In their recent report Jørgensen et al1 raised doubts on the ability of α1-acid glycoprotein (AGP) to bind and inhibit imatinib (STI571), as shown in our previous report.2 We would like to comment on this paper, on some methodological inaccuracies of their paper, and on additional in vivo data that in our opinion strongly indicate an important role for AGP in modulating imatinib bioavailability and pharmacokinetics (PK).
First, it is well known that chromatographically isolated AGP, the one used by Jørgensen et al, show less-efficient binding of drugs in general than chemically isolated AGP, the one used in our paper.3 It is surprising in this respect that Jørgensen et al never used as a control our preparation of AGP.
Second, in their paper the authors state that our AGP preparation, supplied by Sigma, “risks desialylation of the protein.” 1(p714) But the authors fail to acknowledge that such a phenomenon has been associated with a decrease(or with no change at all) in drug binding,3-5 and not with an increase in binding, as their data apparently suggest.
Third, the drug-binding assay shown is misleading. Quenching of AGP fluorescence requires detailed information on a given drug's binding site to AGP, since several binding sites for drugs on AGP are known3; this information was not provided for imatinib. In addition, quenching should be shown using progressively increasing concentrations of the drugs being studied and not, as done by Jørgensen et al, by comparing 2 different drugs, used at a single concentration, which differed in the 2 drugs studied (imatinib at 1 μM, chlorpromazine at 2.5 μM).
Fourth, in vitro experiments using unmanipulated AGP (in the form of sera containing different concentrations of AGP) performed by 2 independent groups2,6 show that the inhibition on imatinib activity is proportional to the content of AGP and can be blocked by the coincubation with erythromycin, a known binder of AGP.
Fifth, additional ex vivo experiments were performed using unseparated blood samples from patients on treatment with imatinib and clinically resistant to it. Although plasma levels of imatinib exceeded 3 μM levels in these patients, Bcr/Abl was highly phosphorylated; short-term incubation (1 h) with erythromycin resulted in almost total (> 85%) phosphorylation inhibition.7
Sixth, in vivo studies in patients treated with imatinib show that there is a significant correlation between AGP levels and some PK data such as Cmax.8 In addition, the coadministration of imatinib and clindamycin, another antibiotic known to bind AGP, resulted in significantly reduced Cmax and AUC and in increased free fraction of imatinib; in particular, clindamycin induced within 5 minutes a fall in plasma imatinib concentrations ranging from 2-fold to 5-fold (Gambacorti-Passerini et al9 and Gambacorti-Passerini, April 9, 2002, manuscript submitted for publication).
For the above-mentioned reasons, the data from Jørgensen et al are difficult to evaluate and their in vivo relevance is questionable.
Further observations on the debated ability of AGP to bind imatinib
We thank Gambacorti-Passerini et al for their comments on our paper, which we note with interest. We are indeed encouraged by the debate provoked by the publication of our brief report1-1and welcome objective discussion from scientific colleagues. But we feel that some of the points made by Gambacorti-Passerini et al potentially arise from a lack of appreciation of the methods utilized for the purification and characterization of glycoproteins, the importance of glycosylation as a secondary modification of proteins, and its implication for drug binding. Our responses to the specific points raised are as follows:
First, historically the majority of techniques for the isolation of human α1-acid glycoprotein (AGP) were chromatographic procedures with strongly acidic buffer. Indeed, the commercial AGP product assayed by Gambacorti-Passerini et al1-2 was isolated according to the process of Hao and Wickerhauser,1-3 which is a combination diethyl-amino ethyl (DEAE)-Sephadex/carboxymethyl (CM)-cellulose chromatographic method at pH 4.7, not a chemical method as stated by Gambacorti-Passerini et al. Thus, both the commercial product and our AGP are chromatographically isolated. Review of the literature, including Kremer et al,1-4 indicates that acidic isolation methods will damage AGP oligosaccharide (principally by desialylation) and polypeptide components. Thus we are satisfied that, by avoiding harsh acidic conditions, our published purification method yields AGP without any structural degradation1-5 and thus gives a valid representation of the actual in vivo presentation of the glycoprotein. Furthermore, the fluorescence data (see the third point below) presented indicates that chlorpromazine, a known AGP binder, effectively binds to our isolated AGP, which is not in keeping with “less-efficient binding.”
As clearly presented in our paper, the main aim was to examine AGP in the CML setting, as it is well documented that the glycoprotein alters both quantitatively and qualitatively in disease. Thus, commercial AGP isolated from normal plasma would not satisfy this requirement. Nonetheless, the most logical approach was to isolate AGP from normal plasma as a control by the same method as for CML-derived AGP.
Second, glycosylation, in the form of oligosaccharide chains covalently bound to protein, is a significant presence on the surface of glycoproteins, such as AGP, and functions to affect the conformation of the underlying polypeptide largely owing to the huge hydrodynamic volume occupied relative to amino acids. Thus, the presence of a particular glycosylation pattern may influence the protein conformation and thus the degree of access to the drug-binding site. Any change in glycosylation, including the removal of sialic acid, may result in a conformational rearrangement that could conceivably increase, decrease, or leave unaltered the access to the drug-binding site. Our isolation method does not involve denaturing steps, such as preliminary acidic preparation and/or exposure to strongly acidic buffers during chromatography, and has been proven not to result in structural degradation. The quoted phrase was included to emphasize that the latter could not be used to explain our results. In other words, our observed lack of binding of imatinib to CML-derived AGP is not an artifact of processing, but rather reflects more truly the nature of the in vivo interaction. We do not read pretense of increased binding into our data.
Third, the fluorescence-quenching experiment, a method utilized to directly study AGP drug binding,1-6 was simply employed to demonstrate the retention of drug-binding potential by our purified glycoprotein, which was amply shown with the known AGP-binder, chlorpromazine. An inability to bind the control substance, chlorpromazine, would have been indicative of loss of AGP structural integrity, which clearly had not been induced by our purification processing. As this assay revealed no interaction between CML-derived AGP and imatinib in the face of proven chlorpromazine binding, we feel we cannot comment further on the nature of the purported imatinib-binding site on AGP. It is not accepted that chlorpromazine and imatinib should be tested at identical concentrations providing clinically relevant concentrations are chosen. (A correlation was observed between quenching and chlorpromazine concentration up to 250 μM; data not shown).
Fourth, we would consider the evidence presented suggesting a stoichiometric interaction specifically between AGP and erythromycin in whole sera to be circumstantial and difficult to evaluate. For this reason, we deliberately isolated AGP from contaminating, nonspecific plasma-protein drug binders. Additionally, can 2 groups be described as independent when they have shared membership?
Regarding the fifth and sixth points, we feel unable to comment on unpublished manuscripts but look forward to scrutinizing the data once in print. We do, however, wish to reiterate the findings of Gorre et al,1-7 exponents of the “cell intrinsic” theory of resistance, which our data supports: (a) cells taken from relapsing patients exhibited reduced sensitivity to imatinib compared with pretreatment cells; (b) relapsing patients did not have significantly reduced imatinib plasma concentrations,1-8 despite a presumed concomitant increase in plasma AGP concentration with disease progression mirroring the findings in our cohort of patients1(Fig1); and (c) dose escalation has not proven successful in inducing remissions in relapsing, resistant patients (our observations, and Gorre et al1-9).
We trust that our responses will facilitate further evaluation of our data.