NEWS AND VIEWS
`
`Next -generation protein drugs
`
`Ian M Tomlinson
`
`Ankyrin repeats generate high -affinity protein binders with biophysical properties that may favor
`therapeutic applications.
`
`Ask any major pharmaceutical company what
`constitutes an ideal drug and the answer
`would probably include the words `specificity,
`affinity, solubility, stability and safety' along
`with the phrases `cheap to manufacture, easy
`to formulate, simple to deliver and the right
`pharmacokinetic profile: Ironically, many
`drugs on the market fail to deliver in one or
`more of these areas because of their sub -opti-
`mal biophysical makeup. Even blockbuster
`biologics, such as therapeutic antibodies',
`suffer from drawbacks, such as the require-
`ment for an expensive mammalian cell pro-
`duction system and the need for intravenous,
`intramuscular or subcutaneous
`injection
`(with molecular weights of around 150,000,
`they are too large to be administered by any
`is room for
`other route). Clearly, there
`improvement. In this
`issue, Binz et al.2
`describe a natural scaffold, ankyrin repeat
`protein, that has promising biophysical prop-
`erties for therapeutic application. Ankyrin
`repeats are one of several new types of scaf-
`fold being developed for a new generation of
`protein therapies.
`An ideal drug would have the following
`qualities: it would have very high affinity and
`exquisite specificity for its target; it could
`be manufactured by the bucket -load in bacte-
`ria or yeast; it would be both incredibly solu-
`ble and remarkably stable;
`it could be
`delivered to any part of the human body by
`any route of administration; and, once there,
`it would hang around long enough to have
`the desired therapeutic effect. Achieving all
`these goals has been particularly difficult for
`protein drugs.
`Currently, protein drugs come in all shapes
`and sizes: some are recombinant human pro-
`teins (for instance, insulin, growth hormone
`
`Ian M. Tomlinson is Chief Scientific Officer of
`Domantis Limited, 3I5 Cambridge Science
`Park, Cambridge CB4 OWG, UK.
`e -mail: ian.tomlinson @domantis.com
`
`Ankyrin repeat protein
`
`Ribosome display selection for
`high -affinity binders to
`maltose- binding protein
`
`i?
`
`,
`
`.
`
`- _-
`
`>
`
`i
`
`-
`
`Maltose-
`binding
`protein
`
```
`
`- - -
`
`-
`
`- J
`1"e5r.
`.-, a..,
`
`Il
`
`q
`
`Ankyrin
`repeat
`protein
`
`i=
`
`and erythropoietin), others are monoclonal
`antibodies (for instance, Remicade (influx-
`imab; Johnson & Johnson, Kenilworth, NJ,
`USA), Rituxan (rituximab; Genentech; S. San
`Francisco, CA, USA) and Erbitux (cetuximab;
`ImClone, New York, NY, USA) and still others
`are viral or bacterial proteins used as vaccines
`to elicit a specific immune response. Nature
`did not evolve proteins for manufacture ex
`vivo. For this reason, many human proteins
`produced in recombinant form are difficult to
`manufacture and some cannot be expressed
`at all in microbial cell culture. Furthermore,
`the serum half -life and tissue distribution of
`endogenously expressed proteins is carefully
`controlled in vivo to optimize their biological
`activity. Most human proteins are not
`
`NATURE BIOTECHNOLOGY VOLUME 22 NUMBER 5 MAY 2004
`
`Figure 1 All in a bind. Binz et al. randomized 6
`of the 33 amino acids (red side chains) in three
`ankyrin repeats (dark blue) and, using ribosome
`display, isolated a range of nanomolar binders to
`mannose- binding protein. The co- crystal structure
`confirms the predicted binding of the engineered
`ankyrin repeat protein to the mannose- binding
`protein target.
`
`designed to be administered from outside the
`body. Recombinant proteins therefore tend to
`be rapidly cleared and thus require frequent
`interest
`(thus, the growing
`injection
`in
`extending the serum half -life by, for example,
`polyethylene glycol conjugation).
`Antibodies have proved useful as human
`protein therapeutics because they exhibit a
`favorable pharmacokinetic profile. After a
`single injection, they can persist for a long
`time in the bloodstream, maintaining their
`biological activity for several weeks. However,
`antibodies have also evolved to be secreted
`from mammalian cells and, for a variety of
`reasons, cannot be expressed in yeast or bac-
`terial cell culture.
`Given the limitations of current protein
`therapies, scientists are starting to develop
`more tailored approaches to drug design
`whereby you first assemble a list of the vari-
`ous properties you want the drug to have and
`then engineer a drug with precisely those
`properties. Over the past three years, several
`new biotech companies have been set up to
`exploit the use of `well- behaved' human pro-
`teins as scaffolds to create a range of designer
`protein drugs that have improved therapeutic
`properties (see Table 1). This approach pro-
`ceeds through the following steps: first, chose
`a human protein that is well expressed in bac-
`teria and /or yeast and has good biophysical
`properties (solubility, stability and others);
`second, create a repertoire by introducing
`diversity into the loop regions of the given
`scaffold, preferably in a way that does not dis-
`rupt the overall structure of the protein; third,
`
`521
`
`Lilly Exhibit 1266
`Eli Lilly & Co. y. Teva
`Phnrmt Inl'l GM111 -1
`
`

`

`NEWS AND VIEWS
`
`Table 1 Selected companies using human proteins as scaffolds to create next -
`generation drugs
`
`Company
`
`BioRexis (King of Prussia, PA, USA)
`Borean Pharma (Aarhus, Denmark)
`Compound Therapeuticsa (Waltham, MA, USA)
`Domantis (Cambridge, UK)
`Dyax (Cambridge, MA, USA)
`Pieris ProteoLab (Freising -Weihenstephan, Germany)
`
`Protein scaffold
`
`Transferrin
`C -type lectins
`Trinectins
`Domain antibodies
`Kunitz domains
`Lipocalins
`
`°On 9 March 2004, Compound Therapeutics announced the acquisition of the intellectual property estate of Phylos
`(Lexington, MA, USA).
`
`strate in vivo efficacy with an engineered
`ankyrin repeat protein, the libraries they have
`created should be a valuable resource for
`the isolation of therapeutically relevant leads
`that are both well expressed and highly
`stable.
`Undoubtedly, there is a big drive for the
`drugs of the future to be much easier to man-
`ufacture and administer to patients. They
`must also be highly efficacious with few, if
`any, side effects. By wiping the slate clean and
`designing potent drugs based on human pro-
`tein scaffolds with good biophysical proper-
`ties, we may find that the ideal drug is closer
`than ever before.
`
`1. Reichert, J.M. Nat. Biotechnol. 19, 819 -822 (2001).
`2. Binz, H.K. et al. Nat. Biotechnol. 22, 575 -582
`(2004).
`3. Scott, J.K. & Smith, G.P. Science 249, 386 -390
`(1990).
`4. Mattheakis, L.C., Bhatt, R.R. & Dower W.J. Proc. Natl.
`Acad. Sci. USA 91, 9022 -9026 (1994).
`5. Ali, S.A., Joao, H.C., Hammerschmid, F., Eder, J. &
`Steinkasserer, A. J. Biol. Chem. 274, 24066 -24073
`(1999).
`6. Jespers, L., Schon, O., James, L.C., Veprintsev, D. &
`Winter, G. J. Mol. Biol. 337, 893 -903 (2004).
`7. Sedgwick, S.G. & Smerdon, S.J. Trends Biochem. Sci.
`24, 311 -316 (1999).
`
`two or three ankyrin repeats. The diversity
`thereby generated (12 or 18 randomized
`residues) is sufficient to isolate a range of
`nanomolar binders
`to mannose- binding
`protein using ribosome display, all of which
`the desirable biophysical properties
`have
`of the parental ankyrin scaffold. Import-
`antly, the authors also showed, by co- crystal-
`lization, that the selected binders have the
`same structural fold as the parental scaffold.
`Although the authors have yet to demon-
`
`use a genotype -phenotype display system
`(such as phage3 or ribosome display) to
`select a range of binders to a given therapeutic
`target; and fourth, use some form of screen to
`identify those leads that have the desired bio-
`logical activity. If all goes according to plan,
`the outcome should be a protein that has all
`the desirable biophysical properties of the
`parental scaffold and the required potency for
`the therapeutic target.
`Of course, it's not always that straightfor-
`ward. In many cases, the mutations intro-
`the
`target
`to enable binding
`duced
`to
`compromise
`biophysical properties
`the
`and/or the three -dimensional structure of the
`parental scaffold. In some cases, it may not
`to achieve the desired
`even be possible
`potency using such a small binding footprint
`to the given target. And of course each start-
`
`ing scaffold has its unique pros and cons -
`
`intracellular or extracellular expression, diff-
`erent binding sites for purification, immuno-
`genicity and so on. Some scaffolds may have
`intrinsically long serum half- livess, whereas
`others may show unusual properties, such as
`the ability to refold reversibly after heat
`denaturation6.
`In the present paper, Binz et al. focus on the
`use of one particular scaffold, based on
`ankyrin repeats, to generate binders with bio-
`physical properties designed for therapeutic
`application. Ankyrins are proteins, first iso-
`lated in mammalian erythrocytes, involved in
`the targeting, mechanical stabilization and
`orientation of membrane proteins to special-
`ized compartments within the plasma mem-
`brane and endoplasmic reticulum. Natural
`ankyrin repeat proteins consist of many 33-
`amino -acid modules, each comprising a (3-
`turn and two anti -parallel a- helices7. They do
`not contain any disulfide bonds and therefore
`can be expressed at very high yields in the
`bacterial cytoplasm. They also seem to be
`both highly soluble and stable.
`The approach used by Binz et al. ran-
`domizes 6 of the 33 amino acids in each of
`
`overcoming the gridlock in
`discovery research
`
`Jonathan Margolis & Greg D Plowman
`
`Chemical screening in a zebrafish mutant has turned up two compounds
`that rescue a heart defect, but will this yield new drugs?
`
`Why is it so hard to find new drugs? A signifi-
`cant amount of time goes into finding and
`validating new drug targets for the develop-
`ment of small -molecule or biotherapeutic
`leads. Why not create in vivo disease models
`and directly screen for compounds that can
`ameliorate the disease state? In this issue,
`Peterson et al.' describe an effort to do just
`that, using a zebrafish mutant with an
`anatomical defect that resembles a malforma-
`tion in the human heart.
`theory, whole -organism
`screening
`In
`should circumvent the need to identify spe-
`cific drug targets, allowing the entire genome
`to be screened in a single, unbiased assay. This
`
`Jonathan Margolis and Greg D. Plowman
`are at Exelixis, 170 Harbor Way, P.O. Box 511,
`South San Francisco, CA, USA.
`e -mail: gplowman @exelixis.com
`
`approach is equivalent to a classical genetic
`mutant suppressor screen, in which one
`searches for secondary mutations that revert
`the abnormal phenotype
`to wild
`type.
`Sensitized cell -based screens have previously
`been used to identify chemical suppressors of
`a disease process -for example, drug leads
`that block the proliferation of carcinoma
`cells2. Peterson et al. extend this strategy to a
`vertebrate organism. They describe corn -
`pounds that rescue the abnormal vascular
`development of a zebrafish mutant and sug-
`gest that this could provide a rapid path to
`drug leads for diseases whose underlying biol-
`ogy is not well understood.
`The gene encoding the gridlock transcrip-
`tion factor is a classic developmental selector
`controlling the choice of angioblasts between
`venous and arterial fates in the developing
`fish heart (Fig. 1). With a partial reduction of
`gridlock activity, the bifurcation of the lateral
`
`522
`
`VOLUME 22 NUMBER 5 MAY 2004 NATURE BIOTECHNOLOGY
`
`