UNC-16 alters DLK-1 localization and negatively regulates actin and microtubule dynamics in Caenorhabditis elegans regenerating neurons

Abstract

Neuronal regeneration after injury depends on the intrinsic growth potential of neurons. Our study shows that UNC-16, a Caenorhabditis elegans JIP3 homolog, inhibits axonal regeneration by regulating initiation and rate of regrowth. This occurs through the inhibition of the regeneration-promoting activity of the long isoform of DLK-1 and independently of the inhibitory short isoform of DLK-1. We show that UNC-16 promotes DLK-1 punctate localization in a concentration-dependent manner limiting the availability of the long isoform of DLK-1 at the cut site, minutes after injury. UNC-16 negatively regulates actin dynamics through DLK-1 and microtubule dynamics partially via DLK-1. We show that post-injury cytoskeletal dynamics in unc-16 mutants are also partially dependent on CEBP-1. The faster regeneration seen in unc-16 mutants does not lead to functional recovery. Our data suggest that the inhibitory control by UNC-16 and the short isoform of DLK-1 balances the intrinsic growth-promoting function of the long isoform of DLK-1 in vivo. We propose a model where UNC-16’s inhibitory role in regeneration occurs through both a tight temporal and spatial control of DLK-1 and cytoskeletal dynamics.

Introduction

Neurons in the adult nervous system have a limited ability to regenerate after injury. Studies have shown that the extent of neuronal regeneration depends on the balance between factors that inhibit and promote neuronal regeneration which can be both extrinsic and intrinsic to the injured neuron. An important factor downstream of the injury signaling cascade is the dynamics of cytoskeletal elements (Hur et al. 2012; Chisholm 2013; Blanquie and Bradke 2018). After neuronal injury, microtubule dynamics increase under the control of an early injury signaling cascade (Lu et al. 2015). It has been shown that regulation of microtubule dynamics and polarity in response to JNK signaling is crucial to initiate regeneration of an injured axon (Stone et al. 2010). The regenerating neurons also form new growth cones which are rich in actin, actin filaments and their associated proteins and are together responsible for both formation and motility of the growth cone (Gomez and Letourneau 2014).

The balance between growth-promoting and inhibiting factors determines the extent to which a neuron regenerates. This can be achieved in multiple ways, for instance several kinases have been reported to regulate the intrinsic growth potential of regenerating neurons such as, Protein Kinase C (Lee et al. 2013), S6 Kinase (Hubert et al. 2014), MAP Kinases such as Erk (Hanz et al. 2003), JNK (Lindwall and Kanje 2005), and Dual Leucine zipper Kinase (DLK) (Hammarlund et al. 2009). Scaffolding molecules can regulate the activity of such kinases in a spatio-temporal manner. In addition, they can bring together several proteins serving to modulate entire signaling cascades. JIPs are a family of classical scaffolding molecules known to be able to switch their roles from growth-promoting to growth-inhibiting based on their levels, extent, and duration of activity (Whitmarsh 2006).

JIP3, a JNK-interacting protein (JIP) family member, assists in the formation of a functional JNK signaling module by assembling specific components of the MAPK cascade which are known to modulate neuronal growth in various systems (Ito et al. 1999; Kelkar et al. 2000; Byrd et al. 2001). JIP3 interacts with the specific kinases of the JNK signaling pathway i.e., JNK, MKK7, and MLK3 (Whitmarsh, Cavanagh et al. 1998). MAPKKKs are the most upstream kinases in the MAPK cascade and play a role in regulating signal specificity (Kyriakis and Avruch 2012). DLK, an MAPKKK, regulates synapse formation and is essential for neuronal regeneration (Nakata et al. 2005; Hammarlund et al. 2009). DLK-1 activation is tightly regulated and occurs only in the presence of specific stimuli such as neuronal injury (Nakata et al. 2005; Yan and Jin 2012). DLK also suppresses neuronal degeneration as DLK activation following a conditioning lesion in Drosophila larval PNS motoneurons protects axons from degeneration upon subsequent injuries (Xiong and Collins 2012). Homodimerisation of DLK is sufficient for its activation (Nihalani et al. 2000) and JIP1, another JIP family scaffold protein, maintains DLK in an inactive state in part by preventing its dimerization (Nihalani et al. 2003). Thus, the interaction of scaffolding molecules and kinases they scaffold can regulate the activity of the kinases involved in neuronal regeneration.

We investigated the role of a JIP family protein in regeneration. We show that UNC-16, the Caenorhabditis elegans JIP3 homolog (Byrd et al. 2001), inhibits regeneration of touch neurons after axotomy. The faster regeneration in unc-16 mutants is a result of quicker initiation of neuronal regrowth and is independent of Kinesin-1, Dynein, and JNK-1 but is dependent on DLK-1. Our data suggest that UNC-16 independently and in a concentration-dependent manner, inhibits regeneration by acting on the regeneration-promoting long isoform of DLK-1. UNC-16’s inhibitory effect on the long DLK-1 isoform is independent of and in addition to the inhibition by the short DLK-1 isoform. UNC-16 negatively regulates actin dynamics dependent on DLK-1 and microtubule dynamics partially dependent on DLK-1. Microtubule and actin dynamics in unc-16 mutants, post-injury, are also partially dependent on CEBP-1, a transcription factor with key roles in DLK-1 dependent neuronal regeneration (Yan et al. 2009; Li et al. 2015).

We also show that the faster regeneration seen upon loss of UNC-16 does not translate into functional recovery, but slower regeneration observed when UNC-16 is over-expressed shows functional recovery compared to wild-type animals. The dual inhibitory control by both UNC-16 and the short isoform of DLK-1 may balance the intrinsic growth-promoting function of the long isoform of DLK-1 in vivo and a slower regeneration rate through modulation of cytoskeletal dynamics may assist in functional recovery of the injured neuron.

Materials and methods

Strains

Caenorhabditis elegans strains were grown on NGM agar plates seeded with Escherichia coli OP50 bacteria at 16°C (Brenner 1974) (Table 1). jkk-1p::T7::unc-16 plasmid was generated by inserting T7-UNC-16 DNA excised from CMVT7p-UNC-16 plasmid (Byrd et al. 2001) into NHjkkp-1p, which was made by replacing jkk-1p::gfp DNA region of MK104p (Kawasaki et al. 1999) with multiple cloning sites. ofm-1p::DsRed-monomer has been described previously (Arimoto et al. 2011). The homozygosity of all MosSCI* (single-copy insertion of a transgene into a defined site) strains was confirmed using PCR before and after completion of the axotomy assays using reported primers (Yan and Jin 2012).

Table 1.

List of C. elegans strains

Strain	Genotype	Reference
SK4005	zdIs5[mec-4p::GFP]	Clark and Chiu (2003)
FF41	unc-116(e2310)	Patel et al. (1993)
CB109	unc-16(e109)	Brenner (1974)
TT130	unc-16(tb109) abbreviated as unc-16(lf)	Choudhary et al. (2017)
CZ3011	unc-16(ju146)	Byrd et al. (2001)
KG1758	unc-16(ce451)	Edwards et al. (2013)
MT1542	unc-16(n730)	Trent et al. (1983) and Byrd et al. (2001)
NM2417	juIs162 (unc-16p::unc-16::gfp)	Byrd et al. (2001)
VC8	jnk-1(gk7)	Villanueva et al. (2001)
CF702	muIs32[mec-7p::gfp]	Ch'ng et al. (2003)
CB57	unc-14(e57)	Brenner (1974)
NM1787	dhc-1(js319)	Koushika et al. (2004)
KU1000	N2; kmEx1000 [jkk-1p::T7::UNC-16]	Described above
CZ18975	juIs338 [mec-4p::ebp-2::GFP]	Ghosh-Roy et al. (2012)
CZ8920	cebp-1(tm2807) abbreviated as cebp-1(0)	Yan et al. (2009)
	wyIs291[unc-86p::GFP::utCH]	Chia et al. (2014)
CZ11327	dlk-1(ju476)zdIs5 abbreviated as dlk-1(Only S)	Yan and Jin (2012)
CZ16351	dlk-1(tm4024)zdIs5 abbreviated as dlk-1(0)
CZ11530	juEx2529 [mec-4p::dlk-1(L)::gfp]
CZ15142*	zdIs5; rgef-1p::dlk-1(L)(juSi50) abbreviated as Si(L)
CZ15144*	dlk-1(ju476)zdIs5; rgef-1p::dlk-1(L)(juSi50)
CZ16440*	dlk-1(tm4024)zdIs5; rgef-1p::dlk-1(L)(juSi50)
CZ15143*	zdIs5; rgef-1p::dlk-1(S)(juSi46) abbreviated as Si(S)
CZ15145*	dlk-1(ju476)zdIs5; rgef-1p::dlk-1(S)(juSi46)
CZ16352*	dlk-1(tm4024)zdIs5; rgef-1p::dlk-1(S)(juSi46)

Laser axotomy

Nanosecond laser (Spitlight 600, Innolas, Munich, Germany) pulses (λ = 355 nm) of energy 0.8 μJ in single-shot mode was used to generate neuronal cuts (Rao et al. 2008). One of the posterior lateral mechanosensory (PLM) neurons expressing soluble GFP was cut approximately 30–50 μm away from the cell body. All experiments were done using larval stage 4 animals grown at 16°C. We chose this temperature as the unc16(tb109) mutants show accelerated growth and we wanted to ensure that we capture the characteristics of neuronal regeneration in unc-16(tb109) mutants vis-à-vis wild type. Our experiments were carried out 9 h post-axotomy as they gave us clear differences between genotypes, data are represented as percentage regeneration that refers to the number of animals that show neuronal regeneration.

Imaging and analysis

Images for analysis were acquired using a 100X objective, 1.4 NA at 1 × 1 binning, and were processed using ImageJ software. Regeneration was represented as percent neuronal regeneration i.e., number of neuronal processes showing regrowth out of total number of axotomized neurons (Hammarlund et al. 2009; Nix et al. 2011). The length of the regrowing process was measured from the cut site (Wu et al. 2007; Zou et al. 2013). Neurons showing >4 µm regrowth were considered as regenerating. Imaging was done 9 h post-axotomy unless otherwise mentioned.

Statistical analysis

The data have been represented as mean ± s.e.m. The data when represented as % regeneration is an average of at least two separate trials. One-way ANOVA was used as a test of significance unless specified otherwise. P-values < 0.05 were considered as significant.

Localization of GFP tagged DLK isoforms

PLM neurons of one-day adult animals were imaged using a Perkin Elmer spinning-disc confocal microscope (488 nm laser, objective 100X, 1.4 NA) with an EMCCD C9100-13 camera.

Microtubule and actin dynamics measurement

Laser axotomy was performed using a ns-pulsed Nd-YAG LASER at 355 nm and energy 0.8 mJ (Litron Nano S60-30) using a 60X, 1.35 NA objective. The axotomy was performed on juIs338 or wyIs291 L4 animals 100 μm from the cell body. The animals were washed with M9 buffer and rescued. At the reported time point, the animals were anesthetized with 5 mM Levamisole and imaged using 488 nm laser excitation with a Yokogawa CSU-X1-A3 spinning disc confocal head and a Hamamatsu C9100-13 EMCCD Camera controlled by Volocity collectively supplied by PerkinElmer. Imaging was done under 100X 1.4 NA objective using 300 ms exposure at 3 frames/s for a total of 3 min for juIs338 and 600 ms exposure at 1.6 frames/s for a total of 3 min for wyIs291. Kymographs were generated using the ImageJ plugin MultiKymograph (Schindelin et al., 2012). We measure the total number of EBP-2::GFP comets which are visualized as sloped lines in kymographs. The sloped lines were also used to calculate the length and direction of the comets. We measured three parameters, viz., the density of immobile actin-rich regions, number of actin trails and actin trail length as described earlier (Sakamoto et al. 2005; Ganguly et al. 2015). EBP-2::GFP comets and GFP::utCH trails with polymerizing length above 0.3 µm were used for total number, speed, and length of EBP-2::GFP comet. The total number of EBP-2::GFP comets and GFP::utCH trails were then normalized to the duration of the image file as well as the length of the ROI in focus to generate the normalized microtubule dynamics graph. Immobile actin-rich regions are visible in distance and time kymograph plots as vertical lines and actin trails as sloped lines.

Behavioral assay

Laser axotomy was performed using a multiphoton IR laser (720 nm) on one PLM 50 µm away from the cell body. Two cuts were introduced 6 µm apart. A worm that is moving backward responds to a gentle touch to its posterior side by reversing its direction of locomotion (Chalfie et al. 1985). This touch assay was performed on both the sides (right and left) at 3, 9, and 24 h post axotomy where only one of the two PLM neurons was axotomized. The reversal response mediated by the side with the uninjured posterior touch neuron serves as an internal control for each animal whose behavior was assessed post-axotomy. Using this assay, one can measure touch sensation with a posterior touch response index (PTRI) (Basu et al. 2017) which has been plotted.

After ascertaining that there is no difference between PTRI of uncut and laser mock cut animals in all the genotypes tested, the touch assay was performed 3, 9, and 24 h post axotomy (Figure 9A). The PTRI was normalized to the unaxotomized PTRI of each genotype (Basu et al. 2017). The morphology assessment of the axotomized PLM was done after 24 h (Figure 9, B and C). For statistics, multiple column comparison of ANOVA, with Neumann Keul’s post-hoc test was used. *P < 0.01, ** P < 0.001, and *** P < 0.0001.

Figure 9 .

unc-16(lf) does not enhance functional recovery. (A) PTRI in axotomized and internal control neurons in wild type, unc-16(tb109) mutants and T7::UNC-16 (expressed by jkk-1p::T7::unc-16). Multiple column comparison of ANOVA, with Neumann Keul’s post-hoc test P-value *<0.01, **<0.001, and ***<0.0001. ns: not significant. (B) Percentage of animals showing neuronal fusion of the proximal and distal end post-axotomy at 9 h in wild type, unc-16(tb109) mutants and T7::UNC-16 animals. One-way ANOVA P-value ***<0.001. Number of animals inside bars. (C) Representative images of neuronal fusion of the proximal and distal end post-axotomy at 9 h in wild type, unc-16(tb109) mutants and T7::UNC-16 animals. Scale bar 10 µm. (D) Suggested model of the mechanism of inhibition of neuronal regeneration by UNC-16.

Results

Loss of UNC-16 leads to an increase in neuronal regeneration

To study UNC-16’s role in regeneration, we compared the post-axotomy regrowth in the posterior lateral microtubule (PLM) neuron of unc-16 mutants and wild-type animals. All six alleles viz., ce451, ju146, e109, tb109, n730, and km75 of unc-16, showed greater number of animals whose neurons regenerate as compared to wild type (Figure 1A) (Byrd et al. 2001; Edwards et al. 2013; Choudhary et al. 2017). This is similar to the inhibitory effect reported for unc-16 mutants in C. elegans motor neuron regeneration (Chen et al. 2011; Nix et al. 2014). tb109, n730, and km75 have stop codons which are predicted to truncate the protein at different lengths (Figure 1A). All six alleles also showed reduced unc-16 RNA levels compared to wild type (Supplementary Figure S1). The inhibitory effect of unc-16 on neuronal regeneration is not allele specific varying from 43% to 88% in different loss of function (lf) alleles. The alleles tb109 and ju146 had strong phenotypes and hence were chosen for further analysis.

Figure 1.

Loss of UNC-16/JIP3 promotes neuronal regeneration. (A) Percentage of neurons regenerating post-axotomy in unc-16 alleles and in wild type and unc-16(tb109) mutants upon expression of UNC-16::GFP (expressed under the unc-16 promoter) and T7::UNC-16 (expressed under the jkk-1 promoter). Inset: UNC-16 domain structure and location of mutations in unc-16 mutant alleles. (B) Percentage of neurons regenerating at different time points post-axotomy in wild type, unc-16(ju146) mutants and unc-16(tb109) mutants. (C) Representative images of the PLM touch neuron process at different time points post-axotomy in wild type and unc-16(tb109) mutants. Arrow indicates axotomy location. Scale bar 10 µm. (D) Length of the regenerated neuronal process at different time points after axotomy. (E) Regrowth rate of the neuronal process in each 3-h interval after axotomy. One-way ANOVA, P-value *<0.05, **<0.005, ***<0.001. Number of animals inside bars in (A). Number of animals is 20 for (B, D, and E).

We found that expression of a transgene that encodes UNC-16::GFP (Byrd et al. 2001) or T7::UNC-16 (jkk-1p::T7::unc-16) in unc-16(tb109) mutants reduces regeneration from 81% to 27% (Figure 1A). Furthermore, expression of UNC-16::GFP or T7::UNC-16 in wild-type animals resulted in 19% of the severed processes regenerating compared to 36% in wild type. This suggests that the presence and levels of UNC-16 is sufficient to inhibit regeneration following injury.

Loss of UNC-16 results in faster initiation of regeneration

To examine the time course of regeneration in unc-16 mutants, we monitored the neuronal processes of unc-16(tb109) and unc-16(ju146) mutants at different time points after axotomy. At 6 h, 38% of unc-16(ju146) and 57% of unc-16(tb109) mutants initiated neuronal regrowth compared to only 16% in wild type (Figure 1, B and C). Furthermore, neuronal outgrowth in both alleles of unc-16 mutants examined, started earlier at 3 h post axotomy, unlike in wild-type injured axons, where it began at 6 h (Figure 1, B and D). At 9 h, unc-16(ju146) and unc-16(tb109) mutants showed a ∼2.6 and ∼6.7 fold significantly greater neuronal process length compared to wild type (Figure 1B). The increase in the number of animals showing regrowth and the greater length of neuronal outgrowth, together, show that loss of unc-16 promotes earlier initiation of regrowth. Over-expression of UNC-16 in neurons did not alter the time when regrowth was initiated, where growth began similar to wild-type animals at 6 h post-axotomy (Figure 1D).

We also examined the rate of regrowth after initiation by measuring the length of the regrowing axon at defined times after axotomy. Six to nine hours post-axotomy, the neuronal processes in unc-16(tb109) mutants grew at the rate of 2.1 µm/h, significantly higher than the wild-type neuronal growth rate of 0.6 µm/h (Figure 1E). The difference in growth rate between wild-type (1.7 µm/h) and unc-16(tb109) (6.4 µm/h) mutants was greatest in the 9–12-h time interval post-axotomy. We also evaluated the growth rate in unc-16(ju146) mutants and found similar trends (Figure 1E). Elevating UNC-16 levels using a transgene expressing T7:: UNC-16 reduces the growth rate (Figure 1E), eventually resulting in shorter process lengths 24 h post-axotomy (Figure 1D).

At 18 h post-axotomy, the neuronal regrowth rate in unc-16(tb109) mutants was reduced and became comparable to wild type (Figure 1E). Twenty-four hours post-axotomy, more than 50% of the processes of unc-16(tb109) mutants show several branches (Figure 1C), some that overlap with the distal process; thus, an accurate assessment of true growth rates is not measurable. Taken together our data suggest that the loss of unc-16 results in a quicker initiation of regrowth and a faster rate of process regrowth up to 12 h post-axotomy. These data suggest that UNC-16 acts as an inhibitory factor in the early steps of axon regeneration and prevents regrowth initiation and slows the rate of axon extension after injury.

Increased regeneration in unc-16 depends on DLK-1

UNC-16 is a scaffolding protein that plays multiple roles in intracellular transport and MAPK signaling. It interacts with the Kinesin-1 motor (Sun et al. 2011), the Dynein motor (Arimoto et al. 2011), an RUN domain protein UNC-14 (Sakamoto et al. 2005), a leucine-rich repeat kinase 2 (LRRK2) (Hsu et al. 2010; Choudhary et al. 2017) and MAP kinases such as JNK-1 (Kelkar et al. 2000) and DLK (Ghosh et al. 2011). We thus examined if regeneration in UNC-16 depends on the above proteins.

We assessed neuronal regeneration in loss-of-function (lf) mutants of the Kinesin heavy chain-1 (unc-116), the JNK MAP kinase (jnk-1), RUN domain protein (unc-14), leucine-rich repeat kinase 2 (lrk-1), Dynein heavy chain (dhc-1), and the Dual-leucine zipper kinase (dlk-1). The single mutants either regenerate similarly to wild type, as seen in the unc-116 and the jnk-1 mutants or regenerate significantly less than wild type as seen in mutants of unc-14, lrk-1, dhc-1, and dlk-1 (Figure 2A, Supplementary Figures S2 and S3 ). In addition, double mutants between unc-16(tb109) and the above-mentioned genes did not alter UNC-16-dependent regeneration (Figure 2A, Supplementary Figures S2 and S3) except in unc-16; dlk-1 double mutants. We examined the effect of loss of UNC-16 in different alleles of dlk-1 (Figure 2A, Supplementary Figure S4). The milder dlk-1 alleles ju477 and ju591 (Yan and Jin 2012) (Supplementary Figure S4) do not affect regeneration on their own, but they inhibit regeneration in unc-16(tb109) mutants. Regeneration in dlk-1; unc-16(lf) double mutants built with stronger alleles of dlk-1, viz. ju476 and tm4024, is completely abolished (Figure 2A). Thus, the inhibitory role of UNC-16 in neuronal regeneration is dependent on DLK-1.

Figure 2.

UNC-16 inhibits DLK-1L. (A) Effect of unc-16(tb109) mutation on regeneration phenotype of various alleles of dlk-1. (B) Genetic analysis of regeneration in unc-16(tb109), the null dlk-1(tm4024) [dlk-1(0)] and dlk-1(ju476) [dlk-1(Only S)] mutants, expressing single copy transgene of MosSci DLK-1L [Si(L)] or MosSci DLK-1S [Si(S)] 9 h post-axotomy. ns: not significant. (C) Genetic analysis of effect of unc-16(tb109) mutation on animals expressing a single copy MosSci DLK-1L in wild type and dlk-1(tm4024) mutant background 6 h post-axotomy. One-way ANOVA, P-value *<0.05, **<0.005, ***<0.001. Number of animals inside bars.

UNC-16 inhibits the regeneration-promoting function of DLK-1L

DLK is an essential MAPKKK that initiates regrowth of neuronal processes after injury and in growth cone formation in many model systems (Hammarlund et al. 2009; Yan et al. 2009; Xiong et al. 2010; Wang and Jin 2011; Shin et al. 2012). dlk-1 in C. elegans encodes two isoforms, DLK-1L (long) and DLK-1S (short), that have antagonistic functions. DLK-1S forms a heterodimer with DLK-1L and inactivates the regeneration-promoting role of DLK-1L (Yan and Jin 2012) (Supplementary Figure S4). JIPs are known to scaffold and regulate MAP kinases and in particular, the family member, JIP3, the mammalian orthologue of UNC-16, is known to form a complex with DLK (Kelkar et al. 2000; Holland et al. 2016). Hence, we investigated whether UNC-16 influences neuronal regeneration through a specific isoform of DLK-1.

Neuronal regeneration is dependent on DLK-1 levels (Hammarlund et al. 2009; Yan et al. 2009) and our observations could be influenced by variable expression, typical of extra-chromosomal transgenic lines. Hence we used animals with single-copy integrated transgenic expression of DLK-1S or DLK-1L made using the Mos-SCI technique (Yan and Jin 2012) to assess the effect of DLK-1S or DLK-1L in unc-16 mutants. Comparable to previous studies, single copy transgenic expression of DLK-1S or DLK-1L in wild-type animals, respectively suppresses or enhances neuronal regeneration even at 9 h post-axotomy (Figure 2B) (Yan and Jin 2012). The allele dlk-1(ju476) does not express DLK-1L but expresses endogenous levels of intact DLK-1S and is referred to as dlk-1(onlyS) and the allele dlk-1(tm4024) (a DLK-1 null) is referred to as dlk-1(0). These alleles are an useful tool to determine whether regeneration in unc-16 mutants is influenced by DLK-1S or DLK-1L.

To determine if DLK-1S and UNC-16 act independently to inhibit DLK-1L mediated regeneration, we assessed regeneration in the following genotypes: single copy DLK-1L with dlk-1(onlyS); unc-16(lf) and dlk-1(0); unc-16(lf). Single copy DLK-1L is introduced so that some regeneration occurs allowing us to assess the inhibitory effects of DLK-1S and UNC-16. unc-16(lf) mutants lacking DLK-1S show significantly higher regeneration compared to animals that contain DLK-1S, suggesting that DLK-1S and UNC-16 can both inhibit DLK-1L (Figure 2B).

To examine whether UNC-16 can act independently of DLK-1S, regeneration was assessed in animals expressing DLK-1L in dlk-1(0) or dlk-1(0); unc-16(lf). We observed greater regeneration in dlk-1(0); unc-16(lf) compared with dlk-1(0) and the difference was greatest at 6 h post-axotomy consistent with UNC-16’s early role in regeneration (Figure 2B). We observed that loss of unc-16 is unable to promote regeneration in animals expressing DLK-1S in wild type, dlk-1(onlyS) or dlk-1(0) mutant backgrounds (Figure 2B). However, animals expressing DLK-1L in dlk-1(onlyS); unc-16(lf) double mutants exhibited increased regeneration compared with animals expressing DLK-1L in dlk-1(onlyS) mutants alone (Figure 2B). In the presence of single copy of DLK-1S, the loss of UNC-16 leads to a twofold increase in regeneration in dlk-1(onlyS) mutants and a further 1.5-fold increase in regeneration in dlk-1(0) mutants (Figure 2B). This suggests that DLK-1S and UNC-16 act independently, to inhibit regeneration. However, in dlk-1(0) mutants, 6 h post-axotomy, these differences were not significant (Figure 2C), and this may be due to the larger amounts of DLK-1L available in the single copy DLK-1L line, soon after axotomy, resulting in faster initiation of regeneration.

Our data show that UNC-16 inhibits the growth-promoting activity of DLK-1L. We have shown that UNC-16 acts independently and in addition to DLK-1S to inhibit regeneration (Figure 2, B and C) and the inhibition by overexpression of DLK-1S is not suppressed by the unc-16(lf) mutation (Figure 2C).

UNC-16 changes the localization of DLK-1L

Vertebrate DLK, C. elegans GFP::DLK-1L, GFP::DLK-1S and endogenous UNC-16 are all shown to localize in punctae along the neuron (Yan and Jin 2012; Edwards et al. 2013; Holland et al. 2016). The punctate localization of vertebrate DLK has been shown to depend on palmitoylation-dependent binding of DLK to JIP3 (Holland et al. 2016). Therefore, we examined the localization of DLK-1L in both unc-16(tb109) and in neurons with increased levels of UNC-16.

The punctate localization of GFP::DLK-1L is reduced in all six unc-16 alleles examined with stops before and after the MAPK binding domain (Figure 3, A and B, Supplementary Figures S1A and S5). Further, upon UNC-16 over-expression, GFP::DLK-1L showed a significant increase in the intensity of puncta in wild type and in dlk-1(0) mutants. This indicates an increase in GFP::DLK-1L in each puncta in the presence of excess UNC-16, however, the density of the puncta remains unaffected at ∼2.5 ± 0.4/10 µm remaining similar to wild type (Figure 3, A and B). Our observations show that the punctate DLK-1L localization is dependent on the presence and levels of UNC-16 (Figure 3, A and B). We see that mutations in multiple places in UNC-16, both preceding and following the MAPK binding domain (Figure 1A, inset), lead to changes in localization of DLK-1.

Figure 3.

UNC-16 regulates DLK-1L localization in a dose-dependent manner. (A) Normalised intensity of DLK-1::GFP in different unc-16(tb109) mutants and animals expressing T7::UNC-16 (expressed under the jkk-1 promoter) in wild type and the null dlk-1(tm4024) mutants. Intensity of mCherry expressed form mec-4p::mCherry was used as a negative soluble marker control. (B) Representative images of GFP::DLK-1L localization quantified in (A) Scale bar 10 µm. (C) Normalized intensity of GFP::DLK-1L in the first 3 µm near the proximal cut end of an axotomized PLM touch neuron and ∼10 µm away from the injury site in wild type, T7::UNC-16 (expressed from jkk-1p::T7::unc-16) and unc-16(tb109) mutants. (D) Representative images of axotomized PLM touch neurons expressing GFP::DLK-1L, 4 min post-axotomy at and away from the site of axotomy. Arrow points to the proximal cut end with respect to cell body. Scale bar 5 µm. One-way ANOVA, P-value *<0.05, ***<0.001. Number of animals inside bars.

The punctate localization of C. elegans GFP::DLK-1S has been shown to depend on its association with DLK-1L (Yan and Jin 2012). GFP::DLK-1S, similar to GFP::DLK-1L, is not punctate in unc-16(lf) mutants (Supplementary Figures S6 and S7). Surprisingly, DLK-1S is punctate upon over-expression of T7::UNC-16 in the absence of DLK-1L, i.e., in a dlk-1(null) background (Supplementary Figures S6 and S7). This indicates that UNC-16 facilitates the punctate localization of DLK-1L and DLK-1S independent of each other, although our experiments cannot address whether DLK-1L and 1S are present together in these punctae.

Upon axotomy, DLK-1L is known to increase locally at the injury site while DLK-1S does not (Yan and Jin 2012). We examined whether the post-axotomy localization of DLK-1L is dependent on UNC-16 and whether the localization of DLK-1L along the axon is altered post axotomy. Comparable to the previous mentioned study, we observe that 4 min after axotomy, the levels of GFP::DLK-1L at the proximal end of the cut site increase (Figure 3, C and D). However, in unc-16(lf) mutants there is a greater accumulation of GFP::DLK-1L at the proximal cut site (Figure 3, C and D) compared to wild type at the same time point. Furthermore, overexpression of UNC-16 leads to a significant reduction of GFP::DLK-1L accumulation at the proximal cut site with the GFP::DLK-1L signal comparable to the GFP alone control (Figure 3, C and D). Thus, the presence and levels of UNC-16 controls the post-axotomy levels of DLK-1L at the cut site. In addition, the intensity of DLK-1L punctae along the neuronal process ∼10 µm away from the injury site is reduced (Figure 3C). The extent of reduction in intensity depends on the presence and levels of UNC-16 (Figure 3C).

The UNC-16-dependent localization of DLK-1L before injury and UNC-16-dependent control of DLK-1L levels post-injury at the cut site and along the neuronal process suggests that UNC-16 may inhibit DLK-1L function along the axon. The release of DLK-1L from the punctae may be stimulated by injury signals such as Ca⁺² known to regulate DLK-1L (Yan and Jin 2012).

Microtubule dynamics after injury in unc-16 mutants are partially dependent on DLK-1

DLK-1 is known to enable neuronal regeneration through two independent pathways, by elevating microtubule dynamics and by regulating transcription through CEBP-1 (Yan et al. 2009; Ghosh-Roy et al. 2010; Nix et al. 2011; Li et al. 2015; Tang and Chisholm 2016) (Figure 9D). We assessed microtubule dynamics using EBP-2::GFP, a marker for the growing plus end of microtubules (Ghosh-Roy et al. 2012).

In wild-type animals, the number of EBP-2::GFP comets pre-immediately and post-axotomy in a 40 µm region adjacent to the axotomy site are similar (Figure 4, A and C, Supplementary Videos S1 and S2). 3 and 6 h post-axotomy the number of EBP-2::GFP comets show a 2- and 2.5-fold significant increase (Figure 4, A and C) comparable to earlier reports (Ghosh-Roy et al. 2012). In wild-type animals, the length of the EBP-2::GFP comet tracks shows a peak increase at 2 h post-axotomy that remains elevated until 6 h post-injury (Figure 4, B and C). In dlk-1(0) mutant neurons, the number of EBP-2::GFP comets are fewer than in wild-type animals pre-axotomy, and do not significantly increase even 6 h post-axotomy (Figure 4C, Supplementary Table S1). We observe a similar trend where EBP-2::GFP track length increases 3 h post-axotomy but at 6 h post-axotomy, there is no change compared to the uninjured neurons of dlk-1(0) mutants.

Figure 4.

UNC-16 inhibits microtubule dynamics. (A) Number of EBP-2::GFP tracks in uncut and axotomized wild-type animals as a function of time, up to 3 h post injury. (B) Change in length of EBP-2::GFP tracks in uncut and axotomized wild-type animals across a 3 h post-injury period. Number of animals inside bars. One-way ANOVA with Bonferroni correction, P-value *<0.05, **<0.01, ***<0.001. (C) Numbers of EBP-2::GFP tracks in kymographs before and after axotomy in wild type, unc-16(tb109), dlk-1(tm4024) and double mutant dlk-1(tm4024); unc-16(tb109) in uncut and different time points post-axotomy. Two-way ANOVA with Bonferroni correction, P-value *<0.05, **<0.01, ***<0.001. Inset: Schematic of the touch neuron indicating the area of ablation and region of imaging. (D) EBP-2::GFP kymographs from uncut and 3 h post-injury neurons in wild type, dlk-1(tm4024), unc-16(tb109), and double mutant dlk-1(tm4024); unc-16(tb109). The length (x-axis) and time (y-axis) scales are 10 μm and 25 s, respectively. Red arrow indicates an EBP-2 trail.

We assessed the effect of unc-16(lf) on microtubule dynamics pre- and post-axotomy. We observe that there is an elevation in the number of EBP-2::GFP comets in uninjured neurons of unc-16(lf) mutant (Figure 4C, Supplementary Table S1) suggesting that UNC-16 downregulates microtubule dynamics. In uninjured neurons of the double mutant dlk-1(0); unc-16(lf) the number of EBP-2::GFP comets continues to have an 1.5-fold significant increase similar to that seen in the unc-16(lf) mutant, suggesting that the change in microtubule dynamics in unc-16 uninjured neurons is independent of DLK-1.

After injury the number of EBP-2::GFP comets show nearly a twofold increase 6 h in unc-16(lf) close to the 2.5-fold increase seen in wild-type axotomized processes at a similar time point (Figure 4C). The length of the EBP-2::GFP comet trails is increased in wild type, dlk-1(0), unc-16(lf) and dlk-1(0); unc-16(lf) mutants, 3 h post-axotomy (Supplementary Figure S8). In all the genotypes tested, there is no significant difference in the direction of microtubule growth compared to wild type (Supplementary Figure S9). Although regeneration in unc-16(lf) mutants is completely dependent on DLK-1 (Figure 2A), microtubule dynamics in unc-16(lf) mutants are only partially dependent on DLK-1 after axotomy. Since regeneration in unc-16(lf) mutants is fully DLK-1 dependent, it suggests that factors other than microtubules could contribute to DLK-1 dependent regeneration in unc-16(lf) mutants.

Actin dynamics and growth cone protrusions post-injury in unc-16 mutants are dependent on DLK-1

Actin, the actin depolymerizing factor (ADF)/cofilin A and nonmuscle Myosin II that powers actin’s retrograde flow, are all important for growth cone formation and for the extension of both developing and regenerating vertebrate neurons (Hur et al. 2011; Flynn et al. 2012; Chisholm et al. 2016; Blanquie and Bradke 2018; Tedeschi et al. 2019). Thus, we followed actin dynamics by using a transgenic line that expresses the GFP tagged calponin homology domain of F-actin-binding protein Utrophin (GFP::utCH) (Burkel et al. 2007; Chia et al. 2014; Sood et al. 2018).

The number of GFP::utCH trails in wild-type neurons increases significantly, 3 h post-axotomy, in both anterograde and retrograde directions (Figure 5B). In the 40 µm region proximal to the injury site, the length of all GFP::utCH trails also increases 2 h post-axotomy (Figure 5A, Supplementary Videos S3 and S4). In wild-type neurons, the density of actin-rich regions reduces significantly 3 h post-axotomy (Figure 5C). Unlike microtubule dynamics that show regional changes (Ghosh-Roy et al. 2012), actin dynamics change throughout the proximal injured neuron 3 h post-axotomy. We observe an increase in the length of all GFP::utCH trails (Figure 6A) and a decrease in actin-rich regions throughout the proximal injured axon (Figure 6B). The density of actin-rich regions reduces the most in the proximal 30–60 µm next to the axotomy site 3 h post-axotomy (Figure 6B). The number of GFP::utCH trails increases the most 60–90 µm away from the site of axotomy (Figure 6C).

Figure 5.

Temporal changes in actin dynamics in the injured neuronal process. (A) Number of GFP::utCH trails in wild type uncut neurons and 40 µm adjacent to the injury site of wild-type neurons up to 3 h post-injury. (B) Change in length of GFP::utCH trails in anterograde and retrograde directions in neurons and 40 µm adjacent to the injury site of wild-type neurons up to 3 h post-injury. (C) Comparison of the density of stationary GFP::utCH stationary clusters in neurons of uncut and axotomized wild-type animals over time period of 3 h post injury. Number of animals inside bars. One-way ANOVA with Bonferroni correction, P-value **<0.01, ***<0.001. (D) Representative kymographs of GFP::utCH in wild type, dlk-1(tm4024), unc-16(tb109), the double mutant dlk-1(tm4024); unc-16(tb109) and cebp-1 (tm2807) mutants, uninjured and 3 h post injury in PLM touch neuron. Blue arrows indicate stationary utCH clusters and green arrows indicate a utCH trail. The length (x-axis) and time (y-axis) scales are 10 μm and 25 s, respectively.

Figure 6.

Actin dynamics are altered in the entire proximal axotomized neuronal process. (A) Change in length of GFP::utCH trails in uncut and axotomized wild-type animals at 3 h post injury over the entire section of the neuron in anterograde and retrograde direction. Inset: Schematic of the touch neuron indicating the area of ablation and regions of imaging. (B) Comparison of the density of stationary GFP::utCH in uncut and axotomized wild-type animals 3 h post injury in various sections of the neuron. (C) Number of GFP::utCH trails in uncut and axotomized wild type 3 h post injury over various sections of the neuron in anterograde and retrograde direction. Number of animals inside bars. Two-way ANOVA with Bonferroni correction (A, B, D, and F), one-way ANOVA with Bonferroni correction (C, E), P-value *<0.05, ***<0.001.

The number and length of anterograde and retrograde actin trails in uninjured neurons of wild type and dlk-1(0) mutants is comparable (Figure 7, B and C). The density of actin-rich regions in uninjured dlk-1(0) mutant neurons is similar to that seen in wild type (Figure 7A). Three hours post-axotomy, there is no statistically significant change in density of actin-rich regions, numbers, or lengths of actin trails in dlk-1(0) mutant neurons compared to dlk-1(0) uninjured neurons (Figure 7, Supplementary Table S2). This shows that changes in actin dynamics post-injury are dependent on DLK-1.

Figure 7.

Actin dynamics in unc-16 mutants are dependent on DLK-1 and CEBP-1. (A) Comparison of the density of stationary GFP::utCH in uncut and 3 h post injury neurons in wild type, unc-16(tb109), dlk-1(tm4024) and cebp-1(tm2807) and double mutants dlk-1(tm4024); unc-16(tb109) and cebp-1(tm2807); unc-16(tb109). (B) Number of GFP::utCH trails in uncut and 3 h post injury neurons in various genotypes in the anterograde and retrograde directions. (C) Comparison of the length of GFP::utCH trails in uncut and 3 h post injury neurons in various genotypes in the anterograde and retrograde direction. Number of animals inside bars. One-way ANOVA with Bonferroni correction (A), two-way ANOVA with Bonferroni correction (B, C), P-value *<0.05, **<0.01, ***<0.001.

We then assessed the GFP::utCH dynamics in unc-16(lf) mutant touch receptor neurons. The number of GFP::utCH trails in uncut unc-16(lf) mutants is similar to that seen in wild-type animals (Figure 7B). However, the length of the GFP::utCH trails increased fivefold in uninjured unc-16(lf) mutant neurons compared to wild type (Figure 7C). The density of GFP::utCH marked actin-rich regions is reduced by ∼50% in uninjured unc-16(lf) mutant neurons compared to wild type, perhaps contributing to the polymerizing F-actin pools in the neuronal process seen in this genotype (Figure 7A).

In unc-16(lf) mutant neurons 3 h after axotomy, the numbers of trails are ∼1.3-fold greater compared to wild type and nearly twofold higher than actin trails seen in uninjured unc-16(lf) mutant neurons (Figure 7B). Three hours post-injury, the length of the GFP::utCH trails in unc-16(lf) mutant neurons is similar to that seen in wild-type neurons at the same time point (Figure 7C). In all the genotypes tested, the increase in GFP::utCH trails occurs in both anterograde and retrograde directions. Interestingly 3 h post-axotomy, unc-16(lf) mutant neurons showed no significant reduction in density of actin-rich regions compared to uninjured unc-16(lf) neurons (Figure 7A). By contrast, in wild-type neurons there is a significant reduction in density of actin-rich regions post-axotomy.

Neurons in dlk-1(0); unc-16(lf) double mutants were indistinguishable from the dlk-1(0) mutant neurons and show no increase in number, length, or density of GFP::utCH trails in both uninjured neurons or 3 h post axotomy (Figure 7) suggesting that the actin dynamics in unc-16(lf) mutant neurons are dependent on DLK-1 both before and after injury.

We also investigated whether the altered actin dynamics changes the structure of the neuronal growth cone. Hence, we counted the number of filipodial protrusions in growing post-axotomized neurons from wild type, unc-16(lf), dlk-1(0), and dlk-1(0); unc-16(lf) mutant animals. We observe that in unc-16(lf) mutants, the genotype with greatest increase in actin dynamics pre- and post-axotomy, also shows an increase in the number of filipodial protrusions after injury compared to all remaining genotypes examined (Supplementary Figures S10 and S11). This elevation in filipodial protrusions in unc-16(lf) mutants begins 6 h post-axotomy (Supplementary Figure S12). Wild-type neurons show greater number of protrusions at the growth cone compared to neurons from dlk-1(0) mutants (Supplementary Figure S11). In dlk-1(0); unc-16(lf) double mutants, the number of filipodial protrusions are similar to the number seen in dlk-1(0) mutant neurons (Supplementary Figure S11). Thus, increased actin dynamics pre- and post-axotomy may alter the growth cone morphology. Actin dynamics likely make significant contributions to the increased regeneration in unc-16 mutants, lack of regeneration in dlk-1 mutants and the lack of regeneration in dlk-1; unc-16 double mutants.

Post-axotomy actin and microtubule dynamics in unc-16 mutants partially depend on CEBP-1

DLK-1 is known to affect neuronal regeneration by promoting translation and stabilization of CEBP-1 mRNA in touch neurons (Yan et al. 2009) (Figure 9D). CEBP-1 along with another transcription factor ETS-4 activates other growth factors necessary for neuronal outgrowth during regeneration (Li et al. 2015). Thus, we examined if the microtubule and actin dynamics seen in unc-16 mutants also depend on CEBP-1.

We measured EBP-2 dynamics in wild type, unc-16(lf), cebp-1(0) and unc-16(lf); cebp-1(0) mutants. Uninjured neurons in unc-16(lf) mutants have an increased number of EBP-2 comets (Figure 8B) compared to wild-type animals while cebp-1(0) and unc-16(lf); cebp-1(0) mutants are similar to wild type. This suggests that microtubule dynamics in uninjured UNC-16 neurons are dependent on CEBP-1.

Figure 8.

CEBP-1 regulates post-injury microtubule dynamics and regeneration. (A) Change in length of EBP-2::GFP tracks in uncut and axotomized wild-type animals, unc-16(tb109), dlk-1(tm4024) and cebp-1(tm2807) and cebp-1(tm2807); unc-16(tb109) 6 h post injury. Number of animals inside bars. (B) Numbers of EBP-2::GFP tracks in kymographs before and after axotomy in wild type and other mutants in uncut and 6 h post injury. One-way ANOVA with Bonferroni correction, P-value *<0.05, **<0.01, ***<0.001. (C) Neuronal regeneration in wild type, unc-16(tb109), cebp-1(tm2807), and cebp-1(tm2807); unc-16(tb109). One-way ANOVA P-value *<0.05, **<0.005, ***<0.001. Number of animals inside bars.

Post-axotomy, the cebp-1(0) mutant neurons do not show any significant change in the number of EBP-2 comets (Ghosh-Roy et al. 2012) (Figure 8B). After injury, unc-16(lf); cebp-1(0) mutant neurons show a significant decrease in the number of EBP-2 comets, compared to unc-16(lf) mutants (Figure 8B). All genotypes show an increase in EBP-2 comet growth length 6 h post-axotomy (Figure 8A). Thus, in unc-16(lf) mutants, basal microtubule dynamics are fully dependent on CEBP-1 but the elevation in microtubule dynamics post-injury is only partially dependent on CEBP-1.

We also examined whether actin dynamics post-injury are dependent on CEBP-1. Actin dynamics in cebp-1(0) mutant neurons are indistinguishable from dlk-1(0) mutant neurons pre- or post-axotomy (Figure 7). Post-axotomy the number of GFP::utCH trails in neurons from unc-16(lf); cebp-1(0) mutants is significantly lower than that seen in unc-16(lf) mutants and significantly higher than that seen in cebp-1(0) mutants (Figure 7B, Supplementary Table S3). The length of the GFP::utCH trails in the cebp-1(0) mutants post-axotomy is greatly reduced compared to wild type and stays unchanged in the unc-16(lf); cebp-1(0) double mutants (Figure 7C). We conclude that regulation of actin dynamics in unc-16 mutants occurs partially through CEBP-1, suggesting that actin regulation can occur independently of changes in microtubule dynamics, potentially through CEBP-1’s action on regulators of actin.

cebp-1(0) mutants show reduced regeneration compared to wild type (Figure 8C). We observed that despite complete absence of CEBP-1, the unc-16(lf); cebp-1(0) double mutants show a twofold higher regeneration compared to cebp-1(0) mutants, 9 hr post-axotomy (P-value 0.04) (Figure 8C). The partial regeneration seen in unc-16(lf); cebp-1(0) double mutants likely arise from the continued small increase in microtubule and actin dynamics seen in these animals.

Functional recovery of the regenerating neuron

Functional regeneration post-axotomy has been observed in C. elegans motor and touch neurons (Ben-Yakar et al. 2009; Abay et al. 2017; Basu et al. 2017; Ding and Hammarlund 2018). More specifically, the axonal fusion of the proximal and distal ends of touch neurons post-injury has been shown to be responsible for the observed functional regeneration in touch neurons (Abay et al. 2017; Basu et al. 2017). Since loss of UNC-16 causes faster regrowth, we examined whether it led to quicker functional regeneration.

The bilaterally symmetric PLM neurons mediate a gentle touch escape response and is represented as PTRI (Basu et al. 2017). In the wild-type worms, 3 h post-axotomy, there is a significant decrease of the touch response on the axotomized side compared to the response on the side of animals with uncut PLM neuron (Figure 9A). This lowered PTRI recovered significantly 24 h post injury and about 38% of the injured wild-type PLM neurons show fusion between the proximal and distal processes (Figure 9, B and C).

In unc-16(lf) mutants, upon axotomy, there is a significant decrease in touch response on the axotomized neuron side compared to the touch responsiveness on the uninjured side leading to a lower PTRI (Figure 9A). However, there is no significant increase in PTRI from 3 to 24 h after injury, thus there is an absence of significant functional recovery after injury despite greater neuronal outgrowth (Figure 9A). Concurrently, we observe that only 14.5% of the injured neurons show fusion between the distal and proximal ends (Figure 9B). In animals expressing T7::UNC-16 we found a significant decrease in PTRI 3 h after injury compared to the uncut worms of the same genotype (Figure 9A). After 24 h there is a significant increase in PTRI from 3 h and we observe about 26.3% fusion at this time point (Figure 9B). Although the number of neurons that show fusion is slightly lower in T7::UNC-16, it is not statistically significant compared to wild type (P-value of 0.13), and the PTRI at 24 h is comparable to wild type.

Taken together these data show that although there is faster growth in unc-16(lf) mutants, it does not translate into increased functional regeneration. In contrast, in animals overexpressing UNC-16, though the growth rate is significantly reduced, slower regrowth is observed to support significant functional recovery 24 h post-axotomy. Thus, rapid rates of regeneration in an injured axon may not necessarily lead to better functional recovery. Furthermore, excessive sprouting in unc-16(lf) mutants (Supplementary Figures S10 and S11) may reduce fusion events important for functional recovery.

Discussion

Our study shows that UNC-16 plays an inhibitory role on regeneration where its presence and levels act early to control both growth initiation and regrowth rate. These effects are dependent on the MAPKKK DLK-1, essential for regeneration in multiple models (Hammarlund et al. 2009; Itoh et al. 2009; Yan et al. 2009; Xiong et al. 2010; Nix et al. 2011; Shin et al. 2012). There are few known inhibitors that act directly on DLK, the master regulator of regeneration (Hammarlund et al. 2009; Valakh et al. 2015; Chisholm et al. 2016; Asghari Adib et al., 2018). Our study suggests that UNC-16/JIP3 acts by making DLK-1 unavailable along the neuronal process to control the levels of DLK-1 at the injury site. This study identifies the role of DLK-1-dependent actin dynamics as an important mediator of axonal regeneration. In addition, UNC-16 dependent microtubule dynamics are regulated partially by CEBP-1 in both uninjured and injured neurons. Our data suggest that UNC-16 acts like a clamp on cytoskeletal dynamics, negatively regulating both microtubule and actin dynamics. Intriguingly, the rapid regrowth, at least partially mediated by altered cytoskeletal dynamics, does not promote functional recovery. Our study identifies a key role for UNC-16 in balancing the rate of axon regrowth to enable functional recovery.

Systematic genetic screens in C. elegans identified that only ∼10% of genes inhibit neuronal regeneration while 83% of genes identified in the screen promote neuronal growth post-injury (Chen et al. 2011). UNC-16/JIP3 is a negative regulator that acts on DLK-1L independently and in addition to the inhibitory isoform DLK-1S (Figure 3C) (Chen et al. 2011; Nix et al. 2014). Interestingly, JIP3 through its binding ability with Kinesin-1 has been shown to be necessary for axonal regeneration in sensory neurons (Watt et al. 2015). However, our data suggest that in the PLM touch neuron, Kinesin-1 by itself or in conjunction with UNC-16, does not appear to significantly influence axon regeneration (Supplementary Figure S2). UNC-16 was also identified as a negative regulator of regeneration in a screen for motor neuron regeneration in C. elegans (Chen et al. 2011; Nix et al. 2014). JIP3, by the virtue of being a scaffold protein is capable of binding to multiple regulators in the process of regeneration and thereby could play pivotal roles in regulating neuronal regeneration (Watt et al. 2015; Zeke et al. 2016).

Although vertebrate JIP3 is also known to form a complex with DLK (Kelkar et al. 2000; Holland et al. 2016), the effects of UNC-16 on DLK-1L in C. elegans may not be direct since several unc-16 alleles independent of their predicted MAPKKK binding ability still influence both DLK-1 localization and regeneration (Figures 1A and 3, A and B). Apart from JIP3, DLK-1S is the only known inhibitor of DLK-1L (Yan and Jin 2012). Although the inhibitory DLK-1S like isoform has been suggested to be present in humans, its presence has not been verified in other organisms (Supplementary Table S4). Other identified negative regulators of regeneration like DIP-2 and NMAT-2 act through other parallel pathways and not directly on DLK-1 (Kim et al. 2018; Noblett et al. 2019). Thus, UNC-16 and its vertebrate homolog JIP3 may act as a general inhibitor of DLK mediated signaling in multiple species by altering its localization in an injury-dependent manner. UNC-16 may merely control DLK-1 availability at the injury site rather than regulate its kinase activity.

Cytoskeletal damage is thought to trigger DLK activity that subsequently initiates regeneration post injury (Valakh et al. 2015). Despite differences in pathways downstream of DLK-1 in regulating both microtubule and actin dynamics, regrowth after injury is likely to depend on the interaction between the actin and microtubule cytoskeletons during growth cone formation (Dogterom and Koenderink 2019). Our data suggest that the initiation of regrowth may require both microtubule and actin dynamics and both dynamics change at similar time scales after injury (Figures 4–7). DLK-1 is known to regulate local microtubule dynamics close to the injury site through the depolymerizing kinesin-13 family member, KLP-7 (Ghosh-Roy et al. 2012). Thus, some of the increased microtubule dynamics observed in unc-16 mutants may be dependent on KLP-7. In addition, unc-16 mutants also show increased microtubule dynamics partially dependent on DLK-1 (Figures 4C and 7A). However, the dependence of increased microtubule dynamics in unc-16 mutants on KLP-7 remains to be investigated.

Cytoskeletal dynamics are increased in unc-16(lf) mutants. The increased microtubule dynamics persist post-axotomy but are largely independent of DLK-1. By contrast, actin dynamics in unc-16(lf) mutants are also increased but are fully dependent on DLK-1. Because regeneration in unc-16(lf) mutants is dependent on DLK-1 (Figure 2A), regulation of actin dynamics via DLK-1 may be critical for the greater regeneration seen in unc-16(lf) mutants. However, the increased microtubule dynamics in unc-16 mutants may also play a DLK-1-independent role in regeneration, for instance in controlling the rate of axon outgrowth.

Actin is thought to be important in the regenerating growth cone (Hur et al. 2011; Flynn et al. 2012; Chisholm et al. 2016; Blanquie and Bradke 2018; Tedeschi et al. 2019). Overexpression of the ADF/cofilin has been shown to induce regeneration in culture as well as in vivo following spinal cord injury (Tedeschi et al. 2019). Inhibition of nonmuscle myosin II, which powers retrograde actin flow in the growth cone, markedly enhances axon growth by reorganizing the growth cone cytoskeleton in cultured vertebrate neurons (Hur et al. 2011). The altered actin dynamics seen in unc-16 mutants may occur through molecules such as Cofilin, Myosin II and actin polymerizers such as Formin and ARP2/3 (Hur et al. 2011; Flynn et al. 2012; Breitsprecher and Goode 2013; Chisholm et al. 2016; Blanquie and Bradke 2018; Tedeschi et al. 2019).

Our data suggest that UNC-16 controls actin dynamics partially through CEBP-1 (Figure 7). The transcription factor C/EBP, in other models, is known to be dependent on MAP kinases other than DLK-1 (Yamamoto et al. 1999). In unc-16(lf); cebp-1(0) double mutants microtubule dynamics are reduced. This suggests that there is a pathway by which CEBP-1 regulates basal microtubule dynamics. Actin dynamics are abrogated in cebp-1(0) mutants and remains so in unc-16(lf); cebp-1(0) double mutants (Figure 7). There is also a modest increase in number of actin trails compared to cebp-1(0) mutants (Figure 7B). Therefore, regeneration is higher than that seen in cebp-1(0) mutants but lower than observed in unc-16(lf) mutants. Thus, CEBP-1 has effects on both actin dynamics and on basal levels of microtubule dynamics in unc-16(lf) mutants.

Based on our findings we think that there are DLK-1 independent and dependent pathways for microtubule dynamics (Figure 9D). The DLK-1-dependent pathway may act through CEBP-1. Actin dynamics are completely dependent on DLK-1 and partially dependent on CEBP-1, while microtubule dynamics are partially dependent on DLK-1 and CEBP-1. A basal level of microtubule dynamics is also controlled through CEBP-1. Cytoskeletal dynamics is controlled in a complex manner where basal cytoskeletal dynamics as well as post-injury cytoskeletal dynamics are controlled partially through DLK-1 and CEBP-1 but also independent of them. Our data suggest a complex regulation of actin and microtubule dynamics might be responsible for controlling rate of regeneration. It also highlights the importance of UNC-16 in the spatial and temporal control of DLK-1.

Based on prior studies and our findings, we propose a model where UNC-16/JIP3 plays its inhibitory role through tight temporal and spatial control of DLK-1 function (Figure 9D). The dual inhibitory control by both UNC-16 and DLK-1S balance the intrinsic growth-promoting function of DLK-1L in vivo. This intrinsic growth ability is likely achieved through a balance between the actin dynamics and microtubule stabilization at the growth cone that could also mediate rapid steering of the growth cone in a regenerating neuron (Hur et al. 2012; Blanquie and Bradke 2018). Our study reveals the role UNC-16 plays in the complex control of both, basal and post-injury cytoskeletal dynamics.

Axonal injury breaks the connection between the neuronal cell body and its post-synaptic targets disrupting its functions and potentially the behaviors that depend on the neuron (He and Jin 2016). In C. elegans there are multiple reports of partial recovery of locomotion following axotomy in multiple motor neurons (Yanik et al. 2004; Neumann et al. 2011; El Bejjani and Hammarlund 2012; Byrne et al. 2016).

We show that the faster regeneration seen in unc-16(lf) mutants does not lead to functional recovery known to depend on fusion of the distal and proximal parts of the injured neuron (Abay et al. 2017; Basu et al. 2017). The faster neuronal growth observed in unc-16(lf) mutants may not be favorable for fusion as it may adversely affect other required steps in restoring function, e.g., cargo transport that is known to depend on UNC-16/JIP3 (Verhey et al. 2001; Sun et al. 2011). The efa-6 mutant that shows faster regrowth and increased microtubule dynamics also does not show enhanced functional recovery (Chen et al. 2015; Basu et al. 2017). Furthermore, slower regeneration seen when UNC-16 is over-expressed shows functional recovery comparable to wild-type animals (Figures 1D and 9A). Thus, merely a faster rate of neuronal outgrowth is insufficient for functional recovery, highlighting the importance of mechanisms that modulate the growth rate of the regenerating neuron to allow for recovery of function.

Data availability

Strains and plasmids are available upon request. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables.

Supplementary material is available at GENETICS online.