Identification of functional marker proteinsin the mammalian growth coneMotohiro Nozumia,b, Tetsuya Toganoa,c, Kazuko Takahashi-Nikia, Jia Lua,c, Atsuko Hondaa,b, Masato Taokad,e,Takashi Shinkawad,e, Hisashi Kogaf,g, Kosei Takeuchia,b, Toshiaki Isobed,e, and Michihiro Igarashia,b,1

Divisions of aMolecular Cell Biology and cOphthalmology, Graduate School of Medical and Dental Sciences, and bTransdisciplinary Research Program, NiigataUniversity, 1-757 Asahi-machi, Chuo-ku, Niigata, Niigata 951-8510, Japan; dDepartment of Chemistry, Graduate School of Science, Tokyo MetropolitanUniversity, 1-1 Minami-Osawa, Hachioji, Tokyo 192-0397, Japan; eCore Research for Evolutional Science and Technology, Japan Science and TechnologyAgency, Sanbancho 5, Chiyoda-ku, Tokyo 102-0075, Japan; and fChiba Industry Advancement Center and gCollaboration of Regional Entities for theAdvancement of Technological Excellence Program, Kazusa DNA Research Institute, 2-6-7 Kazusa-Kamatari, Kisarazu, Chiba 292-0818, Japan

Edited by Lynn T. Landmesser, Case Western Reserve University, Cleveland, OH, and approved August 24, 2009 (received for review April 18, 2009)

Identification of proteins in the mammalian growth cone has thepotential to advance our understanding of this critical regulator ofneuronal growth and formation of neural circuit; however, to date,only one growth cone marker protein, GAP-43, has been reported.Here, we successfully used a proteomic approach to identify 945proteins present in developing rat forebrain growth cones, includinghighly abundant, membrane-associated and actin-associated pro-teins. Almost 100 of the proteins appear to be highly enriched in thegrowth cone, as determined by quantitative immunostaining, and for17 proteins, the results of RNAi suggest a role in axon growth. Mostof the proteins we identified have not previously been implicated inaxon growth and thus their identification presents a significant stepforward, providing marker proteins and candidate neuronal growth-associated proteins.

GAP-43 � proteomics � RNAi � neuronal growth-associated proteins �axon guidance

Neuronal growth cones execute important steps in neural wiring,including axonal growth and pathfinding, and accurate synap-

togenesis (1, 2). Whether or not an injured axon in the adult neuronregenerates or degenerates depends on surrounding factors thateither maintain or inhibit formation of growth cone-like structures,consistent with a critical role for growth cones in neural plasticityand repair of the adult brain (3, 4). It follows, then, that identifying themolecular basis of growth cone behavior will be critical to understand-ing the cellular mechanisms of higher brain functions. A significantbarrier to this understanding, however, is that little is known aboutthe molecular makeup of the mammalian growth cone.

Classical genetic approaches to identifying key players in brainfunction have been informative in model systems such as Caeno-rhabditis elegans and Drosophila. For example, identification ofmutations in specific C. elegans and Drosophila genes contributedgreatly to the discovery and functional characterization of axonguidance molecules in the mammalian brain, such as netrin andsemaphorins (1, 2). But these are not the only useful tools withwhich to dissect growth cone functions, and they have significantlimitations in terms of their ability to further our understanding ofthe complete picture of the molecular machinery that controlsmammalian growth cones (1, 2). Among these limitations are thatsimilar studies are not feasible in mammals and mammalian growthcone functions are thought to be much more complicated than growthcone functions in C. elegans and Drosophila. Additionally, the molecularredundancy involved in growth cone functions in the mammalianCNS is likely to be much larger than in model organisms.

Despite the knowledge gap, however, we know of no previousreport that applies a systematic approach to identification ofmammalian growth cone proteins. However, cell biological studiescombined with pharmacological tools to detect second messengers(Ca2� or cyclic nucleotides) or the cytoskeleton (i.e., F-actin ormicrotubules) have revealed some of the signaling pathways in-volved in control of mammalian growth cone behavior or guidance.Nonetheless, we are far from having a complete understanding of

mammalian growth cone functions, in particular interactions orrelationships among signaling pathways, at least in part because ofan insufficient understanding of what key molecules are importantfor function (1, 2).

Indeed, even what proteins might be present, i.e., molecularmarkers of the growth cone, is a relatively unexplored territory,particularly as compared with our understanding of synaptic mo-lecular marker proteins. In the case of adult synapses, a largenumber of marker proteins localized to various sublocations in thesynapse are known (5, 6). These include proteins localized tosynaptic vesicles, presynaptic and postsynaptic membranes, activezones, and postsynaptic density (5, 6).

Without a doubt, identification of synaptic marker proteins hasmarkedly enriched our knowledge of the molecular machineryunderpinning synaptic structures and functions (5, 6). This standsin contrast to the very small amount of marker protein informationavailable for the growth cone. Indeed, GAP-43 (growth-associatedprotein 43-kDa; neuromodulin) (7) is the only previously identifiedfunctional molecular marker of growth cones (ideally, a ‘‘functionalmolecular marker’’ would be a protein that is both highly concen-trated in the growth cone and involved in axon growth). GAP-43 isconcentrated in the growth cone, highly expressed, transportedwhen a damaged axon can regrow, and involved in sprouting (7).Although it is concentrated in the growth cone, GAP-43 is alsodetectable in the axon (7). Given the severely limited amount ofinformation about functional markers concentrated in or localizedto growth cones, it follows that identification of putative novelfunctional molecular markers of the mammalian growth conewould be extremely valuable to further study.

To help gain a systemwide understanding of the molecularcomponents of growth cones and identify novel molecular markercandidates, we introduced proteome-scale approaches that havebeen used successfully to identify large numbers of proteins presentin other specific cells or tissues, thus contributing to a more globalunderstanding of the functions of those cells or tissues (8–11). Wesucceeded in identifying 945 species of proteins in growth coneparticle (GCP) and/or a growth cone membrane membrane (GCM)(12–14). By combining the results of immunolocalization and RNAistudies with proteomics, we provide evidence that 17 of the proteinswe identified are highly concentrated in the growth cone area andregulate axonal growth, concluding that they are unique functionalmolecular markers of the growth cone.

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

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

1To whom correspondence should be addressed. E-mail: tarokaja@med.niigata-u.ac.jp.

This article contains supporting information online at www.pnas.org/cgi/content/full/0904092106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0904092106 PNAS � October 6, 2009 � vol. 106 � no. 40 � 17211–17216

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

May

23,

202

0

ResultsLarge-Scale Identification of Proteins in Rat Brain Growth Cones. Ourfirst goal was to identify a large number of proteins expressed ingrowth cones, including proteins common to many cell types andproteins involved in growth cone-specific functions. To do this, wefirst separated proteins from the developing rat forebrain viasubcellular fractionation to obtain a GCP fraction (Fig. 1A). Wethen obtained a GCM via hypotonic treatment of the GCP fraction(see details in SI Text) with the goal of identifying minor membraneproteins in the growth cone. Subsequently, we used multidimen-sional liquid chromatography-tandem mass spectrometry (LC/MS/MS) to identify unique proteins in the sample, because the methodhas been shown to be suitable to large-scale protein identification(15, 16). The method proved powerful in this study as well, with atotal of 945 species of proteins identified in GCP and GCM (Fig.1B). Because �50 of the proteins were previously known to beexpressed in the mammalian growth cone, identification of thislarge number of proteins provides a wealth of molecular informa-tion about mammalian growth cones (2, 4, 17). For comparison, wealso analyzed adult synaptosomes, the counterpart of the growthcone, and identified 1,407 species of synaptosomal proteins (i.e.,twice as many as we found for the GCP and GCM). Approximately65% of the synaptosomal proteins we identified are not found in theGCP or the GCM sets (Fig. 1C), which may be because a large

number of proteins are newly synthesized and added to synapticcomponents for synaptic transmission after synaptogenesis. Wesucceeded in identifying 96 and 141 proteins in the GCP or GCM,respectively, that were not found in the synaptosomes (Fig. 1C).

Proteins Identified by Proteomics Analysis. We next used bioinformat-ics analysis to categorize the proteins into functionally relatedgroups (Fig. 1 C–E and Table S1, Table S2, and Table S3). Weincluded synaptosome data in our analysis, because the presynapticaxon terminal is the adult counterpart of the growth cone. Theproteins were in the following functional categories: those requiredfor cytoskeletal reorganization, involved in vesicular trafficking,and related to signal transduction, including protein kinases, phos-phatases, and G proteins (2, 4). As expected, previously identifiedkey players in growth cone function were identified in the study(including MAP1B and myosin II heavy chain). But intriguingly,most proteins in each category were newly identified (i.e., notpreviously reported as GCPs) (2, 4, 15) (Fig. 1 D and E; see TableS1, Table S2, and Table S3).

As expected, the GCP fraction contained both cytosolic andmembrane-bound proteins; the GCM fraction was enriched formembrane and membranous organelle-associated proteins; and theset of proteins found in the GCP fraction but not the GCM fractionwas enriched for cytosolic proteins (Fig. 1D). As shown in Fig. 1D,

GCP(629) GCM (592)

276 316353

A B

C

23.0 14.8

7.3 10 1

0 5 10 15 20 25 30

Metabolic enzymeCytoskeletal

SignalingGTP binding

Number of proteins (%)D

10.1 6.9

5.8 4.4

1.0 5.4

3.8 2.6

4.6 1.5

8.8

GTP-bindingIon transport/Channel

Membrane trafficKinase/Phosphatase

ReceptorChaperone

Proteasome/UbiquitinationCell adhesion

RibosomalProtein translation

Miscellaneous

10

20

30

40

50

60

Nu

mb

er o

f pep

tid

es

GCP

GCM

Synaptosome

12.6 9.4

Metabolic enzymeCytoskeletal

GCP

GCP > GCM

E

20

30

40

50

60

mb

er o

f pep

tid

es

0

7.9 11.2

10.3 8.7

3.5 8.7

4.0 0.9

6.5 4.5

0.9 10.9

ySignaling

GTP-bindingIon transport/Channel

Membrane trafficKinase/Phosphatase

ReceptorChaperone

Proteasome/UbiquitinationCell adhesion

RibosomalProtein translation

Miscellaneous

24 9Metabolic enzyme

GCM

GCP < GCM

0

10

Nu

m

24.9 11.4

10.2 7.6 7.5

6.6 4.9

3.0 2.7 2.5 2.1

0.6 0.6

15.4

Metabolic Cytoskeletal

SignalingGTP-binding

Ion transport/ChannelMembrane traffic

Kinase/PhosphataseReceptor

ChaperoneProteasome/Ubiquitination

Cell adhesionRibosomal

Protein translationMiscellaneous

Synaptosome

Fig. 1. Proteomic analysis of GCPs and adultsynaptosomes. (A) Electron micrograph of GCPs.The GCP fraction was prepared by using themethod described by Gordon-Weeks (14) (see SIText). The criterion for a GCP was a particle with adiameter of �1–2 �m and the presence of smallclear vesicles. (Scale bar: 10 �m.) (B) GCP proteinsand proteins found in a membrane subfraction ofGCP (GCM). In total, 629 GCP and 592 GCM pro-teins were identified, with 276 proteins commonto both. Note that 316 proteins were identifiedonly in the GCM subfraction. (C) Comparison ofthe GCP/GCM proteome with the proteome ofadult synaptosomes. A total of 1,407 synaptoso-mal proteins were identified, for about twice asmany as were identified in GCP or GCM. (D) Cat-egorization of GCP, GCM, and synaptosomal pro-teins. Categories are color-coded as follows: met-abolic enzymes, violet; cytoskeletal proteins,navy; signaling (except GTP-binding or phosphor-ylation), blue; proteins involved in guanine nucle-otide cycling (including GTP-binding proteins),cyan; ion transport/channels, green; membranetraffic, lime; kinases or phosphatases, yellow; re-ceptors, orange; chaperone, brown; proteasome/ubiquitination-related proteins, maroon; cell ad-hesion proteins, red; ribosomal, magenta;proteins involved in translation, olive; miscella-neous proteins (including organelle-specific orundefined), gray. Note that the ratio of cell adhe-sion molecules, receptors, and transporters/channels is higher in GCM than in GCP, consistentwith enrichment of membrane components inGCM. See Table S1, Table S2, and Table S3 for adetailed report. (E) The major proteins of GCP orGCM as compared with synaptosomal proteins asindicated by the number of identified peptides.Among proteins for which 12 or more peptideswere identified, we chose the subset identified aspeptides at least twice as many times in GCP or inGCM than in synaptosomes. The number of pep-tide identifications is shown. (Upper) GCP � GCM.(Lower) GCM � GCP. Color-coded by functionalcategory as in D. Note that the number of identi-fied peptides is an indicator of relative proteinlevels.

17212 � www.pnas.org�cgi�doi�10.1073�pnas.0904092106 Nozumi et al.

Dow

nloa

ded

by g

uest

on

May

23,

202

0

the set of membrane-associated proteins we identified included celladhesion molecules (2.6% in GCP and 6.5% in GCM), receptorsand receptor-like membrane proteins (1.0% in GCP and 8.7% inGCM), and transporters/channels (6.9% in GCP and 10.3% inGCM). Components of ubiquitin proteasome were detected onlyin the GCP fraction (Fig. 1D). Synaptosomes contain a largernumbers of metabolic enzymes, but the percentages of ribosomalproteins and local protein translation proteins detected in synap-tosomes were lower than in GCP (Fig. 1D).

In our shotgun proteomic analysis, we tentatively assumed thatthe number of times a given peptide is identified correlates with theabundance of the protein in cells (18). In Fig. 1E, we report thoseproteins for which peptides were independently identified 12 ormore times in GCP or GCM (together, these comprise 50% of thepeptides identified in GCP or GCM), and for which the abundancein GCP or GCM was 2-fold or more compared with levels insynaptosomes. These appear to be the most highly abundantproteins in the GCP and GCM fractions, and thus we definedproteins appearing 12 or more times as ‘‘major proteins’’ in the GCPor in GCM fractions. These include cytoskeletal components(tubulins, the microtubule-associated protein MAP1B, dyneinheavy chains, and myosin heavy chains), collapsin response medi-ator proteins (CRMPs), catenins, 14-3-3 proteins, and G� proteins(namely, Gq, Gi1, and Gi2). The known or predicted biochemicaland/or biological activities of the major proteins are consistent withfunctional relevance in growth cones. Moreover, identification ofthe translation factor elongation factor 1� is consistent with thefinding that local protein synthesis is important for growth conebehavior (19). Additionally, identification of CRMP family mem-bers (CRMP4b, CRMP1, and CRMP5; Fig. 1E) as the mostabundant GCP proteins in growth cones correlates well with areport that CRMP2 acts as a tubulin adapter protein and is involvedin axon formation (20).

Verification of Growth Cone Localization Suggests a Negligible False-Positive Rate. We next wanted to determine the specificity of ourproteomics approach,; i.e., to determine the false-positive rate.Generally, a comparison of a given fraction should be made toanother biochemical fraction. However, no fractions from thedeveloping brain can be prepared with comparable purity to theGCP or GCM fractions. Thus, to facilitate generation of a com-parison dataset, we performed systematic immunodetection of asubset of the proteins in cultured cortical neurons (see Table S4,Table S5, Table S6, and Table S7). We excluded ubiquitous or

commonly expressed proteins such as metabolic enzymes andmolecular chaperones so that we could instead focus on proteinsthat may be particularly relevant to growth cone-specific functions.In total, we looked at the distributions of 131 proteins (i.e., �15%of the proteins we identified). The data confirm that in cultured ratcortical neurons all of the proteins we tested are detectable in thegrowth cone area (Fig. 2 and Table S4). Indeed, no proteins wetested could be detected in other axonal regions but not in thegrowth cone, suggesting a very low false-positive rate and validatingour overall approach. In addition, our proteomic data in GCP orGCM contained no transcription factors, suggesting that contam-ination with nuclear components is also negligible (Fig. 1D, TableS1, and Table S2).

Proteins More Concentrated in the Growth Cone Than in the Distal Axon.We next sought to determine the extent to which the proteinsidentified in our study are specific or locally specific to the growthcone. To do this, we compared the distributions of the proteins ingrowth cones with their distribution in distal axons. To the extentthat the method is quantitative, we were also able to compare therelative concentrations of proteins in growth cones versus distalaxons (Fig. 2 and Table S4). We defined the growth cone accu-mulation index as the ratio of fluorescence intensity (FI) in thegrowth cone compared with that in the distal axon (Fig. 2, Table S4,SI Text, and Fig. S1). We also used another index, i.e., the area ratio,to examine the distribution patterns of each protein (Fig. S2). Thegrowth cone accumulation index is an indication of the relative levelof protein accumulation in the growth cone. By applying thestatistical Wilcoxon rank-sum test to our results, facilitating strictclassification of the examined proteins as compared with GAP-43,our quantification of the systematic immunostaining approachusing this index revealed that as many as 69 proteins identified byproteomics were detected at higher levels in the growth cone areathan in the distal axon. These proteins appear to be at much higherlevels in growth cones than the currently established growth conemarker protein GAP-43 (7) (Fig. 2, shown in red). We also foundthat for 33 proteins the statistical error areas overlap with that ofGAP-43, thus we categorize this set of proteins as concentrated inthe growth cone to a similar degree as is GAP-43 (Fig. 2, shown inblue).

RNAi Analysis Reveals Relevance to Axon Length and Functional Markersof the Growth Cone. We used an RNAi-based approach to test theroles of marker protein candidates in axonal growth. Activity-

Ank2 Macf1

8

9

10

14-3-3e4.1N

Ap2b1

Ctnna1Ctnna2

Ap180Arp1

Arvcf

BasiginFabp7

Tubb3

Fapb5

Calnexin

Calm

Camkv

Cap1

Capzb

CAR

Cd47

Clasp

Cltc

Clptm1

Cofilin

Cotl1

C 1

Crmp3Crmp4

Crmp5Cyfip1

DestrinDynamin

F-actin

Gnb2

Gng2

Gng3

Gnai2Git1

Gnao

Hist4

Hsp110

Igsf4

Ktn

Lasp1Lphn3

Letm1M6a

Marcksl1

Ripx

Munc-18

Myo5a

Ncdn

Ndrg1

Nrxn1

Rtn4

Odz2

Pacs1

PakPalm

PAR3B

Picalm

Pclo

Pin

PP2A

Ppfia1

PSD-93

Ptprd

PtprsRab18

Rab35Rab3a

Rab9 Gdi1Rac

R

Rcp

Rtn1

Arhgdia

RKM23

Scamp

Scg10

Sept2

Snap25

Snap29

Spg3a

Stmn1Strap

Stx1

Stx7

Sv2a

Sv2b Syn2

Syt

Tau

Inexa

b

Tip120

FishTmod

Tpbg

V-1

Vamp

Atp6v1a

Vcp

Xlas

Farp2

3

4

5

6

7

8

Are

a ra

tio

(GC

/Axo

n)

GAP-43

Ap2a1

Tuba1

Ctnnb1

BIP

Tubb3

Camk2

Csnk2

Caspr2Celsr3

Crmp1

Crmp2Dclk1Dcx

FascinHsp110

Lamp1Lyric

Map1bAtp1a1

NF200

PrphPlxnc1

Rala RasSpg3a

Stx16

Stx8Tmsb4xV 1

0

1

2

0.7 0.8 0.9 1.0 1.1 1.2 1.3 1. 4 1.5 1.6 1.7 1.8 1.9 2.0 2.1

Fluorescent intensity ratio (GC/Axon)

Fig. 2. Immunofluorescence quantification of GCP.Horizontal axis, FI ratio (growth cone/distal axon), Lon-gitudinal axis, area ratio (growth cone/distal axon).The horizontal axis shows the GC accumulation index(a ratio of 1.0; the vertical black dotted line indicatesthat a given protein is evenly distributed in the distalaxon versus the growth cone). The FI ratio for GAP-43(Gap43; 1.315) is shown as a blue dotted line. Proteinsin the upper right region of the graph are more con-centrated in the growth cone than in the axon. Manyof these are actin-binding proteins and proteins in-volved in vesicular trafficking. The FI ratio of eachprotein was tested by using the Kruskal-Willis statisti-cal test (46) with GAP-43 as the comparison point. Theproteins were grouped based on the two-sided 95%confidence interval, i.e., was it higher than, similar to,or less than that of GAP-43 (shown in red, blue, andblack, respectively). Note that as expected, F-actin (asdetected with rhodamine phalloidin; in red), which isconcentrated in the growth cone, and �-tubulin(green), which is at higher levels in the axon than in thegrowth cone, are distant from each other on the graph.See Table S4 for detailed information.

Nozumi et al. PNAS � October 6, 2009 � vol. 106 � no. 40 � 17213

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

May

23,

202

0

inducing axonal growth was assessed by measuring axonal length(see SI Text; for confirmation of knockdown and specificity, see alsoFig. S1). We selected 68 genes for RNAi treatment and found thatdisruption of 17 of them led to shorter axonal length, by applicationof a stringent nonparametric test, i.e., the Kruskal-Willis test (Fig.3A, Table 1, and Table S8). We categorize the 17 proteins asputative functional growth cone markers (Table 1). Considering theresults of the quantitative immunostaining (Fig. 2), these proteinscan be classified into proteins more concentrated than GAP-43(proteins shown in red in Fig. 2) or similar to GAP-43 (proteinsshown in blue in Fig. 2). Of these 17 proteins, there are fivecytoskeletal proteins (Tmod2, Cap1, Cotl1, CapZb, and Sept 2),four membrane trafficking proteins (Pacs1, Stx7, Snap25a, andRtn1), two GTP-binding proteins (Gnai2, Gnao1), two proteinsinvolved in small G protein signaling (Farp2 and Cyfip1), threesignaling adapter proteins (Strap, FABP7, and Crmp1), and onereceptor candidate (Clptm1).

Most of the candidates listed in Fig. 3A revealed previouslyundetected associations with growth cones and were functionallyrelated (Table 1). Sept2, Cap1 (a G-actin-binding protein),Snap25a, and Cyfip1 have been suggested to play roles in growthcone behavior, based on studies in PC12 cells or chick or Drosophilaneurons (21–24), but their precise roles in the mammalian growthcone are not known. For the two GTP-binding proteins (Gnai2,Gnao1) and the regulator (Farp2), involvement in regulation of agrowth cone response to inhibitors has been reported, although itis not known whether they are necessary or indispensable for axonalgrowth (25–27). To the best of our knowledge, none of the other 10proteins were previously reported to be growth cone regulators inmammalian cells or invertebrate model organisms such as C. elegansor Drosophila, although their paralogues may be related to neuritegrowth [for example, syntaxin-1A (28) and syntaxin-3 (29), paral-ogues of Syx7]. We briefly summarize the current informationabout these proteins in Table 1. At most 3 of the 17 proteinsidentified in our study and tested with RNAi have been previouslyimplicated in axonal growth, suggesting that we succeeded inefficient identification of additional molecules involved in growthcone functions. In total, the results of RNAi analysis point to 17proteins with higher or similar FI ratios than GAP-43, making themstrong candidates for novel neuronal growth-associated proteins(nGAPs; refs. 4 and 30).

The growth cone is comprised of morphologically and function-ally distinct regions referred to as the central (C) and peripheral (P)regions (31). The C region is enriched in vesicles and microtubulesand is probably involved in membrane expansion for axonal growth.The P region is enriched in actin filaments and probably generatesmotive force. Using tubulin as a marker for the C region (see Fig.S3), we classified the marker proteins into four groups: group I,localized predominantly in the P region; group II, detected in bothregions; group III, localized predominantly in the C region; andgroup IV, specifically localized in the C region (Fig. 3B; see alsoFigs. S3 and S4).

The patterns of localization we observed were somewhat differ-ent from what might have been predicted. For example, Pacs1 hasbeen reported to be involved in organelle sorting but in axons, Pacs1localizes to the P region, where F-actin is enriched (Fig. 3B). Aputative soluble adapter protein, Strap, which has been reported asdownstream of TGF-� signaling (32), and Clptm1, an unknowntransmembrane protein (33), are also localized in the P region. Incontrast, Sept 2 is detected near the tubulin-positive region despitethe fact that it was previously reported to be found in the P domainin PC12 cells (21). Only Rtn1, an ER protein and putative mem-brane trafficking regulator (34), was specifically localized in the Cregion, making it potentially useful as a C-region marker.

DiscussionThe growth cone is responsible for axon guidance and synaptogen-esis. Thus, understanding growth cone functions at the molecularlevel will contribute to our overall understanding of how neuralnetworks form and are maintained (35). Using a proteomic ap-proach, we identified �1,000 proteins likely to be components ofmammalian growth cones, including actin-associated proteins thatmay be involved in axon growth and motility (Figs. 1 and 2, TableS1, Table S2, Table S3, and Fig. S4). The results of immunostainingsuggest that several of the proteins we identified will serve as usefulmarkers for growth cones in the mammalian brain and validated ourapproach (Fig. 2). Most of the proteins we identified were notpreviously reported as growth cone-associated in mammalian ormodel systems. Our ability to identify such a large number ofputative growth cone marker proteins seems remarkable, particu-larly given that previous to this study, and despite a decade of workby many laboratories, there has been only one marker available forstudies (namely, GAP-43) (7). The results of our analysis shouldhelp to advance our understanding of the molecular machineryunderlying growth cones and their functions (36).

120A

20

40

60

80

100A

xon

al le

ng

th (%

)

113 138 64 75 74 73 410 16969 28 45 55 72 78 30 322 3953560

B

113 138 64 75 74 73 410 16969 28 45 55 72 78 30 322 395356

Fig. 3. Identification of nGAPs by RNAi. (A) Genes affecting axonal growth.Roles for the identified proteins in growth cone activity were assayed bylooking for RNAi-induced reduction of axonal growth. RNAi was performed asdescribed (45). The eight proteins to the left of the hatched line have FI ratiossuperior to GAP-43 as judged statistically by using the Kruskal-Willis test (red;also shown in red in Fig. 2), and the nine proteins to the right of the hatchedline have FI ratios similar to that of GAP-43 as judged by the same test. Theaxonal lengths of EGFP-positive neurons (no siRNA) were used as the control(blue; also shown in blue in Fig. 2). All siRNAs (except GFP) have P � 0.002 inWilcoxon rank-sum test (vs. control) (46). The data are shown as mean � SEM.The number of neurons measured is shown in at the bottom of each bar. (B)Classification of candidate growth cone marker proteins by immunolocaliza-tion. We defined the C and P domains of a growth cone by using quantitativeanalysis of immunostaining images (diagram at top; see SI Text and Fig. S2).We classified the proteins into four groups: group I (ex. Pacs1), predominantlylocalized in the P region (C [dlt ]P); group II (ex. Syx7), detected in both the Cand P regions (C � P); group III (ex. Gnai2), predominantly localized to the Cregion (C �� P); and group IV (ex. Rtn1), specifically localized in the C region(C). In each case, the left diagram is summary of a typical protein distributionfor each group. Immunofluorescence micrographs of anticandidate proteinantibodies detection in cultured rat cortical neurons are shown in A. Magentaand green show antigen protein and �-tubulin views, respectively. The farviews show the merged image. Three groups (groups I-III) are also shown in thelegend for Fig. S4. Note that �-tubulin is primarily detected in the axon,although it is also detectable in the central region of the growth cone. SeeTable 1 for a detailed characterization of each protein and abbreviatednames. (Scale bars: 10 �m.)

17214 � www.pnas.org�cgi�doi�10.1073�pnas.0904092106 Nozumi et al.

Dow

nloa

ded

by g

uest

on

May

23,

202

0

The ability to follow up on initial results with RNAi-based geneknockdown allowed us to rapidly categorize a subset of 17 proteinsas functionally relevant to axonal growth, indicating that proteomicapproaches offer useful alternatives for learning about mammaliangrowth cones instead of or complementing genetic approaches inmodel organisms, and conventional research methods focused on asmall number of proteins. A perhaps surprising and certainlyinteresting result of our study is that most of the proteins weidentified had not previously been implicated in axonal growth inmammals or axon pathfinding in well-studied model organisms,despite the fact that the proteins we identified have been evolu-tionarily conserved (Table 1). The results suggest that a largenumber of genes and a wide variety of biochemical functions arerelevant to axonal growth. For example, among the proteins weidentified, some have been implicated in membrane trafficking,

including Syx7, Rtn1, and Pacs1, suggesting a novel mechanism ofaxonal growth underlying vesicular transport, as has been shown forDrosophila dendrites (37). Combining this finding with the resultsreported here, the list of functional markers for growth cones hasbeen expanded �10-fold, which should have a profound impact onthe ability to study growth cone functions.

The role of some molecules implicated in axon guidance, such asnetrin and semaphorins, has been characterized in the mammalianbrain. These sometimes act as chemoattractants, but in other cases,the same molecules act as chemorepellants (35), which indicatesthat external guidance molecules alone cannot determine the fateof a growth cone; instead, the specific type of response appears tobe determined by the activity of intrinsic signaling pathways. In thiscontext, it is easy to understand the importance of learning whatmolecules are present in growth cones.

Table 1. Novel candidates for nGAPs

Abbreviatedname

Localizationin growth

cone

National Centerfor

BiotechnologyInformation

RefSeq mRNA no.D. melanogaster(FlyBase gene ID)

C. elegans (WormBasegene ID) Putative functions

Tmod2 I NM 031613 FBgn0082582; tmod(tropomodulin)

WBGene00006581;unc-94/tmd-1

Neuron-specific isoform oftropomyosin, blocks theelongation and depolymerizationof actin filaments

Pacs1 I NM 134406 FBgn0020647; KrT95D WBGene00044077;tag-232

Involved in the localization oftrans-Golgi network

Rtn1 IV NM 053865 FBgn0053113; Rtnl1(reticulon1)

WBGene00004336; ret-1 Associated with the endoplasmicreticulum and are involved inmembrane trafficking

Snap25 III NM 030991 FBgn0011288; snap WBGene00004364; ric-4 t-SNARE protein; involved invesicular fusion

Stx7 II NM 021869 FBgn0033583; Syx7 WBGene00009478;F36F2.4

A SNARE protein mediating fusionof late endosomes

Gnai2 III NM 031035 FBgn0001104; G-ia65A WBGene00001648; goa-1 Gi family protein; heterotrimeric Gprotein alpha

Gnao1 III NM 017327 FBgn0001122; Goa47A WBGene00001648; goa-1 Go protein; heterotrimeric Gprotein �

Fabp7 III NM 030832 FBgn0037913; CG6783 WBGene00002258; lbp-6 Brain-type fatty acid-bindingprotein

Cotl1 II NM 001108452 FBgn0030955; CG6891 WBGene00010664;K08E3.4

Essential eukaryotic actinregulatory proteins

Cap1 II NM 022383 FBgn0028388; capulet WBGene00000294; cas-1 G-actin-binding; promotescofilin-induced actin dynamics

Capzb II NM 001005903 FBgn0011570; cpb (cappingprotein beta)

WBGene00000293; cap-2 F-actin capping protein

Sept2 III NM 057148 FBgn0011710; Sep1(Septin-1) WBGene00006795; unc-61 Cytoskeletal componentinteracting with actin-basedstructures

Strap II NM 001011969 FBgn0034876; wmd (wingmorphogenesis defect)

WBGene00001232; eif-3.I WD domain protein in TGF-�signaling

Clptm1 II NM 001106232 FBgn0031590; CG3702 WBGene00016469;C36B7.6

A transmembrane protein with anunknown function

Cyfip1 II NM 001107517 FBgn0038320; sra-1 (specificallyRac1-associated protein 1)

WBGene00001579; gex-2 Rac1-associated protein; link Rac toactin assembly drivinglamellipodia formation

Crmp1 III NM 012932 FBgn0023023; Crmp WBGene00000963; unc-33 May play a role in neuronalplasticity by transduction ofsignals from semaphorins

Farp2 II NM 001107287 FBgn0051536; Cdep WBGene00001490; frm-3(FERM domain family)

Rho GEF protein; involved insemaphorin-signaling

For the proteins listed here, siRNAs directed against the corresponding genes result in significant reduction of axonal length as compared with the control(see Fig. 3A). Markers detected at higher levels than GAP-43 are the first seven shown; the rest were detected at similar levels (Fig. 3A; see also Fig. 2). Thelocalization of each protein in growth cones is as shown in Fig. 3B (see also Fig. S4). Homologues in C. elegans or D. melanogaster are indicated (WormBase orFlyBase ID numbers and gene symbols).

Nozumi et al. PNAS � October 6, 2009 � vol. 106 � no. 40 � 17215

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

May

23,

202

0

In conclusion, we have performed a large-scale identification ofgrowth cone proteins that provides an important starting point forfurther investigation in mammalian axon development, growth, andregeneration (1–4, 38). Genomic microdeletion of Cyfip1, one ofour candidates, has been suggested to be associated with humanschizophrenia (37), further suggesting that our lists might help toprovide a molecular foothold for study of psychiatric disorders based ondisruption of neuronal developmental (38–40). The availability of along list of candidate genes that may be related to growth conefunctions, along with a large set of new marker proteins for thegrowth cone, provides an important resource for further investiga-tion.

MethodsProteomic Analysis. The GCP and GCM fractions were prepared from postnatalday 1–2 rat forebrains, and adult synaptosmal fractions were prepared fromyoung adult rat cortices by subcellular fractionation (41, 42). Both fractions wereS-carbamoylmethylated and digested with trypsin, and the digests were sub-jected to direct nanoflow 2D LC-MS/MS for protein identification (43, 44). Exper-imental conditionsanddataprocessingwereasdescribed (refs. 15,16,and47; seeSI Text). The GCP, GCM, and synaptosomal proteins identified in this work arelisted in Table S1, Table S2, and Table S3.

Cell Culture and Immunostaining. Rat cortical neurons [embryonic day 19 (E19)]were cultured on glass chamber slides for 3 days on polyethylenimine in thepresence of neurobasal medium containing 0.5 mM glutamine and fixed with2.5% glutaraldehyde. For RNAi, dispersed neurons were plated on poly-L-lysine-coated four-well chamber slides made of Permanox (Nunc) and fixed by

using 4% paraformaldehyde (see SI Text). For double-labeling via immuno-fluorescence, the primary antibody was mixed with anti-�-tubulin antibody,and then with a secondary antibody conjugated with Alexa Fluor-488 or -568.

Quantitation of Fluorescence Intensities. An Axiovert 200 microscope (Zeiss) withan ORCA-ER camera (Hamamatsu) and LD Acroplan lens (40 � 0.6 NA) was usedto collect image data. From each image (1,344 � 1,024 pixels), a square growthcone or axon area was digitally excised (100 � 100 pixels), and the fluorescentintensity was calculated per pixel by using ImageJ. Double-labeling with anti-�-tubulin and another antibody, followed by immunofluorescence detection, wasused to determine the subcellular localization of proteins identified in this study.

Functional Analysis Using RNAi. RNAi experiments were done as described (ref. 45;see also SI Text). Chemically synthesized siRNAs (final concentration, 83.3 nM)were applied to cortical neurons derived from E19 ‘‘green rat’’ (SD-Transgenic ratexpressing CAG-EGFP; Japan SLC), together with 83.3 nM siRNA against EGFP(Ambion) by using Lipofectamine 2000 (Invitrogen) just after the neurons weredissociated and plated. After 72 h, we measured axon lengths for neurons inwhich EGFP was not detectable. Any siRNA-dependent decreases in target geneexpression were quantified by using an antibody against the target protein, withanti-GFP (MBL) included as an additional control (Ambion).

See SI Text for additional details.

ACKNOWLEDGMENTS. We thank all those who donated antibodies (see TableS6), Prof. K. Akazawa for statistical analysis, and the Proteomics Committee in theIntegrated Brain Research group for support. This work was supported in part byGrants-in-Aid 16015240 and 17023019 for Scientific Research on Priority AreasfromtheMinistryofEducation,Culture, Sports, Science,andTechnologyof Japan(to M.I.) and Project Research Promoting Grants from Niigata University (to M.I.and M.N.).

1. Chilton JK (2006) Molecular mechanisms of axon guidance. Dev Biol 292:13–24.2. GomezTM,ZhengJQ(2006)Themolecularbasis forcalcium-dependentaxonpathfinding.

Nat Rev Neurosci 7:115–125.3. Gogolla N, Galimbeti I, Caroni P (2007) Structural plasticity of axon terminals in the adult.

Curr Opin Neurobiol 17:516–524.4. Zhou F-Q, Snider WD (2006) Intracellular control of developmental and regenerative axon

growth. Philos Trans R Soc London Ser B 361:1575–1592.5. Bean AJ (2006) Protein Trafficking in Neurons (Academic, Burlington, MA).6. Igarashi M, Ohko K (2009) Proteins involved in the presynaptic functions. Handbook of

Neurochemistry and Molecular Neurobiology: Neural Signaling Mechanisms, ed Miko-shiba K (Springer, New York), pp 47–62.

7. Fishman MC (1996) GAP-43: Putting constraints on neuronal plasticity. Perspect DevNeurobiol 4:193–198.

8. Liao L, McClatchy DB, Yates JR, 3rd (2009) Shotgun proteomics in neuroscience. Neuron63:12–26.

9. Liao L, Park SK, Xu T, Vanderklish P, Yates JR, 3rd (2008) Quantitative proteomic analysisof primary neurons reveals diverse changes in synaptic protein content in fmr1 knockoutmice. Proc Natl Acad Sci USA 105:15281–15286.

10. Emes RD, et al. (2008) Evolutionary expansion and anatomical specialization of synapseproteome complexity. Nat Neurosci 11:799–806.

11. Pertz OC, et al. (2008) Spatial mapping of the neurite and soma proteomes reveals afunctional Cdc42/Rac regulatory network. Proc Natl Acad Sci USA 105:1931–1936.

12. Pfenninger KH, Ellis L, Johnson MP, Friedman LB, Somlo S (1983) Nerve growth conesisolated from fetal rat brain: I. Subcellular fractionation and characterization. Cell 35:573–584.

13. Ellis L, Wallis I, Abreu E, Pfenninger KH (1985) Nerve growth cones isolated from fetal ratbrain. IV. Preparation of a membrane subfraction and identification of a membraneglycoprotein expressed on sprouting neurons. J Cell Biol 101:1977–1989.

14. Gordon-Weeks PR (1987) The cytoskeletons of isolated, neuronal growth cones. Neuro-science 21:977–989.

15. Mawuenyega KG, et al. (2003) Large-scale identification of Caenorhabditis elegans pro-teins by multidimensional liquid chromatography-tandem mass spectrometry. J ProteomeRes 2:23–35.

16. Shinkawa T, et al. (2005) STEM: A software tool for large-scale proteomic data analyses. JProteome Res 4:1826–1831.

17. Pal CW, Flunn KC, Bamburg JR (2008) Actin-binding proteins take the reins in growthcones. Nat Rev Neurosci 9:136–147.

18. Liu H, Sadygov RG, Yates JR, 3rd (2004) A model for random sampling and estimation ofrelative protein abundance in shotgun proteomics. Anal Chem 76:4193–4201.

19. Lin AC, Holt CE (2007) Local translation and directional steering in axons. EMBO J 26:3729–3736.

20. Fukata Y, et al. (2002) CRMP-2 binds to tubulin heterodimers to promote microtubuleassembly. Nat Cell Biol 4:583–591.

21. Vega IE, Hsu SC (2003) The septin protein Nedd5 associates with both the exocyst complexand microtubules and disruption of its GTPase activity promotes aberrant neurite sprout-ing in PC12 cells. NeuroReport 14:31–37.

22. Wills Z, etal. (2002)ADrosophilahomologofcyclase-associatedproteins collaborateswiththe Abl tyrosine kinase to control midline axon pathfinding. Neuron 36:611–622.

23. Schenck A, et al. (2003) CYFIP/Sra-1 controls neuronal connectivity in Drosophila and linksthe Rac1 GTPase pathway to the fragile X protein. Neuron 38:887–898.

24. Osen-Sand A, et al. (1993) Inhibition of axonal growth by SNAP-25 antisense oligonucle-otides in vitro and in vivo. Nature 364:445–448.

25. Igarashi M, Strittmatter SM, Vartanian T, Fishman MC (1993) Mediation by G proteins ofsignals that cause collapse of growth cones. Science 259:77–79.

26. Hasegawa Y, et al. (2004) Promotion of axon regeneration by myelin-associated glycop-roteinandNogothroughdivergent signalsdownstreamofGi/G. JNeurosci24:6826–6832.

27. Toyofuku T, et al. (2005) FARP2 triggers signals for Sema3A-mediated axonal repulsion.Nat Neurosci 8:1712–1719.

28. Igarashi M, et al. (1996) Growth cone collapse and inhibition of neurite growth byBotulinum neurotoxin C1: A t-SNARE is involved in axonal growth. J Cell Biol 134:205–215.

29. Darios F, Davletov B (2006) Omega-3 and omega-6 fatty acids stimulate cell membraneexpansion by acting on syntaxin 3. Nature 440:813–817.

30. Chen Z-L, Yu W-m, Strickland S (2007) Peripheral regeneration. Annu Rev Neurosci30:209–233.

31. Grzywa EL, Lee AC, Lee GU, Suter DM (2006) High-resolution analysis of neuronal growthcone morphology by comparative atomic force and optical microscopy. J Neurobiol6:1529–1543.

32. DattaPK,ChytilA,GorskaAE,MosesHL(1998) IdentificationofSTRAP,anovelWDdomainprotein in transforming growth factor-� signaling. J Biol Chem 273:34671–34674.

33. Yoshiura K, et al. (1998) Characterization of a novel gene disrupted by a balancedchromosomal translocation t(2;19)(q11.2;q13.3) in a family with cleft lip and palate.Genomics 54:231–240.

34. Steiner P, et al. (2004) Reticulon 1-C/neuroendocrine-specific protein-C interacts withSNARE proteins. J Neurochem 89:569–580.

35. CharronF,Tessier-LavigneM(2005)Novelbrainwiringfunctions for classicalmorphogens:A role as graded positional cues in axon guidance. Development 132:2251–2262.

36. Deuel TA, et al. (2006) Genetic interactions between doublecortin and doublecortin-likekinase in neuronal migration and axon outgrowth. Neuron 49:41–53.

37. Ye B, et al. (2007) Growing dendrites and axons differ in their reliance on the secretorypathway. Cell 130:717–729.

38. David S, Lacroix S (2003) Molecular approaches to spinal cord repair. Annu Rev Neurosci26:411–440.

39. Stefansson H, et al. (2008) Large recurrent microdeletions associated with schizophrenia.Nature 455:232–236.

40. Lewis DA, Sweet RA (2009) Schizophrenia from a neural circuitry perspective: Advancingtoward rational pharmacological therapies. J Clin Invest 119:706–716.

41. Phelan P, Gordon-Weeks PR (1997) Isolation of synaptosomes, growth cones, and theirsubcellular components. Neurochemistry: A Practical Approach, eds Turner AJ, BachelardHS (Oxford Univ Press, New York), 2nd Ed, pp 1–38.

42. Igarashi M, Tagaya M, Komiya Y (1997) The SNARE complex in growth cones: Molecularaspects of the axon terminal development. J Neurosci 17:1460–1470.

43. Natsume T, et al. (2002) A direct nanoflow liquid chromatography-tandem mass spec-trometry system for interaction proteomics. Anal Chem 74:4725–4733.

44. IsobeT,YamauchiY,TaokaM,TakahashiN(2003)Automatedtwo-dimensionalLC-MS/MSfor large-scale protein analysis. Proteins and Proteomics, ed Simpson RJ (Cold SpringHarbor Lab Press, Cold Spring Harbor, NY), pp 869–876.

45. Lu J, Nozumi M, Fuji H, Igarashi M (2008) A novel method for RNA interference in theneuron using enhanced green fluorescence protein (EGFP)-transgenic rat. Neurosci Res61:219–224.

46. Petrie A (1987) Lecture Notes of Medical Statistics (Blackwell, Oxford, UK), 2nd Ed.47. Taoka M, et al. (2004) Only a small subset of the horizontally transferred chromosomal

genes in Escherichia coli are translated into proteins. Mol Cell Proteomics 3:780–787.

17216 � www.pnas.org�cgi�doi�10.1073�pnas.0904092106 Nozumi et al.

Dow

nloa

ded

by g

uest

on

May

23,

202

0