EA043899B1 - ANTIGEN-BINDING PROTEINS CHARACTERIZED BY LOW VISCOSITY AND METHODS OF THEIR PRODUCTION - Google Patents

Abstract

The invention relates to a method for producing antigen-binding proteins characterized by reduced viscosity. The method is carried out by replacing residues in variable domain subfamilies characterized by high viscosity with residues from subfamilies correlated with low viscosity. The method further comprises replacing residues in the Fc domain with residues associated with low viscosity and adding charged residues to the C-terminus of the Fc domain. The invention further relates to antigen-binding proteins obtained using this method.

Description

Cross reference to related applications

This application claims priority to U.S. Provisional Patent Application No. 62/546,469, filed Aug. 16, 2017; US Provisional Patent Application No. 62/430773, filed December 06

2016; and U.S. Provisional Patent Application No. 62/401,770, filed September 29, 2016, each of which is incorporated herein by reference in its entirety for all purposes.

Field of technology to which the invention relates

This invention relates to biopharmaceuticals, in particular to therapeutic antigen-binding proteins, methods of their use, pharmaceutical compositions based on them and methods for their preparation. In particular, the present invention relates to antigen binding proteins, in particular to antibodies mutated to reduce viscosity.

Brief description of the sequence listing

Incorporated herein by reference in its entirety is the sequence listing entitled A-2063-WO-PCT_ST25, containing SEQ ID NO: 1 through SeQ ID NO: 383, which includes the nucleic acid sequences and/or amino acid sequences disclosed herein document. The sequence listing is provided herein in ASCII text format via the EFS file system and thus constitutes both paper and machine-readable form. The sequence listing was first generated using PatentIn on August 16, 2017 and is 4.32 MB in size.

Prerequisites for creating the invention

Monoclonal antibodies (mAbs) are currently the most popular type of modern therapeutic proteins on the market and in development. Differences between antibodies primarily lie in the antigen-binding domains or complementarity determining regions (CDRs). These differences in CDR are thought to result in differences in the propensity for transient protein-protein interactions that manifest as bulk solution viscosity. Several groups have described the presence of reversible antibody clusters in viscous antibody solutions (primarily dimers). Several theoretical descriptions of polymer viscosity have been proposed to explain the interactions of these clusters as a mechanism for the behavior of bulk solution viscosity.

Antibodies typically act as antagonists and therefore require large amounts, often in excess of 100 mg per dose, to block unwanted interactions. For patient convenience, a single 1 ml subcutaneous injection is the most preferred route of administration. The need to administer large quantities of mAbs in a relatively small volume requires high concentration formulations equal to or greater than 100 mg/mL. Antibodies are large biopolymers with molecular weight values of approximately 150 kDa, and their high concentrations result in high shear stress and high viscosity due to protein-protein and protein-wall interactions during filtration and passage through injection needles and into the subcutaneous space. High viscosity poses problems during the production of therapeutic antigen-binding proteins, as well as during their administration to patients, including excessively high back pressure during injections leading to malfunction of injection devices, difficulty in manual administration, reduced bioavailability, and patient discomfort.

The development and use of high-concentration therapeutic monoclonal antibody solutions has accelerated as the cost of producing biopharmaceuticals has decreased. In some cases, these antibody solutions have the characteristics of viscous solutions, which may make it difficult to manufacture and administer the intended dose. The differences in the CDRs, which appear to determine whether an antibody is viscous or non-viscous, are likely due to the propensity of the CDRs to control protein-protein interactions.

Significant efforts are being made in the industry to understand the nature of the interactions leading to high viscosity and to reduce the viscosity of highly viscous antibody formulations. The most important parameters affecting the viscosity of antibody formulations include the following.

Intermolecular interactions determined by protein pI and solution pH. Cheng et al. (2013), Linking the solution viscosity of an IgG2 monoclonal antibody to its structure as a function of pH and temperature, J. Pharm Sci., 102:4291-4304.

Charge interactions. Yadav et al. (2012), Viscosity behavior of high-concentration monoclonal antibody solutions: correlation with interaction parameter and electroviscous effects, J. Pharm Sci., 101:998-1011; Yadav et al. (2012), The influence of charge distribution on self-association and viscosity behavior of monoclonal antibody solutions, Mol. Pharm., 9(4):791-802; Singh et al. (2014), Dipole-Dipole Interaction in Antibody Solutions: Correlation with Viscosity Behavior at High Concentration, Pharm Res., 31(9):2549-2558; Chaudhri et al. (2013), The role of amino acid sequence in the self-association of therapeutic monoclonal antibodies: insights from coarse-grained modeling, J. Phys. Chem. In, 117:1269–1279.

Hydrophobic interactions. Guo et al. (2012), Structure-activity relationship for hydrophobic salts as viscosity-lowering excipients for concentrated solutions of monoclonal antibodies, Pharm. Res., 29:3102–3109.

The highest viscosity of the solution was observed under conditions with the most negative diffusion interaction parameter kD, the largest apparent radius and the smallest total charge. Neergaard et al. (2013), Viscosity of high concentration protein formulations of monoclonal antibodies of the IgGl and IgG4 subclass - prediction of viscosity through protein-protein interaction measurements, Eur. J. Pharm. Sci., 49:400–410. The diffusion interaction parameter (kD), a component of the second virial osmotic coefficient (B(2)), correlates well (R>0.9) with the viscosity of concentrated mAb solutions, while the total charge of mAb correlates weakly (R<0.6) , indicating that weak intermolecular interactions are important for controlling the viscoelastic behavior of concentrated mAb solutions. Connolly, et al. (2012), Weak interactions govern the viscosity of concentrated antibody solutions: high-throughput analysis using the diffusion interaction parameter, Biophys. J., 103:69-78. The study reported here used primary sequences associated with 3D structure. See Honegger et al. (2001), Yet another numbering scheme for immunoglobulin variable domains: an automatic modeling and analysis tool, J. Mol. Biol., 309:657-670. The viscosity values of selected mAb molecules were measured to develop a mAb viscosity prediction model using machine learning algorithms. Structural location, charge, and hydrophobicity were the main amino acid parameters used for the model.

The viscosity of monoclonal antibodies was estimated using molecular information from the following articles: Li, L. et al. (2014), Concentration dependent viscosity of monoclonal antibody solutions: explaining experimental behavior in terms of molecular properties, Pharm. Res., 31:3161–3178; and Sharma et al. (2014), In silico selection of therapeutic antibodies for development: viscosity, clearance, and chemical stability, Proc. Natl. Acad. Sci., USA, 111:18601-6.

The end result of interactions between antibodies is either an extended temporary network of interactions (percolating network) that results in the formation of a viscous solution, or the formation of larger oligomers, which then somehow influence the rheology of the solution into larger structures. In the studies reported herein, small quantities of viscous antibodies were used as subjects for biochemical and biophysical analysis in an attempt to identify specific protein-protein interactions that may result in a viscous antibody solution.

The Aho numbering approach has been used in the past to improve stability and other biophysical properties. Ewert et al. (2003), Structure-based improvement of the biophysical properties of immunoglobulin VH domains with a generalizable approach, Biochemistry, 42:1517-1528; Ewert et al. (2003), Biophysical properties of human antibody variable domains, J. Mol. Biol., 325:531-553; Ewert et al. (2004), Stability improvement of antibodies for extracellular and intracellular applications: CDR grafting to stable frameworks and structure-based framework engineering, Methods, 34:184-199; and Rothlisberger et al. (2005), Domain interactions in the Fab fragment: a comparative evaluation of the single-chain Fv and Fab format engineered with variable domains of different stability, J Mol. Biol., 347:773–789. The Aho numbering system has also been used in the past to reduce the tendency for aggregation. Borras et al. (2013), US Patent No. 8545849.

Brief description of the invention

This invention relates to methods for reducing the viscosity of antigen binding proteins by modifying sequences in framework regions and/or the Fc domain that have been shown to be associated with high viscosity.

In the following detailed description of the method, all variable region amino acids are identified according to Aho numbering, all amino acids from conserved regions are identified according to EU numbering. Aho numbering is aligned with and correlates with other major numbering schemes, including EU (Edelman et al. (1969), The covalent structure of an entire gamma immunoglobulin molecule, Proc. Natl. Acad. Sci., USA, 63, 78-85), Kabat (Kabat et al. (1991), Sequences of proteins of immunological interest, Fifth Edition. NIH Publication No. 91-3242), Chothia (Chothia et al., (1992), Structural repertoire of the human VH segments, J. Mol Biol., 227:799-817); (Tomlinson et al., (1995), The structural repertoire of the human V kappa domain, EMBO J., 14:4628-4638). Any of the four numbering systems may be used interchangeably to identify the preferred amino acid substitutions described herein.

If the antigen binding protein is of the germline VH1|1-18 subfamily, the method includes modifying the VH1 sequence to include one or more substitutions selected from 82X ¹ , 94X ² and 95X ³ , where X ¹ represents basic residues (R, K or H) , X ² represents a polar uncharged residue (S, T, N or Q) and X ³ represents a basic residue (R, K or H). All residues are identified according to the Aho numbering system. The preferred germline mutations for the VH1|1-18 subfamily are 82R, 94S, and 95R. The method applied to the germline VH1|1-18 subfamily may further include a 59X ²⁰ substitution, where X ²⁰ is a basic residue (R, K or H), with the 59K mutation being preferred.

If the antigen binding protein is of the germline VH3|3-33 subfamily, the method includes modifying the VH3 sequence to include one or more substitutions selected from ^1X4 , ^17X5 and ^85X6 , where ^X4 is a negatively charged residue (D or E), X ⁵ represents a small hydrophobic residue (G, A, V, I, L or M) and X ⁶ represents a small hydrophobic residue (G, A, V, I, L or M). All residues are identified according to the Aho numbering system.

- 2 043899

The preferred germline mutations for the VH3|3-33 subfamily are 1E, 17G and 85A.

If the antigen binding protein is of the germline VK3|L16 subfamily, the method includes modifying the VK3 sequence to include one or more substitutions selected from ^4X10 , ^13X11 , 76X12, ^78F , ^95X13 , ^97X14 , and 98P, where ^X10 is selected from G , A, V, I, L and M; X ¹¹ is selected from G, A, V, I, L and M, X ¹² is selected from D and E, X ¹³ is selected from R, K and H; and X ¹⁴ selected from D and E. All residues are identified according to the Aho numbering system. The preferred germline mutations for the VK3|L16 subfamily are 4L, 13L, 76D, 95R, 97E, and 98P.

If the antigen binding protein is of the germline VK3|L6 subfamily, the method includes modifying the VK3 sequence to include one or more substitutions selected from 76X ¹² and 95X ¹³ . The preferred germline mutations for the VK3|L6 subfamily are 76D and 95R.

The methods of the present invention further include modifying the Fc domain to include one or more substitutions selected from ^253X15 , ^440X16 and ^439X17 , where ^X15 is a small hydrophobic residue (G, A, V, I, L or M), X ¹⁶ represents a basic residue (R, K or H) and X ¹⁷ represents a negatively charged residue (D or E), where the Fc domain sequence includes only one of 440X ¹⁶ and 439X ¹⁷ . All residues are identified according to the EU numbering system. The preferred Fc domain mutations are 253A, 440K and 439E.

The methods of the present invention further include modifying the C-terminus of the Fc domain sequence to include X ¹⁸ X ¹⁹ , where X ¹⁸ represents one to four amino acids selected from D and E or from H, K and R; and X ¹⁹ is selected from P, M, G, A, V, I, L, S, T, N, Q, F, Y and W and is absent if X ¹⁸ contains D or E, present if X ¹⁸ contains K or R in its C-terminal region, and is present or absent if X ¹⁸ contains H in its C-terminal region. Preferred C-terminal Fc modifications include KP, KKP, KKKP (SEQ ID NO:380), E or EE at the C terminus.

The above methods are preferably applied to the highly viscous antibodies shown in FIG. 1A and 1B below. The portion of the method using the VH1|1-18 sequence is preferably applied to the antibodies AF, AK, AL, AN and AO shown in FIG. 1B. The portion of the method using the sequence VH3|3-33 is preferably applied to the antibodies AQ, AM, AI and AG shown in FIG. 1B. The VK3|L16 sequence portion of the method is preferably applied to the AF and AQ antibodies shown in FIG. 1B. The part of the method that uses the sequence VK3 |L6 is preferably applied to the antibody AJ.

The methods of the present invention further include a method of producing an antigen binding protein that reaches a maximum concentration in serum faster than the parent antibody if the antigen binding protein and the parent antibody are administered at the same concentration, which includes introducing a 440X ¹⁶ sequence modification into the parent antibody, wherein X ¹⁶ is selected from R, K and H. In a preferred method, the sequence-modified antigen binding protein reaches peak serum concentration after subcutaneous injection at least twice as fast as the parent antibody. Also included within the scope of the present invention is a method for producing an antigen binding protein that achieves a maximum concentration in serum after subcutaneous injection that is higher than the concentration of the parent antibody if the antigen binding protein and the parent antibody are administered at the same concentration, which includes introducing a 440X ¹⁶ sequence modification into the parent antibody, wherein X ¹⁶ is selected from R, K, and H. In a preferred method, the sequence-modified antigen binding protein achieves a peak serum concentration that is at least about 25% higher than the concentration of the parent antibody. In each of these methods, the preferred X ¹⁶ is K, and the preferred parent antibody is PCSK9 polypeptide (AK antibody is most preferred).

The present invention further relates to a mutant antigen binding protein that contains one or more sequences selected from

a) a VH1|1-18 germline subfamily sequence containing one or more substitutions selected from 82X ¹ , 94X ² and 95X ³ , where X ¹ is selected from R, K and H; X ² is selected from S, T, N and Q; and X ³ is selected from R, K and H;

b) a germline VH3|3-33 subfamily sequence containing one or more substitutions selected from 1X ⁴ , 17X ⁵ and 85X ⁶ , where X ⁴ is selected from D and E; X ⁵ selected from G, A, V, I, L and M; and X ⁶ is selected from G, A, V, I, L and M;

c) germline subfamily VK3|L16 containing one or more substitutions selected from 4X ¹⁰ , 13X ¹¹ , 76X ¹² , 78F, 95X ¹³ , 97X ¹⁴ and 98P, where X ¹⁰ is selected from G, A, V, I, L them; X ¹¹ is selected from G, A, V, I, L and M; X ¹² is selected from D and E, X ¹³ is selected from R, K and H; and X ¹⁴ is selected from D and E, wherein the mutant antigen binding protein does not contain only the 78F substitution;

d) germline VK3 |IL6 subfamily containing one or more substitutions selected from 76X ¹² and 95X ¹³ ;

e) an Fc domain sequence containing one or more substitutions selected from 253X ¹⁰ ,

- 3 043899

440X ¹¹ and 439X ¹² , where X ¹⁰ is selected from G, A, V, I, L and M; X ¹¹ is selected from R, K and H; and X ¹² is selected from D and E, wherein the antigen binding protein contains at least one of 253X ¹⁵ or modifications selected from a, b, c, d and f, if X ¹⁶ is K; and X ¹⁷ is E and antigen binding protein specifically binds CD20; And

f) an Fc domain sequence containing at the C-terminus X ¹⁸ X ¹⁹ , where X ¹⁸ represents one to four amino acids selected from D and E or from H, K and R, and X ¹⁹ is selected from P, M, G, A, V, I, L, S, T, N, Q, F, Y and W and absent if X ¹⁸ contains D or E, present if X ¹⁸ contains K or R in its C-terminal region, and is present or absent if X ¹⁸ contains an H at its C-terminal region, and wherein the antigen binding protein contains at least one of 253X ¹⁵ or substitutions selected from a) to e), if PGKP ( SEQ ID NO: 381), PGKKP (SEQ ID NO: 382), PGKKKP (SEQ ID NO: 383) or PGE, and the antigen binding protein specifically binds CD20 or CD38. Preferred C-terminal Fc modifications include KP, KKP, KKKP (SEQ ID NO:380), E or EE at the C terminus, wherein the variable region amino acids are numbered according to the Aho numbering system and all amino acids from the conserved regions, including Fc, correspond to EU numbering.

Preferred antigen binding proteins of the present invention are those where the above modifications are applied to the antibodies shown in FIG. 1A and 1B below. Also preferred are antigen binding proteins in which the VH1|1-18 germline subfamily sequence contains one or more substitutions selected from 82R, 94S and 95R, with antigen binding proteins having all of these substitutions being most preferred;

the germline VH3|3-33 subfamily sequence contains one or more of substitutions 1E, 17G, and 85A, with antigen binding proteins having all of these substitutions being most preferred;

the VK3|L16 germline subfamily sequence contains one or more substitutions selected from 4L, 13L, 76D, 95R, 97E and 98P, with antigen binding proteins having all of these substitutions being most preferred;

the germline VK3|L6 subfamily sequence contains one or more substitutions selected from 76D and 95R, with antigen binding proteins having all of these substitutions being most preferred;

the Fc domain sequence contains one or more substitutions selected from 253A, 440K and 439E, with antigen binding proteins having all of these substitutions being most preferred; And

The C-terminus of the Fc domain contains a sequence selected from KP, KKP, KKKP (SEQ ID NO: 380) and E.

All of the above preferred amino acid substitutions in the variable regions are identified according to the Aho numbering system. All residues in conserved regions, including Fc, are identified according to the EU numbering system.

Preferred antigen binding proteins of the present invention include the antibodies AF, AK, AL, AN and AO shown in FIG. 1B, having one or more, most preferably all, of the above germline VH1|1-18 subfamily substitutions;

antibodies AQ, AM, AI and AG shown in FIG. 1B having one or more, most preferably all, of the VH3|3-33 germline subfamily substitutions;

AF and AQ antibodies having one or more, most preferably all, of the VK3|L16 germline subfamily substitutions;

antibody AJ shown in FIG. 1B having one or more, most preferably all, of the VK3 |L6 germline subfamily substitutions; and antibodies BA, AN and AN shown in FIG. 1B having one or more, preferably all, of the Fc substitutions listed above.

Because of the above sequence modifications, the present invention further provides antigen binding proteins that specifically bind to PCSK9 containing a heavy chain sequence selected from SEQ ID NOs: 352, 353, 354, 3666 and 368, preferably also containing a light chain sequence of SEQ ID NO: 351.

Because of the above sequence modifications, the present invention also provides antigen binding proteins that specifically bind c-fms comprising a heavy chain sequence selected from SEQ ID NO: 356, 357 and 358, preferably further comprising a light chain sequence of SEQ ID NO: 355 .

Because of the above sequence modifications, the present invention also provides antigen binding proteins that specifically bind GIPR comprising a heavy chain sequence selected from SEQ ID NOs: 359, 361, 362, 364 and 368, preferably further comprising a light chain sequence selected from SEQ ID NO: 360, 363, 365 and 367.

All modified antigen binding proteins are applicable for the same indications as

- 4 043899 are described previously for unmodified antibodies.

Each of the antigen binding proteins in FIG. 1A and 1B, characterized by mutated heavy chains, are preferred to further include a light chain sequence as indicated in the unmodified parent antibody in FIG. 1A and 1B. Each of the above-mentioned antigen binding proteins characterized by a mutated light chain is preferred to further include a heavy chain sequence as noted above or as observed in the unmodified parent antibody in FIG. 1A and 1B.

The present invention further includes the antigen binding proteins described above which have improved pharmacokinetic properties. The present invention includes an antigen binding protein, optionally characterized by any of the above sequence modifications, wherein

a) the antigen binding protein contains a modification of the 440X ¹⁶ sequence compared to the parent antibody, which does not contain a modification of the 440X ¹⁶ sequence;

b) the antigen binding protein reaches its maximum serum concentration after subcutaneous injection faster than the parent antibody if the antigen binding protein and the parent antibody are administered at the same concentration; And

c) the antigen binding protein reaches a maximum serum concentration after subcutaneous injection that is higher than the concentration of the parent antibody if the antigen binding protein and the parent antibody are administered at the same concentration.

The preferred parent antibody for such an antigen binding protein is PCSK9 binding polypeptide, with antibody AK being most preferred. The preferred X ¹⁶ substituent in such an antigen binding protein is K. Further included within the scope of the present invention is a method of treating hypercholesterolemia using such an antigen binding protein.

The present invention also relates to isolated nucleic acids encoding the antigen-binding proteins of the present invention, as well as vectors containing these nucleic acids, host cells containing these vectors, and methods for producing and using antigen-binding proteins.

In other embodiments, the present invention provides compositions containing antigen binding proteins and kits containing antigen binding proteins, as well as articles containing antigen binding proteins.

Brief description of graphic materials

In fig. 1A and 1B show a table of viscosity values measured for IgG1 and IgG2 monoclonal antibodies formulated at a concentration of 150 mg/ml in formulation buffer containing 20 mM acetate and 9% sucrose at pH 5.2 (without polysorbate). In fig. 1A and 1B show the targets of the antibodies under study, as well as their light and heavy chain types, germline subfamilies, concentration, pI and viscosity. Each antibody in FIG. 1A and 1B have the amino acid sequences indicated in the figures and in the sequence listing. The amino acid sequences of the heavy and light chains are encoded by the nucleic acids having SEQ ID NO immediately preceding them in the sequence listing.

In fig. 1C and 1D show sequence identification numbers (SEQ ID NOs) for the framework and Fc regions of the antibodies in FIG. 1A and 1B. In fig. 1D also shows the BA antibody discussed in FIG. 17.

In fig. 2 shows the pairs of high viscosity and low viscosity subtypes defined herein.

In fig. Figure 3 shows the expression results for the AK and AO antibody molecules and their mutants, including relative values for titer, final viable cell density (VCD), and viability when harvested after 7 days of cell culture. AK values are 100%.

In fig. Figure 4 shows the effectiveness of the original AK antibodies (control AK, AK) and their mutants.

In fig. Figure 5 shows the viscosity of the mutants relative to the original antibodies AK and AO. The average viscosity values for high viscosity VH1|1-18 and low viscosity VH1|1-02 from a panel of 43 antibodies are also shown for comparison.

In fig. Figure 6 shows the measured mAb viscosities for the high viscosity VH1|1-18 and low viscosity VH1|1-02 subtypes compared to the calculated pI values for the entire molecule. Low viscosity mutants of AK and AO antibodies are also shown.

In fig. Figure 7 shows the measured mAb viscosities for the high viscosity VH3|3-33 and low viscosity VH3|3-07 subtypes compared to the calculated pI values for the entire molecule. The low viscosity AQ mutant (1, 17, 85) is also shown.

In fig. Figure 8 shows the measured viscosities of mAbs with the high viscosity VK3|L16 and VK3|L6 subfamilies and the low viscosity VK3|A27 subfamily compared to the calculated pI values for the entire molecule. The low viscosity AQ mutant (4, 13, 76, 95, 97, 98) is also shown.

Fig. 9 is a table showing the general parameters of sequences

- 5 043899 mAb subtypes VH1|1-18 with high viscosity and VH1|1-02 with low viscosity (germ lines). In fig. 9 mAbs sorted by viscosity. The table includes their mAb symbols, measured viscosities, calculated pI values, and VL and VH germlines. For VH germlines, VH1|1-18 with higher viscosity are shown in bold and underlined. Heavy chain scaffold 3 sequences are shown. Residues correlated with high viscosity are shown in bold and underlined. VBase sequences are added to the germline VH1|1-18 and VH1|1-02 subfamilies for comparison to illustrate that the various residues are representative residues for these subfamilies.

Fig. 10 is a table showing the obtained and characterized antibody mutants AK and AO.

Fig. 11 is a table showing the general sequence parameters of thirteen VH3 heavy chain mAbs. In fig. Figure 11 shows mAb symbols, measured viscosities, calculated pI values, HC and LC types, and VL and VH subtypes (germ types). The higher viscosity subfamily VH3|3-33 is shown in bold and underlined. Residues correlated with high viscosity are shown in bold and underlined. VBase sequences are added to the germline VH3|3-33 and VH3|3-07 subfamilies for comparison to illustrate that the various residues are representative residues for these subfamilies.

Fig. 12 is a table showing common sequence parameters of fourteen VK3 light chain mAbs. In fig. The 12 mAbs are sorted by viscosity, including their mAb symbols, measured viscosities, calculated pI values, HC and LC types, and VL and VH subtypes (germ types). The higher viscosity subfamilies VK3|L16 and VK3|L6 are shown in bold and underlined. Light chain residues that are consistently different in the VK3|L16 and VK3|L6 subfamilies compared with the VK3|A27 subfamily are shown on the right side. Residues correlated with high viscosity in the VK3|L16 and VK3|L6 subfamilies are shown in bold and underlined. VBase sequences are added to the germline VK3|L16 and VK3|A27 subfamilies for comparison to illustrate that the various residues are representative residues for these subfamilies.

In fig. 13A and 13B show the average values for measured viscosity and calculated pI values for intact antibody molecules comprising the indicated heavy and light chain germline families and the VH1, VH3 and VK3 germline subfamilies. The X-axis includes families and the number of members in each family and subfamily.

In fig. 14A shows that mutations at the Fc-Fc interface can reduce solution viscosity. The dependence of viscosity on the concentration of the AK antibody and the I253A and S440K mutants of the AK antibody is shown.

In fig. 14B shows that a double mutant that restores wild-type complement activity also restores wild-type viscosity. The figure shows the concentration dependence of viscosity of the AK antibody, the I253A mutant of the AK antibody, the S440K mutant of the AK antibody, and the K439E/S440K double mutant of the AK antibody. The figure also shows the viscosity of the AK antibody mutant K439E, the AK antibody mutant H433A, and the AK antibody mutant N434A, which do not have reduced viscosity relative to the parent AK antibody.

Fig. 15 is a table showing the absolute and relative viscosities of the original AK antibody and various mutants. The parent AK antibody and mutants are used in studies of the pharmacokinetics and pharmacodynamics of low viscosity AK antibodies and mutants in non-human primates.

Fig. 16 is a diagram of chemical cross-linking using EDC (see Example 2).

Fig. 17 is a table showing the viscosities of proteins selected for Fc mutations to produce variants having lower viscosity.

In fig. 18 shows the concentration-dependent formation of antibody system oligomers by chemical cross-linking of antibody AN with EDC.

In fig. 19A shows that the S440K mutation in the Fc region reduces the viscosity of the AQ antibody. (Note that the mutant concentration is actually 150 mg/ml.) FIG. 19B is a scatter plot of the same data with exponential fit. Diamonds in Fig. 19B indicates unmodified AQ antibody at the concentrations indicated, and the square indicates the S440K mutant at 150 mg/ml.

Fig. 20A is a table showing antibody AQ mutants produced and characterized, including measured concentration and viscosity.

In fig. 20B shows the cAMP response of 293/huGIPR cells expressing human GIP receptors, activated by GIP and blocked with anti-GIPR antibodies. cAMP activity in vitro was not equally affected by viscosity mutations. The activity remained unchanged within the analytical error limits.

In fig. 21 shows a summary of the experimental design for a single dose subcutaneous bolus pharmacokinetic study in male cynomolgus monkeys, as described in more detail in the worked example below.

In fig. 22 shows the average pharmacokinetic parameter estimates for AK antibody or low viscosity mutant homologues following subcutaneous administration of 10 mg/kg to male cynomolgus monkeys (N=4 males). Introduction of a mutation in the Fc region, which reduces viscosity, reduces _Tmax and increases _Cmax .

In fig. Figure 23 shows the percentage of LDL-C compared to pretest (CLAB day). *Percentage change is expressed as the individual animal post-dose value divided by the pretest value on day 1. All four antibodies (AK parent antibody, AK Fc mutant, AK Fab mutant, and AK double Fc/Fab mutant) induced decreased LDL-C levels.

In fig. 24A and 24B show the pharmacokinetic profile (μg/ml) with the corresponding concentration profile (mg/dL) of low density lipoprotein (LDL) in plasma. LDL concentrations and serum test subject concentrations are presented as means of four animals. Filled circles with a solid line indicate serum concentrations of the parent AK antibody. Open squares with a dashed line indicate serum concentrations of antibody carrying the Fab mutation. Filled triangles with a solid line indicate serum concentrations of the antibody carrying the Fc mutation. Open diamonds with a dotted line indicate serum concentrations of an antibody simultaneously carrying Fab and Fc mutations. These data indicate that the presence of a mutation in the Fc region, which reduces the viscosity of the antibody formulation, results in a decreased time to reach _Tmax and a higher _Cmax . All mutant forms of the parent antibody retain the ability to reduce serum LDL-C levels (lower panel).

Detailed Description of the Invention

Definition of terms.

In the following description, a number of terms are used extensively. To facilitate understanding of the invention, the following definitions are provided.

Unless the text expressly states otherwise, the singular form and the expression at least one are used interchangeably to mean one or more than one.

The expression antigen binding protein refers to a protein or polypeptide that contains an antigen binding region or antigen binding portion that is characterized by strong affinity for another molecule to which it binds (antigen). Antigen-binding proteins include antibodies, peptibodies, antibody fragments (eg, Fab, Fab', F(ab') ₂ , Fv, single domain antibody), antibody derivatives, antibody analogues, fusion proteins and antigen receptors, including chimeric antigen receptors (CARs).

Antibodies (Ab) and immunoglobulins (Ig) are glycoproteins with the same structural characteristics. Although antibodies exhibit binding specificity for a particular antigen, immunoglobulins include both antibodies and other antibody-like molecules that do not have antigen specificity. The latter type of polypeptides, for example, are produced at low levels by the lymphatic system and at elevated levels in myeloma. Thus, as used herein, the term antibody or antibody peptide(s) refers to an intact antibody, an antibody that competes for specific binding with an antibody disclosed herein, or an antigen binding fragment thereof (e.g., Fab, Fab', F(ab ')2, Fv, single domain antibody), which competes with the intact antibody for specific binding and includes chimeric, humanized, fully human and bispecific antibodies. In certain embodiments, antigen binding fragments are produced, for example, using recombinant DNA techniques. In additional embodiments, antigen binding fragments are produced by enzymatic or chemical digestion of intact antibodies. Antigen binding fragments include, but are not limited to, Fab, Fab', F(ab) ² , F(ab') ² , Fv and single chain antibodies. Examples of antibodies suitable for use in the present invention include, but are not limited to, the antibodies listed in FIG. 1A and 1B, as well as abagovomab, abciximab, actoxumab, adalimumab, afelimomab, afutuzumab, alacizumab, alacizumab pegol, ALD518, alemtuzumab, alirocumab, alemtuzumab, altumomab, amatuximab, anatumomab mafenatox, anrukinzumab, apol izumab, arcitumomab, azelizumab, altinumab, atlizumab, atorolimiumab, tocilizumab, bapineizumab, basiliximab, bavituximab, bektumomab, belimumab, benralizumab, bertilimumab, basilesomab, bevacizumab, bezlotoxumab, biciromab, bivatuzumab, bivatuzumab mertansine, blinatumomab, blosozumab, brentuximab vedo tin, briakinumab, brodalumab, canakinumab, cantuzumab mertansine, cantuzumab mertansine, caplacizumab, capromab pendetide, carlumab, catumaxomab, CC49, cedelizumab, certolizumab pegol, cetuximab, cituzumab bogatox, cixutumumab, clazakizumab, clenoliximab, clivatuzumab tetraxetane, conatumumab, crenezumab, CR6261, dacetuzumab, dac lizumab, dalotuzumab, daratumumab, demcizumab, denosumab, detumomab, dorlimomab aritox, drocitumab, duligotumab, dupilumab, ecromeximab, eculizumab, edobacomab, edrecolomab, efalizumab, efungumab, elotuzumab, elsilimomab, enavatuzumab, enlimomab pegol, enokizumab, enokizumab, enoticumab, enoticumab, ens ituximab, epitumomab cituxetan, epratuzumab, erenumab, erlizumab, ertumaxomab , etaracizumab, etrolizumab, evolocumab, exbi- 7 043899 virumab, exbivirumab, fanolesomab, faralimomab, farletuzumab, fasinumab, FBTA05, felvizumab, fezakinumab, ficlatuzumab, figitumumab, flanvotumab, fontolizumab, foralumab, foravirumab, fresolimumab, fulranumab, futuximab, galiximab, ganitumab , gantenerumab, gavilimomab, gemtuzumab ozogamicin, gevokizumab, girentuximab, glembatumumab vedotin, golimumab, gomiliximab, GS6624, ibalizumab, ibritumomab tiuxetan, icrucumab, igovomab, imciromab, imgatuzumab, inclacumab, indatuxim ab ravtansin, infliximab, intetumumab, inolimomab, inotuzumab ozogamicin, ipilimumab, iratumumab, itolizumab, ixekizumab, keliximab, labetuzumab, lebrikizumab, lemalezomab, lerdelimumab, lexatumumab, libivirumab, ligelizumab, lintuzumab, lirilumab, lorvotuzumab mertansine, lucatumumab, lumiliximab, mapatumumab, maslimomab, mavrilimumab , matuzumab, mepolizumab, metelimumab, milatuzumab, minretumomab, mitumomab , mogamulizumab, morolimumab, motavizumab, moxetumomab pasudotox, muromonab-CD3, nacolomab tafenatox, namilumab, naptumomab estafenatox, narnatumab, natalizumab, nebacumab, necitumumab, nerelimomab, nesvacumab, nimotuzumab, nivolumab, nofetumomab merpen tan, ocaratuzumab, ocrelizumab, odulimomab, ofatumumab, olaratumab , olokizumab, omalizumab, onartuzumab, oportuzumab monatox, oregovomab, orticumab, otelixizumab, oxelumab, ozanezumab, ozoralizumab, pagibaximab, palivizumab, panitumumab, panobakumab, parsatuzumab, pascolizumab, pateklizumab, patritumab, pemtumomab, perakizumab, pertuzumab, pexelizumab, pidilizumab, pintumomab, placulumab, ponezumab, prezalumab, priliximab, pritumumab, PRO 140, quilizumab, racotumomab, radretumab, rafivirumab, ramucirumab, ranibizumab, raxibacumab, regavirumab, reslizumab, rilotumumab, rituximab, robatumumab, roledumab, romosozumab , rontalizumab, rovelizumab, ruplizumab, samalizumab, sarilumab , satumomab pandetide, secukinumab, sevirumab, sibrotuzumab, sifalimumab, siltuximab, simtuzumab, siplizumab, sirukumab, solanezumab, solitomab, sonepcizumab, sontuzumab, stamulumab, sulesomab, suvizumab, tabalumab, takatuzumab tetrax tan, tadocizumab, talizumab, tanezumab, taplitumomab paptox, tefibazumab, telimomab aritox, tenatumomab, tefibazumab, telimomab aritox, tenatumomab, teneliximab, teplizumab, teprotumumab, tezepelumab, TGN1412, tremelimumab, ticilimumab, tildrakizumab, tigatuzumab, TNX-650, tocilizumab, toralizumab, tositum omab, tralokinumab, trastuzumab, TRBS07, tregalizumab, tremelimumab, tucotuzumab celmoleukin, tuvirumab, ublituximab, urelumab, urtoxazumab, ustekinumab, vapaliximab, vatelizumab, vedolizumab, veltuzumab, vepalimomab, vesencumab, vizilizumab, volociximab, vorsetuzumab mafodotin, votumumab, zalutumumab, zanoli mumab, zatuximab, ziralimumab and zolimomab aritox.

As used herein, the term isolated antibody refers to an antibody that has been identified, isolated and/or isolated from a component of its natural environment. Contaminants in its natural environment are materials that will interfere with the diagnostic or therapeutic use of the antibody and may include enzymes, hormones, and other protein or non-protein solutes. In preferred embodiments, the antibody will be purified (1) to greater than 95% by weight of the antibody, as determined by the Lowry method, and most preferably to greater than 99% by weight;

(2) to an extent sufficient to obtain at least 15 residues of the N-terminal or internal amino acid sequence using a rotating beaker sequencer; or (3) to homogeneity by SDS-PAGE under reducing or non-reducing conditions using Coomassie blue or preferably silver staining.

An isolated antibody includes the antibody in situ in recombinant cells, since at least one component of the antibody's natural environment will not be present. Typically, however, the isolated antibody will be obtained through at least one purification step.

The term binding of an antigen or other polypeptide includes, without limitation, binding of a ligand polypeptide of the present invention to a receptor;

binding a receptor polypeptide of the present invention to a ligand;

binding an antibody of the present invention to an antigen or epitope;

binding an antigen or epitope of the present invention to an antibody;

binding an antibody of the present invention to an anti-idiotypic antibody; binding an anti-idiotypic antibody of the present invention to a ligand; binding an anti-idiotypic antibody of the present invention to a receptor; binding the anti-idiotypic antibody of the present invention to a ligand, receptor or antibody, etc.

The term immunoglobulin refers to a protein consisting of one or more polypeptides essentially encoded by immunoglobulin genes. One form of immunoglobulin constitutes the basic structural unit of an antibody. This form is a tetramer and consists of two identical pairs of immunoglobulin chains, each containing one light and one heavy chain. In each pair, the variable regions of the light and heavy chains are jointly responsible for binding to the antigen, and the constant regions are responsible for the effector functions of the antibody.

Full-length immunoglobulin light chains (approximately 25 kDa or approximately 214 amino acids) are encoded by a variable region gene at the NH ₂ -terminus (approximately 110 amino acids) and a kappa or lambda constant region gene at the COOH-terminus. Full-length immunoglobulin heavy chains (approximately 50 kDa or approximately 446 amino acids) are similarly encoded by a variable region gene (approximately 116 amino acids) and one of the other aforementioned constant region genes (approximately 330 amino acids). Heavy chains are classified as gamma, mu, alpha, delta, or epsilon and define the isotype of the antibody as IgG (such as IgG1, IgG2, IgG3, and IgG4), IgM, IgA, IgD, and IgE, respectively. In the light and heavy chains, the variable and constant regions are connected by a J region of approximately 12 or more amino acids, with the heavy chain also containing a D region of approximately 10 or more amino acids. (See generally Fundamental Immunology (Paul, W., ed., 2nd edition, Raven Press, NY (1989)), chapter 7 (incorporated by reference in its entirety for all purposes).

The variable domain or region of an immunoglobulin light or heavy chain contains framework regions (FRs) interrupted by complementarity determining regions (CDRs). Kabat et al. (1991), Seguences of Proteins of Immunological Interest, 5th Edition, Public Health Service, National Institutes of Health, Bethesda, Md. (1991); Chothia et al. (1987), J. Mol. Biol. 196: 901-917 (both of which are incorporated herein by reference). FR residues are those variable domain residues that are not CDR region residues as defined herein. The framework sequences of the various light or heavy chains are relatively conserved within species. Thus, a human framework region is a framework region that is substantially identical (about 85% or more, typically 90-95% or more) to the framework region of naturally occurring human immunoglobulin. The framework region of an antibody, that is, the combined framework regions included in the light and heavy chains, serves to position and align the CDRs. CDRs are primarily responsible for binding to the epitope of the antigen. Accordingly, the term humanized immunoglobulin refers to an immunoglobulin containing a human framework region and one or more CDRs of a non-human (usually mouse or rat) immunoglobulin. The non-human immunoglobulin providing the CDR is called the donor, and the human immunoglobulin providing the scaffold is called the acceptor. Constant regions need not be present, but if present, they should be substantially identical to human immunoglobulin constant regions, that is, at least about 85-90% identical, preferably about 95% or greater. Therefore, all portions of the humanized immunoglobulin, with the possible exception of the CDRs, are substantially identical to the corresponding portions of the natural human immunoglobulin sequences. In addition, one or more residues in the human framework region may be mutated back to the original sequence to maintain optimal antigen-binding affinity and specificity. Thus, certain framework residues from the original non-human antibodies are retained in the humanized antibody to retain the binding properties of the original antibody while minimizing its immunogenicity. As used herein, the term human framework region includes regions with such back mutations. A humanized antibody is an antibody comprising a humanized light chain and a humanized heavy chain of an immunoglobulin. For example, a humanized antibody will not include a typical chimeric antibody as defined below, for example, because the entire variable region of the chimeric antibody is different from human.

Monoclonal antibodies and antibody constructs of the present invention specifically include chimeric antibodies (immunoglobulins) in which a portion of the heavy and/or light chain is identical or homologous to corresponding sequences in antibodies derived from a particular species, or belongs to a particular class or subclass of antibodies, then how the remainder of the chain(s) are identical or homologous to corresponding sequences in antibodies derived from another species or belong to another class or subclass of antibodies, as well as fragments of such antibodies so long as they exhibit the desired biological activity (US Patent No. 4816567; Morrison et al (1984), Proc. Natl. Acad. Sci., USA, 81:6851-6855). Chimeric antibodies of interest herein include primitive antibodies containing antigen binding variable domain sequences derived from non-human primates (eg, marmosets, apes, etc.) and human constant region sequences. Numerous approaches have been described for generating chimeric antibodies. See, for example, Morrison et al. (1985), Proc. Natl. Acad. Sci., USA, 81:6851; Takeda et al. (1985), Nature, 314:452; Cabilly et al., US Patent No. 4816567; Boss et al., US Patent No. 4816397; Tanaguchi et al., EP 0171496; EP 0173494; and GB 2177096.

The terms human antibody and fully human antibody each refer to an antibody that has the amino acid sequence of a human immunoglobulin, including antibodies isolated from human immunoglobulin libraries or from animals transgenic for one or more human immunoglobulins and which do not express endogenous immunoglobulins; for example, Xenomouse® antibodies and the antibodies described by Kucherlapati et al. in US Patent No. 5939598.

The term genetically modified antibodies means antibodies in which the amino acid

- 9 043899 the sequence has been changed relative to the amino acid sequence of the native antibody. Due to the demand for recombinant DNA techniques, the creation of antibodies should not be limited to the amino acid sequences found in natural antibodies; since antibodies can be engineered to achieve desired characteristics. The possible variations are numerous and range from changes in one or a few amino acids to complete remodeling, such as the variable and/or constant region. Changes to the constant region will generally be made to improve or change characteristics such as complement fixation, membrane interaction and other effector functions, as well as processability and viscosity. Variable region changes will be made to improve antigen binding characteristics.

The Fab fragment consists of one light chain and C _H1 and variable regions of one heavy chain. The heavy chain of a Fab molecule cannot form a disulfide bond with another heavy chain molecule.

The Fab' fragment contains one light chain and one heavy chain, which contains more constant region between the C _H1 and CH2 domains, so that an interchain disulfide bond can be formed between the two heavy chains to form an F(ab')2 molecule.

The F(ab') ₂ fragment contains two light chains and two heavy chains containing part of the constant region between the _CH1 and _CH2 domains, so that an interchain disulfide bond is formed between the two heavy chains.

The terms Fv fragment and single chain antibody refer to polypeptides containing antibody variable regions from both the heavy and light chains, but lacking constant regions. Like a whole antibody, it is capable of selectively binding to a specific antigen. With a molecular weight of only approximately 25 kDa, Fv fragments are much smaller than conventional antibodies (150-160 kDa), which consist of two heavy protein chains and two light chains, and even smaller than Fab fragments (approximately 50 kDa, one light chain and half of the heavy chain).

A single domain antibody is an antibody fragment consisting of an Fv unit with a single domain, such as VH or _VL . Like a whole antibody, it is capable of selectively binding to a specific antigen. With a molecular weight of only 12-15 kDa, single-domain antibodies are much smaller than conventional antibodies (150-160 kDa), which consist of two heavy protein chains and two light chains, and even smaller than Fab fragments (approximately 50 kDa, one light chain and half of the heavy chain) and single chain variable fragments (approximately 25 kDa, two variable domains, one from the light and one from the heavy chain). The first single-domain antibodies were constructed from antibodies containing only the heavy chains found in camelids. Although most single-domain antibody research is currently based on heavy chain variable domains, light chain variable domains and light chain-derived nanoparticles have also been shown to specifically bind to target epitopes.

As used herein, the term monoclonal antibody is not limited to antibodies produced by hybridoma technology. The term monoclonal antibody refers to an antibody that is derived from a single clone, including any eukaryotic, prokaryotic or phage clone, and not to the method of its preparation.

In some embodiments, the antigen binding protein of the present invention selectively suppresses a human antigen for the antibody from which it is derived. For example, an antigen binding protein having an AF antibody sequence substituted as described herein will selectively inhibit the antigen of FIG. 1B for AF antibody. An antibody or functional fragment thereof selectively inhibits a specific receptor or ligand over other receptors or ligands when the IC50 of the antibody in an inhibition assay for the specific receptor is at least 50-fold lower than the IC50 in an inhibition assay for another reference ligand or receptor. The IC50 is the dose/concentration required to achieve 50% inhibition of biological or biochemical function. For radioligands, the IC50 is the concentration of the competing ligand that displaces 50% of the specific binding of the radioligand. The IC50 of any particular drug or antagonist can be determined by constructing a dose-response curve and examining the effect of different drug or antagonist concentrations on the reversible activity of the agonist in a specific functional assay. IC50 values can be calculated for a given antagonist or drug by determining the concentration required to inhibit half the maximal biological response of the agonist. Thus, the IC50 value for any anti-PCSK9 antibody or functional fragment thereof, for example, can be calculated by determining the concentration of antibody or fragment required to inhibit half of the maximal PCSK9 biological response upon activation of the human PCSK9 receptor in any functional assay. By an antibody or functional fragment thereof that selectively inhibits a particular ligand or receptor is meant a neutralizing antibody or neutralizing fragment relative to that ligand or receptor. Thus, in some embodiments, an anti-PCSK9 antibody or functional fragment thereof is

- 10 043899 neutralizing antibody or fragment against human PCSK9.

The substituted antigen binding proteins of the present invention can cross-block the unsubstituted antibodies from which they are derived. The terms cross-blocking, cross-blocking, and cross-blocking are used interchangeably herein to refer to the ability of an antigen binding protein to interfere with the binding of other antigen binding proteins (eg, antibodies or binding fragments) to a target (eg, human PCSK9). The extent to which an antibody or binding moiety is able to interfere with the binding of another antibody to its target, and therefore whether it can be said to cross-block, can be determined using competitive binding assays. In some embodiments, the cross-blocking antigen binding protein of the present invention reduces the binding of the reference antibody to the target antigen by about 40-100%, for example, from about 60% to about 100%, especially preferably by about 70-100%, and more particularly preferably by approximately 80-100%. A particularly suitable quantitative assay for detecting cross-blocking uses the Biacore apparatus, which measures the degree of interaction using surface plasmon resonance technology. Another suitable quantitative cross-blocking assay uses a FACS-based approach to measure the competition between antibodies in terms of their binding to a target antigen.

The term nucleic acid or nucleic acid molecule refers to polynucleotides such as deoxyribonucleic acid (DNA) or ribonucleic acid (RNA), oligonucleotides, fragments produced by polymerase chain reaction (PCR), and fragments produced by any of ligation, digestion, effects of endonuclease and effects of exonuclease. Nucleic acid molecules can be composed of monomers that are naturally occurring nucleotides (such as DNA and RNA) or analogues of naturally occurring nucleotides (such as the α-enantiomeric forms of natural nucleotides), or a combination of both. Modified nucleotides may contain modifications to sugar moieties and/or to pyrimidine or purine base moieties. Modifications of sugars include, for example, the replacement of one or more hydroxyl groups with halogens, alkyl groups, amines and azido groups, or sugars can be functionalized to form ethers or esters. Moreover, the entire sugar moiety can be replaced sterically and electronically by similar structures such as azasugars and carbocyclic sugar analogues. Examples of modifications on the base moiety include alkylated purines and pyrimidines, acylated purines or pyrimidines, or other well known heterocyclic substituents. Nucleic acid molecules can be linked by phosphodiester bonds or analogues of such bonds. Phosphodiester linkage analogues include phosphorothioate, phosphorodithioate, phosphoroselenoate, phosphodiselenoate, phosphoroanilothioate, phosphoranilidate, phosphoramidate, and the like.

The term nucleic acid molecule also includes so-called peptide nucleic acids, which contain naturally occurring or modified nucleic acid bases attached to a polyamide backbone. Nucleic acids can be either single-stranded or double-stranded.

The term molecule complementary to a nucleic acid molecule refers to a nucleic acid molecule having a complementary nucleotide sequence and reverse orientation compared to the reference nucleotide sequence.

The term degenerate nucleotide sequence refers to a sequence of nucleotides that contains one or more degenerate codons compared to the reference nucleic acid molecule encoding the polypeptide. Degenerate codons contain different nucleotide triplets but code for the same amino acid residue (i.e., the GAU and GAC triplets each code for Asp).

An isolated nucleic acid molecule is a nucleic acid molecule that is not integrated into the genomic DNA of an organism. For example, a DNA molecule encoding an antibody heavy chain that has been separated from the genomic DNA of a cell is an isolated DNA molecule. Another example of an isolated nucleic acid molecule is a chemically synthesized nucleic acid molecule that is not integrated into the genome of an organism. A nucleic acid molecule that has been isolated from a particular species is smaller than the entire DNA molecule of a chromosome from that species.

A nucleic acid molecule design is a nucleic acid molecule, single-stranded or double-stranded, that has been modified through human intervention to contain nucleic acid segments combined and arranged side by side in an order not found in nature.

Complementary DNA (cDNA) is a single-stranded DNA molecule that is formed from an mRNA template by the enzyme reverse transcriptase. Typically, a primer complementary to parts of the mRNA is used to initiate reverse transcription. Those skilled in the art further use the term cDNA to refer to a double-stranded DNA molecule consisting of such a single-stranded DNA molecule and its complementary strand of DNA. Additionally, the term cDNA refers to a clone of a cDNA molecule synthesized from an RNA template.

A promoter is a nucleotide sequence that directs the transcription of a structural gene. Typically, the promoter is located in the 5' non-coding region of the gene, proximal to the transcription start site of the structural gene. Sequence elements within promoters that function in initiating transcription are often characterized by consensus nucleotide sequences. These promoter elements include RNA polymerase binding sites, TATA sequences, CAAT sequences, differentiation specific elements (DSE; McGehee et al. (1993), Mol. Endocrinol., 7:551), cyclic AMP response elements (CRE), serum response elements (SRE; Treisman (1990), Seminars in Cancer Biol., 1:47), glucocorticoid response elements (GRE), and binding sites for other transcription factors such as CRE/ATF (O'Reilly et al. (1992) , J. Biol. Chem., 267:19938), AP2 (Ye et al. (1994), J. Biol. Chem., 269:25728), SP1, cAMP response element binding protein (CREB; Loeken (1993), Gene Expr., 3:253) and octamer factors (see generally Watson et al. (1987), eds., Molecular Biology of the Gene, 4th edition, The Benjamin/Cummings Publishing Company, Inc.; and Lemaigre et al (1994), Biochem J, 303:1). If the promoter is an inducible promoter, then the rate of transcription increases in response to the inducing agent. In contrast, the rate of transcription is not regulated by the inducing agent if the promoter is a constitutive promoter. Repressible promoters are also known.

A regulatory element is a nucleotide sequence that modulates the activity of the core promoter. For example, a regulatory element may contain a nucleotide sequence that binds to cellular factors that mediate transcription exclusively or predominantly in certain cells, tissues, or organelles. These types of regulatory elements are typically associated with genes that are expressed in a cell-specific, tissue-specific, or organelle-specific manner.

An enhancer is a type of regulatory element that can increase the efficiency of transcription, regardless of the distance or orientation of the enhancer relative to the transcription start site.

Heterologous DNA refers to a DNA molecule or population of DNA molecules that does not naturally exist in a given host cell. DNA molecules heterologous to a particular host cell may contain DNA derived from the host cell species (i.e., endogenous DNA) as long as such host DNA combines with non-host DNA (i.e. exogenous DNA). For example, a DNA molecule containing a segment of non-host DNA that encodes a polypeptide operably linked to a segment of host DNA containing a transcriptional promoter is considered a heterologous DNA molecule. Conversely, a heterologous DNA molecule may contain an endogenous gene operably linked to an exogenous promoter. As another illustration, a DNA molecule containing a gene obtained from a wild-type cell is considered heterologous DNA if said DNA molecule is introduced into a mutant cell that lacks the wild-type gene.

An expression vector is a nucleic acid molecule encoding a gene that is expressed in a host cell. Typically, an expression vector contains a transcription promoter, a gene, and a transcription terminator. Gene expression is usually under the control of a promoter, and such a gene is said to be operably linked to a promoter. Likewise, a regulatory element and a core promoter are operably linked if the regulatory element modulates the activity of the core promoter.

A recombinant host is a cell that contains a heterologous nucleic acid molecule, such as a cloning vector or expression vector. As used herein, an example of a recombinant host is a cell that produces an antagonist of the present invention from an expression vector. Conversely, such an antagonist can be produced by a cell that is a natural source of said antagonist and that does not contain an expression vector.

The terms amino-terminal and carboxy-terminal are used herein to refer to positions within polypeptides. When the context permits, these terms are used with reference to a specific sequence or portion of a polypeptide to indicate proximity or relative position. For example, a particular sequence located carboxy-terminally relative to a reference sequence within a polypeptide is located proximal to the carboxyl-terminus of the reference sequence, but is not necessarily located at the carboxyl-terminus of the full-length polypeptide.

A fusion protein is a fusion protein expressed by a nucleic acid molecule containing nucleotide sequences from at least two genes. For example, the fusion protein may comprise at least a portion of an antibody heavy chain fused to a polypeptide,

- 12 043899 which binds an affine matrix or other target of interest.

The term receptor refers to a cell-associated protein that binds to a bioactive molecule called a ligand. This interaction mediates the effect of the ligand on the cell. Receptors may be membrane-bound, cytosolic, or nuclear; monomeric (eg, thyroid-stimulating hormone receptor, beta-adrenergic receptor) or multimeric (eg, PDGF receptor, growth hormone receptor, IL-3 receptor, GM-CSF receptor, G-CSF receptor, erythropoietin receptor and IL-6 receptor). Membrane-bound receptors are characterized by a multi-domain structure, including an extracellular ligand-binding domain and an intracellular effector domain, which is usually involved in signal transduction. In certain membrane-bound receptors, the extracellular ligand-binding domain and the intracellular effector domain are located in separate polypeptides that make up the fully functional receptor. In general, binding of a ligand to a receptor results in a conformational change in the receptor, which causes an interaction between the effector domain and other molecule(s) in the cell, which in turn leads to a change in cell metabolism. Metabolic events that are often associated with receptor-ligand interactions include gene transcription, phosphorylation, dephosphorylation, increased cyclic AMP production, cellular calcium mobilization, membrane lipid mobilization, cell adhesion, hydrolysis of inositol-containing lipids, and phospholipid hydrolysis.

The term expression refers to the biosynthesis of a gene product. For example, in the case of a structural gene, expression includes transcription of the structural gene to produce mRNA and translation of the mRNA to produce one or more polypeptides.

The term complementary/anticomplementary pair refers to non-identical moieties that, under appropriate conditions, form a non-covalently linked stable pair. For example, biotin and avidin (or streptavidin) are prototypical members of the complementary/anticomplementary pair. Other exemplary complementary/anticomplementary pairs include receptor/ligand pairs, antibody/antigen (or hapten or epitope) pairs, sense/antisense polynucleotide pairs, and the like. When subsequent dissociation of a complementary/anti-complementary pair is desired, the complementary/anti-complementary pair preferably has a binding affinity of less than 10 ⁹ M' ¹ .

The detectable label is a molecule or atom that can be conjugated to an antibody fragment to produce a molecule useful for diagnosis. Examples of detectable labels include chelators, photoactive agents, radioisotopes, fluorescent agents, paramagnetic ions, or other marker moieties.

The term affinity tag is used herein to refer to a segment of a polypeptide that can be attached to a second polypeptide to provide purification or detection of the second polypeptide or to provide sites for attachment of the second polypeptide to a substrate. In principle, any peptide or protein for which an antibody or other specific binding agent is available can be used as an affinity tag. Affinity tags include polyhistidine region, protein A (Nilsson et al. (1985), EMBO J., 1:1075; Nilsson et al. (1991), Methods Enzymol., 198:3), glutathione S-transferase (Smith et al. al. (1988), Gene, 67:31), Glu-Glu affinity tag (Grussenmeyer et al. (1985), Proc. Natl. Acad. Sci., USA, 82:7952), substance P, FLAG® peptide ( Hopp et al. (1988), Biotechnology, 6:1204), streptavidin binding peptide or other antigenic epitope or binding domain. See generally Ford et al. (1991), Protein Expression and Purification, 2:95. DNA molecules encoding affinity tags are available from commercial suppliers (eg, Pharmacia Biotech, Piscataway, NJ).

The terms acidic residue and negatively charged residue refer to amino acid residues that have side chains containing acidic groups. Exemplary acidic or negatively charged residues include D and E.

The term amide residue refers to amino acids that have side chains containing amide derivatives of acid groups. Exemplary amide residues include N and Q.

The term aromatic residue refers to amino acid residues that have side chains containing aromatic groups. Exemplary aromatic moieties include F, Y and W.

The terms basic residue and positively charged residue refer to amino acid residues that have side chains containing basic groups. Exemplary basic or positively charged residues include H, K, and R.

The terms hydrophilic residue and polar uncharged residue refer to amino acid residues that have side chains containing polar groups. Exemplary hydrophilic or polar uncharged residues include C, S, T, N, and Q.

The terms non-functional residue and small hydrophobic residue refer to amino acid residues characterized by side chains that lack acidic, basic or aromatic groups. Exemplary non-functional small hydrophobic residues include M, G, A, V, I, L and norleucine (Nle).

- 13 043899

One aspect of the present invention relates to PCSK9-binding polypeptides. PCSK9 binding polypeptide means a polypeptide that binds the proprotein convertase subtilisin/kexin type 9 (PCSK9) protein. In some cases, PCSK9-binding polypeptide blocks PCSK9 binding to low-density lipid receptors (LDLRs). Such blocking PCSK9-binding polypeptides may be monoclonal antibodies (mAbs) and may be one of the following:

a) mAb containing a heavy chain polypeptide characterized by the amino acid sequence of SEQ ID NO: 13 6, and a light chain polypeptide characterized by the amino acid sequence of SEQ ID NO: 134 (AK antibody, evolocumab), or an antigen binding fragment thereof;

b) a mAb that competes with evolocumab for binding to PCSK9;

c) mAb containing

i) a heavy chain polypeptide comprising the following complementarity determining regions (CDRs): heavy chain CDR1, which is CDR1 of SEQ ID NO: 376 or 378; heavy chain CDR2, which is the CDR2 of SEQ ID NO: 376 or 378; heavy chain CDR3, which is the CDR3 of SEQ ID NO: 376 or 378; and ii) a light chain polypeptide comprising the following CDRs: light chain CDR1, which is CDR1 of SEQ ID NO: 377 or 379; light chain CDR2, which is the CDR2 of SEQ ID NO: 377 or 379; and a light chain CDR3 which is the CDR3 of SEQ ID NO: 377 or 379;

d) an mAb that binds to at least one of the following residues of PCSK9, wherein PCSK9 contains the amino acid sequence of SEQ ID NO: 369: S153, D188, I189, Q190, S191, D192, R194, E197, G198, R199, V200 , D224, R237 and D238, K243, S373, D374, S376, T377, F379, I154, T1897, H193, E195, I196, M201, V202, C223, T228, S235, G236, A239, G244, M247, I369 , S372 , C375, C378, R237 and D238;

e) an mAb that binds to PCSK9 at an epitope on PCSK9 that overlaps with an epitope bound by an antibody that contains

i) a heavy chain variable region characterized by the amino acid sequence of SEQ ID NO: 136; and ii) a light chain variable region having the amino acid sequence of SEQ ID NO: 134; and iii) where the epitope for the mAb further overlaps with the region to which the epidermal growth factor repeat A (EGF-A) domain of the low-density lipoprotein receptor (LDLR) protein binds (Horton, Cohen & Hobbs (2007), Trends Biochem Sci, 32 (2), 71-77, DOI: 10.1016/j.tibs.2006.12.008; Seidah & Prat (2007), J. Mol. Med. (Berl), 85(7), 685-696;

f) an mAb that contains a heavy chain polypeptide containing the following complementarity determining regions (CDRs):

i) heavy chain CDR1, CDR2 and CDR3, characterized by the amino acid sequence of SEQ ID NO: 373, 374 and 375, respectively; and ii) light chain CDR1, CDR2 and CDR3, characterized by the amino acid sequence of SEQ ID NO: 369, 370 and 371, respectively; or

g) an mAb that contains the heavy chain variable region sequence of SEQ ID NO: 378 and the light chain variable region sequence of SEQ ID NO: 379.

Preferred Embodiments

Correlation of global sequence characteristics with viscosity.

The primary objective of the study reported in Example 1 herein was to identify the relationship between viscosity and the amino acid sequence or overall sequence characteristics of IgG monoclonal antibodies with the goal of reducing the viscosity of high-strength monoclonal antibody formulations. For this purpose, the viscosities of 43 different monoclonal antibodies were measured at a concentration of 150 mg/ml, obtaining a wide range of values from 5 to 33 cP (Fig. 1A and 1B). Basic general characteristics of monoclonal antibody sequences, such as light and heavy chain types, their subtypes (germ lines), and pI, were calculated and correlated with viscosity, but did not immediately reveal significant correlations (Fig. 1A and 1B). Polymorphisms (known as allotypes) in IgG isotypes have been described using human-derived serological reagents (Ropartz, S., Schanfield, M.S., Steinberg, A.G. (1976), Review of the notation for the allotypic and related markers of human immunoglobulins, WHO meeting on human immunoglobulin allotypic markers, held 16-19 July 1974, Rouen, France; report amended June 1976, J. Immunogenet., 3, 357-362), and correlate with certain amino acid residues at several specific positions in conserved regions of heavy and light chains (Jefferis & Lefranc (2009), Human immunoglobulin allotypes: possible implications for immunogenicity, mAbs 1, 332-338.) (Vidarsson, G., Dekkers, G., Rispens, T. (2014), IgG subclasses and allotypes: from structure to effector functions, Front Immunol., 5, 1-17). The allotypes introduced several different residues (described below) into the canned food

- 14 043899 otherwise active regions of light and heavy chains. All kappa light chains used in this study had the same (3) allotype (differing by residues A153, V191 according to EU numbering). All IgG2 heavy chains had the same (n-) allotype (different from P189). Four IgG1 heavy chain allotypes have been described (including the following related residues according to EU numbering): f (R214); z(K214); a (D356, L358) and x (G431) (Jefferis & Lefranc, 2009) (Vidarsson et al., 2014). IgG1 heavy chains with alternative residues at positions E356, M358 and A431 do not constitute allotypes, since these amino acid residues are present in other IgG subclasses. The IgG1 (x) allotype was not represented in the study; all IgG1 heavy chains had A431. IgG1 heavy chain allotypes (f), (z), (a) and related residues are shown in FIG. 1A and 1B.

Antibodies in Fig. 1A and 1B are sorted by viscosity. Tables in Figs. 1A and 1B include the name of the monoclonal antibody, measured concentrations, measured viscosities, and general sequence parameters including type, subtype, and calculated pI. The IgG1 type, lambda light chain, and heavy chain subtype VH1 are in bold.

The heavy chains IgG1 and IgG2 and the light chains kappa and lambda were fairly evenly distributed across the viscosity range. Sequential subtype evaluation identified several pairs of high-viscosity and low-viscosity subtypes: VH1|1-18 and VH1|1-02; VH3|3-33 and VH3|3-07 VK3|L16 and VK3|A27, with a random correlation probability of 0.0002; 0.076 and 0.031, respectively, correlating with the viscosity residuals (Fig. 2). The study looked for correlations of viscosity with D and J region sequences but found no significant correlation.

In fig. Figure 2 shows p values that indicate the likelihood of a random correlation with viscosity. In fig. Figure 2 also shows residues in the high viscosity subtype, positions in the Aho numbering, and residues in the low viscosity subtypes.

Among the fourteen IgG molecules of the VH1 subtype, high viscosity was closely associated with the VH1|1-18 subtype and low viscosity with the VH1|1-01 subtype, with a very low probability of being a random coincidence (Fig. 9).

To assess the correlation between the two subtypes and viscosity, the probability of the mean for the same population was calculated using Student's t test for VH1|1-02 versus VH1|1-18, VH3|3-33 versus VH3|3- 07 and VK3|L16 versus VK3|A27 using a two-sample equal variation t test with a two-tailed distribution.

An association was found (t-test, p=0.031) between the light chain VK3|L16 with high viscosity and VK3|A27 with low viscosity (Fig. 2, 12, left). VK3|L16 versus VK3|A27 and VH3|3-33 versus VH3|3-07 are discussed further herein. Due to the strong correlation with viscosity, VH1|1-18 and VH1|1-02 were further evaluated as follows. In the next step, the sequences of the 43 antibody chains were aligned and evaluated as follows.

Sequence alignment and numbering system.

There are several IgG numbering systems, including:

EU--Edelman et al. (1969), The covalent structure of an entire gamma immunoglobulin molecule, Proc. Natl. Acad. Sci., USA, 63, 78-85;

Kabat--Kabat et al. (1991), Sequences of proteins of immunological interest, Fifth Edition, NIH Publication No. 91-3242;

Chothia--Chothia et al. (1992), Structural repertoire of the human VH segments, J. Mol. Biol., 227:799-817; Tomlinson et al., (1995), The structural repertoire of the human V kappa domain, EMBO J., 14:4628-4638;

Aho-Hoenegger et al. (2001), Yet another numbering scheme for immunoglobulin variable domains: an automatic modeling and analysis tool, J. Mol. Biol., 309, 657-670; and others.

All four numbering systems mentioned above are illustrated in FIG. 9 on the right with respect to VH1 subtype heavy chain framework region 3. The Aho scheme was constructed using the spatial positions of amino acid residues obtained from more than 400 crystal structures of the variable domains of various antibodies. The Aho numbering system defined by A. Honegger (cited above) was used in this work because it is a 3D structure based numbering system. This creates an advantage specifically for residues in the CDR: residues with the same numbers are located in the same spatial regions and are comparable across different IgG sequences. Since residue positions in the Aho numbering scheme are associated with tertiary structure, they should be more associated with biophysical and biochemical properties and possibly viscosity. Aho numbering is aligned and correlated with other major numbering schemes shown in several tables in this specification. Any of the four numbering systems can be used interchangeably to identify preferred amino acid substitutions. The heavy chain variable region ends with the following residues for different numbering systems: 149 Aho, 117 EU, 113 Kabat, 113 Chothia. The light chain variable region ends at 149 Aho, 107 EU, 107 Kabat, 107 Chothia. Aho numbering allocates more numbers to CDR regions instead of using letters for CDR remainders as in Kabat and Chothia (e.g. 82b for Kabat and Chothia). As a result, Aho numbers for the same residues are often larger. Each variable region includes three complementarity determining regions (CDRs) and four framework regions (FRs) in the following sequence:

- 15 043899

FR1-CDR1-FR2-CDR2-FR3-CDR3-FR4. Although CDRs provide a wide variety of sequences for the purpose of binding to antigens (CDR3 regions bind most frequently), the sequences

FRs are more conservative and contain only a few differences, some of which are subtype specific.

Correlation of sequences with viscosity.

In addition to assessing overall sequence characteristics, variable region sequences were aligned to identify residues responsible for viscosity differences. In addition to visual observations, a machine learning software algorithm was developed and applied to identify residues that most affect viscosity and to predict antibody viscosity values from their sequences. A predictive model was built using the charge and hydrophobicity of residues at Aho-aligned positions.

Alignment of the VH1 heavy chain sequences and molecular evaluation of five sequences from the high viscosity subtype VH1 1-18 and five sequences from the low viscosity subtype VH1 1-02 revealed that only 4 residues differed between the two subtypes in the scaffolds, all four located in frame 3 (Fig. 9, right). More sequence differences were observed in the CDRs, but these were left out of the study because CDRs are often involved in antigen binding, and engineered (mutated residues in) CDRs to reduce viscosity carry a significant risk of loss of activity. Subtype-related differences in viscosity and residues in FR3 indicate that the following amino acid substitutions could potentially provide a reduction in viscosity according to Aho numbering: T82R, T86I, R94S, and S95R. The software algorithm supported four substitutions and also provided the assumption that the following two substitutions were correlated with a decrease in viscosity: G13V/L light chain in FR1 and S59R/K heavy chain at the edge of CDR2, the latter correlation observed in the VH3 subtype (Fig. 9, on right).

In silico evaluation of proposed amino acid substitutions.

VH T82R. R82 occurs at high frequency in VH1-02. Modeling of the IgG structure has shown that Aho heavy chain position 82 is part of the upper core of the globulin fold and does not normally contact the antigen, but it can directly contact the CDR backbones according to Ewert et al. (2003), Biophysical properties of human antibody variable domains, J. Mol. Biol., 325:531–553. HC82 is characterized by highly conserved backbone-side chain H-bond interactions with the backbone amides CDR1 and CDR2 (Honegger et al., supra). R82 can also coordinate the oxygen atoms of the backbone of the CDR2 loop.

VH R94S and S95R. S94 and R95 occur with high frequency in VH1-02. These positions are located on the surface, away from the antigen-binding domain, and are considered part of the lower core.

T86I. I86 occurs with high frequency in VH1-02. The data suggest replacement of the hydrophobic residue (I) by a hydrophilic residue (T) at the surface, potentially leading to aggregation.

VH S59R/K- within VH1 and VH3, hydrophilic position 59 is associated with lower viscosity. The VH position of S59R/K has high structural and sequence variability, is quite solvent exposed, and is located directly between residues 58 and 60, which are part of the top core and may influence binding. The structure is likely to depend directly on differences in residues 58 and 60 (especially 58 if immersed). R/K59 has a low frequency of occurrence (below 2%) and was not observed in VH1 according to amino acid residue frequency analysis. All R/K59s are in the data set for VH3 except one (VH4).

VL G13V/L - This position is structurally immersed and is part of the lower core variable domain according to Ewert et al., supra. From this point of view, mutation of G13V to a more hydrophobic residue should make the core more durable.

To summarize, in silico sequence analysis showed that the proposed mutations do not introduce any additional glycosylation sites or sites sensitive to rapid degradation under physiological or slightly acidic formulation conditions (NG, NS, NT, DG, DH). The VH mutations T82R, T86I, r94S and S95R will provide a transition from the VH1-18 to VH1-02 subtype in framework region 3, so they should not introduce any unusual or rare motifs. The addition of the VH S59R and VL G13V mutations was suggested by software based on the VH3 subtype outside of VH1. None of the mutation sites are located close to the binding regions, with the exception of S59R, which is located at the edge of CDR2 of the HC and therefore poses a moderate risk of impaired activity/binding. Arginine is a very rare residue at position 59 (R59), making prediction of its effects difficult. T86I was identified as having a high risk of aggregation and was removed from the list of mutations.

The resulting mutants and their expression, activity, chemical modifications, glycosylation and viscosity.

Taking into account the above considerations, several mutants were prepared for two IgG2 antibodies AK and AO of the high viscosity subtype VH1-18, with the aim of reducing viscosity while maintaining activity (Fig. 9B). Fig. 9B includes symbols for monoclonal anti-16 043899 antibody mutants and corresponding mutations in the heavy and light chains.

Very low levels of expression were observed for both AK mutants containing the S59R substitution (marked * in Fig. 3). Viability and viable cell density were also low for one of them, AK (59 82 94 95 13). On the other hand, mutants of the AO antibody containing the S59R substitution were characterized by the production of a titer comparable to that of the original AO molecule. Although statistical evidence was insufficient to draw a general conclusion about heavy chain position 59, the case demonstrated that a single amino acid substitution could significantly alter expression. Chemical modifications, including oxidation, deamidation, isomerization, and glycosylation patterns, were similar among the two parent molecules and their mutants, as measured by peptide mapping in LC-MS analysis.

The activity values of the parent AK antibody and the two well-expressed mutants, measured by binding to PCSK9, were similar (Fig. 4). Finally, the measured viscosities of the AK and AO mutants were significantly lower than those of the parent antibodies, as predicted (Fig. 5). For example, the AK mutant (82 94 95) had only 39% of the viscosity of the original antibody. The viscosity of both AO mutants containing the S59R substitution was even lower, approximately 28% of the viscosity of the parent antibody (Fig. 5). The S59R mutants for the AK antibody were poorly expressed and viscosity could not be measured. Average viscosity values for the VH1|1-18 and VH1|1-02 germline subfamilies were added for comparison. A total of 12 consistent sequence differences were identified between the VH1|1-18 and VH1|1-02 subfamilies, including 8 in the CDRs and 4 in the framework regions (Fig. 9). Three sites, all in framework region 3, were selected for amino acid substitutions from VH1|1-18, which is characterized by high viscosity, to VH1|1-02, which has low viscosity. Three point mutations in the scaffolds were introduced into two mAbs of the VH1|1-18 subfamily (AK and AO) to convert only these residues to the residues present in VH1|1-02. Although the odds of achieving a possible 2-fold reduction in viscosity were theoretically low (3/12), these substitutions quite unexpectedly produced the desired result: with only three substitutions, the viscosity was approximately halved for both antibody molecules.

Viscosity versus pI.

Although the dependence of viscosity on pI was not clear for the entire group of 43 mAbs, the VH1 subset clearly showed that viscosity increased steadily as mAb pI values decreased from pI 8.5 to pI 6.5 in the pH 5.2 formulation (Fig. 6). A transition from mAb VH1|1-18, which has a high viscosity, to mAb VH1|1-02, which has a low viscosity, can also be observed. As expected, the T82R, R94S, and S95R mutants of AK and AO antibodies moved approximately twofold lower on the viscosity scale from the VH1|1-18 to VH1|1-02 region of the graph (Fig. 6). The AO (59, 82, 94, 95) and AO (59, 82, 94, 95, 13) mutants moved to even lower viscosity and a slightly higher pI value, indicating that the S59R substitution, adopted from outside the VH1 group, was effective in further reducing viscosity. Unfortunately, AK mutants containing R59 were poorly expressed, suggesting that the occurrence of a rare arginine residue at position 59 may influence expression.

Increasing the antibody pI in formulations with a pH < p1 (such as the weakly acidic formulation used in this study) generally results in a decrease in viscosity. This result can be explained by Coulomb repulsion of positively charged antibody molecules. Proteins, including antibodies, are known to exhibit poor solubility and high precipitation rates, which affect viscosity at high concentrations. Interestingly, substitutions of VH1|1-18 with VH1|1-02 in framework region 3 resulted in a twofold decrease in viscosity and only a slight increase in the pI values of the antibodies, which suggests that it is not the increase in charge, but some structural changes that may be the cause a sharp decrease in viscosity.

After superimposing the crystallographic structures of hundreds of Fab domains, hydrogen bond interactions were identified at each VH and VL position (Honegger et al. (2001), Yet another numbering scheme for immunoglobulin variable domains: an automatic modeling and analysis tool, J. Mol. Biol. , 309:657-670). Data indicate that the identified positions may be linked to other residues through backbone and side chain hydrogen bonding interactions (for type III AK and AO antibody molecules). For example, 94 has been linked to 77 in some immunoglobulin VH type III structures; 95 from 18; 59 with 67, 66, 65, 61, 60 and VL13 with 146, 148. Therefore, substitutions of residues at these positions may provide changes in interactions and folding of the immunoglobulin. The crystal structure of the AK antibody (Jackson et al., 2007), The Crystal Structure of PCSK9: a Regulator of Plasma LDL-Cholesterol, Structure, 15:545-52) suggests that all three positions 82, 94 and 95 are at the very periphery Fab regions and are exposed to solvent and other antibody molecules. Changes in FR3 positions in VH3 (Fig. 11) and VK3 (Fig. 12) also correlate with viscosity. One explanation for their role in promoting viscosity is that these positions in FR3 are at the periphery of the molecule and actively participate in intermolecular interactions during shear stress associated with movement through injection needles and viscosity measurements.

The lower viscosity of the VK3|A27 antibodies and the higher viscosity of the VK3|L16 antibodies showed an inverse correlation between viscosity and pI (Fig. 8, 12).

It should be noted that pI was not a very effective, but in some cases a useful prognostic marker of viscosity. In general, viscosity decreased with increasing pI, and pI should be taken into account. For example, the viscosity values of VH3|3-33 become lower and similar to those of

VH3|3-07 for antibody molecules with higher pI (Fig. 7). Therefore, pI should be taken into account when predicting the viscosity of VH3|3-33.

Proposed viscosity reduction for the VH3 and VK3 germline families.

Antibodies VH3|3-33, characterized by high viscosity, and antibodies VH3|3-07, characterized by low viscosity, have on average a large difference in viscosity, while they are in the range of similar pI values (Fig. 7). Surprisingly, on average, VH3|3-07 had a lower viscosity as well as a lower pI, which is contrary to the typical behavior reported in the literature. It is possible to reduce the viscosity values of the following VH3|3-33 molecules using mutations: AQ, AM, AI, AG.

Monoclonal antibodies with high viscosity VK3|L16 light chains and low viscosity VK3|A27 light chains also showed a large difference in viscosity, while exhibiting a relatively small difference in pI values, again indicating structural differences between subfamilies ( Fig. 8). The viscosity values of the following VK3|L16 molecules can be reduced using the mutations described in FIG. 11: AQ and AF antibodies.

Analysis of general sequence characteristics revealed the following pairs characterized by high/low viscosity: VH1|1-18/VH1|1-02; VK3|L16/VK3|A27 and VH3|3-33/VH3|3-07 with p values for correlation of 0.0002, 0.031 and 0.076, respectively (Fig. 2). Positions and sequence residues that correlate with differences in viscosity have been identified and can be considered as candidates for point mutations that reduce viscosity (Fig. 2). The above correlation of the VH and VL families with viscosity shows that theoretically antibodies with the following combinations of VH and VL should have the lowest viscosity (FIGS. 13A and 13B): VH2, VH3|3-07, VH1|1-02 in the case of VH and VK3, VL1, VK3|A27 in case of VL. Three antibodies of the configuration VH1|1-02 and VK3|A27 practically took place in the authors' set, and they did demonstrate low viscosity values: B (5.6 cP), J (8.2 cP), Y (12.1 cP) .

Chemical cross-linking studies.

Conducted extensive research on potential protein-protein interactions in viscous antibody solutions using chemical cross-linking at high protein concentrations. Chemical cross-linking is a classical biochemical technique used to demonstrate how specific parts of proteins interact with each other. Here, the authors describe the use of a chemical zero-length cross-linking reagent to identify potential protein-protein interactions in high-concentration viscous antibody solutions. The results of chemical cross-linking of viscous and non-viscous antibodies are used to construct a model of potential protein-protein interactions in solution.

Chemical cross-linking by EDC shows the presence of an overall chemically cross-linking oligomeric distribution. The chemical cross-linking that results in the oligomeric structure is surprisingly not an intermolecular cross-linking, but rather an intramolecular cross-linking. Intramolecular cross-linking between the Fc top and Fab bottom results in an antibody conformation that can facilitate the formation of Fc-Fc mediated antibody oligomers. The crystal structure of the 1HZH antibody contains an antibody hexamer in which the asymmetric unit has one Fab arm attached to the Fc domain to facilitate FcFc interactions critical for the formation of the IgG1 hexamer. The appearance of this conformation may or may not be critical for hexamer formation in the protein crystal. The increased propensity to form a hexamer in solution when intramolecular Fab-Fc chemical cross-linking is present suggests that the Fab fragment attached to an Fc may have an increased propensity to form Fc-Fc-based antibody oligomers.

Reducing the Fc-Fc interaction parameters can reduce the viscosity of the solution.

The scientific literature includes studies regarding the formation of antibody hexamers associated with interaction with Clq1 and CDC activity. Diebolder et al. found that anti-CD20 antibodies with specific Fc point mutations, including K439E and S440K, suppressed CDC activity, but the related double mutant K439E/S440K restored CDC activity. The I253A mutation also reduced CDC activity. Diebolder et al. (2014), Complement is Activated by IgG Hexamers Assembled at the Cell Surface, Science, 343:1260-3. Diebolder et al. did not link the mutants they uncover to an effect on antibody viscosity. Similarly, van den Bremer et al. (2015) found that charged residues at the C-terminus of the antibody may mediate reduced interaction with Clq1 due to a reduced ability to form IgG hexameric structures. The authors do not associate the presence of charged residues at the C-terminus of IgG with an effect on the viscosity of the antibody solution.

The observation of an antibody hexamer in viscous antibody solutions suggests that Fc-Fc interactions that are present in the crystal structure of the antibody hexamer, as well as a structure thought to form prior to Clq1 (complement) recruitment, are likely present in viscous antibody solutions in the absence of chemical cross-linker EDC. To test whether Fc-Fc interactions could influence solution viscosity, Fc mutants were generated at Amgen based on work done by Diebolder et al. The materials were then evaluated in anti-PCSK9 formulation buffer using cone-plate rheology. Comparison of the parent anti-PCSK9 antibody AK and Fc-mutant anti-PCSK9 antibodies showed that the decrease in Fc-to-Fc affinity does result in a decrease in the viscosity of the antibody solution. The point mutants retained the ability to bind to FcRn and there was no change in biological activity. The ability of a double mutant in which wild-type complement activity was restored to return to wild-type viscosity demonstrates that the decrease in viscosity can be reversed if the ability of Fc-Fc interactions is restored to wild-type levels. Along with the observation of Fc-Fc-mediated oligomeric particles providing an increase in solution viscosity, there is a common Fc-mediated protein-protein interaction that contributes to the viscosity of the antibody solution. It should be noted that of the five mutations, S440K, I253A, K439E, H433A, N434A , defined by Diebolder et al. as reducing the activity of CDC and tested in relation to viscosity in this work, only the first two showed a decrease in viscosity in the composition of the anti-PCSK9 antibody with a high concentration, while K439E, H433A and N343A did not provide a decrease in viscosity in the composition of the anti-PCSK9 antibody with a high concentration, which indicates that there is no direct correlation and that those skilled in the art cannot correctly predict lower viscosity based on the information provided by Diebolder et al. The K439E mutant was also evaluated in a sucrose-based, high protein concentration formulation and found that it was less viscous than the parent PCSK9 antibody mutant at the same concentration. The presence of arginine in the anti-PCSK9 antibody may have contributed to charge shielding, which could reduce the effectiveness of the negative charge introduced into the K439E mutant to reduce Fc-Fc interactions. The H433A and N434A mutants do not have the obvious charge shielding sensitivity of the K439E mutant.

Fc-Fc interactions can influence solution viscosity by increasing the number of potential interactions possible in two potential ways. This may provide an increase in the number of interactions per antibody from two CDR-mediated interactions to two CDR-mediated interactions plus 2 Fc-mediated interactions per antibody, or a change in the number of free CDR ends available in the oligomers that are present in solution. Given that the Fc domains of IgG1, IgG2, IgG3 and IgG4 antibodies are very similar, it is likely that Fc-Fc mediated interactions are present in all antibodies. Analysis of non-viscous antibodies shows that there is no intramolecular cross-linking, which indicates the presence of an Fc-Fc interaction, in contrast to the viscous antibody at the same concentration. This suggests that Fc-Fc interactions, although theoretically possible, are not present in non-viscous antibodies. The interaction of the CDR with the next nearest neighbor can influence the relative distance between Fcs as well as the relative orientation with increased Fc-Fc interactions. This may explain why viscous antibodies are characterized by Fc-Fc interaction, and non-viscous antibodies are characterized by no interaction, at room temperature.

The presence of Fc-Fc interaction also increases the likelihood that the antibody oligomer (dimer, trimer, tetramer, etc.) contains the maximum number of free CDR ends. More free CDR ends increase the number of CDR interactions with the next nearest neighbor. This in turn can increase the tendency for network formation and increase the viscosity of the solution resulting in more efficient infiltration.

C-terminal modifications to reduce viscosity.

Certain modifications at the C-terminus of the antibody interfere with Clq binding and complement-dependent cytotoxicity (CDC). Van den Bremer et al. (2015), Human IgG is produced in a pro-form that requires clipping of C-terminal lysines for maximal complement activation, mAbs 7(4):672-80. The authors found that the C-terminal lysine and C-terminal glutamic acid likely provide a reduced propensity for Fc-Fc interactions resulting in the formation of an antibody hexamer that can interact most effectively with Clq1. We constructed anti-CD20 and anti-CD38 mutants having PGKP (SEQ ID NO:381), PGKKP (SEQ ID NO:382), PGKKKP (SEQ ID NO:383) and PGE at the C-terminus. They found that these mutants exhibited greatly reduced or completely lost CDC activity. Thus, we can conclude that the mutations blocked hexamerization, which the authors previously correlated with CDC activity. Taking into account the existing correlation of hexamerization with viscosity, such mutations should also provide a decrease in the viscosity of antigen-binding proteins. Thus, it is reasonable to conclude that placing positively charged or negatively charged amino acids at the C-terminus, whether placed there by adding or replacing existing C-terminal amino acids, will provide a reduction in the viscosity of the antigen binding protein.

Sequence modification to improve pharmacokinetic parameters.

The present invention also includes the discovery of improved pharmacokinetic properties of antigen binding proteins having mutations that can also provide a decrease in viscosity. In particular, the S440K mutations were found to provide improvements in both _Tmax (time after administration

- 19 043899 dose at which the maximum concentration was observed) and C _max (the maximum observed concentration measured after dose administration). AK antibody mutants characterized by S440K, independently with other mutations, were found to have a _Tmax reduced by more than half that of the parent AK antibody after subcutaneous injection of the mutants and the parent antibody at the same concentration. These mutants were also found to have a _Cmax that was 28% or 42% higher after subcutaneous injection of the mutants and the parent antibody. See fig. 22.

Nucleic acids, vectors, host cells.

The present invention further includes isolated nucleic acids encoding the bispecific antibodies of the present invention, which contain, for example, a light chain, a light chain variable region, a light chain constant region, a heavy chain, a heavy chain variable region, a heavy chain constant region, linkers, and any and all components and combinations thereof of the bispecific antibodies described herein. Nucleic acids of the present invention include nucleic acids having at least 80%, more preferably at least about 90%, more preferably at least about 95%, and most preferably at least about 98% homology with the nucleic acids of the present invention. The terms percent similarity, percent identity, and percent homology when referring to a specific sequence are used as set forth in the University of Wisconsin GCG® software. The nucleic acids of the present invention also include complementary nucleic acids. In some cases, sequences will be completely complementary (no mismatches) when aligned. In other cases, sequences may contain up to approximately 20% mismatches. In some embodiments, nucleic acids encoding both the heavy chain and the light chain of an antibody of the present invention are provided.

The nucleic acids of the present invention can be cloned into a vector, such as a plasmid, cosmid, bacmid, phage, artificial chromosome (YAC) or virus, into which another genetic sequence or element (DNA or RNA) can be inserted, thus to allow replication of the attached sequence or element. In some embodiments, the expression vector contains a segment of a constitutively active promoter (such as a CMV restriction, SV40, elongation factor, or LTR sequence) or an inducible promoter sequence, such as a steroid-inducible pIND vector (Invitrogen), such that expression of the nucleic acid can be regulated. The expression vectors of the present invention may further contain regulatory sequences, for example, an internal ribosome binding site. For example, an expression vector can be introduced into a cell by transfection.

In another embodiment, the present invention provides an expression vector containing the following operably linked elements:

transcription promoter;

a first nucleic acid molecule encoding the heavy chain of a bispecific antigen binding protein, antibody or antigen binding fragment of the present invention;

a second nucleic acid molecule encoding a light chain of a bispecific antigen binding protein, antibody or antigen binding fragment of the present invention; and transcription terminator.

In another embodiment, the present invention provides an expression vector containing the following operably linked elements:

first transcription promoter;

a first nucleic acid molecule encoding the heavy chain of a bispecific antigen binding protein, antibody or antigen binding fragment of the present invention;

first transcription terminator;

a second transcription promoter;

a second nucleic acid molecule encoding a light chain of a bispecific antigen binding protein, antibody or antigen binding fragment of the present invention; and a second transcription terminator.

In addition, the secretory signal peptide sequence may optionally be encoded by an expression vector operably linked to the coding sequence of interest such that the expressed polypeptide can be secreted by the recombinant host cell to more easily release the polypeptide of interest from the cell.

In addition, recombinant host cells containing such vectors and expressing the heavy and light chains are provided.

Cleaning.

Methods for purifying antibodies are known in the art and can be used in the preparation of antibodies and bispecific antibodies of the present invention. In some embodiments, methods for purifying antibodies include filtration, affinity column chromatography, cation

- 20 043899 exchange chromatography, anion exchange chromatography and concentration. The filtration step preferably includes ultrafiltration and more preferably ultrafiltration and diafiltration. Filtration is preferably performed at least about 5 to 50 times, more preferably 10 to 30 times, and most preferably 14 to 27 times. Affinity column chromatography can be performed, for example, as PROSEP® affinity chromatography (Millipore, Billerica, MA). In a preferred embodiment, the affinity chromatography comprises PROSEP®-vA column chromatography. The eluate can be washed with a detergent solvent. Cation exchange chromatography may include, for example, cation exchange chromatography using SP-Sepharose. Anion exchange chromatography may include, for example, but is not limited to, fast flow anion exchange using Q-Sepharose. The anion exchange step is preferably non-binding, allowing the removal of contaminants including DNA and BSA. The antibody product is preferably nanofiltered, for example using a Pall DV 20 nanofilter. The antibody product can be concentrated, for example using ultrafiltration and diafiltration. The method may further include a size exclusion chromatography step to remove aggregates. Other cleaning options are provided in the working examples below.

Bispecific antibodies, antibodies or antigen-binding fragments can also be produced by other methods known in the art, for example, by chemically combining antibodies and antibody fragments.

Each manuscript, research article, review article, abstract, patent application, patent, or other publication cited herein is incorporated herein by reference in its entirety.

Working examples

The present invention is further illustrated by the following working examples, which illustrate, but do not limit the scope of the present invention.

Example 1.

Fab mutations.

Materials.

A panel of 43 human and humanized recombinant monoclonal antibody molecules with different targets and different sequences were prepared and purified according to standard procedures (FIGS. 1A and 1B). A group with equivalent purity of >98% was collected using size exclusion chromatography (SEC). Samples were concentrated to a maximum volume of 3 ml using an Amicon Model 8003 stirred ultrafiltration cell (Millipore, Billerica, MA) at 2-8°C with a maximum pressure of 30 ± 10 psi. inch. They were concentrated to 150 mg/ml according to approximate volume loss in formulation buffer containing 20 mM acetate, 9% sucrose at pH 5.2 (no polysorbate), and final concentrations were determined (±10%) using absorbance at 280 nm relative to protein (after dilution ultimately within 0.1-1 absorption unit (AU)) and protein-specific extinction coefficient.

Several low viscosity mutants were prepared, purified, and generated from the two mAbs according to a similar standard procedure. These included an antibody to proprotein convertase subtilisin-kexin type 9 (PCSK9, antibody AK) and an antibody to macrophage colony-stimulating factor (M-CSF, AO).

Viscosity measurements.

Viscosity analysis was performed on a Brookfield LV-DVIII cone-plate instrument (Brookfield Engineering, Middleboro, MA, USA) using a CP-40 spindle and sample cup. All measurements were performed at 25°C while monitoring with a water bath attached to the sample cup. Repeated viscosity measurements were collected manually within a specific torque range (10–90%) by increasing spindle speed. Measurements were averaged to report one viscosity value per sample to simplify the resulting comparison chart.

Sequence alignment.

Structural sequence alignment was performed using the Ab Initio software tool developed using Excel macros downloaded from the Department of Biochemistry, University of Zurich.

Example 2.

Fc mutations.

Expression and purification of mutants.

Materials and methods.

Antibody to C-kit (antibody BA, SEQ ID NO: 174 and 176, encoded by nucleic acids SEQ ID NO: 173 and 175, respectively), antibody AN to sclerostin and antibody AK to PCSK9.

IgG1 and IgG2 to streptavidin.

- 21 043899

1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride (EDC).

Solubilization with n-methyl-2-pyrrolidone.

Size exclusion high performance liquid chromatography (SE-HPLC) with light scattering (LS).

Reduced and alkylated reverse phase high performance liquid chromatography (RA RP-HPLC).

Tryptic peptide map using electrospray ionization mass spectrometry (ESI-MS).

Cone-plate viscometer.

Results.

Identification of oligomer propensity by SE-HPLC cross-linking using EDC of highly concentrated monoclonal antibody solutions

The compound 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride (EDC) was used to chemically cross-link acidic residues in the antibody with primary amines in the antibody (N-terminus and/or Lys residues) and has been used in other studies to determine regions of protein-protein interactions. The proximity of the carboxyl group to the primary amine is critical because an amide bond is formed between the two groups (Figure 16). The cross-links formed are likely salt bridges that are present in the solution. Carraway and Koshland, Jr. (1972), Carbodiimide modification of proteins, Methods Enzymol., 25:616-623. The antibody panel was chemically cross-linked using EDC under identical solution conditions. Previous rheological studies determined that some of the antibodies on the panel were viscous and some were not (FIG. 17). Non-viscous antibodies showed small increases in dimer content, but they did not contain large amounts of higher order oligomers. In contrast, viscous antibodies contained large amounts of dimer as well as higher order oligomers. A summary is presented in Fig. 17. All antibodies that were identified as viscous contained EDC cross-linked particles that were larger than the dimer. The appearance of the larger oligomeric particles depends on the concentration (antibody AN is shown in Fig. 18 as an example). To facilitate further analysis, the chemical cross-linking conditions were modified to bring the cross-linking reaction to completion. The solutions became solid after chemical cross-linking at 200 mg/mL. Solids were redissolved in buffer or buffered 3% NMP solution. SE-HPLC analysis of both samples showed that the samples were similar. The buffered 3% NMP solution dissolved the protein much faster and more material was introduced into the solution. Antibody AN was redissolved and further analyzed as an example.

Size analysis using SE-HPLC with LS in network access.

The cross-linked antibody solutions were analyzed by online light scattering SE-HPLC to determine the size of the eluting particles. SE-HPLC was performed with online light scattering analysis on antibody AH after chemical cross-linking with EDC and re-dissolution with 3% NMP. SE-HPLC revealed three peaks present in UV and RI. The first peak was identified as particles with a mass of 840.5 kDa. This is close to the expected mass of the AN antibody hexamer of 873.2 kDa. The mass of the other particles was 494.6 kDa, which is close to the predicted mass of 436.6 kDa for the AN antibody trimer. The third particles showed a mass of 139.9 kDa, which is close to the predicted mass of 145.5 kDa for the AN antibody monomer.

HPLC with reduced and alkylated reverse phase EDC cross-linked antibody AH.

The cross-linked antibody solutions were analyzed by reduced and alkylated reverse phase high performance liquid chromatography. The recovery of most of both the light chain (LC) and the heavy chain (HC) was unexpected since either the HC or the LC were expected to be cross-linked to each other to form non-native LC-HC peptides or non-native LC-LC peptides or NS-NS, reflecting the cross-linked oligomers observed in SEC analysis. In the case of the AH antibody, the LC eluted at the same location with the same mass as the non-crosslinked AH antibody LC. There were changes in NA in the AN antibody that were concentration dependent. There is a small amount of HC-HC cross-linked material (eluting at approximately 33 minutes, FIG. 21) that had an identified mass of 100 kDa. At 150 mg/mL and 200 mg/mg, the distribution of NS particles is similar. At lower protein concentrations, the distribution includes more particles, which elute after approximately 28.5 min. The distribution of NS particles correlates with the amount of oligomer present in each sample, as analyzed by SEC in the 200 mg/mL sample containing the largest number of hexamers and the 10 mg/mL sample containing very little oligomer. This picture was observed in solutions of other viscous antibodies.

Example 3.

Pharmacokinetics and pharmacodynamics (PKPD) study of the parent anti-PCSK9 antibody AK and low-viscosity mutants in non-human primates.

- 22 043899

Materials: AK antibody and its mutants. All mutations in the heavy chain.

Fab mutant: T82(72)R, R94(84)S, S95(85)R; numbering according to Aho (actual numbering).

Fc mutant S(434)K (S440K in EU numbering).

Double mutant: T82(72)R, R94(84)S, S95(85)R, S(434)K.

Four groups of 4 male cynomolgus monkeys were used in this study. Each group received 1 subcutaneous (SC) dose of 10 mg/kg as follows. Group 1 received parental AK antibody (140 mg/ml); group 2 received Fab mutant (210 mg/ml); group 3 received Fc mutant (210 mg/ml); group 4: Fab/Fc mutant (210 mg/ml). Groups 1 and 2 also received a control dose of SC diluent. The Fabmutant contained the following substitutions at positions T82(72)R, R94(84)S, S95(85)R. The Fc mutant contained a substitution at position S(434)K.

To measure viscosity at 210 mg/ml, the parent antibody and all mutants were formulated at a concentration of 210 mg/ml in 10 mM acetate, 155 mM N-acetylarginine (NAR), 70 mM ArgHCl, pH 5.4, 0.01% polysorbate 80. Viscosity was measured using an ARG2 cone/plate at 1000 s ^-1 and 25°C. See fig. 15.

General design of the study.

groups of 4 male cynomolgus macaques.

Each group received 1 dose of SC (10 mg/kg):

group 1: parent antibody AK (140 mg/ml);

group 2: Fab mutant (210 mg/ml);

group 3: Fc mutant (210 mg/ml);

group 4: Fab/Fc mutant (210 mg/ml).

Groups 1 and 2 also received a control subcutaneous dose of diluent.

A skin biopsy was taken from the injection site 3 days after dosing.

Histopathological analysis was performed.

LDL, HDL, plasma total cholesterol and RA were monitored for 6 weeks post-dose.

Conclusions of the study.

All 4 homologs provided a significant reduction in LDL.

Maximum reduction at 2 weeks post-dose: 1 week later than previously observed for the parent antibody.

The nadir of the effect was not as pronounced as for the Fab/Fc mutant (~78% compared to ~90%).

The return to baseline appears to be slightly more accelerated for the Fc mutant. PK: Mean exposure values were similar (based on _Cmax and AUClast) between all treatment groups (within 1.4-fold).

Fab and/or Fc mutations in the AK anti-PCSK9 antibody did not provide a significant effect on the injection site response (ISR) parameter or pharmacokinetics and pharmacodynamics (PKPD) profile in non-human primates (NHP) (cynomolgus monkeys).

Fab mutant: T82(72)R, R94(84)S, S95(85)R; numbering according to Aho (actual numbering).

Fc mutant: S(434)K (S440K in EU numbering).

Example 4.

Preparation and characterization of low viscosity mutants of the AQ GIPR antibody (2G10.006).

Cloning, expression, purification and high concentration formulation of low viscosity AQ antibody mutants.

The original AQ GIPR antibody (2G10.006) is described in US Provisional Patent Application 62/387486 as 2G10_LC1.006 (SEQ ID NO: 74 of the cited patent application). The above US patent application is incorporated herein by reference. Heavy chain mutant AQ (HC 1, 17, 85) with mutation sites Q1(1)E, R17(16)G, S85(75)A and light chain mutant AQ (LC 4, 13, 76, 95, 97, 98 ) with mutation sites M4(4)L, V13(13)L, A76(60)d, S95(77)R, Q97(79)E, S98(80)P were obtained as follows. Synthetic genes for low viscosity mutants GIPR (2G10.006) (AQ antibody) were produced, digested and ligated into plasmid expression vectors. The constructs were verified by DNA sequencing. Stable cell pools were generated by electroporation of a clonal CHO host cell line. Pools were cultured during selection until viability exceeded 85%. The pools were sown in a fed-batch production culture for 10 days and the centrifuged medium was collected.

The collected supernatants were sterilized by filtration and purified using a three-column chromatographic method involving protein A, cation exchange and anion exchange, similar to the method described previously (Shukla et al. (2007), Downstream processing of monoclonal antibodies Application of platform approaches, J. Chrom. B, 848:28-39). The resulting purified pools were dialyzed into formulation buffer containing 20 mM acetate and 9% sucrose at pH 5.2 (without polysorbate), achieving a final pH of ~5.2, and concentrated to approximately 150 mg/ml using a cutoff filter above 30 kDa using centrifugal ultrafiltration (Fig. 20A).

- 23 043899

Activity measurements.

Activity was measured using an assay using mammalian 293/huGIPR cells expressing the glucose-dependent insulinotropic polypeptide receptor (GIPR). Increasing concentrations of the original anti-GIPR AQ and low-viscosity mutants blocked the interaction of GIP with GIPR, causing cAMP changes monitored during the assay. Application of the assay has been previously described in Tseng S.S. et al. (1996), Postprandial stimulation of insulin release by glucose-dependent insulinotropic polypeptide (GIP), Effect of a specific glucose-dependent insulinotropic polypeptide receptor antagonist in the rat, J. Clin. Invest., 98: 2440-2445.

Viscosity measurements.

Viscosity analysis of AQ and two low viscosity mutants was performed on an Anton Paar viscometer using a CP25-1/TG spindle. All measurements were performed at 25°C while monitoring with a water bath attached to the sample cup. Viscosity measurements were collected manually as shear rates increased from 0 to 2000 rpm. For each sample, 10 viscosity measurements at 1000 1/s shear rate and 10 viscosity measurements at 2000 1/s shear rate were collected and averaged to obtain one viscosity value per sample.

It should be noted that the reliability of viscosity measurements is much higher than the accuracy, since viscosity measurements are sensitive to even minor changes in some parameters such as the condition of the viscometer, room temperature and some other minor parameters during measurements. It is therefore important to measure all samples of interest at one setting or, if samples of interest are measured at two settings, to have the same reference standard at both settings.

Results.

Anti-GIPR antibody AQ (2G10.006) belongs to the germline VH3|3-33 heavy chain and VK3|L16 light chain subfamilies, characterized by high viscosity. Several mutations obtained in Fig. 11 and 12 were performed in scaffolds in an attempt to provide a reduction in AQ viscosity. The viscosity of the original AQ and two mutants, measured with one viscometer setting, showed the following values: AQ - 19.1 cP, AQ (HC 1, 17, 85) - 15.8 cP, AQ (LC 4, 13, 76, 95, 97, 98) - 12.7 cP (Fig. 20). The AQ heavy chain mutant (HC 1, 17, 85) accounted for 83%, and the AQ light chain mutant (LC 4, 13, 76, 95, 97, 98) accounted for 67% of the original AQ. In fig. Figures 7 and 8 show the positions of the mutants in viscosity versus pI plots for members of the VH3 and VK3 families. cAMP activity in vitro was not equally affected by viscosity mutations. The activity remained unchanged within the error limits of the in vitro cell assay (see Fig. 20B). To summarize, the introduced mutations provided a reduction in viscosity without loss of activity.

Example 5.

Light chain mutant V78F (LC V78F according to Aho numbering and LC V62F in linear numbering) of an anti-GIPR antibody characterized by low viscosity.

Anti-GIPR antibody AQ (2G10.006) showed a high viscosity of 23 cP at 150 mg/mL in A52Su. This antibody was characterized by the rare V78 residue according to Aho numbering (V62 in linear numbering) in the kappa light chain (LC V78 Aho). The frequency of V78 is <1%, whereas F78 is >98% in kappa germline-related light chain sequences. The LC V78 remnant attracted attention because it was a covariance offender. Covariance analysis allows the establishment of pairwise positions of conserved residues based on the physicochemical properties of residues in antibody variable regions, identifying misplaced residues (which are often not germline residues). Additionally, using covariance analysis, it is possible to propose the replacement of amino acids in deviant positions with more common germline sequences, which leads to a significant conformational change detected using molecular dynamics simulations (Kannan G., Method of correlated mutational analysis to improve therapeutic antibodies, patent application USA with serial number 61/451929, PCT/US 2012/028596, WO 2012/125495). In an attempt to correct the covariation disorder and increase the percentage of human sequences, the LC V78F mutation was introduced into the anti-GIPR antibody AQ (2G10.006).

Surprisingly, the viscosity of the mutant was reduced by 25% while maintaining similar activity against human GIPR as measured in cAMP (cellular) assays. Both the sequences of the original anti-GIPR antibody AQ (2G10.006) and its mutant LC V78F are described in US Provisional Patent Application 62/387486 as 2G10_LC1.006 (SEQ ID NO: 74 of the cited application) and 2G10_LC1.003 (SEQ ID NO : 71 from the cited application) respectively. The US patent application is incorporated herein by reference. For the first time, it was discovered in the present invention that such a substitution resulted in a reduction in viscosity of approximately 25% due to the LC V78F mutation. Viscosity analysis of AQ GIPR_2G10.006 and its mutant V78F was carried out at 150 mg/ml in a formulation containing 20 mM acetate, 9% sucrose at pH 5.2, 0.01% polysorbate 80, at a shear rate of 1000 and 25 °C using magnetic cone-plate viscometer AR-G2 from TA Instruments - Waters LLC. The size of the cone plate was

- 24 043899 mm in diameter, 1.988° cone angle, it was equipped with a Peltier plate made of 990918 steel and operated using the Flow Sweep procedure. The measured viscosity values were 21 cP for

GIPR_2G10.006 and 15.3 cP for the LC V78F GIPR mutant (2G10.003), which means a 25% reduction in viscosity.

As noted in the previous example, the reliability of viscosity measurements is much higher than the accuracy. Viscosity at 150 mg/ml with 0.01% polysorbate is typically 10% lower than without polysorbate, as observed with GIPR (2G10.006). Its viscosity was 23 cP without polysorbate (as for all 43 antibodies) and 21 cP with polysorbate.

Example 6.

Research on Cynomolgus Macaques.

The antibodies designated AK (control, also known as AMG 145 and evolocumab) and the Fab mutant, Fc mutant and double mutant shown in Example 3 were prepared using the methodology disclosed in Examples 1 and 2. The pharmacokinetic properties of these antibodies were tested in vivo using a single subcutaneous bolus injection in male cynomolgus monkeys.

Study design.

The study was conducted on male cynomolgus macaques. The animals were between 2.7 and 3.8 years old and weighed between 2.9 and 3.8 kg. Animals were acclimatized in the laboratory building for 7 days before dosing. Selection criteria included acceptable results regarding pre-treatment cholesterol levels (including LDL and HDL levels). Before dosing, all animals were randomized and assigned to groups using a computerized randomization procedure.

Test and control drugs were administered subcutaneously in the mid-back area of the respective animals once on day 1. The injection site(s) were shaved before administration and marked with indelible ink. Animals were temporarily restrained for dosing and were not sedated. The dose volume for each animal was based on the most recent body weight measurements. For groups 1 and 2, dosage solutions were administered via 2 subcutaneous injections into the back of each animal (1 with test material, 1 with diluent). The injection sites were spaced at least 5-6 cm apart. The test material was injected to the right of the spine of each animal. Diluent was delivered to the left of the spine of each animal. For groups 3 and 4, dosage solutions were administered via a single subcutaneous injection into the back of each animal. Dose levels and volumes for each group are summarized in FIG. 21.

Blood samples were collected by venipuncture into tubes containing potassium EDTA (K2) at various time points during the intravital portion of this study (43 days). Animals were not fasted before blood collection for serum chemistry.

Samples were refrigerated after blood collection and separated to obtain either serum or plasma. Samples were mixed gently and centrifuged. Blood samples were stored on wet ice immediately after collection until centrifugation (1500-2000xg for approximately 10 min) at approximately 4°C. The resulting plasma or serum was separated and divided into 2 aliquots (primary and reserve), transferred to appropriately labeled polypropylene tubes and stored in a freezer maintained at -80°C until analysis. Plasma samples were used to determine test drug concentrations for pharmacokinetic evaluation, and serum samples were analyzed for cholesterol, HDL, and LDL.

Pharmacokinetics assessment.

Plasma samples were analyzed for the concentration of each antibody tested (AK antibody, AK Fab mutant, S440K AK Fc mutant, and Fab/S440K AK double mutant) using an enzyme-linked immunosorbent assay (ELISA). The assay uses recombinant human PCSK9 as the capture reagent and horseradish peroxidase-labeled anti-human IgG1 antibody as the detection reagent. Standards and quality control (QC) samples are prepared by adding AK antibody or low viscosity homologs to a 100% cynomolgus K2-EDTA pool. Wells of Costar 96-well microplates (Corning Incorporated) were coated with recombinant human PCSK9. After the blocking step, standards, sample blank (NSB), QC (QC) and test samples are loaded into microplate wells after pretreatment with a dilution factor of 100 in Blocker™ BLOTTO in TBS (Thermo Scientific). The AK antibody in the samples is captured by immobilized recombinant human PCSK9 coated on a microplate. Unbound material is removed by washing the microplate wells. After washing, the mouse anti-human IgG antibody, Ab35, an HRP-conjugated detection antibody, is added to the microplate wells to bind the captured AK antibody. Unbound detection antibody is removed by washing the microplate wells. A one-component TMB solution is added to the microplate wells to detect the HRP-bound mouse anti-human IgG Ab35 conjugate. The TMB substrate solution reacts with the peroxide and, in the presence of HRP, produces a colorimetric signal that is proportional to the amount of AK antibody or low viscosity mutant homologue bound by the capture reagent. The development of color is stopped using 2 N. sulfuric acid and measure the color intensity (optical density or

- 25 043899

OD) at 450 nm minus 650 nm. Data is reduced using a data reduction package

Watson SP3 version 7.4 (or later) using a 4-parameter (Marquardt) regression model with a weight of 1.

Pharmacokinetic parameters were assessed using WinNonlin pharmacokinetic software. A non-compartmentalization approach compatible with the subcutaneous route of administration was used to estimate the parameters. All parameters were derived from individual plasma concentrations from day 1. The following parameters were determined:

Tmax (time after dosing at which the maximum observed concentration was observed);

Cm _a x (maximum observed concentration measured after dosing);

AUC (0-t) (area under the concentration-time curve from dosing initiation to the post-dose time at which the last quantitative concentration was observed, using linear or linear/log trapezoidal method);

AUC(0-t)/D (AUC(0-t) divided by dose administered); And

RAUC (area under the curve from T1 to T2 at steady state divided by the area under the curve from T1 to T2 during the baseline dosing interval).

Results and discussion.

Graphs of test drug concentrations versus time are shown in FIG. 24A and 24B (average concentrations for each drug tested, n=4 at each time point). The pharmacokinetic parameters of the four drugs tested are summarized in FIG. 22. Antibodies containing the Fc S440K mutation (both AK antibody (Fc mutant) and AK antibody (Fc and Fab double mutant) show a decrease in Tmax (0.81 and 1 day, respectively, compared with 2.5 days) and increased Cmax (125 and 112 μg/ml, respectively) compared with 87.8 μg/ml) relative to AK antibody (control) and AK antibody (Fab mutant), indicating that antibodies containing an Fc mutation that provides a reduction in viscosity and is distributed more quickly into the bloodstream after subcutaneous injection.

Administration of all low viscosity AK antibody homologs resulted in the expected mild to moderate pharmacological reduction in low-density lipoprotein (LDL) parameters associated with a reduction in total cholesterol concentrations compared to baseline (day -6). The magnitude of the decrease in total cholesterol and low-density lipoprotein cholesterol after administration of the Fab mutant AK and the Fc mutant S440K AK was generally similar to control animals, with a tendency to restore to the initial level in the Fc mutant S440K AK on day 25 and AK ( control) and Fab mutant AK on day 29. The magnitude of the reduction in total cholesterol and low-density lipoprotein cholesterol in the case of the Fab/S440K AK double mutant was generally less pronounced compared to the control (AK antibody at a dose of 10 mg/kg). There was no change in high-density lipoprotein in any case of the low-viscosity AK homologue. The percentage changes in LDL-C from baseline are tabulated in FIG. 23 and plotted against time in FIG. 24A and 24B.

Abbreviations

Abbreviated terms used throughout this document are defined as follows. AEI allelic expression imbalance

AN0VA analysis of variance

AUC area under the curve

BSA bovine serum albumin

DMEM Needle medium, modified Dulbecco's method

DMSO dimethyl sulfoxide

EDC 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride

EDTA ethylenediaminetetraacetic acid

ELISA enzyme-linked immunosorbent assay eQTL expression of quantitative trait loci ESI-TOF electrospray ionization time-of-flight mass spectrometry

FACS fluorescence activated cell sorting

FBS fetal bovine serum

-

Claims (7)

FPLC fast liquid chromatography of proteins FVB line of mice inbred for the Friend's leukemia virus lb (Fvlb) allele H&E hematoxylin and eosin ON hypoxanthine HIC hydrophobic interaction chromatography HPLC high performance liquid chromatography HRP horseradish peroxidase HUVEC Epithelial cells of the human umbilical vein IBD inflammatory intestinal IDMEM DMEM intestinal IFN without glutamine IFN IL interlayer IL MSR chemotaxis monocytes MSD macromolecular structures of RVMS Mononuclear BBS BBS BBS phosphate-solo buffer-lin PEG PEG PEG polyethylene glycol PEI polyethylene qtl locals of quantitative features RPMI environment developed at Roswell Park Memorial Institute SNP single nucleotide polymorphism TEA trifluoroacetic acid TMB 3,3',5,5'-tetramethylbenzidine CLAIM 1. A method of reducing the viscosity of an antibody or an antigen-binding fragment of the VH1|1-18 germline subfamily, which includes introducing into the sequence of the antigen-binding protein one or more modifications selected from modifying the amino acid sequence of VH1 so that it contains one or more substitutions selected from 82X ¹ , 94X ² and 95X ³ , where X ¹ is selected from R, K and H; X ² is selected from S, T, N and Q; and X ³ is selected from R, K and H, wherein the amino acids are numbered according to the Aho numbering system. 2. The method according to claim 1, characterized in that the antibody or its antigen-binding fragment contains, in addition to those subjected to replacement, the amino acid sequences of the heavy chain and light chain of the antibody AF, AK, AL, AN or AO, presented in FIG. 1B (SEQ ID NOs: 114 and 116; 134 and 136; 138 and 140; 146 and 148; and 150 and 152, respectively). 3. The method according to claim 1 or 2, which includes modifying the amino acid sequence of the VH1 antibody or antigen binding fragment thereof so that it includes the 82R, 94S and 95R substitutions. 4. The method according to claim 1 or 2, further comprising modifying the VH1 sequence of the antibody or an antigen-binding fragment thereof so that it includes substitutions 59X ²⁰ , where X ²⁰ is selected from R, K and H. 5. The method according to claim 3 or 4, characterized in that the antibody or its antigen-binding fragment contains, in addition to those subjected to replacement, the amino acid sequences of the heavy chain and light chain of the antibody AF, AK, AL, AN or AO, presented in FIG. 1B (SEQ ID NOs: 114 and 116; 134 and 136; 138 and 140; 146 and 148; and 150 and 152, respectively). 6. The method according to claim 2 or 5, characterized in that the antibody or antigen-binding fragment thereof contains a heavy chain amino acid sequence containing SEQ ID NO: 114, except for the substitutions 82R, 94S and 95R, and a light chain amino acid sequence containing SEQ ID NO: 116. 7. The method according to claim 2 or 5, characterized in that the antibody or antigen-binding fragment thereof contains a heavy chain amino acid sequence containing SEQ ID NO: 134, except for the substitutions 82R, 94S and 95R, and a light chain amino acid sequence containing -