Patent Application
20230279117
PD-L1 was initially cloned as a member (known as B7-H1) of B7 protein family (Dong et al., 1999 Nature Med 5:1365). It binds to Programmed Death-1 (PD-1) receptor and activates negative regulatory signaling pathway, inhibiting T-cell proliferation and activities (Freeman et. al. 2000 J Exp Med 192:1027). Therefore, it was also termed as PD-1 ligand 1 (PD-L1 or CD274). To date, two sequence-related ligands, PD-L1 (B7-H1) and PD-L2 (B7-DC), were identified that interact with PD-1, induce negative signal transduction and inhibit TCR and CD28 mediated T-cell activation, cell proliferation and secretion of growth factors and cytokines such as IL-2 and IFN-y (Riley et. al. 2009 Immunol Rev 229:114).
Human PD-L1 gene encodes a full-length protein of 290 amino acid residues (NCBI accession NP_054862.1) with a leader peptide, which is removed after PD-L1 is expressed on cell surface as a mature protein. The calculated molecular weight of the full length PD-L1 is 33 kD. However, the observed molecular weight is around 50 kD due to glycosylation, based on Western blot data from ours and others.
PD-L1 was found constitutively expressed in human heart, lung, thymus and vascular endothelial cells, and expressed at a low level in many other human tissues and cell types including antigen presenting cells, peripheral blood monocytes and other immune cells (Freeman et. al. 2000 J Exp Med 192:1027; Eppihimer et. al. 2002 Microcirculation 9:133). When stimulated by IFN-y, IL-12 and type I interferons, many of those cell types were found expressing increased level of PD-L1 (Bald et. al. 2014 Cancer Discov 4:674-687; Planes et. al. 2014 J Virol 88:6672-6689).
Aberrant up-regulation of PD-L1 expression in tumor cells were reported in varieties of cancers involved in different types of tissues and organs such as lung (Konishi et. al. 2004 Clin Cancer Res 10:5094), liver (Shi et. al. 2008 Int J Cancer 128:887; Gao et. al., 2009 Clin Cancer Res 15:971), stomach (Wu et. al. 2006 Acta Histochem 108:19), kidney (Thompson et. al. 2004 Proc Natl Acad Sci 101:17174; Thompson et. al. 2007 Clin Cancer Res 13:1757), breast (Ghebeh et. al. 2006 Neoplasia 8:190), ovary (Hamanishi et. al. 2007 Proc Natl Acad Sci 104:3360), pancreas (Nomi et. al. 2007 Clin Cancer Res 13:2151), melanocytes (Hino et. al. 2010 Cancer 116:1757) and esophagus (Ohigashi et. al. 2005 Clin Cancer Res 11:2947). More frequently, the increased expression of PD-L1 in those cancers is associated to poor prognosis in patient survival outcome.
Blockade of PD-L1 engaging PD-1 receptor by B7-H1Ig or anti-PD-L1 antibody stimulated T-cell proliferation and functional activities (Dong et. al. 1999 Nature Med 5:1365; Freeman et. al. 2000 J Exp Med 192:1027; Tamura et. al. 2001 Blood 97:1809; Iwai et. al. 2002 PNAS 99:12293), enhanced immune responses against tumor growth and viral infection (Iwai et. al. 2002 PNAS 99:12293). Those observations suggested that inhibition of PD-L1/PD-1 signaling may activate immune responses not only against cancer cell growth, but also against viral infection and expansion in human. The prevalent hepatocyte infection viruses, HBV and HCV, induce overexpression of PD-1 ligands in hepatocytes and activate PD-1 signaling in T-effecter cells, resulting T-cell exhaustion and tolerance to the viral infection (Boni et. al. 2007 J Virol 81:4215; Golden-Mason et. al. 2008 J Immunol 180;3637). Likewise, the popular HIV infection frequently evades human immune system by similar mechanism. Therapeutic modulation of PD-L1 induced signaling by antagonist molecules may revert immune cells from tolerance, and reactivated to eradicate cancer and chronic viral infection (Blank et. al. 2005 Cancer Immunol Immunother 54:307; Okazaki et. al. 2007 Int Immunol 19:813).
Recently, it is discovered that PD-L1 also specifically interacts to B7-1 (another B7 family member, also known as CD80) besides binding to PD-1 (Butte et. al. 2007 Immunity 27:111). Initial evidences indicated that interaction of PD-L1 to CD80 exerts negative regulation to T-cell function and activity, and blockage of PD-L1 and CD80 interaction in mice elicited stronger immune responses to OVA antigen challenge (Park et. al. 2010 Blood 116:1291). Therefore, simultaneously blocking PD-L1 binding to PD-1 and CD80 may exert additive or synergistic effect against cancer and viral infection.
The invention provides methods and compositions for immune-activation by inhibiting PD-L1-mediated signaling and function. In one aspect, the invention provides an antibody antigen binding domain which specifically binds human PD-L1, and comprises a complementarity determining region (CDR) sequence described herein. The CDRs are amenable to recombination into heavy chain variable region (Vh) and light chain variable regions (Vk) which comprise (CDR-H1, CDR-H2 and CDR-H3) and (CDR-L1, CDR-L2 and CDR-L3) sequences, respectively and retain PD-L1-specific binding and/or functionality.
In particular embodiments, the domain comprises CDR1, CDR2 and CDR3, in a combination selected from (a)-(r) as follows, wherein the antibody (Ab), heavy chain (HC) or light chain (LC) and CDR nomenclature system (Kabat, IMGT or composite) from which the CDR combinations derive are shown in the first column, and residues in bold text are Kabat system, and residues underlined are IMGT system:
GFSLTSYG
VH
VIWAGGSTNYNSALMS
AKPYGNSAMDY
GFSLTSYG
VH
VIWAGGSTNYVDSVKG
AKPYGNSAMDY
GFSLTSYG
VH
VIWAGGSTNYADSVKG
AKPYGNSAMDY
GFSLTSYG
VH
VIWAGGSTNYVDSVKG
AKPYGTSAMDY
GFSLTSYG
VH
VIWAGGSTNYADSVKG
AKPYGTSAMDY
KASQDVGIVVA
WASIRHT
QQYSNYPLYT
In particular embodiments, the domain comprises a heavy chain variable region (Vh) comprising a CDR1, CDR2 and CDR3 combination selected from (a)-(o), and a light chain variable region (Vk) comprising a CDR1, CDR2 and CDR3 combination selected from (p)-(r).
In particular embodiments, the domain comprises CDR1, CDR2 and CDR3, in a combination selected from (c), (f), (i), (1), (o) and (r), as follows:
In particular embodiments, the domain comprises a heavy chain variable region (Vh) or a light chain variable region (Vk), comprising a sequence that is:
In particular embodiments, the domain comprises a heavy chain variable region (Vh) and a light chain variable region (Vk) comprising a sequence that is:
In particular embodiments, the domain comprises comprising a heavy chain variable region (Vh) or a light chain variable region (Vk) comprising:
In particular embodiments, the domain comprises comprising a heavy chain variable region (Vh) and a light chain variable region (Vk) comprising:
In particular embodiments, the domain specifically binds PD-L1 residues: D26 and R113.
The invention also provides antibodies, particularly monoclonal antibodies, and F(ab) or F(ab)2 comprising a subject PD-L1 binding domain.
The invention also provides novel polynucleotides such as cDNAs and expression vectors, encoding a subject PD-L1 antigen binding domain, and cells comprising such polynucleotides, and non-human animals comprising such cells. The polynucleotides may be operably linked to a heterologous transcription regulating sequence for expression, and may be incorporated into such vectors, cells, etc.
The invention provides methods of using the subject domains by administering the domain to a person determined to have cancer or a viral infection or to otherwise be in need of PD-L1 antagonism.
The compositions of the invention are useful for the treatment of cancer, neurodegenerative and infectious, particularly viral, diseases and other conditions in which inappropriate or detrimental expression of the human PD-1 and/or is a component of the etiology or pathology of the condition. Hence, the invention provides methods for treating cancer or inhibiting tumor progression in a subject in need thereof with a subject anti-PD-L1 protein, and the humanized anti-PD-1 mAbs are used as therapeutic agents to treat human diseases that are involved in suppression of immune cells by PD-1 mediated intracellular signaling, leading to disease progression, particularly cancers and viral infections.
The invention further provides the use of subject polynucleotides for the manufacture of a medicament for treating cancer or inhibiting tumor progression in a subject.
The invention includes all combinations of the recited particular embodiments. Further embodiments and the full scope of applicability of the invention will become apparent from the detailed description given hereinafter. However, it should be understood that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from this detailed description. All publications, patents, and patent applications cited herein, including citations therein, are hereby incorporated by reference in their entirety for all purposes.
PD-L1 initiates inhibitory signaling in immune cells when engaged by its ligands, PD-L1 or PD-L2. In the cases of cancer outgrowth and viral infection, the activation of PD-1 signaling promotes immune tolerance, leading to the cancers or virus-infected cells escaping from immune surveillance and cancer metastasis or viral load increase. Inhibition of PD-L1 mediated cellular signaling by therapeutic agents can activate immune cells including T-cells, B-cells and NK cells, and therefore enhance immune cell functions inhibiting cancer cell growth or viral infection, and restore immune surveillance and immune memory function to treat such human diseases.
The invention provides antibodies whose functions are antagonistic to PD-L1-induced cellular signaling in immune cells. Murine anti-PD-L1 antibodies were humanized to a high degree of similarity to human antibodies in the framework regions. The full antibodies made in the modified human IgG variant format have a unique set of features in the aspects of effector functions and physicochemical properties. The disclosed anti-PD-L1 antibodies are suitable for therapeutic uses in cancer treatment, controlling viral infections and other human diseases that are mechanistically involved in exacerbated immune tolerance.
Unless the context indicates otherwise, the term “antibody” is used in the broadest sense and specifically covers antibodies (including full length monoclonal antibodies) and antibody fragments so long as they recognize PD-L1. An antibody molecule is usually monospecific, but may also be described as idiospecific, heterospecific, or polyspecific. Antibody molecules bind by means of specific binding sites to specific antigenic determinants or epitopes on antigens. “Antibody fragments” comprise a portion of a full length antibody, generally the antigen binding or variable region thereof. Examples of antibody fragments include Fab, Fab′, F(ab′).sub.2, and Fv fragments; diabodies; linear antibodies; single-chain antibody molecules; and multispecific antibodies formed from antibody fragments.
Natural and engineered antibody structures are well known in the art, e.g. Strohl et al., Therapeutic antibody engineering: Current and future advances driving the strongest growth area in the pharmaceutical industry, Woodhead Publishing Series in Biomedicine No. 11, October 2012; Holliger et al. Nature Biotechnol 23, 1126 - 1136 (2005); Chames et al. Br J Pharmacol. 2009 May; 157(2): 220-233.
Monoclonal antibodies (MAbs) may be obtained by methods known to those skilled in the art. See, for example Kohler et al (1975); U.S. Pat. No. 4,376,110; Ausubel et al (1987-1999); Harlow et al (1988); and Colligan et al (1993). The mAbs of the invention may be of any immunoglobulin class including IgG, IgM, IgE, IgA, and any subclass thereof. A hybridoma producing a mAb may be cultivated in vitro or in vivo. High titers of mAbs can be obtained in in vivo production where cells from the individual hybridomas are injected intraperitoneally into mice, such as pristine-primed Balb/c mice to produce ascites fluid containing high concentrations of the desired mAbs. MAbs of isotype IgM or IgG may be purified from such ascites fluids, or from culture supernatants, using column chromatography methods well known to those of skill in the art.
An “isolated polynucleotide” refers to a polynucleotide segment or fragment which has been separated from sequences which flank it in a naturally occurring state, e.g., a DNA fragment which has been removed from the sequences which are normally adjacent to the fragment, e.g., the sequences adjacent to the fragment in a genome in which it naturally occurs. The term therefore includes, for example, a recombinant DNA which is incorporated into a vector, into an autonomously replicating plasmid or virus, or into the genomic DNA of a prokaryote or eukaryote, or which exists as a separate molecule (e.g., as a cDNA or a genomic or cDNA fragment produced by PCR or restriction enzyme digestion) independent of other sequences. It also includes a recombinant DNA, which is part of a hybrid gene encoding additional polypeptide sequence.
A “construct” means any recombinant polynucleotide molecule such as a plasmid, cosmid, virus, autonomously replicating polynucleotide molecule, phage, or linear or circular single-stranded or double-stranded DNA or RNA polynucleotide molecule, derived from any source, capable of genomic integration or autonomous replication, comprising a polynucleotide molecule where one or more polynucleotide molecule has been linked in a functionally operative manner, i.e. operably linked. A recombinant construct will typically comprise the polynucleotides of the invention operably linked to transcriptional initiation regulatory sequences that will direct the transcription of the polynucleotide in the intended host cell. Both heterologous and non-heterologous (i.e., endogenous) promoters can be employed to direct expression of the nucleic acids of the invention.
A “vector” refers any recombinant polynucleotide construct that may be used for the purpose of transformation, i.e. the introduction of heterologous DNA into a host cell. One type of vector is a “plasmid”, which refers to a circular double stranded DNA loop into which additional DNA segments can be ligated. Another type of vector is a viral vector, wherein additional DNA segments can be ligated into the viral genome. Certain vectors are capable of autonomous replication in a host cell into which they are introduced (e.g., bacterial vectors having a bacterial origin of replication and episomal mammalian vectors). Other vectors (e.g., non-episomal mammalian vectors) are integrated into the genome of a host cell upon introduction into the host cell, and thereby are replicated along with the host genome. Moreover, certain vectors are capable of directing the expression of genes to which they are operatively linked. Such vectors are referred to herein as “expression vectors”.
An “expression vector” as used herein refers to a nucleic acid molecule capable of replication and expressing a gene of interest when transformed, transfected or transduced into a host cell. The expression vectors comprise one or more phenotypic selectable markers and an origin of replication to ensure maintenance of the vector and to, if desired, provide amplification within the host. The expression vector further comprises a promoter to drive the expression of the polypeptide within the cells. Suitable expression vectors may be plasmids derived, for example, from pBR322 or various pUC plasmids, which are commercially available. Other expression vectors may be derived from bacteriophage, phagemid, or cosmid expression vectors.
Murine anti-human PD-L1 monoclonal antibodies (mAbs) were generated using hybridoma fusion technology (Kohler and Milstein 1975 Nature 256:495-497; Mechetner 2007 Methods Mol Biol 378:1-13) with modifications. MAbs with high binding activities in enzyme-linked immunosorbent assay (ELISA) and fluorescence-activated cell sorting (FACS) assay were selected for further characterization in cell-based functional assays.
The full length human PD-L1 cDNA was synthesized by GeneScript (Nanjing, China) based on published sequence (NCBI reference sequence NM_014143.3) (SEQ. NO. 1 and 2). The extracellular domain consisting of amino acids (AA) 1-239 of human PD-L1 (SEQ. NO. 3 and 4) was PCR-amplified and subcloned into pcDNA3.1-based expression vector (Invitrogen, Carlsbad, CA, USA) with C-terminus fused with either a Fc region of human IgG4 or a His tag, which resulted in two recombinant PD-L1 fusion constructs, PD-L1-ECD/Fc and PD-L1-ECD/His (abbreviated as PD-L1/Fc and PD-L1/His). The schematic diagram of the PD-L1 fusion proteins were shown in
Expression plasmid containing full-length human PD-1 cDNA was obtained from Origene (Cat. No. SC117011, NCBI Accession No: NM_005018.1, Beijing, China). The extracellular domain consisting of amino acid (AA) 1-168 of PD-1 was PCR-amplified, and subcloned in pcDNA3.1-based expression vector (Invitrogen, Carlsbad, CA, USA) with C-terminus fused to the Fc domain of human IgG4 heavy chain, abbreviated as PD-1/Fc.
The human CD80 (B7-1) cDNA was synthesized by GeneScript according to the published sequence (NCBI access number NM_005191.3). The extracellular domain (AA1-242) of CD80 was fused with human Fc at C-terminus and subcloned in a mammalian expression vector similar to the method described previously (Pat. US 8735553). The fusion protein was named as CD80-ECD/Fc or CD80/Fc.
Stable cell line expressing human PD-L1 was established by transfection of pcDNA3.1 plasmid containing PD-L1 into HEK293 (ATCC, Manassas, VA, USA), and followed by selection with media containing 600 micrograms of hygromycin (Cat. No. 10687-010, Invitrogen) per milliliter. Single clones were isolated by picking up single colonies from culture-dish surface. All clones were screened by FACS analysis and Western blot using PD-L1 antibody (Cat. No. 17-5983, eBioscience, San Diego, USA), and the top expression clones were used for FACS binding analyses and functional assays.
Murine anti-human PD-L1 monoclonal antibodies were generated using the hybridoma fusion technology. All animal protocols were reviewed by and performed following BeiGene Animal Care and Use Procedure. Ten to twelve week-old Balb/c mice (HFK Bioscience, Beijing, China) were immunized three times (3 weeks apart between injections) subcutaneously and/or intra-peritoneally, each immunization was done with 100 uL of adjuvant (Cat. No. KX0210041, KangBiQuan, Beijing, China) containing 5-10 microgram of PD-L1/Fc. Two weeks after the 2nd immunization, the mice sera were collected and evaluated for PD-L 1 binding by ELISA and FACS. An example of such assay results were shown in Table 1 and Table 2. The mice with high anti-PD-L1 binding titers in sera were selected and boosted intraperitoneally with 50 micrograms of PD-L1/Fc in PBS. Three days after boosting, the splenocytes were isolated and fused with the murine myeloma cell line, SP2/0 (ATCC), using standard techniques (Mechetner et. al. 2007 Methods Mol Biol 378:1-13) with some modification.
The supernatants of hybridoma clones were initially screened for PD-L1 binding activities by a modified ELISA assay (Flanagan 2007 Methods Mol Biol 378:33-52). Briefly, 50-200 nanograms of PD-L1/His were diluted in 50 microliters of PBS and coated in each well of 96-well ELISA plates (JinCanHua, Shenzhen, China). After blocking with 3% bovine serum albumin in TBST (20 mM Tris, 150 mM NaCl, 0.05% Tween20, pH7.5) and incubating with culture supernatants of hybridoma clones, the HRP-conjugated horse anti-mouse IgG antibody (Cat. No. 7076S, Cell Signaling Technology) and tetramethylbenzidine (TMB) (Cat. No. PA107-01, TianGen, Beijing, China) were used to detect binding signals by a plate reader (PHREAstar, BMG Labtech, Germany) as light absorbance at 450 nm. The ELISA-positive clones were further verified by fluorescence-activated cell sorting (FACS). PD-L1 expression cells, HEK293/PD-L1 (105 cells/well), were incubated with supernatants from hybridoma clones in V-bottom 96-well plates (Cat. No. 3897, Corning). The cell surface bound PD-L1 antibodies were detected with Dylight 649-conjugated goat anti-mouse IgG antibody (Cat. No. 405312, Biolegend, San Diego, CA, USA) and cell fluorescence was monitored in a Guava EasyCyte 8HT flow cytometer (Millipore, USA).
The hybridoma clones that were positive in both ELISA and FACS assays were then tested in human immune cell-based functional assays to identify antibodies with good functional activities. The hybridoma clones with positive functional activities were further subcloned and characterized.
The positive hybridoma clones from primary screening through ELISA, FACS and functional assays were subcloned by limiting dilution. Three subclones from each original clone were selected and confirmed in FACS and functional assays. The subclones selected through functional assays were defined as monoclonal antibody. The top subclones were adapted to grow in the CDM4MAB medium (Cat. No. SH30801.02, Hyclone) with 0-3% fetal bovine serum for purification and further characterizations.
Hybridoma cells or 293-F cells transiently transfected with an antibody expression plasmid (Cat. No. R79007, Invitrogen) was cultured either in CDM4MAb medium (Cat. No. SH30801.02, Hyclone) or in Freestyle™ 293 Expression medium (Cat. No. 12338018, Invitrogen), and incubated in a CO2 incubator for 5 to 7 days at 37° C. The conditioned medium was collected through centrifugation at 10,000 g for 30 minutes to remove all cells and cell debris, and filtrated through a 0.22 µm membrane before purification. Murine or recombinant antibodies containing supernatants were applied and bound to a Protein A column (Cat. No. 17127901, GE Life Sciences) following the manufacturer’s guide, washed with PBS, eluted in an acidic buffer (pH3.5) containing 20 mM citrate, 150 mM NaCl. The eluted materials were neutralized with 1 M Tris pH8.0. The procedure usually yielded antibodies with purity above 90%. The Protein A-affinity purified antibodies were either dialyzed against PBS or further purified using a HiLoad 16/60 Superdex200 column (Cat. No. 17531801, GE Life Sciences) to remove aggregates. Protein concentrations were determined by measuring absorbance at 280 nm or by Bradford assay (Cat. No. 1856210, Thermo Scientific, Rockford, IL, USA) using bovine IgG reference standard (Cat. No. 23212, Thermo Scientific). The final antibody preparations were stored in aliquots in -80° C. freezer.
The binding activities of the purified monoclonal antibodies were evaluated in ELISA and FACS assays as described in previous sections. The dose-dependent binding curves in ELISA and FACS were used to compare mAb potency. The results of two representative murine mAbs were illustrated in
The stable cell lines for human T cell-based functional assays were essentially the same as described in Pat. US8735553. Briefly, a fusion protein expression plasmid, OS8, was generated containing a scFv of anti-human CD3 mAb OKT3 and a C-terminal domain of mouse CD8α which included transmembrane and cytoplasmic domains. OS8 could function as a membrane anchored T cell engager that directly activates T-cell receptor (TCR). A stable cell line that co-expresses both OS8 and PD-L1 was generated by co-transfection of two expression constructs in HEK293 cells followed by hygromycin or G418 selection for 10-14 days. This cell line was named as HEK293/OS8/PD-L1. Similarly, a human T-cell line, HuT78/PD-1, was generated that expresses human PD-1. And a reverse signaling human T-cell line, HuT78/P3Z, was generated by stable transfection with a chimeric PD-1 expression construct (named as P3Z) made by fusing the extracellular and transmembrane domains of human PD-1 to the cytoplasmic region of human CD3ζ chain. In this way, P3Z encoded a membrane bound receptor that would activate T cells upon ligation with PD-1 ligand (PD-L1 or PD-L2). Cell lines were cloned by limiting dilution as described previously (Fuller 2001 Curr Protoc Mol Biol, Chap 11, unit 11.8).
To determine whether anti-PD-L1 antibodies can block the PD-1 signaling induced by PD-L1, HEK293/OS8/PD-L1 cells were pre-incubated with anti-PD-L1 mAbs for 15 minutes prior to co-culture with HuT78/PD-1 cells (1-3 ×104 per well) in a flat bottom plate fed with 200 µl of RPMI1640 growth medium per well at 37° C. After 16-18 hours of co-culture, supernatants were collected. IL-2 was assayed by ELISA using human IL-2 Ready-Set-Go! ELISA kits (Cat. No. 88-7025, eBiosciences, San Diego, CA). In this assay, blockade of PD-L1-PD-1 signaling with anti-PD-L1 antibodies resulted in enhanced TCR signaling and IL-2 production.
As shown in Table 4, supernatants of ELISA and FACS-binding positive hybridoma clones were screened in this functional assay. Although all the tested clones bound to PD-L1 in ELISA and FACS assays, only a few of them could block PD-L1-PD-1 signaling and resulted in increase of IL-2 production. The remaining clones resulted in either no increase or very little increase of IL-2 production compared to the negative control with fresh medium only. In this experiment, an OD450 reading cut off was set at 2.5, i.e. clones that stimulated IL-2 production above this level were considered to have antagonist functions (Table 4). A reference anti-PD-L1 mAb (named Y1) was synthesized based on the variable regions of the published data (US 2010/0203056 A1), and both human and mouse format of Y1 antibodies were generated by fusing Y1 variable regions with mouse or human IgG1κ constant regions to generate Y1-muIgG1 or Y1-huIgG1, respectively (collectively termed asY1-hIgG1). Y1-mulgG1′s function was also confirmed in this assay.
mu36
2.95 ± 0.22
mu37
3.10 ± 0.11
mu39
2.94 ± 0.45
mu324
3.09 ± 0.11
mu327
2.55 ± 0.36
mu328
3.10 ± 0.47
mu333
3.01 ± 0.23
mu334
3.22 ± 0.09
mu335
3.03 ± 0.15
mu336
3.12 ± 0.24
The purified murine anti-PD-L1 mAbs were compared in the same assay for quantitative assessments of the blocking activities.
In HuT78/P3Z cells, PD-1 mediated TCR signaling is reversed by design as described in the previous sections. In this assay, HEK293/PD-L1 cells were pre-incubated with purified PD-L1 antibodies for 15 minutes prior to co-culture with HuT78/P3Z cells in 96-well flat bottom plates at 37° C. After 16-18 hours of co-culture, supernatants were collected and IL-2 production was assayed by ELISA as described above.
Inhibitory activity of murine anti-PD-L1 mAbs was detected directly correlated to the decrease of IL-2 release in dose-dependent fashion. Consistent with the results shown above, mu333 had potent activities inhibiting IL-2 secretion by preventing PD-L1 engagement on P3Z chimeric receptor on HuT78 cells. As showed in Table 6 and
To determine whether anti-PD-L1 antibodies can compete with PD-1 binding to PD-L1, HEK293/PD-L1 cells (1×105 cells per well) were incubated with the mixture of PD-L1 antibodies and biotin-conjugated PD-1/Fc fusion protein in V-bottom 96-well plate. Biotinylation of PD-1/Fc was done using the EZ-Link Sulfo-NHS-LC-Biotin reagent according to manufacturer’s instruction (Cat. No. 21327, Thermo Sci). Inhibition of PD-L1 and PD-1/Fc interaction by antibodies was assayed (Guava easyCyte 8HT Flow Cytometer, Millipore, USA) by mean fluorescence intensity (MFI) readout probed with Streptavidin-APC. Using this method, we evaluated functional strength of anti-PD-L1 mAbs. As shown in
Besides interaction with PD-1, PD-L1 also binds to another B7 family protein, B7-1 or alternatively named as CD80 (Butte M.J. 2007 Immunity 27:111-122). To determine whether the anti-PD-L1 antibodies compete against the binding of CD80 (NCBI accession: NP _005182.1) to PD-L1, HEK293/PD-L1 cells were incubated with the mixture of PD-L1 antibodies and biotin-conjugated CD80/Fc fusion protein. In this assay, blockade of biotin-CD80/Fc binding to PD-L1 by anti-PD-L1 antibodies resulted in decreased binding signals (MFI readings). As shown in
Cloning and sequencing of variable regions from the selected murine hybridoma clones were done based on commonly used methods with some modifications (Kontermann and Dubel, 2010 Antibody Engineering, Vol 1:3-14). Briefly, bybridoma cells were harvested, washed with PBS and collected by centrifugation at 1500 rpm in a swing bucket rotor. Total cellular RNA was isolated using Ultrapure RNA kit (Cat. No. CW0581, CW Biotech, Beijing, China) following the manufacturer’s protocol.
The 1st strand cDNA was synthesized using reverse transcriptase (Cat. No. AH301-02, TransGen, Beijing, China). PCR amplification of heavy chain (Vh) and light chain variable region (Vκ) of murine mAb was performed using PCR reagent kit (Cat. No. AP221-12, TransGen, Beijing, China) and a set of primers specific for cloning of murine Vh and Vκ as described (Brocks 2001 Mol Med 7:461-469). The PCR products were subcloned into the pEASY-Blunt cloning vector (Cat. No. CB101-02, TransGen) and subsequently sequenced by Genewiz (Beijing, China). The amino acid sequences of Vh and Vk were deduced from the DNA sequencing results.
The murine mAbs were analyzed by comparison of sequence homology, and grouped based on both sequence homology and epitope-mapping results (see Example 7). Complementary determinant regions (CDRs) were defined based on the Kabat (Wu and Kabat 1970 J. Exp. Med. 132:211-250) and IMGT (Lefranc 1999 Nucleic Acids Research 27:209-212) system by sequence annotation and by internet-based sequence analysis (http: //www.imgt.org/IMGT vquest/share/texts/index.html). Table 9 listed the CDRs of mu333 (SEQ. NO. 5-14), based on the definitions of Kabat and IMGT systems.
GFSLTSYG
VH
VIWAGGSTNYNSALMS
AKPYGNSAMDY
KASQDVGIVVA
WASIRHT
QQYSNYPLYT
The three dimensional structures were simulated for variable domain of mu333 in order to identify framework residues that might be important for supporting CDR loop structures. Potentially important framework residues were kept as the original murine residues in the first round antibody humanization. The previously established structural modeling method for antibodies (Morea et al. Methods 2000 20:267-279) was adopted to simulate 3D structure of anti-PD-L1 mAb mu333 based on the known canonical structures of antibodies (Al-Lazikani et al. 1997 Journal of Molecular Biology 273:927-948). Briefly, the sequence of each variable domain (Vk and Vh) of mu333 was blasted in the PDB database (Protein Data Bank, http:// blast.ncbi.nlm.nih.gov/) to identify the most homologous antibody sequence with known high resolution structure (resolution less than 2.5 angstrom). Selected structure templates for modeling mu333 (listed in Table 10) had the same classes of canonical loop structures in L-CDR1, L-CDR2, L-CDR3, H-CDR1, and H-CDR2 to the mu333 to be modeled. As the templates for Vκ and Vh came from different immunoglobulins, they were packed together by a least-squares fit of the main chain atoms to form a hybrid structure of Vk-Vh interface residues, which was used as the templates for structural homology modeling by Swiss-model program (Kiefer et al. 2009 Nucleic Acids Research 37, D387-D392). Certain side chain conformation was adjusted while the main chain conformations were retained. At the sites where the parental structure and the modeled structure had the same residue, the side chain conformation was retained. At sites where the residues were different, side chain conformations were modeled on the basis of template structure, rotamer libraries and packing considerations. After homology modeling, PLOP program (Jacobson et al. 2002 Journal of Physical Chemistry 106:11673-11680) was used to refine the homology models to minimize all-atom energy and optimize Vk and Vh interface. This step was performed to improve the stereochemistry, especially in those regions where segments of structures coming from different antibodies had been joined together. The modeled 3D structure of mu333 variable domain was used to guide the structure-based humanization and engineering process.
For humanization of the anti-PD-L1 mAb, human germline IgG genes were searched for sequences that share high degree of homology to the cDNA sequences of mu333 variable regions by blasting the human immunoglobulin gene database in IMGT (http://www.imgt.org/IMGT_vquest/share/textes/index.html) and NCBI (http://www.ncbi.nlm.nih.gov/igblast/)websites. The human IGVH and IGVκ genes that are present in human antibody repertoires with high frequency (Glanville 2009 PNAS 106:20216-20221) and are highly homologous to mu333 were selected as the templates for humanization.
Humanization was carried out by CDR-grafting (Methods in Molecular Biology, Vol 248: Antibody Engineering, Methods and Protocols, Humana Press) and the humanization antibodies (hu333s) were engineered as the human Fab format using an in-house developed expression vector. In the initial round of humanization, mutations from murine to human amino acid residues in framework regions were guided by the simulated 3D structure, and the murine framework residues of structural importance for supporting the canonical structures of CDRs were retained in the 1st version of humanization antibody 333 (hu333-1A, SEQ. NO. 15-16). Specifically, CDRs of mu333 Vk were grafted into the framework of human germline variable gene IGVK1-5, and no murine framework residues were retained (SEQ NO 16). CDRs of mu333 Vh were grafted into the framework of human germline variable gene IGVH3-7, with 4 murine framework residues retained, V24, L67, K71 and V78 (SEQ NO 15). All grafted CDRs were based on the Kabat’s CDR definition in hu333-1A (Table 9 and SEQ. NO. 15-16). In the following hu333 variants, only the N-terminal half of Kabat H-CDR2 was grafted, as only the N-terminal half was considered to be important for antigen binding according to the simulated 3D structure (Table 14).
Hu333-1A were constructed as human Fab format using in-house developed expression vectors that contain human IgG CH-1 and constant region of kappa chain, respectively, with easy adapting subcloning sites. The hu333-1A joined IgG2a-CH1 was tagged at C-terminus with a 8xHis peptide to facilitate purification. The C232S and C233S (Kabat residue numbering, Kabat et al. Sequence of proteins of immunologic interest, 5th ed Bethesda, MD, NIH 1991) mutations were introduced in the IgG2 heavy chain to prevent disulfide bond exchange and stabilize human IgG2 in the IgG2a conformation (Lightle et al. 2010 Protein Sci 19(4): 753-762). Both constructs contained a signal peptide upstream of the Fab mature sequences. Secreted expression of hu333-1A Fab was achieved by co-transfection of the above two constructs into 293-F cells and cultured for 6-7 days before harvest. His8-tagged Fabs were purified from expression culture supernatants using a Ni-sepharose Fast Flow column (Cat. No. 17531801, GE Life Sciences) followed by size exclusion chromatography using a HiLoad 16/60 Superdex200 column (Cat. No. 17106901, GE Life Sciences). The purified Fabs were concentrated to 0.5-5 mg/mL in PBS and stored in aliquots in -80° C. freezer.
For affinity determinations of anti-PD-L1 Fabs, SPR assays were performed using BIAcore™ T-200 (GE Life Sciences). Briefly, human PD-L1/His was coupled to an activated CM5 biosensor chip (Cat. No. BR100530, GE Life Sciences) to achieve approximately 100-200 response units (RU), followed by blocking un-reacted groups with 1 M ethanolamine. A serial dilutions of 0.12 nM to 90 nM Fab samples were injected, mixed into the SPR running buffer (10 mM HEPES, 150 mM NaCl, 0.05% Tween20, pH7.4) at 30 µL/minute, and binding responses on human PD-L1/His were calculated by substracting of RU from a blank flow-cell. Association rates (kon) and dissociation rates (koff) were calculated using the one-to-one Langmuir binding model (BIA Evaluation Software, GE Life Sciences). The equilibrium dissociation constant (Kd) was calculated as the ratio koff/kon.
Functional activities of hu333 Fabs were confirmed in PD-1 competition assays described in previous sections. Data from SPR measurement and functional assays were summarized in Table 11. Hu333-1A Fab had very high affinity (Kd=9.88 pM) to PD-L1 indicated by a fast kon (1.61 × 106 M-1s-1) and very slow koff(1.59 × 10-5 s-1). It was observed that there was very slow or virtually no dissociations of hu333-1A Fab from the coated PD-L1during the 5-15 minutes of dissociation time in this experiment. It was apparent that the affinity of hu333-1A Fab to PD-L1 was close to the detection limit of the SPR technology. Such high affinity of hu333-1A Fab was consistent with the high potencies in all functional assays tested (Table 11).
Following on hu333-1A, we made individual mutations converting the four murine residues in framework region of Vh to corresponding human germline residues, respectively. At same time In order to further improve humanization level, we also changed the C-terminal part of H-CDR2 (Kabat’s definition) from murine sequence to corresponding human germline residues (Table 14, hu333-2B). Specifications of the four humanization Fabs were hu333-2A (V24A in Vh), hu333-2B (L67F in Vh), hu333-2C (K71R in Vh) and hu333-2D (V78L in Vh), which are illustrated in Table 13 with H-CDR2 changes. All humanization mutations were made using primers containing mutations at specific positions and a site directed mutagenesis kit (Cat. No. FM111-02, TransGen, Beijing, China). The desired mutations were verified by sequencing analyses. These hu333 Fabs were expressed, purified and tested in binding and functional assays as described previously. Comparing to hu333-1A, hu333-2A, hu333-2C and hu333-2D had significantly reduced binding affinities and functionalities. Only hu333-2B (SEQ. NO. 16 and 17) had similar binding and functional activities to hu333-1A (Table 11). Taken together, hu333-2B (SEQ. NO. 16 and 17) reached a high level of humanization in the framework regions while maintained potent binding affinity and functional activities.
To explore the best possible Vh and Vk sequence composition for hu333 that could be used as therapeutic antibody in human, we further engineered the hu333 by introducing mutations in CDRs and framework regions in considerations of the antibody’s molecular properties, such as physiochemical stabilities, amino acid compositions, projected isoelectronic points (pIs), expression level, antibody-dependent cell-mediated cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) while maintaining functional activities.
Deamination site NS76-77 in Vh of hu333-1A was mutated to NT76-77 to generate hu333-3A2 (SEQ. NO. 18 and 23). V60 of hu333-3A2 -Vh was mutated to V60A, which is consistent to the consensus sequences of major human IGVH3 genes with reduced a surface-exposure of hydrophobicity. This mutant construct was given the code name hu333-3C2 (SEQ. NO. 19 and 23). Another deamidation site NS73-74 was mutated to TS73-74 on the template of hu333-3C2, which is also consistent to the consensus sequences of major human IGVH3 genes. The latter one was named as hu333-3H2 (SEQ. NO. 20 and 23). As summarized in Table 12, hu333-3A2, hu333-3C2 and hu333-3H2 all retained the potent functional activity, only with slight variations in binding affinity. On the other hand, these engineered hu333 variants have better projected physiochemical properties.
To eliminate the last deamidation site in the CDR3 of Vh, we mutated NS101-102 to TS101-102 on the templates of hu333-3A2 and hu333-3H2, respectively. The resulting humanization mAbs were constructed in human IgG1 Fab format, named as hu333-4A2 (SEQ. NO. 21 and 23) and hu333-4B2 (SEQ. NO. 22 and 23). The results from binding and functional assays indicated both hu333-4A2 and hu333-4B2 were very similar in affinity and functional activities such as blocking the PD-L1 binding to its targets (PD-1 and CD80) and inhibiting the PD-L1 and PD-1 mediated downstream signaling (Table 13 and Table 14). Several mutations made in the processes including hu333-3B2, -3D2, -3E2, -3G2 and -3I2 were dropped from further development for various considerations. The CDRs of the above mAbs were compared to those of mu333 were shown in Table 14.
All the humanization mAbs shown above were also confirmed for functional effect on primary human immune cells, peripheral blood mononuclear cells (PBMCs), which were isolated from healthy donors by density gradient centrifugation using ficoll lymphocyte separation medium (Histopaque-1077; Cat. No. 10771, Sigma, St. Louis, USA) according to manufacturer’s instruction. PBMCs were then stimulated with 40 ng/mL of anti-CD3 mAb OKT3 (Cat. No. 16-0037, eBioscience, San Diego, CA, USA) for 3 days prior to the assay. The activated PBMC population mainly consisted of T-cells (50-70%), B-cells and NK cells (15-30%), and monocytes (2-10%). To better mimic the response of T cells to PD-L1 expressing tumor cells upon engagement of TCR/CD3 complex, the activated PBMCs were co-cultured with HEK293/OS8/PD-L1 cells in each well of 96-well plates. Functional effect of anti-PD-L1 mAbs were tested by adding the mAb to the culture, co-cultured for 15-18 hours before harvesting culture supernatants to assess IFN-y level using Ready-Set-Go! ELISA kits (Cat. No. 88-7316, eBiosciences). As shown in
PD-L1 is expressed on a wide range of normal human cells including hematopoietic cells such as T-cells, B-cells, dendritic cells, macrophages, mesenchymal stem cells and bone-marrow derived mast cells, and nonhematopoietic cells and tissues such as lung, hepatocytes, pancreatic islets, placental synctiotrophoblasts and vascular endothelium (Keir et. al. 2006 J Exp Med 203:883-895, Keir et. al. 2008 Ann Rev Immunol 26:677-704, Mu et. al. 2011 Medical Oncology 28:682-688). It is expected that PD-L1 blocking antibodies linked to human wild type IgG-yFc moieties will induce yFc-mediated effector functions, for examples, antibody-dependent cell-mediated cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC), which might lead to unwanted toxicity to vital organs.
To eliminate effector functions associated with anti-PD-L1mAbs while maintaining optimal physicochemical properties, we constructed hu333-4A2 and hu333-4B2 full antibody by linking the Vh sequences to mutated IgG1 constant regions, and screened for reduced or null Fcγ receptors (FcyRs) binding or C1q binding activities, therefore, attenuating or eliminating ADCC and CDC effector functions. The regions in IgG1 Fc that are involved in interactions with FcyRs and C1q have been studied extensively in the literature (Tao et al. 1993 J Exp Med 178:661-7; Cole et al. 1997 J Immunol 159:3613-21; Armour et. al. 1999 Eur J Immunol 29:2613-2624; Idusogie et. al. 2000 J of Immunol 164:4178-4184; Shields et. al. 2001 J of Biol Chem 276: 6591-6604; Lund et. al. 2002 Immunol Letters 82:57-65; reviewed by Strohl et. al. 2009 Current Opinion in Biotechnology 20:685-691). Taken together, these data have pointed to the essential role of lower hinge region (AA232-238 based on EU nomenclature) for binding to FcyRs and a structurally clustered region (D270, K322, P329 and P331based on EU nomenclature) of CH2 domain for binding to C1q. On the other hand, IgG2 has some sequence variations from IgG1 in the hinge region, which was attributed to weaker binding or no binding to most of the FcyRs except to FcγRIIAH131. Indeed, a IgG1/IgG2 hybrid (IgGlΔb) with most of IgG1 hinge sequence incorporating some IgG2 sequences was demonstrated having significantly reduced the binding activities to most FcyRs and attenuated ADCC and CDC effector functions (Armour et. al. 1999 Eur J Immunol 29:2613-2624).
By rational design of mutagenesis with considerations of good pharmaceutical and physicochemical properties, we generated a number of mutants IgG1 in the hinge and Fc regions described above, which were fused to the variable regions of hu333-4A2 and hu333-4B2, respectively, as full antibodies. Two of the IgG1 mutants, IgG1mc and IgG1mf, with favorable features in functional assays were shown in Table 15 in comparison to wild type IgG. The IgGlmc (SEQ. NO. 28) contains a combination of additional mutations, V234A, G237A and P239A, from the IgG1/IgG2 hybrid described above. The mutations of V234A and G237A were designed to reduce the surface hydrophobic side chain at the yFc/FcyR binding interface to further reduce the binding to FcyRIIA and FcyRIIB, (Lund et. al. 1992 Mol Immunol 29:53-59, Lund et. al. 1996 J Immunol 157:4963-4969, Wines et. al. 2000 J Immunol 164:5313-5318). The P239A mutation was designed to further reduce the C1q binding and CDC (Idusogie et. al. 2000 J of Immunol 164:4178-4184). The IgG1mf (SEQ. NO. 29) was similar to IgG1mc except that the amino acid residue G237 was not mutated. The recombinant full length anti-PD-L1 antibodies, hu333-4A2-IgG1mc (SEQ. NO. 30 and 32), hu333-4B2-IgG1mc (SEQ. NO. 31 and 32) and hu333-4B2-IgG1mf (SEQ. NO. 32 and 33) were expressed in HEK293-F cells and purified as described in previous sections.
P
A
A
A
A
P
A
A
A
It was well documented that IgG mediated effector functions are triggered following on antibody-antigen complex binding to FcyRs or to complement component C1q (Nimmerjahn et. al. 2008 Nature Rev Immunol 8:34-47). For example, ADCC is initiated when an antibody binds to cell surface target protein followed by ligation to FcyRIIIA expressed on effector cells. CDC is activated when an antibody cross-links a cell surface target by binding to C1q protein, which leads to a cascade reaction of complement complex formation and activation and target cell lysis. As proxy of ADCC, CDC and other antibody mediated effector functions, biochemical assays for antibody binding to FcyRs and C1q may serve as the fundamental indicator of ADCC and CDC. We systematically assessed the bindings of the anti-PD-L1 antibodies with modified constant region to all major FcyRs and all known polymorphic variants, including FcyRI, FcγRIIAH131, FcγRIIAR131, FcγRIIIAF158, FcγRIIIAV158, FcγRIIB, and FcγRIIIB.
The extracellular domains of FcyRs were fused to C-terminal His tags as described in previous sections. Recombinant proteins were expressed in 293-F cells by transient transfection and purified using Ni-Sepharose column followed by gel filtration column as described. 2-5 µg/mL of FcyRs were coated on Nunc MaxiSorp ELISA plates (Cat. No. 442404, Nunc, Thermo Fisher) except FcyRIIB and FcyRIIIB, Ni-chelate plates were used (Cat. No. 15242, Pierce, Thermo Fisher). After washing and blocking of the wells, a preformed immune-complex was added to each well and incubated at room temperature for 1-2 hours. The preformed immune-complex contained 60 ng/mL streptavidin-HRP, 60 ng/mL of biotinylated-F(ab′)2 goat anti-human IgG (Cat. No. 109-066-097, Jackson ImmunoRes, West Grove, PA, USA), and 1-5 µg/mL of indicated IgG1 Fc variants fused to the humanized anti-PD-L1 (hu333-4A2 or hu333-4B2) in the blocking buffer. After washing the plate four times, binding signals were detected by chemiluminescence using Immobilon Chemiluminescence Substrate A/B (Cat. No. WBKLS0500, Millipore). Table 16 summarized the results of hu333-4A2-IgG1mc (SEQ. NO. 30 and 32), hu333-4B2-IgG1mc (SEQ. NO. 31 and 32) and hu333-4B2-IgG1mf (SEQ. NO. 32 and 33) binding to various FcyRs. In comparison to the hu333-4A2-IgG1wt, all three IgG1 mutant hu333 mAbs had very low binding activities to FcyRs, which indicated that all three hu333 mAbs above would have significantly reduced effector functions mediated by FcyRs.
Bindings of humanized anti-PD-L1 in various IgG1 formats (wt, IgG1mc, IgG1mf, SEQ. NO. 27-29) to FcyRs were also determined by flow cytometry. In brief, a series of stable HEK293 transfectants expressing human FcyRs were established. These stable cell lines expressed FcyRI, FcγRIIAH131, FcγRIIAR131, FcγRIIB, FcγRIIIAF158 or FcγRIIIAV158, respectively. Multi-subunit FcyRs (i.e., FcγRI and FcγRIIIA) were co-expressed with FcRy subunit. A secondary antibody (goat anti-human IgG F(ab′)2-Alexa Fluor 488, Cat. No. 109-546-097, Jackson ImmunoResearch, West Grove, PA, USA) was used to detect the binding of monomeric anti-PD-L1 mAbs with the IgG1 variants (Table 17) to FcγR expressing HEK293 cells. As expected, hu333-4A2 in IgG1wt format (hu333-4A2-IgG1wt) had strong binding signals (MFI) to FcyRI, FcγRIIAH131 and FcγRIIIAV158, and weak but significant binding signals to FcγRIIAR131, FeγRIIB and FcγRIIIAF158 (Table 17). The modified IgG1 variants (hu333-4A2-1gGlmc, hu333-4B2-IgG1mc and hu333-4B2-IgG1mf, SEQ. NO. 30-33) had significantly reduced binding signals which were close to backgrounds.
It has been shown that antibodies bind to FcγRs with much bigger strength in the forms of immune-complexes, which is due to the multivalent effect (Bruhns et. al. 2009 Blood 113:3716-3725). Such bindings are thought to be more relevant under physiological condition, as the binding strength between monomeric γFc and most of the FcγRs is very weak. Human immune system also takes advantage of this mechanism to avoid non-specific activation of FcγRs by monomeric IgG which are present at high levels in serum. In order to assess the bindings to FcyRs in the form of immune-complexes, 10 µg/mL of humanized 333 mAb as various IgG1 mutant forms were premixed with 3 µg/mL of biotin-PD-L1/His and 1.5 µg/mL of neutravidin (Cat. No. A-2666, Invitrogen) in FACS buffer to form the multivalent immune-complexes, before incubating with FcyR-expressing HEK293 cells. Goat anti-human IgG F(ab′)2-Alexa Fluor 488 (Cat. No. 109-546-097, Jackson ImmunoResearch) was used to detect the bindings. As shown in Table 18, hu333-4A2-IgG1wt in preformed immune-complex bound to the low affinity FcyRs (FcyRIIA, FeγRIIB, and FcyRIIIA) with much better strength than monomeric IgG1 does (Table 18 data vs. Table 17 data). And again, the anti-PD-L1 mAbs in selected IgG1 mutants (hu333-4A2-IgG1mc, hu333-4B2-IgG1mc and hu333-4B2-IgG1mf, SEQ. NO. 30-33) had significantly reduced binding signals which were close to backgrounds. Taken together, the humanized 333 in modified IgG1 formats had very little bindings to FcγRs, therefore they should have little FcyRs-mediated effector functions under physiological conditions.
The ELISA-based C1q binding assay was done by conventional ELISA method with minor modification. Briefly, indicated amounts of the humanized 333 antibodies fused to either wild type or modified IgG1 constant regions were coated onto the Maxisorp ELISA plate. After blocking and washing, the wells were incubated with 2 µg/mL of human C1q (Cat. No. A400, Quidel, San Diego, USA) at room temperature for 2 hours. After washing, the bound C1q was detected using a murine monoclonal antibody against human C1q (Cat. No. A201, Quidel) and HRP conjugated anti-murine IgG (Cat. No. A0168, Sigma, Shanghai, China). As shown in
Classical antibody-dependent cellular cytotoxicity (ADCC) involves activation of NK cells by antibodies engaging to FcyRIIIA (CD16). To test whether humanized anti-PD-L1 antibodies fused to selected human IgG1 variants induce ADCC, NK92MI/CD16V cells, which were generated from NK92MI cells (Cat. No. CRL-2408, ATCC) by co-transducing expression plasmids containing CD16 (V158 allele) and FcRy genes, were used as effector cells, and PD-L1-expressing HEK293 cell line, HEK293/PD-L1, was used as target cells. The effector cells (105 cells/well) were co-cultured with target cells (104 cells/well, E:T=10) in 96-well V-bottom plates in the presence of hu333-IgG1 variants (0.0001-1 µg/ml) for 5h. Cytotoxicity of NK92MI/CD16 cells exerted against HEK293/PD-L1 cells was determined by lactate dehydrogenase (LDH) release assay using the CytoTox 96 Non-Radioactive Cytotoxicity Assay kit (Promega, Madison, WI). Specific lysis was determined by the following equation.
Consistent with the fact that hu333 in IgG1mc and IgG1mf formats had no or significantly reduced bindings to FcyRIIIA (see above section), the ADCC assays showed that both hu333-4B2-IgG1mc (SEQ. NO. 31 and 32) and hu333-4B2-IgG1mf (SEQ. NO. 32 and 33) had only base level of ADCC. In contrast, 333-4A2-IgG1wt with wild type IgG1 Fc induced 20% specific cell lysis at the concentration of 1 µg/mL (
Human IgG1 antibodies, in general, induce significant complement-dependent cytotoxicity (CDC) via classical pathway. Whether the humanized anti-PD-L1 antibodies in selected IgG1 mutant formats (IgG1mc and IgG1mf) trigger CDC was evaluated using a PD-L1-expressing B cell line, Daudi/PD-L1, and fresh human serum from healthy donors, which contains all necessary components for CDC. Cell lysis by CDC was determined by Celltiter glo assay kits (Promega). In brief, Daudi/PD-L1 cells (2×104 cells/well) were incubated in serum-free RPMI1640 (Invitrogen) with anti-PD-L1 Abs (0.001-10 µg/mL) at 37° C. for 15 minutes before adding normal human serum to the final concentration of 16.6% in 96-well flat-bottom plates in a total volume of 120 µL. After overnight incubation at 37° C., cells were lysed and assayed for ATP concentration. Anti-CD20 mAb Rituximab (Roche) was used as a positive control as Daudi cells constitutively express CD20. The amount of ATP is directly proportional to the number of cells present in culture. Fluorescence was read using a 96-well fluorometer (PHERA Star FS, BMG LABTECH). The results are expressed in relative fluoresence units (RFU) that are proportional to the number of viable cells. The percent CDC activity was calculated as follows: % CDC activity = [(RFU test - RFU background) / (RFU at total cell lysis - RFU background)] × 100. As shown in
The three humanized mAbs in modified IgG1 formats described above were characterized in cell-based binding assays and functional assessment. Table 19 summarized the study results about hu333-4A2-IgG1mc, hu333-4B2-IgG1mc and hu333-4B2-IgG1mf (SEQ. NO. 30-33).
FACS binding analysis was performed as described in previous sections. Serial dilutions of antibodies were incubated with HEK293/PD-L1 cells and the bindings were detected using the Goat anti-human IgG F(ab′)2-Alexa Fluor 488 (Cat. No. 109-546-097, Jackson ImmunoResearch). Dose-dependent binding activities were observed for the selected mAbs to native PD-L1 protein expressed on surface of HEK293 cells. As shown in Table 19, hu333-4A2-IgGlmc, hu333-4B2-IgGlmc and hu333-4B2-IgG1mf showed similar dose-dependent binding activities to the HEK293/PD-L1 cells with EC50 (effective concentration at 50% activity) around 0.1 µg/mL.
FACS based competition assays was performed as described earlier. The results shown in Table 19 demonstrated that hu333-4A2-IgGlmc, hu333-4B2-IgG1mc and hu333-4B2-IgG1mf compete off both PD-⅟Fc binding (IC50S of 0.167- 0.174 µg/mL) and CD80/Fc binding (IC50s of 0.078-0.118 µg/mL) to HEK293/PD-L1 cells almost equally well.
The functionalities of the purified anti-PD-L1 mAbs were assessed in the HuT78/PD-1 and HEK293/OS8/PD-L1 co-culture system as described in previous section. As shown in Table 19, the humanized 333 mAbs were potent antagonists of PD-L1/PD-1 signaling in this co-culture system, and induced increased IL-2 secretions. Consistent with the result of FACS-based competition assay, hu333-4A2-IgG1mc, hu333-4B2-IgG1mc and hu333-4B2-IgG1mf showed similar potencies in this assay with very close EC50 (0.075-0.087 µg/mL) and maximum induction of IL-2 levels (287-300 pg/mL).
The functionalities of the purified anti-PD-L1 mAbs were also assessed in the reversed signaling system in which HuT78/P3Z and HEK293/PD-L1 were co-cultured as described. Consistently, the humanized 333 mAbs were potent inhibitor of PD-L1/P3Z signaling in this co-culture system, and inhibited IL-2 secretions induced by PD-L1/P3Z engagement. And again, hu333-4A2-IgG1mc, hu333-4B2-IgG1mc and hu333-4B2-IgG1mf showed similar potencies in the assay, as shown by similar IC50s (0.037-0.045 µg/mL) and maximum inhibition levels (99%) (Table 19).
To verify if the humanized 333 antibodies also exert functional effect on primary human immune cells, we assayed the antibody function using freshly isolated peripheral blood mononuclear cells (PBMCs), which are mainly consisted of T-cells (50-70%), B-cells and NK cells (15-30%), and monocytes (2-10%). Human PBMCs were isolated from healthy donors by density gradient centrifugation using ficoll lymphocyte separation medium (Histopaque-1077; Cat. No. 10771, Sigma) according to manufacturer’s instruction. The human blood collections were done followed the Internal Procedure of BeiGene. PBMCs were then stimulated with 40 ng/mL of anti-CD3 mAb OKT3 (Cat. No. 16-0037, eBioscience, CA) for 3 days prior to the assay. To mimic the response of pre-activated T cells to PD-L1 expressing tumor cells upon engagement of TCR/CD3 complex, PBMCs (1×104 cells) were co-cultured with HEK293/OS8/PD-L1 cells (3×104 cells) in each well of 96-well flat-bottom plates. Indicated concentrations of anti-PD-L1 antibodies were added to the culture. After 15-18 hours of co-culture, culture supernatants were assayed for IFN-y level by ELISA using Ready-Set-Go! ELISA kits (Cat. No. 88-7316, eBiosciences), which is the most prominent indicator of T-cell activation, as well as of other immune cell activation (Thakur et. al. 2012 Vaccine 30:4907-4920). As shown in
Taken together, these data demonstrated that hu333-4A2-IgG1mc, hu333-4B2-IgGlmc and hu333-4B2-IgG1mf were potent antagonists blocking PD-L1/PD-1 interactions and downstream signaling in all the cell line and primary immune cell-based assays. They were very similar in their functional activities and potencies, as they were very similar in sequences (only minor difference in framework regions), shared identical binding epitope and had very similar binding affinities and specificity (see below section).
The binding specificity was studied for the mAbs hu333 (hu333-4A2-IgG1mc, hu333-4B2-IgG1mc and hu333-4B2-IgG1mf) using human, cynomolgus monkey (Macaca fascicularis) and mouse (Mus musculus) PD-L1 as target proteins. The monkey PD-L1/His and murine PD-L1/His were expressed and purified in a similar way to the human PD-L1/His as described earlier. Y1 was a reference functional anti-PD-L1 mAb which was synthesized according to a published patent (US 2010/0203056 A1) and fused to human IgG1mc variant. The synthesized full length mAb was named as Y1-IgG1mc. The ELISA assay was performed essentially in the same way as described in the previous section. Briefly, 200 ng of PD-L1 protein was coated on each well of Nunc MaxiSorp ELISA plate (Cat. No. 442404, Nunc, Thermo Fisher). After washing and blocking, indicated concentrations of anti-PD-L1 mAbs were added and incubated at room temperature for 1 hour. After washing, the bound anti-PD-L1 mAbs were detected using a HRP-conjugated goat anti-human Fc antibody (Cat. No. A0170, Sigma). As shown in
Hu333-4A2 (SEQ. NO. 21 and 23), hu333-4B2 (SEQ. NO. 22 and 23) and the reference antibody Y1 were constructed as human IgG1 Fab format in which the Vh and Vk were fused to the N-terminus of human IgG1-CH1 and constant region of kappa chain, respectively. The IgG1-CH1 was fused to a C-terminal His6 tag to facilitate purification. Expression and purification of recombinant Fabs were performed as described in the previous section.
For affinity determinations of anti-PD-L1 Fabs, SPR assays were conducted using BIAcore™ T-200 instrument (GE Life Sciences, Shanghai, China) as described earlier. Association rates (kon) and dissociation rates (koff) were calculated using the one-to-one Langmuir binding model (BIA Evaluation Software, GE Life Sciences). The equilibrium dissociation constant (Kd) was calculated as the ratio koff/kon.
The SPR-determined binding affinities of anti-PD-L1 Fabs were listed in Table 20. Hu333-4A2 and hu333-4B2 Fabs bind to human PD-L1 with higher affinities than Y1 Fab does, which was indicated by the faster kon, slower koff and much smaller Kdvalue. The hu333-4A2 and hu333-4B2 Fabs bind to monkey PD-L1 almost equally well as their binding to human PD-L1. In contrast, Y1 Fab binds to the monkey PD-L1with about 100-fold lower affinity than its binding to human PD-L1 (Kd of 0.18 nM to human PD-L1 and Kd of 16.2 nM to monkey PD-L1).
The previous reports about the crystal structures of PD-⅟PD-L1 complexes shed light on the critical amino acid (AA) residues of PD-L1 that directly interact with receptor PD-1 (Zhang et. al. 2004 Immunity 20:337-347; Lin et. al. 2008 PNAS 105:3011-3016; Lazar-Molnar et. al. 2008 PNAS 105:10483-10488). Through point mutation analysis, eight AA residues in PD-L1 sequence were identified being required for its binding to PD-1. Based on the information from the structure-guided mutation analysis, we hypothesized that most effective way for the functional mAbs to block PD-L1 mediated signaling is to compete with PD-1 by binding to the eight critical AA residues, therefore, occupying the binding epitopes required for its binding to PD-1 receptor. To explore the hypothesis and to understand the mechanism of action by functional PD-L1 mAbs, we made eight mutants of PD-L1 by replacing each of the eight critical AAs to Ala, individually, i.e. F19A, I54A, R113A, M115A, D122A, Y123A, K124A and R125A (AA residue numbering based on Lin et. al. 2008 PNAS 105:3011-3016). The wild-type PD-L1/His (
ELISA assays using the wild-type and mutant PD-L1/His were performed to study the binding epitopes of the anti-PD-L1 mAbs. For comparison of antibody binding epitopes, several murine mAbs from us and one reference antibody Y1-IgG1 (adapted from US 2010/0203056 A1 and fused to human IgG1kappa constant regions) were included in the study. Equal volume of CM containing wild type or mutant PD-L1/His was coated in 96-well plate for all mAbs in the same ELISA assay. The ELISA results were normalized using the mean ELISA readings of wild type PD-L1 binding signals as the standard. ELISA binding signals to a specific mutant PD-L1 were further normalized against the highest antibody binding read-out (set as 100%) to the specific mutant PD-L1 to even out expression variations between PD-L1 mutants. For convenience of data analysis, when a mAb’s ELISA binding signal for a specific PD-L1 mutant dropped below 50% relative to wild type PD-L1, it was defined that the binding function is significantly impaired due to the corresponding amino acid mutation. Likewise, if a mAb’s ELISA binding signal for a specific mutant dropped below 25%, it was defined to be very significant.
As shown in
Apart from the above key binding epitope mutations, we also made the mutation D26A. ELISA and Western blot results showed the mutation D26A in PD-L1 significantly inhibited the binding activities of all functional anti-PD-L1 mAbs including mAbs mu333, hu333-4B2-IgG1 and Y1-IgG1, but not inhibited the binding of non-functional antibodies, such as mu260 (
Through the epitope mapping study, we have demonstrated that anti-PD-L1 mAbs are capable of binding to different epitope signatures through molecular recognition, which might have profound impact on binding affinity, binding specificity and functional activity, e.g. hu333-4A2 and hu333-4B2 can only bind to human PD-L1 (
In order to check whether mAb mu333 has non-specific binding to human serum proteins, ELISA study was performed using 96-well Nunc Maxisorp ELISA plates coated with 5% human serum (from healthy donors) and various concentrations of PD-L1/His antigen as indicated in
As showed in
All animal studies were performed following BeiGene Animal Care and Use Procedure. Ten to twelve week-old female Balb/c nude mice (18-25 g) were used to study the pharmacokinetics of the humanized mAb hu333-4B2-IgG1mf (SEQ. NO. 32 and 33). Mice were dosed with 10 mg/kg of mAb hu333-4B2-IgG1mf either as a single intravenous (i.v.) or subcutaneous (s.c.) injection. Intravenous injections were administered via a tail vein, and subcutaneous injections were administered in the flank. In each injection group, mice were separated into different subgroups and in each subgroup blood sera were collected at certain time points. For i.v. injection group, serum was harvested 2 days predose, and postdose at 15 min, 30 min, 60 min, 90 min, 6 h, 24 h and once on days 2, 3, 4, 5, 7, 10, 14, 21 and 28. For s.c. injection group, serum was harvested 2 days predose, and postdose at 1.5 h, 6 h, 24 h and once on days 2, 3, 4, 5, 7, 10, 14, 21, and 28.
Serum level of hu333-4B2-IgG1mf was determined by ELISA using human PD-L1/His protein. Briefly, Nunc MaxiSorp ELISA plates (Cat. No. 442404, Nunc, Thermo Fisher) were coated overnight at 4° C. with 100 µL per well of 3 µg/mL human PD-L1/His protein. Plates were blocked with 3% bovine serum albumin, 0.05% Tween 20 in PBS (blocking buffer) at room temperature for 1 hour. After washing, serially diluted serum samples and purified hu333-4B2-IgG1mf standards were added and incubated at room temperature for 1 hour. After washing, the bound hu333-4B2-IgG1mf was detected using a HRP-conjugated goat anti-human Fc antibody (Cat. No. A0170, Sigma) and color developed using TMB substrate (Cat. No. T0440, Sigma). A standard curve was fit using nonlinear regression and the serum concentrations of hu333-4B2-IgG1mf were deduced from the standard curve and dilution factors.
The serum concentrations of hu333-4B2-IgG1mf versus time data were analyzed using the non-compartment model for the i.v. and s.c. doses (WinNonlin, Pharsight). The clearance, volume of distribution, half-lives, mean residence time and bioavailability were deduced from WinNonlin data fitting.
The pharmacokinetics of hu333-4B2-IgG1mf in mice was summarized in Table 22. After i.v. administration, hu333-4B2-IgG1mf (SEQ. NO. 32 and 33) concentrations were cleared from the serum in a biphasic manner. The terminal half-life was about 10-14 days. After an i.v. dose of 10 mg/kg, the clearance was 7.9 mL/day/kg in mice. After s.c. administration, peak concentrations of hu333-4B2-IgG1mf in the serum as approximately 30-50% of that noted after i.v. administration of the same dose. Comparison of the AUC after the 10 mg/kg s.c. and i.v. dose indicated a bioavailability of 90%. All these PK parameters were close to those of typical humanized monoclonal antibodies, indicating that hu333-4B2-IgG1mf (SEQ. NO. 32 and 33) had good in vivo stability in mice.
The pharmacokinetics of hu333-4B2-IgGlmc (SEQ. NO. 31 and 32) was studied in cynomolgus monkeys. As humanized 333 bound to human and monkey PD-L1 with almost identical affinities, the pharmacokinetic profile in cynomolgus monkeys should be very informative and scalable to predict the pharmacokinetic profile in humans. The drug administrations and blood serum collections were done at 3D BioOptima Co. Ltd (Suzhou, China) following 3D BioOptima’s Animal Care and Use Procedure. Briefly, two 3-5 year-old male monkeys were dosed with 10 mg/kg of mAb hu333-4B2-IgG1mc as a single intravenous (i.v.) dose. Blood samples (~1 mL) were collected at 2 days pre-dosing, 5 min, 30 min, 2 h, 6 h, 12 h, 24 h, 36 h, and 2, 3, 5, 7, 10, 15, 22, 30 days post-dosing via cephalic vein into tubes.
ELISA based bioanalyses and pharmacokinetic analyses were performed essentially as described above. At each time point, the averaged serum concentration from 2 monkeys was used for fitting except for the time points of 22 and 30 days post dose, where the data from only one monkey were used, as another monkey showed accelerated clearance and undetectable hu333-4B2-IgG1mc serum levels, presumably due to a monkey anti-drug immune response, at these two time points. The serum concentration of hu333-4B2-IgG1mc versus time data were analyzed using the non-compartment model for the i.v. dose.
The pharmacokinetics of hu333-4B2-IgG1mc in cynomolgus monkeys was summarized in Table 23. After i.v. administration, hu333-4B2-IgG1mc concentrations were cleared from the sera in a biphasic manner. The terminal half-life was about 9 days. After an i.v. dose of 10 mg/kg, the clearance was 6.4 mL/day/kg in cynomolgus monkeys. After i.v. administration, peak concentration of hu333-4B2-IgG1mc was 283 µg/mL at 5 min post dose. These PK parameters indicated that hu333-4B2-IgG1mc had normal pharmacokinetic profile in cynomolgus monkeys, which predicted normal pharmacokinetic behavior in humans (Deng et. al. 2011 mAbs 3:61-66).
The T-cell line and PBMC-based experiments indicated that the anti-PD-L1 mAb might work in mouse cancer models utilizing immune-compromised mice xenografted with human cancer cells, subsequently implanting human PBMCs and applying the mAb treatment to inhibit cancer cell growth in vivo. An allogeneic mouse cancer model was designed as follows. Female NOD/SCID mice (6-7 weeks) were pre-treated with cyclophosphamide. Human peripheral blood mononuclear cells (PBMCs) were isolated from blood of healthy human volunteer, mixed with A431 epidermoid carcinoma cells (Cat. No. CRL-1555 ATCC) and matrigel, and injected subcutaneously into the right front flank of the animals. Starting from day 0 after cell inoculation, animals were randomly assigned into 3 groups with 8 mice per group. Mice were treated twice weekly (BIW i.p.) with vehicle (PBS) or 10 mg/kg hu333-4B2-IgG1mf (SEQ. NO. 32 and 33) for 4 weeks. Individual animal body weight and tumor volume were recorded twice weekly, with mice being monitored daily for clinical signs of toxicity for the duration of the study. Tumor volumes were calculated using the following formula: [D × (d2)]/2, in which D represents the large diameter of the tumor, and d represents the small diameter.
As shown in
| Number | Date | Country | Kind |
|---|---|---|---|
| PCT/CN2014/081581 | Jul 2014 | WO | international |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 16704738 | Dec 2019 | US |
| Child | 18050441 | US | |
| Parent | 15323153 | Dec 2016 | US |
| Child | 16704738 | US |
