Molecular Advances in Pretargeting Radioimunotherapy with Bispecific Antibodies ¹

Targeted delivery of radionuclides for imaging and therapy of cancer, using antibodies against tumor-associated antigens or haptens, is an active area of ongoing investigation at both the experimental and the clinical levels (1). Since the introduction of the terms RAID³ and RAIT, representing the use of radiolabeled antibodies for imaging and therapy, respectively, >20 years ago, the extensive efforts have resulted in the commercialization of several cancer-imaging agents based on RAID. The current prospects of establishing RAIT as a new therapeutic modality of cancer is also very promising because of the encouraging results obtained with radiolabeled antibodies in the management of hematopoietic neoplasms, especially non-Hodgkin’s lymphoma (2).

In contrast to hematopoietic malignancies, solid tumors have been less responsive to RAIT. Because the efficacy of RAIT depends on several factors, including cumulative radiation dose delivered, dose rate, penetration, and tumor radiosensitivity, one major reason for the diminished success of RAIT in solid tumors may be the suboptimal tumor uptake of the targeting antibody. At an accretion of 0.001–0.01% ID of the radiolabeled antibody/gram tumor, a cumulative tumor dose of <1500 cGy is usually delivered, which falls short of the typical >5000 cGy needed to achieve therapeutic responses in most neoplasms, based on external beam irradiation of adenocarcinomas (3). When the accretion is limited, doses needed to achieve therapeutic response are restricted because of other normal organ dose limitations. For RAIT, bone marrow toxicity often determines the maximum tolerated dose that can be administered.

The challenge of treating solid tumors by RAIT has stimulated a number of approaches to improve the radiation dose delivered and to achieve a more uniform distribution of ionizing radiation, with the ultimate goal being the delivery of tumoricidal doses while sparing normal tissues. One of these methods is pretargeting, a strategy conceived in the 1980s initially for improving the selective delivery of radionuclides by antibodies to tumors and later extended to include nonradioactive agents and nonantibody delivery systems. This minireview addresses the various refinements of the pretargeting strategy, with emphasis on those involving the use of bsAb and radioactive agents in the therapy of cancer. The progress of pretargeting research in the realm of nuclear medicine has received periodic appraisals over the years through excellent editorials (4–8) and reviews (9–15).

The inherent incompatibility between a tumor-targeting antibody and the radiation carried by it became evident from early imaging studies in experimental animals and in patients. Intact immunoglobulins directly labeled with a radioisotope were found to clear very slowly from the circulation, resulting in high background activity, thereby limiting the efficacy for both the detection and treatment of tumors (4). The use of antibody fragments, such as Fab or F(ab‘)₂, which clear from the blood more rapidly than whole IgG, improves target:background ratios but at the expense of lower uptake of the radiolabeled antibody in the tumor, when compared with whole IgG (16). Other approaches to reduce background radioactivity while maintaining tumor uptake of a directly radiolabeled antibody agent have been explored and met with variable degrees of success, e.g., administration of antiantibodies (such as an anti-idiotype antibody) can remove radiolabeled antibodies from blood faster than from tumor and is considered particularly applicable to radioiodinated antibodies (17). For antibodies labeled with a radiometal, the insertion of a metabolizable linker between the chelate and protein is also effective in improving tumor:background ratios, despite a modest drop in tumor concentrations (18). The fact that an antibody can be facilely labeled in vivo with a radioactive hapten to which it binds with a high affinity is fundamental to the development of pretargeting strategies, which successfully overcome the problem of low target:background ratios by temporal separation of the slow antibody-targeting step from the delivery of the radionuclide.

The essence of pretargeting involves the use of a macromolecule that is capable of binding with a high affinity to a radioactive agent of low molecular weight (usually M_r <10,000) and can also selectively target a tumor antigen. The macromolecule is administered first, and the radioactive agent (“the effector”) is given at a later time, ideally when the concentration of the macromolecule in the tumor is greater than in other tissues. Favorable tumor:normal tissue ratios of radioactivity are thus achievable, because the small size of the nontargeted radioactive agent permits its rapid elimination from the body. This sequence of events is commonly referred to as the two-step protocol (Fig. 1). The key for the success of the two-step method is that the macromolecule must be cleared sufficiently from the blood and normal tissues, otherwise the radioactive effector molecule will be retained wherever the macromolecule is distributed. An alternative to the two-step protocol is the three-step protocol, in which a specific molecule (“the chaser”) is injected to remove the residual macromolecule from the bloodstream before giving the radioactive agent, so that further enhancement in the tumor:blood ratios of radioactivity can be obtained. To date, the potential value of applying pretargeting strategies to cancer imaging and therapy has been demonstrated in animal models, as well as in clinical trials for many macromolecule/effector systems, with or without the chase step, using radioactive or nonradioactive effectors (12, 14, 19).

To be useful for pretargeting applications, the macromolecule can be just an antihapten monoclonal antibody that recognizes a structural component (hapten) of an effector (20) or even merely a biotin-binding protein like streptavidin (21–23). These molecules can be used because tumors or inflammatory/infectious foci, because of their leaky capillaries, can accumulate macromolecules nonspecifically through passive diffusion, leading to a preferential localization of the macromolecule at the target. In most cases studied, however, the macromolecule of interest consists of a modified antitumor antibody carrying a secondary recognition moiety, in addition to its primary target-binding function. Representative examples of such modified antibodies can be distinguished into five types: (a) bsAb (24) constructed to contain one hapten-binding site and one or two target-binding sites; (b) antibodies conjugated to streptavidin (25) or avidin (26) to enable the binding of biotin; (c) biotinylated antibodies capable of complexing with avidin (27) or streptavidin (28, 29); (d) antibodies conjugated to DNA (30) to promote binding to complementary nucleotide sequences; and (e) antibodies conjugated to enzymes (19) to activate a prodrug at the target site. In general, for each targeting macromolecule selected, a series of effectors may be devised and evaluated under pretargeting conditions for their respective uptake in tumor relative to normal tissues. These effectors, while having the common hapten or structural component for binding to the macromolecule, could differ in other attributes, such as hapten valency (31, 32), lipophilicity (33), dissociation rate (34), in vivo stability (35), binding affinity (36), and optical property (37), each of which may affect the distribution of the effectors in target as well as in normal organs.

A particular attribute that has proved to be influential on the tumor uptake of an effector is hapten valency. Effectors that contain a bivalent hapten are better tumor localizers than their monovalent analogs. The enhancement is believed to be because of the ability of the bivalent hapten to cross-link the pretargeted macromolecule at the tumor site, resulting in the formation of a more stable complex (Fig. 2) and, therefore, a longer tumor residence time. Bivalent haptens pretargeted with bsAb have been patented as the AES (38). As shown in the autoradiographs of Fig. 3, the clear tumor-to-nontumor contrast provided by the AES pretargeting method in an animal model with human small cell lung carcinoma expressing NCAM with a bispecific anti-NCAM × anti-HSG conjugate (NK1NBl1 × 679, Fab‘ × Fab‘) and a radioiodinated (¹²⁵I) di-HSG hapten is striking, compared with that obtained from a directly radiolabeled anti-NCAM antibody (39). Excellent images were also obtained (40) in nude mice bearing human renal cell carcinoma xenografts, by pretargeting with a bispecific antirenal cell carcinoma × anti-indium-DTPA conjugate (G250 × 734, Fab‘ × Fab‘) for 72 h followed by ¹¹¹In-labeled di-DTPA hapten (Fig. 4). Clinical trials (41) using the AES pretargeting method with a bispecific anti-CEA × anti-indium-DTPA conjugate (hMN14 × 734, Fab‘ × Fab‘) and a radioiodinated (¹³¹I) di-indium-DTPA hapten (Pentacea; IBC Pharmaceuticals, Morris Plains, NJ) also produced impressive images of tumor uptake with high target-to-background contrast, as shown in Figs. 5 and 6 for two patients with CEA-expressing cancers.

The different pretargeting systems that have been investigated are grouped into nine categories in Table 1 according to the constituent of the targeting macromolecule. Within each assigned category, they may be further differentiated with respect to the following parameters: molecular size of the macromolecule, molecular form of the antibody component comprising the macromolecule, method of preparing the macromolecule or source of the macromolecule, valency of the macromolecule to the target, valency of the macromolecule to the hapten or biotin, hapten valency of the effector to the macromolecule, and other functional groups carried by the effector. The search for an effective pretargeting system worthy of commercialization necessitates continuous optimization of several or all of these parameters.

Before the term “pretargeted immunoscintigraphy” first appeared in publication (44) in 1987, the feasibility of using a bsAb or an antihapten antibody to increase the tumor concentration of diagnostic imaging agents had been assessed in several seminal studies with encouraging results (45–47). These earlier achievements were facilitated by the availability of a murine monoclonal antibody, CHA255, having a high binding affinity (Kd = 10⁻⁹ to 10⁻¹⁰ m) for indium benzyl EDTA (48) and a synthetic bsAb, ZCE × CHA (49), prepared by linking the Fab‘ of an anti-CEA murine monoclonal antibody (ZCE025) with the Fab‘ of CHA 255. Coadministration of ¹¹¹In benzyl EDTA or derivatives of ¹¹¹In benzyl EDTA containing cobalt-bleomycin (BLEDTA II or BLEDTA IV) with a sufficient amount of CHA255 was found to alter profoundly the pharmacokinetics of these small molecules in tumor-bearing mice (45). Specifically, with CHA255 acting as an in vivo carrier, the biological half-life of these small molecules was increased from minutes to days, and the concentrations of radioactivity in all organs, most notably tumor and blood, were markedly higher, when measured at 24 h. Furthermore, the injection of a “flushing” dose of unlabeled indium benzyl EDTA was effective at reducing more radioactivity in blood than in tumor, resulting in as much as a 50-fold increase in the tumor:blood ratio (from 0.8 to 47.1 for ¹¹¹In-benzyl EDTA) within 3 h. Comparable improvements (46, 47) in tumor:background ratios were observed in tumor-bearing mice for ZCE × CHA, using ¹¹¹In benzyl EDTA or ¹¹¹In BLEDTA IV premixed with a finite amount of ZCE × CHA as a carrier to ensure a higher tumor uptake. In an optimized protocol in which 14 μg of the synthetic bispecific Fab‘ × Fab‘ conjugate were allowed to prelocalize for 24 h before injecting 10μCi of ¹¹¹In BLEDTA IV premixed with 3 μg of ZCE × CHA, the concentration of radioactivity in tumor at 24 h was 7% ID/gram (50), with the tumor:blood ratio being ∼5. Presumably, transferring ¹¹¹In BLEDTA IV from circulating ZCE × CHA to prelocalized ZCE × CHA at the tumor had occurred efficiently to account for the observed in vivo targeting. Additional biodistribution studies (33, 50) undertaken to compare BLEDTA IV with a series of benzyl EDTA analogs concluded that ¹¹¹In hydroxyethylthiourea-benzyl-EDTA, while showing similar uptake in normal tissues as ¹¹¹In BLEDTA IV, had a remarkable increase in tumor concentration, topping 18% ID/gram at 24 h (50).

As the general principles of pretargeting were being demonstrated, a number of key factors that could affect the tumor-targeting potential of a hapten-antibody delivery system was also elucidated. Subsequent efforts to address these factors to the advantage of pretargeting applications have led to various refinements of the original approach exemplified by ZCE × CHA and ¹¹¹In BLEDTA IV (42).

A higher affinity of hapten for antibody is important to pretargeting, because it would enhance the noncovalent association between antibody and hapten, thus reducing the amount of antibody required for capturing hapten at the target site. Because binding constants of antibodies rarely exceed 10⁻¹⁰ m, additional improvements coming from binding constants would have to rely on other existing noncovalent systems that display affinities much higher than 10⁻¹⁰ m. Accordingly, the very high affinity of biotin for avidin or streptavidin (K_d = 10⁻¹⁵ m, about six to seven orders of magnitude above the binding constants usually achievable with antibodies) was distinguished immediately, and the suitability of biotin-avidin or biotin-streptavidin for pretargeting applications was quickly established (26). Pretargeting systems based on the binding of biotin to avidin or streptavidin have been developed, and some tested extensively in the clinic (27, 51, 52). The simplest system involves only streptavidin and a labeled biotin-chelate conjugate. Pretargeting with streptavidin followed by ¹¹¹In labeled EDTA-hydrazino-biotin was shown to be useful for imaging infectious lesions in a mouse model (21). Other systems are more elaborate and may consist of one of the following: (a) pretargeting with a streptavidin-antibody conjugate followed by a radiolabeled biotin-chelate (25); (b) pretargeting with a streptavidin-antibody conjugate followed by a clearing agent and then a radiolabeled biotin-chelate (51–55); (c) pretargeting with an avidin-antibody conjugate followed by a radiolabeled biotin-chelate (26); (d) pretargeting with a biotin-antibody conjugate followed by a radiolabeled streptavidin (29); (e) pretargeting with a biotin-antibody conjugate followed by avidin and then a radiolabeled biotin-chelate (56); and (f) pretargeting with a biotin-antibody conjugate followed by avidin, then streptavidin, and finally a radiolabeled biotin-chelate (57). Thus, procedures can consist of two, three, or more steps, all intended to increase tumor:nontumor ratios. However, both endogenous biotin and the immunogenicity of avidin and streptavidin can affect the results (54, 58). In addition to (strep)avidin–biotin, other recognition systems conferring high specificity and affinity, either based on complementary oligonucleotide binding (30) or enzyme-substrate interaction (19), have also been adapted for pretargeting applications. The preference of higher affinity also applies to the binding of antibody to target antigen. For a bsAb (monovalent Fab‘ × monovalent Fab‘) with a K_d for target antigen already in the 10⁻⁹ to 10⁻¹⁰ m range, additional improvements may still be possible by incurring bivalent binding with a construct like F(ab‘)₂ × Fab‘ or an engineered bispecific protein preserving the bivalency for the target antigen.

Compared with their monovalent analogs, bivalent haptens have consistently shown higher tumor uptake, a phenomenon attributed to cross-linking antibodies that are specifically bound to tumor surface or nonspecifically accumulated in tumor. Bivalent haptens are used today in almost all pretargeting systems, including those involving biotin/(strept)avidin. The binding between a DNA antibody and its antisense oligonucleotides can also be regarded as multivalent. A variant of bivalent hapten is the so-called asymmetric bivalent hapten (59), which contains two different haptens on the same molecule, with each hapten suitable for binding to a different antihapten antibody. Asymmetric bivalent haptens were designed to increase the targeting specificity when used in conjunction with two distinct bsAb, which differ from each other in antigen, as well as hapten specificity (60). This triple-component system for pretargeting would ensure that only cells expressing double antigens could have accrued both bsAb, which in turn could be cross-linked only by a relevant asymmetric hapten. The advantage of bivalency also applies to the binding of antibodies to target antigen. A modified antibody capable of binding bivalently to target antigen is likely to show an increase in target residence time, which may be further extended by using a relevant bivalent hapten.

The amount of macromolecule available for capturing hapten at the target site is usually very small (in the order of nmol/gram tumor in human subjects). Consequently, the amount of hapten deliverable to the target is quite limited. With a hapten labeled at a higher specific activity, increased radiation dose can be imparted per unit mass to effect more damage or easier detection of the target, thereby improving pretargeting efficacy. Specific activities not <1 mCi/nmol was thought necessary for good pretargeting results (6). In clinical studies, biotin-chelates or hapten-chelates labeled to specific activity ranging from 0.05 to 1.5 mCi/nmol have been acceptable.

Two factors pursuant to specific activity are choice of radionuclide and chelate. Radionuclides that have been evaluated for pretargeting applications include ¹¹¹In and ^99mTc for single photon imaging (20–22, 25–27, 31, 32, 42, 50); ⁶⁸Ga for PET imaging (37, 61, 62); ¹³¹I, ⁹⁰Y, and ¹⁸⁸Re for β-particle therapy (36, 51–53, 63–68); and ²¹²Bi for α-particle therapy (13). Table 2 lists several physical features (half-life, energy of emission, type of emission, maximum particle range, and production mode) of radionuclides that are of current interest for RAIT. Additionally listed in Table 2 are characteristic chelating agents capable of forming stable in vivo complexes with each radiometal. For α-emitters, more work may be needed to develop better strategies or chelates for binding them in a stable manner. The subjects of therapeutic radionuclides and choices of their chelates have been reviewed elsewhere (69, 70).

Preferably, a therapeutic radionuclide selected for use with a particular pretargeting system should possess the following properties: (a) a half-life similar to that of the radiolabeled hapten-chelate at the target site; (b) the ability to discharge particles of higher energy for effective cell killing; (c) the availability of a suitable hapten-chelate for efficient labeling and the resultant formation of a stable in vivo complex that exhibits rapid diffusion, little uptake in nontarget tissues, and fast clearance; and (d) the feasibility of production with a generator after easy elution. The most prevalent radionuclide used for therapy in clinical trials today is still ¹³¹I, despite its relatively long half-life (8 days), suboptimum energy of β-particle, and concomitant emission of γ rays, which requires isolation of patients in shielded rooms for several days but provides the benefit of imaging. Compared with ¹³¹I, ⁹⁰Y is almost ideal for therapy with pretargeting approaches, because: (a) it has a half-life (67 h) similar to that of the radiolabeled effector after prelocalization of the targeting macromolecule; (b) it discharges only electrons of higher energy to cause deeper tissue penetration; (c) it can be labeled efficiently with DOTA-bearing haptens, forming a stable in vivo complex that exhibits satisfactory pharmacokinetics; and (d) it can be produced by a generator and available commercially with acceptable purity. Other therapeutic radionuclides with one or more favorable physical properties than ¹³¹I are ¹⁸⁸Re and several α-emitters under active development. Like ⁹⁰Y, ¹⁸⁸Re is available carrier free and, except for a shorter half-life (17 h), has three of the four attributes characteristic of ⁹⁰Y as a promising therapeutic radionuclide. It is noted that short-lived radionuclides (t_1/2 about hours), α-emitters in particular, are necessarily restricted to rapid targeting processes, which are feasible with pretargeting approaches because the radiolabel resides on a fast-diffusing, small molecule. Radioisotopes emitting α-particles offer several advantages over β-emitters for therapy: (a) discharged α-particles with a mass 8,000 times greater than β-particles exhibit high linear energy transfer (100 KeV/μm) and are therefore very cytotoxic, requiring as few as 6 or 7 disintegrations for internalized α-emitters and ∼25 disintegrations for surface-bound α-emitters to kill a cell; (b) the extremely short path length (40–80 μm) of α-particles greatly reduces the nonspecific irradiation of normal tissue around the target cell; (c) the DNA damage caused by α-particles is not easily repaired by the cell; and (d) the cytotoxicity of α-particles is not affected by oxygen, and, therefore, cell killing is effective under hypoxic conditions usually existing in established tumors. However, the lack of suitable chemistry or effective chelates for preparing stably attached α-emitters has hampered their further development as agents for pretargeting applications. Table 3 compares the dosimetry data obtained from animal studies with ¹³¹I, ⁹⁰Y, and ¹⁸⁸Re under specified pretargeting conditions, as reported in a limited number of publications (39, 64, 66–68). In each case, the tumor was shown to receive a radiation dose considerably higher than any of the normal organs.

In addition to those discussed above, numerous other factors have also been considered as important for improving pretargeting efficacy, e.g., issues like molecular size, immunogenicity, target specificity, antigen density, and antigen modulation should be carefully considered in the design and construction of ultimate targeting macromolecules. As for the hapten-chelates or biotin-chelates, molecular size, immunogenicity, clearance rate, elimination route, and in vivo stability deserve proper attention. Pretargeting systems using a chase step may further face optimization issues involving the chase molecule, such as molecular size, valency, immunogenicity, clearance rate (71), and elimination route.

The success of receptor targeting by peptides, as demonstrated by the regulatory approvals of peptide imaging agents, such as ¹¹¹In-DTPA-pentatreotide for human neuroendocrine tumors, has prompted the adaptation of peptides for pretargeting approaches. One possible system described by Goodwin and Meares (5) involves the conjugation of receptor-specific peptides to long-circulating PEG, which is also derived with an avidin or streptavidin for the recognition by a radiolabeled DOTA-biotin dimer. By using PEG with branched structures, multiple peptides and other optional agents can be attached, along with avidin or streptavidin, thereby yielding a targeting macromolecule with many of the desirable features, including multivalency for the target (through multiple peptide), multivalency for the effector (through multiple binding sites on avidin/streptavidin for biotin), reduced immunogenicity (via PEG), prolonged circulation to reach high target concentration (via PEG of suitable molecular size), and the option of rapid clearance with the use of a biotin-PEG conjugate, which should be nonimmunogenic.

The trend at using smaller antibody fragments for in vivo applications has spurred the development of bispecific diabody or bispecific tetrabodies suitable for pretargeting approaches. We have produced two bispecific diabodies (BS1.5 and BS1.5H) for pretargeted delivery of radiolabeled bivalent haptens to tumors expressing CEA (43). BS1.5 (M_r ∼54,000) consists of two heterologous polypeptide chains associated noncovalently to form one binding site for CEA from the variable domains of hMN14 (a humanized anti-CEA antibody from Immunomedics, Inc., Morris Plains, NJ) and one binding site for HSG from the variable domains of 679 (a murine monoclonal antibody specific for HSG). In BS1.5H, humanized 679 V_H and V_K domains were used. The V domains were engineered into a dicistronic Escherichia coli expression vector for directing the synthesis of the two polypeptides, 679V_H-GGGGS-hMN14V_K−6His and hMN14V_H-GGGGS-679V_K−6His, in the periplasmic space of E. coli. The diabody was purified from the soluble fraction of isopropyl-1-thio-β-d-galactopyranoside-induced E. coli culture by Immobilized Metal Ion Affinity Chromatography and ion exchange chromatography. Kinetic analyses on BIAcore using HSG-immobilized sensor chips showed that the binding affinity for HSG was similar for BS1.5, BS1.5H, and chemically linked hMN14 × 679 (M_r ∼100,000, Fab‘ × Fab‘). The bispecificity of BS1.5 or BS1.5H was demonstrated on BIAcore by measuring the additional increase in response units on successive injections of the bispecific diabody followed by CEA. The utility of BS1.5 for tumor pretargeting was evaluated in CEA human tumor-bearing mice using a bivalent HSG hapten (IMP-241) labeled with ¹¹¹In, and the results obtained were compared with those of hMN14 × 679. For BS1.5 injected 8 h before giving the hapten, tumor uptake of ¹¹¹In was 10.3 and 6.3% ID/gram at 3 and 24 h, respectively, with the tumor:blood ratios being 167 at 3 h and 631 at 24 h. For hMN14 × 679 injected 24 h before giving the hapten, the tumor uptakes of ¹¹¹In were 11.3 and 6.9% ID/gram at 3 and 24 h, respectively, with the tumor:blood ratios being 8.1 at 3 h and 16.4 at 24 h. Thus, the concentrations of hapten in the tumor were similar between BS1.5 prelocalized for 8 h and hMN14 × 679 prelocalized for 24 h. However, because the radioactivity in all nontumor organs was higher for hMN14 × 679, the tumor:nontumor ratios were superior for BS1.5. These results, as summarized in Table 4, indicate that BS1.5 is an attractive candidate for use in a variety of pretargeting applications.

The potential of using a radiosensitizer to augment the therapeutic efficacy of RAIT delivered by the AES pretargeting method has also been demonstrated. In one study (72), paclitaxel, but not doxorubicin, was shown to improve the antitumor response in animals given an anti-CEA, anti-indium-DTPA bsAb (F6 × 734), followed by ¹³¹I-labeled bivalent hapten and the chemotherapeutic drugs. A synergistic effect was observed without increasing toxicity.

These promising developments seem to indicate that the future evolution of pretargeting technology will be directed to at least three potential approaches. The first is to use peptides instead of antibodies for specific targeting. The second is to replace chemically linked bsAb conjugates with recombinant fusion proteins of multispecificity and multivalency. The third is to investigate whether a certain chemotherapeutic agent, which is also a known radiosensitizer, can produce synergistic effects when combined with a pretargeting system for treating a particular cancer. In addition, the potential of a hapten or hapten-chelate that also contains a receptor-binding moiety for enhancing the specific uptake in receptor-positive target cells is worth exploring. Needless to say, the urgent need to establish, for each pretargeting system in carefully designed clinical trials, the optimal amount of the targeting macromolecule, the best timing of the second injection with the effector (“two-step protocol”) or the chaser (“three-step protocol”), the maximum tolerated dose, and the therapeutic efficacy for a certain indication continues to exist and should serve as a basis of future experimentation.