The cell cortex-localized protein CHDP-1 is required for dendritic development and transport in C. elegans neurons

Cortical actin, a thin layer of actin network underneath the plasma membranes, plays critical roles in numerous processes, such as cell morphogenesis and migration. Neurons often grow highly branched dendrite morphologies, which is crucial for neural circuit assembly. It is still poorly understood how cortical actin assembly is controlled in dendrites and whether it is critical for dendrite development, maintenance and function. In the present study, we find that knock-out of C. elegans chdp-1, which encodes a cell cortex-localized protein, causes dendrite formation defects in the larval stages and spontaneous dendrite degeneration in adults. Actin assembly in the dendritic growth cones is significantly reduced in the chdp-1 mutants. PVD neurons sense muscle contraction and act as proprioceptors. Loss of chdp-1 abolishes proprioception, which can be rescued by expressing CHDP-1 in the PVD neurons. In the high-ordered branches, loss of chdp-1 also severely affects the microtubule cytoskeleton assembly, intracellular organelle transport and neuropeptide secretion. Interestingly, knock-out of sax-1, which encodes an evolutionary conserved serine/threonine protein kinase, suppresses the defects mentioned above in chdp-1 mutants. Thus, our findings suggest that CHDP-1 and SAX-1 function in an opposing manner in the multi-dendritic neurons to modulate cortical actin assembly, which is critical for dendrite development, maintenance and function.

Introduction A neuron is the structural and functional unit of the nervous system in animals. During axon formation and dendrite branching, the plasma membrane of a neuron continuously changes its shape until the morphogenesis process is finished, which heavily relies on the membranecytoskeleton interactions [1]. The plasma membrane-associated skeleton, also known as the membrane skeleton, consists of actin, spectrin and associated molecules [2]. Analyses from multiple neuronal cells types from different species revealed that many axons and dendrites contain a specialized periodic actin-spectrin-based membrane skeleton (PMS), which can serve as signaling platforms for RTK transactivation and microtubule maintenance [3][4][5][6][7][8][9][10][11]. Mutations in spectrin are associated with numerous human diseases, such as hereditary elliptocytosis and spinocerebellar ataxia [12,13]. Loss-of-function mutations in unc-70/beta-spectrin in C. elegans result in defects in axonal maintenance, cilium biogenesis, neuron migration and dendrite morphogenesis [10,14]. The actin-binding protein, alpha-adducin, has multiple functions in regulating actin cytoskeleton formation. Knock-out of alpha-adducin causes progressive axon enlargement and degeneration [15]. However, compared to what is known for axonal cortical actin assembly, it is still poorly understood how dendritic cortical actin assembly is controlled and whether it is crucial for dendrite development, maintenance and functions.
We previously identified the calponin homology domain-containing protein (CHDP-1) as a critical regulator of cortical actin assembly in C. elegans. Loss of chdp-1 results in defective membrane protrusion formation in the neurite growth cones of BDU and PLM neurons. Using an overexpressed transgene, we showed that CHDP-1 is widely expressed and mainly labels the cell cortex. In BDU and PLM neurons, CHDP-1 promotes cortical actin assembly via recruiting and activating the small GTPase CED-10/Rac1 [16]. It is unclear whether CHDP-1 also regulates cortical actin assembly in multi-dendritic neurons, and if so, whether it is required for dendrite development, maintenance and function.
Here we address these questions using the C. elegans PVD neurons as a model. The two PVD neurons, PVDL and PVDR, locate on the left and right sides of the nematode C. elegans, respectively, and covers the majority of the surface of the body except for the head and neck regions [17]. These two neurons are born in the middle second larval stage (L2), and each grows an unbranched axon and two primary dendrites (1 o ) towards the anterior and posterior, respectively. At the late L2 and early third larval stage (L3), secondary dendrites (2 o ) are formed from the 1 o and grow along the dorsal-ventral axis. When the dendritic tips reach the borders of the outer body wall muscles, the dendrites turn and form T-shaped tertiary branches (3 o ). At the early L4 stage, quaternary dendrites (4 o ) are formed from the 3 o , and together, they form menorah structures in the wild-type animals [18]. This stereotypical morphogenesis is precisely guided by a multi-protein receptor-ligand complex, including two transmembrane proteins DMA-1/LRR-TM and HPO-30/Claudin on the dendritic membranes, two transmembrane proteins SAX-7/L1CAM and MNR-1/Menorin on the epidermal membranes, and one secreted protein LECT-2/LECT2 derived from the body wall muscle cells [19][20][21][22][23][24][25][26]. Notably, the high-ordered dendrites are sandwiched between the epidermis and body wall muscles, which is consistent with the role of the PVD neurons as the proprioceptors to sense the contraction of the muscle cells [17,27]. During dendrite development, DMA-1 and HPO-30 promote actin assembly via recruiting/activating TIAM-1 and WRC, respectively [24]. In the PVD dendrites, filamentous actin (F-actin) is enriched in the high-ordered, while the microtubule cytoskeleton is enriched in the 1 o dendrites [25,28]. The intracellular organelles, such as the endoplasmic reticulum, mitochondria and secretory/endocytic vesicles, are distributed not only in the primary dendrites but also in the high-ordered ones, which are likely regulated by motor proteins moving along the microtubule cytoskeletons [29,30,31]. In the anterior primary dendrite, a growth cone localized non-centrosomal microtubule organizing center generates plus-end-out microtubules in the growth cone and minus-end-out microtubules along outgrowing dendrites [32]. It is not clear how microtubule assembly is controlled in high-ordered dendritic branches.
In this work, we performed a forward genetic screen and identified a loss-of-function mutation in the chdp-1 gene as the PVD dendrite morphology was defective in the mutant animals. CHDP-1 modulates actin assembly in the dendritic growth cones. Intriguingly, loss of chdp-1 also perturbs the microtubule cytoskeleton assembly, and transport of intracellular organelles, such as dense-core vesicles, ER and mitochondria in the high-ordered branches. The proprioceptive function of PVD neurons is abolished by the loss of chdp-1. Knock-out of sax-1, which encodes an evolutionary conserved protein kinase [33,34], rescues the defects mentioned above in chdp-1 mutants, suggesting that SAX-1 is likely a negative regulator of cortical actin assembly. Together, our results suggest that CHDP-1 and SAX-1 regulate cortical actin assembly, which is critical for proper dendrite development, maintenance and function.

Loss of chdp-1 causes abnormal development of PVD dendrites in C. elegans
To identify additional regulators that control dendrite development, we used the multi-dendritic PVD neurons in C. elegans as a model and conducted a large-scale forward genetic screen. Among the isolated mutants, here we focused on the characterization of zac135, in which the dendrite arborization was significantly affected. Compared to the wild-type controls, the zac135 mutants showed significantly more 2 o and 3 o dendritic branches but less 4 o branches (Fig 1A and 1B). In addition, the intensity of the membrane-targeted green fluorescent protein (myristoylated-GFP, expressed by an integrated transgene ser2prom3::myr-gfp) in the 2 o dendritic branches was significantly dimmer in the zac135 mutants, possibly due to a decreased dendritic width or protein diffusion defect (Fig 1C). Using standard genetic mapping and cloning methods, we identified the causative mutation in zac135 as a single base change (ATG to ATA), which disrupted the start codon of CHDP-1 protein (M1I). We also analyzed the dendrite morphology of tm4947, a putative molecular null mutant of chdp-1 [16], and found that both mutants showed similar dendrite branching defects (Fig 1A-1C). Similar abnormal dendrite branching and intensity phenotypes were observed when we used two cytosolic GFP reporters (wdIs51 and otIs138, respectively) to visualize the PVD dendrites [18,35], suggesting that these phenotypes are not specific to the myr-GFP reporter (S1A-S1D Fig). We further analyzed the dendrite width using Stimulated Emission Depletion Microscopy (STED), which offers a higher resolution for imaging [36]. For the 1 o dendrites, the width of chdp-1 mutants was less uniform than that of the wild-type controls. The average diameters of 1 o dendrites are 0.27 μm and 0.34 μm for wild-type and chdp-1 mutants, respectively. Moreover, the average diameters of 2 o branches are 0.17 μm and 0.09 μm for wild-type and chdp-1 mutants, respectively (S1E-S1G Fig).
To understand why chdp-1 mutants grow more 2 o dendritic branches and less 4 o branches, we performed time-lapse recording. We found that chdp-1 homozygous knock-out mutants showing the PVD morphology of wild-type, chdp-1(zac135) and chdp-1(tm4947) mutant animals at the day 1 adult stage. PVD morphology was visualized using a cell-specific fluorescent marker (ser2prom3::myr-gfp). Scale bar: 20 μm.
(B-C) Quantification of (B) the number of 2 o , 3 o and 4 o branches, and (C) the ratio of the intensity of the 2 o branches to that of the primary dendrite in wild-type, chdp-1(zac135) and chdp-1(tm4947) in the 100 μm area anterior to PVD cell body. Error bars: SEM. ��� p < 0.001 by one-way ANOVA with the Tukey correction. n = 20-30 for each genotype.  showed more branch initiation and retraction during early larval stage 3 (L3) when 2 o dendritic branches were formed. The speed of 2 o dendrite outgrowth and retraction were not significantly different between the two genotypes (S2A- S2E Fig). During 4 o branch development, chdp-1 mutants showed less branch initiation and retraction during early larval stage 4 (L4) when 4 o dendritic branches were formed. The speed of 4 o dendrite outgrowth and retraction were not significantly different between the two genotypes (Fig 1D-1H). Together, our data demonstrate that CHDP-1 plays a vital role in dendrite branch formation.

Dendrite maintenance is defective in chdp-1 knock-out animals during the adult stages
To test whether chdp-1 is required for dendrite maintenance during the adult stages, we examined the dendrite morphologies for both the wild-type control and chdp-1 (tm4947) animals at the larval stage 4 (L4), 1 day post L4 stage (day 1), day 3, day 5, day 7 and day 9, respectively. Almost all the animals in the wild-type control groups showed intact anterior primary dendrites from L4 to day 9, while a significant portion of chdp-1 mutant animals displayed dendrite degeneration in the most anterior part from day 1 to day 9. We quantified the number of 2 o and 4 o branches in the anterior half of the PVD neurons, roughly between the OLL cell bodies and the vulva and found that the number of 2 o branches decreased at day 7 and day 9 when chdp-1 was mutated (S3A- S3C Fig). Thus, CHDP-1 is required for both dendrite development and maintenance.

CHDP-1 acts cell-autonomously in the PVD neurons during dendrite branching
Our previous study showed that the exogenously expressed GFP::CHDP-1 driven by the chdp-1 promoter mainly localizes to the cell cortex in many different cell types [16]. To determine the expression pattern and subcellular localization of endogenous CHDP-1, we inserted the coding sequence of gfp into the N-terminus of chdp-1 locus by CRISPR/Cas9-based genome editing [37]. We quantified the number of dendrite branches of PVD neurons and found that the gfp::chdp-1 knock-in strain and the wild-type control strain showed a similar number of 2 o , 3 o and 4 o dendritic branches, suggesting that the gfp insertion does not significantly affect the function of CHDP-1 in dendrite development (S4A and S4B Fig). We confirmed that the endogenously expressed CHDP-1 mainly localizes to the cell cortex and is expressed in many cell types, if not all, including the cell bodies of the PVD neurons. Due to the relatively limited resolution of the spinning-disk confocal imaging, we could not determine whether the endogenous CHDP-1 localizes onto the cell cortex in the PVD dendrites (Figs 2A, 2B, S5A and S5B). Next, to determine which tissue CHDP-1 acts in, we expressed CHDP-1 using cell-type specific promoters, including ser2prom3 for the PVD neurons, Pdpy-7 for the epidermis, and Phlh-1 for the body wall muscle cells. Pchdp-1 was used as a positive control. CHDP-1 expressed from the chdp-1 endogenous promoter or the PVD promoter fully rescued the dendrite branching defects and the faint staining of 2 o branches by the myr-GFP reporter in the chdp-1(tm4947) animals, while those driven by the epidermis or the body wall muscle promoter failed to do so (Fig 2C-2E). To understand when CHDP-1 functions, we generated a single copy transgene to express CHDP-1 under a heat-shock promoter (Phsp-16.48) in the chdp-1(tm4947) genetic background [38]. The increased number of 2 o and 3 o branches could only be rescued when the transgene expression was induced at the L2 stage, but not at earlier or later stages. The decreased number of 4 o branches could be rescued by inducing transgene expression at L2, L3 and L4 stages, but not earlier or later (Fig 2F and 2G). In addition, the faint staining of 2 o branches by the myr-GFP reporter in the chdp-1(tm4947) animals can be fully or partially rescued by inducing CHDP-1 expression at L2 and L3 stages, respectively ( Fig  2H). Together, these data suggest that CHDP-1 functions cell-autonomously and right at the dendrite branching stage.

Structure-function analysis of CHDP-1 in regulating dendrite morphogenesis
To understand how CHDP-1 regulates dendrite development, we sought to determine which domain(s) is essential by performing the structure-function analysis. The full-length CHDP-1 driven by the PVD promoter fully rescued the dendrite branching defects of chdp-1(tm4947) mutants and was used as a positive control. Truncating the P1 motif, P2 motif or the C-terminal part did not affect the rescue ability, suggesting these motifs are not critical for regulating dendrite branching. In contrast, truncating the calponin homology domain (CH) or the helix motif abolished the rescue activity (Fig 3A-3C). We also examined the expression and subcellular localization of the full-length and truncated CHDP-1 proteins tagged by an N-terminal GFP. GFP signals could be detected for all the transgenes. For the full-length, delta P1, delta P2, or delta C transgenes, GFP clearly labeled the cell margin in the cell bodies of the PVD neurons. However, possibly due to improper folding/trafficking, GFP::CHDP-1 delta CH and GFP::CHDP-1 delta helix failed to localize to the cell cortex. They displayed punctate signals and diffused cytosolic distribution, respectively (Fig 3D), consistent with the notion that the cell cortex localization of CHDP-1 is critical for its function in dendrite development.

Actin assembly in the growth cones of high-ordered branches was reduced by the loss of chdp-1
We previously found that CHDP-1 regulates cortical actin assembly during neurite growth cone formation in the BDU and PLM neurons. Thus, we sought to determine whether CHDP-1 plays a similar role during PVD dendrite development. We first examined the localization of the endogenously expressed CHDP-1 in the dendrite growth cones. To avoid the signals derived from other cells, we specifically labeled the endogenously expressed CHDP-1 protein in the PVD neurons using the native and tissue-specific fluorescence (NATF) approach ( Fig  4A) [39]. Briefly, we inserted the seven copies of sequences encoding the GFP11 into the N-terminus of the endogenous CHDP-1 using CRIPSR/Cas9-mediated genome editing and overexpressed GFP1-10 specifically in the PVD neurons using the ser2prom3 promoter from an extrachromosomal array. The animals expressing GFP(7x)::CHDP-1 showed normal dendrite morphology and intensity of the 2 o branches, suggesting that the function of the endogenous CHDP-1 is not perturbed (S4C- S4E Fig). Unlike the cytosolic GFP, which showed even distribution in the 2 o branches, GFP(7x)::CHDP-1 was enriched in the dendritic growth cones. This pattern was reminiscent of the F-actin probe mCherry::moesin actin-binding domain (moesi-nABD) (Fig 4B and 4D) [40]. Our previous study reported that the formation of the highordered branches relies on F-actin assembly [24]. In the wild-type animals, more than 80% of the moesinABD-labeled growth cones of the 2 o branches showed a palm-like shape. In contrast, nearly all the growth cones of the 2 o branches in the chdp-1(tm4947) animals showed a finger-like shape (Fig 4E-4G). We also compared the actin assembly in the dendritic growth cones of wild-type control and chdp-1(tm4947) animals using time-lapse recording and found that knock-out of chdp-1 dramatically decreased both the growth cone size and F-actin assembly in the growth cones of both 2 o and 4 o branches (Figs 4G, 4H, S6A, and S6B). The actin assembly defect of the chdp-1(tm4947) animals was confirmed using another F-actin reporter -Lifeact::GFP (S7A-S7D Fig) [41].

CHDP-1 is required for the proprioceptive function of the PVD neurons
PVD neurons sense the contraction of body wall muscle cells and thus are proprioceptors [17,27]. As loss of chdp-1 caused defects in dendrite branching and actin cytoskeleton organization, we sought to determine whether CHDP-1 is required for the proprioceptive function of the PVD neurons. We measured the moving tracks and body length of the following four strains: wild-type, dma-1(wy686), chdp-1(tm4947) and chdp-1(tm4947) carrying a PVD::chdp-1 transgene. dma-1 encodes a dendritic branching receptor, loss of which severely affects dendrite branching and the proprioceptive function of the PVD neurons ( Fig 5A) [26,27]. Thus, dma-1(wy686) was used as a positive control in these experiments. All the strains showed similar body lengths, making it possible to directly compare the amplitude and wavelength of the moving tracks ( Fig 5B). Interestingly, although chdp-1(tm4947) animals grew more dendrite branches than dma-1(wy686) animals, they displayed similar defects in proprioception as determined by the quantification of the amplitude and wavelength of the moving tracks. Expressing CHDP-1 specifically in the PVD neurons fully restored not only the dendrite branching but also the proprioceptive function of the PVD neurons ( Fig 5A and 5C-5E). These results reveal that although the high-ordered branches could be generated when chdp-1 is mutated, they are non-functional for proprioception.

Neuropeptide release and microtubule assembly are defective in the highordered branches of chdp-1 mutants
Recently Tao et al. reported that the proprioceptive PVD neurons secret neuropeptide NLP-12 from their 3 o branches to modulate neuromuscular junction activity and set muscle tone and movement vigor [27]. To understand how the loss of chdp-1 results in defects in proprioception, we asked whether the dendritic secretion of the neuropeptide NLP-12 is defective in chdp-1 (tm4947) animals. We compared the distribution of NLP-12::Venus in wild-type and chdp-1 (tm4947) animals. In wild-type animals, NLP-12 positive dense-core vesicles are distributed in the primary dendrites and the high-ordered branches. However, in chdp-1(tm4947) animals, these dense-core vesicles can only be found in the primary dendrites (Figs 6A, 6B, S8A, and S8B). Next, to directly compare the local secretion of NLP-12::Venus, we used the neuropeptide trapping assay developed by Tao et al. Briefly, the GFP nanobody, which is also known as the  GFP binding protein (GBP), was fused at the N-terminus of the single transmembrane domain protein SAX-7 and the fused protein was specifically expressed in the epidermal cells using the Pdpy-7 promoter. SAX-7 formed stripes to guide the formation of 3 o and 4 o branches of the PVD neurons. Thus, once NLP-12::Venus was secreted from these branches, the neuropeptide was captured by the locally expressed GBP::SAX-7 due to the high binding affinity between GBP and the Venus protein. We found that loss of chdp-1 completely abolished NLP-12 secretion from the 3 o dendrites (Fig 6C-6E) [27]. Together, these data suggest CHDP-1 is critical for the dendritic secretion of the neuropeptide NLP-12, and this is likely due to a defect in densecore vesicle transport from the primary dendrites into the high-ordered branches.
Next, we sought to determine whether loss of chdp-1 affects the transport of other types of intracellular organelles. The endoplasmic reticulum (ER) distributes in both the primary and some high-ordered dendrites in wild-type animals [31]. However, ER can only be observed in the primary dendrites in the chdp-1(tm4947) animals. Dynamic imaging analysis revealed that ER invaded and retracted in some of the 2 o and 3 o branches in the wild-type animals, while no such events was observed in the chdp-1(tm4947) animals. Similar defects were also observed for mitochondria (S9A- S9F Fig). We also examined DMA-1::GFP positive vesicles, presumably secretory vesicles and endosomes [30]. In wild-type animals, a small number of DMA-1:: GFP positive vesicles could be observed in the high-ordered branches. However, it was extremely difficult for us to identify any vesicles in the high-ordered branches of chdp-1 mutants. DMA-1::GFP strongly labeled the high-ordered branches in wild-type and chdp-1 (tm4947) mutant animals, making it a bit difficult to quantify the number of vesicles. We took advantage of the exoc-8 mutants, in which the docking/fusion of DMA-1 vesicles onto the dendritic membrane was strongly affected and confirmed that CHDP-1 was indeed required for the transport of the secretory vesicles/endosomes from the primary dendrites into the highordered branches (S10A-S10D Fig). Together, these results demonstrate an essential role of CHDP-1 in organelle transport in the high-ordered dendritic branches. Intracellular organelles are transported by motor proteins, dynein and kinesin, which move along the microtubule cytoskeletons. Thus, we asked whether loss of chdp-1 affects the assembly of the microtubule cytoskeleton in the PVD dendrites. To examine all the microtubules, which include both the dynamic and stable ones, we generated a GFP::TBA-1 transgene driven by the PVD promoter [25,28]. GFP signals could be observed in the primary and high-ordered dendrites in the wild-type control group. In chdp-1(tm4947) animals, GFP::TBA-1 strongly labeled the primary dendrites, while the signal was barely detected in the high-ordered branches (Fig 6F and 6G). To examine the dynamic microtubules, we expressed an EBP-2 (a plus-end binding protein)::GFP reporter in the PVD neurons [28,32]. In both strains, we found a similar number of mobile EBP-2 comets moving either towards the growth cone (anterograde) or the cell body (retrograde) in the growth cones of the anterior primary dendrites. We did not detect any difference regarding the microtubules' running length, growth duration or pause (S11A-S11F abnormality, we examined the distribution of the synaptic vesicle maker mCherry::RAB-3. We found no abnormal accumulation of RAB-3 in the dendrites, suggesting that chdp-1 mutants are not defective in microtubule assembly or polarity in the primary dendrites (S11G Fig). In contrast, we found a small number of EBP-2 comets either moving away from the 1 o /2 o branching sites (anterograde) or towards the branching sites (retrograde) in the 2 o branches of wild-type, but not the chdp-1(tm4947) animals (Fig 6H-6J). Thus, in addition to its role in actin cytoskeleton assembly, CHDP-1 is also required for microtubule cytoskeleton assembly in the high-ordered branches.

CHDP-1 may not function through CED-10
We previously reported that CHDP-1 acts through the small GTPase CED-10/Rac1 in the BDU and PLM neurons [16]. To examine whether CHDP-1 regulates dendritic cortical actin assembly in PVD neurons via a similar mechanism, we examined dendrite morphogenesis in n3246, a strong loss-of-function allele of ced-10 [42]. ced-10 (n3246) mutants did not show an increased number of 2 o and 3 o branches or faint staining of the 2 o branches by the myr-GFP reporter. The number of 4 o branches was reduced in ced-10 (n3246) mutants, but to a lesser extent compared to the chdp-1(tm4947) animals. In addition, we found that ER distribution in the high-ordered branches was not affected in ced-10 (n3246) mutants (S12A-S12E Fig). Together, our data suggest that CHDP-1 regulates dendritic cortical actin assembly via a CED-10-independent pathway or CED-10 only plays a minor role.

Loss of sax-1 genetically suppresses defects of chdp-1 knock-out animals
To understand how dendritic cortical actin assembly is regulated, we searched genes of which loss-of-function lead to excess membrane protrusions and thus serve as negative regulators in cortical actin assembly. A previous study reported that loss of sax-1, which encodes a conserved serine/threonine protein kinase, causes ectopic membrane protrusion formation [33]. Interestingly, overexpression of CHDP-1 also causes ectopic membrane protrusion [16]. Thus, we built genetic double mutants between chdp-1(tm4947) and sax-1(ky491) and examined the genetic interaction between the two genes. Interestingly, loss of sax-1 fully suppressed the increased number of 2 o and 3 o branches, decreased number of 4 o branches, and faint labeling of myr-GFP in the 2 o branches. Loss of sax-2, which genetically acts upstream of sax-1, also suppressed the defects mentioned above in chdp-1(tm4947) (Fig 7A-7C) [34]. Conditional knock-out of sax-1 in PVD neurons and other cells derived from the seam cell lineage also suppressed these defects in chdp-1(tm4947), indicating that SAX-1 acts cell-autonomously (S13A-S13C Fig). To gain more insights into the suppression, we also analyzed actin assembly in the dendrite growth cones, microtubule assembly and distribution of NLP-12-positive dense-core vesicles in the high-ordered dendrites. Loss of sax-1 also restored actin assembly, microtubule assembly and dense-core vesicle transport/distribution in the chdp-1(tm4947) mutants (Fig 7D  and 7E). The double mutant animals also showed nearly normal locomotion: both the amplitude and wavelength were similar to the wild-type control group (S13D-S13G Fig). Together, our results suggest that SAX-1 acts opposingly to CHDP-1, and we speculated that it is a negative regulator of cortical actin assembly.

Discussion
Here, we reported that the cell cortex-localized protein CHDP-1 acts cell-autonomously in the multi-dendritic PVD neurons to promote cortical actin assembly, which is critical for dendrite formation, maintenance and microtubule-based organelle transport (Fig 8). Neurons are specialized cell types, usually with a long unbranched axon and multiple highbranched dendrites. For dendrites, it has been known for decades that dendrite morphogenesis relies on actin assembly, which is regulated by Rac and Cdc42 [43]. Cortical actin is an enigmatic subset of the actin cytoskeleton. Assembly of cortical actin requires Rac1 and Arp2/3 [16,44,45]. However, these regulators may localize to both cell cortex and cytosol and thus not specific cortical actin regulators. To understand whether cortical actin assembly plays an important role during dendrite morphogenesis and maintenance, a manipulation that disrupts actin assembly at the cell cortex but not anywhere else is required. Our previous study and this study identified that CHDP-1, a cell cortex-localized actin assembly regulator, fulfils this requirement [16]. Through genetic analyses of the putative null mutants of chdp-1, we found that loss of chdp-1 caused an increased number of 2 o and 3 o branches, and a decreased number intensity of the 2 o dendrite to that of the primary branches, (middle) the ratio of the TBA-1::GFP intensity of the 2 o dendrite to that of the primary dendrites, (right) NLP-12::Venus fluorescence intensity in high-ordered dendrites in the 100 μm area anterior to PVD cell body for the genotypes indicated (normalized to WT). Error bars, SEM. ��� p < 0.001 by one-way ANOVA with the Tukey correction. ns: non-significant. n = 20-30 for each genotype. https://doi.org/10.1371/journal.pgen.1010381.g007

Fig 8. Working model of CHDP-1 in dendrite development and transport.
A cartoon showing how CHDP-1 regulates cortical actin assembly in the multi-dendritic neurons. In the wild-type animals, CHDP-1 localizes to the cell cortex and promotes cortical actin assembly. In the high-ordered dendrites, CHDP-1-dependent cortical actin assembly is likely required for microtubule assembly and microtubule-based transport. In chdp-1 knockout animals, cortical actin assembly in the high-ordered branches is reduced, and microtubule assembly and organelle transport are defective. Motor: kinesin or dynein. DCV: dense-core vesicle.
https://doi.org/10.1371/journal.pgen.1010381.g008 of 4 o branches. Disruption of actin assembly has been reported to cause less dendrite branching [24,25]. Thus, it is somehow unexpected that chdp-1 mutants contain increased 2 o and 3 o branches. From time-lapse recording, it seems this could be explained by increased 2 o initiation (although 2 o branch retraction was also observed). Possibly, deceased cortical actin assembly enables more filopodia formation derived from the primary dendrite. However, for the 4 o branching, disrupted intracellular transport probably affects branch initiation and stabilization. Using the STED super-resolution microscopy, we also observed opposite defective phenotypes for the diameters of 1 o dendrites vs 2 o branches: the former was enlarged and the latter was decreased. In addition, the diameter became less uniform for the 1 o dendrite in chdp-1 mutants, reminiscent of the dendrite width in unc-70/beta-spectrin mutants [10]. Future study is needed to fully understand why defective cortical actin assembly causes different effects for 1 o dendrites and 2 o branches. Loss of chdp-1 also resulted in spontaneous dendrite degeneration in the adults. Interestingly, loss-of-function mutations in unc-70 and alphaadducin also led to axon degeneration phenotypes [15,46]. Thus, our finding is consistent with the notion that membrane skeleton assembly is critical for neurite maintenance.

Cortical actin assembly is likely required for microtubule assembly and organelle transport
In the high-ordered branches, loss of chdp-1 reduced the cortical actin assembly and the microtubule assembly. Two models can explain the decreased microtubule cytoskeleton. (1) CHDP-1 directly acts as a microtubule assembly factor. (2) The effect of CHDP-1 on microtubule assembly is secondary to its function in cortical actin assembly. Actin cytoskeleton often interacts with microtubule cytoskeleton [1,47]. Numerous proteins have been reported to mediate structural interactions between microtubules and actin, such as coronin and Drebrin-EB3 [48,49]. Disrupting of actin cytoskeleton can lead to microtubule assembly defects [50]. Thus, it is conceivable to hypothesize that the cortical actin in the high-ordered branches can serve as a platform to promote microtubule assembly and stabilization. Future studies are needed to tell which model is correct.
The evolutionarily conserved serine/threonine protein kinase SAX-1/NDR1 is a putative negative regulator of cortical actin assembly SAX-1 shares high homology with its homolog in multiple species, including CBK1 in yeasts, Trc in flies, and NDR1/2 in humans [51]. Mutants of sax-1 was initially identified as the cell body of several types of neurons showed ectopic lamellipodia-like protrusions. Disrupting the endogenous RhoA function by expressing a dominant-negative RhoA transgene phenocopied the sax-1 loss-of-function phenotype, indicating that SAX-1 might act to modulate actin assembly [33]. Trc is required for proper cell shape and wing hair initiation in flies. Trc mutant cells contained more F-actin than that of the wild-type cells [52]. Knock-out of Trc in the class IV DA neurons results in ectopic dendrite branching and dendritic tilling defects. For dendritic branching, Trc kinase negatively regulates Rac signaling pathway [53]. However, it is unclear whether Rac is a phosphorylation target of Trc kinase. In the present study, we found that loss of sax-1 suppressed all the defects in the chdp-1mutant animals, including abnormal dendrite branching pattern, faint labeling of the 2 o branches by the myr-GFP reporter, reduced actin assembly in the growth cones of high-ordered branches, decreased microtubule assembly in the 2 o branches, defective organelle transport from the 1 o dendrites into the high-ordered branches and impaired proprioception. Very likely, restoration of the cortical actin assembly is the primary effect of sax-1 knock-out, and other phenotypic restorations are secondary. To the best of our knowledge, this is the first time that the SAX-1/NDR1 kinase family has been proposed as a putative negative regulator for cortical actin assembly. Currently, the direct phosphorylation target of SAX-1 in this process is not clear. Ultanir et al. identified several membrane traffic-related phosphorylated targets for NDR1 kinase using a chemical genetical approach [54]. A future study using a similar strategy will likely uncover the direct downstream player(s) of SAX-1 in the negative regulation of cortical actin assembly.
Together, our results showed that CHDP-1 promotes cortical actin assembly in multi-dendritic neurons. Compromising this process causes defects in dendrite branching, microtubule assembly, organelle transport, neuropeptide secretion and proprioception. Furthermore, our data also suggest that the evolutionarily conserved SAX-1/NDR1 kinase is a putative negative regulator of cortical actin assembly. CHDP-1 and SAX-1 function in an opposing manner to balance cortical actin assembly in neurons and perhaps other non-neuronal cell types. Our study will shed light on the enigmatic mechanisms and functions of neuronal cortical actin assembly.

Confocal imaging of C. elegans
For static imaging, worms were immobilized using 1 mM levamisole solution and placed on 4% agarose pads, and then imaged using an Olympus IX83 fluorescence microscope equipped with a spinning-disk confocal scanner (Yokogawa CSU-W1), an sCMOS camera (Prime 95B), and a 60x oil Apochromat objective (NA: 1.49). Z stack images were processed by the projection of maximum intensity except for Figs 2A and S5A.
Time-lapse imaging was performed as previously described with some modifications [60]. Briefly, 2 μl of 1 mM levamisole solution was added into the center of the glass bottom of a microwell dish, then about 20 worms were transferred into the drop of levamisole solution. Next, a 4% agarose pad was gently added onto the animals. All time-lapse movies were taken using the spinning-disk confocal microscope, and Z stack images were processed by projection of maximum intensity except for S10 and S11 Figs, in which single-layer images were shown.
For Stimulated Emission Depletion Microscopy (STED) imaging, worms were immobilized using 1 mM levamisole solution and placed on 4% agarose pads. A Leica TCS SP8 STED fluorescence microscope equipped with 592/660/775 nm lasers, and a HC PL APO CS2 100×/1.40 oil objective was used for imaging. Z stack images were processed by the projection of maximum intensity.

Split-GFP assay for cell-specific detection of CHDP-1 expression
This assay was performed as described by Siwei He et al. [39]. Briefly, the coding sequence for GFP11 7x was amplified from a previously published plasmid (Addgene #70224). The PCR product was assembled with two homology arms (~500 bp each) amplified from the N2 genomic DNA, and the digested pSM delta plasmid as the backbone via the Gibson assembly protocol to generate the repair template plasmid pZT105. A similar CRISPR knock-in strategy was used as how gfp::chdp-1 was generated, except that pZT105 was used in this experiment. Successful knock-in animals were obtained through PCR-based genotyping, and no additional mutation was found by Sanger sequencing. The coding sequence for GFP1-10 was amplified from a previously published plasmid (Addgene # 70219). The PCR product was cloned into a plasmid in which the PVD-specific promoter ser2prom3 was previously inserted. This plasmid (pZT106, 20 ng/μl), Podr-1::rfp (50 ng/μl) and Pmyo-2::mcherry (2 ng/μl) were injected into the zac283[gfp11(7x)::chdp-1] strain. Stable transgenic lines were isolated via the expression of the co-injection markers and subjected to confocal imaging.

Locomotion assay
This assay was performed following a previous protocol with some modifications [61]. Briefly, 10-20 worms in the late L4 stage were transferred individually into fresh NGM plates, and then the plates were put in a 20˚C incubator for 1-2 hours. Images of the crawl tracks were taken using a Nikon SM218 stereo microscope with a 1x SHR Plan Apo objective (NA: 0.15). The trajectory's amplitude (the distance between opposite peaks) and wavelength (the distance between two successive peaks) were measured using ImageJ. For each strain, the crawling trajectories of 10-20 worms were measured.

Quantification and statistical analysis
For PVD dendrites, the number of dendrites of 2˚, 3˚, and 4˚is the number within the 100 μm region anterior to the PVD cell body. For the fluorescence intensity and width of dendrites, ImageJ is used for statistics. The numerical data that underlies graphs or summary statistics were summarized in S2 Table. The Student's t-test (for the difference between two groups) or one-way analysis of variance with Tukey correction (for the difference between three or more groups) was used for statistical analysis.