3.1. General comments
SaPurB crystallizes in an orthorhombic system and the systematic absences indicated either space group
I222 or
I2
12
12
1. Both space groups have eight asymmetric units per unit cell and a single polypeptide chain, labelled chain
A, constitutes the asymmetric unit, with an estimated solvent content of 60%. The space-group assignment of
I222 was confirmed using default settings in the program
Phaser (McCoy
et al., 2007
![[triangle]](/corehtml/pmc/pmcents/rtrif.gif)
) and the structure was solved by molecular replacement and refined to a resolution of 2.50 Å. The log-likelihood gain score in
I222 was 742, compared with a score of 86 in
I2
12
12
1. As will be described,
SaPurB forms a tetramer in the crystal structure and this rules out space group
I2
12
12
1. The crystallographic statistics and model geometry (Table 1) indicate that the analysis has produced an acceptable medium-resolution model. Continuous and well defined electron density was observed for the entire polypeptide except for residues 401 and 402. AMP, a product of the catalyzed reaction, was present in the crystallization conditions and well defined electron density for the ligand was observed in the active site (data not shown). Electron density compatible with PEG was located near the C-terminal segment of α7 and the C-terminal segment of α5 from a symmetry-related subunit (symmetry operation −
x + 1/2, −
y + 1/2, −
z + 1/2, data not shown). In addition, a strong electron-density feature appeared in the active site which we could not explain on the basis of the known chemical components of the crystallization mixture. Different molecules were considered and refinements attempted. Our tentative interpretation is that this density represents oxalate (Fig. 2), possibly acquired as a contaminant of the PEG 200 used in crystallization. Contamination of PEG and its influence on protein crystallization has been observed previously (Jurnak, 1986
). In our case, since ethylene glycol is a precursor of PEG and like PEG itself is susceptible to contamination with aldehydes and peroxides (Ray & Puvathingal, 1985
), then breakdown of PEG 200 and/or the presence of ethylene glycol and subsequent oxidation may have produced oxalate.
3.2. Structure overview
The SaPurB subunit can be divided into three domains that combine to form an elongated structure. Domains I and III are placed at either end of the long helical bundle formed by domain II (Fig. 3). Domain I comprises residues 1–93, which form six helices (α1–α6). Domain II, consisting of residues 94–349, forms an elongated helical bundle assembled by α7 through α15 as well as a short two-stranded antiparallel β-sheet (β1, β2) positioned between α7 and α8. Domain III comprises five helices (α16–α20) constructed from residues 350–431. This domain carries the only unresolved portion of the structure: residues 401 and 402 that are on a loop linking α18 with α19 are disordered.
Gel filtration indicates that
SaPurB exists as a single species with an approximate mass of 85 kDa in solution, suggesting the presence of a dimer under the conditions used during the final stage of purification. In the crystal structure, a symmetry-related subunit (
B) related by the crystallographic twofold axis along unit-cell edge
a is the likely partner for such a dimeric species. The
A–
B interface covers an area of approximately 3380 Å
2; almost 16% of the total surface area of the subunit. As observed in orthologues, the crystal structure reveals the formation of a tetramer (Fig. 3), which in the case of
SaPurB involves rotation of the
A–
B pair about the crystallographic axis along unit-cell edge
b to provide the partner pair of subunits, which are labelled
C and
D. The
A–
C interface extends for approximately 1750 Å
2 or 8% of the surface area of a subunit, whilst the smallest interface, that formed between subunits
A and
D, covers an area of about 1600 Å
2 or 7.5% of the subunit surface area. Changes in the multimeric state of PurB orthologues have been reported elsewhere. The
B. subtilis enzyme has been shown to exist largely as a dimeric species when the protein concentration is low (~0.1 mg ml
−1); however, as the protein concentration increases the tetrameric species becomes the most abundant form (Palenchar & Colman, 2003
).
3.3. The active site, interactions with ligands and enzyme mechanism
The tetramer contains four active sites, each formed by regions donated from three different subunits. In the following description, A, B or C after the residue numbers serves to identify which subunit they belong to. We consider distances between functional groups involved in hydrogen-bonding interactions to fall within the range 2.5–3.5 Å.
A sequence alignment of PurB proteins for which structures have been determined was carried out and a subset is presented in Fig. 4 together with the secondary structure of SaPurB. The PurB sequences of T. maritima and B. anthracis share 52 and 70% identity, respectively, to that of SaPurB. The remainder, L. pneumophila, E. coli, Pl. vivax, C. elegans and Py. aerophilum, fall within a range from 20 to 33% identity. A high level of conservation in and around the ligand-binding/active site is noted, with all of the residues implicated in the enzyme mechanism being strictly conserved (Fig. 4).
The structure of AMP in the active site of
SaPurB (Fig. 5) is similar to that in the
E. coli PurB–AMP complex (Tsai
et al., 2007
; PDB code
2ptq). The AMP phosphate accepts hydrogen bonds from Arg4
C, Tyr5
C, Asn276
C, Ser306
A, Arg310
A and
via a water-mediated association with Arg326
A. The ribose O2′ donates a hydrogen bond to the Arg67
A carbonyl and O3′ accepts hydrogen bonds from the amide of Asp69
A and Arg4
C NH2. The glycosidic O4′ is 3.2 Å from Asp69
A OD1, suggesting that the latter is protonated (data not shown). The adenine N1 accepts a hydrogen bond from Thr95
A OG1; N6 donates hydrogen bonds to His141
B NE2 and Gln212
A OE1, residues which are themselves positioned by hydrogen-bonding interactions. The loop Gly182-Ala183-Val184-Gly185, which is conserved in PurB sequences (Fig. 4), is positioned such that the main-chain carbonyl of Ala183 can accept a hydrogen bond from Gln212
A NE2. His141 in
SaPurB corresponds to His171 in
E. coli PurB and is the catalytic acid for the mechanism (Tsai
et al., 2007
; Kozlov
et al., 2009
).
The oxalate binds accepting hydrogen bonds donated from Ser94
A OG and Gln259
A NE2 and a water molecule that in turn interacts with the side chain of Thr95 (Fig. 2). The distance between an oxalate oxygen and Ser94 OG is 3.6 Å and this may indicate a hydrogen bond. In contrast, the oxalate is 3.4 Å distant from adenine N6 but the geometry is not optimal for hydrogen-bond formation (Leonard
et al., 1995
). The oxalate occupies a position similar to that observed for fumarate in the active site of human PurB (PDB code
2vd6) and to that of the succinate moiety observed in the
E. coli PurB–adenylosuccinate complex (PDB code
2ptr).
A model of
SaPurB with adenylosuccinate was produced (Fig. 6) using the human PurB complex to aid the analysis. Two threonines, Thr93 and Thr140 in
SaPurB, are proposed to bind the succinyl moiety (Segall
et al., 2007
). In
SaPurB, Thr93 is almost 8 Å distant from N6. There is a water-mediated link between Thr140 and Asn270 and the asparagine is the more likely residue to interact with the succinyl group. Gln212 is placed to interact with AMP N6 and a carbonyl on the modelled SAICAR–adenylosuccinate. In
SaPurB, Gln259 binds the oxalate, is positioned on a loop that displays elevated thermal parameters (data not shown) and could undergo a small conformational change to interact with substrate/products.
Two conserved residues, Ser94 and Ser306 in
SaPurB (Figs. 4 and 5), have previously attracted attention (Segall
et al., 2007
). Ser94 is close to His68 and might influence the p
K to favour the protonated form. In
SaPurB His68 ND1 accepts a hydrogen bond from the Val70 amide and thus NE2 would under normal conditions carry a proton. Our model suggests that Ser94, like Ser306, binds substrate. In
SaPurB Asp69 orients Arg310 so that it can in turn bind the substrate phosphate.
One side of the
SaPurB active site is formed by a mobile loop that links β5 and α12. The electron density for this loop is more diffuse than in other parts of the structure and the thermal parameters are higher (data not shown). In most other PurB structures this loop is completely disordered (Fig. 4). In
E. coli PurB this loop is disordered in the apo structure but ligand binding induces a reorganization of the loop to close the ligand-binding site (Tsai
et al., 2007
).
PurB catalyzes a β-elimination that is thought to occur
via a general acid–base uni–bi mechanism (Bridger & Cohen, 1968
; Casey & Lowenstein, 1987
). Previous studies identified His141, His68 and Ser262 as important residues (Lee
et al., 1997
, 1998
, 1999
; Tsai
et al., 2007
). The conserved Ser262 of
SaPurB is positioned on the mobile loop near the active site and may abstract the C
β proton of the substrate, helping to create a carbanion intermediate. His171 is the likely proton donor to N6 to support breakage of the C—N bond. The placement of the adenine in the
SaPurB active site with respect to His141 suggests it is unlikely that protonation of N1 occurs during catalysis, which was a possibility presented by Tsai
et al. (2007
).
His141 has been suggested to be the catalytic base (Lee
et al., 1997
), forming a charge-relay pair with Glu275 to extract a proton from the substrate (Toth & Yeates, 2000
) or alternatively Ser262 (Tsai
et al., 2007
; corresponding to Ser295 in
E. coli). The latter assignment was made from structural information on the apo form of
E. coli PurB, in which the side chain of His171 (His141 in
SaPurB) was rotated away from the position expected for it to play a role in catalysis. In
E. coli PurB structures the highly conserved Ser295 (Ser262 of
SaPurB) is close to the site of catalysis. This is not the situation in
SaPurB, where the conformation of the active-site loop holds Ser262 distant from the catalytic site (Fig. 7). Lys268 of
SaPurB, a strictly conserved residue (Fig. 4), is positioned with its NZ atom 2.8 Å from a succinyl carbonyl. Mutagenesis of the equivalent lysine in
B. subtilis PurB decreased the enzyme activity significantly (Brosius & Colman, 2002
).
Our model supports the idea that His9 of the
E. coli enzyme, which is equivalent to
SaPurB His68, is not the catalytic base but rather binds and orients the substrate to interact with fumarate (Tsai
et al., 2007
). In
SaPurB the oxalate is 3.8 Å distant from His68, but our model (Fig. 6) suggests that the distance would be about 3 Å in the substrate complex.
3.4. Dimer interface and the potential for inhibitor development
We were intrigued by two observations. Firstly, a tetramer (or dimer of dimers) is required to produce a functional enzyme since three subunits are required to contribute residues to form an active site. Secondly, there is less sequence conservation amongst the PurB residues involved in the assembly of the quaternary structure than those around the active sites. We decided to consider the dimer interface as a potential target for screening and structure-based drug-discovery methods.
Targeting transient protein–protein interactions or the interface of oligomeric assemblies by screening and structure-based methods to find small-molecule drug-like inhibitors presents a significant challenge (Yin & Hamilton, 2005
). Complicating factors include the high degree of specificity and the large areas of interactions involved in complex formation. Nevertheless, PurB exists in dimeric and tetrameric forms, with enzyme activity restricted to the tetrameric form. Therefore, preventing the formation of or destabilizing the tetramer provides a route to enzyme inhibition. Our interest was furthered by the observation that there is a marked reduction in sequence conservation amongst those residues involved in the assembly of the quaternary structure.
In support of such an approach, we note that in excess of 30 point mutations have been identified that cause human adenylosuccinate lyase deficiency, several of which influence the stability of the tetramer (Spiegel
et al., 2006
). For example, the mutation of Lys246 to glutamic acid in the human enzyme is located at the
A–
B or
C–
D subunit interface and results in a predominantly monomeric protein with negligible enzyme activity. In addition, the presence of an Arg194-to-cysteine mutation has been demonstrated to reduce the thermal stability of the
B. subtilis enzyme tetramer, leading to significant impairment of catalytic activity (Ariyananda
et al., 2009
).
Visualization of the electrostatic properties of the dimer–dimer interface involved in tetramer formation for both the human and
S. aureus enzymes was achieved using the
Adaptive Poisson-Boltzman Solver (Baker
et al., 2001
; Fig. 8). In
SaPurB two acidic pockets are formed ~13 Å from either end of the helical coil region formed by domain II of each monomer within the
A–
B dimer. The identical view of the human enzyme reveals a less acidic and smaller pocket. The residues forming the surface of this pocket are contributed from α10, α12 and from the loop linking α12 to α13. The symmetry of the dimer leads to a second pocket being formed 15 Å away and when the symmetry of the tetramer is considered four such pockets are created which combine to form two cavities within the core of the tetramer of approximately 13 × 10 × 10 Å in size. As mentioned perviously, targeting interface regions to disrupt oligomerization presents a significant challenge to drug discovery but this region in
SaPurB may be more attractive owing to the symmetry of the tetramer. The effect of a small molecule targeted to the pocket on the surface of one subunit could be amplified by the proximity of four such pockets, offering the potential to exploit multivalent ligands.
