P-Rex1 - a multidomain protein that regulates neurite differentiation.

The Rac-GEF P-Rex1 promotes membrane ruffling and cell migration in response to Rac activation, but its role in neuritogenesis is unknown. Rac1 promotes neurite differentiation; Rac3, however, may play an opposing role. Here we report that in nerve growth factor (NGF)-differentiated rat PC12 cells, P-Rex1 localised to the distal tips of developing neurites and to the axonal shaft and growth cone of differentiating hippocampal neurons. P-Rex1 expression inhibited NGF-stimulated PC12 neurite differentiation and this was dependent on the Rac-GEF activity of P-Rex1. P-Rex1 inhibition of neurite outgrowth was rescued by low-dose cytochalasin D treatment, which prevents actin polymerisation. P-Rex1 activated Rac3 GTPase activity when coexpressed in PC12 cells. In the absence of NGF stimulation, targeted depletion of P-Rex1 in PC12 cells by RNA interference induced the spontaneous formation of beta-tubulin-enriched projections. Following NGF stimulation, enhanced neurite differentiation, with neurite hyper-elongation correlating with decreased F-actin at the growth cone, was demonstrated in P-Rex1 knockdown cells. Interestingly, P-Rex1-depleted PC12 cells exhibited reduced Rac3 and Rac1 GTPase activity. This study has identified P-Rex1 as a Rac3-GEF in neuronal cells that localises to, and regulates, actin cytoskeletal dynamics at the axonal growth cone to in turn regulate neurite differentiation.

In this study, we have investigated the role that P-Rex1 plays in regulating neuronal morphogenesis in rat primary embryonic hippocampal neurons and in NGF-stimulated PC12 cells. In hippocampal neurons, P-Rex1 localises to the distal tips of developing neurites, the axon shaft and growth cone. Ectopic expression of P-Rex1 and RNAi-mediated targeted depletion of P-Rex1 in PC12 cells and primary hippocampal neurons demonstrate that P-Rex1 inhibits neurite elongation by directing actin cytoskeletal dynamics specifically at the growth cone. Furthermore, we present evidence that P-Rex1 may function as a novel Rac3-GEF in neuronal cells. These studies have revealed that the relatively uncharacterised Rac-GEF, P-Rex1, contributes to the complex network of signalling events that regulate neuronal morphogenesis.

P-Rex1 distribution during neuronal differentiation
To determine the functional role of P-Rex1 in regulating neuronal differentiation, we first investigated the localisation of P-Rex1 in hippocampal neurons and PC12 cells. P-Rex1-specific polyclonal antibodies were raised to a unique P-Rex1 peptide sequence (Fig. 1A) and immune serum affinity-purified on a peptide-coupled thiopropyl Sepharose column. In immunoblots, P-Rex1 antibodies detected a ~180 kDa protein consistent with P-Rex1 in mouse brain lysates, rat E18 hippocampal neurons and PC12 cells (Fig. 1B). Some proteolytic P-Rex1 immunoreactive peptides migrating at 110 and 80 kDa were also detected in hippocampal neurons and PC12 cells. Two PEST sites, predicted to be susceptible to calpainmediated cleavage (Dice, 1987), are predicted in the P-Rex1 sequence. However, mouse brain immediately harvested and boiled in SDS-PAGE reducing buffer exhibited only the single, 180 kDa immunoreactive polypeptide.
Rat pheochromocytoma PC12 cells provide an experimentally accessible model system in which to study neurite outgrowth. NGF stimulation allows PC12 cells to initiate and extend neurites; however, these processes do not form polarised axons or dendrites, or specialised synapses (Greene and Tischler, 1976). Indirect immunofluorescence of undifferentiated serum-starved PC12 cells using affinity-purified P-Rex1 antibodies revealed punctate cytosolic staining (Fig. 1C, top row). No immunoreactivity was detected using Dishevelled, EGL-10, pleckstrin homology (DEP), postsynaptic density, disc-large, ZO-1 homology (PDZ), PEST sequences (P) and 4-phosphatase homology (4ptase) domains. The peptide sequence used as an immunogen to generate the P-Rex1-specific antibody is indicated. (B) Cell lysates from mouse brain, rat E18 hippocampal neurons and undifferentiated PC12 cells (30 μg) were immunoblotted with affinity-purified P-Rex1 antibodies. Immunoreactive endogenous P-Rex1 is indicated by the arrow. (C) PC12 cells were left undifferentiated, or NGF-stimulated as indicated, then immunostained with P-Rex1-specific antibodies (green) and Texas Red-conjugated phalloidin (red). Merged images are shown in the third column and enlarged images of the boxed areas of growth cones of immature (1) or mature (2) neurites are shown in the bottom row. The growth cone peripheral (P) and central (C) zones are indicated in the bottom row. P-Rex1 localisation at the distal tip of filopodia and lamellipodia is indicated by arrows and at growth cones by arrowheads. Undifferentiated PC12 cells were also stained with P-Rex1 peptide-adsorbed antibody as control (right-most panel). Scale bars: 10 μm.
peptide-adsorbed serum (Fig. 1C). Preimmune sera were also non-reactive (not shown). Following brief NGF stimulation (6 minutes), intense P-Rex1 staining was detected at the tips of lamellipodia and filopodia co-localising with phalloidin-stained Factin (Fig. 1C, arrows). Following 8-hour NGFstimulated differentiation, P-Rex1 was detected at the PC12 cell periphery, neurite shaft and at the distal end of the growth cone, the specialised motile tip of a neuronal process that contains an actin-rich peripheral (P) zone and a central (C) zone of bundled microtubules. High-magnification images revealed that P-Rex1 was localised in the central regions of the growth cone (Fig. 1C, see 'C' in bottom row), but was absent from the actinrich P zone (see 'P' in bottom row). P-Rex1 consistently localised to growth cone tips, distal to the actin-rich zone (Fig. 1C, arrowheads, bottom row), the site of actin polymerisation within the P zone of the growth cone (Lin and Forscher, 1995).
Primary hippocampal neurons provide a wellcharacterised model for the development of neuronal polarity, with cells progressing through five developmental stages characterised by the outgrowth of short, unpolarised neurites during stage 2, axonal specification in stage 3 and dendritic differentiation and synaptogenesis during stages 4 and 5 (Dotti et al., 1988). In early stage 2 hippocampal neurons, P-Rex1 was detected at the cell periphery, the developing neurite shaft and at the tips of the nascent growth cone ( Fig. 2A, arrows, growth cone staining). Rac1 exhibited a patchy distribution in the developing neurite, being most intensely localised at the distal ends of the nascent growth cone, and was co-localised with P-Rex1 ( Fig. 2A, arrow in upper row), but not in the body of the emerging growth cone or the neurite shaft. F-actin at the developing growth cone co-localised with P-Rex1, but only at the most-distal end of the neurite tips ( Fig. 2A, arrows in middle row). P-Rex1 exhibited partial co-distribution with βtubulin across the span of the neurite shaft and at the distal neurite tips ( Fig. 2A, arrows in bottom row).
In early stage 3 neurites, P-Rex1 was detected at the cell periphery, prominently in the primary neurite shaft and developing growth cone, as illustrated by the intensity of staining in the neurite shaft (Fig. 2B, open arrows in 'glowover' images). P-Rex1 co-localised with actin at the growth cone and with β-tubulin at the neurite shaft (Fig. 2B,second panel,arrowhead). In differentiated stage 4-5 neurons, P-Rex1 localised to the perinuclear region of the cell body and intensely in the growth cone and shaft of one process (Fig. 2C, 'glowover' images). Intense P-Rex1 antibody staining of Tau1 (Mapt)-positive axons was detected, with only faint staining in MAP2 (Mtap2)positive dendrites, indicating that P-Rex1 exhibits a polarised axonal distribution (Fig. 2B, see arrows for Tau1-positive, MAP2negative axons).
Journal of Cell Science 121 (17) Ectopic expression of P-Rex1 inhibits neurite differentiation via Rac-GEF activity We examined the effect of ectopic expression of P-Rex1 on NGFmediated PC12 neurite differentiation for 3 days. For each construct, embryonic rat hippocampal neurons were immunostained with P-Rex1-specific antibodies (green), Rac antibodies (red), Texas Red-conjugated phalloidin (red) or β-tubulin antibodies (red) as indicated. Higher-magnification images of the boxed area outlining the growth cones are shown in the fourth column. Arrows indicate endogenous P-Rex1, Rac or F-actin at tips of growth cones. Co-localisation in merged images appears yellow. (B,C) Three (B) and 4-7 (C) d.i.v. embryonic rat hippocampal neurons were co-stained with P-Rex1-specific antibodies (green), Texas Red-conjugated phalloidin (red), β-tubulin (red), Tau1 (red) or MAP2 (red) antibodies. The relative intensity of P-Rex1 antibody staining is shown in the left-hand column by 'glowover' images, in which blue indicates high-intensity staining (see scale beneath B). Merged images in the right-hand column demonstrate co-localisation by yellow staining. Open arrows indicate primary neurites. Arrowhead indicates P-Rex1 co-localisation with β-tubulin.
White arrows in C indicate Tau1-positive or MAP2-negative processes. Scale bars: 10 μm in A; 20 μm in B,C. P-Rex1 regulates neurite differentiation the length of the longest neurite was measured and growth cone Factin was examined by Texas Red-phalloidin staining (see Fig. 3A). Cells were classified as differentiated upon the acquisition of one or more neurites that were greater in length than the diameter of the cell body. By this criterion, 50% of cells expressing vector-only controls were differentiated. Ectopic expression of P-Rex1 resulted in a 3-fold decrease in differentiation (Fig. 3C). P-Rex1-expressing cells initiated actin-rich projections, but these did not differentiate into elongated neurites as shown by scoring the number of neurites/actin-rich projections of any length. However, P-Rex1 cells initiated ~2-fold more actin-rich projections than did vector controls (Fig. 3B, upper row, arrows; supplementary material Fig. S1). Cells were treated with low-dose cytochalasin D, which inhibits actin polymerisation. Under these conditions, >40% of HA-P-Rex1expressing PC12 cells exhibited differentiated neurites (Fig. 3C), suggesting that P-Rex1 inhibition of neurite differentiation is mediated by actin. A P-Rex1 mutant (P-Rex1GEFdead) that is GEFdead as a consequence of two point mutations, E56A and N238A, Fig. 3. Ectopic expression of P-Rex1 inhibits NGF-mediated neurite differentiation and is dependent on both the Rac-GEF activity and the 4-phosphatase domain. (A) Schematic of P-Rex1 deletion mutant constructs. PC12 cells were transiently transfected with P-Rex1 mutant constructs and cell lysates were immunoblotted with HA or myc antibodies, as shown beneath. 1, HA-vector; 2, HA-P-Rex1; 3, HA-P-Rex1ΔN; 4, HA-P-Rex1Δ4P; 5, myc-P-Rex1; 6, myc-P-Rex1GEFdead. To the right, neurite outgrowth, neurite F-actin and initiation are summarised for each P-Rex1 mutant protein. For neurite outgrowth: +++, as vector control; +, reduced neurite outgrowth; -, no outgrowth. For F-actin: +, as vector control; +++, increased. For initiation: +, initiation as vector control. (B,C) PC12 cells were transiently transfected with wild-type or mutant P-Rex1 (5 μg, unless otherwise indicated) and NGF-stimulated for 3 days in the presence or absence of 1 μM cytochalasin D. (B) Cells were co-stained with HA or myc antibodies (green) and Texas Red-conjugated phalloidin (red or grey). Merged images are shown (lower row) with neurites or actin-rich projections indicated by arrows. Scale bar: 10 μm. (C) Cells containing actin-rich projections were imaged and the length of the neurite/projection and the cell body diameter determined. Bars indicate the mean ± s.e.m. of cells bearing neurites longer than one cell body diameter. 100 cells were scored for each construct for three independent transfections. *P<0.05; **P<0.01; ***P<0.001.
in the Rac-GEF domain , did not inhibit neurite differentiation indicating that Rac activation is crucial for P-Rex1 function. Mutant P-Rex1 that contained only the central and 4phosphatase homology domains (P-Rex1ΔN) had no effect, when at low dose (1 μg), on neurite differentiation, but at high levels (5 μg) resulted in a similar phenotype to wild-type P-Rex1, with a 3fold inhibition of differentiation resulting in short projections with greatly increased F-actin ( Fig. 3A-C). This might be due to as yet unidentified molecular mechanisms because this construct does not contain the Rac-GEF activating domain, but does contain the central and 4-phosphatase domains. HA-P-Rex1Δ4P, which lacks the 4phosphatase homology domain but contains the Rac-GEF domain, did not inhibit neurite differentiation, although we noted that Factin growth cone content appeared reduced ( Fig. 3A-C). This result suggests that the 4-phosphatase domain might contribute to P-Rex1 inhibition of neurite differentiation and actin polymerisation by unknown mechanisms. In control studies, expression of mutant and wild-type P-Rex1 was shown to be intact by immunoblot analysis (Fig. 3A).
Recent studies have revealed that the Rac proteins, Rac1 and Rac3, although highly related, may exert distinct functions in the developing nervous system (Hajdo-Milasinovic et al., 2007). P-Rex1 activates Rac1 in purified-component assays, but in neutrophils it exhibits preference for Rac2, a haematopoiesis-specific Rac (Dong et al., 2005;Welch et al., 2005). The ability of P-Rex1 to function as a Rac3-GEF has not been reported. To address this, PC12 cells were co-transfected with vector, P-Rex1 or P-Rex1GEFdead, and with Rac1 or Rac3, and an ELISA-based Rac activation assay (G-LISA) performed. In response to NGF, only a small increase in Rac1 activation in the P-Rex1-expressing cells was detected in each of four experiments and this trend only approached significance (P=0.057). By contrast, Rac3 activation in P-Rex1-expressing cells was consistently higher, and statistically significant (P=0.011), in response to NGF as compared with vector controls. The P-Rex1GEFdead construct did not activate either Rac1 or Rac3 (Fig.  4A). P-Rex1 can therefore activate Rac3 in neuronal cells and this is dependent on its Rac-GEF activity. Rac3 localises to the plasma membrane of stimulated cells, but in contrast to Rac1, it also exhibits a perinuclear localisation (Hajdo-Milasinovic et al., 2007). There was minimal co-localisation of P-Rex1 and Rac1 at the plasma membrane of PC12 cells following a 3-minute NGF stimulation ( Fig. 4B, arrowheads, middle row). Interestingly, however, P-Rex1 co-localised with Rac3 in a perinuclear distribution (Fig. 4B, arrow, bottom row).
We also examined the role P-Rex1 plays in the differentiation of primary neurons, by transfecting rat embryonic hippocampal cells after 1 day of differentiation in vitro (1 d.i.v.) with plasmids encoding HA-P-Rex1. However, no neurons expressing P-Rex1 could be identified, for reasons unknown (results not shown). To overcome this, differentiated hippocampal neurons were transfected at 7 d.i.v., then grown for an additional 2 days. HA-P-Rex1 was detected at the cell body, the distal half of the axon shaft and growth cone (Fig.  5A), co-localising with Tau1, an axonal marker (Fig. 5B, arrowhead). Significantly, P-Rex1 expression resulted in enlargement of the Tau1-positive axonal growth cones (2.1-fold) associated with prominent F-actin accumulation; however, MAP2positive dendritic growth cones were not affected (Fig. 5A, lower row, Fig. 5C). Expression of the P-Rex1ΔN mutant, which lacks the Rac-GEF, PH, DEP and PDZ domains, resulted in loss of the enlargement of axonal growth cones seen with wild-type P-Rex1 expression (Fig. 5C).

Journal of Cell Science 121 (17) P-Rex1 RNAi-mediated depletion promotes spontaneous neurite initiation
The effects on neuronal differentiation of RNAi-mediated depletion of P-Rex1 were investigated. Oligonucleotides designed to unique P-Rex1 sequences (P-Rex1 RNAi), or the same sequence scrambled (scram RNAi), were cloned into the psiRNA-hH1 vector, stably transfected into PC12 cells (Fig. 6A) and individual clones isolated. Immunoblot analysis demonstrated a reduction in P-Rex1 protein (58 and 70% in P-Rex1 RNAi clones 1 and 5, respectively), relative to scrambled RNAi clones (3 and 4) (Fig. 6B). RT-PCR analysis confirmed a ~50% reduction in P-Rex1 mRNA, consistent with the immunoblot analysis (Fig. 6C). In the absence of NGF stimulation, conditions not normally associated with neurite formation in controls, P-Rex1 depletion resulted in the loss of PC12 rounded cell morphology (Fig. 6D). Additionally, P-Rex1 RNAi clones Fig. 4. P-Rex1 co-localises with, and promotes the activation of, HA-Rac3 in NGF-stimulated PC12 cells. (A) PC12 cells were co-transfected with myc-Rac1 or HA-Rac3 and vector control, P-Rex1 or P-Rex1GEFdead, and left serum starved or stimulated with NGF for 3 minutes. Duplicate samples of cell lysates were subjected to colorimetric Rac1,2,3 G-LISA Activation Assays and the average Rac activation was calculated relative to serum-starved HAvector-only controls. Equivalent expression of Rac constructs was confirmed by immunoblot analysis using a pan-Rac antibody (not shown). Bars indicate mean ± s.e.m. of relative Rac activation for four independent experiments (*P<0.05). (B) PC12 cells were transfected with HA-P-Rex1 (upper row) or co-transfected with HA-P-Rex1 and myc-Rac1 (middle row) or myc-P-Rex1 and HA-Rac3 (bottom row). Cells were stimulated with NGF for 3 minutes. P-Rex1 (green) and Myc-Rac1 or HA-Rac3 (red) were detected using antibodies to each tag. Co-localisation is indicated by yellow staining in the merged images (right-hand column). Arrowheads indicate P-Rex1/Rac1 membrane colocalisation, with P-Rex1/Rac3 perinuclear co-staining indicated by the arrow. Scale bars: 10 μm. P-Rex1 regulates neurite differentiation exhibited a 3-fold increase in β-tubulin-rich projections, relative to scrambled or vector controls (2.9±0.3-fold and 3.3±0.3-fold increase in P-Rex1 RNAi clones 1 and 5, respectively, relative to scrambled controls, Fig. 6D). PC12 cell morphology was also assessed immediately following NGF treatment (10 minutes) (Fig. 6D, righthand column). Of the cells expressing scrambled RNAi, 60-80% formed lamellipodia that were similar to those of untransfected PC12 cells. However, P-Rex1 RNAi PC12 cells appeared elongated and exhibited projections that contained a β-tubulin core (Fig. 6D, green) surrounded by F-actin (red), rather than lamellipodia around the entire cell periphery. Immediately following NGF stimulation, the minimum cell diameter was 14.5±0.18 μm in scrambled controls, but had decreased to 11.5±0.14 μm and 11.3±0.25 μm in P-Rex1 RNAi clones 1 and 5, respectively. This correlated with a >30% decrease in the area of the cell body of P-Rex1 RNAi (~170 μm 2 ) as compared with scrambled control (~260 μm 2 ) PC12 cells (200 cells scored using Image J analysis as described in Materials and Methods).
We next examined Rac1 versus Rac3 activation in PC12 cells expressing control or P-Rex1 RNAi. Cells were transfected with Rac1 or Rac3 and Rac activation assays performed (Fig. 6E). Recent studies have revealed that in P-Rex1 -/neutrophils, Rac1 and Rac2 peak activation is reduced by 25 and 50%, respectively . In another study, Rac1 activation was not altered in P-Rex1 -/neutrophils, whereas Rac2 activation was almost completely abrogated, revealing that although P-Rex1 can activate both Rac isoforms in vitro, in haematopoietic cells it functions as a Rac2-specific GEF (Dong et al., 2005). In response to NGF stimulation, peak Rac1 activation was decreased by ~20% in P-Rex1 RNAi cells as compared with the control, but this was not statistically significant. By contrast, we noted a significant, ~40% decrease in Rac3 activation in NGF-stimulated P-Rex1 knockdown cells. These studies suggest that the P-Rex1 RNAi-mediated phenotype might be a consequence of decreased Rac3 activity.
P-Rex1 RNAi-mediated depletion enhances neurite elongation P-Rex1 RNAi clones were stimulated with NGF for 3 days. The percentage of cells bearing differentiated neurites increased >1.8-fold upon targeted depletion of P-Rex1, relative to scrambled or vector controls ( Fig. 7A and see supplementary material Fig. S2 for a comparison of vector versus scrambled RNAi). Although the majority (>70%) of differentiated scrambled RNAi cells exhibited short neurites that were less than two cell diameters in length, less than 40% of differentiated P-Rex1 RNAi cells carried short neurites. By contrast, ~30% of P-Rex1 RNAi versus <5% of scrambled RNAi PC12 cells displayed neurites that were longer than three diameters of the cell body, a 6-fold increase (Fig. 7B). Analysis of the length of the longest neurite per cell after 3 days of NGF stimulation revealed that P-Rex1 RNAi neurites were ~1.4-fold longer than those of scrambled RNAi controls (Fig. 7C).
Rac and its effectors regulate axonal guidance and direction (Dickson, 2001;Kubiseski et al., 2003;Ng et al., 2002). Although this cannot be evaluated in PC12 cells, we noted that P-Rex1 RNAi neurites exhibited a 'wandering neurite' phenotype, with a 5-fold increase in the number of neurites showing multiple changes in direction, or one significant change in direction with an angle of less than 130°, compared with scrambled RNAi neurites (Fig. 7A, representative images). Actin instability has been implicated in mediating defects in neuronal elongation and pathfinding (Marsh and Letourneau, 1984;Zhou et al., 2002). Rac and its effectors regulate axon branching and suppress ectopic axon/neurite formation. P-Rex1 RNAi PC12 cells exhibited a >2-fold increase in the number of branch points per neurite (data not shown). Because siRNAs can exert off-target effects, we investigated whether co-transfection of increasing amounts of P-Rex1 cDNA could rescue the phenotype mediated by P-Rex1 RNAi knockdown. To this end, P-Rex1 RNAi clones were transfected with increasing amounts of wild-type HA-P-Rex1 (1, 2 or 5 μg), or vector (5 μg) and neurite length assessed. A dose-dependent reduction in the Journal of Cell Science 121 (17) percentage of cells bearing long neurites correlated with transfection of wild-type P-Rex1, but not vector cDNA (Fig. 7D). P-Rex1 also rescued the 'wandering neurite' phenotype (not shown). The P-Rex1GEFdead construct only partially rescued the hyper-elongated phenotype. Interestingly, unlike wild-type P-Rex1, expression of HA-P-Rex1Δ4P did not rescue the phenotype of P-Rex-1 RNAi- Fig. 6. Targeted depletion of P-Rex1 promotes spontaneous outgrowth of β-tubulin-rich projections (A) The location of the unique nucleotide sequence used to generate P-Rex1 RNAi clones is indicated in the P-Rex1 structure. (B) Cells were stably transfected with a plasmid encoding P-Rex1 RNAi sequence, or a scrambled RNAi control. Cell lysates (40 μg) were immunoblotted with P-Rex1 or actin antibodies. Relevant samples loaded on the same, representative immunoblot are shown. The relative P-Rex1 expression in P-Rex1 clones (1 and 5) was determined by densitometry of P-Rex1-immunoreactive polypeptides from five independent immunoblots, standardised to the actin loading control and expressed as a percentage of that detected in scrambled RNAi clones (3 and 4 combined). The bars indicate the mean ± s.e.m. of five independent experiments. (C) RT-PCR analysis of P-Rex1 RNAi clones. RT-PCR was performed on mRNA extracted from scrambled and P-Rex1 RNAi clones using Gapdh as a control. P-Rex1 expression was calculated and expressed relative to that from Scram (3), which was designated as one. Bars indicate the mean ± s.e.m. of three independent experiments. (D) Cells stably transfected with P-Rex1 RNAi or scrambled RNAi control were left unstimulated (a,b,c) or NGF-stimulated for 10 minutes (d). Cells were stained with Texas Red-conjugated phalloidin (red or grey) or βtubulin antibody (green) to image cell morphology. Representative images of unstimulated P-Rex1 RNAi clones (1 and 5) and scrambled RNAi control (4) are shown (a,b) with merged magnified images of boxed regions (c). β-tubulin-containing projections are indicated (b, arrows). Merged images of cells treated with NGF for 10 minutes are shown (d) with areas of extensive peripheral F-actin indicated by arrowheads. Scale bars: 5 μm. (E) RNAi-mediated depletion of P-Rex1 in PC12 cells reduces Rac3 activation in response to NGF stimulation. Scrambled RNAi or P-Rex1 PC12 RNAi clones were transiently transfected with myc-Rac1 or HA-Rac3, serum starved and NGF-stimulated (3 minutes). Duplicate samples of cell lysates were subjected to colorimetric Rac1,2,3 G-LISA Activation Assay and the average Rac activation was calculated relative to serum-starved scrambled control cells. Equivalent expression of Rac constructs was confirmed by immunoblot analysis using a pan-Rac antibody (not shown). Bars indicate mean ± s.e.m. of relative Rac activation for four independent experiments. **P<0.005; ***P<0.001. P-Rex1 regulates neurite differentiation transfected cells (Fig. 7D); instead, it enhanced the phenotype and an increased number of cells with hyper-elongated neurites was observed (Fig. 7D). Previous studies have indicated that the 4phosphatase domain of P-Rex1 is required to maintain the protein in the cytosol with basal levels of Rac activity, and that the Nterminal GEF and PH domains mediate membrane localisation (Barber et al., 2007;Hill et al., 2005). Loss of the 4-phosphatase domain might allow unimpaired translocation of P-Rex1 to the membrane where the predominant Rac isoform, Rac1, localises. As Rac1 GEF activity promotes neurite outgrowth and, as we have shown, P-Rex1 exhibits GEF activity for both Rac1 and Rac3, this might lead to unopposed neurite elongation. However, we cannot exclude the possibility that this construct might homodimerise with the residual endogenous P-Rex1 and act as a dominant-negative construct. HA-P-Rex1ΔN did not rescue the phenotype in P-Rex1 RNAi cells. Therefore, both the Rac-GEF and 4-phosphatasehomology domains contribute to P-Rex1 regulation of neuronal differentiation.
To assess the effect of P-Rex1 knockdown in hippocampal neurons, cells at 1 d.i.v. were transfected with the P-Rex1 RNAi constructs and with eGFP-expressing vector (ratio of 0.5 μg eGFP:1.5 μg RNAi plasmid to identify RNAi-transfected neurons). Cells were differentiated for 2 days before fixation and staining for F-actin (Fig. 8A, representative images). As with P-Rex1depleted PC12 cells, P-Rex1-depleted neurons exhibited neurites that were 1.3-fold longer than controls, with an average of two neurites per cell regardless of the construct expressed (data not shown). Interestingly, the growth cones of P-Rex1 RNAi neurons were ~30% smaller than those of controls and exhibited less F-actin (Fig. 8A, enlarged images of growth cones), the opposite phenotype to that observed with ectopic P-Rex1 expression. The actin-rich P zone of the growth cone is composed of radially aligned, filopodial bundles of F-actin with lamellipodial-meshed F-actin situated between the bundles. Dynamic pioneer microtubules advance from the C zone into the P domain and, via interactions with F-actin bundles, promote neurite outgrowth and growth cone turning. The advance of dynamic microtubules into the growth cone P domain is inhibited by retrograde flow of Factin (Lin and Forscher, 1995). F-actin is a major intracellular determinant of neurite elongation and axon elongation rates (Letourneau et al., 1987;Marsh and Letourneau, 1984). F-actin is polymerised from monomeric G-actin subunits at the neurite growth cone, the tips of the filopodia and the leading edge of membrane ruffles. The dynamic depolymerisation of F-actin and recycling of actin monomers for further rounds of polymerisation facilitates neurite/axon elongation (Dent and Gertler, 2003). To assess actin polymerisation at the growth cone, cells were fixed and stained with phalloidin and with fluorescently labelled DNaseI, selective probes for F-actin and G-actin, respectively (Fig.  8B). The ratio of fluorescence of growth cone F-actin to G-actin was determined (Torreano et al., 2005). The growth cones of P-Rex1 RNAi neurites appeared smaller than those of scrambled RNAi neurites, with decreased F-actin relative to G-actin (Fig.  8C). Only ~30% of P-Rex1 RNAi growth cones exhibited lamellipodial veils between filopodia compared with ~60% of scrambled RNAi growth cones (Fig. 8B, arrows; Fig. 8D). These results suggest that P-Rex1 knockdown leads to decreased F-actin growth cone content which thereby promotes neurite elongation. Fig. 7. P-Rex1 regulates NGF-stimulated neurite elongation. Cells stably transfected with P-Rex1 RNAi, or a scrambled control RNAi, were NGF-stimulated for 3 days and stained with Texas Red-conjugated phalloidin and β-tubulin antibodies to image neurite morphology. (A) Representative images of NGF-differentiated P-Rex1-depleted or control PC12 cells. Scale bars: 10 μm. (B) The numbers of differentiated PC12 cells with short neurites (length less than two cell body diameters) and long neurites (length greater than three cell body diameters) were expressed as a percentage of all differentiated neurites. Bars indicate mean ± s.e.m. of at least 100 cells scored for each of three independent differentiation experiments. (C) Neurite length was expressed as the fold increase (mean ± s.e.m.) of the length of the longest neurite for P-Rex1 clones 1 and 5 relative to that of the scrambled control (4). At least 60 neurites were scored for each of three independent differentiation experiments. (D) P-Rex1 RNAi clone (5) was transiently transfected with HA empty vector or with plasmids containing HA-P-Rex1 (1, 2 or 5 μg of DNA), myc-P-Rex1GEFdead (5 μg), HA-P-Rex1ΔN (5 μg) or HA-P-Rex1Δ4P (5 μg) and NGF-differentiated for 3 days. Cells were stained with Texas Red-conjugated phalloidin and HA or myc antibodies. The number of cells bearing neurites longer than two cell body diameters was calculated and standardised relative to that of the P-Rex1 RNAi clone (5). Knockdown cells transfected with HA empty vector were scored as 100%. Bars indicate the mean ± s.e.m. for at least 50 cells scored per indicated construct for each of three independent experiments. *P<0.05; **P<0.01; ***P<0.001.

Discussion
The co-ordinated reorganisation of the actin cytoskeleton is required at several stages of neuronal differentiation including neurite initiation, neurite elongation, growth cone guidance and axonal specification. We have shown here that a recently identified Rac-GEF, P-Rex1, localises to sites of actin reorganisation during various stages of neuronal differentiation, including to the hippocampal cell membrane during early stage 2, to the tips of lamellipodia and filopodia of the developing neurites in stage 2 and 3 neurons, to the growth cones of differentiating PC12 cells and to the growth Journal of Cell Science 121 (17) cone and shaft of axons during later stages of hippocampal morphogenesis. P-Rex1 expression inhibited neurite elongation and actin growth cone dynamics, whereas its RNAi-mediated depletion resulted in spontaneous process initiation with hyper-elongation of differentiated neurites. P-Rex1 knockdown reduced the NGFstimulated activation of Rac3, identifying P-Rex1 as a Rac3-GEF. Collectively, this study has identified P-Rex1 as a novel regulator of actin dynamics at the growth cone that inhibits neurite differentiation.

P-Rex1 activates Rac3
In this study, we made the unanticipated finding that P-Rex1 activates Rac3 in neuronal cells. This conclusion is further supported by the fact that the phenotype of the P-Rex1 knockdown neurites resembles that described for Rac3, but not Rac1, knockdown in neuroblastoma cells. Rac1, which is ubiquitously expressed, and Rac3, which is neuron-specific, may function differently in neuronal development despite their closely related sequences (Hajdo-Milasinovic et al., 2007). These functional differences might be a consequence of the divergent sequences in the C-terminal region, which might target distinct effectors. Although Rac3 exhibits a developmentally regulated distribution in mammalian brain, the physiological function of Rac3 in mammalian neurons remains poorly understood. A recent study demonstrated that hippocampal neurons from Rac3 -/mice develop normally in culture, although it was acknowledged that long-term depletion of Rac3 in these neurons may lead to a compensatory upregulation of Rac1 activity (Gualdoni et al., 2007). Initial studies demonstrated that the avian homologue of Rac3 (Rac1B) promoted neurite outgrowth in disassociated avian neurons (Albertinazzi et al., 1998). However, Rho GTPases exhibit species-, cell type-and matrix-dependent effects on neurite outgrowth (Govek et al., 2005). Recent studies have shown that RNAi-mediated depletion of Rac1 promotes cell rounding in unstimulated neuronal N1E-115 cells, whereas Rac3 depletion stimulates the outgrowth of neurite-like extensions, suggesting that Rac3 exhibits an opposing function to Rac1 in promoting neurite outgrowth (Hajdo-Milasinovic et al., 2007). Here, we have presented several lines of evidence that P-Rex1 activates Rac3 in neuronal cells. First, P-Rex1 activated Rac3 when P-Rex1 was ectopically expressed in PC12 cells and co-localised with Rac3 at the peri-Golgi region and plasma membrane. Secondly, RNAimediated depletion of P-Rex1 promoted the spontaneous formation of projections in PC12 cells, a phenotype similar to that obtained upon knockdown of Rac3 in N1E-115 neuroblastoma cells. Consistent with this, P-Rex1-depleted cells exhibited reduced activation of Rac3. Although P-Rex1 activates Rac1 in purifiedcomponent assays, neutrophils from P-Rex1 -/mice showed little change in Rac1-GEF activity, but significant decreases in Rac2-GEF activity . Given that the P-Rex1 RNAi knockdown phenotype mimics the Rac3 knockdown, we propose that P-Rex1 contributes to Rac3 activation in some neuronal cells.
P-Rex1 knockout mice appear healthy with no gross neuronal phenotype (Dong et al., 2005;Welch et al., 2005). However, detailed analysis of the brain and neural tissue has not been described and neither cognitive function nor behavioural analysis of these mice has been reported. Rac1 -/mice are embryonic lethal with no mice surviving to E10 (Sugihara et al., 1998). By contrast, Rac3 -/mice are healthy with no gross brain abnormalities; however, specific behavioural differences to wild type, including superior motor coordination and learning, were noted (Corbetta et al., 2008; Corbetta The mean fluorescence intensity of F-actin relative to G-actin at the growth cone was determined. Bars indicate the mean ± s.e.m. of the F-actin:G-actin ratio for at least ten growth cones per RNAi clone for each of three independent differentiation experiments. (D) Growth clones were scored for the presence of lamellipodial veils by F-actin morphology. Bars indicate the mean ± s.e.m. for the percentage of cells containing growth cone lamellipodial veils for at least ten growth cones per RNAi clone for each of three independent differentiation experiments. *P<0.05. P-Rex1 regulates neurite differentiation et al., 2005). Rac3 gene expression is highest postnatally in areas of the brain that contain projection neurons involved in long and complex neuronal networks, such as the hippocampus and cerebral cortex (Corbetta et al., 2005). Therefore, P-Rex1 might activate Rac3 in a specific subset of neuronal cells.

P-Rex1 regulates actin dynamics at the growth cone
Co-ordinated neurite outgrowth is essential for both normal nervous system development and for nerve regrowth following injury. P-Rex1 expression inhibited neurite differentiation and elongation, a phenotype corrected by inhibition of actin polymerisation. By contrast, P-Rex1-depleted neurites exhibited decreased F-actin, smaller growth cones and a reduction in lamellipodial veils, correlating with neurite hyper-elongation and abnormal, 'wavy' neurites. These results support the contention that targeted depletion of P-Rex1 leads to decreased Rac activation and, thereby, decreased growth cone actin polymerisation and membrane ruffling, which in turn allows the unopposed activity of microtubules to promote neurite elongation. Rac activity regulates the cycling of actin polymerisation and depolymerisation at membrane ruffles, and the balance between too much versus too little Rac activity governs neurite/axon guidance and elongation. Extensive membrane ruffling strongly attenuates neurite expansion, process formation and directional growth as a consequence of the inhibition of microtubulemediated neurite extension (Tanaka and Sabry, 1995). Growth cone F-actin, which is not incorporated into actin bundles, may retard the dynamic advance of microtubules into the peripheral region (Zhou et al., 2002). The density of actin in the nascent growth cone of PC12 cells overexpressing P-Rex1, and the failure of these neurites to differentiate, did not allow clear visualisation of F-actin in the growth cone in these cells. However, P-Rex1-mediated inhibition of neurite differentiation was rescued by low-dose cytochalasin D treatment, which, by decreasing actin polymerisation at membrane ruffles, may enable neurite extension.

P-Rex1 is a multi-domain protein
The relative roles that Rac and its various Rac-GEFs play in neurite outgrowth, guidance and differentiation are dependent on the cell type, stimulus and matrix upon which the neurons differentiate. Many neuronal Rac-GEFs have a multi-domain structure that allows the activation of Rac exchange activity at specific subcellular regions to be regulated by extracellular signalling and to link with appropriate downstream effectors (Rossman et al., 2005). Domains other than the Rac-GEF domain can modulate exchange activity, often acting as scaffolds so that both the Rho GTPase and its effectors are in contact; for example, PAK binding to the GEF α-Pix (Feng et al., 2004). PH domains can regulate Rac-GEF activity by binding directly to the GEF domain, to phosphoinositides or directly to Rac1, as has recently been described for Trio (Chhatriwala et al., 2007). Additionally, many of these proteins have GEF activity-independent functions; for example, kalirin is able to regulate actin to induce lamellipodia via both Rac1-dependent and Rac1-independent mechanisms (Schiller et al., 2005). Here, we examined the P-Rex1 domains that regulate its capacity to inhibit neurite differentiation. Rac-GEF activity was crucial for P-Rex1 function, as a P-Rex1GEFdead construct that could activate neither Rac1 nor Rac3 did not inhibit neurite differentiation when overexpressed. Interestingly, however, this construct could only partially rescue the P-Rex1 RNAi knockdown phenotype, suggesting that other domains might contribute to P-Rex1 function. In this regard, it is of interest that expression of a P-Rex1 mutant lacking the 4-phosphatase domain (P-Rex1Δ4P) in RNAi P-Rex1depleted cells was unable to rescue the hyper-elongation phenotype, but instead exacerbated it. Cells overexpressing a 4-phosphatase domain deletion mutant (P-Rex1Δ4P) developed neurites of a similar length and F-actin content to vector-transfected cells. These results suggest that the 4-phosphatase domain of P-Rex1 might play a role in regulating neurite elongation, by promoting actin polymerisation independent of the Rac-GEF domain and/or by inhibiting de-polymerisation, or by another as yet unidentified mechanism.
Recent studies by Welch's group using in vitro binding and stimulation of purified proteins has revealed that the isolated 4phosphatase domain does not modulate the ability of PtdIns(3,4,5)P 3 or Gβγ subunits to stimulate P-Rex1 Rac-GEF activity . However, in the absence of the 4-phosphatase homology domain, the level of basal and/or stimulated P-Rex1 Rac-GEF activity is significantly attenuated, suggesting a functional interaction between the DH/PH domains and the 4-phosphatase homology domain that prevents P-Rex1 GEF enzyme inactivation at high concentrations of PtdIns(3,4,5)P 3. Indirect evidence also suggests that the 4-phosphatase domain might contain a second PtdIns(3,4,5)P 3 -binding domain . Collectively, our analyses indicate that P-Rex1 regulation of neurite elongation requires both an intact Rac-GEF domain and a 4-phosphatase domain, although the function of the 4-phosphatase domain remains elusive.
In summary, this study has identified P-Rex1 as a Rac3-GEF in neuronal cells that regulates neurite differentiation. As Rac3 knockout mice exhibit enhanced motor skills and learning ability, it would be of great interest in future studies to determine the cognitive function of P-Rex1 -/mice.

Materials and Methods
Reagents DNA-modifying and restriction enzymes were from Fermentas, New England Biolabs or Promega. Texas Red-conjugated phalloidin, Alexa Fluor 488-conjugated phalloidin and Alexa Fluor 594-conjugated deoxyribonuclease I (DNaseI) were from Molecular Probes (Eugene, OR). Antibodies specific for Rac were from Upstate (Lake Placid, NY), β-tubulin from Zymed (San Francisco, CA), MAP2 from Sigma (St Louis, MO), Tau1 from Chemicon (Temecula, CA), actin from Santa Cruz (Santa Cruz, CA) and hemagglutinin (HA) from Covance (Berkeley, CA). All hippocampal neuron culture reagents were purchased from Gibco BRL (Gaithersburg, MD). Oligonucleotides were obtained either from Micromon (Monash University, Australia) or GeneWorks (Adelaide, Australia). PC12 cells were from the American Type Culture Collection. Synthetic peptides were obtained from Chiron Mimotopes (Melbourne, Australia). Unless otherwise stated, all other reagents were from Sigma. pCGN vector was a gift from Tony Tiganis (Monash University, Australia).

Cloning of human P-Rex1
A human P-Rex1 (PREX1) partial cDNA (KIAA1415, residues 359-4890) was a gift from Dr T. Nagase (Kazusa DNA Research Institute, Kisarazu, Japan). A retained intron of 1100 bp contained within this clone was removed using a PCR-based mutagenesis protocol with overlapping primers, 5Ј-CTATGAACCACAGCTTA-CAAGAGTTTAAACAGAAAGAAG and 5Ј-CTTCTTTCTGTTTAAACTCTTG-TAAGCTGTGGTTCATAG. An additional 254 bp of 5Ј sequence was amplified from a human EST (BFI10873) using primers 5Ј-GCTTAGAATTCCCGTGTGCGGCC-CGGGAGTCCG and 5Ј-CTCCTTAAGGAGGAGCGGGTAC. This fragment was ligated into a unique AflII site within the P-Rex1 cDNA. To generate HA-tag fusion constructs, full-length P-Rex1 cDNA, or various P-Rex1 mutants, were amplified by PCR and subcloned into the XbaI site of pCGN in-frame with the N-terminal HA tag.
Production of a P-Rex1-specific anti-peptide antibody A P-Rex1-specific anti-peptide antibody was generated to a peptide corresponding to a unique sequence ( 1346 LGYRYNNNGEYEESS 1360 ) within human P-Rex1. The peptide, conjugated to diphtheria toxoid, was injected into New Zealand White rabbits and the anti-peptide antibodies were affinity purified from immune sera by chromatography using the peptide coupled to thiopropyl-Sepharose according to manufacturer's instructions (Chiron Mimotopes, Melbourne, Australia). Peptide-adsorbed P-Rex1 antibodies were prepared by incubating the P-Rex1-specific antibody with the peptide-coupled column for 2 hours at room temperature.

Preparation of hippocampal neurons
Hippocampal neuron cultures were prepared from pregnant Sprague-Dawley rats (Monash University Animal Ethics Project No. BAM/B/2004/47) at day 18 of gestation as described (Banker and Goslin, 1991), with slight modifications. Briefly, following dissection, hippocampi were digested in 0.25% trypsin in Hanks Balanced Salt Solution at 37°C for 15 minutes. Cells were manually dissociated by trituration using a fire-polished Pasteur pipette and plated onto 0.01% PLL-coated coverslips at a density of 5ϫ10 5 -1ϫ10 6 /cm 2 . Following adherence of cells to coverslips, plating medium (MEM, 10% FCS, 0.6% glucose, 100 units/ml penicillin and 0.1% streptomycin) was replaced with Neurobasal medium supplemented with B27 and 0.5 mM glutamine, and cells were allowed to differentiate for 1-7 days d.i.v. as indicated. For transient transfection of hippocampal neurons, cells were transfected at 1 or 6 d.i.v. with 1-5 μg of DNA using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Neurons were fixed in PBS containing 4% paraformaldehyde and 0.12 M sucrose for 20 minutes.

Immunofluorescence
Cells were fixed in 3% paraformaldehyde in PBS for 20 minutes, permeabilised in 0.1% Triton X-100 in PBS for 2 minutes, washed three times in PBS and then blocked in 1% BSA in PBS for 15 minutes. Following incubation with primary and secondary antibodies as indicated for 1 hour, cells were analysed with a Leica TCS-NT confocal microscope with an Ar-Kr triple-line laser at Monash MicroImaging, Monash University, Australia.