Activation of ULK Kinase and Autophagy by GABARAP Trafficking from the Centrosome Is Regulated by WAC and GM130

1Cellular homeostasis requires apposition of anabolic and catabolic pathways acting in coordination to regulate protein synthesis, trafficking, and secretion. Autophagy, a conserved catabolic pathway, delivers cytosolic cargo, sequestered by autophagosomes, to the lysosome for degradation. Degraded cargo is recycled for protein synthesis and metabolism. Autophagosomes also sequester damaged organelles, misfolded proteins, or pathogens to maintain cell health and prevent infection.

2The formation of autophagosomes occurs from subdomains of the ER, mediated by ATG9-positive vesicles, followed by activation and recruitment of the ULK kinase complex and the class III PtdIns(3) (phosphatidylinositol(3)) kinase complex containing BECLIN 1 and ATG14. Association of ATG14 to the ER facilitates production of PtdIns(3)P and establishes a DFCP1-rich omegasome from which the phagophore, the nascent autophagosome, forms. Further expansion of the phagophore requires recruitment of lipidated Atg8 proteins via attachment of phosphatidylethanolamine (PE) on their C-terminal glycine by a ubiquitination-like pathway. The lipidation machinery is recruited to the phagophore membrane by the PtdIns(3)P-effector WIPI2B.

3The mammalian Atg8 family includes LC3A, LC3B, and LC3C, GABARAP, GABARAPL1, and GABARAPL2 (or GATE-16), which perform three functions in autophagy: first, to mediate expansion and closure of the phagophore membrane; second, as cargo receptors to recruit cytoplasmic cargo through the LIR (LC3-interacting region) motif; and third, as adaptors to recruit signaling and trafficking proteins and the autophagy machinery to the autophagosome.

4To uncover novel regulators of starvation-induced autophagy, we performed a siGenome screen and identified WAC (WW domain-containing adaptor with coiled coil), which is required for starvation-induced autophagy. WAC is in the nucleus and cytoplasm and associates with the Golgi complex. Nuclear WAC binds to RNA Polymerase II and the E3 ligase complex RNF20/40 to facilitate transcription-coupled H2B monoubiquitination at lysine 120, which is associated with sites of active transcription. Golgi WAC interacts with the deubiquitinase VCIP135, which binds VCP/p97. Here, starting from the identification of GM130 (GOLGA2) as a novel interactor of WAC, we show that WAC is tethered to the Golgi by GM130, and while WAC positively regulates autophagy, GM130 is a negative regulator. GM130 is a coiled-coil domain cis-Golgi matrix tethering protein required for vesicle trafficking and Golgi maintenance through mitosis.

5Furthermore, we show that WAC and GM130 interact with GABARAP and control its subcellular localization. GABARAP is on the Golgi and the pericentriolar matrix (PCM): tethering of GABARAP to the Golgi by GM130 inhibits autophagy, while the PCM pool of GABARAP contributes to autophagosome formation. WAC suppresses GM130 binding to GABARAP, regulating the PCM reservoir of GABARAP. GABARAP association with the PCM does not require lipidation. A photoconvertible GABARAP shows that PCM-associated GABARAP traffics to forming autophagosomes and GABARAP trafficking to the autophagosome requires WAC. Lastly, our data show that like WAC, GABARAP is required for ULK kinase activation and has a non-hierarchical function in formation of autophagosomes.

6In confirmation of our previous results, WAC depletion using siRNA reduced LC3 lipidation during starvation and caused an increase in SQSTM1 (better known as p62) levels under basal conditions (Figures S1A–S1C). Decreased LC3-II accumulation after WAC depletion was rescued with siRNA-resistant GFP-WAC (Figure 1A).

7To understand the role of WAC during autophagy, we determined the stage WAC acted to promote autophagy. In nutrient-rich conditions, mTORC1 phosphorylates and represses ULK1 kinase activity. Starvation inhibits mTORC1, allowing ULK1 activation and phosphorylation of the ULK complex member ATG13 on Ser318. A time course of starvation in WAC-depleted cells revealed that mTORC1 inactivation, measured by loss of phosphorylation of ULK1 at Ser757 and S6 at Ser240/244, was unaffected (Figures 1B–1C and Figure S1D). In contrast, ATG13 Ser318 phosphorylation was reduced after WAC depletion, suggesting that ULK1 activation, not mTORC1 deactivation, is impaired by loss of WAC.

8We next assessed WIPI2 puncta formation, which correlates with the amount of PtdIns(3)P produced by the BECLIN 1, ATG14-containing class III PtdIns(3) kinase complex and is a measure of the activation of the autophagy-specific PtdIns(3) kinase. WIPI2 puncta formation, but not protein levels, was reduced by WAC depletion in starved or Torin-treated cells (Figures 1D–1F, and Figures S1E–S1H). Moreover, siRNA-resistant Myc-WAC could rescue WIPI2 puncta formation, and thus PtdIns(3) kinase activation, during starvation (Figure S1I).

9Finally, starvation-induced ATG9 trafficking to forming autophagosomes, which is required for autophagy, was inhibited by WAC depletion. ATG9 remained localized in the Golgi region (Figure 1G) to an extent similar to that of ULK1 depletion. In short, during amino acid starvation WAC promotes ULK1 activation and the initiation of autophagosome formation.

10WAC localizes to the nucleus and the Golgi. To determine if WAC affects autophagy through its nuclear or cytoplasmic function, we performed a microarray analysis after knockdown of the nuclear complex of WAC/RNF20/RNF40. This generated an overlapping set of 301 potential gene targets that were significantly (p < 0.05) down- or upregulated by depletion of the nuclear complex (Figure 1H and Table S1). Analysis of the 301 hits (Table S1) revealed that p53 and p53 target genes (p = 0.00279) were significantly downregulated. This agrees with data that WAC regulates p53 targets and that RNF20 regulates p53 expression. Autophagy, membrane trafficking, or lysosomal biogenesis pathways were not significantly up- or downregulated. Thus, unlike TFEB or ZKSCAN3, the WAC/RNF20/40 complex does not regulate mRNA expression of the autophagy/lysosome pathway. Finally, knockdown of RNF40 did not affect LC3 lipidation or ULK1 signaling to ATG13 (Figures S1J and S1K).

11WAC interacts with Flag-HA-BECLIN 1 and GFP-BECLIN 1, but not endogenous BECLIN 1 (data not shown). To identify novel binding partners of WAC that regulate autophagy, we used pull-down experiments followed by mass spectrometry. A human WAC gene in a bacterial artificial chromosome (BAC) was modified with a C-terminal FLAP tag (see Figure S2A) and used to generate a HeLa cell line expressing WAC-FLAP; all four isoforms of WAC were tagged and expressed at levels similar to those of endogenous WAC (Figures S2B and S2D). WAC-FLAP localizes to the nucleus and Golgi and binds RNF40 (Figures S2C and S2D). WAC immunoprecipitations from HEK293 cells, GFP immunoprecipitations from WAC-FLAP HeLa cells and mass spectrometry identified the cis-Golgi coiled-coil tether GM130 as a WAC interactor (Figures S2E–S2G), as well as the WAC interactors RNF20, RNF40, and UBE2A.

12We confirmed binding of endogenous WAC to GM130 by co-immunoprecipitation (Figure 2A). Upon starvation of HEK293 or HeLa WAC-FLAP cells, the amount of GM130, but not RNF40, interacting with WAC or WAC-FLAP increased (Figures 2B–2E). As WAC is phosphorylated we asked if phosphorylation regulates the WAC-GM130 interaction. Using Phos-tag SDS-PAGE gels after 30 min starvation, WAC migrated faster, suggesting that WAC is dephosphorylated upon starvation (Figures 2F and 2G). To test if dephosphorylation enhanced the WAC-GM130 interaction, we salt washed GFP-WAC to remove bound proteins, including GM130, and treated with lambda phosphatase (Figure 2H). Dephosphorylated salt washed GFP-WAC had enhanced binding to endogenous GM130 compared to untreated salt washed GFP-WAC or N-terminal WAC (see Figure 2J). These results suggest that WAC is dephosphorylated during starvation and this enhances its ability to bind GM130.

13The highly conserved WAC C-terminal coiled-coil (CC) domain was required for interaction with GM130 (Figures 2I–2J and Figure S2H). Truncation analyses of WAC revealed that a 10 amino acid sequence in the WAC CC domain (aa610–620) containing one heptad (aa612–618) was required for GM130 binding (Figure 2K and Figure S2I). RNF20/40 bind the C terminus of WAC including the CC domain. Isoleucine 626 and leucine 629, at the “a” and “d” positions of a CC heptad respectively (I626S/L629S mutant), are essential for binding to RNF40 but not GM130 (Figure 2K). Finally, recombinant GST-WAC bound directly to purified GM130 and required the WAC CC domain (Figure 2L).

14GM130 interacts with proteins involved in Golgi trafficking including USO1 (p115), RAB1B, RAB33B, and GRASP65. Truncations of GM130 showed that full-length WAC, but not WAC ΔCC, binds to the C-terminal of GM130 at aa774–1002. This includes the fifth and sixth CC domains of GM130 as well as the GRASP65 binding site (Figures 2M and 2N).

15WAC colocalizes with GM130, and in addition with the COPI coat protein βCOP and the ERGIC marker ERGIC-53 (Figure 3A), but less with the TGN markers p230 (GOLGA4) and TGN46 (Figure 3B). To test if GM130 is required for Golgi localization of WAC, we knocked down GM130. In cells with no detectable GM130, we never saw juxtanuclear WAC localization, even though the TGN remained intact (Figure 3C). GFP-WAC but not ΔCC WAC, localized to the Golgi region (Figure 3D). In addition, we performed a knockdown rescue experiment. In WAC knockdown cells, siRNA-resistant GFP-tagged full-length WAC but not ΔCC WAC, was detected on the Golgi (Figure S3A). We conclude that GM130 tethers WAC to the Golgi.

16Using the recently developed mitochondrially targeted golgins that can tether vesicles to mitochondria (see Figure S3B), we show that GM130-MAO can target GFP-WAC to mitochondria but GFP-WAC NT remained exclusively nuclear (Figure 3E). We next asked if the Golgi and nuclear pools of WAC are exchangeable. WAC has a putative nuclear export signal (NES) (Figure 2I), so we inhibited CRM-1-dependent nuclear export using leptomycin B (LMB) (Figure S3C). In contrast to control and 2 hr LMB, 24 hr LMB inhibited GFP-WAC targeting to the mitochondria. p62 was retained in the nucleus after LMB treatment, as expected. This suggests that in contrast to the rapid nuclear shuttling of p62, WAC is retained in the cytoplasm by GM130 and export of nuclear WAC is required to maintain the cytoplasmic WAC population.

17Finally, knockdown and rescue experiments showed that the GM130-interacting CC domain of WAC was required to rescue LC3B spot formation under starvation conditions (Figures 3F and 3G), implying that the WAC-GM130 interaction promotes autophagosome formation.

18Loss of GM130 affects cell growth and increases autophagy in tumor cells and in a lung cancer mouse model. Similarly, in HEK293 cells depletion of GM130 increased LC3-II lipidation in both fed and starved cells, increased WIPI2 spots and basal ULK1 activation, and caused a significant decrease of p62 levels (Figures 4A–4F and Figure S4A). Overexpression of HA-GM130 reduced LC3-II levels but had no effect on p62 levels or WIPI2 spots (Figures S4B–S4E). These data imply that GM130 functions as a negative regulator of autophagy.

19To probe the link between GM130 and autophagy, we investigated the localization of GM130 relative to autophagosome markers. We examined GABARAP, which localizes to the Golgi complex, and saw that GM130 in fed cells was distributed around an enlarged GABARAP structure, as well as occasionally colocalized with starvation-induced GABARAP puncta (Figure 4G).

20Do GM130 and GABARAP, or other Atg8 homologs interact? Immunoprecipitation of a panel of mammalian Atg8 homologs revealed that GM130 preferentially bound GFP-GABARAP, whereas less binding was seen with GFP-LC3 family members (Figure 4H). WAC also interacted with the Atg8s including GFP-GABARAP. Importantly, WAC and GM130 interact with endogenous GABARAP (Figure 4I). No direct GABARAP-WAC interaction could be detected (Figures S4F–S4H), however, GM130 and GABARAP interact directly, and increasing amounts of WAC disrupt this interaction (Figure 4J). Additionally, mitochondrially targeted GM130 recruits GFP-GABARAP (Figure 4K) and endogenous GABARAP to the mitochondria (Figure S4I).

21Based on the localization and shape of the enlarged GABARAP-positive structure, we asked if it was positive for γ-tubulin, a marker for the PCM of the centrosome. Using different GABARAP antibodies, we saw colocalization of GABARAP with γ-tubulin in HEK293A cells; we refer to this pool as centrosomal GABARAP (Figure 5A and Figure S5A). In five other cell lines, GABARAP and γ-tubulin also colocalized (Figure 5A). These GABARAP structures were not LC3B positive (see Figure 6C). Centrosomal GABARAP also localized around Centrin-3, a centriolar marker (Figure 5B). Furthermore, in HEK293 cells with an enlarged PCM, knockdown of GABARAP reduced centrosomal GABARAP (Figures 5C and 5D). Using a tetracycline-inducible cell line, we confirmed that GFP-GABARAP colocalized with γ-tubulin (Figure 5E). Centrosomal GABARAP is not an aggresome, as it is negative for ubiquitin and p62, even upon proteasome inhibition with MG132 (Figures S5B and S5C). Wortmannin, an inhibitor of starvation-induced GABARAP-positive autophagosomes, did not affect GABARAP localization at the centrosome (Figure 5F). Moreover, the GABARAP G116A mutant, which is not lipidated (Figure S5D), also localized to the centrosome (Figure 5G).

22We asked if centrosomal GABARAP was stably associated with the PCM. GABARAP associates with microtubules, so we disassembled microtubules with nocodazole (Figure 5H and Figures S5E and S5F) and found that centrosomal GABARAP was significantly (p < 0.0001) reduced (Figure 5I) and there was a concomitant increased colocalization of GABARAP with Golgi mini-stacks (Figure 5H and Figure S5F). As expected, GABARAP relocalized to nocodazole-induced Golgi mini-stacks during wortmannin treatment (Figure S5F). In addition, centrosomal GABARAP dissociates during metaphase (Figure S5G).

23After starvation, we saw a highly significant (p < 0.0001) decrease in centrosomal GABARAP signal and concurrent formation of GABARAP puncta (Figures 5F and 5J), suggesting GABARAP relocalizes from the centrosome to autophagosomes during starvation. In contrast to nocodazole treatment, dissembling the Golgi with Brefeldin A (BFA) (Figure S5H) did not diminish centrosomal GABARAP. Importantly, as BFA does not affect autophagic flux, GABARAP formed starvation-induced puncta even in the presence of BFA.

24We conclude that centrosomal GABARAP is dynamic and is affected by disruption of microtubules or starvation, and that its association with the PCM and the Golgi is not dependent on lipidation but mediated by protein interactions.

25GABARAPs are essential for the final expansion or closure of the autophagosome. Given this, we suggest that centrosomal GABARAP acts as a reservoir for expansion of forming autophagosomes. To probe the dynamics of centrosomal GABARAP, we attached a photoconvertible fusion protein, EosFP, to the N terminus of GABARAP. EosFP is converted from green to red fluorescence upon excitation with UV light. EosFP-GABARAP colocalized with γ-tubulin and WIPI2 in starved HEK293 cells (Figure S6A), and EosFP-GABARAP formed rings around γ-tubulin, which are visible in live cells (Figure 6A, C and Figures S6B and S6C).

26In starved live HEK293 cells, centrosomal green EosFP-GABARAP was photoconverted to red, and GABARAP-positive red spots formed while the intensity of red centrosomal EosFP-GABARAP decreased (Figure 6A, Figure S6B and Movie S1). Translocation of the photoconverted centrosomal EosFP-GABARAP to distal regions was monitored by quantifying fluorescence intensities (Figure 6B). Like autophagosomes, EosFP-GABARAP puncta were highly mobile and additionally made transient interactions with centrosomal EosFP-GABARAP (Movie S2). Strikingly, peripheral green EosFP-GABARAP spots (that existed prior to photoconversion) acquired centrosomally derived red photoconverted EosFP-GABARAP (Figure 6A and Figure S6B), implying that forming and expanding autophagosomes acquire GABARAP derived from the PCM. After time-lapse microscopy, the cells were fixed and stained with LC3 and γ-tubulin (Figure 6C and Figure S6C). Correlative confocal microscopy confirmed that the photoconverted pool of GABARAP was at the centrosome and that photoconverted centrosomal GABARAP had moved to peripheral puncta, many of which were LC3 positive (Figure 6C).

27In contrast to the centrosomal pool, cytosolic EosFP-GABARAP diffused within seconds across the cell (Figures 6D–6E and Figures S6D and S6E), indicating that the red GABARAP-puncta in Figures 6A and S6B are derived from a centrosomal pool and not from unintentional photoconversion of the cytosolic pool.

28Next, we treated live HEK293 cells expressing EosFP-GABARAP with nocadozole; EosFP-GABARAP was lost from the centrosome and dispersed into puncta (Figure 6F), which were not dynamic over a 20 min period after photoconversion. Correlative confocal microscopy revealed that similar to endogenous GABARAP (Figure 5H), EosFP-GABARAP was retained on Golgi mini-stacks (Figure 6G). Thus, without microtubules, centrosomal GABARAP relocalizes to the Golgi, becomes immobile and does not contribute to autophagosome formation.

29Since WAC and GM130 form a complex with GABARAP (Figures 4H–4J), we asked if WAC and GM130 controlled the localization of GABARAP to the PCM or forming autophagosomes. Centrosomal GABARAP co-localized with γ-tubulin in both fed and starved cells, and in starved cells GABARAP-positive autophagosomes appeared (Figures 6H and S6F). In starvation, WAC knockdown caused a striking accumulation of GABARAP on the Golgi and ERGIC and loss of centrosomal GABARAP (Figure 6H and Figures S6F and S6G). To investigate GABARAP dynamics in WAC-depleted cells, we used EosFP-GABARAP. As expected, EosFP-GABARAP accumulated on and around the Golgi (Figures 7A and 7B). Photoconversion and time-lapse microscopy showed that the pool of Golgi-accumulated EosFP-GABARAP is relatively immobile and does not form cytoplasmic GABARAP spots (Figures 7A and 7B, Movie S3). Importantly, in WAC knockdown cells there was an increase in the GM130-GABARAP interaction (Figures 7C and 7D). Thus, in the absence of WAC, GABARAP, likely non-lipidated, relocalizes from the centrosomal pool to the Golgi and ERGIC, is tethered by GM130, and does not contribute to autophagosome formation.

30GM130 regulates centrosome morphology and function, and after GM130 knockdown, we saw an increase in the size and intensity of the centrosomal GABARAP pool (Figure 6H and Figure S6F). These data suggest that WAC and GM130 control the localization of centrosomal GABARAP in a reciprocal manner and that in cells WAC negatively regulates GABARAP-GM130 binding.

31As WAC depletion reduced ATG13 phosphorylation (Figures 1B–1C and Figure S1D), we asked if GABARAP knockdown also attenuates ULK1 activation. ATG13 p-Ser318 levels were significantly decreased after depletion of GABARAP but not LC3B, GABARAPL1, or GATE-16 (Figures 7E–7H and Figure S7A). Centrosomal GABARAP is not lipidated, and non-lipidated GABARAP interacts constitutively with ULK1 (Figure 7I) in the cytosol and on membranes in a LIR-dependent manner (Figure 7J and Figures S7B and S7C). To define the membrane compartment harboring the ULK1-GABARAP complex, we immunoisolated DFCP-1-positive omegasomes and phagophores from starved cells. These DFCP-1 structures were positive for endogenous ULK1, WIPI2, GABARAP, and overexpressed wild-type ULK1, but not ΔLIR ULK1 (Figures S7D and S7E). GABARAP G116A was also found in similar amounts to GABARAP (Figure 7K), and immunoprecipitation of G116A revealed that it interacts with ULK1 on GFP-DFCP1 membranes (Figure 7L).

32Out of the Atg8 family, ULK complex preferentially binds GABARAP through a LIR motif. Overexpression of HA-ULK1 in fed HEK293 cells was sufficient to drive ATG13-FLAG phosphorylation even under basal conditions and when the ULK1 LIR was mutated (Figure 7M). However, in the presence of GFP-GABARAP and the unlipidated G116A mutant, but not GFP-LC3B, the ULK1 LIR was required for maximal phosphorylation of ATG13-FLAG (Figure 7M and Figure S7F). Thus, GABARAP specifically promotes ULK1 activity through the ULK1 LIR motif and independently of lipidation.

33In conclusion, we have shown a specific regulation of the centrosomal GABARAP pool by WAC and GM130 that correlates with initiation of autophagy, suggesting that centrosomally derived GABARAP signals back to autophagy initiation by sustaining ULK1 activity and explains why WAC, which inhibits GABARAP-GM130 interaction, also promotes ULK1 activation.

34WAC and GM130 are two regulators of autophagy whose mechanism of action is unresolved and previously unconnected. Here we show that WAC directly binds to GM130 and that WAC-GM130 binding is required for autophagosome formation. WAC functions by suppressing GM130 binding to GABARAP, allowing the maintenance of a centrosomal GABARAP reservoir. Centrosomal GABARAP and Golgi WAC coordinate and regulate autophagosome formation through ULK1 activation.

35Like GM130, WAC localizes to and co-purifies with the Golgi complex, remaining on the Golgi after 1 M KCl wash. We show that direct binding to GM130 tethers WAC to the Golgi complex and this is likely regulated by dephosphorylation of WAC during starvation. WAC-GM130 binding is independent of the RNF20/40 complex. The 10aa (aa611–620) of WAC are essential for GM130 binding and these 10aa are highly conserved back to D. melanogaster. WAC and GM130 form a stable complex: WAC remains bound to mitochondrially targeted GM130, even after 2 hr with LMB. GM130 has numerous functions, including vesicle tethering, signaling, and controlling mitosis.

36GM130 can cycle between the ERGIC and the cis-Golgi, which explains the localization of GABARAP and WAC to these compartments. The binding of GM130 to GABARAP is direct and not regulated by starvation, and our live-cell imaging data suggest that the Golgi pool of GABARAP does not participate in autophagy. The strong preference of GM130 for GABARAP, and not LC3B for example, indicates a level of specificity that remains to be investigated.

37In basal conditions, the Golgi acts as a hub for autophagy proteins such as BECLIN 1, GATE-16, ATG9, ATG16, GABARAP, and LC3, and Golgi-derived membranes are proposed to contribute to autophagosome biogenesis. Recently, contact sites between the Golgi and the phagophore were identified that could be sites of lipid transfer. Moreover, the Golgi complex is intimately linked with the centrosome, and GM130 controls centrosome morphology and function. Golgi proteins such as AKAP450 (which binds GM130), GMAP-210, and HOOK3 connect the Golgi ribbon to the microtubule network and centrosome. Interestingly, ATG4D, which processes GABARAPL1, is localized to the centrosome. Here, we show an unexpected localization of non-lipidated GABARAP at the PCM.

38We show that GM130 and WAC allow specific control of the centrosomal GABARAP pool. This pool of GABARAP is non-lipidated and how it associates with the PCM remains to be determined. Depletion of GM130 increases GABARAP residency on the centrosome and drives autophagy. In contrast, WAC knockdown results in loss of centrosomal GABARAP and relocalization to the ERGIC and Golgi due to increased GM130 binding. GABARAP directly binds microtubules and we speculate that microtubules may allow GABARAP transfer from the Golgi to the centrosome and from the centrosome to the autophagosome formation site. In the absence of microtubules, GABARAP relocalizes from the centrosome to Golgi ministacks, which are located at ER exit sites. However, the role of microtubules in these events remains to be clarified.

39The Atg8s are critical autophagy proteins. In yeast there is one Atg8 and no GM130 or WAC homologs. In mammals there are 7 Atg8 homologs, the LC3s, and GABARAPs. Although there is evidence of both redundancy and specialization of the LC3 and GABARAP family members, a comprehensive understanding of regulation and function of these proteins is lacking. In addition, many LC3- and GABARAP-interacting proteins have been identified, raising questions on how these interactions are controlled.

40Our work suggests that non-lipidated centrosomal GABARAP is transported to autophagosome formation sites where it localizes to autophagosomal membranes. Although the GABARAP subfamily promotes phagophore closure, it also binds the ULK complex. The ULK complex members ULK1/2, FIP200, and Atg13 bind GABARAP preferentially via LC3-interacting regions (LIRs). The identification of LIRs in the ULK complex suggests that the LC3 and GABARAP subfamilies act as a scaffold for recruitment of the ULK complex to autophagic structures. This is supported by our data that the ULK1 LIR motif is required for ULK1 retention on GFP-DFCP1-positive membranes. Importantly, we show that knockdown of GABARAP specifically attenuates ULK1 activation: LC3B, GABARAPL1, and GATE-16 do not activate ULK1. In contrast, in yeast the Atg1 kinase is delivered to the vacuole via Atg8 to act as a brake on autophagy, and the Atg8-Atg1 interaction does not modulate kinase activity. The ULK1 association with GABARAP is constitutive and not regulated by starvation or lipidation, thus, the large non-lipidated pool of GABARAP at the centrosome may bind and prime ULK1 activity. In support of this, ULK1 activation by GABARAP, but not LC3B, requires the LIR motif but does not require lipidation. Thus, LIR motifs can be regulatory elements, for example, LIR-containing proteins displace Atg12–5-16 from Atg8–PE. In the absence of WAC, aberrant GABARAP tethering to GM130 disrupts the autophagy-contributing centrosomal GABARAP reservoir and hence reduces GABARAP contribution to autophagosome formation and GABARAP-mediated ULK1 activation.

41It has been hypothesized that the LIR-dependent binding of the ULK complex to LC3 and GABARAP proteins functions in the formation and maturation of autophagosomes, however the mechanism is unknown. Supporting this idea, overexpression of an ULK1 LIR mutant results in the accumulation of WIPI2 puncta, suggesting autophagy has been stalled. Similarly, knockdown of the GABARAP family causes accumulation of ATG5 and ATG16L1 puncta, suggesting autophagy is inhibited downstream of ATG12–5-16L1 complex recruitment.

42Based on our data and existing literature, we propose a working model to explain the non-hierarchical (post-initiation) GABARAP-ULK1 function. Subsequent to activation of the BECLIN 1 complex by ULK1, a pool of PtdIns(3)P is formed at the omegasome. This PtdIns(3)P recruits WIPI2, which recruits the ATG12–5-16L1 complex. ATG16L1 can simultaneously bind WIPI2B and FIP200; FIP200 binding is downstream of WIPI2B and not required for LC3-lipidation. Thus, recruitment of FIP200 (and the ULK complex) to the phagophore by ATG16L1 is downstream of PtdIns(3)P-, WIPI2B-, and ATG12–5-16L1-driven lipidation of the LC3 and GABARAP family (see Figure S7G). We hypothesize that the constitutive GABARAP-ULK1 complex, derived from either the centrosome or the cytosol, would be recruited to the phagophore through ATG16L1 binding. The GABARAP-ULK1 complex may also be stabilized by lipidation. GABARAP association maintains ULK1 activation and substrate phosphorylation during the final stages of phagophore formation, until the ULK1 complex dissociates and the phagophore closes.

43Cell lines, transfection protocols, siRNA, DNA constructs, primers, and antibodies are detailed in the Supplemental Experimental Procedures.

44Whole-genome gene expression analysis was performed by the Genome Centre at Barts and the London School of Medicine and Dentistry, Queen Mary University of London (London, UK) using the Illumina human HT-12 v4 beadchip. Gene expression data were analyzed as described in the Supplemental Experimental Procedures.

45Protein complexes were pulled down from fed cells via a GFP tag using GFP-TRAP beads (ChromoTek) or by immunoprecipitation of WAC. Proteins were resolved by SDS-PAGE, fixed, and stained with GelCode, and full lanes were cut into slices for tryptic digestion and mass spectrometry analysis. For more details, see Supplemental Experimental Procedures.