A Genome-wide ER-phagy Screen Highlights Key Roles of Mitochondrial Metabolism and ER-Resident UFMylation

Abstract

Selective autophagy of organelles is critical for cellular differentiation, homeostasis, and organismal health. Autophagy of the ER (ER-phagy) is implicated in human neuropathy but is poorly understood beyond a few autophagosomal receptors and remodelers. By using an ER-phagy reporter and genome-wide CRISPRi screening, we identified 200 high-confidence human ER-phagy factors. Two pathways were unexpectedly required for ER-phagy. First, reduced mitochondrial metabolism represses ER-phagy, which is opposite of general autophagy and is independent of AMPK. Second, ER-localized UFMylation is required for ER-phagy to repress the unfolded protein response via IRE1α. The UFL1 ligase is brought to the ER surface by DDRGK1 to UFMylate RPN1 and RPL26 and preferentially targets ER sheets for degradation, analogous to PINK1-Parkin regulation during mitophagy. Our data provide insight into the cellular logic of ER-phagy, reveal parallels between organelle autophagies, and provide an entry point to the relatively unexplored process of degrading the ER network.

Keywords: CRISPR; ER-phagy; UFMylation; autophagy; endoplasmic reticulum; genome-wide screen; organelle turnover; oxidative phosphorylation; post-translational modification.

Figure 1.. Unbiased Identification of ER-Phagy Regulators by Genome-wide CRISPRi Screening

(A) Schematic of the ER Autophagy Tandem Reporter (EATR) and CRISPR inhibition (CRISPRi) system used for screening. HCT116 cells stably express a doxycycline-inducible EATR construct that consists of mCherry and eGFP fused to ER localized RAMP4. Cells also stably express dCas9-KRAB for sgRNA-targeted transcriptional repression. (B) FACS screening strategy to identify genes whose knockdown enhances or inhibits ER-phagy. HCT116 cells described in (A) are transduced with a genome-wide lentiviral CRISPRi gRNA library. After selection for gRNA expression and removal of essential genes, doxycycline was added to express EATR. Cells were then starved in EBSS for 16 h to induce ER-phagy. The top and bottom quartiles correspond to enhanced and inhibited ER-phagy, respectively. Cells were sorted and processed for next-generation sequencing to identify gRNA representation in each sort bin. (C) Gene ontology analysis identifies autophagy and mitochondrial metabolism as major signatures of ER-phagy. High-confidence ER-phagy genes were defined as having opposite phenotypes in the enhanced and inhibited sort gates and gene level p < 0.01. Ontologies with Benjamini-Hochberg false discovery rate (FDR) <0.05 are shown. (D) Genes involved in ER-phagy form a physical interaction network. For clarity, only interactions between two or more high-confidence hits are shown. Red and blue circles represent genes whose knockdown represses and enhances ER-phagy, respectively. (E) Subcellular classification of high-confidence ER-phagy genes highlights roles in the ER, auto-lysosomes, and mitochondria. See also Figure S1.

Figure 2.. Intact OXPHOS Promotes ER-phagy

(A) Genes that are components of the OXPHOS pathway were top hits in screen (highlighted in blue). Additional mitochondria-related genes are indicated in black and all other targeting sgRNAs are indicated in gray. (B) Knockdown of NDUFB4 and NDUFB2 significantly inhibit ER-phagy. HCT116 CRISPRi EATR cells were transduced with sgRNAs targeting ULK1, NDUFB4, NDUFB2, or ATP5O and starved for 16 h before FACS measurement for ER-phagy. Data are presented as mean ± SD of eight biological replicates. (C) Re-expression of cognate cDNA of each OXPHOS gene rescues ER-phagy. HCT116 CRISPRi EATR cells were transduced with NDUFB4 or NDUFB2 cDNA constructs and then transduced with the indicated sgRNAs. Cells were starved for 16 h before FACS measurement for ER-phagy. Data are presented as mean ± SD of three biological replicates. (D) Small-molecule inhibitors of the different OXPHOS compartments phenocopy the effect of genetic inhibitions. HCT116 EATR cells were treated with rotenone, antimycin A, or oligomycin A and starved for 16 h before FACS measurement of ER-phagy. Data are presented as mean ± SD of three biological replicates. (E) General autophagy proceeds in cells where NDUFB2, NDUFB4, or ATP5O are knocked down. HCT116 CRISPRi cells expressing mCherryeGFP-LC3B were transduced with the indicated sgRNAs. Cells were starved for 4 h before FACS measurement for general autophagy. Data represent mean ± SD of four biological replicates. See also Figures S2 and S3.

Figure 3.. DDRGK1 Specifically Regulates ER-phagy

(A) ER-phagy CRISPRi screen identifies genes that are associated with the ER. Previously reported ER-phagy regulators are highlighted in red. DDRGK1 is highlighted in blue. (B) DDRGK1 depletion results in inhibition of ER-phagy. HCT116 CRISPRi EATR cells were transduced with sgRNAs targeting ULK1 or DDRGK1 and starved for 16 h before FACS measurement for ER-phagy. Data are presented as mean ± SD of three biological replicates. (C) DDRGK1 specifically inhibits ER-phagy but not general autophagy. HCT116 CRISPRi CCER cells were transduced with sgRNAs targeting DDRGK1 and starved for 16 h. Cells were lysed for western blotting of the indicated proteins. (D) HCT116 CRISPRi cells stably expressing mCherry-eGFP-LC3B constructs were transduced with sgRNAs targeting either ULK1, ATG10, or DDRGK1. Cells were starved for 16 h before FACS measurement for general autophagy. (E) DDRGK1 localizes to the ER. HeLa cells were stably transduced with DDRGK1-mCherry construct and immunostained for calnexin (CANX) as an ER marker. Insets represent a 3-fold enlargement of boxed areas. Scale bar represents 10 μm. See also Figure S4.

Figure 4.. DDRGK1-Dependent UFMylation Regulates Autophagy of ER Sheets

(A) Schematic of the three-step enzymatic reaction of the UFMylation cascade. UBA5 (E1) activate UFM1 and UFC1 acts as an E2 enzyme that interacts with the E3 ligase, UFL1. UFL1 recognizes and transfer UFM1 from UFC1 to its target substrate. Asterisk indicates that DDRGK1 is reported as a substrate of UFMylation in the literature. (B) UFL1 knockdown reduces DDRGK1 protein levels. HCT116 CRISPRi cells were transduced with the indicated sgRNAs and then harvested to immunoblot for UFL1 protein levels. (C) UFL1 knockdown phenocopies DDRGK1 knockdown during ER-phagy. The cells generated in (B) were starved for 16 h before FACS measurement for ER-phagy. (D) UFMylation components are required for ER-phagy. HCT116 CRISPRi EATR cells expressing the indicated sgRNAs were starved for 16 h before FACS measurement for ER-phagy. (E) UFL1 controls DDRGK1 protein levels. HCT116 CRISPRi cells were transduced with the indicated sgRNAs and further transduced with either DDRGK1-HA or HA-UFL1. Cell lysates were immunoblotted for DDRGK1 and UFL1. (F) Re-expression of DDRGK1 in UFL1 knockdown cells does not rescue ER-phagy. The cells generated in (E) were starved for 16 h and then subjected to FACS measurement for ER-phagy. Data represent mean ± SD of three biological replicates. (G) DDRGK1 selectively mediates ER sheets degradation. The RAMP4 in EATR system was replaced with either REEP5 (ER tubule marker) or CLIMP63 (ER sheets). The cells were transduced with the indicated sgRNAs and were starved for 16 h before FACS analysis for ER-phagy progression. Data represent mean ± SD of three biological replicates. (H) DDRGK1 depletion selectively affect FAM134B, TEX264, and SEC62-mediated ER-phagy. HCT116 CRISPRi EATR cells were stably transduced with the cDNA for the indicated ER-phagy receptors. ER-phagy induced by overexpression of the ER-phagy receptors at basal state was measured by FACS analysis. Data represent mean ± SD of three biological replicates. (I) DDRGK1 colocalizes with FAM134B and is co-degraded with FAM134B. U2OS cells stably expressing GFP-FAM134B were transiently transfected with mLAMP1-BFP and DDRGK1-mCherry. Cells were then starved for 4 h in the presence of 50 nM folimycin. Scale bar represents 10 μm. Insets represent a 4-fold enlargement of the boxed area. See also Figure S4.

Figure 5.. DDRGK1 Recruits UFL1 to the ER Surface via the PCI Domain

(A) Schematic of DDRGK1 domains and its conserved lysine residues. The reported major lysine residue for UFMylation (K267) is labeled in red. The two truncated forms of DDRGK1 that either lacks the N-terminal transmembrane domain (ΔTM) or the C-terminal proteasome component domain (ΔPCI) are also shown. (B) Post-translational modification of DDRGK1 occurs on lysine residues. Parental or UFL1 knockout HCT116 cells were transfected with either wild-type (WT) or lysine-less (K-less) DDRGK1-HA constructs. Cells were harvested for HA immunoprecipitation. (C) DDRGK1’s role during ER-phagy does not require post-translational modification on any lysine residue. HCT116 CRISPRi EATR cells were transduced with DDRGK1 sgRNA and then rescued using the indicated DDRGK1-HA mutant constructs. Cells were starved for 16 h before FACS ER-phagy measurement. Data represent mean ± SD of three biological replicates. (D) Loss of DDRGK1 relocalizes UFL1 to the cytoplasm. Wild-type or DDRGK1KO HeLa cells were transiently transfected with GFP-UFL1 and mCherry-KDEL for 48 h. Cells were then fixed and immunostained for endogenous DDRGK1. Insets represent a 3-fold enlargement of boxed areas. Scale bar represents 20 μm. (E) DDRGK1 interacts with UFL1 via its PCI domain. Parental HCT116 cells were stably transfected with the indicated DDRGK1-HA mutant constructs. Cells were then harvested for HA immunoprecipitation. (F) DDRGK1 recruits UFL1 to the ER. DDRGK1KO HeLa cells were stably transduced with mCherry-RAMP4 (ER marker) and the indicated DDRGK1-HA mutant constructs. Cells were then transiently transfected with GFP-UFL1 for 24 h. Cells were then fixed and immunostained for HA epitope. Representative images are shown. Insets represent a 3-fold enlargement of boxed areas. Scale bar represents 10 μm. (G) Pearson’s correlation coefficient for (F) was measured between DDRGK1 versus UFL1, DDRGK1 versus RAMP4, and UFL1 versus RAMP4. Data were generated from one biological experiment, and 20–26 cells were analyzed from each condition. (H) DDRGK1’s role during ER-phagy requires both the SP and PCI domains. HCT116 CRISPRi EATR cells with DDRGK1 knockdown were rescued using the indicated DDRGK1-HA mutant constructs. Cells were then starved for 16 h and ER-phagy was measured by FACS analysis. Data represent mean ± SD of three biological replicates. (I) DDRGK1’s determines the subcellular localization of UFL1. DDRGK1 knockout HeLa cells were stably transduced with either TOM20MTS-DDRGK1-dSP-HA (MTS, mitochondrial targeting signal) or PMP34-DDRGK1-dSP-HA and transiently transfected with GFP-UFL1 and the respective mCherry-organelle constructs. Insets represent a 3-fold enlargement of boxed areas. Scale bar represents 10 μm. See also Figure S5.

Figure 6.. DDRGK1 Mediates UFMylation of ER Surface Proteins

(A) Workflow of mass spectrometry identification of DDRGK1 interactome. DDRGK1KO HEK293T cells ± DDRGK1-HA stable expression were starved for 4 h in the presence of 50 nM folimycin. Cell lysates were harvested for HA immunoprecipitation and mass spectrometry identification of co-immunoprecipitated proteins. (B) Workflow of mass spectrometry identification of DDRGK1-dependent UFMylation substrates. UFSP2KO or DDRGK1 and UFSP2 double-knockout (KO) HEK293T cells were transfected with HA-UFM1-ΔCS for 48 h. Cells were starved for 4 h in the presence of 50 nM folimycin. The cells were then lysed and denatured prior to HA immunoprecipitation and mass spectrometry identification of UFMylated proteins. (C) DDRGK1 interacts with several ribosomal subunit proteins. The volcano plot depicts the log2 fold change of the total peptide count of each identified protein between Ctrl (DDRGK1KO) cells and DDRGK1-HA-expressing cells. (D) Selective enrichment of UFMylated proteins in UFSP2KO cells relative to DDRGK1 and UFSP2 double-KO cells. The volcano plot depicts the log2 fold change of the total peptide count of each identified protein between DDRGK1 and UFSP2 KO versus UFSP2KO cells. (E) RPN1 is structurally in close proximity with RPL26. Structural model of the ribosome, oligosaccharide transferase (OST), and SEC61 complex generated from Protein Data Bank deposition 6FTG using PyMol (Braunger et al., 2018). RPN1 (orange) is part of the ER-localized OST complex (blue), whereas RPL26 (red) is a component of the large 60S ribosomal subunit (gray). The OST complex and the ribosome are also closely associated with the SEC61 translocon complex (green). (F) RPL26 and RPN1 are both UFMylated in a DDRGK1-dependent manner. The same experimental setup as in (B) was performed to probe for the indicated proteins. Note that the size shift corresponding to UFMylated RPN1 is not obvious due to the use of MES buffer that better resolves smaller molecular weight proteins, in this case, RPL26. (G) Reverse immunoprecipitation of RPN1-HA showed DDRGK1-dependent UFMylation of RPN1. The same cell lines as in (B) were transfected with the indicated combinations of RPN1-HA and/or UFM1-ΔCS for 24 h. Cells were then lysed for immunoprecipitation of HA epitope. Samples were resolved using MOPS buffer for better molecular weight separation between unmodified and UFMylated RPN1 proteins. See also Figure S6.

Figure 7.. UFMylation-Mediated ER-phagy Represses IRE1α UPR

(A) Dysregulation of UFMylation results in upregulation of UPR. HCT116 cells were transduced with the indicated sgRNAs and harvested for western blotting analysis. The graph represents densitometry measurement of the indicated proteins upon sgRNA knockdown. A representative blot is shown in Figure S7B. Data represent mean ± SD of three biological replicates. (B) Dysregulation of UFMylation transcriptionally upregulates UPR markers except IRE1 α. HCT116 CRISPRi cells were transduced with the indicated sgRNAs. Tunicamyin (0.5 μg/mL; 4 h) was used as a positive control for ER stress. Cells were harvested for qRT-PCR measurement of the indicated ER or UPR genes. Data represent mean ± SD of three biological replicates. (C) Knockdown of IRE1a partially restores ER-phagy in DDRGK1-depleted cells. HCT116 CRISPRi EATR cells transduced with the indicated sgRNAs and starved for 16 h before FACS measurement for ER-phagy. Data represent mean ± SD of three biological replicates. (D) Proposed model for the role of UFMylation during ER-phagy. DDRGK1 acts as an ER surface adaptor that recruits UFL1. At least two ER surface proteins that are in close proximity, RPN1 and RPL26, are UFMylated during ER-phagy. Dysregulation of UFMylation inhibits ER-phagy, and this potentially results in accumulation of ER stress and subsequently activates the IRE1 a-mediated unfolded protein response pathway. See also Figure S7.