antibodies technology feature To affinity and beyond are the researcher’s Swiss army knife: there’s a tool for every purpose, and they’re made in more shapes and sizes than ever. Pete Moore and Julie Clayton report.

ntibodies not only protect us from from investors requires us to have a pretty For therapeutic use in humans, the gold infection, they have been exploited clear clinical focus.” standard is a fully human monoclonal anti- Afor years in the laboratory — in diag- Simple financial incentives drive much of body.Antibodies in which the antigen-bind- nostic tests, to purify and as the the work in therapeutic antibodies. Donald ing site is from a mouse, even when most workhorses of cell biology to detect, locate Drakeman, president and chief executive of of the rest of the is human, as in and identify cellular proteins. Designed by Medarex in Princeton, New Jersey, estimates chimaeric and humanized antibodies, will evolution to recognize and bind virtually that it takes some US$20 million to get a always risk triggering an unwanted immune any molecule that could exist, what more traditional small-molecule drug ready for response that, at best, neutralizes the thera- could we ask of them? Quite a lot, it seems. clinical testing, whereas this could cost only peutic antibody and renders it ineffective. To meet the demands of today’s researchers $2 million for an antibody. Finding the right human antibody has been and drug developers, antibodies are being made much simpler by the advent of genetic cut down to size, tweaked into new shapes, Consulting the library engineering and the development of vast and manufactured in entirely new ways. The business end of an antibody molecule human antibody libraries packed with Antibodies as research reagents are is the antigen-binding site. This is the part potential candidates. These libraries are ubiquitous. Therapeutic antibodies have had of the antibody that latches onto a specific based on phage-display technology and store a bumpier ride, but now they’re back — and antigenic determinant or epitope — a com- and express billions of human variable here to stay.Crucially,the large pharmaceuti- plementary structure on the surface of regions. Each phage carries a DNA sequence cal companies have regained interest. “Anti- another molecule. The antigen-binding site encoding a different variable region,and dis- bodies are causing a lot of excitement because is also known as the variable region, or vari- plays the corresponding on its sur- there are not many new drugs out there and able domain, as it is the part of the antibody face, usually as part of a phage coat protein. 30% of the biotech industry is driven by anti- molecule that varies most among the bil- When a potential drug target is screened body technology,” says Peter Hudson of lions of different antibodies that can be against the library, only phage carrying anti- CSIRO Health Sciences and Nutrition in made, either naturally in the human body bodies with an affinity for the target lock on Melbourne, Australia. Greg Winter of the or by protein engineering. As a large biolog- to it. The captured phage are multiplied in MRC Laboratory of in ical molecule such as a protein can contain bacteria, and the antibody genes re-isolated, Cambridge, UK, and co-founder of antibody many epitopes, you can either produce to be fashioned into genes that can be used to drug-discovery company Domantis, agrees. monoclonal antibodies that target just one produce large amounts of fully functional “A year or two back we were interested in both epitope on a molecule, or raise a collection monoclonal antibodies in mammalian cell clinical and research applications, but it is of antibodies — polyclonal antibodies — cultures, or into smaller antibody fragments very difficult to make money from research that bind to a number of different epitopes that can be produced more cheaply in yeast applications and it is clear that the funding on the molecule. or bacterial culture.

PLAYING WITH THE PIECES

Regeneron of Tarryton, New York, has turned the usual way of using vascular endothelial growth factor (VEGF) is in clinical trials for the treatment antibodies as therapeutics on its head. To produce therapeutic agents that of cancer by preventing the growth of blood vessels into solid tumours and will mop up unwanted cytokines or growth factors, the company has non-Hodgkin’s lymphoma. “Every time we compare them with a developed a hybrid molecule called a ‘Trap’. This is made from the stem conventional antibody they have affinities two or three orders of magnitude of an antibody fused to two receptors for the target molecule. In this higher,” says George Yancopoulos, Regeneron’s chief scientific officer. application, the antibody stem Unlike antibodies, Traps will pick up every molecule that normally binds to gives the molecule a long half-life the receptor, so the VEGF Trap will also bind the related growth factor PLGF. in the circulation, while the Immunotoxins — antibodies linked to cytotoxic drugs — such as receptors replace the variable Wyeth’s Mylotarg, have the potential to avoid some of the toxic side effects regions. Traps therefore bind of traditional cancer chemotherapy by delivering the drug directly to the tightly to their target and can be tumour cell. Seattle Genetics of Washington state is developing SGN-15, given at infrequent intervals. an antibody–drug conjugate composed of a mouse–human chimaeric Regeneron has created an chemically linked to the cytotoxic drug doxorubicin. interleukin-1 (IL-1) Trap that it This is currently in a phase II clinical trial for the treatment of lung cancer. hopes may be effective against The antibody binds to a Lewisy-related antigen expressed on many types rheumatoid arthritis, and is of solid tumour. SGN-35, one of the company’s second-generation developing a dual IL-4/IL-13 Trap immunotoxins now in preclinical trials, incorporates a potent synthetic Regeneron’s VEGF Trap. as a potential treatment for asthma drug and a novel chemical linkage that makes the molecule more stable and other allergies. A Trap for in the blood. P.M.

NATURE | VOL 426 | 11 DECEMBER 2003 | www.nature.com/nature © 2003 Nature Publishing Group 725 antibodies technology feature

A recent success of the phage-display found in the blood of healthy peo- technology is HUMIRA, the first fully ple,but CAT has shuffled the DNA human monoclonal antibody to be approved sequences that code for the vari- for therapy, developed by Cambridge Anti- able region to give even more dif- body Technology (CAT) in Cambridge, UK, ferent types of antigen-binding and Abbott Laboratories in Abbott Park,Illi- site. Having started with human nois.This antibody,which blocks the inflam- genes, CAT believes its antibodies matory cytokine TNF-Ȋ,was approved in the are less likely to create immuno- United States in 2002 and in Europe in Sep- genic responses than libraries tember this year as a treatment for the symp- built with synthetic sequences. toms of rheumatoid arthritis. Antibody discovery and pro- CAT’s phage-display library holds single- duction company BioInvent in chain variable fragments (scFv) which con- Lund, Sweden, has a similar library sist of the variable ends of a light and a heavy called n-CoDeR, which holds chain linked with a short peptide.The library human scFv. Like CAT, BioInvent Going for GOLD with MorphoSys. now boasts more than 100 billion distinct starts with natural human antibody antigen-binding specificities. And in the genes as the source of the DNA sequences, cross-licensing its intellectual property with world of antibody discovery, most people which are then further diversified by making other major players, such as Genentech, think that size matters.The larger the library, variants of the antigen-binding regions. “We Xoma,Affimed,Biosite and CAT.Chief scien- the greater the chance that it contains an have taken evolution beyond nature and the tific officer Clive Wood believes that one of antibody for your given target.“It means you library now has more than 1010 members,”says the company’s strengths is the newness of its can get reasonably high-affinity antibodies Eskil Söderlind, who invented the n-CoDeR antibody libraries.“We are the new kid on the out of your first collection,”says Winter. technology with Carl Borrebaeck. BioInvent block, but because of that we have important The library started as a collection of anti- can screen up to 20,000 clones per day against a advantages.We have been able to incorporate body genes corresponding to antibodies customer’s target antigen on its automated new features into our Fab library,such as hav- system.The scFv fragment genes are then con- ing a mixture of natural and synthetic diversi- verted into complete antibody genes, and the ty that allows us to pull out antibodies with antibody itself produced in larger quantities picomolar to nanomolar affinity.”The largest

BIOINVENT from genetically engineered mammalian cells. Dyax library contains about 37 billion dis- The biopharmaceutical company Dyax in tinct human antibodies, and the company Cambridge, Massachusetts, has phage-dis- has two therapeutic proteins derived from play libraries holding Fab fragments — an phage display in three phase II trials. entire single ‘arm’ of an antibody. Although The Human Combinatorial Antibody the company holds many of the foundation Library (HuCAL) of MorphoSys in Martin- patents for phage-display technology,Dyax is sried, Germany, offers 10 billion totally syn- a relative newcomer in applying phage dis- thetic variable regions modelled on partial play to antibodies,having initially focused on human variable regions and displayed in peptide and protein display. The company phage. MorphoSys has now introduced Eskil Söderlind: 10 billion antibodies. has moved rapidly into the antibody arena by HuCAL GOLD, in which a disulphide bond

GOING INTO PRODUCTION

“Some of the therapeutic antibodies are required now in tens of thousands “There’s a whole category of antibodies that have not been effectively of grams per year,” says John Birch, chief scientific officer of the Lonza explored for therapeutic or preventive use, and for facing down viruses in the Group, based in Basel, Switzerland, which is planning three new 20,000-litre mouth, genito-urinary and respiratory tract,” says Elliott Fineman, president mammalian cell-culture reactors in the United States for monoclonal and chief executive of Planet Biotechnology in Hayward, California. These antibody production, in addition to its present capacity. Antibody harvests are the secretory IgA antibodies, which cannot be produced from mammalian from cultured mammalian cells are around a gram per litre of culture. cell culture in significant amounts. Instead, his company is using transgenic But the cost of scaling-up production using mammalian cells can be tobacco plants to produce CaroRx, a monoclonal secretory IgA against the prohibitive, particularly for a clinical trial of an antibody that may not turn out bacterium Streptococcus mutans, a leading cause of tooth decay. Originally to be worth it. Looking to fill a gap in the market are alternatives such as the created as ‘Guys 13’ by Thomas Lehner and Julian Ma at Guy’s Hospital in transgenic goats of GTC Biotherapeutics in Framingham, Massachusetts, London, CaroRx has shown its ability to block colonization by S. mutans in and the transgenic rabbits produced under a GTC Biotherapeutics licence by phase II clinical trials, and Planet Biotechnology is hoping to apply shortly for BioProtein Technologies of Paris, France. “If I need it, I breed it — the walls of a licence to market the antibody in Europe for topical application by dentists my bioreactor are expandable,” says Thomas Newberry, vice-president of and patients following antiseptic cleansing of teeth. If the company GTC Biotherapeutics. He estimates that the cost of producing a founder goat succeeds, it will be the first ‘plantibody’ to reach the market. is about US$10 million. Although pricey, this is a fraction of the estimated Fineman estimates that building, equipping and validating a facility $200 million–500 million that is needed to build a large-scale mammalian cell- to produce 100 kilograms of antibody from plants would cost between culture bioreactor. The company estimates that a single goat can produce up $30 million and $50 million. Keeping cost down is particularly important for to a kilogram of pure antibody per year. Goat-produced antibodies are in antibodies intended for non-therapeutic use against diseases that are not clinical trials to determine whether they match the safety, efficacy and life-threatening. “Nobody’s going to take a £2 hike in the cost of pharmacokinetics of more traditionally produced counterparts. toothpaste,” says Ma. J.C.

NATURE | VOL 426 | 11 DECEMBER 2003 | www.nature.com/nature © 2003 Nature Publishing Group 727 antibodies technology feature links the antibody to the phage instead of the colleagues at Alexion Antibody Technologies antibody being an integral part of one of the in San Diego, California, to identify high- coat proteins.This makes it easier to separate affinity human monoclonal antibodies to the antibody at a later stage. MorphoSys also critical elements of the anthrax toxin (M. A. offers ‘Antibodies by design’, a service that Wild et al. Nature Biotechnol. 21,1305–1306; will develop recombinant monoclonal 2003). The technology is also applicable to HuCAL antibodies for individual targets. diseases such as smallpox,says Bowdish. “This is especially interesting for researchers As the source of its library, Affitech in in industry and academia,as the turnaround Oslo, Norway, also uses blood from selected time is only about two months and the cost donors known to produce antibodies of starts at about E5,000 [US$6,000],” says interest, such as cancer patients, or people Claudia Gutjahr-Loeser of MorphoSys. who have been vaccinated or are recovering Rather than going for comprehensive- from an infectious disease. Affitech’s chief ness, a human-antibody library can be executive Martin Welschof says that this, Frozen phage: CAT’s library in store. biased in favour of particular targets, for combined with the company’s proprietary example by taking the initial antibody cod- screening system, avoids the need for repeti- advantages,”says Ian Tomlinson,co-founder ing sequences from individuals vaccinated tive rounds of selection and means that it can and chief scientific officer of Domantis. In against a particular disease. Blood from sol- move from a 10-ml blood sample to a high- partnership with Peptech in North Ryde, diers immunized against anthrax was the affinity human monoclonal antibody of the New South Wales, Australia, Domantis is starting point for a Fab antibody library desired specificity within four weeks. developing a domain antibody against TNF- that has allowed Katherine Bowdish and her Ȋ for the treatment of rheumatoid arthritis Small is beautiful and other inflammatory diseases. If a therapeutic antibody has to penetrate Part of the inspiration for pursuing far into body tissues, or into a tumour, for domain antibodies was the natural domain example, smaller may well be better. antibodies found in camels, dromedaries, Although some companies are concentrat- llamas — and sharks.These natural antibod- ing on scFv antibody fragments, UK-based ies are also being explored for their potential Domantis is producing even smaller human applications by start-up biopharmaceutical antibodies from its phage libraries: these company Ablynx of Zwijnaarde,Belgium. consist of just a single variable domain. Like scFv fragments, these ‘domain anti- Mouse power bodies’ can be mass-produced relatively Both Abgenix of Fremont, California, and cheaply in Escherichia coli or yeast. They are Medarex of Princeton, New Jersey, take a also very stable,which opens up new options different approach to making large numbers for therapy, such as a powder that can be of different human antibodies for screening inhaled to treat asthma and other lung dis- — they get mice to do it for them. In eases, or a tablet to be taken orally for gut Abgenix’s XenoMouse and Medarex’s Katherine Bowdish: getting to grips . “We’re only going after areas HuMab-Mouse, murine germ-line antibody with anthrax. where domain antibodies have unique genes have been replaced with human ones.

ANTIBODIES WHERE YOU WANT THEM

Antibodies are normally built inside cells, take up intrabody that has an affinity close to or better than their final form as they are secreted from the cell, the affinity of the interacting partner, you should be and function exclusively outside cells, binding to able to block the interaction using the intrabody”, cell surfaces and extracellular molecules. But Terry says Rabbitts. Rabbitts at the MRC Laboratory of Molecular Intrabodies offer the cell biologist the prospect Biology in Cambridge, UK, wants to use antibodies’ of highly specific tools that can probe highly specific binding properties inside the cell protein–protein interactions in situ more delicately itself. He has engineered cells to make fragments than techniques that simply remove one of the consisting of just a heavy-chain variable domain proteins from the scene by antisense or gene and hold them within the cell. These single-domain knockout. Because an intrabody binds to a specific antibodies can interact with intracellular proteins at site on a protein, it can block a protein’s interaction binding affinities of 10 nanomolar or better. At Iclectus: the inside story. with one partner while allowing it to continue acting Iclectus, an MRC spin-off company based in with another. London, Rabbitts’ colleagues are developing these fragments as intracellular Another use is for the validation of potential drug targets, one of the first antibodies or ‘intrabodies’. and vital steps in drug discovery. Most potential target proteins are modular, They became interested in intrabodies while they were unravelling the so intrabodies that bind to and block the function of individual modules could role of transcription factors involved in producing chromosomal help to determine which part of the protein is key to causing the disease. translocations in tumour cells. Using intrabodies to bind and block these Placing localization signals on an intrabody allows it to be directed to transcription factors inside the cell opened up a new range of research various parts of the cell, such as the nucleus. “With intrabodies you can be possibilities as well as pointing the way to new anticancer therapies. more versatile in your experiments and ask more questions than, say, with Transcription factors operate by binding other proteins, so “if you have an RNA interference,” says Rabbitts. P.M.

NATURE | VOL 426 | 11 DECEMBER 2003 | www.nature.com/nature © 2003 Nature Publishing Group 729 antibodies technology feature

When challenged by a foreign protein, which works really well,” says these mice produce fully human, func- Andrew Bradbury of the Los Alamos tional antibodies. Medarex markets National Laboratory in New Mexico, the HuMAb-Mouse as one part of its who is looking to commercialize the UltiMAb platform. “Buying the technology. UltiMAb system allows our partners to But despite the prospect of novel use any of three different mice to reagents such as fluorobodies and develop products,” explains Drake- intrabodies, polyclonal antibodies man. One is the HuMAb-Mouse. The raised in the traditional way by immu- second is Kirin TC Mouse, initially nizing animals are still in demand, built by Kirin Brewery in Tokyo, Japan. even for the latest protein array tech- This is a transchromosomal mouse nologies. For example, using high- containing the 15-megabase segment throughput techniques, Eurogentec of human chromosome 14 that con- in Seraing, Belgium, will take a cus- tains all the genes needed to build tomer’s mRNA library and produce human antibodies. The third is KM- peptides representing different parts Mouse, a cross-bred version of of the encoded proteins. Polyclonal HuMAb-Mouse and KM-Mouse. antibodies specific for each peptide are SANGER INST. OF GENE EXPRESSION PROJECT, TRUST WELLCOME ATLAS Using this technology, Medarex has Clear view: laminin expression in kidney core. then raised in rabbits. After selection generated 15 antibodies that are cur- to reduce the chance of crossreactions, rently in clinical trials, while Abgenix has Bender MedSystems of Vienna, Austria, the antibodies are spotted onto glass slides to two in its own pipeline and others being offer ‘instant ELISAs’,complete ELISA kits to produce a customized array of up to 1,000 developed by Pfizer and Amgen. which the user only has to add water. These antibody specificities, which can be used to Whether you use a phage or mouse sys- are available for a range of human, mouse screen and compare tissue samples for the tem can be influenced by what you are trying and monkey immune-system cytokines and differential expression of multiple proteins. to achieve.Engineered mice don’t have all the other proteins. Arrays of hundreds of anti- The approach is intended to complement the answers,claims BioInvent’s Söderlind.“Mice bodies are also currently the basis for the ELISA, which screens a single antibody won’t react with all substances, as they may up-and-coming field of ‘protein chips’, the against multiple cell or tissue samples,accord- have innate tolerance to the antigen, and if proteomics equivalent of DNA microarrays ing to chief executive Gottfried Proess. the target is toxic they may die before any (see Nature 424, 581–587;2003). antibodies are formed,”he says. In those cir- Companies such as Sigma-Aldrich in Antibody’s-eye view cumstances, in vitro phage-based systems St Louis, Missouri; Chemicon in Temecula, A renaissance is taking place in cellular offer a good way forward.On the other hand, California;Active Motif in Carlsbad,Califor- pathology, according to Tony Warford of say the owners of engineered mice, phage nia; MabTech in Nacka Strand, Sweden; and the Wellcome Trust’s Sanger Institute in libraries can only display antibodies that do Progen Biotechnik in Heidelberg, Germany, Cambridge, UK. This is thanks to the devel- not kill bacteria. are just some of many that supply off-the- opment of automated high-throughput shelf monoclonal and polyclonal antibodies immunohistochemistry of intact normal or Antibody assays against thousands of cellular targets. diseased tissue samples. Labelled antibodies Despite a preference for human antibodies Interest doesn’t stop at antibodies against are used to stain tissues to identify the in the clinic, they are not the first choice in human and mouse proteins for biomedical expression of particular proteins. As a other settings. Monoclonal antibody tech- research. BD Biosciences Pharmingen senior project leader for John McCafferty’s nology started with mouse antibodies, and has thought it worthwhile to screen its large ‘Atlas of Gene Expression’ at the Sanger many people argue that mice are still in the catalogue of mouse monoclonal antibodies Institute, Warford, is pioneering the linking driving seat. Winter recalls remarking to the to human intracellular structural and together of automated immunohistochem- late César Milstein, who created mouse signalling proteins for good crossreactivity istry techniques. monoclonal technology in the 1970s, that against the homologous Drosophila proteins. By applying phage-produced monoclonal phage display was going to bypass the whole It now offers a list of about 25 high-quality antibodies as probes to composite micro- business of immunizing animals to get monoclonal antibodies selected to give good arrays of tissue samples, he can see which monoclonal antibodies. “He bridled at this results in Drosophila,and is adding more. proteins are expressed, when and where, in a a bit and said ‘you are wrong’ — and I think This wide array of laboratory reagents given cell and tissue type. The tissue micro- he was right,” says Winter. Milstein believed may soon be joined by a useful molecule arrays, which are created using automated there would be a niche for all types of anti- that combines the binding specificity of an technology from Beecher Instruments in bodies. And some of those niches are huge. antibody with built-in fluorescence. These Sun Prairie, Wisconsin, and Chemicon in For when antibodies are being used as ‘fluorobodies’would circumvent the need to Ternecula, California, consist of paraffin detection reagents in the laboratory,there is no add fluorescent tags to conventional anti- wax-embedded blocks of human or animal need for expensive high-tech human antibod- bodies.Fluorobodies consist of a super-fold- tissues containing up to 500 individual mini- ies.Rat,mouse and hamster monoclonal anti- ing version of green fluorescent protein tissue samples (each about 0.6 mm across and bodies derived by well-established hybridoma (GFP) with the minimal antigen-binding 3 mm deep). These can then be sliced up into technology, as well as polyclonal antibodies portions of antibodies, conferring the around 100 individual sections for immuno- raised by immunizing larger animals, are the specific protein-binding ability, grafted into histochemical staining. mainstay of laboratory immunoassays and four loops at one end of the GFP.They func- With automated processing, it will be immunodetection techniques such as ELISA tion as well as conventional antibodies in possible to screen and interpret up to 20,000 (enzyme-linked immunosorbent assay) and ELISAs, gel-shift assays, protein chip and individual samples — all in a day’s work. ■ its variant ELISpot, western blotting, flow flow cytometry assays — and without Pete Moore is a freelance science writer based near cytometry and immunofluorescent staining the need for secondary antibodies to provide Bristol, UK. Julie Clayton is a science writer based in of cells and tissues. a signal. “It’s a once in a lifetime invention Bristol, UK.

NATURE | VOL 426 | 11 DECEMBER 2003 | www.nature.com/nature © 2003 Nature Publishing Group 731