ESF/EU MolTools Workshop:
Ligand Binding Molecules against the Human Proteome
6 - 8 September 2004
Clare College, Cambridge, UK

Organisers:

Mike Taussig, Babraham Institute, Cambridge, UK
John McCafferty, The Wellcome Sanger Institute, Hinxton, UK
Andreas Plückthun, University of Zurich, Switzerland

Report

Scientific Content

Specific ligand binding molecules, including native and recombinant antibodies, protein scaffolds, peptides and nucleic acid aptamers, are essential tools for monitoring protein expression and determining protein function. There is now a recognised need to establish a comprehensive, characterised and standardised collection of ligand binders directed against the individual proteins of the human proteome. This must be accompanied by the development of high throughput binder-based assay systems, and applications in diagnostics and therapeutics. The purpose of the workshop was to explore the possibility of making a collection of binders against the proteome on a genome wide scale and discuss the benefits that would result. Thus major issues on the agenda were:

• What are the project's objectives and obstacles? It was agreed that this project could become one of the largest to emerge from the human genome; to draw an analogy with that effort, we are just at the stage of discussing the best strategies before thinking about how it would be organised.
• Is it realistic at the technical level? How might it be carried out at the technical level, concentrating particularly on key issues of availability of high quality cDNAs and pure protein, types of binders and selection methods, as well as applications.
• How would it be coordinated? If it was agreed that the project is feasible in principle, how would it be organised in practice? Even if funds were available, it might not be easy to implement because of the 'factory' requirements (large scale, high throughput binder selection and production). Would there be one production centre or several, organised as a European consortium with several labs contributing, or 'top down' in one place?
• Could it form the basis of an integrated proposal, e.g. to EC, for funding? A project officer from Brussels was present to describe opportunities under FP6 and further ahead to FP7.

The programme was divided into the following sections:
Session 1 - Magnitude of the problem: What is it all about? The current status of the genome; how many proteins are there; and bioinformatics. After introductions by the organisers and Ulf Landegren (on behalf of the Moltools FP6 consortium), presentations to set the background on the human genome were given by Ian Dunham (Sanger Institute, Hinxton), 'A genome annotation driven approach to cloning the human ORFeome', and Victor Jongeneel (Ludwig Inst for Cancer Research, Lausanne), 'Computing and representing the diversity of the human transcriptome'. Dunham proposed that the ORF cloning approach could provide higher throughput annotation of human genes by supplementing cDNA library approaches, or perhaps replacing them in other organisms, completion of valuable resources for functional genomics studies, and providing cloned alternative transcripts and haplotypes. In order to generate the proteome, the desirables are to obtain a representative ORF clone for each human gene (~30,000) and subsequently an ORF clone for each common alternative splice variant (increasing the number x3) and each common SNP haplotype (increase x5) and ultimately for each rare splice form and haplotype. To obtain these needs appropriate gene annotation at each level.

Session 2 - cDNA collections/expression, i.e. where will the target proteins come from; how complete are cDNA collections, how good are they; what are the best expression systems to use? The first speakers were Bernhard Korn (Heidelberg), 'Content development for protein chip production: comparison of human protein expression in vivo and in vitro'; Martin Yuille (Cambridge), 'Human cDNA clones at MRC geneservice'; and Joshua LaBaer (Boston), 'Harnessing the Human Proteome'. They described the completeness and availability of cDNA collections at different sites (RZPD, MRC Geneservice and Harvard). Mike Dyson (Sanger Institute) spoke on 'Expression of mammalian proteins for high-throughput antibody generation' and Christian Cambillau (CNRS Marseille) on 'Structural proteomics: from 3D structures to ligand/function identification'. Dyson discussed the considerations for human and mouse protein cloning, expression and purification in the Atlas of Gene Expression Project at the Sanger Institute, quantitation of heterologous protein expression, and evaluation of tags and strains for optimal soluble protein expression in E. coli Cambillau pointed out the difficulties that would arise in trying to express recombinant membrane proteins (10-20% of the proteome), members of large complexes (20-30%) and natively unfolded proteins (10-20%). Korn discussed the alternative of cell-free in vitro expression systems based on E. coli and wheat germ lysate to overcome obstacles of insoluble inclusion bodies and toxicity of some gene products in E. coli. The cell free systems offer rapid protein synthesis, compatibility with PCR-generated templates and plasmids, the possibility to express toxic gene products, and to easily optimise codon usage and RNA secondary structure.

Session 3 - binders and selection methods: how to get the binders; what should they be; antibodies, recombinant and nontraditional binders. Andreas Plückthun (Zurich) gave an overview of selection technologies and Michael Feldhaus (Pacific Northwest Lab) spoke on yeast display and molecular evolution of antibody fragments. Federico De Masi (EMBL Heidelberg) described 'Hybridoma Chips: a high throughput microarray platform for the screening of monoclonal antibodies' in which the bottlenecks limiting the efficiency of hybridoma production were identified as immunisation methods generally focused on one antigen per animal, laborious fusion and tissue culture procedures, and conventional screening methods which are not suited to high throughput production. His solutions were immunisation of individual animals with multiple (up to 10) different antigens, a custom designed robot to carry out fusion protocols (8 simultaneously) and all the cell culture methods, and automated microarray screening on antigen-coated glass slides. Continuing the antibody theme, Mathias Uhlen (Royal Institute of Technology, Stockholm) described 'Antibody-based tissue proteomics to study the human proteome', in which the objective is to raise a collection of polyclonal antibodies against every human protein (as a set of non-redundant proteins) within 10 years, and to use these reagents to explore functionally the corresponding proteins, protein variants (isoforms) and protein interactions. Uhlen has established the Swedish Human Proteome Resource programme (July 1, 2003) for antibody-based tissue profiling, with funding from the Wallenberg Foundation and with the aim of analysing 5,000 human proteins (20% of the human proteome) in four years, at a throughput of 5 new antibodies per day. All data will be publically available and it will also be possible to submit genes as collaborative project. For details see www.hpr.se

In the area of display technologies, Mingyue He (Babraham) described Cell-free cloning of antibodies and use of ribosome display and Andrew Bradbury (Los Alamos) described the use of lox recombination in antibody libraries. Serge Muyldermans (Brussels) discussed the advantages of camel heavy-chain antibodies and single-domain antibody fragments. Since the antigen binding site of the dromedary HCAb comprises a single domain, VHH, it is only necessary to clone one single gene fragment from the B-cells in order to obtain the VHH repertoire of an immunised animal, and even relative small immune VHH libraries (e.g. 106 transformants) are enriched for specific antigen binders. The selected recombinant VHH's are well produced in bacteria, very stable and highly soluble. Leaving antibodies for scaffolds, Andreas Plückthun (Zurich) described Ankyrin scaffolds, Per-Åke Nygren (Stockholm) Affibodies, while nucleic acid aptamers were discussed by Larry Gold (SomaLogic, Boulder) and Jean Toulmé (Marseille). SomaLogic uses DNA photoaptamer arrays to quantify proteins in complex biomixtures. Photoaptamers bind to their protein analytes with sub-nM to low-pM Kd's and high specificity, and then, upon photoactivation, covalently attach to their targets. The combination of good equilibrium binding (equivalent to antibodies) and photocrosslinking provides specificity equivalent to antibody sandwiches (ELISA's). Photoaptamers provide an interesting and attractive alternative to protein binders.



Session 4 - problems of cross-reactivity and sensitivity: Could the project work even in principle on such a scale or would cross-reactivity make a genome wide approach impossible? Contributions on the molecular analysis of binder cross-reactivity were from Annemarie Honegger (Zurich), 'Engineering antibody libraries for high stability, high yield and high functional diversity', and Wolfgang Hoehne (Berlin), 'Molecular recognition: examples for cross reactivity and polyspecificity'. Honegger compared specific cross-reactivity, due to the presence of very similar epitopes in different molecules which can result from divergent evolution from a common ancestor, splicing variants, or proteolytic processing variants, with nonspecific cross-reactivity caused by 'sticky' molecules both in the binder libraries and amongst the target molecules, due to poor folding and stability. She found that simply eliminating the less stable members from a natural antibody library, whether by stringent selection for stability, prepanning to remove sticky members of the library or preferential amplification of the antibodies derived from the more stable germline families, signifcantly reduced the diversity of antigen binding topographies. Hoehne compared cross reactivity and polyspecificity, or binding of structurally different ligands to a binding site using different mode of interaction. The practical assessment of cross-reactivity of ligand binders by screening on protein microarrays was discussed by Paul Predki (Invitrogen) 'Assessing antibody cross-reactivity with functional protein microarrays', and Dolores Cahill (Dublin) 'Elucidating antibody specificity and cross-reactivity on high density protein arrays'. Predki advocated functional protein microarrays as a rapid approach to assess and identify cross-reactivity against thousand of proteins due to both structured and unstructured epitopes. Cahill used high density protein microarrays of human proteins to analyse cross-reactivity of monoclonal antibodies, and found that a degree of cross-reactivity was present in many cases.

Session 5 - existing applications of ligand binders, how they work so far and how they can be improved. Ulf Landegren (Uppsala) talked on 'Proximity ligation as a general tool for high throughput protein analyses'. This is a highly specific and sensitive means of using binders to detecting protein ligands and intracellular protein interactions, through amplification of DNA tags by PCR or rolling circle after they have been brought together (into proximity) by the interaction and ligated. Jörg Hoheisel (Heidelberg) discussed 'Selection of functional antibodies' and Carl Borrebaeck (Lund) 'Designing antibody microarrays for high throughput proteomics'. Borrebaeck detailed the challenges associated with antibody microarrays in terms of content (probe instability, sensitivity) and technology (protein incompatibility with surfaces, proteome labeling, and shortage of antigens due to proteomes not being cloned). He described the designed n-CoDeR library of 2-3 x 1010 scFv fragments, based on a single identical framework, the high level of sensitivity achievable on antibody microarrays, and current work using chips with 125 scFvs against molecules involved in immune regulation, angiogenesis, innate immunity, etc. Hugues Bedouelle (Paris) proposed that reagentless fluorescent biosensors could be developed against the proteome at high throughput, by identifying a set of universal positions for the conjugation of fluorophores into existing families of ligand binding molecules and incorporating universal positions into the scaffolds of future families.

In his introduction to protein arrays, Mike Taussig identified a number of distinct ways in which ligand binders would be used, namely in capture arrays, where the binders are immobilised, reverse arrays, where immobilised tissue extracts or tissue sections are probed with the binder, proteome arrays, which can serve as screens for antibody detection and assessment of cross-reactivity, and functional arrays of folded proteins where binders could confirm or block interactions. Taussig and Joshua LaBaer (Harvard) described novel means of creating protein arrays from immobilised DNA using cell free systems, namely PISA and Self Assembling Protein Microarrays (NAPPA) respectively. Labaer described the NAPPA method developed in his group at Harvard Medical School (www.hip.harvard.edu), in which plasmid DNA encoding GST-fusion proteins and immobilised on a glass slide is transcribed and translated in situ and the proteins colocalised using an anti-GST antibody. Advantages of both the PISA and NAPPA systems include the fact there is no need to express and purify protein, and that proteins can be expressed as required for experiments so that protein stability is less of a concern. LaBaer has used NAPPA for impressive protein interaction studies. Finally, John McCafferty (Sanger Institute) described 'Expression profiling by high throughput immunohistochemistry' and Fredrik Ponten (Uppsala) 'Antibody-based tissue profiling as a tool for clinical proteomics'. Both these projects use tissue arrays with the aim of annotating protein expression, the former using phage-display selected antibodies in the Atlas project, and the latter with polyclonals made in the Swedish Human Proteome Resource programme (see above).

In the discussion which followed the formal presentations, information on other proposed projects by the German Proteome Society and HUPO was provided respectively by Marius Ueffing (Neuherberg) and Mathias Uhlen (Stockholm). A number of issues came to the fore, one being whether there was a willingness to make a resource publicly available or whether it would be primarily intended for proteome research by the consortium. Another was the complex problem of how intellectual property would be assigned. Dr Indridi Benediktsson detailed the possible avenues through the current FP6 programme and in future in FP7 for funding of a project to create a comprehensive resource of ligand binders against the human proteome. It was resolved to prepare a proposal for a Coordinated Action within the remit of FP6 Infrastructures, and that a subgroup would be designated to take this forward over the coming months.

The consensus was that the workshop was a considerable success both of content and planning, favoured by the beautiful setting and a fortunate spell of outstanding weather. Thanks go in particular to Cheryl Smythe for overseeing the local arrangements and to the staff of Clare College for their excellent facilities.

List of participants

1 Sabine Baars, RCSI, Dublin
2 Hugues Bedouelle, Institut Pasteur, Paris
3 Indridi Benediktsson, EU, Brussels
4 Carl Borrebaeck, Lund University
5 Andrew Bradbury, Los Alamos
6 Dolores Cahill, RCSI, Dublin
7 Christian Cambillau, CNRS Marseille
8 Ian Dunham, Sanger Institute, Cambridge
9 Mike Dyson, Sanger Insitute, Cambridge
10 Michael Feldhaus, Pacific Northwest National Laboratory, Seattle
11 Paul Ko Ferrigno, Hutchison/MRC research centre, Cambridge
12 Jeremy Gillespie, Invitrogen, Glasgow
13 Larry Gold, Somalogic, Boulder, Colorado
14 Mingyue He, MolTools, Babraham, Cambridge
15 Wolfgang Hoehne, Humboldt University, Berlin
16 Jörg Hoheisel, Deutsches Krebsforschungszentrum (DKFZ), Heidelberg
17 Annemarie Honegger, Dept of Biochemistry, University of Zurich
18 Victor Jongeneel, Ludwig Institute for Cancer Research, Lausanne
19 Farid Khan, MolTools, Babraham, Cambridge
20 Massoud Kamali-M, MolTools, Rudbeck Laboratory, Uppsala
21 Zoltan Konthur, MolTools, Max Planck Institute for Molecular Biology, Berlin
22 Bernhard Korn, German Genome Resource Centre, RZPD, Heidelberg
23 Urpo Lamminmaki, Dept of Biotechnology, University of Turku
24 Ulf Landegren, Rudbeck Laboratory, Uppsala University
25 Sophie Laurenson, Hutchison/MRC research centre, Cambridge
26 Joshua LeBaer, Harvard Institute of Proteomics, Cambridge (USA)
27 Federico de Masi, EMBL, Heidelberg
28 John McCafferty, Sanger Institute, Cambridge
29 Serge Muyldermans, Vrije Universiteit, Brussels
30 Per Ake Nygren, Royal Institute of Technology, Stockholm
31 Nelly Papin, MolTools, CNG, Paris
32 Mathias Paschke, Charite Medical School, Berlin
33 Andreas Plückthun, Biochemisches Institut, Zurich University
34 Fredrik Ponten, Rudbeck Laboratory, Uppsala University
35 Paul Predki, Invitrogen, Branford, Conneticut
36 Timo Pulli, MolTools, DKFZ, Heidelberg
37 Carolina Rydin, MolTools administrator, Rudbeck Laboratory, Uppsala
38 Sascha Sauer, MolTools, Max Planck Institute for Molecular Biology, Berlin
39 Edith Schallmeiner, MolTools, Rudbeck Laboratory, Uppsala University
40 Steffen Schlehuber, Pieris Proteolab AG, Freising
41 Cheryl Smythe, ESF coordinator, Babraham, Cambridge
42 Mike Taussig, Babraham, Cambridge
43 Jean-Jacques Toulme, Institut Européen de Chimie et Biologie, Bordeaux
44 Marius Ueffing, GSF, Neuherberg
45 Mathias Uhlen, Royal Institute for Biotechnology, Stockholm
46 Martin Yuille, GeneExpress, Cambridge

Many thanks to Annemarie Honneger for taking some great pictures. For more photos of the meeting click here.