Bioconlugete Chem. 1992, 3, 510-513
510
Evaluation of Peptide Libraries: An Iterative Strategy To Analyze the Reactivity of Peptide Mixtures with Antibodies James Blake' and Leonora Litzi-Davis Bristol-Myers Squibb Pharmaceutical Research Institute, 3005 First Avenue, Seattle, Washington 98121. Received June 10, 1992 Peptide libraries corresponding to a presumed mixture of 50 625tetrapeptides or 16 777 216 hexapeptides were each prepared in a single assembly by standard solid-phase peptide synthesis. By enzyme-linked immunosorbent assay, the tetrapeptide library was shown to inhibit the binding of an antiserum to FMRF amide with an FLRF capture antigen; the hexapeptide library was shown to inhibit the binding of a monoclonal antibody to a 28 amino acid peptide with the corresponding peptide capture antigen. An iterative strategy of variation was used to determine for each position in the tetra- or hexapeptides which amino acid contributed the most to activity. As a result we were able to logically select out of the tetrapeptide library the sequence FLRF and to select out of the hexapeptide library a sequence that differed from the apparent probable epitope but was twice as active. A single amino acid substitution in the logically derived sequence gave a peptide that was 35 times as active as the hexapeptide sequence in the original 28 amino acid peptide.
INTRODUCTION Recent advances in peptide chemistry have resulted in the development of methods for the preparation and evaluation of peptide libraries. Large numbers of randomly or specifically directed peptides have been synthesized (1-9) and assayed for activity, usually binding to an antibody to determine the epitope. The strategies employed may be divided into two categories. In the first, a mixture of peptides is exposed to a receptor and the resultant binding is used to separate the active peptides from the inactive peptides. The identity of the active peptides is then determined by the techniques of molecular biology (1-3) or peptide chemistry (4). In the second strategy, the synthesis of the peptide analogs is compartmentalized and this knowledge is used to determine the identity of the active peptides (5-9). We now report a strategy wherein a mixture of peptides is synthesized and evaluated, and subsequent iterative variations in the mixture allow deduction about the structure of the active peptide. EXPERIMENTAL PROCEDURES Peptide Synthesis. Peptides and peptide mixtures were synthesized by the solid-phase method (10) on an Applied Biosystems peptide synthesizer, Model 430A, using Boc amino acids. p-Methylbenzhydrylamine resin was the starting resin for all syntheses. All mixtures of Boc amino acids were coupled by reaction with DCC/ HOBT; individual amino acids were coupled as recommended by Applied Biosystems-DCC/HOBT or preformed symmetric anhydride. Boc amino acids used for the preparation of mixtures were purchased from Bachem. Side chain protecting groups were as follows: Ser, Bzl; Thr, Bzl; Glu, Bzl; Asp, Bzl; Tyr, BrZ; Lys, 2-C1Z; Arg, Tos; His, Bom. In the synthesis of peptides 2 and 3 the side chain of cysteine was protected by an ethylcarbamoyl group (11, 12).
* Author to whom correspondence should be addressed.
Abbreviationsused: Boc, tert-butyloxycarbony1;HRP,horse radish peroxidase; DCC,Nfl-dicyclohexylcarbodiimide;HOBt, 1-hydroxybenzotriazole;ELISA, enzyme-linked immunosorbent assay; BSA, bovine serum albumin. l
The Boc amino acids were divided into three groups which were a (L, A, V, T, F, Y, in molar ratios 1, 1,2, 2, 1,1.4), B (G, S, P, D, E in molar ratios of 1,1,1,1,l),and y (K, R, H, N, Q in molar ratios of 1,2,1,1,1). X designates a mixture of all of the amino acids in the indicated proportions. In the experiments with the antiserum to FMRF amide, the Boc amino acid mixtures were dissolved in DMF and aliquots were added to the amino acid cartridges. In the later experiments with the mAb to peptide 3 it was found to be easier to prepare stock mixtures of the solid derivatives which were ground together with a mortar and pestle to give a fine powder and then weighed into the amino acid cartridges. In the P2-y coupling mixtures indicated in the tables, the molar ratios of the B group amino acids were twice those of the y group, and the amino acids of the a group were absent. As an example for other mixtures shown in the tables, KRH3 indicates that the molar ratio of H to K or R was three times greater than it was in the primary X or y mixtures, and Q and N were absent. The peptides were cleaved from the resin and deprotected on their side chains by reaction with 90% HF/10% anisole for 1 h a t 0 "C. After evaporation of HF, the mixture was washed with ethyl acetate, and the peptides were dissolved in 50% acetic acid. The peptide solution was diluted with water and lyophilized. In the case of peptide mixtures the residue from lyophilization was used in immunoassay without further purification. In the case of individual peptides, purification was achieved by preparative HPLC on a Rainin Dynamax 300-A, C-8 column, 22 X 250 mm, by a gradient of acetonitrile in 0.1 % trifluoroacetic acid. The purified peptides were characterized by analytical HPLC, amino acid analysis, and mass spectrometry. Peptide 2 was treated with aqueous sodium hydroxide to remove the ethylcarbamoyl group as previously described (121, and the peptide was coupled through its side chain thiol group to BSA that had been previously reacted with the heterobifunctional reagent sulfosuccinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (Pierce). Amino acid analysis of the resulting conjugate indicated that it contained 15% by weight of peptide 2. 0 1992 American Chemical Society
Bloconlugete Chem., Vol. 3, No. 6, 1992 511
Strategy To Analyze PeptMe Llbrarles
ELISA Procedure. ELISA was performed on Dynatech Immulon 96-well plates. All operations were performed at room temperature. Rabbit antiserum to FMRF amide was obtained as a dry powder from Peninsula Laboratories (Cat. No. RAS8755) and mouse mAb to peptide 3 was obtained by standard procedures. The plates were coated overnight with bicarbonate solutions (pH 9.6) of peptide 2-BSA or peptide 3 at concentrations of 10or 0.2 pg/mL, respectively. The plates were washed with saline/Tween and then incubated with a skimmed milk blocking agent for 1-2 hand then washed again. The next step varied slightly in the two experiments. In the case of the antiserum to FMRF amide, 25-pL aliquots of the peptide mixtures at serially diluted concentrations were added to each of the wells, followed by the addition of 75 pL of antiserum powder dissolved (2.6 mg/mL) in 0.1 7% Triton X-100 (Sigma). For the mAb to peptide 3,75-pL aliquots of the peptide mixtures were preincubated with 25 pL of mAb in PBS, 0.08 pg/mL, for 15min, and then added to the wells. After 1-h incubation in the wells, the plate was washed and incubated with goat anti-rabbit y-globulin-HRP conjugate or goat anti-mouse y-globulin-HRP conjugate for 1h. Subsequent treatment with tetramethylbenzidine gave a signal (absorbance at 450 nm) that was measured on a microplate reader. The concentration of peptide or peptide mixture that gave a 50% reduction in A450nm is shown in the tables. RESULTS AND DISCUSSION As an initial model problem we chose to examine a commercially available antiserum to FMRF amide. The peptides FLRF amide (1) and GCGGGGFLRF amide (2) were synthesized and peptide 2 was conjugated to BSA through its cysteine residue. We substituted leucine for methionine because we had decided to exclude methionine from our peptide library and thus avoid the chemical problems it presents, and because of indications that leucine substitution would not markedly affect the cross reactivity of the peptide with anti-FMRF amide. The peptide 2-BSA conjugate was used as a capture antigen in ELISA and was shown to bind anti-FMRF amide. Peptide 1 was able to compete with the capture antigen for binding to anti-FMRF amide and a concentration of 0.5 pg of peptide l/mL gave a 50% reduction in signal obtained for anti-FMRF amide. A tetrapeptide mix XXXX was then synthesized where each X represented a mixture of 15 amino acids. The amino acids were A, L, V, F, Y (subgroup a),G, S, P, D, E (subgroup p), K, R, H, N, and Q (subgroup y). Methionine, as indicated above, plus cysteine and tryptophan were excluded to avoid the chemical problems that they present. We also believed that valine is a suitable general substitute for isoleucine, and threonine was excluded to simplify the problem. The fifteen aminoacids were used in equimolar mixtures with the exception that the proportions of V, Y, and R were increased because of their lower relative coupling efficiency (see Experimental Procedures). The presumed mixture of 154 (50 625) peptides was shown to competitively inhibit the binding of anti-FMRF amide to the capture antigen by the decreased signal obtained in ELISA (Table I). The problem was then to logically deduce the structure of a single highly active peptide in the mixture of 50 625 peptides. To do this we employed a bogus coin strategy-a strategy wherein one tries to learn which one of a group of coins has a different weight than the others using a minimum number of weighings or trials. This strategy involved dividing the objects (coins or amino acids) into
Table I. Ability of Synthetic Peptide Mixtures To Inhibit Binding of FMRF Amide Antiserum with FLRF-BSA in ELISA
!auencea
Bzr X
X X
x
Bzr X X
X X
X X
>2500 >2500
a a
1450
Bzr
Y
X
a
a a a a
>2500 35 >500
LAV3 FY FY FY FY F F
Bzr
a a
LAV3 a
Y Y Y KRR3
a
a
Y
FY F FY FY FY F
AL AL A AL AL L
KR KR KR
a
LAV3
a a
K KR R
x
44 41 >500 2.8 1.9 20 >70
1.4
F or Y A or L K or R ForY F L R F
0.5
*
All peptides have carboxyl terminus amide. Concentration (fig/ mL) of peptide or peptide mixture with antiserum that gave 50% of the signal obtained with antiserum alone. The amino acids or amino acid mixtures that were deduced to make the greatest contribution to binding activity.
three groups. The proportion of the first group (a)was decreased, the proportion of the second group (0) was increased, and the proportion of the third group (y)was unchanged. The effect on the activity-decreased, increased, or unchanged-indicated which of the groups was contributing the most to the activity. Four peptide mixtures were synthesized in which three positions contained 15 amino acids and one position contained none of the amino acids of subgroup a, approximately twice the proportion of the amino acids of the subgroup @, and approximately the same proportion of the amino acids of subgroup y. The effect of this change is shown in Table I. As an example, the peptide mixture that contained the @zy mix in position 1required a higher concentration for 50% inhibition than the XXXX peptide; hence this altered mixture was much less active than the XXXX peptide. Therefore, at position 1of the peptide the amino acids of the a subgroup contributed the most to the activity of the peptide mixture. The results indicate that for positions 1,2,3,and 4 in the peptide mixture the greatest contribution to activity was by the subgroups a, a,y, and a,respectively. Synthesis of a new tetrapeptide mixture containing only five amino acids at each position, aaya,resulted in a 40-fold increase in activity. The process was repeated and eachsubgroup was divided into three sections. As an example for the a group: F and Y were decreased (i.e,,omitted from the coupling mixture), V was increased; and A and L were unchanged. Four syntheses and subsequent immunoassay reduced to two the number of key amino acids in eachposition. Synthesis of a new mixture with two amino acids in each position gave a 12-fold increase in activity. Single amino acid substitutions led to the final deduction that the amino acids which contributed most to the activity of the original peptide mixture XXXX were F at position 1, L and position 2, R at position 3, and F at position 4. This corresponds to the sequence of the capture antigen and can be safely presumed to be one of the most active peptides in the original mixture of 50 625 peptides. As a second model problem we chose a mAb raised against a 28 amino acid peptide, acetyl-RTPALGPQAGIDTNEIAPLEPDAPPDAC amide (3). Since the epitope had not been previously characterized, this constituted a problem with an unknown answer. Pre-
Blake and Lkl-DavIs
Bloconlugete Chem., Vol. 3, No. 6, 1992
512
Table 11. Ability of Synthetic Peptide Mixtures To Inhibit Binding of the mAb to Peptide 3 with Peptide 3 in ELISA
peptide sequence
AAI X Bzr X X X X X
AA2
AA3
X X
X X ~927 X Bzr X X X X X X X
AAI
AAs
X X X X
X X X X X
@zr
@zr
X X
X
Ai% X
A4d
deducedb keyresidues
0.356 0.336 0.352 0.427 0.232 0.368 0.234 O.74Oc
X X X X X
@zr
Y Y
a
P Y
P
a Signal obtained at a peptide concentration of 7500 WglmL. Determinations were in quadruplicate and the standard deviation for all measurements was 0.01-0.02 absorbance units. See Table I. Signal obtained in the absence of peptide.
liminary results (data not shown) indicated that a hexapeptide mixture was able to block the binding of the mAb to peptide 3 as a capture antigen. Lesser though measurable blocking was obtained with a pentapeptide mixture and no blocking could be detected in a pentapeptide mixture containing only D-amino acids. The hexapeptide mixture contained 16 amino acids (including threonine) at each position and presumably corresponded to 166or 16 777 216 peptides. Inhibition of ELISA signal by XXXXXX and the corresponding /327 substitution mixtures showed (Table 11)that substitution of positions 1,2, and 5 did not affect activity (deduced y group), substitution at positions 4 and 6 increased activity (deduced /3 group),and substitution a t position 3 decreased activity (deduced CY group). A new mixture, yya/3y/3amide, gave increased activity compared to the activity of XXXXXX (Table 111). The iterative process was continued, and the results (Table 111) obtained for the KRH3 substitutions are illustrative. At position 1, KRH3 substitution gave essentially the same activity, and therefore K or R were
deduced to be the key residues; at position 2, KRH3 substitution gave reduced activity and therefore the missing residues, Q or N, were deduced to be the key residues; at position 5, KRH3 substitution gave increased activity and therefore H was deduced to be the key residue. The final deduced peptide was RQVGHD amide (4). A comparison to the sequence of peptide 3 points up the segment PQAGID (51, which is identical to peptide 4 at positions 2,4, and 6 and has nonconservative differences at positions 1,3, and 5. The activity of peptides 4,5, and several analogs is shown in Table IV. The results indicated that (a) the full length of peptide 4 is necessary for high activity; (b) peptide 4 is ca. twice as active as the "natural" sequence, peptide 5; (c) the nonconservative replacements at positions 3 or 5 increase activity compared to the "natural" sequence; and (d) the nonconservative replacement at position 1 decreases activity compared to the *natural" sequence. It is of interest that the substitutions a t positions 3 and 5 gave a peptide, PQVGHD amide, that is 35 times more active than the "natural" sequence peptide 0.
In conclusion, we believe that the method presented here can prove useful in logically searching through a mixture of thousands or even millions of peptides to pick out the peptides that have binding activity or any activity that correlates directly with binding. The principal advantage is that these results can be obtained by conventional solid-phase peptide synthesis on commercially available synthesizers that are designed for the syntheses of single peptides. It remains to be determined whether this strategy can be extended to peptide mixtures longer than six residues. The geometric decrease in the concentration of individual peptides for each extra residue and the limits of detection in most biological assays suggest that six residues may be at or near the maximum useful length.
Table 111. Ability of Synthetic Peptide Mixtures To Inhibit Binding of the mAb to Peptide 3 with Peptide 3 in ELISA
peptide sequence AA1
A& X
AA3
AA2
X
X
X
Y
a
KRHs
Y Y
Y
KRH,
a
Y
Y Y Y
LzAzVT
Y Y Y
KR
K
KR KR KR KR a See Table I.
P P P P
a
GQ NQ
N NQ NQ NQ
a a a
GSP,
VT VT VT V VT VT
GS GS GS GS G GS
AAG
AA5
X
X
Y Y Y Y Y
P P P P P P
13
KRH,
P
Y
GSP3 DE DE DE DE DE E
H H H H H H
Cl/P 6500 860 975 >4500 810 950 580 >4500 14 28 >75
11
7 >75
deduced key residues K or R N or Q V or T GorS H D or E R
Q
V G D
Table IV. Ability of Synthetic Peptides To Inhibit Binding of the mAb to Peptide 3 with Peptide 3 in ELISA peptide sequence AA1
AA2
AA3
R
Q Q Q Q Q Q Q Q
V
R E P R R P a See Table I.
V V V V A V A
A% G G G G G G G G
AAS
H H H H H H I I
AAS D D
c1/2a
1.2
>750 128
D D D D D
14 0.08 4.0 17 2.8
Strategy To Analyze Peptide Llbraries
ACKNOWLEDGMENT
The authors thank Becky Woodworth and Virginia Smith for their skilled technical assistance, Dale Yelton and Ursula Garrigues for the monoclonal antibody, and Ana Wieman for preparation of the manuscript. LITERATURE CITED
(1) Cwirla, S.E., Peters, E. A., Barrett, R. W., and Dower, W. J. (1990)Peptides on phage: A vast library of peptides for identifying ligands. Proc. Natl. Acad. Sci. USA 87, 63786382. (2) Devlin, J. J., Panganiban, L. C., and Devlin, P. E. (1990) Random peptide libraries: A source of specificprotein binding molecules. Science 249, 404-406. (3) Scott, J. K., and Smith, G. P. (1990)Searching for peptide ligands with an epitope library. Science 249,386-390. (4) Lam, K. S., Salmon, S. E., Hersh, E. M., Hruby, V. J., Kazmierski, W. M., and Knapp, R. J. (1991)A new type of synthetic peptide library for identifyingligand-bindingactivity. Nature 354,82-84. (5) Geysen, H. M., Meloen, R. H., and Barteling, S. J. (1984)Use of peptide synthesis to probe viral antigens for epitopes to a resolution of a single aminoacid. Proc. Natl. Acad. Sci. U.S.A. 81, 3998-4002. (6) Geysen, H. M., Rodda, S. J., and Mason, T. J. (1986)The delineation of peptides able to mimic assembled epitopes.
Bloconlugete Chem., Vol. 3, No. 6, lQQ2 519
Synthetic Peptides as Antigens. Ciba Foundations Symposium 119(R. Porter and J. Wheelan, Eds.) pp 131-149,Wiley, New York. (7) Houghten, R. A. (1985)General method for the rapid solidphase synthesis of large numbers of peptides: Specificity of antigen-antibody interaction a t the level of individual amino acids. Proc. Natl. Acad. Sci. U.S.A. 82,5131-5135. (8) Houghten, R. A., Pinilla, C., Blondelle, S. E., Appel, J. R., Dooley, C. T., and Cuervo, J. H. (1991)Generation and use of synthetic peptide combinatorial libraries for basic research and drug discovery. Nature 354,84-86. (9) Fodor, S.P. A., Read, J. L., Pirrung, M. C., Stryer, L., Lu, A. T.,and Solas,D. (1991)Light-directed, spatially addressable parallel chemical synthesis. Science 251,767-773. (10) Merrifield, R. B. (1963)Solid phase synthesis. I. The synthesis of a tetrapeptide. J.Am. Chem. SOC.85,2149-2154. (11) Guttman, St. (1966)Synthesis of glutathione and oxytocin using a new protecting group for the thiol function. Helv. Chim. Acta 49,83-96. (12)Blake, J., Woodworth, B. A., Litzi-Davis, L., and Cosand, W. L. (1992)Ethylcarbamoyl protection for cysteine in the preparation of peptide-conjugate immunogens. Int. J.Pept. Protein Res. in press.
2, 143924-37-2; 3, 143924-38-3; Registry No. 1, 80690-77-3; 4, 143924-39-4;5, 143924-40-7;FMRF amide, 64190-70-1; PQVGHD amide, 143924-41-8.
