Received Date: May 27, 2015; Accepted Date: October 23, 2015; Published Date: October 30, 2015
Citation:Lynch GW, Portelli SS, Wong SW, Costa MR, Lemgruber APM, et al. (2015) Imaging of High and Low Resolution Ebola Envelope GP Structures Composited with in silico Models of Difficult-to-Resolve Sections. J Mol Genet Med 7: 186 doi:10.4172/1747-0862.1000186
Copyright: ©2015 Lynch GW, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Visit for more related articles at Journal of Molecular and Genetic Medicine
The Ebola surface glycoprotein, GP, facilitates receptor binding, cell infection and cytopathology, and is a principal target for immune protection. The structure of the GP central core structure and antibody target for the 1976 Ebola Zaire Mayinga strain of Ebola has been resolved by X-ray crystallography (e.g., pdb-ID: 3CSY). But other important GP regions have defied crystal structure determination. These include the apical mucin-like domain (MLD) that contains immuno-protective sites and is most distal to the membrane, and the membrane proximal C terminus region (meD) where GP is tethered to the Ebola surface. A molecular structure-based strategy in vaccine design necessitates detailed fine-structure knowledge of exposed sites for immuno-targeting. To address the lack of MLD and meD structures we have performed computational modeling of those regions using the programs iTasser, Phyre2 and CABS-flex. Candidate MLD model structures for the 1976 Mayinga strain were screened for continuity and fit with the 3CSY core and tested further for 3D compliance for its fit within the spatial boundaries of a lowresolution GP trimer cryoelectron tomograph (EMDB-ID: EMD-6003). Under these constraints only 1 of 6 initial MLD models generated by iTasser or Phyre2 met the required structural criteria of continuity, orientation and spatial fit. That best-fit structure (MLDm1p1) was additionally evaluated using the MolProbity geometric analysis, as well by the generation of highly similar and compatible structures (e.g., MLDm2c2) by the CABS-flex program. Notably, documented protective antibody sites of the MLD mapped to the MLDm1p1 apex surface, as anticipated for domain functionality or exposure to antibody attack. These studies describe best fit proximate in silico models of the MLD and meD regions of GP to for improved molecular understanding and therapy design. Biochemical and immunologic validation and refinements of the models are continuing, to establish to their biostructural compatibilities and value to inform on GP structure-function and immuno-target potential.
The 2014 Ebola Virus (EBOV) outbreak in West Africa had up to the end of March 2015 claimed more than 10,236 lives from 24,753 infections . Now more than a year on from its emergence the crisis appears to be contained and finally a couple of vaccine constructs have been in field clinical trials, and show some evidence of protection for the EBOV 2014 strain [2,3].
The wider protective potential of vaccines in the pipeline to protect against any other Ebola strains that might emerge remains a significant unknown, and whether they could provide any protection against the more distant and highly lethal Marburg filovirus. There is therefore a high need for a variety of efficacious drugs and vaccines to pre-empt any such future outbreaks of Ebola or Marburg. Structure based drug and vaccine design of recent times has matured into being of value in the development of therapeutics for infectious disease agents. Our structure based studies of the Ebola GP protein follows from an approach developed in the study of Collateral Immunity© for broad-strain seasoned antibody responses and protection of Influenza [4-7]. In this present article we elaborate on an invited medical image published by this journal earlier this year on EBOV GP structure .
Ebola Envelope GP: The envelope surface glycoprotein, GP, protrudes from the Ebola virus surface as a homo-trimer that contains the receptor binding and membrane fusion determinants that define cell tropism, entry and infection [9,10]. Consequently a detailed understanding of GP structure and interactions as the principle target for immune defense and for vaccine development is paramount. Indeed, detailed knowledge of the high-resolution, 3.4 Angstrom, crystal-structure determination of the central core (pdb structure identifier: 3CSY), that contains receptor binding and fusion sites , has been invaluable, and particularly so for the realization of specific target sites for immune antibodies. However, that resolved structure, the most comprehensive of all performed to date, represents only about half of the whole pre-fusion protein sequence of GP. Unfortunately a number of other GP sections are without detailed structural information as they have been recalcitrant to their determination in the crystal. Within those undefined regions are important structural and functional features and for which high resolution information would be invaluable. The focus of this article is on two of those unresolved sections, namely the mucin-like domain (MLD that spans aa 313- 501) of GP1 that contains immunodominant antibody sites for host immune protection [11-13], and a principal focus of this study, and the C-terminal section of GP2 (meD aa 658-676) at the GP-virion surface envelope interface.
Defining protein structure
X-ray crystallography has emerged over time as the principal technology for the description of high-resolution protein structures and remains the most validated procedure. Despite its considerable value it has limitations and can provide variants of the structure not found in nature. Furthermore X-ray crystallography cannot capture all of the possible structures a protein may achieve, and many of these may be temporal and subject to their biologic context and environment (e.g., by pH). With crystallography likely capturing only one of one of many possible structural configurations. Many proteins or sections of proteins can be recalcitrant to X-ray evaluation, as they are either refractory to crystallization or do not contribute to ordered diffraction patterns, due to being heterogeneous in composition, hydrophobic, have mobile elements, or are post-translationally modified. The most notable and substantial section that had remained unresolved in the GP 3CSY structure, despite the presence of its sequence in the protein construct used for crystalisation, is the mucin-like domain (MLD). In the attached native state the MLD is the most membrane-distal segment of the envelope GP and projects out above the central, chalice-like, 3CSY core [9,10]. Characteristically this highly glycosylated mucin-like domain would provide a surface environment compatible for EBOV to mix and circulate in mucosal surfaces, by virtue of its similarity with the heterogeneous mucin family of proteins [14,15], that populate the mucosa, and with which it likely shares a degree of structural similarity. Notably the entire mucin-family is poorly represented in the structural database. The high degree of glycosylation of this glycoprotein family is in itself problematic for crystal structure refinement.
Another region of GP for which there is currently minimal structural knowledge is the membrane-associating region, which anchors GP into the EBOV envelope. Hence, we asked the question, whether useful structural information on the MLD and the membrane associating meD sections could be attained from computer-based in silico modeling . To assist in evaluating plausible in silico simulations, a low-resolution cryoelectron tomograph structure determination of the GP trimer of the EBOV surface has recently been presented , and for our modeling studies provides a necessary structural framework to which acceptable models need to be compatible.
Rational of the study approach
Success in correctly assigning and defining protein structures and as evaluated biennially with the CASP (Critical Assessment of protein Structure Prediction) challenges  has realistically improved the potential and value for determining structural information for proteins that defy description by other means. The GP of MLD is considered to fit within the 'highly motile', 'structurally heterogeneous' or 'inherently disordered' category of proteins, and for which no structures have been obtained. The quest is to visualize and determine the structures the GP MLD domain that are presented for Ab, and might present to cellular interactions and to the immune system as a potential target or for avoidance thereof. Before technologies arise that are capable of determining the basic structure(s) of such regions the point of this study was aimed at attempting to generate feasible models of the MLD of EBOV GP regions.
Templates, programs and procedures
EBOV Study Strain and sequence: For the modeled structures in this study the Ebola Mayinga 1976 strain Zaire subtype (UniProt Taxon ID: 128952) was the template reference strain sequence used for analysis to ensure direct comparison and consistency with the Ebola strain sequence upon which the resolved core GP 3CSY structure was generated. That isolate was: Ebola virus/H.sapiens-tc/COD/1976/ Yambuku-Mayinga, NCBI: NC_002549. With the specific protein studied; the 676 amino acid (aa) envelope GP protein (UniprotKB ID: Q05320; and Proteome ID: UP000007209).
Molecular Modelling Platforms: The molecular modeling of the MLD and meD segments was performed using the Phyre2 [19,20] (http://www.sbg.bio.ic.ac.uk/phyre2); I-Tasser [21,22] (http:// zhanglab.ccmb.med.umich.edu/I-TASSER/); and CABS-flex [23,24] (http://biocomp.chem.uw.edu.pl/CABSflex/), software tools. GP: Our model structure of the mucin-like region (MLD) was achieved using Phyre2  and covered amino acids 313 to 501 across GP1 and included sequence 313-335 overlap with the 3CSY structure . Modelling, rendering or alignment analyses of the GP structures was performed using RASMOL , PyMol , and Swiss-PdbViewer, SPDBV_4.1.0_OSX (http://www.ncbi.nlm.nih.gov) (http://spdbv. vital-it.ch). meD: In addition, the membrane proximal/associating GP region (meD) was studied for GP2 (aa 658-676) and model achieved using Phyre2. The short C-terminus 13 aa Phyre2-derived model was manually aligned with the 3CSY and EMD-9003 structures and rendered in PyMol.
EBOV GP Reference Structures: The template scaffold structures used in this study were the high resolution 3.4 Å 3CSY derived by x-ray crystallography  for a truncated and modified GP sequence of the Mayinga 75 strain, and a low-resolution CET produced from Ebola virus-like particles expressing full-length EBOB GP deposited into the Electron Microscopy Data Bank (EMDB accession number 6003 (EMD-6003) (http://www.emdatabank.org/)  note:1.
Structure Visualisation and Alignment: The best-fit simulations from the in silico structure predictions were mapped against the prefusion crystal structure 3CSY rendered in PyMol. Successful candidate models were further evaluated by alignment and rendering into the low-resolution, trimeric cryo-electron tomography (CET) structure [17,26], to test for additional 3D-orientation and spatial compliance.
Structure Geometry Profiling: The coordinates in pdb format of the best-fit GP protein Phyre2-models experimentally obtained for the mucin-like (Phyre2 and CABS-Flex structures) and C-terminus regions (Phyre2 structure) were submitted for MolProbity analysis (http://molprobity.biochem.duke.edu). Assessments of molecular conflict, and their rotamer and Ramachandran characteristics obtained using the MolProbity profiling tool [27,28], were compared against the comparable properties generated for a range of other well determined protein structures. The proteins used for comparison were; the A3 collagen binding domain of von Willebrand Factor (pdb: 1AO3), Human Serum Albumin (1BM0), Bovine Serum Albumin (4F5S), EBOV GP core (3CSY)(1976 Mayinga Strain), and EBOV GP core (3VE0)(1976 Boniface Strain) protein structures.
Immunomapping: Established sites of protective antibodies that target the MLD region of GP were mapped onto the Phyre2 MLD model (MLDm1p1). The sites assessed were; i) the IEDB epitope ID: 156605 by the neutralising antibody C4G7 [11,29-31] and ii) a protective site of the Zaire Mayinga strain for the non-neutralising antibody H13F6, but that antibody is not protective for other EBOV strains [12,13,32]. These were rendered in PyMol onto the composite image shown in Figure 1.
Structure Refinement: For assessment and profiling of structure flexibility the MLDm1p1 model coordinates were submitted to the CABS-flex program server 23, 24(http://biocomp.chem.uw.edu.pl/ CABSflex).
Note 1: Cryoelectron micrographs (CEM) and cryoelectron tomographs (CET – i.e., from 2D tilted and tiled CEM images) of a protein of interest are obtained as a consensus from the averaging of up to millions of individual molecular images of a highly purified preparation, and consequently could reveal a full spectrum of molecular shapes captured in liquid phase. Therefore the heterogeneity in shapes captured in a protein would dictate how discrete or broad the returned structure would be. This is in marked contrast to the far more defined structures obtained by X-ray crystallography. A particular strength of CEM and CET is the imaging of molecules in their native water-soluble condition, unlike crystallography that requires crystal formation with relatively close packing and intermolecular contacts and treatments that can involve non-biologic, harsh solvents.
Note 2: For visual representation so that the focal point of the figure was on the in silico models, the resolved structures at the left are reduced in both size and color intensity from the central composite and model counterpart images.
Composite imaging of resolved and in silico modeled GP domains and structures
In the absence of defined high resolution structures for the mucinlike (amino acids 313-501) and membrane spanning (amino acids 658- 676) domains of GP we have performed in silico modeling of these regions and composited those models together with the published low EMD-6003 and high 3CSY structures of Ebola GP. Under the imposed structural requirements only 1 of 6 MLD computer simulated models generated by either iTasser or Phyre2 satisfied the spatial fit criteria set above; continuity with 3CSY, and compatibility and 3D positioning with the EMD-6003 trimer footprint. The successful model, ID: MLDm1p1, was modeled using Phyre2. The central composite of Figure 1 shows the X-ray crystal resolved structure, and the in silico MLD and meD models mapped together and within the recent cryo-electron spatial EBOV GP trimer tomograph . Shown on the periphery of that image are the space-filled models of the complete low-resolution cryoelectron tomograph (Cryo-ET) trimer structure of GP (EMD-6003) (in blue, top-left) and the high-resolution X-ray crystal structure of the GP trimer core (pdb 3CSY ). On the right of the composite are the in silico space-filled models from Phyre2 simulations of a GP1 mucin-like domain monomer and the GP2 C-terminus membrane stalk domain trimer. In the centre image each of these structural elements has been assembled as a composite, built in ribbon display, and within the Cryo ET global framework. Together these structures represent a majority (~ 85%) of the GP protein, with the 3 separate mucin-like domains at the very top of the composite (yellow), a central core trimer (white) and at the bottom the membrane domain (yellow). Structural visualizations of the refined and the in silico-modeled structures are from pdbviewer, Rasmol [25,26,33] or PyMol  rendering. Each of the different GP domain sequence determinations were generated for the 1976 Ebola Zaire strain, and imaged to represent a single, prefusion, GP trimer, knob that projects out from the envelope surface. Noting as mentioned this composite of most of the GP is without proposed structures for the GP regions of amino acid sequences: 1-30, 190-213, 238-299 or 599- 658). For direct reference of the respective regions already resolved or modeled, a linear map of the complete GP protein sequence has been included in bar format in Figure 1.
i) Linear sequence bar of GP amino acids (aa): Duplicate bar graph representations of the 1-676 aa of GP are shown at both the top and bottom of the Figure. On the bars, the truncated sequences represented by the refined 3CSY crystal structure, are indicated with gradient purple/aqua shading. Within this core structure yet to be structurally defined are the sections bounding and interspersing those 3CSY-defined regions and these are indicated by grey coloring . Top bar: For the bar at the top of the Figure the GP1 and GP2 regions are designated as well as the furin cleavage site at aa position 501. The red line indicates the MLD-based sequence submitted for molecular simulation. Within this region is a short 10 aa region of overlap with the 3CSY structure sequence and contain the mucin like region of 313- 501 aa. Bottom Bar: Depicted at the far right is a short brown and gold hatched segment that indicated the membrane associated section of GP which sequence was submitted fo modeling.
ii) Resolved structures: Shown on the left periphery of the central figure are the resolved structures, for the low-resolution cryoelectron tomograph (Cryo-ET) trimer (EMD-6003) (in blue, upper-left), and the high-resolution X-ray GP monomer core structure (pdb 3CSY ) (white, lower-left).
iii) MLD and meD models: in silico space-filled models from Phyre2 simulations are shown to the right of the central image. This shows the designated best-fit GP1 mucin-like domain monomer model (in tan at the upper-right), upon which the regions associated with binding of the protective antibodies have been mapped and these indicated for H13F6 in red , and C4G7 antibody (IED B epitope 156605) in green respectively . A trimeric model for the GP2, C-terminus membrane-stalk domain is shown in gold coloring at the lower right.
iv) Solved and modelled GP structure GP composite: In the central image each of these structural elements are assembled as a composite built in ribbon display, within the Cryo ET (blue mesh) global framework. Together these represent the majority of the GP protein, with the 3 separate mucin-like domains at the very top of the composite (in yellow), a central core trimer (in white) and at the bottom the membrane associating stalk domain (yellow). Each of these different GP domain sequence determinations were generated for the 1976 Ebola Zaire strain, and represent an image of a single prefusion, GP trimeric knob that would project out from the envelope surface.
Attempts to validate computationally derived molecular models, without experimental or biological evidence and using only computational tools can have considerable limitations. Therefore we have taken 2 different approaches for validation of the MLD model, one based on biological data 1) an antibody binding site mapping study, and 2) computational assessments.
Relevance to known functional, structural or biologically identified sites
A necessary requirement for any proposed MLD structure is that sites for biologically protective antibodies to the viral envelope GP must be surface exposed and oriented away from the MLD-3CSY interface, which by definition would physically preclude antibody attack at the EBOV envelope surface. When tested for positioning of the sites for the binding of the protective antibodies C4G734 (marked in green) and H13F613(marked in red) these were mapped onto the proximate MLD model (Figure 1), As shown those sites were found to map to the exterior surface and namely to the top and side surfaces of the model, as would be predicted. Whilst this does not definitively confirm this exact conformation is correct it is appropriate. Importantly, if the converse had been observed, that proposed structure would have been considered to be an irrelevant configuration and excluded from further consideration.
Computational assessment, validation and molecular refinement
Molecular flexibility: Test algorithms were applied to evaluate the proximate structures modeled in this study. Because of the inherent characteristics of the MLD and its failure to be resolved by crystallography it was deemed useful to computationally test the flexibility of our Phyre2 MLDm1p1 model, and this performed using the CAB-Flex web tool. That assessment revealed 12 related and similar structures, indicating a reasonable degree in flexibility for the MLDm1p1 structure. The model MLDm2c2 was selected for further study and comparison with the parent Phyre2 MLDm1p1 structure (Table 1). The MLDm2c2 is highly similar, but with notable structural differences to the MLDm1p1, and this is evident upon alignment of the 2 structures (Table 1).
|Table 1A) Protein||VWF A3 Domain (Coll Type I / III bind region)||Human Serum Albumin 1BM0||Bovine Serum Albumin (BSA)||3CSY 3CSYFH|
|EBOV Strain:||N/A||N/A||N/A||Mayinga Zaire 1976|
|Protein size||186 aa||581 aa||581 aa||187 aa|
|Resolution||1.8 Å||2.5 Å||2.47 Å||3.4 Å|
|Table 1B) 3VE0 EBOV Region||meD (Phyre2)||MLDm1p1 (Phyre2)||MLDm2c2 (CABS)||MLDm1p1 Aligned with MLDm2c2|
|Strain||Boniface Sudan 1976||Mayinga Zaire 1976||Mayinga Zaire 1976||Mayinga Zaire 1976||Mayinga Zaire 1976|
|Protein size||240 aa||13 aa||188 aa||188 aa||2 ×188 aa|
Table 1A and 1B Show respectively the Rotamer and Ramachandran Geometry Comparison Molprobity Profiles of Table 1A) the resolved X-ray crystallography structures for the A3 collagen-binding domain of von Willebrand Factor (vwf, pdb 1AO4), human (pdb:1BM0) and bovine (pdb: 4F5S), and the central the Ebola GP protein core structures of pdb: 3CSY (1976 EBOV Zaire strain) and in 2A) the 3VEO (1976 EBOV Sudan strain), and their geometries compared to computational models of the 1976 EBOV Zaire sequence membrane proximal GP 658-676 aa C terminus mpcDm1p1 (modeled in Phyre2), and the 313-501 aa Mucin-like-domain MLDm1pi (modeled in Phyre2) and the MLDm2c12 (modeled in CABS). For this figure each of the crystal or modeled protein structures were visually prepared in PyMol. The very last figure in the sequence showed a combined alignment of the structurally similar Phyre2 and CABS models for the MLD. Note: The displayed structures differ in relative size between panel 1A (at 100%), and panel 1B (at 150%), to enhance visibility of the shorter protein constructs.
Table 1: Computational Geometric Comparative MolProbity analyses of GP Resolved Core structures & the Proximate MLD and meD Models.
Geometry profiles: The MLD models (MLDm1p1 and MLDm1c2) co-ordinates were submitted to the MolProbity program to assess their geometric rotameric and Ramachandran properties [27,28]. To put the measured geometries for the best-fit MLD and meD models in perspective, these were matched and compared with the equivalent parameters obtained in MolProbity for a number of other protein structures considered relevant for comparison (Table 1). These were the A3 collagen binding domain of von Willebrand Factor, human albumin, bovine serum albumin, the GP core 3CSY structure and the MLD (MLD-m1p1 and MLD-m2c2 and meD. The respective properties of those protein structures and the results of the MolProbity assessments of rotamers and Ramachandran Plot analysis are presented in Table 1.
Comparisons of the acceptable rotamer and Ramachandran plot profiles for the MLD and meD models were made against the profiles obtained for the likely semi-rigid (e.g., A3 collagen-binding domain of vwf) and flexible (e.g., albumin) proteins and x-ray crystallography defined 3CSY and 3VE0 structures. As predicted the A3 domain of vwf gave the best adherence to for both rotamer and Ramachandran plot assessments. In general the in silico models gave geometry profiles within the range of established and defined structures. However, the MLDm1p1 had the 2nd poorest rotamer score of 12.25% and by far the worst Ramachandran value showing 20.19% of outliers. Notably the CABS-Flex MLDn2c2 derivative of the MLDm1p1, on geometry profiling gave remarkably improved scores of 1.20% poor rotamers and 4.84% Ramachandran outliers, and far more comparable to the geometries of highly validated structures. This has revealed a similar but more refined MLD structure in the MLDm2C2 model and better compatibility with accepted structure parameters than the MLDm1p1 model. Further studies are needed to fully evaluate and compare the respective structural and biological feasibilities of the two models. Whilst these studies have mainly focused on the MLD domain, models for the C-terminus membrane associating domain have also been made, and a short 13 amino acid model was created. Further studies of this and more extended models of the C-terminus and including models that overlap with segments of the resolved 3CSY and 3VE0 core structures are now needed.
The essential boundaries and constraints placed for the modeling of MLD and meD regions of GP was that they must be compatible with the resolved low (CEM) and high (crystal) resolution structures. Fitting this basic criteria the in silico MLD model MLDm1p1 structure obtained in these studies was shown to complementarily map with i) continuity to the resolved 3CSY structure, and ii) within the 3D-boundaries of the EMD-6003 trimer structure17. In immunology based testing of the model for biologic relevance the G4G7 and H13F6 protective antibody sites were found to surface map onto the MLD1p1, to the most exposed surface and apex of the MLD region, as would be appropriate for functionality and immune attack. Additional confidence in the Phyre2 generated MLDm1p1 and meDm1p1 models, was also achieved using computational geometric validation. This included examination of the structural flexibility of the MLDm1p1 using the program CABSFlex, which returned 12 additional similar but varied model structures and is indicative of considerable structural flexibility for the MLD. A representative MLDm2c2 model was revealed as highly similar, but with notable structural differences to the MLDm1p1, when the two structures were aligned (Table 1). Furthermore, the geometry profiles for the in silico models on MolProbity analysis, broadly gave geometry profiles amongst the range of values for the established and defined structures. But, the MLDm1p1 model had the 2nd weakest score for both rotamers and Ramachandran values. However, it was found that the CABS-Flex MLDn2c2 derivative of the MLDm1p1, on geometry profiling gave scores of 98.71% acceptable rotamers and 95.16% acceptable Ramachandran plotting, and far more comparable to the geometric characteristics of highly validated structures. This has valuably revealed a similar but refined MLD structure (i.e., the MLDm2C2), which, for future studies be used to evaluate and compare further structural and biological feasibility profiles with those of the MLDm1p1. Additional structure validations are being pursued, such as compatibility with carbohydrate mapping. These studies have mainly focused on the MLD domain, but in addition models for the C-terminus membrane-associating domain have also been built, and a short 13 amino acid model presented. Further studies of this and more extended models with overlapping segments with the resolved 3CSY and 3VE0 core structures are needed.
The GP in silico models shown here fit a range of structural and biomapping criteria, but it needs to be appreciated that without additional qualification, they remain as models or proxies of what those domains might be, and serve at present as interim simulations to provoke new approaches for biotesting and en route to the development and determination of more substantiated structures. Certainly related similar structures generated by the CABS-Flex program and with even stronger geometric validation scores immediately offers new refinements upon which the different models can be tested for in biological assays and further quality checks. These structures herein provide a starting point upon which experiments will be designed for the biotesting and identification of new antibody target sites. It will therefore also be valuable to know if such simulations of difficult to resolve domains can for the future be beneficial for vaccine and drug design. And whether synergistic use of molecular modeling and cryoelectron tomography (e.g., as used here) may offer a viable alternative for the fine structure determination of proteins that are resistant to crystal structure refinement. An enticing prospect for the future may be that sub 10 Ångstrom level cryoelectron tomograph determinations [35,36] may overcome this technical obstacle and be used together with ab initio modeling, to give correct and fully verifiable, high-definition assignments of true native structures. To achieve this would be a holygrail in the realization of difficult-toresolve protein structures of biological and therapeutic importance.
Ana Paula Mello Lemgruber and Maria Rocha Costa are medical students and performing research internships, in the Science without Borders program, a study abroad initiative between the Brazilian Government and the University of Sydney. “Collateral Immunity”© GW Lynch, JS Sullivan.
The authors report no conflicts of interest in the performance of this work. RB’s institution The Children’s Hospital at Westmead has received funding from major Pharmaceutical company’s for attendance and presentations at scientific meetings, and honoraria for RB for delivering educational presentations, but this not personally accepted by RB.