SEARCH

SEARCH BY CITATION

Keywords:

  • Mycobacterium tuberculosis;
  • peptide deformylase;
  • thermostability;
  • oxidative-stress stability;
  • site-directed mutagenesis

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Peptide deformylase (PDF) catalyses the removal of the N-formyl group from the nascent polypeptide during protein maturation. The PDF of Mycobacterium tuberculosis H37Rv (MtbPDF), overexpressed and purified from Escherichia coli, was characterized as an iron-containing enzyme with stability towards H2O2 and moderate thermostability. Substitution of two conserved residues (G49 and L107) from MtbPDF with the corresponding residues found in human PDF affected its deformylase activity. Among characterized PDFs, glycine (G151) in motif III instead of conserved aspartate is characteristic of M. tuberculosis. Although the G151D mutation in MtbPDF increased its deformylase activity and thermostability, it also affected enzyme stability towards H2O2. Molecular dynamics and docking results confirmed improved substrate binding and catalysis for the G151D mutant and the study provides another possible molecular basis for the stability of MtbPDF against oxidizing agents.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Proteins evolve by rare mutations that provide functional innovations without affecting the pre-existing global structure and activity (Bowie et al., 1990). As beneficial mutations are rare, the ability of an enzyme to accumulate sequence changes and maintain the required activity for better survival of the host organism is an important aspect of its evolvability (Woycechowsky et al., 2008).

The emergence of multiple drug-resistant strains of Mycobacterium tuberculosis, a synergy between HIV and M. tuberculosis infection, and a need for shortened chemotherapy for tuberculosis treatment have increased the demand for improved drugs with alternative targets.

Peptide deformylase (PDF; EC 3.5.1.31), encoded by the def gene, catalyses the removal of the formyl group from N-terminal methionine following translation. This enzyme, present in all eubacteria and in eukaryotic organelles, is a potential target for discovery of antibacterial agents (Guay, 2007). Its essentiality for survival has been demonstrated for many bacteria, including Mycobacterium bovis (Teo et al., 2006). Most of the PDF inhibitors available are derivatives of the natural deformylase inhibitor actinonin, and many, such as LBM-415, have progressed to preclinical and clinical stages of development (Chen et al., 2000; Butler & Buss, 2006). However, the published structural evidence for similar binding of actinonin to human PDF has complicated the whole drug discovery process based on PDF (Escobar-Avarez et al., 2009). Thus, the available sequence variations between bacterial and human PDFs need to be explored further to identify structural variations between the two for designing novel PDF inhibitors. Characterizing the amino acid sequence variations between the PDF of M. tuberculosis (MtbPDF) and other PDFs might help us to design specific inhibitors targeting MtbPDF.

Here recombinant MtbPDF and its selected substitution mutants were characterized to study the properties of this enzyme and to define the role of substituted residues in its activity and stability.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

All the routine chemicals, reagents, substrates, culture media and antibiotics were purchased from Sigma-Aldrich. PCR primers were obtained from Integrated DNA Technologies. Mycobacterium tuberculosis H37Rv genomic DNA was obtained from Colorado State University. All DNA manipulations were performed using standard protocols (Sambrook et al., 1989).

Cloning of the def gene and construction of site-directed mutants

The def gene (Rv0429c; 594 bp) was PCR-amplified from genomic DNA of M. tuberculosis H37Rv using specific primers (see Supporting information, Table S1) and was cloned into pET28a vector (Novagen) with the N-terminus His-tag. For creating substitution mutants of recombinant MtbPDF, internal primers having corresponding mutations were designed (Table S1). Site-directed mutagenesis was performed on the def∷pET28a construct using the Quick-Change Mutagenesis kit (Stratagene, Germany). All the mutations were confirmed by DNA sequencing (MWG, Bangalore, India).

Expression, purification and refolding of MtbPDF and its mutants

Expression, purification and refolding of recombinant MtbPDF and mutants were performed from Escherichia coli BL21 (DE3) (Invitrogen) as previously reported (Saxena & Chakraborti, 2005a). The protein fraction extracted in 3 M urea buffer was diluted to a final concentration of 0.3 mg mL−1 with 20 mM phosphate buffer, pH 7.4, containing 10 μg mL−1 catalase and 0.2 mg mL−1 bovine serum albumin, prior to refolding by dialysing against 20 mM phosphate buffer, pH 7.4. The refolded proteins were passed through an Ni-NTA column (Qiagen, Germany) and were eluted with 250 mM imidazole. The metal contents of purified recombinant proteins were analysed by atomic absorption spectroscopy (AAS), without any additional incubation with metal ions (Meinnel et al., 1997).

PDF assay

The deformylase assay of MtbPDF and its variants was determined using 73.3 nM enzyme with 2,4,6-trinitro benzene sulfonic acid (TNBSA) as the reagent, as reported elsewhere (Saxena & Chakraborti, 2005a). Deformylase activities were expressed as micromolar free amines produced per minute per milligram of protein. Deformylase activity assays of MtbPDF and its variants were performed on different substrates (N-formyl-Met-Ala-Ser, N-formyl-Met-Leu-Phe and N-formyl-Met) at different conditions. Km and Vmax were determined from slopes of various concentrations of substrate by applying a nonlinear curve fit. Kinetics analysis was performed using graphpad prism version 5.0 (Graphpad software).

Circular dichroism (CD) spectroscopy

The CD spectrum of purified MtbPDF, G151D and G151A proteins were recorded in a Jasco J-810 (Jasco, Japan) spectropolarimeter in the far-UV region (190–300 nm). CD spectroscopy was performed using 0.1 mg mL−1 purified proteins in 20 mM phosphate buffer, pH 7.4, at 25 °C using a cell with path length of 1 cm (Saxena et al., 2008). Each spectrum represented is the average of three separate scans.

Bioinformatics, molecular dynamics (MD) simulations and molecular docking

Multiple alignments of MtbPDF sequences with other bacterial and human PDFs were performed using the clustalw program (http://www.ebi.ac.uk/clustalW/index.html) (Thompson et al., 1997).

The high-resolution (15.6 nm) crystal structure of MtbPDF was retrieved from the Protein Data Bank (PDB ID: 3E3U) (Pichota et al., 2008), and the G151D structure was generated using the program modeller9v6 (http://salilab.org/modeller/) (Fiser & Sali, 2003). The Ni2+ in the crystal structure was replaced with the native metal cofactor of MtbPDF, Fe2+ (Saxena et al., 2008). For all energy minimization and MD calculations, an AMBER03 force field in conjunction with Visual Molecular Dynamics/NAMD program (Humphrey et al., 1996; Phillips et al., 2005) was employed.

Flexible small molecule-rigid protein docking experiments were performed using autodock 4.0 (Morris et al., 1998) with default parameters. The energy-minimized MtbPDF and G151D structure was used with the substrate, N-formyl-Met-Ala-Ser, prepared and geometrically optimized using arguslab (http://www.arguslab.com).

Results and discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Purification and catalytic properties of MtbPDF and site-directed mutants

Based on multiple alignments of the MtbPDF sequence with other characterized PDFs, three residues from the three conserved motifs were selected for site-directed mutagenesis (Fig. 1a). Two of the mutants, L107E and G49C, substituted MtbPDF residues with corresponding residues found in human PDF. G49P was created as a comparison for G49C mutation. Glycine in motif III of MtbPDF was unique to M. tuberculosis among the characterized PDFs, including human PDF. G151D and G151A mutants were created to study the role of this glycine in MtbPDF.

image

Figure 1.  Amino acid sequence variations and site-directed mutagenesis of MtbPDF. (a) Multiple sequence alignment of different prokaryotic and human PDFs with MtbPDF. The three conserved motifs of deformylase are labelled above. The three arginines in the insertion region are shown with an open circle. The residues selected for mutagenesis are shown with an asterisk. (b) Purified and refolded MtbPDF and its site-directed mutants on 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis.

Download figure to PowerPoint

The purified MtbPDF and mutants showed an apparent molecular weight of 29 ± 1 kDa on 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis, as compared with the calculated molecular weight of 22.5 kDa (Fig. 1b). These anomalous migrations have been reported previously for many bacterial PDFs and have been correlated with high proline contents in PDF sequences (Han et al., 2004; Saxena & Chakraborti, 2005a).

Substrate specificity of purified MtbPDF with the tested substrates was in the order N-formyl-Met-Ala-Ser>N-formyl-Met-Leu-Phe>N-formyl-Met (Fig. 2a). All further deformylase assays were carried out using N-formyl-Met-Ala-Ser as the substrate, unless mentioned otherwise. The kinetic parameters for MtbPDF are summarized in Table 1.

image

Figure 2.  Deformylase activity of MtbPDF and its mutants. (a) Substrate specificity of MtbPDF (open bars) and G151D (shaded bars): 73.3 nM of protein was used with 5 mM of each of the substrates (NfMAS: N-formyl Met-Ala-Ser; NfMLF: N-formyl Met-Leu-Phe; NfM: N-formyl Met) at 30°C and pH 7.4. (b) Deformylase activity of mutant proteins against 5 mM NfMAS in a standard 2,4,6-trinitro benzene sulfonic acid assay. (c) Optimum temperature and (d) pH for deformylase activity of MtbPDF (♦) and G151D (▪).

Download figure to PowerPoint

Table 1.   Kinetic parameters for MtbPDF and its active mutants towards N-for-Met-Ala-Ser
ProteinVmax (μM min−1)Km (mM)Kcat (s−1)Kcat/Km (M−1 s−1)
  1. Kinetic parameters were calculated from continuous TNBSA assay involving 0–20 mM NfMAS and 73.3 nM proteins. Values shown are mean ± SD.

PDF15.8 ± 1.24.3 ± 0.93.6 ± 0.6842 ± 9
G151D19.8 ± 0.72.5 ± 1.14.5 ± 0.61786 ± 19
G151A15.6 ± 0.44.4 ± 0.73.5 ± 0.7795 ± 8
G49C8.9 ± 1.07.8 ± 0.82.0 ± 0.3260 ± 9

Among the mutants corresponding to human PDF, G49C retained nearly 36.1 ± 9% activity of MtbPDF, while the G49P mutant was almost completely inactive. L107E retained <10% activity of MtbPDF (Fig. 2b). In the PDF crystal structures both these residues were found to have a role in maintaining the architecture of the peptide binding pockets (Meinnel et al., 1997; Nam et al., 2009). In the MtbPDF structure, G49 and L107 occupy similar positions (Pichota et al., 2008). Substitution at these positions with residues found in human PDF (C49 and E107) might have disturbed the architecture of the substrate binding pocket in MtbPDF.

The G151D mutant showed 1.5 times the activity of MtbPDF against N-formyl-Met-Ala-Ser with a Kcat/Km value of 1786 ± 19 M−1 s−1 (Fig. 2b; Table 1). Catalytic properties of G151D suggested an improved substrate affinity compared with MtbPDF, as evident from the decreased Km values. There was also a significant increase in Kcat for G151D (Table 1). The G151A mutant showed similar catalytic properties as MtbPDF (Fig. 2b; Table 1).

G151D also deformylated N-formyl-Met-Leu-Phe with higher efficiency than MtbPDF (Fig. 2a), which suggests a possible increase in space within the substrate binding site at inline image and inline image positions to accommodate the bulkier side chains on the substrate.

Biochemical properties of MtbPDF and G151D

The optimum temperature and pH for deformylase activity of MtbPDF was 20–30 °C and pH 7.4 (Fig. 2c and d). However, G151D showed twofold higher activity at 50 °C compared with MtbPDF activity at 30 °C. Similarly, the pH optimum for G151D activity was shifted towards 5.5 (Fig. 2c and d). Note that the temperature optimum for deformylase activity of MtbPDF, which is lower compared than all reported PDFs (Bracchi-Ricard et al., 2001; Han et al., 2004), showed a dramatic shift to higher values upon introduction of aspartate in motif III. This highlights the importance of the residue at this position in modulating the thermostability of PDFs. Similarly, the reported ranges of pH optima for deformylase activity of E. coli and Plasmodium falciparum PDFs were 5.5–7.0, with only a slight decrease in activity in the basic range up to pH 9.0 (Rajagopalan et al., 1997a; Bracchi-Ricard et al., 2001). Only a single ionization event (pKa∼5.2) has been assigned to the deprotonation of the metal-bound water/glutamate network in previously studied PDFs, which led to a flat pH profile in the basic range (Rajagopalan et al., 1997a; Bracchi-Ricard et al, 2001). The pKa values for catalytic E149 in the MtbPDF and G151D were predicted by the H++ server as 6.48 and 4.88, respectively, supporting our experimental findings. The optimum temperature (30 °C) and pH (7.4) of MtbPDF was used in all further comparative studies.

MtbPDF was stable at 30 °C with a half-life (t1/2) close to 4.5 h. At 40 °C t1/2 was reduced to 90 min and at 50 °C to 40 min (Fig. 3a). The temperature stability of MtbPDF at 30 °C in our studies was very similar that reported by Saxena et al. (2008), indicating the consistency in enzyme preparations. However, G151D was very stable at 30 °C with little loss of activity up to 6 h. The t1/2 of G151D at 40 °C was >6 h and at 50 °C was 2 h (Fig. 3a). This increase in thermostability was specific for G151D and was absent for G151A (data not shown). Thermostability of a mutant protein reflects the enhanced stability of the structure induced by the mutation.

image

Figure 3.  Comparison of biochemical and biophysical properties of MtbPDF and G151D. (a) Thermostability of MtbPDF and G151D: MtbPDF at 30°C (▴), 40°C (▪), 50°C (○); G151D at 30°C (□), 40°C (diagonals), 50°C (verticals). Enzyme activity at time zero is represented as 100%. (b) Stability of MtbPDF (◆) and G151D (▪) towards H2O2: enzyme activity in the absence of H2O2 is represented as 100%. (c) Inhibition of deformylase activity of MtbPDF (◆) and G151D (▪) by actinonin: 73.3 nM of both enzymes were incubated with 0–10 μM actinonin for 15 min before initiating deformylase assay. Uninhibited control is represented as 100%. (d) Far-UV CD spectrum of MtbPDF, G151D and G151A.

Download figure to PowerPoint

The susceptibility of Fe2+-containing PDFs to oxidation has been established from studies on E. coli and Haemophillus infuenzae PDFs (Rajagopalan et al., 1997b; Rajagopalan & Pei, 1998). The mechanism reported was oxidation of Fe2+ to Fe3+ and/or oxidation of Sγ in metal-coordinating cystein. AAS revealed Fe as a major metal ion in MtbPDF (0.72 ± 0.21 g-atoms Fe g−1 protein) and G151D (0.69 ± 0.23 g-atoms Fe g−1 protein), as reported elsewhere (Saxena & Chakraborti, 2005a). In our inhibition assay, MtbPDF retained 30% activity after incubation with 500 mM H2O2 for 30 min (Fig. 3b). The results on H2O2 stability vary considerably compared with the previous report on MtbPDF (Saxena & Chakraborti, 2005a, b), where the presence of 500 mM H2O2 for 30 min did not affect the deformylase activity. However, our results also suggest that MtbPDF is resistant to oxidative stress, as there was a >1000-fold increase in resistance compared with previously characterized Fe2+-containing E. coli PDF (Rajagopalan et al., 1997b). Interestingly, G151D completely lost its activity upon incubating with 200 mM H2O2 (Fig. 3b). Thus, the increase in thermostability of G151D was accompanied by a decrease in oxidative stress resistance.

The enzyme activity of MtbPDF was completely inhibited by 5 μM of the deformylase inhibitor actinonin, with an IC50 of 120 nM. Under similar assay conditions, G151D was completely inhibited with 10 μM of actinonin with an IC50 of 800 nM (Fig. 3c). This increase in IC50 of actinonin is a reflection of improved substrate affinity in the case of G151D. Other known metalloprotease inhibitors such as bestatin and amastatin did not produce any inhibitory effects in either case (data not shown).

CD spectroscopy of MtbPDF and G151D

To analyse any possible secondary structure alterations induced by substitutions, the CD spectra of MtbPDF, G151D and G151A were compared. The far-UV-CD spectrum of MtbPDF had two typical negative minima at 208 and 222 nm with a crossover point at 198 nm (Fig. 3d), indicating the presence of sheets and coils in addition to the predominant helical structure. The CD spectra of G151D showed a considerable amount of scatter to low mean residue ellipticity (approximately 30%; Fig. 3d). However, no shift in the negative minima at 222 or 208 nm was observed. These results indicated that the G151D mutation produced only restructuring in the less stable scaffolds such as turns and 310 helices, without affecting the α-helical fold. However, the CD spectrum of G151A was almost completely superimposable on that of MtbPDF.

MD simulations of MtbPDF and G151D structures

The overall structure and stability of MtbPDF and G151D were examined by MD simulation. In the G151D model, D151 was not a part of the catalytic site and was located >50 nm from the metal ion (Fig. S1). The main chain root mean square deviation (RMSD) profile for the two structures (Fig. 4a) showed that G151D reached a flat profile after ∼100 ps whereas MtbPDF showed a variable profile during the entire simulation period. This demonstrated the higher stability of the G151D structure compared MtbPDF. The root mean square fluctuation (RMSF) plot of MtbPDF showed higher fluctuations in Loop 1 (T22–D30) and the C-terminal loop (D191–H197) compared with G151D, whereas the latter showed greater fluctuations in Loop 6 (E91–T95) (Fig. 4b). The MtbPDF structure contains three α-helices, seven β-sheets and three 310 helices, forming three motifs and a structurally conserved active site (Pichota et al., 2008). Both MtbPDF and G151D had comparable secondary structures except that, in the latter, the first two 310 helices (12PVL14 and 53ANQI56) were transformed into turns. Additionally, the helix H1 started from A31 in G151D instead of D32 in MtbPDF. The CD spectrum of G151D reflected these variations in the less stable scaffold.

image

Figure 4.  MD simulations and docking studies on MtbPDF and G151D structures. (a) RMSD from energy-minimized starting structure of MtbPDF (magenta) and G151D (blue) during dynamics for 600 ps. (b) Cα-RMSF plot of MtbPDF (magenta) and G151D (blue) structures during dynamics for 600 ps. (c) Movements of side chain atoms of R77–R79 residues in the insertion loop with reference to metal cofactor Fe2+ in MtbPDF (green sticks) and G151D (pink sticks) structures during dynamics. (d) Substrate binding sites of MtbPDF (cyan) and G151D (green) with substrate N-for-Met-Ala-Ser docked in (pink sticks). The hydrogen bonds stabilizing the substrates in the cavity are shown as yellow dotted lines.

Download figure to PowerPoint

The residues vital for metal binding and catalysis (Q56, C106, H148, E149 and H152) were within 30 nm around the metal ion. MD simulations of MtbPDF and G151D structures revealed no significant differences in the positioning of metal-binding residues and their average distance from the Fe2+ ion. This supports the equal Fe content in MtbPDF and G151D, as seen from the AAS results.

The side chains of residues lining the substrate-binding cavity (G49, V50, G51, E104, G105, C106, L107, R144 and M145) of G151D showed slight fluctuations in positioning compared with MtbPDF. The average distance between side chain atoms of M145 with L107 in G151D was increased by 20 nm compared with MtbPDF (Fig. S2). Similarly, the distance between side chain atoms of G49, V50 and G51 with those of 104EGCL107 was increased by 5–10 nm in G151D (Fig. S3). These differences might have contributed to the increase in space within the peptide binding pocket of G151D. These differences were reported to be decreased in the R77-79K mutation of MtbPDF, leading to a reduction in size of the substrate binding site (Saxena et al., 2008).

Three arginines in the insertion sequence (77RRR79) (Fig. 1a) of MtbPDF were reported to be responsible for the observed resistance to oxidative stress (Saxena et al., 2008). The higher sensitivity of the G151D mutant to oxidizing agents led us to look into the structural variations in the loop containing three arginines. During MD simulations, the side chain of R77 in G151D was displaced by 35 nm from Fe2+, losing its stabilization from hydrogen bonding with side chain atoms of D128 (Fig. 4c). This destabilizes the loop containing three arginines, which was reported to interact with the core helix in MtbPDF to provide oxidative stress stability. The predicted mechanism of this interaction was an ‘action-at-distance’, in which the R77-79 present in the loop away from the active site modulates the thermostability and resistance to H2O2 in MtbPDF. Although the arginine side chains are reported to interact and scavenge oxygen (Saxena et al., 2008), the actual mechanism by which these residues prevent Fe2+ and/or metal-coordinating cystein from oxidizing is still not clear. In G151D, destabilization of the loop containing three arginines might have led to increased oxidation of Fe2+ and/or metal-coordinating cystein. More systematic studies on this property would unveil the underlying mechanism of action.

Molecular docking

The free energy of binding of substrate N-formyl-Met-Ala-Ser into MtbPDF was −6.34 kcal mol−1 and for G151D was −7.25 kcal mol−1. Superimposition of the two docked structures indicated that the positioning of residues at the P′ and inline image position of the substrate (formyl group and Met) was essentially the same in both cases. But residues at the inline image and inline image positions of the substrate (Ala and Ser) were better aligned in G151D than in MtbPDF (Fig. 4d). In MtbPDF, the lowest energy complex had substrate rotated almost 90° between inline image and inline image in order to accommodate the side chains of residues at inline image and inline image in the pocket. In the MtbPDF pocket, a single hydrogen bonding between CO of G105 and NH of substrate Met stabilized the substrate, whereas in the G151D pocket, substrate binding was stabilized by increased hydrogen bonding interactions such as the one between NH of substrate Ala and CO of G105, between NH of substrate Met and Nɛ2 of H148, and between OH of substrate Ser and NH of E104 (Fig. 4d). Docking results provided additional evidence for increased space in the peptide binding pocket of G151D, leading to a stable substrate binding environment compared with MtbPDF.

The available variations in sequence and properties of bacterial enzymes compared with their human counterparts will need to be explored for further improvements in inhibitor screening against PDF. The present study explored such sequence variations and highlighted an additional molecular basis for oxidative stress stability in MtbPDF. It was concluded that an aspartate residue in motif III of PDFs plays important role in providing stability to the enzyme and in modulating the protonation of catalytic glutamate side chains. The presence of glycine instead of conserved aspartate in MtbPDF reduces its thermostability, but provides better resistance to oxidative stress, which might be essential for better survival of the organism in the oxidative environment. The present study also describes the subtle variations in the peptide binding pocket of the enzyme associated with the above mentioned substitution, which could be further explored to design inhibitors with specificity towards MtbPDF. Pinpointing the molecular basis of oxidative stress resistance of MtbPDF will provide further opportunities to design mechanistically based inhibitors targeting MtbPDF.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

K.M.N. acknowledges the Department of Biotechnology (DBT), New Delhi, India, for the research grant. S.S.N. thanks CSIR, India, for SRF. We also thank Mr Jino George, Photochemistry division, NIIST, for assistance with CD spectroscopy.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Fig. S1. Superimposed cartoon models of MtbPDF and G151D structures, in complex with substrate N-for-Met-Ala-Ser.

Fig. S2. Distance between side chain atoms of L107 with side chain atoms of R144 and M145 delineating substrate binding site of MtbPDF and G151D structures.

Fig. S3. Distance between side chain atoms of G49, V50 and G51 with side chain atoms of 104EGCL107 delineating the substrate binding site of MtbPDF and G151D structures.

Table S1. Primers used in the study.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
FML_2289_sm_suppl-information.doc2542KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.