Degradation of Phytate by the 6-Phytase from Hafnia alvei: A Combined Structural and Solution Study

Phytases hydrolyse phytate (myo-inositol hexakisphosphate), the principal form of phosphate stored in plant seeds to produce phosphate and lower phosphorylated myo-inositols. They are used extensively in the feed industry, and have been characterised biochemically and structurally with a number of structures in the PDB. They are divided into four distinct families: histidine acid phosphatases (HAP), β-propeller phytases, cysteine phosphatases and purple acid phosphatases and also split into three enzyme classes, the 3-, 5- and 6-phytases, depending on the position of the first phosphate in the inositol ring to be removed. We report identification, cloning, purification and 3D structures of 6-phytases from two bacteria, Hafnia alvei and Yersinia kristensenii, together with their pH optima, thermal stability, and degradation profiles for phytate. An important result is the structure of the H. alvei enzyme in complex with the substrate analogue myo-inositol hexakissulphate. In contrast to the only previous structure of a ligand-bound 6-phytase, where the 3-phosphate was unexpectedly in the catalytic site, in the H. alvei complex the expected scissile 6-phosphate (sulphate in the inhibitor) is placed in the catalytic site.


Introduction
Plants use phytate (myo-inositol hexakisphosphate here called InsP 6 ; Ins = inositol = cyclohexane-1,2,3,4,5,6-hexol) as their main storage form of phosphorous, inositol and a variety of minerals and it accounts for 75-80% of the total phosphorous in seeds such as those of cereals and legumes [1,2]. Apart from sequestering phosphorous, phytate has additional anti-nutritional properties as it can form insoluble complexes with proteins and nutritionally important minerals such as magnesium, zinc, iron and calcium [3]. Phytases (myo-inositol hexakisphosphate phosphohydrolases) catalyse the hydrolysis of phytate into inorganic phosphate (P i ) and lower phosphorylated myo-inositols. While phytases have been isolated from a variety of microorganisms, plants and some animal tissues [4,5,6], monogastric species including poultry and pigs as well as humans lack the ability to produce phytate degrading enzymes or simply produce them in insufficient amounts to enable direct use of phytate from the food chain [7]. Over the past two decades the economic value of phytases has increased significantly reflecting their increasingly standard use as animal feed supplements to release the phosphorous locked up in phytate and make it bioavailable [8]. The addition of phytase reduces the need to supplement the feed with P i , a costly non-renewable resource that is estimated to be depleted within 50 years [9] [10]. Furthermore, it lessens the anti-nutritional effects and decreases the environmental burden from phosphate pollution in areas of intense animal farming by decreasing the amount of phosphate excreted by animals.
Phytic acid (the protonated form of InsP 6 ) is a symmetric molecule with six dihydrogenphosphate substituents on the myoinositol core. Titration experiments indicate that phytic acid can carry up to twelve protons, six with pK a ,2.2, two ,5.7 and four , 9.2 [11,12]. It is thus likely to carry four protons at pH 7 (the main species being InsP 6 82 ) and six at pH 5 (the main species being InsP 6 62 ). In solution the most stable conformation of the cyclohexane ring of InsP 6 is a 4 C 1 chair, with five of the six phosphates in the equatorial orientation, while the sixth (on carbon C 2 ) is axial. We adhere throughout to the convention recommended by the Nomenclature Committee of the International Union of Biochemistry [13] based on the mnemonic rule first proposed by Bernard Agranoff, a pioneer of inositol chemistry [14], and used in most later studies [15,16]. Myo-inositol (or here its hexakisphosphate derivative) is depicted as a turtle with the head (axial phosphate) pointing upward. The 1D numbering starts with the turtle's right flipper (position 1) and goes anticlockwise (Fig. 1a,b) (the 1L alternative has the numbering in the opposing direction).
The phytases are divided into four homology-based families: histidine acid phosphatases (HAP), b-propeller phytases, cysteine phosphatases and purple acid phosphatases [9]. The majority are HAPs [17,18] and are part of a large superfamily of phosphatases [19]. Here we focus on the limited number of HAPs that show activity against phytate, henceforth referred to as HAPPs. While a large number of putative HAPP genes have been identified, only a few crystal structures have been reported (Fig. 2). These include the fungal phytases from Aspergillus ficuum [20], A. fumigatus [21], A. niger (AnPhy) [22] and Debaryomyces castellii [23], and the bacterial phytases from Escherichia coli (EcPhy) [24] and Klebsiella pneumoniae [25]. HAPPs show a typical HAP fold with a large a/b-domain and a small a-domain [20,24,26]. The sequences of Hafnia alvei phytase (HaPhy), Yersinia kristensenii phytase (YkPhy), EcPhy and AnPhy are aligned in Fig. 3-this is a structure-based alignment as detailed in the discussion and the Figure legend. The enzymes share a common catalytic site architecture located at the interface of the two domains involving the conserved N-terminal active site motif RHGXRXP and the C-terminal HD. The two-step reaction involves a nucleophilic attack on the phosphorous by the histidine from the RHGXRXP motif to form a covalent phosphohistidine intermediate while the aspartic acid of the HD motif serves as a proton donor to the oxygen atom of the scissile phosphomonoester bond [27,28,29]. The need for the aspartate carboxylate group to be protonated in order to donate a proton to the leaving group explains the acidic pH optimum [27].
The enzymes are additionally classified into 3-, 5-or 6-phytases (EC 3.1.3.8, EC 3.1.3.72 and EC 3.1.3.26, respectively) based on the carbon position on the inositol ring at which they initiate phosphate hydrolysis. Thus 3-phytases first remove the phosphate group at the C 3 or C 1 position (1L-vs. 1D-convention) while 6phytases do so at the C 6 position (or C 4 in the 1L convention) [30]. Very few 5-phytases have been identified (for example that from Selenomonas ruminantium [31]). However, phytases sequentially remove several phosphate groups from phytate, so a simple definition of substrate specificity is not straightforward. Furthermore, even in the first step different positions can be attacked, e.g. a 6-phytase can produce both the main product Ins(1,2,3,4,5)P 5 but in addition some Ins(1,2,4,5,6)P 5 in the first hydrolysis step [32], where the numbers in the parentheses refer to the positions of the phosphates. Structures of phytate or phytate-analogue complexes have been reported with either the 3-or 5-phosphate group in the catalytic centre. For the 3-position these are EcPhy [22] (PDB code 1dkq) and AnPhy [24] (PDB 3k4q) from the HAPP family, and a B. subtilis b-propeller phytase [33] (PDB 3ams, 3amr); while for the 5-position there is a S. ruminantium cysteine phytase [31] (PDB 1u26).
It has become common practice to add microbial phytases to animal feed since they were first commercialized in the early 1990's. The first commercial phytases were derived from various fungal Aspergillus donors, e.g. from A. niger in the Natuphos product (BASF, Ludwigshafen, Germany) classified as a 3-phytase. Later fungal 6-phytases originating from Peniophora lycii were introduced (RONOZYME P and RONOZYME NP; DSM Nutritional Products, Basel, Switzerland). RONOZYME NP is a P. lycii variant with improved intestinal and thermal stability. During the last ten years a number of products based on EcPhy has been marketed e.g. Phyzyme XP (Danisco Animal Nutrition, Marlborough, United Kingdom), OptiPhos (JBS United, Indiana, US) and Quantum (AB Enzymes, Darmstadt, Germany), the latter being a thermostabilized variant. Recently, a 6-phytase from Citrobacter braakii was commercialized as RONOZYME HiPhos (DSM  The wheel represents a neighbour-joining of pairwise Smith-Waterman alignment scores. A full list of the UniProt entries can be found in the Supporting Information S1. The grey triangles indicate PDB entries, while the red triangle is YkPhy and the blue triangle is HaPhy. doi:10.1371/journal.pone.0065062.g002 The 6-Phytase from Hafnia alvei Ligand Complex PLOS ONE | www.plosone.org Nutritional Products, Basel, Switzerland) [34]. All of these commercial products are HAPPs with activity in the acidic range. The phytases from C. braakii and EcPhy are both from enterobacterial donors and have similar pH profiles with an optimum around pH 3.5-4.5. The fungal AnPhy and P. lycii phytases have pH optima around 5.5 and 4-5, respectively.
Here we report the identification of two HAPPs from the enterobacteria Hafnia alvei (HaPhy) and Yersinia kristensenii (YkPhy), together with biochemical studies of their pH and temperature optima, as well as their phytate degradation pattern, which show that they are 6-phytases. We present structures of both enzymes: apo-HaPhy, its complexes with the phytate analogue myo-inositolhexakissulphate (MIHS-HaPhy) and the competitive inhibitor L(+)-tartrate (tar-HaPhy) and YkPhy with one of its products P i (P i -YkPhy). The two enzymes are closely related and structurally similar to EcPhy. The MIHS-HaPhy complex is the first structure of a 6-phytase showing a phytate analogue with the 6-position in the catalytic site.

Phytase Production and Purification
In a screen for new phytases, two phytase-positive bacterial strains that produce enzymes with activity at acidic pH were isolated from a sample of wet soil collected in South Zealand, Denmark. The strains were identified as isolates of Yersinia kristensenii and Hafnia alvei from their partial 16SrDNA sequences. Details of cloning, expression, protein production and purification are given in SI.

Determination of Phytase Activity
Activity was determined with an end-point assay measuring total released phosphate from a sodium phytate solution adapted from the method of Engelsen et al. [35] to the microtitre well format. In brief, 75 ml phytase solution diluted in varying amounts of 0.25 M sodium acetate pH 5.5,0.005% (w/v) Tween-20 was dispensed in a microtitre plate well (NUNC 269620) before 75 ml substrate [prepared by dissolving 100 mg sodium phytate from rice (Sigma, P0109) in 10 ml 0.25 M sodium acetate buffer pH 5.5] were added. The plate was sealed and incubated for 15 min at 37uC while being shaken at 750 rpm. After incubation, 75 ml stop reagent (prepared by mixing 10 ml molybdate solution [10% (w/v) ammonium hepta-molybdate in 0.25% (w/v) ammonia solution] with 10 ml ammonium vanadate (0.24% commercial product from Bie&Berntsen, Cat.No. LAB17650) and 20 ml of 21.7% (w/v) nitric acid) was added and the absorbance measured at 405nm. The phytase activity was expressed in FYT units, with one FYT being the amount of enzyme that liberates 1 mmol inorganic ortho-phosphate/min under the above conditions. An absolute value for the measured activity was obtained by reference to a standard curve made from dilutions of a phytase enzyme preparation with known activity.

pH and Temperature Stability
The pH-dependent activity profiles were determined at 376C in the pH range 2.0 to 7.5 (in 0.5 pH-unit steps) as described in Section 3.2, except that a buffer cocktail containing 50 mM glycine, 50 mM acetic acid, 50 mM Bis-Tris (pH was adjusted to the appropriate value) was used instead of the 0.25 M sodium acetate pH 5.5 buffer. The temperature profile was determined in the range 20-906C in PCR tubes instead of microtitre plates. After a 15 min reaction period at the desired temperature the tubes were cooled to 206C for 20 sec and 150 ml of each reaction mixture was transferred to a microtitre plate. 75 ml stop reagent was added and the absorbance at 405 nm was measured.

Differential Scanning Calorimetry
An aliquot of purified phytase was dialysed against 26 500 ml 20 mM sodium acetate pH 4.0 at 4uC in a 2-3 h step followed by an overnight step. The sample was filtered (0.45 mm) and diluted with buffer to approximately 2 A 280nm units. The dialysis buffer was used as reference in Differential Scanning Calorimetry (DSC). The samples were degassed using vacuum suction and stirring for approximately 10 min. A scan was performed on a MicroCal VP-DSC at a constant scan rate of 1.5uC/min from 20-90uC. The MicroCal Origin software (version 4.10) was used to estimate the denaturation temperature T d (melting temperature, T m ).

In Vitro Activity of HaPhy and YkPhy
The effect of the two phytases on the degradation of phytate was evaluated in an in vitro system mimicking the passage through the stomach. Incubation was conducted first at 40uC and pH 3.0 (60 min) followed by pH 4.0 (30 min) in the presence of pepsin (3000 U/g feed) and 0.8 g of a model feed consisting of 30% soybean meal and 70% corn with approximately 7 g Ca 2+ /kg feed, while dosing the phytases at 125 and 250 FYT/kg of feed. Following incubation, the inositol phosphates (InsP) were extracted from the in vitro samples following a slight modification of the method of Carlsson et al. [36], (Fig. S2). In brief, 5.6 ml of 1 M HCl were added to each sample (5.6 ml) and the resulting samples were mixed intermittently at 500 rpm for 3 h at 40uC interrupted by a freezing step. Subsequently, the samples were centrifuged (1,8006g, 4uC, 5 min), the supernatants were recovered and filtered by centrifugation (11,0006g, 0uC, 60 min) using ultracentrifugal filter devices (Microcon YM-30, Millipore, Billerica, MA) prior to analysis by High Performance Ion Chromatography.

Analysis of Inositol Phosphates by HPIC
InsP 6 -InsP 3 in samples from the in vitro and degradation pathway studies were analysed using High Performance Ion Chromatography (HPIC; Dionex Corp., Sunnyvale, CA) as described in [37]. InsP 6 -InsP 3 were detected after elution from the column by reaction with 0.1% Fe(NO 3 ) 3 ?9H 2 O in a 20 ml/l solution of HClO 4 , by UV absorbance at 290 nm. A reference sample for the identification of peaks was prepared by dissolving 0.5 g of phytic acid dodecasodium salt hydrate in 50 ml of 0.5 M HCl and autoclaving the solution for 1 h at 124uC. Peaks were quantified according to an InsP 6 standard curve (data not shown). The detector response factors of the lower inositol phosphates are lower than those of InsP 6 and therefore the amounts of InsP 5 , InsP 4 and InsP 3 were estimated using the theoretical correction factors of 1.2 ( = 6/5), 1.5 ( = 6/4), and 2 ( = 6/3). . Structure-based sequence alignment for four representative phytases: HaPhy, YkPhy, EcPhy and AnPhy. The alignment is based on the 3D structures of HaPhy, EcPhy and AnPhy using the T-Coffee Expresso web server [67,68] -three is the maximum number of structures allowed by the server. The secondary structure elements of HaPhy are shown above, and of AnPhy below, the alignment. Fully conserved resides are shown in red boxes. The Figure was generated using ESPript [69,70]. doi:10.1371/journal.pone.0065062.g003 The 6-Phytase from Hafnia alvei Ligand Complex PLOS ONE | www.plosone.org InsP 6 Degradation Pathway InsP 6 degradation products were analysed using an established approach [38,39] previously used for a number of other phytases [40,41,42,43,44,45]. In brief, InsP 6 and degradation products (InsP 5 -InsP 2 ) generated by EcPhy, HaPhy and YkPhy were identified by High Performance Ion Chromatography (Fig. S3) and the degradation pathways deduced (Fig. S4). Degradation was performed at pH 4.0 and pH 5.5 (0.25 M sodium acetate) by mixing 100 ml sodium phytate (10 mM) with 100 ml enzyme sample (0.5 FYT/ml) and incubating at 37uC (thermomixer 1000 rpm). The reaction was stopped at various time points (0-150 min) by addition of 200 ml 1.0 M HCl. The patterns at pH 4.0 and 5.5 were comparable and therefore only data from pH 4.0 are shown (Fig. 4).

Structure Solution
Crystallisation and data collection. Crystallisation was performed at room temperature with commercially available screens using sitting-drop vapour-diffusion. Drops were set up employing a Mosquito Crystal liquid handling robot (TTP LabTeck, UK) with 150 nl protein solution plus 150 nl reservoir solution in 96-well format plates (MRC 2-well crystallisation microplate, Swissci, Switzerland) equilibrated against 54 ml reservoir solution.
Apo-HaPhy. A crystal cluster formed in 0.1 M HEPES pH 7.5, 10% isopropanol, 20% (w/v) PEG 4000 using protein at a concentration of 7.7 mg/ml. Individual crystals were separated from the cluster and cryoprotected by dipping them into a solution containing 25% glycerol for 5 sec before vitrifying in liquid nitrogen. X-ray data were collected using a Rigaku Micromax-007 X-ray generator (Cu Ka, l = 1.54179 Å ) equipped with a MAR345 image plate detector (Marresearch GmbH, Germany) to a resolution of 1.90 Å in space group P6 3 22.
Tar-HaPhy: tartrate complex. In a first attempt to obtain a complex with myo-inositol hexakis-sulphate (MIHS), the protein at a concentration of 7.7 mg/ml containing MIHS at a 3:1 molar ratio of inhibitor to protein was incubated at 4uC overnight. Crystals formed in condition H2 of the Index screen (Hampton Research, USA): 0.2 M potassium/sodium tartrate +20% (w/v) PEG 3350, pH 8.4-9.1 and were vitrified without the addition of a cryoprotectant. X-ray data were recorded at beam line ID23-1 of the European Synchrotron Radiation Facility (ESRF, Grenoble) at a wavelength of 1.0723 Å to a resolution of 1.60 Å and were in space group C222 1 . MIHS was subsequently shown not to be bound.
MIHS-HaPhy complex. MIHS binding to HaPhy was characterised by isothermal titration calorimetry (MicroCal VP-ITC). The protein was dialysed against 50 mM sodium acetate pH 4.5 and the ligand was dissolved in the same buffer. The protein concentration in the cell was 0.055 mM and the ligand concentration in the syringe 0.55 mM, with 18 injections at room temperature. The optimal pH for HaPhy activity is ,4.5 therefore sodium acetate buffer pH 4.5 was used for the titration experiment. MIHS (Sigma) was diluted in sodium acetate buffer Pi-YkPhy. Crystallisation screening was performed using protein at a concentration of 10.8 mg/ml. Large crystals were obtained in condition A2 of the PACT Screen (Molecular Dimensions, UK): SPGS (succinate, phosphate, glycine system) buffer pH 5.0, 25% PEG 1500, and were cryoprotected by dipping into a solution containing 20% glycerol for 5 sec and vitrified. Data were collected using a Rigaku Micromax-007 X-ray generator (Cu Ka, l = 1.54179 Å ) equipped with a MAR345 image plate detector to a resolution of 1.67 Å .
Data processing, structure solution and refinement. Xray data were processed using programs from the CCP4 suite [46]. The images were integrated with MOSFLM [47] and scaled with SCALA [48,49]. Molecular replacement (MR) solutions were obtained using MOLREP [50] with models derived using CHAINSAW [51]. The structure of apo-HaPhy was solved using that of apo-EcPhy (pdb code: 1dkl; which has 49% amino acid sequence identity) as a search model. The MIHS-HaPhy, tar- HaPhy and P i -YkPhy structures were all solved using the refined apo-HaPhy model. YkPhy was rebuilt using BUCCANEER [52]. The structures were completed using iterative cycles of COOT [53] and REFMAC5 [54]. For the MIHS-HaPhy, tar-HaPhy and P i -YkPhy complexes, the refinement included Translation Libration Screw-motion (TLS) [55]. In the later stages, the contribution of the hydrogen atoms to the structure factors was taken into account. The final models showed good stereochemistry when analysed with SFCHECK [56] and RAMPAGE [57]. For the MIHS-HaPhy complex, validation was performed using SFCHECK, PROCHECK [58] and MOLPROBITY [59]. Processing and refinement statistics are given in Table 1.

Protein Preparation and Biochemical Properties
HaPhy, the T308A HaPhy mutant and YkPhy were successfully overexpressed and purified. The pH profiles showed optimal activity of the wild type enzymes at pH 4.0-4.5 (Fig. S1a). The T308A HaPhy mutant possessed only 5% of the specific activity of the wild type and it is therefore not included in Fig. S1a. The activity of HaPhy at pH 7 or above is low, which explains our inability to obtain a crystal complex of HaPhy with MIHS at this  pH. Crystallisation of the complex was achieved when the pH of the screens was changed to ,4.5. The activity was measured as a function of temperature at pH 5.5 (Fig. S1b) with maximum activity at 65uC for HaPhy and 55uC for YkPhy. The increase in activity with temperature, until unfolding occurs, is often seen for hydrolytic enzymes and is in keeping with measurements using DSC, which showed melting temperatures of 70uC for HaPhy and 57uC for YkPhy at pH 4.0 (data not shown).
When HaPhy and YkPhy were incubated in an in vitro system mimicking the passage through the stomach in a monogastric animal using corn and soybean meal as a substrate, there was a clear dose-dependent effect for both enzymes, with HaPhy degrading phytate to a greater extent than YkPhy (Fig. S2). HaPhy appeared to degrade InsP 6 and InsP 5 at a faster rate than InsP 4 , as the InsP 4 accumulated in the degradation process.

The Structure of apo-HaPhy
The apoHaPhy structure at pH 7.5 was refined at a resolution of 1.90 Å (Table 1) with one monomer in the asymmetric unit. The first two residues of the N-terminus, five residues (184-188) in a small disordered loop and three residues (414-416) at the Cterminus were disordered. Six side chains on the surface were modelled as dual conformers, while Glu181, Gln182 and His183 have poorly defined side chains and lie at the start of the disordered loop. The model includes seven glycerol, five acetate and 306 water molecules. The overall fold conforms to that of known HAPs with an a domain (residues 25-45 and 137-264) and a a/b domain (the remaining residues) with two a helices on each side of a twisted b-sheet (Fig. 5a). The two central helices of the a domain form part of the active site, which lies in a pocket at the interface of the two domains. The conserved N-terminal RHGXRXP and C-terminal HD motifs compose the catalytic site [26]. HaPhy possesses four disulphide bridges involving cysteines 79/110, 135/410, 180/189, and 384/393, the latter being highly conserved in all known HAP structures [25,60].

The Structure of the Tar-HaPhy and MIHS-HaPhy Complexes
The first attempt to co-crystallise HaPhy with the substrate analogue MIHS [61], at pH 8.4-9.1 led to the tar-HaPhy crystal. The structure was isomorphous to that of the apo-enzyme with the same disordered residues. The active site contained a tartrate molecule, a component of the crystallisation buffer and a well-known inhibitor of HAPs [62,63]. The tartrate superimposes rather closely with the 6-sulphate moiety in the MIHS-HaPhy complex and the P i in the P i -YkPhy complex described below (Fig. 6a). There is a second tartrate, modelled with an occupancy of 0.25, close to a two-fold rotation axis, where it coordinates with His116 and Gln118 from two symmetry related HaPhy molecules.
Perhaps it was not surprising that MIHS failed to bind at this high pH, since the optimal pH for HaPhy activity is ,4.5 (see 4.1). Subsequently, isothermal titration calorimetry was used to confirm that MIHS bound to HaPhy at pH 4.5 and allowed estimation of the K d as ,160 nM assuming a one-site model (data not shown), although the analysis was complicated by what appeared to be a degree of non-specific binding of additional MIHS. For the next screen, MIHS (Sigma I6005) was diluted in sodium acetate buffer pH 4.5 to 20 mM concentration and mixed with the protein to reach a final concentration of 5 mM. Co-crystals with the wildtype HaPhy grew in clusters of insufficient quality for data collection, but a successful hit was obtained for the 5% active T308A mutant of the enzyme. While the mutant MIHS-HaPhy complex crystallised in a different space group to the apo-enzyme and tar-HaPhy complex, the overall fold is essentially unchanged, with one molecule in the asymmetric unit and residues 184-189 disordered. There is a difference in conformation of the first few N-terminal residues, reflecting differences in crystal packing. In the complex the electron density for residues 202-204 is very poorly defined so these residues are not modelled, and the side chain of Lys207 is disordered. In the apo-and tartrate-bound structures this region is better defined, with only the side chains of Asn206 and Lys207 disordered in both. Flexibility in this region, located at the outer border of the binding pocket, may facilitate the entry of the substrate into the active site.
The active site. The electron density for MIHS is very well defined, Fig. 5b, with the 6-sulphate in the catalytic centre, unambiguously confirming that this is a 6-phytase, in agreement with the degradation profile. MIHS is coordinated by Arg18, His19, Arg22, Thr25, Lys26, His128, Thr219, His306, Asp307 and a number of water molecules. The Ne 2 atom of the catalytic nucleophile, His19, lies 3.2 Å from S6 of the MIHS and is in-line with the S-O bond from the inositol ring. The OD1 atom of Asp307, the proton donor, forms an H-bond with the oxygen at position 6 of the inositol ring, Fig. 5b,c,d. The need for the Asp307 to be protonated explains the low pH optimum for the enzyme. The activity may also be affected by the protonation state of the phytate itself, expected to carry four protons at pH 7.5, but six at pH 5 giving an ion with six negative charges [11]. The partly inactive T308A mutant retains 5% of the activity of the wild type in keeping with the loss of a single H-bond. However, the position of the Thr308 Cb is essentially identical in the apo-wild-type, and it can confidently be assumed that the position of the ligand is closely similar in both.
Additional ligand binding site. There is a second MIHS bound between three symmetry-related phytase molecules in the crystal, in a completely different conformation, with the 2-sulphate in the equatorial orientation and all five others axial (Fig. S5a). It is coordinated by Gln23 and Arg52 from one protomer, Lys45 from a second and Lys347 from the third. A key feature is the presence of a potassium ion, coordinated by four oxygens from three sulphates of the MIHS (1, 3 and 5) with its ligand shell completed by the main chain oxygens of Gly230 and Gly231 (Fig. S5b). A previous phytate structure, with MIHS occupying a secondary, non-functional site, bound between two crystallographicallyindependent subunits was reported for the 5-phytase from S. ruminantium [31] (PDB 1u26) in which both MIHS ligands have one equatorial and five axial sulphates. This confirms the tendency  -c) were drawn using CCP4mg [71], and (d) using LigPlot+ [72]. doi:10.1371/journal.pone.0065062.g005 of negatively-charged phytate to bind non-specifically to proteins and to coordinate metals, stressing again the importance of phytases in the animal feed industry to allow its hydrolysis. In the structure of sodium phytate itself [64] the phytate is in a similar conformation with five axial phosphate moieties, and only the sixth, the 2-phosphate, equatorial. These authors suggested that dipolar and coulombic repulsions between the phosphates were the dominant factor in stabilizing this conformer.
Taken together, these structures suggest that the energy difference between the five-equatorial/one-axial conformation and its inverse is not too great and that the population of the two states is strongly influenced by the local environment. This may well be significant in the sequential removal of the various phosphates from the myoinositol-hexakis-phosphate. However, we note that a predominantly axial MIHS would not fit into the HaPhy active site.

The Structure of the P i -YkPhy Complex
The structure of P i -YkPhy was refined at a resolution of 1.67 Å (Table 1) with two protein monomers in the asymmetric unit, each modelled with residues 6 to 414, and with very similar folds (rmsd = 0.22 Å ). The overall fold is essentially identical to that of apo-HaPhy (rmsd = 0.96 Å over 391 residues) and is shown in Fig.  S6 in the same orientation as for HaPhy in Fig. 5(a). Each monomer contains two P i molecules in its active site (Fig. 6a). The first Pi superposes almost perfectly with the 6-sulphate from the HaPhy-MIHS complex. The second is coordinated by His126 (corresponding to His128 in HaPhy) similar to the 1-sulphate of MIHS, but does not superimpose with any of sulphates of the MIHS ring (Fig. 6b).

Discussion
We have characterised two bacterial HAPPs, investigated their activity and phytate degradation profiles and determined the structure of a phytate analogue complex. The biochemical properties are similar and they can both be classified as predominantly 6-phytases, with HaPhy showing a considerably higher thermostability. The structure of YkPhy is similar to that of HaPhy and so discussion focuses on the latter.

Substrate Preference in the HAPP Active Site
The HAPPs can belong to either the 3-phytase (EC 3.1.3.8) or 6-phytase (EC 3.1.3.26) class, which share similar folds and high identity of the active site residues. There are structures of ligand complexes in the PDB for two a-a/b fold HAPPs, an inactive mutant of the E. coli enzyme (EcPhy) in complex with phytate itself [24] and the fungal A. niger enzyme (AnPhy) in complex with MIHS [22]. The sequence identities and rms differences in Ca positions in comparison to HaPhy are YkPhy (52.4%, 1.01Å for 394 Ca), EcPhy (48.2%, 1.18Å for 391) and AnPhy (21.6%, 2.30Å for 236) and a structure-based sequence alignment is shown in Fig. 3. The main difference in the MIHS-HaPhy complex is in the position of the scissile moiety in the inositol ring. In both EcPhy and AnPhy the scissile phosphate lies in position 3, with the implication that phosphorolysis starts at this position of the ring. The presence of the 3-phytate phosphate in the catalytic site of AnPhy, which is in accordance with its classification as a 3-phytase, was explained by the architecture of the binding site, where His338 (AnPhy numbering) is positioned so as to favour axial orientation of the adjacent sulphate. In contrast, the corresponding residue in HaPhy (His306) is further away and does not prevent the equatorial orientation of the sulphate in position 1.
EcPhy was classified as a 6-phytase [65], and the phytate degradation profile reported here confirms that phosphorolysis does primarily occur at this position, with a lesser activity at the 3position. However, the published structure of the phytate complex with the inactive EcPhy mutant showed the 3-phosphate bound in the scissile position [24], (PDB 1dkp). In this structure there are two alternatively occupied mercury sites positioned close to phytate, which can be presumed to cause perturbations in the ligand geometry, but it is not clear that these are sufficient to explain the unexpected placing of the 3-phosphate in the catalytic site. Why EcPhy should have the 3-phosphate in the catalytic site remains unexplained and it is possible that this arises from an accumulation of small differences rather than any one single change.
However, one important difference between EcPhy and HaPhy is that there are much more pronounced conformational changes upon ligand binding in the former. In HaPhy, ligand binding does not lead to significant conformational changes in residues 22-27 (corresponding to 20-25 in EcPhy). While Arg20 in EcPhy moves significantly to form a contact with the scissile phosphate, in HaPhy the corresponding Arg22 is already in the ''contact-ready'' conformation in the apo-structure, superimposing very well on the ligand-bound arginines from both the EcPhy and HaPhy complexes. Similarly, there are no significant conformational changes in the next five residues, although the side chain of Lys26 does move slightly (N Z-apo to N Z-complex shift of 3.2 Å ) to make a contact with the O4 of the 4-sulphate of MIHS. This is quite different from the more dramatic events in EcPhy, where the main chain of the corresponding Lys24 moves by 4.7 Å , giving a large repositioning of the side-chain (N Z-apo to N Z-complex shift of 15 Å ), to make a contact with the oxygen from the 6-phosphate of the ligand. This position is occupied by the 3-sulphate in the MIHS-HaPhy structure. While His128 in MIHS-HaPhy coordinates the 1-sulphate, in EcPhy the interaction with the axial 2-phosphate (which is closest to the 1-sulphate in MIHS-HaPhy) is provided by a water molecule that forms a bond with the main chain oxygen of Phe125 (Fig. 6b). In AnPhy a similar interaction with the 2sulphate is formed by Asp188 (Fig. 6c).
Thr219, which coordinates the 4-sulphate in MIHS-HaPhy, is equivalent to Met216 in EcPhy, which has no contacts with the ligand. There are no equivalent residues in AnPhy; the 5-sulphate in this structure, which is the closest to 4-sulphate in MIHS-HaPhy, is coordinated by Tyr28 and Lys68 (not shown, so as not to overload Fig. 6c). Thr308 has been mutated to alanine in the MIHS-HaPhy complex, so a potential H-bond to the 1-sulphate is missing: Thr305, the equivalent residue in EcPhy, coordinates the axial 2-phosphate (Fig. 6b). It is unlikely that the loss of this Hbond in HaPhy is responsible for the change in phytate orientation.
Yet another difference from the EcPhy active site structure is that there is no change in the conformation of Glu222 (Glu219 in EcPhy) between the apo and ligand-bound states. The ligandinduced conformational changes in this residue were proposed to be important for catalysis in EcPhy. However, in AnPhy, as in HaPhy, there are essentially no conformational changes, confirming that the movement of this Glu is not required for catalysis in all HAPPs. Indeed the movement of Glu219 in EcPhy may be a result of changes of pH [24]. For EcPhy, the apo-structures were determined at different pHs (4.5, 5.0 and 6.6) and the conformation of Glu219 in the apo-form differs from that of the phytate-bound form only at pHs 4.5 and 5.0, while at pH 6.6 Glu219 adopts the phytate-bound conformation.
The YkPhy structure is very similar to that of HaPhy, with a very close overlap of the active site residues, suggesting that it is also a 6-phytase in accordance with our biochemical data. One of the important residues determining the binding pocket specificity is His306, corresponding to His338 in AnPhy, which determines the axial orientation of the adjacent phosphate. This His is in the same conformation as in HaPhy, suggesting the phytate phosphates will be in the same orientation. Although the structures of both enzymes are closely similar, HaPhy has a melting temperature about 13uC higher than that of YkPhy, Fig. S1b. Thermostability is a key property in animal feed, where a commercial phytase that can survive feed processing, e.g. pelleting at 90uC, has a competitive advantage. However, pelleting stability is a combination of intrinsic stability and formulation, and adequate solid formulation of the final product is still a prerequisite for high survival.

Flexibility and Specificity
Our combined solution and structural study demonstrates that HaPhy and YkPhy show a strong preference for the first phosphorolytic degradation step being the removal of the 6phosphate of the phytate. This is in contrast to the EcPhy complex where the 3-phosphate is in the catalytic centre, in spite of an earlier report of a preference for the 6-position [41]. In sequence comparisons, EcPhy lies close to other bacterial 6-phytases (Fig. 2), and quite far from the bacterial 3-phytase from Klebsiella ASR1 [40], in agreement with the results of our digestion studies, One explanation for the 3-phosphate being in the catalytic site in the EcPhy phytate complex could be the mutation of the catalytic histidine in that structure. However, the 6-and indeed the 3phytases process the phytate sequentially down to the monoinositol phosphate [66], which must require some flexibility in the conformation of the substrates as phosphorolysis proceeds. Indeed, the ability of phytate to take up alternative conformations is evidenced by the unusual structure of the second MIHS bound between three protomers in the crystal with five axial and one equatorial ligand.
In addition, we note that in our measurements 25% of the InsP6 degradation products for EcPhy was Ins(1,2,4,5,6)P5 whereas the corresponding values for HaPhy and YkPhy were 11-14%. This means there is less specificity in the interaction of EcPhy with phytate, and the likelihood of its crystallising in the 3-position was 1:4, in contrast to HaPhy where it was closer to 1:10. We propose, that it may be His128 (coordinating the phosphate(sulphate)-1) which determines the higher specificity for a scissile phosphate in the six-position, because the only space remaining for the axial sulphate is outside the binding pocket.

Conclusion
Our degradation studies in solution of the phytases from H. alvei and Y. kristensenii (and indeed E. coli) identify them as 6-phytases with a preference for removing the 6-phosphate from phytate. The X-ray studies on HaPhy and YkPhy double the number of bacterial HAPPs for which structures are available which in combination with biochemical characterisation data will contribute to better understanding of the structure-function relations in this family. In the structure of the complex of HaPhy with MIHS, the 6-sulphate is bound in the catalytic site as expected for an enzyme functioning as a 6-phytase, the first report of a HAPP 6-phytase complex. The presence of the second MIHS ion in the crystal packing illustrates phytate's capability of binding proteins and minerals in vivo. Our high resolution structure gives a detailed picture of the interactions between the substrate and the enzyme when working as a 6phytase. Protein engineering will now be better informed when exploiting the structural data to optimise various properties (e.g. pH-optimum, specific activity and thermal stability) that are important for industrial applications. Optimisation of these properties is key for future application of phytases since the continuing development requires more and more efficient enzymes. HaPhy with its intrinsic high thermostability is an excellent candidate for such protein engineering and has the potential to progress towards a commercial product in the animal feed industry. Figure S1 (a) The relative activity of the two enzymes as a function of pH (b) The relative activity as a function of temperature. In both a) and b) the values are relative % activity normalized to the value at optimum for each phytase. Supporting Information S1.

(DOCX)
Hart for assistance with data collection, Andrew Leech for help with ITC and Nina Kouzmina for the turtle. We thank Mette Bjørn Egede and Daisy Hilbrands for their skillful technical assistance and Peter Fischer Hallin for producing the Phytase wheel.