MAY 1999
`VOLUME 12, NUMBER 5
`
`© Copyright 1999 by the American Chemical Society
`
`Perspective
`
`New Concepts in Immunology Relevant to Idiosyncratic
`Drug Reactions: The “Danger Hypothesis” and Innate
`Immune System
`
`Jack P. Uetrecht*
`
`Faculties of Pharmacy and Medicine, University of Toronto, Toronto, Ontario, Canada M5S 2S2
`
`Received November 16, 1998
`
`Introduction
`Idiosyncratic drug reactions, sometimes referred to as
`hypersensitivity reactions or type B reactions, are ad-
`verse drug reactions that do not occur in most patients
`at any dose and do not involve known pharmacologic
`properties of the drug. Although they are often referred
`to as rare, with a typical incidence of from 1/100 to
`1/100000, because of the total number of drugs involved
`and the number of patients treated, such reactions are
`actually common. Their unpredictable and serious nature
`makes them a significant clinical problem, and they also
`significantly hamper drug development. If we are ever
`to effectively deal with these adverse reactions, it is
`imperative that we come to better understanding of their
`underlying mechanism. The most prevalent hypothesis
`for the mechanism of idiosyncratic drug reactions is the
`hapten hypothesis (1-3). This hypothesis proposes that
`drugs, or more commonly reactive metabolites of drugs,
`act as haptens and irreversibly bind to proteins or other
`macromolecules. These altered proteins are “perceived”
`as foreign and induce an immune response. In most
`individuals, this immune response is asymptomatic, but
`in a few cases, it leads to pathology.
`There is a large amount of circumstantial evidence that
`supports the hypothesis that reactive metabolites are
`involved in idiosyncratic drug reactions (2, 4-11). For
`
`* To whom correspondence should be addressed. Telephone: (416)
`917-8939. E-mail:
`jack.uetrecht@utoronto.ca.
`
`example, halothane, which is associated with a relatively
`high incidence of serious idiosyncratic liver toxicity, is
`extensively metabolized to the reactive trifluoroacetyl
`chloride, and patients with halothane-induced hepato-
`toxicity have antibodies against trifluoroacetylated pro-
`tein (12, 13). When the structure is modified to isoflurane
`or desflurane, which are metabolized to essentially the
`same reactive metabolite but to a lesser degree, the risk
`of liver toxicity is markedly reduced (14). Furthermore,
`the site of reactive metabolite formation usually cor-
`relates with the site of toxicity. For example, the oxida-
`tion of a C-H bond, as found in halothane, is essentially
`limited to cytochrome P450 with the result that most
`halothane oxidation occurs in the liver and the major
`toxicity is limited to the liver. Oxidation of vesnarinone
`to a reactive iminium ion occurs in neutrophils but not
`in the liver, and the dominant toxicity is agranulocytosis
`(15). Clozapine is oxidized to a reactive metabolite in both
`neutrophils and the liver, and clozapine is associated with
`both agranulocytosis and liver toxicity (16). These ex-
`amples are illustrated in Figure 1. Many other such
`examples could be given.
`Idiosyncratic reactions involving the skin, which is a
`common site of such reactions, pose a conceptual problem
`because the concentrations of the enzymes most com-
`monly involved in the formation of reactive metabolites
`are low, although they are not completely absent (17, 18).
`Furthermore, other cells that enter the skin can metabo-
`lize drugs (19). Another possible explanation is that,
`
`10.1021/tx980249i CCC: $18.00 © 1999 American Chemical Society
`Published on Web 04/08/1999
`
`AstraZeneca Exhibit 2163
`Mylan v. AstraZeneca
`IPR2015-01340
`
`Page 1 of 9
`
`

`
`388 Chem. Res. Toxicol., Vol. 12, No. 5, 1999
`
`Uetrecht
`
`can act as a superantigen by noncovalently interacting
`with a major histocompatibility complex (MHC)1 mol-
`ecule, resulting in a more general immune response (24,
`25). They were able to generate clones of T cells from
`patients with idiosyncratic drug reactions that respond
`by proliferation on exposure to the responsible drug in
`the absence of metabolism. The observation that such
`cells can be found in patients with a recent history of an
`idiosyncratic reaction does not prove that such a nonco-
`valent interaction can initiate an idiosyncratic reaction,
`and once an immune reaction is initiated by a covalent
`adduct of the drug, some of the cells that proliferate
`would probably also recognize drug alone. At this point,
`there is enough circumstantial evidence for the involve-
`ment of reactive metabolites to suggest that they are
`responsible for most idiosyncratic drug reactions. How-
`ever, if evidence can be obtained that such cells are
`actually involved in the initiation of idiosyncratic drug
`reactions, it would provide a second pathway by which
`drugs could induce an immune response.
`There are some difficulties with the second half of the
`hapten hypothesis, specifically, that the damage is medi-
`ated by a classical antibody or cytotoxic T cell response
`to the hapten. Although there are notable examples in
`which antibodies that bind to the reactive metabolite
`acting as a hapten are found in patients with an
`idiosyncratic reaction to a drug (4, 12, 26, 27), in most
`cases such antibodies are not found. Even when such
`antibodies are found, as in the case of halothane hepa-
`titis, it is not clear that these antibodies are pathogenic,
`and such antibodies could simply be a marker for an
`immune response (28).
`Helper T cells should be involved in either an antibody
`response or a cytotoxic T cell response, and this is the
`basis for the lymphocyte transformation test (29). The
`lymphocyte transformation test consists of taking cells
`from patients with a recent history of an idiosyncratic
`drug reaction and incubating them with the drug that is
`responsible for the reaction along with a source of antigen
`presenting cells and radiolabeled thymidine. In principle,
`T cells that are specific for the drug should be stimulated
`to proliferate and take up the thymidine. Except for
`patients with a history of reactions to drugs such as
`penicillin that are chemically reactive, this is an unreli-
`able test with a high incidence of both false negative and
`false positive results (30-33). There are at least two
`possible reasons why this test could be falsely negative.
`One is that what the T cell recognizes is the reactive
`metabolite bound to protein acting as a hapten, and
`unless metabolism occurs in the cell incubation, the
`reactive metabolite would not be formed (34, 35). This
`would also explain why the test works better with drugs,
`like penicillin, that covalently bind without being me-
`tabolized. However, in studies by Kalish, using the
`reactive metabolite instead of the parent drug did not
`substantially increase the number of positive tests (36).
`The other factor that could lead to a false negative test
`is that the number of T cells specific for the drug is small
`(on the order of 1/100000) and it is difficult to detect
`proliferation of these cells in the midst of the other T cells
`(36). This problem can be approached with limiting
`dilution techniques in which varying numbers of lym-
`phocytes are added to the wells of a multiwell plate and
`
`1Abbreviations: MHC, major histocompatibility complex; APC,
`antigen presenting cell; TCR, T cell receptor.
`
`Figure 1. Illustration of the correlation between the amount
`and location of metabolism and the degree and type of idiosyn-
`cratic reaction.
`
`although most reactive metabolites have a short biologi-
`cal half-life, some, such as acyl glucuronides, are reactive
`but freely circulate and may be responsible for idiosyn-
`cratic reactions in the skin (20). Others, such as the
`iminoquinone formed from carbamazepine, readily redox
`cycle and could undergo several such cycles before finally
`binding in the skin (21). Although the reactive metabo-
`lites of many drugs associated with a high incidence of
`idiosyncratic reactions have not been identified, many of
`the pathways can be postulated, and with LC/MS and
`the other sensitive analytical methods that are now
`available, it appears that most drugs form reactive
`metabolites to some degree.
`There seems to be a crude correlation between the
`amount of reactive metabolite formed and the risk that
`a drug will be associated with a high incidence of
`idiosyncratic drug reactions as illustrated by the com-
`parison of halothane and isoflurane; however, this is
`difficult to prove because it is difficult to quantify the
`amount of reactive metabolite that is formed. The risk
`of an idiosyncratic drug reaction is often said to be
`independent of dose. Such reactions may appear to be
`independent of dose because most patients do not have
`an idiosyncratic reaction at any dose and the usual dose
`range is usually quite narrow; however, even with these
`confounding factors, a relationship between the dose and
`the incidence of an idiosyncratic reaction is often observed
`(22, 23). Obviously, if the total dose of a drug is low, the
`total amount of reactive metabolite that can be formed
`is limited, and drugs given at a daily dose of 10 mg or
`less are rarely if ever associated with a high incidence of
`idiosyncratic drug reactions.
`An alternative hypothesis to the reactive metabolite
`hypothesis has been proposed by Pichler in which a drug
`
`Page 2 of 9
`
`

`
`Perspective
`
`incubated with the drug responsible for the reaction (37).
`The cells are cultured for a longer period of time so that
`cells that are not stimulated by the drug die. From the
`number of wells that are positive, i.e., in which colonies
`form, and the number of cells that were originally placed
`in these wells, the incidence of the cells in the blood that
`are specific for the drug can be calculated. Even with this
`refinement, the lymphocyte transformation test is often
`negative; however, there have not been many studies
`published in which the use of reactive metabolite and
`limiting dilution techniques have been combined (36). It
`has been suggested that adding a prostaglandin synthase
`inhibitor increases the sensitivity of the test, but again
`even with the addition of such an inhibitor, almost half
`of the tests were still negative (38).
`There is another characteristic of some idiosyncratic
`drug reactions that strongly suggests that these reactions
`are not mediated by classical antibody- or cell-mediated
`immune reactions. Specifically, we have been puzzled by
`the fact that if a patient with a recent history of
`clozapine-induced agranulocytosis is rechallenged with
`clozapine, it usually takes just as long (about 6 weeks)
`for the onset of agranulocytosis as it did during the initial
`exposure (39, 40). This is very difficult to reconcile with
`an amnestic response of the immune system. Alterna-
`tively, clozapine-induced agranulocytosis could be due to
`direct cytotoxicity. However, the characteristics of the
`reaction are very difficult to explain on the basis of a
`direct cytotoxic reaction. Specifically, the reaction is
`idiosyncratic in humans; it has not been successfully
`reproduced in animals, and there is usually a delay of
`more than 1 month between starting the drug and the
`onset of agranulocytosis. It might be possible to explain
`the delay in onset on a slow buildup of a toxic agent or
`the slow depletion of some critical factor. However, if such
`a mechanism were responsible for the delay in onset, it
`should take a long time for the bone marrow to recover
`when the drug is stopped, but in general, it recovers very
`rapidly. Specifically, the average time it takes for the
`neutrophil count to return to normal in the circulation
`is 1 week (41), and therefore, the marrow must recover
`almost immediately because it takes about 1 week for
`neutrophil precursors in the bone marrow to mature and
`reach the circulation. There also appears to be a delay
`in the onset of agranulocytosis on reexposure to pheno-
`thiazines, and this has been used as evidence that it is a
`toxic reaction (42); however, these reactions are idiosyn-
`cratic, and their other characteristics do not really fit a
`simple cytotoxic mechanism. Many other idiosyncratic
`drug reactions may be associated with a delay on reex-
`posure, but this is not well documented because, in most
`cases, patients are not rechallenged with a drug believed
`to be responsible for an idiosyncratic drug reaction.
`The major characteristic of idiosyncratic drug reactions
`that make them especially difficult to deal with is simply
`their idiosyncratic nature. It has been impossible to
`predict which patients will have an idiosyncratic reaction
`to a specific drug. Although the formation of a reactive
`metabolite appears to be necessary for the reaction, as
`illustrated by halothane,
`it also appears that most
`patients and animals form the relevant reactive metabo-
`lite without a clinically significant adverse reaction (43).
`Furthermore, although it is probably a risk factor in some
`patients, attempts to find differences in the formation
`or detoxication of a specific metabolite that would predict
`who will have an idiosyncratic reaction to a drug have
`
`Chem. Res. Toxicol., Vol. 12, No. 5, 1999 389
`not yet been successful (44-46). In the few cases where
`covalent binding can be detected in humans, it also occurs
`in patients who do not have an idiosyncratic reaction to
`the drug (47).
`It has also been postulated that since the reactions are
`likely immune-mediated, certain MHC genotypes should
`be linked to an increased risk of a specific idiosyncratic
`drug reaction. Here again, although weak associations
`with specific MHC genotypes have been detected, most
`such studies have been disappointing (48-50). Thus, we
`remain unable to predict individual susceptibility to
`idiosyncratic drug reactions, and a better understanding
`of the mechanism of these reactions will be required to
`make progress in this area. If, as is suggested by their
`characteristics, most idiosyncratic drug reactions are
`immune-mediated, it is likely that a better understanding
`of the mechanism will depend on advances in our
`understanding of immune-mediated reactions. The pur-
`pose of this perspective is to highlight relatively new
`concepts from immunology that have implications for the
`mechanisms of these difficult adverse reactions.
`
`Danger Hypothesis
`
`For a long time, immunologists presumed that a major
`function of the immune system was to differentiate “self”
`from “nonself” and to respond to self with tolerance and
`to mount a response against nonself. In some cases, the
`immune system seemed to do it incorrectly and mounted
`a response against self, resulting in an autoimmune
`reaction. More recently, Matzinger (51) argued that it is
`difficult and inefficient to try to differentiate self from
`nonself and proposed an alternative hypothesis in which
`the immune system responded with tolerance to most
`antigens, and what triggers an immune response is
`presentation of an antigen in the context of a “danger
`signal” rather than the foreignness of the antigen. The
`exact nature and range of stimuli that can act as the
`danger signal remain to be determined, but certainly, cell
`damage must be a major stimulus for the production of
`the danger signal.
`In the more traditional view, tolerance of self is
`achieved, in part, by deletion of T cells that recognize
`self-antigen present in the thymus during maturation
`(52, 53). Another aspect of tolerance in the more tradi-
`tional view is that lymphocytes require a second signal
`as well as signal 1, and without signal 2, the system
`becomes tolerant to the antigen. Signal 1 is the “recogni-
`tion” by T cells of antigen. This recognition involves
`interaction between processed antigen imbedded in the
`MHC on antigen-presenting cells (APCs) and the T cell
`receptor (TCR) on T cells (54). The nature of signal 2 is
`not completely defined, but a principal component con-
`sists of the binding of CD28 on T cells to B7 (there are
`actually two B7 molecules, B7-1 and B7-2) on APCs (55).
`Other signaling molecules and soluble cytokines also play
`a role in mediating signal 2 (56). However, the require-
`ment for signal 2 does not help to explain how tolerance
`to self antigens is achieved because self-antigens are
`presented by APCs in the same manner as foreign
`antigens. This view does, however, require that the APC
`be activated to produce an immune response to an
`antigen because activation of APC leads to the release
`of cytokines and the expression of B7 on the surface of
`the APC. This concept is similar to an essential aspect
`of the Danger Hypothesis.
`
`Page 3 of 9
`
`

`
`390 Chem. Res. Toxicol., Vol. 12, No. 5, 1999
`
`A major aspect of the Danger Model proposed by
`Matzinger is the laws of lymphotics (51). The first law
`states that a lymphocyte will die if it receives signal 1
`without signal 2, and the second law states that a
`lymphocyte will accept signal 2 only from APCs. She goes
`on to postulate that thymocytes are unable to receive
`signal 2 from any source, and so if they encounter antigen
`they would be deleted, which is similar to the more
`traditional view. Additional laws are as follows. (1)
`Inexperienced or “virgin” T cells can only receive signal
`2 from “professional” APCs (i.e., dendritic cells), while
`experienced T cells (those that have responded at least
`once to antigen) can also receive signal 2 from other APCs
`such as B cells or macrophages. (2) Resting B cells only
`accept signal 2 from experienced or effector, i.e., acti-
`vated, T cells. (3) Effector B or T cells respond to signal
`1 without requiring signal 2. (4) The effector cells then
`either revert to a resting state or die after a relatively
`short period of time so that the response does not get
`out of hand.
`The central determinant of an immune response in this
`model is the activation of APC, and Matzinger proposes
`that APCs are activated in the presence of “bad death”
`(i.e., necrosis) or cell stress (57). She proposes that the
`default response of the immune system is tolerance, and
`it is “danger” rather than nonself that leads to an
`immune response. In this model, apoptosis is not bad
`death and would lead to tolerance. Although the model
`presented by Matzinger is slightly more complex than
`this, it is likely that the truth will be found to be even
`more complex than presented by the Danger Model.
`However, the Danger Model may be very useful for
`understanding many aspects of idiosyncratic drug reac-
`tions as described below.
`There are many examples in which specific situations
`that could mediate a danger signal are associated with
`an increase in the risk of an idiosyncratic drug reaction.
`For example, it has been known for a long time that when
`ampicillin is given to patients with mononucleosis, there
`is a marked increase in the risk of an idiosyncratic
`reaction that approaches 100% (58). More recently, it has
`been observed that patients who are HIV positive have
`a much higher risk of idiosyncratic reactions to sulfona-
`mides and other drugs (59, 60). Even an influenza
`vaccination appears to increase the risk of idiosyncratic
`drug reactions as revealed by studies of vesnarinone-
`induced agranulocytosis (15). Open-heart surgery, which
`clearly should produce a danger signal, appears to
`increase the risk of procainamide-induced agranulocyto-
`sis by a factor of 10 (61, 62). Many other such factors
`that represent a danger signal probably remain to be
`discovered. Especially relevant to adverse drug reactions,
`since reactive metabolites are necessary for the formation
`of haptens, these reactive metabolites could also cause
`sufficient cell stress or necrosis to result in a danger
`signal without causing serious direct toxicity. It is also
`very important to point out that although surgery ap-
`peared to increase the incidence of procainamide-induced
`agranulocytosis by a factor of 10, 95% of the patients still
`did not develop agranulocytosis. This suggests that there
`is a mechanism or mechanisms by which the immune
`system can downregulate potentially harmful immune
`responses before they do too much damage. Such mech-
`anisms are probably under genetic control.
`As proposed by the Danger Hypothesis, the usual
`response to a hapten is tolerance. Many drugs, e.g.,
`
`Uetrecht
`
`acetaminophen, form reactive metabolites but at usual
`doses are not associated with a significant incidence of
`idiosyncratic drug reactions. It also appears that starting
`with a low dose of a drug and then later increasing to a
`therapeutic dose increases tolerance (63). The lower dose
`may lead to tolerance because the resultant decrease in
`the amount of reactive metabolite formed may result in
`apoptosis rather than in necrosis or simply less cell
`stress, and therefore, there is no danger signal.
`Another observation that can be explained by the
`Danger Hypothesis is the time course of drug-induced
`autoimmune reactions. Drugs, such as R-methyldopa,
`cause an autoimmune hemolytic anemia (64, 65), and
`other drugs, such as procainamide, cause a lupus-like
`syndrome that is associated with a broader range of
`autoantibodies (66). As implied by the term autoantibody,
`these antibodies bind to self-antigens in the absence of
`drug. Therefore, even if the administration of the drug
`is stopped, the antigenic stimulus remains, and one might
`expect the reaction to continue. In fact, although antibod-
`ies may be detected for some time, as soon as the drug is
`discontinued the clinical syndrome usually is resolved
`quite rapidly (67). This can be explained on the basis of
`the Danger Hypothesis where the drug, presumably due
`to a reactive metabolite, is responsible for the danger
`signal. According to the laws of lymphotics, effector cells
`either revert to a resting state or die after a relatively
`short period of time; therefore, in the absence of a danger
`signal, the model predicts rapid resolution. Although the
`immune response may kill cells, cell death mediated by
`the immune system is usually apoptotic, and this should
`not contribute to a danger signal.
`
`Innate Immune System
`
`For several decades, immunologists have concentrated
`on the adaptive immune system in which antigen is
`processed and presented in the context of MHC-II to
`helper T cells by APCs (54). These cells, in turn, stimulate
`B cells to differentiate and proliferate into antibody-
`producing plasma cells and/or stimulate cytotoxic T cells
`that recognize the same antigen, although in the form of
`peptides derived from the antigen presented in the
`context of MHC-I instead of MHC-II. With the aid of gene
`rearrangements, this system is able to respond to an
`almost infinite number of antigens in a very specific
`manner. Almost forgotten during these studies was the
`innate immune system. The innate immune system was
`viewed as a primitive system present in invertebrates
`but less important in mammals, which also possess an
`adaptive immune system. However, invertebrates, which
`do not possess an adaptive immune system, deal quite
`effectively with pathogens. Furthermore, there has been
`a recent realization that the adaptive immune system
`does not operate in isolation and the innate immune
`system is probably also very important in mammals (68).
`The innate immune system can only respond to stimuli
`that have been encoded in an organism’s DNA because
`the gene rearrangements that make an adaptive response
`possible do not occur in the cells of the innate system
`(69). The cells most important for an innate response are
`granulocytes, macrophages, NK cells, and (cid:231)(cid:228) T cells (70).
`The types of constant structures that stimulate an innate
`response are molecules, such as lipopolysaccharides, that
`are present in the cell wall of many bacteria, viral DNA,
`which is hypomethylated, and several nonpeptidic phos-
`
`Page 4 of 9
`
`

`
`Perspective
`
`phoantigens that are derived from various infectious
`agents (69, 70). However, the full range of structures to
`which the innate immune system can respond has yet to
`be determined, and it is an active field of research at the
`present time. The innate system also differs from the
`adaptive system in that antigen can be recognized by (cid:231)(cid:228)
`T cells in a non-MHC-dependent manner (71). However,
`it appears that cell-surface receptors for MHC-I molecules
`do play an important role in the control of the innate
`response (71). Another factor that appears to stimulate
`an innate response is cell stress (72). This would seem
`to be an important feature in invertebrates for the
`detection of cells infected by viruses or other damaged
`cells that could be harmful because the innate system is
`not able to respond to new and varied antigenic struc-
`tures.
`Although the innate immune system can deal with
`pathogens and damaged cells directly, in mammals,
`another important function of the innate system may be
`to determine whether the adaptive immune system will
`respond to a stimulus with tolerance or an active immune
`response (69, 70). Thus, it may be the innate immune
`system that delivers the danger signal described in the
`previous section. Adjuvants often contain agents such as
`lipopolysaccharide that stimulate the innate immune
`system. Janeway referred to adjuvants as the immunolo-
`gist’s “dirty little secret”, because without them pure
`antigens usually produce little or no response (73). The
`innate system has several mechanisms by which it can
`stimulate an adaptive response. An innate response can
`lead to the secretion of cytokines by macrophages and
`NK cells, the attachment of complement to antigen, and
`the expression of B7 on APCs (72, 73). However, there is
`no need to postulate that the innate immune system is
`required to provide the danger signal, and APCs may be
`able to detect danger without the involvement of the
`innate immune system. Furthermore, Matzinger believes
`that the danger signal originated even before the innate
`immune system (57). When the immune system is
`understood in detail, it is likely that the innate and
`adaptive immune systems will be found to be closely
`interdependent, and the distinction between them may
`begin to blur.
`It seems likely that the innate immune system could
`be involved in the mechanism of some idiosyncratic drug
`reactions, yet this does not appear to have been consid-
`ered previously. The characteristics of the innate immune
`system could easily explain the perplexing time course
`of clozapine-induced agranulocytosis on reexposure, i.e.,
`the lack of a rapid amnestic response, because there are
`no memory cells in the innate immune system (73). Since
`stressed cells seem to be able to stimulate the innate
`immune system, it is obvious how a drug that produces
`reactive metabolites could lead to an innate immune
`response. There are several other examples of idiosyn-
`cratic reactions that have characteristics that suggest
`mediation by the innate immune system. For example,
`the aminopenicillins, ampicillin and amoxicillin, cause
`a rash in children that is often described as a “toxic” rash
`(74). However, these rashes are idiosyncratic, and there
`is usually a delay between starting the drug and the
`onset of the rash which makes it unlikely that they are
`true toxic reactions. These rashes are often seen in
`patients who have a viral infection; they often do not
`recur if the patient takes the drug again, and skin tests
`are usually negative (75). In some cases, these may be
`
`Chem. Res. Toxicol., Vol. 12, No. 5, 1999 391
`
`delayed-type hypersensitivity reactions (75), but in our
`experience, even in adults, the skin tests are usually
`negative and the reaction does not usually occur on
`reexposure. Such characteristics may be explained by an
`innate immune response. The adverse reactions of HIV
`positive patients to sulfamethoxazole and other drugs
`also may be mediated by the innate immune system
`because they often do not recur on reexposure and occur
`in the context of a danger signal (76). However, as
`mentioned above, the distinction between an innate
`reaction and an adaptive response may blur because a
`reaction mediated by the innate system may stimulate
`an adaptive response. The adaptive response may con-
`tribute to pathogenesis or may simply be an epiphenom-
`enon.
`A model of idiosyncratic drug reactions that incorpo-
`rates the Danger Hypothesis and the innate immune
`system is shown in Figure 2. In this figure, the drug is
`denoted by the letter R and the electrophilic metabolite
`that binds to protein is denoted by R+. If only a small
`amount of reactive metabolite is formed, a significant
`response is unlikely. If more reactive metabolite is
`formed, when the cell undergoes apoptosis, either because
`of senescence or because the reactive metabolite led to
`apoptosis, the cell will undergo phagocytosis. The hap-
`tenized proteins will be processed and presented as
`haptenized peptides to T cells in the absence of signal 2
`as shown in the left arm of the figure. (The processed
`peptides bound to MHC-II are shown in Figure 2 as a
`shorter structure than the original haptenized protein.)
`Presentation in the absence of signal 2 is likely to lead
`to immune tolerance to the drug, or more accurately,
`tolerance to the reactive metabolite of the drug acting
`as a hapten.
`If the reactive metabolite is more cytotoxic, either
`because of the amount formed or because of the nature
`of the reactive metabolite, it may lead to cell stress or
`necrosis. This would lead to a danger signal and upregu-
`lation of B7 on the APC as illustrated in the central arm
`of Figure 2. Although in the Matzinger Danger Model
`apoptosis leads to tolerance while necrosis induces an
`immune response, there is evidence that antigen from
`apoptotic cells can induce an immune response (77).
`There could also be some environmental agent, such as
`an infection, that acts as a danger signal and upregulates
`B7. Alternatively, cells of the innate immune system
`could detect cell stress or some other danger signal, and
`they could produce cytokines or other factors that could
`upregulate B7 as illustrated in the right arm of Figure
`2. Alternatively, and not shown in Figure 2, the innate
`system-derived factors might provide signal 2 by directly
`stimulating the helper T cells.
`In Figure 2, it is implied that the immune response,
`either antibody-mediated or cell-mediated, would be
`against the haptenized peptide; however, it is common
`to find antibodies against native proteins that have not
`been modified by reactive metabolite (78-80). One mech-
`anism by which this could happen is one in which
`haptenization can lead to a change in the processing of
`proteins and the presentation of peptides not usually
`formed by the protein. When peptides from self-proteins
`are presented to the immune system that have not been
`presented before, or have been presented in much smaller
`quantities that would not have induced tolerance, these
`peptides are called cryptic peptides (81, 82). In the
`Matzinger model, the concept of self-proteins is not used,
`
`Page 5 of 9
`
`

`
`392 Chem. Res. Toxicol., Vol. 12, No. 5, 1999
`
`Uetrecht
`
`Figure 2. Illustration of the proposed mechanisms of idiosyncratic drug reactions that involve the danger hypothesis and the innate
`immune system. See the text for details.
`but the important concept is that these cryptic peptides
`would not have previously induced tolerance and could,
`in the presence of a danger signal, induce an immune
`response. This may lead to an immune response directed
`against the native protein. It is also possible that a
`hapten-modified protein resembles an infectious agent,
`leading to a break in the tolerance to the protein. This is
`called molecular mimicry (46, 83, 84). By either mecha-
`nism, true autoimmune responses could be initiated.
`Finally, the danger signal may stimulate cells of the
`innate immune system, and these cells might directly
`mediate an idiosyncratic reaction. Exactly how this could
`occur is unknown, but a simple hypothesis is that
`macrophages, NK cells, and/or (cid:231)(cid:228) T cells might detect a
`danger signal on cells that have been modified by reactive
`metabolite. These cells would become the target cells,
`
`either because they formed a large amount of reactive
`metabolite or because they were more sensitive to the
`toxic effects of the reactive metabolite. The danger signal
`detected by cells of the innate immune system would not
`have to be the same as that detected by APCs in the
`adaptive immune system. The cells of the innate immune
`system could directly induce apoptosis in the affected
`target cells. Although the involvement of the innate
`immune system in idiosyncratic reactions has not been
`demonstrated, there is evidence that the innate immune
`system may be involved in other types of xenobiotic
`toxicity. Specifically, agents such as acetaminophen and
`carbon tetrachloride cause liver toxicity that is not
`idiosyncratic. Although there is a large amount of evi-
`dence to support the involvement of reactive metabolites
`in such reactions, despite extensive investigation, the
`
`Page 6 of 9
`
`

`
`Perspective
`
`fi