TARGETED THERAPY

Effective cancer therapy requires anti-tumor effects with selective action or selective cancer targeting. This may be accomplished by delivering potentially toxic agents to all cells; selective anti-tumor effects would be accomplished via unique structural or functional tumor characteristics. Alternatively, molecules that selectively target tumors can act as carriers of various toxic moieties. Initial attempts to create tumor-targeting molecules employed murine monoclonal antibodies prepared by hybridoma technology. Such antibodies have been used to target host immune responses to tumors, or to deliver catalytic toxins, radionuclides or chemotherapy agents. Successful therapeutic strategies must overcome a number of obstacles, including the heterogeneous expression of tumor antigen, incomplete binding of administered antibody to antigen expressing tumors, the induction of host anti-murine antibody responses, and the expression of target antigens by nonmalignant cells, which can cause nonspecific toxicity. Unconjugated antibodies designed to focus a host-derived cellular immune response against tumors also must overcome the obstacles of inadequate antibody-mediated effector functions and insufficient migration of relevant effector cells to tumor sites. Despite such obstacles, clinical trials have identified important new monoclonal antibodies with significant antitumor properties in non-Hodgkin's lymphomas and in breast cancer. Two such antibodies are widely employed in clinical practice.

Since antibody-combining sites offer a powerful means for specific tumor targeting, we have concentrated on identifying the structural properties of antibodies that promote optimal selective tumor targeting by antibody-based proteins. The ultimate goal of these efforts remains the induction of inflammatory anti-tumor immune responses initiated by the delivery of antibody binding site-based proteins to tumors.

Bispecific antibodies targeting tumor antigens and effector cell cytotoxicity trigger molecules represent an attempt to specific-ally induce anti-tumor immune responses using antibody therapy. The bispecific antibody, 2B1, is an immunoglobulin (Ig) G antibody prepared by fusion of two previously prepared hybridomas. 2B1 has specificity for the extracellular domain of the HER2/neu (cerbB-2) oncogene product, and binds to an epitope of the extracellular domain of the human Fcg receptor, FcgRIII. 2B1 binds monovalently to each of its antigen targets. Purified 2B1 efficiently promotes the in vitro lysis of HER2/neu-expressing tumor cells by natural killer cells and macrophages expressing FcgRIII at low effector-to-target cell ratios. The antibody has anti-tumor effects against HER2/neu-expressing human tumor xenografts in immunodeficient severe combined immunodeficient (scid) mice. Such mice do not, however, provide an optimal preclinical model since the leukocytes of these mice do not express human FcgRIII.

Treatment of patients with 2B1 leads to the release of a number of cytokines in the blood. Many patients treated with 2B1 develop antibody responses directed against multiple portions of the HER2/neu molecule. These findings indicate that 2B1 treatment-induced tumor lysis leads to subsequent antigen presentation and initiation of adaptive immune responses to this protein. Thus, targeting antigens for lysis via cellular FcgRs has the potential to break tolerance to self-proteins. If so, therapeutic immunization may be accomplished by targeting antigen presentation through FcgRs. Such observations may have important consequences for the future development of vaccines for cancer and other diseases.

To produce more relevant models of bispecific antibody therapy, we have developed mice that are transgenic for human FcgRIII. Mice transgenic for the A isoform of human FcgRIII express this isoform on macrophages and natural killer cells, while mice expressing the B isoform of this receptor bear abundant human FcgRIII on neutrophils. Selective mating of FcgRIIIA and FcgRIIIB mice has led to the establishment of immunocompetent lines and mice transgenic for functional human FcgRIIIA and B with transgene receptor expression patterns that closely mimic those seen in humans.

Chemically-conjugated HER2/neu--anti FcgR proteins are being tested in the transgenic mouse models that we have developed; these studies will test the hypothesis that antigens can be targeted to FcgR-expressing cells to promote immunization against this protein. A kinase-deficient variant of HER2/neu has also been introduced into mice. The resulting mice will be tolerant to human HER2/neu and can be used to examine the abilities of 2B1 and the HER2/neu anti-FcgRIII conjugate to promote immunization against human HER2/neu.

Recent clinical trials employing 2B1 alone or in combination with the cytokines, granulocyte macrophage-colony stimulating factor (GM-CSF) and interleukin (IL)-2 have shown that the bispecific antibody induces substantial toxicity at low doses; these doses are insufficient to reliably induce significant tumor inflammation. We have postulated that improvements in bispecific antibody-based therapy will require that the antibodies be small to facilitate tumor penetration. 1) The antibodies should bind to tumor antigens with very high functional affinity to capture targeted leukocytes as they traffic through the tumor. 2) The affinity for leukocyte targets should be low enough to limit undesirable binding in circulating blood, an event that prematurely activates leukocytes and causes dose-limiting toxicity. 3)The antibody should be of human origin to limit host-protective anti-mouse antibody immune responses and should be an exceptionally efficient mediator of retargeted leukocyte cytotoxicity against antigen expressing tumor cells. And finally, 4) the treatment strategy should include measures to stimulate increased trafficking of leukocytes to tumor sites.

To address these issues we have prepared a series of novel bispecific antibodies as single-chain Fv (scFv) dimers. These bispecific reagents were prepared by isolating novel antibodies from a human phage display library. These dimers comprise the NM3E2 anti-FcgRIII scFv, which is linked to one of several scFv that are reactive with differing affinities for an epitope on the extracellular domain of HER2/neu. The bispecific (scFv)₂ molecules with higher affinities for HER2/neu efficiently promote targeted tumor lysis by FcgRIII-expressing human mononuclear leukocytes; one of these molecules has a 10^-10 M affinity for HER2/neu and a 5x10^-8 M affinity for human FcgRIII. As human proteins, these scFv should be poorly immunogenic and, therefore, could be repeatedly administered. They contain no Fcg domain and, therefore, should not trigger Fcg receptor crosslinking and activation in the peripheral blood. The relatively low affinity of these scFv for FcgRIII should permit them to traffic to tumor sites efficiently without engaging cellular FcgR in the circulating blood. Finally, the high affinity of these scFv for HER2/neu should promote retention at tumor sites and permit efficient in situ tumor lysis. These molecules are the most efficient promoter of cyto-toxicity by natural killer cells of the recombinant scFv dimers that we have produced; however, they are still less potent than 2B1, despite higher affinity for HER2/neu than 2B1. We postulate that the recombinant scFv has insufficient span and flexibility to optimally mediate cytotoxicity. Therefore, we are preparing larger, scFv-based molecules with identical specificities to address these considerations.

The relatively large size of Ig molecules (150 kD) restricts their penetration from the circulating blood into tumors that express relevant tumor antigens. Recombinant scFv contain variable heavy and light chain domains of antibodies. These scFv possess antigen-binding properties and the exquisite tumor-targeting specificity that is desirable for anti-tumor antibodies. Furthermore, since they are far smaller in size (25 kD) than antibodies, scFv efficiently diffuse through tumors and are rapidly cleared from the circulation. We have examined the binding and in vivo distribution properties of a series of human, phage display-derived scFv that bind to the same epitope of the HER2/neu extracellular domain, with affinities ranging from10^-6 M to 10^-11 M. The availability of this unique panel of affinity mutants has permitted study of the effects of affinity on tumor-targeting. Higher affinity molecules exhibit substantially improved in vitro retention on HER2/neu expressing target cells in a manner consistent with their K_D and k_off values. The in vivo tumor-targeting properties of these affinity mutants have been examined in scid mice bearing HER2/neu-expressing human tumor xenografts. Molecules with affinities above 10^-7 M exhibit improved selective tumor targeting properties, but the production of very high affinity molecules actually impedes in vivo intra-tumor retention in scid mice. Immuno-histochemical analysis of tumor samples from mice treated with these molecules shows a loss of intratumoral diffusion from blood vessels with rising scFv affinity. These results confirm the "binding site barrier" hypothesis and show that extremely high affinity antibodies should not be used to achieve efficient in vivo tumor targeting.

Dimeric scFv molecules target tumors more efficiently than do monomeric scFv, particularly when the recombinant dimeric sFv are prepared as fusion proteins joined by short linkers (e.g., diabodies). Diabodies exhibit exceptionally prolonged cell retention and selective in vivo tumor uptake. The cumulative tumor and normal organ radioactivity retention of radioiodinated diabody targeting HER2/neu approaches the specific tumor retention properties needed for effective radioimmunotherapy of solid tumors. For example, tumor:bone marrow ratios for cumulative retention approximate 3:1 for the anti-HER2/neu C6.5 diabody. Thus, it should prove possible to deliver 3600 cGy of radiation to tumor sites, while exposing the bone marrow (which is thought to comprise 25% of all blood activity) to no more than 300 cGy. This is an acceptable bone marrow dose; however, tumor:blood ratios must be improved to at least 4:1 to fulfill criteria that we have established for proceeding to clinical trials development of this treatment strategy. Interestingly, modulating the affinity of the scFv components of diabodies from 10^-7 to 10^-10 M has little effect on the in vivo tumor targeting properties of these molecules.

Phase I Clinical Trial of Repeated, Pharmacodynamically-based Dosing of the SEA-based Immunoconjugate PNU-214565. PNU-214565 is a recombinant E.coli-derived fusion protein of the bacterial superantigen Staphylococcal enterotoxin A (SEA) and the Fab-fragment of the C242 monoclonal antibody that recognizes Lewis^A, a glycoprotein over-expressed in colon and pancreatic carcinoma cells. SEA is a potent T cell-specific mitogen, and can activate a large proportion of T cells following crosslinking of T-cell receptor (TCR) Vb regions with major histocompatibility complex (MHC) class II molecules. This novel immunoconjugate promotes MHC class II-independent lysis of human colon carcinoma cells. Prior Phase I clinical trials have shown that the probability of patient toxicity is determined by the relationship of the administered dose to both the weight of the patient and the molar concentrations of circulating antibodies directed against SEA. Using this information, it is possible to assign a dose with a known probability of inducing systemic immunologic activation and host toxicity. In a completed Phase I clinical trial of this pharmacodynamically-based dosing scheme; the aim was to identify the maximum weight- and anti-SEA-adjusted dose that could be safely administered to patients. This dose range was identified for patients with low levels of circulating anti-SEA antibodies; however, unpredictable and severe toxicities severely limited the doses of intact SEA-containing fusion protein that could be safely administered. Patients with advanced lung and breast cancers are now being treated with a second-generation molecule in two new Phase I clinical trials. The new molecule, designated PNU-214936, contains a mutant SEA that binds to class II MHC molecules with much lower affinity than the parent molecule. This should decrease undesirable systemic leukocyte activation arising from the fusion protein-induced crosslinking of TCR Vb chains and class II MHC expressing cells in the peripheral blood. In an alternate strategy, we also are developing truncated scFv-SEA proteins aimed at delivering incomplete and non-functional SEA proteins for antibody-directed assembly of intact, functional superantigen at tumor sites.

Clinical Trials of 2B1 Bispecific Antibody. E3194 is an Eastern Cooperative Oncology Group (ECOG) Phase IB/II clinical trial of 2B1 therapy in women with metastatic breast cancer. This completed trial did not demonstrate the efficacy of 2B1 therapy in this patient group. The safe dose for this patient group was only 1.0 mg/m² per dose, which is too low to achieve reliable in vivo tumor targeting. However, we continue to characterize the effects of such therapy on anti-HER2/neu immune responses in blood samples obtained from these patients, since the property of FcgR-targeted immunization is intriguing and requires further study. We also are completing a Phase I trial of 2B1 in conjunction with GM-CSF and IL-2 therapy for patients with adenocarcinomas. The purpose of this ongoing trial is to test the hypothesis that GM-CSF will promote 2B1-directed antigen presentation of HER2/neu, and that subsequent IL-2 therapy will expand the anti- HER2/neu T cell response.

Other Clinical Trials. Other clinical trials conducted in the current period include Phase II trials of antibody pre-targeted radioimmunotherapy employing 90-Yttrium in patients with metastatic colorectal carcinoma and renal carcinoma, and radioimmunotherapy of patients with advanced non-Hodgkin's lymphoma employing a 90-Yttrium labeled anti-CD20 antibody. We also are conducting a tumor vaccine trial employing a recombinant canarypox virus (ALVAC) vector that has been modified to express the carcinoembryonic antigen (CEA) and B7.1 genes. Forty patients with CEA expressing malignancies have been treated with this vaccine, and an additional cohort of patients is being treated with the same vector plus subcutaneous GM-CSF to enhance antigen presentation and induce clinically useful anti-CEA cellular immune responses.

ADAMS, G.P., SCHIER, R., MARSHALL, K., WOLF, E.J., MCCALL, A.M., MARKS, J.D., WEINER, L.M. Increased affinity leads to improved selective tumor delivery of single chain Fv antibodies. Cancer Res. 58:485-490, 1998.

ADAMS, G.P., SCHIER, R. Generating and evaluating anti-HER2/neu single chain Fv fragments. In Methods in Molecular Medicine: Ovarian Cancer, edited by J. Bartlett. The Humana Press, Totowa, New Jersey (in press).

ALPAUGH, R.K., SCHULTZ, J., MCALEER, C., GIANTONIO, B.J., PERSSON, R., BURNITE, M., NIELSEN, S.E., VITEK, L., PERSSON, B., WEINER, L.M. Superantigen-targeted therapy: Phase I escalating repeat dose trial of the fusion protein PNU-214565 in patients with advanced gastrointestinal malignancies. Clin. Cancer Res. 4:1903-1914, 1998.

ALPAUGH, R.K., VON MEHREN, M., PALAZZO, I., ATKINS, M.B., SPARANO, J., SCHUCHTER, L., DUTCHER, J.P., WEINER, L.M. Phase IB trial for malignant melanoma using R24 monoclonal antibody, interleukin-2/a-interferon. Med. Oncol. 15:191-198, 1998.

KELLER, S.M., RYAN, L.M., COIA, L.R., DANG, P., VAUGHN, D.J., DIGGS, C., WEINER, L.M., BENSON, A.B. High dose chemoradiation followed by esophagectomy for adenocarcinoma of the esophagus and gastroesophageal junction: Results of a phase II study of the Eastern Cooperative Oncology Group. Cancer (in press).

MOVSAS, B., HANLON, A.L., LANCIANO, R., SCHER, R.M., WEINER, L.M., SIGURDSON, E.R., HOFFMAN, J.P., EISENBERG, B.L., COOPER, H.S., PROVINS, S., COIA, L.R. Phase I dose escalating trial of hyperfractionated pre-operative chemoradiation for locally advanced rectal cancer. Int. J. Radiat. Oncol. Biol. Phys. 42:43-50, 1998.

WEINER LM. Paclitaxel in the treatment of esophageal cancer. Semin. Oncol. (in press).

ADAMS, G.P. Improving the tumor specificity and retention of antibody-based molecules. In Vivo 12:11-22, 1998.

ADAMS, G.P., SCHIER, R., MCCALL, A.M., CRAWFORD, R.S., WOLF, E.J., WEINER, L.M., MARKS, J.D. Prolonged in vivo tumor retention of a human diabody targeting the extracellular domain of human HER2/neu. Brit. J. Cancer 77:1405-1412, 1998.

ALPAUGH, R.K., WEINER, L.M., PERSSON, R., PERSSON, B. Overview of clinical trials employing antibody-targeted superantigens. Adv. Drug Delivery Rev. 31(1,2):143-152, 1998.

McCALL, A.M., AMOROSO, A.R., SAUTES, C., MARKS, J.D., WEINER, L.M. Characterization of anti-mouse FcgRII single-chain Fv fragments derived from human phage display libraries. Immunotechnology, 4:71-87, 1998.

McCALL, A.M., ADAMS, G.P., WEINER, L.M. Antibody-targeted gene therapy. In Gene Therapy of Cancer, edited by E. Lattime, S. Gerson. Academic Press, San Diego, CA, vol. 7, pp. 113-124, 1998.

§   Fox Chase researcher

^a   A. McCall: Present address--Murdoch Institute, Royal Children's Hospital, Victoria, Australia

^b   L. Houston: 327 Pine Needles Dr., Del Mar, CA 92014

^c   J. Huston: Matritech, 330 Nevada St., Newton, MA 02160

^d   J. Marks: University of California, San Francisco General Hospital, San Francisco, CA 94110

^e   M. Kopreski, B. Persson: Pharmacia Oncology/Immunology, Dept. of Clinical Research, S-220 07 Lund, Sweden

^f   I. Horak: Janssen Research Foundation, Titusville, NJ 08560

^g   J. Schlom: National Cancer Institute, Bethesda, MD 20892

Illustrations or unpublished data in these reports should not be used without permission of the author.

Fox Chase Cancer Center

Scientific Report 1998