Multilayered regulation of TORC1 signaling by Ait1, Gcn2, and SEAC/GATOR during nitrogen limitation and starvation

Abstract

The Target of Rapamycin kinase Complex I (TORC1) is a central hub in the cell growth and metabolic control network of eukaryotes. How its upstream regulators cooperate to tune signaling across environmental conditions remains unclear. Here, we combine phosphoproteomics, TORC1 activity assays, and targeted genetic perturbations to dissect TORC1 regulation in Saccharomyces cerevisiae during transitions from a high-quality nitrogen source (glutamine) to a low-quality nitrogen source (proline), and on to complete nitrogen starvation. In proline medium, Ait1 and Gcn2 attenuate TORC1 activity, establishing a partially inhibited “Low Nitrogen Adaptive” state marked by extensive metabolic reprogramming without growth arrest. In contrast, during nitrogen starvation, SEAC, Ait1, and Gcn2 cooperate to drive TORC1 into a fully inhibited state, triggering widespread dephosphorylation of its downstream targets and entry into quiescence. Our results define a multilayered regulatory circuit that governs graded TORC1 control—a design likely conserved across eukaryotes.

Budding yeast use multilayered regulation by Ait1, Gcn2, and the SEAC/GATOR complex to control TORC1 activity. This circuit design ensures that cells can adapt to low-quality nitrogen sources or shut down growth during starvation.

Introduction

The Target of Rapamycin kinase Complex I (TORC1) is a key regulator of cell growth and metabolism in eukaryotes^1–3. In the presence of abundant nutrients and pro-growth hormones, TORC1 is active and drives mass accumulation by phosphorylating proteins involved in ribosome, lipid, and nucleotide synthesis^1–9. In contrast, when nutrient or hormone levels drop, TORC1 is inhibited, triggering the activation of a wide array of stress and starvation pathways, and entry into a quiescent state^10–12.

TORC1 activity is controlled by a complex signaling network that, in humans, includes over 80 proteins. These upstream regulators converge on two key control points: the Rheb GTPase, a direct activator of TORC1^13–15, and the Rag GTPases (RagA/B and RagC/D), a dimeric complex that recruits TORC1 to the lysosomal membrane^13,16–19. Other core components in the network include the TSC complex, an inhibitor of Rheb that responds to low energy, low hormone levels, and cellular stress^15,20–22; the GATOR1/2 complex, which regulates RagA/B GTPase activity in response to amino acid and other nutrient signals^23–25; the FLCN-FNIP complex, a regulator of RagC/D GTPase activity in response to amino acids^26,27; and the Ragulator and KICSTOR scaffolds, adapter complexes that anchor the Rag GTPases and GATOR complexes to the lysosome^28–30.

Amino acid signaling to TORC1 is especially elaborate, and we now know:

1. Leucine activates TORC1 via (i) Sestrin1–3, which directly sense leucine and regulate GATOR1 via GATOR2^31–33; (ii) Sar1b, another leucine-binding protein that acts through GATOR2³⁴; and (iii) Leucyl-tRNA synthetase, which directly binds and regulates the Rags³⁵.

2. Arginine activates TORC1 via (i) CASTOR which directly senses arginine to regulate GATOR1 via GATOR2^36,37, and (ii) SLC38A9, a lysosomal amino acid transporter that, in the presence of arginine, releases its N-terminal domain to activate the Rags^38,39.

3. Methionine activates TORC1 via SAMTOR, a protein that binds and activates GATOR1 when S-adenosylmethionine (SAM) levels are low^40,41.

4. Glutamine and asparagine activate TORC1 through the Arf1 GTPase, part of a Rag-independent signaling pathway^42,43.

5. Global amino acid depletion activates Gcn2, leading to TORC1 inhibition via ubiquitination of the TOR kinase⁴⁴.

6. Small neutral amino acids activate TORC1 through the lysosomal transporter protein Pat1 (and possibly Pat4) via a poorly understood mechanism^45–47.

Amino acid and nutrient signaling is similarly complex in budding yeast, and many components are conserved^48,49. Most notably, TORC1 binds to, and is regulated by, the small GTPases Gtr1 and Gtr2—homologs of mammalian RagA/B and RagC/D, respectively^50,51. These GTPases are anchored to the vacuolar membrane by the EGO complex, the yeast equivalent of the Ragulator, and cycle between active and inactive states depending on nutrient availability^51–54.

Upstream of Gtr1/2, the SEACIT and SEACAT complexes (homologs of GATOR1 and GATOR2)^55–59 and the Lst4/7 complex (homologs of FLCN-FNIP) regulate TORC1 in response to amino acid, nitrogen, and other nutrient signals^60,61. Gtr1/2 activity is further modulated by Ait1, a yeast-specific, GPCR-like protein on the vacuolar membrane that stabilizes the inactive Gtr1–GDP/Gtr2–GTP conformation during leucine and amino acid starvation⁶².

TORC1 is also directly regulated by Pib2, a vacuolar protein that mediates TORC1 activation in response to glutamine and cysteine^63–68. Other upstream regulators include the sensor kinase Gcn2, which represses TORC1 under amino acid starvation (possibly by phosphorylating the TORC1 component Kog1/Raptor)^69,70, and Whi2, a stress-responsive phosphatase regulator that modulates TORC1 activity during leucine limitation^71,72.

Together, the studies in yeast and human cells, outlined above, have shed considerable light on the sophisticated signaling network that acts upstream of TORC1. Yet, critical questions remain: How are signals from the various TORC1 regulators integrated to control TORC1 activity and network function? What roles do the multiple—and in some cases seemingly redundant—amino acid-sensing proteins play in this process? And how do cells tune TORC1 activity in response to subtle nutrient fluctuations, rather than the binary nutrient-replete to starvation transitions studied to date?

Here, to address these questions, we carry out a systematic and quantitative analysis of TORC1 signaling in yeast during transitions from a high- to a low-quality nitrogen source, and on to complete nitrogen starvation. By combining phosphoproteomics, protein abundance profiling, TORC1 activity assays, and targeted genetic perturbations, we define how different upstream regulators—including Gtr1/2, Pib2, Ait1, Gcn2, and SEAC—cooperate to shape TORC1 output across conditions. These experiments reveal that, in low-quality nitrogen medium, Ait1 and Gcn2 drive cells into a Low Nitrogen Adaptive (LoNA) state, defined by selective reprogramming of nutrient transport, metabolism, and gene expression. In contrast, in complete nitrogen starvation, SEAC cooperates with Ait1 and Gcn2 to drive full inhibition of TORC1, broad dephosphorylation of its targets, and the induction of quiescence-associated programs. These results uncover key principles of TORC1 regulation and reveal how eukaryotic cells balance growth and metabolism across a wide range of environmental conditions.

Results

TORC1 signaling in a poor nitrogen source

To study TORC1 signaling in low/intermediate levels of nitrogen, we grew a prototrophic strain of Saccharomyces cerevisiae in synthetic medium containing the high-quality nitrogen source glutamine (SD+Gln), transferred the cells into synthetic medium containing the low-quality nitrogen source proline (SD+Pro)^73,74, and then followed the response over time (0, 30, 60, 120 and 240 min):

As a first step, we measured the phosphorylation of the key TORC1 substrate Sch9—an AGC kinase that regulates protein and ribosome synthesis^6,75,76. These experiments showed that TORC1 is rapidly inhibited in SD+Pro, but then partially reactivated as the cells adapt to growth in a low-quality nitrogen source (Fig. 1 and ref. ⁷⁷). In contrast, TORC1 is completely (or near completely), inactivated in cells that are transferred from SD+Gln, to medium without nitrogen (SD-N; Fig. 1).

Fig. 1. Sch9 phosphorylation during nitrogen limitation.

a Domains in Sch9, including the C2-domain, Kinase domain, Regulatory domain (RD), and C-terminal extension (CE) domain. TORC1 phosphorylates Sch9 at positions S711, T723, T737, S758, S765 (refs. ^75,106). Protein extracts were treated with NTCB prior to analysis, leading to cleavage after cysteine 555 (ref. ⁷⁵). b Anti-HA Western blot of protein extracts from wild-type cells expressing Sch9-3xHA from its native locus as they transition from growth in SD+Gln to growth in SD+Pro (top panel) or SD-N (bottom panel). The portion of the gel shown here only includes the ~30 kDa, C-terminal fragment of Sch9 containing the TORC1 target sites. c Quantification of Sch9 mobility data shown in (b) as well as 72 replicate experiments for SD+Pro and 8 replicate experiments for SD-N. Open circles and error bars show the average and standard deviation for each timepoint; filled circles show the data from individual experiments. SD+Pro samples were collected throughout this study as controls for experiments examining mutant strains (Figs. 5, 6). The standard error for all time-points in the SD+Pro dataset is less than 0.01. Statistical analysis (using a two-sided Welch’s t-test) comparing the two time-courses show that the Sch9 phosphorylation levels are similar in SD+Gln (p-value 6.36E-01) but diverge (p < 0.01) at subsequent timepoints.

Next, to build up a global view of TORC1 signaling, we used mass spectrometry to track the level of ~10,000 phosphopeptides during the transition from SD+Gln to SD+Pro (Supplementary Data 1). This experiment led to the identification of 436 phosphopeptides, from 232 proteins, that were significantly up- or down-regulated in SD+Pro (>3-fold change, Benjamini-Hochberg adjusted p < 0.05; Fig. 2a and Supplementary Data 1). Notably, the vast majority of the phosphopeptide data did not follow the pattern seen for Sch9—namely, a rapid change followed by a dampening of the response (adaptation). Instead, the time-course data fell into four groups (Fig. 2a):

Fig. 2. Global phosphorylation changes during nitrogen limitation.

a Heatmap of the 436 phosphopeptides that change significantly (>3-fold change, BH corrected p < 0.05) during the SD+Gln to SD+Pro transition as measured by mass spectrometry, and split into four groups using hierarchical clustering (n = 4, p-values from an ANOVA test). The full dataset is in Supplementary Data 1. b Line graphs for key proteins from (a) highlighting the timing and stability of the phosphorylation changes. Open circles and error bars show the average and standard deviation for each timepoint; filled circles show the data from individual experiments (n = 4). c, d Venn diagram showing the top Gene Ontology (GO) groups for the 232 proteins in panel (a), and a list of key target proteins. Asterisks mark proteins where a change in abundance accounts for the observed up- or down-regulation of one or more daughter phosphopeptide.

(I) Phosphopeptides in Group I were dephosphorylated/lost within 30 min and then remained depleted through most, or all, of the four-hour time-course (Fig. 2a). This group includes peptides from the amino acid permease Agp1⁷⁸, the stress/starvation response kinases Ypk3 and Npr1^79–83, the TORC1-associated transcriptional regulators Not5 and Gln3^84,85, a TORC1-dependent enzyme controlling the first step in serine synthesis (Ser33)^77,86, and Pyruvate dehydrogenase (Pda1; Fig. 2b)⁸⁷.

(II) Phosphopeptides in Group II were dephosphorylated/lost with a time constant of one to two hours (Fig. 2a), or rapidly after a delay of ~30 min, as seen for the high-affinity glutamine permease Gnp1⁸⁸ (Fig. 2a, b).

(III-IV) Phosphopeptides in Groups III and IV increased over time, with time constants of <30 min (Group III and Vtc5)⁸⁹ or 1–2 h (Group IV and Mep2)⁹⁰, and again, most remained upregulated throughout the time-course (Fig. 2a, b).

It is worth noting, however, that: (i) a small number of peptides followed a trajectory similar to Sch9—rapid dephosphorylation followed by rephosphorylation (red bar, Group I; Fig. 2a). This group included a peptide from Maf1—a well-established Sch9 substrate, and regulator of tRNA synthesis (Fig. 2b)^6,91. (ii) a second small set of peptides followed the opposite trend; rapid phosphorylation followed by dephosphorylation (green bars, Group III and Put4; Fig. 2a, b). (iii) the C-terminal phosphosites on Sch9, monitored in Fig. 1, were not detected in the proteomics dataset.

Phosphorylation and protein level changes in proline medium

Changes in the level of a phosphopeptide can be caused by a phosphorylation/dephosphorylation reaction, or by a change in the level of the parent protein. To control for this, and learn more about the gene expression program in SD+Pro, we also followed protein abundance during the SD+Gln to SD+Pro transition. We were able to track the level of 3555/ ~ 5800 proteins across the time-course, following an average of 8 peptides per protein (Supplementary Fig. 1, Supplementary Data 1). Comparing the protein abundance data to the phosphopeptide data revealed that only 24 of the 232 proteins that had a significant change in the phosphopeptide experiment had a similar change in total protein level (with 80% of the proteins accounted for based on the overlap of the two datasets). We also identified 186 proteins that are significantly up-regulated, and 26 significantly down-regulated, during the SD+Gln to SD+Pro transition (>3-fold change, p < 0.01, FDR < 2%; Supplementary Fig. 1).

Examining the list of proteins that had a significant change in phosphorylation or abundance during the SD+Gln to SD+Pro growth transition revealed many well-known targets of TORC1 (as discussed later), but several functional groups stood out: First, 47/232 proteins that are phosphorylated or dephosphorylated in SD+Pro are transporters, including 22 amino acid and nitrogen transporters (Fig. 2c). Second, 80/232 of the (de)phosphorylated proteins localize to the cell periphery, and 43/232 are part of the vacuole (Fig. 2d), suggesting they play a role in nutrient transport, nutrient storage, and/or signaling. Third, multiple kinases and transcription factors associated with metabolic reprogramming and/or nutrient transport including Ksp1, Npr1, Yck1, Yck2, Ser33, Met2, Gat1, and Gln3^{79–82,84,86,92–94} are (de)phosphorylated in SD+Pro (Fig. 2c). Finally, 54/186 proteins that are up-regulated in SD+Pro are involved in small molecule metabolism, including glutamine synthesis, nitrogen metabolism, and amino acid catabolism (Supplementary Fig. 1b). Therefore, yeast cells reprogram nutrient transport, nutrient storage, amino acid metabolism, and nitrogen metabolism, to support growth in low-quality nitrogen medium (Fig. 2 and Supplementary Fig. 1).

Comparison of signaling in SD+Pro and SD-N

Next, to establish a reference for our SD+Pro data, we measured the phosphorylation and protein abundance changes that occur as yeast transition from SD+Gln to SD-N. As expected, the transition to SD-N triggered dramatic reprogramming of protein phosphorylation, including significant changes in the level of 630 phosphopeptides from 460 proteins (>3-fold change, p < 0.01, FDR < 3.2%; Fig. 3a, Supplementary Data 2). Again, only around 10% of these changes came from the up or down-regulation of the parent protein (Supplementary Fig. 2, Supplementary Data 2).

Fig. 3. Global phosphorylation changes during nitrogen starvation.

a Heatmap of the 630 phosphopeptides that change significantly (>3-fold change, p < 0.01, FDR < 3.2%) during the SD+Gln to SD-N transition, split into three groups to highlight variation in the timing of the response (n = 4, p-values from an ANOVA test). The full dataset is in Supplementary Data 2. b Venn diagram showing the top GO functional groups for the 460 proteins found in (a), and a list of key target proteins. c Scatterplot comparing the change in phosphopeptide abundance during the SD+Gln to SD+Pro and the SD+Gln to SD-N transition after 30 min, for all peptides that change significantly (>3-fold change, p < 0.01) in either condition. Solid and dashed lines show the values for a perfect correlation, and a spread of +/−0.5 on log₂ scale. The red dots highlight phosphopeptides that have a > 3-fold difference between conditions, and the parent proteins are listed below the graph.

The phosphopeptides that are up/down-regulated in SD-N were clearly distinct from those that change in SD+Pro, and include 101 proteins involved in the cell cycle and 115 proteins involved in gene regulation (Fig. 3b), in line with the fact that yeast exit the cell cycle and enter quiescence in complete nitrogen starvation but not in SD+Pro^74,75,95. However, a more precise comparison between the SD+Pro and SD-N response was difficult since the datasets are from separate mass spectrometry experiments and consequently include different sets of peptides. Therefore, to compare the phosphorylation changes that occur in SD+Pro and SD-N in detail, we carried out two additional experiments, examining samples from:

• (i)Cells grown in (a) SD+Gln, and cells transferred from SD+Gln to (b) SD+Pro or (c) SD-N, for 30 min.
• (ii)Cells grown in (a) SD+Gln, and cells transferred from SD+Gln to (b) SD+Pro, (c) SD-N, or (d) SD+Gln + rapamycin, for 4 hours.

These experiments revealed that the phosphorylation programs triggered in SD+Pro and SD-N are nearly identical after 30 minutes of nutrient deprivation/limitation (although with slightly larger changes in -N than Pro; Fig. 3c), but then diverge after four hours (Fig. 4a)—exactly as seen for Sch9 (Fig. 1). Specifically, in the 4-h time-point dataset (which includes nearly 14,000 phosphopeptides; Supplementary Data 3), there are 3243 phosphopeptides that change significantly (>4-fold change, p < 0.001, FDR < 1%) in SD+Pro, SD-N, and/or rapamycin (Fig. 4a). 784 of these phosphopeptides had a > 3-fold larger change in SD-N than in SD+Pro (blue and red stars, Fig. 4a).

Fig. 4. Global phosphorylation changes in nitrogen limitation versus nitrogen starvation.

a Scatterplot comparing phosphopeptide levels after four hours in SD-N and SD+Pro for all peptides that change significantly (>4-fold change, p < 0.001; FDR < 1%) in SD-N, SD+Pro, or rapamycin (3243 total). Each dot shows the mean abundance for a single phosphopeptide (n = 4). The solid and broken lines show the values for a perfect correlation and a spread of +/−0.5 on log₂ scale, respectively. Blue stars mark phosphopeptides that change >3-fold more in SD-N than in SD+Pro, and red stars mark those that change >3-fold more in both SD-N and rapamycin than in SD+Pro. The full dataset is in Supplementary Data 3. b Heatmap showing the data for all 286 phosphopeptides (on 185 proteins) marked by red stars in (a). (c) Raw data for representative proteins from (b). Open circles and error bars show the average and standard deviation for each timepoint; filled circles show the data from individual experiments (n = 4). d List of key proteins from (b).

We reasoned that the amplified response in SD-N could be due to: (i) stronger TORC1 repression in SD-N than SD+Pro, (ii) activation/inactivation of one or more additional signaling pathway in SD-N versus SD+Pro, or (iii) a mixture of models (i) and (ii). To distinguish between these possibilities, and identify the TORC1-dependent changes, we compared the phosphopeptide abundance in SD+Pro, SD-N, and rapamycin: We found 173 phosphopeptides from 124 proteins that had a > 3-fold larger change in -N, than in Pro and rapamycin, suggesting that they are regulated (at least in part) by a nitrogen-dependent kinase/phosphatase outside of the TORC1 pathway (Supplementary Fig. 3a and Supplementary Data 3). We also found 286 phosphopeptides from 185 proteins that had >3-fold larger change in -N and rapamycin, than in Pro, and are therefore only fully phosphorylated/dephosphorylated upon complete inhibition of the TORC1 pathway (red stars, Fig. 4a–b). Almost all of the other phosphorylation changes were similar in SD-N, SD+Pro, and rapamycin (Fig. 4, Supplementary Figs. 3b and 4), indicating that most of the signaling in SD+Pro and SD-N is (i) TORC1 dependent and (ii) fully activated upon partial or complete inhibition of TORC1.

Examining the 286 peptides with strong dephosphorylation/loss in SD-N and rapamycin, but not SD+Pro (Fig. 4a, b), we found sites at the C-terminus of Sch9 (the TORC1 target sites), the key Sch9 target proteins controlling ribosome/protein synthesis (Dot6, Tod6, Stb3, Maf1)^76,91, and numerous proteins involved in cell cycle control, autophagy, global gene regulation, cell polarity, meiosis, and other processes associated with growth control, cell cycle arrest, and quiescence (Fig. 4c, d). Thus, cells growing in SD+Pro activate a Low Nitrogen Adaptive (LoNA) response through TORC1 (by reprogramming nutrient transport and metabolism; Figs. 1, 2), while cells exposed to complete nitrogen starvation activate a LoNA response and simultaneously block cell growth and division via TORC1 and TORC1-Sch9 (Figs. 1, 3, 4).

Finally, to place these responses in the broader context of TORC1 signaling, we compared our phosphoproteomic dataset to those from previous studies examining signaling through TORC1-regulated kinases—including Sch9, Npr1, Yak1, Rim15, and Atg1^6,79,96. This comparison reinforced the conclusions above and shed light on the network’s behavior across nitrogen conditions (Supplementary Fig. 5 and Supplementary Table 1). Specifically, our analysis indicates that (i) Npr1, Yak1, and Rim15—kinases involved in stress and starvation responses—are activated at moderate to high levels in both SD+Pro and SD-N; (ii) Atg1, a key regulator of autophagy, is strongly activated in SD-N but not SD+Pro; and (iii) Sch9, a central activator of cell growth, is strongly repressed in SD-N but not SD+Pro. It therefore appears that, in proline medium, the TORC1 supercomplex is modulated to activate kinases involved in adaptation to stress and starvation—such as Npr1, Yak1, and Rim15—but still restrain Atg1/autophagy and sustain Sch9-dependent growth.

Gtr1/2 and Pib2 activate TORC1 during the SD+Gln to SD+Pro transition

To identify proteins that drive the TORC1 circuit into the LoNA state in SD+Pro, we measured the impact that various upstream regulators in the TORC1 pathway have on Sch9 phosphorylation during the SD+Gln to SD+Pro transition. As a first step, we examined the influence that the two direct TORC1 regulators in budding yeast, Gtr1/2 and Pib2, have on signaling:

In line with previous work, we found that a strain expressing a mutant form of Gtr1/2—thought to lock the complex in the inactive Gtr1-GDP, Gtr2-GTP state (Gtr1[S20L] Gtr2[Q66L]; Gtr1/2^off)—exhibited a mild defect in TORC1-Sch9 signaling (Fig. 5a)^55,77. This initially suggested that Gtr1/2 are inactive in intermediate/low nitrogen conditions. However, the other data we collected told a different story. Specifically, we found that deletion of Gtr1 reduced TORC1 activity in SD+Gln, and completely abolished TORC1 signaling in SD+Pro (Fig. 5a). Furthermore, a strain expressing a second mutant form of Gtr1/2—thought to lock the complex in the active Gtr1-GTP, Gtr2-GDP state (Gtr1[Q65L] Gtr2[S23L]; Gtr1/2^on)⁵⁵—maintained wild-type levels of TORC1-Sch9 activity throughout the SD+Gln to SD+Pro transition (Fig. 5a). Finally, strains missing the Gtr1 activator and GEF, Vam6⁵¹, or a subunit of the Gtr2 activator and GAP, Lst4/7⁶⁰, had large signaling defects in SD+Pro (Fig. 5a). Thus, it appears that (i) Gtr1/2 are active (or partially active) in SD+Pro and are required for TORC1 signaling to Sch9 in low-to-intermediate levels of nitrogen, and (ii) the mutations in the Gtr1/2^off strain do not actually lock Gtr1/2 in an inactive state—at least in SD+Pro.

Fig. 5. Role of Gtr1/2 and Pib2 in TORC1-Sch9 regulation during nitrogen limitation.

a, b Anti-HA Western blots of protein extracts from wild-type and mutant cells expressing Sch9-3xHA from its native locus as they transition from growth in SD+Gln to SD+Pro. Quantification of Sch9 mobility data shown in (a, b) as well as two repeat experiments for each mutant (n = 3) is shown on the right. The open circles and error bars show the average and standard deviation of the difference between the mutant and wild-type data, comparing samples grown, processed, and run together on the same gel (see Supplementary Table 2 for p-values). These values were added to the average wild-type data from Fig. 1 (shown by the broken line) for easy visualization. The filled circles show the difference calculated in each individual experiment.

Deletion of Pib2 also reduced TORC1 activity in SD+Gln, and severely impaired TORC1 activity in SD+Pro, indicating that Pib2 cooperates with Gtr1/2 to activate TORC1 in low-quality nitrogen medium (Fig. 5b). To determine which of the four domains in Pib2^63,97–99 controls TORC1 activity in SD+Pro, we constructed and analyzed prototrophic strains missing: (1) the C-terminal activating domain (pib2CADΔ); (2) the N-terminal inhibitory domain (pib2NIDΔ); (3) the Kog1 binding domain (pib2KBDΔ); (4) the phosphatidylinositol 3-phosphate binding, FYVE domain (pib2FYVEΔ); and (5) the C-terminal and N-terminal domains (pib2NIDΔCADΔ). These experiments revealed that the NID, KBD and CAD domains of Pib2 each play relatively minor roles in TORC1 signaling during the transition from SD+Gln to SD+Pro (Fig. 5b). Instead, TORC1 activation in SD+Pro depends on the FYVE domain, and the combined action of the CAD and NID domains (Fig. 5b). This pattern contrasts with previously published results collected on cells growing in nutrient rich medium containing all 20 amino acids, where the CAD—and to a lesser extent the FYVE domain—are important for TORC1 activation, while the NID domain is largely dispensable^97,98. Thus, our data suggest that there is a rearrangement of the Pib2-TORC1 interface in SD+Pro, but Pib2 continues to activate TORC1 in this low-quality nitrogen source.

Ait1 and Gcn2, but not SEAC, inhibit TORC1-Sch9 during the SD+Gln to SD+Pro transition

Our analysis of TORC1 signaling in Gtr1/2 and Pib2 mutant strains shows that Gtr1/2 exists in its active Gtr1-GTP, Gtr2-GDP state, and cooperates with Pib2 to promote TORC1-Sch9 signaling in SD+Pro (Fig. 5). What, then, drives TORC1 into a partially inhibited state during the transition from SD+Gln to SD+Pro?

Previous studies have identified the Gtr1/2 GAP—called SEAC in yeast and GATOR in humans—as the primary mediator of nitrogen and amino acid starvation signals to TORC1^55–58. To determine whether SEAC drives cells into the slow-growth, LoNA state in SD+Pro, we examined TORC1 signaling during the SD+Gln to SD+Pro transition in strains lacking (i) Npr2, the catalytic subunit of SEACIT⁵⁸, and (ii) Iml1, a key scaffolding subunit in SEACIT⁵⁸. Surprisingly, TORC1-Sch9 activity was nearly identical across the npr2Δ, iml1Δ, and wild-type strains (Fig. 6a).

Fig. 6. Role of Ait1, Gcn2, and Npr2 in TORC1-Sch9 regulation during nitrogen limitation.

a–c Anti-HA Western blots of protein extracts from wild-type and mutant cells expressing Sch9-3xHA from its native locus as they transition from growth in SD+Gln to SD+Pro. Quantification of Sch9 mobility data (n = 3) is shown on the right (see Supplementary Table 2 for p-values). Open circles and error bars show the average and standard deviation for each timepoint; filled circles show the data from individual experiments. The data were analyzed, and the graphs plotted, as described in Fig. 5.

Work from our lab and others has shown that Ait1, Gcn2 and Whi2 inhibit TORC1 during amino acid starvation, but not complete nitrogen starvation^62,69,71,72. Therefore, to determine if these factors regulate TORC1 during a shift to a poor-quality nitrogen source, we examined TORC1-Sch9 signaling in ait1Δ, gcn2Δ, and whi2Δ cells. These experiments revealed that Ait1, but not Gcn2 or Whi2, is required for full TORC1 repression in SD+Pro medium (Fig. 6b). Moreover, when we examined TORC1 signaling in strains missing combinations of Ait1, Gcn2 and Npr2 (Whi2 strains grew slowly, preventing isolation of double mutants), we found that deletion of Ait1 and Gcn2 almost completely blocked TORC1 inhibition in SD+Pro (Fig. 6c). This phenotype was only slightly stronger in a strain missing Ait1, Gcn2 and Npr2 (Fig. 6c), indicating that Ait1 and Gcn2 play the dominant role in inactivating TORC1 in SD+Pro. We also found that deleting Npr2 and Gcn2 led to a weak, early defect in the TORC1-Sch9 signaling, while deleting Npr2 and Ait1 led to an amplification of the late TORC1-Sch9 signaling defect seen in the ait1Δ strain (Fig. 6c). Together, these findings show that Ait1 and Gcn2 cooperate to inhibit TORC1-Sch9 signaling during the SD+Gln to SD+Pro transition: Ait1 is required throughout the transition, Gcn2 is important during the early stage of the response (prior to adaptation), and Npr2/SEAC plays a backup role in the absence of Ait1 or Gcn2.

Ait1 and Gcn2 cooperate with Npr2 to drive metabolic reprogramming in a poor nitrogen source

Next, we asked whether Ait1 and Gcn2, and/or Npr2/SEAC drive the phosphorylation and dephosphorylation of transporters, metabolic enzymes, transcription factors, and other key proteins as cells enter the LoNA state (seen in Fig. 2). To do this, we repeated the phosphoproteomic analysis of the SD+Gln to SD+Pro transition, comparing phosphorylation dynamics in the ait1Δgcn2Δ, npr2Δ and wild-type cells. We focused on the 30-minute time point to minimize secondary effects in the mutant strains.

Consistent with our earlier findings, we identified 427 phosphopeptides from 216 proteins that showed significant changes in SD+Pro at 30 min in the wild-type and/or mutant backgrounds (>3-fold change, p < 0.01; FDR < 3%). Most of these phosphorylation/dephosphorylation events were blocked, or severely limited, in the aitΔgcn2Δ strain (Fig. 7a, Supplementary Data 4). A notable exception was a set of 19 phosphopeptides that still increased more than 4-fold in ait1Δgcn2Δ cells, including peptides from Npr1, its substrate Par32, and several transporter proteins (highlighted with blue stars in Fig. 7a)⁸⁰. Npr2/SEAC was also required for full TORC1 inhibition in SD+Pro, but in line with our earlier TORC1-Sch9 experiments, did not affect signaling to the Sch9 targets Maf1, Stb3, and Tod6 (red stars with gene names listed below Fig. 7a).

Fig. 7. Contributions of Ait1+Gcn2 and Npr2 to TORC1 signaling during nitrogen limitation.

a Scatterplots comparing the phosphopeptide abundance change in wild-type, ait1Δgcn2Δ (left), and npr2Δ (right) strains subjected to nitrogen limitation (SD+Pro) for 30 min. Each dot shows the average data for a single phosphopeptide (n = 4). Peptides with strong regulation in only one mutant strain are highlighted, and the names of their parent proteins are listed below the graphs (with the number of distinct phosphosites in parentheses). The full dataset is in Supplementary Data 4. b Heatmap showing the significant phosphorylation changes (>3-fold change, p < 0.01, FDR < 3%) that occur after 30 min in SD+Pro, in wild-type, ait1Δgcn2Δ and npr2Δ mutant cells. The left columns show the log₂ fold change in each mutant caused by the SD+Gln to SD+Pro transition, the right columns show the ratio of the phosphopeptide abundance in mutant versus wildtype cells in SD+Gln medium. The bars show phosphopeptides that are hypo- (green) or hyper-phosphorylated (red) due to Npr2 activity and nutrient deprivation.

To see if Ait1 and Gcn2, or Npr2/SEAC, regulate TORC1 in the initial, nutrient-rich conditions, we also compared the phosphorylation profiles of aitΔgcn2Δ, npr2Δ, and wild-type cells growing in SD+Gln. Deleting Ait1 and Gcn2 had minimal impact on protein phosphorylation in SD+Gln (Fig. 7b). In contrast, deleting Npr2 had a strong effect, and pushed the phosphorylation program away from that found in low nitrogen conditions (red and green bars, Fig. 7b; Supplementary Fig. 6).

Together, these results indicate that Ait1 and Gcn2 inhibit TORC1 during the SD+Gln to SD+Pro transition, reducing Sch9 signaling and promoting reprogramming of nutrient transport, storage, and metabolism. Npr2/SEAC, on the other hand, is partially active in SD+Gln, and then cooperates with Ait1 and Gcn2 in SD+Pro, to drive the reprogramming of nutrient transport and metabolism, but does not affect TORC1-Sch9 signaling in SD+Gln or SD+Pro.

Multilayered regulation of TORC1 in complete nitrogen starvation

In our final experiments, we asked whether Ait1, Gcn2 and/or Npr2/SEAC regulate TORC1-Sch9 signaling during complete nitrogen starvation. First, we grew cells in SD+Gln and then transferred them into SD-N medium. As expected, we found (i) that npr2Δ cells had a defect in TORC1 inhibition in SD-N at time-points where the SD+Pro and SD-N responses diverge (Fig. 1), and (ii) that deletion of Ait1 or Gcn2 alone, had little impact on TORC1 signaling in SD-N (Fig. 8a). Remarkably, however, the combined deletion of Ait1 and Gcn2 almost completely blocked TORC1 inhibition for 15 min, and this strong phenotype remained largely unchanged in the triple mutant lacking Ait1, Gcn2, and Npr2 (Fig. 8a).

Fig. 8. Ait1, Gcn2, and Npr2 cooperate to regulate TORC1-Sch9 signaling during complete nitrogen starvation.

Anti-HA Western blots of protein extracts from wild-type and mutant cells expressing Sch9-3xHA from its native locus as they transition from (a) growth in SD+Gln to growth in SD+Pro, or (b) growth in SD complete medium to growth in SD-N medium. Quantification of Sch9 mobility data (n = 3) is shown on the right (see Supplementary Table 3 for p-values). Open circles and error bars show the average and standard deviation for each timepoint; filled circles show the data from individual experiments. The data were analyzed, and the graphs plotted, as described in Fig. 5.

To align with previous studies^{59,62,69,95,100}, we also examined the impact of Ait1, Gcn2 and Npr2 on cells growing in SD complete medium and then transferred to SD-N medium. The results were very similar to those for cells starting in SD+Gln, except that Sch9 dephosphorylation/SEAC activation was accelerated and more pronounced in cells transitioning from SD complete to SD-N (Fig. 8b).

Taken together, these results show that Ait1 and Gcn2 are the primary TORC1 regulators during the early stages of complete nitrogen starvation—a role previously masked by their functional redundancy. They also show that SEAC is not the main driver of the response, but instead facilitates the transition from the partial TORC1 inhibition seen in SD+Pro to the full inhibition observed in SD-N. Finally, our results suggest that at least one additional, unknown layer of TORC1 regulation becomes engaged when cells attempt to grow (with fully active TORC1) for an extended period in complete nitrogen starvation.

Discussion

Previous studies of TORC1 regulation in yeast have focused on identifying proteins that drive the response to complete nitrogen or amino acid starvation in auxotrophic laboratory strains^{51,59,62,69,72,95,99}. These studies led to the view that SEAC is the principal driver of TORC1 repression, while Ait1, Gcn2, and Whi2 play minor or subsidiary roles. However, wild-type (prototrophic) yeast—and cells from most other organisms—rarely encounter abrupt and complete nutrient deprivation. Instead, the TORC1 signaling network must interpret and respond to gradually fluctuating nutrient levels, punctuated by occasional episodes of near or complete starvation.

To construct a more physiologically relevant model, we examined TORC1 regulation in a prototrophic strain during (i) nitrogen limitation, triggered by a shift from SD+Gln to SD+Pro, and (ii) complete nitrogen starvation, triggered by a shift to SD-N. We also challenged the prevailing assumption that TORC1 signals uniformly to its downstream targets across conditions by combining standard TORC1 activity assays with phosphoproteomic analysis of global signaling [building on our previous work⁷⁷].

Our analysis revealed that the TORC1 circuit consists of multiple layers that engage progressively as nutrient limitation intensifies (Fig. 9): First, SEAC is active at a low level, even in the high-quality nitrogen source glutamine, where it promotes phosphorylation/dephosphorylation—and likely activation—of select nutrient transporters and components of the nitrogen catabolite repression response (Fig. 7 and Supplementary Fig. 6). Then, as cells transition from glutamine to the lower-quality nitrogen source proline, Ait1 and Gcn2 collaborate with low-level SEAC signaling to trigger transient, partial inhibition of Sch9 and growth, along with sustained and complete reprogramming of nutrient transport and metabolism required for growth under low-nitrogen conditions (the LoNA state; Figs. 1–7). Finally, in complete nitrogen starvation, SEAC becomes fully activated and cooperates with Ait1 and Gcn2 to drive Sch9 and other remaining TORC1-dependent outputs into an inactive state, leading to cell cycle arrest and entry into quiescence (Figs. 4, 8).

Fig. 9. Model of multilayered TORC1 regulation across nitrogen limitation and starvation.

In glutamine, a high-quality nitrogen source, cells grow rapidly with only low-level transporter and metabolic reprogramming relative to SD complete medium. Upon a transition to nitrogen limitation or starvation, Ait1, Gcn2, and—to a lesser extent—SEAC cooperate to inactivate TORC1, causing growth inhibition and metabolic reprogramming. Over time, however, the outcomes diverge: (1) under nitrogen limitation (SD+Pro), cells enter a Low Nitrogen Adaptive (LoNA) state, maintaining metabolic reprogramming while partially reactivating TORC1 to support slow growth; whereas (2) under nitrogen starvation (SD-N), cells sustain strong TORC1 inhibition driven by SEAC together with Ait1 and Gcn2, leading to growth arrest and entry into quiescence. TORC1-dependent kinases that are activated or repressed in each condition are indicated based on the analysis in Supplementary Table 1 and Fig. 1. See text for further details. Figure created in BioRender. Padilla, C. (2025) Biorender.com. s79920z.

One important consequence of this multilayered regulatory circuit design is that yeast mount similar responses during initial phases of nitrogen limitation and nitrogen starvation (30 min time points, Figs. 1 and 3). Specifically, our data suggest that yeast cells initially reprogram nutrient transport and metabolism to try and optimize growth in the low/no nitrogen environment. If these efforts are successful, the cells grow slowly at a rate determined by the Ait1 and Gcn2 dependent attenuation of TORC1-Sch9 signaling. If not, further decreases in overall nitrogen levels trigger full activation of SEAC, to block TORC1-Sch9 signaling and drive cells into quiescence. Thus, yeast deploy (i) a sensitive regulatory circuit—including Ait1 and Gcn2—to tune TORC1 output and control metabolism in response to moderate changes in nitrogen availability/quality, and (ii) a low-sensitivity circuit—including SEAC—that detects complete starvation and enforces full TORC1 inhibition.

Given the similarities between the TORC1 regulatory circuits in yeast, humans and other eukaryotes, we propose that the multilayered regulatory logic described here—featuring signaling modules for adaptive tuning versus full shutdown, and an ability to enter a Low Nitrogen Adaptive (LoNA) state—is broadly conserved.

Methods

Strain construction

All strains used in this study were generated in haploid S. cerevisiae (W303 background) using standard methods^101,102 and are listed in Supplementary Data 5. Prototrophy was restored to all strains using single-copy plasmids encoding the required markers (LEU2, HIS3, and/or URA3) and/or by integrating the TRP1 gene into the genome¹⁰³.

Cell growth and Sch9 mobility shift assays

Cells were grown in conical flasks at 30°C, shaking at 200 rpm, to mid-log phase (OD₆₀₀ = 0.45) in synthetic complete medium with 2% glucose (SD; Fig. 8b) or SD medium lacking ammonium sulfate and amino acids, but containing 3 mM glutamine (SD+Gln; Figs. 1–7 and 8a). At that point, a 47 mL sample was collected, mixed with 3 mL of 100% trichloroacetic acid (TCA), and held on ice for at least 30 min (and up to 6 h). The remaining culture was filtered and transferred to either SD medium lacking ammonium sulfate and amino acids (SD-N), or to SD medium lacking ammonium sulfate and amino acids but supplemented with 3 mM proline (SD+Pro), after two washes with 100 mL of the same medium. Additional samples were then collected and processed as described above.

TCA-precipitated samples were centrifuged at 3500 g for 5 min at 4 °C, washed twice with cold water, twice with acetone, and then disrupted by two 5-s sonication pulses at 15% amplitude. Samples were then centrifuged at 13,500 g for 30 s, dried in a speedvac for 10 min at room temperature, and stored at −80 °C.

Proteins were extracted by bead beating (6 × 1 min, full speed) in urea lysis buffer (6 M urea, 50 mM Tris–HCl pH 7.5, 5 mM EDTA, 1 mM PMSF, 5 mM NaF, 5 mM NaN₃, 5 mM NaH₂PO₄, 5 mM p-nitrophenylphosphate, 5 mM β-glycerophosphate, and 1% SDS) supplemented with protease and phosphatase inhibitor tablets (Roche, cat. #04693159001 and #04906845001). Lysates were then centrifuged at 2000 g for 5 min, resuspended, heated at 65 °C for 10 min, and clarified at 13,500 g for 10 min. Supernatants were stored at −80 °C.

Cleavage of Sch9 was performed using 2-nitro-5-thiocyanatobenzoic acid (NTCB) in 100 mM CHES (pH 10.5) overnight at 33°C in the dark (1 mM NTCB), as previously described^75,95.

For phosphorylation analysis, 50 μg of total Sch9-3xHA protein was separated on a 10% acrylamide gel (3.5 h at 70 V), transferred to a nitrocellulose membrane, and probed using anti-HA antibody (Sigma-Aldrich, cat. #11583816001, 1:1000) and an IRDye 800CW-conjugated secondary (Li-Cor, cat. #926-32210, 1:10,000). Blots were imaged using a Li-Cor Odyssey Scanner and quantified using Image Studio software, by comparing the dephosphorylated band to all higher molecular weight species. Sch9 phosphorylation data were compared via paired t-tests, comparing all mutants to the wild-type strain, and select mutants to each other. The p-values for nitrogen limitation (SD+Pro) and nitrogen starvation (SD-N) experiments are listed in Supplementary Tables 2 and 3, respectively.

Full images of the 94 Western blots from this study are in the Source Data file.

Phosphoproteomics

Cells were grown and harvested by TCA precipitation as described above. All subsequent steps used MS-grade reagents. Pellets were resuspended in 400 μL of MS-urea buffer (8 M urea, 100 mM ammonium bicarbonate, 5 mM EDTA), extracted by bead beating, and quantified using a BCA assay.

200 μg of protein per sample was diluted to 1 μg/μL, reduced and alkylated with 5 mM TCEP and 5 mM iodoacetamide for 30 min at room temperature. Digestion was carried out sequentially: first with LysC (20 ng/μL, NEB P8109S) for 3 h at 37 °C (urea concentration: 5.5 M), then with trypsin (2 μg, Promega V511C) overnight in 1 M urea.

Digestion was quenched with 1% TFA, and peptides were clarified (21,000 g, 5 min), desalted using Sep-Pak Plus C18 cartridges (Waters WAT020515), and eluted with solution B (65% MeCN, 0.1% TFA). Eluates were dried and stored at −80 °C.

Phosphopeptides were then enriched using Ti(IV)-IMAC beads (MagReSyn MR-TIM005), following the manufacturer’s protocol: Beads were equilibrated in loading buffer (0.1 M glycolic acid, 80% MeCN, 5% TFA), incubated with peptides for 20 min, washed sequentially, and eluted in 1% NH₄OH into 10% FA. The flow through (containing dephosphorylated peptides) was also collected during the purification, and then processed and analyzed as described below to measure total protein levels.

Finally, the peptides were desalted using micro spin columns (Nest Group SEM SS18V), eluted in 50% MeCN with 0.1% FA, dried in a speedvac, and stored at −80 °C.

All samples analyzed in this study were prepared as biological quadruplicates (n = 4 per time point or per strain), and analyzed together with samples from the same timecourse and/or with the matching mutant samples, as described in the text.

Mass spectrometry

HPLC-ESI-MS/MS was performed in positive ion mode on a Thermo Scientific Orbitrap Fusion Lumos tribrid mass spectrometer fitted with an EASY-Spray Source (Thermo Scientific, San Jose, CA) as previously described¹⁰⁴. In brief, NanoLC was performed using a Thermo Scientific UltiMate 3000 RSLCnano System with an EASY Spray C18 LC column (Thermo Scientific, 50 cm × 75 µm inner diameter, packed with PepMap RSLC C18 material, 2 µm, cat. # ES903); loading phase for 15 min; mobile phase, linear gradient of 1–47% ACN in 0.1% FA for 106 min, followed by a step to 95% ACN in 0.1% FA over 5 min, hold 10 min, and then a step to 1% ACN in 0.1% FA over 1 min and a final hold for 19 min (total run 156 min); Buffer A = 100% H₂O in 0.1% FA; Buffer B = 80% ACN in 0.1% FA; flow rate, 300 nL/min. All solvents were liquid chromatography mass spectrometry grade. Spectra were acquired using XCalibur, either version 2.1.0 or 4.7.69.37 (Thermo Scientific). A “TopSpeed” data-dependent MS/MS analysis was performed (acquisition of a full scan spectrum followed by collision-induced dissociation mass spectra of the Top N most intense precursor ions within the 3 s cycle time). Dynamic exclusion was enabled with a repeat count of 1, a repeat duration of 30 seconds, an exclusion list size of 500, and an exclusion duration of 40 s.

Phosphoproteomic data analysis

Ion-intensity based label-free quantification was performed using Progenesis QI v2.4 (Nonlinear Dynamics). Raw files were aligned, and an aggregate dataset was created. Only +2, +3, and +4 charged ions were analyzed. MS/MS spectra were exported as.mgf files and searched using Mascot (v2.6, Matrix Science) against the S. cerevisiae S288c UniProt database (6728 entries). Search parameters: 10ppm precursor tolerance, 0.5 Da product tolerance, trypsin digestion with up to 2 missed cleavages; fixed modification: carbamidomethylation (Cys) and variable modification: oxidation (Met) and phosphorylation (Ser, Thr, Tyr); ^13 C = 1; minimum peptide length = 7 amino acids.

Mascot.xml outputs were imported into Progenesis for peptide assignment. Peptides with Mascot ion scores <25 were excluded. Phosphosite positions were determined by mapping peptides to the reference FASTA sequence. Redundant peptide matches were collapsed, and normalized intensities of peptides mapping to the same site were summed.

Protein abundance analysis

Protein quantification and logarithmic transformation was performed in Progenesis using only non-conflicting peptide ions. Protein abundance was calculated by summing the normalized ion intensities of all non-conflicting peptide ions uniquely assigned to a given protein. Peptides shared between multiple proteins were excluded. Precursor ion- abundance values were normalized across all runs using a global scaling factor derived from the total ion intensity of all features.

Significance cutoffs and statistical tests

Phosphoproteomics data were analyzed using ANOVA, with significance thresholds defined by unadjusted p-value and fold-change cutoffs. In most datasets, phosphopeptides with ≥3-fold change and p < 0.01 were considered significant, corresponding to a false discovery rate (FDR) < 2-3% as judged by Benjamini–Hochberg (BH) correction. For the glutamine to proline time course (Fig. 2, Supplementary Data 1), this approach yielded an FDR > 5%; therefore, BH-adjusted p values (≤ 0.05) were used to select significant sites. For experiments with very large numbers of peptides that change (e.g. 4-hour nitrogen starvation, rapamycin, and proline treatments; Fig. 4, Supplementary Data 3), we applied a more stringent cutoff (p < 0.001 unadjusted and >4-fold change; FDR < 1%) to focus on primary signaling events. Global proteomics data (from flowthrough fractions) were analyzed in parallel using ≥3-fold change and p < 0.01, corresponding to BH-adjusted FDR values of ~2% (Supplementary Data 1–4). A summary of all cutoffs, p-values, and FDR estimates is provided in Supplementary Table 4.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

C.P. and J.L. performed the main experiments and prepared the phosphopeptides and bulk peptides for analysis. A.L. and P.L. carried out the mass spectrometry experiments. C.P. and A.C. analyzed the data. C.P. and A.C. wrote the original draft. C.P., J.L., A.L., P.L. and A.C. reviewed and edited the draft.

Peer review

Peer review information

Nature Communications thanks Xinchen Teng and the other anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository¹⁰⁵ with the dataset identifier PXD068726 and 10.6019/PXD068726. All other data from this study are available within the Article and its Supplementary Information files. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-66907-1.