Construction and Characterization of Additional Chimeric Proteins

Example 5

Additional constructs were generated which include: additional human PD-1-Fc-OX40L constructs as well as human hCD172a-Fc-OX40L, hPD1-Fc-TL1A, hBTLA-Fc-OX40L, hTMIGD2-Fc-OX40L, hTIM3-Fc-OX40L, mPD1-Fc-GITRL, mPD1-Fc-41BBL, mPD1-Fc-TL1A, mCD172a-Fc-CD40L, hTIGIT-Fc-OX40L and canine PD-1-Fc-OX40L. Each of these constructs was codon optimized for expression in Chinese Hamster Ovary (CHO) cells, transfected into CHO cells and individual clones were selected for high expression. High expressing clones were then used for small-scale manufacturing in stirred bioreactors in serum-free media and the relevant chimeric fusion proteins were purified with Protein A binding resin columns. FIG. 17A shows a Western blot characterization of various chimeric proteins including hCD172a-Fc-OX40L, hPD1-Fc-TL1A, hBTLA-Fc-OX40L, hTMIGD2-Fc-OX40L, hTIM3-Fc-OX40L, mPD1-Fc-GITRL, mPD1-Fc-41BBL, mPD1-Fc-TL1A, mCD172a-Fc-CD40L, hTIGIT-Fc-OX40L.

Binding assays were carried out to characterize the ability of the various human ECD-Fc-OX40L constructs to bind to hOX40. With respect to hX_ECO-Fc-OX40L, X refers to the ECD of each protein listed in the bracket on the left (with reference to FIG. 17B). FIG. 17B shows a schematic representation of the ELISA method used to detect binding of hX_ECD-Fc-OX40L to hOX40. Recombinant hOX40 fused to human Fc (hOX40-hFc) was used to capture hX_ECD-Fc-OX40L in the culture media. Because the hOX40 fusion protein used to capture the target fusion proteins also contains a hIgG region, blocking was performed using a non-HRP conjugated anti-hIgG prior to incubation with the culture supernatants containing the target fusion proteins. A rabbit polyclonal antibody to hIgG was used to detect the hIgG domain in the chimeric protein and subsequently detected using a horseradish peroxidase (HRP)-conjugated polyclonal antibody to rabbit IgG (H+L).

The binding of SL-279252 to cell surface expressed OX40 on Jurkat cells by flow cytometry was compared to two negative control proteins which are not expected to bind human OX40. These data demonstrate that SL-279252 efficiently binds human OX40 (left panel), while neither human PD1-Fc-TL1A or canine PD1-Fc-OX40L were observed to bind human OX40 (FIG. 17C).

The human CD172a-Fc-OX40L construct was imported into the protein tertiary prediction software RaptorX to determine the tertiary structure. The predicted tertiary structure is shown in FIG. 17D.

The codon-optimized DNA sequence of several chimeric fusion proteins were synthesized and directionally cloned into pVITRO2, pcDNA3.4 and other expression vectors. Vectors were then either transiently or stably transfected into CHO or 293 cells and individual clones were selected for high expression. For example, SL-279252 was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17E).

In another example, CD172a-Fc-OX40L was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17F). In another example, CD172a-Fc-CD40L was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by the Perkin Elmer LabChip system and quantitated as compared to a BCG standard (FIG. 17G). In another example, human TIGIT-Fc-OX40L was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17H).

The binding affinity of human CD172a-Fc-OX40L was evaluated by surface plasmon resonance (SPR) analysis to hCD47, hOX40 and various human Fc receptors (FIG. 17I to FIG. 17M). Specifically, polyhistidine-tagged versions of recombinant human CD47 and human OX40 was bound to ProteOn HTG tris-NTA chips (BIORAD). CD172a-Fc-OX40L was then flowed over the bound ligands over a time course and a relative index of ‘on-rate’ (Ka) and ‘off-rate’ (Kd) was generated to calculate binding affinity (K_D) of CD172a-Fc-OX40L to each partner. Recombinant human CD47-Fc and OX40L-Fc were used as positive controls for binding. These controls have a relatively fast ‘on-rate’ and an equally fast ‘off-rate’, resulting in low nanomolar binding affinities. Consistent with these results, the ‘on-rate’ of CD172a-Fc-OX40L to human CD47 was rapid, however the ‘off-rate’ was much lower, in fact ˜40-fold slower than the ‘off-rate’ of recombinant CD47-Fc, indicating that CD172a-Fc-OX40L bound quickly and stably, with long on-target residence time (FIG. 17I). The K_D of CD172a-Fc-OX40L binding to human CD47 was calculated to be 3.59 nM. CD172a-Fc-OX40L bound with high affinity to human OX40 (869 pM), again with a fast ‘on-rate’ and slow ‘off-rate’ (FIG. 17J).

To further define the molecular characteristics of CD172a-Fc-OX40L, SPR was performed, analyzing the binding affinities of CD172a-Fc-OX40L to chip-bound, Fcγ receptors FcγR1A and to the neonatal receptor, FcRn. The human immunoglobulin IgG1 was shown to bind with the highest affinities to FcγR1A, followed by FcRn, in addition to low-level binding to FcγR2b (FIG. 17K and FIG. 17L). CD172a-Fc-OX40L did not bind to FcγR1A or FcγR2B, but did bind to FcRn at 790 nM affinity (FIG. 17L). Without wishing to be bound by theory, this binding characteristic may be important to the fusion protein because FcRn is involved in IgG recycling to the surface of a cell, thereby avoiding lysosomal degradation, and potentially extending the in vivo half-life of CD172a-Fc-OX40L. Summary data for CD172a-Fc-OX40L binding affinities are including (FIG. 17M).

The codon-optimized DNA sequence of several additional chimeric fusion proteins were synthesized and directionally cloned into pVITRO2, pcDNA3.4 and other expression vectors. Vectors were then either transiently or stably transfected into CHO or 293 cells and individual clones were selected for high expression. For example, canine PD1-Fc-OX40L was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17N). In another example, mouse PD1-Fc-OX40L was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17O). In another example, mouse PD1-Fc-GITRL was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17P). In another example, mouse PD1-Fc-41BBL was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17Q). In another example, mouse PD1-Fc-TL1A was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17R). In yet another example, CD115-Fc-CD40L was produced from a transient transfection from 293 cells, purified by affinity chromatography to Protein A columns and evaluated by Coomassie staining, Western blot and quantitated as compared to a BCG standard (FIG. 17S).

Each purified protein is characterized by ELISA assays to bind to the marker, e.g. the intended inhibitory ligand as well as the intended costimulatory receptor. For example, to test the binding of purified human PD-1-Fc-OX40L, recombinant PD-L1-Fc is adsorbed to microtiter plates and used to capture PD-1-Fc-OX40L. Any bound PD-1-Fc-OX40L is then detected by using recombinant human OX40-Fc linked to biotin, which is then detected in a chromogenic assay through binding with streptravidin-HRP.

In addition, each purified protein has been characterized by flow cytometry to bind the intended inhibitory ligand as well as the intended costimulatory receptor. For example, human tumor cell lines are characterized for endogenous expression of PD-L1, which was found to be particularly abundant on several human melanoma tumor cell lines. These same tumor cell lines were shown to be negative for human OX40L. Following incubation with PD-1-Fc-OX40L, any bound chimeric fusion protein is detected with human OX40L specific antibodies. Similarly, human Jurkat cells were transfected with human OX40 and shown to be negative for human PD-L1. Following incubation with the chimeric PD-1-Fc-OX40L constructs, any bound complex is detected using anti-human PD-L1 specific antibodies. A series of screening cell lines were generated in order to detect specific cell surface binding of each chimeric fusion protein to its respective receptor/ligand, these included: CHO-K1-CD47, CHO-K1-PD-L1, CHO-K1-HVEM, CHO-K1-HHLA2, CHO-K1-VISTA, CHO-K1-Gal9, HeLa-PD-L1, HeLa-CD47, HeLa-HVEM, HeLa-HHLA2, HeLa-VISTA, HeLa-Gal9.

To determine the functional activity of each receptor, in vitro T cell proliferation assays are performed in the presence of inhibitory ligand positive human tumor cells. For example, human melanoma tumor cells expressing PD-L1 are pulsed with peptides specific for hen egg lysozyme (HEL) and incubated with human HEL specific T cells expressing OX40 receptor. The proliferation of these cells is monitored in the presence and absence of the PD-1-Fc-OX40L construct and found to be functionally responsive to the presence of the chimeric constructs. In a similar system, human tumors expressing HVEM, CD47, galectin-9, TIGIT receptors or TMIGD2 receptors are used.

In some experiments, mouse PD-1-Fc-OX40L or mouse PD-1-Fc-TL1A are used to treat murine tumors known to be positive for murine PD-L1 (including B16-F10 melanoma, MC38 colon carcinoma and CT26 colon carcinoma). In these systems, established tumors are treated with purified chimeric fusion proteins as compared to PD-1-Fc fusion proteins, anti-PD-1 or anti-PD-L1 monoclonal antibodies or anti-OX40 or anti-GITR monoclonal antibodies. In these experiments, the activity of the chimeric constructs is observed to lead to enhanced antigen-specific T cell responses and increased rates of tumor rejection as compared to the individual therapeutics. In some experiments, nucleic acid constructs encoding PD-1-Fc-OX40L or PD-1-Fc-TL1A are directly electroporated into established tumors. In these experiments, the chimeric constructs are shown to lead to increased rates of tumor rejection as well as increased tumor antigen specific CD8+ T cell proliferation detected both in the peripheral blood and within established tumors.

To determine the binding of purified chimeric fusion proteins to human tumor explants, fresh frozen human tumor samples are obtained and incubated with each chimeric fusion protein. Any bound fusion protein is detected with anti-human OX40L and controlled against background staining by separate staining with anti-human OX40L.

To determine the molecular characteristics of each fusion protein, purified chimeric fusion proteins are characterized by size exclusion chromatography. This analysis is important because, for example, the OX40L ECD is known to form a homo-trimer, while the Fc region is known to form a homo-dimer, while the inhibitory ligand binding receptor may either be monomeric (e.g. PD-1) or form homo-multimers (e.g. TIM3). Thus, there are several possibilities for the individual species that may be formed by these chimeric constructs. Further molecular characterization by mass spec, thermal stability, pH stability, physical stability, charge profile, hydrophobicity, physical stability, buffer compatibility and solubility up to 100 mg/mL are also performed.