When bacteria are cultured in medium with multiple carbon substrates, they frequently consume these substrates simultaneously. Building on recent advances in the understanding of metabolic coordination exhibited by Escherichia coli cells through cAMP‐Crp signaling, we show that this signaling system responds to the total carbon‐uptake flux when substrates are co‐utilized and derive a mathematical formula that accurately predicts the resulting growth rate, based only on the growth rates on individual substrates.
When cultured in medium containing two carbon substrates, E. coli frequently consumes both simultaneously. A mathematical formula based on simple assumptions accurately predicts the resulting growth rate from the growth rate on each substrate alone.
Catabolite repression by cAMP‐Crp regulates the total carbon uptake flux through a negative feedback loop.
The uptake of one substrate non‐specifically reduces the uptake of a second, co‐utilizable substrate.
A growth‐rate composition formula is derived that accurately predicts the growth rate on two co‐utilized substrates based on the growth rate on either substrate alone.
Bacterial cultures grown in minimal media supplemented with two carbon substrates (i.e., mixed‐substrate media) can exhibit two types of behavior: In some cases, the substrates are consumed sequentially—which under the right conditions results in diauxic growth (Monod, 1942, 1947)—whereas in other cases, they are consumed simultaneously (Monod, 1942). Sequential utilization and diauxie are commonly attributed (Müller‐Hill, 1996; Deutscher et al, 2006; Narang & Pilyugin, 2007) to catabolite repression by the cAMP‐Crp regulatory system (Kolb et al, 1993; Busby & Ebright, 1999), even though specific studies have shown cAMP‐Crp regulation to be either not necessary (Inada et al, 1996) or not sufficient (Okada et al, 1981) for diauxie. We here study the governing role of the cAMP‐Crp in the simpler case of simultaneous substrate utilization.
Recently, You et al (2013) reported a physiological study of E. coli in steady‐state exponential growth in minimal media supplemented with a single carbon substrate. To study the role of cAMP‐Crp, which activates a large number of carbon‐catabolic genes, the expression of the well‐studied lac system (with LacI inactivated by IPTG) was used as a reporter of cAMP‐Crp activity. It was found that LacZ expression level (Ez) exhibits a negative linear correlation with the growth rate (λ) when the carbon substrate in the medium is varied: (1)
Here, λC is the horizontal intercept, as illustrated in Fig 1A. Several other Crp‐activated catabolic operons were studied; they each show a similar linear relation, each with a horizontal intercept λC of 1.1 to 1.2/h. These results suggest that this ‘C‐line’ is a common response pattern exhibited by carbon‐catabolic genes under variation of the carbon influx, mediated by cAMP‐Crp regulation. Indeed, the cAMP excretion rate, a proxy for the intracellular cAMP level, shows the same trend and intercept, and removal of Crp binding obliterates the C‐line.
Functionally, the C‐line reflects the cell's coordination of its proteome in response to the different demands for ribosomes and metabolic enzymes at different growth rates (Scott et al, 2010; You et al, 2013; Chubukov et al, 2014). During fast growth, a large fraction of the cell's proteome must be allocated toward ribosomal proteins and anabolic enzymes; therefore, a reduced expression of carbon‐catabolic enzymes is obligatory. Mechanistically, this reduction results from an inhibitory effect of several α‐ketoacids on the synthesis of cAMP by adenylate cyclase (see Fig 1B) (You et al, 2013). Here, we apply these insights to growth on two carbon substrates to derive a formula that predicts the resulting growth rate.
Results and Discussion
The proposed theory of mixed‐substrate usage is based on three ingredients
First, let the expression level of the catabolic enzymes for the two carbon substrates, C1 and C2, be E1 and E2. Then, at saturating substrate concentrations, the carbon uptake flux Ji for substrate Ci is given by (2)where ki is the kinetic constant.
Second, a larger growth rate requires a larger carbon uptake flux. We expect that both substrates contribute to the production of biomass, such that the resulting growth rate λ12 obeys (3)
Here, the constants ci reflect the carbon efficiency of growth on substrate Ci, that is, the amount of that substrate that must be consumed to support a given growth rate. In writing equation (3), we allow these efficiencies to differ between substrates, but assume that they do not change drastically as a function of the growth rate (over the range of growth rates studied). At very high growth rates, this is likely inaccurate due to increased carbon excretion in the form of, for example, acetate (el‐Mansi & Holms, 1989; Han et al, 1992); however, because the rate of acetate excretion is generally small compared to the rate of carbon uptake (Han et al, 1992), equation (3) should remain a reasonable approximation.
Third, since the driver of cAMP‐Crp regulation is established to be the α‐ketoacids (You et al, 2013), which we hypothesize respond to the total carbon influx regardless of whether it originates from a single or multiple substrates (Fig 1B), we expect the C‐line observed for single‐substrate growth to hold also during growth on two substrates (as validated below). More precisely, we expect that E1(λ) ∞ E2(λ) ∞ Ez(λ), so that (4)for each substrate Ci. The horizontal intercept λC is expected to be the same for all substrates, but the vertical intercept is substrate specific. Equation (4) describes a cAMP‐Crp response that is a function of the growth rate only, regardless of the number of substrates. It assumes that the expression levels are not affected by any regulators other than cAMP‐Crp.
The system described by equations ((2), (3), (4)) is illustrated in Supplementary Fig S1. Each substrate contributes to the total carbon‐uptake flux (gray arrow). The carbon‐uptake flux Ji of substrate Ci is proportional to the expression of the responsible enzymes Ei (equation (2)). If the total carbon flux is increased, for example, by adding a co‐utilizable substrate to the growth medium, this permits a higher growth rate λ (equation (3)). However, an increased growth rate entails a reduced expression of the catabolic enzymes (red inhibitory lines), prompted by the increased α‐ketoacid pools (Fig 1B) and quantified by the C‐line (equation (4)); this reduces both fluxes Ji. Combined, equations ((2)–(4)) describe a cAMP‐Crp‐mediated negative feedback loop; they exploit the C‐line to quantitatively describe the feedback regulation shown in Fig 1B.
To derive an expression for the resulting growth rate, we first manipulate equations ((2), (3), (4)) to obtain (5)For growth on a single substrate Ci (with growth rate λi = ci Ji), the same equations yield (6)Substituting equation (6) into equation (5), we obtain: (7)This growth‐rate composition formula is the primary result of this report. It provides a quantitative prediction of the growth rate on two co‐utilized carbon substrates, based on the growth rates on each substrate alone. Remarkably, it depends on just a single (strain‐dependent) parameter, λC, whose value was already estimated above based on data from single‐substrate growth. The formula therefore has no tunable parameters whatsoever.
We stress that the above model includes cAMP‐Crp signaling as the only mechanism regulating catabolic enzymes; equation (7) can therefore be interpreted as a null expectation, obeyed in the absence of additional layers of regulation. That said, for many pairs of substrates, no additional regulation is known.
To validate this theory of mixed‐substrate growth experimentally, we repeated the physiological study of catabolite repression in E. coli K12 cells as done before (You et al, 2013), but now for growth on 23 pairs of substrates, listed in Table 1. We selected these as follows. We previously noticed that substrates that merge into the upper part of glycolysis (‘upper’ substrates) are often co‐utilized with those entering at the bottom or directly into the TCA cycle (‘lower’ substrates) (You et al, 2013). Therefore, we combined three ‘lower’ substrates (succinate, pyruvate, and oxaloacetate) with five ‘upper’ substrates (mannose, xylose, glycerol, maltose, and glucose) to form 15 combinations, referred to as group A. In all these cases, the growth rate on two substrates was larger than on either substrate alone, strongly suggesting some degree of co‐utilization (see Table 1). For six of these 15 substrate pairs (green entries to the left of the vertical line in Table 1), we also verified their co‐utilization by measuring the uptake of both substrates from the growth medium (see Supplementary Fig S2, first two columns). For comparison, we also paired the five ‘upper’ substrates mentioned above with either glucose or glycerol (group B). Glucose uptake is known to inhibit the uptake of many ‘upper’ substrates (verified in Supplementary Fig S2, 4th column) through the ‘inducer exclusion’ effect mediated by PTS enzyme EIIAglc (Postma et al, 1984). Glycerol uptake is suppressed in the presence of other ‘upper’ substrates through feedback inhibition mediated by fructose‐1,6‐biphosphate, a key glycolytic intermediate (Zwaig & Lin, 1966). As a result of these additional interactions, the pairs in group B are not expected to obey the above theory, while those in group A may, provided the relevant uptake systems are not subject to unknown regulation in addition to cAMP‐Crp.
We first tested the hypothesis that catabolic gene expression remains on the C‐line during growth on multiple substrates (equation (4)); this is expected for both groups A and B if the cAMP‐Crp system indeed responds only to the total carbon flux. Fig 2A confirms this: The expression of LacZ during mixed‐substrate (orange squares) and single‐substrate (yellow circles) growth is well described by a single C‐line. This result is the first direct confirmation of an important regulatory strategy implied by previous findings (You et al, 2013): that cAMP‐Crp responds to the total carbon‐uptake flux of the cell rather than to the availability of particular carbon substrates, as commonly thought.
An important consequence of the global negative feedback illustrated in Fig 1B and Supplementary Fig S1 is that the uptake of one carbon substrate should indirectly reduce the uptake of a second. We tested this for six substrate pairs from group A. In all cases, the measured uptake of each substrate was reduced in the presence of the other (see Fig 2B–E and Supplementary Fig S3A and B). In contrast, this was not the case for examples from group B (Supplementary Fig S3C–E).
In our model, the reduced uptake in the presence of another co‐utilized substrate follows from a reduced expression of the substrate's uptake system. To test this, we constructed two strains (strains NQ360 and NQ1513; see Supplementary Table S1) that report the expression of the glpFp and dctAp promoters, respectively driving transcription of genes encoding the glycerol and succinate uptake systems (Fig 2F). During growth on both glycerol and succinate, reporter expression from each promoter was much lower than during growth on the corresponding substrates alone (black arrows).
Next, we compare the predicted growth rates on mixed substrates to the measured values in Table 1. In Fig 3A, the growth rates on mixed substrates are plotted against the growth rate on the ‘upper’ substrate alone, with the ‘lower’ substrate fixed to be succinate (orange squares), pyruvate (purple circles), or oxaloacetate (green triangles). Solid lines indicate the prediction based on equation (7). There is excellent agreement between theory and experiment. This is also seen in Fig 3B, where all measured growth rates are plotted against their predicted values, for both group A (filled symbols) and group B (open symbols). The data for group A align with the diagonal (their regression line has slope 0.99 ± 0.18 and offset (0.00 ± 0.07)/h (95% CI), with R2 = 0.92), indicating that the theory accurately predicts the measured values. All outliers belong to group B.
An important feature of equation (7) is that the growth rate on mixed substrates λ12 is always smaller than a direct sum of λ1 and λ2 (see Fig 3C). Also, even if the single‐substrate growth rates are already close to λC, the mixed‐substrate growth rate never exceeds λC (horizontal dotted line in Fig 3C), illustrating that λC acts as a ‘speed limit’ for carbon‐limited growth. This limit reflects that, even in the hypothetical situation in which the cell does not need to express any catabolic enzymes to achieve the required carbon uptake flux (Ei = 0), the growth rate would still be finite, limited by the finite rate of protein synthesis, anabolic processes, and the uptake of other elements.
In biotechnological applications, mixed‐substrate growth is often preferred (Harder & Dijkhuizen, 1982; Joshua et al, 2011; Nakashima & Tamura, 2012). The growth‐rate composition formula derived here, taking into account only the effect exerted by cAMP‐Crp regulation on the expression of all catabolic systems (You et al, 2013; Hui et al, 2015), serves as the baseline for this important mode of bacterial growth. Deviations from this baseline suggest the presence of additional interactions affecting growth, such as inducer exclusion (Deutscher et al, 2006) or metabolic feedback inhibition (Zwaig & Lin, 1966; Zwaig et al, 1970) (see also the discussion in the caption of Table 1). Cases where additional interactions occur could therefore be found by identifying substrate pairs that do not obey the theory. Such interactions may be described as modulations on top of the null‐model presented here. This approach could also provide a new understanding of the classical phenomena of diauxie and sequential utilization.
Materials and Methods
Construction of chromosomal dctAp‐lacZ
The dctA promoter (dctAp) region containing the first 20‐amino‐acid sequence (−283 to + 111 relative to the transcriptional start site) was PCR amplified from E. coli MG1655 chromosomal DNA using the PdctA‐Xho‐F and PdctA‐ER‐R oligonucleotides (Supplementary Table S2). The PCR product was cloned as an XhoI‐EcoRI fragment on pKD13‐rrnBt:PLtet‐O1 (Klumpp et al, 2009), yielding pKDT‐dctAp. A tandem array of kan gene, rrnB terminator, and dctAp (kan:rrnBt:dctAp) on this plasmid was amplified by primers PdctA‐Z‐P1 and PdctA‐Z‐P2 (Supplementary Table S2). PdctA‐Z‐P1 contains a 50‐bp region that is homologous to the lacI promoter region, while PdctA‐Z‐P2 contains a 51‐bp region that is homologous to the first 50‐bp region of the lacZ structural gene. The PCR product was used to transform the strain NQ309 (You et al, 2013) by using the λ Red system (Datsenko & Wanner, 2000). Substitution of kan:rrnBt:dctAp for the lacI and lacZ promoters was verified by PCR and subsequent DNA sequencing. NQ1513 was made by moving the construct kan:rrnBt:dctAp:lacZ into NCM3722 by P1 transduction.
Growth conditions and β‐galactosidase activity assays
Batch culture growth and β‐galactosidase activity assays were performed as described previously (You et al, 2013). The culture medium used was N−C− minimal medium (Csonka et al, 1994) supplemented with 20 mM NH4Cl, saturating amounts of either a single carbon substrate or a combination of two carbon substrates, and 1 mM isopropyl β‐D‐1‐thiogalactopyranoside (IPTG). The concentrations of the various carbon substrates used were as follows: 20 mM fumarate, 20 mM mannose, 15 mM succinate, 20 mM fructose, 20 mM xylose, 20 mM pyruvate, 0.4% (v/v) glycerol, 20 mM maltose, 20 mM oxaloacetate, 0.4% (w/v) glucose, and 0.2% (w/v) lactose.
Carbon‐substrate uptake assays
Carbon substrate concentrations in the medium were measured as follows. The cells were grown on N−C− minimal medium supplemented with 20 mM NH4Cl, 1 mM IPTG, and one or two carbon substrates. The concentrations of carbon substrates used were as follows: 6 mM glucose, 8 mM glycerol, 8 mM xylose, 3 mM maltose, 6 mM mannose, 15 mM succinate, and 20 mM pyruvate. (The growth rates at these concentrations are almost identical with those at the concentrations used for Table 1.) A fraction of the exponentially growing culture was subjected to centrifugation at 16,110 g for 1 min; the supernatant was stored frozen at −80°C. Commercially available kits were used to measure glucose (GAHK20; Sigma‐Aldrich; St Louis, MO, USA), glycerol (FG0100; Sigma‐Aldrich), xylose (K‐XYLOSE; Megazyme), mannose (K‐MANGL; Megazyme; Bray, Ireland), succinate (K‐SUCC; Megazyme; Bray, Ireland), and pyruvate (K‐PYRUV; Megazyme; Bray, Ireland) according to manufacturer's instruction. For the maltose assay, maltose was first digested by incubating 5 μl samples with 30 μl of 100 mM potassium phosphate buffer, pH 6.5, containing 16.7 U/ml α‐glucosidase (G0660; Sigma‐Aldrich) and 38.5 mM β‐mercaptoethanol at 37°C for 30 min. The released glucose was measured using the glucose assay described above. The carbon uptake rate was calculated as the slope of the plot of carbon substrate concentrations versus OD600 multiplied by the specific growth rate.
This research is supported by the NIH to TH (R01GM109069) and by the NSF through the Center for Theoretical Biological Physics (PHY0822283). RH was also supported by the NWO (Nederlandse Organisatie voor Wetenschappelijk Onderzoek) (VENI 680‐47‐419).
CY, HO, and TH designed the experimental study. HO, CY, and NW performed experiments and analyzed the data. RH and TH formulated the theory. RH, HO, and TH wrote the paper.
Conflict of interest
The authors declare that they have no conflict of interest.
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure legends
Source Data for Supplementary Figure S2
Source Data for Supplementary Figure S3
This is an open access article under the terms of the Creative Commons Attribution 4.0 License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
- © 2015 The Authors. Published under the terms of the CC BY 4.0 license