CD19 is a pan B-cell surface receptor expressed from pro–B-cell development until its down-regulation during terminal differentiation into plasma cells. CD19 represents an attractive immunotherapy target for cancers of lymphoid origin due to its high expression levels on the vast majority of non–Hodgkin's lymphomas and some leukemias. A humanized anti-CD19 antibody with an engineered Fc domain (XmAb5574) was generated to increase binding to Fcγ receptors on immune cells and thus increase Fc-mediated effector functions. In vitro, XmAb5574 enhanced antibody-dependent cell-mediated cytotoxicity 100-fold to 1,000-fold relative to an anti-CD19 IgG1 analogue against a broad range of B-lymphoma and leukemia cell lines. Furthermore, XmAb5574 conferred antibody-dependent cell-mediated cytotoxicity against patient-derived acute lymphoblastic leukemia and mantle cell lymphoma cells, whereas the IgG1 analogue was inactive. XmAb5574 also increased antibody-dependent cellular phagocytosis and apoptosis. In vivo, XmAb5574 significantly inhibited lymphoma growth in prophylactic and established mouse xenograft models, and showed more potent antitumor activity than its IgG1 analogue. Comparisons with a variant incapable of Fcγ receptor binding showed that engagement of these receptors is critical for optimal antitumor efficacy. These results suggest that XmAb5574 exhibits potent tumor cytotoxicity via direct and indirect effector functions and thus warrants clinical evaluation as an immunotherapeutic for CD19+ hematologic malignancies. [Cancer Res 2008;68(19):8049–57]
CD19 is a 95-kDa transmembrane glycoprotein of the immunoglobulin superfamily containing two extracellular immunoglobulin-like domains and an extensive cytoplasmic tail (1). It is B-lineage–specific and functions as a positive regulator of B-cell receptor signaling in conjunction with CD21 and CD81. CD19 is an attractive target for cancers of lymphoid origin due to its high expression in most non–Hodgkin's lymphomas (NHL) and many leukemias, including acute lymphoblastic leukemia (ALL), chronic lymphocytic leukemia (CLL), and hairy cell leukemia (HCL; refs. 2, 3). NHL is the most prevalent of all lymphoproliferative diseases, and 85% of NHLs are classified as B-cell disorders (4). In the last decade, many B-NHLs and some B-cell leukemias have been successfully treated by combining chemotherapy with rituximab, a chimeric anti-CD20 antibody, demonstrating the utility of immunotherapies in B-cell diseases. CD19 cell surface expression is lower relative to CD20, but it begins earlier and persists longer through B-cell maturation (1). Consequently, the spectrum of lymphoid malignancies expressing CD19 is broader (2). Moreover, although the chemotherapy-rituximab regimen has led to major improvements in response rates and progression-free survival, some B-cell tumors either lack CD20 expression or lose it during the course of rituximab treatment (5, 6). Consequently, some patients do not respond to rituximab, and many of those that respond will relapse (7). Thus, anti-CD19 antibodies may be efficacious against early B-cell malignancies such as ALL or other CD20− tumors, which are not amenable to treatment with rituximab, and as salvage regimens for patients failing rituximab.
CD19 has been a focus of immunotherapy development for over 20 years, and several CD19-specific antibodies have been evaluated for the treatment of B-lineage malignancies in vitro, in mouse models, and in clinical trials. These have included unmodified anti-CD19 antibodies (8, 9), antibody-drug conjugates (10–12), and bispecific antibodies targeting CD19 and CD3 (13) or CD16 (14) to engage cytotoxic lymphocyte effector functions. Although early clinical studies with murine unconjugated CD19 antibodies showed safety and responses in individual patients, these responses were not durable (15). However, recent phase I trials with bispecific antibodies have yielded encouraging results (16). These clinical studies as well as trials with anti-CD19 immunoconjugates have validated CD19 as a target for B-cell malignancies, and the results with bispecific antibodies indicate that enhancing the engagement of immune effector functions may improve therapeutic efficacy.
Immune effector functions which have been shown to contribute to antibody-mediated cytotoxicity include antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cell-mediated phagocytosis (ADCP), and complement-dependent cytotoxicity (CDC). Cytotoxicity may also be mediated via antiproliferative effects. The mechanism of antibody modulation of tumor cell proliferation is poorly understood. However, advances in understanding the interactions of antibodies with Fcγ receptors (FcγR) on immune effector cells have allowed the engineering of antibodies with significantly improved effector function (17). Protein engineering of the Fc domain has been used to improve binding to FcγRs (18–21), and antibodies with up to 100-fold greater affinity for FcγRIIIa relative to native IgG1 have been generated, leading to significant improvements in ADCC (19). These results, in conjunction with studies in mouse models demonstrating that anti-CD19 antibodies deplete B cells in an FcγR-dependent fashion (9, 22), suggested that engineering an anti-CD19 antibody to increase FcγR binding could enhance its cytotoxic potential.
Here, we describe the characterization of XmAb5574, a humanized anti-CD19 antibody with increased FcγR binding that has highly enhanced ADCC against multiple NHL and leukemia cell lines and primary ALL and mantle cell lymphoma (MCL) tumor cells. XmAb5574 also increases ADCP and antiproliferative activity, and inhibits lymphoma growth in Ramos and Raji mouse xenograft models. Therefore, XmAb5574 may represent a promising next-generation immunotherapeutic for B-cell malignancies.
Materials and Methods
Production of antibodies, FcγRs, and FcγRIIIa–glutathione-S-transferase. Variable region genes for mouse anti-CD19 antibody (clone 4G7; ref. 23) or anti-CD22 antibody (RFB4) were ligated into the expression vector pcDNA3.1Zeo (Invitrogen) comprising the human light chain κ and heavy chain constant regions to produce constructs for the corresponding chimeric antibodies. To generate XmAb5574, the Fv of 4G7 was humanized (24) and affinity-matured using library design automation,3
M.J. Bernett, G.L. Moore, S. Karki, I. Vostiar, J.O. Richards, E. Pong, E.A. Zhukovsky, and J.R. Desjarlais, manuscript in preparation.
Cell lines, donors, and patients. SU-DHL-6, Bonna-12, BV-173, SUP-B15, RS4;11, Ramos, and Namalwa cell lines were obtained from DSMZ (German Collection of Microorganisms and Cell Lines). Raji was obtained from American Type Culture Collection, and WaC3CD5 was provided by John Byrd (Ohio State University, Columbus, OH).
Peripheral blood samples were obtained from healthy volunteers or from male patients with ALL or MCL expressing high levels of CD19. All protocols were performed with written informed consent in accordance with the Declaration of Helsinki and were approved by the Ethical Committee of the Christian-Albrechts-University, Campus Kiel. The percentages of tumor cells and CD19 expressions were determined using directly labeled antibodies and analyzed by fluorescence-activated cell sorting on a Cytomics FC500 (Beckman Coulter).
Internalization. XmAb5574, rituximab, or chimeric RFB4 antibodies were labeled with europium using the DELFIA europium-N1 ITC chelate kit (Perkin-Elmer). Raji cells were resuspended at 106 cells/mL in growth medium (RPMI 1640 medium supplemented with 10% fetal bovine serum; FBS) containing 4 nmol/L of europium-labeled antibody and incubated with mixing for 1 h at 4°C. Cells were washed in ice-cold PBS, resuspended in growth medium, and dispensed into 12-well plates. At the indicated times, samples were split into two aliquots; each was centrifuged and supernatant was saved. Pellets were washed in PBS to determine total cell-associated signal (surface bound plus internalized antibody) or in 0.5% acetic acid/125 mmol/L NaCl to determine internalized antibody, resuspended in PBS, mixed with DELFIA Enhancement Solution, and incubated for 1 h. Europium fluorescence was measured in 96-well plates using EnVision 2012 (Perkin-Elmer).
Fc receptor binding. Surface plasmon resonance (SPR) measurements were performed using a Biacore 3000 (Biacore). Antibodies were captured onto an immobilized protein A/G (Thermo Scientific) CM5 biosensor chip (Biacore) generated using a standard primary amine coupling protocol. All measurements were performed in 10 mmol/L Hepes/150 mmol/L NaCl/3 mmol/L EDTA/0.005% polysorbate-20; glycine buffer was used for protein A/G surface regeneration. Antibodies were immobilized on the protein A/G surface for 5 min at 1 μL/min. Fc receptors were injected over the antibody-bound surface for 2 min at 20 μL/min followed by a 2 to 3 min dissociation phase. Data were fit to a 1:1 binding model (Langmuir) using BIAevaluation software (Biacore). Binding curves corresponding to six FcγR concentration series were fitted individually. Kinetic variables were used to calculate the affinity constant (Ka).
CD19 binding. CD19 binding Kas were determined by employing a Raji cell–based competitive binding assay. Cells were centrifuged, washed with dilution buffer (PBS containing 0.1% bovine serum albumin and 0.1 mg/mL non–B-cell reactive IgG), and diluted to a density of 0.5 × 106 cells/mL. XmAb5574 was labeled with Alexa Fluor 647 protein labeling kit (Invitrogen) and the Ka was measured using a direct binding assay. Serial dilutions of labeled XmAb5574 were incubated with cells, washed, and analyzed on a FACSCanto II. Data were analyzed with a 1:1 binding model. Kas of unlabeled anti-CD19 antibodies were determined in a competition binding assay. Serial dilutions of antibodies (starting at 7.5 μg/mL) were mixed with 75 ng/mL of fluorescently labeled XmAb5574, incubated with cells for 1 h, and analyzed on a FACSCanto II. Data were analyzed by the simplified binding model of Cheng and Prusoff (26).
Cell lines ADCC: fluorescence. Isolated interleukin 2 (IL-2)–activated natural killer (NK) cells were used as effector cells. Human peripheral blood mononuclear cells (PBMC) were purified from leukopacks using a Ficoll gradient and NK cells were isolated using NK Cell Isolation kit II (Miltenyi Biotec). NK cells were incubated in RPMI 1640 medium supplemented with 10% FBS and 10 ng/mL of IL-2 at 37°C overnight. Tumor cell line target cells were seeded into 96-well plates at 10,000 cells/well and opsonized with antibodies in triplicate at the indicated final concentration for 30 min at room temperature. Effector cells at an effector to target (E/T) cell ratio of 5:1 were cocultured for 4 h and lactate dehydrogenase (LDH) activity was measured using the Cytotoxicity Detection kit (CytoTox-ONE, Promega). The percentage of maximal lysis was calculated as follows:
Primary tumor cells ADCC: 51Cr release. Fresh unstimulated PBMC from patients and healthy donors were isolated as described (27). Donor PBMC typically contained 60% CD3+ T cells, 10% to 20% CD56+ NK cells, and 10% CD14+ monocytes; PBMC samples from patients with ALL contained 95% tumor cells, and MCL bone aspirates contained 80% tumor cells. The viability of cells tested by trypan blue exclusion was higher than 95%. ADCC was measured with a 51Cr release assay as described (27), except that 100 μCi of 51Cr/106 cells was used, E/T cell ratio was 80:1, and maximal 51Cr release was determined with Triton X-100 (Sigma-Aldrich). The percentage of maximal lysis was calculated according to the equation above.
Phagocytosis. Macrophage ADCP was determined by flow cytometry as described (28), except that RS4;11 or SUP-B15 cells were used as targets and target cells were labeled with PKH67 (Sigma-Aldrich) or 5-(and-6)-carboxyfluorescein succinimidyl ester (Molecular Probes).
Proliferation and apoptosis. Proliferation and apoptosis assays were done using SU-DHL-6 cells in the presence of antibodies cross-linked by FcγRIIIa-GST. Reacti-Bind streptavidin-coated plates (Thermo Scientific) were coated with 0.5 μg/mL of biotinylated anti-GST antibody (U.S. Biologicals). FcγRIIIa-GST (1 μg/mL), antibodies at the indicated concentrations, and 6 × 103 SU-DHL-6 cells were added to a final volume of 100 μL in RPMI 1640/20% FBS. In the proliferation assay, samples were incubated for 72 h at 37°C, and viability (measured as relative ATP luminescence) was detected using the CellTiter-Glo Luminescent Cell Viability assay kit (Promega) on a Topcount luminometer (Perkin-Elmer).
Caspase activation and Annexin V staining assays were used to detect apoptosis. For the caspase activation assay, samples were incubated for 48 h and analyzed with the Homogeneous Caspases assay kit (Roche Diagnostics). Data were acquired on a Victor2 Multi-Label reader (Perkin-Elmer). For the Annexin V assay, cells were labeled with Annexin V-PE using Apoptosis Detection kit Plus (BioVision) and 5 μg/mL of 7-amino-actinomycin D (Invitrogen). Data were analyzed on a FACSCanto II (BD Biosciences).
Relative ATP luminescence (%), caspase activity (%), and Annexin V+ cells (%) were calculated as the ratio of signal elicited at a given antibody concentration to signal with no antibody present.
Lymphoma xenograft models. Prophylactic tumor studies were performed at, and according to procedures approved by, the Southern Research Institute's Institutional Animal Care and Use Committee (IACUC). Six- to 8-week-old female C.B-17 severe combined immunodeficient (SCID) mice (Charles River Laboratories) were implanted s.c. with either 107 Raji or 5 × 106 Ramos lymphoma cells. Mice were randomly placed into five groups (n = 10/group). On day 3 post-tumor cell implant and on days 6, 10, and 13, mice were injected i.p. with 1, 3, or 10 mg/kg of XmAb5574, 10 mg/kg of XmAb isotype control, or PBS. Palpable tumors were measured twice per week with calipers; tumor volumes were calculated as (length × width2) / 2.
Established tumor studies were performed at Xencor according to procedures approved by Xencor's IACUC. Nine-week-old female C.B-17 SCID mice (Taconic) were injected s.c. with 2.5 × 106 Ramos cells. On day 14 post-cell implantation, mice bearing tumors between 30 and 130 mm3 were placed into groups such that each group had the same mean tumor volume (82–84 mm3; n = 10/group). Mice were injected i.p. with 10 mg/kg antibodies or PBS twice per week for 2 weeks; tumors were measured twice per week with calipers and volumes calculated as π/6 (length × width × height).
Data analysis. For in vivo studies, differences between groups were calculated using the nonparametric Mann-Whitney U test. P < 0.05 was considered significant.
We previously reported that Fc-engineered variants of IgG1 optimized for FcγR affinity exhibit markedly enhanced ADCC in the context of several antibodies (anti-HER-2, anti-CD20, and anti-CD52; ref. 19). The Fc variant with S239D/I332E mutations was one of the most effective. Therefore, we applied these Fc mutations to generate a cytotoxic anti-CD19 antibody. The Fv domain from a mouse anti-CD19 antibody (clone 4G7; ref. 23) was humanized and affinity-matured, and combined with a human Fc domain containing the S239D/I332E mutations to increase FcγR binding and enhance ADCC. We refer to this Fc-engineered humanized anti-CD19 antibody as XmAb5574.
XmAb5574 displays increased FcγR binding affinity. To confirm that the S239D/I332E mutations enhance FcγR binding in the context of an anti-CD19 antibody, we determined the FcγR binding affinity of XmAb5574 and compared it to an IgG1 analogue (anti-CD19 IgG1), which contains the same Fv domain as XmAb5574, but has a native human IgG1 Fc. Kas were determined by measuring SPR on a Biacore biosensor (Fig. 1A). The S239D/I332E mutations elicited a profound increase in affinity to all human FcγRs, with the greatest enhancement of binding observed for the activating receptor FcγRIIIa, and particularly its low-affinity 158F allotype. Kas were also determined for an anti-CD19 Fc-KO (G236R/L328R), which was engineered to disrupt Fc receptor interactions for use as a control lacking effector function while retaining Fv-mediated signaling. The Fc-KO showed 630-fold reduced binding to FcγRI and had no detectable binding to FcγRIIa, FcγRIIb, and FcγRIIIa.
To evaluate the antitumor efficacy of XmAb5574 in mouse xenograft models, it was important to assess whether XmAb5574 also increased binding to mouse FcγRs. Similar to our observations with human FcγRs, XmAb5574 exhibited increased affinity for all mouse Fc receptors relative to the IgG1 analogue, and the anti-CD19 Fc-KO showed no detectable binding to any mouse FcγRs (Fig. 1B).
We also characterized the CD19 binding properties of XmAb5574 by determining Kas in a Raji cell–based competitive binding assay. As shown in Fig. 1C, XmAb5574, anti-CD19 IgG1, and the anti-CD19 Fc-KO all bind CD19 antigen with similar Kas, and these are approximately 2-fold higher than the Ka for the parent mouse anti-CD19 antibody (4G7). Thus, XmAb5574 has substantially increased binding affinity for both human and mouse FcγRs and has high affinity to the CD19 antigen.
XmAb5574 induces minimal receptor internalization. Because the reported rate of internalization of CD19 on anti-CD19 binding varies widely, and because internalization of the target antigen may affect antibody effector function (29), we compared the internalization of XmAb5574 to an anti-CD22 antibody and to rituximab (anti-CD20), which have been shown to induce either fast or slow rates of internalization, respectively (30). Antibodies were directly labeled with europium and then incubated with Raji cells, which express high levels of all three antigens (31). Dissociation of antibody from cell surface CD19 is a function of antigen affinity, and as expected, increased with time, resulting in an increased amount of europium-labeled antibody in the supernatant and a concomitant decrease in the total cell-associated fraction (Fig. 2A). Binding of XmAb5574 resulted in very little internalization over 4 h; maximum internalization occurred at about 2 h and was only 6% (Fig. 2B). Anti-CD20 displayed a similar profile and maximum, with <3% internalized (Figs. S1A and 2B). In contrast, anti-CD22 facilitated rapid internalization, exhibiting a maximum of 37% internalized at 30 min (Figs. S1B and 2B). These data indicate that XmAb5574 binding to CD19 induces minimal internalization which should not significantly affect effector functions.
XmAb5574 enhances ADCC and ADCP. To determine whether the increased FcγR affinity of XmAb5574 translates into enhancements in immune effector functions, we first performed in vitro ADCC assays using purified IL-2–activated NK cells and cell lines spanning a broad range of human lymphomas and leukemias [Burkitt's lymphoma, CLL, HCL, CD19+ chronic myeloid leukemia (CML), diffuse large B-cell lymphoma (DLBCL), and ALL] and expressing different levels of CD19 antigen (15,000–105,000 molecules/cell). XmAb5574 markedly enhanced ADCC in all tumor cell lines assayed (Fig. 3A), significantly increasing both efficacy (maximal lysis) and potency (EC50) relative to the IgG1 analogue. EC50s for XmAb5574 ranged from 0.1 to 1.0 ng/mL, representing increases in potency of 100-fold to 1,000-fold over the IgG1 analogue. Maximal lysis for XmAb5574 ranged from 25% to 90%, and no correlation with cell surface CD19 expression was observed; the IgG1 analogue was much less efficacious (consistent with previous observations; ref. 32), exhibiting either no or 2-fold to 3-fold less ADCC (except RS4;11 cells). Notably, XmAb5574 induced substantial lysis even in cell lines in which the IgG1 analogue was virtually inactive (SU-DHL-6 and Raji). ADCC for the anti-CD19 Fc-KO was also assayed and no lysis was observed (Fig. S2), indicating that FcγR engagement is required for antibody effector function. The XmAb isotype control antibody also induced no detectable cell lysis, indicating that increased binding to FcγR alone in the absence of specific Fv-antigen interaction was insufficient to induce ADCC.
To determine whether XmAb5574 mediated increased ADCC not only against cell lines but also against primary tumors, we performed an assay with tumor cells from patients with ALL or MCL using fresh unfractionated and unstimulated PBMCs. The ADCC averages of three PBMC donors are presented (Fig. 3B and C). XmAb5574 enhanced ADCC in both ALL and MCL cells, in striking contrast to the IgG1 analogue, which showed no detectable ADCC. The lack of measurable ADCC for the IgG1 analogue underscores the resilience of primary tumor cells as was also observed in two of the nine tumor cell lines (Fig. 3A).
To assess the effect of XmAb5574 on phagocytosis, we performed ADCP assays with monocyte-derived macrophages and two ALL cell lines. XmAb5574 increased phagocytosis efficacy by approximately 10-fold relative to the IgG1 analogue (Fig. 4), substantially lower than the enhancement of ADCC (1,000-fold). As discussed later, this difference may be a reflection of different FcγRs responsible for NK cell versus macrophage-associated effector functions. Finally, XmAb5574 did not exhibit any CDC activity on Ramos or Raji cells (data not shown).
Thus, XmAb5574 substantially enhanced immune effector function, augmenting ADCC up to 1,000-fold relative to the IgG1 analogue in a broad range of B-lymphoma and leukemia cell lines and in primary tumor cells, and eliciting a 10-fold enhancement in macrophage-mediated ADCP.
XmAb5574 enhances antiproliferative apoptotic activity. Because previous reports showed that anti-CD19 antibodies are antiproliferative (33, 34), we measured the effect of XmAb5574 on tumor cell viability. We also assessed the effect of Fc binding on antiproliferative activity by comparing the effects of XmAb5574 with those of the anti-CD19 IgG1 analogue and the anti-CD19 Fc-KO. Secondary cross-linking antibodies are often used in vitro to mimic the in vivo environment in which antibody cross-linking occurs via binding to FcγRs on immune effector cells (35). As a more direct mimic of these effects, we used an FcγRIIIa-158V-GST fusion receptor as a cross-linker (the GST domain spontaneously dimerizes) instead of the nonspecific antihuman Fc antibodies used in such assays. The cross-linked antibodies were incubated with SU-DHL-6 cells and cell titers were analyzed using a cell viability assay. XmAb5574 exhibited more potent antiproliferative activity than the IgG1 analogue (Fig. 5A), indicating the importance of optimized Fc binding to FcγRs in mediating this effect. However, the activity of the Fc-KO was similar to the IgG1 analogue, suggesting that the antiproliferative activity is not absolutely dependent on cross-linking, but can be amplified by it. XmAb5574 also displayed antiproliferative activity in Raji and Ramos cell lines (data not shown).
We also performed Annexin V staining and caspase activation assays to elucidate the mechanism of antiproliferative effects induced by XmAb5574, and found them to be due to caspase-mediated apoptosis (Fig. 5B and C). As in the proliferation assay, XmAb5574 was more potent than the IgG1 analogue, which was comparable to the anti-CD19 Fc-KO. The enhanced FcγR binding of XmAb5574 may have augmented antibody cross-linking and thus facilitated cell surface CD19 cross-linking, leading to increased apoptosis. Therefore, XmAb5574 showed potent antiproliferative effects due to caspase-induced apoptosis, and these effects were more pronounced than those seen with the IgG1 analogue.
XmAb5574 inhibits lymphoma growth in SCID mouse xenograft models. We examined the in vivo antitumor effect of XmAb5574, administered either prophylactically or in an established tumor model. In prophylactic studies, mice were implanted s.c. with Ramos or Raji Burkitt's lymphoma cells, then injected i.p. with XmAb5574 (1, 3, or 10 mg/kg), XmAb isotype control, or vehicle on days 3, 6, 10, and 13. In both prophylactic studies, all three XmAb5574-treated groups showed a statistically significant inhibition of tumor growth compared with XmAb isotype–treated or vehicle-treated controls (Fig. 6A and B). On day 42 of the Ramos study, XmAb5574 at 1, 3, or 10 mg/kg reduced tumor growth by 44%, 59%, and 69% relative to controls, respectively. In the Raji study, XmAb5574 reduced tumor growth by >80% at all three doses compared with vehicle or XmAb isotype controls, and tumors were completely eradicated in 10% to 30% of the mice. In both studies, we found no difference between the XmAb isotype–treated and vehicle-treated controls, suggesting that the S239D/I332E mutations did not induce nonspecific activation of immune effector cells.
Next, we assessed the efficacy of XmAb5574 in an established tumor model. To determine the contribution of FcγR binding in mediating antitumor effects, we compared the in vivo activities of XmAb5574, the anti-CD19 Fc-KO, and the IgG1 analogue. SCID mice were implanted with Ramos lymphoma cells, and tumors were allowed to grow to a mean volume of ∼80 mm3 before twice weekly treatment with 10 mg/kg was started on day 14. By day 27, all three anti-CD19 antibodies caused a significant reduction in tumor growth compared with vehicle (Fig. 6C and Fig. S3); XmAb5574 elicited the most dramatic effect, reducing tumor growth by 80% on day 27 (P < 0.0001). The fact that the anti-CD19 Fc-KO also significantly decreased tumor growth (26%) indicates that mechanisms other than those mediated through FcγRs contribute to antitumor efficacy. However, as evidenced by the markedly increased inhibition observed for XmAb5574 relative to the anti-CD19 Fc-KO (80% versus 26%) and to a lesser degree for the IgG1 analogue versus the Fc-KO (57% versus 26%), most of the inhibition seems to be Fc-mediated. Finally, a significant difference was also seen between XmAb5574 and the IgG1 analogue (80% versus 57%; P = 0.0029), demonstrating the effect of increased affinity for FcγR due to the S239D/I332E Fc mutations.
These results show that XmAb5574, in addition to having strong antitumor effects in vitro, also has potent antitumor activity in vivo on human lymphoma xenografts. XmAb5574 inhibits tumor growth in mice via both FcγR-dependent and FcγR-independent mechanisms, and this inhibition is significantly enhanced relative to the IgG1 analogue.
B-cell depletion has been a successful strategy in the treatment of B-cell malignancies (36). The success of the anti-CD20 antibody rituximab in combination with chemotherapy has made this regimen the new standard of care for the treatment of B-cell cancers. The broad use of rituximab has shown that primary monotherapy with this agent, or retreatment of relapsed cases, is efficacious in approximately half of the patients (7). These failures of rituximab have been attributed to a variety of mechanisms (37), including a loss of CD20 expression on tumor cells, observed both in relapsed patients (5, 6) and in mouse xenograft models (38). CD19 is an attractive alternative target for the immunotherapy of lymphoproliferative disorders, due to its expression on a wide range of lymphomas and leukemias, including some early B-cell malignancies that do not express CD20 (2, 9). The potential for anti-CD19 antibodies to be effective against a broader range of B-cell tumors, their potential for overcoming rituximab-specific resistance, and the positive early clinical results with CD19 bispecific antibodies (16) make anti-CD19 antibodies especially promising candidates for therapeutic development.
Although early studies showed that unmodified anti-CD19 antibodies induced B-cell depletion, their clinical efficacy was limited (8, 15), and consequently, clinical trials focused on anti-CD19 immunotoxin conjugates (10, 39). To maximize the cytotoxic potential of an unmodified antibody, multiple modes of tumor cell killing must be engaged: immune effector function (ADCC, ADCP, or CDC) and growth inhibition mechanisms (apoptosis or cell cycle arrest). Data from both murine and human studies have shown that FcγR engagement, a major component of the effector function of antibodies, is the critical factor for optimal antitumor activity. Antibodies such as rituximab and trastuzumab have been reported to lose most of their antitumor activity in mice lacking the common γ chain required by all mouse activating FcγRs (40). Patient FcγR polymorphism analyses have shown that the affinity preference of IgG1 antibodies for the FcγRIIIa-158V versus 158F allotype correlates with improved progression-free survival in lymphoma and breast cancer patients treated with antibodies against CD20 (41, 42) or HER-2/neu (43), respectively. Furthermore, a series of studies in mouse models and human clinical trials have corroborated these fundamental observations (44, 45) and collectively indicate that higher FcγR affinity is associated with more efficient tumor clearance.
Previously, we showed that modification of the Fc domain to increase FcγR binding results in a dramatic enhancement in ADCC (19), and others have shown that anti-CD19 antibodies deplete B cells in an FcγR-dependent fashion (9). Here, we report that XmAb5574, an anti-CD19 antibody engineered for high affinity to human FcγRs, not only increases ADCC, but also increases ADCP and apoptosis, and enhances tumor inhibition in mouse xenograft models. XmAb5574 has high affinity to human CD19 and substantially increased affinity for human FcγRIIIa and FcγRIIa (100-fold and 10-fold, respectively). The higher affinity for FcγRIIIa translates directly into substantially enhanced ADCC against a broad range of leukemia and lymphoma cell lines and primary tumor cells. The increased FcγR affinity of XmAb5574 seems to be critical for maximal effectiveness because the IgG1 analogue displayed no or minimal levels of ADCC. The substantial increase in macrophage phagocytosis attributed to the S239D/I332E Fc substitutions is also of significance, given the growing evidence that macrophage-associated effector functions may be essential for in vivo antibody activity (46, 47). The observed difference in the XmAb5574-mediated enhancement of ADCC versus ADCP (1,000-fold versus 10-fold) may be a reflection of different FcγRs responsible for either NK cell or macrophage-associated effector functions. The Fc substitutions increase binding to FcγRIIIa 10-fold more than to FcγRIIa. In contrast to NK cells, which express only FcγRIIIa, macrophages express both of the low-affinity activating receptors, FcγRIIIa and FcγRIIa. Therefore, the more modest increase of ADCP conferred by these mutations may reflect the dominance of FcγRIIa in mediating phagocytosis, consistent with previous observations (28).
Fc engineering of XmAb5574 also resulted in enhanced apoptosis relative to the IgG1 analogue and the anti-CD19 Fc-KO. Although antibody-elicited apoptosis depends on the direct interaction of antibodies with their cell surface targets, these effects may be amplified by antibody cross-linking via FcγR on effector cells (35). Thus, optimization of Fc-binding to FcγRs in XmAb5574 may also result in more efficient cross-linking to mediate the enhanced apoptosis. This mechanism is supported by the observation that no difference in apoptotic activity is detected between XmAb5574 and the IgG1 analogue when cross-linked via a non–FcγR-binding epitope (data not shown). Thus, increased Fc binding to FcγRs substantially enhanced both effector-mediated killing and apoptotic activity of the resultant antibody, affording XmAb5574 with multiple mechanisms of tumor cytotoxicity.
XmAb5574 treatment of established human lymphoma resulted in significant inhibition of tumor growth relative to vehicle and the anti-CD19 Fc-KO, and notably was more inhibitory than the anti-CD19 IgG1 analogue. Although the anti-CD19 Fc-KO exhibited detectable tumor inhibition, the activity was severely attenuated relative to either XmAb5574 or the IgG1 analogue, adding to observations made by others (40, 46–48) and further emphasizing the importance of FcγR engagement for maximal antitumor activity. The S239D/I332E substitutions in XmAb5574 increased affinity for all the murine FcγRs, with a particularly strong enhancement for mouse activating FcγRI and FcγRIV, the latter of which has been strongly implicated in the in vivo activity of antibodies (46). Consistent with its enhanced affinity for murine FcγRs relative to anti-CD19 IgG1, XmAb5574 also showed more tumor inhibition than its IgG1 analogue, emphasizing the importance of Fc engineering for enhanced in vivo efficacy. These observations are similar to those made for an anti-CD32b antibody with enhanced affinity for mouse FcγRIV (21). However, it is worth noting that extrapolation of the mouse results to human clinical efficacy is not straightforward because of disparities between mouse and human immune systems, including differences in FcγRs and cell types responsible for immune function (17). Nevertheless, taken together, these promising results support further development of XmAb5574 for the therapy of human lymphoma and leukemia.
Internalization of CD19 upon antibody binding is a matter of controversy. Several reports conclude that anti-CD19 antibody binding either facilitates (30, 49) or has no effect on internalization (50, 51). These contradictory results may be reconciled by differences in CD21 expression levels, which are reportedly inversely correlated with internalization (31). Because antibody effector function could be affected by internalization of the antibody-antigen complex, we studied internalization facilitated by XmAb5574 and showed that very little occurs. Moreover, it is obvious from the in vivo results that the degree of internalization observed was not sufficient to prevent the dramatic tumor inhibition seen in the mouse lymphoma models.
The humanized and affinity-optimized anti-CD19 antibody XmAb5574 was Fc-engineered to enhance immune cell–mediated effector functions. It also possesses increased apoptotic activity that may be facilitated via its augmented FcγR interactions. Thus, XmAb5574 exhibits multiple modes of potent tumor cytotoxicity that are substantially increased relative to native IgG1 anti-CD19 antibodies. This activity profile should maximize the chances for successful clinical application of XmAb5574 in a diverse patient population.
Disclosure of Potential Conflicts of Interest
The authors are employees of and have ownership interest in Xencor, with the exception of M. Peipp and R. Repp.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Araz Eivazi, Bally Randhawa, and Chi-Ming Yu for technical contributions, Marie Ary for assistance with writing the article, and David Szymkowski for helpful discussions.