Research ArticleCancer

An anti–glypican 3/CD3 bispecific T cell–redirecting antibody for treatment of solid tumors

See allHide authors and affiliations

Science Translational Medicine  04 Oct 2017:
Vol. 9, Issue 410, eaal4291
DOI: 10.1126/scitranslmed.aal4291

Double trouble for solid tumors

Because the endogenous immune response is not enough to clear a patient’s cancer, therapies are being designed to redirect T cells to tumor cells. This can be done by engineering the cells ex vivo, such as in CAR T cell therapy, or in vivo, such as with bispecific antibodies. Ishiguro et al. describe the development and preclinical testing of a bispecific antibody recognizing CD3 and glypican 3, a common antigen on solid tumors. This bispecific antibody was effective in a variety of mouse cancer models, even when treatment was initiated after the tumor was quite large. Treatment also appeared to be safe when administered to monkeys. These results suggest further development of this antibody for therapeutic use in multiple cancer types.


Cancer care is being revolutionized by immunotherapies such as immune checkpoint inhibitors, engineered T cell transfer, and cell vaccines. The bispecific T cell–redirecting antibody (TRAB) is one such promising immunotherapy, which can redirect T cells to tumor cells by engaging CD3 on a T cell and an antigen on a tumor cell. Because T cells can be redirected to tumor cells regardless of the specificity of T cell receptors, TRAB is considered efficacious for less immunogenic tumors lacking enough neoantigens. Its clinical efficacy has been exemplified by blinatumomab, a bispecific T cell engager targeting CD19 and CD3, which has shown marked clinical responses against hematological malignancies. However, the success of TRAB in solid tumors has been hampered by the lack of a target molecule with sufficient tumor selectivity to avoid “on-target off-tumor” toxicity. Glypican 3 (GPC3) is a highly tumor-specific antigen that is expressed during fetal development but is strictly suppressed in normal adult tissues. We developed ERY974, a whole humanized immunoglobulin G–structured TRAB harboring a common light chain, which bispecifically binds to GPC3 and CD3. Using a mouse model with reconstituted human immune cells, we revealed that ERY974 is highly effective in killing various types of tumors that have GPC3 expression comparable to that in clinical tumors. ERY974 also induced a robust antitumor efficacy even against tumors with nonimmunogenic features, which are difficult to treat by inhibiting immune checkpoints such as PD-1 (programmed cell death protein–1) and CTLA-4 (cytotoxic T lymphocyte–associated protein–4). Immune monitoring revealed that ERY974 converted the poorly inflamed tumor microenvironment to a highly inflamed microenvironment. Toxicology studies in cynomolgus monkeys showed transient cytokine elevation, but this was manageable and reversible. No organ toxicity was evident. These data provide a rationale for clinical testing of ERY974 for the treatment of patients with GPC3-positive solid tumors.


Over the past decade, extraordinary clinical benefits of cancer immunotherapies have been demonstrated, such as durable responses even for late-stage cancers. In particular, the clinical outcomes of immune checkpoint inhibitors such as anti–CTLA-4 (cytotoxic T lymphocyte–associated protein–4) and anti–PD-1 (programmed cell death protein–1) antibodies have been impressive, and they are changing the paradigm of cancer treatment (1, 2). Although these therapies have shown promising clinical responses when used for a broad range of solid tumors, the number of patients who benefit from them is limited (3, 4), and it is necessary to predict which patients will be responsive to such treatments. Many ongoing studies to identify predictive biomarkers have suggested the usefulness of tumor mutation-derived antigens (neoantigens) to select patients who can benefit from treatment with checkpoint inhibitors (5). For example, Rizvi et al. (6) reported that neoantigen burden was positively correlated with clinical benefit in patients with non–small cell lung cancer receiving anti–PD-1 antibody. These studies have provided evidence that neoantigen-reactive T cells are the most important players in the responses to checkpoint inhibitors. However, these advances do not address a fundamental problem with checkpoint inhibitors: immune checkpoint inhibitors will not work if endogenous T cells cannot recognize the cancer cells via T cell receptors (TCRs) because of a lack of neoantigens. Tumors with this characteristic are called “nonimmunogenic” tumors, and targeting them is the next challenge for cancer immunotherapy (7).

One strategy to mitigate this problem is T cell–redirecting antibody (TRAB) technology, which bispecifically engages CD3 on T cells and antigens on cancer cells; this binding activates T cells to kill cancer cells (810). TRAB technology is highly potent because it can theoretically redirect all T cells in the body to cancer cells regardless of the intrinsic antigen specificity of the TCR. Moreover, it has been reported that TRABs can harness not only CD8+ T cells but also CD4+ T cells, including regulatory T cells, as effector cells to kill cancer cells (11). The therapeutic potential of TRABs is exemplified by blinatumomab, a bispecific T cell engager (BiTE) targeting CD19 and CD3 for the treatment of hematological malignancies (12). The U.S. Food and Drug Administration granted approval of blinatumomab based on a phase 2 study for treatment of patients with Philadelphia chromosome–negative (Ph) relapsed or refractory B cell precursor acute lymphoblastic leukemia (13). Several other promising TRABs such as anti-CD20/CD3 and anti-CD123/CD3 are being developed to target hematological malignancies (14, 15). TRABs targeting solid tumors are also under clinical development, including solitomab, which targets the epithelial cell adhesion molecule (EpCAM) (16), or those against carcinoembryonic antigen (17) or prostate-specific membrane antigen (18). However, these TRABs are still not at the stage of showing clinical benefits. Given the potent cytotoxic potential of TRABs, the key to achieving clinical benefit is to avoid “on-target off-tumor” toxicity, which is caused by low expression of the target (on-target) in normal tissues (off-tumor). It is well known that classical tumor antigens such as EpCAM, epidermal growth factor receptor (EGFR), and human EGFR 2 (HER2) are not tumor-specific; therefore, on-target off-tumor toxicity could be a problem. The administration of solitomab was limited by toxicity induced by the triggering of a response in EpCAM+ normal tissues such as bowel epithelia and bile ducts (19). In addition, an anti-EGFR BiTE was not tolerated in cynomolgus monkeys (20). Therefore, the development of TRABs targeting solid tumors has been hampered by the lack of an appropriate antigen that has high tumor specificity.

Glypican 3 (GPC3) is a highly tumor-specific antigen. GPC3 belongs to a family of heparin sulfate proteoglycans that are tethered to the cell surface via a glycosylphosphatidylinositol anchor (21). An analysis using GPC3-deficient mice suggested that GPC3 is involved in the regulation of Wnt, hedgehog, and fibroblast growth factor pathways to control cell growth and apoptosis in particular cell types during development (2224). GPC3 is a fetal protein that is expressed in a wide variety of tissues during embryonic development (25, 26); however, its expression is strictly suppressed in most adult tissues (27). In contrast, elevated GPC3 expression has been reported in a wide variety of tumor types such as liver (28, 29), lung (30), gastric (31), ovarian (32, 33), esophageal (34), and others (35). These features suggest that GPC3 is an ideal target for TRABs in these solid tumors.

Here, we generated an anti-GPC3 TRAB, ERY974, which is a fully humanized immunoglobulin G (IgG)–based bispecific antibody. Using this agent, we observed striking antitumor efficacy in vitro and in in vivo xenograft models. We also examined its antitumor efficacy in an immunocompetent mouse model that mimics cancer patients who are insensitive to checkpoint inhibitors. Using this approach, we revealed that ERY974 is effective against nonimmunogenic tumors. In addition, we report the results of pharmacokinetic (PK) and toxicological studies performed in cynomolgus monkeys, which suggested clinical safety for ERY974.


Generation of the anti-GPC3/CD3 bispecific antibody ERY974

ERY974 was generated by combining two antibodies, specifically an anti-GPC3 antibody (clone GC33) that we previously obtained (36) and an anti-CD3 antibody (clone CE115) that was newly obtained by immunizing a rat with human CD3ε (CD3E) recombinant protein (Fig. 1A). A common light chain was generated by complementarity-determining region and framework region (CDR/FR) shuffling as previously described (37). First, light-chain CDRs from both parental antibodies (anti-GPC3 or anti-CD3) were shuffled and combined with human germline FRs to form an assembly of light chains. Screening of the assembly by binding assay identified a lead common light chain that confers an antibody of the strongest affinity against GPC3 or CD3E when combined with the heavy chains from the respective parental antibodies. The variable domains of the two heavy chains were humanized as previously described (38, 39). Various sets of anti-GPC3 and anti-CD3 arms having different affinities against each antigen were generated and produced in a bispecific antibody format. These were assessed by in vitro T cell–dependent cellular cytotoxicity (TDCC) and in vivo antitumor efficacy in xenograft models described below. The combination of arms that showed the strongest antitumor efficacy was finally selected for ERY974 (Fig. 1A). For the heavy-chain constant region, L235R/S239K/N297A (in EU numbering) mutations were introduced into the CH2 region of human IgG4 to abolish binding to Fcγ receptors (Fig. 1A). This was to avoid GPC3-independent cytokine release by engaging Fcγ receptors and CD3. To facilitate heterodimerization of the two heavy chains by changing the CH3 interface electrical charge (40), we introduced E356K and K439E mutations on each heavy chain. A K196Q mutation was also introduced on the heavy chain of the GPC3 arm to reduce the isoelectric point (pI) to facilitate separation of the homodimers of each heavy chain by ion chromatography during the manufacturing process (39). In addition, other mutations to achieve pI difference between the two heavy chains, decrease nonspecific binding, reduce deamidation, and minimize immunogenicity were also introduced to facilitate bispecific antibody manufacturing and improve physicochemical and safety properties (37). The resulting molecule, termed ERY974, was confirmed to specifically bind membranous human GPC3 and CD3E by flow cytometry analysis using cell lines overexpressing either antigen (Fig. 1B).

Fig. 1. Generation of ERY974.

(A) Schematic illustration of ERY974 structure and the introduced mutations. The two Fab arms share a common light chain, depicted in green, with a variable domain. (B) Flow cytometry results showing specific binding of ERY974 to GPC3 and CD3E in SK-HEP-1 (a parental human liver/ascites adenocarcinoma cell line), SK-HEP-1/hGPC3, (SK-HEP-1 cells that overexpress human GPC3), BAF3 (a parental mouse pro–B cell line), and BAF3/CD3ε (BAF3 cells that overexpress human CD3E). The blue lines indicate staining with a control IgG4 antibody, and the red lines indicate staining with ERY974. FITC, fluorescein isothiocyanate.

GPC3 expression in tumors and normal tissues

To investigate the GPC3 protein expression profile in tumors and normal tissues, we performed immunohistochemistry (IHC) using an anti-GPC3 mouse antibody (mGC33, mouse IgG2a) that specifically detects GPC3. Tissue microarray (TMA) cores from 31 hepatocellular carcinomas, 40 lung squamous cell carcinomas, 69 lung small cell carcinomas, 87 esophagus squamous cell carcinomas, 30 cardiac adenocarcinomas, 40 gastric cancers, and 68 head and neck cancers (Fig. 2 and tables S1 to S7), as well as the cores from 30 different normal tissues (Fig. 2 and table S8), were analyzed. In tumor tissues, 90% (28 of 31) of hepatocellular carcinomas were positive for GPC3 in the cytoplasm or plasma membrane. Similarly, 65% (26 of 40) of lung squamous cell carcinomas, 64% (44 of 69) of lung small cell carcinomas, 28% (24 of 87) of esophagus squamous cell carcinomas, 30% (9 of 30) of cardia adenocarcinomas, 20% (8 of 40) of gastric cancers, and 28% (19 of 68) of head and neck cancers appeared positive for GPC3. In contrast, GPC3 was not detected in normal tissues, except for the endometrium and placenta.

Fig. 2. GPC3 expression in normal and tumor tissues.

IHC analysis of an anti-GPC3 mouse antibody (mGC33) in a multi-tissue array. Representative photomicrographs of GPC3 staining in tumor and normal tissues are shown. Cytoplasmic H-scores are indicated in parentheses. H-scores were calculated as values between 0 and 300, defined as [1 × (percentage of cells staining at 1+ intensity) + 2 × (percentage of cells staining at 2+ intensity) + 3 × (percentage of cells staining at 3+ intensity) = H-score]. Photos were taken using a 20× objective lens.

In vitro pharmacological profiles of ERY974

In vitro pharmacological profiles of ERY974 were evaluated using SK-HEP-1 cells, a GPC3-negative cancer cell line, and SK-HEP-1/hGPC3 cells, which were SK-HEP-1 cells transduced to express human GPC3 on their cell surface, with human peripheral blood mononuclear cells (PBMCs) as effector cells. Concentration-dependent TDCC was seen against SK-HEP-1/hGPC3 cells but not against SK-HEP-1 cells (Fig. 3A). The same results were also confirmed by monitoring cancer cell growth for 72 hours by using an xCELLigence Real-Time Cell Analysis system (fig. S1). We also observed that the expression of CD69, an early marker of T cell activation, and that of CD25, a late marker of T cell activation, were both elevated only in T cells incubated with SK-HEP-1/hGPC3 cells (Fig. 3B). The aforementioned GPC3-dependent activation of T cells was also seen when tested with PC-10 cells in which GPC3 is endogenously expressed, but not with SNU-16 cells in which GPC3 expression was not detected (fig. S2). We next examined the characteristics of T cell activation by ERY974 using purified CD4+ or CD8+ T cells from two different donors (Fig. 3C). We found that not only CD8+ T cells but also CD4+ T cells induced clear TDCC, although the degree of CD4+ T cell–mediated TDCC was lower than that with CD8+ T cells. After 3 days of activation by ERY974, the absolute number of CD4+ or CD8+ T cells increased (Fig. 3D), and a carboxyfluorescein diacetate succinimidyl ester (CFSE) cell division assay revealed that ERY974 induces proliferation in almost all populations of T cells, which suggests that ERY974 can induce polyclonal activation of T cells independently of TCR specificity (Fig. 3E).

Fig. 3. GPC3-dependent TDCC and polyclonal T cell activation induced by ERY974.

(A) Target cells and PBMCs were incubated with various concentrations of ERY974 for 24 hours. A keyhole limpet hemocyanin (KLH)–TRAB antibody was used as a negative control. TDCC was measured as described in Materials and Methods. Data represent means ± SD (n = 3). (B) T cell activation was assessed by measuring CD25 and CD69 levels on CD3+ T cells by flow cytometry. Data represent means ± SD (n = 3). (C) TDCC of ERY974 elicited by CD4+ or CD8+ T cells was measured. Data represent means ± SD (n = 3). (D) Target cells and CFSE-labeled PBMCs were incubated with ERY974 (1 μg/ml) or KLH-TRAB for 3 days. Total numbers of viable CD4+ or CD8+ T cells at day 0 (immediately after reaction started) and day 3 were counted by flow cytometry. Data represent means ± SD (n = 3). (E) Red lines represent the CFSE profiles in the presence of ERY974, and blue lines represent the profile in the presence of KLH-TRAB. The histograms show one representative result of triplicate assays.

In vivo antitumor efficacy of ERY974 against various types of tumors

We then examined the in vivo antitumor efficacy of ERY974 against GPC3-positive cell lines derived from various types of cancer (MKN-74 from gastric adenocarcinoma, PC-10 from lung squamous cell carcinoma, TOV-21G from ovarian clear cell carcinoma, HuH-7 from hepatocellular carcinoma, KYSE70 from esophageal squamous cell carcinoma, and SCC152 from hypopharynx squamous cell carcinoma). These six cell lines were shown to have various levels of GPC3 expression by quantitative flow cytometry; PC-10 had the highest (1.21 × 105 molecules per cell) and MKN-74 had the lowest (2.97 × 103 molecules per cell) expression (Fig. 4A). To compare the expression of GPC3 in these two cell lines in in vivo setting with those in clinical tumor samples, we inoculated these cell lines into nonobese diabetic/severe combined immunodeficiency (NOD-SCID) mice and stained xenograft tissues using the mGC33 antibody under the same conditions used for TMA analysis as described above. Staining intensity was consistent with the levels observed by flow cytometry using in vitro cultured cells; stronger GPC3 staining was observed in PC-10 xenograft tissue (H-score in cytoplasm = 240), and staining in MKN-74 xenograft tissue was weaker (H-score in cytoplasm = 20) (Fig. 4B). The staining intensity in these two xenograft tissues was comparable to the range observed in clinical tumors shown in Fig. 2.

Fig. 4. Antitumor efficacy of ERY974 against various cancer types.

(A) GPC3 expression in various cancer cell lines determined by quantitative flow cytometry. (B) IHC results of GPC3 staining in xenograft tissues of indicated cancer cell lines. H-scores in the cytoplasm are indicated in parentheses. Images were taken using a 20× objective. (C) Antitumor efficacy of ERY974 against various xenograft tumors in NOD-SCID mice inoculated with human T cells. Mean tumor volume values are shown + SD (n = 5). *P < 0.05 between vehicle and ERY974 group by Wilcoxon test. Arrows indicate timing of ERY974 administration. (D) Antitumor efficacy of ERY974 against large tumors. When the mean volume of implanted KYSE70 tumors reached more than 600 mm3, vehicle or ERY974 (1 mg/kg) was administered 3 to 5 hours after the T cell injection. Mean tumor volumes in each group are shown + SD (n = 4). Images represent tumor burden over time in one representative animal dosed with ERY974 on day 29. (E) Antitumor efficacy of ERY974 against GPC3-negative cell xenograft tumors. Mean values for tumor volume in each group are shown + SD (n = 5).

For evaluation of the in vivo antitumor efficacy of ERY974, the six aforementioned cell lines were inoculated into NOD-SCID mice. After the tumor volume reached around 200 mm3, the mice were treated with ERY974. Because ERY974 has no cross-reactivity with mouse CD3, human T cells were intraperitoneally injected into mice as effector cells on the same day of ERY974 administration (human T cell–injected model). With a single administration of ERY974, marked tumor growth inhibition (TGI) was observed (Fig. 4C). The TGI (%) values of PC-10, KYSE70, HuH-7, SCC152, TOV-21G, and MKN-74 cells were 119, 127, 106, 97, 107, and 112%, respectively, indicating tumor shrinkage. Using ERY974, marked tumor regression was also observed against relatively large tumors (Fig. 4D). In contrast, antitumor efficacy was not observed against GPC3-negative SK-HEP-1 cells (Fig. 4E).

Antitumor efficacy of ERY974 in immunocompetent human CD3 transgenic mice

We then explored the antitumor efficacy of ERY974 in an immunocompetent mouse model. Because ERY974 is not cross-reactive to murine Cd3ε, we created a human CD3 transgenic mouse in which all three components of the Cd3 complex, Cd3ε, Cd3δ, and Cd3γ, were replaced by their human counterparts, CD3E, CD3D, and CD3G (41). As for target tumor cells to be used for immunocompetent mouse models, we introduced the human GPC3 gene into two mouse tumor cell lines, Hepa1-6 (hepatoma) and LLC1 (Lewis lung carcinoma), and established cell lines that stably express human GPC3 (Hepa1-6/hGPC3 and LLC1/hGPC3, respectively).

We inoculated human GPC3–expressing mouse tumor cells into human CD3 transgenic mice and established syngeneic tumor models. We then evaluated the immunogenicity of the tumors, that is, the ability to induce an adaptive immune response, by examining immunohistochemically the levels of tumor-infiltrating T cells and PD-L1 expression in the syngeneic tumor tissues (Fig. 5A and table S9). IHC analysis in Hepa1-6/hGPC3 tumors (at day 10) revealed high immunogenicity, specifically moderate infiltration of CD3+ T cells and positive PD-L1 expression. In contrast, LLC1/hGPC3 tumors (at day 14) showed low immunogenicity. The levels of tumor-infiltrating T cells were low, and PD-L1 was negative in tumor tissues. We further examined the degree of immunogenicity in these syngeneic tumors with RNA sequencing (RNA-seq)–based gene expression analysis. A group of immune-related genes was selected on the basis of a previous report that investigated the immunogenicity of implanted mouse syngeneic tumors (42). The gene set included markers of immune cell infiltration, indicators of dendritic cell and T cell activation, mediators of immune suppression, vasculature-related factors, and immune function–related transcription factors, as shown in Fig. 5B. Higher expression of these genes suggests higher immunogenicity of the tumor. The expression of these genes in Hepa1-6/hGPC3 tumors was higher than that in LLC1/hGPC3 tumors. This supports the aforementioned IHC results, in which Hepa1-6/hGPC3 tumors exhibited high immunogenicity, whereas LLC1/hGPC3 tumors displayed low immunogenicity.

Fig. 5. Antitumor efficacy of ERY974 in immunocompetent human CD3 transgenic mice.

(A) Histopathological analysis of Hepa1-6/hGPC3 and LLC1/hGPC3 tumors. Tumor tissue samples taken 3 days after administering vehicle or ERY974 (5 mg/kg) were stained as indicated. H&E, hematoxylin and eosin. (B) Gene expression analysis in Hepa1-6/hGPC3 and LLC1/hGPC3 tumors. RNA from tumors treated with vehicle or ERY974 was used for RNA-seq. Each group was tested in triplicate (n = 3). Z scores were calculated using log2-transformed fragments per kilobase of exon per million mapped fragments values for all target genes. (C) Antitumor efficacy of ERY974 and immune checkpoint inhibitors. Values represent means + SD (n = 5). *P < 0.05 between vehicle group and the antibody treatment group at day 25 determined by Dunn’s multiple comparisons test. Arrows indicate the timing of antibody administration.

Next, the antitumor efficacy of four antibodies, ERY974, anti-mouse CTLA-4, anti-mouse PD-1, and anti-mouse PD-L1, was examined using the syngeneic tumor models with human CD3 transgenic mice. In Hepa1-6/hGPC3 tumors, which are predicted to be immunogenic, ERY974 showed marked efficacy (P = 0.0068) because complete remission of tumors was observed in most mice. Treatment with the anti–PD-1 antibody also showed significant efficacy (P = 0.0208), and anti–CTLA-4 and anti–PD-L1 antibodies showed signs of efficacy (Fig. 5C). In contrast, in the nonimmunogenic LLC1/hGPC3 tumors, only ERY974 showed statistically significant antitumor efficacy (P = 0.0003); anti–CTLA-4, anti–PD-1, and anti–PD-L1 antibodies did not show efficacy (Fig. 5C).

To examine the mechanism by which ERY974 changed the tumor microenvironment, we analyzed the histopathological features and gene expression of the aforementioned immune-related gene set in the tumors. As shown in Fig. 5A and table S9, in both Hepa1-6/hGPC3 and LLC1/hGPC3 tumors, a higher degree of immune cell infiltration, such as that of T cells and granulocytes, was observed by histopathological analysis in tumors treated with ERY974 compared to that in tumors treated with vehicle. In addition, a higher degree of PD-L1 expression was observed in tumors treated with ERY974. GPC3 expression and the degree of FoxP3+ cell infiltration were not changed by ERY974 treatment. Consistently, treatment with ERY974 resulted in further up-regulation of immune-related genes in both tumor types when compared with vehicle treatment (Fig. 5B).

Toxicology and PK of ERY974 in cynomolgus monkeys

To extrapolate the PK and safety of ERY974 to humans, the cynomolgus monkey was selected on the basis of similar binding affinities to GPC3 and CD3 (Fig. 6A) and similar median effective concentration values for TDCC activity (human PBMCs, 0.054 nM; cynomolgus monkey PBMCs, 0.11 nM). We first conducted a single-dose study with low doses and slow intravenous infusion (0.1, 1, and 10 μg/kg over 30 min; n = 3 per sex per dose), and the PK parameters were calculated after excluding the data of anti-ERY974–positive animals (6 of 18 animals; Table 1 and fig. S3). ERY974 exhibited biphasic disposition and a linear PK profile in the range of 0.1 to 10 μg/kg. The total body clearance (CLtotal) was independent of dose, and the elimination half-life (t1/2) was 2.89 to 3.87 days. The maximum plasma concentration (Cmax) and the area under the plasma concentration–time curve from zero to infinity (AUCinf) exhibited a dose-proportional increase from 0.1 to 10 μg/kg. The steady-state volume of distribution (Vd,ss) was 104 to 123 ml/kg, which is less than the volume of extracellular fluid in monkeys, suggesting that ERY974 has low tissue penetration (43). On the basis of these properties, by a single-animal species allometric scaling method with a fixed exponent of 0.85 for clearance and 1.0 for volume of distribution (44), t1/2 in humans was predicted to be 5.1 days.

Fig. 6. Safety assessment in cynomolgus monkey.

(A) Binding affinity of ERY974 to human or cynomolgus monkey epitope peptides derived from GPC3 or CD3 was measured by surface plasmon resonance analysis. KD, dissociation constant. (B) Change in serum interleukin (IL)–6 concentration over time in a single-dose study. Five animals per sex per group received a single administration of ERY974. Necropsies were performed on day 22 after the dose. (C) Serum IL-6 concentration (red and black lines) over time in a repeated-dose study. Three animals per sex received doses (blue bars) that increased daily by about threefold to 1.5 mg/kg. Necropsies were performed on day 10.

Table 1. PK parameters of ERY974 in cynomolgus monkeys after a single intravenous infusion of ERY974.

Animals that developed anti-drug antibodies (ADA) were excluded from the analysis. The mean of three animals is shown. MRT, mean residence time.

View this table:

Here, a transient increase in blood cytokines was observed from 1 μg/kg and became more pronounced at a 10-fold higher dose, with IL-6 being the most prominent cytokine (Fig. 6B). Signs of a deteriorating general condition such as red skin, reduced food consumption, and body weight loss were also noted in a dose-dependent manner. The relevant effects on clinical pathology such as elevation of C-reactive protein, decreased red blood cell count, and increased white blood cell numbers were mostly limited to the highest dose. Histopathological findings such as decreased lymphocytes in the thymus and increased immune cell infiltration in multiple tissues were also seen at the highest dose and were consistent with an inflammatory response. However, clear cytotoxic changes were not detected, and the findings were transient, with the animals exhibiting rapid recovery.

In the second study, we evaluated the safety profile at a higher dose. Animals (n = 3 per sex per dose) were treated with a dose of 0.1 μg/kg on the first day, and then the dosage was increased about threefold per day up to 1500 μg/kg (Fig. 6C). In this period, the plasma concentration of ERY974 increased with dosage, and Cmax reached a value that was about 100-fold higher than that at the highest dose (10 μg/kg) in the previous single-dose study. The ERY974-related transient increase in cytokine levels in the blood and the symptoms observed in the single-dose study were transiently noted again during this dosing at about 1 to 10 μg/kg, with the average cytokine peak concentration being about fivefold lower than that after single dosing (Fig. 6C). One male showed similar but relatively severe signs during the period of increasing doses, and this animal was subjected to necropsy after receiving a dose of 100 μg/kg. The blood cytokine levels in this individual were similar to those in others, and clear cytotoxic changes and organ toxicity were not detected. In the other five animals, dosing was tolerated until the target dose of 1500 μg/kg. The main pathological findings were similar to those in the single-dose study, and no clear cytotoxic changes or organ toxicities were evident.

Effect of steroid premedication on antitumor efficacy and cytokine release

Premedication with corticosteroids, as represented by dexamethasone (DEX), is a common practice in the clinic to prevent infusion reactions including cytokine release syndrome (CRS) (45). Thus, we examined the influence of DEX premedication on cytokine release caused by ERY974 treatment and its antitumor efficacy using a humanized NOD/Shi-scid, IL-2Rγnull (huNOG) mouse model, in which CD34+ human hematopoietic stem cells (HSCs) were transplanted to reconstitute human immune cells. About 100 days after HSC transplantation, each humanized mouse was examined for the presence of human CD45+ leukocytes, human CD3+ T cells, and murine CD45+ leukocytes in the mouse peripheral blood by flow cytometry. Humanization was successfully achieved; the percentage of human leukocytes among all leukocytes was 53.9 to 93.8%, and the percentage of human T cells among human leukocytes was 4.4 to 20.7%. Next, we compared ERY974 dose dependency for antitumor efficacy and that for cytokine release (Fig. 7, A and B). ERY974 was administered as a single dose of 1, 0.2, 0.04, or 0.008 mg/kg 4 weeks after inoculation of PC-10 cancer cells. Substantial antitumor efficacy was observed at all dosages. Among the cytokines tested (IL-2, IL-4, IL-6, IL-10, tumor necrosis factor, and interferon-γ), transient induction of IL-6 and IL-2 was notable, similar to that observed in cynomolgus monkeys. Both antitumor efficacy and cytokine release were dose-dependent and showed similar dose responses. We next examined the effect of premedication by administering DEX at 18 hours and 1 hour before the first administration of ERY974. DEX premedication almost completely inhibited cytokine release, whereas it did not show any suppressive effect on the antitumor efficacy of ERY974 (Fig. 7, C and D). After the second administration of ERY974 on day 7, cytokine elevation was not observed, although DEX was not given before that time point (Fig. 7D). These results were the same at a lower ERY974 dose with submaximal efficacy (fig. S4).

Fig. 7. Effect of premedication on antitumor efficacy and cytokine release.

(A) Antitumor efficacy of ERY974 in huNOG mice. Mean tumor volume values are shown + SD (n = 4 or 5). *P <0.05 between vehicle and ERY974 groups by Dunnett’s multiple comparisons test. Arrows indicate ERY974 administration. (B) IL-6 and IL-2 induction. Blood was collected 17 hours before and 2, 6, and 24 hours after ERY974 administration. Plasma concentrations of IL-6 and IL-2 are shown. Values represent means + SD (n = 5). (C) Influence of DEX premedication on antitumor efficacy. huNOG mice bearing tumors were administered vehicle, DEX premedication alone, or ERY974 (1 mg/kg) with or without DEX premedication. Mean tumor volume values are shown + SD (n = 4 or 5). *P < 0.05 between the vehicle and ERY974 groups by Dunnett’s multiple comparisons test. (D) Effect of DEX on induction of IL-6 and IL-2. Blood was collected 19 hours before and 6 and 24 hours after the first dose of ERY974, and 19 hours before and 6 hours after the second dose (149 and 174 hours after the first dose, respectively) of ERY974. Values represent means + SD (n = 5).


Here, we generated ERY974, a bispecific antibody that redirects the cytolytic activity of T cells against tumor cells expressing the highly tumor-specific antigen GPC3. ERY974 has a whole IgG-like structure with two distinct heavy chains and one common light chain. The presence of a common light chain completely circumvents mismatched pairing of the heavy chain and light chain, which is a general issue for manufacturing a bispecific antibody consisting of two distinct heavy and light chains. In addition to the common light chain, another set of mutations that facilitate the heterodimerization of the two heavy chains and modify the pI to allow straightforward separation of mispaired homodimers by ion chromatography was also introduced to ERY974. These mutations help ensure a robust and efficient manufacturing process for a whole IgG-like bispecific antibody. This set of antibody engineering technologies, termed ART-Ig technology, has been applied to our previous bispecific antibody drug named ACE910/emicizumab (37), and similar to that case, large-scale production of good manufacturing practice-grade ERY974 has been successfully accomplished.

ERY974 has a whole IgG-like structure with intact neonatal Fc receptor (FcRn) binding properties of the Fc portion, which allows it to have a longer plasma half-life than non-IgG formats. From the cynomolgus monkey PK study, the plasma half-life of ERY974 in humans was predicted to be 5.1 days; this enables a weekly or biweekly dose regimen in the clinic. This is one of the advantages of ERY974 having a whole IgG-like structure because this is not feasible for bispecific molecules such as BiTE that lack FcRn binding ability and as a result have poor PK. As an alternative approach to a bispecific antibody, several chimeric antigen receptor T cell–based therapies targeting GPC3 are being developed (46, 47). However, considering that a T cell–based therapy requires ex vivo immune cell manipulation, ERY974 is expected to have an advantage over such approaches, especially with respect to manufacturing and access.

Because TRABs induce cytotoxicity against target-expressing cells, one issue that limits their clinical use is undesired cytotoxicity against normal tissues that express the target molecule. It is crucial to select a highly tumor-specific antigen as the target to avoid such effects; however, only a few antigens are currently known to meet the strict criteria required for TRAB use in solid tumors, and this is hampering the clinical success of TRABs. GPC3 is a molecule reported to have highly tumor-specific expression, and this was further confirmed by IHC in this study using a large set of tumor and normal tissues. High expression of GPC3 was observed in multiple solid tumors, whereas only weak expression was observed in a limited range of normal tissues, such as the endometrium and placenta. Maeda et al. (32) observed expression of GPC3 in the endometrium in 8% (1 of 12) of secretory phase samples but not in the proliferative phase (0 of 12) or menstrual phase samples (0 of 7). In accordance with that report, our examined TMA core of the endometrium, which was GPC3-positive, suggests features of the secretory phase: stromal edema, prominent spiral arterioles, and predecidualization with endometrial gland lumens. Although the positive rate was low, administration of ERY974 to premenopausal females requires careful attention. Strong GPC3 expression in the placenta is a well-known observation (25). Therefore, administration of ERY974 to pregnant women should be avoided. Because only a limited number of normal tissues showed GPC3 expression, this highly tumor-specific expression profile suggests acceptable toxicity for the TRAB, as supported by the results in cynomolgus monkeys including female animals to assess toxicity in endometrium in this study. Administration of ERY974 at a single dose (10 μg/kg) did not cause any cytotoxic changes in normal tissues. Furthermore, even after multiple rounds of administration with increasing doses, to a final dose (1500 μg/kg), cytotoxic changes were not detected in normal tissues, and all other abnormalities were transient and reversible.

In studies on cynomolgus monkeys, the most prominent clinical observation was CRS. As observed with other TRABs such as blinatumomab (48), CRS is thought to be a common side effect induced by TRABs. Thus, the development of agents or clinical regimens to minimize CRS is desirable for expanding the therapeutic index of TRABs. Here, we successfully demonstrated that cytokine release can be managed by corticosteroid premedication. Corticosteroids are widely used during clinical cancer treatment, especially at first dosing, to suppress CRS, which occurs through the use of various cytotoxic and biological agents. However, corticosteroids inhibit T cell activation and proliferation, which results in the concern that premedication with corticosteroids might dampen the antitumor efficacy of TRABs. Therefore, in this study, we examined the effects of corticosteroids on the pharmacological action of ERY974 in the huNOG mouse model, in which both cytokine release and antitumor efficacy can be evaluated in the same system. We found that corticosteroid premedication effectively suppressed cytokine release but did not affect antitumor efficacy. This suggests that a corticosteroid, which has a relatively short half-life, could effectively suppress early and transient cytokine release, without a significant effect on ERY974-mediated cytotoxic activity against tumors. An in vitro study to examine the effect of corticosteroids on cytokine release and pharmacological activity of blinatumomab has also been previously reported (49). This study also demonstrated the selective activity of corticosteroids in significantly reducing cytokine production while maintaining the cytotoxic activity of T effector cells.

Here, we examined the antitumor efficacy of ERY974 and checkpoint inhibitors by using two mouse tumor models, namely LLC1/hGPC3 and Hepa1-6/hGPC3, which present nonimmunogenic or immunogenic features, respectively. A previous report suggested that LLC1 tumors have low major histocompatibility complex (MHC) class I expression (42), and as expected, we found that checkpoint inhibitors have marginal antitumor efficacy against LLC1/hGPC3 tumors. In contrast, ERY974 treatment resulted in marked TGI. Because ERY974 can activate T cells in a polyclonal manner and redirect them to GPC3-expressing tumors regardless of the presence of neoantigens in MHC class I molecules, this therapeutic could efficiently elicit an antitumor effect even in tumors with nonimmunogenic features. In contrast, in immunogenic Hepa1-6/hGPC3 tumors, treatment with anti–PD-1 was effective, suggesting that an immunosuppressive environment is created in the tumor tissue by immune checkpoint molecules. Even in the immunosuppressive environment of an immunogenic tumor, ERY974 alone was shown to induce complete tumor regression. In contrast, multiple studies have suggested that the pharmacological action of TRABs is inhibited by immunosuppressive activity induced by the PD-1/PD-L1 axis (50). For example, Junttila et al. (51) reported that PD-L1 expression in tumors limited anti-HER2 TRAB (HER2-TDB) activity, and this resistance could be reversed by anti–PD-L1 treatment. Here, ERY974 treatment increased the number of PD-L1+ cells in both Hepa1-6/hGPC3 and LLC1/hGPC3 tumors. This suggests that combining the treatment with administration of immune checkpoint inhibitors could further amplify the antitumor efficacy of ERY974 in either immunogenic or nonimmunogenic clinical tumors, which represents an attractive approach for future clinical studies.

The current study thus provides preclinical safety profiles of ERY974 and its potent antitumor efficacy not only in immunogenic but also in nonimmunogenic tumors, which are difficult to treat with immune checkpoint inhibitors. This strongly supports the clinical testing of ERY974 for the treatment of GPC3-positive solid tumors. ERY974 phase 1 clinical trials are currently ongoing (NCT02748837).


Study design

The main objective of our study was to evaluate the antitumor efficacy and safety of ERY974 bispecific antibody targeting GPC3 and CD3E. The in vivo antitumor efficacy was assessed in three different tumor-grafted mouse platforms: NOD-SCID mouse inoculated with human T cells, immunocompetent human CD3E/D/G transgenic mouse, and huNOG mouse. Sample size (n = 4 or 5 per group) was determined on the basis of the consistency of tumor growth observed in preliminary experiments as the one that would give statistically significant differences in tumor size between the various treatment groups. Animals were randomly assigned to groups on the basis of tumor size so that each group had the same average size. Tumor volumes were calculated according to the following formula: [(length × width2)/2]. All tumor volume data (mean tumor volume with SD) were plotted except the data of dead animals (two in Fig. 7A and one in Fig. 7C). Animals were sacrificed at the end of the study. Human PBMCs and HSCs were used in those studies upon approval by an Institutional Review Board. The toxicity of ERY974 was assessed using cynomolgus monkeys. Animals judged unsuitable for toxicological evaluation of ERY974 were excluded from the study before grouping. Animals were assigned to each group using a computerized procedure designed to balance body weight equally among groups. The number of animals per group (n = 3 per sex per group) was chosen according to the ICH-S4A guideline as one that would enable a scientific conclusion on the general safety items with the minimum use of animals. For PK analysis, the data from animals that showed ADA production were excluded, regardless of the titer levels. Primary data are located in table S10.

In vitro TDCC and T cell activation assays

Human PBMCs were purified from fresh blood of healthy donors using a conventional Ficoll-Paque PLUS gradient (GE Healthcare). Adherent target cells were detached with Accutase (Innovative Cell Technologies), and 10,000 cells per well were seeded in 96-well U-bottom plates. ERY974 and human PBMCs were added (E/T ratio of 20:1). Target cell killing was assessed after 24 hours at 37°C and 5% CO2 through the quantification of lactate dehydrogenase (LDH) release into cell supernatants by dead cells (LDH cytotoxicity detection kit; Takara Bio). All samples were assessed in triplicate. Maximal lysis of target cells (100%) was achieved by incubation with Triton X-100. Minimal lysis (0%) refers to target cells incubated with effector cells but without ERY974. The percentage of TDCC was calculated as (sample release − spontaneous release)/(maximum release − spontaneous release) × 100. For TDCC assays using purified CD4+ and CD8+ T cells as effectors, CD4+ and CD8+ T cells were purified from PBMCs from two donors using CD4 and CD8 microbead separation systems (Miltenyi Biotec), respectively. All procedures were performed in the same manner as described previously, except for an alteration to the effector/target cell ratio (E/T ratio of 10). A KLH-TRAB antibody (bispecific to CD3 and keyhole limpet hemocyanin) was used as a negative control. After the TDCC assay, the remaining cells were collected and used to measure T cell activation by examining the expression of CD69 and CD25 in the CD3+ population using FACSVerse. Anti-CD3 (allophycocyanin-conjugated; clone SK7, BD Biosciences), anti-CD25 (phycoerythrin-conjugated; clone M-A251, BD Biosciences), and anti-CD69 (FITC-conjugated; clone FN50, BD Biosciences) antibodies were used.

In vivo human T cell–injected mouse model

All mouse studies were performed in accordance with the policies of the Institutional Animal Care and Use Committee (IACUC) at Chugai Pharmaceutical Co., Ltd. NOD-SCID mice (CLEA Japan Inc.) received subcutaneous implants of human cancer cells (1 × 107 cells per mouse). After palpable tumors were established, randomization was performed on the basis of tumor volume and body weight. Human T cells were selectively amplified by culturing human PBMCs using Dynabead Human T-Activator CD3/CD28 (Life Technologies) and were intraperitoneally injected into mice (3 × 107 cells per mouse) as effector cells. ERY974 (1 mg/kg), KLH-TRAB (1 mg/kg), or vehicle was intravenously administered 3 to 5 hours after T cell injection. Tumor size was measured twice per week. TGI was calculated using the following formula: TGI (%) = [1 − (TT0)/(CC0)] × 100, where T and T0 are the mean tumor volumes on a specific experimental day and on the randomization day, respectively, for the experimental groups, and C and C0 are the corresponding mean tumor volumes for the control group. Values > 100% represent tumor shrinkage.

Cynomolgus monkey studies

All cynomolgus monkey studies were conducted at Covance Laboratories Inc. according to the guidelines of the IACUC, using purpose-bred, naïve, cynomolgus monkeys of Chinese origin. For the single-dose study, three cynomolgus monkeys per sex per group were administered a single intravenous infusion of ERY974 (0.1, 1, and 10 μg/kg) for 30 min. For the second study, three cynomolgus monkeys per sex were administered an intravenous infusion dose of ERY974 (1500 μg/kg) after a daily increasing dose regimen (0.1 to 300 μg/kg per dose). Whole-blood samples or tissues were collected at selected time points for clinical pathology and histopathology. Plasma ERY974 concentration was determined by an electrochemiluminescence immunoassay (ECLIA). This method is based on an indirect immunoassay using GPC3 as a solid-phase antigen, with anti-ERY974 antibody binding to CDR for CD3 antigen as a detection antibody. Anti-ERY974 antibodies in cynomolgus monkey plasma were analyzed by ECLIA using biotin-labeled ERY974 and ruthenium-labeled ERY974. PK analysis was performed using WinNonlin Professional Edition computer software, version 6.1 (Pharsight).

Statistical analysis

Data are presented as means ± SD, means + SD, or means only as stated in the figure legends. Statistically significant differences were tested using specific tests as indicated in the figure legends. P < 0.05 was considered statistically significant.


Materials and Methods

Fig. S1. Cancer cell growth inhibition induced by ERY974.

Fig. S2. TDCC and polyclonal T cell activation induced by ERY974 targeting cancer cell lines.

Fig. S3. Plasma concentration–time profiles of ERY974 after a single intravenous administration.

Fig. S4. Influence of DEX pretreatment on the antitumor efficacy of ERY974 at a dose with submaximal efficacy (0.04 mg/kg).

Table S1. H-scores for 31 hepatocellular carcinoma TMA cores staining for GPC3.

Table S2. H-scores for 40 lung squamous cell carcinoma TMA cores staining for GPC3.

Table S3. H-scores for 69 lung small cell carcinoma TMA cores staining for GPC3.

Table S4. H-scores for 87 esophagus squamous cell carcinoma TMA cores staining for GPC3.

Table S5. H-scores for 30 cardiac adenocarcinoma TMA cores staining for GPC3.

Table S6. H-scores for 40 gastric cancer TMA cores staining for GPC3.

Table S7. H-scores for 68 head and neck cancer TMA cores staining for GPC3.

Table S8. H-scores for 30 different normal TMA cores staining for GPC3.

Table S9. Histopathological analysis of Hepa1-6/hGPC3 and LLC1/hGPC3 tumors.

Table S10. Primary data

References (52, 53)


  1. Acknowledgments: We thank the donors and patients who consented to the use of their cells for the studies. We also thank N. Hironiwa, T. Kuramochi, and K. Esaki for their help with antibody generation and characterization; N. Ikeda and A. Kato for pharmacological evaluation; O. Kondoh for facilitating the pharmacology studies; O. Ueda, N. A. Wada, H. Hino, and K. Jishage for generating the human CD3 transgenic mice; A. Shioda and S. Akai for the safety evaluation; and S. Matsuura for review of the manuscript. Funding: This study was funded and supported by Chugai Pharmaceutical Co., Ltd. Author contributions: T. Ishiguro, Y. Sano, S.i.-K., and J.N. wrote the manuscript; T. Ishiguro, Y. Sano, Y. Kinoshita, Y.A., T. Tsunenari, N.O., Y. Kayukawa, Y. Sonobe, K.S., T.F., Y.M., M.N., M.E., E.F., H.M., Y.N., M.T., Y. Kawabe, and M.A. executed the pharmacological studies and analyzed the assay data; S.i.-K., A.K., A.H., W.F., E.N., M.I., and S.C. executed the toxicological and PK studies and analyzed the assay data; L.S. and E.E. executed the GPC3 IHC studies and analyzed the assay data; M.K.-S., H.S., A.N., A.S., T.W., H.K., H.S., T. Tsushima, T. Igawa, K.H., and J.N. designed, produced, and characterized the antibodies; and T. Ishiguro, Y. Sano, S.i.-K., A.K., H.S., and M.K.-S. designed and supervised the studies. Competing interests: Chugai Pharmaceutical Co., Ltd. is developing ERY974 as a clinical compound. T. Ishiguro, Y. Sano, S.i.-K., M.K.-S., A.K., Y. Kinoshita, H.S., Y.A., T. Tsunenari, N.O., Y. Kayukawa, Y. Sonobe, K.S., T.F., Y.M., M.N., M.E., Y. Kawabe, A.H., W.F., E.F., E.N., A.N., A.S., M.T., T. Tsushima, T.W., H.K., H.S., H.M., Y.N., T. Igawa, M.I., S.C., M.A., K.H., and J.N. are employees of Chugai Pharmaceutical Co., Ltd., and L.S. and E.E. are employees of Ventana Medical Systems Inc.; Chugai Pharmaceutical Co., Ltd. has filed patent applications related to this work for ERY974 and TRAB. T. Ishiguro, M.K., H.S., Y.A., A.N., T. Igawa, and J.N. are inventors on patent application (WO/2016/047722) submitted by Chugai Pharmaceutical Co., Ltd. that covers the ERY794 molecule. Data and materials availability: Materials are available from Chugai Pharmaceutical Co., Ltd. under a material transfer agreement.

Stay Connected to Science Translational Medicine

Navigate This Article