How Pseudomonas aeruginosa adapts to various environments: a metabolomic approach

Authors

  • Eliane Frimmersdorf,

    1. Department of Biochemistry and Bioinformatics, Institute for Biochemistry & Biotechnology, Technische Universitaet Braunschweig, 38106 Braunschweig, Germany.
    Search for more papers by this author
    • E.F. and S.H. contributed equally to this work and should be considered as joint first authors.

  • Sonja Horatzek,

    1. Klinische Forschergruppe, OE 6710, Medizinische Hochschule Hannover, Carl-Neuberg-Str. 1, 30625 Hannover, Germany.
    Search for more papers by this author
    • E.F. and S.H. contributed equally to this work and should be considered as joint first authors.

  • Anya Pelnikevich,

    1. Klinische Forschergruppe, OE 6710, Medizinische Hochschule Hannover, Carl-Neuberg-Str. 1, 30625 Hannover, Germany.
    Search for more papers by this author
  • Lutz Wiehlmann,

    1. Klinische Forschergruppe, OE 6710, Medizinische Hochschule Hannover, Carl-Neuberg-Str. 1, 30625 Hannover, Germany.
    Search for more papers by this author
    • D.S. and L.W. contributed equally to this work and should be considered as joint last authors.

  • Dietmar Schomburg

    Corresponding author
    1. Department of Biochemistry and Bioinformatics, Institute for Biochemistry & Biotechnology, Technische Universitaet Braunschweig, 38106 Braunschweig, Germany.
    Search for more papers by this author
    • D.S. and L.W. contributed equally to this work and should be considered as joint last authors.


E-mail d.schomburg@tu-bs.de; Tel. (+49) 531 391 8301; Fax (+49) 531 391 8302.

Summary

In addition to transcriptome and proteome studies, metabolome analysis represents a third complementary approach to identify metabolic pathways and adaptation processes. In order to elucidate basic principles of metabolic versatility of Pseudomonas aeruginosa, we investigated the metabolome profiles of two genetically and morphologically divergent strains, the reference strain PAO1 and the mucoid clinical isolate TBCF10839 in exponential growth and stationary phase in six different carbon sources (cadaverine, casamino acids, citrate, glucose, succinate and tryptone). Both strains exhibited strong similarities in mode of growth; the metabolite patterns were mainly defined by the growth condition. Besides this adaptive response, a basic core metabolism shapes the P. aeruginosa metabolome, independent of growth phase, carbon source and genetic background. This core metabolism includes pathways related to the central energy and amino acid metabolism. These consistently utilized metabolic pathways are closely related to glutamate which represents a dominant metabolite in all conditions analysed. In nutrient-depleted media of stationary phase cultures, P. aeruginosa maintains a specific repertoire of metabolic pathways that are related to the carbon source formerly available. This specified adaptation strategy combined with the invariant basic core metabolism may represent a fundamental requirement for the metabolic versatility of this organism.

Introduction

The Gram-negative ubiquitous microorganism Pseudomonas aeruginosa exhibits extreme metabolic versatility (Ramos, 2004). Pseudomonas aeruginosa colonizes diverse environmental habitats and is able to persist and grow even under nutrient-poor and hostile conditions (Schmidt et al., 1996; OECD, 1997). In addition to environmental habitats, P. aeruginosa strains are frequently isolated from clinical sources. As one of the most relevant causative agents of nosocomial infections (Campa et al., 1993) this opportunistic human pathogen establishes acute or chronic infections in hospitalized and immunodeficient patients (Quinn, 2003; Trautmann et al., 2005). Lifestyles of this organism comprise both planktonic growth and alginate-associated biofilm formation (Davies et al., 1993). Successful adaptation to multiple niches requires extraordinary physiological capabilities and reflects the broad metabolic capacity of this organism, which, particularly in low-nutrient conditions, provides distinct advantages for outgrowing competitors (Kiewitz and Tümmler, 2000).

Based on the P. aeruginosa PAO1 genome sequence, 8% of the 6.3 Mbp genome is predicted to encode transcriptional regulators and two-component regulatory systems (Anba et al., 1990; Stover et al., 2000; Nishijyo et al., 2001; Krieger et al., 2002). Along with a large number of genes for substrate uptake and catalysis, which enable the bacterium to metabolize various compounds as nutrients, high environmental adaptability is granted (Stanier et al., 1966).

Energy generation is mainly based on oxidative substrate catabolism; however, the bacterium is also able to flourish in a facultative anaerobic environment with alternative external electron acceptors or to survive by fermentation of arginine or pyruvate (Wauven et al., 1984; Davies et al., 1989; Eschbach et al., 2004; Filiatrault et al., 2006; Williams et al., 2007). If a preferred nutrient is offered in a mixture along with unfavoured carbon sources, the carbon metabolism of P. aeruginosa is regulated by catabolite repression control, which facilitates catabolism of substrates in an ordered fashion (Suh et al., 2002). Preferentially chosen sources of carbon or nitrogeninclude short-chain fatty acids, amino acids and polyamines (Stanier et al., 1966; Lessie and Neidhardt, 1967; Meile et al., 1982; Tricot et al., 1994). Although efficiently metabolized, sugars represent less preferred substrates for P. aeruginosa, which are degraded via the Entner-Doudoroff pathway (Entner and Doudoroff, 1952; Goldbourt et al., 2007). Growth on n-alkanes and (halogenated) aromatic compounds reflects the metabolic potential of P. aeruginosa to degrade complex xenobiotics (Palleroni, 1986; Hickey and Focht, 1990; Marín et al., 2003). In addition to its remarkable capacity of substrate utilization, P. aeruginosa is proficient in extensive secondary metabolite production (Leisinger and Margraff, 1979; Gallagher and Manoil, 2001).

Metabolic competence of P. aeruginosa has been previously intensively investigated in many transcriptome studies (Palma et al., 2003; Salunkhe et al., 2005; Weir et al., 2008). Furthermore, using nuclear magnetic resonance technology, a metabolomics approach was employed to study the overall metabolic profiles of P. aeruginosa biofilm and planktonic cells (Gjersing et al., 2007).

Our study aims to investigate the global regulatory network of the P. aeruginosa metabolism with respect to the metabolic proficiency of this organism to cope with diverse habitats. We used a gas chromatography-mass spectrometry (GC-MS)-based approach to assess both intra- and extracellular metabolite spectra (Strelkov et al., 2004). Metabolome data were further statistically analysed with pattern recognition tools for the evaluation of metabolic pathway utilization. The resulting global view on the metabolome uncovers new aspects of adaptation processes to environmental conditions and reveals the nature of metabolic versatility in P. aeruginosa.

Results and discussion

Selection of growth conditions

We included two genetically and morphologically divergent strains of P. aeruginosa, the reference strain PAO1 and the mucoid clinical isolate TBCF10839, in our analyses to investigate strain-specific metabolic features, as well as general aspects of P. aeruginosa metabolism. In order to obtain extensive information on metabolic adaptation processes in P. aeruginosa, diverse growth media were selected to simulate distinct lifestyles and to provoke induction of dissimilar metabolic pathways.

Environmental conditions with minimal nutrient supply were imitated by media containing succinate, citrate, glucose or cadaverine as the single carbon source. Out of these media, succinate and citrate are well-established carbon sources for bacteria in in vitro cultures (Miller and Ku, 1978; Kirner et al., 1996) and known to be preferentially metabolized by P. aeruginosa (Ng and Dawes, 1973). Glucose is also widely used as a growth substrate under laboratory conditions, and sugar derivatives are common in the environment; however, glucose represents a less preferred carbon source for P. aeruginosa. The catabolism via the Entner-Doudoroff pathway (Entner and Doudoroff, 1952; Goldbourt et al., 2007) is less efficient than the glycolysis and is under control of catabolite repression, a global regulatory mechanism allowing hierarchical utilization of carbon sources (Wolff et al., 1991; Nishijyo et al., 2001; Silo-Suh et al., 2005). To mimic bacterial lifestyle under nutrient-poor conditions, the primary amine cadaverine was offered as substrate. As the decarboxylated degradation product of lysine, cadaverine commonly occurs in the environment but allows for only slow and poor growth.

In contrast to single carbon source media, full media mimicked conditions where P. aeruginosa inhabits nutrient-rich niches, such as animate hosts or sewage (Mena and Gerba, 2009). In order to analyse adaptation processes to nutrient-rich habitats, which support rapid bacterial growth, these environments were imitated by tryptone and casamino acids media. Both tryptone and casamino acids are derived from casein and have similar amino acid compositions but different complexities. Tryptone is composed of small oligopeptides and requires further hydrolytic degradation prior to metabolization, whereas casamino acids consist of single amino acids.

Specific growth rates of P. aeruginosa

In Fig. 1 the growth curves of three of the six different cultivation conditions are shown. The specific growth rates µ of all conditions were calculated from the exponential growth phase. Specific growth rates in tryptone media were significantly higher [µ = 0.85 (PAO1), µ = 0.98 (TBCF10839)] than in other analysed media. Specific growth rates are comparable in the minimal media succinate, glucose and citrate, whereas in minimal medium with cadaverine specific growth rates for both strains decrease to only 0.17 for PAO1 and 0.15 for TBCF10839. Metabolome analysis was performed during mid-exponential growth and in the stationary growth phase in batch cultures. Samples from mid-exponential growth were taken at OD600 = 1 (exception: cadaverine OD600 = 0.3). The second set of samples was taken 2 h after bacterial cultures had entered stationary growth phase, determined by optical density measurement (Fig. 1).

Figure 1.

Growth of P. aeruginosa and metabolization of the growth media.
A. Growth curves of P. aeruginosa PAO1 on succinate, casamino acids and cadaverine. Growth curves on citrate, tryptone and glucose (data not shown) were comparable to the succinate growth curve. The growth curves of strain TBCF10839 (data not shown) were comparable to the growth curve of PAO1 in the respective medium. In all media, except for cadaverine, both P. aeruginosa strains grew exponentially to optical densities > 2. Sampling points for metabolome analysis are indicated by arrows.
B. Growth rates of P. aeruginosa strains PAO1 and TBCF10839. Both strains exhibited similar growth rates during exponential growth in the same medium. In cadaverine-containing medium, the growth rate was significantly lower (P < 0.001) than in all other analysed media.
C. Metabolization of the growth medium. Supernatant compositions of three different media (succinate, casamino acids and tryptone) incubated with P. aeruginosa PAO1 were analysed as representatives, at six time points [start [0], early exponential phase (0.3 OD [e.e.]), exponential phase (1 OD [e.]), late exponential phase (2.0 OD [l.e.]), early stationary phase (3.5 OD [e.s.]), stationary phase (2 h later [s])] to monitor substrate consumption in different growth phases. Carbon compounds that represented less than 2% of the initially nutrients available were grouped as ‘others’. In complex growth media, P. aeruginosa metabolized nutrients in a specific order. In stationary phase, the succinate- and casamino acid-containing media were depleted of carbon sources, whereas tryptone still contained initial substrates.

Metabolite analysis of P. aeruginosa

Following GC-MS-based analysis of the P. aeruginosa metabolome, 243 compounds were found according to their retention times and mass spectra. Out of those, 144 could be matched to known compounds by comparison with our in-house metabolite libraries. Sixty metabolites were found in all analysed conditions, and 64 metabolites were present in most metabolomes, whereas 46 metabolites were only present in a few growth states and media. Nineteen metabolites were exclusively found in single growth conditions, mostly in both P. aeruginosa strains.

The majority of identified substances (more than 50%) were represented by metabolites from amino acid and sugar metabolism (Table S1). The number of major metabolites was found to depend on the media composition, irrespective of growth state. In cadaverine-grown cells, 90% of the total signal intensity of GC analyses was generated by only 22 metabolites with high peak areas (It has to be mentioned here that the relationship between metabolite concentrations and peak areas is compound-specific, i.e. a compound A with a peak area twice the area of compound B is not necessarily present in a higher concentration. On the other hand, a specific metabolite showing a larger peak area in growth condition B compared with A is present in a higher concentration. During the discussion sometimes concentration instead of peak area is used for easier readability). In contrast, in succinate or tryptone media more than 40 metabolites accounted for 90% of the overall signal, indicating a greater diversity of metabolic compounds. In all analysed conditions, a few metabolites prevailed over the overall metabolite spectrum, with glutamate as the most dominant compound (Fig. 2).

Figure 2.

Box-plot of the concentrations of 49 identified metabolites that were found in all analysed conditions. Glutamate was identified as the compound with the highest mean concentration and variance.

According to the Kruskal–Wallis test (analysing significance and reproducibility) for each of the 60 constitutive metabolites at least one or more significant differences between the different media was found (median Chi-square: 105.7, sigma = 9.9; confidence level 0.001: ± 4.8; median P-value: P = 5 × 10−13; Fig. 3, isoleucine as example). Additionally, for each metabolite, concentration alterations between all analysed conditions were tested pairwise in Mann–Whitney U-tests. Due to the high technical reproducibility of data and the low biological variance, a large number of significant differences were observed (median 64%, P < 0.001 and P < 0.02 after correction for multiple testing). In other words, the variation of measured data points for each metabolite within one experimental condition were significantly lower compared with results of other conditions.

Figure 3.

Box-plot of measured metabolite concentrations for isoleucine (example). The concentration differences between most conditions were significant (P < 0.001; P < 0.02 after correction for multiple testing).

For a detailed comparison of metabolic states, differences in a metabolite concentration between two conditions were included in the analysis, if the change was statistically significant (P < 0.02, corrected Mann–Whitney U-test) and at least fourfold up or down. A table with fold-changes of all identified metabolites (normalized to the median signal intensity) was deposited as Table S1.

The fold-changes of metabolite concentrations in different media and growth phases within the same P. aeruginosa strain were found to be significantly higher than between different strains in the same medium and growth phase (P = 4.45 × 10−10). This shows that the metabolomes were mostly defined by the growth substrate rather than by the strain-specific genetic background (Fig. 4A). Supporting this hypothesis, under all tested conditions no significant differences between the median fold-changes of metabolite concentrations in strain PAO1 versus strain TBCF10839 were observed (P = 0.30) (Fig. 4B).

Figure 4.

Comparison of median fold-change of metabolites.
A. Fold-changes of metabolite concentrations between different growth phases within one strain (PAO1 or TBCF10839) were significantly higher (P = 4.5 × 10−10) than the differences between PAO1 and TBCF10839 in the same condition.
B. The median fold-changes of metabolite concentrations were not significantly different (P = 0.3) for strain PAO1 and TBCF10839 when different growth conditions were compared for the same strain.

Analysis of the growth media

In addition to comprehensive analyses of intracellular metabolite profiles, the compositions of growth media were analysed at five time points (Fig. 1) during cultivation in two rich media (tryptone and casamino acids) and in succinate as a representative for minimal media conditions.

In succinate-containing minimal medium during early exponential growth, P. aeruginosa had only the initial carbon source available. In contrast, at later stages of cultivation, virtually all succinate has been metabolized to cytosolic compounds or complex polymers (peptides, polysugars) that were not accessible for GC-MS analysis.

Casamino acids are a mixture of amino acids derived from fully hydrolysed casein. Growing in this medium, P. aeruginosa preferentially metabolized compounds that can be easily integrated in central metabolic pathways (Ala, Asp, Glu, Gly, Leu, Pro, Ser, Thr). While most of these compounds were completely digested until mid-exponential growth, other amino acids, which require a more complex catabolism, were metabolized at a later growth state (Ile, Leu, Met, Phe, Tyr, Val). Supernatants from the stationary phase were mostly deprived of the initial carbon source components.

Tryptone is a pancreatic digest of casein/milk powder. It consists mostly of amino acids and oligopeptides, but also contains sugars and lipids. In this complex medium P. aeruginosa metabolizes amino acids in the same sequential order as observed in pure casamino acids. Additionally, central carbohydrates (glucose, fructose, glycerol) were also metabolized during the early growth phases, whereas the concentration of other sugars (erythritol, inositol, maltose, mannitol, mannose, ribose, sucrose, trehalose) and fatty acids (dodecanoic acid, hexadecanoic acid, octadecanoic acid) remained nearly unchanged during the whole incubation period. In contrast to other analysed media, tryptone was not completely digested, when P. aeruginosa entered the stationary phase.

Principal component analysis

Principal component analysis (PCA) is an orthogonal linear transformation that transforms the data to a new coordinate system such that the greatest variance by any projection of the data comes to lie on the first coordinate, the second greatest variance on the second coordinate, etc. If the first two coordinates represent a large percentage of the whole data variation it allows a visual representation of clustering states.

When a PCA was performed with all identified metabolites and standard parameters, nearly 60% of the total variance was attributed to the first two principal components (data not shown). This high correlation was based on high concentrations of only a few components that dominate the overall variance of the metabolome (e.g. glutamate as the most dominant metabolite caused more than 80% of the total variance in the first two principal components).

In order to overcome the bias caused by the dominance of high-concentration compounds, logarithmically scaled metabolite concentrations were used for the PCA (Fig. 5). The contribution of several dozen metabolites to the first two components explained only 40% of the total variation. However, the examined growth conditions were clearly grouped by nutrient and growth phase, with both of the P. aeruginosa strains analysed plotted as nearest neighbours.

Figure 5.

PCA plot (principal component 1 versus principal component 2) of logarithmically transformed metabolite concentrations from P. aeruginosa strains PAO1 and TBCF10839 during exponential growth and in stationary phase growing on various carbon sources. The first two principal components account together for 40% of the variance. The variations of a few highly concentrated metabolites cause more than 80% of the observed effect. The first principal component mainly reflected differences in sugar metabolism, whereas the second represented the overall growth. The metabolomes of strains PAO1 and TBCF10839 from corresponding condition were always closely related.

Fifty per cent of the first and second principal components were defined by 18% and 23% of the identified metabolites respectively. These two sets of compounds are virtually non-overlapping.

The Eigenvalues of the first principal component (x-axis) were mainly determined by metabolites of the gluconeogenesis and the central sugar interconversion. In the second principal component (y-axis), metabolites related to growth (aminosugar, nucleotide and lipid metabolism) and degradation of complex amino acids (lysine, aromatic and branched-chain amino acids) were merged (Table 1). A plot of these two principal components represented the overall metabolic state for P. aeruginosa in different media: metabolomes reflecting optimal growth with high replication rates and a high sugar metabolism, as observed in tryptone, were located in the upper left corner. Cadaverine, as non-preferred carbon source offered in low concentration, was associated with a massive decrease of many metabolite concentrations, and led to an extreme location in one corner of the PCA plot. The other analysed growth conditions represented three one-carbon source minimal media, as well as CAA medium, which are all rich in substrate concentrations. The corresponding metabolomes from these experiments were located at comparable positions in relation to the y-axis; however, variations of their positions on the x-axis are indicative of the degree of sugar metabolism in the respective medium.

Table 1.  Metabolites with the highest Eigenvalues (determining > 50% of total variance of the first two principal components), grouped in metabolic pathways.
Principal component 1Principal component 2
Amino acid metabolismAmino acid metabolism
5-Aminopentanoic acid4-Aminobenzoate
AsparagineCadaverine
CitrullineCysteine
HomocysteineGlutamine
Homocysteine
Lipid metabolismLeucine
1-MonostearoylglycerolLysine
Oleic acid amideO-Phosphoserine
Ornithine
Nucleotide metabolismPhosphoethanolamine
XanthinePipecolinate
Putrescine
Sugar metabolismPyrophosphate
2′-Deoxyguanosine 5′-phosphatePyrrole-2-carboxylate
2-PhosphoglycerateSpermidine
ArabinoseTyrosine
Erythritol 
Glucose-1-phosphateLipid metabolism
Glycerone phosphateGlycerol-3-phosphate
Glyoxylic acid3-Phosphoglycerate
IsomaltoseTetradecanol
PhosphoenolpyruvateHexadecanol
PyrophosphateHexadecanoate
Ribulose-5-phosphateDodecanoate
Uridine-5′-monophosphate 
XyluloseNucleotide metabolism
Adenine
Adenosine-5′-monophosphate
Hypoxanthine
(Amino-) Sugar- and Nucleoside metabolism
1,6-Anhydro-beta-D-glucose
Glucono lactone
Glucosamine
Glucose
Mannitol
Mannose
N-Acetyl-glutamic acid
Ribose-5-phosphoric acid
Xylitol
Xylulose-5-phosphoric acid
Central metabolism
Citric acid
Diethylene glycol
Nicotinic acid

Hierarchical clustering analysis

In order to illustrate the basic principles of how P. aeruginosa metabolically adapts to a specific carbon source or growth phase, logarithmized mean values of metabolite concentrations were analysed in an unsupervized hierarchical cluster analysis (HCA) using Pearson correlation (as current metrics) as similarity measure (Fig. 6).

Figure 6.

Unsupervised hierarchical clustering of the logarithmically transformed into (log2) metabolite concentrations of all 243 found metabolites. The metabolites were divided in two groups: (i) metabolites that were constitutively present and (ii) metabolites that appeared in dependence of the growth conditions. Qualitative differences between both analysed P. aeruginosa strains were rare [Growth substrates: cadaverine (Cad); casamino acids (Cas); citrate (Cit); glucose (Glc); succinate (Suc); tryptone (Try). Growth phase: (e)xponential and (s)tationary. Strains: PAO1 and TBCF10839].

Two major groups of concentration patterns for the detected metabolites were identified. The first group comprised metabolites that were found in almost all analysed conditions. The concentrations of each of these metabolites were only slightly adapted to varying substrates and growth phases. This finding indicated that P. aeruginosa has a stable, nearly invariant core metabolism that is independent of carbon source and growth state. The second group represented metabolites that were found in only a few conditions, for example during exponential and stationary phase of cultures grown on the same carbon source. These metabolite spectra were comparable for both strains, which again indicated that variations in the metabolite profile are strongly defined by the environment rather than by the genetic repertoire of different P. aeruginosa strains.

Groups of concordantly high concentration metabolites within a given condition revealed metabolic pathways that are used for adaptation to a specific carbon source or growth phase. Enrichment of these metabolites occurred during metabolism and integration of the carbon source into the core metabolism, which leads to the hypothesis that, despite a diverse supply of nutrients, the core metabolism is kept in a stationary state. However, in cadaverine more than 30% of the metabolites found in other analysed conditions were below the detection limit and metabolite concentrations were in general strongly decreased. This finding indicated a strong downregulation of the overall core metabolism in agreement with the decreased growth rate. To enable P. aeruginosa for at least very low growth rates on this carbon source (Fig. 1), switching off dispensable pathways and restriction on pivotal metabolic processes might be required.

Altogether, the cluster analysis showed that the observed metabolite spectra were mainly defined by the growth medium. The influence of the growth phase seemed to be less important as in all cases the metabolite spectra of exponential growth and stationary phase were nearest neighbours in the HCA. Clonal differences had the least impact on metabolite patterns as under corresponding growth conditions the strains PAO1 and TBCF10838 always displayed the highest degree of similarity. Our results showed that both strains, regardless of their differences in origin and occupied habitats, share a common, P. aeruginosa-specific, carbon source and growth phase-dependent metabolism that is independent of the clonal lineage.

P. aeruginosa core metabolism

To define the P. aeruginosa core metabolism, all 145 identified metabolites were grouped in metabolic pathways (KEGG database; http://www.genome.jp/kegg/kegg2.html). The subgroup of 60 metabolites found in all analysed conditions was considered as representatives for a stably utilized core metabolism. Pathways containing metabolites of this group (significantly above average) were assigned to the core metabolism (Table 2, P-values were calculated by Monte Carlo simulation). As expected, the identified metabolic pathways belonged to the central amino acid and energy metabolism.

Table 2.  Metabolic pathways in the invariant core metabolism.
Overrepresented (in constitutive detected metabolites)P-value
Pathway
  1. Identified metabolites were grouped to metabolic pathways (according to KEGG database, http://www.genome.jp/kegg/kegg2.html), metabolites that were found in all analysed growth conditions (core metabolites) were significantly overrepresented in a set of central metabolic pathways. No pathways were found with a significant underrepresentation of these core metabolites.

Butanoate Metabolism0.00001
Alanine, Aspartate and Glutamate Metabolism0.00098
Cysteine and Methionine Metabolism0.00683
Phenylalanine Metabolism0.00871
Valine, Leucine and Isoleucine Biosynthesis0.00907
Citrate Cycle0.02059
Cyanamino Acid Metabolism0.02466

In the metabolite analysis, glutamate was identified as the compound with the highest concentration and variance in P. aeruginosa, dominating the metabolite spectrum (Fig. 2). This metabolite is part of or closely related to those pathways that were assigned to the core metabolism with high significance. Together with the observation that – independent of the growth condition – the concentrations of most metabolites were kept at nearly constant levels (Fig. 6), these findings suggested that glutamate and the related pathways may play a crucial role for the P. aeruginosa metabolism. The high concentration of glutamate allows P. aeruginosa to interconnect diverse pathways and to use this central metabolite as a universal substrate for various anabolic and catabolic processes.

Adaptation of P. aeruginosa to diverse carbon sources

Metabolic adaptation of P. aeruginosa was analysed by separate normalization and subsequent cluster analysis (Fig. 7). According to the previous HCA result, the metabolite patterns of strains PAO1 and TBCF10839 were defined by the growth condition. The most striking observation was the increase of distinct subsets of metabolites in a growth condition-characteristic manner. These sets of metabolites with high abundance reflected metabolic pathways that were specifically utilized by P. aeruginosa during metabolic adaptation to the respective medium and/or growth phase serving to integrate the available carbon source in a mainly unaltered background metabolism.

Figure 7.

Hierarchical clustering of the normalized metabolite concentrations of all 243 found metabolites. To illustrate the relative concentration changes of each metabolite, a heat map was generated based on relative metabolite concentration of a given sample compared with the mean of the concentration values across all samples for the given growth condition. This plot shows that, in a condition specific fashion, sets of metabolites were increased [Growth substrates: cadaverine (Cad); casamino acids (Cas); citrate (Cit); glucose (Glc); succinate (Suc); tryptone (Try). Growth phase: (e)xponential and (s)tationary. Strains: PAO1 and TBCF10839].

As expected, metabolites that were directly available as carbon sources and those belonging to the central catabolic pathways of the growth substrate were found in higher concentrations during exponential growth. In the stationary phase, with most growth media depleted from initial carbon sources, the metabolite pattern were more complex than during exponential growth. Stationary phase metabolomes were enriched with metabolites associated with sugar metabolism suggesting induced adaptation to stationary growth with, e.g. enhanced production of exopolysaccharides, development of biofilms, and production of rhamnolipids. Additionally, at high cell densities the nutrient-poor conditions led to degradation of less preferred amino acids (e.g. Lys, Phe, Trp, Tyr), the catabolic intermediates of which were overrepresented in the stationary state metabolomes.

This common metabolic pattern is flanked by two extremes: the growth in tryptone and cadaverine. In tryptone, as the most complex and nutrient-rich medium, the largest group of high concentration metabolites was found. Pseudomonas aeruginosa entered the stationary phase prior to metabolizing supplied media components (Fig. 1) and thus still had access to diverse nutrients in late growth phases. Consequently, the bacterial metabolomes of exponential growth and stationary phase were closely related and reflected an adaptation to conditions with a rich and diverse nutrient range. Metabolomes in tryptone were dominated by many highly concentrated metabolites of sugar and amino acid metabolism that superimposed on metabolites of the background metabolism. This effect was not observed in other media and thus resulted in clustering of metabolomes from tryptone-grown cells as outliers in the HCA (Fig. 7).

Cadaverine-grown cells also exhibited closely related metabolomes from exponential growth and stationary phase. However, in this case the insufficient carbon source evoked a fourfold extension of generation time, and these low growth rates were reflected by a metabolome state with the overall lowest metabolite concentrations. The high similarities in the metabolomes of exponential growth and stationary phase were caused by substrate limitations in both growth phases.

Although most analysed media (exception tryptone) reach a similar, nutrient-depleted state in the stationary phase (Fig. 1), the metabolomes of P. aeruginosa strains in this late growth phase were still biased by the initial substrate conditions. Maintaining the substrate-characteristic metabolite patterns under nutrient-poor conditions might reflect the corresponding enzymatic repertoire, which has been retained mainly unchanged, too. It is well known that enzymatic adaptation to a carbon source is often induced by the growth substrate itself (Lee et al., 2008; Boulette et al., 2009). This specified adaptation process appears to be only slowly reversed under food-limiting conditions as indicated by reduced lag phases if inoculates had been grown on the same substrate.

Conclusion

The analysis of the P. aeruginosa metabolome revealed specified metabolic responses that arise in a nutrition-dependent manner. As previously reported for Pseudomonas putida (van der Werf et al., 2008) induction of substrate-specific metabolic pathways (represented by distinct sets of metabolites) indicates the precise regulation of adaptation processes. Beside this adaptive response, a basic core metabolism shapes the P. aeruginosametabolome independent of growth phase, carbon source or clonal variations.

Experimental procedures

Pseudomonas aeruginosa strains

All experiments were performed with two distantly related, genetically and phenotypically well-characterized P. aeruginosa isolates. Strain PAO1 (genotype 0002) was isolated in 1955 in Melbourne, Australia, from an infected wound (Holloway, 1955) and is the sequenced major reference strain widely used in P. aeruginosa research (Stover et al., 2000). Strain TBCF10839 (genotype 3C52) (Wiehlmann et al., 2007a) was isolated in the early eighties in Hannover from a chronically colonized cystic fibrosis patient (Wiehlmann et al., 2007b). It is a mucoid strain, harbouring a mucA nucleotide substitution [A342G, compared with strain PAO1 sequence (http://www.pseudomonas.com), which, however, does not lead to loss of function (J. Klockgether, pers. comm.).

Media and cultivation

For metabolome analysis precultures and main cultures of P. aeruginosa strains PAO1 and TBCF10839 were grown in either minimal medium [2 g l−1 (NH4)2SO4, 4 g l−1 Na2HPO4, 3 g l−1 KH2PO4, 3 g l−1 NaCl] with trace metals (0.49 g l−1 MgSO4 × 7H2O, 0.04 g l−1 CaCl2 × 6H2O, 20 µg l−1 MnCl2 ×  4H2O, 20 µg l−1 CuSO4, 20 µg l−1 ZnSO4 × 7H2O, 10 µg l−1 Na2MoO4 × 2H2O, 2.8 µg l−1 FeSO4 × 7H2O) containing a sole carbon source [40 mM succinate (Roth), 40 mM citrate (Sigma), 40 mM glucose (Roth), 5 mM cadaverine (Sigma) or 1% casamino acids (Roth)] or in complex media containing tryptone (AppliChem) (20 g l−1)/NaCl (10 g l−1).

All cultivations were carried out at 37°C and 250 r.p.m. on a rotary shaker in triplicate.

For adequate aeration baffled shake flasks (500 ml) were used with 150 ml of culture medium.

The specific growth rates µ were calculated according to Berney and colleagues (2006) (µ = Δln OD/Δt).

Metabolome analysis

Intracellular metabolites.  For sampling of intracellular metabolites 150 mg of cells (wet weight) was harvested by centrifugation (5000 g, 4°C, 5 min) and the media supernatant was removed.

Extraction of intracellular metabolites was performed as described by Haddad and colleagues (2009). Briefly, the following steps were carried out at 4°C. Obtained cell pellets were washed twice with 20 ml of 0.9% NaCl solution and finally resuspended in 1.5 ml methanol containing 40 µl ribitol (0.2 g l−1) per ml methanol as internal standard. The cells were disrupted by ultrasonification (15 min, 70°C). 1.5 ml H2O was added and samples were thoroughly mixed for 30 s. The addition of 1 ml chloroform was followed by mixing and subsequent centrifugation (4000 g, 4°C, 10 min) for separation of hydrophilic and hydrophobic phases. One millilitre of the polar phase was transferred to a new reaction tube and dried.

For derivatization dried samples were dissolved in 50 µl pyridine containing methoxyamine hydrochloride (20 mg ml−1) and incubated at 600 r.p.m. and 30°C for 90 min. After adding 80 µl N-methyl-N-trimethylsilyltrifluoracetamide (Chromatographie Service, Langerwehe, Germany) samples were incubated again (600 r.p.m., 37°C, 30 min) followed by another 120 min incubation at 20°C.

Finally, 6 µl of an alkane mix containing decane, dodecane, pentadecane, nonadecane, docosane, octacosane, dotriacontane and hexatriacontane (20 mg ml−1 in cyclohexane each) was added in order to allow retention indices calculations (Vandendool and Kratz, 1963; Strelkov et al., 2004).

A sample aliquot of 2 µl was injected in a Finnigan Trace gas chromatograph (ThermoFinnigan, San Jose, USA) equipped with a DB-5MS column (J&W Scientific, Folsom, USA). Eluted compounds were analysed with a Trace mass spectrometer (ThermoFinnigan, San Jose, USA) after electron impact ionization. Applied parameters for sample injection, gas chromatography and mass spectrometry were described before (Strelkov et al., 2004). The metabolite identification and quantification were previously described (Strelkov et al., 2004; Spura et al., 2009).

Extracellular metabolites.  For studying the extracellular metabolites, 10 ml of supernatants was taken at the particular time points. The metabolites were extracted at 4°C. The bacterial cells were removed by centrifugation (2× 6000 g, 5 min, 4°C) and an aliquot of the upper part of the centrifuged supernatant was used for metabolome analysis.

The supernatant was diluted 20-fold with water. An aliqout of 1 ml was mixed with 1.5 ml of methanol, containing 12 µg of ribitol as internal standard. The samples were mixed for 60 s. 1 ml chloroform was added and hydrophilic and hydrophobic phases were separated by centrifugation (4000 g, 4°C, 5 min); 1 ml of the polar phase was transferred to a new tube and dried.

For derivatization, dried samples were solved in 40 µl pyridine, containing methoxyamine hydrochloride (20 mg ml−1) and incubated at 600 r.p.m. and 30°C for 90 min. After adding 70 µl of N-methyl-N-trimethylsilytrifluoroacetamide samples were incubated again (600 r.p.m., 37°C, 30 min) followed by another incubation (600 r.p.m., 25°C, 120 min) and centrifugation (18 000 g, 20°C, 5 min). The GC-MS analysis was performed with the supernatant (after addition of the above-mentioned alkane standard).

Statistical analysis

Generally, all metabolome analysis experiments were performed with six replicates (tryptone samples were analysed in duplicates as the macromolecular components of this medium may cause damages of the GC column). Metabolite concentrations were scaled to the internal standard ribitol and to the optical density of cell culture and then calibrated to total signal intensity of the GC spectra. Metabolite concentrations below the detection limit were set to the lowest measured signal. Median intensity values of the six replicates were applied to statistical analysis. Statistical tests for significance (Kruskal–Wallis and Wilcoxon–Mann–Whitney tests) were performed with the R package (http://www.r-project.org/; Gentleman et al., 2005; R Development Core Team, 2008). Non-random pathway-metabolite association was tested by Monte Carlo simulations (Sham and Curtis, 1995).

Principal component analysis

Principal component analysis is a non-parametric method used for pattern recognition requiring no a priori knowledge of data structure. This technique is designed to simplify data sets by reducing the dimensionality of the multivariate data. The resulting compressed data preserve the relevant features of the original data set, which are represented by a set of new variables, termed principal components. Briefly, PCA classifies data sets of unknown structure into meaningful groups. In this analysis standard parameters implemented in the TIGR Mev software (http://www.tm4.org/mev, Eisen et al., 1998) were used. The PCA based on logarithmic calculus was performed simultaneously for all data sets. The principle component calculations were performed with a Centering Mode based on mean.

Unsupervised hierarchical clustering (HCA)

This method provides organization of primary data sets without predefined classification. Based on similarity, clustering of samples as well as associated signature features of sample are visualized by dendrograms. Analysis was performed simultaneously for all data sets. The HCA was based upon logarithmic calculus implemented in the TIGR Mev software. Mean centred values were normalized for metabolite/row, and where required for condition/column. Pearson correlation was used as similarity metric for the HCA.

Acknowledgements

We would like to thank Colin Davenport for assistance with statistical analyses; and Burkhard Tuemmler for stimulating discussions during the preparation of this manuscript. S.H. receives a predoctoral Lichtenberg stipend by the Niedersächsisches Ministerium für Wissenschaft und Kultur and financial support by the DFG-funded Hannover Biomedical Research School. L.W. is supported by a competitive research grant provided by Hannover Medical School (HLF04/08).

Ancillary