The TORC1 complex is a key regulator of cell growth and metabolism in Saccharomyces cerevisiae. The vacuole-associated EGO complex couples activation of TORC1 to the availability of amino acids, specifically glutamine and leucine. The EGO complex is also essential for reactivation of TORC1 following rapamycin-induced growth arrest and for its distribution on the vacuolar membrane. Pib2, a FYVE-containing phosphatidylinositol 3-phosphate (PI3P)-binding protein, is a newly discovered and poorly characterized activator of TORC1. Here, we show that Pib2 is required for reactivation of TORC1 following rapamycin-induced growth arrest. Pib2 is required for EGO complex-mediated activation of TORC1 by glutamine and leucine as well as for redistribution of Tor1 on the vacuolar membrane. Therefore, Pib2 and the EGO complex cooperate to activate TORC1 and connect phosphoinositide 3-kinase (PI3K) signaling and TORC1 activity.
The target of rapamycin complex I (TORC1) couples multiple nutritional cues to orchestrate an appropriate cellular growth response. Nutrients, in particular amino acids, activate TORC1 signaling, which results in a multi-pronged anabolic response, including ribosome and protein synthesis, increase of biomass and growth. On nutrient starvation, TORC1 is inactivated, which leads to a coordinated starvation response, including amino acid permease synthesis and transport, amino acid biosynthesis and induction of macroautophagy (Broach, 2012; Loewith et al., 2002; Neufeld, 2010).
TORC1 is a multisubunit complex of ∼2 mDa and consists of Tor1 or Tor2, a PIK-like kinase and the accessory subunits Kog1, Lst8 and the non-essential Tco89 (Loewith et al., 2002; Wedaman et al., 2003). TORC1 appears to be constitutively associated with the vacuolar membrane, independently of nutrient status, although some sequestration to peri-vacuolar foci has been observed (Kira et al., 2014; Sturgill et al., 2008). TORC1 exerts its growth effects via several downstream signaling branches that together constitute the anabolic or catabolic response. TORC1 stimulates protein and ribosome synthesis through several downstream effector kinases including Sch9 and Ypk3 (González et al., 2015; Urban et al., 2007). Simultaneously, active TORC1 inhibits the PP2A (Pph3, Pph21 and Pph22) and PP2A-related (Ppg1 and Sit4) phosphatases, whose downstream effects include responses to nitrogen starvation (Loewith and Hall, 2011). Furthermore, TORC1 inhibits macroautophagy (Kamada et al., 2010). In addition to these main effector branches, TORC1 directly interacts with an extensive array of kinases and phosphatases (Breitkreutz et al., 2010). These include Npr1, a kinase involved in regulating trafficking and localization of amino acid permeases (MacGurn et al., 2011; Merhi and Andre, 2012; Schmidt et al., 1998) and Nnk1, which has been implicated in nitrogen metabolism (Breitkreutz et al., 2010).
Amino acids regulate TORC1 via several mechanisms that largely depend on the ‘escape from rapamycin-induced growth arrest’ complex (the EGO complex) (Peli-Gulli et al., 2015). The EGO complex consists of two small GTPases, Gtr1 and Gtr2, which are recruited to the vacuolar membrane by a scaffold subcomplex (Powis et al., 2015) consisting of Meh1 (also known as Ego1), Ego2 and Slm4 (also known as Ego3). The EGO complex is highly conserved, and the Gtrs have homologs in higher eukaryotes known as the Rag GTPases. The Gtrs form a constitutive heterodimer whose activity depends on their nucleotide-binding status: the heterodimer is active when Gtr1 is GTP-bound and Gtr2 is GDP-bound (Binda et al., 2009; Jeong et al., 2012; Nakashima et al., 1999) and inactive in the opposite configuration. The nucleotide status of the Gtrs is regulated by several complexes that impinge on GTP hydrolysis, loading or dissociation: Vam6, a component of the HOPS complex involved in vacuolar fusion, was demonstrated to be a GEF for Gtr1 (Binda et al., 2009); Lst4–Lst7, which is a GTPase-activating complex (GAP) for Gtr2, which results in activation of TORC1 (Peli-Gulli et al., 2015), and the SEA complex, which is a GAP for Gtr1 that inactivates it (Neklesa and Davis, 2009; Panchaud et al., 2013).
Particularly potent activators of TORC1 via the EGO complex are the amino acids leucine and glutamine. Leucine promotes interaction between GTP-loaded Gtr1 (Gtr1GTP) and Meh1 (Binda et al., 2009), and the leucyl tRNA synthetase Cdc60 was shown to directly interact with Gtr1 in a leucine-dependent manner (Bonfils et al., 2012). Glutamine stimulates interaction of the GAP Lst4–Lst7 with Gtr2, thereby promoting formation of Gtr2-GDP, the active form that can activate TORC1 (Peli-Gulli et al., 2015). Active Gtrs stimulate TORC1 via direct physical interactions: Gtr1GTP interacts with Tco89 (Binda et al., 2009), and the active heterodimer itself interacts with Kog1 (Sekiguchi et al., 2014).
In addition to GTPases, TORC1 is also regulated by signaling via the phosphoinositide 3-kinase (PI3K) Vps34 and its product phosphatidylinositol 3-phosphate (PI3P) in both yeast and mammalian cells. In mammalian cells, signaling dependent on Vps34 (also known as PIK3C3) is well characterized: amino acids activate Vps34, which results in an elevation of PI3P levels (Nobukuni et al., 2005), which, in turn, leads to activation of mTORC1 (Byfield et al., 2005; Yoon et al., 2011). Importantly, the Vps34 pathway is also necessary for the activation of mTORC1 by the mammalian homologs of the Gtr GTPases (the Rag GTPases). Hence, amino acids activate mTORC1 via two necessary mutually interdependent pathways: Rag GTPases and Vps34. In yeast, deletion of Vps34 also results in a strong inhibition of TORC1 (Bridges et al., 2012) but the downstream effectors of Vps34 in activation of TORC1 are unknown. It is also currently unknown how Vps34-dependent and Gtr-dependent activation of TORC1 are integrated.
Recent work has identified Pib2 (phosphatidylinositol-3-phosphate-binding 2) as an additional activator of TORC1 (Kim and Cunningham, 2015; Michel et al., 2017; Tanigawa and Maeda, 2017). Pib2 was initially identified, together with several components of the EGO complex, as a hit in a screen for factors unable to recover from rapamycin exposure (Dubouloz et al., 2005). Later, Pib2 was reported to be required for TORC1 activation and lysosomal membrane permeabilization in the presence of ER stress (Kim and Cunningham, 2015). It has a FYVE domain, a conserved C-terminal tail motif and a series of conserved stretches of amino acids in a region otherwise predicted to be unstructured. The N-terminal region harbors a TORC1 inhibitory function whereas the C-terminal region is important for activation of TORC1 (Michel et al., 2017). Pib2 interacts with vacuoles via its FYVE domain in a PI3P-dependent manner and this depends on Vps34 (Kim and Cunningham, 2015). Thus, we hypothesize that Pib2 integrates Vps34 signaling into Gtr-dependent activation of TORC1.
Here, we report that Pib2 indeed genetically interacts with components of the EGO complex and TORC1 signaling. Pib2 deletion phenocopies simultaneous loss of Gtr1 and Gtr2 in TORC1 reactivation after rapamycin exposure, microautophagy and Gtr-dependent relocalization of Tor1 to perivacuolar foci. Furthermore, Pib2 and the Gtrs are reciprocally required for activation of TORC1 by glutamine and leucine. Our data suggest that Pib2 and the EGO complex function in the same molecular pathway that leads to activation of TORC1. Therefore, our findings provide evidence for a role for Pib2, together with the EGO complex, in the reactivation of TORC1, thus offering insight into how PI3P signaling might be coupled with Gtr-dependent activation of TORC1.
PIB2 genetically interacts with components of the EGO complex and TORC1
Recent studies have identified Pib2 as a regulator of TORC1 but the mechanism of Pib2 action remains unclear (Michel et al., 2017; Tanigawa and Maeda, 2017). To identify functional interaction partners of Pib2 at the genomic level, we performed a synthetic dosage lethality (SDL) screen by overexpressing Pib2 in each member of the non-essential yeast gene deletion collection (Giaever and Nislow, 2014). The premise of SDL is that overexpression of a gene of interest, when combined with a mutant of a functional interaction partner, results in a measurable fitness defect, or, in the extreme case, lethality (Kroll et al., 1996). In contrast, overexpression of the same gene of interest in a wild-type background may result in no observable phenotype. SDL has been used to screen the non-essential deletion collection for novel participants in various cellular processes (Measday et al., 2005). We used selective ploidy ablation (SPA) to efficiently introduce the Pib2 overexpression plasmid, or appropriate controls, into each haploid member of the non-essential gene deletion collection (Reid et al., 2011). The result is rapid introduction of overexpression plasmids into haploid members of the deletion collection. We obtained several strong SDL hits (P<0.0001, when the deletion strain overexpressing Pib2 is compared to the same deletion strain expressing an empty vector or overexpressing EGFP), which included Δmeh1 (Δego1) and Δtor1 (Fig. 1A; Table S1). Since Meh1 (Ego1) is a vacuolar membrane anchor for both Gtr1 and Gtr2, these newly uncovered genetic interactions demonstrate that Pib2 is functionally related to the EGO complex. An additional strong hit (P<7.8×10−7) was Δpar32, a component of the PP2A signaling branch downstream of TORC1, as well as the deletion of YDL172C, which overlaps with the coding sequence of PAR32 (Fig. 1A; Table S1). We also identified a set of genes enriched in endosomal structure and function (for example Δvps30, Δvps27 and Δvps28) (Fig. 1A; Table S1). These results are consistent with an enrichment of Pib2 in PI3P-containing endosomal/vacuolar membranes (Burd and Emr, 1998; Kim and Cunningham, 2015). We also identified several hits in genes known to be involved in the regulation of the cell cycle and amino acid biosynthesis. In this work, we pursued further characterization of the connections between Pib2, the EGO complex and TORC1.
Pib2 is required for reactivation of TORC1 after treatment with rapamycin
Pib2 was initially identified as a hit in a screen for cells that were impaired in recovery from rapamycin, together with constituents of the EGO complex (Dubouloz et al., 2005). It has been shown that the EGO complex is required for reactivation of TORC1 after inactivation by rapamycin (Binda et al., 2009) as well as for a poorly understood subclass of autophagy known as microautophagy (Dubouloz et al., 2005). Given that Pib2 genetically interacts with the EGO complex and Tor1, we compared the phenotypes of cells lacking Pib2 with those lacking components of the EGO complex, specifically the Rag family GTPases Gtr1 and Gtr2. Like cells lacking Gtr1 or Gtr2 (Fig. 1B; Fig. S1), Δpib2 cells do not recover from exposure to rapamycin and fail to resume growth after rapamycin-induced growth arrest. By contrast, cells lacking Atg7, which have a defect downstream of TORC1 (cannot undergo macroautophagy; Xie and Klionsky, 2007), recover from exposure to rapamycin like W303A cells (Fig. 1B).
To assess TORC1 activity, we monitored the phosphorylation status of a well-characterized target, ribosomal protein S6 (Rps6). Yeast Rps6 is phosphorylated at two serine residues at its C-terminus (S232 and S233) in a TORC1-dependent manner (González et al., 2015). Hence, the phosphorylation status of Rps6 at these sites can be used as a faithful readout of TORC1 activity. Rapamycin treatment virtually eliminated phosphorylation of Rps6 at these sites in both wild-type and Δpib2 cells, as expected, upon TORC1 inactivation (Fig. 1C). Following recovery from rapamycin exposure, an increase in Rps6 phosphorylation was observed in wild-type cells, to levels comparable to those seen in untreated cells. By contrast, Rps6 remained dephosphorylated at S232 and S233 in Δpib2 cells, even after 24 h of recovery (Fig. 1C,D; recovering to ∼3.5% of the value seen in the wild-type untreated control, P<0.01). This suggests that cells lacking Pib2 fail to reactivate TORC1 during recovery.
To determine whether the growth defect of Δpib2 cells on recovery from rapamycin exposure is due to a defect in Gtr activation, we introduced constitutively active forms of both Gtr1 and Gtr2 (Gtr1 Q65L, constitutively GTP-bound, and Gtr2 S23L, constitutively GDP-bound) (Gao and Kaiser, 2006) into Δpib2 cells. Cells lacking Pib2 could not be rescued by introduction of constitutively active Gtrs (Fig. 1B). As a control, cells lacking Gtr1 or Gtr2 were fully rescued by introduction of active Gtrs (Fig. S1). Therefore, activation of Gtrs is not the underlying cause of the defect in Δpib2 cells. To eliminate the possibility that Pib2 is required for the recruitment of Gtrs to the vacuolar membrane, or that the Gtrs are mislocalized away from the vacuolar membrane in Δpib2 cells and thus cannot activate TORC1, we compared the localization of Gtr1, Gtr2 and Meh1 (Ego1) in wild-type (W303A) and Δpib2 cells. The cellular distribution of Gtr1, Gtr2 and Meh1 (Ego1) was identical between Δpib2 and W303A cells (Fig. S2; data not shown). Hence, Pib2 is not required for vacuolar localization of the Gtrs.
The recovery defect in Δpib2 cells was TORC1 dependent, as introduction of a TOR1 mutant allele (L2134M, within the kinase domain), previously shown to render Tor1 hyperactive regardless of Gtr activation (Kingsbury et al., 2014; Takahara and Maeda, 2012), into Δpib2 cells resulted in recovery and growth similar to that seen in wild-type cells (Fig. 1B). Vector alone controls are provided in Fig. S3A. Since Δpib2 cells could not be rescued by constitutively active Gtrs, the defect in Δpib2 is not due to a defect in activation of Gtrs. This result also suggests that activated Gtrs require Pib2 for activation of TORC1.
Mutants in components of the EGO complex display a striking vacuolar phenotype after exposure to rapamycin: grossly enlarged vacuoles that cannot return to their pre-exposure size after removal of rapamycin (Dubouloz et al., 2005). This was proposed to be due to a defect in microautophagy. We next asked whether Δpib2 cells display a similar vacuolar morphology defect. We evaluated the size of vacuoles in W303A, Δpib2 and Δgtr1 Δgtr2 cells before, during and after rapamycin treatment. On rapamycin exposure, vacuoles of wild-type cells increased in size, as expected, as a consequence of increased macroautophagy (Chan and Marshall, 2014) (Fig. 1E). During recovery, the vacuolar size returned to pre-exposure levels after 48 h. By contrast, vacuoles of Δgtr1 Δgtr2 cells enlarged on rapamycin exposure and did not recover. Vacuoles of Δpib2 cells likewise enlarged on rapamycin treatment but continued expanding, even during recovery from rapamycin exposure, similar to what was seen in Δgtr1 Δgtr2 cells (Fig. 1E). We quantified these observations by calculating the ratio of the maximal vacuolar cross-sectional area to the maximal cellular cross-sectional area (vac:cell area), to normalize to cell size (Fig. 1F). Untreated W303A cells had a vac:cell area ratio of 0.23±0.05 (mean±s.d.), which increased to 0.47±0.11 after rapamycin treatment (P<0.01), before recovering to 0.21±0.03 after 48 h. Δgtr1 Δgtr2 cells had an untreated vac:cell ratio of 0.36±0.12. Rapamycin treatment increased this to 0.54±0.07 (P<0.01), which increased further to 0.70±0.08 after 48 h recovery (P<0.01). Similarly, Δpib2 cells had an untreated vac:cell ratio of 0.30±0.12, which increased to 0.52±0.06 after rapamycin exposure (P<0.01). As was the case for cells lacking Gtrs, this ratio increased to 0.80±0.07 after 48 h recovery (P<0.01). Hence, vacuolar size and the cell:vac scaling ratio does not recover after rapamycin treatment in Δgtr1 Δgtr2 or Δpib2 cells. These results demonstrate that loss of Pib2 phenocopies loss of the Gtrs. Pib2 is therefore, like components of the EGO complex, involved in vacuolar dynamics and microautophagy.
Pib2 and Gtrs are both required for activation of TORC1 by glutamine and leucine
Glutamine and leucine are known to be the most potent activators of TORC1 (Bonfils et al., 2012; Peli-Gulli et al., 2015), and these activating stimuli require the EGO complex for relay to TORC1 (Binda et al., 2009; Kim et al., 2008; Sancak et al., 2008). If Pib2 indeed acts within the same pathway as the Gtrs, we predict that we would observe a defect in stimulation of TORC1 by glutamine and leucine in cells lacking Pib2. We therefore compared TORC1 reactivation by glutamine and leucine in cells lacking either Pib2 or both Gtr1 and Gtr2. When grown in nutrient-rich medium, both Δpib2 and Δgtr1 Δgtr2 double mutant cells exhibit basal TORC1 activity, as determined by assessing the phosphorylation state of Rps6 (Figs 2A,B and 3A,B, left-most column, no significant differences). Nitrogen starvation resulted in loss of detectable TORC1 activity, as expected (P<0.01 in all cases). Addition of either glutamine (3 mM) or leucine (3 mM) for the indicated times (Figs 2A and 3A) evoked reactivation of TORC1 in wild-type cells (5 min, P<0.01; 30 min, P<0.05) but not in Δpib2 or Δgtr1 Δgtr2 cells. Importantly, expressing activated Gtrs in Δpib2 cells did not rescue the glutamine- or leucine-dependent activation of TORC1 (Figs 2C,D and 3C,D). Pib2 is therefore not required for activation of Gtrs. Active Gtrs cannot overcome the requirement for Pib2 in activation of TORC1. These observations are quantified in Figs 2B,D,F and 3B,D,F, for glutamine and leucine, respectively. We, therefore, conclude that Pib2 and the Gtrs are both required to relay the glutamine and leucine signals to TORC1.
Of note, refeeding nitrogen-starved Δpib2 or Δgtr1 Δgtr2 cells with a mixture of all amino acids results in a robust and full phosphorylation of Rps6 and thus activation of TORC1 (Fig. S3D). The degree of phosphorylation of Rps6 was comparable in each strain and directly comparable to that in W303A cells. This suggests the existence of an additional amino acid signal that stimulates TORC1 in a Gtr1/2- and/or Pib2-independent manner. This serves as a positive control for our readout that demonstrates that the extent of the potential response in Δpib2 or Δgtr1 Δgtr2 cells is comparable to the response in W303A cells when the stimulus is not glutamine or leucine. Therefore, the defect in TORC1 activation in Δpib2 or Δgtr1 Δgtr2 cells is stimulus specific and the activation of TORC1 by leucine and glutamine is dependent on both Pib2 and Gtr1/2.
Expression of activated Gtrs (Gtr1 Q65L and Gtr2 S23L) in Δgtr1 Δgtr2 cells resulted in sustained activity of TORC1, even under starvation conditions, whereas TORC1 remains inhibited by nitrogen starvation in wild-type cells overexpressing active Gtrs (Figs 2C,D and 3C,D, compare W303A with overexpressed Gtrs to Δgtr1 Δgtr2 with overexpressed Gtrs, P<0.01). Since wild-type cells still express endogenous Gtr1 and Gtr2, we conclude that inactive forms of Gtrs (i.e. Gtr1-GDP and Gtr2-GTP) are therefore required for inhibition of TORC1 by nitrogen starvation, as previously observed (Kira et al., 2014).
To further confirm the interdependence of Pib2 and Gtrs in TORC1 activation by glutamine and leucine, we also evaluated the effect of overexpression of Pib2 in Δgtr1 Δgtr2 cells. Overexpression of Pib2 in Δgtr1 Δgtr2 cells did not rescue TORC1 activity, whereas it rescued TORC1 activity in Δpib2 cells (Figs 2E,F and 3E,F; 5 min, P<0.05). These observations again suggest that Pib2-dependent TORC1 activation by glutamine or leucine requires Gtrs. Pib2 overexpression in Δgtr1 Δgtr2 cells repressed even TORC1 basal activity (P<0.01 for both Pib2 overexpressed in W303A versus Δgtr1 Δgtr2, and Pib2 overexpressed in Δpib2 versus Δgtr1 Δgtr2), confirming the existence of a previously reported Gtr-independent inhibitory function of Pib2 on TORC1 (Michel et al., 2017). Repression of TORC1 basal activity is only observed in cells lacking Gtrs and not wild-type cells. We further examined the effects of overexpression of a truncated Pib2 construct lacking its N-terminal 164 amino acids (Pib2 ΔN-term) on TORC1 activation. The N-terminal 164 amino acids of Pib2 were previously reported to harbor an inhibitory function on TORC1 (Michel et al., 2017). Indeed, the Pib2 ΔN-term did not inhibit basal activity of TORC1 in Δgtr1 Δgtr2 cells, confirming the importance of this domain for the observed inhibitory function of Pib2 (Fig. S3C). Taken together, these data strongly suggest a novel dual mode of action of Pib2 on TORC1 activity: in the presence of Gtrs, Pib2 is an activator of TORC1, whereas in their absence it is an inhibitor.
Pib2 regulates Tor1 localization on the vacuolar membrane
Previously, it has been reported that the nucleotide state of Gtr1 affects localization of Tor1 at the vacuolar membrane. Gtr1-GTP appears to be required for dispersion of Tor1 throughout the vacuolar membrane; in its absence, Tor1 accumulates in perivacuolar foci (Kira et al., 2016). Tor1 also localizes to perivacuolar foci in the absence of both Gtrs (Fig. 4A) (Kira et al., 2016). The identity of the puncta remains unknown – previous work has demonstrated that they do not colocalize with Snf7 or Ape1, and, hence, are not endosomal or phagophore assembly sites, respectively (Kira et al., 2014). As our data suggest that Pib2 is required to relay a signal from activated Gtrs to TORC1, we sought to evaluate the role of Pib2 in the localization of Tor1. Our prediction was that Tor1 will redistribute to puncta in Δpib2 cells if Pib2 indeed relays signals from activated Gtrs.
In W303A cells grown in nutrient-rich medium, GFP–Tor1, expressed under control of its native promoter from a centromeric plasmid, had a diffuse vacuolar membrane distribution with some foci associated with the vacuolar membrane (Fig. 4A), as has been observed previously with an integrated genomic copy of GFP–Tor1 (Kira et al., 2014). Simultaneous loss of Gtr1 and Gtr2 resulted in a marked redistribution of GFP–Tor1 into puncta associated with the vacuole: the number of vacuoles with Tor1 puncta increased from 18.6±2.9% (mean±s.d.) in W303A cells to 65.6±5.0% in Δgtr1 Δgtr2 cells (P<0.01). Similarly, loss of Tco89, a component of TORC1 required for relay of the Gtr signal (Reinke et al., 2004), resulted in a redistribution of GFP–Tor1 into the vacuole-associated puncta (63.3±2.4% of vacuoles were associated with puncta, P<0.01 compared to the result for W303A cells). Loss of Pib2 also resulted in an increase in vacuoles associated with GFP–Tor1 puncta (41.5±3.4% of vacuoles associated with puncta; P<0.01) (Fig. 4A,B). Of note, expressing constitutively active forms of Gtr1 and Gtr2 in Δpib2 cells did not change the number of vacuoles associated with Tor1 foci (Fig. S3D). This observation may be explained by two scenarios: either Pib2 and Gtr1/Gtr2 act independently to regulate Tor1 localization, or Pib2 acts directly downstream of the Gtrs in regulating Tor1 localization. Further studies are required to distinguish between these possibilities. Currently, the function of Tor1 foci formation is unknown. To determine whether Tor1 foci formation impinges on TORC1 activity, we analyzed foci formation after nitrogen starvation, when TORC1 activity is known to be repressed. No significant changes in foci formation were observed in W303A, Δpib2 and Δgtr1 Δgtr2 cells (Fig. S4A,B; compare Fig. 4B and Fig. S4B). This suggests that Tor1 foci formation does not correlate with the activity of TORC1. Strikingly, exposure to rapamycin for 3 h, which is also known to inhibit TORC1 activity, resulted in a complete loss of Tor1 foci in all strains, even Δgtr1 Δgtr2 cells (Fig. S4C,D). A mechanistic explanation of this observation awaits further experimentation.
Pib2 has been reported to directly interact with Tor1 and Kog1 (Michel et al., 2017; Tanigawa and Maeda, 2017). We asked, therefore, whether Pib2 changes its localization in response to loss of the Gtrs, as observed for components of TORC1. Indeed, in W303A cells, Pib2 is associated with the vacuolar membrane with some foci. In the absence of Gtrs (Fig. 4C) or Tco89 (data not shown), Pib2 distribution alters with an increased number of vacuoles containing foci (Fig. 4D; in W303A cells 15.8±4.7% of vacuoles had foci compared to 47.5±7.4% for vacuoles in Δgtr1 Δgtr2 cells, P<0.001). These data indicate that Pib2 is likely to follow the Gtr-dependent distribution of TORC1.
Pib2 is not required for and does not regulate macroautophagy
PI3P is required for macroautophagy, and removing Vps34, which is the sole PI3K in yeast, results in inhibition of macroautophagy (Burman and Ktistakis, 2010). Since Pib2 is a PI3P-binding protein and since its recruitment is Vps34 dependent (Kim and Cunningham, 2015), we sought to determine whether Pib2 was an effector of PI3P in regulating autophagy. We therefore used the well-established GFP–Atg8 processing and flux assay in both W303A and Δpib2 cells expressing GFP–Atg8 from its native promoter (Kirisako et al., 1999; Shintani and Klionsky, 2004). Basal GFP–Atg8 expression levels were directly comparable in W303A and Δpib2 cells (Fig. S5). On rapamycin exposure, similar increased expression levels of GFP–Atg8 were observed in both the W303A and Δpib2 cells, a consequence of enhanced microautophagic flux (Fig. S5). Likewise, comparable elevated amounts of free GFP, reflecting processed GFP–Atg8, were observed in both W303A and Δpib2 cells (Fig. S5). Hence, Pib2 is not required for GFP–Atg expression or processing, and cells lacking Pib2 are not impaired in macroautophagy.
Npr1 is constitutively active in Δpib2 cells
TORC1 directly interacts with, and phosphorylates, Npr1, which inhibits it (Breitkreutz et al., 2010; MacGurn et al., 2011; Schmidt et al., 1998). Inhibition of TORC1 activity, through rapamycin treatment or nitrogen starvation, therefore leads to activation of Npr1 that results in a number of downstream effects, including inhibition of Ldb19 (Art1) (MacGurn et al., 2011), phosphorylation of Bul1 and Bul2 (Merhi and Andre, 2012) and trafficking of the tryptophan permease Tat2 from the surface of the cell to the vacuole for degradation (Schmidt et al., 1998). One additional target of active Npr1 is the poorly characterized protein Par32. Active Npr1 results in extensive phosphorylation of Par32 at multiple sites, which leads to a significant change in migration rate (Boeckstaens et al., 2015). Therefore, the migration rate of Par32 can be used as a readout to evaluate the activity of Npr1.
Δpar32 was a hit in our SDL screen using overexpressed Pib2 (P<7.82×10−7). We, therefore, evaluated phosphorylation of Par32 as a readout of Npr1 activity in W303A and Δpib2 cells. As expected, we observe that all of the hemagglutinin-tagged Par32 (Par32–3xHA) expressed in W303A cells from the native PAR32 promoter is shifted to a slower-migrating form on rapamycin treatment or nitrogen starvation, which would be consistent with extensive post-translational modification (Fig. 5A–C). Essentially this shift in migration depends on the presence of Npr1. In Δpib2 cells, the steady-state distribution of Par32 is more shifted towards the slower-migrating species than in W303A cells, indicating that Npr1 is more active than in controls. Treatment with rapamycin maximally shifted Par32–3xHA in Δpib2 cells, indicating further activation of Npr1 (Fig. 5C). All of the shifts in Par32–3xHA migration, in W303A or in Δpib2 cells, were dependent on the presence of Npr1 (Fig. 5C; data not shown). In summary, at steady state, Npr1 is partially active in Δpib2 cells, but not in W303A cells, and can be further activated by additional inhibition of TORC1. The increased phosphorylation of Par32 observed in Δpib2 cells could not be completely reversed by expression of the hyperactive mutant allele of Tor1 (L2134M) (Fig. 5D). Note that in both W303A cells and Δpib2 cells, the extent of enhancement in TORC1 activity on expression of Tor1 L2134M is directly comparable. Thus, Pib2 has an additional function of repressing Npr1 activity independently of TORC1.
Deletion of Npr1 has been reported to suppress the defect in exit from the rapamycin-induced growth arrest of various EGO mutants, including Meh1 (Ego1), Slm4 (Ego3) and Gtr2 (Dubouloz et al., 2005). We therefore tested whether deletion of Npr1 also suppresses the defect in recovery from rapamycin of Δpib2 cells. Δnpr1 cells displayed enhanced growth compared to W303A cells on recovery from rapamycin (Fig. 5E). As before, Δpib2 cells did not recover from treatment with rapamycin. However, simultaneous deletion of Pib2 and Npr1 resulted in recovery from rapamycin (Fig. 5E). Hence, activated Npr1 after rapamycin exposure contributes to the lack of growth of Δpib2 cells, in the same way as was previously observed in Δmeh1, Δslm4 and Δgtr2 cells. Taken together, these findings indicate that Pib2 has a function in downregulation of Npr1 activity, which negatively affects recovery of growth after rapamycin treatment (Fig. 6).
In this work, we provide a detailed characterization of Pib2 and a comparison of its function to that of the EGO complex. We demonstrate that Pib2, whose mechanism of action was ill defined, is required, together with the Gtrs, for activation of TORC1. We identified strong genetic interactions in an SDL screen between Pib2 and components of the EGO complex–TORC1 network. Δpib2 cells behaved identically to cells lacking both Gtrs in many aspects, including recovery from exposure to rapamycin, vacuolar dynamics, response to amino acids and distribution of GFP-Tor1 on the vacuolar surface. We, therefore, conclude that these responses of TORC1 require both Pib2 and the EGO complex (Fig. 6).
Previous reports demonstrated a reduced response to glutamine in cells lacking Pib2 (Michel et al., 2017; Tanigawa and Maeda, 2017). Our data showed that cells lacking Pib2 are unable to activate TORC1 in response to glutamine or leucine. Importantly, the presence of mutants of Gtr1 and Gtr2 that are restricted to activated states did not override the requirement for Pib2, suggesting that the role of Pib2 is not activation of Gtrs. In mammalian cells, leucine is sensed by leucyl tRNA-synthetase (LRS, also known as LARS), which activates mTORC1 via two mutually necessary mechanisms: LRS has GAP activity for RagD (also known as RRAGD, a mammalian homolog of Gtr2) (Han et al., 2012) and LRS directly interacts with and activates Vps34, thus mediating TORC1 activation via the Vps34–PLD1 branch (Yoon et al., 2016). Thus, leucine-dependent activation of TORC1 integrates PI3P- and Rag-dependent signaling pathways. The yeast homolog of LRS, Cdc60, has been reported to regulate the activities of the Gtrs in response to amino acids (Bonfils et al., 2012). However, a connection between PI3P signaling and leucine has not yet been established. We speculate that Pib2 is an integral part of the PI3P signaling pathway that connects leucine stimulation to TORC1 activation.
Overexpression of Pib2 in Δgtr1 Δgtr2 cells did not rescue the response to glutamine or leucine, further highlighting the co-dependence of Pib2 and the Gtrs in activation of TORC1. Previous models of Pib2 function suggested a Gtr-independent role in activation of TORC1 (Kim and Cunningham, 2015; Tanigawa and Maeda, 2017; Stracka et al., 2014), based on the observations of knockouts of Gtr1 alone and synthetic lethality between PIB2 and components of the EGO complex. In these reports, residual activation of TORC1 was detected in Δgtr1 cells. This residual activity was attributed to Pib2, since it is a known activator of TORC1. Based on the fact that we do not detect residual TORC1 activation in Δgtr1 Δgtr2 cells, it may be that the residual activation detected in the single knockout stems from the action of the remaining component of the Gtr dimer. Gtr1 and Gtr2, and their Rag homologs in higher eukaryotes, form heterodimers that, when asymmetrically loaded with GTP and GDP respectively, activate TORC1 (Hatakeyama and De Virgilio, 2016). It is possible that in Δgtr1 cells the presence of Gtr2, combined with endogenous Pib2, and/or the absence of Gtr1-GDP, could have a residual activity on TORC1. It is known that Gtr1 can form homodimers (Nakashima et al., 1999) and it would be interesting to see whether the same is true of Gtr2, especially in cells lacking Gtr1.
If the pathways mediated by Pib2 and Gtr1/2 to activate TORC1 are independent of each other, there should be an intermediate TORC1 response to glutamine or leucine in cells lacking either Pib2 or Gtr1/2. Under our conditions, we do not observe this and we observe activation only in the presence of both Pib2 and Gtr1/Gtr2. One possibility is that TORC1 is generally impaired in either Δpib2 or Δgtr1 Δgtr2 cells, which may dampen an intermediate TORC1 activation response below detection thresholds. Our data suggests otherwise for two reasons. First, basal TORC1 activity is not impaired in either Δpib2 or Δgtr1 Δgtr2 cells (Figs 2, 3 and 5D). Second, we show that Δpib2 or Δgtr1 Δgtr2 cells can activate TORC1 to the same extent as wild-type cells (using a different, Pib2- and Gtr1/2-independent stimulus as reported in Fig. S3B). This serves as a positive control for our readout that demonstrates that the extent of the potential response in Δpib2 or Δgtr1 Δgtr2 cells is comparable to the response of the wild-type (W303A) cells when the stimulus is different. This argues against a generally reduced/impaired TORC1 activity in the knockout strains, either at the basal level or on activation by different stimuli.
The mechanism of TORC1 activation by Pib2 and the Gtr1/2 may be explained by two overarching models: dependent and independent (Fig. S6). The prediction for the independent model (model C in Fig. S6) is that intermediate levels of TORC1 activation would be detected. Model A (dependent hierarchical) postulates that activation of TORC1 requires both Pib2 and the Gtrs that act in some hierarchical manner (upstream/downstream of each other). A prediction of this model is that no intermediate activation of TORC1 will be detected when Pib2 or the Gtrs are missing. A small modification of model A is model B (dependent threshold). In this case, activation of TORC1 depends, again, on both Pib2 and the Gtrs. However, the extent of activation by either Pib2 alone or the Gtrs alone either does not exist or is so low that it cannot be detected by multiple assays. However, a potentiation occurs between Pib2 and the Gtrs, which would result in a full response, and this potentiation implies dependence. Based on our results of activation of TORC1 by glutamine or leucine after nitrogen starvation, we suggest that models A or B are most plausible. Model C might be supported by the observed synthetic lethality between PIB2 and components of the EGO complex. However, synthetic lethality is not necessarily inconsistent with a dependent mechanism of action of Pib2 and Gtr1/2 on TORC1 activation (models A and B). Here, we report that Pib2 has an additional TORC1-independent inhibitory function on Npr1 (Fig. 5). This could provide an alternative explanation for the observed synthetic lethality between PIB2 and components of the EGO complex. Cells lacking both Pib2 and components of the EGO complex will have constitutive Npr1 activity that is toxic (Schmidt et al., 1998). Taken together, the lack of an intermediate response to glutamine and leucine in Δpib2 or Δgtr1 Δgtr2 cells, as well as our detection of an additional function of Pib2 in regulating Npr1, which is an alternative explanation for synthetic lethality, favors a dependent model of Pib2 action on TORC1 (models A or B).
Overexpression of Pib2 in Δgtr1 Δgtr2 cells not only failed to rescue TORC1 activity in response to amino acids but also significantly dampened the basal response of TORC1, suggesting that Pib2 has an additional inhibitory function on TORC1 that is unmasked in the absence of the Gtrs. This inhibitory function is Gtr independent. This supports previous work that identified an inhibitory region at the N-terminus of Pib2 (Michel et al., 2017). Taken together, our observations suggest that Pib2 has two antagonistic functions: activation of TORC1 in a manner that is co-dependent on the Gtrs, and Gtr-independent inhibition of TORC1. Intriguingly, mammalian cells have two PI3P-binding homologs of Pib2 that are yet to be implicated in the regulation of mTORC1: Phafin-1 and Phafin-2 (also known as PLEKHF1 and PLEKHF2, respectively). These both lack the N-terminal supposedly inhibitory regions present in Pib2 and instead have a PH domain. Currently, a link between the Phafins and mTORC1 has not yet been established and thus it is of immediate interest to determine whether indeed Phafins play a role in mTORC1 signaling and, if so, how their mechanism of action differs from that of Pib2.
Although the vacuolar localization of Tor1 in yeast is independent of the nutritional status of the cell, the distribution of the TORC1 complex is dynamically regulated by the nucleotide-bound state of Gtr1 and Gtr2 (Kira et al., 2016). Absence of the active form of Gtr1 or of the Gtrs altogether leads to the accumulation of Tor1 in perivacuolar foci. In Δpib2 cells, we observe a similar accumulation of Tor1 in foci, suggesting that Pib2 also plays a role in Tor1 localization on the vacuolar membrane. In Δgtr1 Δgtr2 cells, Pib2 similarly accumulates in perivacuolar foci. Pib2 physically interacts with Tor1 and Kog1 (Michel et al., 2017; Tanigawa and Maeda, 2017). Pib2 likely therefore associates with TORC1 and follows its distribution in response to signaling via Gtrs.
We observed that cells lacking Pib2 have partially activated Npr1, as assessed by monitoring the phosphorylation status of the direct Npr1 effector Par32. We also observed that Pib2 inactivates Npr1 in parallel to TORC1. Furthermore, loss of Npr1 resulted in growth resumption in cells lacking Pib2 after rapamycin exposure, as has been previously observed for cells lacking components of the EGO complex (Dubouloz et al., 2005). Thus, loss of Npr1 overrides the requirement for Pib2 or the EGO complex in reactivation of TORC1 after rapamycin exposure. This suggests that sustained Npr1 activity during recovery from rapamycin makes reactivation of TORC1 completely dependent on the EGO complex and Pib2. One potential mechanism for this is that Npr1 directly phosphorylates TORC1 components or regulators, preventing activation by all other activators except for activated Gtrs and Pib2. In this context, it is of interest that Npr1 interacts with multiple components of TORC1 (Breitkreutz et al., 2010). Alternatively, the mechanism for Npr1-mediated suppression of the phenotypes of loss of Pib2 and EGO components could be more complex and indirect. Npr1 is a known regulator of the stability, localization and transport of several permeases, including the tryptophan permease Tat2 (Schmidt et al., 1998), the arginine and uracil transporters Can1 and Fur4 (MacGurn et al., 2011), and the general amino acid permease Gap1 (Merhi and Andre, 2012; O'Donnell et al., 2010; Shimobayashi et al., 2013). Hence, Npr1 may regulate the stability or activity of a permease that supplies an amino acid or other nutrient that is capable of activating TORC1 in an EGO complex- and Pib2-independent manner after rapamycin treatment.
In summary, we establish a function for Pib2, a FYVE domain-containing PI3P-binding protein, in Gtr-dependent activation of TORC1, identifying a molecular bridge between PI3P signaling and the EGO complex. Future work will focus on the conservation of function for the Pib2 homologs in mammalian cells.
MATERIALS AND METHODS
Yeast genetic manipulation and molecular biology
Strains used in this work are listed in Table S2. Gene deletions were generated in W303a/α diploids by homologous recombination and complete replacement of the target open reading frame with cassettes amplified from pFA6a-kanMX6, pFA6a-His3MX6 (Longtine et al., 1998) or pFA6-natMX4 (Goldstein and McCusker, 1999) flanked with sequence (30 nt) proximal to the coding sequence of the target gene. Diploids were subsequently sporulated through starvation in SPO medium. Following manual tetrad dissection, knockout haploids were validated by colony PCR, microscopy and, in some cases, sequencing. Strains harboring more than one genomic modification were generated by mating and sporulation of appropriate parental strains, followed by extensive revalidation. The standard PEG 3350/lithium acetate/single-stranded carrier DNA protocol was used for yeast transformation (Gietz and Schiestl, 2007).
YPD (2% yeast extract, 1% peptone, 2% glucose, supplemented with L-tryptophan and adenine) was used for routine growth. Synthetic Complete (SC; yeast nitrogen base, ammonium sulfate, 2% glucose, amino acids) or Synthetic Defined (SD; yeast nitrogen base, ammonium sulfate, 2% glucose, appropriate amino acid dropout) media were used prior to microscopy or to maintain plasmid selection as indicated. For sporulation, cells were successively cultured in YPA (2% potassium acetate, 2% peptone, 1% yeast extract) and SPO (1% potassium acetate, 0.1% yeast extract, 0.05% glucose). For starvation, cells were grown in SD –N (0.17% yeast nitrogen base without amino acids and ammonium sulfate, 2% glucose). For stimulation, cells were treated with SD –N supplemented with glutamine (Gln, 3 mM) or leucine (Leu, 3 mM), or supplemented with a complete dropout mix, and were incubated for the indicated times prior to lysis and processing.
Cloning and plasmids
Plasmids used in this work are listed in Table S3. GFP-S cer. PIB2 was generated by amplifying the PIB2 promoter and a fragment containing the PIB2 coding sequence and terminator from genomic DNA, prepared from W303a/α diploids by using a yeast DNA extraction kit (Thermo Fisher Scientific, Pittsburgh) and appropriate primers. The fragments were assembled with an additional fragment encoding EGFP by an overlap extension PCR. The resulting construct was introduced into pRS316, previously linearized with SacI and ClaI, by Gibson assembly. S cer. PAR32-3xHA and GFP-TOR1 were amplified from genomic DNA and were cloned through a similar approach.
GTR1 Q65L, GTR2 S23L, and TOR1 L2134M, with their respective promoters and terminators, were cloned by overlap extension and Gibson assembly after amplification from W303a/α genomic DNA. All point mutants described in this work were constructed by overlap extension PCR at the site of the mutation using appropriate primers followed by Gibson assembly into the linearized target vector. All primer sequences used in this work are available on request.
Dosage lethality screening
Selective ploidy ablation was used to introduce a control or Pib2 overexpression plasmid into each strain in the non-essential haploid deletion collection (Thermo Fisher Scientific) (Reid et al., 2011). In brief, the plasmid of interest (PGAL1-S cer. PIB2 or the control PGAL1) is introduced into a universal donor strain (UDS), where all chromosomes are conditionally unstable, by standard transformation. Each chromosome in the UDS has both a galactose-inducible promoter and a URA3 counter-selectable marker adjacent to its centromere. The UDS containing the plasmid of interest is mated to each member of the non-essential deletion collection. UDS chromosomes are subsequently eliminated from the diploids by centromere destabilization followed by counter selection (Reid et al., 2011). Destabilization and Pib2 overexpression are simultaneously induced by switching to galactose as a carbon source. After induction of Pib2 overexpression, colony sizes are measured and compared to those in the same strain containing either of two control plasmids (PGAL1, containing only the galatose promoter, or PGAL1-EGFP) and subjected to further analysis.
Yeast colony manipulations were performed by using a BM3 colony processing robot (S&P Robotics Inc., Toronto). The non-essential haploid deletion collection was reformatted into a density of 4×384 colonies per plate, as 32×48 grids, such that each member of the deletion array was present as a tetrad of four colonies. The MATα UDS, containing the plasmid of interest, was pinned into grids of 32×48 colonies per plate on SC –LEU, followed by overnight growth at 30°C. UDS colonies were pinned onto deletion array colonies, followed by 24 h incubation at 30°C, to allow thorough mating.
The diploids were repinned onto SC –LEU+galactose to induce overexpression of Pib2, or the EGFP control, and simultaneous destabilization of the UDS chromosomes. After ∼48 h, the colonies were repinned onto SC –LEU+galactose+5-fluoro-orotic acid (5-FOA, Toronto Research Chemicals Inc., Toronto) and incubated at 30°C. After 72 h, the colonies were repinned onto SC –LEU+galactose+5-FOA (Toronto Research Chemicals Inc., Toronto), followed by an additional ∼72 h incubation at 30°C, prior to colony size measurement.
SDL data analysis
Colony sizes from high-resolution photographs of plates were measured by using SGAtools (Wagih et al., 2013). Colony size data were then visualized using the web interface of the ‘Data Review Engine’ in ScreenMill (Dittmar et al., 2010), to enable manual checking of colonies flagged for attention due to potential pinning errors or those colonies within individual 2×2 arrays that may be suspect. The ‘Data Review Engine’ was also used to normalize colony sizes to the plate median for every plate analyzed, to allow direct comparison of colony sizes between control and experimental plates. Subsequently, the normalized growth values were used to calculate Z-scores and P-values for each member of the deletion collection overexpressing either Pib2 or containing the control plasmid. The results were then analyzed using the ‘Statistics Visualization Engine’ of ScreenMill. All experimental strains with a growth difference compared to the control strains with an implied P-value of <0.0001 were examined further.
Analysis of growth by serial dilution
Following overnight growth in YPD, target cells were diluted and regrown to mid-logarithmic phase in YPD at 30°C [optical density at 600 nm (OD600) of 0.6–0.8]. Cells were then diluted to 0.5 OD600/ml and 1:5 serial dilutions were made in water. 2 µl of each dilution was spotted onto YPD or YPD+2.5 ng/ml rapamycin plates. Where relevant, cells were incubated for the indicated times with YPD supplemented with 200 ng/ml rapamycin at 30°C. After extensive washing, cells were resuspended in fresh YPD and allowed to recover at 30°C for the indicated time prior to plating on YPD. Plates were then incubated at 30°C for 3 days prior to imaging.
Preparation of yeast for microscopy
Cells were grown overnight in YPD or SD medium appropriately supplemented to maintain plasmid selection. Cells were then diluted in YPD and grown to mid-logarithmic phase. Vacuolar membranes were stained with 10 µM FM 4-64 (Thermo Fisher Scientific) for 45 min, followed by washing and incubation in YPD medium without dye for 1 h. For rapamycin treatment, cells in YPD were treated for the indicated time with a final concentration of 200 ng/ml rapamycin (Thermo Fisher Scientific). For recovery from rapamycin exposure, cells were extensively washed and resuspended in fresh YPD and incubated as indicated. Cells were plated onto No. 1.5 glass-bottomed coverdishes (MatTek Corporation, Ashland) previously treated with 15 µl 2 mg/ml concanavalin-A (Sigma-Aldrich).
Protein extracts for western blotting were obtained as described previously (Millen et al., 2009). Briefly, cells were lysed on ice by resuspension in 1 ml ice-cold H2O supplemented with 150 µl 1.85 M NaOH and 7.5% (v/v) β-mercaptoethanol. Protein was precipitated by addition of 150 µl 50% (w/v) trichloracetic acid. Pellets were washed twice with acetone, resuspended in 150 µl 1× SDS-PAGE buffer and incubated for 30 min at 30°C followed by 2 min at 95°C. Antibodies used were as follows: anti-Rps6 (1:1000, ab40820, Abcam, Cambridge), anti-PGK1 (1:1000, ab113687, Abcam), anti-EGFP (1:1000, ab290, Abcam), anti-phospho-Rps6 (1:1000, 4858, Cell Signaling Technology, Danvers) and anti-HA (1:1000, ab9110, Abcam) antibodies. Labeled secondary antibodies were IRDye 680RD-conjugated goat anti-rabbit-IgG antibody (926-68171, Li-Cor, Lincoln) and IRDye 680RD-conjugated goat anti-mouse-IgG (926-68070, Li-Cor). These were detected using the Odyssey system (Li-Cor). Bands were integrated and quantified using the Fiji distribution of ImageJ (Schindelin et al., 2012).
Confocal microscopy and image analysis
Confocal images were acquired on a Nikon (Melville, NY) A1 confocal microscope, with a 100× Plan Apo 100× oil objective. NIS Elements Imaging software was used to control acquisition. Images were further processed using Fiji or NIS Elements software.
The authors would like to thank Suzanne Hoppins and Jeff Brodsky for extensive discussion, and John Dittmar for assistance with ScreenMill.
Conceptualization: N.V.V., M.G.J.F.; Methodology: M.G.J.F.; Validation: N.V.V., M.G.J.F.; Formal analysis: N.V.V., M.G.J.F.; Investigation: N.V.V., M.J.M., M.G.J.F.; Resources: K.A.B., M.G.J.F.; Writing - original draft: N.V.V., M.G.J.F.; Visualization: N.V.V.; Supervision: M.G.J.F.; Project administration: M.G.J.F.; Funding acquisition: K.A.B., M.G.J.F.
This work was supported by the National Institutes of Health [grants GM120102 (M.G.J.F) and ES024872 (K.A.B.)]. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.