To redirect cytotoxic T cells to target a broad range of adenocarcinomas, the authors constructed a novel, recombinant, bispecific antibody, E3Bi, directed at the tumor-associated antigen, epithelial cell adhesion molecule (EpCAM), and the CD3 receptor on T cells.
T cells were prepared from healthy blood donors. The cytotoxicity of activated T cells (ATC) redirected to tumor cells by E3Bi was measured with in vitro 51Cr release assays. In vivo studies were performed in a severe combined immunodeficient (SCID)/Beige mouse xenograft model. Tumor-bearing mice were treated with low doses (1 mg/kg) or high doses (10 mg/kg) of E3Bi along with ATC (2 × 109 cells/kg), and treatment efficacy was evaluated both by ex vivo tumor cell survival assay after in vivo treatments and by in vivo tumor growth delay studies.
In vitro, targeting the EpCAM-overexpressing human tumor cell lines with E3Bi increased specific cytotoxicity of ATC by > 70% at an effector-to-target ratio of 2.5 (P < 0.001); this cytotoxicity was abolished competitively in the presence of an anti-EpCAM monoclonal antibody. In contrast, E3Bi did not enhance ATC cytotoxicity toward the low EpCAM-expressing tumor cell line. In ex vivo tumor cytotoxicity assays, a significant reduction in tumor cell survival (40% with low-dose E3Bi; 90% with high-dose E3Bi) was observed in E3Bi/ATC-treated mice compared with control mice that were treated with ATC only. In addition, SCID/Beige mice xenografted with LS174T tumors demonstrated a significant tumor growth delay (P = 0.0139) after receiving E3Bi/ATC/interleukin 2 (IL-2) compared with mice that received ATC/IL-2 alone.
Epithelial cell adhesion molecule (EpCAM) is overexpressed abnormally in nearly all adenocarcinomas.1 This makes EpCAM (also known as EGP-2, EGP-40, 17-1A, and KSA) an attractive candidate tumor-associated antigen (TAA) to serve as a target for antibody-based immunotherapy. The highest expression levels of EpCAM have been reported in adenocarcinomas of the gastrointestinal tract, lung, and cervix. Over expression also is seen in basal cell and bladder carcinomas as well as gastric, pancreatic, biliary duct, renal cell, thyroid, prostate, ovarian, endometrium, and mammary adenocarcinomas.
A bispecific antibody (BiAb) is comprised of a molecule with two different antigen-specific binding sites. One site of the BiAb recognizes and binds to the T-cell receptor (TCR) on T-cells, and the other site binds to TAA, thus establishing a “bridge” between the two types of cells. When a BiAb bridges a T-cell and a tumor cell, the BiAb will trigger this T cell to become a specific cytotoxic T lymphocyte (CTL) that will bypass major histocompatibility complex (MHC) restrictions and will destroy the tumor cell directly. CTLs are “professional killer cells” involved in antigen-specific recognition and subsequent killing of “foreign” cells, such as infected cells and tumor cells.2 Therefore, BiAbs offer an attractive approach to redirect the patient's own immune system to destroy tumors in an antigen-specific manner. Not surprisingly, the BiAb approach has improved survival rates in animal cancer models3–5 and has produced minor responses in patients.4, 6–8 The latter responses, however, did not meet the high expectations of therapeutic efficacy because of dose-limiting toxicities associated with the administration of whole antibody generated from mouse origin and because of physiologic barriers impacting on delivery of the large BiAb molecules.5, 8–14
The recombinant BiAb (rBiAb) approach utilizes molecular engineering technologies that offer the potential to remove the restrictions imposed by chemically heteroconjugated BiAbs and provides the option for adding functional protein domains for better eradication of tumor cells while removing domains that cause possible toxicity.5, 9, 10, 15–20 We have employed recent molecular cloning techniques to construct a novel, smaller, and less immunogenic rBiAb (named E3Bi) (Fig. 1) for the purposes of 1) facilitating better tumor penetration compared with large and conventional BiAbs and 2) reducing human antimouse antibody (HAMA) responses, cytokine release syndrome (CRS), and other toxicities generally associated with BiAbs. Figure 1 illustrates the relation of E3Bi with its two targets: a T cell and a tumor cell. We report here for the first time the strategy of using an extra-long linker with 63 amino-acid residues derived from the CD8α immunoglobulin (Ig) hinge-like region. The studies reported here demonstrate that E3Bi effectively redirects antigen-nonspecific T cells to target the antigen and specifically to kill EpCAM-overexpressing tumor cell lines both in vitro and in a human tumor xenograft mouse model in vivo.
MATERIALS AND METHODS
Construction of pG1EN-E3Bi
The single-chain fragments of variable region (scFv) cDNA of GA733.2 were a gift from Dr. Mike Kershaw (National Cancer Institute, National Institutes of Health [NIH], Bethesda, MD), and OKT3-scFv was derived from the OKT3 hybridoma (American Type Culture Collection [ATCC], Rockville, MD). GA733.2 is an anti-EpCAM antibody, and OKT3 is an anti-CD3 antibody. A 14-amino-acid linker21 was inserted between the variable light and heavy chains to form the scFv. A 6xHis sequence was encoded in the downstream primer for later affinity purification. Then, the 3′ end of GA733.2 heavy chain was linked to the 5′ end of OKT3 heavy chain through a 63-amino-acid linker comprised of the CD8α Ig hinge-like domain.21 The assembly process took place in a cloning vector (pBluescript II KS; Stratagene, La Jolla, CA). The assembled cDNA for GA733.2 scFv-linker-OKT3 scFv was cut off by NotI and XhoI and inserted into the expression vector pG1EN. The pG1EN vector also was a gift from Dr. Kershaw and originally was generated from the Maloney murine leukemia virus. The E3Bi recombination was confirmed by restriction enzyme digestions and by DNA sequencing analysis in both directions. Finally, a stable transfection of pG1EN-E3Bi into mammalian Chinese hamster ovary (CHO) cells was performed (Fig. 2).
LS174T, a human adenocarcinoma cell line with 99% EpCAM expression by flow cytometry, and the MIA pancreatic cell line (MIA PaCa-2), with 5% EpCAM expression, were obtained from the ATCC. LS174T cells were maintained in minimum essential medium (MEM) (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS) (Invitrogen) and MIA cells in Dulbecco MEM (Invitrogen) supplemented with 2.5% horse serum (Hyclone Laboratories, Logan, UT) and 10% FBS. The FG cell line (100% EpCAM expression) is a highly metastatic subclone22 that was developed from the human pancreatic adenocarcinoma COLO357 (ATCC) and was a generous gift from Dr. M. P. Vezeridis (Department of Surgery, Roger Williams Medical Center, Providence, RI). FG cells were maintained in RPMI-1640 medium (BioWittaker, Walkersville, MD) supplemented with 10% FBS. The hybridoma cells for producing GA733.2 mouse monoclonal antibody (mAb) were kindly provided by Dr. D. Herlyn (The Wistar Institute, Philadelphia, PA) and were cultured in RPMI-1640 medium containing 10% FBS. The CHO cells (Invitrogen) were cultured in RPMI-1640 medium supplemented with 10% γFBS (certified Australian and γ-irradiated; Hyclone Laboratories). No antibiotics were used during routine maintenance of cell lines in culture.
Peripheral blood mononuclear cells (PBMCs) were prepared from heparinized blood using the standard methods as described previously.21 PBMCs from healthy donors were used for this study. All PBMCs were cultured at 37 °C in a 5% CO2 /95% air humidified atmosphere in RPMI-1640 medium supplemented with 10% FBS and activated with 100 international units (IU) of interleukin-2 (IL-2)/mL (Chiron, Emeryville, CA) and 50 ng OKT3/mL (Ortho Pharmaceuticals, Raritan, NJ). By Day 14, nearly 100% of cells were CD3+ activated T-cells (ATCs). ATCs were maintained in a concentration of 1 × 106 cells/mL and freshly supplemented with 100 IU of IL-2/mL every 2–3 days.
Analysis of EpCAM Expression on Tumor Cells
Tumor target cells were examined for the expression of EpCAM protein using GA733.2 (the primary antibody) followed by a fluorescein isothiocyanate (FITC)-conjugated rabbit antimouse Ig (the secondary antibody; Dako, Carpenteria, CA). Cells incubated with the secondary antibody alone were used as controls to exclude nonspecific staining. Subsequently, the cells were analyzed on a FACSCalibur flow cytometer (Becton Dickinson [BD], San Jose, CA), using CELLQuest software (BD). Cell populations were visualized on forward scatter versus side scatter. The lymphocyte population was gated to exclude dead cells. The percentages of fluorochrome-labeled cells were those that expressed EpCAM.
Affinity Purification of E3Bi
The cell culture supernatant fluid that contains the secreted E3Bi was collected from 100% confluent CHO cell cultures in T75 flasks every 2 days. The E3Bi was extracted from the supernatant fluid through its C-terminal 6xHis tail using immobilized metal-affinity chromatography technology with the nickel-charged Ni-NTA resin (QIAGEN). The eluted protein product was buffer-exchanged and concentrated using the Amicon Ultra-15 (Millipore, Bedford, MA) into phosphate buffered saline (PBS) containing 140 mM NaCl and 2.7 mM KCl in 10 mM phosphate buffer, pH 7.4. The final products were filter-sterilized through a 0.2-μm filter and aliquoted at 1.0 mg protein/mL per vial. The protein concentration was measured using a Bio-Rad protein assay kit (BCA-200 Protein Assay Kit; Bio-Rad, Hercules, CA). All procedures, including E3Bi gene transduction, supernatant fluid collections from CHO cell culture, and the affinity purification, were conducted in our good manufacturing practices laboratory. On average, 1 mg E3Bi could be purified from 100 mL supernatant fluid. In an 8–16% gradient polyacrylamide electrophoresis gel, E3Bi migrates as a 65-kilodalton (kD) protein under nonreducing conditions and as a 60-kD protein under reducing conditions.
In Vitro Cytotoxicity Assay (51Cr Release Assay)
The cytotoxicity of Day 14 ATCs against tumor cell targets in the presence or absence of E3Bi was measured by the cytotoxicity assays.21 Briefly, target cells (4 × 104 per well) were seeded into a flat-bottom, 96-well tissue culture dish. After overnight incubation at 37 °C, the cells were labeled with 2 μCi per well of 51Cr (Na251CrO4; 2 mCi/mL) (CN Biomedicals, Irvine, CA) in 100 μL RPMI-1640 medium containing 10% FBS for 4 hours at 37 °C. The ATCs were mixed with E3Bi first and then were applied to labeled target cells in triplicate at indicated effector-to-target (E:T) ratios in a 200-μL final volume. The E3Bi dose was 100 pmole/1 × 106 ATCs or as indicated. Results were expressed as the mean percentage of cytolytic activity ± standard deviation. Statistical analysis was performed by the paired t test using SigmaPlot (Jandel Scientific Software, San Rafael, CA).
Four-week-old, female, severe-combined immunodeficient (SCID)/Beige mice (Taconic Pharm, Germantown, NY) were used for in vivo studies. The animals were maintained in the animal care facility at Roger Williams Hospital in accordance with NIH animal care guidelines for principles of laboratory animal care. These mice carry the SCID mutation, which causes a deficiency of T cells, and B cells, and natural killer cells. For the tumor growth-delay studies, tumor cells (LS174T; 1 × 106) in 0.1 mL culture medium were injected subcutaneously into the left flank of the mice. After 2–3 weeks, when tumors measured approximately 500 mm3 (0.5 cc), as calculated by the formula volume = (length) × (width)2 ÷ 2, mice were divided into 3 groups: Ten mice were treated with E3Bi plus an admixture of IL-2 with ATC, 8 mice were treated with IL-2 and ATC, and 6 mice were treated with IL-2 only. ATC and IL-2 were administered by intratumoral (IT) injection, and E3Bi was administered by intravenous (i.v.) injection. The dose for IL-2 was 1 × 104 IU/kg per injection, the dose for ATCs was 2 × 109 cells/kg per injection, and the dose for E3Bi was 1 mg/kg per injection. Animals were treated twice within 1 week. The progress of each tumor was monitored 3 times per week until it reached a volume of 2 cc, at which point the mouse was killed. Tumor growth delay is represented as the proportion of mice from each treatment group bearing tumors that measure < 2 cc in volume as a function of time. Statistical comparison was performed by the Student t test for paired data using SigmaSTAT software.
The Tumor Excision Assay
The tumor excision assay was performed as described previously,23 with minor modifications. Briefly, LS174T cells (106 cells per animal) were injected subcutaneously into the left flank of 4 SCID/Beige mice. Fourteen days after tumor implantation, when tumors reached a volume of approximately 1 cc, treatments were initiated. Mice were injected IT either with ATCs (2 × 109 cells/kg) or with ATCs followed by i.v. injection of low-dose E3Bi (1 mg/kg) or high-dose E3Bi (10 mg/kg). IL-2 (1 × 104 IU/kg per injection) was present in the ATC mixture. Mice were killed 24 hours after treatments to allow for full expression of treatment cytotoxicity, and their tumors were excised and processed into single-cell suspensions as described previously.23 Tumor cells from each treated mouse were counted and seeded into culture in four concentrations (from 104 to 2 × 105) with five replicates of each concentration. After 1 week of culture, cells were suspended by treatment with trypsin/ethylenediamine tetraacetic acid and were counted with a Coulter particle counter (Coulter Electronics, Inc., Hialeah, FL). Results are represented as the surviving fraction of tumor cells treated with ATCs compared with ATCs plus low or high doses of E3Bi (± standard error). Statistical analysis was performed with InStat software (Graph Pad, San Diego, CA) using the Kruskal–Wallis nonparametric test followed by the Dunn multiple comparisons test.
E3Bi Specifically Binds to EpCAM Overexpressing Tumor Cells and Triggers the Cytolytic Activity of ATCs
Figure 3 shows two photomicrographs that were taken 20 hours after adding E3Bi and ATCs to LS174T cells. The left photomicrograph demonstrates 1) ATC rosetting consistent with T-cell activation induced by E3Bi-mediated, specific binding to the tumor antigens; and 2) the absence of viable tumor cells in the E3Bi and ATC-treated culture. In the right photomicrograph, in the absence of E3Bi, no T-cell aggregation or rosetting is observed, and tumor cells remain viable. The graphs on the bottom in Figure 3 show that E3Bi triggered ATC cytotoxicity in two separate experiments using ATCs from two different healthy donors. These results provide evidence that the E3Bi-mediated ATC cytotoxicity is reproducible when utilizing ATCs from different donors.
To characterize further the specific E3Bi binding to target cells as a function of EpCAM recognition, the EpCAM-overexpressing cell line, FG (100%), and the low-expressing cell line, MIA (5%), were used as the targets for the cytotoxicity assays. The graphs on the left in Figure 4 show histograms of EpCAM expression levels by flow cytometry, and the graphs on the right show the cytotoxicity assay results using these two corresponding target cell types. The results indicate that E3Bi significantly induces the cytolytic function of ATCs on FG cells, even at a low E:T ratio of 1.5, although not on MIA cells, not even up to an E:T ratio of 12.5. These data suggest that E3Bi only targets tumor cells that overexpress EpCAM. Moreover, the addition of an admixture of the unconjugated parental mAbs (OKT3 and GA733.2) failed to trigger any cytolytic activity.
The specific cytotoxicity of E3Bi was evidenced further in a competitive binding assay (Fig. 5). For this assay, 1 μg of parental mAb, GA733.2, was incubated with 4 × 104 LS174T cells in each well for 4 hour before adding ATCs and E3Bi (100 nM) to each well. The addition of mAb GA733.2 blocked EpCAM sites on tumor cells, resulting in a significant reduction in E3Bi-triggered cytotoxicity of ATCs. These results demonstrating that E3Bi only targets cells that overexpress EpCAM are very important, because they imply that E3Bi will lead to tumor cell death without damaging the normal epithelial cells that express lower levels of EpCAM.
Parenterally Administered E3Bi Traffics to Tumors to Redirect ATC Cytotoxicity at EpCAM-Overexpressing LS174 Xenografts
A xenografted mouse model was used to evaluate the cytotoxicity of ATCs directed by E3Bi in vivo. Mice were divided into 3 groups that received IT injection of ATCs (2 × 109 cells/kg) or IT injection of ATC plus an i.v. injection of either low-dose (1 mg/kg) or high-dose (10 mg/kg) E3Bi. In patients, colorectal tumors attract tumor-infiltrating lymphocytes, particularly cytotoxic T cells.24 Because SCID/Beige mice are T-cell deficient, to recreate a representative immune environment, ATCs were injected directly into the tumors just prior to E3Bi treatments. In addition, because E3Bi is designed to bind human T cells, human ATCs prepared from healthy blood donors were used, and IL-2 was included to maintain the viability of ATCs. A control of E3Bi alone was not included in this experiment, because the cytotoxicity data showed that E3Bi alone failed to trigger any tumor-killing activity (Figs. 3, 4). The ability of E3Bi to traffic to and specifically target EpCAM-overexpressing tumors in vivo was demonstrated by a tumor cell survival assay in which LS174T tumors, growing in mice, were treated and excised 24 hours later. The ex vivo viability of in vivo-treated tumor cells was measured and is presented in Figure 6 as the surviving fraction of tumor cells 5 days after seeding excised tumor cells into culture. The administration of low-dose (1 mg/kg) E3Bi in conjunction with ATC treatment produced a 40% decrease in tumor cell survival. Increasing the E3Bi dose to 10 mg/kg significantly decreased the tumor cell survival by 90% (P < 0.05). Taken together with the tumor growth-inhibition studies, these results suggest that systemically delivered E3Bi traffics to, binds, and is effectively cytotoxic to EpCAM-overexpressing tumor cells in vivo. Enhanced survival of cells from IL-2/ATC-treated mice in this experiment may reflect that cell survival in Figure 6 was normalized to IL-2-treated control tumor cells. Other investigators have reported that IL-2, even in nanomolar concentrations,25 inhibits the growth of some human tumor cells, such as head and neck, gastric, and renal cell carcinomas, by inducing cell cycle arrest.26 Preferential uptake of IL-2 by ATCs may reduce IL-2 availability to tumor cells and, thus, provide one possible explanation for the differential survival observed between tumor cells treated with IL-2 in the presence or absence of ATCs.
E3Bi/ATC Delays Tumor Growth Significantly in SCID/Beige Mice Bearing LS174T Human Colorectal Adenocarcinoma Xenografts
The in vivo antitumor response of E3Bi was evaluated in the tumor xenograft model by a tumor growth delay assay (Fig. 7). In SCID/Beige mice bearing established LS174T tumors, tumor growth delay varied significantly between the three treatment groups. Although it was found that ATC/IL-2 treatments produced significant tumor growth delay (P = 0.038), the administration of E3Bi, ATCs, and IL-2 significantly increased the time it took for tumors to reach a volume of 2 cc (P = 0.0139). These results clearly indicate a therapeutic advantage for adding E3Bi to the ATC/IL-2 treatment regimen. Although animals only received 4 treatments within the first 2 weeks, this pilot study suggests that continued dosing with E3Bi may delay tumor growth further.
BiAbs as cancer therapeutic agents have been investigated extensively in both preclinical models and clinical studies;4–7, 9, 24 however, there are certain structural aspects of BiAbs that limit their clinical application. To address these issues, we employed recent advances in molecular engineering to design a molecule that eliminated the immunogenic FcR portion to reduce potential HAMA responses and other toxicity associated with FcR binding and to reduce the molecular size of E3Bi, which may permit greater penetration into tumors. Only the variable light and heavy chains of OKT3 and GA733.2 mAbs were cloned into the two scFvs, resulting in a construct with two scFvs (separated by a defined polypeptide linker) that binds to EpCAM and to the TCR ϵ-subunit. By bridging tumor cells and T cells, this binding triggers T-cell activation, cytotoxicity, and cytokine production similar to that seen in chemically heteroconjugated BiAbs made from native mAbs.
Our strategy of employing a long linker with 63 amino-acid residues is novel. The rationale for developing a long linker is based on our earlier studies of the chimeric TCR (ch-TCR) in which we compared a short-hinge linker (8 amino acids) with the same 63-amino-acid linker that was used in the current study. The results showed that increasing the distance between the anti-EpCAM scFv domain and the T-cell membrane markedly increased binding of scFv to EpCAM positive targets and the cytotoxicity mediated by ATCs expressing the ch-TCR.21 Based on those observations, we introduced this 63-amino-acid linker into the E3Bi construct. Linker length may be critical in the function of rBiAbs as well as in ch-TCR. To explore this possibility, currently, we are comparing linkers with different lengths. To our knowledge, a linker length > 27 amino acids in a rBiAb construct never has been reported.
EpCAM was selected as the target for our rBiAb, because EpCAM is overexpressed widely in adenocarcinomas and is expressed minimally in normal cells. Our in vitro studies reported here using LS174T as the targets demonstrated > 80% cytolytic activity of E3Bi at the E:T ratio of 2.5 in a 20-hour assay. In addition, however, the antitumor effects of E3Bi have been demonstrated in vivo in SCID/beige mice bearing LS174T tumors.
Previous studies using CO17-1A and GA733.2 mAbs targeting EpCAM provided valuable information for the selection of our parental anti-EpCAM mAb. Both mAbs bind to EpCAM but at different epitopes and with different affinities. GA733.2 (5.0 × 108 M−1) binds more avidly than CO17-1A (0.7 × 108 M−1).1, 25 When their scFvs were constructed into ch-TCRs, only the GA733.2 ch-TCRs, and not the CO17-1 ch-TCRs, bound to and redirected transduced T cells to kill EpCAM-overexpressing tumor targets.26 For this reason, we selected the GA733.2 mAb for constructing E3Bi. One potential drawback associated with high-affinity binding of a rBiAb is the toxicity caused by its binding to normal epithelium that expresses low levels of EpCAM.1 Several observations in our study suggest that E3Bi may not be toxic to normal epithelium, because E3Bi did not induce cytotoxicity directed at EpCAM low-expressing MIA tumor cells (Fig. 4).
To avoid the problems of recombinant proteins produced in bacterial or yeast systems,5 such as improper folding, lack of proper glycosylation, or susceptibility to proteolysis,27 we produced E3Bi in a mammalian cell expression system. We produced a fully folded and functional E3Bi that was secreted into the medium. Such a system is compatible with large-scale and concentrated protein production in bioreactors with subsequent affinity purification of clinical-grade material. E3Bi is stable for > 6 months when it is stored at 4 °C in PBS without evidence of proteolytic degradation or loss of function (data not shown).
To be useful clinically, an rBiAb must retain its binding and cytolytic-inducing function and must traffic successfully to the tumor in an animal model. Others have demonstrated the feasibility of using rBiAbs, such as anti-CD3 x anti-BCL1,28 anti-CD3 x anti-CD19,29 anti-CD3 x anti-CD2030 to target tumors in murine and xenograft lymphoma models. The rBiAb, anti-CD3 x anti-MUC-1,31 has been used to target bile duct carcinoma in a xenograft SCID mouse model. In the current study, we have shown that E3Bi administered i.v. traffics to tumors through the blood stream and is capable of penetrating to target tumor cells and redirecting ATCs to lyse tumor cells (Figs. 6, 7). In animal assays, E3Bi combined with ATC and IL-2 treatment enhanced tumor growth delay significantly in human colorectal xenografted mice compared with ATC and IL-2-treated controls. Therefore, the current results show the feasibility and advantage of combining ATCs and IL-2 with E3Bi.
We designed and engineered a rBiAb, E3Bi, which utilizes a novel, long-hinge linker to optimize binding efficiency to EpCAM and TCR. In the current study, we demonstrated that E3Bi binds to its target molecules and specifically redirects ATC cytolytic activity against human tumor cells in vitro. Furthermore, it retains target antitumor activity when it is administered in vivo. These results suggest that E3Bi may be an attractive antitumor agent for the treatment of EpCAM-overexpressing neoplasms.
The authors thank John Tigges for technical help with the flow cytometry, Christine Samek and Nicole Jones for assistance with animal studies, and Daniel Ren-Heidenreich for editing this article. They also thank Drs. Abby Maizel, Jeffrey Ledbetter, and Raymond Frackelton for scientific discussions.