Proteasome granule formation is regulated through mitochondrial respiration and kinase signaling

The efficiency of many cellular processes relies on optimal levels of the proteins involved. To achieve this, there is a network of factors that coordinate protein synthesis, folding, localization and degradation, known as the proteostasis network. Protein degradation is mainly controlled by two pathways in eukaryotes, the ubiquitin-proteasome system (UPS) and the autophagy-lysosome pathway (Finley et al., 2012; Feng et al., 2014; Dikic, 2017). Substrates of the UPS are labeled with one or more ubiquitins in a highly selective process (Hershko and Ciechanover, 1998; Schrader et al., 2009; Finley et al., 2012). In humans, more than 600 ubiquitin ligases are dedicated to the recognition and labeling of specific proteasome substrates (Li et al., 2008). Although many ubiquitinated proteins are recognized and degraded by proteasomes, ubiquitination also serves other functions depending on the type of ubiquitination and cellular state (Fujii et al., 2009; Shaid et al., 2013; Rittinger and Ikeda, 2017). In the autophagy-lysosome system, macro-autophagy, hereafter referred to as autophagy, is the major pathway. Autophagy (literally, self-eating) involves the targeting of proteins to the concave side of de novo-formed double-membrane structures. Upon expansion and closure of these structures, an autophagosome that has engulfed cytosolic cargo is formed (Kabeya et al., 2000; Feng et al., 2014; Dikic, 2017). Fusion of the outer membrane of these autophagosomes with lysosomes (vacuoles in yeast and plants) exposes the inner membrane and engulfed material to acidic hydrolases that degrade the content. Autophagy can be divided into selective and non-selective processes. These are differentially regulated depending on the substrate and cellular conditions. Although there is overlap between proteasome and autophagy substrates, autophagy is uniquely capable of clearing protein aggregates, damaged organelles, viruses and large multi-subunit complexes (Ogata et al., 2006; Kraft et al., 2009; Kamada et al., 2010; Deffieu et al., 2013; Marshall et al., 2015; Mochida et al., 2015; Waite et al., 2016; Bao et al., 2016).

Proteasomes are one of the large multi-subunit complexes that are degraded via autophagy, a process referred to as proteaphagy. Proteaphagy appears to be a selective process as it depends on factors not involved in general autophagy (Waite et al., 2016; Cohen-Kaplan et al., 2016; Marshall et al., 2016; Nemec et al., 2017; Li et al., 2019) and does not occur under several conditions that induce general autophagy (Waite et al., 2022). A striking example of this is carbon starvation, which induces general autophagy (Adachi et al., 2017). Proteasomes, however, localize to cytosolic punctate structures termed proteasome storage granules (PSGs) upon carbon starvation (Laporte et al., 2008). Although several factors that regulate PSG formation or their subsequent dissolution have been identified (Peters et al., 2013; Weberruss et al., 2013; van Deventer et al., 2014; Marshall and Vierstra, 2018; Li et al., 2019), the signals that trigger proteasome relocalization and the mechanisms that regulate it have not been resolved.

Grown in the presence of glucose, the yeast Saccharomyces cerevisiae mainly utilize glycolysis for energy (ATP) and biomass production, while suppressing mitochondrial respiration (Kayikci and Nielsen, 2015). However, upon depletion of glucose, cells adapt and the metabolism switches from mainly glycolysis to mitochondrial respiration. This process is facilitated by autophagy, which provides a source of serine required for mitochondrial one-carbon metabolism (May et al., 2020). Although uniquely adapted to utilize glucose as a carbon source, yeast grown in the presence of other carbon sources, such as raffinose or glycerol, primarily utilize mitochondrial respiration for energy production (Kayikci and Nielsen, 2015; Adachi et al., 2017). Interestingly, yeast do not induce general autophagy when switched from glucose-containing media to carbon starvation media, presumably due to the lack of ATP. However, yeast grown on non-glucose carbon sources prior to starvation do induce general autophagy upon removal of the carbon source (Adachi et al., 2017). This suggests that the cellular respiratory state and ATP levels are determinants in signaling autophagy. However, it should be noted that the difference between glycolytic growth with repressed respiration and growth with active mitochondrial respiration is also accompanied by differences in the expression of numerous genes (Galdieri et al., 2010).

As proteasomes and autophagy both contribute to the replenishment of cellular metabolites, we hypothesized that the metabolic state of cells, like with autophagy, impacts proteasomes as well. Here, we report that the relocalization of proteasomes to granules is restricted when yeast are starved following respiratory growth, but not glycolytic growth (during which respiration is suppressed). Consistent with this, conditions that interfered with mitochondrial function, such as inhibition of different electron transport chain complexes, resulted in the formation of proteasome granules in yeast. This differential regulation based on the initial carbon source involves specific signaling pathways; the AMP kinase Snf1 and a mitogen-activated protein kinase (MAPK) signaling cascade both regulate proteasome granule formation. We propose a model in which proteasome localization is regulated through mitochondrial respiration.

The observation that the growth media prior to carbon starvation influenced the autophagic response in yeast resolved a controversy regarding general autophagy induction following carbon starvation (Takeshige et al., 1992; Lang et al., 2014). Starvation following respiratory growth, such as growth in media containing raffinose or glycerol as the carbon source, promoted autophagy induction, whereas growth in glycolytic or fermenting media, such as media containing dextrose (D-glucose), did not (Adachi et al., 2017). Consistent with this, a link between respiration and autophagy induction during energy deprivation has been established (Yi et al., 2017). Thus, the metabolic state of the cell is an important variable when evaluating carbon starvation responses. We previously showed that carbon starvation does not induce proteasome autophagy in yeast (Waite et al., 2016); however, we did not specifically control for the pre-starvation condition of the cells. Therefore, we used media containing dextrose (yeast extract peptone dextrose or ypd; used for glycolysis), raffinose (yeast extract peptone raffinose or ypr; used for glycolysis and respiration) or glycerol (yeast extract peptone glycerol or ypg; used exclusively for respiration) as sources of carbon to examine proteaphagy following carbon starvation. We monitored proteasome autophagy with a GFP-tag fused to the proteasome regulatory particle subunit Rpn1. Here, the observation of vacuolar fluorescence in cells or the appearance of a faster migrating ‘free’ GFP species on immunoblots are indicative of proteaphagy. No robust proteaphagy was observed upon switching to carbon starvation, independent of the growth medium used prior to starvation. The amount of free GFP detected at 1 and 2 days of carbon starvation was less than free GFP detected following 6 h of nitrogen starvation (Waite et al., 2016) (Fig. 1A). Instead, proteasome granules were induced, as previously reported (Waite et al., 2016; Marshall and Vierstra, 2018). Interestingly, pre-growth in dextrose-containing media resulted in approximately 58% of cells showing granules and 27% showing multiple granules per cell, whereas pre-growth in raffinose-containing media resulted in less than 32% of cells showing granules and 5% showing multiple granules per cell (Fig. 1B). A similar trend was observed when we monitored the proteasome core particle α1 (encoded by SCL1) using α1–GFP as a reporter (Fig. S1A). To ensure that the carbon source was the main determinant in the difference in granule induction, we also grew cells in synthetic defined (SD) complete media with the different carbon sources prior to starvation. Again, more granules were induced upon switching from media containing dextrose than switching from media containing raffinose or glycerol (Fig. S1B). This shows that proteasome localization is affected by the carbon source that cells were grown in prior to starvation.

To further explore the role of nutrients in proteasome localization, we analyzed the induction of proteasome granules following prolonged growth in rich media, another condition that results in granule formation (Laporte et al., 2008; Peters et al., 2016). When granules formed after growth for an extended period, a single dominant granule was observed in cells regardless of the initial carbon source (Fig. 1C). However, we observed that cells grown in rich media containing dextrose for 3 days had more cells forming granules in general than cells grown in media containing raffinose (Fig. 1C). Growth in glycerol-containing media resulted in very few cells forming granules after 3 days (Fig. 1C). This was also observed with α1–GFP as a reporter (Fig. S1C). Consistent with a model in which proteasome granules protect proteasomes from autophagy (Marshall and Vierstra, 2018), the conditions with less granules (i.e. raffinose and glycerol) showed more free GFP on immunoblots at both 2 and 3 days of growth in carbon-containing media (Fig. 1C; Fig. S1D), suggesting an increase in proteaphagy. However, we would expect more proteaphagy following growth in media containing glycerol considering that this condition formed markedly fewer PSGs (Fig. 1C). Interestingly, we observed GFP-positive bands on immunoblots with molecular masses between those of free GFP and Rpn1–GFP following prolonged growth in glycerol-containing media. Such fragmentation patterns were recently observed specifically for GFP-tagged proteasomes degraded through ESCRT-mediated micro-autophagy (Li et al., 2019). Overall, our data suggest that not only proteasome granules, but also the type and magnitude of proteaphagy are dependent on the initial carbon source and the metabolic state of the cell prior to starvation.

Switching from dextrose-containing media to carbon starvation results in the formation of multiple phagophore assembly sites (PASs), which does not occur when carbon starvation is initiated following prior growth in other carbon sources or upon nitrogen starvation. This likely affects autophagy induction and might contribute to the lack of autophagy in this condition (Adachi et al., 2017). The number of proteasome granules observed per cell correlates with increased PASs when switching from dextrose-containing media to carbon starvation, compared to switching from raffinose-containing media to carbon starvation (Fig. 1B; Fig. S1A). To determine whether proteasomes colocalized with PASs during carbon starvation, we used cells expressing GFP–Atg8 (a PAS or autophagy marker) and Rpn1–mCherry. We monitored GFP and mCherry localization upon carbon starvation after pre-growth in media containing dextrose, raffinose or glycerol (Fig. 1D). PAS structures were observed within 1 h of starvation, when no proteasome granules were yet detectable. At 1, 2 or 4 days, when both proteasome granules and PAS could be observed, we did not observe any striking colocalization (Fig. 1D), indicating that these were distinct structures. To test whether proteasome granule formation depended on PAS formation or the signaling pathway that induces an autophagic response upon carbon starvation, we analyzed strains in which different genes required for autophagy induction (ATG1, SNF1 and GGC1) were deleted. These gene products activate the genome integrity checkpoint kinase protein Mec1 in a process that tethers Mec1 to the mitochondrial outer membrane upon carbon starvation. This mitochondrial localization of Mec1, together with active mitochondrial respiration, is a prerequisite for autophagy induction (Yi et al., 2017), and was shown for yeast pre-grown in media containing dextrose. To determine whether Snf1 and Ggc1 are required for carbon starvation-induced autophagy after pre-growth in media containing raffinose or glycerol, we used a cytosolic rosella autophagy reporter (Rosado et al., 2008; Mijaljica et al., 2012). This reporter undergoes cleavage between red fluorescent protein and GFP when targeted to the vacuole. We observed a reduction in cleaved rosella in the Snf1 and Ggc1 deletion mutants compared to that seen in wild-type (WT) strain, indicating that these proteins are required for efficient autophagy in our carbon starvation conditions (Fig. S1E). However, ATG1, SNF1 and GGC1 were not required for proteasome granule formation upon carbon starvation (Fig. S1F), indicating that multiple distinct signals couple the metabolic state of the cells to these two degradative machineries. The majority of cells formed proteasome granules at 24 h when pre-grown in media containing dextrose (∼56%), which was significantly higher compared to cells pre-grown in media containing raffinose or glycerol (26% and 22%, respectively) (Fig. 1D). In summary, conditions in which cells initially had active respiration showed fewer proteasome granules.

To further probe the role of mitochondrial respiration in regulating proteasome localization, we tested other conditions that induce respiration versus those that induce glycolysis. First, we tested the carbon starvation response after growing cells for 4 h (more glycolysis and fermentation) or 24 h (more respiration) in ypd medium. After 24 h of dextrose depletion, yeast switch to respiration for energy production (Galdieri et al., 2010). When we switched cells to carbon starvation after growing in rich media for 24 h, we observed that fewer cells formed proteasome granules compared to cells switched following logarithmic growth (Fig. S2A).

Next, we starved cells for phosphate following growth in media containing different carbon sources. This was based on the rationale that cells lacking phosphate sources would be compromised in maintaining energy production as phosphate is required for ATP production (Ko et al., 1999). Phosphate starvation induces cell cycle arrest and a stress response similar to that induced by carbon starvation (Petti et al., 2011; Secco et al., 2012). Phosphate starvation also induces proteasome autophagy, but to a lesser extent than that induced by nitrogen starvation at 24 h (Waite et al., 2022). Here, we show that prolonged phosphate starvation induced the formation of proteasome granules (Fig. 2A). Similar to proteasome granules induced by carbon starvation, these granules had different properties depending on the initial carbon source. Cells grown in media containing raffinose that were starved for phosphate showed fewer proteasome granules compared to cells grown in media containing dextrose. Even more striking was the almost complete absence of granules when cells were pre-grown in media containing glycerol (Fig. 2A), even though our phosphate starvation media always contained dextrose as a carbon source. Catabolite repression does not appear to play a role here as pre-growth in media containing raffinose and glycerol resulted in distinct phenotypes. Because granules induced by phosphate starvation are evident only at later timepoints (2 days), they could result from prolonged growth, similar to granules formed during stationary phase. However, we found that granules were induced to a much larger extent in SD media lacking phosphate compared to SD complete media, both of which contained dextrose as a carbon source (Fig. S2B). Overall, proteasome granule formation appears to be in part dependent on mitochondrial respiration. Under conditions of limited mitochondrial respiration, proteasomes appear to relocalize to proteasome granules more efficiently.

Cells grown in respiratory media formed fewer proteasome granules when starved for carbon or phosphate, or when grown for 3 days. Thus, increased mitochondrial function appears to limit granule formation. This led to the hypothesis that reducing mitochondrial function has an opposing effect and causes proteasome granule formation. To test this, we inhibited mitochondrial function by targeting different components of the electron transport chain (ETC). Sodium azide, antimycin A and oligomycin A were used to target cytochrome c oxidase, complex III and ATP synthase, respectively. We further utilized the uncoupler 2-[2-(3-chlorophenyl)hydrazinylidene]-propanedinitrile (CCCP) to disrupt ATP synthesis (Heytler and Prichard, 1962; Wikström and Berden, 1972; Gribble et al., 1997; Symersky et al., 2012). All four drugs induced the formation of proteasome granules in rich (dextrose) media (Fig. 2B), further corroborating the link between mitochondrial function and proteasome localization. We monitored proteasomal and mitochondrial localization in these different granule-inducing conditions to determine whether proteasomes and mitochondria colocalize. Although a strain with two fluorescently tagged proteasome subunits (Rpn1–mCherry/Rpn5–GFP) showed clear colocalization, we did not observe colocalization of the mitochondrial markers Cit1–GFP or Tom70–GFP with mCherry-tagged proteasomes in granules (Fig. S2C).

To test whether granules induced through mitochondrial inhibition behaved like proteasome granules induced by carbon starvation, we tested for the reversibility of these granules. The carbon starvation-induced granules, PSGs, unlike other types of proteasome-containing granules, quickly dissipate as proteasomes relocalize to cell nuclei upon re-addition of a carbon source (Laporte et al., 2008; Peters et al., 2013; Weberruss et al., 2013; Waite et al., 2020). Granules induced by mitochondrial inhibitors also quickly dissipated and the GFP signal from proteasomal subunits was predominantly nuclear after cells were washed and re-inoculated into drug-free media (Fig. 2C). This indicates that these granules were not associated with irreversible aggregates but were dynamic structures, consistent with the proposed storage function of PSGs. These data also suggest that proteasome granule formation depends on nuclear export, as the nuclear GFP signal was increased upon drug washout. The magnitude of mitochondrial inhibition-induced granule formation was dependent on the carbon source during initial growth (Fig. 2D), with the exception of that seen in Rpn1–GFP cells treated with antimycin A. Similar to ETC inhibitors, we predicted that the absence of the final electron acceptor, oxygen, should also result in proteasome granule formation by inhibiting oxidative phosphorylation. Indeed, upon growth in anoxic conditions, proteasome granules were induced. Similar to carbon starvation- and mitochondrial inhibitor-induced granules, fewer granules formed under anoxic conditions when cells were grown in media containing raffinose or glycerol, compared to when cells were grown in media containing glucose (Fig. 2E). Overall, these data further indicate a role for mitochondrial respiration in regulating proteasome granule formation.

Intriguingly, we did not detect the induction of proteasome granule formation with all of the mitochondrial inhibitors tested, as potassium cyanide (KCN) did not induce granule formation. Treatment with various concentrations of KCN, which, like sodium azide, targets cytochrome c oxidase, did not result in any significant granule formation (Fig. S2D). KCN-treated cells grew slower than untreated cells (with a growth rate comparable to those of cells treated with other mitochondrial inhibitors), and we observed an increase in cell size at high concentrations (Fig. S2D), indicating that KCN treatment was effective. Thus, mitochondrial stress alone is not sufficient to drive proteasome granule formation. This led us to seek a downstream effector that regulates proteasome relocalization.

Mitochondrial stress can affect levels of reactive oxygen species (ROS), reduce ATP production, and lead to the accumulation of cytosolic mitochondrial proteins. Because the mitochondria inhibitors we utilize can lead to either an increase or decrease in ROS in different organisms (Starkov and Fiskum, 2003; Caldeira Da Silva et al., 2008; Leadsham et al., 2013; Berry et al., 2018; Malina et al., 2018), we focused on the reduction in cellular ATP levels and the cytosolic accumulation of mitochondrial proteins (due to mitochondrial import defects) as potential downstream factors contributing to proteasome granule formation. The mitochondrial inhibitors we used interfere with the electron transport chain, the proton gradient or ATP synthase, disruption of each of which compromises mitochondrial ATP production. A reduction in ATP levels can directly impact proteasomes because these complexes contain a AAA-ATPase hexameric ring that is essential for their function (Finley, 2009; Schrader et al., 2009). Indeed, proteasome complexes are unstable in vitro in the absence of ATP (Liu et al., 2006; Kleijnen et al., 2007). Therefore, we wondered whether ATP maintenance and proteasome stability are key determinants in proteasome granule formation; something that has been postulated previously (Enenkel, 2018; Karmon and Ben Aroya, 2020; Enenkel et al., 2022). To test whether proteasome granule formation was linked to ATP production, we measured ATP under granule-inducing and non-inducing stimuli using the Cell Titer Glo as well as the Enliten ATP assay systems (Nicastro et al., 2015; Adachi et al., 2017). Inhibiting mitochondria with sodium azide, antimycin A, oligomycin A, CCCP or anoxic growth conditions led to a large reduction in ATP levels compared to those of cells grown in rich media (Fig. 3A). This reduction correlated with granule formation. Interestingly, KCN treatment caused a smaller reduction in ATP levels compared to those of the untreated control, and this inhibitor did not induce proteasome granule formation.

We next measured ATP under starvation conditions. We observed a strong reduction in ATP following nitrogen starvation, a condition that does not induce the formation of granules but instead induces proteaphagy. We also observed a strong reduction in ATP detected following 24 h of phosphate starvation, a condition that induces proteaphagy (as well as granules at later timepoints, Fig. 2A) (Waite et al., 2022). Carbon starvation, however, did not cause a significant reduction in ATP levels, even though it robustly induced proteasome granule formation. Thus, with starvation, a reduction in ATP levels by itself is neither essential nor sufficient to trigger proteasome relocalization to granules. A possible explanation is that in addition to a drop in ATP levels, other signals are generated during nitrogen and phosphate starvation that result in proteaphagy.

Mitochondrial inhibition and a consequential drop in mitochondrial membrane potential (or proton gradient) cause mitochondrial protein import defects. This causes cytosolic accumulation of mitochondrial proteins, triggers protein quality control pathways and upregulates proteasomes (Bragoszewski et al., 2013; Wrobel et al., 2015; Boos et al., 2019). To determine whether this feedback mechanism correlated with the appearance of proteasome granules, we expressed a construct that was recently shown to block mitochondrial protein import and result in proteasome upregulation and cytosolic accumulation of mitochondrial proteins (Boos et al., 2019). Expression was induced with 0.5% galactose in cells containing a plasmid encoding dihydrofolate reductase (DHFR) with a mitochondrial targeting sequence (b2Δ_DHFR), which blocks mitochondrial protein import. As controls, we used a plasmid encoding cytosolic DHFR or a plasmid not encoding any open reading frame (empty vector). Methotrexate, a substrate analog of DHFR that also stabilizes folded DHFR, was added to enhance the blocking of mitochondrial import. The induced expression of this transport channel blocker, however, did not induce the formation of proteasome granules, despite the accumulation of the blocker (Fig. 3B,C). Overall, neither the accumulation of mitochondrial proteins in the cytosol nor the reduction in ATP levels was consistently associated with the formation of proteasome granules, suggesting that different or additional regulatory mechanisms are involved.

The differential regulation of proteasome granules observed following respiratory versus non-respiratory growth, as well as the observation that not all mitochondrial inhibitors induce the formation of proteasome granules (Fig. 2), suggests that there is a regulator of this process that is only active under specific conditions. The cell-wall integrity MAPK cascade proteins Mpk1 (encoded by SLT2), Mkk1 and Mkk2 are known to regulate proteasome abundance and proteasome autophagy (Rousseau and Bertolotti, 2016; Waite et al., 2022). This kinase cascade has further been shown to be required for induction of general autophagy when antimycin A or KCN was added to cells in a process that was dependent on Atg32 and Atg11 (Deffieu et al., 2013). When we tested whether deletion of ATG32 disrupted the formation of proteasome granules induced after prolonged yeast growth or following sodium azide or antimycin A treatment, we observed no reduction in the degree of granule formation (Fig. S3A). Similarly, ATG11 was not required for granule formation when cells were treated with sodium azide (Fig. S3B). This shows that the requirements for autophagy induction and granule formation are not identical in these conditions. Next, we wanted to test the role of MAPK signaling itself. To test whether the MAPK Mpk1 was important for granule formation, we deleted the MPK1 gene. In these cells, we observed a striking difference depending on the initial carbon source used. Granule formation was unaffected in a mpk1Δ strain that was starved for carbon following initial growth in media containing dextrose (Fig. 4A–E). However, when yeast were grown in media containing raffinose prior to starvation, Mpk1 was critical to form proteasome granules efficiently, as indicated by both the regulatory particle (RP) and core particle (CP) reporters (Rpn1–GFP and α1–GFP, respectively) (Fig. 4A–E). This suggests that Mpk1 is required for proteasome granule formation in cells that have active mitochondrial respiration. When we monitored granule formation following oligomycin A, antimycin A and CCCP treatment, we found Mpk1 to again be important for cells that were grown in media containing raffinose (Fig. 4B–E). With respect to sodium azide, granules monitored using α1–GFP behaved as above with a reduction in the MPK1 mutant following growth in media containing raffinose (Fig. 4D,E). Granule formation monitored using Rpn1–GFP, however, was restricted when mpk1Δ cells were grown in media containing dextrose but not in media containing raffinose (Fig. 4D,E). This observation further supports the idea of differential regulation of the CP and RP in granule formation, as has been observed previously (Weberruss et al., 2013; Marshall and Vierstra, 2018; Karmon and Ben Aroya, 2020).

As the deletion of Mpk1 abolishes proteasome granule formation, we predicted that granule-inducing conditions would correlate with activated Mpk1. To test this, we monitored Mpk1 phosphorylation levels by immunoblotting under the different assay conditions (Fig. S4) (Mao et al., 2011). Following prolonged growth in dextrose-containing media, we observed lower levels of Mpk1 phosphorylation than those seen for cells grown in raffinose- or glycerol-containing media, despite the higher levels of proteasome granules (Fig. S4A). We also observed more phosphorylated Mpk1 in anoxic conditions when cells were grown in media containing raffinose or glycerol compared to when cells were grown in media containing dextrose, in which more granules are induced (Fig. S4B). This correlation, however, was not seen in the mitochondrial inhibitor conditions. For example, sodium azide and antimycin A treatment showed no phosphorylated Mpk1 at the time that we monitored granule formation (Fig. S4C). Overall, our data show that the absence of Mpk1 has a clear impact on proteasome granule formation, and increased levels of phosphorylated Mpk1 correlate with a reduction in the magnitude of granule formation. There are many potential explanations for this disconnect. For example, kinase signaling is often rapid and transient, and we monitored phosphorylation at later timepoints compared to when we observed proteasome granules. Furthermore, we looked at the aggregate phosphorylation of this kinase, and its regulation is known to be more complex as differential phosphorylation affects activation and activity of Mpk1 (González-Rubio et al., 2021). More detailed studies on the molecular targets responsible for regulating Mpk1-dependent proteasome localization are needed to resolve this. One obvious target is the proteasome chaperone Adc17 (encoded by TMA17). Adc17 is activated under several stress conditions (Hanssum et al., 2014) by Mpk1 (Rousseau and Bertolotti, 2016). However, we found that Adc17 was dispensable for proteasome granule formation, suggesting an alternate route of proteasome regulation by Mpk1 (Fig. S4D).

To determine the involvement of other kinases, we analyzed other major yeast MAPKs from independent pathways (Levin, 2005), as well as proteins within the same cell-integrity pathway. The deletion of FUS3 or HOG1 (genes encoding MAPKs involved in pheromone response and osmoregulation, respectively) did not prevent or substantially reduce proteasome granule formation (Fig. S4D). This indicates a specific role for the cell-integrity kinase pathway. Looking at MAPK kinases upstream of Mpk1, we evaluated Mkk1 and Mkk2. Although individual deletion of these paralogs had little impact on granule formation, a double deletion of MKK1 and MKK2 showed a strong reduction in proteasome granule formation under carbon starvation and sodium azide (NaN₃), antimycin A, oligomycin A or CCCP treatment (Fig. 4F). This reduction was, similar to what was seen for MPK1 deletion, observed when cells were grown in media containing raffinose but not when cells were grown in media containing dextrose prior to treatment. Tracing this MAPK cascade further, we found that the MAPK kinase kinase Bck1 was also required; however, BCK1 mutants formed fewer granules after switching from glucose-containing media than those seen for the MPK1 mutant or MKK1/MKK2 double mutant, although we did observe a more significant reduction compared to WT when cells were cultured in media containing raffinose (Fig. 4G). Upstream of Bck1, the transmembrane activator of the cascade Wsc1 was also required for granule formation (Fig. S5). Other upstream activators of the cell-integrity pathway, namely, Wsc2, Wsc3 and Ack1 (Verna et al., 1997; Kuranda et al., 2006; Krause et al., 2008), were not required under the same conditions (Fig. S5). Our data show that an intact MAPK pathway, starting at Wsc1 and moving downstream to Mpk1, is required for proteasome granule formation when cells are grown in respiration-inducing conditions. We have previously reported that Mpk1, Mkk1 and Mkk2 are required for efficient proteaphagy (Waite et al., 2022). Both conditions require nuclear export of proteasomes, indicating that this kinase cascade regulates the nuclear export of proteasomes, for example, by direct phosphorylation of proteasomes by Mpk1. However, Mpk1 is involved in many cellular processes and this kinase pathway might be indirectly involved in the regulation of proteasome relocalization.

As mentioned above, Snf1 is required for the induction of autophagy upon carbon starvation. This kinase is recruited to mitochondria shortly after the onset of carbon starvation, where it phosphorylates Mec1. Phosphorylated Mec1 recruits Atg1 to mitochondria and both factors are required for maintaining mitochondrial respiration during carbon starvation (Yi et al., 2017). This respiration, in turn, is also required for autophagy induction (Adachi et al., 2017; Yi et al., 2017). We did not observe a reduction in the formation of proteasome granules upon carbon starvation when SNF1 or other factors involved in this pathway were deleted (Fig. S1F), a finding consistent with published data (Li et al., 2019). Snf1 was also not required for proteasome autophagy induced by nitrogen starvation but was required for proteasome autophagy observed following 4 days of growth in SD media with low (0.025%) glucose (Li et al., 2019). Intriguingly, we noticed a different role for Snf1 in proteaphagy under growth with different carbon sources. Compared to WT, we observed reduced cleavage of the general autophagy reporter rosella in SNF1 mutants grown in media containing raffinose or glycerol for 3 days. Little general autophagy was observed when cells were grown in media containing dextrose for the same time period (Fig. S6A). When we monitored proteaphagy, we saw that deletion of SNF1 promoted proteaphagy when cells were grown for 3 days in rich media containing dextrose or glycerol but not raffinose (Fig. 5A). Furthermore, we observed that SNF1 mutants failed to form proteasome granules efficiently in media containing dextrose or raffinose. No granules were induced in WT after 3 days in media containing glycerol (Fig. 5B). For growth in dextrose-containing media, it appeared that proteaphagy was promoted independently of general autophagy induction in SNF1-deleted strains. As little general autophagy was observed, this was inconsistent with a model wherein the granules protect proteasomes from autophagic degradation (Marshall and Vierstra, 2018). Indeed, we also did not observe increased proteaphagy in the Snf1 mutant when cells were grown in media containing raffinose, despite the failure in granule formation and Snf1-dependent general autophagy in this condition (Fig. 5B; Fig. S6A). Here, the GFP signal appeared more nuclear and diffuse in the cytoplasm for the Snf1 mutant than for the WT, which formed proteasome granules (Fig. 5B). As we did not observe a significant difference in colony-forming units between the WT and Snf1 mutants following prolonged growth in these three carbon sources, differences in autophagy induction and granule formation did not result from the death of snf1Δ cells (Fig. S6B). The disparate phenotypes we observed likely reflect the complexity of combined inputs to the metabolic state of the cells and mitochondrial respiration activity in regulating proteasome localization.

To gain more insight into the role of Snf1 in regulating proteasome localization, autophagy and granule formation, we tested the effect of mitochondrial inhibitors in these cells. Unlike granules induced by carbon starvation (Li et al., 2019), those induced with mitochondrial inhibitors were highly dependent on Snf1 (Fig. 5C) for both the RP and CP reporters. Upon mitochondrial inhibition, compared to WT, Snf1 mutants had fewer proteasome granules, more nuclear GFP signal and little to no vacuolar GFP signal. The observation that mutants (snf1Δ and mpk1Δ) that fail to form proteasome granules retain nuclear GFP signal suggests that proteasomes are exported from cell nuclei to form cytosolic granules. Of note, the SNF1 mutant cells were grown in glucose-containing media, in which Snf1 is inactive (Schüller, 2003; Kayikci and Nielsen, 2015). This might indicate an alternative role for Snf1 in regulating proteasome localization outside of its canonical role of non-fermentative gene repression. Importantly, these cells were largely viable as only antimycin A treatment resulted in a significant difference in colony-forming units between the WT and Snf1 mutants (Fig. S6C). Overall, these data suggest that different pathways are involved in the formation of proteasome granules, for example, during carbon starvation versus mitochondria inhibition. This raises the question as to what extent these granules are similar or qualitatively different in function or regulation.

The transcriptional upregulation of proteasomes (by Rpn4 in yeast and Nrf1 in mammals) is critical for cells to respond to various stress conditions. However, not all stresses require a (continuous) upregulation of proteasomes and instead trigger either the degradation of proteasomes or their relocalization into granules. In yeast, proteasome complexes are enriched in the nucleus under optimal conditions, such as during logarithmic growth. However, proteasomes relocalize to cytoplasmic granules upon carbon limitation, and the rapid reversibility of these granules is consistent with condensates formed via liquid–liquid phase separation (LLPS) (Enenkel, 2012; Peters et al., 2013; Waite et al., 2016; Marshall and Vierstra, 2018). When stressed for nitrogen, however, proteaphagy is induced (Marshall et al., 2015; Waite et al., 2016; Nemec et al., 2017). Similarly, in human cells, proteasomes have been reported to undergo proteaphagy in response to amino-acid starvation or upon proteasome inhibition (Cohen-Kaplan et al., 2016; Choi et al., 2020; Goebel et al., 2020). Interestingly, cytoplasmic proteasomes appear to be substrates for proteaphagy upon amino-acid deprivation, whereas nuclear proteasomes undergo LLPS (Uriarte et al., 2021). Osmotic stress has also been shown to induce nuclear LLPS of proteasomes (Yasuda et al., 2020). Thus, both in yeast and humans depending on the condition, proteasomes show distinct fates and localization. To understand how these specific fates of proteasomes are regulated, we need to not only know the factors involved, but also understand the triggers responsible. This can be rather subtle, as it was recently shown that clearly defined fates for proteasomes were observed when comparing cells starved in low levels of glucose or no glucose at all (Li et al., 2019). Here, only the cells starved in low glucose induced Snf1-dependent micro-autophagy of proteasomes. In the current study, we sought to determine in detail what triggers proteasome granule localization and further identified some of the key signaling kinases involved.

Our data show that mitochondrial respiration, at least in part, regulates proteasome localization (Figs 2 and 5). First, we show that several inhibitors of mitochondrial respiration robustly induce proteasome granule formation. Second, removing a carbon source from cells that grew in media that suppresses respiration (i.e. with glucose; Galdieri et al., 2010) resulted in the formation of multiple granules per cell, whereas one dominant granule was present when cells were switched from media that required respiration (i.e. with glycerol or raffinose). This observation shows interesting parallels with the regulation of general autophagy upon carbon starvation, which was recently shown to be dependent on respiration (Adachi et al., 2017). Another study found that a complex of Snf1–Mec1–Atg1 is recruited to the mitochondrial membrane by Ggc1 following the onset of carbon starvation. The formation of this complex is required to maintain active mitochondrial respiration, which, in turn, is required to initiate carbon starvation-induced autophagy (Yi et al., 2017). Surprisingly, the deletion of GGC1 did not impact PSG formation, suggesting additional signaling pathways link respiration to the regulation of proteasome localization.

The extent of granule formation when cells were grown in different carbon sources negatively correlated with the amount of general autophagy induced, as we detected a reduction in the number of proteasome granules under conditions that led to increased general autophagy upon carbon starvation (Adachi et al., 2017). Here, we observed only a small increase in proteaphagy (Fig. 1A) when cells were grown in respiratory media, suggesting the reduced levels of proteasome granules cannot be explained by increased autophagy of proteasomes. This is in contrast to the reported role of proteasome granules in protecting proteasomes from autophagy, as we would expect more proteasome autophagy when there is a reduction in the formation of proteasome granules. Our data suggest that proteasomes are excluded from autophagic degradation without being sequestered into proteasome granules. This may indicate a role for proteasome activity in regulating the autophagic response of respiring cells. Our observations are further supported by the lack of an increase in proteasome autophagy under autophagic conditions in mutants that are defective in granule formation (Figs 3 and 4). The nuclear enrichment of fluorescence we observed under these conditions indicates that the majority of proteasomes remained nuclear. This localization shields proteasomes from general autophagy (Waite et al., 2016; Nemec et al., 2017). Thus, proteasomes can be excluded from autophagic degradation either by proteasome granule formation or by nuclear retention. In line with this, the deletion of autophagy-related genes results in nuclear retention (Waite et al., 2016, 2022). This suggests that proteasome relocalization is tightly regulated. Next, we discuss a kinase cascade we have identified that is involved in this process.

MAPKs and Snf1 signaling pathways both regulate proteasome localization; however, each do so under distinct conditions. The Mpk1 kinase cascade was required following growth in media that induced respiration, whereas Snf1 was required upon mitochondrial inhibition (but not carbon starvation). These pathways are known to cooperate upon cell-wall and oxidative stress (Backhaus et al., 2013; Willis et al., 2018), indicating that they might work together to regulate proteasomes more generally. Indeed, both pathways are also involved in regulating proteasome abundance and mobilization upon different stressors. Mpk1 is required for proteasome upregulation upon endoplasmic reticulum stress (Rousseau and Bertolotti, 2016; Schmidt et al., 2019) as well as required for efficient proteaphagy (Waite et al., 2022), and Snf1 is required for micro-autophagy of proteasomes following growth in low-glucose media (Li et al., 2019). The requirement for these pathways in both proteasome autophagy and proteasome granule formation indicates that they are necessary for efficient nuclear export of proteasomes, a prerequisite for both autophagy and granule formation (Nemec et al., 2017). However, the possibility that they play a less direct role in regulating proteasome localization cannot be excluded. It has been shown that MPK1 is required for general autophagy induction when mitochondria are inhibited with the drug antimycin A (Deffieu et al., 2013). Similarly, it is required for mitophagy and pexophagy induction following nitrogen starvation, even though general autophagy does not depend on Mpk1 (Mao et al., 2011). Here, mitophagy was shown to depend on the complete cell-integrity MAPK cascade (from the transmembrane receptor kinase Wsc1 to Mpk1). We demonstrate that this same cascade is required for proteasome granule formation. These data point to a more general signaling role for MPK1, particularly when mitochondria are affected. However, other factors that are required for mitophagy under this condition, such as Atg32 and Atg11, had no effect on the formation of proteasome granules, suggesting that a general failure in mitophagy does not correlate with a failure in proteasome granule formation. Nevertheless, the Mpk1 kinase cascade is clearly important in regulating proteasome localization, particularly following respiratory growth.

An intact MAPK cascade is not sufficient to promote proteasome granule formation. We report here that deletion of SNF1 restricted proteasome granule formation upon inhibition of mitochondria. It was previously shown that this kinase is not required for granule formation when cells were starved for glucose (Li et al., 2019). The requirement for this protein in these different contexts suggests that proteasome granules induced upon carbon starvation versus those induced by mitochondrial inhibition are distinct. It is intriguing that Snf1 was required for granules formation when cells were grown in glucose-containing media, as this protein is known to be inactive in this condition (Schüller, 2003; Turcotte et al., 2010). This suggests an unidentified role for Snf1 in this context that regulates proteasome localization.

Overall, our data show that proteasomes are specifically regulated under different metabolic conditions. This may indicate that proteasome activity is similarly regulated based on cellular needs. Given that the majority of proteasomes are nuclear, proteasome capacity and activity must be altered when they relocalize. Indeed, the current literature suggests that proteasomes are inactive in PSGs (Gu et al., 2017; Enenkel, 2018). However, it should be noted that LLPS structures induced by osmotic stress in the HCT116 colon cancer cell line appear to be actively degrading ribosomal proteins (Yasuda et al., 2020). Under conditions of mitochondrial respiration during which proteasomes are more nuclear, proteasome activity might be necessary. This observed relationship between mitochondrial respiration and proteasome localization is intriguing, as other studies have demonstrated a functional link between mitochondria and proteasomes (Bragoszewski et al., 2013, 2017; Lavie et al., 2018). This is also evident in Parkinson's disease in which proteasomes play a crucial role in regulating mitochondrial dynamics (Junn et al., 2002; Ciechanover and Brundin, 2003; Webb et al., 2003; Um et al., 2010). Furthermore, proteasomes are required to resolve mitochondrial stress induced upon transporter clogging or accumulation of misfolded mitochondrial proteins in the cytoplasm (Boos et al., 2019). Our data show that proteasomes are responsive to cellular metabolism and are regulated differently depending on the metabolic status of the cell (Fig. 6). Although ATP appears to not be the trigger for proteasome granule formation, at least upon starvation, our data make it clear that mitochondrial function and signaling play an important role.

All S. cerevisiae strains used in this study are reported in Table S1. Our background strains are the W303-derived SUB61 (Matα, lys2-801 leu2-3, 2-112 ura3-52 his3-Δ200 trp1-1), which arose from a dissection of DF5 (Finley et al., 1987). Standard PCR-based procedures (primers and plasmids presented in Table S2) were used to delete specific genes from the genome or to introduce sequences at the endogenous locus that resulted in the expression of C-terminal fusions of GFP or mCherry (Goldstein and McCusker, 1999; Hailey et al., 2002; Janke et al., 2004).

Overnight cultures of yeast were inoculated at an optical density (OD) measured at 600 nm (OD₆₀₀) of 0.5, grown in yeast extract peptone medium supplemented with 2% dextrose (ypd), raffinose (ypr) or glycerol (ypg) as a carbon source, and grown to an OD₆₀₀ of 1.5 (approximately 4 h). To induce starvation, cultures growing logarithmically were centrifuged, washed with the respective starvation medium, re-inoculated at an OD₆₀₀ of 1.5, and incubated at 30°C with constant shaking. Yeast nitrogen base lacking nitrogen, carbon or phosphate sources was used to make the respective starvation media. For drug treatments, cultures were grown to an OD₆₀₀ of 1.5 as above, then treated. Sodium azide (VWR), antimycin A (Sigma-Aldrich), oligomycin A (Cayman Chemical) and CCCP (Sigma-Aldrich) were used at final concentrations of 0.5 µM, 0.1 mM, 2.5 µM and 10 µM, respectively. KCN (Thermo Fisher Scientific) was used at concentrations shown in Fig. 2E. Anoxic growth was performed by transferring 1.8 ml of logarithmically growing culture to a 2 ml culture tube that was sealed and incubated without shaking for 24 h at 30°C.

For experiments involving expression of the DHFR mitochondrial clogger, cells expressing the empty vector, cytosolic DHFR and mitochondria-targeted DHFR (b2Δ_DHFR) were grown overnight in SD media without tryptophan containing lactate as the carbon source. They were re-inoculated in the same media and grown for 4 h before adding galactose to a final concentration of 0.5% with or without 50 µm methotrexate (Millipore Sigma). Next, cells were incubated for indicated time before microscopy or cell lysis.

For western blots, two ODs of cells were collected at the indicated timepoints and treatments, and stored at −80°C. Lysis was completed using previously established methods (Kushnirov, 2000). Following electrophoreses, samples were transferred to polyvinylidene difluoride membranes and immunoblotted with indicated antibodies followed by the appropriate horseradish-peroxidase (HRP)-conjugated secondary antibodies. The antibodies used were anti-GFP (1:500; Roche, 11814460001), anti-Pgk1 (1:10,000; Invitrogen, 459250), anti-Phospho-p44/42 MAPK (1:1000; Cell Signaling Technology, 9101), anti-Mpk1 (1:1000, Santa Cruz Biotechnology, sc-374435) and anti-DHFR (1:1000, Proteintech, 15194-1-AP). HRP activity was visualized using the Immobilon Forte Western HRP substrate (Millipore), and images were acquired using the G-box imaging system (Syngene) with GeneSnap software. Genetools was used to quantify the amount of free GFP and phosphorylated Mpk1 normalized to the Pgk1 loading control. Uncropped western blot images are presented in Fig. S7.

For the Enliten ATP assay, one OD of cells was collected following the indicated treatments and frozen in liquid nitrogen. Pellets were resuspended in 50 µl of 2.5% trichloroacetic acid and boiled for 3 mins. Sample was centrifuged at 16,200 g for 1 min, and 2 µl supernatant was added to 98 µl of 25 mM Tris-HCl (pH 8.8) (1:50 dilution). Then, 10 µl of the 1:50 diluted solution was combined with 40 µl of 25 mM Tris-HCl (pH 8.8) in a white 96-well plate. Finally, 50 µl of rLuciferase/Luciferin reagent (Promega, ENLITEN ATP Assay System) was added to each well and luminescence was monitored using a plate reader.

For the Cell Titer Glo ATP assay, 0.5 OD of cells were centrifuged following the indicated treatments and resuspended in 50 µl of sterile water. Samples were transferred to a black 96-well plate and 50 µl of Cell Titer Glo 2.0 reagent (Promega) was added. This plate was incubated in the dark under constant rotation on an orbital shaker for 4 min. Following rotation, the plate was further incubated in the dark for 10 min at room temperature. Luminescence was then measured using a plate reader.

All microscopy was performed with live yeast in which proteasome subunits Rpn1 or α1 were C-terminally tagged at their endogenous locus with expression driven by the endogenous promoter. GFP–Atg8 was produced as previously described (Li et al., 2015). After the indicated treatments, approximately 2 ODs of cells were pelleted, washed with PBS, and resuspended in 30 µl of PBS. Then, 3 µl of this sample was mounted on 1% soft agar slides as described by Eric Muller (https://www.youtube.com/watch?v=ZrZVbFg9NE8) (Sundin et al., 2004). All imaging by fluorescence microscopy was done within 10 mins following wash to avoid the effects of prolonged incubation on slides. Images were acquired at room temperature using a Nikon Eclipse TE2000-S microscope at 600× magnification with a Plan Apo 60×/1.40 objective equipped with a Retiga R3 camera (QImaging). Images were collected using Metamorph software (Molecular Devices) and analyzed using Fiji. All quantification was performed using Fiji.

The equivalent of one OD of culture was harvested by centrifugation, washed in dH₂O, and resuspended in 1 ml of dH₂O. Cells were then serially diluted to 1:10,000 and 100 µl plated on ypd plates. Colonies were then counted to determine cell viability.

We thank Dr Stella Lee for helpful discussions and feedback on the manuscript. We thank Mandeep Kaur for generation of some of the yeast strains used in this study and feedback on the manuscript. We thank Dr Alicia Burris for sharing her observation of proteasome granule formation under hypoxic conditions. Some of the text and figures in this paper formed part of K.A.W.'s PhD thesis in the Division of Biology at Kansas State University in 2019.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.259778.reviewer-comments.pdf.