Characterization and inactivation of an agmatine deiminase from Helicobacter pylori

Article abstract of DOI:10.1016/j.bioorg.2009.11.004

Helicobacter pylori encodes a potential virulence factor, agmatine deiminase (HpAgD), which catalyzes the conversion of agmatine to N-carbamoyl putrescine (NCP) and ammonia - agmatine is decarboxylated arginine. Agmatine is an endogenous human cell signaling molecule that triggers the innate immune response in humans. Unlike H. pylori, humans do not encode an AgD; it is hypothesized that inhibition of this enzyme would increase the levels of agmatine, and thereby enhance the innate immune response. Taken together, these facts suggest that HpAgD is a potential drug target. Herein we describe the optimized expression, isolation, and purification of HpAgD (10-30 mg/L media). The initial kinetic characterization of this enzyme has also been performed. Additionally, the crystal structure of wild-type HpAgD has been determined at 2.1 A resolution. This structure provides a molecular basis for the preferential deimination of agmatine, and identifies Asp198 as a key residue responsible for agmatine recognition, which has been confirmed experimentally. Information gathered from these studies led to the development and characterization of a novel class of haloacetamidine-based HpAgD inactivators. These compounds are the most potent AgD inhibitors ever described.

Full text of DOI:10.1016/j.bioorg.2009.11.004

Bioorganic Chemistry 38 (2010) 62–73
Contents lists available at ScienceDirect
Bioorganic Chemistry
journal homepage: www.elsevier.com/locate/bioorg
Characterization and inactivation of an agmatine deiminase from Helicobacter pylori q
Justin E. Jones, Corey P. Causey, Leslie Lovelace, Bryan Knuckley, Heather Flick, Lukasz Lebioda,
*
Paul R. Thompson
Department of Chemistry & Biochemistry, University of South Carolina, 631 Sumter Street, Columbia, SC 29208, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 9 October 2009
Available online 29 November 2009
Helicobacter pylori encodes a potential virulence factor, agmatine deiminase (HpAgD), which catalyzes the
conversion of agmatine to N-carbamoyl putrescine (NCP) and ammonia – agmatine is decarboxylated
arginine. Agmatine is an endogenous human cell signaling molecule that triggers the innate immune
response in humans. Unlike H. pylori, humans do not encode an AgD; it is hypothesized that inhibition
of this enzyme would increase the levels of agmatine, and thereby enhance the innate immune response.
Taken together, these facts suggest that HpAgD is a potential drug target. Herein we describe the opti-
mized expression, isolation, and purification of HpAgD (10–30 mg/L media). The initial kinetic character-
ization of this enzyme has also been performed. Additionally, the crystal structure of wild-type HpAgD
has been determined at 2.1 Å resolution. This structure provides a molecular basis for the preferential
deimination of agmatine, and identifies Asp198 as a key residue responsible for agmatine recognition,
which has been confirmed experimentally. Information gathered from these studies led to the develop-
ment and characterization of a novel class of haloacetamidine-based HpAgD inactivators. These com-
pounds are the most potent AgD inhibitors ever described.
Keywords:
Deiminase
Haloacetamidine
Inactivator
Ó 2009 Elsevier Inc. All rights reserved.
1. Introduction
first 5 days, followed by a triple therapy of PPI, clarithromycin, and
tinidazole for the remaining 5 days [4]. While generally successful,
Helicobacter pylori are gram negative bacteria that colonize the
gastrointestinal tract, and infection by this pathogen can lead to
gastritis, peptic ulcer diseases, and even cancer [1,2]. Although
ꢀ50% of the worldwide population is infected by this bacterium
[3], most are asymptomatic, and only a subset (ꢀ15–20%) will
present with the aforementioned disease pathologies [2]. Never-
theless, the high incidence of H. pylori infection, coupled with the
potential for debilitating effects, often necessitates therapeutic
intervention. Current therapeutic approaches include: (i) the pro-
ton pump inhibitor (PPI) triple therapy (i.e., a PPI plus amoxicillin
and clarithromycin), (ii) the bismuth-based triple/quadruple ther-
apy (i.e., bismuth, metronidazole, tetracycline, and a PPI), and (iii)
the sequential therapy, which includes a PPI and amoxicillin for the
there are a number of problems associated with these therapies,
including: antibiotic resistance, adverse side effects (e.g., allergic
reactions), and cost – the latter issue is especially true in develop-
ing countries [4–6]. Thus, novel therapeutic approaches targeting
this pernicious bacterium are needed.
H. pylori strains encode a number of pathogenicity factors that
contribute to disease onset and progression, including: urease, fla-
gella, vaculating cytotoxin A (VacA), and the cytotoxin associated
gene (cag) pathogenicity island [2]. Urease contributes to acid
resistance and flagella allow for bacterial movement. VacA is a
pore-forming toxin that breaks down the gastric epithelium, lead-
ing to gastric ulceration and necrosis. Finally, the cag pathogenicity
island, which encodes a type 4 secretion system, is associated with
increased inflammation, accumulation of phagocytic cells in the
intestines, and release of interleukin (IL)-8 [1,7]. In addition to
these well established pathogenicity factors, H. pylori also encode
an agmatine deiminase (AgD), which may contribute to disease
pathogenesis vide infra.
Agmatine is a naturally occurring molecule that is generated by
the decarboxylation of arginine. Although it is unclear whether hu-
mans synthesize agmatine directly [8], it is clear that this molecule
is present in human tissues [9–11]; likely sources include dietary
sources or even beneficial bacteria present in the intestinal flora.
In bacteria, the in vivo hydrolysis of agmatine by AgD results in
the formation of N-carbamoyl putrescine (NCP) and ammonia
Abbreviations: PPI, proton pump inhibitor; AgD, agmatine deiminase; NCP,
N-carbamoyl putrescine; PAD, protein arginine deiminase; Cit, citrulline;
ABFA, N-(4-aminobutyl)-2-fluoro-ethanimidamide; ABCA, N-(4-aminobutyl)-
2-chloro-ethanimidamide; BAEE, benzoyl-
L
-arginine ethyl ester; BAA, ben-
-arginine methyl ester; DTT,
zoyl- -arginine amide; BAME, N- -Benzoyl-L
L
a
dithiothreitol; GST, glutathione S-transferase; TCEP, Tris(2-carboxy-
ethyl)phosphine hydrochloride; HEPES, N-(2-hydroxyethyl)piperazine-N0-(2-
ethanesulfonic acid); Orn, Ornithine.
q
Coordinates and structure factors have been deposited with the PDB (accession
code 3HVM).
* Corresponding author. Fax: +1 803 777 9521.
E-mail address: thompson@mail.chem.sc.edu (P.R. Thompson).
0045-2068/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.bioorg.2009.11.004

J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
63
(Fig. 1). AgDs are an integral part of the agmatine deiminase sys-
tem (AgDS) in a number of bacterial species (e.g., Pseudomonas
aeruginosa, Listeria monocytogenes, Enterococcus faecalis, Yersinia
pestis, and Streptococcus mutans) [12–14], where they contribute
to polyamine biosynthesis. For example, S. mutans possess a 4 gene
operon (aguBDAC) that encodes for an agmatine–putrescine anti-
porter (aguD), an AgD (aguA), a putrescine carbamoyltransferase
(aguB) that converts NCP into putrescine and carbamoylphosphate,
and a carbamate kinase (aguC) that transfers the phosphate group
from carbamoylphosphate to ADP (Fig. 1) [14]. This system, which
ultimately converts agmatine into putrescine, ATP, ammonia, and
carbon dioxide, is thought to play an important role in acid resis-
tance and competitive survival. It has also been shown that, under
specific growth conditions, exogenously added agmatine slows the
growth of wild type strains and prevents the growth of strains that
lack components of the AgDS [14,15]. Interestingly, H. pylori, and a
number of other human pathogens (e.g., Legionella pneumophila
(strain Philadelphia 1), Streptococcus pneumonia, Vibrio parahaemo-
lyticus RIMD 2210633, Lactococcus lactis) encode AgDs, i.e., the en-
zymes responsible for converting agmatine into NCP, but lack a
complete AgDS. Instead, these bacteria encode an alternate agma-
tine degradation pathway that involves the initial formation of
NCP, which is subsequently hydrolyzed to form putrescine, via
the actions of NCP amidohydrolase. The essential role of putrescine
in polyamine biosynthesis suggests that this alternate agmatine
degradation pathway, and the AgDs in particular, plays an impor-
tant role in the life cycle of these organisms. This pathway may also
contribute to acid resistance and competitive survival of the bacte-
ria, similarly to the AgDS described above.
Interestingly, H. pylori are known to increase the expression of
ornithine decarboxylase in macrophages and, as a consequence, in-
crease the levels of spermine, a polyamine that is known to inhibit
the effector function of macrophages. This activity has been sug-
gested as a bacterial defense response to avoid being killed by
the innate immune response [20].
In humans, it has also been shown that agmatine stimulates the
release of gastrin [21], which in turn, increases the release of pro-
tons into the gut. Additionally, agmatine levels are markedly de-
creased in tumors of colonic origin [22]. These findings are
relevant to H. pylori infections because they suggest that HpAgD
may contribute to tumor cell proliferation and acid resistance by
decreasing the endogenous levels of agmatine. In total, the above
described actions of agmatine in humans suggest that, by decreas-
ing the in vivo concentration of agmatine, HpAgD may, in addition
to other virulence factors, modulate human cell signaling, thereby
providing a mechanism to avoid (or control) detection by the hu-
man immune system. A similar role has been suggested for H. py-
lori arginase: decreased levels of arginine, due to bacterial arginase
activity, attenuate the innate immune response [23]. If this
hypothesis is correct, AgD inhibitors should, in addition to their po-
tential direct effects on bacterial growth, provide a mechanism to
increase the concentration of endogenous agmatine and thereby
alter the virulence associated with infection by these pathogenic
bacteria. The presence of AgD homologues in other pathogenic bac-
teria suggests that AgD inhibitors may show broad utility.
H. pylori AgD (HpAgD) is a 330 amino acid, 38 kDa protein that
is annotated in both genomic and structural databases (PDB ID:
2CMU) as a putative peptidyl-arginine deiminase (PAD). PADs are
members of the guanidinium modifying family of enzymes, a fam-
ily that includes enzymes that hydrolyze the guanidinium groups
of arginine, methylarginine, peptidyl-arginine, and agmatine to
form the ureido containing derivatives [24]. In PAD4, and other
guanidinium modifying enzymes, there are four key catalytic resi-
dues: Asp350, His471, Asp473 and Cys645 (human PAD4 number-
ing) [25]. Structure based alignments of HpAgD and PAD4 indicate
that Asp84, His202, Asp204, and Cys324 are the corresponding res-
idues in HpAgD. Based on this sequence conservation and insights
gained from mechanistic studies of PAD4 [25], Asp84 and Asp204
likely play critical roles in binding and orienting the substrate
guanidinium for nucleophilic attack by the side chain thiolate of
Cys324 (Fig. 2). His202 would then be expected to act as a general
acid, donating a proton to the departing ammonia either before or
In addition to its potential importance to H. pylori growth and
survival, H. pylori AgD (HpAgD) may also contribute to disease
pathogenesis through its ability to degrade agmatine. This is the
case because agmatine seems to play important, albeit incom-
pletely defined, roles in human cell signaling pathways [16]. For
example, agmatine is known to bind to the
a2-adrenergic and imi-
dazoline receptors, possess neuromodulatory activity [9,10], trig-
ger nitric oxide production [16,17], modulate the innate immune
response [17], and decrease blood glucose in type 2 diabetic rats
[18]. Agmatine can also down regulate the synthesis of other poly-
amines. For example, increased levels of agmatine can result in
decreased levels of ornithine decarboxylase, a key enzyme in
putrescine synthesis; this decrease leads to antiproliferative effects
on a number of colonic, hepatic, and neuronal tumor cells [19].
NH2
NH
NH2
NH
NH2
NH
H2N
H2N
O
NH3
ADC
AgD
H2O
CPA
NH4
CO2
+
+
+
NH4
COO
H3N
H3N
H3N
H3N
Pi PTC
NH2
NH3
O
ADP
CK
O
+
NH4
CO2
+
O
P
O
ATP
O
H3N
Fig. 1. Bacterial agmatine degradation pathways. In bacteria, arginine is converted to agmatine via the actions of arginine decarboxylases (ADC). Agmatine is then hydrolyzed
to form N-carbamoyl putrescine (NCP) by agmatine deiminase (AgD). In bacteria encoding an agmatine deiminase system, NCP is then converted into putrescine and
carbamoylphosphate by putrescine transcarbamoylase (PTC). Carbamate kinase (CK) then catalyzes the transfer of the phosphate moiety in carbamoylphosphate to ADP,
forming ATP, as well as ammonia and CO2. Other pathogenic bacteria encode an alternate non-ATP forming agmatine degradation pathway that involves the conversion of
NCP into putrescine, ammonia, and CO2 via the actions of an N-carbamoylputrescine amidohydrolase (CPA).

64
J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
Electrospray ionization mass spectra were acquired using a
Asp204
O-
Waters Micromass quadrapole-time-of-flight (Q-TOF) tandem
mass spectrometer in the positive ion mode. Matrix assisted laser
desorption/ionization (MALDI) mass spectra were acquired with
a Bruker Ultraflex II MALDI-TOF/TOF mass spectrometer. DNA
sequencing was performed at the Synthesis and Sequencing Facil-
ity at Johns Hopkins University School of Medicine.
H
O
H
N
O-
O
N
H
H
N
H
N
H
N
H
N
N
N
1
H
H
H
O
H
S
N
His202
H
O-
S-
Cys324
N
H
O-
H
O
2.2. Construction of an expression vector for HpAgD
agmatine
Asp84
An expression vector encoding HpAgD (PDB ID: 2CMU) was ob-
tained from SGX Pharmaceuticals, Inc. Using this vector as the tem-
plate for PCR reactions, the HpAgD gene was cloned into a modified
pET28 vector using the following primers: 50-AAAAAACATATGA
AAAGAATGTTAGCGGAGTTTG-30 and 50-AAAAAACTCGAGTCAATAA
AGTTGCATCGTTACACAATG-30. The forward primer contains an
NdeI restriction site (underlined) followed by 22 bases that corre-
spond to the 50-coding region of the HpAgD gene. The reverse pri-
mer contains an XhoI restriction site (underlined) followed by 27
bases that correspond to the 30-coding region of HpAgD. The result-
ing pET28 HpAgD construct was sequenced to ensure that no
mutations were incorporated during the PCR amplification.
Although this construct was expressed and purified by immobi-
lized metal ion affinity chromatography (IMAC), little to no activity
was apparent. This lack of activity was also found when HpAgD
was expressed using the original vector. However, active enzyme
could be purified in both high yield and high purity using conven-
tional methods (vide infra). Because IMAC was not required for
enzyme purification, the N-terminal 6ꢂ-His-tag was removed by
sub-cloning the HpAgD gene into the pET21 vector, thereby gener-
ating the p21HpAgD expression construct.
- NH3
- N-carbamoyl putrescine
+ agmatine
4
2
+ H2O
H
N
-O
O
O
O-
H
N
O
H
N
H
3
HOH
HN
O-
N
N
H
S-
H
S
N
N
H
O-
H
O
O
Fig. 2. Proposed catalytic mechanism: Cys324 is the active site nucleophile and
His202 acts as a general acid, donating a proton to the departing amine, leading to
the formation of an S-alkyl thiouronium intermediate. His202 activates a water
molecule for nucleophilic attack on the thiouronium intermediate. This leads to the
formation of a second tetrahedral intermediate that collapses to eliminate the Cys
thiolate and generate NCP.
after collapse of the tetrahedral intermediate. His202 is also
thought to act as a general base, activating a water molecule for
nucleophilic attack, thus leading to hydrolysis of the thiouronium
intermediate generated in the previous step.
2.3. Purification of HpAgD
While HpAgD shows homology to the PADs, and other guanid-
inium modifying enzymes, we demonstrate here for the first time
that this enzyme is actually an AgD, as originally suggested by bio-
informatic analyses [13,26]. We further demonstrate that this is a
hydrolytic reaction and that NCP and ammonia are produced in
equimolar amounts. Site-directed mutagenesis experiments addi-
tionally confirm the key catalytic role of Cys324. The structure of
HpAgD, which has been determined to 2.1 Å resolution, also pro-
vides a molecular basis for the preferential deimination of agma-
tine by this enzyme, and identifies Asp198 as a key residue
responsible for agmatine recognition, which has been confirmed
by site-directed mutagenesis. Finally, we describe the design, syn-
thesis, and characterization of two haloacetamidine-based inacti-
vators of HpAgD, i.e., N-(4-aminobutyl)-2-fluoro-ethanimidamide
(ABFA) and N-(4-aminobutyl)-2-chloro-ethanimidamide (ABCA).
These compounds, which inactivate HpAgD by modifying Cys324,
are the most potent AgD inhibitors described to date.
The methodology to purify HpAgD is outlined in detail in the
Supplemental Data section. Briefly, the p21HpAgD expression con-
struct, transformed into E. coli BL21(DE3) cells, was used for protein
expression. Cultures were grown in a baffled flask (37 °C, 250 rpm)
to an OD600 of 1.0. Expression of HpAgD was then induced by the
addition of IPTG (0.3 mM final) and the cultures were then incu-
bated overnight (16 °C, 250 rpm). Cells were harvested by centrifu-
gation (4400 g, 10 min, 4 °C), resuspended, and lysed. Streptomycin
was then added to clarified cell extracts to afford the precipitation
of contaminating DNA and enhance enzyme binding to a Q Sephar-
ose strong anion-exchange column. Protein recovered from this
step was further purified using a combination of heparin affinity
and size exclusion chromatography. This methodology afforded ac-
tive HpAgD in high yield (10–30 mg/L of E. coli cell culture) with
greater than 95% purity (Table S1 and Fig. S1). Enzyme prepared
in this fashion was stable for several months at ꢁ80 °C.
2. Materials and methods
2.4. Kinetic assay
2.1. Chemicals and general experimental methods
A previously established discontinuous assay that monitors the
production of ureido-containing compounds was adapted for use
as an HpAgD activity assay [28,29]. The steady state kinetic param-
eters for HpAgD were determined by pre-incubating varying
amounts of agmatine in Assay Buffer (50 mM HEPES, pH 8.0,
Agmatine sulfate,
arginine ethyl ester (BAEE) were acquired from Sigma Aldrich (St.
Louis, MO). N- -Benzoyl- -arginine amide (BAA) was obtained
L-arginine, streptomycin, and N-a-Benzoyl-L-
a
L
from Acros (Hampton, NH). Guanidine–HCl was purchased from
60 lL final volume) at 37 °C. After 10 min the reactions were initi-
EMD Bioscience (Darmstadt, Germany). Deuterium oxide was ob-
ated by the addition of enzyme. Under these conditions, enzymatic
activity was linear with respect to both time and enzyme concen-
tration. Standard errors were typically less than 20%. Initial rates
were fit, by a non-linear least squares method, to Eq. (1) using
GraFit (version 5.0.11) [30].
tained from Cambridge Isotope Laboratories (Andover, MA). N-
a-
Benzoyl- -arginine methyl ester (BAME) was purchased from MP
L
Biomedicals (Irvine, CA). Histone H4 was isolated and purified as
previously described [27]. The Ac-H4-5 peptide (Ac-SGRGK-COOꢁ)
was synthesized using the Fmoc-strategy, purified by reverse
phase HPLC, and its structure confirmed by MALDI-MS.
v
¼ Vmax½Sꢃ=ðkm þ ½SꢃÞ
ð1Þ

J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
65
2.5. Substrate specificity
d 4.25 (s, 2H), 3.22 (m, 2H), 2.87 (m, 2H), 1.58 (m, 4H). 13C NMR
(D2O, 75.47 MHz): d 162.98, 42.15, 39.22, 39.06, 24.21, 23.82. HR-
ESI-MS (C6H15N3Cl+): calculated 164.0954, observed 164.0948.
The enzymatic activity of HpAgD was evaluated against several
guanidinium containing compounds, including: L-arginine, strepto-
mycin, BAA, guanidine, BAEE, BAME, the Ac-H4-5 peptide, histone
H4, and agmatine. All substrates, with the exception of the Ac-H4-5
peptide and histone H4, were tested at a final concentration of
10 mM. The Ac-H4-5 peptide and histone H4 were evaluated at
2.10. Synthesis of N-(4-aminobutyl)-2-fluoro-ethanimidamide (ABFA)
ABFA was prepared using the methodology outlined above for
ABCA, except that ethyl-2-fluoroacetimidate (93.5 mg, 0.66 mmol)
was used in place of ethyl-2-chloroacetimidate. Ethyl-2-fluoroace-
timidate was prepared according to previously established meth-
ods [33,34]. The title compound (72 mg, 84% yield) was obtained
final concentrations of 5 mM and 125 lM, respectively. Enzyme
assays were performed as described above, except that the reac-
tions were incubated for 2–3 h to obtain estimates of the reaction
rate.
as a clear residue. 1H NMR (D2O, 300.107 MHz): d 5.11 (d, JH–F
=
45 Hz, 2H), 3.24 (m, 2H), 2.87 (m, 2H), 1.57 (m, 4H). 13C NMR
2.6. Metal inhibition
(D2O, 75.47 MHz): d 162.73 (d, JC–F = 11.1 Hz), 77.75 (d, JC–F =
177.5 Hz), 41.56, 39.04, 24.20, 23.93. HR-ESI-MS ðC6H15
N
þ3F Þ: calcu-
The effects of Ni2+, Zn2+, and Co2+ on HpAgD activity were deter-
mined by pre-incubating HpAgD with varying concentrations of
each metal ion for 15 min prior to initiating the reaction with agm-
atine (5.0 mM final). After an additional 15 min, the reactions were
quenched and NCP production was quantified using the methodol-
ogy outlined above. Initial rates were fit to Eq. (2) using GraFit (ver-
sion 5.0.11) [30], where [I] is the concentration of inhibitor (in this
case the metal ion) and IC50 is the concentration of metal ion that
yields half-maximal activity.
lated 148.1250, observed 148.1255.
2.11. Determination of IC50 values
The effects of ABFA and ABCA on HpAgD activity were deter-
mined by pre-incubating HpAgD with varying concentrations of
each compound in Assay Buffer for 15 min prior to the addition
of agmatine (0.5 mM final). After 15 min, the reactions were flash
frozen in liquid nitrogen and the results analyzed using the meth-
odology outlined above for NCP quantification. Initial rates were fit
to Eq. (2) using GraFit (version 5.0.11) [30].
Fractional Activity ¼ 1=ð1 þ ð½Iꢃ=IC50ÞÞ
ð2Þ
2.7. 18Oxygen incorporation studies
2.12. Inactivation kinetics of ABFA and ABCA
Oxygen incorporation experiments were conducted in Assay
Buffer made up in either 16O- or 18O- labeled water – the mole per-
cent of 18O-label was 70% final. Agmatine (5.0 mM final) and
Varying concentrations of ABFA and ABCA were pre-incubated
with agmatine (0.5 mM) in Assay Buffer at 37 °C. After 10 min,
the reactions were initiated by the addition of HpAgD. Aliquots
HpAgD (2.8 lM final) were added and the reaction mixture was
(60 lL) of the reactions were removed and flash frozen in liquid
incubated at 37 °C for 3 h. The reaction mixtures were desalted
using a C-18 ZipTip (Millipore), and the deiminated products were
analyzed by ESI mass spectrometry.
nitrogen at various time points. NCP levels were quantified using
the methodology outlined above. The non-linear progress curves
were fit to Eq. (3), using GraFit (version 5.0.11) [30], where vi is
the initial velocity, kobs app is the apparent pseudo first order rate
.
2.8. Ammonia production
constant for inactivation, t is time, and [NCP] is the concentration
of NCP produced during the reaction.
ꢄt
Time course experiments that monitor the production of
ammonia and NCP by HpAgD were performed in parallel. For the
ammonia production assay, agmatine (10.0 mM) in Assay Buffer
was pre-incubated at 37 °C for 10 min prior to the addition of
NCP ¼ við1 ꢁ eꢁkobs:app Þ=kobs:app
ð3Þ
To obtain kobs, the apparent pseudo first order rate constant
(kobs app) was multiplied by the transformation (1+[S]/Km). Because
.
HpAgD (2.8
were removed and quenched with iodoacetamide (50 mM,
180 L) at 0, 2, 4, 6, 10, and 15 min. NCP production was measured
lM final). Aliquots (60 lL) of the reaction mixture
saturation was observed in the plots of kobs versus [I], the data were
fit to Eq. (4) using GraFit (version 5.0.11) [30]. In this equation, ki-
l
corresponds to the maximal rate of inactivation and KI is the
nact
using the assay described above, and ammonia concentrations
concentration of I that yields half-maximal inactivation.
were measured fluorometrically [31,32].
Kobs ¼ kinact½Iꢃ=ðKI þ ½IꢃÞ
ð4Þ
2.9. Synthesis of N-(4-aminobutyl)-2-chloro-ethanimidamide (ABCA)
2.13. Substrate protection experiments
To a flask containing dry methanol (8 mL) was added N-1-Boc-1,4-
diaminobutane hydrochloride (Bachem, 75.0 mg, 0.33 mmol), ethyl-
2-chloroacetimidate (104.3 mg, 0.66 mmol), and cesium carbonate
(162.5 mg, 0.50 mmol). Ethyl-2-chloroacetimidate was prepared
according to previously established methods [33,34]. The reaction
mixture was stirred overnight at rt before being concentrated under
reduced pressure. The concentrated sample was loaded onto a col-
umn of silica gel and eluted with 40% methanol in ether to afford
N-1-Boc-ABCA as a clear residue. Boc-deprotection was accomplished
by treating the residue with neat TFA (5 mL) at 0 °C. After 30 min, the
mixture was allowed to warm to rt and stir for an additional 30 min.
The TFA was then removed, and the remaining residue was dissolved
in water (10 ml). This solution was extracted with ether (3 mL) and
the aqueous portion was lyophilized to afford the title compound
(90 mg, 95% yield) as a clear residue. 1H NMR (D2O, 300.107 MHz):
To evaluate the ability of agmatine to protect HpAgD against
inactivation by ABFA and ABCA, progress curves were generated
for reactions, both in the presence and in the absence of inactivator
(ABFA: 500 lM or ABCA: 175 lM), at two different concentrations
of agmatine (0.5 or 1.0 mM). The Assay Buffer was incubated at
37 °C for 10 min before HpAgD was added to initiate the reaction.
Aliquots (60 lL) of the reaction were removed, flash frozen, and
analyzed at 0, 2, 4, 6, 10, 12 and 14 min. The progress curves were
fit to Eq. (3) as described above.
2.14. Irreversibility of inactivation
To demonstrate that ABFA and ABCA irreversibly inactivate
HpAgD, a solution of HpAgD (2.8 lM) was treated with an excess

66
J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
Table 1
of each inhibitor (1.0 mM final) at 37 °C. After 1 h, the reactions
were dialyzed against 50 mM HEPES pH 8.0, 1.0 mM EDTA, and
10% glycerol for 20 h. The residual activity was measured using
our standard assay for NCP production.
Crystallographic data and refinement statistics.
X-ray source
PDB ID
Wavelength (Å)
APS SER-CAT 22ID
3HVM
100000
Space group
Unit cell dimensions:
a (Å)
P3221
2.15. Site-directed mutagenesis
52.08
c (Å)
246.40
50.0–2.10 (2.18–2.10)
0.47
76.0 (59.8)
0.043
65011
16.5
3.6 (3.2)
0.016
The C324A and D198A HpAgD mutants were generated using
the Quik Change site-directed mutagenesis kit (Stratagene) using
pET21-HpAgD as the template. The forward and reverse primers
for the C324A mutant are 50-CAGCATGGAAGCTTGCA TGCTGTA
ACGATGCAACTTTATTG-30 and 50-CAATAAAGTTGCATCGTTACAGCA
TGCAAGCTTCCATGCTG-30, respectively. The forward and reverse
primers for the D198A mutant are 50-CTATTTAAAGGGCGATGCT
ACGGATAGCCATACCG-30 and 50-CGGTATGGCTATCCGTAGCATCG
CCCTTTAAATAG-30, respectively. Both mutated genes were se-
quenced in their entirety to ensure that only the desired mutation
had been incorporated into the HpAgD expression construct. The
mutant enzymes were expressed and purified as described above
for the wild-type enzyme.
Resolution range (Å) (highest shell)
Mosaicity (°)
Completeness (%) (highest shell)
Total linear R-merge
Reflections observed
I over average sigma (I)
Redundancy (highest shell)
Rmsd bonds (Å)
Rmsd angles (°)
1.73
Ramachandran statistics
Residues in most favored regions (%)
Residues in additionally allowed regions (%)
Residues in generously allowed regions (%)
Residues in disallowed regions (%)
R Refmac5
265 (88.6)
32 (10.7)
1 (0.3)
1 (0.3)
0.200
Rfree Refmac5
0.259
2.16. Tryptic digest
To identify the site of inactivator modification, HpAgD (100
M) was incubated at 37 °C with ABFA (132 M, 5 eq.), ABCA
(132 M, 5 eq.), or no inactivator in Assay Buffer for 30 min. A
tryptic digest was then performed using Seq. Grade Modified
Trypsin (Promega) in a final ratio of 1:20 (enz:sub) and incubated
overnight at 37 °C. The digested sample was prepared for
MALDI-TOF MS and MALDI-TOF MS/MS using a C-18 ZipTip.
lg,
Although this construct has previously been used to express and
purify recombinant HpAgD in E. coli (PDB ID: 2CMU) the enzyme
is inactive (data not shown). Due to the proximity of the C-termi-
nus to the active site cysteine (Cys324), we postulated that the lack
of activity may be a result of the C-terminal His-tag and/or the
additional C-terminal residues. In order to mitigate these effects,
the superfluous residues were deleted and the His-tag was moved
to the N-terminus by sub-cloning the HpAgD gene into a modified
pET28 vector. While this new construct, i.e., pET28HpAgD, was
used to purify the enzyme by IMAC, the enzyme also possessed
essentially no activity (data not shown). By contrast, conventional
purification methods yielded active enzyme in both high yield and
high purity. Upon recognizing that the His-tag was unnecessary for
enzyme purification, the HpAgD gene was subcloned into the
pET21 vector to eliminate this tag, thereby enabling the expression
and purification of the untagged wild-type protein, also in both
high yield (10–30 mg/L of E. coli cell culture) and purity (Table S1
and Fig. S1). Consistent with previous studies with Arabidopsis tha-
liana AgD [40] and P. aeruginosa AgD [26], HpAgD also purifies as a
dimer by size exclusion chromatography (not shown).
26
l
l
l
2.17. Crystallization of HpAgD
Crystals of HpAgD were grown from 15% PEG 4 K, 10 mM so-
dium citrate pH 5.6, and 7% isopropanol by vapor diffusion, hang-
ing-drop method. Crystals were transferred to a cryoprotectant
solution, mother liquor plus 15% ethylene glycol, and then flash
frozen in a cryo stream.
2.18. X-ray diffraction data measurement and processing
X-ray diffraction experiments were carried out at the SER-CAT ID
beamline at APS, Argonne National Laboratory. Data were collected
on a single crystal of HpAgD with approximate dimensions of
0.60 ꢂ 0.30 ꢂ 0.25 mm. The data were indexed and processed with
HKL2000 [35]. Data collection and processing statistics are shown
in Table 1.
3.2. Substrate specificity studies
Given the fact that HpAgD is annotated as a PAD, we initially
examined whether this enzyme could deiminate a number of pro-
tein and peptide substrates, including, histone H4, a peptide based
on the N-terminus of histone H4 (i.e., the Ac-H4-5 peptide), as well
as a number of benzoylated arginine derivatives (i.e., BAA, BAEE,
BAME), all of which are known PAD substrates [28]. However, none
of these compounds served as a substrate for the enzyme. Given
this result, a variety of guanidinium containing small molecules
were then tested for their ability to act as HpAgD substrates. These
2.19. Structure determination and refinement
The structure of HpAgD (PDB ID: 2CMU), was used as the search
model in the molecular replacement method carried out with Pha-
ser [36] from CCP4 [37]. Model rebuilding and crystallographic
refinements were conducted using TURBO graphics software [38]
and the Refmac5 [39] component of CCP4 [37], respectively.
3. Results and discussion
potential substrates included: guanidine, L-arginine, asymmetri-
cally dimethylated arginine (ADMA), streptomycin, and agmatine.
A comparison of the kcat/Km values obtained for these various po-
tential substrates revealed that HpAgD modifies agmatine at rates
that are >104-fold faster than the rates observed for all other sub-
strates tested (Table 2). Thus, this enzyme, like other AgDs [40,41],
demonstrates an exquisite preference for agmatine. Consistent
with the fact that agmatine is the preferred substrate, is our finding
3.1. Purification of HpAgD
An expression vector encoding HpAgD (PDB ID: 2CMU) was ob-
tained from SGX Pharmaceuticals Inc. In addition to the HpAgD
open reading frame, this construct encodes two additional N-ter-
minal residues and four additional C-terminal residues that are
linked, in frame, to a sequence encoding a C-terminal 6ꢂ-His-tag.
that the Km for agmatine is in the low lM range (Table 3). It is note-

J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
67
Table 2
kcat/Km values for potential guanidinium containing substrates of wild type and
mutant HpAgD.
Substrate
Wild type
C324A
D198A
(minꢁ1 Mꢁ1
)
(minꢁ1 Mꢁ1
)
(minꢁ1 Mꢁ1
)
H4d,f
68.55 ꢂ 10ꢁ6
61.03
NDa
62.58 ꢂ 10ꢁ7
NDa
68.96
L
-arginineb,f
Streptomycinb,e 62.26
NDa
NDa
NDa
NDa
NDa
NDa
NDa
619.2
61.98
Ac-H4-5c,e
BAAb,e
60.23
64.32
64.87
65.08
66.23
617.2
147 ꢂ 103
61.41
67.71
Guanidineb,e
BAEEb,f
64.07
68.30
BAMEb,f
613.1
ADMAb,f
Agmatine
61.46
61.7 ꢂ 103
a
b
c
d
e
f
ND = not determined.
10 mM substrate.
5 mM substrate.
125 lM substrate.
2 h incubation.
Fig. 3. The reaction catalyzed by HpAgD leads to the production of NCP and
ammonia in equimolar amounts.
3 h incubation.
3.4. Metal inhibition studies
Our inability to purify active HpAgD via IMAC prompted us to
investigate the inhibitory effects of metal ions that are typically
used for IMAC (i.e., Ni2+, Zn2+, and Co2+). IC50 values were therefore
determined for the aforementioned metal ions. The results of these
studies indicate that each of the metal ions inhibit HpAgD with
varying degrees of potency – the IC50 values for Ni2+, Zn2+, and
worthy that this value is nearly identical to that obtained for the E.
faecalis orthologue [12]. Somewhat surprisingly the kcat is quite
slow (ꢀ5 minꢁ1) relative to both the E. faecalis enzyme and other
deiminases. Although the reasons for this slow turnover are not
known, the specificity constant, i.e., kcat/Km (1.47 ꢂ 105 minꢁ1Mꢁ1),
is quite close to that obtained for both P. aeruginosa AgD
(4.2 ꢂ 105 minꢁ1Mꢁ1) [26] and PAD4 (3.0 ꢂ 105 minꢁ1Mꢁ1) [28],
indicating that HpAgD is a functional enzyme with reasonable ki-
netic parameters. Perhaps the unique role of the enzyme in bacte-
Co2+ are 100 3.9, 30 3.0, and 13 5.5
lM, respectively. Interest-
ingly, AgDs from A. thaliana and Chorella virus are readily purified
by IMAC [40,41], suggesting that the active site of HpAgD is un-
iquely susceptible to metal ion inhibition. While the mechanism
by which these metal ions inhibit HpAgD is not known, it is possi-
ble that these cations inactivate the enzyme by catalyzing the oxi-
dation of the active site cysteine or, alternatively, inhibit HpAgD
activity by coordinating with the active site cysteine and histidine.
There is some precedence for the latter possibility as Stone et al.
have shown that zinc coordinates with the corresponding residues
in P. aeruginosa DDAH [42]. Note that exhaustive treatments with
EDTA could not reactivate the enzyme (data not shown).
rial physiology is the explanation for the relatively low kcat
.
3.3. Product specificity of HpAgD
To confirm that the major products of the reaction are ammonia
and NCP, the production of these compounds was measured as a
function of time. These studies were performed for two reasons:
(i) our assay for NCP production actually detects the presence of
ureido groups and (ii) other guanidinium modifying enzyme family
members can catalyze amidino group transfer reactions [13]. Thus
it was conceivable that the first tetrahedral intermediate could
break down to form putrescine and a thiouronium intermediate
that could ultimately be hydrolyzed to form urea. As depicted in
Fig. 3, the results indicate that ammonia and NCP are produced
in approximately equimolar amounts, thereby confirming that
HpAgD catalyzes the production of agmatine to form NCP and
ammonia. To confirm that HpAgD is a hydrolase, HpAgD catalyzed
reactions were performed in 18O-labeled water. Control reactions
in 16O-labeled water were also performed. The observed masses
(i.e., 132 and 134 Da) are consistent with those expected for the
production of NCP (i.e., 132 and 134 Da) in 16O- and 18O-labeled
H2O, respectively. The two unit mass shift observed for the reaction
performed in 18O-labeled water also confirms that HpAgD cata-
lyzes the hydrolytic deimination of agmatine to form NCP.
3.5. Structure and mutagenesis of HpAgD
To aid our understanding of HpAgD catalysis and substrate rec-
ognition, the structure of the enzyme was determined to 2.1 Å res-
olution. The overall structure of HpAgD consists of five repeated
bbab subdomains that are arranged in space to form a pseudo-5-
fold symmetric structure that has the appearance of an
a
/b propel-
ler. The active site is at the ‘boss’, or hub, of the propeller in a deep
pocket that contains key active site residues (Fig. 4A).
The structure of His-tagged selenomethionine-containing
HpAgD has previously been determined to 2.5 Å resolution (PDB
ID: 2CMU). Although this enzyme was purified by IMAC, and pre-
sumably represents a catalytically inactive form of the enzyme,
the overall structure of the enzyme is very similar to that reported
here. The rmsd of C positions between the two structures is
a
0.26 Å; the deviations over 1 Å are only for loop 262–266 with
the largest deviation being 1.3 Å. We have recalculated electron
density maps for the unpublished structure (PDB ID: 2CMU), using
structure factors deposited in the PDB. Although extra electron
density adjacent to the active site Cys is apparent in omit maps
comparing our structure to the 2CMU structure, the maps are
inconclusive. A water molecule placed next to the active site cys-
teine is at a distance of 2.5 Å from the b-sulfur atom of the active
site Cys. This distance is too short for a non-bonded oxygen–sulfur
interaction. As a Ni-column was used for the purification of this en-
zyme, it is possible that the extra electron density corresponds to a
Table 3
Steady-state kinetic parameters for agmatine with wild type and mutant HpAgD.
Enzyme
Substrate
kcat (minꢁ1
)
Km
(
l
M)
kcat/Km (minꢁ1 Mꢁ1
)
Wild type
C324Aa
Agmatine
Agmatinea
Agmatine
4.90 0.29
NDb
0.28 0.01
33 8.8
NDb
170 79.8
147 ꢂ 103
619.2
D198Aa
1.7 ꢂ 103
a
10 mM agmatine.
ND = not determined.
b

68
J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
Fig. 4. Structure of HpAgD: (A) The overall structure of HpAgD. The catalytic side chains, i.e., Asp84, His202, Asp204, and Cys324, are shown. (B) Superposition of the
structures of HpAgD (in blue), PAD4 (in green) and EfAgD (in yellow). Calcium ions, represented by black spheres, and sulfates are from the PAD4 structure. The position of the
active site is marked by BAA (in red). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
nickel ion coordinated to the active site cysteine, however, the
environment of the sulfur ligand does not appear to be consistent
with this hypothesis. The surrounding residues are Asp199 at
3.14 Å and His204 at 3.17 Å from the ligand. These distances are
significantly greater than the expected distances for oxygen and
nitrogen coordinating to nickel ion (ꢀ2.2 Å), but correspond very
well to interactions with a water molecule. Additionally, the elec-
tron density peak for the sulfur of cysteine is higher than the peak
for the hypothetical nickel ion. These findings suggest that there
may be at most partial occupancy for nickel. The observed distance
is not consistent with the alternative hypothesis of sulfur oxida-
tion. However a combination of these two hypotheses with partial
occupancy would explain the observed lack of activity.
(EfAgD). No major differences are observed in the superpositions
of HpAgD with PAD4 and EfAgD (Fig. 4B) with an overall rmsd of
2.66 Å and 1.59 Å, respectively. The large displacement value for
the difference between HpAgD and PAD4 is due to the movement
of flexible loops.
To identify residues that are important for catalysis, a structure
based alignment of PAD4 and HpAgD was generated using the
Matchmaker function in UCSF Chimera [43]. The results of these
computational studies showed that Asp84, His202, Asp204, and
Cys324 are analogous to the key catalytic residues in PAD4 (i.e.,
Asp350, His471, Asp473, and Cys645) [44]. A superposition of
these residues in HpAgD and PAD4 (rmsd = 0.145 Å), using Lsqkab
[45] in CCP4 [37], demonstrates that they are positioned appropri-
ately to catalyze agmatine deimination (Fig. 5A). One major differ-
ence in these two active sites is the presence of a well positioned,
and highly ordered, water molecule that is roughly equidistant
The structure of the enzyme is also quite similar to that ob-
served for other members of the guanidinium modifying family
of enzymes, including both PAD4 and the AgD from E. faecalis
Fig. 5. Superpositions of HpAgD and PAD4: (A) The active site cleft of HpAgD. Residues that line the entrance to the active site are highlighted. (B) Superposition of the active
site residues of HpAgD (in atom color) and PAD4 (in blue) highlighting the residues that are involved in substrate recognition. A water molecule from HpAgD is shown in red.
BAA, from the PAD4 structure, is shown in blue with functional moieties in atom color. (C) Entrance to the active site with HpAgD represented by its molecular surface (gray).
The position of BAA, shown using stick representation, was modeled by superpositioning the PAD4 complex.

J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
69
between His202 and Cys324 and occupies the space otherwise
occupied by the guanidinium carbon of the substrate. Although
the existence of this water molecule could be an artifact of crystal-
lization at pH 5.6, its presence suggests the potential formation of a
water mediated shared ion pair that could facilitate proton transfer
from the Cys324 thiol to the imidazole moiety of His202 (Fig. 5A);
thereby generating the catalytically competent protonation states
of these two residues. The presence of a water mediated shared
ion pair would also be expected to prevent non-specific oxidation
of the active site cysteine. While PAD4 is also thought to possess
a shared ion pair [25], additional studies are necessary to validate
a role for this water molecule in HpAgD catalysis, and determine
whether proton transfer occurs prior to or concomitant with sub-
strate binding.
Based on the above described structural similarity, it is likely
that HpAgD and PAD4 use analogous catalytic mechanisms. Briefly,
nucleophilic attack by the active site thiolate (Cys324 in HpAgD)
on the guanidinium carbon of agmatine results in the formation
of a tetrahedral intermediate. Protonation of the departing amine
and collapse of the intermediate results in the formation of a stable
thiouronium adduct. This thiouronium intermediate can then be
hydrolyzed to form NCP and regenerate the enzyme (Fig. 2).
In order to verify that Cys324 is critical for catalysis, this residue
was changed to an alanine by site-directed mutagenesis. This mu-
tant protein was expressed, isolated, and purified in a manner iden-
tical to that of the wild-type enzyme. Activity assays of this mutant
enzyme indicated only very limited activity (kcat/Km 6 19.2 minꢁ1
Mꢁ1; Table 3). This level of activity represents a P8000-fold reduc-
tion in HpAgD activity over the wild-type enzyme – a result that is
consistent with both the putative role of this residue in nucleophilic
catalysis and with results obtained for other guanidinium modify-
ing enzymes (e.g. PAD4 and ADI) [25,46]. However, it is interesting
to note that the mutation of the corresponding Cys in A. thaliana
AgD decreases activity by 6.7-fold [40]; this discrepancy is most
likely an artifact of the assay employed in those studies.
3.6. Substrate recognition
The overwhelming preference of HpAgD for agmatine can be ex-
plained, at least in part, by comparing the structures of PAD4 and
HpAgD. The active site cleft of HpAgD is lined with several residues,
including: Trp106, Lys109, Tyr110, Thr199, Asp198, Glu263, and
Gln318 (Fig. 5). The presence of these residues results in an active
site cleft that is considerably narrower than the corresponding re-
gion in PAD4 (Fig. 5), likely preventing the binding of peptide or pro-
tein-based substrates. Of these residues, Trp106 appears to play a
major role in excluding arginine-containing peptide substrates, as
the 1.5 Å distance between the carboxamide moiety of BAA and
Trp106, in the modeled complex, would preclude the binding of such
substrates via steric hindrance (Fig. 5). No effect on agmatine bind-
ing would be expected because this substrate lacks the carboxamide
group. Note that the side chain of Trp106 cannot be repositioned to
analternaterotamerwithoutcreatingintramolecularclashes. Asim-
ilar steric hindrance between Trp106 in HpAgD and the carboxylate
of arginine is observed in a superimposition of HpAgD and the
PaADIꢅarginine complex (not shown), thereby providing a mecha-
nism to exclude arginine from the binding pocket of HpAgD.
It is also noteworthy that Trp106 occupies the space normally
occupied by an arginine residue that corresponds to Arg374 in
PAD4 and Arg243 in P. aeruginosa arginine deiminase (PaADI).
The absence of this residue is significant because, in PAD4, it forms
favorable interactions with the backbone carbonyls immediately
surrounding the site of deimination. Similarly, in PaADI, this argi-
nine interacts with the
a-carboxyl group of free arginine. The loss
of these interactions, combined with the narrower active site,
Fig. 6. (A) Structures of ABXA and X-amidine. (B) Proposed mechanisms of inactivation. Both mechanisms lead to the inactivation of HpAgD through a thioether linkage.

70
J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
would therefore explain, in part, why arginine and peptidyl-argi-
nine are poor HpAgD substrates.
IC50 values obtained from this experiment (ABFA: 6.8 0.26
lM;
ABCA: 0.87 0.03 M; Table 4) indicate that ABCA is a significantly
l
The presence of Asp198 in HpAgD also likely contributes to the
exclusion of both arginine and peptidyl-arginine based substrates.
This is the case because overlays of unliganded HpAgD and the
PAD4 complex show that the Asp198 side chain is 2.4 Å from the
carboxamide group of BAA in the PAD4ꢅBAA complex. A similar
arrangement is observed in a superimposition of HpAgD and the
PaADIꢅarginine complex (not shown). Thus, electrostatic repulsions
and/or sterics would combine to exclude arginine and peptidyl-
arginine containing substrates from the HpAgD active site.
Although readily accessible alternate rotamers remove the steric
hindrance, it is likely that Asp198 plays an additional role in select-
ing agmatine over arginine or arginine-containing peptide sub-
more potent inactivator of HpAgD than is ABFA. This result is not
unexpected given the greater leaving group potential of chloride.
To examine the selectivity of these compounds, their ability to
inactivate PAD4 was determined. The results of these experiments
indicate that ABFA and ABCA are very poor inhibitors of PAD4 –
essentially no inhibition was observed at concentrations as high
as 1 mM. Our previously described PAD4 inhibitors, denoted F-
amidine and Cl-amidine (Fig. 6), were also tested to determine if
either of these compounds inhibit HpAgD. The results of these
experiments demonstrate that these latter compounds are quite
poor HpAgD inhibitors – concentrations as high as 500 lM did
not afford HpAgD inhibition. The fact that F-amidine and Cl-ami-
dine are poor HpAgD inhibitors is consistent with the fact that this
enzyme shows an exquisite preference for agmatine as a substrate.
Additionally, the fact that significant selectivity can be achieved,
between the PADs and the AgDs, by simply changing the context
of the reactive moiety shows the versatility of the haloacetamidine
warhead, and bodes well for the future use of this moiety in gener-
ating chemical probes for the diverse members of the guanidinium
strates because, in the overlays, this residue is 3.2 Å from the
a-
amino group of the arginine. Thus, in addition to excluding argi-
nine, this residue may form favorable electrostatic interactions
that promote agmatine binding. Consistent with this notion is
the fact that a glutamate residue, i.e., Glu214, occupies a nearly
identical position in E. faecalis AgD [12]. To verify this possibility
experimentally, an Asp198Ala mutant was constructed, expressed,
and purified to homogeneity. The results of kinetic analyses
(Table 3) show an 86-fold decrease in kcat/Km, a 17-fold decrease
in kcat, and a 5-fold increase in Km, consistent with this residue
playing an important role in substrate binding. Although this
mutation might be expected to ‘open up’ the substrate binding
cleft and remove a residue that would form unfavorable interac-
tions with both arginine and peptidyl-arginine containing sub-
strates, agmatine remained the preferred substrate (Table 2).
3.7. Design and synthesis of AgD inactivators
After confirming that HpAgD is in fact a bona fide AgD, we set
out to develop inhibitors for this enzyme. Because HpAgD pos-
sesses structural and mechanistic features that are similar to
PAD4, and given our previously successful efforts to inhibit PAD4
with haloacetamidine-containing peptidyl-arginine analogs
[33,34], we designed and synthesized haloacetamidine-containing
agmatine analogs. The structures of these compounds (Fig. 6) re-
place the guanidinium moiety of agmatine with either a fluoro-
or chloroacetamidine moiety. These compounds, denoted N-(4-
aminobutyl)-2-fluoro-ethanimidamide (ABFA) and N-(4-aminobu-
tyl)-2-chloro-ethanimidamide (ABCA), were synthesized over two
synthetic steps (reaction of mono-Boc protected putrescine with
a 2-haloacetimidate followed by deprotection to reveal the termi-
nal amine) and isolated in reasonable yields. Similarly to our PAD4
inactivators F- and Cl-amidine [33,34], ABFA and ABCA show excel-
lent stability in aqueous buffer (not shown).
3.8. Inactivation of HpAgD with ABFA and ABCA
To begin to characterize the inhibitory properties of ABFA and
ABCA, IC50 values were determined. For these reactions, inactiva-
tors were pre-incubated with HpAgD for 15 min prior to the addi-
tion of agmatine (0.5 mM) to initiate the deimination reaction. The
Table 4
IC50 values.
Inhibitor
HpAgDa
(lM)
PAD4b
(lM)
ABCA
ABFA
Cl-amidine
F-amidine
0.87 0.03
6.8 0.26
>500
>1000
>1000
5.9 0.3
21.6 0.3
Fig. 7. (A) ABFA and ABCA irreversibly inhibit HpAgD. Both inhibitors were reacted
with HpAgD for 1 h. The ABFA/ABCA:HpAgD complex was dialyzed at which point
the residual activity was measured. (B and C) Substrate protection experiment with
ABFA (B) and ABCA (C). The amount of product was measured as a function of time
with varying amounts of inhibitor and substrate.
>500
a
0.5 mM agmatine.
10 mM BAEE.
b

J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
71
modifying family of enzymes. Consistent with this notion is the
fact that fluorescently-tagged versions of F- and Cl-amidine show
very little non-specific modification of proteins in whole proteo-
mes [47].
close to the calculated value, i.e., 1643.815. Fragmentation of this
ion, and analysis by MALDI-MS/MS, resulted in the production of
a series of y and b ions that are consistent with the modification
of Cys324 (Figs. 8 and S3). In total, these results confirm that ABFA
and ABCA inactivate HpAgD by modifying Cys324.
To confirm that ABFA and ABCA irreversibly inactivate HpAgD,
pre-formed HpAgDꢅinactivator complexes were dialyzed for 20 h,
after which point, the residual activity was measured. As depicted
in Fig. 7, enzyme activity was not recovered, which is consistent
with irreversible enzyme inactivation. Substrate protection exper-
iments were subsequently conducted to demonstrate that enzyme
inactivation was due to the modification of an active site residue.
For these experiments, product formation was measured as a func-
tion of time in both the absence and in the presence of both ABFA
and ABCA at two different concentrations of agmatine (i.e., 0.5 and
1.0 mM). In the absence of added inhibitor, the progress curves are
linear, as expected. However, in the presence of ABFA and ABCA,
the progress curves are non-linear, a result that is consistent with
the irreversible inactivation of HpAgD (Fig. 7). Notably, at higher
concentrations of substrate, the rate of enzyme inactivation is im-
peded, demonstrating that substrate can protect against enzyme
inactivation. This finding is consistent with the modification of
an active site residue.
To confirm that inactivation was due to the modification of the
active site cysteine (i.e., Cys324), MS and MS/MS analyses were
performed. For these experiments, pre-formed HpAgDꢅinactivator
complexes were digested with trypsin and the resultant peptides
characterized by MALDI-MS. Peptide mass fingerprints were then
constructed for both HpAgD treated with ABFA and ABCA
(Fig. S2). Sequence coverage of 45% and 33% was obtained for the
ABFA and ABCA treated HpAgD, respectively. This coverage com-
pared quite favorably to untreated control samples (sequence cov-
erage = 34%). Notably, a tryptic peptide encompassing Cys324, the
active site cysteine, was absent from both the ABFA and ABCA trea-
ted samples. However, a new signal consistent with the mass of the
covalent adduct was apparent in the MALDI-MS spectra obtained
from samples treated with either of these compounds. The ob-
served m/z values obtained from digests of the ABCA and ABFA
treated enzyme, 1643.816 and 1643.860, respectively, are quite
The kinetics of HpAgD inactivation by ABFA and ABCA were
subsequently determined. In these experiments, product forma-
tion, in the presence of varying concentrations of ABCA and ABFA,
was measured as a function of time (Fig. 9). The pseudo first order
rate constants of inactivation (i.e., kobs), obtained from the non-lin-
ear progress curves, after correction for the substrate concentra-
tion, were plotted against inactivator concentration. The
resultant plots, which are hyperbolic (Fig. 9C and D), are consistent
with a two step mechanism of inactivation where the inactivator
first binds to the enzyme and then reacts with Cys324 to afford
the formation of the covalent adduct. The kinact, KI, and kinact/KI val-
ues obtained from analysis of the data in Fig. 9C and D are
530 150
1200 234
l
l
M, 2.1 0.28 minꢁ1, 4000 minꢁ1 Mꢁ1 for ABFA and
M, 15 1.9 minꢁ1, and 12500 minꢁ1 Mꢁ1 for ABCA.
Interestingly, the kinact/KI values for ABFA and ABCA (4000 min-
1 Mꢁ1 and 12500 minꢁ1 Mꢁ1, respectively) are similar to those ob-
tained for the inactivation of PAD4 by F- and Cl-amidine
(3000 minꢁ1 Mꢁ1 and 13000 minꢁ1 Mꢁ1, respectively) despite the
fact that their kcat values differ by greater than 50-fold. This result
is at least partially consistent with the fact that these enzymes
have very similar kcat/Km values. Based on the fact that the side
chain portion of F- and Cl-amidine imparts a greater than 200-fold
enhancement in kinact/KI ([25,48] and unpublished results), the
amino-alkyl portion of ABFA and ABCA likely enhances inhibitor
binding to a similar extent. Further elaboration of the alkyl chain
should lead to the development of inhibitors with improved
potency.
4. Conclusions
The near constant emergence of antibiotic resistance mecha-
nisms to virtually all known antibiotics [49] highlights the priority
Fig. 8. Site of ABCA modification determined by MALDI-TOF MS and MS/MS: (A) Unmodified HpAgD after tryptic digest. (B) ABCA modified HpAgD after tryptic digest. (C)
Sequence of the predicted site of modification. The b and y ions identified by MALDI-TOF MS/MS are shown in grey.

72
J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
Fig. 9. Inactivation kinetics of ABFA and ABCA. The amount of NCP produced with varying amounts of ABCA (A) and ABFA (B) was measured. The pseudo first order rate
constants (kobs) for inactivation was plotted versus the concentration of ABCA (C) and ABFA (D) were plotted.
that must be placed on the development of new and novel antibi-
otics. Advances in genomics have, and continue to, unveil cellular
processes that are unique to bacteria and such pathways represent
novel targets for antibiotic development because they limit poten-
tial side effects. Given that the deimination of agmatine to form
N-carbamoyl putrescine is one such pathway, HpAgD represents
a potentially novel target for antibiotic development. As such, we
set out to characterize both the substrate specificity and the cata-
lytic mechanism of HpAgD, to determine its structure, and to
develop HpAgD inhibitors. The mechanistic and structural studies
described herein demonstrate that HpAgD is a bona fide agmatine
deiminase that catalyzes the hydrolytic conversion of agmatine
to NCP and ammonia, in a fashion that is analogous to mechanism
employed by PAD4.
Although this enzyme is currently annotated in structural and
genomic databases as a peptidyl-arginine deiminase, our results
definitively establish that this enzyme is, in fact, an AgD. This study
highlights how important it is to obtain hard experimental data to
validate or revise (as in this case) automatic annotations based on
amino acid sequence alone. Given that dysregulated PAD activity is
strongly associated with the onset and progression of rheumatoid
arthritis [50,51], the demonstration that HpAgD is not a PAD is sig-
nificant because it rules out this enzyme as a possible trigger for
this disease – a long held, but unproven, hypothesis in the rheuma-
toid arthritis community is that this disease is caused by an infec-
tious agent of unknown etiology [52].
than current AgD inhibitors, e.g. arcaine (1,4-diguanidinobutane)
[12]. The fact that the chloro-containing compound (i.e., ABCA) dis-
plays greater potency than ABFA is most easily attributed to the
higher propensity of chloride to be displaced. These inhibitors rep-
resent chemical probes that can be used to discern the physiolog-
ical function(s) of HpAgD, and possibly represent lead compounds
for the treatment of the disease pathologies associated with H. py-
lori and other pathogenic bacteria that contain an AgD. Future
work will include efforts to increase potency, and explore the
in vivo efficacy of these compounds.
Acknowledgments
The original HpAgD clone was generously provided by SGX
Pharmaceuticals, Inc. We thank members of the Thompson labora-
tory for critical reading of the manuscript.
This work was supported in part by the University of South Car-
olina Research Foundation (PRT) and NIH Grant GM079357 to PRT.
Appendix A. Supplementary material
Supplementary Table S1 and Figs. S1, S2, and S3 are available. A
complete description of the HpAgD purification protocol is pro-
vided in the supplementary material. Supplementary data associ-
ated with this article can be found, in the online version, at
doi:10.1016/j.bioorg.2009.11.004.
Given the mechanistic similarities between these two enzymes,
and given the difference in substrate scope, we set out to design
HpAgD inactivators using the same methodology that we have suc-
cessfully employed to generate PAD4 inactivators, i.e., the use of a
haloacetamidine reactive moiety in place of the guanidinium moi-
ety of a peptidyl-arginine residue [33,34,47]. Based on this meth-
odology, the replacement of the guanidinium moiety of agmatine
with either a fluoro- or chloroacetamidine moiety led to the devel-
opment of HpAgD inhibitors that are significantly more potent
References
[1] A. Lembo, L. Caradonna, T. Magrone, M.L. Mastronardi, D. Caccavo, E. Jirillo, L.
Amati, Curr. Drug Targets Immune Endocr. Metab. Disord. 1 (3) (2001) 199–
208.
[2] J.C. Atherton, Annu. Rev. Pathol. 1 (2006) 63–96.
[3] D.S. Merrell, S. Falkow, Nature 430 (6996) (2004) 250–256.
[4] N. Vakil, F. Megraud, Gastroenterology 133 (3) (2007) 985–1001.

J.E. Jones et al. / Bioorganic Chemistry 38 (2010) 62–73
73
[5] J.M. Meyer, N.P. Silliman, W. Wang, N.Y. Siepman, J.E. Sugg, D. Morris, J. Zhang,
H. Bhattacharyya, E.C. King, R.J. Hopkins, Ann. Intern. Med. 136 (1) (2002)
13–24.
[6] L.A. Fischbach, S. van Zanten, J. Dickason, Aliment. Pharmacol. Ther. 20 (10)
(2004) 1071–1082.
[29] M. Knipp, M. Vasak, Anal. Biochem. 286 (2) (2000) 257–264.
[30] R.J. Leatherbarrow, Grafit Version 5.0.11, Erathicus Software, Staines, UK, 2004.
[31] K. Sugawara, F. Oyama, J. Biochem. (Tokyo) 89 (3) (1981) 771–774.
[32] P.L. Kearney, M. Bhatia, N.G. Jones, Y. Luo, M.C. Glascock, K.L. Catchings, M.
Yamada, P.R. Thompson, Biochemistry 44 (31) (2005) 10570–10582.
[33] Y. Luo, K. Arita, M. Bhatia, B. Knuckley, Y.H. Lee, M.R. Stallcup, M. Sato, P.R.
Thompson, Biochemistry 45 (39) (2006) 1727–11736.
[7] J.L. Telford, A. Covacci, R. Rappuoli, P. Chiara, Curr. Opin. Immunol. 9 (4) (1997)
498–503.
[8] C.S. Coleman, G. Hu, A.E. Pegg, Biochem. J. 379 (Pt 3) (2004) 849–855.
[9] G. Li, S. Regunathan, C.J. Barrow, J. Eshraghi, R. Cooper, D.J. Reis, Science 263
(5149) (1994) 966–969.
[10] S. Regunathan, D.J. Reis, Annu. Rev. Pharmacol. Toxicol. 36 (1996) 511–544.
[11] W. Raasch, S. Regunathan, G. Li, D.J. Reis, Life Sci. 56 (26) (1995) 2319–2330.
[12] J.L. Llacer, L.M. Polo, S. Tavarez, B. Alarcon, R. Hilario, V. Rubio, J. Bacteriol. 189
(4) (2007) 1254–1265.
[13] H. Shirai, Y. Mokrab, K. Mizuguchi, Proteins 64 (4) (2006) 1010–1023.
[14] A.R. Griswold, M. Jameson-Lee, R.A. Burne, J. Bacteriol. 188 (3) (2006) 834–841.
[15] A.R. Griswold, M.M. Nascimento, R.A. Burne, Oral Microbiol. Immunol. 24 (1)
(2009) 79–82.
[34] Y. Luo, B. Knuckley, Y.H. Lee, M.R. Stallcup, P.R. Thompson, J. Am. Chem. Soc.
128 (4) (2006) 1092–1093.
[35] Z. Otwinowski, W. Minor, Methods Enzymol. 276 (1997) 307–326.
[36] A.J. McCoy, R.W. Grosse-Kunstleve, L.C. Storoni, R.J. Read, Acta Crystallogr. D61
(2005) 458–464.
[37] Collaborative Computational Project, Acta Crystallogr. D 50 (1994) 760–763.
[38] A. Roussel, C. Cambillau, Turbo frodo, in: Silicon Graphics Geometry Partners
Directory, Silicon Graphics, Mountain View, CA, 1991, p. 86.
[39] G.N. Murshudov, A.A. Vagin, E.J. Dodson, Acta Crystallogr. D: Biol. Crystallogr.
53 (1997) 240–255.
[40] T. Janowitz, H. Kneifel, M. Piotrowski, FEBS Lett. 544 (1-3) (2003) 258–261.
[41] S. Baumann, A. Sander, J.R. Gurnon, G.M. Yanai-Balser, J.L. Van Etten, M.
Piotrowski, Virology 360 (1) (2007) 209–217.
[42] E.M. Stone, A.L. Costello, D.L. Tierney, W. Fast, Biochemistry 45 (17) (2006)
5618–5630.
[43] E.F. Pettersen, T.D. Goddard, C.C. Huang, G.S. Couch, D.M. Greenblatt, E.C.
Meng, T.E. Ferrin, J. Comput. Chem. 25 (13) (2004) 1605–1612.
[44] K. Arita, H. Hashimoto, T. Shimizu, K. Nakashima, M. Yamada, M. Sato, Nature
Struct. Mol. Biol. 11 (8) (2004) 777–783.
[16] J. Satriano, Amino Acids 26 (4) (2004) 321–329.
[17] M.S. Joshi, T.B. Ferguson Jr., F.K. Johnson, R.A. Johnson, S. Parthasarathy, J.R.
Lancaster Jr., Proc. Natl. Acad. Sci. USA 104 (24) (2007) 9982–9987.
[18] C.H. Su, I.M. Liu, H.H. Chung, J.T. Cheng, Neurosci. Lett. 457 (3) (2009) 125–128.
[19] C. Wolf, M. Bruss, B. Hanisch, M. Gothert, I. von Kugelgen, G.J. Molderings, Mol.
Pharmacol. 71 (1) (2007) 276–283.
[20] F.I. Bussiere, R. Chaturvedi, Y. Cheng, A.P. Gobert, M. Asim, D.R. Blumberg, H.
Xu, P.Y. Kim, A. Hacker, R.A. Casero Jr., K.T. Wilson, J. Biol. Chem. 280 (4) (2005)
2409–2412.
[45] W. Kabsch, Acta Crystallogr. A32 (1976) 922–923.
[21] G.J. Molderings, M. Burian, J. Homann, M. Nilius, M. Gothert, Digest. Dis. Sci. 44
(12) (1999) 2397–2404.
[22] G.J. Molderings, B. Kribben, A. Heinen, D. Schroder, M. Bruss, M. Gothert,
Cancer 101 (4) (2004) 858–868.
[23] A.P. Gobert, D.J. McGee, M. Akhtar, G.L. Mendz, J.C. Newton, Y. Cheng, H.L.
Mobley, K.T. Wilson, Proc. Natl. Acad. Sci. USA 98 (24) (2001) 13844–13849.
[24] H. Shirai, T.L. Blundell, K. Mizuguchi, Trends Biochem. Sci. 26 (8) (2001) 465–
468.
[25] B. Knuckley, M. Bhatia, P.R. Thompson, Biochemistry 46 (22) (2007) 6578–
6587.
[26] Y. Nakada, Y. Itoh, Microbiology 149 (Pt 3) (2003) 707–714.
[27] P.R. Thompson, H. Kurooka, Y. Nakatani, P.A. Cole, J. Biol. Chem. 276 (36)
(2001) 33721–33729.
[28] P.L. Kearney, M. Bhatia, N.G. Jones, L. Yuan, M.C. Glascock, K.L. Catchings, M.
Yamada, P.R. Thompson, Biochemistry 44 (31) (2005) 10570–10582.
[46] X. Lu, L. Li, R. Wu, X. Feng, Z. Li, H. Yang, C. Wang, H. Guo, A. Galkin, O.
Herzberg, P.S. Mariano, B.M. Martin, D. Dunaway-Mariano, Biochemistry 45 (4)
(2006) 1162–1172.
[47] Y. Luo, B. Knuckley, M. Bhatia, P.J. Pellechia, P.R. Thompson, J. Am. Chem. Soc.
128 (45) (2006) 14468–14469.
[48] E.M. Stone, T.H. Schaller, H. Bianchi, M.D. Person, W. Fast, Biochemistry 44 (42)
(2005) 13744–13752.
[49] V.M. D’Costa, K.M. McGrann, D.W. Hughes, G.D. Wright, Science 311 (5759)
(2006) 374–377.
[50] E.R. Vossenaar, W.J. Van Venrooij, Arthritis Res. Ther. 6 (3) (2004) 107–111.
[51] E.R. Vossenaar, A.J. Zendman, W.J. van Venrooij, G.J. Pruijn, Bioessays 25 (11)
(2003) 1106–1118.
[52] F. Liao, Z. Li, Y. Wang, B. Shi, Z. Gong, X. Cheng, Med. Hypotheses 72 (6) (2009)
732–735.