Research ArticleImmunotherapy

Anti-CD20/CD3 T cell–dependent bispecific antibody for the treatment of B cell malignancies

See allHide authors and affiliations

Science Translational Medicine  13 May 2015:
Vol. 7, Issue 287, pp. 287ra70
DOI: 10.1126/scitranslmed.aaa4802


Bispecific antibodies and antibody fragments in various formats have been explored as a means to recruit cytolytic T cells to kill tumor cells. Encouraging clinical data have been reported with molecules such as the anti-CD19/CD3 bispecific T cell engager (BiTE) blinatumomab. However, the clinical use of many reported T cell–recruiting bispecific modalities is limited by liabilities including unfavorable pharmacokinetics, potential immunogenicity, and manufacturing challenges. We describe a B cell–targeting anti-CD20/CD3 T cell–dependent bispecific antibody (CD20-TDB), which is a full-length, humanized immunoglobulin G1 molecule with near-native antibody architecture constructed using “knobs-into-holes” technology. CD20-TDB is highly active in killing CD20-expressing B cells, including primary patient leukemia and lymphoma cells both in vitro and in vivo. In cynomolgus monkeys, CD20-TDB potently depletes B cells in peripheral blood and lymphoid tissues at a single dose of 1 mg/kg while demonstrating pharmacokinetic properties similar to those of conventional monoclonal antibodies. CD20-TDB also exhibits activity in vitro and in vivo in the presence of competing CD20-targeting antibodies. These data provide rationale for the clinical testing of CD20-TDB for the treatment of CD20-expressing B cell malignancies.


The addition of the CD20-targeting antibody rituximab to chemotherapy has greatly improved the outcome for patients with B cell lymphoma and chronic lymphocytic leukemia (CLL). Nevertheless, disease relapse or recurrence will occur in virtually all patients with follicular lymphoma and CLL, and about half of patients with aggressive B cell lymphoma, for example, diffuse large B cell lymphoma. Despite repeated rounds of treatment, most will ultimately die of disease-related sequelae (1). Monoclonal antibodies targeting B cell antigens, including CD19, CD20, CD22, CD30, and CD52, have been explored for treatment of various B cell malignancies, but none has yet definitively demonstrated a safety and efficacy profile superior to rituximab (2). There therefore remains a need to develop B cell–targeting therapeutics with a distinct mechanism of action (MOA) to increase cure rates in B cell malignancies.

Recent clinical reports illustrate the effectiveness of therapies that involve redirection of T cell cytolytic activity against cancer cells. One approach involves the ex vivo manipulation of autologous or allogeneic T cells to express chimeric antigen receptors (CARs) that target lineage-specific surface molecules such as CD19. CAR-expressing T cells (CAR-T cells) have produced deep and durable responses in patients with relapsed/refractory acute leukemia (3, 4). However, toxicities related to severe cytokine release syndrome, the scalability of production to the broader cancer population, and the clinical benefit in disease beyond the acute leukemia remain significant challenges for the development of CAR-T therapies (5).

An alternate approach to T cell therapy involves the use of bispecific molecules that redirect endogenous T cells to recognize tumor cells. Such therapeutics concomitantly bind to CD3 on T cells and a target antigen such as CD19 on target cancer cells. The therapeutic potential of this approach is exemplified by of the recent approval of blinatumomab, a bispecific T cell engager (BiTE) targeting CD19, for treatment of relapsed/refractory B cell acute lymphoblastic leukemia (ALL). In adult patients with relapsed/refractory ALL, treatment with blinatumomab resulted in complete responses in 26 of 36 (72%) patients, with 88% of these responders achieving minimal residual disease (MRD)–negative status, and a median overall survival of 9.0 months (6). Blinatumomab treatment was also effective in achieving MRD negativity in patients who were MRD-positive after induction and consolidation chemotherapy (7). The clinical activity of blinatumomab has demonstrated that T cell responses against tumors were achievable without the need for ex vivo immune cell manipulation, which affords advantages over cell-based therapies, especially with respect to manufacturing and access. Cytokine release syndrome and central nervous system toxicity similar to that observed with CAR-T therapy were observed with blinatumomab (8, 9), suggesting a possible class effect of redirecting T cells to target B cell malignancies. In addition to blinatumomab, other B cell–targeting bispecific proteins with a variety of structural formats have been characterized preclinically and tested in small human clinical trials (8, 1013).

Following on the development of these first-generation T cell–recruiting agents, there remain opportunities to optimize T cell–directed therapies based on the generation, characterization, and use of preclinical models. Here, we describe CD20 T cell–dependent bispecific antibody (CD20-TDB), a full-length, fully humanized immunoglobulin G1 (IgG1) antibody as T cell–recruiting therapeutic for CD20+ malignancies. CD20-TDB is cross-reactive to cynomolgus monkey CD3ε and CD20 antigens, allowing for appropriate preclinical testing. We provide evidence of potent in vitro and in vivo B cell killing activity and characterization of MOA-related pharmacologic activities of CD20-TDB in normal and leukemic cells, three murine models, and cynomolgus monkey. These studies highlight the use of several new preclinical approaches for the characterization of such B cell–targeted therapies and provide the rationale for clinical testing of CD20-TDB.


CD20-TDB is a full-length humanized IgG with conventional antibody pharmacokinetic properties

We have recently described modular production of full-length IgG1 bispecific antibodies with near-native architecture using “knobs-into-holes” technology (14). Anti-CD20/CD3 TDB (CD20-TDB) was produced without detectable level of homodimers and aggregates (fig. S1). As predicted from our previous experience with this molecular format (15), the pharmacokinetic (PK) properties of CD20-TDB in rats are typical of a nonbinding human IgG1 antibody, suggesting that intermittent dosing on a weekly to monthly schedule will be clinically feasible (fig. S2).

CD20-TDB is a conditional T cell agonist dependent on the presence of both T cells and target for activity

We next characterized the MOA for CD20-TDB. Although the molecule was produced in Escherichia coli and therefore expected to be devoid of antibody-dependent cell-mediated cytotoxicity (ADCC) activity, we first confirmed that functional activity was not dependent on the presence of antibody constant domain regions. In vitro potency was assessed against the B cell line BJAB using healthy human donor peripheral blood mononuclear cells (PBMCs) and an F(ab′)2 fragment of CD20-TDB lacking the Fc region. The fragment was found on a molar basis to be equipotent to intact CD20-TDB (Fig. 1A), suggesting that Fc receptor–mediated ADCC does not contribute substantially to CD20-TDB activity. Conversely, depletion of CD3-expressing cells from PBMCs abolished the activity of CD20-TDB (Fig. 1B), indicating that T cells are required for activity.

Fig. 1. Anti-CD20/CD3 TDB activates T cells in the presence of target and mediates B cell killing in a T cell–dependent manner.

(A) BJAB cells and PBMCs isolated from healthy donor (1:10 cell ratio) were incubated with various concentrations of full-length CD20-TDB or F(ab′)2 CD20-TDB for 24 hours (data shown as means ± SD, n = 3). (B) BJAB cells and PBMCs isolated from healthy donor, or PBMCs depleted of CD3+T cells (1:10 cell ratio), were incubated with various concentrations of CD20-TDB for 24 hours. (C to F) BJAB cells and purified CD8+T cells or CD4+ T cells, or CD8+T cells only, were incubated with various concentrations of CD20-TDB for 24 hours. [T cell–to–BJAB cell count ratio was 5:1 in (C), (E), and (F), or 3:1 in (D) as duplicate]. Cell killing and T cell activation marked as CD69+CD25+ were measured and calculated as described in Materials and Methods. Granzyme B (GrB) induction was also detected by fluorescence-activated cell sorting (FACS), and perforin concentration in medium was measured by enzyme-linked immunosorbent assay (ELISA). Assays were done either in triplicate with average and SD value plotted (A) or as single-dose response curve representative of multiple assays of different donors.

Purified CD4+ or CD8+ T cells used as effector cells were activated by CD20-TDB with comparable potency as measured by the induction of both CD69 and CD25 expression (Fig. 1C and fig. S3). T cell activation by CD20-TDB was strictly dependent on the presence of CD20+ target cells (Fig. 1C). T cell proliferation was also observed in vitro only in the presence of target B cells and CD20-TDB (fig. S4). CD8+ T cells appeared to be more potent in BJAB cell killing because CD8+ T cells resulted in a greater extent of cell killing compared to an equal number of CD4+ T cells (Fig. 1D) (fig. S5; P = 0074, paired t test; n = 3). Consistent with these results, we observed significant intracellular granzyme B up-regulation that was more prominent in activated CD8+ T cells compared to CD4+ T cells (Fig. 1E) (fig. S5; P = 0068, paired t test; n = 3), as well as higher levels of perforin (Fig. 1F) (fig. S5; P = 0046, paired t test; n = 3) released into cell culture medium (16).

CD20-TDB is broadly active across healthy donor B cells and lymphoma cell lines

CD20-TDB was tested in dose-response assays across a panel of seven B lymphoma cell lines with various levels of surface CD20 expression and half-maximal effective concentration (EC50) values ranging from 0.22 to 11 ng/ml in vitro (Fig. 2A). CD20-TDB had no activity against SU-DHL1 cells that are devoid of CD20 expression (Fig. 2A). CD20-TDB was also highly active for killing of endogenous B cells by endogenous T cells in PBMCs prepared from healthy donors. Shown in Fig. 2B are representative dose-response curves from 8 healthy donors (of 30), and the summary of EC50 values and the extent of killing from 30 healthy donors demonstrating a mean EC50 for killing of less than 3 ng/ml and a mean extent of killing about 85% after 24 hours in culture. Of the 30 healthy donors tested, 57 to 96% of B cells were killed with antibody (1000 ng/ml) within 24 hours, with EC50s ranging from 0.4 to 140 ng/ml.

Fig. 2. Anti-CD20/CD3 TDB is potent in killing human lymphoma B cell lines and healthy donor primary autologous B cells in vitro.

(A) B cells and PBMCs isolated from a single healthy donor (1:10 cell ratio) were incubated 24 hours with various concentrations of CD20-TDB. SU-DHL1 cells are CD20-negative. CD20 expressions for the B cell lines were shown, with an isotype control shown in gray. (B) PBMCs isolated from healthy donors were incubated for 24 hours with CD20-TDB at the indicated concentrations. Killing curves in each plot shown are from representative assays [of 4 different donors for (A) and more than 30 different donors for (B)]. For dot plots, extent of cell killing is reported for CD20-TDB (1000 ng/ml), and each dot represents a unique donor with a horizontal bar indicating the mean.

CD20-TDB is highly effective in transgenic murine models

To date, the lack of antibody cross-reactivity to nonhuman models has greatly limited preclinical characterization of T cell–recruiting bispecific antibodies (17). We developed a double-transgenic mouse strain expressing both human CD20 and human CD3ε (huCD20-huCD3) to enable detailed studies in a mouse model with an intact and fully functional immune system (18, 19).

In huCD20-huCD3 double-transgenic mice, we found the expression level for both human antigens on mouse B and T cells to be appropriately lineage-restricted but lower than that on healthy human donor B and T cells, suggesting that the model may be challenging with respect to antigen copy numbers (Fig. 3A). In a dose-ranging study to identify a minimal efficacious dose that would cause robust depletion of B cells, we assessed numbers of circulating and splenic B cells after a single administration of CD20-TDB ranging from 0.00005 to 0.5 mg/kg. Whereas a dose of 0.05 mg/kg caused transient depletion of peripheral B cells 1 day after CD20-TDB administration, a dose of 0.5 mg/kg was required to maintain B cell depletion through day 7 in peripheral blood (Fig. 3B and fig. S6). In spleen samples at day 7, the 0.05 mg/kg dose approximated an ED50 (half-maximal effective) dose level, whereas the 0.5 mg/kg dose was sufficient for complete B cell clearance (Fig. 3C).

Fig. 3. Double-transgenic mouse model.

(A) Human CD3ε and CD20 expression levels on mouse B and T cells from huCD20-huCD3 double-transgenic mice were compared to healthy human donor B and T cells by FACS. MFI, mean fluorescence intensity. (B and C) huCD20-huCD3 double-transgenic mice were treated once intravenously on day 0 with various dose levels of CD20-TDB. Mouse blood [days (D) 1 and 7] and spleens (day 7) were collected, and B cells were quantitated by FACS. (D and E) huCD20 single-transgenic (sTG) mice or huCD20-huCD3 double-transgenic (dTG) mice were treated once intravenously on day 0 with antibodies as indicated (10 mg/kg for rituximab and 0.5 mg/kg for CD20-TDB and HER2-TDB). Mouse spleens were collected at day 7 for B cell quantitation by FACS. Bars in the plots indicate mean values, with P values calculated by unpaired t test (n = 4 or 7 mice per group).

To confirm the CD3 and CD20 specificity of these effects, we dosed human CD20 single-transgenic mice with CD20-TDB, HER2 (human epidermal growth factor receptor 2)–TDB, or rituximab (15), and assessed spleen B cell numbers 7 days after administration. In this experiment, CD20-TDB and rituximab bind mouse B cells but not mouse T cells, whereas HER2-TDB binds neither mouse B or T cells nor mouse HER2. As expected, both TDB test agents were inactive as compared to rituximab (Fig. 3D). We next conducted a similar experiment comparing the activity of CD20-TDB, HER2-TDB, and rituximab in huCD20-huCD3 double-transgenic mice. In this experiment, rituximab binds only to B cells, CD20-TDB binds to both B and T cells, and HER2-TDB binds only to T cells. In contrast to the single-transgenic mouse experiment, CD20-TDB was highly active in the double-transgenic mice and showed significantly greater clearance of splenic B cells than rituximab despite a 20-fold lower dose. As expected, the HER2-TDB was inactive in B cell depletion in both single- and double-transgenic mice (Fig. 3E).

We further characterized immune cell dynamics in spleens of the double-transgenic mice over a 2-week period after administration of a single dose of CD20-TDB (0.5 mg/kg). We observed that splenic B cells declined ~50% from baseline by day 1 (24 hours post-dose) and continued to decline to a nadir on day 3, with slight recovery detected at day 14 (Fig. 4A). Coincident with the large decline in B cells on day 1, we observed that most splenic T cells had an activated phenotype on day 1. By day 2, most T cells were no longer CD69-positive, although levels of CD69+CD8+ cells continued to range from 10 to 30% for the remainder of the 2-week observation period (Fig. 4B). Total splenic CD8+ T cell counts were markedly increased by day 2 of the study, and the timing and size of the increase were suggestive of a proliferative expansion (Fig. 4C). We also noted that, on day 3 when splenic B cells were mostly cleared, CD8+ T cell counts started to decrease significantly after the expansion and eventually came down to below the baseline levels.

Fig. 4. Anti-CD20/CD3 TDB is active in dTG mice.

(A) Time course of tissue B cell depletion. (B and C) CD8+T cell activation and cell numbers. After a single dose of CD20-TDB in huCD20-huCD3 double-transgenic mice. huCD20-huCD3 double-transgenic (dTG) mice were treated once intravenously on day 0 with CD20-TDB (0.5 mg/kg). Spleens were collected on the indicated days, and B cell or CD8+T cell counts and T cell activation were measured by FACS. Bars in the plots indicate mean values, with P values calculated by unpaired t test (n = 3 mice per group).

CD20-TDB is highly effective in murine models reconstituted with healthy human donor immune cells

To provide a second independent murine model, we investigated the use of immunocompromised NSG [NOD (nonobese diabetic) SCID (severe combined immunodeficient) gamma] mice engrafted with human CD34+ hematopoietic stem cells (20) and qualified as individual mouse reconstituted with >25% human CD45+ cells in peripheral blood. Mice used in experiments shown in Fig. 5 had 35 to 80% human CD45+ cells in peripheral blood with percentages of viable leukocyte gate of CD4+, CD8+, and CD20+ cells ranging from 12 to 25%, 2 to 9%, and 32 to 60%, respectively (fig. S7A). Levels of CD3 and CD20 expression from these mice were compared also to those of healthy human donors. CD20 levels on B cells were slightly higher, whereas CD3 levels on T cells were lower (CD8+T) or comparable (CD4+T) to those observed on healthy human donors (fig. S7B).

Fig. 5. Anti-CD20/CD3 TDB depletes B cells in humanized NSG mice.

Human CD34+ cell–reconstituted NSG mice (five per group) were treated intravenously on day 0 with 3 weekly doses of vehicle or CD20-TDB (0.5 mg/kg). (A to C) Blood was collected at 5 days before treatment (D−5), and on days 7, 14, and 21 (A and B); spleens were collected at day 21 (C). B and CD8+T cell counts were determined by FACS. Data in blood are shown as mean and SD, or in spleen as individual points with a bar indicating the mean and P values calculated by unpaired t test (n = 5 mice per group).

These mice were randomized to two groups according to human CD45+ cell counts 5 days before dosing (day −5), and treated with vehicle or with three weekly doses of CD20-TDB at 0.5 mg/kg starting on day 0. We observed a strong reduction of B cells in peripheral blood at day 7, with almost no B cells detected at day 21 (Fig. 5A). In peripheral blood, we observed up to a 10-fold increase in CD8+ T cells by day 7 and a return to baseline or lower at days 14 and 21 (Fig. 5B). A similar trend was also observed for CD4+ T cells (fig. S8). Robust B cell depletion was also observed in spleens of the CD20-TDB–treated mice at day 21 (Fig. 5C).

CD20-TDB is potent in killing primary patient CLL B cells in vitro and in vivo

To characterize the CD20-TDB in the setting of B cell malignancies, we next tested the potency against primary cells from CLL patients in vitro and in vivo. CLL was selected as a “high-bar” in vitro experimental system to assess the CD20-TDB because (i) CD20 exhibits particularly low surface copy number on CLL cells (21, 22); (ii) the effector-to-target ratio is skewed because, in relative terms, T cell numbers are decreased and B cell number are increased by disease; and (iii) the T cells that are present are functionally deficient (2326).

PBMCs prepared from peripheral blood of nine CLL patients were treated with CD20-TDB to determine the in vitro potency in the setting of CLL. We first characterized activity of a single high dose of CD20-TDB (1000 ng/ml) and measured B and T cell content after 48 hours of culture. As expected, we found that T cell content varied substantially (between 0.4 and 8% of mononuclear cells). Strikingly, we observed that the extent of tumor cell killing was highly correlated to T cell content (R2 value of 0.9, Fig. 6A). From four donors of low T cell content and low B cell killing with available sample material, we tested whether addition of healthy donor T cells could enhance CD20-TDB activity. This was indeed the case, demonstrating that T cell content of the cultures was the primary determinant of TDB activity in these cultures (Fig. 6B). For two samples, dose-response curves based on B cell killing and CD8+T cell activation based on CD69 and CD25 expression were generated. CD20-TDB activated CD8+T cells from CLL patients with a B cell killing EC50 of 10 to 20 ng/ml, a value similar to that in healthy donor PBMC assays despite the fact that the B cell leukemia tumor burden was 70% of mononuclear cells with only 8% CD8+T cells in patient sample A1837, and 80% tumor burden with 4% CD8+T cells in patient sample A183D. CD20-TDB can therefore achieve efficient B cell killing even at low effector-to-target ratios (1:8 and 1:18, respectively; Fig. 6C).

Fig. 6. Anti-CD20/CD3 TDB kills CLL patient leukemia cells in vitro and in vivo.

(A) PBMCs from CLL patients were incubated with CD20-TDB (1000 ng/ml) for 48 hours, and leukemia cell killing and T cell numbers were quantitated by FACS. (B) PBMCs from CLL patients were incubated with CD20-TDB (1000 ng/ml), either alone or supplemented with CD8+T cells purified from a healthy donor at the indicated ratios. After 48 hours, leukemia cell killing was quantitated by FACS (means ± SD, triplicate). (C) PBMCs from two different CLL patients were incubated with various concentrations of CD20-TDB for 48 hours, and leukemia cell numbers and T cell activation were determined by FACS. (D) Peripheral B cells and T cells isolated from CLL patients were engrafted in NSG mice as described in Materials and Methods. Mice were treated once intravenously with HER2-TDB at 0.5 mg/kg, rituximab at 0.5 mg/kg, or CD20-TDB at 0.1 or 0.5 mg/kg, and spleens were collected for immunohistochemical analyses 14 days after treatment.

To further characterize the potential of CD20-TDB against CLL leukemic cells, we used a patient leukemia cell–derived adoptive transfer mice model that is a modification of a previously described approach (27). Representative examples are shown in Fig. 6D, with vehicle-treated animals demonstrating successful engraftment of B leukemic cells and autologous T cells from CLL patients. CD20-TDB treatment was initiated after confirmation of the leukemic graft. Very few B cells could be detected after a single dose of CD20-TDB treatment at 0.1 or 0.5 mg/kg. B cell depletion was also observed with rituximab treatment, whereas no B cell depletion was detected with HER2-TDB as a nontarget cell binding control.

CD20-TDB potency requires minimal CD20 expression and is active in the presence of rituximab

To explore the impact of CD20 expression on CD20-TDB activity, we tested the potency of CD20-TDB on lymphoma cell lines NALM-6, SC-1, and OCI-Ly19, which express CD20 at very low levels (Fig. 7A). The estimated CD20 copy number for these cell lines were less than 500 per cell, based on FACS binding data in comparison to BJAB cells (28). In contrast to comparable potency of CD20-TDB and rituximab against BJAB cells (fig. S9A), CD20-low cell lines were killed by CD20-TDB with an EC50 <10 ng/ml, whereas no significant activity was observed with rituximab (Fig. 7B). In addition, monovalent binding affinity of CD20-TDB for CD20 is ~50 to 100 nM, which is significantly less than the reported 8.6 nM binding affinity of rituximab (29). Together, these data suggest that the potency of CD20-TDB requires a low receptor occupancy of 10 to 100 molecules (3032).

Fig. 7. Anti-CD20/CD3 TDB is active on CD20-dim cells and active in the presence of rituximab.

(A) Surface CD20 expression was determined on four human lymphoma cell lines by FACS, isotype control in gray. (B) PBMCs isolated from healthy donor were depleted of B cells and used as effector cells against three different cell lines with low cell surface CD20 antigen. B cells and effector cells (1:10 cell ratio) were incubated with various concentrations of CD20-TDB or rituximab for 24 hours, and B cell killing was quantitated by FACS. (C) PBMCs isolated from healthy donor were first incubated with rituximab-DANA (R-D) at the concentration indicated for 1 hour, and then CD20-TDB was added. After 48 hours, B cell killing was determined by FACS. (D) huCD20-huCD3 double-transgenic mice were treated once intravenously at the dose indicated; for combination treatment, mice were pretreated intravenously with rituximab-DANA, and CD20-TDB (0.5 mg/kg) was injected intravenously 30 min later. Spleens were collected at day 7, and B cell counts were determined by FACS. Bars in the plots indicate mean values, with P values calculated by unpaired t test (n = 3 mice per group).

Because rituximab is widely used in the treatment of B cell malignancies, it was important to characterize CD20-TDB activity in the presence of rituximab, because CD20-TDB and rituximab bind to the same CD20 epitope and CD20-TDB activity may be affected by circulating rituximab from prior treatment. We therefore tested CD20-TDB activity in the presence of varying concentrations of rituximab-DANA, an effectorless rituximab variant that binds to CD20 without significant B cell clearance activity (fig. S9B). Pretreatment of healthy donor B cell target cells with rituximab-DANA had a remarkably minimal effect in blunting CD20-TDB activity. At concentrations up to 250 μg/ml rituximab-DANA, representing a 5000-fold excess over the CD20-TDB EC50 dose level, we observed only modest shifts in the dose-response curves in vitro (less than 7-fold shift at the EC50, 42 ng/ml versus 320 ng/ml). The highest concentration of rituximab-DANA in these experiments approximates the peak serum concentration (Cmax) observed with rituximab-containing clinical regimens (33). Although the EC50 for CD20-TDB was increased up to about sevenfold at the highest doses, the extent of B cell killing was not changed appreciably (Fig. 7C). These in vitro observations were recapitulated in vivo in huCD20-huCD3 double-transgenic mice, which were pretreated with rituximab-DANA (2 or 10 mg/kg) and subsequently challenged with CD20-TDB (Fig. 7D). These results are consistent with our previous finding that very low antigen expression level, or very low receptor occupancy, is required for CD20-TDB potency and that cell killing by T cells does not require the formation of a stable mature immunological synapse (30).

Pharmacologic activity of CD20-TDB is confirmed in cynomolgus monkeys

Although experiments in the aforementioned murine models demonstrating the potency of B cell depletion in the setting of robust T cell activation and expansion support the expected MOA for the CD20-TDB, the decline of T cell counts below pretreatment levels was an unexpected finding. A large range of T cell counts have been reported in clinical studies using blinatumomab, yet no such T cell loss has been reported in humans (34). To inform the likelihood of observing T cell loss in the clinic and to see whether the decrease of T cell counts at the end of CD20-TDB treatment in mice would be recapitulated in nonhuman primates, the pharmacologic activity of CD20-TDB was further tested in cynomolgus monkeys.

Three animals were treated intravenously with a single dose of CD20-TDB (1 mg/kg) (Fig. 8A). Complete B cell depletion was observed in spleen and lymph nodes 7 days after antibody administration (representative FACS plots shown in fig. S10). No apparent T cell loss was observed with both CD8+ and CD4+ T cell counts at comparable or higher levels relative to historic controls (35). We subsequently conducted a second repeat-dose study consisting of four weekly doses at 1 mg/kg, followed by an 8-week recovery period. Figure 8B shows B and T cell counts and serum CD20-TDB measurements from four individual animals over 77 days. In all tested animals, B cells were absent in blood shortly after treatment and remained absent as long as CD20-TDB serum concentrations remained above 100 ng/ml (compare animals 1 to 3 versus 4). Similar to the single-dose experiment and in contrast to results in murine models, we did not observe significant declines in T cell numbers. CD8+ T cell counts, as well as CD4+ T cell counts, increased after the first dose and then gradually returned to within 25 to 150% of baseline values over the course of the study.

Fig. 8. Anti-CD20/CD3 TDB depletes B cells in cynomolgus monkeys.

(A) The second and fourth panels show results from three cynomolgus monkeys that were treated once intravenously with CD20-TDB (1 mg/kg). Spleen and lymph nodes (LN) were collected at day 7 to determine B cell and T cell numbers by FACS. CD4+T cells are shown as black bars, CD8+T cells are shown as gray bars, and B cells are shown as hatched bars. The first and third panels show historic vehicle controls as the mean and SD of four vehicle-treated animals. (B) Four cynomolgus monkeys were treated intravenously four times weekly with CD20-TDB (1 mg/kg). Blood and serum were collected, levels of CD20-TDB were determined by ELISA, and B and T cell numbers were determined by FACS.

In cynomolgus monkeys, CD20-TDB exhibited nonlinear pharmacokinetics consistent with target-mediated clearance expected from binding to CD20 or CD3 on target cells. However, the estimated linear clearance subsequent to B cell elimination was consistent with that expected for a human IgG1 monoclonal antibody. CD20-TDB maintained good exposure throughout the treatment period, with an initial clearance of about 17 ml day−1 kg−1, which decreased to about 6 ml day−1 kg−1 by the fourth dose (days 21 to 28). When extrapolated to human, these clearance values are predicted to be consistent with a weekly to monthly intravenous dose schedule.


We describe the generation of a TDB to redirect T cells to CD20-positive B cell malignancies (CD20-TDB) that has biochemical and pharmacological properties characteristic of full-length IgG1 antibodies. We demonstrated the capacity for routine production of pharmaceutical quality drug product free of significant homodimer or multivalent aggregates that could lead to unacceptable safety through nonspecific T cell activation. In vitro B cell depletion experiments described in this report highlighted the salient mechanistic features of the CD20-TDB including the following: (i) broad activity against normal and malignant B cells expressing CD20; (ii) T cell–dependent killing via the granzyme-perforin pathway of target B cells; and (iii) CD20-TDB activity against cell lines with very low CD20 expression levels. Experiments of CD20-TDB on PBMCs collected from CLL patients showed that although the extent of CLL killing was highly correlated to T cell numbers, T cells activated by the CD20-TDB were capable of serial target cell killing, as demonstrated by the killing of CLL cells by autologous T cells at CD8+ T/CD20+ B cell ratios as low as 1:18.

Of particular note to our work, there has been a paucity of in vivo preclinical models available to support robust preclinical development of T cell–dependent antibodies. Previous studies have typically used human cancer cell lines mixed with human immune cell effectors inoculated into immunodeficient mice and immediately dosed with test agent before the formation of any significant tumor (36). We characterized the activity and potency of the CD20-TDB in four distinct in vivo models including huCD20-huCD3 double-transgenic mice, human immune-reconstituted NSG mice, a murine primary human leukemia model, and nonhuman primates. CD20-TDB administration in these models confirmed in vitro experiments, importantly demonstrating that the CD20-TDB is a conditional agonist whereby T cell activation and target cell killing are strictly contingent on CD20 antigen expression on target cells. In both the huCD20-huCD3 double-transgenic mice and cynomolgus monkeys, CD20-TDB administration resulted in rapid T cell activation and an expansion in T cell numbers, followed by a slower contraction phase that differed between species. The observed differences may be attributed in part to differences in inherent host immune biology (37). Nevertheless, rapid expansion of T cells was observed in all preclinical models and is likely to be an important mechanistic feature of the TDB that allows potent target cell killing even with relatively few T cells. In humanized NSG mice, complete B cell depletion after CD20-TDB treatment was achieved, although pretreatment T cell counts were only 10% that of B cells, as a result of an up to 10-fold increase in circulating CD8+ T cell count after CD20-TDB treatment. Finally, CD20-TDB administration in cynomolgus monkeys demonstrated a PK profile typical of IgG1 monoclonal antibodies, which importantly enable intermittent dosing schedules that allow convenient combinability with current standard therapies for B cell malignancies.

Cytokine release syndrome and neurotoxicity have been reported as serious and poorly understood side effects of blinatumomab and CD19-directed CAR-T therapies (9, 10, 12). These adverse events are potentially a class effect of redirecting T cells against B cells, although no neurotoxicity has been reported from small, compassionate-use clinical studies with FBTA05, an anti-CD20/CD3 bispecific antibody (38). In our studies with the double-transgenic mice and cynomolgus monkeys described herein, transient cytokine release, including that of IL-2, IL-6, interferon-γ, and tumor necrosis factor–α, was observed within the first 24 hours after CD20-TDB treatment (fig. S11). The observed cytokine production is only transient despite the longer half-life of the antibody and is only in the presence of target B cells. Although the objective of the studies reported here was assessment of the efficacy and pharmocodynamic effect of CD20-TDB, within the limits of the study designs (low N and limited dosing), CD20-TDB was found to be generally well tolerated. In addition, although we have demonstrated robust activity of CD20-TDB in a number of preclinical models, the clinical activity may be limited by an immunosuppressive tumor microenvironment. The animal models described in this report allow an opportunity to conduct additional studies to more fully characterize the effects of the microenvironment, as well as the impact of CD20-TDB exposure on cytokine release and safety, including exposure-response relationships and exploration of underlying mechanisms of any safety concern.

An important observation with the CD20-TDB is the preservation of activity in the presence of high concentrations of competing rituximab. This observation has several potentially important implications. First, most patients treated with rituximab-containing treatment regimens will ultimately experience disease relapse or progression indicating that acquired resistance to rituximab remains a significant barrier to treatment success. With its distinct MOA that is based on adaptive (T cell) rather than innate (natural killer cell) immune responses, CD20-TDB can potentially circumvent factors contributing to rituximab resistance. This includes potential down-regulation of CD20 expression, because minimal CD20 expression levels were sufficient for CD20-TDB activity. Second, because CD20-TDB is active in the presence of high concentrations of rituximab and requires minimal CD20 expression for activity, CD20-TDB administration into patients will not require a prolonged “wash-out” period from prior rituximab treatment, therefore avoiding treatment delays. Finally, the nonoverlapping MOAs of CD20-TDB and other anti-CD20 targeting antibodies allow for potential combination, enabling flexible treatment strategies to incorporate CD20-TDB into current standard of therapy for B cell malignancies.

In conclusion, in vitro and in vivo results described in this report highlight the potency and specificity of the CD20-TDB. Together with its favorable and predictable PK profile and demonstration of manufacturing feasibility and scalability, characterization of CD20-TDB safety profile in large animal models and clinical assessment for the treatment of B cell malignancies are warranted.


Antibody production

CD20-TDB was produced as full-length human IgG1 in a knobs-into-holes format as previously described (39). The purified antibodies ran as a single peak (>99% of the signal) in gel filtration with less than 0.2% aggregates. No homodimers were detected by mass spectrometry. The CD20 arm in the bispecific is anti-CD20 clone 2H7, whereas the CD3 arm is either clone UCHT1 or a cynomolgus cross-reactive clone GMX3c (for Fig. 8 only).

In vitro B cell killing and T cell activation assays

B tumor cell lines were obtained from American Type Culture Collection, and PBMCs were isolated from whole blood of healthy donors by Ficoll separation. CD4+T and CD8+T cells were separated with Miltenyi kits according to the manufacturer’s instructions. Cryopreserved PBMCs from CLL patients were purchased from Conversant Bio. Cells were cultured in RPMI 1640 supplemented with 10% fetal bovine serum (Sigma-Aldrich). For B cell killing assays, live B cells were gated as PICD19+ B cells by FACS, and absolute cell count was obtained with fluorescein isothiocyanate beads added to the reaction mix as internal counting control. Activated T cells were detected by CD69 and CD25 surface expression. Intracellular granzyme B induction was detected by FACS. Perforin concentration in medium was detected by ELISA (eBioscience). All antibodies were purchased from BD Biosciences.

In vivo efficacy studies in murine models

All in vivo experimental procedures conformed to the guiding principles of the American Physiology Society and were approved by Genentech’s Institutional Animal Care and Use Committee. Humanized NSG mice were purchased from The Jackson Laboratory. Human CD20 transgenic mice and human CD3 transgenic mice were generated as previously described, and huCD20-huCD3 double-transgenic mice were produced by crossing mice containing each of the two single transgenes (18, 19). Whole blood was collected by puncture of the retro-orbital sinus using heparinized pipets and immediately transferred into heparinized tubes, while the animals were under anesthesia, or by terminal cardiac puncture with a heparinized syringe after CO2 euthanasia. Spleens were collected after CO2 euthanasia. For all studies, clinical observations were performed twice per week to monitor the health of the animals. Animal body weights were taken at least once a week. PBMCs were isolated after red blood cell lysis and analyzed by FACS for B cells [murine CD45+CD19+ (muCD45+CD19+)] and T cells (muCD90.2+CD4+ or muCD90.2+CD8+). All antibodies used were purchased from BD Biosciences and eBioscience.

Pharmacokinetic/pharmacodynamic study in cynomolgus monkeys

All cynomolgus monkey studies were conducted at Charles River Laboratories (Reno, NV) using purpose-bred, naïve, cynomolgus monkeys of Chinese origin. For the single-dose study, three male cynomolgus monkeys were administered a single slow bolus intravenous dose of CD20-TDB (1 mg/kg). For the repeat dose study, four cynomolgus monkeys were administered a slow bolus intravenous dose of CD20-TDB (1 mg/kg) once weekly for a total of four doses. Whole blood or tissues were collected at selected time points for B and T cell counts by FACS. Serum was collected and stored at −70°C until assayed using an ELISA to determine the amount of test article in each serum sample. Serum concentration–time profiles from each animal were used to estimate PK parameters using WinNonlin software (Pharsight).

Statistical analyses

Data are presented as means ± SD or means only as stated in the figure legends. Statistically significant differences were tested by specific tests as indicated in the figure legends. P values are calculated by unpaired t test with Prism software version 5.0 (GraphPad). Sample sizes for all studies are summarized in fig. S12.


Fig. S1. Quality control analysis for anti-CD20/CD3 TDB.

Fig. S2. Summary of PK analysis for anti-CD20/CD3 TDB in rats.

Fig. S3. Representative FACS data for Fig. 1C.

Fig. S4. In vitro CD8+ T cell proliferation in the presence of CD20-TDB and BJAB.

Fig. S5. Data with three healthy human donors for statistical analysis for Fig. 1 (D to F).

Fig. S6. FACS gating strategy for B and T cells with blood and spleen samples of huCD20-huCD3 double-transgenic mice.

Fig. S7. Presence of human B and T cells in humanized NSG mice and antigen expression level for human CD20 and CD3.

Fig. S8. CD4+ T cell counts in humanized NSG mice upon CD20-TDB treatment.

Fig. S9. Control studies with rituximab and rituximab-DANA antibodies.

Fig. S10. FACS gating strategy for B and T cells with blood and tissue samples of cynomolgus monkeys.

Fig. S11. (A and B) Cytokine production in huCD20-huCD3 double-transgenic mice (A) and in cynomolgus monkeys (B) upon CD20-TDB treatment.

Fig. S12. Summary of sample sizes for all presented studies.


  1. Acknowledgments: We thank the donors and patients who consented to the use of their cells for the studies; B. Kelley, G. Nakamura, T. Sai, C. Spiess, T. Lombana, A. Yee, and T. Wong for their help in antibody generation; A. Oldendorp for facilitating the studies in cynomolgus monkeys; M. Byrtek for help with statistical analyses; and M. Nicoll and F. De Sauvage for critical reading of the manuscript and helpful discussions. Funding: Supported by Genentech Inc. Author contributions: L.L.S., Y.-W.C., and A.J.E. wrote the manuscript; L.L.S., E.S., H.W., N.D., N.C., T.J., K.T., and A.J.E. designed and supervised the studies; L.L.S., M. Hristopoulos, Y.L., X.Y., C.J., M. Huseni, A.R., E.M., J.Y., R.C., and P.W. executed the studies and analyzed assay data; and L.L.S., A.J.E., D.E., M.M., X.W., X.C., Y.C., J.M.S., and M.S.D. designed and produced the antibodies. Competing interests: L.L.S., M. Hristopoulos, D.E., M.M., Y.L., C.J., M. Huseni, A.R., E.S., E.M., J.Y., X.W., X.C., Y.C., R.C., H.W., N.D., Y-W.C., J.M.S., T.J., K.T., and M.S.D. are all employees of Genentech Inc.; X.Y. and N.C. have no conflict of interest to disclose; A.J.E. was an employee of Genentech at the time the studies were conducted and is currently an employee and shareholder at Juno Therapeutics. Genentech has filed patent applications related to this work on TDBs. Investigators may request materials from Genentech by submitting a request form at
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article