We present a system for cancer targeting based on single-chain Fv (scFv) antibodies selected from combinatorial libraries, produced in bacteria and purified by using an engineered tag. Combinatorial libraries of scFv genes contain great diversity, and scFv antibodies with characteristics optimized for a particular task can be selected from them using filamentous bacteriophage. We illustrate the benefits of this system by imaging patients with carcinoembryonic antigen (CEA)-producing cancers using an iodine-123 labeled scFv anti-CEA selected for high affinity. All known tumor deposits were located, and advantages over current imaging technology are illustrated. ScFvs are produced in a cloned form and can be readily engineered to have localizing and therapeutic functions that will be applicable in cancer and other diseases.

Antibody-directed enzyme prodrug therapy is a targeted therapy in which a prodrug is activated selectively at the tumour site by an enzyme, which has been targeted to the tumour by an antibody (antibody-enzyme conjugate). Previous clinical trials have shown evidence of tumour response, however, the activated drug had a long half-life, which resulted in dose-limiting myelosuppression. Also, the targeting system, although giving high tumour to blood ratios of antibody-enzyme conjugate (10 000 : 1) required administration of a clearing antibody in addition to the antibody-enzyme conjugate. The purpose of this current study therefore was to attempt tumour targeting of the antibody-enzyme conjugate without the clearing antibody, and to investigate a new prodrug (bis-iodo phenol mustard, ZD2767P) whose activated form is highly potent and has a short halflife. Twenty-seven patients were treated with antibody-directed enzyme prodrug therapy using A5CP antibody-enzyme conjugate and ZD2767P prodrug, in a dose-escalating phase I trial. The maximum tolerated dose of ZD2767P was reached at 15.5 mg m 72 6three administrations with a serum carboxypeptidase G2 level of 0.05 U ml 71 . Myelosuppression limited dose escalation. Other toxicities were mild. Patients' quality of life was not adversely affected during the trial as assessed by the measures used. There were no clinical or radiological responses seen in the study, but three patients had stable disease at day 56. Human anti-mouse antibody and human anti-carboxypeptidase G2 antibody were produced in response to the antibody enzyme conjugate (A5CP). The antibody-enzyme conjugate localisation data (carboxypeptidase G2 enzyme levels by HPLC on tumour and normal tissue samples, and gamma camera analysis of I-131 radiolabelled conjugate) are consistent with inadequate tumour localisation (median tumour: normal tissue ratios of antibody-enzyme conjugate of less than 1). A clearance system is therefore desirable with this antibody-enzyme conjugate or a more efficient targeting system is required. ZD2767P was shown to clear rapidly from the circulation and activated drug was not measurable in the blood. ZD2767P has potential for use in future antibody-directed enzyme prodrug therapy systems.