Oxidative stress-related genes in Kawasaki disease: a multi-omics Mendelian randomization study

Abstract

Background

Kawasaki disease (KD) is a vasculitis of unknown etiology. Oxidative stress is hypothesized to play a key role in KD pathogenesis. However, the specific genes and mechanisms underlying this association remain unclear.

Methods

In this study, we employed a two-step strategy combining SMR screening and differential expression gene (DEG) validation to identify oxidative stress-related genes associated with KD. First, we used the Mendelian randomization approach (SMR) to assess causal associations between genes and KD, integrating data from genome-wide association studies (GWAS), blood methylation quantitative trait loci (mQTLs), expression QTLs (eQTLs), and proteomic QTLs (pQTLs) obtained from public databases. Subsequently, we validated the candidate genes through DEG analysis in two independent KD patient cohorts (GSE68004 and GSE100154).

Results

Integrated analysis identified SLC9A1 and RPS6KA1 as candidate risk loci. Genetically predicted upregulation of SLC9A1 expression (OR = 7.25, P = 0.02) and RPS6KA1 protein abundance (OR = 2.82, P = 0.01) was causally associated with increased KD risk. These findings were validated in clinical cohorts, where both genes were consistently upregulated in KD patients across GSE68004 and GSE100154 (all P < 0.05), aligning with SMR predictions. Additionally, APRT emerged as a multi-omics candidate, demonstrating consistent causal evidence across mQTL, eQTL, and pQTL layers, supported by downregulation in the GSE68004 cohort.

Conclusions

This study highlights prioritized SLC9A1 and RPS6KA1 as potential causal drivers of KD, and highlighted APRT as a potential multi-layer regulatory target. These findings provide genetic evidence linking oxidative stress pathways to KD pathogenesis, offering novel targets for therapeutic intervention.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13052-026-02230-9.

Background

Kawasaki disease (KD) is a prevalent vascular inflammatory disease in children, particularly common among those under the age of five [1]. It is characterized by fever, rash, changes in the extremities, and enlargement of cervical lymph nodes [2]. Coronary artery aneurysms represent the most severe complication of KD [3], which can lead to cardiovascular death, myocardial infarction, stroke [4]. Current treatment approaches primarily focus on immunosuppression and anti-inflammatory therapies, but these methods have limitations [5]. Therefore, delving into the molecular mechanisms of KD, especially the genetic factors associated with it, is crucial for enhancing therapeutic efficacy and improving prognosis.

Oxidative stress is defined as an imbalance between reactive oxygen species (ROS) production and antioxidant defenses, leading to cellular injury [6]. An accumulating body of evidence implicates oxidative stress in KD across acute and convalescent phases and in close association with systemic inflammation [7]. Here we emphasize the knowledge gap: observational associations between oxidative stress and KD may be confounded by disease activity, treatment, and reverse causation, leaving the direction and specificity of causal effects unresolved [8]. Mechanistically, ROS can amplify vascular inflammation, impair endothelial function, and contribute to vascular remodeling, potentially fostering coronary artery complications; yet whether these processes represent causes or consequences of KD remains uncertain due to confounding and bidirectional links in observational settings [9–11]. Addressing this uncertainty requires approaches that strengthen causal inference and map regulatory links from DNA to gene expression and protein levels.

To this end, Mendelian randomization (MR) leverages germline genetic variants as instrumental variables to estimate the causal effect of an exposure on an outcome, thereby reducing confounding and reverse causation [12]. This research utilized a multi-omics MR analysis approach, integrating expression Quantitative Trait Loci (eQTLs), methylation QTL (mQTLs), and protein QTL (pQTLs) data, to comprehensively evaluate the causal associations between the expression, methylation, and protein levels of oxidative stress-related genes and KD. We will employ the summary data-based MR (SMR) and HEIDI tests to assess the causal effects of the associations. Additionally, this study conducted colocalization analysis to identify shared genetic determinants between KD and oxidative stress-related genes.

This study systematically assessed the causal association between oxidative stress-related genes and KD through a multi-omics MR approach, elucidating their specific mechanisms of action in the pathogenesis and progression of the disease. The findings are not only expected to reveal the molecular etiology of KD but also provide new targets and theoretical support for the early prevention and precision treatment of KD.

Methods

Study design

This study followed the STROBE-MR reporting guidelines [13] and used a MR approach to investigate potential causal associations between oxidative stress-related genes and KD. The methodology involves a thorough analysis of genetic data, application of SMR and HEIDI tests for assessing associations between oxidative stress-related genes and KD, and colocalization analysis to identify shared genetic determinants. The study design, workflow for genetic variant selection, and analytical approach are detailed in Fig. 1. The study did not require ethical approval because it was based on GWAS data that already had ethical approval. For a more detailed description of the methodology, please refer to the supplementary material.

Fig. 1.

Study design flow chart

Data sources

For our genetic investigation, we sourced GWAS data for KD from the GeneCards database, specifically focusing on the 1,887 protein-coding genes with relevance scores in the top 25%, indicating association to oxidative stress. The primary discovery dataset was obtained from the GWAS Catalog [14] with GWAS ID GCST90013537, which included 400 cases and 6,101 controls. KD cases were defined according to the American Heart Association diagnostic criteria and were recruited from pediatric centers in the USA, UK, the Netherlands, and Finland between 2001 and 2018. Coronary artery involvement was assessed by echocardiography at each participating center, with coronary artery aneurysms (CAA) classified using coronary artery Z scores or absolute luminal diameters, depending on regional clinical practice. Patients with dilated coronary arteries not meeting CAA criteria were excluded from CAA-specific analyses but included in the KD susceptibility GWAS. Control subjects consisted of healthy European infants recruited during routine vaccination programs.

Additionally, blood eQTL summary data were obtained from eQTLGen [15], which included blood gene expression and genetic data from 31,684 individuals. Blood mQTL summary data were derived from a meta-analysis of two European cohorts: the Brisbane Systems Genetics Study with 614 participants and the Lothian Birth Cohorts with 1,366 individuals [16]. Blood pQTL summary data were obtained from Pietzner, et al. [17], which included 10,708 Europeans.

SMR analysis

Using the SMR software tool version 1.3.1, this study aimed to investigate the causal associations between methylation levels, expression patterns, and protein content of oxidative stress-related genes and KD through SMR and HEIDI tests. We emphasized the enhancement of statistical power by performing SMR analysis based on top-associated cis-QTLs in two independent and large cohorts, compared to traditional MR methods. We used a screening mechanism centered on gene probes, setting a window of ± 1,000 kb and a P-value threshold of 5.0 × 10^{− 8} to select significantly associated cis-QTLs [18]. Additionally, we excluded SNPs with allele frequency differences greater than 0.2 across different datasets (including LD reference samples, QTL summary data, and outcome summary data) to ensure study accuracy. For mQTLs, eQTLs, and pQTLs, we set a threshold of 0.05, allowing a certain proportion of SNPs with allele frequency differences greater than 0.2.

Furthermore, this study explored the potential causal associations between QTLs and KD, investigating the causal association between mQTLs (as exposure) and eQTLs (as outcome) and between eQTLs (as exposure) and pQTLs (as outcome) [19]. Based on SMR analysis, we employed a multi-SNP-SMR analysis method (–SMR-multi) to comprehensively assess significant results. We used HEIDI tests to screen out results with P-values greater than 0.01 and no pleiotropy [18].

To further identify shared causal variants between oxidative stress gene-related cis-QTLs and KD, we performed colocalization analysis using the R package “coloc”. When GWAS signals and QTLs colocalized, we inferred that the site on the GWAS signal might influence the phenotype by affecting the biological process of the gene. In the colocalization analysis, we reported five different posterior probabilities corresponding to five exclusive hypotheses, ranging from no association to sharing a single causal variable: H0: No features are genetically associated with the SNP in this region; H1: Only feature 1 is genetically associated with the SNP; H2: Only feature 2 is genetically associated with the SNP; H3: Both features are associated with the SNP but use different causal variables; H4: Both features are associated with the SNP and share a single causal variable. Based on references [20–23], we set colocalization region windows of ± 500 kb, ± 1,000 kb, and ± 1,000 kb for mQTL-GWAS, eQTL-GWAS, and pQTL-GWAS colocalization analysis, respectively, to allow weaker QTL signals to colocalize with GWAS signals. We considered QTL signals and GWAS signals successfully colocalized if they met the following conditions: (1) PP.H4 > 0.5 when P12 = 5 × 10^{− 5}; (2) PP.H3 < 0.5 when P12 = 1 × 10^{− 5}.

Expression validation of SMR candidate genes

To validate and prioritize genes identified by SMR, we performed independent expression analyses using microarray datasets from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Specifically, we downloaded: (i) GSE68004 (76 KD cases, 37 controls) [24] and (ii) GSE100154 (23 KD cases, 21 controls) [25]. In GSE68004, KD cases were defined according to established clinical criteria, with CAA assessed by echocardiography and defined as a Z score ≥ 2.5 for either the right coronary artery or left anterior descending artery. Gene expression profiles were primarily derived from samples collected prior to intravenous immunoglobulin treatment. In GSE100154, KD patients (< 18 years of age) were diagnosed based on the American Heart Association (AHA) criteria, and healthy controls were recruited without a history of KD or inflammatory disease. Differential expression analysis was conducted with the R package limma (v3.62.2) following standard workflows. P values were adjusted using the Benjamini–Hochberg method, and statistical significance was defined as adjusted P (p.adj) < 0.05. This step was designed to validate and screen SMR-discovered candidate genes in whole blood and to assess the consistency of directionality between SMR-inferred effects and differential expression in KD cohorts.

Software and reproducibility

All statistical analyses were performed using R (v4.3.0). The R packages “ggplot2” and “ggrepel” were used for Manhattan plot generation, while “forestplot” was used for forest plot generation. The code for SMRSite Plot and SMREffectPlot was adapted from Zhu, et al. [26].

Result

Multi-omics SMR screening identifies causal candidates for Kawasaki disease

To comprehensively screen for oxidative stress-related genes causally linked to KD, we performed SMR analysis integrating mQTL, eQTL, and pQTL data. The screening initially identified broad signals: 228 methylation sites (124 genes) at the mQTL level, 27 genes at the eQTL level, and 6 proteins at the pQTL level were potentially associated with KD risk (Figs. 2 and 3; Tables S1-S3). From this pool, SMR analysis estimated that genetically predicted higher expression of SLC9A1 was associated with increased KD susceptibility (OR = 7.25, 95% CI: 1.36–38.79, P = 0.02). Similarly, genetically predicted higher protein abundance of RPS6KA1 was associated with elevated KD risk (OR = 2.82, 95% CI: 1.23–6.51, P = 0.01). Both associations were supported by Bayesian colocalization (PPH4 > 0.5), suggesting they are driven by shared causal variants rather than linkage disequilibrium.

Fig. 2.

SMR analysis results for mQTLs in the GCST90013537 cohort (Highlighting Significant Colocalized Genes)- Forest plot

Fig. 3.

Forest plot. A SMR analysis results for eQTLs in the GCST90013537 cohort; B SMR analysis results for pQTLs in the GCST90013537 cohort

Robust validation of SLC9A1 and RPS6KA1 in independent clinical cohorts

To filter the SMR candidates for clinical relevance, we performed differential expression analysis in two independent KD transcriptomic datasets (GSE68004: 76 cases vs. 37 controls; GSE100154: 23 cases vs. 21 controls) (Table 1). Among the 32 genes potential related in eQTL or pQTL SMR analyses, SLC9A1 and RPS6KA1 emerged as the most robust targets, being the only candidates consistently validated across both datasets with directions matching the genetic prediction (Fig. 4). Specifically, SLC9A1 was potentially upregulated in KD patients in both GSE68004 (logFC = 0.96, p.adj < 0.001) and GSE100154 (logFC = 0.61, p.adj < 0.001), consistent with the SMR result that higher expression increases risk. Similarly, RPS6KA1 was consistently upregulated in both cohorts (GSE68004: logFC = 0.65, p.adj < 0.001; GSE100154: logFC = 0.32, p.adj = 0.018), aligning with the pQTL-SMR finding.

Table 1.

Differential expression analysis in GSE68004 and GSE100154 datasets

Gene	logFC	AveExpr	t	P.Value	adj.P.Val	B	Regulation	Dataset
CCNA2	0.1132	6.8258	2.0766	0.0401	0.1055	-5.1686	NS	GSE68004
NTHL1	-0.2058	6.9728	-4.1486	0.0001	0.0003	0.7509	Down	GSE68004
BACH2	-0.6389	8.2252	-5.8370	0.0000	0.0000	7.6606	Down	GSE68004
SLC9A1	0.9602	8.8580	14.4916	0.0000	0.0000	52.2437	Up	GSE68004
MYB	0.2716	7.6641	3.4374	0.0008	0.0035	-1.6504	Up	GSE68004
BOLA1	-0.1529	6.5680	-5.6169	0.0000	0.0000	6.6795	Down	GSE68004
SIRT6	0.1106	6.5138	4.0649	0.0001	0.0004	0.4503	Up	GSE68004
BTRC	-0.0966	6.6027	-3.0500	0.0028	0.0107	-2.8048	Down	GSE68004
MUTYH	-0.2868	7.6042	-4.5103	0.0000	0.0001	2.0998	Down	GSE68004
PCCA	-0.1648	6.9149	-4.7912	0.0000	0.0000	3.2011	Down	GSE68004
OLA1	-0.0905	6.2051	-3.2527	0.0015	0.0060	-2.2151	Down	GSE68004
LOX	-0.0369	6.2029	-1.4898	0.1390	0.2766	-6.1901	NS	GSE68004
BRCA1	0.1774	7.0480	3.9626	0.0001	0.0006	0.0896	Up	GSE68004
TMED4	-0.3042	7.9291	-6.7652	0.0000	0.0000	12.0036	Down	GSE68004
NFU1	-0.1234	7.7333	-2.0016	0.0477	0.1215	-5.3169	NS	GSE68004
ITGB1	0.1730	11.2129	2.5332	0.0127	0.0399	-4.1596	Up	GSE68004
APRT	-0.8193	9.1368	-10.4146	0.0000	0.0000	30.8803	Down	GSE68004
S100A8	0.3314	13.6414	6.8712	0.0000	0.0000	12.5178	Up	GSE68004
NEFM	-0.0200	6.3348	-0.8976	0.3713	0.5522	-6.8906	NS	GSE68004
MMACHC	-0.0447	6.4169	-1.9344	0.0555	0.1374	-5.4452	NS	GSE68004
PKD1	-0.0686	6.7875	-1.6058	0.1111	0.2345	-6.0134	NS	GSE68004
AKR1A1	-0.2120	9.6710	-2.7523	0.0069	0.0235	-3.6120	Down	GSE68004
MAPK11	0.0486	6.5218	1.8700	0.0641	0.1542	-5.5645	NS	GSE68004
ATG4D	0.0694	7.0198	2.2487	0.0265	0.0747	-4.8097	NS	GSE68004
MAP4K4	0.5644	8.5890	8.9530	0.0000	0.0000	23.1261	Up	GSE68004
LDLR	0.4818	8.2942	5.9061	0.0000	0.0000	7.9729	Up	GSE68004
ZEB1	-0.0822	6.3214	-3.1878	0.0019	0.0072	-2.4074	Down	GSE68004
PARK7	0.0983	10.0713	1.6782	0.0961	0.2102	-5.8969	NS	GSE68004
PRSS8	0.0716	6.5313	2.2875	0.0240	0.0689	-4.7252	NS	GSE68004
PSME2	-0.1585	10.6064	-1.6298	0.1059	0.2263	-5.9754	NS	GSE68004
RBP4	0.0405	6.3695	1.6524	0.1012	0.2188	-5.9390	NS	GSE68004
RPS6KA1	0.6472	9.5578	10.0644	0.0000	0.0000	29.0143	Up	GSE68004
CCNA2	1.5343	4.4127	3.8439	0.0004	0.0004	0.0684	Up	GSE100154
NTHL1	-0.8177	5.2729	-1.3527	0.1830	0.1830	-5.1330	NS	GSE100154
BACH2	-0.5261	7.8343	-2.3856	0.0214	0.0214	-3.4324	Down	GSE100154
SLC9A1	0.6140	8.5743	3.9556	0.0003	0.0003	0.3733	Up	GSE100154
MYB	-0.1145	7.0029	-0.1544	0.8780	0.8780	-5.9811	NS	GSE100154
BOLA1	-0.8155	3.5623	-1.3521	0.1832	0.1832	-5.1337	NS	GSE100154
SIRT6	0.6161	2.2610	1.2371	0.2226	0.2226	-5.2711	NS	GSE100154
BTRC	-0.3859	3.1950	-0.5432	0.5897	0.5897	-5.8514	NS	GSE100154
MUTYH	-0.9757	6.6675	-1.2976	0.2012	0.2012	-5.2002	NS	GSE100154
PCCA	-0.2585	5.6236	-1.5430	0.1300	0.1300	-4.8812	NS	GSE100154
LOX	0.1756	-0.5766	0.2636	0.7933	0.7933	-5.9593	NS	GSE100154
BRCA1	0.7104	5.0247	1.9170	0.0617	0.0617	-4.3022	NS	GSE100154
TMED4	-0.0325	8.2100	-0.2212	0.8260	0.8260	-5.9691	NS	GSE100154
NFU1	-0.9525	7.0953	-1.3928	0.1706	0.1706	-5.0823	NS	GSE100154
ITGB1	-0.6696	10.8157	-0.7610	0.4507	0.4507	-5.7165	NS	GSE100154
APRT	-1.4181	8.6063	-1.8698	0.0681	0.0681	-4.3811	NS	GSE100154
S100A8	0.1484	13.9953	1.2686	0.2112	0.2112	-5.2345	NS	GSE100154
NEFM	-0.0242	0.2893	-0.0324	0.9743	0.9743	-5.9921	NS	GSE100154
MMACHC	-0.4173	3.3107	-0.7716	0.4444	0.4444	-5.7088	NS	GSE100154
PKD1	-0.7989	5.6120	-1.3105	0.1968	0.1968	-5.1847	NS	GSE100154
AKR1A1	-0.2122	9.6162	-1.1886	0.2410	0.2410	-5.3257	NS	GSE100154
MAPK11	0.6184	1.9361	0.7787	0.4403	0.4403	-5.7036	NS	GSE100154
ATG4D	-0.5360	4.7045	-0.8993	0.3734	0.3734	-5.6081	NS	GSE100154
MAP4K4	0.2867	8.4736	1.9490	0.0577	0.0577	-4.2476	NS	GSE100154
LDLR	-0.0731	7.4758	-0.0960	0.9240	0.9240	-5.9882	NS	GSE100154
PARK7	0.1231	9.3045	0.6850	0.4969	0.4969	-5.7686	NS	GSE100154
PRSS8	-0.8820	1.7200	-1.1695	0.2485	0.2485	-5.3465	NS	GSE100154
PSME2	-0.6943	9.3545	-1.1056	0.2749	0.2749	-5.4142	NS	GSE100154
RBP4	0.3516	-0.4934	0.5110	0.6119	0.6119	-5.8676	NS	GSE100154
RPS6KA1	0.3208	9.0489	2.4656	0.0176	0.0176	-3.2694	Up	GSE100154

NS, not significant

Fig. 4.

Differential expression analysis of KD patients and controls in the GSE68004 cohort. A Highlighted genes with consistent directions between SMR predictions and DEG results. B Boxplots of representative SMR candidate genes in GSE68004 (including SLC9A1, RPS6KA1, MYB, SIRT6, BRCA1, ITGB1, APRT, AKR1A1, and ZEB1). C Volcano plot of differential expression analysis in the GSE100154 cohort. D Boxplots showing consistent upregulation of key validated genes (SLC9A1 and RPS6KA1) in KD patients across both datasets

Epigenetic and multi-layer regulation of candidate genes

Beyond the primary targets, we explored potential regulatory mechanisms by integrating mQTL, eQTL, and pQTL data, identifying candidate genes with multi-omics support (Tables S4-S5). APRT emerged as a gene of interest, showing a suggestive methylation–expression–protein axis: specific methylation sites were associated with reduced APRT mRNA levels, which in turn correlated with lower adenine phosphoribosyltransferase protein abundance. SMR analyses further indicated that genetically predicted lower APRT expression and protein levels were nominally associated with higher KD risk. However, external validation yielded mixed results: APRT was potentially downregulated in the GSE68004 cohort (consistent with the SMR direction), but this association did not reach statistical significance in the smaller GSE100154 cohort (adjusted p.adj = 0.068). Given this inconsistency across datasets, APRT should be considered a preliminary candidate requiring further replication, rather than a robustly validated target.

Additionally, we found epigenetic regulation for SLC9A1, where methylation at cg15030789 was causally linked to its expression, providing a mechanistic basis for the upregulation of SLC9A1 observed in patients.

Discussion

In this study, we employed an integrative multi-omics MR framework to systematically investigate the causal relevance of oxidative stress-related genes and KD. By combining genetic instruments from methylation, transcriptomic, and proteomic layers with rigorous validation in independent clinical cohorts, we prioritized SLC9A1 and RPS6KA1 as high-confidence candidate drivers of KD susceptibility. Furthermore, our multi-layer integration highlighted APRT as a potential key node regulated by epigenetic mechanisms, demonstrating a coherent causal chain across methylation, gene expression, and protein levels. These findings provide genetic evidence implicating specific oxidative stress pathways in KD pathogenesis and offer prioritized targets for future mechanistic validation and therapeutic exploration.

One of the most intriguing findings of this study was the identification of SLC9A1 (Sodium/hydrogen exchanger 1) as a key gene in KD. SLC9A1 plays a crucial role in cellular pH regulation and ion homeostasis by mediating the exchange of sodium and hydrogen ions across the cell membrane. This gene’s function is essential for maintaining intracellular pH balance, which is critical for proper cell function. SLC9A1 has been implicated in several physiological processes, including cell proliferation, apoptosis, and inflammatory responses [27, 28]. In the context of KD, SLC9A1 could potentially influence the pathological processes through its role in regulating endothelial cell function. KD is characterized by acute vasculitis, and endothelial cell damage is a key feature of the disease. The upregulation of SLC9A1 in KD patients, observed in both the GSE68004 and GSE100154 datasets, suggests that this gene might contribute to vascular inflammation and endothelial cell dysfunction [29]. The dysregulation of SLC9A1 under oxidative stress may further exacerbate endothelial injury and inflammation, both of which are central to KD pathology. While research directly linking SLC9A1 with vascular inflammation is limited, studies have suggested that similar ion transporters are involved in inflammatory vascular diseases such as atherosclerosis and arteritis [30]. Further studies are needed to validate these hypotheses and better understand the functional role of SLC9A1 in KD.

Another finding was the role of RPS6KA1 (p90 ribosomal S6 kinase 1) in KD. RPS6KA1 is a key kinase in the MAPK signaling pathway, which is critical for regulating cellular responses to stress, including inflammation, cell survival, and cell growth. The MAPK pathway, particularly the p38 MAPK and ERK sub-pathways, has been widely studied in the context of immune responses and inflammation [31]. Given the prominent inflammatory nature of KD, the upregulation of RPS6KA1 in KD patients could be linked to the disease’s inflammatory storm. RPS6KA1 regulates the translation of various pro-inflammatory cytokines and immune mediators, which suggests that it could play a direct role in the inflammatory process in KD. Inflammation in KD is not only a response to infection but also involves immune dysregulation, which may be driven by the activation of signaling pathways like MAPK. The overactivation of RPS6KA1 in KD could contribute to endothelial cell damage, vascular smooth muscle remodeling, and immune cell activation, all of which are observed in KD patients. These findings suggest that targeting RPS6KA1 may provide therapeutic benefits in mitigating the excessive inflammation seen in KD [32].

In comparison to SLC9A1 and RPS6KA1, APRT (adenine phosphoribosyltransferase) also represents an important finding supported by multi-omics evidence. APRT participates in the purine salvage pathway and catalyzes the conversion of adenine to adenosine monophosphate, a critical step in maintaining cellular energy balance and nucleotide homeostasis [33, 34]. Given that purine metabolism dysregulation can influence oxidative stress and inflammatory responses, APRT may hypothetically contribute to KD pathogenesis via metabolic–inflammatory crosstalk. In our integrative analysis, methylation at the APRT CpG site cg07947266 was associated with reduced APRT expression and lower protein abundance, and SMR suggested that lower genetically predicted APRT levels were nominally linked to higher KD risk. However, external transcriptomic validation was inconsistent: while APRT was potentially downregulated in the GSE68004 cohort, the same direction of effect in the smaller GSE100154 cohort did not reach statistical significance (p adj = 0.068). Given this limited reproducibility across independent datasets, the role of APRT in KD remains hypothesis-generating rather than confirmatory. Future studies with larger sample sizes and functional assays are needed to determine whether this signal reflects a true biological association or arises from dataset-specific noise.

We observed that some of the candidate genes identified through SMR analysis showed inconsistent or non-significant results in the differential expression validation. This is not surprising, as SMR and DEG analyses address different biological questions. SMR aims to infer the causal association between lifelong genetic predisposition in gene expression and disease susceptibility, uncovering potential etiological clues. On the other hand, DEG analysis captures the complex pathological state during the acute phase of the disease, which can be influenced by factors such as strong inflammatory responses, compensatory mechanisms of the body, and even clinical interventions. Therefore, genes that exhibit consistent directionality across both SMR and DEG analyses, such as SLC9A1 and RPS6KA1, are more likely to play a central role in the development of KD and are considered more reliable candidate targets.

Our study has several strengths. First, the use of a multi-omics MR approach enables a more comprehensive understanding of the causal relationships between oxidative stress-related genes and KD. By integrating multiple sources of data, including mQTL, eQTL, and pQTL, we were able to identify key genes and proteins involved in the disease. Additionally, the two-stage design, with independent validation using GSE68004 and GSE100154, strengthens the reliability of our findings. However, there are also several limitations. A primary limitation of this study is the ancestry mismatch between the utilized datasets and the global epidemiology of KD. While KD has the highest incidence in East Asian populations, the GWAS and QTL data used here were derived from European-ancestry cohorts. We restricted our analysis to European datasets to ensure consistency in linkage disequilibrium patterns between the GWAS and QTL data, which is a prerequisite for the SMR method and the HEIDI test. Consequently, our findings may not be fully generalizable to East Asian populations. Furthermore, the transcriptomic validation was limited by the lack of suitable case-control datasets from East Asian cohorts. Future studies should prioritize the development of large-scale East Asian KD GWAS and ancestry-matched multi-omics resources to validate these candidate genes and uncover population-specific mechanisms. Moreover, blood-based QTLs do not fully represent the expression profiles of coronary artery tissue, which is the primary site of damage in KD. DEG analysis may also be subject to confounding factors, such as treatment effects or disease stage, which could affect the differential expression of genes. Additionally, the large SMR effect sizes (such as OR > 7 for SLC9A1) should be interpreted with caution. SMR estimates reflect the change in disease odds per unit increase in genetically predicted molecular levels (such as methylation or expression), not a direct clinical risk ratio. Such exaggerated effects may arise when the top cis-SNP used as the instrument is strongly colocalized with the GWAS signal, especially given that SMR relies on a single instrumental variant, potentially amplifying point estimates due to limited instrument strength or scaling artifacts. While SLC9A1 exemplifies this pattern, similar considerations apply to other prioritized genes with large effect sizes. These findings highlight loci of genetic concordance rather than quantifying absolute biological risk.

Conclusions

In conclusion, our study identifies SLC9A1 and RPS6KA1 as potential key genes involved in the pathogenesis of KD, and highlights the importance of oxidative stress-related pathways in the disease. These findings provide valuable insights into the molecular mechanisms of KD and suggest potential targets for therapeutic intervention. Future research should aim to validate these findings in diverse populations and explore the functional roles of these genes in KD.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

KD

Kawasaki Disease

MR

Mendelian Randomization

SMR

Summary data–based Mendelian Randomization

GWAS

Genome–Wide Association Study

eQTL

expression Quantitative Trait Locus

mQTL

methylation Quantitative Trait Locus

pQTL

protein Quantitative Trait Locus

DEG

Differential Expression Gene

ROS

Reactive Oxygen Species

SNP

Single Nucleotide Polymorphism

LD

Linkage Disequilibrium

GEO

Gene Expression Omnibus

OR

Odds Ratio

CI

Confidence Interval

PP.H4

Posterior Probability for Hypothesis 4 (colocalization)

Author contributions

JWS and QMZcarried out the studies, participated in collecting data, and drafted the manuscript. ND and DQ performed the statistical analysis and participated in its design. MJW participated in acquisition, analysis, or interpretation of data and draft the manuscript. All authors read and approved the final manuscript.

Funding

This study was supported by the National Natural Science Foundation of China (82470314); Key Research Project of Anhui Provincial Health Commission (AHWJ2022a04); Bengbu Science and Technology Innovation Guidance Project (20220105), Key Project of Bengbu Municipal Health Commission (BBWK2024A102).

Data availability

All data generated or analysed during this study are included in this published article.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The author(s) report no conflicts of interest in this work.