Damage to the sensory neurons of the dorsal root ganglion (DRG) induce major and long lasting changes in the expression of a large number of genes that promote neurite outgrowth and axonal regeneration (for review see Navarro et al. 2007). Under favourable conditions, for instance following a crush injury, most nerve fibres successfully regenerate. However, in many clinically relevant circumstances, reduced or disordered axonal regeneration often results in a loss of sensation and/or the development of chronic neuropathic pain states. The physiological mechanisms that underlie injury-induced axonal regeneration are therefore of considerable scientific and clinical importance.
The principal secreted factors that are involved in the regeneration of peripheral sensory neurons include the neurotrophins, members of the TGFβ superfamily, cytokines and a number of neuropeptides. We have previously demonstrated that the neuropeptide galanin plays a trophic role to the DRG and stimulates neurite outgrowth and regeneration of injured sensory neurons (Holmes et al. 2000; Mahoney et al. 2003). In the adult, galanin is expressed at low levels in < 5% of DRG neurons, which are predominantly small diameter C-fibres (Hokfelt et al. 1987). Following nerve damage, there is a very substantial increase (up to 120-fold) in the levels of galanin in the rodent, primate and human DRG (Hokfelt et al. 1987; Wiesenfeld et al. 1992) and the peptide is then abundantly expressed in ~ 30–40% of sensory neurons (Hokfelt et al. 1994). Following a crush injury to the sciatic nerve regeneration is reduced by 35% in galanin knockout (Gal-KO) mice compared with wildtype (WT) controls (Holmes et al. 2000). This impaired regenerative capacity is paralleled by a similar reduction in neurite outgrowth of dispersed adult mouse DRG neurons (Holmes et al. 2000), which can be rescued by the addition of exogenous galanin (Mahoney et al. 2003). The neuritogenic role played by galanin is mediated via activation of galanin receptor 2 (GalR2) in a protein kinase C (PKC) dependant manner (Hobson et al. 2006).
When an axon is severed it produces a highly motile tip called a growth cone, which senses the surrounding environment and if favourable leads to axonal elongation. This leads to replacement of the distal nerve segment that is lost or damaged, promoting reinnervation of the target organ and thus recovery of function (Allodi et al. 2012). The most distal portion of the growth cone extends both lamellipodia, which are veil like membranous protrusions, and finger like extensions known as filopodia. These protrusions are mainly formed by actin filament bundles which are in equilibrium between actin polymerisation and depolymerisation generating the protrusion force necessary to allow the filopodia to explore the microenvironment. The small GTPases of the Rho subfamily have been shown to be key mediators of the interaction between cell adhesion molecules and the cytoskeleton and are now regarded as major regulators of axonal and dendritic growth (Luo 2000; Auer et al. 2011), integrating upstream signalling cues with downstream cytoskeletal rearrangements.
Rho GTPases are a subfamily of the Ras superfamily of small GTPases which were first found to mediate filopodia and lamellipodial formation in fibroblasts (Ridley et al. 1992; Nobes and Hall 1995). The Rho GTPases act as molecular switches and can exist in two states: an inactive GDP-bound state and an active GTP-bound state. In the active GTP-bound form, the activated GTPase binds its effectors and thus elicits various biological activities. Each of the three major members of the Rho subfamily, Rho, Rac and Cdc42 have been found to play a specific role in axonal and dendritic morphology. RhoA is involved in growth cone retraction in response to collapsing guidance cues (Jalink et al. 1994; Kozma et al. 1997; Thies and Davenport 2003), whilst Rac and Cdc42 promote neurite outgrowth by formation of lamellipodia and filopodia respectively (Ridley et al. 1992; Nobes and Hall 1995). Rac and Cdc42 share a common effector: p21-activated kinase (Pak) (Manser et al. 1994; Burbelo et al. 1995), which may explain some of the common phenotypes seen when the activities of Rac or Cdc42 are perturbed (Manser et al. 1994). Pak phosphorylates LIM kinase (LIMK) (Edwards et al. 1999), which in turn phosphorylates and inhibits the actin-binding protein cofilin. Cofilin can also be phosphorylated by the Rho pathway via Rho-associated coiled-coil-containing kinase (ROCK)1 and ROCK2 phosphorylation of LIMK2 (Sumi et al. 1999).
Cofilin is the principal terminal effector of a number of signalling cascades (that includes many of the superfamily of small GTPases), and its activation leads to cytoskeletal rearrangement. It is expressed at high levels in the growth cones of neurons, colocalises with filamentous (F) actin (Bamburg and Bray 1987) and over-expression leads to increased neurite outgrowth (Meberg and Bamburg 2000). Cofilin is an important regulator of actin dynamics and acts by increasing actin depolymerisation and severing filaments (Chen et al. 2000). Whether cofilin promotes the assembly or disassembly of actin depends on the concentration of cofilin relative to actin as well as the concentration of other binding proteins (Van et al. 2008). The major mechanism by which cofilin activity is regulated is via phosphorylation at Serine 3 (Ser3) which inhibits its binding to actin monomers and thus reduces its actin depolymerising activity (Morgan et al. 1993; Agnew et al. 1995).
In this study, we have characterised an intracellular signalling pathway by which galanin stimulates neurite outgrowth from sensory neurons which has important therapeutic implications for the treatment of peripheral sensory neuropathies. We show here that galanin reduces the activation state of both Rho and Cdc42 GTPases, but does not affect the activation state of Rac. The reduction in the GTP-bound form of these small GTPases leads to a marked increase in the activation of the actin-binding protein cofilin with associated changes in growth cone dynamics which in turn promotes neuritogenesis.
- Top of page
- Materials and methods
- Supporting Information
We have previously shown that the impaired regenerative capacity observed in Gal-KO mice is paralleled by a 35% reduction in neurite length, compared with WT controls (Holmes et al. 2000). The deficits in neurite outgrowth can be rescued by the addition of either exogenous galanin or the GalR2/3 specific agonist Gal2-11 (Mahoney et al. 2003). Our finding that galanin plays a neuritogenic role is supported by work in rat dispersed DRG cultures. Galanin significantly increased both axonal elongation and also the number of branch points (Suarez et al. 2006). Furthermore, treatment of foetal DRG cultures with galanin increased growth cone velocity by nearly twofold compared with controls (Sanford et al. 2008). To confirm which galanin receptor subtypes mediate the effects of galanin, we studied GalR2-KO animals, demonstrating a marked reduction in neurite outgrowth identical to that observed in the Gal-KO animals (Hobson et al. 2006). The addition of galanin (which also activates GalR1 and GalR3) failed to rescue the deficits in the GalR2-KO cultures (Hobson et al. 2006) implying that neither GalR1 nor GalR3 play a major role in galanin-mediated neuritogenesis of adult sensory neurons. All three galanin receptor subtypes couple to Gi/o leading to an inhibition of adenylate cyclase (Habert-Ortoli et al. 1994; Smith et al. 1998). GalR2 additionally signals via Gq/11 to activate phospholipase C and protein kinase C (PKC) (Habert-Ortoli et al. 1994; Wang et al. 1998; Wittau et al. 2000). Consistent with galanin modulation of neurite outgrowth via GalR2 in DRG neurons, the neuropeptide has also been shown to stimulate neurite outgrowth in a PKC dependant manner from PC12 cells and neurospheres derived from adult hippocampal progenitor cells (Mahoney et al. 2003; Hawes et al. 2006; Hobson et al. 2006). The intracellular signalling pathways which mediate the galanin/GalR2-dependent increase in neurite outgrowth in sensory neurons have yet to be elucidated and are the subject of this study.
Here, we demonstrate that galanin decreases activated levels of Rho (Fig. 1c) and Gal-KO animals have significantly higher levels of GTP-Rho than WT controls (Fig. 2b). Rho has been implicated in regulating growth cone collapse and its activation leads to neurite retraction (Jalink et al. 1994; Kozma et al. 1997; Leclere et al. 1997; Lehmann et al. 1999). This effect is thought to be mediated via ROCK1 and ROCK2 phosphorylation of LIMK2 which phosphorylates cofilin, inhibiting its activity (Agnew et al. 1995; Arber et al. 1998; Maekawa et al. 1999; Gehler et al. 2004b). Thus, the finding that Gal-KO animals have higher levels of active Rho than their WT controls may explain in part the deficits seen in regeneration in vivo and neurite outgrowth in vitro (Holmes et al. 2000).
Our finding that galanin also regulates activated levels of Cdc42 (Figs 1b and 2a) is in accordance with published work demonstrating that Cdc42 regulates filopodial formation in response to extracellular cues in non-neuronal cells (Nobes and Hall 1995) and neuroblastoma cells (Kozma et al. 1997). The currently accepted Cdc42 signalling pathway in neurons is via activation of Pak and LIMK which increase the levels of inactive phospho(Ser3)cofilin (Agnew et al. 1995; Arber et al. 1998; Edwards et al. 1999; Maekawa et al. 1999; Govek et al. 2005). Here, we show that galanin decreases the activated levels of Cdc42 (Fig. 1b) and that Gal-KO animals have significantly higher levels of GTP-Cdc42 than WT controls (Fig. 2a). Consistent with these findings, two recent studies using cultured cortical and hippocampal neurons have demonstrated that inhibition of Cdc42 signalling is critical to dendritic branching and neurite outgrowth via activation of cofilin (Peris et al. 2012; Rosario et al. 2012). Of note these findings are contrary to previous published data showing over-expression of constitutively active Cdc42 (CAcdc42) constructs increases filopodial length and number in retinal ganglion cells (Chen et al. 2006) and the number of filopodia and increased rate of neurite outgrowth in chick spinal cord neurons (Brown et al. 2000). However, some studies have demonstrated that Cdc42 (like other GTPases) may need to complete a full GTP-binding/GTP hydrolysis cycle to propagate signals (Luo 2000; Fidyk et al. 2006), whereas CaCdc42 is GTP hydrolysis defective. Results using dominant negative constructs (DNCdc42) have been less clear with some studies showing DNCdc42 decreased neurite outgrowth (Threadgill et al. 1997) whilst others showed no significant effect on neurite outgrowth (Brown et al. 2000). DNGTPase constructs act by sequestering multiple guanine nucleotide exchange factors (GEFs), but most of these GEFs are not specific for a single RhoGTPase, and thus the DN mutant affects other RhoGTPase pathways (Czuchra et al. 2005; Pertz et al. 2008). Furthermore, knockdown of Cdc42 in neuroblastoma cells by siRNA rather than DNconstruct actually led to a significant increase in neurite length (Pertz et al. 2008). Thus, whilst CA and DN constructs have been important in the analysis of RhoGTPase functions in cells, data should be interpreted with caution and it is important to analyse the endogenous RhoGTPase activation states.
Irrespective of the pathways by which Rac, Rho and Cdc42 GTPases modulate cofilin activation, there is good agreement in the literature that it plays a key role in the control of actin dynamics and neurite outgrowth (Meberg and Bamburg 2000). The cofilin family is ubiquitously expressed in eukaryotes and the activity of all vertebrate cofilin proteins is regulated by phosphorylation. Phosphorylation at Ser3 inhibits the binding of cofilin to actin monomers and its actin depolymerising activity (Morgan et al. 1993; Agnew et al. 1995). Expression of cofilin is abundant within neuronal growth cones (Bamburg and Bray 1987) and signals which alter growth cone motility have been shown to alter phosphorylation of cofilin (Bamburg and Bray 1987; Meberg et al. 1998). Actin monomer is abundantly available in DRG growth cones and polymerisation occurs rapidly when free F-actin barbed ends become available for monomer addition by cofilin severing. Thus, activated (unphosphorylated) cofilin positively correlates with neurite outgrowth (Meberg et al. 1998; Kuhn et al. 2000; Meberg and Bamburg 2000). We show here that following treatment with galanin the levels of phospho(Ser3)cofilin fall in DRG and PC12 cells indicating that galanin activates cofilin (Fig. 3b and c). These data are also in keeping with the finding that treatment with other factors which are trophic to neurons such as nerve growth factor and brain-derived neurotrophic factor (BDNF) also decrease cofilin phosphorylation and promote neurite outgrowth (Gehler et al. 2004b; Chen et al. 2006; Endo et al. 2007).
Since galanin signals through the Rho, Cdc42 and cofilin pathways, we then asked whether Gal-KO animals have altered growth cone dynamics compared to WT controls. Gal-KO sensory growth cones extended and retracted their filopodia at a significantly reduced rate compared with WT filopodia (Fig. 4a and b), consistent with a previous study which showed an increased velocity of rat DRG growth cones in the presence of galanin (Sanford et al. 2008). The neurotrophin BDNF has also been shown to increase filopodial extension and retraction rates by reducing activation of RhoA (Gehler et al. 2004a) and activating cofilin (Gehler et al. 2004b). Most recently, knockout of RhoE (a member of the atypical family of Rho proteins) has been shown to increase RhoA and Cdc42 activation in hippocampal neurons, associated with an decrease in cofilin activation and a reduced rate of filopodial extension and retraction (Peris et al. 2012). In contrast to these data, expression of CACdc42, albeit in drosophila, increased extension and retraction rates by 52% (Kim et al. 2002) and Cdc42-KO mice display an inhibition in the amount of filopodia/min formed and retracted (Garvalov et al. 2007). Both these studies are in embryonic animals and the apparent discordance in findings could be as a result of differing roles played by Cdc42 in early as compared with the later stages of development, as suggested by Rosario et al. (2012).
The increase in activation of Rho, Cdc42 and inactive p(Ser3)cofilin in Gal-KO mice and the resulting reduction in the rates of filopodial extension and retraction may explain in part, the observed increase in growth cone area compared with WT controls (Fig. 4a). Of note, growth cone area was also larger in Cdc42-KO embryonic hippocampal neurons than WT controls (Garvalov et al. 2007). Whilst this may seem contradictory with our data at the level of Cdc42, it is consistent with the unexpected increased levels of phospho-LIMK and p(Ser3)cofilin observed in these mice (Garvalov et al. 2007).
In summary, we have characterised an intracellular pathway by which galanin stimulates neurite outgrowth in sensory neurons, demonstrating a role for the neuropeptide in growth cone dynamics and morphology via modulation of the activation states of Rho and Cdc42 GTPases and the actin-binding protein cofilin. These findings have important therapeutic implications for the treatment of peripheral sensory neuropathies.