Dragons ’ Round the Fleece Again : STI 571 Versus 1 Acid Glycoprotein

Ovid recounts the story of Jason and his Argonauts (1), who, after dealing with jealous harpies, wandering rocks, spectral armies, and fire-snorting bulls, approach the Golden Fleece only to find it guarded by a sleepless dragon who complicates the ease with which their prize might be grasped. In the current age, we hope to revolutionize cancer therapeutics by defining drug molecules that target the pathophysiologic basis of cancer: activated oncogenes, deregulated tumor suppressor genes, and the ubiquitously relevant processes of angiogenesis and apoptosis. Drugs bound specifically to these or their related targets are the Fleeces of our quest. STI571 is a selective p210 oncogene tyrosine kinase inhibitor (2). BCR-ABL-related molecules are etiologic for adult chronic myeloid leukemia (CML) (3) and a fraction of acute lymphoblastic leukemias. Therefore, STI571 treatment of BCR-ABL-expressing neoplasms is prototypic in a way that we hope to emulate by use of target-directed drugs in other diseases that are driven by other oncogene-derived kinases. Noteworthy features in STI571’s preclinical development, presented previously in these pages (4), were the curative activity in certain animal models of CML and the lack of activity in tumor models not dependent on this target kinase. Initial clinical returns with the agent have been concordant with those expectations, with evidence of clinical activity in patients with chronic-phase CML even in early phase I testing (5) and with little evidence of serious toxicity at dose levels where benefit has been seen. A few clouds have appeared on this otherwise rather sunny horizon. In initial clinical testing, patients with more advanced blast-phase CML, in an as yet not rigorously quantitated sense, seem to have fewer responses of shorter duration, in comparison to patients with chronic-phase disease (6). Gambacorti-Passerini et al. (7) undertook further animal model experiments, reported in this issue of the Journal, to address the potentially related phenomenon that mice with a large CML tumor burden had greater resistance to STI571 than mice with a smaller tumor burden. Surprisingly, in vitro, CML cells explanted from the larger, resistant tumors were as sensitive to STI571 as parental cells. In contrast, cells exposed in vivo in large tumors did not show inhibition of growth or p210 kinase activity. This result suggested a problem in target “access” by drug in vivo in animals with large, resistant tumors. The culprit was found to be the plasma 1 acid glycoprotein (AGP): Mice with larger tumors had increased AGP concentrations; exogenously added AGP could confer resistance to STI571 in vitro; and erythromycin, a drug that also binds AGP, could restore sensitivity to STI571 in the previously resistant, in vivo, large tumor models presumably by displacing STI571 from AGP. The striking result in this study is that drug resistance occurs on an organismic rather than a cellular basis, in a way that varies with tumor burden. AGP as a modulator of drug effects in humans is certainly not news (8). AGP, also known as orosomucoid, is a plasma protein that is an acute-phase reactant. It is produced by the liver in response to inflammatory cytokines, stress, and other still poorly defined stimuli, including host neoplasms. It is distinctive in that approximately 40% by weight of each molecule is carbohydrate, including sialic acid. The latter negative charges are thought to contribute to its propensity to bind a number of cationic drugs, including propranolol, haloperidol, methadone, tricyclic antidepressants, among others. A capacity to bind acidic and neutral drugs by distinct binding modes has been described previously (9). In humans, there are two AGP genes, AGP1 and AGP2, thought to have arisen by a gene duplication, and a third AGP gene identical to AGP2 is present in some individuals (10). The existence of polymorphisms in the AGP gene sequence, along with variable states of carbohydrate addition and processing, means that there is really not one AGP species circulating in plasma but rather a population of molecules, all likely able to bind candidate drugs with various affinities around some population average and to have various concentrations owing to the state of a patient’s disease and “underlying” condition. AGP has also long been proposed as a modulator of drug clinical effect in a variety of settings [e.g., (11)]. Recently, AGP came to the fore in cancer developmental therapeutics because it avidly binds staurosporine-related compounds UCN-01 and PKC412, interestingly, in a species-dependent fashion. Mouse AGP binds UCN-01 far less avidly than does human AGP (12,13). This difference changes the handling of UCN-01, from a drug with terminal half-lives in various animal species of 5–12 hours to a drug in humans that persists for hundreds to thousands of hours after a single dose. This behavior renders much of the animal model work on which the development of UCN-01 and PKC412 was predicated to be of questionable relevance to humans. In addition (and as pointed out by Gambacorti-Passerini et al. in this issue of the Journal), the predictive power of animal toxicology may diminish because human toxicity may emerge only after drug has bound to all circulating AGP. At the very least, one must now relate clinical phenomena to levels of free drug and not simply to levels of total drug. One must also be concerned that the patient population one studies is normalized with respect to AGP concentration. Although the clinical impact of the present results of Gambacorti-Passerini et al. (7) on humans remains to be clarified, STI571’s use in blast-phase CML patients must take into account that these patients are, on average, more ill than patients with chronic-phase disease. Thus, they may have higher AGP concentrations than chronic-phase