research papers\(\def\hfill{\hskip 5em}\def\hfil{\hskip 3em}\def\eqno#1{\hfil {#1}}\)

Journal logoSTRUCTURAL
BIOLOGY
ISSN: 2059-7983

The structure of a calcium-dependent phosphoinositide-specific phospholipase C from Pseudomonas sp. 62186, the first from a Gram-negative bacterium

aYork Structural Biology Laboratory, Department of Chemistry, University of York, York YO10 5DD, England, bCCP4, STFC Rutherford Appleton Laboratory, Harwell Oxford, Didcot OX11 0QX, England, and cNovozymes A/S, Krogshoejvej 36, 2880 Bagsvaerd, Denmark
*Correspondence e-mail: keith.wilson@york.ac.uk

Edited by R. J. Read, University of Cambridge, England (Received 18 July 2016; accepted 8 December 2016)

Bacterial phosphoinositide-specific phospholipases C (PI-PLCs) are the smallest members of the PI-PLC family, which includes much larger mammalian enzymes responsible for signal transduction as well as enzymes from protozoan parasites, yeast and plants. Eukaryotic PI-PLCs have calcium in the active site, but this is absent in the known structures of Gram-positive bacteria, where its role is instead played by arginine. In addition to their use in a number of industrial applications, the bacterial enzymes attract special interest because they can serve as convenient models of the catalytic domains of eukaryotic enzymes for in vitro activity studies. Here, the structure of a PI-PLC from Pseudomonas sp. 62186 is reported, the first from a Gram-negative bacterium and the first of a native bacterial PI-PLC with calcium present in the active site. Solution of the structure posed particular problems owing to the low sequence identity of available homologous structures. Its dependence on calcium for catalysis makes this enzyme a better model for studies of the mammalian PI-PLCs than the previously used calcium-independent bacterial PI-PLCs.

1. Introduction

Phospholipases of type C (PLCs) cleave phospholipids at the diacylglycerol side of the phosphodiester bond. The members of the phosphoinositide-specific phospholipase C (PI-PLC) family catalyse the cleavage of the membrane lipid phosphatidylinositol (PI) or its phosphorylated derivatives, releasing diacylglycerol and (phosphorylated) myo-inositol. PI-PLCs have been isolated from plants, mammals, yeast, protozoan parasites and bacteria (reviewed in Griffith & Ryan, 1999[Griffith, O. H. & Ryan, M. (1999). Biochim. Biophys. Acta, 1441, 237-254.]; Katan, 1998[Katan, M. (1998). Biochim. Biophys. Acta, 1436, 5-17.]). In plants, they are involved in lipid signalling, regulation of growth and development, and response to nutrient deficiency, as well as biotic and abiotic stress (Singh et al., 2015[Singh, A., Bhatnagar, N., Pandey, A. & Pandey, G. K. (2015). Cell Calcium, 58, 139-146.]). In eukaryotes, they cleave more highly phosphorylated forms of PI, forming myo-inositol 1,4,5-trisphos­phate (Ins 1,4,5-P3) and diacylglycerol, which act as second messengers in a number of receptor-mediated signalling pathways (Suh et al., 1988[Suh, P.-G., Ryu, S. H., Moon, K. H., Suh, H. W. & Rhee, S. G. (1988). Cell, 54, 161-169.]; Rhee et al., 1989[Rhee, S. G., Suh, P.-G., Ryu, S. H. & Lee, S. Y. (1989). Science, 244, 546-550.]; Majerus et al., 1986[Majerus, P. W., Connolly, T. M., Deckmyn, H., Ross, T. S., Bross, T. E., Ishii, H., Bansal, V. S. & Wilson, D. B. (1986). Science, 234, 1519-1526.]; Cocco et al., 2015[Cocco, L., Follo, M. Y., Manzoli, L. & Suh, P.-G. (2015). J. Lipid Res. 56, 1853-1860.]). In Gram-positive bacteria, they have been proposed to contribute to virulence (Heinz et al., 1995[Heinz, D. W., Ryan, M., Bullock, T. L. & Griffith, O. H. (1995). EMBO J. 14, 3855-3863.]), although direct evidence of their role in pathogenicity is unclear (White et al., 2014[White, M. J., Boyd, J. M., Horswill, A. R. & Nauseef, W. M. (2014). Infect. Immun. 82, 1559-1571.]). PI-PLCs are also found in non­pathogenic Gram-positive bacteria (Griffith & Ryan, 1999[Griffith, O. H. & Ryan, M. (1999). Biochim. Biophys. Acta, 1441, 237-254.]).

All PI-PLCs belong to the PI-PLCc_GDPD_SF (catalytic domain of phosphoinositide-specific phospholipase C-like phosphodiesterases superfamily, accession cl14615) superfamily in the NCBI Conserved Domains Database (CDD; Marchler-Bauer et al., 2015[Marchler-Bauer, A. et al. (2015). Nucleic Acids Res. 43, D222-D226.]), together with several other enzyme families, for example, glycerophosphodiester phosphodiesterases (GP-GDEs), which all share a similar mechanism of general base and acid catalysis with conserved histidine residues and hydrolyze the 3′–5′ phosphodiester bonds in different substrates. The active-site features relevant to PI-PLCs and the reaction mechanism will be discussed in more detail below. Briefly, the catalytic domain fold is a (βα)8 TIM barrel, with conserved active-site residues belonging to one half of the barrel (termed the X-box for the mammalian PI-PLCs; Rhee et al., 1989[Rhee, S. G., Suh, P.-G., Ryu, S. H. & Lee, S. Y. (1989). Science, 244, 546-550.]) and the second half responsible for the specificity towards different ligands.

To date, 13 isoforms of mammalian PI-PLCs have been identified, belonging to six families, PLCβ, PLCγ, PLCδ, PLC, PLCζ and PLCη, with specific isoforms being linked to specific signalling pathways. The best studied isoform for which structural information is available is PI-PLCδ from rat (Essen et al., 1996[Essen, L.-O., Perisic, O., Cheung, R., Katan, M. & Williams, R. L. (1996). Nature (London), 380, 595-602.]), which will be used in structure comparisons here. All isoforms have four domains in common (with additional, different domains present for some, depending on their specific functions): a pleckstrin homology domain (PH) involved in membrane binding, an EF-hand domain (with EF-hand calcium-binding motifs, but most probably not linked to calcium regulation in PLCs), a catalytic domain [(βα)8 TIM barrel] and a β-sandwich C2 domain (named after the second conserved domain of protein kinase C, with a possible role in orientation of the catalytic domain relative to the membrane) (Bunney & Katan, 2011[Bunney, T. D. & Katan, M. (2011). Trends Biochem. Sci. 36, 88-96.]; Fukami et al., 2010[Fukami, K., Inanobe, S., Kanemaru, K. & Nakamura, Y. (2010). Prog. Lipid Res. 49, 429-437.]; Williams, 1999[Williams, R. L. (1999). Biochim. Biophys. Acta, 1441, 255-267.]). It has been speculated that the bacterial enzymes evolved from eukaryotic PI-PLCs, losing the additional domains and optimizing the catalytic domain for PI cleavage (Heinz et al., 1998[Heinz, D. W., Essen, L.-O. & Williams, R. L. (1998). J. Mol. Biol. 275, 635-650.]). Indeed, bacterial PI-PLCs exhibit significant structural similarity to the catalytic domain of mammalian PI-PLCs (Heinz et al., 1998[Heinz, D. W., Essen, L.-O. & Williams, R. L. (1998). J. Mol. Biol. 275, 635-650.]; Essen et al., 1996[Essen, L.-O., Perisic, O., Cheung, R., Katan, M. & Williams, R. L. (1996). Nature (London), 380, 595-602.]), and are viewed as convenient model systems for these more complicated eukaryotic signalling molecules (Griffith & Ryan, 1999[Griffith, O. H. & Ryan, M. (1999). Biochim. Biophys. Acta, 1441, 237-254.]). Until recently, the most significant and consistent difference was the presence of an essential, catalytic calcium in the eukaryotic enzymes and its replacement by arginine in all known bacterial structures. As bacterial cell membranes do not normally contain PI (Bishop et al., 1967[Bishop, D. G., Rutberg, L. & Samuelsson, B. (1967). Eur. J. Biochem. 2, 448-453.]; Ames, 1968[Ames, G. F. (1968). J. Bacteriol. 95, 833-843.]), it was proposed that bacteria may use PI-PLCs for interaction with (higher) eukaryotic organisms (Smith et al., 1995[Smith, G. A., Marquis, H., Jones, S., Johnston, N. C., Portnoy, D. A. & Goldfine, H. (1995). Infect. Immun. 63, 4231-4237.]; Goldstein et al., 2012[Goldstein, R., Cheng, J., Stec, B. & Roberts, M. F. (2012). Biochemistry, 51, 2579-2587.]; Marques et al., 1989[Marques, M. B., Weller, P. F., Parsonnet, J., Ransil, B. J. & Nicholson-Weller, A. (1989). J. Clin. Microbiol. 27, 2451-2454.]; Zhao et al., 2013[Zhao, J.-F., Chen, H.-H., Ojcius, D. M., Zhao, X., Sun, D., Ge, Y.-M., Zheng, L. L., Lin, X., Li, L.-J. & Yan, J. (2013). PLoS One, 8, e75652.]; White et al., 2014[White, M. J., Boyd, J. M., Horswill, A. R. & Nauseef, W. M. (2014). Infect. Immun. 82, 1559-1571.]). While eukaryotic enzymes need calcium for fine-tuning signal transduction, the bacterial enzymes are thought to have evolved to use arginine instead of calcium to allow them to act in the intracellular space after uptake by the host cell, where calcium levels are normally low.

The crystallization of a calcium-dependent PI-PLC from Streptomyces antibioticus (SanPI-PLC; Jackson & Selby, 2012[Jackson, M. R. & Selby, T. L. (2012). Acta Cryst. F68, 1378-1386.]) has previously been reported, but while the structure was deposited in the PDB in 2009 (PDB entries 3h4x and 3h4w), there is no associated publication. The purification of SanPI-PLC has been reported and its calcium dependence has been identified (Iwasaki et al., 1994[Iwasaki, Y., Niwa, S., Nakano, H., Nagasawa, T. & Yamane, T. (1994). Biochim. Biophys. Acta, 1214, 221-228.]). The calcium dependence was subsequently confirmed and sequence alignments showed that the enzyme had calcium-binding residues corresponding to those of the eukaryotic enzymes as well as two catalytic histidines (Iwasaki et al., 1998[Iwasaki, Y., Tsubouchi, Y., Ichihashi, A., Nakano, H., Kobayashi, T., Ikezawa, H. & Yamane, T. (1998). Biochim. Biophys. Acta, 1391, 52-66.]; Zhao et al., 2003[Zhao, L., Liu, Y., Bruzik, K. S. & Tsai, M.-D. (2003). J. Am. Chem. Soc. 125, 22-23.]). Inspection of the structure reveals that while there is no calcium ion in the deposited PDB file, this enzyme potentially has calcium-binding residues instead of an arginine in the active site. However, recent research has shown that the reaction mechanism of SanPI-PLC proceeds in an unusual way, with the enzyme capable of cleaving a number of PI analogues, so the exact natural substrates of SanPI-PLC have still to be confirmed (Bai et al., 2010[Bai, C., Zhao, L., Tsai, M.-D. & Bruzik, K. S. (2010). J. Am. Chem. Soc. 132, 1210-1211.]).

All bacterial PI-PLC structures available from the PDB to date have been from Gram-positive bacteria, and none of them are calcium-dependent. The only other bacterial PI-PLC for which a structure with an active-site calcium has been reported is that from Bacillus thuringiensis (PDB entry 1t6m), which has an engineered calcium-binding site resulting from an R69D mutation (Apiyo et al., 2005[Apiyo, D., Zhao, L., Tsai, M.-D. & Selby, T. L. (2005). Biochemistry, 44, 9980-9989.]; Kravchuk et al., 2003[Kravchuk, A. V., Zhao, L., Bruzik, K. S. & Tsai, M.-D. (2003). Biochemistry, 42, 2422-2430.]). We report here the first structure of a wild-type PI-PLC from the Gram-negative bacterium Pseudomonas sp. 62186 (PsPI-PLC), with calcium clearly identified in the active site in both the apo form and in a complex with myo-inositol. This makes an interesting update to an established PI-PLC classification.

2. Experimental

2.1. Cloning, expression, purification and activity assay

The PsPI-PLC gene was inserted into a Bacillus expression plasmid. The DNA encoding the mature polypeptide predicted by SignalP (Bendtsen et al., 2004[Bendtsen, J. D., Nielsen, H., von Heijne, G. & Brunak, S. (2004). J. Mol. Biol. 340, 783-795.]) was cloned in frame to the B. clausii secretion signal BcSP, with the amino-acid sequence MKKPLGKIVASTALLISVAFSSSIASA replacing the native secretion signal sequence with an extra alanine at the C-terminus. This resulted in a recombinant mature polypeptide with an alanine in front of the N-terminus of the mature wild-type sequence. The BcSP fused to the PI-PLC gene was expressed under the control of the promoter system described in Widner et al. (2000[Widner, B., Thomas, M., Sternberg, D., Lammon, D., Behr, R. & Sloma, A. (2000). J. Ind. Microbiol. Biotechnol. 25, 204-212.]). Furthermore, the plasmid contained a terminator sequence and a gene coding for chloramphenicol acetyltransferase, which was used as a selection maker as described in Diderichsen et al. (1993[Diderichsen, B., Poulsen, G. B. & Jørgensen, S. T. (1993). Plasmid, 30, 312-315.]) for B. subtilis. The β-lactamase gene providing ampicillin resistance and the kanamycin-resistance gene were used as cloning selection marker genes for Escherichia coli growth. The plasmid also contained an E. coli origin of replication.

E. coli TOP10 cells were transformed with the plasmid and one clone containing the correct PI-PLC gene sequence was selected. Competent B. subtilis cells were transformed with the plasmid isolated from the selected E. coli clone containing the PI-PLC gene. The final expression plasmid integrated into the B. subtilis chromosome by homologous recombination into the pectate lyase gene locus. Chloramphenicol-resistant transformants were analysed by PCR to verify the correct size of the amplified fragment. A recombinant B. subtilis clone containing the integrated expression construct was selected and cultivated on a rotary shaking table in 500 ml baffled Erlenmeyer flasks each containing 100 ml LB medium supplemented with 34 mg l−1 chloramphenicol. The clone was cultivated for 5 d at 30°C. The enzyme-containing supernatants were separated from the cells by centrifuging the culture broth for 30 min at 15 000g and the enzyme was purified as described below.

The supernatant from 1 l of culture broth (shake flasks) was filtered through a 0.22 µm PES membrane and buffer-exchanged into 50 mM MES pH 6.5 using a packed bed of G-25 (Sephadex G-25 Medium, GE Healthcare). Collected fractions were analysed by SDS–PAGE (NuPAGE 4–12% Bis-Tris gel, reducing conditions) and pooled based on the presence of a band with the expected molecular weight (approximately 33 kDa). The pool was loaded onto a packed bed of Source 15S (GE Healthcare) and bound material was eluted using a linear gradient. PsPI-PLC eluted at 50 mM MES, 0.15 M NaCl pH 6.5, in a gradient from 0 to 0.5 M NaCl (buffer A, 50 mM MES pH 6.5; buffer B, buffer A + 0.5 M NaCl) in ten column volumes. Collected fractions were analysed by SDS–PAGE and pooled based on the presence of a band with the expected molecular weight. The pool was concentrated twofold by the use of centrifugal spinfilters (Vivaspin 20, 10 000 MWCO PES, catalogue No. VS2002, Sartorius AG) to a final concentration of approximately 3.5 mg ml−1 (based on A280, 280 = 1.7 ml mg−1 cm−1).

For the activity assay, the enzyme was diluted to 0.9 and 0.09 mg ml−1 in 100 mM citrate buffer pH 4.0, 5.5 and 7.0. Crude soybean oil (0.25 ml) was pipetted into 2 ml Eppendorf tubes to which the diluted enzyme (25 µl) was added. This resulted in reactions with 10 and 100 mg enzyme per kilogram of oil, 10% water at pH 4.0, 5.5 and 7.0. The mixtures were incubated in a thermoshaker at 50°C for 2 h. 0.500 ml internal standard solution (IS; 2 mg ml−1 triphenyl phosphate in methanol), 0.5 ml chloroform-d (CDCl3) and 0.5 ml Cs-EDTA buffer (0.2 M EDTA adjusted to pH 7.5 with CsOH) were then added. Phase separation was obtained after 30 s shaking followed by centrifugation (3 min, 13 400 rev min−1). For each reaction, as well as a blank (no enzyme), the lower phase was transferred to an NMR tube with a pipette and the 31P NMR spectrum was recorded using a 400 MHz Bruker Avance III HD instrument operating at 300 K with 128 scans and 5 s relaxation delay. All signals were integrated. Chemical shifts were approximately 1.7 p.p.m. (PA), −0.1 p.p.m. (PE), −0.5 p.p.m. (PI), −0.8 p.p.m. (PC) and −17.8 p.p.m. (IS). The concentration of each phospholipid class was calculated as `p.p.m. P', i.e. milligrams of elemental phosphorus per kilogram of oil sample. Hence, p.p.m. P = I/I(IS) × n(IS) × M(P)/m(oil).

2.2. Crystallization

2.2.1. Native protein

Prior to crystallization, the protein was further concentrated to 24 mg ml−1 by ultrafiltration in an Amicon centrifugation filter unit (Millipore) and aliquoted to 50 µl; aliquots that were not immediately set up for crystallization were flash-frozen in liquid nitrogen and stored at −80°C for later use in optimizations/complex formation. The buffer was not changed from the ion-exchange elution conditions 50 mM MES, 0.15 M NaCl pH 6.5. Crystallization screening was carried out using sitting-drop vapour diffusion with drops set up using a Mosquito Crystal liquid-handling robot (TTP LabTech, UK) with 150 nl protein solution plus 150 nl reservoir solution in 96-well format plates (MRC 2-well crystallization microplate, Swissci, Switzerland) equilibrated against 54 µl reservoir solution. Experiments were carried out at room temperature with a number of commercial screens. Initial crystals were obtained in PACT premier HT-96 (Molecular Dimensions) condition H4 (0.2 M KSCN, 20% PEG 3350, 0.1 M BTP pH 8.5). These were optimized in 24-well Linbro trays (hanging drops) using an Oryx8 robot (Douglas Instruments; Shah et al., 2005[Shah, A. K., Liu, Z.-J., Stewart, P. D., Schubot, F. D., Rose, J. P., Newton, M. G. & Wang, B.-C. (2005). Acta Cryst. D61, 123-129.]) by microseeding with a range of seed dilutions and a fine pH grid (between pH 9 and 10). Seeding stock was prepared and seeding was carried out according to published protocols (Shaw Stewart et al., 2011[Shaw Stewart, P. D., Kolek, S. A., Briggs, A. R., Chayen, N. E. & Baldock, P. F. M. (2011). Cryst. Growth Des. 11, 3432-3441.]). Briefly, crystals from the initial successful drop were transferred onto a glass slide, crushed and collected in a Seed Bead (catalogue No. HR2-320, Hampton Research) with 50 µl well solution added, vortexed for 1 min and used as an initial seeding stock. New seeding stock was prepared from the first optimizations. Different protein:well:seeds ratios (1:0.9:0.1, 1:0.99:0.01 and 1:1:0; these are drop sizes in µl) were used for three hanging drops on one cover slip in an automated setup using the Oryx8 robot. Unused seeding stocks were stored at −20°C for later experiments. The best crystals grew at close to pH 10 in the PEG concentration range 24–27%, with 1:200 (1 µl:0.99 µl well:0.01 µl seed) seed dilution.

2.2.2. Myo-inositol complex

Myo-inositol (Merck Millipore, catalogue No. 104507, 99+%) was dissolved in water to make a 500 mM stock. This was mixed with the protein (40 µl protein + 10 µl myo-inositol) to give a final concentration of 100 mM. Crystals were obtained by seeding with native protein crystallization stock in hanging drops with varying protein-to-well solution ratios, using an Oryx-8 robot under conditions similar to those giving native crystals. No back-soaking was carried out, but the excess myo-inositol was possibly removed during cryoprotection.

2.3. Data collection, structure solution and refinement

Computations were carried out using programs from the CCP4 suite (Winn et al., 2011[Winn, M. D. et al. (2011). Acta Cryst. D67, 235-242.]) unless otherwise stated. The final models were validated with MolProbity (Chen et al., 2010[Chen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S. & Richardson, D. C. (2010). Acta Cryst. D66, 12-21.]). Data-processing and refinement statistics for all three structures are provided in Table 1[link]. In the text below and in Fig. 2 and Supplementary Fig. S1, the structure with the smaller unit cell (unit-cell parameters a = b ≃ 96 Å) is referred to as form 1 and that with the larger unit cell (unit-cell parameters a = b ≃ 135 Å), both apo and with bound ligand, is referred to as form 2.

Table 1
Crystallographic statistics

Data set Native form 1 Native form 2 Myo-inositol complex
Data collection
 Beamline I03, Diamond I03, Diamond I03, Diamond
 Wavelength (Å) 0.98 0.98 0.98
 PDB entry 5fyo 5fyp 5fyr
 Space group P4322 P43212 P43212
 Unit-cell parameters (Å) a = 96.08, b = 96.08, c = 113.30 a = 135.21, b = 135.21, c = 112.47 a = 135.59, b = 135.59, c = 113.79
 Resolution range (Å) 48.81–1.50 (1.53–1.50) 67.61–1.17 (1.20–1.17) 48.93–1.45 (1.47–1.45)
 No. of reflections 1097060 4032897 1334118
 Unique reflections 85192 346088 184175
 Monomers in asymmetric unit 2 4 4
 Completeness (%) 99.9 (100) 99.9 (99.8) 99.7 (95.9)
 〈I/σ(I)〉 27.5 (4.8) 17.2 (2.5) 14.5 (2.5)
 CC1/2 0.999 (0.938) 0.999 (0.733) 0.999 (0.786)
 Multiplicity 12.9 (12.4) 11.7 (4.8) 7.2 (5.7)
Rmerge§ 0.049 (0.535) 0.075 (0.531) 0.067 (0.595)
Refinement statistics
 Fraction of free reflections 0.051 0.050 0.050
 Final Rcryst 0.125 0.116 0.144
 Final Rfree 0.166 0.141 0.192
 R.m.s. deviations from ideal geometry (target values are given in parentheses)
  Bond distances (Å) 0.015 (0.019) 0.013 (0.019) 0.015 (0.019)
  Bond angles (°) 1.57 (1.92) 1.58 (1.92) 1.623 (1.923)
  Chiral centres (Å3) 0.132 (0.200) 0.104 (0.200) 0.113 (0.200)
  Planar groups (Å) 0.009 (0.021) 0.011 (0.021) 0.009 (0.021)
 Average main-chain B values (Å2) 21.9 10.1 14.6
 Average side-chain B values (Å2) 25.1 12.6 17.6
 Average B values for Ca2+2) 23.1 13.0 [occupancy 0.8] 16.1
 Average B values for Ins (Å2) N/A N/A 13.8
MolProbity score 0.83 0.87 1.1
 Ramachandran favoured (%) 97.5 97.5 97.4
 Ramachandran outliers (%) 0.2 0.1 0.1
 Clashscore 1.22 0.86 2.17
†Values in parentheses are for the highest resolution shell.
‡CC1/2 values for Imean are calculated by splitting the data randomly in half.
§Rmerge is defined as [\textstyle \sum_{hkl}\sum_{i}|I_{i}(hkl)- \langle I(hkl)\rangle|/][\textstyle \sum_{hkl}\sum_{i}I_{i}(hkl)], where I(hkl) is the intensity of reflection hkl.
¶Ramachandran plot analysis was carried out by MolProbity (Chen et al., 2010[Chen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S. & Richardson, D. C. (2010). Acta Cryst. D66, 12-21.]).
2.3.1. Native structure

Data were collected from two crystal forms at the Diamond Light Source and processed with xia2 (Winter et al., 2013[Winter, G., Lobley, C. M. C. & Prince, S. M. (2013). Acta Cryst. D69, 1260-1273.]). Data-collection statistics are given in Table 1[link]. Data for form 1 were processed to a resolution limit of 1.46 Å in point group 422 and a P lattice with unit-cell parameters a = b = 96.1, c = 113.3 Å, which suggested either one molecule (solvent content of 69%) or two molecules (solvent content of 38.5%) in the asymmetric unit. The second possibility was confirmed since a strong non-origin peak in the Patterson synthesis at (0.50, 0.50, 0.48) indicated twofold pseudotranslation. The low solvent content implied potential problems for molecular replacement in addition to those associated with the presence of pseudotranslation. Data to 1.7 Å resolution were used in molecular replacement and data to 1.5 Å resolution in the final refinement.

The first search model was generated with Sculptor (Bunkóczi & Read, 2011[Bunkóczi, G. & Read, R. J. (2011). Acta Cryst. D67, 303-312.]) from PDB entry 3h4w (sequence identity 27%). Sculptor option 1 was selected to provide maximum truncation of the input model, but nevertheless Phaser (McCoy et al., 2007[McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Storoni, L. C. & Read, R. J. (2007). J. Appl. Cryst. 40, 658-674.]) rejected all of the putative solutions because of clashes. Moreover, there were no solutions which stood out before the rejection, with the log-likelihood gain (LLG; the difference between the log likelihoods for a given model and unrelated models) varying from 27 to 29 for several top solutions, well below the value of 120 that is considered to be an almost certain indication of a correct solution.

There are several pipelines currently available which try various protein fragments or highly truncated homologous structures as search models and then try to extend such partial solutions through density modification and model building. Before undertaking such time-consuming calculations, an attempt was made (that turned out to be successful) to apply this approach at a much simpler level, with a single truncated model, which was manually prepared using Coot (Emsley et al., 2010[Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. (2010). Acta Cryst. D66, 486-501.]) from the PDB entry 3h4w. The truncated model contained helices 90–105 and 133–144, strands 45–51, 114–122 and 183–190, and several adjacent residues (helices 2 and 3 and strands II, III and IV of the X-box; a total of 72 residues out of the 298 in the target sequence), which seemed to form the most compact fragment of the molecule. Phaser now positioned two tNCS-related copies of this model with a LLG of 52 compared with 44 and 42 for the second and third solutions. The best solution confirmed the space group to be P4322. A straightforward attempt to rebuild the MR solution using Buccaneer failed: the resulting R and Rfree of 0.49 and 0.53, respectively, for 493 out of 596 residues that were built indicated incorrect main-chain tracing or sequence assignment. In order to generate a better starting model, two copies of a complete molecule from PDB entry 3h4w were superposed using Coot on the two truncated copies positioned by Phaser. The complete molecules packed almost perfectly, with only one loop being too close to a neighbouring molecule, thus providing strong evidence that the MR solution was correct. The model consisting of complete homologous molecules was rebuilt with a procedure involving refinement with REFMAC5 (Murshudov et al., 2011[Murshudov, G. N., Skubák, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F. & Vagin, A. A. (2011). Acta Cryst. D67, 355-367.]) and removal of incorrect fragments (a total of 300 cycles of refinement, final R of 0.50 and Rfree of 0.53 against all data), density modification with SHELXE (Thorn & Sheldrick, 2013[Thorn, A. & Sheldrick, G. M. (2013). Acta Cryst. D69, 2251-2256.]) (eight cycles, final CC of 48%), refinement with REFMAC5 (24 cycles, R = 0.42, Rfree = 0.44), model extension and sequence docking with Buccaneer (R = 0.25, Rfree = 0.28, 592 residues) and final refinement and model correction (R = 0.13, Rfree = 0.17).

For crystal form 2, diffraction images were merged using xia2 to a resolution of 1.17 Å. Molecular replacement with the model from form 1 using MOLREP identified four monomers per asymmetric unit, all related by pseudotranslation. Manual rebuilding with Coot and refinement with REFMAC5 resulted in a model with an R of 0.12 and an Rfree of 0.14 and with 92.5% of the residues built.

2.3.2. Myo-inositol complex

Data were collected at DLS and processed using XDS (Kabsch, 2010[Kabsch, W. (2010). Acta Cryst. D66, 125-132.]) in space group P43212 to a resolution of 1.45 Å. The structure was essentially isomorphous to the native form 2 and was solved by MR with MOLREP using that structure as a search model. Refinement was performed using REFMAC5 and manual model building/correction in Coot.

2.4. Sequence alignments and structure superpositions

Sequence alignments and structure superpositions were carried out using ClustalX (Larkin et al., 2007[Larkin, M. A., Blackshields, G., Brown, N. P., Chenna, R., McGettigan, P. A., McWilliam, H., Valentin, F., Wallace, I. M., Wilm, A., Lopez, R., Thompson, J. D., Gibson, T. J. & Higgins, D. G. (2007). Bioinformatics, 23, 2947-2948.]) and SSM (Krissinel & Henrick, 2004[Krissinel, E. & Henrick, K. (2004). Acta Cryst. D60, 2256-2268.]) as implemented in CCP4mg (McNicholas et al., 2011[McNicholas, S., Potterton, E., Wilson, K. S. & Noble, M. E. M. (2011). Acta Cryst. D67, 386-394.]), respectively. The results are summarized in Table 2[link].

Table 2
Sequence and structure comparisons of PsPI-PLC with bacterial PI-PLCs and rat PI-PLC

One representative structure per organism (liganded if available) was chosen for comparison with the PsPI-PLC complex with Ins. Sequence identity with PsPI-PLC is shown for the full-length bacterial enzymes and for the region including only the X-box (300–440) and Y-box (487–606) of the mammalian (rat) PI-PLC. Structural alignments were conducted using SSM (Krissinel & Henrick, 2004[Krissinel, E. & Henrick, K. (2004). Acta Cryst. D60, 2256-2268.]) as implemented in CCP4mg (McNicholas et al., 2011[McNicholas, S., Potterton, E., Wilson, K. S. & Noble, M. E. M. (2011). Acta Cryst. D67, 386-394.]) for the X-boxes only, which form the most conserved parts of the proteins and contain all of the residues essential for catalysis. The table shows the input residue ranges and Cα r.m.s.d.s reported by the program. The superposed structures were also used for manual adjustments of ClustalX sequence alignments (Fig. 5[link]) and for the generation of Figs. 4[link] and 7[link] showing superposition of the complete molecules and their active sites, respectively

Species PDB code Ligand Sequence identity to PsPI-PLC Cα r.m.s.d. to PsPI-PLC (Å) for the X-boxes (input residue ranges)
Pseudomonas sp. 62186 5fyr Ins 1.000 0.00 (2–190)
Streptomyces antibioticus 3h4x None 0.271 1.60 (25–223)
Bacillus thuringiensis 1t6m None 0.200 2.22 (5–162)
Bacillus cereus 1ptg Ins 0.185 2.25 (1–166)
Staphylococcus aureus 3v1h§ Ins 0.185 2.27 (7–166)
Listeria monocytogenes 1aod Ins 0.158 2.15 (21–177)
Rattus norvegicus 1djx I3P 0.206 2.10 (300–440)
†Abbreviations for ligands: Ins, myo-inositol; I3P, D-myo-inositol-1,4,5-triphosphate.
‡Structures shown in Figs. 4[link] and 7[link].
§Goldstein et al. (2012[Goldstein, R., Cheng, J., Stec, B. & Roberts, M. F. (2012). Biochemistry, 51, 2579-2587.]).
¶Moser et al. (1997[Moser, J., Gerstel, B., Meyer, J. E. W., Chakraborty, T., Wehland, J. & Heinz, D. W. (1997). J. Mol. Biol. 273, 269-282.]).

3. Results and discussion

3.1. Cloning, expression, purification and assay

The gene encoding PsPI-PLC was amplified by PCR from genomic DNA of the bacterium Pseudomonas sp. 62186, isolated from a seaweed sample collected in Denmark. Sequencing of 16S rRNA from the Pseudomonas species revealed 99.7% identity to P. protegens CHAO (EMBL CP003190). The PsPI-PLC gene was cloned, expressed in Bacillus subtilis and purified as described in §[link]2. The sequence of the gene coding for PI-PLC from Pseudomonas sp. 62186 was deposited in GenBank with accession No. KY078744.

To assay activity and substrate specificity, the purified enzyme in citrate buffer was incubated with crude (non­refined) soybean oil having a natural content of phospholipids. The assay conditions hence resemble those used in enzymatic plant-oil degumming (De Maria et al., 2007[De Maria, L., Vind, J., Oxenbøll, K. M., Svendsen, A. & Patkar, S. (2007). Appl. Microbiol. Biotechnol. 75, 290-300.]). Following an aqueous extraction, the content of nonhydrolyzed phospho­lipids was characterized and quantified by 31P NMR spectroscopy. This sample preparation was inspired by reports documenting that extraction with aqueous EDTA leads to sharper phospholipid NMR signals (Yao & Jung, 2010[Yao, L. & Jung, S. (2010). J. Agric. Food Chem. 58, 4866-4872.]). The result (Fig. 1[link]) provided information about the pH activity profile and substrate specificity of the enzyme. As expected, the enzyme is highly specific towards PI and prefers neutral over acidic conditions.

[Figure 1]
Figure 1
Content of phosphatidic acid (PA), phosphatidylethanolamine (PE), phosphatidylinositol (PI) and phosphatidylcholine (PC) in crude soybean oil before (blank) and after incubation with 10 and 100 mg kg−1 PsPI-PLC, respectively, over 2 h at 50°C at a range of pH values.

3.2. Pseudosymmetry in the two crystal forms

Fig. 2[link](a) illustrates the relation between the symmetries of the two crystal forms, and demonstrates that the diagonal screw twofold axes in form 1 (unit-cell parameters a = b ≃ 96 Å) become coordinate screw twofold axes in form 2 (unit-cell parameters a = b ≃ 135 Å). Accordingly, the HM symbols for the two space groups are different: P4322 and P43212, respectively. The situation is further complicated by the presence of a pseudo-translation in form 1, which acts as a second pseudo-translation in form 2, and results in additional rotational pseudosymmetry elements in both forms. The latter elements coincide in the two forms and are not shown in Fig. 2[link](a). As was shown using Zanuda (Lebedev & Isupov, 2014[Lebedev, A. A. & Isupov, M. N. (2014). Acta Cryst. D70, 2430-2443.]), the pseudo-symmetry space group (the group that includes all crystallographic symmetry elements and all pseudo-symmetry elements) for both structures is I4122, with unit-cell parameters as in form 1. As outlined in the next paragraph, the space groups realised in each of the two forms present one of four possible ways of breaking I4122 symmetry to obtain a tetragonal space group with a P cell (form 1) and a tetragonal space group with a twice larger P cell (form 2). In particular, the actually existing subgroup–supergroup relation between forms 1 and 2 presented in Fig. 2[link](a) is not a priori imposed by the I4122 pseudosymmetry or by the point-group symmetry of the diffraction data and the systematic absences.

[Figure 2]
Figure 2
Subgroup–supergroup relation between crystal forms 1 and 2. (a) Black and orange colours designate crystallographic symmetry and pseudosymmetry elements, respectively, in the form 2 crystal. All of these elements are crystallographic symmetry elements in form 1, which has to be rotated by −45° about c and translated by b/2 from its standard setting for the corresponding symmetry elements to overlay as shown. (Note that both diagonal pseudosymmetry twofold axes and diagonal crystallographic screw twofold axes in form 2 correspond to crystallographic coordinate twofold axes in form 1.) The black rectangle represents the unit cell in form 2 and the orange rectangle represents the rotated and translated unit cell of form 1. (b, c) The Patterson maps for (b) form 1 and (c) form 2 are coloured blue at 0.03 of the height of the origin peaks and the unit cells are shown as yellow rectangles. The full resolution range of the experimental data was used, and hence the Patterson peaks are narrower for form 2, which diffracted to higher resolution. Peaks 1 and 1′ are origin peaks and the heights of other peaks relative to the corresponding origin peaks are 0.30 for peak 2, 0.21 for peak 2′ and 0.10 for peak 3′. All non-origin peaks are split along c and their fractional coordinates along c are 0.5 ± 0.017 for peak 2, 0.0 ± 0.005 for peak 2′ and 0.5 ± 0.017 for peak 3′. Peaks that are not labelled are generated from equivalent peaks by crystallographic translations. Unlabelled images for (b) and (c) were obtained using Coot.

Figs. 2[link](b) and 2[link](c) show the Patterson peaks and thus illustrate the pseudo-translations in the two crystal forms. There are four possible crystallographic space groups (P4122, P41212, P4322 and P43212) for pseudosymmetry space group I4122 provided that the crystallographic translation base vectors are a, b and c, and (a + b + c)/ 2 is a pseudo-translation (as in form 1). There is intensity modulation for reflections k = l = 0 for form 1 (Supplementary Fig. S1a), which, however, are not true systematic absences because the intensities of the most weak reflections with h = 2n + 1 nevertheless significantly exceed their experimental uncertainties (Supplementary Fig. S1c). This observation reduces the number of possible space groups to two (P4122 and P4322). Similarly, there are four possible space groups (HM symbols P4122, P41212, P4322 and P43212) for the pseudosymmetry space group I4122 with crystallographic translation base vectors a′ = a + b, b′ = −a + b and c′ = c (as in the large-cell form 2). Two of these (HM symbols P41212 and P43212) are consistent with the clear systematic absences k = l = 0, h = 2n + 1 in the form 2 data (Supplementary Figs. S1b and S1d). During the structure determination, the remaining ambiguity (enantiomorphic space groups) was resolved at the molecular-replacement step (form 1, highest Phaser LLGs of 52.5 for the correct space group P4322 and 36.5 for the incorrect space group P4122; form 2, highest MOLREP CCs of 0.269 for the correct space group P43212 and 0.259 for the incorrect space group P41212), and this space-group assignment was confirmed by low R factors (below 0.2) and clear electron-density maps for the refined structures.

3.3. Overall fold and comparison to other PI-PLCs

Like the structures of other bacterial PI-PLCs reported to date, PsPI-PLC forms an imperfect (βα)8 TIM barrel (where TIM stands for triosephosphate isomerase; Banner et al., 1975[Banner, D. W., Bloomer, A. C., Petsko, G. A., Phillips, D. C., Pogson, C. I., Wilson, I. A., Corran, P. H., Furth, A. J., Milman, J. D., Offord, R. E., Priddle, J. D. & Waley, S. G. (1975). Nature (London), 255, 609-614.]); bacterial PI-PLCs are family 51699 in the SCOP database (Andreeva et al., 2008[Andreeva, A., Howorth, D., Chandonia, J. M., Brenner, S. E., Hubbard, T. J. P., Chothia, C. & Murzin, A. G. (2008). Nucleic Acids Res. 36, D419-D425.]). As for other bacterial PI-PLCs, some of the β-strands are longer and some are shorter than in a `canonical' (βα)8 barrel (Moser et al., 1997[Moser, J., Gerstel, B., Meyer, J. E. W., Chakraborty, T., Wehland, J. & Heinz, D. W. (1997). J. Mol. Biol. 273, 269-282.]; Heinz et al., 1995[Heinz, D. W., Ryan, M., Bullock, T. L. & Griffith, O. H. (1995). EMBO J. 14, 3855-3863.]), which gives the barrel an irregular shape. In addition, the α-helices are distributed around the barrel in an irregular way, with most clustered on one side. However, in contrast to other bacterial PI-PLC structures, there are connecting α-helices between strands IV and V and between strands V and VI, and the β-barrel itself is more closed, while the C-terminal α8 helix is absent in PsPI-PLC (Figs. 3[link] and 5). Furthermore, PsPI-PLC has two additional short antiparallel β-strands IIb (between strands II and III) and VIIIb (close to the C-terminus).

[Figure 3]
Figure 3
Ribbon representation of the PsPI-PLC fold. The numbering of the β-strands in the (βα)8 barrel is as for bacterial PI-PLCs (I–VIII), but the numbering of the helices is different because of their different arrangement. The helices are instead numbered similarly to rat PI-PLC (1–7), with numbers following that of the previous β-strand (1 after I etc.); when there are several helices between two strands they have the same number followed by a letter. There is a 310-­helix between strands I and II (38–42), coloured purple, with a number 1 in italics. Several very short 310-helices and β-strands are not shown. This and other structure figures were prepared using CCP4mg (McNicholas et al., 2011[McNicholas, S., Potterton, E., Wilson, K. S. & Noble, M. E. M. (2011). Acta Cryst. D67, 386-394.]).

The X-ray structures of five bacterial PI-PLCs are available from the PDB, from Bacillus cereus (BcPI-PLC), B. thuringiensis (BtPI-PLC), Listeria monocytogenes (LmPI-PLC), Staphylococcus aureus (SauPI-PLC) and Streptomyces antibioticus (SanPI-PLC); PDB codes for representative structures are given in Table 2[link]. The sequence identity of PsPI-PLC to the closest homologue, SanPI-PLC, is only 27.1%, and this made molecular replacement a challenge. A comparison of the overall fold of PsPI-PLC and SanPI-PLC is shown in Fig. 4[link](a) together with the outline of the final model. The sequence identity is even lower, between 15.7 and 18.4% (Table 2[link]), for the other bacterial homologues. Nevertheless, the active sites superimpose relatively well, with the ligand in the structures of complexes occupying essentially the same location (Fig. 4[link]). In addition, the bacterial sequences have a degree of similarity to the X-box regions of the catalytic domains of eukaryotic PI-PLCs; the names X-box and Y-box were given to the N- and C-terminal parts of the split (βα)8 barrel (Rhee et al., 1989[Rhee, S. G., Suh, P.-G., Ryu, S. H. & Lee, S. Y. (1989). Science, 244, 546-550.]). While the sequence identity is low (between 13.9 and 19.7% for the X-box and bacterial PI-PLCs from the above list), the catalytic histidines are conserved. Moreover, the calcium-binding residues Asn312, Glu341 and Gln343 in the X-box are present in both calcium-binding PI-PLCs (Fig. 5[link]). This catalytic calcium-binding (or calcium-free, with arginine) motif has been described for the PI-PLC family in an overview of β-barrel proteins (Nagano et al., 2002[Nagano, N., Orengo, C. A. & Thornton, J. M. (2002). J. Mol. Biol. 321, 741-765.]). The structural similarity, as for the bacterial PLCs, is significant: the X-box of the rat enzyme (296–440; PDB entry 1djx; Essen et al., 1997[Essen, L.-O., Perisic, O., Katan, M., Wu, Y., Roberts, M. F. & Williams, R. L. (1997). Biochemistry, 36, 1704-1718.]) superimposes well on the bacterial PI-PLC N-terminal half of the β-barrel. Similar to what has been described for non-calcium-binding bacterial PI-PLCs (Heinz et al., 1998[Heinz, D. W., Essen, L.-O. & Williams, R. L. (1998). J. Mol. Biol. 275, 635-650.]), sequence conservation for the second half of the barrel, corresponding to the Y-box of the mammalian enzyme, is very low, although a number of structural elements superimpose reasonably well (Figs. 4[link] and 5[link]).

[Figure 4]
Figure 4
Ribbon diagrams showing the superposition of the PsPI-PLC complex with myo-inositol (gold) on (a) apo SanPI-PLC (PDB entry 3h4x; green, with the successful molecular-replacement model highlighted in darker green; see §[link]2.3.1), (b) the BcPI-PLC complex with myo-inositol (PDB entry 1ptg; purple) and apo BtPI-PLC with an engineered calcium site (PDB entry 1t6m; ice blue) and (c) the rat PI-PLC complex with Dmyo-inositol-1,4,5-triphosphate (PDB entry 1djx; pink); the enlargement shows details of the active site. Superposition was carried out as detailed in Table 2[link].
[Figure 5]
Figure 5
Sequence alignment of PsPI-PLC with SanPI-PLC and rat PI-PLC, with calcium-binding residues (putative calcium-binding residues for SanPI-PLC) and catalytic histidines outlined in green. Secondary-structure elements are shown as symbols for PsPI-PLC and as background colours for the other two sequences. In both types of presentation, helices are red, 310-helices yellow and β-­strands violet. ClustalX (Larkin et al., 2007[Larkin, M. A., Blackshields, G., Brown, N. P., Chenna, R., McGettigan, P. A., McWilliam, H., Valentin, F., Wallace, I. M., Wilm, A., Lopez, R., Thompson, J. D., Gibson, T. J. & Higgins, D. G. (2007). Bioinformatics, 23, 2947-2948.]) and the structure superpositions described in Table 2[link] were used for the initial alignment of X-boxes and manual adjustment, respectively. The alignment for Y-boxes was obtained in a similar manner but required more substantial structure-based manual correction. The graphical representation of alignment shown here was prepared using ALINE (Bond & Schüttelkopf, 2009[Bond, C. S. & Schüttelkopf, A. W. (2009). Acta Cryst. D65, 510-512.])

3.4. Ligand-binding site

As for all TIM-barrel proteins, the ligand-binding site is formed by residues from the C-terminal regions of the β-strands plus the connecting loop regions, the so-called `activity face' at the C-terminal end of the (βα)8 barrel (Höcker et al., 2001[Höcker, B., Jürgens, C., Wilmanns, M. & Sterner, R. (2001). Curr. Opin. Biotechnol. 12, 376-381.]; Nagano et al., 2002[Nagano, N., Orengo, C. A. & Thornton, J. M. (2002). J. Mol. Biol. 321, 741-765.]). As reported for other PI-PLCs (Essen et al., 1997), there are no major conformational changes upon ligand binding, with the average r.m.s.d. between the subunits of the complex and those of the native structures lying within the range of r.m.s.d. between the individual subunits of the native structures (5fyo, 0.142 Å; 5fyp, 0.276 Å; 5fyr, 0.245 Å; 5fyo/5fyp, 0.233 Å; 5fyr/5fyo, 0.202 Å; 5fyr/5fyp, 0.220 Å). Myo-inositol is coordinated by Asn27, Asn240, Asn189, Glu48, Arg260, three water molecules and, most interestingly, a calcium ion. Trp262 contributes to the ligand-binding pocket. Asn27, Glu48, Asp50, Asp119, the O2 of myo-inositol and a water molecule (Fig. 6[link]a) coordinate the calcium ion. His26 is 3.2 Å from the O2 of inositol, equivalent to the catalytic His32 first described in BcPI-PLC (Heinz et al., 1995[Heinz, D. W., Ryan, M., Bullock, T. L. & Griffith, O. H. (1995). EMBO J. 14, 3855-3863.]). In PsPI-PLC Tyr283 has replaced Asp274, which was proposed to be a second member of the catalytic triad in BcPI-PLC. While Asp274 in BcPI-PLC forms a hydrogen bond to the ND1 atom of His32 (equivalent to His26 in PsPI-PLC), stabilizing its position for catalysis, Tyr283 possibly plays an equivalent role in making a stacking interaction. The inositol OH2 moiety was proposed to complete a catalytic triad. Another histidine, His70, occupies a similar, although not identical, position to His82 in BcPI-PLC, with the CE1 atom ∼5.5 Å from OH1 of inositol, and has been suggested to be essential for catalysis, protonating the leaving group. Histidines corresponding to His26 and His70 are conserved in all (including mammalian) PI-PLCs, and as mentioned in §[link]1 are present in all of the PI-PLCc_GDPD_SF superfamily members. The reaction mechanism is described below.

[Figure 6]
Figure 6
Active site and reaction mechanism of PsPI-PLC. (a) Schematic representation of the hydrogen bonding in the active site. Dashed lines show potential hydrogen bonds made by the enzyme side chains to myo-inositol, Ca2+ and water molecules; hydrogen bonds between the side chains are not shown. Bond distances (Å) are indicated for hydrogen bonds involving myo-inositol and Ca2+. Trp262, shown as a red arch, is involved in a stacking interaction with the myo-inositol ring. His70 is not shown as it makes no hydrogen bonds to the phosphate-free inositol in the PsPI-PLC structure, but it is shown in Fig. 7[link](c), where it occupies a very similar position to His356 of the superposed structure of the rat PI-PLC–D-myo-inositol-1,4,5-trisphosphate complex. (b) The detailed mechanism shown here was initially demonstrated for rat PI-PLC (Essen et al., 1996[Essen, L.-O., Perisic, O., Cheung, R., Katan, M. & Williams, R. L. (1996). Nature (London), 380, 595-602.], 1997[Essen, L.-O., Perisic, O., Katan, M., Wu, Y., Roberts, M. F. & Williams, R. L. (1997). Biochemistry, 36, 1704-1718.]) and adapted for PsPI-PLC based on structural similarity between the mammalian and the bacterial enzymes. His26 (His311 in the mammalian enzyme) or Glu48 (Glu341) acts as a general base and deprotonates the myo-inositol OH2, which then carries out a nucleophilic attack on the 1-phosphate group of the ligand, resulting in a cyclic intermediate. The role of calcium is to stabilize the transition state, while His70 (His356) acts as a general acid and protonates the diacylglycerol leaving group. The cyclic intermediate undergoes hydrolysis in the reverse reaction, with the acid and base swapping their roles. LigPlot+ (Wallace et al., 1995[Wallace, A. C., Laskowski, R. A. & Thornton, J. M. (1995). Protein Eng. Des. Sel. 8, 127-134.]) with a distance threshold of 3.3 Å and ChemDraw (ChemDraw Prime 15.1, PerkinElmer) were used to generate (a) and (b), respectively.

3.5. Active-site comparisons

3.5.1. Streptomyces antibioticus SanPI-PLC

As expected, the ligand-binding site superimposes closely on that of SanPI-PLC, which was used as the molecular-replacement model. However, a number of features are less similar than for the Bacillus or, more interestingly, rat enzymes. This is in agreement with reports that SanPI-PLC is an unusual bacterial PI-PLC (Iwasaki et al., 1994[Iwasaki, Y., Niwa, S., Nakano, H., Nagasawa, T. & Yamane, T. (1994). Biochim. Biophys. Acta, 1214, 221-228.]), subsequently being identified as one of two independent PI-PLCs in S. antibioticus cloned by a shotgun approach (Iwasaki et al., 1998[Iwasaki, Y., Tsubouchi, Y., Ichihashi, A., Nakano, H., Kobayashi, T., Ikezawa, H. & Yamane, T. (1998). Biochim. Biophys. Acta, 1391, 52-66.]). This enzyme was described as `eukaryotic-like' owing to a number of characteristics, one of which was its dependence on calcium, in addition to its lack of activity on GPI-anchored proteins. Two structures of the native enzyme were deposited in the PDB in 2009 (PDB entries 3h4w and 3h4x), but no calcium ion was modelled in the potential calcium-binding site. Subsequently, an unusual mechanism for PI catalysis was proposed, with trans-1,6-cyclic myo-inositol phosphate (1,6-IcP) being formed as an intermediate step (Bai et al., 2009[Bai, C., Zhao, L., Rebecchi, M., Tsai, M.-D. & Bruzik, K. S. (2009). J. Am. Chem. Soc. 131, 8362-8363.]). In addition, SanPI-PLC was shown to be capable of cleaving a number of structural analogues of PI, with its role and natural substrates still under investigation (Bai et al., 2010[Bai, C., Zhao, L., Tsai, M.-D. & Bruzik, K. S. (2010). J. Am. Chem. Soc. 132, 1210-1211.]).

The calcium-binding residues Asn27, Glu48 and Asp50 are all present in SanPI-PLC (see Fig. 5[link] for a sequence alignment and Fig. 7[link]a). A numbering discrepancy should be mentioned here for clarity: the corresponding residues in SanPI-PLC would have been Asn17, Glu39 and Asp41 if numbered as for the mature protein excluding the signal peptide from the entry for the proenzyme in UniProt: B3A043 (Zhao et al., 2003[Zhao, L., Liu, Y., Bruzik, K. S. & Tsai, M.-D. (2003). J. Am. Chem. Soc. 125, 22-23.]). In the deposited structure (PDB entries 3h4w and 3h4x) the numbering starts from the beginning of the expression tag and so adds 21 to the original numbering, making these residues Asn38, Glu60 and Asp62. We use the PDB numbering here to facilitate structure comparisons for the reader. The fourth potential calcium-binding residue Glu129, corresponding to Asp119 in PsPI-PLC, is also present, although not exactly at the same location, but could still bind a calcium ion (Fig. 7[link]a).

[Figure 7]
Figure 7
Comparison of the binding sites for the PI-PLCs. Stereo diagrams show superposition of the binding site of the complex of PsPI-PLC with myo-inositol (gold) on the corresponding regions of (a) SanPI-PLC (PDB entry 3h4x, green), (b) the complex of BcPI-PLC with myo-inositol (PDB entry 1ptg, purple) and the BtPI-PLC R69D mutant (PDB entry 1t6m, ice blue) and (c) the complex of rat PI-PLC with Dmyo-inositol-1,4,5-trisphosphate (PDB entry 1djx, pink). Protein residues are shown as cylinders, ligands in ball-and-stick representation and calcium ions as spheres. Residues used in the superpositions belong to the X-box domains (Table 2[link]). In all panels the catalytic His26 and His70 of PsPI-PLC are in similar locations to their counterparts in the matched structures. The best alignment for His70 (the general acid) is achieved in (c) in the superposition with rat PI-PLC. The planes of His26 (possible general base) in the PsPI-PLC structure and its counterpart (His311) in rat PI-PLC have different orientations, possibly because inositol is not phosphorylated in the former structure, while the 1-phosphate is present and makes a hydrogen bond to His311 in the latter. In (c) Glu48 in PsPI-PLC is seen to align perfectly with Glu341 in rat PI-PLC (a candidate for the general base in the mammalian enzymes; Essen et al., 1997[Essen, L.-O., Perisic, O., Katan, M., Wu, Y., Roberts, M. F. & Williams, R. L. (1997). Biochemistry, 36, 1704-1718.]), but has no clear counterpart in the wild-type bacterial enzymes (a, b). The difference in the position of the inositol O2 between PsPI-PLC and rat PI-PLC can be attributed to the lack of a 1-­phosphate in the former. Ca2+ is present only in PsPI-PLC, rat PI-PLC and the BtPI-PLC R69D mutant and superposes most closely in the first two. Overall, the PsPI-PLC active site has more similarity to the active site of the rat enzyme (c) than the bacterial enzymes (ab), suggesting a mammalian-like catalytic mechanism for PsPI-PLC.

While the calcium-binding residues and catalytic histidines in SanPI-PLC superimpose reasonably well with the corresponding residues in PsPI-PLC, the ligand-binding site as a whole has considerable differences in architecture from PsPI-PLC as well as the Bacillus and mammalian enzymes, with a wider opening, possibly explained by a different reaction mechanism. In PsPI-PLC, Trp262 lines the binding pocket, making a stacking interaction with inositol. There are no equivalent residues in SanPI-PLC at the corresponding location (Fig. 7[link]a). His236 could possibly take part in the lining of the binding pocket in SanPI-PLC. The other bacterial PI-PLCs have tyrosines involved in stacking with inositol, close to, but not exactly superimposing on, Trp262 in PsPI-PLC (Fig. 7[link]b), and the rat enzyme also has a tyrosine at the corresponding location (Fig. 7[link]c). In summary, the binding pocket of the Streptomyces enzyme has significant differences from those of PsPI-PLC and the Bacillus and rat enzymes, as discussed below.

3.5.2. The Bacillus PI-PLCs

BtPI-PLC is a bacterial enzyme in which a R69D variant was designed to engineer a calcium-binding site (Apiyo et al., 2005[Apiyo, D., Zhao, L., Tsai, M.-D. & Selby, T. L. (2005). Biochemistry, 44, 9980-9989.]; Kravchuk et al., 2003[Kravchuk, A. V., Zhao, L., Bruzik, K. S. & Tsai, M.-D. (2003). Biochemistry, 42, 2422-2430.]), and was shown to be activated by calcium. The structure was determined only for the mutant, but the closely similar BcPI-PLC (PDB entry 1ptg; Heinz et al., 1995[Heinz, D. W., Ryan, M., Bullock, T. L. & Griffith, O. H. (1995). EMBO J. 14, 3855-3863.]) can be used to compare the calcium-bound mutant with a wild-type enzyme with an active-site arginine Arg69. BcPI-PLC has 96% sequence identity to BtPI-PLC, only differing by eight amino acids, and was used for comparisons of the mutant with the wild-type enzyme (Apiyo et al., 2005[Apiyo, D., Zhao, L., Tsai, M.-D. & Selby, T. L. (2005). Biochemistry, 44, 9980-9989.]). Here again there is confusion with residue numbering: the authors refer to the BcPI-PLC numbering, even though the amino acids are numbered differently in PDB entry 1t6m (Apiyo et al., 2005[Apiyo, D., Zhao, L., Tsai, M.-D. & Selby, T. L. (2005). Biochemistry, 44, 9980-9989.]). The BcPI-PLC numbering is used here for both Bacillus structures. Fig. 7[link](b) shows that the Arg69 guanidinium group from BcPI-PLC superimposes well on the calcium ion from BtPI-PLC, which in turn is quite close to the calcium ion in PsPI-PLC. The engineered calcium-binding residue Asp69 (the result of the R69D mutation) is positioned between two calcium-binding residues of PsPI-PLC: Asp50 and Asp119. The other two residues involved in calcium coordination in BtPI-PLC, Asp33 and Asp67, superimpose well on Asn27 and Glu48 from PsPI-PLC. The myo-inositol ligand is in a similar location to that in PsPI-PLC, further away from the calcium ion (or the Arg69 guanidinium group) towards the exit from the binding pocket. Tyr200 is involved in stacking interactions with the ligand, similar to Trp262 in PsPI-PLC coming in from a different part of the protein chain. Asp198 of BcPI-PLC occupies a position very close to that of Trp262, making hydrogen bonds to OH3 and OH4 of the ligand. This again differs from PsPI-PLC, where the OH3 and OH4 of myo-inositol are coordinated by Asn189, Asn240 and Arg260, which are all far away from Asp198 (BcPI-PLC).

3.5.3. Rat PI-PLC

The ligand-binding site of the rat PI-PLC catalytic domain (Essen et al., 1997[Essen, L.-O., Perisic, O., Katan, M., Wu, Y., Roberts, M. F. & Williams, R. L. (1997). Biochemistry, 36, 1704-1718.]) superimposes on that of the PsPI-PLC strikingly well, more closely than for the Bacillus enzymes or even SanPI-PLC. The calcium-binding residues all superimpose almost exactly, with the exception of Asp119 in PsPI-PLC, which is Glu390 in rat PI-PLC. Nevertheless, the calcium-coordinating OD2 from Asp119 in PsPI-PLC is in a very similar location to Glu390 OE1 in rat PI-PLC. The calcium ions from the PsPI-PLC and rat PI-PLC structures superimpose very well, much better than the engineered calcium in BtPI-PLC (Figs. 7[link]b and 7[link]c). The ligand in rat PI-PLC is D-myo-inositol-1,4,5-trisphos­phate (Ins1,4,5-P3), which differs from the inositol (Ins) in the PsPI-PLC complex by three phosphates at positions 1, 4 and 5 of the inositol ring. This is a phosphatidyl-less substrate analogue for the mammalian enzymes, which cleave the head groups from more highly phosphorylated species than the bacterial enzymes. The order of substrate preference is phosphatidyl­inositol-4,5-bisphosphate (PIP2) > phosphatidylinositol-4-phosphate (PIP) > PI; bacterial enzymes are only able to cleave PI. There are several ligand complexes in the PDB, but only PDB entry 1djx was used in the comparisons here, since the others are very similar. The inositol ring of Ins from PsPI-PLC superimposes on that of Ins1,4,5-P3 even better than on Ins from BcPI-PLC (where the ligand is exactly the same). A stacking interaction is provided by Tyr551, which is located closer to Phe262 from PsPI-PLC than the tyrosines in the Bacillus enzymes. There is no equivalent stacking interaction in SanPI-PLC at this location, probably because it binds different ligands and cleaves them using a different mechanism.

3.6. Reaction mechanism

For PI-PLCs, both bacterial (BcPI-PLC; Heinz et al., 1995[Heinz, D. W., Ryan, M., Bullock, T. L. & Griffith, O. H. (1995). EMBO J. 14, 3855-3863.], 1998[Heinz, D. W., Essen, L.-O. & Williams, R. L. (1998). J. Mol. Biol. 275, 635-650.]) and mammalian (rat PI-PLC; Essen et al., 1996[Essen, L.-O., Perisic, O., Cheung, R., Katan, M. & Williams, R. L. (1996). Nature (London), 380, 595-602.], 1997[Essen, L.-O., Perisic, O., Katan, M., Wu, Y., Roberts, M. F. & Williams, R. L. (1997). Biochemistry, 36, 1704-1718.]), substrate-assisted catalysis has been proposed, in which the 2-OH group of phosphoinositide acts as a nucleophile. His32 in BcPI-PLC is a general base and His82 is a general acid, donating the proton to the leaving group. For the mammalian (rat) PI-PLC the mechanism may differ in that His311 (analogous to His32) is too distant from 2-OH of inositol in all available structures of complexes with inositol phosphates, so a role in stabilizing a transition state has been suggested for this His, while the role of general base was assigned to Glu341 (Ellis et al., 1998[Ellis, M. V., James, S. R., Perisic, O., Downes, C. P., Williams, R. L. & Katan, M. (1998). J. Biol. Chem. 273, 11650-11659.]). In PsPI-PLC, His26 (analogous to His311 in rat PI-PLC and His32 in BcPI-PLC) is close enough to the 2-OH group of the inositol (3.2 Å) and its role as general base in the first step of the reaction cannot be excluded. Otherwise, the mechanism for PsPI-PLC should be similar to that suggested for the mammalian enzyme (Fig. 6[link]b), and the general base His26 (His311 in rat PI-PLC and His32 in BcPI-PLC) or Glu48 (Glu341 in rat PI-PLC; no direct equivalent in BcPI-PLC) deprotonates 2-OH, leading to its nucleophilic attack on the 1-phosphate group and the formation of a cyclic intermediate. In this first step the O atom of the 1-phosphate group becomes transiently coordinated to calcium. His70 (His356, His82) then acts as a general acid, protonating the diacylglycerol leaving group. In the second step of the reaction, the cyclic intermediate undergoes hydrolysis. At this step, the acid/base roles were suggested to be reversed, so in PsPI-PLC His70 (His356, His82) would activate a water molecule and His26 (His311, His32) or Glu48 (Glu341) would become a general acid.

4. Conclusions

A new member of the bacterial PI-PLC family, the first of its kind for a PI-PLC from a Gram-negative bacterium, has been cloned and expressed and its X-ray structure has been determined. It is only the second bacterial PI-PLC in which calcium binding has been reported, and the first where a calcium ion has been modelled in the structure. The presence of a catalytic calcium makes this new group of bacterial enzymes more suitable models for activity studies on catalytic domains of medically important mammalian PI-PLCs. However, for SanPI-PLC, recent publications report a cleavage mechanism distinct from the `canonical' mechanisum, namely trans-cyclization involving the formation of an inositol 1,6-cyclic phosphate intermediate rather than the cis-cyclization to form inositol 1,2-cyclic phosphate that occurs for all other known phospholipases C (Bai et al., 2009[Bai, C., Zhao, L., Rebecchi, M., Tsai, M.-D. & Bruzik, K. S. (2009). J. Am. Chem. Soc. 131, 8362-8363.], 2010[Bai, C., Zhao, L., Tsai, M.-D. & Bruzik, K. S. (2010). J. Am. Chem. Soc. 132, 1210-1211.]). In addition, comparisons show that the ligand-binding site of PsPI-PLC is strikingly close to that of the mammalian enzymes and less similar to those from the closest bacterial structures. Taking into account these factors, the Pseudomonas enzyme seems to be a better model for the eukaryotic catalytic domain than SanPI-PLC. Besides serving as a model system for mammalian PI-PLCs, the PsPI-PLC structure provides insight into the Gram-negative PI-PLCs.

Finally, phospholipases, including PLCs, are used in a number of industrial applications, for example in the baking, egg-yolk and dairy industries, and in plant-oil degumming (De Maria et al., 2007[De Maria, L., Vind, J., Oxenbøll, K. M., Svendsen, A. & Patkar, S. (2007). Appl. Microbiol. Biotechnol. 75, 290-300.]). Such enzyme-based procedures are much more environmentally friendly than chemical methods. The new structures provide valuable information for further protein engineering to improve the existing applications or create opportunities for new ones.

Acknowledgements

We thank Diamond Light Source for access to beamline I03 (proposal No. mx-9948) that contributed to the results presented here. We thank Johan Turkenburg and Sam Hart for support with the X-ray data collection.

References

First citationAmes, G. F. (1968). J. Bacteriol. 95, 833–843.  CAS
First citationAndreeva, A., Howorth, D., Chandonia, J. M., Brenner, S. E., Hubbard, T. J. P., Chothia, C. & Murzin, A. G. (2008). Nucleic Acids Res. 36, D419–D425.  CrossRef CAS
First citationApiyo, D., Zhao, L., Tsai, M.-D. & Selby, T. L. (2005). Biochemistry, 44, 9980–9989.  Web of Science CrossRef PubMed CAS
First citationBai, C., Zhao, L., Rebecchi, M., Tsai, M.-D. & Bruzik, K. S. (2009). J. Am. Chem. Soc. 131, 8362–8363.  CrossRef CAS
First citationBai, C., Zhao, L., Tsai, M.-D. & Bruzik, K. S. (2010). J. Am. Chem. Soc. 132, 1210–1211.  Web of Science CrossRef CAS PubMed
First citationBanner, D. W., Bloomer, A. C., Petsko, G. A., Phillips, D. C., Pogson, C. I., Wilson, I. A., Corran, P. H., Furth, A. J., Milman, J. D., Offord, R. E., Priddle, J. D. & Waley, S. G. (1975). Nature (London), 255, 609–614.  CrossRef PubMed CAS Web of Science
First citationBendtsen, J. D., Nielsen, H., von Heijne, G. & Brunak, S. (2004). J. Mol. Biol. 340, 783–795.  Web of Science PubMed
First citationBishop, D. G., Rutberg, L. & Samuelsson, B. (1967). Eur. J. Biochem. 2, 448–453.  CrossRef CAS
First citationBond, C. S. & Schüttelkopf, A. W. (2009). Acta Cryst. D65, 510–512.  Web of Science CrossRef CAS IUCr Journals
First citationBunkóczi, G. & Read, R. J. (2011). Acta Cryst. D67, 303–312.  Web of Science CrossRef IUCr Journals
First citationBunney, T. D. & Katan, M. (2011). Trends Biochem. Sci. 36, 88–96.  CrossRef CAS
First citationChen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S. & Richardson, D. C. (2010). Acta Cryst. D66, 12–21.  Web of Science CrossRef CAS IUCr Journals
First citationCocco, L., Follo, M. Y., Manzoli, L. & Suh, P.-G. (2015). J. Lipid Res. 56, 1853–1860.  CrossRef CAS
First citationDe Maria, L., Vind, J., Oxenbøll, K. M., Svendsen, A. & Patkar, S. (2007). Appl. Microbiol. Biotechnol. 75, 290–300.  CrossRef
First citationDiderichsen, B., Poulsen, G. B. & Jørgensen, S. T. (1993). Plasmid, 30, 312–315.  CrossRef CAS
First citationEllis, M. V., James, S. R., Perisic, O., Downes, C. P., Williams, R. L. & Katan, M. (1998). J. Biol. Chem. 273, 11650–11659.  CrossRef CAS
First citationEmsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. (2010). Acta Cryst. D66, 486–501.  Web of Science CrossRef CAS IUCr Journals
First citationEssen, L.-O., Perisic, O., Cheung, R., Katan, M. & Williams, R. L. (1996). Nature (London), 380, 595–602.  CrossRef CAS
First citationEssen, L.-O., Perisic, O., Katan, M., Wu, Y., Roberts, M. F. & Williams, R. L. (1997). Biochemistry, 36, 1704–1718.  CrossRef CAS PubMed Web of Science
First citationFukami, K., Inanobe, S., Kanemaru, K. & Nakamura, Y. (2010). Prog. Lipid Res. 49, 429–437.  CrossRef CAS
First citationGoldstein, R., Cheng, J., Stec, B. & Roberts, M. F. (2012). Biochemistry, 51, 2579–2587.  Web of Science CrossRef CAS PubMed
First citationGriffith, O. H. & Ryan, M. (1999). Biochim. Biophys. Acta, 1441, 237–254.  CrossRef CAS
First citationHeinz, D. W., Essen, L.-O. & Williams, R. L. (1998). J. Mol. Biol. 275, 635–650.  CrossRef CAS
First citationHeinz, D. W., Ryan, M., Bullock, T. L. & Griffith, O. H. (1995). EMBO J. 14, 3855–3863.  CAS
First citationHöcker, B., Jürgens, C., Wilmanns, M. & Sterner, R. (2001). Curr. Opin. Biotechnol. 12, 376–381.  Web of Science PubMed
First citationIwasaki, Y., Niwa, S., Nakano, H., Nagasawa, T. & Yamane, T. (1994). Biochim. Biophys. Acta, 1214, 221–228.  PubMed Web of Science
First citationIwasaki, Y., Tsubouchi, Y., Ichihashi, A., Nakano, H., Kobayashi, T., Ikezawa, H. & Yamane, T. (1998). Biochim. Biophys. Acta, 1391, 52–66.  Web of Science CrossRef CAS PubMed
First citationJackson, M. R. & Selby, T. L. (2012). Acta Cryst. F68, 1378–1386.  CrossRef IUCr Journals
First citationKabsch, W. (2010). Acta Cryst. D66, 125–132.  Web of Science CrossRef CAS IUCr Journals
First citationKatan, M. (1998). Biochim. Biophys. Acta, 1436, 5–17.  CrossRef CAS
First citationKravchuk, A. V., Zhao, L., Bruzik, K. S. & Tsai, M.-D. (2003). Biochemistry, 42, 2422–2430.  CrossRef CAS
First citationKrissinel, E. & Henrick, K. (2004). Acta Cryst. D60, 2256–2268.  Web of Science CrossRef CAS IUCr Journals
First citationLarkin, M. A., Blackshields, G., Brown, N. P., Chenna, R., McGettigan, P. A., McWilliam, H., Valentin, F., Wallace, I. M., Wilm, A., Lopez, R., Thompson, J. D., Gibson, T. J. & Higgins, D. G. (2007). Bioinformatics, 23, 2947–2948.  Web of Science CrossRef PubMed CAS
First citationLebedev, A. A. & Isupov, M. N. (2014). Acta Cryst. D70, 2430–2443.  Web of Science CrossRef IUCr Journals
First citationMajerus, P. W., Connolly, T. M., Deckmyn, H., Ross, T. S., Bross, T. E., Ishii, H., Bansal, V. S. & Wilson, D. B. (1986). Science, 234, 1519–1526.  CrossRef CAS
First citationMarchler-Bauer, A. et al. (2015). Nucleic Acids Res. 43, D222–D226.  Web of Science PubMed
First citationMarques, M. B., Weller, P. F., Parsonnet, J., Ransil, B. J. & Nicholson-Weller, A. (1989). J. Clin. Microbiol. 27, 2451–2454.  CAS
First citationMcCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Storoni, L. C. & Read, R. J. (2007). J. Appl. Cryst. 40, 658–674.  Web of Science CrossRef CAS IUCr Journals
First citationMcNicholas, S., Potterton, E., Wilson, K. S. & Noble, M. E. M. (2011). Acta Cryst. D67, 386–394.  Web of Science CrossRef CAS IUCr Journals
First citationMoser, J., Gerstel, B., Meyer, J. E. W., Chakraborty, T., Wehland, J. & Heinz, D. W. (1997). J. Mol. Biol. 273, 269–282.  CrossRef CAS
First citationMurshudov, G. N., Skubák, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F. & Vagin, A. A. (2011). Acta Cryst. D67, 355–367.  Web of Science CrossRef CAS IUCr Journals
First citationNagano, N., Orengo, C. A. & Thornton, J. M. (2002). J. Mol. Biol. 321, 741–765.  Web of Science CrossRef PubMed CAS
First citationRhee, S. G., Suh, P.-G., Ryu, S. H. & Lee, S. Y. (1989). Science, 244, 546–550.  CrossRef CAS
First citationShah, A. K., Liu, Z.-J., Stewart, P. D., Schubot, F. D., Rose, J. P., Newton, M. G. & Wang, B.-C. (2005). Acta Cryst. D61, 123–129.  Web of Science CrossRef CAS IUCr Journals
First citationShaw Stewart, P. D., Kolek, S. A., Briggs, A. R., Chayen, N. E. & Baldock, P. F. M. (2011). Cryst. Growth Des. 11, 3432–3441.  CrossRef CAS
First citationSingh, A., Bhatnagar, N., Pandey, A. & Pandey, G. K. (2015). Cell Calcium, 58, 139–146.  CrossRef CAS
First citationSmith, G. A., Marquis, H., Jones, S., Johnston, N. C., Portnoy, D. A. & Goldfine, H. (1995). Infect. Immun. 63, 4231–4237.  CAS PubMed Web of Science
First citationSuh, P.-G., Ryu, S. H., Moon, K. H., Suh, H. W. & Rhee, S. G. (1988). Cell, 54, 161–169.  CrossRef CAS
First citationThorn, A. & Sheldrick, G. M. (2013). Acta Cryst. D69, 2251–2256.  Web of Science CrossRef IUCr Journals
First citationWallace, A. C., Laskowski, R. A. & Thornton, J. M. (1995). Protein Eng. Des. Sel. 8, 127–134.  CrossRef CAS Web of Science
First citationWhite, M. J., Boyd, J. M., Horswill, A. R. & Nauseef, W. M. (2014). Infect. Immun. 82, 1559–1571.  CrossRef
First citationWidner, B., Thomas, M., Sternberg, D., Lammon, D., Behr, R. & Sloma, A. (2000). J. Ind. Microbiol. Biotechnol. 25, 204–212.  CrossRef CAS
First citationWilliams, R. L. (1999). Biochim. Biophys. Acta, 1441, 255–267.  CrossRef CAS
First citationWinn, M. D. et al. (2011). Acta Cryst. D67, 235–242.  Web of Science CrossRef CAS IUCr Journals
First citationWinter, G., Lobley, C. M. C. & Prince, S. M. (2013). Acta Cryst. D69, 1260–1273.  Web of Science CrossRef CAS IUCr Journals
First citationYao, L. & Jung, S. (2010). J. Agric. Food Chem. 58, 4866–4872.  CrossRef CAS
First citationZhao, J.-F., Chen, H.-H., Ojcius, D. M., Zhao, X., Sun, D., Ge, Y.-M., Zheng, L. L., Lin, X., Li, L.-J. & Yan, J. (2013). PLoS One, 8, e75652.  CrossRef
First citationZhao, L., Liu, Y., Bruzik, K. S. & Tsai, M.-D. (2003). J. Am. Chem. Soc. 125, 22–23.  Web of Science CrossRef PubMed CAS

© International Union of Crystallography. Prior permission is not required to reproduce short quotations, tables and figures from this article, provided the original authors and source are cited. For more information, click here.

Journal logoSTRUCTURAL
BIOLOGY
ISSN: 2059-7983
Follow Acta Cryst. D
Sign up for e-alerts
Follow Acta Cryst. on Twitter
Follow us on facebook
Sign up for RSS feeds