Protein Engineering vol.5 no.6 pp.527-533, 1992
Altering the association properties of insulin by amino acid replacement
David N.Brems, Leila A.Alter, Michael J.Beckage, Ronald E.Chance1, Richard D.DiMarchi1, L.Kenney Green1, Harlan B.Long1, Allen H.Pekar, James E.Shields' and Bruce H.Frank Parenteral Products Research and Development and 'Diabetes Research. Eli Lilly & Co., Indianapolis, IN 46285, USA
Key words: circular dichroism/mutagenesis/self-association/ size-exclusion/ultracentrifugation
Introduction Insulin association is well documented but not fully understood. It has been studied by ultracentrifugation (Jeffrey and Coates, 1966; Pekar and Frank, 1972; Goldman and Carpenter, 1974; Mark et al., 1987; Mark and Jeffrey, 1990), light scattering (Bohidar and Geissler, 1984), equilibrium dialysis (Grant et al., 1972), difference UV absorbance spectroscopy (Rupley et al., 1967) and CD (Goldman and Carpenter, 1974; Pocker and Biswas, 1981; Melberg and Johnson, 1990; Roy et al., 1990). The insulin association behavior is known to be complex, with the metal-free species exhibiting a pH, ionic strength and protein concentration-dependent association pattern consisting of monomer, dimer, tetramer and higher ordered polymers all in dynamic equilibrium (Fredericq, 1956; Jeffrey and Coates, 1966; Pekar and Frank, 1972; Goldman and Carpenter, 1974; Jeffrey et al., 1976; Pocker and Biswas, 1981). The association constant for the metal-free porcine monomer-dimer equilibrium at pH 7.0 is 1.4-7.5 x 105 NT 1 (Pekar and Frank, 1972; Jeffrey et al., 1976; Pocker and Biswas, 1981; Strazza et al., 1985). Insulin is capable of binding divalent metal ions, with Zn being the most common ligand. Zinc-binding induces a specific © Oxford University Press
The importance of the C terminus of the B chain is further evidenced by removal of B 26 ~ 30 in despentapeptide insulin, which does not significantly alter the rest of the molecule but abolishes dimerization (Bi et al., 1984). In this study the C terminus of the B chain was systematically truncated and it was established that Pro828 is critical to the stabilization of insulin self-association in solution. Amino acid replacement was used to further investigate the role of B28 and B2^ on self-association. The ability to design an active insulin with diminished selfassociation is important for future diabetes therapy. All commercial pharmaceutical formulations contain insulin in the self-associated state and predominantly in the hexamer form (Blundell et al., 1972). Current models propose that the rate limiting step in absorption of insulin from a subcutaneous injection site is the dissociation to monomer which is readily absorbed (Binder, 1983). Indeed, workers at both the Novo Research Institute and the Lilly Research Laboratories have shown that 527
Downloaded from http://peds.oxfordjournals.org/ at Cornell University Library on December 5, 2012
The importance of Pro828 and LysB29 on the self-association of insulin was established by systematically truncating the C terminus of the B chain. The relationship between structure and association was further explored by making numerous amino acid replacements at B and B29. Association was studied by circular dichroism, size-exclusion chromatography and ultracentrifugation. Our results show that the location of a prolyl residue at B28 is critical for high-affinity selfassociation. Removal of Pro 828 in a series of C-terminal truncated insulins, or amino acid replacement of Pro828, greatly reduced association. The largest disruption to association was achieved by replacing LysB29 with Pro and varying the amino acid at B28. Several of the analogs were predominantly monomers in solutions up to 3 nig/ml. These amino acid substitutions decreased association by primarily disrupting the formation of dimers. Such amino acid substitutions also substantially reduced the Zn-induced insulin hexamer formation. The formation of monomeric insulins through amino acid replacements was accompanied by conformational changes that may be the cause for decreased association. It is demonstrated that self-association of insulin can be drastically altered by substitution of one or two key amino acids.
aggregation to hexamers which is strongly dependent on pH. Two classes of Zn binding sites have been identified. One class of binding site demonstrates the strongest binding, with an association constant of 105—106 M~' and a stoichiometry of two Zn atoms per hexamer (Summerell et al., 1965; Grant et al., 1972; Goldman and Carpenter, 1974). The two Zn atoms are hexacoordinate with three coordinations coming from imidazole moieties of His810 from three monomers in different dimers (Adams et al., 1969; Bradbury et al., 19811) and the remaining coordination site being occupied by water or small anions (Dunn et al., 1980). A weak Zn binding site exists with an association constant of 10 3 -10 4 M~' and is capable of binding several atoms of Zn per insulin monomer (Goldman and Carpenter, 1974). The most detailed atomic information concerning association comes from X-ray crystallography. A variety of crystal forms of insulin have been studied. The N and C termini of the B chain undergo significant conformational changes! between the different crystal forms (Bentley et al., 1976; Chothia et al., 1983; Baker et al., 1988; Derewenda et al., 1989; Badger et al., 1991). How the conformational flexibility of insulin, determined from X-ray crystallography, relates to the self-association in solution is of considerable importance. The crystal structure of the 2-Zn hexameric form and thermodynamic studies of insulin dimerization indicate that both hydrophobic interactions and hydrogen bonding are responsible for stabilizing association (Fredericq, 1956; Jeffrey and Coates, 1966; Pocker and Biswas, 1981). Figure 1 illustrates the insulin dimer-forming surface. X-ray crystallography results demonstrate that most of the nonpolar dimer contacts involve the C-terminal end of the B chain, with B 23 " 26 and B28 being the most predominant (Baker et al., 1988). Association of the dimer is secured by a small antiparallel /3-sheet of hydrogen bonds involving residues B24~26. Formation of the dimer packs Pro 828 against residues B2O~23 which are in a /3-turn (Baker et al., 1988). The hexamer is mainly stabilized through the Zn coordination but additional polar and non-polar residues are buried between the dimers as a result of hexamer assembly (Baker et al., 1988).
D.N.Brems et at.
monomeric insulin analogs act more rapidly than current formulations (Brange etal., 1991; DiMarchi etal., 1992). Materials and methods Zinc-free biosynthetic human insulin was obtained from Eli Lilly & Co. (Indianapolis, IN). Most human insulin analogs were prepared by coupling DesB23~30 porcine insulin (Bromer and Chance, 1967) with a synthetic peptide of desired sequence using a trypsin-catalyzed semisynthesis method in a mixed organic solvent system (Kubiak and Cowburn, 1986). w-tertButoxycarbonyl (tert-BOC) amino acids were purchased from Applied Biosystems Inc. (Foster City, CA). All chemicals were analytical grade or higher. Three insulin analogs were prepared by a chain combination method (Chance et al., 1981) and two by digestion of porcine insulin with either carboxypeptidase A (Schmidtt and Gattner, 1978) or pepsin (Gattner, 1975). Protein concentrations were determined by UV absorbance using extinction coefficients of: 1.05 for insulin (Frank et al., 1972) and LysB28ProB29 insulin; 1.06 for Asp82* insulin and LysB28 insulin; 1.07 for Ala828 insulin, Pro 82 ' insulin, Ala828 Pro829 insulin and Asp^Pro 8 2 9 insulin; 1.08 for Des830 insulin; 1.11 for Des B29 " 30 insulin; 1.13 for DesB28~30 insulin; 1.15 for DesB27"30 insulin; and 0.91 for Des 826 " 30 insulin, at 278 nm for a 1 mg/ml solution in a 1 cm path length. Preparation of analogs Insulin analogs are denoted by the replacement amino acid appearing first followed in superscript by its location in the A or B chain and sequence position. AlaB28ProB29 insulin, AspB28Pro829 insulin, LysB28Pro829 insulin, Pro 829 insulin, DesB29~30 insulin, Des B28 " 30 insulin and DesB27~30 insulin were 528
prepared by coupling Des823 30 porcine insulin with appropriate synthetic peptides using trypsin-catalyzed semisynthesis. Ala828 insulin, Asp828 insulin and Lys828 insulin were prepared by chain combination using recombinant DNA-derived human insulin A chain and the appropriate synthetic human insulin B chain analog. An Applied Biosystems Model 430A peptide synthesizer was used to obtain the desired peptides. Des830 insulin was prepared by carboxypeptidase A digestion of porcine insulin and Des 826 " 30 insulin was obtained by a selective pepsin digestion on porcine insulin. Each insulin analog was purified by size-exclusion and reversed-phase HPLC and characterized by quantitative amino acid analysis, fast atom bombardment mass spectrometry, analytical reversed-phase HPLC and size-exclusion chromatography. All analogs were at least 85 % pure according to reversed-phase HPLC and contained 0.9. For comparative purposes, the chromatography results were normalized to a scale of zero to one, with zero representing the smallest slope for the least aggregating analog and one representing the largest slope for the most aggregating. For the experiment containing a mixture of proteins, the chromatography conditions utilized a mobile phase of 0.04 M Tris, 0.01 M NaCl, pH 8.0, a YMC-Pack PVASil-120 column, 0.50 j*l injection volume, 0.5 ml/min flow rate, ambient temperature and detection at 280 nm. Equilibrium ultracentrifugation All ultracentrifuge studies were obtained in 50 mM NaCl, 50 mM Na2HPO4, pH 7.2, and utilized a Spinco Model E analytical ultracentrifuge at 22°C with a photoelectric scanning optical system. Data on the scanner chart recording were converted to concentration versus radius using the internal calibration factors of the scanner. Over-speeding techniques were used to reach equilibrium in 0.5 mg/ml (Summerell et al., 1965; Grant et al., 1972; Goldman and Carpenter, 1974). AspB28ProB29 insulin, LysB28ProB29 insulin and AlaB28ProB29 insulin show little or no Zn-induced association. Pro829 insulin, LysB28 insulin, Asp828 insulin and Ala828 insulin demonstrate significant Zn-induced association, but less than Zn-insulin. The effect of protein concentration on the size-exclusion results for Zn-free LysB28Pro829 insulin and insulin is illustrated in Figure 3(A). In Figure 3(A), insulin and LysB28Pro829 insulin were mixed together in equal proportions and varying concentrations were injected onto the column. At low protein concentrations (
Altering the association properties of insulin by amino acid replacement
David N.Brems, Leila A.Alter, Michael J.Beckage, Ronald E.Chance1, Richard D.DiMarchi1, L.Kenney Green1, Harlan B.Long1, Allen H.Pekar, James E.Shields' and Bruce H.Frank Parenteral Products Research and Development and 'Diabetes Research. Eli Lilly & Co., Indianapolis, IN 46285, USA
Key words: circular dichroism/mutagenesis/self-association/ size-exclusion/ultracentrifugation
Introduction Insulin association is well documented but not fully understood. It has been studied by ultracentrifugation (Jeffrey and Coates, 1966; Pekar and Frank, 1972; Goldman and Carpenter, 1974; Mark et al., 1987; Mark and Jeffrey, 1990), light scattering (Bohidar and Geissler, 1984), equilibrium dialysis (Grant et al., 1972), difference UV absorbance spectroscopy (Rupley et al., 1967) and CD (Goldman and Carpenter, 1974; Pocker and Biswas, 1981; Melberg and Johnson, 1990; Roy et al., 1990). The insulin association behavior is known to be complex, with the metal-free species exhibiting a pH, ionic strength and protein concentration-dependent association pattern consisting of monomer, dimer, tetramer and higher ordered polymers all in dynamic equilibrium (Fredericq, 1956; Jeffrey and Coates, 1966; Pekar and Frank, 1972; Goldman and Carpenter, 1974; Jeffrey et al., 1976; Pocker and Biswas, 1981). The association constant for the metal-free porcine monomer-dimer equilibrium at pH 7.0 is 1.4-7.5 x 105 NT 1 (Pekar and Frank, 1972; Jeffrey et al., 1976; Pocker and Biswas, 1981; Strazza et al., 1985). Insulin is capable of binding divalent metal ions, with Zn being the most common ligand. Zinc-binding induces a specific © Oxford University Press
The importance of the C terminus of the B chain is further evidenced by removal of B 26 ~ 30 in despentapeptide insulin, which does not significantly alter the rest of the molecule but abolishes dimerization (Bi et al., 1984). In this study the C terminus of the B chain was systematically truncated and it was established that Pro828 is critical to the stabilization of insulin self-association in solution. Amino acid replacement was used to further investigate the role of B28 and B2^ on self-association. The ability to design an active insulin with diminished selfassociation is important for future diabetes therapy. All commercial pharmaceutical formulations contain insulin in the self-associated state and predominantly in the hexamer form (Blundell et al., 1972). Current models propose that the rate limiting step in absorption of insulin from a subcutaneous injection site is the dissociation to monomer which is readily absorbed (Binder, 1983). Indeed, workers at both the Novo Research Institute and the Lilly Research Laboratories have shown that 527
Downloaded from http://peds.oxfordjournals.org/ at Cornell University Library on December 5, 2012
The importance of Pro828 and LysB29 on the self-association of insulin was established by systematically truncating the C terminus of the B chain. The relationship between structure and association was further explored by making numerous amino acid replacements at B and B29. Association was studied by circular dichroism, size-exclusion chromatography and ultracentrifugation. Our results show that the location of a prolyl residue at B28 is critical for high-affinity selfassociation. Removal of Pro 828 in a series of C-terminal truncated insulins, or amino acid replacement of Pro828, greatly reduced association. The largest disruption to association was achieved by replacing LysB29 with Pro and varying the amino acid at B28. Several of the analogs were predominantly monomers in solutions up to 3 nig/ml. These amino acid substitutions decreased association by primarily disrupting the formation of dimers. Such amino acid substitutions also substantially reduced the Zn-induced insulin hexamer formation. The formation of monomeric insulins through amino acid replacements was accompanied by conformational changes that may be the cause for decreased association. It is demonstrated that self-association of insulin can be drastically altered by substitution of one or two key amino acids.
aggregation to hexamers which is strongly dependent on pH. Two classes of Zn binding sites have been identified. One class of binding site demonstrates the strongest binding, with an association constant of 105—106 M~' and a stoichiometry of two Zn atoms per hexamer (Summerell et al., 1965; Grant et al., 1972; Goldman and Carpenter, 1974). The two Zn atoms are hexacoordinate with three coordinations coming from imidazole moieties of His810 from three monomers in different dimers (Adams et al., 1969; Bradbury et al., 19811) and the remaining coordination site being occupied by water or small anions (Dunn et al., 1980). A weak Zn binding site exists with an association constant of 10 3 -10 4 M~' and is capable of binding several atoms of Zn per insulin monomer (Goldman and Carpenter, 1974). The most detailed atomic information concerning association comes from X-ray crystallography. A variety of crystal forms of insulin have been studied. The N and C termini of the B chain undergo significant conformational changes! between the different crystal forms (Bentley et al., 1976; Chothia et al., 1983; Baker et al., 1988; Derewenda et al., 1989; Badger et al., 1991). How the conformational flexibility of insulin, determined from X-ray crystallography, relates to the self-association in solution is of considerable importance. The crystal structure of the 2-Zn hexameric form and thermodynamic studies of insulin dimerization indicate that both hydrophobic interactions and hydrogen bonding are responsible for stabilizing association (Fredericq, 1956; Jeffrey and Coates, 1966; Pocker and Biswas, 1981). Figure 1 illustrates the insulin dimer-forming surface. X-ray crystallography results demonstrate that most of the nonpolar dimer contacts involve the C-terminal end of the B chain, with B 23 " 26 and B28 being the most predominant (Baker et al., 1988). Association of the dimer is secured by a small antiparallel /3-sheet of hydrogen bonds involving residues B24~26. Formation of the dimer packs Pro 828 against residues B2O~23 which are in a /3-turn (Baker et al., 1988). The hexamer is mainly stabilized through the Zn coordination but additional polar and non-polar residues are buried between the dimers as a result of hexamer assembly (Baker et al., 1988).
D.N.Brems et at.
monomeric insulin analogs act more rapidly than current formulations (Brange etal., 1991; DiMarchi etal., 1992). Materials and methods Zinc-free biosynthetic human insulin was obtained from Eli Lilly & Co. (Indianapolis, IN). Most human insulin analogs were prepared by coupling DesB23~30 porcine insulin (Bromer and Chance, 1967) with a synthetic peptide of desired sequence using a trypsin-catalyzed semisynthesis method in a mixed organic solvent system (Kubiak and Cowburn, 1986). w-tertButoxycarbonyl (tert-BOC) amino acids were purchased from Applied Biosystems Inc. (Foster City, CA). All chemicals were analytical grade or higher. Three insulin analogs were prepared by a chain combination method (Chance et al., 1981) and two by digestion of porcine insulin with either carboxypeptidase A (Schmidtt and Gattner, 1978) or pepsin (Gattner, 1975). Protein concentrations were determined by UV absorbance using extinction coefficients of: 1.05 for insulin (Frank et al., 1972) and LysB28ProB29 insulin; 1.06 for Asp82* insulin and LysB28 insulin; 1.07 for Ala828 insulin, Pro 82 ' insulin, Ala828 Pro829 insulin and Asp^Pro 8 2 9 insulin; 1.08 for Des830 insulin; 1.11 for Des B29 " 30 insulin; 1.13 for DesB28~30 insulin; 1.15 for DesB27"30 insulin; and 0.91 for Des 826 " 30 insulin, at 278 nm for a 1 mg/ml solution in a 1 cm path length. Preparation of analogs Insulin analogs are denoted by the replacement amino acid appearing first followed in superscript by its location in the A or B chain and sequence position. AlaB28ProB29 insulin, AspB28Pro829 insulin, LysB28Pro829 insulin, Pro 829 insulin, DesB29~30 insulin, Des B28 " 30 insulin and DesB27~30 insulin were 528
prepared by coupling Des823 30 porcine insulin with appropriate synthetic peptides using trypsin-catalyzed semisynthesis. Ala828 insulin, Asp828 insulin and Lys828 insulin were prepared by chain combination using recombinant DNA-derived human insulin A chain and the appropriate synthetic human insulin B chain analog. An Applied Biosystems Model 430A peptide synthesizer was used to obtain the desired peptides. Des830 insulin was prepared by carboxypeptidase A digestion of porcine insulin and Des 826 " 30 insulin was obtained by a selective pepsin digestion on porcine insulin. Each insulin analog was purified by size-exclusion and reversed-phase HPLC and characterized by quantitative amino acid analysis, fast atom bombardment mass spectrometry, analytical reversed-phase HPLC and size-exclusion chromatography. All analogs were at least 85 % pure according to reversed-phase HPLC and contained 0.9. For comparative purposes, the chromatography results were normalized to a scale of zero to one, with zero representing the smallest slope for the least aggregating analog and one representing the largest slope for the most aggregating. For the experiment containing a mixture of proteins, the chromatography conditions utilized a mobile phase of 0.04 M Tris, 0.01 M NaCl, pH 8.0, a YMC-Pack PVASil-120 column, 0.50 j*l injection volume, 0.5 ml/min flow rate, ambient temperature and detection at 280 nm. Equilibrium ultracentrifugation All ultracentrifuge studies were obtained in 50 mM NaCl, 50 mM Na2HPO4, pH 7.2, and utilized a Spinco Model E analytical ultracentrifuge at 22°C with a photoelectric scanning optical system. Data on the scanner chart recording were converted to concentration versus radius using the internal calibration factors of the scanner. Over-speeding techniques were used to reach equilibrium in 0.5 mg/ml (Summerell et al., 1965; Grant et al., 1972; Goldman and Carpenter, 1974). AspB28ProB29 insulin, LysB28ProB29 insulin and AlaB28ProB29 insulin show little or no Zn-induced association. Pro829 insulin, LysB28 insulin, Asp828 insulin and Ala828 insulin demonstrate significant Zn-induced association, but less than Zn-insulin. The effect of protein concentration on the size-exclusion results for Zn-free LysB28Pro829 insulin and insulin is illustrated in Figure 3(A). In Figure 3(A), insulin and LysB28Pro829 insulin were mixed together in equal proportions and varying concentrations were injected onto the column. At low protein concentrations (
