Introduction
Immune surveillance plays a critical role in cancer prevention. However, in situations where tumors develop resistance mechanisms to suppress the host immune system, tumors eventually grow out of control [
1,
2]. One of such resistance mechanism is the up-regulation of the immune check-point ligand, PD-L1, in tumor cells or in tumor-associated immune cells. PD-L1 interacts with PD-1 (programmed cell death-1) on T cells, inhibiting T-cell proliferation and effector functions such as cytokine secretion and tumor cell-killing [
3,
4]. Several PD-1 antagonist antibodies have been tested in clinical trials, and show significant efficacy in the treatment of advanced cancer types [
5‐
9]. Two anti-PD-1 antibodies, nivolumab, and pembrolizumab recently gained regulatory approval [
3].
Antibody drugs exert primary pharmacodynamic effects through specific binding to a target protein and modulating its functional activity via the variable regions. Furthermore, the constant region of an antibody also plays important roles by exerting secondary pharmacodynamic effects through the binding to FcγRs or activation of complement cascade. Each IgG subclass has a unique set of features for binding to effector receptors that elicit profound functional effects on the target cells [
10,
11].
Most of the anti-PD-1 monoclonal antibodies (mAb), including nivolumab and pembrolizumab, have IgG4
S228P heavy chain, which retains effector-binding functions similar to that of wild-type human IgG4 [
11], while it possesses more stable dimeric structure without fab-arm exchange observed in wild-type IgG4 [
12,
13]. It was well documented that human IgG4 has significant binding to high affinity FcγRI through the Fc-hinge regions [
11]. IgG4
S228P antibodies likely retain the binding to FcγRI. In a syngeneic mouse model, an anti-PD-1 mAb with “effector-less” Fc region demonstrated superior anti-tumor efficacy as compared with the one with effector functions [
14]. However, functional consequences of FcγRI engagement by anti-PD-1 mAb through IgG4
S228P have not been well studied.
FcγRI is highly expressed in type 2 macrophages (M2) under inflammatory conditions in certain tumor types [
15]. It is also expressed in myeloid-derived suppressor cells (MDSCs) and type I macrophages (M1). The functions induced by FcγRI engagement span a wide scope of cellular activities including antibody-dependent cell phagocytosis (ADCP), cell proliferation, and production of cytokines depending on the cell type in which FcγRI is activated [
16].
In this report, we studied the functional consequences of an anti-PD-1 mAb with an IgG4S228P heavy chain, for which we generated a pair of anti-PD-1 antibodies with the same variable regions, but with different forms of the IgG4 heavy chain: BGB-A317/IgG4S228P (with a single S228P mutation) and BGB-A317 (lack of FcγR-binding capacity). Comparative characterization of these two mAbs demonstrated that BGB-A317/IgG4S228P binds to human FcγRI with high affinity and mediates crosslinking between PD-1+ T cells and FcγRI+ cells. In addition, the two BGB-A317 mAbs showed profound differences in their capacity to modulate T-cell and macrophage functions in vitro or inhibit tumor growth using xenograft model in vivo.
Materials and methods
Binding affinity assay by SPR
For the characterization of the binding affinity of BGB-A317 or BGB-A317/IgG4S228P to human PD-1, the extracellular domain of the human PD-1 protein, with a His tag (PD-1/His), was coupled to an activated CM5 biosensor chip (Biacore®, GE Healthcare Life Sci). BGB-A317 or BGB-A317/IgG4S228P samples were injected and binding responses to human PD-1/His were calculated by subtracting the response unit (RU) from the values measured for a blank flow cell. Association rates (Kon) and dissociation rates (Koff) were calculated using the BIA Evaluation Software (GE Life Sciences). The equilibrium dissociation constant (KD) was calculated as the ratio of Koff/Kon, and binding affinity was calculated as Kon/Koff.
For the analysis of the binding of BGB-A317 and BGB-A317/IgG4S228P to human FcγRI, human PD-1/His was immobilized on a CM5 chip. FcγRI was injected over a human PD-1/His-captured antibody surface. The sandwich binding of PD-1, BGB-A317/IgG4S228P, and FcγRI was performed by flowing FcγRI protein on top of the PD-1, BGB-A317/IgG4S228P complex bound on the sensor chip. The same procedure was applied to BGB-A317.
Generation of stable cell lines
A Chimeric PD-1 receptor, P3Z, was constructed by fusing the extracellular and transmembrane domains of human PD-1 to the cytoplasmic domain of human CD3ζ chain, which was stably transduced into HuT78 cells (ATCC) to generate HuT78/P3Z cells. The HEK293/PD-L1 and HEK293/FcγRI cell lines were generated by stable transfection. HuT78/P3Z cells were cultured in complete RPMI1640 (Hyclone) supplemented with 10% heat-inactivated FBS (Corning), and incubated at 37 °C with 5% CO2. HEK293-based cell lines were cultured in DMEM (Invitrogen) supplemented with 10% FBS (Corning).
P3Z assays
HuT78/P3Z cells were pre-incubated with BGB-A317 (0.0014-3 µg/mL) for 15 min prior to co-culturing with HEK293/PD-L1 cells in 96-well plates (Costar) containing complete RPMI1640 media. The cells were incubated for 17 h at 37 °C. IL-2 secretion in the supernatants of co-culture was assayed by ELISA using a kit from eBioscience, according to manufacturer’s instructions.
Human PBMC separation
Human peripheral blood mononuclear cells (PBMCs) were isolated from healthy donors by density gradient centrifugation using Histopaque-1077 (Sigma) according to the manufacturer’s instructions. All procedures were approved by the Internal Review Board at BeiGene. Consent agreement forms were signed before blood donation.
ELISA
In the ELISA-based assay, bait proteins (1 µg/mL, 100 ng/well) were coated onto MaxiSorp plates (Thermo Fisher). For BGB-A317 antibody binding to PD-1, the bait protein is PD-1/his. For assessing antibody and FcγR binding, the extracellular domain of FcγR, e.g., FcγRI/his, was coated in the wells. A preformed immune-complex mixture was added to wells and incubated at room temperature for 1–2 h. The preforming immune-complex reaction contained 60 ng/mL streptavidin-HRP, 60 ng/mL of biotinylated F(ab′)2 goat anti-human IgG (F(ab′)2) (Jackson ImmunoRes, USA), and 1 µg/mL of BGB-A317 or BGB-A317/IgG4S228P or human IgG (huIgG) in the blocking buffer. ELISA-binding signals were detected by Immobilon chemiluminescence substrate A/B (Millipore).
Generation of M2 macrophages
Generation of M2 macrophages was performed according to the protocol described by Leidi et al. [
15]. Briefly, human PBMCs were co-cultured in 6-well plates or 100-mm culture dish (Corning) in complete RPMI1640 media supplemented with 30 ng/ml human M-CSF (R&D systems) for 4 days. Adherent cells were retained by gently washing off non- and loose-adherent cells, with half of media replaced, and culture for 2–3 more days. For M2 polarization, 10 ng/ml IL-10 (Peprotech) was added during the last 48 h of culture.
ADCP
HuT78/PD-1 cells were labeled with carboxyfluorescein succinimidyl ester (CFSE) (Life technologies) according to the manufacturer’s instructions. M2 macrophages were detached by Accutase™ and plated in V-bottom 96-well plates with CFSE-labeled HuT78/PD-1 at a ratio of 2:1 (HuT78/PD-1 cell to macrophage) in the presence of PD-1 mAbs for 5 h at 37 °C. After co-culturing, cells were stained with anti-CD11b-APC and subjected to flow cytometry. The percentage of macrophages that underwent ADCP of HuT78/PD-1 was determined by FACS of double positive cells (CFSE+ and CD11b+) following gating on CD11b+ M2 macrophages.
FACS analysis
FACS analysis was performed using Guava® easyCyte 6HT or 8HT (Millipore Merck). For the cell-based-binding assay, FcγRI-transfected HEK293 cells (HEK293/FcγRI) were stained with BGB-A317 or BGB-A317/IgG4S228P or huIgG, followed by detection with AlexaFluor 488-conjugated goat F(ab′)2 anti-human IgG (F(ab′)2) fragment (Jackson ImmunoRes). The cell surface binding signals were quantified as mean fluorescence intensities (MFIs).
For the analysis of tumor infiltrated immune cells from the mouse in vivo cancer models, tumor tissue was cut into small pieces and digested with collagenase type I (1 mg/ml, Sigma) and 100 µg/ml DNase I (Sigma) in RPMI1640 plus 5% FBS for 30 min at 37 °C. Single cell suspension was obtained after passing the digested tissues through a 70 µm cell strainer. The cells were then washed and blocked by human IgG, followed by staining with human CD3 (HIT3a, 4A biotech), CD8 (OKT8, Sungene biotech), PD-1 (MIH4, eBioscience), CD64 (10.1, eBioscience), or mouse CD64 (X54-5/7.1, Biolegend) antibodies at 4 °C. The stained samples were subjected to flow cytometry and analysis using guavaSoft3.1.1 (Millipore, Merck).
In vivo efficacy study
NOD/SCID mice (purchased from Vital River) were pre-treated with cyclophosphamide (150 mg/kg, J&K) intraperitoneally (i.p.) once a day for 2 days. One day after the second dose, animals were injected subcutaneously (s.c.) with 2.5 × 106 A431 cells (ATCC) and 5 × 106 PBMCs (a total of 200 µl cell mixture in 50% matrigel) in the right front flank. Starting from day 0 after cell inoculation, animals were randomly grouped and then treated as indicated. Primary tumor volume was measured twice every week, using a caliper. All experiments were conducted based on the protocols approved by the Animal Care and Use Committee of BeiGene according to the guidelines of the Chinese Association for Laboratory Animal Sciences.
Pharmacokinetics analysis of BGB-A317 and BGB-A317/IgG4S228P
Mouse blood samples were collected from retro-orbital sinus at indicated time points. The concentration of BGB-A317 or BGB-A317/IgG4S228P in serum was determined by ELISA. Briefly, serum samples were added to PD-1/His protein (2 µg/ml)-coated ELISA plates (Nunc), followed by the capture with anti-huIgG-HRP (Sigma) and color development. OD values at 450 nm were detected by a Microplate Reader (SpectraMax® Paradigm®). The results were analyzed with SoftMax Pro software (Molecular Devices).
IHC and immunofluoresence staining
Tumor tissues were harvested and fixed in formalin, dehydrated, embedded in paraffin, sectioned at 3 µm, and placed on polylysine-coated slides. The sections were deparaffinized in xylene and rehydrated in graded ethanol. Antigen retrieval was performed in citrate buffer (pH 6.0) by boiling for 30 min in a microwave and cooling down to room temperature. Then, the sections were blocked by 3% bovine serum albumin in PBS for 1 h and 0.3% H2O2 solution in PBS for 10 min, and afterwards, stained by the antibodies against human CD8 (SP16, ZSGB-Bio), PD-1 (NAT105, Abcam), CD64 (3D3, Abcam), and mouse CD64 (Clone 027, Sino Biological) at 4 °C overnight. The antibodies were detected by HRP conjugated second antibodies and DAB. The immunofluorescence staining was performed using Opal™ 4 or 7-color immunofluorescence staining kits (PerkinElmer). The images were acquired on the Vectra System and were analyzed using inForm software (both from PerkinElmer).
Statistical analysis
Student’s t test was used to analyze differences between groups. P < 0.05 was considered statistically significant. Statistical analysis was done by GraphPad Prism software (GraphPad, La Jolla, CA).
Discussion
In this report, we showed that the anti-PD-1 mAb with IgG4
S228P retained high affinity binding to FcγRI and mediated crosslinking of PD-1 and FcγRI, which brought PD-1
+ T cells and FcγRI
+ macrophages together and induced profound changes in intracellular signaling and biological activities, including the inhibition of PD-1
+ T-cell functions, induction of ADCP and IL-10 secretion by macrophages. It was well documented that macrophages were present with high frequency and abundance in some cancer types [
26]. The high density of macrophages in the TME promote tumor progression, metastasis and resistance to therapies [
27‐
29] and are associated with poor prognosis [
30]. In addition, a significant portion of macrophages have been found to be in close contact with T cells in many human tumors [
31,
32]. Thus, the crosslinking of PD-1 and FcγRI receptors by an anti-PD-1 mAb with IgG4
S228P will very likely occur in TME.
PD-1 is highly expressed in T effector cells (T
effs) [
33]. Immunotherapy with anti-PD-1 antibody is thought to protect PD-1
+ T effector cells from PD-1 ligand engagement and its inhibitory effects. Therefore, to avoid possible killing of PD-1
+ T
effs through ADCC and CDC, it is desirable to have the effector functions of PD-1 mAb removed. Dahan et al. using FcγR-null mouse cancer models demonstrated that the anti-tumor efficacy of anti-PD-1 mIgG2a mAbs is significantly better in FcγR-null than in FcγR-competent mice. Human IgG4 antibodies usually do not induce significant ADCC and CDC, since IgG4 binds to FcγRIII with very low affinity and lacks binding to C1q [
10,
34,
35]. However, IgG4 antibodies still retain high affinity to FcγRI as shown by our SPR assay and others [
10]. In line with these findings, therapeutic IgG4 antibodies have been shown to deplete target cells in both humans and a humanized mouse model [
36,
37]. Our in vitro assay showed that BGB-A317/IgG4
S228P induced significant ADCP than BGB-A317, raising the possibility that FcγRI binding might lead to the killing of PD-1
+ TILs by TAMs and inferior anti-tumor activity. In fact, reduced anti-tumor efficacy was observed in IgG4
S228P-treated groups as compared with BGB-A317 treatment in a xenograft model. Moreover, fewer numbers of CD8
+ and PD-1
+ T cells within tumors were seen after BGB-A317/IgG4
S228P treatment, which was also consistent with the presence of high density of murine CD64
+ cells within tumors. The results of this study clearly demonstrated that anti-PD-1 antibodies with or without FcγRI-mediated effector functions exert dramatically different pharmacodynamics effects in anti-cancer activity when there is a high density of FcγRI
+ cells (primarily macrophages) present in TME.
There is growing evidence suggesting that FcγRI signaling in macrophages may mediate anti-inflammatory effects. Previous studies have shown that ligation of FcγRI on macrophages could promote the production of the anti-inflammatory cytokine IL-10 and dampen the responses to IFN-γ [
19,
38]. In addition, FcγRI is critical for TGFβ2-treated macrophage-induced tolerance and plays an important role in IgG4-induced M2 macrophage generation [
20,
39]. In a comparative study on the activation of IL-10 production in macrophages, IgG4 has shown similar potency as IgG1 [
40].
This study clearly showed that anti-PD-1 antibody with FcγRI-binding activity had significantly reduced anti-tumor efficacy in the NOD/SCID mouse cancer model with allogenic xenograft of cancer cells (human epidermoid cell line, A431) and human PBMCs. NOD/SCID mice are devoid of T, B cells and IgGs, but have myeloid-derived lineage cells such as macrophages [
41]. Therefore, macrophage FcγRI-mediated ADCP could efficiently eliminate PD-1
+ T cells targeted by a regular PD-1 antibody [
42] (also see Fig.
5). In clinic, both nivolumab and pembrolizumab showed clinical efficacy in multiple cancer types [
43]. However, the response rate is typically less than 30%. Lack of T-cell infiltration, PD-L1 expression and presence of other check-point molecules were attributed to the lack of responses in anti-PD-1 Ab therapy in most patients [
43]. We previously compared the efficacy of anti-PD-1 antibodies (BGB-A317, nivolumab and pembrolizumab) using tumor samples from colorectal liver metastasis (CLM) patients [
44]. BGB-A317 demonstrated better activation of TILs from CLM tumors where macrophages were more abundant. All these studies suggested that beyond the reasons described above, prevalence of FcγRI
+ macrophages infiltration in the TME might also play negative roles in anti-PD-1 antibody (such as nivolumab and pembrolizumab) treatment. Arlauckas et al. recently showed in a mouse model that anti-PD-1 antibody could be transferred from PD-1
+ T cells to macrophages via FcγR-dependent manner [
45], supporting our findings that tumor-associated macrophages and FcγRI can negatively impact on anti-PD-1 antibody-mediated anti-tumor efficacy.
In summary, our study demonstrated that an anti-PD-1 IgG4S228P antibody could mediate crosslinking between PD-1+ T cells and FcγRI+ macrophages, resulting in macrophage-mediated phagocytosis of PD-1+ T cells, conversion of PD-1 blockade to activating and induction of IL-10 gene expression, and, therefore, dampening T-cell-mediated immune responses. Together, the ability to bind FcγRI by an anti-PD-1 mAb could significantly impair its anti-tumor activity, especially in the TME where macrophages are highly enriched. However, such negative impact on anti-tumor activity should be limited if the effector functions of an anti-PD-1 antibody is removed. Further clinical studies on PD-1 therapy may shed more light on the issue.