Identification of conditionally essential genes for growth of Pseudomonas putida KT2440 on minimal medium through the screening of a genome-wide mutant library


  • M. Antonia Molina-Henares,

    1. Consejo Superior de Investigaciones Científicas, Estación Experimental del Zaidín, Department of Environmental Protection, C/Prof Albareda, 1, E-18008 Granada, Spain.
    Search for more papers by this author
  • Jesús De La Torre,

    1. Consejo Superior de Investigaciones Científicas, Estación Experimental del Zaidín, Department of Environmental Protection, C/Prof Albareda, 1, E-18008 Granada, Spain.
    Search for more papers by this author
  • Adela García-Salamanca,

    1. Consejo Superior de Investigaciones Científicas, Estación Experimental del Zaidín, Department of Environmental Protection, C/Prof Albareda, 1, E-18008 Granada, Spain.
    Search for more papers by this author
  • A. Jesús Molina-Henares,

    1. Consejo Superior de Investigaciones Científicas, Estación Experimental del Zaidín, Department of Environmental Protection, C/Prof Albareda, 1, E-18008 Granada, Spain.
    Search for more papers by this author
  • M. Carmen Herrera,

    1. Consejo Superior de Investigaciones Científicas, Estación Experimental del Zaidín, Department of Environmental Protection, C/Prof Albareda, 1, E-18008 Granada, Spain.
    Search for more papers by this author
  • Juan L. Ramos,

    Corresponding authorSearch for more papers by this author
  • Estrella Duque

    1. Consejo Superior de Investigaciones Científicas, Estación Experimental del Zaidín, Department of Environmental Protection, C/Prof Albareda, 1, E-18008 Granada, Spain.
    Search for more papers by this author

E-mail; Tel. (+34) 958 181608; Fax (+34) 958 135740.


In silico models for Pseudomonas putida KT2440 metabolism predict 68 genes to be essential for growth on minimal medium. In this study a genome-wide collection of single-gene P. putida KT2440 knockouts was generated by mini-Tn5 transposon mutagenesis and used to identify genes essential for growth in minimal medium with glucose. Our screening of the knockout library allowed us to rescue mutants for 48 different knockouts that were conditionally essential for growth on minimal medium. The in vivo screening showed that 24 of these mutants had a insertion in genes proposed to be conditionally essential based on in silico models, whereas another 24 newly implicated conditionally essential genes have been found. For 10 of the in silico proposed conditionally essential genes not found in the screening, knockout mutants were available at the Pseudomonas Reference Culture Collection. These mutants were tested for conditional growth on minimal medium, but none of them was shown to be essential, suggesting that the in silico proposal was inaccurate. Among the set of identified conditionally essential genes were a number of genes involved in the biosynthesis of certain amino acids and vitamins. Auxotrophs for all amino acids predicted by the in silico models were found and, in addition, we also found auxotrophs for proline, serine, threonine and methionine, as well as auxotrophs for biotin, nicotinate and vitamin B12 that were not predicted in silico. Metabolic tests were performed to validate the mutants' phenotypes. Auxotrophies for l-Arg, l-Leu, l-Pro and l-Cys were bypassed by external addition of the corresponding d-amino acids, suggesting the existence of number of d- to l-amino acid racemases encoded by the KT2440 genome. Therefore, the in vivo high-throughput analysis presented here provides relevant insights into the metabolic cross-road of biosynthetic pathways in this microorganism, as well as valuable information for the fine tuning of current in silico metabolic models.


Pseudomonas putida KT2440 is a soil microorganism (Worsey and Williams, 1975; Timmis, 2002) that colonizes the roots of a number of plants and is used as a saprophytic model microorganism in the study of plant-microbe interactions (Molina et al., 2000; Espinosa-Urgel and Ramos, 2004; Ramos-González et al., 2005; Matilla et al., 2007; Segura et al., 2009). The genome of KT2440 was sequenced by Nelson and colleagues (2002), and more recently the genome sequence of at least four other P. putida strains (W619, GB1, F1 and DOT-T1E) has been reported (Taghavi et al., 2009;; A. Segura, unpublished). The genomic sequences of several strains belonging to different species of the genus Pseudomonas have also been deciphered; and comparison of the sequences between all available Pseudomonas strains revealed that there are around 2000 genes in common (Weinel et al., 2002; Vodovar et al., 2006; Taghavi et al., 2009; García-Valdés et al., 2010). These genes are believed to constitute the core set of genes that make up the basic genome of bacteria of the genus Pseudomonas.

High-throughput mutational analysis in P. aeruginosa PAO1 and PA14 (Jacobs et al., 2003; Liberati et al., 2006), and signature-tagged mutagenesis (Lehoux et al., 2002), revealed that mutants in nearly 8% of ORFs could not be obtained, which corresponded to about 400 genes that were considered to be essential genes for growth and survival (Jacobs et al., 2003; Liberati et al., 2006;Goldberg, 2010). Among this set of genes are those that encode RNA polymerase subunits, sigma factors, proteins involved in DNA metabolism (i.e. replication, some repair systems), t-RNAs, genes encoding a number of chaperones, as well as some essential genes encoding energy-production proteins and central metabolism enzymes, among others (Lehoux et al., 2002; Jacobs et al., 2003; Liberati et al., 2006). Many of these essential genes are present in monocopy in the genome of P. putida, such that it has been proposed that they may also be essential genes (Duque et al., 2007; and J. de la Torre and E. Duque, unpublished).

A series of bioinformatic tools have been developed to predict which metabolic and regulatory genes are required for growth of different microorganisms under given environmental conditions, particularly human pathogens (Edwards et al., 2001; Covert and Palsson, 2002; Segre et al., 2002; Shlomi et al., 2005; Joyce et al., 2006). This information has led in turn to the validation of predictions by testing the role of and conditions under which these genes are essential in vivo (Maeda et al., 2001; Sassetti et al., 2001; Giaever et al., 2002; Thanassi et al., 2002; Gerdes et al., 2003; Jacobs et al., 2003; Kobayashi et al., 2003; Boutros et al., 2004; Song et al., 2005; Baba et al., 2006; Glass et al., 2006). In contrast with the detailed knowledge of conditionally essential genes in pathogens is the paucity of such information for saprophytic microbes. Knowledge of which genes in an organism are conditionally essential is relevant to understand how cellular networks work. The elucidation of the critical steps in such cellular networks provides important information to facilitate the identification of industrial applications of microbes (Böltner et al., 2008; Roca et al., 2008). Recently, two metabolic models of Pseudomonas putida were generated (Nogales et al., 2008; Puchalka et al., 2008). These studies provided information for almost 900 metabolic reactions along with Boolean logic statements linking the genes to protein complexes and protein complexes to reactions (Nogales et al., 2008; Puchalka et al., 2008). These two in silico P. putida models have proposed the existence of 68 putative conditionally essential genes for growth on minimal medium (see Table S1a and b), although only 12 genes were considered essential by both models (Table S1c). Validation of the models was partially achieved by generating in silico mutants (Nogales et al., 2008), and by assaying the ability of KT2440 to oxidize 95 carbon sources in a BIOLOG assay (Puchalka et al., 2008). A limitation of the latter assay is that it only provides data regarding the oxidation of the tested compounds, but does not provide growth data.

The most widely used tool for generation of KT2440 mutants is a series of miniTn5 transposons (Herrero et al., 1990; de Lorenzo and Timmis, 1994). We have previously shown the feasibility of isolating conditional mutants deficient in the biosynthesis of aromatic amino acids after massive miniTn5 mutagenesis and appropriate selection procedures (Duque et al., 2007). This study established that, in P. putida KT2440, aromatic amino acid biosynthesis takes place through an interplay of convergent pathways for phenylalanine and tyrosine and a single highly conserved biosynthetic pathway for tryptophan synthesis (Molina-Henares et al., 2009). However, no detailed global functional analysis has ever been undertaken to specifically identify conditionally essential genes for growth in P. putida, i.e. metabolic genes that are not essential in rich medium, but which are essential for growth in minimal medium.

In this study a genome-wide collection of viable single-gene knockouts of P. putida KT2440 was created and arranged on rich-medium plates for use in high-throughput assays. This collection has been used to identify genes essential for growth in minimal medium with glucose. Comparison of the proposed in silico circuits and the in vivo mutant collection presented here provides important new insights into the metabolic cross-road of biosynthetic pathways in this microorganism.


In silico identification of essential genes in P. putida KT2440

Computational analysis of single-gene inactivation events according to the iJP815 model (Puchalka et al., 2008) identified 26 conditionally essential genes for growth on minimal medium (Table S1b), while the iJN746 model (Nogales et al., 2008) predicted 53 genes (Table S1a). Taking into account the common and uncommon genes predicted by iJP815 and iJN746, a total of 68 genes were identified in silico to be necessary for growth on minimal medium (see the list of 12 common genes predicted by both models in Table S1c). These conditionally essential genes encode proteins that were involved in the biosynthesis of several amino acids [arginine, histidine, phenylalanine, lysine, cysteine, isoleucine, leucine and tryptophan (41 genes)], nucleotide metabolism [pyrimidine and purine biosynthesis, (12 genes and 5 genes of nucleotide biosynthesis and metabolism general)], cofactors (3 genes for biosynthesis of nicotinic acid and pantothenate]), as well as genes involved in general metabolism [triose phosphate isomerase and succinate dehydrogenase complex (5 genes)], and cell wall proteins [NlpD and SurE (2 genes)].

Screening of a mutant library to identify conditionally essential genes on minimal medium with glucose

The approach that we adopted in order to experimentally identify P. putida conditionally essential genes involved the screening of a genome-wide collection of KT2440 mutants that were generated using miniTn5-Km (see Experimental procedures). The resulting library consisted of 7760 independent miniTn5 clones capable of growing on complex Luria–Bertani (LB) medium. Random analysis of 12 of these clones by Southern blot hybridization confirmed that none of them were siblings and hence the library was considered to represent a genome-wide knockout mutant library useful for high-throughput screening. We therefore tested the ability of these mutants to grow on M9 minimal medium with glucose as the sole carbon source in order to identify conditionally essential genes. We found that 79 mutants were reproducibly incapable of growth on glucose minimal medium. Subsequently, the insertion site in all 79 mutants was identified by sequencing the DNA adjacent to the miniTn5. This allowed us to identify 47 independent knockout genes (Table 1).

Table 1.  Knockout mutants of Pseudomonas putida unable to grow on minimal medium.
LocusGenesPredicted in silicoNumber of times the mutant was identified
  1. The knockout ORFs that yield mutants unable to grow on minimal medium with glucose are shown, together with the gene name, whether or not the mutant was predicted in silico and the number of times the mutant was identified in the library through high-throughput screening.

PP3850calcium-binding protein, hemolysin-typeNo2

We then compared the in silico predicted conditionally essential genes with those identified via miniTn5 mutagenesis (i) to determine which genes had been correctly predicted in silico, (ii) to identify those genes that were predicted in silico, but that were not found in the in vivo screen, and (iii) to further explore the function of the in vivo identified genes that were not predicted in silico. We found that only 38% of the genes predicted in silico had a knockout miniTn5 mutant in the screened library (Table 1). Therefore from a functional point of view, 25 mutants were correctly predicted. Table 2 shows the genes predicted in silico for which no conditionally essential mutant was present in the screened mutant library. An alternative mutant collection to the above platform is the Pseudomonas Reference Culture Collection (PRCC) (Duque et al., 2007) that consists of P. putida mutants generated using different approaches and contains over 3400 independent clones, each with a single gene knockout. Although the format of this library does not allow it to be used with high-throughput assays, is useful for specific screening. Ten mutants of the 38 conditionally essential genes predicted in silico but not found in vivo were available at the PRCC and they were tested for growth on minimal medium with glucose as a carbon source (Table 2). All mutants grew with doubling times similar to those of the parental strain, and therefore we concluded that the in silico annotations of these genes as conditionally essential genes were inaccurate.

Table 2.  Mutants predicted in silico not found in vivo and some available or not in the PRCC.
Predicted in silicoAvailable in PCRRAuxotropha
  • a. 

    When the mutant was available in the PRCC, growth on M9 minimal medium with glucose as the sole carbon source was tested. In the column ‘auxotroph’, No means that the strain grew to high cell density (OD660 > 1) in 24 h.


For the 28 genes predicted to be conditionally essential in minimal medium, no mutants were available in any collection and consequently no predictions could be made.

The in vivo screen identified five genes not predicted by the in silico models namely: edd, eda, dnaJ, PP5305 and PP3850. Mutants in PP3850 appeared twice in the collection, which supports the essential role of this gene. The gene product of ORF PP3850 has homology to Ca2+ binding proteins with peptidase activity, but its role remains unidentified. It has been shown previously that glucose in KT2440 is metabolized through a series of convergent pathways that lead to the synthesis of 6-phosphogluconate, which is subsequently metabolized via the Entner-Doudoroff enzymes Edd and Eda (Vicente and Cánovas, 1973; Blevins et al., 1975; del Castillo et al., 2007; Daddaoua et al., 2009). Mutants in the edd or eda genes did not grow on minimal medium with glucose. These two mutants were found in the screen of mutant libraries, although the in silico programmes fail to identify them. These mutants grew on minimal M9 medium when either citrate or lactate was used as a carbon source, as expected. The gene product of PP5305 has been annotated as putative dehydratase although its role is unknown. The inability of the dnaJ mutant to grow on minimal medium was striking because we had previously shown that in P. putida there are many chaperones that functionally replace each other (Segura et al., 2005). Therefore, we analysed the physical organization of the genome to search for potential downstream effects. We found that a gene involved in lysine biosynthesis (dapB) was located 3′ from dnaJ with 13 nucleotides between the stop codon of dnaJ and dapB (Fig. S1) and hypothesized that the incapability of the dnaJ mutant to grow on minimal medium was due to the polar effect on a downstream gene. In this regard we tested whether l-lysine could restore the growth of the dnaJ mutant on minimal medium with glucose as the sole C source, which was found to be the case, indicating that the insertion in dnaJ is dispensable.

Transcriptomic analysis reveals that conditionally essential genes are expressed when cells are grown on minimal medium with glucose

A premise of conditionally essential genes is that they should be expressed in cells growing under the specific condition being tested. To confirm this hypothesis we analysed global transcriptomic profiles of P. putida cells growing on M9 minimal medium with glucose. To this end total RNA was extracted, cDNA was synthesized and labelled with the fluorosphore Cy3 and hybridized against a microarray that contains a 50-mer oligonucleotide that covers all ORFs of KT2440. Expression was recorded by determining the intensity of fluorescence. Subsequently, we compared the expression level of each of the conditionally essential genes that we identified in vivo with two internal calibrators, the gltA gene (citrate synthase; relative level 14500 units), whose expression is essential under all growth conditions for this strict aerobe, and pobA (p-hydroxybenzoate hydroxylase; relative level 920 units), which is a gene known to be induced in response to p-hydroxybenzoate and which is not expressed in cells growing with glucose. We found that the expression levels, given as fluorescence intensities, of all of the experimentally identified conditionally essential genes were higher than that of pobA (Fig. 1) that five genes were expressed at a higher level than the gltA gene. This set of data confirms that the genes identified as conditionally essential under a specific growth condition are indeed transcribed under this specific set of culture conditions. Importantly, this information should not be misinterpreted in regard to the actual expression levels of the genes. In fact, it has been well established that, in yeast growing on glucose, the absolute mRNA levels of genes involved in the biosynthesis of amino acids do not reflect actual fluxes since other processes, such as post-transcriptional regulatory strategies, protein stability, and metabolic fluxes, are utilized to fine-tune protein levels in order to allow for balanced growth of the cells (Moxley et al., 2009).

Figure 1.

Expression level of genes found to be conditionally essential for growth on minimal medium with glucose. Bacterial cells were grown on M9 minimal medium with glucose until their culture reached a turbidity of 0.7 at 660 nm. Then, RNA was extracted and cDNA labelled with the Cy fluorophore, as described under Experimental procedures. Samples were hybridized to a 50-mer DNA chip that contains most ORFs of the P. putida KT2440 genome. Expression levels were recorded as the fluorescence intensity of spots corresponding to the set of genes shown in Table 1. As internal calibrators we used the pobA gene, for which no expression under glucose culture conditions occurred (Ramos-González et al., 2005), and gltA, encoding citrate synthase a gene that is expressed under all growth conditions in Pseudomonas.

Annotations of conditionally essential genes

Most annotations of conditionally essential genes corresponded to biosynthetic pathways for amino acids, nucleotides and vitamins, and consequently these mutants were expected to be auxotrophs. There is no detailed analysis of auxotrophs in P. putida, and only a limited number of auxotrophs have been isolated and studied in other species of the genus Pseudomonas (Crawford and Gunsalus, 1966; Calhoun and Weary, 1969; Cuppels, 1986; Barth and Pitt, 1996; Thomas et al., 2000; Jacobs et al., 2003; Menard et al., 2007). Therefore, we decided to confirm the auxotrophic character of these mutants by submitting them to growth assays using the amino acids and vitamins shown in Table 3. These substrates were chosen by considering the biosynthetic pathways from which proteinogenic amino acids are derived, i.e. those from Krebs cycle intermediates (aspartate and glutamate groups), the pentose phosphate pathway (histidine), as well as central metabolites such as glyceraldehyde 3-phosphate and pyruvate (serine group, aromatic amino acids and short hydrophobic lateral chain amino acids). Figure 2 illustrates the various pathways that exist for the biosynthesis of amino acids.

Table 3.  Supplements added as nutrients for auxotroph characterization.
  1. The composition of media 1–5 are listed vertically in the table. The composition of media 6–10 are listed horizontally. Medium 11 is an assortment of compounds not included in the others, its contents are listed horizontally at the bottom of the table. All the nutrients are used at the final concentrations between 0.2 and 0.6 mM.

 9GlutamineAsparagineUracilAspartic acidArginine
10ThymineSerineGlutamic acidDAPGlycine
11PyridoxineNicotinic acidBiotinPantothenateAlanine
Figure 2.

Biosynthesis of proteinogenic amino acids from glucose. This scheme shows the main intermediates generated in glucose metabolism and the starting chemicals for the biosynthesis of proteinogenic amino acids. More details on these biosynthetic pathways are shown as Supporting information or have been published previously (Molina-Henares et al., 2009). OM, PG and IM stand for outer membrane, periplasmic space and inner membrane respectively.

The results of these assays revealed, as expected, that many of these knockout strains were auxotrophs, and we found auxotrophs for the eight amino acids predicted by the in silico programmes (Table 1). These included auxotrophs for the following amino acids: histidine (4), lysine (1), phenylalanine (1) arginine (6), isoleucine/valine (2), leucine (3) and tryptophan (6). We also found auxothophs for proline (1), serine (2), threonine (1), methionine (1) and tyrosine (1), which were not predicted by the in silico analysis (Table 4). The isolation of auxothrophs for amino acids that were not predicted by the models reflect some of the technical limitations of the in silico models, which are probably due to the incorrect annotation of the genome of KT2440, and the limited number of reactions that the models take into consideration. Therefore, the obtained in vivo information will be useful to reformulate the metabolic programmes in a kind of in vivo to in silico feedback approach. In order to shed light on the implications of these findings, all biosynthetic pathways involved in the synthesis of amino acids are analysed below.

Table 4.  Growth of auxotroph mutants on mineral medium with l- and d-amino acids.
  1. The carbon source in all cases was 16 mM glucose. All of the added nutrients were used at a final concentration of 0.2 mM.

  2. The symbol ‘+’ means that the turbidity at 660 nm of the culture after 24 h was ≥ 1 unit.

  3. The symbol ‘−’ means that the turbidity of the culture was < 0.1 unit at 660 nm.

  4. The symbol ‘+/−’ means that the turbidity of the culture was high as 0.3 unit at 660 nm.

  5. NT means not tested.

Tryptophan auxotrophsM9l-tryptophand-tryptophan
Arginine auxotrophsM9l-arginined-arginine
Leucine auxotrophsM9l-leucined-leucine
Serine auxotrophsM9l-serined-serine
Threonine auxotrophM9l-threonined-threonine
Proline auxotrophM9l-prolined-proline
Methionine auxotrophM9l-methionined-methionine
Isoleucine auxotrophsM9l-isoleucined-isoleucine
Histidine auxotrophsM9l-histidined-histidine
Tyrosine auxotrophsM9l-tyrosined-tyrosine
Phenylalanine auxotrophsM9l-phenylalanined-phenylalanine
Cysteine auxotrophsM9l-cysteined-cysteine
Lysine auxotrophsM9l-lysined-lysine

Six of the auxotrophs identified in this study required vitamins for growth. One mutant required cobalamine (vitamin B12), four of the mutants required biotin (vitamin H) and one required nicotinic acid (vitamin B3). The pathways leading to the biosynthesis of these three vitamins are analysed in detail below. It was predicted by the two in silico models that mutant PP4699 (panB) and PP4700 (panC) would be deficient in the biosynthesis of pantoneate, and would not grow on minimal medium regardless of the carbon source. Neither of these mutants appeared in our screen, but mutant PP4700 was available from the PRCC. Upon testing the growth of this mutant in minimal medium, we found that it is not an auxotroph (Table 2) and, therefore, it appears that redundant enzymes exist for the biosynthesis of pantoneate.

Analysis of different amino acid biosynthetic pathways

To shed further light on why some expected mutants related to the biosynthesis of amino acids were not found and why some mutants not predicted by the in silico programmes appeared with regards to the biosynthesis of certain amino acids, we analysed the different biosynthetic pathways from a physical and transcriptional point of view, and also carried out a series of nutritional tests to confirm the derived genetic information.

Amino acids synthesized from α-ketoglutarate.  Four amino acids are synthesized from α-ketoglutarate, a Krebs cycle intermediate, namely, glutamate, glutamine, arginine and proline (Fig. 2). Arginine and proline auxotrophs were found, but none was found for glutamate or glutamine. In the biosynthesis of proline, glutamate is initially activated in an ATP-dependent reaction catalysed by ProB that yields γ-glutamyl-phosphate (Fig. S2). Proline is then made via two consecutive reactions mediated by ProA and ProC (Fig. S2). Although the two in silico models did not predict proline auxotrophs, we found that a knockout in proA led to proline auxotrophy. The genes for the biosynthesis of proline are unlinked in the host chromosome, with a single copy each of proB (PP0691) and proA (PP4811), and two copies of proC (PP3778 and PP5095). The proB gene overlaps with a 3′ downstream gene that encodes the ObgE GTPase. Therefore, the lack of mutants in proB could be related to polar effects. Failure to isolate auxotrophs in the last step is likely due to the presence of two copies of proC. As expected, growth of a proA mutant on minimal medium was restored by the addition of l-proline. We also found that d-proline could restore growth of the proA mutant strain, suggesting the existence of a d- to l-proline racemase (Table 4).

In regards to the biosynthesis of arginine, glutamate is first converted into N-acetylglutamate (ArgA), followed by seven consecutive steps leading to the production of l-arginine (Fig. 3). Among the mutants with auxotrophy for arginine we found miniTn5 insertions within the argB, argA, argD, argF, argG and argH genes. As expected, growth defects of the argA, argB and argD mutants were restored by arginine, l-ornithine and l-citrulline. In agreement with in silico models, no arginine auxotrophs in argC were found since two argC genes exist in the genome of KT2440 (PP3633 and PP0432).

Figure 3.

Metabolic steps in the arginine biosynthesis pathway and the enzymes responsible for the corresponding reaction. Mutants in most of the steps were isolated (Table 1) and were shown to be arginine auxotrophs. Metabolic feeding assays showed that mutants deficient in argA, argB and argD were able to grow on minimal medium if the culture was supplemented with 0.6 mM ornithine, citrulline or arginine, whereas mutants in argF were able to grow on minimal medium if supplemented with arginine and citrulline but not with citrulline. A mutant deficient in argH grew on minimal medium if and only if arginine or d-arginine was supplied in the culture medium.

With the exception of PP5185/PP5186 (argA/argE), all genes for arginine biosynthesis are dispersed on the host chromosome and organized as monocystronic units or in clusters along with genes unrelated to arginine biosynthesis (not shown). Although argA and argE are contiguous on the genome, they are separated by 124 bp and RT-PCR analysis showed that they are independent transcriptional units (Fig. S3). Upstream of argA is the PP5184 gene that encodes a glutamine synthetase, which is also involved in initial metabolism of glutamate.

A relevant finding derived from metabolic feeding assays is that d-arginine restored growth all of the arg mutants (Table 4), suggesting that a d- to l-arginine racemase system operates within P. putida. Li and Lu (2009) reported a new two-component amino acid racemase (DauA/DauB) responsible for the d-to-L inversion of d-arginine in P. aeruginosa. Homologues of the genes of this racemase system are present in the genome of P. putida KT2440. A dauA mutant (PP2246) was present in the PRCC, which was unexpectedly able to use d-arginine as the sole source of nitrogen, suggesting that in addition to the in silico identified genes other d- to l-racemases may exist in P. putida KT2440.

Glutamate auxotrophs were not expected because of the existence of two enzymes that can lead to its biosynthesis, namely glutamate dehydrogenase (α-ketoglutarate +  NH3 → glutamate) and glutamate synthase (glutamine + α-ketoglutarate → 2 glutamate). Failure to isolate glutamine auxotrophs is the result of the fact that up to eight glutamine synthetases (GS) (PP2178, PP3148, PP4399, PP4547, PP5046, PP5183, PP5184 and PP5299) are present in the genome of this strain (Duque et al., 2007), although the in silico iJP815 model only proposed PP5299 as the true glutamine synthetase. blast analysis of the eight potential GS cited above revealed that they are more than 50% identical. Mutants in five of these GSs were available from the PRCC and none of them exhibited auxotrophy for glutamine, revealing that redundancy for this critical enzymatic activity exists in P. putida KT2440. Array analysis of cells growing on M9 minimal medium with glucose revealed that at least five of the glnA genes (PP2178, PP4399, PP4547, PP5046, PP5184) were expressed at a level that was two- to fivefold higher than that of the pobA gene.

Biosynthesis of hydrophobic amino acids from pyruvate.  Pyruvate is the key compound for biosynthesis of three amino acids, namely, alanine, valine and leucine. Alanine is made via the transamination of pyruvate. Alanine auxotrophs were not found, which suggests that one or more of the 18 broad-range substrate-specificity transaminases encoded in KT2440 can catalyse this reaction (Table S2). This is not surprising given the wide substrate range of these enzymes and their overlapping substrate specificity (Soda and Osumi, 1969; Christoferson et al., 2008)

In terms of valine and leucine biosynthesis, there exists, between them, a series of common reactions (Fig. 2). The process begins with the convertion of pyruvate into 2-oxoisovalerate by a series of decarboxylation and oxidation steps carried out by IlvB, C and D proteins. 2-Oxoisovalerate can then be either transaminated to yield l-valine in a reaction medited by the IlvE transaminase or converted by LeuA, B, C and D to yield 4-methyl-2-oxopentonate, which is then converted by IlvE into l-leucine. In agreement with in silico models, IlvC and IlvD mutants were found, and both required valine and leucine for growth.

The ilvBHC genes form an operon, while the other ilv genes (ilvD and ilvE) are scattered in the host chromosome and are expressed as monocystronic units (data not shown). The leuCDB genes form a single operon, whereas leuA and ilvE are unlinked to other leu genes. Mutants in leuA, leuC and leuD were found, all of which were leucine auxotrophs (Table 1). We found that these mutants were recovered by the external addition of l-leucine, as expected, and also by d-leucine (Table 4), which supports the existence of a d-to-L leucine racemase.

Biosynthesis of histidine from the pentose phosphate cycle.  Histidine is made from ribulose-5-phosphate. The iJN746 model predicts that inactivation of up to seven his genes could lead to histidine auxotrophy. The his genes are organized into four clusters: (i) PP0289 through PP0293 (hisB, hisH, PP0291, hisA and hisF); (ii) PP0965 to PP0967 (hisGDC) and (iii) PP5014/PP5015 (hisIE), and (iv) the monocistronic gene hisZ (PP4890). The three gene clusters gave rise to three independent operons when co-transcription of genes was assayed by RT-PCR (not shown). Both in silico models predicted that mutants in the first two clusters would be auxotrophs for histidine. We found that mutants in genes in the first cluster, i.e. hisF, hisB and hisH mutants (Table 1), were auxotrophs, but we failed to find any within the other two clusters, although a mutant in hisZ, that encodes the regulatory subunit of the ATP phosphoribosyl transferase and involved in the first step in the histidine biosynthetic pathway was indeed found, although it was not predicted by the in silico models (Fig. S4). It is therefore unclear why mutants in hisGDC or hisIE were not found in the general screening. These mutants are not available at the PRCC either and some intermediates, such as l-histidinol or some of the phosphorylated intermediates of histidine metabolism, may be toxic to cells (Fig. S4).

Amino acids derived from oxaloacetate.  Aspartate is produced from the Kreb's cycle intermediate oxaloacetate, which is also a precursor for the biosynthesis of asparagine, lysine, methionine and threonine, which may be further converted to l-isoleucine (Fig. 2). Although only a single ORF (PP1750) is annotated as an asparagine synthase (Nelson et al., 2002), no auxotroph mutants requiring asparagine were isolated. This is most likely due to the availability of multiple amino transferases that can replace each other in vivo, as discussed above.

A mutant in the asd gene was found (Table 1). This gene encodes aspartate semi-aldehyde dehydrogenase, which converts β-aspartyl phosphate into aspartate-β-semi-aldehyde, and is involved in lysine, threonine and methionine biosynthesis. As expected, this mutant exhibited multiple auxotrophy, and simultaneously required all three amino acids, as well as diaminopimelate for growth (Fig. S5). This finding agrees with an earlier report by Ronchel and Ramos (2001).

Aspartate semialdehyde can be converted, through the action of DapA, into dihydrodipicolinate, which is then subsequently converted into meso-diaminopimelic acid and, finally, through the action of LysA1, into l-lysine. We found that mutants for dapA are auxotrophs for l-lysine; however, lysA1 mutants grew in the absence of lysine, suggesting that the lysA-2 gene product might replace LysA-1. Also present within the the KT2440 genome are two copies of dapF (PP5228 and PP3790), which encode a diaminopimelate epimerase. Mutants in these ORFs were found not to be auxotrophs. Although mutants in dapB were not found in the screening, the inactivation of its 5′ end preceding gene (dnaJ) exerted polar effect, as described above, and l-lysine restored its growth supporting the role of DapB in l-lysine biosynthesis.

Although most of the genes needed for lysine biosynthesis in KT2440 are scattered throughout the genome, using RT-PCR assays we found that the lysA2 and dapF genes form an operon. The lysA1 (PP2077) gene is transcribed divergently from PP2078, a LysR family regulator, but whether this regulator influences expression of lysA1 is unknown.

In regards to methionine, threonine and isoleucine synthesis, there exists another metabolic pathway branching off at aspartate-β-semi-aldehyde to yield homoserine, which is the precursor to these amino acids (Fig. S5). The ThrB and ThrC proteins catalyse the conversion of homoserine into threonine. A ThrB mutant was found to be a threonine auxothroph. Three different threonine dehydrates, namely PP3446 (IlvA-1), PP5149 (IlvA-2) and PP4430 (IlvA-3), are present in the chromosome of KT2440, and as a consequence of the redundancy of these paralogous genes, no l-isoleucine mutant was found. The hom gene, which is involved in conversion of aspartate-β-semi-aldehyde to homoserine, is found in an operon with thrC, whereas thrB (PP0121) is transcribed by itself.

Biosynthesis of methionine from homoserine in Pseudomonas putida takes place in three steps (Alaminos and Ramos, 2001). The first step is the acylation of homoserine to yield acyl-l-homoserine. This reaction is catalysed by the products of the metXW genes and is identical to the process as it occurs in enterobacteria, Gram-positive bacteria and fungi, except that in these microorganisms the reaction is catalysed by a single polypeptide (the product of the metA gene in Escherichia coli and the met5 gene product in Neurospora crassa) (Han et al., 2004). A metX mutant is a methionine auxotroph and was found four times in our collection of mutants. In Pseudomonas putida, as in Gram-positive bacteria and certain fungi, the second and third steps consist of a direct sulfhydrylation catalysed by MetZ that converts the O-acyl-l-homoserine into homocysteine (Fig. S5), and further methylation to yield methionine. The latter reaction can be mediated by either of the two methionine synthetases present in the cells.

The metXW genes form an operon as previously shown (Alaminos and Ramos, 2001), whereas metZ, encoding the sulfhydrylase, is present within an operon together with purF (PP2000) (Nelson et al., 2002). One of the B12-dependent methionine synthetases, PP2375, is transcribed divergently from the cti gene (encoding a cis- to trans- isomerization of unsaturated fatty acids), whereas the other methionine syntase (B12-independent) is present in an operon with genes that may not be related to methionine biosynthesis (not shown).

Amino acids made from 3-phosphoglycerate.  In regards to the biosynthesis of serine from 3-phosphoglycerate, two mutants that were serine auxotrophs were isolated. One of these was an serB mutant and the other an serA mutant (Fig. S6). Serine is, in turn, the starting compound for the biosynthesis of glycine, which occurs via a reaction catalysed by GlyA. There are two glyA genes (PP0322 and PP0671) that are present within the KT2440 genome; a mutant for each was available from the PRCC. Both mutants appeared to be able to make glycine from serine confirming their redundancy.

In most microorganisms, the major route for the biosynthesis of cysteine is the sulfate assimilation pathway (Fig. 4). This process involves the uptake and activation of inorganic sulfate followed by its stepwise reduction to sulfide. The overall reaction accounting for cysteine biosynthesis from serine is as follows: l-serine + acetyl-CoA + hydrogen sulfide →l-cysteine + acetate + coenzyme A. Activation of serine takes place via an acetyl-CoA transferase called CysE, while sulfide is made through the activation of sulfate to yield adenosine-5′-phosphosulfate (APS) in a reaction mediated by CysNC and CysD, and the subsequent reduction of APS to give sulfite and sulfide (catalysed by CysH and CysI respectively) (Fig. 4). The final reaction, in which sulfide is condensed with O-acetyl-l-serine, is mediated by two cysteine synthases B (cysM) and cysteine synthaseA (cysK) enzyme (see Fig. 4). The cysteine pathway auxotrophs that we had expected to encounter included one at the step involving the reduction of sulfate to sulfide (CysNC, CysH, CysD, CysI), another at the activation of serine (CysE), but not at the final ligation reaction because of CysM and CysK redundancy. Our screening revealed the existence of cysteine auxotrophs for mutants in CysH and CysI. The cysI gene encodes sulfite reductase and this mutant failed to use sulfate as an S source in agreement with work by Hummerjohann and colleagues (1998). A number of other mutants involved in cysteine biosynthesis showed compromised growth rates. These include a CysNC mutant, which grew slowly in minimal medium, and the CysB mutant, a transcriptional regulator of genes involved in sulfur utilization and sulfonate–sulfur metabolism, was impaired in its ability to grow on cysteine. As expected a mutant in cysM, available at the PRCC, was not a cysteine auxotroph.

Figure 4.

Biosynthesis of cysteine from inorganic sulfur sources and homocysteine. Details are given in the text, as well as the identified genes via isolation of cysteine auxothrophs, blast analysis and feeding experiments.

For some microorganisms it has been proposed that homocysteine can yield cysteine using cystathionine as an intermediate. We tested whether homocysteine could bypass cysteine auxotrophy in the two available mutants, and found that this was the case for CysH and CysI mutants. Homologues to metC (PP0658) and metB (PP0659) were found in the P. putida genome. It should be noted that growth of the metX mutant was not restored by cysteine, suggesting that the route from cysteine to homocysteine does not operate in P. putida in contrast with other microbes (Alaminos and Ramos, 2001).

Biosynthesis of aromatic amino acids

In this study we confirm that auxotrophs requiring tryptophan, phenylalanine and tyrosine can be repetitively isolated. The analysis of these biosynthetic routes was recently described in our group (Molina-Henares et al., 2009), and we refer to this study and to Fig. S7 showing the pathways and locating the knockout mutants that lead to auxotrophy.

Analysis and functional organization of genes involved in vitamin biosynthesis

The biosynthesis of vitamin B12 starts from glutamate yielding uroporphyrinogen III, which in a CobA-mediated reaction is converted to precorrin 2 (Fig. S8). Subsequently, this chemical undergoes a series of well-established steps to successively yield cob(I)yrinate a,c diamine, adenosine-GDP-cobinamide and then vitamin B12 (Fig. S8). Most of the genes encoding enzymes in the pathway have a single copy in the genome; i.e cobO, cobD, cobB, cobC, cobQ, cobN, cobW, cobJ, cobI, cobH, cobL and cobK. Therefore, we would have expected to isolate several mutants requiring B12 for growth. However, we only isolated a single mutant, in which the miniTn5 was inserted in a gene annotated as cobA2, whose gene product seems to be required in the first step of precorrin-2 synthesis (Table 1). Additionally, this result also seemed unexpected since two genes were annotated as cobA. In the PRCC there are mutants available for cobO, cobD, cobQ, cobN, cobW, cobA1, cobJ, cobI and cobL, which we tested for B12 requirement. Our results revealed that none of the mutants exhibits B12 requirement. First, this indicates that cobA1 is not essential for B12 synthesis and, second, that from precorrin-2 a second alternative pathway for B12 biosynthesis must exist.

Analysis of the organization of these genes shows that many cob genes form operons. These include an operon spanning PP4826 to PP4832 encoding genes related to precorrin metabolism, a second operon encompassing PP1672 through to PP1681 containing genes related to cobyric acid synthesis, the PP3507/PP3508 cluster that encodes the cobalt chelase, and an additional operon that contains a cobalamin biosynthesis protein (Fig. S9). The three genes annotated as uroporphyrin-III C-methyltransferase are present as individual ORFs corresponding to PP2090 (CobA-1), PP3999 (CobA-2) and PP0188 (CobA-3). To identify potential alternative genes for the biosynthesis of vitamine B12, we analysed the KEGG database. A set of genes named cbiK, X, H, F, J, C, A are proposed to convert precorrin-2 into cob(I)yrinate a,c deamine, a pathway that operates preferentially in anaerobes. Our search identified cbiD, but no other cbi genes, therefore uncertainties remain on the alternative pathway P. putida KT2440 uses to produce B12.

Biotin is made from pimeloyl-CoA via a pathway consisting of five steps that are catalysed by BioW, BioF, BioA, BioD and BioB respectively (Fig. 5). A single copy each of bioB and bioF exist in the chromosome. Our screen revealed that mutants with knockouts in these genes are biotin auxotrophs. The bioB and bioF genes are part of the bioBFHCD operon, whereas bioA (PP4984) is transcribed on its own. Mutants in bioA and bioH were isolated in our screening and exhibited biotin requirement. The bioD mutant, available at PRCC, grew on minimal medium, indicating the existence of a paralogue for BioD that can convert 7,8-diaminoate into dethiobiotin (Fig. 5).

Figure 5.

Biosynthesis of biotin from pimelate in Pseudomonas putida KT2440. The corresponding genes of the biotin biosynthetic pathway were identified though blast search in different databases. High identity with characterized proteins in other microorganisms served as the basis for the establishment of the pathway. Mutants in bioF, bioH, bioA and bioB were isolated in our in vivo screening and confirmed the mutants required biotin for growth on minimal medium.

Nicotinate, also known as niacine, forms part of essential cofactors in many redox reactions. We identified an auxotroph with a mini-Tn5 insertion at the nadA gene, this mutant required nicotinic acid. NadA (PP1231) is a quinolinate synthetase that catalyses the second step of the de novo biosynthetic pathway of the pyridine nucleotide (Fig. S10). This pathway step involves condensation of dihydroxyacetone phosphate (glycerone phosphate) and iminoaspartate (made from aspartate by NadB) to form quinolic acid (Hüser et al., 2005). The nadA mutant grew on minimal medium when quinolinate was added to the culture medium. Subsequently quinolinate, in a series of reactions mediated by the nadC, nadD (PP4810), nadE (PP4869) and pcnB (PP4868) gene products, yields nicotinic acid. Most of the nad genes are organized as monocystronic units unlinked in the chromosome of KT2440. Failure to isolate mutants in nadB is probably due to the presence of several amino acid oxidases in the genome of KT2440. Further research is in progress to characterize this pathway in greater detail.


The generation of a genome-wide collection of viable single-gene knockouts in P. putida KT2440 combined with the screening of this library in glucose minimal medium is an effective way to identify conditionally essential genes for P. putida. In this study we have identified 48 genes essential for growth in this minimal-medium environment that are not essential in a rich-medium environment. A total of 24 genes identified in the in vivo screening were not predicted by the current in silico models of the strain. Therefore, this approach is useful for the cross-validation of in silico models of this strain providing feedback to gain a deeper understanding of the interplay between metabolic pathways in P. putida. In addition, a number of in silico predicted conditionally essential genes (38) were not found in our screening. To shed light on whether these genes were essential or not, we requested mutants from an alternative collection at the PRCC (Duque et al., 2007). Our results showed that all 10 mutants were able to grow on minimal medium, suggesting that the prediction was not accurate or that the annotation of the KT2440 genome was incomplete. For some of the in silico predicted conditionally essential genes, mutants were not available in either of the mutant collections under screening (see Table 2), so that their status is unconfirmed as of yet. Therefore, a number of discrepancies between the in silico and in vivo results have been identified and we have tried to explain these discrepancies through the analysis of both the physical and transcriptional organization of the genes under dispute. This has allowed us to explain why certain mutants that were predicted as non-viable in minimal medium are indeed viable, and why certain conditionally essential genes were not properly identified by the in silico models.

Among the conditionally essential genes are a number of knockouts that confer auxotrophic character to the strain. The current study provides a wealth of information on Pseudomonas putida KT2440 amino acid and vitamin auxotrophs. This result is, in it self, interesting since the number of auxotrophs in different Pseudomonas sp. strains that have been previously isolated is rather limited (Isaac and Holloway, 1968; Calhoun and Weary, 1969; Buvinger et al., 1981; Cuppels, 1986; Essar et al. 1990Lindow et al., 1993; Ronchel and Ramos, 2001; Rediers et al., 2003). Pseudomonas putida KT2440 auxotrophs, which required arginine, histidine, phenylalanine, lysine, cysteine, isoleucine, valine, tryptophan and leucine, were found as predicted by the iJN746 and iJP815 models. In addition, we found auxotrophs for methionine, proline, tyrosine, threonine and serine, which were not predicted by the models. Methionine, serine, proline, threonine and tyrosine auxotrophy resulted from the inactivation of metX, serB, serA, proC, thrC and tyrA (see Table 1). These genes are in a single copy in the genome of KT2440 and their blast analysis identified them with the expected function. Moreover, feeding experiments further implicated these genes within the corresponding biosynthetic pathways since it was observed that the growth of the metX mutant was restored by methionine; serB and serA mutants growth was restored by serine; proC growth was restored by proline and tyrA growth was restored by tyrosine. Among the 68 genes that were predicted to be conditionally essential genes in minimal medium by in silico models, nine were proposed to lead to histidine requirement when knocked out. However, our screening yielded mutants in only four of the nine genes. The remaining three mutants were not available at the PRCC. The reason for such failures in the in silico predictions remains unexplained, although we speculate that is might be related to the accumulation of toxic-phosphorylated intermediates.

No auxotroph mutant for a limited number of amino acids (Glu, Gln, Asp, Asn, Ala and Gly) was isolated. These missing auxotrophs seem to have resulted from enzyme or pathway redundancy. For aspartate, alanine and aspargine, which are synthesized by the direct transamination of an a-keto acid, the lack of auxotroph isolation seems to be the result of overlapping specificity of the 18 different amino acid transferases encoded in the genome of KT2440 (Table S2). Another example of redundancy is that of glutamine synthetases (up to 8 are annotated in the KT2440 genome). The existence of multiple biosynthetic pathways allows glutamate to be synthesized through both the GS-GOGAT pathway (Caballero et al., 2004) and the glutamate dehydrogenase pathway. In fact, mutants in each of the GOGAT subunits have been previously generated by our group and showed no glutamate requirement.

It has been shown that P. aeruginosa contains a d- to l-arginine racemase, which allows the utilization of d-arginine as a C-source, and which also allows the growth of strains expected to be l-arginine auxotrophs on l-arginine (Li and Lu, 2009). We have also shown that KT2440 can use d-lysine as the sole C-source, which requires racemization of d- to l-lysine (Revelles et al., 2005; 2007). In this study we have tested in all of the generated auxotrophs to see whether the d-isomer bypasses the requirement for the proteinogenic one. We found that d-arginine, d-leucine, d-cysteine and d-proline allowed the growth of all different arg, leu, cys and pro mutants available (Table 4), suggesting that a number of d- to l-racemases are encoded in the genome of KT2440. Our blast analysis has identified four potential racemases involved in the metabolism of amino acids (Table S3), which are currently under study in our laboratory.

The two available in silico models predicted vitamin auxotrophy for pantoneate; however, no such mutant appeared in our screening and when a mutant with a knockout in the panC gene, available at the PRCC, was analysed no auxotrophic trait was associated with the insertion. Instead, we isolated auxotrophs requiring biotin, nicotinic acid and vitamin B12. Biotin and nicotinic acid auxotrophy resulted from the inactivation of single-copy genes in bioF, bioB, bioA and bioH for biotin and nadA in the case of nicotinic acid. For vitamin B12 the current annotation indicated two copies of cobA (cob1/cobA2) and single copy of a number of the cob genes (see Figs S8 and S9). We would have expected mutants requiring B12 in cobO, cobD and other cob genes, but our screen revealed only one B12-requiring mutant, which was knocked out in the cobA2 gene. These results were corroborated by mutants in cobD, cobE and others available at the PRCC, in that they were not found to be auxotrophs. Therefore, this study shows that an alternative pathway for vitamin B12 biosynthesis may exist and that cobA1– being unable to compensate for the loss of cobA2– may not be involved in the conversion of uroporphyrinogen III into precorrin-2, or may not be expressed. The existence of a second set of yet unidentified genes for B12 biosynthesis is fascinating and deserves further study.

In summary, through the comparison of genes required for growth under rich- and minimal-medium conditions, a set of genes enabling growth in limiting environments has been identified. By studying the genes required for growth in glucose minimal medium, we have shown (i) that the inventory of P. putida metabolic genes designated within the current models is not fully comprehensive, at least with respect to the pathways required to support growth on minimal medium; (ii) that only three putative genes of unknown function (PP3850, PP5127 and PP5305) were identified as essential in this phenotyping screen; and (iii) that the comparison of model predictions with results from high-throughput phenotyping data represents a powerful approach to rapidly refine such models, as well as providing critical information that will likely to lead to an enhanced understanding of the biochemical networks within this organism.

Experimental procedures

Bacterial strains, plasmids and culture conditions used in this study

We used P. putida KT2440R, which is a refampicin derivative of P. putida mt-2 (Franklin et al., 1981). All strains were grown at 30°C and shaken at 200 r.p.m. in 100 ml conical flasks with 20 ml M9 minimal medium, and supplemented with goodies solution (Abril et al., 1989) and amino acids (0.2–1 mM) or vitamins (0.3 mM) when indicated (Table 3). Glucose (16 mM) or citrate (25 mM) was used as carbon sources.

Antibiotics were added, when necessary, to the culture medium at the following concentrations (µg ml−1): ampicillin (Ap), 100; chloramphenicol (Cm), 30; kanamycin (Km), 50; rifampicin (Rif), 30.


MiniTn5 (Km) transposon mutagenesis was performed using triparental matings between the recipient (P. putida KT2440R), donor [E. coli CC118(λpir) with pUT-Km] and helper (E. coli HB101 with pRK600). Cultures were incubated overnight in LB with the appropriate antibiotics. After incubation of the recipient at 40°C for 15 min, in order to temporarily inactivate its restriction systems, 0.7 ml of the recipient was mixed with 0.2 ml of the donor cells and 0.1 ml of the helper cells. Cells were collected by centrifugation, suspended in 50 µl fresh LB, spotted on a 0.45 µm filter, and plated on the surface of an LB plate. After 6 h of incubation at 30°C, cells were resuspended in 1 ml 50 mM phosphate buffer, and serial dilutions were plated on M9 minimal medium with Km and Rif, supplemented with 0.6 mM amino acids and 30 µM vitamins. The mutant clones (7760) were ordered in 384-well plates. These plates were supplied with 15% glycerol and kept frozen at −80°C until used for analysis. Two copies of −80°C stocks were generated. One of these copies was thawed for screening purposes, while the other copy was prepared for storage.

DNA techniques

The preparation of plasmid and chromosomal DNA, the digestion of DNA with restriction enzymes (Roche and New England BioLabs), electrophoresis and Southern blotting were done using standard methods (Sambrook et al., 1989; Ausubel et al., 1991). The transposon insertion point in conditionally essential genes was identified by random PCR amplification with Taq polymerase (Amersham Pharmacia), using TNEXT2 (5′-CTTTATTGATTCCATTTTTCACT-3′) and a mixture of random primers followed by direct sequencing with primers TNEXT2 and TNINT (5′-AGGCGatttcagcgaagcac-3′) (Duque et al., 2007). Sequencing was done on an ABIPRISM 310 automated sequencer. Sequences were analysed with Omiga 2.0 software (Oxford Molecular) and compared with data from the P. putida KT2440 genome, obtained from The J. Craig Venter Institute (, and with the GenBank database using blast programs (Altschul et al., 1997).

RNA preparation and RT-PCR

Total RNA from P. putida KT2440 was isolated according to the RNeasy protocol (Qiagen, GmbH). Reverse transcription (RT)-PCR assays were done with the Titan One Tube RT-PCR System. We used the pairs of primers A-1 (5′-CGATGTGCCTGGCGATGAAC-3′) and B-1 (5′-GCCAGTGCGCTTGATCGACT-3′) to co-transcription of mRNA of leuC and leuD, and we used primers B-2 (5′-CTGCTCAACGGTCTGGACGA-3′) C-1 (5′-AACTGGCGTTGGACCACATC-3′) to determine the contiguity of leuD and PP1987, and C-2 (5′-ACGGCAGTATTACCAGATCG-3′) D-1 (5′-TGTCTCGCTCGATCTTGTCC-3′) to determine co-transcription of PP1997 and leuB.

We used the pairs of primers E-1 (5′-GTGGTCGAGCAGCGTATCAA-3′) and F-1 (5′-TTCCTGGGTGAAGCGTGGCA-3′) to determine co-transcription of mRNA of hom and thrC. We used the pairs or primers G-1 (5′-AAGGTGACCTGCTGGCCGTA-3′) and H-1 (5′-TGGATGTGCGCGTGCTGGCT-3′) to determine co-transcription of mRNA of lysA and dapF. We used the pairs of primers J-1 (5′-CCTGATGGTCGAGCAACTGG-3′) and K-1 (5′-TACAGGCTGGCCACGTAAGA-3′) to determine co-transcription of mRNA that contained the hisI and hisE and arg-1 (5′-GTAGCGGTCCCTCGTACTGG-3′) arg-2 (5′-CGACGGTCTGGTAGCACGAC-3′) to determine co-transcription of argA and argE.

Pseudomonas putida microarrays

The genome-wide DNA chip used in this work (printed by Progenika Biopharma) was described in detail previously (Yuste et al., 2006; Duque et al., 2007). It consists of an array of 5539 oligonucleotides (50-mer) spotted in duplicate onto γ-aminosilane-treated slides and covalently linked with UV light and heat. The oligonucleotides represent 5350 of the 5421 predicted ORFs annotated in the P. putida KT2440 genome (Nelson et al., 2002). The chips are also endowed with homogeneity controls consisting of oligonucleotides for the rpoD and rpoN genes spotted at 20 different positions, as well as duplicate negative controls at 203 predefined positions.

For RNA preparation, Pseudomonas putida KT2440 cells were grown in minimal medium with glucose as a carbon source until the early exponential phase was reached (turbidity at 660 nm was about 0.5). Cells from 12 ml culture samples were harvested by centrifugation (7000 g) at 4°C in tubes pre-cooled in liquid nitrogen. Total RNA was isolated with TRI reagent (Ambion; Reference No. 9738), as recommended by the manufacturer, and then subjected to DNase treatment followed by purification with RNeasy columns (Qiagen; Catalogue No. 74104). The RNA concentration was determined spectrophotometrically, and its integrity was assessed by agarose gel electrophoresis.

To prepare fluorescently labelled cDNA, we primed 25 µg of RNA with 7.5 µg of pd(N)6 random hexamers (Amersham; Catalogue No. 27-2166-01). Probes were synthesized at 42°C for 2 h exactly as described before (Duque et al., 2007). Labeling efficiency was checked with a NanoDrop ND1000 spectrophotometer (NanoDrop Technologies). Hybridization conditions were as described before (Yuste et al., 2006; Duque et al., 2007). Images were acquired at 10 µm resolution, and the background-subtracted median spot intensities were determined with GenePix Pro 5.1 image analysis software (Axon Instruments). Signal intensities were normalized by applying the LOWESS intensity-dependent normalization method (Yang et al., 2002) and statistically analysed with Almazen System software (Alma Bioinformatics S.L.). For appropriate statistical analysis of the results, RNA preparations from at least four independent cultures were tested for each strain (Brazma et al., 2001). P-values were calculated with Student's t-test. A particular ORF was considered differentially expressed if (i) the change was at least 1.8-fold and (ii) the P-value was 0.05 or lower.

Bioinformatic analysis

The main sources of information regarding the composition of the metabolic network of Pseudomonas putida KT2440 were various biological databases. Most of the information came from the Kyoto Encyclopedia of Genes and Genomes (KEGG) (Kanehisa et al., 2006), PSEUDOCYC (Romero and Karp, 2003) and SYSTOMONAS (Choi et al., 2007), and biochemical information found in Pseudomonas-specific and biochemical textbooks.


This work was supported by PSYSMO project GEN2006-27750-C5-5-E/SYS and ERANET Pathogenomics projects BIO2008-04419-E/ and CVI-3010. We thank M.M. Fandila and C. Lorente for secretarial assistance and Ben Pakuts for English improvement.