Selective VPS34 inhibitor blocks autophagy and uncovers a role for NCOA4 in ferritin degradation and iron homeostasis in vivo

Subjects

Abstract

Cells rely on autophagy to clear misfolded proteins and damaged organelles to maintain cellular homeostasis. In this study we use the new autophagy inhibitor PIK-III to screen for autophagy substrates. PIK-III is a selective inhibitor of VPS34 that binds a unique hydrophobic pocket not present in related kinases such as PI(3)Kα. PIK-III acutely inhibits autophagy and de novo lipidation of LC3, and leads to the stabilization of autophagy substrates. By performing ubiquitin-affinity proteomics on PIK-III-treated cells we identified substrates including NCOA4, which accumulates in ATG7-deficient cells and co-localizes with autolysosomes. NCOA4 directly binds ferritin heavy chain-1 (FTH1) to target the iron-binding ferritin complex with a relative molecular mass of 450,000 to autolysosomes following starvation or iron depletion. Interestingly, Ncoa4−/− mice exhibit a profound accumulation of iron in splenic macrophages, which are critical for the reutilization of iron from engulfed red blood cells. Taken together, the results of this study provide a new mechanism for selective autophagy of ferritin and reveal a previously unappreciated role for autophagy and NCOA4 in the control of iron homeostasis in vivo.

Main

Cells rely on the autophagy pathway to survive diverse cellular insults such as nutrient depletion, accumulation of protein aggregates, damaged mitochondria or intracellular bacteria1,2,3. Not surprisingly, there is much interest in the roles that autophagy and its substrates have in human disease. For example, small nucleotide polymorphisms in the autophagy genes ATG16L1 and ATG7 have been detected in a subset of Crohn’s disease4,5,6 and Huntington’s disease patients7, respectively, and mice deficient for these genes exhibit phenotypes that resemble the human disorders8,9. Aggregation-prone proteins, such as the Huntington’s disease-associated forms of Huntingtin or the α-Synuclein (A53T) protein found in Parkinson’s disease, can be degraded through autophagy to alleviate their toxic effects10. Additional autophagy substrates probably exist, some of which may have an important role in human disease.

The regulation of autophagy involves several distinct biochemical steps that control the genesis of autophagosomes, substrate recruitment, lysosome–autophagosome fusion and the final degradation of autolysosome contents11. One of the earliest steps involves the redistribution of the ULK1–ATG13–FIP200 protein complex to the surface of the endoplasmic reticulum for recruitment of the vacuolar protein sorting 34 (VPS34)–Vps15–Beclin1–ATG14 complex. VPS34 is also known as class III phosphatidylinositol-3-kinase (PI(3)K) and in Saccharomyces cerevisiae, where it is the only PI(3)K, it is essential for autophagy12,13. VPS34 kinase activity is thought to be responsible for synthesis and deposition of phosphatidylinositol-3-phosphate (PtdIns(3)P) at autophagosome formation sites, which leads to the recruitment of PtdIns(3)P-binding proteins and additional factors14,15,16. For example, the ATG5–ATG12–ATG16L1 protein complex is required for autophagosome membrane extension and to correctly localize microtubule-associated protein 1, light chain 3 (LC3). Through the action of ATG7 and ATG3, LC3 is then directly conjugated to phosphatidylethanolamine on the autophagosomal membrane, which generates lipidated LC3, referred to as LC3-II (ref. 17). Recruitment of substrates into autophagosomes is regulated by substrate receptors such as p62 and NBR1 that bridge interactions between LC3-II and substrates18. Many autophagy substrates are known to be ubiquitin-modified, which can provide a recognition motif for ubiquitin-associated domains (UBA) contained within autophagy receptors19,20. Several studies have shown that multiple cell types from Atg5- or Atg7-deficient mice have marked increases in the levels of ubiquitylated proteins9,21,22, which is consistent for a role of ubiquitin in targeting substrates to autophagosomes.

In this study, we generated a new autophagy inhibitor to demonstrate in mammalian cells that VPS34 enzymatic function is essential for de novo lipidation of LC3 and the degradation of autophagy substrates such as p62 and mitochondria. Affinity purification of the ubiquitin proteome that accumulated in VPS34-inhibited cells led to the discovery of many autophagy substrates including NCOA4, which physically binds to the ferritin protein complex and directs it to autolysosomes for degradation. Ncoa4-deficient mice have a profound accumulation of iron in the spleen indicating an imbalance between iron storage and recycling in vivo.

RESULTS

PIK-III is a highly selective inhibitor of VPS34 catalytic function

We sought to identify low-molecular-weight inhibitors of VPS34 that could be used as negative regulators of autophagy. Following high-throughput screening of compound libraries, a hit containing a bisaminopyrimidine core was identified that exhibited both potent and selective inhibition of VPS34 compared with other related lipid kinases. Medicinal chemistry optimization efforts further improved potency and selectivity leading to the identification of the 4-aminopyridine-containing PIK-III (Fig. 1a). Biochemical profiling demonstrated that PIK-III is at least 100-fold-selective for VPS34 over related lipid kinases such as PI(3)Kα and the protein kinase mTOR (Fig. 1b) as well as an additional 44 protein kinases (Supplementary Table 1). To determine whether PIK-III could modulate PtdIns(3)P-mediated effectors in mammalian cells we monitored the distribution of a PtdIns(3)P-specific lipid-binding domain fused to GFP (GFP–FYVE; ref. 23). PIK-III inhibited the localization of GFP–FYVE with a half-maximum inhibitory concentration (IC50) of 55 nM, which is >10,000 times more potent than the non-selective VPS34 inhibitor 3-MA (Fig. 1c, representative GFP images shown in Supplementary Fig. 1a). We also observed that although the nuclear translocation of GFP–FOXO3A is sensitive to class I PI(3)K inhibitors such as NVP-BKM120 or GDC-0941 (refs 24, 25), it is non-responsive to PIK-III (Fig. 1d, representative GFP images shown in Supplementary Fig. 1b).

Figure 1: PIK-III is a selective inhibitor of VPS34 enzymatic activity.
figure1

(a) Chemical structure of PIK-III. (b) The indicated kinases were incubated with different doses of PIK-III and IC50 values were determined. (c) U2OS cells expressing GFP–FYVE were incubated with the indicated compounds for 2 h, fixed, stained with Hoechst 33342 and imaged by automated epifluorescence microscopy. GFP-positive puncta per cell were calculated and used to determine IC50 values for each compound. Data points represent values from three wells of one experiment. (d) U2OS cells expressing GFP–FOXO3A were incubated with the indicated compounds for 2 h, fixed, stained with Hoechst 33342 and imaged by automated epifluorescence microscopy. The ratio of nuclear to cellular GFP fluorescence intensity was calculated and used to determine IC50 values for each compound. Data points represent values from three wells of one experiment.

To determine why PIK-III demonstrates high selectivity for VPS34, the co-structure of human VPS34 in complex with PIK-III was determined to 2.8 Å (Fig. 2a). The overall topology of human VPS34 shown here is similar to that of PI(3)Kγ and Drosophila melanogaster VPS34 (ref. 26). The active site is relatively narrow, hydrophobic and best accommodates co-planar aromatic groups (Fig. 2b, c), which is typical for both lipid and protein kinases. A prominent hydrophobic pocket, bounded in part by the side chains of Phe 612, Pro 618 and Phe 684, accommodates the cyclopropyl functional group in PIK-III (Fig. 2b). PIK-III forms two hydrogen-bond interactions between the compound’s donor/acceptor moiety and the backbone amide and carbonyl oxygen of residue Ile 685. An extended solvent-mediated hydrogen-bonding network bridges interactions with the PIK-III aminopyrimidine moiety and the side chains of Asp 671 and Asp 644. Both VPS34 and PI(3)Kα have a hydrophobic pocket in the ATP-binding site that is capped by amino acids of the P-loop. This pocket in VPS34 is displaced towards the hinge relative to the PI(3)Kα pocket. In VPS34, Phe 612 is of primary importance as its position permits the cyclopropyl group to fit into the hydrophobic cavity and allows an optimal interaction with the hinge. Superposition of both active sites showed that the PI(3)Kα pocket is wider proximal to the hinge region compared with VPS34 and the critical phenylalanine has been replaced in PI(3)Kα with a methionine residue that does not accommodate the PIK-III cyclopropyl group (Fig. 2d).

Figure 2: Human VPS34 structure and molecular basis for PIK-III selectivity.
figure2

(a) General domain structure of human VPS34 kinase comprising the helical domain (yellow) and the kinase domain (green). The ATP-binding pocket and active site lie between the amino-terminal lobe and the carboxy-terminal lobe of the kinase domain. (b) Close-up view of the VPS34 active site with PIK-III bound in the ATP-binding pocket. PIK-III forms two hydrogen-bond interactions between the compound’s donor/acceptor moiety and the backbone amide and carbonyl oxygen of residue Ile 685. A prominent hydrophobic pocket, bounded in part by the side chains of Phe 612, Pro 618 and Phe 684 accommodates the cyclopropyl functional group. An extended solvent-mediated hydrogen-bonding network bridges interactions between the PIK-III aminopyrimidine moiety and the side chains of Asp 671 and Asp 644. (c) The binding properties surface of the VPS34 active site is superimposed on the solvent-accessible surface. Green surfaces are regions with hydrophobic binding properties; note the hydrophobic nature of the active site floor and the deep hydrophobic pocket in which the cyclopropyl moiety fits. Blue and red surfaces represent hydrogen bond donor and acceptor regions, respectively. PIK-III is shown bound in the active site. (d) Superposition of the PIK-III-bound VPS34 pocket surface with a mesh (white) mapping the human PI(3)Kα active site. The displacement of the PI(3)Kα hydrophobic pocket away from the hinge (relative to that of VPS34) is readily apparent.

PIK-III inhibits autophagy and LC3 lipidation in mammalian cells

To determine whether inhibition of VPS34 function impacts autophagy we monitored LC3 and known autophagy substrates such as damaged mitochondria or the autophagy cargo receptor p62. In H4 cells expressing the mCherry–GFP–LC3 reporter27 PIK-III inhibited the formation of mCherry-positive autolysosomes and increased the cytosolic signal of LC3 under basal conditions and when autophagy was induced with the mTOR inhibitor AZD8055 (Fig. 3a). In a CCCP-induced mitophagy model, PIK-III inhibited the clearance of mitochondria, visualized by Tom20 (Fig. 3b). In addition, PIK-III prevented the turnover of GFP-tagged p62 under basal conditions and when autophagy was activated with the mTOR inhibitor AZD8055 (Fig. 3c). Endogenous levels of the substrate receptors p62, NBR1 and NDP52 also accumulated following treatment with the VPS34 inhibitor (Fig. 3d). PIK-III treatment led to an increase in the levels of LC3-I in H4 and PSN1 cells (Fig. 3d, e) consistent with the role of VPS34 in regulating LC3 lipidation. Interestingly, in Panc10.05 cells (Fig. 3f) PIK-III increased the levels of LC3-II in parallel with LC3-I suggesting a cell type-specific response. To more definitively determine the role of VPS34 activity in regulating de novo LC3 lipidation we overexpressed inducible dominant-negative ATG4B (ref. 28) to prevent LC3 lipidation and then treated these cells with or without PIK-III as the expression of ATG4BC74A was turned off in the presence of bafilomycin A1, an inhibitor of lysosomal function (Fig. 3g). Robust LC3 lipidation was detected after the doxycycline washout and this was completely suppressed by PIK-III (Fig. 3g, compare lanes 5 and 6). This result shows that VPS34 enzymatic function is essential for LC3 lipidation in mammalian cells.

Figure 3: PIK-III is a robust inhibitor of autophagy and LC3 lipidation in mammalian cells.
figure3

(a) H4 cells expressing mCherry–GFP–LC3 were treated overnight with the indicated compounds, fixed, stained with Hoechst 33342 and imaged by automated acquisition. Representative images of an overlay of GFP (green), mCherry (red) and Hoechst (blue) are shown with a scale bar equivalent to 10 μm. Bars represent the mean GFP- and mCherry-positive LC3 puncta per cell and data points represent values from 18 imaged fields in two wells from one experiment. (b) HeLa cells expressing GFP–Parkin were treated with PIK-III for 12 h followed by the addition of CCCP for 12 h, fixed, stained for endogenous Tom20 and imaged. Representative images of an overlay of Tom20 (red) and Hoechst (blue) are shown with a scale bar equivalent to 50 μm. Bars represent the mean cellular Tom20 staining intensity per cell and data points represent values from four wells from one experiment. (c) H4 cells expressing GFP–p62 were treated overnight with the indicated compounds, fixed, stained with Hoechst 33342 and imaged. Representative images of an overlay of GFP (green) and Hoechst (blue) are shown with a scale bar equivalent to 10 μm. Bars represent the mean GFP-positive puncta area per cell and data points represent values from 18 imaged fields in 2 wells from one experiment. (d) Steady-state levels of autophagy markers visualized by western blotting in H4 GFP–p62 cells treated overnight with 0.5 μM AZD8055 or 5 μM PIK-III. (e) Steady-state levels of p62, LC3 and GAPDH visualized by western blotting in PSN-1 cells treated for 48 h with the indicated doses of PIK-III. (f) Steady-state levels of p62, LC3 and GAPDH visualized by western blotting in Panc10.05 cells treated for 48 h with the indicated doses of PIK-III. (g) RKO cells expressing inducible HA–ATG4B-C74A were treated with doxycycline (DOX induction) for five days and then incubated in DOX-free medium for 24 h (DOX washout). Bafilomycin A1 was added with or without PIK-III at the beginning of the washout period (lanes 5 and 6). Steady-state levels of ATG4B, LC3 and GAPDH were visualized by western blotting. Uncropped images of blots are shown in Supplementary Fig. 7.

Ubiquitin proteomics identifies autophagy substrates in VPS34-inhibited cells

On the basis of our observations that PIK-III treatment inhibits autophagy (Fig. 3) and that many known autophagy substrates are ubiquitylated29 we designed a proteomics strategy to discover additional substrates (Fig. 4a). Quantitative liquid chromatography–mass spectrometry analysis of diglycyl-enriched ubiquitylated peptides and global proteome based on total peptides detected in flow-through material (Fig. 4b) identified a quantitative increase (log2 (fold change) >1.5 in two replicates for both ubiquitylated peptides and total protein) in the levels of the well-established autophagy substrate receptors SQSTM1 (p62; ref. 30), NBR1 (ref. 31) and CALCOCO2 (NDP52; ref. 32 and see Supplementary Table 2 for full data sets and experimental statistics). TAX1BP1 is a paralogue of NDP52 and has been recently described as an autophagy substrate receptor33. We also identified TMEM59, a transmembrane protein that localizes to and targets endosomal compartments for autophagic degradation34, and NDFIP2, a short-lived protein that resides in the trans-Golgi network and can be delivered to the lysosome for degradation35. Hence, PIK-III treatment increased the levels of many proteins with known links to autophagy. Interestingly, accumulation of the candidate autophagy substrate NCOA4 was second only to NBR1 and was chosen for detailed follow-up (Fig. 4b). We confirmed that steady-state levels of NCOA4 protein are robustly increased on autophagy inhibition with either PIK-III or bafilomycin A1 (Fig. 4c) or in cells where ATG7 had been disrupted using homologous recombination (Fig. 4d and Supplementary Fig. 2). When cells were nutrient deprived the levels of NCOA4 decreased in a time-dependent manner, similarly to p62 (Fig. 4e). Importantly, steady-state levels of NCOA4 messenger RNA were unchanged in these different contexts (Supplementary Fig. 3). Endogenous NCOA4 co-localized with LC3 and LAMP2 and this was further enhanced with bafilomycin A1 (Fig. 4f). In cells treated with PIK-III, NCOA4 co-localized with LC3 but did not progress to autolysosomes as evidenced by the lack of detection in LAMP2-positive compartments, which is consistent with PIK-III blocking the initiation of autophagy. These data suggest that NCOA4 is an autophagy substrate that is associated with autolysosomes but not other vesicular structures such as EEA1-associated early endosomes (Fig. 4f).

Figure 4: PIK-III and ubiquitin affinity proteomics reveals NCOA4 as an autophagy substrate.
figure4

(a) Schematic of experimental workflow. DLD1 cells were converted to SILAC medium, incubated overnight in the presence or absence of 5 μM PIK-III, lysed, and ubiquitin-modified peptides enriched with anti-diglycyl antibody and quantified as described in the Methods. (b) Proteins represented by unique ubiquitylation sites with >1.5 log2 (fold change) in both replicates after diglycyl-peptide enrichment were searched for >1.5 log2 (fold change) at the proteome level in both replicates. The resulting 15 proteins are listed to the right in descending order of the average protein fold change in the flow-through replicate analyses. Two independent biological replicates are shown in crossplots. (c) Steady-state levels of NCOA4 in DLD1 cells treated with bafilomycin A1 or PIK-III for 20 h. (d) The levels of NCOA4 and p62 in ATG7 wild-type (+/+), heterozygous (+/−) or homozygous knockout (−/−) DLD1 clones 1 and 2 were determined. (e) DLD1 cells cultured in rich medium (Fed) were deprived of serum and amino acids (Starve) for the indicated times (hours, h) and levels of NCOA4, p62, LC3 and GAPDH were determined by western blotting. (f) Confocal microscopy was used to evaluate the cellular distribution of endogenous NCOA4, LC3, LAMP2 and EEA1 in cells treated overnight with bafilomycin A1 or PIK-III. Regions outlined with white dashed lines are magnified to the right of each panel. Scale bars in full panels on left, and zoomed panels on right correspond to 10 and 2 μm respectively. Uncropped images of blots are shown in Supplementary Fig. 7.

NCOA4 is essential for lysosomal targeting and degradation of ferritin

NCOA4 was originally identified as a coactivator of the androgen receptor36 but the autolysosomal distribution described above seems to be at odds with this possibility. Utilizing an unbiased approach to identify NCOA4-interacting proteins to uncover its functions and role in autophagy, we identified the iron-binding proteins ferritin heavy chain (FTH1) and ferritin light chain (FTL) as candidate interactors (Supplementary Fig. 4 and Table 3). Immunoprecipitation of endogenous proteins confirmed that FTH1 and FTL interact with NCOA4 (Fig. 5a) and that FTH1 and FTL are degraded on autophagy activation by nutrient starvation (Fig. 5b). Furthermore, FTH1 and FTL co-localize with LC3 and LAMP2 under basal autophagy conditions and this is enhanced when autophagy is stalled with bafilomycin A1 (Fig. 5c). As for the cellular distribution of NCOA4 (Fig. 4f), no co-localization was detected between FTH1 or FTL and EEA1-positive endosomes (Fig. 5c).

Figure 5: NCOA4 targets ferritin proteins to the lysosome for degradation.
figure5

(a) Endogenous immunoprecipitation of NCOA4 with ferritin heavy (FTH1) and light (FTL) chains from DLD1 cell lysates. (b) Steady-state levels of FTH1 and FTL expressed in DLD1 cells growing in rich medium or starved for the times indicated (hours, h). (c) DLD1 cells were cultured in the presence or absence of bafilomycin A1 overnight, fixed and processed for confocal immunofluorescence to determine the cellular distribution of endogenous FTH1 or FTL with LC3, LAMP2 or EEA1. Regions outlined with white dashed lines are magnified to the right of each panel. Scale bars in full and zoomed panels correspond to 10 and 2 μm respectively. (d) DLD1 parental or NCOA4−/− cells (clones 1 and 2; KO-1, KO-2) were starved for 6 h and the levels of NCOA4, FTH1, FTL and GAPDH determined. (e) DLD1 parental or NCOA4−/− cells were cultured in the presence or absence of bafilomycin A1 overnight, fixed and processed for confocal immunofluorescence to determine the cellular distribution of endogenous ferritin (FTH1–FTL) and LAMP2. Regions outlined with white dashed lines are magnified to the right of each panel. Scale bars of full and zoomed panels correspond to 10 and 2 μm respectively. Uncropped images of blots are shown in Supplementary Fig. 7.

To determine whether NCOA4 is required for starvation-induced degradation of FTH1 and FTL, NCOA4 was selectively disrupted in DLD1 cells using genome editing based on clustered regularly interspaced short palindromic repeats37,38,39 (CRISPR). Two independent guide RNA sequences were used to control for potential off-target effects (Supplementary Table 4). In NCOA4−/− cells the steady-state levels of FTH1 and FTL were increased and there was a complete loss of a faster-migrating FTH1 species (Fig. 5d). Although starvation decreased FTH1 and FTL in control cells this was completely suppressed in NCOA4−/− cells (Fig. 5d). Critically, this is an NCOA4-specific process because FTH1 and FTL are still starvation sensitive in the absence of p62 (Supplementary Fig. 5). Given that FTH1 and FTL were protected from starvation in NCOA4−/− cells we next examined their cellular distribution (Fig. 5e). FTH1 and FTL co-localized with LAMP2 in control but not in NCOA4−/− cells revealing that NCOA4 probably plays a key role in targeting ferritin proteins to autophagosomes (Fig. 5e).

NCOA4 interacts with the high-molecular-weight ferritin complex in an FTH1-dependent manner

To further characterize the nature of the interaction between NCOA4 and ferritin we analysed cell lysates fractionated by size-exclusion chromatography (SEC). In dimethylsulphoxide (DMSO)-treated WT DLD1 cells, NCOA4 co-eluted with the slower-migrating FTH1 band in fractions containing high-molecular-weight complexes with a relative molecular mass of 1,000,000 (Mr 1,000K; Fig. 6a, fractions 18 and 19) and with both FTH1 bands in fractions containing complexes ranging from Mr 800K to 600K (Fig. 6a, fractions 20–22). Interestingly, in lysates from cells treated overnight with bafilomycin A1, we observed NCOA4 in additional fractions that overlapped with the peak eluting fractions of both FTH1 bands (corresponding to Mr 450K iron-bound assembled ferritin), suggesting that the complex present in these fractions is vulnerable to lysosomal activity (Fig. 6a, fractions 23 and 24). In untreated cells, FTL was found in fractions containing both slow- and fast-migrating FTH1 bands, which peaked at Mr 450K corresponding to assembled ferritin (Fig. 6b). Furthermore, NCOA4 was excluded from fractions containing the low-molecular-weight FTH1 and FTL forms, which are probably monomeric in nature (Fig. 6a, b). Analysis of NCOA4−/− cell extracts using SEC revealed that the faster-migrating FTH1 band was absent in all fractions, and that the relative amount of FTH1 and FTL was greater in fractions containing higher- but not lower-molecular-weight complexes (Fig. 6b, compare fractions 23–26 with 29–31). From these experiments, we reasoned that the faster-migrating FTH1 band is a lysosomal degradation product that could incorporate into high-molecular-weight ferritin complexes (Fig. 6a, b). Introducing WT NCOA4 into NCOA4−/− cells allowed recovery of the faster-migrating FTH1 band in a PIK-III- and lysosomal inhibitor-sensitive manner (Fig. 6c). Together these observations suggest that NCOA4 interacts with the high-molecular-weight ferritin complex and is required for the processing of FTH1 in lysosomes. Although FTH1 and FTL both co-precipitate with NCOA4 (Fig. 5a), we sought to determine whether either or both FTH1 and FTL mediate this interaction. In WT DLD1 cells FTH1 and FTL co-precipitated with NCOA4 but in FTH1−/− cells FTL was not detected in immunoprecipitates (Fig. 6d). Interestingly, in FTL−/− cells the slower-migrating FTH1 band still co-precipitated with NCOA4 indicating that this form of FTH1 targets the ferritin complex to autophagosomes for ultimate processing in lysosomes (Fig. 6d). To account for the relative instability of FTL in the absence of FTH1 (Fig. 6d), we immunoprecipitated endogenous NCOA4 from FTH1−/−cells expressing ectopic FTL (Supplementary Fig. 6a), and in FTH1−/−/FTL−/−cells expressing ectopic FTH1 or FTL (Supplementary Fig. 6b) and confirmed that the NCOA4 interaction with ferritin is specific to FTH1.

Figure 6: Accumulation of the high-molecular-weight ferritin complex in NCOA4−/− cells.
figure6

(a) Cell extracts from DLD1 cells treated with DMSO or bafilomycin A1 overnight were resolved using SEC and each fraction was analysed for the presence of NCOA4 and FTH1 by western blot. Proteins elute in high- to low-molecular-weight fractions from 18 to 33, respectively. Dashed lines are for visual comparison only. (b) DLD1 parental or NCOA4−/− cell extracts were resolved using SEC and each fraction was analysed for the presence of NCOA4, FTH1 and FTL by western blot. (c) NCOA4−/− cells were transfected with control or SBP–NCOA4 cDNA for 36 h and incubated overnight in the presence or absence of the indicated inhibitors, lysed and the occurrence of the faster-migrating FTH1 band was determined. (d) Endogenous NCOA4 immunoprecipitations from DLD1 parental, FTH1−/− or FTL−/− (clones 1 and 2 for each genotype) protein extracts. The presence of FTH1 and FTL was detected by western blot. Uncropped images of blots are shown in Supplementary Fig. 7.

NCOA4 regulates iron-dependent turnover of ferritin by autophagy and iron homeostasis in vivo

Iron chelators such as deferasirox (DFX) can promote the degradation of ferritin proteins40. To determine any role for NCOA4 in this context we treated WT DLD1 or NCOA4−/− cells with DFX and found that the steady-state levels of FTH1 and FTL decreased in WT but not NCOA4−/− cells (Fig. 7a, b). The DFX-induced NCOA4-dependent turnover of FTH1 and FTL was blocked with PIK-III suggesting an autophagy-dependent process (Fig. 7a). Many studies have shown that ferritin proteins and iron can be distributed to lysosomes in various cells types including macrophages41,42 and deletion of Atg7 in haematopoietic cells leads to anaemia43. We reasoned that NCOA4 plays a key role in regulating aspects of iron homeostasis through its ability to target FTH1 and FTL for autophagic degradation. To address this possibility Ncoa4-deficient chimaeric mice (Ncoa4−/−) were generated by targeting Ncoa4 for homozygous deletion in mouse embryonic stem cells (mESCs) by CRISPR genome editing. Ncoa4−/− ESCs lacked Ncoa4 protein and as expected expressed only the slower-migrating form of Fth1 (Fig. 7c). Chimaeric mice were generated by blastocyst injection, and F0 progeny exhibiting 85–100% chimaerism were analysed at 11.5 weeks of age for the presence of general abnormalities and for accumulation of iron in heart, pancreas, spleen, liver, brain, bone-marrow and duodenum. In nearly every chimaeric Ncoa4−/− mouse a profound accumulation of intracytoplasmic brown pigment was observed in haematoxylin/eosin-stained splenic red pulp, which was confirmed to be iron deposits by Perls’ Prussian blue stain (Fig. 7d, e). Microscopic analysis indicated that iron deposits were associated with macrophages. Quantitative image analysis revealed a statistically significant increase in iron levels in age-matched Ncoa4−/− animals compared with WT chimaeric animals (Fig. 7f). Taken together these observations suggest that NCOA4 is responsible for the selective targeting of the ferritin complex to autophagosomes and in vivo this plays a critical role in maintaining normal iron levels in the spleen.

Figure 7: NCOA4 regulates iron homeostasis in vivo.
figure7

(a) DLD1 cells were incubated in the presence or absence of DFX (30 μM) and PIK-III (5 μM) and the levels of FTH1 and FTL were analysed using western blotting. (b) DLD parental or NCOA4−/− cells were incubated in the presence or absence of DFX and the levels of FTH and FTL were determined. (c) Ncoa4−/− mouse embryonic stem cells were analysed for Ncoa4 and Fth1 by western blotting. Note the complete absence of the lower Fth1 band consistent with earlier findings. (d) Representative haemotoxylin and eosin staining of spleens from control or Ncoa4−/− mice is shown and brown pigmented regions apparent in Ncoa4−/− sections are indicated (white arrows). (e) The presence of iron in spleens from control or Ncoa4−/− mice was determined using Perls’ Prussian blue staining. Representative micrographs (left panels) or magnifications (right panels) are shown. Scale bars correspond to 100 μm and 20 μm in full and magnified panels respectively. (f) Quantitative image analysis of the data shown in e was performed as described in the Methods. Each data point represents relative iron levels in a single mouse spleen section and horizontal lines are means of each group (n = 6 mice per group; P < 0.0006 for unpaired t-test). Uncropped images of blots are shown in Supplementary Fig. 7.

DISCUSSION

By generating the pharmacological probe PIK-III, we have demonstrated the pivotal role of VPS34 enzymatic function in the initiation of autophagy and ultimate degradation of substrates. Broad profiling of PIK-III across lipid and protein kinases showed that its activity is mainly restricted to VPS34, and in cells PIK-III inhibited PtdIns(3)P-dependent phenotypes but not FOXO3A redistribution, which is regulated by PI(3)Kα and PtdIns(3,4,5)P3. The high-resolution co-crystal structure of VPS34 with PIK-III revealed a relatively narrow active site in VPS34 compared with PI(3)Kα, into which the PIK-III cyclopropyl group can extend. Many studies have relied on 3-MA and wortmannin to causally link VPS34 and autophagy with certain biological processes but these agents also inhibit PI(3)Kα, and therefore AKT–mTOR signalling, which makes data interpretation challenging44,45,46. The present study shows that PIK-III has good selectivity over PI(3)Kα, does not inhibit the PI(3)K–AKT pathway and can be used to inhibit VPS34 enzymatic function with precision thus avoiding the need to use imperfect kinase inhibitors or to genetically ablate PIK3C3 expression, which alters the stoichiometry of VPS34-containing complexes.

In mammalian cells the VPS34–VPS15–Beclin 1–ATG14 complex regulates critical steps in the initiation of autophagy and a similar VPS34 complex functions in Saccharomyces cerevisiae47. However, results using Pik3c3-knockout mouse cells48,49,50,51,52 have generated controversy as to whether VPS34 is actually required for LC3 lipidation. By using an inducible dominant-negative ATG4B to reversibly ablate LC3-II in cells, we demonstrated that VPS34 enzymatic activity is indeed essential for de novo lipidation of LC3. Interestingly, under basal autophagy conditions some cell types treated with PIK-III show an overall increase in LC3-I and LC3-II, which resembles data from reports using Pik3c3-knockout cells. It is possible that these results can be explained by the existence of at least two distinct sources of PtdIns(3)P that are capable of regulating LC3 lipidation. Class II PI(3)K enzymes can contribute a fraction of the total cellular pool of PtdIns(3)P and recent literature suggests that this might in part explain why LC3-II is still present in Pik3c3-knockout cells53. However, studies in Saccharomyces cerevisiae showed that lipidation of the LC3 homologue Atg8 can still occur in the absence of Vps34 (ref. 13), the only PI(3)K in yeast, which argues that PtdIns(3)P may not be involved in this phenotype. Future studies will be needed to fully understand the mechanistic basis for VPS34/PtdIns(3)P-independent LC3 lipidation.

The removal and degradation of cytoplasmic contents through the autophagy pathway was previously considered to be a nonspecific ‘bulk’ process but recent advances suggest that autophagosomes acquire specific substrates through a series of ordered molecular steps involving substrate receptors and post-translational modifications such as ubiquitylation29. In the present study, we used global profiling of the ubiquitin-associated proteome present in PIK-III-treated cells to identify autophagy substrates. One of the substrates identified was NCOA4, which robustly accumulated in PIK-III-treated and ATG7−/− cells, co-localized with autolysosome markers and was degraded in the absence of nutrients. These phenotypes resemble those of well-validated substrate receptors such as p62 and NBR1 (refs 30, 31). NCOA4 affinity pulldowns and immunoprecipitations contained FTH1 and FTL, which constitute the Mr 450K ferritin protein lattice complex where cells store iron and maintain it in a non-reactive state54. NCOA4 co-eluted with FTH1/FTL in high-molecular-weight lysate fractions produced by SEC, suggesting that its interaction is with the ferritin lattice rather than FTH1 and FTL monomers. In NCOA4−/− cells high-molecular-weight forms of FTH1/FTL are more abundant and ferritin accumulates in the cytoplasm suggesting that the number of functional ferritin complexes increases in the absence of NCOA4. Intriguingly, several past studies have shown that ferritin proteins and iron distribute to autophagosomes and lysosomes40,41,42. Taken together with our observations that FTH1 and FTL are degraded in a NCOA4-dependent manner, we propose that NCOA4 is a selective autophagy receptor for the ferritin iron storage complex. Ferritin proteins are known to have extended stability in the presence of lysosomal hydrolases55 but do undergo limited proteolysis. The present work demonstrated that an electrophoretically faster-migrating FTH1 fragment was absent in NCOA4−/− cells or in cells treated with PIK-III, bafilomycin A1 or E64d/pepstatinA. We believe that this fragment is hemosiderin, a previously identified ferritin-derived fragment that is enriched in lysosomes and which has a reduced capacity to bind and retain iron56.

Mechanisms that regulate the storage and recycling of iron are critical for normal tissue homeostasis57. Red blood cell haemoglobin contains over half of the total iron in humans and following red cell engulfment by macrophages, this iron is either stored in the ferritin complex or effluxed out of the cell by ferroportin58. Disruption of murine Ncoa4 revealed that iron levels within the spleen are particularly vulnerable to the lack of NCOA4. The chromatography and in vivo observations suggest that in the absence of NCOA4 an increase in the number of ferritin complexes within splenic macrophages would lead to an imbalance between iron storage and recycling. This model is supported by previous studies showing that autophagic degradation of ferritin proteins is associated with mobilization of cellular iron40,59. Although future studies will be needed to determine whether Ncoa4−/− mice are susceptible to anaemia, the present study has revealed a role for NCOA4 in maintaining physiological levels of iron in vivo. Recently, another group reported NCOA4 as a receptor that mediates autophagic degradation of FTH1/FTL when supraphysiological levels of iron are depleted using chelators60. The present study provides important additional insight by demonstrating that NCOA4 and the ferritin lattice complex interact, and that NCOA4 is essential for lysosomal processing of this complex. Furthermore, results using Ncoa4−/− mice uncovered a previously unappreciated role for NCOA4 and autophagy in the control of iron in vivo. Interestingly, individuals with movement disorders and neurodegeneration have iron-ferritin inclusions and even mutations in FTL (refs 61, 62) and it is tempting to speculate that reduced autophagic turnover of the ferritin complex may contribute to such diseases. □

Methods

Identification of selective VPS34 kinase antagonists.

The human VPS34 cDNA sequence was sub-cloned into the pNAT234 vector that encodes GST/His6/S-tag and PreScission protease cleavage sites at the amino terminus of VPS34 (pNAT234/hVPS34). Recombinant baculovirus encoding GST–His6–VPS34 was used to infect Tn5 cells using standard protocols and 48 h postinfection cells were lysed. Extracts were centrifuged at 37,000g for 1 h and supernatant was added to glutathione beads for 1 h at 4 °C (GE Healthcare). Beads were then incubated with elution buffer containing 20 mM reduced glutathione and 50% (v/v) ethylene glycol. Protein was stable for up to 4 weeks when stored at 4 °C. For high-throughput screening a scintillation proximity assay (SPA) was developed. Briefly, diluted library compounds were added to individual wells of a 384-well assay plate followed by 4 μg ml−1 phosphoinositide and 10 ng VPS34 in a buffer containing 5 μM ATP, 50 nCi [33P]ATP, 40 mM HEPES (p H7), 10 mM MgCl2, 10 mM MnCl2 and 1 mM dithiothreitol (DTT). After incubation for 2 h at room temperature 100 mM EDTA and 6.7 mg ml−1 WGA-coated SPA beads were added to each well and then incubated for 4 h before reading plates on a ViewLux uHTS reader (Perkin Elmer). Screening hits were validated and then profiled for activity against multiple lipid and protein kinases (Fig. 1b and Supplementary Table 1). Hits with the greatest selectivity for VPS34 (that is, >100-fold over class I PI(3)K enzymes) were characterized using medicinal chemistry to enhance potency and selectivity. A more detailed description of the screen hits and structure–activity relationships will be given elsewhere.

Purification of VPS34 for crystallography.

Human VPS34 was expressed in Tn5 insect cells using a construct with an N-terminal his-tag, a spacer, and a TEV cleavage site (comprising a sequence of MSYY-HHHHHH-DYDIPTT-ENLYFQG-) followed by amino acids 293–887 of VPS34. Cell cultures were grown in shaker flasks at 27 °C and collected 2 days post-infection. Cells were pelleted and resuspended in lysis buffer (40 mM Tris-HCl, pH 7.5, 500 mM NaCl, 1 mM MgCl2, 1% (v:v) betaine, 1% (v:v) ethylene glycol, 1 tablet Complete Protease Inhibitor (Roche), 5 mM β-mercaptoethanol, and 1:20,000 (v:v) Benzonase). Cells were pelleted and frozen at −80 °C before processing. The cell pellet was subsequently thawed and the cells were lysed by douncing and the resulting cell debris was removed by ultracentrifugation. The cell supernatant was loaded onto a 5-ml IMAC column equilibrated with Buffer A (40 mM Tris-HCl, pH 7.5, 500 mM NaCl, 1 mM MgCl2, 5 mM imidazole, 1% (v:v) betaine, 1% (v:v) ethylene glycol, 1 tablet Complete Protease Inhibitor (Roche), 5 mM β-mercaptoethanol, and 1:20,000 (v:v) Benzonase) and the column was washed with Buffer A supplemented with 35 mM of additional imidazole until an A280 nm baseline was reached. The bound VPS34 was then eluted by washing the column with Buffer A supplemented with 245 mM of additional imidazole. IMAC fractions were selected for pooling by SDS–PAGE analysis. Pooled fractions were loaded onto a SEC Superdex 200 column pre-equilibrated with Buffer B (20 mM Tris-HCl, pH 7.2, 200 mM NaCl, 1% (v:v) betaine, 1% (v:v) ethylene glycol, 0.02% (v:v) CHAPS, 5 mM DTT). The protein was eluted by washing the column with Buffer B. SEC fractions were selected for pooling by SDS–PAGE analysis. The pooled fractions were concentrated to 7 mg ml−1 in Buffer B for crystallization.

VPS34 crystallography.

For co-crystallization, VPS34 protein and PIK-III were mixed and incubated on ice for 1 h (final PIK-III concentration was 1 mM). Before crystallization, the mixture was passed through a 0.2 μm filter. The protein:ligand complex was crystallized using the hanging-drop vapour diffusion method in Nextal plates: 6 μl of protein solution was mixed with 4 μl of precipitant, which consisted of 20% (w:v) PEG 3350, 100 mM bis-tris propane, and 200 mM Na-K-phosphate. The resulting drop was suspended over a reservoir of 0.3 ml of precipitant and sealed with a screw cap. The crystals grew at 30 °C in approximately 12–24 h. Crystals were incubated in a cryoprotectant consisting of 80% precipitant and 20% glycerol (v:v) for 1 h and then flash-frozen in liquid nitrogen before data collection. Data collection was performed with the Advanced Light Source (ALS) 5.0.2 at the Lawrence Berkeley National Laboratory with a wavelength of 0.97740 Å at 100°K (173.15 °C). The structure of VPS34 was solved using the kinase domain of an in-house structure of human PI(3)Kα as a search model using the program PHASER (ref. 63). The PIK-III ligand was fitted into the difference density observed in an Fo–Fc map and the co-structure was refined to convergence using the PHENIX and BUSTER refinement suite64. The structure was refined to final Ramachandran values of: 515 residues in favoured conformations (98%), 12 residues were in allowed conformations (2%), and 1 residue was an outlier. Structural waters were added as deemed appropriate according to peaks present in 2Fo–Fc and Fo–Fc maps. Relevant data and refinement statistics are presented in Supplementary Table 5. The PDB accession code for the VPS34–PIK-III structure is 4PH4.

Antibodies, chemicals, cDNA, siRNA and CRISPR constructs.

The following antibodies were used: GAPDH (Cell Signaling 2118, western blotting (WB) 1:10,000), FTH1 (Cell Signaling 4393, WB 1:1,000; immunofluorescence (IF) 1:100), FTL (Proteintech 10727-1-AP, WB 1:500; IF 1:100), ferritin (Rockland Immunochemicals 200-401-090-0100, IF 1:100 (used in Fig. 5e)), NCOA4 (Santa Cruz 373739, WB 1:200; 28749, IF 1:200; Bethyl Laboratories A302-272A, IP 4 μg mg−1 of protein), LC3 (Novus NB100-2220, WB 1:1,000; MBL M115-3, IF 1:100), p62 (BD 610833, WB 1:5,000), LAMP2 (Santa Cruz 18822, IF 1:200), EAA1 (BD 610456, IF 1:200), ATG7 (Cell Signaling 8558, WB 1:1,000), ATG5 (Cell Signaling 2630, WB 1:1,000), NBR1 (Abnova H00004077-M01, WB 1:5,000), NDP52 (Cell Signaling 9036, WB 1:1,000), ATG4B (Cell Signaling 5299, WB 1:1,000), TOM20 (Abcam 56783, IF 1:400), SBP-tag (Pierce MAB10764, WB 1:5,000). The following chemicals were used: dimethylsulphoxide (DMSO, Sigma), PIK-III, NVP-BKM120 and deferasirox (Novartis Pharmaceuticals), bafilomycin A1 (Tocris), GDC-0941 and AZD8055 (ChemieTek), CCCP (carbonyl cyanide 3-chlorophenylhydrazone, SIGMA), 3-MA (Calbiochem), wortmannin (Calbiochem) and ferric ammonium citrate (FAC, Sigma). The GFP–FYVE construct was generated by cloning the FYVE domain from Hrs (hepatocyte growth factor-regulated tyrosine kinase substrate; nucleotides 439–669) in tandem, separated with a QGQGS linker, downstream of GFP in the pLEGFP-C1 vector (Clontech). The EGFP–p62 construct was generated by inserting PCR-amplified full-length human SQSTM1 cDNA, N-terminally tagged with EGFP, into pRetroX-Tight-Pur (Clontech) using NotI and EcoRI restriction sites. The HA–ATG4B-C74A construct was generated by subcloning the HA–ATGB-C74A cDNA (ref. 65) into pLKO-TREX-On (ref. 66). SBP-tagged NCOA4 was generated by shuttling codon-optimized NCOA4 from pENTR2.1 (Invitrogen) into pQC-XIPE-N-SBP-attR using LR ClonaseII (Invitrogen). The following small interfering RNA (siRNA) pools were obtained from Dharmacon: scramble (D-001810-10-20) and human NCOA4 (L-010321-00-0010). Two human SQSTM1 siRNAs were obtained from Qiagen and mixed equally (siRNA 1: 5′-tcggaggatccgagtgtgaat-3′, siRNA 2: 5′-ccgaatctacattaaagagaa-3′). Genome editing based on clustered regularly interspaced short palindromic repeats or CRISPR (refs 37, 38, 39) was used to create DLD1-knockout cell lines. Guide RNA sequences were selected using CRISPR.MIT.edu (see Supplementary Table 4 for sequences). For NCOA4 targeting, two high-scoring guide RNA sequences in exon 3 were selected to create two independent knockout lines. For FTL targeting, guide RNAs targeting exon 2 and exon 3 were selected to create two independent knockout lines. For FTH1 targeting, guide RNAs found in intron 1 and the 3′ UTR were used simultaneously to create one knockout line with a 1.1 kb deletion. Oligonucleotides encoding guide sequences and AarI restriction sites were annealed and ligated into pU6-aarIgRNA-nlsSPycas9nls-2acherry for parallel expression of the guide sequence and the Cas9 enzyme.

Mammalian cell culture.

All cells were cultured in a humidified incubator at 37 °C and 5% CO2. Cell culture reagents were obtained from Invitrogen unless otherwise specified. Retroviral transduction was performed as described previously27. U2OS cells expressing GFP–FOXO3A (ref. 67) or GFP–FYVE were maintained in McCoy’s 5A medium containing 10% FBS. U2OS GFP–FYVE cells were generated by retroviral delivery of the GFP–FYVE construct into U2OS cells followed by selection for stable integration with G418. Individual GFP=-FYVE-expressing clones were isolated and characterized for imaging assays. Parental and mCherry–GFP–LC3-expressing H4 cells68 were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS. H4 Tet-Off EGFP–p62 cells were maintained in DMEM supplemented with 10% Tet system approved FBS (Clontech). H4 Tet-Off cells were generated by retroviral delivery of pRetroX-Tet-Off Advanced (Clontech) into H4 cells and selection for stable integration using G418. H4 Tet-Off EGFP–p62 cells were then generated by retroviral delivery of EGFP–p62 in pRetroX-Tight-Pur followed by selection for stable integration using puromycin. H4 Tet-Off EGFP–p62 single-cell clones were isolated by FACS and clone C8 was picked for imaging assays on the basis of homogeneous and doxycycline-sensitive EGFP–p62 expression. HeLa GFP–PARK2 cells were engineered by Vectalys SAS using cell transduction with a pLV EF1GFP WPRE viral vector and cells were maintained in DMEM supplemented with 10% FBS. FACS sorting was used to isolate cell pools expressing homogeneous GFP–Parkin levels. RKO cells expressing inducible HA=-ATG4B-C74A were maintained in EMEM (ATCC) supplemented with 10% Tet system approved FBS (Clontech). RKO HA–ATG4B-C74A cells were generated by retroviral delivery of HA–ATGB-C74A in pLKO TREX-On followed by selection for stable integration using puromycin. The expression of HA–ATG4B-C74A was induced using 100 ng ml−1 doxycycline. Panc10.05, PSN-1 and DLD1 cells lines were maintained in RPMI medium 1640 supplemented with 10% FBS. ATG7 conditional knockout DLD1 cells were generated by Horizon Discovery using a two-step homologous recombination strategy. The first ATG7 allele was disrupted by inserting a neomycin cassette into exon 15 of the first ATG7 allele thereby deleting the Cys 572 catalytic residue. The second ATG7 allele was targeted by inserting a hygromycin cassette immediately downstream of exon 15 and loxP sites flanking the 5′ of exon 15 and the 3′ of the resistance gene. Neomycin- and hygromycin-resistant DLD1 clones (ATG7 conditional knockout) were isolated and validated using q-PCR and western analysis (Supplementary Figs 2 and 3b). Two selected clones were then infected with adenoviral cre recombinase (Vector Biolabs) to delete the second ATG7 allele and generate homozygous knockout cells, which were validated using western blotting for ATG7 and ATG5–ATG12 conjugation (Supplementary Fig. 2). DLD1 SBP–NCOA4 and control cells for affinity proteomics were generated by retroviral delivery of pQC-XIPE-N-SBP-attR-NCOA4 and empty vector followed by selection of stable integration using 1.5 μg ml−1 puromycin. For CRISPR targeting, DLD1 cells were plated and grown to 70% confluence and transfected using Lipofectamine 3000 (Invitrogen). For NCOA4 and FTL targeting, 14 μg of plasmid DNA was transfected per 10-cm dish. For FTH1 targeting, 7 μg of each Cas9/gRNA vector was mixed for transfection. After 24 h cells were trypsinized, resuspended, strained through 40 μm mesh filters (BD Biosciences) and single mCherry positive cells were sorted into 96-well plates with a FACS ARIA (BD Biosciences). Clones were expanded and screened for loss of protein expression by western blotting.

High-content assays and image quantification.

For the mCherry–GFP–LC3 assay, cells were plated into clear-bottom 96-well plates (Corning) and treated the next day for 18 h with the indicated compounds. Cells were then fixed for 1 h at room temperature in 4% PFA containing Hoechst 33342, washed three times with PBS and subjected to automated epifluorescence microscopy using an In Cell Analyzer 2000 (GE Healthcare). Nine different fields were imaged per well using ×20 magnification and DAPI, FITC and TexasRed filter sets. Images were quantified using the Multi Target Analysis module of the In Cell Analyzer Workstation software (GE Healthcare). In brief, nuclei were identified on the basis of the Hoechst 33342 staining, cells were defined using a collar and the number of cellular LC3 puncta was quantified in the FITC and TexasRed images using multi-top-hat segmentation and averaged per well.

For the GFP–p62 assay, cells were plated into clear-bottom 96-well plates (Corning) and treated the next day for 18 h with the indicated compounds. Cells were then fixed for 1 h at room temperature in 4% PFA containing Hoechst 33342, washed three times with PBS and subjected to automated epifluorescence microscopy using an In Cell Analyzer 2000 (GE Healthcare). Nine different fields were imaged per well using ×20 magnification and DAPI and FITC filter sets. Images were quantified using the Multi Target Analysis module of the In Cell Analyzer Workstation software (GE Healthcare). In brief, nuclei were identified on the basis of the Hoechst 33342 staining, cells were defined using a collar and the total area of cellular p62 puncta was quantified in the FITC images using multi-top-hat segmentation and averaged per well.

For the GFP–FYVE assay, cells were plated into clear-bottom 384-well plates (Greiner) and treated the next day for 2 h with the indicated compounds. Cells were then fixed for 1 h at room temperature using Mirsky’s fixative (National Diagnostics) supplemented with Hoechst 33342, washed four times with TBS and subjected to automated epifluorescence microscopy using an In Cell Analyzer 1000 (GE Healthcare). Three different fields per well were imaged using ×20 magnification and DAPI and FITC filter sets. Images were quantified using the granularity module of the In Cell Analyzer Workstation software (GE Healthcare). In brief, nuclei were identified on the basis of the Hoechst 33342 staining, cells were defined using a collar and the number of cellular GFP–FYVE puncta was quantified in the FITC images using top-hat segmentation and averaged per well.

For the GFP–FOXO assay, cells were plated into clear-bottom 384-well plates (Greiner) and treated the next day for 2 h with the indicated compounds. Cells were then fixed for 1 h at room temperature in 4% PFA containing Hoechst 33342, washed three times with TBS and subjected to automated epifluorescence microscopy using an In Cell Analyzer 2000 (GE Healthcare). Three different fields were imaged per well using ×20 magnification and DAPI and FITC filter sets. Images were quantified using the dual object module of the In Cell Analyzer Workstation software (GE Healthcare). In brief, nuclei were identified on the basis of the Hoechst 33342 staining, cells were defined using a collar and the ratio of nuclear to cellular fluorescence intensity was quantified in the FITC images and averaged per well.

For the mitophagy assay, cells were plated into clear-bottom 384-well plates (Greiner) and treated the next day by adding PIK-III for 12 h followed by the addition of CCCP for 12 h. A final concentration of 10 μM CCCP was used and control wells were treated with equivalent volumes of DMSO. Cells were then fixed for 30 min at room temperature in 4% PFA and washed three times with PBS. Fixed cells were blocked and permeabilized with PBS containing 5% FBS and 0.1% Triton for 30 min at 37 °C and then washed 3 times with PBS. The cells were then incubated in primary antibody (mouse anti-TOM20, Abcam, ab56783 IF 1:400) for 2 h at 37 °C. After 3 washes with PBS, cells were incubated for 90 min at 37 °C in secondary antibody (donkey anti-mouse 680, Invitrogen, A10043, 1:6,000) containing Hoechst 33342. Cells were then washed 5 times with PBS containing 0.02% sodium azide. Plates were then subjected to automated fluorescence confocal microscopy using an Opera high-throughput confocal microscope (PerkinElmer). Five different fields were imaged per well using ×20 magnification and DAPI, FITC and Cy5 filter sets. Images were quantified using the Acapella High Content Imaging and Analysis software. Briefly, nuclei were identified on the basis of the Hoechst 33342 staining, cell cytoplasm was defined using the GFP background staining and the fluorescence intensity of cytoplasmic TOM20 was quantified in the Cy5 images and averaged per well.

Immunofluorescence and confocal microscopy.

DLD1 cells were plated on 12 mm round glass coverslips (Chemglass) in 24-well dishes and grown overnight to 60% confluency. Treatments were performed as indicated. Cells were fixed for 10 min in −20 °C methanol and blocked and permeabilized in a solution containing 1:1 Odyssey blocking buffer (LiCor)/PBS (Invitrogen) with 0.1% Triton X-100 (Sigma) and 1% normal goat serum (Invitrogen) for 1 h at room temperature. Primary antibodies were added overnight at 4 °C in blocking buffer described above. After rinsing in PBS, secondary ALEXA-conjugated antibodies (Invitrogen) were diluted 1:1,000 in blocking solution and applied for 1 h at room temperature. Cells were then rinsed in PBS, mounted on glass slides with ProLong Gold Antifade with 4′,6-diamidino-2-phenylindole (DAPI, Invitrogen) and sealed with clear nail polish. Images were acquired with ×63 objective on an inverted microscope (Axiovert 200; Carl Zeiss) equipped with a motorized stage, a Yokogawa CSU-X1 spinning-disc head, and an EMCCD camera (Evolve 512, Photometrics) using Zen software (Zeiss). Images sets were acquired sequentially on the same day with equivalent exposure times. Composite images were exported to Photoshop (Adobe) and adjustments were made equally across images from all treatment groups.

Cell lysis, immunoprecipitation and immunoblotting.

For western blotting cells were lysed in RIPA (Teknova) with sodium dodecyl sulphate (SDS, Sigma) added to 1% final concentration, and protease inhibitor tablets (Roche). Lysates were homogenized by passage through Qiashredder columns (Qiagen), and protein levels were quantified by Lowry DC protein assay (Biorad). Samples were denatured in 4 × LDS (Invitrogen) with 10 × sample reducing reagent (Invitrogen) at 100 °C for 5 min. Equivalent protein amounts were loaded on Novex (Invitrogen) or XT (Biorad) 4–12% Bis-tris SDS–PAGE gels and resolved. Proteins were transferred to nitrocellulose using standard methods and membranes were blocked in 5% non-fat dry milk (Biorad) in PBS with 0.2% Tween-20 (Teknova). Primary antibodies were diluted in blocking solution and were incubated with membranes at 4 °C overnight; HRP-conjugated secondary antibodies were diluted in blocking solution and incubated with membranes at room temperature for 1 h. Western blots were developed using Pico or west femto Super Signal ECL reagents (Pierce) and film (GE Healthcare). For immunoprecipitations, cells were lysed in a buffer containing 10 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM ETDA, 0.4% NP-40 (v/v), and clarified by centrifugation. Antibodies were added to equal amounts of protein lysates and incubated at 4 °C overnight. Immunocomplexes were captured using protein G Sepharose 4 fast flow (GE) at 4 °C for 1 h. Immunocomplexes were eluted with 2× LDS (Invitrogen) and 10× sample reducing reagent (Invitrogen). Cleanblot detection reagent (Pierce) was used where the primary western blot antibody was raised in a species matching the antibody used for immunoprecipitation; otherwise HRP-conjugated secondary antibodies were used.

qPCR.

For quantitative PCR where indicated, all experiments were performed identically to those prepared for western blotting, except cells were lysed and processed with RNeasy Plus Kit (Qiagen) according to the manufacturer’s instructions for RNA purification. cDNA synthesis was performed with TaqMan reverse transcription reagents (Roche) according to the manufacturer’s instructions. qPCR was performed with TaqMan Fast Advanced Master Mix (Invitrogen) using the following TaqMan probes: ATG7 (HS00893759, HS00893758, HS00197348), PPIA (HS04194521), NCOA4 (HS01033772) from Invitrogen. The reactions were run on a ViiA7 real-time PCR machine (Applied Biosystems) and changes in ATG7 and NCOA4 were calculated relative to PPIA by the ΔΔCt method.

Ubiquitin proteomic profiling.

DLD1 cells were grown in DMEM flex media supplemented with 10% dialysed FBS, 2 mg l−1 glucose, 2 mM L-glutamine (Invitrogen). Light media were prepared with 100 μg ml−1 light L-lysine (K0), and 100 μg ml−1 light L-arginine (R0) (Invitrogen). Heavy media were prepared with 100 μg ml−1 heavy L-lysine (K6 [K-13C6]) (Invitrogen), and 100 μg ml−1 heavy L-arginine (R10 [R-13C6, 15N4]) (Cambridge Isotopes). Once incorporation of heavy amino acids was determined to be >94%, cells were expanded to 150 cm dishes (25 dishes per group) in heavy and light media and treated for 20 h with DMSO (Sigma) or 5 μM PIK-III, lysed in 9 M urea and mixed 1:1 according to protein concentration (total of 14 mg) determined with the 660 nm protein assay (Thermo Scientific). Lysates were reduced with 10 mM DTT (Indofine Chemicals) at 30 °C for 30 min followed by alkylation with 25 mM iodoacetamide (Sigma-Aldrich), for 30 min at room temperature in the dark. The lysates were diluted to a final urea concentration of 1 M with 20 mM HEPES, pH 7 and digested 1:75 with trypsin (Pierce) overnight at room temperature, acidified, and desalted with DSC-18 tubes (Supelco). Peptides containing the diglycyl remnant were enriched using the PTMScan Ubiquitin Remnant Motif (K-ɛɛ-GG) antibody (Cell Signaling Technology) essentially following the manufacturer’s directions69 with modifications as described previously70. Before enrichment, the ubiquitin remnant antibody was crosslinked to the agarose beads with dimethyl pimelidate (Sigma) to minimize antibody loss. Peptides were separated using a Dionex Ultimate 3000 fitted with a 6.4 × 150 mm Zorbax C18Extend column with a flow rate of 1 ml min−1. Eighty 1 ml fractions were collected throughout the segmented gradient (0–8%B/7 min, 8–27%B/38 min, 27–31%B/4 min, 31–39%B/8 min, 39–60%B/7 min, 60–0%B/20 min, mobile phase A: 100% H2O; mobile phase B: 100% AcN; mobile phase C (modifier, constant at 4%): 200 mM ammonium formate, pH 10). Fractions were pooled non-contiguously (every eighth fraction) to obtain a total of 8 fractions that were individually subjected to the ubiquitin enrichment protocol. Each fraction was reconstituted in 1.4 ml Immunoaffinity Purification (IAP) buffer containing 50 mM MOPS, 10 mM Na2HPO4 and 50 mM NaCl pH adjusted to 7.2 with NaOH and placed on ice. One aliquot (40 μl packed bead volume) of crosslinked antibody resin was split evenly into 8 aliquots for enrichment. Incubation of sample and beads was performed with gentle end-over-end rotation at 4 °C for 1 h, followed by a 1 min, 2,000g spin to pellet the beads. The supernatant was removed and retained as the flow-through fraction. The antibody beads were washed twice with cold IAP buffer, followed by 2 washes of ice cold water. Ubiquitylated peptides were eluted from the beads with the addition of 50 μl of 0.15% trifluoroacetic acid (TFA) and allowed to stand at room temperature for 5 min. After centrifugation at 2,000g for 2 min the supernatant was carefully removed and retained. Eluted peptides were cleared of any stray beads or potential antibody contamination using C18 Stage Tips (Proxeon). Tips were washed twice with 20 μl 80% acetonitrile in 5% formic acid, followed by equilibration with 2 aliquots of 5% formic acid. Sample was loaded to the tips and flow-through retained. A final wash of 20 μl 30% acetonitrile in 5% formic acid was used to elute any peptides retained by the C18 membrane. Purified peptides were dried to completion and analysed by nanocapillary liquid chromatography-tandem mass spectrometry on an Easy-nLC 1000 HPLC system coupled to a Q-Exactive mass spectrometer (Thermo Scientific) using an in-house fabricated 75 μm ID spraying capillary packed with ReproSil-Pur 120 C18-AQ, 3 μm material (Dr. Maisch GmbH; 150 mm bed length) with a vented trapping column set-up (1 cm Michrom Magic C18AQ, 5 μm). The peptides were eluted with a gradient of 3% Buffer B (70% acetonitrile in 0.1% formic acid) to 45% B in 80 min (0.5%B min−1) delivered at a flow rate of 300 nl min−1 and using a top 12 HCD dynamic data acquisition method. A second 20 μl injection was made and data were acquired with an exclusion list created from the top 20 proteins identified for each fraction from the first injection to increase depth of analysis. Data were analysed with MaxQuant version 1.3.0.5 (Andromeda search engine with MaxQuant quantification71,72 searched against the UniProt database of canonical human protein sequences (version 9 Jan 2013 plus typical laboratory contaminants) with the addition of MaxQuant-generated reversed database to calculate false discovery rates). Default settings were used with the exception of allowing 3 missed cleavages. Carbamidomethyl (C) was selected as a fixed modification and variable modifications of diglycyl-Lys (restricted to internal lysines only), Oxidation (M) and N-terminal acetylation were applied. Search tolerances of 20 ppm (first pass) and 5 ppm (main search) were allowed for precursor ions; a false discovery rate cutoff of 1% was applied at protein, peptide and modification levels. Fold changes were calculated only for proteins that had 2 or more measured ratios of unique or razor peptides, including only unmodified peptides or peptides containing either oxidized methionine or N-terminal acetylation: peptides containing the diglycyl modification were not used in protein quantification calculations. Lists in Supplementary Table 3 are based on GlyGly(K)Sites.txt and ProteinGroups.txt tables from MaxQuant. Data were visualized for further analysis using Spotfire DXP and Graphpad Prism 6.

Affinity proteomics.

DLD1 cells expressing empty vector or N-SBP–FLAG-tagged NCOA4 were pelleted and resuspended with ×2 volume of packed cell pellet in ice-cold lysis buffer (10 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM ETDA, 0.4% NP-40 (v/v) (Thermo Scientific), 1 mM DTT, 1× HALT Protease inhibitor cocktail (Thermo Scientific). Resuspended cell pellets were allowed to incubate on ice for 30 min, after which the cell pellets were dounce homogenized with 10 strokes of a tight-fitting pestle. The lysates were spun at 800g for 10 min at 4 °C and the resulting supernatants were spun at 100,000g for 1 h at 4 °C. The supernatants from the 100,000g spin were collected (S100) and the pellets was discarded. The protein concentration of the S100 lysates was determined using the Pierce 660 nm protein assay kit according to the manufacturer’s instructions. For enrichment of N-SBP–FLAG-tagged NCOA4 and associated proteins from S100 lysates, 25 mg of lysate (at 5 mg ml−1) was incubated with 50 μl (packed bead volume following 200× g for 2 min spin) of streptavidin high-capacity agarose beads (Pierce) with end-over-end agitation overnight at 4 °C. Streptavidin-captured proteins were purified using mini-columns (MobiTec). After passing the lysate–bead mixtures through individual mini-columns, the resins were washed with 3 × 1 ml Wash Buffer 2 (10 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM EDTA, 0.4% NP40, 1 mM DTT) and 3 × 1 ml Wash Buffer 1 (10 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM DTT). To elute N-SBP–FLAG-tagged NCOA4 and associated proteins from the streptavidin agarose beads, 50 μl of ×2 NuPAGE LDS Sample Buffer (Invitrogen) was added to each mini-column (sealed with the provided plug and screw cap) and incubated at 55 °C for 30 min. The eluted proteins were collected by centrifugation at 14,000g for 2 min at room temperature and the material was cleaned-up using SDS Removal Spin Columns (Pierce) according to the manufacturer’s instructions. Proteins were subjected to in-solution trypsinization overnight at 37 °C followed by isobaric labelling using iTRAQ reagents (AB SCIEX) using 114 for N-SBP–FLAG-NCOA4 and 116 for wild-type control. Samples were mixed and separated using high-pH reverse-phase chromatography (Dionex Ultimate 3000 HPLC, Waters Xbridge column (1 mm × 15 cm), mobile phase A: 100% H2O; mobile phase B: 100% AcN; mobile phase C (modifier, constant at 10%): 200 mM ammonium formate, pH 10; flow rate: 250 μl min−1, 60 min effective gradient). Fractions were pooled into 8 samples that were analysed by nanocapillary liquid chromatography–tandem mass spectrometry on an Easy-nLC 1000 HPLC system coupled to a Q-Exactive mass spectrometer (Thermo Scientific), using an in-house fabricated 75-μm-ID spraying capillary packed with ReproSil-Pur 120 C18-AQ, 3 μm material (Dr. Maisch GmbH; 150 mm bed length) with a vented trapping column set-up (1 cm Michrom Magic C18AQ, 5 μm). The peptides were eluted with a gradient of 3% Buffer B (70% acetonitrile in 0.1% formic acid) to 45% B in 80 min (0.5%B min−1) delivered at a flow rate of 300 nl min−1 and using a top 12 HCD dynamic data acquisition method. Peptide mass and fragmentation data were searched against a combined forward–reverse UniProt database of canonical human protein sequences using Mascot (Matrix Science, UniProt version 16 September 2013 plus typical laboratory contaminants). Precursor and fragment ion tolerances were set to 10 ppm and 0.1 Da, respectively, allowing for 2 missed tryptic cleavages. Carbamidomethyl (C) was selected as a fixed modification and iTRAQ4plex (K), iTRAQ4plex (N-term), Oxidation (M) as variable modifications. Peptide and protein validation was done using Transproteomic pipeline v3.3sqall (Institute for Systems Biology; http://tools.proteomecenter.org/software.php) using a false positive threshold of <1% for protein identifications. For each peptide sequence and modification state, reporter ion signal intensities from all spectral matches were summed for each reporter ion type and corrected according to the isotope correction factors given by the manufacturer. Only peptides unique to a given protein within the total data set of identified proteins were used for relative protein quantification. Peptide fold changes were calculated (N-SBP–FLAG-NCOA4 over wild-type control) and protein fold changes were derived as median peptide fold change. P values were calculated using a one-way t-test (arbitrarily set to 1 for non-significant single peptide quantifications) and adjusted using the Benjamini–Hochberg false discovery rate. Data were visualized for further analysis using Spotfire DXP.

Size-exclusion chromatography.

DLD1 cells were grown to 90% confluency in 15-cm dishes, washed twice with cold PBS and lysed in a buffer containing 10 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM ETDA, 0.4% NP-40 (v/v) with a mini EDTA-protease inhibitor tablet (Roche). Lysates were incubated at 4 °C for 20 min to ensure complete lysis and centrifuged at 16,000g for 20 min to pellet insoluble material. Protein amounts were quantified for all cell lysates by Lowry DC protein assay (BioRad) and samples were adjusted to equal concentrations. Immediately before injection, all samples were passed through a 0.22 μM filter. Samples were processed and analysed on a 24-ml Superose 6 10/300 GL column (GE Healthcare) connected to an Akta Avant fast protein liquid chromatography (FPLC) system controlled by Unicorn software. The column was fully equilibrated with the lysis buffer above before running. For a given experiment, an equivalent amount of protein per sample in 300 μl lysis buffer was injected (1–2 mg total). The samples were eluted from the column in 0.5 ml fractions in the lysis buffer described above at a constant flow rate of 0.25 ml min−1. For FTH1 western blots, 4× LDS sample buffer and 10× reducing agent (Invitrogen) were added to an aliquot of the fractions and run on 4–12% bis-tris gels. For NCOA4 and FTL blots, 300 μl of fractions was precipitated with trichloroacetic acid (TCA, Sigma) by standard methods, 50 μl 2× LDS and 10× reducing agent (Invitrogen) were added to precipitates and they separated by SDS–PAGE as above. All fractions were first run on 26-well gels to determine the protein-containing fractions, and then fractions 18–33 were loaded onto 18-well gels for the blots shown in Fig. 6.

Generation and characterization of Ncoa4−/− mice.

All animal procedures employed in this study were approved by the Novartis Institutes for BioMedical Research Institutional Animal Care and Use Committee. Ncoa4-null ESCs were generated by CRISPR-mediated gene targeting in Actb–EGFP-tagged C57BL/6J ESCs. The plasmids expressing Cas9 and sgRNA against the sequence in intron 1 or 6 (see Supplementary Table 4 for guide sequences) were constructed and co-transfected together with a PGK–Puro cassette into ESCs using Lipofectamine 2000 (Life Technologies). Puromycin-resistant ESC clones with deletion between exons 2 and 6 were first screened by PCR, and then subjected to western blot analysis to obtain the clones with the loss of Ncoa4 protein expression. An Ncoa4-null ESC clone and a parental Actb–EGFP-tagged wild-type clone were used to inject B6-albino blastocysts, and high-percentage chimaeric males, based on coat colour, were selected for phenotypic characterization.

To evaluate iron levels, tissues were collected from 11.5-week-old control or Ncoa4-deficient mice at necropsy, fixed in 10% neural buffered formalin for 48 h, transferred to 70% ethanol and processed for haemotoxylin and eosin (H&E) or Perls’ Prussian blue staining. Stained tissues were imaged by automated bright-field microscopy using a Scanscope AT Turbo slide scanner (Leica Biosystems). Quantitative image analysis was performed on digital images to confirm the results of semiquantitative scoring by light microscopy and to precisely and accurately quantify potential differences in the relative amount of tissue iron between control chimaeric and Ncoa4-null chimaeric mice. Quantification of the area of Perls’ staining on digital images was performed using a colour deconvolution algorithm software application (Aperio ePathology Solutions; Leica Biosystems). A positive pen tool was used to outline the region of interest on digital images (that is, the circumference of the spleen). The colour deconvolution algorithm is used to separate a digital image into channels, corresponding to the actual colours of the stains used to stain tissues to enable accurate calculation of the area for each individual stain. For both blue staining (defined as colour 1) for iron in spleen tissues and the eosin counterstain (defined as colour 2), which was used to produce a contrasting background, colour calibration was performed to define the optical density (OD) values for red, green and blue components of each stain; these values were specified in the input parameters to the algorithm. As only two stains were used for tissue staining, the third colour optical density values in the colour deconvolution algorithm were set to zero (colour 3). Colour deconvolution algorithm input parameters and threshold values, used to create the macro that was employed for subsequent quantification of Perls’ staining on digital images, are provided in Supplementary Table 6. The area of Perls’ staining was calculated and expressed as the percentage of medium-positive pixels (orange in mark-up images) plus the percentage of strong-positive pixels (red in mark-up images) per total stained area of analysis.

Statistical analysis and determination of sample size.

For animal experiments, 6 WT and 6 Ncoa4−/− animals were used for a total of 12 animals. No statistical method was used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. Graphpad Prism software was used for statistical analysis. A two-tailed Student’s t-test for unpaired data was used for a single comparison between the two experimental groups. All other numerical data are as described in the figure legends.

Accession numbers.

Primary accessions: the PDB accession code for the VPS34–PIK-III structure is 4PH4.

Accession codes

Accessions

Protein Data Bank

References

  1. 1

    Begun, J. & Xavier, R. J. Autophagy at the crossroads of metabolism and cellular defense. Curr. Opin. Gastroenterol. 29, 588–596 (2013).

    CAS  Article  PubMed  Google Scholar 

  2. 2

    Huang, J. & Brumell, J. H. Bacteria–autophagy interplay: a battle for survival. Nat. Rev. Microbiol. 12, 101–114 (2014).

    CAS  PubMed  Google Scholar 

  3. 3

    Mizushima, N., Levine, B., Cuervo, A. M. & Klionsky, D. J. Autophagy fights disease through cellular self-digestion. Nature 451, 1069–1075 (2008).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  4. 4

    The Wellcome Trust Case Control Consortium, Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature 447, 661–678 (2007).

    Article  CAS  PubMed Central  Google Scholar 

  5. 5

    Hampe, J. et al. A genome-wide association scan of nonsynonymous SNPs identifies a susceptibility variant for Crohn disease in ATG16L1. Nat. Genet. 39, 207–211 (2007).

    CAS  Article  PubMed  Google Scholar 

  6. 6

    Rioux, J. D. et al. Genome-wide association study identifies new susceptibility loci for Crohn disease and implicates autophagy in disease pathogenesis. Nat. Genet. 39, 596–604 (2007).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  7. 7

    Metzger, S. et al. Age at onset in Huntington’s disease is modified by the autophagy pathway: implication of the V471A polymorphism in Atg7. Hum. Genet. 128, 453–459 (2010).

    CAS  Article  PubMed  Google Scholar 

  8. 8

    Cadwell, K. et al. A key role for autophagy and the autophagy gene Atg16l1 in mouse and human intestinal Paneth cells. Nature 456, 259–263 (2008).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  9. 9

    Komatsu, M. et al. Loss of autophagy in the central nervous system causes neurodegeneration in mice. Nature 441, 880–884 (2006).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  10. 10

    Rubinsztein, D. C., Codogno, P. & Levine, B. Autophagy modulation as a potential therapeutic target for diverse diseases. Nat. Rev. Drug Discov. 11, 709–730 (2012).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  11. 11

    Klionsky, D. J. & Schulman, B. A. Dynamic regulation of macroautophagy by distinctive ubiquitin-like proteins. Nat. Struct. Mol. Biol. 21, 336–345 (2014).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  12. 12

    Kihara, A., Noda, T., Ishihara, N. & Ohsumi, Y. Two distinct Vps34 phosphatidylinositol 3-kinase complexes function in autophagy and carboxypeptidase Y sorting in Saccharomyces cerevisiae. J. Cell Biol. 152, 519–530 (2001).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  13. 13

    Obara, K., Noda, T., Niimi, K. & Ohsumi, Y. Transport of phosphatidylinositol 3-phosphate into the vacuole via autophagic membranes in Saccharomyces cerevisiae. Genes Cells 13, 537–547 (2008).

    CAS  Article  PubMed  Google Scholar 

  14. 14

    Axe, E. L. et al. Autophagosome formation from membrane compartments enriched in phosphatidylinositol 3-phosphate and dynamically connected to the endoplasmic reticulum. J. Cell Biol. 182, 685–701 (2008).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  15. 15

    Matsunaga, K. et al. Autophagy requires endoplasmic reticulum targeting of the PI3-kinase complex via Atg14L. J. Cell Biol. 190, 511–521 (2010).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  16. 16

    Polson, H. E. et al. Mammalian Atg18 (WIPI2) localizes to omegasome-anchored phagophores and positively regulates LC3 lipidation. Autophagy 6, 506–522 (2010).

    CAS  Article  PubMed  Google Scholar 

  17. 17

    Mizushima, N., Yoshimori, T. & Ohsumi, Y. Role of the Apg12 conjugation system in mammalian autophagy. Int. J. Biochem. Cell Biol. 35, 553–561 (2003).

    CAS  Article  PubMed  Google Scholar 

  18. 18

    Johansen, T. & Lamark, T. Selective autophagy mediated by autophagic adapter proteins. Autophagy 7, 279–296 (2011).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  19. 19

    Rogov, V., Dotsch, V., Johansen, T. & Kirkin, V. Interactions between autophagy receptors and ubiquitin-like proteins form the molecular basis for selective autophagy. Mol. Cell 53, 167–178 (2014).

    CAS  Article  PubMed  Google Scholar 

  20. 20

    Shaid, S., Brandts, C. H., Serve, H. & Dikic, I. Ubiquitination and selective autophagy. Cell Death Differ. 20, 21–30 (2013).

    CAS  Article  PubMed  Google Scholar 

  21. 21

    Hara, T. et al. Suppression of basal autophagy in neural cells causes neurodegenerative disease in mice. Nature 441, 885–889 (2006).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  22. 22

    Komatsu, M. et al. Impairment of starvation-induced and constitutive autophagy in Atg7-deficient mice. J. Cell Biol. 169, 425–434 (2005).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  23. 23

    Gillooly, D. J. et al. Localization of phosphatidylinositol 3-phosphate in yeast and mammalian cells. EMBO J. 19, 4577–4588 (2000).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  24. 24

    Folkes, A. J. et al. The identification of 2-(1H-indazol-4-yl)-6-(4-methanesulfonyl-piperazin-1-ylmethyl)-4-morpholin-4-yl-thieno[3,2-d]pyrimidine (GDC-0941) as a potent, selective, orally bioavailable inhibitor of class I PI3 kinase for the treatment of cancer. J. Med. Chem. 51, 5522–5532 (2008).

    CAS  Article  PubMed  Google Scholar 

  25. 25

    Maira, S. M. et al. Identification and characterization of NVP-BKM120, an orally available pan-class I PI3-kinase inhibitor. Mol. Cancer Ther. 11, 317–328 (2012).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  26. 26

    Miller, S. et al. Shaping development of autophagy inhibitors with the structure of the lipid kinase Vps34. Science 327, 1638–1642 (2010).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  27. 27

    Nyfeler, B., Bergman, P., Wilson, C. J. & Murphy, L. O. Quantitative visualization of autophagy induction by mTOR inhibitors. Methods Mol. Biol. 821, 239–250 (2012).

    CAS  Article  PubMed  Google Scholar 

  28. 28

    Fujita, N. et al. An Atg4B mutant hampers the lipidation of LC3 paralogues and causes defects in autophagosome closure. Mol. Biol. Cell 19, 4651–4659 (2008).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  29. 29

    Kirkin, V., McEwan, D. G., Novak, I. & Dikic, I. A role for ubiquitin in selective autophagy. Mol. Cell 34, 259–269 (2009).

    CAS  Article  PubMed  Google Scholar 

  30. 30

    Bjorkoy, G. et al. p62/SQSTM1 forms protein aggregates degraded by autophagy and has a protective effect on huntingtin-induced cell death. J. Cell Biol. 171, 603–614 (2005).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. 31

    Kirkin, V. et al. A role for NBR1 in autophagosomal degradation of ubiquitinated substrates. Mol. Cell 33, 505–516 (2009).

    CAS  Article  PubMed  Google Scholar 

  32. 32

    Thurston, T. L., Ryzhakov, G., Bloor, S., von Muhlinen, N. & Randow, F. The TBK1 adaptor and autophagy receptor NDP52 restricts the proliferation of ubiquitin-coated bacteria. Nat. Immu. 10, 1215–1221 (2009).

    CAS  Article  Google Scholar 

  33. 33

    Newman, A. C. et al. TBK1 kinase addiction in lung cancer cells is mediated via autophagy of Tax1bp1/Ndp52 and non-canonical NF-κB signalling. PLoS ONE 7, e50672 (2012).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  34. 34

    Boada-Romero, E. et al. TMEM59 defines a novel ATG16L1-binding motif that promotes local activation of LC3. EMBO J. 32, 566–582 (2013).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  35. 35

    Ohzono, C. et al. Nedd4-interacting protein 2, a short half-life membrane protein degraded in lysosomes, negatively controls down-regulation of connexin43. Biol. Pharm. Bull. 33, 951–957 (2010).

    CAS  Article  PubMed  Google Scholar 

  36. 36

    Yeh, S. & Chang, C. Cloning and characterization of a specific coactivator, ARA70, for the androgen receptor in human prostate cells. Proc. Natl Acad. Sci. USA 93, 5517–5521 (1996).

    CAS  Article  PubMed  Google Scholar 

  37. 37

    Cong, L. et al. Multiplex genome engineering using CRISPR/Cas systems. Science 339, 819–823 (2013).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  38. 38

    Jinek, M. et al. A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity. Science 337, 816–821 (2012).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  39. 39

    Mali, P. et al. RNA-guided human genome engineering via Cas9. Science 339, 823–826 (2013).

    CAS  Article  Google Scholar 

  40. 40

    Asano, T. et al. Distinct mechanisms of ferritin delivery to lysosomes in iron-depleted and iron-replete cells. Mol. Cell. Biol. 31, 2040–2052 (2011).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  41. 41

    Bridges, K. R. Ascorbic acid inhibits lysosomal autophagy of ferritin. J. Biol. Chem. 262, 14773–14778 (1987).

    CAS  PubMed  Google Scholar 

  42. 42

    Ericsson, J. L. Studies on induced cellular autophagy. I. Electron microscopy of cells with in vivo labelled lysosomes. Exp. Cell Res. 55, 95–106 (1969).

    CAS  Article  PubMed  Google Scholar 

  43. 43

    Mortensen, M. et al. Loss of autophagy in erythroid cells leads to defective removal of mitochondria and severe anemia in vivo. Proc. Natl Acad. Sci. USA 107, 832–837 (2010).

    CAS  Article  PubMed  Google Scholar 

  44. 44

    Klionsky, D. J. et al. Guidelines for the use and interpretation of assays for monitoring autophagy. Autophagy 8, 445–544 (2012).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  45. 45

    Seglen, P. O. & Gordon, P. B. 3-Methyladenine: specific inhibitor of autophagic/lysosomal protein degradation in isolated rat hepatocytes. Proc. Natl Acad. Sci. USA 79, 1889–1892 (1982).

    CAS  Article  PubMed  Google Scholar 

  46. 46

    Wu, Y. T. et al. Dual role of 3-methyladenine in modulation of autophagy via different temporal patterns of inhibition on class I and III phosphoinositide 3-kinase. J. Biol. Chem. 285, 10850–10861 (2010).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  47. 47

    Simonsen, A. & Tooze, S. A. Coordination of membrane events during autophagy by multiple class III PI3-kinase complexes. J. Cell Biol. 186, 773–782 (2009).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  48. 48

    Jaber, N. et al. Class III PI3K Vps34 plays an essential role in autophagy and in heart and liver function. Proc. Natl Acad. Sci. USA 109, 2003–2008 (2012).

    CAS  Article  PubMed  Google Scholar 

  49. 49

    Morishita, H. et al. Deletion of autophagy-related 5 (Atg5) and Pik3c3 genes in the lens causes cataract independent of programmed organelle degradation. J. Biol. Chem. 288, 11436–11447 (2013).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  50. 50

    Parekh, V. V. et al. Impaired autophagy, defective T cell homeostasis, and a wasting syndrome in mice with a T cell-specific deletion of Vps34. J. Immu. 190, 5086–5101 (2013).

    CAS  Article  Google Scholar 

  51. 51

    Willinger, T. & Flavell, R. A. Canonical autophagy dependent on the class III phosphoinositide-3 kinase Vps34 is required for naive T-cell homeostasis. Proc. Natl Acad. Sci. USA 109, 8670–8675 (2012).

    CAS  Article  PubMed  Google Scholar 

  52. 52

    Zhou, X. et al. Deletion of PIK3C3/Vps34 in sensory neurons causes rapid neurodegeneration by disrupting the endosomal but not the autophagic pathway. Proc. Natl Acad. Sci. USA 107, 9424–9429 (2010).

    CAS  Article  PubMed  Google Scholar 

  53. 53

    Devereaux, K. et al. Regulation of mammalian autophagy by class II and III PI 3-kinases through PI3P synthesis. PLoS ONE 8, e76405 (2013).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  54. 54

    Crichton, R. R. & Declercq, J. P. X-ray structures of ferritins and related proteins. Biochim. Biophys. Acta 1800, 706–718 (2010).

    CAS  Article  PubMed  Google Scholar 

  55. 55

    Coffey, J. W. & DeDuve, C. Digestive activity of lysosomes. I. The digestion of proteins by extracts of rat liver lysosomes. J. Biol. Chem. 243, 3255–3263 (1968).

    CAS  PubMed  PubMed Central  Google Scholar 

  56. 56

    Andrews, S. C., Treffry, A. & Harrison, P. M. Siderosomal ferritin. The missing link between ferritin and haemosiderin? Biochem. J. 245, 439–446 (1987).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  57. 57

    Ganz, T. & Nemeth, E. Hepcidin and disorders of iron metabolism. Annu. Rev. Med. 62, 347–360 (2011).

    CAS  Article  PubMed  Google Scholar 

  58. 58

    Donovan, A., Roy, C. N. & Andrews, N. C. The ins and outs of iron homeostasis. Physiology (Bethesda) 21, 115–123 (2006).

    CAS  Google Scholar 

  59. 59

    Kidane, T. Z., Sauble, E. & Linder, M. C. Release of iron from ferritin requires lysosomal activity. Am. J. Physiol. Cell Physiol. 291, C445-455 (2006).

    Article  CAS  Google Scholar 

  60. 60

    Mancias, J. D., Wang, X., Gygi, S. P., Harper, J. W. & Kimmelman, A. C. Quantitative proteomics identifies NCOA4 as the cargo receptor mediating ferritinophagy. Nature 509, 105–109 (2014).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  61. 61

    Gregory, A., Polster, B. J. & Hayflick, S. J. Clinical and genetic delineation of neurodegeneration with brain iron accumulation. J. Med. Genet. 46, 73–80 (2009).

    CAS  Article  PubMed  Google Scholar 

  62. 62

    Levi, S., Cozzi, A. & Arosio, P. Neuroferritinopathy: a neurodegenerative disorder associated with L-ferritin mutation. Best Pract. Res. Clin. Haematol. 18, 265–276 (2005).

    CAS  Article  PubMed  Google Scholar 

  63. 63

    McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 (2007).

    CAS  PubMed  PubMed Central  Google Scholar 

  64. 64

    Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. 66, 213–221 (2010).

    CAS  Article  Google Scholar 

  65. 65

    Proenca, C. C. et al. Atg4b-dependent autophagic flux alleviates Huntington’s disease progression. PLoS ONE 8, e68357 (2013).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  66. 66

    Wee, S. et al. PTEN-deficient cancers depend on PIK3CB. Proc. Natl Acad. Sci. USA 105, 13057–13062 (2008).

    CAS  Article  PubMed  Google Scholar 

  67. 67

    Maira, S. M. et al. Identification and characterization of NVP-BEZ235, a new orally available dual phosphatidylinositol 3-kinase/mammalian target of rapamycin inhibitor with potent in vivo antitumor activity. Mol. Cancer Ther. 7, 1851–1863 (2008).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  68. 68

    Nyfeler, B. et al. Relieving autophagy and 4EBP1 from rapamycin resistance. Mol. Cell. Biol. 31, 2867–2876 (2011).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  69. 69

    Rush, J. et al. Immunoaffinity profiling of tyrosine phosphorylation in cancer cells. Nat. Biotechnol. 23, 94–101 (2005).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  70. 70

    Udeshi, N. D., Mertins, P., Svinkina, T. & Carr, S. A. Large-scale identification of ubiquitination sites by mass spectrometry. Nat. Protoc. 8, 1950–1960 (2013).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  71. 71

    Cox, J. & Mann, M. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 26, 1367–1372 (2008).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  72. 72

    Cox, J. et al. Andromeda: a peptide search engine integrated into the MaxQuant environment. J. Proteome Res. 10, 1794–1805 (2011).

    CAS  Article  PubMed  Google Scholar 

Download references

Acknowledgements

We would like to thank E. George, A. Donovan, K. Mansfield, L. Klickstein, D. Glass, R. Xavier and M. Fishman for their input and advice.

Author information

Affiliations

Authors

Contributions

W.E.D., B.N., E.T., S.L., Z.W., S.D’A., H.W., D.W., J.K., P.B., S.U. and L.O.M. performed cell biology experiments. J.N., J.T., D.S., and M.S. performed proteomic experiments. R.A.E., J.C., C.L., D.E.B. and M.S.K. performed protein crystallography studies and analysed the co-structural data. P.F., M.E.D. and F.H., performed high-throughput screens, I.C-T., E.H., A.H. and E.P.K. performed medicinal chemistry experiments. W.E.D., H.L., Q.F. and T.N. generated mice. W.E.D., B.N., R.A.V. and L.O.M. analysed mouse data and provided interpretation. B.N., I.C-T., S.B.H., J.B., J.T., C.J.W., V.E.M., J.A.P., D.B., P.M.F., M.A.L., X.M., L.G.H., B.D.M., T.N., M.S., K.M.K., E.P.K. and L.O.M. provided supervision. W.E.D., B.N. and L.O.M. wrote the manuscript with input from other co-authors. L.O.M. devised the concept and supervised the project.

Corresponding author

Correspondence to Leon O. Murphy.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Integrated supplementary information

Supplementary Figure 1 PIK-III is a selective inhibitor of VPS34 in vitro and in mammalian cells.

(a) U2OS cells expressing GFP-FYVE were incubated with the indicated compounds (2 h), fixed, stained with Hoechst 33342 and imaged by automated epifluorescence microscopy. Representative images of GFP-FYVE are shown and scale bars are equivalent to 10 μm. GFP-positive puncta/cell were determined using an image analysis algorithm and these data were then used to determine IC50 values for each compound as shown in Fig. 1c. (b) U2OS cells expressing GFP-FOXO3A were incubated with the indicated compounds (2 h), fixed, stained with Hoechst 33342 and imaged by automated epifluorescence microscopy. Representative images of GFP-FOXO3A (green) and Hoechst 33342 nuclear staining (blue) are shown and scale bars are equivalent to 10 μm. The ratio of nuclear to cellular GFP fluorescence intensity was determined using an image analysis algorithm and these data were then used to determine IC50 values for each compound as shown in Fig. 1d.

Supplementary Figure 2 Validation of ATG7 conditional knockout DLD1 cells.

ATG7−/− DLD1 cells harbour a conditional knockout of ATG7, with exon 15 of the first allele being disrupted and exon 15 of the second allele being flanked with two loxP sites (see Experimental Procedures). ATG7+/+, ATG7−/+ or two ATG7flox/− DLD1 cell clones were infected with adenoviral-Cre recombinase (+ Adeno-Cre), lysed, and the levels of ATG7 and ATG5 evaluated using western blotting. Note the complete loss of ATG7 in the ATG7−/− clones treated with Adeno-Cre which results in full inhibition of ATG5-ATG12 conjugation. Cells treated with Adeno-Cre were used to generate data shown in Fig. 4d.

Supplementary Figure 3 The effect of autophagy disruption on NCOA4 mRNA.

(a) Total RNA was isolated from cells treated as in Fig. 4C and the relative levels of NCOA4 mRNA determined using quantitative RT-PCR. (b) Total RNA was isolated from cells treated as in Fig 4d and relative levels of NCOA4 and ATG7 mRNA determined using quantitative RT-PCR. Three different PCR probes (11/12, 14/15 and 15/16) were used to amplify ATG7 mRNA with the latter two flanking the targeted exon 15. (c) Total RNA was isolated from cells treated as in Fig. 4e and relative levels of NCOA4 mRNA determined using quantitative RT-PCR. Values are reported are two independent repeats from qPCR performed in quadruplicate.

Supplementary Figure 4 FTH1 and FTL interact with SBP-NCOA4 in DLD1 cells.

Steptavidin agarose beads were incubated with protein extracts from control or SBP-NCOA4-expressing DLD1 cells, bound proteins were eluted and analysed by LC-MS. Proteins enriched in the SBP-NCOA4 over control sample were determined as described in the methods and log10 fold ratios and adjusted P-Values are shown on the y and x axis, respectively. Note the robust enrichment of NCOA4, FTH1, FTL and TMSB10. The complete dataset is shown in Supplementary Table 3. P-values were calculated using a one-way T-test (arbitrarily set to 1 for non-significant single peptide quantitations) and adjusted using the Benjamini-Hochberg False Discovery Rate.

Supplementary Figure 5 Starvation-induced degradation of NCOA4, FTH1 and FTL does not require p62.

DLD1 cells transfected with the indicated siRNAs were left in rich medium (Fed) or nutrient starved (Starve) for the indicated times before cell lysis and western blot analysis.

Supplementary Figure 6 NCOA4 interacts specifically with FTH1.

(a) Endogenous NCOA4 was immunoprecipitated from FTH1−/− cells transfected with empty vector, N-SBP-FTH1 or N-SBP-FTL and the presence of FTH1 and FTL in the immunoprecipitates was determined by western blot. (b) Endogenous NCOA4 was immunoprecipitated from FTH1−/−/FTL−/− double knockout cells transfected with empty vector, N-SBP-FTH1 or N-SBP-FTL and the presence of FTH1 and FTL in the immunoprecipitates was determined by western blot. A specific interaction between NCOA4 and FTH1 only was detected. Asterisk indicates IgG band and arrow highlights location of FTL band.

Supplementary Figure 7 Uncropped scans of key western blots.

Area shown in figures is outlined in red and molecular weight markers are shown on the left.

Supplementary information

Supplementary Information

Supplementary Information (PDF 1154 kb)

Supplementary Table 1

Supplementary Information (PDF 83 kb)

Supplementary Table 2

Supplementary Information (XLSX 5760 kb)

Supplementary Table 3

Supplementary Information (XLSX 186 kb)

Supplementary Table 4

Supplementary Information (PDF 68 kb)

Supplementary Table 5

Supplementary Information (XLSX 8 kb)

Supplementary Table 6

Supplementary Information (PDF 78 kb)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Dowdle, W., Nyfeler, B., Nagel, J. et al. Selective VPS34 inhibitor blocks autophagy and uncovers a role for NCOA4 in ferritin degradation and iron homeostasis in vivo. Nat Cell Biol 16, 1069–1079 (2014). https://doi.org/10.1038/ncb3053

Download citation

Further reading

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing