Our results show that sulfate may, unexpectedly, induce full closure of catalytic site loops whereas, conversely, binding of inhibitor phosphonoacetohydroxamate may leave open a tunnel f
Trang 1revealed by X-ray crystallography and molecular dynamics Marcos V de A S Navarro1,*,‡, Sandra M Gomes Dias1,*,§, Luciane V Mello2,3,*,
Maria T da Silva Giotto1,†, Sabine Gavalda4,–, Casimir Blonski4, Richard C Garratt1
and Daniel J Rigden2
1 Instituto de Fı´sica de Sa˜o Carlos, Universidade de Sa˜o Paulo, Sa˜o Carlos SP, Brazil
2 School of Biological Sciences, University of Liverpool, UK
3 Northwest Institute for Bio-Health Informatics, University of Liverpool, UK
4 Groupe de Chimie Organique Biologique, Universite´ Paul Sabatier, Toulouse, France
Enolase (2-phospho-d-glycerate hydrolase, EC 4.2.1.11)
catalyses the reversible dehydration of
d-2-phospho-glycerate to phosphoenolpyruvate (PEP) and
partici-pates in both glycolysis and gluconeogenesis In common with most glycolytic enzymes, enolases from a wide variety of organisms, including Archaea, Bacteria
Keywords
crystal structure; drug design; enolase;
molecular dynamics; structural flexibility
Correspondence
D J Rigden, School of Biological Sciences,
Crown Street, University of Liverpool,
Liverpool L69 7ZB, UK
Fax: +44 151 7954406
Tel: +44 151 7954467
E-mail: drigden@liv.ac.uk
Website: http://www.liv.ac.uk/biolsci/
*These authors contributed equally to this
work
†Deceased
Present address
‡Laborato´rio Nacional de Luz Sı´ncrotron,
Campinas, SP, Brazil
§Department of Molecular Medicine,
College of Veterinary Medicine, Cornell
University, Ithaca, NY, USA
–Department of Molecular Mechanisms of
Mycobacterial Infections, Institut de
Phar-macologie et de Biologie Structurale, CNRS,
UPS (UMR5089), Toulouse, France
(Received 5 June 2007, revised 25 July
2007, accepted 3 August 2007)
doi:10.1111/j.1742-4658.2007.06027.x
Enolase is a validated drug target in Trypanosoma brucei To better charac-terize its properties and guide drug design efforts, we have determined six new crystal structures of the enzyme, in various ligation states and confor-mations, and have carried out complementary molecular dynamics simula-tions The results show a striking structural diversity of loops near the catalytic site, for which variation can be interpreted as distinct modes of conformational variability that are explored during the molecular dynamics simulations Our results show that sulfate may, unexpectedly, induce full closure of catalytic site loops whereas, conversely, binding of inhibitor phosphonoacetohydroxamate may leave open a tunnel from the catalytic site to protein surface offering possibilities for drug development We also present the first complex of enolase with a novel inhibitor 2-fluoro-2-phos-phonoacetohydroxamate The molecular dynamics results further encour-age efforts to design irreversible species-specific inhibitors: they reveal that
a parasite enzyme-specific lysine may approach the catalytic site more closely than crystal structures suggest and also cast light on the issue of accessibility of parasite enzyme-specific cysteines to chemically modifying reagents One of the new sulfate structures contains a novel metal-binding site IV within the catalytic site cleft
Abbreviations
EV, eigenvector; FPAH, 2-fluoro-2-phosphonoacetohydroxamate; PAH, phosphonoacetohydroxamate; PDB, protein databank; PEP,
phosphoenolpyruvate.
Trang 2and Eukarya, are highly conserved [1] The catalytic site
is particularly well conserved, leading to broadly similar
kinetic parameters for enzymes of different origins [2,3]
The quaternary structure of enolase is typically a
homodimer, although some bacteria apparently contain
octameric enzymes [4,5]
Each subunit of enolase contains an eightfold b⁄ a
barrel domain preceded by an N-terminal a + b
domain [6] The catalytic site is contained completely
within a single subunit and lies at the interface of the
two domains: monomeric enolase is catalytically active
[7] Catalysis results from acid–base chemistry involving
a Lys-Glu dyad [8,9] Also essential is the binding of
two divalent metal ions to distinct sites: the first
‘con-formational’ site being required for substrate binding
and the second ‘catalytic’ site, occupied after substrate
has bound, stabilizing the reaction intermediate [6]
This ordered binding is accompanied by dramatic
rear-rangements of three protein loops lying near the
cata-lytic site Note that, although the conventional loop
nomenclature is maintained here, regular secondary
structure is sometimes present in these regions Briefly,
when the catalytic site is occupied by sulfate, phosphate
or phosphoglycolate, all three loops typically adopt an
open conformation, as seen, for example, in our
previ-ous Trypanosoma brucei enolase structure [10] When
occupied by substrate, or the
phosphonoacetohydroxa-mate (PAH) inhibitor, and two metal ions, the loops
are generally all in a closed conformation, as in some
yeast structures [11] Intermediate semiclosed
confor-mations have been observed when one metal ion is
absent or in some complexes with PEP [12,13]
As well as its key roles in glycolysis and
gluconeo-genesis, enolase, in common with other glycolytic
enzymes [14], has a remarkable number of
‘moonlight-ing’ roles in diverse organisms that are unrelated to its
catalytic activity [15] These include roles in the RNA
degradosome in Escherichia coli [16], as a structure
lens protein (s-crystallin) in the eye [17], as a
transcrip-tion factor in both animals [18] and plants [19] and, on
cell surfaces, as a receptor for plasminogen [15] In this
last role, the expression of enolase on the surface of
streptococci is particularly interesting, where its
inter-action with host plasminogen is presumed to facilitate
entry of the parasite into host tissues [20] Very
recently, the enolase of the trypanosomatid parasite
Leishmania mexicanahas also been detected on the cell
surface [21] A role for enolase as plasminogen
recep-tor in this organism is highly plausible because
inter-action between parasite and plasminogen has been
demonstrated [22]
Our interest in T brucei enolase [2,10] stems from
the promise of the glycolytic pathway as a target for
drugs against parasitic protozoa [23] With few excep-tions, homologues of the enzymes involved are present
in the human host, and a premium is placed on seek-ing and exploitseek-ing structural differences between para-site and host proteins Irreversible inhibition is particularly desirable because it would be impervious
to high substrate levels that could displace competitive inhibitors [23] Using parasite enzyme-specific residues (e.g lysines in both cases), selective inhibitors against aldolase [24] and phosphofructokinase [25] have been developed Despite bearing chemically reactive groups,
by combining high affinity and low reactivity, opti-mized inhibitors of this kind should have minimal effects on other proteins in vivo Indeed, a prodrug version of an aldolase inhibitor kills parasite cells without detectable cytotoxicity against human MRC-5 cells [26]
Like other glycolytic enzymes, T brucei enolase has been validated as a drug target: RNA interference of enolase in the bloodstream form of the parasite leads
to an effect on its growth within 24 h and death com-mences at approximately 48 h [27] Encouragingly, the same study also demonstrated that a reduction in eno-lase activity to approximately 15–20% of its original level was sufficient for cell death to occur This sug-gests that incomplete inhibition of this enzyme in vivo might prove sufficient for effective treatment The pres-ence of homologous enolase isoenzymes in the human host raises the complication of selectivity In this respect, enolase is not the best target because the para-site and host enzymes share 58% sequence identity Nevertheless, modelling showed that there are three particularly interesting T brucei enzyme residues, two cysteines (numbered 147 and 241) and lysine 155, near
to the catalytic site, which are not conserved in the human enzymes [2] (Fig S3) The chemical characteris-tics of the side chains of these residues offer the poten-tial for species-selective permanent target inactivation
by appropriately designed covalent inhibitors The
T brucei crystal structure suggested that the cysteines were almost entirely solvent inaccessible, yet, most sur-prisingly, at least Cys147 could be chemically modified
by iodoacetamide with consequent enzyme inhibition [10] In that crystal structure, Lys155 is pointed away from the catalytic site, being unfavourably positioned
to make additional interactions with a catalytic site-bound inhibitor
In the present study, we present six new enolase structures that enhance our understanding of the struc-tural and dynamic properties of the T brucei enolase catalytic site, which is essential for further drug design The new structures demonstrate that the enzyme can adopt three distinct catalytic site structures in the
Trang 3sulfate-bound form, including one containing a novel
metal binding site Furthermore, they show structural
heterogeneity in their inhibitor-bound forms,
highlight-ing the potential to extend future inhibitors out of
the ligand-binding pocket We also present extensive
molecular dynamics simulations aiming to address how
the apparently buried cysteine residues achieve solvent
accessibility and show that Lys155 may indeed offer a
useful alternative possibility for covalent inhibition
Results and Discussion
Overview of the new structures
Characteristics and statistics of data collection and
refinement for the six new structures are presented in
Table 1 The crystal form is the same in each case,
namely the C2221 form previously reported [10],
although the precipitant used was PEG 1000 rather
than the PEG monomethylether 550 In the subsequent
analyses, we compare these structures with the
previ-ously published sulfate-bound structure, refined to
2.35 A˚, and containing Zn2+ions bound to sites I and
III [10], which we refer to here as sulfate_1 The new
structures, all obtained by co-crystallization, are all of
significantly better resolution than sulfate_1, in
partic-ular a complex with PAH inhibitor that diffracted well
to 1.65 A˚ In common with previous structures, from
T brucei and other organisms, a single Arg residue,
numbered 400 in T brucei, lies in the disallowed region
of the Ramachandran plot [10] Among the three
important catalytic site loops previously described (and
discussed further below), there is only one that makes
a crystal contact This involves Glu272 of loop 3, its
last residue and the most distant from the catalytic
site Thus, we can be confident that the conformations
observed represent readily achieved structures of the
native enzyme, rather than crystal packing artefacts In
our initial sulfate_1 structure, density did not allow for
chain tracing of two stretches, from Thr41-Gly42 and
Thr260-Pro266, regions that are frequently poorly
ordered in other enolase structures With the exception
of sulfate_2, all the structures presented here could be
unambiguously fully traced In sulfate_2, density did
not allow for the tracing of the polypeptide chain
between Asp251 and Gln273 inclusive As with
sul-fate_1, one or two artefactual residues preceding the
N-terminal Met of the natural sequence could be
traced in each new structure, and these result from
thrombin cleavage of the His-tag used in purification
(see Experimental procedures) In the three
inhibitor-bound structures, artefactual Zn2+ ions bound, with
partial occupancy (0.5–0.7), at the crystal packing
interface to residues ‘His0’ and Glu27, and to His283 from a crystal symmetry-related chain
The new structures are diverse in the contents of their catalytic sites, both in terms of substrate⁄ inhibitor and
in terms of bound metal (Table 1) The two new sulfate complexes and the previous sulfate-bound structure were all achieved at highly similar crystallization condi-tions (Table 1) As such, there is no clear explanation why they should differ in conformation (see below) and
we view their structural diversity as being the result of
‘freezing out’ of similarly accessible catalytic site conformations Substrate (PEP) and inhibitor (phos-phonoacetohydroxamate, PAH) [28] bind with their phospho and phosphono groups, respectively, occupy-ing the same position as that occupied by sulfate in the earlier sulfate_1 structure [10] Schematic diagrams of the interactions of inhibitors and metal ions with eno-lase are given in Fig S4 The binding mode of PEP seen is essentially the same fully closed conformation as that seen for yeast enolase [protein databank (PDB) code 1one][29] One PAH structure is also fully closed,
as in an earlier yeast complex (PDB code 1ebg) [11], whereas the second, as discussed below, represents a novel conformation for bound PAH Electron density maps for each complex are given in Fig S5
The novel compound 2-fluoro-2-phosphonoaceto-hydroxamate (FPAH), a derivative with a pKa value more resembling that of the phosphate of substrate PEP, was also synthesized and its complex determined
It is a competitive inhibitor of enolase which, despite its lower pKa value compared to PAH, binds more poorly with a Ki at pH 7.2 of 1.4 lm compared to approximately 15 nm for PAH (see supplementary Doc S1 and Figs S1 and S2) [30] It binds in the same way as PEP and PAH with an electron density suggest-ing that both isomers of the R,S racemic mixture bind equally well (Fig S5) Despite the uniformity of ligand binding, protein structure varies considerably at the active site in the new set of structures Rather than try
to explain their differences in the typical qualitative way (i.e open, closed, semiopen, loose, etc.), we attempt a more quantitative description
As shown in Figs 1 and 2, the principal conforma-tional differences between the structures lie in three catalytic site loops, 1–3, corresponding to those high-lighted in many other studies However, unlike the results obtained in a similar analysis for Saccharo-myces cerevisiae crystal structures (data not shown), a fourth peak for the region from residues 215–220 is present This loop is a neighbour of loop 2 and moves
in a coordinated way in T brucei but not in yeast structures Because loop 4 is distant from the catalytic site, it is not discussed further
Trang 4Rmerge
Rfree
2 )
a See
b Values
c Calculated
Trang 5Borrowing a technique more commonly associated with molecular dynamics studies, we analysed the con-formational differences in the new set of structures using essential dynamics [31] This also allowed us to visualize to what extent the resulting modes of confor-mational variability were explored during molecular dynamics simulations (see later) The positions of the six new structures projected onto eigenvectors (EVs) 1 and 2 are shown in Fig 3A Visual inspection of the maximum and minimum projections of EV1 shows
A
B
Fig 3 (A) Projections of the six new crystal structures and molec-ular dynamics trajectories on to EVs 1 and 2 resulting from the essential dynamics analysis Blue circles are used for sulfate struc-tures [open for sulfate_1* (see Experimental procedures), filled for sulfate_3], green triangles for PAH complexes (open for PAH_1, filled for PAH_2), a magenta square for the PEP structure and an orange diamond for the FPAH complex Black dots mark the PEP + 2 Mg trajectory and red dots the single Mg trajectory start-ing from the same PEP complex protein conformation Dots are shown at 2 ps intervals along the trajectory (B) Path of the single
Mg trajectory, indicated at 20 ps intervals, showing a structural switch at approximately 5 ns from a closed (low values for EVs 1 and 2) to an open structure (high EVs) The trajectory start is marked with a circle whereas the end is indicated by a square.
Fig 2 Comparison of sulfate_1, sulfate_2, PEP and PAH_1
struc-tures, coloured, respectively, in shades of green, blue, magenta
and orange The FPAH ligand position closely resembles that of
PAH_1 A complete cartoon representation of sulfate_2 is shown.
Backbone structure is shown for the other three structures only for
loops 1–4, which are labelled Note the gaps in loops 1 and 3 of
the sulfate_1 structure and the loop 3 gap in the sulfate_2
struc-ture Side chains of Lys155 and His156 are shown as sticks, as are
the structures’ respective ligands showing the overlay of bound
sul-fate with phospho and phosphono groups Zinc atoms are shown
as spheres occupying the labelled sites I–IV Black dashes mark
the hydrogen bonding interactions of His156 with PEP or with
met-als in sites III or IV in the sulfate_1 and sulfate_2 structures,
respectively.
Fig 1 Multi-rms plot of the enolase structures in Table 1 produced
with LSQMAN [56] The multi-rms value is defined as the rms value of
the distances between all unique pairs of Ca atoms for a given
resi-due Loops 1–4 (see text) are labelled Note that the value for a
section of loop 3 is artificially low in a stretch, coloured grey, for
which density did not allow tracing of the chain in either of the
open forms, sulfate_1 or sulfate_2.
Trang 6that it captures the coordinated closure of loops 1–4
over the catalytic site The structure sulfate_2 (high
value of EV1 projection) has the loops in a fully open
conformation whereas, in the other structures, they
close over the active site EV2 splits this group of five
to two sets, with positive projection values signifying
structures in which His156 remains outside the
cata-lytic site, whereas negative values mean that His156
enters the site so that the enzyme achieves a
catalyti-cally competent conformation A comparison of
sul-fate_1, sulfate_2, PEP and PAH_1 structures is shown
in Fig 2
Unexpected variability in inhibitor complex
structures
Comparison of the substrate and inhibitor complexes
shows that the His156-out and His156-in structures are
equally represented, the former by PAH_1 and FPAH
and the latter by PEP and PAH_2 This appears to be
the first time that an enolase-PAH complex has
crys-tallised in a nonfully closed conformation The PAH_1
and PAH_2 structures were crystallised at different pH
values We therefore considered whether varying
charge on the phosphono groups of the substrate, with
which His156 interacts on entering the catalytic site,
could be responsible However, the negative charge on
the PAH phosphono group would be greater at
pH 6.5, at which the His-out structure was obtained,
compared to crystallization at pH 5.0 of the His-in
PAH structure Furthermore, a greater negative charge
would be expected on the phosphono group of FPAH
than of PAH due to the lower pKavalue of the fluoro
derivative, yet the FPAH structure was also His-out
The pKa of His165 is not known experimentally,
although an observed value of 5.9 has been ascribed
to it in the yeast enzyme [32] If this is true, then its
ionization state will also differ at pH values of 5.0 and 6.5 A greater attraction for bound PAH of the more positively charged His165 is consistent with the His-in structure observed at pH 5.0 (PAH_2) and the His-out PAH_1 structure observed at pH 6.5 However, con-sideration of the ionization state of His165 does not explain why the FPAH structure at pH 5.0 should be His-out There is no obvious explanation for this structural difference, leading to the conclusion that the His-in and His-out conformations may be similarly energetically favourable and perhaps only chance leads
to the freezing of one or the other in a given crystal The previously unsuspected existence of His-out inhibitor-bound conformations has important implica-tions for further ligand design In the fully closed, His-in conformation, the ligand is fully enclosed in a substrate-sized cavity with little potential for the design
of a larger inhibitor of better affinity or selectivity By contrast, as shown in Fig 4, the outward pointing His156 conformation leaves a tunnel open leading from the protein surface down to the bound ligand This allows ‘growing room’ for a catalytic site-bound inhibitor, enabling access to a larger number of target residues and hence increasing the chance of achieving selectivity for the parasite enzyme over the human counterpart
EV3 from the essential dynamics analysis splits the two His-out structures, PAH_1 and FPAH (data not shown) The difference between these can be described
as a localized twist of loop 2 containing His156 In this case, an explanation is forthcoming The fluorine atom
of one of the isomers of the racemic FPAH makes a close nonbonded contact (3.0 A˚) with Gln164, induc-ing a small displacement of the entire loop Although evidently non-natural, the existence of this loop con-formation emphasizes just how concon-formationally plas-tic the catalyplas-tic sites loops are
Fig 4 A tunnel leading to the catalytic site
is present in PAH_1 (left) but not in PAH_2 (right) A semitransparent surface is shown, uniformly coloured with the exception of the surface contributions from bound PAH (col-oured magenta), Lys155 (dark grey) and His156 (light grey) These residues and the ligand are shown as sticks.
Trang 7Sulfate complexes and a novel metal binding
site IV
The protein conformation most similar to the sulfate_1
structure [10] is sulfate_2 (Table 1) to which 405 Ca
atoms could be fit with an rmsd of 0.4 A˚ and a
maxi-mum displacement of 2.0 A˚ at position 276, a surface
residue distant from the catalytic site Remarkably,
however, the sulfate_2 structure binds its two zinc ions
differently to sulfate_1 Both have fully occupied I
sites, the so-called conformational site [6], but
although sulfate_1 showed the position of the
inhibi-tory metal site III, the sulfate_2 structure reveals a
further novel metal-binding site IV at the enolase
catalytic site As with site III, the metal in site IV is
ligated by His156 but, whereas metal in site III is also
bound by Gln164, Glu165 and Glu208, His156 is the
only protein ligand of the metal in site IV (Fig 5)
Zinc ions bound by single protein ligands are
compar-atively rare in the Metalloprotein Database [33] but
there are several other examples The Zn2+ ion in
site IV is fully occupied and there appears to be no
doubt regarding the identity of this feature in the
elec-tron density map: there are no other components of
the crystallization solution that could be responsible
Additionally, anomalous scattering maps reveal clear,
although somewhat noisy, density for both metal sites
(Fig S6A) The density is similar to that observed for
the sulfur atoms of cysteine and methionine residues,
which have similar scattering power to zinc at the
wavelength used (1.54 A˚) (Fig S6B) The zinc ion in
site IV is further ligated by five solvent molecules with
interatomic separations of 1.75–2.20 A˚ (Fig 5) The B-factor of the metal ion in site IV of 41.3 is close to that of the ligating nitrogen atom of His156 (39.2) The His156 conformations in the sulfate_1 and sul-fate_2 structures differ by only 21 at the v1 rotation, but by a 180 flip of the imidazole ring because the Ne2 atom is involved in both cases (Fig 2) The occu-pation of site IV is unexpected because site III, with additional, negatively charged ligands, would be expected to have a higher affinity for the metal We can be confident that site III, and not site IV, corre-sponds to the inhibitory metal site characterized kineti-cally because the H156A mutant of the S cerevisiae enzyme retains an inhibitory site with one third of the native enzyme’s affinity [34] Such a mutant would simply lack a site IV because the His side chain con-tributes its only protein coordination Nevertheless, it remains possible that binding to site IV contributes to the inhibition of enolase at elevated metal concentra-tions
The sulfate_3 structure closely resembles the PEP and PAH_2 structures, with loops 1–4 fully closed Its
Ca atoms can be fit to those of the PEP complex to produce an rmsd of 0.23 A˚ Additionally, its two zinc ions in sites I and II superimpose on those of the PEP complex, as does the sulfate on the phospho group of PEP (Fig 2) It is unusual for occupation of the cata-lytic site by a small sulfate or phosphate to be suffi-cient to support full closure This situation was seen in one subunit of the E coli structure, but the influence
of crystal packing was suspected [35] More recently, one subunit of the asymmetric human neuron enolase
Fig 5 Coordination of zinc ions occupying
metal site I and novel site IV (labelled) in the
sulfate_2 structure Side chains and sulfate
are shown as sticks, water molecules and
grey, respectively Electron density from a
10 r is shown in magenta Density in a
1 r in the vicinity of site IV, and at 2 r
around the sulfate and metal site I.
Trang 8was demonstrated to adopt the closed conformation
while containing only phosphate or sulfate [36]
Molecular dynamics simulations
The static description of crystal structures is
incom-plete for many proteins but particularly so in the case
of enolase Not only do multiple structures from
sev-eral species demonstrate large conformational changes
at the catalytic site but also, in the case of the T
bru-ceienzyme, crystal structures show Cys147 and Cys241
to be entirely buried in the second layer of protein
resi-dues below the base of the catalytic site [10] whereas
experimental data show that at least Cys147 can be
modified by iodoacetamide with resulting protein
inactivation [10] To explore this and other issues, we
carried out 10 ns duration molecular dynamics
simula-tions on two fully solvated dimeric enolase structures,
the PEP complex and a PEP structure derivative with
active site contents removed to leave a single divalent
metal ion, the ‘conformational’ ion in site I In each
case, Zn2+ was replaced by the more physiologically
relevant Mg2+
Initial modeling also highlighted Lys155 as a residue
near the catalytic site, present only in enolases from
T brucei and Leishmania major, Euglena gracilis and
Treponema pallidum, which could be a target for
irre-versible modification by a suitable inhibitor Such an
inhibitor would likely occupy the catalytic site; thus,
we assessed how closely the Lys155 side chain
approached ligands in that site In the previous
sul-fate_1 structure [10], the Nf atom of Lys155 was far,
around 12 A˚, from the catalytic site-bound sulfate In
the new PEP, PAH and FPAH structures, the Nf
atom is separated from the phospho(no) group by
approximately 7.5 A˚ Remarkably, although the
posi-tion of its neighbour, His156 varies dramatically
among these structures (Fig 2), the position of the Nf
atom is constant (Figs 2 and 4), making a hydrogen
bond with the backbone carbonyl of Ala39
Encourag-ingly, Lys155 lies at the mouth of the tunnel leading
from the protein surface to the bound ligand in the
His-out structures (Fig 4) As such, it would lie near
to an expanded inhibitor occupying that tunnel The
molecular dynamics results show that it approaches
the catalytic site even more closely The separations of
its Nf atom and the oxygen atoms of the PEP
phos-pho group were monitored throughout the PEP
trajec-tory and reached values as low as 6.5 A˚ Clearly, the
prospects for the exploitation of this parasite-specific
residue are much better than first supposed
To address the issue of Cys solvent accessibility, the
solvent-accessible surface area of Cys147 and Cys241
was monitored in both subunits throughout the molec-ular dynamics simulations It is already known that the presence of PEP or PAH does not affect the chemi-cal modification of the cysteines, suggesting that iodoacetamide and other reagents do not access the cysteines via the catalytic site Examination of the structures shows that the modifiable cysteine(s), and the adjacent conserved buried water molecules [10], lie quite close to the opposite surface of the protein Only the side chain of the penultimate residue, Trp428 sepa-rates them from bulk solvent The Trp side chain remains firmly in place throughout the course of our simulations, and neither buried water molecule exchanges with bulk solvent, but nevertheless transient displacement of the Trp side chain remains the most likely means of access to the cysteines by modifying reagents Given that modification is a slow process [10], it may be that the timescale of our simulations is simply too short for it to be observed Also, the actual presence of the rather hydrophobic reagents, rather than pure bulk solvent, may be necessary to induce the necessary structural alterations that allow access, as seen in other systems [37]
The trajectories were mapped onto the EVs obtained
by analysis of the crystal structures (Fig 3) to deter-mine to what extent these modes of structural variabil-ity are explored The PEP complex simulation remains
in the vicinity of the starting point There is little ten-dency toward the His156-out conformation (high val-ues of EV2) and no evidence at all of coordinated loop opening (high values of EV1) Because we have so far only observed the His156-out conformation with inhib-itors, and not with substrate, it may be that the His156-out is favoured only for the former ligands for reasons that remain unclear (see above) The results for the single Mg trajectory (obtained by removing PEP and the site II metal from the PEP complex struc-ture) are intriguingly different After exploring the neighbourhood of the starting conformation for approximately 5 ns, there is a transition (Fig 3B) and the protein explores an area of much higher values for both EV1 (centred around 0.6) and EV2 (centred around 0.2) This implies that, in the absence of PEP and site II metal, there is a shift towards a more open structure along both the coordinated loop dimension (EV1) and the His156-out dimension (EV2) The tra-jectory reaches the EV1 value of the open sulfate_1* structure (maximum 1.73) and exceeds the EV2 values
of the His156-out inhibitor complexes (maximum 0.45) These results are fully consistent with the pre-vailing notion of an ordered mechanism for enolase [38] With only metal site I occupied, the enzyme adopts an open conformation (high values of EVs 1
Trang 9and 2) but with substrate present, and site II occupied
is stable in a closed conformation (lower values of EVs
1 and 2) in which catalytic residues align precisely for
reaction to occur The mapping of the trajectories onto
EV3 (not shown), confirms that the twisted His156-out
loop induced by the fluorine atoms in the FPAH is a
conformation not explored naturally and therefore an
unfavourable one This suggests that the lower pKa
value of that inhibitor, compared to parent PAH,
comes at an energetic cost, consistent with the higher
experimental Kiof FPAH (Doc S1, Figs S1–S2)
Conclusions
Although previous work has painted a picture of
flexi-ble loops near the enolase catalytic site, the diversity
of structures observed for the T brucei enzyme, in a
single crystal form and at broadly similar pH values, is
impressive Particularly notable are the findings that
sulfate occupation of the catalytic site alone can lead
to full closure of all loops whereas, for reasons
unknown, other sulfate-bound structures are open and
exhibit unexpected diversity of metal binding [10]
Sim-ilarly, occupation with the inhibitor PAH (or our
novel fluorinated PAH derivative) need not lead to full
closure of catalytic site loops, leaving open a tunnel
allowing for the design of enlarged inhibitors
occupy-ing more than the immediate vicinity of the small,
enclosed catalytic site Equally encouraging for future
drug design is the discovery that a potentially
modifi-able Lys155 side chain lies near to this tunnel, not far
from the catalytic site as previously supposed [10] Our
molecular dynamics results fail to demonstrate the
appearance of a channel exposing the modifiable
cyste-ine residue(s) to solvent, consistent with modification
being a slow process In summary, our results
empha-size the importance of a full understanding of the
dynamic properties of a drug target for the effective
design of tight-binding and specific ligands
Experimental procedures
Chemical synthesis
Phosphonoacetohydroxamate, lithium salt (PAH) and the
corresponding fluoro analog (FPAH) were obtained from
the diethylphosphonoacetic acid and the
diethyl-2-fluoro-phosphonoacetic acid, respectively, using an improved
reaction sequence distinct from that previously described
[28] Diethyl-2-fluoro-phosphonoacetic was obtained by
saponification of the triethyl-2-fluoro-phosphonoacetate
Diethylphosphonoacetic acid and the
diethyl-2-fluoro-phos-phonoacetic acid were linked to O-benzylhydroylamine in
the presence of 1-(3-dimethylaminopropyl)-3-ethylcarbodii-mide and 4-di(methylamino)pyridine [39] to obtain the pro-tected form of PAH and FPAH The next step consisted
of the deprotection of these compounds by a catalytic hydrogenation on Pd⁄ BaSO4[40] followed by the deprotec-tion of the phosphonate group by bromotrimethylsilane [41] Neutralization of the resulting acid derivatives with LiOH provided the expected products PAH and FPAH as lithium salts in 28% and 15% overall yield, respectively
X-ray crystallography
Recombinant T brucei enolase was expressed and purified
as previously described [2] Briefly, bacterial cells (E coli BL21(DE3)pLysS strain) harbouring the recombinant pET28a plasmid were grown at 37C until an attenuance (D) at 600 nm of approximately 0.5 was reached and protein expression was induced with 1 mm isopropyl thio-b-d-galac-toside for 20 h at 30C After centrifugation at 5000 g in a Sorvall RC26 plus centrifuge with Sorvall GS-3 rotor, the harvested cells were resuspended in TEA buffer, lysed through alternating cycles of freezing–thawing and then centrifuged The resulting clarified supernatant was directly subjected to nickel–nitrilotriacetic acid (Qiagen, Valencia,
CA, USA) affinity chromatography and the purified fusion protein was treated with thrombin to remove the His-tag Crystallization was carried out based on the reported conditions for T brucei enolase [2], but using PEG1000 as precipitant rather than PEG monomethylether 550 Ortho-rhombic C2221 crystals were obtained by the hanging drop method using a reservoir solution of 10% (w⁄ v) PEG1000, 0.01 m ZnSO4or ZnCl2, and 0.1 m Mes, pH 5.0–6.5, with
or without ligand at a concentration of 20 mm Before data collection, crystals of native T brucei enolase were immersed in the cryo-solution (mother liquor, 20% ethylene glycol) with or without 20 mm of ligand (PEP, PAH or FPAH) for 5 min and flash-cooled X-ray diffraction data were collected from two native crystals (referred to as sul-fate_2 and sulfate_3 in Table 1) and four ligand-cocrystal-lised crystals (referred to as PEP, PAH_1, PAH_2 and FPAH in Table 1) using a Mar345 image plate detector (X-Ray Research GmbH, Norderstedt, Germany) mounted
on a Rigaku UltraX 18 generator (Rigaku Corporation, Tokyo, Japan) or at the MX1 beam line at the Laborato´rio Nacional de Luz Sı´ncrotron (Campinas, Brazil) using Mar-CCD 125 mm detector and radiation at 1.431 A˚ (PAH_1) All data sets were processed and scaled with the software mosflm[42] and scala [43] from the ccp4 suite [44] The structures were straightforwardly solved by molecu-lar replacement with the software molrep [45], using the previously determined T brucei enolase structure (PDB code 1oep) as the search model The resulting molecular replacement solutions were subjected to interactive rounds
of manual rebuilding into 2Fo–Fc and Fo–Fc electron density maps using coot [46] and restrained refinement
Trang 10implemented in refmac [47] The ligands were built into
difference electron density maps using coot As with the
pre-viously reported T brucei structure, intense positive peaks in
the difference maps were observed near the active site and
were modelled as metal ions Their identities and occupancy
were determined based on the resulting maps and B-factors
Water molecules were located automatically with the
pro-gram warp [48] An additional, partially occupied
(occu-pancy 0.5–0.7) and artefactual metal site involving a His
residue at position ‘0’ (i.e immediately preceding the natural
initiator Met within the tail downstream of the thrombin
cleavage site in the N-terminal extension containing the His
tag) was observed at a crystal lattice interface between in the
PAH and FPAH complexes In all cases, final rounds of
refinement were carried out with the entire subunit defined as
a translation⁄ libration ⁄ screw group in the modelling of
anisotropy within refmac Isotropic B-factors were
calcu-lated from the refined translation⁄ libration ⁄ screw parameters
and residual isotropic B-factors with tlsanl [49]
Stereo-chemical parameters were analysed with procheck [50]
Details of the data collection and refinement statistics are
shown in Table 1 The PDB [51] codes for the new structures
are: sulfate_2 (2ptw), sulfate_3 (2ptx), PEP (2pty), PAH_1
(1ptz), PAH_2 (2pu0) and FPAH (2pu1)
Molecular dynamics
Molecular dynamics simulations of 10 ns duration each were
performed on the PEP complex structure, and also a structure
in which the PEP and site II metal had been removed leaving
a single metal ion in site I The molecular dynamics
calcula-tions employed the gromacs simulation suite [52] using the
force field appropriate for proteins in water Sodium ions were
added to the simulation system to compensate for the net
neg-ative charge of the protein The simulation was carried out in
a cubic box with a minimal distance between solute and box
edge of 0.7 nm Periodic boundary conditions were used The
topology file for PEP was built using the small-molecule
topology generator prodrg [53], followed by manual
exami-nation The crystal structures were relaxed by the default
protocol of energy minimization and 100 ps of
position-restrained molecular dynamics, in which the protein is
restrained to its starting conformation, prior to the start of
the simulations proper After approximately 2000 ps of the
simulation proper, both trajectories were stable, fluctuating at
Ca rms deviations from the starting structure of
approxi-mately 0.18 nm (PEP + 2 Mg trajectory) and 0.22 nm (single
Mg trajectory) Monitoring of interatomic distances was
performed using other gromacs programs whereas solvent
accessible surface areas were calculated using dssp [54]
Other methods
Essential dynamics analysis [31] of a set of crystal
struc-tures (Table 1) was performed with programs from the
gromacs package [52] Because the sulfate_2 structure lacked a large number of loop 3 residues, it was omitted from the set However, the well-defined loop 1 of sulfate_2 was used to fill the gap of two residues (Thr41 and Gly42)
in the sulfate_1 structure, leaving only the gap in loop 3 from Thr260-Pro266 The sulfate_1 and sulfate_2 structures are similar overall and in the vicinity enabling a simple splicing of these two residues Limited energy minimization
of residues 40–43 of the result with modeller 9 [55] was carried out to regularize bond lengths and angles The spliced version of sulfate_1, called sulfate_1*, was used in the essential dynamics analysis The essential dynamics method is based on the diagonalization of the covariance matrix of atomic fluctuations, which yields a set of eigen-values and EVs The EVs indicate directions in a 3n-dimen-sional space (where n¼ the number of atoms in the protein) and describe concerted fluctuations of the atoms The eigenvalues reflect the magnitude of the fluctuation along the respective EVs Structural superpositions and other conformational analyses were performed using lsq-man [56] and mustang [57] Structural figures were made with pymol [58]
Acknowledgements
We are grateful to Paul Michels for useful discussions regarding this manuscript An early part of this work was supported by the European Commission through its INCO-DEV programme (contract ICA4-CT-2001-10075)
References
1 Fothergill-Gilmore LA & Michels PA (1993) Evolution
of glycolysis Prog Biophys Mol Biol 59, 105–235
2 Hannaert V, Albert MA, Rigden DJ, da Silva Giotto
MT, Thiemann O, Garratt RC, Van Roy J, Opperdoes
FR & Michels PA (2003) Kinetic characterization, structure modelling studies and crystallization of Try-panosoma bruceienolase Eur J Biochem 270, 3205– 3213
3 Qin J, Chai G, Brewer JM, Lovelace LL & Lebioda L (2006) Fluoride inhibition of enolase: crystal structure and thermodynamics Biochemistry 45, 793–800
4 Brown CK, Kuhlman PL, Mattingly S, Slates K, Calie
PJ & Farrar WW (1998) A model of the quaternary structure of enolases, based on structural and evolution-ary analysis of the octameric enolase from bacillus sub-tilis J Protein Chem 17, 855–866
5 Schurig H, Rutkat K, Rachel R & Jaenicke R (1995) Octameric enolase from the hyperthermophilic bacterium Thermotoga maritima: purification, characterization, and image processing Protein Sci 4, 228–236