Department of Medical Oncology, Fox Chase Cancer Center, 7701 Burholme Avenue, Philadelphia, Pennsylvania, PA 19111, USA

Correspondence to: L M Weiner, Department of Medical Oncology, Fox Chase Cancer Center, 7701 Burholme Avenue, Philadelphia, Pennsylvania, PA 19111, USA

Antibody-based therapy of human cancers has led to several remarkable outcomes, particularly in the therapy of breast cancer and lymphoma. Many solid tumors have proven less responsive, due in part to difficulties in the tumor-selective delivery of antibodies and potential cytolytic effectors. However, antibodies that directly perturb signaling mechanisms in cells derived from epithelial malignancies have shown benefit; examples include antibodies directed against the extracellular domains of HER2/neu and epidermal growth factor receptor. A long-term goal of immunotherapy has been to induce anti-tumor inflammatory responses that can directly cause tumor regression or induce adaptive responses against tumor-related antigens. This review focuses on the use of antibodies to provide a means for initiating anti-tumor immune responses, and on the use of antibodies as delivery vehicles of radionuclides. Oncogene (2000) 19, 6144-6151.

Antibody therapy; radioimmunotherapy; bispecific antibodies

Antibody therapy of human cancer

Antibody-based therapeutics are beginning to realize the promise envisioned by Ehrlich and enabled by the discovery of hybridoma technology more than two decades ago. Unconjugated antibodies directed against the lymphocyte antigen CD20 have significant clinical activity in patients with low-grade lymphomas (Maloney et al., 1994). The FDA has approved one of these antibodies, Rituxan, for clinical use. Radioimmunoconjugates directed against CD20 demonstrate significant clinical activity in patients with chemotherapy-pretreated lymphomas as well (Kaminski et al., 1993; Press et al., 1993). A chemoimmunoconjugate containing an anti-CD33 antibody and calicheamicin has impressive activity in acute myelogenous leukemia (Sievers et al., 1999). Finally, the anti-HER2/neu antibody, Herceptin, has single-agent activity in metastatic breast cancer, and potentiates the anti-tumor effects of taxol chemotherapy (Baselga et al., 1996; Slamon et al., 1998). This antibody also recently received FDA approval for clinical use. These results provide ample evidence that a number of strategies employing unconjugated antibodies or antibody conjugates carrying toxic payloads such as radiation or chemotherapy agents have clinical benefits. It has become clear that targeting signaling elements using unconjugated antibodies can have potent anti-tumor effects, as exemplified by the experiences with Rituxan, Herceptin and the C225 anti-epidermal growth factor receptor antibody. However, further improvements in therapy can be achieved through the conjugation of radionuclides to antibodies, by the appropriate manipulation of antibody structures to improve their targeting properties or biological activity profiles and by the stimulation of host-protective immune responses. This review will focus on some newer approaches that will provide platforms for future initiatives in this field.

Radioimmunotherapy

Antibodies have been long thought to have therapeutic potential as 'magic bullets' to treat cancer (Silverstein, 1994; Pressman and Korngold, 1953). The modern era was ushered in employing polyclonal antisera conjugated to radionuclides (Order et al., 1986; Goldenberg et al., 1978), and has more recently been based on monoclonal antibodies derived by hybridoma or recombinant DNA technology as delivery vehicles for radionuclides. Recently, it has become evident that radioimmunotherapy (RIT) has significant effectiveness in advanced lymphomas (Kaminski et al., 1993; Press et al., 1993; Witzig et al., 1999) and leukemias (Scheinberg et al., 1991). The challenge now is to identify ways to integrate these active approaches into standard care. Efficacy may depend upon the choice of radionuclides. Currently, most approaches use radioiodine, but there is significant interest in employing and emitters such as ²¹³Bismuth and ⁹⁰Yttrium. High energy particles (e.g., from ⁹⁰Yttrium) have long track lengths that can mediate a 'crossfire' phenomenon to deliver lethal radiation doses over multiple cell diameters in a solid tumor, addressing issues such as antigen heterogeneity (Anderson and Strand, 1987) and poor tumor penetration of large antibody molecules. In contrast, particles (e.g., from ²¹³Bismuth or ²¹¹Astataine) deliver a 500-fold greater amount of energy over a very short distance (e.g., one cell diameter). Accordingly, emitters are potentially extremely useful for the treatment of vascular targets or diffuse malignancies.

Influence of structure on antibody targeting

IgG molecules are large proteins of approximately 150 kDa mass; most chemotherapy agents have a molecular weight of less than 1 kDa. Accordingly, intact IgG antibodies would be expected to have significantly slower kinetics of distribution and severely limited tissue penetration properties as compared to small molecules. Indeed, non-uniform uptake of systemically administered antibody is generally observed in biopsied specimens of solid tumors. While inhomogenous tumor antigen expression can be a factor, physiological barriers to antibody penetration bear the greatest responsibility for their limited distribution within a tumor mass. Heterogeneous tumor blood supply limits uniform antibody delivery to tumors and elevated interstitial pressures in the centers of tumors oppose inward diffusion (Jain, 1987). Furthermore, the relatively large transport distances in the tumor interstitium conspire with the above factors to increase the time required for these large macromolecules to reach target cells. For example, the diffusion an intact IgG molecule into a solid tumor is limited to 100 m in 1 h, 1 mm in about 2 days and 1 cm in about 7-8 months (Jain, 1987). These physiologic barriers pose substantial obstacles to antibody penetration in the majority of solid tumors and bulky lymphomas. Thus, it can be anticipated that the therapy of patients with large tumors using Mab will be compromised. These concepts also hold true for potentially cytotoxic leukocytes to accumulate at tumor sites; as a result, physiological barriers can represent a major limitation to the effective clinical exploitation of ADCC.

Contemporary molecular engineering has led to the production of a wide variety of antibody-based targeting structures (Figure 1). These include chimeric human-murine antibodies (LoBuglio et al., 1989), antibodies derived from human hybridomas (Kudo et al., 1991), scFv from murine hybridomas (Huston et al., 1988; Begent et al., 1996; Mack et al., 1995), phage-displayed scFv (Clackson et al., 1991; Schier et al., 1995) and minibodies (Hu et al., 1996), among others. The affinity of these structures for target antigens can be readily altered (Schier et al., 1996; Adams et al., 1998a). Multivalent targeting structures include minibodies (Hu et al., 1996), single-chain dimers (Adams et al., 1993) and diabodies (Holliger et al., 1993; Adams et al., 1998b). Multispecific binding proteins include bispecific antibodies and antibody-based fusion proteins (reviewed below). As this technology has matured and clinical experience has been gained, it is apparent that insufficient selective targeting is a major impediment to the successful use of RIT in the therapy of patients with solid tumors, where impaired vascular access, limited diffusion, and high interstitial tumor pressures conspire to limit the retention of systemically administered protein molecules (Jain, 1987). Moreover, preclinical murine models do not scale up easily to human clinical applications. For example, as much as 30% ID/g selective retention of intravenously administered IgG is routinely observed in human tumor xenografts in mice. However, when the same molecules are tested in human clinical trials, less than 0.1% ID/g tumor retention is commonly observed (Colcher et al., 1988; Gallinger et al., 1993). The effective treatment of human solid tumors using RIT may require at least twofold higher tumor-selective retention of systemically administered radionuclides than is currently seen in murine preclinical models. Such models have been used as predictors for clinical dosimetry by examining cumulative tumor: normal organ radionuclide retention and measuring the ratios of area-under-the-curve (AUC) of tumors and normal organs over the time course of radionuclide distribution. For example, cumulative tumor: bone marrow ratios of at least 16 : 1 (estimated by tumor : blood ratios of at least 4 : 1) in murine models may identify a particular antibody conjugate with promise. If marrow irradiation proves to be dose limiting, as is common, this ratio would provide tumor irradiation of about 4800 cGy with 300 cGy marrow irradiation. Currently available molecules can approach, but probably cannot exceed the threshold needed for the development of successful therapeutic radioimmunoconjugates for solid tumors.

We and others have manipulated antibody structures to improve the selective tumor retention of intravenously administered antibodies. We initially tested the hypothesis that the small size of single-chain Fv (scFv) molecules would lead to an ideal combination of acceptable quantitative tumor retention and highly selective tumor targeting. Selective tumor targeting in the terminal phases of antibody biodistribution was quite good, with tumor : blood ratios of 20 : 1 or higher. However, quantitative tumor retention in murine models was very low, with no more than 1% ID/g retained 24 h following intravenous administration of ¹²⁵Iodine labeled protein. The overall tumor : normal organ AUC values for scFv and IgG molecules are surprisingly similar, with cumulative retention values of 1.1 : 1 for IgG and 2.0 : 1-2.5 : 1 for scFv monomers (Weiner et al., 1995b). Thus, scFv molecules are excellent vehicles for radioimmunodiagnosis, but not for RIT. Manipulations of scFv dose, administration route and labeling procedures did not substantially alter these findings (Adams et al., 1995). ScFv dimers with smaller size than IgG and divalent binding to antigen had minimally improved overall tumor targeting compared with scFv monomers or IgG (Adams et al., 1993). Therefore, the modest alteration of antibody size (e.g., 25 kDa to 50 kDa) did not sufficiently improve overall tumor targeting to facilitate RIT.

We hypothesized that increasing the affinity of antibodies for tumor antigens by prolonging K_off rates would enhance tumor cell surface retention of these proteins, and correspondingly improve selective in vivo tumor retention. For example, prolonging K_off rates to 10^-5/s will translate to cell surface retentions of approximately 18 h. Given baseline endocytosis and protein catabolism rates at tumor sites, this duration probably approximates the maximum retention permitted by physiological constraints. Using a human phage display library, scFv reactive with HER2/neu were isolated and shown to specifically target relevant tumors in vitro and in vivo with similar properties as previously described hybridoma-derived anti-HER2/neu directed scFv molecules (Schier et al., 1996; Adams et al., 1998a). Using chain-shuffling and site-directed mutagenesis of the V_H and V_L CDR3 domains of the C6.5 scFv, a panel of mutants was created with affinities for HER2/neu ranging from 10^-6 M to 10^-11 M, and K_off rates ranging from 10^-1/s to 10^-5/s (Schier et al., 1996). The preparation of this unique panel of affinity mutants allowed the testing of the hypothesis that increased affinity improves in vitro and in vivo tumor retention in murine models. So far, we have shown that there is clearly a 'threshold' affinity requirement for HER2/neu of about 10^-8 M for in vivo tumor targeting. Increasing the affinity to 10^-9 M (K_off ~10^-4/s) further improves quantitative tumor retention, but by less than twofold compared with the parental scFv (Adams et al., 1998a). This panel of affinity variants will allow definitive testing of the predictions of Weinstein, who has postulated the existence of a 'binding site barrier' that inhibits the diffusion of high affinity antibodies through tumors (Juweid et al., 1992).

As noted above, the construction of scFv dimers improves scFv tumor targeting (Adams et al., 1993). This improvement is due to higher valence as opposed to larger size and slower clearance of the dimeric species. The improvements in quantitative tumor retention by dimeric scFv connected by an intermediate length linker (e.g., [G₄S]₃) are modest, and are insufficient to predict for successful RIT. However, the specific dimeric structure significantly influences tumor targeting, as seen with C6.5 diabodies. Diabodies are non-covalently linked recombinant dimers ([V_H-V_L]-[V_H-V_L]) in which the spacer between the two scFv genes contains a very short (usually less than eight amino acids) linker. This forces two scFv to associate (e.g., V_Hx -V_Ly) so that the two binding sites of the dimer are actually formed by these heterologous associations. The resulting molecule is more compact than a conventional dimeric structure, and the binding sites are most likely arrayed in a Janusian fashion (Holliger et al., 1993; Adams et al., 1998b). The C6.5 diabody was shown to have higher functional affinity (K_d=4´10^-10 M) than either the C6.5 scFv monomer or gene-fused dimer. This high affinity translated to prolonged cell surface retention in in vitro assays (T_1/2=5 h) and strikingly improved in vivo tumor retention, with overall targeting specificity equivalent to, and quantitative tumor retentions sixfold higher than that seen with C6.5 scFv monomer (Adams et al., 1998b; Wu et al., 1996). The tumor : bone marrow AUC ratio for this diabody is approximately 12 : 1, and the radiolabeled C6.5 diabody is being tested for anti-tumor efficacy in preclinical RIT studies.

Taken together, these studies indicate that the alteration of antibody structures can have profound effects on quantitative and selective tumor targeting. However, the overall selectivity of antibody targeting is unlikely to exceed that seen during the terminal distribution phase of antibody clearance. Thus, it is appropriate to consider how this specificity can be optimally exploited for therapeutic applications. Antibody pre-targeted therapeutics offer one approach.

Pre-targeted antibody therapy

Although antibodies can target tumor cells in vivo, the peak selective targeting advantage rarely exceeds 5 : 1 (measured as tumor : normal organ ratios) and is rarely more than 1 : 1 over the time course of antibody distribution. This is insufficient specificity for the selective delivery of most highly toxic agents. Pre-targeting of tumors by anti-tumor antibodies permits the systemic clearance of unbound antibodies and so increases the selectivity of tumor targeting by these antibodies. The great advantage of pre-targeting is that the antigen specificity of antibody targeting is optimally exploited, and retention of the antibody conjugate by normal organs is minimized when a subsequent 'activation' step is employed to release a tumor-toxic principle at tumor sites (Figure 2). The initial example of this strategy was antibody-directed enzyme prodrug therapy (ADEPT), in which an anti-tumor antibody-enzyme conjugate is administered systemically, and then allowed to clear from normal tissues. An inactive chemotherapy prodrug is then administered that rapidly clears from the host, but is selectively retained at tumor sites by the antibody-targeted enzyme to liberate the active chemotherapy agent at tumor sites (Bagshawe et al., 1993). Various iterations of this strategy have been tested in preclinical models and in human clinical trials (Bagshawe et al., 1991; Martin et al., 1997; Siemers et al., 1997). More recently, antibody pre-targeted radioimmunotherapy (PRIT) has been tested, primarily employing the anti-Ep-CAM antibody NR-LU-10, conjugated to streptavidin. Following a systemic clearance step employing biotinylated proteins, the tumors and other Ep-CAM expressing normal cells are decorated with streptavidin. Then, biotin-labeled radiometals are administered. These small molecules are rapidly cleared, but are selectively retained at tumor sites containing streptavidin. This strategy exploits the exceptionally high affinity binding of biotin to streptavidin (K_d <10^-14 M) and the multivalent binding of biotin and streptavidin (Axworthy et al., 1996).

In preclinical models, this PRIT strategy yields exceptional, highly selective tumor targeting and impressive anti-tumor efficacy (Martin et al., 1997). These findings have formed the basis for clinical trials that have been associated with substantial toxicity and disappointing efficacy, due to several factors (Knox et al., 2000). The NR-LU-10 antibody targets normal tissues, particularly the gastrointestinal mucosa. Diarrhea was the dose-limiting toxicity in Phase I trials. The large size of the NR-LU-10 antibody-streptavidin conjugate restricts its diffusion into tumor and increases its distribution in normal organs. This reduces therapeutic targeting and increases the non-specific delivery of radiation. The clearing steps are cumbersome and incomplete, so that streptavidin 'leaches' back into circulation after clearance of the circulating streptavidin. This reduces the tumor : blood ratios at the time of radionuclide administration, and lowers targeting specificity. The NR-LU-10 antibody-streptavidin conjugate is highly immunogenic. This precludes the administration of prolonged or repeated courses of therapy. Endogenous biotin in the host is bound by the streptavidin, blocking the tumor accumulation of subsequently administered ⁹⁰Y-DOTA. The clearance step may occupy tumor-associated streptavidin, interfering with subsequent radionuclide accumulation at tumor sites. The NR-LU-10 antibody probably exhibits insufficient selective targeting of tumors to permit therapeutic dosing even if the above factors could be corrected. It is important to stress that these deficiencies do not invalidate the fundamental promise of the PRIT strategy, but suggest avenues for improvement.

For PRIT to achieve its potential the chosen tumor antigen target must have highly restricted expression in normal tissues. The tumor-targeting vehicle must be multivalent and highly tumor-avid, yet small enough to penetrate into tumors from the vasculature and rapidly clear from normal organs. A formal clearance step should be unnecessary. The tumor-targeting vehicle must be non-immunogenic, or nearly so. Ideally, it should be composed of human proteins. The high affinity interactions between the tumor-targeting vehicle (e.g., a human antibody) and the radionuclide should be mediated by non-immunogenic proteins, again ideally of human origin. The high affinity interactions should not involve potential cross-reactivity with host elements such as endogenous biotin.

The desired characteristics can be achieved by employing antibody engineering and phage display technology. The target chelates can be conjugated to radiometals such as the -emitter ⁹⁰Yttrium and the -emitter, ²¹³Bismuth. The capacity to employ multiple radionuclides indicates that this core strategy offers the potential to function as a delivery system for diverse radionuclides and other chelated substances (Goodwin et al., 1988; Paganelli et al., 1988). The general pre-targeting concept can be applied to direct the delivery of chemotherapy agents (Juweid et al., 1992; Wu et al., 1996; Bagshawe, 1993), toxins, cytokines or other tumor-modulatory agents.

Antibody-dependent cellular cytotoxicity

We now shift gears to discuss the prospects for using antibodies to manipulate the host immune response to tumors. At this time, the mechanisms underlying the clinical benefits of unconjugated antibodies such as Rituxan and Herceptin are not understood, but both antibodies mediate antibody-dependent cellular cytotoxicity (ADCC) in vitro and may do so as well in the clinical setting (Maloney et al., 1994; Baselga et al., 1996; Slamon et al., 1998). Recent studies in Fc receptor-deficient mice indicates that antitumor effects depend upon the presence of intact Fc receptor function (Clynes et al., 2000). Many other antibodies have been clinically tested to exploit their ADCC properties, but without consistent clinical benefit in the setting of advanced disease (Weiner et al., 1986; Vadhan-Raj et al., 1988). ADCC continues to be a potent in vitro phenomenon that should provide a powerful anti-tumor mechanism, if it can be replicated or mimicked in the clinical setting (Steplewski et al., 1983). Advances in the understanding of human Fc receptor and T cell receptor structures and functions have made it possible to refine antibody-directed cellular activation strategies, particularly given the improved flexibility afforded by antibody engineering techniques. Human Fc and Fc receptors can trigger cellular cytotoxicity (Garcia de Palazzo et al., 1990; Valerius et al., 1997; Shen et al., 1986), as can elements of the T cell receptor/CD3 complex (Segal and Wunderlich, 1988; Lanzavecchia and Scheidegger, 1987; Link and Weiner, 1993) CD28 on T cells (Renner et al., 1995) and CD44 on natural killer (NK) cells (Sconocchia et al., 1994). It is now possible to create a wide variety of recombinantly-produced antibody binding site-based targeting proteins, including IgG, (Fab')₂ Fab, and single-chain Fv (scFv) formats, with an equally formidable choice of approaches to create multimeric, monospecific or bispecific binding proteins as discussed earlier in this review. The affinity properties of the binding sites can be manipulated through chain-shuffling or site-directed mutagenesis (Schier et al., 1996). The human anti-mouse antibody response can be circumvented by the production of chimeric antibodies (LoBuglio et al., 1989; Kudo et al., 1991), or by the use of human antibodies prepared by fusion techniques or from human phage display libraries (Schier et al., 1995).

Bispecific antibody therapy

Bispecific antibodies

In the nearly 15 years following the pioneering work of David Segal and colleagues numerous bispecific antibodies targeting tumor antigens and effector cell trigger molecules have been developed and shown to redirect cellular cytotoxicity. For example, BsAb can target tumor antigens and human effector cell trigger molecules on T cells via CD3/TcR (Valerius et al., 1997; Shen et al., 1986) and CD28 (Segal and Wunderlich, 1988). BsAb directed against FcR1 (Steplewski et al., 1983) trigger tumor cytotoxicity by neutrophils, monocytes and macrophages. BsAb targeting FcRIII (Weiner et al., 1986; Vadhan-Raj et al., 1988) promote tumor lysis by macrophages and NK cells and BsAb targeting CD44 (Lanzavecchia and Scheidegger, 1987) promote tumor cytotoxicity by NK cells. Typically, BsAb promote in vitro cytotoxicity at low concentrations and have been shown to cause either growth delays or tumor regressions in appropriate animal models (Weiner et al., 1993a, 1994). These properties have led to the testing of several BsAb in human clinical trials.

Clinical toxicities of bispecific antibodies

Several early reports of CD3-directed BsAb suggested possible applications in glioblastoma (Nitta et al., 1990) and ovarian cancer (Canevari et al., 1995). However, subsequent trials of a number of intact CD3-directed BsAb have yielded few objective responses, and treatment has been limited by toxicities reminiscent of the first dose effect seen with OKT3 antibody therapy. The first dose effect is characterized by the release of massive amounts of cytokines such as tumor necrosis factor- (TNF). Treatment with an (Fab')₂ anti-CD3 BsAb, which is bereft of Fc domains and Fc domain-dependent cellular cross-linking properties, also has led to toxicities at low BsAb doses. This finding indicates that the engagement of some activating molecules on circulating T cells can produce unacceptable dose-limiting toxicity (Tibben et al., 1993). We have conducted a series of clinical trials employing the BsAb 2B1, a murine IgG that is dually specific for extracellular domain epitopes on the human tumor antigen HER2/neu and the human Fc receptor, FcRIII (Weiner et al., 1995a). While therapy with 2B1 yielded some clinical responses, BsAb-induced cross-linking of Fc receptors on circulating leukocytes led to massive cytokine release and associated toxicities. While Rituximab can be safely administered at doses of 375 mg/m² per dose, the maximally-tolerated dose of 2B1 is only 1 mg/m² per dose. 2B1-promoted antigen presentation led to the induction of anti-HER2/neu antibodies, but only in patients treated with 2B1 doses exceeding the ultimate MTD (Clark et al., 1997). Taken together, these results indicate that whole IgG BsAb targeting effector cell trigger molecules are unsuitable therapeutic structures, in accord with predictions in selected preclinical murine models (Link et al., 1998). The engagement of T cell receptor V chains and MHC Class II by the bacterial superantigen, staphylococcal enterotoxin A (SEA), also leads to profound leukocyte activation and the attendant dose-limiting toxicity of a multi-specific fusion protein consisting of an anti-tumor Fab fragment and SEA (Giantonio et al., 1997; Alpaugh et al., 1998). These findings suggest that the cellular activation properties of intact BsAb and antibody-superantigen fusion proteins will cause unacceptable toxicities unless these properties can be selectively localized to tumor sites.

Importance of tumor inflammation for effective bispecific antibody therapy

The clinical experience with the MDX-210 BsAb provides a useful additional perspective to the results described above (Valone et al., 1995; Curnow, 1997). MDX-210 targets the same epitope on HER2/neu as does 2B1, but targets human FcRI on neutrophils and mononuclear phagocytes instead of FcRIII. In contrast to 2B1, MDX-210 is a chemically conjugated (Fab')₂. Because it lacks an Fc domain, MDX-210 cannot activate leukocyte Fc receptors in the absence of tumor antigen engagement. Like 2B1, MDX-210 is a potent mediator of in vitro cytotoxicity and phagocytosis. Since much higher doses of this BsAb can be given safely in comparison with 2B1, monomeric FcRI engagement may have different consequences than do FcR aggregation or CD3 engagement. Like 2B1, MDX-210 therapy leads to antigen presentation and the induction of anti-HER2/neu antibodies. Again, as with 2B1, some clinical responses have been seen with BsAb targeting FcRI. Simply increasing the dose of BsAb may not be sufficient to reliably induce clinical responses, so that further modifications to BsAb treatment strategies are needed. Finally, the results obtained with MDX-210 indicate that the Fc domain of a BsAb is not required for Fc receptor-dependent antigen presentation.

Relatively few tumor biopsies have been performed in patients treated with 2B1 or MDX-210. Some tumor infiltration by leukocytes has been seen following MDX-210 therapy (Link et al., 1998), and we have observed the induction of tumor infiltration by leukocytes in several patients treated with 2B1. However, there is no consistent, profound leukocyte infiltration of tumor following BsAb therapy, and preclinical data suggest that BsAb promoted cytotoxicity will require effector to target ratios of at least 1 : 1 for successful therapy (Weiner et al., 1993b). It is probable that effective BsAb treatment strategies must include ways to induce tumor inflammation so that BsAb-mediated potentiation of cytotoxicity can be properly exploited. This may be accomplished by introducing chemotactic or cell proliferation signals into antibody or BsAb constructs.

Strategies for the induction of selective tumor inflammation

A number of strategies can be considered to promote selective tumor inflammation. Experiences with systemic cytokines have shown that the selectivity of effects at tumor sites is usually low, as evidenced by work with high dose interleukin-2 (IL-2) (Fyfe et al., 1995). With IL-2 and other cytokines such as TNF-, toxicities caused by systemic leukocyte activation have limited the doses that can be safely administered. Several groups have attempted to address this by producing immunocytokines, which are composed of engineered antibodies fused to cytokines. Antibody/IL-2 fusion proteins have exhibited promising preclinical results (Gillies et al., 1992; Sabzevari et al., 1994). However, a major potential limitation to this strategy is that these large (e.g., >150 kDa) proteins have relatively prolonged half-lives. As a result, functionally intact IL-2 will circulate predominantly in the vasculature, leading to systemic leukocyte activation that at least partially offsets any targeting advantages conferred by the antibody binding property. As discussed earlier, such systemic leukocyte activation has interfered with the ability to give reasonable doses of fusion proteins containing antibody Fab domains and the bacterial superantigen, staphylococcal enterotoxin A (SEA).

Less progress has been reported regarding the creation of antibody-targeted vasoactive or chemotactic conjugates or fusion proteins. Conjugates containing antibody binding sites and F-M-L-P (Obrist et al., 1988) and RANTES (Chalitta-Eid et al., 1998) have been reported, but have not undergone extensive preclinical testing or clinical trials. One potential advantage of these approaches over the use of antibody-cytokine fusion proteins is that the former agents are not specifically activating, but function to establish gradients that favor the accumulation of the targeted leukocyte population at tumor sites. In the circulating blood, such gradients will not exist, so that toxicity or preferential systemic leukocyte activation would not be expected. Thus, there is reason for optimism that antibody-delivered chemotactic agents can establish gradients that will promote leukocyte infiltration at tumor sites. If such proteins can be selectively retained in tumors, BsAb should be able to more efficiently promote leukocyte-mediated cytotoxicity. High, relatively sustained intratumoral concentrations of GM-CSF can promote the accumulation of mononuclear phagocytes at tumor sites. This can break immune tolerance and induce immune-mediated tumor regressions (Thomas et al., 1998). Current gene therapy approaches are not able to efficiently target vectors to tumor sites. Accordingly, antibody-directed targeting of other proteins remains an important and validated strategy by which cytokines can be selectively delivered to tumors. Since the considerable clinical experience with GM-CSF indicates that this particular cytokine has a favorable toxicity profile when administered systemically, it is reasonable to hypothesize that efficient tumor targeting of GM-CSF by an antibody that also induces tumor cytotoxicity via FcRIII will promote efficient tumor regression.

Characteristics of an optimized bispecific antibody

Over the past few years the working definition of BsAb can be considered to include all antibody-based conjugates or fusion proteins that exploit antibody combining sites to target tumors and alter the tumor microenvironment to promote leukocyte infiltration and mediate antibody-directed leukocyte-mediated cytotoxicity. For such BsAb to be effective, the molecule must selectively target a relevant tumor antigen with high functional binding affinity. This may be achieved by increasing the intrinsic affinity or by increasing the avidity of binding. This property will enhance selective tumor retention of the BsAb. The molecule should have a low functional affinity for a leukocyte activation or cytotoxicity trigger molecule when the BsAb is in the systemic circulation, but must be able to efficiently trigger leukocyte cytotoxicity when bound to tumor cells expressing the target antigen. This will limit toxicity caused by the unwanted systemic activation of leukocytes. The BsAb should efficiently promote high levels of retargeted leukocyte cytotoxicity. The BsAb should be of human origin to minimize the induction of human anti-mouse immune responses that limit repeat dosing. The BsAb should not contain Fc domains that induce toxicity caused by systemic leukocyte activation resulting from crosslinking of multiple leukocyte Fc receptors. Finally, the BsAb treatment strategy should promote selective in situ tumor inflammation by relevant leukocyte populations.

HER2/neu is an appropriate target for testing the above hypotheses. As discussed earlier, antibodies directed against other HER2/neu epitopes inhibit tumor cell proliferation and are synergistic with chemotherapy agents. One such antibody, Herceptin, exhibits clinical utility and was recently approved for use in women with metastatic breast cancer. Other cytotoxicity triggers have been tested as BsAb components; profound T cell activation by anti-CD3 containing BsAb has induced unacceptable toxicity (Weiner et al., 1994).

Future directions

The rapid pace of advances in cell biology, immunology and protein engineering assures the essential futility of any prognostication attempts regarding antibody-based therapy of human cancer. However, it is safe to presume that the major approaches will exploit the capacity of antibodies to selectively accumulate on cells bearing either structurally or functionally significant targets. Pretargeting offers an important new strategy to take optimal advantage of antibody properties. Finally, the ability of antibodies to promote ADCC and inflammation offers an as yet unexploited mechanism for improving cancer therapy.

Acknowledgements

Supported by NIH grants CA06927, CA50633, CA65559, by the Department of Defense, by an appropriation from the Commonwealth of Pennsylvania, by the Frank Strick Foundation and by the Bernard A and Rebecca S Bernard Foundation.

Adams GP, McCartney JE, Tai M-S, Oppermann H, Huston JS, Stafford WF, Bookman MA, Fand I, Houston LL and Weiner LM. (1993). Cancer Res. 53, 4026-4034. MEDLINE

Adams GP, McCartney JE, Wolf EJ, Eisenberg J, Huston JS, Stafford III WF, Bookman MA, Houston LL and Weiner LM. (1995). Cancer Immunol. Immunother. 40, 299-306. MEDLINE

Adams GP, Schier R, Marshall K, Wolf EJ, McCall AM, Marks JD and Weiner LM. (1998a). Cancer Res. 58, 485-490. MEDLINE

Adams GP, Schier R, McCall AM, Crawford RS, Wolf EJ, Weiner LM and Marks JD. (1998b). Br. J. Cancer 77, 1405-1412.

Alpaugh RK, Schultz J, McAleer C, Giantonio BJ, Persson R, Burnite M, Nielsen SE, Vitek L, Persson B and Weiner LM. (1998). Clin. Cancer Res. 4, 1903-1914. MEDLINE

Anderson W and Strand M. (1987). Natl. Cancer Inst. Monogr. 3, 149-151.

Axworthy DB, Beaumier PL, Bottino BJ, Goshorn S, Mallett RW, Stone DM, Su F-M, Theodore LJ, Yau EK and Reno JM. (1996). Tumor Targeting 2, 156-157.

Bagshawe KD. (1993). Adv. Pharmacol. 24, 99-121. MEDLINE

Bagshawe KD, Sharma SK, Springer CJ, Antoniw P, Boden JA, Rogers GT, Burke PJ, Melton RG and Sherwood RF. (1991). Disease Markers 9, 233-238. MEDLINE

Baselga J, Tripathy D, Mendelsohn J, Baughman S, Benz CC, Dantis L, Sklarin NT, Seidman AD, Hudis CA, Moore J, Rosen PP, Twaddell T, Henderson IC and Norton L. (1996). J. Clin. Oncol. 14, 737-744. MEDLINE

Begent RH, Verhaar MJ, Chester KA, Casey JL, Green AJ, Napier MP, Hope-Stone LD, Cushen N, Keep PA, Johnson CJ, Hawkins RE, Hilson AJ and Robson L. (1996). Nature Med. 2, 979-984. MEDLINE

Canevari S, Stoter G, Arienti F, Bolis G, Colnaghi MI, Di Re EM, Eggermont AMM, Goey SH, Gratama JW, Lamers CHJ, Nooy MA, Parmiani G, Raspagliesi F, Ravagnani F, Scarfone G, Trimbos JP, Warnaar SO and Bolhuis RLH. (1995). J. Natl. Cancer Inst. 87, 1463-1469. MEDLINE

Chalitta-Eid P, Abboud CA, Morrison SL, Penichet ML, Rosell KE, Poles T, Hilchey SP, Planelles V and Rosenblatt JD. (1998). J. Immunol. 161, 3729-3736. MEDLINE

Clackson T, Hoogenboom HR, Griffiths AD and Winter G. (1991). Nature 352, 624-628. MEDLINE

Clark JI, Alpaugh RK, von Mehren M, Schultz J, Gralow JR, Cheever MA, Ring DB and Weiner LM. (1997). Cancer Immunol. Immunother. 44, 265-272. MEDLINE

Clynes R, Towers TL, Presta LG and Ravetch JV. (2000). Nature Med. 6, 443-446. Article MEDLINE

Colcher D, Minelli FM, Roselli M, Muraro R, Simpson-Milenic D and Schlom J. (1988). Cancer Res. 48, 4597-4603. MEDLINE

Curnow RT. (1997). Cancer Immunol. Immunother. 4, 210-215.

Fyfe G, Fisher RI, Rosenberg SA, Sznol M, Parkinson DR and Louie AC. (1995). J. Clin. Oncol. 13, 688-696. MEDLINE

Gallinger S, Reilly RM, Kirsh JC, Odze RD, Hay K, Polihronis J, Damani MT, Shpitz B and Stern HS. (1993). Cancer Res. 53, 271-278. MEDLINE

Garcia de Palazzo I, Gercel-Taylor C, Kitson J and Weiner LM. (1990). Cancer Res. 50, 7123-7128. MEDLINE

Giantonio BJ, Alpaugh RK, Schultz J, McAleer C, Newton DW, Guedez Y, Kotb M, Vitek L, Persson R, Kalland T, Dohlsten M, Persson B and Weiner LM. (1997). J. Clin. Oncol. 15, 1994-2007. MEDLINE

Gillies SD, Reilly EB, Lo KM and Reisfeld RA. (1992). Proc. Natl. Acad. Sci. USA 89, 1428-1432. MEDLINE

Goldenberg DM, DeLand F, Kim E, Bennett S, Primus FJ, van Nagell Jr JR, Estes N, DeSimone P and Rayburn P. (1978). New Engl. J. Med. 298, 1384-1386. MEDLINE

Goodwin DA, Meares CF, McCall MJ, McTigue M and Chaovapong W. (1988). J. Nucl. Med. 29, 226-234. MEDLINE

Holliger P, Prospero T and Winter G. (1993). Proc. Natl. Acad. Sci. USA 90, 6444-6448. MEDLINE

Hu S, Shively L, Raubitschek A, Sherman M, Williams LE, Wong JY, Shively JE and Wu AM. (1996). Cancer Res. 56, 3055-3061. MEDLINE

Huston JS, Levinson D, Mudgett-Hunter M, Tai MS, Novotny J, Margolies MN, Ridge RJ, Bruccoleri RE, Haber E and Crea R. (1988). Proc. Natl. Acad. Sci. USA 85, 5879-5883. MEDLINE

Jain RK. (1987). Cancer Res. 47, 3039-3051. MEDLINE

Juweid M, Neumann R, Paik C, Perez-Bacete M, Sato J, van Osdol W and Weinstein JN. (1992). Cancer Res. 52, 5144-5153. MEDLINE

Kaminski MS, Zasadny KR, Francis IR, Milik AW, Ross CW, Moon SD, Crawford SM, Burgess JM, Petry NA, Butchko GM, Glenn SD and Wahl RL. (1993). New Engl. J. Med. 329, 459-465. MEDLINE

Knox SJ, Goris ML, Tempero M, Weiden PL, Gentner L, Breitz H, Adams GP, Axworthy D, Gaffigan S, Bryan K, Fisher DR, Colcher D, Horak ID and Weiner LM. (2000). Clin. Cancer Res. 6, 406-414. MEDLINE

Kudo T, Saeki H and Tachibana T. (1991). J. Immunolog. Methods 145, 119-125.

Lanzavecchia A and Scheidegger D. (1987). Eur. J. Immunol. 17, 105-111. MEDLINE

Link BK, Kostelny SA, Cole MS, Fusselman WP, Tso JY and Weiner GJ. (1998). Interntl J. Cancer 77, 251-256.

Link BK and Weiner GJ. (1993). Blood 81, 3343-3349. MEDLINE

LoBuglio AF, Wheeler RH, Trang J, Haynes A, Rogers K, Harvey EB, Sun L, Ghrayeb J and Khazaeli MB. (1989). Proc. Natl. Acad. Sci. USA 86, 4220-4224. MEDLINE

Maloney DG, Liles TM, Czerwinski DK, Waldichuk C, Rosenberg J, Grillo-Lopez A and Levy R. (1994). Blood 84, 2457-2466. MEDLINE

Mack M, Riethmuller G and Kufer P. (1995). Proc. Natl. Acad. Sci. USA 92, 7021-7025. MEDLINE

Martin J, Stribbling SM, Poon GK, Begent RH, Napier M, Sharma SK and Springer CJ. (1997). Cancer Chemother. Pharmacol. 40, 189-201. MEDLINE

Nitta T, Sato K, Yagita H, Okumura K and Ishii S. (1990). Lancet 335, 368-371. MEDLINE

Obrist R, Schmidli J and Obrecht JP. (1988). Biochem. Biophys. Res. Commun. 155, 1139-1144. MEDLINE

Order SE, Klein JL, Leichner PK, Frincke J, Lollo C and Carlo DJ. (1986). Int. J. Radiat. Oncol., Biol., Phys. 12, 277-281.

Paganelli G, Riva P, Deleide G, Clivio A, Chiolerio F, Scassellati GA, Malcovati M and Siccardi AG. (1988). Interntl J. Cancer 2, (suppl) 121-125.

Press OW, Eary JF, Appelbaum FR, Martin PJ, Badger CC, Nelp WB, Glenn S, Butchko G, Fisher D, Porter B, Matthews DC, Fischer LD and Bernstein ID. (1993). New Engl. J. Med. 329, 1219-1224. MEDLINE

Pressman D and Korngold L. (1953). Cancer 6, 619-623.

Renner C, Jung W, Sahin U, van Lier R and Pfreundschuh M. (1995). Eur. J. Immunol. 25, 2027-2033. MEDLINE

Sabzevari H, Gillies SD, Mueller BM, Pancook JD and Reisfeld RA. (1994). Proc. Natl. Acad. Sci. USA 91, 9626-9639. MEDLINE

Scheinberg DA, Lovett D, Divgi CR, Graham MC, Berman E, Pentlow K, Feirt N, Finn RD, Clarkson BD, Gee TS, Larson SM, Oettgen HF and Old LJ. (1991). J. Clin. Oncol. 9, 478-490. MEDLINE

Schier R, Marks JD, Wolf EJ, Apell G, Wong C, McCartney JE, Bookman MA, Huston JS, Houston LL, Weiner LM and Adams GP. (1995). Immunotechnology 1, 73-81. MEDLINE

Schier R, McCall A, Adams GP, Marshall KW, Merritt H, Yim M, Crawford RS, Weiner LM and Marks JD. (1996). J. Mol. Biol. 263, 551-567. Article MEDLINE

Sconocchia G, Titus JA and Segal DM. (1994). J. Immunol. 153, 5473-5481. MEDLINE

Segal DM and Wunderlich JR. (1988). Cancer Invest. 6, 83-92. MEDLINE

Shen L, Guyre PM, Anderson CL and Fanger MW. (1986). J. Immunol. 137, 3378-3382. MEDLINE

Siemers NO, Kerr DE, Yarnold S, Stebbins MR, Vrudhula VM, Hellstrom I, Hellstrom KE and Senter PD. (1997). Bioconjugate Chem. 8, 510-519.

Sievers EL, Appelbaum FR, Spielberger RT, Forman SJ, Flowers D, Smith FO, Shannon-Dorcy K and Berger MS. (1999). Blood 93, 3678-3684. MEDLINE

Silverstein AM. (1994). Cell. Immunol. 174, 1-6.

Slamon DJ, Leyland-Jones B, Shak S, Paton V, Bajamonde A, Fleming T, Eirmann W, Wolter J, Baselga J and Norton L. (1998). Proc. ASCO 17, A377.

Steplewski Z, Lubeck MD and Koprowski H. (1983). Science 221, 865-867. MEDLINE

Thomas MC, Greten TF, Pardoll DM and Jaffee EM. (1998). Hum. Gene Thera. 9, 835-843.

Tibben JG, Boerman OC, Claessens RAMJ, Corstens FHM, van Deuren M, de Mulder PHM, van der Meer JWM, Keijser KGG and Massuger LFAG. (1993). J. Natl. Cancer Inst. 85, 1003-1004. MEDLINE

Vadhan-Raj S, Cordon-Cardo C, Carswell E, Mintzer D, Dantis L, Duteau C, Templeton MA, Oettgen HF, Old LJ and Houghton AN. (1988). J. Clin. Oncol. 6, 1636-1648. MEDLINE

Valerius T, Stockmeyer B, van Spriel AB, Graziano RF, van den Herik-Oudijk IE, Repp R, Deo YM, Lund J, Kalden JR, Gramatzki M, van de Winkel JG and Fcalpha RI. (1997). Blood 90, 4485-4492. MEDLINE

Valone FH, Kaufman PA, Guyre PA, Lewis LA, Memoli V, Deo Y, Graziano R, Fisher J, Meyer L, Mrozek-Orlowski M, Wardwell K, Guyre V, Morley TL, Arvizu C and Fanger MW. (1995). J. Clin. Oncol. 13, 2281-2292. MEDLINE

Weiner GJ, Kostelny SA, Hillstrom JR, Cole MS, Link BK, Wang SL and Tso JY. (1994). J. Immunol. 152, 2385-2392. MEDLINE

Weiner LM, Holmes M, Adams GP, LaCreta F, Watts P and De Palazzo IG. (1993a). Cancer Res. 53, 94-100.

Weiner LM, Holmes M, Richeson A, Godwin A, Adams GP, Hsieh-Ma ST, Ring DB and Alpaugh RK. (1993b). J. Immunol. 151, 1-9.

Weiner LM, Clark JI, Davey M, Li WS, Garcia de Palazzo I, Ring DB and Alpaugh RK. (1995a). Cancer Res. 55, 4586-4593.

Weiner LM, Houston LL, Huston JS, McCartney JE, Tai M-S, Apell G, Stafford WF, Bookman MA, Gallo JM and Adams GP. (1995b). Tumor Targeting 1, 51-60.

Weiner LM, Steplewski Z, Koprowski H, Sears HF, Litwin S and Comis RL. (1986). Hybridoma 5, (Suppl. 1) 65-77.

Witzig TE, White CA, Wiseman GA, Gordon LI, Emmanouilides C, Raubitschek A, Janakiraman N, Gutheil J, Schilder RJ, Spies S, Silverman DH, Parker E and Grillo-Lopez AJ. (1999). J. Clin. Oncol. 17, 3793-3803. MEDLINE

Wu AM, Chen W, Raubitschek AA, Willams LE, Fischer R, Hu S-Z, Odom-Maryon T, Wong JYC and Shively JE. (1996). Immunotechnology 2, 21-36. MEDLINE

Figure 1 Antibody-based targeting proteins

Figure 2 Antibody-pretargeted therapy