SH2B1 and IRSp53 Proteins Promote the Formation of Dendrites and Dendritic Branches*

Background: Filopodium formation is a prerequisite for neurite initiation. Results: SH2B1 interacts with IRSp53 and promotes neurite outgrowth of hippocampal neurons. Conclusion: SH2B1-IRSp53 complexes promote filopodium formation, neurite initiation, and neurite branching. Significance: These results suggest that SH2B1 may regulate the formation of filopodia required for many cellular functions. SH2B1 is an adaptor protein known to enhance neurite outgrowth. In this study, we provide evidence suggesting that the SH2B1 level is increased during in vitro culture of hippocampal neurons, and the β isoform (SH2B1β) is the predominant isoform. The fact that formation of filopodia is prerequisite for neurite initiation suggests that SH2B1 may regulate filopodium formation and thus neurite initiation. To investigate whether SH2B1 may regulate filopodium formation, the effect of SH2B1 and a membrane and actin regulator, IRSp53 (insulin receptor tyrosine kinase substrate p53), is investigated. Overexpressing both SH2B1β and IRSp53 significantly enhances filopodium formation, neurite outgrowth, and branching. Both in vivo and in vitro data show that SH2B1 interacts with IRSp53 in hippocampal neurons. This interaction depends on the N-terminal proline-rich domains of SH2B1. In addition, SH2B1 and IRSp53 co-localize at the plasma membrane, and their levels increase in the Triton X-100-insoluble fraction of developing neurons. These findings suggest that SH2B1-IRSp53 complexes promote the formation of filopodia, neurite initiation, and neuronal branching.

During development, neural stem cells undergo a morphogenetic series to become mature neurons in response to extrinsic as well as intrinsic factors. One prominent feature of neurogenesis is neurite outgrowth. Although an array of studies have identified the growth factors, receptors, and signaling events that are responsible for neurite outgrowth and guidance, intrinsic regulation for neurite initiation is less clear. For hippocampal and cortical neurons cultured in vitro, the formation of lamellipodia containing filopodia is required for navigation of neurites. Thus, the coordination of actin and microtubule assembly becomes a major determinant of neurite outgrowth.
As several studies have implicated the importance of filopodia in neurite formation, a recent report clearly demonstrates the formation of filopodia preceding neurite initiation (1). Dent et al. (1) show that extracellular matrix laminin promotes the formation of actin-rich protrusions (filopodium-like) and is able to rescue neuritogenesis in vasodilator-stimulated phosphoprotein (VASP) 3 -deficient neurons. This study also reveals the importance of regulators of filopodium formation during neurite outgrowth.
Filopodia are actin-rich membrane protrusions involve in cell migration, neurite initiation, axon guidance in neuronal growth cones, endocytosis, and wound healing (1)(2)(3). Filopodia consist of unbranched F-actin filament bundles that are regulated by many actin-binding proteins such as IRSp53 (insulin receptor tyrosine kinase substrate p53), fascin, Mena/VASP, and formins (3)(4)(5). IRSp53 belongs to Inverse Bin-Amphiphysin-Rvs 167 (I-BAR), also known as IMD (IRSp53-missing in metastasis homology domain), domain-containing superfamily of proteins and is known to drive membrane deformation, the subsequent plasma membrane protrusions, and thus filopodium formation (3, 6 -8). They are retracted by retrograde flow of F-actin and capping protein activity. The dynamic balance of barbed-end actin polymerization and retraction determines the initiation, maintenance, and stability of filopodia. The molecular mechanisms for controlling the initiation of dendritic filopodia are not clear.

MATERIALS AND METHODS
Animal Handling and Ethics Statement-All experiments were conducted in accordance with the guidelines of the Laboratory Animal Center of National Tsing Hua University. Animal use protocols were reviewed and approved by the National Tsing Hua University Institutional Animal Care and Use Committee (approval number 10126).
Antibodies and Reagents-Polyclonal antibody to rat SH2B1 was raised against a glutathione S-transferase fusion protein containing amino acids 527-670 of SH2B1␤, as described previously (27), and was used at a dilution of 1:1000 for Western blotting. Anti-IRSp53, used at a dilution of 1:5000 for Western blotting and 1:200 for immunoprecipitation, and anti-Myc antibodies, used at a dilution of 1:1000 for Western blotting and 1:200 for immunoprecipitation, were purchased from Millipore (Billerica, MA). Anti-MAP2 (used at a dilution of 1:200 for immunostaining), anti-Tau-1 (used at a dilution of 1:200 for immunostaining), anti-␣-tubulin antibodies (used at a dilution of 1:1000 for Western blotting), and protein G-agarose beads were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-␤III tubulin antibody was purchased from COVANCE (Berkeley, CA) and used at a dilution of 1:1000 for immunostaining. Anti-GFP (used at a dilution of 1:1000 for Western blotting and 1:200 for immunoprecipitation) and anti-GAP43 antibodies were purchased from GeneTex (San Antonio, TX) and used at a dilution of 1:1000 for Western blotting. Anti-actin antibody was purchased from Sigma and used at a dilution of 1:200 for immunostaining. IRDye800CW-labeled anti-rabbit and IRDye680CW-labeled anti-mouse secondary antibodies were purchased from LI-COR Bioscience (Lincoln, NE). Alexa Fluor fluorescence-conjugated secondary antibodies (AF350, AF488, AF555, and AF647), minimum essential media, neurobasal medium, and B-27 serum-free supplement were purchased from Invitrogen. Dulbecco's modified Eagle's medium, horse serum, fetal bovine serum (FBS), L-glutamine (Gln), antibiotic-antimycotic, rhodamine phalloidin, 4Ј,6-diamidino-2phenylindole (DAPI), and Lipofectamine 2000 were purchased from Invitrogen. Protease inhibitors, including aprotinin and leupeptin, were purchased from Roche Applied Science. Enhanced chemiluminescence (ECL) reagent was purchased from PerkinElmer Life Sciences. Mammalian transfection system-calcium phosphate reagent was purchased from Promega (Madison, WI).
Primary Neuronal Culture and 293T Cell Culture-Sprague-Dawley rats were purchased from BioLASCO Taiwan Co., Ltd. Cortical or hippocampal neurons were dissociated from hippocampus dissected from rat embryos (18th day of gestation; E18) by treatment with papain solution (10 units/ml). The isolated primary neurons (1-2 ϫ 10 5 /ml) were seeded on a poly-L-lysine (30 ng/ml)-coated dish or coverslip. On day in vitro (DIV) 0, primary neurons were cultured in minimum essential media/high glucose medium supplemented with 5% FBS and 5% horse serum under 5% CO 2 conditions. On DIV 1, cells were cultured in neurobasal medium with B27 (containing additional glutamine (Gln) and 0.025 mM glutamate). On DIV 2, cells were treated with 5 M cytosine-1-␤-D-arabinofuranoside to inhibit the growth of glial cells. On DIV 3, cells were cultured in neurobasal and Gln medium (neurobasal medium with B27 containing additional Gln), and then half of the neurobasal and Gln medium was replaced by fresh medium every 3 days. Lipofectamine 2000 or calcium phosphate reagents were used to transfect primary neurons according to the manufacturer's instruction. 1.5-3 h after transfection, culture medium was replaced with fresh medium. 293T cells were grown in DMEM containing 10% FBS, 1% L-Gln, and 1% antibiotic/antimycotic under 5% CO 2 conditions. Reverse Transcription-Polymerase Chain Reaction (RT-PCR)-TRIzol reagent was used to isolate total RNA from neurons according to the manufacturer's instructions. For reverse transcription, 2 mg of total RNA was converted to cDNA using the reverse transcription kit (Applied Biosystems). SH2B1 isoform primer pairs are as follows: forward 5Ј-TTCGATATGCTTG-AGCACTTCCGG-3Ј and reverse 5Ј-GCCTCTTCTGCCCC-AGGATGT-3Ј. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) primer pairs are as follows: forward 5Ј-ACCACAG-TCCATGCCATGCCATCAC-3Ј and reverse 5Ј-TCCACCAC-CCTGTTGCTGTA-3Ј. The mRNA levels of SH2B1 isoform from RT-PCR were normalized to that of GAPDH.
Western Blotting, Immunoprecipitation, and Immunofluorescence Staining-Cells were lysed with radioimmunoprecipitation assay buffer, containing 1 mM Na 3 VO 4 , 1 mM phenylmeth-ylsulfonyl fluoride (PMSF), 10 ng/ml aprotinin, and 10 ng/ml leupeptin. Equal amounts of proteins were separated by SDS-PAGE and analyzed by Western blotting using the indicated antibodies followed by incubation with IRDye-conjugated secondary antibody. Protein signal was detected using Odyssey infrared imaging system (Odyssey Imager). For immunoprecipitation, cell lysates were incubated with the specific antibody at 4°C overnight followed by incubation with protein A-or G-agarose beads and rotated at 4°C for 1 h to pull down specific antibodies. Samples were centrifuged at 2000 rpm at 4°C for 1 min. The beads were washed three times with radioimmunoprecipitation assay lysis buffer to remove nonspecific binding. The precipitates were dissolved in 1ϫ sample buffer, boiled at 95°C for 10 min, and analyzed by Western blotting. For immunofluorescence staining, cells were fixed by 4% paraformaldehyde, incubated with the indicated primary antibodies, followed by Alexa Fluor-conjugated secondary antibody. DAPI staining was used to mark the location of the nucleus. Cells were then mounted with Prolong Gold reagent, and the fluorescent images were taken using Carl Zeiss Observer Z1 microscope or Carl Zeiss LSM 510 meta confocal microscope.
GST Pulldown Assay-GST-SH2B1␤ constructs were expressed in a BL21 strain, and lysates were harvested. GST-SH2B1␤ proteins were pulled down via glutathione-Sepharose 4B (GE Healthcare). Sepharose beads conjugated with GST-SH2B1␤ were then incubated with 293T cell lysates for 2 h at 4°C. Beads were washed and resuspended in SDS sample buffer for Western blotting.
Neurite Outgrowth of PC12 Cells-For nerve growth factor (NGF)-induced neurite outgrowth, PC12-shLacZ or PC12-shSH2B1 cells were split onto collagen I-coated plates at about 20 -30% confluency in low serum differentiation medium (DMEM containing 2% horse serum, 1% FBS, 1% antibiotic/ antimycotic, and 1% L-Gln). The average length of the longest neurite of cells was determined with ImageJ software. Differentiation of PC12 cells is defined as the length of the neurites being at least twice the diameter of the cell body.
Triton X-100 Cytoskeleton Extraction-Cells were lysed with extraction buffer (0.5% Triton X-100, 100 mM NaF, 50 mM KCl, 2 mM MgCl 2 , 1 mM EGTA, 10 mM KPO 4 , pH 7.5, 0.5 M sucrose) supplemented with protease inhibitors containing 1 mM Na 3 VO 4 , 1 mM PMSF, 10 ng/ml aprotinin, and 10 ng/ml leupeptin and then centrifuged at 13,000 rpm at 4°C for 10 min. The supernatant (detergent-soluble fraction containing the G-actin fraction) was taken for immunoblotting. Cell pellets containing F-actin fractions were scraped into the same extraction buffer of the supernatant, and both fractions were solubilized with the same volume of 5ϫ sample buffer. Equal volumes of proteins from each fraction were resolved by SDS-PAGE and detected by Western blotting analysis using the indicated antibodies.

Measurement of Attachment, End Points, Axon Length, Pearson's Correlation Coefficient, and Co-localization Analysis-
The number of attachment and end points was quantified using the built-in cell-counter in ImageJ software (rsb.info.nih.gov, National Institutes of Health, Bethesda). The attachment points are defined as the location where neurite and filopodia connect to the cell body and end point as the location at the tip of filopodia and neurites (28). The average length of the axon was measured using the Simple Neurite Tracer of Fiji software, a plugin in ImageJ software performing semi-automatic tracing of neurons. The mean pixel value of axon length was measured and converted pixels into micrometers (m). The Pearson's correlation coefficient (r) was used to measure the relationship between two variables. Images were analyzed using Meta-Morph software (Molecular Devices, Sunnyvale, CA). To examine the correlation between IRSp53 and SH2B1␤ in hippocampal neurons, cells expressing GFP-SH2B1␤ were fixed, then FIGURE 1. ␤ isoform of SH2B1 is the major isoform during the development of hippocampal neurons. A, RNA was extracted from E18 hippocampal neurons of DIV 1, 3, 8, and 14 and converted to cDNAs for RT-PCR analysis. GAPDH was used as a loading control. B, lysates from hippocampal neurons of DIV 1, 3, 5, and 7 were extracted for immunoblotting with anti-SH2B1 or anti-␣-tubulin antibody. ␣-Tubulin was used as a loading control. The relative expression of SH2B1 was quantified by using ImageJ software. *, p Ͻ 0.05, compared with DIV 1.
stained with IRSp53, and imaged by using a confocal microscope (Zeiss LSM 510). Both the raw single plane graph of IRSp53 and SH2B1␤ were quantified by "Co-localization Finder" and "Co-localization Colormap" plugins in ImageJ software. To reach the top 2% overlapping pixels, images should restrain selection to pixels with 85%. Correlation between pair of pixels from IRSp53 and SH2B1␤ is displayed as distribution of normalized mean deviation product values and visualized with a color scale. Negative indexes are represented by cold colors. Indexes above 0 are represented by hot colors, indicating co-localization. The correlations of IRSp53 and SH2B1␤ are represented as a color map that is calculated based on Equation 1, Ia indicates intensity for the given pixel in image a; Ia indicates average intensity of image a; Ia max indicates highest pixel intensity in image a; Ib indicates intensity for the given pixel in image b; Ib indicates average intensity of image b; and Ib max indicates highest pixel intensity in image b. Duolink in Situ Proximity Ligation Assay (PLA)-The Duolink in situ PLA assay kit was purchased from Olink Bioscience (Uppsala, Sweden) and was performed according to the manufacturer's instruction. Hippocampal neurons were fixed by 4% paraformaldehyde for 15 min, permeabilized by 0.1% Triton X-100 for 10 min, and incubated with 1% bovine serum albumin in PBS. Neurons were incubated with rabbit anti-SH2B1 antibody and mouse anti-IRSp53 antibodies at 4°C overnight, followed by incubation with Duolink PLA rabbit PLUS and mouse MINUS probes. After incubation, ligation and  MARCH 6, 2015 • VOLUME 290 • NUMBER 10 amplification were followed by using Duolink Detection Reagents Red. Nuclei were stained with DAPI, and cells were mounted using Prolong Gold reagent (Invitrogen). Alexa Fluor 488-phalloidin incubation was used to detect actin structure. Fluorescent signals were obtained by the Carl Zeiss Observer Z1 microscope with a ϫ100 oil objective.

RESULTS
␤ Isoform of SH2B1 Is the Major Isoform during in Vitro Culture of Hippocampal Neurons-To study the role of SH2B1 during the development of the central nervous system, we first determined whether the splice variants of SH2B1 are differentially expressed during the development of neurites. Hippocampal neu-rons were isolated from embryonic day 18 (E18) of rat brain and cultured in vitro. RNA levels of the four splice variants of SH2B1 from DIV 1, 3, 8, and 14 hippocampal neurons were compared via RT-PCR. As shown in Fig. 1A, mRNA level of ␤ isoform increased on DIV 3, remained high through DIV 8, and decreased by DIV 14. Protein levels of SH2B1 were also up-regulated during early development of hippocampal neurons and a more than 2-fold increase on DIV 7 (Fig. 1B). These results indicate that the expression of SH2B1 is increased, and SH2B1␤ is the predominant isoform during neural differentiation.
SH2B1␤ Increases Neurite Branches of Hippocampal Neurons-To provide better analysis of the effect of SH2B1 on morphogenesis of hippocampal neurons, E18 hippocampal The number of end points was counted from 12 to 18 hippocampal neurons per condition. PC12 cells stably expressing shLacZ or shSH2B1 constructs (shSH2B1#3, #4, #5, #6, and #7) were treated with 100 ng/ml NGF for 1 day. Length of the longest neurite of each cell was calculated from three independent experiments. *, p Ͻ 0.05 by paired Student's t test compared with shLacZ control. Lysates from PC12-shLacZ and PC12-shSH2B1 cells were collected and immunoblotted with anti-SH2B1 and anti-␣-tubulin antibodies. B, hippocampal neurons were transiently transfected with either shLacZϩmCherry vector or shSH2B1#3ϩmCherry vector, together with GFP or GFP-SH2B1␤ on DIV 4, and live cell images were taken. The number of end points was counted from a total of 10 -13 hippocampal neurons per condition. PC12-shLacZ or PC12-shSH2B1(#3, #4, #5, #6, and #7) cells were transiently transfected with GFP or GFP-SH2B1␤ constructs. Transfected cells were then treated with NGF for 1 day, and the length of the longest neurite of each cell was measured from three independent experiments. *, p Ͻ 0.05 by paired Student's t test compared with PC12-shLacZϩGFP control or with its respective control. C, hippocampal neurons were transiently transfected with either shLacZ or shSH2B1 together with pEGFP vector on DIV 4, and images of neuronal morphology on days 1-3 after transfection were taken. The number of attachment points or end points on days 1-3 after transfection was measured. Enlarged images of the neurite branches are shown on the right panels. Scale bar, 40 m. A total of 12 hippocampal neurons were counted per condition from three independent experiments. Values are mean Ϯ S.E. from three independent experiments and statistically compared by Student's t test (*, p Ͻ 0.05, shSH2B1 compared with shLacZ). D, hippocampal neurons were transfected with either shLacZ or shSH2B1 together with pEGFP vector on DIV 4, and the images of neuronal morphology of were visualized on DIV 7. Immunofluorescence staining was performed with MAP2 (red), GFP (green), and DAPI (blue) antibodies. Cell lysates from cultured hippocampal neurons transiently transfected with shLacZ or shSH2B1(#3) were collected and analyzed via Western blotting with anti-SH2B1 or anti-␤III tubulin antibody.
neurons were transiently transfected with GFP or GFP-SH2B1␤ on DIV 4, and morphology of neurons was imaged on DIV 5-7. Quantification of the attachment points and end points, symbolic of neurites and neurite branches, respectively (28), is shown in Fig. 2A. Overexpression of SH2B1␤ increased the numbers of attachment points and end points (Fig. 2B). If SH2B1 is required for neurite outgrowth, knocking down SH2B1 would reduce neurite outgrowth. Five shSH2B1 constructs were tested for the efficiency of knocking down SH2B1. Hippocampal neurons were transfected with GFP together with either shLacZ or shSH2B1(numbers 3-7). Endogenous SH2B1 was differentially reduced by these shSH2B1, and the numbers of end points were reduced. The effect of shSH2B1 on neurite length of PC12 cells were also reduced (Fig. 3A). The relative reduction of endogenous SH2B1 protein by each shSH2B1 construct is shown (Fig. 3A, bottom panel). In the following experiments, we used shSH2B1#3 for morphogenesis analysis. To verify that the reduced end points and neurite length resulted specifically from knockdown of SH2B1, rescue experiments were performed. Hippocampal neurons were transiently transfected with shLacZ or shSH2B1 plus mCherry vector to mark the transfected neurons, together with GFP or GFP-SH2B1␤ overexpression to examine the rescue effect. End points were counted and compared. As shown in Fig. 3B, shSH2B1 reduced the number of end points, and expressing GFP-SH2B1␤ rescued the neurite branching back to the level of shLacZ control. Similar experiments were performed using PC12 cells and putting back SH2B1␤ also rescued the reduced neurite length by shSH2B1 (Fig. 3B). By comparing the attachment and end points, knockdown of SH2B1 significantly reduced the number of end points most dramatically, with 75% reduction 3 days after transfection compared with shLacZ control (Fig. 3C). Moreover, knocking down shSH2B1 led to the reduced numbers of primary dendrites as Correlation between paired pixels of IRSp53 and SH2B1␤ was also displayed as normalized mean deviation product color map. Indexes above 0 are represented in red, indicating co-localization. Scale bar, 40 m. D, whole cell lysates and membrane fractionation levels of SH2B1 and IRSp53 on DIV 2, 4, 7, and 10 were quantified (*, p Ͻ 0.05, Student's t test, compared with DIV 2). E, cell lysates from hippocampal neurons of DIV 2, 4, and 6 were collected. Triton X-100 detergent of soluble G-actin and insoluble F-actin fractions was equally subjected to immunoblotting with anti-SH2B1, IRSp53, and actin. The relative expression levels were quantified on DIV 2, 4, and 6 in soluble and insoluble fractions (*, p Ͻ 0.05, Student's t test, compared with DIV 2). MARCH 6, 2015 • VOLUME 290 • NUMBER 10 demonstrated by the staining of the dendritic marker, microtubule-associated protein 2 (MAP2) (Fig. 3D). These findings suggest that SH2B1 is required for the growth of dendrites and their branches.

SH2B1 and IRSp53 Promote Dendritic Branches
SH2B1 Interacts with IRSp53-The formation of actin-rich filopodia precedes the formation of neurites (1). To determine whether SH2B1 may promote dendrite formation by enhancing filopodium formation, we set out to examine the expression of a known filopodium-regulating protein, IRSp53. For hippocampal neurons cultured in vitro, the expressions of endogenous IRSp53 increased. Similarly, the expressions of SH2B1 and the differentiation marker growth-associated protein 43 (GAP43) were increased over time (Fig. 4A). IRSp53 is an IMD domaincontaining protein known to promote membrane protrusion. To examine whether SH2B1 is in the IRSp53-containing complexes, endogenous IRSp53 was immunoprecipitated from the lysate of cortical neurons, and the presence of endogenous SH2B1 was examined. As demonstrated in Fig. 4B, SH2B1 existed in the IRSp53-containing complexes.
Whether SH2B1 actively participates in the IRSp53-dependent membrane protrusion, they should co-localize at the plasma membrane fraction. Hippocampal neurons were transiently transfected with GFP-SH2B1␤, and immunofluorescence staining of IRSp53 was performed to visualize the potential co-localization of SH2B1 and IRSp53. As revealed in Fig. 4C, GFP-SH2B1␤ and IRSp53 co-localize mostly at the branch points as well as the tips of neurites. The co-localization between SH2B1 and IRSp53 was also analyzed by Pearson's correlation coefficient (ϭ 0.8) and showed a high level of co-localization. The top 2% overlapped pixels of GFP-SH2B1␤ and IRSp53 on the neurite branches are shown in white (Fig. 4C,  panel iii). To complement the single cell assays, fractionation  1-670), or truncated SH2B1␤ and Myc vector or myc-IRSp53 constructs were transiently co-transfected to 293T cells. Cell lysates were collected and subjected to immunoprecipitation (IP) with anti-Myc antibody and immunoblotted (IB) with anti-Myc and GFP antibodies. In the reverse co-immunoprecipitation experiment, myc-SH2B1␤ with or without IRSp53 was transfected to 293T cells. Cell lysates were collected and subjected to immunoprecipitation with anti-Myc antibody and immunoblotted with anti-Myc or anti-IRSp53 antibody. GST-SH2B1␤ was purified via glutathione-Sepharose beads. Cell lysates from 293T cells transiently transfected with Myc or myc-IRSp53 were incubated with Sepharose beads only or beads conjugated with GST-SH2B1␤. Pulldown of the GST-SH2B1␤-containing complex was subjected to Western blotting with anti-SH2B1 and anti-IRSp53 antibodies. * indicates the specific bands. B, hippocampal neurons were fixed, permeabilized, and incubated with anti-SH2B1 and anti-IRSp53 antibodies followed by reaction with Duolink PLA probes and reagents. Neurons incubated with Duolink PLA probes served as the negative control. Images were taken using Carl Zeiss Observer Z1 microscope. F-actin and nucleus were visualized by staining with Alexa Fluor 488-phalloidin and DAPI, respectively. Scale bar, 10 m. C, 293T cells were transiently transfected with IRSp53 and HA-Cdc42WT (wild-type Cdc42) or HA-Cdc42(Q61L) (constitutively active form of Cdc42). Lysates were collected and subjected to immunoprecipitation using anti-HA antibody followed by immunoblotting with anti-IRSp53 and anti-HA antibodies. D, 293T cells were transiently transfected with myc-SH2B1␤ and HA-Cdc42WT or HA-Cdc42(Q61L). Lysates were collected and immunoprecipitated with anti-HA antibody followed by immunoblotting with anti-HA and Myc antibodies. experiments were performed to examine the subcellular distribution of endogenous SH2B1 and IRSp53 in the plasma membrane fraction. As the expression of the endogenous SH2B1 and IRSp53 increased during the formation of the neurite network of hippocampal neurons (Fig. 4A), SH2B1 and IRSp53 increased 2-fold at the plasma membrane fraction on DIV 7 and DIV 10 (Fig. 4D, right panel). The initiation and formation of the prospective neurites require the assembly of G-actin to F-actin. To delineate whether SH2B1-IRSp53 complexes may reside with polymerized F-actin, cell lysates of hippocampal neurons from DIV 2 to 6 were collected for fractionation to separate detergent-soluble and -insoluble fractions. The endogenous SH2B1 and IRSp53 were obviously in the detergent-insoluble fraction containing polymerized F-actin on DIV 4 and DIV 6 compared with DIV 2 (Fig. 4E).
The fact that SH2B1␤ and IRSp53 interact and co-localize, we next examined which region(s) of SH2B1␤ is required for their interaction. To this end, SH2B1 or various truncation FIGURE 6. Proline-rich domains of SH2B1␤ are required to enhance filopodium formation and neurite outgrowth in hippocampal neurons. A, schematics of full-length (1-670) and various truncated SH2B1␤ structures. Amino acid numbers are shown. Top two panels, GFP (panel i), full-length (panel ii), or truncated form (panels iii-vi) of SH2B1␤ constructs were transiently transfected to hippocampal neurons on DIV 3, and images of the live neurons were taken on DIV 7 using Zeiss Observer Z1 microscope. Enlarged images of the neurite branches and filopodium formation are shown on the right panels. Scale bar, 20 m. A total of 15 cells were counted per condition from three independent experiments. Values are mean Ϯ S.E. from three independent experiments (*, p Ͻ 0.05, Student's t test, overexpressing full-length and truncated form of SH2B1␤ constructs compared with GFP). Bottom two panels, GFP (panel i), full-length (panel ii), or truncated form (panels iii-vi) of SH2B1␤ constructs were transiently transfected to hippocampal neurons on DIV 4, and the morphology of the neurons was subjected to immunofluorescence staining with anti-MAP2 (light blue) and Tau-1 (red) antibodies together with DAPI (blue) on DIV 7. The arrows show the overexpression of GFP, full-length, or truncated form of SH2B1␤ in hippocampal neurons. Scale bar, 40 m. B, PC12 cells stably expressing shLacZ (PC12-shLacZ) or shIRSp53 (PC12-shIRSp53) were established. Cell lysates from PC12-shLacZ, PC12-shEps8, and PC12-shIRSp53 cells were collected and immunoblotted with anti-IRSp53 and ␣-tubulin antibodies. Left two panels, PC12-shLacZ and PC12-shIRSp53 cells treated with 50 ng/ml NGF for 6 days were subjected to immunofluorescent staining with anti-␤III tubulin (green) antibody together with DAPI (blue). Scale bar, 20 m. Right two panels, PC12-shLacZ and PC12-shIRSp53 cells were transiently transfected with GFP-SH2B1␤ and then treated with 50 ng/ml NGF for 1 day before immunofluorescent staining with anti-␤III tubulin (red) and GFP (green) antibodies together with DAPI (blue). The arrows show overexpression of GFP-SH2B1␤ in PC12-shLacZ and PC12-shIRSp53 cells. Scale bar, 20 m. A total 100 -105 cells were counted per condition from three independent experiments. Values are mean Ϯ S.E. from three independent experiments (*, p Ͻ 0.05, Student's t test, PC12-shIRSp53 compared with PC12-shLacZ). mutants of SH2B1␤, including SH2B1␤(1-150), SH2B1␤(270 -670), SH2B1␤(397-670), and SH2B1␤(505-670), were used to transfect 293T cells together with myc-IRSp53. IRSp53 was immunoprecipitated, and the presence of SH2B1␤ or its mutants was examined. As in Fig. 5A, SH2B1␤ and SH2B1␤(1-150) interacted with IRSp53. SH2B1␤(270 -670), SH2B1␤-(397-670), and SH2B1␤(505-670) did not. This result suggests that SH2B1␤ likely interacts with IRSp53 through its N-terminal proline-rich domains. Reverse co-immunoprecipitation was also performed through immunoprecipitating myc-SH2B1␤ and re-probed with anti-IRSp53. In vitro GST-SH2B1␤ pulldown assay also detected IRSp53 in the same complex (Fig. 5A,  right panels). To confirm the in vitro interaction results, Duolink in situ PLA assays were used to examine the in vivo association between endogenous SH2B1␤ and IRSp53 in hippocampal neurons. Hippocampal neurons were incubated with anti-SH2B1 and anti-IRSp53 antibodies followed by reaction with in situ PLA assay probes and reagents. The interaction between SH2B1 and IRSp53 was detected by red fluorescent puncta. In addition, actin structure was visualized by Alexa Fluor 488-phalloidin staining. As shown Fig. 5B, the SH2B1-IRSp53 complexes existed at the plasma membrane, actin, and filopodium tips. During the elongation of actin filaments, IRSp53 recruits VASP to the site of action, and this recruitment is regulated by Cdc42 (29,30). To examine whether IRSp53 interacts with Cdc42, 293T cells were transfected with IRSp53 together with HA-Cdc42 or the constitutive active form of Cdc42, HA-Cdc42(Q61L). When immunoprecipitating with anti-HA antibody, IRSp53 was in the same complex with HA-Cdc42(Q61L) (Fig. 5C). This result raised the possibility that SH2B1 may interact with Cdc42 or VASP instead of directly with IRSp53. Nonetheless, SH2B1␤ does not interact with VASP (31). Thus, we examined the possible interaction of SH2B1␤ and Cdc42. As shown in Fig. 5D, SH2B1␤ did not interact with either wild-type Cdc42 or the constitutive active form, Cdc42(Q61L).
Proline-rich Domains of SH2B1␤ Are Required to Enhance Filopodium Formation and Neurite Outgrowth in Hippocampal Neurons-To determine which region(s) of SH2B1␤ is required for enhancing filopodium formation and neurite outgrowth, various truncation mutants of GFP-SH2B1␤ were transfected to hippocampal neurons, and the numbers of attachment points and end points were quantified. As shown in Fig. 6A, overexpressing full-length GFP-SH2B1␤ significantly enhanced the numbers of attachment points and end points in hippocam-pal neurons compared with control GFP cells. SH2B1␤ truncation mutants cannot increase the attachment points. In contrast, overexpressing these mutants significantly reduced the number of end points of hippocampal neurons compared with GFP-expressing cells. It is possible that SH2B1␤(1-150) titrates out the limiting amount of endogenous IRSp53, whereas SH2B1␤(270 -670) and SH2B1␤(397-670) may bind to and sequester other filopodia regulating proteins. Overexpression of SH2B1␤(505-670), lacking all three N-and C-terminal proline-rich domains, inhibited both attachment and end points. This finding could result from the SH2 domain only mutant, SH2B1␤(505-670), sequestering the neurotrophin receptors required for neurite outgrowth. Neuronal markers of dendrites and axons were also determined. Fixed neurons were subjected to immunofluorescence staining with anti-MAP2 (dendritic marker), Tau-1 (axonal marker), and DAPI. Overexpressing truncated mutants of SH2B1␤ significantly reduced the dendritic or axonal filopodia and neurite branching compared with overexpressing full-length GFP-SH2B1␤ cells. Interestingly, overexpressing GFP-SH2B1␤ lacking both N-and C-terminal proline-rich domains strongly blocked the filopodium formation and neurite outgrowth. These results suggest that the Nand C-terminal proline-rich domains of SH2B1␤ are required for enhancing the formation of filopodia and neurite branching during neural development. To determine whether SH2B1␤mediated neurite outgrowth depends on IRSp53, the stable cell line PC12-shLacZ and PC12-shIRSp53 were established. As shown in the immunoblot, the protein level of IRSp53 was significantly reduced (Fig. 6B). Neurite outgrowth was inhibited in PC12-shIRSp53 cells and cannot be rescued by putting back GFP-SH2B1␤. These results suggest that IRSp53 is required for SH2B1␤-enhanced filopodium formation and neurite outgrowth.
N-terminal Proline-rich Domains of SH2B1␤ Enhance IRSp53induced Neurite Outgrowth-To examine whether interaction between SH2B1 and IRSp53 is required to enhance IRSp53induced neurite outgrowth, hippocampal neurons were transiently transfected with myc-IRSp53 together with GFP, GFP-SH2B1␤, GFP-SH2B1␤(1-150), GFP-SH2B1␤(1-260), GFP-SH2B1␤(270 -670), or GFP-SH2B1␤(397-670). The morphology of neurons was examined, and the numbers of attachment points as well as end points were quantified. The co-expression of SH2B1 and IRSp53 is shown in Fig. 7A. Overexpressing SH2B1␤ and IRSp53 increased both attachment (1.7-fold) and end (2.5-fold) points, whereas overexpression of FIGURE 7. SH2B1␤ and IRSp53 synergistically regulate the formation of dendritic and axonal filopodia during neurite outgrowth. A, GFP (panel i), full-length (panel ii) or truncated form (panels iii-vi) of SH2B1␤ constructs were transiently co-transfected with myc-IRSp53 to hippocampal neurons on DIV 4, and the images of neuronal morphology were taken on DIV 7. Neurons were subjected to immunofluorescent staining with anti-Myc (red) and GFP (green). Enlarged images of the neurite branches and filopodium formation for GFP channel are shown in the right panels. Scale bar, 40 m. A total of 15 cells were counted per condition from three independent experiments. Values are mean Ϯ S.E. from three independent experiments (*, p Ͻ 0.05, Student's t test). PC12 cells stably expressing GFP or GFP-SH2B1␤(1-150) were immunoprecipitated using anti-IgG or anti-GFP antibodies. Immunoprecipitated complexes were resolved and immunoblotted using antibodies against anti-GFP and anti-SH2B1 antibodies. B, GFP (panel i), full-length (panel ii), or the truncated form (panels iii-vi) of SH2B1␤ constructs were transiently co-transfected with myc-IRSp53 to hippocampal neurons on DIV 4, and the morphology of the neurons was visualized on DIV 7. Enlarged images of the dendritic filopodia and branches are shown in the right panels. Neurons were subjected to immunofluorescence staining with anti-MAP2 (Cy5) and GFP (green) antibodies and F-actin (red). The end points of MAP2ϩ or MAP2-/F-actinϩ signal were quantified to analyze the filopodial and dendritic branches. MAP2 was highlighted in light blue for observation. C, GFP (panel i), full-length (panel ii), or truncated from (panels iii-vi) of SH2B1␤ constructs were transiently co-transfected with myc-IRSp53 to hippocampal neurons on DIV 4, and the morphology of the neurons was visualized on DIV 7. Enlarged images of the axonal filopodia are shown in the right panels. Neurons were subjected with immunofluorescence staining with anti-GFP (green) and Tau-1 (red) antibodies together with DAPI (blue). Scale bar,ϭ 40 m. Panel vii, axonal filopodia were quantified per 100 m. Total 12 cells were counted per condition from three independent experiments. Values are mean Ϯ S.E. from three independent experiments (*, p Ͻ 0.05, Student's t test).

DISCUSSION
Filopodia are generally known to sense environmental cues and to regulate tissue morphogenesis and regeneration. However, exactly how filopodia are initiated at a specific locale remains unclear. IRSp53 has been well documented in promoting membrane protrusion and filopodium formation. In this study, we demonstrate that SH2B1 interacts with IRSp53 and promotes neurite outgrowth and neuronal branching. Based on the known interaction between SH2B1 and neurotrophin receptors at the cell surface during neuronal differentiation (18,19,(23)(24)(25), these results raise the possibility that SH2B1-IRSp53 complexes specify locations for filopodium formation through SH2B1-mediated signaling events. In line with this possibility, overexpression of SH2 domain only mutant, SH2B1␤(505-670), inhibits neurite outgrowth (Fig. 6A) potentially through sequestering neurotrophin receptors.
The process from membrane protrusion (actin-free) to the formation filopodia (actin-rich) is very dynamic (32). Stabilizing the structure of filopodia is essential for neurite initiation. The fact that overexpression of SH2B1␤ and IRSp53 promotes the formation of actin-positive (filopodia) and MAP2-negative (immature dendrites) end points suggests that SH2B1␤ may be recruited or targeted to the plasma membrane to participate in IRSp53-induced membrane protrusion process. Amino acids 150 -260 of SH2B1␤ contain membrane targeting and actinbinding regions (33,34). Nonetheless, SH2B1␤(1-150) and SH2B1␤(1-260) showed a similar effect on dendritic and axonal branches (Fig. 7, B and C). Thus, it is likely that SH2B1 is recruited to the plasma membrane by mechanisms other than using its own membrane targeting region. We show in this study that SH2B1␤(1-150), containing two proline-rich regions as well as the dimerization domain, brings together the endogenous SH2B1 to action (Fig. 7A).
During actin elongation, IRSp53 is known to recruit VASP to the site of action, and this recruitment is regulated by Cdc42 (29,30). Nonetheless, SH2B1␤ does not interact with either VASP (31) or Cdc42 (Fig. 5D). An IRSp53-interacting protein is Eps8 (epidermal growth factor receptor kinase substrate 8). Eps8 can cap barbed ends of actin filaments when binding to Abi-1, whereas the association of Eps8 with IRSp53 induces filopodium formation by cross-linking actin filaments (30,35,36). These findings suggest that Eps8 can induce or inhibit the filopodium formation depending on its binding proteins. Although published reports suggest the involvement of Eps8 mainly in spine formation, it would be of interest to investigate whether Eps8 is in the complex of SH2B1-IRSp53 to promote filopodium formation and thus spine formation.