Molecular Microbiology (2002) 44(1), 49–60
Molecular characterization of phenyllactate dehydratase and its initiator from Clostridium sporogenes Sandra Dickert, Antonio J. Pierik and Wolfgang Buckel* Laboratorium für Mikrobiologie, Fachbereich Biologie, Philipps-Universität, Marburg, Germany. Summary The heterotrimeric phenyllactate dehydratase from Clostridium sporogenes, FldABC, catalyses the reversible dehydration of (R)-phenyllactate to (E)cinnamate in two steps: (i) CoA-transfer from the cofactor cinnamoyl-CoA to phenyllactate to yield phenyllactyl-CoA and the product cinnamate mediated by FldA, a (R)-phenyllactate CoA-transferase; followed by (ii) dehydration of phenyllactyl-CoA to cinnamoyl-CoA mediated by heterodimeric FldBC, a phenyllactyl-CoA dehydratase. Phenyllactate dehydratase requires initiation by ATP, MgCl2 and a reducing agent such as dithionite mediated by an extremely oxygen-sensitive initiator protein (FldI) present in the cell-free extract. All four genes coding for these proteins were cloned and shown to be clustered in the order fldAIBC, which shares over 95% sequence identity of nucleotide and protein levels with a gene cluster detected in the genome of the closely related Clostridium botulinum Hall strain A. FldA shows sequence similarities to a new family of CoAtransferases, which apparently do not form covalent enzyme CoA-ester intermediates. An N-terminal Strep II-Tag containing enzymatically active FldI was overproduced and purified from Escherichia coli. FldI was characterized as a homodimeric protein, which contains one [4Fe-4S]1+/2+ cluster with an electron spin S = 3/2 in the reduced form. The amino acid sequence as well as the chemical and EPR-properties of the pure protein are very similar to those of component A of 2-hydroxyglutaryl-CoA dehydratase from Acidaminococcus fermentans (HgdC), which was able to replace FldI in the activation of phenyllactate dehydratase. Only in the oxidized state, FldI and compo-
Accepted 4 January, 2002. *For correspondence. E-mail
[email protected]; Tel. (+49) 6421 2821527; Fax (+49) 6421 2828979.
© 2002 Blackwell Science Ltd
nent A exhibit significant ATPase activity, which appears to be essential for unidirectional electron transfer. Both subunits of phenyllactyl-CoA dehydratase (FldBC) show significant sequence similarities to both subunits of 2-hydroxyglutaryl-CoA dehydratase (HgdAB). The fldAIBC gene cluster resembles the hadAIBC gene cluster in the genome of Clostridium difficile and the hadABC,I genes in C. botulinum. The four subunits of these deduced 2-hydroxyacid dehydratases (65–81% amino acid sequence identity between the had genes) probably code for a 2-hydroxyisocaproate dehydratase involved in leucine fermentation. This enzyme could be the target for metronidazole in the treatment of pseudomembranous enterocolitis caused by C. difficile. Introduction The strictly anaerobic bacterium Clostridium sporogenes, clostridial cluster I (Collins et al., 1994), ferments Lphenylalanine to ammonia, carbon dioxide, phenylacetate and 3-phenylpropionate. In this pathway, phenylalanine is oxidatively deaminated to phenylpyruvate and further oxidized to phenylacetyl-CoA, from which phenylacetate and ATP are formed. The electron balance of this fermentation is compensated by reduction of two thirds of the phenylpyruvate via (R)-phenyllactate and (E)-cinnamate to 3-phenylpropionate (Bader et al., 1982; Dickert et al., 2000). The most mechanistically demanding step in this reductive branch is the reversible syn-dehydration of (R)phenyllactate to (E)-cinnamate (Pitsch and Simon, 1982), which is catalysed by an oxygen-sensitive phenyllactate dehydratase, designated as FldABC (F from phenylalanine). The heterotrimeric enzyme complex, m = 130 ± 15 kDa, is composed of three subunits: A, 46.44 kDa; B, 45.49 kDa; and C, 37.34 kDa. Subunit A acts as a cinnamoyl-CoA:phenyllactate CoA-transferase, whereas the [4Fe-4S]2+-containing subcomplex FldBC catalyses the dehydration of (R)-phenyllactyl-CoA to cinnamoylCoA (Fig. 1). The dehydratase has to be activated by a yet unidentified component of the cell-free extract in the presence of 0.4 mM ATP, 2.5 mM MgCl2 and 0.1 mM dithionite. To start the dehydration of (R)-phenyllactate,
50 S. Dickert et al.
Fig. 1. Proposed mechanism for the dehydration of (R)phenyllactate. FldA, cinnamoyl-CoA:(R)-phenyllactate CoAtransferase; FldBC, (R)-phenyllactyl-CoA dehydratase; FldI, initiator (component A) of phenyllactate dehydratase. The size of the electron (e) transferred by FldI should indicate its energy content.
catalytic amounts of cinnamoyl-CoA (20 mM) are required. FAD (10 mM) stimulates the dehydratase activity but is not present as a prosthetic group in the enzyme (Dickert et al., 2000). A very similar reaction, the syn-dehydration of (R)-2hydroxyglutaryl-CoA to (E)-glutaconyl-CoA, occurs in Acidaminococcus fermentans, clostridial cluster IX, which ferments glutamate to ammonia, carbon dioxide, acetate, butyrate and molecular hydrogen (Buckel, 1980). The dehydration is catalysed by a heterodimeric enzyme designated as component D (for dehydratase) also called HgdAB, and a dimeric component A (for activator or initiator) also called HgdC. As isolated, component D contains an oxidized [4Fe-4S]2+ cluster and reduced riboflavin-5¢-phosphate (Hans et al., 2000) but has no CoA-transferase activity. The substrate (R)-2hydroxyglutaryl-CoA is generated from acetyl-CoA and (R)-2-hydroxyglutarate by the action of a separate CoAtransferase, GctAB, which is encoded in the same operon as HgdCAB (Mack et al., 1994). Component D requires activation by ATP, MgCl2 and the reducing agent Ti(III)citrate, which is catalysed by the homodimeric, [4Fe4S]1+/2+ cluster-containing component A or HgdC. Owing to its extreme oxygen sensitivity and instability, component A could not be obtained in pure form from A.
fermentans. Overexpression of the hgdC gene in Escherichia coli, however, enabled the purification of large amounts of the recombinant enzyme with StrepTactin affinity chromatography (Hans and Buckel, 2000; Hans et al., 2000). The protein crystallized under strict anaerobic conditions and the three dimensional structure was solved by X-ray crystallography (Locher et al., 2001). 3-Hydroxyacyl-CoA esters, which are involved in the boxidation of fatty acids and clostridial butyrate synthesis, are reversibly dehydrated to enoyl-CoA by acid-base catalysis, as the a-hydrogen is activated by the thiol ester and can be removed as proton by a basic glutamate residue of the enzyme (Hofstein et al., 1999). In contrast, the dehydrations of 2-hydroxyacyl-CoA esters such as (R)-phenyllactyl-CoA or (R)-2-hydroxyglutaryl-CoA are mechanistically ‘difficult’ reactions, as the hydrogen has to be abstracted from the non-activated b-position and the a-hydroxyl group has to be eliminated adjacent to the partially positive-charged carbon of the thiol ester. It has been proposed that the activation or initiation of phenyllactate dehydratase supplies an ‘energy rich’ electron by which the thiol ester is reduced to the ketyl radical anion I (Fig. 1). This acts as a nucleophile and expels the adjacent hydroxyl group. The resulting enoxy radical can now be deprotonated to the second ketyl radical anion II, which is oxidized to the product cinnamoyl-CoA by the next incoming substrate. Thus many turnovers are possible until the electron is inadvertently lost by oxidation and renewed activation by reduction under hydrolysis ATP has to take place (Müller and Buckel, 1995; Buckel, 1996; Buckel and Golding, 1999). In this paper, we describe the characterization of a gene cluster in Clostridium sporogenes, in which the genes coding for phenyllactate dehydratase (FldABC) are arranged in the following order: fldA, fldI, fldB and fldC. The additional gene fldI encodes the initiator, which was identified after overproduction in E. coli. When fldA, fldI, fldB and part of fldC had been cloned and sequenced, it was recognized that the genome of the closely related Clostridium botulinum Hall strain A (clostridial cluster I) also comprises the four genes with 95% sequence identity. The order of these genes is similar to that of the previously identified dehydratase gene cluster of Clostridium difficile (Dickert et al., 2000) and a cluster of genes in C. botulinum, both of which probably code for another 2hydroxyacid dehydratase (had).
Results Cloning of the fldAIBC gene cluster Based on sequence identities of component A (HgdC) from A. fermentans and Clostridium symbiosum (Hans et al., 1999) with those of a putative component A from © 2002 Blackwell Science Ltd, Molecular Microbiology, 44, 49–60
Clostridial phenyllactate dehydratase 51 C. difficile (Dickert et al., 2000), now designated as HadI (Fig. 2), oligonucleotides were designed to amplify most of the fldI gene by polymerase chain reaction (PCR) with chromosomal DNA from C. sporogenes. The obtained 671-bp fragment was used as a dioxygenine-labelled probe to screen a l ZAP Express gene library of C. sporogenes. This procedure led to an E. coli XLOLR/pBK-CMV derivative (pSD19) with a 3303-bp insert flanked by EcoRI restriction sites. The sequence of the insert comprised three genes, which were identified as fldA, fldI and fldB (Fig. 3). A similar approach using sequences from the Nterminus of the CoA-transferase subunit FldA (Dickert et al., 2000), internal sequences of a related deduced protein from C. difficile, now designated as HadA, and carnitine CoA-transferase CaiB from E. coli (Eichler et al., 1994a; Elssner et al., 2001) yielded a 522-bp PCR
product, with which no clone containing the desired genes could be obtained. Sequencing showed, however, that this product (EMBL database accession no. AF420490) most probably corresponds to another CoA-transferase from C. botulinum, which is denoted as HadA (Fig. 3). With an oligonucleotide derived from the N-terminus of FldC (Dickert et al., 2000) and information from the genome sequencing projects of C. difficile and C. botulinum, four PCR products amplified from genomic DNA of C. sporogenes were used to extend the pSD19 sequence from 3303 to 4896 bp (Fig. 3; EMBL database accession no. AF420489). This combined sequence shares ≥95% (nucleic acid and amino acid) identity with a DNA fragment from C. botulinum Hall strain A (in contig 174) and around 50% amino acid identities with genes on another fragment from the same organism (in contig 105), as well
Fig. 2. Amino acid alignment of the initiators of 2-hydroxyacyl-CoA dehydratases: FldI, HgdC and HadI. The C. botulinum and C. difficile data were obtained from the Sanger centre in September 2001 (http://www.sanger.ac.uk/cgi-bin/nph-BLAST_Server.html). The percentages show the sequence identities to FldI from Clostridium sporogenes. Identical amino acids are marked by an asterisk. Conserved cysteines that are responsible for the co-ordination of the [4Fe-4S] cluster are highlighted in black. The sugar kinase ATP-binding motifs, G(I/V/L)D(I/V)G and (V/I)IDIG, are marked in grey. The cluster helix 5 is shown as a box (Locher et al., 2001). © 2002 Blackwell Science Ltd, Molecular Microbiology, 44, 49–60
52 S. Dickert et al.
Fig. 3. Arrangements of genes coding for 2-hydroxy acid dehydratases. The functions of the proteins encoded by the genes fldAIBC from C. sporogenes (Dickert et al., 2000) and this paper, as well as by gctAB, gcdA and hgdCAB from Acidaminococcus fermentans (Mack et al., 1994) and hgdCAB from Clostridium symbiosum (Hans et al., 1999) have been identified. All other genes encode deduced proteins. Genes coding for functionally identical proteins have the same colour. For abbreviations of FldLAIBC, see Fig. 1. AcdA and AcdB, acyl-CoA dehydrogenases; EtfBA and EtfDC, subunits of electron transferring flavoproteins (ETF); FldH, D-phenyllactate dehydrogenase (NAD+); LdhA and LdhB, D-lactate dehydrogenases (NAD+); HadABC, component D, and HadI, component A of a yet unidentified 2-hydroxyacyl-CoA dehydratase, probably 2-hydroxyisocaproate dehydratase; HgdAB, component D, and HgdC, component A of 2-hydroxyglutaryl-CoA dehydratase; FesA, iron sulphur protein of unknown function; GctAB, glutaconate CoA-transferase; GcdA, carboxytransferase subunit of glutaconyl-CoA decarboxylase. The names in parentheses below the genes indicate putative functions of the deduced proteins. PCR, the PCR-product from C. sporogenes that shares 95% sequence identity with hadA from Clostridium botulinum.
as with a fragment from C. difficile (in contig 1039; all three contigs are updated to September 2001; Fig. 3). The cloned fragment from C. sporogenes comprises four complete open reading frames (ORFs) designated as fldA, fldI, fldB and fldC. The fldA, fldI, fldB and fldC genes of C. sporogenes are all preceded by ‘extended’ ribosome binding sites (AAGGAGG, AAGGAAG, AAGGAGC and AAAGGAGG) typical for Gram-positive bacteria. They match the complementary sequence of the 3¢-terminus of the 16S rRNA in C. sporogenes (3¢-UCUUUCCUCC-5¢) (Hutson et al., 1993). In all four genes, translation is initiated with ATG and terminated by the ochre codon (UAA). The short intergenic regions between fldA, fldI, fldB and fldC (1 bp overlap, 4 bp and 1 bp gaps respectively) suggested that these four genes are transcribed together. Upstream of fldA, after a non-coding region of 74 nucleotides (nt), the 5¢-end of the cloned DNA translates into a sequence of 28 amino acids, 27 of which are identical (96%) to the C-
terminus of a deduced CoA-ligase (FldL) from C. botulinum (Fig. 3). FldL shares 25% sequence identities with the CoA-ligase, CaiC, from E. coli involved in carnitine metabolism (Eichler et al., 1994b) and feruloyl-CoA synthetase from Amycolatopsis sp. strain HR167 (Achterholt et al., 2000). The 3¢-end of the cloned DNA translates into a sequence of 60 amino acids, which is 100% identical to the N-terminus of a deduced acyl-CoA dehydrogenase (AcdA) from C. botulinum. AcdA shares 47% identity with the short chain acyl-CoA dehydrogenase from rat (Matsubara et al., 1989). The relatively long intergenic distance of 165 nt between fldC and acdA suggests that transcription of (fldL)fldAIBC is terminated between the two genes (Fig. 3). Analysis of the cloned genes The first ORF of the cloned gene cluster, fldA, encodes the CoA-transferase FldA containing 412 amino acid © 2002 Blackwell Science Ltd, Molecular Microbiology, 44, 49–60
Clostridial phenyllactate dehydratase 53 residues with a predicted molecular mass of 46 388 Da and a pI of 4.89. Experimentally, a molecular mass of 46 440 ± 50 Da has been determined by MALDI-TOF (Dickert et al., 2000). FldA from C. sporogenes is homologous to FldA (95% identity) and to HadA (46%) from C. botulinum, as well as to HadA from C. difficile (47%); the two HadA proteins share 66% sequence identity. The PCR product obtained with the N-terminus of FldA (accession no. AF420490, see above) shares 95% identity with HadA from C. botulinum. Hence C. sporogenes most probably contains, in addition to the four fld genes, all four had genes encoding a second 2-hydroxyacyl-CoA dehydratase system (Had) with a yet unknown substrate (Fig. 3). Identities of FldA to other characterized CoAtransferases are low (21–23%). These comprise carnitine CoA-transferase (CaiB) from E. coli (Elssner et al., 2001) and benzylsuccinate CoA-transferase (BbsE) from Thauera aromatica (Leutwein and Heider, 2001). The second ORF, fldI, encodes a protein of 264 amino acids with a predicted molecular mass of 28 077 Da and a rather high pI = 8.02. As already mentioned above, the protein was identified as activator or initiator of phenyllactate dehydratase through sequence comparisons with the well characterized component A from A. fermentans and related proteins (Fig. 2). The high pI value correlates well with the observation that the labile initiator protein does not bind to anion exchangers at pH 7.5. The third ORF fldB encodes a protein of 407 amino acids with pI = 5.87. The molecular mass predicted from the primary structure (46 239 Da, including the N-terminal methionine) was 749 ± 50 Da higher than the experimentally determined average mass of 45 490 ± 50 Da. Apparently, about six amino acids are lost during posttranslational modification. This result agrees in part with the chemically determined N-terminus, though of bad quality, which starts with the fifth amino acid resulting in a 490 Da lower mass (Dickert et al., 2000). FldB revealed sequence identities to the 2-hydroxyglutarylCoA dehydratases HgdA from A. fermentans (50%) and C. symbiosum (51%), as well as to FldB from C. botulinum (98%), and to HadB from C. botulinum (50%) and C. difficile (49%). The fourth and last ORF fldC encodes a protein with the predicted molecular mass of 43 150 Da. The high quality N-terminal sequence of 28 amino acids (Dickert et al., 2000) perfectly agreed with the DNA-derived sequence minus the N-terminal methionine. The lower molecular mass of the purified FldC as determined by MALDI-TOF (37 240 ± 50 Da) suggests that possibly about 50 C-terminal amino acids are lost by posttranslational modification. Interestingly, the potential cleavage site for this C-terminal peptide coincides with the highly conserved sequence FCDPEEXXYP. FldC from C. sporogenes shows sequence identities to the HgdB © 2002 Blackwell Science Ltd, Molecular Microbiology, 44, 49–60
subunit of the 2-hydroxyglutaryl-CoA dehydratases from A. fermentans (37%) and C. symbiosum (42%), as well as to FldC from C. botulinum (95%), and HadC from C. botulinum (48%) and C. difficile (49%). The organization of the genes coding for 2hydroxyacyl-CoA dehydratases is fairly conserved within the clostridia, as displayed in Fig. 3. The initiators or components A (FldI, HadI and HgdC) are encoded upstream of the two subunits of 2-hydroxyacyl-CoA dehydratase (FldBC, HadBC and HgdAB); only the hadI gene from C. botulinum (contig 105) is shifted downstream by 8.4 kb when compared with C. difficile. It is inserted between two genes coding for putative permeases. Sequence identities between the corresponding had genes from C. botulinum and C. difficile (hadA 66%, hadB 81%, hadC 65% and hadI 68%) indicate that they are closer related to each other than to the fld genes of the same organism (ca. 50%). Contig 105 contains an additional set of dehydratase genes, which are more related to hgd genes coding for 2-hydroxyglutaryl-CoA dehydratase. The CoAtransferase genes fldA (C. sporogenes and C. botulinum) and hadA (C. difficile) are located directly upstream of the initiator genes fldI and hadI. In C. sporogenes and C. botulinum, a CoA-ligase gene fldL is found upstream of fldA, but in both had clusters the region of c. 700 bp upstream from hadA does not seem to code for any protein. Further upstream in C. difficile follows ldhA, whose deduced amino acid sequence has 41% identity to D-lactate dehydrogenase from Lactobacillus plantarum. The fldC and hadC genes are followed by genes coding for shortchain acyl-CoA dehydrogenase (acdA and acdB respectively) and electron-transferring flavoprotein (ETF) b and a subunits (etfBA and etfDC respectively). In previous work, three additional enzymes were purified from C. sporogenes: D-phenyllactate dehydrogenase (FldH), cinnamate reductase (FldZ) and a putative serinehydroxymethyltransferase (Dickert et al., 2000). The Ntermini of the three proteins closely matched those deduced from the genes found in C. botulinum. Identification of FldI as component A of phenyllactate dehydratase The gene fldI was cloned into the expression vector pASK-IBA 3 by PCR amplification with flanking BsaI restriction sites using plasmid pSD19 as template, as described in Experimental procedures. The resulting plasmid, pSD50 (pASK-IBA 3/fldI), was transformed into E. coli MRF¢ and after induction of the tet-promoter with anhydrotetracycline, overproduction of FldI was monitored by SDS–PAGE. The recombinant protein was thereby fused with a C-terminal Strep II-Tag peptide of the sequence NH2-Trp-Ser-His-Pro-Gln-Phe-Glu-Lys-COOH (Schmidt et al., 1996), which allowed a one step purifica-
54 S. Dickert et al. tion of the protein by affinity chromatography on streptavidin linked to a Sepharose matrix (Hans and Buckel, 2000). Upon elution with D-desthiobiotin, the pure recombinant protein was obtained (Fig. 4) with a yield of 20 mg FldI from 10 l of culture broth. During purification and all successive experiments, FldI was stabilized with 1 mM ATP and 5 mM MgCl2 under an atmosphere of 95% N2 and 5% H2. The Strep-Tactin methodology enlarged the predicted molecular mass of FldI from 28 kDa to 29 kDa, which is consistent with SDS–PAGE, in which the FldI band migrates just below the 30 kDa marker protein (Fig. 4). FldI is extremely sensitive towards exposure to oxygen, which resulted in denaturation and precipitation of the
Fig. 4. Purification of recombinant initiator FldI monitored by SDS–PAGE. M, molecular mass markers; 1, cell-free extract of E. coli MRF¢/pSD50; 2 and 3, 7 and 30 mg protein eluted from the Strep-Tactin Sepharose respectively.
protein. The half-life under air is less than 1 min, therefore, all experiments were performed under strictly anaerobic conditions (Hans et al., 2000). In the reduced state, FldI showed low ATPase activity (2 s–1. Similar results have been obtained with component A from A. fermentans (Table 1) (M. Hans and W. Buckel, unpublished results). Chemical analysis of FldI revealed 4.1 ± 0.5 non-haem iron and 2.3 ± 0.5 acid-labile sulphur per homodimer. Significant amounts of acid-labile sulphur could have been lost during the necessary exposure of the protein to air, as an odour of H2S was detected. FldI was able to initiate catalysis of phenyllactate dehydratase FldABC in the presence of ATP, MgCl2, catalytic amounts of cinnamoyl-CoA and dithionite as an in vitro electron donor. (R)-Phenyllactate was dehydrated to cinnamate at a rate of 0.78 s–1, which was detected spectrophotometrically and by high-performance liquid chromatography (HPLC) (Dickert et al., 2000). Interestingly, the initiator FldI can be replaced by the recombinant component A (HgdC) from A. fermentans (Hans et al., 2000) without a significant decrease in the final dehydratase activity (0.69 s–1). Hence, both initiator proteins are effective in activation of (R)-phenyllactate dehydratase component D. For measurements of maximal enzymatic activity, FldABC and FldI were used in 1:1 to 1:2 ratios, which immediately resulted in full activation. Upon decrease of the ratio to 1:0.01, it was found that the dehydratase activity increased within 2 min from initially 0.46 s–1 up to 0.76 s–1, and then remained constant. This implies that one molecule of FldI can activate many dehydratase FldABC molecules. Lag phases of up to 20 min have been observed when a cell-free extract of C. sporogenes was used to initiate catalysis of the phenyllactate dehydratase (Dickert et al., 2000). Apparently, the cellfree extract contained only a small amount of active FldI. The experiments on the activity of recombinant FldI Table 1. Catalytic properties of phenyllactate dehydratase (FldABC) and its initiator (FldI) from Clostridium sporogenes as well as comparison with 2-hydroxyglutaryl-CoA dehydratases (HgdAB) and the initiator (HgdC) from other organisms.
ATPase activity Oxidized initiator FldI Reduced initiator FldI Oxidized HgdC (A. fermentans) Reduced HgdC (A. fermentans)
>2.0 s-1
