Human Genetics

, Volume 132, Issue 6, pp 669–680

Combined analysis of genome-wide-linked susceptibility loci to Kawasaki disease in Han Chinese

Authors

  • Yuanlong Yan
    • Department of Medical Genetics, State Key Laboratory of Biotherapy, West China HospitalSichuan University
  • Yongyi Ma
    • Department of Medical Genetics, State Key Laboratory of Biotherapy, West China HospitalSichuan University
  • Yunqiang Liu
    • Department of Medical Genetics, State Key Laboratory of Biotherapy, West China HospitalSichuan University
  • Hongde Hu
    • Department of Cardiovascular Medicine, West China HospitalSichuan University
  • Ying Shen
    • Department of Medical Genetics, State Key Laboratory of Biotherapy, West China HospitalSichuan University
  • Sizhong Zhang
    • Department of Medical Genetics, State Key Laboratory of Biotherapy, West China HospitalSichuan University
  • Yongxing Ma
    • Department of Medical Genetics, State Key Laboratory of Biotherapy, West China HospitalSichuan University
  • Dachang Tao
    • Department of Medical Genetics, State Key Laboratory of Biotherapy, West China HospitalSichuan University
  • Qing Wu
    • Department of PediatricsSichuan Academy of Medical Sciences and Sichuan Provincial People’s Hospital
    • Department of PediatricsSichuan Academy of Medical Sciences and Sichuan Provincial People’s Hospital
    • Department of Medical Genetics, State Key Laboratory of Biotherapy, West China HospitalSichuan University
Original Investigation

DOI: 10.1007/s00439-013-1279-2

Cite this article as:
Yan, Y., Ma, Y., Liu, Y. et al. Hum Genet (2013) 132: 669. doi:10.1007/s00439-013-1279-2

Abstract

Kawasaki disease (KD) is a dominant cause of acquired heart disease in children due to frequent complicating coronary artery lesions (CALs). Genome-wide association study and linkage analysis have recently identified 6 susceptibility loci at genome-wide significance of P < 5.0 × 10−8 in subjects of Japanese, Taiwanese and European. In present study, we analysed the variants of 6 single nucleotide polymorphisms (SNPs) in the genetic loci to investigate their potential effect on KD susceptibility and outcomes in Han Chinese population. As a result, the risk alleles of rs1801274 and rs2254546 were observed significant effect on KD with higher frequencies in 358 patients than those in 815 controls. The significant role of rs1801274, rs2857151 and rs2254546 in KD was found in the multi-variable logistic regression analysis of the SNPs. Two 2-locus and one 3-locus combinations of the SNPs showed significant effect on KD with stronger association with KD relative to comparable single SNP or 2-locus combinations. Significant susceptibility to CALs was found in KD patients with high-risk genotypes at both rs1801274 and rs2857151. The meta-analyses first revealed significant risk for CALs in KD patients carrying risk allele of rs11340705, and the association of rs28493229 with KD was not observed in the Han Chinese. In conclusion, the findings demonstrated that 5 of the 6 genetic loci influence the risk for KD and 3 of them may be involved in secondary CALs formation in Han Chinese. The additive effects of 3 multi-locus combinations on KD/CALs imply that some loci may participate together in certain unknown gene networks related to KD/CALs. Further function studies of the genetic loci are helpful for better understanding the pathophysiology of KD.

Introduction

Kawasaki disease (KD) is an acute, self-limited vasculitis that primarily affects children younger than 5 years of age. The clinical manifestations of KD include fever, bilateral non-exudative conjunctivitis, erythema of the lips and oral mucosa, changes in the extremities, rash, and cervical lymphadenopathy (Kawasaki 1967). Of the patients with KD, 50 % develop heart damages, and 20–25 % of untreated patients eventually develop coronary artery lesions (CALs) (Kato et al. 1975, 1996). Owing to the frequent presence of the severe cardiovascular complications, the long-term epidemiological surveys of KD have been carried out in many countries and regions since KD was first reported in 1967 (Kawasaki 1967). The data show that KD is more common in ethnic Northeast Asian populations including Japanese, Korean and Chinese (Nakamura et al. 2012; Park et al. 2011; Ma et al. 2010; Huang et al. 2009; Du et al. 2007). Meanwhile, the familial and regional aggregations of KD patients suggest the important role of genetic factors in KD susceptibility and outcomes (Uehara et al. 2003; Cook et al. 1989; Fujita et al. 1989).

Until now, genome-wide association study (GWAS) and linkage analysis of KD have reported 13 genetic loci showing association with KD or secondary CALs in subjects of Japanese, European, Korean and Taiwanese. The loci include ITPKC (inositol 1,4,5-trisphosphate 3 kinase-C), CASP3 (caspase 3), NAALADL2 (N-acetylated α-acidic dipeptidase-like 2), ZFHX3 (zinc finger homeobox 3), DAB1 (disabled homolog 1, Drosophila), PELI1 (pellino homolog 1, Drosophila), COPB2 (coatmer protein complex beta-2 subunit), ERAP1 (endoplasmic reticulum aminopeptidase 1), IGHV (immunoglobulin heavy chain variable region), FCGR2A (Fc fragment of IgG, low affinity IIa, receptor), BLK (B lymphoid kinase), CD40 (CD40) and HLA (major histocompatibility complex) (Onouchi et al. 2008, 2010, 2012; Burgner et al. 2009; Kim et al. 2011; Tsai et al. 2011; Khor et al. 2011; Lee et al. 2012). The outcomes bring the hope of understanding the pathophysiology of KD and searching out the genetic markers for predicting the susceptibility of KD, the treatment effect of intravenous immunoglobulin (IVIG) and secondary CALs formation.

Among the 13 KD susceptibility loci, six including ITPKC, CASP3, FCGR2A, BLK, CD40 and HLA deserve special attention due to that they show much stronger association with KD at genome-wide significance of P < 5.0 × 10−8 relative to other loci. For this study, we analysed 6 susceptibility single-nucleotide polymorphisms (SNPs) to KD in the 6 genetic loci to investigate the correlation between the loci and KD in Han Chinese. Of the SNPs, 4 GWAS-linked SNPs included rs1801274 in FCGR2A, rs2857151 in HLA region, rs2254546 in BLK region and rs4813003 in CD40 region (Khor et al. 2011; Onouchi et al. 2012; Lee et al. 2012). The other 2 SNPs identified by genome-wide linkage analysis were rs113420705 in CASP3 and rs28493229 in ITPKC, which have been firmly established to underlie the genetic background of KD since their risk allele may lead to significant risk for KD occurrence by reducing T cell apoptosis and increasing T cell proliferation, respectively (Onouchi et al. 2008, 2010).

Methods

Study population

A total of 358 unrelated Han patients with typical KD were recruited from Sichuan province of Southwest China. The patients were admitted to the Sichuan Provincial People’s Hospital and West China Hospital of Sichuan University between 2003 and 2012, and were diagnosed according to the criteria of the Research Committee on Kawasaki Disease of Japan (Research Committee on Kawasaki Disease 1984) and the 5th revised edition in 2002 (http://www.kawasaki-disease.org/diagnostic/index.html). This case group included 222 boys and 136 girls (male-to-female ratio: 1.63). The mean age at diagnosis was 1.6 ± 1.9 (0.02–12.6) years, and 291 patients (81.3 %, 291/358) were younger than 5 years old. All patients were treated with a single infusion of IVIG (1 g/kg) and oral aspirin within the first 10 days of the onset of the illness, and 51 of the patients (14.2 %, 51/358) displayed resistance to the IVIG treatment according to the manifestations, including the return of fever and one or more of the initial symptoms that initially led to the diagnosis of KD within 2–7 days of treatment with IVIG (Wallace et al. 2000). All of the patients were detected coronary arteries with echocardiography in a quiet condition during the febrile stage and after hospital discharge. The scanning operation and the measurements of the coronary arteries were performed with a uniform protocol by the same cardiovascular paediatrician and the echocardiogram was blindly interpreted by the pediatric cardiologist. Of the 358 cases, 46 children (12.8 %, 46/358) suffered from CALs according to the observation of coronary arteries that contained an inner diameter of no less than 3 mm (<5 years old) or no less than 4 mm (>5 years old) or an inner diameter that was > 1.5 times that of the adjacent vessel in the echocardiogram 1 month after the onset of KD (Research Committee on Kawasaki Disease 1984). Among the cases, 172 were included in the 223 KD patients in our previous report of the association of ITPKC SNPs with KD in the population (Peng et al. 2012).

The 815 unrelated Han controls were also from Southwest China (male-to-female ratio: 1.68). The controls, including 624 diagnosed with acute bronchitis and 191 diagnosed with pneumonia when they visited, were recruited from West China Hospital, Sichuan University during the period of 2000–2012. Their mean age was 1.7 ± 2.1 (0.04–13.3) years. Of these control subjects, 631 who were younger than 5 years of age at the time of recruitment were followed up until they were 5 years old. None of the controls had KD or any autoimmune disease.

The study was approved by the Ethics Review Board of the Sichuan Provincial People’s Hospital and Sichuan University. Informed consent was obtained from the parents of all the subjects who were studied.

Genotyping of susceptibility SNPs to KD

Genomic DNA was extracted from a peripheral blood sample using a DNA purification kit (TaKaRa Co, Otsu, Japan). A total of 6 SNPs identified association with KD at genome-wide significance of P < 5.0 × 10−8 in previous studies were included in the present study (Onouchi et al. 2008, 2010, 2012; Khor et al. 2011; Lee et al. 2012). The SNPs included rs1801274 (A/G) in exon3 of FCGR2A, rs113420705 (G/A) in exon1 of CASP3, rs28493229 (G/C) in intron1 of ITPKC, rs2857151 (A/G) in HLA region, rs2254546 (A/G) located between BLK and FAM167A, and rs4813003 (C/T) located 4.9 kb downstream of CD40.

Applied Biosystems® TaqMan® SNP Genotyping Assays were used to determine SNP genotype following the manufacturer’s instructions with a 7900 HT Fast Real-time PCR System (Applied Biosystems, Foster City, CA, USA). PCR amplification was performed in a total volume of 5 μl of buffered solution containing 10 ng of genomic DNA, 2.5 μl TaqMan Genotyping Master Mix, 0.25 μl TaqMan genotyping assay mix on the 384-well reaction plate. The reaction profile was as follows: pre-denaturing at 95 °C for 10 min; 40 cycles of 95 °C for 15 s and 60 °C for 1 min. The genotyping data were collected and analysed by the ABI 7900 instrument. The negative, positive and duplicated controls were included in each 384-well test. Of the 1,173 samples, 20 % were performed DNA sequencing with the PCR products to confirm the genotypes of the SNPs using an ABI377A DNA sequencer (Applied Biosystems) and the Dye Terminator method. The SNPs’ genotypes of the samples detected by the two methods were completely concordant.

Statistical analysis

Hardy–Weinberg equilibrium (HWE) testing of case and control group was performed for the SNPs with SNPAlyze Version 6.0 (Dynacom Co, Ltd, Kanagawa, Japan). The frequencies of risk allele and high-risk genotype/genotype combination of the SNPs were compared between the patients with KD and the controls, using the χ2 test with SPSS 11.0 software (SPSS, Inc., Chicago, IL, USA). Cochran-Armitage trend tests of the risk alleles were conducted with SAS 9.1.3 software (SAS Institute Inc., Cary, NC, USA). The logistic regression analyses for the effect of the SNPs on KD, CALs and IVIG treatment were performed in the patients with KD and controls, in the KD patients with and without CALs, and in the KD patients who were resistant to IVIG treatment and those who were responsive with SPSS 11.0 software. The meta-analyses of ITPKC rs28493229 and CASP3 rs113420705 were performed with RevMan 5.1 software (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark). The publication bias was analysed using Begg and Egger test with Stata 11.0 software (StataCorp, TX, USA). The significant levels were corrected with Bonferroni method in multiple comparisons, and a value of P lower than the significant level was considered statistically significant in the analyses.

Statistical power analysis

When considering that the allele and genotype distribution differences of rs28493229 had not been observed between KD patients and controls in this study, we performed statistical power analysis with the Genetic Power Calculator (/http://www.pngu.mgh.harvard.edu/~purcell/gpc) (Purcell et al. 2003). We defaulted the test at 0.05 of type I error rate, and the genotype relative risks of rs28493229 as 1.89, by referring the data reported previously (Onouchi et al. 2008), with an estimated prevalence of KD of 9.81/100,000 (Li et al. 2008).

Results

Association of the SNPs with susceptibility to KD

The genotyping of 6 susceptibility SNPs to KD were performed in 1,173 Han subjects (358 patients with KD and 815 controls). In both the case and the control groups, their genotypes of the SNPs were in Hardy–Weinberg equilibrium. The comparisons of allele frequencies of the SNPs between the cases and the controls are summarized in Table 1. Of the SNPs, the alleles of two SNPs showed significantly different distribution between the cases and the controls, in which the risk allele A of FCGR2A rs1801274 and allele G of BLK rs2254546 were more common in the cases. The results of the trend tests also supported that the risk alleles of rs1801274 or rs2254546 had significant effect on susceptibility to KD (Table 1). Although the P values did not reach significance in the risk allele comparisons of rs113420705, rs2857151 and rs4813003, the meta-analyses by combining the data of this study and the previous reports showed significant associations of the risk alleles of the three SNPs with KD (meta-analysis with fixed effect model, α = 0.05; rs113420705: OR = 1.33, 95 % CI = 1.22, 1.43, P < 0.00001; rs2857151: OR = 1.41, 95 % CI = 1.27, 1.57, P < 0.00001; rs4813003: OR = 1.37, 95 % CI = 1.27, 1.47, P < 0.00001) (Onouchi et al. 2010, 2011, 2012; Kuo et al. 2011; Lee et al. 2012). The results suggested the possible effect of the three SNPs on KD in the population (Barrett et al. 2008). However, similarly to our previous report (Peng et al. 2012), the difference in the risk allele frequency of rs28493229 was not observed in present study with 95.3 % of the statistical power, although the meta-analysis showed that the risk allele carrier was more common in KD group than in control group (meta-analysis with random effect model, α = 0.05: OR = 1.41, 95 % CI = 1.02, 1.93, P = 0.03) (Table 2).
Table 1

Comparisons of allele and genotype frequencies between patients with KD and controls

SNP (Gene)

Allele

Allele frequency

Risk allele distribution

Genotype

Genotype frequency

P trendb (α′ = 0.004)

Patients, n (%)

Controls, n (%)

OR (95 % CI)

P valuea (α′ = 0.004)

Patients, n (%)

Controls, n (%)

rs1801274 (FCGR2A)

A

528 (73.7)

1,106 (67.9)

1.331 (1.094, 1.619)

0.004

AA

192 (53.6)

366 (44.9)

0.0035

G

188 (26.3)

524 (32.1)

  

AG

144 (40.2)

374 (45.9)

      

22 (6.2)

75 (9.2)

rs113420705 (CASP3)

A

218 (30.4)

430 (26.4)

1.222 (1.07, 1.483)

0.043

AA

32 (8.8)

60 (7.4)

0.0436

G

498 (69.6)

1,200 (73.6)

  

GA

154 (43.1)

310 (38.0)

     

GG

172 (48.1)

445 (54.6)

rs2857151 (HLA)

G

544 (76.0)

1,165 (71.5)

1.262 (1.031, 1.546)

0.024

GG

208 (58.1)

410 (50.3)

0.0225

A

172 (24.0)

465 (28.5)

  

AG

128 (35.6)

345 (42.3)

     

AA

22 (6.2)

60 (7.4)

rs2254546 (BLK)

G

590 (82.4)

1,246 (76.4)

1.443 (1.154, 1.805)

0.001

GG

245 (68.4)

482 (59.1)

0.0016

A

126 (17.6)

384 (23.6)

  

AG

100 (28.0)

282 (34.6)

     

AA

13 (3.6)

51 (6.3)

rs28493229 (ITPKC)

C

46 (6.4)

117 (7.2)

0.888 (0.624, 1.264)

0.509

CC

3 (0.8)

5 (0.6)

0.5151

G

670 (93.6)

1,513 (92.8)

  

GC

40 (11.2)

107 (13.1)

     

GG

315 (88.0)

703 (86.3)

rs4813003 (CD40)

C

448 (62.6)

963 (59.1)

1.158 (0.966, 1.387)

0.112

CC

140 (39.1)

276 (33.9)

0.1065

T

268 (37.4)

667 (40.9)

  

CT

168 (46.9)

411 (50.4)

     

TT

50 (14.0)

128 (15.7)

The risk alleles were underlined

aChi-squared test, the significant level was corrected with the formula of α′ = α/12 = 0.004 (12 tests in the table, 6 SNPs × 2 tests/SNP) according to the Bonferroni method

bCochran-Armitage trend test in 358 patients with KD and 815 controls. The significant level was corrected with the formula of α′ = α/12 = 0.004 (12 tests in the table, 6 SNPs × 2 tests/SNP) according to the Bonferroni method

Table 2

The meta-analysis of the association of risk allele carrier of ITPKC rs28493229 with KD

Study or subgroup

KD

Control

P value

Odds Ratio M-H, random (95 % CI)

Odds Ratio M-H, random (95 % CI)

Events

Total

Events

Total

Weight (%)

Chi et al. (2010) (Taiwan)

61

385

150

1,158

20.6

0.153

1.27 [0.92, 1.75]

https://static-content.springer.com/image/art%3A10.1007%2Fs00439-013-1279-2/MediaObjects/439_2013_1279_Figa_HTML.gif

Kuo et al. (2011) (Taiwan)

52

334

150

1,131

20.0

0.283

1.21 [0.86, 1.70]

Lin et al. (2011) (Taiwan)

44

280

38

492

16.8

0.001

2.23 [1.40, 3.53]

Onouchi et al. (2008) (Japan)

261

637

278

1,034

23.5

0.000

1.89 [1.53, 2.33]

Yan et al. (Sichuan)a

43

358

112

815

19.1

0.420

0.86 [0.59, 1.25]

Total (95 % CI)

 

1,994

 

4,630

100.0

 

1.41 [1.02, 1.93]

Total events

461

 

728

  

Heterogeneity: χ2 = 18.6, df = 4 (P = 0.0009), I2 = 78 %

 

Test for overall effect: Z = 2.11 (P = 0.03)

aThe data were from the present study

Furthermore, the single-variable regression analyses of the SNPs showed that the inheritance model of the risk alleles of rs1801274, rs2857151 and rs2254546 was more likely to be recessive (logistic regression analysis, α′ = 0.05/3 = 0.017 with Bonferroni correction; rs1801274 risk allele A: recessive model P = 0.006 vs. dominant model P = 0.082 vs. additive model P = 0.014; rs2254546 risk allele G: recessive model P = 0.003 vs. dominant model P = 0.068 vs. additive model P = 0.048; rs2857151 risk allele G: recessive model P = 0.014 vs. dominant model P = 0.452 vs. additive model P = 0.048). Because the P values did not surpass the Bonferroni threshold in the analyses of inheritance model of rs113420705 risk allele (logistic regression analysis, α′ = 0.05/3 = 0.017 with Bonferroni correction; recessive model P = 0.356 vs. dominant model P = 0.039 vs. additive model P = 0.112), we further carried out a meta-analysis for the association of risk allele carrier with KD by combining the data of this study and previous reports (Onouchi et al. 2010, 2011; Kuo et al. 2011) (Table 3). The result revealed that there was a significantly higher frequency of risk allele carrier of rs113420705 in patients with KD relative to the controls (meta-analysis with fixed effect model, α = 0.05: OR = 1.40, 95 % CI = 1.24, 1.58, P < 0.00001), supporting that the inheritance model of the risk alleles of rs113420705 was dominant. Owing to the absence of the genotype data of rs4813003 in previous reports, it was difficult to analyse the potential inheritance model of the risk allele, while it was still possible that the risk allele effect of rs4813003 was recessive considering the lower P value under the model relative to other model in present analyses (logistic regression analysis, α′ = 0.05/3 = 0.017 with Bonferroni correction: recessive model P = 0.084 vs. dominant model P = 0.445 vs. additive model P = 0.219). Then, we further compared the risk for KD between subjects with high-risk and low-risk genotype under the potential inheritance model of the risk alleles by conditional multi-variable logistic regression analysis of the SNPs except for rs28493229. The significant effect on KD susceptibility was observed in three SNPs (rs1801274, rs2254546 and rs2857151) (Table 4).
Table 3

The meta-analysis of the association of risk allele carrier of CASP3 rs113420705 with KD

Study or subgroup

KD

Control

P value

Odds ratio M-H, fixed (95 % CI)

Odds ratio M-H, fixed (95 % CI)

Events

Total

Events

Total

Weight (%)

 

Kuo et al. (2011) (Taiwan)

176

303

355

690

21.3

0.054

1.31 [1.00, 1.72]

https://static-content.springer.com/image/art%3A10.1007%2Fs00439-013-1279-2/MediaObjects/439_2013_1279_Figb_HTML.gif

Onouchi et al. (2010) (Japan)

443

637

618

1,031

33.7

0.000

1.53 [1.24, 1.88]

Onouchi et al. (2011) (Japan)

227

342

330

566

19.6

0.016

1.41 [1.07, 1.87]

Yan et al. (Sichuan)a

186

358

370

815

25.4

0.038

1.30 [1.01, 1.67]

Total (95 % CI)

 

1,640

 

3,102

100.0

 

1.40 [1.24, 1.58]

Total events

1,032

 

1,673

 

Heterogeneity: χ2 = 1.23, df = 3 (P = 0.75), I2 = 0 %

Test for overall effect: Z = 5.32 (P < 0.00001)

aThe data were from the present study

Table 4

Conditional multi-variable logistic regression analysis for the associations of the SNPs with KD

SNP (Allele)

The potential inheritance model of risk allelea

Genotypeb

OR (95 % CI)

P valuec (α′ = 0.01)

High-risk

Low-risk

 

rs1801274 (A/G)

Recessive model

AA

AG/GG

1.470 (1.138,1.898)

0.003

rs113420705 (G/A)

Dominant model

AA/AG

GG

1.299 (1.007,1.675)

0.044

rs2857151 (A/G)

Recessive model

GG

AG/AA

1.473 (1.140,1.904)

0.003

rs2254546 (A/G)

Recessive model

GG

AG/AA

1.609 (1.229,2.106)

0.001

rs4813003 (C/T)

Recessive model

CC

CT/TT

1.287 (0.990,1.672)

0.059

The risk alleles were underlined

aThe inheritance models of the SNPs were presumed according to the test results of additive, dominant and recessive model of single SNP with single-variable logistic regression method

bThe high-risk and low-risk genotype distinguished by the inheritance model of each SNP was assigned to “1” and “0” in the multi-variable regression analysis, respectively

cMulti-variable logistic regression analysis in 358 patients with KD and 815 controls. The P values were adjusted for gender and age. The significant level was corrected with the formula of α′ = α/5 = 0.01 (5 tests, 5 SNPs × 1 test/SNP) according to the Bonferroni method

Although there is no evidence for the interaction of the susceptibility loci in KD occurrence, we still conducted the association analyses in multi-locus models to investigate the compound effect of the SNPs on KD. For this, a total of ten 2-locus combinations were analyzed by conditional multi-variable logistic regression method. For each 2-locus combination, we performed 3 regression analyses with a different observed variable (SNP1, SNP2 or SNP1 + SNP2) and three common reference variables (SNP3/SNP4/SNP5) in each analysis, so as to compare the effects of the observed variables on KD in the same background. The results showed that two 2-locus combinations (rs1801274 + rs2857151 and rs113420705 + rs2254546) had a significant effect on KD with stronger association with KD relative to comparable single SNPs (Table 5). In the same way, we further investigated the effect of ten 3-locus combinations on KD by comparison with 2-locus combinations abstracted from respective 3-locus combination, and one 3-locus combinations (rs113420705 + rs2857151 + rs4813003) were found to have significant influence on KD with stronger association with KD relative to comparable 2-locus combinations (Table 6).
Table 5

Conditional multi-variable logistic regression analyses for the associations of 2-locus models of the SNPs with KD

Common variables in analysisa

Regression analysis I

Regression analysis II

Regression analysis III

Observed variablea

P valuec

Observed variablea

P valuec

Observed variableb

OR (95 % CI)

P valuec

rs2857151/rs2254546/rs4813003

rs1801274

0.001

rs113420705

0.020

rs1801274 + rs113420705

1.293 (0.973,1.720)

0.077

rs113420705/rs2254546/rs4813003

rs1801274

0.003

rs2857151

0.003

rs1801274 + rs2857151

1.764 (1.330,2.340)

0.0001

rs113420705/rs2857151/rs4813003

rs1801274

0.009

rs2254546

0.002

rs1801274 + rs2254546

1.422 (1.078,1.874)

0.013

rs113420705/rs2857151/rs2254546

rs1801274

0.004

rs4813003

0.081

rs1801274 + rs4813003

1.307 (0.933,1.831)

0.120

rs1801274/rs2254546/rs4813003

rs113420705

0.078

rs2857151

0.005

rs113420705 + rs2857151

1.219 (0.909,1.636)

0.186

rs1801274/rs2857151/rs4813003

rs113420705

0.045

rs2254546

0.001

rs113420705 + rs2254546

1.633 (1.246,2.141)

0.0004

rs1801274/rs2857151/rs2254546

rs113420705

0.045

rs4813003

0.060

rs113420705 + rs4813003

1.607 (1.163,2.221)

0.004

rs1801274/rs113420705/rs4813003

rs2857151

0.008

rs2254546

0.001

rs2857151 + rs2254546

1.506 (1.152,1.969)

0.003

rs1801274/rs113420705/rs2254546

rs2857151

0.004

rs4813003

0.082

rs2857151 + rs4813003

1.662 (1.210,2.284)

0.002

rs1801274/rs113420705/rs2857151

rs2254546

0.0004

rs4813003

0.043

rs2254546 + rs4813003

1.233 (0.917,1.657)

0.165

aThe high-risk and low-risk genotype of each SNP (see Table 3) was assigned to “1” and “0” in the regression analysis, respectively

bThe high-risk genotype at both loci and the low-risk genotype at any locus of the combined SNPs (see Table 3) was assigned to “1” and “0” in the regression analysis, respectively

cMultiple regression analysis in 358 patients with KD and 815 controls. The P values were adjusted for gender and age. The significant level was corrected with the formula of α′ = α/120 = 0.0004 (120 tests in the table, 30 regression analyses × 4 variables/analysis) according to the Bonferroni method

Table 6

Conditional multi-variable logistic regression analyses for the associations of 3-locus models of the SNPs with KD

Common variables in analysisa

Regression analysis I

Regression analysis II

Regression analysis III

Regression analysis IV

Observed variableb

P valuec

Observed variableb

P valuec

Observed variableb

P valuec

Observed variableb

OR (95 % CI)

P valuec

rs2254546/rs4813003

rs1801274 + rs113420705

0.109

rs1801274 + rs2857151

0.0001

rs113420705 + rs2857151

0.137

rs1801274 + rs113420705 + rs2857151

1.578 (1.094,2.276)

0.015

rs2857151/rs4813003

rs1801274 + rs113420705

0.116

rs1801274 + rs2254546

0.008

rs113420705 + rs2254546

0.0003

rs1801274 + rs113420705 + rs2254546

1.415 (1.004,1.994)

0.047

rs2254546/rs2857151

rs1801274 + rs113420705

0.074

rs1801274 + rs4813003

0.075

rs113420705 + rs4813003

0.002

rs1801274 + rs113420705 + rs4813003

1.656 (1.098,2.498)

0.016

rs113420705/rs4813003

rs1801274 + rs2857151

0.0004

rs1801274 + rs2254546

0.018

rs2857151 + rs2254546

0.005

rs1801274 + rs2857151 + rs2254546

1.653 (1.164,2.347)

0.005

rs113420705/rs2254546

rs1801274 + rs2857151

0.0001

rs1801274 + rs4813003

0.134

rs2857151 + rs4813003

0.002

rs1801274 + rs2857151 + rs4813003

1.909 (1.231,2.959)

0.004

rs113420705/rs2857151

rs1801274 + rs2254546

0.013

rs1801274 + rs4813003

0.137

rs2254546 + rs4813003

0.231

rs1801274 + rs2254546 + rs4813003

1.176 (0.773,1.789)

0.449

rs4813003/rs1801274

rs113420705 + rs2857151

0.257

rs113420705 + rs2254546

0.001

rs2857151 + rs2254546

0.004

rs113420705 + rs2857151 + rs2254546

1.512 (1.061,2.155)

0.022

rs1801274/rs2254546

rs113420705 + rs2857151

0.183

rs113420705 + rs4813003

0.006

rs2857151 + rs4813003

0.002

rs113420705 + rs2857151 + rs4813003

2.217 (1.446,3.400)

0.0003

rs1801274/rs2857151

rs113420705 + rs2254546

0.0004

rs113420705 + rs4813003

0.004

rs2254546 + rs4813003

0.185

rs113420705 + rs2254546 + rs4813003

1.630 (1.102,2.411)

0.015

rs1801274/rs113420705

rs2857151 + rs2254546

0.003

rs2857151 + rs4813003

0.003

rs2254546 + rs4813003

0.260

rs2857151 + rs2254546 + rs4813003

1.250 (0.836,1.869)

0.277

aThe high-risk and low-risk genotype of each SNP (see Table 3) was assigned to “1” and “0” in the regression analysis, respectively

bThe high-risk genotype at all loci and the low-risk genotype at any locus of the combined SNPs (see Table 3) was assigned to “1” and “0” in the regression analysis, respectively

cMultiple regression analysis in 358 patients with KD and 815 controls. P values were adjusted for gender and age. The significant level was corrected with the formula of α′ = α/120 = 0.0004 (120 tests in the table, 40 regression analyses × 3 variables/analysis) according to the Bonferroni method

Furthermore, we predicted the risk for KD in the single SNPs and multi-locus models observed significant effect on KD in regression analyses, by comparing the frequency of high-risk genotype or genotype combination between the patients with KD and controls. As a result, the 3-locus model of rs113420705 +rs2857151 + rs4813003 showed the most significant association with KD, and the risk of subjects with high-risk genotype combination of the model for KD increased 1.154 time with respect to subjects with other genotype combination of the three SNPs (Table 7). However, the sensitivity, specificity, positive predictive value and negative predictive value of multi-locus combination did not show more significant predictive value for KD susceptibility relative to single SNP (Supplementary Table 1).
Table 7

The risk prediction of the single SNP and the multi-locus models for KD

SNP

Patients (n = 358)

Controls (n = 815)

OR (95 % CI)

P valuec

High-riska, n (%)

Low-riskb, n (%)

High-riska, n (%)

Low-riskb, n (%)

Single SNP*

 rs1801274

192 (53.6)

166 (46.4)

366 (44.9)

549 (55.1)

1.419 (1.106,1.821)

0.006

 rs2857151

208 (58.1)

150 (41.9)

410 (50.3)

405 (49.7)

1.370 (1.066,1.760)

0.014

 rs2254546

245 (68.4)

113 (31.6)

482 (59.1)

333 (40.9)

1.498 (1.151,1.949)

0.003

2-locus model*

 rs1801274 + rs2857151

115 (32.1)

243 (67.9)

185 (22.7)

630 (77.3)

1.612 (1.223,2.123)

0.0007

 rs113420705 + rs2254546

129 (36.0)

229 (64.0)

216 (26.5)

599 (73.5)

1.562 (1.197,2.038)

0.001

3-locus model*

 rs113420705 + rs2857151 + rs4813003

45 (12.6)

313 (87.4)

51 (6.3)

764 (93.7)

2.154 (1.412,3.284)

0.0003

* The single SNPs and multi-locus models showing significant effect on Kawasaki disease in multi-variable logistic regression analyses

aSubjects with high-risk genotypes shown in Table 3 at all loci

bSubjects with low-risk genotype shown in Table 3 at any locus

cChi-squared test, the significant level was corrected with the formula of α′ = α/6 = 0.008 (6 tests) according to the Bonferroni method

Association of the SNPs with clinical manifestations of KD

In present study, we analysed the association of the SNPs with two important clinical phenotypes including CALs formation and the treatment effect of IVIG in 358 patients with KD. As is shown in Supplementary Table 2, we found no significant association of any single SNP with CALs by conditional multi-variable logistic regression analysis of the SNPs. However, the meta-analyses of the association of risk allele carriers of rs113420705 with CALs first revealed significant higher risk of the carriers for CALs relative to non-carrier of risk allele (Meta-analysis with fixed effect model, α = 0.05: OR = 1.54, 95 % CI = 1.00, 2.37, P = 0.05) (Table 8). In addition, the 2-locus combination model of rs1801274 + rs2857151 was observed the association with CALs when two 2-locus models showing significant effect on KD were analysed (conditional multi-variable logistic regression analysis, α′ = 0.05/24 = 0.002 with Bonferroni correction: P = 0.002), while the association of either rs1801274 or rs2857151 with CALs did not reach significance in regression analyses with rs113420705/rs2254546/rs4813003 as the common variables (conditional multi-variable logistic regression analysis, α′ = 0.002 with Bonferroni correction; rs1801274: P = 0.050; rs2857151: P = 0.162). The risk of subjects with high-risk genotype combination of the model for CALs increased 1.391 time with respect to subjects with other genotype combination of the two SNPs (Chi-squared test, α = 0.05: OR = 2.391, 95 % CI = 1.277, 4.477, P = 0.005). Meanwhile, we found no significant association of the 3-locus model of rs113420705 + rs2857151 + rs4813003 with CALs (multi-variable logistic regression analysis, α’ = 0.004 with Bonferroni correction: P = 0.071). In addition, we also performed the association analysis of the SNPs with IVIG treatment effect, while the association of the SNPs with IVIG unresponsiveness was not observed in the population (Supplementary Table 3).
Table 8

The meta-analysis of the association of risk allele carrier of CASP3 rs113420705 with CALs

Study or subgroup

KD

Control

P value

Odds ratio M-H, fixed (95 % CI)

Odds ratio M-H, fixed (95 % CI)

Events

Total

Events

Total

Weight (%)

Kuo et al. (2011) (Taiwan)

20

29

156

272

26.6

0.228

1.65 [0.73, 3.76]

https://static-content.springer.com/image/art%3A10.1007%2Fs00439-013-1279-2/MediaObjects/439_2013_1279_Figc_HTML.gif

Onouchi et al. (2011) (Japan)

22

30

329

505

28.1

0.359

1.47 [0.64, 3.37]

Yan et al. (Sichuan)a

28

46

158

312

45.3

0.195

1.52 [0.81, 2.85]

Total (95 % CI)

 

105

 

1,089

100.0

 

1.54 [1.00, 2.37]

Total events

70

 

643

    

Heterogeneity: χ2 = 0.04, df = 2 (P = 0.98), I2 = 0 %

Test for overall effect: Z = 1.97 (P = 0.05)

aThe data were from the present study

To obtain the more evidence on the potential role of the genetic loci in KD occurrence, we analyzed the association of the SNPs with age at onset of KD, while we found no evidence showing that the onset of KD is earlier in patients with high-risk genotype of any SNPs relative to other genotypes, which may be due to the minor effect of single gene or the stratified interference caused by the time difference in suffering from infection. Furthermore, we investigated the association of the SNPs with gender in patients with KD, considering that the significantly higher of males for KD may be partially due to the lower threshold relative to females. If so, it is possible to find the different proportion of gender between KD patients with high-risk genotype combination and low-risk genotype combination of multi-locus model. However, the association was not observed in the analyses (Supplementary Table 4).

Discussion

In recent years, the genetic research on Kawasaki disease has made considerable progress. Until now, 6 GWAS for KD have been published (Onouchi et al. 2012; Lee et al. 2012; Kim et al. 2011; Tsai et al. 2011; Khor et al. 2011; Burgner et al. 2009), in which 3 studies reported 4 genetic loci associated with KD at genome-wide significance of P < 5.0 × 10−8 (Khor et al. 2011; Onouchi et al. 2012; Lee et al. 2012). The loci include FCGR2A gene in 1q23, HLA region in 6p21.3, BLK region in 8p23-22 and CD40 region in 20q12-13.2. Their associations with KD were observed in European (FCGR2A), Japanese (HLA/BLK/CD40) and Taiwanese (BLK/CD40). Owing to the lack of replication study, it remains unknown whether there is the similar association of the genetic loci with the risk for KD in other populations.

In present study, we performed the association analysis of six SNPs in 4 GWAS-linked loci and two identified susceptible loci ITPKC and CASP3 with KD in Han Chinese, an ideal population for the replication study due to that its’ incidence of KD is next only to those of Japan and Korea (Nakamura et al. 2012; Park et al. 2011; Ma et al. 2010; Huang et al. 2009; Du et al. 2007). Our results showed that the risk alleles of FCGR2A rs1801274 and BLK rs2254546 were more common in the cases relative to the controls. Further genotype analyses under the potential inheritance model of the SNPs with conditional multi-variable logistic regression method indicated that, besides rs1801274 and rs2254546, HLA rs2857151 also significantly influence the risk for KD. Together with the meta-analysis result of CASP3 rs113420705, the observations strongly support the idea that the genetic loci have a significant effect on KD susceptibility, although the predictive value of single SNP for KD onset is limited.

When considering the minor effect of single genetic locus on KD, we further analysed the compound role of multi-locus models in KD. As a result, two 2-locus and one 3-locus models were observed significant effect on KD with stronger association with KD relative to comparable single SNPs or 2-locus models. The subjects with high-risk genotype combination of 3-locus model rs113420705 + rs2857151 + rs4813003 presented the most significant risk for KD. The findings imply that some SNP combinations may have additive effect on KD susceptibility. However, we also observed different case, for example, in 2-locus model of rs2254546 + rs4813003 the high-risk genotype combination did not show significant effect on KD under the background of rs1801274/rs113420705/rs2857151 (multi-variable logistic regression analysis, α′ = 0.0004 with Bonferroni correction: P = 0.165), whereas the risk allele homozygote of rs2254546 alone had a significant risk for KD under the same background (multi-variable logistic regression analysis, α′ = 0.0004 with Bonferroni correction: P = 0.0004). As another example, the high-risk genotype combination of rs1801274 + rs113420705 + rs2254546 also did not significantly affect KD onset (multi-variable logistic regression analysis, α′ = 0.0004 with Bonferroni correction: P = 0.047), while rs113420705 + rs2254546 showed strong association with KD under the same background consisted of rs1801274/rs113420705 (multi-variable logistic regression analysis, α′ = 0.0004 with Bonferroni correction: P = 0.0003). The results offer additional evidence on the presence of the complex compound effect of the genetic loci on KD occurrence and development, which also is supported by the result that higher risk for CALs formation was found in the risk allele homozygotes at both loci of rs1801274 and rs2857151 relative to individual with other genotype combinations, while the risk was not observed in the risk allele homozygotes of any single SNP.

Although it is rather difficult to speculate the mechanism of the compound effect of the genome-linked genes on KD, we noticed that their protein productions possess a similar functional characteristic, that is, they all are involved in immune function of organism, in which FCGR2A plays an essential role in the protection against foreign antigens by removing antigen–antibody complexes from the circulation, and transduces activating signals into cells via immune receptor when ligated with immune complexes (Nimmerjahn and Ravetch 2008; Hibbs et al. 1988). BLK is a src family tyrosine kinase expressed mainly in B cell lineage and transduces B cell receptor signals with a critical role in B cell activation and antibody secretion (Nemazee and Weigert 2000; Wasserman et al. 1995; Dymecki et al. 1992). CD40 potentially contributes to autoimmune disease process through the selection of auto-reactive T cells and the activation of B cells and T cells. Increased CD40 signals lead to the production of pro-inflammatory cytokines and chemokines, which contributes to tissue destruction and inflammatory cell influx (Iezzi et al. 2009; Akiyama et al. 2008; Jacobson et al. 2007). Therefore, it is possible that some of the genetic loci above together participate in certain immune pathway and their risk allele combinations may have an additive effect on KD/CALs susceptibility relative to single gene background, which was supported by the observation that two of three multi-locus models presenting more significant association with KD relative to single SNP or 2-locus model involved in HLA rs2857151 in the present study, and the latter as a significant immune-related genetic loci has been found to participate simultaneously many immune pathways including immune-complex processing and immune signal transduction (Harley et al. 2009).

In 2008, the first functional SNP rs28493229 (G/C) of KD was identified. The SNP is located in ITPKC gene, whose protein production is a negative regulator of the Ca2+/NFAT pathway in T cells. The ITPKC mRNA level for rs28493229 C allele was 30 % lower than that of the G allele, which may up-regulate T cell activity and significantly increase risk for KD and CALs (Onouchi et al. 2008). After 2 years, another SNP rs113420705 (G/A) of CASP3 gene was found to have a functional influence on KD susceptibility. Caspase-3 may also be a potentially inhibitor to the activity of the Ca2+/NFAT pathway in T cells. CASP3 expression was regulated by binding of the NFATc2 with the sequence surrounding rs113420705. In comparison of G allele of the SNP, the A allele presents weaker binding with NFATc2, resulting in decreased expression of CASP3 and the inhabitation of T cell apoptosis so as to prolong pro-inflammatory state (Onouchi et al. 2010). Further 2-locus model (rs28493229 and rs113420705) study identified the higher risk for CALS and IVIG unresponsiveness in the KD-susceptible-allele carriers relative to non-carriers in Japanese (Onouchi et al. 2011). Similar to our previous report (Peng et al. 2012), the association of rs28493229 with KD still was not observed in present replication study with more subjects. Although the meta-analysis suggested significant association of rs28493229 with KD, we should attach importance to the remarkable heterogeneity of the functional SNP of KD in different populations, such as Korea and now Han Chinese (Khor et al. 2011). This suggests the necessity of further investigating the effect of the SNP on KD in more populations. Meanwhile, in present population, the functional SNP rs113420705 of CASP3 was linked to KD susceptibility in multi-locus models, and the results of meta-analysis first revealed the higher risk for CALs in the risk allele carriers of the SNP, together supporting the idea that CASP3 gene plays an important role in KD occurrence and outcomes.

In conclusion, our results confirmed the association of the genetic loci including FCGR2A, HLA, BLK, CD40 and CASP3 with KD. The significant compound effects of multi-locus models imply that the interaction of the genetic loci may an important modulating factor for KD susceptibility and outcomes, which may be due to that, some loci are involved together in certain unknown immune pathway related to KD/CALs. However, the predictive value for KD/CALs is limited even if the multiple loci are under consideration, highlighting the complexity of genetic effect on KD occurrence and development. Further seeking genome-wide susceptible loci to KD/CALs and investigating the mechanism of the effect of the GWAS-linked genes on KD/CALs will be helpful for better understanding the pathophysiology of KD.

Acknowledgments

This study was supported by the Health Department Program of Sichuan Province, China (No.120079), and the National Key Technologies R&D Program of China (No. 2006BAI05A10).

Supplementary material

439_2013_1279_MOESM1_ESM.docx (33 kb)
Supplementary material 1 (DOCX 32 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2013