The human pathogenic bacterium Clostridium difficile is a versatile organism concerning its ability to ferment amino acids. The formation of p-cresol as the main fermentation product of tyrosine by C. difficile is unique among clostridial species. The enzyme responsible for p-cresol formation is p-hydroxyphenylacetate decarboxylase. The enzyme was purified from C. difficile strain DMSZ 1296T and initially characterized. The N-terminal amino-acid sequence was 100% identical to an open reading frame in the unfinished genome of C. difficile strain 630. The ORF encoded a protein of the same size as the purified decarboxylase and was very similar to pyruvate formate-lyase-like proteins from Escherichia coli and Archaeoglobus fulgidus. The enzyme decarboxylated p-hydroxyphenylacetate (Km = 2.8 mm) and 3,4-dihydroxyphenylacetate (Km = 0.5 mm). It was competitively inhibited by the substrate analogues p-hydroxyphenylacetylamide and p-hydroxymandelate with Ki values of 0.7 mm and 0.48 mm, respectively. The protein was readily and irreversibly inactivated by molecular oxygen. Although the purified enzyme was active in the presence of sodium sulfide, there are some indications for an as yet unidentified low molecular mass cofactor that is required for catalytic activity in vivo. Based on the identification of p-hydroxyphenylacetate decarboxylase as a novel glycyl radical enzyme and the substrate specificity of the enzyme, a catalytic mechanism involving ketyl radicals as intermediates is proposed.
Clostridium difficile is a spore forming, strictly anaerobic bacterium that causes gastrointestinal infections in humans ranging from asymptomatic colonization to severe diarrhoea, pseudomembranous colitis, toxic megacolon, colonic perforation and occasionally death [1–5].
The organism grows on a variety of different media and has been shown to be among the most versatile clostridia concerning its ability to ferment amino acids . With the exception of histidine, glutamate, glutamine and lysine, the proteinogenic amino acids are fermented by C. difficile to yield a complex mixture of volatile fatty acids. This metabolic diversity may allow the organism to survive under nutritional limitations.
The aromatic amino acids, phenylalanine, tryptophan and tyrosine, are predominantly oxidized to yield phenylacetate, indoleacetate and p-hydroxyphenylacetate (pHPA), respectively. Uniquely among clostridial species, C. difficile can further decarboxylate pHPA to yield p-cresol. It has been shown that this compound is formed by C. difficile in significant amounts  and that the organism can survive in media containing up to 35 mmp-cresol . Phenolic antioxidants are known to act bacteriostaticly  and it has been shown that shock loads of p-cresol at 1 mm concentrations reduced the number of cultivable bacteria in sewage plant slurries by three orders of magnitude . Moreover, protozoa and metazoa are completely eliminated. As the concentration of p-cresol in C. difficile-cultures can easily exceed this concentration , the formation of this compound could allow an active suppression of other microorganisms by C. difficile under competitive growth conditions.
The decarboxylation of arylacetates according to Eqn (1) to yield either toluene, skatole (3-methylindole) or p-cresol has been shown for a very limited number of specialized organisms. The decarboxylation of phenylacetate to yield toluene has been reported for Clostridium aerofoetidum and Tolumonas auensis. Clostridium scatologenes, a Lactobacillus strain  and organisms belonging to the genera Rhizobium and Pseudomonas have been shown to produce skatole from indoleacetate. A Lactobacillus strain  and C. difficile are known to produce p-cresol by decarboxylation of pHPA.
Only two of these reactions (the formation of p-cresol by C. difficile and the formation of skatole by a Lactobacillus strain ) have been initially characterized in cell free extracts prepared from the organisms. The enzymatic systems responsible for the decarboxylations were unstable and highly sensitive towards molecular oxygen. The instability of the activity and the requirement of an as yet not characterized low molecular mass cofactor that was lost rapidly during chromatographic steps or by ultra filtration prevented the purification and detailed characterization of the enzymes.
During the last decades the participation of radicals as reactive intermediates in enzymatic catalysis of chemically difficult reactions has become evident, in particular in the metabolism of anaerobes . A mechanism involving one electron oxidation of pHPA has been proposed as initial step of the p-cresol formation by C. difficile: the p-phenoxyacetyl radical thus formed decarboxylates to a ketyl radical, which is protonated to the p-methylphenoxy radical. This intermediate is then reduced to the p-cresolate anion and protonated to yield the final product, p-cresol. This proposed mechanism is attractive as the ‘Umpolung’ of the phenolate enables the stabilization of the negative charge remaining at the aromatic nucleus when carbon dioxide is released, which is not possible for pHPA. The participation of radical intermediates in enzyme-catalysed reactions is well supported by experimental data and an oxidative/reductive radical formation has been envisaged for a variety of reactions . Although such a mechanism could explain how this chemically difficult reaction proceeds, the radical formation is a critical step in this proposal. Buckel and Golding  suggested an oxidation of pHPA mediated by redox-active cofactors such as flavins or iron–sulfur clusters to yield the radical intermediate of the decarboxylation reaction. However, the radical formation could alternatively be mediated by a homolytic abstraction of a hydrogen atom (Fig. 1). Such reactions require starter radicals for catalysis. These radicals are provided either by homolysis of a carbon-cobalt bound in vitamin B12-dependent enzymes or by enzymes that contain a persistent protein-based glycyl radical (reviewed in ).
As the proposal of radicals as catalytic intermediates in the decarboxylation of pHPA was not supported by experimental evidence, we decided to purify and characterize the enzyme. In this communication we report the initial characterization of pHPA decarboxylase as a novel glycyl radical enzyme.
Materials and methods
C. difficile (DSMZ 1296T) was purchased from the Deutsche Sammlung für Mikroorganismen und Zellkulturen (Braunschweig, Germany). All chemicals were purchased either from Sigma-Aldrich (Deisenhofen, Germany) or Lancaster (Mühlheim, Germany) and all were of the highest quality available. p-Hydroxyphenylacetyl-CoA was synthesized from pHPA, CoA and N,N′-carbonyldiimidazole according to the procedure described by Kawaguchi et al. . The product was purified by preparative reversed-phase HPLC and its identity was confirmed by MALDI-TOF MS .
Organisms and cultivation
C. difficile spores were germinated in an anoxic peptone/yeast extract/glucose medium (PYG). This medium was also used for storage of cultures as the organism sporulated readily at 4 °C. Growing cultures were maintained in a casein amino-acid medium as described previously . Cells for enzyme purification were cultivated at 2 L scale for 16–18 h at 37 °C in a defined medium  supplemented with 10 mmpHPA. The sterile media were inoculated with 1% (v/v) of an exponential phase preculture in PYG-medium. Samples were taken at various time points for determination of the optical density (590 nm) and the p-cresol concentration was determined by RP-HPLC. The cells were harvested by centrifugation, washed with homogenization buffer and used immediately.
Reversed-phase HPLC separation of aromatic compounds
Several substituted phenylacetic acids and their predicted decarboxylation products (Table 1) were separated using a LiChroCART™ 250 × 4 mm HPLC cartridge (Merck, Darmstadt) filled with LiChrospher™ 100 RP-18 (5 µm) and operated at a flow rate of 1.2 mL·min−1 at 50 °C. The eluent contained 0.1% (v/v) trifluoroacetic acid in acetonitrile/water. The acetonitrile concentration was adjusted to allow elution of the arylacetic acid and the toluene derivative within approximately 5–10 min. Aromatic amines were analysed using a 3.9 × 250 mm B195A OD-5–100, SAM-C18 column (Ict, Bad Homburg) operated at a flow rate of 1 mL·min−1 at 50 °C with 0.1% (v/v) triethylamine in water/acetonitrile.
Table 1. Reversed-phase HPLC separation of potential substrates and predicted decarboxylation products. The table summarizes the HPLC methods developed for the analysis of potential pHPA decarboxylase substrates. The capacity factors (k) for the substrate and the predicted product are shown. The chromatographic conditions are described in the text. Unless otherwise stated, 30% (v/v) acetonitrile was used as mobile phase and the products were measured at 275 nm. Capacity factor k = (Va–Vt)/Va where Va = elution volume of the analyte and Vt = void volume.
Unless stated otherwise, wet packed washed cells (1–2 g) of C. difficile were suspended in 100 mm Tris/HCl, pH 7.5, 5 mm ammonium sulfate, 1 mm magnesium chloride, 1 mm sodium sulfide, 0.5 mm sodium dithionite (buffer A, 25 mL) and broken by sonification. Cell debris was removed by centrifugation for 30 min at 100 000 g. The enzyme activity was measured by a discontinuous anoxic assay determining the amount of p-cresol formed during the incubation period. Samples were incubated in buffer A supplemented with 25 mmpHPA for 10 min at 30 °C. The reaction was stopped with perchloric acid and neutralized with potassium carbonate containing phenol (1 mm final concentration) as internal standard. Denaturated protein and potassium perchlorate were removed by centrifugation at 4 °C and the samples were analysed by RP-HPLC as described above.
Cell free extracts for pH-stability tests were prepared in Tris/HCl buffer (pH 7.0, 10 mm). Buffer solutions (0.2 vol.) containing 250 mm each boric acid, citric acid, sodium dihydrogenphosphate and Tris (adjusted to the desired pH with 1 m NaOH) were added. The samples were incubated at 0 °C or 30 °C and tested in various time intervals for activity. The pH dependence of p-cresol formation was measured in the same buffer supplemented with 5 mmpHPA and 1 mm sodium sulfide.
Preparation of protein free extracts and influence of coenzymes
Wet packed cells (2 g) of C. difficile were suspended in 50 mL of 25 mm Tris/HCl, pH 7.5, and homogenized. Cell debris was removed by centrifugation. The supernatant was adjusted to a final concentration of 450 mm trichloroacetic acid and stirred on ice for 15 min. The precipitated proteins were removed by centrifugation and the supernatant (5 mL) was applied on a Sep-Pak C18 cartridge (1 mL bed volume, Waters) equilibrated with 0.1% trifluoroacetic acid. The column was washed with 20 mL of 0.1% trifluoroacetic acid and eluted with 5 mL 0.1% trifluoroacetic acid in 50% (v/v) acetonitrile. The solvent was evaporated and the extract was re-dissolved in anoxic potassium phosphate buffer (50 mm, pH 7.0, 1 mL). The extract (50 µL) was added prior the measurements to cell free extract inactivated by gel filtration on Sephadex G25 equilibrated with 50 mm potassium phosphate, pH 7.0, 1 mm dithiothreitol.
The influence of coenzymes and cofactors was addressed adding coenzymes (FAD, FMN, riboflavin, NAD(P), NAD(P)H, CoA, acetyl-CoA, p-hydroxyphenylacetyl-CoA, thiamine-pyrophosphate, thiamine-monophosphate, adenosylcobalamine (vitamin B12) and pyridoxal-phosphate, 50 µm) or adenosine nucleotides (ATP, ADP and AMP, 250 µm) to the assay mixture in presence and in absence of sodium sulfide.
Substrate specificity and kinetics
The substrate specificity of the enzyme in cell free extracts was addressed adding the phenylacetic acid derivatives (5 mm) to the standard assay buffer. The mixtures were incubated for 60 min at 30 °C and the product formation was analysed by HPLC. Positive results were confirmed with pure enzyme.
The Km values were determined for substrate concentrations ranging from 0.5 mm to 5 mm. The products formed were quantified by HPLC. The activity data were fitted to the Michaelis–Menten equation (V = (Vmax*[S])/(Km + [S]) using origin 5.0. The inhibition constants were obtained from K′m determinations at inhibitor concentrations ranging from 0 mm to 2 mm. The Ki values were calculated from the linear regression of the apparent Km values vs. the inhibitor concentration.
Inactivation of pHPA decarboxylase by molecular oxygen
Cell free extracts were adjusted to the desired final concentration of molecular oxygen with air-saturated water (235 µm at 25 °C). The samples were either mixed and immediately analysed for residual activity or incubated for 10 min at 25 °C prior to the measurements.
Purification of pHPA decarboxylase
The purification procedure for pHPA decarboxylase was performed entirely in an anoxic glove box under an atmosphere of N2/H2 (95/5). All buffers contained 100 mm Tris/HCl, pH 7.5, 5 mm ammonium sulfate, 1 mm magnesium chloride and 1 mm dithiothreitol, were boiled and the head-space atmosphere was replaced by N2. Sodium dithionite (0.5 mm final concentration) was added and the buffers were stirred over night. Cell free extract was applied to DEAE-Sepharose Fast Flow (1.6/20, Pharmacia, Freiburg) equilibrated with this buffer (buffer B). The proteins were eluted by a linear gradient from 0 to 500 mm NaCl in 300 mL with a flow rate of 4 mL·min−1. Fractions exhibiting pHPA decarboxylase activity were diluted 1 : 1 with buffer B and applied to a Resource-Q column (Pharmacia). The proteins were eluted by a linear gradient from 0 to 300 mm NaCl in 30 mL with a flow rate of 2 mL min−1. pHPA decarboxylase-containing fractions were concentrated and applied to a Superdex 200 column (1/30, Pharmacia) equilibrated with 150 mm NaCl in buffer B (Flow rate 0.4 mL·min−1). The purity at different purification stages was monitored by SDS/PAGE  and Coomassie staining.
Sequencing and data base searches
Purified pHPA decarboxylase (100 pmol) was subjected to SDS/PAGE and blotted in a wet blot apparatus onto a PVDF membrane. The protein was stained, the polypeptide bands were cut out and subjected to Edman degradation chemistry using a Procise-cLC sequencer (ABI). Internal peptides were generated from pHPA decarboxylase (100 pmol) subjected to SDS/PAGE and stained with Coomassie blue. The polypeptides were cut out and digested with trypsin . The peptides were extracted from the gel and analysed by MALDI-TOF-MS using Voyager-DE/RP-instrument (PE-Biosystems, Wiesbaden) .
Cell free extracts of C. difficile were prepared from cells grown in the presence of 10 mmpHPA in a defined medium containing glucose and leucine as carbon and energy sources . The cells were cultivated in 2 L culture volumes and grew typically with doubling times of around 3 h to yield around 1 g wet packed cells per l medium. As shown in Fig. 2, the formation of p-cresol correlated well with growth. The maximum specific activities of pHPA decarboxylase were found upon harvest of late exponentially/early stationary cells and were significantly higher (up to 50 U per g cell protein) than reported earlier for cells grown on a peptone-yeast extract medium (up to 4 U per g cell protein) . Neither in control medium without pHPA nor in medium supplemented with tyrosine were significant amounts of p-cresol formed within 24 h. Decarboxylase activity was also not detectable in extracts prepared from cells grown in these media. In the presence of pHPA the growth of C. difficile was slightly slower and cell yields were reduced by around 15%. It is worth noting that C. difficile produced toxins and spores in the peptone-yeast extract medium but not in the defined medium. Therefore, activity yields and safety considerations suggested the use of this medium for cell growth.
Initial characterization of the enzyme in cell free extracts
The formation of p-cresol by cell free extracts of C. difficile was measured under anoxic conditions with pHPA as substrate. The amount of p-cresol formed during the incubation period was quantified by reversed-phase HPLC. One unit of pHPA decarboxylase activity corresponds to the formation of 1 µmol of p-cresol per min at 30 °C. The formation of p-cresol by cell-free extracts of C. difficile was not linear in time. The enzyme was rapidly (t1/2 ≈ 3 min) and irreversibly inactivated. The decarboxylase was not inhibited by p-cresol at concentrations produced in the assay. The enzyme was significantly more stable (t1/2 > 15 min at 30 °C) in the absence of substrate.
In preliminary experiments, the cell free extracts were prepared in 50 mm potassium phosphate buffer, pH 7.0, supplemented with 1 mm dithiothreitol, and measured in the same buffer supplemented with 5 mmpHPA . The activity was unstable in this buffer even at 0 °C and losses of more than 50% of the initial activity were observed after 24 h. Initial attempts to purify the protein failed as the enzymatic activity was entirely lost after chromatographic steps. The inactive protein fractions obtained (e.g. by gel filtration) were not reactivated by either coenzymes or adenosine nucleotides (see Materials and methods). The activity was restored by a protein-free extract, which had been prepared by aerobic trichloroacetic acid precipitation and was partially purified by solid-phase extraction on C18-cartridges.
The enzyme was much more stable in 100 mm Tris/HCl, pH 7.5, supplemented with 5 mm ammonium sulfate, 1 mm magnesium chloride, 1 mm dithiothreitol and 1 mm sodium sulfide and more than 90% of the initial activity was recovered after 5 days at 0 °C. Moreover, the enzymatic activity was at least partially recovered after column chomatography. A closer analysis revealed that sodium sulfide (1 mm) was essentially required for activity, whereas ammonium sulfate (5 mm) and magnesium chloride (1 mm) increased the stability of the enzyme. In the presence of these supplements the enzyme was stable in a pH-range from 6 to 9 at 0 °C but decayed with half life times of around 15 min at 30 °C. A broad pH-dependence for the formation of p-cresol was observed, exhibiting a maximum around pH 7 and more than 50% of this activity at either end of the range. The enzyme was only slightly affected by increasing concentrations of sodium chloride. Concentrations of up to 200 mm did not affect the activity and even at 800 mm sodium chloride 50% of the control activity was found.
The substrate specificity of the decarboxylase was addressed adding various phenylacetic acid derivatives (5 mm, Table 1) to the standard assay. The samples were analysed and the predicted decarboxylation products were quantified by HPLC. The positive results thus obtained were confirmed qualitatively with pure enzyme. Only phenylacetates that contained a hydroxyl group in p-position were substrates: pHPA, 3,4-dihydroxyphenylacetate and p-hydroxymandelate were decarboxylated to yield p-cresol, 4-methylcatechol and p-hydroxybenzylalcohol, respectively. The latter substrate is commercially available only as a racemic mixture of the enantiomers. Further experiments will have to establish whether only one of these is used as substrate and whether the other acts as a competitive inhibitor. pHPA and 3,4-di-hydroxyphenylacetate, which is an end product of the intestinal dihydroxyphenylalanine (DOPA) metabolism, were used by the enzyme with similar specificity (Table 2). In addition to the alternative substrates 3,4-dihydroxyphenylacetate and p-hydroxymandelate, p-hydroxyphenylacetamide was found to act as a competitive inhibitor of pHPA decarboxylation. Neither m- or o-hydroxyphenyl-acetate nor p-hydroxybenzoate or p-hydroxyphenylpropionate inhibited the reaction.
Table 2. Substrates and competitive inhibitors of p-hydroxyphenylacetate decarboxylase. The apparent Km values were determined as described in Materials and methods. The inhibition constants (Ki) were determined for the formation of p-cresol from pHPA.
Specific activity (mU·mg−1)
Specific activity/Km (mU·mg−1·mm−1)
The pHPA decarboxylase was inactivated by low concentrations of dissolved molecular oxygen. When cell free extracts were incubated in the presence of increasing concentrations of oxygen, the activity was rapidly and irreversibly lost (Fig. 3). The midpoint inhibitory concentration of molecular oxygen varied with the preparation of the cell free extract and the concentration of protein, but always occurred at low concentrations and within a narrow concentration range. The variable threshold for the inactivation of pHPA decarboxylase was attributed to the oxygen consumption by reduced flavoproteins and iron–sulfur proteins present in cell-free extracts.
Purification of the enzyme
Based on the initial characterization of the pHPA decarboxylase in cell free extracts, a scheme was developed that allowed purification of the enzyme. As summarized in Table 3 and shown in Fig. 4, the enzyme was essentially pure after three chromatographic steps. The purification protocol consists of two subsequent anion exchange columns (DEAE-Sepharose and Resource-Q) followed by gel filtration on Superdex 200. All purification steps were carried out in 100 mm Tris/HCl, pH 7.5, supplemented with 5 mm ammonium sulfate, 1 mm magnesium chloride, 1 mm dithiothreitol and 500 µm sodium dithionite under strictly anoxic conditions.
Table 3. Purification of pHPA decarboxylase from C. difficile. The enzyme activity was measured with pHPA (25 mm) as substrate at pH 7.5. One unit of pHPA decarboxylase corresponds to the formation of 1 µmol p-cresol per min at 30 °C.
Specific activity (mU·mg−1)
Total Activity (mU)
Purification factor (fold)
The enzymatic activity was bound to DEAE-Sepharose and eluted between 220 mm and 260 mm sodium chloride. Around 95% loss of activity was inevitable in this step. Moreover, the enzymatic activity was highly unstable after this column. Therefore, the fractions were stored on ice immediately and diluted rather than dialysed or concentrated prior application on a Resource-Q column. From the Resource-Q column the activity eluted between 140 mm and 160 mm sodium chloride in good yield (around 35%). It is worth noting that the enzyme was much more stable after this column and could be concentrated by membrane centrifugation prior application onto a Superdex 200 column.
Applying this protocol, between 50 µg and 100 µg of enzyme were purified from 2 g wet packed cells. As demonstrated in Fig. 5, the final preparation contained two polypeptides with apparent molecular masses of 110 kDa and 105 kDa, respectively. The Kav value observed by gel filtration (0.76–0.79) was used to calculate a molecular mass of 200 kDa for the native decarboxylase.
The purified protein was subjected to SDS/PAGE and blotted onto a PVDF-membrane. The subunits were cut out and subjected to N-terminal sequencing using Edman-degradation chemistry. Both polypeptides yielded identical N-terminal amino-acid sequences (Fig. 5). These findings suggested that the smaller subunit of the enzyme is a C-terminally truncated form of the polypeptide. Considering the native molecular mass of 200 kDa and the approximately 1 : 1 ratio of both subunits in the SDS/PAGE, an α,α′-subunit composition for the native protein can be proposed.
Data base searches
The N-terminal sequence was used to screen the unfinished genome sequence data of strain 630 (epidemic type X) that were produced by the C. difficile Sequencing Group at the Sanger Centre and can be obtained from http://www.sanger.ac.uk/Projects/C_difficile/blast_server.shtml. The search produced a high-scoring segment pair (100% identity) with the 3′-region of contig963.1 (date: 12/20/00, Length = 35842 bp, frame + 3). The N-terminal amino-acid sequence of the purified decarboxylase was located at the N-terminus (the start methionine was missing) of a protein encoded by an open reading frame for a 902 amino-acid protein with a predicted molecular mass of 101 294 Da. A tryptic in-gel digest of the purified decarboxylase yielded 11 internal peptides with masses identical within the experimental accuracy (deviation < 500 p.p.m.) to predicted tryptic peptides of the protein sequence encoded by the ORF in the genome (Fig. 6).
The translated amino-acid sequence of this ORF was used to perform a PIR BLAST search . The search produced 11 hits, encoding pyruvate formate-lyases (PFL) and similar proteins. Three entries showed a full length overlap (B65202, PFL-2, E. coli;G64819, PFL-3, E. coli;H69430, PFL-2(pflD), Archaeoglobus fulgidus) [28,29]. These proteins showed between 25% and 28% identity to the pHPA decarboxylase from C. difficile. A clustalw multiple sequence alignment was generated for these proteins and the translated C. difficile ORF. As shown in Fig. 6, the conserved amino acids within these sequences were not randomly distributed along the sequences but were clustered, particularly in the C-terminal region. The fingerprint motif highly indicative for glycyl-radical enzymes (I/V-R-I/V-X-G-F/W/Y) which is located close to the C-terminus of these proteins was clearly identified.
Glycyl-radical enzymes are known to be post-translationally activated by an iron–sulfur protein which is encoded by a gene usually located in the close vicinity of their structural genes. Such a gene was identified starting 226 bp downstream of the decarboxylase gene (Fig. 6). Further downstream of the activating enzyme a third ORF was identified which was not similar to any protein of known function in the databases.
The pHPA decarboxylase from C. difficile is the first arylacetate decarboxylase to be purified and characterized in detail. The N-terminal amino-acid sequence of the protein was identical to protein similar to pyruvate formate-lyase (PFL) encoded by an open reading frame in the unfinished genome of C. difficile strain 630.
PFL is a key enzyme in the anaerobic degradation of glucose (e.g. E. coli and other bacteria) and catalyses the reversible reaction of pyruvate with CoA to form formate and acetyl-CoA. The enzyme has been studied intensively during the last decades and was the first example of a radical enzyme in which the radical is located at the polypeptide backbone on a specific glycine residue [30,31]. At the present time, two groups of glycyl radical enzymes other than PFLs are known: anaerobic ribonucleotide reductase (class III RNR) (reviewed in [32–34]) and benzylsuccinate synthase from Thauera aromatica. In contrast to the class III RNRs which are essential for the provision of deoxyribonucleotides for DNA synthesis, benzylsuccinate synthase is a nonubiquitous enzyme which catalyses the highly unusual first step in the anaerobic degradation of toluene, the addition of the methyl-group of toluene to the double bond of fumarate . In addition to these well-characterized enzymes, numerous PFL-like proteins emerged from genome projects. Accurate annotation of these must await identification of their physiological functions. Our functional characterization of the ORF in the C. difficile genome provides further insight into the diversity of reactions which can be carried out with glycyl radical enzymes.
The crystal structures of PFL and RNR have recently been published [37–39] and the mechanisms of both enzymes have been discussed intensively. In the resting state of these enzymes, the radical is located at a specific glycine residue and is stabilized by captodative effects . The glycine residue that carries the radical is located within a fingerprint sequence (I/V-R-I/V-x-G-F/W/Y) close to the C-terminus which was also observed in the genome derived sequence of pHPA decarboxylase. During catalysis the radical moves to an active centre cysteinyl residue [40–42]. PFL and the closely related ketoacid formate-lyase (TdcE ), contain a pair of neighbouring cysteines, which are essential for catalysis and are thought to act as a radical relay (see Fig. 7). Only one of these cysteines can be identified in pHPA decarboxylase (Cys506). A second cysteine residue could be Cys619 which is conserved among pHPA decarboxylase and some similar proteins from other organisms. Site directed mutagenesis experiments addressing the significance of these cysteines for the catalytic action of pHPA decarboxylase must await the cloning of the gene, which is in progress in our laboratory.
The glycyl radical is introduced into the apo-protein of glycyl radical enzymes by a specific activating enzyme [44–46] that requires S-adenosylmethionine and reduced flavodoxin. The genes enconding the activases are usually located in close proximity of the PFL structural genes. In the genomic DNA of C. difficile, a gene encoding a protein similar to such activating enzymes is located starting 226 bp downstream of the pHPA decarboxylase gene.
In addition to these two structural genes of known function, a third ORF was identified in the pHPA decarboxylase operon (Fig. 7). The protein encoded by this gene was not similar to any protein in the data bases and as yet no function can be assigned to this gene product.
The identification of pHPA decarboxylase as a novel glycyl radical enzyme is supported by biochemical evidence. It has been shown that PFL  and other glycyl radical enzymes [35,47,48] are rapidly and irreversibly inactivated by molecular oxygen and that the inactivation leads to a polypeptide backbone cleavage at the position of the glycyl radical. The purified pHPA decarboxylase consisted of two N-terminally identical polypeptides with slightly different molecular masses in an approximately 1 : 1 ratio, suggesting an αα′ oligomeric structure. As it has been shown for PLF that only one polypeptide chain of the homodimeric enzyme contained a glycyl radical  and is therefore susceptible to oxygen induced backbone cleavage, the partial truncation observed for oxygen exposed pPHA decarboxylase is in good agreement with this type of inactivation.
Thus far, the biochemical characterization of the pHPA decarboxylase was performed in cell-free extracts rather than using purified enzyme. Initial attempts to purify the enzyme were not successful as the pHPA decarboxylase was completely inactivated when passed through chromatographic columns. Supporting the observations of D’Ari and Barker , we found that the enzyme is reversibly inactivated removing the low molecular mass fraction of the cell-free extract. All attempts to replace the extract by a variety of coenzymes, cofactors or by adenosine nucleotides, either alone or in combinations, failed. The replacement of the low molecular mass fraction in cell-free extract by sodium sulfide enabled the purification of the enzyme in an active state but severe losses of activity were inevitable throughout the purification. The low activity yields for the purification (≈ 0.5%) and the limited culture volumes that can be handled for pathogenic organisms restrained the purification of large amounts of active protein. Establishing the identity of the cofactor may improve the yield and specific activity of the enzyme. Consequently, our future work will focus on the identification and purification of the in vivo cofactor.
The decarboxylation of pHPA is hard to achieve under physiological conditions. The hydroxyl group increases the electron density at the aromatic ring and the negative charge remaining at the aromatic nucleus is not stabilized when carbon dioxide is released. In order to avoid this mechanistic problem, Buckel and Golding  suggested a one electron oxidation of the substrate to form a p-phenoxyacetyl radical. As this oxidation gives rise to an ‘Umpolung’ of the phenolate, the negative charge is stabilized by a ketyl radical (a radical anion) intermediate. The finding that pHPA decarboxylase is a glycyl radical enzyme supports this general mechanism but requires a revision of the suggested mechanism of radical formation. In the original model an initial deprotonation of pHPA has been proposed to form the phenolate dianion followed by oxidation of the latter mediated by either a flavine or an iron–sulfur centre to yield the p-phenoxyacetyl radical. The participation of an active centre thiyl radical and the essential requirement of a hydroxyl group in the p-position of the phenyl ring suggests a radical formation by homolytic cleavage of the O-H bond (Fig. 1) rather than by oxidation of the phenolate. Hydrogen abstraction by glycyl radical enzymes to form substrate derived radicals is not without precedence in literature. It has been suggested as initial step in the catalytic mechanisms of benzylsuccinate synthase  and of ribonucleotide reductases .
It is remarkable that the abstraction of a hydrogen atom could also activate indoleacetate for decarboxylation. As this reaction is carried out by some organisms [13–16,50] under strict anoxic conditions, it seems reasonable to conclude that an enzyme similar to pHPA decarboxylase is responsible for the reaction. Although this enzyme has not been purified thus far, indoleacetate decarboxylase activity from a Lactobacillus strain  and from C. scatologenes (P. I. Andrei, unpublished results) have been initially characterized in cell-free extracts and have been found to be rapidly inactivated by molecular oxygen.
The formation of p-cresol by growing cells of C. difficile is known for a long time . It has been shown previously that the p-cresol formation is stimulated by the addition of pHPA to the medium but as yet no data are available on the regulation of pHPA decarboxylase activity. Our data clearly indicate that production of p-cresol by growing cells of C. difficile is regulated independently from the tyrosine metabolism. In all experiments no or only negligible amounts of p-cresol were found in media containing tyrosine after 24 h and the activities of the decarboxylase in cell-free extracts were absent or very low. In contrast, in cultures containing pHPA, high p-cresol-concentrations were found and high specific activities of the pHPA decarboxylase were induced.
As it has been shown earlier that C. difficile can survive in media containing up to 35 mmp-cresol  and that phenolic compounds act bacteriostaticly [9,10], it appears likely that the formation and secretion of this toxic compound could provide a significant advantage for C. difficile under highly competitive growth conditions. In cases of antibiotic-associated infection of the human gut, the formation of p-cresol could account for the ongoing suppression of the intestinal microbiota under certain nutritional conditions and therefore affect the seriousness and progression of C. difficile-associated diseases.
We thank W. Buckel, A.J. Pierik, D.J. Darley and A.K. Croft for fruitful and stimulating discussions, M. Khatib and I. Çinkaya for their contribution to the initial work, B. Schmidt and K. Neifer from the Centre for Molecular Biology and Biochemistry (Göttingen) for protein sequencing and the C. difficile Sequencing Group at the Sanger Centre for providing the unpublished genome sequence data. The project belongs to the priority program ‘Radicals in the enzymatic catalysis’ and was supported by the Deutsche Forschungsgemeinschaft.