Growth hormone variants

FIELD OF THE INVENTION

The invention is directed to methods for identifying the active domains and amino acid residues in polypeptides. It is also directed to hormone variants.

BACKGROUND OF THE INVENTION

Polypeptides, i.e., peptides and proteins, comprise a wide variety of biological molecules each having a specific amino acid sequence, structure and function. Most polypeptides interact with specific substances to carry out the function of the polypeptide. Thus, enzymes, such as subtilisin, amylase, tissue plasminogen activator, etc., interact with and hydrolyze specific substrates at particular cleavage sites whereas proteinaceous hormones such as human growth hormone, insulin and the like interact with specific receptors to regulate growth and metabolism. In other cases, the interaction is between the polypeptide and a substance which is not the primary target of the polypeptide such as an immunogenic receptor. Many polypeptides are pluripotential in that they contain discrete regions which interact with different ligands or receptors each of which produces a discrete biological effect. For example, human growth hormone (hGH) is diabetogenic and lypogenic in adults and induces long bone growth in children.

Efforts have been made to modify the primary functional properties of naturally occurring polypeptides by modifying amino acid sequence. One approach has been to substitute one or more amino acids in the amino acid sequence of a polypeptide with a different amino acid. Thus, protein engineering by in vitro mutagenesis and expression of cloned genes reportedly has been applied to improve thermal or oxidative stability of various proteins. Villafranca, J. E., et al. (1983)

Science

222, 782-788; Perry, L. J., et al. (1984)

Science

226, 555-557; Estell, D. A., et al. (1985)

J. Biol. Chem.

260, 6518-6521; Rosenberg, S., et al. (1984)

Nature

(

London

) 312, 77-80; Courtney, M., et al. (1985)

Nature

(

London

) 313, 149-157. In addition, such methods have reportedly been used to generate enzymes with altered substrate specificities. Estell, D. A., et al. (1986)

Science

223, 655-663; Craik, C. S., et al. (1985)

Science

228, 291-297; Fersht, A. R., et al. (1985)

Nature

(

London

) 314, 235-238; Winther, J. R., et al. (1985)

Carlsberg Res. Commun.

50, 273-284; Wells, J. A., et al. (1987)

Proc. Natl. Acad. Sci.

84, 1219-1223. The determination of which amino acid residue should be modified has been based primarily on the crystal structure of the polypeptide, the effect of chemical modifications on the function of the polypeptide and/or the interaction of the polypeptide with various substances to ascertain the mode of action of the polypeptide. In some cases, an amino acid substitution has been deduced based on the differences in specific amino acid residues of related polypeptides, e.g. difference in the amino acid sequence in substrate binding regions of subtilisins having different substrate specificities. Wells, J. A., et al. (1987)

Proc. Natl. Acad. Sci. USA

84, 5767. In other cases, the amino acid sequence of a known active region of a molecule has reportedly been modified to change that sequence to that of a known active region of a second molecule. Wharton, R. P., et al. (1985)

Nature

316, 601-605, and Wharton, R. P., et al. (1984)

Cell

38, 361-369 (substitution of recognition helix of phage repressor with recognition helix of different repressor); Jones, P. T., et al. (1986)

Nature

321, 522-525 (substitution of variable region of a mouse antibody for corresponding region of human myeloma protein). While this approach may provide some predictability with regard to the properties modified by such substitutions, it is not a methodical procedure which would confirm that all regions and residues determinative of a particular property are identified. At best, empirical estimates of the energetics for the strengths of the molecular contacts of substituted residues may be ascertained. In this manner, the strengths of hydrogen bonds (Fersht, A. R., et al. (1985)

Nature

314, 235; Bryan, P., et al. (1986)

Proc. Natl. Acad. Sci. USA

83, 3743; Wells, J. A., et al. (1986)

Philos. Trans. R. Soc. London A.

317, 415), electrostatic interactions (Wells, J. A., et al. (1987)

Proc. Natl. Acad. Sci. USA

84, 1219; Cronin, C. N., et al. (1987)

J.Am. Chem. Soc.

109, 2222), and hydrophobic and steric effects (Estell, D. A., et al. (1986)

Science

233, 659; Chen, J. T., et al. (1987)

Biochemistry

26, 4093) have been estimated for specific modified residues. These and other reports (Laskowski, M., et al. (1987)

Cold Spring Harbor Symp. Quant. Biol.

52, 545; Wells, J. A., et al. (1987)

Proc. Natl. Acad. Sci. USA

84, 5167; Jones, P. T., et al. (1986)

Nature

321, 522; Wharton, R. P., et al. (1985)

Nature

316, 601) have concluded that mutagenesis of known contact residues causes large effects on binding whereas mutagenesis of non-contact residues has a relatively minor effect.

A second reported approach to understand the relationship between amino acid sequence and primary function employs in vivo homologous recombination between related genes to produce hybrid DNA sequences encoding hybrid polypeptides. Such hybrid polypeptides have reportedly been obtained by the homologous recombination of trp B and trp A from

E.coli

and

Salmonella typhimurium

(Schneider, W. P., et al. (1981)

Proc. Natl. Acad. Sci USA

78, 2169-2173); alpha 1 and alpha 2 leukocyte interferons (Weber, H. and Weissmann, C. (1983)

Nuc. Acids Res.

11, 5661); the outer membrane pore proteins ompC and phoE from

E.coli

K-12 (Thommassen, J., et al. (1985)

EMBO

4, 1583-1587); and thermophilic alpha-amylases from

Bacillus stearothermophilus

and

Bacillus lichiniformis

(Gray, G. L., et al. (1986)

J. Bacterial.

166, 635-643). Although such methods may be capable of providing useful information relating to amino acid sequence and function as well as useful hybrid polypeptides, as reported in the case of the hybrid alpha amylases, it is difficult to utilize such methods to systematically study a given polypeptide to determine the precise regions and amino acid residues in the polypeptide that are active with one of the target substances for that particular molecule. This is because the site of crossover recombination, which defines the DNA and amino acid sequence of the hybrid, is determined primarily by the DNA sequence of the genes of interest and the recombination mechanism of the host cell. Such methods do not provide for the predetermined and methodical sequential replacement of relatively small segments of DNA encoding one polypeptide with a corresponding segment from a second gene except in those fortuitous circumstances when crossover occurs near the 5′ or 3′ end of the gene.

The interaction of proteinaceous hormones with their receptors has reportedly been studied by several techniques. One technique uses hormone peptide fragments to map the location of the receptor binding sites on the hormone. The other technique uses competition between neutralizing monoclonal antibodies and peptide fragments to locate the receptor binding site by epitope mapping. Exemplary of these techniques is the work reported on human growth hormone (hGH).

Human growth hormone (hGH) participates in much of the regulation of normal human growth and development. This 22,000 dalton pituitary hormone exhibits a multitude of biological effects including linear growth (somatogenesis), lactation, activation of macrophages, insulin-like effects and diabetagenic effects among others. See Chawla, R. K. (1983)

Ann. Rev. Med.

34, 519; Edwards, C. K., et al. (1988)

Science

239, 769; Thorner, M. O., et al. (1988)

J. Clin. Invest.

81, 745. Growth hormone deficiency in children leads to dwarfism which has been successfully treated for more than a decade by exogenous administration of hGH. There is also interest in the antigenicity of hGH in order to distinguish among genetic and post-translationally modified forms of hGH (Lewis, U. J. (1984)

Ann. Rev. Physiol.

46, 33) to characterize any immunological response to hGH when it is administered clinically, and to quantify circulating levels of the hormone.

hGH is a member of a family of homologous hormones that include placental lactogens, prolactins, and other genetic and species variants of growth hormone. Nichol, C. S., et al. (1986)

Endocrine Reviews

7, 169. hGH is unusual among these in-that it exhibits broad species specificity and binds monomerically to either the cloned somatogenic (Leung, D. W., et al. (1987)

Nature

330, 537) or prolactin receptor (Boutin, J. M., et al. (1988)

Cell

53, 69). The cloned gene for hGH has been expressed in a secreted form in

Eschericha coli

(Chang, C. N., et al. (1987)

Gene

55, 189) and its DNA and amino acid sequence has been reported (Goeddel, et al. (1979)

Nature

281, 544; Gray, et al. (1985)

Gene

39, 247). The three-dimensional structure of hGH is not available. However, the three-dimensional folding pattern for porcine growth hormone (pGH) has been reported at moderate resolution and refinement (Abdel-Meguid, S. S., et al. (1987)

Proc. Natl. Acad. Sci. USA

84, 6434).

Peptide fragments from hGH have been used in attempts to map the location of the receptor binding site in hGH. Li, C. H. (1982)

Mol. Cell. Biochem.

46, 31; Mills, J. B., et al. (1980)

Endocrinology

17, 391. In another report, a fragment consisting of residues 96-133 was isolated after proteolysis of bovine growth hormone. This fragment was reported to bind to a growth hormone receptor. Yamasakin, et al. (1970)

Biochemistry

9, 1107. However, when a larger peptide containing residues 1-133 was produced by recombinant methodology, no detectable binding activity was observed. Krivi, G. G., et al., International Symposium on Growth Hormone; Basic and Clinical Aspects, Abstract I-18, Final Program, sponsored by Serono Symposia, USA, Jun. 14-18, 1987. These results are clearly irreconcilable and demonstrate the potential unreliability of using peptide fragments to map receptor binding sites on a proteinaceous hormone, especially for those where the binding site is composed of two or more discontinuous and/or conformationally dependent epitopes.

The use of neutralizing monoclonal antibodies to locate the receptor binding sites by epitope mapping has similar limitations. For example, a monoclonal antibody was reportedly used in a receptor binding assay to compete with the hGH receptor for a peptide consisting of residues 98-128 of hGH. Even though the peptide 98-128 of the hGH hormone only binds to the neutralizing monoclonal antibody, it was proposed that this region contains the receptor binding site based on these competitive studies. Retegin, L. A., et al. (1982)

Endocrinology

111, 668.

Similar approaches have been used in attempts to identify antigenic sites on the hGH hormone. Epitope mapping of twenty-seven different monoclonal antibodies to hGH by competitive binding reportedly resolved only four different antigenic sites on the hormone. Surowy, T. K., et al. (1984)

Mol. Immunol.

21, 345; Aston, R., et al. (1985)

Pharmac. Ther.

27, 403. This strategy, however, did not locate the epitopes on the amino acid sequence of the hormone.

Another approach to defining antigenic sites has been to test the binding of antibodies to short linear peptides over the protein of interest. Geysen, H. M., et al. (1984)

Proc. Natl. Acad. Sci. USA

81, 3998; Geysen, H. M. (1985)

Immunol. Today

6, 364. However, this approach suffers from the same limitations of using linear peptide fragments to locate receptor binding sites. To be useful, the linear sequence must be capable of adopting the conformation found in the antigen for the antibody to recognize it. Furthermore, based upon the known size of antibody epitopes from X-ray crystallography (Sheriff, S., et al. (1987)

Proc. Natl. Acad. Sci USA

84, 8075; Amit, A. G., et al. (1986)

Science

23, 747) it has been estimated that virtually all antibody combining sites must be, in part, discontinuous (Barlow, D. J., et al. (1986)

Nature

322, 747) and as a result linear fragments may not adequately mimic such structure.

Peptide fragments from hGH have also been studied by non-covalently combining such fragments. Thus, several investigators have reported the analysis of the combination of relatively large fragments of human growth hormone comprising either the natural amino acid sequence or chemically modified peptides thereof. Burstein, S., et al. (1979)

J. of Endo. Met.

48, 964 (amino terminal fragment hGH-(1-134) combined with carboxyl-terminal fragment hGH-(141-191)); Li, C. H., et al. (1982)

Mol. Cell. Biochem.

46 31; Mills, J. B., et al. (1980)

Endocrinology

107, 391 (subtilisin-cleaved two-chain form of hGH).

Similarly, the chemically modified fragment hGH-(1-134) and a chemically modified carboxy-terminal fragment from human chorionic somatomammotropin (also called placental lactogen), (hCS-(141-191)), have been non-covalently combined, as have the chemically modified fragments hCS-(1-133) and hGH-(141-191). U.S. Pat. No. 4,189,426. These investigators reported incorrectly that the determinants for binding to the hepatic growth hormone receptor are in the first 134 amino-terminal residues of growth hormone (Burstein, et al. (1978)

Proc. Natl, Acad. Sci. USA

75, 5391-5394). Clearly, such techniques are subject to erroneous results. Moreover, by utilizing two large fragments this technique is only potentially able to localize the function to one or the other of the two fragments used in such combinations without identification of the specific residues or regions actively involved in a particular interaction. A review of some of the above techniques and experiments on hGH has been published. Nichol, C. S., et al. (1986)

Endocrine Rev.

7, 169-203.

An alternative approach has been reported wherein a 7 residue peptide fragment from the “deletion peptide” of hGH (hGH-32-46) was modified to contain amino acid residues from analogous segments of growth hormone from other mammalian species. The effect, if any, of such substitutions, however, was not reported. See U.S. Pat. No. 4,699,897. Nonetheless, the shortcomings of the use of short peptide fragments are apparent since the linear sequence of such fragments must be capable of adopting the conformation found in the intact growth hormone so that it may be recognized in an in vitro or in vivo assay.

A number of naturally occurring mutants of hGH have been identified. These include hGH-V (Seeberg, P. H. (1982)

DNA

1, 239; U.S. Pat. Nos. 4,446,235, 4,670,393 and 4,665,180) and 20K hGH containing a deletion of residues 32-46 of hGH (Kostyo, J. L., et al. (1987)

Biochemica et Biophysica Acta

925, 314; Lewis, U. J., et al. (1978)

J. Biol. Chem.

253, 2679).

One investigator has reported the substitution of cysteine at position 165 in hGH with alanine to disrupt the disulfide bond which normally exists between Cys-53 and Cys-165. Tokunaga, T., et al. (1985)

Eur. J. Biochem.

153, 445. This single substitution produced a mutant that apparently retained the tertiary structure of hGH and was recognized by receptors for hGH.

Another investigator has reported the in vitro synthesis of hGH on a solid resin support. The first report by this investigator disclosed an incorrect 188 amino acid sequence for hGH. Li, C. H., et al. (1966)

J. Am. Chem. Soc.

88, 2050; and U.S. Pat. No. 3,853,832. A second report disclosed a 190 amino acid sequence. U.S. Pat. No. 3,853,833. This latter sequence is also incorrect. In particular, hGH has an additional glutamine after position 68, a glutamic acid rather than glutamine at position 73, an aspartic acid rather than asparagine at position 106 and an asparagine rather than aspartic acid at position 108.

In addition to the foregoizng, hybrid interferons have been reported which have altered binding to a particular monoclonal antibody. Camble, r. et. al. “Properties of Interferon-α2 Analogues Produced from Synthetic Genes in Peptides: Structure and Function,”

Proceedings of the Ninth American Peptide Symposium

(1985) eds. Deber et. al., Pierce Chemical Co., Chicago, Ill., pp.375-384. As disclosed therein, amino acid residues 101-114 from α-1 interferon or residues 98-114 from γ-interferon were substituted into α-2 interferon. α-2 interferon binds NK-2 monoclonal antibody whereas α-1 interferon does not. This particular region in α-2 interferon apparently was chosen because 7 of the 27 amino acid differences between α-1 and α-2 interferon were located in this region. The hybrids so obtained reportedly had substantially reduced activity with NK-2 monoclonal antibody. When tested for antiviral activity, such hybrids demonstrated antiviral activity on par with the activity of wild type α-2 interferon. Substitutions of smaller sections within these regions were also reported. Sequential substitution of clusters of 3 to 7 alanine residues was also proposed. However, only one analogue [Ala-30,32,33] IFN-α2 is disclosed.

Alanine substitutions within a small peptide fragment of hen egg-white lysozyme and the effect of such substitutions on the stimulation of 2A11 or 3A9 cells have also been reported. Allen, P. M., et. al. (1987)

Nature

327,713-715.

Others have reported that binding properties can be engineered by replacement of entire units of secondary structure including antigen binding loops (Jones, P. T., et al. (1986)

Nature

321, 522-525) or DNA recognition helices (Wharton, R. P., et al. (1985)

Nature

316, 601-605).

The references discussed above are provided solely for their disclosure prior to the filing date of the present application, and nothing herein is to be construed as an admission that the inventors are not entitled to antedate such disclosure by virtue of prior invention or priority based on earlier filed applications.

Given the state of the art as exemplified by the above references, it is apparent that a need exists for a useful method for the systematic analysis of polypeptides so as to ascertain the relationship between structure and function. Accordingly, it is an object herein to provide such methods to identify the active domains within the polypeptide which contribute to the functional activity of the polypeptide.

It is a further object herein to provide methods for determining the active amino acid residues which determine functional activity.

A further object of the invention is to provide methods for systematically identifying the biologically active domains in a polypeptide.

Further, it is an object herein to provide hormone variants having desirable biological, biochemical and immunogenic properties which are different as compared to the same properties of the hormone from which such variants are derived.

Still further it is an object herein to provide hormone variants having diminished activity with one biological function and substantial or increased activity with a second target substance.

Still further it is an object herein to provide hGH variants having modified binding and/or biological activity with the somatogenic receptor for hGH and increased potency.

Still further it is an object herein to provide hGH variants which retain one or more desirable biological properties but which also have decreased diabetogenic activity.

Further, it is an object herein to provide hPRL and hPL variants having an increased binding activity with the somatogenic receptor of hGH.

Further, it is an object herein to provide DNA sequences, vectors and expression hosts containing such vectors for the cloning and expression of polypeptide variants including hGH variants.

SUMMARY OF THE INVENTION

In one aspect, the invention provides methods for the systematic analysis of the structure and function of polypeptides by identifying unknown active domains which influence the activity of the polypeptide with a first target substance. Such unknown active-domains in one aspect of the invention may comprise at least two discontinuous amino acid segments in the primary amino acid sequence of the polypeptide. Active domains are determined by substituting selected amino acid segments of the polypeptide (referred to as the parent polypeptide) with an analogous amino acid segment from an analog to the polypeptide. The analog has a different activity with the target substance as compared to the parent polypeptide. The segment-substituted polypeptides so formed are assayed to determine the activity of each of the segment-substituted polypeptides with the target substance. Such activities are compared to the same activity for the parent polypeptide. Since the structurally analogous amino acid segments are obtained from an analog that has a different interaction with the target substance, a comparison of such activities provides an indication of the location of the active domain in the parent polypeptide.

The method further comprises identifying the active amino acid residues within the active domain of the parent polypeptide. The method comprises substituting a scanning amino acid for one of the amino acid residues within the active domain of the parent polypeptide and assaying the residue-substituted polypeptide so formed with a target substance. The activity of each of the residue-substituted polypeptides is compared to the same activity of the parent polypeptide. These steps are repeated for different amino acids in the active domain until the active amino acid residues are identified.

In another aspect, the invention provides methods to identify different active domains and active amino acid residues for different target substances. Such methods comprise repeating the foregoing methods with a second target.

In accordance with the foregoing method, polypeptide variants are identified which have a different activity with one or more target substance as compared to a parent polypeptide. Such variants are produced based on the identification of the active domains or the identification of the active amino acid residues in the active domain which determine the activity of the parent polypeptide with a target substance.

The invention further comprises growth hormone, prolactin, and placental lactogen variants comprising at least three portions. The first portion corresponds to at least a part of the amino acid sequence of a parent hormone, the third portion corresponds to the amino acid sequence of at least part of the same parent hormone, and the second portion corresponds to an amino acid sequence of an analog to the parent hormone. The second portion is analogous to those amino acid residues of the parent hormone not contained between the first and third portions of the polypeptide variant.

The invention also includes specific human growth hormone,human prolactin and human placental lactogen variants comprising segment-substituted and residue-substituted variants of hGH.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1

depicts the strategy used to identify active domains.

FIG. 2

shows the conserved and variable amino acid residues amongst the amino acid sequences of hGH, hPL, pGH and hPRL.

FIG. 3

shows the putative low resolution structure of hGH and helical wheel projections viewed from the N-terminal start residue for each helix. Hydrophobic, neutral and charged residues are indicated by ◯, ▪ and &Circlesolid; symbols, respectively.

FIG. 4

is a bar graph showing the relative reduction in binding of various segment-substituted hGH variants to the soluble hGH receptor.

FIG. 5

depicts the analogous amino acids in the active domains A, C and F which interact with the somatogenic hGH receptor.

FIG. 6

depicts the relative binding positions of the somatogenic receptor and eight monoclonal antibodies to hGH. The top, middle, and bottom panels in

FIG. 6A

show the binding positions of monoclonal antibodies

1

,

8

, and

7

, respectively. In

FIG. 6B

, the top and bottom panels show the binding positions of monoclonal antibodies

2

and

6

, respectively, while the binding position of the somatogenic receptor is shown in the middle panel. The top, middle, and bottom panels in

FIG. 6C

show the binding positions of monoclonal antibodies

3

,

4

, and

5

, respectively.

FIG. 7

is a bar graph showing the relative increase or decrease in binding to the soluble hGH somatogenic receptor for various alanine-substituted hGH variants. The stippled bar at T175 indicates that serine rather than alanine is substituted. The broken bar at R178 indicates that asparagine rather than alanine is substituted.

FIG. 8

depicts the DNA and amino acid sequence of the hGH gene used in the examples.

FIG. 9

depicts the construction of vector pB0475 which contains a synthetic hGH gene.

FIG. 10

is the DNA sequence of pB0475 showing the amino acid sequence for hGH.

FIG. 11

depicts the construction of vector pJ1446.

FIG. 12

is the DNA sequence for pJ1446 showing the amino acid sequence for the soluble portion of the somatogenic receptor from liver.

FIGS. 13 through 20

depict the epitope binding sites on hGH for each of eight different monoclonal antibodies.

FIG. 21

shows the active amino acids involved in binding to the somatogenic receptor in hGH and helical wheel projections for helices

1

and

4

.

FIG. 22

shows the rat weight gain versus time for hGH and hGH, variants administered at 50 micrograms/kg/day. The top panel shows weight gain in rats treated with wild type hGH (WT), the E174A variant of hGH, and the segment-substituted hGH variant hPRL (88-95). The bottom panel shows weight gain in rats treated with wild type hGH as compared to hPRL.

FIG. 23

is a semilog plot of Kd ratio versus potency for hGH variants as compared to wild-type hGH.

FIG.

24

. Competitive binding curves of [

125

I]hGH and cold hGH to the hGH binding protein isolated from either human serum (◯) or from

E. coli

KS330 cultures expressing the plasmid phGHr(1-238) (&Circlesolid;). Bars represent standard deviations from the mean. Inset shows Scatchard plots that were derived from the competitive binding curves. The concentrations of the binding protein from human serum and

E. coli

were 0.1 and 0.08 nM, respectively.

FIG.

25

. Structural model of hGH based on a folding diagram for pGH determined from a 2.8 Å resolution X-ray structure. Panel A shows a functional contour map of the hGH receptor epitope and Panel B shows that determined here for the hPRL receptor epitope. The size of the closed circles corresponds to the magnitude of the disruptive effect for alanine substitution at these residues. The small circles represent >2-fold disruption whereas the larger circles represent >10-fold disruption. The ▴ in the hGH receptor epitope (Panel A) represents the position of E174A that causes greater than a four-fold increase in binding affinity.

FIG.

26

. Plasmid diagram of pBO760 used for intracellular expression of hPRL in

E. coli.

FIG.

27

. Location of residues in hGH that strongly modulate its binding to the hGH binding protein. Alanine substitutions (serine or asparagine in the case of T175 or R178, respectively) are indicated that cause more than a 10-fold reduction (◯), a 4- to 10-fold reduction (▪), or more than a 4-fold increase (▴) in binding affinity. Helical wheel projections in regions of α-helix reveal their amphipathic quality and the fact that in helix

4

the most important determinants are on its hydrophilic face (shaded).

FIG.

28

. Circular dichroic spectra in the far UV (Panel A) or near UV (Panel B) of hGH (-), wild-type hPRL (- -) and hPRL variant D ( - - - ) (see Table XXIII).

FIG.

29

. Sequence comparison of hGH and HPRL in regions defined by homolog and alanine scanning mutagenesis to be important for binding. Identical residues are shaded and the numbering is based on the hGH sequence. Residues are circled that when mutated cause more than a 4-fold change in binding affinity. Asterisks above residues indicate sites at which mutations cause a 2- to 4-fold reduction in binding affinity.

DETAILED DESCRIPTION OF THE INVENTION

In one embodiment, the method of the invention provides for the systematic analysis of a parent polypeptide, such as human growth hormone or human prolactin, to determine one or more active domains in the polypeptide that are involved in the interaction of the parent polypeptide with a target substance. To employ the method of the invention, one or more analogs to the polypeptide of interest must exist which exhibit a different activity with the target substance of interest.

Accordingly; as used herein, “parent polypeptide” refers to any polypeptide for which an “analog” exists that has a different activity with a target substance as compared to the same activity for the parent polypeptide. Examples of such polypeptides, analogs and target substances are shown in Table I.

TABLE I

Parent

Target or Assay

Polypeptide

Analog

Containing Target

Human growth

Human placental

Receptors for somatogenic,

hormone

lactogen, human

lactogenic, diabetagenic,

prolactin and

lipolytic, nitrogen

porcine growth

retention, macrophage

hormone

activation and insulin-like

effects of hGH; rat tibia

assay, rat weight gain

assay, insulin resistance

assay in OB/OB mice or dog,

receptors on human liver,

adipose, lymphocytes,

thymocytes and ovary tissue

hPRL

pGH

Binding to human prolactin

receptor

Rabbit GH

Human GH

Binding to rabbit GH

receptor

receptor

α-interferon

Related human

Binding to α

1

interferon

interferons and

receptor

animal interferons

human tissue

human TGF-β

2

Human hemopoietic cell

growth factor

or inhibins

growth modulation

(TGF-β

1

)

Epidermal growth

TGF-α

Carotinocyte proliferation

factor (EGF)

Mouse Tumor

Human Tumor

Mouse TNF receptor

Necrosis

Necrosis

activity

Factor (mTNF)

Factor (hTNF)

human granulocyte

mouse granulocyte

Growth and differentiation

macrophage colony

macrophage colony

of human bone marrow stem

stimulating

stimulating

cells

factor (hGMCSF)

factor (mGMCSF)

human CD-4

mouse CD-4

gp-120 from HIV virus

receptor

receptor

Subtilisin

Subtilisin

succinyl-ala-ala-pro-glu-

Bacillus

Bacillus

P-Nitroanilyd

Amylilquifaciens

licheniformis

human

Related human

Activation of human

γ-interferon

interferons and

interferon receptor

animal interferons,

e.g., from mouse

Insulin-like growth

Insulin

IGF-1 receptor growth

factor (IGF-1)

growth modulation receptor

Tissue

Trypsin, urokinase

Plasminogen (cleavage)

Plasminogen

fibrin (binding)

Activator (tPA)

The parent polypeptides, analogs and target substances in Table I, of course, are exemplary only.

Parent polypeptides also include proteinaceous material comprising one or more subunits, e.g. succinyl coenzyme A synthetase, mitochondrial ATPase, aminoacyl tRNA synthetase, glutamine synthetase, glyceraldehyde-3-phosphate dehydrogenase and aspartate transcarbamolase (see, Huang, et al. (1982),

Ann. Rev. Biochem,

51, 935-971). In such multi-subunit parent polypeptides, active domains may span the two or more subunits of the parent polypeptide. Accordingly, the methods as described in more detail hereinafter can be used to probe each of the subunits of a particular polypeptide to ascertain the active domain and active amino acid residues for a particular target which may be partially contained on one subunit and partially on one or more other subunits.

The parental polypeptide and analog typically belong to a family of polypeptides which have related functions. Moreover, such parental polypeptides and analogs ordinarily will have some amino acid sequence identity, i.e., conserved residues. Such sequence homology may be as high as 90% but may range as low as about 15% to 20%.

In addition to primary sequence homology, an analog to a parent polypeptide may be defined by the three-dimensional framework of the polypeptide and analog. Thus, an analog may be divergent from a parent polypeptide in amino acid sequence but structurally homologous to the parent polypeptide based on a comparison of all, or part, of the tertiary structure of the molecules. Chothia, C., et al. (1986)

Embo. J.

5, 823.

In general, tertiary analogs can be identified if the three-dimensional structure of a possible analog is known together with that of the parent polypeptide. By performing a root means squared differences (RMS) analysis of the α-carbon coordinates, (e.g., Sutcliffe, M. J., et al. (1987)

Protein Engineering

1, 377-384), the superposition of regions having tertiary analog y, if any, are identified. If the α-carbon coordinates overlap or are within about 2 Å to about 3.5 Å RMS for preferably about 60% or more of the sequence of the test analog relative to the α-carbon coordinates for the parent polypeptide, the test analog is a tertiary analog to the parent polypeptide. This, of course, would exclude any insertions or deletions which may exist between the two sequences.

Although the above parent polypeptide and analogs disclose naturally occurring molecules, it is to be understood that parent polypeptides and analogs may comprise variants of such sequences including naturally occurring variants and variations in such sequences introduced by in vitro recombinant methods. Variants used as parent polypeptides or analogs thus may comprise variants containing the substitution, insertion and/or deletion of one or more amino acid residues in the parent polypeptide or analog. Such variants may be used in practicing the methods of the invention to identify active domains and/or amino acids or to prepare the polypeptide variants of the invention. Thus, the naturally occurring variants of hGH or the recombinantly produced variant containing the substitution of Cys-165 with Ala may be used as parent polypeptide or an analog so long as they have some activity with a target. Such naturally occurring and recombinantly produced variants may contain different amino acid residues which are equivalent to specific residues in another parent polypeptide. Such different amino acids are equivalent if such residues are structurally analogous by way of primary sequence or tertiary structure or if they are functionally equivalent.

Further, it should be apparent that many of the parent polypeptides and analogs can exchange roles. Thus, non-human growth hormones and their related family of analogs each can be used as a parent polypeptide and homolog to probe for active domains. Further, targets such as the CD-4 receptor for the HIV virus, may be used as a parent polypeptide with analog CD-4 receptors to identify active domains and amino acids responsible for binding HIV and to make CD-4 variants.

As used herein, a “target” is a substance which interacts with a parent polypeptide. Targets include receptors for proteinaceous hormones, substrates for enzymes, hormones for proteinaceous receptors, generally any ligand for a proteinaceous binding protein and immune systems which may be exposed to the polypeptides. Examples of hormone receptors include the somatogenic and lactogenic receptors for hGH and the receptor for hPRL. Other targets include antibodies, inhibitors of proteases, hormones that bind to proteinaceous receptors and fibrin which binds to tissue plasminogen activators (t-PA).

Generally, targets interact with parent polypeptides by contacting an “active domain” on the parent polypeptide. Such active domains are typically on the surface of the polypeptide or are brought to the surface of the polypeptide by way of conformational change in tertiary structure. The surface of a polypeptide is defined in terms of the native folded form of the polypeptide which exists under relevant physiological conditions, i.e. in vivo or under similar conditions when expressed in vitro. The amino acid segments and amino acid residues may be ascertained in several ways. If the three dimensional crystal structure is known to sufficient resolution, the amino acid residues comprising the surface of the polypeptide are those which are “surface accessible”. Such surface accessible residues include those which contact a theoretical water molecule “rolled” over the surface of the three dimensional structure.

The active domain on the surface of the polypeptide may comprise a single discrete segment of the primary amino acid sequence of the polypeptide. In many instances, however, the active domain of a native folded form of a polypeptide comprises two or more discontinuous amino acid segments in the primary amino acid sequence of the parent polypeptide. For example, the active domain for human growth hormone with the somatogenic receptor is shown in FIG.

5

. As indicated, domain A, C and F of the active domain are each located on discontinuous amino acid segments of the hGH molecule. These amino acid segments are identified in

FIG. 4

by the letters A, C and F. Discontinuous amino acid segments which form an active domain are separated by a number of amino acid residues which are not significantly involved in the interaction between the active domain and the target. Typically, the separation between discontinuous amino acid segments is usually at least about five amino acids.

The methods of the invention are directed to the detection of unknown active domains in the amino acid sequence of a parent polypeptide. Except for those few cases where a three dimensional crystal structure of a polypeptide with its target are available, e.g. the crystal structure of enzymes with inhibitors or transition state analogs, most active domains for a vast array of polypeptides remain unknown.

As used herein an “analogous polypeptide segment” or “analogous segment” refers to an amino acid sequence in an analog which is substituted for the corresponding sequence in a parent polypeptide to form a “segment-substituted polypeptide”. Analogous segments typically have a sequence which results in the substitution, insertion or deletion of one or more different amino acid residues in the parent polypeptide while maintaining the relative amino acid sequence of the other residues in the selected segment substituted in the parent. In general, analogous segments are identified by aligning the overall amino acid sequence of the parent polypeptide and analog to maximize sequence identity between them. Analogous segments based on this sequence alignment are chosen for substitution into the corresponding sequence of the parent polypeptide. Similarly, analogous segments from analogs showing tertiary homology can be deduced from those regions showing structural homology. Such analogous segments are substituted for the corresponding sequences in the parent. In addition, other regions in such tertiary homologs, e.g., regions flanking the structurally analogous region, may be used as analogous segments.

The analogous segment should be selected, if possible, to avoid the introduction of destabilizing amino acid residues into the segment-substituted polypeptide. Such substitutions include those which introduce bulkier side chains, and hydrophilic side chains in hydrophobic core regions.

Typically, the amino acid sequences of the parent polypeptide and analog are known and in some cases three-dimensional crystal structures may be available. An alignment of the amino acid sequence of the parent polypeptide with that of one or more analogs readily reveals conserved amino acid residues in the sequences which should not be altered, at least in the preliminary analysis. Sequence alignment will also reveal regions of sequence variation which may include the substitution, insertion or deletion of one or more amino acid residues. Those regions containing such variations determine which segments in the parent may be substituted with an analogous segment. The substitution of an analogous segment from an analog may therefore result not only in the substitution of amino acid residues but also in the insertion and/or deletion of amino acid residues.

If three-dimensional structural information is available, it is possible to identity regions in the parent polypeptide which should not be subjected to substitution with an analogous segment. Thus, for example, the identification of a tightly packed region in a hydrophobic face of an amphiphilic helix in the parent polypeptide should not be substituted with an analogous segment. Residues identified as such should be retained in the polypeptide variant and only surface residues substituted with analogous residues.

Generally, analogous segments are 3 to 30 amino acid residues in length, preferably about 3 to 15 and most preferably about 10 to 15 amino acid residues in length. In some instances, the preferred length of the analogous segment may be attenuated because of the insertion and/or deletion of one or more amino acid residues in the analogous segment as compared to the homolog or parent polypeptide. If a three-dimensional structure is unavailable for the parent polypeptide, it is generally necessary to form segment-substituted polypeptides with analogous segments covering most, if not all, of the parent polypeptide. Segment-substitution of the entire amino acid sequence, however, is not always necessary. For example, fortuitous segment-substitutions covering only a portion of the total amino acid sequence may provide sufficient information to identify the active domain for a particular target. Thus, for example, the segment-substitution of about 15% of the amino acid sequence of the parent polypeptide may provide sufficient indication of the active domain. In most instances, however, substantially more than about 15% of the amino acid sequence will need to be segment-substituted to ascertain the active domain. Generally, about 50%, preferably about 60%, more preferably about 75% and most preferably 100% of the amino acid sequence will be segment-substituted if no structural information is available.

As used herein, “analogous amino acid residue” or “analogous residue” refers to an amino acid residue in an analogous segment which is different from the corresponding amino acid residue in the corresponding segment of a parent polypeptide. Thus, if the substitution of an analogous segment results in the substitution of one amino acid, that amino acid residue is an analogous residue.

Once the parent polypeptide and one or more analogs are identified, the analogous segments from one or more of the analogs are substituted for selected segments in the parent polypeptide to produce a plurality of segment-substituted polypeptides. Such substitution is conveniently performed using recombinant DNA technology. In general, the DNA sequence encoding the parent polypeptide is cloned and manipulated so that it may be expressed in a convenient host. DNA encoding parent polypeptides can be obtained from a genomic library, from cDNA derived from mRNA from cells expressing the parent polypeptide or by synthetically constructing the DNA sequence (Maniatis, T., et al. (1982) in

Molecular Cloning,

Cold Springs Harbor Laboratory, N.Y.).

The parent DNA is then inserted into an appropriate plasmid or vector which is used to transform a host cell. Prokaryotes are preferred for cloning and expressing DNA sequences to produce parent polypeptides, segment substituted polypeptides, residue-substituted polypeptides and polypeptide variants. For example,

E. coli

K12 strain 294 (ATCC No. 31446) may be used as well as

E. coli

B,

E. coli

X1776 (ATCC No. 31537), and

E. coli

c600 and c600hfl,

E. coli

W3110 (F-, γ-, prototrophic, ATCC No. 27325), bacilli such as

Bacillus subtilis,

and other enterobacteriaceae such as

Salmonella typhimurium

or

Serratia marcesans,

and various pseudomonas species. The preferred prokaryote is

E. coli

W3110 (ATCC 27325). When expressed in prokaryotes the polypeptides typically contain an N-terminal methionine or a formyl methionine, and are not glycosylated. These examples are, of course, intended to be illustrative rather than limiting.

In addition to prokaryotes, eukaryotic organisms, such as yeast cultures, or cells derived from multicellular organisms may be used. In principle, any such cell culture is workable. However, interest has been greatest in vertebrate cells, and propagation of vertebrate cells in culture (tissue culture) has become a repeatable procedure (

Tissue Culture,

Academic Press, Kruse and Patterson, editors (1973)). Examples of such useful host cell lines are VERO and HeLa cells, Chinese Hamster Ovary (CHO) cell lines, W138, BHK, COS-7 and MDCK cell lines.

In general, plasmid vectors containing replication and control sequences which are derived from species compatible with the host cell are used in connection with these hosts. The vector ordinarily carries a replication site, as well as sequences which encode proteins that are capable of providing phenotypic selection in transformed cells. For example,

E. coli

may be transformed using pBR322, a plasmid derived from an

E. coli

species (Mandel, M. et al. (1970)

J. Mol. Biol.

53, 154). Plasmid pBR322 contains genes for ampicillin and tetracycline resistance and thus provides easy means for selection. A preferred vector is pB0475. See Example 1. This vector contains origins of replication for phage and

E. coli

which allow it to be shuttled between such hosts thereby facilitating mutagenesis and expression.

“Expression vector” refers to a DNA construct containing a DNA sequence which is operably linked to a suitable control sequence capable of effecting the expression of said DNA in a suitable host. Such control sequences include a promoter to effect transcription, an optional operator sequence to control such transcription, a sequence encoding suitable mRNA ribosome binding sites, and sequences which control termination of transcription and translation. The vector may be a plasmid, a phage particle, or simply a potential genomic insert. Once transformed into a suitable host, the vector may replicate and function independently of the host genome, or may, in some instances integrate into the genome itself. In the present specification, “plasmid” and “vector” are sometimes used interchangeably as the plasmid is the most commonly used form of vector at present. However, the invention is intended to include such other forms of expression vectors which serve equivalent functions and which are, or become, known in the art.

“Operably linked” when describing the relationship between two DNA or polypeptide regions simply means that they are functionally related to each other. For example, a presequence is operably linked to a peptide if it functions as a signal sequence, participating in the secretion of the mature form of the protein most probably involving cleavage of the signal sequence. A promoter is operably linked to a coding sequence if it controls the transcription of the sequence; a ribosome binding site is operably linked to a coding sequence if it is positioned so as to permit translation.

Once the parent polypeptide is cloned, site specific mutagenesis (Carter, P., et al. (1986)

Nucl. Acids Res.

13, 4331; Zoller, M. J., et al. (1982)

Nucl. Acids Res.

10, 6487), cassette mutagenesis (Wells, J. A., et al. (1985)

Gene

34, 315), restriction selection mutagenesis (Wells, J. A., et al. (1986)

Philos. Trans. R. Soc. London SerA

317, 415) or other known techniques may be performed on the cloned parent DNA to produce “segment-substituted DNA sequences” which encode the changes in amino acid sequence defined by the analogous segment being substituted. When operably linked to an appropriate expression vector, segment-substituted polypeptides are obtained. In some cases, recovery of the parent polypeptide or segment-modified polypeptide may be facilitated by expressing and secreting such molecules from the expression host by use of an appropriate signal sequence operably linked to the DNA sequence encoding the parent polypeptide or segment-modified polypeptide. Such methods are well-known to those skilled in the art. Of course, other methods may be employed to produce such polypeptides and segment-substituted polypeptides such as the in vitro chemical synthesis of the desired polypeptide (Barany, G., et al. (1979) in

The Peptides

(eds. E. Gross and J. Meienhofer) Acad. Press, N.Y., Vol. 2, pp. 3-254).

Once the different segment-substituted polypeptides are produced, they are contacted with a target for the parent polypeptide and the interaction, if any, of the target and each of the segment-substituted polypeptides is determined. These activities are compared to the activity of the parent polypeptide with the same target. If the analog has a substantially altered activity with the target as compared to the parent polypeptide, those segment-substituted polypeptides which have altered activity with the target presumptively contain analogous segments which define the active domain in the parent polypeptide.

If the analog has an activity with the target which is greater than that of the parent polypeptide, one or more of the segment-substituted polypeptides may demonstrate an increased activity with that target substance. Such a result would, in effect, identify an active domain in the analog and an appropriate region in the parent polypeptide which can be modified to enhance its activity with that target substance. Such an event is most likely when the region in the analog responsible for the target interaction is contained primarily within one continuous amino acid segment. If the “active domains” of the analog comprise discontinuous regions in the amino acid sequence of the analog, enhanced activity in the segment-substituted polypeptide is less likely since the demonstration of such enhanced activity may require the simultaneous introduction of all active domains from the analog into the segment-substituted polypeptide.

Accordingly, it is preferred that the analog have an activity with the target which is less than that for the parent-polypeptide. In this manner, a loss in activity is observed in the segment-substituted polypeptide. However, once the active domains in a parent polypeptide are determined, that polypeptide may be used as an analog to sequentially or simultaneously substitute such active domains into a second parent polypeptide which lacks activity with the target for the first parent polypeptide.

Active domains in polypeptides are identified by comparing the activity of the segment-substituted polypeptide with a target with the activity of the parent polypeptide. Any number of analytical measurements may be used but a convenient one for non-catalytic binding of target is the dissociation constant Kd of the complex formed between the segment-substituted polypeptide and target as compared to the Kd for the parent. An increase or decrease in Kd of about 1.5 and preferably about 2.0 per analogous residue substituted by the segment-substitution indicates that the segment substituted is an active domain in the interaction of the parent polypeptide with the target.

In the case of catalytic interaction with a target, a suitable parameter to measure activity relative to the parent enzyme is to compare the ratios of kcat/Km. An increase or decrease in kcat/Km relative to the parent enzyme of about 1.5 and preferably 2.0 per analogous residue-substituted indicates that an active domain has been substituted.

As used herein, a “scanning amino acid” is an amino acid used to identify active amino acids within a parent polypeptide. A “residue-substituted polypeptide” is a polypeptide variant containing at least a single substitution of an amino acid in the parent polypeptide with a scanning amino acid. A “residue-substituted DNA sequence” encodes a residue substituted polypeptide. Such DNA and polypeptide sequences may be prepared as described for the preparation of segment-substituted DNA and polypeptides.

The “active amino acid residue” identified by the amino acid scan is typically one that contacts the target directly. However, active amino acids may also indirectly contact the target through salt bridges formed with other residues or small molecules such as H

2

O or ionic species such as Na

+

, Ca

+2

, Mg

+2

or Zn

+2

.

In some cases, the scanning amino acid is substituted for an amino acid identified in an active domain of the parent polypeptide. Typically, a plurality of residue-substituted polypeptides are prepared, each containing the single substitution of the scanning amino acid at a different amino acid residue within the active domain. The activities of such residue-substituted polypeptides with a particular target substance are compared to the activity of the parent polypeptide to determine which of the amino acid residues in the active domain are involved in the interaction with the target substance. The scanning amino acid used in such an analysis may be any different amino acid from that substituted, i.e., any of the 19 other naturally occurring amino acids.

TABLE II

Isosteric

Polypeptide

scanning

Amino Acid

Amino Acid

Ala

Ser, Gly

Glu

Gln, Asp

Gln

Asn, Glu

Asp

Asn, Glu

Asn

Aln, Asp

Leu

Met, Ile

Gly

Pro, Ala

Lys

Met, Arg

Ser

Thr, Ala

Val

Ile, Thr

Arg

Lys, Met, Asn

Thr

Ser, Val

Pro

Gly

Ile

Met, Leu, Val

Met

Ile, Leu

Phe

Tyr

Tyr

Phe

Cys

Ser, Ala

Trp

Phe

His

Asn, Gln

This table uses the following symbols for each amino acid:

Amino Acid

or residue

3-letter

1-letter

thereof

symbol

symbol

Alanine

Ala

A

Glutamate

Glu

E

Glutamine

Gln

Q

Aspartate

Asp

D

Asparagine

Asn

N

Leucine

Leu

L

Glycine

Gly

G

Lysine

Lys

K

Serine

Ser

S

Valine

Val

V

Arginine

Arg

R

Threonine

Thr

T

Proline

Pro

P

Isoleucine

Ile

I

Methionine

Met

M

Phenylalanine

Phe

F

Tyrosine

Tyr

Y

Cysteine

Cys

C

Tryptophan

Trp

W

Histidine

His

H

Most preferably, the scanning amino acid is the same for each residue substituted polypeptide so that the effect, if any, on the activity of the residue-substituted polypeptides can be systematically attributed to the change from the naturally occurring amino acid residue to a uniform scanning amino acid residue.

In some cases, the substitution of a scanning amino acid at one or more residues results in a residue-substituted polypeptide which is not expressed at levels which allow for the isolation of quantities sufficient to carry out analysis of its activity with a target. In such cases, a different scanning amino acid, preferably an isosteric amino acid, can be used.

The most preferred scanning amino acids are relatively small, neutral amino acids. Such amino acids include alanine, glycine, serine and cysteine. Alanine is the preferred scanning amino acid among this group because it eliminates the side-chain beyond the beta-carbon and is less likely to alter the main-chain conformation of the residue-substituted polypeptide. Alanine is also preferred because it is the most common amino acid. Further, it is frequently found in both buried and exposed positions (Creighton, T. E., in

The Proteins

(eds. W.H. Freeman & Co., N.Y.); Chothia, C. (1976)

J. Mol. Biol.

150, 1). If alanine substitution does not yield adequate amounts of residue-substituted polypeptide, an isosteric amino acid can be used. Alternatively, the following amino acids in decreasing order of preference may be used: Ser, Asn and Leu.

The use of scanning amino acids is not limited to the identification of active amino acids in an active domain ascertained by the analysis of segment-substituted polypeptides. If, for example, one or more amino acids in a parent polypeptide are known or suspected to be involved in the interaction with a target, scanning amino acid analysis may be used to probe that residue and the amino acid residues surrounding it. Moreover, if the parent polypeptide is a small peptide, e.g., about 3 to 50 amino acid residues, scanning amino acid analysis may be carried out over the entire molecule.

Once the active amino acid residues are identified, isosteric amino acids may be substituted. Such isosteric substitutions need not occur in all instances and may be performed before any active amino acid is identified. Such isosteric amino acid substitution is performed to minimize the potential disruptive effects on conformation that some substitutions can cause. Isosteric amino acids are shown in Table II.

Active amino acid residues can be identified by determining the activity of the residue-substituted polypeptide with a target as compared to the parent. In general, a two-fold increase or decrease in Kd indicates that the residue substituted is active in the interaction with the target. Similarly, in the case of catalytic interaction with a target, a two-fold increase or decrease in kcat/Km relative to the parent enzyme indicates that an active residue has been substituted.

When a suspected or known active amino acid residue is subjected to scanning amino acid analysis the amino acid residues immediately adjacent thereto should be scanned. Three residue-substituted polypeptides are made. One contains a scanning amino acid, preferably alanine, at position N which is the suspected or known active amino acid. The two others contain the scanning amino acid at position N+1 and N−1. If each substituted polypeptide causes a greater than about two-fold effect on Kd or kcat/Km for a target, the scanning amino acid is substituted at position N+2 and N−2. This is repeated until at least one and preferably four residues are identified in each direction which have less than about a two-fold effect on Kd or kcat/Km or either of the ends of the parent polypeptide are reached. In this manner, one or more amino acids along a continuous amino acid sequence which are involved in the interaction with a particular target can be identified.

The methods of the invention may be used to detect the active domain for more than one target of a particular parent polypeptide. Further, active amino acid residues within the different active domains may be also identified by the methods herein. Once two or more active domains and active amino acid residues are identified for the different targets of a particular polypeptide, various modifications to the parent polypeptide may be made to modify the interaction between the parent polypeptide and one or more of the targets. For example, two active domains on the surface of hGH have been identified for the somatogenic and prolactin receptor. In this particular case, the active domains overlap. Accordingly, there are a number of common active amino acid residues which interact with the somatogenic and prolactin receptors. Various modifications to hGH may be made based on this information as described in more detail hereinafter.

In some instances, the active domain for different targets will not overlap. In such situations, the active amino acids in the parent polypeptide for one receptor can be substituted with different amino acids to reduce or enhance the interaction of that active domain with its target, thus shifting the physiological effect of such a variant.

As used herein, the term “modified interaction” refers to a polypeptide variant wherein one or more active domains have been modified to change the interaction of the variant with a target as compared to the parent polypeptide. A modified interaction is defined as at least a two-fold increase or decrease in the interaction of the polypeptide variant as compared to the interaction between the parent polypeptide and a particular target.

The interaction between a target and a parent polypeptide, polypeptide variant, segment-substituted polypeptide and/or residue-substituted polypeptide can be measured by any convenient in vitro or in vivo assay. Thus, in vitro assays may be used to determine any detectable interaction between a target and polypeptide, e.g. between enzyme and substrate, between hormone and hormone receptor, between antibody and antigen, etc. Such detection may include the measurement of colorimetric changes, changes in radioactivity, changes in solubility, changes in molecular weight as measured by gel electrophoresis and/or gel exclusion methods, etc. In vivo assays include, but are not limited to, assays to detect physiological effects, & weight gain, change in electrolyte balance, change in blood clotting time, changes in clot dissolution and the induction of antigenic response. Generally, any in vivo assay may be used so long as a variable parameter exists so as to detect a change in the interaction between the target and the polypeptide of interest.

Exemplary of the present invention is a preferred embodiment wherein the active domains and active amino acids of human growth hormone which determine its activity with its somatogenic receptor are identified. In carrying out this embodiment of the invention, human growth hormone variants, including segment-substituted and residue-substituted hGH variants, have been made or identified which have different binding interactions with the somatogenic receptor for growth hormone as compared to naturally occurring human growth hormone. At least one of these human growth hormone variants has a higher affinity for the somatogenic receptor and enhanced potency for somatogenesis in rats. Others have a decreased activity with the somatogenic receptor. Such hGH variants are useful as hGH agonists or antagonists and may have a higher potency for stimulating other receptors for human growth hormone since such variants will be freed from substantial interaction with the somatogenic receptor. Further, such variants are useful in immunoassays for hGH as an hGH standard or tracer. In one instance, a variant has been identified which has a significant decrease in reactivity with human and mouse serum containing anti-hGH polyclonal antibodies. Another has the same binding affinity for the somatogenic receptor as hGH but increased potency to stimulate growth.

The method for determining the active domains for human growth hormone which interact with its somatogenic receptor from liver is shown schematically in FIG.

1

. In this approach, segments of hGH were systematically replaced with analogous sequences from analogs of hGH that are known to have greatly reduced affinities for the cloned hGH liver receptor and for monoclonal antibodies raised against hGH. Such analogs for hGH include human placental lactogen (hPL), porcine growth hormone (pGH) and human prolactin (hPRL). These analogs have binding affinities for the cloned hGH receptor that are reduced by about 100 to 10,000-fold for the somatogenic hGH, receptor (hGHr) (Harrington, A. C., et al. (1986)

J. Clin. Invest.

77, 1817; Baumann, G., et al. (1986)

J. Clin. Endocrinol. Metab.

62, 137). Such analogs are used because homologous proteins are known to have similar three-dimensional structures even though they may have a large sequence divergence (Chothia, C., et al. (1986)

EMBO J.

5, 823). In so doing, the likelihood is increased that analogous sequence substitutions will be readily accommodated without grossly disrupting the native folding of the molecule. The amino acid sequences for human growth hormone and the analogs hPL, pGH and hPRL are shown in FIG.

2

. These latter three analogs share a sequence identity with hGH at the level of 85%, 68% and 23%, respectively.

Referring to

FIG. 1

, the overall strategy is shown for identifying one or more active domains in human growth hormone which interact with the somatogenic receptor for human growth hormone (a “target” for hGH). As indicated, hGH has a positive binding activity with the target receptor, in this case, the somatogenic receptor. The hPRL, hPL and pGH analogs, however, have a greatly reduced activity with that target as indicated by the minus sign. Six segment-substituted growth hormones, identified by letters A through F, are formed by substituting a selected amino acid segment of hGH with an analogous amino acid segment from the hPRL analog. Each of these selected segments are different and are chosen to probe either the entire amino acid sequence of the hGH molecule or those regions which are expected to contain the active domains. After the segment-substituted human growth hormones are prepared each is assayed against the hGH somatogenic receptor to determine its activity. The results of such an assay as compared to hGH are indicated by + or − under the segment-modified human growth hormones in FIG.

1

. As can be seen in

FIG. 1

, segment-substituted human growth hormones C and F in this schematic do not bind the somatogenic, receptor. Based on these results, those regions in human growth hormone corresponding to the analogous segments from the analog in the growth hormone variants C and F are identified as active domains involved in the binding of hGH to its somatogenic receptor.

As indicated, it is not necessary to probe the entire amino acid sequence of human growth hormone or other parental polypeptides if structural information or other data are available. Thus, low-resolution or high-resolution structural information from crystallographic studies can provide important information so as to avoid destabilizing substitutions of selected amino acid segments from a homolog. For example, the X-ray coordinates for human growth hormone are not available. However, helix wheel projections from the pGH folding model, based on the low resolution X-ray crystal structure of pGH, reveal that three of the four helices (helix

1

,

3

and

4

) are amphipathic with strong hydrophobic moments. See FIG.

3

. Eisenberg, D., et al. (1984)

J. Mol. Biol.

179, 125. Since the hydrophobic core in polypeptides is very tightly packed (Ponder, J. W., et al. (1987)

J. Mol. Biol.

193, 775), changes in such buried amino acid residues are generally destabilizing (Alber, T., et al. (1987)

Biol. Chem.

26, 3754; Reidhaar-Olson, J. F. (1988)

Science

241, 53).

In addition, regions of high amino acid sequence conservation amongst members of the polypeptide family, for example the human growth hormone family, in general, need not be probed, at least initially. This is because the disruption of such conserved sequences is likely to disrupt the folding of the molecule. Further, other data may suggest that certain regions of the parent polypeptide are not involved in the interaction with a particular target substance. For example, deletion of the N-terminal 13 amino acids of hGH by mutagenesis (Ashkenazi, A., et al. (1987)

Endocrinology

121, 414) and a natural variant of hGH which deletes residues 32 to 46 (the 20 Kd variant; Lewis, U. J., et al. (1980)

Biochem. Biophys. Res. Commun.

92, 5111) have been reported not to affect dramatically the binding properties to the somatogenic receptor. In addition, the production of a two-chain derivative of hGH by limited proteolysis, which deletes some or all of the residues between 134 and 149, does not markedly affect binding to the somatogenic receptor. Li, C. H. (1982)

Mol. Cell. Biochem.

46, 31; Mills, J. B., et al. (1980)

Endocrinology

107, 391.

Based on this information, six segments of the amino acid sequence of hGH were selected for substitution with the corresponding analogous amino acid segments from a number of analogs to hGH. These selected segments are identified as A through F in FIG.

2

. These segments are separated either by disulfide bonds, by borders of secondary structure (see FIG.

4

), by areas of high sequence conservation in the growth hormone family or by regions previously identified as not being involved in binding to the somatogenic receptor. Seventeen segment-substituted hGH variants were prepared which collectively substituted 85 out of the 191 residues in hGH. The regions identified as A through F in FIG.

2

and the segment-substituted hGH variants prepared within each region are summarized in Table III.

TABLE III

Segment-

Actual

K

d

Region

Substituted

Substitution

Mutagenesis

(variant)

probed

hGH Variant

Introduced

method

K

d

(nM)

K

d

(wt)

hGH

None

0.34

1.0

A 11-33

hPL (12-25)

N12H, F25L

r.s.

1

1.4

4.1

pGH (11-33)

D11A, M14V, M18Q

cassette

2

1.2

3.4

R19H, F25A, Q29K,

E33R

hPRL (12-33)

N12R, M14V, L15V,

cassette

3.6

11

R16L, R19Y, F25S,

D26E, Q29S, E30Q,

E33K

hPRL (12-19)

N12R, M14V, L15V,

r.s.

5.8

17

R16L, R19Y

hPRL (22-33)

Q22N, F25S, D26E,

r.s.

0.29

0.85

Q29S, E30Q, E33K

B 46-52

hPL (46-52)

Q46H, N47D, P48S,

r.s.

2.5

7.2

Q49E, L52F

pGH (48-52)

P48A, T50A, S51A,

r.s.

0.94

2.8

L52F

C 54-74

hPL (56-64)

E56D, R64M

cassette

10

30

pGH (57-73)

S57T, T60A, S62T,

cassette

5.8

17

N63G, R64K, E65D,

T67A, K70R, N72D,

L73V

hPRL (54-74)

F54H, S55T, E56S,

cassette

23

69

I58L, P59A, S62E,

N63D, R64K, E66Q,

T67A, K70M, S71N,

N72Q, L73K, E74D

D 88-104

hPRL (88-95)

E88G, Q91Y, F92H,

r.s.

0.47

1.4

R94T, S95E

hPRL (97-104)

F97R, A98G, N99M,

r.s.

0.53

1.6

S100Q, L101D,

V102A, Y103P,

G104E

E 108-136

hPL (109-112)

N109D, V110D,

cassette

0.61

1.8

D112H

hPRL (111-129)

Y111V, L113I,

cassette

0.52

1.5

K115E, D116Q,

E118K, E119R,

G120L, Q122E,

T123G, G126L,

R127I, E129S

hPRL (126-136)

R127D, L128V,

cassette

0.58

1.7

E129H, D130P,

G131E, S132T,

P133K, R134E,

T135N

F 164-190

pGH (164-190)

Y164S, R167K,

hybrid

3

>34

>100

M170L, D171H,

V173A, F176Y,

I179V, V180M,

Q181K, S184R,

I184F, G187S,

G190A

pGH (167-181)

R167K, D171H,

r.s.

9.2

27

I179V, Q181K

1

Restriction selection - Wells, J. A., et al. (1986) Philos. Trans. R. Soc. London SerA 317, 415.

2

Cassette mutagenesis - Wells, J. A., et al. (1985) Gene 34, 315.

3

Recombination mutagenesis - Gray, G. L., et al. (1986) J. Bacteriol. 166, 635.

The segment-substituted hGH variants are generally identified by the analogous segments substituted into the human growth hormone sequence. However, in some instances, not all of the analogous residues in the substituted analogous segment were maintained in a particular construction. Thus, in Table III hPL (12-25) identifies a segment-substituted hGH variant wherein amino acids 12 through 25 of human placental lactogen (hPL) are substituted for amino acid residues 12 through 25 in the parent hGH. The effect of substituting this analogous segment can be determined by comparing the amino acid sequence of hGH and hPL in this region in FIG.

2

. Four amino acid substitutions are generated in an hPL (12-25) variant where no other changes are made. These residues are 12, 16, 20 and 25 for hPL (12-25).

The actual amino acid substitutions in the hPL (12-25) variant and the other segment-substituted variants are shown in Table III. Each substitution is represented by a letter followed by a number which is followed by a letter. The first letter and number correspond to the amino acid at that residue number in the unmodified hGH. The last letter corresponds to the amino acid which is substituted at that position. Thus, N12H indicates that the asparagine at position 12 in hGH is substituted by histidine in the hPL (12-25) variant.

As can be seen, some of the actual substitutions introduced do not correspond to the totality of substitutions indicated by the corresponding segments in FIG.

2

. Thus, hPL (12-25) would contain the four substitutions N12H, R16Q, L20A and F25L if the entire hPL (12-25) segment were substituted. The actual variant made, however, maintained R16 and L20 and therefore incorporated only two of the four substitutions, i.e., N12H and F25L, as shown in Table III. Other segment-substituted variants which maintained one or more resudues of the parent hGH include those covering regions A and E and the segment-substituted variants hPL (46-52) and pGH (167-181).

Each of the segment-substituted human growth hormone variants were assayed in an in vitro system comprising displacement of [

125

I]hGH from the extracellular portion of the cloned soluble hGH receptor to quantify the relative affinities of the segment-substituted variants to the extracellular domain of the somatogenic receptor. Leung, D. W., et al. (1987)

Nature

330, 537. This truncated form of the somatogenic receptor exhibits the same selectivity for hGH as the membrane form of the receptor (Spencer, S. A., et al. (1988)

J. Biol. Chem.

263, 7862) albeit with about a slight reduction in binding affinity (K

d

=0.3 nM).

As will be described in more detail in the examples, selected segments A, C and F, comprising residues 11-19, 54-74 and 164-191, respectively, are active domains in the hGH molecule interactive with the somatogenic receptor. This is based on the observed decrease in Kd of ten-fold or greater for most of the segment-substituted hGH variants containing analogous segments for hGH analogs over these regions. See FIG.

4

. Of course, this does not mean that each of the amino acid residues within these active domains comprises the binding residues for the somatogenic receptor. Rather, such domains define the amino acid sequence within which such active residues can be found.

The active domains A, C and F were further localized. For example, the variant hPRL (12-33) was dissected into the amino and carboxy terminal variants, hPRL (12-19) and hPRL (22-33). The results from this experiment further localized this active domain of hGH to residues 12 through 19. Similarly, the amino terminal portion of region F (pGH (167-181)) exhibits a large reduction in binding affinity. One of the most dramatic effects was the 30-fold reduction in binding caused by hPL (56-64) which introduced only two mutations, E56D and R64M. Although regions A, C and F are widely separated in the primary sequence of hGH, the tertiary folding of the hormone brings them within close proximity. See FIG.

5

. These active domains form a patch that contains the amino terminus of helix

1

(active domain A), the loop from Cys-53 to the start of helix

2

(active domain C) and the central portion of helix

4

(active domain F).

In addition, eight Mabs against hGH were assayed against segment-substituted hGH variants to map the epitopes of hGH. Further, the Mab's were used in a competitive assay with hGH and hGH variants to evaluate the ability of each of the Mabs to block the binding of the hGH receptor to hGH.

The collective results obtained from these experiments provide several lines of evidence that the substitution of analogous segments into hGH do not grossly disrupt the native folding of the molecule and that the observed activity is due to a direct effect on the interaction between the somatogenic receptor and the segment-substituted hGH variants. Firstly, the segment-substituted variants are highly selective in disrupting binding to the somatogenic receptor or the Mabs. Secondly, the somatogenic receptor and Mabs recognize conformation as well as sequence. The receptor and at least four of the Mabs recognize discontinuous epitopes that are sensitive to the protein tertiary structure. Thirdly, circular dichroic spectra of all of the purified variants are virtually identical to wild-type hGH. Fourthly, all of the variants, with the exception of pGH (164-190), were expressed in essentially wild-type amounts. Resistance to proteolysis in vivo has been used as a screen for conformational integrity. Hecht, M. H., et al. (1984)

Proc. Natl. Acad. Sci. USA

81, 5685; Shortle, D., et al. (1985)

Genetics

110, 539.

The alteration in binding activity for segment-substituted hGH variants does not necessarily indicate that the substituted residues in such variants make direct contact with the somatogenic receptor. A disruptive mutation may not only remove a favorable interaction but may introduce an unfavorable one. For example, the N12R mutation in the hPRL (12-19) segment-substituted hGH variant not only changes the hydrogen bonding amide function of Asn12, the Arg substitution also introduces a bulkier side chain that is positively charged. Furthermore, a number of the binding contacts may be conserved between the analogs so that not all contacts, or even regions, may be probed by generating segment-substituted hGH variants. Further, the substitution of analogous segments generates the substitution of multiple amino acid residues in the hGH molecule.

In order to identify the specific active amino acids within the active domains A, C and F in

FIG. 2

, a fine structure analysis of these active domains was performed. In this analysis, residues in these three active domains were replaced sequentially with alanine. A total of 63 single Alanine mutants were made and each of their binding constants were determined for the soluble hGH receptor (shGHr) by Scatchard analysis. Leung, D. W., et al. (1988)

J. Biol. Chem.

263, 7862.

Based on this analysis, the amino acid residues listed in Table IV comprise residues within the hGH molecule which are actively involved in the interaction with the somatogenic receptor. This is based on the more than four-fold effect on the relative dissociation constant caused by the substitution of alanine for these residues as compared to wt hGH. See FIG.

7

. Preferred amino acid substitutions for these residues to form hGH variants are shown.

TABLE IV

Preferred amino

hGH Residue

acid substitution

F10

GEMARQSYWLIV

F54

GEMARQSYWLIV

E56

GMFARQSDNKLH

I58

GEMFARQSVT

R64

GEMFAQSH,KDN

Q68

GEMFARSHKDN

D171

GEMFARQSHKN

K172

GEMFARQSHDN

E174

GMFARQSHDNKL

T175

GEMFARQSVI

F176

GEMARQSYWLIV

R178

GEMFAQSHKDN

C182

GEMFARQS

V185

GENFARQSITLYW

Other amino acid residues which are less active with the somatogenic receptor are listed in Table V. These residues demonstrate generally less than two-fold increase in relative Kd when substituted with alanine.

TABLE V

I4

N12

S55

E66

Q181

P5

M14

S57

K70

R183

L6

L15

P59

S71

G187

S7

R16

S62

K168

R8

R19

N63

I179

Amino acid residues in hGH showing a relative decrease in Kd when substituted with alanine (and consequently greater affinity for the somatogenic receptor) are listed in Table VI.

TABLE VI

P2

E65

S184

T3

Q69

E186

L9

L73

S188

H18

R167

F191

R64

E174

One residue-substituted hGH variant, E174A, surprisingly resulted in a significant decrease (almost five-fold) in the dissociation constant with the somatogenic receptor. This variant, in addition to showing an increased binding affinity for the somatogenic receptor, also exhibited an increased somatogenic potency relative to hGH in a rat weight gain assay. This and other specific residue substitutes that enhance somatogenic binding by >1.4 fold are presented in Table VII.

TABLE VII

hGH variants having enhanced somatogenic binding

Substituted

hGH residues

amino acid

H18

A

R64

K

E65

A

L73

A

E174

A,N,Q,S,G

E186

A

S188

A

F191

A

Other variants containing alanine substitutions not shown in

FIG. 7

are listed in Table VIII.

TABLE VIII

Variant

K

d

(mM)

K

d

(var)/K

d

(wt)

H21A

NE

—

K172A/F176A

201

543

N47A

0.84

2.3

P48A

NE

—

Q49A

0.36

1.0

T50A

0.38

1.0

S51A

Q46A

NE

—

V173A

NE

—

Note:

NE - not expressed in shake flasks at levels which could be easily isolated (i.e., < − 5% of wild-type expression levels).

Once identified, the active amino acid residues for the somatogenic receptor in hGH are analyzed by substituting different amino acids for such residues other than the scanning amino acid used for the preliminary analysis. The residue-substituted variants in Table IX have been made.

TABLE IX

Variant

K

d

(nM)

K

d

(var)/K

d

(wt)

R77V

0.44

1.3

L80D

0.78

2.3

F176Y

3.2

8.6

E174G

0.15

0.43

E174D

NE

—

E174H

0.43

1.2

E174K

1.14

3.1

E174L

2.36

6.4

E174N

0.26

0.7

E174Q

0.21

0.6

E174S

0.11

0.3

E174V

0.28

0.8

E174R

NE

—

R64K

0.21

0.6

E65K

NE

—

E65H

NE

—

K172R

NE

—

I58L

NE

—

F25S

NE

—

D26E

NE

—

Q29S

NE

—

E30Q

NE

—

R178K

NE

—

R178T

NE

—

R178Q

NE

—

I179M

NE

—

D169N

3.6

10.5

Note:

NE - not expressed in shake flasks at levels which could be easily isolated (i.e., <˜ 5% of wild-type expression levels).

In addition to the hGH variants that have been made, Table X identifies specific amino acid residues in hGH and replacement amino acids which are expected to produce variants having altered biological functions.

TABLE X

wT hGH amino

acid residue

Replacement amino acid

S43

GEMFARQHDKN

F44

GEMARQSYWLIV

H18

GEMFARQSKDNY

E65

GMFARQSHDNKL

L73

GEMFARQSIVY

E186

GMFARQSHDNKL

S188

GEMFARQHDNKY

F191

GEMARQSYWLIV

F97

GEMARQSYWLIV

A98

GEMFRQSDNHK

N99

GEMFARQSDKY

S100

GEMFARQHDNKY

L101

GEMFARQSIVY

V102

GEMFARQSITLYW

Y103

GEMFARQSWLIV

G104

EMFARQSP

R19

GEMFAQSHKND

Q22

GEMFARSKKDN

D26

GEMFARQSHKN

Q29

GEMFARSKKDN

E30

GMFARQSHDNKL

E33

GMFARQSHDNKL

In another embodiment, the binding epitope of hGH for the prolactin receptor was determined. hGH can bind to either the growth hormone or prolactin (pRL) receptor. As will be shown herein, these receptors compete with one another for binding to hGH suggesting that their binding sites overlap. Scanning mutagenesis data show that the epitope of hGH for the hPRL receptor consists of determinants in the middle of helix

1

(comprising residues Phe25 and Asp26), a loop region (including Ile58 and Arg64) and the center portion of helix

4

(containing residues K168, K172, E174, and F176). These residues form a patch when mapped upon a structural model of hGH. This binding patch overlaps but is not identical to that C determined for the hGH receptor as disclosed herein and by B. C. Cunningham and J. A. Wells (1989)

Science

244, 1081-1085. By mutating the non-overlap regions of these receptor binding sites on hGH, the preference of hGH was shifted toward the hGH receptor by >2000-fold or toward the hPRL receptor by >20-fold without loss in binding affinity for the preferred receptor. Similarly, by mutating the overlap regions it is possible to reduce binding to both receptors simultaneously by >500-fold. Such receptor selective variants of hGH should be useful molecular probes to link specific receptor binding events to the various biological activities of hGH such as linear growth or lactation.

In a further embodiment, the receptor-binding determinants from human growth hormone (hGH) were placed into the normally nonbinding homolog, human prolactin (hPRL). The alanine scanning mutagenesis disclosed herein and Cunningham, B. C. & Wells, J. A. (1989) Science 244, 1081-1085 identified important residues in hGH for modulating binding to the hGH receptor cloned from human liver. Additional mutations derived from hPRL were introduced into hGH to determine which hPRL substitutions within the hGH receptor binding site were most disruptive to binding. Thereafter, the cDNA for hPRL was cloned and expressed in

Escherichia coli.

It was then mutated to sequentially introduce those substitutions from hGH that were predicted to be most critical for receptor binding. After seven iterative rounds of site-specific mutagenesis, a variant of hPRL containing eight mutations whose association constant was strengthened over 10,000-fold for the hGH receptor was identified. This hPRL variant binds only six-fold weaker than wild-type hGH while sharing only 26% overall sequence identity with hGH. These results show the structural similarity between hGH and hPRL, and confirm the identity of the hGH receptor epitope. More generally, these studies demonstrate the feasibility to borrow receptor binding properties from distantly related and functionally divergent hormones that may prove useful for the design of hybrid hormones with new properties as agonist or antagonist.

The following is presented by way of example and is not to be construed as a limitation to the scope of the invention.

EXAMPLE 1

hGH Mutagenesis and Expression Vector

To facilitate efficient mutagenesis, a synthetic hGH gene was made that had 18 unique restriction sites evenly distributed without altering the hGH coding sequence. The synthetic hGH DNA sequence was assembled by ligation of seven synthetic DNA cassettes each roughly 60 base pairs (bp) long and sharing a 10 bp overlap with neighboring cassettes to produce the 405 bp DNA fragment shown from NsiI to BglII. The ligated fragment was purified and excised from a polyacrylamide gel and cloned into a similarly cut recipient vector, pB0475, which contains the alkaline phosphatase promoter and StII signal sequence (Chang, C. N., et al. (1987)

Gene

55, 189), the origin of replication for the phage f1 and pBR322 from bp 1205 through 4361 containing the plasmid origin of replication and the β lactamase gene. The sequence was confirmed by dideoxy sequence analysis (Sanger, F., et al. (1977)

Proc. Natl. Acad. Sci. USA

74, 5463).

pB0475 was constructed as shown in FIG.

9

. fI origin DNA from filamentous phage contained on a DraI, RsaI fragment 475 bp in length was cloned into the unique PvuII site of pBR322 to make plasmid p652. Most of the tetracycline resistance gene was then deleted by restricting p652 with NheI and NarI, filling the cohesive ends in with DNA polymerase and dNTPs and ligating the large 3850 bp fragment back upon itself to create the plasmid pΔ652. pΔ652 was restricted with EcoRI, EcoRV and the 3690 bp fragment was ligated to a 1300 bp EcoRI, EcoRV fragment from phGH4R (Chang, C. N., et al. (1987)

Gene

55, 189) containing the alkaline phosphatase promoter, STII signal sequence and natural hGH gene. This construction is designated as pB0473. Synthetically derived DNA was cloned into pB0473 in a three-way construction. The vector pB0473 was restricted with NsiI, BglII and ligated to a 240pb NsiI, HindIII fragment and a 1170 bp HindII, BglII fragment both derived from synthetic DNA. The resulting construction pB0475 contains DNA coding for the natural polypeptide sequence of hGH but possesses many new unique restriction sites to facilitate mutagenesis and further manipulation of the hGH gene. The entire DNA sequence of pB0475 together with the hGH amino acid sequence is shown in FIG.

10

. The unique restriction sites in the hGH sequence in pB0475 allowed insertion of mutagenic cassettes (Wells, J. A., et al. (1985)

Gene

34, 315) containing DNA sequences encoding analogous segments from the analogs pGH, hPL and hPRL. Alternatively, the hGH sequence was modified by site specific mutagenesis in the single-stranded pB0475 vector followed by restriction-selection against one of the unique restriction sites (Wells, J. A., et al. (1986)

Philos. Trans. R. Soc. London SerA

317, 415).

The 17 segment-substituted hGH variants in Table III were prepared. Each was secreted into the periplasmic space of

E. coli

at levels comparable to wild-type hGH and at levels that far exceeded the hGH-pGH hybrid described infra. The hGH and hGH variants were expressed in

E. coli

W3110, tonA (ATCC 27325) grown in low phosphate minimal media (Chang, C. N., et al. (1987)

Gene

55, 189).

The hGH and hGH variants were purified as follows. To 200 g of cell paste four volumes (800 ml) of 10 mM TRIS HCl Tris(hydroxymethyl)aminomethane hydrochloride; pH 8.0 was added. The mixture was placed on an orbital shaker at room temperature until the pellets were thawed. The mixture was homogenized and stirred for an hour in a cold room. The mixture was centrifuged at 7000 g for 15 min. The supernatant was decanted and ammonium sulfate was added to 45% saturation (277 g/l) and stirred at room temperature for one hour. After centrifugation for 30 minutes at 11,000 g, the pellet was resuspended in 40 ml 10 mM TRIS HCl Tris(hydroxymethyl)aminomethane hydrochloride; pH 8.0. This was dialyzed against 2 liters of 10 mM TRIS HCl Tris(hydroxymethyl)aminomethane hydrochloride; pH 8.0 overnight. The sample was centrifuged or filtered over a 0.45 micron membrane. The sample was then loaded on a column containing 100 ml of DEAE cellulose (Fast Flow, Pharmacia, Inc.). A gradient of from zero to 300 mM NaCl in 10 mM TRIS HCl Tris(hydroxymethyl)aminomethane hydrochloride pH; 8.0 in 8 to 10 column volumes was passed through the column. Fractions containing hGH were identified by PAGE, pooled, dialyzed against 10 mM TRIS HCl Tris(hydroxymethyl)aminomethane hydrochloride. pH 8.0 overnight. Samples were concentrated to approximately 1 mg/ml by Centri-Prep10 ultrafiltration.

EXAMPLE 2

Homologous Recombinants of hGH and pGH

A random hybrid library containing various N-terminal lengths of hGH linked to the remaining C-terminal portion of porcine growth hormone (pGH) was constructed by the method of random recombination of tandomly linked genes. Gray, G. L., et al. (1986)

J. Bacteriol.

166, 635.

The EcoRI site of pB0475 was removed by restricting the plasmid with EcoRI, filling in the cohesive ends by addition of DNA polymerase and dNTPs, and ligating the plasmid back together. A new EcoRI site was then introduced just following the 3′ end of the hGH gene. This was accomplished by subcloning the 345 bp BglII, EcoRV fragment of hGH-4R, which contains such an EcoRI site, into similarly restricted vector from the EcoRI= pB0475 construction. The pGH gene (Seeburg, P. H., et al. (1983)

DNA

2, 37) was then introduced just downstream and adjacent to the 3′ end of the hGH gene in this construction. This was accomplished by doping an EcoRI, HindIII (filled in) fragment containing pGH cDNA into the large fragment of a EcoRI, EcoRV digest of the construction described above. The resulting plasmid, pB0509, contains an intact hGH gene with a unique EcoRI site at its 3′ end followed by an intact pGH gene reading in the same direction. Due to the homology between the hGH and pGH genes, a percentage of the pB0509 plasmid underwent in vivo recombination, to make hybrid hGH/pGH genes when transformed into

E. coli

rec

+

MM294 (ATCC 31446). These recombinants were enriched by restricting pool DNA with EcoRI to linearize plasmids which had not undergone recombination, resulting in the loss of that EcoRI site. After two rounds of restriction selection and transformation into

E. coli

rec

+

MM294 nearly all the clones represented hybrid hGH/pGH recombinants. Sequence analysis of 22 clones demonstrate that the hGH/pGH hybrids contained amino terminal hGH sequence followed by pGH sequence starting at amino acid residues +19, +29, +48, +94, +105, +123 and +164.

Seven hGH-pGH hybrids having cross-over points evenly distributed over the hGH gene were obtained. However, only the extreme carboxy terminal hybrid (hGH (1-163)-pGH (164-191)) was secreted from

E. coli

at levels high enough to be purified and analyzed. This hGH-pGH hybrid introduces three substitutions (M170L, V173A and V180M) that are located on the hydrophobic face of helix

4

. Accordingly, most of the sequence modifications in the helical regions A, D, E and F in

FIG. 2

were designed to avoid mutations of residues on the hydrophobic face of the helices. For example, the above hybrid hGH-pGH variant was modified to retain M170, V173, F176 and V180 because these residues are inside or bordering the hydrophobic face of helix

4

.

EXAMPLE 3

Expression and Purification of Soluble

Human Growth Hormone Receptor from E. coli

Cloned DNA sequences encoding the soluble human growth hormone receptor shGHr (Leung, D. W., et al. (1987)

Nature

330, 537) were subcloned into pB0475 to form pJ1446 (see FIGS.

11

and

12

).

The vector pC1S.2 SHGHR (Leung, D. W., et al. (1987)

Nature

33, 537) was digested with XbaI and KpnI and the 1.0 kb fragment containing the secretion signal plus the 246 codon extracellular portion of the hGH receptor was purified (Maniatis, T. et al. (1982) in

Molecular Cloning,

Cold Springs Harbor Laboratory, N.Y.). This fragment was ligated into similarly cut M13-mp18 and single-stranded DNA for the recombinant gene was purified (Messing, J. (1983)

Methods in Enzymology,

Vol. 101, p. 20). Site-specific mutagenesis (Carter, P., et al. (1986)

Nucleic Acids Res.

13, 4331) was carried out to introduce an NsiI site at codon +1 using the 18-mer oligonucleotide, 5′-A-AGT-GAT-GCA-TTT-TCT-GG-3′. The mutant sequence was verified by dideoxy sequence analysis (Sanger, F., et al. (1977)

Proc. Natl. Acad. Sci. USA

74, 5463). Double-stranded DNA for the mutant was purified and cut with NsiI and SmaI. The 900 bp fragment was isolated containing the 246 codon extracellular portion of the hGH receptor. pB0475 was cut with NsiI and EcoRV and the 4.1 kb fragment (missing the synthetic hGH gene) was purified. The 900 bp fragment for the receptor and the 4.1 kb vector fragment were ligated and the recombinant clone (pJ1446) was verified by restriction mapping. This was transformed into the

E. coli

KS303 (Strauch, K., et al. (1988)

Proc. Natl. Acad. Sci. USA

85, 1576) and grown in low-phosphate media (Chang, C. N. (1987)

Gene

55, 189) at 30° C. The receptor fragment protein was purified by hGH affinity chromatography (Spencer, S. A., et al. (1988)

J. Biol. Chem.

263, 7862; Leung, D. W., et al. (1987)

Nature

330, 537). The sequence for pJ1446 is shown in

FIG. 12

together with the amino acid sequence of the cloned receptor.

E. coli

W3110, degp (Strauch, K. L., et al. (1988)

PNAS USA

85, 1576) was transformed with pJ1446 and grown in low-phosphate media (Chang, C. N. (1987)

Gene

55, 189) in a fermentor at 30° C. The 246 amino acid hGHr was used to generate preliminary data. A slightly shorter hGHr containing amino acids 1 through 238 was used in the examples herein. The results obtained with that receptor were indistinguishable from those obtained with the 246 amino acid hGHr.

The plasmid phGHr(1-238) (Table X(A)) was constructed to generate a stop codon after Gln238 to avoid the problem of carboxyl terminal heterogeneity. The binding protein from KS330 cultures containing phGHr(1-238) was produced in slightly higher yields and with much less heterogeneity (data not shown) than frog cultures containing phGHr(1-246). Routinely, 20 to 40 mg of highly purified binding protein could be isolated in 70 to 80 percent yield starting from 0.2 kg of wet cell paste (−2 liters high cell density fermentation broth). Both N-terminal sequencing and peptide mapping coupled to mass spectral analysis of the C-terminal peptide confirmed that the product extended from residues 1 to 238.

Site-directed mutagenesis of the phGHr (1-246) template was performed (Carter, et al. (1986)

Nucleic Acids Res.

13, 4431-4443) to produce phGHr (1-240, C241R) using the oligonucleotide

* * * *

5′-ATG-AGC-CAA-TTT-

ACG-CGT

-TAG-GAA-GAT-TTC-3′;

the asterisks are mismatches from the phGHr (1-246) template, underlined is a new unique MluI site) and CGT-TAG directs the C241R mutation followed by a stop codon (Table X(A)).

TABLE X(A)

Sequences of amino- and carboxyl-termini of hGH

binding protein constructions

Plasmid

Termini

Protein/DNA sequence/Restriction sites

−3 −2 −1 +1 +2 +3

phGHr(1—246)

Amino

ALA-TYR-ALA-PHE-SER-GLY

GCC-T

AT-GCA-T

TT-TCT-GGA

NsiI

phGHr(1—246)

Carboxyl

238 239 240 241 242 243 244 245 246

GLN-PHE-THR-CYS-GLU-GLU-ASP-PHE-TYR-AM

CAA-TTT-ACA-TGT-GAA-GAA-GAT-TTC-TAC-TA

G-CGGCCGC

NotI

phGHr

Carboxyl

Gln-Phe-Thr-Arg-AM

(1—240,C241R)

* ** * *

CAA-TTT-

ACG-CGT

-TAG-GAA-GAT-TTC-TAC-TA

G-CGGCCGC

MluI NotI

phGHr(1—238)

Carboxyl

Gln-AM

** ** * *

CAA-TAG-ACA-CGT-TAG-GAA-GAT-TTC-TAC-TA

G-CGGCCGC

NotI

*Indicates mismatches frow the wild-type template

The plasmid, phGHr (1-238), was produced by site-directed mutagenesis on the phGHr (1-240, C241R) template using restriction-selection (Wells, et al., (1986)

Phil. Trans. R. Soc. Lond. A,

317, 415-423) against the MluI site (Table X(A)). Briefly, an oligonucleotide,

* *

5′-AG-ATG-AGC-CAA-TAG-

ACA-CGT

-TAG-GAA-3′

introduced a translation stop codon after Gln238 (CAA triplet) and altered the MluI restriction-site (underlined). After growing up the pool of duplex DNA from the initial transfection with heteroduplex, the DNA was restricted with MluI and retransformed to enrich for the desired phGHr (1-238) plasmid prior to DNA sequencing.

It was subsequently determined by DNA sequencing that the cloned hGH binding proteins in phGHr(1-238) contained a T51A mutation which arose either as a cDNA variant or as a cloning artifact. The A51T revertant was therefore to be identical to the published sequence (Leung, et al., (1987)

Nature

(

London

) 330, 537-543). The purification and binding properties of the proteins containing either Thr or Ala at position 51 were indistinguishable (results not shown). The Ala5l binding protein variant was selected for all subsequent analysis because it had been characterized more thoroughly.

To compare the specificity of the recombinant hGH binding protein from

E. coli

with the natural product isolated from human serum, the affinities were determined for wild-type and various hGH mutants:

TABLE X(B)

K

d

(nM) + S.D.

a

for hGH binding protein from:

K

d

(mut)

b

K

d

(mut)

b

K

d

(human serum)

b

hGH mutant

Human serum

K

d

(wt)

E. coli

K

d

(wt)

K

d

(

E. coli

)

wt

0.55 ± 0.07

—

0.40 ± 0.03

—

1.4

I58A

21 ± 2

38 ± 6

14 ± 1

36 ± 5

1.5

R64A

12 ± 1

22 ± 4

11 ± 1

28 ± 5

1.1

E174A

0.27 ± 0.04

0.49 ± 0.11

0.16 ± 0.01

0.4 ± 0.1

1.7

F176A

71 ± 7

130 ± 20

48 ± 5

120 ± 20

1.5

a

Values of K

d

and corresponding standard deviations (SD) were determined by competitive binding analysis (

FIG. 24

) with wild-type hGH (wt) and a number of mutants of hGH.

b

Reduction in binding affinity calculated from the ratio of dissociation constants for the hGH mutant (mut) and wild-type hGH for each hGH binding protein.

c

Ratio of dissociation constants for the two hGH binding proteins with a given hGH type.

Both proteins formed a specific stoichiometric complex with hGH (FIG.

24

). As can be seen, the affinities for wild-type and mutants of hGH are nearly identical between the two binding proteins (right side column, supra). The recombinant hGH binding protein has a marginally higher affinity compared to the natural protein from human serum. This may reflect the greater purity and homogeneity of the recombinant protein. Both proteins had identical specificities as shown by the changes in binding affinities for four alanine mutants of hGH that disrupt binding to the hGH binding protein (K

d

(mut)/K

d

(wt) supra) The affinity of hGH for the binding protein extending to Tyr246 (K

d

=0.36±0.08 nM) was virtually identical to that terminating after Gln238 (0.40±0.03 nM) indicating the last 8 residues (including the seventh cysteine in the molecule) are not essential for binding hGH.

EXAMPLE 4

Receptor and Monoclonal Antibody Binding Assay

Purified hGH or hGH variants (over 95% pure) were assayed for binding to the soluble hGH receptor of Example 3. Laser densitometric scanning of Coomassie stained gels after SDS-PAGE was used to quantitate the concentration of the purified hormones. These values were in close agreement with concentrations determined from the absorbance at 280 nm (ε2

80

0.1%=

0.93). The dissociation constants (K

d

) were calculated from Scatchard analysis for competitive displacement of [

1251

] hGH binding to the soluble gGH receptor at 25° C. The

125

I hGH was made according to the method of Spencer, S. A., et al. (1988)

J. Biochem.

263, 7862.

An enzyme-linked immunosorbent assay (ELISA) was used to assess the binding of eight different monoclonal antibodies to various segment-substituted and residue-substituted hGH variants. The following are the Mabs used:

Mab

Identity

Source/Method

1

MabA

(*)

2

33.2

Hybritech, Inc.

3

Cat# H-299-01

Medix Biotech, Inc.

4

72.3

Hybritech, Inc.

5

Cat# H-299-02

Medix Biotech, Inc.

6

Mab 653

Chemicon

7

Mab D

(*)

8

Mab B

(*)

(*) Carbone, F. R., et al. (1985) J. Immunol. 135, 2609

Rabbit polyclonal antibodies to hGH were affinity purified and coated onto microtiter plates (Nunc plates, InterMed, Denmark) at 2 μg/mL (final) in 0.005 M sodium carbonate pH (10) at 24° C. for 16-20 h. Plates were reacted with 0.1 μg/mL of each hGH variant in buffer B (50 mM TRIS HCl Tris(hydroxymethyl)aminomethane hydrochloride [pH 7.5], 0.15 M NaCl, 2 mM EDTA, 5 mg/mL BSA, 0.05% TWEEN 20™ brand polyoxyethylene sorbitanmonolaurate 0.02% sodium azide) for two hours at 25° C. Plates were washed and then incubated with the indicated Mab which was serially diluted from 150 to 0.002 nM in buffer B. After two hours plates were washed, stained with horseradish peroxidase conjugated anti-mouse antibody and assayed. Values obtained represent the concentrations (nM) of each Mab necessary to produce half-maximal binding to the respective hGH variant.

Competitive displacement of the hGH receptor from hGH by anti-hGH Mabs was determined as follows. Assays were carried out by immobilization of wild-type hGH in microtiter plates coated with anti-hGH rabbit polyclonal antibodies as described. Receptor (fixed at 10 nM) and given anti-hGH Mab (diluted over a range of 150 to 0.002 nM) were added to the hGH coated microtiter plate for 16-20 hours at 25° C., and unbound components were washed away. The amount of bound receptor was quantified by adding an anti-receptor Mab that was conjugated to horseradish peroxidase which did not interfere with binding between hGH and the receptor. The normalized displacement value was calculated from the ratio of the concentration of Mab necessary to displace 50% of the receptor to the half-maximal concentration of Mab necessary to saturate hGH on the plate. This value was used to compare the relative ability of each Mab to displace the receptor.

EXAMPLE 5

Active Domains for Somatogenic Receptor Binding

The 17 segment-substituted hGH variants described in Example 1 and Example 2 were assayed for binding to the soluble somatogenic receptor of Example 3 and binding to the monoclonal antibodies as described in Example 4. The results of the binding assay to the somatogenic receptor are shown in Table III. As can be seen, the segment substitutions that are most disruptive to binding are within regions A, C and F of

FIGS. 4 and

5. These regions were further directed into smaller segments to further localize the active domains of the hGH molecule involved in binding to the somatogenic receptor. The most significant results from Table III are shown in

FIG. 4

, which is a bar graph showing the relative reduction in binding to the soluble hGH receptor as a consequence of the substitution of the indicated analogous sequences from the analogs hPRL, hPL and pGH as shown. Three active domains were identified as regions A, C and F comprising amino acid residues 12-19, 54-74 and 164-190) respectively. These regions are identified in the three-dimensional representation of the hGH molecule in FIG.

5

.

As can be seen, the three active domains, A, C and F, although discontinuous in the amino acid sequence of hGH, form a continuous region in the folded molecule which defines the somatogenic binding site on hGH.

EXAMPLE 6

Epitope Mapping of hGH

The binding of the eight different monoclonal antibodies to specific segment-substituted hGH variants is shown in Table XI.

TABLE XI

Mab

2

3

4

5

1

Hybr

Medix

Hybr

Medix

6

7

8

hGH Variant

MCA

33.2

1

72.3

2

Chemicon

MCD

MCB

wt hGH

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

hPL(12-25)

0.4

0.4

>75

>50

0.2

0.2

0.08

0.1

pGH(11-33)

0.4

>100

1.5

0.05

0.2

0.2

0.08

0.1

hPRL(12-33)

0.4

>100

>75

>50

0.2

0.2

0.08

0.1

hPRL(12-19)

0.4

>12

>75

>50

0.2

0.2

0.08

0.1

hPRL(22-33)

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

hPL(46-52)

0.4

0.4

0.1

0.05

0.2

0.2

0.40

0.1

pGH(48-52)

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

hPL(56-64)

0.4

0.4

0.1

0.05

0.2

0.8

0.08

0.1

pGH(57-73)

0.4

0.4

0.1

0.05

>200

>200

0.08

0.1

hPRL(54-74)

0.4

0.4

0.1

0.05

0.2

0.6

0.08

0.1

hPRL(88-95)

>400

0.4

0.1

0.05

0.2

0.2

0.08

0.1

hPRL(97-104)

>400

>12

0.1

0.05

0.2

0.2

0.08

0.1

hPL(109-112)

>12

0.4

>75

15

0.2

0.2

0.08

0.1

hPRL(111-129)

>12

0.4

>75

>50

0.2

0.2

0.08

0.1

hPRL(126-136)

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

pGH(164-190)

0.4

0.4

0.5

0.3

>25

12.5

0.20

0.4

pGH(167-182)

hGH(Δ32-46)

0.4

0.4

0.1

0.05

0.2

0.2

>100

>100

N12A

0.4

0.4

>75

>50

0.2

0.2

0.08

0.1

C182A

0.4

0.4

0.1

0.05

2.0

0.2

0.08

0.1

With the possible exception of the pGH (167-190) variant, disruption of binding to each monoclonal antibody was dramatic and highly selective.

FIGS. 13 through 20

localize the epitope for each of the Mabs on the three-dimensional structure of hGH.

FIG. 6

comprises these epitopes to the binding site for the somatogenic receptor.

For example, the hPRL (88-95), hPRL (97-104), hPL (109-112) and hPRL (111-129) variants do not bind to Mab

1

, yet the other segment-substituted hGHs outside of these regions bound as effectively as wild-type hGH. Binding to Mabs

2

,

3

,

4

,

5

and

6

was disrupted by mutations in discontinuous regions in the primary sequence but in close proximity in the folded hormone (see

FIGS. 6 and 14

through

19

). In contrast, Mabs

1

,

7

and

8

were disrupted by mutations defined by a continuous sequence as shown in

FIGS. 13

,

19

and

20

.

The regions disrupting binding to a given monoclonal antibody were further analyzed by dissecting specific segment-substituted hGH variants into subdomains or by analyzing variants that had common substitutions that still bound to the particular Mab. For example, pGH (11-33) retained tight binding to Mab

4

yet hPRL (12-33) disrupted binding. Thus, the disruptive mutations in the hPRL (12-33) variant can be confined to residues not mutated in pGH (11-33): N12, L15, R16, D26 and E30. This set can be further restricted to N12, L15 and R16 because the hPRL (12-19) variant disrupts binding, but the hPRL (22-23) variant does not (see FIG.

16

). The N12H mutation in hPL (12-25) can entirely account for the disruption in binding to Mab

4

because this is the only mutation not in common with pGH (11-33). This was tested by substituting alanine for Asn-12. The binding of Mabs

3

or

4

to the N12A residue-substituted hGH variant was reduced by over 100-fold whereas binding to the other Mabs was unaffected.

Using this set of hGH variants, it was possible to resolve the epitopes from all eight Mabs even though binding for most of these Mabs was blocked by a common set of mutations. For example, although hPRL (12-19) disrupted binding to Mabs

2

,

3

and

4

, other variants indicated that these Mabs recognized different structures. Specifically, Mabs

2

and

3

were blocked by pGH (11-33) yet Mab

4

was unaffected. Binding of Mabs

3

and

4

was blocked by hPL (12-25) yet binding to Mab

2

was unaffected. Thus, the eight antibodies may have epitopes that overlap but none superimposed. Mutations that disrupt binding are present in both helices and loops and are always in close proximity in the folded hormone.

Collectively, the epitopes with a set of eight Mabs cover most of the hormone. However, there are still regions where these Mabs did not bind. For example, three of the 20 variants did not significantly disrupt binding to any of the Mabs tested (hPRL (22-33), pGH (48-52) and hPRL (126-136)).

There are significant differences between the antibody epitopes and the receptor binding site. Firstly, the patch defined by disruptive mutations is larger for the receptor than for any of the Mabs. A second difference is that the receptor has more tolerance to disruptive substitutions in the hormone than do the Mabs. This is evidenced by the fact that the maximum reduction in binding to the receptor for any of the mutants is about 70-fold, whereas almost every antibody has at least one variant that causes more than a 1000-fold reduction in binding, some of which may be the result of single substitutions such as N12A.

EXAMPLE 7

Competitive Binding of Mabs and shGHr

Many of the variants which cause disruption of receptor binding also disrupt the binding of one or more of the Mabs. The ability of each of the eight Mabs to block the binding of the hGH receptor to hGH was therefore evaluated. Results of this assay are shown in Table XII.

TABLE XII

Mab

50% binding to hGH↑

displace 50% of receptor

&AutoRightMatch; [\frac{\begin{matrix} Normalized  displacement \\ conc.  for  50%  displacement \end{matrix}}{conc.  for  50%  binding}]

1

0.4

>150

>375

2

0.4

0.8

2

3

0.1

150

1500

4

0.05

150

3000

5

0.2

0.2

1

6

0.2

0.2

1

7

0.08

0.4

5

8

0.1

>150*

>1500

*Binding of Mab 8 appears to slightly enhance binding of receptor to hGH.

↑Data from Table X for binding of each Mab to hGH.

As can be seen Mabs

5

and

6

are the most efficient at blocking binding of the hGH receptor. This is because these Mabs have antigenic determinants located in the loop from residues 54 through 74 and in helix

4

, closely overlap determinants for the receptor (see

FIGS. 5

,

6

,

17

and

18

). Mab

2

was the next most competitive antibody and it too shared a common disruptive mutation with the receptor (hPRL (12-19)). In contrast, Mabs

3

and

4

were roughly 1000-fold less competitive than Mab

2

yet they also shared overlapping disruptive mutations with the receptor in helix

1

. See

FIGS. 15 and 16

. This apparent discrepancy may be easily reconciled if the mutations in helix

1

that disrupt Mabs

3

and

4

differ from those residues which disrupt binding to Mab

2

or the receptor. Indeed, one such mutant (N12A) disrupts binding of either Mab

3

or

4

without affecting binding to Mab

2

or the receptor.

Mab

7

competes relatively strongly with the receptor for hGH and it is disrupted by segment-substituted hGH variants that cause a minor disruption of receptor binding, e.g., hPL (46-52). Thus, it appears that Mabs

2

and

7

sit on the border of the receptor binding site. Mabs

1

and

8

were unable to give detectable displacement of the receptor, and as expected these contain no overlapping antigenic determinants with the receptor. These competitive binding data taken together with the direct epitope mapping and receptor binding data strongly support the general location of the receptor binding site as shown in FIG.

5

.

EXAMPLE 8

Receptor Active Amino Acid Residues

The analysis of hGH in Examples 5, 6 and 7 implicates the amino terminal portion of helix

1

(residues 11-19) as being of moderate importance to receptor binding. In addition, residues 54-74 and 167-191 were identified as being important to receptor binding. Identification of which amino acids in these domains are active in receptor binding was carried out by analyzing a total of 63 single alanine variants. See Tables XIII, XIV and XV.

TABLE XIII

Amino acid scanning of positions 2-19 in hGH

Variant

K

d

(nM)

K

d

(variant)/K

d

(wt)

wt

0.34

1.0

P2A

0.31

0.90

T3A

0.31

0.90

I4A

0.68

2.0

P5A

0.71

2.1

L6A

0.95

2.8

S7A

0.61

1.8

R8A

0.48

1.4

L9A

0.32

0.95

F10A

2.0

5.9

D11A

NE

—

N12A

0.40

1.2

A13 (WT)

M14A

0.75

2.2

L15A

0.44

1.3

R16A

0.51

1.6

A17 (WT)

H18A

0.24

0.71

R19A

0.37

1.1

TABLE XIV

Amino acid scanning of positions 54-74 in hGH

Variant

K

d

(nM)

K

d

variant/K

d

WT

WT

0.31

1.0

F54A

1.5

4.4

S55A

0.41

1.2

E56A

1.4

4.1

S57A

0.48

1.4

I58A

5.6

17.0

P59A

0.65

1.9

T60A

NE

—

P61A

NE

—

S62A

0.95

2.8

N63A

1.12

3.3

R64A

7.11

21.0

E65A

0.20

0.6

E66A

0.71

2.1

T67A

NE

—

Q68A

1.8

5.2

Q69A

0.31

0.9

K70A

0.82

2.4

S71A

0.68

2.0

N72A

NE

—

L73A

0.24

0.70

E74A

NE

—

TABLE XV

Amino acid scanning of positions 167-191 in hGH

Variant

K

d

(nM)

K

d

variant/K

d

WT

WT

0.34

1

R167A

0.26

0.75

K168A

0.37

1.1

D169A

NE

—

M170A

NE

—

D171A

2.4

7.1

K172A

4.6

14

V173A

NE

—

E174A

0.075

0.22

T175A

NE

—

T175S

5.9

16

F176A

5.4

16

L177A

NE

—

R178A

NE

—

R178N

1.4

4.2

I179A

0.92

2.7

V180A

0.34

1.0

Q181A

0.54

1.6

C182A

1.9

5.7

R183A

0.71

2.1

S184A

0.31

0.90

V185A

1.5

4.5

E186A

0.27

0.80

G187A

0.61

1.8

S188A

0.24

0.7

C189A

NE

—

G190A

NE

—

F191A

0.20

0.60

The substitution of alanine was extended to include residues 2-19 because of uncertainties in the position of the amino-terminal residue (Abdel-Meguid, S. S., et al. (1987)

Proc. Natl. Acad. Sci. USA

84, 6434). Indeed, the most pronounced reduction in binding occurred for F10A (6-fold) followed by alanine substitutions at residues 4-6 at the N-terminus of helix

1

(see FIG.

21

). Substantially larger effects on binding (greater than 20-fold) occurred for specific alanine substitutions within the-54 to 74 loop and the carboxy, terminal sequence 167-191. For several alanine variants, binding was enhanced up to 4.5-fold. The most dramatic example was E174A, which was located in the midst of a number of disruptive alanine mutations. See

FIG. 4

,

7

and

21

.

The most disruptive alanine substitutions form a patch of about 25 Å by 25 Å on the hormone that extends from F10 to R64 and from D171 to V185 (see FIG.

21

). Furthermore, these side chains appear to be facing in the same direction on the molecule. For example, all of the alanine mutants that most affect binding on helix

4

(D171A, K172A, E174A, F176A, I179A, C182A and R183A) are confined to three and one-half turns of this helix, and their side chains project from the same face of the helix (see FIG.

21

). Based upon this model, it was predicted that T175 and R178 should be involved in binding because they occupy a central position as shown in FIG.

21

.

Although the T175A mutant could not be expressed in high enough yields in shake flasks to be assayed, a more conservative mutant (T175S) was. Accordingly, the T175S mutant caused a 16-fold reduction in receptor binding. Similarly, although R178A was poorly expressed, R178N could be expressed in yields that permitted analysis. R178N exhibited a greater than four-fold reduction in binding affinity.

The next most disruptive mutant in the carboxy terminus was V185A. Although V185A is outside of helix

4

, it is predicted by the model to face in the same direction as the disruptive mutations within helix

4

. In contrast, alanine mutations outside the binding patch, or within it facing in the opposite direction from those above (R167A, K168A, V180A, Q181A, S184A, E186A, S188A), generally had no or little effect on receptor binding.

A similar analysis applied to alanine mutants in helix

1

, albeit with more moderate effects on binding. Within the helix, the alanine substitutions that most disrupted binding were at residue 6, 10 and 14 which were located on the same face of the helix. The least disruptive alanine mutations (L9A, N12A and L15A) were located on the opposite face of helix

1

. This is further confirmed by the fact that anti-hGH Mabs

3

and

4

, which do not compete with the receptor for binding to hGH, both bind to Asn-12. See Table XVI.

TABLE XVI

Binding of hGH and alanine variants to eight

different anti-hGH monoclonal antibodies (Mab).

Mab

Hormone

1

2

3

4

5

6

7

8

hGH

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

F10A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

N12A

0.4

0.4

>75

>50

0.2

0.2

0.08

0.1

I58A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

R64A

0.4

0.4

0.1

0.05

0.2

1.6

0.08

0.1

Q68A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

K168A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

D171A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

K172A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

E174A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

F176A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

C182A

0.4

0.4

0.1

0.05

2.0

0.2

0.08

0.1

V185A

0.4

0.4

0.1

0.05

0.2

0.2

0.08

0.1

The relative positions of side chains within the 54-74 loop cannot be fixed in the model as they can be for those within helices

1

and

4

. However, there is a striking periodicity in the binding data in which mutations of even-numbered residues cause large reductions in binding relative to odd-numbered residues. This is especially true for the first part of this region (54-59) and may reflect a structure in which even residues project toward the receptor and odd ones away.

EXAMPLE 9

Conformational Integrity and Binding Energetics of Alanine-Substituted hGH Variants

Several lines of evidence indicate that the alanine substitutions that disrupt the receptor binding do not do so by causing the molecule to be misfolded. Firstly, the eight Mabs react as well with almost all of the alanine mutants that disrupt binding to the receptor as they do with hGH. See Table XII supra.

The exceptions are R64A and C182A which selectively disrupt binding to the anti-hGH Mabs

6

and

5

, respectively. These two Mabs as previously indicated compete with the somatogenic receptor for binding to hGH. In addition, two alanine variants were made which do not affect receptor binding. One of these affects the binding of two Mabs (N12A) and the other affects none of the Mabs (K168A). This data indicates that binding to either the Mabs or receptors is disrupted by a very local perturbation in the structure of the variant. Moreover, the far UV circular dichroic spectra of all the hGH variants tested are virtually identical to wild-type hGH.

About 20% of the alanine mutants (D11A, T60A, P61A, T67A, N72A, E74A, D169A, M170A, V173A, T175A, L177A, K178A, C189A, G190A) were not secreted at high enough levels in shake flasks to be isolated and analyzed. Since genes encoding such variants were expressed in the same vector and expression was independent of the specific alanine codon, variations in steady-state expression levels most likely reflect differences in secretion level and/or proteolytic degradation of the hGH variants. Several of the non-expressing alanine variants in helix

4

are located on its hydrophobic face (M170A, V173A and L177A) as shown in

FIG. 21

wherein the hydrophobic side of the helix is identified by open shading. However, this is not a general effect because several alanine substitutions were tolerated on the hydrophobic face of helix

1

(L6A, L9A and F10A) and helix

4

(F176A and V180A).

In addition, impaired expression of hGH variants was sometimes observed when charged or neutral amino acids were replaced with alanine (D11A, T60A, T67A, N72A, E74A, D169A, T175A, R178A). Mutations such as T175S and R178N, which preserved the hydrogen bonding group at those sites, could be expressed albeit at levels below wild-type. The non-expressing C189A variant disrupts the carboxy-terminal disulfide and its counterpart (C182A) was also expressed at levels far below wild-type. Several other non-expressing alanine mutants (T60A, T61A and T67A) were located in a loop structure. Thus, low levels of expression or non-expression can result from a multitude of structural effects but can be obviated by isosteric or isofunctional substitutions.

The substitutions that cause a ten fold or greater effect upon binding (I58A, R64A, K172A, T175S, F176A) are likely to be directly involved in binding. The strengths of hydrogen bonds or salt bridges present in nature (Fersht, A. R. (1972)

J. Mol. Biol.

64, 497; Brown, L. R., et al. (1978)

Eur. J. Biochem.

88, 87; Malivor, R., et al. (1973)

J. Mol. Biol.

76 123) or engineered by site-directed mutagenesis experiments (Fersht, A. R., et al. (1985)

Nature

314, 225; Bryan, P., et al. (1986)

Proc. Natl. Acad. Sci USA

83, 3743; Wells, J. A., et al. (1987)

Proc. Natl. Acad. Sci USA

84, 5167; Wells, J. A., et al. (1987)

Proc. Natl. Acad. Sci. USA

84, 1219; Cronin, C. N., et al. (1987)

J. Am. Chem. Soc.

109 2222; Graf, L., et al. (1988)

Proc. Natl. Acad. Sci. USA

85 4961) overlap and range widely from 1 to 5 kcal/mole depending upon the microenvironment. For hGH, reductions in binding free energy of 0.8, 1.0, 1.2, 1.6 and 1.8 kcal/mol (ΔΔG

binding

=+RT in Kd (var)/Kd(wt)) occurred for alanine substitutions at E56, Q68, D171, K172 and R64, respectively. The energetics for burial of a hydrophobic side chain into a protein tends to parallel its free energy of transfer into ethanol (Estell, D. A., et al. (1986)

Science

233, 659; Nozaki, Y. et al. (1980) in The Hydrophobic Effect (Wiley, N.Y. pp. 4-21)).

Accordingly, the reductions in binding free energies for F175A, F10A, F54A, I58A, and V185A were 1.6, 1.0, 0.9, 1.7 and 0.9 kcal/mol, respectively. These are slightly below the predicted change in hydrophobic free energy in going from Phe, Ile or Val to Ala of 2.0, 2.4 and 1.0 kcal/mol, respectively. By this analysis the effect of the T175S mutant (ΔΔG

binding

=1.6 kcal/mol) is larger than expected for loss of a γ methyl groups (ΔΔG

hydrophobic

=0.7 kcal/mol). To fully characterize the nature of the molecular contacts between hGH and its somatogenic receptor requires direct structural information. However, the energetics of binding of these alanine mutants shows them to be in the range of previous measurements made on contact residues in entirely different systems. In fact, the sum of binding free energies for these alanine-substituted variants exclusive of C182A that are most disruptive to receptor binding (−13.2 kcal/mol.) is comparable to the total free energy binding between hGH and its receptor (−13 kcal\mol).

EXAMPLE 10

Reactivity of hGH Variants with Anti-hGH Polyclonal Antibodies

The hGH variants hPRL (22-33), E174A and hPRL (88-95) were tested in a rat weight gain assay. The results of that assay are presented in FIG.

22

. As can be seen, all the variants except hPRL (22-33) have a reduced potency after about 14 days of growth. The leveling off of growth is attributed to the development of antibodies to the various growth hormones which neutralize the biological effect. The fact that the hPRL (22-33) variant continues to induce growth suggests that it is not as immunogenic as wild-type hGH or the other variants used.

A comparison of the reactivity of various hGH variants with human and murine serum containing polyclonal antibodies to hGH is shown in Table XVII.

TABLE XVII

Serum Anti-hGH Antibodies Binding to hGH Variants

Average % of Reduction

of Anti-PROTROPIN ® brand hGH

Binding ± SD

% Incidence

Human

Mouse

Human

Mouse

Sera N = 22

Sera (N = 6)

Sera

Sera

hGH

0

0

100

100

pGH 11-33

86 ± 13

65 ± 16

100

100

hPRL 12-33

79 ± 19

52 ± 13

100

100

hPL 12-25

35 ± 19

16 ± 11

81

33

hPRL 12-19

29 ± 20

11 ± 12

71

33

hPRL 22-33

69 ± 15

38 ± 8

100

100

hPL 46-52

6 ± 8

2 ± 4

10

0

pGH 48-52

7 ± 8

4 ± 4

10

0

pGH 57-73

43 ± 15

39 ± 12

95

100

hPRL 54-74

14 ± 9

8 ± 7

24

0

D80

13 ± 15

7 ± 7

14

0

hPRL 88-95

14 ± 22

4 ± 5

19

0

hPL 109-112

10 ± 11

9 ± 9

24

17

hPRL 126-136

8 ± 12

2 ± 2

19

0

C182A

1 ± 5

1 ± 3

5

0

As can be seen, variants containing substitutions within the region from residues 22 to 33 have substantially reduced binding activity, and in some cases no activity, with individual human and mouse anti-serum for wild-type hGH.

Except for the variant pGH 57-73, variants containing substitutions in the other regions shown do not have a significant reduction in reactivity. Since the segment-substituted mutants between residues 11 and 33 retain their ability to bind the somatogenic receptor, such variants demonstrate the production of variants which maintain the ability to promote somatogenesis but have another property which is modified, in this case reactivity with anti-hGH polyclonal antibodies.

EXAMPLE 11

Relationship Between K

d

and Potency

A semi-log plot of the ratio of K

d

(variant)/K

d

(wild type) for specific hGH variants versus the potency of such variants in a rat weight gain assay is shown in FIG.

23

. As can be seen a linear relationship exists which suggests that a decreased-binding affinity for the somatogenic receptor will result in a decrease in potency.

As can be seen, the hGH variant E174A has a higher binding affinity for the somatogenic receptor than the wild-type hGH. Its potency is also greater than that of wild-type hGH by about 12%.

Further, the variant pPRL (97-104) has essentially the same binding constant as wild-type hGH but about a 2.7-fold increase in potency.

EXAMPLE 12

Active Domains in hGH for Prolactin Receptor Binding

Human growth hormone (hGH) elicits a myriad of physiological effects including linear growth, lactation, nitrogen retention, diabetogenic and insulin-like effects, and macrophage activation. R. K. Chawla, J. S. Parks and D. Rudman, Annu. Rev. Med. 34, 519-547 (1983); O. G. P. Isaksson, et al. (1985)

Annu. Rev. Physiol.

47, 483-499; C. K. Edwards, et al., (1988)

Science

239, 769-771. Each of these effects begins with the interaction of hGH with specific cellular receptors. J. P. Hughs, et al. (1985)

Annu, Rev. Physiol.

47, 469-482. Thus far, the only cloned genes whose products bind hGH are the hGH receptor from liver (D. W. Leung, et al., (1987)

Nature

(London) 33, 537-543) and the human prolactin (hPRL) receptor from mammary gland (J. M .Boutin, et al., (1988)

Cell

53, 69-77). Receptor “spillover” of hGH onto the hPRL receptor has clinical precedence in cases where acromegalics, who produce high levels of hGH, develop a hyperprolactinemic syndrome despite having normal levels of hPRL (J. E. Fradkin, et al., (1989)

New Engl. J. Med.

320, 640-644). However, other receptors exist that bind hGH, including the placental lactogen (PL) receptor (M. Freemark, et al., (1987)

Endocrinology

120, 1865-1872). It previously was not known if the binding sites on hGH for these receptors are identical or which receptor (or combination of receptors) is responsible for which pharmacological effect. To begin to address these issues the hGH and hPRL receptor binding sites on hGH were mapped. The results obtained indicate that these receptor binding sites overlap but are not identical. This has allowed the rational design of receptor-specific variants of hGH.

The hGH and hPRL receptors both contain an extracellular hormone binding domains that share 32% sequence identity, single transmembrane domains, and cytoplasmic domains which differ widely in sequence and length. The extracellular binding domain of the hGH receptor has been expressed in

E. coli

and has identical binding properties to that found naturally as a soluble serum binding protein (S. A. Spencer, et al., (1988)

J. Biol. Chem.

26, 7862-7867). Similarly, the extracellular domain of the hPRL receptor has been expressed in

E. coli

and purified. The hPRL receptor fragment extends from residues Gln1 to Thr211 and terminates just before the single transmembrane domain. It retains high binding affinity and specificity that is virtually identical to its full-length receptor. The gene encoding the hPRL receptor used in the experiments was kindly provided by Dr. P. A. Kelly, Laboratory of Molecular Endocrinology, McGill University, Montreal, Canada. This DNA sequence was obtained from a human mammary cDNA library and identified with a probe covering known conserved regions amongst cross-species members of the prolactin receptor family. See e.g., Davies, J. A., et al., (1989)

Mol. Endrocrinology

3, 674-680; Edery, et al. (1989)

Proc. Natl. Acad. Sci. USA

86 2112-2116; Jolicoeur, et al. (1989)

Mol. Endrocinology

3, 895-900. These truncated and highly purified receptors are extremely useful reagents for rapid and accurate assessment of binding affinity for mutants of hGH.

Relationship Between hPRL and hGH Receptor Binding Sites.

To determine if the epitopes for the hGH and hPRL receptors overlapped we analyzed whether or not the hPRL receptor fragment could displace the hGH receptor fragment from hGH (results not shown). Indeed, the hPRL receptor fragment competed for the hGH receptor binding site with an apparent Kd of 1 nM. This is virtually the same affinity as that measured by direct binding of the hPRL receptor to hGH (results not shown).

Eleven of the segment-substituted hGH variants from Table III were used to localize the epitope on hGH for the hPRL receptor. The hGHΔ32-46 variant was also used in this experiment. The approach was similar to that used to determine the epitope on hGH for the hGH receptor as previously described, i.e. by the disruption in binding of variants of hGH except that the receptor was hPRLr rather than hGHs. The results for the above twelve segment-substituted hGH variants are summarized in Table XVIII.

TABLE XVIII

Binding of hGH variants produced by homolog-scanning

mutagenesis to the extracellular domain of the hPRL

receptor (hPRLr). Mutants are named according to

the extremes of segments substituted from the various

hGH homologs: pGH, hPL, or hPRL. The exact

description of the mutations introduced is given by

the series of single mutants separated by commas.

The component single mutants are designated by the

single letter code for the wild-type residue followed

by its codon position in mature hGH and then the

mutant residue. Mutants of hGH were produced and

purified as previously described herein. Binding to

hPRLr was measured essentially as described for the

hGHr (Spencer, S. A. et.al. (1988) J. Biol. Chem.

263, 7862-7867) except that affinity purified rabbit

polyclonal antibodies raised against the hPRLr were

used to precipitate the hPRLr complex with GIBCO ™ brand BSA

(crude) as carrier protein. Standard deviations in

values of K

D

were typically at or below 20% of the

reported value. The relative reduction in binding

affinity (K

D

(mut)/K

D

(hGH)) for the hGHr was taken

from Table III herein. The change in receptor

preference was calculated from the ratios of the

relative reductions in binding affinity for the hGHr

to the hPRLr. WT = wild-type.

Change in

receptor

hPRLr

hGHr

preference

Mutant

Mutations

K

D

(mut)

K

D

(mut)

hGHr

Name

Introduced

K

D

(nM)

{overscore (K

D

(hGH))}

{overscore (K

D

(hGH))}

{overscore (hPRLr)}

WT hGH

none

2.3

(1)

(1)

(1)

pGH

D11A, M14V,

852

370

3.4

110

(11-33)

H18Q, R19H,

F25A, Q29K, E33R

pGH

P48A, T50A, S51A,

2.0

0.9

2.8

0.32

(48-52)

L52F

pGH

S57T, T60A, S62T,

167

73

17

4.3

(57-73)

N63G, R64K, E65D,

T67A, K70R, N72D,

L73V

hGH

Deletion of

14

6.1

ND

(Δ32-46)

residues 32 to 46

hPL

Q46H, H47D, P48S,

4.4

1.9

7.2

0.26

(46-52)

Q49E, L52F

hPL

E56D, R64M

4.1

1.8

30

0.06

(56-64)

hPRL

N12R, M14V, L15V,

3.2

1.4

17

0.08

(12-19)

R16L, R19Y

hPRL

Q22N, F25S, D26E,

168

73

0.85

85

(22-33)

Q29S, E30Q, E33K

hPRL

F54H, S55T, E56S,

2.5

1.1

69

0.02

(54-74)

I58L, P59A, S62E,

N63D, R64K, E66Q,

T674, K70M, S71N,

N72Q, L73K, E74D

hPRL

E88G, Q91Y, F92H,

3.8

1.6

1.4

1.1

(88-95)

R94T, S95E

hPRL

F97R, A98G, N99M,

12.1

5.2

1.6

3.2

(97-104)

S100Q, L101D,

V102A, Y103P,

G104E

hPRL

Y111V, L113I,

2.6

1.1

1.5

0.73

(111-129)

K115E, D116Q,

E118K, E119R,

G120L, Q122E,

T123G, G126L,

R127I, E129S

WT hPRL

none

7.6

3.3

>100,000

—

As can be seen, pGH (11-33) and pGH (57-73) cause large disruptions in hPRL receptor binding affinity, whereas pGH (48-52) has no effect. Unlike the hGH receptor, the hPRL receptor will bind hPRL and hPL but not pGH. As expected, virtually all of the substitutions tested from the binding-competent hormones, hPRL or hPL, did not disrupt binding. The only exception was hPRL (22-33), which caused a >70-fold reduction in binding affinity for the hPRL receptor. Thus, the hPRL receptor is very sensitive to mutations in hGH near the central portion of helix

1

and the loop between residues 57 and 73.

The homolog-scan data also suggest that the hPRL and hGH receptor epitopes are not identical because several segment-substituted variants cause huge changes in receptor binding preference (Table XVIII). For example, the disruption in binding caused by the pGH (11-33) or hPRL (22-33) are about 100-fold greater for the hPRL receptor than for the hGH receptor. In contrast, the hPL (56-64) and hPRL (54-74) have almost no effect on the hPRL receptor, whereas they weaken binding to the hGH receptor by factors of 17 and 69, respectively. These preferential binding effects (along with binding of monoclonal antibodies as previously discussed) further substantiate that reductions in receptor binding affinity are caused by local and not global structural changes in the mutants of hGH.

The specific side-chains in hGH that strongly modulate binding to the hPRL receptor were identified by alanine-scanning mutagenesis and homologous substitutions. The hGH variants shown in Table XIX were prepared. The hPRL substitutions, F25S and D26E, cause the largest reductions in binding affinity (21 and 4.5-fold, respectively) in helix

1

. These residues project from the hydrophilic face of helix

1

(

FIG. 25B

) and are on the same side as other mutations in helix

1

(notably H18A and F10A) that have milder effects on binding.

Four residues in the loop region (54 to 68) known to affect binding of hGH receptor as well as two residues (Q49A and T50A) preceding this region that are nearby and do not affect hGH receptor binding were tested. The most disruptive mutants are I58A and R64A, which reduced binding affinity by 32 and 6-fold, respectively; the other four mutations have negligible effects.

The fact that helix

1

and the loop region (58-64) contain strong binding determinants for the hPRL receptors implicate helix

4

because this helix is wedged between these two structures (FIG.

25

B). Indeed, alanine-scanning of the helix

4

region between a disulfide linked to C165 through V185 reveals strong binding determinants (Table XIX). The most disruptive mutations extend nearly four helical turns, from R167 to R178, and are located on the same hydrophilic face.

TABLE XIX

Binding of single mutants of hGH to hPRL or hGH

receptor fragments (hPRLr or hGHr). Mutants of hGH

were prepared and purified as previously described

except for Q22N, F25S, D26E, Q29S and E33K, which were

produced by site-directed mutagenesis (Cunningham,

B. C. and Wells, J. A. (1989) Science 244, 1330-1335;

Zoller, M. J. and Smith, M. (1982) Nucleic Acids Res.

10, 6487-6499). Receptor binding assays and mutant

nomenclature are described in Table XVIII. Data for

the reduction in binding affinity to the hGHr is

taken from Table III. ND indicates not determined.

Change in

receptor

hPRLr

hGHr

preference

K

D

(mut)

K

D

(mut)

hGHr

Mutant

K

D

(nM)

{overscore (K

D

(hGH))}

{overscore (K

D

(hGH))}

{overscore (hPRLr)}

WT hGH

2.3

(1)

(1)

(1)

P2A

1.3

0.6

0.9

0.7

T3A

3.4

1.5

0.9

1.7

P5A

2.5

1.1

2.1

0.5

L6A

4.0

1.8

2.8

0.6

S7A

1.9

0.8

1.8

0.4

F10A

8.1

3.5

5.9

0.6

N12A

1.9

0.8

1.2

0.7

M14A

1.3

0.6

2.2

0.3

L15A

1.2

0.5

1.3

0.4

H18A

3.9

1.7

1.6

0.6

R19A

1.4

0.6

0.7

2.4

Q22N

2.1

0.9

ND

—

F25S

48

21

ND

—

D26E

10

4.5

ND

—

Q29S

3.2

1.4

ND

—

E33K

1.8

0.8

ND

—

Q49A

1.5

0.7

ND

—

T50A

1.9

0.8

ND

—

F54A

1.8

0.8

4.4

0.2

I58A

73

32

17

1.9

R64A

13

5.7

21

0.3

Q68A

3.1

1.2

5.2

0.3

R167A

7.4

3.2

0.75

4.3

K168A

58

25

1.1

23

D171A

3.6

1.6

7.1

0.2

K172A

143

62

14

4.4

E174A

59

26

0.22

120

F176A

129

56

16

3.5

R178N

2.4

1.0

8.5

0.1

R178K

6.7

2.9

ND

—

I179M

1.3

0.6

2.7

0.2

V185A

3.9

1.7

4.5

0.4

Functional contour maps were derived based upon the location of the mutations in hGH that disrupt binding to the hGH and hPRL receptors (FIG.

28

). The maximal extent of the epitope for the hPRL receptor (

FIG. 25B

) is approximated by mutations having less than a two-fold reduction in binding affinity. By this criteria the epitope for the hPRL receptor is essentially confined to the front face of helix

1

from F10 to Q29, the loop from F54 to Q68, and the hydrophilic face helix

4

from R167 to R178. In contrast, the hGH receptor epitope (

FIG. 25A

) is comprised of residues in the amino terminal region through the front face of helix

1

from I4 through M14, the loop region from F54 through S71, and the hydrophilic face of helix

4

from D171 through V185. Although further mutagenic analysis will be necessary to fill in remaining gaps in the hPRL epitope, it is clear this epitope overlaps but is not identical to that for the hGH receptor. These data suggest that not all of the binding determinants for recognizing hGH are the same in the hGH and hPRL receptors despite them sharing 32% sequence identity in their extracellular binding domains.

Residues that cause large changes in receptor binding affinity may do so by indirect structural effects. However, it is believed that most of these disruptive effects are due to local effects because all of the single mutants tested retain full binding affinity to a panel of 8 hGH monoclonal antibodies and often lead to changes in receptor preference (see Table XIX and infra) and not uniform disruptions in receptor affinity.

Design of Receptor-specific Variants of hGH.

A number of the single hGH mutants cause enormous changes in receptor binding preference (Table XIX). The most notable is E174A, which causes a 4-fold strengthening in affinity for the hGH receptor while weakening binding to the hPRL receptor by more than 20-fold. This represents a 120-fold shift in receptor preference. Other mutations (notably R178N and 1179M) cause hGH to preferentially bind to the hPRL receptor. Typically, the variants that cause the greatest changes in receptor specificity are located in the non-overlap regions of the two receptor epitopes.

It was reasoned that if the changes in receptor binding free energy were additive, it could be possible to design highly specific variants of hGH with only a few mutations. Indeed, when the two most hGH-receptor-selective single mutants (K168A and E174A) are combined, the double mutant exhibits a 2300-fold preference for binding to the hGH receptor (Table XX). As previously indicated, the preference for binding the hPRL receptor can be enhanced by nearly 20-fold by hPL (56-64), which contains only two mutations, E56D and R64M (Table XIII). These hGH variants (K168A, E174A or E56D, R64M) do not substantially reduce the affinity for the preferred receptor, hGH or hPRL, respectively. It is also possible to reduce binding to both receptors simultaneously.

TABLE XX

Binding of double mutants of hGH designed to

discriminate between the hGH and hPRL receptors (hGHr

and hPRLr). Mutants of hGH were prepared by

site-directed mutagenesis, purified, and assayed for

binding to the hGHr or hPRLr as described in Table

XIII. Standard deviations in the determination of K

D

were at or below 20% of the reported value except

where the K

d

is above 10 μM, in which case they

were ± 100% of the reported

value.

Change in

receptor

hPRLr

hGHr

preference

K

D

K

D

(mut)

K

D

(mut)

hGHr

Mutant

(nM)

{overscore (K

D

(hGH))}

K

D

(nM)

{overscore (K

D

(hGH))}

{overscore (hPRLr)}

WThGH

2.3

(1)

0.34

(1)

(1)

K168A,

1950

590

0.09

0.26

2300

E174A

R18N,

I179M

ND

—

ND

—

—

K172A,

˜40,000

˜20,000

190

50

˜40

F176A

For example, combining K172A and F176A, which individually cause large reductions in binding affinity to the hGH and hPRL receptors, produces much larger disruptions in affinity of 550—and 15,000-fold, respectively.

In all these instances the changes in the free energy of binding (ΔΔ

binding

) are strikingly additive (Table XXI). Additive effects of mutations have been observed in enzyme-substrate interactions (P. J. Carter, et al. (1984)

Cell

38, 835-840; J. A. Wells, et al., (1987)

Proc. Natl. Acad. Sci. USA

84, 5167-5171), protease-protease inhibitor interactions (M. Laskowski, et al. in Protease Inhibitors: Medical and Biological Aspects, (1983), eds. N. Katunuma, Japan Sci. Soc. Press, Tokyo, pp. 55-68, and protein stability (D. Shortle, et al., (1986)

Proteins

1, 81-89 (1986); M. H. Hecht, J. M. Sturtevant and R. T. Sauer. Proteins 1, 43-46) and, as disclosed in these references, are most commonly found when the mutant residues function independently and are in contact with each other. This suggests the residues paired in the multiple mutants of hGH function independently. Such additivity creates an extremely predictable situation for engineering variants of hGH with desirable receptor binding affinity and specificity.

TABLE XXI

Additive effects of mutations in hGH upon binding to

the hGH or hPRL receptors (hGHr or hPRLr). The

change in the free energy of binding (ΔΔG

binding

)

for the variant relative to wild-type hGH was

calculated from the reduction in binding affinity

according to: ΔΔG

binding

= RT ln[K

D

(mut)/K

D

(hGH)].

The values of (K

D

(mut)/K

D

(hGH) for the single-or

multiple-mutant hormones were taken from Tables XIII-XX.

Change in binding free

energy, ΔΔG

binding

(kcal/mol)

Mutation

hGH

hPRLr

K168A

+0.04

+1.9

E174A

−0.90

+1.9

K168A, E174A (expected)

−0.86

+3.8

(actual)

−0.80

+3.8

K172A

+2.5

+1.6

F176A

+2.4

+1.6

K172A, F176A (expected)

+4.9

+3.2

(actual)

+5.7

+3.8

Q22N

−0.06

ND

F25S

+1.81

ND

D26E

+0.89

ND

Q29S

+0.20

ND

E30Q

ND

ND

E33K

−0.13

ND

hPRL 22-33 (expected)

+2.7

—

(actual)

+2.6

—

E56A

ND

+0.8

R64M

ND

+1.8

E56A, R64M (expected)

—

+2.6

hPL (56-64) (actual)

—

+2.0

There are a number of other cases like hGH where two or more receptors or receptor subtypes are known to exist such as for the adrenergic receptors (for review see R. J. Lefkowitz and M. G. Caron (1988)

J. Biol Chem.

263, 4993-4996), IGF-I receptors (M. A. Cascieri, et al., (1989)

J. Biol. Chem.

264, 2199-2202), IL-2 receptors (R. J. Robb, et al. (1984)

J. Exp. Med.

160, 1126-1146; R. J. Robb, et al. (1988)

Proc. Natl. Acad. Sci. USA

85, 5654-5658) and ANP receptors (D. Lowe and D. Goeddel, unpublished results). In these situations it is difficult to link specific receptor function to a specific pharmacological effect. However, the use of receptor-specific hormone analogs can greatly simplify this task. For example, catecholamine analogs were used to characterize β-adrenergic receptor subtypes and link receptor function to physiologic responses (for review see R. J. Lefkowitz, et al. (1983)

Annu. Rev. Biochem.

52, 159-186). By analogy, the receptor-specific variants of hGH should provide a key tool for identifying other receptors for hGH, and for probing the role of the hGH and hPRL receptors in the complex pharmacology of hGH. This work represents a systematic approach to identifying receptor binding sites in hormones that permits rational design of receptor-specific variants.

EXAMPLE 13

Engineering Human Prolactin to Bind to Human Growth Hormone

Prolactin (PRL) is a member of a large family of homologous hormones that includes growth hormones (GH), placental lactogens (PL), and proliferins. Nicoll, C. S. et. al. (1986)

Endocrinol. Rev.

7, 169-203. Collectively, this group of hormones regulates a vast array of physiological effects involved in growth, differentiation, electrolyte balance, and others. Chawla, R. K. et.al. (1983)

Ann. Rev. Med.

34, 519-547: Isaksson, O. G. P. et.al. (1985)

Ann. Rev. Physiol.

47, 483-499. These pharmacological effects begin with binding to specific cellular receptors. For instance, hPRL binds to the lactogenic but not somatogenic receptor and stimulates lactation but not bone growth; hGH can bind to both the lactogenic and somatogenic receptors and stimulates both lactation and bone growth. The molecular basis for the differences in receptor binding specificity is not understood.

Cloning and Expression of hPRL.

The cDNA for hPRL was cloned from a human pituitary cDNA library in λgt10 (Huynh, T. V., et al. (1985) in

DNA Cloning Techniques: A Practical Approach,

Vol. 1, D. M. Glover, ed. (Oxford IRL Press) pp. 49-78) by hybridization (Maniatis, T., et al., eds. (1982)

Molecular Cloning A Laboratory Manual

(Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y.)) with oligonucleotide probes corresponding to 5′ and 3′ extremes of the published DNA sequence (Cooke, N. E., et al. (1981)

J. Biol. Chem.

256, 4007-4016). A near full-length cDNA clone was identified and the 720-bp BstII-HindIII fragment, extending from codon 12 to 55 bp past the stop codon, was subcloned into pUC118. The sequence was determined by the dideoxy method (Sanger, F., et al. (1977)

Proc. Natl. Acad. Sci. USA

74, 5463-5467) and matched exactly that previously reported (Cooke, N. E., et al. (1981)

J. Biol. Chem.

25, 4007-4016).

The intracellular expression vector, pBO760 (FIG.

26

), was created in several steps by standard methods (Maniatis, T., et al., eds. (1982)

Molecular Cloning A Laboratory Manual

(Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y.)). The

E. coli

trp promoter derived from pHGH207-1 (deBoer, H. A., et al. (1982) in

Promoters Structure and Function,

eds. Rodriguez, R. L. & Chamberlin, M. J. (Praeger, New York) pp. 462-481) was used to transcribe the hPRL gene. The hPRL coding sequence consisted of a 47-bp XbaI-BstEII synthetic DNA cassette and the 720-bp BstEII-HindIII fragment derived from the hPRL cDNA. The synthetic DNA cassette had the sequence

***

5′-CT-AGA-ATT-ATG-TTA-CCA-ATT-TGT-CCA-GGT-GGT-GCA-GCA-AGG-TGT-CAA

3′-T-TAA-TAC-AAT-GGT-TAA-ACA-GGT-CCA-CCA-CGT-CGT-TCC-ACA-GTT-CAC-TG,

where the initiation codon is indicated by asterisks. The phage f1 origin, pBR322 replication origin, and the pBR322 β-lactamase gene were derived from pBO475 (Cunningham, B. C., et al. (1989)

Science

243, 1330-1335).

E. coli

cells (MM 294) containing pBO760 were grown at 370° C. for 4 hr (or early log phase; A

550

=0.1 to 0.3) in 0.5-L shake flasks containing 100 ml of M9 Hycase media (Miller, J. H. (1972)

Experiments in Molecular Genetics

(Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y.)) plus 15 μg/ml carbenicillin. Indole acrylic acid was added (50 μg/ml final) to induce the trp promoter. Cells were grown an additional 6-8 hr and harvested by centrifugation. Cell fractionation experiments showed the hPRL was located almost exclusively in inclusion particles and represented 2-5% of the total cell protein as analyzed by SDS-PAGE (not shown).

Purification and Refolding of hPRL.

Inclusion particles containing hPRL were isolated essentially as described (Winkler, M. E., et al. (1986)

Biochemistry

25, 4041-4045). Briefly, 50 g of wet cell paste was suspended in 0.25 liters, 10 mM, TRIS HCl Tris(hydroxymethyl)aminomethane hydrochloride (pH 8.0), 1 mm EDTA (TE buffer) and cells were lysed by vigorous sonication. Insoluble material was collected by centrifugation (10,000×g for 15 min) and resuspended in 25 ml of TE buffer. The suspension was layered on a 0.2-liter cushion of 50% glycerol, and centrifuged at 9,000×g for 25 min to pellet the hPRL inclusion particles. The hPRL from the inclusion particles (about 20% pure) was suspended in 5 ml of TE buffer.

The hPRL was refolded by solubilizing the inclusion particles in 156 ml of 8N GnHCl in TE buffer plus 0.3 g reduced glutathione (Sigma). After gentle stirring at room temperature for 30 min, the mixture was chilled to 0° C. and diluted with 844 ml of cold TE buffer plus 0.6 g oxidized glutathione. The solution was stirred slowly overnight at 4° C., and dialyzed with 4 liters of TE buffer that was changed three times over 24 hr. Insoluble material was removed by centrifugation (10,000×g for 20 min).

The refolded and solubilized hPRL was further purified by precipitation with (NH

4

)

2

SO

4

to 45% saturation and stirred 2.5 hr at room temperature. The precipitate was collected by centrifugation (12,000×g for 30 min) and redissolved in 5 ml of TE buffer. After 30 min at room temperature, the solution was clarified (10,000×g for 10 min) and filtered through a millipore filter (0.45 μm). The solution was dialyzed against 0.5 liters of TE buffer overnight at 4° C. The hPRL (85% pure) was finally purified to homogeneity (>95%) by FPLC using DEAE fast-flow matrix essentially as described for purifying hGH (Cunningham, B. C., et al. (1989)

Science

243, 1330-1335).

Mutagenesis and Binding Properties of hGH and hPRL Variants.

Site-specific mutagenesis (Zoller, M. J., et al. (1982)

Nucleic Acids Res.

10, 6487-6500) was carried out with the aid of a methylation repair deficient strain of

E. coli,

Mut L (Kramer, B., et al. (1984)

Cell

38, 879-887). Additional enrichment for mutant clones was obtained by designing mutagenic oligonucleotides to either introduce or eliminate a nearby unique restriction site so that restriction-purification or restriction-selection (Wells, J. A., et al. (1986)

Phil. Trans. R. Soc. Lond.

A 317, 415-423), respectively, could be applied to the first pool of plasmid DNA obtained after transformation of the in vitro-generated heteroduplex. All oligonucleotides were designed to have 12 bp of exact match 5′ to the most upstream mismatch and 10 bp 3′ to the most downstream mismatch. For mutagenesis of hGH, the previously described hGH synthetic gene contained multiple restriction sites and was cloned into the plasmid, pBO475. Variants of hGH were secreted into the periplasmic space of

E. coli

(Chang, C. N., et al. (1987)

Gene

55, 189-196) and purified as previously described.

The K

d

of each analog was determined by competitive displacement of [

125

I]hGH bound to the purified recombinant hGH binding protein as previously described herein and in Spencer, S. A., et al. (1988)

J. Biol. Chem.

263, 7862-7867. The previously described hGH binding protein (containing residues 1 to 238 of the cloned human liver receptor) was secreted and purified from

E. coli

as described in Fuh, G., et al. (1989) (submitted). Displacement curves were generated in triplicate and the standard deviations in the K

d

values were generally at or below 20% of the reported values and did not exceed 50% of the reported value except when Kd values were greater than 10 μM.

The concentrations of hPRL and hPRL mutants were determined by A

280

using a calculated extinction coefficient of ±S(0.1%,280)=0.9 (Wetlaufer, D. B. (1962)

Adv. in Prot. Chem.

17, 303-390). This was adjusted accordingly when variants contained mutations in aromatic residues. Concentration values determined by absorbance agreed to within 10% with those determined by laser densitometry of proteins run on SDS-PAGE and stained with Coomassie blue for hGH. Circular dichroic spectra were collected on an Aviv Cary 60 spectropolarimeter.

In order to probe which of the divergent residues in hPRL were most disruptive for binding to the hGH receptor (FIG.

27

), a number of hPRL residues were first introduced into hGH (Table XXII).

TABLE XXII

Comparison of hPRL and alanine substitutions

introduced into hGH

K

d

(mut)

hGH variant

K

d

(nM)

{overscore (K

d

(hGH))}

WT

0.34

(1)

I58L

0.58

1.7

I58A

5.6

16

R64K

0.20

0.6

R64A

7.1

21

F176Y

2.9

8.6

F176A

5.4

16

R178K

1.7

5.1

R178N

2.9

8.5

Whereas single alanine substitutions in hGH at positions 58, 64, 176 and 178 strongly disrupted receptor binding, substitutions of hPRL residues into hGH at these positions had less of an effect. The largest effects for hPRL substitutions were in the helix

4

residues that included positions 176 and 178. These data suggested that residues in the helix

4

region of hPRL could best account for the lack of binding to the hGH receptor.

The recombinant hPRL retained native-like structural and functional properties. First, the near and far ultraviolet CD spectra (

FIG. 28

) are identical to published spectra of natural hPRL (Bewley, T. A. (1979) in

Recent Progress in Hormone Research,

vol. 35, pp. 155-213, Acad. Press, N.Y.). The far ultraviolet spectrum is similar to that of hGH, suggesting a similar 4-helix bundle structure, although important differences in the mean residue ellipticity at 208 and 224 nm have been noted (Id) These hormones differ markedly in the near ultraviolet CD spectra which reflects variation in number and microenvironment of the aromatic residues between hGH and hPRL. In other studies (not shown), the recombinant hPRL retained full immunological cross-reactivity in an hPRL ELISA, and was equipotent with hGH in causing rat lymphoma Nb2 cells to proliferate (Tanaka, T., et al. (1980)

J. Clin. Endo. Metab.

51, 1058-1063). Upon reduction, the purified hPRL showed a pronounced retardation in mobility by SDS-PAGE (as seen for hGH) suggesting that disulfide bonds had formed (Pollitt, S., et al. (1983)

J. Bacteriol.

153, 27-32). Amino-terminal sequence analysis showed that the intracellularly expressed hPRL retained the amino-terminal methionine; however, as with methionyl-hGH (Olson, K. C., et al. (1981)

Nature

(London) 29, 408-411), this does not apparently affect its structure or function.

Binding of hPRL to the hGH binding protein is reduced by more than 10

5

-fold compared to hGH (Table XXIII), which is below the detection limit of our binding assay.

TABLE XXIII

Engineering residues in hPRL to permit binding to the

hGH binding protein

1

Kd (mut)

hPRL Variant

K

d

(nM)

2

{overscore (K

d

(hGH))}

hPRL WT

>40,000

>100,000

A = H171DN175TY176F

4,900

14,000

B = A + K178R

220

660

B + hGH (184-188)

260

740

hGH (54-74)

25,000

66,000

B + hGH (54-74)

2,000

5,800

B + H54FS56E:L58I:

36

110

E62S:D63N:Q66E

B + H54F:S56E:L58I

670

2,000

C = B + E174A

68

200

D = C + E62S:D63N:Q66E

2.1

6.2

D + H54F

4.4

13

D + S56E

2.5

7.4

D + L58I

3.6

11

D + A59P

2.5

7.4

D + N71S

3.6

11

D + L179I

2.1

6.2

1

Mutants of hPRL were generated, purified and analyzed as described. Multiple mutants are indicated by a series of single mutants (Table XXII) separated by colons. Codon numbering is based upon the hGH sequence (FIG. 2).

2

Average standard errors are at or below 20% of the reported values, except in cases where the K

d

exceeds 1 μM, where errors can be as large as 50%, and errors are much larger still when K

d

exceeds 10 μM.

A combination of three divergent residues in helix

4

from hGH (H171D, N175T, and Y176F) were introduced into hPRl. Alanine-scanning mutagenesis and hPRL substitutions (Table XXII) had shown that these residues were very important for binding hGH to the hGH receptor. This triple mutant of hPRL exhibited detectable binding to the hGH binding protein albeit 14,000-fold weaker than hGH. Installation of another important helix

4

residue (K178R) to produce a tetramutant (called variant B in Table XIII) further strengthened binding to a level now only 660-fold below wild-type hGH. Additional incorporation of hGH residues 184 to 188 into hPRL variant B did not enhance binding to the hGH binding protein. However, introduction of E174A to give hPRL variant C (Table XXIII) caused an additional 3.5-fold increase in binding affinity to the hGH binding protein as was found when E174A was incorporated into hGH.

Having engineered binding with the helix

4

region, the loop region containing residues 54 to 74 was analysed. Complete replacement of the loop region in hPRL with the sequence from hGH (hGH (54-74) in Table XIII) gave barely detectable binding to the hGH binding protein. When this mutant was combined with variant B, the binding affinity increased substantially. However, this new variant [B plus hGH (54-74)] was reduced in binding affinity by almost 10-fold from variant B alone. Thus, it appeared that some of the hGH residues in the 54-74 loop were not compatible with the hGH substitutions in helix

4

. We then selected from the 54 to 74 loop of hGH only those seven residues that were shown by alanine-scanning mutagenesis to most greatly influence binding. Although the R64A mutation in hGH caused more than a 20-fold reduction in binding affinity, the R64K variant of hGH (which is an hPRL substitution) slightly enhanced binding to the hGH binding protein (Table XXII). The Lys64 in hPRL therefore was left unchanged. As a consequence, only six of the seven substitutions from hGH were incorporated into hPRL that were most disruptive when changed to alanine in hGH. This new mutant (B plus H65F:S56E:L58I:E56S:D68N:Q66E) binds fifty-fold stronger than B plus hGH (54-74) and was only 110-fold reduced in binding affinity from wild-type hGH (Table XXIII). However, this represented only a modest improvement (six-fold) over variant B alone, which was less than expected for strongly favorable interactions previously observed in the loop region for hGH. Therefore, the six mutations within the loop were further dissected and revealed that the combination of H54F:S56E:L58I plus variant B bound three-fold weaker than variant B alone. Finally, incorporating the mutations E62S:D63N:Q66E into variant C (to give variant D) produced an analog with highest affinity that was only 6-fold reduced in binding affinity relative to hGH. Additional single mutations (H54F, S56E, L58I, A59P, N71S and L179I) did not enhance the binding affinity of hPRL variant D to the hGH binding protein. The conformation of variant D was virtually indistinguishable from that of native hPRL by CD spectral analysis (

FIG. 28

) or by ELISA reactivity (not shown).

These studies demonstrate the feasibility of recruiting binding properties for distantly related homologs using only functional information derived from site-directed mutagenesis experiments. Alanine-scanning mutagenesis of hGH provided a systematic analysis of side-chains that were important for modulating binding of hGH to its receptor (FIG.

27

).

This information highlighted a number of residues in hPRL that could account for its inability to bind to the hGH receptor (FIG.

29

). However, further analysis showed that the alanine substitutions in hGH were more disruptive than the hPRL substitutions in hGH (Table XXII). Furthermore, some of the hPRL substitutions were considerably more disruptive than others for binding affinity, especially when a larger side-chain was present in hPRL. For example, the conservative (but larger) F176Y mutation in hGH caused an eight-fold reduction in binding affinity with the hGH receptor, whereas the smaller R64K substitution showed slightly enhanced binding affinity. Thus, the analysis of disruptive hPRL substitutions in hGH suggested the introduction of the cluster of divergent residues in helix

4

to initially achieve binding affinity for hPRL. This was very important because no binding to the hGH receptor with wild-type hPRL had been observed, and it was necessary to introduce several hGH substitutions simultaneously into hPRL in order to bring the binding affinity within the range of the assay used (K

d

≦50 μM).

Readily detectable binding affinity was engineered into hPRL by incorporating functionally important residues into helix

4

. However, engineering the loop region between 54-74 turned out to be more difficult. Installing the entire loop from hGH into hPRL produced less enhancement in binding than expected, and was disruptive to binding when combined with the optimized helix

4

variant B. Our data suggest that the 54-74 loop structure in hPRL is supported by other interactions in the protein. This problem was solved in stages. First, only those six loop residues from hGH that the alanine scan together with the hPRL substitutions in hGH had identified to be important were introduced into hPRL. Although this improved the situation, the combination of some of these hGH mutations (narrowed down to H54F, S56E, and L58I) were disruptive to hPRL. These data suggest that some of the residues in the loop are crucial for its structure and are better off being left alone.

A number of iterative cycles of mutagenesis were necessary to converge upon a combination of residues that permitted tight binding of hPRL to the hGH receptor. This strategy relies on the assumption that the mutational effects will be somewhat additive as was, in fact, observed. For example, the E174A mutation enhanced the binding three to five-fold when added to either hPRL variant C or hGH. Moreover, the product of the disruptive effects of the H54F, S56E, and L581 single mutants to variant D (4.4-fold) is about the same as the disruption caused by the combination of all three mutations added to variant B (3-fold).

Even though variant D is only six-fold reduced in binding affinity, there are several other residues that could be incorporated into variant D to try to improve further on the binding, such as V14M and H185V; these are sites where alanine substitutions in hGH cause two to five-fold reductions in binding of hGH (FIG.

29

). Although a high resolution structure would have aided in the design process, it was clearly not essential. The cumulative nature of the mutational effects allows one to converge upon the binding property in much the same way as proteins evolve, by cycles of natural variation and selection.

Previous protein engineering experiments have shown it is possible using highs resolution structural analysis to virtually exchange the substrate specificity of natural variant enzymes by site-directed mutagenesis of substrate contact residues (Wells, J. A., et al. (1987)

Proc. Natl. Acad. Sci.

USA 84, 5167-5171; Wilks, H. M., et al. (1988)

Science

242, 1541-1544). Similarly, others have shown that binding properties can be engineered by replacement of entire regions of secondary structure units including antigen-binding loops (Jones, P. T., et al. (1986)

Nature

321, 522-525) or DNA recognition helices (Wharton, R. P., et al. (1985)

Nature

316, 601-605). However, to recruit the hGH receptor binding properties into hPRL required selective residue replacements within the structural scaffold of hPRL. Furthermore, the CD spectral data show that the overall structure of the hPRL variant D resembles more closely the structure of hPRL not hGH even though it attains binding properties like hGH.

The fact that the binding specificity for the hGH receptor could be incorporated into hPRL confirms the functional importance of particular residues for somatogenic receptor binding. These studies also provide compelling proof for structural relatedness between hGH and hPRL despite them having only 23% identity. This provides a rational approach to access new receptor binding functions contained within this hormone family starting with either a growth hormone, prolactin, proliferin or placental lactogen scaffold. Such hybrid molecules should be useful for distinguishing receptor binding and activation as well as the pharmacological importance of receptor subtypes. These analogs could lead to the design of new receptor-specific hormones having more useful properties as agonists or antagonists.

EXAMPLE 14

Recruitment of Binding Properties of Human Growth Hormone into Human Placental Lactogen

Human placental lactogen (hPL) is reduced over thirty-fold in binding affinity compared to hGH for the hGH receptor (G. Baumann, et al., (1986)

J. Clin. Endocrinol, Metab.

62, 134; A. C. Herington, et al. (1986)

J. Clin. Invest.

77, 1817). Previous mutagenic studies showed the binding site on hGH for the hGH receptor is located primarily in two regions (including residues 54-74,) and 171-185), with some minor determinants near the amino terminus (residues 4-14).

The overall sequence of hPL is 85% identical to hGH. Within the three regions that broadly constitute the receptor binding epitope on hGH, hPL differs at only seven positions and contains the following substitutions: P2Q, I4V, N12H, R16Q, E56D, R64M, and I179M. (In this nomenclature the residue for wild-type hGH is given in single letter code, followed by its position in mature hGH and then the residue found in hPL; a similar nomenclature is used to describe mutants of hGH.) Single alanine substitutions have been produced in hGH at each of these seven positions. Of these, four of the alanine substitutions were found to cause two-fold or greater reductions in binding affinity, including I4A, E56A, R64A, and I179A. Generally, the alanine substitutions have a greater effect on binding than homologous substitutions from human prolactin. Therefore, the effect of some of the substitutions from hPL introduced into hGH were investigated. Whereas the I179A substitution caused a 2.7-fold reduction in affinity the I179M substitution caused only a slight 1.7-fold effect. However, the R64A and R64M substitutions caused identical and much larger reductions (about 20-fold) in binding affinity. Moreover, the double mutant (E56D:R64M) in hGH was even further reduced in affinity by a total of 30-fold (Table I). Thus, E56D and R64M primarily determine the differences in receptor binding affinity between hGH and hPL. The double mutant D56E, M64R in hPL therefore substantially enhances its binding affinity for the hGH receptor. Additional modifications such as M179I and V4I also enhance binding of hPL to the hGH receptor.

EXAMPLE 15

Effect of Amino Acid Replacement at Position 174 on Binding to the Human Growth Hormone

As previously indicated, replacement of Glu174 with Ala (E174A) resulted in more than a 4-fold increase in the affinity of human growth hormone (hGH) for its receptor. To determine the optimal replacement residue at position 174 hGH variants substituted with twelve other residues were made and measured to determine their affinities with the hGH binding protein (Table XXIV). Side-chain size, not charge, is the major factor determining binding affinity. Alanine is the optimal replacement followed by Ser, Gly, Gln, Asn, Glu, His, Lys, Leu, and Tyr.

TABLE XXIV

Side chain

Kd (mut)

Mutant

a

Charge

Size(Å

3

)

b

Kd (nM)

c

{overscore (Kd (wild type))}

E174G

0

9

0.15

0.43

E174A

0

26

0.075

0.22

E174S

0

33

0.11

0.30

E174D

−

59

NE

—

E174N

0

69

0.26

0.70

E174V

0

76

0.28

0.80

wild-type

−

89

0.37

1.0

E174Q

0

95

0.21

0.60

E174H

0

101

0.43

1.2

E174L

0

102

2.36

6.4

E174K

+

105

1.14

3.1

E174R

+

136

NE

—

E174Y

0

137

2.9

8.6

a

Mutations were generated by site-directed mutagenesis (Carter, P., et al. (1986) Nucleic Acid Res. 13, 4431-4443) on a variant of the hGH gene thatcontains a KpnI site at position 178 cloned into pB0475. Oligonucleotides used for mutagenesis had the sequence: 5′-AC-AAG-CTC-NNN-ACA-TTC-CTG-CGC-ATC-GTG-CAG-T-3′, where NNN represents the new codon at position 174 and asterisks indicate the mismatches to eliminate the KpnI site starting at codon 178.

# Mutant codons were as follows: Gln, CAG; Asn, AAC; Ser, AGC; Lys, AAA; Arg, AGG; His, CAC; Gly, GGG; Val, GTG; Leu, CTG. Following heteroduplex synthesis the plasmid pool was enriched for the mutation by restriction with KpnI to reduce the background of wild-type sequence. All mutant sequences were confirmed by dideoxy sequence analysis (Sanger, F., et al. (1977) Proc. Natl. Acad. Sci. USA 74, 5463-5467).

b

Side-chain packing values are from C. Chothia (1984) Annu. Rev. Biochem. 53, 537.

c

Dissociation constants were measured by competitive diplacement of [

125

I]hGH from the hGH binding protein as previously described. NE indicates that the mutant hormone was expressed at levels too low to be isolated and assayed.

EXAMPLE 16

The hGH variants shown in Table XXV were constructed. Their relativity potency as compared to wt-hGH are shown.

TABLE XXV

Relative potency in

hGH mutant

rat weight gain assay

F97A

0.87

S100A

2.12

L101A

3.03

V102A

1.39

Y103A

1.73

T175S

1.21

Having described the preferred embodiments of the present invention, it will appear to those ordinarily skilled in the art that various modifications may be made to the disclosed embodiments, and that such modifications are intended to be within the scope of the present invention.

Number	Name	Date	Kind
4446235	Seeburg	May 1984	A
4665160	Seeburg	May 1987	A
4670393	Seeburg	Jun 1987	A
4699897	Jones et al.	Oct 1987	A
4732973	Borr et al.	Mar 1988	A
4871835	Aviv et al.	Oct 1989	A
4888286	Crea	Dec 1989	A
5085862	Klein et al.	Feb 1992	A
5204244	Fell et al.	Apr 1993	A
5350836	Kopchick et al.	Sep 1994	A
5681809	Kopchick et al.	Oct 1997	A
6238915	Chihara	May 2001	B1

Number	Date	Country
0 089 666	Sep 1983	EP
8807578	Jun 1988	EP
0 320 308	Jun 1989	EP
WO 8807084	Sep 1988	WO
WO 8807578	Oct 1988	WO

	Number	Date	Country
Parent	07/960227	Oct 1992	US
Child	08/190723		US
Parent	07/875204	Apr 1992	US
Child	07/960227		US
Parent	07/428066	Oct 1989	US
Child	07/875204		US

	Number	Date	Country
Parent	07/264611	Oct 1988	US
Child	07/428066		US

Growth hormone variants

Information

Patent Number

Date Filed

Date Issued

Inventors

Original Assignees

Examiners

Agents

CPC

US Classifications

Field of Search

US

International Classifications

Term Extension

Abstract

Description

Claims

Parent Case Info

US Referenced Citations (12)

Foreign Referenced Citations (5)

Non-Patent Literature Citations (49)

Continuations (3)

Continuation in Parts (1)

Entry
Nicoll et al. Endocrine Reviews (1986) 7, 169-203.*
Rudinger et al. Peptide Hormones, University Park Press, Baltimore, pp. 1-7, 1976.*
Abdel-Maeguid, S. S. et al., “Three-dimensional structure of a genetically engineered variant of porcine growth hormone”, 84 Proc. Natl. Acad. Sci. USA 6434-6437 (1987).
Bajt et al., Characterization of a Gain of Function Mutation of Integrin αIIβ3 (Platelet Glycoprotein IIb-IIIa), 267 J. Biol. Chem. 22211-22216 (1992).
Barlow, D. J. et al., “Continuous and discontinuous protein antigenic determinants”, 322 Nature 747-748 (1986).
Bennett et al., “High Resolution Analysis of Functional Determinants of Human Tissue-type Plasminogen Activator”, 266 J. Biol. Chem. 5191-5201 (1991).
Berendt et al., “The Binding Site on ICAM-1 for Plasmodium falciparum-Infected Erythrocytes Overlaps, but is Distinct LFA-1 Binding Site”, 68 Cell 71-81 (1992).
Berlot et al., “Identification of Effector-Activating Residues of G sα”, 68 Cell 911-922 (1992).
Bettler et al., “Immunoglobulin E-binding Site in Fc Receptor FcRII/CD23) Identified by Homolog-scanning Mutagenesis”, 267 J. Biol. Chem. 185-191 (1992).
Boutin, J. M. et al., “Cloning and Expression of the Rat Prolactin Receptor, a Member of the Growth Hormone/ Prolactin Receptor Gene Family”, 53 Cell 69-77 (1988).
Burstein et al., “Immunoreactivity and receptor binding of mixed recombinants of human growth hormone and chorionic somatomammotropin”, 75 Proc. Natl. Acad. Sci. USA 5391-5394 (1978).
Camble, R. et al., “Properties of Interferon-α2 Analogues Produced from Synthetic Genes in Peptides: Strucuture and Function”, Proceedings of the Ninth American Peptide Symposium 375-384 (Deber et al. eds. 1985).
Chang, C. N. et al., “High-level secretion of human growth hormone by Escherichia coli”, 55 Gene 189-196 (1987).
Clayton et al., “Substitution of murine for human CD4 residues identifies amino acids critical for HIV-gp120 binding”, 355 Nature 363-366 (1988).
Cunningham, B., “Improvement in the alkaline stability of subtilisin using an efficient random mutagenesis and screening procedure”, 108 Chemical Abstracts 11 (1988).
Cunningham, B. et al., “Receptor and Antibody Epitopes in Human Growth Hormone Identified by Homolog-Scanning Mutagenesis”, 243 Science 1330-1336 (1989).
Cunningham, B. et al., “High-Resolution Epitope Mapping of hGH-Receptor Interactions by Alanine-Scanning Mutagenesis”, 244 Science 1081-1085 (1989).
Ge et al., “Functional Domains of Bacillus thuringiensis Insecticidal Crystal Proteins”, 266 J. Biol. Chem. 17954-17958 (1991).
Geysen et al., 81 P.N.A.S., USA 3998-4002 (1984).
Goeddel et al., “Direct expression in Escherichia coli of a DNA sequence coding for human growth hormone”, 281 Nature 544-548 (1979).
Gray et al., “Periplasmic production of correctly processed human growth in Escherichia coli: nature and bacterial signal sequences are interchangeable”, 39 Gene 247-254 (1985).
Hotta et al., 149 Biochem. and Biophys. Res. Comm. 531-537 (1987).
Huang et al., 223 FEBS Letters 294-298 (1987).
Jones, P. T. et al., “Replacing the complementarity-determining regions in a human antibody with those from a mouse”, 321 Nature 522-525 (1986).
Kobilka, B. et al., “Chemeric α2-, β2-Adrenergic Receptors: Delineation of Domains Involved in Effector Coupling and Ligand Binding Specificity”, 240 Science 1310-1316 (1988).
Kostyo, J. L. et al., “Biological characterization of purified native 20-kDa human growth hormone”, 925 Biochemica et Biohysica Acta 314 (1987).
Krivi, G. G. et al., “Immunohistochemcial Expression of Insulin-like growth factor I during Skeletal Muscle Regeneration in Normal . . . ”, Abstract I-18, Int'l. Symp. on Growth Hormone; Basic and Clin. Aspects, Final Program, sponsored by Serono Sympsoia, USA (Jun. 14-18, 1987).
Laskowski, M. et al., “Postivie Darwinian Selection in Evolution of Protein Inhibitors of Serine Proteinases”, 52 Cold Spring Harbor Symp. Quant. Biol. 545 (1987).
Leung, D. W. et al., “Growth hormone receptor and serum binding protein: purification, cloning and expression”, 330 Nature 537-543 (1987).
Lewis, U. J. et al., “A Naturally Occurring Structural Variant of Human Growth Hormone”, 253 J. Biol. Chem. 2679-2687 (1978).
Lewis, U. J., “Variants of Growth Hormone and Prolactin and Their Posttranslational Modifications”, 46 Ann. Rev. Physiol. 33-42 (1984).
Li, C. H., “Human growth hormone: 1974-1981”, 46 Mol. Cell. Biochem. 31-41 (1982).
Marseigne et al., 31 J. Med. Chem. 966-970 (1988).
Mills, J. B. et al., “Fragments of Human Growth Hormone Produced by Digestion with Thrombin: Chemistry and Biological Properties”, 107 Endocrinology 319-399 (1980).
Nakashima et al., “Alanine-scanning Mutagenesis of the Epidermal Growth Factor-like Domains of Human Thrombomodulin Identifies Critical Residues for Its Cofactor Activtiy”, 268 J. Biol. Chem. 2888-2892 (1993).
Russell et al., “Recombinant Hormones from Fragments of Human Growth Hormone and Human Placental Lactogen”, 256 J. Biol. Chem. 296-300 (1981).
Seeberg, P. H., “The Human Growth Hormone Gene Family: Nucleotide Sequences Show Recent Divergence and Predict a New Polypeptide Hormone”, 1 DNA 239-249 (1982).
Sourouton, M. C. et al., “Localization of Highly Immunogenic Region on the Acetylcholine Receptor Alpha-Subunit”, 135 Biochem. Biophys. Res. Commun. 82-89 (1986).
Tokunaga, T. et al., “Synthesis and expression of a human growth hormone (somatotropin) gene mutated to change cysteine-165 to alanine”, 153 Eur. J. Biochem. 445-449 (1985).
Venuti et al., “The Impact of Biotechnology on Drug Discovery”, Ann. Reports in Medicinal Chem. 289-298 (Vinick ed. 1989).