Children with KD were recruited from 11 tertiary academic hospitals in Korea, all of which participated in the Korean Kawasaki Disease Genetics Consortium. All patients were diagnosed by pediatricians in accordance with the diagnostic criteria of the American Heart Association.12) A total of 241 patients were selected from the pool of our KD patients, including severe cases with coronary artery aneurysms (n=61) and cases without coronary artery aneurysms (n=180). Among these, 210 had complete KD, with fever lasting 5 days or longer, and at least 4 of the 5 principal clinical features of KD. All patients received single high-dose IVIG (2 g/kg). Two-dimensional echocardiography results were interpreted by pediatric cardiologists and coronary arteries were categorized as normal or abnormal (showing CALs, i.e., dilation or aneurysm). In addition, the coronary artery aneurysms were classified by size: small (internal diameter 3 to <5 mm), medium (5 to <8), or giant (≥8 mm). The study protocol was approved by the Institutional Review Boards (2014-0823) of the involved institutions, and the parents or guardians of all patients provided written informed consent.

Plasma samples from 241 children were collected during the acute (before IVIG treatment; n=56), subacute (within 3 weeks after treatment; n=157), convalescent (between 3 weeks and 6 months after treatment; n=7), or normal (>6 months after treatment; n=21) phase of KD. The concentrations of IgG, IgA, and IgM were measured in a nephelometric immunoassay using the IgG/IgA/IgM Flex Reagent Cartridge kit (Siemens, Munich, Germany) and a Dimension Vista 500 instrument (Siemens). IgE concentrations were measured in a fluorescence enzyme immunoassay (UniCAP Total IgE kit; Phadia AB, Uppsala, Sweden) and the ImmunoCAP 1000 instrument (Phadia AB). Ig levels were classified as low (below the reference range), normal (within the reference range), or high (above the reference range) in accordance with the reference values from the Texas Children's Hospital Clinical Laboratory13) (for IgG, IgA, and IgM) and the ImmunoCAP Total IgE system14) (for IgE). The reference values are presented in Supplementary Tables 1 and 2.

All laboratory tests (except Ig measurements) were performed before initial IVIG treatment; these included a white blood cell count, a neutrophil count, a platelet count, the erythrocyte sedimentation rate, and the concentrations of hemoglobin, C-reactive protein, aspartate aminotransferase, alanine aminotransferase, serum albumin, and total protein.

Genotype data used for GWAS of Ig levels were obtained from our previous study.15) Genotyping methods and quality control steps for GWAS are also described in detail in the previous study.15) Briefly, 296 children with KD were genotyped using the Illumina Human Omni1-Quad Bead Chip (Illumina, San Diego, CA, USA), according to the manufacturer's instructions. Among them, 241 patients with Ig data were included in the analysis. To filter single nucleotide polymorphism (SNP) markers, the following were excluded: 2,553 SNPs with missing call rates >2%, 413 SNP markers with a Hardy-Weinberg Equilibrium p value of <1×10⁻⁶, and 209,342 SNP markers with a minor allele frequency <0.01. After SNP filtration, 718,717 SNPs were included in the genome-wide association analysis.

Statistical analysis was performed using the SPSS Statistics, version 24 (IBM, Armonk, NY, USA). Because Ig data showed a skewed distribution, the data were log transformed. Data are expressed as mean±standard deviation, as the geometric mean (95% confidence interval), as the median (interquartile range), or as percentages as appropriate. For continuous variables, differences between 2 groups were analyzed using Student's t-test or the Mann-Whitney test. Differences among 3 or more groups were analyzed by 1-way analysis of variance (followed by Tukey's test) and the Kruskal-Wallis test. An analysis of covariance test, adjusted for age at the time of sampling, was also used to compare Ig levels among the 4 groups. For categorical variables, the χ² test was used. To determine the effects of Ig levels on CAL, logistic regression analysis adjusted for age, sex, and plasma sampling time point was performed and Pearson correlation coefficient was calculated. A 2-tailed p value <0.05 was considered statistically significant. To analyze the association between Ig levels and SNPs, linear regression analysis adjusted for age at sampling, sex, and plasma sampling time point was performed using PLINK software, version 1.07 (http://pngu.mgh.harvard.edu/~purcell/plink/).16) The levels of each Ig isotype and amount of isotype switching (the ratio of each Ig isotype to IgM) were log transformed before analysis.

To measure the amount of each Ig isotype, we collected plasma samples from 241 patients during the acute phase (n=56), subacute phase (n=157), convalescent phase (n=7), or normal phase (n=21). Significantly higher IgA, IgG, and IgM levels were observed during the subacute phase after IVIG treatment than during the other phases. In particular, and as expected, plasma samples taken during the subacute phase (within 3 weeks after IVIG treatment) contained about 3 times more IgG than plasma samples collected before IVIG treatment (acute phase) (Table 1). By contrast, there were no significant changes in IgE levels during the 4 different clinical phases of KD. However, compared with the reference values, constitutively elevated IgE levels were observed in 73.9% (178/241) of patients during all clinical phases of KD (75.0% in the acute phase, 74.5% in the subacute phase, 71.4% in the convalescent phase, and 66.7% in the normal phase; Figure 1 and Supplementary Table 3).

To examine the effect of each Ig isotype level in KD, we compared inflammatory laboratory parameters and clinical outcomes of KD between the normal and high groups for each Ig isotype (Supplementary Tables 4,5,6,7,8,9). Significantly more CALs were observed in the high IgA group than in the normal IgA group (44.7% vs. 20.8%, respectively; p<0.01; Supplementary Table 4). In particular, we observed 2.2-fold (p=0.038) and 1.7-fold (p <0.001) higher IgA levels in the CAL group than in the non-CAL group during the acute and subacute phases, respectively (Table 2). Less significant increases in IgM levels were observed in the CAL group during the acute phase (1.6-fold, p=0.039) and subacute phase (1.3-fold, p=0.001) than in the non-CAL group (Table 2). Since Ig levels can be affected by compounding factors such as age and sex, we performed logistic regression analysis after adjusting for age, sex, and sampling time point to determine the effects of each Ig isotype on CAL formation in KD patients. High IgA and IgM levels were significantly associated with the incidence of CALs (odds ratio [OR]=2.58, p=0.022 and OR=2.12, p=0.043, respectively) (Table 3). When the same logistic regression analysis was performed separately for the 2 subgroups (acute and subacute), a larger effect size was observed for the acute phase than for the subacute phase (IgA: OR=6.31, p=0.199 vs. OR=2.46, p=0.035, respectively; IgM: OR=25.7, p=0.073 vs. OR=1.84, p=0.113, respectively; Supplementary Table 10). This result suggests that the high IgA and IgM levels in the acute phase are a potential prognostic marker for predicting the risk of CALs. In addition, the concentrations of IgA and IgM increased in line with the size of the coronary artery aneurysms (non-CAL, 74.3 mg/dL; small, 111.3 mg/dL; medium, 159.2 mg/dL; giant, 197.6 mg/dL for IgA, p <0.001; non-CAL, 132.4 mg/dL; small, 170.9 mg/dL; medium, 171.6 mg/dL; giant, 320.9 mg/dL for IgM, p=0.001) (Table 4). Furthermore, correlation analysis revealed that IgA and IgM levels correlated significantly with CAL size (IgA: r=0.435, p<0.001; IgM: r=0.272, p<0.001; Figure 2). However, IgE and IgG levels were not associated with clinical outcomes such as CALs and non-response to IVIG (Supplementary Tables 6,7,8,9). Notably, elevated IgE during all clinical phases of KD was not associated with inflammatory laboratory data or other clinical variables, although the incidence of complete KD in the high IgE group was slightly higher than that in the normal IgE group (90.4% vs. 79.4%, respectively; p=0.040; Supplementary Table 6). This result suggests that elevated IgE levels in patients with KD do not play a crucial role in disease pathogenesis.

To test whether the loci determining Ig levels are associated with known KD susceptibility genes, we performed a GWAS using our previous SNP chip data derived from 241 KD patients. Several potential genes associated with Ig levels were identified (Supplementary Tables 11 and 12). However, with the exception of the ACOXL-BCL2L11 locus, none of the loci associated with Ig levels or specific Ig isotype switching were associated with loci linked to KD susceptibility. One SNP (rs875063; located in the intron of the ACOXL gene) showed a strong association with IgE levels (p=2.73×10⁻⁶) and with IgE isotype switching (IgE/IgM) (p=2.61×10⁻⁷) (Supplementary Table 11). In particular, at the IgE level in patients with KD, the T allele increased by 1.5-fold per allele (GG genotype=38.4 KU/L; GT genotype=57.9 KU/L; TT genotype=119.2 KU/L; p=2.3×10⁻⁵; Table 5). SNP rs875063 is located 65.4 kb upstream from the transcription start site of the BCL2L11 gene, which was identified as a strong KD susceptibility gene, especially in IVIG responders.9) However, the SNP associated with IgE levels (rs875063) was not linked to a previously reported SNP (rs3789065) located in the intron of the BCL2L11 gene, which is associated with KD (r²=0.013; Supplementary Figure 1). This result indicates that the locus in the ACOXL gene determining IgE levels is independent of the KD susceptibility locus in the BCL2L11 gene, although both genes are located very near to each other. In addition, well-known susceptibility genes for KD (rs6993775 in BLK, rs1801274 in FCGR2A, rs3789065 in BCL2L11, rs9378199 in HLA-B) reported by us9)15) did not show an association with IgE levels or IgE isotype switching (Supplementary Table 13). These results suggest that elevated IgE levels in patients with KD do not affect the pathogenesis of KD (at least not directly).