The Integrin Signaling Network Promotes Axon Regeneration via the Src–Ephexin–RhoA GTPase Signaling Axis

Axon regeneration is an evolutionarily conserved process essential for restoring the function of damaged neurons. In Caenorhabditis elegans hermaphrodites, initiation of axon regeneration is regulated by the RhoA GTPase–ROCK (Rho-associated coiled-coil kinase)–regulatory nonmuscle myosin light-chain phosphorylation signaling pathway. However, the upstream mechanism that activates the RhoA pathway remains unknown. Here, we show that axon injury activates TLN-1/talin via the cAMP–Epac (exchange protein directly activated by cAMP)–Rap GTPase cascade and that TLN-1 induces multiple downstream events, one of which is integrin inside-out activation, leading to the activation of the RhoA–ROCK signaling pathway. We found that the nonreceptor tyrosine kinase Src, a key mediator of integrin signaling, activates the Rho guanine nucleotide exchange factor EPHX-1/ephexin by phosphorylating the Tyr-568 residue in the autoinhibitory domain. Our results suggest that the C. elegans integrin signaling network regulates axon regeneration via the Src–RhoGEF–RhoA axis. SIGNIFICANCE STATEMENT The ability of axons to regenerate after injury is governed by cell-intrinsic regeneration pathways. We have previously demonstrated that the Caenorhabditis elegans RhoA GTPase–ROCK (Rho-associated coiled-coil kinase) pathway promotes axon regeneration by inducing MLC-4 phosphorylation. In this study, we found that axon injury activates TLN-1/talin through the cAMP–Epac (exchange protein directly activated by cAMP)–Rap GTPase cascade, leading to integrin inside-out activation, which promotes axonal regeneration by activating the RhoA signaling pathway. In this pathway, SRC-1/Src acts downstream of integrin activation and subsequently activates EPHX-1/ephexin RhoGEF by phosphorylating the Tyr-568 residue in the autoinhibitory domain. Our results suggest that the C. elegans integrin signaling network regulates axon regeneration via the Src–RhoGEF–RhoA axis.


Introduction
The ability of neurons to regenerate damaged axons is essential for functional recovery. The regeneration of axons after injury requires the induction of multiple intracellular changes. To achieve this regeneration, the axon undergoes the following processes: local cytoskeletal reorganization to promote growth cone formation, lipid and protein transport for axon outgrowth, and activation of transcription factors that trigger regenerative programs (He and Jin, 2016). Therefore, manipulation of these processes could be an attractive strategy for therapeutic intervention to improve neuronal regeneration. However, the underlying molecular mechanisms that regulate these processes are not fully understood.
Following nerve injury, the end of the damaged axon is transformed into a growth cone-like structure. This is a crucial step in mounting a successful regenerative response. In particular, the cytoskeletal reorganization that accompanies the generation of the new growth cone is essential for the intrinsic ability to regenerate (Bradke et al., 2012). Growth cone formation and axonal regeneration require a number of changes in actin cytoskeletal dynamics and alterations in microtubule stability (Coles and Bradke, 2015). During axon outgrowth, the Rho family of guanosine triphosphatases (GTPases), such as RhoA, Rac1, and Cdc42, plays an important role in transducing signals that lead to the reorganization of the actin cytoskeleton within the growth cone (Hall, 1998). Rho GTPases are molecular switches that cycle between an inactive GDP-bound form and a GTP-bound active conformation (Boguski and McCormick, 1993). The formation of active Rho GTPases is accelerated by guanine nucleotide exchange factors (GEFs; Cerione and Zheng, 1996). Activated Rho GTPases can interact with effector proteins, which, in turn, trigger a variety of cellular responses. Among RhoA effectors, Rho-associated coiled-coil kinase (ROCK) is known to play a key role in actin organization through myosin activation (Bishop and Hall, 2000).
The nematode Caenorhabditis elegans is a valuable model for elucidating the molecular mechanisms involved in axon regeneration (Yanik et al., 2004). Genetic studies in C. elegans have identified common biological pathways that use conserved molecules to regulate regeneration (Hammarlund et al., 2009;Nix et al., 2011;Bejjani and Hammarlund, 2012). We have recently reported that the C. elegans RhoA homolog RHO-1 promotes axon regeneration of motor neurons by activating the downstream effector LET-502, the C. elegans homolog of ROCK . Activated LET-502 phosphorylates nonmuscle myosin light-chain (MLC) MLC-4. Phosphorylation of MLC activates the Mg 21 -dependent ATPase activity of nonmuscle myosin II, resulting in actin-myosin interaction (Amano et al., 1996). Thus, the RHO-1/RhoA-LET-502/ROCK pathway positively regulates C. elegans axon regeneration through MLC-4 phosphorylation (Fig. 1A). However, it is still unknown how the RHO-1 pathway is activated on axon injury.
Integrin signaling is one well characterized pathway that regulates the activation status of Rho GTPases (Huveneers and Danen, 2009). Integrins are a-b -heterodimeric transmembrane receptors that mediate cell-extracellular matrix (ECM) interactions (Hynes, 1987). Upon binding to ECM proteins, integrins can transduce signals that activate Rho GTPases, thereby initiating cytoskeletal rearrangement. Integrin function is regulated by inside-out and outside-in signaling (Kim et al., 2011). Inside-out signaling converts the extracellular domain into a high-affinity receptor, allowing interaction with a ligand, which alters the activity of intracellular components such as protein kinases and GTPases (outside-in signaling). Thus, integrins are bidirectional signaling receptors that transmit information into and out of cells. A critical step in the activation of inside-out signaling is the binding of the cytoskeletal protein talin to the cytoplasmic domain of b -integrin. This interaction leads to the dissociation of the integrin aand b -transmembrane domains that modulate downstream signaling (Kim et al., 2003). In the mammalian nervous system, integrins are involved in axon growth, synaptogenesis, and axon regeneration (Nikonenko et al., 2003;Eva and Fawcett, 2014). The C. elegans integrins INA-1/ integrin a and PAT-3/integrin b also participate in axon regeneration (Pastuhov et al., 2016;Hisamoto et al., 2019). In this study, we investigated the relationship between integrin signaling and RhoA activation in the regulation of axon regeneration in C. elegans. We found that TLN-1/talin-mediated integrin inside-out activation promotes axon regeneration by activating the RHO-1/RhoA signaling pathway. We showed that nonreceptor tyrosine kinase Src, an important effector downstream of integrin, phosphorylates EPHX-1/ ephexin RhoGEF at the Tyr-568 residue located in the autoinhibitory region. This phosphorylation relieves the autoinhibitory regulation of EPHX-1, leading to its activation. These results suggest that the integrin-Src-RhoGEF cascade activates the RhoA-ROCK pathway in axon regeneration.

Materials and Methods
C. elegans strains. The C. elegans strains used in this study are listed in Table 1. All strains were maintained on nematode growth medium Representative D-type motor neurons in wild-type (WT) and pat-3 mutant animals 24 h after laser surgery. In wildtype animals, severed axons exhibited regenerating growth cones (yellow arrowheads). In pat-3 mutants, the proximal ends of axons failed to regenerate (white arrowheads). Scale bar, 10 mm. C, Percentages of axons that initiated regeneration 24 h after laser surgery in the young adult stage. The numbers of axons examined are shown. Error bars indicate 95% confidence intervals. pppp , 0.001, as determined by Fisher's exact test.
plates and fed with bacteria of the OP50 strain by the standard method, as described previously (Brenner, 1974).
Microscopy. Fluorescent images of transgenic animals were observed under the 100Â objective lens of a fluorescence microscope (model ECLIPSE E800, Nikon) and photographed using a CCD camera (Zyla, Oxford Instruments).
Axotomy. Axotomies were performed as described previously (Li et al., 2012). Animals were subjected to axotomy at the young adult stage. Commissures that displayed growth cones or small branches present on the proximal fragment were counted as "regenerated." Proximal fragments that showed no change after 24 h were counted as "no regeneration." A minimum of 20 individuals with 1-3 one to three axotomized commissures was observed for most experiments.
Biochemical analysis. Transfection of transgenes into COS-7 cells, preparation of the cell lysates, immunoprecipitation, and immunoblotting using anti-FLAG and anti-GFP antibodies has been described previously (Li et al., 2012).

Results
PAT-3/integrin b functions in the RhoA-ROCK-MLC phosphorylation pathway to regulate axon regeneration C. elegans axon regeneration after axon injury is regulated by the RHO-1/RhoA-LET-502/ROCK-MLC-4 phosphorylation pathway ( Fig. 1A; Shimizu et al., 2018). We sought to determine whether integrin functions upstream of the RHO-1 pathway in axon regeneration. We have previously reported that PAT-3/ integrin b is involved in axonal regeneration (Hisamoto et al., 2019). Here, we first confirmed the pat-3 mutant phenotype. We subjected GABA-releasing D-type motor neurons (D neurons) to laser axotomy and subsequently monitored the regrowth of their axons. D neurons extend axons from the ventral to dorsal side of the animal body (Fig. 1B). In young adult wild-type animals, ;70% of severed axons formed growth cones and initiated regeneration within 24 h after injury ( Fig. 1B,C, Table 2). However, in pat-3(gk804163) mutants, the frequency of axon regeneration was substantially reduced (Fig. 1B,C, Table 2). Next, we investigated the relationship between PAT-3 and the RHO-1 pathway. The activation of the RHO-1 pathway is bypassed by the introduction of constitutively active downstream effectors. The kinase activity of LET-502/ROCK is usually tln-1(e259); kmEx1406 [Punc-25::venus::mlc-4(DD)] pat-3(D768R); rap-2(gk11) 60 21 ( autoinhibited by its C-terminal region, and truncation of the C-terminus constitutively activates LET-502 kinase activity . We found that the expression of the LET-502 C-terminal truncated mutant (LET-502DC) from the unc-25 promoter in D neurons could suppress the regeneration defect of pat-3 (gk804163) mutants (Fig. 1C, Table 2). Overexpression of LET-502DC does not promote axon regeneration in wild-type animals . These results suggest that PAT-3 functions upstream of LET-502 in axon regeneration.
RhoA activates myosin II via ROCK, which phosphorylates Thr-18 and Ser-19 of the MLC protein (Amano et al., 1996). Next, we examined whether PAT-3 signaling promotes axon regeneration by inducing the phosphorylation of MLC-4. In C. elegans, MLC-4 is phosphorylated by LET-502 at two highly conserved residues, Thr-17 and Ser-18 . We introduced constitutive phosphomimetic mutations, T17D and S18D, into the mlc-4 gene [mlc-4(DD)] and found that the expression of MLC-4(DD) could suppress the axon regeneration defect observed in pat-3(gk804163) mutants (Fig. 1C, Table 2). Expression of MLC-4(DD) in wild-type animals has no effect on the frequency of axon regeneration . Together, these results suggest that PAT-3 functions as an upstream component in the RHO-1-LET-502-MLC-4 phosphorylation pathway to promote axon regeneration.
In the pat-3(gk804163) mutant, the Pro-790 residue in the membrane-proximal NPxY motif Asn-Pro (790)-Ile-Tyr is replaced with a leucine residue ( Fig. 2A; Hisamoto et al., 2019). Integrin activation requires this NPxY motif in the b -tail to interact with the PTB (phosphotyrosine-binding) domain in the N-terminal portion of talin (Calderwood et al., 2003). Therefore, the pat-3(gk804163) mutant is expected to be defective in integrin activation by talin binding. Functional integrin receptors are heterodimers composed of one a-subunit and one b -subunit. Before integrin activation, they form a salt bridge between Arg (Arg-1044 in integrin a6) in the GFFKR motif and Asp (Asp-759 in integrin b 1) in the KLLxIIHD motif, which stabilizes the inactive form of integrin ( Fig. 2A; Hughes et al., 1996). During integrin inside-out activation, talin binds to the cytoplasmic b -tail and forms a salt bridge with the conserved Asp residue, which disrupts the inhibitory interaction of the b -tail with the conserved Arg residue in a-integrin (Wegener et al., 2007). Therefore, the charge-reversal mutations D759R in integrin b 1 and R1044D in integrin a6, wherein Asp-759 and Arg-1044 are replaced by arginine and aspartic acid residues, respectively, lead to a constitutively active integrin via the disruption of the a-b salt bridge ( Fig. 2A; Hughes et al., 1996;Laursen et al., 2011). Because the conserved motifs GFFKR and KLLtVLHD are also present in INA-1/integrin a (Arg-1114) and PAT-3/ integrin b (Asp-768), respectively ( Fig.  2A), we predicted that ina-1(R1114D) and pat-3(D768R) mutations would result in constitutively active integrins. To test this possibility, we generated ina-1(R1114D) and pat-3(D768R) mutants in the endogenous ina-1 and pat-3 loci, respectively, by CRISPR/Cas9 mutagenesis. We found that the ina-1(R1114D) and pat-3(D768R) mutations were able to suppress the tln-1(e259) phenotype of defective axon regeneration (Fig. 2C, Table  2). These results support the possibility that TLN-1 promotes axon regeneration through integrin inside-out activation.
The gk804163 allele of pat-3 fails to bind to TLN-1 and consequently associates constitutively with INA-1. Thus, the pat-3(gk804163) mutation is predicted to have a dominant-negative effect on axon regeneration rather than act as a loss-offunction mutation. Consistent with this prediction, we found that the introduction of the D768R mutation into the pat-3(gk804163) background could suppress the regeneration defect (Fig. 2D, Table  2). Moreover, the ina-1(R1114D) mutation was able to suppress the pat-3 (gk804163) phenotype (Fig. 2D, Table 2). These results suggest that the pat-3 (gk804163) mutation inhibits axon regeneration by constitutively binding to INA-1.
The RAP-2 GTPase2TLN-1 pathway mediates integrin inside-out activation during axon regeneration What is the mechanism that triggers activation of talin during axon regeneration? Talin is autoinhibited in the cytosol by the interaction of the N-terminal head domain with the C-terminal rod domain, which prevents the interaction of the head domain with the membrane surface and b -integrin cytoplasmic tail (Goksoy et al., 2008). Autoinhibited talin is recruited to the membrane via its freely accessible head domain in a Rap1 GTPase-dependent manner (Zhu et al., 2017). Therefore, Rap1 is essential for integrin inside-out signaling (Fig. 3A). We next examined whether C. elegans RAP is also involved in axon regeneration regulated by talin-mediated integrin inside-out activation. The C. elegans genome contains the rap-1 and rap-2 genes, which encode the  1 À 431), and FLAG-GTP-bound RAP-2(G12V) or GDP-bound RAP-2(S17A) as indicated. Complex formation was detected by immunoprecipitation (IP) with anti-FLAG antibody, followed by immunoblotting (IB) with anti-GFP antibody. Total lysates were immunoblotted with anti-GFP antibody. WT, Wild type; Int, integrin.
Rap1 plays a role in activating talin by recruiting it to the plasma membrane, where the autoinhibitory interaction between the talin N-terminal head and C-terminal rod domains is disrupted, which can expose the integrin b -tail binding site in talin ( Fig. 4A; Gingras et al., 2019). A recent study using nuclear magnetic resonance spectroscopy elucidated the structural basis for talin autoinhibition and revealed that the Met-319 site plays a crucial role in the interaction with the C-terminal rod domain of talin (Goksoy et al., 2008). Therefore, the M319A mutation disrupts the talin autoinhibitory interaction, thereby inducing its constitutively open conformation and activation (Fig. 4A). The Talin is autoinhibited by the interaction of the N-terminal head domain with the C-terminal rod domain, which prevents the interaction of the head domain with the membrane surface and b -integrin cytoplasmic tail. The talin head domain is composed of four subdomains: F0, F1, F2, and F3. Sequence alignment of F3 subdomains between talin 1 and TLN-1 is shown. Identical and similar residues are highlighted with black and gray shading, respectively. The Met-319 site in the talin F3 domain plays a crucial role in interacting with the C-terminal rod domain. The conserved residues Met-319 (in talin 1) and Leu-347 (in TLN-1) are indicated by arrows. B, D, Percentages of axons that initiated regeneration 24 h after laser surgery in the young adult stage. The numbers of axons examined are shown. Error bars indicate 95% confidence intervals. ppp , 0.01, pppp , 0.001, as determined by Fisher's exact test. NS, Not significant. C, Introduction of the RAP-2 CAAX box into the tln-1 locus. Sequence alignment of CAAX boxes (red line) in Rap1 and RAP-2 is shown. Identical and similar residues are highlighted with black and gray shading, respectively. The C-terminal 18 residues of RAP-2 were inserted by CRISPR/Cas9 mutagenesis into the C-terminal site of the tln-1 locus carrying the L347A mutation. E, Partners of TLN-1. TLN-1 interacts with PAT-3 and DEB-1. WT, Wild type. region around the Met-319 site in mammalian talin is conserved in C. elegans TLN-1, with the Leu-347 residue corresponding to Met-319 (Fig. 4A). We speculated that the tln-1(L347A) mutation would constitutively activate TLN-1; however, the tln-1(L347A) mutation failed to suppress the axon regeneration defect observed in rap-2(gk11) mutants (Fig. 4B, Table 2). Because recruitment of talin to the membrane is required for integrin activation (Goksoy et al., 2008), the ability of the tln-1(L347A) mutation to suppress the rap-2(gk11) defect in axon regeneration may be dependent on membrane localization of TLN-1(L347A). Rap1 contains a C-terminal CAAX box that plays an important role in its localization to cellular membranes (Hancock, 2003). To investigate the effect of the membrane localization of TLN-1 (L347A) on the suppression of the rap-2(gk11) phenotype, we used CRISPR/Cas9 mutagenesis to introduce the C-terminal 18 residues of RAP-2, which contain the CAAX box, into the COOH-terminal site at the tln-1 locus carrying the L347A mutation (Fig. 4C). We found that tln-1(L347A)::CAAX was able to suppress the axon regeneration defect in rap-2(gk11) mutants (Fig. 4B, Table 2). These results suggest that TLN-1 (L347A)::CAAX bypasses the requirement for RAP-2 activity to activate integrin in the axon regeneration pathway.
Next, we tested linking RAP-2 to the PAT-3ÀLET-502/ROCKÀMLC-4 phosphorylation pathway in axon regeneration. We found that the regeneration defect in rap-2(gk11) mutants was not suppressed by the constitutively active pat-3(D768R) mutation or expression of LET-502DC in D neurons (Fig. 4D, Table 2). Because RAP-2 regulates axon regeneration by activating TLN-1, TLN-1 may mediate signals through two different pathways to promote axon regeneration. A possible second target for a TLN-1-mediated regeneration pathway might be identified by noting that talin also activates vinculin, an essential linker protein between the actin cytoskeleton and ECM-bound integrins (Izard et al., 2004;Parsons et al., 2010). Loveless et al. (2017) demonstrated that the C. elegans vinculin ortholog DEB-1 is associated with TLN-1 (Fig. 4E). Therefore, we examined whether DEB-1 participates in axon regeneration and observed that deb-1(gk329549) mutants were impaired in axon regeneration (Fig. 4D, Table 2). In contrast to pat-3 (gk804163) mutants, expression of LET-502DC or MLC-4(DD) failed to suppress the deb-1(gk329549) phenotype (Fig. 4D, Table 2). These results suggest that TLN-1 regulates axon regeneration via PAT-3-and DEB-1-mediated pathways. Based on these results, it is possible that the tln-1(e259) mutation is defective in activating the PAT-3 pathway but is capable of activating the DEB-1 pathway.

EPAC-1/Epac activates RAP-2 GTPase to promote axon regeneration
We further evaluated the role of RAP-2 in the axon regeneration pathway. Like other GTPases, Rap exists in inactive GDP-bound and active GTP-bound states. A GEF protein promotes the exchange of GDP for GTP and activates its target GTPase (Boguski and McCormick, 1993). A number of GEFs that mediate the activation of mammalian Rap1 have been identified. An intriguing RapGEF is Epac (exchange protein directly activated by cAMP), because this GEF represents a direct target for cAMP, independent of the classical cAMP target, protein kinase A (PKA; de Rooij et al., 1998;Kawasaki et al., 1998). Indeed, Epac is known to be involved in the control of integrin-mediated cell adhesion (Rangarajan et al., 2003), and cAMP also participates in Error bars indicate 95% confidence intervals. pp , 0.05, ppp , 0.01, pppp , 0.001, as determined by Fisher's exact test. NS, Not significant. C, cAMP-binding sites in EPAC-1. Sequence alignment of the first and second cNBDs between Epac2 and EPAC-1 is shown. Identical and similar residues are highlighted with black and gray shading, respectively. The Gly-114 and Gly-422 sites in Epac2 are essential for its cAMP binding activity. The conserved residues, Gly-114 (in Epac2), Gly-84 (in EPAC-1), Gly-422 (in Epac2), and Gly-440 (in EPAC-1) are indicated by arrows. WT, Wild type. axon regeneration (Bhatt et al., 2004;Pearse et al., 2004;Ghosh-Roy et al., 2010). On the basis of these results, we hypothesized that cAMP-activated Epac is responsible for Rap activation, which leads to talin-mediated integrin activation in axon regeneration (Fig. 5A). The epac-1 gene encodes the C. elegans homolog of Epac ( Fig. 5A; Tada et al., 2012). We therefore examined whether epac-1 is required for axon regeneration, and we found that epac-1(tm3203) mutants exhibited a phenotype defective in axon regeneration (Fig. 5B, Table 2). We then analyzed the genetic interaction of epac-1 with rap-2. We found that epac-1 (tm3203); rap-2(gk11) double mutants were almost as defective in axon regeneration as rap-2(gk11) single mutants (Fig. 5B, Table 2). These results suggest that rap-2 and epac-1 act on the same axis controlling axon regeneration. Furthermore, we found that the expression of constitutively active RAP-2 (G12V), but not inactive RAP-2(S17A) (Fig. 3B), from the unc-25 promoter in Dtype motor neurons could suppress the epac-1 phenotype (Fig. 5B, Table 2), suggesting that EPAC-1 is a GEF for RAP-2 in axon regeneration.
Next, we examined whether cAMP binding is important for EPAC-1 to promote axon regeneration. Epac contains a cyclic nucleotide binding domain (cNBD), which is fused directly to the GEF domain as a single polypeptide chain ( Fig. 5A; Bos, 2003). cNBD folds on the GEF domain and prevents its interaction with downstream effectors. Binding of cAMP to Epac results in a conformational change that allows its GEF domain to interact with Rap, leading to Rap activation and subsequent downstream effects. C. elegans EPAC-1 contains two cNBDs, the first at the N terminus and the second in the middle region ( Fig. 5A; Bos, 2003). The human Epac2(G114E; G422D) mutant, wherein both Gly-114 and Gly-422 were replaced with glutamic acid and aspartic acid residues, respectively, has been shown to be defective in cAMP binding ( Fig. 5C; Ozaki et al., 2000). C. elegans EPAC-1 possesses conserved sites Gly-84 and Gly-440 in the first and second cNBDs, corresponding to Gly-114 and Gly-422 in Epac2, respectively (Fig. 5C). To determine whether the cAMP binding activity of EPAC-1 is important for axon regeneration, we generated epac-1(G84E; G440D) knock-in mutants using the CRISPR/ Cas9 technique. We found that epac-1(G84E; G440D) mutants had significantly reduced axonal regeneration (Fig.   5D, Table 2), indicating that EPAC-1 is required for axon regeneration in a manner dependent on its cAMP binding activity. Altogether, these results suggest that axon injury leads to increased cAMP levels, which, in turn, activate the EPAC-1ÀRAP-2ÀTLN-1-mediated integrin inside-out activation signaling pathway to promote axon regeneration. Figure 6. Src phosphorylation of EPHX-1 RhoGEF is essential for axon regeneration. A, Activation of the RHO-1ÀLET-502 pathway by SRC-1 and RhoGEF in axon regeneration. SRC-1 activates RhoGEF by phosphorylating a specific tyrosine residue, which, in turn, leads to the activation of RHO-1. GTP-bound active RHO-1 activates LET-502, resulting in MLC-4 phosphorylation and promoting axon regeneration. B, Structure of EPHX-1. Schematic diagrams of EPHX-1 and mammalian ephexin1 are shown. DH domain, yellow; PH domain, blue; SH3 domain, green. Conserved autoinhibitory regions of EPHX-1 and RhoGEFs of the Dbl family are shown in red and highlighted in the multiple-sequence alignment. Identical and similar residues are highlighted with black and gray shading, respectively. Tyr-568 of EPHX-1 is indicated by an asterisk. The EPHX-1(557-656) region is shown (underlined). C, Src phosphorylation of EPHX-1. In vitro phosphorylation of EPHX-1 (557-656) by Src is shown. COS-7 cells were transfected with GFP-EPHX-1(557-656) or GFP-EPHX-1(557-656; Y568F), and cell lysates were immunoprecipitated with an anti-GFP antibody. Immunoprecipitates were incubated with active recombinant Src in the presence of [g -32 P]ATP for 20 min at 30°C. Autophosphorylated Src and phosphorylated GFP-EPHX-1(557-656) were resolved by SDS-PAGE (top panel, 32 P). Protein input was confirmed by Coomassie Brilliant Blue (CBB) staining. D, Percentages of axons that initiated regeneration 24 h after laser surgery in the young adult stage. The numbers of axons examined are shown. Error bars indicate 95% confidence intervals. ppp , 0.01, pppp , 0.001, as determined by Fisher's exact test. WT, Wild type.
Integrin activates the RhoA signaling pathway via src and RhoGEF in axon regeneration How does the inside-out activation of integrin regulate the RhoA signaling pathway in axon regeneration? Since RhoA activation depends on RhoGEF activity that catalyzes the GDP-GTP exchange reaction (Cerione and Zheng, 1996), a GEF for the RHO-1 GTPase should function downstream of the integrin signaling pathway in regulating axon regeneration (Fig. 6A). One potential mediator between integrin and RhoA activation is the nonreceptor tyrosine kinase Src. Mammalian Src family kinases can interact directly or indirectly with integrins to activate RhoGEFs of the Dbl family, such as ephexin, through tyrosine phosphorylation (Arias-Salgado et al., 2003;Sahin et al., 2005;Huveneers and Danen, 2009). C. elegans SRC-1/Src interacts directly with INA-1/integrin a and mediates integrin signaling in phagocytic cells (Hsu and Wu, 2010). Furthermore, we have previously demonstrated that SRC-1 and INA-1 are required for axon regeneration (Pastuhov et al., 2016). On the basis of these results, we hypothesized that SRC-1 transduces integrin signaling to activate RHO-1 by phosphorylating, thereby activating ephexin-like RhoGEF (Fig. 6A). The C. elegans genome contains EPHX-1, a homolog of mammalian ephexin, that belongs to the Dbl family (Fig. 6B). We therefore investigated the relationship between SRC-1 and EPHX-1 in the axon regeneration pathway. Interestingly, the Src phosphorylation site in ephexin is also conserved in EPHX-1, corresponding to the Tyr-568 residue (Fig. 6B). To test whether EPHX-1 Tyr-568 is phosphorylated by Src, in vitro kinase assays were performed. Since the full-length recombinant EPHX-1 protein obtained from Escherichia coli was insoluble, a shorter EPHX-1 fragment containing the Tyr-568 site (amino acids 557À656; Fig. 6B) was immunopurified from COS-7 cells expressing GFP-EPHX-1(557À656). We incubated active mammalian Src with EPHX-1(557 À 656) in vitro and observed the phosphorylation of EPHX-1(557À 656; Fig. 6C). Mutating Tyr-568 to phenylalanine in EPHX-1(557À656) substantially reduced its in vitro phosphorylation by Src (Fig. 6C). These results suggest that EPHX-1 Tyr-568 represents a substrate for direct phosphorylation by Src.

EPHX-1 N-terminal domain inhibits interaction with RHO-1
RhoGEFs of the Dbl family possess tandem Dbl homology (DH), pleckstrin homology (PH), and SH3 domains ( Fig. 7A; Schmidt and Hall, 2002). Many Dbl family proteins are autoinhibited by direct binding of a putative helix N terminal to the DH domain, which sterically hinders Rho GTPases and prevents activation ( Fig. 7A; Aghazadeh et al., 2000;Yohe et al., 2008). This A, The autoinhibition model for EPHX-1. The autoinhibitory helix (red) packs against a conserved pocket on the DH domain (yellow), which inhibits the interaction between RHO-1 and the DH domain. The tyrosine phosphorylation of the autoinhibitory helix causes its dissociation from the DH domain, resulting in the interaction with RHO-1. EPHX-1DN lacking amino acids 1-574 can constitutively associate with RHO-1. B, Yeast two-hybrid assays for the interactions of RHO-1 with EPHX-1 and EPHX-1DN. The reporter strain PJ69-4A was cotransformed with expression vectors encoding GAL4 DBD-RHO-1, GAL4 AD-EPHX-1, and GAL4 AD-EPHX-1DN, as indicated. Yeasts carrying the indicated plasmids were grown on selective plates lacking histidine and containing 10 mM 5-aminotriazole for 4 d. C, Percentages of axons that initiated regeneration 24 h after laser surgery in the young adult stage. The numbers of axons examined are shown. Error bars indicate 95% confidence intervals. ppp , 0.01, pppp , 0.001, as determined by Fisher's exact test. WT, Wild type. autoinhibition is relieved by the Src phosphorylation of the tyrosine residue in the autoinhibitory helix, which disrupts the interaction with the DH domain. Because Src phosphorylation of EPHX-1 Tyr-568 was required for its function in axon regeneration (Fig. 6D, Table 2), we expected that the deletion of the N terminus containing Tyr-568 (amino acids 1-574; EPHX-1DN; Fig.  7A) would generate a hyperactive EPHX-1. To demonstrate the influence of the EPHX-1 N-terminal domain in regulating its GEF activity, we examined the ability of EPHX-1DN to interact with RHO-1 using a yeast two-hybrid system. We found that EPHX-1DN formed a complex with RHO-1, whereas full-length EPHX-1 did not associate with RHO-1 (Fig. 7B). These results suggest that the access of RHO-1 to its binding DH domain is restricted by the N terminus of EPHX-1.
The increased activity of EPHX-1DN was also apparent in testing the epistatic relation between src-1 and ephx-1 in the regulation of axon regeneration. The expression of ephx-1DN from the unc-25 promoter could suppress the regeneration defect of src-1(cj293) mutants (Fig. 7C, Table 2). Based on these findings, we propose that the EPHX-1 N-terminal domain has an autoinhibitory function, which is released by SRC-1 phosphorylation of Tyr-568, resulting in the activation of RHO-1. Thus, SRC-1 functions upstream of EPHX-1 in the RHO-1-LET-502-mediated signaling pathway to regulate axon regeneration.

Discussion
Integrin has previously been implicated in axonal regeneration in both the peripheral nervous system (PNS) and CNS (Eva and Fawcett, 2014). Successful regeneration in the PNS following injury has been demonstrated correlating with the upregulation of specific integrin subunits, including a4, a5, a6, a7, and b 1 (Vogelezang et al., 2001;Ekström et al., 2003;Wallquist et al., 2004;Gardiner et al., 2007), suggesting the importance of different types of integrins in neuronal regeneration. In fact, the axonal regeneration of the facial nerve in mice defective in integrin a7 is severely impaired (Werner et al., 2000). In the case of CNS injury, upregulation of inhibitory molecules such as chondroitin sulfate proteoglycans and Nogo-A in a pathologic environment can result in integrin inactivation (Hu and Strittmatter, 2008;Tan et al., 2011), whereas induced activation can allow axons to overcome those inhibitory effects on regeneration (Tan et al., 2011). Thus, it is clear that the activation state of integrin is an important factor in achieving significant regenerative growth. However, the molecular mechanisms by which integrin contributes to axonal regeneration have been poorly described. Here, we show that the nonreceptor tyrosine kinase Src and its target ephexin mediate integrin signaling to promote axon regeneration through the RhoA-ROCK-MLC phosphorylation pathway in C. elegans. Our findings thus provide a valuable molecular insight into the integrin-regulated repair of damaged neurons.
In this study, we demonstrate that integrin activation is essential for creating a functional transmembrane receptor that can induce downstream cellular effects in axon regeneration. A key step in the inside-out activation of integrin signaling is the binding of talin to the cytoplasmic domain of the b -subunit (Kim et al., 2011). When the receptor is in the bent, inactivated state, a salt bridge is formed between the KLLxIIHD motif on the b -subunit and GFFKR motif on the a-subunit. When talin binds to the cytoplasmic tail of the b -subunit, the transmembrane domains of aand b -integrins separate, resulting in a conformational change that modulates downstream signaling. The mutated form of the integrin b -subunit D759R (numbered using the b 1 sequence) leads to constitutive activation of integrin by disrupting the salt bridge between the aand b -subunits (Laursen et al., 2011). The conserved sequences GFFKR and KLLtVLHD are also present in INA-1/integrin a and PAT-3/ integrin b , respectively. This raises the possibility that talinmediated inside-out activation of integrin may be involved in the regulation of axon regeneration. Indeed, we found that the pat-3, ina-1, and tln-1 mutants are defective in axon regeneration and that the pat-3(D768R) mutation, corresponding to the mammalian integrin b 1(D759R) mutation, can suppress the tln-1 mutant phenotype, suggesting that the pat-3(D768R) mutation is constitutively active. These results are consistent with the possibility that the TLN-1/talin-mediated activation of PAT-3/integrin b promotes axon regeneration (Fig. 8).
How is TLN-1 activated in response to axonal injury? Mammalian talin exists in the cytoplasm in a closed, autoinhibited conformation, in which intramolecular interactions between the N-terminal and C-terminal domains prevent integrin binding (Goksoy et al., 2008). Upon stimulation, talin is efficiently recruited to the cell membrane and released from its autoinhibitory conformation to trigger integrin activation (Song et al., 2012). It has been suggested that the Rap1 GTPase plays a critical role in activating talin by recruiting it to the membrane, relieving the autoinhibitory interactions (Zhu et al., 2017). In addition, Rap1 is activated by GEF Epac1, which is a target of cAMP (de Rooij et al., 1998;Kawasaki et al., 1998). The present study provides compelling evidence that the EPAC-1/Epac-RAP-2 GTPase signaling pathway activates TLN-1/talin-mediated integrin activation in axon regeneration (Fig. 8). In the absence of cAMP, mammalian Epac has been reported to exist in an inactive conformation such that access of Rap1 to the GEF domain of Epac is occluded by an intramolecular interaction between the cAMP-binding domain and catalytic region (Rehmann et al., 2003(Rehmann et al., , 2006. Upon cAMP accumulation, nucleotide binding promotes conformational change in Epac that uncouples the cAMPbinding domain from the catalytic region, enabling Rap1 activation. Thus, the Epac-Rap1 pathway represents a PKA-independent cAMP signaling cascade. The secondary messenger cAMP is In response to axon injury, cAMP levels are elevated, resulting in the activation of EPAC-1, which, in turn, activates RAP-2. GTP-bound RAP-2 can interact with and activate TLN-1, which then activates integrin and possibly DEB-1 as well. Next, integrin activation directs the GDP-GTP exchange activity of EPHX-1 toward RHO-1 via SRC-1-mediated phosphorylation of EPHX-1 on Tyr-568. Finally, GTP-bound RHO-1 activates LET-502, leading to MLC-4 phosphorylation, which promotes axon regeneration. involved in axonal regeneration through the activation of PKA (Gao et al., 2004). However, an additional cAMP-dependent mechanism involves EPAC-1, which also responds to physiological changes in cAMP concentration, promoting the activation of the RAP-2 GTPase.
Since the artificial localization of TLN-1 to the membrane can suppress the defect in axon regeneration caused by the rap-2 mutation, the main role of RAP-2 in axon regeneration is to recruit TLN-1 to the membrane. However, the constitutively active pat-3(D768R) mutation or LET-502DC expression is unable to suppress the rap-2 defect. Thus, the PAT-3-LET-502 pathway does not appear to be simply downstream of RAP-2 signaling, but another pathway is also likely to be important for axon regeneration. A possible factor in this other pathway might be vinculin, an essential linker protein between the actin cytoskeleton and integrins bound to the ECM. Upon activation by talin, vinculin binds to both integrin-bound talin and actin filaments, allowing cells to transmit the force generated by actomyosin to the ECM via the vinculin-talin-integrin complex (Zamir and Geiger, 2001;Izard et al., 2004). In fact, we find that DEB-1/ vinculin is involved in axon regeneration. In contrast to pat-3 mutants, the expression of LET-502DC does not suppress the deb-1 mutant defect in axon regeneration. Based on these results, we hypothesize that TLN-1 regulates axon regeneration through PAT-3-and DEB-1-mediated pathways (Fig. 8).
What are mediators of the transduction of integrin signals leading to RHO-1/RhoA GTPase activation? In general, Rho GTPases are activated by RhoGEFs, so we expected that a RhoGEF would be involved in regulating the axon regeneration pathway between integrin and RHO-1. Mammalian integrin a 4 -b 1 acts via the a 4 cytoplasmic domain to activate downstream Src, which, in turn, activates the GEF complex that acts on Rac GTPase to promote cell motility (Hsia et al., 2005). These results, combined with our findings, define a conserved integrin signaling pathway, in which activation of SRC-1 by integrin provides a link to RhoGEF-mediated RHO-1 activation during axon regeneration in C. elegans. We demonstrate that RhoGEF EPHX-1 acts downstream of SRC-1 in the integrin-RHO-1 signaling pathway (Fig. 8). EPHX-1 is a homolog of the mammalian RhoGEF ephexin, which belongs to the Dbl family of proteins. The Dbl family GEFs share the structural motif of the central DH catalytic domain in tandem with the regulatory PH domain. The N-terminal half of Dbl RhoGEFs has been proven to be characterized by a negative regulatory element for the DH-PH functional module (Aghazadeh et al., 2000;Yohe et al., 2007Yohe et al., , 2008Tanegashima et al., 2008). This intramolecular interaction hinders the Rho GTPase access to the DH domain needed to catalyze the guanine nucleotide exchange. Mammalian ephexin is phosphorylated at a tyrosine residue in an N-terminal motif and has considerable sequence identity with the autoinhibitory helix described for the members of the Dbl family (Sahin et al., 2005;Yohe et al., 2008). Upon the phosphorylation of this tyrosine residue in the N-terminal region by Src, Dbl RhoGEFs open to yield an active conformation with an exposed DH domain, relieving this autoinhibition. Indeed, the autoinhibitory helix is conserved between EPHX-1 and Dbl family members, and we observed a similar negative regulation of EPHX-1 by its N-terminal domain, as shown by the hyperactivity of the N-terminal truncation mutant. Src phosphorylation of Tyr-568 in the N-terminal domain of EPHX-1 suggests that this tyrosine phosphorylation induces EPHX-1 activation.
In the mammalian Dbl family neuronal guanine exchange factor (Ngef), Tyr-179 is phosphorylated by Src, and substitution of Tyr-179 with glutamic acid (Y179E) causes constitutive activation of Ngef GEF activity (Yohe et al., 2008). In contrast, the Y568E mutation in EPHX-1 is loss of function, which indicates that tyrosine phosphorylation is important for the GEF activity of EPHX-1. The nucleotide exchange activity of the Dbl family RhoGEFs is autoinhibited by an additional intramolecular interaction between the N-and C-terminal regions (Yohe et al., 2008). Binding of another protein to the N-terminal region would activate the GEF activity by disrupting this intramolecular interaction. One potential binding partner for EPHX-1 might be SRC-1 itself. Here, after SRC-1 phosphorylates EPHX-1 at Tyr-568, the SH2 domain of SRC-1 associates with pTyr-568 in EPHX-1, resulting in the activation of EPHX-1 GEF activity by relieving the autoinhibitory intramolecular interaction. Because SRC-1 associates with INA-1 (Hsu and Wu, 2010), this possibility also suggests that SRC-1 acts on EPHX-1 in the vicinity of the integrin receptor. If so, the interaction of TLN-1 with PAT-3 could induce the binding of SRC-1 to INA-1. Therefore, SRC-1 may not only promote GEF activity of EPHX-1, but also determine its subcellular localization.
We have previously shown that the integrin pathway also promotes axon regeneration through activation of the JNK cascade (Pastuhov et al., 2016;Hisamoto et al., 2019). In this pathway, SRC-1 provides a link between INA-1 and the GEF complex CED-2/CrkII, CED-5/DOCK180, and CED-12/ELMO, which activates CED-10/Rac GTPase. GTP-bound CED-10 interacts with and activates the Ste20-related protein kinase MAX-2, which phosphorylates and activates MLK-1 MAPKKK (Pastuhov et al., 2016). In this JNK pathway, integrin is activated by externalized phosphatidylserine generated by axon severing . Therefore, SRC-1 activates two separate signaling pathways via RhoGEFs, one that leads to activation of the Rac-JNK pathway, and a second that leads to activation of the RhoA-ROCK pathway that is linked to SRC-1 by TLN-1. Moreover, we have recently identified the C. elegans tensin protein TNS-1 as an adaptor protein for the SVH-2 Met-like receptor-JNK signaling pathway (Hisamoto et al., 2019). TNS-1 brings PAT-3 in close proximity to SVH-2 by associating with both proteins. Hence, it is likely that TNS-1 links SRC-1 to integrin-JNK signaling.