Function of cell adhesion molecules in differentiation of ray sensory neurons in C. elegans

Abstract For proper functioning of the nervous system, it is crucial that neurons find their appropriate partners and build the correct neural connection patterns. Although cell adhesion molecules (CAMs) have been studied for many years as essential players in neural connections, we have yet to unravel the code by which CAMs encode synaptic specificity. We analyzed the effects of mutations in CAM genes on the morphology and synapses of a set of sensory neurons in the Caenorhabditis elegans male tail. B-type ray sensory neurons express 10 genes encoding CAMs. We examined the effect on axon trajectory and localization of pre-synaptic components in viable mutants of nine of these. We found axon trajectory defects in mutants of UNC-40/DCC, SAX-3/ROBO, and FMI-1/Flamingo/Celsr1. None of the mutations caused loss of pre-synaptic components in axons, and in several the level even appeared to increase, suggesting possible accumulation of pre-synapses. B-type sensory neurons fasciculate with a second type of ray sensory neuron, the A-type, in axon commissures. We found a CAM expressed in A-type functions additively with a CAM expressed in B-type in axon guidance, and lack of a CAM expressed in B-type affected A-type axon guidance. Overall, single and multiple mutants of CAM genes had limited effects on ray neuron trajectories and accumulation of synaptic components.


Introduction
Since the proposal that the pattern of neural connections is controlled by individual identification tags on the cell surfaces of neurons (Sperry 1963), a number of cell adhesion molecules (CAMs) that regulate axon outgrowth and synapse formation have been identified through forward genetic and other methods (Hedgecock et al. 1990;Zallen et al. 1998;Yamagata et al. 2002;Shen and Bargmann 2003;Salzberg et al. 2013;Carrillo et al. 2015;Kurshan et al. 2018;Philbrook et al. 2018;Südhof, 2018;Kurshan and Shen 2019;Taylor et al. 2021). Despite efforts in the last few decades, we have yet to break the code by which CAMs help neurons find their appropriate partners. This is due in part to the redundant and pleiotropic functions of CAMs. There is limited information on expression patterns of CAMs in circuits of known connectivity, which has limited us in studying these parallel functions.
The nematode Caenorhabditis elegans has been a well-studied model organism for neuronal development because of its invariant developmental pattern and fully reconstructed neural and synaptic connectivity (White et al. 1986;Cook et al. 2019). Because of this feature, C. elegans has played an important role in studies of the relationship between CAMs and development (Yamagata et al. 2002;Shen et al. 2004;Salzberg et al. 2013;Kim and Emmons 2017;Shah et al. 2017;Lázaro-Peña et al. 2018;Philbrook et al. 2018;Taylor et al. 2021) In this study, we focused on a set of male-specific sensory neurons to investigate the relationship between CAMs, neurite trajectory, and synapse formation.
The C. elegans male has a set of nine bilateral pairs of sensory structures known as rays projecting from the tail embedded within a cuticular fan (Fig. 1a). Each ray has a similar structure consisting of the sensory endings of two sensory neurons of two types designated A-type and B-type (Sulston et al. 1980). In addition, each of the nine ray pairs has a distinct identity affecting the neurotransmitters, synaptic targets, and functions of the constituent neurons (Baird et al. 1991;Chow and Emmons 1994;Lints et al. 2004). As part of a large cohort of male-specific neurons and muscles, the ray sensory neurons are born during the L3 larval stage and differentiate through the L4 and into young adulthood (Sulston et al. 1980). Their cell bodies are located in a pair of bilateral lumbar ganglia. To reach their synaptic targets, ray neurons extend axons through circumferential commissures and into the pre-anal ganglion, where their growing processes branch considerably and synapse onto a number of post-synaptic targets ( Fig. 1a and b). In this study, we focus on the role of CAMs in determining the trajectories and synapse formation of the B-type ray neurons (RnB, n = 1-9, designating the nine rays).
We have determined the expression patterns of the cell adhesion genes encoded in the C. elegans genome across the C. elegans posterior nervous system (in preparation). In our dataset, the B-type ray neurons collectively express 10 different CAMs, while the A-types altogether express 21 (Fig. 1c). The pattern is left/right symmetric-in each case, the left/right homologous neurons had the same expression, and it is ray and neuron-type specific-only two of the genes, unc-40 and pan-1, are expressed in both A and B types. In this study, we examined the effects of mutations in the B-type-expressed genes as well as several of the A-type-expressed genes on the ray B-type axon trajectories and synapse formation, using reporters that allow us to simultaneously score both the neurite trajectory and the localization of pre-synaptic components. In a previous forward genetic screen, it was found that mutations in unc-40/DCC and sax-3/ROBO prevented the outgrowth of the B-type axons through the circumferential commissures (Jia and Emmons 2006). In a second study, B-type neuron synapse formation was affected by mutations in CAMs neurexin and neuroligin as well as by mutation in a modification enzyme that affects the sulfation pattern of sugar residues on the matrix protein glypican (Lázaro-Peña et al. 2018). In this study, we demonstrated the effects of mutations in fmi-1/flamingo and sax-7/ L1CAM on B-type neuron trajectory. As sax-7 is expressed in A-type but not in B-type neurons, the sax-7 result indicates that sax-7 functions in A-type to affect B-type neuron morphology. Overall, however, in our nearly complete survey, we found limited effects of both single and multiple mutations in most of the expressed CAM genes, suggesting the robustness of the mechanisms underlying neuronal development. , a fluorescent image of RnB neuron processes marked by pkd-2 prom ::GFP (middle), and a schematic image of a male ray structure and RnB processes (bottom). b) Schematic image of a lateral view of the male tail (middle), schematic to explain distribution of RnA, RnB, and RnSt (ray structure cells) (top), and lateral view of the processes of R1-R9BL, as well as pre-synapses (red spheres) and gap junctions (green spheres) constructed from Electron Microscope (EM) images (https://www.wormwiring.org/). Each sphere size represents the number of EM sections containing each synapse. (bottom). The broken circle marks LG (lumber ganglion) where the cell bodies of ray neurons are located, and the broken square marks PAG (pre-anal ganglion) where the ray neurons make many synaptic connections to other neurons. Dendritic regions are shown in light lines and axonal regions are shown in dark lines. c) Expression of transcriptional reporters for CAMs in the RnA and RnB neurons. One hundred CAM genes have been examined and the genes having expression in the ray neurons are shown as red boxes.

Animal maintenance
All animals were maintained using standard methods (Brenner 1974). Briefly, all animals were grown on NGM plates seeded with Escherichia Coli OP50 at 20°C. All strains contain the him-5(e1490) mutation to increase the male population in selfing worms (Broverman and Meneely 1994) unless otherwise specified. For all strain information, see the Supplementary file.

Statistical analyses
All data were analyzed by Fisher's exact test or multiple comparison test using Graph-pad Prism 9.0 statistical software (RRID: SCR_002798).

Reporter genes used in this study
Two integrated transgenes were used to visualize the trajectories and synapses of ray B-type sensory neurons (Fig. 2a). Both utilized the promoter of the polycystin-homolog gene pkd-2 (Barr and Sternberg 1999) to drive expression of fluorescent markers in the RnB neurons. In addition to the RnB's, this promoter drives expression in tail sensory neuron HOB and four CEM sensory neurons in the male head (Barr and Sternberg 1999).

Microscopy and image analysis
Animals were anesthetized with 20 mM levamisole and mounted on their back on 5% agar pads on glass slides. We used 1-day-old males to observe RnB axon morphology. To observe anteriorposterior and circumferential axons, animals were observed with fluorescence microscopy (Zeiss Axio Imager A1) with 100× magnification. To observe diagonal commissures, we utilized Nikon CSU-W1spinning disk confocal microscopy with 100× magnification.

RnB synapse imaging
To visualize the pre-synaptic pattern of RnB neurons, we utilized EM1554 bxIs30[pkd-2 prom ::gfp; pkd-2 prom ::RAB-3-Mcherry; ttx-3p::GFP] II; him-5(e1490)V and EM1890 bxIs33[pkd-2 prom :: 3xnovoGFP-CLA1, pkd-2p::RAB-3-Mcherry, pkd-2p::BFP] I; him-5(e1490)V. Animals were anesthetized with 20 mM levamisole and mounted on their back on 5% agar pads on glass slides. Images were obtained with Nikon CSU-W1spinning disk confocal microscopy with 100× magnification with 5 nm thickness. All images were obtained with same exposure time for the same wave length. (For EM1554, 488 nm laser power 15%, exposure time 50 ms, 561 nm laser power 25%, exposure time 500 ms. For EM1855, 405 nm laser power 30%, exposure time 600 ms, 488 nm laser power 20%, exposure time 200 ms.) To quantify the protein levels of pre-synaptic markers, all planes for each animal were z-stacked. The circle ROI, or region of interest, contained 205 pixels and was set on the synaptic ring, followed by measuring mean signal intensity. We also quantified the signal intensity of the cytoplasmic fluorescent markers in the same ROI and calculated the pre-synapse/cytoplasmic signal intensity ratio.

Axon trajectory phenotypes of strains carrying single mutations in CAM genes
The trajectories of the RnB neuron processes in wild type, visualized by the cytoplasmic marker, are shown in Fig. 1a and schematically in Fig. 2b. Based on the characteristics of RnB axon morphology, we classified the trajectory defects in mutants into one of four categories ( The CAM genes and mutations studied in this work are shown in Table 1. We determined the effect of mutations in all of the B-neuron-expressed genes except pan-1, which is embryonic lethal. We observed trajectory defects in three of the genes tested, unc-40, sax-3, and fmi-1. Unc-40 mutants had significant abnormalities in the dorsoventral guidance system for the formation of all the ray commissures ( Fig. 2e, f and h), as reported previously (Hedgecock et al. 1990;Jia and Emmons 2006). Significant defects were also observed in anteroposterior guidance and posterior commissures (also referred to as diagonal commissures in Fig. 1b, the commissures containing the processes of ray 8 and ray 9 neurons) ( Fig. 2d and g). Although posterior commissures were reported to be unaffected in a previous paper (Jia and Emmons 2006), some were missing when we observed with a confocal microscope. Sax-3 mutants had abnormal anteroposterior and circumferential axonal projections ( Fig. 2d-f). The abnormal anterior axonal projection was often seen in combination with a deletion of the circumferential axon or R1B (Fig. 2h). The lack of circumferential R1B axon seems to be due to a failure of turning of the anterior-posterior axon from the cell body. In sax-3 mutants, the posterior commissures from rays 8 and 9 were also frequently missing ( Fig. 2g and h). Mutation in fmi-1/flamingo resulted in a deficit of the circumferential commissures (Fig. 2e, f, and h), and most notably, an absence of the posterior commissures in almost all animals ( Fig. 2g and h).

Axon trajectory phenotypes of strains carrying multiple mutations in CAM genes
In view of the lack of effects of single mutations in most of the genes, we asked whether this was due to redundancy of their activities by examining double and multiple mutants. We found that this does not appear to be the case. Since rig-3 and lron-7 loci are close and we could not create this double, we created two quintuple mutants (hmr-1; cdh-12 lron-13; nrx-1; lron-7 and hmr-1; cdh-12 lron-13; nrx-1; rig-3). In all of the strains carrying multiple mutations in genes with no single effect, there were no significant abnormalities (Fig. 3a-d).
Next, we crossed the strains with observable phenotypes, sax-3 and fmi-1, into the strains with no obvious phenotype to look for genetic interactions (enhancement or suppression) (Fig. 3e-j). We found that the hmr-1 mutant significantly enhanced the axon guidance defects in the circumferential commissures in a sax-3 mutant ( Fig. 3f and g), indicating that hmr-1 and sax-3 have redundant functions. In addition, lron-13 and lron-7 suppressed axon guidance defects in sax-3 and fmi-1, respectively ( Fig. 3e and i). These mutants have not been reported to have any significant phenotypes, but they may be involved in the regulation of axon guidance in cooperation with other CAMs.

Effects of mutations in genes expressed in RnA neurons on axon trajectories of RnB neurons
RnA and RnB neuron processes run side by side from the dendrites in the rays through circumferential commissure fascicles until they enter the pre-anal ganglion (Fig. 1b) (Jia and Emmons 2006;Jarrell et al. 2012;Cook et al. 2019). We, therefore, examined whether CAMs expressed in RnA neurons affect axon guidance in RnB neurons. Although many CAMs are expressed in the RnA neurons (Fig. 1c), we focused on the effects of four genes (casy-1, bam-2, sax-7, and tol-1) that are widely expressed in most of the RnA neurons. Single mutations in casy-1, bam-2, and sax-7 showed no abnormalities in the morphology of the RnB neurons ( Fig. 4a-d). In the tol-1 mutant, there was a significant increase in the number of commissures from Ray2-7 (Fig. 4c). This result suggests that tol-1 may be involved in the fasciculation of the circumferential commissures. SAX-7 has been reported to interact with SAX-3 in axonal fasciculation (Chen et al. 2019). We therefore asked whether these proteins acted together in a single pathway for ray fasciculation or separately by determining whether mutation in sax-7, expressed in RnA, would enhance the defect of RnB axon trajectory in a double mutant with sax-3, expressed in RnB (Fig. 4e-g). We found the sax-7 mutation significantly enhanced the RnB abnormalities in the circumferential commissure of the sax-3 mutant (Fig. 4f). This suggests that these two proteins have independent and apparently somewhat redundant functions in promoting fasciculation of their respective neuron processes in the commissures.
In view of this possibility, we tested for a possible reciprocal effect on RnA trajectories of mutations in those RnB-expressed genes that affected RnB trajectories. We visualized a subset of RnA neurons using DAT-1p::GFP expressed in R5A, R7A, and R9A (Fig. 4h). We observed abnormalities in RnA neurite trajectories in both sax-3 and fmi-1 mutants similar to those observed in the RnB neurons ( Fig. 4i and j). Thus, each of the two neuron types appears to promote the trajectories of the other type, consistent with their co-fasciculation in the outgrowing commissures.

Effects of mutations in CAM genes on synapse formation
CAMs are involved not only in axon guidance but also in synapse formation (Shen and Bargmann 2003; Dalva et al. 2007; Kim and Emmons 2017;Kurshan et al. 2018;Lázaro-Peña et al. 2018;Philbrook et al. 2018;Südhof 2018). We visualized RnB synapses with fluorescent markers localized respectively to synaptic vesicles (RAB-3) and pre-synaptic densities (CLA-1) (Fig. 2a). RAB-3 signals and CLA-1 signals are mostly co-localized but sometimes are seen in different distributions. This suggests that synaptic vesicles and active zone proteins may be localized separately. Both reporter genes showed a pattern of fluorescence consistent with RnB pre-synapses as described by electron microscopy (Jarrell et al. 2012) (Fig. 5a and  b). Most RnB synapses are in the pre-anal ganglion, where the densest fluorescence appears as a ring, dubbed here the "synaptic ring." In addition, consistent with the mixed axonal/dendritic character of most C. elegans neurons, they receive a similar amount of input from other neurons. These post-synapses are not visualized here. Multiple pre-synapses were also observed within the pre-anal ganglion along the midline axon and in the lumbar ganglia, but few in circumferential commissures and, in the case of the RAB-3 reporter, almost none inside ray structures. By contrast, the GFP::CLA-1 reporter indicated the presence of pre-synaptic densities inside ray structures (Fig. 5b). We examined EM images of the rays and found synapse-like structures with high electron density within the ray dendrites in the rays (Fig. 5c). Altogether, in the EM reconstruction of a single adult male, the RnB neurons form 1,143 chemical synapses onto 123 other sensory neurons, including other ray neurons, interneurons, and motor neurons, most of which are in the pre-anal ganglion (Jarrell et al. 2012) (Fig. 1b).
To assay the amount of synapse formation, in order to control for different levels of gene expression we determined the ratio of signal intensity between the synaptic signals (RAB-3::mCherry for bxIs30II, RAB-3::mCherry and GFP::CLA-1 for bxIs33I) and the cytoplasmic signals (GFP for bxIs30II., BFP for bxIs33I) (Fig. 5d-f). In unc-40 and sax-3 mutants, both pre-synaptic marker ratios were increased in circumferential commissures (Fig. 5, d and g). This suggests accumulation of synaptic proteins in processes that have failed to reach their targets. In the unc-40 mutant, the RAB-3-MCherry ratio decreased in the synaptic ring region but the GFP-CLA1 ratio increased ( Fig. 5e and h). Thus CLA-1 protein may accumulate within those few processes that reach the target area. In the hmr-1(gk3258) mutant, the RAB-3::mCherry signal ratio was decreased in the synaptic ring, whereas there was no significant increase or decrease within circumferential commissures ( Fig. 5d-f). (The CLA-1 reporter could not be tested because of proximity of the mutation to the transgene insertion site.) Thus, hmr-1 may be involved in accumulation of synaptic vesicles. Besides these, there were no other deviations from wild type ratios in the single and multiple mutants. In none of the mutants was there any significant presynapse signal ratio change in midline axons (Fig. 5f).
These results indicate a limited effect of mutations in RnB-expressed CAM genes on accumulation of pre-synaptic components within the neuronal processes. Whether this indicates these CAM genes play no role in recognition of synaptic targets and formation of synapses or that synaptic components accumulate in the absence of target recognition and synapse formation cannot be determined.

SAX-3 regulates circumferential axon guidance in a slt-1 independent manner
For two of the genes we studied, sax-3 and fmi-1, interacting partners or pathway components are known in other systems. For SAX-3/ROBO, it has been demonstrated in many species that Slit, a secreted LRR protein, functions as a ligand (Brose et al. 1999;Zallen et al. 1999;Hao et al. 2001  muscles. We tested whether slt-1 is involved in RnB axon guidance and found that slt-1 null mutants showed abnormal anterior axon projections and loss of circumferential axons, which were also observed in sax-3 mutants ( Fig. 6a and b). The absence of slt-1 did not enhance the sax-3 mutant phenotype in a double mutant, consistent with SLT-1 functioning as a ligand for SAX-3. By contrast, slt-1 single mutation did not show any significant abnormality in the number of commissures from R2-7B (Fig. 6c). This result and the lesser overall severity of the slt-1 null mutant phenotype compared to the sax-3 null mutant phenotype indicate that molecules other than SLT-1 may function as ligands for SAX-3 in RnB axon guidance.

FMI-1 regulates RnB axon guidance both in the PCP pathway and PCP-independent pathway
In Drosophila, Flamingo, a homolog of fmi-1, was previously reported to work together with Van Goh and Frizzled to regulate the PCP (planar cell polarity) pathway (Usui et al. 1999). We tested whether the PCP pathway genes are involved in RnB axon guidance ( Fig. 6d-f). The lack of the circumferential axon from R1B was occasionally seen in a vang-1 mutant (homolog of Van Goh) (Fig. 6e). This phenotype was not enhanced by fmi-1 null mutant (Fig. 6e), indicating fmi-1 and vang-1 work in the same pathway.
On the other hand, the fmi-1; vang-1 double mutant showed a significantly more severe phenotype than the fmi-1 mutant alone in circumferential commissures from R2-7 (Fig. 6f). These results suggest that fmi-1 and vang-1 regulate the morphology of RnB neurons within the same pathway as well as in parallel pathways.

Discussion
C. elegans ray sensory neurons provide an excellent opportunity to identify the factors that determine neuronal wiring selectivity. Genes involved in formation of the two neuron types and that specify ray neuron identities have been described, including transcription factors and signaling pathways (Baird et al. 1991;Chow and Emmons 1994). Regulatory genes must act by controlling the expression of effector genes that function during morphogenesis. Here, we examined the class of genes thought to specify the cell surface labels that govern cell-cell recognition for synapse formation. Like other neurons in the C. elegans nervous system, ray sensory neurons express multiple CAMs, including many conserved across species (Hahn and Emmons 2003;Taylor et al. 2021;this laboratory unpublished). Reflecting an aspect of ray identity, the combination expressed by the neurons in each ray pair differs (Fig. 1c). The same genes are expressed in other combinations across the entire set of neurons and muscles in the posterior nervous system, indeed, across the entire animal in all tissues, reflecting their presumptive function as neutral cell labels, that is, labels not associated with any particular cell type or cell fate (Taylor et al. 2021; this laboratory unpublished). , and number of circumferential commissures from R2-7B (c) for sax-3, slt-1, and sax-3; slt-1 mutants. d-f) Percentage of defective anterior-posterior axon guidance (d), percentage of defective circumferential R1B axon guidance (e), and number of circumferential commissures from R2-7B (f) for fmi-1, vang-1, and fmi-1; vang-1 mutants. ****P < 0.0001, **P < 0.01, *P < 0.05, n.s P > 0.05, Fisher's exact test.
Our primary result reported here is that mutations in the CAM genes expressed by the ray neurons have surprisingly limited effects on two aspects of ray neuron phenotype: axon trajectories and accumulation of pre-synaptic proteins in axonal processes. General absence of effects on axon trajectories reflects the robustness of the mechanisms that guide formation of the structure. For example, we found that SAX-3/ROBO and SAX-7/L1CAM each appear to independently contribute to fasciculation of the RnB and RnA circumferential processes.
In this study, we analyzed the known interacting partners for sax-3 and fmi-1, and found that the mutant of slt-1, a well-known SAX-3 interacting partner, showed a weaker phenotype compared to sax-3 mutants, strongly suggesting the existence of ligands of SAX-3 other than SLT-1. The NELL protein has recently been reported as a ligand for ROBO (Yamamoto et al. 2019;Pak et al. 2020), but no nematode homolog has been reported so far. We also found that lack of vang-1/VanGoh enhanced the fmi-1 phenotype in circumferential commissures from R2-7, suggesting that vang-1 and fmi-1 function in a parallel manner and the existence of a ligand for FMI-1 other than VANG-1. Further research on ray neuron development thus may lead to the discovery of novel ligands for sax-3 and fmi-1.
We observed increased synaptic proteins in circumferential commissures in some mutants. The accumulation of synapse proteins was examined in three different compartments. The amounts of synaptic proteins at the three locations did not coincide. For example, in the sax-3 mutant, synapse protein accumulation was increased in the circumferential commissure, but not in the synaptic ring and longitudinal commissure (Fig. 6d-f). This may suggest that the degree of contribution of CAMs to synaptogenesis varies from compartment to compartment within the same nerve.
There is precedent for synapses between sensory dendrites as we report here between ray sensory neuron dendrites. Pre-synaptic densities are observed within the sensory amphid bundles in the head (Cook et al. 2019). The functions of such dendrite-dendrite synapses are unknown but may simply be a part of the complex set of connections among sensory neurons seen broadly across the connectome. Alternatively, they could represent sites of release of extracellular vesicles, which occurs from the ray neurons as well as other C. elegans sensory dendrites (Wang et al. 2014).
Ray sensory neurons send axonal growth cones into a preexisting, complex neuropil where they seek out and find their postsynaptic partners. The question we have addressed here is how molecular cell-cell recognition occurs to guide this synaptogenesis. Previously we have shown that elimination of a specific sugar modification on the heparan sulfate proteoglycan glypican affected the formation of synapses between the ray B-type neurons and three of their major post-synaptic targets, the three malespecific EF interneurons (Lázaro-Peña et al. 2018). In a background containing a mutation in a gene encoding a sugar modification enzyme, the amount of synapse formation to the EF neurons was reduced although not eliminated. Intriguingly, the level of fluorescently labeled RAB-3 within the RnB target area increased in this background. Effects on accumulation of synaptic proteins probably reflects the fact that assembly and transport of such components to sites of expected synapse formation may be to some degree independent of target recognition and synapse formation (Kurshan and Shen 2019). The correlation of apparent accumulation of pre-synaptic components with loss of synapse formation in this previous example suggests that the apparent accumulation observed here could likewise indicate a loss of synapses. The question can be addressed with the methods that demonstrate interaction of pre-and post-synaptic cells, GRASP or iBLINC (Feinberg et al. 2008;Desbois et al. 2015).