Growth on monomeric sugars
In order to manipulate the intracellular level of glycolytic intermediates, and G6P in particular, PGI1 gene was deleted in three recombinant d-xylose-fermenting strains carrying fluorescent sugar signaling biosensors. The strains carried either the HXT1p-GFP biosensor (TMB3752), the SUC2p-GFP biosensor (TMB3755) or the TPS1p-GFP (TMB3757) biosensor, reporting on the Snf3p/Rgt2p, the SNF1/Mig1p, and the cAMP/PKA sugar signaling pathways, respectively. Details on strain construction and validation can be found in Additional file 1: S1.
Saccharomyces cerevisiae strains with inactive Pgi1p are known to lose their ability to grow on single monomeric sugars such as d-glucose and d-fructose [24, 25]. Abolishment of growth on d-glucose has been attributed to three potential factors: (1) the imbalance caused when ATP is consumed during initial glucose phosphorylation but not regenerated further down in glycolysis [26], (2) an accumulation of toxic levels of G6P [26], and (3) the inability to channel carbon into the lower glycolysis for precursor production and NADPH reoxidation as the flux through the PPP is considered too low [20, 25, 27]. Similarly, growth on d-fructose as sole carbon source has not yet been reported, likely because G6P is needed for anabolic processes and can only be generated by gluconeogenesis via Pgi1p (Fig. 1). Indeed, PGI1 deletants have been reported to grow on d-fructose supplemented with a small amount (1 g L−1) of d-glucose [28]. In order to screen the growth response of the PGI1 deletants to a larger number of carbon source combinations, including d-xylose, micro-scale cultivations were carried out in 96-well microplates (Fig. 2).
Cultivation in microplates is typically limited by incomplete aeration, medium evaporation, and non-linear correlations between OD620 and biomass. Our experiments were performed in quadruplicates and the aforementioned limitations did not appear to have interfered with the determination of lag phase duration and growth in microtiter plates (Fig. 3). The longer lag phase seen for the PGI1 deletants in aerobic shake flasks containing rich YPFG medium (Additional file 1: Fig. S2) was also observed in the microscale cultures in defined YNB-FG medium, albeit with larger variations (Fig. 2A). In the PGI1-wildtype strains, exponential growth on single fermentable sugars (d-fructose, d-glucose and d-galactose) started almost immediately (Figs. 2B–D, 3). Growth was also recorded on d-xylose for the PGI1-wildtype strains (engineered with the XR/XDH pathway), albeit with a longer lag phase (22.7 h) compared to other monomeric sugars (8.3–13 h) and with a linear growth pattern (Figs. 2E, 3).
Overall, the PGI1 deletants displayed a significant increase in lag phase duration compared to the PGI1-wildtype strains. Surprisingly, and in contrast to previous reports, the PGI1 deletants were able to grow on monomeric d-glucose, d-galactose and d-fructose, albeit with prolonged lag phases compared to PGI1-wildtype strains. Additionally, an increased heterogeneity between biological replicates was observed (Figs. 2B–E, 3), as not all biological replicates of the PGI1 deletants initiated growth. A possible explanation for the growth of PGI1 deletants on monomeric sugars may lie in the genetic makeup of the PGI1-wildtype strains: the strains have been engineered to ferment d-xylose by introduction of heterologous XR and XDH genes, overexpression of a d-xylose-selective mutated Gal2p transporter [29], and by overexpression of the PPP genes TKL1 and TAL1 [30]. The latter modification is likely of importance for the growth on monomeric d-glucose since the PPP flux is thought to be one of the limiting factors in PGI1 deletants [28]. In fact, a doubling of the activities of 6-phosphogluconate dehydrogenase (GND), ribulose-5-phosphate-3-epimerase (RPE) and transaldolase (TAL) (Fig. 1) has been shown to be one of the mechanisms used in suppressor mutants of PGI1 deletant strains [31]. Additionally, PGI1 deletants of the yeast Kluyveromyces lactis were able to grow on d-glucose only, but lost this ability upon additional disruptions in the PPP [32]. Similarly, the overexpressed PPP might explain the growth seen in d-galactose media. d-Galactose enters the yeast central metabolism at the level of G6P through isomerization, phosphorylation, UDP-transfer and phosphate transfer (Fig. 1). This assimilation requires UDP-d-glucose, which has been shown to accumulate in PGI1 deletants pre-grown on YPFG [33]. In conjunction with higher flux through the PPP, this may also explain the growth observed on sole d-galactose in the PGI1 deletants.
In contrast, upregulation of PPP genes cannot explain growth on d-fructose in the d-xylose-engineered strains of the present study, since G6P cannot be generated under these conditions. In a previous study, Corominas and colleagues observed that during early exponential growth on YPFG media, PGI1 deletants accumulated G6P, UDP-d-glucose and the storage carbon glycogen [33]. Since the pre-cultures of the present study all consist of rich YPFG medium, it is possible that the cells being inoculated into d-fructose medium had sufficient storage of glycogen to be broken down into d-glucose-6-phosphate (via d-glucose and d-glucose-1-phosphate; Fig. 1) to support growth on d-fructose. The growth phase of the pre-culture may be crucial as glycogen was shown to decrease fourfold between early and late exponential phase [33], which might offer an explanation, together with the benefit of the upregulated PPP, to why other studies also employing YPFG as pre-culture did not observe growth on d-fructose for their PGI1 deletants.
In the cultivations containing d-xylose as a single carbon source, no growth was recorded for the PGI1 deletants (Fig. 2E), whereas the addition of 1 g L−1 d-glucose enabled some growth in two of the replicates, but with a lag phase of over 60 h (Fig. 2F). d-Xylose is expected to enter glycolysis at the F6P and G3P nodes (Fig. 1), implying PGI1 deletants could theoretically grow on d-xylose through the same mechanisms as on d-fructose. However, the lower carbon flux and limited glycogen production might exceed the maintenance needs and prevent growth on d-xylose.
Targeted metabolite profiling reveals systemic changes in intracellular sugar phosphate levels after PGI1 deletion
To further study the phenotype of PGI1 deletants assimilating various carbon sources, concentrations of intracellular sugar phosphates were determined in the PGI1-wildtype and PGI1 deletion strains carrying each biosensor. This analysis focused on sugar phosphates from the glycolysis since several of these have been implicated in the regulation of sugar metabolism S. cerevisiae [19, 22]. The metabolite profiling also included other sugar phosphates with possibly unknown regulatory roles from closely related pathways such as the pentose phosphate pathway, the Leloir pathway, and the trehalose pathway.
First, a principal component analysis (PCA) was conducted to generate an overview of alterations in metabolite levels associated with the PGI1 deletion and variation in the carbon source. This analysis revealed systematic changes in intracellular sugar phosphate accumulation in the PGI1-wildtype and deletant strains on the two different carbon sources. The first and second principal components (PCs), PC1 and PC2, accounted for 39.1% and 33.4% of the variation in the data, respectively (Fig. 4; Scree plot is shown in Additional file 1: Fig. S3). The score scatter plot for the two first PCs revealed a clear impact of both the sugar used during the incubation (d-glucose vs. d-xylose) and the genotype of the strain (PGI1-wildtype vs. PGI1 deletants) on levels of sugar phosphates (Fig. 4).
Next, specific alterations in intracellular metabolite levels were investigated. In general, the PGI1-wildtype strains showed lower levels of intermediates when grown on d-xylose compared to d-glucose (Fig. 5A), as expected from the lower flux generally observed through the d-xylose pathway compared to the glycolysis. For the PGI1 deletant strains, a media-dependent effect on metabolite levels was observed, with higher levels of metabolites upstream of Pgi1p on d-glucose (G6P; T6P; G1P, d-glucose-1-phosphate) and higher levels of metabolites downstream of Pgi1p on d-xylose (F6P; S7P, sedoheptulose-7-phosphate; DHAP, dihydroxyacetone phosphate) (Fig. 5B).
In 20 g L−1 d-glucose, PGI1 deletants showed a hyper-accumulation of G6P and accumulation of intermediates formed from G6P through glycogen and trehalose synthesis pathways (G1P and T6P, respectively) (Fig. 5C). Elevated levels of 6-phosphogluconate (6PG) and erythrose-4-phosphate (E4P) were detected in the d-glucose media compared to the d-xylose media (Fig. 5B). Formation of these two intermediates (6PG and E4P) indicates that the PPP is active, since they cannot form via the inactivated Pgi1p, which supports the proposed mechanism mentioned earlier whereby the increased expression of PPP in our engineered strains is enabling the unprecedented growth on monomeric sugars for PGI1 deletants. The deletion strains also showed decreased concentrations of intermediates downstream of Pgi1p (F1,6bP; F6P; DHAP; G3P; 3PG, 3-phosphoglycerate) which could be expected as a direct consequence of the PGI1 deletion (Fig. 5C). Similarly, a decrease in 6PG, as well as downstream intermediates such as S7P and E4P, was also observed, confirming that the oxidative PPP flux is limited in G6P-accumulating PGI1 deletants [26, 27]. Although the data showed that concentrations of intermediates downstream of the Pgi1p reaction were lower in the deletant strains compared to the PGI1-wildtype strains, the compounds were not fully depleted (Additional file 1: Fig. S4).
In both the PGI1-wildtype and deletion strains, d-xylose was converted via the XR/XDH pathway, followed by the non-oxidative PPP after which the carbon entered the lower glycolysis at the F6P and G3P nodes (Fig. 1). Consequently, PGI1-wildtype strains were expected to form upstream intermediates (G6P, T6P, G1P, 6PG) from F6P via Pgi1p when incubated on d-xylose, while deletant strains were not. Indeed, formation of these upstream sugar phosphates was observed on d-xylose for the PGI1-wildtype strain but not for the deletants (Fig. 5D). Accumulations of F6P and S7P were recorded in the deletants on d-xylose, which could result from the overall decreased glycolytic flux leading to accumulation of intermediates and the ability of the PGI1-wildtype strain to convert F6P into intermediates upstream of the Pgi1p reaction such as G6P. Curiously, d-galactose-1-phosphate (Gal1P) was observed to accumulate in the PGI1 deletant strains. Given the decrease in G6P levels, deletant strains would be expected to also consume Gal1P via the Leloir pathway as the flux is directed to G6P formation via G1P (Fig. 1). The persistence of Gal1P in the deletant strains might be explained by a lack of UDP-d-glucose, which is required for its integration into the glycolysis. Previous studies have reported the accumulation of glycogen in PGI1 deletants [33], which consumes UDP-d-glucose and thus may partly explain the Gal1P accumulation seen on d-xylose. Unfortunately, it was not possible to determine the UDP-d-glucose levels in this experiment as it could not be distinguished from UDP-d-galactose.
Sugar signaling responses and further indications of d-xylose not being recognized as fermentable
The yeast sugar signaling response was explored in the PGI1-wildtype and PGI1 deletant strains by recording the activity of biosensors previously constructed to report on the three main sugar signaling routes [17, 18]. We first attempted to plot the biosensor responses directly as a function of the intracellular sugar phosphate levels, using combined data from both strains and carbon sources. The biosensor responses to d-glucose were found to have similarities to the biosensor responses for G6P (Fig. 6). However, correlations between biosensor responses and other intracellular sugar phosphates proved difficult to explain (Additional file 1: Fig. S5), likely due to the time separation between metabolite sampling (30 min of incubation) and optimal GFP expression (6 h).
To further understand the sugar signaling responses, the strains were instead evaluated in d-glucose, d-xylose, d-fructose, and d-galactose at various concentrations, similar to a previous experiment performed by Osiro et al. (2018) [18]. The strains were incubated in the sugar concentrations previously evaluated for growth (20 g L−1) as well as in concentrations that have been used in previous sugar sensing studies (50 g L−1 and 1 g L−1) [17, 18, 34]. No significant differences in fluorescent intensities were recorded between 20 and 50 g L−1 d-xylose for any of the strains and conditions (see Fig. 7), indicating that 20 g L−1 was sufficient to elicit a similar signaling response as the one previously observed at 50 g L−1 [18]. Notably, since 50 g L−1 d-xylose elicits signals resembling those on low levels of d-glucose [18], 1 g L−1 d-glucose and 1 g L−1 d-galactose were also included as conditions in this study for comparison. Although we made attempts to repress the sensors prior to the signaling assays (to produce fold change values rather than fluorescent intensities), repression of the SUC2p and TPS1p sensors was not achievable due to the d-glucose toxicity seen in PGI1 deletants. Consequently, all strains were pre-incubated on a mixture of 20 g L−1 d-fructose with 1 g L−1 d-glucose prior to inoculation into the various conditions. Although minor subpopulations were observed in some fluorescence histograms for the PGI1-wildtype strains (Additional file 1: Fig. S6), they did not alter the interpretation of the results and were included in the average fluorescent intensities shown in Fig. 7.
In the PGI1-wildtype strains, the expected biosensor responses were recorded, i.e. HXT1p induction at high d-glucose concentrations (20 g L−1) as well as repression at low-to-no d-glucose; full induction of SUC2p at low concentrations of d-glucose (1 g L−1), with repression at higher levels (> 5 g L−1) and only basal induction in carbon-free media; and TPS1p repression in response to preferred carbon sources such as d-glucose, d-fructose, and d-galactose (Fig. 7) [16]. Additionally, we observed the previously reported induction of SUC2p on d-xylose [18] as well as TPS1p induction, indicating both a carbon starvation response and a decrease in PKA activity, which further supports the notion that d-xylose is not recognized as a fermentable carbon source [35,36,37,38]. The combination of 20 g L−1 ethanol with 20 g L−1 glycerol resulted in a biosensor expression pattern that was also strikingly similar to the one observed on 20 g L−1 d-xylose (repressed HXT1p, induced SUC2p, and induced TPS1p). In both conditions, the assimilated carbon sources enter glycolysis downstream of the Pgi1p reaction (F6P and G3P for d-xylose, DHAP and PEP for glycerol and ethanol), which could indicate a connection between the level of, or flux through, some of these downstream intermediates and the observed response. Although in opposition of this, the condition containing 20 g L−1 d-fructose, which also enters glycolysis downstream of the Pgi1p reaction, instead showed a SUC2p and TPS1p response that was closer to that seen on 20 g L−1 d-glucose rather than d-xylose. Notably, the low HXT1p response seen on 20 g L−1 d-fructose aligned with previous data which indicated weaker PKA-induced phosphorylation of Rgt1p on d-fructose than on d-glucose [39].
Deletion of PGI1 alters the sugar signaling response
The PGI1 deletion resulted in changes to the biosensor response for nearly all conditions (Fig. 7). Induction of the HXT1p reporter was maintained at high d-glucose level, but it was now observed at low d-glucose levels as well, which points towards a role of G6P and/or T6P on HXT1p induction. Even higher induction of HXT1p was observed in the combinations of low d-glucose with d-fructose or low d-glucose with d-xylose, whereas neither high d-fructose nor high d-xylose alone led to induction. One possible explanation is that a certain combination of intermediates (either at certain levels or at certain rates of formation) upstream (G6P/T6P) and downstream (F6P/F1,6bP) of the Pgi1p reaction is necessary to achieve this induction.
The TPS1p biosensor remained repressed by d-fructose in the PGI1 deletants but a weaker repression was observed on both d-glucose and d-galactose. TPS1p expression is repressed by PKA activity [16, 40] and consequently the persistence of repression in the d-fructose condition likely reflects an active PKA. The cAMP/PKA pathway has been hypothesized to be controlled directly by intracellular metabolite concentrations, rates of formation, and metabolic fluxes in addition to the established impact of the extracellular Gpr1p sensor and the intracellular Ras1/2p components [16, 41]. For instance, PKA has been shown to be activated by F1,6bP [19], which is formed from d-fructose via F6P but cannot be formed in PGI1 deletants on d-glucose or d-galactose. The formation of F1,6bP could explain why TPS1p remains repressed when supplied with d-fructose, but not when given d-glucose or d-galactose. Interestingly, the TPS1p and HXT1p biosensors both show the highest increase in fluorescence in response to the mix of 1 g L−1 d-glucose with 50 g L−1 d-xylose and to the mix of 5 g L−1 d-glucose with 20 g L−1 d-fructose in the PGI1 deletants. The simultaneous induction of both these sensors at once is quite unexpected since the full induction of HXT1p is dependent on hyperphosphorylation of the Rgt1p transcription factor by active PKA; hence full HXT1p induction is expected only when TPS1p is repressed by PKA activity [16, 40, 42].
The SUC2p biosensor showed an overall increase in fluorescence in the PGI1 deletant compared to the PGI1-wildtype strain, with the exception of the 1 g L−1 d-glucose condition where a lower signal was observed. Compared to the condition lacking a carbon source, the PGI1 deletants still showed repression in response to high concentrations of d-glucose (20 g L−1), and now also displayed the same repression in response to low concentrations of the sugar (1 g L−1) likely due to the accumulation of G6P. These findings are in line with the proposed role of G6P as a key regulator of glucose repression: G6P acts via the SNF1/Mig1p pathway, likely by dephosphorylating SNF1 through the Reg1p-Glc7p phosphatase via an unknown signaling mechanism, which ultimately leads to the repression of SUC2p and other genes [16, 22, 43, 44]. As such, the increased concentration of intracellular G6P seen for PGI1 deletants in response to d-glucose is expected to lead to SUC2p repression. Counterintuitively, this was not observed in growth conditions including d-galactose, however this may be due to the lower concentration (1 g L−1) and the less rapid accumulation of G6P via the Leloir pathway [45]. The SUC2p repression was relieved on d-xylose and ethanol-glycerol media, further confirming that d-xylose is not sensed as a repressing fermentable sugar. Perhaps even more interestingly, repression was also relieved in the d-glucose/d-xylose medium (1 g L−1 + 50 g L−1) and maintained in the d-fructose (20 g L−1) medium. This hints towards a role of metabolites downstream of G6P on catabolite repression, as G6P is not expected to form during the utilization of d-fructose nor d-xylose in the PGI1 deletant.
The cause of the TPS1p induction on mixtures of d-glucose with either d-xylose or d-fructose is not known, but it might indicate that TPS1p has become partially deregulated from PKA activity in PGI1 deletants. This would also be in line with the results showing the curious co-induction of HXT1p and TPS1p mentioned earlier. It has been hypothesized that an increase in trehalose synthesis, and consequently TPS1p expression, might act as a way for the cell to free up inorganic phosphates (bound in the form of G6P) in these deletants [46]. However, given that d-glucose and d-galactose are both expected to result in G6P accumulation in the PGI1 deletant, this does not explain why TPS1p is induced in the mixed 50 g L−1 d-xylose media containing 1 g L−1 d-glucose, but not in the mixed media containing 1 g L−1 d-galactose. Possibly, the elevated TPS1p induction seen on d-glucose relies partly on the activation of the extracellular Gpr1p d-glucose sensor and partly on the intracellular G6P levels. Alternatively, there may be differences in the flux rates and final metabolites formed when incubated in d-glucose compared to d-galactose (for instance the G6P formation rate may be too low on d-galactose). Additional experiments investigating the signaling response and changes in metabolite levels while varying concentrations of sugars that enter upstream and downstream of Pgi1p might shed more light on the causes of this peculiar signaling state.
The putative role of d-fructose-bisphosphate regulation on d-xylose utilization
The PGI1 deletants have previously been reported to accumulate F1,6bP from d-fructose [43]. However, in the metabolite profiling of the present study (Fig. 5D) accumulation of F6P from d-xylose was instead observed. Hence, it is possible that the differences in sugar signaling seen between d-xylose and d-fructose may be linked to d-fructose phosphate levels or formation rates (in addition to the repressive effect of G6P). F1,6bP is synthesized from F6P by phosphofructo-1-kinase (Pfk1p) and can be converted back to F6P by d-fructose-1,6-bisphosphatase (Fbp1p) [47]. The activities of both Pfk1p and Fbp1p are allosterically regulated by d-fructose-2,6-bisphosphate (F2,6bP), a metabolite synthesized by Pfk26/27p, which activates Pfk1p and inactivates Fbp1p [47,48,49]. It has been shown that the PFK27 gene is induced by d-fructose but not by d-xylose [49], which could lead to differences in F2,6bP levels in these two conditions. In extension, the decreased levels of F2,6bP might cause the decrease in F1,6bP levels on d-xylose via Pfk1p/Fbp1p regulation. Interestingly, increased levels of F1,6bP have been shown to enhance PKA activity [19]. Thus, the variation in F1,6bP levels might also affect the regulation of PKA differently between the two conditions (Fig. 8). This is of particular interest, since low PKA activity has been pointed out as a possible component that results in poor d-xylose utilization [16, 34, 37, 50]. Additionally, activation of PKA by F1,6bP in the d-fructose condition may lead to the inactivation of the SNF1 kinase and thus the observed SUC2p repression. Conversely, the lack of F1,6bP may lead to the induction of SUC2p seen on d-xylose, and consequently the expression of suboptimal catabolic genes which further impedes strain performance. To investigate the potential role of F2,6bP on poor d-xylose utilization, one possibility would be to deregulate PFK26/27 genes in d-xylose-utilizing strains and measure both the F2,6bP levels and the sugar signaling over time on different carbon sources. An increased PFK26/27 expression on d-xylose would be expected to result in a similar sugar signaling state as seen on d-fructose, and potentially also lead to an improved performance on d-xylose. Indeed, in a previous study by Shen and colleagues, PFK27 overexpression was found to be one of the changes that arose during adaptive laboratory evolution for improved d-xylose utilization [51]. However, reintroduction of PFK27 overexpression in the parental strain did not show improved growth in the studied conditions and strain [51], and further testing of the potential benefits of PFK27 has yet to be performed. Future experiments exploring this topic could investigate the hypothesized decrease of F1,6bP levels in d-fructose media after PFK27 deletion in PGI1 deletants, and the potential increase of F1,6bP levels in d-xylose media upon PFK27 overexpression. Additionally, since we hypothesize that the cAMP/PKA pathway will show increased activity upon overexpression of PFK27, the sugar signaling and d-xylose utilization of PGI1-wildtype strains carrying this mutation should be examined as well.