The goal of cancer treatment is the elimination of tumor cells while inflicting the least amount of harm to normal cells. Because approximately half of all cancers are not cured using conventional therapies, new strategies are needed. One emerging therapeutic approach is the targeted delivery of highly toxic substances.
Certain bacterial and plant toxins [e.g., diphtheria toxin, ricin, and pseudomonas exotoxin (PE)] are among the most toxic substances known. Because these toxins bind to cell surface receptors that are expressed on many normal cell types, they would, unmodified, be useless as therapeutics. Therefore, to convert toxins into useful agents, their binding to normal tissues has to be eliminated and replaced with tumor-selective binding. The strategies for modifying toxin activity and engineering novel binding specificities are discussed. The combination of a modified toxin with a binding ligand is often called an immunotoxin or chimeric toxin and the term "recombinant" usually signifies that molecules have been expressed in E. coli.
II. IMMUNOTOXIN DESIGN AND TESTING
The genes for diphtheria toxin, ricin, and PE have been cloned and sequenced. In addition, structural data are available from protein crystals of each toxin. Structure-function studies have indicated which part of each gene encodes sequences of a particular function. Based on this knowledge, recombinant immunotoxins have been generated by deleting the DNA encoding the binding portion of the toxin and replacing it with an antibody, growth factor, or other ligand that will bind selectively to cancer cells. Once the immunotoxin is made, its usefulness is assessed in a series of preclinical tests. A favorable outcome leads to the initiation of clinical trials, which determine the properties of immunotoxins in cancer patients.
Conventional DNA cloning techniques are used to delete the sequences in toxin genes that are responsible for toxin binding. cDNA sequences encoding tumor-binding ligands are then inserted in place of these sequences. Immunotoxin sequences are cloned in an expression plasmid downstream of the T7 promoter. Plasmids are transformed into an expression host, typically E. coli stain BL21 (λDE3), developed by Studier. This strain produces T7 polymerase in response to induction of the lactose operon. BL21 (λDE3), transformed with the appropriate plasmid, is grown in Luria-Bertani broth with ampicillin (100 μg/ml). At an absorbance of 0.5-2.0 (600 nm), the culture is induced with isopropyl-β-D-thiogalactopyranoside to bring about high-level expression.
Recombinant immunotoxins are usually produced within the E. coli cell and are later recovered from inclusion bodies by denaturation and renaturation. The renatured material is then typically purified by Q Sepharose, Mono Q, and HPLC size-exclusion chromatography.
Purified immunotoxins are tested for cytotoxicity by measuring their ability to inhibit protein synthesis. Typically, immunotoxins in the range of 0.1-100 ng/ml are added to both target and nontarget cells. After an overnight incubation at 37°C, protein synthesis is measured by incubating cells for 1 hr with [3H]leucine. Immunotoxins that exhibit IC50 values below 10 ng/ml for target cells and above 1000 ng/ml for nontarget cells are considered potent and selective enough for testing in tumor models. Mice are injected with the appropriate tumor xenograft. After several days to allow the tumor to get established, immunotoxin treatments are given (immunotoxin injections on days 5, 7, and 9 are typical). Antitumor activity and toxicity to mice are assessed. A therapeutic window is established whereby the therapeutic dose is compared with the toxic dose (the LD50 is divided by the dose giving complete regressions of tumor). Immunotoxins with a good therapeutic window are then evaluated in monkeys for toxicity to normal primate tissue. On completion of these preclinical evaluations, compounds with an acceptable therapeutic profile go forward to phase I trials in patients harboring cancers that have failed to respond adequately to conventional treatment.
FIGURE 1 PE structure-function. PE is composed of three structural domains. Domain I encodes the binding domain. Domain II has translocating activity and contains the site of proteolytic processing. Cleavage is between arginine 279 and glycine 280. Domain III has ADP-ribosylating activity and a ER retention sequence at the C terminus.
III. PSEUDOMONAS EXOTOXIN: STRUCTURE-FUNCTION
Because our group has worked almost exclusively with PE, most working examples are drawn from that experience. PE, which is a three-domain protein (Fig. 1), binds and enters cells by receptor-mediated endocytosis (Fig. 2). Specifically, PE binds via its N-terminal domain (termed domain I) to the heavy chain of the low-density lipoprotein receptor-related protein (LRP). LRP carries the toxin to the endocytic compartment. Once there, the toxin is cleaved by a furin-like protease. Cleavage, which is necessary for toxicity, has an optimum of pH 5.5 and because of this it most likely occurs in the endosomal compartment.
The cleavage site is located in the middle domain (termed domain II) between arginine 279 and glycine 280 (see Figs. 1 and 2). Proteolytic cleavage followed by the reduction of the disulfide bond that joins cysteines 265 and 287 produces a N-terminal fragment of 28 kDa and a C-terminal fragment of 37 kDa (Fig. 2). The 28-kDa fragment is composed of all of domain I and a small portion of domain II. The 37-kDa fragment is composed of most of domain II and all of domain III. Besides serving as the site of cleavage, domain II has sequences necessary to translocate domain III to the cell cytosol. Domain III, which is located at the C terminus, has ADP-ribosylating activity and it is this activity that mediates cell killing. Specifically, once in the cell cytosol, domain III ADP ribosylates elongation factor 2 and shuts down the synthesis of new cellular protein (Fig. 2). Domain III also has an endoplasmic reticulum retention sequence that is located at the very C terminus of PE and is composed of the following five amino acids: REDLK. The terminal lysine is probably removed by carboxy peptidases, leaving REDL as the last four amino acids of PE. This sequence closely resembles KDEL, which is the authentic ER retention sequence. An ER retention sequence is necessary for toxin-mediated inhibition of protein synthesis, presumably because it directs PE to the ER where it can translocate to the cytosol (Fig. 2).
FIGURE 2 Pathway for PE as it enters the cell cytosol of mammalian cells. Binding is followed by internalization to the endosomal compartment. Cleavage by a furin-like enzyme is essential for toxicity. A C-terminal fragment of 37 kDa is generated, which translocates from the ER to the cell cytosol where it inactivates protein synthesis by ADP-ribosylating elongation factor 2.
IV. CONSTRUCTION OF RECOMBINANT IMMUNOTOXINS
Because it binds LRP, which is present on the surface of most cells and tissues, native PE exhibits no selective cytotoxic activity for tumor cells. However, selective binding can be engineered by eliminating toxin binding to LRP and redirecting domains II and III to the surface of tumor cells. Routinely, the DNA encoding domain I is deleted and replaced with cDNAs encoding binding ligands such as tumor growth factor α (TGFα), interleukin (IL)-2, IL-4, IL- 6, and single chain antibodies that bind surface determinants on various human malignancies (Fig. 3).
Ligand-toxin and single chain antibody-toxin fusion proteins are expressed in E. coli, purified, and tested for cytotoxic activity on appropriate target lines. A similar strategy can be used to construct ligand-toxin fusion proteins with diphtheria toxin (DT). For DT, the sequences encoding toxin binding are located at the 3' end of the structural gene of the toxin. Many of the same ligands that have been fused to the 5' end of PE have also been fused to the 3' end of DT.
FIGURE 3 Construction of fusion proteins with domains II and III of PE. The binding domain of PE is removed and replaced with binding ligands to direct the remainder of the toxin to cancer cells expressing particular antigens or receptors. Usually, the targeting ligand is placed at the N terminus of the PE fusion protein. The single asterisk represents ligands that include growth factors and cytokines. The single chain Fv of a monoclonal Ab is shown by the double asterisk. The variable portion of light and heavy chains is joined by a flexible peptide linker.
Most ligand-toxin fusions exhibit potent toxicity for target cell lines and little or no cytotoxicity for nontarget lines. The activity of TGFαPE38 (PE38 is a truncated version of PE that is composed of domains II and III) for a variety of target lines is provided in Table I. TGFα binds to and is internalized by the epidermal growth factor (EGF) receptor. TGFαPE38 is not toxic for CHO cells, which express no EGF receptors on their cell surface. Like native PE, TGFαPE38 is cleaved within cells by a furin-like protease (data not shown). The mutation of arginine 279 to glycine makes this chimeric toxin refractory to cleavage and nontoxic for cells (Table I). Similar kinds of results were obtained when key basic residues at the DT cleavage site were mutated to nonbasic amino acids.
TABLE I Cytotoxic Activity of TGFα-Toxins for Various Cell LinesCell line Type TGFαPE38
IC50 (ng/ml) TGFαPE38 Gly279
IC50 (ng/ml) MCF7 Breast 1.1 >1000 HT29 Colon 2.4 ~2000 KB Epidermoid 0.1 310 A431α Epidermoid 0.28 10 CHO Ovary >1000 >1000
αA431 IC50 (PE) = 5 ng/ml. IC50 (PEGly279) > 1000 ng/ml.
Because recombinant immunotoxins are made by gene fusion technology, certain constraints are inherent in their construction. A ligand fused with domains II and III of PE has to be placed at the N ter- minus of the construction. This means that the C terminus of the ligand is joined via a peptide bond with the N terminus of domain II. If the ligand requires a free C terminus to mediate receptor binding, this kind of construction is likely to exhibit diminished binding activity. Another problem that can arise stems from the need for toxin cleavage at the Arg279-Gly280 bond. If the ligand does not transport domains II and III of PE into a furin-containing compartment, there will be little or no cleavage, the 37-kDa fragment will not be generated, and there will be no toxicity. Similarly, if cells fail to express sufficient furin, there will be no cleavage and no toxicity. Single chain antibodies are composed of the variable light and variable heavy chains of a monoclonal antibody. Routinely, to hold these two chains together, a 15 amino acid peptide is used to tether the C terminus of one chain with the N terminus of the second chain. While this approach facilitates the expression of the recombinant antibody from a single transcript, the product is not always stable and binding activity may be reduced.
FIGURE 4 Possible ligand placements. The binding ligand (e.g., TGFα) is placed routinely at the N terminus of domains II and III. It can, however, be placed near the C terminus. This only produces an active cytotoxic agent if the ER retention sequence is present as the last four amino acids. The need for furin-mediated cleavage can be bypassed by initiating translation of domain II at residue 280 (glycine is replaced by methionine). (A) TGFα is placed at the N terminus and cell-mediated cleavage is required. (B) TGFα is placed near the C terminus and cleavage is required. (C) TGFα is placed near the C terminus with no cleavage necessary.
To address each of the three problems mentioned earlier, novel engineering strategies had to be developed. The placement of the binding ligand in the immunotoxin construction can be modified in one of two ways. The ligand can be moved from its N-terminal location and inserted near the C terminus of domain III of PE (Figs. 4A and 4B). Unfortunately, the KDEL sequence must still be placed C-terminal to the ligand, but at least there are fewer structural constraints than when the ligand is placed in front of domain II. This kind of construct has been made with TGFα (Fig. 4B). Another strategy is used to deal with molecules such as human IL-4, which retain very little binding activity when placed N-terminal to domain II (Fig. 5). Because the N and C termini of human IL-4 are close together, it was possible to link them with a flexible peptide linker. Then new termini were generated artificially by "opening up" the molecule at one of two different loops that are not involved in ligand binding (Fig. 5). This strategy generates circularly permuted molecules. Circularly permuted IL-4 beginning at either residue 38 or residue 105 and fused to domains II and III of PE exhibited much better binding activity than the same construct with the linear version of IL-4 (Fig. 5).
The need for furin-mediated proteolysis can be overcome by initiating translation of the recombinant immunotoxin at residue 280 and placing the ligand near the end of domain III (Fig. 4C). The example given is of TGFα. When assayed on a number of different cell lines, this construct proved more active than the conventional chimeric toxin shown in Fig. 4A. From this we concluded that furin cleavage can be rate limiting and that strategies to bypass this step may prove useful in the design of more potent molecules.
Finally, unstable single chain antibodies can be stabilized by the introduction of novel disulfide bonds into the framework segments of the variable chains (Fig. 6). Residues that are opposed and separated by the appropriate distance were modified to create novel cysteine residues, one each in the light and heavy chains. The light chain with a free sulfhydryl could then be linked by a disulfide bond with a heavy chain construct to form a disulfide-stabilized Fv immunotoxin. Such an approach has been used successfully to generate several very stable recombinant immunotoxins.
Experiments in nude mice have been performed to test the antitumor activity of many PE-derived recombinant immunotoxins. For example, TGFα-toxin and B3Fv-toxin were both shown to inhibit the growth of human A431 tumors. (B3Fc is derived from the B3 monoclonal antibody, which binds to carbohydrate antigens displayed on the surface of many human adenocarcinomas.) TGFα-toxin caused a profound reduction in the rate of tumor growth but did not produce complete regressions. B3Fv-toxin had a larger therapeutic window and, at dose levels of 0.75 mg/kg × 3, caused complete regression of tumors and produced long-lasting cures. Variants of both of these immunotoxins are currently in clinical trials.
It is possible to design and produce wholly recombinant immunotoxins composed of truncated toxin molecules joined with tumor-targeting agents that exhibit cytotoxic activity for target cell lines and antitumor activity in model systems. Results of ongoing clinical trials will determine the utility of these agents.
Robert J. Kreitman
National Cancer Institute, Bethesda, Maryland
ANTI-VASCULAR ENDOTHELIAL GROWTH FACTOR-BASED ANGIOSTATICS ; HEMATOPOIETIC GROWTH FACTORS ; INSULINLIKE GROWTH FACTORS ; TARGETED TOXINS
endoplasmic reticulum (ER) retention sequence Specific amino acids located at the C termini of proteins that retain them in, and/or retrieve them to, the ER.
Fv fragment The portion of a monoclonal activity that is composed of the variable domain of the heavy chain linked with the variable domain of the light chain.
immunotoxin An antibody-toxin or a ligand-toxin hybrid protein.
ligand Either a growth factor or a cytokine that binds to a cell surface receptor.
recombinant Signifying that molecules have been expressed in Escherichia coli.
translocation Toxins have the unusual property of being able to move from one side of a cell membrane to the other side. This movement is called translocation.
Brinkman, U., Pai, L. H., FitzGerald, D. J., Willingham, M. C., and Pastan, I. (1991). B3(Fv)-PE38KDEL, a single chain immunotoxin that causes complete regression of a human carcinoma in mice. Proc. Natl. Acad. Sci. USA 88, 8616-8620.
Chaudhary, V. K., Jinno, Y., FitzGerald, D., and Pastan, I. (1990). Pseudomonas exotoxin contains a specific sequence at the carboxyl terminus that is required for cytotoxicity. Proc. Natl. Acad. Sci. USA 87(1), 308-312.
Chiron, M. F., Fryling, C. M., and FitzGerald, D. J. (1994). Cleavage of pseudomonas exotoxin and diphtheria toxin by a furin-like protease prepared from beef liver. J. Biol. Chem. 269, 18167-18176.
Jung, S. H., Pastan, I., and Lee, B. (1994). Design of interchain disulfide bonds in the framework region of the Fv fragment of the monoclonal antibody B3. Proteins 19(1), 35-47.
Kounnas, M. Z., Morris, R. E., Thompson, M. R., FitzGerald, D. J., Strickland, D. K., and Saelinger, C. B. (1992). The alpha 2-macroglobulin receptor/low density lipoprotein receptor-related protein binds and internalizes pseudomonas exotoxin A. J. Biol. Chem. 267, 12420-12423.
Kreitman, R. J., Puri, R. K., and Pastan, I. (1994). A circularly permuted recombinant interleukin 4 toxin with increased activity. Proc. Natl. Acad. Sci. USA 91(15), 6889-6893.
Murphy, J. R. (1988). Diphtheria-related peptide hormone gene fusions: A molecular genetic approach to chimeric toxin development. Cancer Treat. Res. 37(123), 123-140.
Ogata, M., Fryling, C. M., Pastan, I., and FitzGerald, D. J. (1992). Cell-mediated cleavage of pseudomonas exotoxin between Arg279 and Gly280 generates the enzymatically active fragment which translocates to the cytosol. J. Biol. Chem. 267(35), 25396-25401.
Pai, L. H., Gallo, M. G., FitzGerald, D. J., and Pastan, I. (1991). Anti-tumor activity of a transforming growth factor α-Pseudomonas exotoxin fusion protein (TGFα-PE40). Cancer Res. 51, 2808-2812.
Pai, L. H., and Pastan, I. (1993). Immunotoxin therapy for cancer. JAMA 269(1), 78-81.
Pastan, I., Chaudhary, V., and FitzGerald, D. J. (1992). Recombinant toxins as novel therapeutic agents. Annu. Rev. Biochem. 61(331), 331-354.
Reiter, Y., Brinkmann, U., Jung, S. H., Lee, B., Kasprzyk, P. G., King, C. R., and Pastan, I. (1994). Improved binding and antitumor activity of a recombinant anti-erbB2 immunotoxin by disulfide stabilization of the Fv fragment. J. Biol. Chem. 269(28), 18327-18331.
Reiter, Y., Brinkmann, U., Kreitman, R. J., Jung, S. H., Lee, B., and Pastan, I. (1994). Stabilization of the Fv fragments in recombinant immunotoxins by disulfide bonds engineered into conserved framework regions. Biochemistry 33(18), 5451-5459.
Strom, T. B., Anderson, P. L., Rubin, K. V., Williams, D. P., Kiyokawa, T., and Murphy, J. R. (1990). Immunotoxins and cytokine toxin fusion proteins. Semin. Immunol. 2(6), 467-479.
Studier, F. W., and Moffatt, B. A. (1986). Use of bacteriophage T7 polymerase to direct selective expression of cloned gene. J. Mol. Biol. 189, 113-130.
Theuer, C. P., FitzGerald, D., and Pastan, I. (1992). A recombinant form of pseudomonas exotoxin directed at the epidermal growth factor receptor that is cytotoxic without requiring proteolytic processing. J. Biol. Chem. 267, 16872-16877.
Theuer, C. P., FitzGerald, D. J., and Pastan, I. (1993). A recombinant form of Pseudomonas exotoxin A containing transforming growth factor alpha near its carboxyl terminus for the treatment of bladder cancer. J. Urol. 149(6), 1626-1632.
Williams, D. P., Wen, Z., Watson, R. S., Boyd, J., Strom, T. B., and Murphy, J. R. (1990). Cellular processing of the interleukin- 2 fusion toxin DAB486-IL-2 and efficient delivery of diphtheria fragment A to the cytosol of target cells requires Arg194. J. Biol. Chem. 265(33), 20673-20677.