Mesaconase/Fumarase FumD in Escherichia coli O157:H7 and Promiscuity of Escherichia coli Class I Fumarases FumA and FumB

Mesaconase catalyzes the hydration of mesaconate (methylfumarate) to (S)-citramalate. The enzyme participates in the methylaspartate pathway of glutamate fermentation as well as in the metabolism of various C5-dicarboxylic acids such as mesaconate or L-threo-β-methylmalate. We have recently shown that Burkholderia xenovorans uses a promiscuous class I fumarase to catalyze this reaction in the course of mesaconate utilization. Here we show that classical Escherichia coli class I fumarases A and B (FumA and FumB) are capable of hydrating mesaconate with 4% (FumA) and 19% (FumB) of the catalytic efficiency k cat/K m, compared to the physiological substrate fumarate. Furthermore, the genomes of 14.8% of sequenced Enterobacteriaceae (26.5% of E. coli, 90.6% of E. coli O157:H7 strains) possess an additional class I fumarase homologue which we designated as fumarase D (FumD). All these organisms are (opportunistic) pathogens. fumD is clustered with the key genes for two enzymes of the methylaspartate pathway of glutamate fermentation, glutamate mutase and methylaspartate ammonia lyase, converting glutamate to mesaconate. Heterologously produced FumD was a promiscuous mesaconase/fumarase with a 2- to 3-fold preference for mesaconate over fumarate. Therefore, these bacteria have the genetic potential to convert glutamate to (S)-citramalate, but the further fate of citramalate is still unclear. Our bioinformatic analysis identified several other putative mesaconase genes and revealed that mesaconases probably evolved several times from various class I fumarases independently. Most, if not all iron-dependent fumarases, are capable to catalyze mesaconate hydration.


Introduction
Fumarase (EC 4.2.1.2) catalyzes the reversible hydration of fumarate to (S)-malate (or L-malate in D/L nomenclature) ( Fig 1A) and participates in the tricarboxylic acid (TCA) cycle and in a number of other metabolic processes. Two phylogenetically unrelated classes of fumarases are known today [1]. The class I enzymes are thermolabile homodimers with a molecular mass of 120 kDa. They contain an oxygen-sensitive catalytic [4Fe-4S]-cluster acting as a Lewis acid to We have recently shown that B. xenovorans class I fumarase is equally efficient in the hydration of fumarate to (S)-malate and mesaconate to (S)-citramalate (L-citramalate in D/L nomenclature) ( Fig 1C) and participates in mesaconate utilization in this bacterium (Fig 2) [10]. Despite the high similarity of fumarate and mesaconate (methylfumarate), the finding was unexpected, as homologous class I fumarases of E. coli and Syntrophobacter fumaroxidans were already tested for mesaconase activity, and the obtained results were negative [1,12,15,16].
Mesaconase was first identified in the methylaspartate pathway of glutamate fermentation in Clostridia. In this pathway, glutamate is converted in the B 12 -dependent glutamate mutase reaction to methylaspartate, which is deaminated to mesaconate by methylaspartate ammonialyase. Mesaconase further converts mesaconate to (S)-citramalate, which is cleaved by citramalate lyase to pyruvate and acetate. Mesaconase from Clostridium tetanomorphum was purified and characterized in the 1960 th and was shown to be an oxygen-sensitive and Fe-dependent heterodimer possessing also fumarase activity [17,18,19]. Furthermore, the participation of mesaconase in acetate assimilation in Rhodospirillum rubrum was proposed [20,21], although the corresponding pathway still needs to be proven [22]. Mesaconase from a soil Bacillus sp. participates in L-threo-β-methylmalate and mesaconate metabolism [23,24] and has properties similar to those of class I fumarase from Burkholderia [10]. These data suggest that the presence of mesaconase activity may be a general feature of class I fumarases, being in line with the notion that enzymes, transporters and sensors interacting with different C 4 -dicarboxylates have often a broad specificity over their substrate (e.g., [25]) In this work, we have cloned and expressed the genes of E. coli fumarases A and B and characterized the corresponding proteins. We show that these enzymes are promiscuous catalyzing also mesaconate hydration with 10% of catalytic efficiency, compared to fumarate. Furthermore, we have found that most of enterohemorrhagic E. coli (EHEC) strains possess an additional gene for class I fumarase. It is clustered with genes for the key enzymes of the methylaspartate pathway of glutamate fermentation and encodes a hydratase highly specific for mesaconase as a substrate. Our bioinformatic analysis suggests that mesaconases in other species also belong to class I fumarases or are phylogenetically related to them.

Microbial strains and growth conditions
Escherichia coli strains K12 W3110 (DSMZ 5911) and ATCC 700728 (DSMZ 17076) were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ). The E. coli K-12 AG1 strains containing the respective plasmid-encoded fumarase genes fumA (strain JW1604), fumB (strain JW4083) and fumC (strain JW1603) from E. coli K-12 W3110 were obtained from the E. coli ASKA library [26]. The growth of the K-12 strain was tested in the presence of mesaconate (20 mM) under both oxic and anoxic conditions in the minimal medium described Kronen et al. [10]. For some experiments, the cells of the ATCC 700728 strain were grown under oxic conditions at 37°C in the same minimal medium with acetate (50 mM) or propionate (50 mM) and citrate (2 mM). Furthermore, we attempted to grow the ATCC 700728 strain in this minimal medium under anaerobic conditions in the presence of glutamate (80 mM) or yeast extract (0.5%) plus glutamate (80 mM) as well as in the glutamate/ yeast extract medium described by Kato and Asano [27]. In addition, we grew these cells in the lysogeny broth (LB) medium (bactotryptone, 10 g l -1 , yeast extract, 5 g l -1 , NaCl, 10 g l -1 ) and in the LB-glutamate (80 mM) medium. All anaerobic growth tests were done in the presence and in the absence of cobalamins. Three different forms of cobalamin were tested: hydroxocobalamin (0.2 and 2 μM), cyanocobalamin (0.2 and 2 μM) and adenosylcobalamin (0.2 and 2 μM).

Preparation of cell extracts
For the cell extract enzyme assays, cell extracts were prepared under oxic conditions using a mixermill (MM200, Retsch, Haare, Germany). Cells (100-150 mg) were suspended in 0.5 ml of 20 mM 3-(N-morpholino)propanesulfonic acid (Mops)/KOH, pH 6.9, 0.1 mg ml -1 DNase I and 0.5 mM dithiothreitol (DTT) in 2.2 ml Eppendorf vials. Then, 1 g of glass beads (diameter 0.1-0.25 mm) was added to the suspension, and the cells were treated in the mixer-mill for 10 min at 30 Hz. This was followed by a centrifugation step (14,000 g, 4°C, 10 min), and the supernatant (cell extract) was used for enzyme assays. The protein content of the cell extract was 6 to 20 mg ml -1 .

Enzyme assays
Spectrophotometric enzyme assays were performed in 0.5 ml cuvettes at 37°C, unless otherwise indicated. Anaerobic assays were done in an anaerobic chamber. One unit of enzyme activity (U) corresponds to 1 μmol substrate converted per minute.
Citramalate lyase was measured as described for citrate lyase, but with (S)-citramalate and lactate dehydrogenase (Sigma) instead of citrate and malate dehydrogenase.

Reactivation of fumarases
As exposure to oxygen (partly) inactivated the FeS clusters of class I fumarases, the enzymes had to be reactivated by activation with DTT and Fe 2+ . The reactivation was performed, as described previously [10].

Cloning of fumarase D genes from E.coli ATCC 700728
The fumD gene (ecatcc700728_0849) was amplified by PCR from E.coli O157:H7 ATCC 700728 chromosomal DNA with the Q5 High-Fidelity DNA polymerase (New England Biolabs) according to manufacturer's instructions with the following primers: FumD_fwd (CATATCGAAGGTCGTCATATGTCCAAACCATTTATCTGGC), and FumD_rev (GACAGCTTATCATCGATATTAATGTCCCGCAGGGCA). Primers were constructed with NEB Builder (http://nebuilder.neb.com/) and contained adapter sequences for the cloning vector pET16b (underlined) and NdeI restriction site (in bold). Thirty-four cycles were conducted in an automated thermocycler under the following conditions: a 20 s (3 min in the first cycle) denaturation at 98°C, annealing for 30 s at 63.8°C, and primer extension for 1.5 min (5 min in the final cycle) at 72°C. The PCR products were purified and ligated into the pET16b vector (restricted with HindIII and NdeI), using the Gibson Assembly Cloning Kit (New England Biolabs, Frankfurt) according to the manufacturer's protocol.
Fumarase D was produced in E. coli lysY cells (New England Biolabs, Frankfurt) that had been transformed with the corresponding plasmid. The cells were grown aerobically at 37°C in LB medium supplemented with 100 μg ml -1 ampicillin, 100 μM FeSO 4 and 100 μM Fe(II) citrate. Expression was induced at OD 578 of 0.4-0.5 with 0.8 mM IPTG. After the induction, the culture was transferred to a sterile Schott bottle that was shut tightly for anaerobic conditions and slightly shaken at 20°C. After additional growth for 12-15 h, the cells were harvested under anaerobic conditions and stored in an anaerobic Schott bottle at -20°C until use.

Purification of recombinant enzymes
Frozen E. coli cells from the respective homologously expressed fumarases were suspended in a buffer containing 40 mM Tris/HCl (pH 7.8), 0.1 mg ml -1 DNase I and 0.5 mM DTT (1.5 ml resuspension buffer per 1 g cell mass). The suspensions were passed twice trough a chilled French pressure cell at 137 MPa, and the cell lysate was centrifuged for 1 h at 100,000 × g and 4°C. The supernatant was loaded onto a 1 ml Protino Ni-NTA column (Macherey-Nagel) previously equilibrated with buffer A (50 mM Tris/HCl, pH 7.8, 300 mM NaCl, 5 mM MgCl 2 ). Purification was carried at a flow rate of 1 ml min -1 . The His-tagged proteins were eluted at different concentrations of imidazole (25, 50, 300, 500 mM) obtained by mixing buffer A with buffer B (buffer A with 500 mM imidazole). The enzymes were concentrated by ultrafiltration (Amicon YM 10 membrane, Millipore) and stored at -20°C with 50% (v/v) glycerol. For fumarase D, all purifications steps had to be performed under anaerobic conditions.

Molecular biological techniques
The in silico cloning steps were performed with the program Clone Manager 7.11 (Scientific & Educational Software). Standard protocols were used for purification, preparation, cloning, transformation and amplification of DNA [33,34,35]. Plasmid DNA isolation, purification of PCR products and plasmids were performed using Qiagen kits according to the manufacturer's specifications.

Database search and phylogenetic analysis
Query sequences were obtained from the National Center for Biotechnology Information (NCBI) database. The BLAST searches were performed via NCBI BLAST server (http://www. ncbi.nlm.nih.gov/BLAST/) [36]. The amino acid sequences were aligned with sequences from GenBank using CLUSTALW [37] implemented within BioEdit software (http://www.mbio. ncsu.edu/BioEdit/bioedit.html). The phylogenetic tree was reconstructed using a neighborjoining algorithm [38] in the TREECONW program package [39]. The positions with gaps were not taken into account in the phylogenetic reconstructions.
Other methods DNA sequence determination of purified plasmids was performed by GATC Biotech (Konstanz, Germany). SDS-polyacrylamide gel electrophoresis (SDS/PAGE; 12.5%) was performed using the Laemmli method [40]. Proteins were visualized by Coomassie blue staining [41]. Protein concentration was measured according to the Bradford method [42], using BSA as standard.

Characterization of fumarases A, B and C from E. coli
To test the activities of fumarases of a classical E. coli K-12 strain (W3110) with mesaconate, we homologously produced class I fumarases FumA and FumB as well as a class II fumarase FumC. The corresponding plasmids were obtained from the ASKA library [26]. The proteins containing N-terminal His-tags were aerobically purified using a Ni-NTA column. SDS-PAGE analysis revealed the presence of a major band with an apparent molecular mass of 60 kDa (for FumA and FumB) and 50 kDa (for FumC) in the 50 mM (for FumA) or 300 mM imidazole fractions (for FumB and FumC) (Fig 3), which were used to study the catalytic properties of the enzymes.
All three enzymes catalyzed the reversible conversion of fumarate into (S)-malate with very similar catalytic efficiencies k cat /K m ( Table 1). Class I enzymes were oxygen-sensitive and required anoxic conditions for maximal activity; incubation of the enzyme with Fe 2+ and thiol restored full enzymatic activity. On the opposite, class II fumarase FumC remained stable under oxic conditions, and its activity did not increase upon incubation with Fe 2+ and DTT. All three enzymes were active with (2S,3S)-tartrate, but much less than with (S)-malate ( Table 1). FumB had the highest catalytic efficiency with this substrate, corroborating the published results [6] and the requirement of FumB for (2S,3S)-tartrate metabolism [14]. The enzymes were not able to convert (R)-malate, (2R,3R)-tartrate, meso-tartrate or (R)-citramalate. Unlike FumC, both class I fumarases catalyzed the reversible conversion of mesaconate into (S)-citramalate, with FumB being a better catalyst for mesaconase reaction (10/19% of catalytic efficiency with (S)-citramalate/mesaconate, compared to (S)-malate/fumarate) ( Table 1). The catalytic efficiency of FumA with (S)-citramalate/mesaconate was~4% of that with fumarate or (S)-malate. Identification of putative mesaconase genes in bacteria possessing the methylaspartate pathway for glutamate fermentation As the methylaspartate pathway of glutamate fermentation involves mesaconase reaction (Fig  2), we decided to check the presence of class I fumarase homologues in the vicinity of the genes for glutamate mutase and methylaspartate ammonia-lyase in the genomes of bacteria possessing both of these enzymes (as described in [32]).
Clostridium tetanomorphum as well as many other Clostridia possess the gene cluster containing genes for glutamate fermentation (e.g., C. tetani, C. carboxidivorans, C. ljungdahlii, Thermobrachium celere). Besides glutamate mutase and methylaspartate ammonia-lyase, this cluster encodes genes for the putative citramalate lyase (annotated as citrate lyase genes; note that the citramalate and citrate lyase reactions are very similar and most probably are catalyzed by homologous enzymes [43]). Furthermore, there are also two genes encoding homologues of N-and C-terminal domains of class I fumarase in this genes cluster (Fig 4), which were putatively designated as mesaconase genes for C. tetani [44]. Similar gene clusters can be found in the genomes of Blautia hansenii (Firmicutes, Clostridia), Alkaliphilus metalliredigens (Firmicutes, Clostridia), Thermosinus carboxydivorans (Firmicutes, Negativicutes), Dictyoglomus thermophilum (Dictyoglomi), Desulfitobacterium hafniense (Firmicutes, Clostridia) (Fig 4). In Treponema denticola (Spirochaetes), class I fumarase genes are located together with citrate lyase genes, whereas methylaspartate ammonia-lyase and glutamate mutase genes are located elsewhere in the genome. In Clostridium kluyveri, only a rudimentary glutamate fermentation gene cluster is present (Fig 4), which contains, apart from a putative mesaconase gene, only genes encoding one of the subunits of glutamate mutase and an enzyme involved in citrate (citramalate) lyase biogenesis. Some Archaea (haloarchaea) possess genes for glutamate mutase and methylaspartate ammonia-lyase as well. However, these enzymes participate in haloarchaea in the methylaspartate cycle of acetate assimilation rather than in glutamate fermentation. In this cycle, mesaconate formed in the methylaspartate ammonia-lyase reaction is metabolized through the activation to mesaconyl-CoA rather than via the mesaconase reaction [32,45]. Interestingly, 14.8% of all sequenced Enterobacteriaceae strains (26.5% of all E. coli strains, 90.6% of E. coli O157:H7) possess the genes for methylaspartate ammonia-lyase and glutamate mutase. In all these cases, these genes are co-localized with an additional copy of class I fumarase (fourth fumarase gene in E. coli O157:H7) (Fig 5). We designated the corresponding protein as fumarase D (FumD) and decided to characterize it biochemically.

Characterization of fumarase D from E. coli O157:H7
The fumD gene from E. coli O157:H7 (ATCC 700728) was cloned and expressed in E. coli lysY. The resulting protein was purified using a Ni-NTA column and analyzed using SDS-PAGE. A major band with an apparent molecular mass of~60 kDa was revealed in the 300 mM imidazole fraction (Fig 3). Aerobic purification resulted in inactive protein, which could only partly be reactivated with Fe 2+ and thiol (38 U mg -1 protein, measured as mesaconate hydration). However, if all operations (e.g., protein expression, purification, reactivation and enzyme Note that the gene encoding a protein of unknown function (DUF1446) downstream of mal is often annotated as methylaspartate ammonia-lyase, for unknown reason. A homologous gene is present downstream of mal in Dictyoglomus thermophilum (Fig 4). Note also that glutaminase is known to contribute to the acid resistance that is an important feature for enterobacterial virulence [49,50].
doi:10.1371/journal.pone.0145098.g005 assays) were performed under anoxic conditions, significantly higher enzymatic activity could be measured ( Table 2). In fact, the enzyme has higher catalytic efficiency k cat /K m for mesaconate than for fumarate and therefore appears to be a true mesaconase rather than fumarase. The clustering of fumD with methylaspartate ammonia-lyase and glutamate mutase genes (Fig 5) implies the involvement of FumD in glutamate conversion to (S)-citramalate.

Presence of glutamate fermentation gene cluster in pathogenic enterobacteria
All enterobacteria carrying this cluster appear to be (opportunistic) pathogens. Indeed, homologs of E. coli O157:H7 glutamate fermentation gene cluster can be found in e.g. enteropathogenic E. coli O55:H7 strains, shigatoxigenic E. coli O145:H28 strains, Klebsiella oxytoca, Citrobacter freundii, Enterobacter cloacae, Shigella flexneri, Salmonella enterica, E. fergusonii, Yersinia enterocolitica, Morganella morganii, among others. The presence of the glutamate fermentation cluster in these pathogenic species suggests a possible role of glutamate fermentation in pathogenicity. Glutamate is known to be present in the intestinal tract in considerable amounts (of millimolar range [51]) and could therefore serve as an important carbon and energy source for pathogens. For that, (S)-citramalate should be further converted to pyruvate and acetate/acetyl-CoA via citramalate or citramalyl-CoA lyase reactions. As most enterobacteria do not possess citramalyl-CoA lyase genes [52], we first suggested that the citramalate cleavage proceeds through the activity of citrate lyase which is present in many Enterobacteriaceae [53].
Fate of (S)-citramalate produced from glutamate E. coli ATCC 700728 possesses a set of genes encoding citrate lyase which, taking into account the similarity of citrate and citramalate lyases, may catalyze both reactions. In order to test this hypothesis, we grew E. coli ATCC 700728 in the presence of citrate and tested it for citrate and citramalate lyase activities. Although we measured high activity of citrate lyase under these conditions (0.48 ± 0.05 U mg -1 protein), no activity of citramalate lyase could be detected. Further possible candidates for catalyzing the citramalate cleavage were isocitrate lyase (0.79 ± 0.04 U mg -1 protein in acetate-grown cells) and methylisocitrate lyase (involved in the methylcitrate pathway of propionate assimilation in E. coli [54,55]; the activity of the key enzyme of the pathway, methylcitrate synthase, was 0.62 ± 0.10 U mg -1 protein in propionategrown cells). However, no citramalate lyase activity could be detected in both acetate-and propionate-grown cells (measurement limit: 1 mU mg -1 protein). Therefore, the fate of (S)citramalate produced from glutamate is still unclear.
All our attempts to find conditions under which E. coli ATCC 700728 expresses the glutamate fermentation genes failed. No growth was detected with glutamate as a sole carbon source under anoxic conditions, and glutamate did not stimulate anaerobic growth of E. coli ATCC 700728 on the LB medium. Furthermore, we could not detect any activity of methylaspartate ammonia-lyase in these cells (measurement limit: 0.1 mU mg -1 protein), suggesting that the enzyme is not produced under these conditions. Previously, the activity of methylaspartate ammonia-lyase was shown to be widespread among enterobacteria, although none of the strains was capable to grow on glutamate or methylaspartate alone [27,56]. However, E. coli ATCC 700728 strain cultivated under the same conditions as by Kato and Asano [27] did not possess any detectable activity of this enzyme either (data not shown).

Phylogenetic analysis of class I fumarases
Data base search showed that the sequences identified as (potential) mesaconases are distributed along different phylogenetic clusters in the tree (Fig 6). Heterodimeric class I fumarases and mesaconases are located in one cluster with the genes encoding (2R,3R)-tartrate dehydratases, with the latter being most closely related to the archaeal proteins. These data imply that mesaconase evolved from various class I fumarases several times during the evolution. Together with intrinsic mesaconase activity of class I fumarases of E. coli ( Table 1) and B. xenovorans [10], these data strongly suggest that most, if not all iron-dependent fumarases, are capable to catalyze mesaconate hydration.

Discussion
This study shows that classical fumarases A and B of E. coli possess mesaconase activity, and that many (opportunistic) pathogenic Enterobacteriaceae have an additional class I fumarase homologue adapted for catalyzing mesaconase reaction. Although the function of this mesaconase in enterobacteria is not known, this enzyme is present in (opportunistic) pathogens and may therefore be related to pathogenicity. Thus, the transcription of the glutamate fermentation genes in E. coli ATCC 700728 could be dependent on one of the (unknown) stimuli of pathogenicity. Even if so, the further citramalate metabolism in this species is unclear. We cannot exclude the possibility that (S)-citramalate is not further metabolized but secreted by pathogens. Interestingly, citramalate, probably of bacterial origin, was detected in plasma from trauma patients [57]. In fact, this compound may inhibit certain enzymes of host metabolism and thus contribute to virulence; for example, citramalate is an inhibitor of mitochondrial aconitase in plants [58]. Altogether, the function of glutamate fermentation cluster in Enterobacteriaceae requires further investigations. Interestingly, the studied E. coli O157:H7 strain (ATCC 700728) is classified to Biosafety level 2 and is therefore less pathogenic than most other E. coli O157:H7 strains (Biosafety level 3). Therefore, the strain we chose in our work might be deficient in some pathogenicity-related features, and further studies have to be done either with classical O157:H7 strains or with enterobacteria that have already been shown to express genes involved in the glutamate fermentation [27,56].
Many prokaryotes use the citramalate pathway for isoleucine biosynthesis [59,60,61], and it is possible that citramalate synthesized from glutamate can be used for the formation of isoleucine. However, this process proceeds via the (R)-stereoisomer of citramalate [62,63], whereas (S)-citramalate in formed in the mesaconase reaction. Therefore, the synthesis of isoleucine from glutamate with participation of FumD would require an epimerase interconverting (R)and (S)-citramalate.
Mesaconic, itaconic, methylsuccinic and other C 5 -dicarboxylic acids are regarded as important platform chemicals [64,65,66], and their bioproduction is currently being optimized [65,66,67,68]. The promiscuity of class I fumarases can therefore be important for the optimization of production of these compounds, but also for engineering of (S)-citramalate producing strains. Interestingly, E. coli strain producing up to 7 g/l mesaconate has been engineered recently [66]. It is of particular interest to test the effect of the deletion of class I fumarases on mesaconate production by this strain.
Our bioinformatic analysis identified several putative mesaconases involved in glutamate fermentation through the methylaspartate pathway. The phylogenetic analysis of identified mesaconases and fumarases suggests that mesaconases do not construct a monophyletic cluster on the tree and therefore probably evolved several times from various iron-dependent fumarases during evolution.