Nutrient starvation elicits an acute autophagic response mediated by Ulk1 dephosphorylation and its subsequent dissociation from AMPK

Abstract

Macroautophagy (herein referred to as autophagy) is an evolutionarily conserved self-digestive process cells adapt to starvation and other stress responses. Upon starvation, autophagy is induced, providing cells with needed nutrient supplies. We report here that Unc-51-like kinase 1 (Ulk1), a key initiator for mammalian autophagy, undergoes dramatic dephosphorylation upon starvation, particularly at serine 638 and serine 758. Phosphorylations of Ulk1 are mediated by mammalian target-of-rapamycin (mTOR) kinase and adenosine monophosphate activated protein kinase (AMPK). AMPK interacts with Ulk1 in a nutrient-dependent manner. Proper phosphorylations on Ulk1 are crucial for Ulk1/AMPK association, as a single serine-to-alanine mutation (S758A) at Ulk1 impairs this interaction. Compared to the wild-type ULK1, this Ulk1-S758A mutant initiates starvation-induced autophagy faster at an early time point, but does not alter the maximum capacity of autophagy when starvation prolongs. This study therefore revealed previously unnoticed acute autophagy response to environmental changes.

Eukaryotic cells have evolved various signaling cascades and cellular processes in response to rapid environmental changes. Among these, macroautophagy (herein referred to as autophagy) is an evolutionarily conserved self-digestive process cells adapt to nutrient starvation (1, 2). Autophagy plays crucial roles in development, innate immune defense, protein quality control, tumor suppression, and cell death (3, 4). During autophagy, portions of cytoplasmic materials are engulfed into specialized double-membrane structures to form autophagosomes, which then fuse with lysosomes to degrade their cargos and regenerate nutrients (5, 6). This process is highly inducible and tightly regulated. Under normal growth conditions when nutrients are abundant, autophagy is kept at a basal level mainly for house-keeping purposes such as degradation of long-lived proteins and turn-over of damaged cellular organelles; under stress conditions like nutrient starvation, autophagy is further induced to provide cells with additional internal nutrient supplies. This induction is largely due to inhibition of TOR (target-of-rapamycin) complex 1 (TORC1), a kinase complex whose activity is regulated through integrating upstream PI3K (phosphatidylinositol 3-kinase)/AMPK (adenosine monophosphate activated protein kinase) activities and other nutrient-sensing signalings (1, 7, 8).

Initiation of autophagy has been extensively studied in budding yeast Saccharomyces cerevisiae. Atg1 kinase actively participates in cytoplasm-to-vacuole targeting pathway under nutrient-rich condition and switches to induce autophagy upon starvation by forming an autophagy-initiating complex with Atg13, Atg17, Atg2, and Atg31 (2). This event is composed of a signaling cascade including inhibition of yeast TORC1, dephosphorylation of Atg13, increased association between Atg13 and Atg1, and finally, increased Atg1 kinase activity to trigger downstream events. In higher eukaryotes including mammals, though similar to their yeast counterpart, detailed machinery and regulation involved in autophagy initiation have not been investigated until recently. Unc-51-like kinase 1 (Ulk1), as well as less-studied Ulk2, are two mammalian functional homologs of yeast Atg1 kinase. Same as yeast Atg1, Ulk1 kinase interacts with several autophagy-related partners, including mammalian Atg13 (mAtg13), Atg101, and FIP200/RB1CC1 (focal adhesion kinase family interacting protein of 200 kD, or retinoblastoma 1-inducible coiled-coil 1) (9–15).

Despite these similarities between yeast and mammalian autophagy-initiating complexes, significant differences, however, remain between these two systems. FIP200/RB1CC1 and Atg101 are not present in budding yeast S. cerevisiae. As for FIP200/RB1CC1, it may functionally overlap with both yeast Atg11 and yeast Atg17 (16). Most importantly, mAtg13 constantly associates with Ulk1 regardless of nutrient availability, which is different from the starvation-induced association between Atg1 and Atg13 in S. cerevisiae (10–12, 14). Because this dynamic interaction between Atg1 and Atg13 is a key regulatory step for yeast autophagy initiation and is lacked in Ulk1 complex, the mammalian autophagy needs to be initiated in a way that, at least partially, differs from that in yeast.

Functional outputs of autophagy are generally considered as accumulative and relatively slow processes. Variety of assays monitors autophagic readouts in wide time windows, normally hours after treatments that perturb cellular signalings. Yet, upstream regulations of autophagy are mainly through kinase cascades that are inherently prompt. It is therefore conceptually plausible that autophagy can be detected at earlier time points, especially when cultured cells are treated with harsh conditions such as total medium/serum withdrawal.

In this report, we address these issues by focusing on the phosphorylation status of Ulk1. Our data reveal Ulk1 is globally dephosphorylated upon starvation. 13 phosphorylation sites were mapped. AMPK, a key energy-surveillance kinase complex that is considered to act upstream of mTOR, directly associates with Ulk1 complex in a nutrient-dependent manner. Ulk1/AMPK association is determined by Ulk1 phosphorylation and consequently, by nutrient availability. This interaction may serve as a sequestering reservoir to confine Ulk1 during normal growth condition, and prepares cells to promptly initiate autophagy upon nutrient deprivation.

Results

Ulk1 Is Rapidly Dephosphorylated upon Starvation.

Recent studies show mammalian TORC1 (mTORC1), the upstream regulator of autophagy, may directly interact with Ulk1 complex under nutrient-rich (fed) conditions and possibly prevent autophagy initiation through inhibitory phosphorylations (10, 12, 14). More specific information including critical phosphorylation residues has not been provided and functional consequences not characterized. We used SILAC (stable isotope labeling with amino acids in cell culture) to quantitatively monitor phosphorylation changes of Ulk1 complex in response to nutrient availability. Unexpectedly, mAtg13 showed little change before and after starvation. Total phosphorylation levels of mAtg13 were low when cells were fed and stayed largely unaltered after starvation (Table S1). In contrast, Ulk1 was extensively phosphorylated at many residues under nutrient-rich condition and dramatically dephosphorylated upon medium withdrawal (Table 1). These results suggest that instead of mAtg13, Ulk1 could be the major regulatory component during mammalian autophagy initiation.

Table 1.

Quantification of the Ulk1 phosphorylated peptides reveals that phosphorylation of Ulk1 is globally decreased after starvation

Phospho-sites	*Phosphorylation ratio (fed/starved)	SD
Serine 341	1.61	0.29
Serine 405	0.59	0.06
Serine 450	3.28	0.05
Threonine 468	2.85	0.14
Serine 469	1.14	0.02
Serine 479	5.01	0.07
Serine 505	2.15	0.04
Serine 533	0.67	0.04
Serine 556	6.94	0.28
Serine 623	0.93	0.07
Serine 638	11.55	N/A
Serine 758	12.31	0.21
Serine 775	3.63	0.09

Thirteen phosphorylated residues were identified. Quantitative changes of phosphorylation level were recorded by SILAC. Serine 638 and serine 758 showed more than 10-fold decrease in phosphorylation level upon starvation. See also Fig. S3.

^*Calculated from ratio of identified phospho-peptide amount, and corrected to the amount of Ulk1 nonphosphorylated backbone peptides. SD, standard deviation.

In particular, SILAC experiment observed more than 10-fold decreases of phosphorylation at serine 638 and serine 758 of Ulk1, the most significant changes among all phosphorylation sites identified. Phospho-specific antibodies against these two sites were generated and the phosphorylation statuses of these two sites in response to nutrient availability were confirmed. Both serine 638 and serine 758 of Ulk1 were dephosphorylated after starvation as verified by Western blotting (Fig. S1A). Mutating these two residues to alanine (S638/758A) mimics dephosphorylation status, as phospho-specific antibodies could no longer pick up signal on mutant proteins. Mutating these two residues to aspartic acid (S638/758D) could not properly mimic phosphorylation status of these two sites; the S638/758D mutant behaved similarly to the S638/758A mutant as tested by the phospho-specific antibodies (Fig. S1B).

Phosphorylations of Ulk1 at Serine 638 and Serine 758 Are Differentially Regulated.

Time course experiments were performed to monitor dynamics of phosphorylation changes at serine 638 and serine 758 in response to nutrient availability. Serine 638 was found to respond faster than serine 758. As shown in Fig. 1A, 5 min after medium withdrawal, serine 638 was largely dephosphorylated. For serine 758, this process took 30 min. Similarly, when cells were replenished with medium, serine 638 was rapidly rephosphorylated in 5 min, while it took 30 min for serine 758 to be phosphorylated again.

Fig. 1.

Phosphorylations of Ulk1 at serine 638 and serine 758 were regulated differently. (A) Dynamics of Ulk1 phosphorylation. U2OS cells were starved in total starvation medium (HBSS solution containing 1% rich medium), and then replenished with rich medium (DMEM plus 10% dialyzed FBS) for durations as indicated. Cell extracts were analyzed by Western blotting to visualize phosphorylations of Ulk1 at serine 638 and serine 758. (B) U2OS cells were treated for 2 h as indicated. For serum withdrawal, cells were cultured in DMEM without FBS; for DMEM withdrawal, cells were cultured using 10% dialyzed FBS in HBSS. Phosphorylations of Ulk1 at Ser638 and Ser758 were analyzed using phospho-specific antibodies against these two sites. (C) Calcium is crucial for phosphorylation of Ulk1 at serine 638, but not serine 758. U2OS cells are treated for 2 h as indicated. Lane 3: calcium was added into total starvation medium to reach the same calcium concentration as in regular DMEM (0.2 g/L); Lane 4: regular DMEM medium (containing 10% dialyzed FBS) with single drop-out of calcium. See also Fig. S1.

Phosphorylations at serine 638 and serine 758 exhibited different kinetics, suggesting these two sites are regulated differently. We next investigated what are the specific nutritional triggers for phosphorylations at these two sites. As shown in Fig. 1B, DMEM (Dulbecco’s modified eagle medium) withdrawal remarkably decreases phosphorylation at serine 638, while serum withdrawal has a minor effect on this residue. Nevertheless, ULK1 protein was downshifted. In contrast, phosphorylation at serine 758 was severely affected by either serum or DMEM withdrawal.

Each component from DMEM was further screened for its role in impact ULK1 phosphorylation (for DMEM formula, see Table S2). Unexpectedly, among all components in DMEM, calcium showed the biggest effect on serine 638 phosphorylation. Simply adding calcium to total starvation medium largely restored phosphorylation of Ulk1 at this site (Fig. 1C, compare lane 3 with lane 2). Single calcium drop-out from DMEM led to dephosphorylation at serine 638, but not to the same extent as in total starvation medium. These results suggest calcium is critical for proper phosphorylation of Ulk1 at serine 638, yet there are other nutritional factors in DMEM which also contribute. In contrast, phosphorylation of Ulk1 at serine 758 is not influenced by calcium availability.

mTOR Mediates Phosphorylation at both Serine 638 and Serine 758, While AMPK only Responsible for Serine 638 Phosphorylation.

The fact that both serine 638 and serine 758 are regulated by multiple nutrients made us suspect the involvement of mTOR kinase, which can integrate various nutritional signals and has been reported to interact with Ulk1. Indeed, rapamycin treatment induced rapid dephosphorylation at serine 638 and serine 758. Both sites started to be dephosphorylated 10 min after treatment (Fig. 2A). Furthermore, knockdown of mTOR also led to dephosphorylation at these two sites (Fig. 2B). We conclude that mTOR signaling is required for phosphorylations at both serine 638 and serine 758 of Ulk1.

Fig. 2.

Phosphorylations of Ulk1 at serine 638 and serine 758 were mediated differently by mTOR and AMPK. (A) Rapamycin treatment induced dephosphorylation of Ulk1 at both serine 638 and serine 758. U2OS cells were cultured in rich medium and rapamycin was added (100 nM). Cells were then collected at given time points after rapamycin treatment. (B) Knockdown of mTOR induced dephosphorylation of Ukl1 at both serine 638 and serine 758. U2OS cells were transfected with either control siRNA (Luc) or mTOR siRNA oligos. 72 h after transfection, cells were collected and extracts analyzed by Western blotting. (C) Knockdown of AMPK induced dephosphorylation of Ulk1 at serine 638, but not serine 758. U2OS stable cell lines that inducibly knock down AMPKα (α1 and α2) or AMPKβ (β1 and β2) upon addition of doxycycline (Dox) were generated. 72 h after addition of Dox, phosphorylations of Ulk1 at Ser638 and Ser758 were visualized by Western blotting. (D) AMPK activity correlates with phosphorylation of Ulk1 at serine 638 but not serine 758 in response to various nutrient conditions. U2OS cells were treated for 2 h as indicated and cell extracts were analyzed by Western blotting.

Calcium signal only affects phosphorylation at serine 638, suggesting that this residue is also under regulation of other kinase(s) besides mTOR. We noticed that knockdown of either AMPKα1/α2 or AMPKβ1/β2 subunits led to dephosphorylation at serine 638 but not at serine 758 (Fig. 2C). Moreover, in any nutrient condition tested, there is a fine correlation between serine 638 phosphorylation and the kinase activity of AMPK as indicated by the phosphorylation at threonine 172 of AMPKα, and at serine 79 of acetyl-CoA carboxylase (ACC), a well studied AMPK substrate (Fig. 2D). Taken together, these results indicate that besides mTOR, AMPK signaling is also required for the phosphorylation of Ulk1 at serine 638.

Nutrient-Dependent Phosphorylation of Ulk1 Is Required for Ulk1/AMPK Interaction.

We next investigated whether phosphorylation statuses of Ulk1 influences its association pattern. Flag-Hemagglutinin (HA) tandem immunoprecipitation (IP) assays were performed in HeLa cells that stably express Flag-HA-Ulk1. The precipitants were subjected to silver staining followed by MS, which identified mAtg13, FIP200, and Atg101 to be constantly associated with Ulk1 (Fig. 3A, lane 1 and lane 2). Surprisingly, AMPK subunits (AMPKα and AMPKγ1) were identified to be associated with Ulk1 only under nutrient-rich condition and dissociated from Ulk1 after starvation. Another AMPK subunit AMPKβ did not stain with silver but was indentified there as well by Western blotting (see below). We also performed Flag-HA tandem IP in HeLa cells stably expressing Flag-HA-mAtg13. As expected, Ulk1, FIP200, and Atg101 were successfully pulled down; however, mAtg13 did not interact with AMPK subunit under either fed or starved condition (Fig. 3A). Similar association patterns were confirmed in HEK 293T cells: Ulk1 interacted strongly with all three AMPK subunits (AMPKα, AMPKβ, and AMPKγ), while mAtg13 did not associate with any subunit of AMPK complex (Fig. S2, compare A with B).

Fig. 3.

Ulk1 interacts with AMPK in response to nutrient availability. (A) Only Ulk1, but not mAtg13, associates with AMPK when cells are fed. In HeLa cells, Flag-HA-Ulk1, and Flag-HA-Atg13 stable lines were generated. Cells were treated for 2 h with either rich medium or total starvation medium. Flag-HA tandem IP assays were performed using cell extracts prepared from above treatments. Proteins in the final eluates were separated with 4–12% SDS-PAGE gel and were visualized by silver staining. Arrows indicates major coimmunoprecipitated proteins identified by MS. (B) Dynamics of Ulk1/AMPK interaction in response to nutrient availability. Cells were starved and repleted with rich medium for durations as indicated. Flag IP assays were then performed. Ulk1/AMPK association patterns were analyzed by Western blotting. (C) Phosphorylation of Ulk1 at serine 758 was required for Ulk1/AMPK association. HEK 293T cells were transfected with either wild-type Flag-Ulk1 or Flag-Ulk1 mutants as indicated, and treated for 2 h in rich or total starvation medium. Cell extracts were immunoprecipitated with anti-Flag beads, and the eluates were analyzed by Western blotting. (D) Phosphorylation of Ulk1 at serine 638 was not required for Ulk1/AMPK association. Experiments were performed the same as in (C) with either wild-type Ulk1 or indicated mutants. See also Fig. S2.

We then used SILAC to further confirm these observed changes in ULK1 complexes. Amount of Ulk1-associated AMPKα and AMPKγ1 decreased the most when medium was withdrawn (Table S3). Association between Ulk1 and mAtg13 only slightly increased. Associations of Ulk1 with FIP200 and Atg101 stayed largely unchanged.

Ulk1 did not interact with constitutively active form of AMPKα1 (aa1 to aa312) (Fig. S2C). Neither did Ulk1 interact with C-terminal part of AMPKα1 (aa312—C-terminal end), suggesting full-length AMPKα is required for its association with Ulk1 (Fig. S2 D and E). Kinase activity of AMPK seems not required for its interaction with Ulk1, as Ulk1 was able to interact with kinase-dead form of AMPKα in the same nutrient-dependent manner (Fig. S2F). As to Ulk1, kinase domain of Ulk1 (aa1—a278) is not required for its interaction with AMPK (Fig. S2G). Association between AMPK and the kinase-domain-deleted Ulk1 (aa279—C-terminal end) was still regulated by nutrient availability (Fig. S2G, compare lane 6 with lane 5).

Interestingly, AMPK started to dissociate from Ulk1 5 min after starvation, and completely reassociated with Ulk1 within 30 min after medium replenishment (Fig. 3B). This result is in accordance with the time course experiment for Ulk1 phosphorylation. A series of Ulk1 mutants were then generated to test whether Ulk1/AMPK dissociation upon medium withdrawal is due to dephosphorylation of Ulk1 at these sites. As shown in Fig. S3, mutations at most of residues did not alter Ulk1/AMPK interaction. However, phosphorylation of Ulk1 at serine 758, one of the most regulated sites as identified by SILAC, is critical for AMPK association. Single mutation at this site (Ulk1-S758A) impaired Ulk1/AMPK interaction, but the residue interaction was still regulated by starvation (Fig. 3C). In contrast, single mutation at serine 638 (Ulk1-S638A), the other most regulated site, did not alter Ulk1/AMPK interaction, but it helped further dissociate Ulk1/AMPK when combined with the S758A mutation (Fig. 3D).

Ulk1/AMPK Dissociation Primes Cells for Faster Response to Starvation-Induced Autophagy.

Both SILAC and Western blotting data show AMPK associates with Ulk1 but not mAtg13. Recent studies also reported that various AMPK subunits are able to interact with mammalian autophagy-initiating factors such as Ulk1/2, FIP200, and Atg101, but not mAtg13 (17). These observations raise a possibility that Ulk1 may exist in two mutually exclusive complexes: both will have core components such as Ulk1/2, FIP200 and Atg101, and either mAtg13 or AMPK bind to these core factors. Both may contribute to autophagy induction and it is therefore difficult to characterize the functional outputs from one particular complex in the presence of the other. Indeed, for autophagy level as tested by long-lived protein degradation (LLPD) assay, no significant difference was observed between wild-type Ulk1 and the S758A mutant defective in AMPK association (Fig. S4 D–F).

We tried to resolve this issue by knocking down mAtg13 and therefore eliminating the activity generated from the mAtg13 complex. Interestingly, absence of mAtg13 led to dramatic Ulk1 destabilization (Fig. S4 A and B; also see in ref. 12). We then used this cellular background (absence of both Ulk1 and mAtg13) to study Ulk1/AMPK interaction and performed rescue experiments with either wild-type Ulk1 or Ulk1-S758A mutant. Autophagy is dampened if protein level of mAtg13 and/or Ulk1 is decreased. Nonetheless, LLPD assay showed that in the absence of mAtg13, autophagy could still be greatly induced when cells were starved. In the first 10 min during starvation, U2OS cells expressing Ulk1-S758A mutant induced much more protein degradation than those expressing wild-type Ulk1 (Fig. 4A). Thirty minutes after starvation, the mutant still exhibited more than 20% excess of activity compared to the wild type (Fig. S4G). This difference in LLPD between wild-type Ulk1 and S758A mutant disappeared after prolonged starvation for 120 min (Fig. S4H). These results correlate with the dynamics between Ulk1 and AMPK interaction, as it takes about 30 min of starvation for wild-type Ulk1 to dissociate with AMPK. After that time point, wild-type Ulk1 should behave the same as the mutant in terms of AMPK-association pattern and consequent functional outputs.

Fig. 4.

Functional outputs of Ulk1 phosphorylations at serine 638 and serine 758. (A) LLPD assays showed Ulk1-S758A initiates starvation-induced autophagy faster compared to wild-type Ulk1 (*P = 0.003 in one-tailed Student’s t-test with equal variances, n = 3). Error bar represents SD. U2OS stable cell line that inducibly knock down mAtg13 upon addition of Dox was generated. Absence of mAtg13 also led to absence of Ulk1. 72 h after addition of Dox, cells were transfected with either wild-type Ulk1 or Ulk1-S758A construct, and cultured for another 48 h in rich medium containing no ³H-labeled but excessive cold leucine. Cells were than cultured in either rich or total starvation medium for 10 min. Percentages of ³H-labeled leucine released from cells into medium were calculated. (B) In either fed condition or 10 min after starvation, cells expressing Ulk1-S758A exhibited higher hVps34 activity compared to the wild type. Using same cells and same transfections as in (A), kinase activity of hVps34 was assayed by quantifying conversion from PI to PI3P in given time as described in Materials and Methods. Error bar represents SD. (C) Upper: LC3II formation in U2OS cells expressing either Ulk1-S758A mutant or wild-type Ulk1, in fed condition (0 min) and 30 or 120 min after starvation; lower: Quantitative analysis of the ratio of LC3II to LC3I (n = 3). (D) In Ulk1-S638A background, serine 758 could not be properly rephosphorylated 60 min after medium replenishment, while in wild-type Ulk1 background this process took less than 15 min. U2OS stable cell line that inducibly knock down Ulk1 upon addition of Dox was generated. 72 h after addition of Dox, cells were transfected with either wild-type Ulk1 or Ulk1-S638A construct, and cultured for another 48 h in rich medium to allow expression. Cells were then treated as indicated. (E) In S638A background, Ulk1/AMPK reassociation is slowed down. Using same cell line and transfections as in (D), cells were starved for 1 h and recovered for 20 min. Extracts were immunoprecipitated with anti-Flag beads, and the eluates were analyzed by Western blotting. See also Fig. S4.

The Class III PI3K, hVps34, phosphorylates the inositol ring of phosphatidylinositol (PI) at the D3 position to generate PI3P, a step essential for autophagosome formation. The hVps34 proteins from cells expressing Ulk1-S758A mutant has higher in vitro activity compared to those from cells expressing wild-type Ulk1 protein under fed condition (Fig. 4B). This result indicates that the Ulk1 mutant defective in AMPK association may better prime cells for autophagy induction with higher Class III PI3K activity. Upon starvation, the PI3K activity of hVps34 further increased, yet differences between the wild type and the mutant became less obvious as starvation prolonged (Table S4).

Microtubule associated protein light chain 3 (LC3) is the mammalian homologue of yeast Atg8 protein. During autophagy, cytoplasmic LC3 (LC3I) is translocated to autophagosomes, where LC3II is generated by proteolysis and lipidation at its C terminus. This conversion of LC3I to LC3II represents activation of autophagy. Thirty minutes after starvation, cells expressing Ulk1-S758A mutant induced more LC3II conversion as indicated by increased ratio of LC3II to LC3I, while the wild-type control stays largely unchanged (Fig. 4C). Prolonged starvation (120 min) abolishes this difference of LC3II/LC3I ratio between the wild type and the mutant.

Taken together, these data show Ulk1-S758A mutant defective in Ulk1/AMPK interaction initiates starvation-induced autophagy faster at an early time point, but does not change the maximum capacity.

Proper Phosphorylation at Serine 638 Facilitates Phosphorylation at Serine 758 and Proper Ulk1/AMPK-Association.

Alteration in serine 638, the other most regulated residue besides serine 758, does not affect apparent Ulk1/AMPK-association. Consequently, it is difficult to observe the functional impacts of serine 638 phosphorylation toward autophagy using LLPD or other assays. Instead, because serine 638 always responds to nutrient signals faster compared to serine 758 as shown in Fig. 1A, we asked whether phosphorylation at serine 638 facilitates such change at serine 758. The experiment was carried out in U2OS cells in which endogenous Ulk1 was stably knocked down. The cells were then rescued with either wild-type Ulk1 or Ulk1-S638A mutant. No obvious difference was observed between the wild type and the mutant in terms of dephosphorylation rate at serine 758 upon starvation (Fig. 4D). However, when cells were replenished with rich medium, rephosphorylation at serine 758 was much stronger in the wild-type background compared to that in the S638A mutation background. This result indicates proper phosphorylation at serine 638 promotes faster recovery of phosphorylation at serine 758 (Fig. 4D, compare lane 9 and 10 with lane 4 and 5). In accordance with these dynamics of phosphorylation, reassociation between Ulk1 and AMPK was also decreased in S638A mutation background (Fig. 4E, compare lane 6 with lane 5). Therefore, serine 638 may indirectly contribute to better regulation of autophagy by helping proper phosphorylation of serine 758 and Ulk1/AMPK interaction.

Discussion

We discovered Ulk1 undergoes dramatic dephosphorylation upon starvation, particularly at serine 638 and serine 758. Phosphorylation of Ulk1 is regulated by mTOR and AMPK, and is crucial for Ulk1/AMPK association. Ulk1 dissociates with AMPK when cells are deprived of nutrients. A single serine-to-alanine mutation (S758A) on Ulk1 impairs Ulk1/AMPK interaction. Upon starvation, this mutant Ulk1 induces autophagy much faster compared to the wild type.

Our report presented here reveals several previously unknown regulatory steps occurred on Ulk1 during starvation-induced autophagy. As shown schematically in Fig. 5, Ulk1 is hyper-phosphorylated at many serine/threonine residues including S638 and S758. Upon starvation, serine 638 is firstly dephosphorylated then serine 758 follows. Dephosphorylation at serine 758 leads Ulk1 to dissociate from AMPK and become more active in autophagy induction. When cells are replenished with nutrients, mTOR is reactivated and phosphorylates Ulk1 at multiple sites such as S638 and S758. Proper phosphorylation of Ulk1 then leads to Ulk1/AMPK association. Though kinase activity of AMPK is considered relatively low under nutrient-rich condition, when in close proximity, AMPK may help maintain phosphorylation of Ulk1 at serine 638 and strengthen its association with Ulk1.

Fig. 5.

Model of Ulk1 phosphorylation in response to nutrients. When cells are fed, Ulk1 is hyper-phosphorylated at serine 638 and serine 758. Upon starvation, serine 638 is firstly dephosphorylated; then followed by dephosphorylation at serine 758. Dephosphorylation of Ulk1 leads to dissociation of Ulk1/AMPK. When cells are replenished with rich medium, mTOR is activated; it phosphorylates serine 638 and serine 758. The phosphorylation of Ulk1 at serine 758 then leads to reassociation between Ulk1 and AMPK. When in close proximity, AMPK functions to maintain phosphorylation Ulk1 at serine 638.

AMPK has basal level activity when cells are fed and will be further activated upon amino acid starvation or glucose starvation (Fig. 2D, compare lane 6 and 7 with lane 1). Interestingly, calcium seems to be crucial for AMPK to properly exhibit its kinase activity. Basal AMPK activity will be further decreased if calcium is removed from rich medium (Fig. 2D, compare lane 5 with lane 1). This result explains why total medium withdrawal, the harshest starvation condition, leads to lower AMPK activity, because calcium was also removed under this circumstance (Fig. 2D, compare lane 2 with lane 1). The kinase activity of AMPK may also contribute to autophagy induction besides classical AMPK-TSC1/2-mTOR signaling, as suggested recently by J W Lee et al. that AMPK can lift the inhibitory effect of mTOR on Ulk1 by phosphorylating raptor, the key adaptor in mTORC1 (18). Taken together, we envision AMPK may have dual roles toward autophagy regulation: it promotes autophagy when cells are starved as previously well documented, while suppressing autophagy when cells are fed by forming a complex with and confining portions of Ulk1.

Functional readouts of Ulk1/AMPK interaction cannot be well detected in the presence of high “background noise” due to the existence of too much Ulk1/mAtg13 complex. We resolved this issue by knocking down mAtg13. Paradoxically, it seems the presence of mAtg13 facilitates efficient phosphorylations of Ulk1 (Fig. S4C). Therefore knockout (or knockdown) of mAtg13 may impair Ulk1/AMPK interaction due to less phosphorylation of Ulk1 at serine 758. This result leads to a situation where presence of mAtg13 brings high background noise, while absence of mAtg13 brings less Ulk1/AMPK interaction. In either way, it is difficult to pinpoint the actual contribution of Ulk1/AMPK interaction toward autophagy induction. Moreover, Ulk1/AMPK interaction is promptly regulated. The rapid kinetics makes most of qualitative assays unsuitable for monitoring autophagic effects caused by Ulk1/AMPK interaction. If nutrient deprivation prolongs, the differences between wild-type Ulk1 and AMPK-association-defective Ulk1 mutant (Ulk1-S758A) gradually disappears. Time windows for studying function of Ulk1/AMPK interaction is limited, only at the very early time point of a given environmental change. In our hands, it is less than 30 min after starvation. Fortunately, despite all these difficulties, the functional differences between wild-type and S758A mutant Ulk1 can be distinguished by using more quantitative approaches, such as LLPD assay and ELISA for hVps34 activity.

Though autophagy is in general conserved from yeast to human (particularly in later stages including membrane expansion, autophagosome formation, fusion with lysosome, and recycling), regulations of autophagy initiation in mammals are much more complicated and differ from yeast in many aspects. In this study, we try to elaborate this complexity by discovering a more rapid form of autophagy induction. For prolonged starvation, there is little difference between mutant Ulk1-S758A and the wild type, indicating interaction between Ulk1/AMPK mainly responds to acute nutritional changes. This acute mechanism not only functions as prompt response to nutrient deprivation, but may also be a crucial exit strategy to effectively down-regulate autophagy when cells leave harsh environments and need to resume proper growth/proliferation. In summary, with this layer of regulation, mammalian autophagy is capable of responding to environmental changes more rapidly than previously considered.

Materials and Methods

Mass Spectrometry Analysis.

The protein gel bands were digested in gel with sequencing grade trypsin (10 ng/μL trypsin, 50 mM ammonium bicarbonate, pH8.0) overnight at 37 °C. Peptides were extracted with 5% acetic acid/50% acetonitrile and 0.1% acetic acid/75% acetonitrile sequentially and then concentrated to ∼20 μL. The extracted peptides were separated by a homemade analytical capillary column (50 μm 10 cm) packed with C18 reverse phase material (YMC 5 μm spherical particles). An Agilent 1100 series binary pump was used to generate HPLC gradient as follows: 0%–5% B in 5 min, 5%–40% B in 25 min, and 40%–100% B in 15 min (A = 0.1 M acetic acid in water, B = 0.1 M acetic acid/80% methanol). The eluted peptides were sprayed into a QSTAR XL mass spectrometer (MDS SCIEX) equipped with a nano-ESI ion source. The mass spectrometer was operated in Information Dependent Acquisition mode. Ion spray voltage was 2.1 KV. The MS scan was from m/z 400 to 2,000. From each MS scan, top three most abundant peaks were selected for MS/MS (tandem mass spectrometry) fragmentation. Each scan was accumulated for 1 sec. The dynamic exclusion time was set as 20 sec.

Protein Database Search and Peptide Quantification.

The mass spectra were searched against IPI-human database on an in-house Mascot server (Version 2.2, Matrix Science Ltd.). Carbamidomethyl on cysteine was set as fixed modification. Variable modifications included oxidation on methionine, phosphorylation on serine, threonine, and tyrosine. Quantification mode was SILAC double labeled [13C6, 15N2]-lysine and [13C6, 15N4]-arginine. A maximum of three miscleavages was allowed for search. Mass tolerance was 0.2 Da for precursor ion, 0.25 Da for fragment ions. Mascot search results were imported to the open source software MSQuant (http://msquant.sourceforge.net) to calculate ratios of the heavy/light peptide pairs. Quantification results from each peptide were manually checked to ensure their correctness.

Long-Lived Protein Degradation Assay.

Detailed procedure has been described previously (19). In brief, U2OS cells were plated in 6-well dish with a density of 40,000 cells per well and let grow for 3 d. The growth medium was leucine-drop-out DMEM (USBiological) supplied by addition of 65 μM cold leucine and 1 μCi/mL ³H-labeled leucine (Perkin Elmer). In this assay, dialyzed FBS was used to eliminate leucine from other source. The cells then were washed with regular DMEM medium and let grow for another 48 h in regular DMEM medium containing 2 mM cold leucine. On the next day, the cells were washed and treated with either regular DMEM or total starvation medium for time durations as indicated (10 min, 30 min, or 120 min). After treatment, for the growth medium, we collected 1 mL from each well and added with 112 μL of 100% trichloroacetic acid (TCA) to reach 10% TCA concentration; we centrifuged the samples at 12,000× g for 2 min; and took 400 μL of supernatant to measure ³H readout with scintillation counter. For the cells, we first washed with DPBS, then 1 mL of 10% TCA was added to each well and we incubated the cells at room temperature for 5 min; after fixation, the cells were washed with 10% TCA and dissolved in 0.2 M NaOH; and we took 400 μL of the lysates to measure ³H readout. We calculated the total ³H readout in the medium and total ³H readout within the cells. Then we calculated the ratio of released ³H in the medium to the total ³H readout; this represents the degradation percentage in a given time duration.

Phosphatidylinositol 3-Phosphate (PI3P) Quantification (ELISA for hVps34 Activity).

For details, refer to the protocol for class III PI3-Kinase kit (96-well ELISA assay for detection of PI3P) by Echelon Inc. In brief, endogenous hVps34 proteins were immunoprecipitated and incubated with PI substrate for 2 h to catalyze the synthesis of PI3P. The reaction (containing synthesized PI3P product) was then added to 96-well coated with PI3P. Detector proteins that recognize PI3P were also added to the wells. In the next 2 h, the PI3P immobilized on the 96-well plate competed with free PI3P in the reaction for detector binding. The amount of detector proteins bound to the plate is determined through colorimetric detection at 450 nm absorbance. The reading was inversely proportional to the amount of PI3P produced by hVps34.