Anomalies of the anaerobic tricarboxylic acid cycle in Shewanella oneidensis revealed by Tn-seq


For correspondence. E-mail; Tel. 612 626 6496; Fax 612 625 5780.


The availability of increasingly inexpensive sequencing combined with an ever-expanding molecular biology toolbox has transported classical bacterial genetics into the 21st century. Whole genome genetic fitness analysis using transposon mutagenesis combined with next-generation high-throughput sequencing (Tn-seq) promises to revolutionize systems level analysis of microbial metabolism. Tn-seq measures the frequency of actual members of a heterogeneous mutant pool undergoing purifying selection to determine the contribution of every non-essential gene in the genome to the fitness of an organism under a given condition. Here we use Tn-seq to assess gene function in the Gram negative γ-proteobacterium Shewanella oneidensis strain MR-1. In addition to being a model environmental organism, there is considerable interest in using S. oneidensis as a platform organism for bioremediation and biotechnology, necessitating a complete understanding of the metabolic pathways that may be utilized. Our analysis reveals unique aspects of S. oneidensis metabolism overlooked by over 30 years of classical genetic and systems level analysis. We report the utilization of an alternative citrate synthase and describe a dynamic branching of the S. oneidensis anaerobic tricarboxylic acid cycle, unreported in any other organism, which may be a widespread strategy for microbes adept at dissipating reducing equivalents via anaerobic respiration.


Microorganisms possess unparalleled metabolic diversity and play a dominant role in the web of life. A complete understanding of microbial metabolism is critical not only for understanding biology, but for the discovery, optimization and industrialization of useful metabolic pathways for clean renewable sources of industrial and pharmaceutical chemicals as well as biofuels. To expand our knowledge of microbial metabolism and harness novel organisms for biotechnology there is a need to quickly and accurately define metabolic processes. Traditional genetic screens based on growth phenotypes are powerful genetic tools but have a number of inherent limitations. Manipulation and evaluation of individual colonies is tedious and impractical for many microorganisms and growth conditions. Utilization of transposon mutagenesis coupled with Illumina sequencing technology allows the measurement of genetic fitness on a genome wide scale through massively parallel sequencing of transposon–genome junctions (Tn-seq) (Gawronski et al., 2009; Goodman et al., 2009; Langridge et al., 2009; van Opijnen et al., 2009; Gallagher et al., 2011). Variations of Tn-seq have been performed but the basic experimental outline is the same. Briefly, a saturating transposon library is generated in an organism of interest, mutants are pooled en masse, and the resulting population is subjected to a growth condition. Genomic DNA from the parent population (t 1) as well as the selected populations (t 2, t 3, etc.) is isolated and the location and frequency of each transposon insertion is determined. The change in frequency of each insertion mutant within the population is calculated, reflective of the effect the insertion has on strain fitness. The fitness for each gene in an entire genome for any given condition can therefore be determined in a number of days using standard protocols available to any molecular biology lab. Traditionally, whole genome analyses have been performed using microarray technology or, more recently, TagModule mutagenesis coupled with microarray technology (Warner et al., 2010; Deutschbauer et al., 2011). Tn-seq is a direct measure of the relative proportion of transposon mutants before and after selection, eliminating a number of inherent assumptions and/or biases introduced by other methods. Importantly Tn-seq requires only a genome sequence and the ability to deliver a transposon randomly into the genome, greatly expanding the range of organisms in which detailed systems level genetics can be performed.

Shewanella oneidensis strain MR-1 (MR-1) is a Gram negative γ-proteobacterium capable of reducing, among many other substrates, insoluble metals and electrodes (Nealson and Scott, 2006). MR-1 has served as a model environmental organism and there is growing interest in the use of Shewanella sp. for specific applications ranging from bioremediation to synthetic biology (Hau and Gralnick, 2007). A wealth of systems level analysis has been performed using MR-1; however, validation of the models produced has been limited to a handful of targeted gene deletions (Tang et al., 2007b; Fredrickson et al., 2008; Pinchuk et al., 2010; Deutschbauer et al., 2011; Flynn et al., 2012). Here we report the use of Tn-seq to determine the effect each gene in the MR-1 genome has on fitness in oxic or anoxic liquid culture, closely examining genes involved in carbon metabolism. The phenotypes explored in this manuscript were unexpected, deviating from canonical models of bacterial central metabolism, and therefore could not have been predicted using modelling alone. MR-1 uses an alternative citrate synthase with significant activity when grown on a number of carbon sources. We also characterize a dynamic branching of the anaerobic tricarboxylic acid (TCA) cycle that has not been described for any other organism to date.


Library selection and analysis of sequencing data

In order to assess the fitness of insertion mutants en masse, a library of ∼ 50 000 transposon mutants was generated on oxic Shewanella basal medium (SBM) (Covington et al., 2010) agar plates with 40 mM lactate and 80 mM fumarate, and all subsequent growth experiments were performed in this medium unless otherwise noted. Following transposition, the entire mutant library was pooled, aliquoted and frozen. To perform a selection experiment, a library aliquot was thawed and genomic DNA was extracted from a portion of the culture. The remaining culture was back diluted to a final optical density (OD600) of 0.005 and grown shaking at 30°C either with oxygen or anaerobically with fumarate. When culture OD600 reached 0.2, genomic DNA was extracted resulting in populations under selection for ∼ 5–6 generations. We deliberately chose an OD600 of 0.2 because under both oxic and anoxic conditions cultures still exhibit exponential growth and are not limited for carbon source, electron donor, or electron acceptor (Fig. S1). One of the primary advantages of Tn-seq is the semi-quantitative detection of subtle changes in fitness that are commonly overlooked by other methods. Accordingly, we limited the number of doublings during selection to only 5–6 with the hope of maintaining the ability to differentiate subtle and strong phenotypes.

Tn-seq relies on the sequencing of transposon–genome junctions en masse to determine the relative abundance of each transposon mutant in the population, before and after selection. MR-1 genomic DNA from selection experiments was prepared as described in Experimental procedures (Opijnen and Camilli, 2010) and subjected to Illumina sequencing. Genomic DNA flanking each transposon was mapped to the MR-1 genome to determine the site of transposition, insertion sites were parsed to individual genes, and the number of reads found within a given gene were normalized and tabulated. Our data set derives from a library of 26 793 unique insertion mutants and each coding region has on average five insertions within the central 89% portion of the gene. Fitness values were calculated by dividing the number of reads mapped to each gene at the end of a selection by the number mapped to the same gene in the starting pool, followed by a logarithmic transformation. The entire selection experiment was independently performed twice and values derived from sequencing, as well as fitness calculations, are reported for each experiment individually (Table S1). Fitness values are only calculated for genes with at least three independent insertion sites to reduce variability that can result in misleading fitness calculations. Genes with a fitness value of ≤ −0.5 in both experiments for a given condition are considered to cause a fitness defect when inactivated. To validate our results we compared Tn-seq fitness data (Table S1) with studies of genes known to be important for aerobic or anaerobic growth of MR-1 using lactate and fumarate, limiting our comparison to studies in which actual mutants were analysed using growth assays (Table 1). In every case Tn-seq fitness data correctly measured a fitness defect under the conditions previously reported (see Table 1 for specific genes and references).

Table 1. Previously characterized genes important for growth of MR-1 with lactate and fumarate.
Locus tagGene nameTn-seq fitness valuesReference
Replicate 1Replicate 2
  1. aReference listed did not characterize the specific mutant; however, gene is included as it is part of the pathway characterized by the authors.
  2. ND, not detected; no reads detected after selection.
Previously reported to have a fitness defect only when grown aerobically on lactate
SO_0424 aceE −1.960.06−2.59−0.33Flynn et al. (2012)
SO_1518 lldG −0.81−0.28−1.10−0.40Pinchuk et al. (2009)
SO_1519 lldF −0.570.29−0.76−0.27Pinchuk et al. (2009)
Previously reported to have a fitness defect only when grown anaerobically on lactate and fumarate
SO_0624 crp −0.42ND−0.55−1.35Saffarini et al. (2003)
SO_1329 cyaC −0.13−0.99−0.12−0.41Charania et al. (2009)
SO_0970 fccA 0.08−4.340.11−1.65Myers and Myers (1997)
SO_2916 pta 0.34−4.640.52−2.12Hunt et al. (2010)
SO_2915 ackA 0.19−5.150.05−3.75Hunt et al. (2010)
SO_2912 pflB −0.01−2.87−0.01−2.48Flynn et al. (2012)
SO_1910 menA 0.39ND0.04NDMyers and Myers (1993)a
SO_4713 menF 0.08−3.06−0.02−1.70Myers and Myers (1993)a
SO_4739 menB 0.12−3.230.43−2.11Myers and Myers (1993)a
SO_4573 menD 0.10−2.980.35−2.84Myers and Myers (1993)a
SO_4574 menH 0.02−1.150.01−0.56Myers and Myers (1993)a
SO_4575 menC 0.15−3.050.16−2.57Myers and Myers (1993)a
SO_4576 menE 0.16−3.050.29−1.76Myers and Myers (1993)a
SO_0259 ccmE −0.23ND−0.39−2.78Bouhenni et al. (2005)a
SO_0261 ccmC −0.49−4.29−0.45NDBouhenni et al. (2005)
SO_0263 ccmA −0.47−3.63−0.26NDBouhenni et al. (2005)a
SO_0266 ccmF-1−0.51−3.70−1.05NDBouhenni et al. (2005)
SO_0268 ccmH −0.44−1.010.32−0.10Bouhenni et al. (2005)a
Previously reported to have a fitness defect when grown aerobically or anaerobically on lactate and fumarate
SO_0274 ppc −2.88−1.38−2.95−1.24Flynn et al. (2012)

In the current study we use Tn-seq analysis to gain new insights into the central metabolic pathways of MR-1. Here we report fitness values for 3030 of the 4467 genes encoded by the MR-1 genome and megaplasmid. Figure 1 shows a map of the central carbon metabolism of MR-1, with genes colour coded based on fitness data provided by Tn-seq analysis. The effect of inactivation of most genes was consistent with recent metabolic models for MR-1 (Tang et al., 2007b; Pinchuk et al., 2010; Flynn et al., 2012). Inactivation of a few genes, however, resulted in an unexpected fitness defect, or lack thereof, and we chose these genes to study further.

Figure 1.

Graphic representation of fitness values for the genes of central carbon metabolism in MR-1 determined by Tn-seq. Solid arrows denote single reactions while dotted arrows denote multiple reactions. The inactivation of a gene is determined to have a fitness cost for a condition if the average fitness value falls below −0.5 (Table S1) and given the corresponding colour (see inset). Genes which did not contain insertions in the parent library (†) or did not meet the filtering requirements outlined in Experimental procedures (*) are indicated. A white box denotes genes for which a fitness value was not determined.

Lactate metabolism and regulation of lactate utilization

MR-1 is thought to form syntrophic partnerships with fermentative microbes, oxidizing fermentation end-products (e.g. lactate, propionate, formate, hydrogen, and some amino acids) for anaerobic respiration (Nealson and Scott, 2006). Lactate is of particular interest as it is used as a carbon and energy source for the vast majority of studies involving Shewanella sp. grown in a minimal medium, and Fig. 1 shows a model for lactate metabolism in MR-1. Following uptake, lactate is oxidized to pyruvate by membrane associated flavin adenine dinucleotide (FAD)-dependent lactate dehydrogenases specific for the l- or d-isomer (LldEFG or Dld respectively) (Pinchuk et al., 2009). MR-1 grows somewhat faster on l-lactate than d-lactate in the presence of oxygen (Pinchuk et al., 2009), and accordingly, insertions in lldF and lldG resulted in a mild fitness defect in oxic SBM (Table 1 and Fig. 1). Interestingly, insertions in dld had a fitness defect under both oxic and anoxic conditions. These data are consistent with d-lactate inhibition of l-lactate utilization in MR-1, a hypothesis confirmed by phenotypic analysis of deletion mutants and lactate dehydrogenase activity assays (E.D. Brutinel and J.A. Gralnick, manuscript in preparation).

Subsequently, the fate of pyruvate is dependent on the availability of oxygen. In the presence of oxygen pyruvate is oxidized to acetyl coenzyme A (acetyl-CoA), which then enters the TCA cycle, by the pyruvate dehydrogenase complex (AceEF/LpdA) (Scott and Nealson, 1994; Tang et al., 2007a). Under anoxic conditions pyruvate is converted to acetate and formate in a stepwise fashion by the action of pyruvate–formate lyase (PflAB), phosphate acetyltransferase (Pta), and acetate kinase (AckA) (Scott and Nealson, 1994; Tang et al., 2007c; Hunt et al., 2010). In agreement with the above model, transposon insertions in aceE result in a fitness defect in the presence of oxygen while insertions in pflA, pflB, pta, and ackA result in a fitness defect under anoxic conditions (Table 1 and Fig. 1). Insertions in aceF had a fitness defect under both conditions (Table S1), possibly due to polar effects on the expression of the downstream lpdA gene that we infer to be essential based on the absence of insertion mutants in lpdA in our Tn-seq data set.

Growth on lactate necessitates gluconeogenesis to generate essential precursor metabolites. Accordingly, we did not detect transposon insertions in genes required for gluconeogenesis with the exception of pgi (encoding phosphoglucose isomerase) and gapA1 [encoding glyceraldehyde-3-phosphate dehydrogenase (GAPDH)]. Transposon insertions in pgi do not appear to be lethal but result in a reduced fitness under both oxic and anoxic conditions (Fig. 1 and Table S1). Transposon insertions in the gapA1 gene (SO_0538) appeared to have no effect on fitness under either condition, in disagreement with a model predicting that gapA1 is essential when MR-1 is grown on lactate (Flynn et al., 2012). Genomic sequence analysis previously predicted two paralogues of gapA1 in the genome of MR-1, gapA2 (SO_2345) and gapA3 (SO_2347) (Serres and Riley, 2006), whose protein products share 43% and 34% identity respectively with that of gapA1. While transposon insertions in gapA2 did not appear to effect fitness, insertions in gapA3 had a pronounced effect on fitness under both oxic and anoxic conditions (Fig. 1 and Table S1). This result was unexpected and we wanted to determine if each of the predicted gapA paralogues encode an enzyme with GAPDH activity. To this end we performed NAD- and NADP-dependent GAPDH activity assays on cell extracts derived from MR-1 strains expressing gapA1, gapA2, or gapA3 from a plasmid and the results are reported in Table 2. GAPDH activity in cells expressing gapA1 was not significantly different from a strain carrying a vector control, consistent with the lack of a phenotype for insertions in gapA1, and suggesting that gapA1 was either not expressed or inactive under the conditions tested. We detected a significant increase in NAD- or NADP-dependent GAPDH activity in cells expressing gapA2 or gapA3 respectively. It appears that, when grown in minimal medium supplemented with lactate and fumarate, MR-1 primarily generates glyceraldehydes-3-phosphate in an NADP-dependent reaction catalysed by gapA3. This result highlights the utility of Tn-seq in evaluating the contribution of predicted isoenzymes to an essential biochemical reaction under a given condition. The roles played by gapA1 and gapA2 in the physiology of MR-1 remain unclear, though we speculate that these paralogues must be important under conditions not tested in this manuscript.

Table 2. Glyceraldehyde-3-phosphate dehydrogenase activity in cell lysates from MR-1 expressing gapA1, gapA2 or gapA3 from a plasmid.
 OD340 min−1 mg−1
  1. Reported error represents the standard error of the mean for at least three independent experiments.
MCS23.36 ± 0.100.31 ± 0.02
gapA1 3.79 ± 0.750.24 ± 0.09
gapA2 10.52 ± 0.340.57 ± 0.09
gapA3 4.93 ± 0.212.00 ± 0.20

MR-1 uses the non-oxidative branch of the pentose phosphate pathway for anaerobic metabolism

The pentose phosphate pathway is required for the generation of three key metabolites (ribulose-5-phospphate, erythrose-4-phosphate, and sedoheptulose-7-phosphate) and has an oxidative and a non-oxidative branch. Transposon insertions in a number of genes in the pentose phosphate pathway did not affect the fitness of MR-1 under oxic or anoxic conditions (Fig. 1). Two of the dispensable genes were in the oxidative branch (zwf and gnd), essential for the conversion of glucose 6-phoshate to ribulose-5-phosphate coupled to the generation of NAD(P)H (Fig. 1). In contrast, the genes of the non-oxidative branch are critical for growth of MR-1 on lactate, with the notable exception of the gene coding for transaldolase (tal). Transposon insertions in the gene encoding ribulose-5-phosphate 3-epimerase (rpe) drastically reduced fitness under aerobic and anaerobic conditions, and transposon insertions were not detected in the remaining genes involved in the non-oxidative branch of the pentose phosphate pathway (rpiA and tkt), consistent with the essential role of generating key metabolic intermediates. Under the conditions tested, MR-1 is able to generate ribulose-5-phosphate, erythrose-4-phosphate, and sedoheptulose-7-phosphate via the non-oxidative branch of the pentose phosphate pathway with the reactions catalysed by Tkt, RpiA, and Rpe. This conclusion is supported by the fact that the growth rate of a strain lacking gnd is identical to that of MR-1 grown in oxic or anoxic SBM (Flynn et al., 2012). Our data demonstrate that the pentose phosphate pathway of MR-1 functions in a similar manner as that of Escherichia coli, not with genomics and modelling based predictions, but with phenotypic data.

Branching of the MR-1 anaerobic TCA cycle is dynamic and utilizes an additional citrate synthase

MR-1 preferentially uses oxygen as a terminal electron acceptor and, in the presence of oxygen, acetyl-CoA enters the TCA cycle and is completely oxidized to CO2 generating both energy and reducing equivalents (Tang et al., 2007a). Accordingly, transposon insertions that disrupt TCA cycle genes should result in a decreased growth rate in the presence of oxygen, and fitness values measured by Tn-seq are in agreement with this assertion (Fig. 1). Transposon insertions in the genes encoding aconitate hydratase (SO_0432; acnB) and the α-ketoglutarate dehydrogenase complex (SO_0930, SO_0931; sucAB) were not present in the parent library suggesting disruption of these genes is extremely deleterious in the presence of oxygen, consistent with published E. coli and S. oneidensis mutants lacking aconitate hydratase or α-ketoglutarate dehydrogenase activity (Creaghan and Guest, 1978; Gruer et al., 1997; Pinchuk et al., 2011). An acnB deletion strain was unable to grow in the presence of oxygen and required glutamate for growth in SBM without casamino acids under anoxic conditions demonstrating that the acnB gene product, a predicted bifunctional aconitate hydratase II/2-methylisocitrate dehydratase, is the predominant source of aconitate hydratase activity under either condition tested. Significant fitness defects are not observed for insertions in either of the genes annotated to encoding an isocitrate dehydrogenase (icd or SO1538) suggesting that both genes are expressed and functional under oxic and anoxic conditions. Our data are consistent with a complete suite of TCA cycle reactions in MR-1 grown under aerobic conditions.

Under anaerobic conditions many organisms do not complete the TCA cycle, instead operating with an oxidative and reductive branch, primarily for anaplerotic reactions to replenish α-ketoglutarate, oxaloacetate, and succinyl-CoA pools. Numerous reports have suggested that this is the case for MR-1 using transcriptomics, metabolic labelling, enzymatic assays, and computer modelling to infer the expression/functionality of key enzymes in the TCA cycle in the absence of oxygen (Scott and Nealson, 1994; Beliaev et al., 2002; Tang et al., 2007c,c; Pinchuk et al., 2010). The effect of an insertion mutation in each gene of the anaerobic TCA cycle was measured by Tn-seq and a number of unexpected phenotypes were observed.

Citrate synthase (GltA) catalyses the first reaction of the oxidative arm of a branched anaerobic TCA cycle required to generate α-ketoglutarate for biosynthesis of glutamate. Accordingly, E. coli strains lacking a functional gltA gene require exogenous glutamate for growth in minimal medium (Lakshmi and Helling, 1976). Our parent library contained transposon insertions in the gene that encodes citrate synthase (SO_1926; gltA), despite being grown in SBM, and under anoxic conditions the fitness of cells with transposon insertions in gltA was the same as MR-1 (Table S1 and Fig. 1). The growth rate of a gltA deletion strain was substantially slower than wild type in the presence of oxygen (Fig. 2A), and indistinguishable from wild type under anoxic conditions, even in the absence of the 0.01% casamino acids normally present in SBM (Fig. 2A). The following two possibilities could explain the observed phenotype of the gltA mutant: (i) MR-1 uses an alternative pathway to synthesize α-ketoglutarate/glutamate, or (ii) another protein provides sufficient citrate synthase activity under anaerobic conditions. The former possibility seems unlikely because inactivation of the genes encoding the α-ketoglutarate dependent glutamate synthase (SO_1325, SO_1324; gltBD) has a drastic effect on fitness under oxic and anoxic conditions (Table S1 and Fig. 1). A candidate for the latter possibility is 2-methylcitrate synthase (encoded by prpC; SO_0344), which, while predominantly catalysing the reaction of oxaloacetate and propionyl-CoA to form 2-methylcitrate, has been shown to possess citrate synthase activity in E. coli and S. enterica (Gerike et al., 1998; Horswill and Escalante-Semerena, 1999). To evaluate the contribution of 2-methylcitrate synthase to the lactate metabolism of MR-1 we generated a prpC deletion strain as well as a gltA prpC double mutant. While growth of the prpC mutant on lactate was indistinguishable from wild type, the gltA prpC double mutant required exogenous glutamate for growth in SBM without casamino acids (Fig. 2B). The glutamate requirement of the gltA prpC double mutant was abated by expression of either gltA or prpC from a plasmid. Citrate synthase activity assays were performed on cell extracts derived from MR-1 strains grown in the presence and absence of oxygen and the results are reported in Table 3. While citrate synthase activity in the gltA mutant grown under oxic or anoxic conditions was significantly lower than MR-1, the activity measured was well above that of the double mutant, demonstrating a significant contribution by 2-methylcitrate synthase to the formation of citrate in MR-1.

Figure 2.

The gltA and prpC gene products contribute to the production of citrate in MR-1.

A. Growth curves of MR-1, the gltA mutant, and the prpC mutant were performed in SBM under oxic or anoxic conditions in the presence and absence of casamino acids. The growth curve of the prpC mutant did not differ significantly from that of MR-1 and is represented by the same line for clarity.

B. Growth curves of the gltA prpC double mutant in SBM without casamino acids under oxic or anoxic conditions in the presence and absence of glutamate. The reported values are the averages of three independent experiments and error bars represent the standard error of the mean (SEM).

Table 3. Citrate synthase acticity in cell lysates from cultures grown under oxic and anoxic conditions.
 OD412 min−1 mg−1
  1. Reported error represents the standard error of the mean for at least three independent experiments.
MR-14.92 ± 0.131.76 ± 0.34
ΔgltA 0.34 ± 0.020.19 ± 0.05
ΔprpC 4.76 ± 0.151.60 ± 0.23
ΔgltA ΔprpC 0.03 ± 0.010.06 ± 0.04

In the canonical branched anaerobic TCA cycle α-ketoglutarate is derived from citrate and the oxidative branch while oxaloacetate and succinyl-CoA are derived from phosphoenolpyruvate (PEP) and the reductive branch. It has been suggested that MR-1 uses the above pathways based on transcriptional data and enzyme activity measured in cell extracts (Scott and Nealson, 1994; Beliaev et al., 2002); however, definitive genetic analysis has not been performed. Our Tn-seq experiments suggest that under anoxic conditions inactivation of the genes encoding the succinyl-CoA synthase (sucCD) and fumarate reductase (frdABCD) complexes did not affect the fitness of MR-1 (Table S1 and Fig. 1), a surprising result considering the requirement of succinyl-CoA for biosynthesis. E. coli strains lacking succinyl-CoA synthase activity are prototrophic in the presence of oxygen but require exogenous lysine and methionine for anaerobic growth on glucose (Herbert and Guest, 1968; Mat-Jan et al., 1989). In SBM without casamino acids the growth rate of a strain derived from MR-1 with a sucCD deletion was measurably slower than MR-1 in the presence of oxygen, and indistinguishable from MR-1 under anoxic conditions (Fig. 3A). These results indicate that under anoxic conditions succinyl-CoA is not solely derived from succinate and instead derives either partially or entirely from α-ketoglutarate and the oxidative branch in MR-1. Unfortunately, our Tn-seq experiments did not detect any insertions in the sucAB genes (encoding α-ketoglutarate dehydrogenase). In agreement, E. coil strains lacking α-ketoglutarate dehydrogenase activity are unable to grow in the presence of oxygen (our library was generated on oxic SBM agar), but exhibit prototrophic growth in the absence of oxygen (Herbert and Guest, 1969; Creaghan and Guest, 1978). A strain derived from MR-1 with a sucB deletion was unable to grow in the presence of oxygen, even in rich medium, but exhibited the same growth rate as MR-1 in SBM without casamino acids under anoxic conditions (Fig. 3B).

Figure 3.

The MR-1 anaerobic TCA cycle can produce succinyl-CoA via succinyl-CoA synthase (SucCD) or α-ketoglutarate dehydrogenase (SucAB). Growth assays under aerobic and anaerobic conditions comparing MR-1 to (A) the sucCD mutant or (B) the sucB mutant were performed in SBM without casamino acids. The reported values are the averages of three independent experiments and error bars represent the standard error of the mean (SEM).


The rise of next-generation sequencing has enabled the relatively small genomes of microorganisms to be sequenced cheaply and quickly. The wealth of genomic data available has led to increased efforts to understand cellular processes by systems level analysis. Computer-based metabolic modelling, while an extremely useful tool for the generation of testable hypotheses, is limited by both the accuracy of gene annotation as well as the collective understanding of gene function. Additionally, uncharacterized and hypothetical genes cannot be integrated into models built using computational analysis alone. Critical for the rational design of genetic manipulation intent on changing the metabolic output of MR-1 is an accurate assessment of gene function. We subjected a pooled library of 26 793 unique mutants to selection and the effect of inactivation on fitness was determined for 3030 genes (Table S1). While the coverage within our pooled library was extensive, the oxic conditions under which the library was generated limit genes which can be inactivated by a transposon and result in a viable cell. The sucB mutant clearly demonstrates this as it was not present in our pooled library generated under oxic conditions (Table S1 and Fig. 1), but has the same growth rate as the wild type under anoxic conditions (Fig. 3B). In the future combining pooled libraries generated under both oxic and anoxic conditions would alleviate this problem and increase the number of genes for which our analysis can assign a fitness score.

Our data set contains a variety of genes reported to be essential by studies which relied on individually isolated transposon mutants (Deutschbauer et al., 2011), or the construction of a metabolic model using elementary mode analysis (Flynn et al., 2012) (Table S2). Most of the formerly ‘essential’ genes have slow growth phenotypes under aerobic conditions (Table S1), likely the reason they were incorrectly labelled as essential. This highlights a decided advantage of Tn-seq, where transposon mutants are pooled en masse, over high-throughput methods that rely on the isolation/manipulation of single colonies. Making sense of this massive data set presented a daunting task and accordingly we discuss only the analysis of genes involved in the reactions of central carbon metabolism. In spite of this rather narrow focus, we have included our entire data set in this report with the hopes that other researchers may find the data useful (Table S1).

MR-1 couples the oxidation of fermentation end-products to the reduction of a wide variety of electron acceptors (Nealson and Scott, 2006). Here we demonstrate that 2-methylcitrate synthase, involved in propionate metabolism, provides citrate synthase activity in MR-1 in the absence of propionate (Fig. 2). In Salmonella enterica propionyl-CoA is the preferred substrate of 2-methylcitrate synthase, although acetyl-CoA can be used, and prpC is only expressed in the presence of propionate (Tsang et al., 1998; Horswill and Escalante-Semerena, 1999). Unlike MR-1, S. enterica with a gltA deletion requires glutamate for growth (Horswill et al., 2001). Significant citrate synthase activity in a gltA prpC + strain of MR-1 alludes to a much greater capacity for catabolism of propionate. Expression of prpC in the absence of propionate could represent a metabolic strategy tailored to the ecological niche occupied by MR-1. In an environment with rapid transitions between oxic and anoxic conditions, organisms prepared to oxidize the end-products of fermentation for anaerobic respiration may have an advantage when anoxia inevitably returns. Alternatively, the 2-methylcitrate synthase of MR-1 could have an altered affinity for acetyl-CoA.

Our Tn-seq data led us to investigate the prototrophic nature of a sucCD mutant and genetic data presented here highlights the ability of MR-1 to generate succinyl-CoA from either the reductive or oxidative branch of an incomplete anaerobic TCA cycle. Facultative anaerobic microorganisms are traditionally thought to generate succinyl-CoA from the reductive branch, a model based largely on work done in E. coli and S. typhimurium (Neidhardt, 1987). The same model has been proposed for MR-1 based on a reduction of sucB transcription and α-ketoglutarate dehydrogenase activity under anoxic conditions (Scott and Nealson, 1994; Beliaev et al., 2002). Without genetic data this conclusion was tenuous as both studies saw significant reductions in the transcription/activity of a number of other TCA cycle genes, as would be expected. The direct measurement of mutant fitness by Tn-seq demonstrated that the above conclusions were incorrect and genetic data confirmed the ability of MR-1 to use either branch of the anaerobic TCA cycle. Use of the oxidative branch to generate succinyl-CoA significantly changes the number of electrons consumed/produced as compared with the reductive branch, and Fig. 4 outlines our calculations for each pathway. The oxidative branch produces eight additional electrons as compared with the reductive branch when MR-1 is grown on lactate or a carbon source that enters central metabolism at or above the level of PEP. The disparate redox balances obtained by each branch likely determines the route used by an organism or under a certain condition. For example, fermentative organisms like E. coli must internally balance all redox reactions, necessitating use of the reductive route to avoid the unnecessary generation of NAD(P)H. To date we are unaware of any other examples of a facultative anaerobic microorganism which derives succinyl-CoA from the oxidative branch of an incomplete anaerobic TCA cycle. The use of this pathway to generate reducing equivalents, able to enter the respiratory chain and contribute to proton motive force, may be a widespread strategy for organisms specialized for anaerobic respiration. The decision to use the reductive or oxidative branch by MR-1 is likely a dynamic process dictated by environmental conditions, and ongoing efforts are aimed at elucidating the regulatory mechanisms that control the biosynthetic route to succinyl-CoA.

Figure 4.

Model of the canonical and alternative branch point of the incomplete anaerobic TCA cycle. The indicated NAD(P)H and ATP consumption/generation was calculated for MR-1 oxidizing two molecules of lactate or two molecules of PEP using the reductive (red) or oxidative (ox) branch of an incomplete anaerobic TCA cycle (boxed; lower right corner). In the case of production of succinyl-CoA via the reductive route, the lactate or PEP not used to generate oxaloacetate for citrate synthesis was assumed to be converted to acetate. Conversion of ATP to AMP by PEP synthase is calculated as a cost of two ATP. Formate generated during the anaerobic conversion of pyruvate to acetyl-CoA by PflAB was assumed to be oxidized to CO2 by one or more formate dehydrogenases, generating an additional pair of electrons.

In summary, this report is a functional demonstration of the power of Tn-seq to unravel the complexities of microbial central metabolism. Tn-seq does not rely on gene annotations or computer models, rather a classical genetic approach is applied at the organismal level, validating predictions, testing hypotheses, and revealing unexpected phenotypes.

Experimental procedures

Bacterial strains and culture conditions

Strains and plasmids used in this study are listed in Table S3 and oligonucleotide primers are listed in Table S4. E. coli strains were maintained on Luria–Bertani (LB) agar plates containing 50 μg ml−1 kanamycin and/or 250 μM 2,6-diaminopimelic acid as necessary. During routine manipulation and strain construction MR-1 was maintained on LB agar containing 50 μg ml−1 kanamycin as necessary. For growth experiments MR-1 was grown in/on SBM (Covington et al., 2010) containing 5 ml l−1 vitamin mix (Balch et al., 1979), 5 ml l−1 mineral mix (Marsili et al., 2008), 0.01% casamino acids (Difco), 40 mM sodium dl-lactate, 80 mM sodium fumarate, and 4 mM glutamate or 50 μg ml−1 kanamycin when required. For growth curves strains stored in glycerol at −80°C were freshly streaked onto LB agar plates and incubated for ∼ 16 h at 30°C after which single colonies were inoculated into 2 ml of LB medium and shaken at 30°C for 6 h. LB cultures were diluted 1/100 into 2 ml of SBM and shaken for ∼ 16 h at 30°C after which cultures were diluted 1/100 into fresh SBM. Anaerobic cultures were stoppered with butyl rubber and flushed with nitrogen gas for 15 min following inoculation. For enzymatic assays cultures grown as above were pelleted by centrifugation and frozen at −20°C. Starter cultures of acnB and sucB mutant strains were grown in anoxic SBM supplemented with 10% LB.

Transposon mutant library generation and selection

MmeI restriction enzyme recognition sites were introduced in both inverted repeats of the pMiniHimar RB1 (Bouhenni et al., 2005) transposon delivery vector (5′-CCGGGGACTTATCAGCCAACCTGT-3′ to 5′-CCGGGGACTTATCATCCAACCTGT-3′) using Quikchange mutagenesis (Stratagene) resulting in the pEB001 transposon delivery vector. Transposon mutagenesis was performed by conjugal transfer of pEB001 to MR-1 as previously described (Bouhenni et al., 2005). Mating mixes were scraped from LB agar plates, resuspended in SBM, and plated on SBM agar plates containing 50 μg ml−1 kanamycin. After incubating aerobically for 48 h at 30°C, ∼ 50 000 colonies were resuspended in SBM and pooled cells were aliquoted and frozen at −80°C in glycerol. Library selection was performed by thawing an aliquot, isolating genomic DNA from half of the culture, and back diluting the other half in SBM to a final OD600 of 0.005. Cultures were harvested by centrifugation when the OD600 reached ∼ 0.2 and genomic DNA was extracted using the Wizard genomic DNA purification kit (Promega) per manufacturers instructions.

Illumina sequencing and data analysis

Genomic DNA was prepared for sequencing as previously described (Opijnen and Camilli, 2010). Briefly, genomic DNA samples were digested with MmeI and an adapter with a 3′-NN overhang was ligated to the resulting fragments. Adapters contained a unique 4 bp bar code enabling multiple samples to be mixed into a single sequencing lane. Inverted repeat-specific and adapter-specific oligonucleotide primers were used to amplify the intervening genomic region. The 5′-end of each primers contained sequences required for Illumina sequencing. Sequence analysis (single read, 50 bp) was performed on an Illumina Genome Analyzer IIx (Tufts University Genomics Center or the University of Minnesota BioMedical Genomics Center). All downstream analysis of sequence data was performed using the Galaxy server (Giardine et al., 2005; Blankenberg et al., 2010; Goecks et al., 2010) maintained by the Minnesota Supercomputing Institute (MSI, University of Minnesota). Raw sequencing reads were separated by bar code with a roughly equal number of reads obtained for each bar code (2.45, 2.85, and 3.03 × 107 reads each for experiment one; 1.11, 0.97, and 0.96 × 107 reads each for experiment two). Reads were stripped of the bar code and adapter sequence and filtered based on quality and length (16–17 bp). Reads were mapped to the MR-1 genome and megaplasmid (NC_004347 and NC_004349 respectively) using Bowtie (Langmead et al., 2009; Langmead, 2010), discarding reads without 100% identity to a unique position. Insertion sites with less than 10 reads were ignored. The location of each transposon insertion site was determined and parsed based on the J. Craig Venter Institute annotation of the MR-1 genome (Rockville, MD). Insertions that fell in the first 1% or the last 10% of a coding sequence were omitted. Due to variation in the total reads returned from different sequencing runs, reads per insertion sit were normalized by total reads in millions. Fitness values were not calculated for genes with less than three insertion sites due to a high degree of variability. The above processed data set is provided (Table S1).

Deletion and complementation

In-frame gene deletions in MR-1 were generated using homologous recombination as previously described (Coursolle et al., 2010). Plasmids for gene deletions or complementation were constructed using standard laboratory molecular biology protocols. For deletion constructs regions flanking deletion targets (1 kb) were amplified by polymerase chain reaction (PCR) and cloned into the pSMV3 suicide vector. For complementation constructs gene coding regions and 30 bp upstream were amplified by PCR and cloned into pBBR1MCS-2 (Kovach et al., 1995). Complementation constructs were moved into deletion strains by conjugation. All plasmid-based constructs and gene deletions were verified by PCR and sequencing.

Citrate synthase activity assay

Crude cell extracts were generated from strains grown to mid-exponential phase in SBM under either oxic or anoxic conditions. Cultures were pelleted by centrifugation, washed once with fresh SBM, and resuspended in ice-cold 50 mM Tris-HCl (pH 8.0). Citrate synthase activity was determined as previously described (Bond et al., 2005). Control reactions were performed by omitting oxaloacetate (Sigma).

Glyceraldehyde-3-phosphate dehydrogenase activity assay

Crude cell extracts were generated from strains grown to stationary phase in SBM with 50 μg ml−1 kanamycin under oxic conditions. Cultures were pelleted by centrifugation, washed once with fresh SBM, and resuspended in ice-cold lysis buffer [20 mM Tris-HCl (pH 8.0), 2 mM β-mercaptoethanol, and 2 mM EDTA] (Hillman, 1979). Cell suspensions were disrupted by sonication and cleared by centrifugation at 15 000 g for 15 min at 4°C. GAPDH activity in crude cell extracts was determined as previously described (Ferdinand, 1964), except 2 mM of dl-glyceraldehydes-3-phosphate (Sigma) was used. Control reactions were performed by omitting dl-glyceraldehydes-3-phosphate.


We would like to thank Dr Andrew Camilli and Dr Tim van Opijnen (Tufts University) for protocols and advice regarding the Tn-seq methodology. We would like to thank Kip L. Bodi (Tufts Genomics Core Facility), Nicole Peterson, and Aaron Becker (University of Minnesota BioMedical Genomics Center) for assistance with Illumina sequencing. We would like to thank James E. Johnson of the Minnesota Supercomputing Institute for programming support. We would like to thank Dr Dan Coursolle for generation of the plasmid used to delete gltA. This work was supported by seed grants from the Graduate School and the Microbial and Plant Genomics Institute at the University of Minnesota.