The action of many extracellular guidance cues on axon pathfinding requires Ca2+ influx at the growth cone (Hong et al., 2000; Nishiyama et al., 2003; Henley and Poo, 2004), but how activation of guidance cue receptors leads to opening of plasmalemmal ion channels remains largely unknown. Analogous to the chemotaxis of amoeboid cells (Parent et al., 1998; Servant et al., 2000), we found that a gradient of chemoattractant triggered rapid asymmetric PI(3,4,5)P3 accumulation at the growth cone's leading edge, as detected by the translocation of a GFP-tagged binding domain of Akt in Xenopus laevis spinal neurons. Growth cone chemoattraction required PI(3,4,5)P3 production and Akt activation, and genetic perturbation of polarized Akt activity disrupted axon pathfinding in vitro and in vivo. Furthermore, patch-clamp recording from growth cones revealed that exogenous PI(3,4,5)P3 rapidly activated TRP (transient receptor potential) channels, and asymmetrically applied PI(3,4,5)P3 was sufficient to induce chemoattractive growth cone turning in a manner that required downstream Ca2+ signaling. Thus, asymmetric PI(3,4,5)P3 elevation and Akt activation are early events in growth cone chemotaxis that link receptor activation to TRP channel opening and Ca2+ signaling. Altogether, our findings reveal that PI(3,4,5)P3 elevation polarizes to the growth cone's leading edge and can serve as an early regulator during chemotactic guidance.
Soon after discovering the axonal growth cone, Ramón y Cajal suggested that factors secreted from target cells may attract developing axons in a manner similar to cell chemotaxis (Ramón y Cajal, 1972). Many secreted axon guidance factors and their cognate receptors have since been identified (Tessier-Lavigne and Goodman, 1996), but key elements of the transduction mechanisms underlying growth cone chemotaxis remain outstanding. Several guidance cues induce Ca2+ signals in the growth cone, and asymmetric Ca2+ elevation elicits growth cone turning (Zheng, 2000; Henley and Poo, 2004; Henley et al., 2004). In particular, both growth cone attractive turning and Ca2+ influx induced by brain-derived neurotrophic factor (BDNF) and netrin-1 require activation of TRP channels (Li et al., 2005; Wang and Poo, 2005). The activation of TRP channels can be regulated by many mechanisms, including membrane insertion (Bezzerides et al., 2004), channel phosphorylation (Ramsey et al., 2006), phosphoinositides (Kwon et al., 2007), and proline isomerization (Shim et al., 2009). However, the signaling intermediates linking guidance receptor activation and TRP channel opening at the growth cone surface has remained unknown.
The phosphoinositides are important signaling lipids. Phosphoinositide 3-kinase (PI3K) phosphorylates PI(4,5)P2 to cause a dynamic increase in the relatively lower level of PI(3,4,5)P3 in the inner leaflet of the plasmalemmal bilayer (Di Paolo and De Camilli, 2006). Elevated PI(3,4,5)P3 elicits the recruitment and activation of a number of effector proteins, including the downstream kinase Akt. Polarized elevation of PI(3,4,5)P3 mediates the detection of chemoattractant gradients during chemotaxis of Dictyostelium and neutrophils (Parent et al., 1998; Servant et al., 2000; Wang et al., 2002). Intriguingly, perturbing PI3K function disrupts axon guidance (Ming et al., 1999; Chang et al., 2006; Wolf et al., 2008; Akiyama and Kamiguchi, 2010). However, the spatiotemporal profile of PI(3,4,5)P3 elevation in the growth cone, the role of downstream effectors like Akt, and the link between PI(3,4,5)P3 and Ca2+ signaling during chemotactic growth cone guidance has remained completely unknown.
Here we report for the first time that polarized PI(3,4,5)P3 elevation and Akt signaling mediate growth cone detection of chemoattractive guidance cues. In cultures of Xenopus spinal neurons, chemoattractive turning of growth cones induced by netrin-1 and BDNF required Akt activity, and a gradient of BDNF rapidly triggered the accumulation of PI(3,4,5)P3 at the growth cone's leading edge, as revealed by the translocation of a GFP-tagged PI(3,4,5)P3-binding domain of Akt (PHAkt-GFP). A gradient of exogenous PI(3,4,5)P3 was also sufficient to induce attractive growth cone turning. Uniform elevation of Akt activity in the nervous system of Xenopus embryos disrupted axon pathfinding in vivo. Growth cone guidance by PI(3,4,5)P3 required both TRP channel function and downstream Ca2+ signaling. Furthermore, patch-clamp analysis revealed that PI(3,4,5)P3 and Akt positively regulate TRP channel-dependent Ca2+ influx at the growth cone. Together, these findings reveal the essential role of asymmetric PI(3,4,5)P3 elevation and Akt activation in the transduction mechanism linking guidance receptors to cytoplasmic Ca2+ signaling during growth cone chemotaxis.
Materials and Methods
Primary cell culture of Xenopus spinal neurons.
We maintained wild-type Xenopus laevis (Nasco and Xenopus One) in approved animal facilities (UC Berkeley and Mayo Clinic) according to institutional guidelines. In vitro fertilization and dissociated cell culture from stage 22 embryos of either sex were described previously (Zheng et al., 1994; Henley et al., 2004). We plated cells onto coverglass 14 h before experimentation. Reagents were from Sigma unless indicated otherwise.
Quantitative assay of growth cone turning.
Calibrated micropipettes produced microscopic gradients resulting in a 1000-fold concentration decrease at the growth cone compared to the solution in the pipette, as described previously (Zheng et al., 1994). The micropipettes contained netrin-1 (5 μg/ml; M. Tessier-Lavigne, Genentech, South San Francisco, CA), BDNF (50 μg/ml; Peprotech), synthetic PI(3,4,5)P3 with neomycin carrier (400 and 266 μm, respectively; Echelon), or ionomycin (1 μm; Calbiochem). Pharmacological agents were applied as stated in the figure legends for 15 min before the start of the assay at the following concentrations: 3.33 μm LY294002, 5 μm Akti, or 1 μm BAPTA-AM. We monitored neurite growth for 15 min to determine the initial direction of extension, and the micropipette was positioned at a 45° angle relative to this initial direction of extension. After 1 h, we measured the change in direction of extension relative to the initial trajectory.
Quantitative immunofluorescence analyses of Akt function.
Spinal neuron cultures were first fixed in PBS with 4% formaldehyde, permeabilized with 0.1% Triton X-100, and blocked with 5% goat serum. Cultures were then stained with primary antibodies against phospho-Akt (16.7 μg/ml, Rockland, 600-401-268), phospho-Akt substrate (10 μg/ml, Cell Signaling Technology, 9611), and/or HA (10 μg/ml, Cell Signaling Technology, 2367) and the appropriate Alexa dye-labeled secondary antibodies (4 μg/ml, Invitrogen). Last, we stained for total protein using 5-(4,6-dichlorotriazinyl)aminofluorescein (20 μm, DTAF, Invitrogen). We obtained images using a Zeiss 5-LIVE with 100× 1.4 NA objective. ImageJ (NIH) software was used to determine the mean thresholded fluorescence intensity within a region of interest containing the growth cone. CA-Akt expression was determined based on HA fluorescence. Values for phosphorylated Akt (pAkt) and Akt substrate phosphorylation site (pSub) were normalized to DTAF values in the same region of interest to control for fluctuations in protein levels in the growth cone. All values were normalized to the appropriate control condition.
Embryo injections and live-cell imaging.
We injected embryos at the two- to four-cell stage with ∼10 nl of DNA encoding the PI(3,4,5)P3 biosensor PHAkt-GFP (200 ng/ml; T. Balla, National Institutes of Health, Bethesda, MD). Some embryos were coinjected with rhodamine-dextran (250 μm; Invitrogen) as a general cytoplasmic tracer. Embryos with PHAkt-GFP-expressing spinal cords were selected for culture at stage 22. We plated neurons on coverglass bottom dishes for confocal imaging (Zeiss 5-LIVE and Leica TCS SP) and collected images at 20 s intervals throughout the experiment, starting 2 min before treatment with BDNF, exogenous PI(3,4,5)P3, or control solutions as indicated in the figure legends. Overexpression of PHAkt-GFP can act as a dominant-negative inhibitor of Akt function, which precludes its utility for functional assays (Akiyama and Kamiguchi, 2010). We chose growth cones with low PHAkt-GFP expression that had maintained filopodial activity for the imaging experiments. Gradient applications were from a micropipette positioned at a 45° angle relative to the long axis of the growth cone to recapitulate the conditions used for the quantitative assays of growth cone turning. For image analysis (ImageJ, NIH), we thresholded the gray level above background levels and measured the mean fluorescence intensity for pixels in regions of interest containing the third of the growth cone nearest (FN) and farthest (FF) from the pipette, respectively (Henley et al., 2004). We then calculated the asymmetry index (FN/FF × 100) for each time point for every growth cone.
Assays of axon guidance in vivo.
Messenger RNA was transcribed in vitro (mMessage machine, Ambion) from linearized plasmid pSP64 DNA (16244; Addgene) encoding the HA-CA-Akt sequence (Zhou et al., 2000) and pT3TS plasmid DNA encoding DsRed (S. Ekker, Mayo Clinic, Rochester, MN). A single cell of four- to eight-cell stage embryos was injected with ∼10 nl mRNA encoding either HA-CA-Akt (1.5 μg/ml) or DsRed (1.5 μg/ml) so that expression was limited to one side of the dorsal spinal cord. Spinal cords were prepared for imaging as previously described (Moon and Gomez, 2005). Stage 27 embryos were fixed overnight in a PBS solution containing 4% formaldehyde and 4% sucrose. After dissection, the spinal cords were permeabilized and washed before incubating overnight with primary antibodies against neurofilament-associated antigen (1:25; 3A10, Iowa Hybridoma Database) and anti-HA (4 μg/ml; ab9110, Abcam). Secondary antibodies (8 μg/ml; Alexa 488 and 594, Invitrogen) were applied for 2 h. Spinal cords were pinned and imaged with a Bio-Rad confocal microscope using a 20× NA 0.5 water-immersion objective. Antibody 3A10 labels commissural axons that originate in both sides of the spinal cord and then cross the midline to the contralateral side. Therefore, 3A10-immunopositive axons that were double labeled with anti-HA or DsRed, respectively, were selected for analysis since these axons originated from one side only (the injected side) and crossed the midline in the same direction. Images were analyzed using ImageJ software (NIH). The approach angle and crossing angle were measured as the angle an axon deviated from a line perpendicular to the midline in a 30 μm region approaching the midline, or a 60 μm region centered on the midline, respectively. For these measurements, a line perpendicular to the midline was set as 0°, and axons that had turned anteriorly were defined as positive, whereas those that had turned posteriorly were defined as negative. For Figure 4D, the mean absolute value of the approach angles is shown to demonstrate the change in deviation from perpendicular.
Whole-cell recordings using amphotericin B perforation were made on the growth cone of Xenopus spinal neurons, using a patch-clamp amplifier (Axopatch 200B, Molecular Devices). Recording pipettes had resistances of 5–8 MΩ and were filled with an internal solution consisting of (in mm) 125 K+ gluconate, 20 KCl, 10 HEPES, 0.2 EGTA, 1 MgCl2, and 1 NaCl, adjusted to pH 7.4 with KOH and an osmolarity of 290–296 mOsm. We added amphotericin B (200 μg/ml) to the internal solution immediately before recording. The bathing solution contained the following (in mm): 140 NaCl, 35 TEA-Cl, 1 CaCl2, 10 HEPES. In experiments that determined the I/V relationship, we isolated the putative TRP current with a modified internal solution. The modified internal solution replaced K+ gluconate and KCl with CsCl, and the following blockers of voltage-dependent Na+ and Ca2+ (VDCC) channels were added in the bath: TTX (1 μm; Na+ channel), calcicludine (200 nm; L-, N-, P-type VDCC; Calbiochem) (Schweitz et al., 1994), Ni2+ (50 μm; T-type VDCC), mibefradil (5 μm; T-type VDCC), and pimozide (70 μm; T-type VDCC). For measurement of I/V characteristics, step depolarizations (10 mV, 100 ms) from a holding membrane potential of −60 to +90 mV or a 100 ms voltage ramp from −60 to +40 mV were generated using P-Clamp 8.0 (Molecular Devices). Data were collected at 10 kHz and filtered at 2 kHz. A second micropipette was used for focal application of netrin-1 or synthetic PI(3,4,5)P3 mixed with the neomycin sulfate carrier solution (resulting in concentration of 5 ng/ml and 400 nm at the growth cone, respectively).
A morpholino oligonucleotide designed to target a 25 nt sequence in the 5′UTR of the reported gene sequences of Xenopus TRP-1 (Bobanovic et al., 1999; Brereton et al., 2000) (5′-CTCTGATAAAGAGCAGCCATGATGA) and a control morpholino oligonucleotide based on this sequence with 5 randomized (misprimed) nucleotides (5′-CTgTcATAAAGAcCAGgCATcATGA) or standard control morpholino (Ctrl MO) (5′-CCTCTTACCTCAGTTACAATTTATA) were from Genetools. These morpholino oligonucleotides were conjugated to lissamine for verification using fluorescence microscopy. The neurons were loaded with these morpholino oligonucleotides by early embryo injection following the method previously described (Wang and Poo, 2005).
Data were analyzed with GraphPad Prism software (v5). The figure legends state the statistical tests used. Statistical comparisons involving turning assay experiments and in vivo turning measurements used the Mann–Whitney U test due to the nonparametric distribution of the data. For all other data with a normal distribution, statistical comparisons used the two-tailed Student's t test (see Figs. 2C, 6) and repeated-measures two-way ANOVA (see Fig. 2B).
Requirement of PI(3,4,5)P3 and Akt signaling during growth cone chemoattraction
We examined the role of PI(3,4,5)P3 and Akt signaling in growth cone chemoattraction induced by extracellular guidance cues, using cultured Xenopus spinal neurons and a quantitative assay of growth cone turning and extension (Zheng et al., 1994). Consistent with previous findings (Song et al., 1997; Ming et al., 1999), pulsatile ejection of netrin-1 or BDNF from a calibrated micropipette produced a microscopic gradient of these factors that led to significant growth cone chemoattraction compared to a control solution (Fig. 1A–E). Uniform bath application of a cell permeant inhibitor of Akt (Akt inhibitor 3, abbreviated “Akti”; 5 μm) (Kozikowski et al., 2003) abolished growth cone chemoattraction induced by BDNF and netrin-1 (Fig. 1B–E). The rate of axon extension was unaffected by the Akt inhibitor at the concentration used for turning assays (Fig. 1F), but higher doses of Akti slowed outgrowth (data not shown). These results indicate that growth cone chemoattraction induced by these guidance factors requires PI(3,4,5)P3 production and Akt activation.
Polarization of PI(3,4,5)P3 during growth cone chemoattraction
To determine the spatiotemporal pattern of PI(3,4,5)P3 signals during growth cone chemoattraction, we expressed a biosensor fusion protein comprising the PI(3,4,5)P3 binding PH-domain of Akt fused to GFP (Várnai and Balla, 1998) in developing Xenopus embryos (Fig. 2). Confocal imaging of cultured myocytes revealed that the expressed fusion protein exhibited a diffuse cytoplasmic distribution that translocated to the plasmalemma in response to the elevated PI(3,4,5)P3 in the membrane induced by insulin (Toker and Cantley, 1997; Servant et al., 2000; Wang et al., 2002) (data not shown). We used this approach to monitor PI(3,4,5)P3 in spinal neuron growth cones. Uniform bath application of BDNF triggered translocation of the diffuse cytoplasmic PHAkt-GFP to the plasmalemma within minutes, indicating accumulation of plasmalemmal PI(3,4,5)P3. The PHAkt-GFP translocation showed a global plasmalemmal distribution with stochastic fluctuations (Fig. 2A, Movie 1). When a microscopic gradient of BDNF was applied to individual growth cones, the cytoplasmic PHAkt-GFP appeared to translocate to the surface membrane in a polarized manner: the PHAkt-GFP translocation fluctuated but in general showed a preferential elevation on the side nearer the source of the BDNF gradient (Fig. 2C,E, Movie 2). Growth cones expressing PHAkt-GFP and double labeled with the fluorescent cytoplasmic tracer rhodamine-dextran (Rd-dx) confirmed that the PHAkt-GFP signal became less cytoplasmic and more enriched in the plasmalemma during the BDNF gradient (Fig. 2E). Imaging for a longer time period revealed that the preferential elevation on the side nearer to the gradient source persisted for >25 min (Fig. 2E).
To analyze the distribution of the PHAkt-GFP biosensor across the growth cone, we made quantitative ratiometric measurements and compared the mean fluorescence intensity at the near (FN; closer to the source of BDNF) versus the far (FF) side of the growth cone (Fig. 2F, see Materials and Methods). Analysis of an example growth cone during uniform BDNF application showed a relatively even distribution of the PHAkt-GFP biosensor on both sides of the growth cone (Fig. 2B). In contrast, a growth cone treated with a BDNF gradient had a higher elevation of PHAkt-GFP on the side of the growth cone nearer the gradient source (Fig. 2D). Similar analysis of all growth cones treated with uniform or gradient application of BDNF confirmed a significant difference in the asymmetry of the PHAkt-GFP signal between the treatment conditions (Fig. 2G) (repeated-measures two-way ANOVA). Measurements of peak asymmetry (FN:FF, see Materials and Methods) for each growth cone demonstrated a significant difference in the spatial distribution of PI(3,4,5)P3 signals induced by a gradient of BDNF (127.0 ± 6.9% SEM) compared to a uniform concentration of BDNF (100.9 ± 4.1% SEM) or before gradient application (100.6 ± 1.4 SEM; p < 0.01, Student's t test) (Fig. 2H). Thus, analogous to chemotaxis of amoeboid cells, PI(3,4,5)P3 accumulation in the growth cone plasmalemma is polarized toward the source of chemoattractant gradient.
Growth cone steering by exogenous PI(3,4,5)P3
Is asymmetric accumulation of PI(3,4,5)P3 alone sufficient to induce growth cone turning? We addressed this question directly by applying a microscopic gradient of exogenous PI(3,4,5)P3 in a solution containing a carrier molecule that enables its incorporation into the inner leaflet of the plasmalemma (see Materials and Methods) (Ozaki et al., 2000; Wang et al., 2002). When a gradient of exogenous PI(3,4,5)P3 together with the carrier was applied to a growth cone expressing the PI(3,4,5)P3 biosensor PHAkt-GFP, we observed a marked asymmetric translocation of PHAkt-GFP to the plasmalemma (Fig. 3A). This indicates that the exogenous PI(3,4,5)P3 is able to incorporate into the cytoplasmic side of the growth cone plasmalemma. Similar application of a solution containing the carrier alone caused no detectable PHAkt-GFP translocation (data not shown). Remarkably, application of the exogenous PI(3,4,5)P3 gradient triggered significant growth cone attraction, whereas a gradient of the carrier alone had no effect (Fig. 3B–D,F,G). A gradient of exogenous PI(3,4,5)P3 without the carrier caused no preferential growth cone turning, indicating that incorporation into the cytoplasmic side of the plasmalemma is necessary to induce growth cone attraction. The related phosphoinositide PI(4,5)P2, when applied exogenously together with the carrier, also caused no preferential growth cone turning (Fig. 3G). This demonstrates that PI(3,4,5)P3-induced chemoattraction is distinct rather than a general characteristic of all phosphoinositides. Application of either phosphoinositide as a gradient had no effect on the rate of axon extension (Fig. 3H). Thus, asymmetric PI(3,4,5)P3 signaling is not only necessary for chemotactic growth cone turning induced by BDNF and netrin-1, but also sufficient by itself to elicit growth cone chemotaxis.
We next tested whether growth cone turning to the exogenous PI(3,4,5)P3 gradient required Akt and PI3K activities. Application of Akti completely abolished the PI(3,4,5)P3-induced attraction (Fig. 3E–G). Treatment with the PI3K inhibitor LY294002 attenuated the growth cone attraction induced by PI(3,4,5)P3 (Fig. 3G), but the attenuation did not reach statistical significance (Mann–Whitney U test) when compared to the turning induced by either the PI(3,4,5)P3 gradient or the vehicle control alone. Neither treatment affected the rate of axon extension (Fig. 3H). These findings indicate the critical role of downstream Akt activation during PI(3,4,5)P3-induced growth cone turning.
Axon guidance in vivo requires asymmetric Akt activity
To further examine the role of Akt signaling in axon guidance, we used a molecular approach to dysregulate Akt function. We expressed an HA-tagged constitutively active form of Akt (CA-Akt) (Zhou et al., 2000) in the nervous system of Xenopus embryos by injection of mRNA (see Materials and Methods). Expression of CA-Akt caused no gross neural tube defects such as those reported after expressing a dominant-negative form of Akt (Peng et al., 2004) (data not shown). To demonstrate the effect of CA-Akt expression on spinal neuron growth cones, we used immunofluorescence to stain for pAkt and pSub (Fig. 4A). Quantitative measurements of both antibody staining patterns compared to a total cellular protein stain demonstrated a significant increase in phosphorylation (Fig. 4B). The functional growth cone turning assay revealed that expression of CA-Akt completely abolished growth cone attraction to a gradient of BDNF (Fig. 4C,D). Expression of CA-Akt had no effect on the rate of growth cone extension (Fig. 4E). Together, these results validate that expression of CA-Akt dysregulates Akt signaling in the growth cone, and support our findings with the Akt inhibitor.
To test whether dysregulated Akt signaling affects axon pathfinding in vivo, we assessed the guidance of commissural interneuron axons at the ventral midline during embryonic development. In this well defined model system, axons project ventrally toward the midline, guided by gradients of chemoattractant cues like netrin-1 (Fig. 5A,B) (Serafini et al., 1996; Kennedy et al., 2006). Commissural axons project toward the midline, which they cross at an angle slightly rostral of perpendicular. After midline crossing, they make a sharp turn rostrally and project toward targets in the brain. We expressed either the fluorescent protein DsRed or CA-Akt (Zhou et al., 2000) in one side of the nervous system of Xenopus embryos by ipsilateral injection of mRNAs (see Materials and Methods). Immunostaining with the antibody 3A10 identified commissural axons specifically (Moon and Gomez, 2005).
We first measured the approach angle of axons toward the ventral midline (0°: perpendicular to the midline; see Materials and Methods). For control axons expressing DsRed alone or noninjected wild-type axons, the mean approach angle was nearly perpendicular to the midline (8.4° ± 0.75 SEM and 7.3° ± 0.76, respectively) (Fig. 6A,C,D). In contrast, the mean approach angle of axons expressing CA-Akt (17.1 ± 1.1°) showed significant deviation from the control groups, with a significantly wider distribution than control axons (p < 0.05, F test) (Fig. 6B–D). The wider range of paths taken by CA-Akt-expressing axons demonstrates that these axons are less well guided to the midline. We next measured the angle of axon midline crossing in a 60 μm region centered at the midline (0°: perpendicular to the midline, anterior angles as positive, see Materials and Methods). The mean crossing angle of control DsRed and wild-type axons was anterior (8.9 ± 2.0° and 7.1 ± 1.1°, respectively) (Fig. 6A,E,F). In contrast, the axons expressing CA-Akt had a significantly posterior mean crossing angle (−7.3° ± 2.3), with a wider distribution of midline crossing angles compared to the controls (p < 0.05, F test) (Fig. 6B,E,F). Expression of CA-Akt also disrupted the guidance of postcrossing axons, with only 45.3% turning rostrally after midline crossing compared to 96.6% of the control DsRed-expressing axons. Thus uniform elevation of Akt activity caused significant guidance errors at the midline, supporting the notion that detection of endogenous guidance cues requires regulated Akt activation.
PI(3,4,5)P3 and Akt signaling act upstream from Ca2+ signals
Intracellular Ca2+ signals mediate the effects of many chemoattractant cues, including BDNF and netrin-1 (Hong et al., 2000; Li et al., 2005; Shim et al., 2005; Wang and Poo, 2005), and asymmetric Ca2+ signals in the growth cone can directly induce attractive turning (Zheng, 2000; Henley et al., 2004). However, the link between guidance receptor activation and the induction of Ca2+ signals during chemoattraction remains unclear. We therefore examined whether PI(3,4,5)P3 and Akt signaling is upstream or downstream of Ca2+ signals. Treatment of these cultured spinal neurons with the cell permeant Ca2+-buffer BAPTA-AM (1 μm) abolished the chemoattractive growth cone turning induced by a gradient of exogenous PI(3,4,5)P3 applied in a carrier solution (Fig. 7A,B,E,F). In contrast, growth cone turning resulting from an asymmetric cytoplasmic Ca2+ signal, via application of a gradient of the Ca2+-ionophore ionomycin in a bath solution containing defined Ca2+ concentration (Henley et al., 2004), persisted in the presence of Akti (Fig. 7C–F). Neither treatment affected the rate of axon extension (Fig. 7G). These findings suggest that PI(3,4,5)P3 and Akt signaling act upstream of Ca2+ signals during growth cone chemotaxis.
PI(3,4,5)P3 and Akt signaling activate growth cone TRP channels
Activation of Ca2+-permeant TRP channels and the resultant Ca2+ influx at the growth cone of Xenopus spinal neurons are necessary for growth cone chemotaxis induced by netrin-1 and BDNF (Li et al., 2005; Shim et al., 2005; Wang and Poo, 2005). We therefore examined whether TRP channel activity and Ca2+ influx are causally linked to PI(3,4,5)P3 signaling. We first tested whether Xenopus TRPC1 (xTRPC) is required for PI(3,4,5)P3-induced chemoattraction. It has been shown previously (Wang and Poo, 2005) that injection of a specific antisense morpholino oligonucleotide in the Xenopus embryo (see Materials and Methods) efficiently downregulates the expression of xTRPC1 and abolishes the TRP current induced by netrin-1 in these cultured spinal neurons. We used this same antisense morpholino approach to downregulate the expression of xTRPC1 during the growth cone turning assay. Downregulating the expression of xTRPC1 by the xTRPC1 morpholino (TRPC MO) abolished growth cone attraction induced by a gradient of exogenous PI(3,4,5)P3 plus carrier, whereas a control morpholino (Ctrl MO) had no effect (Fig. 7F). Neither morpholino affected the rate of axon extension (Fig. 7G).
To better understand the link between PI(3,4,5)P3 and TRPC, we monitored plasmalemmal ion channel activity directly by performing whole-cell patch recordings from the growth cone (Fig. 8A). Perforated patch-clamp recordings of membrane potential in current-clamp mode revealed a significant depolarization induced by local application of netrin-1 from a micropipette (Fig. 8A,B), as was reported previously (Li et al., 2005; Wang and Poo, 2005). Similar local application of exogenous PI(3,4,5)P3 in the carrier solution caused a transient membrane potential depolarization of a lower amplitude, whereas application of the carrier solution alone had no effect (Fig. 8B). The PI(3,4,5)P3-induced depolarization was abolished by treatment with the TRP channel blocker SKF-96365 (25 μm) (Fig. 8B) (Zhu et al., 1998).
To control for potential off-target effects of SKF-96365 on other channels, we isolated TRP currents with pharmacological agents that blocked most K+ channels, Na+ channels, and L-, N-, P-, and T-type Ca2+ channels (see Materials and Methods). We then applied a voltage ramp to determine the current–voltage (I/V) relationship, and observed a small, putative TRP current in untreated growth cones (Fig. 8C,D). Local application of exogenous PI(3,4,5)P3 in the carrier solution markedly elevated this current to a level similar to that induced by netrin-1 applied either locally by a micropipette or uniformly in the bath solution. Subsequent addition of SKF-96365 reduced the current to a level below that found before the application of PI(3,4,5)P3. The carrier solution alone caused no change in the current compared to untreated controls (Fig. 8D). Local application of PI(3,4,5)P3 in the presence of uniform bath-applied netrin-1 caused no further elevation of the putative TRP current (Fig. 8D), suggesting that the same channels are activated by both netrin-1 and PI(3,4,5)P3. Using the morpholino knockdown approach (KD), we found that downregulation of xTRPC1 inhibited the putative TRP current in the growth cone induced by locally applied PI(3,4,5)P3, whereas a control misprimed oligonucleotide (MP) had no effect on the PI(3,4,5)P3-induced Ca2+ current (Fig. 8C,D). Downregulating xTRPC1 expression also attenuated the basal putative TRP currents that were detected in growth cones from the untreated and MP-treated embryos (Fig. 8D). These results indicate that PI(3,4,5)P3 alone is sufficient to activate TRP channels in these spinal neurons, and thus is capable of eliciting Ca2+ influx that mediates chemoattractive growth cone turning.
We further examined whether the PI(3,4,5)P3-induced TRP current requires PI3K and Akt activities. Treatment of the cultured spinal neurons with either LY294002 or Akti significantly attenuated the TRP current induced by a gradient of exogenous PI(3,4,5)P3 applied in the carrier solution compared to the respective inhibitor alone controls (Fig. 8C,D). Treatment with Akti, SKF, or morpholino knockdown of xTRPC1 significantly attenuated the PI(3,4,5)P3-induced TRP current when compared to growth cones treated with PI(3,4,5)P3 alone (Fig. 8C,D). Treatment with LY294002 appeared to attenuate both the basal and PI(3,4,5)P3-induced currents but was not significantly different from the respective no-inhibitor conditions (Fig. 8D). Altogether, these findings reveal that PI(3,4,5)P3-induced activation of Akt elicits xTRPC1-dependent Ca2+ influx in the growth cone.
Our findings support the following model of early signal transduction at the growth cone induced by netrin-1 and BDNF (Fig. 9): An environmental gradient of these guidance cues induces a gradient of activation of their cognate receptors at the growth cone surface membrane. Activated receptors in turn trigger the activation of PI3K (Chao, 2003; Round and Stein, 2007) and local production of PI(3,4,5)P3, which accumulates asymmetrically at the plasmalemma, leading to asymmetric activation of Akt. The activated Akt in turn leads to potentiation of TRPC channels in the membrane, allowing the influx of Ca2+ required for triggering chemoattractive growth cone turning. Thus, asymmetric PI(3,4,5)P3 signals are essential and instructive for chemotactic growth cone guidance by netrin-1 and BDNF. Furthermore, to our knowledge, this is the first demonstration that Akt signaling is critical for axon guidance. Analogous to the chemotaxis of amoeboid cells, the asymmetric membrane accumulation of PI(3,4,5)P3 represents an early signal downstream of guidance receptor activation that polarizes to the leading edge during growth cone chemotaxis. Furthermore, our discovery of the PI(3,4,5)P3/Akt/TRPC signaling cascade marks a novel mechanism for controlling Ca2+ influx. These findings on PI(3,4,5)P3 and Akt signaling in the growth cone during gradient detection underscore Ramón y Cajal's vision of the growth cone as a chemotaxing amoeba and support the notion that PI(3,4,5)P3 and Akt signaling is a conserved mechanism for chemotactic guidance among diverse cell types.
PI(3,4,5)P3 regulation of TRPC
Phosphoinositides have been shown to bind and directly regulate multiple TRP channels (Kwon et al., 2007). Evidence also suggests that the BDNF receptor TrkB can gate TRP channels by a direct interaction (Bobanovic et al., 1999; Chao, 2003). Our finding that the PI(3,4,5)P3-induced potentiation of TRPC channels in the growth cone requires Akt activity identifies a new mechanism. Xenopus TRPC lacks the Akt phosphorylation consensus sequence RXRXXS/T (Alessi et al., 1996), suggesting that Akt may regulate TRPC function through an intermediary. Although the molecular details remain to be defined, PI3K-dependent insertion of TRPC5 into the plasmalemma has been demonstrated in hippocampal neurons (Bezzerides et al., 2004). However, this same study showed that TRPC1, which is most similar to xTRPC by sequence homology (Bobanovic et al., 1999), is not regulated in the same manner. Whether BDNF and netrin-1 regulate xTRPC distribution at the plasmalemma awaits further investigation.
Our patch-clamp recordings demonstrated that maximal PI(3,4,5)P3-induced activation of TRPC channels required PI3K activity, suggesting that a PI(3,4,5)P3–PI3K positive feedback loop may regulate TRPC currents in the growth cone. We also found that inhibiting PI3K trended toward attenuating growth cone turning to a gradient of exogenous PI(3,4,5)P3, but this did not reach statistical significance. The potential role of a feedback loop would provide a robust mechanism for breaking symmetry in the process of developing polarized growth cone responses. Indeed, cell polarization and chemotaxis in neutrophils (Weiner et al., 2002; Inoue and Meyer, 2008) requires a PI(3,4,5)P3–PI3K positive feedback loop. However, we cannot conclusively demonstrate that such a feedback loop is required for TRPC activation and growth cone chemoattraction. It has recently been shown that photoactivated release of caged Ca2+ in the growth cone leads to growth cone attraction that is blocked with the PI3K inhibitor LY294002 (Akiyama and Kamiguchi, 2010). Together with our findings that PI(3,4,5)P3 leads to Ca2+ influx through TRPC in the growth cone, these findings raise the possibility of a PI(3,4,5)P3–Ca2+ feedback mechanism during growth cone guidance.
PI(3,4,5)P3 as a signaling node
The early PI(3,4,5)P3 elevation at the growth cone's leading edge is poised to function as a signaling node that coordinates the localization and activity of a diverse set of cellular machinery during growth cone chemotaxis. For example, the PI(3,4,5)P3/Akt transduction pathway can regulate target of rapamycin (TOR)-dependent protein synthesis (Potter et al., 2002), which is required for retinal ganglion cell axon guidance (Campbell and Holt, 2001). A rapamycin-insensitive TOR complex can also affect actin dynamics in many cell types (Jacinto et al., 2004; Sarbassov et al., 2004). Therefore, whereas our findings demonstrate the Akt-dependent regulation of TRPC, Akt is likely to play additional roles in axon guidance.
It is known that PI(3,4,5)P3 can also regulate multiple effector proteins independent of Akt. For example, PI(3,4,5)P3 can bind and activate the detector of cytokinesis (DOCK180) family of GTP exchange factors (GEFs), which positively regulates Rac activation at the plasmalemmal leading edge in chemotaxing neutrophils (Côté et al., 2005; Nishikimi et al., 2009). Activated Rac in turn can regulate actin and membrane dynamics (Ridley, 2006), and is required for bidirectional growth cone guidance (Yuan et al., 2003). Recently, it has been demonstrated that PI(3,4,5)P3 controls microtubule extension in the growth cone, providing yet another mechanism by which PI(3,4,5)P3 can regulate the cytoskeleton. Thus, PI(3,4,5)P3 can potentially function as a signaling node that directly regulates key effector proteins in addition to Akt, although the specific molecular interactions in the growth cone remain to be defined.
Removal of the 5′ phosphate from PI(3,4,5)P3 yields another key second messenger, PI(3,4)P2. Lamellipodin binds PI(3,4)P2 and recruits Ena/VASP family members, which regulate actin dynamics (Krause et al., 2004). Null mutations of lamellipodin and Ena/VASP disrupt axon guidance in Caenorhabditis elegans (Chang et al., 2006). A key 5′ phosphatase that catalyzes the conversion of PI(3,4,5)P3 to PI(3,4)P2 is the Src homology 2 domain-containing inositol-5′-phosphatase (SHIP), which is required for neutrophil chemotaxis (Nishio et al., 2007). Thus, the regulated conversion of PI(3,4,5)P3 to PI(3,4)P2 could extend the signaling network that originates from PI(3,4,5)P3 during growth cone chemotaxis.
Removal of the 3′ phosphate by phosphatase and tensin homolog (PTEN) converts PI(3,4,5)P3 to PI(4,5)P2 and suppresses PI(3,4,5)P3/Akt signaling (Di Paolo and De Camilli, 2006). Chemotaxis in Dictyostelium requires PTEN, which is recruited to the plasmalemma and functions to polarize signals by excluding PI(3,4,5)P3 from the lagging edge (Iijima and Devreotes, 2002). The role of PTEN in the spatial and temporal regulation of PI(3,4,5)P3 signals in the growth cone remains unknown. Intriguingly, the collapsing factor semaphorin 3A induces the recruitment of PTEN to the growth cone plasmalemma (Chadborn et al., 2006), which may suppress growth-promoting PI(3,4,5)P3 signals at the leading edge (Markus et al., 2002). Whether PTEN is recruited to the plasmalemma and facilitates polarization of PI(3,4,5)P3 guidance signals during growth cone chemotaxis awaits further investigation.
Together, our findings have revealed that PI(3,4,5)P3 elevation polarizes to the leading edge and may serve as an early regulator during growth cone guidance. Furthermore, we have identified PI(3,4,5)P3 and Akt signaling as a novel link to Ca2+ influx in the growth cone. Potentiating Akt signaling (Namikawa et al., 2000) or downregulating PTEN expression (Park et al., 2008) can promote regenerative axon outgrowth at sites of injury. These results, together with our finding that a gradient of exogenous PI(3,4,5)P3 can direct growth cone chemotaxis, suggest that selective manipulation of the PI(3,4,5)P3/Akt signaling pathway may hold therapeutic promise for enhancing functional axon regeneration.
Supplemental material for this article is available at http://www.youtube.com/user/HenleyLab and consists of additional movies of the PHAkt-GFP biosensor associated with the paper. These include PHAkt-GFP translocation induced by insulin, BDNF, and exogenous synthetic PI(3,4,5)P3. This material has not been peer reviewed.
This work was supported by a John M. Nasseff, Sr., career development award from the Mayo Clinic (J.R.H.), the National Institutes of Health (NIH) (J.R.H.; M.-m.P.), and a Robert D. and Patricia E. Kern Predoctoral Fellowship award (S.J.H.). We thank S. Wong (University of California, Berkeley) for discussions, T. Balla (NIH) for the gift of the PHAkt-GFP construct, S. Ekker (Mayo Clinic) for the DsRed RNA expression vector, T. Gomez (University of Wisconsin, Madison) and G. Sieck (Mayo Clinic) for assistance with in vivo imaging, M. Tessier-Lavigne (Genentech) for providing netrin-1, and A. Windebank (Mayo Clinic) for sharing laboratory space.
The authors declare no competing financial interests.
- Correspondence should be addressed to John R. Henley, Department of Neurologic Surgery, Mayo Clinic, 200 First Street SW, Rochester, MN 55905.