Snf1/AMPK fine-tunes TORC1 signaling in response to glucose starvation

  1. Marco Caligaris
  2. Raffaele Nicastro
  3. Zehan Hu
  4. Farida Tripodi
  5. Johannes Erwin Hummel
  6. Benjamin Pillet
  7. Marie-Anne Deprez
  8. Joris Winderickx
  9. Sabine Rospert
  10. Paola Coccetti
  11. Jörn Dengjel
  12. Claudio De Virgilio

  • Curated by eLife

    eLife logo

    eLife assessment

    This is an important study that shows how Snf1/AMP Kinase fine-tunes TORC1 signaling in response to glucose starvation. Their observation that Snf1 phosphorylation of the TORC regulator Pib1 and the TORC effector kinase Sch9 provides new mechanistic information on this important pathway involved in cell growth. The combination of phosphoproteomics, genetic, biochemical, and physiological experiments is generally convincing, although the results with the Pib2 SA and SE mutants are somewhat inconsistent.

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

The AMP-activated protein kinase (AMPK) and the target of rapamycin complex 1 (TORC1) are central kinase modules of two opposing signaling pathways that control eukaryotic cell growth and metabolism in response to the availability of energy and nutrients. Accordingly, energy depletion activates AMPK to inhibit growth, while nutrients and high energy levels activate TORC1 to promote growth. Both in mammals and lower eukaryotes such as yeast, the AMPK and TORC1 pathways are wired to each other at different levels, which ensures homeostatic control of growth and metabolism. In this context, a previous study (Hughes Hallett et al., 2015) reported that AMPK in yeast, that is Snf1, prevents the transient TORC1 reactivation during the early phase following acute glucose starvation, but the underlying mechanism has remained elusive. Using a combination of unbiased mass spectrometry (MS)-based phosphoproteomics, genetic, biochemical, and physiological experiments, we show here that Snf1 temporally maintains TORC1 inactive in glucose-starved cells primarily through the TORC1-regulatory protein Pib2. Our data, therefore, extend the function of Pib2 to a hub that integrates both glucose and, as reported earlier, glutamine signals to control TORC1. We further demonstrate that Snf1 phosphorylates the TORC1 effector kinase Sch9 within its N-terminal region and thereby antagonizes the phosphorylation of a C-terminal TORC1-target residue within Sch9 itself that is critical for its activity. The consequences of Snf1-mediated phosphorylation of Pib2 and Sch9 are physiologically additive and sufficient to explain the role of Snf1 in short-term inhibition of TORC1 in acutely glucose-starved cells.

Article activity feed

  1. Version published to 10.7554/elife.84319 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Public Review):

    1. The authors show that there are several classes of Snf1 targets (Fig. 3e), most notably some that are phosphorylated immediately after Snf1 activation by glucose (<5 min) and others that are only phosphorylated after 15 min. In a simple view, all direct Snf1 targets should be phosphorylated immediately after Snf1 activation. Is that the case? What is the overlap between the direct targets found using the OBIKA assay and the slow and fast responding in vivo targets? What about the phosphorylation motif, does it differ between the groups? These points are not discussed in the text except to point out that the direct Snf1 target Msn4 is among the slowly phosphorylated group.

    This is a very good point and we have performed the suggested analysis, which resulted in an interesting finding that we describe now in the text as follows:

    “Notably, of the 145 confirmed target sites, 81 (i.e. 72%) were significantly regulated after both 5 min and 15 min. Of the remaining 64 sites, 32 responded only after 5 min, while the other 32 responded only after 15 min. Some of the former residues are located within Snf1 itself, the -subunit of the Snf1 complex (i.e. Sip1), the Snf1-targeting kinase Sak1, or Mig1, while some of the latter are located within the known Snf1-interacting proteins such as Gln3, Msn4, and Reg1. These observations indicate that Snf1-dependent phosphorylation initiates, as expected, within the Snf1 complex and then progresses to other effectors. Interestingly, based on the residues that responded exclusively after 5 min, we retrieved a perfect Snf1 consensus motif (i.e. an arginine residue in the -3 position and a leucine residue in the +4 position; Supplementary figure 2A). The one retrieved for the residues that respond exclusively at 15 min, in contrast, significantly deviated from this consensus motif (Supplementary figure 2B). The slight temporal deferral of Snf1 target phosphorylation may therefore perhaps in part be explained by reduced substrate affinity due to consensus motif divergence.”

    1. The data showing that Snf1-dependent phosphorylation of Pib2 plays a key role in triggering inhibition of TORC1 is convincing but is entirely dependent on a rescue of the TORC1 inhibition defect seen in cells where Snf1 is inhibited. That is, TORC1 is normally inactivated during glucose starvation; this does not occur when Snf1 is inhibited by 2nm-pp1 but does occur when Snf1 is inhibited in a strain carrying a phosphomimetic version of Pib2 (Pib2SESE). This indicates that Pib2 phosphorylation is sufficient to replace Snf1 signaling and inhibit TORC1 during glucose starvation. However, in a simple model, a phosphodead version of Pib2 (SASA) should have the opposite effect. That is TORC1 should remain active during glucose starvation in the Pib2SASA strain-but that is not the case (Fig. 4g). This point is not discussed in the paper; why do the authors think that TORC1 is inhibited normally in the SASA mutant inhibits TORC1 normally?

    We fully agree with this statement and have highlighted and discussed this issue now in the last paragraph of the results section (where we think this fits best) as follows:

    “In contrast, the separated and combined expression of Sch9S288A and Pib2S268A,S309A showed, as predicted, no significant effect in the same experiment. Unexpectedly, however, the latter combination did not result in transient reactivation of TORC1, like we observed in glucose-starved, Snf1-compromised cells. This may be explained if TORC1 reactivation would rely on specific biophysical properties of the non-phosphorylated serines within Sch9 and Pib2 that may not be mimicked by respective serine-to-alanine substitutions. Alternatively, Snf1 may employ additional parallel mechanisms (perhaps through phosphorylation of Tco89, Kog1, and/or other factors; see above) to prevent TORC1 reactivation even when Pib2 and Sch9 cannot be appropriately phosphorylated. While such models warrant future studies, our current data still suggest that Snf1-mediated phosphorylation of Pib2 and Sch9 may be both additive and together sufficient to appropriately maintain TORC1 inactive in glucose-starved cells”

    Reviewer #2 (Public Review):

    1. Because PIB2 is a major focus of the manuscript, I was surprised that it was not discussed in the introduction. I think it would be appropriate to discuss prior evidence linking this protein to TORC1.

    We thank the reviewer for this suggestion. Pib2 and its role in TORC1 control is now described in the introduction.

    1. The authors introduce mutations into PIB2 at two sites determined to be phosphorylated by SNF1, at S268 and S309. Somewhat confusing results are obtained, in that the PIB2 null and phosphomimic mutants (S268E and S309E) confer a similar TORC1 phenotype, compared to the S268A S308A mutant. These results require further explanation than simply that "TORC1 inactivation defect in SNF1-compromised cells is due to a defect in PIB1 phosphorylation". This is particularly intriguing given that the opposite results are observed with the SCH9 mutants, where the null and alanine mutants confer a similar phenotype compared to the S to E mutants.

    The finding that both loss of Pib2 and expression of the phosphomimetic allele yield the same phenotype is indeed counterintuitive. Hence, we fully agree with the criticism put forward here. We believe that the underlying reason for our observation is based on the unique property of Pib2 in having both a C-terminal TORC1-activating domain (CAD) and an-N-terminal TORC1-inhibitory domain (NID). We have addressed this point briefly in the discussion ("Our current data favor a model according to which Snf1-mediated phosphorylation of the Kog1-binding domain in Pib2 weakens its affinity to Kog1 and thereby reduces the TORC1-activating influence of Pib2 that is mediated by the C-terminal TORC1-activating (CAD) domain via a mechanism that is still largely elusive"), but now also address this issue in the results section as suggested.

    1. The authors conclude, based on the co-IP data in Figure 4H, that interactions between KOG1 and PIB2 are direct. However, it remains possible that interactions between these proteins are mediated by other components of TORC1 or within cells. This should be addressed.

    Please note that the Kog1-Pib2 interaction has previously been demonstrated by different methods. Accordingly, Pib2 has not only been shown to interact with Kog1 (or TORC1) in co-IP studies in vivo (PMID: 30485160, PMID: 29698392), but also by co-IP studies in vitro (PMID: 29698392, PMID: 28483912, PMID: 34535752). In addition, the interaction between Kog1-Pib2 has also been dissected (down to defined domains) by classical two hybrid analyses (PMID: 28481201). All of these studies are cited now in the introduction where Pib2 is discussed.

    1. The authors demonstrate convincingly that the PIB2 and SCH9 SNF1-specific phospho-site mutants have a detectable effect on TORC1, primarily by examining TORC1-dependent phosphorylation of SCH9. What is unclear is whether phosphorylation at these sites has a significant physiological impact on cells. It appears that the rapamycin hyper-sensitivity displayed in Figure 6E is the only data presented to address this question. It would be appropriate for the authors to comment further on the significance of SNF1-dependent phosphorylation of these two substrates.

    To further address the physiological role of the Snf1-dependent phosphorylation of Sch9 and Pib2 combined, we newly assessed the growth rate of the strain that expresses the Sch9SE and Pib2SESE alleles combined. Accordingly, we found the snf1as pib2SESE sch9SE strain to exhibit a significantly higher doubling time than the snf1as strain on both low-nitrogen-containing media and standard synthetic complete media. This is now included in the text (results section).

    Reviewer #3 (Public Review):

    1. Conceptually, the manuscript shows that Snf1 activity is important for the acute inhibition of TORC1 during glucose starvation. However, this is mainly restricted to 10 and 15 minutes of glucose starvation. After 20 minutes, TORC1 is inhibited by some unknown mechanisms independent of Snf1 (Hughes Hallet et al). This raises concern regarding the physiological relevance of Snf1-mediated TORC1 inhibition during acute glucose stress. The authors show that this regulation is important for the survival of cells under TORC1 inhibition. How do the authors envision that the acute role of Snf1 plays an important long-term physiological relevance during rapamycin treatment? Providing more support for the physiological relevance of this regulation will make this study of interest to a broad readership.

    Please see our response to point 4 of reviewer #2.

    1. Another major concern of the manuscript is the inconsistencies between the various representative immunoblots and their quantifications. The effect of AMPK activity on TORC1 signaling under glucose starvation seems very subtle. A few specific concerns are mentioned below:

    a) In figure 1A, the increase in TORC1 activity upon inhibition of analogue sensitive Snf1as by 2NM-PP1 is very marginal. Although quantification shows a significant increase, a representative western blot figure should be shown.

    We have replaced the original immunoblots with more representative ones in Figure 1A.

    b) Does deleting Snf1 itself have any effect on TORC1 activity? Lane 4 of figure 1A shows reduced activity compared to lane 1.

    TORC1 activity is generally assessed as the ratio between phosphorylated Sch9 and total Sch9 (see also below under (e)). Accordingly, based on the quantification of 6 blots (we added two more experiments to address this point; Figure 1B), loss of Snf1 has no significant impact on TORC1 activity in exponentially growing cells, as we expected.

    c) To show the effect of Snf1 on the repression of TORC1, the time-course experiments are run on two separate gels in figure 1C. Hence, it is difficult to compare the effect of Snf1 on unscheduled reactivation of TORC1 under glucose starvation.

    Please note that the data of the two blots were cross-normalized to the sample from exponentially growing cells (labeled “Exp”; i.e. the same sample was loaded on the two blots) in order to compare and quantify the effects of Snf1.

    d) In figure 1E, the effect of Reg1 deletion on TORC1 activity seems minor as both phospho- and total levels of Sch9 are reduced.

    As correctly pointed out by this reviewer, we consistently found the total Sch9 levels to be lower in reg1Δ cells when compared to wild-type cells. To assess TORC1 activity, we therefore always determine the ratio between phosphorylated Sch9 and total Sch9, and the respective ratio is significantly different in reg1∆ cells when compared to wild-type cells. We speculate that the reduced Sch9 levels in this mutant are caused by the reduced growth rate (PMID: 22140226) and hence lower protein synthesis rate (to which translation of SCH9 mRNA may be specifically sensitive).

    Since further mechanistic insights are based on these initial findings of figure 1, solidifying these observations is very important.

    1. In figure S1, the analogue sensitive Snf1as shows significant reduction in its activity (reduced S79 phosphorylation of ACC1-GFP). This raises the concern of whether this genetic background is an ideal system to resolve the mechanism of TORC1 suppression.

    The Snf1as allele is indeed hypomorphic, which we acknowledge appropriately in the text. We would like to point out however, that we took great care in each experiment to include the DMSO control that allowed us to unequivocally assign any observed effects to the specific drug-mediated inhibition of Snf1as. Importantly, we think that the hypomorphic nature of the Snf1as allele (which allows normal growth on non-fermentable carbon sources) represents a minor trade-off when compared to the advantages that this allele provides over the use of a snf1∆ strain, which exhibits a fundamentally reprogrammed transcriptome/proteome (PMID: 17981722). Accordingly, this allele allows the assessment of Snf1 inhibition on very short time scales while minimizing confounding large-scale proteome rearrangements that may indirectly affect the studies. Moreover, use of the Snf1as allele also allowed us to compare our results more directly with other phosphoproteome studies that used the same allele (PMID: 25005228, PMID: 28265048). Finally, please also note that our main conclusions (on Snf1-mediated control of TORC1) are corroborated by additional genetic data such as the ones in Figure 1A/E where we use snf1∆ and reg1∆ cells.

    1. In figure 2, during glucose restimulation, there is increased retention of Snf1as-pThr210 in the presence of 2NM-PP1. This suggests that the upstream glucose sensing pathway as well as Snf1 might be more active than in DMSO-treated cells. This also raises concerns regarding the suitability of the genetic background for the study. Can authors comment on why this phosphorylation persists? Does the phosphoproteomic analysis give any hint for this phenotype?

    This is a very good point. In fact, we forgot to mention in the text that the observed effect of the 2NM-PP1 treatment on Snf1-Thr210 phosphorylation has already been studied and mechanistically explained earlier (PMID: 23184934). Accordingly, the entry of the drug into the broader catalytic cleft of the Snf1as mutant causes the catalytic domain to be stabilized in a conformation, which prevents dephosphorylation of pThr210 by the dedicated Glc7-Reg1 phosphatase heterodimer. This can be observed each time when we compared 2NM-PP1- and DMSO-treated cells and probed for Snf1-Thr210 phosphorylation. This is, in fact, an independent control for proper 2NM-PP1 functioning. We have now added a sentence (including reference) that pinpoints this issue in the text.

    1. In figure 4H, where authors claim reduced binding of Kog1 to Pib2SESE, levels of Kog1 in input are also reduced. Can authors provide further support using colocalization studies? Also, does Pib2SESE has any defect in forming Kog1 bodies?

    We took great care to load equal amounts of IPed Pib2-myc variants and then normalized the co-IPed Kog1-HA on the IPed Pib2-myc variant levels. The Kog1-HA input levels vary a bit between the 4 experiments, but they are on average not significantly lower in Pib2SESE-myc-expressing cells when compared to WT cells. In addition, in our Co-IP experiments, the beads are saturated with Pib2-myc variants and Kog1-HA levels are generally not limiting. We therefore deem it fair to say that the Pib2SESE has a reduced affinity for Kog1. Based on our experience with other co-localization studies of membrane-bound proteins and protein complexes (e.g. TORC1 versus EGOC), we find it extremely difficult to quantify local interactions by fluorescence microscopy (unless they are close to all or nothing). In this case, where we have a partial defect in the interaction between Kog1 and Pib2SESE, we anticipate that such analyses will not allow us to draw additional conclusions.

    Regarding the issue of Kog1/TORC1-body formation: all of our mutations in PIB2 and SCH9 were introduced (by CRISPR-Cas9) in the genome of our snf1as strain, which was used throughout this study. To analyze Kog1/TORC1-bodies, we have therefore first tried to C-terminally tag KOG1 with GFP in the genome of our strain background (similarly as was done in the original description of Kog1 bodies; PMID: 26439012). However, because all our attempts failed to create KOG1-GFP in our strain, we assumed that this construct may be lethal in our strain background. This is not completely unexpected, as it is known that the Kog1-GFP allele is hypomorphic and temperature sensitive (PMID: 19144819). In an alternative approach, we have therefore set out to study TORC1 body formation in our strains by using a GFP-TOR1 allele that can be integrated into the genome and that expresses functional TORC1 (PMID: 25046117). As we have described earlier, the respective GFP-Tor1 construct localized on vacuolar membranes and on foci that we previously have shown to correspond to signaling endosomes (PMID: PMID: 30732525, 30527664). Unexpectedly, however, when we starved the respective cells for glucose, the number of GFP-Tor1 foci did only marginally increase (20%) in our strain background over a period of up to 1 hour. Given these various unexpected issues, we prefer to not include any of these preliminary data in the current version of our manuscript, but to rather follow up on these observations in a separate study. We deem this particularly justified as the current literature on TORC1-body and TOROID formation also appears controversial and may need further clarification. For instance, while TORC1-body formation has been suggested to represent a Snf1-dependent process that is dispensable for TORC1 inhibition (PMID: 30485160), TOROID formation has been suggested to represent a Snf1-independent process that is mechanistically linked to TORC1 inhibition (PMID: 28976958).

    1. In figure 5F, where the authors claim the Sch9SE mutant has lower TORC1 activity, the difference is very minor. Furthermore, corresponding lanes also show reduced levels of Snf1as expression. Hence, improved blots are required here. Also, an in vitro kinase assay with full-length Sch9 KD with and without the Ser288 mutation could solidify the observation that phosphorylation of Ser288 indeed affects TORC1-mediated phosphorylation.

    We have replaced the blots in Figure 5F with an alternative set that more clearly highlights the (statistically significant) differences, while also exhibiting more equal levels of Snf1as levels. Regarding the in vitro kinase assays: we have repeatedly tried to perform TORC1 kinase assays on full length Sch9KD without success. We currently believe that proper TORC1-mediated phosphorylation of Sch9 may have to occur on membranes to which both TORC1 and Sch9 are tethered through phospholipid interactions (PMID: 29237820). We are trying to set up such a system on liposomes, but we assume that this will be a major effort that cannot be resolved in due time.

    1. In figure 6E, the Sch9SE mutant shows no effect in the presence of rapamycin. Thus, in vivo, phosphorylation at Ser288 may not be perturbing the phosphorylation of Sch9 by TORC1.

    When cells are grown on glucose where TORC1 is highly active (as in Fig. 6E or 6A/B in Exp), expression of Sch9SE has no significant effect indeed. However, in glucose-starved cells, where TORC1 activity is low, expression of the Sch9S288E allele clearly and significantly contributes to inhibition of Sch9-Thr737 phosphorylation by TORC1 (Figure 6A/B and Figure 5F/G).

    1. According to the author's proposed mechanism, TORC1 activity in Pib2SASA or Pib2SASA/Sch9SA backgrounds should be higher during glucose starvation compared to the control strains. However, glucose starvation shows a similar level of reduction in TORC1 activity in these backgrounds. This raises concern regarding the proposed mechanism. The authors mainly base their conclusions on Ser to Glutamate mutants. The authors should be cautious that Ser to Glutamate changes may also affect the protein structure which can confer similar phenotypes. How do the authors justify this discrepancy?

    Please see our response to point 2 of reviewer #1.

  3. eLife

    eLife assessment

    This is an important study that shows how Snf1/AMP Kinase fine-tunes TORC1 signaling in response to glucose starvation. Their observation that Snf1 phosphorylation of the TORC regulator Pib1 and the TORC effector kinase Sch9 provides new mechanistic information on this important pathway involved in cell growth. The combination of phosphoproteomics, genetic, biochemical, and physiological experiments is generally convincing, although the results with the Pib2 SA and SE mutants are somewhat inconsistent.

  4. eLife

    Reviewer #1 (Public Review):

    Caligaris et al set out to uncover the mechanism(s) underlying AMPK/Snf1-dependent regulation of TORC1 during glucose starvation. As a first step, they map the global phosphorylation changes that occur when cells are shifted from high glucose to low glucose concentrations (to activate Snf1) both in the presence and absence of the Snf1 inhibitor 2NM-pp1. They then follow this up by carrying out an on-bead in vitro kinase assay (OBIKA) to identify direct Snf1 targets. Together these very high-quality data lead to the identification of nearly 1300 Snf1-dependent phosphorylation sites and at least 150 direct Snf1 target sites. The target proteins are involved in a variety of processes--such as transcription, ribosome biogenesis, signaling, and vesicle trafficking-and will serve as an important resource for those studying Snf1/AMPK-dependent regulation of a wide variety of cellular programs.

    Among the Snf1 targets the authors identify several proteins involved in TORC1 signaling and follow up on two new connections; Snf1-dependent phosphorylation of (i) the recently identify TORC1 regulator Pib2, and (ii) the key TORC1 substrate Sch9. Using a set of rigorous experiments, they confirm that Pib2 is a direct target of Snf1, and then show that cells carrying a phosphomimic version of Pib2 rescue the TORC1 inhibition defect seen in cells with no Snf1 activity. They then carry out a very similar set of experiments with Sch9 and show that Snf1-dependent phosphorylation of Sch9 is also important for the inhibition of TORC1 signaling during glucose starvation, and furthermore that impact of the Snf1-dependent phosphorylation of Pib2 and Sch9 is additive.

    This paper provides some of the first mechanistic insights into the way that glucose starvation triggers inhibition of TORC1 (particularly in yeast) and will serve as an important resource for those interested in AMPK/Snf1-dependent regulation of a variety of other pathways and processes. The paper also provides the clearest picture yet of the regulation of Pib2, an important but poorly understood TORC1 regulator in yeast and likely beyond.

    This important and rigorous work is very well presented but there are two areas that could do with further discussion and clarification.

    (1) The authors show that there are several classes of Snf1 targets (Fig. 3e), most notably some that are phosphorylated immediately after Snf1 activation by glucose (<5 min) and others that are only phosphorylated after 15 min. In a simple view, all direct Snf1 targets should be phosphorylated immediately after Snf1 activation. Is that the case? What is the overlap between the direct targets found using the OBIKA assay and the slow and fast responding in vivo targets? What about the phosphorylation motif, does it differ between the groups? These points are not discussed in the text except to point out that the direct Snf1 target Msn4 is among the slowly phosphorylated group.

    (2) The data showing that Snf1-dependent phosphorylation of Pib2 plays a key role in triggering inhibition of TORC1 is convincing but is entirely dependent on a rescue of the TORC1 inhibition defect seen in cells where Snf1 is inhibited. That is, TORC1 is normally inactivated during glucose starvation; this does not occur when Snf1 is inhibited by 2nm-pp1 but does occur when Snf1 is inhibited in a strain carrying a phosphomimetic version of Pib2 (Pib2SESE). This indicates that Pib2 phosphorylation is sufficient to replace Snf1 signaling and inhibit TORC1 during glucose starvation. However, in a simple model, a phosphodead version of Pib2 (SASA) should have the opposite effect. That is TORC1 should remain active during glucose starvation in the Pib2SASA strain-but that is not the case (Fig. 4g). This point is not discussed in the paper; why do the authors think that TORC1 is inhibited normally in the SASA mutant inhibits TORC1 normally?

  5. eLife

    Reviewer #2 (Public Review):

    This study by De Virgilio and co-workers examines the role of SNF1, the AMPK ortholog in yeast, in regulating TORC1 activity and cell growth. The authors combine the use of an analog-sensitive allele of SNF1 and SILAC to characterize the yeast phospho-proteome under nutrient-complete and glucose-starved conditions. Using this approach, in addition to confirming previously identified targets of SNF1, many potential new substrates were identified. The authors follow up on two of these: PIB2, a known upstream regulator of TORC1 linked previously to glutamine signaling; and SCH9, an important downstream target that is directly phosphorylated by TORC1. The authors use a combination of mutagenesis and in vitro and in vivo assays to demonstrate that SNF1 is a bona fide kinase for PIB2 and SCH9 and that phosphorylation of these targets impacts TORC1 kinase activity. In general, the data are thorough and convincing, and these findings will be appreciated by the wide readership of this journal.

    Specific points to consider:

    1. Because PIB2 is a major focus of the manuscript, I was surprised that it was not discussed in the introduction. I think it would be appropriate to discuss prior evidence linking this protein to TORC1.

    2. The authors introduce mutations into PIB2 at two sites determined to be phosphorylated by SNF1, at S268 and S309. Somewhat confusing results are obtained, in that the PIB2 null and phosphomimic mutants (S268E and S309E) confer a similar TORC1 phenotype, compared to the S268A S308A mutant. These results require further explanation than simply that "TORC1 inactivation defect in SNF1-compromised cells is due to a defect in PIB1 phosphorylation". This is particularly intriguing given that the opposite results are observed with the SCH9 mutants, where the null and alanine mutants confer a similar phenotype compared to the S to E mutants.

    3. The authors conclude, based on the co-IP data in Figure 4H, that interactions between KOG1 and PIB2 are direct. However, it remains possible that interactions between these proteins are mediated by other components of TORC1 or within cells. This should be addressed.

    4. The authors demonstrate convincingly that the PIB2 and SCH9 SNF1-specific phospho-site mutants have a detectable effect on TORC1, primarily by examining TORC1-dependent phosphorylation of SCH9. What is unclear is whether phosphorylation at these sites has a significant physiological impact on cells. It appears that the rapamycin hyper-sensitivity displayed in Figure 6E is the only data presented to address this question. It would be appropriate for the authors to comment further on the significance of SNF1-dependent phosphorylation of these two substrates.

  6. eLife

    Reviewer #3 (Public Review):

    Caligaris and colleagues show a new mechanism by which AMPK/Snf1 inhibits TORC1 signaling during glucose starvation. They propose that under glucose starvation, Snf1 inhibits the unscheduled activation of TORC1 by phosphorylating Pib2 and Sch9, upstream and downstream effectors of TORC1. The study also provides a resource for novel substrates of Snf1 which can be useful for future studies. Specific comments are below.

    1. Conceptually, the manuscript shows that Snf1 activity is important for the acute inhibition of TORC1 during glucose starvation. However, this is mainly restricted to 10 and 15 minutes of glucose starvation. After 20 minutes, TORC1 is inhibited by some unknown mechanisms independent of Snf1 (Hughes Hallet et al). This raises concern regarding the physiological relevance of Snf1-mediated TORC1 inhibition during acute glucose stress. The authors show that this regulation is important for the survival of cells under TORC1 inhibition. How do the authors envision that the acute role of Snf1 plays an important long-term physiological relevance during rapamycin treatment? Providing more support for the physiological relevance of this regulation will make this study of interest to a broad readership.

    2. Another major concern of the manuscript is the inconsistencies between the various representative immunoblots and their quantifications. The effect of AMPK activity on TORC1 signaling under glucose starvation seems very subtle. A few specific concerns are mentioned below:

    a) In figure 1A, the increase in TORC1 activity upon inhibition of analogue sensitive Snf1as by 2NM-PP1 is very marginal. Although quantification shows a significant increase, a representative western blot figure should be shown.

    b) Does deleting Snf1 itself have any effect on TORC1 activity? Lane 4 of figure 1A shows reduced activity compared to lane 1.

    c) To show the effect of Snf1 on the repression of TORC1, the time-course experiments are run on two separate gels in figure 1C. Hence, it is difficult to compare the effect of Snf1 on unscheduled reactivation of TORC1 under glucose starvation.

    d) In figure 1E, the effect of Reg1 deletion on TORC1 activity seems minor as both phospho- and total levels of Sch9 are reduced.

    Since further mechanistic insights are based on these initial findings of figure 1, solidifying these observations is very important.

    3. In figure S1, the analogue sensitive Snf1as shows significant reduction in its activity (reduced S79 phosphorylation of ACC1-GFP). This raises the concern of whether this genetic background is an ideal system to resolve the mechanism of TORC1 suppression.

    4. In figure 2, during glucose restimulation, there is increased retention of Snf1as-pThr210 in the presence of 2NM-PP1. This suggests that the upstream glucose sensing pathway as well as Snf1 might be more active than in DMSO-treated cells. This also raises concerns regarding the suitability of the genetic background for the study. Can authors comment on why this phosphorylation persists? Does the phosphoproteomic analysis give any hint for this phenotype?

    5. In figure 4H, where authors claim reduced binding of Kog1 to Pib2SESE, levels of Kog1 in input are also reduced. Can authors provide further support using colocalization studies? Also, does Pib2SESE has any defect in forming Kog1 bodies?

    6. In figure 5F, where the authors claim the Sch9SE mutant has lower TORC1 activity, the difference is very minor. Furthermore, corresponding lanes also show reduced levels of Snf1as expression. Hence, improved blots are required here. Also, an in vitro kinase assay with full-length Sch9 KD with and without the Ser288 mutation could solidify the observation that phosphorylation of Ser288 indeed affects TORC1-mediated phosphorylation.

    7. In figure 6E, the Sch9SE mutant shows no effect in the presence of rapamycin. Thus, in vivo, phosphorylation at Ser288 may not be perturbing the phosphorylation of Sch9 by TORC1.

    8. According to the author's proposed mechanism, TORC1 activity in Pib2SASA or Pib2SASA/Sch9SA backgrounds should be higher during glucose starvation compared to the control strains. However, glucose starvation shows a similar level of reduction in TORC1 activity in these backgrounds. This raises concern regarding the proposed mechanism. The authors mainly base their conclusions on Ser to Glutamate mutants. The authors should be cautious that Ser to Glutamate changes may also affect the protein structure which can confer similar phenotypes. How do the authors justify this discrepancy?

  7. Version published to 10.1101/2022.10.18.512649v1 on bioRxiv