INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Access to complete genome sequence information for a number of microorganisms now motivates the development and application of experimental techniques for phenotype characterization (such as transcriptome and proteome analyses), providing arrays of data that can be directly mapped to corresponding arrays of genes (14, 36). The physiological counterpart to such composition arrays is the array of fluxes (reaction rates on a per-unit cell volume or per-unit cell mass basis) for all of the reactions that occur in the organism, for which we use, by analogy, the term fluxome. Approximate fluxome access for certain subsets of metabolism can be attained by methods of metabolic flux analysis, which require data on uptake and efflux rates of certain metabolites outside the cell and which assume a corresponding network of metabolic pathways in the cell (39). Alternatively, by use of more recently introduced methodology based on isotopic imprinting of amino acids by their precursors, the active central carbon pathways and the ratios of their fluxes can be directly determined from two-dimensional (2D) nuclear magnetic resonance (NMR) analysis of hydrolyzed cell protein (30-33). This method, for which we introduce the term METAFoR (metabolic flux ratio) analysis, offers a relatively high throughput access to these key fluxome elements, enabling physiological data arrays to be acquired over a broad range of genetic and environmental conditions.

Specifically, METAFoR analysis quantifies the relative abundance of intact carbon bonds originating from uniformly isotopically labeled source molecules by use of proton-detected 2D ¹³C-¹H correlation NMR spectroscopy (COSY) (30, 34, 42). Such 2D NMR analysis of amino acids obtained from hydrolyzed cell protein permits quantitative analysis of the relative abundance of intact, contiguous fragments in the precursor metabolites of central metabolism, because the carbon backbone of these molecules is conserved in the amino acids. Typically, fractional ¹³C labeling of amino acids is achieved by growing cells with a mixture of 85 to 90% natural-abundance glucose and 10 to 15% [U-¹³C₆]glucose (22, 27, 30-32, 34). Because alternative pathways leading to common intermediates or products produce different intact fragments originating from a single glucose source molecule (30-32), specific multiplet patterns in the ¹³C fine structures that reflect the in vivo usage of reactions are generated. Probabilistic equations relate the determined intensities of the multiplet components to the relative abundance of intact carbon fragments (30) and thus allow derivation of intracellular carbon flux ratios (30-33). These data provide not only comprehensive insight into cellular metabolism but also inherent flux indications that can provide critical information for metabolic (net) flux analysis (27, 32).

The active pathways and the flux distribution in central carbon metabolism are critical components of a multidimensional physiological representation of the organism, since this central backbone of metabolism provides energy, cofactor regeneration, and building blocks for biomass synthesis and controls the extent and nature of by-product excretion. A wide array of regulatory responses are embedded in this network on the transcriptional level as well as the protein level. The purpose of this complex regulatory structure is not yet fully elucidated, but the observed insensitivity of growth rates and extracellular fluxes to the overexpression of key enzymes suggests a homeostatic objective of the regulatory system (4, 8, 38).

In this study, we used METAFoR analysis to examine, at the level of flux ratios and operational pathways, how the central carbon physiology of Escherichia coli responds to genetic and environmental manipulations. In addition, we show the extent of variation in these facets of the central carbon network fluxome in several different standard laboratory strains. These METAFoR data show in detail how and under what conditions the E. coli central carbon metabolic network maintains flux ratio homeostasis and when significant alterations arise in both active pathways and flux ratios.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains, plasmids, and media. Strains and plasmids used in this study are listed in Table 1. All batch cultivations were performed with a minimal medium containing 5 g of glucose per liter, 48 mM Na₂HPO₄, 22 mM KH₂PO₄, 10 mM NaCl, and 30 mM (NH₄)₂SO₄. The following components were sterilized separately and then added (per liter of final medium): 1 ml of 1 M MgSO₄, 1 ml of 0.1 mM CaCl₂, 1 ml of 1 mg of vitamin B₁ per liter (filter sterilized), and 10 ml of trace element solution containing (per liter) 0.55 g of CaCl₂ · 2H₂O, 1 g of FeCl₃, 0.1 g of MnCl₂ · 4H₂O, 0.17 g of ZnCl₂, 0.043 g of CuCl₂ · 2H₂O, 0.06 g of CoCl₂ · 6H₂O, and 0.06 g of Na₂MoO₄ · 2H₂O. To ensure maintenance of plasmids, ampicillin was added to a final concentration of 25 mg/liter. The medium fed into the glucose-limited chemostat had the same composition as the batch medium, with the following exceptions (per liter): 3.6 g of glucose, 4.7 g of Na₂HPO₄ · H₂O, 3.0 g of KH₂PO₄, 0.5 g of NaCl, and 1 g of NH₄Cl. To enforce nitrogen limitation, the concentration of glucose was increased to 4.5 g/liter, and the concentration of the sole nitrogen source, NH₄Cl, was reduced to 0.7 g/liter. Chemostat media were sterilized by passage through a 0.2-µm-pore-size filter, and 10-fold-diluted trace element solution was added after filtration to prevent losses via precipitation.

Batch and chemostat cultivations. All batch cultivations were performed at 30°C. Aerobic batch cultures were grown in 1-liter baffled shake flasks with 150 ml of medium on a gyratory shaker at 200 rpm. Anaerobic batch cultivations were performed with rubber-sealed glass flasks previously flushed with N₂ and incubated in a gyratory water bath (G76D; New Brunswick). Chemostats were operated at 37°C in a 1.5-liter bench-top fermentor (Bioengineering) with a working volume of 1.0 liter and a constant dilution rate (D) of 0.2 h¹, meaning that the feed rate was 0.2 liter/h. The working volume was kept constant by removal of effluent from the center of the culture volume by use of a weight-controlled pump. The pH of the culture was maintained at 7.0 by automatic addition of 2.0 M NaOH with a pH controller and was verified periodically by off-line measurements. The airflow was maintained at 1 liter/min with filter-sterilized air by use of a volume flow meter, and the agitation speed was set to 1,200 rpm.

Analytical procedures. Cell growth during the cultivations was monitored by measuring the optical density at 600 nm (OD₆₀₀). For cellular dry weight (cdw) determination in selected cases, a known volume of fermentation broth was centrifuged for 10 min in preweighed glass tubes at 4°C and 3,000 × g, washed once with water, and dried at 90°C for 24 h to a constant weight. Samples for extracellular metabolite analysis were centrifuged for 1 min at maximum speed in an Eppendorf tabletop centrifuge to remove the cells. Glucose and ethanol concentrations were determined enzymatically (Synchron CX5CE apparatus; Beckman) with kits supplied by the manufacturer. Acetate (and ethanol, in selected cases) was measured by gas chromatography (5890E chromatograph; Hewlett-Packard) with a Carbowax MD-10 column (Macherey-Nagel) and butyrate as an internal standard. Concentrations of oxygen and carbon dioxide in the feed medium and off gas of bioreactor fermentations were determined with a mass spectrometer (Prima 600; Fisons Instruments).

Determination of physiological parameters. In batch cultures, the exponential growth phase was identified by log-linear regression of biomass concentration versus time, with growth rate (µ) as the regression coefficient. The biomass yield on the substrate (Y_X/S) was determined as the coefficient of a linear regression of biomass concentration (X) versus substrate concentration (S) during the exponential growth phase. A predetermined correlation factor (OD₆₀₀, 0.33) was used to convert the OD₆₀₀ values into cell concentrations for the calculation of specific conversion rates. The specific consumption rate for a substrate (q_S), e.g., glucose and O₂defined as the differential change in S with time (t) normalized to the biomass concentrationwas obtained as the coefficient of a linear regression of S (the change in S) versus X divided by µ, on the basis of the relationship S_t1  S_t2 = S = X_t2(q_S/µ). The same relationship holds for the specific rate of formation of products (P), e.g., acetate and CO₂. This relationship is linear provided that µ and q_S are constant. In a steady-state chemostat, µ is constant and equals D. In batch cultures, maximum µ was constant during the exponential growth phase, and all specific rates from batch experiments reported here refer to the exponential phase.

NMR sample preparation. For network topology and flux ratio analysis by NMR, a specified amount of culture was harvested and cells were centrifuged at 1,200 × g for 10 min at 4°C. The cell pellet was washed once with 20 mM Tris-HCl (pH 7.6) and centrifuged again. Washed pellets from chemostat cultures were resuspended in the above buffer, and cells were disrupted by sonication on ice three times for 45 s each time at 20% output (XL-2020 sonicator; Hert Systems). Cell debris was removed by centrifugation for 20 min at 9,000 × g. Sonication and centrifugation were repeated until cell lysis was virtually complete, as determined by visual inspection with a microscope. Small debris particles were removed by ultracentrifugation for 30 min at 33,000 × g. Cellular protein in the supernatant was precipitated overnight at 20°C after the addition of 60% (vol/vol) ethanol. The precipitate was resuspended in 6 ml of 6 M HCl and hydrolyzed by incubation in sealed Pyrex glass tubes for 24 h at 110°C. The hydrolysate was filtered through a 0.2-µm-pore-size filter and lyophilized. The dried material was dissolved in 600 µl of 20 mM deuterium chloride (DCl) in D₂O, incubated for 2 h at room temperature, centrifuged, and used for the NMR measurements. Washed pellets from batch cultures were directly resuspended in 6 M HCl and hydrolyzed.

NMR spectroscopy and data analysis. Proton-detected 2D ¹³C-¹H heteronuclear single-quantum COSY was performed with the pulse sequence of Bodenhausen and Ruben (3), which ensures that ¹H-¹³C scalar couplings do not affect the ¹³C-¹³C scalar coupling fine structure along the chemical shift ₁(¹³C) (34). Pulsed-field gradients were used for coherence pathway rejection (2, 40), and a 2-ms spin-lock pulse (24) was used to purge the magnetization arising from ¹²C-bound protons and the residual ²HOH signal. ¹³C decoupling during data acquisition was achieved by use of the composite pulse decoupling scheme GARP (28), and quadrature detection in ₁ was accomplished with States-TPPI (18). The spectra were recorded at a ¹H resonance frequency of 500 MHz by use of a Bruker DRX500 spectrometer; the sample temperature was 40°C. For each sample, two spectra were measured: one spectrum for the aliphatic resonances, with the ¹³C carrier set to 42.5 ppm relative to 2,2-dimethyl-2-silapentane-5-sulfonate sodium salt, and one spectrum for the aromatic resonances, with the ¹³C carrier set to 125.9 ppm. The spectra of the aliphatic resonances were folded along ₁(¹³C) with a sweep width of 33.8 ppm. The measurement time was 4.5 h per spectrum (1,706 × 256 complex points; t_1max, 402 ms; t_2max, 102 ms; relaxation delay between scans, 2 s). The spectra of the aromatic resonances were recorded in about 2.5 h (920 × 512 complex points; t_1max, 392 ms; t_2max, 87 ms; relaxation delay between scans, 2 s). Before Fourier transformation with the program PROSA (12), the time domain data were multiplied in t₁ and t₂ with sine-bell windows shifted by /2 (5). The digital resolutions after zero filling were 1.0 Hz/point along ₁ and 2.4 Hz/point along the second frequency axis ₂ for spectra of the aliphatic resonances and 0.6 Hz/point along ₁ and 5.8 Hz/point along ₂ for spectra of the aromatic resonances. One-dimensional ¹H NMR spectra (t_max, 1.022 s; 10-s relaxation delay between scans) were recorded to determine the overall degree of ¹³C labeling in the amino acids from the satellites of isolated proton peaks, which corresponds to P₁ in the probabilistic equations of Szyperski (30).

Biochemical reaction master network of E. coli. As an initial step for METAFoR analysis, a biochemical master network that comprises all currently known reactions of central carbon and amino acid metabolism for the organism under investigation is constructed (30). For E. coli, this network (Fig. 1) was compiled from textbooks (11, 20) and Internet-accessible metabolic databases (16). Inspection of observed intact carbon fragments in the amino acids subsequently allows identification of active biosynthetic pathways (27, 30-32). Such analysis cannot distinguish between trioses that are generated via the methylglyoxal bypass or via the Entner-Doudoroff pathway, since these pathways generate fragment patterns that are indistinguishable from those emerging from the glycolysis and pentose phosphate (PP) pathways, respectively (30). Because the methylglyoxal bypass and the Entner-Doudoroff pathway were reported to be inactive for E. coli cells grown with glucose (11), they were not considered in the presently used network.

View larger version (74K): [in this window] [in a new window]   FIG. 1.   Biochemical master network with reactions for identified genes or enzyme activities in E. coli. The information was compiled from the EcoCyc database (17) and other sources (11, 20). The reaction sets of glycolysis, the PP pathway, the TCA cycle (including the glyoxylate shunt), and C1 metabolism are shaded, and enzymes catalyzing key reactions are indicated in italics. The hatched arrow highlights the TCA cycle-replenishing anaplerotic reaction, and the broken arrows indicate the anaerobic pyruvate formate-lyase, which is interrupted in strain KO20. The grey arrows indicate the reactions catalyzed by the enzymes encoded on the pet operon of Z. mobilis. The intact carbon fragment patterns of boxed metabolites were directly determined by 13C-1H COSY of proteinogenic amino acids. Abbreviations: F6P, fructose-6-phosphate; Ru5P, ribulose-5-phosphate; X5P, xylulose-5-phosphate; E4P, erythrose-4-phosphate; S7P, seduheptulose-7-phosphate; PGA, 3-phosphoglycerate; SER, serine; GLY, glycine; AAD, acetaldehyde; CIT, citrate; ICT, isocitrate; OGA, oxoglutarate; SUC, succinate; GOX, glyoxylate; and DH, dehydrogenase.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Analysis of glucose- and ammonia-limited chemostat cultures of wild-type E. coli MG1655. Continuous cultivation was performed with aerobic chemostats at a D (volumetric flow rate/working volume) of 0.2 h¹ under glucose- or ammonia-limited conditions, representing two largely different bioenergetic regimens. Carbon-sufficient (i.e., ammonia-limited) cultures are known to exhibit metabolic behavior that differs from that of carbon-limited cultures with respect to specific substrate consumption rate, maintenance requirements, and by-product secretion (21). This fact is reflected by the physiological data from the cultures described here (Table 2). When the ammonia-limited and the glucose-limited cultures were compared, marked increases were found in the specific glucose consumption rate (q_glc) and in the specific rates of production of acetate and pyruvate, while the specific oxygen consumption rate and CO₂ evolution rate varied little between the two different conditions; these results indicated that there are only minor changes in respiratory metabolism. The low steady-state biomass concentration in the ammonia-limited culture results from the low ammonia concentration used (13.1 mM). The residual glucose concentration in this culture was slightly above 1 g/liter.

View this table: [in this window] [in a new window]   TABLE 2.   Aerobic growth parameters of glucose- or ammonia-limited chemostat cultures of E. coli MG1655 at a D of 0.2 h1

View larger version (42K): [in this window] [in a new window]   FIG. 2.   Origins of metabolic intermediates (A to P) during aerobic growth of E. coli MG1655 in glucose-limited or ammonia-limited chemostats. In certain cases, the NMR data permit the determination only of upper bounds (ub) or lower bounds (lb) on the origin of intermediates. The experimental error (error bars) was estimated from the analysis of redundant 13C scalar coupling fine structures and the signal-to-noise ratio of the 13C-1H COSY spectra by use of the Gaussian law of error propagation. The fraction of the total pool for a particular metabolite quantifies the ratio of this metabolite derived from a specified substrate to the sum of all other substrates that contribute to the pool of this metabolite. In cases where only two reactions contribute to one metabolite, e.g., OAA from PEP and PEP from OAA, the remaining fraction of the total pool can be attributed to the competing reaction. Abbreviations are explained in the text and in the legend to Fig. 1; rev., reversibly.

Analysis of wild-type batch cultures harvested at different growth phases. Chemostat experiments are the most suitable method of analyzing cells under steady-state conditions. On the other hand, a physiological steady state is also attained during the exponential growth phase in batch cultures, which are characterized by unrestricted growth at the maximum specific rate possible under the applied conditions. Because batch cultures enable more efficient parallel analysis of different strains, we adapted batch cultivations for METAFoR analysis of genetic effects on central carbon metabolism.

View larger version (13K): [in this window] [in a new window]   FIG. 3.   Growth of E. coli ATCC 11303 in aerobic batch cultures. The line represents the best fit to the exponential growth phase data, and the arrows indicate the times of biomass sampling for the METAFoR analysis.

Analysis of aerobic batch cultures of wild-type and mutant E. coli strains during exponential growth. Two commonly used but genetically different E. coli strains, JM101 (a K-12 strain) and ATCC 11303 (a B strain), were grown in shake flask cultures under aerobic conditions for direct comparison of their carbon metabolism. Although ATCC 11303 grew somewhat faster (Table 4), the fluxome data for the two wild-type strains were almost identical (Fig. 4), suggesting that central metabolism is very similar in the two E. coli strains. There were, however, small changes in the origins of metabolites involved in the nonoxidative part of the PP pathway, from ribose-5-phosphate (R5P) to triose-3-phosphate (T3P) (Fig. 4B, C, and E), indicating that ATCC 11303 cells exhibit a higher degree of exchange through transketolase II.

View larger version (43K): [in this window] [in a new window]   FIG. 4.   Origin of metabolic intermediates (A to P) during aerobic exponential growth of various E. coli strains. For more details, see the legend to Fig. 2.

Carbon metabolism during exponential growth in aerobic batch cultures of E. coli strains engineered for ethanol production. We chose ethanol-producing E. coli KO20 (23) as the host for a series of plasmids used for the expression of homologous and heterologous pyruvate kinases and phosphofructokinases (6). The expression of both genes was induced by use of isopropyl--D-thiogalactopyranoside (IPTG) (0.01 mM), leading to in vitro activities two- to ninefold above the control level, which was represented by the empty expression vector pTrc99a (Pharmacia). In terms of their growth physiology, all strains showed similar behaviors under aerobic conditions in batch cultures (Table 4). However, by-product formation was altered in the overexpression strains, with KO20::pPPec generating more acetate and KO20::pPYKbs generating more ethanol than KO20. Again, the METAFoR patterns of all strains derived from ATCC 11303 were identical, within experimental error, to those of KO20 in Fig. 4 (data not shown). Thus, expression of the pet operon, medium-copy-number plasmid maintenance, and the overexpression of either a heterologous pyruvate kinase or the homologous phosphofructokinase do not appear to have a pronounced influence on the flux ratios accessible by current METAFoR analysis.

Analysis of anaerobic batch cultures of wild-type and mutant E. coli strains during exponential growth. The physiological differences among the wild-type and mutant E. coli strains under anaerobic conditions were larger than those observed in aerobiosis (Table 5). The two wild-type strains, JM101 and ATCC 11303, showed significant differences with respect to glucose consumption as well as acetate production rates, both of which were higher in ATCC 11303. Accordingly, the apparent biomass yield was lower in this strain. While pyruvate kinase-deficient PB25 grew slower than its parent, JM101, ethanol-producing KO20 grew fastest of all the strains investigated. Although the specific glucose consumption rates of JM101 and PB25 were identical, the latter was severely impaired in its ability to form the normal anaerobic end products ethanol and acetate. On the other hand, KO20 produced about fivefold more ethanol than any other strain and exhibited a higher glucose consumption rate and a lower biomass yield on glucose than its parent, ATCC 11303. 

View larger version (44K): [in this window] [in a new window]   FIG. 5.   Origin of metabolic intermediates (A to N) during anaerobic exponential growth of various E. coli strains. For more details, see the legend to Fig. 2.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

This study provides novel insights into global metabolic network behavior. Variations resulting from different growth conditions and genetic backgrounds in E. coli are monitored by combined physiological and fluxome analyses. A particular strength of the METAFoR analysis used here is the ability to decipher the relative fluxes connecting the lower part of glycolysis with the TCA cycle, namely, the anaplerotic reaction and certain futile cycles which dissipate ATP. On the other hand, flux ratios of the oxidative versus the nonoxidative PP pathways (R5P from G6P; Fig. 2A, 4A, and 5A) and glycolysis versus the PP pathway (PEP molecules derived through at least one transketolase reaction; Fig. 2B, 4B, and 5B) are accessible only as upper and lower bounds, because the pool of pentoses is rapidly and reversibly turned over by transaldolase and transketolase. The key findings are as follows. (i) Intracellular carbon flux ratios in the central metabolism of E. coli are affected only a little by enzyme overexpression and are flexible toward gene disruption. (ii) Of all central carbon fluxes, those in the TCA cycle change most significantly in response to changes in environmental conditions. (iii) Reactions mediated by the malic enzyme and PEP carboxykinase and previously considered to be absent during growth on glucose were identified. (iv) A novel regulation phenomenon in which futile cycling through at least one set of reactions is increased under conditions of a very low extracellular glucose concentration was evident.

Interstrain differences. The METAFoR pattern of exponentially growing aerobic E. coli revealed surprisingly few interstrain differences (Fig. 4), although major changes in physiological parameters are documented for the various strains used here. One would expect such physiological differences to be reflected in fluxes through central metabolism (Fig. 1), which provides energy, cofactor regeneration, and building blocks for biosynthesis. Strain KO20, for example, which expresses the pet operon of Z. mobilis alcohol dehydrogenase II and pyruvate decarboxylase and has pyruvate formate-lyase knockout mutations, was previously described to exhibit significantly altered by-product formation, with ethanol as the major product (23). Furthermore, overexpression of pyruvate kinase and phosphofructokinase, which are major control enzymes in the glycolytic pathway, were shown to have a profound effect on glucose catabolism in resting E. coli KO20. Specifically, an increased glucose consumption rate was found for KO20::pPYKbs harvested from aerobic precultures (13), and a large shift from ethanol to lactate formation was described for KO20::pPPec harvested from anaerobic precultures (6).

Anaplerosis and the TCA cycle. METAFoR data provide information about the fraction of OAA molecules that originate from PEP (Fig. 2F, 4F, and 5F); this information quantifies the contribution of the TCA cycle-replenishing anaplerotic reaction to OAA generation, relative to that of MAL dehydrogenase in the TCA cycle (30-32). In the aerobic cultures studied here, the relative flux through anaplerotic PEP carboxylase is about 40%; the value for the ammonia-limited chemostat culture is increased to about 70% (Fig. 2F). Consistent with the extensive overflow metabolism seen in the ammonia-limited culture (Table 2), this increased anaplerosis indicates that, compared with the situation for the glucose-limited culture, a larger portion of the TCA cycle flux is used for biomass formation instead of energy generation. In the anaerobic regimen, the TCA cycle is reduced to a two-branch pathway (30) and OAA is generated by anaplerosis only (Fig. 5F).

Futile cycling. In the glucose-limited chemostat, a significant fraction of PEP molecules were found to originate from OAA (Fig. 2G). This result indicates in vivo activity of gluconeogenic PEP carboxykinase, an enzyme that is required for growth on carbon sources that are metabolized via the TCA cycle and that has previously been considered to be inactive in cells grown on glucose (11). In principle, PEP molecules could also originate from OAA via the reverse reaction of anaplerotic PEP carboxylase, but based on thermodynamic considerations and the absence of ¹⁴CO₂ exchange with OAA in enzyme assays (37), this notion is highly unlikely. Hence, we have evidence of an ATP-dissipating futile cycle via PEP carboxylase and ATP-consuming PEP carboxykinase. At the same D under ammonia-limited conditions in the chemostat, this futile cycle appears to be significantly less active (Fig. 2G). Similar activity levels for this cycle were seen with Bacillus subtilis (27) grown in a chemostat under glucose limitation at a D of about 0.1 h¹. In faster growing cells of B. subtilis in this chemostat, however, the contribution of this futile cycle was reduced, and the batch data presented here show it to be absent in exponentially growing E. coli (Fig. 4). On the other hand, for Corynebacterium glutamicum, a similar exchange between the PEP-PYR and OAA-MAL pools was described, not only for glucose-limited chemostat cultures (41) but also for batch cultures (29); however, it is not clear whether or not that pool exchange involved dissipation of ATP via a futile cycle. These data provide evidence for a metabolic regulation phenomenon in E. coli and B. subtilis in which futile cycle activity is less tightly controlled under extreme glucose limitation than under glucose excess, as in slow-growing chemostat cultures. It is tempting to speculate that this reduced control is caused by the extremely low extracellular glucose concentration and a concomitant reduction in catabolite repression. This hypothesis is also supported by the observation that PEP carboxykinase expression in E. coli is repressed by glucose (10).

Exchange reactions. METAFoR analysis affords a qualitative assessment of several exchange fluxes (30). In the experiments analyzed here, the fraction of R5P molecules originating from T3P and a C₂ unit via the transketolase reaction was usually about 70 to 80% (Fig. 2C and 4C) and, under anaerobic conditions, even as high as 90% (Fig. 5C). In contrast, a much lower fraction of R5P molecules originated from erythrose-4-phosphate (Fig. 2D, 4D, and 5D), representing either an exchange via transaldolase or a recycling of PP pathway-generated fructose-6-phosphate to G6P and on to R5P. Similar insights into the PP pathway have previously been reported for batch cultures of E. coli K-12 and B strains by mass spectrometric analysis of ¹⁸O-labeling patterns in the ribose moiety-containing nucleotides (15) and by METAFoR analysis (30). Rapid exchange of metabolite pools in the PP pathway was also described for C. glutamicum (19, 41). In B. subtilis, the exchange mediated by transketolase appears to be less significant, since about 50% of the R5P in slow-growing, glucose-limited chemostat cultures was found to contain intact C₅ fragments from the source glucose (27).