The construction of antibody fragments has the potential to reduce the high cost of therapeutic antibody production, but the structures of these fragments, with monovalency and the lack of an Fc region, can lead to reduced function. Multimerization is one strategy for recovering function that also yields better tumor-to-blood ratios than IgGs or monomeric antibody fragments because of rapid tumor uptake and clearance. Here, we constructed single-chain variable fragment (scFv) multimers by modifying the linker length and domain order of humanized anti-(epidermal growth factor receptor) IgG 528 (h528) and tested their ability to inhibit tumor growth. h528 scFv multimers, expressed using a bacterial expression system, were successfully fractionated and inhibited cancer growth in a multimerization-dependent manner, whereas the h528 scFv monomer showed no inhibition. h528 scFv trimers with variable heavy–light domain order and no linkers showed the highest in vitro and in vivo antitumor effects, which were comparable with those of the approved anti-(epidermal growth factor receptor) therapeutic IgG Cetuximab and Panitumumab. The trimers were also structurally stable in vitro and in vivo, which may be attributable to a strong interaction between the variable heavy and variable light domains of h528 Fv. Thus, h528 scFv multimers, especially trimers, are attractive as the next generation of anti-(epidermal growth factor receptor) therapeutic IgG and offer the possibility of low-cost production.
The epidermal growth factor receptor (EGFR), a transmembrane tyrosine kinase receptor, is widely expressed in various solid tumors, and its expression level is correlated with malignancy, metastatic phenotype and poor prognosis [1-3]. For these reasons, EGFR is considered an important target for cancer therapeutic reagents, including therapeutic antibodies. Two anti-EGFR therapeutic IgGs, Cetuximab and Panitumumab, are currently approved by the US Food and Drug Administration, and others are undergoing clinical trials . However, such intact IgG has several limitations including high production costs, because of the requirement for a mammalian expression system, and poor penetration into tumor tissues [5, 6].
Advances in recombinant technology have made it feasible to generate antibody fragments, such as variable fragments (Fv) and single-chain Fvs (scFv). These compact structures contribute to low immunogenicity, high tumor penetration and the potential for large-scale preparation through bacterial expression systems. However, this downsizing results in too rapid clearance from the blood and involves a decrease in valence and removal of the Fc region. Consequently, Fvs and scFvs can have low affinity for the target and fail to induce secondary immune functions such as antibody-dependent cellular cytotoxicity and complement-dependent cytotoxicity.
The multimerization of small recombinant antibodies is one strategy for improving their pharmacokinetic and binding affinity. In scFvs, the length and composition of the polypeptide linker between variable heavy (VH) and light (VL) domains strongly influence the formation of the multimeric structure. A linker of 15 amino acid residues leads to the formation of an scFv monomer, but reducing the linker length to 8–12 residues causes the scFvs to assemble into dimers, so that diabodies are formed. Further reduction to fewer than five residues or elimination of the linker leads to the formation of more higher order multimers such as scFv trimers, tetramers (known as triabodies, tetrabodies), etc. [7-11]. These increases in molecular size are thought to prolong serum half-lives, which may bring better tumor-to-blood ratios than either IgG or scFv monomers because of the rapid tumor uptake and clearance of scFv multimers . In addition, multimerization generally achieves multivalency against the target antigen; this may contribute not only to increases in binding affinity, because of an avidity effect, but also to improvements in function, which may compensate for the lack of induction of secondary immune functions.
We previously reported the humanization of anti-EGFR IgG 528, which inhibits tumor growth by blocking the phosphorylation of a tyrosine kinase, and also reported the construction of recombinant antibodies based on humanized 528 (h528) Fv [13-16]. Here, we prepared scFv multimers using two types of h528 scFvs each with a different domain order, because the domain order of VH and VL in scFv may influence multimerization [8, 9]. The h528 scFv multimers were successfully fractionated from dimers to pentamers, and the h528 scFv trimers with the VH–VL order and no linker showed the highest in vitro and in vivo antitumor effects, comparable with those of Cetuximab and Panitumumab. To our knowledge, this is the first report of antitumor effects of humanized anti-EGFR scFv multimers. These molecules may translate into the next generation of anti-EGFR therapeutic IgGs because they offer the possibility of low-cost production through bacterial expression systems.
Preparation of h528 scFv multimers with different domain orders
To produce anti-EGFR scFv multimers, we constructed h528 scFv expression vectors with a shorter or no linker and a different domain order, as described in Materials and Methods and summarized in Fig. 1 (A). Purified h528 scFvs from bacterial soluble fractions, including the original h528 scFv with a long linker, were applied to gel-filtration columns for further purification. Each h528 scFv in the VH–VL order, HLG3, HLG1 and HLG0, predominantly formed monomers, dimers and trimers, respectively (Fig. 1B). In the VL–VH order, LHG1 also predominantly formed dimers, whereas a broad, mixed peak representing several molecular species was found for LHG0 (Fig. 1C). The LHG0 trimers, tetramers and pentamers were fractionated by molecular calibration and their monodispersity was confirmed by repeat chromatography (Fig. 1D). Thus, we successfully prepared h528 scFv monomers and six types of h528 scFv multimers and evaluated their functions by using the peak fractions indicated by the arrows (Fig. 1B,C).
Effect of the multimerization of h528 scFv on growth inhibition
To evaluate the effect of the multimerization of h528 scFv on the inhibition of human carcinoma cell growth, we analyzed h528 scFvs and compared them with anti-EGFR IgGs by using 3-(4,5-dimethylthiazole-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium inner salt. By contrast to the marginal effects of the HLG3 monomer, both scFv multimers, the HLG1 dimer and the HLG0 trimer, effectively inhibited cancer growth; in particular, the HLG0 trimer was more effective than the parental 528 IgG and Cetuximab, an approved therapeutic anti-EGFR IgG, at low concentrations (Fig. 2A). The scFv multimers in the VL–VH order (i.e. the LHG1 dimer and the LHG0 trimer), also inhibited cancer growth, but to a lesser extent than the multimers in the VH–VL order (Fig. 2B). To compare the LHG0 multimers, we calculated the concentrations of each multimer by using the molecular masses of the trimers because the exact molecular species were unknown; this meant that if these multimeric forms were identical to the predicted ones, the 100 nm tetramer and pentamer would correspond to 75 and 60 nm, respectively. The comparative effects observed among the LHG0 multimers might indicate that higher multimeric forms can induce greater growth inhibition (Fig. 2C), although the VL–VH order had no additional effects beyond those achieved by IgG.
We next compared the h528 scFv multimers in the VH–VL order, adjusted by weight concentration, with Panitumumab, another approved therapeutic anti-EGFR IgG characterized by high affinity and strong neutralization activity but, unlike Cetuximab, not by antibody-dependent cellular cytotoxicity induction. Both multimers, especially the HLG0 trimer, inhibited cancer growth more effectively than did Panitumumab in A431 cells (Fig. 2D). The h528 scFv multimers also inhibited the growth of TFK-1 cells, which express moderate levels of EGFR; however, they were less effective than Panitumumab (Fig. 2E). Thus, depending on the EGFR expression level and/or the characteristics of the target cells, these results show that the h528 scFv multimers have potential beyond that of intact anti-EGFR IgG.
Evaluation of binding kinetics with surface plasmon resonance
To evaluate the effect of the multimerization of h528 scFv on apparent affinities, we analyzed the binding kinetics for immobilized soluble EGFR (sEGFR) using SPR. The resultant sensorgrams of the h528 scFvs against EGFR showed similar association phases, but slower dissociation phases depending on the multimerization (Fig. 3). Consequently, the dissociation rates of the HLG1 dimer and the HLG0 trimer were 12- and 46-fold lower than that of the HLG1 monomer, and these decreases contributed to their 19- and 118-fold lower dissociation constant, respectively (Table 1). With increasing valency, the chance increases that at least one of the remaining arms will find a target before the monovalently bound complex dissociates; this often results in a slow off-rate in a multivalent-dependent manner [17, 18]. Multimerization of scFvs substantially increased their apparent affinities, which might provide the h528 scFv multimers with growth inhibition effects.
Table 1. Binding and pharmacokinetic parameters. Kinetic parameters were calculated by means of a global fitting analysis with the assumption of a 1 : 1 Langmuir binding model
kon (× 105·m−1·s−1)
koff (× 10−3·s−1)
KD (× 10−8m)
Inhibition effect of the multimerization of h528 scFv on EGFR phosphorylation
To evaluate the effect of the multimerization of h528 scFv on the inhibition of phosphorylation, we analyzed tyrosine-phosphorylated EGFR using western blotting. After preculturing with each antibody, A431 cells were stimulated in culture medium containing EGF. The HLG1 monomer showed a certain level of inhibitory effect compared with the control without antibody, whereas all of the multivalent antibodies including the IgGs more intensely inhibited EGFR phosphorylation. These results suggest that the multivalencies of the anti-EGFR IgGs and antibody fragments are responsible for their inhibitory effects on EGFR phosphorylation, which, in turn, are responsible for their growth inhibition effects (Fig. 4).
Confirmation of the in vitro and in vivo stability of h528 scFv multimers
Structural stability is a critical factor for potential therapeutic recombinant proteins; however, there is the concern that scFv multimers made by noncovalent association might dissociate under lower concentrations such as in vivo or during long-term storage. For the h528 scFv multimers, effective growth inhibition, comparable with that achieved with Cetuximab, which stabilized by intermolecular disulfide bonds, was observed even at subnanomolar concentrations (Fig. 5A). The stability of the assembled structures was also evaluated by gel-filtration analysis using fractionated scFvs that had been stored for 9 months. A single, monodisperse peak was found for all of the scFvs including the HLG3 monomer, indicating that the multimeric structure of h528 scFvs is stable during long-term storage (Fig. 5B). Further, to evaluate in vivo stability, we measured the AUC using radioiodine-labeled h528 scFvs. Compared with the HLG3 monomer, the AUC(1.5–8 h) was increased 2.5-fold for the HLG1 dimer and 3.9-fold for the HLG0 trimer, depending on the calculated molecular mass (Fig. 5C and Table 1). These results suggest that the multimeric structures of h528 scFvs are also maintained in vivo.
Comparison of the in vivo antitumor effects of h528 scFvs
To evaluate the influence of the multimerization of the h528 scFvs on their in vivo antitumor effects, we transplanted TFK-1 cells into severe combined immunodeficient mice, which we then treated for 4 days with 20 μg of these reagents. We have used this protocol to evaluate the in vivo antitumor effect of hEx3, a bispecific diabody with monovalent binding to EGFR and CD3, in the presence of activated T cells . Thus, we chose this well-tried method to compare with IgGs and also with hEx3. Similar to the NaCl/Pi control, the HLG3 monomer and hEx3 showed no growth inhibition effect. By contrast, the HLG1 dimer and the HLG0 trimer significantly (P < 0.05) inhibited tumor growth in severe combined immunodeficient mice (Fig. 6A). In this in vivo model, no major differences were observed among Cetuximab, Panitumumab and the HLG0 trimer, which was the most effective of the h528 multimers (Fig. 6B). Our results show that the h528 scFv multimers are effective in vivo, depending on the degree of multimerization, and that the HLG0 trimer, which had effects comparable with those of the approved anti-EGFR IgGs, is an attractive candidate for novel anticancer drugs that can be prepared by using a bacterial expression system.
We have developed bispecific antibodies with specificity for EGFR and CD3 as a next-generation format for EGFR-targeting therapeutic antibodies. These antibodies included IgG-like bispecific antibodies that contain the human Fc region, and studies to assess their usefulness for clinical therapy are currently underway [15, 20, 21]. However, the recruitment of T cells by bispecific antibodies has led to concerns about side effects, including effects on the central nervous system, and this has led to the permanent discontinuation of study drug in some cases . In addition, although IgG-like bispecific antibodies are an attractive format, they are difficult to express in bacteria, which leads to high production costs.
Among anti-EGFR therapeutic IgGs, Panitumumab is characterized by its high affinity; its proposed mechanisms of action include downregulation of EGFR expression resulting from receptor internalization, induction of apoptosis via inhibition of EGFR signaling pathways and the induction of cell-cycle arrest, induction of autophagy and inhibition of angiogenesis. Unlike Cetuximab, Panitumumab is not characterized by the induction of antibody-dependent cellular cytotoxicity . Further, there have been reports of the induction of strong agonist activities following the conversion to the diabody format from the parental IgGs, which have little to no agonist activity; differences in accessibility to, or in the proximity of, target antigens might contribute to this change [24, 25]. Given these findings and the tumor growth inhibitory effect of humanized anti-EGFR IgG 528, we tried to prepare scFv multimers with high affinity due to multivalency by using a bacterial expression system. The molecular sizes of the scFv multimers were also expected to yield better tumor-to-blood ratios than either IgG or scFv monomers due to their rapid tumor uptake and clearance .
To prepare the h528 scFv multimers, we designed four types of scFvs with two different domain orders, with and without a short linker (Fig. 1A), and successfully fractionated them from dimers to pentamers (Fig. 1B–D). By contrast to the lack of effects of the h528 scFv (HLG3) monomer, all of the h528 scFv multimers inhibited cancer growth in a multimerization-dependent manner, with more intense effects being observed with the VH–VL domain order (Fig. 2). Multimerization of scFvs is thought to be affected by not only the length of the linker, but also the domain order [8, 9, 26], however, there have been no reports on the influence of domain order on the antitumor effects of scFv multimers. We recently reported a cytotoxic enhancement by converting the VH–VL order to the VL–VH order in a bispecific diabody; structural superiority for cross-linking target cells may have contributed to this enhancement . Here, conversely, the configuration of the VH–VL order in the h528 scFv multimers may have been superior to the VL–VH order for cross-linking or simultaneous binding to EGFRs, suggesting that it is important to consider the domain order when constructing functional antibody fragments.
The ability of scFv multimers to exhibit multivalent binding and to cross-link adjacent surface receptors in vitro and in vivo obviously depends on the flexibility between the Fv modules and the density and orientation of the antigen-binding sites, as well as the domain order and the structure of the receptor . For example, the anti-CD19 scFv trimers were predominantly inactive and showed only monovalent binding to the target cell surface [26, 28], and in the case of anti-Lewisy scFv, the trimers dissociated into inactive monomers in a rapid equilibrium . By contrast, the HLG0 trimers in this study showed the highest growth inhibition effects and higher apparent affinity compared with the HLG1 dimers (Figs 2 and 3, Table 1), suggesting that HLG0 forms functional trimers with three active binding sites and that EGFR is an ideal target for therapeutics using specific scFv multimers. Further, the specificity against EGFR of the HLG1 dimers and the HLG0 trimers was confirmed by competitive inhibition assay using FITC-labeled parental Fab and flow cytometric analysis (data not shown).
The trimeric structure of HLG0 was retained at lower concentrations and in long-term storage, and also showed prolonged blood-retention time corresponding to the calculated molecular mass (Fig. 5, Table 1). The stability of the multimeric structure of scFv may dependent on the affinities of the individual VH and VL domains to associate with each Fv . A strong interaction between VH and VL in h528 Fv has been implied by crystallographic analysis , and we previously reported that this strong interdomain interaction drives the formation of the homogeneous active bispecific diabody based on h528 Fv . In this study, the stability of h528 Fv might also contribute to the homogeneous multimeric structure of h528 scFv. Consequently, inhibitory effects of h528 scFv on EGFR phosphorylation were enhanced by multimerization (Fig. 4), and h528 scFv multimers showed tumor growth inhibitory activity in vitro and in vivo that was comparable with that of approved anti-EGFR IgGs without any other immune functions (Fig. 6).
To date, several alternative binding scaffolds have been developed [31, 32], and one of which designed ankyrin repeat proteins (DARPins) are a particularly promising example . Designed ankyrin repeat proteins can be expressed very well in the cytoplasm of E. coli, and recently it was reported that tumor growth inhibition effects of designed ankyrin repeat proteins targeting EGFR with detailed inhibitory mechanism . By contrast, scFv multimers have the advantage that they can be prepared from the existing well-characterized scFvs and also from IgGs. Although the current yields of the h528 scFv multimers were low (< 1 mg·L−1), large-scale fermentation of scFv multimers has been reported . We are now working to increase the yield of the h528 scFv multimers, especially that of the HLG0 trimer.
Taken together, although more detailed in vitro experiments such as analyses of downstream signaling and in vivo experiments such as biodistribution are needed, our data suggest that anti-EGFR scFv multimers are attractive candidates for the next generation of anti-EGFR therapeutic IgGs and offer the possibility of low-cost production.
Materials and methods
Construction of expression vectors for anti-EGFR scFv multimers
The bacterial expression vector for h528 scFv in the VH–VL order with a 16-amino-acid, glycine-rich linker (designated as HLG3 in this report) was constructed as described previously . Based on this vector, we constructed the vectors for h528 scFv with a six-amino-acid linker and without a linker (designated as HLG1 and HLG0, respectively) by use of overlap extension PCR to produce anti-EGFR scFv multimers. The vectors for h528 scFv in the VL–VH order with a five-amino-acid linker and without a linker (designated as LHG1 and LHG0, respectively) were similarly constructed. The sequences of the linkers and the neighboring sequences are shown in Fig. 1(A).
Preparation of anti-EGFR scFv multimers
scFvs were prepared by using the bacterial expression system described previously . Briefly, the constructs were expressed individually in E. coli strain BL21 (DE3) (Life Technologies, Carlsbad, CA, USA) and purified through immobilized metal-affinity chromatography from bacterial supernatant and periplasmic fractions. Gel-filtration analysis with a Hiload Superdex 200-pg column (26/60; GE Healthcare Bio-Science, Piscataway, NJ, USA) was used for further purification. The column was equilibrated with NaCl/Pi, and then 5 mL of purified scFv was loaded onto the column at a flow rate of 2.0 mL·min−1. Gel-filtration analysis with a Hiload Superdex 200-pg column (10/300; GE Healthcare) was used to confirm the monodispersity of each fractionated LHG0 and to evaluate the long-term stability of the scFvs in storage. For these experiments, the column was equilibrated with NaCl/Pi, and then 0.25 mL of scFv was loaded onto the column at a flow rate of 0.5 mL·min−1.
Human bile duct carcinoma (TFK-1) and human epidermoid cancer (A431) cell lines were used in this study. The TFK-1 cell line was established in our laboratory ; A431 cells were obtained from the Cell Resource Center for Biomedical Research (Institute of Development, Aging, and Cancer, Tohoku University, Sendai, Japan). These cell lines were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum, 100 U·mL−1 penicillin, and 100 μg·mL−1 streptomycin.
In vitro growth inhibition assay
In vitro growth inhibition of cancer cells was assayed with a 3-(4,5-dimethylthiazole-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium inner salt assay kit (CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay; Promega, Madison, WI, USA) according to previous reports [20, 38, 39]. In brief, cancer cells (2000 cells in 200 μL of culture medium containing 0.5% fetal bovine serum), were added to a 96-well plate, which was incubated overnight to allow the cells to adhere to the well. After the culture medium was removed by aspiration, scFvs or control IgGs were added to each well, and the cells were incubated for 96 h at 37 °C. Each well was then washed three times with NaCl/Pi, and detection reagent was added to each well. The cell viability of the target cells was calculated according to the following equation:
The interactions between sEGFR and scFvs were analyzed by using SPR spectroscopy (Biacore 2000; GE Healthcare). The methods for the expression and purification of sEGFR have been described previously . sEGFR was immobilized on the cells in a CM5 sensor chip up to 4066 resonance units. Various concentrations of scFv multimers in NaCl/Pi containing 0.005% Tween 20 were allowed to flow over the bound sEGFR at a flow rate of 20 μL·min−1 at 25 °C. The surface was regenerated with 10 mm glycine–HCl (pH 2.0) with no loss of activity. The data were referenced by subtracting the response of a blocked blank cell. biaevaluation software (GE Healthcare) was used to analyze the data. Kinetic parameters were calculated by means of a global fitting analysis with the assumption of a 1 : 1 Langmuir binding model.
Western blot analysis
Tyrosine phosphorylation of EGFR was evaluated by western blot analysis. A431 cells were precultured with each antibody at 100 nm for 1 h at 37 °C and then stimulated in culture medium containing 1 μg·mL−1 EGF for 15 min. Cell lysates were then separated by SDS/PAGE under reducing conditions and transferred to nitrocellulose by electroblotting. Tyrosine-phosphorylated EGFR were detected by using Phospho-EGF Receptor (Tyr1068) (D7A5) XP Rabbit mAb (Cell Signaling Technology Japan, Tokyo, Japan). Specific binding was visualized by using, horseradish peroxidase-linked anti-(rabbit IgG) (CST Japan) and the ECL Detection System (GE Healthcare). Horseradish peroxidase-linked monoclonal rabbit anti-(β-actin) (13E5) (CST Japan) was used as the loading control.
Radiolabeling of scFvs
An iodogen tube was prepared by coating a microfuge tube (1.5 mL) with Iodogen (100 μg·tube−1; Thermo Fisher Scientific, Waltham, MA, USA) in-house and used to radiolabel scFvs with [125I]NaI (74 MBq·0.1 mL−1; Perkin–Elmer, Wellesley, MA, USA). ScFvs (600–900 μL, 89–108 μg) and [125I]NaI (20–25 μL, 26–37 MBq) were placed in the iodogen tube and incubated for 15 min at room temperature with vortex mixing. The 125I-labeled scFv was then separated from the unreacted [125I]NaI by size-exclusion chromatography with a Bio-Gel P-6 Desalting Cartridge (10 mL; BioRad, Hercules, CA, USA), eluting with NaCl/Pi-Tween 20 (0.05%) at a flow rate of 1.5 mL·min−1. Radiochemical purities of the isolated scFvs ranged from 94 to 98%.
The Ethics Committee for Experimental Research in Animals of Tohoku University (Sendai, Japan) approved the study protocol. Male imprinting control region mice (6 weeks old, 27–31 g) were injected in the lateral tail vein with the 125I-labeled scFv (2 μg, 150–180 kBq) in a NaCl/Pi solution (0.2 mL). A portion of blood (~ 10 μL) was collected from the contralateral tail vein at 1.5, 3, 5 and 8 h post injection (n = 5 at each time point). Radioactivity and weight of the blood were measured with a gamma counter (AccuFLEXγ7000; Hitachi Aloka Medical, Tokyo, Japan). Blood radioactivity was expressed as the standardized uptake value (SUV), which was defined as follows: SUV = (blood radioactivity/blood weight)/(injected radioactivity/body weight).
In vivo tumor models
For each mouse, 5 × 106 TFK-1 cells in a final volume of 0.15 mL of NaCl/Pi were injected subcutaneously into the dorsal thoracic wall of female, 5-week-old severe combined immunodeficient mice (CLEA Japan, Tokyo, Japan). Then, five mice per group were treated intravenously with 20 μg of antibodies or NaCl/Pi on day 10 after tumor inoculation. This treatment was repeated once daily for three consecutive days. Tumor size was measured weekly by caliper, and the approximate tumor volume (V, in mm3) was calculated from linear measurements of the width (A, in mm) and length (B, in mm) as follows: V = (A2 × B)/2. Experiments involving mice were reviewed by the Ethics Committee for Experimental Research in Animals of Tohoku University and were performed under the Guidelines for Animal Experiments of Tohoku University and according to the laws and notifications of the Japanese government.
We would like to thank Mr Yuichi Ito for his technical assistance. This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Science, Sports, and Culture of Japan (R.A. and I.K.) and by grants from the New Energy and Industrial Technology Development Organization (NEDO) of Japan. Additional support was provided through the the Advanced research for medical products Mining Programme of the National Institute of Biomedical Innovation (NIBIO).