PIGN prevents protein aggregation in the endoplasmic reticulum independently of its function in the GPI synthesis

In eukaryotic cells, newly synthesized secretory proteins must pass quality control by chaperones for proper folding in the endoplasmic reticulum (ER) before they are secreted and become functional. Misfolded proteins that fail to pass this quality control form aggregates that cause ER stress. To mitigate this ER stress, the unfolded protein response (UPR) attenuates translation and induces expression of ER residential chaperones (Schröder and Kaufman, 2005; Walter and Ron, 2011). If the ER stress is not resolved within a certain time period, cells undergo apoptosis (Tabas and Ron, 2011; Walter and Ron, 2011). Therefore, aggregated proteins in the ER are harmful for cellular viability.

The PIGN protein was originally identified as an enzyme that synthesizes the glycosylphosphatidylinositol (GPI)-anchor (Gaynor et al., 1999). The GPI-anchor is a post-translational modification found in ∼60 proteins in yeast and 150 proteins in humans. GPI is synthesized and transferred to proteins in 12 stepwise reactions, which are catalyzed by ER-associated enzymes encoded by more than 20 genes – the so-called called PIG genes – including PIGN (Ferguson et al., 2009; Kinoshita et al., 2008). PIGN catalyzes the transfer of phosphoethanolamine (EtNP) to the first mannose residue at precursors of the GPI-anchor (Gaynor et al., 1999; Hong et al., 1999). Autosomal-recessive mutations in the human PIGN gene were identified in patients with multiple congenital anomalies-hypotonia-seizures syndrome 1 (MCAHS1, Online Mendelian Inheritance in Man, OMIM 614080) (Maydan et al., 2011). In addition to the original mutations reported in patients with this disease, many recessive missense PIGN mutations have also been identified in patients with MCAHS1 (Brady et al., 2014; Couser et al., 2015; Fleming et al., 2015; Khayat et al., 2015; Nakagawa et al., 2015; Ohba et al., 2014). Recently, some patients with Fryns syndrome (OMIM 229850), which causes lethality during the neonatal period through congenital diaphragmatic hernia (CDH) and other associated malformed features, have also been reported to harbor recessive mutations in the PIGN gene (McInerney-Leo et al., 2016).

By screening for mutants of Caenorhabditis elegans that conferred defects in protein secretion, we identified an os156 mutation in the pign-1 gene encoding a PIGN homolog. Although the pign-1(os156) mutant possessed normal GPI-anchor biosynthetic activity, protein aggregation was observed in the ER. Two other mutations equivalent to those observed in patients with MCAHS1 and that affect the GPI anchor biosynthesis also caused similar protein aggregation. In a mammalian cell line that lacks the PIGN gene, accumulation of protein aggregates – as in C. elegans – was observed Our findings demonstrate a conserved function for PIGN-1/PIGN in the regulation of protein quality control in the ER, and the lack of it might be involved in symptoms of patients with PIGN mutations.

To identify genes required for the efficient secretion of basement membrane (BM) proteins, we used a transgenic line of C. elegans that expresses type IV collagen::mCherry (emb-9::mCherry) (Ihara et al., 2011). EMB-9::mCherry is secreted from body wall muscle (BWM) cells and localizes to the BM in vivo (Fig. 1A, Fig. S1A). We conducted a forward genetic screen using EMB-9::mCherry and succeeded in isolating the recessive mutant os156, which exhibits a fully penetrant phenotype that is characterized by accumulation of EMB-9::mCherry in BWM cells (Fig. 1B). In wild-type animals, EMB-9 is continuously secreted from BWM cells (Graham et al., 1997) and localizes primarily to the BM covering most tissues throughout larval development (Fig. 1A,C; Fig. S1A,B; Movie 1). In contrast, in the os156 mutant, accumulation of EMB-9::mCherry fluorescence was observed in all BWM cells (Fig. 1B, D; Fig. S1C; Movie 2). Compared to wild-type animals, the os156 mutant had three times more EMB-9::mCherry localized in the BWM cells in the head region (Fig. 1I). In addition, the fluorescence intensity of EMB-9::mCherry at a region of the BM in the pharynx (along the dashed lines in Fig. S1B,C) was decreased ∼30% in the os156 mutant compared to wild type (Fig. S1D), indicating defects in protein secretion. Transgenic animals expressing secreted soluble GFP (ssGFP) (Fares and Greenwald, 2001), laminin::GFP (lam-1::GFP) (Kao et al., 2006) and secreted metalloprotease::Venus (mig-17::venus) (Nishiwaki et al., 2000) from BWM cells showed similar dot-like localizations (Fig. 1E–H and Fig. S1E,F). In contrast, transmembrane proteins (MOM-5/Frizzled::GFP and PAT-3/ β-integrin::GFP) showed normal localization on the cell surface in the os156 mutant (Fig. S1G–J), suggesting that the os156 mutation causes a global defect in the secretion of soluble but not transmembrane proteins – at least in BWM cells.

We mapped the os156 mutation between SNPs located at 991k and 2819k of C. elegans chromosome I. Whole-genome sequencing revealed that os156 harbors a missense mutation of the arginine residue at position 78 of Y54E10BR.1, which we named pign-1 (see below) (Fig. S2). Similar to the os156 mutant, protein accumulation is also present in the pign-1 deletion mutants (ok1118 and os162) that additionally show a larval-arrest phenotype, in contrast to os156 with the missense mutation (Fig. 2A,C). The expression of the GFP::PIGN-1 fusion protein in the BWM cells (myo-3p::GFP::PIGN-1) rescued the phenotype of the os156 and ok1118 mutants, confirming that pign-1 is responsible for the phenotype and also showing that PIGN-1 functions cell autonomously (Fig. 2B,C). The predicted amino acid sequence of PIGN-1 indicates that it is homologous to human PIGN, with 14 transmembrane domains and a large hydrophilic ER luminal domain containing three conserved motifs (Fig. 2D; Fig. S2). The PIGN enzyme adds phosphoethanolamine (EtNP) to the first mannose residue of GPI precursors (Fig. 2E) (Hong et al., 1999). Because the proteins (ssGFP, EMB-9, LAM-1 and MIG-17) that showed abnormal accumulation in the pign-1(os156) mutant are non-GPI-anchored proteins, these results indicate that pign-1 is important for secretion of non-GPI-anchored proteins in the ER.

A pign-1 reporter (pign-1p::GFP) containing a 1.9 kb promoter sequence upstream of the pign-1 start codon was expressed at the highest levels in the pharynx and BWM cells and at lower levels in the intestine and hypodermis (Fig. 2F). The subcellular localization of PIGN-1 in BWM cells was analyzed by using myo-3p::GFP::PIGN-1. Consistent with previous studies that have used yeast and human cells (Gaynor et al., 1999; Hong et al., 1999), GFP::PIGN-1 localized to the ER membrane (as determined by colocalization with an ER marker, RFP::SP12) (Fig. 2G). GFP::PIGN-1 with the R78K mutation as in the os156 mutant also localized normally to the ER membrane (Fig. 2H), indicating that the os156 mutation affects the function but not localization of PIGN-1.

The accumulation of secretory proteins (EMB-9::mCherry, ssGFP, laminin::GFP, and MIG-17::venus) in BWM cells in pign-1(os156) led us to hypothesize that proteins can amass in organelles, such as lysosomes, ER or the Golgi complex. To determine these organelles, we established transgenic animals that express EMB-9 fused with the pH-sensitive green fluorescent protein pHluorin, whose fluorescence is quenched at vesicular pH, e.g. pH 5.5 (Dittman and Kaplan, 2006). In the pign-1 (os156) mutant, colocalization of the EMB-9::mCherry and EMB-9::pHluorin was detected in BWM cells (arrowheads in Fig. 3A) but not in coelomocytes (arrows), i.e. scavenger cells with acidic organelles. These results indicate that those proteins accumulated in organelles, such as the ER or Golgi complex, that are near neutral pH. To examine organelles with accumulated proteins, we simultaneously visualized EMB-9::mCherry, the ER marker TRAM::YFP (Rolls et al., 2002) or the Golgi marker MIG-23::GFP (Nishiwaki et al., 2004). Although the ER in wild type showed the typical morphology of the reticular network, the ER in the pign-1(os156) mutant did not show normal network morphology and the fluorescence of EMB-9::mCherry was confined to the lumen of vesicle-like exaggerated ER membranes (Fig. 3B,C). In contrast, the morphology of the Golgi complex, which was observed by using MIG-23::GFP, was not significantly different between wild type and pign-1(os156), and the marker did not colocalize with the EMB-9::mCherry signals (Fig. S3A,B).

Transmission electron microscopy of cross-sections of BWM cells of the pign-1(os156) mutant expressing EMB-9::mCherry showed large electron-dense structures surrounded by membrane studded with ribosomes that are likely to be protein aggregations in the rough ER (Fig. 3D–F). The largest of these expanded vesicles in our observation was ∼2.5 µm (Fig. S3C). Similar enlarged structures with ribosomes were observed without the EMB-9::mCherry transgene (Fig. 3G), indicating that protein aggregation within the ER lumen of the pign-1 mutants was not caused by the overexpression of EMB-9::mCherry.

It has been reported that protein aggregates within the ER often contain aberrantly formed disulfide bonds (Braakman et al., 1992; Marquardt and Helenius, 1992). Therefore, we examined the formation of aberrant disulfide crosslinks in EMB-9::mCherry by SDS-PAGE under non-reducing conditions. Western blot analyses with an anti-mCherry antibody showed bands of high molecular mass (>500 kDa) at the top of the gel in a sample from the pign-1(os156) mutant but not from wild type (Fig. S3D). Under reducing conditions, the bands of high molecular mass were not detected and a major band of nearly 250 kDa corresponding to full-length EMB-9::mCherry was observed in both samples (Fig. S3E) (Graham et al., 1997). These observations confirmed protein aggregation with aberrant disulfide crosslinking in the ER lumen of the pign-1(os156) mutant.

Protein aggregation in the ER should activate the UPR pathway. In fact, the expression of HSP-4, a homolog of a human chaperone and the UPR stress sensor BiP/GRP78 (Shen et al., 2001), was increased in the pign-1(os156) mutant, as determined by hsp-4p::GFP, indicating the activation of the UPR pathway in response to protein aggregation (Fig. S3F). Even upon UPR activation, protein aggregation is not resolved in the pign-1 mutant. These results show that PIGN-1 is required for quality control to prevent accumulation of protein aggregates in the ER.

To determine whether the role of pign-1 in protein quality control is conserved in mammalian cells, we expressed ssGFP in wild-type and PIGN-knockout HEK293 cells (Ohba et al., 2014). High levels of ssGFP puncta were observed in PIGN-knockout HEK293 cells but not in wild-type cells (Fig. 4A, B). These puncta colocalized with mCherry-ER-3 (calreticulin) – an ER luminal chaperone (Fig. 4F) – indicating that the puncta represent aggregation of ssGFP in the ER. Transient transfection of HEK293 cells with vector expressing wild-type human PIGN abrogated the protein aggregation in PIGN-knockout HEK293 cells (Fig. 4C–E). These observations strongly suggest that the regulation of protein quality control by PIGN is evolutionarily conserved from C. elegans to humans.

To exclude the possibility that the aggregation of ssGFP is caused by exit defects from ER rather than defects of protein quality control, we examined colocalization of Sec-31A-TagRFP-T (a marker for ER exit sites and COPII vesicles) (Shibata et al., 2010) and ssGFP in PIGN-knockout HEK293 cells that stably expressed ssGFP. Punctate signals of Sec-31A-TagRFP-T did not colocalize with ssGFP (Fig. 4G). Furthermore, in the C. elegans sec-16(tm5375) deletion mutant, which is defective in the initiation of the coat protein II (COP-II) assembly that is required for transport of vesicles from the ER (Espenshade et al., 1995; Supek et al., 2002), EMB-9::mCherry localized normally in the basement membrane around the pharynx (Fig. S3G). These results indicate that the protein aggregation is caused by defects in quality control within the ER rather than the disruption of COP-II vesicle function.

The phenotype of the pign-1 mutants suggested that mutations in other genes involved in GPI biosynthesis cause similar phenotypes. We analyzed deletion mutants of enzymes involved in the GPI biosynthesis: piga-1(tm2939), pigp-1(ok3557), pigo-1(tm1906), pigk-1(tm2739) and pgap-2(gk285) (Murata et al., 2012). These mutants shared the sterile phenotype, which is likely to be caused by defects in the GPI-anchor synthesis. Unexpectedly, in all of these deletion mutants, EMB-9::mCherry did not form dot-like aggregates and was normally detected in the BM of the pharynx (Fig. 5A–F). These results indicate that only PIGN-1 among the GPI biosynthetic enzymes regulates protein quality control in the ER.

Because PIGN functions in the middle of the GPI synthesis pathway (Kinoshita et al., 2008), it is possible that incorrectly formed GPI-anchor intermediates disrupt protein quality control in the pign-1 mutants. If so, mutation of piga-1, which catalyzes the first step of the GPI synthesis, should suppress the pign-1 phenotype. However, the pign-1(os156); piga-1(tm2939) double mutant showed aggregation of EMB-9::mCherry, similar to the pign-1(os156) single mutant (Fig. 5G, H). These results strongly suggest that PIGN-1 regulates protein quality control independently of the GPI synthesis.

To further examine the relationship between the GPI-anchor and protein quality control, we analyzed the effects of the os156 mutation on the GPI-anchor proteins (GPI-AP). We first analyzed the localization of EGFP::WRK-1 (GPI-AP) in pign-1 mutants. Consistent with previous reports (Murata et al., 2012), EGFP::WRK-1 expressed in BWM cells by the myo-3 promoter (myo-3p::EGFP::WRK-1) localized to the plasma membrane in wild type (Fig. 5I). In the pign-1(ok1118) or piga-1(tm2739) deletion mutants, as expected, EGFP::WRK-1 was not detected on the plasma membrane (Fig. 5J, L). Surprisingly, in the pign-1(os156, R78K) mutant, EGFP::WRK-1 localization was normal (Fig. 5K). We next used PIGN-knockout HEK293 cells expressing C. elegans PIGN-1 with or without the os156 mutation (R78K) and measured the cell surface expression of CD59 (GPI-AP) as a surrogate for the enzymatic activities by using flow cytometry. Consistent with the normal localization of EGFP::WRK-1 in the pign-1(os156) mutant, expression of C. elegans PIGN-1 (R78K) and wild-type PIGN-1, as well as human PIGN restored CD59 location to the plasma membrane at similar levels (arrows in Fig. 5M). These observations suggest that the PIGN-1 (R78K) protein possesses normal enzymatic activity (EtNP transferase) despite its compromised function in protein quality control.

The above results show that the GPI synthesis activity is independent from its role in protein quality control. However, it is possible that PIGN catalyzes substrates other than GPI to regulate protein quality control. PIGN has a large loop between the first and second transmembrane domain that contains three conserved motifs present in phosphodiesterases, nucleotide pyrophosphatases and sulfatases, each of which is believed to be important for the PIGN catalytic activity in the EtNP transferase reaction (Fig. 2D; Fig. S2) (Gaynor et al., 1999). Mutations in each putatively conserved motif of human PIGN were generated by replacing conserved histidine residues with alanine. The mutations strongly (H263A in motif 3) or moderately (H218A in motif 2) affected GPI synthesis of PIGN, as determined by the surface expression of CD59 (GPI-AP) in PIGN-knockout HEK293 cells (Fig. 6A). Unexpectedly, the H98A mutation in motif 1 had normal enzymatic activity. Surprisingly, all of the PIGN mutants efficiently rescued the protein aggregation phenotype in PIGN-knockout HEK293 cells (Fig. 6B,C). Furthermore, C. elegans PIGN-1 constructs with equivalent mutations (H97A, H214A, H258A) efficiently rescued protein aggregation in the C. elegans pign-1(ok1118) mutant (Fig. S4). These results clearly show that PIGN-1/PIGN mediates efficient quality control and secretion of proteins from the ER, independently of their catalytic activities, including that of EtNP transferase but through a so-far-unknown activity, which we named and hereafter refer to as the ‘non-canonical function’ of PIGN-1/PIGN.

Multiple congenital anomalies-hypotonia-seizures syndrome 1 (MCAHS1) is caused by recessive missense mutations in the PIGN locus (Maydan et al., 2011). Patients with MCAHS1 exhibit developmental delay, hypotonia and epilepsy combined with multiple congenital anomalies that may lead to early death (Maydan et al., 2011; Ohba et al., 2014). Although these PIGN mutations affect the GPI-anchor biogenesis in mammalian cells (Maydan et al., 2011; Ohba et al., 2014), they may also affect the non-canonical function of PIGN-1 in protein quality control. We used the CRISPR/Cas9 system to generate missense mutations in the C. elegans pign-1 gene (R679Q and S265P) that are identical to those present in MCAHS1 patients (designated as os163 and os164, respectively) (Fig. 2D, Fig. S2). In the pign-1(os163 and os164) and the deletion mutants, we observed the defect in EGFP::WRK-1 (GPI-AP) localization (Fig. 7A–C), confirming that these mutations affect GPI biosynthesis. In terms of aggregation of EMB-9::mCherry, the phenotypes of pign-1(os163 and os164) were indistinguishable from that of pign-1(os156) (Fig. 7D–F). To exclude the possibility that the phenotype of the pign-1 (os163 and os164) mutants is caused by abnormal localizations or unstable expression of the mutated proteins rather than their abnormal functions, we expressed GFP::PIGN-1 constructs with the corresponding mutations (S265P and R679Q) (Fig. 7G–I). Although some signals of GFP::PIGN-1(S265P) and GFP::PIGN-1(R679Q) showed abnormal localization that did not overlap with the ER marker (arrows in Fig. 7H′, I′), the majority of these proteins still localized to the ER (Fig. 7H, I). Therefore, the protein aggregation phenotype in these pign-1 mutants is likely to be caused by defects in the non-canonical function rather than localization. Taken together, our observations raise the possibility that some symptoms of MCAHS1 are caused by defects of the non-canonical function of PIGN for protein quality control in the ER.

In this study, we revealed an evolutionarily conserved and unexpected role of PIGN-1/PIGN in protein quality control in the ER, which we designated as the non-canonical function. In the C. elegans pign-1 mutants, but not in those of other genes belonging to the pig family, several proteins failed to be secreted and, instead, formed abnormal aggregates in the ER. PIGN-1/PIGN with mutations in the catalytic motifs that disrupt its enzymatic activity for the GPI-anchor biogenesis, efficiently rescued protein aggregation in PIGN-knockout HEK293 cells and in pign-1 mutants, indicating that the enzymatic activity of PIGN-1 is not required to prevent protein aggregation. In contrast, the os156 (R78K) mutation in pign-1, which causes the protein aggregation phenotype in C. elegans, does not affect EtNP transferase activity of the GPI biogenesis pathway. These results clearly show that PIGN-1/PIGN is a bi-functional protein: an EtNP transferase enzyme (canonical function) and a mediator of protein quality control within the ER (non-canonical function).

The PIGN-1 protein possesses 14 transmembrane domains in addition to a large first ER luminal loop. This loop contains residues that are required for canonical and non-canonical functions: R78 mutated in os156 is required only for the non-canonical function, H214 and H258 in the catalytic motifs are only required for canonical functions, whereas S265 associated with MCAHS1 is required for both of the functions (Fig. 2D). Therefore, it is plausible that PIGN cannot fulfill both the functions simultaneously and that these functions might be competitive. For example, the canonical activity might inhibit its non-canonical function. In addition, it appeared to be reasonable that ER stress causes dysfunction of PIG proteins (e.g. PIGA) in the ER, resulting in the decrease of PIGN substrates (intermediates of the GPI synthesis). The lack of the substrates of the canonical activity might stimulate the non-canonical activity of PIGN. If this is the case, PIGN might sense ER stress through dysfunction of the GPI pathway to switch its function to the non-canonical role preventing protein aggregation.

How does the non-canonical function of PIGN-1/PIGN prevent protein aggregation in the ER? One possibility is that activities of chaperones are regulated directly or indirectly by PIGN-1/PIGN. BiP, one of the most abundant ER chaperone that refolds misfolded proteins is transcriptionally activated following activation of the UPR pathway in response to ER stress (Hendershot et al., 1995; Zhao et al., 2005). In pign-1 mutants, we observed enhanced reporter expression of HSP-4/BiP, indicating that BiP is activated but fails to ameliorate protein aggregation in the pign-1 mutants, suggesting that the function of BiP is abrogated. Although involvements of other chaperons regulated by PIGN-1/PIGN are possible, PIGN-1/PIGN might be required for correct activity of BiP by directly regulating its activity, its cofactors or its correct localization in the ER. Interestingly, it has been reported that the expression of an ATPase-inactive BiP mutant in cultured cells or the disruption of mouse Sil1 (also known as BAP), which encodes an adenine nucleotide exchange factor for BiP, result in an expanded ER filled with protein aggregates (Hendershot et al., 1995; Zhao et al., 2005). Protein aggregates in the ER and a block of protein secretion were also observed following artificial depletion of ATP in cultured cells (Braakman et al., 1992; de Silva et al., 1993; Dorner et al., 1990; Verde et al., 1995). These studies emphasize the importance of ATP for chaperone functions in resolving protein aggregation within the ER. Interestingly, it has been reported that overexpression of the S. cerevisiae PIGN homolog MCD4 can facilitate ATP transport – mostly into the Golgi complex and weakly into the ER (Zhong et al., 2003). Therefore, it is tempting to speculate that PIGN-1/PIGN facilitates transport of ATP into the ER, which is required for correct chaperone function in order to prevent protein aggregation. Biochemical analyses of the roles PIGN has in the activity of chaperones and/or transport/metabolism of ATP will be an important step toward a better understanding of its non-canonical function.

Recently, in addition to the original mutation (Maydan et al., 2011), many new recessive PIGN mutations, mostly missense mutations, have been reported in patients with MCAHS1 (Brady et al., 2014; Couser et al., 2015; Fleming et al., 2015; Khayat et al., 2015; Nakagawa et al., 2015; Ohba et al., 2014). In addition to MCAHS1, PIGN mutations with premature stop codons have been identified in patients suffering from Fryns syndrome, a condition that shares some symptoms, e.g. congenital anomalies, with MCAHS1 but also shows more-severe symptoms, such as congenital diaphragmatic hernia and neonatal lethality (McInerney-Leo et al., 2016). Since mutations of other ‘PIG’ genes involved in the GPI-anchor biosynthesis have been identified in patients with similar symptoms (e.g. PIGA mutations in MCAHS2 or PIGT mutations in MCAHS3), most symptoms for MCAHS1 seem to be explained by defects in the canonical PIGN function. Nonetheless, the defects in the non-canonical function might affect some symptoms of MCAHS1 and Fryns syndrome. Since it is known that protein aggregation decreases cell viability in several cultured cells (Bucciantini et al., 2002; Rao and Bredesen, 2004), disruption of the non-canonical PIGN function might affect symptom of other diseases. Future research on the non-canonical function of PIGN/PIGN-1 and the regulation of its function is likely to contribute to the development of novel therapeutic strategies in a variety of diseases.

The N2 Bristol strain of C. elegans was used as wild-type strain (Brenner, 1974). Animals were raised at 20°C. The following alleles and transgenes were used in this study: osEx454 [mig-17p:: MIG-17::Venus; unc-119(+)], osEx502 [myo-3p::GFP::PIGN-1; unc-54p::RFP::SP12; unc-119(+)], osEx508 [emb-9p::EMB-9::pHluorin; emb-9p::EMB-9::mCherry; unc-119(+)], osEx512 [myo-3p::GFP::PIGN-1(R78K); unc-54p::RFP::SP12; unc-119(+)], osEx517 [myo-3p::GFP::PIGN-1; sur-5::gfp(+)], osEx524 [mom-5p::MOM-5::GFP; unc-76(+)], osIs54 [pign-1p::GFP; unc-119(+)], osIs60[unc-54p::MIG-23::GFP; unc-119(+)], osIs64[myo-3p::YFP::TRAM; unc-119(+)], osIs66[myo-3p:: EGFP::WRK-1; unc-119(+)], osIs77 [unc-54p::RFP::SP12; unc-119(+)]; arIs37 [myo-3p::ssGFP], pign-1&arx-7 (ok1118), pign-1(os156), pign-1(os162), pign-1(os163), pign-1(os164), pigo-1(tm1906), piga-1(tm2939), pigp-1 (ok3557), pgap-2(gk285), unc-119(ed4), pigk-1(tm2739), sec-16 (tm5375), qyIs10 [lam-1p::LAM-1::GFP], qyIs42 [pat-3::GFP;genomic ina-1], qyIs44 [emb-9p::EMB-9::mCherry], and zcIs4 (hsp-4p::GFP). The pign-1(os156) mutant was identified through the abnormal localization of EMB-9::mCherry after mutagenesis of qyIs44 animals with ethyl methanesulfonate. The pign-1(os156) and pign-1(os163) mutants were viable and did not exhibit any obvious phenotype, although the growth rate was less and brood numbers were slightly lower than those of wild-type animals. The pign-1(ok1118 and os162), and the deletion mutants of other ‘PIG’ genes (tm1906, tm2939, ok3557, gk285, and tm2739) shared the sterile phenotype, which is likely to be caused by defects in GPI-anchor synthesis.

To construct myo-3p::gfp::pign-1, we first generated the gfp::pign-1 plasmid. A cDNA fragment from yk573C9 corresponding to the entire PIGN-1 protein (residues 1-906) was amplified and inserted immediately after gfp in pPD95.75 (pPD vectors are gifts of A. Fire) in frame with the gfp sequence with a linker sequence (5′-GGATCCGGAGGAGGATCCGGAGGAGGATCC-3′) between gfp and pign-1. For the construction of myo-3p::gfp::pign-1, a gfp::pign-1 fragment was amplified by PCR from the gfp::pign-1 plasmid and inserted between the myo-3 promoter and the unc-54 3′ UTR of pPD136.64 by using the In-Fusion technique (Clontech). To create the mutations R78K, H97A, H214A and H258A in pign-1 cDNA, a QuikChange site-directed mutagenesis kit (Stratagene) was used with primers containing the corresponding nucleotide changes. To generate the pign-1p::gfp plasmid, the putative promoter sequence encompassing 1960 bp immediately upstream the first methionine codon was PCR-amplified and inserted into the SphI-PstI sites of pPD95.75. The emb-9p::emb-9::pHluorin plasmid was generated by replacing mCherry sequences with pHluorin (Dittman and Kaplan, 2006) in emb-9p::emb-9:mCherry (Ihara et al., 2011) by using the In-Fusion technique. For myo-3p::egfp::wrk-1, a egfp::wrk-1 fragment was amplified by PCR from the egfp::wrk-1 plasmid (Murata et al., 2012) and inserted between the myo-3 promoter and the unc-54 3′ UTR of pPD136.64 by using the In-Fusion technique. For rescue experiments, plasmids myo-3p::gfp::pign-1, myo-3p::gfp::pign-1 (H97A), myo-3p::gfp::pign-1 (H214A) or myo-3p::gfp::pign-1 (H258A) were injected into pign-1 (os156) or pign-1 (ok1118)/hT2 hermaphrodite gonads at 10 µg/ml with 50 µg/ml of unc-119 plasmid pDP#MM016B (Maduro and Pilgrim, 1995) and 90 µg/ml of digested salmon sperm DNA. Plasmids pign-1p::gfp, myo-3p::gfp::pign-1, myo-3p::gfp::pign-1 (R78K), myo-3p::egfp ::wrk-1 (10 µg/ml each) and myo-3p::emb-9::pHluorin (2 µg/ml) were each injected into the gonads of unc-119 (ed4) with 50 µg/ml of pDP#MM016B and 95 µg/ml of digested salmon sperm DNA. Microinjection was performed as previously described (Mello et al., 1991).

Images of C. elegans were acquired by using a Yokogawa spinning disk confocal scan head mounted on a Zeiss Axioplan 2 microscope equipped with a 63× Plan Apochromat oil objective lens that was controlled by using Andor iQ (Andor Technology). Images were also acquired by using a confocal laser scanning microscope (Olympus FV1200) equipped with 20× and 40× objective lenses. Images were optimized and superimposed using Photoshop CS5 Extended (Adobe Systems), and three-dimensional (3D) projections were generated using IMARIS 7.4 (Bitplane). The dots representing accumulated protein aggregates (fluorescence signals of EMB-9::mCherry in C. elegans and those of secreted soluble GFP (ssGFP) in HEK293 cells) were counted by using the MeasurementPro function of Imaris. All images were acquired by using identical settings and counted under the same threshold conditions.

Animals were grown at 20°C and prepared for TEM as previously described (Hall et al., 2012). We prepared cross-sections of the pharyngeal region of L3-stage animals. Fig. 2D shows the region of the cross-sections that was observed. Samples were observed by using a JEM1010 electron microscope (JEOL, Tokyo, Japan).

Mixed populations of transgenic worms expressing emb-9::mCherry were collected and then disrupted with glass beads by using Micro Smash MS-100 (Tomy) in 100 mM Tris–HCl pH 7.4, 150 mM NaCl, protease inhibitor cocktail (Roche) and 1% (w/v) Triton X-100. After disruption, the lysates were rotated at 4°C for 30 min, followed by centrifugation at 17,400 g for 20 min at 4°C. The supernatants were boiled in SDS–PAGE sample buffer with or without 100 mM DTT, separated by 4-12% SDS–PAGE and then transferred to nitrocellulose membranes. After blocking with PBS (supplemented with 3% skimmed milk) for 2 h at room temperature, membranes were immunoblotted with 1:1000 diluted rabbit anti-RFP (which reacts with mCherry, MBL) for 2 h at room temperature. The filter was washed three times with Tris-buffered saline containing 0.05% Tween 20 for 10 min each and then incubated with 1:1000 diluted horseradish-peroxidase-conjugated anti-rabbit IgG (Santa Cruz) at room temperature for 1 h.

The pign-1 deletion (os162) was generated from the EMB-9::mCherry transgenic line using the CRISPR/Cas9 method (Dickinson et al., 2013). We used the Streptococcus pyogenes cas9 gene with its codons optimized for C. elegans and the single-guide RNA (sgRNA) expression plasmid pDD162 (Addgene, Cambridge, MA). The pDD162 (Peft-3::Cas9+Empty sgRNA) was a gift from Bob Goldstein (Addgene plasmid # 47549). The 5′-ACGGGTCCATATCTTGTGG-3′ sequence was used to identify the sgRNA target sequence. We selected candidate animals for each mutation based on abnormal localization of EMB-9::mCherry and verified the mutations by determining the nucleotide sequences of the all exons of the pign-1 genomic locus.

To measure the phosphoethanolamine transferase activity of C. elegans pign-1, pME 3HA::PIGN-1 and pME 3HA::PIGN-1(os156:R78K) were constructed by replacing the human PIGN cDNA with the corresponding cDNAs amplified by PCR from myo-3p::gfp::pign-1 and myo-3p::gfp::pign-1(R78K), respectively, in the pME 3HA::PIGN plasmid in which PIGN cDNA comprising three hemagglutinin (HA) tags at the N terminus is driven by the simian virus 40 (SV40)-derived SRα promoter (Ohba et al., 2014). To generate site-specific mutations in the PIGN conserved domains for pME 3HA-human PIGN, the QuikChange Site-Directed Mutagenesis kit (Agilent) was used with primers containing the appropriate nucleotide changes (H98A, H218A, and H263A). PIGN-knockout HEK293 cells were transiently transfected with wild-type and various mutants of PIGN-containing plasmids, and flow cytometric analyses of GPI-AP expressed on the cell surface were performed as previously described (Ohba et al., 2014).

HEK293 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum and antibiotics. To express secreted soluble GFP (ssGFP) in HEK293 cells, the signal sequence of CD59 (1-30 aa) was PCR-amplified and ligated to the CMV promoter-driven expression vector of EGFP using the In-Fusion technique to produce CMVp::ssGFP, which was then transfected into wild-type and PIGN-knockout HEK293 cells by using Lipofectamine (Invitrogen). Three clones each for wild-type and PIGN-knockout HEK293 cells, which stably expressed ssGFP, were selected in a medium containing 500 µg/ml of G418 (Sigma-Aldrich). The results of only one clone each for wild-type and PIGN-knockout cells are shown in Fig. 4, and equivalent results were obtained using the other clones. The fluorescence of ssGFP expressed in HEK293 cells was captured by using an Olympus IX71 microscope equipped with a 40× objective lens. To analyze the rescue of aggregated ssGFP, PIGN-knockout HEK293 cells were transiently transfected with the SRα-promoter-driven expression plasmid pME 3HA::PIGN (Ohba et al., 2014). The localization of ssGFP was determined 3 days after transfection. To express ER luminal marker and ERES marker, PIGN-knockout HEK293 cells, which stably express ssGFP, were transiently transfected with CMVp::mCherry-ER-3 or CMVp::Sec31A-TagRFP-T by using Lipofectamine. mCherry-ER-3 was a gift from Michael Davidson (Addgene plasmid # 55041). These localizations were analyzed using a confocal laser scanning microscope (Olympus FV1200).

We thank Dr Yuji Kohara (National Institute of Genetics) for yk573C9; Dr Kazuya Nomura (Kyusyu University) for the WRK-1::GFP and mCherry::SP12 vectors; Dr Kiyoji Nishiwaki (Kwansei Gakuin University) for the transgenic lines MIG-17::venus, mig-23::GFP, and myo-3p::TRAM::YFP; Dr Sawako Yoshina and Dr Shohei Mitani (Tokyo Women’s Medical University) for the venus::SP12 and aman-2::venus vectors; Dr Joshua M. Kaplan (Massachusetts General Hospital) for pHluroin vector; Dr Hideki Shibata (Nagoya university) for CMVp::Sec31A-TagRFP-T vector; the Caenorhabditis Genetic Center (funded by the NIH Office of Research Infrastructure Programs; P40 OD010440), the C. elegans Gene Knockout Consortium and National Bioresource Project for reagents and strains. We thank Dr Masato Kanemaki (National Institute of genetics) for advice on imaging and the luciferase assay. We thank Akane Oishi for technical assistant for Transmission electron microscopy. We thank Dr Akira Nozawa (Ehime Univeristy) for technical advice. We thank Dr David R. Sherwood (Duke University) for comments on the manuscript.