Construction and Characterization of Novel, Recombinant Immunotoxins Targeting the Her2/neu Oncogene Product: In vitro and In vivo Studies

The Her2/neu proto-oncogene encodes a 185-kDa transmembrane glycoprotein kinase with extensive homology to the epidermal growth factor receptor (HER1; refs. 1–3). Amplification of the gene and overexpression of the Her2/neu protein product in tumor cells have been well described in numerous human cancers, including mammary and ovarian carcinomas and gastric and lung tumors (4–6). Because Her2/neu plays a central role in malignant transformation and growth, it provides an attractive target for focused therapeutic approaches.

A number of approved immunotherapeutic agents are directed at tumors that express high levels of Her2/neu, such as the monoclonal antibody trastuzumab (Herceptin), and small-molecule inhibitors such as gefitinib (Iressa) have shown promising results, but the development of resistance to treatment remains a well-known problem (7, 8). To enhance its clinical potential, cell-surface Her2/neu has been targeted using antibody-drug conjugates (9) or immunotoxins, composed of plant or bacterial toxins linked with a targeting molecule composed of monoclonal antibodies or antibody fragments (10, 11). Previously, a recombinant, murine anti-Her2/neu single-chain antibody (scFv), designated e23, has been fused to catalytic toxins such as Pseudomonas exotoxin A and diphtheria toxin (DT) to specifically target Her2/neu-expressing cells (12, 13). A major drawback of such proteins is their potential for immune response after repeated administration. Further complications could result from nonspecific binding of foreign proteins to vascular endothelial cells, leading to vascular leak syndrome and, ultimately, interstitial edema and organ failure (14, 15).

The development of immunotoxins containing human or humanized components may circumvent these problems. Such immunotoxins may display reduced immunogenicity although antibodies to the toxin components may still limit prolonged therapy (16). We previously reported in vitro characterization and in vivo antitumor efficacy studies of an immunotoxin composed of the human chimeric anti-Her2/neu antibody (BACH-250) chemically conjugated to recombinant gelonin (rGel; ref. 17). rGel is a 29-kDa ribosome-inactivating plant toxin with a potency and mechanism of action similar to that of ricin toxin A-chain, but with improved stability and reduced toxicity (18, 19). The BACH-250/rGel conjugate showed potent and specific cytotoxicity against Her2/neu-overexpressing human tumor cells in culture and against SKOV3 tumor xenografts. However, the treatment of solid tumors presents a potential problem because full-length antibodies must diffuse into the tumor against a hydrostatic pressure gradient and into disordered vasculature (20, 21).

Marks and colleagues previously described an anti-Her2/neu scFv, designated C6.5, which was selected from a human scFv phage display library and affinity-matured in vitro (22). Using scFv C6.5, McCall and colleagues constructed and characterized a bispecific scFv composed of C6.5 and anti-CD16 scFv, displaying a high level of in vitro tumor cell cytotoxicity and in vivo tumor targeting (23). Studies by Park and colleagues generated immunoliposomes containing doxorubicin, which were targeted to tumor cells using antibody C6.5. These constructs showed selective enhancement of the therapeutic index of doxorubicin chemotherapy (24). Most recently, Adams and colleagues successfully used a C6.5 diabody construct as a radioimmunotherapeutic agent containing (²¹¹At) for the treatment of Her2/neu-positive solid tumors in xenograft models, showing that scFv C6.5 could be used effectively in vehicles for targeted radioimmunotherapy by using powerful, short-lived α-emitting radioisotopes (25).

In the present study, we describe the construction and characterization of several rGel-based chimeric toxins composed of the scFv e23 or C6.5 and using various linker configurations to examine how different antibodies and linker choices affect the in vitro and in vivo efficacy of fusion constructs.

The gene encoding murine scFv e23 was obtained from Oncologix, Inc., and human scFv C6.5 was kindly supplied by Prof. James D. Marks (University of California, San Francisco, San Francisco, CA). Illustrations of the immunotoxin constructs are shown in Fig. 1A. Recombinant immunotoxins containing rGel and either e23 or C6.5 and the flexible linker (GGGGS, designated “L”) were constructed by overlapping PCR. These proteins were designated e23-L-rGel and C6.5-L-rGel, respectively. In addition, we generated two different linkers (TRHRQPRGWEQL, designated “Fpe,” and AGNRVRRSVG, designated “Fdt”) containing furin cleavage sites from Pseudomonas exotoxin A and DT, incorporated between the C6.5 and rGel components. The intent was that furin cleavage would allow more facile release of the toxin from the complex once internalized by the endosomal compartment. The resulting proteins were designated C6.5-Fpe-rGel and C6.5-Fdt-rGel, respectively. All the fusion genes were cloned into a T7 promoter–based E. coli expression vector, pET-32a(+).

To express the recombinant fusion proteins, bacterial cultures were incubated at 37°C in LB growth medium with strong antibiotic selection (400 μg/mL ampicillin, 70 μg/mL chloramphenicol, and 15 μg/mL kanamycin) and grown to log phase (A_{600 nm} = 0.8). The target protein expression was induced at 18°C with 0.1 mmol/L isopropyl-l-thio-β-d-galactopyranoside for 16 h. Induced bacterial cultures were then centrifuged and stored frozen at −80°C.

Frozen bacterial pellets containing different immunotoxins were purified as follows: The pellets were allowed to thaw with the addition of 50 mmol/L Na-phosphate (pH 7.6), 300 mmol/L NaCl, 5 mmol/L imidazole. The bacterial suspension was further lysed by passing the material through a microfludizer (Microfluidics). The solution was then clarified by centrifugation (40,000 rpm), and the supernatants were loaded onto columns (2.5-cm internal diameter × 13 cm) containing cobalt-charged IMAC resins. After washing with 50 mmol/L Na-phosphate (pH 7.6), 300 mmol/L NaCl, 15 mmol/L imidazole, the bound protein was eluted with 50 mmol/L Na-phosphate (pH 7.6), 300 mmol/L NaCl, 300 mmol/L imidazole. Fractions containing immunotoxin were dialyzed in 20 mmol/L Tris-HCl (pH 7.4) and 150 mmol/L NaCl, followed by digestion with recombinant enterokinase. The purified immunotoxins were dialyzed in PBS, filter sterilized, and stored at 4°C.

The binding affinity and specificity of rGel-based immunotoxins containing either scFv e23 or C6.5 were evaluated by ELISA on Her2/neu-positive SKOV3 cells and Her2/neu-negative MCF7 cells. Rabbit anti-rGel antibody and horseradish peroxidase–conjugated goat anti-rabbit IgG was used as a tracer in this assay as described previously (26).

Immunofluorescence-based internalization studies were also done on SKOV3 and MCF7 cells (27). Cells treated with immunotoxins or rGel were subjected to immunofluorescent staining with anti-rGel antibody (FITC-conjugated secondary antibody). Nuclei were counterstained with propidium iodine. Visualization of immunofluorescence was done with a confocal laser scanning microscope (Zeiss LSM510, Carl Zeiss).

The rGel-induced inhibition of [³H]leucine incorporation into protein in a cell-free protein synthesizing system after the administration of various doses of immunotoxins was done as described previously (28).

Various C6.5/rGel fusions containing different linkers were treated with recombinant human furin (NEB) at various pH. For pH 5.4, 50 mmol/L sodium acetate buffer was used, and for pH 7.2, 10 mmol/L HEPES buffer was used. A dose of 2 units of purified furin was applied to 25 μg of fusion protein in each reaction. After incubation for 16 h at room temperature, the proteins were analyzed by Western blot with anti-rGel antibody. The rate of cleavage by furin was calculated using AlphaEaseFC software (version 4.0.1) from the following formula: cleavage rate = value of gray scale of cleaved protein / value of gray scale of total protein.

Log-phase cells were seeded (∼5 × 10³ per well) in 96-well plates and allowed to attach overnight. Cells were further incubated with various concentrations of immunotoxins, rGel, or medium at 37°C for 72 h, and cell viability was determined by using the crystal violet staining method followed by solubilization of the dye in Sorensor's buffer as described previously (26).

The intracellular rGel release from C6.5/rGel fusion constructs was analyzed in SKOV3 cells. After incubation for various times, the cells were treated with acidic glycine buffer [500 mmol/L NaCl and 0.1 mol/L glycine (pH 2.5)] to strip cell-surface–bound proteins and lysed in lysis buffer [10 mmol/L Tris-HCl (pH 8.0), 60 mmol/L KCl, 1 mmol/L EDTA, 1 mmol/L DTT, 0.2% NP40]. Cytosolic fractions were further quantified and analyzed by Western blot using anti-rGel antibody.

SKOV3 cells treated with 200 nmol/L immunotoxins were pelleted and lysed. Proteins from each cell lysates were analyzed by Western blot with antibodies that recognized poly(ADP-ribose) polymerase (PARP), β-actin, high mobility group box 1 (HMGB1; Santa Cruz Biotechnology), and microtubule-associated protein 1 light chain 3 (MAP LC3; Novus Biological). For HMGB1 release assay, the medium was harvested, concentrated, and further analyzed by Western blot.

BALB/c nude mice were injected s.c. with SKOV3 cells (5 × 10⁶ per mouse). Once tumors were measurable (∼40–50 mm³), animals were treated (i.v. via tail vein) with PBS or rGel as control or with different C6.5/rGel fusion proteins, every other day for 10 d. Animals were monitored and tumors were measured for an additional 40 d.

Twenty-four hours after i.v. injection of C6.5-L-rGel or rGel, the mice were sacrificed and tumor samples were collected and frozen immediately for sectioning. The sample slides were fixed in 3.7% paraformaldehyde, permeabilized with PBS containing 0.2% Triton X-100, and blocked in 5% nonfat milk in PBS. After incubation with anti-rGel antibody, the samples were subjected to immunofluorescent staining with FITC-conjugated secondary antibody and nuclear counterstaining with propidium iodine. The slides were mounted and delivered for the immunofluorescence observation under a Nikon Eclipse TS-100 fluorescence microscope (Nikon).

The initial rGel-based immunotoxins consisted of a flexible L linker (GGGGS) tethering the COOH terminus of e23 or C6.5 to the native rGel NH₂ terminus. V_H/V_L orientations determined the best binding activity of V_L-V_H for e23 and V_H-V_L for C6.5 (data not shown). The C6.5/rGel construct was further engineered by incorporating two different enzymatically sensitive furin cleavage linkers between the scFv and rGel toxin components. The two furin sensitive sequences are designated Fpe (TRHRQPRGWEQL) and Fdt (AGNRVRRSVG; Fig. 1A). Several biochemical studies have shown that the serine protease furin efficiently cleaves proteins containing these recognition sequences (29, 30).

Following purification, all the rGel-based immunotoxins migrated on SDS-PAGE at the expected molecular weight of 55 kDa (Fig. 1B). However, with the introduction of sensitive furin linker, C6.5-Fdt-rGel, but not C6.5-Fpe-rGel, displayed cleavage bands to some extent after rEK digestion. The cleavage was found to be occurring precisely at the predicted furin cleavage site producing the 27- to 28-kDa fragments of scFv and rGel. Further analysis indicated that the yields for each protein (per liter of bacterial culture) were 1.55 mg for e23-L-rGel, 1.05 mg for C6.5-L-rGel, 1.08 mg for C6.5-Fpe-rGel, and 0.70 mg for C6.5-Fdt-rGel.

To ensure that immunotoxins retained antigen binding ability, the fusion proteins were compared in an ELISA-based binding assay (Fig. 2A) using Her2/neu-positive SKOV3 and Her2/neu-negative MCF7 cells. The equilibrium dissociation constant, K_d, was further calculated (GraphPad Prism, v4.03). The affinity of e23-L-rGel (K_d = 8.5 nmol/L) for SKOV3 cells was similar to that of C6.5-L-rGel (K_d = 12.6 nmol/L). The K_d values were consistent with those previously measured in an in vitro live cell assay using scFv itself (22). In addition, both immunotoxins showed significant specificity based on the background of binding to MCF7 cells. ELISA assay suggested that the human scFv C6.5 displayed similar binding specificity compared with the murine e23.

The biological activity of toxins can be severely compromised when incorporated into fusion constructs. To examine the N-glycosidic activity of the rGel component of the immunotoxins, these materials were added to an in vitro protein translation assay using [³H]leucine incorporation by isolated rabbit reticulocytes. Inhibition curves for the fusion constructs e23-L-rGel and C6.5-L-rGel and native rGel were compared (Fig. 2B), and IC₅₀ values for the three molecules were found to be virtually identical (15.41, 15.52, and 10.6 pmol/L, respectively), suggesting that no loss of toxin activity occurred in the fusion molecules.

We next examined whether the e23-L-rGel and C6.5-L-rGel fusions could specifically internalize into target cells. Immunofluorescence staining was done on SKOV3 and MCF7 cells after exposure to the constructs. As shown in Fig. 2C, the rGel moiety of both fusions was observed primarily in the cytosol of SKOV3 after treatment, but not in MCF7 cells, showing that both constructs were comparable in efficient cell binding and rapid internalization and delivery of rGel toxin to the cytoplasm after exposure to Her2/neu-positive cells.

The e23-L-rGel and C6.5-L-rGel constructs and rGel were tested against a number of different tumor cell lines (Table 1). The SKBR3 cells with the highest level of Her2/neu expression were killed most efficiently by both antibody fusion constructs, with IC₅₀ values of 6.0 and 9.1 nmol/L for e23-L-rGel and C6.5-L-rGel, respectively. IC₅₀ values for rGel toxin were ∼200-fold higher (1,671 nmol/L). For the other Her2/neu-positive cells, both immunotoxins also showed similar IC₅₀ values, showing that the two fusion proteins possess very similar cell killing activity and specificity. Furthermore, MCF7 and 4T1 cells, which express relatively low levels of Her2/neu, showed little to no specific cytotoxicity of the fusion constructs compared with rGel itself, clearly showing that the presence of higher levels of cell-surface Her2/neu is required for specific cytotoxicity of the constructs.

From the in vitro study, it was evident that no significant differences were observed between murine e23– and human C6.5–based fusion constructs. Therefore, we focused on C6.5/rGel for further studies by incorporation of proteolytically cleavable linkers (Fpe and Fdt) to examine whether this change would improve killing efficiency. To investigate the susceptibility of various chimeric toxins to proteolytic cleavage, purified fusions were subjected to proteolysis with recombinant furin (Fig. 3A). At pH 7.2, cleavage of Fpe (18.5% of total) and Fdt (100%) was observed. At pH 5.4, Fpe was cleaved less efficiently (4.5% of total), but Fdt still displayed high cleavage efficiency (100%). In contrast, fusion with L linker was found to be highly stable and could not be cleaved at either pH. As indicated, the Fdt linker was the most sensitive to cleavage among all the constructs. In contrast, cleavage of the molecule containing the Fpe linker was highly dependent on pH. The L linker was found to be comparatively resistant to protease action without regard to the pH.

To investigate the kinetics of cytotoxicity by different C6.5/rGel fusions, their cell killing activities were assessed against SKBR3, SKOV3, Calu3, and MDA MB435S cells at various time points (Supplementary Table S1). Interestingly, the cell lines showed no differences in overall sensitivity to the fusion constructs with furin cleavage linkers compared to those with flexible L linker. All the fusions showed potent cytotoxicity after 48 hours and exerted highly potent cell killing at 72 hours. This suggests that the cleavage efficiency of different linkers for these chimeric toxins was not a major determinant of the overall cytotoxic effects observed with different linkers. Surprisingly, the cytotoxic kinetics of the constructs therefore seemed to be independent of the sensitivity of the constructs to proteolytic cleavage.

The intracellular release of rGel after endocytosis of various C6.5/rGel fusion constructs was assessed by Western blot with an anti-rGel antibody (Fig. 3B). During the treatment of SKOV3 cells, rGel release was found to be maximal at 2 hours after treatment with C6.5-L-rGel and 4 hours after exposure to C6.5-Fpe-rGel. For C6.5-Fdt-rGel, the rGel component was released within 1 to 2 hours and degraded simultaneously, corresponding to the status of full-length protein. The decreasing intracellular level of full-length C6.5-Fdt-rGel could be ascribed to rapid instability of the construct after internalization. Although the maximal rGel release of different fusions was achieved at different time points, the absolute amounts of delivered rGel found in the cytosol were virtually identical. Therefore, these data confirm the observation that introduction of an unstable furin cleavage linker does not improve the intracellular rGel release of the constructs.

The linkers between C6.5 and rGel showed a differential sensitivity to protease action, which may result in different clearance and metabolic kinetics in vivo (31, 32). To estimate the stability of various C6.5/rGel fusions, we incubated the purified proteins at 37°C for varying times in the presence of human plasma before testing cellular Her2/neu binding to SKOV3 cells (Fig. 3C). Our results showed that in the presence of human plasma, the C6.5-Fdt-rGel construct displayed a reduction in binding activity within 6 hours of incubation and a 20% loss of binding activity after 72 hours of incubation. In contrast, the C6.5-L-rGel and C6.5-Fpe-rGel fusion constructs showed only 9% and 12% reductions, respectively, after 72 hours of incubation.

In addition, the immunotoxins were evaluated for cytotoxic activity following incubation in human plasma for 0, 24, 48, and 72 hours (Fig. 3C). For the C6.5-Fdt-rGel construct, the cell killing activity was reduced more than 2-fold after 48 hours, as indicated by increasing IC₅₀ values of 20 versus 48 nmol/L. However, this was not the case for C6.5-L-rGel and C6.5-Fpe-rGel, which retained most of its cytotoxic activity even after 48 hours and displayed a little influence on IC₅₀ after 72 hours of incubation in plasma (16 versus 22 nmol/L and 17 versus 25 nmol/L for each construct). This functional stability analysis indicated that compared with the L and Fpe linkers, the Fdt linker was much more unstable in human plasma, and this may reduce the in vivo potency of potential therapeutic applications using constructs containing this linker design.

The cytotoxic effects mediated by C6.5/rGel fusions were analyzed to evaluate whether the cytotoxic mechanisms of the constructs observed included elements of apoptosis, necrosis, or autophagy in SKOV3 cells. As shown in Fig. 4A, C6.5/rGel fusions did not show activation of caspase-dependent apoptosis in SKOV3 cells and showed no cleavage of caspase substrate PARP. The terminal deoxyribonucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) results (Supplementary Fig. S1) confirmed that the cytotoxic effects of the rGel-based fusions were not mediated by apoptosis and DNA fragmentation.

To assess whether the necrotic cell death was induced, we examined lactate dehydrogenase (LDH) release, which is a marker of abrupt membrane lysis (33). In this case, treatment of SKOV3 cells with Triton X-100 serves as a positive control causing LDH release (Supplementary Fig. S2). In contrast, treatment with the fusion constructs failed to show LDH release, indicating that the observed cytotoxicity did not seem to be a result of necrosis.

We next asked if the immunotoxins activate autophagic signaling in SKOV3 cells. MAP LC3-I, known to be usually present in the cytosol, is palmitoylated during autophagy to form membrane-bound LC3-II and is associated with autophagosomes (34). As shown in Fig. 4B, the ratio of LC3-II formation to the β-actin control was increased after treatment with the fusion constructs, showing that autophagic flux was induced by C6.5/rGel fusions in SKOV3 cells. In addition, autophagy induction by C6.5/rGel fusions was further validated by the selective release of HMGB1 (Fig. 4C). Tumor cells that are dying by autophagy selectively release the nuclear HMGB1 protein without displaying characteristics of necrosis (35). These data indicated that the observed cytotoxic effects of C6.5/rGel fusions on SKOV3 cells seemed to be mediated not through an apoptotic or necrotic mechanism but by the efficient induction of autophagic cell death.

We evaluated the ability of various C6.5/rGel fusion constructs to inhibit the growth of established SKOV3 tumor xenografts in nude mice after systemic administration. Tumors were induced in nude mice by s.c. injection of SKOV3 cells on day 0, and treatment was initiated on day 9 post-injection when the tumors were well established. Treatment consisted of five i.v. injections every other day. Groups of mice were treated at doses of 40 and 20 mg/kg for each fusion construct. Control mice were treated with PBS or 20 mg/kg rGel only. As shown in Fig. 5A and B, treatment with C6.5-L-rGel exhibited a significant antitumor effect. Mice treated with the 40 mg/kg dose of C6.5-L-rGel showed a long-lasting antitumor effect that lasted more than 1 month until the animals were sacrificed. In mice treated with the 20 mg/kg dose, tumor growth was, in most cases, arrested for the duration of the treatment and resumed a couple of weeks after its completion. Otherwise, treatment of mice with 40 mg/kg C6.5-Fpe-rGel resulted in a significant delay in tumor growth. This was similar to the effect observed with the same dose of C6.5-L-rGel, but no significant effect could be observed at the lower (20 mg/kg) dose level. In contrast, mice treated with C6.5-Fdt-rGel at either dose (40 or 20 mg/kg) showed no specific antitumor effect above that observed with rGel alone.

We next examined the localization of C6.5-L-rGel and rGel after administration to mice bearing SKOV3 tumors. Immunofluorescence staining confirmed that C6.5-L-rGel localized specifically in tumor tissue, but no staining was observed in tumors after administration of rGel itself (Fig. 5C). This suggests that the fusion construct C6.5-L-rGel can effectively target tumor cells overexpressing Her2/neu in vivo and can show significant tumor growth–suppressive effects in the absence of observable toxicity.

The development of recombinant immunotoxins has markedly affected the field of targeted therapeutics by allowing numerous molecular design approaches to engineer molecules with improved in vitro and in vivo performance characteristics (36, 37). Detailed investigations to design protein-based recombinant molecules with reduced immunogenicity without affecting efficacy, toxicity, or specificity have not generally focused on anti-Her2/neu agents (26, 38, 39). To the best of our knowledge, this is one of the first comprehensive examinations of the effects of immunotoxin design on the ability of constructs to effectively target in vitro and in vivo cells expressing the Her2/neu oncogene product.

Previous studies have shown the potent in vitro and in vivo xenograft efficacy of immunotoxins containing murine scFv e23 (12, 13). However, in cancer patients, i.v. administration of e23-based fusion constructs was shown to result in severe liver toxicity, and effective doses could not be achieved (40, 41). The current study shows that the human scFv C6.5 isolated and characterized by Marks and colleagues binds to Her2/neu with high affinity (K_d of 2.0 × 10⁻⁸ mol/L; ref. 22). rGel-based immunotoxins containing either e23 or C6.5 were shown to have virtual identical patterns of binding (ELISA), protein synthesis inhibitory activity (rabbit reticulocyte lysate assay), and internalization against both Her2/neu-positive and Her2/neu-negative cell lines. Evaluation of the in vitro cytotoxic effects of e23 and C6.5 fused to rGel showed that human scFv C6.5 possessed a comparable affinity and specificity to the murine e23 and could be regarded as an effective replacement.

We further described the construction and functional characteristics of novel C6.5/rGel constructs, with the introduction of three different linker sequences. In addition to the flexible L linker (GGGGS), two other furin cleavage linkers, namely, Fpe (TRHRQPRGWEQL) from Pseudomonas exotoxin A and Fdt (AGNRVRRSVG) from DT, were incorporated between the scFv and the toxin.

Furin is a cellular endoprotease and has been implicated in the proteolytic activation of large numbers of secreted proteins (29, 30). Inclusion of furin-cleavable linkers in fusion constructs containing ribotoxin, caspase-3, or granzyme B has resulted in a significant improvement in specific toxicity compared with constructs containing stable linkers (42, 43). Our studies clearly showed that incorporation of furin linkers into rGel-based constructs resulted in constructs with increased protease cleavage of the rGel from the scFv carrier. However, this did not seem to make a significant difference in the total levels of intracellular rGel, and the in vitro specific cytotoxicity of the construct was not improved compared with the more stable L linker. Intracellular trafficking of rGel-based constructs does not seem to be consistent with the model concept that the enzymatic dissolution of the link between C6.5 and rGel by furin (or other enzymes) is a rate-limiting step in determining the cytotoxic effects of rGel fusion toxins.

Depending on the lethal stimulus, tumor cells can die by distinct cell death mechanisms including apoptosis, necrosis, and autophagy (44, 45). Although rGel-based immunotoxins have been used in the treatment of malignant cancers for more than 15 years, the actual mechanisms behind the induction of cell death have remained unclear (46, 47). In this study, we showed that C6.5/rGel fusions had no effect on apoptosis-related mechanisms (PARP cleavage, TUNEL, and DNA fragmentation) or necrosis-related mechanisms (LDH release). However, we clearly showed that the constructs induced autophagic cell death on SKOV3 cells as assessed by LC3-I to LC3-II conversion and HMGB1 selective release. Further studies will be needed to determine whether the C6.5/rGel fusions induced general autophagic cell death among different cell types, and whether autophagic cell death can be attributed to other rGel-based immunotoxins.

Our initial goals for this study were to create an optimal construct targeting Her2/neu on tumor cell surfaces with relatively low immunogenicity, thus providing an opportunity for long-term, repeated administration, and with significant selectivity and stability. To that end, we performed a stability study in human plasma, and the data showed that the C6.5-Fdt-rGel construct was the least stable in binding activity and cytotoxicty after incubation because of the highest cleavage sensitivity of the Fdt linker. On the other hand, C6.5-L-rGel showed the highest degree of stability after incubation in plasma of all molecules tested. The in vivo antitumor study further confirmed that C6.5-L-rGel was the most active construct against established tumor xenografts. The two companion molecules with the furin-cleavable linkers showed significantly less activity in vivo. The enzymatic stability of the linker between antibody and toxin can seriously affect the pharmacokinetics of immunotoxins and can apparently affect loss of targeting function and in vivo efficacy.

The animal study presented highlights an additional point regarding the on-target toxicity for the immunotoxins in future clinical applications. On-target toxicity, or the toxicity of the construct to normal tissues as a result of expression of the target antigen, is a common problem that is addressed in clinical trials (48). In this case, the C6.5 antibody does not cross-react with the murine Her2/neu analogue. Therefore, the toxicity of the C6.5/rGel fusions may be underestimated in these murine studies.

In conclusion, we have designed and developed several novel immunotoxins containing the human scFv C6.5 and the toxin rGel. These agents exhibit efficient cytotoxicity against Her2/neu-overexpressing tumor cells, and the human antibody seems to be virtually identical as an effective carrier of rGel toxin to the murine e23. The introduction of a furin-cleavable linker between C6.5 and rGel did not result in improved intracellular rGel release and cytotoxic effects in vitro, despite showing more sensitivity to protease cleavage. In addition, the incorporation of a furin-cleavable linker resulted in a decrease in in vivo antitumor efficacy compared with a noncleavable linker. These studies clearly show the highly individualized nature of some payloads and targeted constructs, and that observations about similar types of toxin payloads do not necessarily translate to other payloads.

No potential conflicts of interest were disclosed.

Grant support: This research work was conducted, in part, by the Clayton Foundation for Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Walter N. Hittelman for assistance with confocal microscopy techniques.