HLAs associated with perampanel-induced psychiatric adverse effects in a Korean population

Perampanel (PER) is a new-generation antiepileptic drug that has an occasional but significant shortcoming, psychiatric adverse effects (PAEs). Recently, antiepileptic drug-related adverse reactions, such as skin rash and even PAEs, have been discovered to be correlated with certain human leukocyte antigen (HLA) types. Thus, we aimed to analyze specific HLA alleles as risk factors for PER-PAEs. We prospectively enrolled 17 patients with epilepsy who were prescribed PER between May 2016 and Jul 2018 at Seoul National University Hospital and developed PAEs while taking PER. Their HLA types were analyzed compared to those of 19 patients in the PAE-tolerant group and the general Korean population. In silico docking was performed with two different computational programs, AutoDock Vina and SwissDock, to theoretically evaluate the binding affinity of PER in the grooves of the specific HLA alleles. The HLA-DQB1*06:01, DRB1*08:03, and B*54:01 alleles were significantly associated with the patients who developed PER-PAEs compared with the general Korean population (odds ratio [OR] 3.94, p = 0.008, OR 9.24, p = 0.037, and OR 3.25, p = 0.041, respectively). As a haplotype, the combination of the three alleles was significantly more frequent in the PER-PAE group than in both the PER-tolerant group and the general Korean population. DQB1*06:01 and B*54:01 also demonstrated higher docking scores with PER than other alleles. This is the first study to analyze the association of PER-PAEs with specific HLA genotypes. Our results suggest that an HLA-associated genetic predisposition and a possible immunological mechanism are involved in the occurrence of PER-PAEs.

Ethical publication statement. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Result
Characteristics and psychiatric manifestations of patients with PER-PAEs. A total of seventeen patients developed PER-PAEs (Table 1). Of the patients with PER-PAEs, seven (41%) were male, and the median age was 33 years [interquartile range 28-43 years]. Temporal lobe epilepsy was the most common indication for the administration of PER (13, 76%). The patients had suffered epilepsy for 17.6 ± 8.2 years, and until the onset of PAEs, PER was administered for 2.6 ± 2.3 months with a median dose of 4 mg [interquartile range 4-6 mg].
In silico docking. The binding affinity of the PER molecule to HLA alleles calculated by two different software programs is displayed in Table 3. HLA alleles causing PER-PAEs tended to have higher docking scores than alleles prevalent in the PER-tolerant group. In detail, PER was predicted to be docked into the groove of DQB1*06:01 ( Fig. 1) with a docking score of 10.7, which was higher than the docking scores of DQB1*05:03 and DQB1*04:01 (docking scores (− kcal/mol) 7.9 and 9.5, respectively, in AutoDock Vina). B*54:01 had a higher binding affinity than B*15:01 (docking scores (− kcal/mol) 9.6 vs. 9.2 in AutoDock Vina and 7.8 vs. 7.1 in Swiss-Dock).
However, DRB1*08:03, the allele significantly more frequent in the PER-PAE group than in both the PERtolerant group and the general Korean population, did not show stronger binding than DRB1*04:05 (docking scores (kcal/mol) 8.2 vs. 8.1 in AutoDock Vina and 7.6 vs. 7.6 in SwissDock).

Discussion
This is the first study to analyze the association of PER-PAEs with specific HLA genotypes. We demonstrated that the HLA-DQB1*06:01, DRB1*08:03, and B*54:01 alleles were associated with PER-PAEs. As a haplotype, the combination of the three alleles was significantly more frequent in the PER-PAE group than in both the PER-tolerant group and the general Korean population. Among them, DQB1*06:01 might be the allele most susceptible to PER-PAEs, since it was more frequent in the patients with more severe PAEs and had a higher docking score with PER than other alleles. Our research implies that HLA-associated genetic susceptibility could be involved in the occurrence of PER-PAEs.
In our study, the PER-PAEs were categorized according to the modified version of the Psychiatric Symptoms and Behavior Checklist of the Vanderbilt-Kennedy Center. Aggression was the most common PAE observed in our patients taking PER. This is consistent with previous reports 6,7 , and aggression dramatically increases the burden of epilepsy on patients and caregivers. Impulsivity, one of the second most common symptoms, becomes extremely dangerous when it is accompanied by self-injurious behavior 17,18 . Indeed, all three patients who attempted self-injurious behavior simultaneously showed impulsivity, and one male patient even attempted suicide, which resulted in his admittance to an intensive care unit. Therefore, in regard to the patients' safety and caregivers' burden, the importance of researching the risk factors for PER-PAEs cannot be emphasized enough.
Specific HLA genotypes are known to be risk factors for several AED adverse reactions. For instance, we have demonstrated the association between the cutaneous adverse reaction induced by lamotrigine and oxcarbazepine   www.nature.com/scientificreports/ with certain HLA types 12,19 . By showing in silico docking of the AED in the HLA groove, we have suggested the possibility of oxcarbazepine acting as a hapten in the allergic reaction. In the same regard, it is remarkable that PER-PAEs had significant associations with the HLA-DQB1*06:01, DRB1*08:03, and B*54:01 alleles, as LEV-PAEs were correlated with the specific HLA genotype A*11:01 9 . This implies that PAEs that develop in the brain could also be idiosyncratic adverse reactions induced by an immunological response. In patients with autoimmune encephalitis such as N-methyl-d-aspartate receptor-encephalitis and leucine-rich glioma-inactivated protein1-antibody encephalitis, it has been demonstrated that psychiatric manifestations can be induced by an autoimmune mechanism in the brain 10,20,21 . It might not be a coincidence that an association of house-dustmite-sensitive allergic rhinitis and the same HLA genotypes, DQB1*06:01 and DRB1*08:03, was demonstrated in Chinese subjects 22 . Moreover, in the Korean population, B*54:01 was associated with Kawasaki disease, an acute systemic vasculitis causing heart disease in children 23 . The in silico docking analysis supports this hypothesis. Remarkably, the docking scores of DRB1*06:01 and B*54:01 were higher than those of the other alleles that were prevalent in the PER-tolerant group. Since the docking score indirectly indicates a binding affinity of an antigen to HLA grooves, the higher scores imply stronger binding of PER to the particular HLA as the antigen or hapten causing the immunological reaction. It is notable that two different programs, AutoDock Vina and SwissDock, showed similar results for the docking scores. Thus, our analysis would reflect more objective and realistic results. On the other hand, the score of DRB1*08:03 was not much higher than those of the other alleles. DRB1*08:03 might be just a bystander of DQB1*06:01 or B*54:01, not the main cause of the immunological reaction that induces PER-PAEs. Indeed, both alleles, DQB1*06:01 and B*54:01, have high phenotypic frequencies (14% for HLA-DQB1*06:01-DRB1*08:03 and 2% for HLA-B*54:01-DQB1*06:01-DRB1*08:03, respectively) when composing the haplotype with DRB1*08:03 in the Korean population 11 . In our study, the haplotype analysis further demonstrated that HLA-DQB1*06:01-DRB1*08:03 and HLA-B*54:01-DQB1*06:01-DRB1*08:03 were concomitantly associated with PER-PAEs.
According to the high docking scores, the alleles DQB1*06:01 or B*54:01 could be the main risk factors involved in the PAE pathomechanism. However, it is remarkable that the two alleles have different antigen processing; DQB1*06:01 is a major histocompatibility complex (MHC) class II allele, interacting with CD4 + T cells through extracellular antigen processing, and B*54:01 is an MHC class I allele, interacting with CD8 + T cells through intracellular processing. In the clinic, CD4 + T cell involvement might be more plausible, in that PER-PAEs are fully reversible if the AED is discontinued, than a cytotoxic CD8 + T cell response, which results in irreversible damage to the target. Actually, for AED-related cutaneous adverse drug reactions, MHC class II molecules are associated with relatively mild maculopapular eruption, but MHC class I molecules are significantly correlated with an irreversible skin rash such as Stevens-Johnson syndrome and toxic epidermal necrolysis 12,24 . Although DQB1*06:01 was associated with more severe clinical manifestations, it would be noteworthy that the patients with PER-PAEs fully recovered without any sequelae after the discontinuation of PER.
Taken together, among the three alleles, we cautiously suggest that DQB1*06:01 could be the allele most susceptible to PER-PAEs. In addition to its high docking score and high frequency in the patients with PER-PAEs, DQB1*06:01 was more common in the patients with more severe psychiatric manifestations, such as selfinjurious behavior or psychosis, and the pathomechanism of MHC class II molecules might be more plausible than that of MHC class I molecules in that the clinical manifestation was fully reversible. Nevertheless, we should not completely exclude the possibility that HLA-B*54:01 and DRB1*08:03 are involved in the pathomechanism of PER-PAEs; DRB1*08:03 had a higher odds ratio in the PER-PAE group than in the PAE-tolerant group. Moreover, as previous results have shown, LEV-PAEs were associated with MHC class I molecules 9 , and it is not yet known exactly how AED-related cerebral adverse effects occur at the cellular level. The immune reaction in the central nervous system would be much more complex than in the skin. Beside the antigen presentation and CD8 + T cell activation, MHC class I molecules are known to perform nonclassical functions in association with multiple cytokines and may be engaged in controlling neuronal plasticity [25][26][27][28][29] . We hypothesize that the nonclassical function of MHC class I molecule could also be associated with a reversible phenotype, but this should be investigated in further studies. Future analysis with genome-wide association studies might also play a critical role in the identification of which HLA alleles would be important for PER-PAEs.
The study is limited by the small number of patients who were included in the PER-PAE group and the PAE-tolerant group. In addition, the analysis was not controlled for co-medications. However, to overcome this limitation, patients with PER-PAEs were strictly selected according to certain criteria, and the PER-tolerant group was qualified to be an extreme phenotype with a daily dose of 10 mg per day for 6 months. Moreover, the PER-PAEs were analyzed using an objective tool, the modified version of the Psychiatric Symptoms and Behavior Checklist of the Vanderbilt-Kennedy Center. The results will need to be validated in an independent large cohort. Lastly, the HLA allele frequency differs according to ethnicity so that further investigation is required in multiple ethnic groups.
In summary, our study shows that PER-PAEs were associated with HLA-DQB1*06:01, DRB1*08:03, and B*54:01 and their combinations as haplotypes in the Korean population. Among the alleles, DQB1*06:01 could be the most likely HLA type correlated with the immunological reaction of PER-PAEs. As HLA-related genetic risk factors for each ethnicity are important in and of themselves, the results of this study would be helpful as a reference to investigate potential screening methods for Korean patients who will be prescribed PER. Moreover, our results of the HLA-related genetic susceptibility of PER-PAEs suggest that psychiatric symptoms induced by adverse drug effects could be induced by an immunological reaction in the brain. Therefore, further research should be performed with a larger number of patients in terms of the pathomechanism of PAEs in the near future.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.