Idiosyncratic drug toxicity (IDT) is a rare toxic drug reaction characterized by the delayed onset of symptoms, a dose- and duration-independent occurrence, and an unpredictable nature. It is often life-threatening and, thus, one of the major reasons for drug withdrawals or black box warnings Clinical investigations have proposed that the mechanisms of most IDTs are immune-mediated and there is a very strong association with specific human leukocyte antigen (HLA) genes for certain reactions [1
) is an orally-available anticoagulant agent that directly inhibits thrombin and was developed for the prevention and treatment of thromboembolism. Although ximelagatran was generally well tolerated in short-term use (<12 days), an elevated serum alanine aminotransferase (ALT) level of >3× upper limit of normal was found to develop during long-term treatment (>35 days) in 7.9% of patients [2
]. Due to the concerns for its potential liver toxicity, the development of ximelagatran was terminated and it was withdrawn from the world market in 2006. A retrospective case-control pharmacogenetic study of ALT elevation during long-term treatment of ximelagatran revealed a strong genetic association between elevated ALT and the HLA alleles DRB1*07 and DQA1*02, suggesting a possible immune-mediated pathogenesis [3
]. Furthermore, a competitive binding assay revealed that ximelagatran was able to inhibit the binding of the ligand peptide to HLA-DRB1*07:01, supporting the specific involvement of HLA-DRB1*07:01 in the idiosyncratic ximelagatran-induced hepatotoxicity [3
In recent years, the mechanisms of immune stimulation by small molecules without involving covalent binding to macromolecules, such as the “pharmacological interaction with immune receptors” [4
], and “altered self-repertoire” concepts [6
], have been reported. There also have been several reports showing that small molecules can modify the antigen loading onto HLA molecules [5
], and the interaction modes of some compounds with HLA molecules have been evaluated by in silico docking simulation [5
], MD simulation [9
], or X-ray crystalline structure analysis [6
]. We previously reported that in vitro lapatinib enhances the binding of the ligand peptide to HLA-DRB1*07:01, which is strongly associated with lapatinib-induced liver injury, and MD simulations indicated the allele specific modification of the structure of the peptide binding groove of HLA-DRB1*07:01 by lapatinib [10
]. Ximelagatran also showed idiosyncratic hepatotoxicity in clinical settings, which was strongly associated with the same HLA allele, DRB1*07:01. Interestingly, the same in vitro HLA class II binding assay conducted by EpiVax, Inc. (Providence, RI, USA), resulted in the opposite effects of two drugs on the binding of the same ligand peptide to HLA-DRB1*07:01 [3
]. Therefore, in the present study, we evaluated the possible interaction modes of ximelagatran and its active metabolite, melagatran (Figure 1
) with three HLA-DR molecules in silico to compare them with those of lapatinib. Furthermore, the amount of ximelagatran bound to HLA-DR molecules when it showed the inhibitory effect was evaluated in vitro by liquid chromatography tandem mass spectrometry (LC-MS/MS).
In order to evaluate the possible direct interaction of ximelagatran with the peptide binding groove of HLA-DRB1*07:01, a series of in silico simulations and in vitro measurement of ximelagatran bound to HLA-DR molecules were performed.
The first docking studies indicated that ximelagatran and its active metabolite, melagatran, have high potential to interact with the peptide binding groove of HLA molecules compared with other IDT causing drugs, nevirapine [12
] and allopurinol [13
]. Ximelagatran showed higher binding affinities than melagatran to all three HLA-DR molecules. Considering that 40% to 70% of the oral dose of ximelagatran was absorbed in humans [14
], the liver would be exposed to a high concentration of ximelagatran after oral administration, although it is rapidly metabolized (t1/2
= 0.34 h after oral administration) in humans [14
]. Therefore, not melagatran (the predominant compound in human plasma), but rather ximelagatran is likely to be the primary compound responsible for ximelagatran-induced hepatotoxicity.
Additionally, MD simulations were conducted in order to see the possible effects of ximelagatran on the structure of the peptide binding groove of HLA-DRB1*07:01 and the conformation of bound ligand peptide, and to compare them with those of AdCaPy (one of the major histocompatibility complex loading enhancers [9
]) and lapatinib [10
]. The interaction modes of ximelagatran in the absence and presence of the ligand peptide were similar to those found in the docking study. Ximelagatran binds along the center of the peptide binding groove of HLA-DRB1*07:01 and makes its deepest contact in or near the P4 pocket. In the absence of the ligand peptide, ximelagatran keeps the peptide binding groove in an open state (Figure 4
d and Figure 5
a) similarly to AdCaPy, which enhances the loading efficiency of ligand peptides onto several HLA-DR molecules [9
]. However, unlike AdCaPy, which is small and interacts with the P1 pocket of HLA-DR molecules, ximelagatran interacts along a relatively large part of the peptide binding groove of HLA-DRB1*07:01. As a result, the binding of the N- and C-terminal portions of the ligand peptide appears to be disrupted and, therefore, it does not seem to find a stable conformation on top of ximelagatran (Figure 5
e). These results are consistent with the fact that ximelagatran functioned as a peptide competitor in vitro [3
]. Although lapatinib is a much larger molecule and also interacts along a large part (P1–P6 pockets) of the peptide binding groove of HLA-DRB1*07:01, it could induce a tightly-closed binding groove structure that might be able to stabilize the binding of ligand peptides even in an irregular fashion [10
]. MD simulations of ximelagatran and lapatinib indicated that small molecules can alter the antigen loading by affecting the conformation of bound peptides [6
], as well as the structure of the peptide binding groove [10
Finally, ximelagatran bound to three HLA-DR molecules was detected by LC-MS/MS in the absence of the ligand peptides (Table 3
). This result was consistent with the docking study result in Table 1
, which indicates the high potential of ximelagatran to interact with the peptide binding groove of these HLA-DR molecules. In contrast, this result seemingly sounds contradictory to the inhibitory effect of ximelagatran in vitro, which is HLA-DRB1*07:01 allele specific [3
]. One possible reason might be seen in the comparison of the concentration of ximelagatran in the presence of HLA-DR molecules and the ligand peptides with that found in the presence of HLA-DR molecules and absence of the ligand peptides, in which the extent of the decrease of ximelagatran bound to the HLA-DR molecules by the addition of the ligand peptides was different between the three HLA-DR alleles. According to these results, one hypothesis could be assumed that ximelagatran interacts with a relatively broad range of HLA-DR alleles, but its affinity is not enough to affect the binding of their ligand peptides, and it is driven out from the peptide binding groove by the ligand peptides, except for in the case of HLA-DRB1*07:01. Ximelagatran was indeed simulated to bind to the peptide binding groove of each HLA-DR in a different manner (Figure 2
), thus, it would be reasonable that its effects on the binding of ligand peptides are also different depending on the HLA alleles. For HLA-DRB1*07:01, the MD simulation indicated the possible disruption of the binding of the ligand peptide (Figure 5
e), which is consistent with the in vitro results [3
]. Similar MD simulations or docking studies with the ligand peptides for HLA-DRB1*01:01 and DRB1*15:01 are subjects for future study. Furthermore, this hypothesis might also be supported by the future optimization of the LC-MS/MS assay system whose current sensitivity might not be enough to detect a tiny amount of ximelagatran bound to HLA-DR molecules clearly. If the sensitivity of this assay system were optimized enough, it would be useful as a new IDT risk assessment system.
Taken together, these results indicate that ximelagatran directly interacts with the peptide binding groove of HLA-DRB1*07:01 and competes with ligand peptides for this binding site. In the cases of abacavir for HLA-B*57:01 [6
], and lapatinib for HLA-DRB1*07:01 [10
], in both of which drugs increase the loading of ligand peptides, it is easy to imagine the immune stimulation by the enhanced antigen presentation and/or the presentation of neo-antigens. Unlike these cases, it may be somewhat difficult to associate the inhibited antigen presentation by ximelagatran with the immune stimulation. However, considering the simulated binding mode of ximelagatran at the peptide binding groove of HLA-DRB1*07:01, there is a possibility that there are some kinds of peptides that are loaded onto HLA-DRB1*07:01 de novo and/or in a different manner in the presence of ximelagatran. Although further investigations are needed to reveal the conclusive mechanism, the present study strongly supports the interaction between ximelagatran and the peptide binding groove of HLA-DRB1*07:01, which could alter the immune response in some way and lead to the idiosyncratic ximelagatran-induced hepatotoxicity. In vitro studies using hepatocyte co-cultured with macrophages or lymphocytes carrying various HLA alleles will support our hypothesis and give us additional information to understand the detailed mechanism of ximelagatran-induced hepatotoxicity.