Báo cáo khoa học: Subunit composition of the glycyl radical enzyme p-hydroxyphenylacetate decarboxylase doc

Subunit composition of the glycyl radical enzyme
p
-hydroxyphenylacetate decarboxylase
A small subunit, HpdC, is essential for catalytic activity
Paula I. Andrei
1
, Antonio J. Pierik
1
, Stefan Zauner
2
, Luminita C. Andrei-Selmer
3
and Thorsten Selmer
1
1
Laboratorium fu
¨
r Mikrobiologie, Fachbereich Biologie, Philipps-Universita
¨
t, Marburg, Germany;
2
Institut fu
¨
r Zellbiologie und
angewandte Botanik, Fachbereich Biologie, Philipps-Universita
¨
t, Marburg, Germany;
3
Institut fu
¨
r Klinische Immunologie und

Transfusionsmedizin, Justus-Liebig Universita
¨
t Giessen, Germany
p-Hydroxyphenylacetate decarboxylase from Clostridium
difficile catalyses the decarboxylation of p-hydroxyphenyl-
acetate to yield the cytotoxic compound p-cresol. The three
genes encoding two subunits of the glycyl-radical enzyme
and the activating enzyme have been cloned and expressed
in Escherichia coli. The recombinant enzymes were used
to reconstitute a catalytically functional system in vitro.In
contrast with the decarboxylase purified from C. difficile,
which was an almost inactive homo-dimeric protein (b
2
), the
recombinant enzyme was a hetero-octameric (b
4
c
4
), cata-
lytically competent complex, which was activated using
endogenous activating enzyme from C. difficile or recom-
binant activating enzyme to a specific activity of 7 UÆmg
)1
.
Preliminary results suggest that phosphorylation of the small
subunit is responsible for the change of the oligomeric state.
These data point to an essential function of the small subunit
of the decarboxylase and may indicate unique regulatory
properties of the system.
Keywords: Clostridium difficile; cresol; glycyl radical

enzymes; S-adenosyl-methionine radical enzymes; Tanne-
rella forsythensis.
Clostridium difficile is a spore forming, strict anaerobic
bacterium that causes gastrointestinal infections in humans
ranging from asymptomatic colonization to severe diar-
rhoea, pseudomembranous colitis, toxic megacolon, colon
perforation and occasionally death [1]. C. difficile-associated
diarrhoea is very common in hospitalized patients, partic-
ularly after the normal intestinal flora has been disturbed by
an antibiotic or an antineoplastic treatment [2,3]: the normal
gut microbiota has to be disrupted before C. difficile
infection can become established. The production of toxic
fermentation end products may allow an ongoing suppres-
sion of the endogenous microflora and therefore may play an
important role in the progression of the disease.
The formation and tolerance of p-cresol by C. difficile as
the end product of tyrosine fermentation is well known [4,5].
The enzyme responsible is p-hydroxyphenylacetate decarb-
oxylase (Hpd, E.C. 4.1.1 ) [6] which was previously purified
in an almost inactive state [7]. Based on the N-terminal
amino acid sequence of the protein, an ORF was detected
in the unfinished genome of C difficile strain 630 provided
by the C. difficile Sequencing Group at the Sanger Center.
The encoded 902-amino acid protein was most similar to
pyruvate formate lyase-like proteins of unknown function
and showed a typical glycyl radical consensus sequence
motif (VRVAGF) in the C-terminal region. Moreover, the
decarboxylase gene (hpdB) was located in a putative operon
together with a gene encoding an activating enzyme (hpdA),
which is required to form the kinetically stable glycyl radical

in the active enzyme.
In this communication we report the identification of a
hitherto unknown small subunit (HpdC) of the decarboxy-
lase, which is essential for catalytic activity and may provide
the unique regulatory properties of the Hpd system.
Materials and methods
Materials
C. difficile (DSM 1296
T
) was purchased from the German
Collection of Micro-organisms and Cell Cultures (DSMZ,
Braunschweig, Germany). Escherichia coli strains and
plasmids were obtained from commercial sources. Chemi-
cals were purchased from commercial sources and were of
the highest grade available.
Organisms and cultivation
C. difficile was cultivated as described previously [7]. Unless
otherwise stated, E. coli strains were grown on Luria–
Bertani (LB) agar plates or LB media supplemented with
the required antibiotics at 28–30 °C.
Correspondence to T. Selmer, Laboratorium fu
¨
r Mikrobiologie,
Fachbereich Biologie, Philipps-Universita
¨
t, Karl-von-Frisch Str.,
D-35032 Marburg, Germany.
Fax: + 49 6421 2828979, Tel.: + 49 6421 2825606,
E-mail: selmer@staff.uni-marburg.de
Abbreviations: LB, Luria–Bertani; Hpd, p-hydroxyphenylacetate

decarboxylase; SAM, S-adenosyl-methionine; TFA, trifluoroacetic
acid; RBS, ribosome binding sites.
Enzymes: p-Hydroxyphenylacetate decarboxylase (EC 4.1.1 );
pyruvate formate lyase (EC 2.3.1.54); ribonucleotide reductase
(EC 1.17.4.1); benzylsuccinate synthase (EC 4.1 ).
(Received 8 January 2004, revised 15 March 2004,
accepted 6 April 2004)
Eur. J. Biochem. 1–6 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2004.04152.x
Cloning and sequencing of the genes
The individual cloning steps were carried out in E. coli
DH5a or GM 2159 strains. Genomic DNA of C. difficile
strain DSM 1296
T
was used as a template for PCR
amplification of the hpd-genes. The primers were deduced
from the genomic sequence provided by the Sanger Centre
C. difficile Sequencing Group ( />Projects/C_difficile/blast_server.shtml). KpnIandClaI
endonuclease cleavage sites were introduced upstream and
downstream of the coding sequence in order to facilitate
cloning (Table 1). To minimize PCR errors, a Hi-Fidelity
DNA polymerase (Hi-Fidelity-PCR Enzyme mix, Abgene)
was used. The amplified hpdA and hpdB genes were cloned
into pBluescript II SK(+) (Stratagene). Three clones of
three individual PCR products were sequenced from
double-stranded DNA in order to obtain the type strain
sequences. The hpdC gene was sequenced directly using the
PCR product as template.
In order to allow an in-frame cloning of the individual
genes in expression vectors, mutagenic primers were used
to introduce suitable cleavage sites at the start codon and

downstream of the stop codon. These primers were used to
amplify the desired gene from genomic DNA by PCR and
the fragments were inserted into pET11a, pET11d (Nov-
agen) or pASK-IBA7 vectors (Institut fu
¨
r Bioanalytik,
Go
¨
ttingen, Germany). The resulting clones were sequenced
on both strands and PCR artefacts were replaced with
corresponding DNA from other clones. The resulting clones
were designated pET-D3 (hpdB), pET-A4 (hpdA), pASK-
A2 (hpdA) and pET-X1 (hpdC).
The vector for the coexpression of hpdBandhpdC (pET-
DX4) was obtained from a PCR of genomic DNA using
the primers sDecNheI and InterCterBamHI (Table 1). The
fragment was cloned in pET11a vector and DNA without
mutations in the 3¢-end of hpdB and the entire hpdC gene
wasusedtoreplaceaStuI/BamHI fragment in pET-D3.
Production of HpdB and HpdC in
E. coli
The expression plasmids were used to transform several
E. coli host strains including DH5a [8] GM2159 [9],
BL21
TM
(DE3) Codon Plus-RIL (Stratagene), Tuner
TM
(DE3)pLysS and Rosetta
TM
(DE3)pLysS (Novagen). The

growth conditions (media, temperature, oxygen) and the
concentration of the inducers were varied in order to
establish optimal conditions for the production of soluble
proteins.
HpdB/C and HpdC were produced by E. coli BL21
TM
(DE3) Codon Plus-RIL harbouring pET-DX4 and pET-X1
plasmids, respectively. The cells were grown aerobically in
LB medium supplemented with glucose (0.2%), carbenicil-
lin (100 lgÆmL
)1
) and chloramphenicol (50 lgÆmL
)1
)at
28–30 °C (pET-DX4) and 37 °C (pET-X1). Isopropyl
thio-b-
D
-galactoside was added (1 m
M
)atD
578
of 0.5–0.8.
After 3 h the cells were collected by centrifugation and
stored frozen at )20 °C.
Purification of HpdB and HpdB/C
All purification steps were performed in an anoxic chamber
(Coy Laboratories, Ann Arbor, MI, USA) in a N
2
/H
2

(95%/5%) atmosphere at 15–20 °C.
Hpd was purified from freshly prepared cells of C. diff-
icile (2–2.5 g) essentially as described previously [7]. How-
ever, the DEAE-Sepharose column used in the earlier
preparations was omitted and the Resource-Q column was
replaced by a Source 15Q (1.6/20 cm) column.
The HpdB/C complex was purified from aerobically
induced E. coli BL21
TM
(DE3) Codon Plus-RIL/pET-DX4.
The cell pellet (3.0 g) was washed with buffer A [100 m
M
Tris/HCl pH 7.5, 5 m
M
(NH
4
)
2
SO
4
,1m
M
MgCl
2
,0.5m
M
sodium dithionite, 20 l
M
ATP] and resuspended in 25 mL
uffer A. The cell suspension was sonicated at 50 W, for

4 · 5 min with a Branson sonifier (Branson Ultrasonics,
Danbury, CT, USA) on ice. Cell debris and membranes
were removed by centrifugation (100 000 g,60min).The
supernatant was loaded on a Resource-Q column (6 mL),
which was equilibrated with buffer A. The decarboxylase
complex was eluted in a linear gradient of 0–450 m
M
NaCl
in 120 mL buffer A. The fractions containing HpdB/C were
concentrated using a Vivapure device with a 100 kDa cut-
off membrane (Vivascience, Germany) and loaded on a
prepacked Superdex 200 HR 10/30 gel filtration column.
The column was run in 150 m
M
NaCl in buffer A.
The elution and purity of the proteins were monitored by
Table 1. Amplification primers. s, Sense-strand; as, antisense-strand.
Name Nucleotide sequence (5 fi 3¢) Recognition site
DCNtermKpnI ATTCTGGTACCGTTTTATACTAATTATAGAAAGATTAAGG KpnI
DCCtermClaI AAATTCAATCGATCACTATGCTTTCTCATATTTTACACC ClaI
sDecNheI GAAATGGCTAGCAGTAAAGAAGACAAAATAAG NheI
asDCBamHI ATGCTTTCTCGGATCCTACACCCCTTCATACTCTGTTCTAGC BamHI
ActNtermKpnI TGTGTTGGTACCGAGAAAGAAGACTGGG KpnI
ActCtermClaI TATTTATCGATAAAAACCACATAAAAAAGG ClaI
ActNterNheI GTGTTTAATGGCTAGCCAAAAGCAATTAGAAGGC NheI
ActCterBamHI CTTGTATTGGATCCTAGAAAGCTGTCTCATGACC BamHI
ActSacII GGACATCCGCGGTATGAGTAGTCAAAAGC SacII
InterNterm CAATTAATGGTACCTGTTGCTGGGTTTACTCAATATTGG KpnI
InterCterm CCTTCTAATCGATTTTGACTACTCATTAAACACATGCC ClaI
InterNterNcoI GTGTAAACCATGGGAAAGCATAGTGATTGTATG NcoI

InterCterBamHI CTTCTGGATCCTTTTGACTACTCATTAAAC BamHI
2 P. I. Andrei et al. (Eur. J. Biochem.) Ó FEBS 2004
SDS/PAGE with Coomassie blue staining and the purified
enzyme was stored anaerobically at 4 °C.
Production of HpdA
HpdA was produced anaerobically by E. coli GM 2159
harbouring the pASK-A2 plasmid in LB medium supple-
mented with 0.2% glucose and carbenicillin (100 lgÆmL
)1
)
at 30 °C. The production of HpdA was induced at D
(578nm) of 0.35–0.5 with anhydrotetracycline (50 lgÆL
)1
)
for 20 h. During the first 6 h of induction, the pH was
maintained at 7.5 by addition of 10
M
NaOH. Cells were
harvested and the protein was immediately purified.
Purification of recombinant HpdA
HpdA was purified in an anoxic chamber using 5 mL
StrepTactin
Ò
Sepharose columns. Wet packed cells (3.5–
5.0 g) were suspended in buffer B (20 mL) containing
100 m
M
Tris/HCl pH 8, 150 m
M
NaCl, 5 m

M
dithiothre-
itol, and homogenized by sonication. Cell debris was
removed by centrifugation (60 min at 100 000 g). The clear
supernatant was applied onto the column equilibrated with
buffer B. Non-bound protein was washed off with buffer B
prior to elution of HpdA with desthiobiotin (2.5 m
M
)in
buffer B. The enzyme was concentrated in 2 mL Vivaspin
centrifugation devices with a cut-off of 10 kDa (final
concentrations > 1 mgÆmL
)1
) and stored anoxically at
)20 °C.
Reconstitution of the Hpd activity
in vitro
All reconstitution steps were carried out under strict anoxic
conditions. The recombinant HpdB/C was tested for
catalytic competence using endogenous HpdA in cell-free
extracts of C. difficile. Therefore, cell-free extracts from
C. difficile and E. coli Bl21
TM
(DE3) Codon Plus-RIL/pET-
DX4 (85 lgÆmL
)1
and 28 lgÆmL
)1
total protein, respect-
ively) were incubated at 30 °C in 100 m

M
Tris/HCl pH 7.5,
5m
M
(NH
4
)
2
SO
4
,1m
M
MgCl
2
,0.5m
M
sodium dithionite,
with 20 m
M
p-hydroxyphenylacetate and 0.23 m
M
S-aden-
osyl-methionine (SAM). At defined time points samples
were withdrawn and analysed for p-cresol formation as
described previously [7]. Controls omitting any of the
essential components (HpdA, HpdB/C or SAM) were
analysed in parallel assays.
p-Hpd activity was also reconstituted with purified
recombinant proteins. Pure HpdA (3.8 lg) was reduced
for 3 h in the presence of 0.5 m

M
sodium dithionite,
0.57 m
M
SAM and 17 lg HpdB/C in a final volume of
100 lL100m
M
Tris/HCl pH 7.5, 5 m
M
(NH
4
)
2
SO
4
,5m
M
dithiothreitol, 1 m
M
MgCl
2
(100 lL) at 4 °C. To follow
the decarboxylation, 25 m
M
substrate (1 mL) was added.
Aliquots were taken at defined time points and assayed for
p-cresol formation by HPLC as described previously [7].
MALDI-TOF MS of HpdC
Partially purified decarboxylase from C. difficile and recom-
binant HpdB/C were acidified with trifluoroacetic acid

(TFA). The supernatant was subjected to solid phase
extraction using 50 mg-Sep-PakÒ Vac C18-cartridges
(Waters) equilibrated with 0.1% TFA. The columns were
washed twice with 1 mL 0.1% TFA and eluted with 0.1%
TFA/67% acetonitrile. The samples thus obtained (1 lL)
were mixed on a gold-plated target with 1 lL of a saturated
solution of sinapinic acid in 0.1% TFA/67% acetonitrile
and dried under air. The samples were analysed using a
Voyager-DE/RP-MALDI-TOF MS in reflector mode.
Determination of relative molecular masses
of the native enzymes
The apparent molecular masses of the native 4-Hpd was
determined by gel filtration on a Superdex 200 HR 10/30
prepacked column, equilibrated with 150 m
M
NaCl in
buffer A. Ribonuclease A from bovine pancreas (13.7 kDa),
chymotrypsinogen A from bovine pancreas (27 kDa),
ovalbumin from hen egg (43 kDa), albumin from bovine
serum (67 kDa), aldolase from rabbit muscle (158 kDa),
ferritin from horse spleen (440 kDa) and thyroglobulin
from bovine thyroid (669 kDa) were used as molecular
mass marker proteins (Amersham Biosciences, Germany.)
Other methods
Protein concentrations were determined using the Bradford
procedure [10].
Results
Based on the N-terminal amino acid sequence of purified
HpdB, a putative operon was identified in the genome of
C. difficile strain 630, which encoded both the glycyl radical

subunit of the decarboxylase (HpdB) and its activating
enzyme (HpdA). A detailed analysis of the sequence taking
into account putative ribosome binding sites (RBS) estab-
lished a third ORF (hpdC) located between hpdB and
hpdA (Fig. 1). During the initial purification of the decar-
boxylase from C. difficile this small protein (85 amino acids,
9.5 kDa) was overlooked. However, the low activity yield of
the purification was attributed to the loss of a low molecular
mass cofactor [6,7].
Based on the genomic DNA sequence of C. difficile strain
630, specific primers were deduced in order to amplify the
genes encoding the two putative decarboxylase subunits and
its activase by PCR from genomic DNA of the type strain
DSM 1296
T
. The type strain sequences have been deposited
in the EMBL Nucleotide Sequence Database under the
accession numbers AJ543425 (hpdB), AJ543426 (hpdC)and
AJ543427 (hpdA). Within the hpdB gene, nine nucleotides
were exchanged between the type strain and strain 630, but
only two of these replacements changed the amino acid
sequence (M670I and E806D). The gene of the small
subunit (hpdC) contained two exchanged nucleotides, which
were silent at the amino acid level. In hpdA, one nucleotide
differed, leading to one amino acid exchange (I165V).
The recombinant proteins were produced in E. coli from
inducible expression vectors. Suitable endonuclease cleavage
sites were introduced by PCR mutagenesis in order to allow
in-frame cloning of the genes. While the BamHI sites
introduced in the 3¢-UTR of the genes did not affect the

resulting amino acid sequences, the introduction of NheIor
NcoI sites next to the start codon altered the N-terminal
Ó FEBS 2004 Subunit composition of p-Hpd (Eur. J. Biochem.)3
sequences of the resulting proteins (MSQS to MASS in
HpdB, MRKH to MGKH in HpdC and MSSQ to MASQ
in HpdA). In contrast with the hpdC gene product, which
was produced in a variety of host cells as a soluble protein,
HpdB and HpdA were synthesized only in Rosetta
TM
(DE3)pLysS and BL21
TM
(DE3) Codon Plus-RIL cells.
These hosts produce additional rare tRNAs and therefore
support the expression of AT-rich genes such as hpdAand
hpdB. Although strongly induced by isopropyl thio-b-
D
-galactoside, the proteins were produced as inclusion
bodies. Neither variations in induction procedure nor media
nor temperature yielded soluble protein (data not shown).
The coexpression of hpdB and hpdC was achieved from
a pET11a-derived expression clone, which contained both
genes. In order to obtain this clone, the 3¢-region of hpdB
and the hpdC gene were introduced into the existing clone of
hpdB in pET11a. The resulting plasmid contained the
vector-derived RBS and 5¢-UTR in front of the hpdB gene
but the clostridial RBS and 5¢-UTR in front of hpdC.
Expression of this construct in E. coli BL21
TM
(DE3)
Codon Plus-RIL was efficient for both polypeptides and

resulted in soluble protein.
Though no formation of p-cresol was observed in cell-free
extracts from E. coli coexpressing hpdB and hpdC,a
catalytically competent protein was produced. As shown
in Fig. 2, the decarboxylase was rapidly activated by HpdA
from cell-free extracts of C. difficile yielding a specific
activity of 90 mUÆmg
)1
corrected for an almost negligible
background activity of the C. difficile extract, demonstra-
ting the production of a functional recombinant enzyme.
While strict anoxic conditions were required to achieve
activation, the process was equally effective for cell-free
extracts containing HpdB/C prepared from aerobically or
anaerobically grown cells. No p-cresol formation was
detected in the Resource-Q fractions containing separated
subunits HpdB or HpdC. These results show that HpdC is
essential for p-cresol formation and establish this polypep-
tide as a subunit of p-Hpd.
The decarboxylase was originally purified from cell-free
extracts of C. difficile by successive anion exchange chro-
matography on DEAE-Sepharose and Resource-Q fol-
lowed by size exclusion chromatography on Superdex 200
[7]. The DEAE Sepharose column led to a loss of 95% of
the activity and was therefore omitted throughout this work
without significantly affecting purity. The presence of both
HpdB and HpdC in these preparations was demonstrated
both on SDS/PAGE and by MALDI-TOF MS (see below)
and < 10% of the initial activity was lost in the first step.
Gel filtration, however, led to a separation of the subunits

and the final activities were found to be similar to those
reported previously (< 0.5 UÆmg
)1
).
The recombinant decarboxylase eluted from the anion
exchange column in a similar position to the enzyme from
C. difficile. In contrast, the behaviour of this enzyme on
the size exclusion chromatography column was surprisingly
different: The nonactivated homo-dimeric decarboxylase
(eluting at  15 mL) was found only in poor yields and
essentially free of HpdC (Fig. 3). The majority of the
recombinant decarboxylase eluted at  12 mL, indicating
a native molecular mass of 460 kDa. As judged by SDS/
PAGE, this protein was composed of HpdB and HpdC
polypeptides. The relative intensities of the bands were
estimated by scanning and quantified using the Molecular
Dynamic
IMAGEQUANT
5.2 program. The ratio of HpdB(b)/
HpdC(c) was corrected for the different sizes and found
to be 1 : 1, indicating an b
4
c
4
composition in a hetero-
octameric complex.
Both the protein preparations from C. difficile and the
recombinant ones were analysed by MALDI-TOF MS.
Since the fully purified enzyme from C. difficile was
essentially free of HpdC, partially purified enzyme obtained

from Source 15Q anion exchange column was subject
to a solid phase extraction procedure. The molecular
mass observed for HpdC in these preparations was
Fig. 2. Activation of the recombinant HpdB/C complex. Cell-free
extract from C. difficile (85 lgÆmL
)1
)wasincubatedat30°Cinthe
presence of 20 m
M
pHPA, in the absence (d)orpresence(s)of
0.23 m
M
SAM. In the presence of E. coli extract containing HpdB/C
(n), a rapid activation of the recombinant decarboxylase was
observed. The activation was strictly dependent on the endogenous
HpdA from C. difficile and SAM (data not shown).
Fig. 1. Location of the hpdC gene. Potential
clostridial ribosomal binding sites are boxed
and the start codons are shaded grey. The
amino acid sequence of HpdC is shown in
bold letters together with the C-terminal end
of HpdB and the N-terminal of HpdA.
4 P. I. Andrei et al. (Eur. J. Biochem.) Ó FEBS 2004
9508 ± 9 Da. This value is in good agreement with the
predicted molecular mass of 9504 Da for HpdC. The mass
spectrum of recombinant HpdC was strikingly different:
in addition to a signal indicating a molecular mass of
9510 ± 9 Da, a second signal with equal intensity was
observed at 9590 ± 9 Da. The 80-Da mass increment
between the two signals suggests a possible phosphorylation

of the small subunit in this enzyme, which could account for
the different oligomeric states.
The recombinant activase was produced as an
N-terminally streptavidin-tagged protein and purified to
apparent homogeneity by affinity chromatography on
StrepTactinÒ Sepharose under strict anoxic conditions
(Fig. 4). The final preparation was deep brown, indicating
the expected presence of iron–sulfur clusters in the enzyme.
Size exclusion chromatography showed HpdA to be a
monomeric enzyme that was determined to contain 7–8 mol
of iron and 6–7 mol of acid labile sulfur per mol of enzyme.
The E. coli extracts containing the recombinant activator
catalytically activated the recombinant decarboxylase to
yield specific activities of > 7 UÆmg
)1
.However,after
purification of the recombinant enzymes, the efficiency of
the activating process dropped dramatically yielding specific
activities of below 1.5 UÆmg
)1
. This level of activity was
achieved only with a large molecular excess of the activating
enzyme (HpdA/HpdB/C ¼ 10 : 1).
Discussion
Initial attempts to purify p-Hpd from C. difficile [6]
were unsuccessful. Later, the decarboxylase was isola-
ted ) though almost inactive ) as a homodimer of the
HpdB subunits [7]. In both cases it was suggested that the
low activity yield was due to the loss of a low molecular
mass fraction of < 10 kDa during the purification.

A closer analysis of the DNA sequence provided by the
Sanger Center, taking into account the possible ribosomal
binding sites, showed a third ORF, hpdC, located between
the decarboxylase and the activase genes. The hpdC gene
starts directly downstream of the decarboxylase gene (hpdB),
and overlaps the 5¢-region of hpdA. It encodes an 85-amino
acid, cysteine rich polypeptide (9.4% cysteine). In contrast
with the well-studied glycyl radical enzymes pyruvate
formate-lyase and class III ribonucleotide reductase, which
are homodimeric enzymes (for review see [11,12]), the
findings reported in this paper suggest a hetero-oligomeric
structure of the decarboxylase in the catalytically competent
enzyme. A hetero-oligomeric structure has been described for
the benzylsuccinate synthase of Thauera aromatica and
related organisms [13,14]. Whereas the small subunits form
a stable complex with the glycyl radical subunit in benzyl-
succinate synthase and therefore copurify, the complex of
HpdB and HpdC is apparently much weaker and rapidly
dissociates during the purification.
The first evidence for an important structural function of
HpdC in the decarboxylase arose from the observation that
HpdB produced by E. coli/pET-D3 exclusively yielded
insoluble protein, whereas coexpression of hpdB and hpdC
gave a soluble, catalytically competent enzyme, which was
smoothly activated by cell-free extracts of C. difficile.
An important, probably regulatory function of HpdC
became evident during the purification of the recombinant
protein. The molecular mass data immediately suggested a
phosphorylation of HpdC, which could affect the oligo-
meric structure and complex stability of the decarboxylase.

Since recombinant HpdC is not phosphorylated when
produced by E. coli/pET-X1 plasmid (data not shown), it
seems very likely that its phosphorylation is a catalytic
property of the HpdB subunit. Indeed, a Prosite motif scan
of the HpdB amino acid sequence revealed the presence of a
P-loop ATP-binding motif (PS00017, [AG]-x(4)-G-K-[ST])
comprising amino acids 181–188 (AKEWVGKS) [15].
However, at present it is not possible to exclude an artificial
origin for this modification in E. coli and further analysis of
this putative phosphorylation will be required in order to
establish its functional relevance.
A correlation between glycyl radical concentration and
Ævolume activity for partially purified enzyme from C. diff-
icile, containing the HpdC subunit, allowed an estimate for
the specific activity of about 50 UÆmg
)1
in a fully active
Fig. 4. Purification of HpdA from E. coli cells expressing Strep-tagged
hpdA. The samples were separated by SDS/PAGE and stained with
Coomassie blue. Cell-free extracts of noninduced cells (NI), induced
cells (I) and the affinity purified product (HpdA) are shown
(S: standard proteins).
Fig. 3. Purification of HpdB/C from E. coli coexpressing hpdB and
hpdC. The SDS/PAGE analysis of an elution profile from a Superdex
200 column is shown. Consecutive fractions of the 450-kDa (15–17)
and the 200-kDa (24–27) region were analysed by SDS/PAGE and
Coomassie blue staining. Cell-free extracts of noninduced (NI) and
induced (I) cells and partially purified enzyme from the Resource-Q
column (R) are shown for comparison. Molecular masses of standard
proteins (S) and the positions of HpdB and HpdC are indicated.

Ó FEBS 2004 Subunit composition of p-Hpd (Eur. J. Biochem.)5
decarboxylase with one radical site per HpdB dimer. The
recombinant, octameric complex of HpdB and HpdC is
smoothly activated using either endogenous HpdA from
C. difficile extracts or recombinant HpdA from E. coli
extracts. The specific activity of the recombinant decarb-
oxylase is higher (7 UÆmg
)1
) than the highest activity
observed for the homo-dimeric enzyme purified from
C. difficile (< 0.5 UÆmg
)1
), but is still significantly lower
than the estimated maximum value. These findings suggest
that the recombinant decarboxylase and the recombinant
activating enzyme are functional; however, the reconstitu-
tion of the system using individually purified enzymes
in vitro is difficult due to an essential requirement for an as
yet unknown factor present in the cell extracts of both
C. difficile and E. coli. Apparently, this factor is lost during
purification and limiting when cell-free extracts are used to
restore activity (P. I. Andrei and M. Blaser, unpublished
data), suggesting that higher specific activities might be
obtained by providing the missing compound.
Interestingly, a small fraction of the recombinant decarb-
oxylase dissociated during the purification. While the
resulting homo-dimers of nonactivated HpdB thus obtained
remained soluble and have been purified, all attempts to
detect activity in these preparations failed. It will be
interesting to establish whether the glycyl radical formation

by HpdA is possible with this form or whether it inhibits this
reaction with the functional complex.
The function of HpdC remains to be established, but the
data presented strongly suggest that this small subunit is
essential for catalytic activity and may play an important role
in regulation of the decarboxylase system. Indeed, the
presence of this small subunit distinguishes the 4-Hpd from
all other groups of glycyl radical enzymes. While several
hundreds of putative glycyl radical enzymes are found in the
finished and unfinished genomes of microbes, only one
additional putative arylacetate decarboxylase has been found
so far in the unfinished genome of the human pathogen
Tannerella forsythensis ATCC 43037 (formerly named
Bacteroides forsythus), the sequence for which can be
obtained from />base ¼ b_forsythus. The gene encoded amino acid
sequence of a putative glycyl radical enzyme of this organism
is highly similar (57% identity, 88% similarity) to HpdB and
also has a small ORF encoding an 86-amino acid protein
directly downstream. The amino acid sequence of this small
protein is 33% identical and 58% similar to HpdC. Since
T. forsythensis has very complex growth requirements and is
therefore not accessible for metabolic testing in vivo, recom-
binant production of this enzyme will be performed in order
to study the properties of this new system and to compare its
properties with those of the Hpd from C. difficile.
Acknowledgements
We are very grateful to Prof. Dr W. Buckel for his constant support
throughout the project and to Dr Dan Darley for proof-reading the
manuscript. We also like to thank the Max-Planck Institute for
Terrestrial Microbiology for the access to MALDI-TOF MS and EPR.

This work was supported by grants from the priority program ÔRadicals
in Enzymatic CatalysisÕ of the Deutsche Forschungsgemeinschaft
(DFG) and is dedicated to Prof. Dr Achim Kro
¨
ger. Prof. Dr Kro
¨
ger
was a member of the reviewing panel of the priority program and died
on 11 June 2002.
References
1. Spencer, R.C. (1998) Clinical impact and associated costs of
Clostridium difficile-associated disease. J. Antimicrob. Chemother.
Suppl. C41, 5–12.
2. Fedorko, D.P., Engler, H.D., O’Shaughnessy, E.M., Williams,
E.C., Reichelderfer, C.J. & Smith, W.I. Jr (1999) Evaluation of
two rapid assays for detection of Clostridium difficile toxin A in
stool specimens. J. Clin. Microbiol. 37, 3044–3047.
3. Borriello, S.P. & Wilcox, M.H. (1998) Clostridium difficile infec-
tions of the gut: the unanswered questions. J. Antimicrob.
Chemother. Suppl. C41, 67–69.
4. Elsden, S.R.H.M.G. & Waller, J.M. (1976) The end products of
the methabolism of aromatic amino acids by Clostridia. Arch.
Microbiol. 107, 283–288.
5. Hafiz, S. & Oakley, C.L. (1976) Clostridium difficile:isolationand
characteristics. J. Med. Microbiol. 9, 129–136.
6. D’Ari, L. & Barker, H.A. (1985) p-Cresol formation by cell-free
extracts of Clostridium difficile. Arch. Microbiol. 143, 311–312.
7. Selmer, T. & Andrei, P.I. (2001) p-Hydroxyphenylacetate
decarboxylase from Clostridium difficile.Anovelglycylradical
enzyme catalysing the formation of p-cresol. Eur. J. Biochem. 268,

1363–1372.
8. Sambrook, J. & Russell, D.W. (1989) Molecular cloning.
A laboratory Manual, 3 edn, Cold spring harbor laboratory press,
New York.
9. Parker, B. & Marinus, M.G. (1988) A simple and rapid method to
obtain substitution mutations in Escherichia coli: Isolation of a
dam deletion/insertion mutation. Gene 73, 531–535.
10. Bradford, M.M. (1976) A rapid and sensitive method for the
quantitation of microgram quantities of protein utilizing the
principle of protein-dye binding. Anal. Biochem. 72, 248–254.
11. Fontecave, M. (1998) Ribonucleotide reductases and radical
reactions. Cell Mol. Life Sci. 54, 684–695.
12. Sawers, G. & Watson, G. (1998) A glycyl radical solution: oxygen-
dependent interconversion of pyruvate formate-lyase. Mol.
Microbiol. 29, 945–954.
13. Coschigano, P.W., Wehrman, T.S. & Young, L.Y. (1998) Identi-
fication and analysis of genes involved in anaerobic toluene
metabolism by strain T1: putative role of a glycine free radical.
Appl. Environ. Microbiol. 64, 1650–1656.
14. Leuthner, B., Leutwein, C., Schulz, H., Horth, P., Haehnel,
W., Schiltz, E., Schagger, H. & Heider, J. (1998) Biochemical
and genetic characterization of benzylsuccinate synthase from
Thauera aromatica: a new glycyl radical enzyme catalysing the
first step in anaerobic toluene metabolism. Mol. Microbiol. 28,
615–628.
15. Walker, J.E., Saraste, M., Runswick, M.J. & Gay, N.J. (1982)
Distantly related sequences in the alpha- and beta-subunits of
ATP synthase, myosin, kinases and other ATP-requiring enzymes
and a common nucleotide binding fold. EMBO J. 1, 945–951.
6 P. I. Andrei et al. (Eur. J. Biochem.) Ó FEBS 2004