wrk-1 and rig-5 control pioneer and follower axon navigation in the ventral nerve cord of Caenorhabditis elegans in a nid-1 mutant background

Abstract

During nervous system development, neurons send out axons, which must navigate large distances to reach synaptic targets. Axons grow out sequentially. The early outgrowing axons, pioneers, must integrate information from various guidance cues in their environment to determine the correct direction of outgrowth. Later outgrowing follower axons can at least in part navigate by adhering to pioneer axons. In Caenorhabditis elegans, the right side of the largest longitudinal axon tract, the ventral nerve cord, is pioneered by the AVG axon. How the AVG axon navigates is only partially understood. In this study, we describe the role of two members of the IgCAM family, wrk-1 and rig-5, in AVG axon navigation. While wrk-1 and rig-5 single mutants do not show AVG navigation defects, both mutants have highly penetrant pioneer and follower navigation defects in a nid-1 mutant background. Both mutations increase the fraction of follower axons following the misguided pioneer axon. We found that wrk-1 and rig-5 act in different genetic pathways, suggesting that we identified two pioneer-independent guidance pathways used by follower axons. We assessed general locomotion, mechanosensory responsiveness, and habituation to determine whether axonal navigation defects impact nervous system function. In rig-5 nid-1 double mutants, we found no significant defects in free movement behavior; however, a subpopulation of animals shows minor changes in response duration habituation after mechanosensory stimulation. These results suggest that guidance defects of axons in the motor circuit do not necessarily lead to major movement or behavioral defects but impact more complex behavioral modulation.

Introduction

Correct development of any nervous system involves precise navigation of developing axons as they navigate through the surrounding substrate. Early axons use environmental guidance cues to navigate, pioneering pathways that later axons can follow. This “pioneer” and “follower” relationship is highly conserved and is seen in a wide variety of organisms including grasshoppers, zebrafish, and mammals, (Klose and Bentley 1989; Chitnis and Kuwada 1991, p. 199; Rash and Richards 2001). Follower–pioneer adhesion leads to the development of fasciculated nerve bundles within which axons can build en passant synapses, a synapse form that is common in a wide range of organisms including mammals, amphibians, and invertebrates like C. elegans (White et al. 1986; Roberts et al. 1999; McAllister 2007; Wong et al. 2012).

Many genes that regulate axon guidance have been identified in a variety of organisms including C. elegans. Among these are members of the immunoglobulin superfamily (IgCAMs), a diverse family of highly conserved membrane-bound proteins defined by extracellular immunoglobulin domains. While the many roles of IgCAMs in follower axon guidance have been analyzed by this and other groups (Katidou et al. 2008; Schwarz et al. 2009), little is known of their roles in pioneer axon guidance.

The ventral nerve cord (VNC) of C. elegans is the major anteroposterior axon pathway of the organism, consisting of a smaller left and a larger right axon tract. The right tract is pioneered by the AVG axon, which extends posteriorly from the retrovesicular ganglion (rvg) at the anterior end of the VNC into the tail region (Fig. 1). Later, command interneuron (CI) axons and motoneuron neurites extend along the path first forged by the AVG axon (White et al. 1986). When AVG is ablated before its axon extends, these following neurites make migrational errors as they extend (Durbin 1987; Hutter 2003). Similar follower navigation defects are seen in lin-11 mutants (Hutter 2003). LIN-11 is a transcription factor required at a late stage in the differentiation of several neurons including AVG (Hobert et al. 1998; Sarafi-Reinach et al. 2001). The VNC axon guidance defects observed in lin-11 mutants are mostly attributable to the absence of a functional pioneer in the right VNC (Hutter 2003).

Fig. 1.

Schematic representation of relevant neurons and axons in the ventral nerve cord (VNC). RIFR and RIFL axons pioneer the path from the retrovesicular ganglion (rvg) toward the nerve ring in the head. The AVG axon pioneers the right VNC from the rvg, and the PVPR axon pioneers the left VNC from the posterior. CI axons from the nerve ring follow RIFR and RIFL into the rvg, then follow the AVG to continue posteriorly along the right VNC. Motoneurons (DD and VD, representing two of many cells in each class) extend neurites anteriorly and posteriorly within the right VNC after the AVG axon has begun extending. A, anterior, P, posterior.

Within the right VNC fascicle, CI synapse with each other and with motoneurons en passant (White et al. 1986), forming the circuits that will later regulate movement and behavior (Wicks and Rankin 1995; Von Stetina 2006). C. elegans exhibit a wide variety of spontaneous and elicited behavioral responses and learning (Ardiel and Rankin 2010). Incorrect VNC pathfinding may disrupt the formation of the en passant synapses that coordinate essential behaviors.

Given the crucial role AVG plays in setting the groundwork for the development of the right VNC, it is likely that several partially redundant pathways have evolved to ensure correct AVG axon navigation. This redundancy may explain difficulties researchers face when trying to identify genes regulating AVG axon extension (Hutter 2003; Moffat et al. 2014; Bhat et al. 2015). To surmount these obstacles, we previously conducted an enhancer screen in a sensitized nid-1/Nidogen mutant background to identify novel genes regulating AVG axon guidance (Bhat and Hutter 2016). Nidogen is an evolutionarily conserved basement membrane protein that may facilitate the localization of guidance factors along the developing VNC. nid-1/Nidogen null mutants exhibit a low penetrance of AVG axon guidance defects but are otherwise healthy (Kim and Wadsworth 2000) facilitating the identification of genes controlling AVG axon navigation. Two genes identified in this screen are aex-3 (Bhat and Hutter 2016) and ccd-5 (Feresten et al. 2022), both of which genetically interact with a series of vesicle-associated genes (rab-3, unc-18, ida-1) and unc-5 (the netrin receptor). Here, we expand the set of genes required for AVG axon navigation in a nid-1 mutant background, adding the IgCAMs wrk-1 and rig-5.

wrk-1 nid-1 and rig-5 nid-1 double mutants have substantial AVG and follower axon navigation defects. We find that follower CI and DD/VD motoneuron axons are more likely to cross the midline with the AVG pioneer axon in the double mutants. wrk-1 and rig-5 act in different genetic pathways, and wrk-1 is in a genetic pathway with vab-1, the sole C. elegans Eph receptor. Behavioral studies revealed no significant defects in free movement but subtle defects in the tap response habituation of a subpopulation of rig-5 nid-1 double mutants. Taken together, our study identifies wrk-1 and rig-5 as members of two different pathways controlling AVG axon navigation in the absence of nid-1.

Materials and methods

Nematode strains and alleles used

The following marker strains were used for phenotypic analysis: otIs182[inx-18::GFP] IV for AVG; hdIs30[glr-1::dsRed2] I and hdIs32[glr-1::dsRed2] for CI; hdIs24[unc-129::CFP, unc-47::DsRed2] for DD/VD. The following alleles were used for phenotypic analysis: rig-5(hd48), wrk-1(ok695), nid-1(cg119), unc-18(e234), cdk-5(gm336), vab-1(dx31), and ccd-5(gk5256). All alleles in this study are expected to be null alleles. Details for genotyping mutant strains are included in Supplementary Table 1. All strains were cultured and maintained at 20°C under standard conditions (Brenner 1974). Double and triple mutants were constructed as described (Iwasaki et al. 1997).

Phenotypic analysis of neuronal defects and microscopy

Animals were immobilized with 10 mM sodium azide in an M9 buffer for 1 h and mounted on 2% agar pads before analysis. Axonal defects were scored with a Zeiss Axioscope (40× objective) in adult animals expressing fluorescent markers in respective neurons. Only animals positioned ventral side up were scored, so that the entire VNC was visible. Normal VNCs were tallied, while VNCs with axons that deviated from their known path were drawn by hand onto a schematic as accurately as possible. Any animals in which the VNC appeared distorted by eggs or degenerating, or where the researcher was not confident that the VNC was anatomically normal, were excluded from the analysis. Researchers were not blinded when scoring, and scoring occurred across multiple days and times. The drawings were scored en masse once desired N had been reached and could be rescored if necessary. Images were obtained using a Nikon digital camera through the Zeiss Axioscope 40× objective and assembled using ImageJ and Microsoft Powerpoint.

Behavioral methods

Habituation assays

Basic locomotion and habituation responses to mechanosensory tap stimuli of C. elegans strains were assessed using the Multi-Worm Tracker, MWT (Swierczek et al. 2011). Petri plates containing animals for the behavioral experiments were prepared by evenly spreading 50 μl of E. coli OP50 on the plate using a sterilized Pasteur pipette, 48 h before tracking. Age-synchronized populations of each genotype were acquired by bleaching 15–30 egg-bearing gravid animals onto the Petri plate. This method generated 60–120 animals per plate. Plates were then parafilmed and stored on a vibration-proof suspension shelf in a temperature and humidity-controlled room (20°C and 40%, respectively) until the tracking experiment 4 days (∼96 h) later.

At the start of each tracking experiment, a plate was mounted onto the tracking platform, and a plate lid was gently put on. Once the lid was on, the tracking session started. Worms moved freely on the Petri plate for 600 s, during which morphological and locomotor data were collected. Starting from 600 s, 30 mechanosensory stimuli were delivered by a solenoid tapping to the side of the Petri plate every 10 s, and tap response data were collected. Each tracking session was terminated 10 s after the last tap. During the entire tracking session, animals’ behavior was recorded at a rate of ∼20 frames per second. Tracking of different strains was rotated to minimize the impact of time-sensitive factors. Researchers were not blinded, as the MWT uses machine-vision to record behavior objectively. Responses were averaged for each of 13 plates per strain, and plates for which 10 or more of the 30 tap responses fell outside of 2 SD were excluded from analysis. For each tap, we assessed probability, reversal speed, and reversal duration. The MWT software (version 1.2.0.2) was used to record animals’ behavior, deliver stimuli.

The MWT data were first analyzed using Choreography (Swierczek et al. 2011). In Choreography (version 1.3.0_r1035), a “–shadowless, –minimum-move-body 2, –minimum-time 20” filter was applied to restrict the analysis to animals that traveled at least two body lengths and were tracked for at least 20 s. The MeasureReversal plugin was used to identify responses occurring within 1 s (dt = 1) of stimulus onset. Custom Matlab scripts were used to organize the data and compute population statistics. Animals’ locomotion was binned together every 10 s, and the proportions of time worms spent in forward locomotion and average locomotion speed were calculated.

Animals typically respond to mechanosensory taps by initiating an episode of backward locomotion before they change direction and move forward again. We assessed the probability, duration, and speed of the reversal response to taps. Habituation scores for response probability, duration, and speed were calculated by the percent change from initial response level to the final response level, represented by an average of the last three responses.

Identification of subpopulations in the habituation assays

For each strain, naïve response (response to the 1st stimulus) duration and final response (responses to the 28th–30th stimuli) for all individual worms from each of 10–12 plates were grouped together, respectively. Outliers that were two SD above the overall means were removed (for all strains, <3% outliers were removed; there were no outliers below the mean). Histograms of response duration for each strain were plotted with a bin size of 0.1 s. To evaluate the population distributions, we first applied a nonparametric, Kernel-based fit-curve method, fitdist (dataset, “Kernel”), to visualize the shape of the response duration distribution. This technique relies on moving averages of the raw data instead of creating and testing a predictive model of the data. In four of five strains, the corresponding fit curve of the histograms had two peaks. We then assumed that a two-peaked distribution is composed of two normally distributed subpopulations, each with a different mean (μ),SD (σ), and proportion contribution factor (φ). The combined distribution of each histogram can be described as follows:

• combined_distribution = φ₁ * normal_distribution(μ1, σ₁) + φ₂ * normal_distribution(μ₂, σ₂);

with, μ₁, σ₁, and μ₂, σ₂ determining the shapes of the two normally distributed subpopulations, and φ₁ and φ₂ describing the proportion of each subpopulation contributing to the combined population distribution with φ₁ + φ₂ = 1.

To determine what parameters best define each combined population distribution, we used a maximum likelihood estimation method. The parameter estimate analysis began with an initial assumption of an even distribution between two identical normal curves, followed by a series of adjustments based on our data sets. The parameter starting values of φ₁ and φ₂ were 0.5; the starting values of μ₁ and μ₂ were assigned the values of the 1st and 3rd quartiles of the range of the data; the starting values of σ₁ and σ₂ were determined by the formula for the variance of a combined distribution in terms of the mean and variance of each component, calculated as such:

$${\sigma }_1={\sigma }_2=\surd (varianceofthecombineddistribution-0.25*({\mu }_2-{\mu }_1{)}^2)$$

Then, a maximum likelihood estimates function, mle(), was used with a maximum of 300 iterations (‘MaxIter’ = 300). For each parameter, upper and lower boundaries were defined, such that the mle() function finds estimated parameters within realistically meaningful boundaries ([0, 1] for p, [− ∞, +∞] for μ, and [0, + ∞] for σ).

In each iteration, the mle() function: (1) obtains a set of parameters, (2) calculates the probability of occurrence of each data point with this set of parameters, (3) computes the log-likelihood function by summing the logarithms of the probabilities, (4) compiles the parameters and the log-likelihood value into a list, and (5) updates the estimated parameters with corrections. Each iteration of the function repeats the above steps to generate a log-likelihood function for a set of parameters, and iterations are terminated after the 300th iteration. The mle() function selects from the list the parameters that result in the max log-likelihood and returns the estimated parameters for the combined distribution for each strain (Supplementary Table 2). Curves plotted using the estimated parameters were reasonably accurate fit curves for the histograms of associated data sets, delineating two peaks representing a longer duration response subpopulation (∼3 seconds) and a shorter duration response subpopulation (∼0.8 s; Supplementary Fig. 1a, red curve).

To differentiate between the average and SD of our data set and those of the modeled fit curves, we refer to the properties of modeled fit curves by their Greek abbreviations (mean = μ, SD = σ, proportion contribution factor =φ). We used these parameters to assess distribution between subpopulations with longer vs shorter response durations.

Statistical analysis

We used chi-square tests (χ²) to assess the statistical significance of differences in penetrance on axon guidance defects. Two-tailed student’s T-tests were used to assess the significance of variation in behavioral responses between lines. Paired student's T-tests were used to assess habituation of tap responses. Significant variation of modeled fit-curve parameters μ₁ or μ₂ between lines was defined as differences greater than σ₁ or σ₂, respectively.

Results

rig-5 nid-1 and wrk-1 nid-1 animals exhibit VNC axon guidance defects

We have previously documented VNC axon guidance defects in rig-5 and wrk-1 single mutants (Schwarz et al. 2009). Here, we used cell-type-specific fluorescent markers to assess the effects of these mutations in a nid-1 null background on AVG and its follower DD/VD and CI axons. We also combined multicolor cell-type-specific markers to determine whether observed follower axon defects were correlated with pioneer axon defects. Figure 2 illustrates percent total penetrance of axon defects in AVG, DD/VD, and CI axons within the VNC.

Fig. 2.

wrk-1 and rig-5 affect axon guidance in nid-1 null background. Overall penetrance of midline crossover defects in nid-1, wrk-1 nid-1, and rig-5 nid-1 mutant strains. N > 100, *P < 0.05 (chi-square test) when compared to nid-1. Markers used: otIs182 (AVG), hdIs24 (DD/VD), hdIs30, and hdIs32 (CI).

AVG defects

In wild-type, the AVG cell body is located in the retrovesicular ganglion (rvg), at the anterior end of the VNC. The AVG axon extends posteriorly along the right VNC, terminating in the tail. Table 1 shows results for wrk-1 nid-1 and rig-5 nid-1 alongside nid-1 single-mutant results. We previously observed that parental marker strains did not exhibit any midline CO defects, nor did rig-5 or wrk-1 single mutants (Schwarz et al. 2009).

Table 1.

Penetrance of AVG defects in nid-1, wrk-1 nid-1, and rig-5 nid-1 mutants.

Genotype	Branch	Leave	Cell pos	Cell lateral	Start left	Cross left	Multiple CO	Total defective	Total N
nid-1	2 (3)	1 (2)	13 (22)	1 (1)	5 (8)	28 (45)	9 (14)	38 (62)	163
wrk-1 nid-1	2 (3)	15* (20)	19 (26)	7* (10)	7 (10)	46* (62)	15* (20)	72* (96)	134
rig-5 nid-1	4 (9)	8* (16)	14 (29)	3 (5)	4 (9)	43* (89)	6 (13)	57* (119)	209

Values in the table are percentages, with number of animals (n) in brackets. wrk-1 and rig-5 single mutants’ total AVG defects less are than that of nid-1, as is the that of the parental marker strain (data not shown). The asterisk (*) indicates values that are significantly different from nid-1 with P < 0.05 (chi-square test). Marker used: otIs182.

We observed three significant axon defects; the “CO,” “leave,” and “branch” phenotypes. Midline CO defects were the most prevalent defect type in both rig-5 nid-1 (43% affected) and wrk-1 nid-1 (46% affected) animals (Fig. 2, Left). In these cases, the AVG axon crossed from the right into the left VNC and extended along the left VNC before returning. In the “leave” phenotype, AVG left the VNC and extended in parallel before returning. This occurred in 15% in wrk-1 nid-1 individuals, significantly more than seen in nid-1 single mutants (1%, P = 0.002). In the “branch” phenotype, the AVG axon branches and an independent branch extends before terminating seemingly at random. We observed AVG branching in 4% of rig-5 nid-1 individuals, a low but significant (P = 0.04) rate as this defect is never observed in nid-1 single mutants. AVG axon branches were observed in 2% of wrk-1 nid-1 animals (not significant).

We observed two significant AVG cell positioning defects; the “cell posterior” and “cell lateral” phenotypes. The AVG cell body was located posterior to the rvg in 19% of wrk-1 nid-1 and 14% of rig-5 nid-1 animals. Both represent a significant increase in this phenotype (P > 0.0005) when compared with the 5% defect rate observed in nid-1 single mutants. These posterior AVG cells invariably extend apical neurites, presumably to the rvg. Occasionally, AVG cell bodies were positioned lateral to the rvg in wrk-1 nid-1 animals. This defect was rare (7%) but significant (P = 0.019 when compared with 1% in nid-1). These lateral cell bodies project neurites both to the right VNC and the rvg. Neurites aimed for the rvg usually extend from the cell body but can also branch off from the axon within the right VNC.

Follower axon guidance defects

CI axons and DD/VD motoneuron axons grow out after the AVG axon and require AVG to navigate correctly. CI cell bodies are in the head close to the nerve ring. CI axons extend into the VNC bilaterally, with about half initially extending into the left VNC (see Fig. 1: red). At the rvg, all CI axons in the left VNC cross the midline and enter the right VNC where they remain as they extend posteriorly. In contrast, the cell bodies of the DD and VD motoneurons are located along the ventral midline (Fig. 1: purple). These cells extend neurites exclusively into the right VNC where they synapse with other motoneurons and CIs.

We noticed a variety of follower axon navigation defects in our mutant strains but chose to only analyze midline CO defects because this was the most common defect. The midline CO defects we observed were highly variable with respect to number and location of CO events. This phenomenon is most clearly observable in CI axons (Supplementary Fig. 2), but was also apparent in DD/VD and where the number of neurites extending from DD/VD cell bodies into the left axon tract was also highly variable in individual animals (Supplementary Fig. 3). This phenotypic variability makes it difficult to place mutant animals into a small number of clearly defined phenotypic groups.

To address the question of whether follower axons cross the midline as a group, we distinguished between fasciculated and defasciculated COs (Fig. 3a). We define “fasciculated CO” as COs where all axons from a class of neurons cross the midline together and “defasciculated CO” as COs where only some of the axons cross, leading to a defasciculation. In an individual animal, COs of both types can occur, so that the total number of COs is larger than the number of animals with COs. The terms “all,” “some,” and “both” are used in figures and tables describing these defects. Tables 2 and 3 show penetrance of different COs types for DD/VD and CI axon COs, respectively, in wrk-1 nid-1, rig-5 nid-1, and nid-1 mutants. Figure 3, b and c shows these values as percentages of total DD/VD and CI COs, respectively. We previously observed that rig-5 and wrk-1 single mutants have a total of 2 and 5% DD/VD defects, and 4 and 16% CI defects, respectively (Schwarz et al. 2009). Due to the generally low penetrance of overall VNC defects, we did not further examine the types of follower COs in these single mutants.

Fig. 3.

wrk-1 and rig-5 mutations change distributions of follower axon CO types in a nid-1 null background. a) Schematic and example images showing CO defect types in follower axons. All images are of rig-5 nid-1 mutants with CI (hdIs32) fluorescent marker, captured at 40× magnification. Dashed lines note the bounds of the left (top) and right (bottom) VNC where not occupied by fluorescing axons. Arrows point to the anterior bound of depicted CO(s). Scale bar is 5 microns. b, c) Distribution of DD/VD (b) and CI (c) axon CO types in nid-1, wrk-1 nid-1, and rig-5 nid-1 mutant lines. Percentages are out of total animals with COs, not all animals counted. *P < 0.05 (chi-square test) when compared with nid-1. Colors in schematics (a: top) correspond to colors used in (b) and (c). “All” is the proportion of animals with COs in which all of those COs involve all axons of that type at that location. “Some” is the proportion of animals with COs in which all of those COs involve only a subset of axons of that type at that location. “Both” is the proportion of animals with COs in which at least one CO involved all axons of that type in that location, and at least one CO involved only a subset of axons of that type at that location.

Table 2.

Penetrance of DD/VD CO defects in nid-1, wrk-1 nid-1, and rig-5 nid-1 mutants.

Genotype	CO of all local axons			CO of some local axons			Animals with both CO types			Total animals with CO
	% of CO	% of total	N	% of CO	% of total	N	% of CO	% of total	N	% of total	N	Total N
hdIs24	5	1	1	100	19	14	5	1	1	19	14	72
nid-1	28	13	27	42	19	40	29	13	27	44	95	216
wrk-1 nid-1	23	11	12	29*	13	15	48*	22*	25	46	52	113
rig-5 nid-1	29	16	28	34	19	33	37*	21*	36	57*	97	171

“% of CO” values are percentages of total animals with CO defects. “% of total” values are percentages of total animals observed. “N” values are the raw count of individuals exhibiting each phenotype, or phenotype combination in the case of “both CO types.” The asterisk (*) indicates values that are significantly different from nid-1 with P < 0.05 (chi-square test). Marker used: hdIs24.

Table 3.

Penetrance of CI CO defects in nid-1, wrk-1 nid-1, and rig-5 nid-1 mutants.

Genotype	CO of all axons			CO of some axons			Animals with both CO types			Total animals with CO
	% of CO	% of total	N	% of CO	% of total	N	% of CO	% of total	N	% of total	N	Total N
nid-1	51	10	21	29	6	12	20	4	8	19	41	215
wrk-1 nid-1	23	28*	37	19*	11*	14	32*	18*	24*	58*	75	130
rig-5 nid-1	48	27*	34	21	12	15	31*	17*	22	56*	71	124

“% of CO” values are percentages of total animals with CO defects. “% of total” values are percentages of total animals observed. “N” values are the raw count of individuals exhibiting each phenotype, or phenotype combination in the case of “both CO types.” The asterisk (*) indicates values that are significantly different from nid-1 with P < 0.05 (chi-square test). Markers used: hdIs30 and hdIs32.

DD/VD CO defects

DD/VD axons cross the midline in 44% of nid-1 single mutants (Table 2 and Fig. 2, middle), a significant (P = 0.005) increase compared with the parental marker strain (19%). Furthermore, 28% of DD/VD COs in nid-1 animals are fasciculated (Fig. 3b), a significant increase from 5% seen in the parental strain (P < 0.0001). While the overall penetrance of wrk-1 nid-1 animals with DD/VD COs is similar to those of nid-1 single mutants, these animals are more likely to have multiple COs, involving both all and a subset of axons instead of just one type of CO (Fig. 3b, middle). The penetrance of DD/VD COs is significantly elevated in rig-5 nid-1 animals when compared to nid-1 single mutants (57%, P < 0.0001). This is the result of a greater number of animals with both types of CO and animals, where all DD/VD axons at a specific location cross the midline together (Fig. 3b, middle). This results in a significant change in the distribution of CO types that resembles the shift observed in wrk-1 nid-1.

CI CO defects

The penetrance of CI axon COs is significantly elevated in both rig-5 nid-1 (56%) and wrk-1 nid-1 double mutants (58%) compared with nid-1 single mutants (19%, P < 0.0001; Fig. 2, Right). In all three strains, about half of the individuals with CI COs only have fasciculated COs (Fig. 3c). In nid-1 mutants, 29% of individuals with COs have only defasciculated COs and 20% have both types of CO. This ratio is inverted in wrk-1 nid-1 and rig-5 nid-1: 19% of wrk-1 nid-1 animals with CI COs have only defasciculated COs, and 32% have both CO types, significant differences from nid-1 (P < 0.001). We observed similar differences in rig-5 nid-1 but to a slightly lesser extent.

CI and DD/VD axons follow AVG axons across the midline

As described above, the distribution of follower axon defect type is significantly different in both wrk-1 nid-1 and rig-5 nid-1 when compared to nid-1 single mutants. Follower axon defects might arise from CI and DD/VD axons following misguided AVG axons, or they could be the direct result of the mutations, and independent of AVG. To determine to what extent this shift in follower defects is the result of increased AVG pioneer defects, we assessed whether midline crossing defects in each follower axon type corresponded spatially with AVG midline crossing defects. We previously demonstrated that CI (and to a lesser extent DD/VD) CO defects in ccd-5 nid-1 mutants were attributable to axons following AVG across the midline (Feresten et al. 2022). Here, we assess the relationship between AVG and follower axons in wrk-1 nid-1 and rig-5 nid-1 mutants.

As illustrated in Fig. 4, we again focus on the frequency of fasciculated vs defasciculated CO defects; a “fasciculated CO” (where all axons from a class of neurons cross the midline together) can coincide spatially with an AVG CO, or it can occur independent of AVG. Similarly, a “defasciculated CO” (where only some of the axons cross leading, to a defasciculation) can occur with an AVG CO, or it can occur independent of AVG. In an individual animal, COs of both types can occur, so that the total number of COs is larger than the number of animals with COs. This principle holds true for follower axons crossing with AVG vs independently; one individual can have multiple follower axon COs, some coinciding with an AVG CO and others occurring independently of AVG. Figure 5 shows follower axon association with AVG COs (AVG CO). More detailed CO defect scores are available in Supplementary Tables 3 and 4.

Fig. 4.

Illustration of crossover phenotypes assessed. Example images illustrate the different combinations of follower CI axons (red) and AVG axon (green) crossovers assessed. We observed many variations on the common theme of “all,” “some,” or “no” follower axons crossing in conjunction with AVG. When all axons cross the midline together as a single fascicle, it is a “fasciculated CO.” When some axons leave the VNC fascicle to cross the midline, it is a “defasciculated CO.” AVG can cross independent of follower axons, and follower axons can cross independent of AVG. Many individuals had multiple instances of midline COs. An example is provided of a stretch of VNC where axons cross the midline multiple times, in a variety of ways. All images are of rig-5 nid-1 mutants with CI (hdIs32; red) and AVG (otIs182; green) fluorescent markers, captured at 40× magnification. Dashed lines note the bounds of the left (top) and right (bottom) VNC where not occupied by fluorescing axons. Arrows point to the anterior bound of depicted CO(s). Scale bar is 5 μm.

Fig. 5.

wrk-1 and rig-5 mutations increase proportion of followed AVG COs in nid-1 null background. a, b) Distribution of DD/VD (a) and CI (b) axon associations with AVG COs in wrk-1 nid-1, rig-5 nid-1, and nid-1 mutant strains. Out of all animals with AVG COs, some have only AVG COs that are followed by all follower axons observed, some only have AVG COs followed by a subset of the follower axons observed, and some only AVG COs that are completely unfollowed. Lower panel contains schematic representations of each CO type. Color blending in (a) and (b) indicates more than one type of follower axon response to AVG COs within the same animal. For example, “All and none” indicates the proportion of animals with AVG COs in which at least one AVG CO was followed by all follower axons in that location, and at least one AVG CO was not followed by any follower axons. *P < 0.05 (chi-square test) when compared with nid-1. For N breakdown, see Supplementary Tables 3 and 4.

DD/VD follow AVG across the midline

Figure 5a shows DD/VD axon association with AVG COs in nid-1, wrk-1 nid-1, and rig-5 nid-1 mutants. wrk-1 nid-1 and rig-5 nid-1 double mutants showed a significant relative increase in fasciculated COs compared to nid-1 single mutants, and a commensurate decrease in AVG COs unaccompanied by DD/VD axons, only reaching significance in rig-5 nid-1. This suggests that both wrk-1 and rig-5 provide guidance information preventing axon crossing when the pioneer is misguided. There were no significant differences in the relative rate of defasciculated COs, in which only some of the available DD/VD axons follow AVG across the midline.

CI follow AVG across the midline

Figure 5b shows CI axon association with AVG COs in nid-1, wrk-1 nid-1, and rig-5 nid-1 mutants. As seen with DD/VD, a significantly higher proportion of AVG CO were followed by fasciculated CI in wrk-1 nid-1 and rig-5 nid-1 double mutants compared to nid-1 single mutants. In wrk-1 nid-1, we also saw an increase in defasciculated COs, reducing the fraction of AVG COs with no CI following to 10%. This suggests that wrk-1 has a guidance function for several classes of axons when the pioneer is misguided. We did not observe this shift in defasciculated COs in rig-5 nid-1, instead seeing an even larger increase in fasciculated COs. The increase in the fraction of follower axons following the misguided pioneer provides an explanation for the general increase in follower axon defects in the double mutants.

Taken together, our data show that follower axons tend to follow the pioneer axon even when the pioneer is misguided. However, since not all followers follow the misguided pioneer, there must be additional guidance cues available to the follower axons. Our data suggest that both wrk-1 and rig-5 are part of these additional guidance systems but are not required for maintaining fasciculation within axon types.

Behavioral analysis

The navigation errors described above result in VNC disorganization, which may impact VNC neurites’ ability to establish en passant synapses. This would cause defects in neuronal circuit formation and might have wide-reaching impacts on basic motor function and behavior in animals. We chose to focus on the behavior of rig-5 and rig-5 nid-1 mutants, as some wrk-1 nid-1 mutant animals have obvious morphological defects which could cause behavioral defects independent of axonal defects. We did not observe any obvious defects in the ability of animals to properly move forward or backward on culture plates. To detect more subtle movement defects, we used an automated tracking system, the Multi-Worm Tracker or MWT (Swierczek et al. 2011). The MWT allowed us to assess spontaneous movement of an entire population of animals and capture a large number of movement parameters. In addition, we were able to assess a more complex nervous system function in the form of responses to mechanical stimuli, employing the well-established mechanosensory “tap” response assay (Swierczek et al. 2011). C. elegans responds to tap by ceasing forward movement and reversing for a few seconds before moving forward again. This behavior habituates, meaning the likelihood and magnitude of an animal’s response to tap decreases as a result of repeated stimulation. This is a form of simple learning (Ardiel and Rankin 2010). The underlying circuit of the tap response has been mapped through laser ablation studies (Wicks and Rankin 1995) and is dependent on en passant synapses between CI and motoneurons within the right VNC. These neurons have axon guidance defects in rig-5 nid-1 double mutants and rely heavily on AVG for guidance. If these guidance defects interfere with the ability of CI and motoneuron axons to find their synaptic partners, we would expect to see disruptions in tap response and/or tap response habituation. In lin-11 mutants, the AVG neuron fails to differentiate, i.e. is completely absent functionally, leading to the highest observed rate of AVG-dependent guidance errors in CI and DD/VD motoneuron axons (Hutter 2003). Therefore, lin-11 mutants can serve as a reference for the largest behavioral defects that can be expected solely due to AVG-related axon guidance errors.

The data for control, lin-11 and nid-1, are taken from our previous study assessing the effects of ccd-5 on behavior (Feresten et al. 2022). The behavioral experiments described here were conducted concurrently with those of ccd-5 and all associated controls. Here we assess this dataset in the context of rig-5.

rig-5 has a limited effect on behavior of free-moving animals

To assess whether rig-5 is required for locomotion in the absence of nid-1, we recorded 30 seconds of baseline activity following acclimatization and assessed the average proportion of time spent moving forward and forward locomotion speed (Fig. 6). rig-5 single mutants moved slightly but significantly faster than control (rig-5: 0.9 mm/s; control: 0.7 mm/s; P = 0.008), whereas nid-1 single mutants and rig-5 nid-1 double mutants did not differ significantly from control (Fig. 6a). Neither nid-1 nor rig-5 significantly affected the proportion of the 30 s animals spent moving forward (Fig. 6b). The large SD suggest that individual animals vary substantially in their baseline locomotion. In contrast, lin-11 mutants moved significantly less frequently and more slowly than all other mutants. These results confirmed observations of the movement of animals on a plate; lin-11 mutants are noticeably less active than control, while the other strains’ behaviors are indistinguishable from control to the casual observer.

Fig. 6.

Free movement of rig-5 mutants. Behavior of: parental control strain (Control), nid-1, rig-5, rig-5 nid-1, and lin-11. a) Average forward movement speed. b) Proportion of animals engaged in forward locomotion at each timepoint, averaged across timepoints. Individual points mark within-plate averages. N > 10 plates for each line. Error bars mark the upper and lower limits of the 95% confidence interval. *P < 0.05, **P < 0.005, ***P > 0.0005 when compared via two-tailed T-test with the strain sharing a color with the asterisk.

rig-5 mutation increases nid-1-associated defects in tap response and its habituation

To evaluate tap response and habituation, we delivered 30 consecutive “tap” stimuli spaced 10 s apart and recorded the responses of animals on a minimum of 10 plates per line tested. We assessed the proportion of animals responding to tap, the speed of their reversal, and the duration of the reversal as described by (McDiarmid et al. 2020).

Response probability

Control and rig-5 single mutants respond to the initial tap at similar rates (75 and 70%, respectively; Fig. 7a left panel), whereas nid-1 single mutants are significantly less responsive (46%, P < 0.005). rig-5 nid-1 double mutants are less responsive than nid-1 single mutants (32%, P = 0.02). lin-11 (29%) was less likely to respond to tap than control (P < 0.005) or nid-1 (P = 0.006) but did not significantly differ from rig-5 nid-1. These differences in tap response are largely preserved over the 29 subsequent taps (Fig. 7a, middle panel). The response probability of control and rig-5 mutants habituated to similar extents (56 and 48%, respectively), while nid-1 habituated significantly less (33%, P < 0.005). rig-5 nid-1 and lin-11 habituate even less (20 and 12%, respectively), and their habituation rate curves are noticeably different (Fig. 7a, middle panel): rig-5 nid-1 response probability drops steeply and reaches its lowest point (10%) already by tap 6 while lin-11 has a more gradual and linear decrease in response probability. These differences in habituation pattern suggest that the degree to which rig-5 nid-1 can habituate may be limited by its low initial responsiveness. These results show the profound impact of nid-1 on the likelihood of animals to respond to a tap stimulus. Although the rig-5 null mutation slightly amplifies nid-1 associated decreases in response to the initial tap, we did not observe any additional significant effects of the rig-5 null mutation with or without nid-1.

Fig. 7.

rig-5 mildly exacerbates nid-1-associated decreases in responsiveness to tap and habituation of that response. Behavior of: parental control strain (Control), nid-1, rig-5, rig-5 nid-1, and lin-11. a) Average probability of response to each consecutive tap. b) Average speed of reversal for each consecutive tap. c) Average duration of reversal in response to each consecutive tap. Left: Box and whisker plots show quartile data distribution of initial tap responses. Center: scatter plot of response averages at each stimulus, spaced 10 s apart. 20 s moving average of every two values is shown as a solid line. Right: Box and whisker plots show quartile data distribution of habituation, with individual plate averages shown as points. Habituation was calculated as the average first response minus the average of the last three responses. N > 10 plates. *P < 0.05, **P < 0.005, ***P > 0.0005 when compared via two-tailed student's T-test with the strain sharing a color with the asterisk.

Reversal speed

rig-5 and nid-1 single mutants both reversed significantly more slowly than control in response to tap (Fig. 7b left panel; P < 0.0005). rig-5 nid-1 reversed slightly but significantly more slowly than rig-5 (P < 0.0005), at a speed similar to that of nid-1. This indicates that the effects of rig-5 and nid-1 on initial reversal speed are not additive. The reversal speed of rig-5 habituated significantly less than control (Fig. 7b, middle and right panels; P = 0.002). This is not due to a physical inability to reverse more slowly, since lin-11 mutants reversed much more slowly at all time points. We observed previously that the nid-1 null mutation decreases reversal speed and prevents its habituation (Feresten et al. 2022). rig-5 nid-1 double mutants also failed to habituate, consistent with this known effect of the nid-1 mutation.

Here we observed that rig-5 decreases reversal speed and its habituation. Since the effects of rig-5 and nid-1 on initial reversal speed are not additive, it suggests that both genes affect the same pathway required for tap response.

Reversal duration

rig-5 single mutants initially reverse for a slightly longer period of time than control in response to tap (3.3 s and 2.85 s, respectively. P = 0.02), but they habituate normally (Fig. 7c). nid-1 single mutants initially reverse for less time than control (2.08 s, P = 0.004). nid-1 single mutants’ response duration habituates (P = 0.002), slightly but not significantly less than control (P = 0.03). Initial response duration for rig-5 nid-1 was significantly shorter than that for rig-5 (P = 0.01), but not significantly different from control or nid-1. rig-5 nid-1 did not significantly habituate. While we have observed some effect of rig-5 on response duration, large variations in nid-1 and rig-5 nid-1 response durations prevent us from drawing clear conclusions.

Two subpopulations with different reversal durations are distinguishable

The large variability observed in reversal durations in this dataset was unexpected. Over a large number of experiments, we have observed that response duration is less prone to experimental variations than are response probability or speed (e.g. McDiarmid et al. 2020). This unusually large distribution of reversal durations may indicate real differences within the population. As the axon guidance defects in all mutants discussed here are partially penetrant and only affect a subpopulation of animals, we hypothesized that the aggregate data might be obscuring the presence of subpopulations with different tap response characteristics; behavioral phenotypes may also be partially penetrant. To investigate this possibility, we plotted histograms of response durations at each time point for each strain. We found that a nonparametric Kernel-based fit-curve suggested two subpopulations in the mutant strains but lin-11. We then conducted a parameter estimate analysis in Matlab (R2020a; MathWorks) assuming two normally distributed subpopulations, each with a different mean (μ), SD (σ), and population contribution (φ). For details see Methods section, Supplementary Fig. 1, Supplementary Tables 2 and 6. We excluded lin-11 from this analysis as the parameters estimated did not reliably fit the observed data.

In response to the initial stimulus, a small proportion of control strain animals responded for a short duration (μ₁ = 0.81 seconds; φ₁ = 0.13; Fig. 8a and c), and a larger proportion of animals responded for a much longer duration (μ₂ = 3.21 seconds; φ₂ = 0.87; Fig. 9b and c). Initial responses of mutant strains deviated from the control to various degrees. In nid-1, more animals had shorter duration responses and consequently fewer animals had longer duration responses (Fig. 8c). While the mean of the longer duration response did not significantly differ between control and nid-1 (Fig. 8b), the mean of the shorter duration response was slightly shorter in the mutant (Fig. 8a). rig-5 mutants showed a shift in the opposite direction: slightly fewer animals with shorter duration responses and slightly more animals with longer duration responses. As in nid-1, there was a small reduction in the shorter response duration. The population distribution in rig-5 nid-1 double mutants was similar to control (Fig. 8c). The double mutant's shorter duration responses were even shorter than those of the rig-5 and nid-1 single mutants (Fig. 8a).

Fig. 8.

rig-5 mutation affects μ₁ and μ₂ differently. For parental control strain (Control), nid-1, rig-5, and rig-5 nid-1, (a) and (b) show changes in modeled response duration parameters for from first (solid) to last (cross-hatched) tap responses for μ₁ (a), and μ₂ (b). Error bars in (a) and (b) are + and—the model generated standard deviation (σ for each respective μ). Black asterisk (*) notes significant (P < 0.05) decrease between timepoints within a strain. Colored asterisk (*) above values that are significantly (P < 0.05) lower than that of a different strain (sharing the color with the asterisk) at the same time point. NS = not significantly different between timepoints. c) Population distribution (φ) between the shorter response group (φ₁) and longer response group (φ₂) for the first and last tap responses in each line. d) Percent of total population lost in φ₂ and gained in φ₁ between first and last taps, as calculated by first φ₁ minus last φ₁. The inverse values are true for φ₂. All modeled parametric values are included in Supplementary Table 6. Fit values of the modeled parametric curves are available in Supplementary Table 2.

Fig. 9.

rig-5 and wrk-1 amplify nid-1-associated AVG CO defects, and genetically interact with previously described genetic interaction pathways. The figure shows penetrance of AVG CO defects in various single, double, and triple mutants. To ensure consistency, these mutant strains were all scored in the same period and independently of the phenotypic analysis presented in earlier figures. a) AVG CO penetrance of nid-1, rig-5, and wrk-1 single, double, and triple mutants, demonstrating that rig-5 and wrk-1 increase AVG navigation defects in a nid-1 null background. b) rig-5 and wrk-1 double mutants genetically interact with vesicle transport pathway genes ccd-5, cdk-5, and unc-18. c) wrk-1 and vab-1 double and triple mutants genetically interact in a nid-1 null background. Colored asterisk (*) notes significant (P < 0.05, N > 100, chi-square test) difference between strain below asterisk and the strain sharing a color with the asterisk.

When comparing the initial to the last response (Fig. 8a and b), we found that the mean response duration of each subpopulation in the control strain became significantly shorter over time (i.e. habituated). The subpopulations of all mutant strains except rig-5 nid-1 habituated as well. Also notable in the control strain was a shift in the size of the subpopulations over time: 31% of the population shifted from the longer response μ₂ group to the shorter response μ₁ group over the course of 30 taps (Fig. 8d). This population shift was lower in all mutants: 26% of rig-5 single mutants, 21% of nid-1 single mutants, and 12% for rig-5 nid-1 double mutants. In keeping with those initial observations, rig-5 nid-1 response durations did not significantly habituate (Fig. 9a and b). Furthermore, the population shift from a longer to a shorter duration response we saw in the control strain was largely absent in rig-5 nid-1 mutants. This further supports our hypothesis that normal behavior is impaired in rig-5 nid-1 mutants beyond that seen in nid-1 or rig-5 single mutants.

Taken together, these analyses revealed that animals respond to a tap stimulus with two distinct responses, defining two subpopulations: one with a shorter duration response and another with a longer duration response. Mutants varied in both the size of the subpopulations and the magnitudes of the response compared to control: the nid-1 mutant had a shift in the size of the subpopulations and a decreased magnitude of the shorter duration response; rig-5 and rig-5 nid-1 mutants did not differ much from control in the size of the subpopulations but had a decreased mean duration for the shorter duration response and an increased mean duration for the longer duration response. These observations explain the high variability in the reversal duration of these mutant strains and indicate that the current practice of averaging the behavioral data of entire populations may be obscuring a broad range of mutant phenotypes.

Genetic interaction analysis

rig-5 and wrk-1 act in different genetic pathways

As we have shown previously (Schwarz et al. 2009), rig-5 and wrk-1 single mutants do not have significant AVG axon navigation defects. However, the penetrance of AVG CO defects was significantly greater in double mutants of rig-5 and wrk-1 with nid-1 (Fig. 9a), indicating that these IgCAMs are involved in AVG axon navigation. To determine, if the genes act in the same pathway, we compared a wrk-1 rig-5 nid-1 triple mutant with corresponding double mutants and a nid-1 parental strain. Defects in the triple mutant were significantly elevated compared to the corresponding double mutants (Fig. 9a) indicating that rig-5 and wrk-1 act in different pathways.

rig-5 and wrk-1 genetically interact with the ccd-5 and cdk-5 pathways

We previously identified a genetic interaction cluster in which ccd-5 and cdk-5 interact with a series of vesicle-associated genes (rab-3, unc-18, ida-1) and unc-5 (the netrin receptor) to facilitate AVG axon guidance (Bhat and Hutter 2016; Feresten et al. 2022). To determine whether the two pathways discussed here—rig-5 and wrk-1—fit into this previously identified pathway, we created the following triple mutants: rig-5 ccd-5 nid-1, rig-5 unc-18 nid-1, and wrk-1 cdk-5 nid-1. It is of note that ccd-5, unc-18, and wrk-1 are all located on the X chromosome, which prevents us from easily assessing their genetic interactions. The AVG defect penetrance rates of these triple mutants were compared to that of their respective nid-1 null double mutants. The penetrance of defects in all three triple mutants was significantly lower compared to at least one of their respective double mutants (Fig. 9b and Supplementary Table 5). These data suggest that wrk-1 and rig-5 gene products interfere with the pathways affected by ccd-5, cdk-5 and unc-18 mutations such that the presence of wrk-1 or rig-5 increases the defects.

wrk-1 interacts with vab-1 Eph receptor

Previous authors (Boulin et al. 2006) have demonstrated a relationship between wrk-1 and vab-1, the sole C. elegans Eph receptor, in guiding follower axons. In this model, wrk-1 expressed on embryonic motoneuron cell bodies interacts in trans with vab-1 to prevent PVQ and HSN axons from crossing the midline (Boulin et al. 2006). To determine whether vab-1 and wrk-1 genetically interact in the context of AVG axon guidance, we compared AVG defect penetrance between vab-1 nid-1, wrk-1 nid-1, and vab-1 wrk-1 nid-1 mutants. As shown in Fig. 9c, the vab-1 wrk-1 nid-1 triple mutant does not have an increased penetrance of AVG axon guidance defects when compared to the vab-1 nid-1 double mutant. This indicates that wrk-1 and vab-1 act in the same pathway with respect to AVG navigation as well.

Discussion

In this study, we identified rig-5 and wrk-1 as enhancers of nid-1-associated AVG pioneer axon guidance defects. rig-5 nid-1 and wrk-1 nid-1 double mutants also showed a substantial increase in follower axon navigation defects. This is largely due to an increase in the proportion of follower axons following the misguided AVG pioneer. We found that wrk-1 and rig-5 act in different pathways. Since the misguided axons are key components of the motor circuit, we assayed the mutant animals for behavioral defects. Using a novel analytical approach to the well-established tap response behavioral assay, we found subtle defects in elicited movement behavior and habituation in rig-5 nid-1 double mutants.

rig-5 and wrk-1 are required for AVG and follower axon navigation in a nid-1 mutant background

We found that rig-5 and wrk-1 mutants enhance AVG axon navigation defects in a nid-1 mutant background. rig-5 and wrk-1 single mutants show no AVG axon navigation defects (Schwarz et al. 2009), suggesting that these two GPI-anchored IgCAMs have an indirect modulatory function on AVG axon navigation. rig-5 nid-1 and wrk-1 nid-1 double mutants also show increased CI and DD/VD follower axon guidance defects. This increase can largely be explained by follower axons following the misguided pioneer, suggesting that much of the observed follower axon defects are secondary to increased AVG axon defects. rig-5 and wrk-1 might not have a direct role in follower axon navigation.

Our data suggest that nid-1 acts in a parallel pathway to both wrk-1 and rig-5, and that the amplification or appearance of defects in double mutants with nid-1 is the result of a compounding loss of guidance components. NID-1 is a basement membrane protein localized close to body wall muscle cells, which flank the VNC (Kang and Kramer 2000), and might facilitate the localization of guidance cues. How nid-1 mutations cause CO defects at the ventral midline is currently unclear.

The insect homolog of rig-5, lachesin, plays a role in neuronal differentiation in grasshoppers and fruit flies (Karlstrom et al. 1993) and acts through homophilic adhesion to allow neuronal cell aggregation in vitro (Nurre 2007). Since the AVG axon in C. elegans is the first axon to grow into the right VNC axon tract, RIG-5 is unlikely to act as a homophilic adhesion molecule between the AVG axon and other axons or cells. We previously determined the expression of rig-5 using promoter-GFP constructs (Schwarz et al. 2009). We did not notice rig-5 expression in either AVG or embryonic motoneurons. We also did not see expression in muscle cells or hypodermal cells, which are close to the VNC and could provide guidance cues. However, our reporter construct might lack important control elements for gene expression, so that we currently cannot rule out expression in AVG and a potential cell-autonomous action in AVG. A genome-scale analysis of gene expression at the tissue level in the embryo using RNAseq (Boeck et al. 2016) shows strong expression of rig-5 in neuronal tissue and little or no expression in muscle or hypodermis. This suggests that rig-5 likely acts in neurons. One possible site of action is AVG, suggesting a cell-autonomous role. Alternatively, rig-5 could act in embryonic motoneurons acting as a repulsive cue like wrk-1 to prevent midline crossing. Several experiments are required to determine the cellular focus of rig-5. Insertion of a reporter into the genomic locus of rig-5 should provide the complete expression pattern of rig-5 at the time of AVG axon outgrowth in the embryo. Depending on the observed expression, rescue experiments with cell-type-specific promoters are required to confirm the anticipated site of action. With our current data, we cannot distinguish between the two models for rig-5 action described above.

wrk-1 single mutants show midline CO defects in a small number of neurons with axons in the VNC including CI, but the majority of VNC axons do not show CO defects in wrk-1 single mutants (Boulin et al. 2006; Schwarz et al. 2009). The penetrance of CI axon defects in wrk-1 nid-1 double mutants is more than additive compared with the single mutants, suggesting a synergistic relationship between wrk-1 and nid-1 in CI axon navigation. wrk-1 is expressed on the surface of embryonic motoneurons located at the ventral midline. Boulin et al. (2006) have shown that wrk-1 prevents CO of follower axons PVQ and HSN through genetic and physical interactions with the Eph receptor VAB-1, expressed in neurons. These authors propose that wrk-1 expression facilitates an Eph-mediated repulsive interaction between extending VNC axons and embryonic motoneuron cell bodies, preventing axons from crossing the midline. We found that wrk-1 and vab-1 also act in the same pathway for AVG axon navigation suggesting that this mechanism may more broadly affect both pioneer and follower axons.

Follower defects are correlated with pioneer defects

When the AVG pioneer neuron is ablated before axon outgrowth, various classes of follower axons including CI and DD/VD axons show CO defects (Durbin 1987; Hutter 2003). The penetrance of these defects is less than 50%, indicating that there are pioneer-independent cues that follower axons can use to navigate. AVG ablation experiments cannot address the question of whether follower axons can develop CO defects by following a misguided pioneer. We recently found that some but not all follower axons follow the misguided VNC pioneer AVG in ccd-5 nid-1 double mutants (Feresten et al. 2022). ccd-5 encodes a cdk-5 interacting protein and genetically interacts with the netrin receptor unc-5, but it is not clear which guidance pathways require ccd-5.

In this study, we characterized the role of wrk-1 and rig-5 in AVG axon navigation. Both genes encode GPI-anchored cell surface receptors/adhesion molecules of the IgCAM family and provide a direct molecular link to guidance pathways. We found that in nid-1 mutants more than 50% of follower axon COs were following the misguided pioneer. This percentage increased dramatically in wrk-1 nid-1 and rig-5 nid-1 double mutants, up to over 90% for CIs. This allows several conclusions: (1) follower axons tend to follow the pioneer, even if it is misguided. (2) Since followers do not consistently follow the pioneer across the midline in nid-1 mutants, there must be additional pioneer-independent cues able to counter incorrect pioneer guidance. (3) Since followers follow the misguided pioneer much more consistently in wrk-1 nid-1 and rig-5 nid-1 double mutants, we can surmise that both wrk-1 and rig-5 are part of these additional guidance cue pathways.

Through our genetic interaction studies, we showed that wrk-1 and rig-5 are in different pathways. We have therefore identified two distinct pathways that help follower axons ignore misguided pioneers. As mentioned above, wrk-1 is proposed to act as a negative cue expressed at the midline to prevent axon crossing. Its absence would allow more axons to cross the midline and follow the pioneer. The site of action for rig-5 and its molecular pathway are currently unknown.

rig-5 nid-1 double mutants show subtle behavioral defects

Tap response is regulated by two opposing circuits moderated by a turning circuit (Wicks and Rankin 1995; Zhen and Samuel 2015; Wang et al. 2020). The forward and backward locomotion circuits habituate independently, affecting the timing of the switch from reverse back to forward locomotion (Wicks and Rankin 1995; Kitamura et al. 2001; Ardiel and Rankin 2010; Sordillo and Bargmann 2021). The synapses forming this circuitry are largely en passant synapses within the right VNC (White et al. 1986). Axon CO events may disrupt the formation of these synapses, which could lead to movement and behavioral defects.

We previously demonstrated that nid-1 single mutants are on average less responsive to tap than a parental control strain, and they habituate less (Feresten et al. 2022). We also found that ccd-5 nid-1 double mutants, which show substantial CI and DD/VD axon CO defects, do not have movement defects that are more severe than nid-1 single mutants. At that time, we proposed that preserved fasciculation in ccd-5 nid-1 mutants was preserving behavioral function. rig-5 nid-1 double mutants’ axon navigation defects were also largely fasciculated, and we did not find any dramatic exacerbation of nid-1-associated movement defects, further supporting this hypothesis.

A detailed analysis of rig-5 nid-1 tap response revealed defects in the habituation of response duration. This analysis also revealed that each line tested had two subpopulations: a smaller fraction with a shorter response duration and a larger fraction with a longer response duration. Both groups habituate, with their mean response durations declining over time. However, habituation also involves a shift in population from the longer response group to the shorter response group. Compared with nid-1 single mutants, half as many rig-5 nid-1 animals shifted from longer to shorter response durations over 30 taps. The fraction of animals that failed to shift their behavioral response type is smaller than the fraction of animals with elevated axonal defects in rig-5 nid-1 double mutants when compared to nid-1 single mutants, suggesting that there is no simple correlation between axonal and behavioral defects.

A simple correlation between axonal and behavioral defects is not expected for several reasons. First, not all axonal defects necessarily cause defects in synapse formation. Minor navigation defects might have a limited effect on synapses. In the case of motoneurons, misplaced axons can still establish functional neuromuscular synapses (Hedgecock et al. 1990; Gally and Bessereau 2003; Alexander et al. 2009). Second, the mutants might have additional neuronal defects. We already know that nid-1 mutants have defects in synaptic morphology and function that likely affect the function of neuronal circuits (Ackley et al. 2003, 2005). The axonal defects described here are partially penetrant and highly variable. Animals with AVG navigation defects show variable follower axon defects with respect to the number and location of COs and the number of axons crossing. The range of axonal defects prevents us from placing individual mutant animals into a small number of clearly defined groups with distinct axonal defects.

To establish a direct correlation between axonal and behavioral defects, we would need to test several mutant subpopulations with specific axonal defects for behavioral defects. Because behavioral responses are variable even in wild-type, a large number of animals (ideally more than 100) must be tested from each phenotypic group. Axonal defects can be detected and classified only under high magnification with a fluorescence microscope so this proposed experiment would require mounting animals on coverslips, observing, and classifying their phenotypes, and then manually recovering hundreds to thousands of individual animals with the targeted axonal defects in order to get a large number of animals in each subgroup. This procedure is not only difficult but also time-consuming and the MWT must assess age-synchronized animals, further complicating this already convoluted experimental process. Despite these limitations, our behavioral analysis suggests that even severe axonal defects, which are expected to significantly affect neuronal circuit formation, have only mild effects on the functioning of the neuronal circuit. This suggests that axonal defects can be partially compensated either at the level of circuit formation or at the level of circuit function.

It is surprising that even our control strain showed two subpopulations with respect to response duration, since the strain does not show axonal navigation defects. We observed a bimodal distribution of the tap response also in other wild-type strains, suggesting that this is a novel aspect of normal habituation to tap, which can be explored further in the future. Previous authors have suggested that multiple levels of CI circuitry may both interact and compete when regulating elicited reversal (Zhen and Samuel 2015; Wang et al. 2020). It is possible that, like the forward and reverse circuits, these different levels of reverse locomotion regulation also habituate at distinct and independent rates. Further understanding of this phenomenon will require detailed assessment of a variety of wild-type strains in the future.

wrk-1 and rig-5 genetic interaction studies

Our genetic interaction studies revealed that rig-5 and wrk-1 act in different pathways indicating that we have identified two different pathways implicated in AVG axon navigation.

We previously identified aex-3, an activator of the GTPase RAB-3, as important for AVG axon navigation in a nid-1 mutant background (Bhat and Hutter 2016). RAB-3 is implicated in the control of vesicle trafficking in the axon, so we performed genetic interaction studies with additional genes involved in vesicle trafficking and vesicle release. We found that aex-3 was in a common genetic pathway with all the genes tested. More recently, we found that ccd-5, a cdk-5 binding partner, is also part of this pathway (Feresten et al. 2022). We performed genetic interaction studies with rig-5 and wrk-1 to determine if they too interact with this vesicle transport pathway. We produced triple mutants with rig-5 or wrk-1 and either ccd-5 nid-1, cdk-5 nid-1, or unc-18 nid-1 (unc-18 promotes vesicle exocytosis) and found that AVG axon navigation defects are reduced in all triple mutant combinations. This result indicates that AVG axon defects associated with ccd-5, cdk-5, or unc-18 mutations are exacerbated by wild-type wrk-1 and rig-5. This unexpected result shows that wrk-1 and rig-5 are not simply part of the pathway we discovered previously, but instead interact with or are regulated by this pathway. Further studies are required to clarify the relationship between these genes.

Conclusion

In this study, we characterized the role of two members of the IgCAM family, rig-5 and wrk-1, in AVG pioneer axon navigation and follower navigation. We found that both rig-5 and wrk-1 increase defects of AVG axon navigation defects in a nid-1 mutant background. Increased follower axons defects are due to an increased number of follower axons following the misguided pioneer. The two IgCAMs establish two additional pathways contributing to AVG axon navigation. Misguided axons are part of the motor circuit controlling movement of the animals so axonal defects should lead to functional defects in the motor circuit. Our initial behavioral analysis revealed no obvious defects in free movement and only minor defects in response habituation to mechanosensory stimuli. A more detailed analysis showed two distinct subpopulations responding differently to tap in control and mutant strains. The response pattern differed in various mutants compared to control suggesting that some animals with axonal defects have defects in neuronal circuit formation leading to subtle changes in their behavioral response pattern.

Funding

This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) grant RGPIN- 2017-03942 awarded to H.H.

Conflicts of interest

None declared.

Data availability

Strains 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.

Supplemental material available at GENETICS online.