Derlin family members participate in the retrotranslocation of endoplasmic reticulum (ER) lumen proteins to the cytosol for ER-associated degradation (ERAD); however, the proteins facilitating this retrotranslocation remain to be explored. Using CRISPR library screening, we have found that derlin-2 and surfeit locus protein 4 (Surf4) are candidates to facilitate degradation of cyclooxygenase-2 (COX-2, also known as PTGS2). Our results show that derlin-2 acts upstream of derlin-1 and that Surf4 acts downstream of derlin-2 and derlin-1 to facilitate COX-2 degradation. Knockdown of derlin-2 or Surf4 impedes the ubiquitylation of COX-2 and the interaction of COX-2 with caveolin-1 (Cav-1) and p97 (also known as VCP) in the cytosol. Additionally, COX-2 degradation is N-glycosylation dependent. Although derlin-2 facilitates degradation of N-glycosylated COX-2, the interaction between derlin-2 and COX-2 is independent of COX-2 N-glycosylation. Derlin-1, Surf4 and p97 preferentially interact with non-glycosylated COX-2, whereas Cav-1 preferentially interacts with N-glycosylated COX-2, regardless of the N-glycosylation pattern. Collectively, our results reveal that Surf4 collaborates with derlin-2 and derlin-1 to mediate COX-2 translocation from the ER lumen to the cytosol. The derlin-2–derlin-1–Surf4–Cav-1 machinery might represent a unique pathway to accelerate COX-2 degradation in ERAD.

In the endoplasmic reticulum (ER)-associated degradation (ERAD) pathway, retrotranslocation of ER proteins from the lumen to the cytosol for proteasomal degradation plays a crucial role in maintaining ER protein homeostasis (Krshnan et al., 2022). The derlin family proteins, homologs of yeast Der1 and Dfm1, are multi-spanning membrane proteins that are candidates for the retrotranslocation channels facilitating initial retrotranslocation of ER proteins from the ER lumen to the cytosol for ubiquitylation and degradation (Chang et al., 2004; Lilley and Ploegh, 2004; Neal et al., 2018; Ye et al., 2001). Derlins recruit caveolin-1 (Cav-1) and the p97 (also known as VCP) complex to facilitate ER protein extraction and proteasomal degradation (Chen et al., 2013; Katiyar et al., 2005; Lilley and Ploegh, 2005; Oda et al., 2006; Olzmann et al., 2013; Ye et al., 2005). Various factors participate in ER protein degradation and might interact with each other to form a complex system in ERAD. However, specific factors are required to facilitate the degradation of a particular protein. (Krshnan et al., 2022). Studies suggest that derlin-1 and derlin-2 function independently in mediating ER protein degradation. Derlin-1 mediates US11-dependent dislocation and degradation of major histocompatibility complex class I heavy chains and facilitates the retrotranslocation of cholera toxin (Bernardi et al., 2008; Chang et al., 2004; Lilley and Ploegh, 2004, 2005). Derlin-2, but not derlin-1 or derlin-3, facilitates HRD-1 (SYVN1)-mediated sonic hedgehog retrotranslocation (Huang et al., 2013). Although derlin-1 interacts with derlin-2 in mediating protein degradation (Lilley and Ploegh, 2005), whether and how derlin-1 and derlin-2 collaborate in mediating protein retrotranslocation in ERAD, and whether there are other proteins that participate in the derlin-mediated retrotranslocation of ER lumen proteins, remains to be explored.

Surfeit locus protein 4 (Surf4), an ER membrane protein with five putative transmembrane domains, is the homolog of yeast Erv29 and is expressed ubiquitously in most cell types (Belden and Barlowe, 2001; Reeves and Fried, 1995). As a cargo receptor, Surf4 contains a protein-binding motif that recognizes ER proteins and recruits them into COPII-coated vesicles to facilitate their secretion and export (Belden and Barlowe, 2001; Reeves and Fried, 1995; Saegusa et al., 2018; Yin et al., 2018). Furthermore, Surf4 governs ER–Golgi structural integrity and facilitates protein transportation between the ER and Golgi (Mitrovic et al., 2008). However, whether Surf4 participates in the translocation of ER proteins to the cytosol for proteasomal degradation remains unknown.

Cyclooxygenase (COX) is the rate-limiting enzyme in the production of eicosanoids, which regulate the homeostasis of various physiological functions (Smith et al., 2011). In humans, two functional COX isozymes have been identified: cyclooxygenase-1 (COX-1, also known as PTGS1) and cyclooxygenase-2 (COX-2, also known as PTGS2). COX-1 is relatively stable and is constitutively expressed in most mammalian cells, whereas levels of COX-2 are strictly regulated and can be induced to a high level or decreased to a barely detectable level in a matter of hours (Chen et al., 2010, 2013; Mbonye et al., 2006; Smith et al., 2011; Yazaki et al., 2012). Upregulated expression of COX-2 has been reported in a number of malignant tumors, and inhibition of COX-2 expression or activity can suppress tumor growth (Hashemi Goradel et al., 2019; Pu et al., 2021). COX-2 is an N-glycosylated, ER lumen-resident homodimeric enzyme that interacts with the ER membrane via a hydrophobic region (Spencer et al., 1999). COX-2 contains four N-glycosylation sites (Mbonye et al., 2006); however, the roles of these N-glycosylation modifications in protein interactions and COX-2 degradation through ERAD remain to be examined.

In this study, we used a CRISPR–Cas9 screening approach to identify proteins potentially involved in COX-2 degradation. Our findings indicate that Surf4 and derlin-2 are candidates in mediating COX-2 degradation. We examined the interplay of derlin-2, derlin-1 and Surf4, as well as the role of COX-2 N-glycosylation, in COX-2 translocation and degradation. Our results show that Surf4 collaborates with derlin-2 and derlin-1 to mediate COX-2 retrotranslocation in ERAD.

Surf4 mediates COX-2 protein degradation

To screen for genes that are potentially involved in COX-2 degradation, we constructed a HEK293 cell line overexpressing superfolder green fluorescent protein (sfGFP)-tagged COX-2 (sfGFP–COX-2) and Cas9, and used it for CRISPR-based single guide RNA (sgRNA) library screening via lentivirus infection (Fig. S1). We identified Surf4 and derlin-2 as candidates for having a role in COX-2 degradation (Fig. S2A). We first examined whether Surf4 mediates COX-2 degradation. A549 cells were infected with lentivirus carrying shRNAs to suppress Surf4 mRNA and protein expression (Fig. 1A,B). COX-2 mRNA levels were not altered, whereas COX-2 protein levels were upregulated in Surf4-knockdown cells (Fig. 1A,B). Furthermore, transfection of Surf4 siRNA upregulated COX-2 levels in interleukin (IL)-1β-induced human fibroblast HS-68 cells (Fig. 1C). Moreover, lentiviral transduction of Surf4 shRNA increased the half-life of endogenous COX-2 from less than 1.5 h to ∼3 h in A549 cells (Fig. 1D). In addition, HA-tagged Surf4 (HA–Surf4) co-precipitated with endogenous COX-2 in HA–Surf4-expressing H1299 cells (Fig. 1E). Association of Surf4 and COX-2 was detected in reciprocal immunoprecipitation assays using H1299 cells co-transfected with COX-2–Flag and HA–Surf4 (Fig. 1F,G). Furthermore, confocal microscopy revealed partial colocalization of COX-2 and Surf4 in the perinuclear and ER regions in both HS-68 and A549 cells, with Manders' overlap coefficients (MOCs) of 0.98 and 0.85, respectively (Fig. 1H). These results suggest that Surf4 facilitates COX-2 degradation.

Fig. 1.

Surf4 interacts with COX-2 and mediates COX-2 degradation. (A,B) A549 cells were infected with lentivirus carrying shRNA targeting Surf4 (S266 and S412; Table S1) or a control (Con) vector. The mRNA levels (A) and protein levels (B) of COX-2 and Surf4 were determined by RT-PCR and western blotting, respectively. Quantification of western blotting is shown on the right. KD, knockdown. (C) Western blot analysis of COX-2 levels in fibroblast HS-68 cells that were untransfected (−) or transfected with either siRNA targeting Surf4 or control siRNA. Quantification is shown on the right. (D) Western blot analysis and quantification of COX-2 levels in A549 cells infected with lentivirus carrying Surf4 shRNAs (S266 and S412) or vector control and then treated with cycloheximide (CHX; 20 µg/ml) for 0 to 3 h, as indicated. Relative intensity levels of COX-2 protein were quantified and are shown on the right (n=3 per group). (E) H1299 cells were transfected to express COX-2 and HA–Surf4 for an immunoprecipitation (IP) assay with anti-HA antibody. (F,G) H1299 cells were transfected to express COX-2–Flag and HA–Surf4 for reciprocal immunoprecipitation assays with anti-Flag (F) and anti-HA (G) antibodies. (H) Confocal microscopic analysis of endogenous COX-2 and Surf4 localization in HS-68 and A549 cells. Cells were double-stained with anti-COX-2 (green, FITC) and anti-Surf4 (red, rhodamine) antibodies. MOC was 0.98 and 0.85 in HS-68 and A549 cells, respectively. The fluorescence intensity profiles across the transect marked by a red line for both green and red channels are shown below the images. Blue: nucleus (DAPI). Scale bars: 10 µm. In A–E, β-actin is shown as a loading control. In B–G, molecular masses are indicated in kDa. In E–G, input lanes show 10% of the total lysate. Images in A and E–H are representative of three experiments. Quantitative data are presented as mean±s.d. (n=3–4) and are normalized to the control-shRNA group or untreated control group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Fig. 1.

Surf4 interacts with COX-2 and mediates COX-2 degradation. (A,B) A549 cells were infected with lentivirus carrying shRNA targeting Surf4 (S266 and S412; Table S1) or a control (Con) vector. The mRNA levels (A) and protein levels (B) of COX-2 and Surf4 were determined by RT-PCR and western blotting, respectively. Quantification of western blotting is shown on the right. KD, knockdown. (C) Western blot analysis of COX-2 levels in fibroblast HS-68 cells that were untransfected (−) or transfected with either siRNA targeting Surf4 or control siRNA. Quantification is shown on the right. (D) Western blot analysis and quantification of COX-2 levels in A549 cells infected with lentivirus carrying Surf4 shRNAs (S266 and S412) or vector control and then treated with cycloheximide (CHX; 20 µg/ml) for 0 to 3 h, as indicated. Relative intensity levels of COX-2 protein were quantified and are shown on the right (n=3 per group). (E) H1299 cells were transfected to express COX-2 and HA–Surf4 for an immunoprecipitation (IP) assay with anti-HA antibody. (F,G) H1299 cells were transfected to express COX-2–Flag and HA–Surf4 for reciprocal immunoprecipitation assays with anti-Flag (F) and anti-HA (G) antibodies. (H) Confocal microscopic analysis of endogenous COX-2 and Surf4 localization in HS-68 and A549 cells. Cells were double-stained with anti-COX-2 (green, FITC) and anti-Surf4 (red, rhodamine) antibodies. MOC was 0.98 and 0.85 in HS-68 and A549 cells, respectively. The fluorescence intensity profiles across the transect marked by a red line for both green and red channels are shown below the images. Blue: nucleus (DAPI). Scale bars: 10 µm. In A–E, β-actin is shown as a loading control. In B–G, molecular masses are indicated in kDa. In E–G, input lanes show 10% of the total lysate. Images in A and E–H are representative of three experiments. Quantitative data are presented as mean±s.d. (n=3–4) and are normalized to the control-shRNA group or untreated control group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Derlin-2 functions upstream of derlin-1 and Cav-1 in COX-2 degradation

Next, we examined the role of derlin-2 in COX-2 degradation. Suppression of derlin-2 expression by siRNA upregulated COX-2 levels in A549 cells and IL-1β-induced HS-68 cells. (Fig. 2A,B). Also, suppression of derlin-2 increased the half-life of endogenous COX-2 from ∼1.5 h to more than 3 h in A549 cells (Fig. 2C) and of exogenously expressed COX-2 from ∼4.5 h to more than 8 h in H1299 cells (Fig. 2D). Moreover, suppression of derlin-2 reduced the ubiquitylation of exogenously expressed COX-2 in H1299 cells (Fig. 2E). Additionally, confocal microscopy revealed partial colocalization of COX-2 and derlin-2 in the perinuclear and ER regions in A549 cells, with a MOC of 0.92 (Fig. 2F). These results suggest that derlin-2 facilitates COX-2 degradation.

Fig. 2.

Derlin-2 mediates COX-2 degradation. (A) Western blot analysis and quantification of COX-2 levels in A549 cells transfected with control (Con) siRNA or siRNA targeting derlin-2 (Der-2). (B) Western blot analysis and quantification of COX-2 levels in HS-68 cells transfected with control siRNA or siRNA targeting derlin-2 for 24 h, with or without IL-1β treatment as indicated. (C) Western blot analysis and quantification of COX-2 levels in A549 cells transfected with control siRNA or siRNA targeting derlin-2 for 24 h then treated with cycloheximide (CHX) (20 µg/ml) for 0 to 3 h (n=4 per group). (D) Western blot analysis and quantification of COX-2 levels in H1299 cells transfected to express COX-2 and either control siRNA or siRNA targeting derlin-2 for 24 h then treated with cycloheximide for 0 to 8 h (n=4 per group). (E) HEK293 cells were transfected to express COX-2 and His-tagged ubiquitin (His-Ub) plus control or derlin-2 siRNA for 24 h, and then treated with MG132 (20 mM) for 6 h. His-conjugated proteins were isolated for an in vivo ubiquitylation assay, then underwent western blot analysis (IB). (F) Confocal microscopic analysis of COX-2 and derlin-2 localization in A549 cells double-stained with anti-COX-2 (red) and anti-derlin-2 (green) antibodies. MOC=0.92. The fluorescence intensity profile across the red line for both green and red channels is shown below the images. Blue: nucleus (DAPI). Scale bars: 10 µm. In A–E, β-actin is shown as a loading control and molecular masses are indicated in kDa. Images in E and F are representative of three experiments. Quantitative data are presented as mean±s.d. (n=3–4) and are normalized to the untreated control group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Fig. 2.

Derlin-2 mediates COX-2 degradation. (A) Western blot analysis and quantification of COX-2 levels in A549 cells transfected with control (Con) siRNA or siRNA targeting derlin-2 (Der-2). (B) Western blot analysis and quantification of COX-2 levels in HS-68 cells transfected with control siRNA or siRNA targeting derlin-2 for 24 h, with or without IL-1β treatment as indicated. (C) Western blot analysis and quantification of COX-2 levels in A549 cells transfected with control siRNA or siRNA targeting derlin-2 for 24 h then treated with cycloheximide (CHX) (20 µg/ml) for 0 to 3 h (n=4 per group). (D) Western blot analysis and quantification of COX-2 levels in H1299 cells transfected to express COX-2 and either control siRNA or siRNA targeting derlin-2 for 24 h then treated with cycloheximide for 0 to 8 h (n=4 per group). (E) HEK293 cells were transfected to express COX-2 and His-tagged ubiquitin (His-Ub) plus control or derlin-2 siRNA for 24 h, and then treated with MG132 (20 mM) for 6 h. His-conjugated proteins were isolated for an in vivo ubiquitylation assay, then underwent western blot analysis (IB). (F) Confocal microscopic analysis of COX-2 and derlin-2 localization in A549 cells double-stained with anti-COX-2 (red) and anti-derlin-2 (green) antibodies. MOC=0.92. The fluorescence intensity profile across the red line for both green and red channels is shown below the images. Blue: nucleus (DAPI). Scale bars: 10 µm. In A–E, β-actin is shown as a loading control and molecular masses are indicated in kDa. Images in E and F are representative of three experiments. Quantitative data are presented as mean±s.d. (n=3–4) and are normalized to the untreated control group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Derlin-1 and Cav-1 have been reported to mediate COX-2 degradation through ERAD (Chen et al., 2013). Hence, we examined the interplay of derlin-1 and derlin-2 in COX-2 degradation. Suppression of derlin-1 or derlin-2 expression by siRNA led to an increase in the level of exogenously expressed COX-2 in H1299 cells (Fig. 3A). However, there was no synergistic effect of derlin-1 and derlin-2 in COX-2 degradation. Indeed, similar COX-2 levels were detected in cells transfected with siRNA targeting derlin-1, derlin-2, or derlin-1 and derlin-2 combined. These results reveal that derlin-1 and derlin-2 are in the same pathway in mediating COX-2 degradation. Additionally, confocal microscopy revealed partial colocalization of derlin-2 and HA–derlin-1 in the perinuclear and ER regions, with a MOC of 0.95 (Fig. 3B).

Fig. 3.

Derlin-2 functions upstream of derlin-1 and Cav-1 in mediating COX-2 degradation. (A) H1299 cells were transfected to express COX-2 alone (−) or with either control (Con) siRNA, derlin-1 (Der-1) siRNA, derlin-2 (Der-2) siRNA, or both derlin-1 and derlin-2 siRNAs (Der-1+Der-2) for western blot analysis. Relative COX-2 levels were quantified and normalized to that of the untreated control. Quantitative data are presented as mean±s.d. (n=4) normalized to the untreated control group. *P<0.05 (one-way ANOVA followed by post-hoc Bonferroni test). (B) Confocal microscopic analysis of A549 cells double-stained with anti-HA (derlin-1, red) and anti-derlin-2 (green) antibodies. MOC=0.95. Blue: nucleus (DAPI). Scale bars: 10 µm. The fluorescence intensity profile across the red line for both green and red channels is shown on the right. (C–E) HEK293 cells were transfected to express HA–derlin-1 (C) or HA–derlin-2 (D and E) along with control siRNA or siRNA targeting derlin-2 (C), derlin-1 (D) or Cav-1 (E) for 24 h. Cell lysates were immunoprecipitated (IP) with anti-HA antibody for western blot analysis with the indicated antibodies. In A and C–E, β-actin is shown as a loading control and molecular masses are indicated in kDa. In C–E, input lanes show 10% of the total lysate. Images in B–E are representative of three experiments.

Fig. 3.

Derlin-2 functions upstream of derlin-1 and Cav-1 in mediating COX-2 degradation. (A) H1299 cells were transfected to express COX-2 alone (−) or with either control (Con) siRNA, derlin-1 (Der-1) siRNA, derlin-2 (Der-2) siRNA, or both derlin-1 and derlin-2 siRNAs (Der-1+Der-2) for western blot analysis. Relative COX-2 levels were quantified and normalized to that of the untreated control. Quantitative data are presented as mean±s.d. (n=4) normalized to the untreated control group. *P<0.05 (one-way ANOVA followed by post-hoc Bonferroni test). (B) Confocal microscopic analysis of A549 cells double-stained with anti-HA (derlin-1, red) and anti-derlin-2 (green) antibodies. MOC=0.95. Blue: nucleus (DAPI). Scale bars: 10 µm. The fluorescence intensity profile across the red line for both green and red channels is shown on the right. (C–E) HEK293 cells were transfected to express HA–derlin-1 (C) or HA–derlin-2 (D and E) along with control siRNA or siRNA targeting derlin-2 (C), derlin-1 (D) or Cav-1 (E) for 24 h. Cell lysates were immunoprecipitated (IP) with anti-HA antibody for western blot analysis with the indicated antibodies. In A and C–E, β-actin is shown as a loading control and molecular masses are indicated in kDa. In C–E, input lanes show 10% of the total lysate. Images in B–E are representative of three experiments.

We further examined the interaction and interplay of derlin-2, derlin-1 and Cav-1 in COX-2 degradation using immunoprecipitation assays. To minimize potential variations in the immunoprecipitation assays, we adjusted the amount of plasmid used for transfection to achieve similar COX-2 expression levels. Derlin-2 knockdown by siRNA decreased the interaction of derlin-1 with COX-2 but had no significant effect on the interaction of derlin-1 with Cav-1 and p97 in HEK293 cells (Fig. 3C), which suggests that derlin-2 facilitates the interaction of derlin-1 with COX-2. Moreover, derlin-1 knockdown by siRNA increased the interaction between derlin-2 and COX-2, and decreased the interaction of derlin-2 with Cav-1 and p97 (Fig. 3D), suggesting that derlin-1 acts downstream of derlin-2 in COX-2 degradation. Additionally, Cav-1 knockdown using siRNA decreased the interaction between derlin-2 and p97 but had no significant effect on interactions between derlin-2, derlin-1 and COX-2 (Fig. 3E), which suggests that Cav-1 facilitates the interaction of derlin-2 and derlin-1 with p97. The interplay of derlin-2, derlin-1 and Cav-1 in COX-2 degradation was also examined without adjustment of COX-2 levels, with similar results (Fig. S3A–C). Since Cav-1 is located on the cytoplasmic face of the ER membrane (Parton and Simons, 2007) and p97 is in the cytoplasm, these results suggest that COX-2 degradation might be sequentially mediated through the derlin-2–derlin-1–Cav-1–p97 pathway.

Surf4 interacts with derlin-1 and derlin-2 to facilitate COX-2 translocation to the cytosol for ubiquitylation and interaction with the p97–Ufd1 complex

We then investigated the role of Surf4 in derlin-2- and derlin-1-mediated COX-2 degradation. HA–Surf4 co-precipitated with derlin-1, derlin-2 and Cav-1 in HA–Surf4-expressing A549 cells (Fig. 4A), suggesting that Surf4 interacts with derlin-2, derlin-1 and Cav-1. Additionally, confocal microscopy revealed partial colocalization of derlin-1, derlin-2, COX-2 and GFP–Surf4 in the perinuclear and ER regions, with MOCs of 0.98 (COX-2 versus derlin-1), 0.97 (GFP–Surf4 versus derlin-1), and 0.96 (GFP–Surf4 versus derlin-2), respectively (Fig. 4B).

Fig. 4.

Surf4 interacts with derlin-1 and derlin-2 to facilitate COX-2 ubiquitylation and the interaction of COX-2 with Cav-1 and p97. (A) A549 cells were transfected to express HA–Surf4 and then immunoprecipitated (IP) with anti-HA antibody for western blot analysis with the indicated antibodies. Der-1, derlin-1; Der-2, derlin-2. (B) Confocal microscopic analysis of A549 cells transfected to express GFP–Surf4 (green) and then double-stained with anti-COX-2 (red) and either anti-derlin-1 (pink, left) or anti-derlin-2 (pink, right) antibodies, with the MOC shown in the figure. Blue: nucleus (DAPI). The fluorescence intensity profiles across the red line for each channel are shown on the right. Scale bars: 10 µm. (C–G) A549 cells (C–F) or HEK293 cells (G) were transduced with lentivirus carrying control (Con) shRNA or shRNA targeting Surf4 (S266). Cells were then transfected to express COX-2–Flag (C), HA–p97 (D), or HA–p97 and Flag–Ufd1 (E,F) for immunoprecipitation assays with anti-Flag or anti-HA antibody, as indicated. Lanes marked with an asterisk show a cell lysate mixture of control and assay samples used for IgG-binding controls. (G) Cells were transfected to express COX-2 and His-tagged ubiquitin (His-Ubi) for 24 h then treated with MG132 for 6 h. His-conjugated proteins were isolated for an in vivo ubiquitylation assay, then underwent western blot analysis (IB). In A, C and D, β-actin is shown as a loading control. In A and C–G, molecular masses are indicated in kDa. In A and C–F, input lanes show 5% of the total lysate. Images in A, B and G are representative of three experiments. Quantitative data showing immunoprecipitated protein levels are presented as mean±s.d. (n=3) normalized to the control-shRNA group. *P<0.05; ***P<0.001 (unpaired, two-tailed Student's t-test ).

Fig. 4.

Surf4 interacts with derlin-1 and derlin-2 to facilitate COX-2 ubiquitylation and the interaction of COX-2 with Cav-1 and p97. (A) A549 cells were transfected to express HA–Surf4 and then immunoprecipitated (IP) with anti-HA antibody for western blot analysis with the indicated antibodies. Der-1, derlin-1; Der-2, derlin-2. (B) Confocal microscopic analysis of A549 cells transfected to express GFP–Surf4 (green) and then double-stained with anti-COX-2 (red) and either anti-derlin-1 (pink, left) or anti-derlin-2 (pink, right) antibodies, with the MOC shown in the figure. Blue: nucleus (DAPI). The fluorescence intensity profiles across the red line for each channel are shown on the right. Scale bars: 10 µm. (C–G) A549 cells (C–F) or HEK293 cells (G) were transduced with lentivirus carrying control (Con) shRNA or shRNA targeting Surf4 (S266). Cells were then transfected to express COX-2–Flag (C), HA–p97 (D), or HA–p97 and Flag–Ufd1 (E,F) for immunoprecipitation assays with anti-Flag or anti-HA antibody, as indicated. Lanes marked with an asterisk show a cell lysate mixture of control and assay samples used for IgG-binding controls. (G) Cells were transfected to express COX-2 and His-tagged ubiquitin (His-Ubi) for 24 h then treated with MG132 for 6 h. His-conjugated proteins were isolated for an in vivo ubiquitylation assay, then underwent western blot analysis (IB). In A, C and D, β-actin is shown as a loading control. In A and C–G, molecular masses are indicated in kDa. In A and C–F, input lanes show 5% of the total lysate. Images in A, B and G are representative of three experiments. Quantitative data showing immunoprecipitated protein levels are presented as mean±s.d. (n=3) normalized to the control-shRNA group. *P<0.05; ***P<0.001 (unpaired, two-tailed Student's t-test ).

The p97–Ufd1 complex interacts with ubiquitylated proteins to facilitate retrotranslocation of ER lumen proteins to the cytosol for proteasomal degradation (Lilley and Ploegh, 2005; Ye et al., 2005). We examined whether Surf4 affects the interaction of COX-2 with p97 and Ufd1. Suppression of Surf4 expression reduced the interaction of COX-2 with p97 (Fig. 4C,D). However, suppression of Surf4 had no significant effect on the interaction of p97 with Ufd1, as assayed using reciprocal immunoprecipitation assays (Fig. 4E,F). Moreover, suppression of Surf4 expression led to a decrease in the level of ubiquitylated COX-2 (Fig. 4G). Despite the absence of data supporting the physical localization of COX-2, these results imply that Surf4 interacts with derlin-1 and derlin-2 to facilitate COX-2 translocation to the cytosol for ubiquitylation and to interact with the p97–Ufd1 complex.

Surf4 functions downstream of derlin-2 and derlin-1 but upstream of Cav-1 in COX-2 degradation

Derlin-1 and derlin-2 participate in the retrotranslocation of ER proteins from the ER to the cytosol (Chang et al., 2004; Lilley and Ploegh, 2005), and Surf4 has been suggested to mediate ER protein export (Saegusa et al., 2018). We therefore examined the role of Surf4 in derlin-2-, derlin-1- and Cav-1-mediated COX-2 translocation and degradation. Surf4 knockdown by siRNA decreased the interaction of Cav-1 with COX-2 in A549 cells (Fig. 5A,B). Cav-1 knockdown did not affect the interaction between Surf4 and COX-2 (Fig. 5C,D). These results suggest that Cav-1 acts downstream of Surf4 in the COX-2 degradation pathway. Moreover, suppression of Surf4 by siRNA had no effects on either the interaction between derlin-1 and COX-2 (Fig. 5E,F), or the interaction between derlin-2 and COX-2 (Fig. 5G,H). However, knockdown of either derlin-1 or derlin-2 by siRNA decreased the interaction of Surf4 with COX-2 (Fig. 5I–L). Taken together, these data suggest that Surf4 acts downstream of derlin-2 and derlin-1 to mediate COX-2 translocation from the ER lumen to the cytosol for ubiquitylation and to interact with Cav-1 and the p97 complex.

Fig. 5.

Surf 4 functions downstream of derlin-2 and derlin-1 but upstream of Cav-1 in mediating COX-2 degradation. (A–H) A549 cells were transfected with control (Con) siRNA or siRNA targeting either Surf4 or Cav-1, as indicated. Cells were then transfected to express COX-2–Flag and HA–Cav-1 (A,B), COX-2–Flag and HA–Surf4 (C,D), COX-2–Flag and HA–derlin-1 (Der-1; E,F) or COX-2–Flag and HA–derlin-2 (Der-2; G,H) for immunoprecipitation (IP) assays with anti-HA or anti-Flag antibodies, as indicated. (I–L) A549 cells were transfected with control siRNA or siRNA targeting derlin-1 (I,J) or derlin-2 (K,L), as indicated. Cells were then transfected to express COX-2–Flag and HA–Surf4 for immunoprecipitation assays with anti-HA (I,K) or anti-Flag (J,L) antibodies. β-actin is shown as a loading control, and molecular masses are indicated in kDa. Input lanes show 10% of the total lysate. Quantitative data showing immunoprecipitated protein levels are presented as mean±s.d. (n=4) and are normalized to the control-siRNA group. **P<0.01; ***P<0.001 (unpaired, two-tailed Student's t-test).

Fig. 5.

Surf 4 functions downstream of derlin-2 and derlin-1 but upstream of Cav-1 in mediating COX-2 degradation. (A–H) A549 cells were transfected with control (Con) siRNA or siRNA targeting either Surf4 or Cav-1, as indicated. Cells were then transfected to express COX-2–Flag and HA–Cav-1 (A,B), COX-2–Flag and HA–Surf4 (C,D), COX-2–Flag and HA–derlin-1 (Der-1; E,F) or COX-2–Flag and HA–derlin-2 (Der-2; G,H) for immunoprecipitation (IP) assays with anti-HA or anti-Flag antibodies, as indicated. (I–L) A549 cells were transfected with control siRNA or siRNA targeting derlin-1 (I,J) or derlin-2 (K,L), as indicated. Cells were then transfected to express COX-2–Flag and HA–Surf4 for immunoprecipitation assays with anti-HA (I,K) or anti-Flag (J,L) antibodies. β-actin is shown as a loading control, and molecular masses are indicated in kDa. Input lanes show 10% of the total lysate. Quantitative data showing immunoprecipitated protein levels are presented as mean±s.d. (n=4) and are normalized to the control-siRNA group. **P<0.01; ***P<0.001 (unpaired, two-tailed Student's t-test).

COX-2 degradation is N-glycosylation dependent

In our CRISPR–Cas9-based screen, we also identified a number of glycosylation-related genes (Fig. S2), suggesting that glycosylation might play an important role in COX-2 degradation. Protein N-glycosylation is involved in derlin-mediated protein degradation (Oda et al., 2006). However, how N-glycosylation of COX-2 affects its interaction with ERAD components remains to be examined. COX-2 features four N-glycosylation sites (Fig. 6A) (Mbonye et al., 2006). We therefore examined the role of N-glycosylation in COX-2 degradation through ERAD. Constructs for expression of mutant COX-2 with disrupted N-glycosylation at the first to third sites (N53, N130 and N396), first to fourth sites (N53, N130, N396 and N580) or fourth site alone (N580) were created and designated as COX-2-m1–3, COX-2-m1–4 and COX-2-m4, respectively (Fig. 6B; Fig. S4A). COX-2-m1–3, COX-2-m1–4 and COX-2-m4 showed different molecular mass shifts in western blotting (Fig. S4B). After endoglycosidase H (Endo H) or peptide:N-glycosidase F (PNGase F) treatment, COX-2-m1–3, COX-2-m1–4 and COX-2-m4 were observed to have a molecular mass similar to that of treated wild-type COX-2 and non-treated COX-2-m1–4 (Fig. S4B), which confirmed the N-glycosylation mutations in these constructs. Confocal microscopy showed colocalization of mutated COX-2 with glucose-regulated protein 94 (GRP94, also known as HSP90B1), an ER marker (Fig. S4A), which suggests that mutation of the N-glycosylation residues does not alter the ER residency of COX-2. Moreover, mutation of the N-glycosylation residues significantly increased the half-life of COX-2-m1–3 and COX-2-m4 from ∼3 h to more than 9 h (Fig. 6B). Nevertheless, the half-life of COX-2-m1–4 was further increased (Fig. 6B). Similar results were obtained showing that inhibition of glycosylation by NGI1 (which targets oligosaccharyltransferase) significantly increased the half-life of COX-2 (Fig. S4C,D). Hence, these results reveal that COX-2 degradation is N-glycosylation dependent.

Fig. 6.

Interaction of derlin-2 with COX-2 is independent of COX-2 N-glycosylation. (A) Map of COX-2 N-glycosylation sites (see GenBank accession number NM-000963). (B) H1299 cells were transfected to express COX-2 or the indicated COX-2 mutants (m4, COX-2-m4; m1-3, COX-2-m1–3; m1-4, COX-2-m1–4) for 24 h, then treated with cycloheximide (CHX; 20 µg/ml) for 0 to 9 h. Western blot analysis of COX-2 protein levels and quantification are shown. (C) H1299 cells were transfected to express COX-2 with HA–derlin-2 (Der-2) and then treated with tunicamycin (Tuni; 1 µg/ml) for 24 h before an immunoprecipitation (IP) assay using anti-HA antibody. (D) H1299 cells were transfected to express wild-type (WT) COX-2 or COX-2 mutants, as indicated, along with HA–derlin-2 for an immunoprecipitation assay using anti-HA antibody. (E) H1299 cells were transfected to express wild-type COX-2–Flag or COX-2Δ32aa–Flag (Δ32aa) along with HA–derlin-2 for an immunoprecipitation assay using anti-Flag antibody. (F) HEK293 cells were transfected to express wild-type COX-2, COX-2-m4 or COX-2-m1–4 with control (Con) siRNA or derlin-2 siRNA. Whole-cell lysates were used for western blot analysis. In B–F, β-actin is shown as a loading control, and molecular masses are indicated in kDa. Input lanes in C–E show 5% of the total lysate. Images in C and E are representative of three experiments. Quantitative data are presented as mean±s.d. (n=3–4), normalized to the WT COX-2 group or control-siRNA group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Fig. 6.

Interaction of derlin-2 with COX-2 is independent of COX-2 N-glycosylation. (A) Map of COX-2 N-glycosylation sites (see GenBank accession number NM-000963). (B) H1299 cells were transfected to express COX-2 or the indicated COX-2 mutants (m4, COX-2-m4; m1-3, COX-2-m1–3; m1-4, COX-2-m1–4) for 24 h, then treated with cycloheximide (CHX; 20 µg/ml) for 0 to 9 h. Western blot analysis of COX-2 protein levels and quantification are shown. (C) H1299 cells were transfected to express COX-2 with HA–derlin-2 (Der-2) and then treated with tunicamycin (Tuni; 1 µg/ml) for 24 h before an immunoprecipitation (IP) assay using anti-HA antibody. (D) H1299 cells were transfected to express wild-type (WT) COX-2 or COX-2 mutants, as indicated, along with HA–derlin-2 for an immunoprecipitation assay using anti-HA antibody. (E) H1299 cells were transfected to express wild-type COX-2–Flag or COX-2Δ32aa–Flag (Δ32aa) along with HA–derlin-2 for an immunoprecipitation assay using anti-Flag antibody. (F) HEK293 cells were transfected to express wild-type COX-2, COX-2-m4 or COX-2-m1–4 with control (Con) siRNA or derlin-2 siRNA. Whole-cell lysates were used for western blot analysis. In B–F, β-actin is shown as a loading control, and molecular masses are indicated in kDa. Input lanes in C–E show 5% of the total lysate. Images in C and E are representative of three experiments. Quantitative data are presented as mean±s.d. (n=3–4), normalized to the WT COX-2 group or control-siRNA group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Interaction between COX-2 and derlin-2 is independent of COX-2 N-glycosylation

Derlin-2 facilitates degradation of N-glycosylated proteins (Oda et al., 2006). We examined the role of N-glycosylation of COX-2 in the interaction between COX-2 and derlin-2. Inhibition of N-glycosylation using tunicamycin resulted in increased mobility of COX-2 in western blotting, whereas the levels of COX-2 co-precipitated with derlin-2 were not significantly changed as compared with the levels observed for the untreated control (Fig. 6C). Moreover, similar levels of COX-2, COX-2-m1–3, COX-2-m1–4 and COX-2-m4 were co-precipitated with derlin-2 (Fig. 6D). Furthermore, deletion of the C terminus of COX-2 harboring the fourth N-glycosylation site (COX-2Δ32aa) did not alter the level COX-2 co-precipitated with derlin-2 (Fig. 6E). Therefore, interaction of derlin-2 with COX-2 is independent of COX-2 N-glycosylation.

We next examined the role of N-glycosylation in COX-2 degradation mediated by derlin-2. Similar to previous results, mutation to interrupt N-glycosylation in COX-2 resulted in increased COX-2-m4 and COX-2-m1–4 levels (Fig. 6F; lane 1 versus lane 3 and lane 1 versus lane 5). Suppression of derlin-2 by siRNA further upregulated COX-2-m4 levels but not COX-2-m1–4 levels as compared with levels in the relative control (Fig. 6F; lane 3 versus lane 4 and lane 5 versus lane 6, respectively). Although interaction of derlin-2 with COX-2 is independent of N-glycosylation, these results imply that derlin-2 is involved in the degradation of N-glycosylated COX-2 but not of non-glycosylated COX-2.

Derlin-1, p97 and Surf4 preferentially interact with non-glycosylated COX-2

Next, we investigated the role of COX-2 N-glycosylation in the interaction of COX-2 with derlin-1, p97 and Surf4. Inhibition of N-glycosylation by tunicamycin increased the levels of COX-2 co-precipitated with HA–derlin-1 (Fig. 7A). A substantial level of low- or non-glycosylated COX-2 was co-precipitated with HA–Derlin-1 (Fig. 7A, lower band in lane 1), which was barely detectable in the input sample. Furthermore, mutation at the N-glycosylation residues in COX-2 enhanced the interaction of COX-2 with derlin-1 (Fig. 7B). The levels of COX-2 co-precipitating with derlin-1 were higher when more COX-2 N-glycosylation sites were mutated, with co-precipitation of COX-2-m1-4 greater than that of COX-2-m1–3, which in turn was greater than co-precipitation of COX-2-m4. These results indicate that derlin-1 preferentially interacts with non-glycosylated COX-2. Contrary to the results for suppression of derlin-2, suppression of derlin-1 by siRNA led to an increased level of COX-2-m1–4, but not of COX-2-m4, as compared with the relative control (Fig. 7C; lane 5 versus lane 6 and lane 3 versus lane 4). These results imply that derlin-1 preferentially interacts with non-glycosylated COX-2 and facilitate its degradation. Similar to the results for derlin-1, we found that p97 prefers to bind with non-glycosylated COX-2, in that HA–p97 co-precipitated more COX-2-m1–4 than COX-2 (Fig. 7D, lane 3 versus lane 1). Additionally, we observed that Surf4 preferentially interacts with non-glycosylated COX-2, in that inhibition of N-glycosylation by tunicamycin increased the level of low- or non-glycosylated COX-2 co-precipitated with HA–Surf4 (Fig. 7E). Furthermore, the levels of COX-2 co-precipitated with HA–Surf4 were increased when more COX-2 N-glycosylation sites were mutated (Fig. 7F).

Fig. 7.

Derlin-1, Surf4 and p97 preferentially interact with non-glycosylated COX-2. (A) HEK293 cells were transfected to express HA–derlin-1 (Der-1) along with COX-2. Cells were treated with tunicamycin (Tuni; 1 µg/ml) for 24 h. Cell lysates were immunoprecipitated (IP) with HA antibodies for western blot analysis. (B) HEK293 cells were transfected to express HA–derlin-1 (Der-1) along with wild-type (WT) COX-2 or the indicated COX-2 mutants (m4, COX-2-m4; m1-3, COX-2-m1–3; m1-4, COX-2-m1–4). Cell lysates were immunoprecipitated with HA antibodies for western blot analysis. Quantification of immunoprecipitated COX-2 is shown on the right. (C) HEK293 cells were transfected to express wild-type COX-2, COX-2-m4 or COX-2-m1–4 along with control siRNA or siRNA targeting derlin-1. Whole-cell lysates were used for western blot analysis. Quantification of COX2 is shown on the right. (D) HEK293 cells were transfected to express HA–p97 along with wild-type COX-2 or various levels of COX-2-m1–4 for an immunoprecipitation assay with anti-HA antibody. Cells were transfected with increasing amounts of COX-2-m1–4 plasmid, as indicated by the triangle symbol. (E,F) HEK293 cells were transfected to express HA–Surf4 along with COX-2 or COX-2 mutants, as indicated, for immunoprecipitation assays with anti-HA antibody. Where indicated, cells were treated with tunicamycin (1 µg/ml) for 24 h (E). β-actin is shown as a loading control, and molecular masses are indicated in kDa. Input lanes in A, B and D–F show 10% of the total lysate. Images in A, D and E are representative of three experiments. Quantitative data are presented as mean±s.d. (n=3–4) and are normalized to the WT COX-2 group or WT control-siRNA group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Fig. 7.

Derlin-1, Surf4 and p97 preferentially interact with non-glycosylated COX-2. (A) HEK293 cells were transfected to express HA–derlin-1 (Der-1) along with COX-2. Cells were treated with tunicamycin (Tuni; 1 µg/ml) for 24 h. Cell lysates were immunoprecipitated (IP) with HA antibodies for western blot analysis. (B) HEK293 cells were transfected to express HA–derlin-1 (Der-1) along with wild-type (WT) COX-2 or the indicated COX-2 mutants (m4, COX-2-m4; m1-3, COX-2-m1–3; m1-4, COX-2-m1–4). Cell lysates were immunoprecipitated with HA antibodies for western blot analysis. Quantification of immunoprecipitated COX-2 is shown on the right. (C) HEK293 cells were transfected to express wild-type COX-2, COX-2-m4 or COX-2-m1–4 along with control siRNA or siRNA targeting derlin-1. Whole-cell lysates were used for western blot analysis. Quantification of COX2 is shown on the right. (D) HEK293 cells were transfected to express HA–p97 along with wild-type COX-2 or various levels of COX-2-m1–4 for an immunoprecipitation assay with anti-HA antibody. Cells were transfected with increasing amounts of COX-2-m1–4 plasmid, as indicated by the triangle symbol. (E,F) HEK293 cells were transfected to express HA–Surf4 along with COX-2 or COX-2 mutants, as indicated, for immunoprecipitation assays with anti-HA antibody. Where indicated, cells were treated with tunicamycin (1 µg/ml) for 24 h (E). β-actin is shown as a loading control, and molecular masses are indicated in kDa. Input lanes in A, B and D–F show 10% of the total lysate. Images in A, D and E are representative of three experiments. Quantitative data are presented as mean±s.d. (n=3–4) and are normalized to the WT COX-2 group or WT control-siRNA group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Cav-1 preferentially binds to N-glycosylated COX-2

The Cav-1 scaffolding domain binds to proteins containing aromatic residues (Couet et al., 1997). We examined whether N-glycosylation of COX-2 affects its interaction with Cav-1. Inhibition of N-glycosylation by tunicamycin reduced the interaction between COX-2 and Cav-1 (Fig. 8A). Furthermore, mutations to disrupt N-glycosylation decreased the interaction of COX-2 with Cav-1, in that the levels of COX-2 co-precipitated with Cav-1 were positively associated with the number of intact COX-2 N-glycosylated residues (wild-type COX-2>COX-2-m4>COX-2-m1–3>COX-2-m1–4; Fig. 8B). Additionally, we examined whether the N-glycosylation pattern of COX-2 affects its interaction with Cav-1 by using CHO-Lec1 cells, which have non-detectable N-acetylglucosaminyl-transferase I activity and bear mannose-rich carbohydrates at N-glycosylation sites. Cav-1 and COX-2 were exogenously expressed in CHO-K1 (wild type) and CHO-Lec1 cells. In the input samples, the COX-2 level was higher in CHO-Lec1 cells than in CHO-K1 cells (Fig. 8C), suggesting that changes to glycosylation in CHO-Lec1 cells hinders COX-2 degradation. Moreover, similar levels of COX-2 were co-precipitated with Cav-1 in CHO-K1 cells and CHO-Lec1 cells (Fig. 8C), implying that the interaction of Cav-1 with COX-2 is independent of N-glycosylation patterns.

Fig. 8.

Cav-1 preferentially interacts with N-glycosylated COX-2. (A,B) H1299 cells were transfected to express HA–Cav-1 along with wild-type (WT) COX-2 or COX-2 mutants (m4, COX-2-m4; m1-3, COX-2-m1–3; m1-4, COX-2-m1–4), as indicated. (A) H1299 cells were treated with tunicamycin (Tuni; 1 µg/ml) for 24 h. (B) Interaction of Cav-1 and COX-2 mutants. (C) CHO-K1 and CHO-Lec1 cells were transfected to express HA–Cav-1 and various levels of COX-2. Cells were transfected with decreasing amounts of COX-2 plasmid, as indicated by the triangle symbol. (D) H1299 cells were transfected to express wild-type HA–Cav-1 or HA-tagged Cav-1 lacking the scaffolding domain (HA–Cav-1-ΔSD) along with COX-2. (E) H1299 cells were transfected to express wild-type COX-2 or COX-2-m1–4 along with wild-type HA–Cav-1 or HA–Cav-1-Y14F, as indicated. (A–E) Cell lysates were immunoprecipitated (IP) with anti-HA antibodies for western blot analysis. β-actin is shown as a loading control, and molecular masses are indicated in kDa. Input lanes show 10% of the total lysate. Images in A and C–E are representative of three experiments. Quantitative data showing levels of immunoprecipitated COX-2 in B are presented as mean±s.d. (n=3) and are normalized to the WT COX-2 group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

Fig. 8.

Cav-1 preferentially interacts with N-glycosylated COX-2. (A,B) H1299 cells were transfected to express HA–Cav-1 along with wild-type (WT) COX-2 or COX-2 mutants (m4, COX-2-m4; m1-3, COX-2-m1–3; m1-4, COX-2-m1–4), as indicated. (A) H1299 cells were treated with tunicamycin (Tuni; 1 µg/ml) for 24 h. (B) Interaction of Cav-1 and COX-2 mutants. (C) CHO-K1 and CHO-Lec1 cells were transfected to express HA–Cav-1 and various levels of COX-2. Cells were transfected with decreasing amounts of COX-2 plasmid, as indicated by the triangle symbol. (D) H1299 cells were transfected to express wild-type HA–Cav-1 or HA-tagged Cav-1 lacking the scaffolding domain (HA–Cav-1-ΔSD) along with COX-2. (E) H1299 cells were transfected to express wild-type COX-2 or COX-2-m1–4 along with wild-type HA–Cav-1 or HA–Cav-1-Y14F, as indicated. (A–E) Cell lysates were immunoprecipitated (IP) with anti-HA antibodies for western blot analysis. β-actin is shown as a loading control, and molecular masses are indicated in kDa. Input lanes show 10% of the total lysate. Images in A and C–E are representative of three experiments. Quantitative data showing levels of immunoprecipitated COX-2 in B are presented as mean±s.d. (n=3) and are normalized to the WT COX-2 group. *P<0.05; **P<0.01; ***P<0.001 (one-way ANOVA followed by post-hoc Bonferroni test).

The Cav-1 scaffolding domain is involved in binding with target proteins (Parton and Simons, 2007). We examined whether the Cav-1 scaffolding domain is involved in the interaction of Cav-1 with COX-2. Cav-1 co-precipitated with derlin-1 and derlin-2 in H1299 cells (Fig. 8D), but the interaction was markedly reduced when the Cav-1 scaffolding domain was deleted, which suggests that the scaffolding domain is involved in the interaction of Cav-1 with COX-2 and ERAD components during COX-2 degradation. Moreover, phosphorylation of Tyr14 in Cav-1 regulates Cav-1 activity (Li et al., 1996). Mutation of Tyr14 to Phe (Y14F) to abolish this phosphorylation hampered the interaction of Cav-1 with COX-2 (Fig. 8E; lane 1 versus lane 2). Mutation at all N-glycosylation residues, COX-2-m1–4, further suppressed the interaction of COX-2 with Cav-1-Y14F (Fig. 8E; lane 2 versus lane 4). These results imply that Cav-1 interacts with N-glycosylated COX-2 through its scaffolding domain and that the Tyr14 phosphorylation in Cav-1 enhances binding with COX-2.

Retrotranslocation of ER lumen proteins to the cytosol is a critical step for protein ubiquitylation and subsequent degradation in ERAD. In this study, we demonstrate a novel degradation machinery involving several components. Surf4 collaborates with derlin-2 and derlin-1 to mediate COX-2 translocation to the cytosol to interact with Cav-1 for ubiquitylation and, subsequently, with the p97 complex for degradation (Fig. S5). The protein export function of Surf4 is conserved across species. As a cargo receptor or protein export mediator, Erv29, the yeast homolog of Surf4, mediates sex-pheromone precursor and vacuolar protease export (Belden and Barlowe, 2001). SFT-4, the Caenorhabditis elegans homolog of Surf4, facilitates export of the yolk protein VIT-2 (Saegusa et al., 2018). In mammalian cells, Surf4 promotes the ER protein export and secretion of soluble proteins (Emmer et al., 2018; Yin et al., 2018). This functional conservation suggests that Surf4 participates and plays important roles in ER protein export. Conversely, our study reveals a novel function of Surf4 as a component of the ERAD machinery that facilitates translocation of COX-2, a membrane-associated ER lumen protein, to the cytosol for degradation.

Additionally, the results obtained from our CRISPR screening revealed the presence of several glycosylated proteins expressed in the Golgi. This finding suggests that transfer of COX-2 to the Golgi for glycosylation and then back to the ER is a critical step for COX2 degradation through ERAD. The requirement of COX-2 transport between the ER and Golgi for ERAD is supported by the function of the yeast homolog, Erv29, which is required for quality control of ER proteins (Caldwell et al., 2001). Although derlin-2 and derlin-1 are candidates for protein retrotranslocation, our results reveal that Surf4 acts downstream of both derlins to facilitate translocation of COX-2 to the cytosolic face of the ER membrane for ubiquitylation and to interact with Cav-1 and p97. Furthermore, in addition to its role in translocation of soluble proteins, our results demonstrate that Surf4 also facilitates translocation of membrane-associated proteins, such as COX-2, from the ER lumen to the cytoplasm.

In mammalian cells, three derlins, derlin-1, derlin-2 and derlin-3, have been identified, and they are thought to oligomerize in the ER membrane and participate in formation of a retrotranslocation channel (Chang et al., 2004; Lilley and Ploegh, 2004, 2005; Ye et al., 2005). Derlin-1 and derlin-2 are expressed ubiquitously, but derlin-3 mRNA expression is restricted to several tissues (Oda et al., 2006). In contrast to the independent function of derlin-1 and derlin-2 in facilitating ER protein retrotranslocation (Bernardi et al., 2008; Chang et al., 2004; Huang et al., 2013; Lilley and Ploegh, 2004, 2005; Oda et al., 2006; Sugiyama et al., 2011), our results suggest a collaborative role of derlin-2 and derlin-1 in interacting with Surf4 to mediate COX-2 degradation. Derlins contain four transmembrane domains and have been suggested to collaborate with other transmembrane proteins to form a translocon for ER protein retrotranslocation (Lilley and Ploegh, 2005). This hypothesis is supported by our results that derlin-2, derlin-1 and Surf4 interact with each other to mediate COX-2 translocation from the ER to the cytosol. Due to the rapid degradation of COX-2 shortly after induction, our findings suggest that derlin-2, derlin-1 and Surf4 constitute a distinct complex that facilitates the swift degradation of COX-2.

N-glycosylation plays an important role in regulating ER protein folding and degradation (Oda et al., 2006; Xu and Ng, 2015). In mammals, EDEM family proteins facilitate glycoprotein degradation and mannose trimming (Hirao et al., 2006; Oda et al., 2006; Olivari et al., 2006), and EDEM1 interacts with derlin-2 and derlin-3 to mediate degradation of glycosylated ER proteins (Oda et al., 2006). Our results confirm that N-glycosylation has a crucial role in COX-2 degradation, as previously suggested (Chen et al., 2013; Mbonye et al., 2006). Although N-glycosylation facilitates COX-2 degradation, our results reveal that derlin-1 and Surf4 preferentially bind to non-N-glycosylated COX-2. Moreover, we found that the interaction of derlin-2 with COX-2 is independent of COX-2 N-glycosylation, whereas derlin-2 preferentially facilitates degradation of N-glycosylated COX-2 (Fig. 7D). These results imply that derlin-2 might interact with lectins, such as EDEM2 and EDEM3 (also detected in our CRISPR screen; Fig. S2A), which interact with the N-glycosylated COX-2, transfer COX-2 to derlin-2, and then recruit derlin-1 and Surf4 for translocation and degradation (Fig. S5). At the cytosolic face of the ER, Cav-1 interacts with the derlin-2–derlin-1–Surf4 machinery and the N-glycosylated COX-2, and recruits the p97 complex to accelerate translocation of COX-2 to cytosol for ubiquitylation and proteasomal degradation, as previously suggested (Chen et al., 2013). The derlin-2–derlin-1–Surf4–Cav-1 machinery might represent a unique pathway for ER protein degradation. We note that derlin-2, derlin-1, Surf4, Cav-1 and p97 interact with non-glycosylated COX-2 (Figs 57). However, these interactions do not effectively facilitate the degradation of non-glycosylated COX-2. These results imply that binding of these factors to non-glycosylated COX-2 alone is not sufficient to mediate its degradation.

Cav-1 has been suggested to interact with specific motifs containing aromatic residues via its scaffolding domain (Couet et al., 1997). Our results indicate that Cav-1 prefers to interact with N-glycosylated COX-2 independently of N-glycosylation patterns, because Cav-1 was seen to interact with COX-2 despite the dissimilar N-glycan patterns in CHO-K1 and CHO-Lec1 cells. Since the interaction is independent of the aromatic residues of COX-2, Cav-1 might interact with the specific secondary structure of N-glycosylated COX-2. Moreover, our data suggest that the Cav-1 scaffolding domain and phosphorylation of Cav-1 at Tyr14 are involved in the interaction between COX-2 and Cav-1.

In summary, our results demonstrate that Surf4 interacts with derlin-2, derlin-1 and Cav-1 to facilitate COX-2 translocation to the cytosol for degradation through the derlin-2–derlin-1–Surf4–Cav-1–p97 pathway (Fig. S5). Although COX-2 degradation is N-glycosylation dependent, derlin-2, derlin-1 and Surf4 do not prefer to interact with N-glycosylated COX-2. Cav-1 interacts with derlin-2, derlin-1 and Surf4, and facilitates translocation of N-glycosylated COX-2 to the cytosol. The derlin-2–derlin-1–Surf4–Cav-1 machinery might represent a unique pathway to accelerate COX-2 degradation in ERAD.

Cell culture

A549, HEK293, CHO-K1, CHO-Lec1 and HS-68 cells were obtained from the American Type Culture Collection (ATCC) and cultured in DMEM (Invitrogen) supplemented with 10% FBS (Hyclone) and 100 units/ml penicillin-streptomycin (PS; Invitrogen). H1299 cells (ATCC) were maintained in RPMI medium (Invitrogen) containing 10% FBS and PS. To induce COX-2 expression, HS-68 cells were treated with IL-1β (1 ng/ml; Calbiochem) in culture medium for 6 h.

Antibodies and reagents

Antibodies against derlin-1, derlin-2, β-actin, HA and Flag were from Sigma-Aldrich. Antibodies against Cav-1 and p97 were from Santa Cruz Biotechnology (Santa Cruz, CA, USA). COX-2 mouse monoclonal antibodies were from Cayman Chemical (Ann Arbor, MI, USA). Antibodies against GRP94 and Surf4, and anti-rabbit IgG and anti-mouse IgG were from GeneTex. Horseradish peroxidase-conjugated secondary antibodies were from Jackson ImmunoResearch. Fluorophore-conjugated secondary antibodies and DAPI were from Thermo Fisher Scientific and Sigma-Aldrich, respectively. IL-1β and cycloheximide were from Calbiochem (La Jolla, CA, USA). MG-132 and tunicamycin were from Sigma-Aldrich. Cav-1 and derlin-1 siRNAs were from Dharmacon. derlin-2 siRNAs were from Sigma-Aldrich. Surf4 siRNAs were from MDBio (sequence shown in Table S1). Lentiviral vectors encoding Cas9 nuclease and either Surf4 shRNA or control shRNA were obtained from the RNA Technology Platform and Gene Manipulation core (RNAi core, Academia Sinica, Taipei, Taiwan). Additional details of reagents are included in Table S1.

CRISPR–Cas9 screening for potential genes mediating COX-2 degradation

The sfGFP–COX-2 chimeric cDNA was constructed by PCR amplification to insert the sfGFP sequence into the sequence encoding the N terminus of COX-2 cDNA, after the signal peptide sequence (Fig. S1A), and subsequent subcloning into pcDNA3.1 (Invitrogen). sfGFP–COX-2 was stably expressed in HEK293 cells by transfection and selection using G418 (InvivoGen, 800 μg/ml). The Cas9 nuclease was stably expressed in sfGFP–COX-2 expressing HEK293 cells by lentiviral transduction and selection using blasticidin (InvivoGen, 10 μg/ml). Then, 2×107 sfGFP–COX-2- and Cas9-expressing cells were transduced with a pooled sgRNA library at multiplicity of infection of 0.2 and selected for puromycin resistance (InvivoGen, 2 μg/ml) for 5 days. The sgRNA library was kindly provided by the RNAi core (Academia Sinica, Taiwan). The GFP-high cells were enriched by two rounds of fluorescence-activated cell sorting, and genomic DNA was extracted from the sorted cells (Fig. S1B,C). The amplicons from two independent genomic DNA samples for next-generation sequencing were prepared using PCR with different combinations of primers (Table S1). The PCR amplicons were sequenced on an Illumina MiSeq system by the National Center for Genome Medicine (NCGM), Academia Sinica, and the results were analyzed by Welgene Biotech Company (Taiwan). The genes involved in ER protein degradation or dislocation, or targeted by at least two different sgRNAs, are listed in Fig. S2A. The genes related to protein glycosylation are listed in Fig. S2B.

Expression vector construction and lentivirus-mediated shRNA transduction

Expression vectors of pCOX-2, pHA-p97, pHA-Derlin-1, pHA-Cav-1 and pUfd-1-FLAG were generated as described previously (Chen et al., 2013). COX-2 mutants were generated by PCR mutagenesis. Coding sequences of derlin-2 and Surf4 were amplified by reverse transcription PCR (RT-PCR) (see Table S1), and then subcloned into pXJN-HA (provided by Dr S. Y. Shieh, Academia Sinica) or GFP-C3 (Clontech) to generate pXJN-HA-Derlin-2, pXJN-HA-Surf4 and GFP-Surf4. Cells were transfected with plasmids or siRNAs using Lipofectamine 2000 (Invitrogen) or PepMute siRNA transfection reagent (SignaGen Laboratories) according to the manufacturer's instructions. COX-2 mutants were constructed by PCR amplification using the primers listed in Table S1.

The amounts of pCOX-2 and pCOX-2 mutants for transfection were adjusted to ensure similar COX-2 expression levels for immunoprecipitation assays. To avoid inconsistent protein expression levels, for co-transfection of a plasmid with siRNA, the plasmid was transfected for 6 h, then cells were trypsinized and subcultured in smaller dishes overnight, followed by siRNA transfection for another 24 h, as described previously (Chen et al., 2013). For shRNA transduction, A549 cells were infected with lentivirus carrying shRNA for 24 h and then selected with puromycin (InvivoGen, 2 μg/ml) for 5 days. The shRNA sequences for Surf4 and control (luciferase) are listed in Table S1.

Western blot analysis

Pulldown lysates from immunoprecipitation assay or cell lysates were resolved on SDS-polyacrylamide gels and then transferred to PVDF membranes, which were blocked in 5% skim milk in PBS containing 0.1% Triton X-100 (PBST), incubated with primary antibodies for 2 h, then washed three times with PBST. The blots were incubated with HRP-conjugated secondary antibodies for 1 h. After three washes, signals were visualized by using enhanced chemiluminescent reagent (Millipore). β-actin was an internal control. The relative levels of protein expression were quantified using Gel-Pro Analyzer 3.2, and the intensity of the input group in western blots was used as control to quantify immunoprecipitation results. All experiments were performed as minimum of three biological replicates. Antibody dilutions are given in Table S1. Fig. S6 shows larger portions of all images for blot transparency.

RT-PCR

Total RNA was extracted with Trizol reagent (Invitrogen). 10 μg RNA was used for cDNA synthesis with oligo-dT21 and SuperScript III reverse transcriptase (Invitrogen) according to the manufacturer's protocols. cDNAs were used for PCR amplification with human-specific primers (Table S1).

Immunoprecipitation assays

Cells were lysed using RIPA buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100 and protease inhibitor cocktail (Roche)], and lysates were sonicated and centrifuged at 4°C. The supernatant was incubated with antibodies for 2 h at 4°C and then 20 µl of 50% A+G agarose beads (Amersham Biosciences) were added for 1 h at 4°C. Antibody details are provided in Table S1. The immunoprecipitates were washed three times with RIPA buffer and then boiled with protein sample buffer (50 mM Tris-HCl pH 6.8, 2% SDS, 0.02% Bromophenol Blue, 10% glycerol and 1% β-mercaptoethanol). The bound proteins were analyzed by western blot analysis. In part of experiments (in Fig. 4C,E,F), the cell lysate mixture of control and assay samples were used to assess the unspecific binding with IgG to save effort.

Immunofluorescence and confocal microscopy

Cells were fixed in 4% paraformaldehyde in PBS for 15 min, permeabilized with methanol for 15 min at room temperature, and blocked in PBS with 10% FBS for 30 min. Cells were incubated with primary antibodies for 1 h, washed three times and incubated with fluorescently labeled antibodies for 1 h. Antibody details are provided in Table S1. Cells were washed and analyzed by confocal microscopy (Carl Zeiss LSM 700 with 63× oil objective). Colocalization analysis and MOC were calculated using Zeiss Zen Black software. Calculated values ranging from 0.6 to 1 were indicated as colocalization (Zinchuk and Zinchuk, 2008).

Ubiquitylation assay

Cells were co-transfected to overexpress COX-2 and His-tagged ubiquitin (pcDNA3-His-ubiquitin was provided by Dr S. Y. Shieh, Academia Sinica) for 24 h, then treated with MG-132 (10 mM) for another 16 h, followed by sonication in buffer A (6 M guanidine-HCl, 0.1 M Na2HPO4/NaH2PO4, 10 mM imidazole, pH 8.0). After centrifugation, cell lysates were incubated with 20 ml of 50% Ni-beads (Invitrogen) at room temperature for 1 h. The precipitated beads were washed once in buffer A, once in 1:3 ratio mixed buffer (buffer A:buffer TI) and three times in buffer TI (20 mM imidazole, 0.2% Triton X-100, 25 mM Tris-HCl pH 6.8). Bound proteins were analyzed by western blot analysis.

Statistical analysis

Data are presented as mean±s.d. One-way ANOVA followed by post-hoc Bonferroni test was used to account for multiple testing. Unpaired two-tailed Student's t-test was used to account for two group testing. Statistical analyses involved use of GraphPad Prism 8 (GraphPad Software, Inc., La Jolla, CA, USA). Statistical significance was set at P<0.05.

We thank the Flow Cytometry Core Facility, DNA Sequencing Core Facility and Light Microscopy Core Facility of the Institute of Biomedical Sciences, Academia Sinica for technical support, and the RNAi core of Academia Sinica for providing the sgRNA library. We also thank Laura Smales (BioMed Editing, Toronto, Canada) for help with language editing.

Author contributions

Conceptualization: S.-K.S.; Methodology: S.-F.C., K.T., S.-K.S.; Software: S.-F.C., C.-H.W., Y.-M.L., J.-Y.L.; Validation: S.-F.C., S.-K.S.; Formal analysis: S.-F.C.; Investigation: S.-F.C., S.-K.S.; Resources: S.-F.C.; Data curation: S.-F.C., Y.-M.L., K.T., J.-Y.L.; Writing - original draft: S.-F.C., S.-K.S.; Writing - review & editing: S.-F.C., C.-H.W., Y.-M.L., K.T., J.-Y.L., S.-K.S.; Visualization: S.-F.C., C.-H.W., S.-K.S.; Supervision: S.-K.S.; Project administration: S.-K.S.; Funding acquisition: S.-K.S.

Funding

This study was supported by grants from the National Science and Technology Council, Taiwan (NSC 94-2320-B-001-038, MOST 103-2321-B-001-072, MOST 104-2321-B-001-035, MOST 105-2321-B-001-021 and MOST 108-2320-B-001-014-MY3); the Institute of Biomedical Sciences, Academia Sinica (CRC104-P04); and Academia Sinica (20y3-y311005).

Data availability

All relevant data can be found within the article and its supplementary information.

Belden
,
W. J.
and
Barlowe
,
C.
(
2001
).
Role of Erv29p in collecting soluble secretory proteins into ER-derived transport vesicles
.
Science
294
,
1528
-
1531
.
Bernardi
,
K. M.
,
Forster
,
M. L.
,
Lencer
,
W. I.
and
Tsai
,
B.
(
2008
).
Derlin-1 facilitates the retro-translocation of cholera toxin
.
Mol. Biol. Cell.
19
,
877
-
884
.
Caldwell
,
S. R.
,
Hill
,
K. J.
and
Cooper
,
A. A.
(
2001
).
Degradation of endoplasmic reticulum (ER) quality control substrates requires transport between the ER and Golgi
.
J. Biol. Chem.
276
,
23296
-
23303
.
Chang
,
D.-M.
,
Shyue
,
S.-K.
,
Liu
,
S.-H.
,
Chen
,
Y.-T.
,
Yeh
,
C.-Y.
,
Lai
,
J.-H.
,
Lee
,
H.-S.
and
Chen
,
A.
(
2004
).
Dual biological functions of an interleukin-1 receptor antagonist-interleukin-10 fusion protein and its suppressive effects on joint inflammation
.
Immunology
112
,
643
-
650
.
Chen
,
S. F.
,
Liou
,
J. Y.
,
Huang
,
T. Y.
,
Lin
,
Y. S.
,
Yeh
,
A. L.
,
Tam
,
K.
,
Tsai
,
T. H.
,
Wu
,
K. K.
and
Shyue
,
S. K.
(
2010
).
Caveolin-1 facilitates cyclooxygenase-2 protein degradation
.
J. Cell Biochem.
109
,
356
-
362
.
Chen
,
S. F.
,
Wu
,
C. H.
,
Lee
,
Y. M.
,
Tam
,
K.
,
Tsai
,
Y. C.
,
Liou
,
J. Y.
and
Shyue
,
S. K.
(
2013
).
Caveolin-1 interacts with Derlin-1 and promotes ubiquitination and degradation of cyclooxygenase-2 via collaboration with p97 complex
.
J. Biol. Chem.
288
,
33462
-
33469
.
Couet
,
J.
,
Sargiacomo
,
M.
and
Lisanti
,
M. P.
(
1997
).
Interaction of a receptor tyrosine kinase, EGF-R, with caveolins. Caveolin binding negatively regulates tyrosine and serine/threonine kinase activities
.
J. Biol. Chem.
272
,
30429
-
30438
.
Emmer
,
B. T.
,
Hesketh
,
G. G.
,
Kotnik
,
E.
,
Tang
,
V. T.
,
Lascuna
,
P. J.
,
Xiang
,
J.
,
Gingras
,
A. C.
,
Chen
,
X. W.
and
Ginsburg
,
D.
(
2018
).
The cargo receptor SURF4 promotes the efficient cellular secretion of PCSK9
.
Elife
7
,
e38839
.
Hashemi Goradel
,
N.
,
Najafi
,
M.
,
Salehi
,
E.
,
Farhood
,
B.
and
Mortezaee
,
K.
(
2019
).
Cyclooxygenase-2 in cancer: A review
.
J. Cell Physiol.
234
,
5683
-
5699
.
Hirao
,
K.
,
Natsuka
,
Y.
,
Tamura
,
T.
,
Wada
,
I.
,
Morito
,
D.
,
Natsuka
,
S.
,
Romero
,
P.
,
Sleno
,
B.
,
Tremblay
,
L. O.
,
Herscovics
,
A.
et al. 
(
2006
).
EDEM3, a soluble EDEM homolog, enhances glycoprotein endoplasmic reticulum-associated degradation and mannose trimming
.
J. Biol. Chem.
281
,
9650
-
9658
.
Huang
,
C. H.
,
Hsiao
,
H. T.
,
Chu
,
Y. R.
,
Ye
,
Y.
and
Chen
,
X.
(
2013
).
Derlin2 protein facilitates HRD1-mediated retro-translocation of sonic hedgehog at the endoplasmic reticulum
.
J. Biol. Chem.
288
,
25330
-
25339
.
Katiyar
,
S.
,
Joshi
,
S.
and
Lennarz
,
W. J.
(
2005
).
The retrotranslocation protein Derlin-1 binds peptide:N-glycanase to the endoplasmic reticulum
.
Mol. Biol. Cell
16
,
4584
-
4594
.
Krshnan
,
L.
,
Van De Weijer
,
M. L.
and
Carvalho
,
P.
(
2022
).
Endoplasmic reticulum-associated protein degradation
.
Cold Spring Harb Perspect Biol
.
14
,
a041247
.
Li
,
S.
,
Seitz
,
R.
and
Lisanti
,
M. P.
(
1996
).
Phosphorylation of caveolin by src tyrosine kinases. The alpha-isoform of caveolin is selectively phosphorylated by v-Src in vivo
.
J. Biol. Chem.
271
,
3863
-
3868
.
Lilley
,
B. N.
and
Ploegh
,
H. L.
(
2004
).
A membrane protein required for dislocation of misfolded proteins from the ER
.
Nature
429
,
834
-
840
.
Lilley
,
B. N.
and
Ploegh
,
H. L.
(
2005
).
Multiprotein complexes that link dislocation, ubiquitination, and extraction of misfolded proteins from the endoplasmic reticulum membrane
.
Proc. Natl. Acad. Sci. USA
102
,
14296
-
14301
.
Mbonye
,
U. R.
,
Wada
,
M.
,
Rieke
,
C. J.
,
Tang
,
H. Y.
,
Dewitt
,
D. L.
and
Smith
,
W. L.
(
2006
).
The 19-amino acid cassette of cyclooxygenase-2 mediates entry of the protein into the endoplasmic reticulum-associated degradation system
.
J. Biol. Chem.
281
,
35770
-
35778
.
Mitrovic
,
S.
,
Ben-Tekaya
,
H.
,
Koegler
,
E.
,
Gruenberg
,
J.
and
Hauri
,
H. P.
(
2008
).
The cargo receptors Surf4, endoplasmic reticulum-Golgi intermediate compartment (ERGIC)-53, and p25 are required to maintain the architecture of ERGIC and Golgi
.
Mol. Biol. Cell
19
,
1976
-
1990
.
Neal
,
S.
,
Jaeger
,
P. A.
,
Duttke
,
S. H.
,
Benner
,
C.
,
Glass
,
C. K.
,
Ideker
,
T.
and
Hampton
,
R. Y.
(
2018
).
The Dfm1 Derlin is required for ERAD retrotranslocation of integral membrane proteins
.
Mol. Cell
69
,
915
.
Oda
,
Y.
,
Okada
,
T.
,
Yoshida
,
H.
,
Kaufman
,
R. J.
,
Nagata
,
K.
and
Mori
,
K.
(
2006
).
Derlin-2 and Derlin-3 are regulated by the mammalian unfolded protein response and are required for ER-associated degradation
.
J. Cell Biol.
172
,
383
-
393
.
Olivari
,
S.
,
Cali
,
T.
,
Salo
,
K. E.
,
Paganetti
,
P.
,
Ruddock
,
L. W.
and
Molinari
,
M.
(
2006
).
EDEM1 regulates ER-associated degradation by accelerating de-mannosylation of folding-defective polypeptides and by inhibiting their covalent aggregation
.
Biochem. Biophys. Res. Commun.
349
,
1278
-
1284
.
Olzmann
,
J. A.
,
Kopito
,
R. R.
and
Christianson
,
J. C.
(
2013
).
The mammalian endoplasmic reticulum-associated degradation system
.
Cold Spring Harb. Perspect. Biol.
5
,
a013185
.
Parton
,
R. G.
and
Simons
,
K.
(
2007
).
The multiple faces of caveolae
.
Nat. Rev. Mol. Cell Biol.
8
,
185
-
194
.
Pu
,
D.
,
Yin
,
L.
,
Huang
,
L.
,
Qin
,
C.
,
Zhou
,
Y.
,
Wu
,
Q.
,
Li
,
Y.
,
Zhou
,
Q.
and
Li
,
L.
(
2021
).
Cyclooxygenase-2 inhibitor: a potential combination strategy with immunotherapy in cancer
.
Front. Oncol.
11
,
637504
.
Reeves
,
J. E.
and
Fried
,
M.
(
1995
).
The surf-4 gene encodes a novel 30 kDa integral membrane protein
.
Mol. Membr. Biol.
12
,
201
-
208
.
Saegusa
,
K.
,
Sato
,
M.
,
Morooka
,
N.
,
Hara
,
T.
and
Sato
,
K.
(
2018
).
SFT-4/Surf4 control ER export of soluble cargo proteins and participate in ER exit site organization
.
J. Cell Biol.
217
,
2073
-
2085
.
Smith
,
W. L.
,
Urade
,
Y.
and
Jakobsson
,
P. J.
(
2011
).
Enzymes of the cyclooxygenase pathways of prostanoid biosynthesis
.
Chem. Rev.
111
,
5821
-
5865
.
Spencer
,
A. G.
,
Thuresson
,
E.
,
Otto
,
J. C.
,
Song
,
I.
,
Smith
,
T.
,
Dewitt
,
D. L.
,
Garavito
,
R. M.
and
Smith
,
W. L.
(
1999
).
The membrane binding domains of prostaglandin endoperoxide H synthases 1 and 2. Peptide mapping and mutational analysis
.
J. Biol. Chem.
274
,
32936
-
32942
.
Sugiyama
,
T.
,
Shuto
,
T.
,
Suzuki
,
S.
,
Sato
,
T.
,
Koga
,
T.
,
Suico
,
M. A.
,
Kusuhara
,
H.
,
Sugiyama
,
Y.
,
Cyr
,
D. M.
and
Kai
,
H.
(
2011
).
Posttranslational negative regulation of glycosylated and non-glycosylated BCRP expression by Derlin-1
.
Biochem. Biophys. Res. Commun.
404
,
853
-
858
.
Xu
,
C.
and
Ng
,
D. T.
(
2015
).
Glycosylation-directed quality control of protein folding
.
Nat. Rev. Mol. Cell Biol.
16
,
742
-
752
.
Yazaki
,
M.
,
Kashiwagi
,
K.
,
Aritake
,
K.
,
Urade
,
Y.
and
Fujimori
,
K.
(
2012
).
Rapid degradation of cyclooxygenase-1 and hematopoietic prostaglandin D synthase through ubiquitin-proteasome system in response to intracellular calcium level
.
Mol. Biol. Cell
23
,
12
-
21
.
Ye
,
Y.
,
Meyer
,
H. H.
and
Rapoport
,
T. A.
(
2001
).
The AAA ATPase Cdc48/p97 and its partners transport proteins from the ER into the cytosol
.
Nature
414
,
652
-
656
.
Ye
,
Y.
,
Shibata
,
Y.
,
Kikkert
,
M.
,
Van Voorden
,
S.
,
Wiertz
,
E.
and
Rapoport
,
T. A.
(
2005
).
Recruitment of the p97 ATPase and ubiquitin ligases to the site of retrotranslocation at the endoplasmic reticulum membrane
.
Proc. Natl. Acad. Sci. USA
102
,
14132
-
14138
.
Yin
,
Y.
,
Garcia
,
M. R.
,
Novak
,
A. J.
,
Saunders
,
A. M.
,
Ank
,
R. S.
,
Nam
,
A. S.
and
Fisher
,
L. W.
(
2018
).
Surf4 (Erv29p) binds amino-terminal tripeptide motifs of soluble cargo proteins with different affinities, enabling prioritization of their exit from the endoplasmic reticulum
.
PLoS Biol.
16
,
e2005140
.
Zinchuk
,
V.
and
Zinchuk
,
O.
(
2008
).
Quantitative colocalization analysis of confocal fluorescence microscopy images
.
Curr. Protoc. Cell Biol.
Chapter 4
,
Unit 4 19
.

Competing interests

The authors declare no competing or financial interests.

Supplementary information