The neuronal Arf GAP centaurin α1 modulates dendritic differentiation

Centaurin α1 is an Arf GTPase-activating protein (GAP) that is highly expressed in the nervous system. In the current study, we show that endogenous centaurin α1 protein is localized in the synaptosome fraction, with peak expression in early postnatal development. In cultured dissociated hippocampal neurons, centaurin α1 localizes to dendrites, dendritic spines and the postsynaptic region. siRNA-mediated knockdown of centaurin α1 levels or overexpression of a GAP-inactive mutant of centaurin α1 leads to inhibition of dendritic branching, dendritic filopodia and spine-like protrusions in dissociated hippocampal neurons. Overexpression of wild-type centaurin α1 in cultured hippocampal neurons in early development enhances dendritic branching, and increases dendritic filopodia and lamellipodia. Both filopodia and lamellipodia have been implicated in dendritic branching and spine formation. Following synaptogenesis in cultured neurons, wild-type centaurin α1 expression increases dendritic filopodia and spine-like protrusions. Expression of a GAP-inactive mutant diminishes spine density in CA1 pyramidal neurons within cultured organotypic hippocampal slice cultures. These data support the conclusion that centaurin α1 functions through GAP-dependent Arf regulation of dendritic branching and spines that underlie normal dendritic differentiation and development.

Of the six Arf genes identified, Arf1 and Arf6 are the best characterized. Arf1 is involved in vesicular trafficking in the Golgi, trans-Golgi network (TGN) and endosomal compartments whereas Arf6 regulates trafficking in the plasma membrane/recycling endosomal compartment and cytoskeletal organization at the cell periphery (Donaldson and Jackson, 2000;Nie and Randazzo, 2006;D'Souza-Schorey and Chavrier, 2006). Arfs facilitate coat protein assembly onto budding membranes, and activate PtdIns(4)P 5-kinase and phospholipase D (PLD). Arf6 has also been shown to directly affect the Rho-GTPase family member, Rac1 (D' Souza-Schorey and Chavrier, 2006;Cotton et al., 2007). Arfs are regulated by GTP-exchange factors (GEFs), which facilitate GTP binding, and by Arf GAPs which stimulate GTP hydrolysis (Donaldson and Jackson, 2000;Nie and Randazzo, 2006). Arf GAPs have also been proposed to function with Arfs in coat assembly and disassembly (Nie and Randazzo, 2006).
In addition to its characterized roles in trafficking, Arf6 has been implicated in neuronal differentiation. Expression of Centaurin ␣1 is an Arf GTPase-activating protein (GAP) that is highly expressed in the nervous system. In the current study, we show that endogenous centaurin ␣1 protein is localized in the synaptosome fraction, with peak expression in early postnatal development. In cultured dissociated hippocampal neurons, centaurin ␣1 localizes to dendrites, dendritic spines and the postsynaptic region. siRNA-mediated knockdown of centaurin ␣1 levels or overexpression of a GAP-inactive mutant of centaurin ␣1 leads to inhibition of dendritic branching, dendritic filopodia and spine-like protrusions in dissociated hippocampal neurons. Overexpression of wild-type centaurin ␣1 in cultured hippocampal neurons in early development enhances dendritic branching, and increases dendritic filopodia and lamellipodia. Both filopodia and lamellipodia have been implicated in dendritic branching and spine formation. Following synaptogenesis in cultured neurons, wild-type centaurin ␣1 expression increases dendritic filopodia and spine-like protrusions. Expression of a GAP-inactive mutant diminishes spine density in CA1 pyramidal neurons within cultured organotypic hippocampal slice cultures. These data support the conclusion that centaurin ␣1 functions through GAPdependent Arf regulation of dendritic branching and spines that underlie normal dendritic differentiation and development.
To date, the requirement for an endogenous Arf GAP in neuronal differentiation has not been established. In the current study, we have characterized the developmental expression of endogenous centaurin ␣1 and tested its requirement for differentiation in cultured neurons and neuronal slices. Centaurin ␣1 is expressed in the brain during prenatal and early postnatal development. It localizes to developing dendrites and following differentiation is localized in dendritic spines and synaptic regions. Knocking down centaurin ␣1 levels by siRNA or expression of a GAP-inactive mutant centaurin ␣1 leads to a dramatic decrease in dendritic branching and dendritic spines. Overexpression of centaurin ␣1 shows the opposite phenotype with increased dendritic arborization and spine-like protrusions. The results reported herein support the function for centaurin ␣1 in Arf-dependent regulation of dendritic differentiation.

Results
Developmental expression and localization of endogenous centaurin ␣1 in rodent brain In the present study, we investigated the developmental expression and localization of endogenous centaurin ␣1 in rodent brain and cultured neurons. Centaurin ␣1 protein expression is detectable as early as E16 and increases during prenatal and postnatal brain development (Fig. 1A). Peak expression is observed between 2 and 4 weeks postnatally in brain and expression persists in the adult at approximately half the peak developmental levels. This protein expression is in concordance with published mRNA and protein expression data (Aggensteiner and Reiser, 2003) and corresponds to a period of robust neuronal differentiation and synaptogenesis in the hippocampus (Melloni and DeGennaro, 1994). We detected two closely migrating protein bands, irrespective of the specific antibody or lysis protocol used, which showed variable resolution. Whether this doublet results from post-translational processing, alternative splicing, or proteolysis is currently under investigation.
In fractionated P23 rat brain, endogenous centaurin ␣1 is present in the supernatant (S2) and crude microsomal and synaptosomal fraction (P2) in approximately equal levels, consistent with previous results showing both cytosolic and membrane association for centaurin ␣1. Further fractionation of P2 was performed to yield the purified synaptosome fraction, an enriched synaptic fraction that contains both presynaptic proteins, including syntaxin and synapsin, and post-synaptic proteins such as PSD95 (Fig. 1B) (Gordon-Weeks, 1987). Centaurin ␣1 is also present in purified synaptosomes, with an enrichment over P2 comparable with the other synaptic proteins examined (Fig. 1B). The centaurin ␣1 protein associated with synaptosomes showed a peak developmental expression at P28, similar to the expression of total centaurin ␣1 protein (Fig. 1C). Centaurin ␣1 expressed in the adult is also associated with the synaptosome fraction. This result is in agreement with immunohistochemistry results that show centaurin-␣1/p42IP4 localizes to synaptic regions in neuronal sections (Kreutz et al., 1997;Sedehizade et al., 2002) and is consistent with the identification of centaurin ␣1 in the post-synaptic density fraction by proteomic analysis (Yoshimura et al., 2004).
Expression of endogenous centaurin ␣1 in dissociated cultured hippocampal neurons Centaurin ␣1 is expressed in the cell soma and processes of hippocampal neurons in the CA1, CA3 and dentate gyrus regions ( Fig. 1D and data not shown). Cultured hippocampal neurons provide an excellent model system to study neuronal differentiation and synaptogenesis (Goslin and Banker, 1989;Papa et al., 1995), therefore we examined the expression and localization of centaurin ␣1 in primary hippocampal cultures. Neurons were dissociated from E18 rat brain and cultured from 3 to 21 days in vitro (DIV). Centaurin ␣1 protein expression, measured by immunoblot analysis, could be detected at 3 DIV and expression increased in culture with maximal expression at the latest time examined, 21 DIV ( Fig. 2A). Thus, the expression of centaurin ␣1 in cultured neurons paralleled its in vivo expression. Furthermore, the expression of centaurin ␣1 in developing cultured hippocampal neurons also mirrored that of the dendritic spine marker spinophilin, the presynaptic marker synaptotagmin and the postsynaptic marker PSD95 ( Fig. 2A).
Journal of Cell Science 120 (15) Fig. 1. Developmental expression and localization of endogenous centaurin ␣1 in the rodent brain. (A) Immunoblot analysis of centaurin ␣1 protein from lysates (25 g total lysate per lane) from whole brain at different developmental stages from E10 to P42 (=adult). (B) Immunoblot analysis of centaurin ␣1 and synaptic markers in fractionated P23 whole rat brain. H, whole brain homogenate; P2, crude microsomes and synaptosomes; S2, supernatant; Syn, purified synaptosomes; 20 g of each fraction was loaded per lane. (C) Immunoblot analysis of centaurin ␣1 in the synaptosome fraction prepared from the developing brain. 10 g of each fraction was loaded per lane. (D) Immunohistochemistry with anti-centaurin ␣1 antibody in hippocampus from P28 rat. DG, dentate gyrus.
At 3 DIV, neurons extend short immature processes. In these neurons, centaurin ␣1 localized in a punctate pattern in the cell soma, as well in the numerous short immature processes ( Fig.  2B and supplementary material Fig. S1). Punctate staining in undifferentiated neurons is consistent with association of centaurin ␣1 with membrane compartments. By 7 DIV, after processes have begun to differentiate into the lone axon and multiple dendrites, centaurin ␣1 was localized in puncta throughout the cell soma, and in the numerous dendritic-like processes, but not in the axon. Dendrites typically branch repeatedly to form a complex tree-like structure (arbor) that received inputs from other neurons at the synapse. After synapse formation by 14 and 21 DIV, centaurin ␣1 was localized throughout the processes, and often localized in large puncta, suggestive of a spine or synaptic location (see below).
Localization of endogenous centaurin ␣1 in dissociated cultured hippocampal neurons The localization of centaurin ␣1 was compared with wellcharacterized neuronal marker proteins, the dendritic marker MAP2, the axonal marker tau, two presynaptic markers: synaptotagmin and SV2, the spine marker spinophilin, and the excitatory postsynaptic marker PSD95 (Huber and Matus, 1984). Levels of colocalization between centaurin ␣1 and these marker proteins in dendritic segments were quantified (Sailer et al., 2004;Swanwick et al., 2004;Zhang et al., 2006). At 7 and 14 DIV, centaurin ␣1 localized to both the cell soma and to MAP2 positive dendrites (Fig. 3). Interestingly, although centaurin ␣1 localized within the MAP-2 positive dendrites, only an average 21% of centaurin ␣1 staining colocalized with MAP2. Centaurin ␣1 was usually found at the lateral regions of the dendrite and to a lesser extent in the central core region where microtubules, stained with beta-tubulin and MAP2 are located ( Fig. 3 and supplementary material Fig. S2) (Kaech et al., 2001). The centaurin ␣1 that did colocalize with MAP2 was often localized near dendritic branch points (Fig. 3). By contrast, centaurin ␣1 did not appear to localize to axonal processes marked by tau (Fig. 3). Approximately 15% colocalization with tau was observed (supplementary material Fig.  S2) (Binder et al., 1985). We estimate an approximate 5-7% nonspecific or background colocalization between proteins known to reside in different compartments. The 15% level of colocalization is above background but probably represents the overlap of closely apposed staining between dendrites and axons (see below). Centaurin ␣1 staining within MAP2positive dendrites was punctate and found closely juxtaposed to, but not usually colocalizing with, the presynaptic markers synaptotagmin and the synaptic vesicle-2 protein (SV2) (   apposition between centaurin ␣1 (found in dendrites) with the presynaptic markers was suggestive of a postsynaptic localization for centaurin ␣1. We observed approximately 18% colocalization between centaurin ␣1 and either of the presynaptic markers in dendritic segments. A similar extent of colocalization (15%) has been reported between presynaptic and postsynaptic markers (Zhang et al., 2006). When centaurin ␣1 was compared with the postsynaptic marker PSD95 or the spine marker spinophilin, many puncta showed exact colocalization (Fig. 3). We measured an overall 28% colocalization with PSD95 and 30% colocalization with spinophilin ( Fig. 3 and supplementary material Fig. S2) (Hunt et al., 1996;Allen et al., 1997) in dendritic segments. This level of colocalization was similar to that reported for other synaptic and extrasynaptic proteins (Sailer et al., 2004;Swanwick et al., 2004), and consistent with the observation that centaurin ␣1 localizes to more than one compartment, throughout the dendrite, dendritic spines and the post-synaptic region.
Knockdown of centaurin ␣1 by siRNA in dissociated cultured hippocampal neurons The developmental expression and dendritic localization highlighted the possibility that centaurin ␣1 might function in dendritic differentiation in neurons. To address this, endogenous centaurin ␣1 levels were knocked down in hippocampal neurons using double-stranded siRNA transfection by Amaxa nucleofection, which yields >90% transfection efficiency of siRNA (Gresch et al., 2004) (Fig. 4). Following centaurin-␣1specific siRNA treatment, there was a significant decrease in centaurin ␣1 protein levels of 60% at 3 DIV and a more moderate decrease of 40-45% at 7 DIV as assessed by immunoblot analysis (Fig. 4A). Centaurin ␣1 siRNA transfection resulted in a reduction of centaurin ␣1 staining by immunofluorescence, with a substantial depletion of centaurin ␣1 from dendritic processes but a relative sparing of some perinuclear centaurin ␣1 staining, suggesting that different pools of centaurin ␣1 may have different turnover rates (Fig. 4C).
At 3 S3). We measured axonal and dendritic length and the number of terminal tips -a well-established measure of branching (Uylings and van Pelt, 2002;Kumar et al., 2005). Decreasing centaurin ␣1 levels at 3 DIV inhibited dendritic branching and length but did not affect the axon (Fig. 4C). Knockdown of centaurin ␣1 produced an average 70% decrease in the number of terminal tips and a 73% inhibition of dendritic length at 3 DIV ( Fig. 4D). At 7 DIV, the terminal tips were reduced by 46%, and the dendritic length was reduced by 34%, suggesting that these processes were recovering as the centaurin ␣1 levels recovered. As with 3DIV, no statistically significant effect on axonal length was observed at 7DIV. By 14 DIV, the levels of centaurin ␣1 protein had returned to control (Fig. 4A), and the dendritic morphology appeared normal, indicating that the effects of centaurin ␣1 knockdown were reversible (supplementary material Fig. S3). Knockdown of centaurin ␣1 at 3 and 7 DIV also inhibited lamellipodia and filopodia (Fig. 4). Lamellipodia are thin, broad, veil-like extensions that contain a dynamic array of branched actin filaments. Filopodia are finger-like projections containing tight bundles of actin. Both structures are crucial to the growth and differentiation of neuronal processes, and are shown to be involved in dendritic branching and spine formation (reviewed by Luo, 2002;Chen and Ghosh, 2005).
To demonstrate the specificity of the siRNA effects, we tested whether expression of an siRNA-resistant cDNA plasmid could rescue the siRNA-dependent inhibition of dendritic outgrowth and branching. The human centaurin ␣1 cDNA is divergent from the rat cDNA in several nucleotides and therefore is more resistant to the rat siRNAs. Expression of the human centaurin ␣1 plasmid in siRNA-treated neurons produced an overexpression of human centaurin ␣1 of approximately three-to fivefold at 3 and 7 DIV (Fig. 4B). This expression fully restored dendritic length and branching in the siRNA-treated neurons at both 3 and 7 DIV (Fig. 4D,E). In fact, the rescued neurons showed a statistically significant 34% and 91% increase in number of terminal tips at 3 and 7 DIV, respectively, and a 64% and 102% increase in dendritic length at 3 and 7 DIV, respectively, compared with control scrambled siRNA-treated neurons (Fig. 4E). There was not a statistically significant change in axon length. These data support the conclusion that centaurin ␣1 is required for early dendritic differentiation in cultured neurons.
We next assessed the effects of overexpressing wild-type and mutant rat centaurin ␣1 on neuronal morphology. Amaxa nucleofection yields approximately 50% transfection efficiency of mammalian expression plasmids in neurons (Gresch et al., 2004). Immunoblot analysis was used in parallel cultures to determine the levels of overexpression of centaurin ␣1 at 3, 7 and 14 DIV following transfection (supplementary material Fig. S4A). Centaurin ␣1 levels were normalized to the actin levels. Given the 50% transfection efficiency, we expect a nine-to tenfold increase of centaurin ␣1 levels at 3DIV, and a fourfold increase at 7 and 14 DIV in the transfected neurons (supplementary material Fig. S3 and Fig. 4).
Overexpression of wild-type centaurin ␣1 in developing dissociated hippocampal neurons The staining pattern of expressed centaurin ␣1, assessed with an anti-Flag antibody, was similar to the staining of the endogenous protein with localization in the cell body and dendrites in a punctate pattern. However, unlike the endogenous protein, expressed centaurin ␣1 also localized to the axon. Overexpression of centaurin ␣1 at 3 DIV led to an increase in filopodia and branches arising from the cell soma and the dendrites (Fig. 5A and supplementary material Fig.  S4). At 3 DIV, expression of centaurin ␣1 led to a statistically significant 22% increase in terminal tips (Fig. 5C). A statistically significant 15% increase in total dendritic length was also observed. By 7 DIV, centaurin ␣1 overexpression led to a 54% increase in the number of terminal tips and a 17% increase in the dendritic length. Expression of centaurin ␣1 also increased the size and number of lamellipodia and filopodia at 3 and 7 DIV (supplementary material Fig. S4). At 7 and 14 DIV, the expressed Flag-centaurin ␣1 colocalized with spine and synaptic markers to a similar extent as the endogenous protein (data not shown). Although localized to the axon, no effects on the axonal processes were observed at 3 or 7 DIV. By 14 DIV, centaurin ␣1 levels remained elevated, and there was also a clear increase in dendritic branching and filopodia (Fig. 5B); however, as the neuronal processes were so long and complex at this stage, we were unable to quantify these effects. Thus, expressed centaurin ␣1 localizes to the same compartments as endogenous centaurin ␣1 and enhances dendritic branching and filopodia in developing cultured neurons. Interestingly, rat centaurin ␣1 was not as effective as human centaurin ␣1 in enhancing dendritic outgrowth and branching in rat hippocampal neurons, which could be a result of differences in the proteins or their expression levels.

Overexpression of a GAP-inactive mutant centaurin ␣1
Centaurin ␣1 is an Arf GAP (Venkateswarlu et al., 2004). Mutation of the conserved arginine in the GAP domain has been shown to eliminate GAP activity in all Arf GAPs examined, including centaurin ␣1. Furthermore, expression of GAP mutants can result in inhibition of endogenous Arf GAPs (Randazzo and Hirsch, 2004;Venkateswarlu et al., 2004). Similarly to the expressed and endogenous wild-type protein, the GAP mutant localized to dendrites in a punctate pattern (Fig. 6) and colocalized with markers of spines and synapses (data not shown).
Overexpression of the GAP mutant R49K centaurin ␣1 in hippocampal neurons led to a significant decrease in dendritic branching and dendritic length at 3, 7 and 14 DIV ( Fig. 6A and supplementary material Fig. S5). Expression of R49Kcentaurin ␣1 at 3 DIV produced a 35% decrease in the number of terminal tips and a 30% decrease in dendritic length (Fig. 6B). At 7 DIV, a 30% decrease in terminal tips and a 22% decrease in dendritic length was observed (Fig. 6B). These effects were statistically significant (P<0.05). No significant effects on the axon length were observed at 3 DIV or 7 DIV. Likewise, the GAP inactive mutant did not appear to inhibit the size of lamellipodia. At 14 DIV, there was a discernible inhibition of dendritic branching, but as described above, the processes were so long and complex that the effects could not be quantified. Although not as robust, the inhibitory effects of the GAPinactive mutant were similar to those observed with siRNA, and consistent with the GAP-inactive mutant acting as a dominant-negative inhibitor. These results suggest that the GAP activity of centaurin ␣1 is required for its regulation of dendritic branching.
Overexpression of wild-type and GAP-inactive centaurin ␣1 At 14 DIV the centaurin ␣1 expression remained elevated about fourfold (supplementary material Fig. S4A). Thus, we could address whether centaurin ␣1 functions during later stages of dendritic differentiation, during dendritic filopodia and spine formation. Dendritic spines are regions where excitatory synapses form when filopodia are contacted by the axon (Tada and Sheng, 2006). At 14 DIV, neurons display mature spines, immature spines and dendritic filopodia. We classified all three structures as dendritic protrusions for analysis (Kumar et al., 2005). Neurons overexpressing wildtype centaurin ␣1 for 14 DIV showed a dramatic (73%) increase in the number of dendritic protrusions (Fig. 7). In addition, there was an increase in the number of long and branched filopodia in the centaurin ␣1 expressing neurons. By contrast, expression of the GAP-inactive mutant R49K centaurin ␣1 produced a statistically significant reduction in dendritic protrusions of 15% (Fig. 7). By this stage of development, few lamellipodia were observed in either control Journal of Cell Science 120 (15) or transfected neurons. These data suggest that centaurin ␣1 may function at later stages of dendritic differentiation in spine development or spine maintenance.
Centaurin ␣1 localizes to dendritic spines and regulates dendritic spine density in CA1 neurons To address this further, we overexpressed centaurin ␣1 or the GAP-inactive mutant in organotypic hippocampal slice cultures, which provide another valuable experimental preparation for investigating dendritic development (Gahwiler et al., 1997). An advantage of slice cultures is the preservation of neuronal integrity and circuitry, allowing neurons to undergo development in culture that parallels in vivo development (Lo et al., 1994;Alonso et al., 2004). Slices were cotransfected with eYFP and either wild-type centaurin ␣1 or R49K centaurin ␣1, or with dsRed as a control. CA1 pyramidal cells were identified by their characteristic morphology and location within the hippocampus (Fig. 8). In agreement with its endogenous localization in dissociated hippocampal cultures, expressed centaurin ␣1 was detected in dendritic regions and in dendritic spines (Fig. 8). Over 98% of cells expressing eYFP also expressed the cotransfected plasmid (data not shown), consistent with results showing that over 90% of neurons in slices transfected with two plasmids express both proteins (Klimaschewski et al., 2002;Wirth et al., 2003).
Expression of wild-type centaurin ␣1 in P9 slices for 36 hours did not affect dendritic spine density in CA1 pyramidal neurons compared with the dsRed control (8.5±0.5 spines/10 m with dsRed vs 8.0±0.5 spines/10 m with centaurin ␣1) (Fig. 8). Thus endogenous centaurin ␣1 levels may facilitate maximal differentiation in the P9-11 slice cultures. However, expression of the GAP-inactive mutant R49Kcentaurin ␣1 significantly reduced spine density by 39% (5.2±0.6 spines/10 m; P<0.00016 compared with dsRed control) (Fig. 8). Inhibition of dendritic spines by R49K centaurin ␣1 expression in organotypic slices was more robust than its effects in dissociated neurons and provides further evidence that centaurin ␣1 functions through its GAP activity to regulate spine formation or stability.

Discussion
The results reported herein demonstrate that centaurin ␣1 is a developmentally expressed Arf GAP that is required for dendritic differentiation in developing neurons. In the brain, centaurin ␣1 protein expression peaks during the first few postnatal weeks and is enriched in the purified synaptosome fraction. In developing hippocampal neurons, centaurin ␣1 localizes to dendrites, dendritic spines and synapses. Knockdown of centaurin ␣1 levels by siRNA in cultured dissociated hippocampal neurons significantly reduces dendritic branching and length, but the axon is unaffected. This inhibition is specific because it could be rescued by expression of siRNA-resistant human centaurin ␣1. Likewise, expression of a dominant-negative GAP mutant centaurin ␣1 in cultured neurons and organotypic hippocampal slices inhibits dendritic branching and spine density, respectively. By contrast, overexpression of wild-type centaurin ␣1 increases dendritic branching and dendritic protrusions. As a modulator of dendritic arborization and spine density, centaurin ␣1 may directly impact the number and dynamics of synapses in the developing brain.
Centaurin ␣1 protein was detectable in brain at E16, a stage of rapid proliferation, differentiation and migration of neuronal  cells. Centaurin ␣1 protein levels peaked between 2 and 4 weeks postnatally, in accordance with data reported for centaurin ␣1 mRNA (Aggensteiner and Reiser, 2003). Centaurin ␣1 expression increased in developing cultured neurons between 3 and 21 DIV. This temporal expression is similar to that reported for the spine marker spinophilin, the early presynaptic markers synapsin I and bassoon, and postsynaptic marker PSD-95 (Melloni and DeGennaro, 1994;Allen et al., 1997;Zhai et al., 2001;Bresler et al., 2001). This corresponds to a broad period of neuronal differentiation, including dendritic arborization, spine formation and synaptogenesis in the developing rat cortex and hippocampus (Gaarskjaer, 1981;Melloni and DeGennaro, 1994).
The results from siRNA knockdown and GAP mutant expression, and its localization in dendrites and spines during development support a role for centaurin ␣1 in both dendritic branching and spine regulation. There is now a strong precedent for proteins serving a dual function in both dendritic branching and spine dynamics. For example, members of the Rho-GTPase family, Rac1, Cdc42 and Rnd1/2 have been shown to function in both dendritic branching and spine development (see Govek et al., 2005;Watabe-Uchida et al., 2006;Ishikawa et al., 2006). PSD-95 is a protein involved in dendritic spine maturation that plays an additional role in regulating dendrite outgrowth and branching (Charych et al., 2006). The origin recognition core (ORC) complex protein is required for dendritic branching and initiating dendritic spine formation (Huang et al., 2005). In addition, EFA6, an Arf GEF, and Arf6 have been reported to regulate the development of dendrites and dendritic spines (Sakagami et al., 2004;Miyazaki et al., 2005;Choi et al., 2006). Many of these proteins, including centaurin ␣1, are also expressed in the adult brain and are localized to synapses. As regulators of dendritic spine dynamics, these proteins and centaurin ␣1 could also function in synaptic remodeling in the adult brain.
One way that centaurin ␣1 could serve a dual role in development in branching and spine dynamics and later in synaptic remodeling, is through the regulation of lamellipodia and filopodia (see Luo, 2002;Jaffe and Hall, 2005;Dijkhuizen and Ghosh, 2005;Lippman and Dunaevsky, 2005;Newey et al., 2005). siRNA knockdown of centaurin ␣1 in cultured neurons reduced lamellipodia and filopodia, whereas overexpression of wild-type centaurin ␣1 enhanced these structures. Lamellipodia are found on the cell body, growth cones, and spines, and are thought to be the precursors of primary dendrites and possibly additional branch formation. Filopodia are formed on primary dendrites and when stabilized, become the source for further branch addition. Later in development, filopodia are thought to serve as the precursors of immature spines, which, after contact by an axon, can become mature spines where excitatory synapses are formed. Studies are underway to determine whether centaurin ␣1 is required for the initiation/formation, growth or stability of filopodia and lamellipodia.
Arf GAP activity is required for centaurin ␣1 regulation of dendritic branching, dendritic filopodia and spines. Our results correspond well with studies in which increasing Arf6-GDP levels enhances dendritic branching and spine density, whereas increasing Arf6-GTP levels inhibits these processes (Hernandez-Deviez et al., 2002;Miyazaki et al., 2005). Our results are also consistent with centaurin ␣1 functioning with Arf6 to regulate the formation or stabilization of filopodia. Regulation of dendritic filopodia, thought to be the precursors of spines, may directly impact spine density. Interestingly, overexpression of the Arf6 GEF EFA6A also increases mature spine density at the expense of filopodia, suggesting that Arf6 functions in both the production or stablization of emerging filopodia and also in the conversion of filopodia to mature spines (Choi et al., 2006). How Arf6 regulates filopodia and spines has not yet been elucidated, but may involve control of lipid-modifying enzymes, vesicular trafficking and/or regulation of the Rho GTPase family (Radhakrishna et al., 1999;Santy and Casanova, 2001;Hernandez-Deviez et al., 2002;Krauss et al., 2003;Miyazaki et al., 2005;Cotton et al., 2007).
The regulation of centaurin ␣1 localization and activity in neurons are important issues for future studies. The observation that Arf GAPs, such as centaurin ␣1 and Arf-GEFs, such as EFA6 and ARNO are potentially regulated by PtdIns(3,4,5)P 3 , and found in dendritic and synaptic regions, lends support to the hypothesis that PI 3-kinase coordinately regulates GEFs and GAPs that control the Arf pathways required for dendritic development. In addition, centaurin ␣1 has multiple identified binding partners in addition to PtdIns(3,4,5)P 3 , including protein kinases and cytoskeletal proteins, which could provide input to centaurin ␣1 regulation from a variety of signaling pathways (Dubois et al., 2001;Dubois et al., 2003;Zemlickova et al., 2003;Thacker et al., 2004;Venkateswarlu et al., 2005;Striker et al., 2006). In turn, centaurin ␣1 may also provide cross-talk to other signaling pathways such as the Ras-MAP kinase pathway, which was shown to be diminished by reducing centaurin ␣1 expression (Hayashi et al., 2005).

Antibodies
A polyclonal antibody against centaurin ␣1 was previously described (Hammonds-Odie et al., 1996). Goat polyclonal anti-centaurin ␣1 was from Abcam. Anti-Flag M5 monoclonal antibody and anti-Flag polyclonal antibody and mouse antisyntaxin were from Sigma. Goat anti-synapsin-1 was from Santa-Cruz Biotechnology. Mouse anti-synapsin-1 monoclonal antibody and mouse antisynaptophysin monoclonal antibody were from Chemicon International. Goat anti-Neurabin II (A-20) polyclonal antibody, antibodies mouse anti-MAP-2 and rabbit anti-MAP-2 were from Zymed Laboratories. Mouse anti-PSD-95, clone K28/43 was from Upstate Biotechnology, Lake Placid, NY. Oregon Red-X phalloidin, Alexa Fluor 488 anti-rabbit, Cascade Blue anti-rabbit and anti-mouse secondary antibodies were from Molecular Probes. Texas-Red anti-mouse and anti-rabbit secondary antibodies were from Vector Laboratories. Mouse anti ␤-tubulin, anti-SV2 and antiactin were from Developmental Studies Hybridoma Bank, University of Iowa, IA. Mouse anti-tau was a gift from Gail Johnson, University of Alabama at Birmingham, AL.

Subcellular fractionation
Synaptosomes were prepared from 2-to 3-week-old rats as described (Blackstone et al., 1992). Briefly, forebrains were homogenized with a glass homogenizer, centrifuged at 1000 g for 15 minutes at 4°C, and the resulting post-nuclear supernatant (S1) was centrifuged at 10,000 g for 15 minutes to yield the crude synaptosomal and microsomal pellet (P2). The washed crude synaptosomal/ microsomal P2 pellet was resuspended in 0.8 M sucrose, layered onto 1.2 M sucrose, and centrifuged at 230,000 g for 15 minutes. The gradient interface was collected, and layered onto 0.8 M sucrose and centrifuged at 230,000 g for 15 minutes. The pellet contained pure synaptosomes (Syn). Synaptosomes were prepared from different developmental ages as previously described (Patterson and Skene, 1999) by a modification of the Ficoll flotation method (Gordon-Weeks, 1987).

Dissociated neuron transfections
For studies in early development (3-7 DIV) after dissociation, neurons were transfected via Nucleofection (Amaxa Biosystems). Dissociated neurons (3ϫ10 6 cells) were centrifuged and resuspended in 100 l rat neuron Nucleofector solution (Amaxa Biosystems). 3 g plasmid DNA or 3 g of each siRNA was added to the neuronal suspension in a 2 mm electroporation cuvette and electroporated with program O-03. NBMplus medium was added and neurons were plated at a density of 8ϫ10 4 cells per well in a 12-well plate at 37°C in 5% CO 2 . For studies in 14 DIV cultures, hippocampal neurons were transfected at 1 DIV using a modified calcium phosphate protocol (Kohrmann et al., 1999). 3-5 g of DNA was added in 250 mM CaCl 2 . 2ϫ BBS (250 mM NaCl, 1.5 mM Na 2 PO 4 , 50 mM BES (Sigma), pH 7.1) was added to the Ca-DNA mixture. The transfection solution was added immediately to the coverslips and incubated at 2.5% CO 2 at 36.5°C. After 30-45 minutes, coverslips were washed twice with HBS (135 mM NaCl, 4 mM KCl, 1 mM Na 2 HPO 4 , 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, 20 mM HEPES, pH 7.35).

Immunocytochemistry
Neurons were fixed in 4% formaldehyde (Tousimis Research) in phosphate buffer (23 mM NaH 2 O 4 , 80 mM Na 2 HPO 4 ) for 20 minutes, permeabilized with 0.25 % Triton X-100 in 4% formaldehyde in PBS for 10 minutes, washed three times with 1ϫ PBS, then blocked in 10% BSA in 1ϫ PBS for 1 hour at 37°C. Coverslips were incubated overnight with primary antibodies, washed three times in 1ϫ PBS and incubated with secondary antibody for 2 hours at 25°C. Coverslips were washed three times with 1ϫ PBS, once with ddH 2 O, then dried and mounted with Vectashield (Vector Laboratories, CA).

Neuronal imaging
Images were acquired with an Olympus IX 70 inverted microscope with epifluorescence optics using the 40ϫ and 100ϫ objective with a 1.5ϫ tube lens attached giving a resolving power of 150ϫ. Images were captured using Retiga 1300 cooled CCD, firewire, high resolution, monochromatic camera (QImaging; Burnaby, British Columbia, Canada), and images were deconvolved using IPLab microtome extension and merged with IPLab Spectrum software from (Scantalytics). The percentage colocalization in dendritic segments from cultured neurons was measured using the IPLab extension, Fluorescence CV. Images were processed using Adobe Photoshop.

Evaluation of neuronal morphology
Images were obtained using the 40ϫ objective and analyzed and quantified using image Pro Plus 4.1 software (Media Cybernetics). The number of dendritic terminal tips, which is a measure of the dendritic complexity, not including the longest extension (axon) and its branches, was determined as described (Uylings and van Pelt, 2002;Hernandez-Deviez et al., 2002;Gerecke et al., 2004;Kumar et al., 2005). The length of each dendritic extension was measured by tracing along its length and the lengths were subsequently summed for the total dendritic length. A range of 25-50 neurons was measured for each condition. Results are expressed as means of treatments normalized to means of control. Values were compared by Student's t-test (two-tailed) with P<0.05 considered statistically significant.

Dissociated neuron protrusion analysis
Images were acquired with an 100ϫ objective with a 1.5ϫ tube lens attached giving a resolving power of 150ϫ. Spines and filopodia were identified as small protrusions <7 m in length at right angles from dendrite and had no further branches according to (Papa et al., 1995;Kumar et al., 2005). The total number of protrusions on the dendritic length was measured using ImageJ software (National Institute of Health, MD). Protrusion density was calculated by counting the number of spines per unit length of parent dendrite, and normalized to 10 m of dendrite. The total dendritic length for protrusion density analysis was as follows: GFP control was 910.2 m, Flag-centaurin ␣1 777.1 m and Flag-centaurin ␣1-R49K 822.1 m. Values were compared by Student's t-test (two-tailed) with P<0.05 considered statistically significant.

Organotypic slice cultures
Transverse hippocampal slices (500 m) were prepared from postnatal day 7-9 rats with a custom-designed wire slicer (California Fine Wire Company, Grover Beach, CA) and maintained as previously described (Tyler and Pozzo-Miller, 2001;Tyler and Pozzo-Miller, 2003). All tissue culture reagents were purchased from Life Technologies.

Biolistic transfections and Immunocytochemistry
After 4 DIV, culture medium was completely changed to culture medium containing antibiotic and antimycotic. Cultured slices were transfected 4 hours later using the Helios Gene Gun (Bio-Rad) as described (Alonso et al., 2004). Gold particles (25 mg of 1.6 mM) (Bio-Rad) were coated with 26.7 g pDsRed2-N1 (dsRed) (Clontech), as a control plasmid and 26.7 g enhanced yellow fluorescent protein (eYFP) (Clontech) to visualize dendrites and spines, or 26.7 g eYFP and either 53.3 g Flag-centaurin ␣1 or Flag-R49K centaurin ␣1. A time of 36 hours was used for transfection because this provides reliable slice viability as previously characterized (Tyler and Pozzo-Miller, 2001;Tyler and Pozzo-Miller, 2003). Cultured slices were fixed in 4% paraformaldehyde in 100 mM phosphate buffer (PB) for 40 minutes at RT, then rinsed in PBS. Slices were incubated overnight at 4°C in blocking buffer (1ϫ PBS, 0.25% Triton X-100, 5% normal goat serum), followed by overnight incubation at 4°C in blocking buffer with anti-Flag polyclonal antibody (1:1000). After washing in PBS, slices were incubated for 3 hours at 25°C in blocking buffer with Alexa Fluor 546 anti-rabbit secondary (1:2500). Slices were rinsed in PBS and mounted with Vectashield (Vector, CA).

Quantitative spine density analysis by laser-scanning confocal microscopy
Fluorescent images of CA1 pyramidal neurons were acquired with a laser-scanning confocal microscope (LSCM; Olympus Fluoview 300, NY), using an oil-immersion 60ϫ (1.2 NA) or 100ϫ (1.65 NA) objective lens (Olympus). eYFP was excited with the Ar laser, dsRed was excited with the Kr laser and detected with standard FITC filters. Optical z-sections were acquired at 0.5 m steps through the apical dendritic tree of the neurons. Measurements were performed on secondary and tertiary dendrites. Dendritic projections with lengths between 1 and 3 m were identified as spines and counted off-line using ImageJ software (National Institutes of Health, Bethesda, MD) as described previously (Tyler and Pozzo-Miller, 2003;Alonso et al., 2004). Spine counts were obtained from a total apical dendritic length of 891 m in control (eYFP) slices (five cells from four slices), 1855 m in centaurin-␣1transfected slices (nine cells from six slices) and 1214 m in R49 centaurin-␣1transfected slices (seven cells from five slices). Data were normalized per 10 m of dendritic length.