Therapeutic control of B cell activation via recruitment of Fcγ receptor IIb (CD32B) inhibitory function with a novel bispecific antibody scaffold


  • All authors own stock or stock options in MacroGenics, Inc.



To exploit the physiologic Fcγ receptor IIb (CD32B) inhibitory coupling mechanism to control B cell activation by constructing a novel bispecific diabody scaffold, termed a dual-affinity retargeting (DART) molecule, for therapeutic applications.


DART molecules were constructed by pairing an Fv region from a monoclonal antibody (mAb) directed against CD32B with an Fv region from a mAb directed against CD79B, the β-chain of the invariant signal-transducing dimer of the B cell receptor complex. DART molecules were characterized physicochemically and for their ability to simultaneously bind the target receptors in vitro and in intact cells. The ability of the DART molecules to negatively control B cell activation was determined by calcium mobilization, by tyrosine phosphorylation of signaling molecules, and by proliferation and Ig secretion assays. A DART molecule specific for the mouse ortholog of CD32B and CD79B was also constructed and tested for its ability to inhibit B cell proliferation in vitro and to control disease severity in a collagen-induced arthritis (CIA) model.


DART molecules were able to specifically bind and coligate their target molecules on the surface of B cells and demonstrated a preferential simultaneous binding to both receptors on the same cell. DART molecules triggered the CD32B-mediated inhibitory signaling pathway in activated B cells, which translated into inhibition of B cell proliferation and Ig secretion. A DART molecule directed against the mouse orthologs was effective in inhibiting the development of CIA in DBA/1 mice.


This innovative bispecific antibody scaffold that simultaneously engages activating and inhibitory receptors enables novel therapeutic approaches for the treatment of rheumatoid arthritis and potentially other autoimmune and inflammatory diseases in humans.

Activation-inhibition coupling, the pairing of a positive signal with an inhibitory loop, controls the magnitude and duration of many biologic processes (1–3). In B lymphocytes, recognition of an antigen by the clonotypic B cell receptor (BCR) induces a signal that can direct clonal expansion, differentiation, the release of cytokines, and, ultimately, Ig production. Uncontrolled activation is prevented by exhaustion of the activating stimulus as well as by the triggering of a negative feedback loop that involves the engagement of an inhibitory Fcγ receptor (FcγR), FcγRIIb (CD32B) (4). The latter mechanism is triggered when the BCR recognizes immune-complexed antigen, resulting in the concomitant engagement of CD32B by the Fc domain of the complex-bound IgG, thus preventing the expansion of B cell clones that share the same specificity as that recognized by the soluble IgG. Attesting to its critical role in immune regulation, CD32B-knockout mice on the Th1-prone C57BL/6 background develop a lupus-like glomerulonephritis (5, 6). We have focused on CD32B inhibitory signaling in B lymphocytes as a model system for the development of an alternative class of biologics that exploit activation-inhibition coupling for the control of immune activation. In addition to Ig production, B lymphocytes play a central immunologic function as regulators of the adaptive immune response, as shown by the clinical success of B cell–depleting therapies, such as anti-CD20 monoclonal antibody (mAb) interventions (e.g., rituximab), in autoimmune diseases (7, 8).

A successful negative regulatory strategy should recapitulate the antigen-driven proximity of the activating and inhibitory receptors that form the molecular basis for the negative signaling loop. We have developed a platform of dual-affinity retargeting (DART) molecules in which the Fv regions of 2 distinct antibodies are paired by heterodimerizing sequences and covalently linked by a carboxyl-terminal disulfide bond. DART molecules designed to modulate B cell function were constructed by pairing an Fv region from a mAb directed against CD32B (9) with an Fv region from a mAb directed against CD79B, the β-chain of the invariant signal-transducing dimer of the BCR complex (10). We show here that coligation of activating and inhibitory receptors by CD32B × CD79B DART molecules can alter the response of human B lymphocytes to BCR stimuli, resulting in signal disruption as well as in inhibition of calcium mobilization, cell proliferation, and Ig secretion in vitro. Furthermore, treatment of mice with a DART molecule directed against the mouse orthologs controls the development of collagen-induced arthritis (CIA) in a susceptible strain. This novel approach to biologic response modulation has potential as an intervention in autoimmune diseases.


Cells and antibodies.

B lymphocytes from healthy donors and mouse splenic B cells were isolated with Dynal B-Cell Negative Isolation Kits (Invitrogen). The CHO-CD32B cell line was previously described (9). Daudi and RAMOS cell lines were from American Type Culture Collection (ATCC). Anti-human CD32B mAb, 2B6 and 3H7, were described previously (9); mAb 8B5 was generated in a similar manner. Monoclonal antibody CB3.1 (anti-human CD79B) (10) was obtained from Dr. Max Cooper (University of Alabama at Birmingham). Monoclonal antibodies 2.4G2 (anti-mouse CD32/CD16), HM79 (anti-mouse CD79B), and 4-4-20 (antifluorescein) were from ATCC. Monoclonal antibodies 2B6, 8B5, and CB3.1 were humanized (IgG1) by complementarity-determining region grafting. Monoclonal antibodies 2.4G2 and HM79 were chimerized to mouse IgG1, and mAb 4-4-20 was chimerized to a human IgG1.

DART molecules.

DART molecules are bispecific diabodies (11) in which the 2 chains are linked via a disulfide bond formed between Cys residues at the C-terminus of the 2 respective chains (Figure 1a). In some cases, we introduced domains prior to the C-terminal cysteines to drive assembly of the desired heterodimeric form. Control (CONTR) DART molecules contained a 4-4-20 (antifluorescein) Fv region as the irrelevant arm. The CD32B(2B6) × CD79B, CD32B(2B6) × CONTR, CD32B(8B5) × CONTR, and CONTR × CD79B DART molecules were engineered with an E-coil domain (chain A) and a K-coil domain (chain B). A second CD32B(2B6) × CONTR DART molecule was constructed with chain A appended with FNRGEC (residues 209–214 of human Cκ) and chain B appended with VEPKSC (residues 216–221 of human Cγ1). CD32B(8B5) × CD79B and murine (m) CD32 × mCD79B were built with both chains terminating with LGGC. CD32B(2B6) × CONTR, CD32B(8B5) × CONTR, CONTR × CD79B, and CD32B(2B6) × CD79B DART molecules were expressed by transient transfection of the 2 chains, each driven by cytomegalovirus immediate early promoter, into HEK 293H cells. CD32B(8B5) × CD79B and mCD32 × mCD79B DART molecules were expressed using the GS System (Lonza) in CHO-S cells.

Figure 1.

Structure, biochemical characterization, and binding properties of CD32B × CD79B dual-affinity retargeting (DART) molecules. a, Schematic representation of linear sequences assembled into covalently linked DART molecules. The 2 polypeptide chains are depicted with an N-terminal VL domain, an amino acid linker, a mismatched C-terminal VH domain, and a Cys residue for covalent linkage. b, Analysis of purified DART molecules by sodium dodecyl sulfate–polyacrylamide gel electrophoresis under reducing and nonreducing conditions. Lane 1, CD32B(2B6) × CONTR (control); lane 2, CD32B(8B5) × CONTR; lane 3, CONTR × CD79B; lane 4, CD32B(2B6) × CD79B; lane 5, CD32B(8B5) × CD79B; lane 6, murine (m) CD32 × mCD79B. c, Kinetic parameters obtained by surface plasmon resonance analysis of DART molecules binding to the corresponding immobilized receptor ectodomains. Binding constants were measured in 2 independent experiments. The difference between measurements was <20%. d, Results of CD32B-CD79B bispecific enzyme-linked immunosorbent assay. Data were fitted to a sigmoidal dose-response curve by using Prism software (GraphPad Software) and are representative of 3 independent experiments. Values are the mean ± SEM. RLU = relative luminescence units.

All DART molecules were purified by antigen affinity chromatography using Sepharose 4B (GE Healthcare) coupled to one of the DART antigens: fluorescein-conjugated IgG, human soluble (hs)CD79A/B-Fc-Agly (a fusion protein linking the CD79A and CD79B ectodomains to an aglycosyl [N297Q] human IgG1 domain), or a mouse soluble CD16-Fc fusion protein (msCD16-Fc-Agly, the mouse CD16 ectodomain fused to mouse N297Q IgG1 Fc; note that 2.4G2 cross-reacts with mCD16). This was followed by size-exclusion chromatography. The resulting proteins were >95% monomeric heterodimer with <5% high molecular weight species. Dual antigen recognition was tested by enzyme-linked immunosorbent assay (ELISA) using immobilized hsCD79A/B-Fc-Agly for capture and biotinylated hsCD32B (hsCD32B-Fc) (9) plus horseradish peroxidase (HRP)–conjugated streptavidin for detection. Data were analyzed by using Prism software (GraphPad Software).

Surface plasmon resonance (SPR) analysis.

Binding of DART molecules to hsCD32B and hsCD79A/B was analyzed by SPR using a BIAcore 3000 biosensor. Human soluble CD32B or hsCD79A/B was immobilized on the CM-5 sensor chip using the amine coupling kit (GE Healthcare). Briefly, the carboxyl groups on the sensor chip surface were activated, and the antigen was then injected over the activated CM-5 surface in 10 mM sodium acetate, pH 5.0, at a flow rate of 5 μl/minute until an immobilization level of ∼500 reference units was reached, followed by 1M ethanolamine for surface deactivation. Binding experiments were performed in HEPES buffered saline with EDTA and Surfactant P20 (BIAcore). DART molecules were injected in duplicate at concentrations of 0, 6.25, 12.5, 25, 50, and 100 nM and at a flow rate of 30 μl/minute for 120 seconds, followed by a dissociation time of 180 seconds. Data were analyzed using BIAevaluation 3.1 software (BIAcore) after subtraction of the buffer sensogram. Kinetic constants, ka and kd, were estimated by global fitting analysis of the association/dissociation curves to the 1:1 Langmuir interaction model, with the equilibrium dissociation constant (KD) calculated as KD = kd/ka.

Flow cytometric analysis.

For the detection of DART molecule–mediated cell clustering, CHO-CD32B cells, RAMOS cells (CD79B+) (a B cell lymphoma line), Daudi cells (CD32B+CD79B+) (a Burkitt's lymphoma cell line), or purified human B cells were labeled with CellTrace 5,6-carboxyfluorescein succinimidyl ester (Invitrogen) or with PKH26 (Sigma) following the manufacturers' instructions. Cells (5 × 106/ml) were resuspended in phosphate buffered saline, and pairs were combined in a 1:1 ratio in the presence of 2 nM DART molecules for 1 hour at room temperature. Cells were analyzed by flow cytometry on a FACSCalibur flow cytometer (Becton Dickinson).

Expression of CD32B and CD79B was detected with Alexa 488–conjugated 2B6 and phycoerythrin-conjugated CB3.1, respectively. Cell-bound DART molecules, used at a concentration of 1 μg/ml, were detected using a polyclonal rabbit antibody directed against the E-coil/K-coil domain located within the DART molecule heterodimerization domain (anti-DART linker) followed by detection with 0.2 μg/ml of allophycocyanin-conjugated F(ab′)2 fragments of donkey anti-rabbit IgG (Jackson ImmunoResearch).

Western blotting.

CD32B immunoprecipitation from cleared lysates (from 1 × 107 cells) was performed with 2B6 prebound to protein A/G beads. Washed immunoprecipitates or whole-cell lysates (from 1 × 106 cells) were resolved by 9–12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis, transferred onto polyvinylidene difluoride membranes, blotted with 1 μg/ml of anti-CD32B (3H7), anti–phospho-CD32B, anti–phospho-Syk, anti–phospho–ZAP-70, anti–phospho-Akt, anti–phospho–ERK-1/ERK-2 (all from Cell Signaling Technology), anti–phospho-SHIP (Stem Cell Technology), or anti-GAPDH (Sigma), and visualized using enhanced chemiluminescence.

Intracellular calcium mobilization.

B cells were loaded for 30 minutes at 37°C with 4 μM Fluo-4AM (Invitrogen) in Hanks' balanced salt solution containing 0.02% Pluronic F-127 and 1 mM probenecid (Invitrogen). Changes in intracellular Ca2+ levels were recorded as Fluo-4 fluorescence emission in fluorescence channel 1 of a FACSCalibur instrument (BD Bioscience) running Flow-Jo software (Tree Star).

B cell proliferation and Ig production.

Human or mouse B cells (5 × 104/ml) were seeded in complete RPMI 1640 medium in 96-well plates (100 μl/well). For human B cell proliferation assay, cells either were activated with 2.5 μg/ml of mouse monoclonal anti-human μ-chain (Southern Biotechnology) and 10 μg/ml of goat anti-mouse antibodies in the presence of increasing concentrations of DART molecules or were treated with DART molecules or with control antibodies alone. For mouse B cell proliferation assay, cells were activated with goat anti-mouse μ-chain (30 μg/ml) in the presence of increasing concentrations of DART molecules or control antibodies. One microcurie per well of 3H-thymidine was added to the wells after 48 hours for a 16–18-hour pulse before harvesting. Incorporation of 3H-thymidine was measured by liquid scintillation counting. For Ig measurements, resting human B cells were left untreated or were activated with goat anti-human μ-chain (30 μg/ml) in the presence of 20 nM DART molecules or control antibodies. Cells were incubated for 72 hours at 37°C, supernatants were harvested, and secreted Ig were measured by ELISA. Briefly, microtiter plates were coated with 2 μg/ml of goat anti-human Ig to capture the secreted Ig. Bound Ig were detected using HRP-conjugated goat anti-human κ-chain antibody (Invitrogen) with subsequent colorimetric development.

Pharmacokinetic analysis.

The level of mCD32 × mCD79B in mouse serum was determined in a bispecific ELISA using msCD79A/B-Fc-Agly for capture and msCD16-Fc-Agly-biotin (2.4G2 binds mouse FcγRIII) plus streptavidin–alkaline phosphatase for detection. Concentrations of mCD32 × mCD79B were calculated using a 4-parameter algorithm fitting the data against a standard curve, and pharmacokinetic calculations were performed using WinNonlin Professional 5.1 (Pharsight). Parameters were determined by noncompartmental analysis based on an intravenous (IV) injection model (Model 201) and the linear trapezoidal method.

Mouse model of CIA.

The mouse strains used in the CIA study were generated from CD16−/−, hCD32A-transgenic, and hCD16A-transgenic mice obtained from Dr. Jeffrey Ravetch (Rockefeller University, New York, NY). Arthritis in DBA/1 mouse strains was induced with a single intradermal (ID) injection of 200 μl of a 1-mg/ml suspension of bovine type II collagen in modified Freund's complete adjuvant (CFA) (both from Chondrex), and, when indicated, the animals were immunized with 2 additional ID injections of bovine type II collagen in Freund's incomplete adjuvant (Sigma) on days 21 and 42. Disease severity was scored as footpad swelling recorded with an electronic caliper 3 times a week. Mice from the same litter were randomly assigned to treatment and control groups and were generated from >10 backcrosses. All experiments were conducted under the approval of the Institutional Animal Care and Use Committee.


Construction, characterization, and binding properties of CD32B- and CD79B-specific DART molecules.

CD32B × CD79B DART molecules were constructed by using humanized variable chains from 2 anti-CD32B mAb and those from a single anti-BCR Fv region (Figures 1a and b). One DART molecule incorporated the humanized version of 2B6, a high-affinity anti-human CD32B mAb capable of blocking immune complex binding to the inhibitory receptor (9). A second DART molecule was built on humanized 8B5, an antibody that binds a linear epitope outside of the IgG binding site of CD32B and does not compete for immune complex binding. Control DART molecules were built by substituting one or the other arm with an antifluorescein Fv region from mAb 4-4-20. Each purified DART molecule migrates as a single species under nonreducing conditions and can be resolved into its constituent 2 chains of the predicted molecular mass (based on its Fv regions and linker components) under reducing conditions (Figure 1b).

The affinity of the CD32B arm for its antigen was comparable with that of the CD79B arm, as in the case of the CD32B(8B5) × CD79B DART molecule, while the CD32B(2B6)-based DART molecules showed greater affinity for CD32B than the CD32B(8B5)-based versions, as anticipated (Figure 1c). In both cases, the on-rate for CD32B binding was faster than that for CD79B binding. Both CD32B × CD79B DART molecules interacted with their antigens, as shown in a dual-specificity ELISA in which soluble recombinant CD32B was immobilized on the plate and soluble CD79B was used as detecting reagent for bound DART molecules (Figure 1d). The shift in the CD32B(8B5) × CD79B DART molecule binding curve compared with that of the CD32B(2B6) × CD79B DART molecule was consistent with the lower affinity of its CD32B arm.

While control DART molecules reacting with only 1 of the 2 antigens went undetected in this ELISA format, fluorescence-activated cell sorting (FACS) analysis demonstrated binding to the corresponding antigens on live cells for all DART molecules (Figure 2a). CHO cells transfected with the human inhibitory receptor (CHO-CD32B) were used to detect CD32B-specific binding, while RAMOS cells, a B cell lymphoma line that expresses undetectable levels of CD32B, were used to detect CD79B-specific binding. Daudi cells, a Burkitt's lymphoma cell line that expresses both CD32B and CD79B, were used as double-positive cells, while nontransfected CHO cells represented the negative control. Bound DART molecules were detected by using a polyclonal rabbit antibody raised against a linker region shared by DART molecules. CHO-CD32B cells were stained exclusively by DART molecules that included the anti-CD32B specificity (CD32B × CONTR and CD32B × CD79B); conversely, only anti-CD79B–bearing DART molecules (CONTR × CD79B and CD32B × CD79B) bound RAMOS cells, while all DART molecules bound Daudi cells, confirming that each molecule performed as expected.

Figure 2.

Cell binding properties of CD32B × CD79B DART molecules. a, Left, CD32B and CD79B were detected on the indicated cell lines (see Materials and Methods) expressing CD32B, CD79B, or both CD32B and CD79B, using Alexa 488–conjugated 2B6 and phycoerythrin-conjugated CB3.1, respectively. Right, DART molecules were detected using the anti-DART linker rabbit antibody and allophycocyanin-conjugated F(ab′)2 fragments of donkey anti-rabbit IgG. b, Cells were labeled with 5,6-carboxyfluorescein succinimidyl ester (CFSE) or with PKH26, and pairs were combined as indicated in the leftmost diagram in a 1:1 ratio in the presence of the indicated DART molecules. Cell–cell clustering was detected in the upper right quadrants, and the results are representative of at least 3 experiments. PBS = phosphate buffered saline (see Figure 1 for other definitions).

A critical requirement for signal modulation via inhibitory coupling is that both receptors be coligated on the same cell rather than on adjacent cells; consistent with this notion, independent crosslinking of CD32B, even if simultaneous with BCR activation mediated by CD79B, was found to be insufficient to trigger negative signaling (further information is available at Given that DART molecules are monovalent binders for each target, we anticipated that avidity would favor binding to cells coexpressing both antigens (cis-binding mode) rather than binding either antigen on separate cells (trans-binding mode).

Detection of DART molecule–mediated cell clustering of cells bearing one, the other, or both specificities with different combinations of ligands was performed by FACS analysis of mixtures of cells loaded with 2 independent fluorochromes. When mixtures of CHO-CD32B cells and RAMOS cells (CD79B+) were used as targets, the CD32B × CD79B DART molecule mediated at most 26% clustering of the 2 cell types under optimal conditions (2 nM) (Figure 2b), while no clustering was observed with control DART molecules (Figure 2b). In contrast, when Daudi cells (CD32B+CD79B+) were used as targets, the CD32B × CD79B DART molecule did not exhibit trans binding activity. Furthermore, the CD32B × CD79B DART molecule did not mediate trans binding even when added to a mixture of Daudi and CHO-CD32B or RAMOS cells, indicating that the cis-bound DART molecule is not available to mediate heterologous cell clustering. The preferential cis-binding mode of the CD32B × CD79B DART molecule also occurred with human peripheral blood B cells (Figure 2b).

Treatment with a CD32B × CD79B DART molecule disrupts BCR-induced calcium mobilization and signal transduction.

CD32B inhibitory function is the result of the phosphorylation of an immunoreceptor tyrosine–based inhibition motif (ITIM) in its intracellular tail (12–15). Treatment of human peripheral blood B lymphocytes with the CD32B(8B5) × CD79B DART molecule resulted in the phosphorylation of the CD32B ITIM motif (Tyr292) in a B cell activation–dependent manner (aggregation of cell surface IgM by goat anti-human μ-chain was used as a surrogate for antigen-mediated BCR engagement) (Figure 3a, left panels), confirming that CD32B phosphorylation is dependent upon the activation of BCR-linked kinases (16). Importantly, the DART molecule did not trigger significant ITIM phosphorylation in the absence of BCR activation; therefore, coligation of CD79B with CD32B by a bispecific ligand whose arms are each functionally monovalent for their respective antigens is insufficient to trigger a signal. The phosphorylated CD32B ITIM recruits the cytoplasmic SHIP1, which in turn acts as a kinase substrate (13).

Figure 3.

CD32B × CD79B DART molecules activate an inhibitory pathway in human B cells. a, Left, Lysates from B cells treated with goat anti-human μ-chain (GAHμ; anti-μ) alone or in combination with CD32B(8B5) × CD79B DART molecules for 10 minutes were subjected to immunoprecipitation (IP) with anti-CD32B (clone 2B6), resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, immunoblotted with pCD32B Y292 (top), and stripped and reprobed with anti-CD32B (clone 3H7) (bottom). Right, Western blot (WB) analysis of whole-cell lysates from purified B cells that were treated as shown at left at different time points. The blots were probed with pCD32B Y292 (top) and anti–pSHIP1 Y1020 (middle). Equivalent loading was confirmed by stripping and reprobing with anti-GAPDH (bottom). b, Ca2+ mobilization was measured as mean fluorescence emission of Fluo-4AM–loaded B lymphocytes treated with phosphate buffered saline (PBS) or 1 μg/ml of the indicated DART molecules and stimulated for 140 seconds with 30 μg/ml of goat anti-human μ-chain followed by 1 μM ionomycin for an additional 140 seconds. The ionomycin response indicates equivalent loading. c and d, Whole-cell lysates from human B cells treated as described and for the indicated time intervals were immunoblotted with anti–pAkt S473 (c) or with anti–pZAP-70 Y319, anti–pSyk Y525/526, or anti–pERK-1/2 T202/Y204 (d). Immunoblots were stripped and reprobed with anti-GAPDH as a loading control. All results are representative of 3 independent experiments. See Figure 1 for other definitions.

Consistent with CD32B phosphorylation, phosphorylation of SHIP1 Tyr1020, a docking site for the Ras inhibitory adaptor, DOK1, was observed only in the presence of the CD32B × CD79B DART molecule and exclusively in response to BCR stimulation (Figure 3a, right panels). SHIP1 counteracts phosphatidylinositol 3-kinase (PI 3-kinase) signaling by dephosphorylating phosphatidylinositol 3,4,5-trisphosphate, a phospholipid that enables membrane translocation of proteins possessing pleckstrin homology domains. One such molecule, phospholipase Cγ, generates a second messenger that controls calcium mobilization (17). Consistent with SHIP1 activation, both CD32B × CD79B DART molecules attenuated BCR-induced Ca2+ mobilization (Figure 3b), while control DART molecules had no effect. Treatment with a CD32B × CD79B DART molecule did not affect BCR-induced phosphorylation of ZAP-70 or Syk (Figure 3d), 2 CD79B proximal kinases. In contrast, the phosphorylation of ERK-1/2 and Akt, 2 molecules downstream of Ca2+ mobilization and PI 3-kinase activation, was reduced compared with that of samples activated in the absence of the DART molecule (Figures 3c and d). These data indicate that by recruiting CD32B within the activation complex, treatment with a CD32B × CD79B DART molecule biases B cells toward inhibitory coupling of BCR activation.

Treatment with CD32B × CD79B DART molecules inhibits BCR-induced proliferation and Ig production.

Treatment with CD32B × CD79B DART molecules inhibited BCR-induced proliferation of human peripheral blood B cells in a dose-dependent manner (Figure 4a, upper and lower panels). Proliferation was unaffected by control DART molecules or the anti-CD79B or anti-CD32B mAb. Omitting the anti–μ-chain mAb resulted in no proliferation; under these conditions, the addition of CD32B × CD79B DART molecules did not result in detectable activation (Figure 4a, upper panel). Delaying DART molecule addition to prestimulated cultures for up to 24 hours resulted in nearly maximal attenuation of proliferation (Figure 4b, upper and lower panels). B cells treated with CD32B × CD79B DART molecules also showed greatly diminished Ig secretion upon stimulation with goat anti-human μ-chain compared with cells treated with antibodies of the individual specificities (Figure 4c, upper and lower panels). In multiple experiments, the 2B6-based DART molecule exhibited higher potency than the 8B5-based DART molecule, consistent with the differential affinity for the inhibitory receptor (Figure 1c). Inhibition was not associated with increased cell death or apoptosis (data not shown). These data show that CD32B × CD79B DART molecule–mediated coligation of the activating and inhibitory receptors biased the cells toward hyporesponsiveness via negative signaling.

Figure 4.

B cell treatment with CD32B × CD79B DART molecules inhibits B cell receptor–induced proliferation and Ig secretion. a, Purified human B cells were left untreated (phosphate buffered saline [PBS]) (top) or stimulated with mouse monoclonal anti-human μ-chain and goat anti-mouse antibodies (MAHμ + GAM) together with increasing concentrations of the indicated DART molecules (top) or with increasing concentrations of the indicated DART molecules and chimeric (Ch) monoclonal antibodies (mAb) (bottom). Values are the mean ± SEM 3H-thymidine (3H-TdR) incorporation. b, The indicated DART molecules (40 nM = 1 μg/ml) were added to purified B cells at the indicated time points before or after stimulation. After 48 hours, proliferation was assessed as 3H-thymidine incorporation during an 18-hour pulse. Values are the mean and SEM of triplicate cultures. c, Purified B cells were left untreated (PBS) or incubated for 72 hours with 30 μg/ml of F(ab′)2 fragments of goat anti-human μ-chain (GAHμ) in the presence of 20 nM of the indicated DART molecules or mAb (mouse mAb were used in the assay to avoid interference with the detection assay). Total Ig secretion in the supernatants was detected by enzyme-linked immunosorbent assay. Values are the mean and SEM of triplicate cultures. All results are representative of at least 3 independent experiments. See Figure 1 for other definitions.

Treatment with CD32B × CD79B DART molecules reduces disease severity in a mouse model of CIA.

To determine whether a CD32B × CD79B DART molecule–based intervention could control autoimmunity, we engineered a surrogate molecule directed against murine CD32B and CD79B, since the anti-human CD32B and anti-human CD79B mAb do not cross-react with the corresponding mouse antigens (data not shown). Mice express only the inhibitory form of FcγRII (mCD32), which is homologous to human CD32B. As in humans, mouse B lymphocytes express mCD32, where it functions as a negative regulator of BCR-induced activation. The mAb 2.4G2 is a rat mAb that binds mCD32; however, 2.4G2 also reacts with mCD16 (FcγRIII), an activating Fc receptor (18). To eliminate mCD16 reactivity, a confounding factor, mice deficient for mCD16 were generated on the autoimmune DBA/1 background (mCD16−/− DBA/1 mice) to generate a CIA model that could be used to test the therapeutic effect of DART molecules, since this model has been shown to be dependent on B cells, because B cell–deficient mice do not develop arthritis (19). Contrary to what occurs in wild-type DBA/1 mice, however, collagen immunization of these mice did not result in arthritis (20) (Table 1), expanding the observation of a previous study that showed that DBA/1 mice lacking the activating FcγR common γ-chain are resistant to CIA (21).

Table 1. Role of mCD16 and the hCD16A or the hCD32A activating Fcγ receptor transgene in the induction of CIA in DBA/1 mice*
Mouse lineCIA
  • *

    Arthritis was induced and assessed in mice as described in Materials and Methods. mCD16 = murine CD16; hCD16A = human CD16A; CIA = collagen-induced arthritis.

  • Animals received 3 immunizations with bovine type II collagen.

  • Animals received 1 immunization with bovine type II collagen.

mCD16−/− DBA/1No
mCD16−/− hCD16A+/+ DBA/1No
mCD16−/− hCD32A+/+ DBA/1Severe
mCD16−/− hCD16A+/+ hCD32A+/+ DBA/1Severe

In order to reconstitute CIA in mCD16−/− DBA/1 mice, the human CD16A or the human CD32A activating FcγR was incorporated as a transgene, resulting in mCD16−/− hCD32A+/+ DBA/1 mice, mCD16−/− hCD16A+/+ DBA/1 mice, and mCD16−/− hCD16A+/+ hCD32A+/+ DBA/1 mice. The human CD16A was unable to reconstitute CIA in mCD16−/− DBA/1 mice even after 3 injections of collagen; in contrast, when the hCD32A transgene was incorporated in mCD16−/− DBA/1 mice or in mCD16−/− hCD16A+/+ DBA/1 mice, severe arthritis was observed after a single injection with bovine type II collagen (Table 1). Human CD16A does not bind murine IgG (22), and this characteristic may explain the inability to reconstitute CIA in mCD16−/− DBA/1 mice. The human activating CD32A receptor binds murine IgG (23) (data not shown), and it can substitute for the mouse CD16 in the mCD16−/− DBA/1 mice.

The anti-CD79B arm of the mCD32 × mCD79B DART molecule was engineered from HM79, a hamster anti-mouse CD79B (24). The DART molecule was qualified for performance characteristics consistent with those of its human-reactive counterpart. These included the ability of the mCD32 × mCD79B DART molecule to inhibit BCR-induced proliferation of murine splenic B cells (Figure 5a) and a preferential cis-binding modality with double-positive cells (data not shown). Pharmacokinetic analysis following a single IV injection of mCD32 × mCD79B DART molecules in mCD16−/− C57BL/6 mice showed single elimination kinetics with a serum half-life of 9.1 hours (Figure 5b).

Figure 5.

A mouse-specific CD32 × CD79B DART molecule reduces disease severity of murine collagen-induced arthritis. a, Proliferation of purified splenic mCD16−/− hCD32A+/+ DBA/1 mouse B cells stimulated in the presence of increasing concentrations of mCD32 × mCD79B DART molecules or anti-mCD32 or anti-mCD79B monoclonal antibodies (mAb). The inset shows mCD32 and mCD79B expression in the B cell population used in the assay. Values are the mean ± SEM 3H-thymidine (3H-TdR) incorporation. Data are representative of 3 separate experiments. b, Pharmacokinetics analysis of mCD32 × mCD79B DART molecules in mCD16−/− C57BL/6 mice following a single intravenous injection at 2 μg/gm. Blood was collected from 5 male and 5 female mice at the indicated time points, and mCD32 × mCD79B DART molecule serum levels were determined by bispecific enzyme-linked immunosorbent assay. Values are the mean ± SEM. Pharmacokinetic (PK) parameters are shown in the inset. T½ = half-life; Tmax = time to maximum concentration of drug in serum; AUC = area under the curve; Cmax = maximum concentration. c and d, Induction and recording of arthritis in female mCD16−/− hCD32A++ DBA/1 mice as described in Materials and Methods. Treated mice received 2 μg/gm of mCD32 × mCD79B DART molecules intravenously daily for 5 consecutive days (arrows). Values are the mean ± SEM of the average of all 4 paws of 11 phosphate buffered saline–treated mice (untreated) or 12 DART molecule–treated mice per group. The difference between untreated mice and mice treated on days 0–4 (c) or on days 14–18 (d) was significant (P < 0.05 by analysis of variance followed by paired 2-tailed t-test). The difference in d remained significant after Bonferroni correction for multiple comparisons. See Figure 1 for other definitions.

To ascertain the DART molecule therapeutic activity, arthritis was induced in female mCD16−/− hCD32A+/+ DBA/1 mice with an ID injection of bovine type II collagen in modified CFA on day 0. Approximately 2 weeks after the induction, untreated mice developed signs of arthritis (inflamed paws and reduced mobility) that were associated with increased paw thickness (Figures 5c and d). Mice that received 2 μg/gm of mCD32 × mCD79B DART molecules IV daily for 5 consecutive days starting on day 0 (Figure 5c) showed delayed arthritis onset. More importantly, mice treated with 2 μg/gm of mCD32 × mCD79B DART molecules IV for 5 consecutive days starting on day 14 after induction, which is around the time of disease onset, showed disease regression with diminished disease severity compared with untreated mice (Figure 5d). In a separate experiment, a DART molecule reacting only with CD32B failed to ameliorate disease (data not shown). These data indicate that treatment with an inhibitory DART molecule capable of coligating CD32B and the BCR may be used to control autoimmune processes.


The physiologic role of inhibitory receptors in limiting cell activation makes them appealing targets for controlling pathogenic processes underlying autoimmune diseases. In B lymphocytes, the antigen itself serves as a scaffold for the formation of an immune complex that can link the stimulatory BCR and the inhibitory FcγR (25). We transduced this paradigm into a tool with pharmacologic potential by engineering a new class of bispecific diabodies, DART molecules, with specificities that directly ligate inhibitory and stimulatory receptors on the B cell surface. Treatment of B lymphocytes with a DART molecule recognizing CD32B and CD79B negatively regulated B cell activation with disruption of signal transduction and attenuation of BCR-induced calcium mobilization, proliferation, and Ig secretion. Furthermore, a mouse-specific CD32B × CD79B DART molecule demonstrated activity in a murine arthritis model. To our knowledge, this is the first example of pharmacologic manipulation of inhibitory signaling in autoimmunity.

Attempts to manipulate CD32B negative signaling have previously focused on approaches with a common Fc domain–based strategy. To attenuate basophil and mast cell antigenic responses (26, 27), proteins were engineered with an IgE Fc (the GE2 molecule) or with FelD, a cat allergen (the GFD molecule) fused to an IgG Fc domain. More recently, Chu et al (28) developed an anti-CD19 antibody with an Fc domain engineered with increased affinity for CD32B as a tool to coengage this receptor. Compared with its wild-type Fc domain counterpart, treatment with this antibody resulted in apoptosis of primary human B cells. Either strategy corecruited CD32B to the activating receptor by means of an IgG1 Fc domain. This approach is limited for applications in vivo, since IgG1 Fc-based fusion proteins will likely bind FcR other than CD32B. Even in the case of the Fc-engineered anti-CD19, binding to FcR other than CD32B is likely to occur, since the mutant demonstrated affinities for CD64 (FcγRI) and the H131 polymorphism of CD32A (FcγRIIA) that were identical to those of a native IgG1 (28). Furthermore, the interaction with the inhibitory receptor in an Fc domain–based strategy will be driven by proximity via primary binding to the activating receptors.

By engineering our scaffold with anti-CD32B mAb, we have obviated the limitations of Fc domain–based strategies together with increasing the specificity for the intended inhibitory target. Furthermore, the relative affinity of either arm of the scaffold can be selected or engineered in ways to introduce bias for one or the other antigen. To favor inhibitory receptor recognition, CD32B arms were engineered with faster on-rates and affinities equal to or greater than those of the CD79B arm, properties expected to favor primary recognition of CD32B. Furthermore, in selecting 8B5 as an alternate component of the CD32B × CD79B DART molecule, we have shown the feasibility of employing a binding entity that does not block the inhibitory receptor, which may be preferable to avoid interference with physiologic inhibitory mechanisms mediated through CD32B. Future studies will address the ability of CD32B × CD79B DART molecules to inhibit B cell responses in patients with autoimmune diseases, including ascertaining the inhibitory activity in relation to the CD32B Ile232Thr signaling polymorphisms (29).

Several reports have suggested that B cell apoptosis can be triggered by homoligation of CD32B. Accounting for up to 10% apoptosis of ex vivo mouse plasma cells or in vitro plasmablasts (30) or for up to ∼15% apoptosis of splenic mouse B cells (31), this phenomenon is physiologically important but of limited pharmacologic magnitude. Following treatment with anti-CD32B mAb and secondary crosslinking, we have observed apoptosis of human peripheral blood B cells of a magnitude similar to that reported (∼15%), although no effect on BCR-induced proliferation was seen (data not shown). Conversely, even extensive CD32B crosslinking in the absence of coligation with CD79B was ineffective in attenuating B cell proliferation regardless of the concomitant activation status (data not shown), and no apoptosis was observed after CD32B × CD79B DART molecule treatment irrespective of BCR activation (data not shown). Therefore, simultaneous coengagement of CD32B with the activating receptor is an absolute requirement for B cell inhibition.

This requirement is driven by a molecular constraint exploited by the CD32B × CD79B DART molecule—the need for CD32B to leverage a kinase associated with the activating receptor in order to phosphorylate its intracellular tail, since the inhibitory receptor is not coupled to a kinase (14). CD32B inhibitory function is therefore conditional on its proximity to, and the activation status of, the tyrosine kinase–coupled receptor. By loading the BCR complex with CD32B in resting B cells, a CD32B × CD79B DART molecule sensitizes the antigen receptor to deliver an inhibitory signal if antigen recognition occurs. Furthermore, in stimulated cells the inhibitory signal predominates over the activation signal, effectively dampening ongoing activation, as shown by inhibition of proliferation after delayed DART molecule addition and the DART molecule's ability to control ongoing inflammation in mouse CIA, a model that displays immunologic and histopathologic similarities to rheumatoid arthritis and the severity of which correlates with CD32B expression (32, 33).

The exploitation of CD32B inhibition coupling by DART molecules is defined by a unique set of features. By ligating each component in an essentially monovalent fashion, DART molecules appear to have no intrinsic activation properties. Furthermore, these DART molecules function as activation-dependent inhibitors, exerting their activity only in the context of antigen receptor signaling. CD32B-based DART proteins are simple to engineer, capable of faithfully preserving the binding properties of the parental Fv regions, and potent; these characteristics make them suitable not only as investigational tools but also for pharmacologic applications.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Bonvini had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Veri, Moore, Koenig, Johnson, Bonvini.

Acquisition of data. Veri, Burke, Huang, Li, Gorlatov, Tuaillon, Rainey, Ciccarone, Zhang, Shah, Jin, Ning, Minor.

Analysis and interpretation of data. Veri, Tuaillon, Koenig, Johnson, Bonvini.


We would like to thank Robert Whitener and Amanda Zhang for technical assistance with antibody and DART molecule purification, Weili Wang, Arin Whiddon, and Wanhua Yan for help with protein expression, Wenjun Zhang for in vitro assays, and Shelley Butler for assistance with animal modeling. We thank Jeffrey L. Nordstrom, Kathryn E. Stein, and Timothy J. Mayer for discussion and critical review of the manuscript and Melinda Hanson for editorial assistance.