Identification of functional marker proteins in the mammalian growth cone

Abstract

Identification of proteins in the mammalian growth cone has the potential to advance our understanding of this critical regulator of neuronal 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 945 proteins present in developing rat forebrain growth cones, including highly abundant, membrane-associated and actin-associated proteins. Almost 100 of the proteins appear to be highly enriched in the growth cone, as determined by quantitative immunostaining, and for 17 proteins, the results of RNAi suggest a role in axon growth. Most of the proteins we identified have not previously been implicated in axon growth and thus their identification presents a significant step forward, providing marker proteins and candidate neuronal growth-associated proteins.

Neuronal growth cones execute important steps in neural wiring, including axonal growth and pathfinding, and accurate synaptogenesis (1, 2). Whether or not an injured axon in the adult neuron regenerates or degenerates depends on surrounding factors that either maintain or inhibit formation of growth cone-like structures, consistent with a critical role for growth cones in neural plasticity and repair of the adult brain (3, 4). It follows, then, that identifying the molecular basis of growth cone behavior will be critical to understanding the cellular mechanisms of higher brain functions. A significant barrier to this understanding, however, is that little is known about the molecular makeup of the mammalian growth cone.

Classical genetic approaches to identifying key players in brain function have been informative in model systems such as Caenorhabditis elegans and Drosophila. For example, identification of mutations in specific C. elegans and Drosophila genes contributed greatly to the discovery and functional characterization of axon guidance molecules in the mammalian brain, such as netrin and semaphorins (1, 2). But these are not the only useful tools with which to dissect growth cone functions, and they have significant limitations in terms of their ability to further our understanding of the complete picture of the molecular machinery that controls mammalian growth cones (1, 2). Among these limitations are that similar studies are not feasible in mammals and mammalian growth cone functions are thought to be much more complicated than growth cone functions in C. elegans and Drosophila. Additionally, the molecular redundancy involved in growth cone functions in the mammalian CNS is likely to be much larger than in model organisms.

Despite the knowledge gap, however, we know of no previous report that applies a systematic approach to identification of mammalian growth cone proteins. However, cell biological studies combined with pharmacological tools to detect second messengers (Ca²⁺ or cyclic nucleotides) or the cytoskeleton (i.e., F-actin or microtubules) have revealed some of the signaling pathways involved 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 or relationships among signaling pathways, at least in part because of an insufficient understanding of what key molecules are important for function (1, 2).

Indeed, even what proteins might be present, i.e., molecular markers of the growth cone, is a relatively unexplored territory, particularly as compared with our understanding of synaptic molecular marker proteins. In the case of adult synapses, a large number of marker proteins localized to various sublocations in the synapse are known (5, 6). These include proteins localized to synaptic vesicles, presynaptic and postsynaptic membranes, active zones, and postsynaptic density (5, 6).

Without a doubt, identification of synaptic marker proteins has markedly enriched our knowledge of the molecular machinery underpinning synaptic structures and functions (5, 6). This stands in contrast to the very small amount of marker protein information available for the growth cone. Indeed, GAP-43 (growth-associated protein 43-kDa; neuromodulin) (7) is the only previously identified functional molecular marker of growth cones (ideally, a “functional molecular marker” would be a protein that is both highly concentrated in the growth cone and involved in axon growth). GAP-43 is concentrated in the growth cone, highly expressed, transported when a damaged axon can regrow, and involved in sprouting (7). Although it is concentrated in the growth cone, GAP-43 is also detectable in the axon (7). Given the severely limited amount of information about functional markers concentrated in or localized to growth cones, it follows that identification of putative novel functional molecular markers of the mammalian growth cone would be extremely valuable to further study.

To help gain a systemwide understanding of the molecular components of growth cones and identify novel molecular marker candidates, we introduced proteome-scale approaches that have been used successfully to identify large numbers of proteins present in other specific cells or tissues, thus contributing to a more global understanding of the functions of those cells or tissues (8–11). We succeeded in identifying 945 species of proteins in growth cone particle (GCP) and/or a growth cone membrane membrane (GCM) (12–14). By combining the results of immunolocalization and RNAi studies with proteomics, we provide evidence that 17 of the proteins we identified are highly concentrated in the growth cone area and regulate axonal growth, concluding that they are unique functional molecular markers of the growth cone.

Results

Large-Scale Identification of Proteins in Rat Brain Growth Cones.

Our first goal was to identify a large number of proteins expressed in growth cones, including proteins common to many cell types and proteins involved in growth cone-specific functions. To do this, we first separated proteins from the developing rat forebrain via subcellular fractionation to obtain a GCP fraction (Fig. 1A). We then obtained a GCM via hypotonic treatment of the GCP fraction (see details in SI Text) with the goal of identifying minor membrane proteins in the growth cone. Subsequently, we used multidimensional liquid chromatography-tandem mass spectrometry (LC/MS/MS) to identify unique proteins in the sample, because the method has been shown to be suitable to large-scale protein identification (15, 16). The method proved powerful in this study as well, with a total of 945 species of proteins identified in GCP and GCM (Fig. 1B). Because <50 of the proteins were previously known to be expressed in the mammalian growth cone, identification of this large number of proteins provides a wealth of molecular information about mammalian growth cones (2, 4, 17). For comparison, we also analyzed adult synaptosomes, the counterpart of the growth cone, and identified 1,407 species of synaptosomal proteins (i.e., twice as many as we found for the GCP and GCM). Approximately 65% of the synaptosomal proteins we identified are not found in the GCP or the GCM sets (Fig. 1C), which may be because a large number of proteins are newly synthesized and added to synaptic components for synaptic transmission after synaptogenesis. We succeeded in identifying 96 and 141 proteins in the GCP or GCM, respectively, that were not found in the synaptosomes (Fig. 1C).

Fig. 1.

Proteomic analysis of GCPs and adult synaptosomes. (A) Electron micrograph of GCPs. The GCP fraction was prepared by using the method described by Gordon-Weeks (14) (see SI Text). The criterion for a GCP was a particle with a diameter of ≈1–2 μm and the presence of small clear vesicles. (Scale bar: 10 μm.) (B) GCP proteins and proteins found in a membrane subfraction of GCP (GCM). In total, 629 GCP and 592 GCM proteins were identified, with 276 proteins common to both. Note that 316 proteins were identified only in the GCM subfraction. (C) Comparison of the GCP/GCM proteome with the proteome of adult synaptosomes. A total of 1,407 synaptosomal proteins were identified, for about twice as many as were identified in GCP or GCM. (D) Categorization of GCP, GCM, and synaptosomal proteins. Categories are color-coded as follows: metabolic enzymes, violet; cytoskeletal proteins, navy; signaling (except GTP-binding or phosphorylation), blue; proteins involved in guanine nucleotide cycling (including GTP-binding proteins), cyan; ion transport/channels, green; membrane traffic, lime; kinases or phosphatases, yellow; receptors, orange; chaperone, brown; proteasome/ubiquitination-related proteins, maroon; cell adhesion proteins, red; ribosomal, magenta; proteins involved in translation, olive; miscellaneous proteins (including organelle-specific or undefined), gray. Note that the ratio of cell adhesion molecules, receptors, and transporters/channels is higher in GCM than in GCP, consistent with enrichment of membrane components in GCM. See Table S1, Table S2, and Table S3 for a detailed report. (E) The major proteins of GCP or GCM as compared with synaptosomal proteins as indicated by the number of identified peptides. Among proteins for which 12 or more peptides were identified, we chose the subset identified as peptides at least twice as many times in GCP or in GCM than in synaptosomes. The number of peptide identifications is shown. (Upper) GCP > GCM. (Lower) GCM > GCP. Color-coded by functional category as in D. Note that the number of identified peptides is an indicator of relative protein levels.

Proteins Identified by Proteomics Analysis.

We next used bioinformatics analysis to categorize the proteins into functionally related groups (Fig. 1 C–E and Table S1, Table S2, and Table S3). We included synaptosome data in our analysis, because the presynaptic axon terminal is the adult counterpart of the growth cone. The proteins were in the following functional categories: those required for cytoskeletal reorganization, involved in vesicular trafficking, and related to signal transduction, including protein kinases, phosphatases, and G proteins (2, 4). As expected, previously identified key 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., not previously reported as GCPs) (2, 4, 15) (Fig. 1 D and E; see Table S1, Table S2, and Table S3).

As expected, the GCP fraction contained both cytosolic and membrane-bound proteins; the GCM fraction was enriched for membrane and membranous organelle-associated proteins; and the set of proteins found in the GCP fraction but not the GCM fraction was enriched for cytosolic proteins (Fig. 1D). As shown in Fig. 1D, the set of membrane-associated proteins we identified included cell adhesion molecules (2.6% in GCP and 6.5% in GCM), receptors and receptor-like membrane proteins (1.0% in GCP and 8.7% in GCM), and transporters/channels (6.9% in GCP and 10.3% in GCM). Components of ubiquitin proteasome were detected only in the GCP fraction (Fig. 1D). Synaptosomes contain a larger numbers of metabolic enzymes, but the percentages of ribosomal proteins and local protein translation proteins detected in synaptosomes were lower than in GCP (Fig. 1D).

In our shotgun proteomic analysis, we tentatively assumed that the number of times a given peptide is identified correlates with the abundance of the protein in cells (18). In Fig. 1E, we report those proteins for which peptides were independently identified 12 or more times in GCP or GCM (together, these comprise 50% of the peptides identified in GCP or GCM), and for which the abundance in GCP or GCM was 2-fold or more compared with levels in synaptosomes. These appear to be the most highly abundant proteins in the GCP and GCM fractions, and thus we defined proteins appearing 12 or more times as “major proteins” in the GCP or in GCM fractions. These include cytoskeletal components (tubulins, the microtubule-associated protein MAP1B, dynein heavy chains, and myosin heavy chains), collapsin response mediator proteins (CRMPs), catenins, 14-3-3 proteins, and Gα proteins (namely, Gq, Gi1, and Gi2). The known or predicted biochemical and/or biological activities of the major proteins are consistent with functional relevance in growth cones. Moreover, identification of the translation factor elongation factor 1α is consistent with the finding that local protein synthesis is important for growth cone behavior (19). Additionally, identification of CRMP family members (CRMP4b, CRMP1, and CRMP5; Fig. 1E) as the most abundant GCP proteins in growth cones correlates well with a report that CRMP2 acts as a tubulin adapter protein and is involved in axon formation (20).

Verification of Growth Cone Localization Suggests a Negligible False-Positive Rate.

We next wanted to determine the specificity of our proteomics approach,; i.e., to determine the false-positive rate. Generally, a comparison of a given fraction should be made to another biochemical fraction. However, no fractions from the developing brain can be prepared with comparable purity to the GCP or GCM fractions. Thus, to facilitate generation of a comparison dataset, we performed systematic immunodetection of a subset 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 and molecular chaperones so that we could instead focus on proteins that 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 rat cortical neurons all of the proteins we tested are detectable in the growth cone area (Fig. 2 and Table S4). Indeed, no proteins we tested could be detected in other axonal regions but not in the growth cone, suggesting a very low false-positive rate and validating our overall approach. In addition, our proteomic data in GCP or GCM contained no transcription factors, suggesting that contamination with nuclear components is also negligible (Fig. 1D, Table S1, and Table S2).

Fig. 2.

Immunofluorescence quantification of GCP. Horizontal axis, FI ratio (growth cone/distal axon), Longitudinal 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 indicates that a given protein is evenly distributed in the distal axon versus the growth cone). The FI ratio for GAP-43 (Gap43; 1.315) is shown as a blue dotted line. Proteins in the upper right region of the graph are more concentrated in the growth cone than in the axon. Many of these are actin-binding proteins and proteins involved in vesicular trafficking. The FI ratio of each protein was tested by using the Kruskal-Willis statistical test (46) with GAP-43 as the comparison point. The proteins 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, and black, respectively). Note that as expected, F-actin (as detected with rhodamine phalloidin; in red), which is concentrated in the growth cone, and β-tubulin (green), which is at higher levels in the axon than in the growth cone, are distant from each other on the graph. See Table S4 for detailed information.

Proteins More Concentrated in the Growth Cone Than in the Distal Axon.

We next sought to determine the extent to which the proteins identified in our study are specific or locally specific to the growth cone. To do this, we compared the distributions of the proteins in growth cones with their distribution in distal axons. To the extent that the method is quantitative, we were also able to compare the relative concentrations of proteins in growth cones versus distal axons (Fig. 2 and Table S4). We defined the growth cone accumulation index as the ratio of fluorescence intensity (FI) in the growth 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). The growth cone accumulation index is an indication of the relative level of protein accumulation in the growth cone. By applying the statistical Wilcoxon rank-sum test to our results, facilitating strict classification of the examined proteins as compared with GAP-43, our quantification of the systematic immunostaining approach using this index revealed that as many as 69 proteins identified by proteomics were detected at higher levels in the growth cone area than in the distal axon. These proteins appear to be at much higher levels in growth cones than the currently established growth cone marker protein GAP-43 (7) (Fig. 2, shown in red). We also found that for 33 proteins the statistical error areas overlap with that of GAP-43, thus we categorize this set of proteins as concentrated in the growth cone to a similar degree as is GAP-43 (Fig. 2, shown in blue).

RNAi Analysis Reveals Relevance to Axon Length and Functional Markers of the Growth Cone.

We used an RNAi-based approach to test the roles of marker protein candidates in axonal growth. Activity-inducing axonal growth was assessed by measuring axonal length (see SI Text; for confirmation of knockdown and specificity, see also Fig. S1). We selected 68 genes for RNAi treatment and found that disruption of 17 of them led to shorter axonal length, by application of a stringent nonparametric test, i.e., the Kruskal-Willis test (Fig. 3A, Table 1, and Table S8). We categorize the 17 proteins as putative functional growth cone markers (Table 1). Considering the results of the quantitative immunostaining (Fig. 2), these proteins can be classified into proteins more concentrated than GAP-43 (proteins shown in red in Fig. 2) or similar to GAP-43 (proteins shown in blue in Fig. 2). Of these 17 proteins, there are five cytoskeletal proteins (Tmod2, Cap1, Cotl1, CapZb, and Sept 2), four membrane trafficking proteins (Pacs1, Stx7, Snap25a, and Rtn1), two GTP-binding proteins (Gnai2, Gnao1), two proteins involved in small G protein signaling (Farp2 and Cyfip1), three signaling adapter proteins (Strap, FABP7, and Crmp1), and one receptor candidate (Clptm1).

Fig. 3.

Identification of nGAPs by RNAi. (A) Genes affecting axonal growth. Roles for the identified proteins in growth cone activity were assayed by looking for RNAi-induced reduction of axonal growth. RNAi was performed as described (45). The eight proteins to the left of the hatched line have FI ratios superior 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 hatched line have FI ratios similar to that of GAP-43 as judged by the same test. The axonal 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 in Wilcoxon 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 immunolocalization. We defined the C and P domains of a growth cone by using quantitative analysis of immunostaining images (diagram at top; see SI Text and Fig. S2). We classified the proteins into four groups: group I (ex. Pacs1), predominantly localized in the P region (C [dlt ]P); group II (ex. Syx7), detected in both the C and P regions (C ≃ P); group III (ex. Gnai2), predominantly localized to the C region (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 distribution for each group. Immunofluorescence micrographs of anticandidate protein antibodies detection in cultured rat cortical neurons are shown in A. Magenta and green show antigen protein and α-tubulin views, respectively. The far views show the merged image. Three groups (groups I-III) are also shown in the legend for Fig. S4. Note that α-tubulin is primarily detected in the axon, although it is also detectable in the central region of the growth cone. See Table 1 for a detailed characterization of each protein and abbreviated names. (Scale bars: 10 μm.)

Table 1.

Novel candidates for nGAPs

Abbreviated name	Localization in growth cone	National Center for Biotechnology Information RefSeq mRNA no.	D. melanogaster (FlyBase gene ID)	C. elegans (WormBase gene ID)	Putative functions
Tmod2	I	NM 031613	FBgn0082582; tmod (tropomodulin)	WBGene00006581; unc-94/tmd-1	Neuron-specific isoform of tropomyosin, blocks the elongation and depolymerization of actin filaments
Pacs1	I	NM 134406	FBgn0020647; KrT95D	WBGene00044077; tag-232	Involved in the localization of trans-Golgi network
Rtn1	IV	NM 053865	FBgn0053113; Rtnl1 (reticulon1)	WBGene00004336; ret-1	Associated with the endoplasmic reticulum and are involved in membrane trafficking
Snap25	III	NM 030991	FBgn0011288; snap	WBGene00004364; ric-4	t-SNARE protein; involved in vesicular fusion
Stx7	II	NM 021869	FBgn0033583; Syx7	WBGene00009478; F36F2.4	A SNARE protein mediating fusion of late endosomes
Gnai2	III	NM 031035	FBgn0001104; G-ia65A	WBGene00001648; goa-1	Gi family protein; heterotrimeric G protein alpha
Gnao1	III	NM 017327	FBgn0001122; Goa47A	WBGene00001648; goa-1	Go protein; heterotrimeric G protein α
Fabp7	III	NM 030832	FBgn0037913; CG6783	WBGene00002258; lbp-6	Brain-type fatty acid-binding protein
Cotl1	II	NM 001108452	FBgn0030955; CG6891	WBGene00010664; K08E3.4	Essential eukaryotic actin regulatory proteins
Cap1	II	NM 022383	FBgn0028388; capulet	WBGene00000294; cas-1	G-actin-binding; promotes cofilin-induced actin dynamics
Capzb	II	NM 001005903	FBgn0011570; cpb (capping protein beta)	WBGene00000293; cap-2	F-actin capping protein
Sept2	III	NM 057148	FBgn0011710; Sep1(Septin-1)	WBGene00006795; unc-61	Cytoskeletal component interacting with actin-based structures
Strap	II	NM 001011969	FBgn0034876; wmd (wing morphogenesis defect)	WBGene00001232; eif-3.I	WD domain protein in TGF-β signaling
Clptm1	II	NM 001106232	FBgn0031590; CG3702	WBGene00016469; C36B7.6	A transmembrane protein with an unknown function
Cyfip1	II	NM 001107517	FBgn0038320; sra-1 (specifically Rac1-associated protein 1)	WBGene00001579; gex-2	Rac1-associated protein; link Rac to actin assembly driving lamellipodia formation
Crmp1	III	NM 012932	FBgn0023023; Crmp	WBGene00000963; unc-33	May play a role in neuronal plasticity by transduction of signals from semaphorins
Farp2	II	NM 001107287	FBgn0051536; Cdep	WBGene00001490; frm-3 (FERM domain family)	Rho GEF protein; involved in semaphorin-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). The localization 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 or FlyBase ID numbers and gene symbols).

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

The growth cone is comprised of morphologically and functionally distinct regions referred to as the central (C) and peripheral (P) regions (31). The C region is enriched in vesicles and microtubules and is probably involved in membrane expansion for axonal growth. The P region is enriched in actin filaments and probably generates motive 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 both regions; group III, localized predominantly in the C region; and group IV, specifically localized in the C region (Fig. 3B; see also Figs. S3 and S4).

The patterns of localization we observed were somewhat different from what might have been predicted. For example, Pacs1 has been reported to be involved in organelle sorting but in axons, Pacs1 localizes to the P region, where F-actin is enriched (Fig. 3B). A putative soluble adapter protein, Strap, which has been reported as downstream of TGF-β signaling (32), and Clptm1, an unknown transmembrane protein (33), are also localized in the P region. In contrast, Sept 2 is detected near the tubulin-positive region despite the fact that it was previously reported to be found in the P domain in PC12 cells (21). Only Rtn1, an ER protein and putative membrane trafficking regulator (34), was specifically localized in the C region, making it potentially useful as a C-region marker.

Discussion

The growth cone is responsible for axon guidance and synaptogenesis. Thus, understanding growth cone functions at the molecular level will contribute to our overall understanding of how neural networks form and are maintained (35). Using a proteomic approach, we identified ≈1,000 proteins likely to be components of mammalian growth cones, including actin-associated proteins that may be involved in axon growth and motility (Figs. 1 and 2, Table S1, Table S2, Table S3, and Fig. S4). The results of immunostaining suggest that several of the proteins we identified will serve as useful markers for growth cones in the mammalian brain and validated our approach (Fig. 2). Most of the proteins we identified were not previously reported as growth cone-associated in mammalian or model systems. Our ability to identify such a large number of putative growth cone marker proteins seems remarkable, particularly given that previous to this study, and despite a decade of work by many laboratories, there has been only one marker available for studies (namely, GAP-43) (7). The results of our analysis should help to advance our understanding of the molecular machinery underlying growth cones and their functions (36).

The ability to follow up on initial results with RNAi-based gene knockdown allowed us to rapidly categorize a subset of 17 proteins as functionally relevant to axonal growth, indicating that proteomic approaches offer useful alternatives for learning about mammalian growth cones instead of or complementing genetic approaches in model organisms, and conventional research methods focused on a small number of proteins. A perhaps surprising and certainly interesting result of our study is that most of the proteins we identified had not previously been implicated in axonal growth in mammals or axon pathfinding in well-studied model organisms, despite the fact that the proteins we identified have been evolutionarily conserved (Table 1). The results suggest that a large number of genes and a wide variety of biochemical functions are relevant to axonal growth. For example, among the proteins we identified, some have been implicated in membrane trafficking, including Syx7, Rtn1, and Pacs1, suggesting a novel mechanism of axonal growth underlying vesicular transport, as has been shown for Drosophila dendrites (37). Combining this finding with the results reported here, the list of functional markers for growth cones has been expanded >10-fold, which should have a profound impact on the ability to study growth cone functions.

The role of some molecules implicated in axon guidance, such as netrin and semaphorins, has been characterized in the mammalian brain. These sometimes act as chemoattractants, but in other cases, the same molecules act as chemorepellants (35), which indicates that external guidance molecules alone cannot determine the fate of a growth cone; instead, the specific type of response appears to be determined by the activity of intrinsic signaling pathways. In this context, it is easy to understand the importance of learning what molecules are present in growth cones.

In conclusion, we have performed a large-scale identification of growth cone proteins that provides an important starting point for further investigation in mammalian axon development, growth, and regeneration (1–4, 38). Genomic microdeletion of Cyfip1, one of our candidates, has been suggested to be associated with human schizophrenia (37), further suggesting that our lists might help to provide a molecular foothold for study of psychiatric disorders based on disruption of neuronal developmental (38–40). The availability of a long list of candidate genes that may be related to growth cone functions, along with a large set of new marker proteins for the growth cone, provides an important resource for further investigation.

Methods

Proteomic Analysis.

The GCP and GCM fractions were prepared from postnatal day 1–2 rat forebrains, and adult synaptosmal fractions were prepared from young adult rat cortices by subcellular fractionation (41, 42). Both fractions were S-carbamoylmethylated and digested with trypsin, and the digests were subjected to direct nanoflow 2D LC-MS/MS for protein identification (43, 44). Experimental conditions and data processing were as described (refs. 15, 16, and 47; see SI Text). The GCP, GCM, and synaptosomal proteins identified in this work are listed 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 the presence of neurobasal medium containing 0.5 mM glutamine and fixed with 2.5% 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 immunofluorescence, the primary antibody was mixed with anti-α-tubulin antibody, and then with a secondary antibody conjugated with Alexa Fluor-488 or -568.

Quantitation of Fluorescence Intensities.

An Axiovert 200 microscope (Zeiss) with an ORCA-ER camera (Hamamatsu) and LD Acroplan lens (40 × 0.6 NA) was used to collect image data. From each image (1,344 × 1,024 pixels), a square growth cone or axon area was digitally excised (100 × 100 pixels), and the fluorescent intensity was calculated per pixel by using ImageJ. Double-labeling with anti-α-tubulin and another antibody, followed by immunofluorescence detection, was used 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 rat expressing CAG-EGFP; Japan SLC), together with 83.3 nM siRNA against EGFP (Ambion) by using Lipofectamine 2000 (Invitrogen) just after the neurons were dissociated and plated. After 72 h, we measured axon lengths for neurons in which EGFP was not detectable. Any siRNA-dependent decreases in target gene expression were quantified by using an antibody against the target protein, with anti-GFP (MBL) included as an additional control (Ambion).

See SI Text for additional details.