Selective Ligands for the μ, δ, and κ Opioid Receptors Identified from a Single Mixture Based Tetrapeptide Positional Scanning Combinatorial Library*

A combinatorial library of 6,250,000 tetrapeptides in the mixture based positional scanning format was screened in binding assays for the three opioid receptors, μ, δ, and κ. Three different binding profiles were found. Individual peptides were synthesized representing all possible combinations of the active amino acids identified from the screening data. New, highly active peptides selective for each of the three receptors were chosen. This study demonstrates the power of mixture-based combinatorial libraries to identify distinctly different ligands for closely related receptors.

Synthetic combinatorial libraries have gained widespread acceptance for the rapid identification of new drug leads. Complete combinatorial libraries contain all possible arrangements of building blocks used in their synthesis. Peptides were the first compound class to be used in constructing such libraries (1)(2)(3). This laboratory first demonstrated the application of soluble mixture based combinatorial libraries through the identification of the peptide Ac-DVPDYA-NH 2 from a library of 52 million N-acetylated hexapeptides. The sequence corresponds to the antigenic determinant recognized by the antibody 17DO9 (1). Concurrently, a second combinatorial library containing 52 million non-acetylated hexapeptides was subsequently used to identify ligands for the opioid receptors, which were closely related to the natural ligands methionine-and leucine-enkephalin (YGGFM, YGGFL) (4). In later studies, six distinctly hexapeptide sequences that specifically bound to the opioid receptor were identified from the acetylated and nonacetylated libraries (5).
The combinatorial libraries described above are all mixturebased. Many questions have arisen concerning the use of mixture-based libraries. For example, how many different families of compounds for a particular target exist and can they be identified in a given library? What are the chances of missing the most active compound? What other compounds are present or could be identified from the data that are not immediately obvious? Are similar building blocks universally replaceable? Peptides, because of their ease of synthesis, represent the most convenient compound class for use in the study of mixturebased combinatorial libraries. By achieving an appreciation for the behavior of large mixtures through the use of peptide libraries, we believe these same principles can be applied to virtually all other compound classes (i.e. heterocycles and other small molecules).
The opioid receptors represent a convenient system to investigate the power of combinatorial libraries to identify distinctly different ligands for related receptors. There are three primary opioid receptors: mu (), delta (␦), and kappa (). All three receptors have recently been cloned, and they belong to the seven-transmembrane G-protein-coupled family of receptors and have approximately 60% amino acid sequence homology. Screening of the same combinatorial library in separate assays selective for each of the three receptors provides not only new ligands for these receptors but yields insights into the ability of combinatorial libraries to discriminate between closely related receptors.
A combinatorial library of 6,250,000 tetrapeptides, made using 50 different amino acids, was prepared in the positional scanning format (6,7). The use of positional scanning synthetic combinatorial libraries (PS-SCLs) 1 enables the most active amino acids at each position of a peptide or non-peptide to be determined directly from the initial screening data. This information can then be used to synthesize highly active individual compounds. A PS-SCL of tetrapeptide amides used in the current study consists of four separate sublibraries, each having a single defined position (O) and three mixture positions (X) as follows: O 1 XXX-NH 2 , XO 2 XX-NH 2 , XXO 3 X-NH 2 , and XXXO 4 -NH 2 . The defined positions of the mixtures making up each of the four separate sublibraries address a single position in the tetrapeptide. It should be noted that each of the four positional sublibraries are made up of the same 6,250,000 tetrapeptides. Screening the four sets of mixtures in the three separate opiate specific assays yielded information about the most important amino acids of each position in the tetrapeptide and led to the identification of three different series of active individual tetrapeptides selective for the , ␦, and receptors.

Preparation of the Tetrapeptide PS-SCL
The PS-SCL used in this study is composed of 6,250,000 tetrapeptides and contains four sublibraries, in which one of the four positions is defined with a single amino acid (O) and the three remaining positions are a mixture of 50 different L-, D-, and unnatural amino acids (X). The tetrapeptides were synthesized using the solid phase simultaneous multiple peptide synthesis approach (8)  polystyrene resin using t-butoxycarbonyl-protected amino acids. Mixture resins (X) were prepared using mixtures of t-butoxycarbonyl-protected amino acids at each coupling step. Each of the amino acids was present in a concentration that yielded close to equimolar coupling of each amino acid. The ratio of the concentrations of the individual amino acids used to yield this approximate equimolar coupling was pre-deter-mined using reverse phase-high pressure liquid chromatography to compare mixture profiles relative to standard mixtures synthesized using the divide, couple, and recombine method (1) as detailed in Ref. 9. Coupling completion was determined using Kaiser's ninhydrin test (10). Side chain deprotection and cleavage from the resin support were achieved using low hydrogen fluoride (11) and high hydrogen fluoride FIG. 1. Screening of the non-acetylated tetrapeptide PS-SCL for the ability to inhibit the binding of selective radiolabels to the , ␦, and receptors. The receptor was labeled using [ 3 H]DAMGO, and the ␦ receptor was labeled using [ 3 H]DSLET. Both assays were carried out using rat brain homogenates. The receptor was labeled using [ 3 H]U69,593 and guinea big brain homogenates. Each panel represents one of the four positional SCLs (i.e. position one SCL is O 1 XXX-NH 2 ). Each bar within a panel represents percent inhibition by a peptide mixture defined in the O position with one of 50 amino acids. Amino acids are listed in the footnote to Table II. While these graphs illustrate active peptide mixtures, the choice of amino acids for the synthesis of individual peptides was based on IC 50 values. (12) procedures. The 200 peptide mixtures were individually extracted with water, lyophilized, and resuspended in water at a final concentration of 10 mg/ml.

Synthesis of Individual Peptides
Peptides were synthesized using Fmoc (N-(9-fluorenyl)methoxycarbonyl)/t-butyl chemistry on a COMPAS 242 multiple peptide synthesizer (Spyder Instruments, San Diego). The peptides were synthesized on TentaGel resin in polypropylene mesh packets using inclusion volume synthesis. Reagent and wash solutions were removed by centrifugation (13,14).

Receptor Binding Assays
Receptor Assay-Membrane homogenates were prepared from rat brains. Brains were homogenized in 40 ml of Tris-HCl buffer, 50 mM, pH 7.4, 4°C (buffer A), and centrifuged (Beckman J2-HC, 35,300 ϫ g) for 10 min. The pellets were resuspended in fresh buffer and incubated at 37°C for 40 min. Following incubation, the suspensions were centrifuged as before, the resulting pellets resuspended in 100 volumes of Tris buffer, and the suspensions combined. Membrane suspensions were prepared and used on the same day. Protein content of the crude homogenates was determined by the method described by Bradford (15). Each assay tube contained 0.5 ml of membrane suspension, 3 nM [ 3 H-D-Ala 2 ,MePhe 4 ,Gly 5 -ol]enkephalin (DAMGO) and 0.08 mg/ml mixture, and 50 mM Tris-HCl in a final total volume of 0.65 ml. Assay tubes were incubated for 1 h at 25°C. Unlabeled DAMGO was used as a competitor to generate a standard curve and determine nonspecific binding. The reaction was terminated by filtration through GF-B filters on a Tomtec harvester (Orange, CT). The filters were subsequently washed with 6 ml of 50 mM Tris-HCl, pH 7.4 (buffer B), at 4°C. Bound radioactivity was counted on a Wallac Beta-plate Liquid Scintillation Counter (Piscataway, NJ).
␦ Receptor Assay-Rat brains were homogenized as described above using 40 ml of 50 mM Tris-HCl buffer, 100 M phenylmethylsulfonyl fluoride, 5 mM MgCl 2 , 100 nM Ac-rfwink-NH 2 (16), 1 mg/ml bovine serum albumin, pH 7.4, 4°C (buffer A). Homogenates were centrifuged and incubated as above. Each assay tube contained 0.5 ml of membrane suspension 3 nM tritiated [D-Ser 2 ,Leu 5 ,Thr 6 ]enkephalin (DSLET) in a total volume of 0.65 ml. Assay tubes were incubated for 2.5 h at 25°C. The assay was terminated, filtered, and counted as above. Unlabeled DSLET was used as a competitor to generate a standard curve and determine nonspecific binding.
Receptor Assay-Guinea pig cortices and cerebella were homogenized in 40 ml of buffer A. Homogenates were centrifuged and incubated as above. Each assay tube contained 0.5 ml of membrane suspension, 3 nM tritiated U69,593 in a total volume of 0.65 ml. Assay tubes were incubated for 2.5 h at 25°C. The assay was terminated, filtered, and counted as above. Unlabeled U50,488 was used as a competitor to generate a standard curve and determine nonspecific binding.

Adenylyl Cyclase Assay
The human SH-SY5Y neuroblastoma cell line was a generous gift from Dr. David K. Grandy (Vollum Institute for Advanced Biomedical Research, Portland, OR). The R1.G1 mouse thymoma cell line was obtained from ATCC (Rockville, MD) and has been shown to express the opioid receptor but not the or ␦ receptors (17). The cells were cultured in RPMI 1640 medium, buffered with 12.5 mM HEPES, pH 7.2, and containing 300 g/ml L-glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, 50 M 2-mercaptoethanol, 60 M 2-ethanolamine, and 10% iron-supplemented bovine calf serum in 5% CO 2 at 37°C. SH-SY5Y neuroblastoma cells were cultured in media containing 10 M retinoic acid for 6 days before harvesting in order to differentiate the cells as described previously (18). Cell membranes were prepared for use in the adenylyl cyclase assays as described previously (19). After the initial centrifugation at 200 ϫ g for 15 min at 4°C, the cells were resuspended in sucrose buffer (0.32 M sucrose, 40 mM HEPES, 2 mM EGTA, pH 7.6). Cells were centrifuged again at 200 ϫ g and then homogenized in sucrose buffer with five strokes of a Dounce homogenizer. Membranes were centrifuged at 22,000 ϫ g for 20 min at 4°C, followed by resuspension in sucrose buffer. The protein concentration was determined by the method of Bradford (15) using bovine serum albumin as standard. Cell membranes at a protein concentration of 1-4 mg/ml were stored at Ϫ80°C until use.
Membranes were incubated in a final volume of 100 l of 40 mM HEPES, containing 15 units of creatine phosphokinase, 20 mM phosphocreatine, 1 mM 1,10-o-phenanthroline, 60 M isobutylmethylxanthine, 50 M ATP, 50 M GTP, 3 mM MgCl 2 , and 100 mM NaCl. Agonists and antagonists were included at final concentrations as stated in the text. Naloxone, ICI 174,864, and nor-binaltorphimine were used to block , ␦, and opioid receptors, respectively. The reaction was initiated by the addition of 36 g of membrane protein. After 15 min at 30°C, the reaction was stopped by the addition of 40 l of cold 30% potassium bicarbonate, and then the membranes were centrifuged at 12,00 ϫ g for 4 min at 4°C in a microcentrifuge.
The amount of cyclic AMP present in 100 l of the supernatant,

RESULTS
The tetrapeptide PS-SCL was screened in each of the three separate opioid receptor binding assays described above. Each mixture contained 125,000 tetrapeptides (50 3 ). The mixtures were initially screened at a fixed concentration (0.08 mg/ml) at this mixture concentration each peptide was present at a concentration of .06 nM. Each of the 200 mixtures (50 for each of the four positions) making up the library was screened for its ability to inhibit binding of the tritiated ligand to brain homogenates.
Receptor-The library was screened again at a 10-fold lower concentration (0.008 mg/ml) (Fig. 1), since too many mixtures inhibited Ͼ90% of [ 3 H]DAMGO binding in the initial screening. IC 50 values were subsequently calculated for all mixtures that inhibited Ͼ90% of radioligand binding from each of the four positions (Fig. 2). The most active mixture from the series in which the first position was defined was YXXX-NH 2 (IC 50 ϭ 638 nM). This mixture was 3-fold more active than the second most active mixture (L-Nal)XXX-NH 2 (IC 50 ϭ 1946 nM) and 7-fold more active than the third most active mixture FXXX-NH 2 (IC 50 ϭ 4526 nM). There was a 25-fold difference in activity between the most active and least active mixture tested.
IC 50 values were calculated for seven mixtures in which the second position was defined (XOXX-NH 2 ). The most active mixtures found were all defined with D-amino acids. These mixtures exhibited a smaller range between the most and least active than for the first position. X(D-Nve)XX-NH 2 (IC 50 ϭ 690 nM) was the most active mixture found. The second, third, and fourth most active mixtures were within a 2-fold difference in activity: X(D-Nle)XX-NH 2 (IC 50 ϭ 1112 nM), yXXX-NH 2 (IC 50 ϭ 1378 nM), and rXXX-NH 2 (IC 50 ϭ 1507 nM) (where y indicates D-tyrosine and r indicates D-arginine).
IC 50 values were calculated for seven mixtures in which the third position was defined (XXOX-NH 2 ). Six of the mixtures exhibited activities below 2,000 nM, and all six contained Lamino acids in the defined position. The three most active mixtures found were XXFX-NH 2 (IC 50 ϭ 824 nM), XXGX-NH 2 (IC 50 ϭ 1119 nM), and XXWX-NH 2 (IC 50 ϭ 1227 nM).
IC 50 values were calculated for seven mixtures in which the fourth position was defined (XXXO-NH 2 ). The most active mixture found was XXX(L-Nal)-NH 2 (IC 50 ϭ 279 nM). The second most active mixture was 3-fold less active (XXXW-NH 2 ; IC 50 ϭ 850 nM) than the most active mixture, and the third most active mixture found (XXXF-NH 2 ; IC 50 ϭ 1545 nM) was over 5-fold less active than the most active mixture. The amino acids chosen to make 32 (1 ϫ 4 ϫ 4 ϫ 2) individual peptides are listed in Table I. K i values obtained for these peptides in the receptor binding assay are given in Table II. All 32 of the peptides were found to have high affinity for the receptor (K i values were Ͻ 15 nM). Three of the amino acids in the most active peptide Tyr-(D-Nve)Gly(L-Nal)-NH 2 (K i ϭ 0.4 nM) were the most active amino acids found for their particular position, whereas glycine in the third position was the second most active amino acid found for that position. The four D-amino acids chosen for the second position were found to be replaceable (i.e. peptides which differed only by the amino acid at this position had very similar activities). Whereas all four amino acids chosen in the third position yielded active peptides, those with small amino acid side chains (glycine and alanine) were more active than those with aromatic side chains. L-Naphthylalanine and L-tryptophan were found to be replaceable at the fourth position. The general motif of active peptides identified at the receptor was Tyr-(D-amino acid)(L-amino acid with small side chain) (L-aromatic)-NH 2 . Since all of the peptides synthesized were found to be active, additional -selective peptides are likely to be identified from this library. The selectivity ratios (K i of peptide at ␦ or receptor/K i value of peptide at the  receptor) of the 32 peptides are illustrated in Fig. 3A. All 32 peptides were found to be selective; none of the ratios were less than 1. The most -selective peptide found was YrAW-NH 2 (Rank 13 in Table II). All peptides with D-arginine at the second position exhibited excellent selectivity. These -selective peptides were generally more active at the ␦ receptor than at the receptor.
␦ Receptor-After an initial screening (0.08 mg/ml) in a ␦-selective receptor binding assay using [ 3 H]DSLET as radioligand, the tetrapeptide PS-SCL was screened again at a 10-fold lower concentration (Fig. 1). IC 50 values were calculated for those mixtures that inhibited Ͼ60% of [ 3 H]DSLET binding for positions 1 and 2 and Ͼ70% of [ 3 H]DSLET binding for positions 3 and 4 (Fig. 4). The most active of the 10 mixtures tested at position 1 included YXXX-NH 2 (IC 50 ϭ 2468 nM), WXXX-NH 2 (IC 50 ϭ 6250 nM), wXXX-NH 2 (IC 50 ϭ 7906 nM; w indicates D-tryptophan). The most active mixture was approximately 3-fold more active than the second most active mixture. Although the active mixtures identified were similar to those found in the receptor assay, they were less active, e.g. the mixture YXXX-NH 2 was 4-fold more active in the versus the ␦ receptor assay.
The 10 mixtures tested in which the third position was defined were found to be less active than those of the remaining three positions, and no mixture was found to have an IC 50 value lower than 5,000 nM. The four most active mixtures were XX(aAba)X-NH 2 (IC 50 ϭ 6474 nM), XXGX-NH 2 (IC 50 ϭ 7115 nM), XX(L-Cha)X-NH 2 (IC 50 ϭ 8462 nM), and XXMX-NH 2 (IC 50 ϭ 9085 nM). There was very little difference in affinity between the most active mixtures found at the third position; the 10 mixtures had IC 50 values between 6,000 and 10,000 nM.
IC 50 values were calculated for nine mixtures in which the fourth position was defined. Unlike the results found in the receptor assay, the two most active amino acids found at the fourth position were positively charged: XXXR-NH 2 (IC 50 ϭ 4529 nM) and XXXK-NH 2 (IC 50 ϭ 9026 nM). Mixtures ranked third and fourth at this position were aromatics, as was also found in the receptor assay [XXXW-NH 2 (IC 50 ϭ 6966 nM) and XXXF-NH 2 (IC 50 ϭ 7395 nM)]. The amino acid combinations chosen for the synthesis of 60 individual peptides are listed in Table III. Peptides that had K i values below 500 nM in the ␦-selective assay are shown in Table IV. Only three of the peptides were found to have activity under 10 nM. The ␦ selectivity of the peptides is shown in Fig. 3B. Twelve of the peptides had greater activity in the assay than in the ␦-selective assay (ratio of less than 1). This is not altogether surprising as many of the mixtures with defined amino acids chosen for the ␦ peptides were more active in the receptor assay. It is also not surprising that the most ␦-selective peptide found, Wy(aAba)R-NH 2 , (Rank 2 in Table III) contained L-arginine in the fourth position, since this amino acid was ranked first in the ␦ receptor screening and ranked 17 in the receptor screening. Many of the peptides tested were virtually inactive at the receptor at the highest concentration tested (10,000 nM). These data points are indicated by an asterisk in Fig. 3B.
Receptor-For the receptor, the library was screened at an initial concentration of 0.08 mg/ml in guinea pig brain homogenates using [ 3 H]U69,593 as radioligand (Fig. 1). IC 50 values were subsequently calculated for those mixtures which FIG. 3. Selectivity ratios of K i values for peptides found from screening the library in the receptor assay (A), the ␦ receptor assay (B), and the receptor assay (C). The selectivity ratio was determined using the following formula: K i value at receptor A/K i at receptor B, where A is either of the two alternate receptors, and B is the receptor for which the peptides were synthesized. The ratios are expressed in log scale; the higher the value the greater the selectivity of the compound for the particular receptor. Values less than 0 represent ratios of less than 1, which indicates that the peptides bind preferentially to another receptor. An asterisk (*) indicates that the peptide was inactive at the highest concentration tested, and K i values could not be determined.
Eleven mixtures were tested in which the second position was defined. As observed for the first position, D-amino acids were also favored. The most active mixture, X(D-Nal)XX-NH 2 (IC 50 ϭ 1545 nM), was 4-fold more active than the second most active mixture XfXX-NH 2 (IC 50 ϭ 4115 nM).
The most active mixtures found for the third position were XX(D-Nle)X-NH 2 (IC 50 ϭ 2346 nM) and XXlX-NH 2 (IC 50 ϭ 3019 nM). Eight of the nine mixtures tested at this position were found to have IC 50 values of less than 10,000 nM. The mixture ranked third at this position contained L-tryptophan, XXWX-NH 2 (IC 50 ϭ 5115 nM), whereas all other mixtures contained D-amino acids.
Seven mixtures were tested for the fourth position. As was found in the ␦ receptor assay, positively charged amino acids were found to have the greatest activity in the fourth position [XXXr-NH 2 (IC 50 ϭ 1526 nM) and XXXXk-NH 2 (IC 50 ϭ 2013 nM)]. The third most active mixture was XXX(D-Cha)-NH 2 (IC 50 ϭ 2929 nM). The amino acids chosen for inclusion in the synthesis of individual peptides are shown in Table V. Twenty four peptides were synthesized, and their K i values are given in Table VI. Fourteen of the 24 peptides had K i values below 50 nM, 11 of which were below 10 nM. The most active peptide was found to be ff(D-Nle)r-NH 2 (K i ϭ 1.2 nM). D-Phenylalanine and D-norleucine were replaceable at the first position. Surprisingly, D-phenylalanine and D-naphthylalanine were replaceable at the second position. D-Norleucine and D-isoleucine were replaceable at the third position; however, none of the peptides containing L-tryptophan in the third position had a K i value below 1000 nM. This suggests the existence of another family of ligands containing L-tryptophan in the third position. The active peptides (those ranked 1-16) were also highly selective for the receptor (Fig. 3C). The most selective peptides found had / and /␦ ratios of greater than 5,000.
Adenylyl Cyclase Assay-Opioid agonists inhibit adenylyl cyclase activity, resulting in reduced levels of cyclic AMP (cAMP) (21,22). An adenylyl cyclase assay using SH-SY5Y neuroblastoma or R1G1 thymoma cell line membranes was used to rapidly determine whether a peptide was an opioid agonist or antagonist. The opioid receptors expressed on the SH-SY5Y cell line after culturing in the presence of 10 M retinoic acid are approximately 85% of the type and 15% of the ␦ type (18). 2  receptor subtype. A reduction of cyclic AMP levels to less than 70% of the basal cAMP levels was regarded as being indicative of an opioid agonist effect, provided that the inhibition of cAMP was blocked by an opioid antagonist. Naloxone was used for receptors and nor-binaltorphimine was used for receptors. Their ability to inhibit the accumulation of cAMP was similar to that of DAMGO. Inhibition of cAMP was antagonized by naloxone but not the ␦-specific antagonist ICI 174,864, indicat-ing that the reduction in cAMP was mediated by receptors. The most active peptides found for the and assays have been tested for ability to inhibit cAMP accumulation (Table  VII). All peptides tested were found to be agonists at their respective receptor. The percent inhibition of cAMP was similar to the standard ligands, indicating that their efficacy at the receptor is similar to standard ligands. DISCUSSION The opiate receptor systems used in the current study have been described in our earlier work (4,7,16,23). In this laboratory studies involving combinatorial libraries for the identification of opioid ligands have focused on the receptor. These studies have a dual purpose as follows: first to find new ligands and expand our knowledge of the opioid receptors, and second to explore the use of combinatorial libraries made up of large mixtures. The ability to identify highly active individual li-   (23) and an all D-amino acid agonist (Ac-rfwink-NH 2 ) (16), were found. More recently we have used the receptor assay to demonstrate how a number of different sequences could be identified from such libraries. In the present study, the same PS-SCL containing 6,250,000 tetrapeptides was screened in receptor assays for the , ␦, and opioid receptors. The screening data obtained using this same library in three closely related receptors yielded three different binding profiles. The mixtures screened were found to have the greatest activity in the receptor relative to their activities at the ␦ and receptors. The general motif of active peptides identified for the receptor was Tyr-(D-amino acid)(L-amino acid)(aromatic side chain)-NH 2 . Tyrosine was the only amino acid chosen for the first position due to the excellent specificity found for this amino acid. The L-amino acids at the third position were either glycine or the aromatic amino acids phenylalanine or tryptophan. This peptide motif is similar to the truncation analogs of dermenkephalin and deltorphin described previously, YOFG where O ϭ D-methionine, D-alanine, or D-tyrosine (24). The peptide YmFG-NH 2 was previously identified from a tetrapeptide library using an iterative deconvolution process (25) but was not identified here because D-methionine was not included in this library. The most -selective peptides found in this study contain D-arginine at the second position and are similar to peptides reported previously (YrFK-NH 2 (DALDA) (26)). There are other reported tetrapeptides with high affinity for the receptor that were not identified in this report (e.g. YPWF-NH 2 (27) and WWPR-NH 2 (5)). YPWF-NH2 (27) would have been identified using a tetrapeptide library made up only of 20 L-amino acids or if more amino acids were chosen at each position. We are currently working on deconvolution strategies that would minimize the peptides required to be synthesized while maximizing the number of amino acids chosen at each position. In all three assays, the number of amino acids chosen for the synthesis of individual peptides was restricted since the number of combinations rises exponentially with the number of amino acids at each position (i.e. 81 tetrapeptides for 3 amino acids are chosen at each position, 256 tetrapeptides if 4 amino acids are chosen at each position, and 625 tetrapeptides if 5 amino acids are chosen at each position). This restriction clearly results in a limitation in the identification of additional active sequences. This can be seen in the current report in which active sequences for the receptor were identified when the same library was screened against the ␦ receptor. It should be noted that these sequences would have been identified from the screening data if a greater number of combinations had been synthesized. Furthermore, due to limited resources a subjective choice of amino acids is involved. In choosing the amino acids at each position, one endeavors to compromise between covering the greatest range in chemical diversity and avoiding the choice between two similar amino acids which are not replaceable. The majority of the combinations identified from the data for the ␦-selective assay were found to be more active in the receptor assay. A common motif for ␦-selective peptides was WOOR-NH 2 , as can be seen in Table VIII, which shows the effect on K i values at the and ␦ receptors of substitution analogs of the ␦-selective peptide Wy(aAba)R-NH 2 . There is no simple correlation between chemical similarities in amino acids and selectivity. Replacement of tryptophan at position 1 by tyrosine retains selectivity, but replacement of tyrosine by tryptophan at position 2 results in a substantial reduction of selectivity. Replacement of L-norvaline at position 3 by L-cyclohexylalanine results in reduction of selectivity, whereas a double replacement of tyrosine at position 1 and by L-cyclohexylalanine at position 3 completely reverses the selectivity, yielding a -selective peptide. This illustrates the dangers inherent in making broad statements on the replaceability of amino acids of similar chemical character.
The similarities between the peptides found to have activity and ␦ receptors appear to indicate a closer relationship between the two receptors, as opposed to the peptides found to be active at the receptor. DSLET is also known to bind to receptors, but we have included 100 nM Ac-RFWINK-NH 2 (16), a highly selective peptide, in the assay buffer in order to adequately block the radioligand from binding to receptors. Furthermore, the library was also screened using [ 3 H]Naltrindole as a ␦-selective radiolabel, and no significant library profiles differences were observed.
The sequences found from screening the tetrapeptide PS-SCL in an assay selective for the receptor are perhaps the most interesting of this study. They are unlike any peptides reported to have activity at receptors and would not have been identified by classical structure-activity studies. A general motif is not easily identified. The sequence clearly appears to favor D-amino acids at all four positions. Also, any peptide synthesized containing L-tryptophan at position 3 had poor activity. The first position accepts an aromatic side chain, D-phenylalanine, or an aliphatic side chain, D-norleucine, but whether this may be generalized to all similar amino acids is not known. Peptides with K i values below 10 nM were found with D-arginine in the fourth position. Sequences that differed only in the fourth position were always more active with Darginine than with D-cyclohexylalanine. Active peptides identified for the kappa receptor were highly selective. A study involving the synthesis of 1,000 -3,000 individual tetrapeptides based on the data presented in this study is underway.
In this study, highly active individual compounds and highly and -selective compounds were rapidly identified from a large mixture-based positional scanning combinatorial library. All of the most active peptides tested were found to be agonists.
It has yet to be determined if these peptides are capable of crossing the blood-brain barrier, but it is expected that the presence of D-amino acids in their sequences will prolong their biological half-lives. The peptides identified may prove useful for pain management, as compounds have come into focus in recent reports as having greater efficacy in analgesia for women (28). On the other hand, if these peptides do not cross the blood-brain barrier, they may be useful in attenuating the pain and/or progression of adjuvant arthritis (29). This study illustrates not only the power of the positional scanning concept for the rapid identification of new ligands but also how distinct ligands may be rapidly identified for closely related receptors.