Abstract
Glycolysis is central to energy metabolism in most organisms, and is highly regulated to enable optimal growth. In the yeast Saccharomyces cerevisiae, feedback mechanisms that control flux through glycolysis span transcriptional control to metabolite levels in the cell. Using a cellobiose consumption pathway, we decoupled glucose sensing from carbon utilization, revealing new modular layers of control that induce ATP consumption to drive rapid carbon fermentation. Proton pumping and regulation of amino acid biosynthesis mediated by extracellular glucose sensors independently contribute to maintenance of energy homeostasis. Controlling the upper bound of cellular ATP levels may be a general mechanism used to regulate energy levels in cells, via a regulatory network that can be uncoupled from ATP concentrations under perceived starvation conditions.
Introduction
Most microorganisms favor glucose as their primary carbon source, as reflected in their genetic programs hard-wired for this preference. Central to carbon metabolism is glycolysis, which is finely tuned to the dynamic state of the cell due to the fact that glycolysis first consumes ATP before generating additional ATP equivalents. To avoid catastrophic depletion of ATP, the yeast Saccharomyces cerevisiae has evolved a transient ATP hydrolysis futile cycle coupled to gluconeogenesis (van Heerden et al, 2014). Glycolysis in yeast is also tightly coupled to glucose transport into the cell, entailing three extracellular glucose sensing mechanisms and at least one intracellular glucose signaling pathway (Youk & van Oudenaarden, 2009).
Synthetic biology and metabolic engineering of yeast holds promise to convert this microorganism into a “cell factory” to produce a wide range of chemicals derived from renewable resources or those unattainable through traditional chemical routes. However, many applications require tapping into metabolites involved in central carbon metabolism, a daunting challenge as living cells have numerous layers of feedback regulation that fine-tune growth to changing environments. Cellular regulation evolved intricate networks to maintain and ensure cell survival. For example S. cerevisiae has evolved to rapidly consume high concentrations of glucose through fermentation, while repressing the expression of other carbon consumption pathways, an effect termed glucose repression. When perturbed genetically, regulatory networks such as those in glucose repression often generate undesirable or unexpected phenotypes.
For yeast to be useful in producing large volumes of renewable chemicals or biofuels, it will be important to expand its carbon utilization to include multiple sugars in the plant cell wall. One promising approach that helps overcome glucose repression and allows simultaneous utilization of different sugars is cellobiose consumption (Ha et al, 2011). Cellobiose is a disaccharide with two units of glucose linked by a ß-1,4 glycosidic bond. Cellobiose consumption using a minimal additional pathway in yeast— a cellodextrin transporter (CDT-1) and intracellular ß-glucosidase (Galazka et al, 2010)– avoids glucose repression by importing carbon in the form of cellobiose instead of glucose. The cellodextrin transporter allows cellobiose to enter the cell where it is hydrolyzed to glucose and consumed via the native glycolytic consumption pathway. By moving glucose production into the cell, the N. crassa-derived cellobiose consumption pathway is nearly the minimal synthetic biological module imaginable in S. cerevisiae, comprised of just two genes. Nevertheless, in S. cerevisiae the cellobiose consumption pathway is inferior to consumption of extracellular glucose in terms of rate and results in a prolonged lag phase (Lin et al, 2014). Previous efforts to understand the impact of cellobiose consumption on the physiology of S. cerevisiae using transcriptional profiling revealed that cellobiose improperly regulates mitochondrial activity and amino acid biosynthesis, both of which are tightly coupled to the transition from respiration to fermentation (Lin et al, 2014).
Since glycolytic enzymes are regulated mostly at the post-transcriptional level (Daran-Lapujade et al, 2007), we probed cellobiose consumption in S. cerevisiae at the metabolite level. We found that key metabolites in glycolysis are highly imbalanced, leading to low flux through glycolysis and slow fermentation. We also found that excess ATP levels drive the imbalance, and identified a new regulatory role of glucose sensors in cellular ATP homeostasis.
Results
Metabolite profiling of cellobiose utilizing S. cerevisiae
S. cerevisiae cells engineered with the cellobiose consumption pathway ferment cellobiose more slowly than glucose and often exhibit a prolonged lag phase when transitioning to consumption of this disaccharide (Lin et al, 2014). We hypothesized that cellobiose consumption results in an ATP deficit in glycolysis relative to glucose, due to the fact that the cellodextrin transporter (CDT-1) in the cellobiose utilizing pathway is a proton symporter, requiring ATP hydrolysis for cellobiose uptake (Kim et al, 2014). Moreover, under anaerobic conditions, ATP production is limited to substrate-level phosphorylation, further restricting ATP availability. We measured the steady-state concentrations of ATP and other metabolites in central carbon metabolism in yeast fermenting cellobiose compared to glucose. Of the 48 compounds analyzed, the abundance of 25 compounds was significantly different (p-value < 0.01) between cellobiose and glucose-fed samples (Figure 1A). Surprisingly, ATP levels increased by 6-fold in the cellobiose grown cells (Figure 1B). The relative abundance of compounds in glycolysis—fructose 6-phosphate (F6P), glucose 6-phosphate (G6P), glucose and pyruvate—increased by 444-, 81-, 7- and 3-fold, respectively, while that of phosphoenolpyruvate decreased by 2-fold (Figure 1B). These results suggest that the yeast cells underwent drastic physiological changes, reflected in metabolite levels, when cellobiose was provided in place of glucose.
Phosphofructokinase-1 inhibition by excess ATP
Given the dramatic buildup of glucose, G6P and F6P intermediates prior to the phosphofructokinase (PFK1) reaction in glycolysis (Figure 2A), we reasoned that Pfk1 might be a major bottleneck in cellobiose consumption. Pfk1 catalyzes the phosphorylation of F6P, using one ATP and yielding fructose 1,6-bisphosphate (F1,6BP) as a product. As the second committed step in glycolysis, Pfk1 is heavily regulated—with ATP acting as an allosteric inhibitor and AMP and fructose 2,6- bisphosphate serving as activators (Bañuelos et al, 1977; Avigad, 1981; Nissler et al, 1983).
To test whether allosteric inhibition of Pfk1 by ATP limited cellobiose fermentation, a mutation in Pfk1 that makes the enzyme ATP-insensitive (P722L in the Pfk1 beta subunit, note the original publication referred to this mutation as P728L (Rodicio et al, 2000)) was introduced into the chromosomally encoded PFK1 gene (mutation here termed pfk1m) in the cellobiose utilizing strain. This mutation was previously shown to reduce ATP inhibition but also AMP and fructose 2,6-bisphosphate activation of Pfk1 in S. cerevisiae (Rodicio et al, 2000). We chose this mutation over an ATP-insensitive, AMP/F2,6BP-sensitive mutant phosphofructokinase (Nissler et al, 1983) because the latter’s phenotype has not been evaluated in S. cerevisiae. Consistent with allosteric inhibition of Pfk1 by ATP, the cellobiose consumption efficiency (Ec) of the pfk1m strain increased by 33% in comparison to the strain with wild-type Pfk1 (Figure 2B). Furthermore, the relative abundance of G6P and F1,6BP decreased by 47% and 34%, respectively, while that of ATP remained relatively unchanged (Figure 2C). These results indicate that the 6-fold increase in cellular ATP concentrations allosterically inhibited Pfk1, resulting in accumulation of glucose, G6P and F6P, which eventually slowed down cellobiose consumption.
Limited activity of plasma membrane ATPase
Although the pfk1m strain partially increased the rate of cellobiose fermentation, cellular ATP remained elevated relative to glucose fermentation. It is unlikely that ATP production was the cause of the difference, as fermentation is limited to substrate-level phosphorylation under anaerobic conditions regardless of carbon source. We therefore tested whether the activity of one of the major ATP sinks in yeast, the plasma membrane ATPase (Pma1) was responsible for the ATP buildup. Pma1 hydrolyzes 25- 40% of cellular ATP in yeast (Gradmann et al, 1978) and is heavily regulated by glucose (Serrano, 1983).
A constitutively active mutant form of PMA1 (pma1-∆916, here abbreviated pt) (Mason et al, 2014) was introduced into the endogenous PMA1 locus in the cellobiose utilizing strain. This mutation results in high Pma1 ATPase activity even in carbon starvation conditions (Mason et al, 2014). We observed a 66% decrease in cellular ATP levels in the pt strain in comparison to the wild-type control (WT, i.e. cellobiose pathway only) (Figure 2C). Phenotypically, the Ec of the mutant strain was 4 times that of the control (Figure 2B). In addition, the concentrations of G6P and F1,6BP decreased by 58% and 93%, respectively, relative to strains expressing wild-type PMA1. Notably, these concentrations dropped more than when the ATP-insensitive PFK1 mutant was introduced (Figure 2C). These results suggest that increased Pma1 ATPase activity improved cellobiose fermentation, by alleviating ATP accumulation that allosterically inhibited Pfk1 in glycolysis.
With decreased intracellular ATP concentrations in the pma1-∆916 strain, ATP inhibition at the Pfk1 step is unlikely to limit glycolysis. Consistent with this hypothesis, the cellobiose consumption profile of a pfk1m-pt double mutant was identical to that of the pt strain (Figure 2A). Surprisingly, the levels of G6P, F1,6BP and ATP in the double mutant strain were higher than in the pt strain (Figure 2C). In fact, their levels were similar to those in the pfk1m strain (Figure 2C). This may be because pfk1m has reduced activation by AMP and fructose 2,6-bisphosphate, in addition to reduced inhibition by high ATP concentrations (Rodicio et al, 2000). Taken together, these data suggest that the effect of reduced Pfk1 allosteric activation may only be evident in the absence of Pfk1 inhibition by ATP.
Carbon starvation-like state of the plasma membrane ATPase
We wondered whether the limited Pma1 activity in cellobiose-fed cells is due to the absence of extracellular glucose in the media. Transcriptionally, the presence of glucose increases PMA1 mRNA levels by 2-4 times via the regulation of Rap1, Gcr1 and Sir2 (Rao et al, 1993; García-Arranz et al, 1994; Kang et al, 2015). Consistent with the requirement for extracellular glucose sensing, previous RNA sequencing experiments revealed a 40% decrease in PMA1 transcript levels when cellobiose was provided in place of glucose (Lin et al, 2014). However, although transcriptional regulation of PMA1 is important, it is slower than post-transcriptional regulation and results in smaller changes (Serrano, 1983; Eraso et al, 2006). In the presence of glucose, phosphorylation of Ser-899 decreases Pma1’s Km and Ser-911/Thr-912 increases Pma1’s Vmax for ATP, respectively (Serrano, 1983; Portillo et al, 1991; Lecchi et al, 2007). Given the excess amount of ATP observed in cellobiose utilizing conditions (Figure 1), we hypothesized that Vmax-determining phosphorylation states of Ser-911 and Thr-912 might play a major role in establishing the efficiency of cellobiose fermentation. We did not investigate the phosphorylation of Ser-899, because the 6-fold excess of ATP likely overcame the potential problems associated with Km values (Eraso & Portillo, 1994).
Combinatorial mutations of Ser-911/Thr-912 to alanine and aspartic acid were introduced into the endogenous PMA1 gene to prevent or mimic phosphorylation, respectively. We were unable to obtain strains with pma1 S911A T912A and pma1 S911A T912D, potentially because the combinations were lethal. All mutant strains whose Pma1 S911 position was mutated to aspartic acid consumed cellobiose more efficiently in comparison to when the S911 position remained unchanged (Figure 3A). By contrast, mutating Pma1 T912 to aspartic acid did not show a correlation with the cellobiose consumption phenotype. These results suggest that phosphorylation of Pma1 at S911 was lacking when cellobiose was provided as a sole carbon source.
Positive effects of extracellular glucose sensor deletions
According to the above mutational analysis, the Pma1 phosphorylation state of cellobiose-fed cells was similar to that in carbon starvation conditions (Figure 3B) (Lecchi et al, 2007). Based on the relatively high level of intracellular glucose present in the cellobiose-grown cells (Figure 1), extracellular glucose is likely to be essential for full activation of Pma1 through S911 phosphorylation, as high intracellular glucose concentrations (Figure 1) or glucose metabolism (Rolland et al, 2002) was unable to satisfy this role.
Snf3, Rgt2 and Gpr1 are the three known sensors of extracellular glucose in S. cerevisiae. Snf3 and Rgt2 mainly regulate glucose transport while Gpr1 controls cell physiology via an interaction with Gpa2 to activate protein kinase A and cAMP synthesis (Rolland et al, 2002). To mimic the presence of extracellular glucose, constitutively active mutations (snf3 R229K, rgt2 R231K and gpa2 R273A) were introduced into the endogenous loci to probe the role of each glucose-sensing pathway (Ozcan et al, 1996; Xue et al, 1998). Surprisingly, the cellobiose consumption efficiency of all the three mutant strains decreased by ~25% (Figure S1A). We then inverted the genetic modifications by deleting SNF3, RGT2 and/or GPA2. Notably, the triple glucose-sensing deletion strain (snf3∆rgt2∆gpa2∆, or srg∆) showed a 275% increase in Ec (Figure 4A).
Combinatorial deletions revealed that the Gpr1 pathway did not contribute to improved cellobiose fermentation, but combining the SNF3 and RGT2 deletions (sr∆) was necessary and sufficient to replicate the Ec of the triple deletion strain (Figure 4A, Figure S1C). Consistent with the observed Ec values, the intracellular ATP levels of srg∆ and sr∆ decreased by 41% and 18% respectively, while those in the individual-deletion strains snf3∆ and rgt2∆ remained unchanged (Figure 4A). These results reveal a negative correlation between Ec and cellular ATP levels (Figure 4B) and that Snf3 and Rgt2 acted together to regulate cellular ATP levels, in addition to regulating glucose transport.
Although the additional deletion of GPA2 (gpa2∆) in the sr∆ strain did not further improve Ec (Fig 4A), it reduced the relative abundance of ATP by 28%, implying that the Gpr1 pathway has a separate mechanism to control cellular ATP levels that does not directly affect carbon metabolism. Consistently, gpa2∆ had a negative or neutral impact on cellobiose consumption (Figure S1). The decrease in ATP level may be a result of altered cellular activities, controlled by Gpr1 via the Tor and cAMP-PKA-Ras pathways (Rolland et al, 2002). The relationship between Gpr1-regulated ATP levels and carbon metabolism remains to be discovered. Since gpa2∆ did not have a direct effect on Ec, it was not investigated further in this study.
Snf3/Rgt2 regulation of cellular ATP levels
To examine whether Snf3/Rgt2 regulated the cellular ATP level in cellobiose fermentations via Pma1, an snf3∆ rgt2∆ pma1-∆916 strain (sr∆-pt) was constructed. Notably, the Ec of the sr∆-pt strain increased more than 4 times in comparison to the wild-type control (Figure 4A, C). The improvement was additive, within the range of the ∆Ec summation of sr∆ and pma1-∆916 strains relative to wild-type (Figure 4C). Although ATP levels decreased in a nearly linear fashion as a function of Ec (Figure 4B), it is not presently possible to ascertain whether the sr∆ and pma1-∆916 mutations act entirely independently due to limitations in measurement accuracy (Figure 4C).
To further determine the relationship between Snf3/Rgt2 and Pma1, the vanadate-specific ATPase activity of Pma1 (Viegas & Sá-Correia, 1991) from different strains consuming cellobiose were analyzed (Figure 5). Consistent with the constitutively active nature of the pt mutation, the activities of Pma1 in the pt and sr∆-pt strains were higher than those observed in the WT or sr∆ strains, respectively. Notably, the ATPase activitiy of Pma1 in sr∆ was comparable to the WT strain suggesting that Snf3 and Rgt2 in combination do not regulate Pma1 transcriptionally or post-translationally in a straightforward manner.
Transcriptional impact of Snf3/Rgt2 Deletion
Given the lower cellular ATP concentration (Figure 4A) but nearly unchanged Pma1 ATPase activity in the sr∆ strain relative to WT (Figure 5), Snf3/Rgt2 likely regulate ATP homeostasis through an additional mechanism other than Pma1. A systems-level analysis of yeast transcriptomes from the WT, s∆, r∆, and sr∆ strains identified 137 genes that were differentially expressed and unique to sr∆ (p-value < 0.001 and absolute fold change > 2 for sr∆ compared pairwise with WT, s∆ and r∆) (Dataset 1.3). Notably, 14 genes (10.2%) encoding ATP binding proteins and those involved in ATP catabolic processes were upregulated (Figure S2, Dataset 1.4). The 137 genes were further categorized into five clusters (Dataset 1.5), revealing upregulation of genes involved in cellular amino acid biosynthesis, metabolic processes and ATP catabolic processes, and downregulation of those in phospholipid biosynthesis, phospholipid metabolism, mitochondrial structure, negative regulation of gluconeogenesis and translation initiation factor eIF1 (SUI1) (Dataset 1.5, Figure S2). A more comprehensive analysis of genes involved in amino acid biosynthesis revealed a high correlation with upregulation of Gcn4, the central transcriptional activator of genes in this pathway (Figure S3, Dataset 1.7) (Moxley et al, 2009; Natarajan et al, 2001).
Discussion
To identify the effects of a minimal alteration to carbon metabolism in yeast, we chose a cellobiose-consumption pathway composed of two genes and analyzed its cellular metabolite profiles in comparison to cells provided with glucose, yeast’s preferred carbon source (Figure 6). Cellobiose consumption efficiency (Ec) linearly correlated with ethanol production rate, while ethanol yield remained mostly unchanged (Figure S4).
More than half of the metabolites changed abundance significantly when cellobiose was provided in place of glucose. The buildup of G6P, F6P and ATP in S. cerevisiae fermenting cellobiose suggested that Pfk1 was one of the bottlenecks in the process. Pfk1 is subjected to complex allosteric regulation, including inhibition by ATP and activation by AMP and fructose 2,6-bisphosphate (F2,6BP) (Bañuelos et al, 1977; Avigad, 1981; Nissler et al, 1983). The Pfk1 bottleneck was partially relieved in cells expressing an ATP/AMP/F2,6BP-insensitive PFK1 allele, while the ATP level remained elevated. These results contrast with previous studies that identified ATP depletion and the buildup of fructose-1,6-bisphosphate–the metabolite immediately downstream of Pfk1–as a weak link in glycolysis (van Heerden et al, 2014). Although we did not investigate the effect of AMP and F2,6BP activation since their changes between glucose and cellobiose conditions were less than 2 fold and not statistically significant (p-value ≥ 0.01), it is possible that they could influence cellobiose consumption efficiency.
Although ATP is central to a cell’s energy currency, too much ATP is not necessary beneficial (Pontes et al, 2015; Browne, 2013). In fact, we observed a negative correlation between cellular ATP levels and cellobiose consumption efficiency (Figure 4B). A similar correlation has been reported for glucose as a carbon source, suggesting metabolic uncoupling of energy homeostasis in yeast cells (Larsson et al, 1997). We propose that intracellular glucose concentrations–generated by cellobiose hydrolysis in our experiments–and glucose metabolism (Rolland et al, 2002) are insufficient to trigger glucose activation of key metabolic pathways and enzyme activity. For example, we found that the ATP-dependent proton pump Pma1 existed in a carbon-starvation like state during cellobiose fermentation, and was partially responsible for the aberrant accumulation of ATP. These results suggest that neither intracellular glucose nor glucose metabolism are sufficient to fully activate Pma1. A previous study showed the lack of phosphorylation of S899 and S911/T912 in Pma1, in a hexokinase/glucokinase deletion strain (hxk1∆ hxk2∆ glk1∆) provided with glucose, suggesting that phosphorylation of these residues requires glucose metabolism (Mazón et al, 2015). Together with our results, we propose that the activation of Pma1 through S911 phosphorylation requires both extracellular glucose and glucose metabolism. Our results reveal that the cellobiose utilization system allows uncoupling of glucose metabolism and intracellular glucose from extracellular glucose signaling. Future experiments will be required to reveal why ATP was not consumed by other cellular processes triggered under starvation (Thomsson et al, 2003).
Cytosolic pH is also a key regulator of carbon utilization (Dechant & Peter, 2014), and is likely to be impacted by the use of the proton symporter CDT-1 for cellobiose import and the resulting low activity of Pma1. High cytosolic pH is necessary and sufficient to activate Tor-Ras-PKA activities, which are downstream of Gpr1 glucose sensing pathway (Dechant & Peter, 2014). By contrast, the proton symporter CDT-1 and low activity of Pma1 may result in a low cytosolic pH. However, cytosolic pH alone is unlikely to determine the cellobiose consumption efficiency (Ec) as the strain with an ATP/AMP/F2,6BP-insensitive PFK1 allele showed improved Ec but unaltered cellular ATP levels. Furthermore, the inactivation of the Gpr1 pathway resulted in decreased cellular ATP but unaltered Ec. Glucose storage, i.e. in the form of trehalose, may be interconnected through the Gpr1 pathway because Ras-cAMP activates trehalase required to break down trehalose—a phenomenon observed when gluconeogenesis is switched to glycolysis (Thevelein & Hohmann, 1995). Trehalose cycling has been shown to lead to an imbalanced state of glycolysis (van Heerden et al, 2014). The relationship between carbon storage and Ec will require future studies to examine this relationship.
We also found that the well-studied extracellular glucose sensors Snf3/Rgt2 exhibited a novel role in cellular ATP homeostasis through an unknown mechanism, which seems to be independent of the major plasma membrane ATPase Pma1. Deletion of extracellular glucose sensors (Snf3/Rgt2) increased cellobiose consumption efficiency and partially restored ATP levels. It is known that deletion of SNF3 and RGT2 slows down glucose consumption (Choi et al, 2015), due to the inability of these strains to degrade Mth1/Std1, which block the promoter regions of hexose transporters required for optimal glucose import (Moriya & Johnston, 2004; Flick et al, 2003; Lakshmanan et al, 2003; Jouandot et al, 2011). Unlike glucose, cellobiose does not signal Mth1 degradation even with intact Snf3/Rgt2 (Jouandot et al, 2011). Thus, genes downstream of Mth1 regulation, including hexose transporters, are not expected to be responsible for the improved Ec and decreased cellular ATP levels observed in sr∆. Based on transcriptome profiling, the lack of Snf3 and Rgt2 increases ATP-requiring cellular activities through regulatory nodes other than Mth1. Consistent with this model, no growth defect is observed in a mth1∆ strain growing on glucose, suggesting that Snf3/Rgt2 has additional regulatory nodes other than Mth1 (Choi et al, 2015).
Regulation of ATP levels by the combined action of Snf3 and Rgt2 seems to involve transcriptional activation of amino acid biosynthetic pathways (Figure S3), although Snf3 and Rgt2 are glucose sensors (Rolland et al, 2002). In yeast, general amino acid control (GAAC) requires transcription factor Gcn4, which directly activates the amino acid biosynthetic pathways, as well as pathways for vitamins associated with amino acid synthesis in conditions of amino acid starvation or stress (Natarajan et al, 2001; Moxley et al, 2009). Notably, the transcription profile of the srΔ strain compared to WT reveals that pathways directly regulated by Gcn4 are induced during cellobiose fermentation, indicating that Gcn4 is activated in these conditions (Figure S3). Interestingly, we found that patterns of transcription activation of wild-type GCN4 relative to gcn4∆ strains correlate highly with srΔ relative to WT in cellobiose fermentations (correlations of 0.52 with (Moxley et al, 2009) and 0.68 with (Natarajan et al, 2001),respectively) (Figure S3).
The present results indicate that Snf3 and Rgt2 repress Gcn4 levels in the absence of glucose. GCN4 mRNA levels remain unchanged in srΔ relative to WT, indicating that Gcn4 repression by Snf3 and Rgt2 also likely occurs translationally (Hinnebusch, 1997).These results come as a surprise, as previous experiments showed glucose limitation induced GCN4 mRNA translation as a starvation response (Yang et al, 2000). Glucose limitation was shown to activate GCN4 mRNA translation by stimulating Gcn2 kinase activity (Yang et al, 2000), which is responsible for phosphorylating translation initiation factor eIF2 and depleting active ternary complexes of eIF2 with initiator tRNA and GTP (Hinnebusch, 1997). By contrast, a chemical genetics approach identified the kinase Snf1, the yeast homologue of human AMP-activated protein kinase (AMPK), as a repressor of GCN4 mRNA translation in conditions with a non-glucose sugar and abundant amino acid availability (Shirra et al, 2008), conditions more in line with those used here.
As a homolog of the important energy sensor AMPK that maintains energy homeostasis in the cells (Hardie, 2011), it is tempting to speculate that Snf1 plays a role in ATP regulation downstream of Snf3/Rgt2, and potentially connects to Gcn4. Snf1 is required for expression of glucose-repressed genes and may be regulated by mitochondrial starvation signaling (Li et al, 2013). Mitochondrial activation observed in cellobiose fermentation compared to glucose fermentation (Lin et al, 2014) may be a readout of the Snf1 regulation. The direct relationships between Snf3/Rgt2, Snf1, Gcn4, mitochondrial activation and cellular ATP remains to be investigated.
One possible additional mechanism of activation of GCN4 mRNA translation could be depletion of translation initiation factor 1 (eIF1, encoded by SUI1). In the srΔ strain, levels of SUI1 mRNA decreased nearly 2.6-fold. Consistent with a direct role for eIF1, previous results demonstrated that activation of GCN4 translation decreases with increased functional eIF1 levels (Hinnebusch, 1997; Herrmannová et al, 2012). Notably, SUI1 transcript levels were also lower in GCN4 cells when compared to gcn4Δ strains in starvation conditions (Natarajan et al, 2001). Future experiments will be required to demonstrate a direct link between eIF1 levels and GCN4 mRNA translation activation.
The present systems-level study of a minimal synthetic biology pathway for cellobiose consumption revealed the dramatic impact of decoupling extracellular and intracellular glucose sensing, resulting in an overabundance of ATP in cells. The inability of S. cerevisiae to catabolize ATP for cellular processes in the presence of intracellular glucose and glucose metabolism but in the absence of extracellular glucose resulted in slow fermentation. Thus, ATP levels must be kept in a relatively narrow range for optimal fermentation and to allow robust startup of glycolysis, yet yeast seems to lack a direct mechanism to monitor ATP concentrations. For example, a dynamic model showed that a small concentration difference of inorganic phosphate, a product of ATP hydrolysis, could alter cell fate from a stable glycolytic steady state to an imbalanced dead-end state (van Heerden et al, 2014). Here, we found that the extracellular glucose sensing by Snf3/Rgt2 required for optimal glucose fermentation (Youk & van Oudenaarden, 2009) can be uncoupled from the role of these receptors in regulating ATP homeostasis under carbon starvation conditions. It will be important in the future to map the fully regulatory pathways of ATP homeostasis leading from Snf3/Rgt2 and, independently, terminating in Pma1, along with the mechanisms connecting carbon utilization to the translation of the master transcriptional regulator Gcn4.
Materials & Methods
Yeast strains, media and anaerobic fermentation
The S. cerevisiae background strain used in this study was S288C ura3::PPGK1- cdt-1 N209S/F262Y-TADH1 lyp1::PTDH3-gh1-1 (codon optimized)-TCYC1 derived by chromosomal DNA library selection (Ryan et al, 2014). The strain was subjected to further modifications using the CRISPRm system (Ryan et al, 2014). The list of strains constructed, CRISPRm guides and primers used are included in Table S1.
Seed cultures for cellobiose fermentation experiments were grown in 20 g/L glucose modified optimal minimal media (oMM) (Lin et al, 2014) and harvested at mid-log phase. All cellobiose fermentation experiments were conducted under strict anaerobic conditions, in 80 g/L cellobiose oMM media at an initial OD600 of 20, using 10 mL serum flasks containing 5 mL fermentation in 2-5 biological replicates. The flasks were incubated at 30°C, 220 rpm. The cellobiose consumption efficiency (Ec) was defined as the inverse of the area under the curve of extracellular cellobiose concentration over time.
Analytical analysis of yeast metabolites
Extracellular cellobiose concentrations were determined by high performance liquid chromatography on a Prominence HPLC (Shimadzu, Kyoto, Japan) equipped with Rezex RFQ-FastAcid H 10 × 7.8 mm column. The column was eluted with 0.01 N of H2SO4 at a flow rate of 1 mL/min, 55°C.
For the metabolite profiling comparison between glucose and cellobiose (modified from (Benjamin et al, 2014)), equal amounts of yeast cells at mid-exponential phase of anaerobic sugar consumption (10 g/L cellobiose or glucose) were harvested (final pellet OD600 equivalent to 5). The samples were quenched in 180 µL of 40:40:20 acetonitrile:methanol:water. Following the addition of 10 nmols of d3 serine (as an internal standard), the mixtures were vortexed and centrifuged at 13,000 rpm for 10 minutes. The supernatants were injected onto an Agilent 6460 QQQ LC-MS/MS and the chromatography was achieved by normal phase separation with a Luna NH2 column (Phenomenex) starting with 100% acetonitrile with a gradient to 100% 95:5 water acetonitrile. 0.1% formic acid or 0.2% ammonium hydroxide with 50 mM ammonium acetate was added to assist with ionization in positive and negative ionization mode, respectively. Five biological replicates were used for each sample analyzed.
For targeted intracellular metabolite comparisons, yeast cells equivalent to 20 OD600 units were harvested and filtered through a 0.8 µm nylon membrane, prewashed with 3 mL water, followed by another 3 mL water wash after cell filtration. The membranes were placed in 1.5 mL extraction solution (0.1 M formic acid, 15.3 M acetonitrile) flash-frozen in liquid nitrogen, and stored at −80°C. Before analysis, the extracts were vortexed for 15 minutes and centrifuged to collect the supernatants at 4°C. Glucose 6-phosphate and fructose 1,6-bisphosphate were separated and identified using a 1200 Series liquid chromatography instrument (Agilent Technologies, Santa Clara, CA). 1 µL of each sample was injected onto an Agilent Eclipse XDB-C18 (2.1 mm i.d., 150 mm length, 3.5 µm particle size) column with a Zorbax SB-C8 (2.1 mm i.d., 12.5 mm length, 5 µm particle size) guard column and eluted at 25 °C and a flow rate of 0.2 mL/min with the following gradient (modification from (Luo et al, 2007)): 15 min isocratic 100% buffer A (10 mM tributylamine/15 mM acetic acid), then in 15 min with a linear gradient to 60% buffer B (methanol), 2 min isocratic 60% B, then 10 min equilibration with 100% buffer A. The eluent from the column was introduced into a mass spectrometer for 25 minutes after the first 10 minutes. Mass spectrometry (MS) was performed on an LTQ XL ion trap instrument (Thermo Fisher Scientific, San Jose, CA) with an ESI source operated in negative ion mode. The MS settings were capillary temperature 350 °C, ion spray voltage 4.5 kV, sheath gas flow: 60 (arbitrary units), auxiliary gas flow 10 (arbitrary units), sweep gas flow 5 (arbitrary units). For the MS/MS product ion scan, the scan range was m/z 80 to m/z 300. The compounds G6P at m/z 259.1 and F16BP at m/z 339.1 were isolated with an m/z 2 isolation width and fragmented with a normalized collision-induced dissociation energy setting of 35% and with an activation time of 30 ms and an activation Q of 0.250.
Plasma membrane isolation
Strains were subjected to cellobiose fermentation under anaerobic conditions. Yeast cells with an OD600 equivalent to 40 were harvested at mid-log phase and flash frozen in liquid nitrogen. Membrane fractions were extracted based on the protocol published in (Kaiser et al, 2002).
Pma1 ATPase activity assay
The ATPase assay described in (Viegas & Sá-Correia, 1991) was modified as follows. 30 µg of the isolated membrane fraction was incubated in assay buffer (50 mM MES pH 5.7, 10 mM MgSO4, 50 mM KCl, 5 mM NaN3, 50 mM KNO3) with and without 3 mM orthovanadate for 25 minutes at 30 °C. 1.8 mM ATP was added to start the 100 uL reactions. The reactions were incubated at 30 °C for 15 minutes, then the membranes were isolated from the reactions by centrifugation at 13,000x g for 10 minutes at 4°C. The released inorganic phosphate was measured in the supernatant using the ATPase/GTPase Activity Assay Kit (Sigma-Aldrich) based on the manufacturer’s protocol. The specific Pma1 ATPase activities were calculated by subtracting the concentration of released inorganic phosphate in reactions provided with orthovanadate from those without.
RNA Sequencing
Total RNA was extracted from yeast cells equivalent to 40 OD600 units harvested at mid exponential phase by the RiboPure Yeast kit (Thermo Fisher Scientific). Multiplexed libraries were then derived from the total RNA using TruSeq RNA Library Preparation Kits v2 (Illumina Sequencing). The quality of the library was evaluated by Agilent Bioanalyzer 2000 (Functional Genomics Laboratory, UC Berkeley) and sequenced with Illumina HiSeq2500 RAPID platform (Vincent J. Coates Genomic Sequencing Laboratory, UC Berkeley). The reads were assembled and analyzed in CLC Genomics Workbench 9 (Denmark). RPKM (Reads Per Kilobase of exon model per Million mapped reads) was chosen as the expression value output, which was further normalized by the total expression values in a sample. Functional classification was performed using FunSpec (Robinson et al, 2002).
Yeast cell-based cellobiose uptake assay
The cell-based cellobiose uptake assay was modified from (Li et al, 2015). Yeast strains were grown to mid-exponential phase in 2% oMM glucose, washed with assay buffer (5 mM MES, 100 mM NaCl, pH 6.0) three times and resuspended to a final OD600 of 10. Equal volumes of the cell suspension and 200 µM cellobiose were mixed to start the reactions, which were incubated at 30°C with continuous shaking for 15 minutes. The reactions were stopped by adding 150 µL of supernatants to 150 µL 0.1 M NaOH. The concentrations of the remaining cellobiose were measured using an ICS-3000 Ion Chromatography System (Dionex, Sunnyvale, CA, USA) equipped with a CarboPac® PA200 carbohydrate column. The column was eluted with a NaOAc gradient in 100 mM NaOH at a flow rate of 0.4 mL/min, 30°C.
Author contributions
All authors contributed to the design of the experiments. KC and DIB carried out the experiments. DKN and JHDC aided in interpretation of the systems-level experiments. KC wrote the manuscript, with editing from DIB, DKN, and JDHC.
Acknowledgements
The authors thank Dr. Xin Li, Dr. Ligia Acosta-Sampson, Dr. Yuping Lin and Dr. Matt Shurtleff for helpful discussions. This work was supported by funding from Energy Biosciences Institute to JHDC and from National Institutes of Health (R01CA172667) to DKN.
Footnotes
KC: kchomv{at}gmail.com DIB: dbenjami{at}stanford.edu DKN: dnomura{at}berkeley.edu JHDC: jcate{at}lbl.gov
Conflict of Interest
The authors declare that they have no conflict of interest.