The rate and direction of axon and dendrite growth depend on multiple guidance signals and growth factors. Semaphorin 3A (Sema3A) acts as a repellent for axons and attractant for dendrites. Here, we show that the requirement for integrin engagement distinguishes the response of axons and dendrites to Sema3A in hippocampal neurons. Sema3A promotes the extension of hippocampal dendrites by a pathway that requires focal adhesion kinase (FAK). The stimulation of dendrite growth and FAK phosphorylation by Sema3A depend on integrin engagement. Unlike their function as a target of Sema3A during the collapse of axonal growth cones, integrins facilitate the stimulation of dendrite extension. Conditional inactivation of the genes encoding β1 integrin or FAK blocks the growth-promoting effect of Sema3A but not the collapse of axonal growth cones. Our results demonstrate that different pathways mediate the stimulation of dendrite growth and the collapse of axonal growth cones by Sema3A.
The ephrins, netrins, Slit proteins and semaphorins are four families of guidance signals that direct the extension of axons and dendrites, and serve as attractive or repellent cues during the establishment of neuronal connections (Dickson, 2002; Goldberg, 2004; Huber et al., 2003; Jan and Jan, 2003; Kim and Chiba, 2004). The effect of these signals depends on the composition of receptor subunits and the intracellular concentration of second messengers such as cAMP and cGMP (Corset et al., 2000; Höpker et al., 1999; Nishiyama et al., 2003; Song et al., 1998; Song and Poo, 1999; Yu and Bargmann, 2001). Sema3A acts as a repellent for many types of axons and induces the collapse of growth cones by initiating the depolymerisation of actin filaments, promoting endocytosis and reducing integrin-mediated adhesion (Fournier et al., 2000; Jin and Strittmatter, 1997; Kuhn et al., 1999; Mikule et al., 2002). Manipulating the concentration of cGMP can convert the repellent Sema3A into an attractive cue (Campbell et al., 2001; Song et al., 1998). The repellent effects of the secreted class-3 semaphorins are mediated by a receptor complex that contains neuropilin-1 (Nrp-1) or Nrp-2 as the ligand-binding subunit and an A-type plexin as the signal-transducing subunit (Kruger et al., 2005; Püschel, 2007; Tran et al., 2007). Plexins act as GTPase-activating proteins (GAPs) for R-Ras and antagonise integrin-dependent cell adhesion by reducing the local concentration of GTP-bound R-Ras (Kinbara et al., 2003; Kruger et al., 2005; Oinuma et al., 2004a; Oinuma et al., 2004b). In addition to its repellent effects, Sema3A also acts as an attractive or growth-promoting signal for the dendrites of pyramidal neurons from the mouse cortex (Fenstermaker et al., 2004; Gu et al., 2003; Polleux et al., 2000; Sasaki et al., 2002). Sema3A directs the orientation of axons and dendrites by pyramidal neurons of cortical layer 5 (Polleux et al., 1998; Polleux et al., 2000; Sasaki et al., 2002). In Sema3a–/– mice, apical dendrites deviate from their normal orientation perpendicular to the pial surface (Polleux et al., 2000). In vitro assays suggest that a Sema3A gradient attracts apical dendrites to the pial surface by a signalling pathway that depends on cGMP while directing axons in the opposite direction through repulsion (Polleux et al., 1998; Polleux et al., 2000). Sema3A acts through the same receptor complex to repel axons and attract dendrites but it is unknown if the same signalling pathways are involved for both effects.
Another important factor that determines the response to guidance signals is the extracellular matrix. The integrin ligand laminin converts the attractive effect of netrin-1 on Xenopus spinal-cord axons to a repulsive response (Höpker et al., 1999). The genetic interaction between mutants for integrins or their ligands and Slit suggests that integrin-dependent signalling influences the sensitivity of axons to the repellent guidance signal Slit in the Drosophila CNS (Stevens and Jacobs, 2002). A modulatory role of laminin and fibronectin was also reported for mammalian Slit-2 (Nguyen-Ba-Charvet et al., 2001). The tyrosine kinase focal adhesion kinase (FAK) is a central component of the signalling pathway regulated by integrins (Mitra et al., 2005; Parsons, 2003). Integrin activation leads to an increase in FAK activity by the ordered phosphorylation at multiple tyrosine residues. FAK activation is initiated by the autophosphorylation at Y397, which enables the binding and activation of Src-family kinases (SFKs). SFKs in turn phosphorylate the kinase domain of FAK at Y576/577. In addition to integrins, FAK is also regulated by axon-guidance signals such as netrin-1 and ephrin-A1 (Bourgin et al., 2007; Carter et al., 2002; Li et al., 2004; Liu et al., 2004; Miao et al., 2000; Ren et al., 2004). FAK and SFKs directly interact with the netrin receptor DCC and are activated by netrin-1. Their activity is required for the attraction of axons by netrin-1.
Here we show that Sema3A promotes dendrite growth by stimulating FAK phosphorylation and increasing the intracellular concentration of cGMP. The activation of FAK and the stimulation of dendrite extension but not the collapse of axonal growth cones by Sema3A depend on the simultaneous activation of integrins. The stimulatory effect of Sema3A but not growth-cone collapse is blocked by conditional deletion of the β1 integrin (Itgb1) and FAK (Fak) genes. Thus, integrin activation facilitates the effect of Sema3A on dendrites through a signalling pathway that is not required for collapse of the growth cone.
Aberrant development of hippocampal dendrites in Sema3a-knockout mice
To investigate the role of Sema3A in the development of hippocampal neurons, the Thy1-YFPH transgene, which drives YFP expression in pyramidal neurons of the hippocampus (Feng et al., 2000), was crossed into a Sema3a-knockout strain (Taniguchi et al., 1997). Expression of the Thy1-YFPH transgene was detectable only in a few cells at early postnatal stages but became visible in a larger number of neurons at postnatal day 15 (P15) (Fig. 1). The morphology of pyramidal neurons was analysed in 60-μm sections from brains of P15 mice. The structure of the hippocampus was normal in Sema3a–/– mice, as was reported previously (Pozas et al., 2001). However, the morphology of apical dendrites showed defects in Sema3a mutants. In wild-type mice, pyramidal neurons extended apical dendrites from the stratum pyramidale through the stratum radiatum and formed branches mainly in the stratum lacunosum-moleculare or in the stratum radiatum close to the border with the stratum lacunosum-moleculare. The dendrites usually extended parallel to each other in the stratum radiatum until they reached the stratum lacunosum-moleculare, where they spread out. In Sema3a-deficient mice, this well-ordered organisation was disrupted (Fig. 1A). The majority of apical dendrites was stunted and branched soon after entering the stratum radiatum. When the relative position of the first branch point of apical dendrites in the stratum radiatum was measured, a significant difference to wild-type hippocampus was obvious (Fig. 1B). In total, 81% of the apical dendrites branched in the stratum radiatum of Sema3a–/– mice, whereas branches in the stratum radiatum were seen only in 45% of wild-type neurons (n=108 neurons each; three animals). In addition, branches were significantly closer to the stratum pyramidale in Sema3a mutants (Fig. 1). Although the orientation of apical dendrites appeared to be normal, their branches formed a disorganised dendritic arbour. The branches of apical dendrites did not extend parallel to each other as in wild-type hippocampi and showed more irregular trajectories in the stratum radiatum of Sema3a mutants. It is unlikely that the extensive disorganisation of apical dendrites in the Sema3a mutant is an indirect effect of axonal misprojections. In wild-type mice, entorhino-hippocampal projections are restricted to the stratum lacunosum-moleculare. In Sema3a mutants, only a few entorhino-hippocampal fibres extend into inappropriate layers such as the stratum radiatum (Pozas et al., 2001). These minor guidance defects are unlikely to cause the pronounced disorganisation of dendrites observed in the mutant hippocampus.
Sema3A stimulates dendrite growth
The abnormal morphology of apical dendrites in Sema3a-deficient mice might reflect an inhibitory function of Sema3A that prevents premature branching in the stratum radiatum. Alternatively, Sema3A could influence the growth of dendrites as an attractive or growth-stimulating signal as described for cortical neurons (Fenstermaker et al., 2004; Gu et al., 2003; Polleux et al., 2000; Sasaki et al., 2002). To investigate how Sema3A regulates the morphology of dendrites, we used cultures of dissociated hippocampal neurons from embryonic day 18 (E18) mouse or rat embryos with fibronectin as a substrate. At 3 days after plating [3 days in vitro (d.i.v.)], Sema3A was added for 16-20 hours and the development of dendrites was analysed after staining with an anti-MAP2 antibody. When added to cultured neurons from mouse embryos, Sema3A increased the average length of dendrites by more than 50%, from 51±4.5 μm in controls to 79±2 μm (Fig. 2) (n=30–50 neurons, three experiments). By contrast, Sema3a–/– mouse hippocampal neurons formed shorter dendrites than wild-type neurons, with an average length of 37±1 μm (Fig. 2B). Incubation of Sema3a–/– neurons with Sema3A stimulated the growth of dendrites and rescued the loss of Sema3A. The branching of dendrites was not affected by Sema3A treatment during the observed time period (data not shown). We also investigated the effect of Sema3A on neurons prepared from embryonic rat hippocampus (supplementary material Fig. S1A,B). Compared with mouse neurons, the average length of dendrites from rat neurons was slightly shorter at 4 d.i.v. (control: 33±1 μm) and was increased by the addition of Sema3A to 57±3 μm (Fig. 3) (n=3 with 75 neurons each). As already observed for the stimulation of neurite extension by PC12 cells (Schwamborn et al., 2004), a prolonged incubation with Sema3A was required to stimulate dendritic growth. When Sema3A was replaced by control medium after incubation for only 1 hour and the culture continued for 16 hours, no effect on dendritic growth was observed (data not shown). When neurons were cultured on the nonspecific adhesive poly-L-ornithine (PO), Sema3A was unable to stimulate dendrite growth and the length of dendrites from wild-type and Sema3a–/– neurons did not differ significantly (Fig. 2B). These results indicate that Sema3A does not act as an inhibitory signal for dendrite growth or branching but stimulates dendrite elongation in a substrate-dependent manner.
The stimulation of dendrite growth by Sema3A depends on integrin engagement
Sema3A increased the length of dendrites only when neurons were cultured on fibronectin but not on the nonspecific adhesive PO (Fig. 2B; Fig. 3). When rat hippocampal neurons were cultured on PO, the addition of Sema3A did not have any significant effect on the extension of dendrites (control: 32±1 μm; Sema3A: 33±1 μm; n=4 with 75 neurons each). In the presence of control medium, dendrites had an average length of approximately 30 μm on PO or fibronectin (control: 32±1 μm; n=4 with 75 neurons each). This result suggests that the stimulation of dendrite extension by Sema3A depends on the ligation of integrins by fibronectin. To test the role of integrins, we used Mn2+ ions as an activator of integrins, which stabilises the active conformation, and an RGD peptide as a specific inhibitor (Gailit and Ruoslahti, 1988; Mould et al., 1995; Takagi et al., 2002). Incubation of neurons with Mn2+ ions had no effect on dendrite length by itself but could replace fibronectin when Sema3A was added to neurons cultured on PO (Fig. 3). Blocking the interaction of fibronectin and integrins by an RGD peptide prevented the stimulatory effect of Sema3A. This indicates that Sema3A stimulates dendrite elongation only when integrins are activated.
The repellent activity of Sema3A is mediated by a receptor complex that consists of Nrp-1 and A-type plexins (Tran et al., 2007). To test whether these receptor subunits are also required for the stimulatory activity of Sema3A on dendrites, we transfected rat hippocampal neurons at 2 d.i.v. with vectors for dominant-negative Nrp-1 or plexin-A1, which interfere with signalling through Nrp-1 and A-type plexins, respectively. At 3 d.i.v., neurons were treated with Sema3A for 16 hours. Both constructs blocked the effect of Sema3A on dendrites (supplementary material Fig. S1C). By contrast, expression of a dominant-negative mutant for plexin-B1, the receptor for Sema4D, did not interfere with the stimulation of neurite extension by Sema3A.
Sema3A has a potent growth cone collapsing activity for most types of neurons. It also induced the collapse of growth cones from hippocampal axons (supplementary material Fig. S2). When hippocampal neurons were cultured on fibronectin, 19±4% (n=3) of the axonal growth cones were collapsed under control conditions. Incubation with Sema3A for 1 hour increased the percentage of collapsed growth cones to 60±7%. The collapse of axonal growth cones by Sema3A was independent of the substrate and was not affected by the addition of RGD peptides or Mn2+ ions (supplementary material Fig. S2). Thus, the response of axonal growth cones to Sema3A is not influenced by integrin engagement.
Nearly all fibronectin-binding integrins belong to the β1- or αv-integrin subfamilies (Hynes, 2002). Hippocampal neurons express β1 integrin (Belvindrah et al., 2007; Huang et al., 2006). Staining of cultured hippocampal neurons for β1 integrin showed a punctate distribution in axons, dendrites and growth cones (supplementary material Fig. S3A). The immunofluorescence signals were weak compared with non-neuronal cells present in these cultures. We failed to detect αv integrins in hippocampal neuron using different antibodies, fixation protocols and blocking methods. Non-neuronal cells that are also present in these cultures showed a staining pattern typical for focal-adhesion complexes with some of the tested antibodies (data not shown). Thus, αv integrin is not expressed in neurons or present only at low levels, and is not concentrated at sites of adhesion.
To investigate the role of fibronectin receptors including the β1-integrin subunit, we used hippocampal neurons prepared from a conditional knock-out (Itgb1flox/flox) (Potocnik et al., 2000). The cells were cultured on fibronectin and transfected at 3 d.i.v. with an expression vector for EGFP-Cre (pBS505) to inactivate Itgb1 (Fig. 4). After 24 hours, the cells were incubated with Sema3A or control medium for 16 hours. The staining with an antibody specific for β1 integrin showed that the immunofluorescence signal was strongly reduced in cells expressing EGFP-Cre (supplementary material Fig. S3B). The length of dendrites extended by these neurons was significantly shorter (49±2 μm) compared with control neurons (59±1 μm; n=15-25 neurons, three experiments) and the stimulatory effect of Sema3A on dendrite growth was completely blocked (Fig. 4). The length of axons deficient for β1 integrins was similar to that of wild-type neurons (supplementary material Fig. S4). Expression of EGFP-Cre by itself did not affect neuronal differentiation. The length of dendrites formed by wild-type or Ingb+/flox neurons transfected with the vector for EGFP-Cre did not differ from neurons transfected with vector for EGFP (data not shown).
The stimulation of dendrite growth by Sema3A depends on cGMP production
It has been shown previously that activation of the nitric oxide (NO)-dependent guanylate cyclase can switch the repellent activity of Sema3A to an attractive effect (Castellani et al., 2000; Polleux et al., 2000). To test whether NO production and cGMP synthesis are required for the stimulatory effect of Sema3A, we inhibited NO-dependent guanylate cyclase pharmacologically by incubation of rat hippocampal neurons with 1H-[1,2,4]oxadiazolo[4,3-a]quinoxalin-1-one (ODQ). ODQ completely blocked the stimulation of dendrite growth by Sema3A (Fig. 5A,B). The average length of dendrites after treatment with Sema3A for 16 hours was 45±13 μm (n=100 neurons, three experiments). Addition of ODQ together with Sema3A reduced dendrite length to that of control neurons (∼24±6 μm). An inhibitor for NO synthase (L-NAME) also blocked the effect of Sema3A on dendrites. ODQ and L-NAME did not affect the extension of dendrites on PO or fibronectin by themselves. Thus, an increase in the cGMP concentration is required for the stimulation of dendrite extension.
To test whether cGMP is responsible for the modulation of the response to Sema3A by integrins, we directly measured the cGMP concentration after addition of Sema3A. Sema3A induced a tenfold increase in the intracellular cGMP level within 5 minutes that lasted for at least 1 hour (Fig. 5C). After 12 hours of culture in the presence of Sema3A, the cGMP concentration returned to control levels. The increase in cGMP was completely blocked by L-NAME and ODQ. To analyse the time course of cGMP production in more detail, the level of cGMP was measured between 1 minute and 1 hour after the addition of Sema3A (Fig. 5D). The intracellular cGMP concentration increased within 1 minute after application of Sema3A and remained constant for at least 1 hour. However, the induction of cGMP production was independent of integrin engagement. We observed a similar increase in cGMP levels when neurons were cultured on PO or fibronectin. Staining with an antibody specific for cGMP showed that Sema3A stimulated cGMP production in both dendrites and axons (supplementary material Fig. S5). Treatment only with the cell-permeable cGMP analogue 8-bromo-cGMP had no effect on dendrite length, irrespective of the substrate on which neurons were grown (Fig. 5E). cGMP could rescue the block of neurite growth by ODQ, confirming the specificity of the inhibitor. These results show that the production of cGMP is a direct effect of Sema3A and independent of the substrate. cGMP is required for the stimulation of dendrite growth by Sema3A but does not appear to be responsible for mediating the effect of integrins on the Sema3A signalling pathway.
Sema3A stimulates FAK activity
Activated integrins mediate the transmission of signals from the extracellular matrix to the cytoskeleton during cell adhesion, spreading and migration (Giancotti and Ruoslahti, 1999; Mitra et al., 2005). A central component in this pathway is FAK, which is phosphorylated at several tyrosine residues upon recruitment to focal adhesions (Mitra and Schlaepfer, 2006). The axon-guidance signals netrin-1 and ephrin-A1 can also stimulate FAK phosphorylation (Carter et al., 2002; Li et al., 2004; Liu et al., 2004; Ren et al., 2004). To investigate a possible involvement of FAK, Sema3A was added at 3 d.i.v. to mouse hippocampal neurons cultured on fibronectin (Fig. 6). After 20 minutes, cells were lysed and protein phosphorylation investigated with an anti-phospho-tyrosine (anti-pY) antibody. Sema3A increased the phosphorylation of several proteins, including a major band at a molecular weight of 125 kDa (Fig. 6A). Immunoprecipitation of phosphorylated proteins with the anti-pY antibody showed an increase in FAK phosphorylation by more than twofold in western blots with an anti-FAK antibody (Fig. 6A,C). The same result was observed when FAK was immunoprecipitated and phosphorylated proteins detected with an anti-pY antibody.
FAK autophosphorylation at Y397 initiates the binding of SFKs (Mitra et al., 2005). The subsequent phosphorylation of FAK at Y576/577 is required for its full activation. Using antibodies specific for individual phosphorylation sites, FAK phosphorylation can be detected at both Y397 and Y576/577 in neuronal cultures on fibronectin as a substrate (Fig. 6B,C). Incubation with Sema3A increased autophosphorylation at Y397 and phosphorylation at Y576/577 approximately twofold (Fig. 6C) (Y397: increase by a factor of 1.9±0.2; Y576/577: 1.8±0.2; n=3). The increase in FAK phosphorylation at Y397 was also observed by immunofluorescence (supplementary material Fig. S6). To investigate whether the increase in FAK phosphorylation depends on integrin activation, neurons were cultured on PO. Sema3A had no effect on FAK phosphorylation when PO was used as a substrate (Fig. 7). When neurons were cultured on PO in the presence of Mn2+ ions, Sema3A significantly increased phosphorylation at Y397 and Y576/577. FAK phosphorylation at Y576/577 was only slightly increased when the neurons were treated with Mn2+ ions alone. These results show that the stimulation of FAK activity by Sema3A depends on the simultaneous activation of integrins.
Sema3A requires FAK to stimulate dendrite growth
To test whether FAK is required for the extension of dendrites or the stimulation of dendrite growth by Sema3A, hippocampal neurons from Fakflox/flox mice (Beggs et al., 2003) were cultured on fibronectin and transfected with an expression vector for EGFP-Cre (pBS505) at 3 d.i.v. to generate FAK-deficient cells and a vector for mCherry to visualise neuronal morphology (Fig. 8). After 24 hours, Sema3A was added for 16 hours and cells were stained with an anti-MAP2 antibody. The development of dendritic arbours was not reduced in untreated FAK-deficient neurons in comparison to wild-type neurons (control: 58±2 μm; Fak–/–: 54±5 μm). However, the stimulation of dendrite growth by Sema3A was blocked in FAK-deficient neurons (Fig. 8) (control: 73±4 μm; Fak–/–: 51±5 μm). To investigate whether FAK is also required for the collapse of axonal growth cones by Sema3A, the percentage of collapsed axonal growth cones was determined after incubation of FAK-deficient neurons with Sema3A for 1 hour at 3 d.i.v. (Fig. 8C; supplementary material Fig. S7). Sema3A induced the collapse of FAK-deficient hippocampal growth cones to an extent comparable to that of control neurons. It has been previously reported that FAK is required for the collapse of growth cones from cortical axons (Bechara et al., 2008). These results were based on the effect of overexpressing an inactive FAK mutant (FAK-R454). When we expressed this mutant in murine hippocampal neurons, we also observed that FAK-R454 blocked growth-cone collapse by Sema3A (supplementary material Fig. S8). Because neurons from Fak-knockout mice still show a collapse of growth cones after incubation with Sema3A, FAK-R454 does not appear to act by inhibiting endogenous FAK. Thus, FAK is not required for the effect of Sema3A on axonal growth cones but is essential for the stimulation of dendrite growth by Sema3A.
Sema3A acts as a repellent for axons and an attractant or growth-promoting signal for dendrites (Campbell et al., 2001; Fenstermaker et al., 2004; Gu et al., 2003; Polleux et al., 1998; Polleux et al., 2000; Sasaki et al., 2002; Song et al., 1998). However, it has been unclear so far whether both activities of Sema3A impinge on the same pathway, or whether different signalling cascades are involved in its effects on axons and dendrites (Kruger et al., 2005). Here we show that the requirement for integrin engagement distinguishes the response of axons and dendrites to Sema3A in hippocampal neurons. Sema3A induces the collapse of axonal growth cones but stimulates the extension of dendrites. Defects in the formation of dendritic arbours were observed in the hippocampus of Sema3a–/– mice and in cultures of Sema3a–/– neurons. The addition of Sema3A to cultured neurons stimulates dendrite growth. This effect depends on β1 integrins and FAK, which both are dispensable for the collapse of axonal growth cones by Sema3A. These results reveal a bidirectional interaction of integrins and semaphorin signalling. β1 integrins facilitate the effect of Sema3A on dendrites, whereas integrins are inhibited by Sema3A during growth-cone collapse (Kruger et al., 2005; Mikule et al., 2002; Toyofuku et al., 2005) and in endothelial cells (Barberis et al., 2004; Serini et al., 2003). Sema3A stimulates dendrite extension only when neurons are cultured on the integrin ligand fibronectin. Fibronectin could be substituted by Mn2+ ions, which stabilise the active conformation of integrins (Gailit and Ruoslahti, 1988; Mould et al., 1995; Takagi et al., 2002), allowing Sema3A to promote dendrite extension on the non-specific adhesive PO. A soluble RGD peptide that blocks the interaction with integrins and deletion of the Itgb1 gene abrogated the stimulatory effect of Sema3A. By contrast, the collapse of axonal growth cones was independent of integrin engagement. Unlike the effect of cyclic nucleotides on the response to axon-guidance signals (Campbell et al., 2001; Song et al., 1998), β1-integrin engagement does not modulate the response of dendrites because Sema3A had no effect on dendrite length after inactivation of Itgb1 or Fak. Our results show that the effect of Sema3A on dendrite growth requires a signalling pathway that is distinct from that involved in collapse of the growth cone.
Sema3A and β1-integrin signalling converge on FAK to stimulate dendrite growth. Sema3A increases FAK phosphorylation at Y397 and Y576/577 only when integrins are activated simultaneously. Deletion of Fak blocked the stimulation of dendrite growth but did not prevent the collapse of growth cones. The phosphorylation of FAK is mediated by SFKs such as Fyn (Mitra et al., 2005). Sema3A activates SFKs and Cdk5, which could mediate the activation of FAK also in hippocampal neurons (Brown et al., 2004; Mitsui et al., 2002; Morita et al., 2006; Sasaki et al., 2002; Uchida et al., 2005). Fyn is also required for the effect of Sema3A on cortical dendrites and Cdk5 has been shown to phosphorylate FAK in neurons (Morita et al., 2006; Sasaki et al., 2002; Xie et al., 2003). A stimulation of FAK has also been reported for netrin-1 and ephrin-A1 (Carter et al., 2002; Li et al., 2004; Liu et al., 2004; Ren et al., 2004). However, the effect of netrin-1 did not require integrin function.
The adhesion molecule L1 can serve as an alternative signalling receptor for Sema3A (Castellani et al., 2002). Recently, it was reported that Sema3A activates FAK in an L1-dependent manner in cortical neurons during the collapse of axonal growth cones (Bechara et al., 2008). In these experiments, expression of an inactive FAK mutant in cortical neurons blocked growth-cone collapse by Sema3A. Our results show that, although expression of the inactive mutant FAK-R454 blocks growth-cone collapse also in hippocampal neurons, inactivation of the Fak gene by a conditional knockout does not. The difference in the effects of FAK-R454 expression and genetic inactivation of Fak demonstrates that FAK-R454 does not block the function of endogenous FAK. The effect of FAK-R454 might result from sequestering and inhibiting the function of FAK-interacting proteins that have a second, FAK-independent function to mediate growth-cone collapse downstream of the Sema3A receptor.
It has been shown previously that an increase in the concentration of cGMP can convert the repellent effect of Sema3A to an attraction or block the collapse of sensory growth cones (Castellani et al., 2002; Song and Poo, 1999). In addition, an involvement of cGMP synthesis by the guanylyl cyclase Gyc76C was demonstrated genetically in Drosophila, in which Gyc76C facilitates the repulsion of plexin-A-expressing axons by Sema1a (Ayoob et al., 2004). In cultured Xenopus spinal-neuron growth cones, Sema3A triggers cGMP production, which activates Ca2+ channels (Togashi et al., 2008). Our results show that Sema3A induces the production of cGMP in both axons and dendrites from hippocampal neurons, which is necessary but not sufficient for the response of hippocampal dendrites to Sema3A. However, cGMP production does not depend on integrin activation and, unlike Mn2+ ions, cGMP is not able to substitute for integrin activation. Thus, cGMP acts as a second messenger but not as a modulator of the response to Sema3A in dendrites.
In Sema3a–/– mice, apical dendrites branch prematurely in the stratum radiatum and these branches show irregular trajectories. However, we did not observe a significant effect on the orientation of apical dendrites before they branch, suggesting that Sema3A probably does not act as an attractant in the hippocampus. The abnormal organisation of apical dendrites could result from the absence of a repellent signal in the stratum radiatum that restricts branching to the stratum lacunosum-moleculare. However, we did not observe any inhibitory effect of Sema3A on dendritic branching in cultured hippocampal neurons. In addition, Sema3A does not show a layer-specific expression in the hippocampus (Chedotal et al., 1998; Pozas et al., 2001; Skaliora et al., 1998). Therefore, we think that the absence of Sema3A reduces the growth of apical dendrites resulting in a shift in the position of the branching point. The Sema3a–/– phenotype in the hippocampus is similar to that reported for cortical neurons in some aspects (Polleux et al., 1998; Polleux et al., 2000; Sasaki et al., 2002). Apical dendrites are disorientated and deviate from their normal orientation of being perpendicular to the pial surface in the cortex of Sema3a–/– mice (Polleux et al., 2000). However, there is no evidence that Sema3A acts as an attractant signal in the hippocampus.
In summary, our results provide evidence for a new integrin- and FAK-dependent signalling pathway that mediates the stimulation of dendrite growth by Sema3A but is not required for growth-cone collapse.
Materials and Methods
The vectors for dominant-negative Nrp-1, plexin-A1 or plexin-B1 were described previously (Rohm et al., 2000b; Serini et al., 2003). Neurons were transfected with pEGFP-N3 (Clontech) or pmCherry to visualise their morphology. The expression vector for monomeric Cherry (pmCherry) was constructed by introducing BamHI and NotI restriction sites into the sequence of mCherry by PCR using pRSET-b-mCherry (Shaner et al., 2004) as template and cloning the resulting fragment into pEGFP-N3 (Clontech) to replace the coding sequence of EGFP. The vector for EGFP-Cre (pBS505; Addgene plasmid 11955) has been described before (Le et al., 1999).
Mouse strains with floxed alleles of the genes encoding FAK (Ptk2tm1Lfr; MMRRC 009967-UCD) and β1 integrin (Itgb1) have been described previously (Beggs et al., 2003; Potocnik et al., 2000). Sema3a+/– mice were backcrossed to CD1 mice for more than five generations and genotyped as described (Serini et al., 2003; Taniguchi et al., 1997). Although most Sema3a–/– mutants die around birth in the C57/B6 background, a few mutant mice survive for several weeks in the CD1 background. Tg(Thy1-YFPH)2Jrs/J mice (Thy1-YFPH) (Feng et al., 2000) were obtained from the Jackson Laboratory (Bar Harbor, MA) and crossed with Sema3a+/– mice to obtain Thy1-YFPH+/+;Sema3a+/– mice. For the analysis of neuronal morphology, the brains were dissected from P15 Thy1-YFPH+/0;Sema3a–/– mice, fixed overnight in 4% paraformaldehyde/phosphate-buffered saline (PBS), and coronal sections (60 μm) cut with a vibratome (Leica). Genotyping was performed as described (Beggs et al., 2003; Feng et al., 2000; Potocnik et al., 2000; Schwarting et al., 2000).
Neuronal cultures, transfection and immunofluorescence
Cultures of dissociated hippocampal neurons were prepared and transfected as described previously (Schwamborn et al., 2006). Briefly, the hippocampus was dissected from E18 rat or mouse embryos, dissociated, and neurons were plated onto coverslips coated with PO (Sigma) or fibronectin (Roche) at a density of 40,000 cells per coverslip in a 24-well plate and cultured in Neurobasal medium (Invitrogen) supplemented with B27 (Invitrogen). A stably transfected HEK293 cell line (cell line 293-717) was used to produce AP-Sema3AP1b (Zanata et al., 2002). The 293-717 cells were incubated with serum-free medium for 48 hours and the conditioned medium was concentrated with 100 kDa centrifugal filter devices (Millipore). The concentration of Sema3AP1b was measured by an alkaline phosphatase assay with para-nitrophenol (Sigma) as a substrate (Rohm et al., 2000a). Hippocampal neurons were incubated with equal amounts of Sema3AP1b as quantified with the phosphatase assay with a final concentration of approximately 10 nM. Medium conditioned by HEK293T cells was used as control medium. DMSO, ODQ, Nω-Nitro-L-arginine methyl ester (L-NAME), 8-Br-cGMP, MnCl2 and the RGD peptide (all Calbiochem) were added directly to neuronal cultures 15 minutes before conditioned medium containing the inhibitors at the same concentration was added. Neurons were cultured for 16-20 hours in the presence of conditioned medium.
Neurons were transfected at 2 days in culture (2 d.i.v.) using Lipofectamine 2000 (Invitrogen) or by calcium phosphate co-precipitation (Schwamborn et al., 2006). After incubation for 2 hours, the transfection medium was replaced by Neurobasal medium (supplemented with B27, 0.5 mM glutamine and 100 U/ml penicillin/streptomycin; Invitrogen). Neurons were fixed at 4 d.i.v. with 4% paraformaldehyde and 15% sucrose in PBS for 20 minutes at 4°C, permeabilised with 0.1% Triton X-100 in PBS, blocked with 10% FCS or normal goat serum in PBS, and stained with an anti-MAP2 and the Tau-1 antibody (Chemicon) as markers for dendrites and axons, respectively, rhodamine-phalloidin (Molecular Probes), or an anti-α-tubulin antibody (Sigma), and Alexa-Fluor-594-, Alexa-Fluor-488- or Alexa-Fluor-350-conjugated secondary antibodies (Molecular Probes). β1 integrins were detected using a monoclonal anti-β1-integrin antibody (MAB1997, Chemicon).
Images were taken with the Zeiss Axioskop 40 FL equipped with a Spot Insight Camera Mono 15.0 (Diagnostic Instruments). The morphology of neurons was analysed using the WASABI software (Hamamatsu) and Adobe Photoshop. The length of dendrites was determined using the Spot software (Diagnostic Instruments). The Student's t-test was used to test statistical significance.
Measurement of cGMP levels
The intracellular cGMP concentration was measured by the cGMP Direct Biotrak enzyme-linked immunoassay (EIA) system (Amersham) according to the manufacturer's instructions. Briefly, cell lysates were incubated with cGMP-coupled peroxidase and an immobilised anti-cGMP antibody. The concentration of intracellular cGMP was calculated based on the amount of bound cGMP-coupled peroxidase.
In Figs S1-S8 in the supplementary material, neurons were incubated at 3 d.i.v. with medium containing CMFDA (16 μM, Invitrogen) for 30 minutes, fixed with 4% paraformaldehyde and stained with an anti-cGMP antibody (Chemicon). The cGMP immunofluorescence signals for cGMP and CMFDA in the growth cones of dendrites and axons were measured using ImageJ (NIH) and the signal for cGMP normalised to that of CMFDA to determine the relative level of cGMP in arbitrary units (a.u.). To analyze FAK phosphorylation, neurons were stained with an anti-phospho-FAK (Y397) antibody (BD Bioscience).
Immunoprecipitation and western blot
Mouse hippocampal neurons from E18 were cultured on fibronectin (Roche) or PO (Sigma). After 4 days in culture the cells were incubated with Sema3AP1b for 20 minutes. In some cases the neurons were incubated with 1 mM MgCl2 before and during treatment with Sema3A. The neurons were lysed in 50 mM HEPES buffer, pH 7.4, 1% NP-40, 1% Na-deoxycholate, 0.1% SDS, 150 mM NaCl, 1.5 mM MgCl2 and 10% glycerol including protease inhibitor cocktail (complete, Roche) and phosphatase inhibitors (1 mM Na3VO4 and 50 mM NaF). The cleared lysates were diluted 1:1 with 50 mM HEPES, pH 7.4, 1% NP-40, 150 mM NaCl, 1.5 mM MgCl2 and 10% glycerol including the inhibitors, and proteins were immunoprecipitated for 1 hour at 4°C with anti-FAK (BD Bioscience) or anti-phospho-tyrosine (Cell Signaling) antibodies, and analysed by western blot using anti-FAK, anti-phospho-FAK (Y397) (BD Bioscience), anti-phospho-FAK (Y576/577), anti-phospho-tyrosine (Cell Signaling), anti-α-tubulin (Sigma) and horseradish-peroxidase-coupled secondary antibodies (Dianova). Chemiluminescence (Uptilight HRP Blot Chemiluminescent Substrate, Uptima) was detected using the LAS-1000 luminescent image analyser (Fuji). The signal intensity was determined for each band using the Image Gauge 3.12 software (Fujifilm). The background was measured for an area of identical size and subtracted from the value for each band. The corrected signal intensity for phosphorylated FAK was normalised to the signal for total FAK or α-tubulin. The value for the control experiment was set to 1.0 and changes in the level of phosphorylated FAK calculated relative to the value of the control.
We thank Takeshi Yagi and Masahiko Taniguchi for providing us with Sema3a+/– mice, Roger W. Tsien for providing the pREST-mCherry, David D. Schlaepfer for the FAK-R454 construct, and Lena Will for constructing pCherry. The GFP-Cre (pBS505) construct generated by Brian Sauer was obtained through addgene. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB 492-B14) to A.W.P. and a fellowship from the Boehringer Ingelheim Fonds (J.C.S.).