SURO-2/TMEM39 Facilitates Collagen Secretion through the Formation of Large COPII Vesicles

Fibrosis of various tissues is a typical disease caused by excessive production and secretion of extracellular matrix. We used Caenorhabditis elegans to investigate the formation of large transport vesicles to understand collagen secretion, a critical factor in brosis formation. The suro-2 mutant displays obvious defects in collagen secretion and cuticle structure including a rupture phenotype in early adults. Transmission electron microscopy exhibited that the cuticle thickness of the suro-2 mutant was severely reduced. SURO-2/TMEM39 has 8 transmembrane domains and localizes in the endoplasmic reticulum (ER) membrane. SURO-2 interacts directly with NPP-20/Sec13, a component of the coat protein II (COPII) complex responsible for ER-to-Golgi transport. SURO-2 and NPP-20 localized at the same large puncta, a large COPII vesicle enough to accommodate collagens. We report here that SURO-2/TMEM39 is highly conserved among animal species and is a specialized regulator of bulky collagen secretion rather than general transport in C. elegans. A large-scale yeast two-hybrid screen for protein networks and a genome-wide prediction of C. elegans genetic interactions proposed associations between D1007.5 (SURO–2) and nucleopore protein–20 (NPP–20) 41, 42 , the C. elegans orthologue of human protein Sec13. Sec13 is a component of the COPII complex originally discovered in S. cerevisiae and participates in the early collagen secretory pathway 43, 44 . Another role of Sec13 relates to its function as a nucleopore protein during nuclear envelope reassembly 45 and is associated with the GATOR2 complex that regulates the mTORC1 complex 46 . Amidst varied functions, we focused on its potential role in the collagen secretory pathway to understand the relationship between SURO–2 and NPP–20, because SURO–2 is localized at the ER and participates in cuticle formation. We examined the knockdown phenotype of npp–20 via RNAi to compare its phenotype with that of the suro–2 mutant. Like suro–2 mutants, npp–20 RNAi suppressed the Rol phenotype of jgIs4 (Fig. 3a), implying a close relationship between suro–2 and npp–20. Therefore, we performed RNAi-mediated knockdowns of other COPII components to examine their functions in cuticle formation. Knockdown of most COPII genes resulted in Rol suppression. Knockdown of sar–1 and tfg–1 also resulted in Rol suppression (Fig. 3b). SAR–1 is the orthologue of human Sar1 GTPase which is the key regulator of COPII vesicle formation 19 . TFG–1 associates with SEC–16 and also functions in early collagen secretion 47 . Collectively, these results suggest that SURO–2 and COPII may function in the same secretory pathway. We compared COL–19::GFP to identify in

copies of COPII genes except for Sec13 12 . Collagen transport is known to rely on several gene products including Sec23A 13, 14 , Sec24D 15, 16 , and Sec13 17 . Collagen secretion is also affected by the depletion of Sar1b, a small GTP-binding protein that initiates COPII coat assembly in the ER membrane 12,18,19 .
The Sar1 cycle regulator Sedlin 20 , and the Sec16-interacting protein TFG1 also contribute to collagen secretion 21 .
Several studies have demonstrated potential mechanisms for the formation of large transport vesicles, including collagen packaging and COPII vesicle size increases. The single transmembrane domain protein TANGO1 plays an important role in procollagen VII transport and formation of large vesicles 22 . TANGO1 and cTAGE delay COPII vesicle formation until large enough to accommodate collagen packaging through direct binding of procollagen VII at the ER lumen and interaction with the Sec23-Sec24 heterodimer in the cytoplasm 22, 23, 24, 25 . Sec31 ubiquitination by the E3 ligase Cullin 3 and its speci c adapter protein KLHL12 contributes to an increase in COPII coat size 26 . Recruitment and interaction of these proteins with COPII-related vesicles may be su cient for the export of bulky collagen from ER 11 . A recent study reported that TMEM131, ranging from nematodes to humans, commonly plays an important role in regulating collagen secretion 27 . However, the understanding of the underlying mechanism of collagen secretion through evolutionarily large vesicles is still limited. For instance, simple animals like Caenorhabditis elegans do not have the TANGO1 homologue.
There are two main types of collagen in C. elegans: cuticle and basement collagen. Basement collagen contains two types of IV and one XVIII collagen, but cuticle collagen contains more than 170 cuticle collagen genes that support the rapid molting process associated with the synthesis and secretion of new cuticles at each stage of development 28,29 . C. elegans cuticle mutants exhibit a broad range of phenotypes such as the lethal (Let), abnormal embryogenesis (Emb), dumpy (Dpy), long (Lon), blister (Bli), squat (Sqt) and roller (Rol) phenotypes 28,29 . Genetic analysis of these cuticle mutants has provided valuable information regarding collagen function and ECM organization. We performed a suppressor screen using a transgenic strain of C. elegans with the Rol phenotype to identify genes involved in cuticle formation, processing, and secretion. In this study, we characterized suro-2 (suppressor-of-rolling-2) mutants in C. elegans and determined a cooperative role of SURO-2/TMEM39 and COPII in the formation of large ER vesicles essential for collagen secretion and cuticle formation.

Results
The suro-2 mutant has serious defects in collagen secretion and cuticle structure The suro-2 mutant was isolated from a suppressor screen of C. elegans with the Rol phenotype. The suro-2 mutation completely inhibits the Rol phenotype of the jgIs4 expressing mutant ROL-6collagen derived from the rol-6 (su1006) mutant 30, 31, 32 . The suro-2 mutant is a small and sick Dpy, and suro-2 jgIs4 exhibits a severe Dpy phenotype compared to a single suro-2 mutant (Fig. 1a).
Another interesting suro-2 mutant phenotype is young adult lethality. Wild-type adults begin to die after laying eggs, usually 6-7 days of adult life, but retain their body shape. However, some suro-2 mutants were ruptured (Fig. 1b) and approximately 18% of suro-2 mutants died between 1-3 days (Fig. 1c). Since this burst phenotype may be related to cuticle abnormalities or disorders, we used electron microscopy to observe the cuticle structure in the suro-2 mutant. The cuticle thickness of the adult suro-2 mutant was reduced to nearly one third of the wild type (Fig. 1d). The average cuticle thickness was 604 nm in the wild type and decreased to 233 nm in the suro-2 mutant. These results suggest that SURO-2 is required for proper collagen secretion and normal cuticle formation in C. elegans.
We further examined the effect of SURO-2 on the cuticle structure using two collagen markers, ROL-6::GFP 33 and COL-19::GFP 34 . These fusion proteins exhibit a striped pattern in the wild-type cuticle. Collagen-GFP expression was signi cantly reduced in the suro-2 mutant and the wild-type stripe pattern disappeared from the cuticle (Fig. 1e). Therefore, SURO-2 is required for the secretion of collagen, and a rupture phenotype of the suro-2 mutant may occur due to insu cient collagen supply.

SURO-2 is a homologue of mammalian transmembrane protein 39
We isolated two suro-2 alleles, jg35 and jg92. Sequencing of the D1007.5 using suro-2 (jg35) genomic DNA revealed a point mutation that resulted in the substitution of amino acid 187 with a stop codon (W187*). The wild-type suro-2 gene putatively encodes 477 amino acids (Supplementary g. 1), while the suro-2 (jg35) mutant encodes a shorter protein composed of only 186 amino acids whose suro-2 mRNA may be degraded by nonsense RNA-mediated decay.
The suro-2 (jg92) mutantdisplays a mild Dpy phenotype compared with suro-2 (jg35). Genomic DNA sequencing of suro-2 (jg92) revealed an A>T mutation located in the noncoding region between rps-10 and suro-2 (Fig. 2a). Because D1007.4, suro-2, and rps-10 are putatively transcribed as a single operon, we compared the mRNA expression level of these genes using RT-PCR. We found that suro-2 mRNA expression decreased signi cantly compared to the other two genes (Supplementary g. 1e). Because knockdown of each gene using RNAi resulted in different phenotypes (Supplementary g. 1b) and suro-2 (jg92) showed a mild phenotype compared with the putative null suro-2 (jg35) mutant, the suro-2 (jg92) mutation and associated reduction of suro-2 mRNA expression are reliable. We speculate that jg92 allele has a problem in trans-splicing after transcription. In fact, 70% of C. elegans genes are trans-spliced after transcription using SL1 or SL2 leader sequences 35 . Although recent studies revealed that conserved sequences are required for trans-splicing 36,37 in C. elegans, we could not link the suro-2 (jg92) mutation to any of these sequences.

SURO-2 is a putative ER membrane protein
We generated GFP reporter constructs using the suro-2 promoter to examine the suro-2 expression pattern. Because the transgenic strain with a proximal upstream of suro-2 coding region did not express GFP, we generated another transgenic strain with a distal promoter upstream of the rps-10 coding region, and observed GFP expression in many tissues, such as the pharynx, hypodermis, body wall muscles, vulva, spermatheca, several neurons, and the intestine (Fig. 2b). We generated several GFP::SURO-2 fusion constructs using the Y37A1B.5 promoter (Y37A1B.5p) that enables to express gene products in the hypodermis 38 . Because we could not detect GFP expression from transgenic strains with terminal SURO-2::GFP fusions, we generated three additional GFP::SURO-2 constructs in which GFP was inserted between the cytoplasmic domains of SURO-2. The pJG720 plasmid displayed the best GFP expression, which showed uorescence around the nuclei and relatively speci c uorescence in the cytoplasm We also generated several organelle markers by expressing GFP fusion proteins in the hypodermis to determine the exact subcellular localization of SURO-2. TRAM-1::GFP was used for the ER marker, and AMAN-2::GFP for the Golgi apparatus marker. TRAM-1 is the C. elegans orthologue of the translocating chain-associated membrane protein 1 39 , and AMAN-2 is an alpha-mannosidase II which functions at the Golgi apparatus 40 . The pJG735 plasmid is a tdTomato version of pJG720 and expresses tdTomato::SURO-2 in the hypodermis. Most tdTomato::SURO-2 proteins overlapped exactly with TRAM-1::GFP, but not with AMAN-2::GFP puncta (Fig. 2c). Because other ER markers such as GFP::KDEL and GFP::PISY-1 also overlapped with tdTomato::SURO-2 (Supplementary g. 2B), we concluded that SURO-2 mainly localized to the ER membrane.
Because suro-2 encodes two isoforms via alternative splicing, we examined the subcellular localization and functional differences of these two proteins. To examine the expression of SURO-2A and SURO-2B in the hypodermis, we generated a transgenic strain expressing tdTomato::SURO-2A and GFP::SURO-2B simultaneously. We found that these two proteins overlapped in most area (Fig. 2d), suggesting SURO-2A and SURO-2B function at the same ER membrane. Next, we performed rescue experiments using the suro-2 jgIs4 strain to compare protein function involved in cuticle formation. We crossed suro-2 jgIs4 with transgenic strains expressing GFP::SURO-2A or GFP::SURO-2B. The suro-2 jgIs4 strain expressing the GFP::SURO-2A transgenic product recovered the Rol phenotype, whereas the suro-2 jgIs4 strain expressing GFP::SURO-2B did not (Fig. 2e). Therefore, hypodermal SURO-2A expression is required and su cient for proper cuticle formation. SURO-2 is closely related to COPII in relation to collagen secretion A large-scale yeast two-hybrid screen for protein networks and a genome-wide prediction of C. elegans genetic interactions proposed associations between D1007.5 (SURO-2) and nucleopore protein-20 (NPP-20) 41,42 , the C. elegans orthologue of human protein Sec13. Sec13 is a component of the COPII complex originally discovered in S. cerevisiae and participates in the early collagen secretory pathway 43,44 . Another role of Sec13 relates to its function as a nucleopore protein during nuclear envelope reassembly 45 and is associated with the GATOR2 complex that regulates the mTORC1 complex 46 . Amidst varied functions, we focused on its potential role in the collagen secretory pathway to understand the relationship between SURO-2 and NPP-20, because SURO-2 is localized at the ER and participates in cuticle formation. We examined the knockdown phenotype of npp-20 via RNAi to compare its phenotype with that of the suro-2 mutant. Like suro-2 mutants, npp-20 RNAi suppressed the Rol phenotype of jgIs4 (Fig. 3a), implying a close relationship between suro-2 and npp-20. Therefore, we performed RNAi-mediated knockdowns of other COPII components to examine their functions in cuticle formation. Knockdown of most COPII genes resulted in Rol suppression. Knockdown of sar-1 and tfg-1 also resulted in Rol suppression (Fig. 3b). SAR-1 is the orthologue of human Sar1 GTPase which is the key regulator of COPII vesicle formation 19 . TFG-1 associates with SEC-16 and also functions in early collagen secretion 47 . Collectively, these results suggest that SURO-2 and COPII may function in the same secretory pathway.
Following Rol suppression by COPII knockdown, we determined collagen secretion was affected by COPII RNAi. We compared COL-19::GFP expression in the cytoplasm to identify potential de ciencies in collagen secretion when SURO-2 or NPP-20 was depleted by RNAi. COL-19::GFP was evenly distributed in the wild-type background, but COL-19::GFP accumulated in the cytoplasm by SURO-2 and NPP-20 knockdowns (Fig. 3c). We observed larger COL-19::GFP puncta in SURO-2-depleted worms than in NPP-20-depleted worms, implying that SURO-2 is more specialized in collagen secretion than NPP-20. Other components of COPII were depleted by RNAi and also resulted in COL-19::GFP accumulation in the cytoplasm (Fig. 3d). COPII plays a major role in early secretion of collagens from ER-to-Golgi apparatus 17, 23 , and our experimental data supported this role as described in previous studies. Therefore, SURO-2 and COPII are closely related with respect to collagen transport and cuticle formation.
SURO-2 interacts directly and is located together with NPP-20 SURO-2 and COPII are collectively involved in collagen secretion, and hypodermal expression is important for cuticle formation. We performed a glutathione S-transferase (GST) pulldown analysis to examine the direct interaction of SURO-2 and NPP-20. Because we were unable to express a SURO-2 protein including its TM domains in E. coli, we used its two long loop domains exposed to cytoplasm ( Fig  4A). We found that NPP-20 and SURO-2 interacted directly through the 6th loop (L6) of SURO-2 using in vitro binding analysis (Fig. 4b).
We constructed 11 plasmids expressing series of YFP::SURO-2 deletion proteins to examine in vivo expression and function of SURO-2 domains (Supplementary g. 3A). Most YFP::SURO-2 proteins exhibited similar expression patterns to intact protein excepting SURO-2C3, which has only C-terminus of SURO-2 and expressed like simple GFP expression in the cytoplasm and nucleus, while some proteins yielded strong punctate uorescence. In particular, YFP::SURO-2 proteins having the L6 domain with the C-terminus formed large bright puncta (Supplementary g. 3B). These results correlate with in vitro interactions of SURO-2 and NPP-20 and suggest the importance of the L6 domain of SURO-2 for in vivo complex formation.
Super resolution images using Airyscan determined the identity of the puncta formed by NPP-20::GFP and tdTomato::SURO-2. The processed images of both small and large puncta appeared as circular forms. As a result, each punctum can be assumed to be a uorescence image of the vesicle. The small puncta composed of NPP-20::GFP was around 100 nm in diameter ( Fig. 4f left panel). Whereas, large puncta of NPP-20::GFP and tdTomato::SURO-2 were often larger than 400 nm in diameter ( Fig. 4f right  panel). This result implies that SURO-2 is required and facilitates large COPII vesicle formation.
Large COPII vesicles were actively produced during molting We observed uorescence over time from mid L4 larvae to investigate when large vesicles formed. In the mid L4 stage, tdTomato::SURO-2 expressed at ER membrane and NPP-20::GFP formed small puncta. From the late L4 stage, large puncta composed of tdTomato::SURO-2 and NPP-20::GFP began to form, and increased until young adults. The peak time of large vesicle formation represented by NPP-20::GFP puncta size and intensity is around 24 hours from the mid L4 stage. Passing the gravid adult stage, tdTomato::SURO-2 expression disappeared in the hypodermis and NPP-20::GFP expressed as small puncta (Fig. 5a). This observation implies that large vesicle formation is highest when new cuticles form, and most SURO-2 proteins in the hypodermis are consumed for making large vesicle in contrast to NPP-20 which existed as small vesicles in late adults.
Following strong interaction of SURO-2 and NPP-20, we examine the relationship of these proteins by depleting SURO-2 or NPP-20 using RNAi. SURO-2 RNAi resulted in reduction of NPP-20::GFP expression, both of intensity and large puncta formation in compared to control RNAi. When NPP-20 was depleted by RNAi, tdTomato::SURO-2 disappeared completely (Fig. 5b). However, small NPP-20::GFP puncta existed after SURO-2 depletion. These results suggest that SURO-2 is specialized for large vesicles rather than general COPII vesicles. The mutual stabilization and cooperation of SURO-2 and NPP-20 proteins seem to form large vesicles to facilitate collagen transport from the ER.
We also investigated if COPII regulators are required for large vesicle formation as their knockdown suppressed the Rol phenotype (Fig. 3b). We performed sar-1 RNAi using the jgIs56 strain expressing tdTomato::SURO-2 and NPP-20::GFP. To observe young adults, we transferred L4 larvae to sar-1 RNAi plates because sar-1 RNAi from early larvae resulted in larval lethality. After 24 hours from L4, sar-1 knockdown resulted in signi cant reduction of large vesicles compared with control RNAi. The mean vesicle size and intensity of NPP-20::GFP decreased by SAR-1 depletion (Fig. 5c). From this result, we know that SURO-2 associated large vesicle formation is basically dependent on COPII and its regulator.
We further investigated to nd out important regions of L4 and L6 domains by sequence alignment of SURO-2 and human TMEM39A and B. The posterior region of L4 domain and the anterior region of L6 domain are highly conserved compared with other regions. Seven out of ten amino acids from 232 to 241 and eight out of eleven amino acids from 318 to 328 are positive among three proteins (Fig. 6b). We generated 10 aa deletion (tdTomato::SURO-2Δ10) or 11 aa deletion (tdTomato::SURO-2Δ11) expression constructs to examine the importance of these conserved amino acids in L4 and L6 domains. In the wildtype background, tdTomato::SURO-2Δ10 expression was stronger than tdTomato::SURO-2A similar to tdTomato::SURO-2ΔL4 expression. tdTomato::SURO-2Δ11 and tdTomato::SURO-2Δ10Δ11 did not express well similar to tdTomato::SURO-2ΔL6 (Supplementary g. 4B). In the suro-2 mutant background, any tdTomato::SURO-2 deletion proteins did not express well. As expected, the size and intensity of NPP-20::GFP puncta when expressed together with tdTomato::SURO-2Δ10 or tdTomato::SURO-2Δ11 decreased signi cantly in the suro-2 mutant (Fig. 6c).

SURO-2 is required for SURO-1/Carboxypeptidase A secretion
SURO-1 is a carboxypeptidase A (CPA) and is required for normal cuticle formation. Since the suro-1 mutant exhibited a mild Dpy compared with the suro-2 mutant that is a severe small Dpy, SURO-1 is one candidate cargo protein of the SURO-2 vesicle. In particular, SURO-1::DsRed is also secreted to the cuticle as a large vesicle 32 . To examine that SURO-2 is required for SURO-1 secretion, the jgIs32 integration line expressing SURO-1::DsRed was crossed with the suro-2 mutant. SURO-1::DsRed exhibited a typical cuticle pattern in wild type, but it was barely detected in the suro-2 mutant excepting marginal regions of the cuticle. Whole uorescence intensity of SURO-1::DsRed also decreased signi cantly in the suro-2 mutant (Fig. 7a). Next, we generated a transgenic strain expressing SURO-1::GFP and tdTomato::SURO-2 to examine that SURO-1 is localized at the SURO-2 vesicle. Most SURO-1::GFP proteins were exactly localized at the same puncta of tdTomato::SURO-2 (Fig. 7b). These results indicate that one of the collagen-modifying enzymes, SURO-1/CPA, is secreted by large vesicles composed of SURO-2.
In conclusion, the results of this work are summarized in a simple model (Fig. 7c). Soluble proteins are transported from the ER by conventional COPII vesicles with an average diameter of 80 nm. On the other hand, many proteins involved in ECM should be transported from the ER using large vesicles. Both types of vesicles, large and small, contain COPII and commonly require regulators such as Sar1. The conserved protein SURO-2/TMEM39 is necessary for the formation of this large vesicle and may function as the basic mechanism of bulky secretion, from simple animals to mammals.

Discussion
We have identi ed novel genes involved in collagen secretion and ECM remodeling using C. elegans to provide basic knowledges for solving brotic diseases. SURO-2/TMEM39, a suro-2 gene product, is a highly conserved protein found in animals. The secretion level of collagen proteins was drastically reduced in the suro-2 mutant, indicating SURO-2, rather than TANGO1/cTAGE5, may play a fundamental role in bulky collagen secretion like recently reported TMEM131 27 . Since TANGO1/cTAGE5 have only been found in higher animals, the TANGO1-cTAGE system may have evolved to enable transport of diverse collagens. In fact, collagen VII, a main cargo of TANGO1-associated COPII vesicles, does not exist in C. elegans. However, we cannot dismiss the possible existence of protein other than SURO-2 that is directly involved in bulky collagen secretion because the suro-2 (jg35) mutant phenotype was less severe than that of a single essential collagen mutant. In addition, the cuticle still formed without SURO-2, indicating cuticle formation is not wholly depended on SURO-2 (Fig. 1d). Nevertheless, our data suggest that SURO-2 is a common key component for collagen secretion among diverse animal species.
Two TMEM39 genes exist in the human genome, as opposed to a single gene in C. elegans. TMEM39A seems to be closely associated with the human immune system, as several genome-wide association studies have reported relationships between TMEM39A polymorphisms and several autoimmune diseases, such as multiple sclerosis (MS) 48 and systemic lupus erythematosus 49 . For example, the rst genome-wide association study of MS patients identi ed KIF21B and TMEM39A as MS susceptibility loci 48 . MS, which affects the central nervous system, is likely to be linked to the collagen secretion function of TMEM39A in the fact that it is clinically indistinguishable from collagen disorders 50,51 . KIF21B is a kinesin motor protein that functions in Golgi-to-ER retrograde transport mediated by COPI vesicles 52 . Likewise, Arf1 which is a key regulator of COPI vesicle formation, seems to be required for proper cuticle formation, because knockdown of arf-1, the C. elegans Arf1 orthologue, resulted in Rol suppression (our unpublished data). Defects in either TMEM39A or KIF21B could lead to perturbation of the ER-to-Golgi or Golgi-to-ER secretion of certain proteins involved in the immune system and MS. No functional study of TMEM39B has been published to date, although one previous report indicates TMEM39B is highly expressed in diffuse large B-cell lymphomas 53 . Thus, our results regarding the functionality of SURO-2/TMEM39 could provide preliminary information to guide studies of TMEM39A/B and their roles in human disease.
COPII and its associated gene products are required for collagen transport from the ER in vertebrates 12 .
In C. elegans, SEC-23/Sec23 was reported to function in collagen secretion during embryogenesis 54 . Our RNAi experiments showed that most COPII genes suppressed the Rol phenotype and induced collagen accumulation in the cytoplasm as observed in suro-2 knockdown. However, SURO-2 depletion exhibited more obvious phenotypes than the COPII gene knockdown in collagen secretion (Fig. 3). Because COPII is essential for the export of general proteins including soluble and membrane proteins, COPII depletion resulted in more pleiotropic and detrimental phenotypes than SURO-2 was depleted. The suro-2 (jg35) mutant, which is expected to be a null mutation, is alive, while most of the COPII gene RNAi leads to embryonic death. Thus, SURO-2 may speci cally facilitate bulky secretion of collagens and collagen processing enzymes, rather than of general proteins. However, SURO-2 may be involved in secretion of other proteins besides collagens, as suro-2 was expressed throughout various cell types that do not produce collagens (Fig. 2b).
SURO-2 protein has eight putative TM domains and two long cytoplasmic loop domains. We determined that the L6 cytoplasmic domain directly binds NPP-20/Sec13. Due to its eight TM domains, SURO-2/TMEM39's movement might be more restricted than that of NPP-20/Sec13, a soluble cytoplasmic protein. Therefore, SURO-2 may recruit Sec13-Sec31 outer coatomers through direct interaction with Sec13 when a large vesicle is forming. SURO-2 and NPP-20 localization patterns also suggest that SURO-2 recruits NPP-20 to large collagen transport vesicles. Large puncta of NPP-20::GFP localization overlapped exactly with those of tdTomato::SURO-2, although smaller puncta of NPP-20::GFP did not (Fig. 4). Secretion of SURO-1, a CPA involved in collagen processing, was also dependent on SURO-2 function (Fig. 7a). SURO-1::DsRed proteins were secreted as large puncta in the cytoplasm and initially considered to be secretory granules because of their large size 32 . Taken together, SURO-1 function is interpreted as a subset of SURO-2 because SURO-1 and SURO-2 localized at the same large puncta (Fig. 7b), and the suro-1 mutant has a much weaker phenotype than the suro-2 mutant.
Cytoplasmic regions of SURO-2 are composed of 5 domains including N-terminus, L2, L4, L6 and Cterminus (Fig. 4a). We tried to clarify the function around L4 and L6 domains, because they are long enough to have roles compared with the short L2 domain. Both L4 and L6 domains are required for SURO-2 expression, and particularly, L6 domain binds NPP-20/Sec13 directly. Because large vesicles are composed of SURO-2 and COPII, NPP-20 interacting L6 domain has a critical role in collagen secretion. SURO-2 and human TMEM39 proteins have highly conserved regions in L4 and L6 domains. These conserved 10 amino acids in L4 and 11 amino acids in L6 are required for L4 and L6 function to express SURO-2 itself and form large vesicles. The greatest difference between L4 and L6 was that the L4 deletion form increased in the wild type but was not expressed well in the suro-2 mutant. When endogenous SURO-2 and L4-de cient SURO-2 coexist, there appears to be a complex interaction between SURO-2 proteins.
Because the secondary protein structure and binding partner of SURO-2 are unlike TANGO1, a single-TM protein that interacts with Sec23-Sec24 inner coatomers 24, 41 , mechanisms of collagen-packaging vesicle formation between the two proteins may differ. TANGO1 has a long tail and delays COPII vesicle formation by interacting with collagen VII directly 41 . However, both termini of SURO-2/TMEM39 are exposed to the cytoplasm, and only four short loop regions (L1, L3, L5 and L7) composed of less than 20 amino acids are located on the luminal side of the ER (Fig. 4a). We propose that SURO-2 directly incorporates into the COPII membrane and reduces its exibility via SURO-2's eight TM domains and four luminal loop domains. This complex of ER membrane and SURO-2 proteins may look like sewn buttons on fabric. The in exible domains of the COPII membrane formed by SURO-2 may inhibit the formation of standard-sized COPII vesicles and instead induce spontaneous formation of a larger COPII-SURO-2 vesicle. This hypothesis requires additional investigation using biochemical, structural, and biophysical approaches, although we propose a model for early collagen secretion based on our data presented here (Fig. 7c). In mammals, HSP47, which functions as a collagen chaperone, is present in the ER lumen 55 , and the presence of HSP47 orthologue has not been reported in nematodes. Recently reported TMEM131 recruits premature collagen monomers through the bacterial PapD chaperone-like domain and functions in collagen assembly and secretion. The C-terminal region of TMEM131 interacts with TRAPPC8, a component of the TRAPP tethering complex, and functions in collagen secretion 27 . In this regard, SURO-2/TMEM39 and TMEM-131/TMEM131 seem to cooperate in collagen secretion.
In summary, we elucidated novel functions of SURO-2/TMEM39 in C. elegans, thus providing valuable insights into bulky protein secretion, matrix biology, and TMEM39A-associated diseases including cancer brosis in higher animals.

Mapping and cloning of suro-2 mutants
We used a SNP mapping method to identify suro-2 (jg35) mutation as previously described 56 . More than 1,000 F2 progeny that exhibited Dpy and non-Rol phenotypes were selected from the cross of suro-2 (jg35) jgIs4 and Hawiian CB4856 male. Several F3 worms from each F2 plate were lysed, and their genomic DNA was used for PCR templates. PCR products were digested by restriction enzymes speci c to each SNP and analyzed by 2% agarose gel electrophoresis. The RNAi clones for 12 candidate genes in the mapped region were originated from the Ahringer library 57 . Several L4-stage jgIs4 worms were transferred to each RNAi plate, and F1 progeny were observed. The WRM067cA7 fosmid and 8 kb PCR products, including three genes in the same operon, were used for the rescue experiment. We generated transgenic worms and mated each transgenic line with suro-2 mutants because direct microinjection ruptured suro-2 mutants. Finally, PCR-ampli ed D1007.5 from the suro-2 mutant was sequenced to detect molecular resin (Cosmogenetech, Seoul, South Korea). Whole-genome sequencing of suro-2 (jg92) was performed by Macrogen (Seoul, South Korea).

RNA interference
Most RNAi clones that were originated from the Ahringer library 57 were kindly provided by Dr. J. Lee (Seoul National University). Additionally, we constructed RNAi clones for suro-2 using its cDNA and L4440 vector. The 1.4 kb suro-2 cDNA was ampli ed by PCR using the wild-type N2 cDNA as templates.
The PCR fragments were cloned into L4440 vector using SalI and BamHI sites. Primers used for suro-2 RNAi clone were D1007.5-03 (5ATG TCG ACA TGC CGC CTC GAA GAC G 3) and D1007.5-23 (5 TGG ATC CTT ATC GTT CTT GTT GAA GTT GAT GC 3). We transferred different staged worms for each gene RNAi according to the extent of RNAi effect. L4 stage worms were transferred for control, sec-31 and suro-2 RNAi, and gravid adult worms for npp-20 and sec-16 RNAi. F1 progeny were observed in these cases. However, the effect of sec-12, sec-23, sec-24.1, sar-1 and tfg-1 RNAi was very severe, and L1-L2 larvae were transferred and observed when they grew up. The empty vector L4440 was used for control.

GST pull-down and western blot analyses
The pGEX4T1 GST vector system (GE helthcare, Wauwatosa, WI, USA) or GST-SURO-2 and the pRSET vector system (Thermo-Fisher  Cuticle malformation of suro-2 mutants. a Dpy and Rol suppression phenotype of suro-2 mutants. The suro-2 jgIs4 strain is more severe than a single suro-2 mutant. Scale bar= 100 µm. b Dead worms of wild type and suro-2 mutants (D6= adult day 6). The suro-2 mutant worms often ruptured at the young adult stage (D2= adult day 2). The ruptured suro-2 mutant released its contents including the intestines and gonads (arrow). Scale bar= 150 µm. c Approximately 18% of suro-2 mutants ruptured and died within 3 days of becoming adults. Error bar= a standard error. d Electron microscopy of adult cuticle. Twenty adult animals of wild type or suro-2 mutants were analyzed. The average cuticle thickness in wild type and suro-2 mutants was 604 ± 23.2 nm and 233 ± 18.7 nm, respectively. Black H lines indicate cuticle thicknesses. C= cuticle, M= body wall muscle, and H= hypodermis. ± sign is standard deviation. Scale bar= 500 nm. e Expression of two collagen markers, ROL-6::GFP and COL-9::GFP in adult cuticle. Scale bar= 10 µm. The lower panel quanti es GFP uorescence along thin white lines in each gure to represent ROL-6::GFP and COL-19::GFP as wild type and suro-2 mutants. In the suro-2 mutant, ROL-6::GFP decreased from 31.6 ± 9.93 to 15.9 ± 5.95, and COL-19 :: GFP decreased from 30.8 ± 8.98 to 19.9 ± 5.23 compared to the wild type. ± sign is standard error.     When new cuticle components were synthesized and secreted, SURO-2 forms large vesicles with NPP-20 / SEC13 from the ER membrane. SAR-1/Sar1 is required for both conventional COPII vesicle and large vesicle formation.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.