Bench philosophy: Non-immunoglobulin protein scaffolds
Beginning of the End for Traditional Antibodies?
by Steven Buckingham, Labtimes 02/2015
Until recently, we have taken advantage of natural ways to make antibodies, by raising immunity in living animals and harvesting the antibodies. But the scientific community is losing patience with the limitations of this approach.
Handy things, proteins. You could say they are nature’s favourite all-purpose tool, judging by the bewildering array of purposes for which she uses them. All due to the practically infinite combinations of the 20 common amino acids that form the proteins found in living things. With this simple alphabet of 20 letters, all the varieties of proteins with their diverse functions are formed: enzymes, receptors, scaffolds, lenses, signalling molecules, hormones and fingernails.
The secret of course is in the folding. Given the right sequence of amino acids, and a little help from some other proteins, the polypeptide will fold into a specific three-dimensional shape. The right mix of alpha chains, beta sheets, loops, bends and random coils sculpts the protein such that, coupled with just the right distribution of charge, hydrogen bonding and van der Waals’ Forces, you end up with a precisely-sculpted surface that does the job. This surface might bind tightly and specifically with another molecule (protein, DNA fragment, ion), or perhaps catalyse some sort of molecular change.
The usual picture of monoclonal antibodies binding always specifically to the right targets is far too optimistic. Photo: Wellcome Library, London
Think of that old lab workhorse, the antibody. Antibody molecules contain a variable domain, which adapts the molecule to bind with very high specificity to its target. Now, art imitates nature nowhere as much as in a biology lab, where using antibodies to label a target is standard. And nature has proved remarkably obliging in offering its own service to design bespoke antibodies. All we mere mortals have to do is to inject a rabbit with whatever we want an antibody to, and the adaptive immune response does the manufacturing on our behalf.
But the limitations of the obliging molecular workers that nature kindly seconds to us are beginning to exhaust the patience of the men in white coats and there is an increasing chorus, calling for a change in how we do things. First, there is the problem of specificity. The traditional way of making antibodies yields not one but a whole slew of antibodies, raising the danger that some of them will also latch on to something other than to that what you had intended. This lack of specificity was the driving force behind the development of monoclonal antibodies. These are produced by a single clonal line of cells descended from a unique parent cell, increasing uniformity. Yet even with monoclonals, perfect specificity is not guaranteed: they remain, after all, a cocktail of diverse molecules.
In the case of clinical trials, the stringent (but very costly) quality control exerted by the supplier companies has reduced this problem to a large extent. But this has not been the case in academic research. The lack of guidelines or standardisation in the production and packaging of antibodies for research has largely been to blame. Some have estimated that around a half of commercially-supplied antibodies are lacking in specificity or in some other way do not do what it says on the packet. To get an idea of the scale of the problem, C. Glenn Begley and co-workers (Begley, C. G. & Ellis, L. M. Nature 483, 531-33, 2012) found that they could not replicate more than six out of 53 landmark preclinical studies, which they attributed largely to antibody quality.
Another major issue with antibodies is the question of reproducibility. The problem is that when you inject a rabbit with the target protein, you don’t get the same set of antibodies each time, even if you use the same rabbit. And even if you opt for monoclonals, your research might come unstuck when the hybridomas (the cells that make the antibody) die or for some reason stop making the antibody, as can so often happen.
As if that weren’t enough, the size of an antibody is another problem − they are just too big for many applications. No wonder then that many leading scientists are calling for change. In February of this year, Andrew Bradbury and Andreas Plückthun rounded up a posse of 110 co-signatories and marched on the doors of the journal, Nature, demanding a shake-up of antibody practice, including a move towards better standardisation (Nature, 518, 27-8).
But there is a new trend emerging, as the focus shifts away from borrowing the adaptive immune response and turns increasingly towards engineering antibodies directly. Indeed, this was one of the developments strongly endorsed by Bradbury and Plückthun. The idea is to introduce a plasmid encoding the antibody into cells via transient transfection. That way you control exactly what gets expressed and you side-step all the variability introduced by the natural immunity mechanisms. A bit of a drudge for your average lab, perhaps. That is why companies like Absolute Antibodies offer to do it for you. And they are building up a collection of ready-made antibodies to sell off-the-peg. Increased purity, reproducibility and faster production time are a consequence of this approach.
But why use antibodies in the first place? If we are going to choose the engineering route, why not go all the way and reinvent our own alternatives from scratch? This is not fantasy − the methodology is already developing rapidly. How is it done? First, you need a scaffold. That is to say, you need a protein with a stretch of amino acids that you can mess around with, without completely changing the overall 3D structure. After all, isn’t that what the variable domain of an antibody is? There are plenty on offer already and if the scaffold you have in mind already binds a protein, all the better.
A wide choice of good scaffolds is already available. Take the fibronectin type III (FN3) domain, for instance. The great thing about FN3 is that it is structurally homologous to the Ig domain, so you can leverage what we already know about how the sequence of Ig makes it work, to decide how you are going to go about changing the FN3 scaffold.
Another handy scaffold to start from is our old friend, Green Fluorescent Protein. The big plus about GFP is that it does its own reporting: once it is bound, you can just read off the fluorescent image. But can you mess around with GFP without breaking its celebrate fluorescent properties? Until recently, people would have said, no, based on disappointing attempts to do so. But it turns out you can, after all: Eric Shustas group at the University of Wisconsin-Madison, USA, used a directed evolution approach coupled with yeast surface display, to screen the products to yield GFP-based binders with nanomolar affinity without losing fluorescence (Pavoor T.V., et al., 2009 Proc. Natl. Acad. Sci. USA 106:11895-900).
Can’t find a scaffold? Not to worry. Molecular modelling is at such a stage now that you can find one in a protein that doesn’t have any binding properties of its own. What matters, is to find a stretch that is exposed to the surface of the protein and allows the alteration, or insertion, of residues without changing the overall structure.
Before we go any further, a quick aside. Since we are engineering, we can do things our way, so while we are at it, let’s dump the problematic requirement of having disulphide bridges. Traditional antibodies use these to keep them in the right shape. The problem is, it means you can’t conveniently assemble the molecule in the strongly reducing cytosol of mammalian cells. Our molecule is going to be a single subunit, so we don’t need disulphide bridges.
Okay, so you have your scaffold. Now this is where the fun starts. With the crystallography and molecular modelling toolkit we have nowadays, we can easily identify potential binding motifs and possible candidate regions or residues for mutagenesis. You can start hacking the sequence of the scaffold, building a library with site-directed mutagenesis or by randomising using synthetic oligonucleotides and degenerate codons.
This will give you a nice big pool of variants, but which ones are worth chasing up? The next thing to be done, then, is to screen the proteins we have made. Doing it individually will take too much time. So instead, you should set up a high-throughput screen based on one of the many display technologies that somehow tie the mutant protein with its corresponding sequence. This becomes standard when you incorporate the protein into a viral coat, of course, because the sequence is wrapped up in the capsid. Or you can strap a receptor onto a cell membrane, so that as soon as the protein is secreted it gets captured. However it is done, the aim is to select for proteins with desired activity. Follow it up with a round or two of random mutagenesis (perhaps with error-prone PCR at the core) and you’ll end up with a fine-grained space of protein structures and their matching activities.
And it works. Non-Ig antibody alternatives being spawned with this kind of methodology are smaller and more specific than their organic, home-grown equivalent. And being directly engineered, there is more scope for creativity. For instance, many labs are playing around with alternative labels with not one but two recognition sequences, boosting the power of the labelling. This seems to be working as a potential treatment for cancer. The basic idea is to use an antibody with two business ends: one sticks to an antigen on a cancer cell, the other sticks to a cancer-killing effector cell, bringing hunter and prey together. The only problem is: it doesn’t work. At least, it hasn’t until now. The difference now is that engineered antibodies are better at the job and the result is that therapies based on them are now in clinical trials (Muller D., Kontermann R.E., 2010 BioDrugs 24:89-98).
Is this the end of the organic antibody as we know it? No. Well, not according to Peter Bell, Senior Director of Research and Development of Protein Biology at Thermo Fisher Scientific. “The research community will be slow to switch to a new framework for detection,” thinks Bell. “Traditional IgGs will continue to be used in immunoassays such as Western blotting, ELISA, Immunohistochemistry, and Immunofluorescence.” Part of the problem is confidence. “Diagnostic companies will continue to use validated monoclonal antibodies with only a slow transfer to non-immunoglobulin molecules due to validation requirements for switching,” says Bell. “Alternatives to immunoglobulins may enter the therapeutic market faster than they are likely to replace antibodies in the research market.”
All the same, antibody suppliers are starting to explore the alternative binder market. Francesca Axe of Abcam tells us that while “traditional” antibody sales continue at a healthy pace, they are keeping a close eye on new binder platforms, as well as offering a few themselves. Stefan Pellenz, Project Manager Independent Validation for Antibodies-Online, agrees that while polyclonals have had their day, recombinant monoclonals will be with us for a while yet. “That being said,” muses Pellenz, “antibodies are likely to be phased out as more and more, better alternatives are becoming available − at least in certain applications like intracellular staining or imaging of solid tumours. So, the proteomics market is going to change in the long run. And as the market changes, companies such as ours will change accordingly, to satisfy the needs of our customers.”
Bradbury and Plückthun speak of one billion dollars being required to produce the necessary recombinant antibodies. Clearly, there is a lot of money to be made in this business and we note that some of the co-signatories are from researchers that have just launched an antibody engineering start-up and have an interest pushing this field. Two factors working together may well seal the fate of good ol’ antibodies. On the one hand there is a growing concern over the dangers of bad antibodies. On the other hand there is a Precambrian explosion of new tools, some of which may well emerge as some sort of new standard.
Last Changed: 25.03.2015