Alteration of protein localization is an important strategy for cells to regulate protein homeostasis upon environmental stresses. In the budding yeast Saccharomyces cerevisiae, many proteins relocalize and form cytosolic granules during chronological aging. However, the functions and exact components of these protein granules remain uncharacterized in most cases. In this study, we performed a genome-wide analysis of protein localization in stationary phase cells, leading to the discovery of 307 granule-forming proteins and the identification of new components in the Hsp42-stationary phase granule (Hsp42-SPG), P-bodies, Ret2 granules and actin bodies. We further characterized the Hsp42-SPG, which contains the largest number of protein components, including many molecular chaperones, metabolic enzymes and regulatory proteins. Formation of the Hsp42-SPG efficiently downregulates the activities of sequestered components, which can be differentially released from the granule based on environmental cues. We found a similar structure in the pre-whole genome duplication yeast species, Lachancea kluyveri, suggesting that the Hsp42-SPG is a common machinery allowing chronologically aged cells to contend with changing environments when available energy is limited.
Changing subcellular localization of proteins enables cells to efficiently regulate intracellular enzymatic activities or protein interactions in response to environmental changes. Such physiological adjustments are often crucial for cell survival or competition in fluctuating natural environments. One mode of spatial regulation is the formation of dynamic granule structures, such as stress granules or processing bodies (P-bodies). These RNA-protein granules are conserved from yeast to mammals and are suggested to control translation or degradation of mRNA under specific conditions or subcellular localizations (Decker and Parker, 2012; Parker and Sheth, 2007). Interfering with the formation or correct localization of these granules can lead to reduced fitness or diseases (Buchan, 2014; Lavut and Raveh, 2012; Nostramo et al., 2016; Ramachandran et al., 2011).
Recently, several granule structures have been observed in stationary phase yeast cells (Laporte et al., 2008; Liu et al., 2012; Narayanaswamy et al., 2009; Sagot et al., 2006; Shah et al., 2014), raising the possibility that granule formation represents a general mechanism for cells to combat energy-limiting conditions. During entry into stationary phase, gene expression profiles are greatly altered, involving upregulation of stress response genes, including many heat shock proteins (Hsps; Gasch et al., 2000; Werner-Washburne et al., 1993). In addition, the rates of transcription and translation are generally reduced to conserve energy (Choder, 1991; Fuge et al., 1994). Genetic studies of the physiological changes in stationary phase yeast cells have provided important insights into the general mechanism of chronological aging (Bitto et al., 2015; Kaeberlein, 2010). However, our knowledge of the regulation and function of stationary phase granules remains limited.
Since the sources of energy and materials are restricted in stationary phase cells, formation of protein granules is often regarded as a mechanism to store crucial proteins that will later be used for mitosis re-entry (Laporte et al., 2008; Narayanaswamy et al., 2009; Sagot et al., 2006). However, a few recent studies indicate that some granule structures only appear in cells with inferior physiological states (Lee et al., 2016; Vasicova et al., 2015), suggesting that the storage model is oversimplified. In log phase cells, misfolded or damaged proteins are often collected to specific compartments or deposition sites. Upon inhibition of the proteasome, misfolded proteins induced by heat shock are sequestered to the insoluble protein deposit (IPOD), which is located beside vacuoles and contains amyloidogenic aggregates, and cytosolic (CytoQ) as well as intranuclear quality control compartments (INQ), which include amorphous aggregates (Kaganovich et al., 2008; Miller et al., 2015). In the absence of proteasome inhibition, misfolded proteins are transported to multiple dynamic foci, which are named Q-bodies, in the cytosol under stress conditions (Escusa-Toret et al., 2013). These cytosolic structures probably function as transient storage compartments for cells to deal with the overwhelming proteotoxicity generated by stress-induced misfolded proteins. It is unclear whether, in the energy-limited stationary phase cells, the granule structures also serve as collecting sites for misfolded or damaged proteins.
Small Hsps, such as Hsp42, can assemble into versatile dynamic oligomers with diverse subunit numbers and architectures (Haslbeck et al., 2005; Sun and MacRae, 2005). In log phase cells, Hsp42 positively regulates the formation of protein foci containing misfolded proteins upon heat stress, and the absence of Hsp42 reduces cell fitness at high temperature or during recurring heat shock cycles (Escusa-Toret et al., 2013; Grousl et al., 2018; Specht et al., 2011). During replicative aging, Hsp42 is also required for the formation of age-associated protein deposits and influences the aging process of mother cells (Saarikangas and Barral, 2015). In stationary phase, Hsp42 is crucial for the assembly of some proteins that have a role in epigenetics (Liu et al., 2012). Moreover, Hsp42-containing stationary phase granules (Hsp42-SPGs) have been shown to be enriched in long-lived quiescent cell populations, suggesting that Hsp42-SPGs help quiescent cells to combat various stresses during stationary phase, although the detailed mechanisms are still unclear (Lee et al., 2016).
In this study, we used mutant firefly luciferase as a model enzyme to elucidate the molecular function of Hsp42-SPGs. In stationary phase, Hsp42 granule formation sequestered the luciferase protein, thereby reducing its cytosolic activity. Since Hsp42 and Hsp104 were involved in granule assembly and disassembly, respectively, it suggested that the proteins sequestered to the granule are partially misfolded. However, when nutrients were replenished, the sequestered enzymes were quickly discharged to restore luciferase activity. A systematic screen of the yeast GFP-fusion collection (Huh et al., 2003) revealed that Hsp42-SPGs contained at least 61 endogenous components, including Hsps and many metabolic enzymes. Interestingly, only molecular chaperones were released under heat stress conditions, indicating that the granule structure is dynamic. Finally, we found that a distant yeast species Lachancea kluyveri also developed similar granule structures during stationary phase, supporting the functional importance of Hsp42-SPGs.
A mutant luciferase protein is sequestered to Hsp42-SPGs in chronologically aged yeast cells
Previously, Hsp42-SPGs were found to be enriched in long-lived quiescent yeast cells (Lee et al., 2016; Liu et al., 2012), suggesting that formation of Hsp42-SPGs influences cell physiology during chronological aging. When the viability of stationary phase cells was examined in rebudding assays, cells without Hsp42-SPGs exhibited significantly reduced viability in 1-month cultures (46±1.6% of wild-type cells vs 26±0.3% of hsp42Δ cells, two-tailed t-test, P=0.0049) (Fig. 1A). However, in order to understand the biochemical states of Hsp42-SPG components during granule formation, a sensitive functional assay of protein folding and enzymatic activities is needed.
We found that a mutant form of the firefly luciferase-EGFP fusion protein might serve as a model enzyme to tackle this issue (Gupta et al., 2011). This luciferase mutant spontaneously formed cytosolic granules that colocalize with Hsp42-SPGs in stationary phase cultures even without heat shock (Fig. 1B). In contrast, the wild-type luciferase was rarely recruited to the granule (3.3±0.9% compared to 81.3±3.4% in the cells carrying the luciferase mutant). Moreover, the granule formation of luciferase strictly depended on Hsp42; the luciferase protein was distributed evenly in the cytosol in stationary phase hsp42Δ cells (Fig. 1C). The behavior of luciferase is similar to the previously identified Hsp42-SPG components Hos2 and Mca1 (Liu et al., 2012). Since this luciferase contains mutations that will readily induce protein misfolding, it raises the possibility that only proteins prone to misfolding were collected to Hsp42-SPGs. Nonetheless, it remains unclear whether the sequestered proteins are permanently damaged or can be reactivated later.
Formation of Hsp42-SPGs allows cells to regulate protein activities in stationary phase
Since the mutant luciferase was collected to Hsp42-SPGs gradually (Fig. 1C), it is possible that Hsp42-SPGs only sequester the luciferase-GFP protein that is completely misfolded or damaged during stationary phase. To test this hypothesis, we grew granule-less hsp42Δ mutant and wild-type cells and monitored the activity of luciferase at different time points. If only completely misfolded or damaged and, therefore, inactive proteins are collected to Hsp42-SPGs, wild-type and hsp42Δ mutant cells should exhibit similar luciferase activities. In contrast, wild-type cells will show much lower luciferase activities if Hsp42-SPGs can actively collect fully or partially functional proteins. In 1-day cell cultures, both strains had similar levels of total cellular luciferase activity (Fig. 1D). However, the difference in luciferase activities between hsp42Δ and wild-type cells gradually increased at later time points when luciferase granules started to form (Fig. 1C and D, ∼3-fold and ∼10-fold differences in 7-day-old and 14-day-old cells, respectively).
We also performed western blots to examine the total amount of luciferase protein. In both wild-type and hsp42Δ mutant cells, the levels of luciferase protein were gradually reduced – possibly due to the activated autophagy in stationary phase cells (Wang et al., 2001). Nonetheless, abundance of luciferase protein in hsp42Δ mutants was slightly less compared to that in the wild-type strain (Fig. 1E), negating the possibility that decreased luciferase activities in the wild-type cells was caused by reduced protein amounts. Another possible explanation for the higher luciferase activity in hsp42Δ cells is that more molecular chaperones were available in the cytosol of mutant cells. We performed an in vitro luciferase refolding assay to test this possibility (Glover and Lindquist, 1998). Cell lysates prepared from 14-day-old wild-type and hsp42Δ cells, which did not carry the luciferase-GFP construct, were measured for their ability to reactivate denatured luciferase. The result showed that hsp42Δ cells had a slightly lower level of refolding activity than wild-type cells (Fig. S1), indicating that hsp42Δ cells did not have more chaperones in the cytosol. Taken together, our results show that sequestering a specific protein to Hsp42-SPGs enables a cell to downregulate its enzymatic activity without lowering the protein level.
In stationary phase, cells encounter many challenges that are similar to stress conditions. Studies in log phase cells have revealed that cells can collect damaged or misfolded proteins into specific compartments, such as IPOD, CytoQ or Q-body, under stress conditions, and that these proteins are subsequently degraded once the stress is relieved (Escusa-Toret et al., 2013; Kaganovich et al., 2008; Miller et al., 2015). We tested whether the protein components in Hsp42-SPGs are targeted for degradation or can be refolded back into functional conformation when cells exit the quiescent state. Stationary phase cells were supplied with fresh medium containing cycloheximide to inhibit translation of new proteins. Cycloheximide treatment ensured that the detected luciferase activity came from pre-existing luciferase in stationary phase cells. Our previous study has shown that, under this condition, Hsp42-SPGs disassemble and release their components (Liu et al., 2012). Interestingly, cytosolic luciferase activity drastically increased when the protein was released from Hsp42-SPGs (Fig. 1F). In contrast, luciferase activity remained at a similar level in the granule-less hsp42Δ mutants before and after nutrient feeding. These data provide direct evidence that specific proteins stored in Hsp42-SPGs can be refolded and reactivated for later use upon re-entry into the cell cycle.
Identification of the protein components in known granule structures within stationary phase cells
Our result, showing that proteins sequestered to Hsp42-SPGs could be reactivated at later stages, prompted us to search for the endogenous components of Hsp42-SPGs. To obtain a comprehensive list of Hsp42-SPG components, we performed a genome-wide screen of the subcellular localization of yeast proteins in stationary phase using the yeast GFP-fusion collection (Huh et al., 2003). In the first round, 4071 strains from the collection were grown in YPD for 5 days to enter stationary phase, and fluorescence images of cells in both log and stationary phase were analyzed (see Materials and Methods for details). The localization patterns of GFP fusion proteins in stationary phase were manually divided into six categories, i.e. no specific pattern, granule, punctate, cell periphery, nuclear periphery and fibril (Table 1 and Fig. S2A; see also data on FigShare available at https://doi.org/10.6084/m9.figshare.6958307). Most of the proteins were evenly distributed in the cytosol, nucleus or vacuole in stationary phase cells and, therefore, fell into the category no specific pattern. Interestingly, more than 600 yeast proteins formed dot-like structures in stationary phase, falling into the category punctate or granule (Table 1).
Next, we focused on the 307 proteins that formed only one or two individual cytosolic dots, since this category included Hsp42 and because other proteins within the same category were more likely to be Hsp42-SPG components (Fig. S2C). When the log-phase localization patterns of these granule-forming proteins were examined, only a small proportion of them (17.9%) exhibited punctate patterns and the majority were evenly distributed in the cytosol (45.0%) or the nucleus (31.6%) (Fig. S2D). In total, more than 200 yeast proteins radically changed their original localization to form cytosolic granules during stationary phase, suggesting that granule formation is a specific response to starvation stress or aging effects in stationary phase cells.
To further identify the Hsp42-SPG components, mCherry-tagged Hsp42 was used as a marker of Hsp42-SPGs to see whether it colocalized with other proteins (see Materials and Methods for details). A plasmid containing the mCherry-tagged HSP42 gene was first transformed into each of the 307 strains that carry GFP-fusion granule-forming proteins, and the transformants were induced into stationary phase to examine the localization patterns of mCherry and GFP signals (Fig. S3A). Of the 307 strains, 61 displayed colocalization of GFP and Hsp42-mCherry signals (one example is shown in Fig. 2A), and were defined as components of Hsp42-SPGs (Table 2 and Table S1). Interestingly, different components of Hsp42-SPGs were collected to granules in order (Fig. S3B and Table S2), suggesting that the Hsp42-SPG has a specific structure.
An alternative approach was also used to identify the components of Hsp42-SPGs. We tagged Ssa1, an Hsp42-SPG component, with the C-terminal TAP-tag (Puig et al., 2001) and used it as the bait to pull down Hsp42-SPGs in stationary phase cells (Fig. S3C,D). Combining co-immunoprecipitation (co-IP) and mass spectrometry (see Materials and Methods for details), we identified 532 proteins with a detected intensity that was at least 10-fold higher than that in the log phase control (Table S3). The Hsp42-SPG components identified by our GFP colocalization screen were significantly enriched in the co-IP proteins (32 out of 61 components were present in the co-IP list; Fisher's exact test, P=3.101×10−16). The missing Hsp42-SPG components in the co-IP result were probably due to low abundance or low association stability of the proteins in the granule. Alternatively, the colocalization of some GFP-tagged Hsp42-SPG components might depend on their fusion to GFP. More experiments will be needed to distinguish these possibilities. Nonetheless, both colocalization and co-IP experiments revealed that Hsp42-SPGs are huge complex structures that contain more than 30 different proteins. To gain more information regarding the function of Hsp42-SPGs, the Hsp42-SPG components identified by the colocalization screen were used for our further analyses.
Several granule-like structures have previously been reported in stationary phase cells (Laporte et al., 2008; Narayanaswamy et al., 2009; Sagot et al., 2006; Shah et al., 2013). However, the components of these structures are not fully identified. To know whether the rest of GFP-fusion granule-forming proteins belong to these structures, we repeated the colocalization experiments by using other marker proteins. The P-body component Edc3, the actin-binding protein Abp1, the proteasome subunit Pre6, and the coatomer complex subunit Ret2 were chosen as marker proteins to identify the components of these protein granules (Laporte et al., 2008; Sagot et al., 2006; Shah et al., 2013). By screening the same set of 307 GFP strains, we found 11, 21, 15 and 12 proteins to be the components of actin bodies, P-bodies, proteasome storage granules and Ret2 granules, respectively (Fig. 2B and Table S1). Most components of proteasome storage granules and actin bodies are proteasome subunits and proteins related to actin organization (Table S1). No new components were identified in the proteasome storage granule. In P-bodies, 14 of the 21 components are known components of log-phase cells (Table S1), and the remaining 7 genes encode protein kinases and proteins involved in RNA metabolism. Collectively, our data show that the colocalization experiment is a reliable method to identify the components of stationary phase granule structures. Moreover, ∼64% of granule-forming proteins (197 out of 307) did not colocalize with any known protein granules, indicating that, in stationary phase cells, there are many other granule structures not yet characterized.
The Hsp42-SPG contains enzymes comprising various functions
Electron microscopy experiments were conducted to further examine the size of Hsp42-SPGs. Stationary phase yeast cells were fixed and immunogold-labeled using anti-Hsp42 antibodies. The electron microscopy data showed that Hsp42-SPGs were at least 260 nm in diameter (major axis=365±14 nm; minor axis=260±10 nm; n=54). Also, unlike protein aggregation structures, such as IPOD or INQ, in log phase cells (Kaganovich et al., 2008; Miller et al., 2015), Hsp42-SPGs did not specifically lie next to the vacuole or nucleus (Fig. 2C). To further examine the relationship between Hsp42-SPGs and IPOD, Rnq1-GFP was used as the marker of IPOD (Kaganovich et al., 2008) and overexpressed in the cell carrying Hsp42-Ruby2. Consistent with previous observations (Escusa-Toret et al., 2013), only few cells contained Hsp42 dots in log phase, and when Hsp42 dots were observed, they did not colocalize with the IPOD marker (Fig. S4). However, in 3-day-old cultures, Hsp42-SPGs and Rnq1-GFP dots were formed in most cells, and these two structures partially overlapped (90.5% in cells harboring both Hsp42-Ruby2 and Rnq1-GFP dots; n=158; Fig. S4). High-resolution imaging revealed that Hsp42-SPGs and IPOD were located adjacent to each other in these overlapping dots (Movie 1), suggesting that regulation of Hsp42-SPGs and IPOD is tightly associated. It will be interesting to see whether IPOD is regulated differently in stationary phase cells.
The functions of the Hsp42-SPG components are quite diverse, and can generally be divided into several categories (Table 2). Molecular chaperones (including yeast Hsp104, Hsp40, Hsp70, Hsp90 co-chaperones and small Hsps) and metabolic enzymes involved in various biosynthesis pathways are two main groups. Among those 15 metabolic enzymes, many of them (Ald6, Cys4, Glk1, Gly1, Ham1, Rib4, Sam2, Shm2 and Yhb1) have been shown to be important for survival in stationary phase or mitosis re-entry (Fabrizio et al., 2010; Garay et al., 2014; Powers et al., 2006). Therefore, sequestering these metabolic enzymes is consistent with the storage model. In contrast, it is less clear why molecular chaperones were also sequestered to Hsp42-SPGs. One possibility is that they were collected to form a protein quality control center so they could work more efficiently. We performed additional experiments to address this issue (see next section).
Our luciferase experiments showed that only the mutant luciferase was collected in Hsp42-SPGs in stationary phase, suggesting that the endogenous components of Hsp42-SPGs also have a similar biochemical property. A previous proteomic study has identified 319 yeast proteins that had a tendency to misfold and form aggregates in log phase cells after heat shock (Ruan et al., 2017). We compared these two lists and found that approximately half of the Hsp42-SPG components (31 out of 61=51%) belonged to this group of misfolding-prone proteins (Fisher's exact test, P=2.20×10−16; Table S3). These proteins are likely to be protected by molecular chaperones in Hsp42-SPGs to prevent further misfolding and damage during chronological aging.
Hsps are released from Hsp42-SPGs when cells encounter heat stress
We have shown that fresh nutrients could trigger the restoration of luciferase activity without translation of new proteins in stationary phase cells. Similar to the luciferase protein, endogenous Hsp42-SPG components were also released from granules when fresh carbon sources were supplied. Nonetheless, the role of molecular chaperones in Hsp42-SPGs needs to be further examined. We used heat stress to address this question. If the chaperones are simply stored like other metabolic enzymes or if they form structural components of the granule, their localization should not change after heat shock. In contrast, the chaperones should be released to deal with heat-induced denatured proteins in cytosol if their function is to help cells in stationary phase to control proteostasis.
Stationary phase cells were subjected to heat shock at 50°C and then monitored by using microscopy. After heat shock, many Hsps – including members of the Hsp70 family, such as Ssa1, Ssa2, Ssa3 and Ssa4, as well as Sis1 and Sti1 – were released from Hsp42-SPGs (Fig. 3A, Table S2), presumably to deal with denatured proteins or stressed cytosolic environments. In contrast, all metabolic enzymes were stably maintained in the granules (Fig. S5A and Table S2). Interestingly, Hsp42 and Hsp104 were also maintained in Hsp42-SPGs after heat shock. Earlier studies have shown that Hsp104 is targeted to protein aggregates with the help of Hsp40 and/or Hsp70 chaperones (Kłosowska et al., 2016; Winkler et al., 2012). Since we have ruled out the possibility that the Hsp104-GFP fusion protein is not functional (Fig. S5B), this result implies that the recruitment of Hsp104 to Hsp42-SPGs is different compared with typical protein aggregates. The functions of Hsp42 and Hsp104 are further addressed later (see next two sections).
We also performed western blots to show that total amounts of Hsps were constant before and after heat shock, indicating that the GFP signals observed in the cytosol were not from newly synthesized proteins (Fig. 3B). The heat shock experiment was repeated following addition of cycloheximide and similar results were observed (Fig. S5C), confirming that protein synthesis is not required. Furthermore, cell viability assays showed that cells had a similar viability before and after heat shock (Fig. 3C). Thus, the diffuse GFP signals were not caused by cell death. Most interestingly, the diffused chaperones were later re-assembled to Hsp42-SPGs (Fig. 3A and Fig. S4C), probably transferring the damaged proteins to Hsp42-SPGs for further processing or storage. It might also indicate alleviation of the cellular crisis.
Finally, we tested the effect of Hsp42-SPG formation on heat stress resistance. One-week-old stationary phase cells were heat shocked at 50°C and their viability was measured in rebudding assays. In 1-week-old cultures, cells with or without Hsp42-SPGs exhibited similar viabilities before heat shock. However, viability of hsp42Δ cells was significantly reduced after heat shock (73±1.8% of wild-type cells vs 57±2.5% of hsp42Δ cells, two-tailed t-test, P=0.008) (Fig. 3D). The heat shock-induced difference in viability was not observed in log-phase cells (Fig. S5D) (Gasch et al., 2000; Haslbeck et al., 2004), suggesting that it was not simply due to HSP42 deletion. Together, these results provide evidence that some molecular chaperones, such as Ssa1, Ssa2, Ssa3, Ssa4, Sis1 and Sti1, in Hsp42-SPGs are dynamic and can be used to protect misfolding-prone enzymes in the granule but are released to deal with denatured proteins under environmental stresses. This might be crucial for cell survival in energy-limited stationary phase.
Hsp42 facilitates efficient granule formation of most Hsp42-SPG components
In the heat stress experiments, Hsp42 and Hsp104 behaved differently from many other molecular chaperones. Several lines of evidence suggest that the Hsp42 has a critical role in the structure of Hsp42-SPGs. Of all Hsp42-SPG components, Hsp42 is the earliest one to form the granule structure (Table S2). In the luciferase experiment, we found that granule formation of the mutant luciferase strictly depended on Hsp42 (Fig. 1C). When cells encountered heat shock, many Hsps were released from the granules but Hsp42 remained unreleased. Moreover, deletion of HSP42 significantly reduced the rebudding frequency and heat resistance of stationary phase cells (Figs 1A and 3D). To examine how Hsp42 influenced the granule formation of other Hsp42-SPG components, we deleted HSP42 in strains carrying GFP-tagged genes encoding Hsp42-SPG components and compared percentages of cells forming granules in the wild-type and mutant strains. Our results showed that 41 out of 60 Hsp42-SPG components were significantly influenced by deletion of HSP42 (Fig. 3E and Table S2), suggesting that Hsp42 is a crucial component for Hsp42-SPG formation. In contrast, deletion of HSP104 did not affect the assembly of Hsp42-SPGs (Fig. 4A and Fig. S6).
The disaggregase Hsp104 is required for efficient disassembly of Hsp42-SPG components
Our luciferase activity assays showed that proteins sequestered to Hsp42-SPGs can be reactivated when they were released from the granule (Fig. 1F). To understand how Hsp42-SPG components are reactivated, we investigated whether Hsp104 was involved in the disassembly process of Hsp42-SPGs. Hsp104 is the disaggregase required for efficient re-folding or clearance of misfolded proteins (Glover and Lindquist, 1998; Parsell et al., 1994) and a component of Hsp42-SPGs. An earlier study showed that Hsp104 was also required for efficient disassembly of heat-induced stress granules (Cherkasov et al., 2013). We measured granule disassembly rates of nine proteins, Ade16, Gdh2, Gln1, Sam1, Sam2, Sis1, Uba4, Vma2 and Yhb1, which were sequestered to Hsp42-SPGs at different time points. Although each component started to disassemble at different time points, disassembly of all examined granule structures were significantly delayed in the absence of Hsp104 (Fig. 4A and Fig. S6). The effect of Hsp104 on granule disassembly is specific to Hsp42-SPGs. When the granule disassembly rates of proteasome storage granules and P-bodies were examined in wild-type and hsp104Δ cells, no obvious differences were observed (Fig. 4B).
To further confirm the effect of Hsp104, time-lapse microscopy was used to record the time points of granule disassembly in individual cells (Fig. 4C). Again, the data showed that the Gln1-GFP granule started to disassemble at much later time points in hsp104Δ mutant cells (median disassembly times of 60 min and 150 min in wild-type and hsp104Δ cells, respectively. Mann–Whitney U-test, P=1.347×10−9). Together, our data suggest that Hsp104 plays a vital role in the disassembly of Hsp42-SPGs and, perhaps, also acts to restore Hsp42-SPG components.
Hsp42-SPGs are conserved structures that also exist in a pre-whole-genome duplication yeast species
Our current data suggest that the Hsp42-SPG is a complex structure allowing cells to sequester various proteins and help cell survival during stationary phase. We wondered whether this structure is a general mechanism for cells to deal with fluctuating environments or is a recent innovation specific to fermentative yeasts. A protoploid yeast species, Lachancea kluyveri, was selected to address this topic. L. kluyveri diverged from the S. cerevisiae lineage before the whole-genome duplication (WGD) event that occurred in the latter lineage more than 100 million years ago (Souciet et al., 2009; Wolfe and Shields, 1997). It is well known that WGD leads to many new innovations in the lifestyles of yeast and that pre-WGD species differ radically from post-WGD species in their physiologies (Jiang et al., 2008; Scannell et al., 2007).
In L. kluyveri cells, the Hsp42 ortholog (Lk-Hsp42), SAKL0H14454p, was tagged with a fluorescent protein (mCherry or BFP) and then monitored at different cell stages. Interestingly, despite the protein sequence identity between Lk-Hsp42 and S. cerevisiae Hsp42 being far below the average for the whole proteome (44% compared to 57%), Lk-Hsp42 exhibited similar expression and localization patterns to S. cerevisiae Hsp42. Expression levels of Lk-Hsp42 increased when L. kluyveri cells started to enter stationary phase and granule formation could be observed in most cells after 1 day of growth in YPD (Fig. 5A). Next, we checked the granule formation of orthologs of other Hsp42-SPG components; an essential metabolic enzyme Lk-Gln1 (SAKL0F13838p) and a GET complex component Lk-Sgt2 (SAKL0E01210p). Both orthologs formed granule structures and colocalized to Lk-Hsp42 (Fig. 5B), indicating that they are also components of Hsp42 granules in stationary phase L. kluyveri cells. Since a similar granule structure can be found in both pre-WGD and post-WGD species, we postulate that formation of Hsp42-SPGs evolved in ancient times and that they play a critical role in enabling cells to adapt to environmental change.
Distinct cytosolic granule structures have been widely observed in various cell types and organisms when cells encounter environmental or physiological changes (An et al., 2008; Decker and Parker, 2012; Jain et al., 2016; Jourdain et al., 2016; Narayanaswamy et al., 2009; Parker and Sheth, 2007; Patel et al., 2004; Sagot et al., 2006; Xu and Chua, 2011). A recent study also showed that stress-triggered phase separation allowed cells to form granule structures and adapt to environmental stresses (Riback et al., 2017). In natural habitats, the stationary phase represents one of the most stringent challenges to the cell, with nutrients being limited and many cellular functions needing to be shut down in order for cells to conserve energy (Werner-Washburne et al., 1993). Thus, it is crucial for there to be a global adjustment in metabolic pathways and cellular processes in order to survive during stationary phase and to exit it efficiently. Our systematic screen revealed that at least 307 proteins formed granule structures during stationary phase, suggesting that granule formation represents a general mechanism for cells to regulate a variety of cellular functions and to maintain homeostasis.
Previously, it had been proposed that granule formation in stationary phase cells functions as a means of storage for essential proteins to facilitate efficient cell cycle re-entry once nutrients become available (Laporte et al., 2008; Sagot et al., 2006). The observation that many stationary phase granules are reversible structures provides indirect evidence that supports this hypothesis. However, as observed in log phase cells, some granule structures also collect permanently damaged proteins under stress conditions and release them for degradation when conditions recover (Escusa-Toret et al., 2013; Kaganovich et al., 2008; Miller et al., 2015). This type of regulation is probably crucial for protein homeostasis when the proteasome is compromised and misfolded proteins cannot be efficiently degraded (Miller et al., 2015; Specht et al., 2011). When yeast cells enter stationary phase, proteasome activity is also gradually decreased (Bajorek et al., 2003). Therefore, it is possible that stationary phase cells use Hsp42-SPGs to collect partially misfolded proteins to prevent further damage or perturbation. An earlier study has shown that overexpression of Hsp42 can prevent the formation of toxic [URE3] prion amyloids by sequestering prion seeds (Wickner et al., 2014). Moreover, protein substrates associated with small Hsps can retain near-native conformations and, thus, have a chance to be efficiently refolded (Ungelenk et al., 2016). By using a firefly luciferase mutant that is prone to misfolding, we have demonstrated that sequestration of the luciferase protein to Hsp42-SPGs allows cells to downregulate cytosolic luciferase activities during stationary phase and to quickly restore this activity when cells re-enter mitosis. These results suggest that the Hsp42-SPG can work as a center that controls both protein quality and quantity in stationary phase.
Apart from protection of crucial components, downregulation of undesirable metabolic activities by Hsp42-SPGs can prevent ‘metabolic fluctuations’ that would waste cellular energy and might accidentally lead to disastrous outcomes, such as re-entry into the cell cycle under poor conditions, thereby shortening the chronological lifespan of the cell (Petrovska et al., 2014). We have further shown that stationary phase cells can release several Hsps from Hsp42-SPGs into the cytosol under heat stress and that the cells without Hsp42-SPGs were more sensitive to heat stress. Thus, the materials stored in Hsp42-SPGs are not necessary only for cell cycle reentry but also help stationary phase cells to counteract environmental challenges (Fig. 6).
Hsp42 is a small Hsp that interacts with various proteins to prevent them from forming irreversible aggregates (Haslbeck et al., 2005; Stengel et al., 2010; Van Montfort et al., 2001). The expression level of HSP42 is highly induced during diauxic shift before cells enter stationary phase (Gasch et al., 2000; Haslbeck et al., 2004). Our results strongly suggest that Hsp42 is a basic structural unit of Hsp42-SPGs that allows other proteins to be organized in the granule without losing their refolding ability. In contrast, Hsp104 is an ATP-dependent protein disaggregase that helps proteins to refold. Our data show that deletion of HSP104 does not affect the formation of Hsp42-SPGs. However, when nutrients are replenished, Hsp104 allows cells to release Hsp42-SPG components more efficiently, which is crucial for niche competition in natural habitats.
Hsp42-SPGs shared some components with the stress granule. However, heat-induced stress granules in log phase cells are morphologically distinct from those formed during stationary phase (Liu et al., 2012). Although Hsp104 is involved in efficient disassembly of both heat-induced stress granules (Kroschwald et al., 2015) and Hsp42-SPGs, the assembly of stress granules does not require small Hsps (Cherkasov et al., 2013). Recently, Jain and colleagues have used a proteomic approach to identify components of stress granules in log phase cells (Jain et al., 2016), providing us with the chance for a detailed comparison between these two structures. Among those 159 proteins in the extended stress granule proteome, 17 of them were shared by the Hsp42-SPG (Fisher's exact test, P=3.11×10−13; Fig. 5C and Table S3). The significant overlap in the components suggests that Hsp42-SPGs share similar functions with stress granules in log phase cells or that Hsp42-SPGs even represent a subtype of stress granules.
Another interesting granule-like structure is the hypoxia-induced glycolytic body (G body) that is important for the adaptation to hypoxia and conserved between yeast and humans (Jin et al., 2017). G bodies share 12 common components with Hsp42-SPGs (Fisher's exact test, P=1.34×10−7; Table S3). However, the formation of G bodies under hypoxia requires the presence of glucose and cannot be triggered by glucose starvation (Jin et al., 2017). Most of the shared components are chaperones, such as Ssa1-4, Ydj1, Hsp26 and Hsp42, suggesting that these common components are the core proteins that interact with other different proteins to form granule structures under distinct stresses.
Our data reveal that Hsp42-SPG formation not only allows cells to regulate protein activities during stationary phase and its exit, but also provides a protein quality control center for cells to deal with unexpected stresses (Fig. 6). Of the 307 granule-forming proteins we identified in the current study, only approximately one third fall into the category of known granule structures. Characterizing the functions of other so far unknown stationary phase granules will provide further information on how homeostasis is achieved during chronological aging.
MATERIALS AND METHODS
Yeast strains and growth conditions
All S. cerevisiae strains used in this study were derived from the S288C strain background. All L. kluyveri strains used in this study were derived from the Y159 (GRY1183) strain background (Gojkovic et al., 2001; Weinstock and Strathern, 1993). Yeast strains containing C-terminal GFP-tagged proteins were obtained from the chromosomal GFP-tagged yeast collection (Huh et al., 2003). The GFP, mCherry, yomRuby2 and yomTagBFP2 tags (Addgene, Cambridge, MA) were inserted in-frame at the C-terminus of the coding region of a gene, as described previously (Howson et al., 2005; Lee et al., 2013). All fusion proteins were expressed under their endogenous promoters. Deletion strains were constructed using PCR-amplified DNA fragments that contain deletion cassettes from the yeast deletion collection (Giaever et al., 2002; Winzeler et al., 1999) and by transforming the DNA fragments into the BY4741 parental strain. Cells were cultured at 28°C with aeration in homemade liquid YPD medium before being examined by microscopy. Distilled water was added regularly into the liquid culture to compensate for water loss from evaporation.
Rebudding assay and viability measurement
Agarose (2%) was dissolve in synthetic medium (6.7 g/l yeast nitrogen base without amino acids, 2% glucose, 20 mg/l methionine, 120 mg/l leucine, 20 mg/l histidine, 20 mg/l uracil) to prepare rebud agarose pads that had been spliced to the size of 4.5×4.5 mm. Yeast cells were loaded onto the surface of spliced agarose pads. After 3 min of air-drying at room temperature, the pads with cells were turned upside down and put into the wells of a Glass Bottom ViewPlate-96F (PerkinElmer, Waltham, MA). Time-lapse images were acquired using the ImageXpress Micro XL system (Molecular Devices, Sunnyvale, CA) every 10 min and analyzed manually using ImageJ. In the case of budded cells found at the beginning of the assay, the first rebudding time was counted when the bud started to enlarge. For single cells found at the beginning of the assay, the first rebudding time was counted when a small bud started to emerge. Cells that did not rebud within 10 h were classified as non-reproductive cells, and the rebudding frequencies were calculated.
Monitoring of protein localization
Yeast cells were diluted with PBS and loaded into a Glass Bottom ViewPlate-96F coated with concanavalin A (C2010, Sigma-Aldrich, St Louis, MO). Then, the plates were centrifuged to attach cells to the glass bottom and images were obtained using the ImageXpress Micro XL system at 28°C. All images were analyzed manually using ImageJ (http://rsbweb.nih.gov/ij). A previous study has shown that incubation of log-phase cells in water for 1 h is sufficient to induce granule formation of some metabolic enzymes, such as Gln1 and Gdh2 (Narayanaswamy et al., 2009). To rule out the possibility that the granules observed in stationary phase cells were induced by the PBS treatment, stationary phase cells carrying Gln1-GFP or Gdh2-GFP were monitored in PBS or exhausted YPD medium treated with activated charcoal (C3345, Sigma-Aldrich) to reduce auto-fluorescence (Thayer et al., 2014). There was no significant difference in the frequency of granule formation between cells treated with PBS and the exhausted YPD (Fig. S2B). In addition, the granule formation of Gln1-GFP and Gdh2-GFP was not further induced during prolonged treatment of PBS (Fig. S2B). For our experiments, all images were taken within 30 min after the cells were loaded.
Luciferase activity assay
The thermolabile luciferase mutant used in this study carried two substitutions (R188Q and R261Q) (Gupta et al., 2011). Cells carrying the wild-type or mutant luciferase-EGFP-containing plasmid (Lee et al., 2016) were cultured in synthetic complete medium without uracil (SC-URA) for one day and the medium was replaced by the same volume of YPD to avoid the quick-dying phenotype of cells grown in synthetic medium. The day we changed the medium to YPD was defined as day 0; thereafter cells were maintained in the same medium for one month. At different time points, cells from 1 ml of cell culture were collected, washed once with PBS and re-suspended in 0.7 ml of ice-cold PBS with 1 mM phenylmethylsulfonyl fluoride (PMSF). Samples were kept on ice before detection of luciferase activity. A volume of 0.5 ml of glass beads (0.5 mm diameter, BioSpec, Bartlesville, OK) was added to samples and cells were lysed by four rounds of 15-s bead-beating in 2-min intervals on ice using a Mini-Beadbeater-16 (BioSpec). The lysate was cleared by centrifugation at 1000 g for 10 min and the supernatant was defined as the total cell lysate. Protein concentrations of total cell lysates were quantified using the Bradford assay (B6916, Sigma-Aldrich). Luciferase activities of total cell lysates were measured using the ONE-Glo Luciferase Assay System (Promega, Madison, WI) on an EnSpire Multimode Plate Reader (PerkinElmer).
Total cell lysates were mixed with a half-volume of 3×SDS-loading buffer (188 mM Tris-Cl pH 6.8, 3% SDS, 30% glycerol, 0.01% Bromophenol Blue, 15% β-mercaptoethanol) and boiled for 5 min. The same total protein amount of cell lysate was loaded and SDS-PAGE was performed according to a standard protocol (Laemmli, 1970). Immunoblotting with homemade anti-Hsp104, anti-Hsp42, anti-Ssa1/2 (all used at 1:20,000), as well as anti-GFP (1:2000; sc-9996, Santa Cruz Biotechnology, Inc., Dallas, TX), and anti-glucose-6-phosphate dehydrogenase (1:5000; anti-G6PDH; Sigma-Aldrich) was performed according to the recommended protocols of the manufacturers.
Luciferase refolding assay
Luciferase refolding assay was performed following the protocol described by Glover and Lindquist, 1998, with some modifications. In brief, firefly luciferase (L9420, Sigma-Aldrich) was dissolved in stabilizing buffer (25 mM Tricine-HCl pH 7.8, 8 mM MgSO4, 0.1 mM EDTA, 10 mg/ml BSA, 10% glycerol, 0.25% Triton X-100) to make 10 μM stock solution. Luciferase was unfolded by diluting the luciferase solution 10-fold in 8 M urea in 25 mM HEPES-KOH pH 7.6, 150 mM KOAc, 10 mM Mg(OAc)2, and 10 mM DTT, followed by incubation at 30°C for 30 min. Solution of unfolded luciferase was 10-fold diluted in cold lysate buffer [20 mM HEPES-KOH pH 7.4, 100 mM KOAc, 2 mM Mg(OAc)2, 20% glycerol, 2 mM DTT] and kept on ice before reaction. Cells from 2-week-old cultures were lyzed in lysate buffer by bead beating. The lysate was centrifuged at 20,000 g for 20 min at 4°C. For the refolding reaction, 100 μl of lysate buffer containing 150 μg cell lysate, 3 mM ATP (A26209, Sigma-Aldrich), 10 mM phosphocreatine (P1937, Sigma-Aldrich), 3.5 U creatine phosphokinase (C3755, Sigma-Aldrich) and 10 nM luciferase was incubated at room temperature. At different time points, 10 μl of reaction solution was used to measure the luciferase activity using the ONE-Glo Luciferase Assay System on an EnSpire Multimode Plate Reader. Native control was performed following identical procedures omitting the urea unfolding step. For negative control, no cell lysate was added in the reaction.
Genome-wide screening of protein localization patterns in stationary phase cells and identification of granule components
Cells from the yeast GFP-fusion collection (Huh et al., 2003) were grown on 96-well plates for 5 days to acquire images of stationary phase cells. The acidity of YPD medium remained higher than pH 6 after 5 days of inoculation (from pH 6.53±0.02 in the beginning to pH 6.24±0.01 on day 5), which prevented cell death induced by acidic medium (Burtner et al., 2009). Image acquisition was performed following the method described above under ‘Monitoring of protein localization’. For each strain, at least 100 cells were analyzed and strains were divided into specific categories (i.e. granule, punctate, cell periphery, nuclear periphery or fibril) when >30% of the cells displayed the same localization pattern (e.g. 1 or 2 dot-like structures in the cytosol). When multiple localization patterns were observed in the population for a protein, the population was classified according to the most frequently observed pattern. The whole screening process was performed twice and only strains with consistent results are reported.
To rule out the possibility that granule structures were inside the nucleus, the granule-forming GFP strains were mated with strains carrying a marker of the nucleus Htb1-mCherry. The zygotes were then grown to stationary phase and colocalization of GFP and mCherry signals was monitored. Proteins that localized to the nucleus were excluded from our granule-forming gene list. We also used Kgd1-mCherry to indicate mitochondrial localization, with 205 of 384 punctate-forming proteins colocalizing with mitochondria.
To identify protein components of different granules, we cloned the mCherry-tagged marker genes (HSP42, EDC3, PRE6, ABP1 and RET2 for Hsp42-SPGs, P-bodies, proteasome storage granules, actin bodies and Ret2 granules, respectively) into the pRS416 vector. The resulting plasmids were transformed into the granule-forming GFP strains. Then the transformants were grown in SC-URA medium for 1 day before being sub-cultured in SC-URA or YPD for 5 days before microscopic examination. For each strain, at least 50 cells containing both green and red dots were examined. When >50% of cells showed colocalization of GFP and mCherry granules, the GFP-tagged protein was classified as a component of a specific granule. The localization pattern of the endogenous Hsp42 protein was confirmed by using immunofluorescence. The granules formed by endogenous Hsp42 colocalized to Hsp104 (Fig. S3A), demonstrating that the mCherry tag did not influence the localization of Hsp42.
Formation kinetics of Hsp42-SPG components
We examined whether the endogenous Hsp42-SPG components were sequestered to Hsp42-SPGs at different stages. The 61 strains carrying GFP-tagged proteins were grown in YPD before GFP images were taken at different time points to analyze the percentages of cells with GFP granules. In general, the frequency of GFP granule-forming cells did not increase after day 5 so the day 5 data were used as the maximum level. Next, we used the time point when granule-forming cells were first seen to exceed 50% of the maximum level to indicate their granule-forming kinetics (Fig. S3B and Table S2). The small Hsp 42 (Hsp42), one of the Hsp40 co-chaperones Sis1 (Lu and Cyr, 1998), and a dynamin-like GTPase Vps1 (Vater et al., 1992) were the earliest components of Hsp42-SPGs. Twelve and 15 proteins started forming granules in the populations on day 2 and day 3, respectively. Thirty one proteins constituted the late components that entered Hsp42-SPGs on day 5. These late components included most of the kinases, proteins involved in translation and transcription, and several metabolic enzymes.
Co-immunoprecipitation of Hsp42-SPGs
Three components of the Hsp42-SPG, Gln1, Hsp42 and Ssa1, were tagged with the C-terminal TAP-tag (Puig et al., 2001) and used to isolate Hsp42-SPGs. However, Hsp42-SPGs (marked with Hsp42-BFP or Ssa1-BFP) could only be co-immunoprecipitated by Ssa1-TAP (Fig. S3C). Thus, we used Ssa1-TAP as bait to pull down Hsp42-SPGs following the protocol used for stress granule purification (Jain et al., 2016). In brief, yeast cells carrying Ssa1-TAP were grown in 20 ml YPD for 5 days, harvested and resuspended in 8 ml ice-cold lysis buffer [50 mM Tris HCl pH 7.4, 100 mM KOAc, 2 mM Mg(OAc)2, 0.5 mM DTT, 50 μg/ml heparin (H3393, Sigma-Aldrich), 0.5% NP40, 1:5000 Antifoam B emulsion (A5757, Sigma-Aldrich)]. The same amount of log phase cells (OD600=0.4−0.6) was used as a negative control. Cells were lyzed by bead-beating and the Hsp42-SPGs were enriched by differential centrifugation at 4°C (Jain et al., 2016). The enriched lysate was incubated with magnetic beads (Dynabeads Pan Mouse IgG, Thermo Fisher Scientific, Waltham, MA) at 4°C for 1 h. The magnetic beads were then washed 3 times with lysis buffer, followed by 1× wash with lysis buffer containing 2 M urea, 1× wash with lysis buffer containing 300 mM KOAc and 1× wash with lysis buffer. Washed magnetic beads were resuspended in 2× SDS-loading buffer and boiled for 10 min. Purified proteins in the supernatant were separated using SDS-PAGE gels that were stained with Imperial protein stain (Thermo Fisher Scientific) to visualize proteins (Fig. S3D). Finally, gels were sliced into pieces and in-gel trypsin digestion was performed to prepare the samples for mass spectrometry (Shevchenko et al., 2006).
NanoLC-nanoESi-MS/MS analysis was performed on a Thermo UltiMate 3000 RSLCnano system connected to a Thermo Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific) equipped with a nanospray interface (New Objective, Woburn, MA). Peptide mixtures were loaded onto a 75 μm ID, 25 cm length PepMap C18 column (Thermo Fisher Scientific) packed with 2-μm particles with a pore width of 100 Å and were separated using a segmented gradient during150 min from 5–35% solvent B (0.1% formic acid in acetonitrile) at a flow rate of 300 nl/min. Solvent A was 0.1% formic acid in water. The mass spectrometer was operated in the data-dependent mode. Briefly, survey scans of peptide precursors from 350–1600 m/z were performed at 120 K resolution with a 2×105 ion count target. Tandem MS (MS2) was performed by isolation window at 2 Da with the quadrupole, CID fragmentation with normalized collision energy of 30, and rapid scan MS analysis in the ion trap. The MS2 ion count target was set to 104 and the maximal injection time was 50 ms. Only those precursors with a charge state 2–6 were sampled for MS2. The instrument was run in top speed mode with 3 s cycles; the dynamic exclusion duration was set to 60 s with a 10 ppm tolerance around the selected precursor and its isotopes. Monoisotopic precursor selection was turned on.
Database search and protein quantification
Mass spectra and MS2 spectra were analyzed by Proteome Discoverer (version 2.2) (Thermo Fisher Scientific) using SEQUEST algorithm and the yeast ORF database. The enzyme specificity was set to trypsin (full) allowing for two missed cleavages and precursor mass tolerances were 10 ppm for parent ions and 0.6 Da for fragment ions. Dynamic modifications were set for methionine oxidation (+15.99492 Da), asparagine and glutamine deamidation (+0.98402 Da) and protein N-terminal acetylation (+42.03670 Da). A maximum of 3 dynamic modifications were allowed per peptide and a static modification of +57.02147 Da was set for carbamidomethyl cysteine. The Percolator node within Proteome Discoverer was used to filter the peptide spectral match (PSM) false discovery rate to 1% and peptide intensity was calculated using the Precursor Ions Quantifier node. The unique and razor peptides were used for protein quantitation. Precursor intensity abundance was subsequently used to compare the protein and peptides across the replicates and datasets. Technical replicates (two injections of the same sample) were used to evaluate technical variation. To compare the protein amount between 5-day-old and log phase samples, the intensity of human IgG from magnetic beads was used as an internal control.
Yeast cells were fixed and stained following previously described protocols (Baggett et al., 2003) with modifications. In short, cells from stationary phase cultures were fixed with 3.7% formaldehyde for 2 h. Fixed cells were washed with PBS and treated with 160 μg/ml Zymolyase 100T (Seikagaku America Inc., St Petersburg, FL) for 1 h in the spheroplasting solution (1.2 M sorbitol, 0.1 M K2HPO4/KH2PO4, pH 7.4, 0.2% 2-mercaptoethanol). Spheroplasts were permeabilized with 1% Triton X-100 in the spheroplasting solution and blocked for 1 h in WT buffer (50 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 0.1% [v/v] Tween-20, 5% [w/v] BSA). Spheroplasts were then incubated with the homemade rabbit anti-Hsp42 antibody (1:2000 dilution in WT buffer) for 2 h at room temperature. The secondary antibody (711-165-152, Jackson Immunoresearch, West Grove, PA) was diluted 500-fold with WT buffer and incubation was performed for 1 h at room temperature.
Heat shock and granule component release
Yeast cells carrying GFP-tagged Hsp42-SPG components were grown in YPD for one week, and then treated with heat shock at 50°C for 10 min. Cells were then spotted on YPD plates to measure their viability and altered localization of Hsp42-SPG components was immediately monitored using the protocol described in “Monitoring of protein localization” above. Cells from the same culture were monitored again 24 h after heat shock.
Immunogold electron microscopy
Stationary phase yeast cells (1 week old) were pelleted and transferred to a 3-mm metal carrier for immediate high-pressure freezing and electron microscopy using a Leica EM HPM 100 (Leica Microsystems GmbH, Wetzlar, Germany). Freeze substitution was conducted using a Leica EM ASF (Leica Microsystems GmbH) as follows. (1) The frozen yeast cells were freeze-substituted in anhydrous ethanol containing 0.2% glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA) and 0.1% uranyl acetate (Polysciences, Niles, IL) at −85°C for 72 h. Then, the temperature was elevated at a rate of 1°C/h to −70°C, and the samples were kept for 24 h at this temperature. (2) The temperature was then increased at a rate of 1°C/h to −20°C and the samples were kept for 24 h at this temperature. Then the temperature was elevated at a rate of 5°C/h to 0°C, and the samples were kept for 12 h at this temperature. (3) Finally, the temperature was elevated at a rate of 5°C/h to 20°C. The samples were transferred to 1.5 ml tubes and then substituted with four changes of anhydrous ethanol over 24 h. (4) Samples were infiltrated with London Resin Gold (Electron Microscopy Science) at room temperature over 6 days. (5) Polymerization was conducted with UV light at −20°C for 24 h and then at room temperature for 48 h. Ultrathin sections were put on nickel grids. The grids were first blocked with 5% normal goat serum (Jackson Immunoresearch) in 50 mM Tris-Cl pH 7.5 at room temperature for 1 h, and then incubated with homemade rabbit anti-Hsp42 antibodies (dilution 1:4000) in washing buffer (50 mM Tris-Cl pH 7.5 containing 1% normal goat serum) at room temperature for 1 h. The grids were then washed in washing buffer, incubated with 12-nm gold-conjugated goat anti-rabbit IgG (dilution 1:30; Jackson Immunoresearch), and washed with washing buffer and then with distilled water. The sections were further stained with uranyl acetate. Images were obtained by a Tecnai G2 Spirit TWIN transmission electron microscope (FEI, Hillsboro, OR) equipped with a CCD camera (794.10.BP2 MultiScan; Gatan, Pleasanton, CA) and the acquisition software Digital Micrograph (Gatan).
Examination of the relative localization of Hsp42-SPGs and IPOD
To examine the relative localization of Hsp42-SPGs and IPOD, the IPOD marker Rnq1-GFP (Kaganovich et al., 2008) was cloned into the plasmid pCM189 and the expression of Rnq1-GFP was driven by the tet-off promoter (Garí et al., 1997). The plasmid was transformed into the cell with Hsp42 C-terminally tagged with Ruby2. The transformants were grown in CSM-URA medium at 28°C. At different time points, the z-stack images of the cells were acquired using a DeltaVision platform (Applied Precision-GE Healthcare Life Sciences, Pittsburgh, PA) with an Olympus IX70 wide-field microscope equipped with a 100×, 1.4 numerical aperture Plan-Apochromat objective and a Cascade II 512 electron-multiplying charge-coupled device camera (Photometrics, Tucson, AZ). To optimize the accuracy of protein colocalization, each z-stack with spacing of 0.2 μm was acquired with GFP signals immediately followed by Ruby2 signals. The images were deconvolved using softWoRx software (Applied Precision-GE Healthcare Life Sciences). The 3-dimentional animation was generated using Imaris software (Bitplane AG, Zurich, Switzerland).
Granule formation kinetics of Hsp42-SPG components
To quantify the kinetics of granule formation of Hsp42-SPG components, GFP strains were grown in YPD for 5 days. At different time points, we took GFP images of the strains. For each strain, we analyzed at least 100 cells in each field and data from three different fields were averaged.
Kinetics of granule disassembly of Hsp42-SPG components
Regrowth of stationary phase cells carrying GFP-fusion Hsp42-SPG components were monitored using the protocol described in “Monitoring of protein localization” above, except that PBS was replaced by low-autofluorescence YPD medium (Thayer et al., 2014). Time-lapse fluorescence images were taken every 10 min. For the time-course monitoring of granule disassembly of Yhb1-GFP, Gln1-GFP and Gdh2-GFP, we analyzed at least 100 cells in each field at specified time points. The proportions of cells displaying cytosolic granules from three different fields were averaged.
Transformation of L. kluyveri cells
L. kluyveri cells were transformed by electroporation as described previously (Hsu et al., 2011) with some modifications. Cells from 5 ml of the overnight culture in YPD medium were collected and resuspended in lithium buffer (10 mM Tris-Cl pH7.5, 1 mM EDTA, 25 mM dithiothreitol [DTT], 0.1 M lithium acetate pH 7.5). After incubation at 28°C for 2 h with shaking, the cells were washed sequentially with 40 ml of ddH2O, 25 ml of ice-cold ddH2O, and 5 ml of ice-cold 1 M sorbitol. Then the cells were resuspended in 0.5 ml ice-cold 1 M sorbitol and kept on ice. Approximately 1 μg of linear DNA fragments was mixed with 100 μl of electrocompetent cells, and the cells were transformed using the ECM 630 High Throughput Electroporation System (BTX, Holliston, MA) with the following settings: 1800 V, 200 Ω, and 25 μF. After electroporation, the cells were immediately mixed with 1 ml of 1 M sorbitol, pelleted and resuspended in 1 ml fresh YPD medium. The cells were recovered at 28°C for 4 h and then plated on appropriate agar plates to select successful transformants.
We thank Zdena Palkova and members of the Leu lab for helpful discussion and comments on the manuscript. We thank the staff of the Human Disease Modeling Center at the First Core Labs, National Taiwan University College of Medicine, for sharing bioresources. We also thank Cheng-Hsilin Hsieh from the Institute of Molecular Biology (IMB) Genomics Core for technical assistance, the IMB Imaging Core for microscopy support, and John O'Brien for manuscript editing. MS data acquisition with the Orbitrap Fusion mass spectrometer were obtained using the Academia Sinica Common Mass Spectrometry Facilities at IBC.
Conceptualization: J.-Y.L.; Methodology: H.-Y.L., J.-C.C., K.-Y.C., J.-Y.L.; Validation: H.-Y.L.; Formal analysis: H.-Y.L., J.-Y.L.; Investigation: H.-Y.L., J.-C.C., K.-Y.C., J.-Y.L.; Resources: J.-Y.L.; Data curation: H.-Y.L.; Writing - original draft: H.-Y.L., J.-Y.L.; Writing - review & editing: H.-Y.L., J.-Y.L.; Visualization: H.-Y.L.; Supervision: J.-Y.L.; Project administration: J.-Y.L.; Funding acquisition: J.-Y.L.
This work was supported by Academia Sinica of Taiwan (grant number: 2316-1050-300) and the Taiwan Ministry of Science and Technology (grant number: 106-2321-B-001-018).
The images from genome-wide screening of protein localization patterns in stationary phase, including two repeats, have been deposited into FigShare (https://doi.org/10.6084/m9.figshare.6958307)
The authors declare no competing or financial interests.