Remodeling of Carbon Metabolism during Sulfoglycolysis in Escherichia coli

ABSTRACT

Sulfoquinovose (SQ) is a major metabolite in the global sulfur cycle produced by nearly all photosynthetic organisms. One of the major pathways involved in the catabolism of SQ in bacteria such as Escherichia coli is a variant of the glycolytic Embden-Meyerhof-Parnas (EMP) pathway termed the sulfoglycolytic EMP (sulfo-EMP) pathway, which leads to the consumption of three of the six carbons of SQ and the excretion of 2,3-dihydroxypropanesulfonate (DHPS). Comparative metabolite profiling of aerobically glucose (Glc)-grown and SQ-grown E. coli cells was undertaken to identify the metabolic consequences of the switch from glycolysis to sulfoglycolysis. Sulfoglycolysis was associated with the diversion of triose phosphates (triose-P) to synthesize sugar phosphates (gluconeogenesis) and an unexpected accumulation of trehalose and glycogen storage carbohydrates. Sulfoglycolysis was also associated with global changes in central carbon metabolism, as indicated by the changes in the levels of intermediates in the tricarboxylic acid (TCA) cycle, the pentose phosphate pathway (PPP), polyamine metabolism, pyrimidine metabolism, and many amino acid metabolic pathways. Upon entry into stationary phase and the depletion of SQ, E. coli cells utilize their glycogen, indicating a reversal of metabolic fluxes to allow glycolytic metabolism.

IMPORTANCE The sulfosugar sulfoquinovose is estimated to be produced on a scale of 10 billion metric tons per annum, making it a major organosulfur species in the biosulfur cycle. The microbial degradation of sulfoquinovose through sulfoglycolysis allows the utilization of its carbon content and contributes to the biomineralization of its sulfur. However, the metabolic consequences of microbial growth on sulfoquinovose are unclear. We use metabolomics to identify the metabolic adaptations that Escherichia coli undergoes when grown on sulfoquinovose versus glucose. This revealed the increased flux into storage carbohydrates through gluconeogenesis and the reduced flux of carbon into the TCA cycle and downstream metabolism. These changes are relieved upon entry into stationary phase and reversion to glycolytic metabolism. This work provides new insights into the metabolic consequences of microbial growth on an abundant sulfosugar.

INTRODUCTION

Sulfoquinovose (SQ) is a 6-sulfonated analogue of glucose (Glc) that is produced by most photosynthetic organisms (Fig. 1a) (1, 2). SQ occurs primarily as the polar head group of sulfoquinovosyl diacylglyceride (SQDG), a major glycolipid in plant chloroplasts and cyanobacteria (1, 3). The global level of production of SQ is estimated to be in the order of 10 billion metric tons per annum (4). Consequently, SQ is a major compound in the biogeochemical sulfur cycle. While the biosynthesis of SQ and SQDG occurs within plants and cyanobacteria (1, 5), its catabolism appears to be mediated exclusively by bacteria.

FIG 1.

The glycolytic Embden-Meyerhof-Parnas pathway and the sulfoglycolytic Embden-Meyerhof-Parnas pathway (a) and diagrammatic representation of central carbon metabolism in glycolytic and sulfoglycolytic E. coli (b). Glc, glucose; G6P, glucose-6-phosphate; F6P, fructose-6-phosphate; FBP, fructose-bisphosphate; GAP, glyceraldehyde-3-phosphate; DHAP, dihydroxyacetone phosphate; PYR, pyruvate; SQ, sulfoquinovose; SF, sulfofructose; SFP, sulfofructose-1-phosphate; SLA, sulfolactaldehyde; DHPS, 2,3-dihydroxypropanesulfonate; PGI, phosphoglucose isomerase; PFK, phosphofructose kinase; TPI, triose phosphate isomerase; acetyl-CoA, acetyl coenzyme A; TCA, tricarboxylic acid.

The pathways for the catabolism of SQ are termed “sulfoglycolysis” (6–8). The first sulfoglycolytic pathway was described in Escherichia coli (9) and was coined the sulfoglycolytic Embden-Meyerhof-Parnas (sulfo-EMP) pathway owing to its similarity to the “investment phase” of glycolysis or the upper glycolytic EMP pathway (Fig. 1a), in which two molecules of ATP are consumed to convert Glc to two molecules of glyceraldehyde-3-phosphate (GAP). Upper glycolysis comprises the steps catalyzed by hexokinase (phosphorylation of Glc to glucose-6-phosphate [G6P]), phosphoglucose isomerase (conversion of G6P to fructose-6-phosphate [F6P]), phosphofructose kinase (phosphorylation of F6P to form fructose-bisphosphate [FBP]), FBP aldolase (cleavage of FBP into two 3-carbon [3C] fragments, GAP and dihydroxyacetone phosphate [DHAP]), and triose phosphate (triose-P) isomerase (interconversion of GAP and DHAP). During the “payoff phase,” or lower glycolysis, two molecules of GAP are converted into two molecules of pyruvate over five steps, producing energy as ATP and reducing power as NADH. In contrast, the sulfo-EMP pathway converts SQ to DHAP and sulfolactaldehyde (SLA) using a dedicated SQ-SF isomerase (conversion of SQ to sulfofructose [SF]), an SF kinase (phosphorylation of SF to give sulfofructose-1-phosphate [SFP]), and an SFP aldolase (cleavage of SFP into DHAP and sulfolactaldehyde [SLA]) (Fig. 1a). As in glycolysis, DHAP is isomerized to GAP and further catabolized in lower glycolysis, with the production of ATP and NADH, while the other 3C fragment, SLA, is reduced to 2,3-dihydroxypropanesulfonate (DHPS), generating NAD⁺, and excreted quantitatively from the cell (9, 10). The sulfo-EMP pathway therefore has balanced NAD⁺/NADH production but yields half the ATP and carbon per hexose sugar compared to the glycolytic EMP pathway (11). The sulfo-EMP pathway is present in the majority of commensal and pathogenic E. coli strains as well as a wide range of other Enterobacteriaceae (9). A variant of the sulfo-EMP pathway (termed sulfo-EMP2) identified in Gram-positive Bacillus spp. (6) and Arthrobacter spp. (12) has steps identical to those of the E. coli pathway and leads to the production of DHAP but contains an SLA dehydrogenase (instead of an SLA reductase) that leads to the quantitative excretion of sulfolactate (SL).

E. coli shows a clear preference for growth on Glc versus SQ. Glc-adapted E. coli K-12 strains take 2 weeks to adapt to growth on SQ and, once adapted, grow slower (1 to 3 days) than on Glc (9). The maximum biomass of E. coli grown on SQ is also approximately half that of cultures grown on Glc (9). The reduced growth and biomass on SQ could reflect the lower yields of carbon, ATP, and NADH in the sulfo-EMP pathway and supplies of energy and precursors for anabolic processes. The sulfo-EMP pathway also bypasses the steps in glycolysis that produce G6P and F6P, which are needed to supply precursors for the pentose phosphate pathway (PPP) and sugar nucleotide and cell wall biosynthesis, respectively (Fig. 1b). SQ-grown E. coli must therefore additionally divert the triose-P DHAP/GAP produced by sulfoglycolysis into G6P and F6P (13, 14) while maintaining sufficient flux into lower glycolysis and pyruvate production to sustain the tricarboxylic acid (TCA) cycle and fatty acid biosynthesis. The switch from glycolysis to sulfoglycolysis therefore requires large-scale metabolic changes that may be responsible for the slow transition from Glc to SQ. While these changes have yet to be characterized, limited studies by Burrichter and coworkers provided evidence of novel mixed-acid fermentation and the conversion of SQ-derived triose-P to succinate, acetate, and formate when E. coli was grown on SQ under anaerobic conditions (15).

In this work, we investigate the metabolic changes occurring in E. coli grown with Glc or SQ as the sole carbon source under aerobic conditions using comparative metabolite profiling. Unexpectedly, we find evidence for the extensive diversion of SQ-derived carbon into glycogen and trehalose storage carbohydrates in SQ-grown E. coli during logarithmic phase, which is suggestive of limitation of other pathways required for growth. These storage carbohydrates are utilized following the commencement of stationary phase, which is suggestive of a reverse diauxic shift whereby bacteria temporarily switch back to glycolysis following the exhaustion of their primary carbon source. Untargeted metabolite analysis reveals further changes in metabolite levels across a range of metabolic pathways during sulfoglycolytic metabolism; these include clear perturbations to the TCA cycle, the PPP, polyamine metabolism, pyrimidine metabolism, and many amino acid metabolic pathways. This work shows that metabolic adaptation to sulfoglycolytic growth in E. coli requires the simultaneous operation of gluconeogenesis and lower glycolysis, leading to the accumulation of storage carbohydrates. We propose that the accumulation of storage carbohydrates during SQ metabolism is associated with primary deficits in energy (as ATP), reducing power, and carbon-based building blocks.

RESULTS

Growth on SQ leads to major changes in E. coli metabolism.

E. coli BW25113 cells were grown to mid-log phase on Glc (4 mM) and SQ (8 mM) as the sole carbon sources. The SQ concentration chosen was double that of Glc to account for the fact that E. coli cells excrete half of the internalized SQ (carbons 4 to 6 and the sulfonate group) as DHPS; this approach was also utilized previously by Burrichter and coworkers (15). This doubling of the SQ concentration did not appear to have a substantial impact on the growth rate: the growth rate was approximately 5-fold lower than that of Glc-grown E. coli (compared with 3.8-fold for E. coli MG1655 in a previous study [9]). Cells were harvested and metabolically quenched, and intracellular metabolites were extracted into methanol (MeOH)-H₂O (3:1, vol/vol) (16), derivatized, and then analyzed by gas chromatography–electron ionization–triple-quadrupole mass spectrometry (GC-EI-QqQ-MS) (17). Metabolites were detected by multiple-reaction monitoring (MRM) (521 targets, representing approximately 350 metabolites, with two MRM transitions [qualifier and quantifier] per target). Data were manually inspected and curated prior to more detailed analyses.

In total, 146 metabolites were detected under at least one growth condition, 117 of which had statistically significant differences in abundances in bacteria grown on the two carbon sources (false detection rate [FDR]-adjusted P value of <0.05). As shown using a volcano plot, 36 metabolites had lower abundances in SQ-grown than in Glc-grown E. coli, and 81 had higher abundances (Fig. 2a). Fold changes spanned several orders of magnitude: lysine was >4,000-fold less abundant, several metabolites were >100-fold less abundant, and several were >10-fold more abundant in SQ-grown E. coli (Fig. 2b and c). There were also large increases in hexose-based sugars in SQ-fed E. coli.

FIG 2.

Fold changes (FC) in metabolite abundances in E. coli cells grown on Glc versus SQ. (a) Volcano plot of the results, with selected metabolites labeled; (b) 15 largest decreases in abundance in SQ-grown E. coli cells; (c) 15 largest increases in abundance in SQ-grown E. coli cells.

Amino acid pools are perturbed during sulfoglycolysis.

Many amino acid pools underwent very large perturbations upon the change from growth on Glc to growth on SQ (Fig. 2b and c; see also Table S1 in the supplemental material). Tyrosine was 900-fold less abundant in SQ-grown E. coli, yet phenylalanine and its derivatives phenyllactate and phenylpyruvate (18–20) showed no statistically significant changes (Fig. 2b; Table S1). Lysine was >4,000-fold less abundant in SQ-grown E. coli (Fig. 2b; Table S1). Alanine, methionine, threonine, valine, arginine, and asparagine were also less abundant (Table S1). In contrast, serine, cysteine, glycine, isoleucine, leucine, glutamate, aspartate, and proline were all more abundant in SQ-grown E. coli (Table S1).

Changes were also observed for intermediates in amino acid biosynthetic and degradation pathways. The cysteine degradation product 3-sulfino-l-alanine (18–20) was 17-fold less abundant in SQ-grown E. coli, while the lysine degradation products 5-aminopentanoate, glutarate, 2-hydroxyglutarate, and α-ketoglutarate (18–20) and the valine degradation product 3-hydroxyisobutyrate (18–20), were all more abundant (Table S1). The leucine precursors 2-isopropylmalate and ketoleucine (18–20), and the common leucine and valine precursor α-ketoisovalerate (18–20), were more abundant in SQ-grown E. coli, while the valine, leucine, and isoleucine common precursor α-ketobutyrate (18–20) was less abundant (Table S1).

Putrescine levels were 1.5-fold higher in mid-log-phase SQ-grown than in Glc-grown E. coli cells (Fig. 3a). There was also 1.5-fold more proline and glutamate but 10-fold less arginine in SQ-grown E. coli (Fig. 3a). Putrescine is derived either from arginine via agmatine (which releases urea) or from glutamate or proline via ornithine, with the latter being more direct (Fig. 3c) (18–23). It has been reported that in E. coli, urea is produced almost exclusively by putrescine biosynthesis from arginine via agmatine (21, 24). The observation of 3-fold less urea in SQ-grown E. coli may indicate less dependence on the agmatine pathway. Putrescine has two possible fates in E. coli: it can be oxidized to γ-aminobutyric acid (GABA) or acetylated to N-acetyl putrescine (Fig. 3c) (18–20). There was 300-fold less N-acetyl putrescine and 2.4-fold more GABA in SQ-grown E. coli (Fig. 3a), suggesting that more oxidation and less acetylation of putrescine occur during sulfoglycolysis. The lower levels of N-acetyl putrescine may indicate lower levels of acetyl-CoA production.

FIG 3.

(a) Heat plot of selected metabolites detected in mid-log-phase E. coli cells grown on Glc and SQ depicting fold changes (FC). (b to d) Selected metabolic pathways in E. coli (18–20) grown on minimal medium, with fold changes indicated by colors. (b) Trehalose, maltose, and glycogen biosynthesis (22, 23, 47, 48); (c) polyamine metabolism (21); (d) central carbon metabolism. Metabolites highlighted in green are more abundant in SQ-grown E. coli cells, and those highlighted in red are less abundant. Metabolites shown in gray were not statistically different (FDR-adjusted P value of >0.05). Underlined metabolites were not detected.

Sulfoglycolysis perturbs purine and pyrimidine metabolic pools.

Adenosine was approximately 6-fold less abundant and xanthine was >360-fold less abundant in SQ-grown than in Glc-grown E. coli cells (Table S1), which is indicative of a perturbation of purine metabolism. Pyrimidine metabolism was also perturbed: in SQ-grown E. coli, there were >95-fold-lower levels of orotic acid as well as lower levels of the orotic acid precursors N-carbamoyl-l-aspartate (>100-fold lower) and dihydroxyorotate (230-fold lower) (Table S1). In E. coli BW25113, orotic acid is a metabolic end product (25); thus, the lower levels of orotic acid may suggest a decreased input into pyrimidine metabolism in sulfoglycolytic E. coli. Consistent with this observation, broader effects on pyrimidine metabolism were evident: there were 1.3-fold-lower levels of uracil, 9-fold-lower levels of the uracil degradation product β-alanine, and 2.7-fold-higher levels of thymine in sulfoglycolytic E. coli. The cytosine pool was not significantly perturbed (Table S1).

Pools of redox mediators are perturbed upon the switch to sulfoglycolysis.

Glutathione and NAD(P)H help maintain the redox status of E. coli and are cofactors for a wide range of oxidoreductases. In SQ-grown E. coli, there were 7-fold-higher levels of glutathione and 1.7-fold-higher levels of nicotinate and nicotinamide than in glycolytic E. coli (Table S1), showing that sulfoglycolysis perturbs these key species in redox biochemistry.

SQ-grown E. coli cells accumulate carbohydrates during mid-log phase.

Various nonphosphorylated sugars (Glc, galactose, and mannitol) and oligosaccharides (trehalose and maltose) accumulated in SQ-grown E. coli during mid-log phase (Fig. 2c), suggesting that a significant fraction of the triose-P generated during SQ catabolism is diverted into the synthesis of hexose phosphates (hexose-P) via the reverse action of FBP aldolase and the dedicated gluconeogenic enzyme fructose-bisphosphatase (FBPase). The accumulation of Glc (>39-fold increase in SQ-fed versus Glc-fed bacteria) was unexpected and may reflect the activity of a hexose-P phosphatase (Fig. 2c and Fig. 3a) or the constitutive cycling of glycogen pools via glycosidases as well as glycogen phosphorylase. The conversion of SQ to Glc via the gluconeogenic pathways was confirmed by labeling E. coli with [¹³C₆]SQ (ratio of labeled/unlabeled SQ of 1:1). The presence of a prominent +3 isotopomer in the free Glc pool reflects the incorporation of triose-P into hexose synthesis.

Quantitative analysis reveals that there are higher levels of GAP but lower levels of phosphoenolpyruvate, 3-phosphoglycerate (3PG), and 2-phosphoglycerate in mid-log-phase SQ-grown E. coli cells (Fig. 3a). In contrast, higher levels of pyruvate and lactate were present in SQ-grown E. coli; this may indicate the slower consumption and/or production of these metabolites through degradative pathways during adaptation to sulfoglycolysis and may also indicate a change in the redox status. Similarly, the TCA cycle was also perturbed by the switch from growth on Glc to growth on SQ. α-Ketoglutarate was 1.6-fold more abundant and fumarate and citrate were 1.6-fold less abundant in SQ-grown E. coli (Fig. 3a). There were no statistically significant differences in the malate, succinate, and cis-aconitate pools. The perturbations of the TCA cycle and lower glycolytic metabolites are consistent with the diversion of a significant portion of the triose-P produced by sulfoglycolysis into gluconeogenesis, away from these downstream metabolic pathways.

The PPP metabolites d-glucono-1,5-lactone, d-gluconate, d-ribose-5-phosphate, d-ribose, d-sedoheptulose-7-phosphate, d-ribulose, d-xylulose, and d-arabitol were between 1.5-fold and 3-fold more abundant in SQ-grown E. coli (Fig. 3a), consistent with generally increased flux into sugar-phosphate synthesis under these growth conditions.

Sulfoglycolysis perturbs cell wall biosynthesis in E. coli.

The levels of N-acetylglucosamine, a key component of the peptidoglycan cell wall and outer membrane of E. coli (26, 27), and N-acetylmannosamine, which can interconvert with the former (18–20), were >5-fold higher in log-phase SQ-grown E. coli (Fig. 3a). The pools of fatty acids, key precursors required for the lipopolysaccharide outer membrane (28), were also perturbed; all detected fatty acid pools, except for linoleic acid, were 1.4-fold to 1.9-fold larger in SQ-grown E. coli (Table S1). This may reflect the lower growth rate of SQ-grown E. coli (9) and, hence, the slower consumption of these cell wall precursors. Alternatively, while the cell pellets harvested from Glc-grown and SQ-grown E. coli were of similar sizes, the SQ-grown pellets were more difficult to resuspend in the extraction solvent, which may reflect differences in the cell wall structures of glycolytic and sulfoglycolytic E. coli cells, such as thicker cell walls and/or cell walls that lead to greater cell adhesion.

E. coli switch from sulfoglycolysis to glycolysis during stationary-phase adaptation.

The finding that mid-log-phase E. coli cells accumulate trehalose and maltose, as well as intermediates of the PPP and amino-sugar synthesis when grown on SQ, indicated that a substantive proportion of triose-P is diverted into hexose/pentose phosphate synthesis via the final steps of gluconeogenesis or the PPP transketolase/transaldolase enzymes (Fig. 2c and Fig. 3a; Table S1). To assess whether growth on SQ also resulted in the accumulation of glycogen, the major E. coli carbohydrate reserve (29–31), Glc- and SQ-grown E. coli cells were collected at five time points along the growth curve and differentially extracted to recover low-molecular-weight oligosaccharides and glycogen. Maltose and trehalose were recovered in the methanol-water extract, and total protein and glycogen were recovered in the insoluble pellet. The total carbohydrate content of the soluble fraction was also quantitated after methanolysis and the conversion of Glc and fructose to their trimethylsilyl (TMS) derivatives. The storage carbohydrate content was normalized to the total protein content for analysis (Fig. 4; Fig. S1 and S2).

FIG 4.

Storage carbohydrate contents of Glc-grown and SQ-grown E. coli cells across the growth curve. (a) Storage carbohydrate content overlaid on the growth curve of Glc-grown E. coli; (b) storage carbohydrate content overlaid on the growth curve of SQ-grown E. coli; (c) glycogen content (measured as the total Glc present in the insoluble fraction) of Glc-grown and SQ-grown E. coli cells; (d) trehalose content of Glc-grown and SQ-grown E. coli cells; (e) Glc content of the soluble fractions of Glc-grown and SQ-grown E. coli cells. (ci to ei) early log phase; (cii to eii) transition to stationary phase; (ciii to eiii) beginning of stationary phase; (civ to eiv) 60 to 90 h into stationary phase. The data shown are per cell pellet and are averages from two independent replicates (means ± standard deviations [SD]) (n = 2), with the exception of data for SQ shown in panels eii and eiii (n = 1).

In Glc-grown E. coli, glycogen levels rose during log phase, peaking at the transition to stationary phase (or in late log phase) before declining as stationary phase progressed (Fig. 4a and c). This is broadly consistent with a previous report of this strain of E. coli (BW25113) in M9 minimal medium containing 0.2% Glc (compared with 0.072% Glc in this work) in which the glycogen levels peaked at around late log phase or the transition to stationary phase and declined as stationary phase progressed (32). SQ-grown E. coli also produced glycogen, which accumulated during log phase, peaked at the beginning of stationary phase, and then fell as stationary phase progressed (Fig. 4b and c). Peak levels were similar in SQ- and Glc-grown E. coli cells. Collectively, these data show that E. coli cells grown on SQ continue to accumulate glycogen and low-molecular-weight oligosaccharides at the expense of carbon flow into lower glycolysis. The accumulated glycogen appears to function as an alternative carbon source once extracellular Glc or SQ is depleted and cells enter stationary phase.

To verify that the depletion of glycogen during stationary phase in SQ-grown E. coli was due to catabolism and not conversion to smaller storage carbohydrates such as the disaccharides trehalose/maltose, we analyzed the carbohydrate content of the soluble fraction of stationary-phase cells. No maltose was detected under either growth condition, and only very low levels of trehalose (an order of magnitude lower than glycogen levels) were detected in SQ-grown E. coli, with levels decreasing as stationary phase progressed (Fig. 4b to d). Only trace amounts of trehalose were detected in Glc-grown E. coli during all stages of growth. In the total monosaccharide analysis, we targeted Glc and fructose. Only Glc was detected in both Glc- and SQ-grown E. coli, with levels being an order of magnitude higher than those of glycogen in SQ-grown E. coli (Fig. 4b, c, and e), indicating that most carbohydrate in Glc- and SQ-grown E. coli is present as free Glc and/or unidentified Glc-containing disaccharides. Glc levels rose across log phase in SQ-grown E. coli and then fell as stationary phase progressed. Taken together, these results suggest that the glycogen that accumulates during the log-phase growth of E. coli on SQ is consumed upon the transition to stationary phase, indicating a reversal in carbon flux from sulfoglycolytic/gluconeogenic to the upper glycolytic pathway.

DISCUSSION

The metabolism of SQ through the sulfo-EMP pathway leads to the formation of triose-P, bypassing key steps in upper glycolysis that generate G6P and F6P for the PPP and the synthesis of cell wall intermediates. As a result, triose-P generated by SQ metabolism must be diverted into upper gluconeogenesis for hexose/pentose phosphate production, as well as being further catabolized in ATP-generating steps in lower glycolysis. This requirement imposes a double bioenergetic burden on SQ-grown bacteria: sulfoglycolysis provides only one 3C fragment and consumes one molecule of ATP. Furthermore, while upper glycolysis produces no reducing power, the sulfo-EMP pathway consumes one molecule of NADH (for the reduction of SLA to DHPS) per SQ molecule. E. coli is thought to derive most of its ATP from oxidative phosphorylation when utilizing Glc as a carbon source, and the markedly reduced growth rate of SQ-grown E. coli compared to that of Glc-grown E. coli (9) indicates that the bioenergetic burden associated with SQ metabolism is substantial.

To partially compensate for the reduced carbon yield associated with SQ versus Glc metabolism, E. coli cells were cultivated in 8 mM SQ and 4 mM Glc. This ensured that the quantity of carbon (and, potentially, energy) available to the cells was essentially the same for both growth substrates. Nonetheless, a substantially lower (approximately 5-fold) growth rate was observed for SQ-grown E. coli than for Glc-grown E. coli. Comparative analysis of metabolite abundances in mid-log-phase E. coli cells grown on SQ and Glc revealed that under SQ-replete conditions (compared to Glc-replete conditions) there were large perturbations in the levels of amino acids, various sugars, assorted cellular metabolites of a wide range of pathways, and, possibly, cell wall peptidoglycan. These changes are likely secondary effects that arise from large differences in the channeling of carbon that occurs during the switch from growth on Glc (and the use of both upper and lower glycolysis) to the use of the sulfo-EMP pathway (the use of sulfoglycolysis, lower glycolysis, and upper gluconeogenesis); that is, these changes arise due to a switch from glycolytic to gluconeogenic metabolism.

Unexpectedly, we show that log-phase sulfoglycolytic cells produce large quantities of hexoses, disaccharides, and the storage polysaccharide glycogen. Thus, under both Glc- and SQ-replete conditions, E. coli diverts some of the G6P produced through upper glycolysis/gluconeogenesis into storage carbohydrates. Glycogen is typically produced in E. coli when excess carbon is present but growth is limited by a deficiency in an essential nutrient(s) required for growth (29). Similarly, cultivation on SQ may lead to reduced flux into lower glycolysis and/or the TCA cycle and the reduced synthesis of multiple anabolic intermediates needed for biomass (protein, nucleic acid, and lipid) accumulation, leading to the diversion of the resultant excess triose-P into carbohydrate synthesis.

Alternatively, the unbalanced diversion of triose-P into hexose-P synthesis during SQ metabolism may lead to reduced flux into lower glycolysis, oxidative phosphorylation, and downstream metabolic pathways. Reduced downstream flux will result in less ATP, reducing power, and carbon-based building blocks being available to sulfoglycolytic E. coli, which may be the main origin of the adjustments undertaken by E. coli to maintain balanced growth and, thus, the lower rate of growth on SQ. In accordance with this analysis, products of the TCA cycle were perturbed, suggesting adaptation to reduced carbon input, and putrescine acetylation was depressed, potentially due to a lack of acetyl-CoA (Fig. 1b and Fig. 3c). Lower levels of pyrimidine biosynthesis metabolites may also arise from a deficiency in carbon building blocks. Amino acid levels were generally lower in sulfoglycolytic cells, which could reflect the overall energy state of the cell given that protein synthesis is one of the most energy-intensive cellular processes (Fig. 2b and c). Similarly, in SQ-grown E. coli, putrescine biosynthesis appears to prefer the more energetically favorable pathway directly from glutamate and/or proline via ornithine, with the alternative pathway from arginine via agmatine seemingly being disfavored (Fig. 3c) (18–21). The levels of glutathione and other redox-active metabolites were perturbed and the oxidation of putrescine to GABA appeared to increase during sulfoglycolytic metabolism, potentially reflecting an altered redox state. Loewen showed that the glutathione pool increases in stationary phase in Glc-grown E. coli, which was hypothesized to be in proportion to the size of the available amino acid pool (33). Possibly, the differences seen here also relate to differences in the sizes of the available amino acid pools in Glc- and SQ-grown E. coli.

Organisms regulate the levels of intracellular metabolites to prevent their accumulation to toxic levels. This can be especially pronounced during logarithmic growth when metabolite production can far exceed cellular needs, leading to spillover. For example, Saccharomyces cerevisiae possesses an energy-consuming futile cycle that consumes ATP through the production of trehalose from the sugar nucleotide UDP-Glc and Glc, while no ATP is produced in the hydrolysis of trehalose to Glc, even under conditions such as heat shock where trehalose accumulates (34). The formation of glycogen in SQ-grown E. coli is a resource-intensive process: the formation of one G6P molecule from triose-P through gluconeogenesis requires the investment of two ATP molecules and the consumption of two SQ molecules; UTP must also be invested to form UDP-Glc. In Glc-grown E. coli, the diversion of Glc into glycogen via G6P is less energy-intensive, requiring the investment of only one ATP and one UTP molecule. The futile cycling of cellular glycogen and other storage polysaccharides provides a mechanism to consume overflow ATP via the conversion of G1P to UDP-Glc to make glucosidic linkages and then release this carbon as G1P by phosphorolysis. Possibly, the accumulation of cellular glycogen occurs to provide a substrate for an energy-consuming futile cycle.

Analysis of the storage polysaccharide contents of E. coli cells grown on Glc or SQ at different stages of growth indicates that upon the transition to stationary phase, carbohydrates stored during log phase are consumed. In the case of SQ-grown E. coli, the depletion of SQ and triose-P appears to trigger a reverse diauxic shift whereby cells switch from limited gluconeogenic metabolism to canonical glycolytic metabolism as Glc/G6P is mobilized from glycogen/trehalose/maltose breakdown. Thus, growth on SQ requires multiple changes in the direction of the flux of central carbon metabolism depending on the growth state. The requirement for a change in flux from the glycolytic to the gluconeogenic direction upon the switch to growth on SQ may partly contribute to the slow adaptation of E. coli to SQ (9); in E. coli, switching from glycolytic to gluconeogenic metabolism also induces a long lag phase (35, 36). In contrast, in E. coli, the switch from gluconeogenic to glycolytic metabolism occurs quickly (35, 36), so the reversal upon SQ exhaustion is expected to be rapid.

Most enzymes in upper glycolysis/gluconeogenesis are reversible, and glycolysis/gluconeogenesis is almost thermodynamically neutral. As such, only limited changes in protein expression are required for a change in the flux direction, assisted by the presence of concentration gradients as one substrate is depleted and another becomes dominant, as well as allosteric effects. Growth on Glc provides a strong gradient promoting flux in the glycolytic direction, while growth on SQ leads to the production of large amounts of triose-P, reversing the concentration gradient and providing a driving force for upper gluconeogenesis. The expression of the class I FBP aldolase of E. coli, which is induced by growth on gluconeogenic substrates and presumed to be utilized for gluconeogenesis (as opposed to the constitutively expressed class II aldolase presumed to be utilized for glycolysis) (37), may be required upon a switch to SQ. In contrast, it has been shown that the switching of E. coli from growth on Glc to growth on the gluconeogenic substrate acetate results in just 2-fold changes in the expression levels of the upper gluconeogenic enzyme FBPase (36), a key regulatory enzyme of gluconeogenesis (38). FBPase is subject to complex allosteric regulation, with AMP and G6P serving as allosteric inhibitors (38, 39), while TCA cycle intermediates (such as citrate and isocitrate) (40), phosphorylated 3C carboxylic acids (such as phosphoenolpyruvate [PEP] [41] and 3PG [40]), sulfate (41), and phosphate (40) are allosteric activators. The key allosteric interactions appear to be activation by citrate and PEP (38, 40) and synergistic inhibition by AMP and G6P (38). There is likely an interplay of the production of storage carbohydrates from hexose-P generated by gluconeogenesis in SQ-grown E. coli and the inhibitory effect of G6P on FBPase.

Conclusions.

We provide evidence that the growth of E. coli on SQ imposes a significant bioenergetic cost on the bacteria, because of the lower yield of carbon from each molecule of SQ catabolized, and the need to divert a significant fraction of the triose-P produced into hexose/pentose phosphate synthesis. As a result, the reduced rate of growth of adapted E. coli on SQ likely stems from the reduced flux of carbon into the TCA cycle and downstream metabolism, producing lower levels of carbon building blocks, NADH, and ATP, thereby triggering large-scale changes in cell metabolism to maintain balanced growth. Further work is needed to define the factors that regulate the partitioning of carbon between upper gluconeogenesis and lower glycolysis under SQ-grown conditions, while the mobilization of low- and high-molecular-weight carbohydrates following the depletion of SQ indicates that the accumulation of these molecules is an important adaptive response. This work highlights the ascetic nature of growth on SQ and provides insights into the metabolic adaptations of E. coli to growth on this widespread but poorly studied organosulfur sugar.

MATERIALS AND METHODS

Reagents.

SQ and [¹³C₆]SQ were synthesized according to methods described previously (42). scyllo-Inositol (catalogue number I8132), trehalose (catalogue number T9531), and fructose (catalogue number F0127) were purchased from Sigma-Aldrich. Maltose (catalogue number 29131) was purchased from BDH. Glucose (catalogue number GA-018) was purchased from ChemSupply Australia.

Comparative metabolite profiling of E. coli on Glc and SQ.

E. coli BW25113 (adapted to growth on the relevant substrate) was used to inoculate a 5-mL starter culture of M9 minimal medium containing 4 mM Glc or a 3-mL starter culture of M9 minimal medium containing 8 mM SQ. The Glc starter culture was grown to an optical density at 600 nm (OD₆₀₀) of 0.0680 (6 h 20 min) and the SQ starter culture was grown to an OD₆₀₀ of 0.1864 (41 h) at 37°C with shaking (250 rpm) and used to inoculate six experimental cultures for each substrate. These experimental cultures contained either 4 mM Glc or 8 mM SQ. The culture volumes were 50 mL for Glc and 20 mL for SQ, and the inoculant volumes were 298 μL for Glc and 118 μL for SQ. The Glc cultures were grown at 30°C with shaking (250 rpm) for 11 h and then at 37°C with shaking (250 rpm) until mid-logarithmic phase was achieved (OD₆₀₀ of approximately 0.50). The SQ cultures were grown at 37°C with shaking (250 rpm) until mid-logarithmic phase was achieved (OD₆₀₀ of approximately 0.44). One 5-mL aliquot was harvested per culture. Representative growth curves are shown in Fig. S3 and S4 in the supplemental material.

Aliquots of cell culture media were metabolically arrested by infusing them with ice-cold phosphate-buffered saline (PBS) (3× aliquot volume) and placing them in an ice water slurry for 10 min. The cell suspension was centrifuged (4,000 rpm for 10 min at room temperature [RT]), and the supernatant was discarded. Cells were washed 3 times by resuspending them in ice-cold PBS, pelleting them by centrifugation (14,000 rpm for 1 min at RT), and discarding the supernatant. Cell pellets were then centrifuged again (14,000 rpm for 1 min at RT) to remove residual culture medium and PBS prior to metabolite extraction.

Extraction, derivatization, and analysis of metabolites for comparative metabolite profiling.

Cell pellets were resuspended in a chilled extraction solution (500 μL) comprised of 3:1 MeOH-H₂O and [¹³C₅,¹⁵N]valine (1 mM; 0.5 μL) and [¹³C₆]sorbitol (1 mM; 0.5 μL) as the internal standards. The suspensions were subjected to freeze-thaw cycles to facilitate the lysis of the cells (30 s in liquid N₂ and 30 s in a dry ice-ethanol [EtOH] bath for 10 cycles) and shaken (9,000 rpm for 10 min at 2°C). The samples were then centrifuged (12,700 rpm for 5 min at 1°C) to remove cell debris and precipitated macromolecules (16).

The cell lysate supernatant was transferred into glass inserts and dried in a rotational vacuum concentrator. To remove all residual H₂O, all samples were washed with MeOH (50 μL). The glass inserts were transferred to 2-mL autosampler vials. Online derivatization was conducted using an autosampler robot (AOC6000; Shimadzu). All samples were methoximated with a methoxyamine hydrochloride solution (30 mg/mL in pyridine at 20 μL) for 2 h at 37°C, followed by trimethylsilylation in N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) plus 1% trimethylchlorosilane (TMCS) (20 μL) for 1 h at 37°C with continuous mixing. Samples were incubated at RT for 1 h prior to GC-MS analysis.

Metabolite profiles were acquired on a Shimadzu 2010 GC system coupled to a TQ8040 QqQ mass spectrometer. The inlet temperature was held at 280°C, and helium was used as a carrier gas (purge flow of 5.0 mL/min and column flow of 1.1 mL/min). One microliter of the derivatized sample was injected into the GC-QqQ-MS system in splitless mode. Chromatographic separation was achieved using a J&W DB-5 capillary column (17) (30 m by 0.25 mm by 1.00 μm). The oven had a starting temperature of 100°C, which was held for 4 min, ramped to 320°C at 10°C/min, and held for 10 min. The transfer line temperature was 280°C, and the ion source temperature was 200°C. Compounds were fragmented using electron ionization (EI). Argon was used as the collision-induced dissociation gas. Metabolite detection was performed using the Shimadzu Smart MRM (multiple-reaction monitoring) database, which contains up to 521 targets representing ~350 metabolites with two MRM transitions (quantifier and qualifier) per target, including the precursor ion, product ion, collision energy, retention index, and time, with a minimum dwell time of 2 ms set up for the acquisition method. The automatic adjustment of retention time (AART) function in GCMSsolution software (V 4.42; Shimadzu) and a standard alkane series mixture (C₇-C₃₃; Restek) were used to correct retention time shifts in the acquisition method when the column was cut or replaced.

Data analysis for comparative metabolite profiling.

Using Shimadzu Lab Solutions Insight, each target was visually inspected and manually integrated if required. Targets that were lacking in either the qualifier or the quantifier MRM or that had a quantifier at <50% of the abundance of the qualifier or had inconsistent retention times (either between the qualifier and quantifier or between samples) were discarded. This provided a data matrix for downstream data analysis, which was conducted using Metaboanalyst software (https://www.metaboanalyst.ca/MetaboAnalyst/home.xhtml). Peak areas were submitted to Metaboanalyst, normalized by the median, and log transformed. A two-sample t test was performed. Reported P values were false detection rate (FDR) adjusted using the Benjamini-Hochberg procedure (43). FDR-adjusted P values of <0.05 were considered significant.

KEGG Mapper Search & Colour Pathway (https://www.genome.jp/kegg/tool/map_pathway2.html [reference organism, E. coli K-12 W3110]) was used to assist in drawing conclusions. In cases where more than one target corresponding to a single metabolite was detected, the fold changes of the targets were averaged to give the fold change for the metabolite.

Metabolite analysis of E. coli grown on [¹³C₆]Glc and [¹³C₆]SQ.

E. coli BW25113 (adapted to growth on the relevant substrate) was used to inoculate a 5-mL starter culture of M9 minimal medium containing 4 mM Glc or a 3-mL starter culture of M9 minimal medium containing 8 mM SQ. The Glc starter culture was grown to an OD₆₀₀ of 0.0808 (8 h) and the SQ starter culture was grown to an OD₆₀₀ of 0.1964 (41 h) at 37°C with shaking (250 rpm) and used to inoculate the experimental cultures. These experimental cultures were either unlabeled (4 mM [¹²C₆]Glc or 8 mM [¹²C₆]SQ), labeled for the entire growth period (2 mM [¹²C₆]Glc plus 2 mM [¹³C₆]Glc or 4 mM [¹²C₆]SQ plus 4 mM [¹³C₆]SQ), or labeled for a portion of the growth period (4 mM [¹²C₆]Glc, resuspended into 2 mM [¹²C₆]Glc plus 2 mM [¹³C₆]Glc 2.75 h before harvest, or 8 mM [¹²C₆]SQ, resuspended into 4 mM [¹²C₆]SQ plus 4 mM [¹³C₆]SQ 4 h before harvest). There were two experimental replicates for each medium type, for totals of 6 replicates for Glc and 6 replicates for SQ. The culture volumes were 50 mL for Glc and 20 mL for SQ, and the inoculant volumes were 200 μL for Glc and 112 μL for SQ. The Glc cultures were grown at 30°C with shaking (250 rpm) for 11 h and then at 37°C with shaking (250 rpm) until mid-logarithmic phase was achieved (OD₆₀₀ of approximately 0.50). The SQ cultures were grown at 37°C with shaking (250 rpm) until mid-logarithmic phase was achieved (OD₆₀₀ of approximately 0.44). Three 5-mL aliquots were harvested per culture.

Cell pellets were resuspended in a chilled extraction solution (500 μL) comprised of 3:1 MeOH-H₂O and scyllo-inositol (1 mM; 0.5 μL) as an internal standard. Cell suspensions were subjected to freeze-thaw cycles to facilitate the lysis of the cells (30 s in liquid N₂ and 30 s in a dry ice-EtOH bath for 10 cycles) and shaken (9,000 rpm for 10 min at 4°C). The samples were then centrifuged (14,000 rpm for 5 min at 1°C) to remove cell debris (16).

The cell lysate supernatant was transferred into glass inserts and dried in a rotational vacuum concentrator. To remove all residual H₂O, all samples were washed with MeOH (50 μL). The glass inserts were transferred to 2-mL autosampler vials. Online derivatization was conducted with a Gerstel MSP2 XL autosampler robot. All samples were methoximated with a methoxyamine hydrochloride solution (30 mg/mL in pyridine at 20 μL) for 2 h at 37°C, followed by trimethylsilylation in BSTFA plus 1% TMCS (20 μL) for 1 h at 37°C with continuous mixing. Samples were incubated at RT for 1 h prior to GC-MS analysis.

Metabolite profiles were acquired on an Agilent 7890A gas chromatograph coupled to a 5975 mass spectrometer as a detector. One microliter of the derivatized sample was injected in splitless mode into a split/splitless inlet set at 250°C. Chromatographic separation was achieved using a J&W VF-5ms capillary column (30 m by 0.25 mm by 0.25 μm plus a 10-m Duraguard column). The oven had a starting temperature of 35°C, which was held for 1 min, ramped to 320°C at 25°C/min, and held for 5 min. Helium was used as the carrier gas at a flow rate of 1 mL/min. Compounds were fragmented using electron impact ionization and detected across a mass range of 50 to 600 atomic mass units (amu) with a scan speed of 9.2 scans/s (44).

Chromatograms produced from unlabeled samples were used for pool size comparisons. Using Agilent’s Mass Hunter Quantitative Analysis software for GC-MS, metabolites contained within an in-house Metabolomics Australia library, with the addition of some metabolites detected and identified in the chromatograms using Agilent ChemStation for GC-MS and the Fiehn L metabolite library, were identified, and representative target ion areas were integrated. Each detected metabolite in each chromatogram was visually inspected and manually integrated if required. These formed an output data matrix for downstream data analysis (44), which was conducted using Metaboanalyst software (http://www.metaboanalyst.ca/faces/ModuleView.xhtml). Data were normalized by the median and log transformed. P values of <0.05 and fold changes of <1.5 were considered significant.

To detect labeled metabolites, chromatograms were processed using Nontargeted Tracer Fate Detection software (https://ntfd.bioinfo.nat.tu-bs.de/) (peak threshold of 5, minimum peak height of 5, deconvolution width of 5 scans, minimum percent label of 5%, maximum percent label of 100%, minimum R² of 0.95, maximum fragment deviation of 0.20, required number of labeled fragments of 1, and M₁ correction of 0.0109340), followed by manual verification by visual inspection using Agilent ChemStation for GC-MS. Identification was achieved by comparison with the unlabeled samples, from which metabolites were identified using Agilent ChemStation for GC-MS and the Fiehn L metabolite library.

Storage carbohydrate analysis.

E. coli BW25113 (adapted to growth on the relevant substrate) was used to inoculate a 5-mL starter culture of M9 minimal medium containing 4 mM Glc or a 3-mL starter culture of M9 minimal medium containing 8 mM SQ. The Glc starter culture was grown to an OD₆₀₀ of 0.0630 (5 h 15 min) and the SQ starter culture was grown to an OD₆₀₀ of 0.2524 (41 h) at 37°C with shaking (250 rpm) and used to inoculate the experimental cultures (two each for SQ and Glc) that contained either 4 mM Glc or 8 mM SQ. The culture volumes were 50 mL for Glc and 20 mL for SQ, and the inoculant volumes were 322 μL for Glc and 87 μL for SQ. The Glc cultures were grown at 30°C with shaking (250 rpm) for 11 h and then at 37°C with shaking (250 rpm). The SQ cultures were grown at 37°C with shaking (250 rpm). Bacteria were harvested at 5 different time points, with two aliquots harvested per culture at each time point. The harvest time points for the Glc cultures were 960 min (OD₆₀₀ values of 0.2380 and 0.24120), 1,060 min (OD₆₀₀ values of 0.7528 and 0.7504), 1,090 min (OD₆₀₀ values of 0.9512 and 0.9436), 1,330 min (OD₆₀₀ values of 1.0444 and 1.0472), and 5,040 min (OD₆₀₀ values of 1.0100 and 0.9464). The harvest time points for the SQ cultures were 33 h (OD₆₀₀ values of 0.3988 and 0.3636), 41 h (OD₆₀₀ values of 0.7172 and 0.6736), 59 h (OD₆₀₀ values of 1.0012 and 0.8600), 78 h (OD₆₀₀ values of 1.0376 and 0.9088), and 147 h (OD₆₀₀ values of 0.8920 and 0.8100). The culture volume harvested contained the same number of cells (by the OD₆₀₀) as 500 μL of the culture at an OD₆₀₀ of 0.5.

Aliquots of cell culture media were metabolically arrested by infusing them with ice-cold PBS (800 μL, except for the 960-min harvest for Glc and the 33-h harvest for SQ, which used 400 μL and 700 μL, respectively) and placing them in an ice water slurry for 5 min. The cell suspension was centrifuged (14,000 rpm for 5 min at RT), and the supernatant was discarded. Cells were washed 3 times by resuspending them in ice-cold PBS (0.2 mL), pelleting them by centrifugation (14,000 rpm for 1 min at RT), and discarding the supernatant. Cell pellets were then centrifuged again (14,000 rpm for 1 min at RT) to remove residual culture medium and PBS prior to glycogen extraction. Cell pellets were thawed in 100 μL of MilliQ water. A total of 375 μL of 1:2 CHCl₃-MeOH was then added. The samples were incubated at RT for 1 h and vortexed regularly. The samples were then centrifuged (15,000 rpm for 10 min at RT) to separate soluble and insoluble metabolites.

Analysis of glycogen content.

Insoluble metabolites were resuspended in 100 μL of MilliQ water. Ten microliters of each sample was transferred to a glass insert. Each insert also contained 1 nmol of scyllo-inositol. The following standards were also prepared: 1 nmol scyllo-inositol only, 5 nmol Glc, 2.5 nmol Glc, 1 nmol Glc, 0.5 nmol Glc, and 0.25 nmol Glc. All inserts were dried in a rotational vacuum concentrator (1 h). Fifty microliters of 2 M trifluoroacetic acid (TFA) in water was then added, and the samples and standards were incubated for 2 h at 100°C and dried under N₂ (1 h). Ten microliters of MeOH was added, and the inserts were dried in a rotational vacuum concentrator to remove all H₂O (25 min). Fifty microliters of 0.5 M HCl in MeOH was added, and the samples and standards were incubated for 17 h at 80°C. Ten microliters of pyridine was added to neutralize the HCl, and the inserts were dried in a rotational vacuum concentrator (30 min). Twenty microliters of pyridine was added, followed by 20 μL of BSTFA plus 1% TMCS.

Samples and standards were analyzed on an Agilent 7890A gas chromatograph coupled to a 5975 mass spectrometer as a detector. The inlet temperature was 250°C, and helium was used as a carrier gas (purge flow of 20 mL/min and column flow of 1 mL/min). One microliter of the derivatized sample was injected. Chromatographic separation was achieved using a DB-5 capillary column (30 m by 0.25 mm by 1.00 μm). The oven had a starting temperature of 70°C, which was held for 1 min and then ramped to 295°C at 12.5°C/min and then to 320°C at 25°C/min. The transfer line temperature was 250°C, and the ion source temperature was 230°C. Compounds were fragmented using electron ionization. Spectra were acquired over the range of 50 to 500 m/z.

Data were analyzed using Agilent MassHunter software. The total ion chromatogram for each sample and standard was integrated. Sugars were identified based on the GC retention time and mass spectra of authentic standards. The ratio of the minor β-anomer of Glc (16.4 min) to scyllo-inositol (18.7 min) was determined, with that peak chosen as it was well resolved compared to the major α-anomer. A Glc calibration curve was constructed using the β-Glc/scyllo-inositol ratio from the 5 Glc standards and was used to determine the quantity of Glc in each of the samples. The quantity of Glc found in the scyllo-inositol blank was subtracted, and this value was used to determine the quantity of Glc in each cell pellet.

Analysis of disaccharide content.

One hundred twenty microliters of the soluble fraction of each sample and 1 nmol of scyllo-inositol were transferred to a glass insert (45). Standards were prepared for trehalose and maltose at the following concentrations: 100 nmol, 50 nmol, 25 nmol, 12.5 nmol, 6.25 nmol, 3.125 nmol, 1.563 nmol, 0.781 nmol, 0.391 nmol, and 0.195 nmol. All inserts were dried in a rotational vacuum concentrator (1 h). Twenty microliters of 20-mg/mL methoxyamine hydrochloride in pyridine was added, and the samples and standards were incubated with continuous mixing for 14 h at 25°C. Twenty microliters of BSTFA plus 1% TMCS were added, and samples and standards were incubated for 1 h at 25°C.

Samples and standards were analyzed on an Agilent 7890A gas chromatograph coupled to a 5975 mass spectrometer as a detector. The inlet temperature was kept at 250°C. Chromatographic separation was achieved using an Agilent CP9013 VF-5ms column (30 m by 0.25 mm by 0.25 μm) with a 10-m EZ-guard column. The oven had a starting temperature of 70°C, which was held for 2 min, ramped to 295°C at 12.5°C/min and then to 320°C at 25°C/min, and then held at 320°C for 3 min. The transfer line temperature was 280°C. Compounds were fragmented using electron ionization.

Data were analyzed using Agilent MassHunter software. The total ion chromatogram for each sample and standard was integrated. Sugars were identified based on the GC retention time and mass spectra of authentic standards. The ratio of trehalose (19.8 min) to scyllo-inositol (15.1 min) was determined. Calibration curves were constructed using the disaccharide/scyllo-inositol ratio from 7 of the standards (25 nmol, 50 nmol, and 100 nmol were excluded due the to loss of linearity) and used to determine the quantity of trehalose in each of the samples. This value was used to determine the quantity of trehalose in each cell pellet.

Monosaccharide derivatization and analysis.

Eighty microliters of the soluble fraction of each sample was transferred to a glass insert that also contained 1 nmol of scyllo-inositol (45). Standards were also prepared for glucose and fructose at the following concentrations: 100 nmol, 50 nmol, 25 nmol, 12.5 nmol, 6.25 nmol, 3.125 nmol, 1.563 nmol, 0.781 nmol, 0.391 nmol, and 0.195 nmol. All inserts were dried in a rotational vacuum concentrator (1 h). Fifty microliters of 0.5 M HCl in MeOH was added, and the samples and standards were incubated for 3 h at 80°C. Ten microliters of pyridine was added to neutralize the HCl, and the inserts were dried in a rotational vacuum concentrator (30 min). Twenty microliters of BSTFA plus 1% TMCS were added, and samples and standards were incubated for 1 h at 25°C.

Samples and standards were analyzed on an Agilent 7890A gas chromatograph coupled to a 5975 mass spectrometer as a detector. The inlet temperature was kept at 250°C. Chromatographic separation was achieved using an Agilent CP9013 VF-5ms column (30 m by 0.25 mm by 0.25 μm) with a 10-m EZ-guard column. The oven had a starting temperature of 70°C, which was held for 2 min, ramped to 295°C at 12.5°C/min and then to 320°C at 25°C/min, and then held at 320°C for 3 min. The transfer line temperature was 280°C. Compounds were fragmented using electron ionization.

Data were analyzed using Agilent MassHunter software. The total ion chromatogram for each sample and standard was integrated. Sugars were identified based on the GC retention time and mass spectra of authentic standards. Due to the high level of variation in internal standard (scyllo-inositol) peak areas across the standards and samples, peak areas were normalized to the lowest detected internal standard peak area. The ratio of the normalized peak area of the first-eluting Glc peak (13.96 min) to the area of the scyllo-inositol peak (15.1 min) was determined. Calibration curves were constructed using the Glc/scyllo-inositol ratio from 4 of the standards (6.25 nmol to 50 nmol) and used to determine the quantity of Glc in each of the samples. This value was used to determine the quantity of Glc in each cell pellet.

BCA protein assay.

Sixty microliters of insoluble metabolites suspended in MilliQ water from each sample was dried in a rotational vacuum concentrator (1.5 h at 50°C). Sixty microliters of 0.5% SDS in distilled water (dH₂O) was added, and the samples were vortexed, boiled for 5 min, and centrifuged (16,100 rpm for 5 min at RT) to pellet the cell debris. The supernatant and standards containing bovine serum albumin (BSA) were analyzed using a bicinchoninic acid (BCA) protein assay kit (Sigma).

Data availability.

Metabolomics data are available through the Metabolomics Workbench (46) under project ID: PR001508.