Classification of mechanisms and role of reactive metabolites
In view of the different chemical and pharmacological properties of the causative agents, and the heterogeneity of the clinical presentations, it is not surprising that idiosyncratic reactions involve a broad range of mechanisms, and more than one mechanism may be involved in any single event. A classification which has didactic value, but may not be always easily applicable to individual cases, distinguishes three broad mechanisms: (1) direct cytotoxicity, whereby a drug or a metabolite cause direct cellular damage; (2) immune-mediated hypersensitivity reactions; (3) off-target pharmacology, whereby a drug or a metabolite interact directly with a system other than that for which the drug is intended.
Many idiosyncratic reactions are initiated by reactive drug metabolites, which bind covalently to macromolecules and either cause direct cell damage or trigger an immune response (Guengerich, 2006). Reactive metabolites often have a very short half life, which explains why their sites of formation, and the liver in particular, are also the major targets of tissue damage (Ju and Uetrecht, 2002). Extrahepatic damage may be seen when reactive metabolites are formed at multiple sites, when long-lived metabolites travel from the liver to other organs, or when a locally initiated immune response spreads systemically.
There are many examples of AEDs producing idiosyncratic reactions via formation of toxic metabolites. For instance, the ability of CBZ to cause liver toxicity, blood dyscrasias, skin reactions and multiorgan hypersensitivity syndromes seems to be related, at least in some cases, to reactive metabolites such as carbamazepine-2,3-epoxide (Madden et al., 1996) and an iminoquinone which is sufficiently long-lived to be detectable in the urine of patients treated with this drug (Ju and Uetrecht, 1999). Covalent binding of CBZ metabolites to proteins has been observed in vitro using both liver microsomal and myeloperoxidase activation systems (Naisbitt et al., 2003b), whereas in vivo most of the reactive epoxides are detoxified to dihydrodiols by microsomal epoxide hydrolase 1 or to glutathione conjugates by glutathione transferase (Lillibridge et al., 1996). A reactive arene oxide intermediate is also known to be formed during the conversion of PHT to its primary para-hydroxy-phenyl-metabolite (p-HPPH): although this intermediate has never been isolated from plasma or urine, presumably because it is too unstable, it is considered to be involved in PHT-induced idiosyncratic reactions affecting the liver, the blood and other organs (Browne and Leduc, 2002). Similar reactive intermediates produced by cytochrome P450 (CYP) enzymes may play a role in hypersensitivity reactions associated with PB (Knowles et al., 2000) and lamotrigine (LTG) (Maggs et al., 2000; Schaub and Bircher, 2000). In the case of LTG, most of the drug is cleared by glucuronide conjugation and only minor amounts are converted by CYP enzymes to an arene oxide intermediate. Since valproic acid (VPA) inhibits LTG glucuronidation, in patients comedicated with VPA a higher percentage of the LTG dose is converted through the alternative CYP-mediated pathway to the oxide intermediate, which may explain the greater susceptibility of these patients to LTG-induced skin rashes (Anderson, 2002). As discussed in the next section, the hepatotoxicity of VPA and felbamate (FBM) is also related to formation of toxic metabolites.
Variability in the rate of formation and detoxification of reactive metabolites can explain why some reactions only occur in susceptible individuals (Glauser, 2000). Susceptible individuals may produce excessive amounts of reactive metabolites, for example, as a result of intake of high doses of the drug and/or abnormally high activity of bioactivating enzymes, or they may have impaired cellular defense mechanisms, for example, abnormally low levels of detoxifying enzymes such as epoxide hydrolases or substrates such as glutathione (Johnson, 2003; Lee et al., 2004; Gerber and Pichler, 2006; Guengerich, 2006). Variability in response is also related to the fact that not all covalent binding to macromolecules is pathogenic, and some may even play a protective role by sequestering and/or inactivating reactive species. In particular, covalent binding to serum proteins is less likely to lead to idiosyncratic reactions than binding to membrane proteins (Ju and Uetrecht, 2002; Seguin and Uetrecht, 2003).
Some idiosyncratic reactions appear to be caused by a direct cytotoxic effect of the drug or its metabolites, without pathogenetic involvement of the immune system (Ju and Uetrecht, 2002). As far as AEDs are concerned, the best example of such reactions is probably VPA-induced hepatotoxicity. While a direct role of the parent drug in causing or contributing to liver damage cannot be excluded, there is experimental and clinical evidence for a direct cytotoxic effect of two metabolites, namely 4-en VPA and its ß-oxidation derivative 2,4-dien VPA (Sadeque et al., 1997). The formation of 4-en VPA is largely catalyzed by CYP2C9, whose activity is inducible and is higher in young children (Johnson, 2003), which may explain why the risk of VPA-induced liver toxicity is highest in infants comedicated with enzyme inducing AEDs. 4-en VPA is further metabolized in mitochondria to 2,4-dien VPA (Walgren et al., 2005), which is a reactive species capable of causing inhibition of ß-oxidation and mitochondrial dysfunction.
Considerable evidence indicates that FBM-induced liver and bone marrow toxicity is mediated by the reactive metabolite atropaldehyde (Thompson et al., 1996, 1997). Both atropaldehyde and another FBM metabolite, alcohol carbamate, have been shown to inhibit glutathione transferase and to cause cytotoxicity in human hepatocytes (Kapetanovic et al., 2002). Likewise, FBM metabolites have been shown to form covalent adducts with human serum albumin (Walgren et al., 2005). Since the half-life of the atropaldehyde precursors CPPA (3-carbamoyl-2-phenylpropionic acid) and 4-hydroxy-5-phenyl-(1,3)-oxazinan-2-one is in the order of hours, it has been suggested that these FBM metabolites may travel from the liver and release atropaldehyde to other sites such as the bone marrow (Dieckhaus et al., 2001a; Walgren et al., 2005). Whether immune mechanisms play an important role in the toxicity of FBM metabolites is unclear, but their involvement is suggested by experimental studies on the immunogenic potential of reactive FBM metabolites (Popovic et al., 2004) and by the observation that patients with a history of hypersensitivity reactions and autoimmune disease are at greater risk of developing FBM-induced aplastic anemia (Pellock et al., 2006). Evidence for a key role of reactive metabolites in FBM toxicity provides a rationale for the development of fluorofelbamate, a FBM analogue that is not converted to atropaldehyde and is currently under clinical evaluation as a potentially safer AED (Bialer et al., 2007).
Immune-mediated hypersensitivity reactions, which result from an evolutionary derangement of the main defense mechanisms against infectious agents, involve abnormal humoral- or cell-mediated responses. AEDs may initiate these responses by interacting with cells of adaptive immunity through incompletely understood mechanisms. Fig. 1 summarizes known types of immune-mediated drug reactions, and their main clinical correlates. These reactions can be broadly divided into two classes, namely those involving an interaction of B cells, which are able to recognize antigenic determinants through B cell receptors (BCRs), and those whereby T cells recognize, through T-cell receptors (TCRs), molecules that have been phagocytized, modified and presented by antigen presenting cells (APCs). These two arms of adaptive immunity are cooperative, since B cells can act as APCs or recognize processed antigens, and T cells can act as helper cells towards B cells.
Figure 1. Classification of reactions responsible for immune-mediated hypersensitivity to drugs according to Coombs and Gell (1968), as updated by Pichler (2003). The different subtypes of type IV reactions may yield similar clinical phenotypes. BCR, B-cell receptor; IFN, interferon; IL, interleukin; TCR, T-cell receptor; Th1, T helper type 1 lymphocyte; Th2, T helper type 2 lymphocyte; and CTL, cytotoxic T lymphocyte.
Download figure to PowerPoint
To overcome the limitation of being too small to trigger immune responses, the drug or a metabolite need to behave as haptens (from Greek haptein, to fasten), for example, they have to covalently bind and modify a macromolecule (usually, a self-peptide) to become immunogenic (Landsteiner and Jacobs, 1935; Park, 1998). Alternatively, electrophilic metabolites can react with nucleophilic groups on proteins without covalent binding (Park et al., 1987). The drug-peptide complex, which is recognized as foreign, is thus processed by APCs which, in turn, can trigger B- or T cell-mediated responses.
Type I, II, and III reactions involve activated B cells (plasma cells) which produce antibodies directed against antigenic determinants located on the drug itself or generated by an interaction of the drug (or a reactive metabolite) with macromolecules of the host organism (Coombs and Gell, 1968). In type I reactions, which include anaphylactic reactions (a rare event with AEDs) and some urticarioid skin rashes, an antigen to which the organism is sensitized binds to IgE antibodies at the surface of mast cells and basophils, resulting in their degranulation and release of inflammatory factors. Among the released factors, histamine is responsible for vasodilation and leakage of fluids in the interstitial space, with chemoattraction of T helper 2 (Th2) cells (Bryce et al., 2006) and upregulation of proallergic, Th2-related cytokines. Type II reactions, conversely, include complement-mediated cytotoxic effects triggered by an interaction of IgG and/or IgM antibodies with antigenic determinants at the surface of target cells in the tissue affected by the reaction, such as the blood or the bone marrow (Parr and Doukas, 1999). In Type III reactions (serum sickness-like reactions), the interaction of the antigen with IgG and/or IgM antibodies results in the formation of immunocomplexes, whose accumulation in the affected tissue results in vasculitic changes and tissue damage (Calabrese and Duna, 1996).
Type I to III reactions seem to occur less frequently than previously suspected. In fact, many immune-mediated reactions to AEDs, including the large majority of those affecting the skin, consist in delayed (type IV) hypersensitivity reactions mediated by different T cell subpopulations (Krauss, 2006). Histopathological examination of these skin lesions shows that CD4+ T cells predominate in dermis, and CD8+ T cells in epidermis (Barbaud et al., 1997). Interestingly, these subpopulations include phenotypes, such as T helper 1 (Th1) cells, with predictable protective roles in allergy (Woodfolk, 2006). These observations weaken the “Th1/Th2 paradigm” (Maggi, 1998), in which T cells and related mediators such as interleukins (ILs) and interferons (IFNs) are differentiated into proallergic (Th2 cells, IL-4, and IL-5) and antiallergic (Th1 cells, IFN-γ, and IL-12) subtypes.
The discovery of T cells with specific reactivity for antibiotics (Yawalkar et al., 2000), CBZ (Naisbitt et al., 2003a), and LTG (Naisbitt et al., 2003b) in the blood of patients hypersensitive to the respective drug helped in identifying the mechanisms by which delayed hypersensitivity reactions occur (Fig. 2). Beside the classical model of APC-T cell interaction, which is characterized by a covalent binding between major histocompatibility complex (MHC) molecules and the exposed peptide, followed by priming of naive T cells, an alternative mechanism by which delayed hypersensitivity may occur is outlined by the so-called “p-i concept,” that is, pharmacological interaction with immune receptors (Pichler, 2002). According to this concept, drugs or metabolites can interact first with T cells and then, through noncovalent binding, with APCs, without previous uptake and intracellular processing. In this case the reaction does not involve naive T cells but, instead, crossreactive memory T cells, which can account for allergic reactions without antecedent drug exposure. The “p-i concept” can also explain drug-induced skin reactions that occur a few hours after administration (Christiansen et al., 2000; Gerber and Pichler, 2006) and were previously ascribed to IgE-mediated responses. Experiments on mouse T-cell hybridomas transfected with drug-specific human TCRs seem to confirm the “p-i concept” (Schmid et al., 2006). However, further confirmations from in vivo studies are awaited, especially on the postulated existence of TCRs with double specificity for a drug and a self-peptide. The reason why the skin is the organ most commonly affected by these reactions is unclear, but studies on cutaneous lymphomas suggest that Langerhans cells, which act as APCs in the skin, play a pivotal role in the epidermotropism of lymphocytes (Twersky and Nordlund, 2004). The presence of skin-homing molecules, such as cutaneous lymphocyte antigen (CLA), has been reported in peripheral blood mononuclear cells of patients with hypersensitivity reactions to both CBZ (Leyva et al., 2000) and LTG (Naisbitt et al., 2003a).
Figure 2. Possible pathways of T-cell priming and activation in immune-mediated hypersensitivity reactions to drugs. To trigger immune-mediated inflammation, the drug or a metabolite has to interact with one of the indicated pathways. The classical pathway involves covalent binding to a macromolecule (self-peptide) and antigen processing and presentation (upper part). An alternative pathway exploits noncovalent antigen binding after antigen recognition by crossreactive T cells. Additional “danger signals” from the environment, which act on antigen presenting cells (e.g., substances derived from a cytotoxic effect of the drug or a metabolite) or simultaneously on antigen presenting cells and T cells (e.g., a concomitant viral infection with production of cyto/chemokines), may be needed to trigger immune-mediated inflammation. Locally released cyto/chemokines foster the amplification of the immune response. T cells may express the skin-homing receptor CLA (cutaneous lymphocyte antigen) (see text for details). Ag, antigen; APC, antigen presenting cell; CCL, chemokine (C-C motif) ligand; CD28 and CD80: T-cell activation antigens CD28 and CD80; CXCL, chemokine (C-X-C motif) ligand; iAPC, immature APC; IL, interleukin; MHC, major histocompatibility complex; EBV, Epstein-Barr virus; HIV, human immunodeficiency virus; nT cell, naive T cell; mT cell, memory T cell; TCR, T cell receptor; and TNF, tumor necrosis factor.
Download figure to PowerPoint
Irrespective of the pathogenetic pathway, the role for APCs in immune-mediated hypersensitivity reactions is crucial (Pirmohamed et al., 2002). These cells contribute to inflammation through production of specific cytokines and chemokines that can boost or even suppress the process, depending on the role of co-stimulatory stimuli. Boosting stimuli involve an interaction of APCs with additional, incompletely defined environmental “signals.” Such signals can possibly derive from cells that have been damaged by the drug or a metabolite, or from the immune activation that follows infections or nonspecific “cellular stress” (in the latter case, T cells are also involved as targets) (Fig. 2). The occurrence of boosting stimuli, which is part of the “danger hypothesis” (Uetrecht, 1999; Matzinger, 2002), would explain both the low incidence of hypersensitivity reactions in the general population as well as the increased risk of such reactions under stressful conditions (surgery, viral infections, certain associated disorders). Some of these mechanisms may be reciprocally reinforcing. For example, evidence has been provided that the AED-induced syndrome of drug-related rash with eosinophilia and systemic symptoms (DRESS) may trigger latent virus reactivation and massive nonspecific immune-inflammatory responses, leading to sensitization to other drugs administered during the course of the reaction (Gaig et al., 2006).
The immune-inflammatory responses triggered by the processes described above are polymorphic and show predominant infiltration by specific T cell subtypes (particularly CD4+ though CD8+ may be prevalent in severe clinical presentations), and scattered monocytes and eosinophils (Pichler et al., 2002). The cellular heterogeneity mirrors the complex and overlapping production of cytokines and chemokines (Fig. 2). Eotaxin (CCL11) and IL-5 act as key factors as attractants and activators of eosinophils. IL-8, a neutrophil-attracting chemokine that can also be produced by T cells, is particularly upregulated in SJS, where it contributes to the severe clinical manifestations and intense leukocytosis (Greenberger, 2006). The role of regulatory T cells, which are important in autoimmunity and allergy, has been little studied in immune-mediated hypersensitivity to AEDs. These cells can exert suppressive functions in hapten-allergic individuals, mainly through production of IL-10 (Girolomoni et al., 2004). Advances in knowledge regarding their role in drug-related hypersensitivity bear promise for application in specific immunotherapies.
Certain idiosyncratic adverse reactions cannot be explained by the mechanisms discussed above. In such cases, the pathogenesis must involve alternative events which, while diverse at molecular level, share as a common feature an unusual interaction of the drug (or a metabolite) with the host organism. By definition, these reactions cannot be explained by the primary pharmacological properties of the offending agents, and one must consider or postulate the presence in the affected organism of specific peculiarities, which result in unexpected effects.
On some occasions, the mechanism underlying these reactions are explained by genetically or disease-mediated alterations in susceptible individuals. Examples include the precipitation of hemolytic attacks by several therapeutic agents in patients with 6-phosphate dehydrogenase deficiency (favism) (Mehta et al., 2000), or the induction of porphyric attacks by a variety of AEDs in patients with AIP (Hahn et al., 1997). In most cases, however, the pathogenic mechanism is unknown, and these reactions stand out for their unpredictability, low frequency and, at times, dramatic presentation. Many unusual CNS adverse effects fall within this category: examples include the precipitation of choreoathetoid reactions by PHT (Zaccara et al., 2004), Parkinsonian symptoms by VPA (Masmoudi et al., 2006), and severe psychiatric reactions by AEDs not commonly associated with such effects (Wong et al., 1997). While the classification of some of these reactions as idiosyncratic may be questioned, their unpredictability, low frequency, occurrence (at times) at low dosages and uncertain pathogenesis, possibly related to interaction with altered neuronal circuitries, are reminiscent of idiosyncratic mechanisms.