RED BLOOD CELL TARGETED FACTOR VIII AND METHOD OF USING THE SAME
Targeted Factor VIII molecules comprising a Factor VIII linked with at least one domain that specifically binds to a membrane protein on a red blood cell is provided. The disclosed targeted coagulation factors prolong their duration of action and thus, are an improvement for the treatment of hematological diseases such as hemophilia A.
Provided herein are Factor VIII (FVIII) compounds targeted to red blood cells having an improved half-life and/or increased efficacy. Further provided are methods of treating patients suffering from a coagulation factor deficiency disorder by selectively targeting coagulation factors to their biological sites of action, such as by targeting Factor VIII (FVIII) to red blood cells. In addition, provided herein are single chain antibodies to glycophorin A on red blood cells.
BACKGROUNDThe effectiveness of biological drugs is often limited by their duration of action in patients, particularly when the disease requires continuous modulation by the drug. Consequently, enhancement of pharmacokinetic properties is often more critical to the success of a therapeutic agent in the clinic than is optimization of the drug's potency. One approach to protect drugs from various mechanism of clearance so as to prolong the half-life is to add targeting domains that promote drug binding to long-lived proteins in circulation such as matrix proteins, or to the surface of cells, such as blood cells or endothelial cells. For example, localization of therapeutic peptides or proteins to blood cell surfaces has been shown to prolong their circulation half-life by preventing normal clearance mechanisms (Chen, et al., Blood 105(10):3902-3909, 2005). A wide variety of molecules may be used as the targeting domain.
In another instance, when the Kunitz-type protease inhibitor (KPI) domain of tick anticoagulant protein was linked with an anionic phospholipid, phosphatidyl-L-serine (PS) binding protein, annexin V (ANV), the fusion protein (ANV-KPI) was shown to be more active and possess higher in vivo antithrombotic activities than the non-fusion counterpart (Chen, et al., 2005). Because ANV has strong affinities for PS and phosphatidylethanolamine (PE), it is hypothesized that the fusion protein ANV-KPI can be specifically targeted to the PS/PE-rich anionic membrane-associated coagulation enzyme complexes present at sites of thrombogenesis. Similarly, Dong, et al., reported fusing the fibrin-selective Desmodus rotundus salivary PA α1 (dsPA α1) to a urokinase (uPA)/anti-P-selectin antibody (HuSZ51) to produce a fusion protein that is fully functional with similar antithrombotic activities as the non-fusion counterpart in in vitro assays. Furthermore, the fusion protein HuSZ51-dsPA α1 was shown to bind to thrombin-activated human and dog platelets (Dong, et al., Thromb. Haemost. 92:956-965, 2004).
Other efforts have been made in targeting anticoagulants to prevent clots and to reduce mortality associated with thrombotic diseases (see, e.g., WO 94/09034). A more recent development is demonstrated by Stoll, et al., (Arterioscler. Thromb. Vasc. Biol. 27:1206-1212, 2007), in which a Factor Xa (FXa) inhibitor, tick anticoagulant peptide (TAP), was targeted to ligand-induced binding sites (LIBS) on GPIIb/IIIa, a glycoprotein abundantly expressed on the platelet surface, via an anti-LIBS single-chain antibody (scFvanti-LIBS). The fusion protein scFvanti-LIBS-TAP was shown to possess an effective anticoagulation activity even at low doses at which the non-targeted counterpart failed.
The aforementioned targeted anticoagulants were fusion proteins designed to target specific cells. According to Stoll, et al., the targeted anticoagulant should be a small molecule with a highly potent coagulation inhibition activity that is retained while fused to an antibody. The release of the anticoagulant from the fusion proteins in its targeted sites was not discussed.
Muzykantov et al (U.S. Pat. No. 8,333,973) discloses fusion proteins of a single chain antigen-binding domain to an anti-thrombotic agent, specifically thrombomodulin. While Muzykantov et al also suggest targeting glycophorin A with the scFv, only a murine antibody to glycophorin A is disclosed. No other antibodies or scFvs were disclosed.
Additionally, others have made efforts to target other blood cells. For example, Hilden et al (US2016/0263230A1) discloses targeting of FVIIa to platelets by fusion to an monoclonal antibody or fragment thereof that binds to TREM-like transcript 1 protein (TLT-1). The antibody was attached to FVIIa either by conjugation to a glycosylation site or via direct genetic fusion to FVIIa.
The present disclosure focuses on targeting therapeutic proteins for the treatment of hematological diseases such as hemophilia. For example, current treatment of hemophilia A patients with FVIII concentrates or recombinant FVIII is limited by the high cost of these factors and their relatively short duration of action. Hemophilia A patients are currently treated by intravenous administration of FVIII on demand or as a prophylactic therapy administered several times a week. For prophylactic treatment, FVIII is administered three times a week. Unfortunately, this frequency is cost prohibitive for many patients. Because of its short half-life in man, FVIII must be administered frequently. Despite its large size of greater than 300 kD for the full-length protein, FVIII has a half-life in humans of only about 11-18 (average 14) hours (Gruppo, et al., Haemophila 9:251-260, 2003). For those who can afford the frequent dosing recommended, it is nevertheless very inconvenient to frequently intravenously inject the protein. It would be more convenient for the patients if a FVIII product could be developed that had a longer half-life and therefore required less frequent administration. Furthermore, the cost of treatment could be reduced if the half-life were increased because fewer dosages may then be required. It is therefore desirable to have more efficient forms of FVIII that can lower the effective dose or have a prolonged duration of action to significantly improve treatment options for hemophiliacs.
Also, a sustained plasma concentration of targeted FVIII may reduce the extent of adverse side effects by reducing the trough to peak levels of FVIII, thus eliminating the need to introduce super-physiological levels of protein at early time-points. Therefore, it is desirable to have forms of FVIII that have sustained duration and a longer half-life than current marketed forms. It is also desirable to maintain trough levels of FVIII above 5%, or above 10%, or ideally above 15% of normal FVIII levels because these levels will reduce or eliminate breakthrough bleeds that can occur when FVIII levels are between 1% and 5% FVIII.
An additional disadvantage to the current therapy is that about 25-30% of patients develop antibodies that inhibit FVIII activity (Saenko, et al., Haemophilia 8:1-11, 2002). Antibody development prevents the use of FVIII as a replacement therapy, forcing this group of patients to seek an even more expensive treatment with high-dose recombinant Factor Vila (FVIIa) and immune tolerance therapy. A less immunogenic FVIII replacement product is therefore desirable.
One approach in improving the treatment for hemophiliacs involves gene therapy. Ectopically targeting FVIII to platelets by directing FVIII expression in platelets can have therapeutic effects in the treatment of hemophilia A in animal models (Shi, et al., J. Clin. Invest. 116(7):1974-1982, 2006).
It is an object to provide targeted coagulation factors that have prolonged duration of action, greater efficacy, fewer side effects, and less immunogenicity compared to the untargeted protein.
Another object is to reduce side effects associated with therapeutic protein administration by having the protein targeted to the specific site of desired action and thereby reducing the exposure of the protein to other potential biologically active sites that may result in undesired side effects.
A further object is to obtain further advantages by designing targeted therapeutic coagulation factors in which the therapeutic protein is released from the targeting domain in the immediate vicinity of its site of action in vivo. A high local concentration of the non-fusion, activated proteins may be achieved. Thus, the therapeutic efficacy of the proteins is enhanced.
SUMMARYProvided herein are recombinant fusion proteins comprising a functional Factor VIII polypeptide, at least one binding domain that specifically binds to a membrane protein on a red blood cell.
In some embodiments, the functional Factor VIII polypeptide is a full-length Factor VIII or a B-domain deleted Factor VIII. In some embodiments, the membrane protein is glycophorin A or Band3.
In some embodiments, the binding domain is an antibody, an antibody fragment, a scFv, a peptide, a peptide mimetic, or a small molecule. In further embodiments, the binding domain is a scFv and the membrane protein is glycophorin A.
In some embodiments, the scFv comprises a heavy chain selected from the group consisting of SEQ IS NOS: 33, 35, 37, 39, 41, 43, 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65 and 67, and/or the scFv comprises a light chain selected from the group consisting of SEQ IS NOS: 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64 and 66.
In some embodiments, the scFv may be strategically located in the Factor VIII. In some embodiments, the scFv is inserted into the functional Factor VIII polypeptide in the B-domain and most of the B-domain is removed. In some embodiments, the scFv is fused at the N-terminus of the functional Factor VIII polypeptide. In other embodiments, the scFv is fused at the C-terminus of the functional Factor VIII polypeptide.
In some embodiments, the functional Factor VIII polypeptide comprises a reduced or no binding to vWF. To achieve this, in some embodiments, the functional Factor VIII polypeptide comprises a deletion of the a3 domain of the functional Factor VIII polypeptide. In some embodiments, the deletion of the a3 domain of the functional Factor VIII polypeptide comprises deletion of residues 1652 to 1682 of Factor VIII.
In some embodiments, the functional Factor VIII polypeptide comprises a composition of predominantly a 1 chain form. The 1 chain form can be generated by removing the furin proteolytic cleavage site at residue 1648 of the functional Factor VIII or by deleting residues 1645-1648 of the functional Factor VIII or by deleting residues 1637-1651 of the functional Factor VIII.
In specific embodiments, the recombinant fusion protein of claim 1, wherein the recombinant fusion protein comprises the sequence selected from the group consisting of SEQ ID NO: 113-133.
Also disclosed are pharmaceutical compositions comprising a therapeutically effective amount of the recombinant fusion protein as described above and a pharmaceutically acceptable excipient or carrier. Further disclosed are methods for treating hematological diseases comprising administering an effective amount of the recombinant fusion protein to a patient in need thereof.
While a majority of this disclosure focuses on Factor VIII, it is envisioned that such targeting domains may be used to extend the half-life of any protein where half-life extension is desired. Thus, also disclosed are recombinant fusion proteins comprising a protein wherein extension of circulating half-life would be beneficial to a patient, and at least one binding domain that specifically binds to a membrane protein on a red blood cell. In some embodiments, the membrane protein is glycophorin A or Band3. In some embodiments, the binding domain is an antibody, an antibody fragment, a scFv, a peptide, a peptide mimetic, or a small molecule. In some embodiments, the binding domain is a scFv and the membrane protein is glycophorin A.
In specific embodiments, the scFv comprises a heavy chain selected from the group consisting of SEQ IS NOS: 33, 35, 37, 39, 41, 43, 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65 and 67, and/or comprises a light chain selected from the group consisting of SEQ IS NOS: 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64 and 66.
The present disclosure is directed to targeting a coagulation factor to blood cells in order to prolong the circulating half-life and thus reduce the frequency of administration of the coagulation factor. In one embodiment, a targeted coagulation factor is provided that is specifically targeted to erythrocytes through linking the factor to at least one domain that binds to a membrane protein on the surface of erythrocytes.
Erythrocytes, also referred to as red blood cells are very abundant, composing about 50% of total blood volume with a cell density of about 5×109 per ml in normal human blood. In addition erythrocytes have a long lifespan in circulation of about 120 days in humans, 95 days in primates and about 30 days in mice. The abundance and long life-span make erythrocytes an attractive target cell to which to bind a coagulation factor in order to prolong it's circulating half-life. In addition erythrocytes are a common component of clots such that coagulation factors present on the surface of erythrocytes are likely to be available a sites of injury to promote stable clot formation.
The cell surface protein on the erythrocyte to which the coagulation factor is targeted should ideally be an abundant protein in order to facilitate efficient binding of the targeted coagulation factor. In addition, binding to this cell surface protein should not interfere with the function of the erythrocyte. Examples of proteins with these characteristics are glycophorin A and Band 3 (AE1). Approximately 0.5×106 to 1×106 copies of the glycophorin A protein are present on each erythrocyte. Glycophorin A (also referred to GPA in this disclosure) is a structural component of the erythrocyte membrane and also contributes to the overall negative charge of the erythrocyte surface. Similarly, Band 3 is a structural component of the erythrocyte membrane and is present at about 106 copies per erythrocytes. While Band 3 is a structural component of the erythrocyte, it also functions as an anion transporter.
The domain for targeting the coagulation factor to the blood cell may be without limitation an antibody fragment, an antibody, a peptide, a receptor ligand, a carbohydrate, or a small molecule that has a high affinity to a membrane protein on the surface of the blood cell. The blood cell for example is a red blood cell or a platelet.
As used herein, “coagulation factor” refers to a protein that is involved in the coagulation cascade and has predominantly procoagulant activity. Coagulation factors are well known in the art and include without limitation coagulation factors I, II, V, VI, VII, VIII, IX, X, XI, XII, and XIII, and protein S. The coagulation factors may be concentrated from plasma or may be recombinantly produced. If recombinantly produced, the coagulation factors may have an amino acid structure that varies from the natural structure as long as sufficient procoagulant activity is maintained such that the variant is therapeutically useful. In one embodiment, the coagulation factor is a functional FVIII polypeptide, such as without limitation a FVIII concentrate from plasma or recombinantly produced FVIII, or Factor IX (FIX).
“Functional FVIII polypeptide” as used herein denotes a functional polypeptide or combination of polypeptides that are capable, in vivo or in vitro, of correcting human FVIII deficiencies, characterized, for example, by hemophilia A. FVIII has multiple degradation or processed forms in the natural state. These are proteolytically derived from a precursor, one chain protein. A functional FVIII polypeptide includes such single chain protein and also provides for these various degradation products that have the biological activity of correcting human FVIII deficiencies. Allelic variations likely exist. The functional FVIII polypeptides include all such allelic variations, glycosylated versions, modifications and fragments resulting in derivatives of FVIII so long as they contain the functional segment of human FVIII and the essential, characteristic human FVIII functional activity. Those derivatives of FVIII possessing the requisite functional activity can readily be identified by straightforward in vitro tests described herein. Furthermore, functional FVIII polypeptide is capable of catalyzing the conversion of Factor X (FX) to FXa in the presence of Factor IXa (FIXa), calcium, and phospholipid, as well as correcting the coagulation defect in plasma derived from hemophilia A affected individuals. From the published sequence of the human FVIII amino acid sequence and the published information on its functional regions, the fragments that can be derived via restriction enzyme cutting of the DNA or proteolytic or other degradation of human FVIII protein will be apparent to those skilled in the art. Specifically included within functional FVIII polypeptides without limitation is full-length human FVIII (e.g., SEQ ID NO: 1 and SEQ ID NO: 2) and B-domain deleted factor VIII (e.g., SEQ ID NO: 3 and SEQ ID NO: 4) and having the amino acid sequences as disclosed in WO 2006/053299.
The primary polypeptide chain of both full length and B-domain deleted FVIII are normally partially proteolytically cleaved during their expression to produce a mixture of molecules composed of heavy and light chains held together by a non-covalent interaction and a smaller proportion of un-cleaved single chain FVIII. A predominantly single chain FVIII can be generated by mutation of the two residues R1313 and R1648 within full length human FVIII as described previously (Pittman et al, J. Biol Chem (1994) vol 269, p 17329-17337). Mutation of residue R1313 inactivates a proteolytic cleavage site within the B-domain of FVIII while mutation of residue R1648 inactivates a proteolytic cleavage site the lies between the end of the B-domain and the start of the a3 domain. Thus it can be inferred that in the context of a B-domain deleted FVIII molecule the inactivation of the proteolytic cleavage site at R1648 will be sufficient to generate a predominantly single chain FVIII protein. The proteolytic cleavage at R1648 is thought to be catalyzed by an enzyme of the subtilisin family of proteases. The sequence RHQR (residues 1645 to 1648) within full-length human FVIII that borders residue R1648 is similar to that of the furin protease and thus has been called the furin cleavage site. Thus it is well described that inactivation of this furin cleavage site either by mutation or deletion will lead to the generation of a primarily single chain FVIII protein. A potential advantage of a single chain FVIII protein is improved stability because the heavy and light chains will be held together by a covalent bond as well as by non-covalent interactions. A predominantly single chain FVIII protein is also a more homogeneous molecule which is a preferred property of therapeutic products.
FVIII binds with high affinity (KD about 0.5 nM) to von Willebrand Factor (vWF), a protein found in the blood of humans and mammalian species. Thus, FVIII circulates in blood as a complex with vWF. The interaction with vWF reduces clearance of FVIII via the liver presumably because vWF reduces binding to the clearance receptor. The binding site for vWF within FVIII lies primarily within the acidic a3 domain, a 43 amino acid domain between the B-domain and the A3 domain (Fay, Int. J Hemat (2006) 83, 103-108). Deletion of the a3 domain was reported to prevent binding to vWF and result in a shorter half-life in mice (Tang et al 2013, Hemophilia 19, 539-545). While vWF binding to FVIII prolongs circulation half-life it is possible that vWF binding might interfere with the binding of a targeted FVIII molecule to erythrocytes by sterically masking the ability of the targeting moiety to bind to its target on the erythrocyte.
“Procoagulant activity” of FVIII refers to the activity of FVIII in the coagulation cascade. FVIII itself does not cause coagulation, but plays an essential role in the coagulation cascade. The role of FVIII in coagulation is to be activated to FVIIIa, which is a catalytic cofactor for intrinsic FX activation (Thompson, Semin. Thromb. Hemost. 29:11-22, 2003). FVIII is proteolytically activated by thrombin or FXa, which dissociates it from von Willebrand factor (vWf) and activates its procoagulant function in the cascade. In its active form, FVIIIa functions as a cofactor for the FX activation enzyme complex in the intrinsic pathway of blood coagulation, and it is decreased or nonfunctional in patients with hemophilia A.
“FIX” means coagulation factor IX, which is also known as human clotting factor IX, or plasma thromboplastin component.
As used herein, the term “targeted coagulation factor” refers to a coagulation factor that is coupled with at least one domain that specifically binds to a membrane protein on a blood cell. The targeted coagulation factor should bind potently to the blood cells, for example, with an affinity of less than 10 nM. Binding should be specific to the targeted blood cells, for example, through binding to membrane proteins selectively expressed on the targeted cell.
“Targeting domain” as used herein refers to a moiety that has a high affinity for membrane proteins on target cells. Targeting domains suitable for the present invention include, but are not limited to, antibodies, antibody fragments, such as single chain antibodies (scFv) or FAB fragments, antibody mimetics, and peptides or small molecules with high affinity for membrane proteins on the surface of the blood cells.
The coagulation factor can be coupled with the domain either chemically or by recombinant expression of a fusion protein. Chemical linkage can be achieved by linking together chemical moieties present on the coagulation factor and the targeting domain, including chemical linkages using moieties such as amino, carboxyl, sulfydryl, hydroxyl groups, and carbohydrate groups. A variety of homo- and hetero-bifunctional linkers can be used that have groups that are activated, or can be activated to link to attach these moieties. Some useful reactive groups on linker molecules include maleimides, N-hydroxy-succinamic esters and hyrazides. Many different spacers of different chemical composition and length can be used for separating these reactive groups including, for example, polyethylene glycol (PEG), aliphatic groups, alkylene groups, cycloalkylene groups, fused or linked aryl groups, peptides and/or peptidyl mimetics of one to 20 amino acids or amino acid analogs in length. For example, the domain may be linked with the coagulation factor in such a way that in vivo a functional form of the coagulation factor would be released from its targeted domain or the release occurs at or near the site of biological activity of the coagulation factor in the body.
In another method, the coagulation factor can be fused to a single chain antibody fragment or a peptide, wherein its coding sequence can be genetically linked with the FVIII coding sequence to produce a fusion protein using recombinant technology. Use of scFv may avoid cross-linking of binding sites or determinants thereby avoiding potentially harmful cell membrane modification and cell aggregation.
Accordingly, in one embodiment of the invention, a targeted coagulation factor is provided wherein the linkage attaching the coagulation factor to the targeting domain for targeting the coagulation factor to the blood cell can be cleaved or degraded thereby releasing the coagulation factor from the conjugate.
The targeted coagulation factor as described herein can be prepared by linking (fusing) the above-described scFv capable of binding a determinant expressed on the surface of a red blood cell to the coagulation factor, e.g., FVIII. Moreover, genetic engineering allows the design and synthesis of targeted coagulation factor which can be cleaved by pathophysiologically relevant enzymes that are generated at the site of disease that cannot be attained using chemical conjugation.
As noted above, linkers may also be utilized to join variable heavy and variable light chain fragments. A linker as used herein refers to a chain of as short as about 1 amino acid to as long as about 100 amino acids, or longer. In a further embodiment, the linker is 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 amino acids in length. In one embodiment, the linker is 13 amino acids in length.
Further, a cleavage sequence, such as the thrombin-sensitive cleavage sequence or other enzyme cleavage sequence, can be inserted in the linker to provide for release of the drug when the RBC to which it is targeted encounters the appropriate cleaving enzyme at the site of the pathological condition, e.g., upon active coagulation. This cleavage sequence may be located within a linker or at a terminus thereof.
In another embodiment, antibody-derived scFv with a thrombin releasing site can be cloned by an upstream primer, which anneals to the carboxy terminus and introduces the sequence including a short peptide linker with the thrombin cleavage site. In still another embodiment, the cleavage site is internal to the targeted coagulation factor itself.
The release of the coagulation factors from their conjugate form (i.e., from the targeted coagulation factor) can be achieved by linking the targeting domain to a site on the coagulation factor that is removed during its activation process, or by using a linker that degrades in a controlled manner by enzymes in the blood. For example, sugar polymers or peptides can be used that are susceptible to general blood proteases or hydrolases. A variety of such technologies is known in the art and has been used to make pro-drugs. The linker could be further engineered to be cleaved specifically at sites where the coagulation factors are most needed, such as sites of inflammation or blood coagulation triggered through trauma. For example, the linker may be susceptible to specific proteases produced only at the desired site of action, such as proteases released by the inflammation process or generated by the blood coagulation cascade. This selective release of the therapeutic protein may lower the potential for side effects and increase the efficiency of the protein at its site of action.
Targeting FVIII and FIXTargeting FVIII and FIX to the surface of blood cells, such as platelets or red blood cells, may serve to slow the clearance of these coagulation factors by preventing mechanisms that normally eliminate these molecules from circulation. Such mechanism of protein clearance may include the interaction of the free protein in plasma with one or more specific endocytic receptors; proteolysis; secretion via kidneys; and extravasation. For FVIII, LRP1 and the asialoglycoprotein receptor (ASGPR) on liver cells are important means of factor FVIII clearance. The role of proteolysis on the duration of action of FVIII is uncertain. FVIII bound to RBCs, however, may potentially protect against proteolysis by the action of its dense cell surface layer of glycans as has been documented previously for other molecules. Since most plasma FVIII forms complexes with vWF, clearance mechanisms for vWF may also be an important route of plasma FVIII clearance. The mechanisms of vWF clearance are not well characterized. Targeting FVIII to the surface of red blood cells is of particular interest because of their long circulation time. In humans the average life span of red blood cell is about 120 days and the abundance in blood is 5×109 cells per ml. By comparison the average life span of a platelet is only 7 to 10 days in humans and their abundance is between 1.5×108 to 4×108/ml. The longer life span and higher abundance of red blood cells compared to platelets make red blood cells more attractive target cells for increasing the half-life of targeted coagulation factors.
For targeted FVIII to promote coagulation, the molecule must be capable of being processed to a functional form (FVIIIa). Additionally, it may be beneficial if the FVIII has been released from its binding site, e.g. glycophorin A. However it is also possible that activated FVIII that remained bound would also be functional in promoting coagulation via complex formation with FIXa and Factor X. In one embodiment, this is achieved by linking a glycophorin A targeting domain to FVIII in place of the B-domain of FVIII (see
In other embodiments, the targeting domain can be linked to the N-terminus or C-terminus of FVIII in which case a linker sequence containing a protease cleavage site such as a thrombin cleavage site to enable the release of FVIII from the red blood cell surface in a pro-coagulant environment (See
The linkage between FVIII and the targeting domain can be achieved by covalently binding the targeting domain to reactive groups on FVIII, including amino, sulfhydryl, carboxyl groups and carbonyl groups using cross-linking approaches described herein. Targeting domains can also be coupled to carbohydrate present mostly on the B-domain of the FVIII molecule. For example, mild oxidation of FVIII with periodate produces aldehydes on carbohydrate chains, which can then be reacted with amines or hyrazides, followed optionally by reduction to form more stable linkages.
Free cysteine can be selectively generated on the B-domain of recombinant FVIII through mild reduction with Tris(2-carboxyethyl)phosphine (TCEP), allowing specific linking of the B-domain with a targeting domain that reacts with a free cysteine, such as a domain containing a thiol, triflate, tresylate, aziridine, oxirane, S-pyridyl, or maleimide moiety. Furthermore, FVIII can be modified to replace an amino acid residue with cysteine to provide a specific location for attachment to a targeting domain. If a B-domain deleted FVIII is used, a variety of cysteine muteins of B-domain deleted FVIII, such as those disclosed in WO 2006/053299, can be used to link FVIII with a targeting domain through chemical binding at a surface cysteine residue. Examples of amino acid residues that may be modified to replace an amino acid residue with cysteine include, but not limited to, 81, 129, 377, 378, 468, 487, 491, 504, 556, 570, 1648, 1795, 1796, 1803, 1804, 1808, 1810, 1864, 1911, 2091, 2118, and 2284 (the amino acid residue is designated by its position in the sequence of full-length FVIII).
The coagulation factor may also be coupled to the targeting domain using recombinant technology. Host cells may be transfected with a vector comprising a fusion protein of FVIII and the targeting domain. In one embodiment, the targeting domain may be inserted in place of the B-domain of B-domain deleted FVIII. As illustrated in
The host cell line may be any cell known to those skilled in the art as useful for producing a coagulation factor such as without limitation for FVIII CHO cells, HEK cells, BHK cells, and HKB11 cells (a hybrid of a human embryonic kidney cell line, HEK293 and a human Burkitt B cell lymphoma line, 2B8).
A number of domains can be linked chemically to FVIII, or recombinantly expressed with FVIII, to target FVIII to glycophorin A on the surface of red blood cells. Examples of such domains include, but are not limited to, antibodies against glycophorin A, peptide mimetics, or small molecule mimetics targeting glycophorin A. Antibodies, such as single chain antibodies (scFv) or FAB fragments targeting glycophorin A, are particularly useful as targeting domains.
It has been shown that the B-domain of FVIII can be removed without loss of FVIII function. Additionally, it has been also shown that various B-domain truncated forms of FVIII and B-domain fusions with other protein domains can yield functionally active FVIII. In one aspect, the invention involves targeting domains that can be engineered to insert into, replace, or partially replace the B-domain of FVIII without blocking the normal processing of the molecule to yield active FVIII. For example, using recombinant DNA technology, a FVIII molecule can be produced in which single chain antibody fragments are fused to the C-terminus of the B-domain of FVIII. Alternatively, scFv fragments can also be used to replace the whole or a part of the B-domain of FVIII. This can be achieved through insertion of the DNA sequence encoding the scFv fragments, in frame, after the B-domain coding sequence, or replacing some or all of the B-domain coding sequence. This strategy will preserve thrombin cleavage sites required for normal proteolyic activation of FVIII.
Use of an Antibody as the Targeting DomainAs used herein, an “antibody” refers to a whole antibody and any antigen binding fragment (i.e., “antigen-binding portion”) or single chain thereof. The term includes a full-length immunoglobulin molecule (e.g., an IgG antibody) that is naturally occurring or formed by normal immunoglobulin gene fragment recombinatorial processes, or an immunologically active portion of an immunoglobulin molecule, such as an antibody fragment, that retains the specific binding activity. Regardless of structure, an antibody fragment binds with the same antigen that is recognized by the full-length antibody. For example, an anti-glycophorin A monoclonal antibody fragment binds to an epitope of glycophorin A. The antigen-binding function of an antibody can be performed by fragments of a full-length antibody. Examples of binding fragments encompassed within the term “antigen-binding portion” of an antibody include (i) a Fab fragment, a monovalent fragment consisting of the VL, VH, CL and CH1 domains; (ii) a F(ab′)2 fragment, a bivalent fragment comprising two Fab fragments linked by a disulfide bridge at the hinge region; (iii) a Fd fragment consisting of the VH and CH1 domains; (iv) a Fv fragment consisting of the VL and VH domains of a single arm of an antibody, (v) a dAb fragment (Ward et al., (1989) Nature 341:544-546), which consists of a VH domain; and (vi) an isolated complementarity determining region (CDR). Furthermore, although the two domains of the Fv fragment, VL and VH, are coded for by separate genes, they can be joined, using recombinant methods, by a synthetic linker that enables them to be made as a single protein chain in which the VL and VH regions pair to form monovalent molecules (known as single chain Fv (scFv); see e.g., Bird et al. (1988) Science 242:423-426; and Huston et al (1988) Proc. Natl. Acad. Sci. USA 85:5879-5883). Such single chain antibodies are also intended to be encompassed within the term “antigen-binding portion” of an antibody. These antibody fragments are obtained using conventional techniques known to those with skill in the art, and the fragments are analyzed for utility in the same manner as are intact antibodies.
If an antibody is used as the targeting domain, a single chain fragment of the antibody, such as scFv or Fab fragment, can be used.
Furthermore, it is contemplated that an antigen binding fragment can be encompassed in an antibody mimetic. The term “antibody mimetic” or “mimetic” as used herein is meant a protein that exhibits binding similar to an antibody but is a smaller alternative antibody or a non-antibody protein. Such antibody mimetic can be comprised in a scaffold. The term “scaffold” refers to a polypeptide platform for the engineering of new products with tailored functions and characteristics.
The terms “monoclonal antibody” or “monoclonal antibody composition” as used herein refer to a preparation of antibody molecules of single molecular composition. A monoclonal antibody composition displays a single binding specificity and affinity for a particular epitope. Accordingly, the term “human monoclonal antibody” refers to antibodies displaying a single binding specificity which have variable and constant regions derived from human germline immunoglobulin sequences. The human antibodies of the invention may include amino acid residues not encoded by human germline immunoglobulin sequences (e.g., mutations introduced by random or site-specific mutagenesis in vitro or by somatic mutation in vivo).
An “isolated antibody,” as used herein, is intended to refer to an antibody which is substantially free of other antibodies having different antigenic specificities (e.g., an isolated antibody that binds to glycophorin A is substantially free of antibodies that bind antigens other than glycophorin A). An isolated antibody that binds to an epitope, isoform or variant of human glycophorin A may, however, have cross-reactivity to other related antigens, e.g., from other species (e.g., glycophorin A species homologs). Moreover, an isolated antibody may be substantially free of other cellular material and/or chemicals.
As used herein, “specific binding” refers to antibody binding to a predetermined antigen. Typically, the antibody binds with an affinity of at least about 105 M−1 and binds to the predetermined antigen with an affinity that is higher, for example at least two-fold greater, than its affinity for binding to an irrelevant antigen (e.g., BSA, casein) other than the predetermined antigen or a closely-related antigen. The phrases “an antibody recognizing an antigen” and “an antibody specific for an antigen” are used interchangeably herein with the term “an antibody which binds specifically to an antigen.”
As used herein, the term “high affinity” for an IgG antibody refers to a binding affinity of at least about 107M−1, in some embodiments at least about 108M−1, in some embodiments at least about 109M−1, 1010M−1, 1011M−1 or greater, e.g., up to 1013M−1 or greater. However, “high affinity” binding can vary for other antibody isotypes. For example, “high affinity” binding for an IgM isotype refers to a binding affinity of at least about 1.0×107M−1. As used herein, “isotype” refers to the antibody class (e.g., IgM or IgG1) that is encoded by heavy chain constant region genes. High affinity binding for a scFv refers to binding affinity of at least 109M−1 or greater.
“Complementarity-determining region” or “CDR” refers to one of three hypervariable regions within the variable region of the heavy chain or the variable region of the light chain of an antibody molecule that form the N-terminal antigen-binding surface that is complementary to the three-dimensional structure of the bound antigen. Proceeding from the N-terminus of a heavy or light chain, these complementarity-determining regions are denoted as “CDR1,” “CDR2,” and “CDR3,” respectively. CDRs are involved in antigen-antibody binding, and the CDR3 of the heavy chain comprises a unique region often of particular importance for specific antigen-antibody binding. An antigen-binding site, therefore, may include six CDRs, comprising the CDR regions from each of a heavy and a light chain V region.
As used herein, “conservative substitutions” refers to modifications of a polypeptide that involve the substitution of one or more amino acids for amino acids having similar biochemical properties that do not result in loss of a biological or biochemical function of the polypeptide. A “conservative amino acid substitution” is one in which the amino acid residue is replaced with an amino acid residue having a similar side chain. Families of amino acid residues having similar side chains have been defined in the art. These families include amino acids with basic side chains (e.g., lysine, arginine, histidine), acidic side chains (e.g., aspartic acid, glutamic acid), uncharged polar side chains (e.g., glycine, asparagine, glutamine, serine, threonine, tyrosine, cysteine), nonpolar side chains (e.g., alanine, valine, leucine, isoleucine, proline, phenylalanine, methionine, tryptophan), beta-branched side chains (e.g., threonine, valine, isoleucine), and aromatic side chains (e.g., tyrosine, phenylalanine, tryptophan, histidine). It is envisioned that the antibodies of the present invention may have conservative amino acid substitutions and still retain activity.
For nucleic acids and polypeptides, the term “substantial homology” indicates that two nucleic acids or two polypeptides, or designated sequences thereof, when optimally aligned and compared, are identical, with appropriate nucleotide or amino acid insertions or deletions, in at least about 80% of the nucleotides or amino acids, usually at least about 85%, preferably about 90%, 91%, 92%, 93%, 94%, or 95%, more preferably at least about 96%, 97%, 98%, 99%, 99.1%, 99.2%, 99.3%, 99.4%, or 99.5% of the nucleotides or amino acids. Alternatively, substantial homology for nucleic acids exists when the segments will hybridize under selective hybridization conditions to the complement of the strand. The invention includes nucleic acid sequences and polypeptide sequences having substantial homology to the specific nucleic acid sequences and amino acid sequences recited herein.
The percent identity between two sequences is a function of the number of identical positions shared by the sequences (i.e., % homology=# of identical positions/total # of positions×100), taking into account the number of gaps, and the length of each gap, which need to be introduced for optimal alignment of the two sequences. The comparison of sequences and determination of percent identity between two sequences can be accomplished using a mathematical algorithm, such as without limitation the AlignX™ module of VectorNTI™ (Invitrogen Corp., Carlsbad, Calif.). For AlignX™, the default parameters of multiple alignment are: gap opening penalty: 10; gap extension penalty: 0.05; gap separation penalty range: 8; % identity for alignment delay: 40. (further details found at http://www.invitrogen.com/site/us/en/home/LINNEA-Online-Guides/LINNEA-Communities/Vector-NTI-Community/Sequence-analysis-and-data-management-software-for-PCs/AlignX-Module-for-Vector-NTI-Advance.reg.us.html).
Another method for determining the best overall match between a query sequence (a sequence of the present invention) and a subject sequence, also referred to as a global sequence alignment, can be determined using the CLUSTALW computer program (Thompson et al., Nucleic Acids Research, 1994, 2(22): 4673-4680), which is based on the algorithm of Higgins et al., (Computer Applications in the Biosciences (CABIOS), 1992, 8(2): 189-191). In a sequence alignment the query and subject sequences are both DNA sequences. The result of said global sequence alignment is in percent identity. Preferred parameters used in a CLUSTALW alignment of DNA sequences to calculate percent identity via pairwise alignments are: Matrix=IUB, k-tuple=1, Number of Top Diagonals=5, Gap Penalty=3, Gap Open Penalty=10, Gap Extension Penalty=0.1. For multiple alignments, the following CLUSTALW parameters are preferred: Gap Opening Penalty=10, Gap Extension Parameter=0.05; Gap Separation Penalty Range=8; % Identity for Alignment Delay=40.
The nucleic acids may be present in whole cells, in a cell lysate, or in a partially purified or substantially pure form. A nucleic acid is “isolated” or “rendered substantially pure” when purified away from other cellular components with which it is normally associated in the natural environment. To isolate a nucleic acid, standard techniques such as the following may be used: alkaline/SDS treatment, CsCl banding, column chromatography, agarose gel electrophoresis and others well known in the art.
A goal of the antibody generation campaign was to identify antibodies that bound to cell surface proteins on red blood cells from humans that also cross reacted with other mammalian species, e.g. monkeys. Antibodies that cross react to humans and monkey are essential to enable testing in the monkey before proceeding to human testing. In addition antibodies that bind only to red blood cells and not to other cells types such as peripheral blood mononuclear cells and endothelial cells is essential to avoid off target effects. Furthermore, antibodies that specifically bind to glycophorin A (GPA) are highly desirable because GPA is red blood cell specific, very abundant on the red blood cell surface (0.5 to 1×106 copies per cell) and because knock out of GPA in mice as well as humans who naturally lack GPA have no pathology. Thus antibodies that bind to GPA are unlikely to have undesirable side effects. Although it is known that antibodies against GPA can cause lysis of red blood cells (hemolysis) this requires complement activation which requires large numbers of antibodies to be bound to a given red blood cell. At therapeutic doses of a red blood cell targeted Factor VIII molecule the number of copies per cell is estimated to be less than 100 and thus not sufficient to activate the complement system. Another potentially suitable target on red blood cells is Band3, a transmembrane protein that is also red blood cell specific and very abundant on the red blood cell surface. Therefore candidate antibodies were screened for binding to GPA or Band3.
Two approaches were used to generate antibodies against human red blood cell surface proteins that cross react to monkey red blood cell surface proteins. In one method mice were immunized with intact red blood cells. In a second approach phage display libraries of human antibodies were panned against intact red blood cells. The use of intact red blood cells ensures that the cell surface proteins to which antibodies may be generated are in their native physiologic conformation. This is especially important for cell surface proteins that contain membrane spanning regions and are typically hard to express recombinantly in their native conformation. Details of specific embodiments are provided in the Examples below.
Pharmaceutical Compositions and UsesThe invention also concerns pharmaceutical compositions comprising therapeutically effective amounts of the targeted coagulation factors of the invention and a pharmaceutically acceptable excipient or carrier. “Pharmaceutically acceptable excipient or carrier” is a substance that may be added to the active ingredient to help formulate or stabilize the preparation and causes no significant adverse toxicological effects to the patient. Examples of such excipients or carriers are well known to those skilled in the art and include water, sugars such as maltose or sucrose, albumin, salts, etc. Other excipients or carriers are described, for example, in Remington's Pharmaceutical Sciences (Mack Publishing Co., Easton, Pa., 20th edition, 2000). Such compositions will contain an effective amount of the targeted coagulation factors together with a suitable amount of excipients or carriers to prepare pharmaceutically acceptable compositions suitable for effective administration to a patient in need thereof.
For example, the conjugate may be parenterally administered to subjects suffering from hemophilia A at a dosage that may vary with the severity of the bleeding episode.
In one embodiment, the present invention concerns a method for treating hematological diseases comprising administering an therapeutically effective amount of the aforementioned targeted coagulation factor to a patient in need thereof.
As used herein, “therapeutically effective amount” means an amount of a targeted coagulation factor that is need to provide a desired level of the targeted factor (or corresponding unconjugated factor released from the targeted form) in the bloodstream or in the target tissue. The precise amount will depend upon numerous factors, including, but not limited to the components and physical characteristics of the therapeutic composition, frequency of dosing, intended patient population, individual patient considerations, and the like, and can readily be determined by one skilled in the art.
As used herein, “patient” refers to human or animal individuals receiving medical care and/or treatment.
The polypeptides, materials, compositions, and methods described herein are intended to be representative examples of the invention, and it will be understood that the scope of the invention is not limited by the scope of the examples. Those skilled in the art will recognize that the invention may be practiced with variations on the disclosed polypeptides, materials, compositions and methods, and such variations are regarded as within the ambit of the invention.
The following examples are presented to illustrate the invention described herein, but should not be construed as limiting the scope of the invention in any way.
EXAMPLESIn order that this invention may be better understood, the following examples are set forth. These examples are for the purpose of illustration only, and are not to be construed as limiting the scope of the invention in any manner. All publications mentioned herein are incorporated by reference in their entirety.
Example 1: Anti-GPA Antibody Generation: Mouse HybridomasAntibodies against cell surface proteins of human and monkey red blood cells (RBC) were generated by immunization of mice with human red blood cells followed by a boost immunization with monkey red blood cells. Hybridomas were generated from the spleen cells of these mice and these hybridomas were screened first by fluorescence activated cell sorting (FACS) against isolated human and monkey (Cynomolgus monkey and Rhesus monkey) red blood cells (RBC) and then by ELISA against purified human glycophorin A protein (GPA). Isolated RBC were prepared from whole blood from normal human, Cynomolgus monkey, and beagle dog, were anticoagulant with sodium heparin (Bioreclamation). 2-3 mL of whole blood was loaded gently on the top of 3 mL Ficoll Plus (Sigma Aldrich) then centrifuged at 500 g for 20 min at room temperature (RT). The RBC pellets were washed 3 times with 10 mL of PBS, and centrifuge 400 g for 10 min at RT. RBC pellets were re-suspended in PBS at the same total volume as the original blood volume to achieve the same RBC cell density as in whole blood and stored at 4° C. Purified human glycophorin A protein was obtained from Sigma Aldrich and coated on ELISA plates at Zug/ml. Binding of antibodies to human GPA was detected using appropriate secondary antibodies using methods well known in the art.
Clones that bound to both human and monkey RBC and were positive for GPA binding by ELISA were selected. Antibody clones that bound to human peripheral blood mononuclear cells (PBMC) and human endothelial cells were excluded. A summary of the hybridoma screening cascade is shown in
Twelve clones were identified as positive for binding to human and monkey red blood cells by FACS and to human GPA by ELISA as shown in Table 1 below. Only background binding was observed on human umbilical vein endothelial cells (HUVEC) and very low binding was seen on dog red blood cells. Hybridoma supernatants were used as a source of the IgG form of the antibodies. Clones 10B7 and 8G8 were excluded from further analysis due to inability to generate VH sequences from these clones.
The alignment of the sequences of the antibody variable regions from the 10 hybridoma clones that bound to both human and monkey red blood cells demonstrated that 9 out of the 10 antibodies contained unique but closely related complementarity determining regions (CDR), see
Nine of the 10 antibodies were recombinantly expressed in mammalian cells as a single chain Fv format fused to a mouse Fc fragment at the C terminus of the protein to improve expression and stability of the protein. The antibody clone identifier (ID) was shortened to the first identifier, for example “6C12B8/H8” was shortened to ‘6C12”. These scFv-Fc proteins were assessed for binding to human, Rhesus, Cynomolgus and Dog red blood cells by FACS as shown in Table 2. The results indicated that all but two of the antibodies from the “6C12 cluster” bound similarly to Human, and Rhesus RBC and to RBC isolated from 2 out of 3 individual Cynomolgus monkeys. Clones 1B3 and 1H5 had no binding to RBC which may relate to the quality of these protein preparations. The 7G4 antibody in the scFv-Fc format bound to human but not to monkey or dog RBC.
Shown in Table 3 is a summary of the sequences of the CDR regions of 7 of the antibodies that were positive in FACS using the scFv-Fc format. Differences to the sequences of the CDRs of 13G7 antibody are shown (an asterisk [*] indicates an identical residue at that position; and a dash [-] indicates a missing residue).
Based on these data, two antibodies from the 6C12 cluster were selected, namely 13G7 and 6C12, together with 7G4 (the human GPA specific antibody) for more detailed analysis. The IgG forms of these 3 antibodies were purified to homogeneity then tested in ELISA against different preparations of human GPA. The “Sigma GPA” is a crude preparation of human GPA extracted from human red blood cells and is therefore representative of the highly glycosylated form of the native protein. The kinetics of binding to the recombinantly expressed ectodomain of human GPA was determined using surface plasmon resonance. The ectodomain is composed of the 72 amino acids that comprise the extracellular region of GPA:
The lack of binding of the 7G4 IgG antibody to the de-sialylated human GPA (asialo-GPA) and to recombinant GPA (which was poorly glycosylated, data not shown) suggests that the epitope for this antibody contains a glycan component (Table 4). In contrast, the binding of 6C12 and 13G7 antibodies was not dependent on the sialylation status of the human GPA.
An ELISA was run against different preparations of human GPA, see Table 4. EC 50 values were determined from a dilution curve of the antibody. Sigma GPA is partially purified GPA extracted from human red blood cells and purchased from Sigma Chemical Company (cat #G5017), asialo hGPA is human GPA that was desialylated and ECD is the recombinantly expressed ectodomain which is the extracellular portion of GPA.
Surface plasmon resonance was performed to determine the kinetics of binding of the IgG forms of three antibodies to recombinant human GPA ectodomain as shown in Table 5.
To further confirm that antibodies 6C12 and 13G7 bind specifically to GPA, Chinese hamster ovary (CHO) cells (QMCF1 cells) were transfected with an expression plasmid encoding human GPA. While the 6C12 and 13G7 antibodies failed to bind to the non-transfected parental QMCF1 cells, both antibodies bound well to the QMCF1 cells transfected with the human GPA expression plasmid (
Based on these data, the 6C12 antibody sequence was selected for sequence humanization and affinity optimization. However, it should be recognized that the 6C12 antibody is representative of all of the nine antibodies identified in the 6C12 cluster since these antibodies all have closely related CDR sequences and similar binding to red blood cells.
Provided in SEQ ID NOs: 6-14 are the sequences of the murine hybridoma derived antibodies as scFv-Fc formats. The domain structure of these sequences is (from N to C terminus): Signal peptide, light chain variable domain, artificial linker (GGGGSGGGGSGGGGS, SEQ ID NO: 15), heavy chain variable domain, 230 amino acid Fc fragment domain. The artificial linker sequence between VH and VL domains is underlined. Bold text shows the heavy chain variable domain (located at the N-terminal side of the linker) and the light chain variable domain (located at the C-terminal side of the linker). The C-terminal 230 amino acids is the common antibody Fc domain that was added to each antibody to improve expression and stability.
Example 3: Humanization and Germlining of Select AntibodiesIn order to humanize the anti-GPA antibody 6C12 the mouse framework regions are replaced by human framework regions. While the variable regions remain unchanged the framework sequences can affect the ability of the antibody to bind to its epitope. Thus it is important to test a number of different human framework sequences for both the heavy and light chains. Humanization of antibody 6C12 was performed by generating a number of variants containing different pairs of heavy chain (columns) and light chain (rows) framework sequences as outlined in Table 6 and 7. A total of 23 variants were generated using 8 different heavy chain framework sequences and 3 different light chain framework sequences.
The antibodies in the above table were expressed in the IgG format and their binding to human glycophorin A was tested using an ELISA assay and the results are in Table 8.
The binding of the humanized variants to intact red blood cells (RBC) from humans (Hu01 to 03; 3 individuals), Rhesus monkey (Rh01 to 03; 3 individuals) and Cynomolgus monkey (Cy01 to 02; 2 individuals) was measured by fluorescence activated cell sorting (FACS). The results, shown in Table 9, indicate that some variants (−3, −7, −19, −20, −22, −23, −24) retained binding to both human and monkey RBC. Some variants (−3, −5, −6, −11, −12, −14, −15, −16) had reduced binding to human RBC. Some variants (−1, −4, −9, −10, −13, −17, −18, −21) lost all binding to human RBC. There was a good correlation between the GPA ELISA results and the binding to human RBC. In particular, variants 19, 20, 22, 23, 24 maintained the highest levels of binding to both human and monkey RBC and so were selected as possible therapeutic candidates. Humanized variants 20, 23 and 24 were preferred due to the differences to germline sequences. The light chain sequence of Hu6C12-20, Hu6C12-23 and Hu6C12-24 are identical. The heavy chain sequences of Hu6C12-20, Hu6C12-23 and Hu6C12-24 differ from each other by only 1 to 3 amino acids. The humanized antibodies that retained binding to human GPA and human RBC contained combinations of 3 different light chains (VK.1, VK.1a and VK.1b) and six different heavy chains (VH.1b, VH.1c, VH.1d, VH.1e, VH.1f, VH.1g), the sequences of which are included in Appendix A. The most promising candidates Hu6C12-20, Hu6C12-23 and Hu6C12-24 are composed of the same light chain VK.1b combined with the heavy chains VH.1c, VH.1f and VH.1g respectively. In particular Hu6C12-24 is preferred based on the lowest number of amino acid differences from human germline sequences.
Comparison of the sequences of the heavy chain variable domains of humanized variants of 6C12 (
Generation of germlined variants of Hu6C12-24 and testing of the binding to GPA was performed. Germlining is a process in which residues that differ from human germline sequences are changed back to the germline residue. This is performed to further reduce the risk of immunogenicity but can also provide information about which residues are critical for binding of the antibody to its target. Comparison of the sequence of the variable domains of Hu6C12-24 with sequences of human germline identified amino acids that differ from germline.
For the heavy chain variable domain 22 germline deviations were identified (7 in framework regions, 15 in CDR regions). For the light chain variable domain 16 germline deviations were identified (2 in framework regions, 14 in CDR regions). Each residue that deviated from germline was individually changed to the germline residue and the protein was expressed as a Fab format in E. coli. GPA is a transmembrane protein and therefore the correct physiologic structure of GPA may only be present when the protein is present within a cell membrane. In order to evaluate binding of antibodies to GPA in the native context of the red blood cell membrane we made use of isolated red blood cell membranes, also called red blood cells ghosts. These membranes are prepared by first isolating red blood cells from citrated whole blood. Citrated whole blood is layered on to Ficol and centrifuged at low speed for 20 minutes to pellet the red blood cells. The supernatant containing plasma and other blood cells is discarded and the red blood cell pellet is subsequently washed 3 times in 10 cell volumes of phosphate buffered saline (PBS). A 10 ml pellet of purified red blood cells is resuspended in 30 ml of cold low osmorality 20 mM phosphate buffer (pH7.4) and incubated for 30 mins to break open the RBC. After Centrifugation at 10,000 rpm for 30 min at 5° C. the hemoglobin containing supernatant is discarded. To the membrane pellet another 30 ml of 20 mM phosphate buffer is added and the sample centrifuged again at 10,000 rpm for 30 min at 5° C. to pellet the membrane fraction. This wash and centrifugation step is repeated another 4 times. The red blood cell membranes are then resuspended in 20 mL of PBS to achieve an equivalent of approximately 2.5E9 cells/mL and stored frozen at −20° C. Binding of antibodies to RBC ghost can be measured by ELISA in which the RBC ghost suspension (5×106 in 100 ul PBS) was added to each well of flat bottom 96-well ELISA microtiter plates (2HB, Immunolon) and the plates incubated overnight at 4° C. The plates were blocked with Starting blocking buffer (Thermo Fisher) for 2 hours at room temperature (RT). Antibodies or other test molecules were added to each well and incubated for 1.5 hours at RT. After 4 washes 100 ul of an appropriate detection antibody was added to each well and incubated for 1 hour at RT. After 4 washes the detection antibody was measured using appropriate detection reagents.
The germlined variants were assayed for binding to GPA protein and to RBC ghost membranes by ELISA and the results are summarized in Table 10. Binding and expression is relative to that of the hu6C12-24 parental antibody. The Rel Expn column indicates the expression level of the Fab protein relative to that of the parental Fab TPP-4935. The hRBC column is binding to human RBC membranes. The natGPA and recGPA columns are binding to native human GPA protein and recombinant human GPA ectodomain protein respectively.
These results demonstrate that for many of the non-germline residues the reversion to germline had a negative effect on binding to GPA protein and/or red blood cells. In particular amino acid changes in the heavy chain of T33Y, Y50I, R59S, and T98R significantly reduced binding demonstrating that these residues are important for recognition of the epitope. In particular amino acid changes in the light chain of Y33A, T50A, H88Q, R90L, F93Y significantly reduced binding demonstrating that these residues are also important for recognition of the epitope. However, 6 residues in the heavy chain (E1Q, 148M, D56G, D66G, L70M, T72R) and 5 residues in the light chain (S24R, A26S, S49A, P540, F95Y) were identified that either had no effect on binding or increased binding. Residues D56G, S49A and P54Q significantly increased the binding to GPA and to RBC membranes.
It was noted that the CDR-L1 loop has an untypical length of 10 amino acids and that repair of this loop length by insertion of a serine at position 30 abrogated binding to GPA by about 70 to 80% indicating the importance the atypical CDR-L1 loop configuration.
Critical residues where more than 80% of the binding to GPA was inhibited when mutated to germline were identified in all CDRs. A predominant role of CDR-L3 was revealed by the mutations H88Q, R90L and F93Y within CDR-L3 that dramatically reduced the binding to GPA. These data also confirm the finding from the humanization that the un-typical heavy chain CDR3 residue T98 is critical for binding.
Combining individual germline reversions that have no effect on binding or improve binding within a single protein was also examined. The results of these combinations is shown in Table 11 (Rel Expn: relative expression level of the Fab protein; Hu-nat GPA: native human GPA protein; Hu-rec GPA; recombinant human GPA ectodomain protein; hRBC: human RBC membranes; Cyno_rec GPA: Cynomolgus monkey recombinant GPA ectodoamin; Cyno RBC: Cynomolgus monkey RBC membranes. Proteins or membranes were coated on plates and binding of antibodies was determined by ELISA. Binding is reported as relative to the parental antibody. Blank cells indicate that binding was not tested).
In particular when 10 or 11 of the germline reversions were included in to a single protein as in TPP-5906 and TPP-5907 the binding to human GPA protein was increased 7-fold to 13-fold and the binding to human RBC membranes was increased 90-fold. Binding to Cyno GPA protein and Cyno RBC membranes was also increased for these variants by about 1.5 to 2.5-fold. These data demonstrate that mutations VL: (S24R, A26S, S49A, P540, F95Y); VH: (148M, D56G, D66G, L70M, T72R) or VL: (S24R, A26S, S49A, P540, F95Y); VH: (148M, S54G, D56G, D66G, L70M, T72R) could improve the binding to human and cyno monkey GPA and red blood cells while reducing the sequence differences to human germline. Sequences of the humanized 6C12 parental Fab antibody and the final germlined 6C12 Fab antibodies is shown in Table 12.
Kinetic analysis of the antibody binding to human red blood cell membranes was also performed using surface plasmon resonance (SPR) using a Biacore instrument. In this assay membranes prepared from human red blood cells were attached to a L1 sensor matrix and antibodies were flowed over the RBC membranes. Data from single cycle kinetics as shown in
Additional SPR analysis of the different germlined variants of hu6C12-24 as shown in Table 13 confirmed the results seen by ELISA assay. In particular TPP-5906 and TPP-5907 exhibited affinities for human RBC membranes of 0.9 and 0.8 nM which is improved by about 200-fold compared to the parental antibody. Thus the Hu6C12 antibody variants TPP 5906 and TPP-5907 represent attractive lead candidates for binding to GPA on human red blood cells.
Surface plasmon resonance (SPR) analysis of the humanized and germlined 6C12 antibody TPP-5906 binding to human RBC membranes indicated that this antibody had a relatively fast on rate (Ka=4 E+06) but a fast off rate (Kd=3.4 E-03). Thus despite having an affinity (KD) of around 0.9 nM the estimated time for 50% dissociation from the RBC membranes was only a few minutes. A FVIII fusion protein containing this parental antibody inserted in place of the B-domain was able to bind to intact human red blood cells, but only about 60% binding was observed after about 1 hour in a plasma matrix. Therefore an antibody with a slower off rate would be beneficial at increasing the binding of FVIII-antibody fusions to red blood cells. To improve the off rate of the 6C12 antibody an affinity maturation process was used. The starting point for affinity maturation was the humanized 6C12-24 antibody that recognizes human GPA and cross reacts to monkey GPA. A CDR focused library containing amino acid substitutions at all positions within the CDR regions was generated. This library was screened for changes in binding to both human and Cyno monkey RBC membranes and recombinant glycophorin A protein. This enabled the identification of positions within the CDRs that either improved binding, reduced binding or had no effect when substituted with a specific amino acid. Based on these results 11 positions within the CDRs were selected which when altered to the specified amino acid resulted in improved binding. The improved binding for these positions was confirmed using FACS analysis of human and Cyno monkey RBC. A combinatorial library was then created on the backbone of the humanized and germlined 6C12 antibody variant TPP-5906 where each of these 11 positions was varied from the parental residue resulting in a theoretical complexity of 2048 different CDR sequences. In total this library contained approximately 700 unique antibodies. This library was screened using ELISA assay for binding to human and Cyno monkey RBC membranes and recombinant glycophorin A protein. Based on these results a set of 24 antibodies with the most improved binding to human and Cyno RBC membranes and recombinant GPA proteins was identified. Table 14 provides the sequences for these 24 antibodies.
The binding kinetics of the affinity matured Fab antibodies was also measured. The top 24 affinity matured antibodies identified from the high throughput screen of the library of approximately 700 unique variants of the humanized and germlined 6C12 antibody were evaluated for their binding to human RBC membranes using surface plasma resonance in a Biacore instrument. Fifteen of the Fabs generated interpretable data, another 4 had slow off rates but kinetic values could not be assigned, and another 6 did not give interpretable data. TPP-5906 is the parental humanized and germlined antibody before affinity maturation. The results are summarized in Table 15 as the on rate (Ka), off rate (kd), affinity (KD) and estimated time to 50% dissociation (T½).
The binding of affinity matured Fab antibodies to intact human RBC was also measured. Human RBC were isolated from fresh anti-coagulated normal human blood by centrifugation through a ficol gradient and washing in PBS buffer. RBC diluted to 1/10th of normal hematocrit were incubated with purified Fab antibodies at a final concentration of 7 ug/ml for 1 hour at room temperature on a rotating shaker after which a sample was taken and the RBC were pelleted by centrifugation and the supernatant was collected. The remaining RBC suspension was diluted 200-fold in PBS buffer and incubated for a total of 2 hours at room temperature on a rotating shaker and samples were removed at 0, 5 min, 20 min, 1 h and 2 h and the RBC pelleted and the supernatant collected. The amount of Fab in the supernatants was then measured using an ELISA assay and the results used to calculate the percentage of the input Fab that was bound to the RBC. Following dilution it is expected that antibodies with a faster off-rate and/or overall lower affinity will dissociate from the RBC and reach a new equilibrium binding. The percentage that remains bound after 200-fold dilution of the RBC suspension is an indication of the overall strength of the binding to human RBC. The results are summarized in Table 16 below. It can be seen that compared to the parental antibody (TPP-5906) the affinity matured antibodies exhibited significantly improved binding as indicated by the observation that binding is retained following 200-fold dilution. By contrast the parental antibody dissociated rapidly from the RBC such that <5% remained bound after 20 minutes. In general there was a good correlation between the binding to intact RBC and the affinity (KD) measured by SPR. For example TPP-7792, 7797, 7793, 7794, 7790 and 7783 which exhibited the strongest binding to intact RBC also had the highest affinities. From these results TPP-7796 was eliminated as a potential candidate. The remaining 10 antibodies (“the top 10 set”) were reformatted as single chain antibodies (scFv) with a C-terminal 6×HIS tag and expressed in mammalian HEK-293 cells and purified by affinity chromatography on NiNTA columns.
The binding of affinity matured Fab antibodies to intact Cynomolgus monkey RBC was measured. Cynomolgus monkey RBC were isolated from fresh anti-coagulated blood by centrifugation through a ficol gradient and washing in PBS buffer. RBC diluted to 1/10th of normal hematocrit were incubated with the Fab at a final concentration of 7 ug/ml for 1 hour at room temperature on a rotating shaker after which a sample was taken and the RBC were pelleted by centrifugation and the supernatant was collected. The remaining RBC suspension was diluted 200-fold in PBS buffer and incubated for a total of 2 hours at room temperature on a rotating shaker and samples were removed at 0, 5 min, 20 min, 1 h and 2 h and the RBC pelleted and the supernatant collected. Following dilution it is expected that antibodies with a faster off-rate and/or overall lower affinity will dissociate and reach a new equilibrium binding. The amount of Fab in the supernatants was then measured using an ELISA assay and the results used to calculate the percentage of the input Fab that was bound to the RBC. The results are summarized in Table 17. It can be seen that the affinity matured antibodies exhibited strong binding to Cynomolgus RBC as indicated by the observation that binding is retained following 200-fold dilution. In particular antibodies TPP-7792 and TPP-7790 remained 89% and 85% bound at 2 hours after 200-fold dilution. The other antibodies exhibited less tight binding to Cynomolgus monkey RBC as indicated by their reduced binding after dilution. Based on these results antibodies TPP-7792 and TPP-7790 were selected as antibody candidates that exhibit the tightest binding to both human and Cynomolgus monkey RBC. An antibody with tight binding to Cynomolgus monkey RBC is desirable to enable testing of drug candidates in Cynomolgus monkeys.
Human RBC were isolated from fresh anti-coagulated normal human blood by centrifugation through a ficol gradient and washing in PBS buffer. RBC diluted to 1/10th of normal hematocrit were incubated with the set of “Top 10” purified scFv antibodies at a final concentration of 7 ug/ml for 1 hour at room temperature on a rotating shaker after which a sample was taken and the RBC were pelleted by centrifugation and the supernatant was collected. The remaining RBC suspension was diluted 200-fold in PBS buffer and incubated for a total of 2 hours at room temperature on a rotating shaker and samples were removed at 0, 5 min, 20 min, 1 h and 2 h and the RBC pelleted and the supernatant collected. The amount of scFv antibody in the supernatants was then measured using an ELISA assay and the results used to calculate the percentage of the input scFv that was bound to the RBC. Following dilution it is expected that scFv antibodies with a faster off-rate and/or overall lower affinity will dissociate from the RBC and reach a new equilibrium binding. The percentage that remains bound after dilution is an indication of the overall strength of the binding to human RBC. The results are summarized in Table 18. The parental scFv (TPP-5906) exhibited poor binding to human RBC after dilution. Although all of the parental scFv was bound to RBC after the initial incubation with RBC at 1/10th normal hematocrit (when the effective concentration of the target protein is high), only 16% remained at 5 minutes after 200-fold dilution and all of the scFv had dissociated from the RBC after 20 minutes. By comparison all of the top 10 affinity matured scFv exhibited significantly improved binding to human RBC. All 10 of the affinity matured scFv retained 76% or more binding at 2 hours after dilution. Some of the scFv, for example TPP-7792, 7797, 7783, 7781 retained >90% binding at 2 hours.
Initial peptide mapping studies were performed in which a series of overlapping 20 residue peptides derived from the human GPA extracellular domain were bound to plates and then incubated with the antibody (data not shown). These experiments identified a 20 amino acid peptide (called GP5) containing residues 42 to 61 the human glycophorin A extracellular domain that was bound by the 6C12 IgG antibody. In order to determine if the affinity matured antibodies bound to the same peptide and further map the specific amino acids in the peptide required for binding a series of overlapping 8 amino acid long peptides were synthesized covering the sequence of the GP5 peptide as shown in Table 19.
These peptides were attached to the surface of a streptavidin coated plate via a biotin moiety that was present at the N-terminus of each peptide. Two concentrations of two of the affinity matured Fabs (TPP-7792 and TPP-7790) together with the parental humanized and germlined 6C12 Fab (TPP-5906) were incubated on the peptide coated plate and after a series of washing steps with PBS/0.05% tween the amount of antibody bound was detected using a secondary antibody against the 6×HIS tag present on the Fabs. The results are summarized Table 20. The affinity matured antibodies TPP-7792 and TPP-7790 bound to the 21 amino acid peptide GP5 that contains the epitope for the parental antibody 6C12. The affinity matured antibodies also bound to the 8 mer GP8 peptide, but only weakly to the neighboring peptide GP9, and did not bind to any of the other 8 amino acid overlapping peptides. The binding to the 8 amino acid GP8 peptide appeared to be weaker than to the 20 amino acid GP5 peptide as evidenced by the weaker signal at the lower antibody concentration. The parental antibody failed to bind to any of the 8 amino acid peptides likely because of the lower affinity of this antibody. These results indicate that the core epitope recognized by the affinity matured antibodies lies within the 8 amino acid sequence of peptide GP8, that being VRTVYPPE. Peptide GP7 (SVRTVYPP) was not recognized by the affinity matured antibodies indicating that the Glutamic acid (E) at the C-terminus of GP8 is essential for binding. Similarly peptide GP9 (RTVYPPEE) was recognized only weakly by the affinity matured antibodies indicating that the valine (V) at the C-terminus of the GP8 peptide is important for binding.
To further define the epitope of the affinity matured antibodies a series of peptides containing the sequence corresponding to the 20 amino acid GP5 peptide were synthesized in which each peptide contained 1 residue that was substituted with alanine (A) as shown in Table 21.
These alanine scanning peptides which also contained biotin at the N-terminus were attached to streptavidin plates and then incubated with affinity matured Fab antibodies TPP-7790 and TPP-7792 or the parental Fab TPP-5906. After extensive washing with PBS/0.05% tween buffer the bound antibody was detected using a secondary antibody against the 6×HIS tag present on the Fab and the results are summarized in Table 22.
The same epitope mapping study on the 20 alanine scanning peptides was performed for a total of 7 affinity matured Fabs (TPP #7782, 7783, 7778, 7790, 7792, 7793, 7797). A similar trend in binding was seen. The binding to peptides GP5-1, 5-2, 5-3, 5-4, 5-5, 5-6, 5-7 and 5-8 was similarly strong indicating that substitution of the first 8 residues for alanine had no effect upon binding. In order to visualize the average binding of all 7 affinity matured Fabs to the different peptides the mean binding to each peptide as a percentage of the average of the binding to GP5-1, 5-2, 5-3, 5-4, 5-5, 5-6, 5-7 and 5-8 was calculated and is plotted in the graph as shown in
The alanine substitution present in the peptides GP5-11 (EVSEISVRTVAPPEEETGER) and GP5-14 (EVSEISVRTVYPPAEETGER) reduced binding of the affinity matured Fabs by 100% and 90% respectively while the alanine substitution present in GP5-10 (EVSEISVRTAYPPEEETGER) reduced binding by about 20%. Therefore residues at positions 11 (Tyrosine) and 14 (glutamic acid), as defined by counting from the N-terminus of GPS, are essential for binding to the affinity matured Fabs. The residue at position 10 (valine) plays a minor role. These results are consistent with the observation described above in which the peptide scanning analysis defined the core epitope as an 8 amino acid sequence present in GP8 (VRTVYPPE) which span from residues 8 to 14 of GP5 and thus contains residues 11 and 14 that were defined as critical by the alanine scanning.
The pattern of binding of the paternal 6C12 Fab antibody to the alanine scanning peptides was different to that seen for the affinity matured Fabs. The binding of the parental Fab was sensitive to alanine substitution at an additional 5 positions. Binding was reduced by >95% by alanine substitution at positions 10, 11, 12, 14, and 15 and was reduced by 70% and 50% by substitution at residues 13 and 16. This defines the epitope for the parental antibody as spanning residues 10 to 16 which corresponds to the 7 amino acid sequence VYPPEEE in which residues valine, tyrosine, proline and glutamic acid (underlined) are critical. While overlapping with the “core” sequence recognized by the affinity matured Fabs (VRTVYPPE) the parental Fab appears to require the additional residues EE at positions 15 and 16. The finding that substitution of additional residues interferes with binding of the parental antibody is likely related to the lower affinity of this antibody such that binding can be prevented more readily.
In summary the core epitope of the affinity matured antibodies is the sequence VRTVYPPE within which the tyrosine (Y) and the glutamic acid (E) are essential and thus are likely critical contact points for the antibody.
Habib et al (Analytical Biochemistry 2013, vol 438 p 82-89) to mapped the epitope of the anti-GPA nanobody IH4 and the results are summarized in Table 23 in comparison to the epitope mapping data for the affinity matured 6C12 Fabs on the same GPA peptide sequences. While the IH4 nanobody bound well to the GP9, GP11 and GP12 peptides, the 6C12 Fabs did not bind to any of these peptides. The strongest binding of the IH4 nanobody was to the GP11 peptide that was not bound by 6C12. The peptide mapping data for the affinity matured 6C12 Fabs indicate that they require the sequence VRTV that is present at the N-terminus of GP8 and this VRTV sequence is not present in GP11 peptide that was the strongest binder for IH4. Therefore the epitopes for 6C12 Fabs and the IH4 Nanobody are distinct.
DNA encoding various fusions between B-domain deleted human FVIII and different scFV antibodies were constructed using standard molecular biology techniques which are well known in the art. A combination of synthetically synthesized DNA fragments, PCR, and restriction enzyme based cloning techniques was used. The native human FVIII signal peptide was used in all constructs, for example in constructs in which the scFv was fused at the N-terminus of FVIII, the scFv was inserted between the native FVIII signal peptide and the start of the mature FVIII protein. The resulting cDNA sequences encoding human FVIII-scFv fusion proteins were transferred to the mammalian expression vector pSS207 (pSS207 is described in Mei et. al, 2006 Molecular Biotechnology, 34:165-178). The mammalian cell line HKB11 was stably transfected with the expression vectors, and clones were selected that express the corresponding FVIII-scFv fusion proteins by measuring FVIII chromogenic activity in the cell culture supernatant. To produce FVIII scFv fusion proteins, stably transfected HKB11 cells were cultured at a 10 liter scale using Wave fermenters and condition media containing the secreted FVIII protein was harvested, concentrated between 3 and 5-fold then frozen at −80° C.
Example 8: Purification of Factor VIII-scFv Fusion ProteinsFVIII-scFv fusion proteins were purified using methods well known in the art. Fractions containing the peak of FVIII activity were pooled, sucrose was added at final concentration of 1% and protein concentration measured by soloVPE. Aliquots were flash frozen and stored at −80° C. Purity of purified FVIII fusion proteins was assessed by SDS PAGE, and analytical SEC. FVIII activity was assessed using the two stage Coatest FVIII chromogenic assay.
Example 9: Generation of Hemophilia A Mice Expressing Human Glycophorin A on their Red Blood CellsThe most widely used animal model to evaluate FVIII based therapeutics is the Hemophilia A mouse in which the mouse FVIII gene was disrupted resulting in the complete absence of mouse FVIII protein. These mice exhibit a hemophilic phenotype as typified by inability to form clots and rapid death after tail injury. Because the anti-GPA antibodies described herein do not bind to mouse GPA we created a transgenic mouse expressing the human GPA protein. To do this a bacterial artificial chromosome (BAC) containing the human glycophorin A gene genomic sequence and surrounding regulatory regions selected from among BAC ID's RP11-50O5 (size of 176 Kb with the human GYPA gene in the middle of the BAC). After removal of the cloning vector portion the BAC was used for pro-nuclei of fertilized oocytes from C57BL/6 mice. Founder mice that carried the BAC inserted in to their genome were identified by PCR analysis with primer pairs homologous to the 5′ end of the BAC, the 3′ end of the BAC and to exon 5 of the glycophorin A gene. Only mice that were positive for all 3 sets of PCR primers were selected. The founder mice were screened for expression of human GPA on the surface of their red blood cells by FACS analysis using the antibody BRIC256 that binds to human GPA. The Ter119 antibody that recognizes mouse GPA was used as a control. One founder animal was identified that expressed human GPA on its RBC at levels similar to that measured on normal human RBC. This animal was bred with non-transgenic C57BL/6 mice to generate F1 animals which were screened for the presence of the transgene by PCR. PCR positive F1 mice were screened for expression of human GPA on their RBC by FACS. Several F1 mice with human GPA expression were identified and one animal (M9) with human GPA levels of between 15 and 30% of that on normal human RBC (depending on whether the BRIC256 antibody or the 6C12 antibody was used for FACS analysis) was selected.
The M9 F1 mouse was then bred to the Hemophilia A mouse strain (also in C57BL/6 background) to generate Hemophilia A mice expressing human GPA on their RBC. Mice homozygous for the human glycophorin A transgene could not be generated and this was presumed to be because the presence of the transgene was homozygous lethal. This is likely to have been caused by integration of the BAC in to an essential gene, but the precise cause was not investigated. The hemizygous huGPA/HemA mice were evaluated for the expression of human GPA on the surface of their RBC by FACS using the 7792 anti-GPA scFV (PPB-6676) and compared to the signal on normal human RBC (
An ideal FVIII-scFv fusion protein will have the properties of rapid and tight binding to red blood cells while maintaining sufficient coagulation activity. While some of these properties can be evaluated in vitro it is also appreciated that given the complexity of the coagulation system and the natural clearance mechanisms for FVIII, evaluation in vivo will also be needed. A number of design variables were considered when designing FVIII-scFv fusions. These included (i) the specific anti-GPA scFv to be selected from among the affinity matured scFv described in example 4; (ii) the location within FVIII at which the scFv is fused to FVIII; (iii) the length and composition of the polypeptide linker to fuse the scFv to FVIII; (iv) the vWF binding capability of the molecule as determined by the presence or absence of the a3 domain; and (v) the chain composition of the fusion protein as defined by the presence or absence of the furin cleavage site. The first step in the design of FVIII-scFv fusion proteins with optimal properties was to define the specific scFv antibody that provides the longest duration of action in vivo. It was envisaged that differences in the binding kinetics between different affinity matured variants might result in different RBC binding characteristics when the scFv is fused to FVIII.
Based upon a combination of the binding kinetics on isolated human RBC membranes and the binding to intact Human and Cyno RBC, the scFv variants of antibody 6C12 called 7790, 7792 and 7793 were selected for fusion to FVIII. These scFv variants were also selected based on their degree of net positive charge. The parent scFv, 6C12, is net positively charged, and all of the affinity-matured scFv variants are even more net positively charged. GPA is net negatively charged. However, in order to minimize the risk of non-specific binding to undesired, negatively charged moieties, such as proteins other than GPA, cells other than RBCs, or to avoid intramolecular association with negatively charged regions of FVIII, scFv variants with minimally increased net positive charge were selected.
Variants 7790 and 7792 were among the best binders to intact Human and Cyno RBC and 7790 had amongst the lowest KD values of the different scFv variants. The binding kinetics of the scFv variant 7792 to human RBC membranes was not quantifiable but was observed to have low off rates that were too slow to be quantified by the Biacore instrument. The scFv 7793 exhibited a faster on rate than the 7790 variant but also a faster off-rate such that the overall affinity was less (KD 2.46E-11 as compared to an estimated KD of 1.56 E-15 for 7790). In addition the 7793 variant exhibited lower binding strength to intact human RBC but similar binding strength to intact Cyno RBC. Thus 7790 and 7792 were selected as examples of scFv with very tight binding including slow off-rates whereas 7793 was selected to evaluate if an scFv with a faster off-rate but a faster on-rate might be more beneficial in targeting FVIII to RBC in vivo. All 3 scFv were fused to B-domain deleted FVIII in place of the B-domain. The order of the VH and VL domains within the fusion protein was established as N-terminus-VL-linker-VH-C-terminus. This order of the heavy and light chains appeared to enable more effective purification of the FVIII-scFv fusion due to reduced levels of aggregation (data not shown).
In order to maintain correct folding of the VH and VL regions in the context of a scFv it has been well established in the art that a linker of not less than 15 amino acids composed of GGGGS repeats is required (Huston et al 1988, PNAS vol 85, 5879; Huston et al, Methods in Enzymology 1991, vol 203 p 46-88). The use of linkers composed of glycine and serine residues is well established in the art due to the structural flexibility of such sequences and the low risk of immunogenicity in patients established with protein drugs containing such linkers. The use of the repeating motif GGGGS is well known in the art as described in Huston et al, Methods in Enzymology 1991, vol 203 p 46-88. It is also well appreciated in the art that longer linkers for example of 20 amino acids or more may be beneficial in the case of certain scFv molecules to provide improved folding and stability (for example see Albrecht et al, J. of Immunological Methods 2006, vol 310 p 100-116). Preliminary experiments with FVIII fusions to the scFv format of antibody 6C12 that incorporated a 15 amino acid linker (GGGGSGGGGSGGGGS, SEQ ID NO: 106) between the VH and VL regions and the chain order VH-VL suggested that the protein exhibited unexpectedly high levels of aggregation during purification that was not seen previously using a similar FVIII fusion to a scFv derived from the mouse GPA specific antibody Ter119. When a longer 20 amino acid linker (GGGGSGGGGSGGGGSGGGGS, SEQ ID NO: 107) was used and the order of the heavy and light chains of the scFv was changed to VL-VH the level of protein aggregation was significantly reduced and the yields of purified FVIII-scFv fusion protein was increased. Based on these observations all subsequent FVIII-scFv fusion proteins incorporated the scFv with the domain order VL-VH (N-term to C-term) with the VL and VH domains separated by the 20 amino acid linker with the sequence; GGGGSGGGGSGGGGSGGGGS.
When fusing an RBC targeting moiety such as an scFv to a normal 2 chain form of FVIII-BDD in place of the B-domain, the scFv could be placed either on the N-terminal side of the furin site or on the C-terminal side of the furin site resulting in a mature 2 chain protein in which the scFv is fused either on the C-terminus of heavy chain (composed of A1-A2 domains of FVIII) or on the N-terminus of the light chain (composed of A3-C1-C2 domains of FVIII), respectively. In both cases the fusion protein can be designed to include to native FVIII thrombin sites such that the FVIII will be released from the scFv in the presence of thrombin or other related proteases of the coagulation system such as Factor Xa (FXa) or Factor XIa (FXIa). Such a fusion protein design provides for the release of an essentially normal FVIIIa molecule from the RBC surface upon activation of the coagulation cascade at the site of an injury. While not proven experimentally it is envisioned that the release of FVIIIa from the RBC bound scFv would provide for normal control of bleeding. However it is also possible that activated FVIII that remained bound to RBC would also be functional in promoting coagulation via complex formation with FIXa and Factor X. When fusing a RBC targeting moiety such as an scFv at the N-terminus or the C-terminus of either full length FVIII or B-domain deleted FVIII a linker sequence may be included between the targeting moiety and the FVIII. It is well appreciated in the art that linkers can be beneficial when fusing two different proteins or protein domains because they provide for structural flexibility to ensure correct folding, steric accessibility and function of the two proteins or protein domains. Typical linkers used in the art of protein engineering consist of flexible sequences of between 4 and 20 amino acids. Linkers are typically made up of combinations of glycine and serine residues as described in Huston et al, Methods in Enzymology 1991, vol 203 p 46-88. In order to enable the release of a FVIII-scFv fusion from the surface of RBC upon activation of coagulation a cleavage site for the proteases thrombin and/or FXa and/or FXIa can be included. Thrombin, FXa and FXIa are all generated during the activation of the coagulation cascade and thus are good candidates as proteases to enable the release of an RBC targeted FVIII protein. The sequences of the cleavage sites for thrombin, FXa and FXIa are well known in the art and the use of these protease cleavage sites within fusion proteins has been described (Jenny et al 2003, Protein Expression and Purification vol 31, p 1-11). Thrombin and FXa are serine proteases that cleave on the C-terminal side of a basic amino acid residue, typically arginine (R) referred to as the P1′ residue. Cleavage sites for thrombin and FXa and FXIa can be derived from proteins known to be cleaved by these enzymes, including FVIII itself. Alternatively a consensus cleavage site that is derived by comparing the cleavage sites within different proteins can be derived. Additionally, the substrate specificity of proteases has been studies in vitro to further define the recognition sequences (Backes et al 2000, Nature Biotechnology vol 18, p 187-193, Gallwitz et al 2012, PLoS ONE 7(2):e31756.doi:10.1371/journal.pone.0031756)
A commonly used thrombin cleavage site is LVPRGS (SEQ ID NO: 108) in which the R residue is the P1′ residue. The sequence IEGR taken from prothrombin has been commonly used as the cleavage site for FXa. It is well known in the art that various different sequences can function as cleavage sites for thrombin or FXa or FXIa. In theory any sequence that functions as a cleavage site for these proteases could be used. These cleavage sites are preferentially incorporated within a flexible glycine serine linker having at least 4 but up to 15 or more glycine or serine residues either side of the protease recognition sequence. The protease site may be located in the middle of the flexible linker, for example the sequence GGGSGGGGSGLVPRGSGGGSGGGGSG (SEQ ID NO: 109) which consists of 10 and 10 amino acid glycine serine flexible linker either site of the thrombin recognition sequence LVPRGS. Alternatively more than one protease sites may be located at the N-terminus of the linker for example the sequence GGEGRTATGGGSGLVPRGSGGGSGGGGSG (SEQ ID NO: 110) which consists of a FXa site (GEGRTAT, SEQ ID NO: 111) followed by a 5 residue flexible linker, a thrombin recognition sequence (LVPRGS) followed by a 10 residue flexible linker. Alternatively the protease site may be located at the C-terminal end of a flexible linker for example in the sequence GGGGSGGGGSGGLVPRGSGGGG (SEQ ID NO: 112) which contains a 12 residue flexible linker at the N-terminus followed by a thrombin recognition sequence and a short 4 residue flexible linker at the C-terminus. Depending on the location of the protease site within the linker and relative to the flanking protein domains in the fusion protein, different linker sequences will be left on the flanking sequences after cleavage. In the case of fusions of a scFv at the N-terminus of FVIII the linker GGGSGGGGSGLVPRGSGGGSGGGGSG will leave the sequence GSGGGSGGGGSG attached at the N-terminus of FVIII. In the case of fusions of a scFv at the C-terminus of FVIII the linker GGGSGGGGSGLVPRGSGGGSGGGGSG will leave the sequence GGGSGGGGSGLVP attached at the C-terminus of FVIII. Many different linker sequences could theoretically provide for correct folding of the fusion protein and efficient cleavage by the selected protease.
Using standard molecular biology techniques a series of FVIII-scFv fusion gene expression cassettes were constructed that encode proteins with the features as listed in Table 24. The proteins encoded by these constructs were expressed in HKB11 cells and purified.
Using standard molecular biology techniques a series of FVIII-scFv fusion gene expression cassettes were constructed that encode proteins with the features as listed in Table 24. The proteins encoded by these constructs were expressed in HKB11 cells and purified.
Example 11: In Vitro Partition Assay of RBC Targeted FVIII Molecules in Whole BloodWhole blood was drawn from huGPA/Hem A mice in to anticoagulant with 4% sodium citrate (Sigma, St Louis, Mo.). B-domain deleted FVIII (BDD) and FVIII-scFv fusion molecules were spiked into anticoagulated fresh whole blood from human GPA/Hem A mice at 1 IU/mL, and incubated at room temperature for 2-3 hours with gentle rotation. Plasma was prepared from an aliquot of blood by centrifugation at 1000 g for 3 min after incubation for 0, 5, 15, 30, 45, 60, 120, 180 min and the FVIII activity remaining in the plasma was determined using the chromogenic assay. By subtracting the amount of FVIII activity in the plasma from the amount of FVIII activity spiked in, the amount of the FVIII-scFv fusion bound to RBC was calculated.
To evaluate partition to Cynomolgus monkey or human red blood cells, whole blood was draw from Cynomolgus monkey or normal human volunteers in to anticoagulant with 4% sodium citrate (Sigma, St Louis, Mo.). The red blood cells were purified by layering the whole blood on to ficoll followed by low speed centrifugation. The supernatant containing plasma, platelets and white blood cells was discarded and the red blood cell pellet was resuspended in 10 pellet volumes of PBS (50 mM Phosphate buffer, pH7.4, 150 mM NaCl). The red blood cells were then pelleted by slow speed centrifugation and the red blood cells re-suspended in 10 volumes of PBS. This washing step was repeated 4 more times and the red blood cells finally resuspended in a volume of plasma from Hemophilia patients with less than 1% of normal levels of FVIII (GeorgeKing Bio-Medical, Overland Park, Kans.) to generate a total volume equal to the initial blood volume such that the RBC density was equal to that of normal whole blood. BDD and RBC targeted FVIII molecules were then spiked into RBC suspension in HemA plasma at 1 IU/mL, and incubated at room temperature for 2-3 hours with gentle rotation. After incubation for 0, 5, 15, 30, 45, 60, 120, 180 min a sample of the suspension was removed and centrifuged to pellet the RBC and the plasma was collected and frozen. An additional control was performed in which BDD or RBC targeted FVIII molecules were added at 1 IU/ml into the same human hemophilia A plasma and incubated for 5 minutes and 180 mins. These samples were used to determine the input FVIII activity in the presence of plasma which was used to calculate the percentage of the molecule that was bound to RBC. All plasma samples were then assayed together for FVIII activity using the Coatest SP assay (DiaPharma, Columbus, Ohio). The assay was performed following the manufacturer's instructions in a 96-well plate format. Briefly, 50 uL of WHO8 calibrator, plasma sample diluted 100-fold with 1× Coatest buffer and 50 uL of a mixture of activated FIX/FX/phospholipid were added to each well, followed by 25 uL of 25 mM CaCl2 and 50 uL of chromogenic substrate S-2765/1-2581. The plate was incubated at 37° C. for 10 minutes with shaking between each reagent addition. After the final addition of chromogenic substrate, the kinetics of FXa generated reflecting FVIII activity (Vmax, mIU/min) in each well were read at 405 nm at 37° C. for 10 min with 30 sec interval. The amount of FVIII activity in the plasma reflects the amount of unbound FVIII. The FVIII activity of the same molecule spiked in to Hemophilia A plasma alone (no RBC) and incubated for 5 minutes was normalized to 100%.
The data (see Table 25) demonstrated that the amount of BDD present in the plasma did not change significantly over time, indicating that BDD did not bind to RBC, as expected. However, increased binding of the scFv-FVIII fusion proteins over time was observed indicating that the anti-GPA scFv was mediating the binding of the scFv-FVIII fusions to the human RBC (Table 26). TPP-8798 in particular showed very fast association with human RBC when the experiment was performed in vWF deficient plasma. TPP-8798 consists of the 7792 scFv fused to the N-terminus of FVIII-BDD that contains the a3 domain and is primarily a 2 chain molecule. The observation that TPP-8798 exhibits no binding to human RBC in the presence of human Hem A plasma that contains vWF provides evidence that vWF interferes with binding of scFv-FVIII fusions to RBC. TPP-8741 and TPP-8798 differ only in the site at which the 7792 scFv is fused to FVIII with TPP-8741 having the scFv fused in place of the B-domain and TPP-8798 having the scFv fused at the N-terminus of FVIII. While both TPP-8741 and TPP-8798 achieved similar levels of RBC binding after 120 minutes, TPP-8798 bound to the RBC more rapidly with 76% bound after 1 minute compared to 35% for the B-domain fusion TPP-8741. These data demonstrated that a FVIII molecule in which the scFv is fused at the N-terminus of FVIII results in a molecule with more rapid binding to RBC, a characteristic that is preferred as it should enable more complete binding to RBC in vivo, especially in a situation where vWF or other plasma proteins may interfere with the binding to RBC. The protein TPP-9423 which consists of the 7792 scFv fused at the N-terminus of FVIII-BDD containing a deletion of the a3 domain and the furin site also exhibited rapid and complete binding to human RBC. TPP-9049 in which the 7792 scFv is fused at the C-terminus of FVIII-BDD containing the a3 domain and the furin site also exhibited slower binding to RBC than the corresponding N-terminal fusion with the same scFv (TPP-8798) suggesting that the C-terminus of FVIII is a less favorable location for fusion to a RBC targeting scFv.
Partitioning to human RBC in human hemophilia A plasma, which most closely mimics the situation in humans, is shown in Table 26. TPP-8741 that consists of the 7792 scFv fused in place of the B-domain in FVIII-BDD containing the a3 domain and furin site exhibited slow and incomplete binding to human RBC, reaching only 52% bound after 120 minutes. The slow binding of TPP-8741 is due to the presence of vWF in the Hemophilia A plasma. TPP-9424 that consists of the 7792 scFv fused at the N-terminus of FVIII-BDD containing a deletion of the a3 domain and the furin site exhibited very rapid and complete binding to human RBC in the presence of human hemophilia A plasma with 76% bound within 1 minute and reaching saturation binding of about 95% by 15 minutes. Interestingly, scFv-FVIII fusions containing 2 copies of the 7792 scFv exhibited slower binding to human RBC in hemophilia A plasma than did TPP-9424 that contains a single copy of the scFv. It might have been expected that including 2 copies of the scFv in one scFv-FVIII fusion might have increased the affinity for GPA via an avidity effect given that each scFv contains only a single antigen binding site. Surprisingly these data indicate that scFv-FVIII fusions containing 2 copies of the GPA targeting scFv exhibited slower binding than scFv-FVIII fusion containing one copy of the same scFv at the N-terminus. Among the three scFv-FVIII fusions containing 2 copies of the 7792 scFv, TPP-9976 that contains 2 tandem copies of the scFv in place of the B-domain of FVIII exhibited the fastest and most complete binding to human RBC. Interestingly, among FVIII-scFv containing two copies of the scFv, TPP-9976 also exhibited the longest persistence in vivo in the huGPA/HemA mice.
Specific scFv-FVIII fusion proteins also bound to RBC from Cynomolgus (Cyno) monkey in human HemA plasma as shown in Table 27. The fastest and most complete binding was observed for TPP-9424 indicating that this molecule was a good candidate for evaluation in the Cyno monkey. TPP-9711, TPP-9161 and TPP-9976 also exhibited good binding to Cyno RBC. It was apparent that scFv-Fusions that contained the a3 domain such as TPP-8741 and TPP-8798 and thus were capable of binding vWF with high affinity exhibited little or no binding to Cyno RBC in human Hem A plasma that contains vWF. These same proteins (TPP-8741 and TPP-8798) did bind to Cyno RBC in vWF deficient plasma (Table 28) achieving 70 to 80% bound after 30 minutes.
It can also be appreciated that the binding of a given scFv-FVIII fusion protein to human RBC was faster and more complete than to Cyno RBC in the same human HemA plasma matrix. This is likely due to a lower affinity of the anti-GPA scFv for Cyno GPA compared to human GPA. This is expected given that the sequence of GPA in the region of the epitope for the 6C12 antibody (monkey; GRTHYPPEE and human; VRTVYPPEE) are not identical.
Example 12: Kinetics of Binding of FVIII-scFv Fusion Proteins to Glycophorin a Peptide GP5The binding kinetics of FVIII-scFv fusions to glycophorin A may be evaluated using surface plasmon resonance (SPR), a technique well known in the art. Attempts to perform SPR analysis of FVIII-scFv fusion proteins binding to membrane preparations from human RBC were not successful for technical reasons. However, binding data could be obtained by using a peptide that was identified in Example 6 to contain the epitope for the anti-GPA antibody 6C12 used in the FVIII-scFv fusions. The peptide GP5 (EVSEASVRTVYPPEEETGER) was attached to the probe of the OCTET instrument (ForteBio/Pall Corp) which was dipped in to a solution of the purified FVII-scFv fusion protein and the binding kinetics were measured by the instrument and fitted to a 1:1 binding model. The results are summarized in Table 29 for selected FVIII-scFv fusion proteins. The results demonstrate that the affinities to the GPA peptide ranged from 13.3 to 1.3 nM with off-rates ranging from 2.7 E-3 to 6.6 E-4 and on-rates ranging from 3.5 E+5 to 1.27 E+6. The true affinity to an intact, correctly folded GPA protein in the RBC membrane is expected to be higher than the values obtained for binding to a 20 residue peptide. This was supported by the observation that the affinity of binding of the affinity matured anti-GPA 6C12 antibodies to RBC membranes was higher than for the GP5 peptide by about 10-fold
The majority of optimized FVIII-scFv fusion molecules will bind quickly to RBC after injection into the circulation such that low levels of FVIII activity remains in plasma. Therefore, traditional plasma based chromogenic and aPTT assays are not appropriate for monitoring RBC targeted FVIII levels. Therefore the ROTEM® whole blood clotting assay was used to assess the FVIII activity after spiking of RBC targeted FVIII molecules into HemA whole blood. To perform the assay, citrated whole blood from huGPA/HemA mice drawn via vena cava was mixed with equal IU of RBC targeted FVIII or rFVIII-BDD (based on the Coatest® chromogenic assay) at room temperature. Samples were decalcified by dispensing 300 μL treated blood with an automated pipette into ROTEM® cups with 20 μL Star-tem® (200 mM CaCl2, Munich, Germany) without exogenous activator (NATEM). Clotting was initiated immediately after the last pipetting, and blood clot formation was continuously monitored for 2 hours at 37° C. ROTEM® analysis parameters include clotting time (CT), clot formation time (CFT), and α-angle. The same assay was performed in normal human whole blood by using the anti-FVIII neutralizing monoclonal antibody BO2C11 to neutralize the endogenous human FVIII. After addition of 100 ug/ml of BO2C11 to normal human whole blood and analysis by ROTEM, no clot formation was observed (“Induced HemA” in
The pharmacodynamics of RBC targeted FVIII proteins were assessed in huGPA/HemA mice. As a comparator we used Eloctate which is a marketed FVIII based drug in which human FVIII is genetically fused to a human Fc domain in order to prolong the circulating half-life of the protein. Eloctate has a 2-fold longer T½ in mice than un-modified FVIII and protects HemA mice from bleeding for about twice as long as un-modified FVIII protein (Dumont et al, Blood 29, p 3024-3030, 2012). RBC targeted FVIII molecules or Eloctate were administered into HuGPA/HemA mice by tail vein injection at 200 IU/kg. At different times after injection ranging from 2 h to 16 days, 400 uL of whole blood were drawn via vena cava from 4 mice with 25G syringe filled with 40 uL of 4% sodium citrate. Clotting time was measured by ROTEM by immediately transferring 300 uL of the whole blood and 20 uL of Star-tem® (0.2 mol/lCaCl2) to the cuvette of the ROTEM® delta machine. The median clotting time of the 4 mice per time point was calculated. To correlate clotting time to the level of FVIII activity, different amounts of un-modified FVIII ranging from 0.6% to 60% FVIII activity (based on units from chromogenic assay and assuming I IU/ml of plasma equals 100% FVIII) were added to anti-coagulated whole blood from the huGPA/HemA mice and clotting time measured by ROTEM. These results indicated that 0.6% FVIII was equivalent to clotting time of about 1450 sec, 2% FVIII was equivalent to clotting time of about 1150 sec, 6% FVIII was equivalent to clotting time of about 800 sec, 60% FVIII was equivalent to clotting time of about 750 sec. Un-treated whole blood from the huGPA/HemA mice that contains no FVIII had a clot time in ROTEM of 2250 seconds or longer
Example 15: Affinity Maturation of the scFv Antibody Prolongs the Pharmacodynamics of a scFV-FVIII Fusion In VivoTo confirm that affinity maturation of the anti-GPA targeting antibody 6C12 translates to improved duration of action when fused to FVII, the molecules TPP-8277 and TPP-9711 were produced. Both TPP-8277 and TPP-9711 contain a variant of the 6C12 scFv inserted in place of the B-domain of FVIII from which the a3 domain and the furin site were deleted. TPP-8277 contains the humanized and germlined version of the 6C12 scFv (antibody TPP-5906) while TPP-9711 contains the affinity matured scFv variant 7792. An additional comparator molecule is TPP-9161 that contains the affinity matured 7792 scFv fused to FVIII in place of the B-domain and a deletion of the a3 domain but is a predominantly 2 chain molecule. At different time points after injection in to the huGPA/HemA mice the whole blood clotting time was measured by ROTEM as described in example 14 and the results are shown in
To identify the specific variant of the affinity matured 6C12 scFv antibody that when fused to FVIII provided the longest persistence in the circulation of mice, the fusion proteins TPP-8820, 8743 and 8744 that contain the 7790, 7792 and 7793 scFv variants, respectively fused in place of the B-domain of FVIII from which the a3 domain was deleted were injected into huGPA/HemA mice and their pharmacodynamics (PD) profiles were measured using ex vivo ROTEM as the read out and compared to that of Eloctate. As shown in
While it is known that binding of vWF to FVIII prolongs the circulation time of un-modified FVIII, it is possible that vWF may interfere with the ability of a RBC targeted FVIII molecule to bind to RBC. vWF is a large multimeric protein which might stearically interfere with the ability of an anti-GPA scFv fused to FVIII to bind to RBC. In order to determine if vWF binding to a RBC targeted FVIII has a positive or negative impact on the circulation time in vivo we compared molecules TPP-8741 and TPP-8743 both of which are composed of the 7792 scFv fused in place of the B-domain of FVIII. While TPP-8741 contains the complete FVIII sequences found in FVIII-BDD, TPP-8743 contains a deletion of residues 1652 to 1682 that comprise the a3 domain of FVIII that is known to contain the major vWF binding region of FVIII.
It can be appreciated that the location within FVIII at which the anti-GPA scFv is placed might impact the kinetics of binding to RBC. The location of the scFv within FVIII might impact the ability of the scFv to access and bind to its epitope on GPA in the correct conformation. Other factors such as the local net charge of the protein in the region of the scFv might also impact the ability of the scFv-FVIII fusion to interact with GPA on the highly negatively charged surface of an RBC. The molecules TPP-8741, 8798, and 9049 are composed of the optimal scFv 7792 fused to FVIII-BDD either in place of the B-domain (TPP-8741) or at the N-terminus of FVIII (TPP-8798) or at the C-terminus of FVIII (TPP-9049). For fusion at the N-terminus and the C-terminus a linker containing a consensus thrombin cleavage site (GGGSGGGGSGLVPRGSGGGSGGGGSG) was placed between the scFv and FVIII to enable removal of FVIII from the RBC surface upon local activation of the coagulation system. All three of these molecules contain the a3 domain and are thus capable of binding to vWF.
While endogenous FVIII proteins exist primarily as a 2 chain molecule derived from the proteolytic cleavage of the primary polypeptide at the furin site at residue 1648, a 1 chain FVIII molecule might be more stable in vivo. A 1 chain molecule might also be more homogenous and thus be less complex to manufacture. To generate predominantly 1 chain forms of scFv-FVIII fusion proteins we deleted the furin site (RHQR) located between residues and 1645 and 1648. Molecule TPP-9161 is composed on the 7792 scFv fused to FVIII in place of the B-domain in which the a3 domain was deleted but the furin site remains intact. Molecule TPP-9711 is identical to TPP-9161 except that the sequence SQNPPVLKRHQREIT from residues 1637 to 1651 that contains the furin site (RHQR) have been deleted. The scFv-FVIII fusions TPP-6195 and TPP-8277 both contain the humanized and germlined 6C12 scFv (so6C12) in place of the B-domain and both are deficient in vWF binding via deletion of the a3 domain in the case of TPP-8277 or by mutation of the residue Y1680 to F in the case of TPP-6195. Mutation of Y1680 to F is known to significantly reduce vWF binding (Leyte et al 1991, J. Biol Chem. Vol 266, p 740-746). While TPP-6195 contains the furin site and is primarily a 2 chain molecule, TPP-8277 lacks the furin site and is composed of about 90% 1 chain molecule.
Examples 15 to 19 demonstrated that the longest persistence in the huGPA/HemA mice was provided by (i) the 7792 affinity matured anti GPA scFv; (ii) fusion of the scFv at the N-terminus of FVIII; (iii) deletion of the a3 domain of FVIII to prevent or reduce vWF binding to FVIII. In addition a single chain form of the scFv-FVIII fusion protein may be preferred due to greater protein homogeneity and does not negatively impact the PD profile as compared to a predominantly 2 chain form of the protein.
Therefore the molecule TPP-9424 was constructed that is composed of the 7792 scFv fused at the N-terminus of a B domain deleted FVIII that also has the a3 domain deleted and the furin site deleted. A variant of this molecule called TPP-9423 was also generated that is identical to TPP-9424 except that it contains a point mutation F2196K within the FVIII portion of the protein. The F2196K mutation makes the FVIII resistant to neutralization by the anti-FVIII antibody BO2C11. This mutation was introduced specifically for the purpose of enabling the clotting activity of the molecule to be detected in the blood of normal animals such as non-human primates that contain normal levels of endogenous FVIII. Neutralization of the endogenous FVIII by addition of the BO2C11 antibody will render the blood hemophilic and thus enable the clotting activity of an introduced FVIII-F2196K variant protein to be measured.
The PD profiles of TPP-9424 and TPP-9423 were evaluated in the huGPA/HemA mice and the results are shown in
A scFv form of an antibody contains a single antigen binding domain composed of the VL and VH domains and thus is monomeric in terms of the binding to the epitope. Full length antibody formats such as IgG contain two antigen binding domains and as such the apparent binding strength to a surface bound epitope can be increased via an avidity effect in which both binding domains engage adjacent epitopes simulataneously. Thus it was envisaged that a scFv-FVIII fusion protein containing 2 copies of the anti-GPA scFv might have increased binding strength to GPA on the surface of RBC. Such increased binding strength would be expected to be beneficial in terms of increasing the speed of binding to RBC and prolonging the time that the scFv-FVIII fusion protein remains bound to the RBC surface. An avidity effect could be particularly important in reducing the off rate from the RBC surface which should translate to a longer persistence in circulation, the main goal of this invention. Therefore scFv-FVIII fusion proteins were constructed in which 2 copies of the 7792 scFv were fused to FVIII-BDD containing the deletion of the a3 domain and the furin site. The fusion sites in these molecules were; (i) TPP-9900; N-terminus and C-terminus, (ii) TPP-9901; B-domain and C-terminus, (iii) TPP-9976; 2 tandem scFv in place of the B-domain, and (iv) TPP-9977; N-terminus and B-domain. The proteins TPP-9900, 9901, and 9976 were purified and their PD profiles in huGPA/HemA mice were determined as shown in
Eight (8) week old male mice were randomized into different treatment groups by their body weight. Mice were dosed by tail vein injection at different times prior to the tail vein transection. Before the tail vein transection, mice were anesthetized (IP) with a cocktail containing 70 μg/kg of ketamine and 0.7 mg/kg of medetomidine. The tail was marked at the location where the diameter is about 2.7 mm. The anesthetic effect of medetomidine was reversed with 0.7 mg/kg of atipamezole by IP injection. The tail vein was transected with a scalpel blade where it was the marked. The tail was then submerged into 37° C. saline tube for 3 minutes to rinse away the blood from the cut. The mouse was then returned to a clean cage with white paper bedding placed on top of a 4×8 inch heating pad. The observation for morbidity was made hourly for the next 9 hours followed by a final observation at 24 hr post transection. The survival rate at 24 h after the TVT injury was plotted in
A non-human primate such as the Cynomolgus (Cyno) monkey represents a species more closely related to humans and as such is more likely to be predictive of patient responses to a FVIII molecule. In order to evaluate the persistence of scFV-FVIII fusion proteins in the Cyno monkey, one amino acid change, F2196K, was made in the FVIII portion of the protein. This single amino acid change blocks the neutralization of FVIII activity by the monoclonal antibody (mAb) BO2C11 while not impacting the coagulation activity of FVIII itself (Lin et al 2015, PLOS ONE, 10(1) e01 16577. doi:10.1371). The BO2C11 mAB is also able to neutralize monkey FVIII. To confirm that the effect of the F2196K mutation was translatable to scFv-FVIII fusions the protein TPP-9423 was tested in a whole blood clotting assay using the Hemochron Signature Elite instrument (Accriva Diagnostics). Whole blood was collected from Cyno monkey into 3.2% citrate at a 9:1 ratio (blood to citrate) and stored at 4° C. up to 48 h until needed. Anti-coagulated Cyno whole blood was incubated with BO2C11 mAB at different concentrations or left untreated then added to the Hemochron cuvette and clotting time was measured using the aPTT program setting. In the absence of BO2C11 the Cyno whole blood clotted rapidly with a clot time of about 80 seconds as would be expected given that the blood was taken from a normal Cyno with normal coagulation system including normal FVIII levels (
To evaluate the resistance of TPP-9161 (7792 scFv fused in place of the B-domain of FVIII-BDD with a3 domain deleted and F2196K) to inhibition by BO2C11, anti-coagulated Cyno whole blood was mixed with 100 ug/ml of BO2C11 and increasing concentrations of TPP-9161. The clotting time was then measured in the Hemochron device. There was a linear dose response to TPP-9161 (
Prior to dosing, a dose response curve for TPP-9423 (7792 scFv fused at the N-terminus of FVIII-BDD with a3 domain and furin sites deleted and F2196K) was generated in whole blood drawn from the four adult male Cyno monkeys used in the PD study using the same Hemochron assay. The assay was run in triplicate samples from two separate blood draws and the means from the 2 blood draws are plotted in
The same four Cyno monkeys used to generate the dose response curves to TPP-9423 were dosed with TPP-9423 at 200 IU/kg (units based on chromogenic assay) and whole blood samples were drawn in to 3.2% citrate (1:9 ratio of citrate to blood) at different time points. After adding 100 ug/ml of B02C11 to neutralize the endogenous monkey FVIII the clotting time of the whole blood was measured in the Hemochron using the aPTT program. Duplicate or triplicate assays were performed on each blood sample and the mean value was calculated. The results are summarized in
While TPP-9424/9423 exhibited desirable pharmacodynamics in both the huGPA/HemA mouse and in the Cyno monkey, this protein as expressed and purified from HKB11 cells is composed of a mixture of about 65% 1 chain molecule and 35% 2 chain molecule. While the percentage of 1 chain molecule is higher than is typically seen for a un-modified FVIII molecule, a characteristic attributable to the deletion of the furin site in TPP-9424/9423, it might be desirable for manufacturing purposes to further increase the percentage of the 1 chain form of the molecule to >70%, or >80%, or >90%, or >95%, or >98% or 100%. The normal cleavage of FVIII at the furin site (RHQR) that generates the heavy and light chains of FVIII is believed to be catalyzed by the protease Furin (also called PACE) or a similar protease. This cleavage occurs during the post-translational modification of FVIII in the golgi-ER system. Post-translational modifications are often cell type dependent and thus it is possible that the percentage of the 1 chain form of the scFv-FVIII fusion will vary dependent upon the cell type used for expression and even between stable clones generated in the same cell type. Thus it is possible that by using BHK21 or CHO cells, which are typically used in the manufacture of FVIII based drugs, that the percentage of 1 chain scFV-FVIII fusion will be altered from what was obtained in HKB11 cells used to produce the scFv-FVIII fusion proteins described herein.
The mechanism by which TPP-9424 and TPP-9423 are cleaved to generate the heavy and light chain despite the deletion of the furin site has not been fully elucidated. Analysis of the N-terminal sequences present in TPP-9423 identified both the expected N-terminus of the 1 chain form of the protein and the N-terminus of a second species with the sequence SFSQNDE (residues 741 to 747 in full length FVIII), indicating that the cleavage site was between residue R740 and 5741 in the sequence EPRSF (residues 738 to 742 of full length FVIII). Residue R740 is the naturally occurring thrombin sites between the A2 domain and the B domain in full length FVIII. Because serine proteases such as Furin and Thrombin cleave proteins at related sequences it is possible that in the absence of the normal furin cleavage site that furin is instead cutting at the R740 thrombin site which lies within 10 amino acids of where the furin site would have been within TPP-9423 and TPP-9424.
We have also observed that when the scFv is fused in place of the B-domain in the context of a FVIII protein from which the furin site was deleted, the molecule was produced as a greater than 90% 1 chain molecule, for example as in TPP-9711. This suggests that the presence of additional amino acid sequences between the R740 thrombin site and the where the furin site would normally be found in FVIII prevents furin from cleaving efficiently at R740.
Based on these observations two additional scFv-FVIII fusion molecules were designed in order to reduce cleavage by furin. The designs incorporate deletions of the R740 thrombin site and insertion of flexible, non-immunogenic linkers in place of the B-domain. While specific sequence designs are provide as examples it should be appreciated that alternative sequence designs that incorporate deletion of part or all of the R740 thrombin recognition sequence (IEPRSF) and/or different linker sequences that may be longer or shorter or incorporate amino acids other than G or S may serve the same purpose and are hereby included. The molecules TPP-10297 and TPP-10298 were designed. TPP-10297 contains a 20 residue flexible linker (bold text) between the heavy and light chains of FVIII resulting in the sequence: . . . SKNNAIEPRSFSQNGGGGGSGGSGGSGGSGGGGGDENQSPR . . . which spans residues 732 to 1689 of full length FVIII. TPP-10298 contains a deletion of the R740 thrombin site and a 6 residue flexible glycine linker (bold text) such that the sequence between the heavy and light chains of FVIII is: . . . SKNNAIEPGGGGGGDENQSPR . . . which spans residues 732 to 1689 of full length FVIII.
Both TPP-10297 and TPP-10298 would be expressed in HKB11 or another suitable mammalian cell type and the expressed scFv-FVIII fusion proteins purified and the chain composition of the proteins would be evaluated by SDS-PAGE analysis. In the event that either of these proteins was predominantly a 1 chain molecule, for example >70% or >80% or >90% or >95% or >98% 1 chain then the PD profile of these proteins would be evaluated in the huGPA/HemA mice as described in the earlier examples. Incorporation of the F2196K amino acid change in to TPP-10297 and TPP-10298 would enable their PD profiles to be evaluated in Cyno monkey.
Example 26: Additional Variants of TPP-9424 Designed to be Comprised of Predominantly a 2 Chain MoleculeWhile a predominantly 1 chain form of the scFv-FVIII fusion proteins such as TPP-9423 and 9424 described herein have extended durations of action in animals models, it is also appreciated that a predominantly 2 chain form is expected to also have a long duration of action in vivo. Example 19 demonstrates that the 1 chain form of one of the scFv-FVIII fusions did not provided a significant improvement in the duration of action in the huGPA/HemA mice. While a 2 chain form of the scFv-FVIII fusion protein might be less desirable for manufacturing due to potentially greater heterogeneity it is also appreciated that existing un-modified recombinant FVIII drugs are composed of mostly 2 chain FVIII with a variable amount of one chain FVIII, typically in the range of 5 to 20%. Two chain forms of all of the optimized scFv-FVIII fusion proteins described herein are expected to have desirable pharmacologic properties, in particular a significantly longer circulation time as compared to current marketed FVIII based drugs, and thus are explicitly included.
TPP-10299 represents an example of a variant of TPP-9424 in which the native FVIII furin site is retained in its normal location within the FVIII portion of the protein. This molecule contains the 7792 scFv fused at the N-terminus of FVIII-BDD via the same thrombin cleavable linker. In addition the a3 domain has been deleted to prevent or reduce vWF binding. At the junction between the heavy and light chains of FVIII (where the B-domain and a3 domain have been removed) the sequence is: . . . SKNNAIEPRSFSQNPPVLKRHQREITDENQSPR . . . (furin site underlined) which spans residues 732 to 1689 of full length FVIII.
TPP-10299 would be expressed in HKB11 or another suitable mammalian cell type and the expressed scFv-FVIII fusion proteins purified and the chain composition of the proteins would be evaluated by SDS-PAGE analysis. In the event that the protein was predominantly a 2 chain molecule, for example >70% or >80% or >90% or >95% or >98% 2 chain then the PD profile would be evaluated in the huGPA/HemA mice as described in the earlier examples. Incorporation of the F2196K amino acid change in to TPP-10299 would enable the PD profile to be evaluated in Cyno monkey.
Example 26: Additional SequencesProvided below are additional sequences:
While the present invention has been described with reference to the specific embodiments and examples, it should be understood that various modifications and changes may be made and equivalents may be substituted without departing from the true spirit and scope of the invention. The specification and examples are, accordingly, to be regarded in an illustrative rather then a restrictive sense. Furthermore, all articles, patent applications and patents referred to herein are incorporated herein by reference in their entireties.
Claims
1-21. (canceled)
22. A recombinant fusion protein comprising a protein wherein extension of circulating half-life would be beneficial to a patient, and at least one binding domain that specifically binds to a membrane protein on a red blood cell.
23. The recombinant fusion protein of claim 22, wherein the membrane protein is glycophorin A or Band3.
24. The recombinant fusion protein of claim 22, wherein the binding domain is an antibody, an antibody fragment, a scFv, a peptide, a peptide mimetic, or a small molecule.
25. The recombinant fusion protein of claim 24, wherein the binding domain is a scFv and the membrane protein is glycophorin A.
26. The recombinant fusion protein of claim 25, wherein the scFv comprises a heavy chain selected from the group consisting of SEQ IS NOS: 33, 35, 37, 39, 41, 43, 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65 and 67.
27. The recombinant fusion protein of claim 25, wherein the scFv comprises a light chain selected from the group consisting of SEQ IS NOS: 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64 and 66.
28. The recombinant fusion protein of claim 25, wherein the scFv comprises a heavy chain selected from the group consisting of SEQ IS NOS: 33, 35, 37, 39, 41, 43, 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65 and 67, and a light chain selected from the group consisting of SEQ IS NOS: 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64 and 66.
Type: Application
Filed: Jun 18, 2021
Publication Date: Mar 24, 2022
Inventors: Alan Brooks (Clayton, CA), Richard Feldman (El Cerrito, CA), Jian-Ming GU (Lafayette, CA), Shaun Lippow (San Carlos, CA), Shujun Yuan (Daly City, CA), Peter Bringmann (Concord, CA)
Application Number: 17/351,527
