Overflow metabolism provides a selective advantage to Escherichia coli in mixed cultures

Abstract

Purpose

It has previously been shown that organic acids produced by Escherichia coli suppress the growth of Pseudomonas aeruginosa in co-cultures under conditions of glucose excess, due to overflow metabolism. Inactivation of genes involved in central carbon metabolism favours fermentation of glucose over respiration and therefore increases production of organic acid by-products such as acetate and lactate. We sought to extend and refine the list of genes known to contribute to the metabolic balance between respiration and fermentation, to better understand the role of overflow metabolism in competitive survival of E. coli.

Methods

We confirmed the previous finding that E. coli excludes P. aeruginosa from co-cultures by producing organic acids in the presence of glucose. Using a genome-wide transposon screen we identified E. coli genes that are important for survival in co-cultures with P. aeruginosa, both with and without glucose supplementation.

Results

Central carbon metabolism was the dominant gene function under selection in our experimental conditions, indicating that the observed inhibition is a side-effect of overflow metabolism adopted by E. coli as a response to high glucose concentrations. The presence of a competing species increased the selective pressure for central carbon metabolism genes, with 31 important for growth in the presence of P. aeruginosa and glucose, while only 9 were significant for pure E. coli cultures grown with glucose. In our experiments, each transposon mutant was competed against all others in the pool, suggesting that overflow metabolism provides benefits to individual E. coli cells in addition to competitive inhibition derived from acidification of the growth medium.

Conclusion

Co-culture assays using transposon mutant libraries can provide insight into the selective pressures present in mixed species competition. This work demonstrates central carbon metabolism is the dominant gene function under selection in E. coli for aerobic growth in glucose and a side-effect of this is overflow metabolism which can inhibit growth of bystander species.

Introduction

Overflow metabolism, also known as the Warburg effect (Warburg 1956) or the Crabtree effect (Crabtree 1929; Deken 1966), is the tendency for cells to utilise a mixture of respiratory and fermentative metabolism under aerobic conditions when carbon sources are in plentiful supply (Holms 1996). This strategy is commonly observed in cells from all domains of life but a well-studied example is the bacterium Escherichia coli, which employs a partly fermentative metabolic strategy at high growth rates, leading to the formation of acetate and other organic acids (Wolfe 2005; Eiteman and Altman 2006).

Respiration maximises the yield of chemical energy, in the form of NADH and ATP, by oxidising every C-C bond in a carbon source. However, the ubiquity of overflow metabolism shows that absolute efficiency is not the only factor affecting the metabolic balance of cells. Models to explain and predict the seemingly sub-optimal mixture of metabolic pathways in overflow metabolism therefore continue to be suggested. These include the effects of cytoplasmic molecular crowding (Vazquez and Oltvai 2016), the idea that fermentative pathways place fewer demands on the cells’ protein expression machinery (Basan et al. 2015), the limited surface area of the inner membrane for respiratory enzymes (Szenk et al. 2017), ATP dynamics (Lin and Jacobs-Wagner 2022), and combinations of these and other factors (Swain and Fagan 2019; de Groot et al. 2020; Regueira et al. 2021). However, a full understanding of the causes and regulation of overflow metabolism remains elusive.

E. coli is a pioneer coloniser of the infant human gut (Mackie et al. 1999), and certain probiotic strains like E. coli Nissle 1917 can help to maintain a healthy balance of species and outcompete pathogenic bacteria through regulated processes such as the production of microcins and signalling compounds including indole (Sassone-Corsi et al. 2016; Zarkan et al. 2020). One effect of the production of organic acids during overflow metabolism is that E. coli is able to inhibit the growth of other bacteria such as Pseudomonas aeruginosa (Christofi et al. 2019) and Vibrio cholerae (Sengupta et al. 2017; Nag et al. 2018). This inhibition, therefore, is essentially a side-effect of overflow metabolism as a response to high-glucose environments, in which E. coli is better able to tolerate the consequent acidic conditions (Christofi et al. 2019).

P. aeruginosa is a versatile coloniser of habitats ranging from natural settings including soil, fresh water and marine environments (Kidd et al. 2012; Diggle and Whiteley 2020) to built environments such as drains, sewers, and surfaces within households and hospitals (Bédard et al. 2016; Gilbert and Stephens 2018; Crone et al. 2020). In humans, P. aeruginosa is rarely pathogenic for healthy carriers but is a prodigious opportunistic pathogen of immunocompromised people. In some cases, the gut is thought to act as a reservoir of infection that facilitates migration to the respiratory and urinary tracts (Kerr and Snelling 2009; Streeter and Katouli 2016; Cornforth et al. 2018; Wheatley et al. 2022).

Christofi et al (2019) used a panel of defined knockout mutants to investigate the inhibition of P. aeruginosa by organic acids produced by E. coli. Each mutant was competed individually against P. aeruginosa. While this is useful for identifying the effects of individual genes, their approach does not allow for genetic heterogeneity within a population. It is also unclear whether other mechanisms are important for E. coli survival and competition against P. aeruginosa under the same conditions.

We used a library of more than 800,000 E. coli transposon-insertion mutants to assay the effects of mutations in every E. coli gene on overflow metabolism in the presence of glucose. In this Transposon-Directed Insertion Sequencing with Expression (TraDIS-Xpress) library, the transposon encodes an outward-transcribing, inducible promoter to modulate gene expression in addition to conventional inactivation. Importantly, this allows the interrogation of all genes, including essential genes that would otherwise not be assayed (Yasir et al. 2020). We grew co-cultures consisting of the transposon mutant library and a single strain of P. aeruginosa, with and without glucose. This approach, in which all mutants are competed against each other and P. aeruginosa simultaneously, effectively rules out the contribution of an individual mutant population towards the accumulation of organic acids in the medium because all cells experience the same conditions no matter their contribution to establishing their environment. Our hypothesis was that using TraDIS-Xpress would allow us to broaden the panel of genes identified by Christofi et al., with a focus on the intracellular mechanisms required to thrive in high-glucose environments and in complex competitive situations.

Materials and methods

Strains and growth conditions

The TraDIS-Xpress mutant library, described previously by Yasir et al. (2020), was constructed by electroporating a customised mini-Tn5 transposon into E. coli strain BW25113, with selection of consequent transposon mutants on Miller’s formulation of lysogeny broth agar (LBA) containing kanamycin (30 μg.mL⁻¹). Aliquots of the transposon library were recovered from storage at − 70 °C and approximately 10⁷ cells (estimated by OD₆₀₀ with 1 OD₆₀₀ unit = 10⁹ CFU) were inoculated into 5 mL of Miller’s lysogeny broth (LB) in a screw-capped 50 mL tube. For co-cultures, an equal number of P. aeruginosa strain PA14 cells from an overnight culture in LB were added. An identical co-culture was grown in parallel, except that sterile glucose was added to the medium at 4% (w/v). Cultures consisting of only the E. coli transposon library were also grown under the same conditions, in LB with and without glucose. Each culture was grown in duplicate. All cultures were incubated for 24 h at 37 °C with shaking at 200 rpm. Cultures reached stationary phase at a density of approximately 3 × 10⁹ cells.mL⁻¹ having experienced approximately 10 generations.

Quantifying numbers of each species

A single colony of P. aeruginosa strain PA14 or approximately 10⁷ cells from the E. coli strain BW25113 TraDIS-Xpress library were inoculated into 5 mL of LB and grown overnight to reach stationary phase. Equal numbers of cells from each overnight culture were then transferred into fresh LB, with or without 4% glucose, as described above. Pure cultures of E. coli and P. aeruginosa and co-cultures of both species were grown in parallel. After 24 h growth at 37 °C with shaking at 200 rpm, tenfold serial dilutions were performed on each culture and 100 µL of the 10⁻⁶ and 10⁻⁷ dilutions were spread on LBA plates containing kanamycin (for selective growth of E. coli) and trimethoprim (for selective growth of P. aeruginosa). The plates were incubated overnight at 37 °C and number of colony-forming units (CFU) were determined from plates with approximately 10 – 100 colonies. Each culture was grown and enumerated in triplicate.

Sequencing of transposon insert sites

DNA was extracted from the cell pellets of each culture using a Quick-DNA fungal/bacterial 96 extraction kit (Zymo Research). Transposon-containing DNA fragments were prepared for sequencing using a Nextera DNA library preparation kit (Illumina) with custom i5 primers Yasir et al (2020). The sequencing libraries were selected for a size range of 300 – 500 bp and sequenced using the Illumina NextSeq 500 platform with NextSeq 500/550 high output v2 kit for 75 cycles as described by Yasir et al (2020).

TraDIS-Xpress data analysis

Nucleotide sequence data were converted to fastq format using bcl2fastq v. 2.20 (Illumina) then reads were mapped to the E. coli strain BW25113 reference genome sequence (CP009273) using the Burrows-Wheeler Aligner (Li and Durbin 2009) in Bio-TraDIS v. 1.4.1 (Barquist et al. 2016) to generate maps of insertion site frequency. Frequency of insertions for test conditions (in duplicate) were compared to the E. coli-only control using AlbaTraDIS v. 0.0.5 (Page et al. 2020), generating measures for log₂ fold-change (log₂FC) in sequence reads (as a proxy for number of cells containing each mutation in test conditions versus the control) and a p-value of statistical significance adjusted for the false discovery rate (q). Significantly changed mutants were defined as those with − 1 > log₂FC > 1 and q < 0.01. The full output of AlbaTraDIS is provided in Supplementary Tables 1, 2 and 3.

Determination of pH and organic acid concentrations in culture supernatants

Cultures were centrifuged and then filtered (0.2 µm cellulose acetate) to remove cells. The supernatant pH was measured using Whatman Panpeha universal indicator paper strips, with a pH range of 0 – 14 and sensitivity of 0.5 pH units.

Samples of supernatant (100 µL) were then mixed with internal standard (100 µL of 100 µg.mL⁻¹ D5-propionic acid in methanol) and derivatised to the nitrophenylhydrazine by addition of 50 µL each of 50 mM 3‑nitrophenylhydrazine hydrochloride in 80% methanol, 50 mM dimethylaminopropyl)‑N‑ethylcarbodiimide hydrochloride in 80% methanol and 7% pyridine in methanol. The mixture was shaken and heated at 40 °C for 30 min before terminating the reaction by addition of 0.5% aqueous hexanoic acid (250 µL). Liquid chromatography coupled to mass spectrometry (LC–MS/MS) analysis was conducted using a Waters Acquity UPLC coupled to a TQ-Absolute triple quadrupole mass spectrometer. The analytical column was a Waters Premiere HSS T3 C18 100 × 2.1 mm × 1.7 µm at 35 °C. Mobile phases were solvent A: water with 0.1% formic acid and solvent B: acetonitrile with 0.1% formic acid with a constant flow of 0.5 mL.min⁻¹. A gradient of solvent B was applied as 20% at injection for 1 min, 55% at 7 min, 100% at 7.5 min and held for 2 min, then back to initial conditions for 3 min. Sample injections were 0.2 µl. Samples were undiluted for measurements of butyric acid, formic acid, lactic acid and propionic acid, and diluted 50-fold for measurement of acetic acid. Quantification was made against standard preparations of target compounds (5 – 100 µg.mL⁻¹). Results for formic acid are reported as concentrations above the background level of approximately 20 µg.mL⁻¹.

Determination of pyocyanin concentrations in culture supernatant

Pyocyanin concentrations were measured spectrophotometrically using a modified version of the method of Essar et al. (1990). Centrifuged, filtered supernatants (2.5 mL) were vortex mixed with chloroform (1.5 mL) then centrifuged at 8,000 × g for 5 min. The chloroform layer was taken (1.4 mL) and added to 0.5 mL of 0.2 N HCl then re-extracted by vortex mixing and centrifuged again. The acid fraction (0.45 mL) was mixed with a further 0.45 mL of 0.2 N HCl and the absorbance was measured at 520 nm. The concentration, measured as µg.mL⁻¹ was calculated by A₅₂₀ × 34.144.

Results

Disruption of respiration pathways facilitates survival during overflow metabolism

Following growth of the transposon mutant library in LB containing 4% glucose, 12 genes were identified where the abundance of transposon insertions significantly changed (− 1 > log₂FC > 1, and q < 0.01) compared to growth in LB without glucose (Table 1). Nine of these 12 genes code for respiratory metabolic functions, and contained significantly more mutants within the gene. These mutations are likely to shift metabolism towards fermentation under aerobic conditions. Therefore, under high glucose concentrations, a selective advantage was conferred by mutations that hinder or prevent respiration, even without additional competition with P. aeruginosa. The remaining 3 genes encode functions in amino acid metabolism (dapF) and re-structuring of the cell envelope during cell division (ftsN and prc).

Table 1 Significant loci for the growth of E. coli with 4% glucose

E. coli inhibition of P. aeruginosa requires high concentrations of glucose

Co-cultures of E. coli and P. aeruginosa grown in LB without glucose permitted growth of both species to approximately the same density as when they were grown in isolation (Fig. 1). During overnight growth, the co-culture became a blue-green colour due to production of the antimicrobial compound pyocyanin by P. aeruginosa (Lau et al. 2004). Pyocyanin is known to inhibit E. coli growth (Christofi et al. 2019) but did not significantly reduce the number of E. coli in co-cultures without glucose. P. aeruginosa produced the highest concentration of pyocyanin (9.1 µg/mL) when grown alone without glucose (Fig. 2). Pyocyanin production decreased to 4.4 µg/mL when glucose was added, and dropped to 2.0 µg/mL when P. aeruginosa was competed against E. coli without glucose. When glucose was introduced to the competitive culture, pyocyanin production could not be detected. Previous reports have suggested a MIC of pyocyanin against E. coli of at least 20 µg/mL (Shouman et al. 2023).

Fig. 1

Comparison between growth of E. coli BW25113 TraDIS-Xpress library and P. aeruginosa PA14, either alone or in co-cultures and in the presence or absence of 4% glucose. A Cultures (n = 3) were grown in LB for 24 h, as described in Materials and Methods then diluted and spread on LBA plates supplemented with appropriate antibiotics. Colonies were counted after overnight incubation at 37 °C. Bars indicate mean values and points indicate individual data. The p-value result of a two-way T-test comparing average CFU/mL of E. coli and P. aeruginosa under each condition is given above each pair of bars. B The colour of a culture after 24 h incubation provide an approximate indication of relative P. aeruginosa cell numbers. Cultures dominated by E. coli remained yellow-beige, while those with higher numbers of P. aeruginosa became blue-green due to production of pyocyanin

Fig. 2

Concentrations of pyocyanin and organic acids in culture supernatants after 24 h growth. E. coli BW25113 and P. aeruginosa PA14 were grown in LB either alone or in co-culture, and with or without 4% glucose. Bars indicate mean values (n = 3) and points indicate individual data. Red points indicate samples in which no pyocyanin or acid was detected. Concentrations of acids below 5 µg.mL⁻¹ were recorded as trace amounts

In co-cultures grown in LB + 4% glucose, E. coli grew to a higher density and inhibited the growth of P. aeruginosa nearly 40-fold (Fig. 1), from 2.66 × 10¹⁰ CFU/mL when grown alone to 7.11 × 10⁸ CFU/mL when grown together (mean values; n = 3). Pyocyanin production was also no longer visible. The inclusion of glucose led to a reduction of E. coli cell density at 24 h (Fig. 1) in both pure and co-cultures. This is likely to be due to acidification of the culture medium as the abundant glucose was fermented to short-chain organic acids (El-Mansi 2004). All conditions without glucose had a final supernatant pH of 7.5. With 4% glucose, pure P. aeruginosa culture supernatant pH was 7.0, indicating a small amount of acid production, while E. coli with 4% glucose, either alone or in competition, decreased the supernatant pH to 4.5. This drop in pH was caused primarily by production of acetic acid (Fig. 2). Small quantities of formic acid were produced in all cultures, while butyric acid, lactic acid and propionic acid were also detected in some cultures.

Disruptions to respiration provide a competitive advantage for E. coli against P. aeruginosa, but only in the presence of glucose

The TraDIS-Xpress screen identified 57 loci in which insertional inactivation led to significantly changed abundances (− 1 > log₂FC > 1 and q < 0.01) of E. coli mutants in co-cultures with P. aeruginosa compared to the E. coli mutants grown alone (Fig. 3). Both with and without glucose, the gene with the largest log₂FC and most significant q-values was tolC (Tables 2 and 3), which encodes an outer membrane channel used by several export systems. Insertional inactivation of tolC was deleterious to survival irrespective of glucose addition, which suggests that a competitive factor of P. aeruginosa inhibits E. coli but is at least partly counteracted by a TolC-dependent export process. The most probable identity for this factor is pyocyanin as we established that high levels of pyocyanin were produced when P. aeruginosa dominated the co-cultures. Mutants of E. coli lacking TolC exhibit increased sensitivity to pyocyanin and decreased metabolic output in glucose minimal medium so would be expected to be less competitive against P. aeruginosa (Sulavik et al. 2001; Dhamdhere and Zgurskaya 2010).

Fig. 3

Overview of significant loci identified for E. coli BW25113 grown with 4% glucose, and during competition with P. aeruginosa PA14 with and without 4% glucose. A Venn diagram containing names of all significant genes (− 1 > log₂FC > 1; q < 0.01) under each condition. Red gene names indicate those for which transposon insertions within the coding region reduced competitive fitness (negative log₂FC). Underlined genes are involved in central carbon metabolism. Three unnamed short repeat regions were identified in the no glucose condition, which are represented by the range of their base pair position in the genome. B Number of significant genetic loci for each of the important gene functions identified for each condition

Table 2 Significant genetic loci for competition of E. coli with P. aeruginosa in the absence of glucose

Table 3 Significant genetic loci for competition of E. coli with P. aeruginosa in the presence of 4% glucose

In co-cultures without glucose, 14 loci with a range of functions (Fig. 3 B) were important for E. coli competition with P. aeruginosa (Table 2, Fig. 3B). Three of these loci led to reduced fitness when they were disrupted (negative log₂FC), whilst disruption to 11 loci increased fitness. This included 3 repetitive extragenic palindromic elements, which are referred to here by their base pair coordinates in the reference genome (Fig. 3A). The function of such elements is not understood fully, but they can regulate gene expression when they occur within 15 bp downstream of a stop codon (Liang et al. 2015). Only the region between bp 736409 – 736515 met this criterion out of those identified in this study. It is situated 15 bp downstream of dtpD on the reverse strand, which encodes a proton-dependent dipeptide symporter and was not itself identified as significant. However, the small size (approximately 100 bp) of these palindromic elements could lead to increased experimental variation as they contain fewer unique transposon insertions than typical coding regions.

For the co-culture of E. coli and P. aeruginosa with 4% glucose, TraDIS-Xpress identified 46 genes where disruptive transposon insertions led to a significant fold change in abundance, although only 3 genes were shared with the LB co-culture without glucose (Fig. 3). Of these 46 genes, a competitive disadvantage was conferred by 8, while mutations in the other 38 genes were advantageous. Notably, the most frequently identified gene function under either growth condition was central carbon metabolism (CCM). However, while only 5 CCM genes (36%) were significant without glucose, 31 CCM genes (67%) were significant with glucose. The much larger number of significant genes when glucose was included demonstrates a clear competitive advantage for these mutants in the presence of abundant glucose.

From all the genes identified as significant in any condition (Fig. 3A), 34 were involved in aerobic CCM, and insertional inactivation conferred a selective advantage in all but arcB (E. coli monoculture without glucose, log₂FC =  − 1.01; q = 8.98 × 10^–49), which encodes the kinase of the ArcAB two-component system that, under anaerobic conditions, represses transcription of genes coding for aerobic metabolism (Iuchi et al. 1989). The fact that arcB mutants are selected against highlights the importance of the ability to suppress aerobic respiration and so promote overflow metabolism to facilitate competition in these conditions.

In co-cultures with glucose, mutants of genes encoding TCA cycle and oxidative phosphorylation functions were over-represented, with 29 significant genes including citrate synthase (gltA), aconitate hydratase 2/2-methylisocitrate dehydratase (acnB), 2-oxoglutarate dehydrogenase (sucAB), succinyl-CoA synthetase (sucCD), succinate:quinone oxidoreductase (sdhA-E) and malate dehydrogenase (mdh) from the TCA cycle; aspA and aspC, which code for functions that derive energy from L-glutamate by generating TCA cycle intermediates; and NADH:quinone oxidoreductase (nuoBC,F–H,J,L-M) and proton translocating F₀F₁ ATP synthase (atpA-H) from oxidative phosphorylation (Fig. 4). Disruptions to all these genes are likely to increase metabolic flux towards the fermentative mixed organic acid pathways.

Fig. 4

Genes coding for enzymes of the tricarboxylic acid cycle, NADH:quinone oxidoreductase and proton translocating F_OF₁ ATP synthase, which are all involved in respiratory metabolism. Insertional inactivation of the 31 genes shown in bold, green typeface conferred a selective advantage in the presence of P. aeruginosa PA14 and 4% glucose. Insertional inactivation of the 9 underlined genes also conferred a selective advantage when E. coli was grown alone with 4% glucose

The other significant loci were divided between multiple functions, although no other functions had more than 5 significant loci for either condition and there were no obvious links between those functions and competition with P. aeruginosa. We did not identify any regulatory genes other than global gene/protein regulators such as lon and fis. This suggests that no specific regulatory or sensing mechanisms are important for the production of organic acids in response to P. aeruginosa. We also did not identify genes involved in production or resistance to other potentially inhibitory chemicals such as indole.

Discussion

Taken together, our results both support and extend the findings of Christofi et al. (2019). However, while Christofi et al. grew each mutant of interest individually in competition with P. aeruginosa with and without glucose, our TraDIS-Xpress assay included an ensemble of all mutants in the entire library, thus competing each against the pool and against P. aeruginosa. Our approach more closely resembles the likely conditions encountered in nature by a spontaneous mutant arising within a population of bacteria, in which the mutant must compete not only with clonal siblings but also with its’ parent strain, other mutants and, probably, other species. The probability of any mutant multiplying to form a significant sub-population is proportional to its relative fitness in those conditions; from a deleterious mutation that is likely to die out to a highly advantageous mutation that might quickly become the dominant strain. Therefore, TraDIS-Xpress co-culture assays could provide fresh insight into the evolution of overflow metabolism and many other processes.

Crucially, in our screen each mutant constitutes only a tiny minority of the population—probably too few for individual CCM mutants to have a discernible effect on the acid concentration of the culture, and their increased acid production (if it has any effect) should benefit all the E. coli mutants present by repressing growth of P. aeruginosa. Correspondingly, mutants that significantly decrease acid production also make up only a small proportion of the population. They are, however, able to benefit from the P. aeruginosa-inhibiting effect of acids produced by most of the E. coli and therefore might not experience a significant negative selective pressure. The overall environmental conditions are determined by the majority of cells, which do not have transposon insertions that significantly affect acid production or survival. Under our experimental design, these majority cells may have contributed a proportion of their acid production through anaerobic respiration or fermentation in the absence of oxygen, rather than overflow metabolism sensu stricto, at higher cell densities. However, due to their specific transposon insertions, mutants that significantly increased in abundance possessed a selective advantage in these conditions. Therefore, while screening mutants in isolation (c.f. Christofi et al. 2019) identified genes involved in mixed acid fermentation, our TraDIS-Xpress screen identified mutants with altered fitness for the complex mixture of stresses imposed by the acidic environment and growth within a heterogeneous population.

Transposon insertions in 12 E. coli genetic loci conferred a selective advantage during monoculture growth in 4% glucose. 9 of these genes code for either the NADH oxidoreductase or ATP synthase complexes, in which the mutants are expected to be deficient in aerobic and anaerobic respiration. This suggests that fermentation under these conditions confers a selective advantage over respiring E. coli cells. Since an abundant supply of glucose, as in our experiment, will promote overflow metabolism, it may be the case that the CCM mutants identified here shift the metabolic balance further towards fermentation (even completely preventing respiration) and thus confer a slight survival advantage over cells that retain wild-type metabolism due to the limitations of cellular physiology mentioned in the introduction (de Groot et al. 2020). Therefore, while bulk production of organic acids benefits the E. coli population as a whole, mutants that are genetically ‘primed’ for overflow metabolism gain an additional individual selective advantage. When 4% glucose and P. aeruginosa were present, as well as additional genes coding for NADH oxidoreductase or ATP synthase, many mutations in genes coding for TCA cycle functions were identified as advantageous. Therefore, although P. aeruginosa growth was supressed, it did exert a competitive stress upon the E. coli mutants, potentiating the selective pressure for CCM mutants that favour fermentation. Since the same pathways were identified with and without P. aeruginosa, the benefits of overflow metabolism are not simply due to the accumulation of organic acids in the medium.

The overall small log₂FC values (Tables 2 and 3) indicate that the fitness advantages experienced by these mutants were slight. Nevertheless, the data were highly reproducible between duplicates and indicated extremely high levels of significance for many of the genes. TraDIS-Xpress results can comprise data from tens to hundreds of transposon inserts per gene, all of which are assayed simultaneously. This provides a degree of internal validation and increases the resolution of the assay compared to those approaches using single gene knockouts. Small, but statistically significant,phenotypic changes, such as we observed here, may not be observable by other, less sensitive, assay methods. In this experiment, we observed that mutations in genes coding for aerobic respiration possess a selective advantage under the test conditions. We hypothesise, therefore, that inactivating mutations in the suppressor of aerobic respiration functions, arcB, would be detrimental under these test conditions. The data indeed show that arcB mutants do have a significant disadvantage. Thus, the TraDIS-Xpress data are very robust and often include their own validating data, precluding the need for additional testing of defined mutants.

Mixed acid fermentation is a common trait within E. coli and a wide variety of other bacteria, so it is likely that, under appropriate conditions, many members of the mammalian gut microbiome will be able to outcompete P. aeruginosa and other species in a similar manner. However, due to the wide range of habitats occupied by P. aeruginosa, adaptation and evolution for specific environmental conditions have been observed. For example, total cell growth varies for P. aeruginosa strains isolated from different locations, even when grown under the same in vitro conditions (Diaz et al. 2018). Therefore, P. aeruginosa strain source and history might affect their susceptibility to lactic and acetic acids. Acetic acid has been suggested previously as an antibiotic-free decontamination method for P. aeruginosa biofilms in hospital drains (Stjärne Aspelund et al. 2016) and for burn wounds (Halstead et al. 2015). This raises the tempting prospect of using probiotic bacteria to modulate the microbiome for therapeutic purposes by production of organic acids, which may be more resilient than antibiotics against the development of resistance.

Conclusions

Our results enrich our understanding of overflow metabolism in E. coli, its impact on individuals in a population, and its role in competition with P. aeruginosa. We identified CCM genes that contribute towards the balance between respiration and fermentation during overflow metabolism and highlighted the importance of tolC for survival in the presence of pyocyanin produced by P. aeruginosa. The TraDIS-Xpress data adds a further quantitative dimension to the importance of genes and sensitivity in their detection. The density of transposon insertions across the genome for all the mutants combined means that it is likely that 99% of the E. coli genome was screened. Therefore, it is unlikely that any other gene functions are involved under the experimental conditions employed.