Background
CXCR4 (chemokine C-X-C motif receptor 4), also known as CD184, is a chemokine G protein coupled receptor [
1], expressed in different cell types, including normal B cells [
2‐
6]. CXCR4 is overexpressed in a variety of cancers including chronic lymphocytic leukemia (CLL), acute myeloid leukemia (AML), myeloma, lymphomas, and solid tumors [
7]. CXCL12 (chemokine C-X-C motif ligand 12), also known as stromal cell-derived factor 1 (SDF-1), is CXCR4 sole ligand. It is a homeostatic chemokine [
8], highly expressed in the lymph nodes, bone marrow (BM), liver, and lung [
9]. CXCL12 regulates hematopoietic cell trafficking and their homing to the BM [
5]. Chemotaxis driven by CXCR4 and CXCL12 interactions has been shown to control various biological functions including cell adhesion, migration, and invasion [
10].
CLL is the most prevalent adult leukemia and is characterized by accumulation of dysfunctional B-lymphocytes in the lymph nodes and BM [
11]. Stromal cells secrete CXCL12 and promote B-cell progenitors and CLL cell survival through CXCR4 signaling [
8,
12,
13]. Thus, activation of the CXCL12/CXCR4 axis plays an important role in stromal cell-dependent resistance to therapy in CLL patients, including cytotoxic drugs [
4] or steroids [
14], thereby promoting minimal residual disease [
15]. These observations support the rationale for targeting CXCR4 for the treatment of CLL.
In the last decade, a number of agents targeting CXCR4 have been developed. These include small molecules, peptides, and monoclonal antibodies. The role of CXCR4 in hematopoietic stem cell (HSC) retention and trafficking led to the development of agents used for HSC mobilization. AMD3100 (Plerixafor), a small molecule inhibitor of CXCR4, was approved for mobilization of HSCs prior to autologous transplantation; however, this compound has limited application for sustained treatment due to toxicity [
16,
17]. BL-8040 (BKT140), a peptide inhibitor of CXCR4, has robust cell mobilization capacity [
18,
19], similarly to other CXCR4-specific antagonist peptides (T-140, TN-14003, TC-14012), which were shown to inhibit CXCR4-CXCL12 signaling in CLL-B cells [
4]. However, these peptides show limited in vivo exposures. Recently, two CXCR4 human IgG4 antagonist antibodies, ulocuplumab [
20‐
22] and LY2624587 [
23], were described. Ulocuplumab is currently in phase 1 clinical studies, and it was shown to have prolonged pharmacokinetic exposure compared to small molecules or peptide inhibitors [
20‐
22].
PF-06747143 is a novel and potentially first in class humanized IgG1 anti-CXCR4 antibody that recently entered into clinical studies (NCT02954653). Here, we show that it potently binds to CXCR4 and inhibits CXCL12-driven calcium flux. Moreover, it induces cell death in malignant CLL-B cells via two main mechanisms of action: (1) bivalency-dependent mechanism, involving generation of reactive oxygen species (ROS) and independent of caspases and (2) Fc region-driven cytotoxicity, including complement-dependent cytotoxicity (CDC) and antibody-dependent cell-mediated cytotoxicity (ADCC) activity. Importantly, we show that PF-06747143 triggers cell death in B-CLL patient-derived primary leukemia cells, in spite of the presence of stromal cells, mimicking the leukemia microenvironment in vitro. The antibody also synergizes with conventional CLL treatment agents such as bendamustine, rituximab, fludarabine (F-ara-A), and ibrutinib, significantly improving their cytotoxicity in combination. Furthermore, we show that PF-06747143 inhibits tumor burden and improves survival as a monotherapy or in combination with bendamustine, in a CLL xenograft tumor model. Based on these unique mechanisms of action, PF-06747143 has a promising therapeutic potential in CLL patients and other hematological malignancies dependent on the CXCR4 axis.
Methods
Isolation of PBMCs from CLL patients
The CLL-B cells were collected from blood samples at the Moores-UCSD Cancer Center in compliance with the Declaration of Helsinki and after approval of the UC San Diego Institutional Review Board (IRB) [
24].Peripheral blood mononuclear cells (PBMC) from CLL patients were isolated using Ficoll-Hypaque gradient density centrifugation (Cat# 17-1440-03, GE Healthcare Life Science). For caspase activation assays, the CLL-B cells were purified by positive selection using Dynabeads CD19 pan B (Cat# 11143D, Invitrogen) and DETACHaBEAD CD19 (Cat# 12506D, Invitrogen) according to the manufacturer’s protocol. For the other assays, fresh or frozen PBMCs were used and cells were stained with CD19/CD5 antibodies for detection of double positive CLL-B cells.
CLL-B cells co-culture to mimic CLL microenvironment
Primary leukemic cells from CLL patients were cultured in RPMI supplemented with 10% heat-inactivated FBS (fetal bovine serum, Catalog # FB-02, Omega Scientific, Tarzana, CA) and 1% antibiotic at a density of 3 × 10
5 cells per milliliter at 37 °C and 5% CO
2. The cells were either cultured in 96-well round bottom plates (Catalog # 3596, Corning, NY) alone or co-cultured with NK-tert stromal cells (RIKEN, Yokohama, Japan) at a ratio of 20:1 (CLL: stroma-NK-tert) in RPMI with 1% Penn-Strep and 10% FBS [
25].
CXCR4 expression by flow cytometry
The CXCR4 phenotyping of CLL-B, stroma-NK-tert, normal B, and T cells was done by flow cytometry using a 1:50 dilution with rat anti-human CD184 (CXCR4) PE Mab (Catalog # 551966, clone:2B11, BD Biosciences). The isotype control antibody was PE Rat IgG2b, κ (Catalog # 12-4031-83, clone: eB149/10H5, eBioscience).
CXCR4 antibody generation
The parental CXCR4 antagonist antibody, m15, was derived from immunization of Balb/c mice with CHO cells transfected with human CXCR4. The heavy and light chain variable domains of m15 were then cloned into human IgG1 or hinge stabilized IgG4 and light κ backbone, to generate chimeric m15-IgG1 and m15-IgG4. m15 was subsequently humanized by CDR grafting/affinity maturation and cloned into human IgG1/κ constant domains to create PF-06747143.
Binding kinetics and affinity
Experiments were performed on a BiacoreTM T200 surface plasmon resonance biosensor (GE Life Sciences). The binding to human CXCR4 was determined using human CXCR4-enriched lipoparticles (Integral Molecular) compared to null particles. Lipoparticles were diluted into 10 mM HEPES, 150 mM NaCl, 1 mg/mL BSA, pH 7.4 buffer to concentrations between 0.015 to 0.04 units/mL and captured for 5 min onto flow cells. A threefold dilution series of Fab was evaluated and dissociation was monitored for 10 min. The data were fit to a 1:1 Langmuir with mass transport model using Biacore T200 Evaluation Software Version 2.0.
PF-06747143 binding to tumor cells by flow cytometry
Cell suspensions (n = 3/group) were stained with 20 μg/mL of either a human IgG1 ĸ Phycoerthrin (PE)-labeled antibody (isotype control) (Southern Biotech) or with PF-06747143 PE-conjugated antibody, labeled using the SiteClick™ R-PE Kit (Molecular Probes, Life Technologies). Flow cytometric acquisition and analysis was conducted using FACS LSRII™ flow cytometer (Beckman Dickinson).
Calcium flux functional assay
The ability of PF-06747143 or m15-IgG1 to inhibit CXCL12-induced calcium flux was evaluated in human T cell leukemia Jurkat cells using the Fluo-NW Calcium assay kit (Life Technologies). Cells were plated in 384-well plates at 70,000 cells per well in quadruplicates and incubated with m15-IgG1 parent antibody and PF-06747143, upon stimulation with CXCL12 at 8 nM (EC80) (Invitrogen), for 110 min. Calcium flux was then measured for 95 s using a FLIPR Tetra (Molecular Devices).
Cell death
Cell death was evaluated by flow cytometry analysis using CD19/CD5/Annexin V antibodies [
26]. Specific induced cell death (SICD) calculation was used in order to discriminate the antibody/compound-specific induced cell death from background or spontaneous cell death observed in the vehicle-treated groups. The calculation of % SICD was performed using the following formula: % SICD = (Compound-induced cell death − Vehicle spontaneous cell death)/(100 − Vehicle spontaneous cell death) × 100.
Cell death in combination with CLL standard of care agents
m15-IgG1 was tested in combination with different standard of care (SOC) agents currently used for treatment of CLL. F-ara-A, bendamustine, rituximab, and ibrutinib were evaluated in combination with 200 nM of m15-IgG1. CLL-B cells were treated for 48 h at 37 °C either cultured alone or co-cultured with stroma-NK-tert cells. The combination data and level of synergism was analyzed using CompuSyn software (ComboSyn, Inc., NJ, USA). The data derived from this analysis were expressed as combination index (CI), which offers definition for additive (CI = 1), synergism (CI < 1), and antagonism (CI > 1) in drug combination [
27].
Antibody-dependent cellular cytotoxicity (ADCC) assay
For analysis of ADCC in B-CLL patient primary cells, the ADCC Reporter Bioassay kit from Promega (Catalog #G7010) was used, per instructions from the manufacturer. The ADCC Reporter Bioassay uses engineered Jurkat cells stably expressing the FcγRIIIa receptor, V158 (high affinity) variant, and a NFAT (nuclear factor of activated T cells) pathway response element driving expression of firefly luciferase as effector cells. The transfected Jurkat cell line was grown in RPMI containing G-418 sulfate solution (Catalog # V8091) and hygromycin (Catalog # 10687010, 50 mg/mL solution). The ADCC buffer (99.5% RPMI 1640 with L-glutamine and 0.5% super low IgG FBS) was prepared using RPMI supplemented with super low IgG defined fetal bovine serum (catalog # SH30898, Hyclone). The luciferase assay system was used as a readout (Catalog # G7940, Promega). Different concentrations of antibodies IgG1 control, PF-06747143, rituximab and obinutuzumab were added to the effector/target cell 1:1 ratio mixtures. A total of 75,000 for effector and target cells were incubated for 6 h at 37 °C in a humidified CO2 incubator. Following incubation, the plate was equilibrated to ambient temperature for 15 min. Bio-Glo™ luciferase assay reagent was added and incubated at room temperature for 30 min. The luminescence was detected using an Infinite 200 Microplate Reader (Tekan), and the results are expressed in relative light units (RLU).
ADCC activity of PF-06747143 and m15-IgG1 parent antibody was evaluated in JVM-13 CLL tumor cell line, in presence of the NK-92 FcγRIIIA 158V (NK92 158V) cell line as effector cells (Conkwest). Antibodies were incubated for 4 h, with tumor cells and effector cells (1:10 ratio) (n = 4/group). ToxiLight bioluminescent cytotoxicity assay (Cat # LT07-117 Lonza) was used to detect cell lysis.
CDC assay
PF-06747143 was added to CLL-B cells (1 × 10
6/mL) in RPMI media with 5% active human serum [
28,
29] or inactivated human serum, which was incubated at 56 °C for 30 min. The heat-inactivated/normal human serum-treated cells were incubated for 4 h at 37 °C with increasing concentrations of PF-06747143. Cytotoxicity was determined by flow cytometry using CD19/CD5/Annexin V staining. % SICD was calculated according to the following formula: 100 × (% viable cells with inactivated serum − % viable cells with native serum)/(% viable cells with inactivated serum).
Inhibition of actin polymerization
Cytoskeletal reorganization (F-actin polymerization) was evaluated in CLL samples activated by CXCL12 and treated with PF-06747143 or control agents [
4].
Inhibition of migration of cells in a transwell assay
PF-06747143 was assessed for its ability to inhibit CXCL12-induced chemotaxis in primary CLL-B cells derived from CLL patients using a transwell migration assay [
30].
Caspase activity assay
To evaluate the mechanism of cell death induced by PF-06747143, CLL-B cells were purified from patient-derived PBMCs and tested for caspase activation including caspases 3, 8, and 9 using the ApoTarget Caspase Colorimetric Protease Assay Sampler kit (Cat # KHZ1001, Invitrogen, Frederick, MD) according to the manufacturer’s instructions. Z-VAD-FMK, a caspase inhibitor (Cat # G7231, Promega Corporation), was used as control [
31].
Detection of reactive oxygen species (ROS) by flow cytometry
CLL-B cells were seeded at 2.5 × 10
5/mL in RPMI media and treated with antibodies for 4 h at 37 °C and 5% CO
2 in 24-well plates. The generation of ROS was detected using dihydroethidium (HE) staining (Catalog # D1168, Sigma-Aldrich, St. Louis, MO) as described previously [
32]. The samples were then analyzed by flow cytometry followed by data analysis using FlowJo software.
In vivo efficacy study
JVM-13 tumor cell line [
33,
34], purchased from ATCC, was stably transfected with the luciferase gene. The cells were cultured in RPMI media with 10% FBS. To establish a JVM-13 disseminated model, 1 × 10
6 cells per mouse were implanted via tail vein injection in female SCID beige mice (Charles River). Tumor burden was monitored via bioluminescence imaging (BLI) (IVISÒ 200) throughout the study. When the tumor burden (mean BLI) reached 7.2 × 10
6 photons/s, on day 19, mice were randomly assigned into four groups and treated with (1) IgG1 negative control Ab or (2) PF-06747143, dosed subcutaneously at 10 mg/kg, once a week, for a total of 6 doses; (3) bendamustine, dosed intraperitoneally at 30 mg/kg, on days 19 and 20, followed by another 2-day treatment cycle 28 days later; and (4) combination of PF-06747143 and bendamustine. Mice were euthanized according to the IACUC guidelines once they developed disease-related symptoms such as hind leg paralysis.
Statistical analysis
The statistical analysis was carried out using GraphPad Prism software (v. 5.0c; San Diego, CA). The statistical differences for the mean values were analyzed using one way ANOVA and are indicated with *, p < 0.05; **, p < 0.01; ***, p < 0.001; and ****, p < 0.0001. Tumor model survival analysis was performed using Kaplan-Meier followed by a long-rank (Mantel-Cox) test.
Discussion
CXCR4 overexpression was shown to correlate with poor prognosis in CLL patients [
23]. Activation of CXCR4 induces cell trafficking and homing of malignant cells to the BM and lymph nodes [
49,
50], where CXCL12 is highly expressed, leading to retention of these cells in a microenvironment that provides growth signals, induces proliferation, and contributes to drug resistance, leading to poor prognosis and relapse [
51]. In this study, we assessed the effect of inhibition of the CXCR4-CXCL12 pathway by a novel CXCR4 antagonist IgG1 antibody, PF-06747143, and its parental antibody m15-IgG1.
We showed that the surface expression of CXCR4 is at least tenfold higher in CLL patients than in normal T and B cells, in concordance with previous studies [
52‐
54]. Similar CXCR4 expression levels were observed in high-risk or low-risk prognostic CLL patients. Importantly, both CLL risk groups had comparable sensitivity to CXCR4 antibody-induced cell death, while normal T and B cells, or the CXCR4-negative MEC-1 cell line, were largely spared. This indicates that the cell death mechanism is dependent on the level of CXCR4 surface expression. Notably, cells from CLL patients bearing the
TP53mut/Del(17p) genotype, which is known to be resistant to F-ara-A therapy, were significantly sensitive to cell death induced by CXCR4 antibody treatment. This is of particular relevance in the treatment of refractory patients, who, in general, present with abnormal
TP53 status [
55]. Furthermore, the similar levels of cell death observed in CLL-B cells cultured in presence or absence of stromal cell support suggest that the CXCR4 antibody has the potential to overcome the protection provided by the microenvironment. These findings may have significant clinical implications in the treatment of patients presenting with minimal residual disease in the BM and lymph nodes.
The CXCR4 antibody significantly synergized with CLL SOC agents (rituximab, ibrutinib, F-ara-A, and bendamustine), by increasing CLL cell death. Furthermore, we demonstrated that PF-06747143 improves survival as a monotherapy and its activity is increased when combined with bendamustine in the JVM-13 mouse xenograft disseminated staged model. These findings illustrate the potential of PF-06747143 to be used in combination with these agents in the clinic.
PF-06747143 blocked CXCL12-induced cytoskeletal changes and migration of primary CLL-B cells. Although to a lesser extent, these events were also inhibited by AMD3100, the CXCR4 small molecule inhibitor. However, the lack of cell death activity shown by ADM3100 in this study, and by CXCR4 peptide inhibitors described in the literature [
4,
53,
56], suggests that just binding to CXCR4 and inhibiting CXCR4 signaling pathways is not sufficient to trigger this phenomenon. These differences may be explained by the observation that PF-06747143 induction of cell death required antibody bivalency. Bivalent binding is a property inherent to antibodies, due to their two binding regions, which are not a characteristic of small molecules or peptides. Other anti-CXCR4 antibodies, including ulocuplumab, LY2624587, and hz515H7, were also shown to induce tumor cell death upon binding to CXCR4 [
23,
46,
57]; however, the role of antibody bivalency in this process was not described in these studies.
In further characterizing the mechanisms involved in the cell death triggered by PF-06747143, we demonstrated that this process did not involve caspase activation. The PF-06747143 caspase-independent cell death mechanism is similar to that described for ulocuplumab, an IgG4 CXCR4 antibody [
46].
Recently, antibodies binding to CD20, CD74, CD47, or HLA-DR have been shown to directly induce programmed cell death (PCD), without involvement of caspases or the need for hyper-cross-linking of the antibody [
32,
58]. This novel cell death mechanism is dependent on homotypic cell adhesion triggered upon antibody binding, followed by actin redistribution and ROS activation [
59,
60]. Although it remains unclear which proximal events trigger PCD, the process results in loss of plasma membrane integrity and non-apoptotic cell death. A role for antibody bivalency in PCD has not been previously described; however, our results suggest that antibody bivalency might play a role in the initiating steps, by inducing homotypic cell-cell adhesion through binding to antigens expressed in adjacent cells simultaneously. We have also shown that PF-06747143 treatment generated ROS production in CLL-B cells, in association with cell death. The pattern of ROS production and cell death induced by PF-06747143 was similar to that observed with other antibodies shown to induce ROS-dependent cell death, such as CD20 (obinutuzumab), TAG-A1 [
61], and 1D10 [
62]. Taken together, our results suggest that upon binding to CXCR4 receptors in a bivalent manner, PF-06747143 triggers cell death through a caspase-independent and bivalency-dependent mechanism, similar to PCD.
In addition to signaling blockade and induction of bivalency-dependent cell death, PF-06747143 showed potent Fc effector-mediated ADCC and CDC activity in CLL-B cells. Of note, PF-06747143 and m15-IgG1 ADCC activity was significantly greater than that of the m15-IgG4 antibody in the ADCC assay. Similarly to m15-IgG4, ulocuplumab, which is a human IgG4 CXCR4 antibody, was recently reported to have no Fc-driven cytotoxic activity [
46,
57]. The lack of Fc effector function-driven cytotoxicity observed for the IgG4 antibodies is expected, based on the human IgG4 inherently lower affinity for the proteins involved in the process [
63]. We also showed that PF-06747143 cytotoxic activity was comparable to that of obinutuzumab and rituximab [
64]. The importance of ADCC or CDC activity has been clinically demonstrated for obinutuzumab and rituximab, as well as other antibodies successful therapeutic IgG1 antibodies approved for the treatment of cancers, including alemtuzumab (CD52), trastuzumab (HER2), cetuximab (EGFR), and daratumumab (CD38) [
48,
65].
Therapeutic CXCR4 antagonists currently available lack desirable potency, cytotoxicity, safety, or adequate exposure for prolonged treatment. The small molecule AMD3100 approved for stem cell mobilization in autologous transplantation has significant safety issues that limit its chronic use [
16]. In addition, we showed that AMD3100 does not induce significant cell death in B-CLL cells. Peptide antagonists of CXCR4 have been recently evaluated in clinical trials as mobilizing agents. LY2510924 was evaluated as a single agent in a phase 1 dose escalation trial in advanced metastatic cancers [
66], and BKT140 (BL8040/TN14003) was evaluated in a phase 1 clinical trial in multiple myeloma (MM) patients [
19]. Treatment with both peptides induced rapid mobilization of stem cells but failed to reduce tumor burden. In addition, as for peptide therapeutics in general, they had short half-lives and required frequent administration, which makes peptides a challenging modality for sustained treatment [
19,
66,
67]. The CXCR4 antagonist humanized IgG4 antibody, ulocuplumab, was recently evaluated in phase 1 clinical trials [
20‐
22] and, as expected for an antibody, it exhibited a longer half-life than that of small molecules and peptides. In contrast to small molecule and peptide inhibitors of CXCR4, ulocuplumab, as well as another IgG4 CXCR4 antibody, LY26245587, and an IgG1 CXCR4 antibody hz515H7 [
68], can induce tumor cell death via a mechanism reminiscent of PCD, similarly to PF-06747143.
However, the IgG4 antibodies do not induce tumor cell death via ADCC or CDC [
23,
46], as expected for human IgG4 antibodies. PF-06747143 and hz515H7 are the first IgG1 CXCR4 antibodies to be described. A key role for Fc effector functions ADCC and CDC was demonstrated when a mutation in the Fc region of hz515H7, abrogating the Fc effector function, resulted in significantly decreased efficacy in a mouse tumor model [
68]. Taken together, these data suggest that the CXCR4 antibody Fc effector cytotoxic functions, ADCC and CDC, play a key role in vivo and they may contribute to efficacy enhancement in the clinic.
Conclusions
The novel CXCR4 antagonist IgG1 antibody PF-06747143 binds to CXCR4 with high affinity and blocks CXCL12-induced mechanisms including calcium flux and cell migration. The CXCR4 antibody mediates CLL-B cell death via a bivalency-dependent mechanism, involving generation of reactive oxygen species (ROS), with no caspase activation requirement, reminiscent of PCD. Moreover, PF-06747143 induces B-CLL cell death regardless of patient prognostic risk factor or the presence of stromal cells, indicating that it may be valuable in the treatment of resistant disease. PF-06747143 synergizes with CLL SOC agents such as bendamustine, rituximab, fludarabine, and ibrutinib. Moreover, in a CLL xenograft tumor model, PF-06747143 causes tumor growth inhibition, with increased survival, both as a monotherapy and in combination with bendamustine. Differently from other CXCR4 antagonists in the clinic, PF-06747143 induces potent cell death via Fc-driven cytotoxicity, through ADCC and CDC. Taken together, our data support the development of PF-06747143 for the treatment of CLL patients.
Acknowledgements
Not applicable.