Báo cáo khoa học: The activity of Plasmodium falciparum arginase is mediated by a novel inter-monomer salt-bridge between Glu295–Arg404 doc

The activity of Plasmodium falciparum arginase
is mediated by a novel inter-monomer salt-bridge
between Glu295–Arg404
Gordon A. Wells
1
, Ingrid B. Mu
¨
ller
2
, Carsten Wrenger
2
and Abraham I. Louw
1
1 Department of Biochemistry, University of Pretoria, South Africa
2 Department of Biochemical Parasitology, Bernhard Nocht Institute for Tropical Medicine, Hamburg, Germany
The polyamines putrescine, spermine and spermidine
are near ubiquitous polycationic aliphatic amines
required for a number of essential cellular processes,
particularly in organisms undergoing rapid prolifera-
tion [1–3]. These processes involve the stabilization of
macromolecules [4–6] and progression through the cell
cycle [7]. Additionally, certain secondary metabolites,
such as the post-translationally modified amino acid,
hypusine [1,8], and the glutathione analogue, trypano-
thione, in Trypanosoma [9], require polyamines for
their biosynthesis. Polyamine biosynthesis has been
identified as a possible therapeutic target for various
parasitic diseases [10,11], cancers [12] and even HIV
via the requirement for hypusine [13]. Putrescine is
synthesized by the decarboxylation of ornithine (orni-
thine decarboxylase) and serves as substrate for the

addition of aminopropyl groups to form spermidine
and spermine. The aminopropyl groups are donated
Keywords
arginase; malaria; metal; modelling; trimer
Correspondence
A. I. Louw, Department of Biochemistry,
University of Pretoria, Lynwood Road,
Pretoria 0002, South Africa
Fax: +27 (0)12 362 5302
Tel: +27 (0)12 420 2480
E-mail:
(Received 27 January 2009, revised 26
March 2009, accepted 23 April 2009)
doi:10.1111/j.1742-4658.2009.07073.x
A recent study implicated a role for Plasmodium falciparum arginase in the
systemic depletion of arginine levels, which in turn has been associated with
human cerebral malaria pathogenesis. Arginase (EC 3.5.3.1) is a multimeric
metallo-protein that catalyses the hydrolysis of arginine to ornithine and
urea by means of a binuclear spin-coupled Mn
2+
cluster in the active site.
A previous report indicated that P. falciparum arginase has a strong depen-
dency between trimer formation, enzyme activity and metal co-ordination.
Mutations that abolished Mn
2+
binding also caused dissociation of the
trimer; conversely, mutations that abolished trimer formation resulted
in inactive monomers. By contrast, the monomers of mammalian (and
therefore host) arginase are also active. P. falciparum arginase thus appears
to be an obligate trimer and interfering with trimer formation may there-

fore serve as an alternative route to enzyme inhibition. In the present
study, the mechanism of the metal dependency was explored by means of
homology modelling and molecular dynamics. When the active site metals
are removed, loss of structural integrity is observed. This is reflected by a
larger equilibration rmsd for the protein when the active site metal is
removed and some loss of secondary structure. Furthermore, modelling
revealed the existence of a novel inter-monomer salt-bridge between
Glu295 and Arg404, which was shown to be associated with the metal
dependency. Mutational studies not only confirmed the importance of this
salt-bridge in trimer formation, but also provided evidence for the indepen-
dence of P. falciparum arginase activity on trimer formation.
Abbreviations
NP, constant number of atoms and constant pressure; NPT, constant number of atoms, constant pressure and temperature; PfArg,
Plasmodium falciparum arginase.
FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS 3517
from decarboxylated S-adenosylmethionine, the prod-
uct of S-adenosylmethionine decarboxylase [2,14].
Arginase (EC 3.5.3.1) catalyses the hydrolysis of
l-arginine to l-ornithine and urea. The arginase fold is
part of the ureohydrolase superfamily, which also
includes agmatinase [15,16], histone de-acetylase and
acetylpolyamine amidohydrolase [17]. Agmatine is
formed by decarboxylation of arginine (arginine decar-
boxylase) and is converted by agmatinase to putrescine
and urea. Arginase is thus part of one of two alterna-
tive biosynthetic routes to putrescine. Polyamine bio-
synthesis enzymes characterized in the malaria
parasite, Plasmodium falciparum, include the bifunc-
tional S-adenosylmethionine decarboxylase ⁄ ornithine
decarboxylase [18–22], spermidine synthase [23] and

arginase [24]. In P. falciparum, the agmatinase route
to putrescine has not been identified, thus making
arginase the sole biosynthetic route to putrescine in
the malaria parasite [24].
In mammals, two isoforms of arginase have been
identified: arginase I is cytosolic and largely hepatic
where it catalyses the final step of the urea cycle
[25,26]; arginase II is nonhepatic and occurs in the
mitochondrial matrix [27–29] and is involved in
homeostasis of ornithine for the further production of
proline and glutamate [30]. Both isoforms have been
implicated in regulating NO biosynthesis as a result of
competition for the common substrate arginine with
inducible NO synthase [31]. In bacteria, there is a
single arginase isoform, whereas more than one exists
in yeast [32]. The yeast isoforms have been linked
to glutamate accumulation during germination and
asexual spore development [33,34].
Arginase is a multimeric metallo-enzyme with a
binuclear spin-coupled Mn
2+
cluster in each active site
that is restricted to a single a ⁄ b monomer. The metal
cluster is located in a 15 A
˚
deep cleft with the Mn
2+
atoms 3.3 A
˚
apart and bridged by a solvent molecule

[35]. Structures from the bacteria Bacillus caldovelox
[36,37], human arginase I [38] and II [39], rat arginase
I [35,40] and Thermus thermophilus [Protein Data
Bank (PDB) codes: 2EF4, 2EF5 and 2EIV] have been
determined. In the mechanism proposed by Kanyo
et al. [35], the metal bridging solvent is an activated
hydroxyl, which attacks the guanidium carbon of
arginine, followed by collapse of the tetrahedral inter-
mediate to release ornithine and urea. However, ab
initio modelling of the active site suggests that the
bridging solvent may be a neutral water molecule
instead [41].
Eukaryotic arginases are trimeric, whereas bacterial
arginases are hexameric [42]. However, the trimeric
arginase from Saccharomyces cerevisiae forms a
regulatory complex with trimeric ornithine transcarba-
moylase, thus forming a hexameric complex [43]. In
mammals, monomers retain substantial activity if tri-
mer formation is disrupted [44,45]. In recent studies,
disrupting metal binding has no reported effect on the
quaternary structure [40,46,47]. However, previously, it
was reported that oligomerization of human arginase I
could be disturbed by chelating out Mn
2+
but sub-
stantially similar kinetics could be restored to nylon-
immobilized monomers after reincubation with Mn
2+
[48,49]. Removal of the active site metals in yeast not
only abolishes enzyme activity, but also affects the

maintenance of the quaternary structure and sensitivity
to temperature [43,50].
To date, the strongest metal dependency has been
reported for the arginase from P. falciparum, where
mutations that abolish metal binding or removal of
the metal ions cause dissociation of the trimer into
inactive monomers. Conversely, a mutation that abol-
ishes a conserved inter-monomer interaction located
away from the active site results in inactive monomers
[24]. These host–parasite differences may thus provide
a novel non-active site based strategy for inhibiting
P. falciparum arginase (PfArg). Essentially, disturbing
trimer formation may serve as a novel means of inhib-
iting PfArg. Thus, the mechanism of this structural
dependency was investigated by homology modelling
and molecular dynamics, aiming to establish an
in silico system for exploiting this dependency.
Results
Sequence alignment and homology modelling
Searching online Plasmodium genome resource, Plas-
moDB [51], revealed the arginase sequences for Plas-
modium vivax, Plasmodium yoelii, Plasmodium knowlesi
and Plasmodium berghei, in addition to the previously
characterized PfArg. From the automated alignments,
two parasite-specific inserts were revealed in the vari-
ous Plasmodium arginases (Fig. 1). In both reference
alignments, the positions of the inserts do not differ
markedly. In the final model of the study, insert 1 is
predicted to run from residues 77–151 (75 residues),
and insert 2 from residues 377–388 (12 residues). The

exact positions vary slightly depending on the align-
ment used for modelling. Insert 1 varies considerably
in sequence and length between different Plasmodium
species, ranging from approximately 100 residues
(P. vivax) to only 15 residues (P. berghei). It is pre-
dicted to lie between the second b-strand and second
a-helix of the model on the outer edge of the trimer.
By contrast, insert 2 is highly conserved in length and
Novel interaction in plasmodial arginase G. A. Wells et al.
3518 FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS
sequence in all Plasmodium species. Insert 2 is located
between the last b-strand and the last a-helix. The
sequence identity between the P. falciparum and tem-
plates was approximately 35%, 30% and 27%, respec-
tively, for the bacterial, rat and human arginases. Even
a slight difference in the alignment used for insert 2
was found to have a significant effect on its conforma-
tion. Modelling with insert 2 shifted one residue
upstream (fugue derived alignment; see Experimental
procedures) caused the insert to fold away from
the trimer interface, interacting with its respective
monomer (Fig. 2).
The model preserves standard active site residues
observed in other arginase structures. All Mn
2+
coor-
dinating residues (discussed below) previously identi-
fied are conserved in the model. The only substitutions
are in second shell ligands when compared with the
bacterial template, where Ser176 and Glu268 (B. cald-

ovelox) are replaced by Asp272 and Asp365 (P. falci-
parum), respectively. Residues implicated in substrate
binding are also highly conserved. There is only one
conservative substitution compared to the mammalian
templates, and none compared to the bacterial tem-
plate. Thr135 (rat) is replaced by Ser227 in the model.
In the model, Mg
2+
was modelled instead of Mn
2+
as a result of limitations of the forcefield, which
was not parameterized for Mn
2+
(see Experimental
procedures).
Visual inspection of the model suggested that a
novel inter-monomer salt-bridge forms between
Glu295
x
and Arg404
y
(where subscripts designate dif-
ferent monomers) (Fig. 3). In multiple sequence
alignments, Glu295 aligns with conserved acidic resi-
dues in the bacterial and mammalian templates.
P. falciparum Glu295 aligns with an Asp in mam-
mals (human arginase II: Asp223; rat arginase I:
Asp204), fungi and bacteria (B. caldovelox arginase:
Asp199). In the other Plasmodium species, Glu295
Fig. 1. Alignment used for modelling. P. falciparum (pfam), H. sapiens (human), R. norvegicus (rat) and B. caldovelox (bacc). The positions of

the Plasmodium-specific inserts are indicated. Identical residues are shaded dark grey, and similar residues indicated by lighter shades. Heli-
ces are indicated by rods, and b-strands by arrows.
Fig. 2. Effect of alignment on conformation of insert 2. When mov-
ing the insert one residue upstream, the insert folds away from the
trimer interface (yellow) compared to making contact (red). The
active site Mg
2+
atoms are indicated in green. Monomers are dis-
tinguished by different shades of blue. The image was generated
using
PYMOL.
G. A. Wells et al. Novel interaction in plasmodial arginase
FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS 3519
aligns either with Asp (P. yoelii and P. berghei)or
Glu (P. knowlesi and P. vivax). The only exception is
in plants, where Glu295 aligns with Ser instead
(Arabidopsis thaliana arginase I: Ser247). In the
model, Glu295 forms an interaction with the
adjacent monomer via partner residues that do not
align in sequence in the mammalian and bacterial
templates. In mammalian arginases, the Asp223 ⁄ 204
x
(rat arginase I ⁄ human arginase II) cognate forms an
inter-monomer salt-bridge with Arg308 ⁄ 327
y
. This
salt-bridge nucleates considerable inter-monomer
interactions, characterized by an S-shaped C-terminus
[35,39], which is absent in the Plasmodium sequences.
Arg308 from rat arginase I aligns with Ile in

Plasmodium (408: P. falciparum; 368: P. knowlesi;
436: P. vivax; 353: P. berghei; 376: P. yoelii). The
P. falciparum Arg404
y
salt-bridge partner to Glu295
x
aligns with small and ⁄ or hydrophilic residues in
other organisms (e.g. Ser, Thr, Cys, Ala and Glu).
In the bacterial structure, the Asp199
x
cognate forms
an inter-monomer bridge with Glu256
y
that is medi-
ated either by urea or by free arginine, depending
on the crystallization conditions [36].
To determine other possible interactions, the salt-
bridge analysis tool of vmd [52] was employed to
search for all possible salt-bridges in the protein, using
co-ordinates prior to sampling. All salt-bridges with a
hydrogen bond donor ⁄ acceptor distance less than
3.2 A
˚
were identified. Only one other interaction
between adjacent monomers was found between
Glu400
x
and Lys340
y
. However, this interaction was

not stable during molecular dynamics. This instability
was observed both with and without the active-site
metal, Mg
2+
. Thus, this interaction is likely to be only
of secondary importance in maintaining quaternary
structure, and attention was focused on the Glu295
x
–
Arg404
y
interaction instead.
Protein stability
Before proceeding with detailed analysis, it was neces-
sary to ensure that gross changes to the protein neces-
sitated by the modelling dificulties would not
compromise the interpretation of the results. The omis-
sion of parts of the protein is potentially problematic
in that it introduces an unnatural chain break and
therefore potential instability. The deletion of insert 1
creates a protein fragment on the outer edge of the tri-
mer complex that does not interact extensively with
any neighbouring monomers, and largely makes intra-
monomer contacts. This fragment was stable for at
least 50 ns of simulation and, apart from the loss of
some secondary structures in this region (described
below), remained in contact with the rest of the
protein.
The protein stability of PfArg was monitored by the
change in C

a
rmsd during equilibration and sampling
compared to the starting co-ordinates. In both cases,
with and without Mg
2+
, there is an increase in rmsd,
which typically equilibrates after approximately 20 ns
(Fig. 4). However, without Mg
2+
, the C
a
rmsd equili-
brates at approximately 1 A
˚
more than in the presence
of Mg
2+
, which was usually observed by 20 ns in both
the constant number of atoms and constant pressure
(NP) and constant number of atoms, constant pressure
and temperature (NPT) ensembles and persisted up to
50 ns in the NPT simulations.
The effect of removing the metal on conservation of
secondary structure during sampling was also moni-
tored. In general, a greater loss of secondary structural
integrity was observed for nonmetal systems in both
the NP and NPT ensembles. In the absence of Mg
2+
,
a complete loss of secondary structure is observed for

certain elements. Combining both NP and NPT
ensembles gives a total of six simulations of the mono-
mer, which can be used to observe any general loss of
Arg308/327
Asp204/223
Fig. 3. Salt-bridges in PfArg. Template residue numbers are shown
in italics (rat ⁄ human). The conserved interaction between Arg346
x
and Glu347
y
(P. falciparum) is indicated. Glu295
x
aligns with acidic
residues in the template structures (Asp204 ⁄ Asp223) but forms a
novel interaction with Arg404
y
compared to the template residues
that interact with Arg308 ⁄ Arg327 in the C-terminus. Monomers are
shown in green, light blue and mauve. Template backbones are
transparent, and template salt-bridge residues are depicted
in lighter shades of red (acidic) and blue (basic). The image was
generated using
VMD.
Novel interaction in plasmodial arginase G. A. Wells et al.
3520 FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS
secondary structure. However, these data are not suffi-
cient to determine possible co-operative effects between
monomers. The monomers ⁄ chains are arbitrarily desig-
nated A, B and C. In chain B of the NP simulation,
the first and second b-strands are both lost, whereas

the second half of the first a-helix is lost in chain C of
the NPT simulation. The first half of the third a-helix
is also lost in chains B and A of the NP and NPT sim-
ulations, respectively. All of these secondary structural
elements align with cognate elements in all of the tem-
plates. The sixth helix (3
10
) of the model is lost in
some chains of both the metal and metal-free simula-
tions. Whether this element is a helix is uncertain
because it only aligns with a 3
10
-helix in the bacterial
template. In both P. falciparum and the bacteria, the
N-terminal residue is a proline, which often forms the
N-terminal cap of both a-helices and 3
10
-helices, and is
also over-represented as the helix capping residue when
followed by a b-strand [53]. The absence of helical
structure in the mammalian templates indicates the
possibility of another conformation in the P. falcipa-
rum structure.
During sampling, insert 2 moves considerably and
does not retain its interaction as predicted by the origi-
nal homology model prior to molecular dynamics.
Furthermore, there are some noticeable differences
between the metal and nonmetal simulations. In both
the NP and NPT ensembles, insert 2 partially occupies
the interface between two adjacent monomers, which is

more pronounced, however, when Mg
2+
is included
(results not shown).
Similar effects on protein stability were observed for
simulations of five mutants (Glu295 Ala, Glu295 Arg,
Arg404 Ala, Glu295
x
Ala ⁄ Arg404
y
Ala, Glu347 Gln)
of the PfArg model. For all mutations, there is also a
greater increase in rmsd (described above) compared
to the wild-type model with Mg
2+
. The largest
increase is observed for Glu347 Gln, which is approxi-
mately double that of wild-type enzyme without
Mg
2+
. The mutations Glu295 Ala, Glu295 Arg,
Glu295
x
Ala ⁄ Arg404
y
Ala show a similar increase to
wild-type enzyme without Mg
2+
. The smallest effect is
observed for Arg404 Ala, which is similar to the wild-

type enzyme without Mg
2+
for part of the 50 ns run
(results not shown).
Stability of inter-monomer salt-bridges
In all arginases studied to date, there is a conserved
inter-monomer salt-bridge represented in P. falciparum
by Arg346
x
–Glu347
y
(Fig. 3). The cognate salt-bridge
in the templates used is between Arg255 ⁄ 274 ⁄ 249
x
–
Glu256 ⁄ 275 ⁄ 250
y
(rat, human and bacterial templates,
respectively). These residues align unambiguously and
the salt-bridge forms reliably during modelling.
Considering the established importance of this inter-
action, its integrity was monitored during modelling
and simulation.
In the sampling runs, the Arg346
x
–Glu347
y
interac-
tion was generally stable and intact for both the Mg
2+

and Mg
2+
-absent cases. One inter-monomer bridge
did break in the presence of Mg
2+
in the NP ensem-
ble. In the NPT ensembles, the interactions remain
intact with and without Mg
2+
but there is an increase
in the average standard deviation of the salt-bridge dis-
tance in the absence of Mg
2+
(Fig. 5). This suggests
that the Arg346
x
–Glu347
y
interaction is susceptible to
removal of Mg
2+
, even though the interaction
remained intact.
As described above, visual inspection of the homol-
ogy models suggested a further interaction between
Glu295
x
and Arg404
y
. Although not fully formed in

the homology models, the salt-bridge distance did
0 2 4 6 8 101214161820
Time (ns)
0
1
2
3
4
5
6
7
8
RMSD (Å)
C
α
deviation
(NP ensemble)
RMSD (Å)
01020304050
Time (ns)
0
1
2
3
4
5
6
7
C
α

deviation
(NPT ensemble)
Fig. 4. Effect of removing Mg
2+
on backbone C
a
rmsd. A running
average was calculated using a window of 500 frames (250 fs per
frame). With Mg
2+
, removed ( ), a greater increase is observed
than with Mg
2+
(+) included, implicating Mg
2+
in the structural
stabilization of the enzyme.
G. A. Wells et al. Novel interaction in plasmodial arginase
FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS 3521
adopt standard values (± 4 A
˚
) during minimization
and heating of the systems. The integrity of this inter-
action was found to be more susceptible to removal of
Mg
2+
than the Arg346
x
–Glu347
y

interaction. Between
the NP and NPT ensembles, there are six Glu295
x
–
Arg404
y
salt-bridges. In the Mg
2+
-free systems, the
salt-bridge was broken in half of these (Fig. 6). In the
NP ensemble, two interactions are broken between
chain A and B, and chain B and C by the end of
20 ns. The third interaction was transiently broken
(chain C and A). In the NPT ensemble, only one of
these interactions (chain A and B) was broken by the
end of 50 ns.
The alignment used to model insert 2 also affected
the Arg346
x
–Glu347
y
interaction. In models where
insert 2 was predicted to interact at the trimer inter-
face, the interaction was broken during in vacuo simu-
lations in charmm [54]. It was also observed during
simulation with charmm that models built without the
imposition of symmetry on the internal co-ordinates
tended to disturb the Arg346
x
–Glu347

y
interaction.
The stability of this interaction was improved by using
models built with symmetry imposed on internal
co-ordinates (i.e. with perfectly super-imposable
monomers) (results not shown).
During the simulation of the mutant enzymes, there
was no effect on the integrity of Arg346
x
–Glu347
y
interaction for mutations directed at Glu295
x
and ⁄ or
Arg404
y
. For the simulation of Glu347 Gln, however,
an interesting result was obtained: two out of three of
the Glu295
x
–Arg404
y
interactions were broken during
the sampling run. This indicates that Glu347 Gln may
also effect trimer destabilization through disturbing
Glu295
x
–Arg404
y
as well by the loss of the Arg346

x
–
Glu347
y
interaction. Conversely, the results do not
indicate that disturbing the Glu295
x
–Arg404
y
interac-
tion affects the Arg346
x
–Glu347
y
salt-bridge.
Coordination geometry of Mn
2+
/Mg
2+
Highly conserved residues are involved in a specific
co-ordination pattern by donation of free electron
pairs for the binuclear Mn
2+
cluster in existing crys-
tal structures. In rat I arginase, the more deeply bur-
ied ion (Mn
2+
A
) is co-ordinated by His101, Asp124,
Asp128, Asp232 and the bridging solvent in a square

pyramidal geometry. The respective residues in
P. falciparum are His193, Asp216, Asp220 and
Asp323. The second metal, Mn
2+
B
is co-ordinated
by His126, Asp124, Asp232, Asp234 and the bridg-
ing solvent in a distorted octahedral geometry in
rat I arginase (His218, Asp216, Asp323, Asp325 in
P. falciparum).
0 10 20 30 40 50
Time (ns)
0.1
0.2
Standard deviation (
Å
)
Distance (Å)
Arg346–Glu347 salt bridge
0
1
2
3
4
5
6
7
8
9
Fig. 5. Effect of removing Mg

2+
on backbone Arg346
x
–Glu347
y
salt-bridge. A running average was calculated using a window of
500 frames (250 fs per frame). Distances for all interactions from
the NP and NPT ensembles are plotted against the right y-axis. In
the NP ensemble, including Mg
2+
, one of the salt-bridges is broken
between chains A and B (d). All other salt-bridges remain intact for
both the NP and NPT ensembles with (black) and without (purple)
Mg
2+
. The average standard deviation for the NPT ensemble is
plotted against the left y-axis with (black) and without (red) Mg
2+
.
Solid lines indicate the average sum for all three chains, and the
mean over the entire run is indicated by the dashed lines. With
Mg
2+
removed, there is an increase in the standard deviation.
0 1020304050
Time (ns)
0
5
10
15

20
Distance (Å)
Glu295–Arg404 salt bridge
Fig. 6. Effect of removing Mg
2+
on backbone Glu295
x
–Arg404
y
salt-
bridge. A running average for the interaction between each chain-pair
combination was calculated using a sliding window of 500 frames
(250 fs per frame). The NP simulation was terminated after 20 ns. In
the absence of Mg
2+
(purple) in the NP ensemble, all three interac-
tions were broken (chains: d, AB; r, BC;
, CA), albeit that between
chains CA only transiently. In the NPT ensemble, only the bridge
between chains A and B (
) was broken in the absence of Mg
2+
.
The interactions remained intact with Mg
2+
present (black).
Novel interaction in plasmodial arginase G. A. Wells et al.
3522 FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS
During modelling, the conformations adopted by the
co-ordinating residues did not entirely conform to

known crystal structures from homologues. The most
notable difference is Asp323, which is expected to form
a monodentate bridging interaction between the two
ions. During the simulations, it formed a bidentate
bridge instead. All other expected co-ordinating atoms
were oriented close enough to interact with the ions.
The only other missing interaction was that of the
bridging OH
–
because no attempt was made to intro-
duce the bridging solvent molecule. The Mg
2+
–Mg
2+
distance was also approximately 0.6 A
˚
greater than the
known Mn
2+
–Mn
2+
distance. The Mg
2+
ions
remained in the active site during the simulations and
restricted the movement of the interacting ligands.
When Mg
2+
is removed, considerable movement is
observed in the co-ordinating residues in both the

NPT and NP simulation.
Site-directed mutagenesis of Glu295
x
and
Arg404
y
Modelling predicts that Glu295
x
–Arg404
y
is necessary
for trimer formation. The existence of a structural-
metal dependency between trimer formation and activ-
ity in P. falciparum arginase supports the involvement
of the Glu295
x
–Arg404
y
interaction. The effects of
mutating Glu295
x
and Arg404
y
were therefore deter-
mined in the recombinantly expressed enzyme and are
summarized in Table 1. PfArg was found to be more
susceptible to mutations introduced at Glu295
x
than at
Arg404

y
. Mutating Glu295 to Ala or Arg considerably
reduces enzyme activity (by 96% and 73%, respec-
tively) under standard assay conditions, which is also
the case for the double mutant Glu295 Ala ⁄ Arg404
Ala. (95%). However, single mutations of Glu295 to
Arg and Arg404 to Ala leads to altered K
m
values of
146 mm and 45 mm for l-Arg, respectively, which is
up to 11-fold higher compared to the wild-type argi-
nase. The catalytic activity (as k
cat
) of these mutants
were 7 and 11 s
)1
, respectively, and were thus 27%
and 46% of that for the wild-type enzyme. The result-
ing efficiencies expressed as k
cat
⁄ k
m
values are reduced
in both mutants to 1.6% and 46%, respectively, com-
pared to the wild-type. The elution profile of all
mutants analysed by gel filtration revealed monomeric
forms, except for Glu295 Ala, which is partially tri-
meric (Fig. 7). By contrast, trimer formation is more
susceptible to mutation of Arg404. The partial activity
of Arg404 Ala is the first clear evidence of active

monomers for PfArg.
Discussion
From the multiple sequence alignment used for model-
ling, two parasite-specific inserts were identified in the
P. falciparum sequence. Proteins from Plasmodium fre-
quently have large inserts relative to sequences from
homologues in other organisms [55]. These inserts are
often characterized by low complexity [56,57] and ⁄ or
have a strong amino acid bias towards small and
hydrophilic residues. Apart from possible global func-
tions [55–57], it has been demonstrated that inserts
may have local functions relative to their enzymes
[22,58,59]. Plasmodium-specific inserts can be difficult
to delineate in sequences of low homology. Thus,
Table 1. Comparison of kinetic parameters for wild-type (WT) and
mutant arginases. The results are derived from at least three inde-
pendent assays with standard deviations. ND, not detectable.
V
max
(lmolÆmin
)1
Æmg
)1
)
K
m
(mM)
k
cat
a,b

(s
)1
)
k
cat
⁄ K
m
a,b
(mM
)1
Æs
)1
)
WT 31
c
13
c
24.8
(100%)
1.9
(100%)
Glu295 Ala 1.3 ± 0.3
d
ND 1.0 ± 0.2
d
(4%)
ND
Glu295 Arg 8.4 ± 0.9 146 ± 6 6.7 ± 0.7
(27%)
0.03

(1.6%)
Arg404 Ala 14.3 ± 0.9 45 ± 3 11.4 ± 0.7
(46%)
0.25
(13%)
Glu295 Ala ⁄
Arg404 Ala
1.6 ± 0.1
d
ND 1.3 ± 0.1
(5%)
ND
a
Percentage of WT value is shown in parentheses.
b
Calculated
from 48 kDa per monomer.
c
Values, without standard deviation,
are taken from Mu
¨
ller et al. [24].
d
£ 5% of WT activity in standard
assay.
Fig. 7. Effect of Glu295
x
–Arg404
y
salt-bridge mutations on trimer

formation. Recombinant proteins were separated on a Superdex
S-200 gel sizing column (1 · 30 cm) using a buffer containing
50 m
M Tris–HCl, pH 8, 1 mM dithiothreitol and 1 mM MnCl
2
.
Aliquots of 100 lL of the elution fractions (0.5 mL) were analysed
by western dot-blotting using monoclonal anti-Strep-tag serum
(Institut fu
¨
r Bioanalytik) at a dilution of 1 : 5000. The corresponding
molecular masses are given above the dots.
G. A. Wells et al. Novel interaction in plasmodial arginase
FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS 3523
where possible, other Plasmodium sequences were
included to assist with the insert delineation. Because
of its length, most of insert 1 was left unmodelled (resi-
dues 81–147 removed; Fig. 1). Insert 2 is considerably
shorter and more conserved and was therefore retained
for ab initio modelling. The choice of alignment was
found to have a considerable effect on the conforma-
tion of insert 2. Because a small change in alignment
had a substantial effect on insert 2, it is important to
justify the choice of reference alignment used. In the
fugue [60] derived alignment (see Experimental proce-
dures), insert 2 was predicted to fold away from the
trimer interface, compared to the clustalw derived
alignment. The fugue alignment was favoured,
however, because the fugue software makes use of
environment-specific substitution tables and structure-

dependent gap penalties, and is thus generally expected
to give a more accurate starting alignment for model-
ling purposes. The function of the inserts in PfArg has
yet to be established.
Comparing the active site of the model with the
templates revealed only a small number of substitu-
tions. The high conservation of the active site suggests
that inhibitors specific to the P. falciparum active site
will be difficult to find. Thus, an alternative means of
inhibition may be necessary if PfArg is to be of poten-
tial therapeutic value. Therefore, attention was direc-
ted at the inter-monomer interactions. A novel inter-
monomer salt-bridge forms between Glu295
x
and
Arg404
y
. Although Glu295 aligns with conserved
acidic residues in the templates, its interacting partner
does not align. In mammalian structures, the acidic
equivalent nucleates considerable inter-monomer inter-
actions by means of an S-shaped C-terminus by form-
ing a salt-bridge with Arg308. The importance of the
S-shaped tail is still in doubt because products trun-
cated after Arg308 can still form active trimers [61]. In
bacterial structures, an interaction is formed with
another acidic residue (Glu256) that is mediated by
either urea or free arginine. Finally, in the P. falcipa-
rum model, Glu295 is predicted to interact with
Arg404, which does not align with mammalian Arg308

or bacterial Glu256. Thus, there appears to have been
evolutionary pressure to establish a strong inter-mono-
mer interaction in this region of the monomer-mono-
mer interface. The differences between the
P. falciparum model and templates suggest this salt-
bridge as a possibly unique interaction and was
therefore subjected to scrutiny using molecular dynam-
ics and site-directed mutagenesis.
The deletion of insert 1 for modelling did not
adversely affect the stability of the model. Although
potential problems with respect to introducing a chain
break could have been avoided by ligating the ends of
the gap, this would also be unnatural. Because the
fragment was apparently stable and closing the gap
unligated is less parsimonious, the break was left in.
The equilibration of C
a
rmsd during molecular
dynamics at a larger distance for the Mg
2+
-free
systems suggests that removing the active site metals
has a detrimental effect on protein stability. It was
previously reported that removing Mn
2+
, either by
dialysis and chelation with EDTA, or by mutagenesis
of Mn
2+
co-ordinating residues in the active site, of

PfArg not only abolished enzyme activity, but also
promoted dissociation of the trimer, which could be
reversed by addition of Mn
2+
[24]. The general loss of
secondary structure further mirrors the increase in
rmsd upon removing the metal and confirms the
necessity of the active site metals for protein stability.
Removing the active site metals also affected the
conformation of insert 2 during molecular dynamics,
which generally remained more solvent exposed and
made less inter-monomer contacts. This suggests that
insert 2 may also be involved in maintaining the trimer
and thus part of the structural metal dependency.
The temperature of the NP ensemble was allowed
to increase (310 to 332 °K) by not coupling it to a
temperature bath. Although it is usual to apply some
means of keeping temperature constant (isothermal
ensemble), sampling at higher temperatures allows
the system to overcome energy barriers faster. In the
present study, the increase in temperature accelerates
the effects of removing Mg
2+
. In the NPT ensemble,
only one Glu295
x
–Arg404
y
interaction is broken after
20 ns, whereas, in the NP ensemble with increasing

temperature, all three have been broken before 20 ns.
The effect of the increasing temperature is also
reflected in the rmsd, which is more pronounced and
more rapid in the NP ensemble. The increasing
temperature may be detrimental, however, as
reflected by breaking an Arg346
x
–Glu347
y
interaction
in the NP ensemble with Mg
2+
. For this reason,
subsequent simulations were carried out in the NPT
ensemble.
Because most simulations based on classical
mechanics do not model metal co-ordination, the
conformations adopted by the co-ordinating residues
did not entirely conform to known crystal structures
from homologues. The larger distance between the
Mg
2+
atoms compared to Mn
2+
in known structures
is partly a result of the inability of the software to
recognize co-ordination chemistry natively as well as
the larger van der Waals radius of Mg
2+
compared to

Mn
2+
. Because of the stability of the Mg
2+
cluster, it
was considered unnecessary to introduce artificial
Novel interaction in plasmodial arginase G. A. Wells et al.
3524 FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS
restraints to replicate metal-co-ordination. Because the
presence of Mg
2+
was able to stabilize the co-ordina-
ting residues by electrostatic interactions alone, this
approach appears to be viable for investigating the
structural metal dependency. These results also suggest
that structural metal dependency involves free move-
ment of the metal co-ordinating residues.
The existence of the inter-monomer salt-bridge
between Glu295
x
and Arg404
y
was corroborated by
site-directed mutagenesis of the recombinant enzyme.
All mutants tested promoted trimer dissociation, with
incomplete dissociation for Glu295 Ala but contrasts
with Glu295 Arg, which led to complete dissociation.
Mutating Glu295 to Arg is expected to be more drastic
compared to Ala because this would introduce a
positive charge and thus an electrostatic repulsion in

the vicinity of the Glu295
x
–Arg404
y
interaction. Inter-
estingly, this mutation leads to active but less efficient
monomers with a 11-fold increased K
m
value of
146 mm for l-arginine, indicating altered substrate
binding. By contrast, mutating Glu295 to Ala reduced
the activity to 4% of the wild-type enzyme but with its
trimeric conformation partially retained. The K
m
value
for the Glu295 Ala mutant was not measurable
because it was not saturated up to 200 mm arginine.
Mutating Arg404 to Ala abolished trimer formation.
However, this mutant enzyme shows 46% activity (as
k
cat
) and 13% efficiency (as k
cat
⁄ K
m
) and its K
m
value
is approximately three-fold increased compared to the
wild-type enzyme. This result is similar to the previ-

ously reported behaviour of the rat liver arginase
Arg308 mutants, which, as monomers, still had a resid-
ual activity of 41% and an efficiency in the range
13–17% [44]. Size-exclusion chromatography therefore
suggests that certain mutations abolish trimerization,
although the enzymatic data suggests that trimeriza-
tion is not absolutely necessary for activity. However,
the possibility that a weakened trimer can form under
enzyme assay conditions cannot be excluded. Such a
possibility is suggested by rat arginase, where the
Arg308 Lys mutant is apparently active as a monomer,
but nonetheless crystallizes as a trimer [44]. Although
it has been demonstrated that disturbing the oligomer
via the conserved interaction between Arg346
x
and
Glu347
y
largely inactivates the enzyme, it still has 10%
residual activity [24]. The results of the Arg404 Ala
mutation indicates that it is possible to produce active
monomers and, furthermore, that certain mutations
can partially compensate for induced structural insta-
bility of monomerization by long range allosteric
effects. Although there is a dependency between trimer
formation and enzyme activity, these results indicate
that it is not complete. This incompleteness was sug-
gested by previous results where mutating His193 in
the active site also results in an inactive trimer [24] as
was also found for the Glu295 Ala mutation in the

present study. Mutations that disturb the Arg346
x
–
Glu347
y
and Glu295
x
–Arg404
y
interactions both result
in decreased activity. Furthermore, during simulations
of the Glu347 Gln mutant, the Glu295
x
–Arg404
y
inter-
action is also disturbed. These observations suggest
that disruption of both interactions may provide a
novel means of inhibiting PfArg. These results suggest
that formation of the Glu295
x
–Arg404
y
salt-bridge is
necessary for trimer formation, and that the hypothesis
that the enzyme can be inhibited via disturbing the
trimer warrants further investigation.
It is expected that disturbing the interactions
involved in trimer formation mediate their effects via
the co-ordination of Mn

2+
in the active site, which is
required for the arginase chemistry. This is reflected by
the increased equilibrium rmsd during molecular
dynamics, which should ultimately translate into lost
co-ordination of Mn
2+
in the active site. The loss of
Mn
2+
under such conditions, however, has yet to be
observed directly. Nonetheless, the decreased activity
of monomeric mutants and the increased equilibrium
rmsd of modelled mutants suggests that this may be
the case.
It has been demonstrated that rat arginase I loses
some activity (33–41% of k
cat
) when the trimerization
is disturbed by mutagenesis [44]. This has not been
observed for human arginase I, where fully functional
monomers have been obtained [45,48,62]. Despite the
high sequence similarity (87%) between arginase I
from rat and human [63], they differ in their kinetic
properties. Human arginase I has a substantially lower
K
m
for arginine compared to rat arginase I. Further-
more, the K
d

values for the inhibitors S-(2-borono-
ethyl)-l-cysteine and 2-amino-6-boronohexanoic acid
are one order of magnitude less than for the rat coun-
ter part [38,46,64]. This suggests that it may be possi-
ble to inhibit PfArg via disturbing oligomerization
without affecting the human counterpart.
Arginine levels in in vitro cultures of P. falciparum
are depleted by PfArg, although the relevance of
arginase as a malaria drug target remains to be dem-
onstrated [65]. Hypoargininaemia has been linked to
the progression of severe malaria and may be related
to the requirement of arginine for NO biosynthesis
[66]. It has been speculated that low host NO bene-
fits the parasite by causing increased expression of
host intracellular adhesion molecule-1, which is used
by parasatized red blood cells to adhere to the vas-
cular endothelium and thus avoid spleen clearance.
Arginase knockouts of the rodent malaria parasite,
G. A. Wells et al. Novel interaction in plasmodial arginase
FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS 3525
P. berghei (ANKA strain), are viable and show simi-
lar growth behaviour ex vivo and in infected mice
[65], although this has yet to be established for the
human parasite.
Experimental procedures
Sequence alignments
Reference multiple alignments were generated using clu-
stalw 1.82 and the fugue server. The clustalw align-
ment included eukaryotic arginases types I and II, and
bacterial arginases. The sequences used for the clustalw

alignment (Entrez accession number are given in brackets
for non-Plasmodium species, PlasmoDB reference numbers
are used for Plasmodium sequences) were: A. thaliana
(P46637, Q9ZPF5), Schizosaccharomyces pombe (P37818,
Q10066), Xenopus laevis (Q91553, Q91554, Q91555,
P30759), Homo sapiens (P78540, P05089), Mus musculus
(O08691, Q61176), Rattus norvegicus (O08701, P07824),
Agrobacterium tumefaciens (P14012), B. caldovelox
(P53608), Bacillus subtilis (P39138), Brucella melitensis
(Q59174), Coccidioides immitis (P40906), Emericella nidu-
lans (Q12611), Neurospora crassa (P33280), Rana catesbei-
ana (P49900), Glycine max (O49046), Staphylococcus aureus
(P60086), S. cerevisiae (P00812), P. knowlesi (PKH_
070380), P. vivax (Pv098770), P. falciparum (PFI0320w),
P. yoelii (PY03443) and P. berghei (PB000787.03.0). Seq-
uences for P. knowlesi, P. vivax, P. yoelii and P. berghei
were obtained using blast (available at: http://plasmodb.
org) [51] with the P. falciparum sequence as query.
Although the reference alignments were often highly redun-
dant, all sequences were retained to offset the bias of
including five Plasmodium sequences.
Homology modelling
modeller 8v0 [67,68] was used to build the homology
models. Trimeric models were constructed on the rat argi-
nase I (PDB code: 1RLA[abc]), human arginase II (PDB
code: 1PQ3[abc]) and B. caldovelox (PDB code: 1CEV[abc])
templates. Superimposable monomers were constructed by
imposing symmetry restraints on the internal coordinates of
all atoms during the model building process. The effect of
various sequence alignments was determined by generating

multiple models with different random number seeds and
monitoring the effect on the number of residues in dis-
allowed regions of the Ramachandran plot and on the
overall G-factor score from procheck [69]. Problem areas
were identified as residues that frequently fell in disallowed
regions. Models that minimized residues with poor phi ⁄ psi
values and maximized the G-factor were used for molecular
dynamics. In the final run, one model was chosen from a
total of 33 for molecular dynamics.
Molecular dynamics
Hydrogen atoms were added automatically using charmm
32b1 [54,70]. All residue positions are given relative to their
own sequence. The histidine protonation scheme adopted
was based on the requirements for co-ordination of metal
atoms in the active site in known structures. Thus, His193
and His218 (which align with His101 and His126 in rat
arginase I, respectively) were protonated on N
e
, and His233
was protonated on both nitrogen atoms. His233 aligns with
His141 in human arginase I. The double protonation of
His233 was based on the high resolution (1.29 A
˚
) crystal
structure of human arginase I [38], for which hydrogen
positions were also determined, as well as previous specula-
tion concerning activity [35]. All remaining histidines were
protonated on the N
d
atom. Glutamate, aspartic acid and

lysine residues were charged. Because the current common
protein forcefields (charmm, amber, gromos) were not
parameterized for Mn
2+
, the Mg
2+
ion was used instead.
It was thus assumed that any effects of the metal on trimer
formation were largely electrostatic in nature.
Although charmm was initially used for molecular dynam-
ics, it was found to perform too slowly on the available hard-
ware (Pentium IV Beowulf cluster with GigaBit Ethernet).
namd 2.6 [71] was used instead, which provided better scaling.
pymol [72] was used for visualization, as well as vmd, which
was also used for analysis of molecular dynamics trajectories
[52]. stride software, as provided with vmd, was used to
assign secondary structure. The multiseq plugin [73] was used
to visualize secondary structure alignments. Salt-bridges
between arginine and glutamate residues were measured
between the C
f
and C
d
atoms, respectively. A distance between
these atoms of 4 A
˚
corresponds to the typical distance of
2.8 A
˚
between the hydrogen bond donor (PDB naming: NH

x
)
and acceptor atoms (PDB naming: OE
x
) respectively.
The system was explicitly solvated (transferable inter-
molecular potential water model) using the solvate plugin
of vmd. The protein was padded with solvent for 12 A
˚
in
the x- and y-axis (xy being coplanar with the trimer) and
10 A
˚
in the z-axis. NaCl counter ions were added with the
autoionize plugin from vmd to an ionic strength of 50 mm
with charge balancing to create a net system charge of zero.
The system with metals (Mg
2+
) included 14 Na
+
ions and
11 Cl
–
ions, whereas the system without metal finally
contained 20 Na
+
ions and 5 Cl
)
ions.
Nonbonded interactions were shifted to zero from 10 A

˚
to a cut-off of 12 A
˚
. All nonbonded interactions connected
by more than four covalent bonds were included. Prior to
solvation, the trimer complex was minimized in namd
in vacuo for 450 steps. All steps including the complete
system with solvent and ions were simulated with periodic
boundary conditions using particle mesh Ewald Sums for
the electrostatic calculations. A typical cell size was approx-
imately 118x · 113y · 79z A
˚
. A particle mesh Ewald
sub-A
˚
ngstrom grid size of 120 · 120 · 90 was used.
Novel interaction in plasmodial arginase G. A. Wells et al.
3526 FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS
Multiple protocols were used to verify the general robust-
ness of the system. Two different heating protocols were
used. In the first protocol, the solvent of the entire system
was minimized for 2000 steps, followed by the solvent and
nonbackbone atoms (backbone = C,N,O,C
a
) for another
2000 steps. The entire system was then heated over 20 000
steps (20 ps) from 60 °Kat10°K increments every 500
steps using velocity reassignment. All atoms were then min-
imized (2000 steps), followed by another heating step as
described, but for 200 ps. In the second protocol, all steps

were the same except that, during the first heating, only
solvent atoms were allowed to move. During heating, time-
steps were 1 fs and a Langevin piston (piston period 100 fs,
piston decay 50 fs) was used to maintain pressure at one
atmosphere.
Two different protocols were used during the production
runs. In the first protocol, a Langevin piston was used to
maintain pressure at one atmosphere (NP ensemble) with a
piston period of 200 fs and a piston decay of 100 fs,
whereas temperature was left to fluctuate. In the second
protocol, Langevin dynamics was used to also maintain
temperature at 310 °K (NPT ensemble) with a damping
constant of 5 ps
)1
. The first protocol was used in conjunc-
tion with the first heating protocol and the second sampling
protocol was used with the second heating protocol
described above. Total run lengths for this stage were in
the range 20–50 ns. In accordance with the results of the
initial simulations and site-directed mutagenesis, various
mutants of the model were also simulated. Mutations were
introduced using scwrl 3 [74] into the homology model
followed by addition of hydrogens and molecular dynamics
according to the second sampling protocol described above
with Mg
2+
included.
A number of Linux clusters were used for molecular
dynamics simulations. These clusters include a 64 processor
Gentoo Linux cluster of Pentium IV processors (University

of Pretoria), Clusters of Intel Xeon
Ò
, AMD Opteron
Ò
or
Intel Itanium2
Ò
processors (Bio-Medical Informatics Centre,
Meraka Institute, Council for Scientific and Industrial
Research) running Scientific Linux. In both cases, the inter-
connect comprises GigaBit Ethernet. During the set-up
phases of a new national supercomputer at The Centre for
High Performance Computing (Cape Town, South Africa),
temporary access was granted to the iQudu cluster. The
iQudu hardware comprises Multicore AMD Opteron
Ò
pro-
cessors with InfiniBand interconnect.
Activity and oligomeric status of mutant Glu295
x
and Arg404
y
arginase
Simulations with the arginase model suggested further
experiments to be performed on the recombinant enzyme.
The mutations Glu295 Ala, Glu295 Arg, Arg404 Ala, as
well as the double mutation Glu259
x
Ala ⁄ Arg404
y

Ala,
were introduced into the recombinant PfArg by
site-directed mutagenesis; activity and oligomeric status
were detected as described previously [24]. Briefly, PfArg
was cloned into the pASK-IBA3 expression vector with a
C-terminal Strep-tag (Institut fu
¨
r Bioanalytik, Go
¨
ttingen,
Germany) for affinity purification. The construct was
transformed and expressed in Escherichia coli BL21-Codon-
Plus
Ô
(DE3)-RIL (Stratagene, Amsterdam, The Nether-
lands). A single colony was picked and grown overnight in
LB medium. The bacterial culture was diluted 1 : 50 and
grown at 37 °C until A
600
of 0.5 was reached. The expres-
sion was initiated with 200 ngÆmL
)1
of anhydrotetracycline
and the cells were grown for 4 h at 37 °C before being
harvested. The cell pellet was resuspended in 100 mm
Tris–HCl, 150 mm NaCl, pH 8.0, containing 0.1 mm
phenylmethanesulfonyl fluoride, sonicated, and centrifuged
at 100 000 g for 1 h at 48 °C. Strep-tag fusion protein was
purified according to the manufacturer’s recommendations
(Institut fu

¨
r Bioanalytik).
Site-directed mutagenesis was carried out according to
the manufacturer’s recommendations (QuikChange proto-
col; Stratagene, Amsterndam, The Netherlands). All muta-
tions were verified by nucleotide sequencing using the
Sanger dideoxy chain termination reaction for double-
stranded DNA [75]. The purification and expression of the
protein variants were carried out as described above.
The arginase activity was assayed by measuring the for-
mation of urea in a colorimetric method with a-isonitroso
propiophenone at a wavelength of 540 nm as described pre-
viously [76]. The standard assay was carried out in 50 mm
Tris–HCl, pH 8.0, 1 mm dithiothreitol, 1 mm MnCl
2
and
30 mm arginine in a total volume of 750 mL. l-arginine lev-
els up to 200 mm were used to determine the K
m
values. The
molecular size and oligomeric state of PfArg were assessed
by subjecting the affinity-purified protein to FPLC on a cali-
brated Superdex S-200 column (GE Healthcare, Munich,
Germany) equilibrated with 50 mm Tris–HCl, pH 8.0,
150 mm NaCl, 1 mm dithiothreitol, 1 mm MnCl
2
. PfArg
was detected in aliquots of each fraction using Western dot-
blotting, and a monoclonal antibody against Strep-tag
(Institut fu

¨
r Bioanalytik) at a dilution of 1 : 5000. Anti-
mouse-horseradish peroxidase-coupled secondary serum
(Invitrogen, Karlsruhe, Germany) was used for detection at
a dilution of 1 : 5000.
Acknowledgements
Temporary access was granted to the iQudu cluster at
the Centre for High Performance Computing, Meraka
Institute, Council for Scientific and Industrial
Research, Cape Town, South Africa. Access was also
granted to the computer clusters at the BioMedical
Informatics Centre, Meraka Institute, Council for
Scientific and Industrial Research, Pretoria, South
Africa. This work was funded by the National
G. A. Wells et al. Novel interaction in plasmodial arginase
FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS 3527
Research Foundation of South Africa (PhD Bursary
GUN: 2066842), as well as in part by the international
office of the Federal Ministry of Education and
Research, Germany.
References
1 Tabor CW & Tabor H (1984) Polyamines. Annu Rev
Biochem 53, 749–790.
2 Tabor CW & Tabor H (1985) Polyamines in micro-
organisms. Microbiol Rev 49, 81–99.
3 Cohen SS (1998) A Guide to the Polyamines. Oxford
University Press, New York, NY.
4 Childs AC, Mehta DJ & Gerner EW (2003) Polyamine-
dependent gene expression. Cell Mol Life Sci 60, 1394–
1406.

5 Wallace HM, Fraser AV & Hughes A (2003) A perspec-
tive of polyamine metabolism. Biochem J 376, 1–14.
6 Janne J, Alhonen L, Pietila M & Keinanen TA (2004)
Genetic approaches to the cellular functions of poly-
amines in mammals. Eur J Biochem 271, 877–894.
7 Ackermann JM, Pegg AE & McCloskey DE (2003)
Drugs affecting the cell cycle via actions on the
polyamine metabolic pathway. Prog Cell Cycle Res 5,
461–468.
8 Park MH, Wolff EC & Folk JE (1993) Is hypusine
essential for eukaryotic cell proliferation? Trends Bio-
chem Sci 18, 475–479.
9Mu
¨
ller S, Liebau E, Walter RD & Krauth-Siegel RL
(2003) Thiol-based redox metabolism of protozoan par-
asites. Trends Parasitol 19, 320–328.
10 Mu
¨
ller S, Coombs GH & Walter RD (2001) Targeting
polyamines of parasitic protozoa in chemotherapy.
Trends Parasitol 17, 242–249.
11 Heby O, Persson L & Rentala M (2007) Targeting the
polyamine biosynthetic enzymes: a promising approach
to therapy of African sleeping sickness, Chagas’ disease,
and leishmaniasis. Amino Acids 33, 359–366.
12 Wallace HM (2007) Targeting polyamine metabolism: a
viable therapeutic ⁄ preventative solution for cancer?
Expert Opin Pharmacother 8, 2109–2116.
13 Schafer B, Hauber I, Bunk A, Heukeshoven J, Dusedau

A, Bevec D & Hauber J (2006) Inhibition of multidrug-
resistant HIV-1 by interference with cellular S-adenosyl-
methionine decarboxylase activity. J Infect Dis 194,
740–750.
14 Tabor CW & Tabor H (1984) Methionine adenosyl-
transferase (S-adenosylmethionine synthetase) and
S-adenosylmethionine decarboxylase. Adv Enzymol
Relat Areas Mol Biol 56, 251–282.
15 Ouzounis CA & Kyrpides NC (1994) On the evolution of
arginases and related enzymes. J Mol Evol 39, 101–104.
16 Ahn HJ, Kim KH, Lee J, Ha JY, Lee HH, Kim D,
Yoon HJ, Kwon AR & Suh SW (2004) Crystal struc-
ture of agmatinase reveals structural conservation and
inhibition mechanism of the ureohydrolase superfamily.
J Biol Chem 279, 50505–50513.
17 Dowling DP, Costanzo LD, Gennadios HA & Chris-
tianson DW (2008) Evolution of the arginase fold and
functional diversity. Cell Mol Life Sci 65, 2039–2055.
18 Mu
¨
ller S, Da’dara A, Luersen K, Wrenger C, Das GR,
Madhubala R & Walter RD (2000) In the human
malaria parasite Plasmodium falciparum
, polyamines are
synthesized by a bifunctional ornithine decarboxylase,
S-adenosylmethionine decarboxylase. J Biol Chem 275,
8097–8102.
19 Krause T, Luersen K, Wrenger C, Gilberger TW, Mu
¨
ller

S & Walter RD (2000) The ornithine decarboxylase
domain of the bifunctional ornithine decarboxylase ⁄
S-adenosylmethionine decarboxylase of Plasmodium
falciparum: recombinant expression and catalytic
properties of two different constructs. Biochem J 352
Pt 2, 287–292.
20 Wrenger C, Luersen K, Krause T, Mu
¨
ller S & Walter
RD (2001) The Plasmodium falciparum bifunctional
ornithine decarboxylase, S-adenosyl-L-methionine
decarboxylase, enables a well balanced polyamine syn-
thesis without domain-domain interaction. J Biol Chem
276, 29651–29656.
21 Birkholtz L, Joubert F, Neitz AW & Louw AI (2003)
Comparative properties of a three-dimensional model of
Plasmodium falciparum ornithine decarboxylase.
Proteins 50, 464–473.
22 Birkholtz LM, Wrenger C, Joubert F, Wells GA,
Walter RD & Louw AI (2004) Parasite-specific inserts
in the bifunctional S-adenosylmethionine decarboxyl-
ase ⁄ ornithine decarboxylase of Plasmodium falciparum
modulate catalytic activities and domain interactions.
Biochem J 377 Pt 2, 439–448.
23 Haider N, Eschbach ML, Dias SS, Gilberger TW, Walter
RD & Luersen K (2005) The spermidine synthase of the
malaria parasite Plasmodium falciparum: molecular and
biochemical characterisation of the polyamine synthesis
enzyme. Mol Biochem Parasitol 142, 224–236.
24 Mu

¨
ller IB, Walter RD & Wrenger C (2005) Structural
metal dependency of the arginase from the human
malaria parasite Plasmodium falciparum. Biol Chem 386,
117–126.
25 Spector EB, Rice SC, Moedjono S, Bernard B &
Cederbaum SD (1982) Biochemical properties of arginase
in human adult and fetal tissues. Biochem Med 28, 165–
175.
26 Spector EB, Rice SC, Kern RM, Hendrickson R &
Cederbaum SD (1985) Comparison of arginase activity
in red blood cells of lower mammals, primates, and
man: evolution to high activity in primates. Am J Hum
Genet 37, 1138–1145.
27 Grody WW, Dizikes GJ & Cederbaum SD (1987)
Human arginase isozymes. Isozymes Curr Top Biol Med
Res 13, 181–214.
Novel interaction in plasmodial arginase G. A. Wells et al.
3528 FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS
28 Gotoh T, Sonoki T, Nagasaki A, Terada K, Takiguchi
M & Mori M (1996) Molecular cloning of cDNA for
nonhepatic mitochondrial arginase (arginase II) and
comparison of its induction with nitric oxide synthase
in a murine macrophage-like cell line. FEBS Lett 395,
119–122.
29 Morris SM Jr, Bhamidipati D & Kepka-Lenhart D
(1997) Human type II arginase: sequence analysis and
tissue-specific expression. Gene 193, 157–161.
30 Morris SM Jr (2002) Regulation of enzymes of the urea
cycle and arginine metabolism. Annu Rev Nutr 22,

87–105.
31 Li H, Meininger CJ, Hawker JR Jr, Haynes TE,
Kepka-Lenhart D, Mistry SK, Morris SM Jr & Wu G
(2001) Regulatory role of arginase I and II in nitric
oxide, polyamine, and proline syntheses in endothelial
cells. Am J Physiol Endocrinol Metab 280, E75–E82.
32 Borkovich KA & Weiss RL (1987) Purification and
characterization of arginase from Neurospora crassa.
J Biol Chem 262, 7081–7086.
33 Marathe S, Yu YG, Turner GE, Palmier C & Weiss
RL (1998) Multiple forms of arginase are differentially
expressed from a single locus in Neurospora crassa.
J Biol Chem 273, 29776–29785.
34 Turner GE & Weiss RL (2006) Developmental expres-
sion of two forms of arginase in Neurospora crassa. Bio-
chim Biophys Acta 1760, 848–857.
35 Kanyo ZF, Scolnick LR, Ash DE & Christianson DW
(1996) Structure of a unique binuclear manganese clus-
ter in arginase. Nature 383, 554–557.
36 Bewley MC, Jeffrey PD, Patchett ML, Kanyo ZF &
Baker EN (1999) Crystal structures of Bacillus caldovel-
ox arginase in complex with substrate and inhibitors
reveal new insights into activation, inhibition and catal-
ysis in the arginase superfamily. Structure. 7, 435–448.
37 Bewley MC, Lott JS, Baker EN & Patchett ML (1996)
The cloning, expression and crystallisation of a thermo-
stable arginase. FEBS Lett 386, 215–218.
38 Di Costanzo L, Sabio G, Mora A, Rodriguez PC,
Ochoa AC, Centeno F & Christianson DW (2005) Crys-
tal structure of human arginase I at 1.29-A resolution

and exploration of inhibition in the immune response.
Proc Natl Acad Sci USA 102, 13058–13063.
39 Cama E, Colleluori DM, Emig FA, Shin H, Kim SW,
Kim NN, Traish AM, Ash DE & Christianson DW
(2003) Human arginase II: crystal structure and physio-
logical role in male and female sexual arousal. Biochem-
istry 42, 8445–8451.
40 Scolnick LR, Kanyo ZF, Cavalli RC, Ash DE & Chris-
tianson DW (1997) Altering the binuclear manganese
cluster of arginase diminishes thermostability and cata-
lytic function. Biochemistry 36, 10558–10565.
41 Ivanov I & Klein ML (2004) First principles computa-
tional study of the active site of arginase. Proteins 54,
1–7.
42 Jenkinson CP, Grody WW & Cederbaum SD (1996)
Comparative properties of arginases. Comp Biochem
Physiol B Biochem Mol Biol 114, 107–132.
43 Green SM, Eisenstein E, McPhie P & Hensley P (1990)
The purification and characterization of arginase
from Saccharomyces cerevisiae. J Biol Chem
265, 1601–
1607.
44 Lavulo LT, Sossong TM Jr, Brigham-Burke MR, Doyle
ML, Cox JD, Christianson DW & Ash DE (2001) Sub-
unit-subunit interactions in trimeric arginase. Genera-
tion of active monomers by mutation of a single amino
acid. J Biol Chem 276, 14242–14248.
45 Sabio G, Mora A, Rangel MA, Quesada A, Marcos
CF, Alonso JC, Soler G & Centeno F (2001) Glu-256 is
a main structural determinant for oligomerisation of

human arginase I. FEBS Lett 501, 161–165.
46 Cavalli RC, Burke CJ, Kawamoto S, Soprano DR &
Ash DE (1994) Mutagenesis of rat liver arginase
expressed in Escherichia coli: role of conserved histi-
dines. Biochemistry 33, 10652–10657.
47 Cama E, Emig FA, Ash DE & Christianson DW (2003)
Structural and functional importance of first-shell metal
ligands in the binuclear manganese cluster of arginase I.
Biochemistry 42, 7748–7758.
48 Carvajal N, Venegas A, Oestreicher G & Plaza M
(1971) Effect of manganese on the quaternary structure
of human liver arginase. Biochim Biophys Acta 250,
437–442.
49 Carvajal N, Martinez J, de Oca FM, Rodriguez J &
Fernandez M (1978) Subunit interactions and immobi-
lised dimers of human liver arginase. Biochim Biophys
Acta 527, 1–7.
50 Green SM, Ginsburg A, Lewis MS & Hensley P (1991)
Roles of metal ions in the maintenance of the tertiary
and quaternary structure of arginase from Saccharo-
myces cerevisiae. J Biol Chem 266, 21474–21481.
51 Bahl A, Brunk B, Crabtree J, Fraunholz MJ, Gajria
B, Grant GR, Ginsburg H, Gupta D, Kissinger JC,
Labo P et al. (2003) PlasmoDB: the Plasmodium
genome resource. A database integrating experimental
and computational data. Nucleic Acids Res 31,
212–215.
52 Humphrey W, Dalke A & Schulten K (1996) VMD:
visual molecular dynamics. J Mol Graph 14, 33–38.
53 Pal L, Dasgupta B & Chakrabarti P (2005) 3(10)-Helix

adjoining alpha-helix and beta-strand: sequence and
structural features and their conservation. Biopolymers.
78, 147–162.
54 Brooks B, Bruccoleri R, Olafson D, States D, Swami-
nathan S & Karplus M (1983) Charmm: a program for
macromolecular energy, minimization, and dynamics
calculations. J. Comput. Chem. 4, 187–217.
55 Aravind L, Iyer LM, Wellems TE & Miller LH (2003)
Plasmodium biology: genomic gleanings. Cell 115,
771–785.
G. A. Wells et al. Novel interaction in plasmodial arginase
FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS 3529
56 Pizzi E & Frontali C (2001) Low-complexity regions in
Plasmodium falciparum proteins. Genome Res 11,
218–229.
57 Xue HY & Forsdyke DR (2003) Low-complexity seg-
ments in Plasmodium falciparum proteins are primarily
nucleic acid level adaptations. Mol Biochem Parasitol
128, 21–32.
58 Yuvaniyama J, Chitnumsub P, Kamchonwongpaisan S,
Vanichtanankul J, Sirawaraporn W, Taylor P,
Walkinshaw MD & Yuthavong Y (2003) Insights into
antifolate resistance from malarial DHFR-TS
structures. Nat Struct Biol 10, 357–365.
59 Jean L, Withers-Martinez C, Hackett F & Blackman MJ
(2005) Unique insertions within Plasmodium falciparum
subtilisin-like protease-1 are crucial for enzyme
maturation and activity. Mol Biochem Parasitol 144,
187–197.
60 Shi J, Blundell TL & Mizuguchi K (2001) FUGUE:

sequence-structure homology recognition using environ-
ment-specific substitution tables and structure-depen-
dent gap penalties. J Mol Biol 310, 243–257.
61 Mora A, del Ara RM, Fuentes JM, Soler G & Centeno
F (2000) Implications of the S-shaped domain in the
quaternary structure of human arginase. Biochim
Biophys Acta 1476, 181–190.
62 Carvajal N, Martinez J & Fernandez M (1977) Immobi-
lised monomers of human liver arginase. Biochim
Biophys Acta 481, 177–183.
63 Di Costanzo L, Moulin M, Haertlein M, Meilleur F &
Christianson DW (2007) Expression, purification, assay,
and crystal structure of perdeuterated human arginase
I. Arch Biochem Biophys 465, 82–89.
64 Kim NN, Cox JD, Baggio RF, Emig FA, Mistry SK,
Harper SL, Speicher DW, Morris SM Jr, Ash DE, Traish
A et al. (2001) Probing erectile function: S-(2-borono-
ethyl)-L-cysteine binds to arginase as a transition state
analogue and enhances smooth muscle relaxation in
human penile corpus cavernosum. Biochemistry 40,
2678–2688.
65 Olszewski KL, Morrisey JM, Wilinski D, Burns JM,
Vaidya AB, Rabinowitz JD & Llinas M (2009) Host-
parasite interactions revealed by Plasmodium falciparum
metabolomics. Cell Host Microbe 5, 191–199.
66 Weinberg JB, Lopansri BK, Mwaikambo E & Granger
DL (2008) Arginine, nitric oxide, carbon monoxide, and
endothelial function in severe malaria. Curr Opin Infect
Dis 21, 468–475.
67 Sali A (1993) Comparative protein modelling by satis-

faction of spatial restraints. J Mol Biol 234, 779–815.
68 Marti-Renom MA, Stuart AC, Fiser A, Sanchez R &
Melo F (2000) Comparative protein structure modeling
of genes and genomes. Annu Rev Biophys Biomol Struct
29, 291–325.
69 Morris AL, MacArthur MW & Hutchinson EG (1992)
Stereochemical quality of protein structure coordinates.
Proteins 12, 345–364.
70 MacKerel J, Brooks IC, Nilsson L, Roux B, Won Y &
Karplus M (1998) Charmm: the energy function and its
parameterization with an overview of the program. The
Encyclopedia of Computational Chemistry 1, 271–277.
71 Phillips JC, Braun R, Wang W, Gumbart J, Tajkhors-
hid E, Villa E, Chipot C, Skeel RD, Kale L & Schulten
K (2005) Scalable molecular dynamics with NAMD.
J Comput Chem 26, 1781–1802.
72 DeLano W (2002) The PyMOL Molecular Graphics
System. DeLano Scientific, Palo Alto, CA.
73 Roberts E, Eargle J, Wright D & Luthey-Schulten Z
(2006) MultiSeq: unifying sequence and structure
data for evolutionary analysis. BMC Bioinformatics 7,
382.
74 Canutescu AA, Shelenkov AA & Dunbrack RL Jr
(2003) A graph-theory algorithm for rapid protein side-
chain prediction. Protein Sci 12, 2001–2014.
75 Sambrook J, Fritsch EF & Maniatis D (1989) Molecu-
lar Cloning: A Laboratory Manual, 2nd edn. Cold
Spring Harbor Press, Cold Spring Harbor, NY.
76 Archibald RM (1945) Colorimetric determination of
urea. J Biol Chem 157, 507–518.

Novel interaction in plasmodial arginase G. A. Wells et al.
3530 FEBS Journal 276 (2009) 3517–3530 ª 2009 The Authors Journal compilation ª 2009 FEBS