUDEPIERRE ET AL.

Briefly, it consists in a two-step selec-tion method to obtain a highly pureneuronal preparation: the first stepremoves microglia using an anti-mac-rophage antibody (Sigma). Then, inthe second step, binding of RGCs tothe culture plate is achieved using anantibody raised against Thy1, a GPI-linked protein strongly expressed onthe cell surface of RGCs. Temporaland nasal retina were digested for 45min at 37°C in D-PBS (Gibco/Invitro-gen) containing 160 U ml�1 papain,200 U ml�1 DNAse, then mechani-cally triturated in D-PBS containing0.15% trypsin inhibitor (Roche Diag-nostics, Meylan, France), 650 U ml�1

DNAse and 1:75 rabbit anti-rat mac-rophage antibody (Sigma). The cell-suspensions were then incubated for30 min on subtraction plates (Ø150-mm Petri dishes, Falcon, BDBio-sciences/VWR, Fontenay sous Bois,France) coated with goat anti-rabbit-IgG (Jackson Immunoresearch Labo-ratories/Beckman Coulter, Marseille,France) to remove microglial cells andtransferred for 1 hr at room temper-ature on the selection plate (Ø100-mm Petri dish) precoated withgoat anti–rat-Ig mu chain (Jack-son Immunoresearch Laborato-ries) and 0.2 �g/ml mouse IgM anti-Thy1.2 (MCA01, Serotec, Cergy Saint-Christophe, France). Nonadherentcells were thoroughly washed off andadherent RGCs were released bytrypsination (12,000 U ml�1 in EBSSfor 10 min in 5% CO2 at 37°C). Afterinactivation of trypsin by 30% fetalcalf serum (Gibco/Invitrogen), RGCswere spun down and resuspended inchemically defined culture medium(see Steinmetz et al., 2006). Prepara-tions were highly pure containing98.2% of RGCs. All experiments wereperformed on 1- and 8-day-old mice,according to the protocol approved bythe Animal Care and Use Committeeof INSERM.

Semiquantitative RT-PCR

Total RNA was prepared from 2-day-old primary cultures of RGCs immu-noisolated from P1 and P8 mice usingRNeasy Mini Kit (Qiagen). RNA wasreverse-transcribed using SuperScriptIII First Strand (Invitrogen) accordingto the manufacturer’s protocol. PCRfragments were generated using PCR

SuperMix (Invitrogen) for 12 cycles(94°C, 30 sec, 57°C, 30 sec, 72°C, 1min; primer sequences available uponrequest). PCR fragments were hybrid-ized using 32P-labeled specific probesgenerated by random priming (Ran-dom Primed DNA labeling Kit, RocheApplied Science). Signals were quan-tified using PhosphorImager (Bio-Rad) and results were normalized toGAPDH transcripts.

LAP/RAP Assays, LacZStaining

For LAP experiments, retinae weredissected, washed in PBS/Hepes 10mM, blocked for 1 hr in PBS/10 mMHepes containing 20% fetal calf serumand incubated at room temperaturefor 4 hr with Sema3F-AP conditionedmedium from transiently transfectedHEK cells. After incubation, retinaewere washed in PBS/Hepes, fixed for 3min on ice in 60% acetone/3%PFA inPBS and washed in PBS/Hepes. Toinactivate endogenous phosphatases,retinae were heated at 65°C for 1 hr,then washed in PBS/Hepes before APrevelation using NBT/BCIP (Nitro-Blue Tetrazolium Chloride/ 5-Bromo-4-Chloro-3�-Indolyphosphate p-Tolu-idine) overnight. The following day,retinae were washed in PBS/ethyl-enediaminetetraacetic acid 5mM andstored in 2% formaldehyde at 4°C be-fore Vibratome sectioning.

RAP experiments on SC were per-formed on 100-�m Vibratome sec-tions. Before incubation with Npn-2–conditioned medium, colliculi werefixed in 2% formaldehyde for 10 minand then sectioned. AP staining wasperformed as described for retinae.See below for Sema3F-AP and Npn-2-AP containing medium.

Revelation of the LacZ activity wasperformed as described previously(Supp. Fig. 2 in Reber et al., 2004).

In Situ Hybridization/SignalMeasurement

In situ hybridizations were performedas described (Reber et al., 2004) ex-cept for SC hybridizations where 0.1%Triton-100X was added to the prehy-bridization washes. Probes were gen-erated by RT-PCR (primers sequencesavailable upon request) and clonedinto PGEM-T easy expression vector

(Promega). In vitro transcription wasperformed using the Maxiscript kit(Ambion) and cRNA probes were la-beled with digoxigenin (DIG)-UTP.Revelation of the AP activity was per-formed using DIG nucleic acid detec-tion kit according to manufacturer’sinstructions (Roche). In situ signalanalysis was performed as described(Reber et al., 2004). Briefly, digitalgray scale images (eight-bit) of in situhybridizations were acquired at1,300 � 1,300 pixels. The NT axis ofthe RGC layer of the retina was di-vided into 10 equal segments and sig-nal intensities for each segment werequantified with Scion Software (NIHImage version for PC). Positive in situhybridization gray scale signal wascounted after subtraction of the back-ground signal, which was determinedfrom control sense probe using “den-sity slicing”. Pixel counts are given inunits of “integrated signal density”(I.D.). The I.D. corresponds to the sumof gray scale values for all pixels abovethe lower gray detection threshold ina given segment (background sub-tracted). Within a segment, only pix-els with signal values above thisthreshold were scored.

Immunocytochemistry

RGCs cultures were processed for im-munocytochemical staining as de-scribed (Steinmetz et al., 2006) usinganti–Npn-2 antibodies (1/100, SantaCruz), anti-PlxA1 (1/500, gift from G.Roussel) and Alexa Fluo 488 labeledsecondary antibody (1/600, MolecularProbes).

Conditioned Medium andCollapse Assay

Growth cone collapse assays were per-formed based on previously describedtechniques (Campbell et al., 2001;Goldberg et al., 2004; Steinmetz et al.,2006). Transfections of human embry-onic kidney 293 cells (HEK-293) witheither Sema3F-AP or Npn-2-AP plas-mids were performed using lipo-fectamine 2000 (Invitrogen) as recom-mended by the manufacturer. Controlsupernatant corresponds to MOCKtransfected cells. Twenty-four hoursafter transfection, the medium was re-placed by fresh culture medium andconditioned medium was harvested

NEUROPILIN 2/SEMAPHORIN 3F AND RETINOCOLLICULAR MAPS 3401

after an additional 48 hr. The pres-ence of Sema3F-AP and Npn-2-AP inthe media was estimated based on APactivity measured by chemolumines-cence (not shown).

For the collapse assay, half of theRGC culture medium was replaced bySema-3F conditioned medium. After 1hr, RGCs were fixed by adding 4%formaldehyde directly in the culturemedium. After 5 min, the medium wasremoved and cells were incubated foradditional 10 min in 4% formalde-hyde. RGCs were permeabilized inPBS-BSA 3% - Triton X-100 0.1% for 5min and after PBS washing stepsstained with phalloidin-TRITC (In-vitrogen) overnight at 4°C. We per-formed 6 independent experiments inwhich a minimum of 250 growth conesper condition were analyzed. Col-lapsed growth cones were counted bytwo different experimenters blind tothe experimental condition. A col-lapsed growth cone was defined aspresenting 2 or less filopodia and theabsence of lamellipodia (Campbel etal., 2001; Goldberg et al., 2004). Per-centage of collapse is the ratio be-tween RGCs collapse in Sema3F-treated culture and RGC collapse incorresponding control (MOCK-trans-fected) treatment.

ACKNOWLEDGMENTSWe thank M. Lindberg for excellenttechnical assistance, Dr. A. Kolodkinfor Npn-2 fusion construct, Dr. A. Che-dotal for Sema3F fusion construct andDr. G. Roussel for the anti-PlxA1 an-tibody. We also thank Dr. S. Reibel-Foisset and N. Lethenet for animalcare and mouse facility maintenance.T.C. is supported by Retina France,E.K. is supported by French Ministryof Research. Project is funded by IN-SERM (M.R., D.B.) and UniversityLouis Pasteur (M.R.).

REFERENCES

Atwal JK, Singh KK, Tessier-Lavigne M,Miller FD, Kaplan DR. 2003. Sema-phorin 3F antagonizes neurotrophin-in-duced phosphatidylinositol 3-kinase andmitogen-activated protein kinase kinasesignalling: a mechanism for growth conecollapse. J Neurosci 23:7602–7609.

Bagnard D, Lohrum M, Uziel D, PuschelAW, Bolz J. 1998. Semaphorins act as at-tractive and repulsive guidance signalsduring the development of cortical projec-tions. Development 125:5043–5053.

Bagnard D, Thomasset N, Lohrum M, Pus-chel AW, Bolz J. 2000. Spatial distribu-tions of guidance molecules regulate che-morepulsion and chemoattraction ofgrowth cones. J Neurosci 20:1030–1035.

Barres BA, Silverstein BE, Corey DP,Chun LL. 1988. Immunological, morpho-logical, and electrophysiological varia-tion among retinal ganglion cells puri-fied by panning. Neuron 9:791–803.

Brown A, Yates PA, Burrola P, Ortuno D,Vaidya A, Jessell TM, Pfaff SL, O’LearyDD, Lemke G. 2000. Topographic map-ping from the retina to the midbrain iscontrolled by relative but not absolutelevels of EphA receptor signalling. Cell102:77–88.

Buhusi M, Schlatter MC, Demyanenko GP,Thresher R, Maness PF. 2008. L1 inter-action with ankyrin regulates mediolat-eral topography in the retinocollicularprojection. J Neurosci 28:177–188.

Callander DC, Lamont RE, Childs SJ, Mc-Farlane S. 2007. Expression of multipleclass three semaphorins in the retinaand along the path of zebrafish retinalaxons. Dev Dyn 236:2918–2924.

Campbell DS, Regan AG, Lopez JS, Tan-nahill D, Harris WA, Holt CE. 2001.Semaphorin 3A elicits stage-dependentcollapse, turning, and branching in Xe-nopus retinal growth cones. J Neurosci21:8538–8547.

Chen H, Chedotal A, He Z, Goodman CS,Tessier-Lavigne M. 1997. Neuropilin-2, anovel member of the neuropilin family, isa high affinity receptor for the Sema-phorins Sema E and Sema IV but notSema III. Neuron 19:547–559.

Cloutier JF, Sahay A, Chang EC, Tessier-Lavigne M, Dulac C, Kolodkin AL, GintyDD. 2004. Differential requirements forSemaphorin 3F and Slit-1 in axonal tar-geting, fasciculation, and segregation ofolfactory sensory neuron projections.J Neurosci 24:9087–9096.

de Winter F, Cui Q, Symons N, VerhaagenJ, Harvey AR. 2004. Expression ofclass-3 Semaphorins and their receptorsin the neonatal and adult rat retina. In-vest Ophthalmol Vis Sci 45:4554–4562.

de Wit J, Verhaagen J. 2003. Role of Sema-phorins in the adult nervous system.Prog Neurobiol 71:249–267.

Erskine L, Herrera E. 2007. The retinalganglion cell axon’s journey: insightsinto molecular mechanisms of axon guid-ance. Dev Biol 308:1–14.

Feldheim DA, Kim YI, Bergemann AD,Frisen J, Barbacid M, Flanagan JG.2000. Genetic analysis of ephrin-A2 andephrin-A5 shows their requirement inmultiple aspects of retinocollicular map-ping. Neuron 25:563–574.

Feldheim DA, Nakamoto M, Osterfield M,Gale NW, DeChiara TM, Rohatgi R, Yan-copoulos GD, Flanagan JG. 2004. Loss-of-function analysis of EphA receptors inretinotectal mapping. J Neurosci 24:2542–2550.

Flanagan JG, Vanderhaeghen P. 1998. Theephrins and Eph receptors in neural de-velopment. Annu Rev Neurosci 21:309–345.

Frisen J, Yates PA, McLaughlin T, Fried-man GC, O’Leary DD, Barbacid M. 1998.Ephrin-A5 (AL-1/RAGS) is essential forproper retinal axon guidance and topo-graphic mapping in the mammalian vi-sual system. Neuron 20:235–243.

Gariano RF, Hu D, Helms J. 2006. Expres-sion of angiogenesis-related genes dur-ing retinal development. Gene Expr Pat-terns 6:187–192.

Giger RJ, Cloutier JF, Sahay A, PrinjhaRK, Levengood DV, Moore SE, PickeringS, Simmons D, Rastan S, Walsh FS,Kolodkin AL, Ginty DD, Geppert M.2000. Neuropilin-2 is required in vivo forselective axon guidance responses to se-creted Semaphorins. Neuron 25:29–41.

Goldberg JL, Vargas ME, Wang JT, Man-demakers W, Oster SF, Sretavan DW,Barres BA. 2004. An oligodendrocyte lin-eage-specific Semaphorin, Sema5A, in-hibits axon growth by retinal ganglioncells. J Neurosci 24:4989–4999.

Goodhill GJ. 1997. Diffusion in axon guid-ance. Eur J Neurosci 9:1414–1421.

Goodhill GJ, Baier H. 1998. Axon guid-ance: stretching gradients to the limit.Neural Comput 10:521–527.

Hindges R, McLaughlin T, Genoud N, Hen-kemeyer M, O’Leary DD. 2002. EphBforward signalling controls directionalbranch extension and arborization re-quired for dorsal-ventral retinotopicmapping. Neuron 35:475–487.

Huber AB, Kolodkin AL, Ginty DD, Clou-tier JF. 2003. signalling at the growthcone: ligand-receptor complexes and thecontrol of axon growth and guidance.Annu Rev Neurosci 26:509–563.

Huerta MF, Frankfurter A, Harting JK.1983. Studies of the principal sensoryand spinal trigeminal nuclei of the rat:projections to the superior colliculus, in-ferior olive, and cerebellum. J CompNeurol 220:147–167.

Lemke G, Reber M. 2005. Retinotectalmapping: new insights from moleculargenetics. Annu Rev Cell Dev Biol21:551–580.

Liu Y, Berndt J, Su F, Tawarayama H,Shoji W, Kuwada JY, Halloran MC.2004. Semaphorin3D guides retinal ax-ons along the dorsoventral axis of thetectum. J Neurosci 24:310–318.

Luo Y, Shepherd I, Li J, Renzi MJ, ChangS, Raper JA. 1995. A family of moleculesrelated to collapsin in the embryonicchick nervous system. Neuron 14:1131–1140.

Matsunaga E, Nakamura H, Chedotal A.2006. Repulsive guidance molecule playsmultiple roles in neuronal differentia-tion and axon guidance. J Neurosci 26:6082–6088.

May P. 2006. The mammalian superior col-liculus: laminar structure and connec-tions. Prog Brain Res 151:321–378.

McLaughlin T, O’Leary DD. 2005. Molecu-lar gradients and development of retino-topic maps. Annu Rev Neurosci 28:327–355.

McLaughlin T, Hindges R, O’Leary DD.2003a. Regulation of axial patterning ofthe retina and its topographic mapping

3402 CLAUDEPIERRE ET AL.

in the brain. Curr Opin Neurobiol 13:57–69.

McLaughlin T, Hindges R, Yates PA,O’Leary DD. 2003b. Bifunctional actionof ephrin-B1 as a repellent and attract-ant to control bidirectional branch exten-sion in dorsal-ventral retinotopic map-ping. Development 130:2407–2418.

Meyer-Franke A, Kaplan MR, Pfrieger FW,Barres BA. 1995. Characterization of thesignalling interactions that promote thesurvival and growth of developing reti-nal ganglion cells in culture. Neuron 15:805–819.

Murakami Y, Suto F, Shimizu M, ShinodaT, Kameyama T, Fujisawa H. 2001. Dif-ferential expression of plexin-A subfam-ily members in the mouse nervous sys-tem. Dev Dyn 220:246–258.

Niederkofler V, Salie R, Sigrist M, Arber S.2004. Repulsive guidance molecule(RGM) gene function is required for neu-ral tube closure but not retinal topogra-phy in the mouse visual system. J Neu-rosci 24:808–818.

Pascual M, Pozas E, Soriano E. 2005. Role ofclass 3 Semaphorins in the developmentand maturation of the septohippocampalpathway. Hippocampus 15:184–202.

Pasterkamp RJ, Kolodkin AL. 2003. Sema-phorin junction: making tracks towardneural connectivity. Curr Opin Neuro-biol 13:79–89.

Polleux F, Giger RJ, Ginty DD, KolodkinAL, Ghosh A. 1998. Patterning of corticalefferent projections by Semaphorin-neu-ropilin interactions. Science 282:1904–1906.

Polleux F, Morrow T, Ghosh A. 2000.Semaphorin 3A is a chemoattractantfor cortical apical dendrites. Nature 404:567–73.

Reber M, Burrola P, Lemke G. 2004. Arelative signalling model for the forma-tion of a topographic neural map. Nature431:847–853.

Sahay A, Molliver ME, Ginty DD, KolodkinAL. 2003. Semaphorin 3F is critical fordevelopment of limbic system circuitryand is required in neurons for selectiveCNS axon guidance events. J Neurosci23:6671–6680.

Sakai JA, Halloran MC. 2006. Semaphorin3d guides laterality of retinal ganglioncell projections in zebrafish. Develop-ment 133:1035–1044.

Schmitt AM, Shi J, Wolf AM, Lu CC, KingLA, Zou Y. 2006. Wnt-Ryk signalling me-diates medial-lateral retinotectal topo-graphic mapping. Nature 439:31–37.

Shah RD, Crair MC. 2008. Retinocollicularsynapse maturation and plasticity areregulated by correlated retinal waves.J Neurosci 28:292–303.

Sperry RW. 1963. Chemoaffinity in the or-derly growth of nerve fiber patterns and

connections. Proc Natl Acad Sci U S A50:703–710.

Steffensky M, Steinbach K, Schwarz U,Schlosshauer B. 2006. Differential im-pact of Semaphorin 3E and 3A on CNSaxons. Int J Dev Neurosci 24:65–72.

Steinbach K, Volkmer H, Schlosshauer B.2002. Semaphorin 3E/collapsin-5 inhib-its growing retinal axons. Exp Cell Res279:52–61.

Steinmetz CC, Buard I, Claudepierre T,Nagler K, Pfrieger FW. 2006. Regionalvariations in the glial influence on syn-apse development in the mouse CNS.J Physiol 15:249–261.

Takahashi T, Strittmatter SM. 2001.Plexina1 autoinhibition by the plexinsema domain. Neuron 29:429–439.

Thaler JP, Koo SJ, Kania A, Lettieri K,Andrews S, Cox C, Jessell TM, Pfaff SL.2004. A postmitotic role for Isl-class LIMhomeodomain proteins in the assign-ment of visceral spinal motor neuronidentity. Neuron 41:337–350.

Yamagata M, Weiner JA, Dulac C, RothKA, Sanes JR. 2006. Labelled lines in theretinotectal system: markers for reti-norecipient sublaminae and the retinalganglion cell subsets that innervatethem. Mol Cell Neurosci 33:296–310.

Yu HH, Houart C, Moens CB. 2004. Clon-ing and embryonic expression of ze-brafish neuropilin genes. Gene Expr Pat-terns 4:371–378.

NEUROPILIN 2/SEMAPHORIN 3F AND RETINOCOLLICULAR MAPS 3403