Systems biology of bacterial nitrogen fixation: High-throughput technology and its integrative description with constraint-based modeling
© Resendis-Antonio et al; licensee BioMed Central Ltd. 2011
Received: 10 March 2011
Accepted: 29 July 2011
Published: 29 July 2011
Bacterial nitrogen fixation is the biological process by which atmospheric nitrogen is uptaken by bacteroids located in plant root nodules and converted into ammonium through the enzymatic activity of nitrogenase. In practice, this biological process serves as a natural form of fertilization and its optimization has significant implications in sustainable agricultural programs. Currently, the advent of high-throughput technology supplies with valuable data that contribute to understanding the metabolic activity during bacterial nitrogen fixation. This undertaking is not trivial, and the development of computational methods useful in accomplishing an integrative, descriptive and predictive framework is a crucial issue to decoding the principles that regulated the metabolic activity of this biological process.
In this work we present a systems biology description of the metabolic activity in bacterial nitrogen fixation. This was accomplished by an integrative analysis involving high-throughput data and constraint-based modeling to characterize the metabolic activity in Rhizobium etli bacteroids located at the root nodules of Phaseolus vulgaris ( bean plant). Proteome and transcriptome technologies led us to identify 415 proteins and 689 up-regulated genes that orchestrate this biological process. Taking into account these data, we: 1) extended the metabolic reconstruction reported for R. etli; 2) simulated the metabolic activity during symbiotic nitrogen fixation; and 3) evaluated the in silico results in terms of bacteria phenotype. Notably, constraint-based modeling simulated nitrogen fixation activity in such a way that 76.83% of the enzymes and 69.48% of the genes were experimentally justified. Finally, to further assess the predictive scope of the computational model, gene deletion analysis was carried out on nine metabolic enzymes. Our model concluded that an altered metabolic activity on these enzymes induced different effects in nitrogen fixation, all of these in qualitative agreement with observations made in R. etli and other Rhizobiaceas.
In this work we present a genome scale study of the metabolic activity in bacterial nitrogen fixation. This approach leads us to construct a computational model that serves as a guide for 1) integrating high-throughput data, 2) describing and predicting metabolic activity, and 3) designing experiments to explore the genotype-phenotype relationship in bacterial nitrogen fixation.
Biological nitrogen fixation carried out by Rhizobiaceas represents nearly 70 percent of the entire nitrogen transformation required for maintaining life in our biosphere. Simultaneously, nitrogen fixation driven by these bacteria constitutes an appealing and natural strategy for developing sustainable agricultural programs due to its cost-effectiveness in crop improvement and its more environmentally friendly effects in comparison to those produced by chemical fertilizers . Based on these fundamental and practical issues, the study of bacterial nitrogen fixation is one active line of research that in the post-genomic era demands new paradigms capable of surveying in a systematic fashion the metabolic organization by which this process occurs in nature.
At a molecular level, symbiotic nitrogen fixation arises as a consequence of the coordinated action of a variety of genes, proteins and metabolites that in turn activate signal transduction cascades and transcriptional factors inside bacteroids. At the end of the day, the consequences are the activation and repression of certain metabolic pathways whose end products are required for counteracting the microenvironmental conditions prevailing inside nodules [2–4]. The advent of high-throughput technologies has fostered the genome scale analysis for bacterial nitrogen fixation, and the output data constitute valuable material in deciphering their metabolic organization at different biological layers [5, 6]. Although some significant results have been achieved in interpreting the high-throughput data, their overwhelming numbers and heterogeneous composition represent a challenge for inferring biological knowledge in a coherent and systematic fashion. This challenge is, indeed, a central issue in systems biology, and its solution demands integrative efforts among genome scale data, physiological knowledge and computational modeling [7–11].
With the purpose of contributing to this integrative challenge, in this paper we present a systems biology description in bacterial nitrogen fixation. In particular, it integrates high-throughput technology and flux balance analysis in order to explore the metabolic activity of Rhizobium etli bacteroids while they fix nitrogen in symbiotic association with Phaseolus vulgaris (common bean plant) . To survey the bacterial phenotype and sketch the genetic and metabolic profile during nitrogen fixation, transcriptome and proteome technologies were carried out for R. etli bacteroids selected at 18 days after inoculation with root plants of P. vulgaris (see details in experimental procedure and methods). We selected this interval of time based on experimental knowledge that has indicated it as an average for maximum enzymatic activity of nitrogenase in R etli bacteroids. To identify those genes with a significant role in nitrogen fixation, we accomplished a comparative analysis between the gene expression profile at the nitrogen fixation stage and under free-living conditions in R etli, this last condition mainly defined by succinate and ammonia as carbon and nitrogen sources, respectively (see methods). Simultaneously, the protein profile inside bacteroids was obtained, also at 18 days after plant inoculation. A set of genes with significant participation in bacterial nitrogen fixation was defined by combining those genes differentially expressed in the two physiological conditions--free life and nitrogen fixation-- and those codifying for the proteins detected inside bacteroids. This same set of genes served as our benchmark for extending the metabolic reconstruction for R. etli metabolism (iOR 363) and evaluating the consistency of the metabolic capacities inferred by the in silico analysis . To assess the predictive scope of the model, we qualitatively compared the metabolic activity predicted by constraint-based modeling against that which was deduced from the high-throughput data obtained for R etli. Overall, our study represents a significant effort toward the reconstruction of a systems biology platform for studying metabolic activity in bacterial nitrogen fixation. It is characterized by its capacity to integrate and describe high-throughput data and predict the metabolic mechanism underlying bacterial nitrogen fixation.
High-throughput technology to guide the Metabolic Reconstruction
To give a broader view of the biological activity inside the bacteroid, proteome analysis was conducted for R. etli bacteroids similarly recollected from nodules selected at 18 days after inoculation in root plants of P. vulgaris, see experimental procedure and methods. In total, proteome studies led us to identify and characterize 415 spot proteins in the bacteroids that suggested the expression of 293 genes during nitrogen fixation, see Figure 1 (B) and Additional File 2.
Both technologies--transcriptome and proteome--contributed to supply a broader biological landscape regarding bacterial nitrogen fixation. However, it is necessary to be aware of some differences in the experimental design underlying both technologies in order to integrate and interpret this data in a coherent fashion. While microarray technology resulted from a comparative analysis of two physiological conditions (free life and nitrogen fixation stages), proteome data identified the most abundant proteins present exclusively during nitrogen fixation stages. As Figure 1 (C) and Additional File 3 show, a scarce overlapping between the genes identified by both data sets is observed due to the experimental distinctness inherent in each technology. Thus, in order to identify those genes and enzymes with a relevant role in bacteroid metabolism and, in turn, form a set of genes that serve as a benchmark for computational assessment, we followed an integrative, rather than, selective strategy. Taking into account both sources of data, we hypothesized that up-regulated genes identified by microarray data and those genes that codify for the identified proteins potentially reveal those genes with a major role in nitrogen fixation. Under this assumption, both technologies led us to integrate a total of 948 genes that have a role in supporting bacterial nitrogen fixation, see Figure 1 and Additional File 1 and 2.
Functional classification of this set of genes ranged from enzymes participating in central metabolism and amino acid production to those maintaining specific pathways of nitrogen fixation such as glycogen and poly-β-hydroxybutyrate (PHB) biosynthesis. In addition, we identified enzymes participating in catabolism and anabolism of amino acids, chemotaxis, ribosome composition, RNA polymerase, DNA replication, nucleotide repairs, secretion systems and fatty acids metabolism. Moreover, a significant number of proteins participating as transporters reflects the intense metabolic crosstalk between plant and bacteroid; for instance, proteins participating in transport of small molecules, such as carbon, hydrogen, phosphate and sugar, fall under this classification, see panels (A) and (B) in Figure 1. We also identified proteins participating in the regulatory mechanism in nitrogen fixation, two components systems, transport and cell surface structure, energy transfer, cellular protection, and the transport and synthesis of polysaccharides. An extended discussion of the functional analysis that emerged from both technologies and its implication at a metabolic level can be reviewed in the Additional File 4.
Expanding Rhizobium etli metabolic reconstruction and selecting pathways for its experimental assessment
The data generated by high-throughput technology constitutes a cornerstone in moving toward a descriptive analysis of nitrogen fixation. Despite the fact that this top-down scheme represents a valuable contribution to monitoring cell activity at a genome scale, complementary descriptions are required to integrate these data and survey how genetic perturbations affect nitrogen fixation in a systematic and quantitative fashion (bottom-up scheme). Among these quantitative schemes, constraint-based modeling is an appropriate formalism for exploring the cellular metabolic activity and guiding experiments to improve cellular behavior in a rational, coherent and optimal fashion [7, 8, 19, 20]. In order to construct a bottom-up scheme for bacterial nitrogen fixation, our strategy consisted of three steps: 1) metabolic reconstruction for R. etli; 2) in silico modeling of nitrogen fixation, and 3) a cyclic assessment of computational predictions and experimental results.
In terms of metabolic reconstruction, proteome and transcriptome data were used to elaborate on the previous report for R. etli, thereby making some metabolic improvements and including new metabolic pathways absent in the previous version. To visually identify these metabolic reactions, we proceeded to represent the set of genes identified by high-throughput data and those from iOR363 reconstruction into each metabolic pathways defined in KEGG database. A comparative analysis among each pathway led us to visualize and highlight their differences. Consistent with the previous metabolic reconstruction, certain reactions were identified in the experimental set of data, while others led us to postulate the activity of new metabolic pathways that were absent in the previous reconstruction . Specifically, high-throughput data strongly indicated the biological activity of fatty acid metabolism, and we therefore included this pathway in the metabolic reconstruction, see supplementary material. Overall, a set of 405 reactions and 450 genes made up the new metabolic reconstruction for R. etli (i0R450) with which in silico simulations and analysis were carried out. Topological properties that emerged from the updated metabolic reconstruction are shown in Figure 1 (D).
To evaluate the concordance between the metabolic activity predicted in silico and that interpreted from high-throughput technology, we selected 22 KEGG metabolic pathways  that had the highest number of genes experimentally detected by high-throughput data see Figure 1 (F). According to the KEGG database, these 22 metabolic pathways contain 311 genes for R. etli of which 76.7% were included in the metabolic reconstruction iOR450. This set of genes and their corresponding enzymes constituted the central core for evaluating the coherence between in silico predictions and high-throughput data interpretations. Even though in silico assessment relies on the activity of 22 metabolic pathways, in silico analysis of nitrogen fixation took into account all the reactions included in the metabolic reconstruction. This latter procedure will be valuable especially for exploring and predicting the metabolic role that additional pathways have on nitrogen fixation.
Constraint-based modeling: evaluating the descriptive and predictive capacities of the metabolic reconstruction
where glycogen, lysine, poly-hydroxybutyrate, alanine, aspartate and ammonium are denoted as glycogen[c], lys[c], phb[c], ala[e], asp[e] and nh4[e], respectively. All these metabolites are required to support an effective symbiotic nitrogen fixation , and their spatial location is indicated by [c] and [e] for cytoplasm and external compound. As a result of this simulation, we obtained a consistency coefficient of η Genes = 0.6835 for genes and η Enzymes = 0.702 for enzymes. Notably, this numerical value implied that 68.35% of the genes and 70.2% of the enzymes predicted in silico were consistently identified by high-throughput technology. To evaluate the statistical significance of this correlation, a hypergeometric test was applied in each case. In terms of enzymes, the coefficient reflected that of 74 enzymes predicted in silico, 52 were identified by high-throughput data. Meanwhile, the gene consistency coefficient indicated that of 237 expressed genes, 162 were identified experimentally. In both cases we concluded that these correlations were statistically significant: p-value = 8.59 × 10-35 and p-value = 4.9 × 10-64 for genes and enzymes, respectively.
Improving predictability capacity of constraint-based modeling
where boldface letters indicate those metabolites that were added to the previous objective function. Taking into account this implementation and simulating the flux distribution through FBA as described above, we obtained the following results during nitrogen fixation: η Genes = 0.6948 and η Enzymes = 0.7683, see Figure 1(E). In terms of enzyme activity this numeric value indicates that of the 82 metabolic reactions predicted in silico, 63 of them were consistently justified by high-throughput data (p-val = 3.05 ×10 -64 ). Meanwhile the gene consistency coefficient indicated that of 249 expressed genes, 173 were identified by high-throughput data (p-value = 4.9 ×10-64).
Given this improvement, a detailed comparison between computational predictions and high-throughput data of the 22 metabolic pathways defined in Figure 1 (E) led us to distinguish three possible cases: the presence of 1) genes (enzymes) that were predicted in silico but not detected experimentally, 2) genes (enzymes) that were consistently observed in both schemes, and 3) genes (enzymes) that were experimentally detected but not predicted in silico, see Figure 1 (E). As explained in the methods section, η is related to the fraction of genes (enzymes) that were consistently observed in both schemes and constitutes the backbone of our modeling assessment. However, the biological explanation for the discrepancies described above (in cases 1 and 3) requires feedback assessment between modeling and experiments. For instance, these discrepancies could be reflecting the presence of post-transcriptional and post-translational regulation during nitrogen fixation and the design of proper experiments will be fundamental to discarding or accepting this hypothesis.
Citric acid cycle
Glycolysis, gluconeogenesis and pentose phosphate pathways
A common metabolic trait for some Rhizobiaceas is the intense activity of gluconeogenesis pathway . In agreement with this finding, a significant number of gluconeogenic and glycolytic enzymes were identified by high-throughput technology, and constraint-based modeling consistently concluded that gluconeogenesis pathway was actively participating in nitrogen fixation.
Multiple isoforms of PEP carboxykinase (pckA) were detected by proteome technology, see Additional File 2, mirroring their pivotal role in nitrogen fixation and bacteroid differentiation. Thus, R. etli CE3 pckA mutant produces few nodules into which the infection threads do not appear to penetrate . In qualitative agreement with this report, in silico mutation suggests that pckA is an essential gene for accomplishing nitrogen fixation in R. etli, see Figure 3.
In addition, 6-phosphogluconolactonase (pgl), glucose 6-phosphate dehydrogenase (Zwf1), its chromosomal homolog (designated by zwf2) and one transaldolase (Tal) were detected by proteome, supplying evidence that pentose phosphate pathways can be actively participating in nitrogen fixation. Consistent with this finding, in fast-growing Rhizobiaceas, there is evidence that pentose phosphate and Entner-Doudoroff pathways work in coordinate action as the probable major routes for the metabolism of sugars .
As mentioned before, some glycolytic genes were identified by high-throughput data: two triosephosphate isomerases (TpiAch and TpiAf), one glyceraldehyde 3-phosphate dehydrogenase (Gap), one pyruvate kinase II (PykA), one 2-phosphoglycerate dehydratase (enolase), phosphoglycerate mutase (pgm), and the bisphosphate aldolase (fbaB), see Additional File 1 and 2. Furthermore, there is experimental evidence that the genetic silence of fbaB in R. etli causes the development of sparse, empty nodules on root beans . Consistent with this fact, computational gene deletion analysis carried out with this gene confirms that fbaB has a crucial role in supporting the metabolism of bacterial nitrogen fixation , see Figure 3(B). Even though these findings were not enough to postulate an active glycolytic cycle, they may suggest the metabolism of sugar intermediates via other pathways. For example, the presence of a specific transporter for glycerol-3-phosphate (ugpAch1, induced 3.88-fold by microarray analysis) indicates that this may be an important source for generating glycolytic intermediates. Similarly, the expression of 6-phosphogluconate dehydrogenase (Gnd) suggests the presence of an active pentose pathway, which is another potential channel for the metabolism of glycolytic intermediates.
Myo-inositol catabolic pathway
Myo-inositol is one of the most abundant compounds in the soybean nodule, and accordingly, high-throughput technology successfully detected the presence of myo-inositol 2-dehydrogenase proteins (IdhA and IolB) encoding a myo-inositol protein in catabolism . In agreement with this fact, computational analysis of the metabolism in R. etli suggests that a decrease of myo-inositol inside the nodule can reduce its capacity to fix nitrogen, see Figure 3(B). This result supports the hypothesis that the presence of myo-inositol in the nodule is essential for growth and maturation of the bacteroid and its metabolic inhibition can lead to both a nonfunctional bacteroid and the reduction of nitrogen-fixation activity .
Poly-β-hydroxybutyrate and glycogen accumulation
While most of the bacteroid carbon supplied by the plant is channeled into energy production to fuel nitrogen reduction, in certain types of nodules some carbon is diverted by the bacteroids into the production of intracellular storage polymers composed of either glycogen or poly-β-hydroxybutyrate (PHB). Our simulations produced PHB, and consistent with our predictions, high-throughput analysis led us to identify the presence of three components related to its metabolic pathway: the polymerase PhbC (poly-beta hydroxybutyrate polymerase protein), a putative polyhydroxybutyrate depolymerase protein (detected by transcriptoma, see Additional File 1) and the acetyl-CoA acetyltransferase (beta-ketothiolase, phbAch) detected by proteome. Other reports confirm that metabolic fluxes in PHB and glycogen pathways are such that inhibition of one results in accumulation of the other, a property that was consistently observed by in silico modeling [8, 33, 34]. The precise role of PHB and glycogen during infection, nodulation, and nitrogen fixation and the factors that induce their accumulation are not yet determined. Future experiments dealing with these pathways are necessary to elucidate their role in bacterial nitrogen fixation.
To ensure the proper production of the ammonium required to establish an optimal bacterial-plant symbiosis, constraint-based modeling concludes that central genes involved in nitrogen fixation (nif and fix genes) are required for an optimal activity. Consistent with this fact, NifH, NifD and NifK were identified in proteome data and detected up-regulated in transcriptome analysis. In addition, an up-regulated gene expression was observed for nifE (nitrogenase reductase iron-molibdenum cofactor synthesis truncated protein), nifN (nitrogenase reductase iron-molibdenum cofactor synthesis protein), nifX (iron-molibdenum cofactor processing protein) and nifB (FeMo cofactor biosynthesis).
In R. etli, the iscN gene (Fe-S cofactor nitrogenase synthesis protein) is co-transcribed with nifU and nifS, and in conjunction, these genes were significantly up-regulated in bacteroids in comparison to bacteria under free-life condition (10.82, 3.92 and 1.99-fold, respectively). Furthermore, the iscN mutant in R. etli showed a significant reduction in nitrogen fixation . Consistent with this report, in silico gene deletion analysis of those genes codifying for nitrogenase mostly reduces nitrogen fixation, see Figure 3.
Amino acid metabolism and transport
A previous report suggests that Rhizobiaceas require the availability of 20 amino acids to establish an effective symbiosis with legumes . Some amino acids are synthesized by Rhizobiaceas whereas the remaining are supplied by the host plant, a condition that appears to be plant-type specific. High-throughput analysis led us to identify certain proteins required for the synthesis of arginine, tyrosine, tryptophan, phenylalanine and lysine, the latter participating in the objective function defined in constraint-based modeling. On the other side, from the ABC-transporter proteins founded in nodule bacteria, thirteen were involved in amino acid transport, it strongly suggests that the uptake of amino acid is of particular importance in nitrogen fixation. The general amino acid ABC-transporter protein for AapJ (substrate binding protein) was detected by proteome analysis: the aapJ gene is part of the aapJQMP operon that exists in many Rhizobiaceas and has been described in detail in R. leguminosarum. BraC1 and braC2, of the branched-chain amino acid ABC transporter, were detected in bacteroid by proteome and transcriptome technologies (2.85 fold). In R. leguminosarum braDEFG is required for alanine, histidine, leucine and arginine uptake  (two of which form part of the objective function associated with the metabolism of nitrogen fixation in our in silico model). Alternately, in R. leguminosarum, braC mutants are effective in alanine uptake (but are lacking in the uptake of the other three amino acids) . Phenotype behavior for braC mutants has not been studied in R. etli, but there is evidence that braD and braH mutants were found to be deficient in glutamine uptake and respiration but proficient in nodulation and nitrogen fixation .
Purine and pyrimidine pathways are important during the nodulation processes given that most purine or pyrimidine auxotrophs in Rhizobiaceas are ineffective in symbiotic nitrogen fixation because they elicit pseudo-nodules devoid of infection threads . Thus, for instance, the purB and purH gened in Mesorhizobiumi loti are involved in infection thread formation and nodule development in Lotus japonicus. In addition, purB and purH mutants exhibited purine auxotrophy and nodulation deficiency in L. japonicus. As Figure 2(A) and Additional File 5 panel (C) shows in the supplementary material, constraint-based modeling concludes that some enzymes in purine and pyrimidine pathways are actively participating in reaching an optimal symbiotic nitrogen fixation. Supporting this finding, several key enzymes were identified in bacteroids by proteome technology. Among them, we identified: phosphoribosylamine-glycine ligase protein (PurD), adenylosuccinate lyase protein (PurB), phosphoribosylformylglycinamidine synthetase protein (PurL), adenylosuccinate synthetase protein (PurA), IMP cyclohydrolase/phospho-ribosylaminoimidazole-carboxami-deformyltransferase protein (PurH), adenylate kinase (ATP-AMP transphosphorylase, Adk) and nucleoside-diphosphate-kinase protein (Ndk).
In the presence of adenine, only the purH mutant induced nodule formation, and the purB mutant produced few infection threads, suggesting that 5-aminoimidazole-4-carboxamide ribonucleotide biosynthesis catalyzed by PurB is required for the establishment of symbiosis. In addition, purL mutants in S. fredii HH103 strain does not grow in minimal medium unless the culture is supplemented with thiamin and adenine or an intermediate of purine biosynthesis . Furthermore, gene expression of purC1, phosphoribosylaminoimidazole-succinocarboxamide (SAICAR) synthetase protein, purUch (formyltetrahydrofolate deformylase protein), gmk2 (guanylate kinase (GMP kinase protein) and pyrE (orotate phosphoribosyltransferase protein) were up-regulated inside bacteroids between 2.3 to 6.35 fold. In S. meliloti, nodule development in the case of pyrE/pyrF mutants did not reach the extent observed in the parental strain. These results suggest that some of the intermediates and/or enzymes of the pyrimidine biosynthetic pathway play a key role in bacteroid transformation and nodule development , information that should be taken into account for constructing an improved objective function and ensuring a proper computational description in future analysis.
Fatty acids metabolisms
According to high-throughput data, metabolism of fatty acid can play a significant role in bacterial nitrogen fixation, this being in contrast to the drastic reduction of lipid biosynthesis observed in B. japonicum. Thus, a variety of fab genes and proteins participating in fatty acid biosynthesis were detected by both methodologies (proteome and transcriptome). For instance, we detected by proteome the MccB subunit of methylcrotonyl-CoA carboxylase protein, acyl-CoA thiolase protein (FadA), enoyl-CoA hydratase protein (FadB1), enoyl-[acyl-carrier-protein] reductase (NADH) protein (FabI2) and S-malonyltransferase protein (FadD); and by transcriptome fadB2 was induced 3.09-fold. As these findings suggest, fatty acid metabolism could play an important role in bacteroid metabolism given that it can supply a variety of precursors such as components of the rhizobial membrane, lipopolysaccharides and coenzymes required in signal transduction. As opposed to the process in other Rhizobiaceas where fatty acids can be supplied by the host plant , we supply experimental evidence that bacteroids of R. etli synthesize and metabolize their fatty acids. The assessment of this hypothesis and the biological implications on bacterial nitrogen fixation constitute an avenue to experimentally verify in the future.
In this study we present a systemic metabolic description of bacterial nitrogen fixation carried out by R. etli in symbiosis with P. vulgaris, at present the most complete study made in Rhizobiaceas. Collectively, high-throughput data suggest the following significant clues: 1) R. etli bacteroids are capable of synthesizing several amino acids through integrated carbon and nitrogen metabolisms. In addition, we observe the participation of some minor metabolic pathways such as myo-inositol catabolic pathway, degradation and synthesis of poly-b-hydroxybutyrate and glycogen. 2) Gene expression in bacteroids suggests the presence of a specialized transport system for sugars, proteins and ions. 3) An antioxidant defense mechanism based on peroxiredoxine, regulated by nifA, prevails during nitrogen fixation, as opposed to in free-living condition, where the mechanism is rooted in catalases . 4) R. etli over-expresses genes and enzymes required in fatty acid and nucleic acid metabolism, contrary to other studies in bacteroids. Finally, 5) this study contributes a computational model that serves as a useful framework for integrating data, designing experiments and predicting the phenotype during bacterial nitrogen fixation, see Figure 3.
This systemic and integrative approach constitutes a valuable effort toward a systems biology description of the metabolism in bacterial nitrogen fixation; however, to increase our understanding and predictive accuracy some issues should be addressed in the future. Thus, particular attention should be directed toward those enzymes that were predicted metabolically active in silico but were not detected experimentally, and conversely, those enzymes that were detected experimentally but not in silico, see Figure 1(E). We expect that the study of these differences will be fundamental in postulating, verifying and uncovering mechanisms of regulation, while simultaneously confirming or improving hypotheses derived through in silico predictions.
Notably, even though the simulations have been carried out without a detailed numerical description of the coefficients c i in the objective function--see methods section--we have shown that the in silico model is capable of qualitatively predicting the activity of classic metabolic pathways and successfully describing some phenotype behavior in bacterial nitrogen fixation. Even though this represents a significant advance toward a systems biology description of bacterial nitrogen fixation, some improvements should be addressed in future. For instance, additional metabolites with a biological role in nitrogen fixation should be considered in order to obtain a more proper objective function that contributes to uncovering the role that less known metabolic pathways, such as nucleotides and fatty acid metabolisms, have on this biological process. As described here, these improvements will be guided by high-throughput data and the cyclic crosstalk between model and theory, a needed step in integrating, interpreting and generating biological hypotheses in a more accurate fashion.
Overall our study contributes to establishing the bases toward a systems biology platform capable of integrating high-throughput technology and computational simulation of bacterial nitrogen fixation. In particular, we envision that this metabolic reconstruction for R. etli (iOR450) will contribute to the rational design of optimal experiments that help us understand biological principles and identify those molecular mechanisms in order to improve this biological process, all this from a systems biology perspective.
Bacterial strains, growth conditions
Three-day-old Phaseolus vulgaris cv. Negro Jamapa seedlings were inoculated with R. etli CFN42 strains as previously described by Peralta et al. . After 18 days post-inoculation (dpi), nodules were picked out from the roots, immediately frozen in liquid nitrogen and stored at -70°C until further use. Bacteria were isolated from nodules and their identities verified by their antibiotic resistances.
RNA isolation and microarray hybridization
Microarray experiments were carried out using three independently isolated RNA preparations from independent cultures and set of plants. Approximately 3 g of nodules were immersed in liquid nitrogen and macerated. Total RNA was isolated by acid hot-phenol extraction as described previously by de Vries et al. For microaerobic free-living conditions, 50 ml of bacterial cell cultures were collected and total RNA isolated using a RNeasy Mini Kit (QIAGEN, Hilden, Germany). RNA concentration was determined by measuring the absorbance at 260 nm. The integrity of RNA was determined by running samples on a 1.3% agarose gel. 10 μg of RNA was differentially labeled with Cy3-dCTP and Cy5-dCTP using a CyScribe First-Strand cDNA labeling kit (Amersham Biosciences). Pairs of Cy3- and Cy5-labeled cDNA samples were mixed and hybridized to a Rhizobium_etli_CFN42_6051_v1.0 DNA microarray as described by Hegde et al. [48, 49]. After washing, the arrays were scanned using a pixel size of 10 μm with a Scan Array Lite microarray scanner (Perkin-Elmer, Boston, MA). Three biological replicates with one dye swap were performed. We used real-time quantitative PCR to provide an independent analysis of gene expression for selected genes. Primer sequences and additional experimental protocols are reported in the supplementary material section.
DNA microarray analysis
Spot detection, mean signals, mean local background intensities, image segmentation, and signal quantification were determined for the microarray images using the Array-Pro Analyzer 4.0 software (Media Cybernetics, L.P). Statistical treatment of microarray data was accomplished with bioconductors software (http://www.bioconductor.org/). Specifically, microarray normalization was carried out by applying the maNormMain function in the marray library. MA-plots before and after normalization are depicted in Additional File 5. Having normalized the gene expressions in the three experimental replicates, differentially expressed genes were identified by the following procedure. First, we calculate the average log-ration for each gene obtained from the three experimental replicates. Then, we obtained the standardized z-score of the log-ratio associated to each gene. The set of genes differentially expressed during nitrogen fixation was selected as those genes with a z-score higher than 1.65, see Additional File 5. The complete dataset used in the transcriptome analysis can be downloaded from GEO (http://www.ncbi.nlm.nih.gov/geo) with accession numbers: GPL10081 for Rhizobium etli platform and GSE21638 for free-life and symbiosis data.
Verification by RT-PCR
We used real-time quantitative PCR to provide an independent assessment of gene expression for selected genes. The cDNA used for microarrays or freshly prepared cDNA was used as a template for Real-time PCR. Primer sequences used were as follows: fabI2-RECH000938f (5'-GTA TTG CCA AGG CCA TTC AT-3'), fabI2-RECH000938r (5'-CCC ACA GTT TTT CGA CGT TT-3') for the fabI2 gene. idhA-RECH003170f (5'-TTT CTT CAT GAC CCG CTA CA-3'), idhA-RECH003170r (5'-TTG ATC AGC TTG CCT TCC TT-3') for the idhA gene. ppK-RECH001491f (5'-TCC TGG CAC TGA ACA CTC TG-3'), ppK-RECH001491r (5'-GAG AAG GAA CTG GAC CAC CA- 3') for the ppK gene. hisD-RECH000581f 5'GAT CTG AAG CAA GCC ATT CC 3', hisD-RECH000581r (5'-ACA TAA TCG CCG ATG ACC TC-3') for the hisD gene. nifH-REPD00202f (5'-CCT CGG GCA GAA GAT CCT GA-3'), nifH-REPD00202r (5'-CAT CGC CGA GCA CGT CAT AG-3') for the nifH gene. fixA-REPD00224f (5'-ACA TCA ATG GGC GCG AGA TT-3'), fixA-REPD00224r (5'-TGT CGA TCT GCT CCG CCT TT-3') for the fixA gene. cpxP2-REPD00252f (5'-TCC GTG CCA TTT CAA AGA CC-3'), cpxP2-REPD00252f (5'-CCG CCA AAT GAG AAG ATT GC-3') for the cpxP2 gene. hisC-RE1SP0000233f (5'- CGA TGG CGA GAC AGC TAA AT-3'), hisC-RE1SP0000233r (5'-ATC ATC GCA ACG CTA TCT CC-3') for the hisCd gene. Each reaction contained 12.5 μl SYBR green PCR mastermix (Applied Biosystems), 3.5 μl H2O, forward and reverse primers in a volume of 5 μl, and template in a volume of 4 μl. PCR reactions were run with the ABI Prism 7700 sequence detection system (Applied Biosystems) using the following steps: 50°C for 2 min, 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The dissociation protocol was 95°C for 15 s, 60°C for 20 s, followed by ramp from 60°C to 95°C for 20 min. The transcript of the histidinol phosphate aminotransferase protein (hisCd) was used as an internal (unregulated) reference for relative quantification. This gene was selected as a reference because its expression is constitutive in all tested conditions (free live and symbiosis). Results of RT-PCR in real time were analyzed using the ΔΔCT method  and the data was presented like relative expression. All reactions were done by triplicate.
Bacteroids purification, protein extraction and two dimensional gel electrophoresis were done as previously described in . Briefly, bacteroids were purified from root nodules by centrifugation through self-generated Percoll gradients. Bacteroid proteins were obtained by sonication at 24 kHz 1 min ON/1 min OFF for 5 cycles at 4°C in a Vibra Cell (Sonics, USA) in the presence of a protease inhibitor (Complete tablets, Roche Diagnostics GmbH, Mannheim, Germany). To further limit proteolysis, protein isolation was performed using phenol extraction. Two dimensional gel electrophoresis (2D-PAGE), was performed like previously described. Gels were stained with Coomasie Blue G-250, scanned with PDI image analysis system, and analyzed with PD-Quest software (Bio-Rad Laboratories, Inc, Hercules, CA.). Selected spots from preparative 2-D gels were excised, digested and the proteins were identified by PMF MALDI-TOF using a Bruker Daltonics Autoflex, following the same methodology mentioned in . The experiments were performed three times. Selected spots from Coomassie stained preparative 2-D gels were excised and processing automatically using the Proteineer SP spot picker and DP digestion robots (Bruker Daltonics, Billerica MA). Mass spectra were obtained using a Bruker Daltonics Autoflex (Bruker Daltonics Bellerica, Mass. USA) operated in the delayed extraction and reflectron mode. Spectra was externally calibrated using a peptide calibration standard (Bruker Daltonics 206095). Peak lists of the tryptic peptide masses were generated using FlexAnalysis1.2vSD1Patch2 (Bruker Daltonics). The search engine MASCOT server 2.0 was used to compare fingerprints against Rhizobium etli CFN42, NC_007761.1, pA, NC_007762.1, pB, NC_007763.1, pC, NC_007764.1, pD, NC_004041.2, pE, NC_007765.1, pF, NC_007766.1 with the following parameters: one missed cleavage allowed, carbamidomethyl cysteine as the fixed modification and oxidation of methionine as the variable modification. We accepted those proteins with scores greater than 50 and a p < 0.05. Proteome data associated with this manuscript can be downloaded from http://ProteomeCommons.org Tranche using the following hash:
BY/eCcVjwTWN1+m+2ArvJ0QVnesGx5Ekgd4wUOASACfm/ueNl7YI3iLf4xz0lnGsepV5LkpMWOQOrZtjYExlNpQkIBcAAAAAAAABjA = =
High-throughput technology and its use for extending metabolic reconstruction and simulating nitrogen fixation
With the purpose of establishing an integrative description between modeling and experimental data, we extended the metabolic reconstruction for R. etli by including those reactions whose enzyme activity were supported by high-throughput data. Thus, the fatty acids metabolism was included in the metabolic reconstruction, and some metabolic improvements were made along the network. Additional File 6 enlist the main abbreviations used along this paper. Additional File 7 in supplementary material contains a detailed description of the reactions included in this new metabolic version (iOR450). Overall, the updated metabolic reconstruction for R.etli consists of 377 metabolites and 450 genes codifying for enzymes participating in 405 metabolic reactions. The gene-protein reaction association for the entire metabolic reconstruction, lower and upper bounds and reversibility information associated to each reaction are shown in Additional File 7.
where S i,j represents the entries of the stoichiometric matrix, v j is the metabolic flux of the j-th reaction and α j and β j account for thermodynamic and enzymatic constraints, see Additional File 7. Linear programming was carried out using the Tomlab optimization package called from COBRA toolbox in Matlab.
External metabolites considered for flux balance analysis
In order to explore the phenotype capacities of the bacteria metabolism, we included in the reconstruction certain exchange and sink reactions for limiting our metabolic modeling and representing the microenvironmental conditions in the plant nodules. In general, these can be classified as one of two categories. Class I includes those metabolites that can be exchanged between the bacteroid membrane and the plant environment. Among them, we included carbon dioxide (CO2), water (H2O), oxygen (O2), malate (mal-L) and glutamate (glu-L). In addition, exchange reactions for nitrogen (n2), alanine (ala-L), aspartate (asp-L), succinate and ammonium (NH4) were included in the reconstruction for representing their possible bidirectional exchange from plant to bacteroids. On the other hand, metabolites in class II include those that contribute to the defining of internal frontiers in the bacteroids. Importantly, these sink reactions were included as a representation of metabolites originating from metabolic processes currently absent in the metabolic reconstruction. Thus, phosphate (pi), myo-inositol (inost), L-histidinol phosphate (hisp), palmitoyl-CoA (pmtcoa), dodecanoyl-CoA (dodecoa), decanoyl-CoA (decoa), octanoyl-CoA (otcoa) and hydrogen (h) fall in this classification.
Definition of consistency coefficient
Both ratios range from zero to one and constitute our central parameter to assess and quantify the degree of coherence between constraint-based modeling and experimental data.
In silico gene deletion analysis
Computational gene deletion analysis was used to quantify the effects that gene silencing has in supporting bacterial nitrogen fixation. Thus, once the gene to be switched off was selected, we identified its gene-protein reaction association and selected as zero its upper and lower bound in flux activity. Having made this adjustment, we applied flux balance analysis and obtained the new resulting objective function. In order to quantify the participation of this metabolic reaction in bacterial nitrogen fixation, we calculated the percentage of reduced activity of the mutated strain in comparison to the wild type, see Figure 3.
The authors thank Prof Jaime Mora Head of the Program of Functional Genomics of Prokaryotes at the Center of Genomic Sciences-UNAM for his support and comments. OR-A also thanks to Prof. B.Ø. Palsson for his guiding and encouraging support during the progress of this study. We thank Oliver Castillo, J. L. Zitlalpopoca, and Hadau Sánchez for plant experiments and greenhouse support, and María del Carmen Vargas for technical assistance. Finally, the authors appreciate the valuable comments and suggestions from two anonymous referees during the review process. This work was supported by combined grants from National Council of Science and Technology CONACyT-Mexico, grants 83461 (OR-A) and 60641 (SE), and from PAPIIT-DGAPA-UNAM through grants IN222707 (SE) and IN203809-3 (OR-A).
- Diaz RJ, Rosenberg R: Spreading dead zones and consequences for marine ecosystems. Science. 2008, 321: 926-929. 10.1126/science.1156401View ArticlePubMedGoogle Scholar
- Deakin WJ, Broughton WJ: Symbiotic use of pathogenic strategies: rhizobial protein secretion systems. Nat Rev Microbiol. 2009, 7: 312-320.PubMedGoogle Scholar
- Lodwig E, Poole P: Metabolism of Rhizobium bacteroids. Critical Reviews in Plant Sciences. 2003, 22: 37-78. 10.1080/713610850.View ArticleGoogle Scholar
- Prell J, Poole P: Metabolic changes of rhizobia in legume nodules. Trends Microbiol. 2006, 14: 161-168. 10.1016/j.tim.2006.02.005View ArticlePubMedGoogle Scholar
- Sarma AD, Emerich DW: Global protein expression pattern of Bradyrhizobium japonicum bacteroids: a prelude to functional proteomics. Proteomics. 2005, 5: 4170-4184. 10.1002/pmic.200401296View ArticlePubMedGoogle Scholar
- Oehrle NW, Sarma AD, Waters JK, Emerich DW: Proteomic analysis of soybean nodule cytosol. Phytochemistry. 2008, 69: 2426-2438. 10.1016/j.phytochem.2008.07.004View ArticlePubMedGoogle Scholar
- Feist AM, Palsson BO: The growing scope of applications of genome-scale metabolic reconstructions using Escherichia coli. Nat Biotechnol. 2008, 26: 659-667. 10.1038/nbt1401PubMed CentralView ArticlePubMedGoogle Scholar
- Resendis-Antonio O, Reed JL, Encarnacion S, Collado-Vides J, Palsson BO: Metabolic reconstruction and modeling of nitrogen fixation in Rhizobium etli. PLoS Comput Biol. 2007, 3: 1887-1895.View ArticlePubMedGoogle Scholar
- Zhang Y, Thiele I, Weekes D, Li Z, Jaroszewski L, Ginalski K, Deacon AM, Wooley J, Lesley SA, Wilson IA, Palsson B, Osterman A, Godzik A: Three-dimensional structural view of the central metabolic network of Thermotoga maritima. Science. 2009, 325: 1544-1549. 10.1126/science.1174671PubMed CentralView ArticlePubMedGoogle Scholar
- Covert MW, Knight EM, Reed JL, Herrgard MJ, Palsson BO: Integrating high-throughput and computational data elucidates bacterial networks. Nature. 2004, 429: 92-96. 10.1038/nature02456View ArticlePubMedGoogle Scholar
- Gonzalez V, Santamaria RI, Bustos P, Hernandez-Gonzalez I, Medrano-Soto A, Moreno-Hagelsieb G, Janga SC, Ramirez MA, Jimenez-Jacinto V, Collado-Vides J, Davila G: The partitioned Rhizobium etli genome: genetic and metabolic redundancy in seven interacting replicons. Proc Natl Acad Sci USA. 2006, 103: 3834-3839. 10.1073/pnas.0508502103PubMed CentralView ArticlePubMedGoogle Scholar
- Dixon R, Kahn D: Genetic regulation of biological nitrogen fixation. Nat Rev Microbiol. 2004, 2: 621-631. 10.1038/nrmicro954View ArticlePubMedGoogle Scholar
- Tatusov RL, Koonin EV, Lipman DJ: A genomic perspective on protein families. Science. 1997, 278: 631-637. 10.1126/science.278.5338.631View ArticlePubMedGoogle Scholar
- Tatusov RL, Fedorova ND, Jackson JD, Jacobs AR, Kiryutin B, Koonin EV, Krylov DM, Mazumder R, Mekhedov SL, Nikolskaya AN, Rao BS, Smirnov S, Sverdlov AV, Vasudevan S, Wolf YI, Yin JJ, Natale DA: The COG database: an updated version includes eukaryotes. BMC Bioinformatics. 2003, 4: 41- 10.1186/1471-2105-4-41PubMed CentralView ArticlePubMedGoogle Scholar
- Barnett MJ, Toman CJ, Fisher RF, Long SR: A dual-genome Symbiosis Chip for coordinate study of signal exchange and development in a prokaryote-host interaction. Proc Natl Acad Sci USA. 2004, 101: 16636-16641. 10.1073/pnas.0407269101PubMed CentralView ArticlePubMedGoogle Scholar
- Becker A, Berges H, Krol E, Bruand C, Ruberg S, Capela D, Lauber E, Meilhoc E, Ampe F, de Bruijn FJ, Fourment J, Francez-Charlot A, Kahn D, Küster H, Liebe C, Pühler A, Weidner S, Batut J: Global changes in gene expression in Sinorhizobium meliloti 1021 under microoxic and symbiotic conditions. Mol Plant Microbe Interact. 2004, 17: 292-303. 10.1094/MPMI.2004.17.3.292View ArticlePubMedGoogle Scholar
- Uchiumi T, Ohwada T, Itakura M, Mitsui H, Nukui N, Dawadi P, Kaneko T, Tabata S, Yokoyama T, Tejima K, Saeki K, Omori H, Hayashi M, Maekawa T, Sriprang R, Murooka Y, Tajima S, Simomura K, Nomura M, Suzuki A, Shimoda Y, Sioya K, Abe M, Minamisawa K: Expression islands clustered on the symbiosis island of the Mesorhizobium loti genome. J Bacteriol. 2004, 186: 2439-2448. 10.1128/JB.186.8.2439-2448.2004PubMed CentralView ArticlePubMedGoogle Scholar
- Encarnacion S, Hernandez M, Martinez-Batallar G, Contreras S, Vargas Mdel C, Mora J: Comparative proteomics using 2-D gel electrophoresis and mass spectrometry as tools to dissect stimulons and regulons in bacteria with sequenced or partially sequenced genomes. Biol Proced Online. 2005, 7: 117-135. 10.1251/bpo110PubMed CentralView ArticlePubMedGoogle Scholar
- Feist AM, Herrgard MJ, Thiele I, Reed JL, Palsson BO: Reconstruction of biochemical networks in microorganisms. Nat Rev Microbiol. 2009, 7: 129-143.PubMed CentralView ArticlePubMedGoogle Scholar
- Hyduke DR, Palsson BO: Towards genome-scale signalling-network reconstructions. Nat Rev Genet. 11: 297-307.Google Scholar
- Kanehisa M, Goto S: KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000, 28: 27-30. 10.1093/nar/28.1.27PubMed CentralView ArticlePubMedGoogle Scholar
- Segre D, Vitkup D, Church GM: Analysis of optimality in natural and perturbed metabolic networks. Proc Natl Acad Sci USA. 2002, 99: 15112-15117. 10.1073/pnas.232349399PubMed CentralView ArticlePubMedGoogle Scholar
- de las Nieves, Peltzer M, Roques N, Poinsot V, Aguilar OM, Batut J, Capela D: Auxotrophy accounts for nodulation defect of most Sinorhizobium meliloti mutants in the branched-chain amino acid biosynthesis pathway. Mol Plant Microbe Interact. 2008, 21: 1232-1241. 10.1094/MPMI-21-9-1232View ArticleGoogle Scholar
- Mahadevan R, Schilling CH: The effects of alternate optimal solutions in constraint-based genome-scale metabolic models. Metab Eng. 2003, 5: 264-276. 10.1016/j.ymben.2003.09.002View ArticlePubMedGoogle Scholar
- Becker SA, Feist AM, Mo ML, Hannum G, Palsson BO, Herrgard MJ: Quantitative prediction of cellular metabolism with constraint-based models: the COBRA Toolbox. Nat Protoc. 2007, 2: 727-738. 10.1038/nprot.2007.99View ArticlePubMedGoogle Scholar
- Dunn MF: Tricarboxylic acid cycle and anaplerotic enzymes in rhizobia. FEMS Microbiol Rev. 1998, 22: 105-123. 10.1111/j.1574-6976.1998.tb00363.xView ArticlePubMedGoogle Scholar
- Thony-Meyer L, Kunzler P: The Bradyrhizobium japonicum aconitase gene (acnA) is important for free-living growth but not for an effective root nodule symbiosis. J Bacteriol. 1996, 178: 6166-6172.PubMed CentralPubMedGoogle Scholar
- McDermott TR, Kahn ML: Cloning and mutagenesis of the Rhizobium meliloti isocitrate dehydrogenase gene. J Bacteriol. 1992, 174: 4790-4797.PubMed CentralPubMedGoogle Scholar
- Soto MJ, Sanjuan J, Olivares J: The disruption of a gene encoding a putative arylesterase impairs pyruvate dehydrogenase complex activity and nitrogen fixation in Sinorhizobium meliloti. Mol Plant Microbe Interact. 2001, 14: 811-815. 10.1094/MPMI.2001.14.6.811View ArticlePubMedGoogle Scholar
- Tate R, Ferraioli S, Filosa S, Cermola M, Riccio A, Iaccarino M, Patriarca EJ: Glutamine utilization by Rhizobium etli. Mol Plant Microbe Interact. 2004, 17: 720-728. 10.1094/MPMI.2004.17.7.720View ArticlePubMedGoogle Scholar
- Romanov VI, Hernandez-Lucas I, Martinez-Romero E: Carbon Metabolism Enzymes of Rhizobium tropici Cultures and Bacteroids. Appl Environ Microbiol. 1994, 60: 2339-2342.PubMed CentralPubMedGoogle Scholar
- Jiang G, Krishnan AH, Kim YW, Wacek TJ, Krishnan HB: A functional myo-inositol dehydrogenase gene is required for efficient nitrogen fixation and competitiveness of Sinorhizobium fredii USDA191 to nodulate soybean (Glycine max [L.] Merr.). J Bacteriol. 2001, 183: 2595-2604. 10.1128/JB.183.8.2595-2604.2001PubMed CentralView ArticlePubMedGoogle Scholar
- Cevallos MA, Encarnacion S, Leija A, Mora Y, Mora J: Genetic and physiological characterization of a Rhizobium etli mutant strain unable to synthesize poly-beta-hydroxybutyrate. J Bacteriol. 1996, 178: 1646-1654.PubMed CentralPubMedGoogle Scholar
- Encarnacion S, del Carmen Vargas M, Dunn MF, Davalos A, Mendoza G, Mora Y, Mora J: AniA regulates reserve polymer accumulation and global protein expression in Rhizobium etli. J Bacteriol. 2002, 184: 2287-2295. 10.1128/JB.184.8.2287-2295.2002PubMed CentralView ArticlePubMedGoogle Scholar
- Dombrecht B, Tesfay MZ, Verreth C, Heusdens C, Napoles MC, Vanderleyden J, Michiels J: The Rhizobium etli gene iscN is highly expressed in bacteroids and required for nitrogen fixation. Mol Genet Genomics. 2002, 267: 820-828. 10.1007/s00438-002-0715-0View ArticlePubMedGoogle Scholar
- Randhawa GS, Hassani R: Role of rhizobial biosynthetic pathways of amino acids, nucleotide bases and vitamins in symbiosis. Indian J Exp Biol. 2002, 40: 755-764.PubMedGoogle Scholar
- Walshaw DL, Poole PS: The general L-amino acid permease of Rhizobium leguminosarum is an ABC uptake system that also influences efflux of solutes. Mol Microbiol. 1996, 21: 1239-1252. 10.1046/j.1365-2958.1996.00078.xView ArticlePubMedGoogle Scholar
- Hosie AH, Allaway D, Galloway CS, Dunsby HA, Poole PS: Rhizobium leguminosarum has a second general amino acid permease with unusually broad substrate specificity and high similarity to branched-chain amino acid transporters (Bra/LIV) of the ABC family. J Bacteriol. 2002, 184: 4071-4080. 10.1128/JB.184.15.4071-4080.2002PubMed CentralView ArticlePubMedGoogle Scholar
- Newman JD, Diebold RJ, Schultz BW, Noel KD: Infection of soybean and pea nodules by Rhizobium spp. purine auxotrophs in the presence of 5-aminoimidazole-4-carboxamide riboside. J Bacteriol. 1994, 176: 3286-3294.PubMed CentralPubMedGoogle Scholar
- Okazaki S, Hattori Y, Saeki K: The Mesorhizobium loti purB gene is involved in infection thread formation and nodule development in Lotus japonicus. J Bacteriol. 2007, 189: 8347-8352. 10.1128/JB.00788-07PubMed CentralView ArticlePubMedGoogle Scholar
- Buendia-Claveria AM, Moussaid A, Ollero FJ, Vinardell JM, Torres A, Moreno J, Gil-Serrano AM, Rodriguez-Carvajal MA, Tejero-Mateo P, Peart JL, Brewin NJ, Ruiz-Sainz JE: A purL mutant of Sinorhizobium fredii HH103 is symbiotically defective and altered in its lipopolysaccharide. Microbiology. 2003, 149: 1807-1818. 10.1099/mic.0.26099-0View ArticlePubMedGoogle Scholar
- Vineetha KE, Vij N, Prasad CK, Hassani R, Randhawa GS: Ultrastructural studies on nodules induced by pyrimidine auxotrophs of Sinorhizobium meliloti. Indian J Exp Biol. 2001, 39: 371-377.PubMedGoogle Scholar
- Sarma AD, Emerich DW: A comparative proteomic evaluation of culture grown vs nodule isolated Bradyrhizobium japonicum. Proteomics. 2006, 6: 3008-3028. 10.1002/pmic.200500783View ArticlePubMedGoogle Scholar
- Salazar E, Diaz-Mejia JJ, Moreno-Hagelsieb G, Martinez-Batallar G, Mora Y, Mora J, Encarnacion S: Characterization of the NifA-RpoN regulon in Rhizobium etli in free life and in symbiosis with Phaseolus vulgaris. Appl Environ Microbiol. 2010, 76: 4510-4520. 10.1128/AEM.02007-09PubMed CentralView ArticlePubMedGoogle Scholar
- Encarnacion S, Dunn M, Willms K, Mora J: Fermentative and aerobic metabolism in Rhizobium etli. J Bacteriol. 1995, 177: 3058-3066.PubMed CentralPubMedGoogle Scholar
- Peralta H, Mora Y, Salazar E, Encarnacion S, Palacios R, Mora J: Engineering the nifH promoter region and abolishing poly-beta-hydroxybutyrate accumulation in Rhizobium etli enhance nitrogen fixation in symbiosis with Phaseolus vulgaris. Appl Environ Microbiol. 2004, 70: 3272-3281. 10.1128/AEM.70.6.3272-3281.2004PubMed CentralView ArticlePubMedGoogle Scholar
- Vries S, Hoge H, Bisseling : Isolation of total and polysomal RNA from plant tissues. Plant Molecular Biology Manual. 1988, 6: 1-13.Google Scholar
- Hegde P, Qi R, Abernathy K, Gay C, Dharap S, Gaspard R, Hughes JE, Snesrud E, Lee N, Quackenbush J: A concise guide to cDNA microarray analysis. Biotechniques. 2000, 29: 548-550. 552-544, 556 passim,PubMedGoogle Scholar
- Livak KJ, Schmittgen TD: Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods. 2001, 25: 402-408. 10.1006/meth.2001.1262View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.