Proteome census upon nutrient stress reveals Golgiphagy membrane receptors

Abstract

During nutrient stress, macroautophagy degrades cellular macromolecules, thereby providing biosynthetic building blocks while simultaneously remodelling the proteome^1,2. Although the machinery responsible for initiation of macroautophagy has been well characterized^3,4, our understanding of the extent to which individual proteins, protein complexes and organelles are selected for autophagic degradation, and the underlying targeting mechanisms, is limited. Here we use orthogonal proteomic strategies to provide a spatial proteome census of autophagic cargo during nutrient stress in mammalian cells. We find that macroautophagy has selectivity for recycling membrane-bound organelles (principally Golgi and endoplasmic reticulum). Through autophagic cargo prioritization, we identify a complex of membrane-embedded proteins, YIPF3 and YIPF4, as receptors for Golgiphagy. During nutrient stress, YIPF3 and YIPF4 interact with ATG8 proteins through LIR motifs and are mobilized into autophagosomes that traffic to lysosomes in a process that requires the canonical autophagic machinery. Cells lacking YIPF3 or YIPF4 are selectively defective in elimination of a specific cohort of Golgi membrane proteins during nutrient stress. Moreover, YIPF3 and YIPF4 play an analogous role in Golgi remodelling during programmed conversion of stem cells to the neuronal lineage in vitro. Collectively, the findings of this study reveal prioritization of membrane protein cargo during nutrient-stress-dependent proteome remodelling and identify a Golgi remodelling pathway that requires membrane-embedded receptors.

Fig. 1. Proteome census reveals Golgi and ER proteins as prioritized clients during macroautophagy.

a, Method for quantifying proteome alterations through autophagy in response to withdrawal of amino acids (−AA). UT, untreated. b, Violin plots for proteins identified as CAPs (n = 684) in WT and FIP200^−/− HEK293 cells without or with amino acid withdrawal (12 h). Navy dashed lines: median value for known autophagy proteins. c, Top ten Gene Ontology terms for CAPs from cells subjected to amino acid withdrawal. P values were calculated using two-sided Fisher’s exact test and adjusted for multiple comparisons using the Benjamini–Hochberg method. d, Frequency of proteins with the indicated subcellular localization for the CAPs (amino acid withdrawal). e, Schematic depicting selective autophagy within the macroautophagy pathway. See text for details. f, Among CAPs, percentage of total protein copy numbers lost upon amino acid withdrawal. g, Percentage of all protein copies lost from CAPs (purple) or other mechanisms (green) by amino acid withdrawal for subcellular compartments (1.2829 × 10⁶ total). h, Golgi proteins from CAPs coloured by FIP200-dependent turnover during amino acid withdrawal.

Fig. 2. Orthogonal proteomics for Golgiphagy receptor identification.

a, TMT pipeline to measure relative protein abundance during nutrient stress (EBSS) with or without ATG7. b, Plots of ATG7^−/− log₂[FC(EBSS/UT)] − WT log₂[FC(EBSS/UT)] versus WT log₂[FC(EBSS/UT)], in which FC represents fold change, for HeLa cells treated with EBSS for 18 h with priority for individual proteins scaled on the basis of the inset colour code. Full plots are shown in Extended Data Fig. 5g,h. c, Ten-plex TMT APEX2–ATG8 pipeline to capture autophagy receptors during nutrient stress (EBSS + BafA1, 4 h) with or without LDS. At 3 h post-nutrient stress, cells were supplemented with biotin phenol (1 h) and then treated with H₂O₂ for 1 min followed by quenching (Methods). d, APEX2 proximity labelling plots of GABARAPL2(Y49A/L50A) log₂[FC((EBSS + BafA1)/UT) − WT log₂[FC((EBSS + BafA1)/UT)] versus WT log₂[FC((EBSS + BafA1)/UT)] in which priority for individual proteins is scaled on the basis of the inset colour code. Full plots are shown in Extended Data Fig. 6e,f. e, Top ranked proteins (n = 30) on the basis of summed individual rankings for global proteomics and ATG8 proximity biotinylation (Methods) displayed on the basis of their subcellular localization, autophagy involvement and known or candidate LIR motif. Known autophagic cargo receptors are in bold font.

Fig. 3. LIR-containing YIPF3 and YIPF4 undergo autophagic flux and associate with autophagy machinery during macroautophagy.

a, Domain structures of YIPF3 and YIPF4 showing the locations of transmembrane segments and N-terminal candidate LIR motifs. Transmembrane domains (TM1–TM5) are shown in grey; single-letter code amino acid sequences for WT and mutant LIR motif labelled are indicated below. b, Colabfold model for the YIPF3–YIPF4 complex. Candidate LIR motifs are shown in red. c, Keima–YIPF3 HEK293 cells (± FIP200) were untreated or subjected to nutrient stress for 16 h before flow cytometry. Frequency distributions of 561 nm/405 nm excitation ratios are shown (n = 10,000 cells per condition). d, Bar graph of median values of the biological duplicate experiments for 561 nm/405 nm excitation ratios for Keima–YIPF3 or Keima–YIPF4 with or without FIP200. e, APEX2 proximity labelling plots of YIPF3(F47A) log₂[FC((EBSS + BafA1)/UT)] − WT log₂[FC((EBSS + BafA1)/UT)] versus WT log₂[FC((EBSS + BafA1)/UT)] in which priority for individual proteins is scaled on the basis of the inset colour code. Full plots are shown in Extended Data Fig. 7h,i. f, RFP–Trap immunoprecipitates (IP) of WT and LIR-mutant GFP–YIPF3 and mCherry (mCh)–YIPF4 in HEK293 YIPF3^−/−YIPF4^−/− cells that were untreated or starved of amino acids (2 h + BafA1) were immunoblotted (IB) with the indicated antibodies. L1, GABARAPL1; L2, GABARAPL2. This experiment was repeated in biological triplicate with similar results. Source Data

Fig. 4. YIPF4 mobilization into autophagosomes during nutrient stress.

a,b, Confocal micrographs of HEK293 cells expressing endogenous mNEON–YIPF4 co-stained with GOLGB1 (magenta) with (b) or without (a) nutrient stress (3 h with BafA1). Hoechest 33342 labels nuclei (cyan). Scale bars, 5 μm (right) and 10 μm (left). c, Cells as in b immunostained with anti-LAMP1 (magenta). Line scan region indicated with dashed yellow line, LAMP1-positive mNEON–YIPF4 puncta indicated with yellow arrowheads (left). Line scans for LAMP1 and mNEON signal as a histogram (right). Scale bars, 1 μm (right) and 10 μm (left). d, As in c but using FIP200^−/− cells. e, Number of mNEON–YIPF4 puncta per cell for the indicated treatments in cells ± FIP200. Each dot represents one image in which mNEON and nuclei were counted. ***P  < 0.05 (two-tailed Mann–Whitney test); Left to right: P > 0.9999, P = 0.0238, P = 0.7, P = 0.318. Lines, mean values; error bars, s.d. NS, not significant. n = total number of cells analyzed for each condition. f, mNEON–YIPF4 puncta in cells treated as in b but with or without addition of VPS34i were quantified as in e. Two-tailed Mann–Whitney P values from left to right: P = 0.8857, P = 0.0095, P = 0.4, P = 0.6095, P = 0.0022. Lines, mean values; error bars, s.d. NS, not significant. n = total number of cells analyzed for each condition. g, Cells treated as in b were immunostained with anti-LC3B (magenta). Yellow arrowheads indicate YIPF4⁺ puncta overlapping LC3B⁺ structures. Scale bars, 1 μm (right) and 10 μm (left). h,i, HEK293 cells expressing mNEON–YIPF4 and mCherry–LC3B were subjected to live-cell confocal microscopy 2 h post EBSS treatment and single confocal slices through cells are shown. The time series in i shows coincident movement of mNEON and mCherry signal over successive frames (arrowheads). Scale bars, 1 μm. j, Number of mNEON–YIPF4 puncta per cell for the indicated treatments in cells 3 h post EBSS, quantified as in e. Lines, mean values; error bars, s.d. NS, not significant. ***P < 0.05 (two-tailed Mann–Whitney test). From left to right: P = 0.1, P = 0.0238, P = 0.0121, P > 0.9999. n = total number of cells analyzed for each condition. For all micrographs (a–d,g,i), all experiments were carried out in biological triplicate with similar results. Source Data

Fig. 5. YIPF3 and YIPF4 mediate the autophagy-based recycling of Golgi proteins during nutrient stress and neuronal differentiation in vitro.

a, Method for global proteome alterations through YIPF4 or FIP200 in response to nutrient stress. b, Correlation plot of CAPs for alterations in protein abundance for the indicated subcellular compartments during amino acid withdrawal for YIPF4^−/− − WT cells (y axis) versus FIP200^−/− − WT cells (x axis). Points are the median of each distribution, and lines represent the 25–75% quantile. c, Classification of Golgi proteins that exhibit YIPF4- or FIP200-dependent degradation in response to amino acid withdrawal (12 h), with the number of transmembrane segments for each membrane protein, as well as Golgi-associated proteins, shown. Grey density scale, FIP200 dependence; colour scale, YIPF4 dependence. d, Workflow for analysis of ATG12^−/− and YIPF4^−/− iNeurons (iN). e, Correlation plot of CAPs for alterations in protein abundance for the indicated subcellular compartments during in vitro differentiation for YIPF4^−/− − WT iNeurons (y axis) versus ATG12^−/− − WT iNeurons (x axis). Points are the median of each distribution, and lines represent the 25–75% quantile. f, Heatmap of log₂[FC] values from ATG12^−/− and YIPF4^−/− iNeurons for the indicated proteins identified as Golgi CAPs in response to nutrient stress. g, Model of YIPF3- and YIPF4-mediated Golgiphagy upon nutrient starvation. Aspects of how YIPF3, YIPF4 and other Golgi cargo are selected for capture as well as how autophagic Golgi vesicles are formed remain to be delineated, as indicated by a question mark.

Extended Data Fig. 1. Experimental approach for identification of candidate autophagy proteins via quantitative proteomics.

a, Immunoblot for HEK293 control and ATG7^−/− cells with or without EBSS starvation for 12h in duplicate with the indicated antibodies. Independently cultured replicate samples were loaded in adjacent lanes. b, Immunoblot for HEK293 control, FIP200^−/− and YIPF4^−/− cells with or without AA withdrawal for 12h in duplicate with the indicated antibodies. Independently cultured replicate samples were loaded in adjacent lanes. c, Immunoblot for HEK293 control, FIP200^−/− and YIPF4^−/− cells probed with the indicated antibodies. Dotted lines indicate separate lanes on samples analyzed on the same gel. This experiment was performed in biological triplicate with similar results. d, List of proteins that demonstrate autophagy dependent degradation during nutrient starvation. e, Dot plot of proteins from panel d in WT or ATG7^−/− HEK293 cells treated with EBSS for 12h. Navy dashed line represents median protein abundance. f, Dot plot of proteins from panel d in WT or ATG7^−/− HEK293 cells treated with AA withdrawal for 12h. Navy dashed line represents median protein abundance. g, Workflow for calculating RMSE of all proteins in HEK293 Control or ATG7^−/− cells treated with EBSS for 12h, and HEK293 control, FIP200^−/−, and YIPF4^−/− cells treated with AA withdrawal for 12h.

Extended Data Fig. 2. Golgi and ER are enriched in candidate autophagy proteins.

a, Violin plots for proteins identified as candidate ‘autophagy’ proteins in WT and ATG7^−/− HEK293 cells with or without EBSS (12h). Navy dashed lines: median value for known autophagy proteins. b, RMSE plot for HEK293 cells treated with 12h of EBSS. CAPs are shown along with all autophagy machinery, nuclear, and ribosomal proteins. c, RMSE plot for HEK293 cells treated with 12h of AA Withdrawal. CAPs are shown along with all autophagy machinery, nuclear, and ribosomal proteins. d, Top 10 Gene Ontology terms identified for candidate ‘autophagy’ proteins from cells subjected to EBSS. e, Frequency of proteins with the indicated sub-cellular localizations for the candidate ‘autophagy’ proteins or all other proteins for cells subjected to EBSS treatment. f, Frequency of proteins with the indicated sub-cellular localizations for the candidate ‘autophagy’ proteins or all other proteins for cells subjected to AA withdrawal. g, Frequency of proteins with the indicated sub-cellular localizations for either overlapping or non-overlapping proteins. h, RMSE for each compartment shown from HEK293 EBSS experiment. i, RMSE for each compartment shown from HEK293 AA withdrawal experiment. j, RMSE plots for EBSS and −AA with known ERGIC proteins indicated in magenta.

Extended Data Fig. 3. Subcellular localization analysis of candidate autophagy proteins.

a, Enrichment of Golgi-membrane and Golgi-associated proteins in the candidate ‘autophagy’ list and all other proteins for AA withdrawal and EBSS treatment. b, Enrichment for the cytosolic proteins in the candidate ‘autophagy’ list and all other proteins for AA withdrawal and EBSS treatment. c, Venn diagrams indicating the overlap of proteins identified in common within candidate ‘autophagy’ lists for AA withdrawal and EBSS treatment. Numbers within the diagram indicate the number of proteins present. d, Violin plots for Log₂FC (−AA/UT) for control, FIP200^−/−, or YIPF4^−/− HeLa cells displayed for 38 proteasome and 84 ribosomal proteins as well as proteins annotated as cytosolic. Median values are indicated by solid bold line. e, TMT-scaled MS1 ranked plots. Protein copy number estimates for CAPs in HEK293 cells (black) in rank order. Among CAPs, the number of protein copies after loss by autophagy during amino acid starvation for each compartment as determined using protein abundance fold changes (AA withdrawal – untreated). f, Rank plot for cytoplasmic, ER and Golgi localized proteins. g, Model for possible selectivity of macroautophagy at the organelle level. Abundance rank change (ΔRank) between proteins in the ‘autophagy’ candidate list – all other proteins for each organelle, scaled to number of total proteins in both scaled TMT and DIA experiments. For each compartment, p-values are listed and organelles with significant differences are in bold. Interestingly, cytosolic CAPs display a bias toward less abundant proteins, while CAPs annotated as ER or endosomal are biased for more abundant proteins. In contrast, CAPs annotated as Golgi proteins do not present a significant bias toward more or less abundant proteins. p-values from two-sided Wilcoxon Rank Sum test.

Extended Data Fig. 4. Proteome census analysis during nutrient stress.

a, DIA ranked plots. Protein copy number in the untreated condition for candidate ‘autophagy’ proteins in HEK293 cells (black) in rank order. The number of protein copies after loss by autophagy during amino acid starvation for each compartment as determined using protein abundance fold changes (AA withdrawal – untreated) by DIA. b, Among the candidate autophagy proteins, percentage of total protein copy numbers lost via amino acid withdrawal (3.0161 × 10⁷ total). c, Percentage of all protein copies lost from ‘autophagy’ candidate list (purple) or other mechanisms (green) by amino acid withdrawal for subcellular compartments based on DIA values with histone-based proteome ruler values. d,e, Same as panels b and c, respectively, but based on DIA FC values mapped onto proteome ruler values from Wisniewski et al. (9.77 × 10⁶ total). f, Correlation with DIA protein copy number estimates against Wisniewski et al. protein copy numbers. R = 0.85, p < 2.2 × 10⁻¹⁶. Statistics are Pearson correlation. g,h, Same as panels b and c, respectively, based on TMT-scaled FC values mapped onto proteome ruler values from Wisniewski et al. (7.1573 × 10⁶ total). i, Correlation plots for TMT-scaled MS1 protein signals against Wisniewski et al. copy number. R = 0.81, p < 2.2 × 10⁻¹⁶. Statistics are Pearson correlation. j, Correlation plots for TMT-scaled MS1 protein copy numbers and DIA protein copy numbers. R = 0.79, p < 2.2 × 10⁻¹⁶. Statistics are Pearson correlation. k, Distribution of proteins with roles in coatamer (COPI/II) function in response to EBSS treatment in WT and ATG7^−/− HeLa cells. These proteins do not display properties of CAPs. l, Distribution of proteins with roles in coatamer (COPI/II) function in response to EBSS treatment in WT and ATG7^−/− HEK293 cells. These proteins do not display properties of CAPs.

Extended Data Fig. 5. Identification of candidate autophagy proteins using total proteome analysis in ATG7^−/− HEK293 or HeLa cells.

a, Western blot showing markers of starvation (ULK1, 4EBP dephosphorylation) and ATG7 in WT and ATG7^−/− HEK293 cells grown in EBSS for 12h. Independently cultured replicate samples were loaded in adjacent lanes. b, Violin plots of relative total (8258), Golgi (160), or ER (344) protein abundance in response to EBSS treatment (12h) in WT and ATG7^−/− HeLa cells. c-f, Volcano plots [WT Log₂(12h EBSS/UT) versus -Log₁₀(q-value)] for HeLa (panel c) or HEK293 (panel d) or analogous plots for ATG7^−/− HeLa (panel e) or HEK293 (panel f) cells. g, h, Plots of ATG7^−/− Log₂(EBSS/UT) - WT Log₂(EBSS/UT) versus WT Log₂(EBSS/UT) for HeLa cells (panel g) and ATG7^−/− Log₂(EBSS/UT) – WT Log₂(EBSS/UT) versus WT Log₂(EBSS/UT) for HEK293 cells (panel h) where priority for individual proteins is scaled based on the color code inset.

Extended Data Fig. 6. Proximity biotinylation of ATG8 proteins MAP1LC3B or GABARAPL2 in response to nutrient stress.

a-d, Volcano plots [WT Log₂(4h EBSS+BafA1/UT) versus −Log₁₀ (q-value)] for APEX2-GABARAPL2 (panel a) or APEX2-MAP1LC3B (panel b) or analogous plots for APEX2-GABARAPL2^Y49A/L50A (panel c) or APEX2-MAP1LC3B^K51A/G120A (panel d) in HeLa cells. e,f, Plots of GABARAPL2^Y49A/L50A Log₂(EBSS+BafA1/UT) – WT Log₂(EBSS+BafA1/UT) versus WT Log₂(EBSS+BafA1/UT) (panel e) and MAP1LC3B^K51A/G120A Log₂(EBSS+BafA1/UT) – WT Log₂(EBSS+BafA1/UT) versus WT Log₂(EBSS+BafA1/UT) (panel f) where priority for individual proteins is scaled based on the color code inset.

Extended Data Fig. 7. Proximity biotinylation of YIPF3 and YIPF4.

a, Immunoblotting of WT, YIPF3^−/−, and YIPF4^−/− HeLa cells probed in duplicate with the indicated antibodies. Anti-HSP90 was used as a loading control. Independently cultured replicate samples were loaded in adjacent lanes. b, Scheme outlining Keima-YIPF3/4 as reporters for Golgiphagic flux. c, Experimental scheme for proximity biotinylation using APEX2-YIPF3/YIPF4 (or LIR mutants) in response to nutrient stress (EBSS, 4h). d-i, Volcano plots [WT Log₂(4h EBSS+BafA1/UT) versus −Log₁₀ (q-value)] for APEX2-YIPF3 (panel d) or APEX2-YIPF4 (panel e) or analogous plots for APEX2-YIPF3^F47A (panel f) or APEX2-YIPF4^F17A/V20A (panel g) in HeLa cells. h,i, Plots of APEX2-YIPF3^F47A Log₂(EBSS+BafA1/UT) WT Log₂(EBSS+BafA1/UT) versus WT Log₂(EBSS+BafA1/UT) (panel h) and APEX2-YIPF4^F17A/V20A Log₂(EBSS+BafA1/UT) WT Log₂(EBSS+BafA1/UT) versus WT Log₂(EBSS+BafA1/UT) (panel i) where priority for individual proteins is scaled based on the color code inset.

Extended Data Fig. 8. Association of YIPF3/4 with ATG8 proteins.

a, Immunofluorescence of WT and LIR mutant YIPF3 and YIPF4 exogenous expression constructs in DKO (YIPF3/4) HEK293 cells. Cells were imaged using confocal microscopy and co-stained with the Golgi marker GOLGA2 (yellow). Scale bars is 10 microns. b, FLAG-LC3B (WT or LDS mutant) was stably expressed in HEK293 cells expressing mCherry-YIPF3 and GFP-YIPF4 (WT or LIR motif mutants) via lentivirus and anti-FLAG immune complexes isolated with or without a 2h EBSS+BafA1 treatment. Immune complexes were subjected to immunoblotting and probed with the indicated antibodies. This experiment was performed once. c, Immunoblotting and probing of indicated antibodies in ATG8 KO cell lines subjected to nutrient stress with EBSS. HKO (deletion of LC3A, B, C, GABARAP, L1, L2), ΔLC3 (deletion of LC3A, B, C), ΔRAP (deletion of GABARAP, L1, L2). Results are representative of experiments performed twice.

Extended Data Fig. 9. Analysis of YIPF4 localization in response to nutrient stress.

a, Immunoblot showing mNEON-YIPF4 endogenous tagging results at a higher molecular weight indicative of the total fusion protein length (~50 kDa). This experiment was performed in biological triplicate with similar results. b, HEK293 cells and HEK293 FIP200^−/− cells expressing endogenous YIPF4 tagged on its N-terminus with mNEON (green) imaged using confocal microscopy and co-stained with cis (GOLGA2) or trans (TGN46) Golgi markers. Scale bars 20 microns. This experiment was performed in biological duplicate with similar results. c, HEK293 cells untreated or starved for 3h with EBSS in the presence of BafA1 (100 nM) expressing endogenous YIPF4 tagged on its N-terminus with mNEON (green) imaged using confocal microscopy and co-stained with an antibody against YIPF3. Scale bars 10 microns. This experiment was performed in biological triplicate with similar results. d, HEK293 cells expressing endogenous YIPF4 tagged on its N-terminus with mNEON (green) imaged using confocal microscopy and co-stained with LAMP1 (magenta). Cells were either left untreated (top left) or subjected to nutrient stress +BafA1 and VPS34i (3h) (top right) or subjected to nutrient stress +BafA1 and an E1 inhibitor (TAK243) (3h) (bottom) prior to imaging. Nuclei were labeled with Hoechst33342 dye (cyan). Scale bars 10 microns. This experiment was performed in biological triplicate with similar results. e, Line scan of HEK293 cells expressing endogenous YIPF4 tagged on its N-terminus with mNEON and MAP1LC3B show colocalization upon EBSS+BafA1 treatment for 3h. f, HeLa cells lacking OPTN, NDP52, SQSTM1, NBR1, and TAX1BP1 (Penta KO) were subjected to the indicated treatment for 12h and extracts probed with the indicated antibodies. Blots from two independent experiments were quantified based on LiCor intensities, then normalized to the UT sample for each genotype. The lower panel shows the individual values for each replicate, and error bars are S.D. Source Data

Extended Data Fig. 10. Contribution of YIPF3/4 to Golgi turnover by autophagy during nutrient stress.

a, Violin plots for Log₂(-AA/UT) for control, FIP200^−/−, or YIPF4^−/− HeLa cells displayed for various classes of proteins with the indicated sub-cellular localizations for either the ‘autophagy’ candidates or all other proteins. Median values are indicated by solid bold line. N = 3 in biological replicates. Unpaired two-sided t-test; *, p < 0.05. ns, not significant. P-values for FIP200^−/− vs WT CAPs from left to right: p = 2.14 × 10⁻²⁵, p = 0.0005, p = 4.12 × 10⁻¹¹, p = 5.98 × 10⁻¹⁸, p = 0.004, p = 3.48 × 10⁻¹⁷, p = 8.97 × 10⁻⁶, p = 3.32 × 10⁻¹¹, p = 0.012, p = 0.065, p = 0.0446, p = 0.053, p = 0.0392, p = 0.043. p.values for YIPF4^−/− vs WT CAPs from left to right: p = 0.494, p = 0.303, p = 0.765, p = 0.002, p = 0.712, p = 0.0013, p = 0.765, p = 0.911, p = 0.967, p = 0.504, p = 0.486, p = 0.349, p = 0.910, p = 0.863. p-values for FIP200^−/− vs WT other proteins from left to right: p = 0.000015, p = 0313, p = 0.0331, p = 0.0733, p = 0.784, p = 0.005, p = 0.328, p = 0.0003, p = 0.148, p = 0.602, p = 0.011, p = 9.06 × 10⁻⁸, p = 0.241, p = 0.415. p-values for YIPF4^−/− vs WT other proteins from left to right: p = 0.792, p = 0448, p = 0.557, p = 0.488, p = 0.680, p = 0.519, p = 0.503, p = 0.841, p = 0.919, p = 0.658, p = 0.723, p = 0.227, p = 0.916, p = 0.944. b, Violin plots for Log₂(−AA/UT) for control, FIP200^−/−, or YIPF4^−/− HeLa cells displayed for various classes of proteins with the indicated sub-cellular localizations for ‘autophagy’ candidates. Median values are indicated by solid bold line. n = 3 in biological replicates. Unpaired two-sided t-test; *, p < 0.05. ns, not significant. Data is extracted from panel a. p-values for FIP200^−/− vs WT from left to right: p = 2.14 × 10⁻²⁵, p = 5.98 × 10⁻¹⁸, p = 0.004, p = 3.48 × 10⁻¹⁷. p-values for YIPF4^−/− vs WT from left to right: p = 0.494, p = 0.002, p = 0.712, p = 0.0013. c, Western blot showing Golgi protein levels in WT, FIP200^−/−, or DKO (YIPF3^−/−/YIPF4^−/−) HEK293 cells in response to AA withdrawal (12h). d, Quantification of western blots for the indicated Golgi proteins in HEK293 control, FIP200^−/−, and DKO (YIPF3/YIPF4) either untreated or starved for amino acids for 12h (as in panel c) performed in biological triplicate. Unpaired two-sided t-test; *, p < 0.05. ns, not significant. Bars are mean values and error bars represent S.D. p-values for control UT vs -AA from left to right: p = 0.000739, p = 0.03188, p = 0.1489, p = 0.006017, p = 0.027868. p-values for FIP200^−/− UT vs -AA from left to right: p = 0.503841, p=0.456941, p = 0.540851, p = 0.11076, p = 0.733579. p.values for DKO (YIPF3/YIPF4^−/−) UT vs -AA from left to right: p = 0.344926, p = 0.555756, p = 0.398393. e, Correlation plot for alterations in protein abundance for proteins in the indicated sub-cellular compartments in HeLa cells after 18h of EBSS for YIPF3^−/−/WT or YIPF4^−/−/WT cells (y-axis) versus FIP200^−/−/WT cells (x-axis). f, GALNT2-Keima expressing HEK293 cells (WT, FIP200^−/−, DKO) were left untreated or subjected to nutrient stress for 16h and then analyzed by flow cytometry. Frequency distributions of 561/405 nm ex. ratios are shown (n = 10,000 cells per condition). Source Data

Extended Data Fig. 11. Analysis of YIPF3/4^−/− cells for ATG9 vesicles and proteomic analysis reveals no obvious role for CALCOCO1 in Golgi turnover in HeLa cells.

a, HEK293 cells of the indicated genotypes were subjected to immunofluorescence with α-ATG9 (magenta) and α-GOLGB1 (yellow) antibodies. Nuclei were stained with Hoechst (cyan) Scale bars are 10 microns. ATG9 puncta (objects/cell) were quantified using Cell Profiler (lower panel), n = 4 or 5 as indicated by dots. Two-Tailed Mann-Whitney test, *, p-value < 0.05. ns, not significant. Number of cells analyzed for WT, YIPF3/4^−/−, and FIP200^−/− genotypes were 80, 108, and 100, respectively. Bars are mean values and error bars represent S.D. p-values from left to right: p = 0.0317, p = 0.0286, p = 0.5556. b, Immunoblots of whole cell extracts from the indicated HeLa control and mutant cells in duplicate either left untreated or subjected to EBSS for 18h using the indicated antibodies. Independently cultured replicate samples were loaded in adjacent lanes and indicated by “Replicate #”. α-PCNA was used as a loading control. These blots were visualized with chemiluminescence. We used densitometry of blots of different exposures to estimate signal intensities for YIPF3, YIPF4, CALCOCO1, and TEX264. Signal intensities were averaged across replicates and normalized to untreated cells of the same genotype, and relative values are provided under the corresponding samples. c, Violin plot for Golgi-membrane protein Log₂ FC with or without 18h of EBSS in control, YIPF4^−/− or CALCOCO1^−/− HeLa cells. Mean abundance is indicated by bold line. Source Data

Extended Data Fig. 12. Role of YIPF4 in Golgi remodeling during differentiation of human ES cells to iNeurons.

a, Validation of CRISPR/Cas9 mediated deletion of YIPF4 in ES cells. b, Log₂FC (ATG12^−/− iN – control iN) and Log₂FC (YIPF4^−/− iN – control iN) values for proteins localized in individual subcellular compartments (12 day differentiation). Each sample/condition represents triplicate independent cultures. c, Heatmap of relative increase or decrease in the abundance of stem cell or iNeuron marker, comparing iNeurons versus ES cells. Each sample/condition represents triplicate independent cultures. d, Quantification of the abundance of YIPF3 in ATG12^−/− or YIPF4^−/− ES cells. TMT intensities for triplicate analyses are shown. Unpaired two-sided t-test, p < 0.05, ***. Error bars represent S.D. n.s., not significant. Bars are mean values and error bars represent S.D. p-values from left to right: p = 0.631, p = 0.0002, p < 0.0001. e, Violin plots for Log₂FC (YIPF4^−/−/WT) iNeurons (day 12) for the indicated sets of proteins. The extent of Golgi-membrane protein stabilization in YIPF4^−/− iNeurons is similar to that seen in ATG12^−/− iNeurons. Source Data

Extended Data Fig. 13. CARGO: an interactive website to interrogate Cellular Autophagy Regulation and Golgiphagy data from this work.

The website can be found at: https://harperlab.connect.hms.harvard.edu/CARGO_Cellular_Autophagy_Regulation_GOlgiphagy/. a, Example of visualization data combining orthogonal proteomics methods to create a priority list of putative autophagy factors. b, Example of visualization data for CAPs and subcellular compartment analysis. c, Example of visualization tools for mapping Golgiphagy and CAPs.