Rufy3, a protein specifically expressed in neurons, interacts with actin-bundling protein Fascin to control the growth of axons

Authors

  • Zhe Wei,

    1. Key Laboratory of Stem Cell Biology, Institute of Health Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences/Shanghai JiaoTong University School of Medicine, Shanghai, China
    2. University of Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Ming Sun,

    1. Key Laboratory of Stem Cell Biology, Institute of Health Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences/Shanghai JiaoTong University School of Medicine, Shanghai, China
    2. University of Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Xinyi Liu,

    1. Department of Pathophysiology, Key Laboratory of Cell Differentiation and Apoptosis of Chinese Ministry of Education, School of Medicine, Shanghai Jiao Tong University, Shanghai, China
    Search for more papers by this author
  • Jian Zhang,

    1. Department of Pathophysiology, Key Laboratory of Cell Differentiation and Apoptosis of Chinese Ministry of Education, School of Medicine, Shanghai Jiao Tong University, Shanghai, China
    Search for more papers by this author
  • Ying Jin

    Corresponding author
    1. Key Laboratory of Stem Cell Biology, Institute of Health Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences/Shanghai JiaoTong University School of Medicine, Shanghai, China
    2. Shanghai Stem Cell Institute, Institute of Medical Sciences, Shanghai JiaoTong University School of Medicine, Shanghai, China
    • Address correspondence and reprint requests to Ying Jin, Institute of Health Sciences, 225 South Chongqing Road, Shanghai, 200025, China. E-mail: yjin@sibs.ac.cn

    Search for more papers by this author

Abstract

For our nervous system to function properly, each neuron must generate a single axon and elongate the axon to reach its target. It is known that actin filaments and their dynamic interaction with microtubules within growth cones play important roles in inducing axon extension. However, it remains unclear how cytoskeletal dynamics is controlled in growth cones. In this study, we report that Rufy3, a RUN domain-containing protein, is a neuron-specific and actin filament-relevant protein. We find that the appropriate expression of Rufy3 in mouse hippocampal neurons is required for the development of a single axon and axon growth. Our results show that Rufy3 specifically interacts with actin filament-binding proteins, such as Fascin, and colocalizes with Fascin in growth cones. Knockdown of Rufy3 impairs the distribution of Fascin and actin filaments, accompanied by an increased proportion of neurons with multiple axons and a decrease in the axon length. Therefore, Rufy3 may be particularly important for neuronal axon elongation by interacting with Fascin to control actin filament organization in axonal growth cones.

image

We propose that Rufy3 may control mouse neuron axon development through its specific interaction with Fascin and Drebrin. Over-expression of Rufy3 (Rufy3 OE) leads to longer axons and expands the distribution of Drebrin to almost the entire growth cone. In contrast, knockdown of Rufy3 (Rufy3 RNAi) results in shortened axons and enhanced the percentage of mutipolar neurons. Moreover, silencing of Rufy3 reduces and restricts the expression of Fascin and F-actin to the edge of the growth cone. These findings provide new insights into the molecular regulation of axonal outgrowth and cell polarization in neurons.

Abbreviations used
bFGF

basic fibroblast growth factor

BME

β-mercaptoethanol

BSA

bovine serum albumin

DIC

differential interference contrast microscope

DMSO

dimethyl sulfoxide

EGF

epidermal growth factor

EGFP

enhanced green fluorescent protein

ESCs

embryonic stem cells

FBS

fetal bovine serum

GST

glutathione-S-transferase

IB

immunoblotting

IP

immunoprecipitation

IRES

internal ribosome entry sites

LIF

leukemia inhibitory factor

Map2

microtubule associated protein 2

MS

mass spectrometry

NEAA

non-essential amino acid

NP-40

nonyl phenoxypolyethoxylethanol

NPCs

neural progenitor cells

PBS

phosphate-buffered saline

PCR

polymerase chain reaction

RNAi

RNA interference

Rufy3 OE

Rufy3 over-expression

SDS–PAGE

dodecyl sulfatesodium salt–polyacrylamide gel electrophoresis

A properly connected and functional nervous system is dependent on the ability of neurons to develop polarity through forming a single axon and multiple dendrites. Knowledge of how the axon forms and elongates is critical for understanding the development of the nervous system. Cultured neurons from the mammalian hippocampus have been widely used to study the regulation of neuronal axon development (Craig and Banker 1994). In culture, these neurons form several morphologically similar neurites before polarization occurs. One of the neurites then grows significantly faster than others and becomes an axon, whereas others become dendrites (Dotti and Banker 1987; Goslin and Banker 1989). At the tip of a growing axon, there is a highly dynamic, sensory-motile structure, specialized for elongation and steering, known as a growth cone. It is known that neuronal growth cones play critical roles in guiding axons to their appropriate targets. Operationally, the growth cone has been divided into three primary domains. The central domain, located at the end of a growing axon, is the most proximal part of the growth cone and is filled with abundant microtubules and vesicles. The peripheral domain is at the most distal area of the growth cone and is motile; it is composed of bundles of actin filaments (F-actin) and fringed with lamellipodia and filopodia. The transitional domain is located at the junction between the central and peripheral domains. It contains overlapping networks of F-actin and microtubules (Lin and Forscher 1993; Geraldo and Gordon-Weeks 2009). A previous study demonstrated that local instability of the actin network in a single growth cone determines whether it becomes an axon or a dendrite, suggesting that the higher rate of actin turnover may underlie neuronal polarization (Bradke and Dotti 1999). Therefore, considerable attention in recent years has been centered on understanding how actin cytoskeleton and its interaction with microtubules are precisely modulated in the growth cone.

There are two kinds of cytoskeletal filaments, F-actin and microtubules, in the growth cone. The actin cytoskeleton, composed of actin polymers and a large variety of associated proteins, is a highly dynamic network (Schmidt and Hall 1998). Actin bundles distributed throughout the lamellipodium extend beyond the leading edge of the growth cone to form the core of filopodia (Bentley and Toroian-Raymond 1986). Although how exactly filopodia form is not entirely clear, some actin-associated proteins have been found in growth cones and are known to play an important role in the regulation of growth cone morphology and behavior. For instance, Fascin, a conserved actin-bundling protein, has been reported to play a role in defining the dendrite morphology in neurons of Drosophila larvae (Nagel et al. 2012), forming filopodia of tumor cells (Sun et al. 2011) as well as forming and maintaining actin bundles in Helisoma neuronal growth cone (Cohan et al. 2001). Moreover, the F-actin side-binding protein, Drebrin, has been shown to localize at the transitional domain of growth cones and is involved in dendritic (Hayashi and Shirao 1999; Mizui et al. 2005) and axonal development (Mizui et al. 2009). Furthermore, Drebrin was reported to inhibit the bundling activity of Fascin. Exploring how these F-actin-binding proteins participate in axon growth and identifying new actin-associated factors are important for our understanding of the molecular mechanism underlying the formation and elongation of neuronal growth cones.

In this study, we characterize the role of Rufy3, a RUN domain-containing protein, in the morphogenesis of neural axons and explore the potential molecular basis underlying its function in primary mouse hippocampal neuron culture. Rufy3 was chosen as the focus of this study because of the reported role of its rat ortholog, named Singar 1, for suppressing surplus axon formation in a PI3K activity-dependent manner in rat embryonic hippocampal neurons (Mori et al. 2007). Nevertheless, it remains unknown whether Rufy3 would exhibit additional roles in neurons besides suppressing surplus axons, and if so, how the function is executed. Here, we report that Rufy3 is exclusively expressed in the neurons of the mouse developing brain. Knockdown of Rufy3 leads to shortened axons in addition to an increased fraction of neurons with multiple axons. Conversely, Rufy3 over-expression results in an increase in the axon length. Moreover, we find that Rufy3 interacts with Fascin and Drebrin, codistributing with F-actin in the axonal growth cone. Silencing of Rufy3 disrupts the normal distribution of Fascin and F-actin. The findings are significant for understanding how neuronal axon formation and growth cone morphogenesis are controlled at a molecular level.

Materials and methods

Animal, cell culture, and drug treatment

Mice of the C57/BL6 strain used in this study were purchased from SLAC, Shanghai, China. All animal procedures were performed according to the guidelines approved by the Shanghai JiaoTong University School of Medicine (20080050) and conformed to the NIH guideline for the use of animals in research. Every effort was made to minimize animal suffering and reduce the number of animals used.

HEK293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (Gibco/Life Technologies, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (Hyclone/ GE Healthcare, Piscataway, NJ, USA), 2 mM l-glutamine, 1 unit/mL penicillin, and 100 μg/mL streptomycin, and passaged every 2 days. After transfection with plasmids for 48 h, cell lysates were collected for coimmunoprecipitation (Co-IP) experiments. Mouse hippocampal and cortical neurons prepared from post-natal day 0 (P0) C57/BL6 mouse embryos were seeded on coverslips coated with poly-d-lysine (Millipore, Temecula, CA, USA) and laminin (Invitrogen/Life Technologies, Grand Island, NY, USA), and cultured in Neurobasal medium (Gibco) supplemented with B27 supplement (Invitrogen), 2 mM l-glutamine (Gibco) and 1% non-essential amino acid (NEAA, Gibco), without a glial feeder layer, as described previously (Ha et al. 2009). The D3 mouse embryonic stem cells (ESCs) were maintained in DMEM supplemented with 10% fetal bovine serum (Hyclone), 1% NEAA, 100 μM β-Mercaptoethanol (Gibco), 2 mM l-glutamine, 1 unit/mL penicillin, 100 μg/mL streptomycin, and 1000 units/mL leukemia inhibitor factor (LIF, Millipore) on mouse embryonic fibroblast (MEF), and passaged using 0.05% trypsin every 3 days, as described previously (Tremml et al. 2008). For cytochalasin D (CytoD, Sigma, St Louis, MO, USA) treatment, hippocampal neurons cultured for 48 h were incubated with 1 μM of CytoD for 2 h before fixation.

DNA constructs

Full-length cDNA of the mouse Rufy3 was amplified by PCR using cDNA from a mouse E10.5 brain as a template. The cDNA was then subcloned into the pCAG-IRES-EGFP plasmid described previously (Lorsbach et al. 2004). For the study of protein interaction, the cDNA was transferred into a Flag-pPy vector as described previously (Li et al. 2009). Full-length cDNA of mouse Drebrin was amplified by PCR using cDNAs from a mouse E18.5 brain as a template and the longest transcript was subcloned into the HA-pPy vector. Full-length cDNA of mouse Fascin, the N terminus of Rufy3, and the C terminus of Rufy3 were amplified in the same way as for Drebrin. For RNA interference, two shRNA duplexes (Thermo Scientific, Wilmington, DE, USA) against Rufy3, shRufy3-1: 5′- AGAAATGGAACGAGTTAAA -3′ and shRufy3-2: 5′- TAAACAACCTTCAGACAAA -3′, were, respectively, cloned into a pLKO.3G vector purchased from the Addgene (named Ri-1# and Ri-2#, respectively), and the empty pLKO.3G vector was used as a control. The sequences of all constructs were verified by DNA sequencing.

Transfection and coimmunoprecipitation

Plasmid transfection was performed using Lipofectamine 2000 (Invitrogen). For Co-IP experiments with exogenously expressed proteins, cell extracts were collected 48 h after transfection and immunoprecipitated by anti-Flag antibody, as described previously (Li et al. 2007). For Co-IP experiments with endogenous proteins, whole-brain extracts from C57/BL6 P0 mice were extracted in Co-IP buffer containing 50 mM Tris-HCl pH 7.5 (Sangon, Shanghai, China), 150 mM NaCl (Sangon), 2 mM EDTA, pH 8.0 (Sangon), 10% glycerol (Sangon) and, 0.5% nonyl phenoxypolyethoxylethanol (NP-40, Sangon), then immunoprecipitated by antibodies against Rufy3, Drebrin or Fascin, respectively.

Mouse ESC neural differentiation

Neural differentiation from D3 mouse ESCs was induced as previously described (Watanabe et al. 2005). ESCs were cultured in suspension in a serum-free medium for 7 days for the formation of embryoid bodies (EBs). The EBs were then plated onto dishes coated with Matrigel and passaged every 3 days. After differentiation for 7–10 days, many NPCs and a few neurons appeared. Subsequently, the number of neurons increased dramatically with the concomitant decrease in the number of NPCs. At 19 days of differentiation, the number of neurons decreased with an increase in the number of glia cells.

Primary neuron differentiation

The brain cortex of the C57/BL6 E14.5 mouse was dissected and digested. The dissociated cells were suspended as neurospheres in the DMEM/F12 medium (Gibco) supplemented with N2 (Invitrogen) and Neurobasal medium plus B27 (N2B27 medium) with bFGF (R&D, 10 ng/mL) and EGF (R&D, 10 ng/mL) for 3 day as described previously (Wu et al. 2001). On the fourth day, the spheres were digested and plated onto Matrigel-coated (BD) dishes in the N2B27 medium without bFGF and EGF for 6 days. Samples were collected at day 0 (the day 3 – neurosphere stage), day 1 (the first day when the cells were plated onto dishes), day 2, and day 6.

Antibodies

To generate the antibody that specifically recognizes Rufy3, we expressed a GST-fused Rufy3 fragment (residues 567-807) and purified the fusion proteins from E. coli BL21 (DE3) to immunize rabbits. Anti-serum was affinity purified as previously reported (Xu et al. 2004). The following antibodies were purchased from commercial sources: α-Tublin from Sigma, Nestin from Millipore, Map2 from Millipore, Tuj1 from Invitrogen, Drebrin from Thermo, Fascin from Epitomics Inc, SMI312 from Abcam (Cambridge, MA, USA), Gapdh from Boster (Wuhan, Hubei, China), enhanced green fluorescent protein (EGFP) from Roche (Indianapolis, IN, USA) and Flag from Sigma.

Immunofluorescence staining, western blot, and microscopic analyses

Cells were fixed with 4% paraformaldehyde and 0.1% glutaraldehyde (Sangon) for 10 min, permeabilized with a solution containing 0.5% Triton X-100 for 3 min, and blocked with 3% bovine serum albumin for 30 min at 25°C. Then, the cells were incubated with antibodies against Rufy3 (1 : 500 dilution), Fascin (1 : 200 dilution), Drebrin (1 : 1000 dilution) or EGFP (1 :500 dilution), respectively. After extensive washing with phosphate-buffered saline for three times, the cells were incubated with FITC-conjugated or CY3-conjugated secondary antibodies (Proteintech, Chicago, IL, USA) at 37°C for 1 h. F-actin was labeled with phalloidin from Invitrogen and the nuclei were counter-stained with 4, 6-diamidino-2 phenylindole. Finally, images of cells were captured (TCS SP5; Leica Microsystems, Wetzlar, Germany). Brain sections were stained as the cells were, but longer fixation times and antibody incubation times were used. Western blot analyses were conducted as described previously (Liao and Jin 2010).

Mass spectrometry

To identify proteins interacting with Rufy3, whole-brain proteins from C57/BL6 P0 mice were extracted using the Co-IP buffer. About 4 mg of brain extract was incubated with Rufy3 antibodies overnight. On the next day, the mixture was incubated with protein A beads for 3 h at 4°C to precipitate Rufy3 and associated proteins. IgG was used as a negative control for the Rufy3 antibody. Endogenous Rufy3 and proteins forming protein complexes with Rufy3 in brain cells were separated by SDS–PAGE and visualized by Coomassie Brilliant Blue staining. The gels were cut into 7–10 pieces for both IgG- and Rufy3 antibody-precipitated samples. Proteins in the gels were then sequenced by mass spectrometry in BIDMC Proteomics Center and Dana Farber/Harvard Cancer Center, Cancer Proteomics Core in Beth Israel Deaconess Medical Center at Harvard Medical School. The mass spectrometry spectra were acquired using the mass spec analysis program called Protein Pilot, version 3.0 reported in a previous study (Shilov et al. 2007). To find out relevant functions and pathways from the identified protein list, the enrichment analysis was carried out using the DAVID website. Firstly, various annotations were extracted automatically by the tool. Then annotation clustering analysis was performed by choosing the three categories of Gene Ontology (GO), including Biological Process (BP), Cellular Component (CC), and Molecular Function (MF), which could reveal common functions among the proteins. Finally, 18 KEGG pathways were mapped for the protein list according to the previous study (da Huang et al. 2009).

Statistical analyses

The significance for the difference between two groups was determined by an unpaired Student's t-test. A neurite was considered as an axon when it was longer than 100 μm and its length was twice that of other neurites. The axon length was measured by the Image-Pro Plus. All experiments were conducted at least three times independently. Data are shown as mean ± SD.

Results

Rufy3 is detected in the neuron-rich regions of the brain

Our previous study found that RUFY3 had an expression pattern similar to that of neuronal markers, such as TUJ1 and MAP2, in the human embryo during human embryonic days 20–32, suggesting that Rufy3 might be expressed in the human developing nervous system (Figure S1) (Fang et al. 2010). In addition, its ortholog was detected in the rat brain (Mori et al. 2007). However, its temporal and spatial expression pattern in the mouse remains largely unclear. To address this issue, we generated an affinity-purified polyclonal antibody against Rufy3. The antibody specifically recognized endogenous Rufy3 proteins, but not its other family members, as it detected only a single protein band when western blot analysis was conducted using whole-cell lysates of various mouse organs (Fig. 1a). Among four members of the mouse Rufy3 family, we found that Rufy3 was only expressed in the brain, whereas Rufy1 and Rufy2 were found in other mouse organs in addition to the brain, and Rufy4 was not detected in the brain (Figure S2a and b). At P0, mouse Rufy3 proteins were only detected in the brain and not in the other organs tested, including the heart, lung, liver, and kidney, verifying it as a brain-specific protein (Fig. 1a). During embryonic development, Rufy3 expression was detected in the brain from embryonic day 13.5 (E13.5) to adult, peaking around P0 and decreasing gradually with development (Fig. 1b). To gain more information of its spatial expression profile, embryonic sections of different developmental stages were used in the immunofluorescence staining. In sections of the whole embryo at E11.5, we found expression of Rufy3 in the cortex of the central nervous system, where it appeared restricted to the upper layers, but was not found in the ventricular zone (VZ) (Fig. 1c). Moreover, results from double staining of brain sections of E18.5 embryos with antibodies against Rufy3 and Tuj1 (a neuronal marker) or Nestin (a neural progenitor cell, NPC, marker) indicated that Rufy3 coexpressed with Tuj1 in the upper layers of the cortex and in other brain regions where neurons were abundant, but did not coexpress with Nestin (Fig. 1d). In the mouse E18.5 cortex, cells with high Rufy3 expression were distributed in the regions from the intermediate zone (IZ) to the cortical plate (CP). The expression of Rufy3 exhibited identical distribution to that of Tuj1 expression, but was different from that of Nestin. Furthermore, results from double staining using antibodies against Rufy3 and Map2 (a neuronal dendrite marker) showed that Rufy3 proteins were also present in dendrites containing Map2 proteins (Fig. 1e). These data clearly indicate that Rufy3 is a brain-specific protein predominantly expressed in the neuron-rich regions of the central nervous system.

Figure 1.

The expression pattern of Rufy3 in mouse brains. (a) Rufy3 antibody recognizes a single Rufy3 band in brain extracts. Whole tissue extracts from different organs of post-natal day 0 (P0) mice were prepared and immunoblotted with the affinity-purified Rufy3 polyclonal antibody. Rufy3 proteins were only detected in the brain. (b) Protein levels of Rufy3 in the brain at different stages of mouse development determined by western blot analysis. (c) The distribution of Rufy3 proteins in the brain of an E11.5 mouse detected by immunofluorescence staining. The right panel is an enlarged image of the inbox shown in the left panel. The Rufy3 antibody was used to stain endogenous Rufy3 proteins (red). (d) The distribution of Rufy3 in the brain of an E18.5 mouse was determined by double staining with antibodies against Rufy3 (red) and Tuj1 (green) or Nestin (green). Merged images with higher magnification are shown in the right panels. (e) Immunofluorescence images to show the distribution of Rufy3 in the cortex of an E18.5 mouse brain. The brain sections were double stained with antibodies against Rufy3 (red) and Nestin (green), or Tuj1 (green) or Map2 (green). Thin white lines are used to highlight the borderline of the cortical plate (CP), intermediate zone (IZ), and ventricular zone (VZ) of the brain. Merged images with higher magnification are shown in the right panels. Scale bars: 500 μm in the left panel of (c); 200 μm in the right panel of (c) 100 μm in (d); 250 μm in the left four panels of (e); 100 μm in the right panel of (e).

Rufy3 is specifically expressed in mouse primary cortical neurons and embryonic stem cell-derived neurons

To test whether Rufy3 is indeed a neuron-specific protein, we examined the expression of Rufy3 in neurons differentiated in vitro from cortical NPCs of E14.5 mouse embryos, as neurogenesis becomes abundant in mouse embryos around E13 (Gardette et al. 1982). When neurospheres, formed by suspension culture were digested into single cells and plated onto Matrigel-coated dishes, many NPCs appeared on day 1 of the plating (The last day of neurosphere suspension culture was defined as differentiation day 0). From day 2 of the plating, the NPCs began to differentiate into neurons efficiently. Around day 6, neuronal death took place and the cell number declined (Fig. 2a). Results of western blot analysis showed that Rufy3 proteins were hardly detectable at differentiation day 0, increasing from day 2 to day 4 and decreasing at day 6. A low level of Tuj1 proteins was observed at day 0, displaying a similar pattern as Rufy3 from day 2 to day 6. Unlike Tuj1 and Rufy3, proteins of NeuN (a mature neuron maker) were not detected until day 4 and continued to increase at day 6, whereas Sox2 (a stem cell marker) had a highest level at day 0 and decreased drastically after day 2 (Fig. 2b). This result suggests that Rufy3 might be highly expressed in newly generated neurons with decreased expression during neuronal maturation. Next, we examined Rufy3 expression at a single-cell level using P0 mouse primary cortical neurons and found that Rufy3 was widely expressed in all types of Tuj1+ cortical neurons. To determine whether Rufy3 would be detected in the axon, in addition to the cell body and dendrite, we carried out immunostaining for the neurons with longer axons using SMI 312 (a pan-axon marker) antibody to define axons. Our result showed that Rufy3 could be detected in the axon (Figure S3b). Taken together, Rufy3 proteins distributed in a spotty fashion in the cell body as well as in the outgrowth of primary mouse cortical neurons (Fig. 2c).

Figure 2.

Expression of Rufy3 in mouse primary neurons and during mouse ESC neural differentiation. (a) The cell morphology of neuronal differentiation from primary cortical neural progenitor cells (NPCs). NPCs were isolated from E14.5 mouse brain cortex and suspended as neurospheres in culture. On the third day, spheres were digested and plated onto Matrigel-coated dishes and differentiated for 6 days. Images of the bright field were taken at day 0 (before plating onto dishes, 3 day of neurosphere culture), day 1 (the next day after plating onto dishes), day 2, and day 6. (b) Protein levels of Rufy3, Tuj1, Sox2, and NeuN during primary cortical NPC differentiation were determined by western blot analysis. Arrows indicate the specific signal for NeuN. (c) The expression of Rufy3 in mouse primary post-natal day 0 (P0) cortical neurons were examined by double staining using antibodies against Rufy3 (red) and Tuj1 (green). The upper panel shows images of neurons with shorter outgrowth, while the bottom panel shows the images of neurons with longer axons. (c') is an enlarged image of the inbox shown in the bottom panel of C for Rufy3 staining. (d) The sketch map of mouse embryonic stem cell (ESC) neural differentiation process. After 7 days of embryoid body formation under a serum-free condition (SFEB), embryoid bodies were plated onto Matrigel-coated dishes and passaged every 3 days. (e) Expression patterns of different markers during the mouse ESC neural differentiation process were determined by qRT-PCR. The mRNA level in undifferentiated ESCs was set as I. Three independent experiments were performed. Data are shown as mean ± SD. (f) Expression levels of Rufy3 during the mouse ESC neural differentiation process were examined by qRT-PCR and western blot analyses. Data of mRNA levels are from three independent experiments. One representative result from three independent western blot analyses is shown. (g) The cell types expressing Rufy3 specifically were determined by immunofluorescence double staining in mouse ESC-derived neural cells, using antibodies against Rufy3 (red) and Nestin (green), or Tuj1 (green) or GFAP (green). The Nucleus was stained by 4, 6-diamidino-2 phenylindole (DAPI). Scale bars: 100 μm in (a); 50 μm in (c); 0.4 μm in C'; 25 μm in (g).

To obtain additional evidence that Rufy3 is a neuron-specific protein, we examined the expression pattern of Rufy3 during neural differentiation of mouse ESCs using a previously reported protocol (Watanabe et al. 2005), as illustrated in Fig. 2d. The neural differentiation process was characterized by the expression of various markers determined by real-time quantitative PCR (qRT-PCR). For example, Nestin was highly expressed at differentiation days 10 and 13, and its expression decreased afterward. The expression of Tuj1 reached to the peak at differentiation day 13, whereas the expression of glial markers Olig2 and Gfap did not reach their maximum levels until the late stages of the differentiation process (Fig. 2e). The sequential generation of NPCs, neurons and glial cells from ESCs matched the embryonic neural differentiation process. When expression of Rufy3 was examined using the same neural differentiation model, we found that it had an expression pattern identical to that of Tuj1 (Fig. 2f, upper panel). The transcript level of Rufy3 was barely detectable in undifferentiated ESCs, reached its peak at differentiation day 13 and declined afterward. Similarly, the protein level of Rufy3 was up-regulated until day 16 of differentiation and then decreased as the cells differentiated (Fig. 2f, lower panel). The identical temporal expression pattern of Rufy3 and Tuj1 is consistent with the notion that Rufy3 might be highly expressed in neurons but not other cell types. To further confirm this hypothesis, immunofluorescence staining was carried out in mouse ESC-derived neural cells (Fig. 2g). We found that Nestin and Rufy3-positive staining occurred exclusively in the same cells when antibodies against Nestin and Rufy3 were used to doubly stain the cells. Moreover, Rufy3 was not expressed in GFAP+ radial glia or astrocytes. Notably, every Tuj1-positive neuron expressed Rufy3, indicating that Rufy3 was coexpressed with Tuj1 in neurons. Therefore, we clearly demonstrate that Rufy3 is a neuron-specific protein.

Rufy3 controls the axonal growth and polarization process in hippocampal neurons

The next question was the function of Rufy3 in neurons. We addressed the question using both over-expression and RNA interference (RNAi) strategies. First, constructs containing the Rufy3-IRES-EGFP sequence or the IRES-EGFP control sequence were transfected into mouse P0 hippocampal neurons before plating. Two days later, the axon length of EGFP+ neurons and the percentage of EGFP+ neurons with multiple axons were measured. The morphology of neurons was revealed by EGFP expression and the axon was defined by SMI 312 staining (Fig. 3a). We found that the axonal length of neurons over-expressing Rufy3 was significantly longer than that of neurons transfected with a control plasmid (= three independent cultures, 128 neurons examined; < 0.05) (Fig. 3b). Moreover, the percentage of neurons with multiple axons seemed reduced by the expression of exogenous Rufy3, although the difference was not statistically significant (Fig. 3c). Thus, we concluded that over-expression of Rufy3 enhanced axonal growth. Second, we silenced Rufy3 expression by transfection of hippocampal neurons with two shRNA plasmids, which targeted two different regions of the Rufy3 mRNA (Ri-1#, Ri-2#), respectively, and contained a separate cassette for constitutive EGFP expression. The silencing efficiency of the shRNA plasmids was verified by western blot analysis of endogenous Rufy3 protein levels in neurons (Figure S3a). In hippocampal neurons transfected with both shRNA plasmids, knockdown of Rufy3 expression led to significant shortening of axons (n = three independent cultures; 96 neurons transfected with Ri-1# plasmid, < 0.01, when compared with the control plasmid; 91 neurons transfected with Ri-2# plasmid, < 0.01, when compared with the control plasmid) (Fig. 3d and e). Moreover, the percentage of EGFP+ neurons bearing more than one (two or three) axons increased in Rufy3-knocked down neurons (< 0.001) (Figs. 3d and f). Therefore, appropriate expression of Rufy3 is required for both establishment of normal polarity and maintenance of axonal growth in mouse hippocampal neurons.

Figure 3.

Rufy3 plays an important role in axon morphogenesis in mouse hippocampal neurons. (a) The morphology of mouse hippocampal neurons transfected with control (Ctrl) and Rufy3 over-expression (Rufy3 OE) plasmids. All plasmids contained a CAG promoter-driven EGFP cassette. Cells were stained with SMI312 antibody to label axons (red). (b) The axon length of mouse hippocampal neurons transfected with the Rufy3 OE plasmid were compared with that of control cells. (c) The percentage of mouse hippocampal neurons with multiaxons was compared between Rufy3 OE and control cells. (d) The morphology of mouse hippocampal neurons transfected with control and Rufy3 shRNA plasmids (Ri-1# and Ri-2#). The plasmids contained a PGK promoter-driven EGFP expression cassette. (e) The axon length of mouse hippocampal neurons transfected with Rufy3 shRNA plasmids was compared with that of control cells (Ri-1# with control, Ri2# with control, respectively). (f) The percentage of mouse hippocampal neurons with multiaxons was compared between control and Rufy3 shRNA cells (Ri-1# with control, Ri2# with control, respectively). *< 0.05; **< 0.01; ***< 0.001. Scale bars: 75 μm. Data are shown as mean ± SD.

Rufy3 interacts and colocalizes with F-actin-binding proteins Fascin and Drebrin in neuronal growth cones

Having defined the role of Rufy3 in the axonal growth of mouse hippocampal neurons, we went on to investigate how Rufy3 influenced axon growth at a molecular level. As Rufy3 contains RUN and coil–coil domains, which could mediate protein–protein interactions, we assumed that identification of Rufy3-associated proteins in neurons would facilitate our understanding how it executes its functions. To this end, immunoprecipitation experiments were carried out using tissue extract of the whole-mouse P0 brain and the Rufy3 antibody to purify endogenous Rufy3-associated protein complexes. Precipitated proteins were separated on SDS–PAGE gels, which were then cut off for mass spectromic (MS) analysis (Supplementary Information, Figure S4). The name of proteins specifically precipitated by Rufy3 antibodies is listed in Table 1. Interestingly, cytoskeletal proteins such as Spectrin, Filamin A, Ankyrin-2, Fascin, and Drebrin were enriched in the list. Consistent with that result, the gene ontology (GO) analysis also showed that Rufy3-associated proteins might be involved with the control of cytoskeletons (Fig. 4a). The finding suggests that Rufy3's role in neuronal morphogenesis may be related to the regulation of cytoskeletal organization. Of these proteins, we focused on Fascin and Drebrin, as both are known F-actin-binding proteins and are involved in the growth of neuronal cones.

Table 1. The list of Rufy3 interacting proteins identified by mass spectrometry
AccessionNamePeptides (95%)
  1. The proteins in the Rufy3 (Rufy3 antibody IP) group having a value at a 95% confidence interval at least two folds greater than that in the control (IgG IP) group were selected and are listed.

sp|P16546|SPTA2_MOUSESpectrin alpha chain, brain OS = Mus musculus GN = Sptan1 PE = 1 SV = 427
sp|P57780|ACTN4_MOUSEAlpha-actinin-4 OS = Mus musculus GN = Actn4 PE = 1 SV = 126
sp|Q7TPR4|ACTN1_MOUSEAlpha-actinin-1 OS = Mus musculus GN = Actn1 PE = 1 SV = 116
sp|Q62261|SPTB2_MOUSESpectrin beta chain, brain 1 OS = Mus musculus GN = Sptbn1 PE = 1 SV = 216
sp|Q8BTM8|FLNA_MOUSEFilamin-A OS = Mus musculus GN = Flna PE = 1 SV = 510
sp|P46660|AINX_MOUSEAlpha-internexin OS = Mus musculus GN = Ina PE = 1 SV = 29
sp|P63260|ACTG_MOUSEActin, cytoplasmic 2 OS = Mus musculus GN = Actg1 PE = 1 SV = 18
sp|P60710|ACTB_MOUSEActin, cytoplasmic 1 OS = Mus musculus GN = Actb PE = 1 SV = 18
sp|Q61553|FSCN1_MOUSEFascin OS = Mus musculus GN = Fscn1 PE = 1 SV = 46
sp|Q9D394|RUFY3_MOUSEProtein RUFY3 OS = Mus musculus GN = Rufy3 PE = 1 SV = 16
sp|Q9WUM4|COR1C_MOUSECoronin-1C OS = Mus musculus GN = Coro1c PE = 1 SV = 23
sp|Q8BH44|COR2B_MOUSECoronin-2B OS = Mus musculus GN = Coro2b PE = 2 SV = 23
sp|Q9QXS6|DREB_MOUSEDrebrin OS = Mus musculus GN = Dbn1 PE = 1 SV = 43
sp|Q99K48|NONO_MOUSENon-POU domain-containing octamer-binding protein OS = Mus musculus GN = Nono PE = 1 SV = 33
sp|Q8C8R3|ANK2_MOUSEAnkyrin-2 OS = Mus musculus GN = Ank2 PE = 1 SV = 22
sp|Q62167|DDX3X_MOUSEATP-dependent RNA helicase DDX3X OS = Mus musculus GN = Ddx3x PE = 1 SV = 32
sp|Q62095|DDX3Y_MOUSEATP-dependent RNA helicase DDX3Y OS = Mus musculus GN = Ddx3y PE = 1 SV = 22
sp|Q9QXZ0|MACF1_MOUSEMicrotubule-actin cross-linking factor 1 OS = Mus musculus GN = Macf1 PE = 1 SV = 22
sp|P16381|DDX3L_MOUSEPutative ATP-dependent RNA helicase Pl10 OS = Mus musculus GN = D1Pas1 PE = 1 SV = 12
sp|Q8CAQ8|IMMT_MOUSEMitochondrial inner membrane protein OS = Mus musculus GN = Immt PE = 1 SV = 11
sp|Q99KE1|MAOM_MOUSENAD-dependent malic enzyme, mitochondrial OS = Mus musculus GN = Me2 PE = 2 SV = 11
sp|Q61656|DDX5_MOUSEProbable ATP-dependent RNA helicase DDX5 OS = Mus musculus GN = Ddx5 PE = 1 SV = 21
sp|Q7TMM9|TBB2A_MOUSETubulin beta-2A chain OS = Mus musculus GN = Tubb2a PE = 1 SV = 11
sp|Q9CWF2|TBB2B_MOUSETubulin beta-2B chain OS = Mus musculus GN = Tubb2b PE = 1 SV = 11
sp|P68372|TBB4B_MOUSETubulin beta-4B chain OS = Mus musculus GN = Tubb4b PE = 1 SV = 11
Figure 4.

Rufy3 interacts with Drebrin and Fascin, colocalizing with F-actin in the growth cone of mouse hippocampal neurons. (a) The GO analysis for Rufy3 interacting proteins identified by mass spectrometry based on the biological process. The p-value was converted into –log10. The biological process with a smaller p-value is more significant than others. (b) Endogenous coimmunoprecipitation (Co-IP) between Rufy3 and Drebrin or Fascin using extracts of post-natal day 0 (P0) mouse brains. The symbol * indicates the heavy chain of IgG. (c) Exogenous Co-IP between Rufy3 and Fascin in 293T cells. (d) A sketch map of different domains for Rufy3 proteins. These domains were constructed into a Flag-pPy plasmid and expressed in 293T cells. (e) Protein interactions between different domains of Rufy3 and Fascin in 293T cells. The cDNA of full-length Fascin was constructed into a HA tagged plasmid. The symbol # indicates non-specific signals in the western blot analysis. The arrows indicate specific HA-Fascin signals. The symbol * indicates the heavy chain of IgG. (f) Subcellular distribution of Drebrin, Rufy3, and Fascin in growth cones of mouse hippocampal neurons was detected by immunofluorescence staining. The neurons were double stained with Rufy3 (green) and Drebrin (red) or Rufy3 (red) and Fascin (green) antibodies. Growth cone and the distal part of axon are shown. (g) The localization of Drebrin, Rufy3, and Fascin was compared to that of F-actin in growth cones. The hippocampal neurons were double stained with antibodies against Drebrin (green) and F-actin (red), or Fascin (green) and F-actin (red) or Rufy3 (green) and F-actin (red). Only growth cones are shown in images. Scale bars: 10 μm. IB: immunoblotting. IP: immunoprecipitation.

Specific interactions between Rufy3 and Fascin or Drebin were verified by Co-IP assays using tissue extracts from the P0 mouse brain. As shown in Fig. 4b, Drebrin antibodies could pull down Rufy3 and Fascin proteins in addition to Drebrin proteins; proteins of Fascin, Rufy3, and Drebrin could be detected in the complexes precipitated by Rufy3 antibodies; and Fascin antibodies coprecipitated Drebrin and Rufy3 as well as Fascin proteins. As negative controls, IgG could not bring down any of these three proteins. Therefore, we show here for the first time that endogenously expressed Rufy3 interacts with Fascin and Drebrin specifically in the brain cells under physiological conditions.

To further characterize the interaction between Rufy3 and Fascin or Drebrin, we over-expressed Rufy3 and HA-tagged Fascin or HA-tagged Drebrin in 293T cells. Interestingly, we were unable to detect the interaction between exogenously expressed Rufy3 and HA-tagged Drebrin in 293T cells, although the specific interaction between Rufy3 and HA-Fascin was readily observed (Fig. 4c). This phenomenon suggests that the specific interaction between endogenous Rufy3 and Drebrin detected in neurons might not be direct, but mediated through factor (s) absent from 293T cells; or the interaction only occurs under certain conditions. As exogenously expressed Rufy3 and Fascin could interact, we next determined the domain of Rufy3 mediating its interaction with Fascin. Flag-tagged full-length Rufy3, the RUN domain in its N-terminus, or the coil–coil domain in its C-terminus was coexpressed with or without HA-tagged Fascin in 293T cells and Co-IP experiments were performed (Fig 4d). Our results indicated that Fascin interacted with the full-length and the N-terminus of Rufy3, but not with the C-terminus (Fig. 4e). As the RUN domain is the only known domain in the N-terminus of Rufy3, we proposed that it was the RUN domain of Rufy3 responsible for its association with Fascin.

Both Drebrin and Fascin have been reported to be distributed in growth cones (Edwards and Bryan 1995; Geraldo et al. 2008). Moreover, our study indicated that they were also highly expressed during neural differentiation of ESCs and in the developing brain (Figure S2c and d). However, whether they are codistributed with Rufy3 in growth cones is unknown. To answer this question, immunofluorescence staining was performed in mouse P0 hippocampal neurons using the specific Rufy3 antibody. We found that Drebrin was mainly expressed at the transitional domain, as reported previously (Geraldo and Gordon-Weeks 2009), whereas Rufy3 and Fascin were distributed primarily at filopodia and lamellipodia, although they were also seen at the transitional domain and central domain of the growth cone (Fig. 4f). Rufy3 staining appeared punctate and coincided with Fascin entirely in the whole growth cone, only partially overlapping with Drebrin at the transitional domain. In addition, Rufy3 displayed a similar distribution pattern to F-actin at the peripheral and transitional domains (Fig. 4g), while Drebrin was partially coexpressed with F-actin at the transitional domain. Our results suggest that Rufy3 may be a new member of the F-actin binding and bundling protein family.

Rufy3 and F-actin are mutually dependent for normal distribution in the axonal growth cone

As Rufy3 colocalized with F-actin in the growth cone, we tested whether disruption of F-actin would affect distribution of Rufy3 in growth cones by treating cultured hippocampal neurons at day 2 for 2 h with 1 μM of cytochalasin D (CytoD), an inhibitor of actin polymerization (Nair et al. 2008). After CytoD treatment, cultured neurons formed axons without large growth cones, as reported previously (Bradke and Dotti 1999). Simultaneously, F-actin contents decreased drastically in the peripheral and transitional domains, with a small amount of F-actin scattered along the edges of filopodia and lamellipodia. Interestingly, CytoD treatment also led to a decrease in Rufy3 staining, which exhibited a distribution pattern identical to F-actin (Fig. 5a), indicating that the localization of Rufy3 in growth cones depended on the F-actin organization. In addition, disruption of F-action brought about an almost complete loss of Fascin and Drebrin in the transitional and central domains, with a small amount of these two proteins detected at the edge of the growth cone (Fig. 5b and c), in accordance with previous reports (Cohan et al. 2001; Mizui et al. 2009). Notably, F-actin staining at the edge of growth cones invariably colocalized with the staining of Rufy3, Drebrin, and Fascin after CytoD treatment. These observations indicate that Rufy3 is closely associated with F-actin in growth cones, as are Drebrin and Fascin.

Figure 5.

Normal distribution of Rufy3 in the growth cone relies on the polymerization of F-actin. Mouse hippocampal neurons used in (a–c) were cultured for 2 days and then treated with 1 μM of cytochalasin D (CytoD) for 2 h before fixation. Typical bright field and immunofluorescence staining images are shown. Images of double staining of Rufy3 (green) and F-actin (red) with or without CytoD treatment are shown in (a). Arrows indicate some signals of F-actin and Rufy3 after adding CytoD. Images of double staining of Fascin (green) and F-actin (red) are shown in (b). Arrows indicate some signals of F-actin and Drebrin after adding CytoD. Images of double staining of Drebrin (green) and F-actin (red) are shown in (c). Arrows indicate some signals of F-actin and Fascin after adding CytoD. Scale bars: 10 μm.

We then asked whether appropriate expression of Rufy3 was required for normal organization of F-actin. Hippocampal neurons of P0 mice were nucleofected with Rufy3 over-expression plasmid (OE) or shRNA plasmids (Ri-1#, Ri-2#), both containing a constitutive EGFP expression unit. The EGFP expression cassette was used to illustrate the morphology of the growth cones. Over-expression of Rufy3 did not alter the localization of F-actin (Fig. 6a). However, knockdown of Rufy3 reduced F-actin staining intensity in the transitional domain and F-actin appeared to concentrate at the edge of the growth cone (Fig. 6b). Thus, Rufy3 played an important role in the normal function of F-actin and the proper structure of growth cones.

Figure 6.

Normal distribution of F-actin in the growth cone of mouse hippocampal neurons requires the presence of Rufy3. (a) Typical bright field images and images of double staining of F-actin (red) and enhanced green fluorescent protein (EGFP) (green) in mouse hippocampal neurons transfected with control (ctrl) or Rufy3 over-expression plasmid (Rufy3 OE) are shown. (b) Typical bright field images and images of double staining of F-actin (red) and EGFP (green) in mouse hippocampal neurons transfected with control (ctrl) or Rufy3 shRNA plasmid (Ri-1#, Ri-2#). Arrows indicate some signals of F-actin after Rufy3 shRNA plasmids were transfected. Scale bars: 9 μm.

Rufy3 is required for the normal localization of Drebrin and Fascin in axonal growth cones

Given that normal expression of Rufy3 is crucial for proper organization of F-actin in axonal growth cones and that disruption of F-action structure could affect the distribution of Fascin and Drebrin, we anticipated that aberrant expression of Rufy3 would also alter the localization of Fascin and Drebrin. To verify this assumption, we conducted the same over-expression and RNAi experiments to test whether Rufy3 could affect the distribution of Fascin and Drebrin. Over-expression of Rufy3 seemed to increase the staining intensity of Drebrin in the growth cone, which extended to the peripheral domain from its normal position in the central and transitional domains (Fig. 7a). In contrast, the localization of Fascin was not significantly affected by Rufy3 over-expression (Fig. 7b). However, when the expression of Rufy3 was knocked down, the intensity of Fascin staining decreased with restricted localization at the edge of the growth cone (Fig. 7d), while the distribution pattern and staining intensity of Drebrin did not change (Fig. 7c). These data clearly demonstrate that there are functional interactions among Rufy3, Fascin, Drebrin, and F-actin, and that Rufy3 is an important player in cytoskeletal organization during axon development in mouse neurons.

Figure 7.

Rufy3 controls distribution of Fascin and Drebrin in the growth cone of mouse hippocampal neurons. (a) Typical bright field images and images of double staining of Drebrin (red) and enhanced green fluorescent protein (EGFP) (green) in mouse hippocampal neurons transfected with control (ctrl) or Rufy3 over-expression plasmid (Rufy3 OE). Arrows indicate some signals of Drebrin after Rufy3 over-expression plasmids were transfected. (b) Typical bright field images and images of double staining of Fascin (red) and EGFP (green) in mouse hippocampal neurons transfected with control (ctrl) or Rufy3 over-expression plasmid (Rufy3 OE). (c) Typical bright field images and images of double staining of Drebrin (red) and EGFP (green) in mouse hippocampal neurons transfected with control (ctrl) or Rufy3 shRNA plasmid (Ri-1#, Ri-2#). (d) Typical bright field images and images of double staining of Fascin (red) and EGFP (green) in mouse hippocampal neurons transfected with control (ctrl) or Rufy3 shRNA plasmid (Ri-1#, Ri-2#). Arrows indicate some signals of Fascin after Rufy3 shRNA plasmids were transfected. Scale bars: 10 μm.

Discussion

In this study, we proposed that Rufy3 is a new member of F-actin-associated proteins specifically expressed in mouse neurons and important for neural axonal morphogenesis. Rufy3 colocalized with F-actin entirely in the peripheral and transitional domains of axonal growth cones in mouse hippocampal neurons. Importantly, the distribution of Rufy3 and F-actin in growth cones was dependent on each other. Actin bundles are important not only for the formation of growth cones through supporting filopodial extension but also for cone motility by controlling protrusion of lamellipodia and filopodia at the leading edge. The bundling of F-actin is regulated by different actin binding and bundling proteins (Bartles 2000). Thus, identification of Rufy3 as a new F-actin-associated protein is significant for understanding how axons form and elongate. In particular, we found that Rufy3 was exclusively expressed in neurons of mouse brain tissues, but not in NPCs or glia cells, suggesting its unique role in neuronal development. Indeed, in this study, Rufy3 over-expression enhanced the axon length significantly and reduced the percentage of neurons with multiple axons. Conversely, Rufy3 knockdown reduced the axonal length substantially and increased the neurons with surplus axons in mouse primary hippocampal neurons, suggesting that Rufy3 is a crucial player for controlling axon formation and elongation. Consistent with our finding, a previous study showed that Singar 1 was important for suppressing surplus axons in rat hippocampal neurons (Mori et al. 2007). However, in that study, silencing Singar 1 did not alter the axonal length, and Singar 1 over-expression did not affect the axon formation or axon length during normal polarization process. Therefore, it appears that Rufy3 proteins from the rat and mouse may play similar and different roles in neuronal morphogenesis. Our discovery of Rufy3 as a neuron-specific protein functionally relevant to F-actin places Rufy3 at an important position in the control of axon formation and elongation.

Additional evidence to support a close relationship between Rufy3 and F-actin came from the identification of Rufy3-associated protein complexes. The highly specific Rufy3 antibody developed in our laboratory provided us an opportunity to purify endogenous proteins interacting with Rufy3 in a physiological context. There were several cytoskeleton-associated proteins in the Rufy3 protein complexes. Among these proteins, we verified the specific interactions between Rufy3 and Drebrin or Fascin. Similar to Rufy3, Drebrin has been shown to be expressed throughout neurons. However, its highest levels are in the cell soma and growth cone, where it localizes to the transitional domain and the proximal region of filopodia where microtubules insert (Geraldo et al. 2008). Geraldo et al. demonstrated that the specific interaction between the Drebrin bound to the proximal F-actin bundle of filopodia and microtubule and plus-tip protein EB3 located at the tip of microtubules invading filopodia is essential for growth cone formation and neurite extension. We found that Rufy3 formed protein complexes with Drebrin in neurons, but not in non-neural 293T cells, suggesting that the interaction between these two proteins might be indirect or occur only under certain conditions. Interestingly, over-expression of Rufy3 caused Drebrin proteins to localize to a more peripheral region, while Rufy3 knockdown did not affect its distribution. As over-expression of Rufy3 did not alter localization of F-actin and Fascin, it is unlikely that the aberrant localization of Drebrin mediated by Rufy3 over-expression was through F-actin or Fascin. In fact, we found that the distribution of Rufy3 in growth cones resembled Fascin. Fascin was shown to incorporate into F-actin from the beginning of growth cone formation and to localize to and associate with radially oriented actin bundles in lamellipodia and filopodia except at the proximal part adjacent to the central domain, where actin disassembling takes place. The changes in Fascin's actin-bundling activity, which is inhibited by phosphorylation, could alter the actin bundle organization (Cohan et al. 2001). Similar to Fascin, Rufy3 was found in the cell body, axon, and growth cones. Punctate Rufy3 staining could be observed in both lamellipodia and filopodia, entirely colocalizing with Fascin and partially with Drebrin. Co-IP results further supported the close relationship between Rufy3 and Fascin. Notably, knockdown of Rufy3 reduced Fascin and F-actin expression in the proximal part of growth cones, although Drebrin expression was not affected by the silencing of Rufy3. Thus, we favor the notion that Rufy3 plays an important role in the maintenance of F-actin structure and axonal growth, probably through its interaction with Fascin. Nevertheless, we do not rule out the possibility that other proteins identified in the Rufy3 protein complexes are also involved in the regulatory effect of Rufy3 on axon elongation. Moreover, it remains unclear whether Rufy3 suppresses the surplus axons and regulates the axon growth through the same or different mechanisms. It is obvious that the normal distribution of Drebrin, Fascin, and Rufy3 all require the presence of F-actin. Further investigation is needed to understand how the three proteins interact, in particular how Rufy3 influences Fascin.

During human embryo development, the expression of RUFY3 was found to increase during embryonic days 20–32, with an expression pattern similar to those of neural markers such as TUJ1 and MAP2, but different from those of mesoendoderm markers, such as GATA4 and GATA6. This suggests that RUFY3 may also be expressed in neural tissues of human embryos. Considering the function of Rufy3 in the mouse and the rat, the function of Rufy3 in human embryonic development should be studied using human ESC neural differentiation as an in vitro model. In addition, actin-binding proteins are known to regulate the morphological changes of neurons in neurodegenerative diseases. For instance, Drebrin has been implicated in Alzheimer's diseases (Shim and Lubec 2002). It will be interesting to examine whether Rufy3 is also associated with the development of human neural disorders.

Acknowledgments and conflict of interest disclosure

We would like to thank Dr Zhenge Luo for his helpful discussion and Dr Yongchun Yu for his technical support in the brain section experiments. In addition, authors thank Erbo Xu for his editing of the manuscript. This study was supported by grants from the Chinese Academy of Sciences (XDA01010102), the National Natural Science Foundation (91019023), and the National High Technology Research and Development Program of China (2010CB945200, 2011DFB300100, 2011CB965101, and 2009CB941103). No competing interests exist.

All experiments were conducted in compliance with the ARRIVE guidelines.

Ancillary