HLA: A Pharmacogenomics Success Story

Figure 1. Mechanism of an immunoallergic adverse drug reaction according to the HLA–drug (hapten) theory.

The drug or a drug metabolite is processed by an APC and presented by a class I or class II HLA protein. The HLA–drug (hapten) complex occurs only in the presence of a specific HLA allele. The presentation of the hapten to the naive CD4 lymphocyte via its TCR triggers the immunologic reaction. Depending on the cytokine immunologic environment, the activation is oriented either via a TH1 reaction (in the presence of TH1 ILs such as IL-2, IL-12 or IFN) or TH2 reaction (in the presence of TH2 interleukins such as IL-4 or IL10). Alternatively, the drug or the hapten processed by an HLA class I protein can directly stimulate cytotoxic CD8 lymphocytes.

APC: Antigen-presenting cell; IFN: Interferon; IL: Interleukin; TCR: T-cell receptor.

Adverse drug reactions (ADRs) represent a major health problem, accounting for 6–7% of hospitalization cases and approximately 100,000 deaths a year in the USA [1]. They also represent the primary cause of drug withdrawal from the market [2]. Most ADRs are considered type A, that is, predictable, dose dependent and related to the pharmacological action of the drugs. However, 5–20% of ADRs are unpredictable (type B), mainly related to immunoallergic reactions. From a clinical point of view, immunoallergic ADRs (IADRs) can be divided into two categories: immediate allergic events such as urticaria, anaphylaxis, collapsus and angioedema and delayed allergic events such as Stevens–Johnson syndrome (SJS), toxic epidermal necrolysis (TEN), drug rash with eosinophilia and systemic symptoms, or isolated organ injury. The former occur during the first hours after drug administration and are mainly mediated by drug-specific IgE antibodies. The latter occur several days or weeks after drug initiation and are primarily mediated by T-lymphocyte cytotoxicity.

Immunoallergic ADR symptoms are nonspecific, therefore the causal relationship with a drug is frequently difficult to establish, even when taking into account the delay between drug prescription and the event, the improvement after drug withdrawal, and recurrence after re-exposure (positive rechallenge).

The most prevalent hypothesis used to explain drug immunoallergic reactions is the hapten/prohapten theory: according to this hypothesis, the drug (or a drug metabolite) is processed by an antigen-presenting cell (APC) and further expressed at the cell membrane in the groove of a class I (MHC1) or a class II HLA (MHC2) molecule. The HLA–drug (hapten) complex is presented to a naive lymphocyte via its T-cell receptor, which will initiate different types of immunological responses, depending on the type of HLA molecule expressed by the APC and the cytokine environment (Figure 1). Two other hypotheses, involving either a direct interaction of the drug with the T-cell receptor and the drug [3] and the cellular release of danger biological signals [4], have been proposed. All three are not mutually exclusive. However, many factors that trigger the drug-related immunoallergic reaction are still missing. As a consequence, until recently there were only a few unspecific biomarkers such as eosinophilia to signal the immunoallergic nature of an ADR. Skin tests, specific IgE and in vitro lymphocyte proliferation tests can be used after the event, but their specificity or sensitivity remain disappointing [5].

On the other hand, it has been known for a long time that drugs can induce immunological diseases [6], such as systemic lupus erythematosus or vasculitis, which occur in susceptible patients expressing specific HLA epitopes. This indicates a genetic susceptibility to developing IADRs, restricted to HLA class I (A, B and C) or class II alleles. In addition, more and more IADRs have been identified as being associated with HLA loci since the 1970s.

However, the real ‘drug–HLA story‘ started at the beginning of the century with abacavir: in 2002, two independent groups observed that abacavir hypersensitivity syndrome was restricted to carriers of the HLA-B*5701 allele with the highest odds ratio ever met (>100) [7,8]. Following this discovery, GlaxoSmithKline (London, UK) conducted the largest international pharmacogenetic randomized clinical trial ever performed to date, which was released in 2008 [9]. They demonstrated that screening for HLA-B*5701 before introducing abacavir and the exclusion of the patients carrying this allele resulted in the disappearance of the hypersensitivity syndrome related to this drug, which usually occurs in 5% of the patients treated with abacavir during the first weeks of treatment. This pharmacogenetic test is now routinely used in many different countries before introducing abacavir [10–12].

During the same period, screening of HLA alleles in patients hospitalized for life-threatening cutaneous ADRs (SJS and TEN), allowed for the identification of additional HLA alleles as potent risk factors for IADR. Expressing the HLA-B*1502 allele exposes patients of Asian origin to a 2504- and 36-fold higher risk of SJS to carbamazepine [13] and phenytoin [14], respectively (Table 1). Since that time, many different HLA loci have been demonstrated to be highly predictive of specific type B ADRs related to specific drugs (Table 1). HLA alleles have high negative predictive values (Table 1), but low positive values to predict ADRs, indicating that these genetic biomarkers are necessary but not sufficient for the triggering of such immunoallergic events. According to the hapten theory, it is believed that the hapten complex, which triggers the immunoallergic reaction, only occurs in the presence of a specific HLA allele. Therefore, prospective HLA screening should prevent some patients from receiving a certain drug, if they express a specific risk allele, in order to avoid a potentially life threatening IADR.

The technological screening facilities mainly explain the recent and sudden identification of different HLA alleles as potent risk factors of IADRs: simultaneous screening for a large number of HLA epitopes has been performed for a long time in organ transplantation and also allowed for the early identification of HLA-restricted drug-related diseases [6]. Recent genomic technologies offered the opportunity to simultaneously screen, in so-called genome-wide association studies (GWAS), most of the thousands of alleles [15] of the extended human MHC [16]. Furthermore, owing to the extraordinarily high risks conferred by several HLA alleles (odds ratio ranging from 10 to 2500), GWAS can be performed with less than 100 subjects and still allow for the discovery of positive pharmacogenomic hits [17]. Using this agnostic genome-wide strategy, Daly et al. recently identified HLA-B*5701 as a potent risk factor for drug-induced liver injury (DILI) owing to flucloxacillin from the Drug Induced Liver Injury and Gene (DILIGEN) study, which included 51 cases and 289 controls [18]. In a retrospective study performed in 74 cases and 130 controls, AstraZeneca (London, UK) was able to identify HLA-DRB1*0701 as the major genetic risk factor associated with ximelagatran DILI, which led to its withdrawal soon after its introduction on the market. This underlines the crucial need for systematic DNA banking during clinical drug development. DNA biobanks collected during drug development or in the post-marketing phase allows one, if an IARD is identified, to rapidly perform a retrospective GWAS and identify a predictive genomic biomarker. The systematic use of the latter will certainly limit the indication of the drug, but will on the other hand help to avoid its withdrawal from the market.

An increasingly more affordable genome-wide screening approach, combined with the high risk predicted by some HLA alleles, allows us to hope that in the near future, more and more severe IADRs might be prevented by pharmacogenomics testing [19]. However, looking forward to that day, several issues remain to be addressed. They will be discussed here.

In order to be used in routine medical practice, the positive or negative predictive values of the HLA biomarker must reach 100%. Some of the HLA alleles listed in Table 1 fulfill this condition. Even so, the cost–effectiveness of systematic screening before drug prescription must be positive. For a frequent IADR, as observed with abacavir, this condition has been verified [20], but for rare events, such as allopurinol-induced SJS, it remains to be established. In addition, one should also keep in mind that a HLA risk allele may be ethnically restricted: HLA-B*1502 is highly predictive of carbamazepine-induced SJS in different Asian populations. However, in a European survey of SJS/TEN related to carbamazepine, only four out of 12 patients were carriers of the HLA-B*1502; interestingly these four patients had an Asian ancestry [21]. HLA-restricted immunological response is also specific to a certain type of adverse event for a specific drug: HLA-B*1502 is a strong predictor of carbamazepine-related SJS/TEN but not of carbamazepine isolated skin rashes [22], which are more frequently observed in carriers of the HLA-A*3101[22]. Conversely, the same HLA allele can be predictive of different IADRs of different drugs such as HLA-B*5701, which is a strong risk factor of abacavir HSS, as well as flucloxacillin DILI.

However, despite these limits underlying the need for further research into the mechanisms of IADRs, we now have strong evidence that some IADRs are linked to specific HLA alleles and can be prevented by prospective pharmacogenomics testing. The pharmacogenomics of HLA is a recent field that has been rapidly translated into the clinic and is already improving the way we prescribe drugs. It will probably become more and more important over the next few years.

Table 1. Pharmacogenomic risk factors for adverse drug reactions.

Drug	Gene target	Adverse drug reaction	Odds ratio	Predictive values (%)	Ref.
Carbamazepine	HLA-B*1502	SJS (Asian [21])	2504	PPV = 94 NPV = 100	[13]
Carbamazepine	HLA-A*3101	Rash	17	PPV = 60 NPV = 92	[22]
Carbamazepine	MLN Ω	HSS	7	N.A.	[22]
Phenytoin	HLA-B*1502	SJS (Asian)	36	PPV = 33 NPV = 100	[14]
Allopurinol	HLA-B*5801	SJS	580	PPV = 77 NPV = 100	[23]
Abacavir	HLA-B*5701	HSS	33	PPV = 47 NPV = 100	[9]
Flucloxacillin	TNF-α (-238G/A)	DILI	31	N.A.	[24]
Flucloxacillin	HLA-B*5701	DILI	81	PPV = 91 NPV = 88	[18]
Flucloxacillin	HLA-DRB1*0107-DQB1*0103	DILI	17	N.A.	[24]
Nevirapine	HLA-DRB1*0101	HSS	18	N.A.	[25]
Nevirapine	HLA Cw8-B14	HSS	15	PPV = 75 NPV = 83	[26]
Ximelagatran	HLA-DRB1*0701	DILI	4	PPV = 61 NPV = 73	[27]
Betalactams	TNF (-308G>A)	IgE-mediated HSS	2.5	PPV = 59 NPV = 73	[28]
Aminopenicillin	HL A-A2 HLA-DRW52	Delayed HSS	7 9		[29]
Co-amoxiclav	HLA-DRB1*1501	DILI	9	PPV = 34 NPV = 95	[30]
Clozapine	HLA-DRB5*0201	Neutropenia	22	PPV = 100 NPV = 65	[31]
Sulfomethoxazole	HLA-B*3802	SJS/TEN	76	PPV = 50 NPV = 99	[32]
Lamotrigine	HLA-B*3801	SJS/TEN	5	PPV = 5 NPV = 99	[32]
Oxicams	HLA-B*7301	SJS/TEN	152	PPV = 50 NPV = 99	[32]
Aspirin	HLA-DRB1*1302-DQB1*0609-DPB1*0201	Urticaria	4	PPV = 67 NPV = 68	[33]

Ω: Motilin, MHC II gene; DILI: Drug-induced liver injury; HSS: Hypersensitivity syndrome; N.A.: not available; NPV: Negative predictive value; PPV: Positive predictive value; SJS: Stevens–Johnson syndrome; SLE: Systemic lupus erythematosus; TEN: Toxic epidermal necrolysis.

Financial & competing interests disclosure

Laurent Becquemont has received consulting fees from Sanofi-Aventis, Pfizer, Servier and lecture fees from Genzyme, GlaxoSmithKline, Bristol–Myers Squibb, Wyeth and Merck Sharp and Dohme. His wife works for Sanofi-Aventis. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript

Additional information

Funding

Laurent Becquemont has received consulting fees from Sanofi-Aventis, Pfizer, Servier and lecture fees from Genzyme, GlaxoSmithKline, Bristol–Myers Squibb, Wyeth and Merck Sharp and Dohme. His wife works for Sanofi-Aventis. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.