Axotomy-induced HIF-serotonin signalling axis promotes axon regeneration in C. elegans

The molecular mechanisms underlying the ability of axons to regenerate after injury remain poorly understood. Here we show that in Caenorhabditis elegans, axotomy induces ectopic expression of serotonin (5-HT) in axotomized non-serotonergic neurons via HIF-1, a hypoxia-inducible transcription factor, and that 5-HT subsequently promotes axon regeneration by autocrine signalling through the SER-7 5-HT receptor. Furthermore, we identify the rhgf-1 and rga-5 genes, encoding homologues of RhoGEF and RhoGAP, respectively, as regulators of axon regeneration. We demonstrate that one pathway initiated by SER-7 acts upstream of the C. elegans RhoA homolog RHO-1 in neuron regeneration, which functions via G12α and RHGF-1. In this pathway, RHO-1 inhibits diacylglycerol kinase, resulting in an increase in diacylglycerol. SER-7 also promotes axon regeneration by activating the cyclic AMP (cAMP) signalling pathway. Thus, HIF-1-mediated activation of 5-HT signalling promotes axon regeneration by activating both the RhoA and cAMP pathways.

A n axon's ability to regenerate after injury depends on both its intrinsic growth capacity and the external environment. Most invertebrate neurons, as well as neurons in the mammalian peripheral nervous system are able to regenerate. Although intrinsic regeneration signals can influence regenerative success, the variables determining a neuron's intrinsic regrowth capacity remain poorly defined at the molecular level. The nematode Caenorhabditis elegans has recently emerged as a tractable genetic model for studying neuron regenerative processes 1,2 . Two recent screens have identified candidate genes that may be involved in axon regeneration: Chen et al. 3 screened a collection of mutants of C. elegans, and Nix et al. 4 screened using an RNA interference (RNAi) library. The candidate genes identified include those encoding both growth-promoting and growth-inhibiting factors. Several of these genes have been implicated previously in regeneration and others define new and conserved pathways of interest.
In the present study, we investigate the roles of the tph-1 and rga-5 genes in the regulation of axon regeneration. The tph-1 and rga-5 genes were identified as candidate genes in a screen of mutants and by an RNAi screen, respectively 3,4 . The tph-1 gene encodes an enzyme involved in 5-HT biosynthesis 5 , while the rga-5 gene encodes a homolog of the mammalian p190 Rho GTPase-activating protein. We show that the hypoxia-inducible transcription factor (HIF-1) induces expression of the tph-1 gene in axotomized neurons, which are not serotonergic. 5-HT subsequently promotes axon regeneration by signalling through the serotonin receptor SER-7. We further demonstrate that the RhoA signalling pathway is important for neuronal regeneration via its ability to inhibit diacylglycerol (DAG) kinase, thus increasing DAG concentrations. These two effects are linked, as C. elegans RhoA is activated by SER-7 and by G12a through RhoGEF encoded by the rhgf-1 gene. SER-7 is also required for axon regeneration via its ability to activate the cyclic AMP (cAMP) signalling pathway. Thus, HIF-1mediated activation of 5-HT synthesis and consequent autocrine signalling promotes the neuron regenerative response by activating two separate signalling pathways.

5-HT is required for axon regeneration.
It has been suggested that axon injury may represent a type of cellular stress. In response to specific types of stress, neurotransmitters are deployed to effect the appropriate physiological and behavioural modifications [6][7][8] . In C. elegans, 5-HT modulates a variety of stress responses 9,10 . However, little is known about how specific types of stress induce 5-HT-mediated signal transduction in the nervous system. Interestingly, Chen et al. 3 recently carried out a systematic screen of mutants looking for defects in axon regeneration and identified tph-1 among several genes that positively regulate axon regeneration in a glutaminergic touch sensory neuron 3 . The tph-1 gene encodes tryptophan hydroxylase 5 , which is required for 5-HT biosynthesis (Fig. 1a). We confirmed that a loss-of-function mutation (mg280) of tph-1 caused a defect in axon regeneration in touch sensory posterior lateral microtubule (PLM) neurons (Fig. 1b,c and Supplementary  Table 1). Furthermore, the mg280 and n4622 mutations of tph-1 were also defective in axon regeneration in GABA-releasing D-type motor neurons ( Fig. 1d and Supplementary Table 2). Exogenous 5-HT treatment restored axon regeneration in tph-1 mutants. These results suggest that 5-HT is generally required during axon regeneration.
We next searched for the 5-HT-producing neurons essential for axon regeneration. To match serotonin levels to the normal range for each neuron, we used a system recently developed by Flavell et al. 11 A single floxed copy of the tph-1 gene, including the promoter, exons and introns, was inserted into a defined location on chromosome IV (Fig. 1e). This inserted gene was able to rescue the tph-1 defect in axon regeneration of D-type motor neurons (Fig. 1f and Supplementary Table 2). This was combined with cell-specific transgenes expressing Cre recombinase in amphid sensory neuron, dual cilia, designation F. (ADF), neurosecretory motor neuron (NSM) or hermaphrodite specific neuron (HSN) neurons, which are known to be 5-HT-producing neurons 5 , or in D-type neurons. This system allowed the excision of the inserted tph-1 gene and elimination of 5-HT production in these individual neurons. We found that the tph-1 defect was still rescued when Cre was expressed in ADF, NSM or HSN neurons ( Fig. 1f and Supplementary Table 2). In contrast, the tph-1 defect was not rescued when Cre was expressed in D-type motor neurons. Furthermore, expression of tph-1 in ADF neurons by the srh-142 promoter was unable to rescue the tph-1 defect ( Fig. 1d and Supplementary Table 2). These results suggest that axon regeneration in D-type motor neurons requires 5-HT production by D neurons. To confirm this, we ectopically expressed tph-1 in D-type motor neurons in a tph-1(mg280) background using the unc-25 promoter. We found that expression of tph-1 in D neurons did indeed suppress the defect in axon regeneration of tph-1 mutants ( Fig. 1d and Supplementary Table 2). These results demonstrate that TPH-1 acts to promote regeneration of the damaged neuron in a cell-autonomous manner.
HIF-1 upregulates 5-HT expression in injured neurons. The identification of D-type neurons as the 5-HT-producing neurons essential for axon regeneration was unexpected, as D neurons are known to be GABAergic and the expression of 5-HT synthesis enzymes in these neurons has not been found 5,[12][13][14] . We therefore addressed the possibility that axon severing might induce transcription of the tph-1 gene in D-type neurons. To test this, we monitored expression of the Ptph-1::gfp transgene, which consists of the tph-1 promoter driving the fluorescent protein green fluorescent protein (GFP). We found that expression of tph-1 was altered in D-type neurons by axon injury. In control animals, Ptph-1::gfp was robustly expressed in the head of animals in the NSM and ADF neurons ( Supplementary Fig. 1), and was not expressed in D neurons. However, after laser surgery of D-type neurons, Ptph-1::gfp expression was strongly upregulated in these neurons ( Fig. 2a,b). This expression became visible at 30 min after surgery, and returned to undetectable levels by 24 h. Thus, axon injury induces the transient expression of Ptph-1::gfp in severed D-type neurons. 5-HT immunofluorescence staining of animals treated with laser surgery also revealed induction of 5-HT expression, compared with control animals ( Supplementary Fig. 2). Expression of Punc-47::mcherry, which reflects the expression of vesicular GABA transporter, was constant over this time period (Fig. 2a). Therefore, the severed GABAergic neurons still keep the GABAergic characteristics, and do not transdifferentiate into serotonergic, non-GABAergic neurons. Furthermore, axon severing also induced 5-HT and Ptph-1::gfp expression in PLM neurons ( Supplementary Figs 2 and 3), which are glutamatergic, non-serotonergic touch sensory neurons 15 . Our results demonstrate that axotomized 'non-serotonergic' neurons temporarily gain the ability to synthesize serotonin in response to axon injury.
What determines expression of the ectopic tph-1 gene in response to axon injury? Interestingly, the promoter region of the tph-1 gene contains hypoxia response elements, which are binding sites for HIF-1 (ref. 16). The HIF-1 protein is a transcription factor containing a bHLH-PAS domain that mediates transcriptional responses to hypoxic stress 17 . Indeed, it has been demonstrated that HIF-1 binds to the tph-1 promoter via hypoxia response elements to drive serotonin production under hypoxic conditions in C. elegans 16 . We therefore examined whether HIF-1 is involved in tph-1 expression in response to axon injury. The hif-1(ia4) mutation had no effect on expression of Ptph-1::gfp in the head of animals (Supplementary Fig. 1) but decreased its induction in response to laser surgery in D-type neurons (Fig. 2a,b), indicating that induction is dependent on HIF-1 function. In addition, expression of Ptph-1::gfp in PLM neurons following laser surgery decreased in hif-1 mutants ( Supplementary Fig. 3). Thus, HIF-1 is required for induction of 5-HT expression in injured neurons in response to axon injury.
These results raised the possibility that hypoxic stress might enhance axon regeneration. In normoxic conditions, HIF-1 is degraded through the E3 ubiquitin ligase VHL-1 (von Hippel-Lindau tumour suppressor protein)-dependent degradation pathway 18 . Degradation is initiated through hydroxylation of a specific proline residue in HIF-1 by the oxygen-dependent prolyl 4-hydroxylase EGL-9 (ref. 18). VHL-1 recognizes the hydroxylated proline and targets HIF-1 for ubiquitin-mediated proteasomal degradation 19 . Hypoxic conditions inhibit the hydroxylation of HIF-1, stabilizing the protein and allowing it to transcriptionally regulate expression of its target genes 19,20 . Thus, stabilization of HIF-1 through removal of the HIF-1 degradatory pathway phenocopies the effects of hypoxia seen in neurons. We examined the effect of HIF-1 stabilization on tph-1 expression in D-type motor neurons. For this purpose, we used vhl-1(ok161) mutants and a mutated form of HIF-1, HIF-1(P621A), in which the proline in the conserved LXXLAP motif required for degradation of HIF-1 has been altered to an alanine 18,21 . As observed previously 16 Fig. 4). These results suggest that the mechanism that activates HIF-1 in response to axon injury is different from the HIF-1-stabilizing mechanism that is involved in the response to hypoxia.
Next, we examined whether HIF-1 is required for 5-HT-driven axon regeneration. We found that the frequency of axon regeneration in D neurons was reduced significantly, in hif-1(ia4) and hif-1(ok2564) mutants, and was restored by exogenous 5-HT treatment ( Fig. 2c and Supplementary Table 2). This is consistent with the possibility that HIF-1 regulates axon regeneration by driving 5-HT production. Furthermore, the hif-1(ia4) mutation was also defective in axon regeneration in PLM neurons following laser surgery ( Fig. 2d and Supplementary Table 1). These results suggest that HIF-1 is generally required during axon regeneration.
The biological effects of 5-HT are mediated through its interactions with G protein-coupled receptors 22 . Since SER-7 is one of the 5-HT receptors, we examined whether SER-7 participates in axon regeneration. We found that two mutations (tm1325 and tm1548) in the ser-7 gene caused defects in axon regeneration ( Fig. 3a and Supplementary Table 2). In contrast to ser-7, the ok345 null mutation of ser-1 encoding another serotonin receptor had no effect on axon regeneration. Thus, SER-7 plays a specific role in axon regeneration. To test whether SER-7 can act in a cell-autonomous manner, we expressed ser-7 from the unc-25 or mec-7 promoters in ser-7(tm1325) mutants.
The ser-7 defect in axon regeneration was rescued by expression of ser-7 in D-type motor neurons by the unc-25 promoter but not by expression in sensory neurons by the mec-7 promoter (Fig. 3a  and Supplementary Table 2). These results demonstrate that SER-7 acts to promote regeneration in a cell-autonomous manner in the damaged neuron. Exogenous 5-HT treatment was unable to efficiently suppress the ser-7 defect ( Fig. 3a and Supplementary Table 2). What functions downstream of SER-7 to regulate axon regeneration? SER-7 is coupled to GPA-12, a C. elegans homologue of the a subunit of the heterotrimeric G12 protein (Fig. 3b) 23 . We found that the gpa-12(pk322) and gpa-12(gk766855) mutations significantly reduced axon regeneration after axon injury (Fig. 3a and Supplementary  Table 2). To further address whether ser-7 and gpa-12 function in the same pathway, we constructed ser-7(tm1325); gpa-12(pk322) double mutants. We found that the gpa-12 mutation did not enhance the axon regeneration defect of ser-7 mutants (Fig. 3a and Supplementary Table 2), suggesting that SER-7 and GPA-12 function in the same pathway.
In mammals, the RhoGEF LARG protein specifically interacts with G12a via its regulator of G protein signalling (RGS) domain 24 . C. elegans RHGF-1, a homolog of mammalian LARG (Fig. 3c), binds to GPA-12 and acts as a regulator of neurotransmitter release at a subset of cholinergic motor synapses 25,26 . These observations raised the possibility that RHGF-1 may act downstream of GPA-12 in the axon regeneration pathway (Fig. 3b). Indeed, we found that the rhgf-1(ok880) and rhgf-1(gk217) mutants were defective in axon regeneration ( Fig. 3d and Supplementary Table 2). Taken together, these results suggest that SER-7 activates axon regeneration through GPA-12 and RHGF-1.
RHO-1 activates axon regeneration by inhibiting DGK-1. The activity of Rho proteins is modulated by their cycling between the active GTP-and inactive GDP-bound states. Rho activity is further regulated by other proteins, including GAPs, which enhance GTP hydrolysis and the consequent inactivation of Rho, and guanine nucleotide exchange factors (GEFs), which promote the binding of GTP and, thus, activation of Rho 27 . RHO-1 also cycles between the GTP-bound active and the GDP-bound inactive forms. To confirm that RHGF-1 acts as a GEF for RHO-1 and to test if it thereby controls axon regeneration, we expressed a mutant RHO-1 that is locked in the GTP-bound state, RHO-1(G14V), in D-type motor neurons using the unc-25 promoter. We found that expression of RHO-1(G14V) was able to suppress the rhgf-1 defect in axon regeneration, whereas a GDP-bound RHO-1(T19N) expressed from the unc-25 promoter was not ( Fig. 3d and Supplementary Table 2).
What functions as a RhoGAP in the regulation of axon regeneration? One of the genes identified in our previous RNAi screen in b-spectrin mutant animals (unc-70 ; ref. 28) was rga-5, the product of which belongs to the Rho GTPase-activating protein (GAP) family 4 . RGA-5 has an N-terminal GTP binding domain (GBD) and a C-terminal GAP domain. RGA-5 displays strong homology across its GBD and GAP domains to the mammalian p190 RhoGAP (Fig. 4a and Supplementary Fig. 5). Since the GAP domain in p190 RhoGAP is specific for the Rho family of GTPases, particularly RhoA (ref. 29), we expected that RGA-5 would specifically act on RHO-1. To test this prediction, we performed yeast two-hybrid binding assays to test for interactions between RGA-5 and two RHO-1 variants that are locked in the GTP-and GDP-bound states (G14V and T19N, respectively). We found that the RGA-5 GAP domain (1118-1341 amino acids) specifically interacted with the GTP bound, but not GDP-bound version of RHO-1 (Fig. 4b), consistent with the property of GAP proteins to associate with the GTP-bound version of GTPases. The RGA-5 GAP domain failed to bind to GTP-bound CDC-42 or CED-10, two other Rho/Rac family members, demonstrating that the RGA-5-RHO-1 interaction is specific.
Next, we determined the function of RGA-5 in axon regeneration. First, we examined the rga-5(ok2241) mutation, which deletes the GAP domain in RGA-5 (Fig. 4a). After laser surgery, axons in rga-5(ok2241) mutant animals regenerated normally. However, overexpression of the RGA-5 GAP domain reduced regeneration (Fig. 4c,d and Supplementary Table 2). It has been confirmed that the GAP domain of p190 RhoGAP possesses GAP activity and that the conserved Lys-1311 (Lys-1223 in RGA-5) residue is essential for this activity ( Supplementary Fig. 5b) 29,30 . Overexpression of the mutant GAP(K1223A), in which the Lys-1223 in the RGA-5 GAP domain was mutated to alanine, did not cause reduced regeneration (Fig. 4d and Supplementary Table 2). These results suggest that the RGA-5 GAP domain stimulates the conversion of RHO-1 to the GDP-bound inactive form, resulting in the inhibition of axon regeneration. In this manner, RGA-5 acts as a potent inhibitor of axon regeneration. We further examined possible genetic interactions between rga-5 and rhgf-1, and found that the rga-5(ok2241) mutation was able to suppress the rhgf-1(ok880) defect ( Fig. 4d and Supplementary Table 2). This can be explained because RHO-1 can still be weakly activated in the absence of its GEF, thus concomitant loss of GAP affects the phenotype.
How does RHO-1 regulate axon regeneration? Mammalian RhoA has been shown to bind and inhibit the DAG kinase DGKy (refs 31,32). DAG kinases remove DAG by phosphorylating the membrane-bound secondary messenger DAG, converting it to phosphatidic acid 33,34 , which implicates RhoA in regulating DAG levels 31 . In addition, C. elegans RHO-1 is able to bind to DGK-1, the C. elegans ortholog of DGKy (ref. 35). Thus, DGK-1 is a possible target of negative regulation by RHO-1 (Fig. 3b). If RHO-1 modulates axon regeneration through the inhibition of DGK-1, then we would predict that a dgk-1 loss-of-function mutation should suppress the regeneration defect by inactivation of RHO-1. As expected, we found that the dgk-1(ok1462) null mutation suppressed the defect caused by the rhgf-1(ok880) deletion mutation or overexpression of the RGA-5 GAP domain ( Fig. 4d and Supplementary Table 2). Our results are most consistent with a signal transduction cascade involving activation of RHO-1, inhibition of DGK-1, and a consequent increase in DAG (Fig. 3b).

SER-7 promotes axon regeneration via the cAMP pathway.
If the SER-7-GPA-12 pathway acts upstream of RHO-1 to promote axon regeneration, an rga-5 or dgk-1 mutation should suppress the defect in axon regeneration seen in ser-7 and gpa-12 mutants. As expected, the rga-5(ok2241) and dgk-1(ok1462) mutations were each able to suppress the gpa-12 defect (Fig. 5a and Supplementary Table 2). However, in contrast to the gpa-12 mutation, the ser-7 defect was not suppressed by the rga-5(ok2241) or dgk-1(ok1462) mutation ( Fig. 5a and Supplementary Table 2). These results suggest that another regulatory pathway exists downstream of SER-7, in addition to the GPA-12-RHO-1 pathway, that is required for axon regeneration.
Since the SER-7 5-HT receptor has also been reported to be coupled to GSA-1 Gsa (Fig. 5b) 23,36 , we hypothesized that an additional pathway of SER-7-mediated axon regeneration could involve the GSA-1 signalling pathway. Since Gsa is coupled to adenylyl cyclase, Gsa signalling can be monitored by intracellular cAMP levels. As observed previously 37 , axon regeneration was inhibited by overexpression of pde-4, which encodes a cAMPphosphodiesterase and reduces intracellular cAMP levels ( Fig. 5c and Supplementary Table 2). Furthermore, a loss-of-function mutation of acy-1, which encodes adenylyl cyclase, was defective in axon regeneration ( Fig. 5c and Supplementary Table 2). These results suggest that activation of GSA-1 signalling is necessary to activate axon regeneration. To confirm that ser-7 and acy-1 function in the same pathway, we constructed ser-7(tm1325); acy-1(nu329) double mutants. The ser-7 mutation did not enhance the axon regeneration defect of acy-1 mutants (Fig. 5c and Supplementary Table 2), supporting the possibility that SER-7 functions in the GSA-1 signalling pathway to activate axon regeneration.
Since SER-7 activates both GPA-12 and GSA-1 signalling pathways (Fig. 5b) 23 , we speculated that activation of both pathways are required for axon regeneration. If so, activation of the GSA-1 signalling pathway should suppress the ser-7 defect in axon regeneration in a manner dependent on the activation of the GPA-12 signalling pathway. Treatment of animals with forskolin is expected to cause an increase in cAMP levels by activating adenylyl cyclase 37,38 . We found that forskolin treatment failed to suppress the single mutant ser-7(tm1325) defect in axon regeneration ( Fig. 5a and Supplementary Table 2). However, forskolin treatment was able to suppress the axon regeneration defect in ser-7; rga-5 double mutants, in which the GPA-12 signalling pathway is activated (Fig. 5a and Supplementary Table 2). These results suggest that SER-7 activates GPA-12 and GSA-1, which act in concert to activate axon regeneration.

Discussion
The neurotransmitter 5-HT plays multiple roles in the enteric, peripheral and central nervous system , and its most prominent biological function is as a messenger transmitting signals from pre-to post-synaptic neurons. However, 5-HT has been reported to play other roles, such as the regulation of neurite growth 39,40 , suggesting that 5-HT also plays a prominent role in regulating the neuronal cytoarchitecture 41,42 . In C. elegans, 5-HT modulates cognitive behaviours such as chemosensation, learning and memory in addition to a variety of stress responses 10,13,16 . Here our results demonstrate that C. elegans 5-HT is synthesized in axotomized 'non-serotonergic' neurons and is required for axon regeneration. This effect is likely due to a direct action of 5-HT on ARTICLE injured neurons through the SER-7 5-HT receptor, which activates axon regeneration in a cell-autonomous manner. Neurons are known to be able to produce distinct sets of neurotransmitters 43 , and some studies have shown that certain stimuli can induce neurotransmitter switching 44,45 . In mammals, leukaemia inhibitory factor and ciliary neurotrophic factor have been shown to induce the differentiation of developing peripheral sympathetic neurons, which normally release noradrenaline, into cholinergic neurons 46,47 . Another study revealed that the balance of the two hippocampal granule cell neurotransmitters, glutamate and GABA, changes during post-embryonic development, as well as with seizures and long-term potentiation 48 . Recently, Dulcis et al. 49 reported that the neurotransmitters produced by hypothalamic neurons switch between dopamine and somatostatin in response to photoperiod stimuli. Although these studies were largely carried out using populations of neuronal cells that could include different types of cells such as reserve pool neurons, they indicate that neurotransmitter production varies dynamically in neurons. Our findings in C. elegans demonstrate that a single, differentiated GABAergic D-type motor neuron or glutamatergic PLM sensory neuron that is never serotonergic under standard laboratory conditions can be induced to produce 5-HT as a result of axotomy in a manner dependent on HIF-1. This represents a novel case in which a switch in neurotransmitter expression can be regulated. Hypoxia also stimulates expression of 5-HT via HIF-1 in glutamatergic ASG neurons 16,50 . However, stabilization of HIF-1, which phenocopies the hypoxia-induced neuronal effects, is sufficient to induce tph-1 expression in ASG neurons, but not in D or PLM neurons. Therefore, HIF-1-dependent tph-1 expression in severed neurons is regulated by a mechanism different from the HIF-1stabilizing mechanism.
In mammals, the expression and stability of HIF-1, as well as the expression of its target genes, are all increased after spinal cord injury 51,52 . Thus, activation of HIF-1 by axotomy may be conserved among animals. Furthermore, the HIF-induced upregulation of 5-HT in specific regions of the vertebrate brain 53 suggests that the mechanisms coupling axon regeneration and 5-HT signalling to produce an alteration in neuronal responses may be conserved among animals. Therefore, it would be interesting to know whether axotomy of vertebrate neurons also stimulates expression of 5-HT or other neurotransmitters that influence regeneration. The inability of axons to regenerate in the adult mammalian central nervous system is a major problem in a number of diseases, and understanding the multi-functional roles of 5-HT signalling on neurite growth may lead to the development of novel neuroregenerative therapies by targeting serotonin-regulated signalling pathways.
The biological effects of 5-HT are mediated through interactions with G protein-coupled receptors 22 . Here we have shown that the activated SER-7 5-HT receptor promotes axon regeneration in injured neurons by activating G12a and Gsa signalling pathways in a cell-autonomous manner. How do G12a and Gsa signalling cascades activate axon regeneration? Several studies suggested that DAG and cAMP act downstream of the G12a and Gsa cascades, respectively [54][55][56][57] . In mammals, G12a activates an RGS-containing RhoGEF, which in turn activates RhoA. Several lines of evidence suggest that this pathway also acts within C. elegans motor neurons 26 . The axon regeneration pathway is regulated by both of these components acting in the G12a pathway, as either a mutation in rhgf-1 encoding the RGS-containing RhoGEF or inactivation of RHO-1 by RhoGAP RGA-5 inhibits axon regeneration. In addition, RHO-1 activation is able to suppress the defect in axon regeneration caused by the gpa-12 mutation. Thus, the genetic data support a model in which the G12a-RhoGEF-RhoA signalling pathway is conserved in the control of axon regeneration in C. elegans. Consistent with this, GPA-12 and RHGF-1 are both expressed in motor neurons and physically interact 25 . In C. elegans, activated RHO-1 causes an increase in DAG by directly binding to and inhibiting DGK-1 (ref. 35). High levels of DAG activate the PKC homolog TPA-1, leading to activation of the JNK pathway by phosphorylation of MLK-1 MAPKKK 58 . These results raise the possibility that the G12a pathway activates axon regeneration by inducing an increase in DAG, resulting in the activation of the JNK pathway.
The Gsa cascade involves the activation of adenylyl cyclase by Gsa, and the consequent increase in cAMP production leads to the activation of PKA. Our recent results suggest that in response to axon injury, PKA upregulates transcription of the svh-2 gene, which encodes a growth factor receptor functioning in the JNK pathway 59,60 . Thus, one effector of the Gsa pathway that stimulates axon regeneration is a transcription factor that activates transcription of the svh-2 gene, resulting in activation of the JNK pathway. However, the Gsa pathway likely has additional targets. Taken together, these results suggest that the G12a and Gsa pathways converge to activate the JNK pathway in axon regeneration.

Methods
C. elegans strains. C. elegans strains used in this study are listed in Supplementary  Table 3. All strains were maintained on nematode growth medium (NGM) plates and fed with bacteria of the OP50 strain, as described previously 61 .
Axotomy. All animals were subjected to axotomy at the young adult stage at 20°C. Young adult hermaphrodites were immobilized with 20 mM sodium azide in M9 buffer on 2% agarose pad under a cover slip. D-type motor neurons expressing GFP were imaged with a fluorescence microscope. Selected commissural axons, mainly posterior D-type neurons, were cut using a 440-nm MicroPoint ablation Laser System from Photonic Instruments. After surgery, animals were allowed to recover on an agar plate and remounted for fluorescent imaging B24 h after surgery. For measurements of D neurons, imaged commissures that exhibited growth cones or small branches on the proximal fragment were counted as 'regenerated'. The proximal fragments that showed no change after 24 h were counted as 'no regeneration'. A minimum of 50 individuals with 1-3 axotomized commissures were observed for most experiments. For PLM touch neurons, extended process lengths were measured from maximum transparency projections of a single z-stack, by using the Keyence BZ-X700 analysis software.
Drug experiments. 5-HT and forskolin were obtained from Wako and Focus Biomolecules, respectively. For 5-HT treatment, worms were grown from egg to adult, up until axotomy, on NGM plates containing 6.5 mM 5-HT and OP50 Escherichia coli. After axotomy, the worms were plated on NGM plates containing 5-HT and OP50 E. coli for 24 h. For forskolin treatment, the mixed stages of worms were incubated with M9 buffer containing 100 mm forskolin and heat-killed OP50 for 12 h at 20°C, and the worms at young adult stage after the treatment were used.
Immunofluorescence. 5-HT immunofluorescence was essentially performed as described previously 5 . Briefly, the worms axotomized either D neurons or PLM neurons were fixed by 4% formaldehyde in PBS buffer at 4°C for overnight, washed three times by PBS and then treated by PBS containing 5% b-mercaptoethanol and 1% Triton X-100 for 24 h. Then, the worms were washed by PBS and digested in 100 mM Tris-HCl (pH7.5), 1 mM CaCl 2 and 5 mg ml À 1 collagenase (Wako, 034-10533). After digestion, a rat anti-serotonin antibody (Chemicon, Cat. No. MAB352), 100X diluted with Can Get Signal Immunostain (TOYOBO), was used as the first antibody solution. A chicken anti-rat Alexa 594 (Molecular probes), 200X diluted with Can Get Signal Immunostain (TOYOBO), was used as the second antibody solution.
Microscopy. Standard fluorescence images of transgenic worms were observed and photographed, using either a Zeiss Plan-APOCHROMAT Â 100 objective of a Zeiss Axioplan II fluorescence microscope with a Hamamatsu 3CCD camera, or a Â 100 objective of a Keyence BZ-X700 all-in-one fluorescence microscope. Confocal fluorescence images were taken on an Olympus FV500 and FV1000 confocal laser scanning microscopes with Â 60 and Â 100 objectives.
Quantification of GFP expression. Expression of GFP fluorescence was quantified using the ImageJ program (NIH). The cell bodies of severed D or PLM neurons were outlined with closed polygons and the fluorescence intensities within these areas were determined (I s ). Then I s was divided by the area to derive intensity per area (I sa ). To determine the background intensity of each cell, the same polygon was placed in an area neighbouring the cell body, fluorescence was measured (I b ), and this was divided by area (I ba ). The relative signal intensity per area (I ra ) was calculated as I sa À I ba . Cells having I ra 47 (for D neuron) or I ra 44 (for PLM neuron) were categorized as 'expressed', respectively. Statistical analysis. Statistical analyses were carried out as described previously 58 . Briefly, confidence intervals (95%) were calculated by the modified Wald method and two-tailed P values were calculated using Fisher's exact test (http://www.graphpad.com/quickcalcs/). The Welch's t -test was performed by using a t-test calculator (http://www.graphpad.com/quickcalcs/ttest1/).