Self-avoidance is a conserved mechanism that prevents crossover between sister dendrites from the same neuron, ensuring proper functioning of the neuronal circuits. Several adhesion molecules are known to be important for dendrite self-avoidance, but the underlying molecular mechanisms are incompletely defined. Here, we show that FMI-1/Flamingo, an atypical cadherin, is required autonomously for self-avoidance in the multidendritic PVD neuron of Caenorhabditis elegans. The fmi-1 mutant shows increased crossover between sister PVD dendrites. Our genetic analysis suggests that FMI-1 promotes transient F-actin assembly at the tips of contacting sister dendrites to facilitate their efficient retraction during self-avoidance events, probably by interacting with WSP-1/N-WASP. Mutations of vang-1, which encodes the planar cell polarity protein Vangl2 previously shown to inhibit F-actin assembly, suppress self-avoidance defects of the fmi-1 mutant. FMI-1 downregulates VANG-1 levels probably through forming protein complexes. Our study identifies molecular links between Flamingo and the F-actin cytoskeleton that facilitate efficient dendrite self-avoidance.
Dendrite self-avoidance prevents crossover between sister dendrites from the same neuron, maximizing coverage of the sensory territory of individual dendrites with minimal overlapping (Zipursky and Grueber, 2013). In mammals, defects in dendrite self-avoidance are associated with abnormal gait patterns or impaired directional sensitivity in vision, suggesting that self-avoidance is essential for the proper functioning of the neural circuitries (Gibson et al., 2014; Kostadinov and Sanes, 2015). Many molecules have been identified as important regulators for dendrite self-avoidance in diverse species, including Down's syndrome cell adhesion molecules (Hughes et al., 2007; Matthews et al., 2007; Soba et al., 2007), γ-protocadherins (Lefebvre et al., 2012), integrins (Han et al., 2012; Kim et al., 2012), the atypical cadherin Flamingo and the transmembrane polarity protein Van Gogh (Matsubara et al., 2011), the immunoglobulin superfamily protein Turtle (Long et al., 2009), the secreted proteins Netrin and Slit (Gibson et al., 2014; Smith et al., 2012; Sundararajan et al., 2019), the furin-like protease KPC-1 (Salzberg et al., 2014) and the Wnt-secretory factor Wntless (Liao et al., 2018). The signaling pathways through which these molecules control self-avoidance are incompletely defined.
A crucial observation made by live imaging in the nematode Caenorhabditis elegans indicates that dendrite self-avoidance occurs through contact-dependent dendrite retraction (Smith et al., 2012). This finding is consistent with the fact that the majority of self-avoidance molecules are cell-membrane proteins and several of these display homophilic interactions (Goodman et al., 2016; Rubinstein et al., 2015; Wojtowicz et al., 2007). It is less clear what drives subsequent dendrite retraction after the initial transient contact between sister dendrite branches. Recently, F-actin assembly has emerged as a potential cytoskeletal platform that integrates signaling from the plasma membrane to promote self-avoidance. In C. elegans, UNC-6/Netrin signaling and Wntless engage F-actin assembly to promote dendrite self-avoidance (Liao et al., 2018; Sundararajan et al., 2019). The Netrin receptor UNC-40/DCC (deleted in colorectal cancer) contains a conserved motif for direct interaction with the WAVE regulatory complex (WRC), providing a molecular link between Netrin signaling and F-actin assembly (Chen et al., 2014). Wntless genetically interacts with neural Wiskott–Aldrich syndrome protein (N-WASP), a conserved actin regulator (Derivery and Gautreau, 2010), further strengthening the idea that F-actin dynamics play an important role in mediating contact-dependent dendrite repulsion (Liao et al., 2018).
Flamingo is an atypical cadherin that has seven transmembrane domains, making it structurally reminiscent of a G protein-coupled receptor (GPCR) (Langenhan et al., 2016). In Drosophila, Flamingo promotes the self-avoidance of sister dendrites from the same larval class IV dendrite arborization (da) neuron (Matsubara et al., 2011). Flamingo binds the LIM domain protein Espinas and also genetically interacts with the small GTPase RhoA and Van Gogh, molecules that govern planar cell polarity (PCP) across epithelial tissues (Matsubara et al., 2011). As RhoA is a well-studied regulator of F-actin dynamics, it is tempting to speculate that Flamingo promotes dendrite self-avoidance by engaging F-actin assembly. In this study, we provide experimental evidence that FMI-1, the C. elegans Flamingo, genetically interacts with WSP-1/N-WASP and probably regulates dendrite self-avoidance by orchestrating spatially and temporally defined F-actin activity at the dendrite tips. Our data further suggest that FMI-1 antagonizes, rather than collaborates with, VANG-1/Van Gogh in dendrite repulsion. These findings provide insights into the molecular mechanisms by which Flamingo shapes the fine architecture of dendrite arborization.
C. elegans FMI-1/Flamingo regulates dendrite self-avoidance
To understand how dendrite self-avoidance is regulated at the molecular and cellular level, we focused on PVD, a bilaterally symmetric multidendritic nociceptive neuron in C. elegans that has extensive dendrite arborization (Fig. 1A). PVD elaborates peripheral branches orthogonal to the more proximal dendrites, forming a stereotyped, menorah-like dendrite morphology (Albeg et al., 2011; Oren-Suissa et al., 2010; Smith et al., 2010). The horizontal 3° branches display robust self-avoidance, leaving gaps of variable length with minimal contact or crossover between sister 3° dendrites (Smith et al., 2012). The C. elegans gene fmi-1 encodes Flamingo, an atypical cadherin with enormous extracellular domains, seven transmembrane domains and a short cytoplasmic tail (Fig. 1B) (Steimel et al., 2010). The fmi-1(rh308) allele, a nonsense mutation predicted to truncate most of the FMI-1 protein (Steimel et al., 2010), displayed increased self-avoidance defects compared with those of the control, with 3° dendrites in contact with each other and missing gaps between them (Fig. 1B,C). Another nonsense allele, hd121 (Steimel et al., 2010), showed similar phenotypes (Fig. 1B,D), suggesting that self-avoidance defects in these mutants are probably caused by loss of fmi-1 gene activity. Trans-heterozygotes between fmi-1(rh308) and nDf42, a deficiency chromosome that deletes the entire fmi-1 locus, showed self-avoidance defects similar to those of the fmi-1(rh308) homozygotes (Fig. 1D), suggesting that rh308 is a null allele of fmi-1. Gross PVD dendrite morphology of the fmi-1(rh308) mutant was indistinguishable from that of the control (Fig. S1). As rh308 is a putative null mutation, we focused the rest of our investigation on this allele.
Previous studies suggest that growing PVD 3° dendrites briefly contact each other and promptly retract, typically within 3 to 5 min after the initial contact (Liao et al., 2018; Smith et al., 2012). To gain insight into the cellular basis of self-avoidance defects in the fmi-1 mutant, we performed live imaging by spinning-disk confocal microscopy in third-stage larvae (L3) when dendrite self-avoidance is being established (Smith et al., 2012). Consistent with the published literature, we found that contact of 3° dendrites was resolved by dendrite retraction within 3 min in more than 60% of the contact events, with more than 90% of contact events resolved in 10 min (Fig. 1E,F, Movie 1). By contrast, in the fmi-1 mutant, 3° dendrites remained in contact for 10 min or longer in more than 60% of the events (Fig. 1E,F, Movie 2). As defects in self-avoidance scored in late L4 fmi-1 animals are considerably less penetrant than dendrites in contact in our live-imaging experiments, it is possible that some contacting events are resolved at later time points beyond our live-imaging sessions. Possible interference of dendrite development might arise from worm manipulation, immobilization or phototoxicity during live imaging. Low concentrations of levamisole and microbeads were used to minimize toxicity possibly caused by immobilization, and we tried to reduce phototoxicity by using spinning-disk confocal microscopy, which markedly shortened image acquisition time. Taken together, these data suggest that fmi-1 regulates self-avoidance by promoting efficient retraction after transient contact of the 3° dendrites.
Consistent with previous reports (Huarcaya Najarro and Ackley, 2013; Steimel et al., 2010), we found that fmi-1 was expressed throughout the C. elegans nervous system at all stages of development, as revealed by a high-dose translational FMI-1::GFP reporter that contained the 2615 bp fmi-1 promoter and the genomic fmi-1 sequence injected at a DNA concentration of 30 ng/µl (twnEx458; Fig. 2A). FMI-1::GFP was enriched in the nerve ring and the ventral nerve cord (Fig. 2A), highlighting the role of FMI-1 in axon and synapse development. These findings do not exclude the possibility that fmi-1 expression levels in non-neural tissues were lower than the detection threshold of our transgene. FMI-1::GFP showed membrane enrichment in the cell body of PVD (Fig. 2B). Punctate FMI-1::GFP could also be found in the 1° dendrites and occasionally in the 2° or 3° dendrites (Fig. 2C). The low signal-to-noise ratio of FMI-1::GFP in peripheral PVD branches precluded verification of possible FMI-1 localization in the tips of growing 3° dendrites. Even at such marginal signal intensity, this transgene caused self-avoidance defects in an otherwise wild-type genetic background [control, median=2.9% (2.3-7.9%), n=11; twnEx458, median=8.2% (0-21.7%), n=12, P<0.05, Mann–Whitney U test]. Another Pfmi-1::FMI-1::GFP transgene that was expressed at a lower dosage (10 ng/µl of DNA injected; twnEx423) fully rescued self-avoidance defects in the fmi-1 mutant (Fig. 2D), and it did not cause self-avoidance defects in the control background. Because the 3° dendrites of PVD are not associated with other neuronal processes (Dong et al., 2013), and fmi-1 is specifically expressed in neurons, including PVD, we speculate that FMI-1 acts autonomously to promote PVD dendrite self-avoidance. As a definitive test, we expressed the fmi-1a transcript in PVD, and confirmed that the dendrite self-avoidance defects of the fmi-1 mutant were completely rescued (Fig. 2E). These data suggest that fmi-1 acts cell-autonomously in PVD to regulate dendrite self-avoidance.
FMI-1 is required for local F-actin assembly during the establishment of dendrite self-avoidance
Recent reports show that contact and retraction of PVD 3° dendrites during self-avoidance is associated with local F-actin assembly at dendrite tips (Liao et al., 2018; Sundararajan et al., 2019). This spatially and temporally defined F-actin activity is significantly diminished in mutants that display self-avoidance defects, such as mig-14/Wntless and wsp-1/N-WASP (Liao et al., 2018). Cell-adhesion molecules and surface receptors engage the actin cytoskeleton to regulate axon branching and synapse formation (Chen et al., 2014; Chia et al., 2014). Therefore, we tested whether FMI-1 promotes dendrite self-avoidance through F-actin assembly. To monitor F-actin activity in live animals, we expressed EGFP-tagged LifeAct, a small peptide that binds F-actin and reports F-actin dynamics (Riedl et al., 2008) in the PVD neuron, and performed time-lapse spinning-disk confocal microscopy. We identified dendrite contacts by quantifying mCherry fluorescent signals that labeled PVD dendrites, as reported previously (Fig. 3A, Fig. S2; see also Materials and Methods) (Liao et al., 2018). Briefly, a 5 μm region centering at the dendrite contact point was defined as the contact site, and another 5 μm region along the same 3° dendrite but away from the contact point was selected as the control non-contact site (Fig. 3B). Using this method, we quantified LifeAct::EGFP fluorescent signal intensity and confirmed that F-actin activity increased at the contact sites that lasted for the duration of dendrite contact and diminished before dendrite retraction (Fig. 3C,D, Movie 3). No such F-actin activity was observed at the non-contact sites (Fig. 3C,D), suggesting that dendrite self-avoidance is associated with temporally and spatially defined F-actin dynamics. In the fmi-1 mutant, F-actin assembly at the contact sites during self-avoidance events was markedly decreased, whereas F-actin dynamics at non-contact sites were comparable with those in the control (Fig. 3A,C,D, Movie 4). These observations indicate that FMI-1 regulates dendrite self-avoidance by promoting F-actin assembly at dendrite tips that coincides with the self-avoidance event.
FMI-1 promotes F-actin dynamics through WSP-1/N-WASP
To explore how FMI-1 regulates F-actin dynamics, we next tested wsp-1, which encodes the C. elegans homolog of N-WASP known as a crucial regulator of F-actin assembly (Derivery and Gautreau, 2010). We previously showed that wsp-1 promotes PVD dendrite self-avoidance probably through facilitating local F-actin assembly at dendrite tips (Liao et al., 2018). gm324 is a deletion allele that produces no detectable wsp-1 transcripts or WSP-1 protein, and is thus a putative null wsp-1 allele (Withee et al., 2004). Defects of self-avoidance were comparable in the fmi-1 and wsp-1 single mutants and were not further increased in the wsp-1; fmi-1 double mutant, suggesting that these two genes act in a common pathway (Fig. 4A). PVD-specific expression of wsp-1 rescued self-avoidance defects in the wsp-1 and fmi-1 mutants (Fig. 4A), and it did not cause self-avoidance defects in the wild-type background [control, median=2.9% (2.3-7.9%), n=11; Pser-2.3::mCherry::WSP-1, median=5.15% (0-9.7%), n=12, P=0.51, Mann–Whitney U test]. By contrast, fmi-1 overexpression from the fmi-1 promoter failed to rescue the wsp-1 mutant. These results indicate that wsp-1 acts cell-autonomously in PVD and probably downstream of fmi-1. We next tested how FMI-1 interacts with WSP-1. Diffuse cytosolic signal of an mCherry::WSP-1 transgene expressed in PVD made it difficult to conclude whether mCherry::WSP-1 colocalizes with FMI-1::GFP. We were unable to detect WSP-1 protein distribution in PVD, possibly due to a low expression level, in a strain in which the endogenous wsp-1 gene was tagged with GFP using the CRISPR/Cas9 technique (Zhu et al., 2016).
Our model suggests that the inhibition of F-actin disassembly might offset the deleterious effects of fmi-1 or wsp-1 mutations on F-actin dynamics. The C. elegans gene unc-60 encodes cofilin, a protein that promotes F-actin disassembly (Ono and Benian, 1998). An unc-60 mutation suppressed self-avoidance defects in the fmi-1 mutant but did not cause dendrite defects in an otherwise control background (Fig. 4B). Taken together, our data indicate that FMI-1 leverages F-actin dynamics to promote dendrite self-avoidance.
To further understand whether fmi-1 regulates dendrite self-avoidance by interacting with known signaling pathways, we made double mutants that contained fmi-1(rh308) and either the unc-40 or mig-14 mutations. We showed in previous work that mig-14 and unc-40 act independently to regulate dendrite self-avoidance in PVD (Liao et al., 2018). We found that self-avoidance defects were more severe in the unc-40; fmi-1 and the mig-14; fmi-1 double mutants, compared with those of the unc-40, mig-14 and fmi-1 single mutants (Fig. S3A,B). These observations suggest that fmi-1 acts in a genetic pathway distinct from that of mig-14 and unc-40 to regulate dendrite self-avoidance.
FMI-1 antagonizes the polarity protein VANG-1/Vangl2 to regulate dendrite self-avoidance
Flamingo is a component of the PCP molecular cascade, which specifies planar tissue polarity in Drosophila and mammals (Devenport, 2014). VANG-1, the C. elegans homolog of the PCP component Vangl2/Strabismus/Van Gogh, regulates several aspects of neural development, including neuronal migration and neurite branching (Chen et al., 2017; He et al., 2018; Mentink et al., 2014; Sanchez-Alvarez et al., 2011). We recently showed that VANG-1 acts in the Wnt-Frizzled pathway to specify neurite branching sites in the PLM mechanosensory neuron by restricting F-actin dynamics (Chen et al., 2017). In the vang-1 mutant, F-actin assembly increased and was distributed to ectopic sites. This observation raises the intriguing possibility that vang-1 mutations might suppress dendrite self-avoidance defects of the fmi-1 mutant by upregulating F-actin assembly.
To test this hypothesis, we examined two vang-1 mutants (Fig. 5A,B). In addition to the commonly used tm1422 deletion allele, we generated a new vang-1 allele, twn2, using CRISPR/Cas9 gene editing. twn2 contains a premature stop codon predicted to truncate the majority of the VANG-1 protein and thus represents a putative null mutation (Fig. 5A). Dendrite self-avoidance in the vang-1(tm1422) and vang-1(twn2) mutants was indistinguishable from the control (Fig. 5B). Interestingly, both vang-1 mutations significantly suppressed self-avoidance defects of the fmi-1 mutant (Fig. 5B). Expression of vang-1 specifically in PVD of the fmi-1; vang-1 double mutant restored self-avoidance defects to the level observed in the fmi-1 mutant, indicating that vang-1 acts cell-autonomously (Fig. 5C). Of note, PVD-specific vang-1 expression did not cause self-avoidance defects in the wild-type background [control, median=2.9% (2.3-7.9%), n=11; Pser-2.3::mCherry::VANG-1, median=2.65% (0-7.3%), n=14, P=0.23, Mann–Whitney U test]. We failed to detect vang-1 expression in PVD using the transgene syIs202, which contains 3 kb of the vang-1 promoter and part of the vang-1-coding sequence to drive YFP expression (Green et al., 2008). However, PVD-specific transcriptional profiling by mRNA tagging and sequencing had detected vang-1 expression in PVD (Smith et al., 2013).
We hypothesize that fmi-1 antagonizes vang-1 to ensure efficient dendrite self-avoidance. To test this model, we first investigated the subcellular localization of VANG-1 by expressing VANG-1 tagged with GFP or mCherry at the N-terminus in PVD. Punctate GFP::VANG-1 or mCherry::VANG-1 signals were found in PVD soma and infrequently in dendrites (Fig. 5D-F). We found that some GFP::VANG-1 puncta in peripheral PVD dendrites showed colocalization with mCherry-fused COR-1/coronin, an F-actin-binding protein (Fig. 5E), although we were not able to identify VANG-1 signals at the tips of growing 3° dendrites with confidence. In the fmi-1 mutant, signal intensity of mCherry::VANG-1 in the PVD soma was significantly upregulated (Fig. 5F,G). The frequency of mCherry::VANG-1 puncta in peripheral PVD dendrites was also dramatically increased in the fmi-1 mutant compared with the control background (Fig. 5H,I). Overexpression of FMI-1 did not further decrease mCherry::VANG-1 level or frequency, suggesting that the VANG-1 level in PVD was minimal in the control (Fig. 5F-I). Using a transgenic C. elegans strain that expresses FMI-1::GFP and HA::VANG-1 broadly in the nervous system (from the fmi-1 and rgef-1 promoter, respectively), we showed that FMI-1 co-immunoprecipitated VANG-1 (Fig. 5J, Fig. S4). Taking these data together, we conclude that FMI-1 forms protein complexes with VANG-1 in C. elegans neurons, and it decreases VANG-1 levels in PVD.
FMI-1 regulates PVD axon extension both autonomously and non-autonomously
In addition to dendrite development, FMI-1 is important for the extension and guidance of certain axons in the C. elegans ventral nerve cord (VNC) that require pioneer axons (Steimel et al., 2010). The AVG axon is the first nerve process that establishes the right bundle of the embryonic VNC (Durbin, 1987). The PVP axons subsequently join the embryonic VNC and serve as pioneers or scaffolds for the following PVQ axons, which enter the embryonic VNC later (Durbin, 1987; Wadsworth and Hedgecock, 1996; White et al., 1986). Previous studies show that the navigation of follower PVQ axons requires both cell-autonomous and non-autonomous (from PVP pioneers) fmi-1 activities (Steimel et al., 2010). Although axons of the HSN motor neurons, which join the VNC at L3, show navigation and guidance errors in the fmi-1 mutant, the site of action for fmi-1 in postembryonic VNC axon navigation remains incompletely understood. The axon of PVD extends and enters the VNC at late L2 to L3 stages, making it a good model for studying the autonomous requirement of fmi-1 in postembryonic axon development.
The axons of left and right PVDs fasciculate after entering the right bundle of the VNC (Fig. 6A,B). In the control, PVD axons navigated anteriorly to terminate at or beyond the vulval region (Fig. 6A,B). In the fmi-1 mutant, PVD axons displayed either premature truncation or posterior misrouting, or both, in more than 90% of the animals (Fig. 6A,B, Table 1). Genomic fmi-1 sequences expressed from the fmi-1 promoter significantly rescued PVD axon defects when expression in PVD could be confirmed with a cell-specific marker, but not in mosaic transgenic animals that lost fmi-1 expression from PVD (Fig. 6C, Table 1). These observations indicate that fmi-1 activity in PVD is necessary to facilitate PVD axon navigation in the VNC, but do not rule out the possibility that non-autonomous fmi-1 activity is also required. In contrast to its rescue of self-avoidance defects, PVD-specific expression of fmi-1a failed to rescue axon phenotypes. As the PVD axons fasciculate with other VNC axons, this result suggests that fmi-1 needs to be present in both PVD and its fasciculating partners for anterior projection, although it might also be possible that other fmi-1 isoforms are necessary for complete functional complement of fmi-1 loss.
We noted that some fmi-1 animals still displayed PVD axon defects even with the Pfmi-1::FMI-1::GFP transgene expressed in PVD (Fig. 6C, Table 1). This raises the possibility that fmi-1 expression outside PVD is also important for PVD axon navigation. PVD axons form chemical synapses with several command interneurons that express the GLR-1 glutamate receptor, including AVA, AVB and AVD (Goodman, 2006), which also express fmi-1. fmi-1 might function in one or several of these neurons and signal in a non-autonomous fashion to regulate PVD axon guidance. To test this, we achieved cell-specific fmi-1 RNAi by expressing a fmi-1 RNA duplex in glr-1(+) neurons (Fig. S5A,B). Pglr-1::fmi-1(RNAi) significantly reduced the intensity of FMI-1::GFP in the nerve ring, which contains axons of AVA, AVB and AVD, suggesting that the Pglr-1::fmi-1(RNAi) transgene effectively decreases fmi-1 expression in these neurons (Fig. S5B,C). Dimmer signal intensity and susceptibility to photobleaching precluded the use of FMI-1::GFP signals in the VNC for evaluating the efficiency of fmi-1 RNAi. Silencing fmi-1 in glr-1(+) interneurons triggered PVD axon navigation defects in ∼20% of the wild-type animals (Fig. 6D, Table 1), implying that, in addition to its major, cell-autonomous function, fmi-1 can also regulate PVD axon development in a non-autonomous manner.
In this study, we present evidence indicating that the C. elegans Flamingo FMI-1 promotes dendrite self-avoidance by orchestrating spatially and temporally defined F-actin assembly at the dendrite tips. This intricate regulation of F-actin dynamics by FMI-1 is probably mediated through two molecules (Fig. 7A). First, FMI-1 facilitates F-actin assembly by genetically interacting with N-WASP, which is well-known for promoting F-actin polymerization through Arp2/3 (Derivery and Gautreau, 2010). Second, FMI-1 antagonizes VANG-1/Van Gogh, which has recently been found to restrict F-actin assembly in C. elegans neurites (Chen et al., 2017). These two molecules might act in similar or distinct genetic pathways to ensure precise F-actin assembly at the tips of contacting sister dendrites for efficient retraction. Together with recent studies on the effects of MIG-14/Wntless and Netrin signaling in PVD dendrite morphogenesis, our work reinforces the notion that F-actin is a central cytoskeletal component for dendrite self-avoidance in C. elegans. Validation in other systems, such as insects and mammals, is important to establish this as a conserved mechanism that sculpts non-overlapping dendrite morphology.
FMI-1 is required for contact-dependent dendrite retraction
Cadherins play important roles in different aspects of neural development, such as neurite extension, axon fasciculation and synapse formation (Takeichi, 2007). The role of Flamingo, an atypical cadherin, in dendrite self-avoidance was first discovered in Drosophila (Matsubara et al., 2011). Given its multiple extracellular domains characteristic of an adhesion molecule, such as cadherin repeats and laminin G domains, it was somewhat unexpected that the absence of Flamingo activity resulted in ectopic dendrite contact. Matsubara et al. (2011) showed that Flamingo genetically interacts with the LIM-domain protein Espinas through the juxtamembrane domain A (JM-A). LIM domain proteins, such as C. elegans UNC-115, regulate cytoskeletal elements, including actins (Gitai et al., 2003; Smith et al., 2014). We performed amino acid sequence analysis of FMI-1 and Drosophila Flamingo, but we did not find FMI-1 sequences homologous to the Drosophila Flamingo JM-A. Flamingo also interacts with RhoA, a well-established F-actin regulator. These observations are consistent with our findings that F-actin assembly mediates the effects of Flamingo on dendrite self-avoidance. FMI-1 lacks the conserved WIRS (WAVE regulatory complex interacting receptor sequence), a hexapeptide motif in the cytosolic tail of cell-membrane receptors that directly bind WRC (Chen et al., 2014). We therefore suspect that the interaction between FMI-1 and the actin regulatory machinery is indirect. The identification of FMI-1-interacting proteins will shed light on the molecular link between Flamingo, N-WASP, WRC, Arp2/3 and other actin regulators.
Antagonism between Flamingo and Van Gogh in dendrite self-avoidance
The core PCP component Van Gogh acts with Flamingo in a number of neurodevelopmental processes (Tissir and Goffinet, 2013). The antagonism between VANG-1/Van Gogh and FMI-1 during dendrite self-avoidance was thus unexpected and highlighted the complex interaction between different PCP molecules. FMI-1 forms protein complexes with VANG-1 and lowers the level of VANG-1 in neurons, which raises the possibility that the retention of VANG-1 by FMI-1 in the protein complex facilitates VANG-1 turnover. Our previous work on neurite branching revealed that VANG-1 restricts F-actin assembly in the C. elegans PLM mechanosensory neuron (Chen et al., 2017). Consistent with this, we find that vang-1 mutations suppressed self-avoidance defects of the fmi-1 mutant, probably through restoring F-actin formation in the dendrites. It remains unclear how VANG-1 inhibits F-actin assembly, although some VANG-1 proteins appear to colocalize with F-actin in PVD neurons. A recent study in the rat testis found that Vangl2, the mammalian Van Gogh, bound actin and modulated F-actin configuration (Chen et al., 2016). Loss or overexpression of Vangl2 changed F-actin organization and subcellular distribution in Sertoli cells, with knockdown of Vangl2 enhancing F-actin and N-cadherin levels. The elucidation of the protein structure of Van Gogh, which remains largely unexplored, is necessary to understand how Van Gogh modulates F-actin organization, as well as how Flamingo antagonizes Van Gogh.
F-actin assembly and dendrite retraction
Previous studies using LifeAct::EGFP as a reporter for F-actin dynamics suggest that F-actin assembly at dendrite tips increases around the time of contact and retraction of sister dendrites in PVD (Liao et al., 2018; Sundararajan et al., 2019). The spatial resolution of time-lapse confocal microscopy does not allow us to distinguish whether F-actin assembly occurs after physical contact of dendritic membrane, or whether it occurs when sister dendrites are close but still separate. Dendrite extension occurs normally in the fmi-1 and wsp-1 mutants but their retraction is impaired, together with diminished F-actin assembly. These observations imply that F-actin assembly is required for dendrite repulsion, a notion further supported by the observation that unc-60/cofilin mutations suppress dendrite self-avoidance defects. How does F-actin assembly drive dendrite retraction? Semaphorin 3A has been shown to collapse the axon growth cone and induce axon retraction in cultured chicken dorsal root ganglion neurons in an F-actin-dependent manner (Gallo, 2006). The addition of semaphorin 3A diminishes F-actin in the growth cone but increases F-actin assembly in an axonal segment behind the growth cone. Inhibition of RhoA or Rho-dependent kinase (ROCK) decreases intra-axonal F-actin and significantly suppresses semaphorin 3A-induced axon retraction. It is thus hypothesized that intra-axonal F-actin drives axon retraction in the presence of repulsive signals, probably through engaging myosin II (Gallo, 2006). Interestingly, a recent study found that mutations in the C. elegans gene nmy-1, which encodes non-muscle myosin II, resulted in PVD dendrite self-avoidance defects (Sundararajan et al., 2019). One possibility is that the activation of membrane receptors, such as Flamingo, Wntless or DCC, leads to a redistribution of F-actin assembly towards dendrite tips. Retrograde actin flow driven by myosin II, along with enhanced membrane retrieval, leads to axon retraction (Gallo, 2006; Yang et al., 2012). The termination of receptor activation following dendrite separation from the initial contact decreases F-actin assembly and stops dendrite repulsion, which might explain the transient nature of dendrite retraction during self-avoidance. F-actin polymerization may also facilitate membrane retrieval at distal dendrites, which contributes to dendrite retraction. We speculate that F-actin dynamics orchestrate concerted cytoskeletal shortening and retrieval of the dendrite membrane, resolving sister dendrites in transient contact during self-avoidance. However, spatial and temporal resolution of fluorescent imaging in this study was not sufficient for documenting fine F-actin features in the dynamic dendritic branches during self-avoidance. Super-resolution microscopy that offers temporal resolution at the time scale of a hundred milliseconds, and an F-actin reporter with improved signal-to-noise ratio, should help to clarify the correlation between F-actin dynamics and the distinct steps in dendrite self-avoidance.
Autonomous and non-autonomous FMI-1 functions in axon guidance
In addition to dendrite self-avoidance, we found that FMI-1 is also required for navigation of the PVD axon in the VNC. The modest self-avoidance defects in the fmi-1 mutant suggest that other pathways act in parallel to ensure the robustness of dendrite repulsion. By contrast, most fmi-1 mutant animals show PVD axon guidance defects, implicating FMI-1 as an essential factor in PVD axon development. FMI-1 is required for the development of multiple C. elegans neuronal classes whose axons travel in the VNC, including PVP, PVQ, HSN and glr-1(+) command interneurons (Steimel et al., 2010). The axon phenotypes are complex in the fmi-1 mutant and seem to depend on the neuronal type. Axons of the left and right PVP travel in the right and left fascicles of the VNC, respectively. Axons of the left and right PVQ use the PVP axons as a guidepost and follow them in the VNC. In the absence of fmi-1, the pioneer PVPR axon often defasciculates from the left VNC fascicle and crosses to the contralateral right VNC fascicle. As a result, the follower PVQ axons display penetrant crossover defects and join inappropriate VNC fascicles in the fmi-1 mutant. fmi-1 expression in the pioneer PVP axons rescues crossover defects in both PVP and the follower PVQ, whereas fmi-1 expression in PVQ rescues defects of the PVQ but not those of the pioneer PVP axons. These observations suggest that fmi-1 controls PVQ axon development both cell-autonomously and non-autonomously. By contrast, premature truncation of PVQ axons in the fmi-1 mutant could only be rescued when fmi-1 is expressed in PVQ, suggesting a cell-autonomous requirement of fmi-1 in axon extension. Analysis of fmi-1 mutant animals mosaic for PVD-specific fmi-1 rescue, using the Pfmi-1::FMI-1::GFP transgene, suggests that fmi-1 activity in PVD is essential for axon development. However, expression of fmi-1a in PVD fails to rescue axon defects in the fmi-1 mutant. Moreover, reducing fmi-1 activity in glr-1(+) command interneurons, with which PVD axons make synapses, results in defective PVD axon navigation. These data are also consistent with the model that FMI-1 acts both autonomously and non-autonomously to control PVD axon development. It is also possible that more than one fmi-1 isoform is necessary to fully rescue the axon defects. Crossover defects in axons of command interneurons are found in 10-20% of the fmi-1 mutant animals (Steimel et al., 2010), implying that PVD axon defects in some fmi-1 animals are probably the consequence of disruption of the development of interneuron axons. This also raises the possibility of heterophilic FMI-1 interaction with other membrane-tethered or diffusible ligands. Candidates include molecules with cadherin repeats, EGF or laminin domains. The identification of FMI-1 ligands will advance our understanding of how Flamingo signaling controls axon extension and navigation to sculpt the connectivity of neuronal circuitries.
MATERIALS AND METHODS
C. elegans strains
All strains were cultured and maintained as described previously (Brenner, 1974). The alleles and integrated transgenes used in this study were: LG I, unc-40(n324); LG II, mig-14(ga62); LG III, unc-119(ed3); LG IV, wsp-1(gm324), cas723[gfp::wsp-1a knock-in]; LG V, fmi-1(rh308), fmi-1(hd121), nDf42, unc-60(su158), syIs202(Pvang-1::YFP, Pmyo-2::DsRed); LGX, vang-1(tm1422), vang-1(twn2), wdIs52[F49H12.4::GFP, unc-119(+)]. Extrachromosomal arrays used were: twnEx382[Pser-2.3::LifeAct::EGFP, Pser-2.3::mCherry, Pgcy-8::GFP]; twnEx422[Pfmi-1::FMI-1::GFP (30 ng/μl), Pser-2.3::mCherry, Pgcy-8::mCherry]; twnEx423[Pfmi-1::FMI-1::GFP (10 ng/μl), Pser-2.3::mCherry, Pgcy-8::mCherry]; twnEx444[Pegl-17::GFP::VANG-1, Pser-2.3::mCherry, Pgcy-8::mCherry]; twnEx445[Pser-2.3::mCherry::WSP-1, Pttx-3::GFP]: twnEx446[Pglr-1::fmi-1(RNAi: sense/anti-sense), Pglr-1::mCherry, Pgcy-8::mCherry]; twnEx450[Pser-2.3::mCherry::VANG-1, Pgcy-8::mCherry]; twnEx451[Pfmi-1::FMI-1::GFP, Pser-2.3::mCherry::VANG-1, Pgcy-8::mCherry]; twnEx452[Pser-2.3::GFP::VANG-1, Pser-2.3::COR-1::mCherry, Pgcy-8::mCherry]; twnEx458[Pfmi-1::FMI-1::GFP (30 ng/μl), Pgcy-8::mCherry] (‘high-dose FMI-1::GFP’); twnEx483[Pglr-1::mCherry]; twnEx488[Prgef-1::HA::VANG-1, Pfmi-1::FMI-1::GFP(30 ng/μl), unc-119(+)]; twnEx571[Pser-2.3::FMI-1a::gfp, Pgcy-8::mCherry]; and twnEx572[Prgef-1::HA::VANG-1].
Plasmid construction and germline transformation
Constructs used in this study were generated by standard molecular biological techniques. Constructs used to generate the twnEx series of transgenes were in the pPD95.77 vector, except for Pfmi-1::FMI-1::GFP, which was kindly provided by Harald Hutter (Simon Fraser University, Burnaby, Canada). The fmi-1a cDNA was a generous gift from Georgia Rapti and Shai Shaham (The Rockefeller University, New York, NY, USA). Germline transformation by microinjection was performed as described previously (Mello et al., 1991).
CRISPR/Cas9 genome editing
CRISPR/Cas9 mutagenesis was performed by germline transformation using pDD162(Peft-3::Cas9, Addgene #47549) inserted with vang-1 sgRNA (GACACGAGGAGTTGCGTT). An unc-22 sgRNA construct was co-injected as a co-CRISPR marker to select F1 animals whose genomes were successfully edited at the unc-22 locus to indicate possible editing at the vang-1 locus. Confirmation of editing of vang-1 was also sought by performing T7-endonuclease I digestion of the PCR products of target vang-1 sequence from F1 (Mashal et al., 1995; Shen et al., 2014), and was verified by DNA sequencing.
Scoring of PVD axon projection defects
PVD axon defects were scored with the transgene wdIs52, which expresses soluble GFP from the regulatory sequence of the C. elegans gene F49H12.4 in PVD, AQR and a cell in the tail (Smith et al., 2010). For characterizing PVD axon projection, L4 animals were collected and immobilized by 1% sodium azide and were manually oriented with the ventral side up. PVD axons were imaged using the 40× objective of the Zeiss AxioImager M2 system. Axons that projected anteriorly beyond the vulva were defined as ‘normal’. Axons that projected both anteriorly and posteriorly were defined as ‘axon bifurcation’. Axons that projected anteriorly but failed to reach the midpoint between the vulva and the PVD soma were classified as ‘premature truncation’. Axons that projected posteriorly were classified as ‘posterior misrouting’. Unbundling of the left and right PVD axons was defined as ‘defasciculation’.
Quantification of dendrite self-avoidance defects
For time-lapse imaging of dendrite self-avoidance, laid embryos were cultivated at 20°C for ∼30 h to obtain L3 larvae. Animals were immobilized in 2 μl of 1 mM levamisole with 2 μl of polystyrene beads (0.1 μm, Polysciences) on 10% agar pads. Images were taken using the 40× objective of the Carl Zeiss Cell Observer SD equipped with a Yokogawa CSU-X1 spinning disk and EMCCD Qimaging Rolera EM-C2 at 2 frames/min for at least 30 min per animal. For F-actin imaging during dendrite contact, L3 larvae expressing LifeAct::EGFP were collected and immobilized in 1 mM levamisole and polystyrene beads (as described earlier) on 10% agar pads. The imaging apparatus and condition were the same as those for time-lapse imaging of dendrite self-avoidance.
Fluorescence confocal microscopy
For analyzing dendrite self-avoidance defects and protein localization, L4 hermaphrodites were anesthetized in 1% sodium azide and mounted on 5% agar pads. To assess fmi-1 RNAi knockdown efficiency, FMI::GFP intensity in the nerve ring of L2 hermaphrodites was quantified. C. elegans protein expression pattern and neuronal morphology were imaged using a Zeiss LSM700 or LSM880 Airyscan Imaging system.
Quantification of fluorescent signal intensity
To evaluate fmi-1 RNAi knockdown efficiency, z-stack projection images of the nerve ring were acquired using a Zeiss LSM700 Confocal Imaging System and FMI-1::GFP signals were quantified using ImageJ. For analyzing LifeAct::EGFP signals in PVD dendrites, two 5 μm regions (contact site and non-contact site) were selected and quantified using ZEN software (2011 blue edition). For analyzing mCherry::VANG-1 signals in PVD soma, single optical sections of a z-stack series, obtained using a Zeiss LSM700 Confocal Imaging System, were quantified using ImageJ and then summed to derive mean signal intensity, which is total mCherry pixel intensity divided by the total area of the PVD soma. To quantify mCherry::VANG-1 puncta in PVD dendrites, z-stack projection images were obtained using a Zeiss LSM700 Confocal Imaging System. Each single mCherry::VANG-1 punctum in PVD dendrites was first quantified using Zen software (2010 black edition). We defined discrete mCherry::VANG-1 puncta with an area larger than 1 μm2 and an intensity higher than 10 a.u. as valid mCherry::VANG-1 puncta.
Worm immunoprecipitation and western blotting
The constructs Pfmi-1::FMI-1::GFP and Prgef-1::HA::VANG-1, together with a rescue construct PCG150 (Punc-119::UNC-119 rescue fragment), were co-injected into unc-119(ed3). The resulting transgenic strain expressed both FMI-1::GFP and VANG-1::HA expression throughout the nervous system. The Bristol N2 strain and strains with only twnEx458[Pfmi-1::FMI-1::GFP (30 ng/μl), Pgcy-8::mCherry] or twnEx572[Prgef-1::HA::VANG-1] served as controls. Animals were cultivated on 10 cm nematode growth media plates spread with OP50 E. coli at 20°C for 4 to 5 days until 80% confluence was reached. Animals from 20 such plates were collected by washing with M9 buffer for five times, followed by a wash with 1× PBS and lysis buffer [25 mM Tris-HCl (pH 7.5), 100 mM NaCl, 1 mM EDTA], and buffer for worm lysis containing 25 mM Tris-HCl (pH 7.5), 100 mM NaCl, 1 mM EDTA, 1 mM Na3VO4, 10 mM NaF and protease inhibitor cocktail (Roche, 04693132001). A worm pellet of ∼0.25 ml was then transferred to a screw cap tube (DOT Scientific) with 0.5 ml of 1.0 mm zirconia beads (BioSpec Products) and 1 ml of lysis buffer/protease inhibitor cocktails. Tubes were then placed on a FastPrep-24 5G Benchtop Homogenizer (MP Biomedicals) and homogenized with three 30-s pulses at maximal speed. Proteins were collected by spinning the lysates at 15,500 g for 30 min to remove the insoluble material. A total of 10 mg protein with 30 μl of GFP nanobody (GFP-Trap, Chromotek) or 30 μl of monoclonal anti-HA-agarose (Sigma-Aldrich, A2095) were used to perform the immunoprecipitation experiments. We used the following primary antibodies: rabbit polyclonal anti-HA (1:1000, Abcam, ab71113), rabbit polyclonal anti-GFP (1:1000, Santa Cruz Biotechnology, sc-8334) and anti-beta actin (1:2000, Santa Cruz Biotechnology, sc-47778). The secondary antibodies used in this study were peroxidase-conjugated AffiniPure goat anti-rabbit IgG (1:5000, Jackson ImmunoResearch, 111-035-003) and HRP goat anti-mouse IgG (1:5000, BioLegend, 405306). See supplementary Materials and Methods for further details regarding antibody validation.
We thank Gian Garriga, Harald Hutter, Georgia Rapti, Shai Shaham and the C. elegans Genetics Center (CGC) for the provision of worm strains and plasmids. We thank Chun-Hao Chen for the vang-1(twn2) allele; Chun-Wei He and Ya-Wen Liu for advice on biochemistry; and Hwa-Man Hsu for assistance with confocal imaging experiments. We thank Chun-Wei He for comments on the manuscript. CGC is funded by the National Institutes of Health Office of Research Infrastructure Programs (P40 OD010440).
Conceptualization: H.-W.H., C.-P.L., Y.-C.C., C.-L.P.; Methodology: H.-W.H., C.-P.L., Y.-C.C., R.-T.S., C.-L.P.; Validation: H.-W.H., C.-L.P.; Formal analysis: H.-W.H., C.-P.L., Y.-C.C., R.-T.S., C.-L.P.; Investigation: H.-W.H., C.-P.L., Y.-C.C., R.-T.S., C.-L.P.; Data curation: C.-L.P.; Writing - original draft: H.-W.H., C.-L.P.; Writing - review & editing: C.-L.P.; Supervision: C.-L.P.; Project administration: C.-L.P.; Funding acquisition: C.-L.P.
This study was supported by the Center of Precision Medicine from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project administered by the Ministry of Education (NTU-109L901402A to C.-L.P.) and the Ministry of Science and Technology, Taiwan (MOST 106-2320-B-002-051-MY3, MOST 108-3017-F-002-004 and MOST 109-2634-F-002-043 to C.-L.P.).
The authors declare no competing or financial interests.