Summary Studies with a conjugate of carboxypeptidase G2 (CPG2) and the F(ab')2 fragment of monoclonal anti-CEA antibody, A5B7, have shown specific localisation in a human colon tumour xenograft, LS174T, growing in nude mice. The conjugate reaches a peak concentration in the tumour within 24 h but enzyme activity in blood remains above a critical value for therapeutic purposes for several days. Here we describe a new monoclonal antibody, SB43, raised against CPG2 which is capable of reducing enzyme activity in blood. In vitro studies demonstrated specific binding of SB43 to CPG2 causing inactivation. Moreover, in the nude mouse model SB43 was also capable of inactivating the enzyme in the circulation within minutes of administration. Radiolabelled native SB43 persisted in blood for several days and appreciable non-specific uptake into the xenograft was also observed. Uptake of SB43 by the tumour, with possible inactivation of CPG2 at this site, could be limited by first coupling the antibody to galactose. This ensured recognition and excretion of SB43 and SB43-enzyme complexes via the liver and their rapid removal from the circulation. Galactosylation had no effect on the ability of SB43 to inactivate the enzyme.Anti-cancer agents such as drugs, toxins and radioisotopes have been linked directly to antibodies to achieve greater anti-tumour selectivity but this approach has so far proved to be of restricted therapeutic value. Major limitations resulting from the use of directly coupled antibodies are the amount of anti-cancer agent that can be attached to the antibody molecule without loss of immunological activity and the amount of conjugate that can be specifically targeted to tumour. A novel approach of converting a prodrug into a potent anticancer agent by a targeted enzyme (ADEPT) aims to overcome these difficulties. As originally envisaged (Bagshawe, 1987) this system employed two phases in which an enzyme conjugated to an anti-tumour antibody was first allowed to localise and clear from the circulation before injection of the prodrug. Our preliminary studies (Bagshawe et al., 1988) were carried out with carboxypeptidase G2 (CPG2), a folate depleting bacterial enzyme, conjugated to the F(ab')2 fragment of a monoclonal anti-hCG antibody and capable of converting a glutamyl benzoic acid mustard prodrug into a potent benzoic acid mustard. In the choriocarcinoma xenograft system tested, the target antigen, hCG, was present in the plasma and clearance of the antibody-enzyme conjugate from the circulation occurred within 72 h allowing administration of the prodrug and elimination of the tumour. A similar approach, in which placental alkaline phosphatase was used to convert the prodrug etoposide phosphate into etoposide, was shown to be effective in a colon xenograft model (Senter et al., 1988). We have now extended our studies to the human colonic carcinoma xenograft model (LS174T) in which we have targeted CPG2 conjugated to the monoclonal antibody fragment, A5B7-F(ab')2, directed at carcinoembryonic antigen (CEA...