We maintained C. elegans strains under standard laboratory conditions on nematode growth medium agar (NGM: 0.25% Tryptone, 0.3% Sodium Chloride, 1 mM Calcium Chloride, 1 mM Magnesium Sulfate, 25 mM Potassium Phosphate (pH 6.0), 5 μg/ml Cholesterol, 1.7% Agar) plates with OP-50-1 Escherichia coli bacterial lawn at 15°C (Stiernagle, 2006), without antibiotics. All animals used in the experiments were hermaphrodites.

We acquired semaphorin and plexin mutants from Caenorhabditis Genetics Center (CGC) or the C. elegans National Bioresource Project of Japan (NBRP): ev778 (mab-20, null), tm729 (plx-2, null), ev715 (smp-1, null), ev709 (smp-2, null), tm10697 (plx-1, null), and evIs111 ([F25B3.3:GFP + dpy-20 (+)], pan-neural GFP expression). To allow imaging and microsurgery, we crossed males of NW1229 (evIs111), induced by 10-min exposure of L4 larvae to 10% ethanol (Lyons and Hecht, 1997), with null-mutant hermaphrodites to obtain knockout animals expressing GFP in the entire nervous system: TOL55 (ev715, evIs111, outcrossed x6), TOL57 (ev709, evIs111, outcrossed x6), TOL59 (tm10697, evIs111, outcrossed x1), and TOL62 (tm729, evIs111, outcrossed x1). All strains were verified by PCR upon arrival, after crosses, and at the end of the study. All generated strains and primer sequences for genotyping will be deposited with the CGC.

The reporter strain for plx-1p:EGFP (NW2339, 2,621 bp sequence immediately 5′ to the ATG start codon cloned into the multiple cloning site of pPD95_77; Dalpé et al., 2004) and plx-2p:GFP (NW1693, 4,529 bp sequence immediately 5′ to the ATG start codon cloned into the multiple cloning site of pPD95.75) were generous gifts from Dr Joseph Culotti (University of Toronto, Mt Sinai Hospital) and Dr Richard Ikegami (UC Berkeley), respectively. For unambiguous identification, we crossed each reporter strain with a NeuroPAL transgenic strain (OH15495; Yemini et al., 2021).

We tracked locomotion behavior of multiple animals over an agar surface (1.7% in NGM buffer), without food, as well as in liquid (NGM buffer). We recorded videos with a static multi-worm tracker, composed of three major parts, from top to bottom: 1) a CMOS camera (acA4024-29um, Basler) mounted with a fixed focal length lens (C Series 5 MP 35 mm 2/3″, Edmund Optics), and an infrared cut-off filter (SCOTT-KG3 M25.5 × 0.5, Edmund Optics); 2) a specimen stage for plates or slides; 3) a collimated Infrared LED light source (M850L3 and COP1-B, Thorlabs).

One day before the experiment, we transferred animals of the fourth larval stage (L4) onto a new plate with healthy OP-50-1 bacterial lawn. Ten to fifteen minutes before tracking, animals were transferred onto a 30 mm agar plate with no food or a 150 µL drop of NGM buffer, placed on a microscope slide. During tracking, animals moved freely, and we recorded multiple 25 Hz 15-s videos using Pylon Viewer (Pylon Camera Software Suite, Basler). We analyzed the videos with Tierpsy worm-tracker (Javer et al., 2018) that can track multiple animals and extract up to 726 features for each tracked trajectory. We used the Tierpsy post-processing user interface to merge tracked sections (trajectories) if those were erroneously split by the automatic tracking, and we rejected any trajectory shorter than 3 s, as well as ambiguous cases of animal proximity. Recording and Tierpsy analysis were done by undergraduate researchers, blinded to the animals’ genotype and injury condition. We analyzed the HDF5 output file produced by Tierpsy with a MATLAB script (code available upon request) to collect the mean speed and frequency values for each trajectory and then plotted the data and estimated confidence intervals between each group and its control with a freely available software for Estimation Statistics (https://www.estimationstats.com; Ho et al., 2019); that focuses on the magnitude of the effect (the effect size) and its precision. We also present statistical significance calculated with a two-sided permutation t-test to compare sham vs. injured groups, or one-way ANOVA with Tukey’s multiple comparisons test post hoc to compare genotypes (GraphPad Prism v9.2), included as p values in the text and as asterisks that denote levels of significance. We routinely use this tracking system to evaluate and compare wild type, injured, and uncoordinated mutant strains. We tracked all the knockout, transgenic, and wild type strains without injury to assess their baseline locomotion parameters. Further, we tracked locomotion to assess recovery 6, 12, and 24 h after microsurgery. For comparison, we also quantified locomotion parameters of sham-surgery groups for each genotype and time point. We treated the sham-surgery groups through the same protocol (including cooling and immobilization, see below), except for the exposure to the laser beam.

To reduce autofluorescence and straighten the animals we incubated fourth stage larvae (L4) in M9 buffer for 90 m and washed in the same buffer three times, incubated in 1 mM Levamisole (a paralytic nicotinic agonist, Sigma Aldrich) for 15 m, and fixed overnight at 4°C in 10% formalin solution, neutral buffered (SIGMA), then washed and mounted with Fluoromount-G (EMS), and allowed the slides to dry for at least 24 h before imaging. We used a laser scanning confocal microscope (Leica SP8; microscope: DM6000CS; objectives: Leica ×40/NA1.30 HC PL APO oil or Leica 63x/NA1.40 HC PL APO oil, with lateral resolutions of 223 nm and 207 nm respectively; laser lines: 405 nm, 561 nm, and 488 nm). We collected multiple optical slices (thickness optimized by the confocal software, ranging 0.343–0.345 µm for the ×63 objective, and 0.410–0.422 µm for the ×40 objective). To analyze morphology and cellular expression we constructed the maximum intensity projections for at least 10 animals of each strain and, in some cases, processed images to reduce background noise via the Leica Application Suite (LASX) software.

For unambiguous identification of VNC motoneuronal expression, we crossed each transcriptional reporter strain with a NeuroPAL transgenic strain and imaged the F1 progeny that express both transgenes. The NeuroPAL strains express an invariant color map across individuals, where every neuron is uniquely identified by its color and position (Yemini et al., 2021). We identified 29 motoneurons in three animals and rejected three motoneurons that expressed GFP but their location and NeuroPAL colors were ambiguous.

For laser microsurgery and associated microscopy, we mounted C. elegans hermaphrodites at L4 stage by placing them in a drop of ice cold, liquid 36% Pluronic F-127 with 1 mM levamisole solution and pressed them between two #1 coverslips (Melentijevic et al., 2017). We brought the coverslips to room temperature, to solidify the Pluronic F-127 gel and immobilize the animals. We used a Yb-fiber laser (100 pulses at 10 kHz repetition rate) to cut a single neurite with submicron precision and no discernable collateral damage (Harreguy et al., 2020; Harreguy et al., 2022). We took images immediately before and after the lesion to visually verify the microsurgery. In some cases, multiple laser exposures were necessary to disconnect a neurite. We disconnected the ventral-dorsal commissures (White et al., 1976) of all motoneurons that we were able to identify by their relative position (at least six per animal), at about 45 μm away from the VNC. We assessed neuronal regeneration 24 h (following most regeneration studies in C. elegans, since Yanik et al., 2004) after microsurgery on the same microscope and imaging system in at least six neurons per animal in at least 15 animals for each condition. We considered neurites regrown when a new branch was observed extending from the proximal segment of the injury site (Harreguy et al., 2020; Harreguy et al., 2022). When the branch extended to the distal segment or the target of the pre-injury neurite, we considered it regrown and reconnected. We used Fisher Exact on 2 × 3 contingency table to compare the fraction of observed neurites that regrew or reconnected. We used ImageJ (FIJI v.1.52) and LASX (Leica) for image processing and visualization, and Prism (GraphPad v.9.2.0) for statistical analysis and plotting.

We analyzed the contribution to locomotor behavior of each of C. elegans three semaphorins and two plexins (Figure 1A) by comparing the speed and frequency of locomotion of knockout (ko) mutant strains to that of wild type animals. During crawling on agar (Figure 1B), all strains translocated significantly slower compared to 204 ± 54 μm/s of wild type (speed and p values were: plx-1 123 ± 37, p < 0.0001; smp-1 83 ± 33, p < 0.0001; smp-2 123 ± 35, p < 0.0001; plx-2 168 ± 41, p = 0.0011; mab-20 186 ± 51, p = 0.0016); and the undulation frequency of all strains was reduced compared to 0.43 ± 0.08 Hz of wild type (frequency and p values were: plx-1 0.29 ± 0.07, p = 0.0497; smp-1 0.19 ± 0.09, p < 0.0001; smp-2 0.25 ± 0.06, p < 0.0001; plx-2 0.36 ± 0.09; mab-20 0.36 ± 0.08). Relative to crawling, swimming speed and frequency were less affected by the absence of plexins or semaphorins (Figure 1B), only plx-1(ko) and plx-2(ko) animals translocated slower than 243 ± 88 μm/s of wild type (speed and p values were: plx-1,196 ± 63, p = 0.003; smp-1,209 ± 94; smp-2 234 ± 63; plx-2 172 ± 38, p < 0.0001; mab-20 277 ± 52); only smp-1(ko) animals undulated at higher frequency compared to 1.34 ± 0.27 Hz of wild type (frequency and p values were: plx-1 1.53 ± 0.55; smp-1 1.62 ± 0.44, p = 0.0014; smp-2 1.17 ± 0.29; plx-2 1.14 ± 0.22; mab-20 1.23 ± 0.22). The largest reduction of crawling speed and frequency was in smp-1(ko) animals that were also the only genotype to exhibit a change (increase) in undulation frequency during swimming.

We focused further analysis on the plexins (plx-1 and plx-2), because as the only receptors, segregating membrane-bound and secreted pathways, they provide a comprehensive and specific manipulation of these pathways, as well as the identity of the cellular targets (Fujii et al., 2002).

We used confocal microscopy to image at least five intact four instar (L4) larvae of each plexin-knockout and wild type strain, expressing pan neuronal green fluorescent protein (GFP), with emphasis on neuron-rich areas around head, tail, the ventral nerve cord, pharynx, and vulva, and particularly at the commissures of motoneurons (Figure 2). We did not observe any morphological differences between mutant and wild type animals in any of these regions.

We imaged transcriptional reporters for plx-1p and plx-2p in order to identify their neuronal expression in the ventral nerve cord (VNC). GFP under the plx-1p promoter (Figures 3A,B) was mostly expressed in non-neuronal tissue including the pharyngeal muscle, the body-wall muscle in the head and along the body, and vulva muscle. We did not find expression in the nervous system of plx-1p:GFP, although a translational reporter was reported to express in the axon of a motoneuron at the base of the tail, namely DA9, of the embryo and L1 larva (Mizumoto and Shen, 2013). GFP under the plx-2p promoter was expressed by neurons in the head and tail (Figure 3C), as well as in motoneuron in the VNC (Figure 3D). Most expressing motoneurons were AS and DA classes (14 and 9, respectively, from three animals), six motoneurons of other classes, namely DB (3), VA (2), and VB (1) also expressed GFP. Both AS and DA extend commissures that were the targets for microsurgery, from the VNC to the dorsal nerve cord on the opposite side of the animal.

We disconnected 156 commissural neurites of motoneurons of wild type and plexin knockout mutant animals with laser microsurgery (Harreguy et al., 2020; Harreguy et al., 2022). These lateral processes extend to connect the ventral and dorsal nerve cords and when multiple processes are disconnected, locomotion is impaired (Yanik et al., 2004). When we examine the same neurite after 24 h, some regrew by sprouting a growth cone from the proximal segment and some of those reconnected to the distal segment or the dorsal nerve cord (Figure 4A). In the wild type, 38 of 73 neurites regrew (0.52 ± 0.11) and only five of those (0.07 ± 0.058) reconnected (Figure 4B). The plexin knockout mutants exhibited significantly more regrowth (p = 0.049), 33 of 47 (0.7 ± 0.13) for plx-1(ko) and 26 of 36 (0.72 ± 0.15) for plx-2(ko). Reconnection happened significantly more (p < 0.0001) in the plexin knockout strains: in plx-1(ko), 13 of the regrown neurites (0.28 ± 0.13) and in plx-2(ko), 20 of the regrown neurites reconnected (0.56 ± 0.16).

Six hours after microsurgery, wild type and mutant animals moved slower than sham-treated animals of the same genotype (sham vs. injured: WT 79 ± 39 vs. 41 ± 20 μm/s, p = 0.004; plx-1(ko) 72 ± 18 vs. 31 ± 25 μm/s, p < 0.0001; plx-2(ko) 70 ± 22 vs. 35 ± 21 μm/s, p = 0.0001; Figure 4C, top). Twelve hours after microsurgery, the mean locomotion speed of wild type animals has recovered to levels comparable to sham-treated, while mutant animals moved slower than their sham-treated controls (sham vs. injured: WT 79 ± 28 vs. 63 ± 36 μm/s; plx-1(ko) 84 ± 22 vs. 58 ± 26 μm/s, p = 0178; plx-2(ko) 106 ± 25 vs. 51 ± 26 μm/s, p = 0.0001; Figure 4C, middle). Subsequently, 24 h after microsurgery, mean locomotion speed has recovered to levels comparable to sham-treated animals for all groups (sham vs. injured: WT 115 ± 45 vs. 145 ± 49 μm/s; plx-1(ko) 111 ± 44 vs. 146 ± 51 μm/s; plx-2(ko) 120 ± 35 vs. 121 ± 20 μm/s; Figure 4C, bottom).