Date post: | 18-Sep-2016 |
Category: |
Documents |
Upload: | philip-martin |
View: | 214 times |
Download: | 1 times |
Structure, Vol. 9, 1071–1081, November, 2001, 2001 Elsevier Science Ltd. All rights reserved. PII S0969-2126(01)00672-4
Insights into the Structure, Solvation,and Mechanism of ArsC Arsenate Reductase,a Novel Arsenic Detoxification Enzyme
be reduced to arsenite before it is extruded. ArsC onthe R773 plasmid is an arsenate reductase that usesreduced glutathione (GSH) to convert arsenate to arse-nite via the reaction H3AsO4 � 2GSH → H3AsO3 � GS-SG � 2H2O. ArsC is a 141 residue protein (14.8 kDa)
Philip Martin,1 Srini DeMel,1 Jin Shi,1
Tatiana Gladysheva,1 Domenico L. Gatti,1
Barry P. Rosen,1 and Brian F.P. Edwards1,2
1 Wayne State University School of MedicineDepartment of Biochemistry
with a redox active cysteine residue in the active siteand Molecular Biology[3, 4]. Any one of the three glutaredoxins (Grx) in E.540 E. Canfieldcoli can participate in the catalytic cycle [5], althoughDetroit, Michigan 48201glutaredoxin 2 (Grx2), which has the highest affinity forArsC, is probably the preferred hydrogen donor in thereaction [5].
Summary Two other families of arsenate reductases have beenidentified that are unrelated to R773 ArsC. The best-
Background: In Escherichia coli bearing the plasmid characterized member of the second bacterial family isR773, resistance to arsenite, arsenate, antimonite, and the ArsC enzyme of Staphylococcus aureus plasmidtellurite is conferred by the arsRDABC plasmid operon pI258 [6], which is homologous to low-molecular-weightthat codes for an ATP-dependent anion pump. The prod- acid phosphatases [3]. Both the R773 and pI258 reduc-uct of the arsC gene, arsenate reductase (ArsC), is re- tases have catalytic cysteine residues but the latter usesquired to efficiently catalyze the reduction of arsenate thioredoxin (Trx) instead of glutaredoxin in its reactionto arsenite prior to extrusion. [6]. The third arsenate reductase, Acr2p, has been identi-
fied in the eukaryote Saccharomyces cerevisiae [7] andis related to the Cdc25A family of phosphotyrosyl phos-Results: Here, we report the first X-ray crystal struc-phatases [8].tures of ArsC at 1.65 A and of ArsC complexed with
Mutational studies of R773 ArsC have identified Cys-arsenate and arsenite at 1.26 A resolution. The overall12 as the catalytic residue and have shown that it isfold is unique. The native structure shows sulfate andactivated by a nearby positive charge [3], which wassulfite ions binding in the active site as analogs of arse-postulated to be His-8 [9]. A preliminary NMR studynate and arsenite. The covalent adduct of arsenate withof ArsC suggested that its overall fold was structurallyCys-12 in the active site of ArsC, which was analyzedhomologous to those of E. coli glutaredoxin, thiol trans-in a difference map, shows tetrahedral geometry with aferases, and glutathione S-transferase, with some “kin-sulfur-arsenic distance of 2.18 A. However, the corre-ship” to low-molecular-weight tyrosine phosphatasessponding adduct with arsenite binds as a hitherto un-based upon having a common anion binding motif [10].seen thiarsahydroxy adduct. Finally, the number of
ArsC forms an active quaternary complex with GSH,bound waters (385) in this highly ordered crystal struc-arsenate, and glutaredoxin 1 (Grx1) [4]. The three ligandsture approaches twice the number expected at this reso-must be present simultaneously for reduction to occur,lution for a structure of 138 ordered residues.although other oxyanions can replace arsenate in form-ing the quaternary complex. Blocked GSH derivatives,
Conclusions: Structural information from the adduct of such as S-methyl and S-hexyl glutathione, still form theArsC with its substrate (arsenate) and with its product quaternary complex but not as effectively as GSH. Ap-(arsenite) together with functional information from mu- parently, the GSH thiol group is required for catalysis buttational and biochemical studies on ArsC suggest a not for oxyanion binding. Moreover, GSH participatesplausible mechanism for the reaction. The exceptionally directly in the reaction, as evidenced by the inability ofwell-defined water structure indicates that this crystal dithiothreitol or �-mercaptoethanol to support catalysis.system has precise long-range order within the crystal Studies using mutants of ArsC and Grx1 suggest thatand that the upper limit for the number of bound waters Grx1 reduces a mixed disulfide in the reaction and thatin crystal structures is underestimated by the structures a GSH-dependent reaction occurs after the binding ofin the Protein Data Bank. the substrate, arsenate [4]. The reduction requires a free
thiol on both ArsC and GSH.This paper describes the novel structure of arsenateIntroduction
reductase and the structures of its complexes with sub-strate and product, namely arsenate and arsenite, atEscherichia coli has a chromosomal resistance system,high resolution. The latter complex has an unusual thiar-
encoded by the arsRBC operon, for the detoxificationsahydroxyl derivative attached to Cys-12 in the active
of arsenate, arsenite, and antimonite [1]. It may alsosite. These three structures identify the key active site
have a plasmid-coded system, of which the arsRDABC residues, delineate their catalytic roles, and suggest aoperon on the R733 plasmid is the best characterized reaction pathway in which the final step is the release(reviewed in [2]). Both systems transport arsenite and of arsenite.antimonite out of the cell. Arsenate, however, must first
Key words: arsenite; arsenate; crystal structure; water structure;reductase; X-ray crystallography2 Correspondence: [email protected]
Structure1072
Figure 1. The ArsC Structure
The overall fold of ArsC is shown as a topol-ogy diagram from TOPS [37] (top left), as aribbon diagram rendered by MOLMOL [38](top right), and as a labeled, C� stereo plot(bottom). The fold has no identical homologsalthough the lower ��� substructure hassome homology to both glutaredoxin andcrambin.
Results database using several different algorithms, includingboth sequence and structural homology, found no sig-nificant, global similarity to any known protein. AlthoughThe Protein Fold
The initial electron density map at 2.5 A resolution had ArsC is too small to have separate domains [14], it con-tains elements of super secondary structure that arean average figure of merit of 0.826 after density modifica-
tion. The program ARP/wARP fit the entire sequence present in other proteins. The four-stranded � sheet andthe two helices below it in the topology diagram (Figureautomatically except for two disordered residues at the
N terminus, two cis-proline residues at positions 95 and 1) resemble crambin [14], while the same sheet andhelices plus the last two strands resemble glutaredoxin108, which had to be fit manually, and a single disor-
dered residue at the C terminus. The native crystal struc- [15, 16]. The C� tracing in Figure 1 shows the relativelocation of residues 3–140, including the four residuesture has an R value of 0.126 (Rfree � 0.184) to 1.65 A for
138 residues. Analysis of the final structure with PRO- that are known to be required for arsenic resistance,namely Cys-12, Arg-60, Arg-94, and Arg-107 ([9] andCHECK [11] shows that 91.6% of the residues are in
the most favored region, and 8.4% are in additionally unpublished data).allowed regions. A Luzzati plot (data not shown) givesan estimate of 0.10 A for the mean coordinate error [12]. Hydration and Water Structure
A striking feature of this crystal structure is the extensiveThe E. coli ArsC has no detectable sequence similarityto any known enzyme, including the S. aureus ArsC canopy formed by 385 well-defined water molecules
that cover more than 96% of the accessible surface ofencoded by plasmid pI258 [3, 6]. The overall fold (Figure1) is of the � � � class [13]. A search of the structural the enzyme, and are, in places, three layers deep. Over
Arsenate Reductase1073
half of the waters were visible in the initial MIR map at charge attracts three sulfate anions (Figures 3b and 4a),which have sufficient occupancy to show a significant2.5 A; the rest were added during refinement. Of the 385
water molecules in the native structure, 169 are within anomalous signal (data not shown). One anion (SO4-1)is in the active site and has an oxygen atom within3 A of the surface of the protein, 205 are in the 3 to 6 A
shell, and 11 are in the shell greater than 6 A. The average hydrogen bonding distance (3.5 A) of the thiol group ofCys-12. It is stabilized further by a hydrogen bond (2.8temperature factors (B values) for these solvent shells
are 38.6 A2, 47.6 A2, and 52.7 A2, respectively. By compar- A) to the amide nitrogen of Gly-13 (Figure 4a). SO4-2 ison the “outside” of the active site and forms a hydrogenison, the average temperature factors for the protein back-
bone and side chain atoms are 24.8 A2 and 29.5 A2, respec- bond (3.4 A) with Arg-107 NH2. SO4-3, which has thebest anomalous signal, is further removed from the ac-tively. These solvent shells cover 96% of the accessible
surface area of ArsC (5658 A2), excluding the surface tive site and forms three strong hydrogen bonds withThr-14 OG1 (2.5 A), Ser-109 OG (2.7 A), and Lys-126 NZinvolved in crystal contacts (1455 A2) and the termini of
long hydrophilic side chains, namely Glu, Gln, Asp, Asn, (2.9 A).The electron density map of the native structureLys, and Arg (483 A2).
A recent analysis of water molecules observed in the shows a fourth anion in the active site. The electrondensity does not match a sulfate or acetate anion, butprotein crystal structures determined at different tem-
peratures and resolutions predicts a maximum of 212 does fit a sulfite anion (Figure 4a)—actually a bisulfiteanion at pH 4.8—in a strong hydrogen bond with Arg-water molecules for a structure of this size at 1.65 A
[17]. The ArsC structure shows that the upper limit for 94 (2.5 A). Evidently, with 1.2 M sulfate anion and 5 mMdithiothreitol (DTT) in the crystallizing medium, ArsC canthe amount of bound water that can be seen in X-ray
crystal structures is higher than previously estimated. reduce enough sulfate anions to sulfites over severalweeks to allow significant binding of sulfite to theenzyme.The Structural Cesium Ion
A cesium ion, with crystallographic coordinates 0,1/2,1/12, lies on a crystallographic two-fold axis and plays Reaction with Arsenate: Thiarsenato-Cys12-ArsC
When arsenate is diffused into ArsC crystals, it reactsan essential role in stabilizing this crystal system. Theion is at the center of a distorted, octahedral bipyramid with Cys-12 in a mechanism analogous to the formation
of a thiophosphate by phosphatases (reviewed in [18]),formed by the carbonyl oxygens of Leu-113, Asp-114,and Leu-116, a structural water molecule, and the sym- i.e., formation of a pentacoordinate arsenic followed by
collapse to a covalent, tetrahedral, thiol intermediate.metry mates of these ligands. Consequently, surrogatemother liquors must be doped with at least 0.1 M cesium This is the first step in the reaction pathway. Soaking
an ArsC crystal in 0.5 M arsenate in the standard cryo-or the crystals deteriorate rapidly. Crystals soaked forup to a week in these mother liquors still diffracted to protectant at pH 4.8 for two days produced an appar-
ently unaltered enzyme with a mean fractional isomor-at least 1.8 A, although roughly half of the cesium hadleached out, based upon refined occupancies. phous difference of only 4.3% (Table 1), indicating that
the native and arsenate-treated crystal are very similarAdditional, loosely bound cesium sites were locatedin the native structure by anomalous difference Fourier to one another. However, the difference electron density
map (Farsenate � Fnative) shows concerted shifts of electronmaps. Seven sites were located with the MIR phasesfrom the platinum and gold derivatives and subse- density within the active site (Figure 5). The density for
the sulfate ion, SO4-1, decreases, and new density cor-quently used in phasing (Table 1). After refinement, threeadditional sites were located. All but two of the adventi- responding to the S-AsO3 moiety appears, as predicted
by earlier mutational studies [4]. The Cys-12 thiol shiftstious sites are located close to negatively chargedaspartic or glutamic acid side chains, or to one of the to form the new S-As bond. The guanidino group of Arg-
94 moves out of the active site to avoid colliding withsulfate ions, which act as counter ions. The notableexception is the structural cesium ion; it has no charged the newly formed S-AsO3 group, while the side chain of
Arg-60 moves closer to the adduct. Arg-107 remains inneighbors in the electron density maps. Apparently, thecesium ion is adequately neutralized by the electronega- place. The S-AsO3 adduct is stabilized by three strong
hydrogen bonds, two to the side chains of Arg-60 andtive carbonyl atoms that bind it, or one of its “water”ligands is actually a hydroxide ion. 107 and one to the amide nitrogen of Gly-11.
Reaction with Arsenite:The Active SiteThe catalytic residue Cys-12 is held rigidly at the end Thiarsahydroxy-Cys12-ArsC
When an ArsC crystal is soaked in 0.09 M arsenite inof the first � strand in the middle of two consecutivetype I � turns as the strand turns and inserts into the the standard cryoprotectant, an adduct is formed at
Cys-12 (Figure 4b). The new electron density increasesfirst � helix (Figures 1 and 2). The Cys-12 N is the lastatom in the first � turn, and the carbonyl oxygen is the as the arsenite concentration is increased to 0.4 M and
can be removed by “back-soaking” the crystal in cryo-first atom in the second � turn. The side chain of His-8is 7.2 A away from Cys-12 and makes a side chain protectant without arsenite (Figures 6a–6d). The pres-
ence of an arsenic atom attached to the Cys-12 sulfurhydrogen bond (2.7 A) to the OG of Ser-15, therebystabilizing the geometry of the active site loop. atom is confirmed by diffraction data collected at a
wavelength of 1.0 A, which shows a significant anoma-Cys-12 is sandwiched between Arg-94 and Arg-107,with Arg-60 nearby. This concentration of positive lous difference peak adjacent to the sulfur atom of Cys-
Structure1074
Tab
le1.
Sum
mar
yS
tatis
tics
Dat
aP
hasi
ng
Mea
sure
dU
niq
ueP
erce
ntP
hasi
ngP
hasi
ngD
ata
Set
�(A
)d
min
Ref
lect
ions
Ref
lect
ions
Ob
serv
edS
itesa
Rsy
mM
FID
bR
anom
RC
ullis
RK
raut
Rat
io
Nat
ive
Ars
Cc
1.54
1.65
1,35
7,85
655
,121
d96
.97
Cs
.137
–.0
30–
.017
0.91
K2P
tCl 4
1.54
2.50
137,
853
9,42
599
.63
Pt
.085
.110
–.5
60.1
521.
36K
2PtC
l 4an
o1.
542.
50–
––
––
–.0
30–
.212
0.91
KA
uBr 4
1.54
2.30
418,
370
12,1
3499
.91
Au
.107
.190
–.6
08.2
741.
09K
AuB
r 4an
o1.
542.
30–
––
––
–.0
32–
.436
1.27
Cs 2
SO
4(2
.7M
)1.
001.
6032
6,36
832
,034
97.9
15C
s.0
61.1
22–
.587
.160
1.40
Cs 2
SO
4an
o1.
001.
60–
––
––
–.0
29–
.241
0.59
0.09
MA
sO3�
31.
541.
651,
011,
969
30,0
9698
.4–
.096
.076
.031
––
–0.
27M
AsO
3�3
1.54
1.65
888,
328
53,2
40d
94.1
–.0
95.0
70.0
26–
––
0.40
MA
sO3�
31.
001.
2636
3,77
151
,735
90.1
–.0
59.1
04.0
34–
––
Bac
kso
ak1.
541.
8076
3,15
924
,745
98.0
–.1
45.1
79.0
30–
––
0.50
MA
sO4�
31.
541.
7572
2,59
822
,762
84.8
–.1
07.0
43.0
25–
––
Ref
inem
ent
with
Dat
afr
om
20A
toth
eR
eso
lutio
nLi
mit
for
Eac
hC
ryst
al
Pro
tein
So
lven
tE
stim
ated
Err
or
RM
SM
od
elR
eR
free
eA
tom
s�B
�A
tom
s�B
��
fLu
zzat
iB
ond
sA
ngle
sgP
lane
sC
hira
l
Ars
CN
ativ
e0.
129h
0.18
510
8627
.2A
238
543
.6A
2–
0.10
A0.
015
A0.
036
A0.
031
A0.
066
A3
0.09
MA
sO3�
30.
176
0.22
010
8819
.0A
237
438
.3A
2.0
76A
0.18
A0.
010
A0.
027
A0.
028
A0.
066
A3
0.27
MA
sO3�
30.
113h
0.17
610
8827
.2A
238
346
.2A
2.0
70A
0.08
A0.
016
A0.
038
A0.
030
A0.
058
A3
0.40
MA
sO3�
30.
128h
0.16
610
8816
.2A
231
840
.3A
2.1
04A
0.08
A0.
013
A0.
032
A0.
0–30
A0.
074
A3
Bac
kso
ak0.
189
0.24
410
8626
.7A
236
745
.3A
2.1
79A
0.20
A0.
008
A0.
023
A0.
029
A0.
049
A3
0.50
MA
sO4�
30.
163
0.19
210
8619
.9A
225
134
.2A
2.0
43A
0.14
A0.
012
A0.
031
A0.
031
A0.
081
A3
aT
heP
t,A
u,an
dC
sd
eriv
ativ
esal
soha
dth
e7
Cs
site
sse
enin
the
nativ
ecr
ysta
lsth
atw
ere
tran
sfer
red
bri
efly
toth
ecr
yop
rote
ctan
t(0
.27
MC
s 2S
O4)
.T
hear
seni
tecr
ysta
ls,
whi
chw
ere
soak
edfo
r2–
4d
ays
inth
ecr
yop
rote
ctan
tw
ithad
ded
arse
nite
,ha
d5–
6o
fth
ena
tive
Cs
site
s.bM
ean
frac
tiona
liso
mo
rpho
usd
iffer
ence
on
|F|f
or
20–2
.5A
dat
a.cT
hem
ean
figur
eo
fm
erit
for
the
initi
alp
hase
sca
lcul
ated
by
mul
tiple
iso
mo
rpho
usre
pla
cem
ent
was
0.64
0b
efo
red
ensi
tym
od
ifica
tion
and
0.82
6af
terw
ard
s.dN
umb
ero
fun
ique
refle
ctio
nsw
ithF
ried
elp
airs
cons
ider
edas
sep
arat
ere
flect
ions
.eV
alue
sar
efo
rth
e4�
(F)
dat
a.W
hen
allt
hed
ata
are
incl
uded
,th
eva
lues
for
the
nativ
ed
ata
set
are
R�
0.13
4an
dR
free
�0.
191.
f Ro
ot-
mea
n-sq
uare
dev
iatio
no
fal
lato
ms
com
par
edto
nativ
eA
rsC
.gS
HE
LXre
po
rts
ang
led
evia
tions
as1–
3d
ista
nces
.hF
ulla
niso
tro
pic
refin
emen
tw
ithth
ead
diti
on
of
“flo
atin
g”
hyd
rog
ens
(see
text
).
Arsenate Reductase1075
Figure 2. Active Site Loop
The catalytic residue Cys-12 is held in a rigidloop formed by two type I � turns that bracketthe cysteine. The loop is stabilized by fourinternal hydrogen bonds including one be-tween the side chains of His-8 at the begin-ning of the loop and Ser-15 at the end of theloop.
12 (Figure 7c). At this wavelength, arsenic has an anoma- the above equation implies that the ratio of labeled ArsCto free ArsC will decrease dramatically as the pH in-lous signal of 3.5 electrons, which distinguishes it from
the other atoms in the structure. The small but real den- creases.sity about the sulfur atom is also notable. It attests tothe outstanding quality of the anomalous data collected Discussionfrom this crystal, since sulfur has an anomalous signalof only 0.25 electrons at this wavelength. In this communication, we report the first three-dimen-
sional structure of a novel arsenic-detoxifying enzyme.The thiarsahydroxy adduct is the result of a simpleaddition reaction which is the reverse of the last step A previous report determined most of the secondary
structure of ArsC by NMR methods but could not mea-in the mechanism of ArsC: Cys-SH � H3AsO3 → Cys-S-As�-OH � H2O � OH�. A simple equilibrium constant sure sufficient NOE distances to calculate a reliable ter-
tiary structure [10].for the reaction can be written as follows: Keq � [E-Cys-S-As�-OH] [H2O] [OH�] / [E-Cys-SH] [H3AsO3], where E Although the NMR topology map was missing two
helices between � sheet strands 2 and 3 and two shortis the concentration of enzyme in the crystal. Since theconcentration of water is constant, we can combine strands of the � sheet at the C terminus, the topology
map did identify E. coli glutaredoxin (PDB code 1EGO),Keq / [H2O] into a new constant, Keq. The hydroxide ionconcentration at pH 4.8 is 6.31 10�10 M. The concentra- glutathione S-transferase (PDB code 2GSR), and thiol-
transferase (PDB code 1KTE) as the nearest homologs oftion of enzyme is calculated to be 0.0264 M from thevolume of the unit cell volume (755,678 A3) and the 12 ArsC [10]. A similar analysis with the ArsC X-ray structure
shows that glutaredoxin and glutathione S-transferaseequivalent positions. Finally, the concentration of theadduct is the enzyme concentration multiplied by the are still the closest relatives of the ArsC X-ray structure
in the TOPS database [19]. However, they both haverefined occupancy of the arsenic atom, q, that is calcu-lated by SHELXL. It follows that the concentration of a rank of 13, where a value below 10 is considered
significant. Thioltransferase, with a rank of 15, is clearlythe unlabeled enzyme is 1 � q times the enzyme concen-tration. Substituting the above values reduces the equi- more distant from ArsC.
Previous studies have also suggested that the activelibrium expression to Keq � q [OH�] / (1 � q) [H3AsO3].The refined occupancies of the arsenic atom, q, for arse- site of ArsC might resemble that of the low-molecular-
weight tyrosine phophatases [10], with His-8 depressingnite concentrations of 0.09 M, 0.27 M, and 0.40 M are0.150, 0.350, and 0.381, respectively. These values imply the pK of the active site residue, Cys-12, by more than
two pH units [9]. The latter suggestion is not supportedan average value of 1.1 � 0.1 10�9 for the equilibriumconstant at pH 4.8. The presence of hydroxide ion in by the crystal structure, which clearly shows that Cys-
Figure 3. The Electrostatic Surface of ArsC
(a) The electrostatic surface of ArsC is shownin the absence of the sulfate anions. The cata-lytic residue Cys-12 is in a strongly positiveregion (blue) due to the close proximity ofthree arginines (60, 94, and 107).(b) The electrostatic surface of ArsC is shownwith the bound surface anions included in thecalculations. The binding of three sulfate ionsand one bisulfite ion in and near the activesite overwhelms the positive charge with alarger negative charge (red) that drasticallychanges the electrostatic characteristics onthe surface of the enzyme. The electrostaticsurfaces were calculated with GRASP [39].
Structure1076
Figure 4. The Active Site of ArsC
(a) The structure without arsenite shows hy-drogen bonds between a sulfate anion(SO4-1) and the Cys-12 SH and the Gly-13NH groups. Sulfate is isosteric with arsenateand consequently is an analog of the sub-strate. A putative bisulfite anion, which isisosteric with arsenite and therefore is an an-alog of the product, makes a hydrogen bondwith the guanidino group of Arg-94.(b) Arsenite has reacted with the active siteCys-12 thiol to produce a thiarsahydroxy de-rivative (see text). The OH group on the arse-nic atom is stabilized by a hydrogen bondto Arg-107. The putative �1 charge on thearsenic atom is neutralized by the nearby sul-fate ion, which has shifted slightly from itsposition in the apo-enzyme to accommodatethe arsenic atom. An occupancy of approxi-mately 0.5 for the arsenite is deduced fromthe SHELX refinement. The sulfate ion(SO4-1) was refined in two conformations,one corresponding to that observed in thenative structure (orange bonds) and onemoved slightly to accommodate the arsenicatom (yellow bonds).
12 is activated by hydrogen bonds from Arg-94 and Arg- of a mixed � sheet, precludes a direct relationship toArsC.107. His-8, which is more that 7 A from Cys-12, stabilizes
the active site loop by forming a side chain hydrogen The crystal structure shows that the active site of ArsChas no similarity to that of the known phosphatases.bond with Ser-15 (Figure 2). The suggested link to the
phosphatases was based primarily upon their character- When the catalytic cysteine residues of low-molecular-weight tyrosine phosphatase and its bound sulfate anionistic CX5R anion binding motif, since their topology,
which contains a four-stranded parallel � sheet instead ([20]; PDB code 1PHR) are superimposed on Cys-12 and
Figure 5. The ArsC-Arsenate Complex
The FAsV � Fapo difference electron density inthe active site of ArsC complexed with arse-nate is shown in stereo. Negative (red) andpositive (blue) contours are at the 3� level.Sulfate, Cys-12, Arg-60, and Arg-94 move asindicated to accommodate the formation ofthioarsenate. Atoms whose positions are un-changed from the apo-enzyme structure, areshown in light blue; the arsenic atom attachedto Cys-12 is shown in a medium blue.
Arsenate Reductase1077
tions as a general acid and base in the mechanism ofthe phosphatases [21, 22].
Given that pI258 ArsC and Acr2p arsenate reductaseare both homologs of protein phosphatases, it is possi-ble that the R773 ArsC also evolved from a currentlyunidentified phosphatase and retains structural similari-ties to the ancestral active site. ArsC has no measurablephosphatase activity in vitro (R. Mukhopadhyay, B.P.R.,unpublished data). ArsC may have diverged so far fromits ancestral phosphatase that it has lost all recognizablehomology and all phosphatase activity. Alternatively, thesuperficial resemblance of the active site to that of aphosphatase could be the result of structural conver-gence driven by using the same chemistry, namely nu-cleophilic attack by a cysteine sulfur atom on an oxy-anion.
Reaction MechanismThe first mechanism proposed for ArsC began with anucleophilic attack on an arsenate anion by ArsC Cys-12 to give a {Cys-12}S-AsV intermediate [4]. A glutathionemolecule then attacked the {Cys-12}S-As bond to pro-duce the {Cys-12}S-S{glutathione} mixed disulfide inter-mediate and free arsenite (AsIII). In the penultimate step,
Figure 6. The Addition and Removal of Arsenite glutaredoxin removed the oxidized glutathione moleculeThe electron density (2Fo � Fc) at Cys-12 is shown for ArsC crystals from ArsC. The catalytic cycle was completed when thetreated with arsenite. glutathione molecule was abstracted from {glutare-(a) The crystal was soaked in 0.09 M arsenite for 1 hr. doxin}-S-S-{glutathione} by a second glutathione mol-(b) The crystal was soaked in 0.27 M arsenite for 4 days.
ecule.(c) The crystal was soaked in 0.40 M arsenite for 2 days. The redBased upon these studies, we propose a revisedcontours show the anomalous difference density calculated from
Friedel pairs collected at a wavelength of 1.0 A. mechanism for ArsC with four steps and three intermedi-(d) The crystal was soaked in 0.4 M arsenite for 4 days and then ates (Figure 7). The first step, which involves the forma-“back-soaked” in cryoprotectant without arsenite for 3 days. The tion of the thioarsenate binary adduct of ArsC (Interme-diffraction data for (a), (b), and (d) were collected at a wavelength diate I), is the same as in the original mechanism. Theof 1.54 A (Table 1).
second step in which a {ArsC Cys-12}S-AsV-S{glutathi-one} tertiary complex is formed (Intermediate II) is in-ferred from biochemical studies using a fluorescentSO4-1 of ArsC, the CX5R motif diverges greatly from themutant of ArsC (A11W) to follow the reaction [4]. TheCX4R sequence of ArsC (not shown). Moreover, Arg-16,spectroscopic signal from the inserted tryptophan resi-the ArsC homolog of the invariant arginine residue indue indicates that GSH reacts only after arsenate binds.the CX5R motif of phosphatases that is required for sub-Moreover, the reaction requires a free thiol on GSH andstrate binding and transition state stabilization, is outArsC to proceed. In the third step, arsenate is reducedof the active site of ArsC and does not interact withto arsenite in a quaternary complex with glutaredoxineither arsenate or arsenite. The active site of ArsC alsothat dissociates into the thiarsahydroxy adduct of ArsClacks a homolog of the aspartic acid residue that func-(Intermediate III) and a mixed disulfide complex of gluta-thione and glutaredoxin. In the fourth step, the ArsC-arsenite bond is hydrolyzed (Figure 6d). Eukaryotic arse-nate reductases probably employ the same mechanism,as evidenced by the observation that yeast Acr2p arse-nate reductase is fully active with E. coli glutaredoxinand that the R773 ArsC can utilize yeast glutaredoxin [7].
The existence of Intermediate I is demonstrated bythe difference electron density in Figure 5. IntermediateIII is seen in Figures 4b and 6a–6c, and its facile hydroly-sis is demonstrated in Figure 6d. The participation ofglutaredoxin as a second source of electron has beenestablished previously [4, 5]. So far, Intermediate II has
Figure 7. The Reaction Mechanism of ArsCeluded our efforts to image it by X-ray crystallography.
The mechanism shown here is consistent with the crystal structures Although there is spectroscopic evidence for a ternaryof the arsenate complex (Intermediate I) and the arsenite complex
complex between ArsC, arsenate, and glutathione, the(Intermediate III) of ArsC. Intermediate II and the interaction withelectron density maps calculated for crystals soaked inglutaredoxin (GrxSH) are inferred from previous biochemical studies
(see text). arsenate and then glutathione show only the arsenate.
Structure1078
Apparently, the ternary complex is too short lived to be groups enough for it to become a good leaving group.Once the hydroxyl group has departed, resonance andcaptured by routine X-ray crystallographic methods. The
structure of the quaternary complex with glutaredoxin the close proximity of a negatively charged oxyanion(sulfate, phosphate, or arsenate) stabilize the speciesalso remains to be determined. However, it has been
shown in vitro that glutaredoxin binds to ArsC only if we see here. As evident in the binding constant, thethiarsahydroxy adduct is much less stable than mostglutathione and arsenate or a similar oxyanion (phos-
phate, sulfate) is present [4]. The fact that multiple oxya- cysteine-arsenite complexes—which is essential if arse-nite is not to function as a suicide inhibitor of ArsC.nions bind to ArsC and dramatically change the electro-
static surface of ArsC (Figure 4) might explain whyoxyanions are required for ArsC to bind glutaredoxin. The Sulfate/Sulfite Anions
There are three sulfate anions bound to ArsC in thissystem. The first anion, SO4-1, is oriented in the activeThe Thiarsahydroxyl Adductsite like a substrate analog. An arsenate molecule at itsThe arsenite adduct in Intermediate III, which has onlyposition could be attacked by the Cys-12 thiol to formtwo atoms linked to the arsenic atom, has two possiblethe different intermediates discussed earlier. The anom-structures. It could be Cys-S-As�O, a trivalent arsenicalalous difference Fourier map, calculated from data mea-with a double-bonded oxygen similar to that found insured on the arsenite adduct at a wavelength of 1.0 Acompounds such as phenylarsine oxide, or it could beand 0.4 M arsenite, shows an anomalous signal (dataCys-S-As�-OH, which has no precedent. A search ofnot shown) at the sulfur position of SO4-1. This indicatesthe Cambridge database of small molecule structuresthat SO4-1 can be replaced by arsenite. The binding[23] revealed that organometallic As(III), with one excep-constant for arsenite must therefore be significantlytion, has three single bonds. The exception is benzo-higher than for sulfate, since the mother liquor contains1,3,2-thiarsolium tetrachloroaluminate, an aromatic ar-2.7 M sulfate. The three bound anions neutralize thesenical with a single positive charge on the AsIII atompositive electrostatic potential of the active site (Fig-[24]. All the arsenic(III) derivatives in the Protein Dataure 4).Bank also have three single bonds. For example, the
The electron density for the sulfite anion that is hydro-crystal structure of HIV integrase (PDB code 1BHL),gen bonded to Arg-94 in the structure of native ArsCwhich was crystallized in the presence of 100 mM caco-(Figure 4) is also present in the structures of all of thedylate and 5 mM DTT, contains dimethylarsenic cysteinethiarsahydroxy adducts, as well as in the crystal treateddue to the reduction of cacodylate by DTT and subse-with arsenate. In the back-soaked crystal, however, thequent reaction with exposed cysteine residues [25].electron density is much flatter, although it is still trigonalSince the positions of two covalent ligands of the(data not shown). Consequently, we have fit the densityarsenic atom in the arsenite adduct are known, the pos-with an acetate molecule, which was 100 mM in thesible positions of a putative third ligand, assumed tosoaking solution, instead of a sulfite anion. Apparently,be a hydroxyl group, are easily calculated. The modelthe sulfite—or added arsenite—anion bound at this siteindicates that a collision would occur on one side withnext to Arg-94 leaches out of the crystal during the back-the bound sulfate-1 and on the other side with the sidesoaking experiment and is replaced by acetate.chain of Arg-94 (data not shown). However, it is unlikely
that steric hindrance would drive the loss of the putativehydroxyl group, since the two blocking groups are The Solvent Structure
There is no telling argument for why the number of well-mobile.Although a positively charged arsenic atom is a sur- defined water molecules in the ArsC structure exceeds
the expected value by almost 100%. At first, we sus-prise, the arsenic-oxygen bond in the arsenite adductis definitely a single bond. Small molecule structures pected an artifact of the crystallization conditions. Per-
haps the added cesium in some way promoted the hy-give values of 1.83 A for an As-O bond and 1.67 A fora As�O bond [26, 27]. A final cycle of full matrix least dration of ArsC. This hypothesis was discounted by
adding cesium to other crystals we are currently examin-squares refinement of the structure of the adduct givesa length of 1.86 � 0.04 A for the bond between the ing. Neither the maximum resolution nor the number of
observed water molecules increased. Another possibil-arsenic and oxygen atoms. It is possible that the posi-tively charged intermediate is stabilized by resonance ity is the addition of 20% trehalose as a cryoprotectant.
However, it likewise has not affected the structure ofas follows: �CH2-S-As�-OH � - - - - - - �CH2-S��As-OH. If the above sulfonium ion species with its double the hydration layer in other systems where we have used
it. Other variables, such as the acetate buffer at pH 4.8,bond character is significant, then the C�-S�-As-O groupshould be planar and the S-As bond somewhat shorter. are common to many other crystal systems that do not
exhibit such extensive layers of solvent.Indeed, the four atoms deviate by a maximum of only0.016 A from a least squares plane, although they were The surface of ArsC is unexceptional. It has patches
of positive and negative charge dispersed throughoutnever restrained to be planar at any point in the refine-ment. The S-As bond in the adduct is 2.20 � 0.05 A, as the surface (Figure 4) as seen in other proteins. The
packing in the crystal is also unexceptional. The ArsCcompared to 2.24 A for a single bond [26].A chemical and tautological argument can be made molecules occupy approximately 30% of the crystal vol-
ume and have packing contacts with five neighbors.as to why the thiarsahydroxy species forms in the activesite. The close proximity of Arg-94, perhaps aided by However, a recent Laue photograph (not shown) of the
native crystal at room temperature suggests that theArg-107 and Arg-60, polarizes one of the hydroxyl
Arsenate Reductase1079
mosaic spread may be closer to 0.1�. Thus, the elevated that the reaction is completely reversible. By combiningthis new information with the biological work done pre-number of water molecules that we see in these struc-
tures is probably due to the exceptional, long-range viously, we propose a reaction pathway for this enzymethat incorporates these intermediates and the other co-order in this crystal system. Neutron diffraction, which
yields a stronger signal from water molecules than does factors (glutathione, sulfate and/or phosphate, glutare-doxin) into the reaction scheme.X-ray diffraction, supports our conclusion that the num-
ber of water molecules bound to proteins is commonly Finally, all six ArsC structures that were independentlyrefined at high resolution (1.8–1.26 A) in this study haveunderestimated. The neutron crystallographic structure
of concanavalin A at 2.4 A resolution includes 160 water almost twice the number of waters normally observedon protein surfaces. This observation suggests that themolecules (PDB code 1C57 [28]), while the X-ray struc-
ture determined at 2.15 A resolution has only 86 water number of statically bound waters on protein surfacesis dramatically higher than previously suspected frommolecules (PDB code 1DQ1 [29]).
The structural cesium atom, which sits at a special other X-ray crystallographic structures.position defined by the symmetry of the crystal space
Experimental Proceduresgroup, contributes to this long-range order. It forms athree-dimensional “chain link fence” throughout the lat-
The protein was produced as previously described [3]. Crystals weretice. Consequently, every ArsC molecule is attached togrown at 5�C by hanging-drop vapor diffusion from 1.2 M Cs2SO4,a neighboring ArsC very precisely at the special position. 100 mM ammonium acetate (NH4OAc) (pH 4.8) at a protein concen-
Since each cesium atom must sit on a symmetry ele- tration of 30 mg/mL. They belong the hexagonal space group P6122,a � b � 86.73 A, c � 116.17 A, � � � � 90�, � � 120�. Because thement, the three-dimensional order is replicated very pre-elevated cesium concentration gave an unacceptably high back-cisely—more so than in other crystal systems whichground in the X-ray diffraction data, all crystals except the “cesiumhave higher static disorder due to less-constrainedderivative” were transferred to cryoprotectant mother liquor com-packing interactions between the proteins. This argu-posed of 70% saturated (NH4)2SO4, 0.27 M Cs2SO4, 20% w/v treha-
ment suggests that proteins have significantly more lose, and 100 mM NH4OAc (pH 4.8) before data collection (this wasbound water molecules than are generally reported for the “standard” cryoprotectant). The arsenite/arsenate derivatives
were soaked in the standard cryoprotectant containing added arse-the crystal structures in the Protein Data Bank.nite/arsenate for up to four days (Figure 6). The “back-soaked”arsenite was kept in 0.4 M arsenite in the standard cryoprotectantfor 4 days, then soaked in the standard cryoprotectant without arse-Biological Implicationsnite for three days. All data were collected at �180�C. The structurewas phased with three derivatives, Pt, Au, and Cs. Both the platinum
Arsenic is a public health problem in many parts of the and gold derivatives were prepared by soaking crystals in cryopro-world [30]. Among its various health effects, it has been tectant solution containing the heavy metal for 1–2 days, while the
cesium derivative was stored in cryoprotectant solution containingassociated with a high incidence of several forms of2.7 M Cs2SO4.cancer. The World Health Organization recommends an
Data were measured for the native, platinum, gold, arsenite (ex-upper limit of 10 �g/l in drinking water, but many coun-cept the 0.4 M arsenite derivative), and arsenate derivatives on atries, including the United States, have not met this stan-Raxis IV image plate data collection system mounted on a Rigaku
dard because of the enormous cost of arsenic remedia- RU-200 X-ray generator running at 50 kV and 100 mA. The Cu K�tion. Thus, biological mechanisms of arsenic resistance radiation was focused with Yale mirrors before passing through a
0.0015 mm Ni filter and a 0.5 mm collimator. Frames of data at 1�are of considerable importance, and the structure of anwere collected for 5–10 min. The native data were collected overarsenic detoxifying enzyme is extremely timely.477�, the platinum data over 293�, the gold data over 398�, the 0.09In this study, we report the first structure of the en-M arsenite data over 332�, the 0.27 M arsenite data over 317�, thezyme arsenate reductase (ArsC) that helps an organism0.5 M arsenate data over 350�, and the back-soaked data over 296�.
survive in an environment contaminated with arsenic. The Pt and Au data were reduced to structure amplitudes withThe overall fold of this protein is unique. Its biological DENZO and scaled with SCALEPACK [31]; the rest of the data sets
were processed with XGEN [32].function is to reduce arsenate, the species of arsenicData for the cesium and 0.4 M arsenite derivative were collected atmore prevalent in our oxidizing atmosphere, to arsenite,
the Advanced Photon Source (IMCA beam line ID-17) at the Argonnewhich can then be extruded by the bacterium.National Laboratory. For the cesium derivative, 90� of data wereAt high resolution (1.65 A), we see that the nativecollected at 0.5�/frame on a MAR CCD at a wavelength of 1.000 A.
structure utilizes the chemistry of the Cys-12 thiol group, Frame exposure time was 1.5 s. For the arsenite derivative (0.4 Min concert with at least three arginines (60, 94, and 107), arsenite in the standard cryoprotectant, 2 days), 120� of data were
collected at 0.5�/frame for 1 s, with the detector offset to collectfirst to trap the arsenate in three binding pockets. Thisdata to 1.2 A. These data sets were reduced to structure amplitudesstep changes the overall surface charge at the activewith XGEN.site from positive to negative and facilitates the binding
The first heavy-atom position in the platinum derivative was lo-of other cofactors and enzymes in later steps in thecated with SOLVE [33] by using data from 12 to 2.5 A in space group
reduction process. The active site residues move dra- P6122. A difference Fourier map located additional platinum sites,matically to accommodate the first intermediate of the and an anomalous difference Fourier map located the seven cesium
sites in the native data set. Difference Fourier maps located thereaction, ArsC-Cys-12-thioarsenate. The third interme-gold and cesium sites for their respective derivatives. The structurediate in the reaction, Cys-12-thiahydroxy-ArsC, whichwas phased and the map solvent leveled with the PHASES packagewas formed in this study by mixing ArsC with its product,[34]. The program ARP/WARP [35] fit almost all of the main chainarsenite, has a novel chemical structure with only twoautomatically into the MIR map in a single pass and extended the
atoms bonded to the AsIII atom in a planar arrangement. resolution to 1.65 A (Table 1). Automated side chain fitting with ARP/We have calculated an equilibrium constant using struc- WARP and manual rebuilding completed the structure. ARP/WARP
fit 265 solvent positions.tures at varying concentrations of arsenite and shown
Structure1080
SHELX [36] was used to finish the refinement, adding additional 15. Nordstrand, K., Sandstrom, A., Aslund, F., Holmgren, A., Otting,G., and Berndt, K.D. (2000). NMR structure of oxidized glutare-water molecules. The various structures were refined on F2 with all
of the data from 20 A to the limit of resolution for each crystal. doxin 3 from Escherichia coli. J. Mol. Biol. 303, 423–432.16. Sodano, P., et al., and Wuthrich, K. (1991). Sequence-specific 1HFriedel pairs were not merged in the diffraction data from the native
crystal, the 0.27 M arsenite crystal, and the 0.5 M arsenate crystal. NMR assignments and determination of the three-dimensionalstructure of reduced Escherichia coli glutaredoxin. J. Mol. Biol.This gave an observation-to-parameter ratio large enough to permit
a full anisotropic refinement with the addition of “floating” hydrogen 221, 1311–1324.17. Carugo, O., and Bordo, D. (1999). How many water moleculesatoms in the last cycles, an option that does not add extra parame-
ters to the refinement. The 0.4 M arsenite crystal had data all the can be detected by protein crystallography? Acta Crystallogr.D 55, 479–483.way to 1.26 A. There were sufficient data in this set to perform full
anisotropic refinement without having to treat the Friedel structure 18. Fauman, E.B., and Saper, M.A. (1996). Structure and functionof the protein tyrosine phosphatases. Trends Biochem. Sci. 21,factors as separate observations. Unfortunately, the back-soaked
crystal diffracted to only 1.8 A after being subjected to multiple 413–417.19. Westhead, D.R., Slidel, T.W., Flores, T.P., and Thornton, J.M.manipulations, so a full anisotropic refinement was not possible.
(1999). Protein structural topology: automated analysis and dia-The overall statistics for the data sets collected during this studygrammatic representation. Protein Sci. 8, 897–904.are shown in Table 1.
20. Su, X.D., Taddei, N., Stefani, M., Ramponi, G., and Nordlund,P. (1994). The crystal structure of a low-molecular-weight phos-Acknowledgmentsphotyrosine protein phosphatase. Nature 370, 575–578.
21. Alhambra, C., Wu, L., Zhang, Z.-Y. and Gao, J. (1998). Walden-We thank M. Doyle and B. Prince for helping to crystallize ArsC; A.inversion-enforced transition-state stabilization in a protein ty-Howard and J. Chrzas for assistance at the synchrotron; S. Moba-rosine phosphatase. J. Am. Chem. Soc. 120, 3858–3866.shery for reviewing the mechanism; and S. Stevens, E. Zuiderweg,
22. Kolmodin, K., Nordlund, P., and Aqvist, J. (1999). Mechanismand L. Lee for early access to their NMR results. This work wasof substrate dephosphorylation in low Mr protein tyrosine phos-supported in part by National Institutes of Health grants AI43918phatase. Proteins 36, 370–379.(D.L.G.) and GM52216 (B.P.R.).
23. Allen, F.H., and Kennard, O. (1993). 3D search and researchusing the Cambridge Structural Database. Chem. Des. Auto.Received: June 4, 2001News 8, 31–37.Revised: October 3, 2001
24. Burford, N., Parks, T.M., Royan, B.W., Richardson, J.F., andAccepted: October 10, 2001White, P.S. (1992). Benzothiarsolium cations: evidence for p�-bonding between arsenic and sulfur in the first structurally
Referencescharacterized arsenium system. Can. J. Chem. 70, 703–710.
25. Maignan, S., Guilloteau, J.P., Zhou-Liu, Q., Clement-Mella, C.,1. Carlin, A., Shi, W., Dey, S., and Rosen, B.P. (1995). The ars and Mikol, V. (1998). Crystal structures of the catalytic domain
operon of Escherichia coli confers arsenical and antimonial re- of HIV-1 integrase free and complexed with its metal cofactor:sistance. J. Bacteriol. 177, 981–986. high level of similarity of the active site with other viral integ-
2. Rosen, B.P. (1999). Families of arsenic transporters. Trends rases. J. Mol. Biol. 282, 359–368.Microbiol. 7, 207–212. 26. Cruse, W.B.T., and James, M.N.G. (1972). The crystal structure
3. Gladysheva, T.B., Oden, K.L., and Rosen, B.P. (1994). Properties of the arsenate complex of dithiothreitol. Acta Crystallogr. Bof the arsenate reductase of plasmid R773. Biochemistry 33, 28, 1325–1331.7288–7293. 27. Cameron, T.S. (1976). The crystal and molecular structure of
4. Liu, J., and Rosen, B.P. (1997). Ligand interactions of the ArsC the 2:1 complex between triphenylarsine oxide and selenousarsenate reductase. J. Biol. Chem. 272, 21084–21089. acid. Acta Crystallogr. B 32, 2094–2097.
5. Shi, J., Vlamis-Gardikas, A., Aslund, F., Holmgren, A., and Ro- 28. Habash, J., et al., and Helliwell, J.R. (2000). Direct determinationsen, B.P. (1999). Reactivity of glutaredoxins 1, 2, and 3 from of the positions of the deuterium atoms of the bound water inEscherichia coli shows that glutaredoxin 2 is the primary hydro- concanavalin A by neutron Laue crystallography. Acta Crys-gen donor to ArsC-catalyzed arsenate reduction. J. Biol. Chem. tallogr. D 56, 541–550.274, 36039–36042. 29. Bouckaert, J., Dewallef, Y., Poortmans, F., Wyns, L., and Loris,
6. Ji, G., Garber, E.A.E., Armes, L.G., Chen, C.M., Fuchs, J.A., and R. (2000). The structural features of concanavalin A governingSilver, S. (1994). Arsenate reductase of Staphylococcus aureus non-proline peptide isomerization. J. Biol. Chem. 275, 19778–plasmid pI258. Biochemistry 33, 7294–7299. 19787.
7. Mukhopadhyay, R., Shi, J., and Rosen, B.P. (2000). Purification 30. Alpert, M. (2001). A touch of poison. Sci. Am. 284, 20–21.and characterization of ACR2p, the Saccharomyces cerevisiae 31. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray dif-arsenate reductase. J. Biol. Chem. 275, 21149–21157. fraction data collected in oscillation mode. In Methods in Enzy-
8. Fauman, E.B., et al., and Saper, M.A. (1998). Crystal structure mology, Volume 276, C.W. Carter, Jr. and R.M. Sweet, eds. (Newof the catalytic domain of the human cell cycle control phospha- York: Academic Press), pp. 307–326.tase, Cdc25A. Cell 93, 617–625. 32. Howard, A.J. (2001). Data processing in macromolecular crys-
9. Gladysheva, T., Liu, J., and Rosen, B.P. (1996). His-8 lowers tallography. In Crystallographic Computing 7: Proceedings fromthe pKa of the essential Cys-12 residue of the ArsC arsenate the Macromolecular Crystallographic Computing School, P.E.reductase of plasmid R773. J. Biol. Chem. 271, 33256–33260. Bourne and K.D. Watenpaugh, eds. (Oxford: Oxford University
10. Stevens, S.Y., Hu, W., Gladysheva, T., Rosen, B.P., Zuiderweg, Press), in press.E.R., and Lee, L. (1999). Secondary structure and fold homology 33. Terwilliger, T.C., and Berendzen, J. (1999). Automated MAD andof the ArsC protein from the Escherichia coli arsenic resistance MIR structure solution. Acta Crystallogr. D 55, 849–861.plasmid R773. Biochemistry 38, 10178–10186. 34. Furey, W., and Swaminathan, S. (1997). PHASES-95: a program
11. Laskowski, R.A., MacArthur, M.W., Moss, D.S., and Thornton, package for the processing and analysis of diffraction data fromJ.M. (1993). PROCHECK: a program to check the stereochemi- macromolecules. In Methods in Enzymology, Volume 277, C.W.cal quality of protein structures. J. of Appl. Crystallogr. 26, Carter, Jr. and R.M. Sweet, eds. (New York: Academic Press),283–291. pp. 590–620.
12. Luzzati, V. (1952). The statistical treatment of errors in crystal 35. Perrakis, A., Sixma, T.K., Wilson, K.S., and Lamzin, V.S. (1997).structures. Acta Crystallogr. 5, 802–810. wARP: improvement and extension of crystallographic phases
13. Levitt, M., and Chothia, C. (1976). Structural patterns in globular by weighted averaging of multiple refined dummy atomic mod-proteins. Nature 261, 552–558. els. Acta Crystallogr. D 53, 448–455.
14. Orengo, C.A., and Thornton, J.M. (1993). Alpha plus beta folds 36. Sheldrick, G.M., and Schneider, T.R. (1997). Shelxl: high-resolu-tion refinement. In Methods in Enzymology, Volume 277, C.W.revisited: some favored motifs. Structure 1, 105–120.
Arsenate Reductase1081
Carter, Jr. and R.M. Sweet, eds. (New York: Academic Press),pp. 319–343.
37. Westhead, D.R., Hatton, D.C., and Thornton, J.M. (1998). Anatlas of protein topology cartoons available on the World-WideWeb. Trends Biochem. Sci. 23, 35–36.
38. Koradi, R., Billeter, M., and Wuthrich, K. (1996). MOLMOL: aprogram for display and analysis of macromolecular structures.J. Mol. Graphics 14, 51–55.
39. Honig, B., and Nicholls, A. (1995). Classical electrostatics inbiology and chemistry. Science 268, 1144–1149.
Accession Numbers
The coordinates and structure factors of ArsC and of its complexeswith arsenate and arsenite have been deposited in the Protein DataBank under accession codes 1I9D, 1JZW, and 1J9B, respectively.