Macrophage and NK-mediated Killing of Precursor-B Acute Lymphoblastic Leukemia Cells Targeted with a-Fucosylated Anti-CD19 Humanized Antibodies

Abstract

This work reports the tumoricidal effects of a novel investigational humanized anti-CD19 monoclonal antibody (Medi-551). An a-fucosylated antibody with increased affinity for human FcγRIIIA, Medi-551 is shown to mediate both antibody-dependent cellular cytotoxicity (ADCC) and phagocytosis (ADCP). Medi-551/CD19 complexes internalize slowly (>5 hrs) and thus remain accessible to effector cells for prolonged periods. We evaluated in vitro ADCC and ADCP activities of primary human natural killer (NK) cells and macrophages against pre-B ALL cell lines and pediatric patient blasts. Fluorescent imaging studies document immunological synapses formed between anti-CD19-bound target leukemia cells and effector cells and capture the kinetics of both NK-mediated killing and macrophage phagocytosis. Genetic polymorphisms in FcγRIIIA-158F/V modulate in vitro activities of effector cells, with FcγRIIIA-158V homo- or heterozygotes showing the strongest activity. Medi-551 treatment of SCID mice engrafted with human pre-B cells led to prolonged animal survival and markedly reduced disease burden in blood, liver and bone marrow. These data show that anti-CD19 antibodies effectively recruit immune cells to pre-B ALL cells and support a move forward to early phase trials in this disease.

INTRODUCTION

Precursor-B acute lymphoblastic leukemia (pre-B ALL) is the most common malignancy in children and young adults. Although cure rates approach 80%¹, survivors may develop one or more therapy-related toxicities after receiving dose-intensified treatment. Leukemic relapse occurs in approximately 20% of patients² and for this subset novel treatments are critically sought. There is a clear need for less toxic, targeted therapies capable of engaging the immune system as opposed to causing its compromise¹.

Immunotherapy using antibodies to leukemia cell surface antigens has potential for extending the spectrum of treatment options to most patients with pre-B ALL. Monoclonal antibodies involve various mechanisms to target cancer cells, including NK-mediated killing via FcγRIIIA (CD16) and phagocytosis by macrophages through their repertoire of Fcγ receptors (FcγRI, FcγRIIA, FcγRIII)^3-8. Antibody-based therapies can also involve complement-dependent cytotoxicity (CDC) or directly modify the activity or ligand capabilities of cell surface receptors on target cells^3,9. In B-cell diseases, the anti-CD20 monoclonal antibody rituximab has improved outcome in B-cell lymphomas^10-12. However, CD20 surface expression is typically low or absent during early stages of B-cell development and has poor response in ALL¹³.

CD19 is a B lymphocyte-specific surface antigen expressed by early pre-B cells from the time of heavy chain rearrangement¹⁴ and it can serve as a co-receptor for the BCR^15,16. Because CD19 is typically detected on neoplasms of B-cell origin, including many B-cell derived leukemias and lymphomas, it is an attractive target for antibody-mediated therapy. Several CD19 specific antibodies, including unconjugated and bispecific reagents such as anti-CD19/CD3 or anti-CD19/CD16, are under development thus far for the treatment of B cell-mediated diseases^17-21.

Medi-551 is an affinity-optimized, humanized anti-CD19 investigational monoclonal antibody produced in eucaryotic cells that lack α1,6-fucosyltranserase²². As a consequence, the antibody's N-linked carbohydrates lack fucosylation and exhibit ~9 fold increased affinity to human FcγRIIIA and mouse FcγRIV²³. Prior work has shown Medi-551 effectively binds several classes of B-lineage neoplastic cells, including primary chronic lymphocytic leukemia (CLL) and acute lymphoblastic leukemia (ALL) blasts from adult patients and is devoid of CDC activity²⁴. Medi-551 inhibited tumor growth in a SCID mouse xenograft model of human B-cell lymphoma and stimulated NK-mediated killing of adult leukemia cells in vitro. The combination of Medi-551 with rituximab resulted in prolonged suppression of tumor growth (~60 days)²⁴. In vivo studies in human CD19/CD20 transgenic mice support continued development of Medi-551 for autoimmune disease and primarily implicated macrophages in the clearance of B-cells in mice ²².

Here we report preclinical studies of Medi-551 using as targets both pre-B ALL cell lines and blasts from pediatric patients, and primary human effector cells. We found significant variability in the killing capacity of NK cells from different human donors, linked to genetic polymorphisms in FcγRIIIA-158 that affect binding to a-fucosylated IgG^25,26. Human macrophages express additional activating Fcγreceptors (FcγRI and FcγRIIA)^5,27, rendering them less dependent on high affinity FcγRIIIA binding for phagocytosis of opsonized leukemia cells. Importantly, treatment of SCID mice engrafted with pre-B ALL cells led to significant reduction in tumor burden and prolonged mice survival with no observable complications. Taken together, results suggest that further development of Medi-551 is warranted in support of early phase trials in relapsed, pediatric precursor-B malignancies.

MATERIALS AND METHODS

Antibodies

Medi-551 was produced at MedImmune, Gaithersburg, MD according to good manufacturing practices, using a fucosyltransferase-deficient producer CHO cell line (BioWa Potelligent® Technology, BioWa Inc. Princeton, NJ). A-fucosylated R347 IgG1 (R347aFuc) served as a negative isotype-matched control. Antibody Labeling Kits (Invitrogen, Carlsbad, CA) were used for Alexa dye conjugation. Mouse anti-human CD137, CD16, CD32, CD64 were from Abcam (Cambridge, MA). Mouse anti-human granzyme, perforin and CD107a were from BioLegends (San Diego, CA). Secondary antibodies were Alexa Fluor-488 F(ab)'2 of anti-mouse IgG (Invitrogen) or DyLight488 AffiniPure F(ab)'2 of anti-rabbit IgG (Jackson Laboratories, West Grove, PA).

Cells and reagents

Pre-B ALL cell lines (697, MHH-Call3, Nalm6, RS4;11) were cultured in RPMI-1640 medium, 10% fetal bovine serum (FBS) (20% for MHH-Call3), 50 U/ml penicillin-streptomycin, 2 mM L-glutamine. Peripheral blood mononuclear cells were isolated from buffy coats of normal donors (United Blood Services, Albuquerque, NM) by centrifugation in a Ficoll-Paque (GE Healthcare) density gradient. Primary NK cells and monocytes were negatively isolated using Dynabeads Untouched Human NK Cells or Monocytes (Invitrogen). NK cells were cultured in Iscove's Modified Dulbecco's Medium (IMDM), 20% FBS, 10% AB-human serum (3H Biomedical), 50 U/ml Pen-Strep, 2 mM L-glutamine, 1x non-essential amino acids, 1 mM sodium pyruvate, 50 μM β-mercaptoethanol (Sigma, St. Louis, MO) and stimulated with interleukin-2 (IL-2, 100 U/ml; Peprotech, Rocky Hill, NJ). Macrophages were differentiated from monocytes by culturing 7 days in RPMI-1640, 10% FBS, Pen-Strep, L-glutamine, recombinant human macrophage colony stimulating factor (rhM-CSF; 250 ng/ml; Peprotech). Bone marrow samples were acquired at diagnosis from pediatric patients with written, informed consent (HRPO-05-435). Leukemia blasts were enriched by Ficoll-Paque centrifugation²⁸.

Quantitation of cell surface expression

CD19 receptors were enumerated by flow cytometry using anti-hCD19^FITC antibody (BD Pharmingen, Sparks, MD) and Quantum Simply Cellular anti-Mouse IgG beads by manufacturer's protocol (Bangs Laboratories, Fishers, IN). Bead populations were labeled to saturation with anti-hCD19^FITC antibody, providing a calibration curve for staining of pre-B ALL cells. The number of CD19 on cell surface was calculated from regression curve associating fluorescence channel value to the antibody binding capacity of beads.

Genetic analysis

Polymorphism of FcγRIII-158F/V was determined by nested PCR-based allele-specific restriction assay^10,29.

Acid stripping internalization assay

Pre-B ALL cells and patient blasts (0.5x10⁶/ml) were incubated in duplicate with 1μg of Medi-551^Alexa488 or R347aFuc^Alexa488 for specified times. One sample per set was resuspended in 2% paraformaldehyde (PFA; Electron Microscopy Sciences, Fort Washington, PA) for 15 min fixation. Other samples were incubated for 3 min in stripping buffer (1mM NaCl, 0.2M glacial acetic acid, pH 2.7) to remove non-internalized antibody, followed by fixation in 2% PFA. Samples were analyzed on a BD-FACSCalibur flow cytometer.

Antibody-dependent cellular cytotoxicity (ADCC)

Pre-B ALL cells or patient blasts were plated into 48-well plates (5x10⁵/well) and incubated (1 hr, 37°C) with Medi-551 or R347aFuc antibodies. Cells were resuspended in IMDM/5% FBS, plated in triplicate into 96-well plates (5x10⁴/well) and co-incubated with primary NK cells at target:effector (T:E) ratio of 1:3 for 4hr at 37°C. Target cell lysis was based on LDH release using the CytoTox⁹⁶ Non-Radioactive Cytotoxicity Assay (Promega, Madison, WI).

Antibody-dependent cellular phagocytosis (ADCP)

Monocyte-derived macrophages (7.5x10⁴) were dispensed into 8-chambered borosilicate coverglasses (Thermo Scientific). Pre-B ALL cell lines or patient blasts were stained with Cell Trace Violet from Invitrogen's Cell Proliferation Kit, then incubated with Medi-551 or R347aFuc as described above. Cells were resuspended in RPMI-1640/2% FBS and co-incubated with primary macrophages at T:E=1:3 for 4hr at 37°C in triplicate. Cells were fixed in 2% PFA for 10 min and phagocytized cells scored on a Zeiss LSM510 confocal microscope.

Live cell imaging

Pre-B ALL 697 cells or patient blasts (1x10⁶ cells/ml) were incubated with Medi-551^Alex594 or Medi-551^Alex488 (1 μg/ml, 1 hr, 37°C). NK cells were stained with SP-DiOC₁₈ (Invitrogen), then resuspended in phenol free IMDM/5% FBS Monocyte-derived macrophages were stained with Orange Cell Mask (1 μg/ml) and Hoechst 33342 (0.5 μg/ml) (Invitrogen) as indicated, then resuspended in phenol free RPMI-1640/2% FBS. Effector and pre-B ALL cells were dispensed into 8-well chambers in 3:1 ratio. Imaging was performed on a Zeiss LSM510 confocal microscope equipped with a 63x water objective heater.

Immunofluorescence

697 cells were treated with Medi-551^{Alexa 594} (1 μg/ml), for 1 hr at 37°C. Primary NK cells or macrophages were incubated with 697 cells at T:E=1:3 for 0.5-1 hr at 37°C in IMDM/5% FBS or RPMI-1640/2% FBS, respectively. Cells were dropped into poly-L lysine coated slides (15 min; Polysciences Inc.) and fixed in 4% PFA for 15 min. For intracellular proteins, cells were permeabilized with 0.5% Triton X-100/PBS (Sigma-Aldrich) for 3 min. After blocking with 3% BSA/PBS (30 min), samples were incubated with primary antibodies (1:200, 1 hr at RT or overnight at 4°C) followed by secondary antibodies (1:250, 30 min, RT). For surface antigens, chilled samples were labeled with antibodies, then fixed in PFA. Slides were mounted in ProLong Gold antifade reagent (Invitrogen) and analyzed by confocal microscopy.

Murine xenograft model of pre-B ALL

With local institutional approval (ARF-100510), female SCID mice (10-12 week-old; NCI, Rockville, MD) were housed in a specific pathogen free, AAALAC accredited facility as described²⁸. Mice were engrafted with 5x10⁶ 697 cells by tail vein injection on day zero. After five days, mice were divided into groups that received a regimen of Medi-551 injections (3 mg/kg), the same dose of control antibodies (R347aFuc) or normal saline alone. These treatments were administered twice weekly for total of five doses. To assess tumor burden, groups of 6-7 mice for each condition were sacrificed at 21 days post-engraftment. Disease burden was assessed in blood smears and slides prepared from formalin-fixed, paraffin embedded liver and bone marrow utilizing Wright and H&E stains, and immunohistochemistry-based detection of human hCD10²⁸. To evaluate survival, groups of 10 mice were treated with Medi-551, R347aFuc and saline control. Criteria for humane sacrifice were significant weight loss up to 15%, reduced activity and moribund condition.

RESULTS

Medi-551/CD19 complexes persist on the cell surface for prolonged periods

The rate of internalization of antigen-antibody complexes can be a key factor in effectiveness of therapeutic antibodies that mediate ADCC and ADCP, since only antibodies bound to the cell surface are accessible to effector cells. Fig. 1AB shows the slow internalization rate of Medi-551/CD19 complexes into pre-B ALL cells and patient blasts, as measured in vitro. Cells were incubated with Medi-551^{Alexa 488} for intervals up to 5 hrs. Endocytosis in pre-B ALL cells was quantified, based on the comparison of median fluorescent intensity of acid-stripped samples (the pool of internalized Medi-551/CD19) to unstripped samples (the total of surface-bound and internalized antibodies). Less than 10% of complexes internalized within 60 min, increasing to only 20% by 300 min. Microscopic images of 697 cells incubated with Medi-551^{Alexa 488} are shown in Fig. 1C, confirming that most of Medi-551/CD19 complexes remain on the surface of pre-B ALL cells for hours. No binding or antibody uptake was observed for pre-B ALL cell lines and primary cells incubated with control antibodies (R347aFuc; Supplemental Fig. 1)

Figure 1. Medi-551/CD19 complexes slowly internalize into pre-B ALL cell lines and patient-derived blasts in vitro.

(A) pre-B ALL cell linses and (B) blasts from four patients were co-incubated with 1 μg of Medi-551^Alexa488 for indicated times followed by acid stripping and/or fixation with 2% PFA. Flow cytometry was used to analyze the fluorescence intensity of internalized Medi-551/CD19 (stripped, red) as compared to the intensity representing the total of non- and internalized mAb (unstripped, blue). Bars represent standard deviations (SD). (B) Medi-551^{Alexa 488} was present on the surface of 697 cells during prolonged incubation. Cells were coated with 1 μg of Medi-551^{Alexa 488} for different intervals, fixed with 2% PFA and analyzed on a Zeiss LSM 510 META confocal microscope using a 63X/1.4 oil objective and Zen 2009 software. All measurements were performed in triplicate.

Supplemental Table 1 reports the characteristics of experimental cell lines that represent several common pre-B ALL genotypes. Three of the cell lines (Nalm6, 697, MHH-Call3) have a similar range of CD19 surface expression (>170,000/cell). Only RS4;11 cells express low levels of CD19 (~4500/cell). Table 1 reports characteristics including NCI risk assessment, genetic abnormalities and phenotypic markers of leukemia cells. CD19 expression levels in this group ranged from a low of ~60,000 to a high of ~145,000.

FcγRIIIA-158 F/V polymorphism affects Medi-551 ADCC activity against pre-B ALL cell lines and patient-derived blasts

A-fucosylated Medi-551 was shown to have no direct effect on viability or proliferation of all pre-B ALL cell lines, as compared to both untreated and isotype-matched controls (Supplemental Fig. 2AB). Similarly, no direct killing was observed for patient blasts incubated for up to 24 hrs with Medi-551 (Supplementary Fig. 2CD). Since CD19 has been linked to PI3-kinase activation³⁰, we also tested whether crosslinking with Medi-551 altered levels of AKT phosphorylation downstream of PI3K (Supplemental Fig. 2E). We observed no change, indicating that in the absence of effector cells Medi-551 does not induce detectable changes in CD19-mediated signaling.

Next, Medi-551 was tested for its ability to promote NK-mediated killing of pre-B ALL cells. Fig. 2A reports results of LDH assays for target cell cytotoxicity. IL2-stimulated NK cells from eight healthy donors were incubated with 697 cells opsonized with Medi-551. Two conclusions were drawn: first, optimal killing was achieved at doses ranging from 0.1-10 μg/ml Medi-551; second, when incubated under the same conditions, the killing response of NK cells from different healthy donors varied considerably. While NK cells from several donors killed 60-80% of target cells, three of eight NK preparations exhibited relatively poor killing capability (<20%).

Figure 2. FcγRIIIA-158F/V polymorphism affects NK-mediated cytotoxicity of pre-B ALL cell lines and patient-derived blasts opsonized with Medi-551.

ADCC was measured by LDH assay. Target pre-B ALL cells were preincubated 1 hr at 37°C with indicated concentrations of Medi-551, then washed and co-incubated for 4 hr with IL2-stimulated primary NK cells as effectors (T:E ratio 1:3). (A) Variability in antibody-mediated cytotoxicity initiated by a range of 8 healthy NK cell donors. (B, C, D) Comparison of antibody-mediated cytotoxity using four pre-B ALL cell lines as targets (697-blue, Nalm6-black, MHH-Call3-red, RS4;11-green) and NK cells from donors characterized for FcγRIIIA-158 polymorphism status as indicated by labels. (E, F, G) Comparison of antibody-mediated cytotoxity using blasts from six patients (Table 2) as targets and NK cells from donors characterized for FcγRIIIA-158 polymorphism status as indicated by labels. FcγRIIIA-158 polymorphism by nested PCR as in Methods. Measurements were performed in triplicate and data are presented as mean ± SD.

We speculated that the variability in NK-mediated killing efficiency might be linked to genetic polymorphisms in the gene encoding FcγRIIIA receptor, based upon reports that allotypes with V/V, F/V or F/F at amino acid position 158 differ in IgG-binding affinity and immune cell effector function^11,20. A PCR-based assay was implemented to determine the status of FcγRIII-158 polymorphisms for each NK donor. Of the pool of 23 donors, 11 individuals were determined to be heterozygous at this position (FcγRIIIA-158F/V). 9 were homozygous for the weak binding 158F/F allotype and only 2 were homozygous for the strong binding 158V/V allotype.

We evaluated the in vitro killing efficiency of NK cells from this donor pool, using 4 pre-B ALL cell lines as target cells (Fig. 2B-D). Results clearly link NK-mediated killing of leukemia cells with FcγRIIIA allelic variation, with the trend of 158V/V>F/V>F/F. Fig. 2B shows representative results using NK cells from donors homozygous for FcγRIIIA-158F/F, which proved capable of killing up to 30% of 697, Nalm6 or MHH-Call3 cells during the 4 hr incubation period. However, NK cells from 158F/F donors were ineffective at killing RS4;11 cells that have low CD19 levels. NK cells from donors homozygous for FcγRIIIA-158V/V were most effective, reaching 40-80% in vitro killing of 697, Nalm6 and MHH-Call3 cells and up to 33% of killing of RS4;11 cells despite low numbers of CD19 surface expression. NK cells from heterozygous donors were also effective in the in vitro cytotoxicity assay, indicated that the expression of at least one FcγRIIIA-158V form ensures highly effective ADCC activity.

NK-mediated cytotoxicity results were next confirmed using primary cells isolated from bone marrow of pediatric patients with precursor-B ALL. Fig. 2E-G show that the killing efficiency of NK cells against patient blasts bound to Medi-551 followed similar trends to that of the pre-B ALL cell lines, where efficient ADCC activity is linked to FcγRIIIA-158V/V and -158F/V polymorphisms, but not with the -158/F/F variant. We note that CD19 expression was not a limiting factor on this group of patients, since 50% killing was observed even for blasts having ~60,000 CD19 molecules/cell (patient #045-11) (Fig. 2F). As expected, incubation of pre-B ALL cell lines and patient samples with R347aFuc control antibodies did not mediate in vitro ADCC activity, regardless of FcγRIIIA genetic polymorphism (Supplemental Fig. 3).

Immunological synapses between Medi-551-bound pre-B ALL and NK effector cells mediate killing

To evaluate the interactions between NK and target cells that result in leukemia cell lysis, we employed live cell fluorescence imaging. Images in Fig. 3A,B document the formation of immune synapses between 697 cells (Fig. 3A) or patient blasts (Fig. 3B) and primary NK cells. For these experiments, target cells were pre-incubated with red fluorescent anti-CD19 antibodies (Medi-551^Alexa594) while NK cells were labeled with the green fluorescent dye, SP-DiOC₁₈. Formation of the immune synapse is accompanied by accumulation of CD19-Medi-551^Alexa594 complexes to the site of contact (red arrows in Fig. 3AB), reflecting their interactions with FcγRIIIA receptors on the NK surface. The image in Fig. 3C shows that this interaction promotes CD19-Medi-551^Alexa594 internalization (red internal vesicles in Fig. 3C) and typically results in a dramatic swelling of the target cell membrane that precedes cell lysis and cell death as reported before^31,32. Notably, we observed that NK cells from FcγRIIIA-158F/F donors were capable of forming synapses despite their lower killing efficiency (Supplemental Fig. 4), suggesting that the lower affinity interaction can support binding but is less efficient at initiating FcγRIII-mediated degranulation.

Figure 3. Medi-551 mediated ADCC involves formation of immunological synapses between Medi-551-bound pre-B ALL and NK cells as well as effector cell activation.

Pre-B ALL cells were preincubated with Medi-551^Alexa594 (1 μg/ml, 1 hr) and co-incubated with IL2- stimulated primary NK FcγRIIIA-158F/V cells at T:E ratio 1:3. NK cells were stained with SP- DiOC₁₈. A-C) Live cell imaging was performed in imaging medium (IMDM phenol free, 5% FBS) on a Zeiss LSM510 META confocal microscope equipped with 63x/1.2 water objective heater (Bioptics). Formation of immune synapses between target and effector cells, and concentration of Medi-551 at synapse, was observed in (A) 697 cells and (B) patient-derived blasts (1 hr). Over a 3-4 hr period, this process was resulted in target cell death, seen as cell swelling (C), loss of membrane integrity and lysis (C). (D-I) For immunofluorescent staining, Medi-551^{Alexa 594}bound 697 cells were co-incubated with NK cells at T:E = 1:3 for 1 hr. For surface staining (CD16, CD137, CD107a), samples were chilled and labeled with antibodies followed by PFA fixation. To label intracellular antigens in NK lytic granules (granzyme, perforin), samples were fixed and briefly permeabilized with 0.5% Triton-X100 before incubation with primary antibodies. For all cases, secondary antibodies were anti-mouse IgG Alexa Fluor 488 F(ab)'2 fragments as in Methods. Images in D-F show that the formation of immune synapse between antibody-bound target cells and NK cells promotes migration of distributed NK lytic vesicles (D, labeled here for granzyme) to the synapse (E, labeled for perforin), followed by detection of granzyme released into the cytoplasm of the target cell (F). The presence of FcγRIIIA (CD16) at synapses (G) confirms the involvement of Fcγ receptor in antibody-mediated ADCC. The expression of CD137 (H) and CD107a (I) on the surface of NK cells provides positive indication of effector cell activation.

Fig. 3D-F illustrate the steps in the NK-mediated killing process. Initially, NK lytic granules are well distributed (granzyme label, Fig. 3D), followed by migration to the synapse (perforin label, Fig. 3E). Subsequently, fusion of the lytic granules results in release of perforin into the synaptic cleft and uptake of granzyme into the cytosol of the target cell (Fig. 3F), where red endocytic vesicles identify the CD19⁺ leukemia cell. Fig. 3G confirms expression of FcγRIIIA in primary human NK cells based upon immunolabeling with anti-CD16 (green). As expected, overlap of red fluorescent CD19-Medi-551^Alexa594 and green fluorescent label for CD16 occurs only at the synapse between the two cells (yellow region in Fig. 3G).

Prior studies have demonstrated that activated NK cells upregulate a unique profile of cell surface antigens, including CD107a and CD137^33,34. Micrographs in Fig. 3H-I show that the interaction between NK and target cells has a reciprocal effect that activates NK cells, as indicated by the marked membrane labeling of CD107a and CD137.

Medi-551 induces potent in vitro ADCP activity against pre-B ALL cell lines and patient blasts

We developed an assay for phagocytosis of antibody-coated leukemia target cells by human monocyte-derived macrophages (MΦ), based upon microscopic analysis. This assay is illustrated in Fig. 4A, where target cells were labeled with Cell Trace Violet, followed by incubation with primary human macrophages. Under these conditions, individual macrophages were often observed to ingest multiple target cells, typically ranging from 2-5. The graph in Fig. 4B reports scoring macrophage and 697 leukemia cell interactions over a 4 hr incubation period. Initial binding within the first half hour resulted in a synapse-like contact, followed by phagocytic cup development and uptake of the target cell. At a T:E ratio of 1:3, the majority of 697 cells were completely engulfed within 3 hours. Supplemental Movie 1 shows this process in action (sped up 20 times). Similar results were obtained for primary mouse monocyte-derived macrophages and Medi-551-coated 697 cells, further confirming that murine macrophages can recognize Medi-551-opsonized target cells for phagocytosis (Supplemental Figure 5)²².

Figure 4. Medi-551 recruits macrophage Fcγ receptors to mediate phagocytosis of target cells.

(A) Microscopic technique used for evaluation of ADCP. 697 cells were stained with Cell Trace Violet followed by preincubation with Medi-551 (1 hr). Cells were then co-incubated with primary human monocyte-derived macrophages (MΦ) at T:E = 1:3 for 4 hrs. After fixation with 2% PFA, cells were enumerated on a Zeiss LSM510 META laser-scanning confocal microscope (63x/1.4 lens) and Zen 2009 Software. In this micrograph, 4 target cells are engulfed by one MΦ cell. (B) The kinetics of target cell binding (“synapses”) and internalization by MΦ. 697 cells were coated with Medi-551^Alexa488 (1 μg/ml; 1 hr) and nucleoli of these cells were stained with Hoechst 33342. Next, 697 cells were co-incubated with MΦ from a FcγRIIIA-158F/V donor at T:E ratio 1:3 for indicated intervals. Cells were fixed with 2% PFA and analyzed by confocal microscopy. Images in C-E represent data from live cell imaging, performed in RPMI 1640 phenol free medium, 2% FBS. Target cells were preincubated with green fluorescent antibody (1 μg/ml Medi-551^Alexa488 for 1 hr) and MΦ were stained with Orange Cell Mask. Target and effector cells (FcγRIIIA-158F/V) were co-incubated at a T:E ratio of 1:3 for up to 30 min while imaging. Macrophages were seen to form “synapse-like” structures with antibody-bound pre-B ALL (C) cells and patient blasts (D, E). Images in F-H report results of immunofluorescence labeling for Fcγ receptors on the surface of MΦ in contact with target cells. For these surface antigens, samples were chilled and labeled with antibodies followed by PFA fixation. The presence of FcγRIIIA (F, CD16), FcγRII (G, CD32), FcγRI (H, CD64) suggest the involvement these receptors in antibody-mediated phagocytosis of target cells. Secondary antibodies used here were either anti-mouse IgG Alexa Fluor 488 or DyLight488 Affini Pure F(ab)'2 fragments.

Images in Fig. 4C-E document that, like the NK synapse, Medi-551/CD19 complexes (green) accumulate at the contact site between target cells and macrophages labeled with Orange Cell Mask (red). Results were identical using cultured pre-B ALL cells (697 cells; Fig. 4C) or patient blasts (Fig. 4D-E) preincubated with Medi-551^{Alexa 488}. Figures 4F-H show results of immunolabeling macrophages for all classes of IgG receptors, FcγRI (CD64; Fig. 4F) and FcγRII (CD32; Fig. 4G) in addition to FcγRIIIA (CD16; Fig. 4H).

Having confirmed that the macrophage effector cells expressed multiple Fcγ receptors in these experiments, we next sought to determine the relative importance of FcγRIIIA-158 polymorphisms in the phagocytosis process (Fig. 5). Results document the relative phagocytic activity for macrophages derived from donors with all three FcγRIIIA-158 (F/F, F/V, V/V) allotypes , using our panel of pre-B ALL cell lines (Fig. 5A-C) or patient blasts (Fig. 5D-F) as targets. Robust phagocytic activity (80-100%) was observed when the combination included 1) macrophages with at least one high affinity allele of FcγRIIIA-158V and 2) target cells with CD19 expression in the range of 60,000-185,000 (Tables 1,2). It is remarkable that significant phagocytosis was observed for RS4;11 target cells, with only ~4500 cell surface CD19 receptors, provided the macrophages were either homozygous or heterozygous for FcγRIIIA-158F. Macrophages from donors homozygous for FcγRIIIA-158F/F were capable of taking up 697, Nalm6 and MLL-Call3 pre-B ALL cells, as well all patient blast samples, with efficiencies of up to 40%. We conclude that maximal phagocytosis occurs when effector cells expressing FcγRIIIA-158V are presented with leukemia target cells bearing abundant Medi-551-CD19 complexes (≥60,000 CD19 molecules for this patient group). Nevertheless, the presence of additional activating forms of FcγR (FcγRIIA/CD32 and FcγRI/CD64) on macrophages can mediate significant phagocytosis of leukemia target cells. Isotype-matched R347aFuc control antibodies did not mediate phagocytosis of pre-B ALL and patient blasts (Supplemental Fig. 6).

Figure 5. Medi-551 exhibits potent in vitro ADCP activity against pre-B ALL cell lines and patient derived blasts.

697 cells were stained with Violet Trace followed by incubation with stated concentrations of Medi-551 for 1 hr. After a PBS wash, cells were co-incubated with primary human monocyte derived macrophages (MΦ) at T:E = 1:3 for 4 hrs. After fixation with 2% PFA, cells were enumerated on a Zeiss LSM510 META laser-scanning confocal microscope. Percent phagocytosis was determined; at least 300 target cells were scored per condition. (A-C) Summary of phagocytosis results using four cell lines (697, Nalm6, MHH-Call3, RS4;11), as a function of FcγRIIIA-158 polymorphism on donor macrophages. (D-F) Summary of phagocytosis results using six patient blasts, as a function of FcγRIIIA-158 polymorphism on donor macrophages. Effector cells were analyzed for FcγRIIIA-158 polymorphism by nested PCR as in Methods. Data represent mean ± SD for triplicate samples.

Medi-551 has anti-tumor activity in murine xenograft model of human pre B ALL

Our final goal was to evaluate the anti-tumor ability of Medi-551 in an in vivo model, using SCID mice engrafted with human 697 pre-B ALL cells (Fig. 6, Table 2). While NK cells in these mice lack FcγRIV (orthologous of human FcγRIII) their macrophages are expect to be fully functional against Medi-551 opsonized cells. Post-euthanasia necropsies of liver and bone marrow from mice sacrificed at 21 days post-engraftment revealed the significant decrease (p<0.0005; Fisher's exact test) of leukemia burden in experimental group, as compared to the both control groups that had profound histologic evidence of disease progression (Fig. 6A, Table 2). The antitumor efficacy of Medi-551 was further investigated by assessing endpoint survival. As shown in the Kaplan-Meier plot in Figure 6B, Medi-551 treatment prolonged survival in engrafted SCID mice when compared to PBS-vehicle and R347aFuc controls (log rank, p<0.005).

Figure 6. Medi-551 lowers leukemia burden and prolongs animal survival in murine xenograft model of human pre B ALL.

SCID mice were engrafted with 697 cells and received 5 doses of Medi-551 twice weekly starting at day 5. Mice were sacrificed at day 21. Tissue sections were paraffin-embedded and processed. (A-B) H&E staining of liver tissue from control and treated mice 3 weeks after engraftment with 697 cells. The image in (A) is typical for control mice, showing a large nodule of pre-B ALL blasts in the tissue. The image (B) represents liver tissue from a Medi-551 treated mouse, with healthy tissue free of blast infiltration. (C-D) Immunohistochemical staining of bone marrow sections for human CD10 was performed to distinguish human blast cells from any potential leukocyte progenitors in this immunocompromised mouse. (C) hCD10⁺ staining of bone marrow from a control mouse shows blasts have filled essentially the entire marrow space. (D) By comparison, this image from a treated mouse shows normal bone marrow architecture and a marked reduction in hCD10+ staining of leukemia blasts. (E) Kaplan-Meier plot of animal survival in each treatment group reveals the increase in mice survival after treatment with Medi-551.

DISCUSSION

The targeted recruitment of immune effector cells, particularly macrophages and NK cells is important for therapeutic efficacy of monoclonal antibodies. Critical factors include the binding affinity and stability of the complex formed between mAb-coated target cells and Fcγ receptors on effector cells, since stable direct contacts are required for both ADCC- and ADCP-mediated killing of cancer cells^35,36. Therefore, current research in the field has focused on engineering the Fc fragment of mAbs to improve binding to effector cells^{9,25,26,35,37}.

Here, we describe the first preclinical data for investigational Medi-551 therapeutic antibody in a pediatric setting. We provide evidence that Medi-551: 1) induces ADCC of pediatric pre-B ALL cells in vitro, with the highest efficiency linked to an NK genotype of FcγRIIIA-158V/V or 158F/V; 2) mediates potent ADCP activity against pre-B ALL cells in vitro, with the caveat that maximal phagocytic activity depends on both the expression of FcγRIIIA-158V and high levels of antigen; and 3) lowers leukemia burden to prolong survival in pre-B ALL xenograft model.

Previous studies have linked the clinical efficacy of anti-CD20 rituximab antibodies to patients with FcγRIII-158V/V genotype^10,38. Genetic variations at this locus apparently do not distribute with expected Mendelian frequencies, since we and others^10,38 find that only 10-15% of the human population are homozygous for the high affinity FcγRIIIA-158V/V. Glycoengineered mAbs represent one strategy to make immunotherapy accessible to a wider group of patients, since even the weaker FcγRIIIA-158F allotype has improved binding to a-fucosylated IgG^10,38. The concept that Fc engineering can enhance immune effector function of mAbs is supported by data with other mAbs such as anti-CD19 (XmAb5574, MDX1342)^17,18,39,40, anti-CD20⁴¹, anti-CD40 (XmAbCD40)³⁷ and anti-CD70⁹.

Live cell imaging of antibody-dependent killing of leukemia cells by NK cells and macrophages is a novel aspect of our study. We demonstrate that Medi-551/CD19 complexes are initially uniformly distributed on the leukemia cell surface and remain available to effector cells for many hours in vitro. The slow endocytic rates are consistent with prior reports^42,43. Remarkably, tight binding to effector cells is associated with redistribution of Medi-551/CD19 complexes to synapses between the engaged cells. In the context of NK cells, the synapse architecture has been proposed to critically control the polarized secretion of cytotoxic proteins, selectively killing target cells and protecting bystander cells^44,45. We found that NK cells from all donors, regardless of FcγRIIIA polymorphism status, could form synapses with antibody-bound target cells. However, efficient killing was strongly associated with expression of FcγRIIIA-158V on the NK cell. This suggests that FcγRIIIA receptors accumulating at the synapse must remain stably bound to IgG-Fc on the target cell in order to efficiently signal to polarized degranulation. Once productively engaged, NK cell activation leads to up-regulation of the cell surface markers CD107a and CD137. This observation may be useful way to monitor mobilization of NK cells in patients treated with therapeutic antibodies that rely on the ADCC pathway. Our data agree with Ward et al.²⁴, who concluded that the number of CD19 receptors present on the surface of B-cells is important for ADCC. While these observations are likely to be relevant to other anti-CD19 antibodies, we note that ADCC activity may be influenced by higher rates of CD19 internalization under specific conditions, as suggested by studies of CLL cells treated with XmAb5574¹⁸.

Our microscopy studies also shed light on factors that regulate phagocytosis of Medi-551-opsonized leukemia cells. Microscopic analysis provided an accurate method to quantitate phagocytosis by primary human and mouse monocyte-derived macrophages, showing that individual cells could engulf multiple target cells over a matter of hours. Using this assay, we demonstrated that blasts from children bound sufficient anti-CD19 antibodies to induce highly efficient macrophage activity. Although the highest efficiency for antibody-mediated phagocytosis was seen for macrophages with the FcγRIIIA-158V/V or 158F/V genotypes, significant internalization of patient samples was also observed for macrophages expressing FcγRIIIA-158F/F. The poor uptake of the RS4;11 cells, with very low CD19 expression, suggests that the quantitation of antigen levels on patient blasts may be the most important criteria for stratifying those patients who are more likely to respond to anti-CD19 immunotherapy.

Effector functions of macrophages relay on interplay between all Fcγ receptors, including the activating forms of FcγRI, FcγRIIA and FcγRIII as well as the inhibitory receptor, FcγRIIB present on the cell surface^5,6,8,27. Richards et al.⁶ showed that human macrophages express high levels of FcγRIIA and FcγRIIB, lower but significant levels of FcγRIIIA, and low levels of FcγRI. Based upon incomplete reliance on high affinity FcγRIIIA expression for phagocytosis of Medi-551-bound leukemia cells (Fig. 5), we expect that the other FcγR receptors detected by immunolabeling are active participants (Fig. 4G-I). In particular, FcγRIIA is known to mediate phagocytosis of antibody-bound target cells^5,6,46. Another anti-CD19 antibody, XmAb5574, was engineered by introducing two point mutations in the Fc fragment for increased affinity to activating FcγRIIIA, FcγRIIA and FcγRI but was also found to bind inhibitory FcγRIIB; it has been shown to mediate ADCC and ADCP against NHL and CLL^17,18. Like Medi-551, MDX-1342 is an example of an antibody specifically engineered for increased affinity to FcγRIIIA and shows enhanced ADCP and ADCC activity against human lymphoma cells⁴⁰. However, clinical development of MDX-1342 has been discontinued. Thus, it is reasonable to conclude that a-fucosylated anti-CD19 antibodies elicit particularly strong phagocytic activity from FcγRIIIA-bearing macrophages, but also evoke participation of other Fcγ receptors. Medi-551 was generated by humanization and affinity maturation which improved the kinetics of binding associated with decreased internalization rate and prolonged persistence on the target cell surface, further improving in vitro potency relative to the parental antibody (Herbst, unpublished). By comparison, other anti-CD19 targeted therapies (SAF3419, BiTE) differ considerably in their mechanism of action by relying on either toxin delivery or T-cell engagement to kill target cells^47,48.

Mouse xenograft models are most informative for macrophage-mediated responses to humanized antibodies, since murine NK cells lack FcγRIV, the orthologue of human FcγRIIIA. Our study in SCID mice, engrafted with human 697 pre-B ALL cells, also shows that macrophages alone can reduce leukemic progression. We confirmed that only five doses of therapeutic antibody significantly lowered leukemia burden and extended mouse survival, if administered within 5 days of engraftment before complete bone marrow involvement. Therapeutic effect was seen at a dose of 3 mg/kg, previously established as an optimal dose for depletion of endogenous and human B-cells in mice^22,24. Others have pointed out that the different types and distribution of Fcγ receptors in mice vs. human can cause difficulties in extrapolating in vivo results from mice to potential clinical application in humans^25,37. Nevertheless, strong in vitro ADCC activity makes a compelling case that NK cell recruitment may enhance therapeutic activity in patients.

The strong potential of Medi-551 as an anti-leukemia and lymphoma agent merits further development to support moving this mAb into early phase trials in relapsed pre-B ALL, where CD19 expression is typically maintained at high levels⁴⁹. Based on the results of this study, we predict clinical responses when administered as post-induction therapy, after the initial reduction in tumor burden and followed by post-remission recovery effector cells. Consideration should be given to combination with other mAbs, chemotherapeutic drugs or immunomodulatory agents, as suggested by prior work with other drugs in this class.^50-52 Correlative studies, particularly that link genetic polymorphisms at FcγRIIIA-158 with outcome, should provide insight into the importance of NK cells in clinical response.