Phosphorylation switches protein disulfide isomerase activity to maintain proteostasis and attenuate ER stress

Abstract

Accumulated unfolded proteins in the endoplasmic reticulum (ER) trigger the unfolded protein response (UPR) to increase ER protein folding capacity. ER proteostasis and UPR signaling need to be regulated in a precise and timely manner. Here, we identify phosphorylation of protein disulfide isomerase (PDI), one of the most abundant and critical folding catalysts in the ER, as an early event during ER stress. The secretory pathway kinase Fam20C phosphorylates Ser357 of PDI and responds rapidly to various ER stressors. Phosphorylation of Ser357 induces an open conformation of PDI and turns it from a “foldase” into a “holdase”, which is critical for preventing protein misfolding in the ER. Phosphorylated PDI also binds to the lumenal domain of IRE1α, a major UPR signal transducer, and attenuates excessive IRE1α activity. Importantly, PDI‐S359A knock‐in mice display enhanced IRE1α activation and liver damage under acute ER stress. We conclude that the Fam20C‐PDI axis constitutes a post‐translational response to maintain ER proteostasis and plays a vital role in protecting against ER stress‐induced cell death.

Fam20C‐dependent phosphorylation of protein disulfide isomerase (PDI) turns PDI activity from “foldase” into “holdase”, thereby protecting mouse liver from ER stress‐induced cell death.

Introduction

The endoplasmic reticulum (ER) is the main cellular organelle for protein folding and secretion, calcium storage, and lipid synthesis. Numerous folding enzymes and molecular chaperones in the ER guide the secretion of properly folded proteins while retaining misfolded proteins or targeting their degradation (Sitia & Braakman, 2003). An accumulation of unfolded/misfolded proteins or alternation of redox and Ca²⁺ states in the ER results in imbalanced proteostasis and ER stress (Balch et al, 2008; Walter & Ron, 2011; Karagoz et al, 2019; Ushioda & Nagata, 2019). Chronic ER stress plays a central role in various human pathologies, including cancer, diabetes, cardiovascular diseases, and neurodegenerative diseases (Oakes & Papa, 2015; Wang & Kaufman, 2016).

Endoplasmic reticulum proteostasis is governed by a dynamic signaling network, the unfolded protein response (UPR), which consists of three distinct arms in mammals defined by ER transmembrane sensors—IRE1α, PERK, and ATF6 (Walter & Ron, 2011). The output of UPR signaling is to induce the expression of genes involved in ER protein quality control and/or translational repression of global protein synthesis to counteract proteostatic perturbations in the ER. The IRE1α branch is the most conserved arm of the UPR. ER stress induces the dimerization/oligomerization of IRE1α and its autophosphorylation, which leads to XBP1 mRNA splicing and generates the functional spliced XBP1 (XBP1s) transcription factor regulating multifarious targets (Yoshida et al, 2001; Calfon et al, 2002; Lee et al, 2003; Shoulders et al, 2013). Nevertheless, hyperactivation of IRE1α under prolonged ER stress results in the transition from an adaptive UPR to a terminal proapoptotic program by surpassing the oligomerization threshold that expands the RNase substrate repertoire to many other mRNAs or precursors of apoptosis‐inhibitory microRNAs (Han et al, 2009; Hollien et al, 2009; Upton et al, 2012; Ghosh et al, 2014). Moreover, IRE1α can trigger cellular apoptosis through the TRAF2‐ASK1‐JNK pathway (Urano et al, 2000; Nishitoh et al, 2002). Therefore, IRE1α functions as a central adjustor of cell fate under ER stress, and its activity is precisely and timely regulated by multiple regulatory elements (Hetz et al, 2006; Lisbona et al, 2009; Rodriguez et al, 2012; Eletto et al, 2014; Sepulveda et al, 2018).

However, the inherent latency of the UPR limits its responsiveness to the fluctuation of protein status in the ER, particularly in dedicated secretory cells with heavy protein folding burden. The post‐translational regulation of BiP, a key chaperone in the ER, emerges as a rapid and an economic way to regulate protein folding capacity during the early stage of ER stress (Preissler & Ron, 2018). However, whether post‐translational regulation of other preexisting components contributes to the early response to ER stress remains unclear.

The reversible phosphorylation of proteins is central to the regulation of most aspects of cell function. Fam20C, a secretory pathway protein kinase, which recognizes S‐x‐E/pS motifs within the substrate proteins, generates the majority of the secreted phosphoproteome (Tagliabracci et al, 2012, 2015; Cui et al, 2015). Functional mutations in Fam20C cause a rare and often lethal osteosclerotic bone dysplasia called Raine syndrome (Raine et al, 1989). Recently, the critical roles of Fam20C in fine‐tuning ER redox homeostasis (Zhang et al, 2018) and Ca²⁺ homeostasis (Pollak et al, 2018) have been revealed. However, it is still an open question whether Fam20C is involved in ER proteostasis regulation and UPR signaling under ER stress.

In this study, unbiased proteomics analysis allows the identification of protein disulfide isomerase (PDI) as a substrate of Fam20C kinase under ER stress. PDI, also known as collagen prolyl 4‐hydroxylase subunit β (P4HB), is one of the most abundant enzymes in the ER. PDI is a versatile protein acting as both a thiol‐disulfide oxidoreductase and a molecular chaperone (Hatahet & Ruddock, 2009; Wang et al, 2015). We show that Fam20C phosphorylates serine 357 (Ser357) of PDI and responds rapidly to various ER stressors. Importantly, phosphorylation of Ser357 in the x‐linker region induces an open conformation of PDI and turns it from a “foldase” to a “holdase”, which is critical for preventing protein misfolding in the ER. Furthermore, we demonstrate that phosphorylated PDI directly interacts with IRE1α to attenuate its signaling amplitude, which is critical for protecting against ER stress‐induced liver damage, as illustrated in a mouse model. Overall, our results unravel Fam20C‐induced PDI phosphorylation as a rapid post‐translational mechanism for integrating ER proteostasis sensing and cell fate control.

Results

Fam20C depletion sensitizes the IRE1α branch of the UPR under ER stress

To assess the role of Fam20C in the UPR, the activation of three UPR sensors, IRE1α, PERK, and ATF6, was evaluated in control (shCtrl) and Fam20C stable knockdown (shFAM20C) HepG2 cells. When ER stress was induced by the sarco/endoplasmic reticulum Ca²⁺‐ATPase (SERCA) calcium pump inhibitor thapsigargin (Tg), depletion of Fam20C significantly elevated the phosphorylation of IRE1α but had little effect on the phosphorylation of PERK and its downstream substrate eIF2α or the nuclear localization of the N‐terminal cleavage product of ATF6 (Fig 1A). These results indicated that Fam20C could modulate the IRE1α branch of the UPR under ER stress. Consistent with this notion, Fam20C‐depleted cells displayed a higher amplitude of XBP1 mRNA splicing (Figs 1B and EV1A). As the duration of splicing was similar (Fig EV1A), Fam20C may participate in the early response to ER stress. Ectopic expression of Fam20C wild‐type (WT) decreased the XBP1 mRNA splicing level in Tg‐treated HepG2 cells compared with its inactive D478A mutant (DA) (Fig 1C). Similar results were observed when HepG2 cells were treated with another ER stress inducer tunicamycin (Tm), which inhibits protein N‐glycosylation (Fig 1D and E). Again, we showed that Fam20C can also negatively regulate IRE1α activity in HeLa cells by using Fam20C knockout (KO) cells generated by CRISPR/Cas9 technology previously reported (Zhang et al, 2018; Fig EV1B and C).

Figure 1. Fam20C depletion sensitizes the IRE1α branch of the UPR under ER stress.

• A
  Detection of the activation of IRE1α, PERK, and ATF6 UPR branches by immunoblotting in control (shCtrl) and Fam20C stable knockdown (shFAM20C) HepG2 cells treated with or without 5 μM Tg for 30 min. ATF6‐FL: full‐length ATF6; ATF6‐N: N‐terminal cleavage product of ATF6; Asterisk: unspecific background bands. Fam20C knockdown was verified by protein immunoblotting of Concanavalin A‐Sepharose (Con A)‐enriched culture medium. Ponceau staining was shown as a loading control.
• B
  (Left) Detection of spliced XBP1 (S) and unspliced XBP1 (U) mRNA in shCtrl and shFAM20C HepG2 cells treated with or without 5 μM Tg for 1 h. (Right) Quantification of XBP1 mRNA splicing levels.
• C
  (Left) Detection of XBP1 mRNA splicing in shFAM20C HepG2 cells expressing RNAi‐resistant codon‐altered Fam20C wild‐type (WT) or its inactive D478A mutant (DA) treated with or without 5 μM Tg for 1 h. Fam20C expression levels were shown by protein immunoblotting. (Right) Quantification of XBP1 mRNA splicing levels.
• D, E
  (Left) Detection of XBP1 mRNA splicing in shCtrl and shFAM20C HepG2 cells (D) or shFAM20C HepG2 cells expressing Fam20C WT or DA (E) treated with or without 2 μg/ml Tm for 8 h. (Right) Quantification of XBP1 mRNA splicing levels.
• F
  Volcano plot depicting the log2 of fold change versus – log₁₀ (P‐value) for Fam20C‐interacting proteins in Tg‐treated HepG2 cells. The red dots represent proteins showing statistically significant increase of binding with Fam20C (log₂ fold change ≥ 1; −log₁₀ (P‐value) ≥ 1.3); the blue dots represent proteins showing statistically significant decrease of binding with Fam20C (log₂ fold change ≤ −1; −log₁₀ (P‐value) ≥ 1.3); the green dot is Fam20C.
• G
  HepG2 cells expressing FLAG‐tagged Fam20C were treated with or without 5 μM Tg for 30 min. FLAG immunoprecipitates were analyzed by PDI immunoblotting.

Data information: In (B–E), data were shown as mean ± SEM of three independent experiments. *P < 0.05, **P < 0.01 (two‐tailed Student's t‐test).Source data are available online for this figure.

Figure EV1. Identification of Fam20C interactome under ER stress.

1. (Left) shCtrl and shFAM20C HepG2 cells were treated with or without 5 μM Tg for 1 h. Aliquots of Tg‐treated cells were then washed with culture medium and chased for indicated times. XBP1 mRNA splicing was assayed as described in Fig 1B. (Right) Quantification of XBP1 mRNA splicing levels.
2. (Left) Detection of XBP1 mRNA splicing in WT and Fam20C knockout (KO) HeLa cells treated with or without 5 μM Tg for 1 h. Fam20C KO was verified by protein immunoblotting of Concanavalin A‐Sepharose (Con A)‐enriched culture medium. Ponceau staining was shown as a loading control. (Right) Quantification of XBP1 mRNA splicing levels.
3. (Left) Detection of XBP1 mRNA splicing in Fam20C KO HeLa cells expressing Fam20C WT and its inactive D478A mutant (DA) treated with or without 5 μM Tg for 1 h. Fam20C expression levels were shown by protein immunoblotting. (Right) Quantification of XBP1 mRNA splicing levels.
4. Schematic diagram illustrating the workflow used for identification of Fam20C interactome under ER stress.
5. HepG2 cells expressing FLAG‐tagged Fam20C were treated with or without 5 μM Tg for 30 min, and FLAG immunoprecipitates were analyzed by SDS–PAGE and Coomassie blue staining. Bands were excised for MS analysis except for heavy chain, light chain, and Fam20C bands. h.c.: heavy chain; l. c.: light chain.
6. Venn diagram depicting the number of ER and Golgi proteins bound to Fam20C identified in three independent experiments based on DAVID GO term analysis. A total of 173 proteins identified in all three experiments were analyzed as shown in Fig 1F.

Data information: Data were shown as mean ± SEM of three independent experiments. In (A), **P < 0.01 (two‐way ANOVA, the post hoc Tukey's HSD test). In (B, C) *P < 0.05 (two‐tailed Student's t‐test).Source data are available online for this figure.

To explore the mechanism by which Fam20C controls IRE1α signaling, we set to identify the Fam20C interactome by co‐immunoprecipitation (co‐IP) and mass spectrum (MS) analysis (Fig EV1D and E). Three independent experiments led to the identification of a total of 173 ER and Golgi proteins, which were based on DAVID GO term analysis (Dataset EV1 and Fig EV1F). Among these Fam20C‐interacting proteins, five proteins, including PDI (P4HB), P4HA1, P4HA2, TRIM68, and FKBP9, showed statistically significant increased binding with Fam20C after Tg treatment, whereas two proteins, SERCA2 and ERGIC2, showed statistically significant decreased binding (Fig 1F). The increased binding between PDI and Fam20C under ER stress was further confirmed by co‐IP and immunoblotting (Fig 1G). We thereafter focused on PDI because it is the most enriched protein among the Fam20C interactome under ER stress and is critical for protein folding and quality control in the ER.

PDI Ser357 is a genuine phosphosite of Fam20C

To determine whether PDI is phosphorylated by Fam20C during ER stress, HepG2 cells were treated with Tg, and endogenous PDI was immunoprecipitated and subjected to MS/MS analysis (Fig EV2A). Three phosphorylation sites (Ser357, Ser331, and Ser427) were identified upon Tg treatment (Figs 2A and EV4D and E). Since phosphorylation of Ser331 and Ser427 had little effect on the conformation and activity of PDI (see below), in this study, we focused on the phosphorylation effect of PDI Ser357.

Figure EV2. Fam20C catalyzes PDI Ser357 phosphorylation.

1. (Upper) Immunoprecipitation of endogenous PDI with anti‐PDI mouse monoclonal antibody from HepG2 cells treated with DMSO and 5 μM Tg for 10 min, respectively. (Lower) Successful immunoprecipitation was confirmed by immunoblotting with rabbit anti‐PDI serum. The corresponding PDI band on Coomassie blue gel was excised for mass spectra identification for PDI phosphosites. h.c.: heavy chain; l. c.: light chain.
2. Non‐phosphopeptide (NP‐Pep) and phosphopeptide (P‐Pep) used for generating non‐phospho‐antibody (NP‐Ab) and phospho‐antibody (P‐Ab). NP‐Pep and P‐Pep were spotted on membranes separately, followed by dot blot analysis with NP‐Ab and P‐Ab, respectively.
3. pS³⁵⁷‐PDI immunoblotting of HepG2 cell lysates treated with gradient concentration of λ‐phosphatase (λ‐p'tase) as indicated.
4. Generation of PDI KO HepG2 cells using CRISPR/Cas9 genome editing. The targeting locus of exon 1 in the human PDI gene is represented. The sequences of the mutated alleles in PDI clone 1 (C1) and clone 2 (C2) depicting the insertions/deletions (INDELs, red) are shown. The INDELs are predicted to cause frameshift mutations producing inactive truncations of the protein.
5. Detection of PDI expression in WT and PDI KO HepG2 cells by immunoblotting with rabbit anti‐PDI serum.
6. Detection of PDI (red) expression in WT and PDI KO HepG2 cells by immunofluorescence. ER and nucleus are indicated by CRT (calreticulin, green) and Hoechst 33258 (blue) staining, respectively. Scale bar = 10 μm.
7. Detection of pS³⁵⁷‐PDI in WT and PDI KO HepG2 cells transfected without or with HA‐tagged PDI WT or S357A by immunoblotting.
8. Detection of Fam20C expression in the Con A‐precipitated culture medium from WT and Fam20C KO HeLa cells (C3 and C5) by immunoblotting.
9. Detection of pS³⁵⁷‐PDI in WT and Fam20C KO HeLa cells transfected without or with FLAG‐tagged Fam20C WT or D478A by immunoblotting.

Source data are available online for this figure.

Figure 2. PDI Ser357 is phosphorylated by Fam20C during ER stress.

• A
  Representative MS/MS fragmentation spectrum of tryptic phosphorylated PDI peptide (Ile351‐Lys370) depicting PDI Ser357 phosphorylation enriched from HepG2 cells treated with 5 μM Tg for 10 min.
• B
  (Upper) Domain organization of the human PDI molecule. The residue numbering is for human PDI with signal sequence. The “CGHC” motifs in the a and a’ domains are active sites; carboxyl‐terminal KDEL motif is the ER retention sequence. Ser357 located in the x‐linker region is marked in red. (Lower) Amino acid sequence alignments of PDI x‐linker in several species. Residue positions are indicated by numbering with signal sequences. Human PDI Ser357 and its counterparts are shown in red.
• C
  Time‐dependent incorporation of phosphate group into PDI WT or S357A catalyzed by recombinant Fam20C. Phosphorylated PDI was determined by anti‐pS³⁵⁷‐PDI immunoblotting on both regular and phostag gels.
• D
  Phosphorylated PDI catalyzed by Fam20C was analyzed by dual‐color immunoblotting with anti‐PDI (red) and anti‐pS³⁵⁷‐PDI (green) antibodies. The yellow signal depicts the phosphorylated PDI species.
• E–G
  Detection of pS³⁵⁷‐PDI in HepG2 cells treated with 5 μM Tg (E), 2.5 μg/ml Tm (F), or 1 mM DTT (G) for indicated times.
• H, I
  (Left) Detection of p‐PDI in WT and PDI KO HepG2 cells (H) or in WT and Fam20C KO HeLa cells (I) treated with 5 μM Tg for indicated times by protein immunoblotting on both regular and phostag gels. (Right) Quantification of p‐PDI percentage based on band intensities on phostag gels.

Data information: Data were shown as mean ± SEM of three (H) or four (I) independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 (one‐way ANOVA, the post hoc Tukey's HSD test).Source data are available online for this figure.

Figure EV4. Phosphorylation of PDI Ser357 induces its functional switch.

• A, B
  Limited proteolysis of PDI WT and S357E by trypsin (A) and chymotrypsin (B), respectively.
• C
  Intrinsic fluorescence spectra of recombinant PDI proteins as indicated. Dotted lines indicate the maximal emission wavelength. A.U.: arbitrary units. Data were shown as mean from five independent experiments each performed in three technical replicates.
• D, E
  Representative MS/MS fragmentation spectra of tryptic phosphorylated PDI peptides (Tyr327‐Arg338, D) and (Met425‐Lys436, E), depicting phosphorylated PDI Ser331 and Ser427 enriched from HepG2 cells treated with 5 μM Tg for 10 min.
• F
  Schematic (Upper) and structural (PDB: 4EL1, Lower) representation of PDI phosphosites. Ser331 (blue), Ser357 (red), and Ser427 (cyan) are located in the b’ domain, x‐linker, and a’ domain, respectively.
• G
  Far UV circular dichroism spectra of recombinant PDI proteins as indicated. Data were shown as mean from three independent experiments each performed in five technical replicates.
• H
  Quantification of the enhancement factors of PDI proteins recorded by ANS fluorescence spectra.
• I, J
  Chaperone activities of PDI proteins were measured using GAPDH (I) and rhodanese (J) as substrate, respectively. The chaperone activity of PDI WT was taken as 100%.
• K, L
  Reductase (K) and isomerase (L) activities of PDI proteins. The activity of PDI WT was taken as 100%.

Data information: In (H–L), data were shown as mean ± SEM from three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 (one‐way ANOVA, the post hoc Tukey's HSD test).Source data are available online for this figure.

Protein disulfide isomerase comprises four thioredoxin (Trx) domains: two catalytic domains a and a’, both containing a CGHC active site, separated by two homologous noncatalytic domains, b and b’, and a C‐terminal tail. In addition, there is a short linker region “x” between the b’ and a’ domains. Ser357 is located in the x‐linker of PDI, is well conserved from yeast to human in eukaryotes, and lies within a Fam20C S‐x‐E/pS recognition motif (Fig 2B). Using a generated antibody specifically recognizing phosphorylated Ser357 of PDI (pS³⁵⁷‐PDI) (Fig EV2B), in vitro kinase assays showed that recombinant PDI was phosphorylated by purified Fam20C protein in a time‐dependent manner (Fig 2C). Phosphorylation of PDI WT by Fam20C resulted in a band migrating slightly slower on a regular gel, and this phosphorylation was confirmed to be pS³⁵⁷‐PDI using dual‐color immunoblotting (Fig 2D). Notably, p‐PDI migrated drastically slower on a phostag gel, which can be recognized by anti‐pS³⁵⁷‐PDI antibody. Fam20C also caused the slow migration of PDI S357A mutant on both the regular and phostag gels and was consistent with MS results showing that PDI can be phosphorylated at other sites (Fig 2C). Interestingly, it seems that S357A is a better substrate for Fam20C than the WT PDI, and the lack of intermediate forms on the phostag gel suggests that the phosphorylation of PDI's multiple sites is cooperative.

The pS³⁵⁷‐PDI signal was detected in untreated HepG2 cells and was remarkably weakened when treated with λ‐protein phosphatase (Fig EV2C). Moreover, in two CRISPR/Cas9‐edited clones of PDI KO HepG2 cells (Fig EV2D–F), the pS³⁵⁷‐PDI signal diminished and was restored by expressing HA‐tagged PDI WT but not PDI S357A (Fig EV2G). By using two clones of Fam20C KO HeLa cells (Fig EV2H), we showed that phosphorylation of PDI Ser357 was strictly dependent on Fam20C kinase activity (Fig EV2I). Convincingly, Fam20C is the bona fide kinase catalyzing PDI Ser357 phosphorylation.

Phosphorylation of PDI Ser357 is induced by ER stress

Next, we used different ER stress inducers to investigate the correlation between ER stress and the phosphorylation of PDI. Tg, Tm, and dithiothreitol (DTT, a reducing reagent that reduces protein disulfide bonds) induced PDI phosphorylation in a dose‐dependent manner (Fig EV3A–C). Notably, PDI phosphorylation was triggered by pharmacological ER stress within minutes, while the total amount of PDI protein did not change (Fig 2E–G). The rapid phosphorylation of PDI under ER stress preceded the upregulation of the BiP chaperone, the canonical UPR target, suggesting that PDI phosphorylation is an early response to ER stress. Similarly, Brefeldin A (BFA, an inhibitor of ER‐Golgi traffic) and MG132 (an inhibitor of protein degradation), which also trigger ER stress, can induce PDI phosphorylation (Fig EV3D and E). We also used the reversible SERCA inhibitor cyclopiazonic acid (CPA) and showed that CPA induces PDI Ser357 phosphorylation in a dose‐dependent manner (Fig EV3F).

Figure EV3. PDI Ser357 is phosphorylated by various ER stress inducers.

• A
  Detection of pS³⁵⁷‐PDI in HepG2 cells treated with Tg at the indicated concentrations for 30 min.
• B
  Detection of pS³⁵⁷‐PDI in HepG2 cells treated with Tm at the indicated concentrations for 2 h.
• C
  Detection of pS³⁵⁷‐PDI in HepG2 cells treated with DTT at the indicated concentrations for 30 min.
• D, E
  Detection of pS³⁵⁷‐PDI in HepG2 cells treated with 2.5 μg/ml BFA (D) or 2 μM MG132 (E) for indicated times.
• F
  Detection of pS³⁵⁷‐PDI and p‐IRE1α in HepG2 cells treated with CPA at the indicated concentrations for 12 h.
• G
  Detection of pS³⁵⁷‐PDI and p‐IRE1α in HepG2 cells after 25 μM CPA treatment and washout for indicated times.

Source data are available online for this figure.

The Tg‐induced pS³⁵⁷‐PDI signal was detected neither in PDI KO nor in Fam20C KO cells by immunoblotting, further indicating that phosphorylation of PDI under ER stress conditions was specifically dependent on Fam20C kinase (Fig 2H and I). PDI phosphorylation stoichiometry was quantified by phostag gels, which showed that more than half of the total PDI had been phosphorylated after Tg treatment for 30 min, either in HepG2 cells or in HeLa cells (Fig 2H and I). Considering that PDI is one of the most abundant folding catalysts in the ER lumen, the high stoichiometry of PDI phosphorylation suggests that this post‐translational modification could be functionally important to regulate ER stress.

Phosphorylation of PDI depends on the retention of Fam20C in the ER

Approximately 90% of Fam20C is secreted to the extracellular space, and the intracellular protein is mainly localized in the cis‐Golgi apparatus (Tagliabracci et al, 2012). Therefore, Fam20C‐catalyzed protein phosphorylation is believed to occur in the Golgi lumen or the extracellular space (Tagliabracci et al, 2013; Zhang et al, 2018). To decipher the mechanism of Fam20C‐catalyzed phosphorylation of PDI, we constructed PDI‐mEmerald‐KDEL and Fam20C‐mApple to detect the spatiotemporal interaction between PDI and Fam20C. As expected, in unstressed cells, PDI displayed ER localization, and Fam20C colocalized with the cis‐Golgi resident protein GM130. Interestingly, after a Tg treatment of 30 min, the ER distribution was observed for a significant portion of Fam20C, which showed colocalization with PDI (Pearson correlation coefficient = 0.51 ± 0.02; Fig 3A and B). Importantly, most of the Fam20C was relocated from the ER to the Golgi at 60 min after Tg washout (Fig 3A) and was accompanied by the disappearance of the pS³⁵⁷‐PDI signal (Fig 3C). Similarly, the pS³⁵⁷‐PDI signal disappeared as quickly as 30 min after the washout of CPA, a reversible SERCA inhibitor, suggesting that PDI is highly sensitive to SERCA activity (Fig EV3G). Thus, it seems that Fam20C could sense ER stress and be retained in the ER to phosphorylate PDI. To further test this concept, we pretreated HepG2 cells with cycloheximide (CHX, an inhibitor of protein synthesis) to reduce the protein folding load in the ER and then introduced Tg. Tg‐induced pS³⁵⁷‐PDI levels gradually decreased with the extension of CHX treatment (Fig 3D). Simultaneously, Fam20C no longer showed ER distribution in CHX‐pretreated cells upon Tg introduction (Fig 3E).

Figure 3. Phosphorylation of PDI depends on the retention of Fam20C in the ER .

1. Fluorescent photomicrographs of HepG2 cells co‐transfected with PDI‐mEmerald‐KDEL (green) and Fam20C‐mApple (red). The cells were treated with DMSO (Upper) or 5 μM Tg (Middle) for 30 min. After Tg treatment, the cells were washed with PBS and then returned to culture medium for 60 min (Lower). GM130 was immunostained (blue) as a cis‐Golgi marker. Scale bar = 10 μm.
2. Pearson correlation coefficient calculated from (A). Data were shown as mean ± SEM of N > 80 cells from four independent experiments. ***P < 0.001 (one‐way ANOVA, the post hoc Tukey's HSD test).
3. Detection of pS³⁵⁷‐PDI and UPR signaling in HepG2 cells after Tg treatment and washout for indicated times.
4. HepG2 cells were treated with 25 ng/μl CHX for indicated times and analyzed by immunoblotting for pS³⁵⁷‐PDI and UPR signaling. 5 μM Tg or DMSO was added 30 min before cell harvest.
5. Fluorescent photomicrographs of HepG2 cells co‐transfected with PDI‐mEmerald‐KDEL (green) and Fam20C‐mApple (red) pretreated with 25 ng/μl CHX for 7.5 h, followed by introduction of DMSO (Upper) or 5 μM Tg (Lower) for 30 min, respectively. GM130 was immunostained (blue) as a cis‐Golgi marker. Scale bar = 10 μm.
6. Fluorescent photomicrographs of HepG2 cells co‐transfected with PDI‐mEmerald‐KDEL (green) and Fam20C‐mApple‐KDEL (red) treated with DMSO (Upper) or 5 μM Tg (Lower) for 30 min, respectively. Scale bar = 10 μm.
7. HepG2 cells were transfected with empty vector (–), FLAG‐tagged Fam20C, or Fam20C‐KDEL followed by introduction without or with 5 μM Tg for 30 min, respectively. PDI phosphorylation was detected by protein immunoblotting on both regular and phostag gels. p‐PDI percentage was quantified based on band intensities on phostag gels.

Source data are available online for this figure.

To provide further evidence that Fam20C can catalyze PDI phosphorylation in the ER lumen, we engineered a KDEL ER retrieval sequence to the C‐terminus of Fam20C‐mApple, as this sequence can be sequestrated by the cis‐Golgi‐anchored KDEL receptor. Fam20C‐mApple‐KDEL displayed perfect colocalization with PDI‐mEmerald‐KDEL in the ER at basal state (Fig 3F). pS³⁵⁷‐PDI immunoblotting showed that Fam20C‐KDEL was more efficient at catalyzing PDI phosphorylation than Fam20C WT (Fig 3G, compare lane 5 to lane 3). Interestingly, Fam20C‐mApple‐KDEL formed granular structures 30 min after Tg treatment, and at least some of the PDI proteins were recruited into these condensates (Fig 3F). Accordingly, the phosphorylation level of PDI was largely increased after Tg treatment in cells expressing Fam20C‐KDEL (Fig 3G, compare lane 6 to lane 4 and lane 2), supporting the notion that the Fam20C and PDI‐containing condensates formed during ER stress facilitate the phosphorylation process. Altogether, these results suggested that during ER stress, Fam20C is retained in the ER to catalyze efficient phosphorylation of PDI.

Phosphorylation of PDI induces its functional switch from an oxidoreductase to a molecular chaperone

To address how phosphorylated Ser357 of PDI regulates the conformational change, we performed molecular dynamics (MD) simulation using the previously solved crystal structure of PDI (PDB code 4EL1). After the introduction of phosphorylated Ser357, PDI underwent a significant domain rearrangement and showed a more open conformation (Movie EV1). When the last snapshot of the simulation was superimposed with the crystal structure based on the bb’ domain, the a’ domain rotated around the x‐linker outward toward the other three thioredoxin domains, making p‐PDI an extended “L”‐shaped molecule compared with the previous “U”‐shaped structure (Fig 4A). Time‐resolved analysis showed that the distance between domains a and a’ increased to ~70 Å from the original ~50 Å, and the angle among domains b, b’, and a’ markedly increased to ~170° from the original ~140° (Fig 4B, red). The open conformation of p‐PDI is distinct from the previous PDI simulations with compact conformations, which most likely represent the intrinsic stable substrate‐free state (Fig 4B, black; Yang et al, 2014). Therefore, this open form of p‐PDI may have an important functional influence.

Figure 4. Phosphorylation of PDI induces its functional switch from an oxidoreductase to a molecular chaperone.

• A
  The last snapshot of the molecular dynamic simulation of pS³⁵⁷‐PDI. PDI is composed of four thioredoxin‐like domains a (orange), b (green), b’ (blue), a’ (cyan), and the x‐linker (red). The initial structure of PDI (PDB code 4EL1) is shown in gray. pSer357 and Ser357 are shown in stick.
• B
  The time course of the distance between the geometrical centers of domains a and a′ and the angle among domains b, b′, and a′ from the simulation of pS³⁵⁷‐PDI (red) and PDI (black; Yang et al, 2014).
• C
  (Left) ANS fluorescence spectra of PDI WT and S357E. Controls with ANS alone were shown. A.U.: arbitrary units. (Right) Enhancement factor was quantified.
• D
  Limited proteolysis of PDI WT (Left) and S357E (Right) proteins by proteinase K (Pro K) for indicated times. Main digested fragments were indicated on the right margin. Band abb'x* is a digested product with cleavage after Met356 in the x‐linker region (Wang et al, 2010).
• E
  shCtrl and shFAM20C HepG2 cells were treated without or with 5 μM Tg for 30 min, and in some aliquots of cells, Tg was removed for 30 min. Cell lysates were subjected to proteinase K digestion for 3 min and immunoblotting analysis.
• F
  HepG2 cells expressing HA‐tagged PDI and FLAG‐tagged Fam20C with different ratios of plasmid concentrations were subjected to proteinase K digestion for 21 min and immunoblotting analysis.
• G–J
  Chaperone activities of PDI WT and S357E (G, H) or phosphorylated PDI WT and S357A by Fam20C for indicated times (I, J). Denatured and reduced GAPDH (G, I) and rhodanese (H, J) were used as substrates, respectively. The chaperone activity of PDI WT was taken as 100%.
• K, M
  Reductase activities of PDI WT and S357E (K) or phosphorylated PDI WT and S357A by Fam20C for indicated times (M) measured by insulin reduction assay. The activity of PDI WT was taken as 100%.
• L, N
  Isomerase activities of PDI WT and S357E (L) or phosphorylated PDI WT and S357A by Fam20C for indicated times (N) measured by scrambled RNase A reactivation assay. The activity of PDI WT was taken as 100%.
• O
  Schematic diagram illustrating phosphorylation of PDI Ser357 by Fam20C induces its functional switch from a “foldase” to a “holdase”.

Data information: All data were shown as mean ± SEM of three independent experiments. In (C, G, H, K, L), **P < 0.01, ***P < 0.001 (two‐tailed Student's t‐test). In (I, J, M, N), ***P < 0.001 (two‐way ANOVA, the post hoc Tukey's HSD test).Source data are available online for this figure.

There is a continuous hydrophobic surface on the inner side of the structure of PDI, which can be probed by 1‐anilino‐8‐naphthalene sulfonate (ANS; Wang et al, 2010). The phosphorylation mimic PDI S357E showed much higher ANS fluorescence intensity compared to PDI WT, indicating that a more hydrophobic surface was exposed (Fig 4C). In a limited proteolysis assay, full‐length PDI S357E was quickly digested by proteinase K around the x‐linker region, whereas PDI WT was more resistant, indicating that PDI S357E tends to adopt an open conformation (Fig 4D). Similar results were observed when PDI WT and S357E were digested by trypsin and chymotrypsin (Fig EV4A and B). Moreover, we measured the intrinsic fluorescence of PDI by employing the PDI W128F/W407F mutant, in which the fluorescence is dominated by Trp364 in the x‐linker as an intrinsic reporter of microenvironment changes (Wang et al, 2010). We observed a dramatic fluorescence intensity increase and a red shift of maximum emission in the intrinsic tryptophan fluorescence spectrum of PDI S357E compared to PDI WT, demonstrating that conformational change occurs around the x‐linker in this phosphorylation mimic (Fig EV4C). To investigate the conformational change driven by the phosphorylation of PDI Ser357 in cells, DMSO‐ or Tg‐treated HepG2 cells were lysed and subjected to proteinase K digestion. Clearly, PDI in Tg‐treated cells was more susceptible to proteolytic digestion (Fig 4E). The increased proteinase K sensitivity on PDI is dependent on the presence of Fam20C, and Tg washout can reverse this feature (Fig 4E). Moreover, along with the increase in Fam20C expression, the phosphorylation level of PDI gradually increased and became more susceptible to attack by proteinase K (Fig 4F). All these results indicated that the phosphorylated PDI adopts an open conformation both in vitro and in cells.

We next investigated the effect of phosphorylation on PDI activities. PDI S357E showed significantly higher potency than PDI WT in suppressing the aggregation of denatured GAPDH and rhodanese upon dilution (Fig 4G and H), in line with the previous observation that open forms of PDI display a higher chaperone activity (Wang et al, 2012). We also prepared genuine‐phosphorylated PDI by using the Fam20C kinase assay and observed that PDI WT, but not the S357A mutant, exhibited dramatically increased chaperone activity as the phosphorylation level increased (Fig 4I and J). To our surprise, PDI S357E displayed a remarkable decrease in both reductase and isomerase activities compared to PDI WT as measured by insulin reduction assay (Fig 4K) and scrambled RNase A reactivation assay (Fig 4L). Consistently, Fam20C‐catalyzed phosphorylation largely suppressed the reductase (Fig 4M) and isomerase (Fig 4N) activities of PDI WT but not PDI S357A. We also examined the effects on PDI activities of the other two phosphosites identified by MS, Ser331 (Fig EV4D), and Ser427 (Fig EV4E), which are located in the b’ domain and a’ domain, respectively (Fig EV4F). The Ser331 and Ser427 phosphorylation mimicking mutant showed similar overall structural conformations (Fig EV4G and H), chaperone activities (Fig EV4I and J), and enzymatic activities (Fig EV4K and L) to PDI WT. Taken together, phosphorylation of Ser357 induces PDI to be captured in an open conformation and triggers its functional switch from an oxidoreductase to a molecular chaperone (Fig 4O).

Phosphorylation of PDI Ser357 safeguards ER proteostasis

The fact that phosphorylation of PDI on Ser357 promotes its chaperone activity inspired us to test the possibility that pS³⁵⁷‐PDI regulates ER proteostasis during ER stress. To visualize and detect ER proteome stress in living cells, we generated a Halo‐tag mutant (K73T/L172Q) prone to ER‐localized aggregation based on a previous screen (Liu et al, 2017), by introducing an ER targeting signal sequence and a KDEL ER retrieval sequence (hereafter referred to as AgHalo_ER). The AgHalo_ER sensor was labeled with our developed solvatochromic fluorogenic probe (P1), which turns on fluorescence only upon its misfolding and aggregation. In addition, the AgHalo_ER probe can be combined with commercially available always‐fluorescent ligand (TMR) to enable dual‐color imaging, allowing for direct visualization of the AgHalo_ER sensor in cells both before and after stress conditions (Fig 5A). It is worth noting that fluorescence resonance energy transfer did not occur in two‐color imaging of AgHalo labeled by P1 and TMR (Liu et al, 2018). Confocal fluorescence microscopic imaging confirmed that AgHalo_ER was correctly localized in the ER lumen (Fig 5B). Live cell imaging of HepG2 cells expressing AgHalo_ER labeled by P1 (green) and TMR (red) demonstrated that the AgHalo_ER was well‐folded with little green fluorescence signal under unstressed conditions and formed granular green fluorescent structures after Tg treatment for 30 min (Fig 5C). Thus, the fluorogenic AgHalo_ER‐P1 conjugate provides a direct readout for the facile detection of ER proteome stress.

Figure 5. Phosphorylation of PDI Ser357 safeguards ER proteostasis.

1. (Upper) Schematic representation of ER‐located fluorogenic probe Halo‐tag K73T/L172Q mutant (AgHalo_ER). SP: signal peptide. The carboxyl‐terminal KDEL motif is the ER retention sequence. (Lower) Diagram of AgHalo_ER that detects ER proteostasis in live cells.
2. Fluorescent photomicrographs of HepG2 cells expressing AgHalo_ER (TMR‐labeled, red). Endogenous PDI was immunostained (green) as an ER marker. Scale bar = 10 μm.
3. Confocal live cell imaging of HepG2 cells expressing AgHalo_ER labeled by P1 (green) and TMR (red), followed by introduction of DMSO or 5 μM Tg for 30 min. Scale bar = 10 μm.
4. Confocal live cell imaging of WT and PDI KO HepG2 cells co‐transfected with AgHalo_ER labeled by P1 (green) and TMR (red) and empty vector (–), PDI WT, S357E, or S357A. Scale bar = 10 μm.
5. Confocal live cell imaging of shCtrl and shFAM20C HepG2 cells expressing AgHalo_ER labeled by P1 (green) and TMR (red). Scale bar = 10 μm.
6. (Upper) Cell viabilities of WT and PDI KO HepG2 cells co‐transfected with AgHalo_ER and empty vector (–), PDI WT, S357E, or S357A. Cells were treated by 5 μM Tg for indicated times and visualized by crystal violet staining. (Lower) Quantification of surviving fraction of cells. Data were shown as mean ± SEM of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 versus WT HepG2 (two‐way ANOVA, the post hoc Tukey's HSD test).

To examine the role of PDI phosphorylation in regulating ER proteostasis, we first detected AgHalo_ER signaling in PDI KO HepG2 cells. Depletion of PDI resulted in the accumulation of granular structures of AgHalo_ER with green fluorescence, suggesting that PDI is vital for the correct folding of the AgHalo_ER sensor. Replenishment with PDI WT or S357E, but not the S357A mutant, suppressed the granule formation in PDI KO cells, highlighting the importance of Ser357 phosphorylation in maintaining ER proteostasis (Fig 5D). Furthermore, AgHalo_ER‐P1 fluorescence could also be detected in shFAM20C HepG2 cells, albeit the puncta sizes were smaller than those observed in PDI‐depleted cells (Fig 5E), in line with the observation that cells lacking Fam20C were more sensitive to ER stress (Fig 1). Since the accumulation of misfolded proteins in the ER will imbalance ER proteostasis and even induce cell death if ER stress cannot be resolved, we further studied whether phosphorylation of PDI Ser357 could affect cell viability. A crystal violet staining assay showed that AgHalo_ER‐expressing PDI KO HepG2 cells were more vulnerable than WT cells when subjected to long‐term Tg treatment. Replenishment with PDI WT or S357E in KO cells completely rescued cell viability to the extent of WT cells, whereas expression of S357A had little effect (Fig 5F). Altogether, these results indicated that phosphorylation of PDI Ser357 is critical to maintain ER proteostasis and reduce cell death under ER stress.

Phosphorylated PDI interacts with IRE1α and attenuates IRE1α signaling during ER stress

Because Fam20C phosphorylates PDI and suppresses IRE1α signaling under ER stress, we decided to test the possibility that phosphorylated PDI can directly modulate IRE1α signaling. We first evaluated the levels of XBP1 mRNA splicing in WT and PDI KO HepG2 cells. Compared with control cells, PDI KO cells displayed a higher amplitude of XBP1 mRNA splicing after Tg treatment (Fig 6A). Expression of PDI WT or S357E, but not S357A, decreased XBP1 mRNA splicing in PDI KO cells (Fig 6B). Similar results were obtained when cells were treated with Tm (Fig EV5A and B). Again, PDI S357E selectively suppressed the phosphorylation of IRE1α but did not affect PERK or ATF6 signaling (Fig 6C).

Figure 6. Phosphorylated PDI interacts with IRE1α and attenuates IRE1α signaling during ER stress.

1. (Left) Detection of XBP1 mRNA splicing in WT and PDI KO HepG2 cells treated with or without 5 μM Tg for 1 h. PDI KO was verified by protein immunoblotting. (Right) Quantification of XBP1 mRNA splicing levels. Data were shown as mean ± SEM of three independent experiments. *P < 0.05 (two‐tailed Student's t‐test).
2. (Left) Detection of XBP1 mRNA splicing in PDI KO HepG2 cells expressing PDI WT or its mutants treated with or without 5 μM Tg for 1 h. PDI expression levels were shown by protein immunoblotting. (Right) Quantification of XBP1 mRNA splicing levels. Data were shown as mean ± SEM of three independent experiments. *P < 0.05, **P < 0.01 (one‐way ANOVA, the post hoc Tukey's HSD test).
3. Detection of IRE1α, PERK, and ATF6 activation by immunoblotting in PDI KO HepG2 cells expressing PDI S537A or S357E treated with 5 μM Tg for 30 min. Asterisk indicates unspecific background band.
4. HepG2 cells were treated without or with 5 μM Tg for 1 h. Endogenous IRE1α immunoprecipitates were analyzed by PDI and BiP immunoblotting.
5. HepG2 cells expressing FLAG‐tagged IRE1α were treated without or with 5 μM Tg for 1 h. An aliquot of Tg‐treated cells was then washed with culture medium and chased for 4 h. FLAG immunoprecipitates were analyzed by PDI and BiP immunoblotting.
6. shCtrl and shFAM20C HepG2 cells were treated without or with 5 μM Tg for 1 h. Endogenous IRE1α immunoprecipitates were analyzed by PDI and BiP immunoblotting.
7. Co‐immunoprecipitation of FLAG‐tagged IRE1α and HA‐tagged PDI in HepG2 cells expressing V5‐tagged Fam20C WT or DA.
8. Co‐immunoprecipitation of FLAG‐tagged IRE1α, HA‐tagged PDI WT, or S357A in HepG2 cells expressing V5‐tagged Fam20C.
9. Co‐immunoprecipitation of HA‐tagged PDI and FLAG‐tagged IRE1α^NLD in HepG2 cells.
10. Recombinant GST‐PDI was phosphorylated by Fam20C protein for 2 h and then used to pulldown His‐IRE1α^NLD protein, as visualized by Coomassie blue staining. Successful phosphorylation of GST‐PDI was verified by pS³⁵⁷‐PDI immunoblotting.
11. Co‐immunoprecipitation of FLAG‐tagged IRE1α^NLD WT or its mutant C109/148/332S with HA‐tagged PDI (Left) or MYC‐tagged P5 (Right) in HepG2 cells.
12. Co‐immunoprecipitation of FLAG‐tagged IRE1α and HA‐tagged PDI WT or its binding mutant W128I/L403W in HepG2 cells expressing V5‐tagged Fam20C or not.

Source data are available online for this figure.

Figure EV5. Phosphorylated PDI interacts with IRE1α and attenuates IRE1α signaling during ER stress.

1. (Upper) Detection of XBP1 mRNA splicing in WT and PDI KO HepG2 cells treated with or without 2 μg/ml Tm for 8 h. (Lower) Quantification of XBP1 mRNA splicing levels. Data were shown as mean ± SEM of three independent experiments. *P < 0.05 (two‐tailed Student's t‐test).
2. (Upper) Detection of XBP1 mRNA splicing in PDI KO HepG2 cells expressing PDI WT or its mutants treated with or without 2 μg/ml Tm for 8 h. (Lower) Quantification of XBP1 mRNA splicing levels. Data were shown as mean ± SEM of three independent experiments. ***P < 0.001 (one‐way ANOVA, the post hoc Tukey's HSD test).
3. Co‐immunoprecipitation of FLAG‐tagged IRE1α and HA‐tagged PDI in HepG2 cells.
4. Co‐immunoprecipitation of FLAG‐tagged IRE1α and HA‐tagged PDI S357A or S357E in HepG2 cells.
5. GST‐pulldown of recombinant His‐IRE1α^NLD protein with GST‐PDI or GST protein.
6. (Left) Co‐immunoprecipitation of FLAG‐tagged IRE1α and HA‐tagged PDI WT or its substrate binding mutants. (Right) Domain organization of the human PDI molecule. Mutations that destroy substrate binding are indicated.

Source data are available online for this figure.

Next, we performed co‐IP assays to analyze the interaction between PDI and IRE1α. The PDI‐IRE1α interaction was detected by reciprocal co‐IP in cells co‐expressing HA‐PDI and IRE1α‐FLAG (Fig EV5C). Importantly, the interaction between endogenous PDI and IRE1α was also detected in cells and was induced under ER stress (Fig 6D). Interestingly, the binding of PDI to IRE1α was first induced by Tg and then reduced after removal of Tg, correlating with the release of BiP from IRE1α and rebinding (Fig 6E). The ER stress‐related PDI‐IRE1α association matches PDI phosphorylation dynamics (Fig 3C), and PDI S357E showed a stronger interaction with IRE1α than PDI S357A (Fig EV5D), implying that phosphorylation of PDI promotes its binding with IRE1α. Consistent with this idea, depletion of Fam20C greatly reduced the Tg‐induced PDI‐IRE1α interaction (Fig 6F). In addition, overexpression of Fam20C WT, but not the inactive D478A mutant, enhanced the phosphorylation level of PDI and the interaction between PDI and IRE1α (Fig 6G). Importantly, the enhancement of the PDI‐IRE1α interaction relied on the phosphorylation of PDI Ser357 by Fam20C (Fig 6H). PDI is an ER lumen protein; co‐IP (Fig 6I) and GST‐pulldown (Fig EV5E) assays supported that PDI directly binds to the N‐terminal lumenal domain of IRE1α (IRE1α^NLD). Furthermore, we generated phosphorylated GST‐PDI by the Fam20C kinase assay and found that the binding affinity of PDI and IRE1α^NLD is largely increased by phosphorylation (Fig 6J).

PDI can interact with client proteins through disulfide bridges or hydrophobic interactions. To explore whether PDI interacts with IRE1α in a cysteine‐dependent manner, we substituted all three cysteines in the lumenal domain of IRE1α with serine residues. The IRE1α C109/148/332S mutations did not affect the interaction with PDI, suggesting that the PDI‐IRE1α association is cysteine‐independent (Fig 6K, Left). By contrast, another PDI family member, P5, interacted with IRE1α in a cysteine‐dependent manner as previously reported (Eletto et al, 2014; Fig 6K, Right). There are three binding sites in PDI domains, a, b’, and a’, contributing to efficient collagen prolyl 4‐hydroxylase tetramer assembly (Koivunen et al, 2005) and platelet α_IIbβ₃ integrin binding (Wang et al, 2019). We then employed PDI binding mutants that have been previously used to study the molecular details of the PDI‐IRE1α interaction. The single mutants W128I and L403W (with the binding site in the a and a’ domains being disrupted, respectively) rather than I289A (with the binding site in the b’ domain being disrupted) impaired the interaction between PDI with IRE1α, and the W128I/L403W double mutant almost completely abolished the PDI‐IRE1α interaction (Fig EV5F). Importantly, the PDI W128I/L403W double mutant was still unable to interact with IRE1α when Fam20C was co‐expressed, albeit Ser357 in the double mutant was sufficiently phosphorylated (Fig 6L). Collectively, these results suggest that upon ER stress, phosphorylated PDI binds to the lumenal domain of IRE1α via noncovalent hydrophobic interactions and suppresses IRE1α activation.

Phosphorylation of PDI protects against ER stress‐induced liver damage in vivo

The above finding that p‐PDI negatively regulates IRE1α signaling could be physiologically important because excessive activity of IRE1α under ER stress can lead to cellular apoptosis and diseases. To reveal the potential protective role of p‐PDI under pathophysiological conditions, we generated a Pdi ^S359A/S359A knock‐in (KI) mouse by CRISPR/Cas9‐mediated genome editing (Figs 7A and EV6A). Ser359 of mouse PDI is homologous to Ser357 of human PDI (Fig 2B), and S359A KI was validated by DNA sequencing (Fig 7B). The Pdi ^S359A/S359A homozygotes were born at Mendelian ratios and had no phenotypic abnormalities from birth through adulthood, in line with the results that the human PDI S357A is as active as the WT protein (Fig 4). For acute ER stress, mice were injected intraperitoneally with a concentration of 50 ng Tm per gram of animal (Sepulveda et al, 2018). Kinetic analysis of Xbp1 mRNA splicing in the liver revealed that both the strength and duration of IRE1α signaling were increased in the Pdi ^S359A/S359A mice compared to the littermate Pdi ^+/+ mice (Fig 7C). Higher Xbp1 mRNA splicing levels were also observed in isolated primary hepatocytes from Pdi ^S359A/S359A mice after treatment with Tg (Fig 7D) and Tm (Fig 7E). In accordance, IRE1α was phosphorylated to a higher degree in primary hepatocytes from PDI S359A KI mice under ER stress (Fig EV6B).

Figure 7. Phosphorylation of PDI protects against ER stress‐induced liver damage in vivo .

• A
  Schematic diagram illustrating the generation of Pdi ^S359A/S359A knock‐in (KI) mice by CRISPR/Cas9‐mediated genome editing.
• B
  Sanger sequencing results of Pdi ^+/+ (WT) and Pdi ^S359A/S359A (KI) mice.
• C
  (Left) Detection of Xbp1 mRNA splicing in mice liver after injection of Tm (50 ng/g, i.p.) for indicated times. Each lane represents independent animals. (Right) Quantification of Xbp1 mRNA splicing levels at indicated times.
• D, E
  Detection of Xbp1 mRNA splicing in primary hepatocytes isolated from WT and KI mice treated with 0.2 μM Tg for 2 h (D) or 0.5 μg/ml Tm for 2 h (E).
• F–K
  Detection of Bip (F), Pdi (G), Erdj4 (H), Edem1 (I), Il6 (J), Chop (K) mRNA levels by real‐time PCR in the liver of the same animals as in (C) after 24 h of Tm injection.
• L
  Detection of serum alanine transaminase (ALT) levels of the same animals as in (C) after 24 h of Tm injection.
• M
  Hematoxylin and eosin staining of liver tissue of the same animals as in (C) after 24 h of Tm injection (magnification × 400). Immune cell infiltration was indicated with a black arrow. Scale bar = 50 μm. Representative images were shown from five animals per group analyzed.
• N
  (Left) TUNEL assay of the same liver as in (M). Scale bar = 40 μm; zoom box, scale bar = 5 μm. (Right) Quantification of apoptotic cells. 10 random fields were counted for each TUNEL‐stained tissue sample.
• O
  Proposed model for the role of p‐PDI. (Left) Under basal conditions, PDI is responsible for the catalysis of oxidative protein folding, and IRE1α is maintained by BiP chaperone in a monomeric inactive state. (Middle) Upon ER stress, Fam20C kinase is retained in the ER for rapid phosphorylation of PDI. Phosphorylated Ser357 induces an open conformation of PDI and turns it from a “foldase” to a “holdase”, which is critical for preventing protein misfolding in the ER. Meanwhile, BiP dissociates and IRE1α is activated for XBP1 splicing. p‐PDI directly binds to IRE1α to control the amplitude of IRE1α activity for adaptive UPR signaling. (Right) In PDI S357A mutant, protein aggregates form under ER stress due to incapable phosphorylation of PDI, and sustained ER stress induces hyperactivation of IRE1α, which leads to terminal UPR and apoptosis.

Data information: All data were shown as mean ± SEM from five biological replicates. In (C, N), *P < 0.05, ***P < 0.001 (two‐tailed Student's t‐test). In (F–L), *P < 0.05, ***P < 0.001 (two‐way ANOVA, the post hoc Tukey's HSD test).Source data are available online for this figure.

Figure EV6. Generation and characterization of Pdi ^S359A/S359A mice.

• A
  Schematic diagram illustrating the single‐guide RNAs (sgRNAs) targeting intron 4 and intron 10 of mouse Pdi gene locus.
• B
  Detection of IRE1α, eIF2α, and PDI phosphorylation by immunoblotting in primary hepatocytes isolated from WT and Pdi ^S359A/S359A knock‐in (KI) mice treated with 5 μM Tg for 30 min.
• C, D
  Detection of serum creatinine (Cr) (C) and blood urea nitrogen (BUN) (D) levels of the same animals as in Fig 7C after 24 h of Tm injection. Data were shown as mean ± SEM from five biological replicates.
• E, F
  Hematoxylin and eosin staining of liver tissue of the same animals as in Fig 7C after 24 h of Tm injection (magnification × 200). Liver congestion was indicated with arrowheads (F). Scale bar = 100 μm. Representative images were shown from five animals per group analyzed.

Source data are available online for this figure.

We also compared the expression of other downstream UPR signaling genes in Pdi ^S359A/S359A and Pdi ^+/+ mice. The mRNA levels of XBP1 targeting genes Bip (Fig 7F), Pdi (Fig 7G), Erdj4 (Fig 7H), and Edem1 (Fig 7I) were less affected under basal conditions but were sharply induced after 24 h of Tm injection, confirming enhanced IRE1α signaling in PDI S359A KI mice. The proinflammatory cytokine Il6 (Fig 7J) and proapoptotic transcription factor Chop (Fig 7K) were also upregulated in PDI S359A KI mice after Tm administration, implying that more cell apoptosis occurs. Indeed, the serum alanine transaminase (ALT) level, an indicator of liver injury, was significantly elevated in Pdi ^S359A/S359A mice under ER stress (Fig 7L). Although the kidney injury biomarkers creatinine (Cr) (Fig EV6C) and blood urea nitrogen (BUN) (Fig EV6D) were not significantly different between PDI S359A KI and WT mice, there seems to be a tendency of kidney injury in Pdi ^S359A/S359A mice under ER stress. The extent of liver damage was further analyzed by hematoxylin and eosin staining of liver sections from Tm‐challenged mice. Massive hepatocyte damage was observed, including edema and vacuolization (Fig EV6E), inflammatory cell infiltration (Fig 7M), and hemorrhage (Fig EV6F). The levels of cell apoptosis were also monitored by using the TUNEL assay. As expected, only occasional apoptotic nuclei were seen in WT mice with the low dose of Tm administrated, whereas a significant increase in apoptotic cells (> 15%) was observed in the livers of Tm‐challenged Pdi ^S359A/S359A mice (Fig 7N). In summary, phosphorylation of PDI attenuates the excessive activity of IRE1α and plays a prosurvival role in protecting against ER stress‐induced liver damage.

Discussion

In this study, we demonstrate that PDI, a key folding catalyst in the ER, is rapidly phosphorylated by the secretory pathway kinase Fam20C under ER stress, and the p‐PDI with increased chaperone activity is important for maintaining ER proteostasis and tuning the amplitude of IRE1α signaling (Fig 7O). In contrast to the canonical UPR pathways with inherent latency acting at transcriptional and translational levels, post‐translational modification promptly responds to the proteostatic perturbations in the ER through preexisting components (Preissler & Ron, 2018). Recent reports on the regulation of the ER Hsp70 chaperone BiP at the post‐translational level have set such an example. BiP is AMPylated when the folding requirement in the ER is low, and the reversible deAMPylation recruits BiP back into the chaperone cycle under ER stress (Preissler et al, 2015, 2017a,b). Thus, both the phosphorylation of PDI and the deAMPylation of BiP are early responses to ER stress by simply increasing the protein folding capacity at the post‐translational level. Because PDI and BiP are both high‐abundance proteins in the ER, their post‐translational regulation can well match the accumulating load of misfolded/unfolded proteins. However, whether there is a crosstalk between post‐translational regulation of PDI and BiP is still an open question.

Phosphorylated PDI not only suppresses the aggregation of misfolded proteins but also binds to the lumenal domain of IRE1α and limits its activation. The amplitude and duration of IRE1α activation determines cell fate under ER stress, and the precise control of IRE1α activity is of vital importance. Previous studies revealed that IRE1α activity can be modulated either from the lumenal side by BiP (Bertolotti et al, 2000; Kimata et al, 2003; Bakunts et al, 2017), P5 (Eletto et al, 2014), and Hsp47 (Sepulveda et al, 2018) chaperones or from the cytosolic side by BAK and BAX (Hetz et al, 2006), BAX inhibitor‐1 (Lisbona et al, 2009), PUMA and BIM (Rodriguez et al, 2012), ABL (Morita et al, 2017), and MITOL (Takeda et al, 2019). Our results show that the recruitment of p‐PDI to IRE1α is induced by Tg addition and then reduced after Tg removal, coinciding with the release and binding between BiP and IRE1α. Thus, it seems that p‐PDI acts as a reservoir for the precise control of IRE1α signaling from the lumenal side, adding an additional layer of complexity of IRE1α regulation.

Protein disulfide isomerase has been recognized as both an oxidoreductase and a molecular chaperone for a long time (Wang & Tsou, 1993; Wang et al, 2015). Nevertheless, the molecular mechanism for the adjustment of the two activities is elusive. Here, we report that phosphorylation of Ser357 in the x‐linker region of PDI induces PDI from a “foldase” to a “holdase”. This functional switch is important to prevent proteotoxicity under ER stress, probably because the prevention of misfolded protein aggregation is of higher priority than normal oxidative protein folding in the stressed ER. Indeed, similar mechanisms have been reported not only for phospho‐regulated chaperones (Velasco et al, 2019) but also for redox‐regulated chaperones and acid‐activated chaperones (Voth & Jakob, 2017). Our MD and biochemical analyses indicate that the functional switch of PDI is based on the relocation of the x‐linker region and that p‐PDI tends to adopt an open conformation distinct from the relatively closed conformation of nonphosphorylated PDI. This open conformation has a more exposed hydrophobic surface that favors unfolded/misfolded substrate binding but may disrupt the complete catalytic cycle of PDI. Indeed, MD simulation revealed that the distance between the two active sites in p‐PDI increased to 70 Å from the original 40.3 Å, and this increased distance may impair the cooperation of the two active sites during catalysis. The phosphorylation‐induced conformational change is in line with previous studies that the x‐linker of PDI can be trapped in either a “capped” or an “uncapped” conformation by mutagenesis (Nguyen et al, 2008; Wang et al, 2010) or small molecule ligand (Bekendam et al, 2016). Collectively, we propose that PDI is a novel phospho‐regulated chaperone against ER stress.

In this study, we demonstrate that Fam20C plays an important role in regulating ER proteostasis and UPR signaling, in addition to its previously identified regulatory role in regulating redox homeostasis (Zhang et al, 2018) and Ca²⁺ homeostasis (Pollak et al, 2018). Thus, Fam20C‐mediated phospho‐regulation could be a general mechanism to maintain ER homeostasis. Fam20C‐catalyzed protein phosphorylation was believed to occur in the lumen of Golgi and/or extracellularly (Tagliabracci et al, 2013; Zhang et al, 2018). However, we show that Fam20C is at least partially retained in the ER during ER stress and could form condensates with PDI, resulting in higher local concentrations for catalyzing efficient phosphorylation processes. Moreover, ER stress also induces an elevated ATP level in the ER via AXER (an ATP/ADP exchanger in the ER membrane; Klein et al, 2018), which may also facilitate Fam20C‐catalyzed phosphorylation. An unresolved question is how Fam20C senses the environmental changes in the ER to phosphorylate certain clients. We notice from our Fam20C interactome data that the interaction between Fam20C and ERGIC2 significantly decreased after Tg treatment. ERGIC2, together with ERGIC3, functions as a cargo receptor in protein trafficking between the ER and Golgi (Otte et al, 2001; Orci et al, 2003). Thus, we speculate that the transport of Fam20C from the ER to Golgi might be mediated by ERGIC2/3 and that Fam20C is retained in the ER under ER stress, possibly due to its decreased interaction with ERGIC2/3. Moreover, we observed that the ER‐retained Fam20C is rapidly transported to the Golgi after Tg washout and that p‐PDI diminishes simultaneously. The decrease in p‐PDI could either be catalyzed by an unidentified protein phosphatase in the ER or achieved by selective degradation through the ER‐associated degradation or the autophagic pathway (Cha‐Molstad et al, 2015).

Our finding that the Fam20C‐PDI axis negatively regulates IRE1α signaling is physiologically relevant because terminal UPR is associated with many diseases, such as diabetes, cancer, and neurodegeneration (Oakes & Papa, 2015). Indeed, under Tm‐induced acute ER stress, the Pdi ^S359A/S359A mice had elevated levels of the proinflammatory cytokine Il6 and the proapoptotic transcription factor Chop as well as liver damage. In line with our results, CHOP mRNA is robustly induced by Tg stimulation in Fam20C‐depleted U2OS cells, and cardiomyocyte‐specific Fam20c knockout causes accelerated deterioration of cardiac function in mice following various pathological stimuli (Pollak et al, 2018). Recently, Fam20C has been reported to be significantly elevated in the islet β cells of diabetic mice (Kang et al, 2019). Thus, future efforts to validate the function of the Fam20C‐PDI axis under pathophysiological conditions, such as diabetes, are particularly important for providing a new therapeutic strategy to combat ER stress‐associated diseases.

Materials and Methods

Antibodies, chemicals, peptides, recombinant proteins, and DNA

The sources and identifiers of antibodies, chemicals, peptides, recombinant proteins, and DNA used in this paper can be found in Table EV1.

Cell culture

HepG2 cells were cultured in RPMI medium modified without calcium nitrate with 2.05 mM l‐glutamine (HyClone) supplemented with 10% fetal bovine serum (Gibco). HeLa cells were cultured in Dulbecco's modified Eagle's Medium (DMEM) (HyClone) supplemented with 5% fetal bovine serum. All media were supplemented with 100 μg/ml streptomycin and 100 U/ml penicillin (Invitrogen), and the cells were cultured at 37°C with 5% CO₂. Plasmid transfection was accomplished by using ViaFect (Promega) according to the manufacturer's instructions.

Construction of PDI S359A knock‐in mice

Mice harboring PDI S359A mutation were generated by CRISPR/Cas9‐mediated genome editing (Biocytogen). In brief, the S359A mutation was introduced in Pdi exon 8 using an overlap extension‐PCR method. Homology regions covering ~2 kb upstream of Pdi exon 5 and ~2 kb downstream of exon 10 were subcloned into the targeting vector. Two single‐guide RNAs (sgRNAs) targeting intron 4 (5′‐TATGGTTTAGGCAATGACAA‐3′) and intron 10 (5′‐AGATCTATACCTAGGAAGCT‐3′) were designed. Cas9/sgRNA and targeting vector were microinjected into C57BL/6J oosperms and implanted into pseudopregnant females. Mice (Pdi ^+/S359A) carrying the recombined allele containing the S359A mutation were genotyped by PCR followed by sequencing using primers (5′‐GTACTAGCCTAGCCATGCACCAAGG‐3′, 5′‐AAGGGGCATCTGAAAGAGAGCAGTG‐3′). Southern blot was performed to identify the proper integration of recombined loci in targeted mice. Heterozygotic mice (Pdi ^+/S359A) were intercrossed to generate homozygotic mice (Pdi ^S359A/S359A). All these mice were maintained in specific pathogen‐free conditions and housed separately during the experiments. For acute ER stress, 16‐ to 20‐week‐old male mice were injected intraperitoneally with tunicamycin (Tm, Abcam) or dimethylsulfoxide (DMSO, Sigma‐Aldrich) for different times. Blood was collected via the inferior vena cava (IVC) in anesthetized mice for serum biochemical analysis. The liver was frozen at −80°C for biochemical analysis, and the right major lobe of the live was fixed in 4% paraformaldehyde (NOVON) for 72 h for histological analyses. Animal experiments were conducted with the approval of the Institutional Biomedical Research Ethics Committee of the Institute of Biophysics, Chinese Academy of Science.

Primary hepatocyte isolation and culture

Mice were anesthetized, and abdominal cavity was dissected. The IVC was cannulated with a 24‐gauge 3/4‐inch angiocatheter (BD Biosciences), and the portal vein was cut. The liver was perfused via the IVC with 20 ml KRG buffer (120 mM NaCl, 20 mM NaHCO₃, 20 mM glucose, 5 mM HEPES, pH 7.4, 5 mM KCl, 1 mM MgSO₄, 1 mM KH₂PO₄) with 0.5 mM EGTA at 37°C, followed by perfusion with 20 ml of 25 mg/ml collagenase type 4 (Sigma‐Aldrich) in KRG buffer. After the liver was digested, it was cut to release the hepatocytes, passed through a 75‐μm cell strainer, and washed with ice‐cold DMEM until the supernatant appear clear after spinning. The pellet was suspended with DMEM supplemented with 10% fetal bovine serum and cultured at 37°C with 5% CO₂.

Generation of PDI knockout HepG2 cell lines

CRISPR/Cas9 genome editing was performed to generate PDI knockout HepG2 cell lines. Briefly, the guide sequence targeting human PDI (5′‐GCGGAAAAGCAACTTCGCGG‐3′) was designed using the CRISPR design tool (http://chopchop.cbu.uib.no/). HepG2 cells were transfected with pSpCas9 (BB)‐2A‐GFP vector (Laboratory of Feng Zhang, Addgene) containing the sgRNA for 48 h, and GFP‐positive cells were single‐cell‐sorted into a 96‐well plate format containing RPMI medium by fluorescence‐activated cell sorting (BD Influx). Expanded single clones were screened for PDI knockout by protein immunoblotting and DNA sequencing. The following primers were used to amplify the region surrounding the protospacer adjacent motif (PAM): 5′‐CAGGATTTATAAAGGCGAGGC‐3′, 5′‐CTCACAGAACTCCACCAGCA‐3′.

RNA interference

To generate stable Fam20C knockdown HepG2 cells, cells were transfected with pSUPER plasmid (OligoEngine) expressing the short‐hairpin RNA (shRNA) targeting FAM20C sequence 5′‐GAGCTGTACTCCAGACACA‐3′ by ViaFect according to the manufacturer's instructions. Puromycin (InvivoGen) was added into the culture medium to a final concentration of 2 μg/ml to kill the negative cells. Cells and culture medium were harvested and analyzed by immunoblotting to confirm the knockdown efficiency.

RNA isolation, RT–PCR, and real‐time PCR

Total RNA was isolated from cells and tissues using TRIzol (Invitrogen), and RNA samples were then reverse‐transcribed into cDNA using GoScript Reverse Transcription System (Promega). Quantitative real‐time PCRs employing SYBR Select Master Mix (Applied Biosystems) were performed in the QuantStudio 7 Flex machine (Applied Biosystems), following the manufacturer's instructions. The relative amounts of mRNAs were calculated from the values of comparative threshold cycle by using Actb as a control. The methods for the XBP1 mRNA splicing assay were performed as previously described (Upton et al, 2012). In brief, XBP1 mRNA was reverse‐transcribed, PCR‐amplified, and then resolved on 2.5% agarose (Sigma‐Aldrich) gels. PCR primers were described in Table EV2.

Protein purification

Recombinant human PDI protein and its mutants were expressed in Escherichia coli BL21 (DE3) cells and purified as described (Wang et al, 2010). Recombinant IRE1α^NLD (24–390) protein was expressed in Escherichia coli BL21 (DE3) cells and purified as described (Zhou et al, 2006). Human Fam20C (141–578) protein was expressed in Hi5 insect cells and purified from the conditioned medium (Xiao et al, 2013). Proteins were quantified spectrophotometrically at 280 nm and stored at −80°C in aliquots.

In vitro kinase assay

1 mg/ml recombinant PDI WT or S357A protein and 20 μg/ml Fam20C protein were incubated in HEPES, pH 7.0, 10 mM MnCl₂ at 30°C. The reactions were initiated by adding ATP (Sigma‐Aldrich) to a final concentration of 2 mM and were terminated at the indicated time points by adding SDS‐loading buffer containing 15 mM EDTA. Reaction products were separated by SDS–PAGE and visualized by Coomassie blue staining and immunoblotting.

λ‐phosphatase assay

HepG2 cell lysates were treated with indicated units λ‐protein phosphatase (NEB) in 50 mM HEPES, pH 7.5, 100 mM NaCl, 2 mM dithiothreitol (DTT, Sigma‐Aldrich), and 0.01% Brij 35 with 10 mM MnCl₂ at 30°C for 1 h, followed by SDS–PAGE and immunoblotting.

Fluorescence spectra

Intrinsic fluorescence spectra of 5 μM PDI or its mutants in 50 mM Tris–HCl, pH 7.6, 150 mM NaCl were recorded at 310–400 nm at 25°C with excitation at 290 nm. For ANS fluorescence spectra, 50 μM ANS was incubated with or without 5 μM PDI proteins in 50 mM Tris–HCl, pH 7.6, 150 mM NaCl for 20 min at 25°C in the dark. ANS fluorescence emission spectra at 400‐650 nm were measured with excitation at 370 nm. The concentration of ANS was determined using an extinction coefficient at 350 nm of 5,000 M⁻¹ cm⁻¹. Enhancement factor = [F (protein + ANS + buffer) – F (protein + buffer)]/[F (ANS + buffer) – F (buffer)], where F is the fluorescence intensity at 480 nm. Fluorescence spectra were recorded by using RF‐5301PC Spectrofluorophotometer (SHIMADZU).

Circular dichroism spectra

Circular dichroism spectra of 2.5 μM PDI or its mutants in 20 mM sodium phosphate buffer, pH 7.5, were measured at 190–260 nm at 25°C by Chirascan Plus (Applied Photophysics). Scanning speed was 2.5 s/dot.

Chaperone activity assay

Rabbit muscle glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH) was purified as described before (Cai et al, 1994). Denaturation of GAPDH was carried out by incubation of 0.14 mM GAPDH in 3 M guanidine hydrochloride with 1 mM DTT overnight at 4°C. Refolding was initiated by 50‐fold dilution of the denatured GAPDH into 100 mM sodium phosphate, pH 7.4, 2.5 mM EDTA with or without 7 μM PDI or its mutants at 25°C. GAPDH aggregation was monitored by recording the light scattering at 488 nm for 30 min. Denaturation of rhodanese (Sigma‐Aldrich) was carried out by incubation of 45 μM rhodanese in 6 M guanidine hydrochloride with 10 mM DTT for 1 h at room temperature (RT). Refolding was initiated by 100‐fold dilution of the denatured rhodanese into 200 mM sodium phosphate, pH 7.5, with or without 1 μM PDI or its mutants at 25°C. Rhodanese aggregation was monitored by recording the light scattering at 350 nm for 30 min. Light scattering was recorded by using RF‐5301PC Spectrofluorophotometer (SHIMADZU). The chaperone activities of PDI proteins were calculated as (I ₀ − I)/(I ₀ − I ₁) × 100%, where I ₀, light scattering intensity determined in the absence of PDI; I and I ₁, the values determined in the presence of PDI mutants and PDI WT, respectively.

Reductase activity assay

130 μM insulin (Sigma‐Aldrich) was added to 0.1 M potassium phosphate buffer, pH 7.5, 2.5 mM EDTA, 0.1 mM DTT in the absence or presence of 0.5 μM PDI or its mutants to initiate the reaction, and the absorbance at 650 nm that represents light scattering from reduced and precipitated insulin B chains was immediately recorded at 25°C using UV‐2700 Spectrophotometer (SHIMADZU). The reductase activity was calculated as described (Wang et al, 2010) and normalized to that of PDI WT.

Isomerase activity assay

Preparation of scrambled bovine pancreatic RNase A (Sigma‐Aldrich) (sRNase A) was according to Lyles and Gilbert (1991). 8 μM sRNase A was reactivated in 100 mM Tris–acetate, pH 8.0, 50 mM NaCl, 1 mM EDTA, 1 mM GSH, and 0.2 mM GSSG in the absence or presence of 3 μM PDI proteins at 25°C. The reactivation of sRNase A was assayed quantitatively by monitoring the absorbance increase at 296 nm due to the hydrolysis of 2′,3′‐cyclic CMP (Sigma‐Aldrich), and the concentration of reactivated RNase A was calculated as described (Wang et al, 2009) and normalized to that of PDI WT.

Limited proteolysis assay

In vitro

Proteinase K, trypsin, or chymotrypsin (Amresco) was added to a final concentration of 5 μg/ml with 1 mg/ml PDI or its mutants in 50 mM Tris–HCl, pH 7.6, for digestion at 25°C. The reactions were terminated at different times by adding 0.5 mM phenylmethanesulfonylfluoride (PMSF) and analyzed by SDS–PAGE and Coomassie blue staining.

In cells

HepG2 cells were lysed by 50 mM Tris–HCl, pH 7.6, 150 mM NaCl, 5 mM KCl, 1% NP‐40 with phosphatase inhibitor cocktail (Roche) for 30 min on ice. Proteinase K was added to a final concentration of 100 μg/ml with 6 mg/ml total lysates. The reactions were terminated at indicated times by adding 0.5 mM PMSF and analyzed by SDS–PAGE and immunoblotting.

Immunoblotting and phostag gels

Harvested cells were washed with ice‐cold phosphate‐buffered saline (PBS) and then lysed in radioimmunoprecipitation assay (RIPA) lysis buffer (50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 0.25% deoxycholic acid, 1% NP‐40, 1 mM EDTA) (Millipore) containing phosphatase and protease inhibitor cocktails (Roche) for 30 min on ice. To remove cell debris, cell lysates were centrifuged at 16,000 × g for 15 min. Protein concentration of cell lysates was quantified by BCA method (Beyotime), and the same amount of proteins was loaded and separated by SDS–PAGE and then transferred to polyvinylidene fluoride (PVDF) membranes (pore size 0.45 μm). The membranes were blocked in TBST buffer (50 mM Tris pH 8.0, 150 mM NaCl, 0.05% Tween‐20) containing 5% (w/v) skimmed milk or 5% (w/v) bovine serum albumin (BSA) (for phosphor‐protein immunoblotting). After blocking, the membranes were washed in TBST and incubated with various primary antibodies at 4°C overnight and followed by incubated with horseradish peroxidase (HRP) or fluorescent‐labeled secondary antibodies at RT for 2 h and then visualized by using a ChemiScope Mini imaging system (Clinx Science) with enhanced chemiluminescence (Thermo Fisher) or by an Odyssey CLx infrared imager (LICOR).

For phostag gels, 50 μM phostag acrylamide (NARD) and 100 μM MnCl₂ were included in the gel recipe according to the manufacturer's instructions. Phostag gels were washed in transfer buffer supplemented with 10 mM EDTA for three times before transferring to PVDF membranes.

Immunofluorescence

Transfected and treated HepG2 cells were harvested and seeded on a glass bottom cell culture dish (NEST), washed with PBS for three times and fixed with 4% paraformaldehyde for 15 min, then permeabilized with 0.3% Triton X‐100 for 5 min and blocked with 5% BSA for 30 min at RT. Cells were incubated with primary antibodies at 4°C overnight and fluorescent‐conjugated secondary antibodies in dark at 37°C for 1 h. The cells were rinsed with Hank's balanced salt solution and analyzed by confocal laser scanning microscopy (Zeiss, LSM710).

ER proteostasis assay

HepG2 transfectants expressing ER‐localized aggregation‐prone Halo‐tag mutant (K73T/L172Q) (AgHalo_ER) without or with PDI WT or S357A or S357E were seeded on a glass bottom cell culture dish. After 36 h, the spent medium was replaced with fresh RPMI medium containing 5 μM P1 or TMR ligand to label AgHalo_ER protein for 30 min. In case needed, cells were washed with PBS in three times, then cultured with RPMI medium, and further treated with DMSO or 5 μM thapsigargin (Tg, Sigma‐Aldrich) for additional 30 min. Confocal images were obtained using an Olympus FluoView™ FV1200 confocal microscope. The P1 signal was visualized with a blue argon laser (488 nm), and the TMR signal was visualized using a green HeNe laser (568 nm).

Crystal violet staining

HepG2 transfectants expressing AgHalo_ER without or with PDI WT or S357A or S357E were counted and seeded on 6‐well cell culture plates, and treated with 5 μM Tg for indicated times. The cells were washed gently with warmed PBS and then incubated with crystal violet solution for 30 min at RT. The plates were washed by distilled water for three times and then imaged.

Immunoprecipitation and pulldown assay

Transfected and treated HepG2 cells were lysed in RIPA lysis buffer containing phosphatase and protease inhibitor cocktails. For immunoprecipitation of IRE1α‐FLAG, HA‐PDI, and endogenous IRE1α, cell lysates were incubated with anti‐FLAG (Sigma‐Aldrich), anti‐HA (Sigma‐Aldrich), and anti‐IRE1α (CST) antibodies at 4°C overnight, respectively, followed by incubation with protein A+G beads (Beyotime) for 1 h. Beads were washed three times with RIPA buffer and then analyzed by SDS–PAGE and immunoblotting.

For GST‐pulldown assay, 2 μM GST‐PDI protein on glutathione sepharose (GE Healthcare) was incubation with 4 μM IRE1α^NLD protein at RT for 2 h or 4°C overnight. Beads were washed three times with PBS and then analyzed by SDS–PAGE and Coomassie blue staining.

Histological evaluation and TUNEL assay

Liver tissue specimens were fixed in 4% paraformaldehyde, embedded in paraffin, sectioned (5 μM thick), and stained with hematoxylin–eosin or fluorescein‐dUTP by DeadEnd Fluorometric TUNEL System (Promega).

Serum biochemical analysis

Blood samples were kept at RT for 1 h and centrifuged at 2,000 × g for 30 min to obtain sera. Sera were frozen at −80°C for further assay. Serum alanine aminotransferase (ALT), creatinine (Cr), and blood urea nitrogen (BUN) were measured using commercial kits (Nanjing Jiancheng Bioengineering Institute, China) according to the manufacturer's instructions.

Mass spectrometry (MS)

Sample preparation

To identify the Fam20C interactome during ER stress, HepG2 cells were transfected with FLAG‐tagged Fam20C for 36 h and treated with DMSO or 5 μM Tg for 30 min. For PDI phosphosite identification, HepG2 cells were treated with DMSO or 5 μM Tg for 10 min. The cell extracts were incubated with anti‐FLAG or monoclonal mouse anti‐PDI antibody overnight at 4°C, followed by the addition of protein A+G beads for 2 h. The immunoprecipitates were washed by PBS for five times, separated by SDS–PAGE, stained by Coomassie blue, and immunoblotted with polyclonal rabbit anti‐PDI in parallel.

In‐gel digestion of proteins

For Fam20C interactome identification, all bands below Fam20C excluding immunoglobulin chains were excised and cut into small plugs; for PDI phosphosite identification, the band corresponding to PDI was excised. After destaining, reduction (10 mM DTT in 25 mM NH₄HCO₃ for 45 min at 56°C), and alkylation (40 mM iodoacetamide in 25 mM NH₄HCO₃ for 45 min at RT in the dark), the plugs were washed twice with 50% acetonitrile, dried using a SpeedVac, and digested with trypsin in 25 mM NH₄HCO₃ overnight at 37°C to allow complete digestion. The reaction was terminated by adding formic acid to a 1% final concentration.

Liquid chromatography (LC)–MS/MS analysis

The digested peptides were separated on an Acclaim PepMap RSLC C18 capillary column (Reprosil‐Pur C18‐AQ, 3 μm; Dr. Maisch GmbH). A linear acetonitrile gradient was used to elute the bounded peptides at a flow rate of 300 nl/min. The eluate was electrosprayed at a 2.0 kV voltage directly into a Q Exactive mass spectrometer (Thermo Fisher Scientific). In the data‐dependent acquisition mode, the MS data were acquired at a high resolution of 70,000 (m/z 200) across a mass range of 300–1,600 m/z. The top 20 precursor ions were selected from each MS full scan with isolation width of 2 m/z for fragmentation in the HCD collision cell. Subsequently, MS/MS spectra were acquired at a resolution of 17,500 (m/z 200). The dynamic exclusion time was 40 s.

Protein identification

The raw data from Q Exactive were analyzed with Proteome Discoverer 2.2.0.388 (Thermo Fisher Scientific) using SEQUEST HT search engine for protein identification and Percolator for false discovery rate (FDR, < 1%) analysis against a UniProt human protein database (updated 10‐2017). The peptide mass tolerance was set to 10 ppm and the MS/MS mass tolerance to 0.02 Da. The peptide confidence was set as high for peptide filter. Label‐free quantification (LFQ) analysis was performed using consensus mode, and parameters were set as follows: unique and razor used for peptide quantification; precursor abundance based on intensity; total peptide amount used for normalization mode; pairwise ratio for ratio calculation; maximum allowed fold change as 100. For Fam20C interactome identification, proteins with unique peptides ≥ 2 were selected and three independent experiments were performed. Based on DAVID GO term analysis, a total of 173 proteins localized in the ER and Golgi were identified in all three experiments. For PDI phosphosite identification, we selected phosphorylation for serine, threonine, or tyrosine and methionine oxidation as variable modifications and the cysteine carbamidomethylation as a fixed modification. The tandem mass spectra of the matched phosphorylated peptides were manually checked for their validity.

Molecular dynamics simulation

Molecular dynamics (MD) simulations follow similar procedures as reported earlier (Yang et al, 2014). Briefly, the crystal structure of hPDI in the oxidized (PDB code 4EL1) state (Wang et al, 2013) was used as the initial structure. This structure was processed with VMD (Humphrey et al, 1996) to generate the necessary files for MD simulation, and phosphorylated Ser357 residue was used. The protein system was solvated in 128 × 96 × 96 Å³ water box and maintained with 150 mM NaCl salt concentration. The energy minimization and simulations were performed with NAMD (Phillips et al, 2005) and CHARMM forcefield for proteins (MacKerell et al, 1998). The simulation protocol is the same as that in previous study (Yang et al, 2014).

Quantification and statistical analysis

Band intensities on gels and blots were quantified using the software ImageJ. All data were presented as mean, and error bars represent standard error of mean (SEM) from ≥ 3 biological replicates. Statistical analyses were done using GraphPad Prism (v.7.0) or Origin 7.0. The two‐tailed Student's t‐test was used to analyze data between two groups, and the one‐way and two‐way ANOVAs followed by the post hoc Tukey's HSD test were used when more than two groups were present. Statistical significance levels were defined as *P < 0.05; **P < 0.01; ***P < 0.001. All statistical details including number of biological or technical replicates can be found in each figure legend.

Author contributions

JY and TL designed and performed experiments, analyzed the data, and prepared the figures. YL created the AgHalo_ER sensor and P1 ligand. XW and XEW generated PDI and Fam20C KO cells. JZ analyzed the MS data. GS contributed to the histopathological analysis. JL performed the MD simulation. LKW participated in the designs and data analysis. C‐CW and LW designed and supervised the experiments. LW conceived the project, interpreted the data, and wrote the manuscript with the input from JY and TL. All authors approved the final version of the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Dataset EV1

Movie EV1

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 6

Source Data for Figure 7

The EMBO Journal (2020) 39: e103841

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez‐Riverol et al, 2019) partner repository with the dataset identifier PXD016619 (http://www.ebi.ac.uk/pride/archive/projects/PXD016619).