In Vitro and in Vivo Cytotoxic Activities of Recombinant Immunotoxin 8H9(Fv)-PE38 against Breast Cancer, Osteosarcoma, and Neuroblastoma Free

Immunotoxins (ITs) are composed of monoclonal antibodies (MAbs) or fragments of MAbs attached to a toxin. Initially, ITs were produced by chemically conjugating whole MAbs to plant or bacterial toxins such as ricin, saporin, pokeweed antiviral protein, diphtheria toxin, or Pseudomonas exotoxin A (PE). More recently ITs have been produced using genetic and protein engineering, in which the Fv of a MAb is fused to the toxin and the recombinant protein produced in Escherichia coli (1). Our laboratory has focused on the development of recombinant ITs (RITs) in which the Fv of a MAb is fused to a 38-kDa truncated mutant form of PE. These RITs are made by deleting the cell-binding domain of PE (amino acids 1–252) and replacing it with the Fv portion of a MAb. During the past several years, we have made a variety of RITs using different MAbs (2, 3, 4, 5, 6, 7). Five of these RITs have been evaluated or are now being evaluated in Phase I clinical trials in patients with cancer (8, 9). During the past several years, we have completed Phase I clinical trials with two of these ITs. One is BL22, which targets CD22 on B-cell malignancies and has produced complete remissions in the majority of patients with drug-resistant hairy cell leukemia. The other one is LMB-2 [anti-Tac(Fv)-PE38], which has produced one complete remission in hairy cell leukemia and partial remissions in Hodgkin’s disease, cutaneous T-cell lymphoma, and adult T-cell leukemia as well as in three patients with hairy cell leukemia. These clinical successes have encouraged us to find new targets for RIT therapy.

MAb 8H9 is a murine IgG1 derived from the fusion of mouse myeloma SP2/0 cells and splenic lymphocytes from BALB/c mice immunized with a human neuroblastoma. Using immunohistochemistry, MAb 8H9 was shown to be highly reactive with human brain tumors, childhood sarcomas, and neuroblastomas (10). In contrast, 8H9 is not reactive with normal human tissues. Immunofluorescence studies show that the 8H9 antigen is present on the external surface of tumor cell membranes. The antigen is not yet fully characterized but has the properties of a glycoprotein (10). The presence of the antigen on the surface of cancer cells makes it a promising target for IT therapy.

In the current study, we have constructed a RIT containing the 8H9 single-chain Fv (scFv) and shown that 8H9(scFv)-PE38 selectively kills cells that react with the MAb 8H9 and produces regressions of two human cancers growing in severe combined immunodeficient (SCID) mice that express the 8H9 antigen. In addition, we have constructed a disulfide-linked Fv (dsFv) RIT that is more suitable for clinical development because dsFv RITs are more stable and are produced in better yields during refolding and purification (11). 8H9(dsFv)-PE38 is cytotoxic to MCF-7 cells, produces tumor regressions in nude mice, and is well tolerated by monkeys.

Human neuroblastoma cell lines were provided by Dr. Robert Seeger (LA-N-1), Children’s Hospital of Los Angeles (Los Angeles, CA), and by Dr. Shuen-Kuei Liao (NMB-7), McMaster University (Ontario, Canada). Cell lines were cultured in 10% fetal bovine serum in RPM1 1640 with l-glutamine, penicillin, and streptomycin. The human osteosarcoma cell line OHS was established at the Norwegian Radium Hospital. It was maintained for several passages in DMEM supplemented with 10% fetal bovine serum and penicillin-streptomycin. OHS-M1 is a subline of OHS, isolated from a tumor growing s.c. in SCID mice. L428 (from Dr. C. S. Duckett; NIH, Bethesda, MD) is a Hodgkin’s lymphoma cell line.

DNA encoding the 8H9 Fv in a single-chain form has been described previously (12). Primers were designed to clone the DNA fragment encoding 8H9 Fv into the PE38 expression vector. The V_H 5′ primer introduced a NdeI restriction site (underlined), and the V_L 3′ primer introduced a HindIII restriction site (underlined) to facilitate cloning of the single-chain antibody variable domain (scFv) into the expression vector. Because the cloned Fv contained an uncommon residue at position 3(K) (Kabat number) in the V_H, Fv was designed and produced as follows. Lysine at position 3 of the V_H was substituted with Q. The following primers were used to make the scFv: VL3′, 5′-CTC-ggg-ACC-TCC-ggA-AgC-TTT-CAg-CTC-CAg-CTT-ggT-CCC-AgC-3′; and VH5′K3Q, 5′-AgC-TgC-Tgg-ATA-gTg-CAT-ATg-CAg-gTC-CAA-CTg-CAg-CAg-TCT-ggg-gCT-gAA-CTg-3′. PCR fragments were digested with NdeI and HindIII restriction enzyme and cloned into the NdeI-HindIII site in the expression vector (Ref. 13; Fig. 1). With regard to the making of dsFv, the positions of disulfides for the stabilization of B3(Fv) were identified using a computer-modeled structure of B3(Fv) that was generated by mutating and energy minimizing the amino acid sequence and structure of McPC603, as described previously (14). The amino acid sequences of 8H9(Fv) were simply aligned with those of B3(Fv) to determine the positions to insert cysteine residues. For the construction of 8H9(dsFv) fragments, cysteine residues were introduced in the V_H and V_L using PCR as described previously (15). The following primers were used to make the dsFv: STUVH, 5′- Tgg-gTg-Agg-CAg-Agg-CCT-gAA-CAg-TgT-CTT-gAg-Tgg-ATT-ggA-Tgg-ATT-TTT-3′; HinH, 5′-gCC-TgA-ACC-gCA-AgC-TTg-TgA-ggA-gAC-ggT-gAC-CgT-ggT-CCC-3′; PNDEL, 5′-TCT-ggC-ggT-ggC-CAT-ATg-gAC-ATC-gAg-CTC-ACT-CAg-TCT-CCA-ACC-ACC-3′; and EcoL, 5′-CTC-ggg-AgA-ATT-CTA-TCA-TTT-CAg-CTC-CAg-CTT-ggT-CCC-ACA-ACC-gAA-CgT-gAg-Cgg-AAA-gCT-gTg-3′. The primers STUVH and EcoL replaced Gly⁴⁴ in the V_H chain and Ala¹⁰⁰ in the V_L chain with cysteines, respectively (in bold). These primers introduce restriction enzyme sites (underlined) for easy cloning of the V_L chain into NdeI-EcoRI site and easy cloning of the V_H chain into StuI-HindIII site in the expression vector.

8H9(scFv)-PE38 or the two components of 8H9(dsFv)-PE38 (V_L and V_H -PE38) were expressed in E. coli, BL21(λDE3) and accumulated in inclusion bodies, as described previously (13). Inclusion bodies were solubilized in guanidine hydrochloride solution, reduced with dithioerythritol, and refolded by dilution in a refolding buffer containing arginine to prevent aggregation and oxidized and reduced glutathione to facilitate redox shuffling. Active monomeric protein was purified from the refolding solution by ion-exchange and size-exclusion chromatography (16, 17). Protein concentration was determined by Bradford Assay (Coomassie Plus; Pierce, Rockford, IL). For the primate study, a special batch of 8H9(dsFv)-PE38 was produced using precautions to remove endotoxin. The endotoxin content is less than 6 endotoxin units (EU)/mg.

The specific cytotoxicity of each IT was assessed by inhibition of protein synthesis by cells exposed to various concentrations of IT. Protein synthesis was measured as cellular incorporation of [³H]leucine (13, 17). Cells, at a concentration of 1.6 × 10⁴ cells/well, were plated in 96-well plates and incubated overnight. IT was diluted in PBS/0.2% BSA to desired concentrations and added to target cells in triplicate. The cells were incubated for 20 h at 37°C before the addition of 2 μCi [³H]leucine/well and further incubation for 2 h at 37°C. Cells were frozen, thawed, and harvested onto glass fiber filter mats using automated harvester. The radioactivity associated with the cells was counted in an automated scintillation counter. For competition experiments, excess 8H9 MAb or T6 MAb was added 15 min before the addition of the IT (15.5 ng/ml).

Groups of 5–10 female NIH Swiss mice were given single injections of escalating doses of ITs i.v. through the tail vein, as described previously (16). Animal mortality was observed over 2 weeks. The LD₅₀ was calculated with the Trimmed Spearman-Karber program version 1.5, from the Ecological Monitoring Research Division, Environmental Monitoring Systems Laboratory, United States Environmental Protection Agency.

The monkey studies were performed at the National Cancer Institute under an approved protocol (LMB-045). For the toxicology studies, one 9-kg monkey received injection with 8H9(dsFv)-PE38 (0.1 mg/kg, i.v., QOD × 3), and the other 5-kg monkey received injection with 8H9(dsFv)-PE38 (0.2 mg/kg, i.v., QOD × 3). Plasma samples were obtained 10 min after each dose for blood level measurements and on days 1, 5, 8, and 15 for blood chemistry measurements. To determine the blood levels of the RIT in monkeys, plasma samples diluted 200–400× were incubated with MCF-7 cells overnight in cytotoxicity assay, which is described in “Cytotoxicity Assay,” and active IT was quantitated by interpolation on a standard curve made from the cytotoxicity of purified IT (16).

The antitumor activity of RITs was determined in SCID mice bearing human cancer cells. MCF-7 cells (2 × 10⁶) were injected s.c. into SCID mice on day 0. Tumors (about 0.05 cm³ in size) developed in animals by day 4 after tumor implantation. Starting on day 4, animals were treated with i.v. injections of each of the RITs diluted in 0.2 ml of PBS/0.2% human serum albumin (HSA). Therapy was given once every other day on days 4, 6, and 8; treatment groups consisted of 5 or 10 animals. Tumors were measured with a caliper every 2 or 3 days, and the volume of the tumor was calculated by using the following formula: tumor volume (cm³) = length × (width)² × 0.4. Two days before implanting MCF-7 cells, 17β-estradiol pellets (0.72 mg, 60-day release; Innovative Research of America, Sarasota, FL) were implanted s.c. because MCF-7 cells are estrogen dependent for growth. For the osteosarcoma model, 1.5 × 10⁶ OHS-M1 cells were planted s.c. without implanting 17β-estradiol pellets and treated using the identical protocol.

Tumor sizes in animal experiments are expressed as mean ± SD. For comparison between the two experimental groups, Mann-Whitney test was used. P < 0.05 is considered statistically significant.

To determine whether 8H9(scFv) could target a cytotoxic agent to antigen-positive cells, we constructed two different RITs. Initially, we made a single-chain IT in which the Fv portion of MAb 8H9 is fused to PE38, a truncated form of PE. In the Fv, lysine at position 3 of the V_H is mutated to glutamine because glutamine is the most frequent amino acid in this position, and the yields are often improved by this mutation (17, 18). Because the yield of the scFv IT was low (Table 1), we also constructed a more stable disulfide-linked IT (dsFv RIT), in which the light and heavy chains are linked by a stable disulfide bond. This procedure not only increases stability but often has the further advantage of increasing recombinant protein yield (11). Both types of ITs were produced in E. coli and purified by ion-exchange and size-exclusion chromatography after renaturation from inclusion bodies as described previously (17). Each RIT eluted as a monomer on TSK gel filtration chromatography, and each migrated as a single band of about 62 kDa in SDS-PAGE (Fig. 2). IT 8H9(scFv)-PE38 was prepared from a 1-liter culture of E. coli. After extensive washing, we recovered inclusion body protein (100 mg) that was used to make IT. The final yield was 1.7 mg or 1.7%. In contrast, 8H9(dsFv)-PE38 is prepared by combining inclusion body protein from cells grown separately that express the V_L protein or the V_H-PE38 protein. When 33 mg of V_L protein were combined with 67 mg of V_H-PE38 protein, we recovered 16 mg of purified IT or a 16% yield (Table 1; Ref. 19). Because of this high yield, the dsFv molecule is better suited for further preclinical development.

The ability of the 8H9(Fv)-PE38 to inhibit protein synthesis was used as a measure of its cytotoxic effect. We exposed a variety of antigen-positive cell lines and two antigen-negative cell lines to the RIT for 20 h and then measured [³H]leucine incorporation. MCF-7 cells, which react strongly with the 8H9 antibody, were the most sensitive to 8H9(scFv)-PE38 with an IC₅₀ of 5.0 ng/ml (Fig. 3; Table 2). On two other breast cancer cell lines, BT-474 and ZR-75-1, which also react with MAb 8H9, the IC₅₀ values were 20 and 35 ng/ml. Three osteosarcoma cell lines, U2OS, CRL1427, and OHS-M1, were also sensitive. The IC₅₀ values were 30, 50, and 20 ng/ml, respectively. U2OS, CRL1427, and OHS-M1 are known to react with MAb 8H9. Also, three neuroblastoma cell lines, NMB-7, LA-N-1, and SK-N-BE(2), are sensitive to 8H9(Fv)-PE38 with IC₅₀ values of 9.0, 12.5, and 90 ng/ml, respectively. On two cell lines that do not react with MAb 8H9, there was no cytotoxic effect at 1000 ng/ml.

After completing studies with the single-chain IT, we prepared the disulfide-linked Fv molecule and tested it on the MCF-7 cell line. The IC₅₀ of 8H9(dsFv)-PE38 is 5 ng/ml, which is similar to the cytotoxic activity of the scFv molecule (Table 1).

To determine whether the cytotoxic activity of 8H9(scFv)-PE38 is specific and requires binding to the antigen recognized by MAb 8H9, we performed several control experiments. The results in Table 2 show that L428 and SP2/0 cells, which do not react with MAb 8H9, are not sensitive to 8H9(scFv)-PE38 (IC₅₀ > 1000 ng/ml). In addition, M1(dsFv)-PE38, an IT that targets CD25, the α subunit of the interleukin 2 receptor, is not cytotoxic to cell lines killed by 8H9(scFv)-PE38. More evidence of specificity is shown in Fig. 4. The cytotoxic activity of 8H9(scFv)-PE38 was competed by an excess amount of MAb 8H9, but not with MAb T6 that reacts with CD30 (20). Finally, MAb 8H9 alone was not cytotoxic. Thus, specific binding to the 8H9 antigen and the toxic activity of PE38 are necessary for the cytotoxic activity of 8H9(Fv)-PE38.

8H9(dsFv)-PE38 was evaluated for its nonspecific toxicity in mice. Groups of 5 or 10 mice received i.v. injection once with varying doses of IT and were observed for 2 weeks. Almost all of the deaths occurred within 72 h after treatment. The mortality data are shown in Table 3. The LD₅₀ of 8H9(dsFv)-PE38 is 0.78 mg/kg (95% confidential range, 0.66–0.93 mg/kg) calculated with the Trimmed Spearman-Karber Program version 1.5.

Because 8H9(dsFv)-PE38 is being considered for use in humans, we evaluated its toxicity in two cynomolgus monkeys because there is a similar reactivity of monkey and human tissues with 8H9 (10). One monkey received 0.1 mg/kg, QOD × 3, and the second received 0.2 mg/kg, QOD × 3. In this pilot study, both monkeys tolerated 8H9(dsFv)-PE38 well, with only mild laboratory abnormalities (Table 4). There was a slight decrease in albumin that was more pronounced in the high-dose monkey. The hepatic abnormalities were a borderline elevated alanine aminotransferase on days 3, 5, and 8 in the high-dose monkey, and the lactate dehydrogenase was borderline elevated on day 5 in the low-dose monkey. The major toxicity observed in both monkeys was loss of appetite. Thus, 8H9(dsFv)-PE38 could be administered safely to cynomolgus monkeys, and the high dose used (0.2 mg/kg) was higher than that needed to cause tumor regression of a human cancer xenograft in mice (0.15 mg/kg).

Serum levels of 8H9(dsFv)-PE38 were determined in each of the two monkeys 10 min after each of the three doses. The levels were determined by cytotoxicity assay so that only intact cytotoxic protein would be measured. As shown in Fig. 5, the levels of 8H9(dsFv)-PE38 were 5.0–5.4 μg/ml at 10 min after administration of 0.1 mg/kg and 11.0–13.0 μg/ml at 10 min after administration 0.2 mg/kg. These blood levels are 1000-fold higher than the IC₅₀ of the IT on MCF-7 cells in cell culture.

To determine the antitumor activity of 8H9(Fv)-PE38, several different doses of both the single-chain and the disulfide-linked Fv IT were administered to SCID mice bearing MCF-7 tumors or OHS-M1 tumors. The mice developed tumors of about 50 mm³ in size by day 4 and were treated on days 4, 6, and 8. Fig. 6,B shows tumor sizes in mice treated with 0, 0.075, or 0.15 mg/kg 8H9(scFv)-PE38. In both groups of mice treated with 8H9(scFv)-PE38, tumor regressions were observed, with the higher dose producing a larger effect. The control group received PBS/0.2% HSA. To determine whether antitumor activity was specific, we treated mice with a control IT that does not react with MCF-7 cells. We chose M1(dsFv)-PE38, an IT directed at CD25, the α subunit of the interleukin 2 receptor (16). Mice received injection with 0.15 mg/kg × 3 of M1(dsFv)-PE38. No responses were noted with this treatment (Fig. 6,A). M1(dsFv)-PE38 has previously been shown to produce complete regression of tumors expressing CD25 (16). To show that the effects of 8H9(scFv)-PE38 were reproducible, we carried out the antitumor experiments a total of three times and observed specific tumor regressions in all of the experiments (data not shown). In the second set of animal experiments, we evaluated the antitumor activity of 8H9(dsFv)-PE38 at 0.075 and 0.15 mg/kg × 3. Both doses were effective, producing statistically significant and prolonged tumor regressions (Fig. 6,C). In these experiments, 8H9(scFv)-PE38 and 8H9(dsFv)-PE38 showed similar antitumor activities at 0.15 mg/kg (Fig. 6 D).

We also investigated the effects of 8H9(scFv)-PE38 on osteosarcoma cells. The OHS-M1 cell line forms tumors in SCID mice. Mice received injection with 1.5 × 10⁶ cells on day 0. The mice developed tumors of about 50 mm³ in size by day 4 and were treated with IT i.v. on days 4, 6, and 8. Fig. 6, E and F, shows tumor sizes before and after treatment with 0.075 or 0.15 mg/kg 8H9(Fv)-PE38. Although treatment with 0.075 mg/kg 8H9(Fv)-PE38 had little effect, tumor regressions were observed using 0.15 mg/kg 8H9(Fv)-PE38. The average size of the tumors, which is indicated by the asterisk, was statistically different between the control group and the IT-injected group (P < 0.05) for MCF-7 cells and for OHS-M1 cells. In comparison with the MCF-7 breast cancer tumors, the osteosarcoma tumors are less responsive to 8H9(Fv)-PE38. This is consistent with the difference in IC₅₀ values observed in cell culture experiments (MCF-7 = 5 ng/ml, OHS-M1 = 20 ng/ml).

In the current study, we have prepared a single-chain and a disulfide-linked IT with the Fv portion of the 8H9 MAb and shown that both are specifically cytotoxic to cell lines reacting with the 8H9 antibody and produce substantial tumor regressions in mice at doses that do not produce significant animal toxicity (Fig. 6). The 8H9 antibody was chosen for IT development because it reacts with an antigen present on the cell surface of a variety of human cancers and does not appear to be expressed on the cell surface of normal tissues (10).

Several ITs containing PE38 as the toxic component have been evaluated in clinical trials (8, 9, 21, 22). The most striking responses have been with IT BL22 in drug-resistant hairy cell leukemia. More than 50% of the patients achieved a complete remission at dose levels ranging from 30 to 50 μg/kg with acceptable side effects (9). This IT targets the CD22 antigen that is expressed on B cells but not on other cell types. We have also studied an IT (Erb38) that targets the Her2/neu protein expressed on breast cancers (6). Erb38 was evaluated in a Phase I trial in patients with breast cancer, but it produced liver toxicity at very low doses (2 μg/kg) due to the presence of Her2/neu on human liver cells (21). This toxicity was not predicted in mouse studies. In fact, both BL22 and Erb38 had similar toxicities (LD₅₀ values) in mice whose cells do not react with either of these antibodies. In general, mouse toxicity studies have not been useful in predicting human toxicities, especially when the toxicities are due to killing of normal cells bearing the target antigen, as observed with Erb38. Monkey studies can potentially be useful in predicting toxicities, if the antibody reacts equally well with human and monkey tissues. In a previous study, weak nonspecific staining of the 8H9 MAb was noticed in the cytoplasm of normal human pancreas, stomach, liver, and adrenal cortex (10). Normal tissues from cynomolgus monkeys also demonstrated a similar reactivity, with nonspecific staining observed in stomach and liver (10). To evaluate possible liver toxicity or other toxicities due to this cytoplasmic staining observed in immunohistochemical studies, we did a preliminary toxicology study using two cynomolgus monkeys. The injection of 0.1 mg/kg of 8H9(dsFv)-PE38 did not produce any increase in the level of liver enzymes in the blood of these monkeys, and the injection of twice the dose produced only a small increase in liver enzymes. These data indicate that 8H9(dsFv)-PE38 has low toxicity for liver and is worthy of further clinical development. Because normal human brain tissue sections including frontal lobe, spinal cord, pons, and cerebellum are completely negative for staining in immunohistochemical studies, we may also be able to use 8H9(dsFv)-PE38 for intrathecal therapy of leptomeningeal carcinomatosis from a wide spectrum of human solid tumors.

The 8H9 ITs were tested against a panel of cell lines known to react with the 8H9 antibody, and many were killed by the IT. The most sensitive was the breast cancer cell line MCF-7, and this could be used for animal studies because it grows in immunodeficient mice. When tested against MCF-7 tumors, both ITs produced substantial tumor regressions when given at 0.15 mg/kg. It is interesting to point out that tumor regressions of the same magnitude were obtained with similar doses of IT BL22 using the CD22-positive CA46 cell line (4) and that BL22 was subsequently found to induce complete regressions in many patients with drug-resistant hairy cell leukemia (9). In leukemias, the cancer cells are readily accessible to ITs, whereas in solid tumors, particularly large tumor masses, entry into the tumor is impeded. An ideal setting for the use of ITs would be small-volume disease after tumor reduction by surgery and chemotherapy.

As shown previously with other RITs, the yield of the more stable disulfide-linked IT molecule was much higher than that of the single-chain molecule (16% compared with 1.7%; Table 1), and 8H9(dsFv)-PE38 will be used for further preclinical development. The highest activity of 8H9(dsFv)-PE38 was observed on the MCF-7 cell line, where the IC₅₀ is 5 ng/ml (0.8 × 10⁻¹⁰ m). One major factor determining the IC₅₀ is the affinity of the Fv for the target antigen. We have shown that it is possible to increase the affinity and activity of other ITs by 5–20-fold using site-directed mutagenesis to alter amino acids in the complementarity-determining regions of the Fvs (23, 24, 25). We plan to do such studies with 8H9(dsFv)-PE38 as the next step in preclinical development.

In summary, we have produced and evaluated two RITs, 8H9(scFv)-PE38 and 8H9(dsFv)-PE38, which have a specific cytotoxic activity against cell lines derived from breast cancer, osteosarcoma, and neuroblastoma. Both ITs showed specific antitumor activity using mouse xenograft models for human breast cancer and osteosarcoma. Also, cynomolgus monkeys tolerated the injection of the RIT 8H9(dsFv)-PE38 without laboratory abnormalities. Because 8H9(dsFv)-PE38 is stable, active, and can be produced in large amounts, it is a good candidate for further development for cancer therapy.

We thank Anna Mazzuca for expert editorial assistance and Maria Gallo for reading the manuscript.