Risk of drug-induced liver injury in multiple system atrophy patients treated with KM-819 is potentially associated with the HLA-DPB1*05:01:01:01 allele

Abstract

KM-819, a first-in-class inhibitor of Fas-associated factor 1 (FAF1), was investigated in a Phase 2 clinical trial in Korean patients with multiple system atrophy (MSA). Among 34 KM-819-treated patients, 11 met the case definition for suspected DILI, and 9 of these (26%) were adjudicated by an independent hepatic adjudication committee (HAC) as KM-819-related DILI; no cases occurred among the 34 patients receiving placebo. Blood samples for genetic analysis were available from only 6 of the 9 adjudicated DILI cases; therefore, an exploratory comparative genetic analysis was performed using DNA samples from a total of 16 patients treated with KM-819, including the 6 available DILI cases and 10 patients without DILI. Whole-genome sequencing (WGS) identified the HLA-DPB1*05:01:01:01 allele in all 6 patients with DILI, compared to only 1 of the 10 patients without DILI (p = 0.0009), suggesting a potential association between HLA-DPB1*05:01:01:01 and DILI due to KM-819. In this discovery cohort, HLA-DPB1*05:01:01:01 showed 100% sensitivity, 90.0% specificity, a positive predictive value (PPV) of 85.7%, and a negative predictive value (NPV) of 100% for DILI due to KM-819, albeit with wide confidence intervals reflecting the small sample size. Given the small sample size, exploratory nature of the analysis, and use of the same cohort for both discovery and performance estimation, these findings should be considered hypothesis-generating. Therefore, HLA-DPB1*05:01:01:01 should be regarded as a candidate risk marker specific to this Korean cohort, requiring validation in larger, independent, and ethnically diverse cohorts before its application in clinical risk stratification.

Introduction

KM-819 is a small-molecule inhibitor of Fas-associated factor 1 (FAF1), a key mediator in promoting α-synuclein accumulation and neuronal cell death [1–3]. By targeting FAF1, KM-819 exerts neuroprotective effects through inhibition of both caspase-dependent and caspase-independent cell death pathways, thereby protecting both oligodendrocytes and neurons. In addition, KM-819 facilitates autophagic clearance of α-synuclein, a process otherwise suppressed by FAF1, suggesting its potential as a disease-modifying therapy for multiple system atrophy (MSA). MSA is a rare, progressive neurodegenerative disorder characterized by autonomic failure, parkinsonism, and cerebellar ataxia [4]. As there are no approved disease-modifying treatments for MSA, clinical care remains limited to symptomatic management such as medications for parkinsonism and autonomic failure. Thus, there is a major medical need in this condition which is associated with severe morbidity and mortality. Given the therapeutic potential of KM-819 for MSA, a clinical trial was conducted to evaluate its efficacy as a novel treatment option.

Study KMCP-819-K102 was a randomized, double-blind, placebo-controlled phase 2 clinical trial conducted in South Korea to evaluate the efficacy and safety of KM-819 in patients diagnosed with MSA (ClinicalTrials.gov ID: NCT05695378). During the trial, elevated liver enzymes, including aspartate aminotransferase (AST), alanine aminotransferase (ALT), and total bilirubin, were observed in 11 of the 34 patients who received KM-819, whereas no such elevations were observed in the placebo group, raising concerns about potential drug-induced liver injury (DILI) defined as an elevation of ALT or AST to ≥3-times upper limit of normal (ULN) or of alkaline phosphatase (ALP) or total bilirubin to ≥2-times ULN.

Following an independent review by a hepatic adjudication committee (HAC) comprising hepatology experts, 9 of the 11 cases were adjudicated as KM-819-related DILI: One case did not fully meet the criteria for DILI, while the other lacked essential blood biochemistry data, including bilirubin, for adjudication.

A comprehensive evaluation was conducted to identify potential causes of DILI, including gender, concomitant medication use, and duration of KM-819 exposure. However, none of these factors including gender, any specific concomitant medication, or treatment duration, showed meaningful differences in their correlation with DILI in either the overall cohort or the KM-819 treatment group. Given that there were no clear associations with the occurrence of KM-819-related DILI, to elucidate the underlying mechanisms of KM-819-related DILI, whole genome sequencing (WGS) was performed. Because DILI is known to be associated with multiple biological mechanisms, including immune and metabolic processes [5], the WGS data were analyzed to evaluate their association with known mechanisms of DILI. Due to a progressive worsening of their clinical status or refusal to provide consent, blood samples were available for genetic testing from only 6 of the 9 MSA patients who developed DILI (DILI patients). For comparison, their data were compared with those from 10 of the 25 patients who did not exhibit DILI (non-DILI patients).

Materials and methods

Clinical trial design

This was a multi-center double-blind randomized controlled trial based in South Korea between 2023–2024. The study protocol for biospecimen collection and genetic analysis was reviewed and approved by the Institutional Review Board (IRB) of CHA Bundang Medical Center, CHA University (Seongnam-si, Korea, IRB No. 2025-02-061-005). All patients provided written informed consent. Eligible patients were adults aged 30–80 years with probable or possible MSA, able to attend clinic visits during the study, and with stable antiparkinsonian medications for at least one month prior to enrollment. Additional inclusion requirements included a BMI of 18.5–30 kg/m², cognitive ability to provide informed consent, and adherence to acceptable contraceptive methods as appropriate. Patients requiring wheelchair assistance were permitted if accompanied by a caregiver. Exclusion criteria included drug-induced parkinsonism, pregnancy or lactation, recent suicidal ideation or history, febrile or symptomatic infections, clinically significant abnormalities at screening, cardiac conduction issues (e.g., QTc prolongation), positive serology for hepatitis B, C, A, or HIV, hypersensitivity to KM-819 or its components, serious medical or surgical conditions, inability to understand the consent form, inability to attend clinic visits, and prior brain surgery.

Following a 4-week screening period, participants exhibiting clear signs of parkinsonism, regardless of the MSA subtype, were enrolled in the KMCP-819-K102 study and randomized in a 1:1 ratio to receive either KM-819 (400 mg) or placebo. The study was conducted at a single site (CHA Bundang Medical Center). All participants were of Korean descent. A total of 68 participants enrolled with 34 subjects allocated to each treatment arm. The blinded treatment period was planned to be 36 weeks. KM-819 or placebo was administered orally once daily (QD), commencing at the baseline visit. Efficacy assessments were scheduled to be conducted at weeks 12, 24, and 36 ( ± 14 days) to evaluate the primary and secondary endpoints and safety labs were assessed at each time point. A safety follow-up visit was planned for week 40. Based on the potential safety findings, all participants were withdrawn from treatment and ultimately the study was terminated.

Definition and assessment of DILI

In patients enrolled to the clinical trial with normal or near normal liver tests at baseline, an increase of serum ALT or AST to ≥3-times ULN or ALP or total bilirubin to ≥2-times ULN was used as the case definition of suspected DILI [6]. In addition to standard clinical care, retrospective causality assessment was performed by HAC. Each case was independently assessed by three adjudicators using all available clinical information and documentation, including treatment allocation. Severity of DILI was assessed according to the International DILI Expert Working Group classification [7]. When all adjudicators reached the same conclusion, the determination was considered final. If their assessments differed, a reconciliation discussion was held.

Next generation sequencing

After written informed consent was obtained from all patients, peripheral blood samples were collected into EDTA tubes at CHA Bundang Medical Center, and subsequently transferred to Macrogen (Seoul, South Korea) for genomic DNA extraction and analysis. Genomic DNA was extracted using the Maxwell® 16 Blood RSC DNA Purification Kit (Promega, Madison, USA). WGS was performed on genomic DNA first using short-read next generation sequencing (NGS) technology (NovaSeq platform, Illumina, San Diego, USA) [8]. One microgram of genomic DNA was fragmented using Covaris systems (Covaris, Woburn, USA) to generate 150–300 base pair (bp) double-stranded DNA fragments. The prepared libraries were then sequenced on an Illumina platform according to the manufacturer’s protocols.

Short-read NGS-based WGS was initially performed as an exploratory analysis to screen for genetic variants potentially associated with DILI. During the initial screening, an HLA locus of interest was identified. To achieve more precise allele-level characterization in this highly polymorphic and structurally complex region [9], the HLA gene region was PCR-amplified and further analyzed using long-read NGS technology (Revio platform, PacBio, Menlo Park, USA) [10]. For long-read NGS, DNA samples were amplified at 11 HLA loci (HLA-A, -B, -C, -DPA1, -DPB1, -DQA1, -DQB1, -DRB1, -DRB3, -DRB4, and -DRB5) using the NGSgo-MX11-3 primer mix (GenDx, Utrecht, the Netherlands), generating 2.5–6.7 kilobases (kb) amplicons. SMRTbell libraries (PacBio, Menlo Park, USA) were prepared and sequenced on the PacBio Revio platform [11].

Variant analysis

DNA sequences were aligned to the GRCh38.p14 human reference genome to identify genetic variants [12]. Variant calling and filtration were performed using the Genome analysis toolkit (GATK) variant filtration module to accurately detect genetic variants and reduce technical errors [13]. Identified variants were annotated with SnpEff [14]. Additional filtering was applied using publicly accessible variant databases including dbSNP and the 1000 Genomes Project, to distinguish known polymorphisms from potentially novel or rare variants relevant to the study [15].

For statistical analysis, numerical variables were compared using a two-tailed Welch’s t-test, while categorical variables were evaluated using two-tailed Fisher’s exact test. Differences in characteristics between the DILI and non-DILI patient groups were considered statistically significant at a p-value threshold of < 0.05. For genetic analysis, variants identified by short-read NGS were compared between DILI and non-DILI patients using a two-tailed Fisher’s exact test. Odds ratios (ORs), 95% confidence intervals (CIs), and p-values were calculated using R (version 4.4.2). CIs for sensitivity and specificity were calculated using the exact Clopper–Pearson method [16]. CIs for positive predictive value (PPV) and negative predictive value (NPV) were calculated using the standard logit method [17], except when PPV or NPV was 0% or 100%, in which case the exact Clopper–Pearson method was used. A dominant genetic model was assumed for each variant to determine the statistical significance of allele frequency differences between the two groups. Subsequently, long-read NGS identified the allele with even greater precision, down to the fourth field.

Results

A total of 68 participants were enrolled in the KMCP-819-K102 study and randomized in a 1:1 ratio to receive either KM-819 or placebo, with 34 patients assigned to each group. Patients with positive serology for hepatitis B, C, A, or HIV, or with serious medical conditions were excluded during the screening period although no further testing was performed subsequently. Table 1 shows the demographic data for the participants who enrolled in the study.

Table 1.

Demographics of patient participants enrolled in study KMCP-819-K102.

Characteristics	KM-819 group (N = 34)	Placebo group (N = 34)	Total (N = 68)
Age (years)
Mean ± SD	62.0 ± 6.92	61.8 ± 7.52	61.9 ± 7.17
Median	63.5	62.5	63.0
Min-Max	44–77	49–75	44–77
Age group
30 - <65 years (%)	20 (58.8)	22 (64.7)	42 (61.8)
65 - 80 years (%)	14 (41.2)	12 (35.3)	26 (38.2)
Sex
Female (%)	22 (64.7)	16 (47.1)	38 (55.9)
Male (%)	12 (35.3)	18 (52.9)	30 (44.1)
Race
Asian	34 (100)	34 (100)	68 (100)
Height at screening (cm)
Mean ± SD	159.8 ± 9.50	161.6 ± 8.60	160.7 ± 9.04
Median	159.5	162.5	160.0
Min-Max	143–176	146–183	143–183
Weight (kg)
Mean ± SD	60.49 ± 11.502	61.37 ± 9.859	60.93 ± 10.641
Median	59.00	61.75	60.35
Min-Max	39.8–86.7	42.2–83.2	39.8–86.7
BMI (kg/m2)
Mean ± SD	23.56 ± 3.001	23.41 ± 2.820	23.49 ± 2.891
Median	23.45	22.90	23.10
Min-Max	18.7–29.2	19.4–29.5	18.7–29.5
MSA type
MSA-P (%)	24 (70.6)	23 (67.6)	47 (69.1)
MSA-C (%)	10 (29.4)	11 (32.4)	21 (30.9)

Age was calculated based on the date of sample collection. MSA-P parkinsonian subtype, MSA-C cerebellar subtype, BMI body mass index, SD standard deviation.

Among the 34 patients treated with KM-819, 11 had abnormal liver function tests (LFTs; ALT or AST > 3× the ULN). Considering potential safety findings, study medication (KM-819 or placebo) was discontinued for all participants. These 11 patients underwent hepatitis B virus testing, and all results were negative. Participant characteristics at the time the study was halted are shown in Table 2.

Table 2.

Demographic data of patient participants at termination of study KMCP-819-K102.

Disposition	KM-819 Group N (%)	Placebo Group N (%)	Total N (%)
Started the main study	34 (100)	34 (100)	68 (100)
On study at time of study discontinuation	25	31	56
Discontinued from the main study	9 (26.5)	3 (8.8)	12 (17.6)
Primary reason for discontinuation*
1. Entry criteria violation	0	1 (33.3)	1 (8.3)
2. Adverse event	4 (44.4)	0	4 (33.3)
3. Investigator’s decision	1 (11.1)	0	1 (8.3)
4. Death	2 (22.2)	2 (66.7)	4 (33.3)
5. Withdrawal by Subject	2 (22.2)	0	2 (16.7)
Last Visit Completed at Time of Study Discontinuation
Baseline	19 (55.9)	14 (41.2)	33 (48.5)
Week 12	10 (29.4)	15 (44.1)	25 (36.8)
Week 24	2 (5.9)	2 (5.9)	4 (5.9)
Week 36	3 (8.8)	3 (8.8)	6 (8.8)

Values are presented as number of patients (N) and percentage (%).*Percentages for reasons are based on the number of subjects discontinued from the main study.

Eleven patients met the case definition for suspected DILI among participants who received KM-819 (n = 34), whereas no cases of DILI were observed in the placebo group (n = 34) (Table 3). Following HAC adjudication, 9 of the 11 suspected DILI cases were considered related to KM-819.

Table 3.

Clinical data of 11 patients with elevated liver enzyme levels.

Case No.	Age/Sex	BMI (kg/m2)	Highest fold elevation*				DILI latencya	KM-819 W/Da	Steroid Txa	Day to Recoveryb	HAC adjudicationc	Final outcome
			AST	ALT	ALP	Total Bilirubin				AST/ALT
2013d	60/M	28.5	13.9	16.3	3.4	12.8	144	NA	150	179/179	Y, PR, S	-
2016	67/F	23.6	3.2	4.2	3.4	0.6	91	155	-	204/91	Y, PR, M	-
2025	51/F	29.2	16.4	11.6	4.1	1.7	84	84	-	105/105	Y, PR, M	-
2041	62/F	23.4	2.4	1.3	1.8	0.7	85	98	-	132/132	N, PR, M	-
2054e	64/M	22.1	22.4	37.1	2.1	3.5	68	74	-	-/112	Y, PR, MO	D
2064	56/F	24.1	4.1f	6.5f	2.1	0.6	103	117	148	166/166	Y, PR, M	-
2066	56/F	18.8	18.1	36.5	4.0	3.9	31	30	6	-/-	Y, PR, F	D
2067	65/F	20.8	23.7	33.3	0.3	1.2	77	77	84	91/91	Y, PR, M	-
2071	70/M	23.5	25.9	36.2	4.1	15.8	74	74	77	89/89	Y, PR, S	-
2075	65/F	23.7	13.2	8.5	1.5	0.4	74	74	87	94/94	Y, PR, M	-
2082	61/F	19.1	5	1.4	1.0	0.4	28	30	72	71/-	ID, PR, ID	D

NA not applicable, Y adjudicated as DILI, N not adjudicated as DILI, PR partial related, ID insufficient data, S severe, M mild, MO moderate, F fatal, D deceased.

*Aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP) and total bilirubin levels are expressed as the highest fold elevation above the upper limit of normal (ULN).

^aValues indicate the days when liver injury was diagnosed (DILI latency), KM-819 was withdrawn (KM-819 W/D), and steroid treatment was initiated (Steroid Tx).

^bThe first value represents the day when AST/ALT levels returned to <3 × ULN.

^cThe columns represent DILI adjudication, causality, and severity, respectively.

^dCase 2013 was treated with pulse steroid in another hospital without knowledge of his participation in the KM-819 trial. His wife, caretaker during hospitalization, kept giving him KM-819. It was on day 272 when the study team was informed of the incident and the participant discontinued KM-819. He was initially suspected to have pneumonia but was subsequently adjudicated as hepatitis.

^eParticipant 2054 was diagnosed with DILI prior to the diagnosis and death from COVID-19 pneumonia. However, the cause of death, as assessed by the PI, was pneumonia, not liver injury. Participant continued to receive KM-819 until the Investigator was informed about participant situation (quarantined at another hospital for COVID pneumonia).

^fHighest values observed at Day 149.

To investigate potential genetic factors associated with KM-819-related DILI, peripheral blood samples were collected for genomic DNA extraction. However, samples were available from only 6 of the 9 adjudicated DILI patients (Cases 2013, 2016, 2025, 2064, 2067, and 2071), as two patients were unable to provide blood due to declining health (Cases 2054 and 2066) and one patient declined blood collection (Case 2075). Among the six available DILI cases, the mean latency to DILI onset was 95.5 days. The mean peak AST, ALT, ALP, and total bilirubin levels were 14.5, 18.0, 2.9, and 5.5 × ULN, respectively. The mean recovery times, defined as the intervals from the initial DILI diagnosis until AST and ALT levels returned to <3 × ULN, were 139.0 and 120.2 days, respectively. No alternative causes of liver injury were identified in any of the six cases.

The first DILI case was detected on day 40 after treatment initiation. At that time, LFTs were performed for all patients, and eight additional cases were identified. The latent period of DILI among these patients ranged from 28–144 days. Only one patient had a latent period of 144 days (longer than 3 months), while the others developed DILI by or before the 3-month mark. To reduce the risk of misclassifying patients who might later develop DILI, non-DILI cases were restricted to those who had received KM-819 treatment for at least 3 months. Accordingly, patients who discontinued KM-819 within 3 months due to skin rash (n = 2), pneumonia (n = 2), or spontaneous withdrawal (n = 3), as well as those who discontinued treatment within 3 months following the IRB decision (n = 5), were excluded, leaving 13 eligible non-DILI patients for genetic analyses. Among these 13 patients, 10 were selected for WGS by simple random sampling [18] (Fig. 1). Clinical characteristics, including age, sex, MSA subtype, and BMI, did not differ significantly between the selected and non-selected non-DILI patients (Supplementary Table 1).

Fig. 1. Patient selection flowchart for genetic analysis in the KMCP-819-K102 trial.

DNA was genotyped from 16 participants marked with *, including 10 randomly selected non-DILI cases and 6 available DILI cases.

The clinical characteristics of the 16 patients included in the genetic analysis are summarized in Table 4, which also shows that cases 2013 and 2071 developed jaundice.

Table 4.

Clinical characteristics of 16 KM-819-treated subjects who underwent genetic analysis.

Group	Case No.	Age/Sex	MSA type	Duration of KM-819 (weeks)	Highest fold elevation of ULN			Jaundice
					AST	ALT	Total bilirubin
DILI	2013	60/M	P	32.7	13.9	16.3	12.8	Yes
	2016	67/F	P	22.1	3.2	4.2	0.6	No
	2025	51/F	P	11.9	16.4	11.6	1.7	No
	2064	58/F	C	15.6	4.1	6.5	0.6	No
	2067	67/F	P	10.9	23.7	33.3	1.2	No
	2071	72/M	P	10.4	25.9	36.2	15.8	Yes
Non-DILI	2002	64/M	C	36.9	0.7	1.0	0.8	-
	2007	61/F	P	35.7	0.7	0.9	0.6	-
	2011	59/M	P	35.3	0.4	0.4	0.7	-
	2014	64/M	C	33.1	3.8	2.6	0.6	-
	2033	70/M	P	26.1	0.7	0.3	0.8	-
	2043	45/F	P	22.9	1.2	1.7	0.7	-
	2050	53/F	C	21.9	0.6	0.9	1.0	-
	2052	66/F	C	21.1	0.5	0.3	0.3	-
	2068	64/M	P	16.3	0.6	1.0	0.5	-
	2081	54/M	C	15.7	1.3	0.8	0.7	-

The clinical characteristics of each patient are listed, including age, sex, DILI status, MSA subtype (P: MSA-P, C: MSA-C), and KM-819 treatment duration (in weeks) from baseline. Age was calculated based on the date of sample collection. * AST, ALT and total bilirubin levels are expressed as the highest fold elevation above the ULN.

An analysis of the clinical characteristics of the patients did not reveal any association with age, sex, MSA subtypes, BMI, KM-819 duration, and follow-up duration between groups (Table 5).

Table 5.

Clinical characteristics in KM-819-treated DILI vs. non-DILI patients.

Characteristics	DILI (n = 6)	non-DILI (n = 10)	p-value
Age (years)	62.5 ± 7.6	60.0 ± 7.4	0.535
Sex (Female/Male)	4/2	4/6	0.608
MSA type (P/C)	5/1	5/5	0.307
BMI (kg/m2)	24.9 ± 3.2	24.8 ± 2.6	0.923
Duration of KM-819 (weeks)	17.3 ± 8.7	26.5 ± 8.2	0.062
Follow-up duration (weeks)	29.5 ± 11.2	32.0 ± 9.0	0.649

Age, BMI, duration of KM-819 administration and follow-up duration are presented as mean ± standard deviation values for each group. Differences between DILI and non-DILI groups were assessed using Welch’s t-test. The number of sex and MSA type were compared between groups using Fisher’s exact test.

Exploratory WGS was performed using genomic DNA from the 16 patients with MSA. Variants with significant functional impacts, such as missense mutations and frameshift variants, were identified. Among these, variants with a significantly higher frequency in DILI patients were selected for further analysis. Fisher’s exact test was employed to assess allele frequency differences between the DILI and non-DILI patient groups, and variants were ranked by p-value. Genes harboring variants significantly enriched in the DILI group, defined as occurring in at least 5 of the 6 DILI patients but in no more than 1 of the 10 non-DILI patients, were ranked by p-value using a p-value cut-off of 0.002 (Table 6). Six genes exhibited a p ≤ 0.002 in the DILI group compared to the non-DILI group (Table 6), including ankyrin repeat and ubiquitin domain containing 1 (ANKUB1), human leukocyte antigen DP beta 1 (HLA-DPB1), COMM domain containing protein 2 (COMMD2), peptidyl arginine deiminase 3 (PADI3), secernin 3 (SCRN3), and phospholipase A2 group IVF (PLA2G4F).

Table 6.

Significantly enriched gene variants in MSA Patients with KM-819-related DILI identified by short-read NGS.

Gene symbol	HGVS.c	HGVS.p	DILI / non-DILI*	p-value	OR (95% CI)
ANKUB1	c.1156 A > G	p.Ser386Gly	6/0	0.0001	Inf (6.25-Inf)
HLA-DPB1	c.700 G > A	p.Val234Met	6/1	0.0009	Inf (3.33-Inf)
COMMD2	c.337 A > C	p.Ile113Leu	6/1	0.0009	Inf (3.33-Inf)
PADI3	c.154 A > G	p.Ile52Val	5/0	0.0014	Inf (3.00-Inf)
SCRN3	c.1230_1242delTTTATCAGTCAAA	p.Asn410fs	5/0	0.0014	Inf (3.00-Inf)
PLA2G4F	c.2396 C > G	p.Pro799Arg	5/0	0.0014	Inf (3.00-Inf)

Gene variants significantly enriched in the DILI group compared to the non-DILI group ranked by p-value. Variants with p-values below 0.002 were considered key candidates. For each variant, the table lists the gene name, HGVS.c (Human Genome Variation Society, DNA-level change), HGVS.p (protein-level change), and the p-value. HGVS nomenclature is used to describe DNA and protein sequence variants [49]. *The number of patients carrying each variant in the DILI or non-DILI group.

The six candidate genes identified from the WGS analysis were further evaluated based on their reported associations with DILI. ANKUB1 exhibited the lowest p-value (0.0001), suggesting the strongest statistical association with DILI. However, no previous studies have implicated ANKUB1 in drug-induced liver injury or other drug-related adverse effects. Similarly, no evidence linking COMMD2, PADI3, SCRN3, or PLA2G4F to drug-induced liver injury was identified in the literature. In contrast, HLA-DPB1, which showed the second lowest p-value (0.0009) in Table 6, was identified as a gene with known associations with drug-induced hepatotoxicity in the ClinPGx database which provides clinical annotations summarizing the relationship between genetic variants and drug responses [19]. Specifically, HLA-DPB1, with the Val234Met (700 G > A) variant identified in this study, has been previously reported to be associated with drug-induced hepatotoxicity [20].

Because immune processes are known to play a prominent role in DILI, the six candidate genes were further evaluated using ImmPort, an immunology bioinformatics database [21]. Among these six candidate genes, only HLA-DPB1 was associated with immune-related genes. Combined with its known association with DILI, HLA-DPB1 was selected as the candidate gene for subsequent analyses.

The HLA-DPB1 gene is located within the human leukocyte antigen (HLA) haplotype, a cluster of genes spanning ~3.8 Mbp on chromosome 6 that is typically inherited together [22]. The HLA haplotype encompasses class I and class II HLA genes, along with more than 120 non-HLA genes. Class I HLA molecules present endogenous antigens derived from intracellular proteins, whereas class II HLA molecules present exogenous antigens originating from extracellular sources. Class I HLA molecules are composed of a heavy chain which contains the peptide-binding cleft, and a light chain, β2-microglobulin, which provides structural stability [23]. In contrast, class II HLA molecules consist of two HLA chains, an α and a β chain, which together form the peptide-binding cleft. HLA genes have been implicated in DILI in over 40 previous reports [19]. However, the HLA haplotype exhibits structural complexity, including repeat elements and gene duplications, while the high degree of polymorphism among HLA genes presents challenges for accurate sequencing [22]. In particular, the HLA-DPB1 gene exhibits extensive polymorphism, with 2 890 nucleotide-level alleles identified. After excluding synonymous variants, these alleles correspond to 112 distinct protein isoforms [24]. Given the extensive polymorphism of HLA genes and the challenges in accurately resolving HLA alleles from WGS data, including potential ambiguity arising from homologous genomic regions, we performed targeted amplification of 11 HLA loci followed by long-read NGS to achieve precise, high-resolution allele characterization at the fourth-field level (Table 7 and Table 8). HLA alleles with different nucleotide sequences can encode identical amino acid sequences [24]. In accordance with HLA nomenclature, HLA alleles are named using four fields separated by colons. The first field denotes the allele group, typically corresponding to a serologically defined antigen. The second field specifies coding sequence differences that result in amino acid changes. The third field indicates synonymous nucleotide substitutions that do not alter the amino acid sequence. The fourth field represents differences in non-coding regions such as introns or untranslated regions. All 6 DILI patients carried at least one HLA-DPB1*05:01:01:01 allele, with patient case No. 2025 being homozygous for this allele. Among the non-DILI patients, case No. 2002 also carried one HLA-DPB1*05:01:01:01 allele (Table 7). Thus, the HLA-DPB1*05:01:01:01 was the only allele consistently present in all 6 DILI patients, whereas this allele was found in just 1 of the 10 non-DILI patients.

Table 7.

HLA-DPB1 alleles in MSA patients treated with KM-819, identified via long-read NGS.

Group	Case No.	HLA-DPB1 allele
DILI	2013	05:01:01:01	02:01:02:01
	2016	05:01:01:01	02:02:01:01
	2025	05:01:01:01	05:01:01:01
	2064	05:01:01:01	02:02:01:01
	2067	05:01:01:01	17:01:01:01
	2071	05:01:01:01	09:01:01:01
non-DILI	2002	05:01:01:01	02:01:02:01
	2007	02:01:02:01	04:02:01:02
	2011	02:01:02:01	02:02:01:01
	2014	02:01:02:01	13:01:01:02
	2033	02:01:02:01	13:01:01:01
	2043	02:01:02:01	13:01:01:05
	2050	02:01:02:11	04:01:01:01
	2052	02:02:01:01	04:02:01:02
	2068	02:01:02:73	13:01:01:02
	2081	02:01:02:32	04:01:01:01

HLA typing was performed on 16 MSA patients using long-read NGS. For each patient, the two HLA-DPB1 alleles are shown in the table, with alleles resolved down to the fourth field.

Table 8.

HLA typing results of DILI patients by long-read sequencing.

Case No.	HLA-A	HLA-B	HLA-C	HLA-DPA1	HLA-DPB1	HLA-DQA1	HLA-DQB1	HLA-DRB1	HLA-DRB3	HLA-DRB4	HLA-DRB5
2013	02:01:01:01	15:18:01:02	01:02:01:01	02:02:02:01	02:01:02:01	01:02:01:01	06:01:01:01	08:03:02:01			01:01:01:02
	02:06:01:01	40:06:01:01	08:01:01:01	02:02:02:01	05:01:01:01	01:03:01:03	06:02:01:01	15:01:01:01
2016	11:01:01:01	40:06:01:01	08:01:01:01	02:02:02:01	02:02:01:01	01:02:01:01	04:02:01:07	08:02:01:01			01:01:01:02
	31:01:02:01	51:01:01:01	14:02:01:01	02:02:02:11	05:01:01:01	04:01:01:01	06:02:01:01	15:01:01:01
2025	02:01:01:01	15:01:01:01	01:02:01:01	02:02:02:01	05:01:01:01	01:04:01:01	04:01:01:03	04:05:01:01	02:02:01:03	01:03:01:05
	02:01:01:01	48:01:01:01	08:22:01:01	02:02:02:01	05:01:01:01	03:03:01:03	05:02:01:01	14:54:01:01
2064	02:01:01:01	15:01:01:01	03:03:01:01	01:03:01:01	02:02:01:01	03:02:01:01	03:03:02:02	09:01:02:01		01:03:02:01
	24:02:01:01	40:02:01:01	03:03:01:01	02:02:02:01	05:01:01:01	03:02:01:02	03:03:02:03	09:01:02:01
2067	02:06:01:01	13:02:01:01	06:02:01:01	02:01:01:03	05:01:01:01	01:04:01:01	02:02:01:01	07:01:01:01	02:02:01:03	01:03:01:01
	11:01:01:01	51:01:01:01	14:02:01:01	02:02:02:01	17:01:01:01	02:01:01:01	05:03:01:01	14:05:01:01
2071	01:01:01:01	51:01:01:01	03:04:01:02	02:01:01:02	05:01:01:01	01:02:01:01	03:03:02:01	07:01:01:01		01:03:01:02 N	01:01:01:02
	31:01:02:01	57:01:01:01	06:02:01:01	02:02:02:01	09:01:01:01	02:01:01:01	06:02:01:01	15:01:01:01
2002	11:01:01:01	14:01:01:01	08:02:01:02	01:03:01:01	02:01:02:01	03:01:01:01	03:02:01:01	08:02:01:01		01:03:02:01
	30:04:01:01	51:01:01:01	14:02:01:01	02:02:02:01	05:01:01:01	03:02:01:01	03:03:02:02	09:01:02:01
2007	30:01:01:01	13:02:01:01	03:03:01:01	01:03:01:01	02:01:02:01	03:01:01:01	03:01:01:12	04:06:01	03:01:03:02	01:03:01:01
	31:01:02:01	40:02:01:01	06:02:01:01	01:03:01:05	04:02:01:02	06:01:01:01	03:02:01:01	12:02:01:01
2011	02:06:01:01	40:02:01:01	01:02:01:01	01:03:01:01	02:01:02:01	01:03:01:03	03:01:01:03	08:03:02:01	02:02:01:02
	24:02:01:01	46:01:01:01	03:04:01:02	02:02:02:01	02:02:01:01	05:05:01:01	06:01:01:01	11:01:01:01
2014	02:06:01:01	15:01:01:01	03:03:01:01	01:03:01:01	02:01:02:01	01:02:01:08	03:02:01:01	04:03:01:01	03:01:01:01	01:03:01:01
	26:02:01	44:03:01:01	14:03:01:01	02:01:01:01	13:01:01:02	03:01:01:01	06:04:01:01	13:02:01:02
2033	03:01:01:01	44:03:02:01	07:06:01:01	02:01:01:06	02:01:02:01	01:02:01:01	02:02:01:01	07:01:01:01		01:03:01:01	01:01:01:02
	33:03:01:01	51:01:01:01	14:02:01:01	02:02:02:01	13:01:01:01	02:01:01:01	06:02:01:01	15:01:01:01
2043	24:02:01:01	40:01:02:01	01:02:01:01	01:03:01:01	02:01:02:01	01:01:01:01	04:01:01:03	01:01:01:01		01:03:01:05
	24:02:01:01	54:01:01:01	03:04:01:02	02:01:01:01	13:01:01:05	03:03:01:03	05:01:01:03	04:05:01:01
2050	02:01:01:01	40:01:02:01	03:04:01:02	01:03:01:01	02:01:02:11	01:02:01:08	03:01:01:03	11:01:01:01	02:02:01:02
	33:03:01:01	44:03:01:01	14:03:01:01	01:03:01:01	04:01:01:01	05:05:01:01	06:04:01:01	13:02:01:02	03:01:01:01
2052	02:01:01:01	14:01:01:01	06:02:01:01	01:03:01:01	02:02:01:01	01:05:01:01	04:02:01:13	04:04:01:02		01:03:01:01
	30:04:01:01	37:01:01:01	08:02:01:02	01:03:01:05	04:02:01:02	03:03:01:01	05:01:01:05	10:01:01:01
2068	02:03:01:01	13:02:01:01	06:02:01:01	01:03:01:01	02:01:02:73	01:01:01:10	02:02:01:01	07:01:01:01		01:03:01:01	01:02:01
	30:01:01:01	38:02:01:01	07:02:01:01	02:02:02:01	13:01:01:02	02:01:01:01	05:01:24:01	15:02:01:02
2081	24:02:01:01	58:01:01:01	01:02:01:01	01:03:01:01	02:01:02:32	03:03:01:03	02:01:01:01	03:01:01:01	02:02:01:01	01:03:01:05
	33:03:01:01	59:01:01:01	03:02:02:03	01:03:01:04	04:01:01:01	05:01:01:03	04:01:01:03	04:05:01:01

PCR amplification of the HLA genes HLA-A, -B, -C, -DPA1, -DPB1, -DQA1, -DQB1, -DRB1, -DRB3, -DRB4, and -DRB5 was performed for each patient, and high-resolution HLA typing was conducted using long-read NGS. DRB3, DRB4, and DRB5 are closely linked to specific DRB1 alleles, so each patient may carry up to two of these genes or none depending on their two DRB1 alleles [50]. Thus, the presence of DRB3, DRB4, and DRB5 alleles is determined by the patient’s DRB1 genotype.

When the diagnostic performance of the HLA-DPB1*05:01:01:01 allele for DILI was calculated based on Table 7, the sensitivity was 100% (95% CI = 54.1–100%) and the specificity was 90.0% (95% CI = 55.5–99.7%). The PPV and NPV were 85.7% (95% CI = 48.3–97.5%) and 100% (95% CI = 66.4–100%), respectively.

Furthermore, 11 highly polymorphic HLA loci (HLA-A, -B, -C, -DPA1, -DPB1, -DQA1, -DQB1, -DRB1, -DRB3, -DRB4, and -DRB5) were also examined to assess whether specific alleles are associated with the HLA-DPB1*05:01:01:01 allele (Table 8). Other loci, including pseudogenes or HLA genes with low polymorphism that contributed little to the evaluation of DILI associations, were excluded from the analysis [22, 25]. Particular attention was given to HLA-DPA1, as it forms a heterodimer with HLA-DPB1. All patients with HLA-DPB1*05:01:01:01 carried HLA-DPA1*02:02:02:01 (Table 8). This co-occurrence is consistent with the previous haplotype studies showing that HLA-DPB1*05:01 is most commonly linked to HLA-DPA1*02:02 across multiple populations [26, 27]. Although patients with HLA-DPA1*02:02:02:01 did not invariably carry HLA-DPB1*05:01:01:01 in this study, it is noteworthy that all 6 DILI patients carried the HLA-DPA1*02:02:02:01 allele, compared with only 4 of 10 non-DILI patients (Table 8).

Discussion

In this study, we identified HLA-DPB1*05:01:01:01 as a candidate genetic risk marker for KM-819-related DILI. Although the analysis was conducted in a small, exploratory cohort, the association observed supports a potential role for HLA-DPB1*05:01:01:01 in genetic predisposition to KM-819-induced toxicity. These findings, together with the biomarker’s apparent predictive performance within this discovery cohort, are hypothesis-generating and warrant further investigation in larger, independent cohorts before any clinical application.

HLA-DPB1 gene locus is highly polymorphic [24]. HLA-DPB1*05:01 allele has several suballeles with variations at the third- and fourth-field levels that may encode identical proteins. Given that alleles identical down to the second field are expected to encode the same protein, differences only in the third and fourth fields may retain equivalent antigen-presenting functions. To identify individuals at risk for DILI, analyzing down to the second field and using HLA-DPB1*05:01 as a biomarker might be appropriate. Therefore, our study employed allele analyses at both the second- and fourth-field levels, the former for biomarker identification and the latter for study resolution.

The Val234Met variant is located outside the peptide-binding groove and is unlikely to directly influence peptide or drug binding. Therefore, this variant may serve as a marker for a linked HLA haplotype or other genetic variants associated with KM-819-induced DILI. If so, other HLA-DPB1 alleles carrying the Val234Met variant, including HLA-DPB1*19:01, HLA-DPB1*22:01, and HLA-DPB1*29:01 [24], could also be associated with an increased risk of DILI. Alternatively, susceptibility to KM-819-induced DILI may be primarily mediated through HLA-dependent antigen presentation. In that case, individuals carrying HLA alleles belonging to the same P group as HLA-DPB1*05:01, which encode identical amino acid sequences within the peptide-binding groove [24], such as HLA-DPB1*135:01, may also be at increased risk of developing DILI. However, none of the HLA-DPB1 alleles sharing either the same P group or the Val234Met variant were identified in our cohort. Therefore, larger independent cohorts are needed to determine whether these alleles are also associated with KM-819-induced DILI. In addition, functional studies, such as T-cell activation assays, will be valuable for elucidating the underlying mechanism.

Our results suggest that the HLA-DPB1*05:01 allele may have diagnostic potential as a predictive biomarker for KM-819-induced DILI. The observed 100% sensitivity and 90.0% specificity are comparable to those of established genetic biomarkers that have prompted the adoption of pre-treatment screening protocols. For instance, genetic screening for HLA-B*57:01 is mandatory prior to prescribing abacavir, a treatment for acquired immune deficiency syndrome, due to its association with hypersensitivity reactions and its predictive performance (100% sensitivity, 96.9% specificity) [28]. Similarly, HLA-B*15:02 screening is recommended before initiating carbamazepine for epilepsy, given its strong link to severe cutaneous adverse reactions and its corresponding sensitivity and specificity of 100 and 97%, respectively [29]. However, we would like to emphasize that the observed association between HLA-DPB1*05:01 and KM-819-induced DILI is based on an exploratory analysis with a limited sample size compared with the above-mentioned studies, and requires validation in larger, independent cohorts before clinical implementation.

Even when certain HLA alleles show high sensitivity and specificity for predicting drug-related adverse events, their application in patient selection can be limited by low PPV [30]. HLA-DPB1*05:01 shows an 85.7% PPV and a 100% NPV for KM-819-induced DILI in patients with MSA, which is higher than the 47.9% PPV for HLA-B*57:01, the HLA typing-based selection marker that is mandatory when using abacavir [28, 31]. This higher apparent PPV likely reflects the enriched case-control design of this study, in which DILI and non-DILI patients were included at a relatively balanced proportion, rather than the true prevalence of DILI in a general KM-819-treated population. Because PPV is highly sensitive to disease prevalence [32], it would be expected to be substantially lower when applied as a pre-treatment screening marker in a population with a lower incidence of DILI. These findings suggest that HLA-DPB1*05:01 could serve as a potential selection marker for KM-819-related DILI, although further validation in larger cohorts is warranted.

We evaluated whether the presence of HLA-DPB1*05:01:01:01 in all DILI patients could occur by chance based on the general allele frequency. For the 10 non-DILI patients, the HLA-DPB1*05:01:01:01 allele frequency was 5.0% (1 out of 20 alleles), which is significantly lower than the 35.1% frequency of the HLA-DPB1*05:01 allele reported in the study of 339 Koreans [33]. This difference is unlikely to have occurred by chance (p = 0.0034, OR = 0.097, 95% CI = 0.0023–0.6209). By contrast, the HLA-DPB1*05:01:01:01 allele frequency in DILI patients was 58.3% (7 out of 12 alleles), appearing higher than the 35.1% reference frequency; however, the statistical significance was much weaker (p = 0.127, OR = 2.584, 95% CI = 0.6975–10.4412), likely due to the small sample size.

This study was not powered to detect differences in genetic locations and is thus exploratory in nature for this purpose. Therefore, a Bonferroni adjustment was not utilized as a Bonferroni adjustment can result in a large Type II error (missing a potentially important correlation due to lack of power) with small sample sizes and is therefore not recommended [34, 35]. While the sample size is limited, given that the population controls match the ethnicity and overall demographics of the DILI patients, a small sample size may be sufficient to detect the high effect size of the HLA on T-cell-mediated immune adverse reactions. Nevertheless, validation in an independent replication cohort will be important for robust confirmation of these findings.

The frequency of HLA-DPB1*05:01 exhibits significant variation across global populations, based on the gold standard population data from the Allele Frequency Net Database [36]. The allele is particularly prevalent in South-East Asia (39%) and Oceania (46%). In contrast, the allele is much less common in Europe (1.9%), North America (2.3%), South and Central America (1.6%), and Sub-Saharan Africa (0.0%). A similar trend is observed across ethnic groups in the US, where the allele is rarely found among individuals of Caucasian (2.04%) or Hispanic (1.19%) ancestry, but is more prevalent in those of Asian descent (10.19%) [37]. This marked disparity underscores the importance of population ancestry when interpreting HLA-DPB1*05:01 as a candidate risk marker, particularly in Asian and Oceanian populations.

HLA-DPB1 alleles have been implicated in drug-induced hepatotoxicity for several agents [19]. Specifically, HLA-DPB1*01:01 and HLA-DPB1*10:01 have been associated with DILI caused by amoxicillin and infliximab, respectively [38, 39]. Of particular relevance, a recent Korean study identified HLA-DPB1*05:01 as a risk factor for hepatotoxicity induced by isoniazid (INH) and rifampin (RIF) [20]. Although the predictive performance for INH- and RIF-induced hepatotoxicity was modest (sensitivity 50.0%, specificity 83.0%, PPV 55.6%, NPV 79.6%), HLA-DPB1*05:01 nonetheless showed a significant association with the hepatotoxicity group compared to the non-hepatotoxicity group. This prior evidence is consistent with a potential role of HLA-DPB1*05:01 as a genetic risk factor, supporting its association with KM-819-induced DILI. In addition, HLA-DPB1*05:01 has been associated with chronic hepatitis B infection and auto-immune conditions such as systemic lupus erythematosus and rheumatoid arthritis [40, 41], and none of the patients with DILI had these conditions prior to enrolling in the study. Therefore, the association between HLA-DPB1*05:01 and DILI appears to be independent of patients’ pre-existing autoimmune or infectious conditions, suggesting a potential KM-819–induced immune-mediated mechanism.

One possible mechanism for KM-819-induced DILI is an immune response triggered by reactive metabolites. However, a recent study indicates that KM-819 is highly stable across rat, dog, and human systems [42]. Although glucuronidation was found to be the predominant metabolic pathway, its overall metabolic contribution remains limited due to the small amount. The high metabolic stability of KM-819 suggests that KM-819 itself may contribute to immune-mediated responses; however, this hypothesis requires further experimental validation, such as peptide-binding and T-cell activation studies.

As HLA-DPB1 forms a heterodimer with HLA-DPA1, specific DPA1-DPB1 haplotypes may influence susceptibility to KM-819-related DILI. HLA-DPA1*02:02, HLA-DPA1*02:01, and HLA-DPA1*01:03 are the major alleles reported to form haplotypes with HLA-DPB1*05:01 across multiple populations [26, 43]. In our cohort, all six DILI patients carried the HLA-DPB1*05:01 allele and at least one copy of HLA-DPA1*02:02. The paired HLA-DPA1 alleles were HLA-DPA1*02:02 in three patients, HLA-DPA1*02:01 in two patients, and HLA-DPA1*01:03 in one patient. Further studies are required to determine which HLA haplotypes may contribute to susceptibility to KM-819-related DILI.

We reviewed studies investigating HLA-DPB1*05:01-related haplotypes in the Korean population [43]. These studies reported nine haplotypes encompassing HLA-DPB1*05:01 with a frequency greater than 1%. Among them, the HLA-DRB1*15:01:01-DQA1*01:02:01-DQB1*06:02:01-DPA1*02:02:02-DPB1*05:01:01 haplotype appeared in three DILI cases (Cases 2013, 2016, and 2071) shown in Table 8. The other haplotypes observed in DILI Cases 2025, 2064, and 2067 were not shared by any other cases. This suggests that the HLA-DRB1*15:01:01-DQA1*01:02:01-DQB1*06:02:01-DPA1*02:02:02-DPB1*05:01:01 haplotype may confer a higher predisposition to KM-819-induced DILI compared with other haplotypes that include HLA-DPB1*05:01.

The Ser386Gly variant of ANKUB1 also showed a notable association with DILI (Table 6). According to the gnomAD database [44], the Ser386Gly variant of ANKUB1 is relatively common across diverse populations: Middle Eastern (46.7%, n = 2 996), European (39.2%, n = 573 458), East Asian (22.0%, n = 20 449), and African (20.2%, n = 36 557). Despite its high prevalence, no studies have examined the functional consequences of this variant. The Ser386Gly substitution may have a significant impact on ANKUB1 protein function, as the loss of a serine residue could potentially alter phosphorylation. In addition, ANKUB1 contains an ankyrin repeat domain [45], which is structurally similar to the ankyrin-like repeat domain previously reported to promote α-synuclein aggregation [46]. These observations suggest that it may be worthwhile to investigate whether the Ser386Gly variant of ANKUB1 influences α-synuclein aggregation.

The Ile113Leu substitution of COMMD2, a negative regulator of NF-κB signaling [47], shows the same p-value (0.0009) as HLA-DPB1*05:01:01:01 (Table 6). This variant is also relatively common, with allele frequencies of 48.0% (n = 3 016) in Middle Eastern, 41.5% (n = 586 885) in European, 22.7% (n = 22 332) in East Asian, and 9.7% (n = 37 451) in African populations [44]. Despite its prevalence, no experimental evidence is currently available to clarify its functional impact. Since overactivation of the NF-κB signaling pathway may contribute to inflammation-induced liver tissue damage [48], further research is needed to determine whether the Ile113Leu variant of COMMD2 affects DILI risk or severity through NF-κB regulation.

The Ile52Val variant of PADI3, the Asn410fs variant of SCRN3, and the Pro799Arg variant of PLA2G4F are relatively common in the East Asian population, with allele frequencies of 15.7% (n = 22 415), 22.6% (n = 22 268), and 11.9% (n = 22 441), respectively [44]. Despite their relatively high frequencies, no functional characterization of these variants has been reported. In addition, no published studies have linked PADI3, SCRN3 or PLA2G4F to DILI. Therefore, the current evidence supports no direct role for these variants in the pathogenesis of DILI. Rather, they may serve as markers of other linked genetic variants associated with DILI susceptibility, and further studies are warranted to clarify their potential biological significance.

This study has several limitations, including a small sample size, incomplete genotyping of all DILI and non-DILI patients, possible selection bias in the choice of non-DILI controls, restriction to a Korean population, and the lack of functional validation. Because the same 16-patient cohort was used both to identify HLA-DPB1*05:01:01:01 and to estimate its sensitivity, specificity, PPV, and NPV, these performance estimates may be optimistic and have not yet been confirmed in an independent validation cohort; the wide confidence intervals around each estimate reflect this uncertainty. Furthermore, given the exploratory nature of the study and its limited sample size, no correction for multiple testing was applied. Despite these limitations, our findings suggest that HLA-DPB1*05:01 may represent a candidate genetic risk marker for KM-819-induced DILI, and this hypothesis-generating association warrants confirmation in larger, independent, and ethnically diverse cohorts, together with functional studies, before clinical application.

Author contributions

Jeong-Hoon Han, Kyuchan Kim, Carrolee Barlow, and Eunhee Kim conceived the study. Chong Sik Lee recruited the patients and collected the samples. Jeong-Hoon Han and Eunhee Kim analyzed and interpreted the data. Jeong-Hoon Han, Chong Sik Lee, Carrolee Barlow, and Eunhee Kim prepared the manuscript. All authors contributed to the editing and review of the manuscript and approved the final version.

Funding

This study was funded by Kainos Medicine, Inc.

Data availability

The raw genome sequencing data generated in this study are not publicly available due to patient privacy and ethical restrictions. However, de-identified data are available from the corresponding author upon reasonable request and subject to IRB approval.

Competing interests

Jeong-Hoon Han, Suyoung Go, Kyuchan Kim, Carrolee Barlow, and Eunhee Kim are employees of Kainos Medicine, Inc., the company developing KM-819. Simona Stankeviciute is an employee of Parexel, a contract research organization. Raúl J. Andrade, Einar Stefán Björnsson, James Hendricks Lewis, M Isabel Lucena, Guruprasad P. Aithal, and Chong Sik Lee declare no competing interests.

Ethics approval and consent to participate

All procedures were performed in accordance with the principles of the Declaration of Helsinki and relevant guidelines and regulations. This study was approved by the IRB of CHA Bundang Medical Center, CHA University (IRB No. 2025-02-061-005). Informed consent was obtained from all participants.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41397-026-00431-3.