LETTERS TO NATURE
`
`3. Frasch, M. EMBO 110, 1225-1236 (1991).
`4. Aebi, M., Clark, M. W., Vijayraghavan, U. & Abelson, J. Molec. gen. Genet. 224, 72-80 (1990).
`5. Clari<, K. L. & Sprague, Jr. G. F. Molec. cell. Biol. 9, 2682-2694 (1989).
`6. Kai, R., Ohtsubo, M., Sekiguchi, M. & Nishimoto, T. Molec. cell. Biol. 6, 2027 -2032 (1986).
`7. Ohtsubo, M. et al. Genes Dev. 1, 585-593 (1987).
`8. Ohtsubo, M., Okazaki, H. & Nishimoto, T. 1 Cell. Biol. 109, 1389-1397 (1989).
`9. Bischoff. F. R., Maier, G., Tilz, G. & Ponstingl, H. Proc. natn. Acad Sci. US.A. 87, 8617-8621 (1990).
`10. Bischoff, F. R. & Ponstingl, H. Proc. natn. Acad. Sci. US.A. (in the press).
`11. Drivas, G. T., Shih, A., Coutavas, E., Rush, M. G. & D'Eustachio, P. Mo/ec. cell. Biol. 10, 1793-1798
`(19901.
`12. Valencia, A., Chardin, P., Wittinghofer, A. & Sander. C. Biochemistry 30, 4637-4648 (1991).
`13. Crechet, J.-B. et al. Science 248, 866-868 (1990).
`14. Wolfman, A. & Macara, I. G. Science 248, 67-69 (1990).
`15. West, M., Kung, H.-F. & Kamat.a, T. FEBS Lett. 259, 245-248 (1990).
`16. Bourne, H. R., Sanders, D. A. & McCormick, F. Nature 349, 117-127 (1991).
`17. John, J., Frech, M. & Wittinghofer, A. J. biol. Chem. 263, 11792-11799 (1988).
`18. Matsumoto, T. & Beach, D. Cell 66, 347-360 (1991).
`19. Nishimoto, T., Eilen, E. & Basilica, C. Cell 15, 4 75-483 (19781.
`20. Chardin. P. Cancer Cells 3, 117-126 (19911.
`
`ACKNOWLEDGEMENTS. We thank R. H. Himes for discussions, J. Kretschmer for technical assistance.
`T. Nishimoto for a gift of anti-RCC1-antibody, and A. Wittinghofer for a gift of Ha-ras protein and
`for discussions. This work was supported by the Deutsche Forschungsgemeinschaft.
`
`A new type of synthetic
`peptide library for
`identifying ligand-binding
`activity
`Kit S. Lam, Sydney E. Salmon, Evan M. Hersh,
`Victor J. Hruby*, Wieslaw M. Kazmierskit
`& Richard J. Knappt
`Arizona Cancer Center and Department of Internal Medicine,
`College of Medicine, Tucson, Arizona 85724, USA
`* Department of Chemistry, Faculty of Science, University of Arizona,
`Tucson, Arizona, USA
`t Selectide Corporation, 10900 N. Stallard Place, Tucson,
`Arizona 85737, USA
`
`OUR aim was to improve techniques for drug development by
`facilitating the identification of small molecules that bind with
`high affinity to acceptor molecules (for example, cell-surface recep(cid:173)
`tors, enzymes, antibodies) and so to mimic or block their interaction
`with the natural ligand 1
`2
`, Previously such small molecules have
`'
`been characterized individually on a serial basis. The systematic
`synthesis and screening of peptide libraries of defined structure
`represents a new approach. For relatively small libraries, predeter(cid:173)
`mined sequence variations on solid-phase supports have been
`used3.4, and large libraries have been produced using a bac(cid:173)
`teriophage vector into which random oligodeoxynucleotide sequen(cid:173)
`ces have been introduced5
`-B, but these techniques have severe
`limitations. Here we investigate an alternative approach to syn(cid:173)
`thesis and screening of peptide libraries. Our simple methodology
`greatly enhances the production and rapid evaluation of random
`libraries of millions of peptides so that acceptor-binding ligands
`of high affinity can be rapidly identified and sequenced, on the
`basis of a 'one-bead, one-peptide' approach.
`Our method involves creating a large peptide library consist(cid:173)
`ing of millions of beads, with each bead containing a single
`peptide and with the complete collection representing the uni(cid:173)
`verse of possible random peptides in roughly equimolar propor(cid:173)
`tion. It is clearly not enough to use a random mixture of activated
`amino acids in a peptide synthesis protocol, because the widely
`different coupling rates of different amino acids will lead to
`unequal representation and because each bead will contain a
`mixture of different peptides. Our solution was to use a 'split
`synthesis' approach. The first cycle consisted of distributing a
`pool of resin beads into separate reaction vessels each with a
`single amino acid, allowing the coupling reactions to go to
`completion, and then repooling the beads. This cycle was
`repeated several times to extend the peptide chain (Fig. la). In
`this fashion, each bead should contain only a single peptide
`species.
`82
`
`We then developed a rapid approach for screening the library
`to find beads containing peptides able to bind to any particular
`acceptor molecule. Acceptor molecules were coupled to an
`enzyme (alkaline phosphatase) or to fluorescein and added in
`soluble form to the peptide-bead library. Typically, a few beads
`were intensely stained and were visible to the naked eye and
`easily seen with a low-power dissecting microscope (bead
`diameter of 100-200 µm) against a background of colourless,
`nonreactive beads (Fig. 2). With the aid of tiny forceps coupled
`to a micromanipulator, the intensely staining beads could be
`removed for analysis (Fig. lb). We washed each bead with 8M
`guanidine hydrochloride to remove the acceptor complex, and
`then determined the peptide sequence contained on the bead
`by placing it on a glass filter which was inserted into a peptide
`microsequencer (model 477 A, Applied Biosystems). A library
`containing several million beads could be screened in 10-15
`Petri dishes in an afternoon. Afterwards, it could be washed
`with 8M guanidine hydrochloride and subsequently reused for
`screening new acceptors.
`Our sequencing studies established that the peptide present
`on any given single bead is sufficient for unambiguous sequence
`analysis as each bead sequenced contained 50-200 pmol of
`peptide (the lower limit of sensitivity of the instrument is in the
`range of 5 pmol). Furthermore, as measured by preview analysis 9
`for several dozen individual beads from this library, the coupling
`reactions in the split synthesis procedure were virtually complete
`as most individual beads displayed peptides that were over 99%
`pure (range 97-100%). At least three pentapeptide beads were
`sequenced daily using the microsequencer.
`We applied the process to produce a library of millions of
`pentapeptides and screen it against two well-studied acceptor
`molecules. Using the split synthesis approach with 19 reaction
`vessels, we synthesized pentapeptide libraries incorporating all
`the natural amino acids except for cysteine (for simplicity to
`eliminate disulphide crosslinking). The random incorporation
`of 19 amino acids into pentapeptides can produce a total of up
`to 2,476,099 (19 5
`) individual peptides of differing sequence with
`any one sequence represented on at least one solid-phase resin
`bead (of course, the number of beads of any given sequence
`will follow a Poisson distribution, and many multiples of this
`minimal number of beads are required to assure that the
`maximum theoretical number of possible peptide entities are
`approximated in the library). Libraries were readily synthesized
`in a few days.
`We studied a monoclonal antibody against /3-endorphin with
`high affinity (Ki= 17.5 nM) for the epitope sequence YGGFL
`(single-letter amino-acid code). A total of six reactive beads
`were retrieved from about 2 million beads screened from the
`pentapeptide library. One peptide ligand sequence retrieved,
`YGGFQ, had an affinity (Ki= 15.0 nM) nearly identical to the
`native epitope. Two of the other peptide ligands retrieved had
`Kis of less than 37 nM (Table 1). These affinities were more
`than 50-fold better than those obtained for the identical mono(cid:173)
`clonal antibody by Cwirla et al., who used a phage library
`method 7
`.
`
`TABLE 1 Affinity of anti-,8-endorphin ligands: comparison of the natural
`ligand and peptide ligands identified from a large pentapeptide library
`
`Native I igand
`Peptide ligand
`
`Sequence
`YGGFL
`YGGFQ
`YGGFA
`YGGFT
`YGGLS
`YGALQ
`YGGMQ
`
`K,(nM)
`17.5±3.2
`15.0±1.7
`32.9±2.0
`36.9± 7.7
`726±134
`1980±303
`8780±1500
`
`Affinity constants for the peptide ligands determined with competitive
`radioligand binding assays using tritiated YGGFL as the standard.
`
`NATURE · VOL 354 · 7 NOVEMBER 1991
`
`© 1991 Nature Publishing Group
`
`

`
`LETTERS TO NATURE
`
`TABLE 2 Amino-acid sequences of individual pentapeptide beads that
`interacted with streptavidin
`
`HPQFV
`HPQGP
`HQPAG
`
`LHPQF
`FHPQG
`GHPQN
`THPQN
`QHPQG
`IHPQG
`GHPQG
`
`WNHPM
`WTHPM
`VHPMA
`MHPMA (2)
`
`MYHPQ
`REHPQ
`IQHPQ
`GNHPQ
`TVHPQ
`IGHPQ
`WMHPQ
`GAHPQ
`PLHPQ
`AIHPQ
`AAHPQ
`TPHPQ (2)
`
`All sequences listed above were found on single beads except for TPHPQ
`and MHPMA for which two beads were obtained. The first three columns
`list the sequences found with HPQ located in the amino terminus, central
`region or carboxy terminus respectively. The fourth column lists the HPM
`sequences retrieved.
`
`biotin. These studies establish proof of concept for the process,
`and show that our synthetic libraries can be effectively recycled.
`Other investigators have attempted to develop peptide
`libraries for similar purposes. For example, Geysen and col(cid:173)
`leagues3 synthesized peptides of known amino-acid sequence
`on plastic pegs in 96-well plates. This approach permitted the
`synthesis of several thousand peptides 4
`• A related technique
`using complex instrumentation, photochemistry, and computer(cid:173)
`ized inventory control reported by Fodor et al. 4 permitted syn(cid:173)
`thesis of known arrays of at least 1,024 peptides on an individual
`microscope slide. Finally, Smith and colleagues pioneered the
`concept of using a recombinant bacteriophage incorporating
`random nucleic acids to produce phage displaying millions of
`random peptides5
`. This approach is innovative but faced with
`the inherent limitations of synthetic and selection biases of
`biological systems. Various investigators have used these tech(cid:173)
`niques to identify ligands for several monoclonal antibodies and
`streptavidin3-8
`. On the basis of comparisons with this published
`information, it is clear that our process is far simpler and more
`rapid than any of the alternative methods and unlike these
`methods has identified peptide ligands with affinities virtually
`identical to those of the native ligands. Additionally, our
`approach has far greater potential for applying the richness of
`well-established peptide chemistry to synthesize libraries incor(cid:173)
`porating D-amino acids or unnatural amino acids as well as
`specific secondary structures including cyclic peptides. All of
`this can be accomplished without need to keep records of the
`synthetic products as our interest is focused just on those pep(cid:173)
`tides which provide a strong interaction signal with the acceptor.
`On the basis of the combinatorial possibilities and chemical
`procedures available, it is feasible for us to develop far larger
`
`a
`Coupling Step
`
`(9 dipeptides)
`
`Ill
`
`(27 tripeptides)
`
`b
`
`Al
`
`A~
`
`Al
`
`AA~
`AG~
`AV~
`
`Al
`
`AAA~
`AAG~
`AAV~
`
`AGA~
`AGG~
`AGV~
`AVA~
`AVG~
`AVV~
`
`~
`
`Gi
`
`G~
`
`{randomize, split)
`
`Gi
`
`GA~
`GG~
`GV~
`
`(randomize, split)
`
`Gi
`
`GAA~
`GAG~
`GAV~
`
`GGA~
`GGG~
`GGV~
`GVA~
`GVG~
`GVV~
`
`vl
`
`v~
`
`vl
`
`VA~
`VG~
`vv~
`
`vl
`
`VAA~
`VAG~
`VAV~
`VGA~
`VGG~
`VGV~
`VVA~
`VVG~
`vvv~
`
`1. Stained bead removed and washed
`
`2. Peptide microsequenced from a single bead
`
`FIG. 1 Steps in peptide library synthesis and screening. a, Flow diagram of
`a simplified example of solid phase 'split synthesis' of tripeptides consisting
`of alanine (A), glycine (G) and valine (V) using standard solid-phase peptide
`synthesis methods with Fmoc or Boe chemistry. After each coupling step,
`the beads from each of the three reaction vessels are combined for
`randomization and then split again to the three vessels for the next coupling
`step. After three such steps the 27 possible peptide sequences (33
`) are
`all represented on separate beads. b, A single bead binding an acceptor
`molecule tagged with an enzyme is identified after being stained by enzymatic
`reaction on a dye substrate. The stained bead is physically removed from
`the colourless beads remaining in the library, washed free of the acceptor
`complex, and subjected to microsequencing. Once identified, the reactive
`peptide is then synthesized in larger quantities for confirmatory binding and
`biological studies.
`
`We also used the same pentapeptide bead library to find
`peptides binding streptavidin (chosen as a receptor-like target).
`Of 75 reactive beads retrieved, 28 were sequenced and a triplet
`consensus sequence of HPQ was found in either the amino or
`carboxy terminus or the central portion of 23 of the recovered
`pentapeptides (Table 2). The five other peptide ligands sequen(cid:173)
`ced contained the triplet sequence HPM in the pentapeptide.
`When beads containing LHPQF were synthesized, competitive
`binding studies established that the HPQ sequence was recogn(cid:173)
`ized at the same binding site on streptavidin as the native ligand
`
`FIG. 2 Low- and high-power photomicrographs of a peptide ligand library
`screening in which a reactive (dark) bead stained with the alkaline phos(cid:173)
`phatase reaction can be easily identified in a background of many thousands
`of nonreactive (colourless) beads.
`
`NATURE · VOL 354 · 7 NOVEMBER 1991
`
`83
`
`© 1991 Nature Publishing Group
`
`

`
`LETTERS TO NATURE
`
`libraries of longer or more diverse peptides should they be
`required for any given application.
`We have expanded the applications of our peptide library
`approach by modifying the synthesis procedure to incorporate
`cleavable linkers on each bead. After exposure to the cleaving
`agent, such beads can then release a portion of their peptides
`into solution for biological assay while still retaining sufficient
`peptides on the beads for subsequent structure determination.
`The one-bead, one-peptide concept and its applications dis(cid:173)
`cussed above demonstrate that this approach provides important
`new tools with which to search for specific ligands of potential
`diagnostic or therapeutic value. Such information should also
`enhance fundamental understanding of interactions between
`ligands and acceptor molecules.
`O
`
`Received 30 May; accepted 19 September 1991
`
`1. Hruby. V. J, Al-Obeid1. F. & Kazmierski. W Biochem J 268, 246-262 (1990)
`2. Hruby. V. J. & Sharma. S. 0. Curr Opin. 81otech 2, 599-605 (1991).
`3. Geysen. H. M .. Melven. R. H. & Barteling, S. J Proc. natn. Acad. Sci. USA. 81, 3998-4002 (1984).
`4. Fodor, S. P et al. Science 251, 767-773 (1991).
`5 Parmley. S. F. & Smith. G. P Gene 73, 305-318 (19881.
`6. Scott. J. K. & Smith. G. P. Science 249, 386-390 (1990)
`7. Cwirla, S. E., Peters, E. A .. Barrett. R W. & Dower. W. J. Proc. natn. Acad Sci. U.S.A. 87, 6378-6382
`(1990)
`8. Devlin. J. J.. Panganiban. L. C. & Devlin. P. E Science 249, 404-406 (1990).
`9. Niall, H. D .. Tregear. G. W. & Jacobs, J. in Chemistry and Biology of Peptides. (ed. Meienhofer, J.),
`695 (Ann Arbor. M1ch1gan. 1972).
`
`ACKNOWLEDGEMENTS. We thank F. Al-Obeidi. M. Ross and R. Hirschmann for valuable suggestions
`and E. Lander for comment on the manuscript. K.S.L. is a Special Fellow of the Leukemia Society of
`America. The peptide synthesis and screening process is referred to as the Selectide Process and
`is the subject of patent This work was supported by the National Institutes of Health. the Arizona
`Disease Control Research Commission and the Selectide Corporation.
`
`Generation and use of
`synthetic peptide
`combinatorial libraries
`for basic research and
`drug discovery
`Richard A. Houghten, Clemencia Pinilla,
`Sylvie E. Blondelle, Jon R. Appel, Colette T. Dooley
`& Julio H. Cuervo
`
`Torrey Pines Institute for Molecular Studies, 3550 General Atomics Court,
`San Diego, California 92121, USA
`
`EXISTING methods for the synthesis and screening of large num(cid:173)
`bers of peptides are limited by their inability to generate and
`screen the requisite number (millions) of individual peptides 1
`-4
`and/or their inability to generate unmodified free peptides in
`quantities able to interact in solution4--8. We have circumvented
`these limitations by developing synthetic peptide combinatorial
`libraries composed of mixtures of free peptides in quantities which
`can be used directly in virtually all existing assay systems. The
`screening of these heterogeneous libraries, along with an iterative
`selection and synthesis process, permits the systematic identi(cid:173)
`fication of optimal peptide ligands. Starting with a library com(cid:173)
`posed of more than 34 million hexa-peptides, we present here the
`precise identification of an antigenic determinant recognized by a
`monoclonal antibody as well as the straightforward development
`of new potent antimicrobial peptides.
`The initial synthetic peptide combinatorial library (SPCL)
`prepared and used in this work consisted of six-residue peptide
`sequences with acetylated N terminals and amidated C ter(cid:173)
`minals. The first two positions in each peptide were individually
`and specifically defined, whereas the last four positions consisted
`of equimolar mixtures of 18 of the 20 natural L-amino acids
`(for ease of synthesis, cysteine and tryptophan were omitted in
`
`84
`
`this initial library). Such libraries can be generally represented
`by the sequence Ac-0 10 2 XXXX-NH 2 (where Ac represents
`acetyl) (see legend to Fig. 1).
`Using a competitive enzyme-linked immunosorbent assay
`(ELISA), each of the 324 different peptide mixtures of the SPCL
`(Ac-0 10 2XXXX-NH 2 ) was assayed to determine its ability to
`inhibit the interaction of a monoclonal antibody with a larger
`13-residue peptide (Ac-YPYDVPDYASLRS-NH 2 ; single-letter
`amino-acid code). Of the 324 peptide mixtures examined (Fig.
`1), Ac-DVXXXX-NH 2 caused the greatest inhibition of anti(cid:173)
`body binding (Table 1). Twenty new peptide mixtures were then
`synthesized in which the third position of the peptide mixture
`Ac-DVXXXX-NH 2 was defined (Ac-DVOXXX-NH2, tryp(cid:173)
`tophan now included in the X positions). Each new peptide
`mixture contained 6,859 (193
`) individual peptides (137,180 in
`total). The most effective inhibiting peptide mixture was Ac(cid:173)
`DVPXXX-NH2 (50% inhibitory concentration, IC 50 = 41 µM;
`Table 1 b ). The above; iterative process, which reduces the num(cid:173)
`ber of peptide sequences by 20-fold each time it is repeated,
`was then carried out for the remaining three positions (Table
`1, c-e). It should be noted that on defining the fifth position
`Id), the IC 50 found for Ac(cid:173)
`(Ac-DVPDOX-NH 2, Table
`DVPDYX-NH2 (0.38 µM) was at least 3,500-fold lower than
`any of the other 19 peptide mixtures. Also, the peptide mixtures
`Ac-DVPDXX-NH2 and Ac-DVPXXX-NH 2 had IC 50 values
`lower than all of the peptide mixtures with the fifth position
`defined, with the exception of Ac-DVPDYX-NH2. This clearly
`
`TABLE 1
`
`Identification of the antigenic determinant recognized by mono-
`clonal antibody 19810
`
`Peptide mixture
`
`(a)
`Ac-DVXXXX-NH2
`Ac-DIXXXX-NH2
`Ac-DMXXXX-NH2
`Ac-DLXXXX-NH2
`
`(b)
`Ac-DVPXXX-NH2
`Ac-DVEXXX-NH2
`Ac-DVQXXX-NH2
`Ac-DVXXXX-NH2
`Ac-DVMXXX-NH2
`Ac-DVRXXX-NH2
`Ac-DVAXXX-NH 2
`Ac-DVCXXX-NH 2
`
`(c)
`Ac-DVPDXX-NH2
`Ac-DVPXXX-NH2
`Ac-DVPAXX-NH2
`
`(d)
`Ac-DVPDYX-NH2
`Ac-DVPDXX-NH2
`Ac-DVPDAX-NH2
`
`IC50
`(µM)
`
`250
`318
`752
`>1,400
`
`41
`146
`215
`250
`451
`906
`1,107
`>1,400
`
`4.4
`41
`>1,400
`
`0.38
`4.4
`>1,400
`
`Peptide
`
`(e)
`Ac-DVPDY A-NH2
`Ac-DVPDYS-NH2
`Ac-DVPDYX-NH2
`Ac-DVPDYC-NH2
`Ac-DVPDYV-NH2
`Ac-DVPDYT-NH2
`Ac-DVPDYG-NH2
`Ac-DVPDYE-NH2
`Ac-DVPDYl-NH2
`Ac-DVPDYM-NH2
`Ac-DVPDYQ-NH2
`Ac-DVPDYH-NH2
`Ac-DVPDYL-NH2
`Ac-DVPDYR-NH2
`Ac-DVPDYF-NH2
`Ac-DVPDYN-NH2
`Ac-DVPDYK-NH2
`Ac-DVPDYY-NH2
`Ac-DVPDYP-NH2
`Ac-DVPDYW-NH2
`Ac-DVPDYD-NH2
`
`IC50
`(µM)
`
`0.03
`0.27
`0.38
`0.90
`1.10
`1.50
`1.60
`4.06
`5.29
`7.70
`8.18
`8.99
`9.98
`10.90
`12.02
`15.56
`17.60
`22.48
`26.14
`32.14
`48.00
`
`The IC50s of the most effective inhibitory peptide mixtures obtained at
`each iterative step are illustrated for: a, peptide mixtures from the initial
`screening of the SPCL; b, the third position defined (AC-DVOXXX-NH2); c, the
`fourth position defined (Ac-DVPOXX-NH2 ); d, the fifth position defined (Ac(cid:173)
`DVPDOX-NH2); and e, the sixth position defined (Ac-DVPDYO-NH2). The IC50
`of the peptide mixture derived from the previous iterative step is in bold
`for comparison. Peptide mixtures were assayed by competitive ELISA (see
`Fig. 1). The concentration of each peptide mixture necessary to inhibit 50%
`of the antibody binding to the control peptide on the plate was obtained by
`serial dilutions of the peptide mixture. The IC50s were calculated using the
`software GRAPHPAD (ISi, San Diego). The four-step iterative screening and
`synthesis process takes approximately 4 weeks. This time frame will vary
`depending on the assay being used and the number of cases moved forward
`at each iterative step.
`
`NATURE · VOL 354 · 7 NOVEMBER 1991
`
`© 1991 Nature Publishing Group