Introduction

Stevens–Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are severe cutaneous adverse reactions (SCARs) characterized by extensive necrosis and detachment of full-thickness epidermis and erosions of the skin and mucous membranes.1 Although these adverse reactions are rare, they are life-threatening conditions with high morbidity and mortality. The incidence rates of TEN and SJS are generally low (0.4–1.2 and 1.2–6 cases per million population per year, respectively),2 but mortality is as high as 1–5% for SJS and 20–30% for TEN.2, 3

Methazolamide is a sulfonamide derivative used as a carbonic anhydrase inhibitor to reduce intraocular pressure elevations associated with glaucoma and other ocular disorders. Sulfonamides are a class of drugs with a high risk of inducing SCARs.4, 5 Sulfonamide derivatives that are potent inhibitors of carbonic anhydrase have been reported to cause SJS and TEN,6, 7 albeit less frequently than antimicrobial sulfonamides.

The first cases of methazolamide-induced SJS were reported in two Japanese-American women.7 Five more cases were subsequently reported in Japan.8, 9 Shirato et al.8 identified three of four patients as HLA-B*59 carriers. A later series of reports highlighted the risk of SJS/TEN induced by methazolamide treatment and provided further evidence for the association of HLA-B*59 with adverse reactions in the Korean population.10, 11, 12, 13 Kim et al.14 recently performed allele-level genotyping of HLA-A, -B and -C genes and observed a strong association between methazolamide-induced SJS/TEN and HLA-B*59:01 in five Korean patients.

Case reports of methazolamide-induced SJS/TEN have primarily been limited to patients of Japanese and Korean descent. To date, only two cases of SJS in glaucoma patients following methazolamide treatment have been reported in domestic journals of China, and no genotyping was performed.15, 16 In the present study, we diagnosed a total of eight Han Chinese patients with methazolamide-induced SJS/TEN and characterized the HLA class I genotypes of each patient to identify risk alleles for adverse reactions in the Han Chinese population. To our knowledge, this is the first association study of HLA genes with methazolamide-induced SJS/TEN in Han Chinese.

Materials and methods

Subjects

Study subjects were recruited during January 2008 through November 2014 from the dermatology and emergency wards of Huashan Hospital in Shanghai, China. Eight patients who fulfilled the diagnostic criteria for methazolamide-induced SJS or TEN were enrolled in the study. All patients came from the southern region of China. Diagnosis was performed by two independent dermatologists who examined photographs, pathological slides and medical records. All clinical records were reviewed to exclude cases of immunobullous diseases, viral infections and toxic eruptions. In accordance with the clinical morphology described by Roujeau et al.,17 SJS was defined as epidermal detachment involving <10% of the total body surface area, overlap SJS/TEN as epidermal detachment of 10–30%, and TEN as >30%. Diagnoses of SJS and TEN were confirmed by skin biopsies showing full-thickness necrosis of the epidermis on pathological examination. Staphylococcal-scalded skin syndrome and other less likely differential diagnoses were excluded using these procedures. Methazolamide was considered the offending drug if the onset of SJS or TEN symptoms occurred within the first 2 months of exposure and the symptoms resolved after withdrawal of the drug. Patients with an absence of symptoms after re-exposure to methazolamide were excluded. Concomitant medications were used in five of the eight patients studied, but no adverse drug reactions (ADRs) related to prescribed drugs other than methazolamide were found in the medical records of these patients.

Two control groups were enrolled in the study. The general population control group consisted of 283 healthy southern Han Chinese subjects from the human major histocompatibility complex database.18 The methazolamide-tolerant group contained 30 patients who did not develop any cutaneous manifestations at least 3 months after the methazolamide treatment. All subjects in case and control groups were unrelated Han Chinese. This study was approved by the ethics committee of Fudan University, and informed consent was obtained from all participants.

HLA genotyping

Peripheral blood samples from all participants were collected using ethylenediaminetetraacetic acid-containing tubes. Genomic DNA was extracted using the AxyPrep Blood Genomic DNA Miniprep Kit (Axygen Biotechnology, Union City, CA, USA). HLA-A, -B and -C alleles were genotyped using the PCR sequencing of specific oligonucleotide (PCR-SSO) method and the LABT SSO Kit (One Lambda, Los Angeles, CA, USA). Genotyping was performed by Mentality Bio-Tech (Beijing, China).

Statistical analysis

Comparisons of the parametric data and categorical data were conducted using Student’s two-tailed t-test and Fisher's exact test, respectively. Odds ratios (ORs) were calculated using Haldane’s modification, which adds 0.5 to all cells to avoid zero counts.19 The statistical significance level was set at P<0.05. All statistical analyses were performed using R software version 3.0.2; The R Foundation for Statistical Computing, Vienna, Austria.

The power for detecting significant differences in HLA allele frequencies between patients and controls was estimated using the R-package statmod (version 1.4.20) (http://cran.r-project.org/web/packages/statmod). The proportion of methazolamide–SJS/TEN patients with HLA alleles (p1) and that of tolerant or general population controls (p2) were compared. Samples sizes were the numbers of patients (n1) and controls (n2). The significance level was set to 0.05. The alternative hypothesis was a one-tailed test because HLA alleles were assumed to be enriched in patients but not controls.

Molecular docking and dynamics simulation

To evaluate the potential interaction pattern between HLA-B*59:01 and methazolamide, molecular docking was performed between the two molecules followed by molecular dynamics simulation. The protein structure of HLA-B*59:01 was modeled based on existing crystal structure 1E27 (ref. 20) with 97% identity using MODELLER 9.11.21 The entire binding groove of HLA-B*59:01 protein was defined as the binding pocket using AutoDock Tools.22 Autodock Vina23 was implemented for in silico docking simulation using default parameters. The docked HLA–methazolamide complex structure models were solvated in a rectangular box of ~20 000 TIP3P water molecules,24 and the boundary of the box was at least 9 Å away from any solute atom. Counter Na+ ions were placed based on the Columbic potential to keep the entire system neutral. Particle mesh Ewald was employed to consider the long-range electrostatic interactions.25 Before molecular dynamics (MD) simulation, a series of minimizations was preformed: all hydrogen atoms were first minimized, followed by all water molecules with counter ions, and then a cycle of minimization was performed to relax all atoms without constraints. The maximum number of minimization steps was set to 5000, and the convergence criterion for the energy gradient was 0.1 kcal mol−1 Å−1. The SHAKE algorithm26 was employed to constrain all bonds involving hydrogen atoms in MD simulation, and the time step was set to 2 fs. The MD simulation consisted of a gradual temperature increase from 0 to 300 K over 100 ps and a 2-ns simulation for equilibration followed by a 3-ns simulation for data collection. All the above computations were performed using AMBER ff10 force field27 in AMBER1128 package.

Results

Of the eight patients enrolled in this study, six were diagnosed with TEN and two with SJS. Table 1 shows the clinical characteristics and HLA genotypes of the patients. The mean age of the patients was 50.8±11.1 years (range 33–67 years), and 62.5% of the patients (n=5) were male. The mean duration of methazolamide exposure was 22.0±16.9 days and ranged from 2 to 58 days. The latent period before the onset of symptoms varied from 12 to 58 days with a mean of 23.0±15.7 days.

Table 1 Clinical characteristics and HLA genotypes of patients with methazolamide-induced SJS/TEN
Full size table

To investigate whether the risk of methazolamide-induced SJS/TEN might be associated with specific HLA alleles, patients who were prescribed methazolamide without developing SCARs for more than 3 months after treatment were recruited as tolerant controls. Table 2 shows demographic and clinical variables of patients with methazolamide-induced SJS/TEN and tolerant patients. There were no significant differences in age, gender and methazolamide exposure between the methazolamide–SJS/TEN group and the methazolamide-tolerant group. The case patients received methazolamide for the treatment of postoperative intraocular pressure elevation and glaucoma, and all tolerant patients were prescribed methazolamide for the treatment of glaucoma.

Table 2 Demographic and clinical variables of patients with methazolamide-induced SJS/TEN and methazolamide-tolerant patients
Full size table

The association of HLA alleles with the occurrence of methazolamide-induced SJS/TEN was first assessed between methazolamide–SJS/TEN patients and tolerant controls. Significant associations were subsequently validated in a second, independent cohort of 283 healthy southern Han Chinese from the general population. HLA alleles that showed significantly different frequencies between cases and both control groups are listed in Table 3. Notably, HLA-B*59:01 was present in 87.5% (7/8) of the patients with methazolamide-induced SJS/TEN, 0% (0/30) of the methazolamide-tolerant controls (OR=305.0; 95% confidence interval (CI), 11.3–8259.9; P=6.3 × 10−7), and 0.35% (1/283) of the general population (OR=1974.0; 95% CI, 111.8–34 868.4; P=2.0 × 10−12). Compared with HLA-B*59:01, HLA-C*01:02 showed a significant but weaker association with methazolamide-induced SJS/TEN. HLA-C*01:02 is a common HLA allele in the Han Chinese population, and it was present in 87.5% (7/8) of the methazolamide–SJS/TEN patients, a significantly higher frequency than the 36.7% (11/30) of the methazolamide-tolerant patients (OR=12.1; 95% CI, 1.3–111.7; P=0.016) and 31.1% (88/283) of the general population (OR=15.5; 95% CI, 1.9–128.0; P=2.0 × 10−3). The distribution of HLA haplotypes in case and control groups was also analyzed. HLA-C*01:02 was closely linked to HLA-B*59:01, which formed the HLA-B*59:01-C*01:02 haplotype. A significant association was found between the occurrence of methazolamide-induced SJS/TEN and the HLA-B*59:01-C*01:02 haplotype, which was present in 87.5% of methazolamide–SJS/TEN patients but in none of tolerant patients and 0.35% of healthy Han Chinese.

Table 3 Frequencies of HLA-B*59:01, C*01:02 and the haplotype and association analyses with methazolamide-induced SJS/TEN
Full size table

The predictability of HLA alleles as markers for methazolamide-induced SJS/TEN was examined. HLA-B*59:01 exhibited high sensitivity (87.5%, 95% CI, 0.47–0.99) and specificity (100%, 95% CI, 0.86–1.00) in predicting the risk of SJS/TEN caused by methazolamide, whereas HLA-C*01:02 exhibited high sensitivity (87.5%, 95% CI, 0.67–0.99) but low specificity (63.3%, 95% CI, 0.44–0.79). In addition, the positive predictive value and negative predictive value of HLA-B*59:01 were 100% (95% CI, 0.56–1.00) and 96.8% (95% CI, 0.56–1.00), respectively.

In silico docking was conducted to characterize the interaction between HLA-B*59:01 and methazolamide, and a 5-ns atomistic MD simulation was addressed on the complex structure model of HLA-B*59:01 with methazolamide created by the molecular docking protocol. The minimum docking score of all eight possible binding poses of methazolamide and HLA-B*59:01 was −6.2. The receptor HLA and the ligand methazolamide exhibited a slight fluctuation over the simulation during the MD process (Figure 1a). However, the relative movement of methazolamide toward HLA-B*59:01 over the simulation was minor, which suggests a relatively stable binding pose of the HLA-B*59:01–methazolamide complex. Quantitative MM-PB/SA (molecular mechanics Poisson–Boltzmann surface area) analysis of the free-energy changes during HLA–methazolamide interaction showed that the total interaction energies ΔGtotal of methazolamide with HLA-B*59:01 was −26.9 kcal mol−1, which was further decomposed into van der Waals energy (−27.1 kcal mol−1), electrostatic potential (−21.4 kcal mol−1) and solvent effects (21.6 kcal mol−1). Figure 1b shows the binding pose of methazolamide and HLA-B*59:01 following MD simulation. Methazolamide tended to bind close to the A pocket of HLA-B*59:01, and two hydrogen bonds were formed from methazolamide toward residues TYR-9 and TYR-159 on HLA-B*59:01.

Figure 1
figure 1

The binding mode of methazolamide to HLA-B*59:01. (a) Superposition of HLA-B*59:01–methazolamide complex structures before and after molecular dynamics simulation. (b) The interaction details of methazolamide with HLA-B*59:01. Hydrogen bonds between methazolamide and amino acid residues on HLA-B*59:01 are indicated by black dashed lines.

PowerPoint slide

Full size image

Discussion

The highly polymorphic MHC genes in humans, also known as the HLA genes, encode cell-surface proteins that bind and present antigenic peptides to T cells, which trigger the acquired immune response. Strong associations between serious adverse events induced by certain drugs and specific HLA alleles were demonstrated, and these associations led to changes in drug labeling by the U.S. Food and Drug Administration (FDA) (http://www.fda.gov/Drugs/ScienceResearch/ResearchAreas/Pharmacogenetics/ucm083378.htm). For example, the FDA recommends pretherapy screening for the presence of HLA-B*15:02 and HLA-B*57:01 as risk factors for carbamazepine-induced SJS/TEN and abacavir-induced hypersensitivity, respectively.29, 30 Screening patients for risk HLA alleles before the initiation of drug therapies may provide an efficient way to reduce the risk of adverse drug events.

The HLA-B59 serotype/HLA-B*59:01 has been associated with methazolamide-induced SJS/TEN in Japanese and Korean populations.8, 12, 13, 14 The HLA-B59 serotype has also been detected in SJS caused by acetazolamide, another carbonic anhydrase inhibitor, in the Korean population.31, 32 Few cases of methazolamide-induced SJS have been reported in China, and the risk factors have not been identified. Our study is the first to perform high-resolution HLA genotyping in Han Chinese patients with methazolamide-induced SJS/TEN. We identified HLA-B*59:01 as a strong genetic risk factor for methazolamide-induced SJS/TEN in the Han Chinese population. The HLA-C*01:02 allele was also found to be significantly associated with methazolamide-induced SJS/TEN in this study. However, considering that this allele was closely linked with HLA-B*59:01, it could be encompassed by the association signal of the causal allele. Kim et al.14 demonstrated that HLA-C*01:02 and A*24:02 alleles were closely linked to HLA-B*59:01, and the B*59:01-C*01:02-A*24:02 haplotype had a significantly higher frequency in Korean patients with methazolamide-induced SJS/TEN than in the general population. Inconsistently, our study showed that HLA- A*24:02 was present in only 25% (2/8) of case patients, but 89 of 283 (31.45%) subjects in the general Han Chinese population carried the HLA-A*24:02 allele. Therefore, no significant association was found between HLA-A*24:02 and methazolamide-induced SJS/TEN, which suggests that HLA-A*24:02 may not contribute to the etiology of SJS and TEN caused by methazolamide treatment.

The racial/ethnic differences in associations between HLA genotypes and drug hypersensitivity may depend primarily on the frequencies of risk HLA alleles among ethnic groups. The allele frequency of HLA-B*59:01 (0.1–0.2%) in Han Chinese is ~10-fold lower than the frequency (1–2%) in Korean and Japanese populations.33 This difference may explain why cases of methazolamide-induced SJS/TEN are rarely reported in the Chinese population. Additionally, HLA-B*59:01 has rarely been found in other ethnic populations, including people of Caucasian and African descent.33 Despite the low frequencies of the risk allele HLA-B*59:01, there is still an urgent need to develop preventive strategies to minimize the risk of SCARs related to methazolamide, especially in Asian populations, considering their life-threatening potential.

For genetic studies of ADRs, especially rare ADRs, it would be difficult to collect large numbers of cases within a reasonable time frame. However, some ADRs are associated with large-effect genetic risk factors that can be identified using small numbers of affected individuals in large-scale genomic screens.34, 35, 36 For instance, as few as six cases and 200 population controls is sufficient to achieve an 80% power at a 2 × 10−5 significance level to detect the large genetic effect of HLA-B*15:02 on carbamazepine-induced SJS in genome-wide analyses.35, 37 Additionally, it has been suggested that the use of population controls in the study of rare ADRs (incidence <1%) does not have a significant impact on statistical power.35 In our study, we used 283 population controls, 30 clinical controls and 8 cases of methazolamide-induced SJS/TEN for the identification of genetic risk factors. Strong associations of methazolamide-induced SJS/TEN with HLA-B*59:01 were identified, which could be attributed to the large effect size of this rare allele. Our power analysis showed that the power and probability of detecting a true difference in the frequency of HLA-B*59:01 between patients and controls was 100%.

Although the mechanism by which HLA-B*59:01 contributes to SJS and TEN in patients using methazolamide has not been elucidated, it is noteworthy that the pathogenesis of drug-induced SCARs can be mediated by multiple determinants, including the drug, HLA allotypes, T-cell receptors, HLA-bound peptides and other co-stimulatory molecules.38 In the current study, the positive and negative predictive values of HLA-B*59:01 were 100% and 96.8%, respectively. These results suggest that HLA-B*59:01 is sufficient but not necessary to trigger methazolamide-induced SJS and TEN. None of the tolerant patients carried HLA-B*59:01, but there was also one case patient who did not carry the allele, which indicates that patients can develop methazolamide-induced SJS/TEN via other HLA factors.

The mechanism of how methazolamide interacts with HLA-B*59:01 remains unknown. It is unclear whether the interaction is mediated via methazolamide or any of its metabolites, and whether any drug-specific peptides are involved in the interaction. Although specific peptides or drug metabolites were not considered in the MD simulation,39 the results suggest a potential interaction mechanism of HLA-B*59:01 with methazolamide. Previous studies focusing on other drug–HLA interactions demonstrated that abacavir binds closely to the F pocket of HLA-B*57:01 as an altered repertoire,40, 41 and carbamazepine binds around the A–D pockets of HLA-B*15:02.40 The binding pattern of HLA-B*59:01 and methazolamide is more similar to that of carbamazepine and HLA-B*15:02. Because carbamazepine may directly interact with HLAs to activate T-cell receptors in accordance with the p-i concept,42 further studies can be designed to determine whether methazolamide directly interacts with HLA-B*59:01 to trigger immunological reactions using wet-lab approaches, such as binding experiments and in vitro immunological assays.42, 43

To our knowledge, this study is the first report of an association between HLA-B*59:01 and methazolamide-induced SJS/TEN in Han Chinese patients. The findings of our study and the case reports in Korea and Japan suggest that HLA-B*59:01 is a strong predictive marker for methazolamide-induced SJS and TEN in patients of Asian ancestry. Screening for HLA-B*59:01 before initiation of methazolamide therapy may reduce the risk of methazolamide-induced SCARs in Asians.