Class I RNRs have a strict requirement for oxygen, whereas the class II enzymes are indifferent to the Keywords crystal structure; flavin mononucleotide; flavodoxin; NrdI; ribonucleotide
Trang 1in oxidized and reduced states – an unusual flavodoxin
Structural biology
Renzo Johansson1, Eduard Torrents2, Daniel Lundin3, Janina Sprenger1, Margareta Sahlin3,
Britt-Marie Sjo¨berg3and Derek T Logan1
1 Department of Biochemistry and Structural Biology, Lund University, Sweden
2 Cellular Biotechnology, Institute for Bioengineering of Catalonia, Barcelona, Spain
3 Department of Molecular Biology and Functional Genomics, Stockholm University, Sweden
Introduction
The ribonucleotide reductase (RNR) system is essential
for genome replication and repair in all free-living
organisms, comprising the enzymes that carry out the
first committed step of synthesis of the building blocks
of DNA, namely the conversion of ribonucleotides to
deoxyribonucleotides RNRs have a highly diverse set
of radical generation, storage and transfer strategies, and are divided into three classes on this basis [1–3] Class I RNRs have a strict requirement for oxygen, whereas the class II enzymes are indifferent to the
Keywords
crystal structure; flavin mononucleotide;
flavodoxin; NrdI; ribonucleotide reductase
Correspondence
D T Logan or B.-M Sjo¨berg, Department
of Biochemistry and Structural Biology, Lund
University, Box 124, S-221 00 Lund;
Department of Biochemistry of Molecular
Biology and Functional Genomics,
Stockholm University, S-106 91 Stockholm,
Sweden
Fax: +46 46 222 4692; +46 8 16 64 88
Tel: +46 46 222 1443; +46 8 16 41 50
E-mail: derek.logan@mbfys.lu.se; britt-marie.
sjoberg@molbio.su.se
Website: http://www.mps.lu.se;
http://www.molbio.su.se.
Database
Structural data for oxidized and reduced
NrdI are available in the Protein Data Bank
under the accession numbers 2XOD and 2XOE
(Received 12 July 2010, revised 10 August
2010, accepted 18 August 2010)
doi:10.1111/j.1742-4658.2010.07815.x
The small flavoprotein NrdI is an essential component of the class Ib ribo-nucleotide reductase system in many bacteria NrdI interacts with the class
Ib radical generating protein NrdF It is suggested to be involved in the rescue of inactivated diferric centres or generation of active dimanganese centres in NrdF Although NrdI bears a superficial resemblance to flavo-doxin, its redox properties have been demonstrated to be strikingly differ-ent In particular, NrdI is capable of two-electron reduction, whereas flavodoxins are exclusively one-electron reductants This has been suggested
to depend on a lesser destabilization of the negatively-charged hydroqui-none state than in flavodoxins We have determined the crystal structures
of NrdI from Bacillus anthracis, the causative agent of anthrax, in the oxidized and semiquinone forms, at resolutions of 0.96 and 1.4 A˚, respec-tively These structures, coupled with analysis of all curated NrdI sequences, suggest that NrdI defines a new structural family within the flavodoxin superfamily The conformational behaviour of NrdI in response
to FMN reduction is very similar to that of flavodoxins, involving a pep-tide flip in a loop near the N5 atom of the flavin ring However, NrdI is much less negatively charged than flavodoxins, which is expected to affect its redox properties significantly Indeed, sequence analysis shows a remarkable spread in the predicted isoelectric points of NrdIs, from approximately pH 4–10 The implications of these observations for class Ib ribonucleotide reductase function are discussed
Abbreviations
MR, molecular replacement; PDB, Protein Data Bank; RNR, ribonucleotide reductase; RNRdb, Ribonucleotide Reductase Database.
Trang 2degree of aerobicity and the class III RNRs are strictly
anaerobic Class I RNRs are further divided into class
Ia and Ib based on differences in operon structure,
allosteric activity regulation and domain structure
[4–6], and recent phylogenetic studies demonstrate that
class Ib is restricted to the bacterial kingdom [7] The
class Ia RNRs use a di-iron-oxo metal centre to
gener-ate a stable tyrosyl radical in protein NrdB (R2),
which is then reversibly tranported through a
con-served radical transfer pathway to the active site on
protein NrdA (R1) when required for catalysis [1] The
class Ib homologue of NrdA is NrdE (or R1E) and
the class Ib protein NrdF (or R2F) is equivalent to
NrdB in class Ia Class Ia RNRs are allosterically
reg-ulated with regard to both overall activity and
sub-strate specificity [2,3] Class Ib RNRs are not regulated
for overall activity [5], and there is some ambiguity
concerning the nature of their metal centres The first,
manganese-containing, RNR from
Corynebacte-rium ammoniagenes[8] was later shown to be a class Ib
RNR that was also functional in an Fe-containing
form [9,10] Recently, the class Ib RNR from
Escheri-chia coli was also shown to be enzymatically active,
both as a Mn- and as a Fe-containing enzyme [11,12]
NrdI is a small flavodoxin-like protein whose gene is
found in all organisms where class Ib ribonucleotide
reductases (RNR) are present It was first identified
in the mid-1990s as part of the nrdEF gene cluster in
Escherichia coli and Salmonella typhimurium [13] In
these enterobacteria, it was found to code for a small
protein of 136 amino acids with a molecular mass of
15.3 kDa, normally forming an nrdHIEF operon
struc-ture Subsequently, NrdI was shown to be involved in
the activity of class Ib RNR [14], having a stimulatory
effect on NrdEF activity A decade later, NrdI was
demonstrated to be essential for class Ib RNR activity
in Streptococcus pyogenes [15], which contains two
redundant and simultaneously expressed class Ib gene
clusters: NrdHEF and NrdF*I*E* The latter system
was not active in enzymatic assays in vitro but, in
com-plementation experiments, NrdF*I*E* was able to
restore lost class Ib RNR activity in a
temperature-sensitive E coli strain This led to the first proposal
that NrdI could be essential for maintenance of class
Ib RNR activity [15]
Recently, a thorough investigation of the potential
roles of NrdI in function of class Ib RNR in E coli
has been carried out Two non-mutually exclusive
hypotheses have been proposed [11,12] In the first
sce-nario, NrdI is suggested to be involved in rescue of
active NrdF proteins whose FeIII-FeIII-Tyr centres
have been reduced by one electron to produce the
inactive FeIII-FeIII-Tyr (met) form This rescue would
be effected by the injection of two electrons in rapid succession into the FeIII-FeIII centre to produce a reduced FeII-FeII centre, which would then react with molecular oxygen according to the well-characterized assembly pathway [16] to regenerate active NrdF Importantly, NrdI was shown to differ significantly in its redox properties from previously characterized Flds, which typically alter the redox potentials for the ox⁄ sq and sq⁄ hq couples of their FMN cofactors in such a way that the flavin group becomes a one-electron reductant Flds normally stabilize near stoichiometric amounts of the neutral sq form of FMN by shifting the redox couple Esq⁄ hq from)172 mV for free FMN [17] to between )370 and )450 mV for the bound form [18] and Eox⁄ sq from)238 mV for the free form
to between)50 and )220 mV for the bound form [18]
By contrast, the protein environment of E coli NrdI maintains the redox potentials of the two couples at very similar values, namely Eox ⁄ sq=)264 mV and
Esq ⁄ hq=)255 mV, respectively [11] In this way, FMN bound to E coli NrdI may be made capable of injecting two electrons in rapid succession into NrdF The interaction with NrdI was also shown to be spe-cific for NrdF because no effect was seen on the class
Ia NrdB protein
Given the ambiguity as to the nature of the redox-active metal species in class Ib RNRs, an alternative scenario has been investigated in which NrdI is involved in the assembly of an active MnIII-MnIII-Tyr cofactor in E coli NrdF [12] The two proteins were found to form a tight complex during nickel–nitrilotri-acetic acid affinity chromatography Aerobic incuba-tion of fully-reduced NrdI with MnII-reconstituted NrdF led to the formation of active NrdF with 0.25 tyrosyl radicals per dimer This was suggested to occur through the reaction of the MnII centre with two equivalents of HO2) produced by two successive one-electron reductions of O2 by NrdIhq bound to NrdF
By contrast, aerobic incubation of NrdF reconstituted with FeII in the presence of NrdI led to a species with only 13% of the specific activity, although it had 0.2 tyrosyl radicals per dimer It was thus proposed that NrdI is involved in the assembly of a MnIII-MnIII-Tyr cofactor in E coli and that this is the true cofactor
in vivo However, this hypothesis does not exclude the possibility that the cofactor is FeIII–FeIII-Tyr under some growth conditions, and that NrdI could be involved in maintenance of the cofactor under these circumstances
E coli expresses a class Ia RNR during aerobiosis that cannot be substituted for by its chromosomally encoded class Ib RNR This is in contrast to many bacterial species that are dependent upon their class Ib
Trang 3RNR for aerobic growth It is therefore of interest to
study the structural and functional properties of class
Ib RNR from organisms such as the Bacillus cereus
group and, in the present study, we present the
struc-ture of NrdI from the human pathogen Bacillus
an-thracis, baNrdI [19] Although the baNrdI protein is
highly similar to the B cereus protein recently reported
in partially photoreduced forms [20], the previous
study concentrated on the structural effects on the
flavin of photoreduction during data collection Given
the functional disparities between NrdI and normal
Flds, it is important to study the structural basis for
NrdI function In the present study, we present the
crystal structures of baNrdI in the oxidized and
chemi-cally-reduced semiquinone forms NrdI is shown to
have an unusually compact Fld fold, defining a new
structural class within the Fld family The electrostatic
potential surface of baNrdI is shown to be strikingly
different to that of Flds A bioinformatic analysis of a
large number of NrdI sequences shows that this effect
is general; indeed, on average, NrdI proteins are
signif-icantly basic and their electrostatic and redox
proper-ties can be expected to vary to a surprising degree
Results
The crystal structure of baNrdI has been solved to
0.96 A˚ resolution with the FMN cofactor in its
oxi-dized state and to 1.4 A˚ with chemically-reduced
FMN B anthracis NrdI is unambiguously a member
of the Fld superfamily The fold consists of a
five-stranded parallel b-sheet flanked by two a-helices on
each side (Fig 1) However, a search using the dali
server [21] indicates that NrdI is a structural outlier
within the Fld family The closest structural neighbour
is Fld from Desulfovibrio desulfuricans [dsFld; Protein Data Bank (PDB) code: 3F6R] [22], with a rmsd of 2.4 A˚ for 113 alignable Ca atoms out of 117 in baNrdI and 147 in dsFld Very similar statistics are obtained for a wide variety of Flds from diverse organisms, whether short-chain or long-chain How-ever, B anthracis NrdI is 30 residues shorter than a typical short chain Fld and displays a more compact fold The truncations occur principally on the side of baNrdI furthest from the flavin binding site Helix a1
is shorter by five residues (seven versus 13), strand b2
by four residues (three versus seven) and strand b3 by four residues (five versus nine) (Fig 1 and Table S1)
In addition, the loops between a1 and b2; a3 and b4; and a4 and b5 are shortened compared to dsFld Analysis of 199 NrdI sequences extracted from the Ribonucleotide Reductase Database (RNRdb) [7] (Fig S1) shows that NrdIs are fairly homogeneous in length, with a median value of 141 and a SD of 13 There is no division into short- and long-chain vari-ants as there is for Fld The minimum structural core, with shortest loops, is apparently represented by the Corynebacterium striatum sequence at 109 amino acids (Fig S1) Variations in length are essentially limited to the termini and the loops between a1 and b2, between b2 and b3, and the loop (residues 42–49) that interacts with the flavin moiety of FMN The first two variable loops are distant from FMN The variable length of this ‘‘40s loop’’ is discussed below By contrast, beyond the FMN-binding loop, the NrdI structure is extremely well conserved (Fig S1), with no significant insertions or deletions
The electron density for the FMN cofactor is excel-lent in both structures, and the high resolution of the oxidized form allows confirmation of the protonation
40s loop
70s loop
β1
α1
α2
α4
50s loop
90s loop
β1
α1
α2
α4
β2 β3
β4 β5
β2 β3 β4
β5
Fig 1 (A) Overall structure of NrdI The
car-toon is coloured as a rainbow from blue at
the N-terminus to red at the C-terminus to
emphasize the topology For clarity, helix 1
is semi-transparent The FMN cofactor is
shown in a stick representation Lengths of
the secondary structure elements are given
in Table S1 (B) Structure of the most
struc-turally homologous standard flavodoxin, the
short chain protein from D desulfuricans
(PDB ID 3F6R), for comparison The
repre-sentation is as shown in (A) Prepared using
PYMOL
Trang 4state In electron density maps calculated without
inclusion of explicit hydrogen atoms on the cofactor,
the difference electron density can clearly be seen at
2.5 r for many of the aliphatic hydrogen atoms of the
cofactor (Fig 2A) By contrast, no hydrogen atoms
can be seen on N5, confirming the oxidized state of
FMN The phosphate group of FMN is bound by the
N-terminus of a1 and the preceding P-loop The flavin moiety of FMN binds in a pocket formed by loops at the C-terminal ends of b-strands 3 and 4 (loops 42–49 and 71–79, respectively), known in Fld as the W and
Y loops, or the 50s and 90s loops [23] We refer to these as the 40s and 70s loops, respectively, in baNrdI The flavin moiety is sandwiched between Trp74 on one face and the side chain of Thr42 and the main chain atoms of residues 42–44 on the other (Fig 2A) The isoalloxazine ring is completely buried and anchored through seven hydrogen bonds between its carbonyl (O2, O4) and amide (N3) groups and main-chain car-bonyl and amide groups in the 40s and 70s loops By contrast, the dimethylbenzene ring is solvent-exposed The flavin binding pocket is capped by Phe45, whose side chain lies perpendicular to the flavin moiety and also makes an edge-on interaction to the stacking Trp74
The flavin environment in NrdI is considerably less negatively charged than in Fld For example, Clostrid-ium beijerinckii Fld (cbFld) has a net charge of )14, whereas in baNrdI it is only )4 (in the present study the net protein charges always refer to the protein component only) Figure 2B shows the preponderance
of acidic side chains in the vicinity of the flavin in cbFld, including two in the 50s loop itself There are three acidic residues within 6 A˚, and a further four within 10 A˚ Overall, 26 negative charges are compen-sated by only 12 positive charges By contrast, in baNrdI,
18 negative charges are compensated for by 14 positive ones The closest of these to the flavin, Asp76
in the 70s loop and Asp83 in helix a4, are 9 A˚ distant from the flavin (Fig 2A) This has a remarkable effect
on the electrostatic energy landscape of NrdI com-pared to Fld (Fig 3) To test the generality of this observation, we carried out an analysis of the length, amino acid composition and calculated isoelectric point of a representative set of 199 NrdI sequences extracted from the RNRdb A parallel analysis of 38 manually reviewed flavodoxin sequences from the UniRef100 database (http://www.uniprot.org) was performed for comparison, confirming that NrdI differs strongly from flavodoxins with respect to pI The median pI for NrdI sequences is 9.0 with a SD of 1.7 (Fig 4A) However, the spread in values is wide, ranging from 4.2 for Eubacterium biforme to 10.4 for NrdI1 of S pyogenes M1 Net charges vary remarkably, from)15 to +15 The pI distribution is approximately bimodal, with a major peak at pH 9.0–9.5 and a broader peak at pH 5.0–5.5 By contrast, the 38 representative flavodoxin sequences that have been analyzed are much more homogeneous in pI: the median value is 4.5 with a
SD of only 0.6 (Fig 4B)
70s loop
90s loop 40s loop
A
B
50s loop
W74 2.9 3.4
3.1
3.2 2.8 D83
E98 D105
F45 T43
α1
D76
G44
W90
D92
D58 D59
D57 M56
D98
E101 E120
E62
E63 E65
Fig 2 (A) Electron density for the FMN cofactor and the 40s loop.
The grey mesh shows a r A -weighted 2|F o | ) |F c | map to 0.96 A ˚
cal-culated using SHELXL and contoured at 1.2 r around the FMN
cofac-tor and the 40s loop Residues within 4 A ˚ of FMN that make
interactions with it are shown as thin lines Particularly relevant
side chains, including all acidic side chains in the view, are shown
as sticks Hydrogen bonds are shown as dashed lines The green
mesh shows a similarly calculated F o ) F c map to 1.1 A ˚ resolution
contoured at 2.5 r For calculation of this map, hydrogen atoms
were included for the protein but omitted from the cofactor (B)
The flavodoxin from C beijerinckii in its oxidized state (PDB code:
4NLL) in the same orientation as baNrdI shown in (A) The
repre-sentation is identical to that shown in (A) The overall details of
FMN binding are very similar to baNrdI; however, note the
prepon-derance of acidic residues in the vicinity of the FMN binding site.
Prepared using PYMOL
Trang 5Reduced baNrdI-FMN
The crystal structure of fully reduced baNrdI-FMN
was obtained by chemical reduction of crystals of
oxi-dized baNrdI using 500 mm sodium dithionite The
largest conformational change in the protein is a pep-tide flip between residues 44 and 45, resulting in the orientation of the Gly44 carbonyl group towards N5, which is protonated in the neutral sq radical form (Fig 5A) This peptide flip is accompanied by a slight rearrangement of the whole loop from Thr42 to Asn47 Interestingly, Thr42 undergoes a small shift, and difference density appears at a level of approxi-mately 4 r between its side chain and the flavin ring (Fig 5B) This density is also present at approximately
6 r in the maps of the chemically-reduced B cereus NrdI [20] as generated by the Electron Density Server [24] (http://eds.bmc.uu.se/eds/), although the authors responsible for this entry did not interpret it The density is too close to Thr42 (1.7 A˚ to Oc) to be a water molecule, although it could be a loosely coordi-nated metal ion There is also a rotamer change in Thr43 and a slight movement of Phe45 upon flavin reduction This confirms that the conformational response of NrdI to reduction is very similar to that observed in flavodoxins [18,25–27]
Flavin photoreduction as a result of X-ray exposure
Røhr et al [20] recently noted the need to take into consideration the effects of radiation damage on the geometry of flavin cofactors when analysing structures where data were collected using synchrotron X-ray sources A significant distortion of the flavin was noted
in both NrdIox and NrdIsq from B cereus after esti-mated radiation doses of 9 and 10 MGy, respectively Quantum mechanics simulations of the flavin geometry
in the protein context coupled to resonance Raman experiments on the crystals suggested that both NrdIox and NrdIsq had been reduced by one electron during X-ray exposure, such that the flavins were now in the FMN) and FMNH) states respectively With this in mind, we investigated the effect of radiation damage in the almost identical B anthracis system Using suitable parameters for the I911-3 (NrdIox) and I911-5 (NrdIsq) beamlines at MAX-lab, Lund, Sweden), we arrived at a dose estimate of approximately 2–4 MGy for baNrdIox and 5–6 MGy for baNrdIsq, using the software rad-dose [28] In the last cycle of refinement, no restraints were used in the refinement of the FMN geometry for oxidized baNrdI The ‘butterfly angle’ between the fla-vin ring planes is 5.7, which compares well with the value of 4.8 reported for the oxidized B cereus protein after photoreduction, indicating that one-electron reduction has also occurred in baNrdI The resolution
of the sq form was not sufficiently high to allow unre-strained refinement, and so a similar comparison is not
0
3.50–3.99 4.00–4.49 4.50–4.99 5.00–5.49 5.50–5.99 6.00–6.49 6.50–6.99 7.00–7.49 7.50–7.99 8.00–8.49 8.50–8.99 9.00–9.49 9.50–9.99
10.00–10.49
5
10
15
20
25
0
10
20
30
40
50
B
Fig 4 Histograms of the distributions of predicted isoelectric point
for (A) NrdI and (B) Fld sequences The sequences are colour-coded
according to the phylum of the source organisms The pI groups to
which the NrdIs with determined structures belong are labelled:
Bant, B anthracis; Bcer, B cereus; Bsub, B subtilis; Ecol, E coli.
Produced using Google Docs (http://docs.google.com/).
Fig 3 Electrostatic potentials for (A) baNrdI and (B) C beijerinckii
flavodoxin The potentials at the solvent accessible surface were
calculated using APBS [53] and mapped onto the molecular surface
using PYMOL The colour scale in both panels runs from deep red at
)5 kTe )1to blue at +5 kTe)1 The FMN molecule is shown in a
space-filling representation The molecular surface is
semi-transpar-ent and a grey cartoon of each molecule is shown for orisemi-transpar-entation.
The direction of view is into the side of the flavin plane Prepared
using PYMOL
Trang 6meaningful However, distortion of the flavin geometry
as a result of accumulated photoreduction during data
collection can be seen in the anisotropic B-factors of
the flavin atoms in both oxidized and sq forms (Fig 6)
With the isoallazine ring being fixed by its interactions
with the protein, the flavin distortion is concentrated
on the dimethylbenzene ring, which has greater
free-dom to distort from planar geometry
Discussion
The crystal structures of NrdI from B anthracis have
been solved in two functional states: in complexes with
oxidized and semiquinone FMN The structures reveal
that NrdI is unambiguously a member of the Fld
superfamily, although it has the most compact fold of
a Fld seen to date, being shorter than the average
short-chain Fld It can thus be considered to define a
new family within the Fld superfamily The NrdI
fam-ily is not divided into short- and long-chain
subfami-lies: the region in and around the final strand b5,
where the insertion defining the long-chain Flds
occurs, is extremely highly conserved with regard to
secondary structure in NrdI sequences (Fig S1)
Despite being an outlier in the Fld family, the
FMN-binding regions are more conserved than the rest of
the structure
Correctly-folded baNrdI for functional and
struc-tural studies could only be obtained by including
FMN in the growth medium (in our case LB medium)
at a concentration of 60 lm, when overexpressed in
E coli In the absence of FMN NrdI was misfolded and produced irreversibly in inclusion bodies Without FMN, NrdI from S typhimurium, C ammoniagenes,
B anthracis and Deinococcus radiodurans also form inclusion bodies during heterologous overexpression in
E coli (E Torrents, unpublished results) The observa-tion is also in agreement with previously published studies on E coli NrdI, in which significant quantities
of functional protein could only be obtained by refold-ing from inclusion bodies in the presence of FMN [11]
A general requirement for FMN in NrdI folding would contrast with the behaviour of traditional Flds The dependence of Fld folding on FMN has been studied, and the binding of FMN to native apo-Fld was found to constitute the last step [29] The autono-mous formation of native apo-Fld is essential during holo-Fld folding, and FMN does not act as a nucle-ation site for folding FMN can be removed from Fld
by acid treatment [30], despite affinity in the sub-nanomolar range [29], also resulting in a stable apo-protein Conformational differences between apo- and holo-Flds are small and confined to the 50s and 90s loops [31,32] Further experiments are required to establish whether NrdI has a general requirement for FMN for correct folding during overexpression
NrdI is remarkably less negatively charged than normal flavodoxins
The major function of the protein environment in flavodoxins is modification of the redox potentials
W74
F45 G44
2.8 3.4 2.8
T43
T42
O2 FMN
S69 3.2
2.1 2.4
Fig 5 (A) Conformational change upon chemical reduction of baNrdI to the semiquinone state The grey mesh shows a r A -weighted 2|Fo| ) |F c | map to 1.4 A ˚ resolution calculated using REFMAC5 and contoured at 1.2 r around the FMN cofactor and the 40s loop Reduced baNrdI is shown in green and the 40s loop of the oxidized form is shown in blue for comparison Hydrogen bonds between FMN and the 40s loop are shown as dashed lines (B) The strong difference density that appears between the flavin and Thr42 in the crystal structure of reduced baNrdI, which is also present in bcNrdI (Fig S2) A 2|F o | ) F c | map contoured at 1.0 r and an |F o | ) |F c | map contoured at 3.0 r are shown in grey and green, respectively The height of the difference map peak is approximately 4 r A marker atom has been placed in the electron density to show the distances to potential coordinating atoms in the vicinity Prepared using PYMOL
Trang 7ox⁄ sq and sq ⁄ hq couples from the rather similar values
of)172 mV and )238 mV, respectively, in the free
fla-vin in FMN [17] to the widely different values of)50
to 260 mV and)370 to )450 mV, respectively, for the
bound form [18] Thereby, flavodoxins are made into
effective one-electron donors The effects on redox
potentials occur primarily through (a) stabilization of
the sq form via a hydrogen bond between the N5 atom
and a carbonyl group in the 50s loop [18,25,26,33–36]
and (b) destabilization of the negatively-charged hq
form (FMNH)) through the lack of solvation in the
protein environment coupled to highly negative
elec-trostatic field, which reduces the protein’s association
rate with the hq form of FMN [18,36,37] Replacement
of acidic residues by neutral or positively-charged ones
tends to increase the sq⁄ hq potential [18,38,39],
although theoretical studies have shown that
compen-satory (de)protonation effects on other charged
resi-dues can make the effect difficult to predict [37] By
contrast, modification of the conformational properties
of the 50s loop affects the ox⁄ sq potential to a greater
degree [26,35,36]
The remarkably similar redox potentials of the
ox⁄ sq and sq ⁄ hq couples in E coli NrdI (ecNrdI) have
been attributed to a lesser destabilization of the sq form, FMNH), than what is normally the case in flavodoxins [11] The structure of baNrdI confirms this hypothesis, and our bioinformatic analysis extends it, with few exceptions, to the whole NrdI family The conformational changes observed between the oxidized and reduced states are limited to a peptide flip between residues 44 and 45 and a small rearrangement of the 40s loop The high similarity of these changes to those observed in Fld from several species strongly suggests that the unusual redox potentials of NrdI are governed more by protein electrostatics than by specific hydro-gen bonds or other direct interactions with the flavin However, the wide spread in predicted pI values for NrdI sequences is unexpected The data point to the interesting possibility of two distinct functional groups with acidic and basic characters, respectively, which obviously will have quite different effects on the poten-tials of the FMN redox couples Alternatively, the charge variation in NrdI may be correlated with a sim-ilar variation in the electrostatic properties of the respective NrdF proteins, in particular at the interac-tion area The reason for the wide spread in predicted
pI is not obvious Figure 4A shows the distribution colour-coded by taxonomy It can be seen that NrdIs from the a- and c-proteobacteria belong almost exclu-sively to the high-pI group, whereas the firmicutes and actinobacteria occupy a wide range of pI values Sev-eral organisms in the latter two groups, including some Bacillusspecies, encode more than one class Ib operon, whereas the proteobacteria all encode only one In general, the NrdI pI values differ substantially within organisms encoding two different nrdI genes A phylo-genetic tree of 91 representative NrdI sequences (Fig S3) shows that the NrdI phylogeny is not signifi-cantly different from that based on NrdF sequences [19] or on 16S rRNA, although actinobacteria and fir-micutes, with two class Ib operons, generally have NrdIs divided into two separate clusters However, the presence of a sequence in a genome provides no infor-mation regarding if and when the protein is expressed, and so further experiments are required to establish the reason for the wide spread in electrostatic proper-ties in an otherwise structurally conserved family
E coli NrdI, with pI = 9.4, belongs to the major group of NrdI sequences with basic character, whereas baNrdI, with pI = 5.4, belongs to the minor, acidic group These proteins also differ in their ability to sta-bilize an FMN sq radical: a maximum of 28% can be detected in ecNrdI, whereas, in baNrdI, the amount is
up to 60% (M Sahlin & B.-M Sjo¨berg, unpublished results) Interestingly, this correlates with the predicted pIs: that of baNrdI (net charge )4) is much closer to
A
B
Fig 6 Depiction of the anisotropic movements of atoms in the
FMN molecule as represented by their anisotropic B-factors (A)
NrdI ox , 0.96 A ˚ resolution, refined using SHELXL (B) NrdI sq , 1.4 A ˚ ,
refined using REFMAC5 The anisotropic B-factors are represented by
thermal ellipsoids at 50% probability The B-factors are coloured
from dark blue at 5.0 A˚2to bright red at 12.8 A˚2in (A) and from
7.7 A ˚ 2 to 18.7 A ˚ 2 using the same colour scheme in (B) Prepared
using PYMOL
Trang 8that of a normal flavodoxin than that of ecNrdI (net
charge +4)
In E coli NrdI, a neutral sq radical is produced when
NrdI is titrated anaerobically with dithionite [11] In the
presence of NrdF, whether in the apo- or
Mn-contain-ing forms, an anionic sq radical is produced instead
This behaviour was described as being potentially more
similar to that of flavoprotein oxidases than flavodoxins
[12] However, the stabilization of an anionic sq form
would be favoured by the presence of a
positively-charged protein residue in the vicinity of the N1-C2-O2
atoms [40], as in glycolate oxidase [41] No such residue
can be found in baNrdI and, indeed, the formation of
anionic sq should be disfavoured by the overall negative
charge of the protein This points to influence of NrdF
on the flavin environment in the NrdF⁄ I complex,
although the flavin N1-C2-O2 atoms would remain
inaccessible to residues from NrdF in the complex,
unless NrdF induces a conformational change in the 90s
loop to expose the flavin We have identified the
possi-bility that a metal ion is trapped close to the N1-C2-O2
locus in chemically-reduced baNrdI and bcNrdI, in the
space made available by the conformational
rearrange-ment in the 40s loop induced by the peptide flip of
Gly44 (Fig 5B) This may help to compensate the
negative charge in an anionic sq form
The FMN-binding loop is a major site of
sequence variation in NrdI
NrdI sequences vary in length from 103 to 188 amino
acids Intriguingly, one of the most variable loops in
NrdI is the one that interacts with the flavin ring of
FMN, namely the 40s loop In the baNrdI structure, the
loop extends from Thr42 to Pro49 (i.e eight residues)
At the other extreme, in mycobacteria, the loop is up to
15 residues longer, containing a high proportion of Pro
and Gly (Fig S1) The role of this highly variable length
in a critical area of the structure is not clear Short- and
long-chain flavodoxins differ in the presence or absence
of a 20 residues loop that splits the fifth b-strand [23],
although this is not the case in NrdI The 50s loop is
slightly longer in long-chain Flds than in short-chain
ones, and it has been proposed that the inserted loop
stabilizes the longer 50s loop [42] However, an extended
40s loop in NrdI does not appear to be correlated with
any other insertion It might be hypothesized that the
extended loop contributes to increased affinity for
NrdI’s interaction partner NrdF because, although the
KD for the baNrdI⁄ baNrdF complex is only 50 lm,
the affinity of the E coli complex (with a loop longer
by four residues) is so high that interactions are
maintained during nickel–nitrilotriacetic acid affinity
chromatography [12] The capping of the FMN bind-ing site by Phe45 appears to be specific for Bacillus and a few other species, and is correlated with the exceptionally short loop found in baNrdI This residue
is Gly in almost all other NrdI sequences Thus, the extended loop may close off the flavin binding site by folding back over the core of the protein
When the present paper was in revision, the crystal structures of the complex between NrdI and NrdF from E coli in the oxidized and hq forms were pub-lished [43] This work confirmed our prediction that the variable 40s loop will contribute to the different affinities of NrdI for NrdF in various organisms The equivalent loop of NrdI in the E coli complex forms
an important part of the molecular interface and undergoes more significant conformational changes upon reduction than in the Bacillus species, although this may be influenced by its interactions with NrdF
In summary, we have identified the NrdI protein as a structural outlier within the flavodoxin family, having a significantly more compact fold, although the structure close to the flavin binding site is more conserved The 40s loop is an important site of sequence variation in the NrdI family A very wide distribution in predicted
pI values has been identified, which may imply different functional roles for NrdI in different organisms
Materials and methods
Cloning of the nrdI gene The B anthracis nrdI was amplified by PCR from strain Sterne 7700 pXO1)⁄ pXO2) genomic DNA as described previously [19] using BanrdIup 5¢-ACATATGTTAGTTG CCTATGATTCTATG-3¢ and BanrdIlw 5¢-AAAGCTTAT TCAGTTCAATGTGTC-3¢, as a forward and reverse prim-ers containing NdeI and HindIII restriction sites, respectively (underlined) The PCR product was cloned in the pGEM-T easy vector (Promega, Madison, WI, USA) After digestion with NdeI and HindIII, the nrdI fragment (380 bp) was ligated into pET22b generating plasmid pETS153
Expression and purification
E coli Rosetta(DE3) cells (Novagen, Madison, WI, USA) containing pETS153 were grown in LB medium (Difco, Franklin Lakes, NJ, USA) at 37C with 100 lgÆmL)1 ampicillin, 17 lgÆmL)1 chloramphenicol and 60 lm FMN (Sigma, St Louis, MO, USA) until a A550 of 0.5 was reached, induced with 1 mm isopropyl thio-b-d-galactoside for 4 h, collected by centrifugation, and disrupted in an X-Press (BioX AB, Gothenburg, Sweden) in buffer 50 mm Tris-HCl (pH 7.6), 30 mm KCl and protease inhibitors
Trang 9(Roche Applied Science, Basel, Switzerland) All the
pro-tein purification steps were carried out at 4C After
high-speed centrifugation, the protein concentration was
adjusted to 10 mgÆmL)1 and the supernatant solution was
first precipitated with streptomycin sulfate (final
concentra-tion 1%) and, after a second centrifugaconcentra-tion, with solid
ammonium sulfate to 45% saturation After centrifugation,
the precipitate was dissolved in buffer A (50 mm Tris-HCl,
pH 7.6, and 30 mm KCl) and desalted by dialysis against
2 L of buffer A for 16 h The dialyzed solution was diluted
with buffer A to 6 mgÆmL)1 protein concentration and
loaded on HiLoad 16⁄ 10 Q-Sepharose High Performance
column (GE Healthcare, Milwaukee, WI, USA) on a
Bio-Logic DuoFlow System fast protein liquid chromatography
instrument (Bio-Rad, Hercules, CA, USA) previously
equil-ibrated with ten volumes of buffer A NrdI protein was
eluted with a linear gradient of KCl (30–400 mm,
3 mLÆmin)1) in buffer A Fractions containing the NrdI
protein were pooled, concentrated using Centricon-10
(Mil-lipore, Billerica, MA, USA) and loaded at 0.5 mLÆmin)1
on a 24-ml Superdex-75 column equilibrated and eluted
with buffer 50 mm Tris-HCl (pH 7.6) and 200 mm KCl
Each fraction was analyzed by PhastGel electrophoresis
(GE Healthcare), and fractions with the highest purity
(strong yellow colour) were pooled, concentrated using
Centricon-10 (Millipore), and finally freed from KCl by
washing with buffer A
Crystallization and data collection
The protein solution used for crystallization was at
8 mgÆmL)1 in 50 mm Tris-HCl (pH 7.6) Screening for
ini-tial crystallization conditions was performed at the
crystalli-zation facility at MAX-lab The PACT Premier and
JCSG+ screens (Molecular Dimensions Ltd, Newmarket,
UK) were carried out at 20C in 100 + 100 nL drops in
Greiner low-profile 96-well plates (Greiner Bio-One GmbH,
Frickenhausen, Germany) A hit in condition 55 (E7) of the
JCSG+ screen was refined using manual setups The
crys-tal used for data collection on the oxidized form was grown
from a drop consisting of 1 lL of protein solution and
1 lL of a reservoir solution consisting of 10% (v⁄ v)
2-pro-panol, 0.2 m Zn acetate, 0.1 m Na cacodylate buffer (pH
6.5) Crystals appeared after one day and reached full size
after approximately 1 week The crystal used for data
col-lection was approximately 0.2· 0.1 · 0.1 mm in size
Crys-tals were taken directly from the drop and flash-cooled in
the gas stream from an Oxford Diffraction CryoJet (Oxford
Diffraction Ltd, Oxford, UK) Data on the oxidized form
were collected to 0.96 A˚ resolution at station I911-3 of the
MAX-II synchrotron (MAX-lab), using a 225 mm
marMo-saic CCD detector (Rayonix LLC, Evanston, IL, USA)
The X-ray wavelength was 0.7300 A˚ Data were collected,
in two passes, to 1.9 and 0.96 A˚, respectively The crystal
belonged to space group P212121 with cell dimensions as
shown in Table 1 For determination of the structure of reduced NrdI, a crystal was soaked in a solution consisting
of 500 mm sodium dithionite in crystallization mother liquor for 10 min before flash cooling During this time, the crystal colour changed from yellow to dark blue Data were collected at station I911-5 of the MAX-II synchrotron, with
an X-ray wavelength of 0.9789 A˚, in two passes, to 1.8 and 1.4 A˚, respectively Diffraction data were integrated using xds and scaled using xscale [44] Data were further pro-cessed using software from the ccp4 suite [45] The crystals contain one NrdI molecule in the asymmetric unit, resulting
in a Matthews volume of 2.05 A˚3ÆDa)1 and a solvent con-tent of 40.0%
Structure solution and refinement The structure of oxidized baNrdI was solved by molecular replacement (MR) using the unpublished coordinates of
Table 1 Data and structure quality statistics Figures in parenthe-ses refer to the highest resolution bin.
Unit cell dimensions (A ˚ ) a = 42.80,
b = 45.62,
c = 56.33
a = 42.83,
b = 45.26,
c = 55.66 Data collection
(0.98–0.96)
23.7–1.4 (1.44–1.4)
Refinement
(1.0–0.96)
26.7–1.4 (1.46–1.40)
1 · Zn
1 · cacodylate,
3 · Zn Mean isotropic B-factor (A˚2 ) 13.4 (protein) 14.3 (protein),
34.2 (water) 39.2 (water) Rmsd from ideal geometry
Ramachandran plot quality
a From DANG restraints in SHELXL
Trang 10NrdI from Bacillus subtilis (PDB code: 1RLJ), which has
48% sequence identity to baNrdI The automated MR
pipe-line mrbump [46] was used The best search model was
gen-erated by truncation of nonconserved side chains using
molrep from the ccp4 suite and the solution was found
using molrep as the MR search engine The MR solution
had an R-factor of 36.0% and a free R-factor of 37.3%
(cal-culated using 3% of the data) The model was rebuilt by
iterated rounds of model building in coot [47] and
ment in refmac5 [48] The high-resolution limit for
refine-ment was increased gradually from 2.0 A˚ to 1.1 A˚
Anisotropic B-factors were introduced at 1.5 A˚ After
con-vergence of refinement in refmac5, Rmodel and Rfree were
15.6% and 18.1%, respectively At this point, the refinement
software was switched to shelxl [49] Restraints for FMN
were generated from coordinates in the HIC-UP database
[50] using the prodrg server [51] To ensure convergence,
the resolution was reduced to 1.8 A˚ and gradually increased
to 0.96 A˚ Riding hydrogen atoms were used throughout
Anisotropic B-factors were introduced at 1.6 A˚ resolution
After shelxl refinement, Rmodel and Rfree were 13.5% and
16.3%, respectively Water molecules were introduced in
peaks over 4.0 r in the difference map fulfilling reasonable
distance and hydrogen bonding criteria to protein residues or
other water molecules Refined water molecules were removed
if they had excessively high B-factors or electron density in
2Fo) Fcmaps under 1.0 r
The structure of reduced baNrdI was solved by direct
refinement of the oxidized structure against the dataset
from a reduced crystal After the first round of refinement,
strong difference electron density was observed, indicating
a flip in the peptide bond between residues 44 and 45 The
solvent structure was modelled according to the same
crite-ria as for the oxidized protein and the coordinates were
refined using refmac5 [48] and phenix.refine [52]
Electrostatic potential calculations
Electrostatic potentials were calculated using the apbs
plu-gin [53] to pymol (http://www.pymol.org) using default
parameters throughout
Sequence analysis
A set of 199 unique NrdI sequences was extracted from the
RNRdb database: sequences annotated as containing only
an NrdI fragment were immediately discarded Sequences
from different strains of the same organism were then
removed to decrease redundancy In cases where different
strains contained either one or two NrdI sequences, the
strain containing two sequences was retained The remaining
sequences were aligned using clc sequence viewer 6.3
(CLC Bio, A˚rhus, Denmark; http://www.clcbio.com) Four
sequences that were not marked as fragments in the
data-base, but which were evidently too short because they
lacked the first a-helix and b-strand, were then removed Isoelectric points, sequence lengths and amino acid compo-sitions were calculated and tabulated using clc sequence viewer 6.3 and overall statistics calculated using Microsoft Excel (Microsoft Corp., Redmond, CA, USA) For compar-ison, an analysis of 38 flavodoxin sequences extracted from the UniProt100 database (http://www.uniprot.org) was also carried out
Phylogenetic reconstruction From the 277 unique NrdI sequences in the RNRdb [7], 91 representative sequences were chosen and aligned using probcons, version 1.10 [54] The sequences were chosen to represent the full diversity of NrdI sequences except for highly divergent sequences A maximum likelihood tree was estimated from 100 well-aligned positions using phyml, ver-sion 3.0, the LG substitution model and four gamma cate-gories [55] Branch confidence was calculated using the SH-like algorithm [56]
Estimation of isoelectric points
pI values of protein sequences were estimated using the
pI⁄ Mw tool at the expasy web server (http://www.expasy ch/tools/pi_tool.html)
Acknowledgements
This work was supported by grants from the Swedish Research Council to B.M.S and D.L E.T was supported by grants from the Spanish Ministerio de Ciencia e Innovacio´n (PI081062), the CONSOLIDER (CSD2008-00013) and ERANET Pathogenomics.We wish to thank Maria Ha˚kansson for help at the MAX-lab crystallization facility and the staff at beamline I911 at MAX-lab for assistance with data collection
We thank Ilya Borovok for stimulating discussions
References
1 Eklund H, Uhlin U, Fa¨rnega˚rdh M, Logan DT & Nordlund P (2001) Structure and function of the radical enzyme ribonucleotide reductase Prog Biophys Mol Biol
77, 177–268
2 Nordlund P & Reichard P (2006) Ribonucleotide reduc-tases Annu Rev Biochem 75, 681–706
3 Torrents E, Sahlin M & Sjo¨berg BM (2008) The ribonu-ceotide reductase family: genetics and genomics In Ribonucleotide Reductase(Andersson KK ed), pp 17–77 Nova Science Publishers, New York
4 Jordan A, Pontis E, Aslund F, Hellman U, Gibert I & Reichard P (1996) The ribonucleotide reductase system
of Lactococcus lactis Characterization of an NrdEF