Proc. Natl. Acad. Sci. USA
`Vol. 90, pp. 10922-10926, December 1993
`Chemistry
`
`Complex synthetic chemical libraries indexed with molecular tags
`(combinatorial chemistry/encoded libraries/peptides/antibody recognition)
`MICHAEL H. J. OHLMEYER*, ROBERT N. SWANSONt, LAWRENCE W. DILLARD*, JOHN C. READER*,
`GIGI ASOULINEt, RYUJI KOBAYASHIt, MICHAEL WIGLERt, AND W. CLARK STILL*
`*Department of Chemistry, Columbia University, New York, NY 10027; and tCold Spring Harbor Laboratory, Cold Spring Harbor, NY 11724
`Contributed by Michael Wigler, August 4, 1993
`
`Combinatorial methods of chemical synthesis
`ABSTRACT
`allow the creation of molecular libraries having immense
`diversity. The utility of such libraries is dependent upon
`identifing the structures of the molecules so prepared. We
`describe the construction of a peptide combinatorial library,
`having 117,649 different members, synthesized on beads and
`indexed with inert chemical tags. These tags are used as a
`binary code to record the reaction history of each bead. The
`code can be read directly from a single bead by electron capture
`capillary gas chromatography. We demonstrate the correct
`selection of members of the library on the basis of binding to
`a monoclonal antibody.
`
`With advances in the detailed knowledge of the molecular
`basis of biology and the development of techniques to isolate
`virtually any component ofa biological system, it has become
`increasingly feasible to search for biologically active com-
`pounds by screening for natural or synthetic ligands to
`biological molecules of known importance. Ligands discov-
`ered in this manner can be useful agents ifthey mimic or block
`natural ligands, or if they interfere with the naturally occur-
`ring interactions of the biological target. They can also
`provide a starting point for the engineering of molecules with
`more desirable properties. Since the chance of finding valu-
`able ligands will increase with the number of compounds
`screened, the success of the search will be best with massive
`libraries of compounds. Such libraries can have many
`sources. Plant and animal extracts, for example, provide a
`rich source of molecular diversity, though finding and iden-
`tifying biologically active molecules at parts-per-million-to-
`billion levels can be problematic.
`One of the most promising approaches to the synthesis of
`large collections of diverse molecules is known as combina-
`torial chemistry (1, 2), in which vast libraries of molecules
`having different chemical compositions are synthesized si-
`multaneously. Combinatorial methods entail a series of
`chemical steps with multiple choices of chemical reagents for
`each step. The complexity, or number of members in a
`combinatorial library, is given by the product of the number
`of reagent choices for each step of the synthesis and can
`therefore be quite large. The challenge in using combinatorial
`libraries is the characterization ofmembers ofthe library with
`particular desired properties. Several solutions to this prob-
`lem have been described in the literature. Members of the
`library can be synthesized in spatially segregated arrays, but
`to date this has resulted only in relatively small libraries (3).
`Alternatively, in the multivalent synthesis method, a library
`of moderate complexity can be produced by pooling multiple
`choices of reagents during synthesis (4, 5). Once a given pool
`is shown to have an interesting property, it is resynthesized
`iteratively with lower and lower complexity until a single
`compound having the desired property is identified. The
`
`The publication costs of this article were defrayed in part by page charge
`payment. This article must therefore be hereby marked "advertisement"
`in accordance with 18 U.S.C. §1734 solely to indicate this fact.
`
`multivalent method is not practical for construction of mas-
`sive libraries because the concentration of any individual
`member of the library decreases with complexity. Moreover,
`cumbersome resyntheses are required to isolate individual
`compounds. Another approach, the split synthesis method,
`involves combinatorial synthesis on solid particles such as
`Merrifield synthesis beads (t, §, and refs. 6-10). Through a
`protocol of separating and mixing beads during the synthesis,
`each bead in the final library has a product from a single,
`specific reaction sequence chemically bound to it and that
`product is likely to differ from that bound to another bead.
`After selecting a particular bead having some desirable
`property, the identity of the attached compound is deter-
`mined by analytical chemistry. Thus the split synthesis
`method can be employed only to synthesize compounds that
`can be readily elucidated by microscale sequencing, such as
`oligonucleotides and certain oligopeptides.
`Yet another solution, the cosynthesis method, attempts to
`solve the structure elucidation problem by cosynthesizing a
`sequenceable tag that encodes the series of steps and re-
`agents used in the synthesis of each library element. The tag
`and the corresponding library element are associated by a
`chemical bond. Once a library element is selected, the
`procedure used to synthesize it can be read by sequencing the
`tag. Oligonucleotide and oligopeptide tags have both been
`proposed (11, 12). The main problem with the cosynthesis
`method is that the tagging structures are chemically labile and
`incompatible with many of the reagents normally associated
`with synthetic organic chemistry. Additional limitations fol-
`low from the constraint of compatible protecting groups
`which allow the alternating cosynthesis of tag and library
`element. Moreover, the oligonucleotide or peptide tags may
`themselves associate selectively with biological receptors
`and confuse the assay.
`We have devised an alternative method that is not plagued
`by these problems and that allows the construction of large
`chemically diverse libraries. As in the original split synthesis
`method, we synthesize library elements on microsphere
`beads (see also ref. 12). During each step of synthesis,
`however, we attach to the beads tagging molecules that
`encode both the step number and the chemical reagent used
`in that step. The array of tags used forms a binary record of
`the synthetic steps for each bead. Our tagging molecules are
`not sequentially connected, hence no cosynthesis is required.
`With only 20 such tags, we can uniquely encode 220 =
`1,048,576 different syntheses. In the following paragraphs,
`we describe the use of this method to prepare and study a
`
`Abbreviations: DMF, dimethylformamide; EC, electron capture;
`mAb, monoclonal antibody.
`:Furka, A., Sebestyen, M., Asgedom, M. & Dibo, G., 14th Inter-
`national Congress on Biochemistry, July 10-15, 1988, Prague,
`Czechoslovakia, Vol. 5, p. 47 (abstr.).
`§Furka, A., Sebestyen, M., Asgedom, M. & Dibo, G., 10th Inter-
`national Symposium on Medical Chemistry, August 15-19, 1988,
`Budapest, Hungary, p. 288 (abstr.).
`
`10922
`
`

`
`Chemistry: Ohlmeyer et al.
`chemically encoded combinatorial library of 117,649 pep-
`tides.
`
`MATERIALS AND METHODS
`Typical Tag Linker Preparation. To a solution of 8-bromo-
`1-octanol (0.91 g, 4.35 mmol) and 2,4,6-trichlorophenol (1.03
`g, 5.22 mmol) in dimethylformamide (DMF) (5 ml) was added
`cesium carbonate (1.70 g, 5.22 mmol), resulting in the evo-
`lution of gas and the precipitation of a white solid. The
`reaction mixture was stirred at 80°C for 2 hr. The mixture was
`diluted with toluene (50 ml), washed with 0.5 M NaOH (twice
`with 50 ml), 1 M HCl (twice with 50 ml), and water (50 ml),
`and the organic phase was dried (MgSO4). Removal of the
`solvent by evaporation gave 1.24 g (87% yield) of tag alcohol
`as a colorless oil.
`The above tag alcohol (0.81 g, 2.5 mmol) was added to a 2
`M solution of phosgene in toluene (15 ml) and stirred at room
`temperature for 1 hr. The excess phosgene and the toluene
`were removed by evaporation and the resulting crude chlo-
`roformate was dissolved in CH2Cl2 (5 ml) and pyridine (0.61
`ml, 7.5 mmol). tert-Butyl-4-(hydroxymethyl)-3-nitroben-
`zoate (13) (0.5 g, 1.98 mmol) was added and the reaction
`mixture was stirred at room temperature for 3 hr. The
`solution was diluted with ethyl acetate (75 ml) and poured
`into a separatory funnel. After washing with 1 M HCl (three
`times with 35 ml), saturated NaHCO3 (twice with 35 ml), and
`saturated NaCl (35 ml), the organic phase was dried (MgSO4).
`The solvent was removed by evaporation and the residue was
`purified by chromatography on silica gel (5-7.5% ethyl
`acetate in petroleum ether), affording 0.95 g (79% yield) of the
`tag-linker tert-butyl ester as a clear oil.
`Trifluoroacetic acid (3 ml) was added to a solution of the
`tag-linker tert-butyl ester (0.95 g, 1.57 mmol) in CH2Cl2 (30
`ml) to deprotect the linker acid and the solution was stirred
`at room temperature for 7 hr. The mixture was then evapo-
`rated to dryness and the residue was redissolved in CH2Cl2
`(30 ml). The solution was washed with saturated NaCl (20 ml)
`and the organic phase was dried (MgSO4). Removal of the
`solvent by evaporation gave 0.75 g (87% yield) of the tag-
`linker acid (6B) as a pale yellow solid (see Fig.
`1 and
`Generation of a Large Encoded Library below for nomen-
`clature).
`Typical Encoded Library Synthesis Step. Na-Fmoc-
`Glu(tBu)-Glu(tBu)-Asp(tBu)-Leu-Gly4-NH-Merrifield resin
`was suspended in DMF (20 ml) and shaken for 2 min. After
`filtering, 1:1 diethylamine/DMF (40 ml) was added to remove
`the Fmoc (fluoren-9-ylmethoxycarbonyl) protecting groups
`and the resin was shaken for 1 hr. The resin was separated by
`filtration and washed with DMF (twice with 20 ml, 2 min
`each), 2:1 (vol/vol) dioxane/water (twice with 20 ml, 5 min
`each), DMF (three times with 20 ml, 2 min each), and CH2Cl2
`(three times with 20 ml, 2 min each) then dried under reduced
`pressure at 25°C.
`Portions (150 mg) of the resin were placed in seven
`Merrifield vessels and suspended in CH2Cl2 (5 ml). The
`appropriate linker-tag acids were activated as their acyl
`carbonates as follows (for the first coupling): 10A (6.6 mg,
`0.0098 mmol) was dissolved in anhydrous ether (2 ml) and
`pyridine (10 IlI) was added. Isobutyl chloroformate (1.3 ul,
`0.0096 mmol) was added as a solution in anhydrous ether (0.1
`ml). The resulting mixture was stirred at 25°C for 1 hr, during
`which time a fine white precipitate formed. The stirring was
`stopped and the precipitate was allowed to settle for 30 min.
`Solutions of the acylcarbonates of 9A and 8A were prepared
`in the same way. Aliquots (0.25 ml) of the supernatant
`solution of activated linker-tags were mixed to give the
`appropriate 3-bit binary tag codes as described in the text,
`and the appropriate coding mixtures of tags were added to
`each of the seven synthesis vessels. The vessels were shaken
`
`Proc. Natl. Acad. Sci. USA 90 (1993)
`
`10923
`
`in the dark for 12 hr, and then each was washed with CH2Cl2
`(four times with 10 ml, 2 min each). A solution of the
`symmetrical anhydride (ref. 14, pp. 80-83) of an Nc-Fmoc
`amino acid in CH2Cl2 (3 eq in 10 ml) was then added to the
`correspondingly coded batch of resin and shaken for 20 min.
`Five percent N,N-diisopropylethylamine in CH2Cl2 (1 ml)
`was added and the mixture was shaken until the resin gave a
`negative Kaiser test. The resin batches were filtered, com-
`bined, and then washed with CH2Cl2 (four times with 20 ml,
`2 min each), isopropyl alcohol (twice with 20 ml, 2 min each),
`and CH2Cl2 (four times with 20 ml, 2 min each). The next
`cycle of labeling/coupling was initiated by Fmoc deprotec-
`tion as described above.
`After Fmoc deprotection of the residues in the last position
`of the peptide, the side chain functionality was deprotected
`by suspending the resin in CH2Cl2 (10 ml), adding thioanisole
`(2 ml), ethanedithiol (0.5 ml), and trifluoroacetic acid (10 ml)
`then shaking for 1 hr at 25°C. The resin was then washed with
`C12CH2 (six times with 20 ml, 2 min each) and dried.
`Electron Capture (EC) Gas Chromatography (GC) Reading
`of Code. A single selected synthesis bead was placed in a
`Pyrex capillary tube and washed with DMF (five times with
`10 ,ul). The bead was then suspended in DMF (1 AL) and the
`capillary was sealed. The suspended bead was irradiated at
`366 nm for 3 hr to release the tag alcohols, and the capillary
`tube was subsequently placed in a sand bath at 90°C for 2 hr.
`The tube was opened and bis(trimethylsilyl)acetamide (-0.1
`,ul) was added to trimethylsilylate the tag alcohols. After
`centrifuging for 2 min, the tag solution above the bead (1 ,u)
`was injected directly into an EC detection, capillary gas
`chromatograph for analysis.
`Antibody-Afmity Methods. The anti-c-MYC monoclonal
`antibody (mAb) 9E10 has been described (15, 16). To test
`beads for binding to 9E10, beads were incubated in TBST (20
`mM Tris HCl, pH 7.5/500 mM NaCl/0.05% Tween-20) con-
`taining 1% bovine serum albumin (BSA) to block nonspecific
`protein binding sites. The beads were then centrifuged,
`resuspended in a 1:200 dilution of 9E10 ascites fluid in
`TBST/1% BSA, and incubated overnight at 4°C. Beads were
`subsequently washed three times in TBST and incubated for
`90 min at room temperature in alkaline phosphatase-coupled
`goat antibodies against mouse IgG (Bio-Rad), diluted 1:3000
`in TBST/1% BSA. After the beads had been washed twice in
`TBST and once in phosphatase buffer (100 mM Tris HCl, pH
`9.5/100 mM NaCl/5 mM MgCl2), they were incubated 1 hr at
`room temperature in phosphatase buffer containing 0.01 part
`each of AP color reagents A and B (Bio-Rad). To stop the
`reaction, the beads were washed twice in 20 mM NaEDTA,
`pH 7.4. Solution phase affinities between 9E10 and various
`peptides were determined by a modification of the compet-
`itive ELISA previously described (17), using a recombinant
`fusion protein containing at its amino terminus the antigenic
`peptide EQKLISEEDL, kindly provided by A. Polverino
`(Cold Spring Harbor Laboratory). The concentration of each
`peptide necessary to inhibit mAb binding by 50%o (IC50) was
`determined in at least three independent assays.
`
`RESULTS AND DISCUSSION
`A Binary Encoding Scheme Using Chemically Inert Tags. A
`simple binary code can be used to describe an organic
`synthesis; it is best illustrated by example. Imagine carrying
`out a combinatorial synthesis using any of seven different
`reagents in each of N steps. Such a combinatorial synthesis
`would yield 7N different final products. Let us designate the
`various reagents which can be used in any step as binary 001
`(reagent 1), 010 (reagent 2), 011 (reagent 3),
`111 (reagent
`7). We can now write a binary synthesis code describing any
`complete N-step synthesis using 3 x N binary digits. For
`example, if we used reagent 3 in the first step, the binary
`
`

`
`10924
`
`Chemistry: Ohlmeyer et al.
`numerical description would be "011." If we next used
`reagent 1 in the second step, the description would be "001
`011." And if we finally used reagent 6 in the third step, we
`would obtain "110 001 011." This 9-bit binary synthesis code
`describes the synthesis and can be read from right to left in
`3-bit blocks to decode the reagents used in each step of the
`synthesis. More bits per step could be used to encode more
`reagent designations. To represent such a synthesis code
`chemically, we use a set of distinguishable, sensitively de-
`tectable molecules as tags and the presence ofa particular tag
`to represent a binary 1 for the corresponding bit. Using a set
`of nine tagging molecules, T9-T1, where T9 represents the
`leftmost binary bit and Ti represents the rightmost bit, the tag
`mixture containing only T9, T8, T4, T2, and Ti would
`represent the 110 001 011 synthesis code.
`Various methods can be used to analyze minute quantities
`of organic tagging molecules. Capillary GC is a convenient
`analytical technique for separating and identifying tags.
`When tagging molecules with unusually electrophoric func-
`tionality (e.g., fluoro- or chlorocarbons) are used, the con-
`junction of EC detection (18) with GC separation enables us
`to analyze tags from a single 50-,m microsynthesis bead
`without significant interference from contaminants. We em-
`ploy a photocleavable linkage (see below) between the tag
`and the bead so that the tag may be selectively liberated from
`the bead.
`Generation of a Large Encoded Library. We prepared a set
`of 18 GC-separable tags with linkers that allowed them to be
`attached to, and then detached from, the synthesis beads.
`The linker/tagging molecules we used are summarized in Fig.
`1. By using different lengths of the hydrocarbon chain (N =
`1-10) and three different aromatic electrophores (Ar), we
`were able to prepare more than 20 tagging molecules which
`were well-separated by capillary GC and selectively detected
`by EC at levels <1 pmol. We designate these tags as NAr,
`where N is the length of the hydrocarbon chain and Ar is the
`identity of the electrophore as shown in Fig. 1. Thus tag 2B
`has the structure with N = 2 and Ar = B. The linker segment
`of these molecules incorporates a carboxylic acid for attach-
`ment to the synthesis beads and a photochemically labile
`ortho-nitrobenzylcarbonate for subsequent detachment of
`the tags.
`To relate our tagging molecules to the binary bits of the
`synthesis code, we arrange them by their GC elution order.
`Thus the tag called Ti is retained the longest on our GC
`column and designates the rightmost bit of the binary syn-
`thesis code number. The next-longest retained tag we call T2,
`and so on. Using a 0.2 mm x 20 m methylsilicone capillary
`GC column with different combinations of tag chain lengths
`and electrophoric halobenzenes, we assembled 18 well-
`resolved tagging molecules (Tl-T18) whose chemical com-
`positions were, respectively, 1OA, 9A, 8A, 7A, 6A, SA, 4A,
`3A, 6B, 2A, SB, 1A, 4B, 3B, 2B, 1B, 2C, and 1C.
`To test our encoding method in the context ofscreening for
`binding to a biological receptor, we synthesized an encoded
`combinatorial library of 117,649 peptides. In the standard
`single-letter codes for amino acids, this library had the
`sequence H2N-XXXXXXEEDLGGGG-Bead, where the
`variable residue X was D, E, I, K, L, Q, or S. This library
`included H2N-EQKLISEEDLGGGG-Bead among its se-
`
`Proc. Natl. Acad. Sci. USA 90 (1993)
`
`quences, and EQKLISEEDL is known to be bound by 9E10,
`a mAb directed against the human c-MYC protein (15, 16).
`The four glycines served as a molecular spacer to separate the
`epitope from the bead. Three binary bits were sufficient to
`represent the seven alternative reagents for each step. We
`assigned the following 3-bit, binary codes as follows: 001 =
`S (serine), 010 = I (isoleucine), 011 = K (lysine), 100 = L
`(leucine), 101 = Q (glutamine), 110 = E (glutamate), and 111
`= D (aspartate).
`We synthesized our library by first preparing the constant
`segment ofthe library (H2N-EEDLGGGG-Bead) on 1.05 g of
`50- to 80-gtm Merrifield polystyrene synthesis beads, using
`standard solid-phase methods based on tBu side-chain pro-
`tection and Fmoc main-chain protection (ref. 14, pp. 80-83).
`After removing the N-terminal Fmoc protecting group, we
`divided the beads into seven 150-mg portions and processed
`each as described in Materials and Methods, attaching first
`the tags and then the corresponding amino acids to each
`portion. The tags were attached via their carboxylic acids to
`the synthesis beads by activating the linker carboxyl groups
`as mixed carbonic anhydrides and then adding an amount of
`activated tag corresponding to -1% of the free amino groups
`on the beads. In the process of this coupling, approximately
`0.5% of the growing peptide chains were terminated for each
`tag added. The remaining free amino groups were then
`coupled in the usual way with the corresponding protected
`amino acids as their symmetrical anhydrides. After washing,
`the seven fractions were combined. After Fmoc deprotec-
`tion, the beads were again divided into seven portions and
`processed as before except that in place of tags representing
`the first step (Ti, T2, T3), tags representing the second step
`(T4, T5, T6) were used. At this point the library had 72
`members. By repeating this procedure four more times using
`tags T7-T18 analogously, the entire encoded library of 76 =
`117,649 different peptides was prepared.
`Given any bead, the attached tags could be detached by
`UV irradiation and unambiguously decoded by ECGC (e.g.,
`see Fig. 3). To verify that the codes corresponded to the
`actual peptide sequence present on the beads, two beads
`were picked at random, the tags present on each were
`released and read by ECGC, and then the peptide sequence
`present on each was determined by microsequencing. We
`observed complete concordance between the synthesis code
`and the peptide sequence in each case.
`Screening the Library with a mAb. To pick out those
`members of our library that bound to the anti-c-MYC mAb,
`we mixed the bead library with the antibody and stained those
`beads that bound antibody by using alkaline phosphatase-
`coupled secondary antibodies (Fig. 2). When viewed under a
`low-power microscope, dark-staining beads could be easily
`distinguished from the vast majority of nearly colorless
`unstained beads and were individually picked by using a
`manual micropipetter. From two different antibody stainings
`of 30-mg samples of the peptide library, we picked out 40
`dark-colored beads for decoding. We found that the stained
`beads had reaction histories leading to the presence of either
`the MYC epitope (EQKLISEEDL) or sequences that dif-
`fered by one or two substitutions among the three N-terminal
`residues. In most cases, these sequences were found multiple
`
`HOOC
`
`\
`
`Ar
`
`Ci
`
`Cl
`
`Ci
`
`H
`
`CI
`
`CI
`
`H
`
`H
`
`C
`
`N02
`
`Unker
`
`FIG. 1.
`
`Ar-= QCi
`CI
`CI
`
`Electrophoric Tag
`Molecular tags which create a binary synthesis code.
`
`A
`
`CI
`CB H
`B
`
`

`
`Chemistry: Ohlmeyer et al.
`
`Proc. Natl. Acad. Sci. USA 90 (1993)
`
`10925
`
`Peptide library beads stained with mAb 9E10. The beads shown are approximately 50-80 ,um in diameter. A stained bead is visible
`FIG. 2.
`in the middle, 4 cm from the right edge of the photograph.
`times. The synthesis codes of the stained beads that were
`solution, the affimities of the antibody for free peptides
`picked are listed in Table 1.
`derived from five reactive (dark-staining) bead sequences
`were measured as IC5o values by using a competitive ELISA.
`Reading the binary synthesis code ofa single bead by using
`ECGC was a straightforward process, as the reader may see
`From entries 1-5 of Table 1, it can be seen that all of the
`from an actual gas chromatogram of a stained EQKLIS bead
`peptides derived from stained bead sequences are bound with
`taken from the antibody-binding experiment (Fig. 3). Peaks
`high affinity. A control peptide derived from a nonstaining
`other than those of the tags (T18-T1) come from impurities
`bead sequence (DKISSLEEDL) showed no detectable bind-
`ing (IC5o > 500 ,uM).
`in the solvent or silylating agent. Such impurity peaks occur
`at the same retention times (±0.05 min) in every chromato-
`Since all of the peptides we identified on stained beads
`gram, and hence the synthesis codes of single beads can be
`contained the sequence LIS, we decided to measure the
`read unambiguously in nearly all instances.
`affinity of the mAb 9E10 for MYC epitope-derived peptides
`To test whether the peptides identified in our solid-phase
`that contained substitutions at these positions. As shown in
`binding experiment could also be bound to mAb 9E10 in free
`entries 13-15 of Table 1, substitution of isoleucine for leu-
`Solution and solid-phase binding of peptide library elements to mAb 9E10 -
`Table 1.
`Synthesis code*
`Entry
`Stain*
`Sequencet
`110 101 011 100 010 001
`EQKLIS
`Yes
`100 101 011 100 010 001
`LQKLIS
`Yes
`101 101 011 100 010 001
`Yes
`QQKLIS
`110 101 101 100 010 001
`EQQLIS
`Yes
`110 110011 100 010 001
`EEKLIS
`Yes
`111 110 011 100 010 001
`DEKLIS
`Yes
`100 110 011 100 010 001
`LEKLIS
`Yes
`11 100 010 001
`111 101
`DQKLIS
`Yes
`111 110 011 100 010 001
`Yes
`QEKLIS
`110 111011 100 010 001
`10
`EDKLIS
`Yes
`100 101 101 100 010 001
`11
`LQQLIS
`Yes
`111 011 010 001 001 100
`12
`No
`DKISSL
`Bead not found
`EQKIIS
`13
`Bead not found
`EQKLLS
`14
`Bead not found
`15
`EQKLID
`*Synthesis code of selected library beads.
`tAmino acid residues indicated in boldface are those which differ from those found in the c-MYC
`protein epitope against which mAb 9E10 was raised (15, 16).
`*Staining with mAb 9E10.
`§Peptides with the indicated sequence plus EEDL were synthesized and their ICso values for binding
`were determined, except where indicated by ND.
`
`IC50,§ PM
`1.31 ± 0.05
`1.36 ± 0.51
`1.15 ± 0.03
`23.3 ± 2.3
`4.67 ± 0.85
`ND
`ND
`ND
`ND
`ND
`ND
`>500
`163.4 ± 20.6
`40.1 ± 5.0
`166.2 ± 38.8
`
`123456789
`
`

`
`10926
`
`Chemistry: Ohlmeyer et al.
`
`Proc. Natl. Acad. Sci. USA 90 (1993)
`
`Til
`
`TI0
`
`TI5
`
`T13
`
`T18
`
`TI7
`
`T5
`
`Tl
`
`T9
`
`Injection
`
`Binarynhes Code:
`
`1
`
`_
`
`1
`
`0
`
`.
`
`Step 6
`E
`
`js1KL
`
`1
`
`%
`
`v
`
`0
`Step 5
`a
`
`,'%
`
`, %
`
`v
`
`0
`1 01 11 0
`Step4 Step 3
`K
`L
`
`0 1
`Step 2
`l
`
`0
`
`O
`
`0
`
`1
`
`Step 1
`S
`
`FIG. 3. Gas chromatogram of tags from EQKLISEEDLGGGG-Bead. The synthesis code from one stained synthesis bead was read by
`releasing and analyzing tags as described in the text.
`
`Cancer Institute (to M.W.). M.W. is an American Cancer Society
`Professor.
`
`cine, leucine for isoleucine, or aspartic acid for serine led to
`more weakly binding peptides having IC50 values of 40-166
`,uM. We were surprised to find that the conservative substi-
`tutions ofleucine and isoleucine produced such large changes
`in IC50. Since none of these substitutions were found in a
`stained bead sequence (though such sequences should have
`been present in the library), and since the stained bead
`sequence EQQKLISEEDL (IC50 = 23 ,uM) was found, we
`have an indication of the minimum affinity required for
`detection of antibody binding to bead-supported peptides
`under the conditions described here.
`The results above establish that a chemically inert, multiple
`tag labeling scheme can be useful for the practical generation
`of large encoded combinatorial libraries. While we demon-
`strate the method here by generating an encoded library of
`peptides, the method can be applied to other library types.
`The most exciting applications will likely involve the con-
`struction of similarly encoded small molecule organic librar-
`ies whose chemical elements cannot be sequenced the way
`peptides can. Studies of the detection limits of the molecular
`tags we have employed suggest that we can create encoded
`libraries having as many as 109 different members on approx-
`imately 1 cm3 of microsphere synthesis beads. Such com-
`plexity should be attainable by using 10- to 20-,um beads,
`which can carry tags at the readily detectable level of 0.1
`pmol. With the standard 50- to 80-,um beads used in our
`current work, libraries having more than 106 members per
`cm3 are readily available. While the assays we used here to
`select beads were conducted with library members chemi-
`cally bound to the synthesis beads, one can readily imagine
`schemes in which cleavable linkers between the beads and
`the library members allow off-bead solution-phase assays.
`Access to such encoded combinatorial libraries should pro-
`vide a substantial benefit to those searching for new organic
`compounds having desirable properties.
`
`The authors thank Richard Axel, Paul Bartlett, and David Beach
`for helpful discussions and Gilbert Stork for careful readings of the
`manuscript. This work was supported by National Science Founda-
`tion Grant CHE92 08254 (to W.C.S.) and grants from the National
`
`1.
`
`2.
`
`3.
`
`4.
`
`5.
`
`6.
`
`8.
`
`9.
`
`Jung, G. & Beck-Sickinger, A. G. (1992) Angew. Chem. Int.
`Ed. Engl. 31, 367-383.
`Pavia, M. R., Sawyer, T. K. & Moos, W. H. (1993) Bioorg.
`Med. Chem. Lett. 3, 387-3%.
`Fodor, S. P., Read, J. L., Pirrung, M. C., Stryer, L., Lu, A. T.
`& Solas, D. (1991) Science 251, 767-773.
`Houghten, R. A., Pinilla, C., Blondelle, S. E., Appel, J. R.,
`Dooley, C. T. & Cuervo, J. H. (1991) Nature (London) 354,
`84-86.
`Houghten, R. A. & Dooley, C. T. (1993) Bioorg. Med. Chem.
`Lett. 3, 405-412.
`Furka, A., Sebestyen, M., Asgedom, M. & Dibo, G. (1991) Int.
`J. Pept. Protein Res. 37, 487-493.
`7. Lam, K. S., Salmon, S. E., Hersh, E. M., Hruby, V. J., Kaz-
`mierski, W. M. & Knapp, R. J. (1991) Nature (London) 354,
`82-84.
`Zuckermann, R. N., Kerr, J. M., Siani, M. A. & Banville,
`S. C. (1992) Int. J. Pept. Protein Res. 40, 498-507.
`Sebestyen, F., Dibo, G., Kovacs, A. & Furka, A. (1993)
`Bioorg. Med. Chem. Lett. 3, 413-418.
`10. Lam, K. S., Hruby, V. J., Lebl, M., Knapp, R. J., Kazmierski,
`W. M., Hersh, E. M. & Salmon, S. E. (1993) Bioorg. Med.
`Chem. Lett. 3, 419-424.
`Brenner, S. & Lerner, R. A. (1992) Proc. Natl. Acad. Sci. USA
`89, 5381-5383.
`Kerr, J. M., Banville, S. C. & Zuckermann, R. N. (1993) J.
`Am. Chem. Soc. 115, 2529-2531.
`Barany, G. & Albericio, F. (1985) J. Am. Chem. Soc. 107,
`4936-4942.
`Stewart, J. M. & Young, J. D. (1984) Solid Phase Peptide
`Synthesis (Pierce Chem. Co., Rockford, IL), 2nd Ed., pp.
`80-83.
`Evan, E. I., Lewis, G. K., Ramsay, G. & Bishop, M. (1985)
`Mol. Cell Biol. 5, 3610-3616.
`Munro, S. & Pelham, H. R. B. (1987) Cell 48, 899-907.
`Harlow, E. & Lane, D. (1988) Antibodies: A Laboratory
`Manual (Cold Spring Harbor Lab. Press, Plainview, NY), pp.
`570-573.
`Grimsrud, E. P. (1992) in Detectors for Capillary Chromatog-
`raphy, eds. Hill, H. H. & McMinn, D. G. (Wiley, New York),
`pp. 83-107.
`
`11.
`
`12.
`
`13.
`
`14.
`
`15.
`
`16.
`17.
`
`18.