Proteotoxic stress disrupts epithelial integrity by inducing MTOR sequestration and autophagy overactivation

ABSTRACT Macroautophagy/autophagy, an evolutionarily conserved degradation system, serves to clear intracellular components through the lysosomal pathway. Mounting evidence has revealed cytoprotective roles of autophagy; however, the intracellular causes of overactivated autophagy, which has cytotoxic effects, remain elusive. Here we show that sustained proteotoxic stress induced by loss of the RING and Kelch repeat-containing protein C53A5.6/RIKE-1 induces sequestration of LET-363/MTOR complex and overactivation of autophagy, and consequently impairs epithelial integrity in C. elegans. In C53A5.6/RIKE-1-deficient animals, blocking autophagosome formation effectively prevents excessive endosomal degradation, mitigates mislocalization of intestinal membrane components and restores intestinal lumen morphology. However, autophagy inhibition does not affect LET-363/MTOR aggregation in animals with compromised C53A5.6/RIKE-1 function. Improving proteostasis capacity by reducing DAF-2 insulin/IGF1 signaling markedly relieves the aggregation of LET-363/MTOR and alleviates autophagy overactivation, which in turn reverses derailed endosomal trafficking and rescues epithelial morphogenesis defects in C53A5.6/RIKE-1-deficient animals. Hence, our studies reveal that C53A5.6/RIKE-1-mediated proteostasis is critical for maintaining the basal level of autophagy and epithelial integrity. Abbreviations: ACT-5: actin 5; ACTB: actin beta; ALs: autolysosomes; APs: autophagosomes; AJM-1: apical junction molecule; ATG: autophagy related; C. elegans: Caenorhabditis elegans; CPL-1: cathepsin L family; DAF: abnormal dauer formation; DLG-1: Drosophila discs large homolog; ERM-1: ezrin/radixin/moesin; EPG: ectopic P granule; GFP: freen fluorescent protein; HLH-30: helix loop helix; HSP: heat shock protein; LAAT-1: lysosome associated amino acid transporter; LET: lethal; LGG-1: LC3, GABARAP and GATE-16 family; LMP-1: LAMP (lysosome-associated membrane protein) homolog; MTOR: mechanistic target of rapamycin kinase; NUC-1: abnormal nuclease; PEPT-1/OPT-2: Peptide transporter family; PGP-1: P-glycoprotein related; RAB: RAB family; RIKE-1: RING and Kelch repeat-containing protein; SLCF-1: solute carrier family; SQST-1: sequestosome related; SPTL-1: serine palmitoyl transferase family.


Introduction
Polarized membrane transport, a process that sorts and delivers transmembrane proteins and lipids to either apical or basolateral membrane domains, is fundamental for the maintenance of epithelial polarity and integrity. As a sorting platform in this process, the endosomal system acts in both exocytic trafficking, by directing polarized proteins emanating from trans-Golgi network (TGN) to their destined membrane domains, and in endocytic trafficking, by recycling endocytosed cargos from the plasma membrane for redistribution or degradation [1][2][3]. Depending on the sorting signals they carry, some cargo proteins may require specific adaptors for their trafficking. As an example, some glycoproteins are recognized, sorted, and retained at the plasma membrane by galectins through the formation of galectin-glycan lattices [4,5].
The actin cytoskeleton plays an important role in polarized membrane transport. Canonically, actin filaments can function as tracks for the myosin-driven movement of vesicles [6,7], facilitating directional sorting. Alternatively, actin can also regulate membrane budding and fission through dynamin and/or CDC-42 and its downstream effectors, contributing to the biogenesis of transport carriers [8]. Finally, actin can also function to generate a local mechanical force, facilitating the movement of transport carriers [9].
Autophagy is a unique membrane trafficking process which is believed to contribute to epithelial homeostasis by converging with the endosomal system in multiple processes, including formation and expansion of the phagophore, biogenesis of autophagosomes (APs) and subsequent fusion of the APs with lysosomes (autolysosomes, ALs) for degradation [10][11][12].
Conceivably, endosome-mediated membrane trafficking will be affected by autophagy-dependent lysosomal degradation, and vice versa. Autophagy occurs at a basal level in most cells to maintain cellular homeostasis and fitness [13][14][15][16]. Moreover, autophagy can be induced by many types of stress, at least in part, through the inhibition of MTOR (mechanistic target of rapamycin kinase) [17]. The autophagic response needs to be tightly regulated because its level is often linked with beneficial or detrimental outcomes [18]. A persistent and overactivated autophagic response promotes organelle damage and cell death [19][20][21][22]. These effects can be partially restored by autophagy inhibition [18,21,[23][24][25]. The stimuli that trigger autophagy overactivation and its cytotoxic effect in vivo remain largely unknown [26]. A mechanistic understanding of autophagy overactivation and its impact is critical to precisely control the level of autophagy for therapeutic intervention in many diseases [27][28][29].
Our exploration of the mechanisms underlying autophagy overactivation began with our interest in C. elegans intestinal development. Given that the actin cytoskeleton has diverse functions in polarized membrane trafficking and that multiple Kelch repeat family members have been shown to regulate cytoskeletal organization [30], we specifically investigated whether Kelch repeat-containing proteins modulate intestinal morphogenesis. Here we show that animals lacking the Kelch repeat-containing protein RIKE-1, encoded by a previously uncharacterized gene, C53A5. 6, exhibit severe defects in intestinal morphology and integrity accompanied with mislocalization of intestinal membrane components and excessive endosomal degradation. We identify that autophagy overactivation is the underlying cause of excessive endosomal degradation. Blocking autophagosome formation prevents endosomal degradation and partially rescues intestinal morphology defects in animals with compromised RIKE-1 function. Further study reveals that autophagy overactivation is caused by imbalanced proteostasis, which consequently leads to cytoplasmic aggregation of proteins, including LET-363/ MTOR. Attenuating the level of DAF-2 insulin/IGF1 signaling alleviates the LET-363/MTOR aggregation and autophagy overactivation, which in turn restores epithelial morphology in RIKE-1-deficient animals. Together, our findings suggest a role of RIKE-1 in maintaining the basal level of autophagy and epithelial integrity through shaping the cellular proteome.

Loss of RIKE-1 results in defects in intestinal morphogenesis and integrity
To investigate possible roles of Kelch repeat-containing proteins in modulating intestinal morphogenesis, we carried out RNAi knockdown of 11 C. elegans proteins containing Kelch repeats (based on the SMART database) using worms expressing the lumenal membrane marker ACT-5::GFP. We identified that the previously uncharacterized gene C53A5.6 shows fully penetrant L1 larval lethality with severe intestinal morphogenesis defects upon standard RNAi treatment [31] ( Figure 1 and Table S1). The protein encoded by C53A5.6 is distinct from other Kelch repeat proteins owing to the presence of an N-terminal RING domain; therefore, we named this gene rike-1 (RING-and Kelch-containing protein). Given the severe phenotypes caused by standard rike-1 RNAi knockdown where RNAi is initiated in L4-stage worms and the progeny are evaluated for phenotypes, most of the RNAi experiments were performed using a conditional knockdown approach where RNAi is initiated from L1s and the same generation animals are scored. Detailed analysis revealed that loss of RIKE-1 upon standard or mild RNAi caused cytoplasmic mislocalization and aggregation of multiple apical membrane-associated proteins, including the cortical actin ACT-5, the membrane-cytoskeleton linker ERM-1, the intermediate filament protein IFB-2, and the integral membrane protein PGP-1 ( Figure 1B-D). The apical junction component DLG-1 was disorganized ( Figure 1E), whereas AJM-1 entirely disappeared from the apicolateral junctures and accumulated in the cytoplasm ( Figure 1F) in intestinal cells of rike-1(RNAi) animals. Moreover, the basolateral protein LET-413/SCRIB formed cytoplasmic aggregates with partial displacement to the subapical domain ( Figure 1G). Of note, the distribution of the apical peptide transporter PEPT-1/OPT-2 and the basolateral pyruvate transporter SLCF-1 was less affected ( Figure  1H,I). Transmission electron microscopy (TEM) analysis revealed striking lumenal structural defects. These included loss of microvilli and disruption of the terminal web, a cytoskeletal structure on the apical surface of intestinal cells ( Figure  1L and S1A). These results suggest that loss of rike-1 causes intestinal atrophy and defects in assembly and maintenance of intestinal membrane components at designated positions ( Figure 1M).

Loss of RIKE-1 promotes endosomal enlargement and fusion with lysosomes
In intestinal epithelial cells, apicobasal polarity and integrity are maintained through constant, polarized trafficking of both protein and lipids [32,33]. Loss of microvilli and mislocalization of membrane components is often associated with disordered endosomal trafficking [34,35]. Indeed, we observed that loss of the apical recycling endosomal component RAB-11 induced the mislocalization of ACT-5::GFP ( Figure S2A). We therefore examined the effect of rike-1 knockdown on endosomal markers. Intriguingly, all types of endosomes examined, including RAB-5-labeled early endosomes (EEs), RAB-7-marked late endosomes (LEs), and RAB-10-, RME-1and RAB-11-positive recycling endosomes (REs), were dramatically enlarged and clustered in intestinal cytoplasm at an early stage (48 h; see Materials and Methods for staging) after rike-1 RNAi exposure (Figure 2A-I and S2B-D). Unexpectedly, we found a strong reduction in fluorescence of GFP-labeled endosomes at a later stage (72 h) after rike-1 RNAi (Figure 2A-G). Compared with mCherry signals, GFP fluorescence is easily quenched in the acidic lysosomal environment [36,37], so this result suggests that the abnormally enlarged endosomes were likely fused with lysosomes. As expected, we observed strong colocalization of GFP-positive endosomal clusters with the lysosomal marker LMP-1:: mCherry in early stage (48 h) rike-1(RNAi) animals ( Figure  2J and S2E). Moreover, the enlarged RAB-7 LEs and RAB-10 REs were enclosed within LAAT-1::GFP-labeled lysosomes in rike-1-deficient worms at the later stage (72 h) ( Figure 2K and S2F). Of note, LAAT-1::GFP was not easily quenched like the endosomal GFP signals, because LAAT-1 is localized to lysosomal membranes [38]. These data indicate that endosomelysosome fusion upon RIKE-1 loss facilitates endosomal degradation. Several lines of evidence further support this notion. First, we detected a significant decrease of protein levels of RAB-5, RAB-7 and RAB-11 in late-stage rike-1 (RNAi) animals (Figure 2A-I). Second, we observed dramatic degradation of the GFP-tagged recycling cargo IL2RA/hTAC in rike-1-deficient worms ( Figure S2G and 2H). Third, in rike-1(RNAi) animals, both the number of vesicular lysosomes and the total volume of lysosomes tagged by NUC-1::Cherry increased significantly ( Figure 2L,M). Finally, we examined the processing of CPL-1 (CathePsin L family) from the inactive propeptide to the active mature form, which is an indicator for the degradation activity of lysosomes [39]. We found that more mature CPL-1 was produced in rike-1-deficient worms ( Figure 2N,O). Together, these results suggest that RIKE-1 loss promotes enlargement of trafficking endosomes and likely forces them to undergo lysosomal degradation.
Elevated autophagy activity could result from accelerated autophagosome biogenesis and/or blockade of the downstream autophagosome-lysosome fusion and degradation events. To distinguish between the two possibilities, we examined the fusion of autophagosomes with late endosomes and lysosomes. Strong colocalization of RAB-7, RAB-10 and LMP-1 with enlarged GFP::LGG-1 puncta was observed ( Figure 3G, S3E and S3F), suggesting that autophagosome-lysosomes fusion is not compromised. To further corroborate these findings, we used a tandem-tagged mCherry::GFP::LGG-1 reporter that labels APs yellow (positive for both GFP and mCherry) and acidic ALs red (positive for mCherry only, due to the GFP signal being quenched in an acidic environment) [37]. At the early stage of rike-1 RNAi (48 h), the number and intensity of APs (yellow [green/red] puncta) and ALs (red puncta) were significantly elevated, and tubular structures extruding from ALs were also observed ( Figure 3H-J). Later (72 h), only ALs (red puncta) were detected ( Figure 3H-J), indicating a normal degradation process. Moreover, we tested the effect of RIKE-1 loss on degradation of autophagy substrates [40]. The intestinal level of W07 G4.5::mCherry was significantly reduced in rike-1(RNAi) animals ( Figure 3K, L). Another target of autophagy, the SQSTM1/p62 homolog SQST-1, accumulates in rpl-43 mutants owing to impaired protein synthesis and can be removed by autophagy activation [41]. The number of SQST-1::GFP puncta was also significantly reduced at 72 h after rike-1 RNAi ( Figure  S3G and S3H). It is worth noting that there is no contradiction between the observed degradation of GFP::LGG-1 and SQST-1::GFP puncta at 72 h after conditional rike-1 RNAi initiated from the L1 stage and the observed upregulation of LGG-1 and SQST-1 in arrested L1s under standard RNAi. Rather, these results respectively reflect a mild and prolonged response versus a strong and acute response. Last, to measure autophagic flux, we treated the worms with the lysosomal inhibitor bafilomycin A 1 , and found that bafilomycin A 1 treatment led to a further increase in the number of GFP::LGG-1 puncta ( Figure 3M). Together, these findings argue strongly for an overactivated autophagic flux under rike-1 RNAi.
To rule out the possibility that autophagy overactivation is secondary to lethality, or defective epithelial morphogenesis and/or membrane trafficking, we knocked down several essential genes with known function in intestinal morphogenesis and/or trafficking, such as act-5 [42], sptl-1 [43] and rab-11 [44], to check whether they cause enlarged APs. As expected, knocking down each of these genes resulted in highly penetrant larval arrest phenotypes, but not in the accumulation of enlarged GFP::LGG-1 puncta ( Figure S3I). This provides further evidence that loss of RIKE-1 specifically induces autophagy upregulation.
Considering that excessive degradation of endosomes may underlie the intestinal morphology defects in rike-1-deficient worms, we next examined whether the disrupted intestinal morphogenesis is also relieved upon autophagy inhibition. A striking restoration of lumen width was observed in rike-1-deficient worms carrying the atg-18(gk378) mutation ( Figure S4G). The mislocalization of ACT-5::GFP in rike-1(RNAi) animals was partially alleviated by atg-18(gk378) mutation ( Figure 5D, E). Similar rescuing effects were also observed for ERM-1 and AJM-1 ( Figure 5F,G, S4H and S4I). However, the mislocalization of p-glycoprotein PGP-1 was not rescued in response to autophagy inhibition ( Figure 5H,I). The discrepancy in the rescuing effects on the mislocalization of PGP-1 and other apical domain components (such as ACT-5, ERM-1 and AJM-1) lies in the fact that PGP-1 is an integral membrane protein, and its sorting and transport may also depend on the interaction of its transmembrane domains with sorting receptors and adaptors or direct partitioning into lipid domains [48] in addition to endosomes. Support for this explanation comes from a previous study showing that PAR-5 depletion alters the apical localization of F-actin and recycling endosome RAB-11, but has no effect on the localization of the apical membrane proteins PGP-1 and PEPT-1 [44]. This suggests that RIKE-1 probably acts to maintain PGP-1 at the apical membrane through an autophagy/ endosome-independent mechanism. Thus, RIKE-1 loss leads to excessive degradation of endosomes due to hyperactivated autophagy, which, in turn, misroutes membrane components and disrupts epithelial integrity.

RIKE-1 loss results in proteostasis imbalance and MTOR aggregation
Activation of autophagy and enlargement of lysosomes are hallmarks of proteotoxic stress. To explore the origin of the stress, we examined the expression of several heat shock protein (HSP) genes. The mRNA level of the cytosolic unfolded protein response gene hsp-70 was increased about 10-fold in rike-1 RNAi animals compared to empty vector control animals, whereas the mRNA levels of an ER chaperone (hsp-3 and hsp-4) and mitochondrial chaperones (hsp-6 and hsp-60) were not increased ( Figure 6A). Taken together, these results indicated that rike-1 knockdown appears to disrupt proteostasis in the cytoplasm, consequently stimulating autophagy. Imbalanced proteostasis is often accompanied with widespread protein aggregation. To determine whether this is also the case in rike-1 knockdown animals, we extracted insoluble proteins at the early stage of rike-1 RNAi (48 h), a time point when autophagy has been induced. Intriguingly, we found that, compared to control animals, the insoluble protein fractions in rike-1 RNAi animals were significantly increased while total protein levels remained constant ( Figure 6B). Given that the MTOR complex is the master regulator of autophagy [12] and that LET-363/MTOR deficiency in C. elegans causes larval arrest and intestinal atrophy similar to rike-1-deficient worms [49], we specifically examined whether the LET-363/MTOR formed protein aggregates. Indeed, in rike-1-deficient worms, the LET-363 level was significantly higher in the insoluble fraction, yet was almost unaltered in total protein ( Figure 6C,D). LET-363 aggregation upon rike-1 knockdown was also observed by immunofluorescence analysis ( Figure 6E). Furthermore, LET-363 aggregates were enclosed in GFP::LGG-1 autophagosomes, implying an eventual degradation of LET-363 aggregates ( Figure 6F). Next, we assessed the level of the LET-363/ MTOR substrate p-RSKS-1/RPS6KB1/p70 S6K, and found that rike-1 RNAi resulted in a decreased p-RSKS-1 level ( Figure 6G,H), in line with decreased LET-363/MTOR activity derived from MTOR sequestration. Additionally, we investigated the localization of another direct target of MTOR, HLH-30/TFEB, which can translocate to the nucleus to activate autophagy-related genes in response to inhibition of MTOR signaling [50]. As expected, rike-1 RNAi resulted in significant nuclear translocation of HLH-30::GFP ( Figure 6I,J). To . Arrows indicate nuclear HLH-30::GFP signals. Scale bars: 10 μm. (K and L) atg-18(gk378) mutation does not suppress FLAG::LET-363 aggregation induced by rike-1 RNAi, as evidenced by western blotting and protein level quantification (n = 3 independent biological replicates). All statistical analyses were performed using two-tailed unpaired t-tests. Error bars indicate mean ± SEM. *P < 0.05, ***P < 0.001, ****P < 0.0001. ns, not significant. determine whether LET-363 aggregation preceded hyperactivated autophagy in rike-1-deficient worms or vice versa, we crossed FLAG-LET-363 into atg-18(gk378) mutants. We found that FLAG-LET-363 aggregation was not suppressed in atg-18(gk378);rike-1(RNAi) animals ( Figure 6K,L), which indicates that autophagy upregulation is one of the downstream consequences of proteotoxic stress. These findings suggest that RIKE-1 loss inhibits the MTOR signaling pathway to activate autophagy, at least in part, through inducing MTOR complex aggregation.

Discussion
We have demonstrated that the novel protein RIKE-1 contributes to epithelial membrane morphogenesis, at least in part, by controlling protein homeostasis. Imbalanced protein homeostasis triggers aggregation of the LET-363/MTOR complex as well as many other proteins, which subsequently overactivates autophagy, degrades endosomes and derails membrane trafficking (Figure 8). Emerging evidence suggests that protein aggregation is often part of the cellular defense against imbalanced protein homeostasis [55][56][57][58]. Similarly, autophagy helps the cell adapt to and cope with stress by eliminating misfolded proteins and damaged organelles [59][60][61][62]. Protein aggregation and autophagy upregulation are both supposed to be reversible. Paradoxically, in response to RIKE-1 loss-induced proteotoxic stress, it seems that the insoluble proteins are terminally deposited as aggregates and autophagy is constitutively induced (until the worms die). A consequence is the sequestration of MTOR into aggregates, leading to MTOR inhibition followed by autophagy upregulation. Alleviating proteotoxicity by reducing DAF-2 insulin/IGF-1 signaling prevents protein aggregation and inhibits autophagy overactivation, but not vice versa ( Figure 7A-D and S5A-D). This demonstrates that hyperactivated autophagy indeed occurs downstream of proteotoxic stress. Excessive autophagy depletes endosomes, probably by using them for autophagosome biogenesis. Exploitation of the endosomal system compromises its roles in membrane trafficking and disrupts plasma membrane composition. The aggravated cellular stress caused by the damaged membrane may serve as an alternative explanation for the irreversible response. Taking these lines of evidence together, we envision that RIKE-1 is a major factor that shapes cellular proteostasis.
RIKE-1 contains a RING domain and Kelch repeats, and is homologous to the mammalian actin-binding protein IPP based on the similarity of its Kelch domain. Members of the Kelch repeat protein family have a known role in regulating cytoskeletal organization [30]. However, disruption of the actin cytoskeleton cannot lead to cellular stress that is comparable with the stress that occurs under rike-1 deficiency. This suggests that the interaction of RIKE-1 with actin, if any, is only part of the story. RIKE-1 also carries a Cys3HisCys4-type Zinc finger domain at the N-terminus and this group of RING domain proteins includes many E3 ubiquitin ligases [63]. In fact, E3 ligases are implicated in both proteome balance and epithelial morphogenesis [64]. Further work is required to elucidate the detailed molecular mechanism of how RIKE-1 controls protein homeostasis by preventing the accumulation of surplus and aberrant protein species.

RNAi knockdown experiments
RNA interference (RNAi) experiments were performed by the feeding method as previously described [43]. For standard RNAi, about 10 L4 larvae were picked and cultured on RNAi plates (NGM containing 2 mM IPTG [Sigma-Aldrich, I6758] and 50 μg/ml carbenicillin [Sigma-Aldrich, C3416]) seeded with bacterial clones of target genes and F1 progeny larvae were examined after 72 h. Standard rike-1 RNAi leads to fully penetrant L1 lethality. We therefore also performed mild RNAi, starting from the L1 stage, for better evaluation of changes in physiological processes of rike-1-deficient larvae. Briefly, eggs from over 50 bleached adults were allowed to hatch on rike-1 RNAi plates. After 48 h, most of the worms developed into L3-L4 larvae and were scored for early phenotypes. At 72 h, most of the larvae were arrested at the L3-L4 stages and scored for late phenotypes. Even milder RNAi conditions can be achieved by mixing the rike-1 RNAi clone with the RNAi bacteria containing the L4440 vector plasmid (Addgene, 1654; Andrew Fire) at different proportions when necessary. Worms were fed with RNAi bacteria containing the L4440 empty vector plasmid as a control treatment (EV RNAi). All RNAi clones were confirmed by sequencing.

Gene editing by CRISPR-Cas9
The mutant rike-1(g55t;c72a) carries two premature stop codons in close proximity (E19X,C24X). Design of CRISPR-Cas9 plasmids to create this mutant was performed as previously described [66]. The 20 bp gRNAs for creating DNA double-strand breaks in the rike-1 genomic locus were cloned into pPD162 Peft-3::Cas9;PU6::empty sgRNA vector (Addgene, 47549; Bob Goldstein). An exogenous donor oligonucleotide for homologous repair was designed and synthesized, including ~50 bp of flanking homology on either side and additional synonymous mutations to ablate the gRNA cleavage site and introduce an FspI restriction site for convenient screening. Fifty ng/μl gRNAs and 500 nM donor oligonucleotides were used for injection. A plasmid containing gRNA and repair oligonucleotide for the dpy-10(cn64) mutation, which confers a dominant rolling phenotype, was co-injected at a concentration of 50 ng/μl as a screening marker [40]. Roller animals screened from F1 progeny were singled and, after laying eggs, were picked out for singleworm PCR and restriction enzyme digestion for validation of the rike-1(g55t;c72a) mutation. Non-roller F2 heterozygotes were singled and screened for homozygosity. After confirming that rike-1(g55t;c72a) mutation causes lethality, rike-1(g55t; c72a) mutants and the rike-1(g55t;c72a) derivatives were maintained as heterozygotes. To test heterozygotes, 10-20 animals were singled out for each generation, and the plates containing dead worms were picked for phenotypic scoring and genotyping. For genotyping, the arrested animals were used as templates for single-worm PCR followed by sequencing.
Similarly, to generate the knockout mutant rike-1(syb1165), two pairs of gRNAs were inserted into Cas9-sgRNA plasmids and co-injected with pRF4 rol-6(su1006) [67] as an injection marker at a concentration of 50 ng/μl each. Roller F1 progeny were isolated, and their progeny were examined by PCR. The rike-1(syb1165) knockout mutation was generated and balanced by SunyBiotech (China). To generate the rike-1-(syb1165) mutants carrying GFP::LGG-1 and ACT-5::GFP, rike-1(syb1165) mutants were crossed with the reporter strains followed by crossing back with the nT1 balancer. The specificity of mutations was confirmed by sequencing. The sequences of gRNAs and repair oligonucleotides are provided in Table S3.

Quantitative real-time PCR (qRT-PCR)
Total RNA was extracted using TRIzol reagent (TaKaRa, 9108) following the standard protocol. RNA was quantified based on its OD at a wavelength of 260 nm and RNA purity was assessed by the absorbance ratios of 260:280 and 260:230. Reverse transcription was performed using a PrimeScript TM RT Reagent Kit (TaKaRa, RR047A). Quantitative RT-PCR was performed using Fast SYBR Green Master Mix (Thermo Fisher Scientific, 4385616) and the ABI7500 Fast Real-Time PCR Detection System. Relative mRNA expression levels were normalized to tba-1 and analyzed using the ΔΔC t method. Primers used for the RT-PCR experiments are listed in Table S3.

Extraction of insoluble proteins
Insoluble proteins were extracted from worm lysates as previously described [68] with some modifications. Briefly, worms were lysed with RIPA buffer supplemented with protease inhibitor. After mild sonication for 15 min at 4°C, worm lysates were centrifuged at 400 g for 5 min to remove debris. To serve as total protein,100 µl of the protein suspension was saved and mixed with 2X urea-SDS buffer (2% SDS, 50 mM dithiothreitol, 50 mM Tris, pH 8.0, 16 M urea [Sigma-Aldrich, U5378]) at RT. The rest was centrifuged at 21,000 g for 20 min at 4°C to remove detergent-soluble protein fractions. The pellet was resuspended in 75 μl urea-SDS buffer (8 M urea) at RT and used as the insoluble protein fraction. Both total and insoluble protein samples were kept at −80°C for later examination or resolved on a 10% SDS-PAGE gel followed by Coomassie Brilliant Blue (Bio-Rad, 1610436) staining.

Transmission electron microscopy (TEM) analysis
For TEM analysis, 10-30 living L3-L4 larvae from control and rike-1(RNAi) groups were picked into a 200-μm specimen carrier filled with E. coli for rapid freezing in a highpressure freeze (HPF) device (Leica HPM100, Germany). Following HPF, the fast-frozen samples were immersed into a cryovial containing 2% osmium tetroxide in 98% acetone/ 2% water for a freeze substitution (FS) procedure in an FS unit (Leica EM AFS, Germany) with the following parameter settings: T1 = −90°C for 72 h, S1 = 5°C/h, T2 = −60°C for 12 h, S2 = 5°C/h, T3 = −30°C for 10 h, followed by a slow warming procedure to 10°C (5°C/h). After completion of FS, samples were rinsed extensively with acetone, then stained with 0.5% uranyl acetate dissolved in 90% acetone/10% methanol for 2 h in the dark at RT. After another series of extensive rinses with acetone, the samples were successively infiltrated with a mixture of resin (Electron Microscopy Sciences, EMbed 812) and acetone at ratios of 1:2, 1:1 and 3:1 at RT for 5 h, overnight, and 12 h, respectively. To ensure complete infiltration, the worms underwent 3-4 additional changes with 100% resin over the next five days with a tissue rotator. In 100% resin, the worm plus E. coli can easily come out of the holder; therefore, the worms were repositioned for the convenience of cross sectioning using a stereoscope in an embedding mold with fresh resin containing 1.5% benzyldimethylamine (Sigma-Aldrich, 185582). The block was allowed to polymerize at 60°C for 48 h before being trimmed and sectioned on an ultramicrotome (UC6, Leica Biosystem, Germany) with diamond knives (Diatome, Switzerland). Serial sections of worm samples were automatically collected using the Auto CUTS device and were observed with a JEM-1400 (JEOL) operating at 80 kV.

Fluorescence microscopy
Live worms were paralyzed in 5 μl 10 mM sodium azide (Sigma-Aldrich, S2002) on glass slides and observed directly under a Zeiss LSM710 confocal microscope (Carl Zeiss MicroImaging) equipped with a Plan-Apochromat DIC 63x/ 1.4 oil objective. For comparison of fluorescence intensity in Figure 6(b,c), the paralyzed worms were mounted in order on 2% agarose pads for direct detection using a LSM710 microscope with a 10x/0.3 dry objective. Single plane images were taken at 0.2 μm or smaller intervals along the z axis and maximum intensity projection images were generated by integrating 6-10 sections of single plane images. Multi-channel images were acquired by sequential scanning of individual channels to eliminate bleed-through between different channels. To distinguish the signals of interest from intestinal autofluorescence, the DAPI channel (405 nm excitation) was used to identify autofluorescent granules [44]. The images were processed using Adobe Photoshop CS6.

Statistical analysis
GraphPad Prism 6 software was used to perform statistical tests and generate graphs. Statistical differences were determined by two-tailed unpaired Student's t-test and bar graphs are shown as mean ± standard error of the mean (SEM). A P value less than 0.05 was considered statistically significant. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, no significance. Statistical parameters including the definitions, exact values of n, and statistical significance are indicated in the figures and figure legends. All experiments were independently repeated at least twice with similar results. The density of immunoblot bands and fluorescence intensity of microscopy images were quantified using ImageJ software.