SEARCH

SEARCH BY CITATION

Keywords:

  • penicillin-binding protein;
  • dd-carboxypeptidase;
  • Escherichia coli ;
  • beta-lactam antibiotics;
  • deacylation;
  • modeller

Abstract

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

Of the five dd-carboxypeptidases in Escherichia coli, only PBP5 demonstrates its physiological significance by maintaining cell shape and intrinsic beta-lactam resistance. DacD can partially compensate for the lost beta-lactam resistance in PBP5 mutant, although its biochemical reason is unclear. To understand the mechanism(s) underlying such behaviour, we constructed soluble DacD (sDacD) and compared its biophysical and biochemical properties with those of sPBP5, in vitro. Unlike sPBP6, sDacD can deacylate Bocillin significantly, which is very similar to sPBP5. sDacD shows weak dd-carboxypeptidase activity, although lower than that of sPBP5. Bioinformatics analyses reveal a similar architecture of sPBP5 and sDacD. Therefore, based on the obtained results we can infer that biochemically DacD and PBP5 are more closely related to each other than to PBP6, enabling DacD and PBP5 to play a nearly similar physiological function in terms of recovering the lost beta-lactam resistance.


Introduction

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

Of the five dd-carboxypeptidases (DD-CPases) in Escherichia coli, PBP4, PBP5, PBP6, DacD and AmpH (Holtje, 1998; Ghosh et al., 2008; Sauvage et al., 2008; Gonzalez-Leiza et al., 2011), only PBP5 has been studied thoroughly concerning enzymology, structure and physiological aspects (Nelson & Young, 2000; Nelson & Young, 2001; Chowdhury et al., 2010; Sarkar et al., 2010, 2011). However, other DD-CPases are mostly characterized on the basis of their kinetic properties in vitro (Korat et al., 1991; Chowdhury et al., 2010; Gonzalez-Leiza et al., 2011), the structure–function relationship drawn based on the crystal structure (Davies et al., 2001; Kishida et al., 2006; Chen et al., 2009) and by molecular modelling (Chowdhury et al., 2010; Chowdhury & Ghosh, 2011). Therefore, in vitro enzymatic activities of DD-CPases with their physiological functions have rarely been correlated.

Likewise, little is known about E. coli DacD (PBP6b), a product of dacD gene, which is expressed at the mid-logarithmic phase (Baquero et al., 1996). DacD shares 47% amino acid (aa) identity with PBP5 sequence (Sarkar et al., 2011; Baquero et al., 1996). Unlike PBP6, overexpression of DacD can partially compensate the lost beta-lactam resistance in PBP5 deletion mutants (Sarkar et al., 2011). In addition, DacD overexpression reduces the proportion of muramyl pentapeptides in vivo (Baquero et al., 1996). It is therefore plausible that the above-mentioned functions of DacD are mediated through a similar enzymatic action as that of PBP5. To corroborate the assumption in vitro, the biochemical properties of DacD need to be evaluated. To address this, we have constructed in this study, a soluble cytoplasmic form of DacD (sDacD) and assessed the enzymatic activity in order to correlate this with the observed physiological properties. The structure–function relationship of sDacD was further investigated by bioinformatics analyses to determine the basis of its differential behaviour from its nearest homolog.

Materials and methods

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

Bacterial strains, plasmids and chemicals

Escherichia coli BL21 star (Stratagene, TX) was used to express recombinant proteins for purification. Escherichia coli CS109 (K12 variant) (Denome et al., 1999) was used for sDacD gene amplification. Unless otherwise specified, enzymes for molecular analysis were purchased from New England Biolabs (Ipswich, MA) and other reagents from Sigma-Aldrich, (St. Louis, MO).

Materials for in silico analysis

DacD sequence (aa) was obtained from NCBI (accession no. AAC75071.2). However, the sDacD sequence was used for model building. The software used for in silico analysis is described in the section 3D model building.

Construction of sDacD

The gene of sDacD was amplified using oligonucleotide primers (5′-CTCTCTGGATCCATGGCGGAAAACATTCCTTTTTCACCTCAGCC-3′ and 5′- CTCTCTAAGCTTTCAATAATCACTCAGGCGAGAAAACATGCTGCC-3′) in such a way that the resulting gene would express protein devoid of its signal peptide (21 aa from N-terminus) and C-terminal amphipathic anchor (5 aa). The conditions for amplification with Deep vent DNA polymerase was: 94 °C for 5 min (initial denaturation), 94 °C for 1 min, 60 °C for 1 min and 72 °C for 1 min (for 30 cycles), followed by a final extension of 72 °C for 7 min. The amplicon (1.1 kb) was cloned in pT7-7K vector (Tabor & Richardson, 1985) at the BamH1 and HindIII sites (underlined) creating the plasmid, pTADacD-3K, by screening it on Luria–Bertani (LB) agar containing kanamycin (50 μg mL−1) and sequenced using commercial services (Eurofins MWG Operon, Bengaluru, India).

Expression and purification of sDacD

A 12 h old culture was diluted 1:100 in 1 L LB containing kanamycin (50 ug mL−1) and grown at 37 °C (OD600 nm ~ 0.5). Isopropyl-β-d-thiogalactopyranoside (IPTG) was added to a final concentration of 0.5 mM and the culture was incubated for a further 6 h. Cells were harvested and lysed in a lysis buffer as described earlier (Chowdhury et al., 2010). Cell debris and the membrane vesicles were removed from the cell lysate by ultracentrifugation. Supernatant collected was subjected to ampicillin-affinity chromatography as described previously (Nicholas & Strominger, 1988; Chowdhury et al., 2010). The purified protein was eluted from the ampicillin-coupled resin with 1 M NH2OH, 0.5 M Tris–HCl at neutral pH, and the pooled fractions were dialyzed with three changes of buffer (20 mM Tris–HCl and 150 mM NaCl, pH 7.5). The purified sDacD was used for biophysical and biochemical analyses.

Fluorescent penicillin labelling

The activity of purified protein was checked by labelling sDacD with fluorescent penicillin, Bocillin-FL (Invitrogen Inc., Carlsbad, CA) at 35 °C for 30 min as described previously (Zhao et al., 1999; Chowdhury et al., 2010). After denaturation by boiling, the protein was analysed through 12% SDS-PAGE and visualized under UV using the GelDoc-It 310 system (UVP, UK).

Circular dichroism (CD) analyses

The far UV circular dichroism (Far UV CD) spectrum of sDacD was determined using a Jasco J-810 spectropolarimeter (Easton, MD). In brief, spectral data of sDacD were collected by placing the sample in a quartz cell (path length = 0.2 cm) at 25 °C with a 0.2-nm step resolution, 1-s time constant, 10 milli-degree sensitivity at a 2.0 nm spectral bandwidth, with a scanning speed of 50 nm min−1. Corrected spectra were obtained by subtracting the solvent spectrum. Secondary structure was estimated by K2d software (Andrade et al., 1993).

Kinetic analysis of the interaction of sDacD with beta-lactams and peptide substrates

The constant k2/K (rate of formation of the acyl-enzyme complex at sub-saturating concentrations of substrate) was determined from the time courses of enzyme–substrate complex formation with Bocillin-FL as described earlier (Chowdhury et al., 2010). In brief, the sDacD was incubated with Bocillin-FL at different concentrations for 30, 60, 90 and 120 s. The reaction was stopped by adding denaturing buffer and boiling for 5 min, and the samples were analysed in 12% SDS-PAGE. The labelled sDacD was quantified by densitometric scanning (Chambers et al., 1994; Chowdhury et al., 2010).

The k3 value (deacylation rate constant) was determined from semi-log plots of the percentage of Bocillin-FL remaining vs. time (Di Guilmi et al., 2000; Chowdhury et al., 2010). In brief, the purified sDacD was incubated with Bocillin-FL (50 μM) for 15 min at 37 °C. At t = 0, penicillin G was added to 3 mM, and the fluorescent intensity of the protein was determined by removing aliquots at various times.

The DD-CPase activity of sDacD was assessed for artificial peptide, Nα,Nε-diacetyl-l-Lys-d-Ala-d-Ala and pentapeptide substrate, l-Ala-γ-d-Glu-l-Lys-d-Ala-d-Ala (Chowdhury et al., 2010). Free d-alanine generated was detected and compared with a standard d-alanine solution using a Multiskan Spectrum-1500 Spectrophotometer (Thermo Scientific, Switzerland) at 460 nm (Frere et al., 1976). Kinetic parameters for the DD-CPase assay were deduced from the linear regression of the double reciprocal plot (Lineweaver & Burk, 1934).

3D model building

A restraint based program modeller 9v1 (Sali & Blundell, 1993) was used for generating the three-dimensional (3D) model of sDacD. Initially, sDacD aa sequence was allowed to search for potentially related sequences. The sDacD sequence was aligned with the corresponding template, and the 3D model was calculated based on the lowest value of modeller objective function (Sali & Blundell, 1993). sDacD model was improved through energy minimization (EM) using the charmm version 22 (Brooks et al., 1983) available in the discovery studio software suite (Version 1.5; Accelrys Software Inc., San Diego, CA). The models were further refined by adding explicit water molecules to the model for molecular dynamics (MD) simulation at 300 K using gromacs (Van Der Spoel et al., 2005) for 300 ps. The resulting model was subjected to procheck (Laskowski et al., 1993) and verify3d (Luthy et al., 1992) to evaluate the model folding and the stereochemistry. As the volume of the active-site groove influences the binding of the substrate molecule and hence the catalysis, the volume of the groove associated with the active-site motifs was measured by surface topography analysis (CASTp) (Dundas et al., 2006; Chowdhury & Ghosh, 2011). The secondary structure of sDacD was identified using three independent algorithms, predict protein (Rost et al., 2004), psipred (Jones, 1999), and stride (Heinig & Frishman, 2004).

Results and discussion

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

Expression and purification of sDacD

To simplify the purification procedure, soluble DacD (sDacD) containing 363 aa was constructed and purified by ampicillin-affinity chromatography (final concentration ~ 0.9 mg mL−1). The average molecular weight of sDacD was ~ 40 kDa. The protein was stable and active after purification, as observed by Bocillin-FL labelling (Fig. 1).

image

Figure 1. SDS-PAGE analysis of purified sDacD. Lane 1: Protein molecular weight marker. Lane 2: Unpurified sDacD. Lane 3: Purified sDacD. Lane 4: Purified sDacD labeled with Bocillin-FL.

Download figure to PowerPoint

Penicillin acylates sDacD at a moderate rate

To understand how efficiently sDacD binds penicillin, we assessed the interaction of sDacD with fluorescent penicillin, Bocillin-FL. The acylation rate constant (k2/K) of sDacD was determined for different time intervals assuming a pseudo-first order reaction (Chowdhury et al., 2010). The acylation rate constant, 450 ± 45.9 M−1 s−1 (Table 1), indicates considerable beta-lactam binding efficiency of sDacD. However, the rate of acylation was a little lower than that of sPBP5 (Chowdhury et al., 2010).

Table 1. Kinetic parameters with Bocillin-FL
Soluble PBPsk2/K (M−1 s−1)k3 (s−1) × 10−5
  1. a

    Courtesy from Chowdhury et al. (2010) FEMS Microbiol Lett 303: 76-83.

sDacD450 ± 45.912.9 ± 1.1
sPBP5a800 ± 50.3a18.9 ± 1.7a
sPBP6a1900 ± 65.5a1.7 ± 0.1a

sDacD possesses significant deacylation activity with beta-lactams

The deacylation reaction, in which inactive beta-lactam was released from the covalent adducts, was described by first-order rate constant k3. The calculated deacylation rate of labelled sDacD (See Table 1) revealed a moderate k3 value, which indicates a fair deacylation efficiency of sDacD.

sDacD exhibits weak DD-CPase activity

The interaction with penicillin did not reflect the whole enzymatic activity of DacD. Therefore, the DD-CPase activity of sDacD was determined with artificial substrate, Nα,Nε-diacetyl-l-Lys-d-Ala-d-Ala and with pentapeptide substrate, l-Ala-γ-d-Glu-l-Lys-d-Ala-d-Ala. The lower kcat values with both the substrate indicate weak DD-CPase activity of sDacD (Table 2). It is worth mentioning that sPBP6, which is the next nearest homolog of DacD, is inactive on pentapeptide substrate (Chowdhury et al., 2010).

Table 2. Kinetic parameters with peptide substrates
ProteinNα,Nε-Diacetyl-l-Lys-d-Ala-d-Alal-Ala-γ-d-Glu-l-Lys-d-Ala-d-Ala
Km (mM)kcat (s−1)Km (mM)kcat (s−1)
  1. a

    Courtesy from Chowdhury et al. (2010) FEMS Microbiol Lett 303: 76-83.

  2. b

    ND, not detected.

sDacD4.7 ± 1.50.25 ± 0.022.95 ± 0.120.13 ± 0.02
sPBP5a8.98 ± 1.79a2.70 ± 0.30a1.38 ± 0.43a1.41 ± 0.11a
sPBP6a1.28 ± 0.26a0.56 ± 0.03aNDa,bNDa,b

DacD is a β-sheet-rich protein

The crystal structures of sPBP5 and sPBP6 (Nicholas et al., 2003; Chen et al., 2009) show a similar secondary structure with no gross architectural differences. In the absence of crystal structure, CD spectral analysis would be of utmost importance to elucidate the biophysical characteristics of sDacD. It was evident from the CD spectra that purified protein was in a native conformation with characteristics of the molecular spectra of alpha and beta structures, indicating the protein was active and stable at room temperature. Unlike sPBP5 and sPBP6 (Chowdhury et al., 2010), more beta-sheets were detected in sDacD (Table 3, Supporting Information, Fig. S1). The occurrence of a larger amount of β-sheet structure in sDacD may cause some structural alteration, which might exert different biological activity than PBP5.

Table 3. Percentage distribution of secondary structures obtained with different methods
Soluble PBPsMethods followed% of α-Helix% of β-sheet% of Random coil
sPBP5 (1NZO)Crystal structure23.027.849.2
sDacDCircular dichroism (CD)13.637.449.0
sDacDPREDICT PROTEIN14.036.050.0

Molecular modeling of sDacD resembles the obtained secondary structure

Because DacD shared a high level of aa identity with PBP5, homology modelling (or comparative protein structure modelling) could be applied to generate the three-dimensional conformation of sDacD. For model building, the program modeller 9v1 was used with the pdb coordinate, 3BEC chain A (crystal structure of E. coli PBP5 in complex with a peptide-mimetic cephalosporin; Sauvage et al., 2008) as template. The secondary structure prediction by predict protein and psipred suggested that sDacD was a αβ protein with a larger amount of β-sheet structure (Table 3 and Fig. S2), which was consistent with the results obtained from CD spectroscopic analyses. The model of lowest energy value had 94.9% residues in the most favoured region in the Ramachandran plot and 98.35% residues had an average 3D-1D score above 0.2, as obtained through verify3d profile (Fig. 2), which affirms a well derived model. The model has been deposited to the PMDB server (ID PM0076504).

image

Figure 2. Homology model of sDacD. (a) The ribbon structure and (b) the surface structure of sDacD (top view, only active site highlighted). The residues forming the active-site groove are coloured green and are space-filled. The volume of active-site groove was measured through CASTp calculation.

Download figure to PowerPoint

Similar topology of active-site residues explains functional relatedness of DacD with PBP5

Like sPBP5, the sDacD model is composed of two Domains placed perpendicular to each other. Domain II is β-sheet-rich, whereas Domain I is composed of both α-helices and β-sheets (Fig. 2a). There is a relative increase in beta-sheet in Domain I of sDacD as compared with sPBP5. Comparison of the calculated secondary structure of the sDacD model generated by stride with that of sPBP5 indicates that residues Gln 38-Arg 39 and His158-Ser159-Ser160 of sDacD create a beta-sheet structure, whereas the respective positions of sPBP5 create coils and turns. Moreover, the Glu 230 and Met 233 of sPBP5 Domain I form turns, whereas the corresponding residues (Gln 229 and Arg 232, respectively) in the sDacD model adopt a beta-conformation. Therefore, both similarities and the differences exist when we take a closer view at the active-site of sDacD and sPBP5.

The crystal structure of sPBP5 shows that all the residues present in the signature motifs are oriented in such a way as to create an intermolecular nexus (Davies et al., 2001; Nicholas et al., 2003) (Fig. 3a). The serine residue of SXXK motif is the most important catalytic residue at the active-site which binds both beta-lactam and peptide substrate. Mutation of active-site serine residue causes severe impairment of DD-CPase activity and beta-lactam binding (van der Linden et al., 1994). The serine residue of SXN motif helps in the hydrolysis of peptide substrate by polarizing water molecule (Nicola et al., 2005). The histidine residue in the Ω-type loop is functionally analogous to Glu166 of TEM-1 beta-lactamase (Davies et al., 2001) and promotes hydrolysis of beta-lactams.

image

Figure 3. Active-site environment of sPBPs: (a) sPBP5 (1NZO); (b) stereo-overlay of sPBP5 (1NZO) active-site residues on that of sDacD (RMSD value of sPBP5 vs. sDacD is 0.66 Å). sPBP5 residues are coloured grey and are labelled in black; the residues of sDacD are coloured and labelled in green.

Download figure to PowerPoint

Superimposing the active-site of sDacD model onto sPBP5 [1NZO, (Nicholas et al., 2003)] (Fig. 3) reveals that the orientations of the relevant active site residues of SXN motif are nearly identical (Ser 110 and Asn 112 of sPBP5 vs. Ser 109 and Asn 111 of sDacD). The serine residue of SXXK motif of sDacD adopts a similar orientation to that of sPBP5 (Ser 43 of sDacD vs. Ser 44 of sPBP5). The His 150 of Ω-type loop and Arg197 of sDacD also clearly overlap with that of sPBP5 (His 151 and Arg 198 of sPBP5) (Fig. 3b). The close resemblance in the orientation topology of the active-site residues of sDacD with sPBP5 may explain the similarity in enzymatic activities during deacylation.

Reasons for low acylation efficiency of sDacD

In the proposed sDacD model, Lys 46 of SXXK motif shifts away from Ser 43, making the distance between these two residues 5.14 Ǻ, which is probably too big to form hydrogen bond (Fig. 3b) (the distance between Lys 47 and Ser 44 of SXXK motif in sPBP5 is 3.15 Ǻ). In all DD-CPase PBPs, the lysine of the SXXK tetrad acts as a proton acceptor for a nucleophilic attack by serine that facilitates the formation of an acyl-enzyme intermediate (Nicholas et al., 2003; Zhang et al., 2007; Chowdhury & Ghosh, 2011). Therefore, the large distance between Ser 43 and Lys 46 probably weakens the nucleophilicity of the active-site serine and hence lowers the acylation rate. It is worth mentioning that during acyl-enzyme complex formation, the terminal d-Ala is removed from the pentapeptide. Therefore, the larger distance between lysine and serine of SXXK possibly decreases the affinity of sDacD toward beta-lactams and reduces its DD-CPase activity. In addition, SXN and KTG motifs might influence DD-CPase activity in sDacD. The lysine residue in KTG motif is known to stabilize the acyl-enzyme complex (Zhang et al., 2007; Chowdhury & Ghosh, 2011). An increase in the distance between the Lys (KTG) and Ser (SXN) has a significant effect on the DD-CPase activity, as observed in the Lys213Arg mutant of E. coli PBP5 (Malhotra & Nicholas, 1992). In the sDacD model, the lysine of KTG motif twists farther from serine of SXN motif, creating a distance of 3.05 Ǻ, whereas it is 2.7 Ǻ for sPBP5, which, although not large, is accountable (Fig. 3b). Therefore, we speculate that the combined effect of the orientation differences between the active-site residues may be responsible for the altered biochemical and physiological functions seen in PBP5 and DacD.

As predicted through surface topology analysis (CASTp), the groove volume at the active-site signature motifs of sDacD is 326.1 Ǻ3 (Fig. 2b), whereas that of sPBP5 is 960.8 Ǻ3 (Chowdhury & Ghosh, 2011). The smaller groove of sDacD possibly affects the binding of pentapeptide and, therefore, may decrease DD-CPase activity. However, activity toward smaller substrates such as Bocillin-FL may not be impaired. It is noteworthy that although the active-site groove volume of sDacD is nearly three times smaller than PBP5, it is about double the size of that of sPBP6 (161.5 Ǻ) (Chowdhury & Ghosh, 2011), which may explain why sDacD exerted better DD-CPase activity than sPBP6 towards pentapeptide substrate (Table 2).

Unlike other DD-CPases, PBP5 mutant sensitizes E. coli to beta-lactam antibiotics and complementation of PBP5 restores the resistance (Sarkar et al., 2010). The reason for the PBP5-mediated beta-lactam resistance lies in its typical enzymatic properties. PBP5 deacylates beta-lactam more rapidly than PBP6 does (Chowdhury et al., 2010), even though PBP5 does not possess any beta-lactamase activity (Sarkar et al., 2010) at physiological pH, which is in disagreement with earlier claims (Georgopapadakou, 1993; Davies et al., 2001). It is proposed that PBP5 may behave as a trap for beta-lactams and provide a shielding effect over the lethal targets, which in turn protects the essential PBPs from being inhibited (Sarkar et al., 2010). This may be due to the high deacylation efficiency and the high copy number of PBP5, and both factors taken together may act such that the effective pool of PBP5 remains available to bind beta-lactams. On the other hand, PBP6 due to its low deacylation efficiency cannot reverse the lost beta-lactam resistance in PBP5 mutants, even when it is overexpressed (Sarkar et al., 2010, 2011).

In contrast to PBP6, DacD can rescue the lost beta-lactam resistance in E. coli PBP5 mutant, at least partially (Sarkar et al., 2011). Our results reveal that sDacD possesses a higher rate of deacylation activity toward beta-lactams (~ 65% of PBP5) compared with PBP6. Therefore, it makes sense that DacD can partially substitute the loss of PBP5 in terms of maintaining intrinsic beta-lactam resistance when expressed in mid-logarithmic phase. These observations imply that the cellular function of DacD is more closely related to PBP5 than with PBP6. In silico analyses of sDacD also reveals a possible structural relatedness with PBP5. Nevertheless, little differences in the orientation of the active-site residues exist, which probably cause these two proteins to act differently. The identical topology of sDacD and PBP5 at the Ω-type loop region predicts a high deacylation efficiency of sDacD. However, DacD possesses comparatively weak DD-CPase activity, possibly due to a far-reaching change in the orientation of Lys 46 from the active-site serine residue (Ser 43). It can be speculated that the higher β-sheet structures that predominated in the Domain I (as obtained from CD, psipred, and stride analyses; Table 3 and Fig. S2) may influence the packing of active-site residues, probably changing the orientation of the lysine residue (Lys 46) and hence modifying the substrate specificities.

Based on the in vitro characterization and in silico predictions concerning DacD function, it can be speculated that three homologous proteins – PBP5, PBP6 and DacD – possibly exert different or partially overlapping cellular activity at different time points of the growth phases.

Acknowledgements

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

This work is supported in parts by two different grants from the DBT and CSIR, Govt. of India to A.S.G. A.K. is supported by a fellowship from UGC, Govt. of India. There is no conflict of interest to declare.

References

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

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results and discussion
  6. Acknowledgements
  7. Authors' contribution
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
fml12009-sup-0001-FigS1.pdfapplication/PDF74KFig. S1. CD spectrum for sDacD.
fml12009-sup-0002-FigS2.pdfapplication/PDF202KFig. S2. Calculation of secondary structure of sPBP5, and sDacD using psipred v2.6.

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.