RTKN-1/Rhotekin shields endosome-associated F-actin from disassembly to ensure endocytic recycling

Yan et al. identify C. elegans RTKN-1 as a novel endocytic recycling regulator that impedes UNC-60A/cofilin-mediated endosomal actin disassembly. Further evidence indicates that the self-binding capacity of RTKN-1 is indispensable for this functionality and SDPN-1/Syndapin acts to direct the proper residency of RTKN-1 in recycling endosomes.


Introduction
Endocytosed membrane proteins are usually delivered to early endosomes, where initial cargo sorting occurs. After that, some membrane proteins will be sent back to the cell surface via the recycling pathway (Grant and Donaldson, 2009;Hsu and Prekeris, 2010;Cullen and Steinberg, 2018;Liu et al., 2018). Increasing evidence has suggested that actin structures play crucial roles in concerted regulation of endocytic recycling (Wang et al., 2016;Delevoye et al., 2016;Puthenveedu et al., 2010;Gong et al., 2018;Gleason et al., 2016). For instance, when EGFR and MT1-MMP are mutated and thus lose actinbinding capacity, their recycling transport is severely impaired (MacDonald et al., 2018). Conversely, the insertion of an actinbinding module into a cargo destined to lysosomes leads to the missorting of the cargo into the recycling route (MacDonald et al., 2018). In agreement with these observations, WASH and Arp2/3 can be localized onto endosomes to administer local actin dynamics, therefore promoting the biogenesis of tubular membrane carrier (Puthenveedu et al., 2010;Simonetti and Cullen, 2019;Gong et al., 2018). WASH-mediated actin assembly was also suggested to assist dynamin-mediated membrane remodeling (Derivery et al., 2009). Our recent mechanistic inquiry showed that CAMSAPs/Patronin homologue PTRN-1 resides in the sorting endosomes and unlocks the self-inhibitory configuration of CYK-1, hence facilitating actin polymerization during recycling (Gong et al., 2018). Actin disassembly factor ADF/cofilin has also been implicated in endocytic trafficking. During metastasis, a decrease of cofilin phosphorylation disturbed the endocytosis of EGFR, and this defect could be alleviated by the overexpression of a cofilin phosphomimicking (S3E) mutant (de Graauw et al., 2014). In Caenorhabditis elegans, UNC-60A/cofilin promotes actin dynamics in the invadopodia of anchor cells. In the absence of UNC-60A, invadopodia membrane recycling is undermined (Hagedorn et al., 2014). Various actin-binding proteins, including myosin and tropomyosin, have been shown to compete with cofilin for F-actin binding and thus affect the severing efficacy of cofilin (Ngo et al., 2016;Ono and Ono, 2002;Christensen et al., 2017). Specifically, the motor domain of myosin inhibits the generation of cofilin clusters along F-actin (Ngo et al., 2016). Likewise, tropomyosin forms head-to-tail multimers along F-actin and displaces cofilin (Li et al., 2011;Meiring et al., 2018;Jansen and Goode, 2019).
Rhotekin/RTKN was initially identified as a binding protein of RhoA and RhoC in mice (Reid et al., 1996). Rhotekin contains an N-terminal Rho-binding domain (RBD) and a central pleckstrin homology (PH) domain, followed by two proline-rich motifs (PRMs) and a PDZ domain recognizing X(S/T)XV-COOH consensus motif (Ito et al., 2018). The PH domain possesses an affinity for phosphoinositide and aids in localizing proteins to specific membranous structures (Garcia et al., 1995;Mao et al., 2009). Recently, Rhotekin was implicated in the interaction with the Src homology 3 (SH3) domain of vinexin/SORBS3 (Ito et al., 2018). The vinexin SH3 domain interacts with the PRM and recruits Rhotekin to the midbody during cell division (Nagata et al., 2009;Chang and Huang, 2017). Also, through interaction between the C-terminus and the PDZ domain of LIN-7B, Rhotekin can be located at the junctions of MDCK cells and participate in establishing or maintaining cell polarity (Sudo et al., 2006). It is worth noting that RTKN2/Rhotekin was upregulated in colon cancer cells. In the absence of RTKN2, the proliferation rate slowed down, and the cell cycle progression was affected (Pang et al., 2017). Furthermore, the Rho-Rhotekin signaling pathway has been involved in septin cytoskeleton assembly, which is closely correlated to cancer development (Ito et al., 2005).
Here, we found that lack of RTKN-1/Rhotekin undermined basolateral recycling transport and actin integrity in the C. elegans intestine. Specifically, in the absence of RTKN-1, newly formed membrane carriers still emanated from recycling endosomes but failed to move along actin remnants and finally retracted. Subsequent assays revealed that RTKN-1 was capable of impeding UNC-60A-mediated actin disassembly. Accordingly, RTKN-1 deficiency-induced recycling defect was alleviated by UNC-60A knockdown. Further testing showed that the interaction between RBD and the C-terminal segment (CT) is required for the oligomerization of RTKN-1. The overexpressed RTKN-1ΔCT could neither defer actin disassembly nor alleviate the recycling defects in rtkn-1 mutants. Finally, we identified that recycling regulator SDPN-1/Syndapin interacted with PRM and thus directed the proper dwelling of RTKN-1 in recycling endosomes. Together, these findings led to a model of actin homeostasis during endocytic recycling where RTKN-1 stabilizes endosome-associated F-actin, promoting the formation and directional movement of recycling membrane carriers.

RTKN-1 is required to sustain actin integrity in the intestinal epithelia
Dynamic actin architecture facilitates membrane carrier biogenesis and movement during endocytic recycling (Campellone and Welch, 2010;Wang et al., 2016;Gong et al., 2018). To better understand the mechanisms underlying recycling endosomeassociated actin dynamics, we performed a two-step RNAi genetic screen in C. elegans, attempting to identify new regulators in this process (see Materials and methods for detail). We used recycling cargos to follow the recycling aberration in the intestinal epithelia (Chen et al., 2006). Next, we sought to recognize actin regulators among the subset of genes whose expression knockdown leads to recycling defects. To this end, we identified C07B5.4 as a candidate, which encodes a homologue of Rhotekin. Notably, although Rhotekin is implicated in the septin assembly (Ito et al., 2005), the interaction between Rhotekin and actin dynamics remains unknown.
Here, we named C07B5.4 as RTKN-1 with reference to the nomenclature of its mammalian homologue and employed an rtkn-1(ok1404) null allele bearing a 1,345-nucleotide intragenic deletion (outcrossed three times), which led to an early nonsense mutation (Fig. 1 A). By measuring growth time from the L1 larval stage to the young adult stage, we observed that 60 h after the L1 stage, ∼80% of rtkn-1 mutants were still at the L4 stage (Fig. S1, A-A"), highlighting that RTKN-1 could play a role during development. Indeed, the intestine of rtkn-1 mutant adults appeared to be thinner (Fig. S1 B). To present the location of fluorescent proteins, the imaging plane close to the basal membrane of the intestinal cells was defined as "top," and the imaging plane where the apical membrane and lumen can be observed was defined as "middle" (deep cytosol; Fig. 1 B). The utrophin calponin homology (CH) domain (Utrophin-CH) bears a binding specificity for F-actin and has been adopted as a faithful actin filaments probe in C. elegans (Mizumoto and Shen, 2013;Chia et al., 2012). We validated that Utrophin-CH-labeled actin structures near the basolateral membrane were reduced in rtkn-1 mutants, leaving a few cytoplasmic punctate structures ( Fig. 1, C-C"). Phalloidin has been used to label endogenous actin in C. elegans (Gong et al., 2018). Likewise, while the level of actin in rtkn-1 mutant intestine is comparable to that of WT animals (Fig. S1 D), phalloidin-stained actin structures were severely reduced upon loss of RTKN-1 (Fig. S1, C and C'), corroborating that RTKN-1 plays a critical role in maintaining the integrity of actin structures in the intestinal epithelia.
RTKN-1 is broadly expressed in C. elegans (Birchall et al., 1995;Spencer et al., 2011), including intestine and coelomocyte. We prepared a CRISPR-based 2xFlag knockin strain to examine RTKN-1 level via Western blot (Table S1). The results showed that the endogenous RTKN-1 could be readily detected (Fig. S4  E). Moreover, we observed that endogenous RTKN-1 resides on the actin structures in the intestine (Fig. S1,E and E'). To obtain an ideal fluorescent signal-to-noise ratio, we deployed the integrated low-expression transgenic strains (intestinal vha-6 promoter driven) in the subsequent studies. Consistent with the localization of the endogenous protein, RTKN-1-mCherry maintained a notable dwelling on the actin structures near the basolateral membrane (Fig. 1, D-D").
To probe the details underlying the actin diminishment phenotype, we tracked the dynamic changes of Utrophin-CH-labeled actin structures. In this respect, we noted that intestinal actin structures are relatively stable, with occasional attachment, detachment,and wobble (Fig. 1,E and E'). However, in RTKN-1-depleted cells, actin structures are very dynamic and frequently elongate, retract, break, and fade out (Fig. 1, E and E'), consolidating the role of RTKN-1 in maintaining actin integrity.

Loss of RTKN-1 leads to basolateral recycling defects
To assess the specificity of RTKN-1 in the recycling pathway, we examined the steady-state distribution of various cargo proteins in rtkn-1 mutants (Chen et al., 2006;Shi et al., 2007). We first validated the intracellular itinerary of hTAC (clathrin-independent recycling cargo, the α-chain of the human interleukin 2 receptor) by assaying the distribution of hTAC-GFP in rab-7(RNAi) animals. In the absence of RAB-7, the distribution of hTAC-GFP was unaffected, and clathrin-dependent nonrecycling cargo GFP-CD4-LL overaccumulated in the cytosol (Fig. S3, A-B'; Gu et al., 2013). These results suggested that hTAC-GFP undergoes continuous internalization and recycling with negligible sorting to late endosomes and lysosomes. Likewise, TfR (clathrindependent recycling cargo, the human transferrin receptor [hTfR]) has been reported to be continuously internalized and recycled, and its transit to late endosomes is negligible (Sönnichsen et al., 2000). Next, we assessed the recycling implication of RTKN-1 and observed that lack of RTKN-1 caused aggregation of hTAC-GFP and hTfR-GFP in the cytosolic structures (Fig. 3, A and A'; and Fig. S3, C and C'), whereas the distribution of the clathrin-dependent retrograde cargo MIG-14-GFP or nonrecycling cargo GFP-CD4-LL was not affected (Fig. S3, F-G'; Gu et al., 2013;Shi et al., 2009), indicating that RTKN-1 is implicated in the recycling regulation of both clathrin-dependent and clathrin-independent cargos.
To clarify whether RTKN-1 is implicated in apical recycling, we examined the distribution of low-density lipoprotein receptor-related protein 1 (LRP-1)-GFP. LRP-1 localizes to the apical plasma membrane of C. elegans hypodermis and can be recycled from endosomes to the plasma membrane (Yochem et al., 2015). In the intestine, LRP-1-GFP mainly resides on the punctate structures around the lumenal apical membrane, and it is rarely observed on the basal side (Fig. S3,H and H'). In the absence of RTKN-1, there was no significant aberration of LRP-1-GFP distribution (Fig. S3, H and H'). Together, these results suggested that RTKN-1 is explicitly required for basolateral recycling.
To assess the functional step of RTKN-1 in the recycling pathway, we compared the localization of recycling cargos with that of the early endosome marker mCherry-RAB-5 and the recycling endosome marker mCherry-RME-1 Gao et al., 2020). In the WT animals, hTAC-GFP and hTfR-GFP primarily appeared in mCherry-RME-1-labeled tubular and punctate structures and, to a lesser extent, in mCherry-RAB-5-positive endosomes (Fig. 3, B-C'; and Fig. S3, D-E'). In the absence of RTKN-1, the labeling of hTAC-GFP or hTfR-GFP in recycling endosomes was reduced, while their dwelling in early endosomes was increased (Fig. 3, B-C'; and Fig. S3, D-E'), implicating RTKN-1 in the transit between early endosomes and recycling endosomes. By using live animal imaging, we checked the dynamics of endosomal structures. Normally, hTAC-GFP-labeled membrane tubules are highly dynamic, with frequent extension and detachment (Fig. 3,D and D'). In contrast, in RTKN-1-depleted cells, hTAC-GFP-positive tubules were still able to extend but often retracted (Fig. 3, D and D'), indicative of the impaired biogenesis of recycling membrane carrier (Gao et al., 2020).

RTKN-1 acts independently of the Rho family of GTPases during endocytic recycling
Studies in mammals identified Rhotekin as an effector of the Rho family of GTPases during cell polarity establishment and neuronal synapse formation (Reid et al., 1996;Ito et al., 2018). Here, we sought to assay the interaction of RTKN-1 with canonical Rho family members, including RHO-1/RhoA, CDC-42/Cdc42, and CED-10/Rac1. The GST pulldown assay revealed specific binding of RTKN-1 to active RHO-1(GMP-PNP) (Fig. 4 A). We then examined whether the depletion of these GTPases affects the localization of RTKN-1. To circumvent embryonic lethality, synchronized L1 larvae were fed with double-stranded RNA-producing bacteria and scored at the day 1 adult stage ( Fig. 4, B-E'). Unexpectedly, the distribution of RTKN-1-GFP was not affected by the loss of these GTPases (Fig. 4, F and F'), implying that RTKN-1 may not act as an effector of the Rho family of GTPases in intestinal cells. Consistently, the localization of RTKN-1-GFP did not show significant variations in cells Rhotekin has been shown to sequester active Rho GTPase (Fu et al., 2000). Also, Rhotekin has been suggested to enhance RhoA activity and stress fiber formation (Pusapati et al., 2012). Per this inference, loss of Rho GTPases should disturb recycling, as observed in RTKN-1-deficient cells. However, there were no distribution irregularities of hTAC-GFP in RHO-1-or CDC-42-depleted cells (Fig. 4, G and G'). Likewise, the subcellular localization of hTAC-GFP was not affected by the overexpression of RHO-1(G14V) or RHO-1(T19N) (Fig. 4, G and G'). Notably, CDC-42 deficiency induced intracellular overaccumulation of hTfR-GFP (Fig. 4, H and H'). Since the loss of CDC-42 failed to disturb the distribution of recycling endosomes (Gao et al., 2020), we speculated that this phenotype could be due to the blockage of ESCRT-mediated degradative traffic (Macara and Spang, 2006;Lu and Drubin, 2020). It should also be noted that CED-10 is capable of recruiting RAB-5-GAP/TBC-2 onto endosomes, thus shutting down RAB-5 activity to facilitate the recycling (Sun et al., 2012). Given the intact localization of RTKN-1-GFP in CED-10-depleted cells (Fig. 4, F and F'), the recycling implication of RTKN-1 is unlikely to be directed by CED-10. Taken together, these results suggested that RTKN-1 functions independently of the Rho family of GTPases to facilitate recycling in the intestine and that Rho/Rhotekin signaling is unlikely involved in this process.

RTKN-1 impedes UNC-60A/cofilin-mediated actin disassembly
Our results introduced RTKN-1 as a novel regulatory protein involved in recycling transport and actin integrity. To directly evaluate the requirement of F-actin for recycling flow in the C. elegans intestine, we assayed the dynamic correlation between Utrophin-CH-and hTAC-labeled structures. In a WT background, newly formed hTAC-positive membrane carriers emanated from recycling endosomes, moved along the actin filaments, and finally fused with the opposite endosomal structures (Fig. 5, A and A'; and Video 1). However, in the absence of RTKN-1, some hTAC-positive tubules were still extending but eventually retracted, and Utrophin-CH-labeled actin structures barely overlapped with these membrane tubules (Fig. 5, A and A'; and Video 2). These results corroborated the notion that endosome-associated F-actin is the physical support for the formation, stabilization, and movement of tubular membrane carriers (Simonetti and Cullen, 2019). It is also worth noting that Utrophin-CH preferentially labels Arp2/3-dependent branched F-actin, such as endocytic actin patches and synaptic actin network in C. elegans HSN neuron (Chia et al., 2014). Thus, we speculated that the highly dynamic recycling endosome is mainly associated with branched F-actin structures. Indeed, the branched F-actin has been reported to play roles in intracellular membrane dynamics, such as shaping and fission (Picco et al., 2018;Kessels and Qualmann, 2021;Ripoll et al., 2018;Akamatsu et al., 2020).
Given the above findings, we probed whether RTKN-1 can interact with F-actin using an F-actin cosedimentation assay. We noticed that the presence of increasing concentrations of F-actin gradually boosted the proportion of GST-RTKN-1 in the precipitate (Fig. 5,B and B';and Fig. S4,A and A'). We validated the F-actin interaction of RTKN-1 via coimmunoprecipitation (Fig.  S4 B). Actin filaments often undergo additional assembly (branching and bundling) and disassembly (severing and depolymerization) to establish a network, fulfilling diverse functions (Paavilainen et al., 2008;Pollard, 2016). The actin structure decrease in rtkn-1 mutants raised the possibility that RTKN-1 could promote actin polymerization directly. Alternatively, RTKN-1-mediated repression of actin depolymerization can reconcile this phenotype. To distinguish these possibilities, we examined the effects of RTKN-1 on actin in vitro (F-actin 10 µM). The presence of RTKN-1 (0.25-1 µM) had no discernible effect on actin polymerization (Fig. 5 C and Fig. S4 C). Also, RTKN-1 failed to restrain the spontaneous F-actin depolymerization (Fig. 5 D). Considering that PTRN-1/CAMSAPs promote actin polymerization during endocytic recycling (Gong et al., 2018), our data prompted that the positive effect of RTKN-1 on endosomal actin integrity is indirect and that the actin disassembly factor could be the regulatory object of RTKN-1.
Given the actin-binding potential of RTKN-1, we reasoned that RTKN-1 could mitigate the competence of UNC-60A. To test this idea, we first prepared CRISPR-based knockin strains UNC-60A::2xFlag and RTKN-1::2xFlag. Through Western blotting, we estimated that the endogenous RTKN-1/UNC-60A molar ratio is ∼1.136 (Fig. S4 E). Next, we assayed the efficiency of F-actin disassembly accordingly. As expected, UNC-60A (1 µM  In the presence of RTKN-1 (0.5, 1, and 2 µM), the depolymerization was substantially deferred, and this effect was concentration dependent (Fig. 5 E). Regarding this observation, one explanation is that the interaction between RTKN-1 and actin could interfere with the association of UNC-60A with F-actin. UNC-60A was shown to associate with F-actin for a short period, quickly disassemble F-actin, and enter the supernatant (Ono and Benian, 1998). This dynamic allocation precludes us from examining the actin-binding status of UNC-60A in vitro. To evaluate this possibility, we turned to determine whether the residency of UNC-60A in actin structures can be affected by RTKN-1 in vivo. Phalloidin is likely to compete with cofilin for F-actin binding (Nishida et al., 1987), causing falsepositive results. Thus, we deployed Utrophin-CH and noticed that although overexpressed RTKN-1 led to an increase of actin structures, the localization of UNC-60A in actin structures was reduced (Fig. 5, F-F"). In the absence of RTKN-1, actin structures near the basolateral membrane were decreased, and UNC-60A could be found on actin remnants (Fig. 5, F and F'). Importantly, we observed no evident overlap between RTKN-1-GFP and UNC-60A (Fig. 5, G and G'). Likewise, we did not identify the binding of RTKN-1 to UNC-60A (Fig. S4 H). Together these findings suggested the existence of an RTKN-1-mediated impediment toward the F-actin association of UNC-60A and that this process does not involve an RTKN-1-UNC-60A interaction.
Loss of UNC-60A alleviates recycling defects in RTKN-1-depleted cells Our results indicated that RTKN-1 could antagonize UNC-60A-mediated endosomal actin disassembly. To determine the reciprocal influence of UNC-60A on RTKN-1-assisted actin integrity, we assayed the effects of the UNC-60A deficiency in rtkn-1 mutants. In line with the canonical role of cofilin, actin structures overaccumulated in UNC-60A-depleted cells and diminished upon overexpression of UNC-60A-mCherry (Fig. 6, A and A'), which also indicated that transgenic UNC-60A-mCherry effectively disassembled F-actin in vivo. Of note, in RTKN-1-depleted cells, loss of UNC-60A augmented the level of actin structures (∼38% recovery; Fig. 6, A and A'), suggesting that actin decrease in rtkn-1 mutants is due at least in part to the enhanced efficacy of UNC-60A. By contrast, the subcellular localization of RTKN-1 was not affected by UNC-60A depletion (Fig. S4, I-J'), implying that the endosomal targeting of RTKN-1 is independent of UNC-60A.
To determine whether RTKN-1 deficiency-induced elevation of UNC-60A potency is implicated in the recycling defects, we examined the distribution of recycling cargos in rtkn-1;unc-60a double knockdown animals. In agreement with the variation of actin structures, the overaccumulation of hTAC-GFP or hTfR-GFP in RTKN-1-deficient cells was mitigated by the concomitant loss of UNC-60A (Fig. 6, B-C'). Further highlighting the significance of actin integrity in recycling transport, ectopic expression of UNC-60A-mCherry led to recycling cargo overaccumulation in the cytosol (Fig. 6, B-C').
RBD and PH-CT are required for the functionality of RTKN-1 RTKN-1 contains an RBD, a PH domain, and a C-terminus bearing a PRM. RBD mediates interactions with the Rho family of GTPases (Dvorsky et al., 2004;Ito et al., 2018), while the PH domain often presents a phospholipid affinity (Someya et al., 2001;Xu et al., 2018). We split RTKN-1 into an RBD segment and a portion comprising the PH-CT (Fig. 7 A). We observed that RBD still interacted with F-actin (Fig. 7, B and B'). Live cell imaging showed that RBD-GFP appeared diffusively in the cytosol (Fig. 7, C and C'). Surprisingly, PH-CT-GFP did not display a meshwork pattern and often resided in the intracellular puncta (Fig. 7, C and C'), suggesting that the RBD-actin interaction contributes to RTKN-1 localization and/or maintenance of recycling endosomal architecture.
We then probed the functional relevance of RBD and PH-CT. As expected, the expression of PH-CT-mCherry failed to relieve the actin diminishment phenotype in RTKN-1-depleted cells (Fig. 7, D and D'). Unexpectedly, the presence of RBD did not restore the level of actin structures (Fig. 7, D and D'), indicating that the RTKN-1-assisted actin integrity requires the synergism of RBD and the downstream region. Consistently, overexpression of RBD-mCherry or PH-CT-mCherry did not complement the recycling defects (Fig. 7, E-F'), establishing the implication of both RBD and PH-CT in recycling modulation.
CT is imperative for RTKN-1 oligomerization Tropomyosin can be joined end to end to form a cable that extends along the sides of the actin filaments (Christensen et al., 2017;Skoumpla et al., 2007), thus competing with cofilin for actin association (Kuhn and Bamburg, 2008;Ono and Ono, 2002). Our evidence suggested that RTKN-1 could interfere with the association of UNC-60A with endosomal actin structures. An intriguing scheme arises as to whether RTKN-1 is capable of self-associating and thus shields F-actin from cofilin. To test this idea, we first identified the self-binding capacity of RTKN-1 (Fig. 8, A and B), which is mediated by the RBD and CT beyond the PH domain (Fig. 8, C and C'; and Fig. S5 A). Accordingly, the RTKN-1 fragment lacking the CT (RBD-PH) lost the self-binding potential (Fig. 8 D). Of note, the C-terminal PRM is not required for this self-association (Fig. 8 E). Next, we determined whether RTKN-1 could form oligomers via this RBD-CT interaction. In SDS-PAGE gel, purified HA-RTKN-1 yielded a band around 59 kD (Fig. 8 F). However, after protein was covalently crosslinked, the HA-RTKN-1 band shifted to a high molecular mass range, and a prominent band of ∼669 kD appeared in the gel (Fig. 8 G). In contrast, HA-RBD-PH remained in the form of the monomer (Fig. 8 G). Likewise, uncrosslinked HA-RTKN-1 presented a band of ∼669 kD in the native PAGE gel (Fig. S5 B). Together these results suggested that RTKN-1 could form an oligomer, and CT is indispensable for this assembly.
Our previous study showed that actin-bundling protein EHBP-1 could assist in recycling endosomal tubulation (Wang et al., 2016). Nonetheless, by implementing a low-speed cosedimentation assay, RTKN-1 did not show actin-bundling potency (Fig. S5, D-E'), suggesting that oligomerization does not endow RTKN-1 with the ability to promote actin bundling. Recently, PTRN-1/CAMSAP was found to promote actin polymerization during endocytic recycling (Gong et al., 2018). In the absence of PTRN-1, the distribution of RTKN-1-GFP was intact (Fig. S5, F and F'), indicating that the subcellular localization of RTKN-1 does not require PTRN-1. Indeed, these results are consistent with the identification of PTRN-1 as a resident protein in punctate early/sorting endosomes, modulating actin network architecture there.

CT is a determinant for actin integrity during basolateral recycling
To explore the significance of the C-terminus of RTKN-1, we first analyzed the actin-binding capability of the RBD-PH. As expected, RBD-PH maintains an interaction with F-actin (Fig. 9, A and A'). However, RBD-PH-GFP only exhibited punctate labeling (Fig. 9, B and B'), suggesting that the CT is essential for the generation and/or maintenance of recycling endosomal architecture. Consistently, the recycling endosomal meshwork labeled by EHBP-1-GFP collapsed in rtkn-1 mutants (Fig.  S5, G and G'). We next assessed whether CT is required for actin integrity. Remarkably, the UNC-60A-assisted actin depolymerization was not affected by the presence of RBD or RBD-PH ( Fig. 9 C and Fig. S5 C), indicating that RBD-actin interaction is not sufficient to defer actin disassembly. After that, we assayed the effect of RBD-PH overexpression on colocalization between UNC-60A and actin structures. The overlap between UNC-60A and actin was not significantly affected (Fig. 9, D and D'). Similar to the rtkn-1 mutants, RTKN-1-depleted cells coexpressing RBD-PH-mCherry exhibited a distinguished decrease in Utrophin-CH-labeled actin structures (Fig. 9, E and E'). Also, the overexpression of RBD-PH-mCherry failed to alleviate the overaccumulation of recycling cargos in rtkn-1 mutants (Fig. 9, F-G'). Taken together, these results highlighted the importance of the RTKN-1 C-terminus as an actin integrity determinant during recycling transport.

SDPN-1 acts to direct the proper residency of RTKN-1 in recycling endosomes
This study noticed that RTKN-1 colocalizes well with the recycling endosomal actin regulators SDPN-1 and EHBP-1, whereas it overlaps with a recycling endosome-associated dynamin-like protein RME-1 to a lesser extent. This distinct phenotype prompts a correlation between RTKN-1 and specific recycling regulators. It is noteworthy that the SH3 domain has been identified as a PRM binding partner (Mayer, 2001;Lim et al., 1994). Therefore, we scanned the sequences of the major recycling regulators and found that SDPN-1/Syndapin and AMPH-1/Amphiphysin carry an SH3 domain. We tested the interaction and observed that RTKN-1 solely binds to SDPN-1 (Fig. 10, A and B). Further analyses revealed the necessity of PRM in the RTKN-1-SDPN-1 interaction (Fig. 10 C).
To further elaborate on the functional significance of this interaction, we examined the distribution of actin structures in RTKN-1-or SDPN-1-deficient cells. As expected, the residence of GFP-Utrophin-CH-labeled F-actin on Tubby-PH (R332H) -positive recycling endosomes was reduced by the loss of RTKN-1 or SDPN-1 (Fig. 10, G and G'). Consistent with this observation, although RTKN-1ΔPRM overexpression could somewhat alleviate the actin diminishment in rtkn-1 mutants (Fig. S5, H and H'), it was not sufficient to fully rescue the hTAC-GFP recycling defects (Fig. 10, H and H'). Similarly, overexpression of RTKN-1 or RTKN-1ΔPRM could not effectively relieve the overaccumulation of hTAC-GFP in sdpn-1 mutant animals (Fig. 10, H  and H'). Altogether, these results suggested that SDPN-1 acts upstream to direct the proper association of RTKN-1 with recycling endosomes and promote the integrity of endosomeassociated actin structures.

Discussion
Here, we identified RTKN-1/Rhotekin as a requisite for the basolateral recycling in the intestine. Particularly, RTKN-1 mitigated the efficacy of the actin disassembly factor UNC-60A/cofilin. Further analyses suggested that SDPN-1 acts to sequester RTKN-1 onto recycling endosomes, thereby stabilizing endosome-associated F-actin and facilitating membrane carrier formation and movement. The Rho GTPases have been involved in regulating the actin architecture (Lee and Dominguez, 2010;Camera et al., 2003;Leung et al., 1996;Mott et al., 1999;Bishop and Hall, 2000). Cdc42 exposes the binding surface within N-WASP and activates branched actin nucleator Arp2/3 (Rohatgi et al., 2000). Through binding to mDia/formin GTPase binding domain, RhoA disrupts the self-inhibiting conformation of mDia (Higashida et al., 2004). In addition, RhoA promotes phosphorylation of LIMKs via effector ROCK. Active LIMKs further phosphorylate cofilin, thus inhibiting F-actin disassembly (Geneste et al., 2002). Notably, RhoA effector Rhotekin was observed to affect the activity of RhoA via a feedback mechanism. The overexpression of an S435E subcellular localization of RTKN-1-GFP, RBD-GFP, and PH-CT-GFP in intestinal cells. Box-and-whisker plots (n = 18 cells): 10-90th percentile; dots, outliers; red midline, median of WT; boundaries, quartiles. ***, P < 0.001 by one-way ANOVA followed by Dunn's post hoc multiple comparison test. (D-F9) Confocal images showing the distribution of GFP-Utrophin-CH, hTAC-GFP, and hTfR-GFP in intestinal cells. Box-and-whisker plots (n = 18 cells): 10-90th percentile; dots, outliers; red midline, median of WT; boundaries, quartiles. ***, P < 0.001 by one-way ANOVA followed by Dunn's post hoc multiple comparison test. The black asterisks in the image panels indicate intestinal lumen. Scale bars, 10 µm. Ctrl, control; MC, mCherry; P, pellet; S, supernatant.  Rhotekin caused an increase in active RhoA level and the stress fiber formation (Pusapati et al., 2012). We validated the interaction of RTKN-1/Rhotekin with active RHO-1/RhoA. However, basolateral recycling was not impaired by loss of RHO-1, suggesting that RHO-1 does not act as a regulator of RTKN-1 during endocytic recycling in the intestine. Furthermore, unlike the way in which ROCK directs cofilin activity via phosphorylation (Geneste et al., 2002), RTKN-1 could attenuate the potency of UNC-60A/cofilin by impeding the F-actin association of UNC-60A. Coronin promotes the association of cofilin with F-actin followed by recruitment of AIP1, which synergistically acts on F-actin for rapid depolymerization (Kueh et al., 2008;Ishikawa-Ankerhold et al., 2010). UNC-78 and AIPL-1 are homologues of AIP1 in C. elegans, both of which were perceived to function together with UNC-60B/cofilin and have no engagement with the UNC-60A/cofilin (Mohri and Ono, 2003;Ono and Ono, 2014). The worm homologues of coronin, COR-1 and POD-1, are predicted to be expressed in the intestine (WormBase version WS241). We thus examined the effect of COR-1 and POD-1 on actin depolymerization. Nonetheless, UNC-60A-mediated actin disassembly was not affected by COR-1 or POD-1. Given that UNC-60B was found to be solely expressed in muscle (Ono and Ono, 2002), these findings suggested that although coronin could synergize with AIP1 and UNC-60B in specific tissues or biological processes, the acting scenario of UNC-60A in intestinal epithelia appears to be quite different.
Our analysis prompts that RTKN-1 could adopt two distinct conformations, one of which is a self-inhibiting configuration. Given the distinct interacting partners of the RBD and PH domain (F-actin and membrane, respectively), the self-inhibiting configuration would hinder the interaction of RTKN-1 with the corresponding partners. An alternative conformation could be the formation of oligomers, which could extend on F-actin, thereby limiting the presence of cofilin in actin structures. Further biochemical and structural analysis will be required to justify this interpretation. Regarding the positioning mechanism of RTKN-1/Rhotekin, previous studies showed that the midbody residency of Rhotekin during cell division is governed by vinexin (Chang and Huang, 2017). Vinexin belongs to the sorbin adaptor family, and C. elegans SORB-1 is most homologous to vinexin (Loveless et al., 2017). Nevertheless, in sorb-1(RNAi) animals, the distribution of RTKN-1-GFP or hTAC-GFP was not affected (Fig. S5, I-K), suggesting that the regulatory role of RTKN-1 during recycling is likely independent of SORB-1. Indeed, we found that the recycling endosomal residence of RTKN-1 requires the presence of SDPN-1.
Utrophin-CH preferentially labels Arp2/3-dependent branched F-actin (Chia et al., 2014). Given the reduced endosomal labeling of the Utrophin-CH in sdpn-1 mutants (Fig. 10, G and G'), we speculated that the recycling endosome is mainly associated with branched F-actin. Indeed, SDPN-1/Syndapin has been shown to regulate recycling by promoting the formation of a branched F-actin network (Giridharan et al., 2013;Sherlekar et al., 2020;Gleason et al., 2016). Moreover, while demonstrating the functional context of RTKN-1-mediated actin integrity modulation, our data further suggested that SDPN-1 could act as an endosomal actin organization coordinator, which orchestrates assembly and stabilization of the endosome-associated actin and thus promotes the formation and directional movement of recycling membrane carriers. Thus, in addition to previously described endosomal actin dynamics (Simonetti and Cullen, 2019, our study provides new insight into actin organization regulation during endocytic recycling.

General methods and strains
The C. elegans strains were cultured at 20°C on nematode growth medium plates with Escherichia coli strain OP50 as the food source. RNAi-mediated interference was performed by feeding the C. elegans strains with E. coli strain HT115 carrying the L4440 vector expressing double-stranded RNA (Timmons and Fire, 1998). If not specifically indicated in the RNAi experiments, animals in the adult stage were cultured, and the next generation was scored at the day 1 adult stage.

Two-step genome-wide RNAi genetic screen
The rab-10 mutant is superficially normal in growth and development (Shi et al., 2010;Wang et al., 2016). However, rab-10 mutants accumulate recycling cargo hTAC-GFP in intestinal cells (Chen et al., 2006). To identify recycling regulators, a genome-wide RNAi screen (19,763 genes) was performed using rab-10(RNAi) as the positive control. After the primary screen, the knockdown of 54 candidates was found to lead to intracellular hTAC-GFP aggregation, including RTKN-1. A small-scale screen for actin dynamics regulators was then conducted within these candidates, and loss of RTKN-1 exhibited an adverse effect on GFP-Utrophin-CH-labeled actin structures.

Plasmids and transgenic strains
To construct transgenes expressed in the intestine, vha-6 promoter-driven destination vectors with a Gateway cassette inserted upstream or downstream of the GFP or mCherry coding intestinal cells. Box-and-whisker plots (n = 18 cells): 10-90th percentile; dots, outliers; red midline, median of WT; boundaries, quartiles. ***, P < 0.001 by oneway ANOVA followed by Dunn's post hoc multiple comparison test. The black asterisks in the image panels indicate intestinal lumen. Scale bars, 10 µm. MC, mCherry; P, pellet; S, supernatant.
Yan et al.

Phalloidin labeling
Animal dissection was performed as previously described (Grant et al., 2001). Dissected samples were fixed in 4% paraformaldehyde (Beyotime) in PBS for 30 min at room temperature, and three times for 10 min with PBS followed by 10 min of treatment with PBS containing 0.5% Triton X-100. Alexa Fluor 488 phalloidin (final 330 nM in PBS; Thermo Fisher Scientific) was used for whole-mount staining for 1 h at room temperature. The specimens were washed three times for 10 min in PBS to remove unbound phalloidin and then subjected to confocal microscopy.
Yan et al. Journal of Cell Biology examine fluorescence. The "Integrated Morphometry Analysis" function was used to measure the fluorescence area (total area) and the fluorescence intensity (total intensity) within unit regions. A total of 18 cells from nine animals of each genotype were collected for analysis. For colocalization assay, images were captured on young adults; DAPI channel was used to indicate broad-spectrum autofluorescence. From a total of 12 animals of each genotype, Mander's coefficients and Pearson's correlation coefficients were acquired using a plugin, JACoP, of the opensource Fiji (ImageJ) software (Schindelin et al., 2012). For tracking the dynamics of endosomal carriers and actin structures, animals were loaded to a Stellaris 5 confocal microscope (Leica) equipped with a 63× (NA 1.4) oil objective and white light lasers. Dynamic images were captured using LAS X software (Leica). Synchronized day 1 adults were selected for imaging (GFP and mCherry channels).

GST pulldown assay and protein purification
The Gateway BP and LR recombination system were employed to construct expression vectors. The PCR products of rtkn-1a, rtkn-1a(1-231aa), rtkn-1a(232-515aa), rtkn-1a(346-515aa), rtkn-1a(386-515aa), rtkn-1a(232-385aa), rtkn-1a(1-345aa), unc-60a, gsnl-1, cor-1 and pod-1 coding sequences were introduced in frame into entry vector pDONR221 by BP reaction. pGEX-2T destination vector (GE Healthcare Life Sciences) with a Gateway cassette was used for prokaryotic expression. pcDNA3.1 destination vector (Invitrogen) with a 2xHA tag and a Gateway cassette was used for in vitro transcription and translation. LR reactions were conducted to transfer the coding sequences into the above destination vectors. HA-tagged proteins were synthesized in vitro using pcDNA3.1 vectors and a TNT Coupled Transcription/Translation System (Promega). The pGEX-2T vectors were transformed into the ArcticExpress strain of E. coli (Stratagene) to express GST fusion proteins. The bacteria were pelleted and lysed in lysis buffer (50 mM Hepes [pH 7.5], 400 mM NaCl, 1 mM DTT, 1 mM PMSF) supplemented with Complete Protease Inhibitor Mixture Tablets (Sigma). Extracts were cleared by centrifugation (10,000 rpm for 60 min), and supernatants were incubated with 300-3,000 µl glutathione sepharose 4B beads (GE Healthcare Life Sciences) at 4°C overnight. Beads were washed five times with cold STET buffer (10 mM Tris-HCl [pH 8.0], 150 mM NaCl, 1 mM EDTA, 1mM DTT, 0.2% [vol/vol] NP-40) and three additional times with 50 mM Tris-HCl (pH 8.0). GST-tagged proteins were eluted by using 50 mM Tris-HCl (pH 8.0) containing 10 mM glutathione. After centrifugation (13,000 rpm for 30 min at 4°C), eluted proteins were further purified with the AKTA system equipped with Superdex 200 or Superose 6 size exclusion chromatography (GE Healthcare Life Sciences). For GST pulldown assays, 20µl beads were incubated with purified proteins and washed five times with cold STET buffer, then incubated with HA-tagged proteins overnight at 4°C. Beads were washed five times in cold STET buffer to remove unbound proteins; 20 µl 2xSDS-PAGE loading buffer was then added to each sample. After boiling, eluted proteins were resolved on SDS-PAGE (10% [wt/ vol] polyacrylamide) and blotted onto nitrocellulose. The blot was probed with anti-HA or anti-GST antibody.
F-actin cosedimentation assay F-actin binding potential was assayed using an Actin-Binding Protein Biochem Kit: Non-Muscle Actin (BK013; Cytoskeleton). G-actin was prepared by dilution with general actin buffer to 1 mg/ml. Actin polymerization buffer was added into G-actin preparations. F-actin was obtained by incubation of the reaction cocktail for 1 h at room temperature. The purified GSTtagged proteins were centrifuged at 150,000 xg for 1 h at 4°C, and the supernatant was used for F-actin cosedimentation. GSTtagged proteins were incubated with an equal volume of general actin buffer or F-actin for 30 min and then centrifuged at 150,000 xg for 1.5 h at room temperature. For actin-bundling detection, F-actin was incubated with purified proteins for 30 min and centrifuged at 15,000 xg for 1 h at room temperature. The pellet/supernatant ratio was quantified by densitometry using FluorChem FC3 version 3.4.0 (ProteinSimple).

F-actin polymerization and depolymerization assays
Actin Polymerization Biochem Kit: Muscle Actin (BK003; Cytoskeleton) was used for both polymerization and depolymerization assays. The fluorescence intensity of the pyrene-conjugated actin was measured to monitor actin polymerization or depolymerization dynamics. In actin polymerization assay, 10 µM pyrene-G-actin was prepared by the standard protocol. The purified GST-tagged proteins were added into pyrene-G-actin preparation. The fluorescence intensity was recorded every minute over 20 min using a Synergy 2 modular multimode reader (BioTek Instruments) at an excitation wavelength of 350 nm and an emission wavelength of 407 nm. The purpose of this step is to detect whether the protein itself has a polymerization activity. Then, actin polymerization buffer was added to the reaction, and the fluorescence intensity was recorded for an additional 100 min. For actin depolymerization assay, 5 µM pyrene-G-actin was used for the preparation of F-actin. F-actin was diluted to 1 µM, and the fluorescence intensity was recorded every minute over 3 min to obtain the baseline. The purified GST-tagged proteins were then added into pyrene-F-actin preparation, and the fluorescence signal was recorded for an additional 60 min.

Crosslinking protein interaction analysis and native PAGE
The cleavage site of PreScission protease (LEVLFQGP) followed by the HA coding sequence was inserted behind the GST in the pGEX-2T vectors via overlapping PCR. GST fusion proteins with the PreScission protease cleavage site were expressed in the ArcticExpress strain of E. coli (Stratagene). Bacterial pellets were lysed, and supernatants were collected and incubated with glutathione sepharose 4B beads as described previously. After five 10-min washes with cold STET buffer, the beads were further washed three times with PreScission cleavage buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM DTT [pH 7.5]). A total of 50 µg GST-PreScission Protease was added into 300-µl beads per sample. The cleavage reaction was performed at 4°C overnight. The supernatant was acquired by centrifugation (13,000 rpm for 30 min at 4°C), and the protein concentration was assessed. After centrifugation (13,000 rpm for 30 min at 4°C), eluted proteins were further purified with the AKTA system equipped with Superdex 200 (for HA-RBD-PH) or Superose 6 (for HA-RTKN-1) size exclusion chromatography (GE Healthcare Life Sciences). For crosslinking analysis, proteins were covalently crosslinked for 30 min at 25°C in PBS (disuccinimidyl glutarate, #20593; Thermo Fisher Scientific) and incubated with 20 mM Tris for 15 min to terminate the reaction. Proteins were separated on SDS-PAGE using a 3%, 5%, and 8% (wt/vol) gradient resolving gel and stained with Coomassie blue. For native gel analysis, proteins (20 µg) were mixed with 5× Native Gel Sample Loading Buffer (Beyotime). Samples were separated on native PAGE (3-12% Bis-Tris Protein Gel; Thermo Fisher Scientific) and stained with Coomassie blue.

Statistical analysis
Statistical analyses in this study were performed and plotted using GraphPad Prism 8.02 software (GraphPad Software). Biochemical datasets were analyzed by t test, and Mann-Whitney U test was used for most imaging data sets. For comparison of more than two data sets, one-way ANOVA followed by a Dunn's post hoc multiple comparison test was applied.