Supplementary Information for:

Fitness and stability of obligate cross-feeding interactions that emerge upon gene loss in bacteria

Samay Pande1, Holger Merker1, Katrin Bohl1,2,3, Michael Reichelt4, Stefan Schuster2, Luís F. de Figueiredo2,5, Christoph Kaleta3, Christian Kost1,6

1 Experimental Ecology and Evolution Research Group, Department of Bioorganic Chemistry, Max Planck Institute for Chemical Ecology, Jena, Germany, 2 Department of Bioinformatics, Friedrich Schiller University Jena, Jena, Germany, 3 Research Group Theoretical Systems Biology, Friedrich Schiller University Jena, Jena, Germany, 4 Department of Biochemistry, Max Planck Institute for Chemical Ecology, Jena, Germany, 5 Cheminformatics and Metabolism, European Bioinformatics Institute (EBI), Welcome Trust Genome Campus, Hinxton, Cambridge, United Kingdom, 6 Institute of Microbiology, Friedrich Schiller University Jena, Germany. Correspondence and requests for materials should be addressed to C. Kost (email: )

Supplementary Figure S1. Amino acid production of wild type (WT) and all single and double deletion mutants as quantified by using auxotrophs as biosensors. All genotypes were cocultured together with each of four E. coli auxotrophs (1:1) and the productivity of auxotrophs was determined as the number of CFUs x 107 after 24 h minus the initial density. Combinations with matching amino acid auxotrophies were excluded. (a) Amino acid production given as the mean productivity (±95%CI) of auxotrophs. (b) Expected versus observed amino acid production of cross-feeding mutants. The expected production levels are the sum of the measurements of the two corresponding single gene deletion mutants (i.e. auxotroph and overproducer) that were combined in one genetic background (i.e. cross-feeder). Asterisks indicate a significant difference (***paired t-test, P=9x10-14, n=8).

Supplementary Figure S2. Fitness of wild type and two-membered consortia. The table (left) summarizes the genetic backgrounds of the two cocultured genotypes: wild type (WT), auxotrophs (Aux), overproducers (Ov), cross-feeding consortia (CF), and cross-feeding consortia (CF*) that included DnuoNDleuB. ‘Mutation 1’ specifies whether or not a strain is auxotroph for either arginine (DargH), tryptophan (DtrpB), leucine (DleuB), or histidine (DhisD). ‘Mutation 2’ indicates whether a strain carries one of three deletion mutations causing amino acid overproduction (i.e. DnuoN, Dmdh, Dppc). Four genotypes that cover a representative spectrum of consortium-level fitness values are marked in bold and were used for subsequent experiments. Shown is the median (±range) fitness (i.e. Malthusian parameter) of WT and all two-membered consortia. The region delimited by dashed lines marks the range of WT fitness. All fitness values were significantly different from WT levels (FDR-corrected two-sample t-test, P<0.05, n=8), except the one marked with ‘ns’.

Supplementary Figure S3. Competitive fitness of the four amino acid auxotrophic mutants relative to WT. Single-gene deletion mutants were competed against WT for 24 h in minimal medium, to which the one amino acid has been added (100 mM) the auxotrophs needed for growth. The dashed line indicates equality in fitness between WT and the corresponding competitors. Asterisks indicate fitness values that were significantly different from 1 (i.e. WT fitness, one sample t-test, ***P<0.001, n=10).

Supplementary Table 1. Meta-analysis and phenotypic predictions of 32 different Escherichia coli strains to overproduce amino acids and/ or to be auxotroph for certain amino acids.

Predicted phenotype / Total auxo-trophies
AA over- / AA auxotrophyc
Strain IDa / productionb / ile / pro / trp / lys / his / thr / met / leu / arg / tyr
FRIK 920 / / / / / / / / / / / / 6
FRIK2001 / / / / / / / / / / / / 4
LRH13 / / / / / / / / / / / / 3
EC20000964 / / / / / / / / / / / / 3
F1095 / / / / / / / / / / / / 2
63154 / / / / / / / / / / / 1
FRIK1990 / / / / / / / / / / / / 5
FRIK1999 / / / / / / / / / / / / 4
Zap0046 / / / / / / / / / / / / 3
EC20000948 / / / / / / / / / / / 3
EC20000703 / / / / / / / / / / / / 2
K12MG1655 / / / / / / / / / / / / 1
71074 / / / / / / / / / / / / 1
EC20030338 / / / / / / / / / / / / 1
EC20011339 / / / / / / / / / / / / 1
EC20000958 / / / / / / / / / / / / 1
E2328 / / / / / / / / / / / / 1
F1299 / / / / / / / / / / / / 1
EC2000623 / / / / / / / / / / / / 1
EC970520 / / / / / / / / / / / / 1
59243 / / / / / / / / / / / / 0
LRH6 / / / / / / / / / / / / 0
H4420 / / / / / / / / / / / / 0
58212 / / / / / / / / / / / / 0
97701 / / / / / / / / / / / / 0
ECI-634 / / / / / / / / / / / / 0
EDL933 / / / / / / / / / / / / 0
Sakai / / / / / / / / / / / / 0
F5 / / / / / / / / / / / / 0
R1797 / / / / / / / / / / / / 0
EC20020119 / / / / / / / / / / / / 0
FRIK1985 / / / / / / / / / / / / 0
Total / 6 / 12 / 7 / 7 / 5 / 3 / 3 / 3 / 2 / 2 / 1
(%) / 19 / 37 / 22 / 22 / 16 / 9 / 9 / 9 / 6 / 6 / 3

a Microarray-based comparative genomic hybridization; data from Zhang et al. (2007); b Absence ( )/ presence ( ) of the ppc gene; c Strains predicted to be auxotroph ( ) or prototroph ( ) for certain amino acids. Predictions based on the presence/ absence of genes encoding key steps (Kanehisa & Goto 2000, Karp et al 2002) in the biosynthesis of amino acids as well as published in silico predictions (Tepper & Shlomi 2011) and experimental data (Bertels et al 2012, Joyce et al 2006, Orth et al 2011)(in the latter: essential biosynthetic genes (i.e. no growth in minimal medium) were declared as giving rise to an auxotrophy when lost).

Supplementary Materials and Methods

CASOP-GS

In order to identify target genes that can be knocked out to increase amino acid production we used CASOP-GS. CASOP-GS (Bohl 2010) is based on the CASOP method (Hadicke & Klamt 2010), which was extended for application to genome-scale metabolic networks. The method provides a ranking of the reactions in a metabolic network with respect to how much they contribute to the synthesis of a certain product of interest and is based on the concept of elementary flux modes (Schuster et al 2000). Since the entire set of elementary flux modes cannot be enumerated for most genome-scale metabolic networks (Klamt & Stelling 2002), CASOP-GS uses a linear programming-based sampling procedure that computes a subset of all elementary flux modes to obtain scores for reactions used in CASOP. Another extension of CASOP-GS over CASOP is that a measure to assess the approximate increase of production of a particular metabolite of interest for a given set of gene deletions is provided. This measure thus allows one to rank gene deletions according to their increase in the production of the target metabolite.

As metabolic network we used the most recent reconstruction of E. coli metabolism, iAF1260 (Feist et al 2007) with external conditions corresponding to the cultivation media. For taking into account regulatory interactions, we used the Boolean network provided by Gianchandani et al. (Gianchandani et al 2009) For each production scenario we sampled two sets of elementary flux modes: one that accounted for regulatory interactions and one that did not. For each case, one million elementary flux modes were sampled. Regulatory interactions were taken into account by testing for each elementary flux mode whether any of the Boolean rules of the regulatory network were violated for the given external conditions. Such a violation could be, for instance, that a reaction is used by an elementary flux mode that is not active under the given environmental condition (i.e., it is set to off by the Boolean regulatory network).

In order to identify potential deletion targets for amino acid overproduction we used CASOP-GS as described before (Bohl 2010) (with γ=0.0 for the wild-type, γ=0.9 for the production scenario and the weighting factor k=5). For each of the four amino acids considered in this study (Arg, Trp, Leu, and His), all 1,260 genes of iAF1260 were ranked according to how much their deletion would increase the production of the corresponding amino acid (with and without considering regulation). Thus, we obtained eight lists of genes with a total of about 80 knockout target predictions. From these lists, the ten top ranking genes (excluding genes participating in the same enzyme complex) were chosen for each scenario. Excluding genes with rather poor growth characteristics as documented in the Keio collection (Baba et al 2006) led to a list of 54 candidate deletion targets for amino acid overproduction. After inserting these 54 deletion alleles into a common genetic background (E. coli BW25113), their production characteristics with respect to the wild type were characterized by measuring amino acid concentrations in the supernatants as well as determining cell count after 24 hours of growth. Subsequently, all mutants were ranked according to total amino acid produced per cell and three of the top ranking mutants (DnuoN, Dmdh, Dppc) were chosen for further examination.

Amino acid quantification

5 µl from a preculture diluted to OD600nm 0.1 were inoculated into 1 ml MMAB medium. To quantify the amount of amino acids the overproducing strains released into the medium, 24 h-old cultures were sterile-filtered (0.2 µm) and amino acids in the supernatant analysed by LC/MS/MS. 100µl of the culture was sampled before this step in order to estimate the numbers of colony-forming units (CFUs) by plating on LB agar plates. These numbers were used to normalize the total amino acid production of the culture in 1 ml of medium per individual CFU.

Initially, we intended to analyse the non-derivatised amino acids in the culture medium directly following a protocol modified from Jander et al (2004). However, it turned out that the high concentrations of the media components led to a strong quenching effect in the ionisation of the mass spectrometer due to coelution of a number of amino acids with the media components.

Therefore, we first derivatised the amino acids with 9-fluorenylmethoxy-carbonyl chloride (FMOC-Cl) (Fluka, Germany) in order to convert them into less polar derivatives. 100 µl of the sterile-filtered medium was mixed with 100 µl of borate buffer (0.8 M, pH 8.0), spiked with internal standards amino acid mix (13C-, 15N-labelled amino acids (algal amino acids 13C, 15N, Isotec, Miamisburg, US) at a concentration of 20 µg of the algal amino acid mix per mL of borate buffer). 200 µl of FMOC-Cl reagent (30 mM in acetonitrile) was added to the samples and mixed. After 5 minutes, 800 µl of Hexane was added to stop the reaction and to remove excess FMOC-Cl reagent, mixed and let stand for the separation of liquid phases. 200 µl of the aqueous liquid phase was then transferred to a fresh 96 deep-well plate for chromatographic analysis.

Chromatography was performed on an Agilent 1200 HPLC system (Agilent Technologies, Böblingen, Germany). 10 µl of derivatised sample was injected and separation was achieved on a Zorbax Eclipse XDB-C18 column (50x4.6 mm, 1.8 µm, Agilent Technologies, Germany). Formic acid (0.05%) in water and acetonitrile were employed as mobile phases A and B respectively. The elution profile was: 0-1 min, 90%A; 1-4.5 min, 10-90%B in A; 4.51-5 min 100% B and 5.1-8 min 90% A. The mobile phase flow rate was 1.1 ml min-1. The column temperature was maintained at 20°C. The liquid chromatography was coupled to an API 3200 tandem mass spectrometer (Applied Biosystems, Darmstadt, Germany) equipped with a turbospray ion source operated in negative ionization mode. The instrument parameters were optimized by infusion experiments with pure FMOC-derivatized standards (amino acid standard mix, Fluka, St. Louis, USA). The ionspray voltage was maintained at -4.5 keV. The turbo gas temperature was set at 700°C. Nebulising gas was set at 70 psi, curtain gas at 35 psi, heating gas at 70 psi and collision gas at 2 psi. Multiple reaction monitoring (MRM) was used to monitor analyte precursor ion → product ion (see Supplementary Table 2).

Both Q1 and Q3 quadrupoles were maintained at unit resolution. Analyst 1.5 software (Applied Biosystems, Darmstadt, Germany) was used for data acquisition and processing. Linearity in ionization efficiencies was verified by analyzing dilution series of FMOC-derivatized standard mixtures (amino acid standard mix, Fluka plus Gln, Asn and Trp, also Fluka). All samples were spiked with 13C-, 15N-labelled amino acids (see above). The concentration of the individual amino acids in the 13C-, 15N-labelled amino acids mix had been determined by classical HPLC-fluorescence detection analysis after pre-column derivatisation with ortho-phthaldialdehyde-mercaptoethanol using external standard curves made from standard mixtures (amino acid standard mix, Fluka plus Gln, Asn and Trp, also Fluka). Individual amino acids in the sample were quantified by the respective 13C, 15N-labelled amino acid internal standard, except for tryptophan, which was quantified using 13C, 15N-Phe applying a response factor of 0.42.