Neuroligin 3 Is a VertebrateGliotactin Expressed in the Olfactory

Ensheathing Glia, a Growth-Promoting Class of Macroglia

MARY GILBERT,1 JEFF SMITH,1 ANGELA-JANE ROSKAMS,2AND VANESSA J. AULD1*

1Department of Zoology, University of British Columbia, Vancouver, Canada2Center of Medical and Molecular Therapeutics, University of British Columbia,

Vancouver, Canada

KEY WORDS olfactory ensheathing glia; astrocytes; spinal cord; retina; Schwanncells; development; neuroligin; gliotactin

ABSTRACT The molecular mechanisms that drive glia–glial interactions and glia–neuronal interactions during the development of the nervous system are poorly under-stood. A number of membrane-bound cell adhesion molecules have been shown to playa role, although the precise nature of their involvement is unknown. One class ofmolecules with cell adhesive properties used in the nervous system is the serine-esterase-like family of transmembrane proteins. A member of this class, a glia-specificprotein called gliotactin, has been shown to be necessary for the development of the glialsheath in the peripheral nervous system of Drosophila melanogaster. Gliotactin isessential for the development of septate junctions in the glial sheath of individual andneighboring glia. Mutations that remove this protein result in paralysis and eventuallydeath due to a breakdown in the glial-based blood–nerve barrier. To study the role ofgliotactin during vertebrate nervous system development, we have isolated a potentialvertebrate gliotactin homologue from mice and rat and found that it corresponds toneuroligin 3. Using a combination of RT-PCR and immunohistochemistry, we havefound that neuroligin 3 is expressed during the development of the nervous system inmany classes of glia. In particular neuroligin 3 is expressed in the olfactory ensheathingglia, retinal astrocytes, Schwann cells, and spinal cord astrocytes in the developingembryo. This expression is developmentally controlled such that in postnatal and adultstages, neuroligin 3 continues to be expressed at high levels in the olfactory ensheathingglia, a highly plastic class of glia that retain many of their developmental characteristicsthroughout life. GLIA 34:151–164, 2001. © 2001 Wiley-Liss, Inc.

INTRODUCTION

The molecular mechanism that triggers and main-tains the glia ensheathed around axons is poorly un-derstood. Yet this is a crucial step in normal develop-ment and regeneration after injury or disease. In thedeveloping PNS of Drosophila melanogaster, the glialprotein gliotactin is essential for the formation of theglia sheath (Auld et al., 1995). Drosophila gliotactin isfirst expressed in peripheral glia as they migrate fromthe border of the CNS along the axons of the developing

PNS and expression peaks as the peripheral glia wraptheir associated nerve bundles. In Drosophila, as wellas other invertebrates and primitive vertebrates, glia

Grant sponsor: Howard Hughes Medical Institute; Grant sponsor: Rick Han-sen Neurotrauma Foundation; Grant sponsor: British Columbia Health Re-search Foundation; Grant sponsor: CCMT/Merck Frosst.

*Correspondence to: Vanessa J. Auld, Department of Zoology, University ofBritish Columbia, 6270 University Blvd., Vancouver, BC, Canada V6T 1Z4.E-mail: auld@zoology.ubc.ca

Received 30 November 2000; Accepted 6 March 2001

GLIA 34:151–164 (2001)

© 2001 Wiley-Liss, Inc.

Figure 1.

152 GILBERT ET AL.

are responsible for the formation of the blood–brainand blood–nerve barriers. Gliotactin plays an essentialrole in the establishment of the blood–nerve barrier inDrosophila as mutations that remove the gliotactinprotein disrupt glial wrapping and result in a break-down of the blood–nerve barrier and paralysis (Auld etal., 1995). Drosophila gliotactin is also expressed in awide range of glia in the adult including the glia of thewing and leg, the retina and the olfactory system. Theperipheral glia of Drosophila share many morphologi-cal characteristics with vertebrate nonmyelinatingSchwann cells and olfactory ensheathing glia (Sepp etal., 2000). In addition the glia in Drosophila also ex-press a number of transmembrane proteins conservedwith those found in vertebrate glia.

In order to study the formation and consolidation of theglial sheath during development and regeneration in ver-tebrates, the gliotactin homologue was investigated inboth mice and rats. For these studies, we focused on glialcells during the late embryonic and early postnatal stagesof development (before and at the start of myelination), aswell as in adults. In particular we concentrated on theolfactory system given the role of the olfactory ensheath-ing glia which wrap, guide and protect the olfactory neu-rons throughout out the life of the animal (Ramon-Cuetoand Avila, 1998). Olfactory ensheathing glia exhibit char-acteristics found in both Schwann cells and astrocytes.Like Schwann cells, the olfactory ensheathing glia arecapable of promoting continued regeneration in the olfac-tory bulb. Transplanted olfactory ensheathing glia havealso been shown to promote neuronal regeneration andsome functional recovery in the PNS and CNS (Li et al.,1997, 1998; Navarro et al., 1999; Ramon-Cueto et al.,2000). These glia continue to express throughout life themolecules necessary to promote continued axonal elonga-tion such as the cell adhesion proteins L1/Ng-CAM andNCAM (Doucette, 1990). The homologues of both theseproteins are expressed in Drosophila glia and we proposethat the vertebrate gliotactin protein may represent an-other cell adhesion protein expressed in the olfactoryensheathing glia during development.

The first step in determining the role of vertebrategliotactin during olfactory glia development was to isolateand characterize the vertebrate gliotactin homologue.Our investigations show that neuroligin 3 is a vertebratehomologue of gliotactin. Even though the neuroliginswere initially described as CNS specific neuronal pro-teins, we show that in the olfactory epithelium, neuroli-gin 3 is exclusively expressed in the ensheathing glia. Insupport of these observations, we also find neuroligin 3expression during both PNS and CNS development in awide range of glia, which include Schwann cells, retinalastrocytes, and spinal cord astrocytes.

MATERIALS AND METHODSGliotactin/Neuroligin RT-PCR and Cloning

Total RNA was isolated from P3 Sprague-Dawley ratbrain, spinal cord, dorsal root ganglia (DRG), and cul-tured Schwann cells or CD1 mouse olfactory epithe-lium and olfactory bulb (Chomczynski and Sacchi,1987). Contaminating DNA was removed from all sam-ples by treatment with DNase I (Pharmacia). All ex-periments were accompanied by negative controls forreverse transcription that included 5 mg RNase A.

For the amplification of gliotactin homologues, a de-generate primer based on the gliotactin serine-esterasedomain (MG4) (50 pmole) was used to prime cDNAsynthesis (from 2 mg of total RNA) using Superscript IIreverse transcriptase (Gibco-BRL)-polymerase chainreaction (RT-PCR) to amplify the highly conserved re-gion of gliotactin/neuroligins was performed on 1/10 volof each RT reaction following the manufacturer’s(Pharmacia) protocols with 2.5 mM MgCl2, 1 pmoleMG1 degenerate primer, 1 pmole MG4 degenerateprimer. PCR conditions used previously to successfullyisolated mammalian homologues of Drosophila geneswere followed (Kolodkin et al., 1993). PCR productswere subcloned (pGEM-T, Promega) and all clones se-quenced (Amersham-USB).

For the amplification of neuroligin-specific products,50 pmole oligo-dT12–18 (Pharmacia) was used in first-strand cDNA synthesis. Primer pairs specific to eachneuroligin were designed using the unique sequencesfound in the C-terminal region and 39UTR of each gene.Neuroligin-specific PCR was performed on 1/10 vol ofeach RT reaction using the manufacturer’s (Pharma-cia) suggested conditions, 1.5 mM MgCl2, and 2 pmoleof each primer pair. Primers for the amplification of thecyclophilin gene were included to control for the levelsof RNA in each reaction. As a negative control, eachcDNA was replaced with an equal volume of H2O.Thermocylcing conditions were 70°C for 5 min, fol-lowed by 32 cycles at 94°C for 1 min, 60°C for 1 min,and 72°C for 1 min, then 72°C for 5 min. All neuroligin-specific PCR products were purified, subcloned intopGEM-T vector (Promega), and sequenced (Amersham-USB) to confirm the sequence identity.

Sequences and regions spanned by the primers usedabove are as follows: MG1 59-CCATCGATGATGGAY-

Fig. 1. Gliotactins and neuroligins form a distinct subfamily withinthe serine esterase family of proteins. A: The gliotactin amino acidsequence was aligned to other members of the subfamily of serine-esterase-like proteins. Only the region spanning the first and secondloops disulfide bonds the active site serine (which is converted to G orD in this subfamily) is shown as a result of space constraints. Aminoacid identities are boxed; similarities are in gray. (rNLIG3, rat neu-roligin 3; hHNL3l, human neuroligin 3 long; hHNL3s, human neu-roligin 3 short; hNLIG1, human neuroligin 1; rNLG2, rat neuroligin 2;rNLG1, rat neuroligin 1; dGliotactin, D. melanogaster gliotactin;ceF55D10.3, C. elegans, gliotactin; dNL, D. melanogaster neuroliginCG13772; ceC40C9.5, C. elegans neuroligin; dGlutactin, D. melano-gaster glutactin; dNeurotactin, D. melanogaster neurotactin). B: Evo-lutionary tree built from the cladistic analysis of the gliotactins andrepresentatives of other serine esterase family members. Thickbranch lines indicate the gliotactin subfamily. Gliotactins and neu-roligins from insects, metazoans, and vertebrates are more closelyrelated because only one branch point separates gliotactins fromneuroligins. In contrast, gliotactins and neuroligins are much lessclosely related to other esterases from the same species. The firstletter of the sequence name identifies the organism from which itoriginates: d, Drosophila melanogaster; ce, Caenorhabditis elegans; r,rat; m, mouse; h, human.

153VERTEBRATE GLIOTACTIN EXPRESSION

Fig. 2. Neuroligins are expressed in the peripheral, olfactory, andcentral nervous systems. Reverse transcription-polymerase chain re-action (RT-PCR) analysis was used to determine the distribution ofeach neuroligin in various parts of the nervous system and in culturedSchwann cells. All products were subcloned and sequenced to verifyidentity. A: Neuroligins are differentially expressed in different re-gions of the nervous system. RT-PCR using specific primers was usedto detected expression of neuroligin 1 (510 bp, lane 1), neuroligin 2(595 bp, lane 2), neuroligin 3 (564 bp, lane 3), or cyclophilin (300 bp,lane C) in RNA isolated from rat brain (brain), rat spinal cord (SpC),rat dorsal root ganglia (DRG), mouse olfactory bulb (OB), and mouse

olfactory epithelium (OE). Products of the expected size were detectedwhen compared with the 100-bp ladder (BRL). Cyclophilin (C) wasused to control for RNA levels. Lane R contains the negative controlfor reverse transcription; lane P contains the negative control forPCR. B: Neuroligins 2 and 3 are expressed in Schwann cells. RT-PCRusing specific primers was used to detected expression of neuroligin 1(1), neuroligin 2 (2), neuroligin 3 (3), or cyclophilin (C) in RNA isolatedfrom rat brain (B) or cultured Schwann cells (S). Products of theexpected size were detected when compared with the 100-bp ladder(BRL). Cyclophilin (C) was used to control for RNA levels. Lane Rcontains the negative control for reverse transcription.

154 GILBERT ET AL.

GARGAYTGYYTITA-39 MG4 59-CCATCGATGRTCIG-GCCARAAIGTRTACAT-39 which amplify a 629-bp frag-ment corresponding the conserved serine esterasedomain of gliotactin; NL1 sense 59-AATTCCACTCCAGT-CACATCAGC-39 and NL1 antisense 59-CAGTGTATTAT-TCTGTCCTCCAGT-39 which amplify a 510-bp fragmentcorresponding to nucleotides 2825-3335 at the 39 end ofthe rat neuroligin 1 cDNA; NL2 sense 59-CACGCTGGC-CACCTCGCACACC-39 and NL2 antisense 59-GTTGTT-GTGGCTGGTAGCAGTAGG-39 which amplify a 595-bpfragment corresponding to nucleotides 2430-3025 at the39 end of the rat neuroligin 2 cDNA; NL3 sense 59-AAT-GAGAATGCTCCTGGGTCCTGG-39 and NL3 antisense59-GTTTGTGGATCTGGGAGGGAGG-39 which amplify564-bp fragment corresponding to nucleotides 2339-2903at the 39 end of the rat neuroligin 3 cDNA; Cyclophilinsense 59- TGGTCAACCCCACCGTGTTCTT-39 and anti-sense 59- CCATCCAGCCACTCAGTCTTG -39 which am-plifies a 300-bp fragment.

Sequence Alignment and Evolutionary Analysis

Sequences were aligned first using Clustal V (Hig-gins and Sharp, 1988). Evolutionary analysis was per-formed on truncated sequences using PAUP 3.1 (Si-nauer) specifically, heuristic maximum parsimony withbootstrap analysis. Sequences used were: Drosophilaacetylcholinesterase (accession 113036), Drosophilaglutactin (accession 462182), Drosophila neurotactin(accession 128570), Drosophila gliotactin (accessionL39083), Drosophila “neuroligin” CG13772 (accessionAF251479), Drosophila EST6 (accession U23761), Cae-norhabditis elegans F55D10.3 (accession U40948), C.elegans C40C9.5 (accession CAA94208), C. elegansest-1 (accession Q04457), C. elegans ace-1 (accessionX75331), rat acetylcholinesterase (accession 584716),rat carboxylesterase (accession CAA55241), rat neu-roligin 1 (accession U22952), rat neuroligin 2 (acces-sion U41662), rat neuroligin 3 (accession U41663), hu-man cholinesterase (accession 4557351), humanneuroligin 1 (accession AI056353), human neuroligin3s (accession AAF71231), human neuroligin 3l (acces-sion NP061850), human butylcholinesterase (accessionM16541), human carboxylesterase (accession I61085).

Culturing of Primary Schwann Cells andOlfactory Ensheathing Glia

The isolation of pure Schwann cell cultures was car-ried out essentially as described (Brockes et al., 1979).These cultures were determined to be free of neuronalcontamination using antibodies to neuron-specific eno-lase (NSE) (Sigma) and RT-PCR with NSE primers.Olfactory ensheathing glia were cultured essentially asdescribed (Au and Roskams, 2001).

Immunohistochemistry

Frozen tissue sections from E17, P1, P5, or adultmice were generated as described previously (Roskamset al., 1996, 1998). Primary antibodies used in thisstudy were olfactory marker protein polyclonal diluted1:5,000 (a gift of F.L. Margolis), glial fibrillary acidicprotein (GFAP) polyclonal diluted 1:5 (Incstar Prod-ucts), neural-specific tubulin bIII monoclonal diluted1:500 (BAbCO), NCAM polyclonal diluted 1:500(Chemicon), neuroligin 1/3 monoclonal diluted 1:250(Synaptic Systems), S-100b monoclonal diluted 1:2,500(Sigma). For immunofluorescence studies, fluores-cently labeled secondary antibodies used in this studywere: donkey anti-rabbit Cy2 F(ab9)2 fragments, rabbitanti-mouse Texas Red, rabbit anti-mouse Cy2, rabbitanti-goat Texas Red (Jackson ImmunoResearch Labo-ratories), and goat anti-mouse Alexa 546, goat anti-Mouse Alexa 488 (Clontech). All sections were mountedin Vectashield. For immunostaining, sections were in-cubated in biotinylated anti-rabbit or anti-mouse sec-ondary antibodies (Vector Laboratories) diluted 1:200in conjunction with the Vectastain Elite B Kit (VectorLaboratories) protocol. For using mouse monoclonalswith mouse tissues, a MOM blocking solution was used(M.O.M kit, Vector Laboratories).

Imaging Using Confocal and StandardImmunofluorescence Microscopy

Some tissue sections were examined using a ZeissAxioskop 2 MOT microscope. Images were capturedwith a SPOT digital camera (Diagnostic Instruments)and Northern Exposure software on a Pentium II PC.Some tissue sections were imaged using a Zeiss Axio-vert S-100 TV microscope fitted with Bio-Rad RadiancePlus confocal hardware and LaserSharp software run-ning on a Dell Pentium II PC. Confocal Z-series wereprocessed using NIH Image software version 1.62(Wayne Rasband, National Institutes of Health) andimported into Adobe Photoshop 5.5 for colorization anddetermination of signal co-localization.

RESULTSGliotactin and Neuroligins Belong to a

Subfamily of Serine-Esterase-like Proteins

In order to investigate the role of the gliotactinprotein in glial sheath formation during vertebratedevelopment, the gliotactin homologue was isolatedusing RT-PCR and degenerate primers. Gliotactin,which is a member of the serine-esterase-like family,has the greatest similarity to the vertebrate neuroli-gin family by both sequence and predicted structure(Botti et al., 1998). A comparative examination of theamino acid sequences of gliotactin and the vertebrateneuroligins indicated that all three subtypes of neu-roligin had a high degree of identity (20 –21%) with

155VERTEBRATE GLIOTACTIN EXPRESSION

Figure 3

Fig. 4. Cultured olfactory ensheathingglia express neuroligin 3. Cultured olfac-tory ensheathing glia were immunostainedwith glial fibrillary acidic acid (GFAP)(green, A,D) and neuroligin 1/3 (red, B,E)antibodies. C,F: Overlap of GFAP and neu-roligin 3 staining as yellow. All OEG invitro express S-100b, while only a subsetexpress GFAP (Franceschini and Barnett,1996). Neuroligin 3 are expressed in bothsubsets of OEG in vitro. The cell bodies(double arrows) and the projections (ar-rows) coexpress both proteins with a tracesurrounding the nucleus (arrowheads).Neuroligin 3 is most concentrated in thecell surface overlying the intermediate fil-aments (C,F).

156 GILBERT ET AL.

gliotactin over the entire length of the protein (Fig.1A). To further analyze this similarity, cladisticanalysis was used to examine the evolutionary rela-tionship between gliotactin, the neuroligins, andother members of the serine esterase family (Fig.1B). The results from heuristic maximum parsimonyanalysis show that Drosophila gliotactin; DrosophilaCG13772, also referred to as Drosophila neuroligin(G. Boulianne; personnel communication); and thevertebrate neuroligins form a distinct subfamilywithin the serine-esterase family. This analysis alsoidentified two new members of this family from theC. elegans sequence database (F55D10.3 andC40C9.5), the former being more related to gliotactinand the later more closely related to CG13772. Fromour analysis, the Drosophila CG13772 and C. elegansC40C9.5 genes are ancestral to both the gliotactinand vertebrate neuroligin genes. Other Drosophilaproteins were predicted from the Drosophila genometo be similar to neuroligins, CG5030 for instance(Adams et al., 2000), but these encode small, non-transmembrane predicted proteins with the homol-ogy localized to small regions rather than over thelength of the protein. The high degree of similaritybetween the gliotactins and neuroligins and theirvery close evolutionary relationship suggests thatthese proteins may perform very similar functions intheir respective organisms and cell types.

Neuroligin 3 Is a Potential Homologue ofDrosophila Gliotactin

Based on the sequence and predicted structural sim-ilarities, our strategy to isolate the vertebrate homo-logue of gliotactin involved identification of highly con-served regions of similarity among the members of thisfamily from Drosophila, Caenorhabitis elegans, rat,and human. Potential vertebrate gliotactin clones weregenerated using RT-PCR and degenerate oligonucleo-tides based on the conserved sequences. This strategywas intentionally unbiased so that all candidate glio-tactin orthologues would be isolated. To select for bothCNS and PNS sources, clones were generated from ratbrain, spinal cord, and dorsal root ganglion (DRG)RNA. Sequence analysis of the generated clones re-vealed that all the clones isolated from the DRG wereidentical to neuroligin 3. This was an unexpected resultgiven that all three neuroligins were believed to beexpressed only in CNS neurons (Ichtchenko et al.,1995; Ichtchenko et al., 1996). Moreover, all clonesisolated corresponded to two of the three possible neu-roligin 3 splice variants; the variant that has an exonconserved with neuroligin 1 was absent (data notshown).

Further PCR analysis was carried out to confirm theexpression of neuroligin 3 in the DRG and to determinewhich cell types express this gene. Oligonucleotideprimers were designed to the C-terminal region that isunique to each neuroligin. These neuroligin-specificprimers were used for RT-PCR assays of RNA from ratbrain, spinal cord (SpC), and dorsal root ganglia (DRG),mouse olfactory bulb (OB), and epithelium (OE) (Fig.2A). As expected all three neuroligin genes are ex-pressed in RNA isolated from brain, spinal cord, andolfactory bulb. This result also confirms the ability ofthe primers to detect neuroligin expression in both ratand mouse tissues. Only neuroligins 2 and 3 are ex-pressed in RNA derived from the DRG. Furthermore,only neuroligin 3 is expressed in olfactory epitheliumRNA. Neuroligin 1 appears to be expressed exclusivelyin the CNS. To test whether peripheral glia expressneuroligins, RT-PCR using RNA isolated from purecultures of Schwann cells was carried out (Fig. 2B).Our results show that both neuroligins 2 and 3 areexpressed in Schwann cells in vitro, further suggestingthat the neuroligin family is not exclusively expressedin neurons.

Neuroligin 3 Is Expressed in the OlfactoryEnsheathing Glia But Not in the

Olfactory Neurons

Next, the distribution of the neuroligin 3 protein wasdetermined to confirm the PCR results and to confirmthat neuroligin 3 is expressed in glial cells. If mamma-lian neuroligins perform a similar function as Drosoph-ila gliotactin, then one or more of them should bedevelopmentally expressed during glial sheath forma-

Fig. 3. Neuroligin 3 is expressed in the olfactory ensheathing glia(OEG). Sections from E17 mice were stained with the neuroligin 1/3antibody (green), glial fibrillary acidic protein (GFAP) antibody (red;A,B,D), or OMP antibody (red; C) and regions of overlap are yellow. A:The nerve rootlets form channels (arrows) projecting to the olfactorybulb (OB) from the olfactory epithelium (OE) through the cribriformplate (CP). Neuroligin 3 is localized to the olfactory ensheathing gliafound in the OB and OE, as seen by the extensive overlap (yellow) ofGFAP (red) and neuroligin 1/3 (green) staining. A glia limitans (ar-rowhead), composed primarily of overlapping OEG projections and afew astrocytes, forms a barrier through which only the axons of theolfactory neurons can penetrate into the OB. Both GFAP and neuroli-gin 3 appear to be particularly concentrated at the glia limitans of theOB. The area of fuzzy staining along the cribriform plate is due to theedge of the cartilage that stains positive in our control sections.B:Closer view of the outer layer of the olfactory bulb. The glomerularlayer (GL), which lies just beneath the glia limitans (arrowhead), iswhere the axons of the olfactory neurons terminate. Most GFAP-positive OEG projections (red) also contain neuroligin 3 (green), asshown by the yellow resulting from overlapping staining. A fewGFAP-positive projections (red) that extend away from the glia limi-tans contain lower levels of neuroligin 3 (double arrow). C: Transversesection through a turbinate of olfactory epithelium. OMP identifiesmature olfactory neurons in the outer layer of the olfactory epithelium(ONL) and axonal projections through the lamina propria (LP) intonerve rootlets. Neuroligin 3 (green) and OMP (red) only overlap (yel-low) in regions where the OEG make extensive contacts with theaxonal projections in the lamina propria (concave arrows) and nerverootlets (arrow). Small groups of axons are seen projecting into the LP(concave arrows). The strongest-staining region for neuroligin 3 ap-pears to be at the glial barrier that separates the olfactory nerve layerfrom the lamina propria. D: Closer view of the lamina propria of theolfactory epithelium. The lamina propria (LP) is stained with neuroli-gin 3 (green) throughout and specifically in the OEG projections withGFAP (red). No staining is seen in the olfactory neurons (ONL). Smallbundles of axons are wrapped by glial projections (arrowheads) inwhich GFAP and neuroligin 3 staining overlap (yellow). These bun-dles combine to form larger rootlets (arrows). A,C: 320; B,D: 363.

157VERTEBRATE GLIOTACTIN EXPRESSION

tion. Because neuroligin 3 alone is expressed in theolfactory epithelium, and given the role of the olfactoryensheathing glia during development and regenera-tion, we investigated the cellular distribution of neu-roligin 3 throughout the olfactory neuraxis (Fig. 3). Forthis purpose, a monoclonal antibody that recognizesboth neuroligin 3 and 1 was used to investigate thecellular distribution of these neuroligins (Song et al.,1999). Two stages of mouse development, E17 and P5,were investigated which correspond to the start andpeak of glial cell ensheathing activity. Neuroligin 3 isexpressed in the lamina propria of the olfactory epithe-lium where the olfactory ensheathing glia wrap bun-dles of axons as they exit the olfactory epithelium andproject as rootlets to the olfactory bulb (arrows, Fig.3D). Staining is also seen in the glia associated withthe rootlets that emerge from the lamina propria andpass through the cribriform plate (arrows) and into theglomerular layer of the olfactory bulb (Fig. 3A). Theolfactory epithelium was then double labeled with glialand neuronal antibodies to identify the cell type ex-pressing the neuroligin 3 protein. Neuroligin 3 stainingoverlaps extensively with the GFAP labeling of theolfactory ensheathing glia in both the olfactory epithe-lium and bulb (Fig. 3B,D). The neuroligin 3 stainingonly overlaps with the neuronal (NCAM) staining fromthe point where olfactory axons enter the lamina pro-pria (arrows, Fig. 3C) to their termination in the glo-merular layer of the olfactory bulb. In these regions,the olfactory ensheathing glia are tightly associatedwith the neurons and extend processes around theseaxons. It is these GFAP-positive processes that containthe neuroligin 3 protein. The expression of neuroligin 3is most intense where ever glial to glial connections areformed. For instance, the glial limitans, the boundaryformed by overlapping OEG projections at the bulb, hasintense staining in the GFAP-positive array of glialprocesses (Fig. 3B). Many GFAP processes that extendaway from the boundary are lightly or not stained(double arrows). The expression of neuroligin 3 is alsomore intense at the glial boundary between the basalcell layer and the lamina propria of the olfactory epi-thelium (Fig. 3D). There is no neuroligin 3 staining inthe neuronal layer of the olfactory epithelium (ONL)(Fig. 3C) which indicates that neuroligin 3 is expressedin olfactory glia, but not in the olfactory neurons. Thissame pattern of expression is maintained in adultstages were neuroligin 3 is found in the OEG formingthe glial limitans of the olfactory epithelium but not inthe olfactory neurons (data not shown).

Olfactory ensheathing glia isolated from the olfac-tory epithelium of P5 mice were cultured and tested forneuroligin 3 expression in the absence of neurons (Fig.4). Neuroligin 3 staining is seen over the surface of thecultured olfactory ensheathing glia (Fig. 4B,E). Thestaining appears to be more intense in the glial projec-tions (arrows) coinciding with the GFAP staining (Fig.4C,F). These results indicate that neuroligin 3 is ex-pressed by olfactory ensheathing glia independent ofpresence of neurons.

Neuroligin 3 Is Expressed in Retinal Astrocytes

Given that olfactory ensheathing glia are a uniqueclass of glia that retain their developmental character-istics throughout life, we decided to extend our analysisto other classes of glia in both the developing CNS andPNS. For this analysis, E17 or P1 mice were used todetermine the pattern of neuroligin 3 expression at thebeginning of glial wrapping (at embryonic day 17) aswell as postnatally during the initial stages of myeli-nation and the formation of many glial derived bound-aries (postnatal days 1 or 5). We used the neuroligin 1/3monoclonal antibody in combination with GFAP toidentify other glial cell populations that express neu-roligins during these developmental stages. The retinapossesses two types of glia: Muller cells and astrocytes.The astrocytes originate in the optic stalk and invadethe retina starting after birth (Huxlin et al., 1992; Linget al., 1989). In the P5 retina, the astrocytes which arefound in the outer plexiform layer (OPL) and opticnerve fiber layer (ONFL) stain for GFAP (Sarthy et al.,1991) (Fig. 5A). The Muller glia, which are unique tothe retina and extend through all layers of the retina,do not stain for GFAP. Neuroligin 1/3 staining is seenin the optic nerve fiber layer (ONFL) (Fig. 5B, E) andthe outer plexiform layer (OPL) (Fig. 5B, C) overlap-ping with GFAP-positive astrocytes, and not the Mullerglia (Fig. 5B). Neuroligin 1/3 is expressed in the hon-eycomb-like array of astrocytes that is closely associ-ated with the retinal ganglion cells and the blood ves-sels in the ganglion cell layer (Fig. 5D) (Ramirez et al.,1996). Neuroligin 1/3 also labels the astrocytic pro-cesses found in the nerve fiber layer (NFL) arrangedalong the blood vessels and surrounding bundles ofaxons, as well as the astrocytic processes of the innernerve layer (INL) as they follow the capillaries.

Neuroligin 3 Is Expressed in Developing SpinalCord Astrocytes

Given the expression of neuroligin in the retinal as-trocytes, we investigated the expression of neuroliginin the developing spinal cord to determine whetherother glial populations express neuroligin. We com-pared the pattern of the neuroligin 1/3 monoclonalantibody staining with S-100b and GFAP-stained as-trocytes in the spinal cord at E17 and P1 (Fig. 6).Neuroligin 1/3 staining overlaps with both, S-100b, amarker for immature glia and GFAP, a marker fordifferentiated astrocytes. Neuroligin 1/3 staining is de-tected in the glia of the midline raphe (Fig. 6A,B, ar-rowhead), which is a transient population of S-100bpositive, GFAP2 glia (Fig. 6C) (Van Hartesveldt et al.,1986). Neuroligin 1/3 expression overlaps both GFAPand S-100b in the astrocytes that form the glial limi-tans surrounding the spinal cord (Fig. 6D,E, arrows),but is again present in all S-100b-positive cells, but notall GFAP-positive cells (Fig. 6D,E, arrowheads). At P1neuroligin expression is still present in the glia of the

158 GILBERT ET AL.

Fig. 5. Retinal astrocytes express neuroligin 1/3. Sections throughthe retina of P5 mice were tested for neuroligin 1/3 expression (redimmunofluorescence) and glial fibrillary acidic acid (GFAP) expres-sion (green immunofluorescence or immunoperoxidase). A: Trans-verse retinal section labeled with GFAP (purple immunoperoxidasereaction) to identify the astrocytes in the outer plexiform layer (OPL)and optic nerve fiber layer (ONFL). Other layers of the retina are:outer neuronal layer (ONL), inner neuronal layer (INL), inner plexi-form layer (IPL), and ganglion cell layer (GL). B: Transverse retinalsection labeled with neuroligin 1/3 in red and GFAP in green. Themajority of the astrocytes are found in the ONFL, as is most neuroli-gin staining. The low level of staining observed in the neuronal layers

is reproducible and specific and potentially reflects low levels of neu-roligin 1/3 expression in the retinal neurons or Muller glia. C: Highermagnification of the astrocytes in the outer plexiform layer shown inB. Arrowheads indicate astrocytic projections extending horizontallyin the outer plexiform layer. D: Higher magnification of the astrocytesin the optic nerve fiber layer from a coronal section. Arrows indicateprojections extending throughout the optic nerve fiber layer (ONFL)and extending through the ganglion cell layer (GL) to the innerplexiform layer (IPL). E: Transverse section similar to D, which moreclearly shows the extent of astrocytic projections throughout the opticnerve fiber layer (arrows). A, B: 320; C,E: 340; D: 363.

159VERTEBRATE GLIOTACTIN EXPRESSION

Figure 6.

160 GILBERT ET AL.

midline raphe as well as the astrocytes at the edges ofthe spinal cord though at reduced levels with the great-est concentrations in the astrocytic endfeet (Fig. 6G,H). At both E17 and P1 no other cells in our prepara-tions show expression of the neuroligin protein andgiven the expression of neuroligin 3 in olfactory andperipheral cells, we ascribe this pattern to neuroligin 3not to neuroligin 1.

Neuroligin 3 Is Expressed in DevelopingPeripheral Glia

The expression of neuroligin 3 was examined in theglia of the developing DRG (satellite cells) and thesensory nerves (Schwann cells). Our RT-PCR resultsdetected expression of neuroligins 2 and 3, but not ofneuroligin 1, in mRNA isolated from DRG and fromcultured Schwann cells (Fig. 2A,B). To confirm the PCRresults and to determine the expression pattern of neu-roligin 3, tissue sections were isolated at E17 and P1and labeled using the neuroligin 1/3 and S-100b anti-bodies (Fig. 7). Because we could not detect neuroligin1 expression in these tissues, the expression patternsin the peripheral glia are due to the presence of theneuroligin 3 protein. The satellite cells of the develop-ing DRG show strong co-expression of neuroligin 3 andS-100b (Fig. 7C, D). At late embryonic and early post-

natal stages, most of the Schwann cells present in thenerve are in an immature precursor stage and expressS-100b (Jessen et al., 1994, 1990). Confocal analysisconfirms that the S-100b-positive Schwann cells asso-ciated with the nerves exiting the DRG also expressneuroligin 3 (Fig. 7A,B,C).

DISCUSSION

The search for a vertebrate gliotactin has led us to asurprising candidate—the neuroligin family and inparticular neuroligin 3. Because the known function ofthe neuroligins is at the neuronal synapse, we chose todetermine which neuroligins could be of glial originand when these proteins are expressed during devel-opment. At the molecular level the Drosophila and C.elegans gliotactin and the C. elegans, Drosophila, andmammalian neuroligins form a distinct subfamilywithin the serine esterase family. In other words, glio-tactins and neuroligins are more closely related to eachother across species than any one subfamily member isto other serine esterases from the same species. Thesemembers of the serine esterase family contain a highlyconserved extracellular domain. When quantitativelyanalyzed, the esterase domain is conserved at thestructural level such that this family of proteins hasbeen named “electrotactins” (Botti et al., 1998).

In order for neuroligin 3 to be the equivalent ofgliotactin it must meet three criteria: expression inglia, expression during development and expression inretinal, peripheral and olfactory glia. We determinedthat neuroligin 3 is expressed in a wide range of glialcells, including Schwann cells, retinal astrocytes, spi-nal cord astrocytes, and olfactory ensheathing glia. Therange of neuroligin 3 expression mirrors that seen withgliotactin in Drosophila. Gliotactin is expressed in thedeveloping peripheral glia as they migrate from thelateral most edges of the CNS to their positions in theperiphery. These glia share many developmental char-acteristics and express proteins conserved with thosefound in nonmyelinating Schwann cells (Auld, 1996;Sepp et al., 2000). Gliotactin is also expressed in theretinal glia which are a class of glia that migrate fromthe developing optic stalk into the Drosophila retina(Choi and Benzer, 1994; Rangarajan et al., 1999). Thisglial type is similar to the astrocytes that migrate fromthe optic nerve into the retina in vertebrates. Finallygliotactin is expressed in the glia associated with theolfactory neurons from the maxillary palp and anten-nae (T. Rice and V. Auld, unpublished data).

We chose to focus our analysis on neuroligin 3 in theolfactory neuraxis due to the unique nature of theolfactory ensheathing glia. These glia share propertieswith both astrocytes and Schwann cells and maintaintheir developmental characteristics throughout life.The olfactory ensheathing glia are involved in the con-stant remodeling of the olfactory primary projectionsand are responsible for shepherding their axons acrossthe PNS–CNS transitional zone (Doucette, 1991). This

Fig. 6. Neuroligin is expressed in S-100b and glial fibrillary acidicprotein (GFAP)-positive glia during spinal cord development. Trans-verse sections from E17 and P1 mice were stained with antibodies toneuroligin 1/3 (red in A,B,D,E,G,H) and either GFAP (green in A,C-,D,F,G,I) or S-100b (red in C,F,I; green in B,E,H). All panels areoriented dorsal side up. A–F: Neuroligin 3 expression coincides withS-100b in E17 spinal cord. A: GFAP-positive (green) astrocytes arefound predominantly in the peripheral layers of the spinal cord. Eachpanel is a confocal image of E17 spinal cord sections at 320 (A–C) or363 (D–F). By contrast, neuroligin-positive (red) cells are found pre-dominantly in the midline raphe (arrowhead), and to a lesser extent inthe peripheral layers of the spinal cord (arrow and yellow whereneuroligin and GFAP staining overlap). B: Neuroligin (red) andS-100b (green) staining patterns completely overlap in the midlineraphe (arrowhead) and the peripheral layers of the spinal cord (ar-row), resulting in yellow. C: S-100b (red) predominantly stains themidline raphe (arrowhead) overlapping with GFAP (green) primarilyin the periphery of the spinal cord (arrow). D: Neuroligin (red) stain-ing overlaps extensively with the GFAP-positive (green) cells in theperipheral layers of the spinal cord (yellow and arrows), but not allneuroligin-positive cells are GFAP-positive (arrowheads). E: All neu-roligin-positive (red) cells are S-100b-positive (green) in the periph-eral layers of the spinal cord (arrow), as well as in the midline raphe.Staining with both antibodies extends from the cell body (arrowhead)to the ends of the projections (three are seen extending from this cellbody). F: S-100b (red) expression partially overlaps GFAP (green)expression in the periphery of the spinal cord (yellow staining andarrows). G–I: Neuroligin 3 staining is reduced in the spinal cord gliaat postnatal day 1. All panels are confocal images of P1 spinal cordsections at 363. G: Neuroligin (red) staining is still found in manyGFAP-positive (green) astrocytes, but at much lower levels. Somestaining is still seen in the projections (arrows), but the highest levelsremaining are in the astrocytic endfeet (arrowhead). H: Neuroligin(red) and S-100b (green) are still expressed in the same cells at P1, butthe levels of both have decreased significantly. Astrocytic projectionslightly stain with both antibodies (arrows) and endfeet more strongly(arrowhead). I: S-100b (green) staining is decreased relative to GFAP(red). Co-staining processes are labeled with arrows, and an astrocyticendfoot with an arrowhead. This increase in GFAP and decrease inS-100b is characteristic of astrocytes as they mature.

161VERTEBRATE GLIOTACTIN EXPRESSION

ability to guide and wrap axons in both PNS and CNSenvironments is thought to contribute to the success oftransplanted OEG in promoting the regeneration ofseveral different classes of CNS neurons (Kafitz andGreer, 1999; Li et al., 1997, 1998; Navarro et al., 1999;Sonigra et al., 1999). Our results show that neuroligin3 is the only neuroligin expressed in the olfactory epi-

thelium and its expression is confined to the olfactoryensheathing glia. We found that in the olfactory systemthe level of neuroligin 3 expression is increased atpoints of glia–neuronal contact as the glia wrap theaxons in the nerve root (Fig. 3C). In addition, neuroli-gin 3 expression increases where glial–glial contactsare formed as glia connect to generate the glia limitans

Fig. 7. Immature peripheral glia in the dorsal root ganglion (DRG)express neuroligin 3. The expression pattern of neuroligin 3 wasassessed in the developing DRG. A transverse section of mouse DRGat P1 was labeled with neuroligin (A) and S-100b (B). The overlap inexpression pattern was assessed by confocal microscopy at E17 (C,D).Neuroligin 1 is not expressed in these cells; thus, the staining pat-terns reflects neuroligin expression only. A: Neuroligin 3 is expressedin the satellite cells of the DRG and in the Schwann cells (arrow) of

the nerve root. B: S-100b is expressed both the satellite cells of theDRG and the Schwann cells of the nerve root (arrow). C: A low-magnification overview of the DRG showing coexpression (yellow) ofneuroligin 3 (green) and S-100b (red) in the satellite cells of the DRGand in Schwann cells of the nerve root (arrow). D: Higher magnifica-tion of C, showing the overlap of neuroligin (green) and S-100b (red)in the satellite glia (arrowhead) surrounding neuronal cell bodies(arrow).

162 GILBERT ET AL.

of the olfactory bulb (Fig. 3B) and the lamina propria ofthe olfactory neuraxis (Fig. 3C). Given the function ofgliotactin in Drosophila during glial sheath formation,the function of neuroligin 3 in developing glia and adultolfactory glia maybe in the formation of cellular con-tacts or junctions that form between these cells duringdevelopment. Previous studies on members of the neu-roligin family support the hypothesis that gliotactin/neuroligin may work to establish the formation of junc-tional complexes at the point of glia–glia and glia–neuronal contact. Neuroligin 1, for instance, is foundconcentrated at excitatory synaptic junctions in theCNS, where it is associated with large multimeric com-plexes (Song et al., 1999). Both neuroligin 1 and 2 havebe shown to capable to triggering the formation ofsynapses even when expressed in non-neuronal cells(Scheiffele et al., 2000).

Further support for this role comes from our obser-vations of neuroligin expression in the retina and spi-nal cord. In the retina, neuroligin expression is locatedin the GFAP-positive astrocytes, which are found onlyin the ganglion cell and nerve fiber layers. GFAP-pos-itive retinal astrocytes migrate from the optic stalkinto the retina starting at about postnatal day 1 in themouse (Sarthy et al., 1991). At early postnatal stages,neuroligin expression is found associated with theGFAP-positive processes of the astrocytes in the hon-eycomb-like array of glial processes that surround theganglion cells and in the lamina propria of the retinalnerve layer (Fig. 5D,E). It is in the neuroligin 3 positiveregions that extensive connections are made betweenthe astrocytes and Muller cells, and between adjacentastrocytes (Ramirez et al., 1996). In the developingspinal cord, at both the embryonic and early postnatalstages, neuroligin is strongly expressed in the midlineraphe glial structure (Fig. 6B). The expression ofS-100b in the midline glial structure is transient fromE15 to P7-8 in the midline of the midbrain, hindbrain,and spinal cord of rats (Van Hartesveldt et al., 1986).This region does not express GFAP and has also beenreported to express vimentin, another intermediateprotein found in immature glia (Oudega and Marani,1991). As development proceeds, more S-100b fibersare added to the midline glial structure, leading to aproposed role in axon guidance and as a barrier toaberrant decussation of growing axons (Takano andBecker, 1997). This is suggestive of a role for neuroliginin mediating the cellular associations that must occurbetween the glial cells in the formation of these struc-tures.

We also find that neuroligin 3 expression is develop-mentally regulated in a manner similar to that ofS-100b. For instance at the midline raphe and theedges of the spinal cord S-100b and neuroligin arecoexpressed in exactly the same pattern (Fig. 6E). Theexpression patterns of both proteins only overlaps asubset of the GFAP-positive astrocytes and perhapsrepresent a discrete step in the development of astro-cytes. S-100b along with vimentin are found in imma-ture astrocytes, whereas GFAP expression is thought

to mark astrocytes as they mature (Tardy et al., 1989).There is a gradual transition in expression of differentintermediate filament proteins from intermediate fila-ment-associated protein (IFAP) to vimentin to GFAPexpression as glia develop (Oudega and Marani, 1991;Yang et al., 1993). And it appears that as vimentin andS-100b expression is reduced so is neuroligin 3 expres-sion in these glia. Conversely in the olfactory neuraxisneuroligin 3 and S-100b continue to be co-expressed insubsets of adult ensheathing glia (data not shown).

The fact that neuroligin 3 and Drosophila gliotactinare conserved at the molecular level both in their struc-ture and in their expression patterns suggests that thefunction of these proteins will also be conserved. Thisgives rise to the hypothesis that neuroligin 3 and glio-tactin play a role in the formation of the cellular junc-tions that mediate glia–glia or glia–neuron interac-tions during nervous system development. The truecomparison with gliotactin and testing of the role ofneuroligin 3 in the development of vertebrate glia willhave to await the analysis of the neuroligin 3 knockoutmutation.

ACKNOWLEDGMENTS

The authors thank Dr. Brose for the source of theneuroligin1/3 monoclonal antibody. We thank WolframTetzlaff for his comments and suggestions. This work isfunded by grants from the Howard Hughes MedicalInstitute, the Rick Hansen Neurotrauma Foundation,and the British Columbia Health Research Foundation(to V.J.A.) and by the CCMT/Merck Frosst (to A.J.R.).

REFERENCES

Adams MD, Celniker SE, Holt RA, Evans CA, Gocayne JD, Amanati-des PG, Scherer SE, Li PW, Hoskins RA, Galle RF, et al. 2000. Thegenome sequence of Drosophila melanogaster. Science 287:2185–2196.

Au E, Roskams AJ. 2001. Culturing olfactory ensheathing glia fromthe mouse olfactory epithelium. In: Zigova, T, Sanchez-Ramos, J.editors. Methods in molecular medicine: stem cells—approachesand protocols. Cambridge: Humana Press (in press).

Auld VJ. 1996. The role of glia in the development of the insectnervous system. In: Jessen JR, Richardson WD, editors. Glial celldevelopment. Oxford: Bios Scientific. p 229–250.

Auld VJ, Fetter RD, Broadie K, Goodman CS. 1995. Gliotactin, a noveltransmembrane protein on peripheral glia, is required to form theblood–nerve barrier in Drosophila. Cell 81:757–767.

Botti SA, Felder CE, Sussman JL, Silman I. 1998. Electrotactins: aclass of adhesion proteins with conserved electrostatic and struc-tural motifs. Protein Eng 11:415–420.

Brockes JP, Fields KL, Raff MC. 1979. Studies on cultured ratSchwann cells. I. Establishment of purified populations from cul-tures of peripheral nerve. Brain Res 165:105–118.

Choi KW, Benzer S. 1994. Migration of glia along photoreceptor axonsin the developing Drosophila eye. Neuron 12:423–431.

Chomczynski P, Sacchi N. 1987. Single-step method of RNA isolationby acid guanidinium thiocyanate-phenol-chloroform extraction.Anal Biochem 162:156–159.

Doucette R. 1990. Glial influences on axonal growth in the primaryolfactory system. Glia 3:433–449.

Doucette R. 1991. PNS–CNS transitional zone of the first cranialnerve. J Comp Neurol 312:451–466.

163VERTEBRATE GLIOTACTIN EXPRESSION

Franceschini IA, Barnett SC. 1996. Low-affinity NGF-receptor andE-N-CAM expression define two types of olfactory nerve ensheath-ing cells that share a common lineage. Dev Biol 173:327–343.

Higgins DG, Sharp PM. 1988. CLUSTAL: a package for performingmultiple sequence alignment on a microcomputer. Gene 73:237–244.

Huxlin KR, Sefton AJ, Furby JH. 1992. The origin and development ofretinal astrocytes in the mouse. J Neurocytol 21:530–544.

Ichtchenko K, Hata Y, Nguyen T, Ullrich B, Missler M, Moomaw C,Sudhof TC. 1995. Neuroligin 1: a splice site-specific ligand forbeta-neurexins. Cell 81:435–443.

Ichtchenko K, Nguyen T, Sudhof TC. 1996. Structures, alternativesplicing, and neurexin binding of multiple neuroligins. J Biol Chem271:2676–2682.

Jessen KR, Brennan A, Morgan L, Mirsky R, Kent A, Hashimoto Y,Gavrilovic J. 1994. The Schwann cell precursor and its fate: a studyof cell death and differentiation during gliogenesis in rat embryonicnerves. Neuron 12:509–527.

Jessen KR, Morgan L, Stewart HJ, Mirsky R. 1990. Three markers ofadult non-myelin-forming Schwann cells, 217c (Ran-1), A5E3 andGFAP: development and regulation by neuron–Schwann cell inter-actions. Development 109:91–103.

Kafitz KW, Greer CA. 1999. Olfactory ensheathing cells promoteneurite extension from embryonic olfactory receptor cells in vitro.Glia 25:99–110.

Kolodkin AL, Matthes DJ, Goodman CS. 1993. The semaphorin genesencode a family of transmembrane and secreted growth cone guid-ance molecules. Cell 75:1389–1399.

Li Y, Field PM, Raisman G. 1997. Repair of adult rat corticospinaltract by transplants of olfactory ensheathing cells. Science 277:2000–2002.

Li Y, Field PM, Raisman G. 1998. Regeneration of adult rat cortico-spinal axons induced by transplanted olfactory ensheathing cells.J Neurosci 18:10514–10524.

Ling TL, Mitrofanis J, Stone J. 1989. Origin of retinal astrocytes inthe rat: evidence of migration from the optic nerve. J Comp Neurol286:345–352.

Navarro X, Valero A, Gudino G, Fores J, Rodriguez FJ, Verdu E,Pascual R, Cuadras J, Nieto-Sampedro M. 1999. Ensheathing gliatransplants promote dorsal root regeneration and spinal reflex res-titution after multiple lumbar rhizotomy. Ann Neurol 45:207–215.

Oudega M, Marani E. 1991. Expression of vimentin and glial fibrillaryacidic protein in the developing rat spinal cord: an immunocyto-chemical study of the spinal cord glial system. J Anat 179:97–114.

Ramirez JM, Trivino A, Ramirez AI, Salazar JJ, Garcia-Sanchez J.1996. Structural specializations of human retinal glial cells. VisionRes 36:2029–2036.

Ramon-Cueto A, Avila J. 1998. Olfactory ensheathing glia: propertiesand function. Brain Res Bull 46:175–187.

Ramon-Cueto A, Cordero MI, Santos-Benito FF, Avila J. 2000. Func-tional recovery of paraplegic rats and motor axon regeneration intheir spinal cords by olfactory ensheathing glia. Neuron 25:425–435.

Rangarajan R, Gong Q, Gaul U. 1999. Migration and function of gliain the developing Drosophila eye. Development 126:3285–3292.

Roskams AJ, Bethel MA, Hurt KJ, Ronnett GV. 1996. Sequentialexpression of Trks A, B, and C in the regenerating olfactory neuro-epithelium. J Neurosci 16:1294–1307.

Roskams AJ, Cai X, Ronnett GV. 1998. Expression of neuron-specificbeta-III tubulin during olfactory neurogenesis in the embryonic andadult rat. Neuroscience 83:191–200.

Sarthy PV, Fu M, Huang J. 1991. Developmental expression of theglial fibrillary acidic protein (GFAP) gene in the mouse retina. CellMol Neurobiol 11:623–637.

Scheiffele P, Fan J, Choih J, Fetter R, Serafini T. 2000. Neuroliginexpressed in nonneuronal cells triggers presynaptic development incontacting axons. Cell 101:657–669.

Sepp K, Schulte J, Auld VJ. 2000. Developmental dynamics of periph-eral glia in the embryo of Drosophila melanogaster. Glia 30:122–133.

Song JY, Ichtchenko K, Sudhof TC, Brose N. 1999. Neuroligin 1 is apostsynaptic cell-adhesion molecule of excitatory synapses. ProcNatl Acad Sci USA 96:1100–1105.

Sonigra RJ, Brighton PC, Jacoby J, Hall S, Wigley CB. 1999. Adult ratolfactory nerve ensheathing cells are effective promoters of adultcentral nervous system neurite outgrowth in coculture. Glia 25:256–269.

Takano T, Becker LE. 1997. Developmental change of the nestin-immunoreactive midline raphe glial structure in human brainstemand spinal cord. Dev Neurosci 19:202–209.

Tardy M, Fages C, Riol H, LePrince G, Rataboul P, Charriere-Ber-trand C, Nunez J. 1989. Developmental expression of the glialfibrillary acidic protein mRNA in the central nervous system and incultured astrocytes. J Neurochem 52:162–167.

Van Hartesveldt C, Moore B, Hartman BK. 1986. Transient midlineraphe glial structure in the developing rat. J Comp Neurol 253:174–184.

Yang HY, Lieska N, Shao D, Kriho V, Pappas GD. 1993. Immunotyp-ing of radial glia and their glial derivatives during development ofthe rat spinal cord. J Neurocytol 22:558–571.

164 GILBERT ET AL.