Structure and function comparison of Micropechis ikaheka snake venom phospholipase A2 isoenzymes

Authors


Kunchithapadam Swaminathan, Institute of Molecular and Cell Biology, 61 Biopolis Drive, Singapore 138673
Fax: +65 67791117
Tel: +65 65869697
E-mail: nathan@imcb.a-star.edu.sg

Abstract

Comparison of the crystal structures of three Micropechis ikaheka phospholipase A2 isoenzymes (MiPLA2, MiPLA3 and MiPLA4, which exhibit different levels of pharmacological effects) shows that their C-terminus (residues 110–124) is the most variable. M-Type receptor binding affinity of the isoenzymes has also been investigated and MiPLA4 binds to the rabbit M-type receptor with high affinity. Examination of surface charges of the isoenzymes reveals a trend of increase in positive charges with potency. The isoenzymes are shown to oligomerize in a concentration-dependent manner in a semi-denaturing gel. The C-termini of the medium (MiPLA4) and highly potent (MiPLA2) isoenzyme molecules cluster together, forming a highly exposed area. A BLAST search using the sequence of the most potent MiPLA2 results in high similarity to Staphylococcus aureus clotting factor A and cadherin 11. This might explain the myotoxicity, anticoagulant and hemoglobinuria effects of MiPLA2s.

Abbreviation
sPLA2

secreted phospholipase A2

Secreted phospholipase A2 (sPLA2) is a stable and compact enzyme (about 125 residues) with five to eight intramolecular disulfide bridges. It is found in animal venoms and mammalian tissues [1]. It hydrolyzes phospholipids at the sn-2 position of the glycerol backbone releasing free fatty acids and lysophospholipids. Its catalytic activity is dependent on the presence of the calcium ion. In addition, sPLA2s from snake venoms have been shown to cause a wide variety of pharmacological activities, such as neurotoxic, cardiotoxic, myotoxic, anticoagulant and antiplatelet effects.

sPLA2s are divided into different groups based on their primary structures [2]. sPLA2s from group I and II are very similar in terms of their secondary and tertiary structures. Group IB sPLA2s differ from group IA sPLA2s by the presence of five extra residues forming a pancreatic loop at position 62. Group IB sPLA2s are found in both mammalian tissues and various snake venoms (Micropechis ikaheka[3], Ophiophagus hannah[4]). The structures involved in hydrolytic activity have been well studied [5]. The group IB sPLA2 structure consists of three α-helices and a distinct backbone loop (residues 12–39, Fig. 1C). The N-terminus, together with the pancreatic loop (residue 62–66), has been shown to be involved in interfacial binding [6]. A hydrophobic cleft, which holds the substrate, is also present. The outer wall of this hydrophobic channel contains the calcium binding loop (residues 25–33) and Asp49, which are essential for carrying out the phospholipase activity.

Figure 1.

(A) Amino acid sequences of the four sPLA2 isoenzymes isolated from the venom of M. ikaheka[3]. Highlighted residues have charge differences. The α-helix is represented by a cylinder and the β-strand is represented by an arrow. (B) The 2FoFc map contoured at the 1.2 σ level for residues 1–10 of the MiPLA4 molecule. (C) Comparison of the three phospholipase A2 isoenzyme (MiPLA2, MiPLA3 and MiPLA4) structures. Overlap of the Cα backbone structures is colored gray while the regions that are different are highlighted in magenta (MiPLA2), blue (MiPLA3) and green (MiPLA4). Residues that contribute to differences in charge among the isoenzymes are shown as yellow diamonds. The interfacial binding sites, hydrophobic channel and residues involved in catalytic activity are circled in blue, green and red, respectively.

Regardless of their small molecular size, toxic group IA and IB sPLA2s are able to induce different pharmacological effects in animal models. Most of the secondary structures in sPLA2 have been identified to be involved in the hydrolytic activity leaving very few sites for any pharmacological determinants (Fig. 1C). So how does sPLA2 signal for diversity? Most of the structures required for hydrolysis are α-helices and these structures have been shown to be very sensitive to residue changes [7]. In that case, are loops, which are more tolerant to mutation, the key to diversity?

Kini and Chan [8] have shown that the evolution of venom sPLA2 isoenzymes has little preference for higher toxicity and specific pharmacological activity. In a single venom, more than 10 isoenzymes can be isolated with different levels of toxicity. Therefore examination of these naturally occurring sPLA2s may lead to the identification of plausible pharmacological determinants and receptor binding sites. However, comparison of the solved sPLA2 structures deposited in the Protein Data Bank are not appropriate for analysis because the solved structures are from different snakes, have not been fully characterized or have different pharmacological effects.

The aim of this study is to compare the structures of three toxic Group IB sPLA2s, isolated from the venom of M. ikaheka (small eye snake, Papua New Guinea), with their pharmacological effects and their receptor binding affinity in order to identify the structural determinants that are likely responsible for these activities. We have isolated [3] four isoenzymes (MiPLA1, MiPLA2, MiPLA3 and MiPLA4) from the snake venom. These isoenzymes belong to the rare group IB sPLA2s with sequence identity ranging from 87 to 92% (Fig. 1A). They exhibit pharmacological effects such as hemoglobinuria (hemoglobin in the urine) in mice, myotoxicity, and anticoagulant activity at different levels (Fig. 2), thus allowing direct sequence, structural and functional comparison.

Figure 2.

(A) Hydrolytic and (B) hemoglobinuria, (C) anticoagulant and (D) myotoxicity profile of the isoenzymes MiPLA2, MiPLA3 and MiPLA4 [3]. (A) The phospholipase A2 activity was determined by titration with 20 mm NaOH at room temperature and pH 8.0. An aqueous emulsion of 20 mm phosphatidylcholine in the presence of different concentration of CaCl2 was used as substrate. (B) The potency of the isoenzymes in causing hemoglobinuria was determined by injecting different doses of isoenzymes into mice tail vein and the minimal dose that causes dark urine induction is plotted. (C) Prothrombin time assay was used to determine the anticoagulant activity. Different concentrations of isoenzymes were added to citrated plasma in the presence of Tris buffer and clotting was initiated by addition of thromboplastin with calcium (Sigma). (D) Myotoxicity was determined by measuring the creatine kinase level in the blood serum of mice injected with the isoenzymes. The creatine kinase level is increased when myonecrosis occurs.

Results and Discussion

Receptor binding studies

Several group IA and IB sPLA2s including OS1, and OS2 from the Australian Taipan snake and mammalian pancreatic sPLA2 have been shown to bind to the M-type receptor from rabbit and mouse species [9,10]. The M-type receptor has been classified as a member of the mannose receptor family. The structure of this receptor consists of a single transmembrane domain, a short cytoplasmic tail, and a large extracellular region made of an N-terminal cysteine-rich domain, a fibronectin-like type II domain and a tandem repeat of eight distinct carbohydrate recognition domains [11].

Because the diet of M. ikaheka includes rabbit and mouse, the ability of the three isoenzymes (MiPLA2, MiPLA3 and MiPLA4) to bind to the rabbit and mouse M-type receptors has been investigated by competitive binding studies with 125I-labelled OS1 to rabbit and mouse M-type receptors expressed in COS cells (Fig. 3). OS1 binds to both the receptors with very high affinities of 0.07 and 0.08 nm. MiPLA4 has binding similar to that of OS1 for the rabbit M-type receptor (IC50 = 0.11 nm) while the other isoenzymes have much weaker affinities. On the other hand, MiPLA4 displays a much lower binding affinity to the mouse M-type receptor (IC50 = 20 nm) and the other isoenzymes do not bind to this receptor. The ability of MiPLA4 to bind to the mouse M-type receptor indicates a presence of a specific receptor binding site in addition to the structures involved in hydrolysis.

Figure 3.

Competition between 125I-labelled OS1 and unlabeled isoenzymes (MiPLA2, MiPLA3 and MiPLA4) for binding to (A) rabbit and (B) mouse M-type receptor. Membranes from COS cells expressing the rabbit or mouse M-type receptor, 125I-labelled OS1 ligand and unlabeled M. ikaheka isoenzymes were incubated at 20 °C in 0.5 mL of binding buffer and filtered after 60 min through GF/C glass fiber filters presoaked in 0.5% polyethyleneimine to discard the unbound 125I-labelled OS1. (C) Tabulation of the molar concentration of isoenzymes required to replace 50% binding of 125I-labelled OS1 to the M-type receptor.

Comparison of structures

Structures of MiPLA2 (PDB code: 1PWO), MiPLA3 (PDB code: 1OZY), and MiPLA4 (PDB code: 1P7O) have been solved at 2.6, 2.7 and 2.8 Å resolution, respectively (Table 1). The β-wing (residues 74–85) and C-terminus (residues 110–124) in the three isoenzymes have the highest structural differences (Fig. 1C). The structure of the β-wing from both MiPLA2 and MiPLA4 is very similar. As the pharmacological effects and M-type receptor binding affinities of MiPLA2 and MiPLA4 are different, the highly similar β-wing structure is unlikely to contain any pharmacological determinant. Several lines of evidence suggest that the C-terminus could be pharmacologically important. First, both MiPLA1 and MiPLA2 have significantly higher levels of pharmacological effects compared to MiPLA3 and MiPLA4 (Fig. 2). Thus residues that are the same in both MiPLA1 and MiPLA2 but different to MiPLA3 and MiPLA4 are important. Residues 24, 112, 120 and 122 fall into this category and most of these residues are within the C-terminus. In addition, there is a proline bracket in the C-terminus in the most potent MiPLA2 (PYIEANNHIDP). Proline residues have been observed to flank the interaction sites and they serve to protect the integrity and conformation of the active site [12].

Table 1.  Crystal parameters, data-collection, processing and refinement statistics.
 MiPLA2MiPLA3MiPLA4
  • a

     Rsym = ΣhklΣi[| Ii(hkl) – < I(hkl) > |]/Σhkl I(hkl).

  • b

     R-factor = Σhkl||Fo(hkl)| − |Fc(hkl)||/Σhkl|Fo(hkl)|.

Unit-cell parameters
 Space groupC2P41 (hemihedral twinning)I41
 a (Å)129.18653.52288.038
 b (Å)100.64353.52288.038
 c (Å)88.307148.272262.941
 β (°)133.17  
 Matthews' coefficient3.743.793.03
 Percentage solvent66.8467.3159.12
Number of molecules in ASU426
Data collection
 X-ray sourceAPS (BM19)ALS (BL5.0.1)Spring8 (BL40B2)
 DetectorCCD SBC3CCD ADSCCCD ADSC
 Wavelength (Å)0.9780.991.00
 Resolution (Å)2.62.72.3
 Total observations53 79379 401452,468
 Unique reflections25 37111 48767,558
 Completeness (%)97.898.798.7
 Redundancy2.26.96.7
 Rsyma0.0600.0860.074
Refinement
 R-factor/Rfreeb0.22/0.270.22/0.280.22/0.27
Final model
 Non-hydrogen atoms387218805,844
 Waters32691503
 Average B-factors (Å2)
  Protein40.852.038.7
  Water62.844.169.6
 RMSD in bond length (Å)0.0100.0090.007
 RMSD in bond angles (Å)1.91.921.4

Examination of the M-type receptor binding affinity and the differences in the primary structure also suggest that the C-terminus might contain the M-type receptor-binding motif. As MiPLA4 binds to the M-type receptor while others bind very weakly, the residues in MiPLA4, which is different from all the other three isoenzymes, should be important. Residues 7, 88, 112, 115, 118, 120, 121 and 122 are such amino acids and most of them are located at the C-terminus. Lambeau et al. 1995 [9] have performed mutagenesis studies on mammalian phospholipase A2 and found that amino acid residue 7 is not important for M-type receptor binding. Residues that are close to or within the Ca2+ binding loop have been mapped to be involved in M-type receptor binding [9]. Coincidently, the C-termini of the isoenzymes are located near the Ca2+ binding loop via a disulfide bridge.

Surface charge analysis

Surface charges can alter protein to protein interactions. The front view (Fig. 4A, orientated as in Fig. 1C) of the surface residues shows that although MiPLA2 has significantly higher catalytic activity, it does not contain any unique charge difference when compared to MiPLA3 and MiPLA4 at the sites that are known to be involved in hydrolytic activity. Positive charges on sPLA2 are important for binding to the negatively charged surface of phospholipids. MiPLA2 and MiPLA3 have more positive charges than MiPLA4. MiPLA3 has the lowest enzymatic activity (Fig. 2). Comparison of MiPLA3 with MiPLA2 and MiPLA4 shows that MiPLA3 has an additional positively charged amino acid at position 56 that is located at close proximity to the catalytic site, which could interfere with the catalytic activity. However, this contradicts the mutagenesis study on bovine pancreatic sPLA2[6] that shows a reduction in catalytic activity when Lys56 is substituted by a glutamic acid. Examination of the front view of surface charges of other Group 1B sPLA2s (bovine, porcine and Ophiophagus hannah) shows that Lys56 in bovine and O. hannah sPLA2 and the equivalent Lys55 in porcine sPLA2 is either located farther or is facing away from the catalytic site.

Figure 4.

(A) Front view (arranged similar to Fig. 1C) of surface charge distributions (blue for positive and red for negative charge) of the sPLA2 proteins arranged in the order of increasing potency, (i) the least potent MiPLA3, (ii) MiPLA4 and (iii) the most potent MiPLA2. (B) Surface charge distributions viewed from the C-terminus of the sPLA2 proteins arranged in the order of increasing potency, (i) the least potent MiPLA3, (ii) MiPLA4 and (iii) the most potent MiPLA2.

Examination of the surface charge on the C-terminal face of the isoenzymes (Fig. 4B) shows a pattern of increasing number of positive charge with potency. MiPLA3 has a significantly higher number of negative charges. The MiPLA4 structure has less negative charges and two additional positive charges at positions 16 and 124 while the most potent MiPLA2 has positive charges at position 16, 122 and 124. Site-directed mutagenesis studies on the cationic and aromatic residues of the C-terminus of a myotoxic Lys49 phospholipase A2 (which has no or very low phospholipase activity) have shown that the substitution of lysine and arginine residues with alanine in the region 117–122 results in significant reduction of the myotoxic activity [13]. The C-terminal peptide (residue 115–129) of Agkistrodon piscivorus piscivorus sPLA2 LYS49 has exhibited direct lysis of skeletal muscle C2C12 cells in vitro and causes myotoxicity in vivo[14]. However, this was not observed with C-terminal peptides of ASP49 sPLA2 enzymes. Lomonte et al. [15] have also found that the C-terminus is important for the pharmacological effect. The C-terminus (residues 115–129) together with Lys36 and Lys38 of another class II sPLA2 (myotoxin II) has binding affinity to heparin of at least six repeated saccharide units. The cationic/hydrophobic surface on sPLA2 is attributed to be important for this binding.

Oligomerization of the phospholipase A2 isoenzymes

Multimerization of molecules is also different among the isoenzymes. Higher potency isoenzymes have higher levels of oligomerization. The C-termini of the molecules are placed close together forming a highly exposed area. MiPLA3, which is nontoxic, exists as dimers and the C-termini of the molecules are located far apart (Fig. 5A). MiPLA4, which has higher pharmacological activity than MiPLA3 exists as a dimer of trimers. The C-termini of the trimers are located at close proximity to each other (Fig. 5B). The most potent MiPLA2 exists as a tetramer where the C-termini form a highly exposed bulge (Fig. 5C). Kini and Chan [8] suggest that the highly exposed portion of proteins could play a critical role in ligand–receptor interactions. Clustering of the C-termini forming a highly exposed region has also been observed in other sPLA2s, e.g. three sPLA2 from the Indian cobra venom (PDB code: 1A3F, 1MH2 and 1PSH), two from Agkistrodon halys Pallas (PDB Code: 1JIA and 1B4W), one from Ophiophagus hannah (PDB Code: 1GP7) and one from the Crotalus atrox venom (PDB Code: 1PP2).

Figure 5.

Oligomerization assemblies of the isoenzymes. (A) MiPLA3 molecules pack as dimers and their C-termini are located far away from each other. (B) The stereodiagram of one of the trimers in MiPLA4. Notice that the C-termini are located close together forming a highly exposed area. (C) The most potent MiPLA2 molecule forms a dimer of dimers. The stereodiagram of the side view clearly shows that the C-termini of two monomers are protruding and located at close proximity to each other. (D) Semi-denaturing gel analysis of the isoenzymes showing concentration-dependent oligomerization. Lane 1 contains fully denatured sample, lanes 2, 3 and 4 contain semi-denatured isoenzymes at 1, 5 and 10 mg·mL−1 concentration, respectively.

Multimerization can also be a way for small proteins to increase avidity to their receptors, which have repeated units, such as heparin sulfate (contains multiple saccharide units). Oligomerization of the isoenzymes has been analyzed by gel filtration chromatography and the proteins are found to exist as monomers (data not shown). The gel filtration is only suitable for analyzing samples at less than 1 mg·mL−1. However, in a semi-denaturing SDS/PAGE gel where analysis of proteins at high concentrations is possible, concentration-dependent oligomerization has been detected (Fig. 5D). Using the semi-denatured isoenzyme at 1 mg·mL−1 concentration as a control, when the protein concentration is increased (5 and 10 mg·mL−1), the protein movement in the gel has been retarded indicating increased oligomerization with concentration. The exact molecular weight of the multimers is not known as the protein structure is not disrupted and there could be some interference from the internal charges of the protein. Previous studies have also shown that multimerization is concentration dependent [16,17]. Cobra venom sPLA2 exists as a monomer in dilute solution and forms higher order aggregates when concentrated on the surface of a phospholipid substrate. It is difficult to distinguish whether the monomer or the aggregated form is the representation of the functional state. The crystal structure might thus indicate the possible arrangement at high concentrations on the surface of target cells and also in the highly concentrated venom.

Implication of pharmacological effects

Pharmacological effects can be caused by the products released by the phospholipase activity or by specific pharmacological determinants (which are totally independent of hydrolysis) or is a combined result. Therefore, we examined the pharmacological characteristics of the isoenzymes in relation to their hydrolytic activity to identify the presence of any plausible pharmacological determinants. Gao et al. [3] had attempted to identify the relationship of the catalytic activity to the pharmacological effect by chemical modification of His48 using p-bromophenacyl bromide. The chemical modification stopped all pharmacological effects but the effect could not be attributed to the modification of His48 alone as there are other histidines (in the most potent isoenzyme MiPLA2 positions 48 and 117 are histidines). Therefore, the relationship of the pharmacological effects and the hydrolytic activity cannot be verified by this method. In our attempt to verify the presence of a pharmacological site in the MiPLA isoenzymes, we observe the relative differences in their hydrolytic activities and the pharmacological effects of the isoenzymes (Fig. 2). If the pharmacological effects are the direct result of hydrolytic activity, then the relative difference in the pharmacological activities and hydrolytic activity among the isoenzymes should be the same. However, this is not true among the isoenzymes and is especially obvious in MiPLA1. The isoenzyme is less active than MiPLA2 and MiPLA4 in the hydrolytic assay, but in the prothrombin assay (testing for anticoagulant activity) and muscle necrosis assay (testing for myotoxicity), MiPLA1 and MiPLA2 are far more active than both MiPLA3 and MiPLA4. Also, these sPLA2s have the unique ability to cause hemoglobinuria, which has not been observed for sPLA2s from other snake venoms even with high hydrolytic activity. This evidence indicates that the isoenzymes serve as good models for the identification of a pharmacological site.

BLAST search using residues 110–124 of MiPLA2 identified similar motifs on Clotting factor A (clf A) of Staphylococcus aureus, the cadherin 11 molecule and hemoglobin α-chain (Table 2). The motif on cadherin 11 is conserved in mouse, rat, Homo sapiens, chick and Xenopus, while the motif on hemoglobin α-chain is conserved across birds (turkey, chicken gray francolin, emperor penguin, pigeon, blue and yellow macaw, black-head gull, great cormorant, magpie goose, turtle dove, Japanese quail, etc).

Table 2.  BLAST search for motifs similar to MiPLA2 C-terminal residues 110–122 (conserved residues are in bold).
ProteinResiduesProtein sequence
MiPLA2 PYI_EANN_HIDPNR
Clotting factor A (Staphylococcus aureus)486–498PYIVVVNG_HIDPN
Cadherin 11 (mouse)347–356  I_EAANVHIDP
Hemoglobin α-chain (turkey)67–74  I_EAAN_HID

Clotting factor A (Clf A) of S. aureus is known to attach to fibrinogen. Fibrinogen plays central roles in hemostasis and coagulation. Coincidently, the MiPLA2 C-terminal motif is similar to the minimal fibrinogen-binding region (residues 221–559) of Clf A [18]. This region of MiPLA2 is also similar to the extracellular domain III of cadherin 11, which is present on the surface of mesodermal cells. The cadherin 11 molecule is used to assemble nearby mesodermal cells through their inter-cadherin interactions. Residual mesodermal cells from the developmental process are still present in the adult organs such as, blood, kidney and muscle [19]. Mesenchyme cells in the adult kidney and muscle do not play any significant functional role. However, the kidney has a repair scheme whereby epithelial cells can revert to mesenchyme cells when damage is detected [20]. This scheme could be exploited by the destructive mechanism of MiPLA2, which provides more target cells for MiPLA2. The sequence similarity of MiPLA2 to Clf A and cadherin 11 might explain for the anticoagulant and hemoglobinuria effects, respectively. Gao et al. [3] have shown that MiPLA2 does not cause direct hemolysis and therefore damages in the kidney are postulated. The binding activity of both Clf A and cadherin 11 requires calcium ions. Coincidentally, the C-terminus in all sPLA2 has been observed to be located very close to the calcium binding loop via a disulfide bridge between the cysteine residues 27 and 123.

BLAST search has also identified the C-terminal sequence to be similar to the hemoglobin α-chain of birds. However, there is no recorded interaction between hemoglobin and sPLA2 and pharmacological effects of MiPLA2 have not been tested on birds.

Experimental procedures

Purification of MiPLA2, MiPLA3 and MiPLA4

The isoenzymes have been purified as reported earlier [3]. The venom of M. ikaheka was resuspended in 50 mm Tris/HCl (pH 7.4) and passed through a Superdex 30 gel filtration column (Pharmacia). Fractions containing proteins with molecular masses around 14 000 Da were then passed through the Uno-S column (Biorad). MiPLA4 did not bind to the Uno-S column and therefore was present in the flow-through fraction. MiPLA2 and MiPLA3 were eluted by using a linear salt gradient with 50 mm Tris/HCl (pH 7.4), 1 m NaCl as the elution buffer and later separated using a C18 reverse-phase column with a linear gradient of elution buffer (80% acetonitrile). All isoenzymes were eluted from the column between 20 to 30% elution buffer. The flow through fraction containing MiPLA4 was dialyzed in 50 mm sodium acetate (pH 5.0), reapplied to the Uno-S column and eluted using a linear salt gradient with 50 mm sodium acetate (pH 5.0), 1 m NaCl. All the proteins were concentrated to more that 10 mg·mL−1 and stored at −80 °C.

M-Type receptor binding studies

Competition binding assays between unlabeled isoenzymes (MiPLA2, MiPLA3 and MiPLA4) and iodinated OS1 were performed on recombinant rabbit and mouse M-type receptors expressed in COS cells. Briefly, membranes from COS cells expressing the rabbit or mouse M-type receptor, 125I-labelled OS1 ligand and unlabeled M. ikaheka isoenzymes were incubated at 20 °C in 0.5 mL of binding buffer [140 mm NaCl, 0.1 mm CaCl2, 20 mm Tris/HCl (pH 7.4) and 0.1% bovine serum albumin]. Incubations were initiated by addition of membranes and filtered after 60 min through GF/C glass fiber filters presoaked in 0.5% polyethyleneimine [21]. 125I-labelled OS1 and membranes expressing M-type receptors were prepared as described previously [22].

Crystallization and data collection

Crystals of these sPLA2s were first observed in the ammonium sulfate [(NH4)2SO4] grid screen (Hampton Research) by using the sitting drop vapor diffusion method at 293 K. Two microliters of the protein was mixed with 2 µL of precipitant solution. After optimization, the following conditions formed suitable crystals for X-ray diffraction studies: MiPLA2: 0.1 m Mes (pH 6.5), 2.4 m (NH4)2SO4; MiPLA3: 0.1 m Mes (pH 6.5), 1.8 m (NH4)2SO4; MiPLA4: 0.1 m Tris/HCl (pH 8.5), 2 m (NH4)2SO4. The same precipitating solutions with 25% glycerol were used as the cryo-protectant for freezing the PLA2 crystals in liquid nitrogen. The X-ray diffraction data were collected using synchrotron radiation and processed using denzo, hkl2000 and scalepack[23].

Structure determination and refinement

All three structures were determined by the method of molecular replacement using the molrep program [24]. To solve the structure of MiPLA3, bovine sPLA2 (PDB code: 1UNE) was used as the initial search model. The MiPLA3 crystal was later found to be hemihedrally twinned (twinning fraction: 0.474). Model building and refinement were carried out using the o[25] and cns[26] programs, respectively. The MiPLA4 structure was solved using MiPLA3 as the search model and the MiPLA2 structure was in turn solved by using the MiPLA4 molecule. Geometry of the structures was checked with procheck[27]. Table 1 shows the data and refinement statistics and Fig. 1(B) shows the quality of the electron density around residues 1–10 of MiPLA4.

Semi-denaturing gel analysis of oligomerization

This method has been described previously [28]. Proteins were analyzed in both denaturing and semi-denaturing conditions. Under the denaturing condition, samples were exposed to sodium dodecyl sulfate and dithiothreitol, and boiled. In the semi-denaturing condition, samples (at different concentrations 1, 5 and 10 mg·mL−1) were only exposed to SDS and not boiled. All samples were analyzed using SDS/PAGE.

Acknowledgements

We thank R.M. Kini, National University of Singapore for his critical review of the manuscript and Zhengjiong Lin and Lichuan Gu for their help and advice in the refinement of the MiPLA3 structure. This work was supported by the funding of Agency for Science, Technology and Research (A*STAR), Singapore to KS.

Ancillary