The UFM1 system regulates ER-phagy through the ufmylation of CYB5R3

Abstract

Protein modification by ubiquitin-like proteins (UBLs) amplifies limited genome information and regulates diverse cellular processes, including translation, autophagy and antiviral pathways. Ubiquitin-fold modifier 1 (UFM1) is a UBL covalently conjugated with intracellular proteins through ufmylation, a reaction analogous to ubiquitylation. Ufmylation is involved in processes such as endoplasmic reticulum (ER)-associated protein degradation, ribosome-associated protein quality control at the ER and ER-phagy. However, it remains unclear how ufmylation regulates such distinct ER-related functions. Here we identify a UFM1 substrate, NADH-cytochrome b5 reductase 3 (CYB5R3), that localizes on the ER membrane. Ufmylation of CYB5R3 depends on the E3 components UFL1 and UFBP1 on the ER, and converts CYB5R3 into its inactive form. Ufmylated CYB5R3 is recognized by UFBP1 through the UFM1-interacting motif, which plays an important role in the further uyfmylation of CYB5R3. Ufmylated CYB5R3 is degraded in lysosomes, which depends on the autophagy-related protein Atg7- and the autophagy-adaptor protein CDK5RAP3. Mutations of CYB5R3 and genes involved in the UFM1 system cause hereditary developmental disorders, and ufmylation-defective Cyb5r3 knock-in mice exhibit microcephaly. Our results indicate that CYB5R3 ufmylation induces ER-phagy, which is indispensable for brain development.

Fig. 1. CYB5R3 is a bona fide substrate for UFM1.

a Immunoblot analysis. Wild-type or UFSP2^−/− HEK293T expressing UFL1 and UFBP1 were fractionated into microsomal (M) and cytoplasmic (C) fractions followed by immunoblot analysis. Data shown are representative of three separate experiments. b An experimental approach to identify ufmylated proteins. c Domain structure of CYB5R3. Membrane anchor, FAD-binding site, and lysine residue for UFM1 conjugation are indicated. d Immunoblot analysis. The indicated constructs were transfected into parental or UFSP2^−/− HEK293T. 48 h after the transfection, the cell lysates were pulled down with Ni-NTA agarose under denaturing conditions followed by immunoblot analysis. Data shown are representative of three separate experiments. e Immunoblot analysis. A series of CYB5R3KR mutants together with MYC-UFM1, MYC-UFL1, and UFBP-MYC were expressed in UFSP2^−/− HEK293T. The cell lysates were analyzed as described in d. Data shown are representative of three separate experiments. f Alignment of the region around K214 of CYB5R3. g In vitro conjugation assay. Recombinant UFM1, UBA5, and UFC1 from E. coli and FLAG-UFL1 and UFBP1-MYC from UFC1-knockout HEK293T cells were incubated with a microsomal fraction prepared from UFBP1-deficient HEK293T cells in the presence or absence of ATP for 90 min, and the mixture was subjected to SDS-PAGE followed by immunoblot analysis with CYB5R3 antibody. Data shown are representative of three separate experiments. h Immunoblot analysis. The indicated constructs were transfected into UFSP2^−/− HEK293T. 48 h after transfection, cell lysates were subjected to immunoblot analysis. Data shown are representative of three separate experiments. Source data are provided as a Source Data file.

Fig. 2. Ufmylation of CYB5R3 on the ER.

a Immunoprecipitation assay. HEK293T expressing UFL1 and UFBP1 were fractionated into microsomal (M) and cytoplasmic (C) fractions. The fractions were denatured with TNE with SDS (final 1%), diluted 10-fold with TNE, and then immunoprecipitated with CYB5R3 antibody followed by immunoblot analyses. Data shown are representative of three separate experiments. Bar graphs show the relative value of the ratio of endogenous ufmylated CYB5R3 to CYB5R3 was 1 when neither overexpression nor knockdown of any E3 components was performed. Data are means ± s.e. Statistical analysis was performed by two-sided Welch’s t test. b Immunoprecipitation assay. UFL1 or UFBP1 in UFSP2^−/− HEK293T were knocked down, and the cells were fractionated into microsomal (M) and cytoplasmic (C) fractions. The fractions were analyzed as described in a. Data shown are representative of four separate experiments. Bar graphs show the relative value of the ratio of endogenous ufmylated CYB5R3 to CYB5R3 was 1 when neither overexpression nor knockdown of any E3 components was performed. Data are means ± s.e. Statistical analysis was performed by Šidák’s multiple comparison test after one-way ANOVA. c Immunoblot analysis. The indicated constructs were transfected into CYB5R3^−/− UFSP2^−/− HEK293T. 48 h after transfection, the cell lysates were fractionated into microsomal (M) and cytoplasmic (C) fractions followed by immunoblot analysis. Data shown are representative of three separate experiments. Asterisk indicates a degradative product of CYB5R3-His-3xFLAG. Since the CYB5R3 in the cytoplasmic fraction (indicated by asterisk) is almost as mobile as ΔN26, it is considered to be the N-terminal truncated form. d Immunofluorescence analysis. GFP-tagged wild-type (CYB5R3-GFP) or ER-localization-defective CYB5R3 (CYB5R3ΔN26-GFP) were transfected into CYB5R3^−/− HeLa. 48 h after transfection, the cells were immunostained with anti-PDI antibody. Bars: 10 μm. Data shown are representative of three independent experiments. Source data are provided as a Source Data file.

Fig. 3. Loss of CYB5R3 enzymatic activity after ufmylation.

a Left, crystal structure of human CYB5R3 (PDB 1UMK). FAD and NADH domains are colored yellow and salmon pink, respectively, whereas bound FAD is colored green. Right, close-up view of the domain interface that recognizes FAD. Broken lines indicate possible hydrogen bonds. b Top, Successive HS-AFM images of CYB5R3. Scale, 10 nm; Height, 0–4 nm. Middle, Histograms of the distances between two globular lobes from 194 frames of 1 molecule. Representative images of closed and open conformations with height profiles are shown right. Bottom, Height profiles of simulated AFM images from the crystal structure (PDB ID: 1UMK) (left) and the manually arranged structure (right) with the simulated images. c The closed conformation (left) corresponds to the crystal structure (1UMK) while the open conformation (right) was manually prepared based on the HS-AFM data. CYB5R3 and FAD are shown with surface and space-filling models, respectively. d Color comparison of purified free CYB5R3 with ufmylated CYB5R3. Recombinant CYB5R3 and UFM1-conjugated CYB5R3 were purified in large quantities as described in Fig. S6. e In vitro enzymatic activity of recombinant CYB5R3 (n = 3) and UFM1-conjugated CYB5R3 (n = 3) were measured as described in Methods section. Data are means ± s.e. Statistical analysis was performed by Šidák’s multiple comparison test after one-way ANOVA. f Left panel: CYB5R3^−/− UFSP2^−/− HEK293T expressing indicated constructs were lysed and then subjected to immunoblot analysis with anti-CYB5R3 antibody. Data shown are representative of three separate experiments. Right panel: the reductase activity of microsomal fractions prepared from the aforementioned cells was measured as described in the Methods section. Data are means ± s.e. Statistical analysis was performed by Šidák’s multiple comparison test after one-way ANOVA. g, h The amounts of saturated, monounsaturated, and polyunsaturated phosphatidylcholine and of phosphatidylethanolamine (g) (n = 3) and of cholesterol (h) (n = 4) in CYB5R3^−/− UFSP2^−/− HEK293T expressing UFM1, UFL1, UFBP1, and CYB5R3 or CYB5R3^K214R. Data are means ± s.e. Source data are provided as a Source Data file.

Fig. 4. Ufmylated CYB5R3 interacts with UFBP1.

a CYB5R3^−/− UFSP2^−/− HEK293T expressing indicated constructs were lysed and then subjected to immunoprecipitation. b–c In vitro pull-down assay. The amount of each protein bound to GST-UFBP1 was estimated by immunoblot (b). The faster-migrating ufmylated CYB5R3ΔN26 proteins are most likely degradation products due to long-term storage and freeze–thawing. UFM1 and UFL1 binding to GST-UFBP1 and each mutant was estimated by immunoblot (c). d Top, analysis of the secondary structure of UFBP1 (116–214) using PSIPRED/DISOPRED^, and COILS. Bottom, alignment of UFIM of UFBP1. e Crystal structure of UFM1 complexed with the UFIM of UFBP1 (left) or the UFIM of UBA5 (right, PDB 5HKH). The side chains of the residues involved in the UFM1-UFIM interaction are shown with a stick model. Broken lines indicate possible hydrogen bonds. f Structural comparison between UFBP1 and UBA5 UFIMs complexed with UFM1. The figure was prepared by superimposing the structure of the UFIM moiety of UFBP1-UFM1 onto that of UBA5 UFIM-UFM1. g In vitro pull-down assay. The amount of each protein bound to GST-UFBP1 was estimated by immunoblot. h–i Wild-type (h) and UFBP1^−/− HEK293T (i) expressing indicated constructs were lysed and then subjected to immunoprecipitation. j HEK293T expressing indicated constructs were lysed and then subjected to immunoblot. Bar graphs show the relative value of the ratio of endogenous ufmylated CYB5R3 to CYB5R3 was 1 and the relative value of the ratio of ufmylated UFBP1 to UFBP1 as 1 when UFL1 and UFBP1 were overexpressed, respectively. Data are means ± s.e. Statistical analysis was performed by two-sided Welch’s t test. k HEK293T expressing indicated constructs were lysed and then subjected to pull-down assay with Ni-NTA agarose under denaturing conditions, followed by immunoblot. Bar graph shows the relative value of the ratio of MYC-UFM1-conjugated CYB5R3-His-FLAG to CYB5R3-His-FLAG in crude lysates as 1 when UFL1 and UFBP1 were overexpressed. Data are means ± s.e. Statistical analysis was performed by Šidák’s multiple comparison test after one-way ANOVA. Data for immunoblots presented in this Figure are representative of three separate experiments. Source data are provided as a Source Data file.

Fig. 5. The regulation of ER-phagy by ufmylation of CYB5R3.

a Schematic representation of the CYB5R3-mCherry-GFP (CYB5R3-CG) reporter for monitoring lysosomal degradation of CYB5R3. b Fluorescence microscopic analysis. CYB5R3-CG or CYB5R3^K214R-CG (KR) together with UFL1 and UFBP1 were co-transfected into CYB5R3^−/− HeLa. Forty-eight hours after transfection, the cells were cultured under nutrient-rich or deprived conditions for 9 h in the presence or absence of Bafilomycin A₁. The cells were fixed and observed by confocal microscopy. The number of single mCherry-positive punctae per cell was determined using a Benchtop High-Content Analysis System and CellPathfinder software without bias. The numbers of cells used to count the mCherry-positive punctae were 401, 226, 117, 136, and 129 from lanes 1 to 5. Bars: 10 μm. c Fluorescence microscopic analysis. CYB5R3-CG together with UFL1 and UFBP1 or UFBP1^{F196A V198A} (FA, VA) were co-transfected into UFBP1^−/− HeLa. The cells were analyzed as described in b. The numbers of cells used to count the mCherry-positive punctae were 335, 176, 90, 123, 114, and 102 from lanes 1 to 6. Bars: 10 μm. d Fluorescence microscopic analysis. CYB5R3-GFP together with UFL1 and UFBP1 were co-transfected into CYB5R3^−/− HeLa. Forty-eight hours after transfection, the cells were cultured under nutrient-rich or deprived conditions for 9 h. The cells were fixed, immunostained with anti-FIP200, anti-WIPI2, and anti-LC3 antibodies, and observed by confocal microscopy. Bars: 10 μm. The number of punctae positive for GFP and FIP200, WIPI, or LC3 per cell (n = 30) was determined. e Fluorescence microscopic analysis. CYB5R3-CG together with UFL1 and UFBP1 were co-transfected into parental or ATG7-deficient HeLa cells. The cells were observed as described in b. The numbers of cells used to count the mCherry-positive punctae were 245, 150, 186, and 144 from lanes 1 to 4. Bars: 10 μm. For the box plots, horizontal bars indicate medians, boxes the interquartile range (25th–75th percentiles), and whiskers 1.5× the interquartile range; outliers are plotted individually. Data are means ± s.e. Statistical analysis was performed by Šidák’s multiple comparison test after one-way ANOVA. Source data are provided as a Source Data file.

Fig. 6. CDK5RAP3 is indispensable for ufmylation-mediated ER-phagy.

a–b UFSP2^−/− (a) or CDK5RAP3^−/− UFSP2^−/− HEK293T (b) expressing indicated constructs were cultured under nutrient-rich conditions (a) in the presence or absence of Bafilomycin A₁ for 12 h (b)._. Lysates were subjected to immunoblot analysis. Data shown are representative of three separate experiments. Bar graphs show the results of quantitative densitometric analysis of the indicated proteins relative to free CYB5R3 or free UFBP1. Data are means ± s.em. Statistical analysis was performed by Šidák’s multiple comparison test after one-way ANOVA. c Fluorescence microscopic analysis. CDK5RAP3^−/− HeLa expressing indicated constructs were analyzed as described in Fig. 5b. The numbers of cells used to count the mCherry-positive punctae were 337, 235, 191, 127, 113, and 92 from lanes 1 to 6. Bars: 10 μm. Horizontal bars indicate medians, boxes in the interquartile range (25th–75th percentiles), and whiskers 1.5× the interquartile range; outliers are plotted individually. Data are means ± s.e. Statistical analysis was performed by Šidák’s multiple comparison test after one-way ANOVA. d Dorsal view of the brains of 4-month-old wild-type, Cyb5r3^K214R/+, and Cyb5r3^K214R/K214R mice. Graphs show the axial length and the maximal lateral length of brains of mice of the indicated genotypes. Data are means ± s.e. of wild-type (n = 3), Cyb5r3^K214R/+ (n = 10) and Cyb5r3^K214R/K214R (n = 11) mice. Statistical analysis was performed by Šidák’s multiple comparison test after one-way ANOVA. Bars: 2 mm. e H&E staining of brain sections from 4.5–5 month-old Cyb5r3^K214R/+ and Cyb5r3^K214R/K214R mice. Boxed regions are magnified and shown on the right. The graph shows ratios of the thickness of the auditory cortex to the distance from the aqueduct to the corresponding lateral edge. Data are means ± s.e. of wild-type (n = 1 from one mouse, black) and Cyb5r3^K214R/+(n = 6 from three mice, blue), and Cyb5r3^K214R/K214R (n = 7 from four mice, pink) mice. Statistical analysis was performed on the wild-type and Cyb5r3^K214R/+ mice together as controls, and these were compared to the Cyb5r3^K214R/K214R mice. Bars: 2 mm. Source data are provided as a Source Data file.