Transforming Growth Factor-β Signaling Pathway in Patients with Kawasaki Disease

Abstract

Background

Transforming growth factor (TGF)-β is a multifunctional peptide that is important in T-cell activation and cardiovascular remodeling, both of which are important features of Kawasaki disease (KD). We postulated that variation in TGF-β signaling might be important in KD susceptibility and disease outcome.

Methods and Results

We investigated genetic variation in 15 genes belonging to the TGF-β pathway in a total 771 KD subjects of mainly European descendent from the US, UK, Australia and the Netherlands. We analyzed transcript abundance patterns using microarray and RT-PCR for these same genes and measured TGF-β2 protein levels in plasma. Genetic variants in TGFB2, TGFBR2 and SMAD3 and their haplotypes were consistently and reproducibly associated with KD susceptibility, coronary artery aneurysm formation, aortic root dilatation, and intravenous immunoglobulin treatment response in different cohorts. A SMAD3 haplotype associated with KD susceptibility replicated in two independent cohorts and an intronic SNP in a separate haplotype block was also strongly associated (A/G, rs4776338) (p=0.000022, OR 1.50, 95% CI 1.25-1.81). Pathway analysis using all 15 genes further confirmed the importance of the TGF-β pathway in KD pathogenesis. Whole blood transcript abundance for these genes and TGF-β2 plasma protein levels changed dynamically over the course of the illness.

Conclusions

These studies suggest that genetic variation in the TGF-β pathway influences KD susceptibility, disease outcome, and response to therapy and that aortic root and coronary artery Z scores can be used for phenotype/genotype analyses. Analysis of transcript abundance and protein levels further support the importance of this pathway in KD pathogenesis.

Kawasaki disease (KD, Table 1 for abbreviations) is an acute, systemic vasculitis associated with cardiovascular sequelae that include coronary artery aneurysms (CAA), aortic root (AoR) dilatation, and myocardial fibrosis.^{1, 2} CAA develop in 25% of untreated patients and 3-5% of patients treated with intravenous immunoglobulin (IVIG) within the first 10 days after fever onset.³ AoR dilatation occurs in approximately 15% of treated patients.^{4, 5} Histological examination of tissues from KD autopsies has revealed myocardial fibrosis, dysregulation of angiogenesis and coronary artery thrombosis, stenosis, and calcification.^{6, 7} The etiology of KD remains unknown although an infectious trigger affecting genetically susceptible hosts is suspected.^8-12 Genetic and immunologic studies suggest that T-cell activation and regulation play important roles in the pathogenesis of KD.^12-14

Table 1. Abbreviations used in this paper.

TGF	Transforming growth factor
SMAD	Mothers against decapentaplegic, drosophila, homolog of
ACVRL1	Activin A receptor, type II-like 1
FURIN	Furin
ENG	Endoglin
MAP3K7	Mitogen-activated protein kinase kinase kinase 7
EMILIN1	Elastin microfibril interfacer 1
FoxP3	Forkhead box P3
NFAT	Nuclear factor of activated T cells
TAF1B	TATA box binding protein-associated factor, RNA polymerase I, B
KD	Kawasaki disease
CAA	coronary artery aneurysm
RCA	right coronary artery
LAD	left anterior descending coronary artery
AoR	aortic root
IVIG	intravenous immunoglobulin
TDT	transmission disequilibrium test
SNP	Single nucleotide polymorphism
CA Zmax	Maximum Z score of any coronary artery segments at any time point
LD	Linkage disequilibrium
FDR	False discovery rate

Transforming growth factor (TGF)-β is a multifunctional peptide that regulates proliferation, differentiation, apoptosis, and migration in many cell types. In the cardiovascular system, TGF-β can induce neoangiogenesis, cardiomyocyte hypertrophy, calcification and fibrosis.^{15, 16} Alteration of TGF-β signaling (Figure 1) has been implicated in the pathophysiology of several vascular disorders including Marfan and Loeys Dietz syndromes, which are associated with CAA and AoR dilatation, and vascular complications of sickle cell anemia. ^{17, 18 19 20} In the immune system, TGF-β modulates the balance of pro-inflammatory and anti-inflammatory T-cells through a complex set of interactions. The importance of cardiovascular remodeling and immune activation in KD suggest that TGF-β may play an important role in KD pathogenesis.^{7, 8, 12-14}

Figure 1.

TGF-β signaling pathway TGF-β is secreted in a latent form and activation is mediated by several molecules including furin and emilin. Active TGF-β peptides (TGFB1, 2, 3) bind to the Type II receptor (TGFBR2), which recruits and activates Type I receptors such as TGFBR1 or ACVRL1. The activated Type I receptor phosphorylates receptor-specific SMAD molecules: TGFBR1 phosphorylates SMAD2 and 3, and ACVRL1 phosphorylates SMAD1, 5 and 8. These activated SMADs form a larger complex with Smad4 and translocate to the nucleus where they regulate gene transcription. Accessory receptors such as ENG have a role in the balance of ACVRL1 and TGFBR1 signaling to regulate endothelial cell proliferation. In addition to the classical SMAD signaling pathway, a SMAD-independent pathway using alternative molecules such as MAP3K7 can also mediate TGF-β signaling. TGFB: Transforming growth factor-β, TGFBR: Transforming growth factor-β receptor, EMILIN1: Elastin microfibril interfacer 1, SMAD: SMAD family member, MAP3K7: Mitogen-activated protein kinase kinase kinase 7, ENG: Endoglin, ACVRL1: Activin A receptor type II-like 1, FURIN: Furin

We investigated this hypothesis by analyzing genetic variation and transcript abundance patterns for genes in the TGFβ pathway. We also measured levels of TGF-β2 in acute and convalescent KD plasma samples. These studies revealed that genetic variation in the TGFβ pathway influences KD susceptibility, severity, and treatment response.

Methods

Subjects

The recruitment of KD patients and the details of their clinical presentation and diagnosis have been previously described.⁹ A total of 771 KD subjects were genotyped and divided into different cohorts for analysis (Supplemental Table 1, Figure 2, Supplemental Methods). IVIG resistance was defined as previously described.²¹ The Institutional Review Boards of the participating centers reviewed and approved this study and parental consent and assent as appropriate were obtained from parents and participants.

Figure 2.

Cohorts and phenotypes analyzed in genetic association studies

Blood samples

RNA was isolated from whole blood collected in PAXgene tubes from US KD subjects (Supplemental Table 2, Supplemental Figure 1) as previously described.²¹ Plasma (sodium citrate) for ELISA was collected from KD patients during acute (pre-IVIG) and convalescent (≥1 month after fever onset and normalization of erythrocyte sedimentation rate and platelet count) phases. Blood samples were processed within 48 hours of collection and plasma was stored at -70°C.

Echocardiography

All KD subjects had echocardiographic assessment during the acute and convalescent stages of their illnesses. For all cohorts, coronary artery (CA) lesions (aneurysm or ectasia) were defined according to the Japanese Ministry of Health criteria. ²² The coronary artery status of the US subjects (Cohort 4) was assessed by echocardiography during the acute and subacute phase and serially thereafter as dictated by patient status. Measurements of the internal diameters of the proximal right (RCA) and left anterior descending (LAD) coronary arteries were normalized for body surface area and expressed as standard deviation units from the mean (Z scores).⁵ The worst-ever Z-score for either the RCA or LAD at any time point or the Z score of the largest aneurysm was used for the continuous variable analysis (CA Z max).

AoR dimensions (annulus and sinus) were measured at only two US sites (Rady Children's Hospital San Diego and Boston Children's Hospital). Measurements performed within 1 year from the onset of KD were normalized for body surface area and expressed as Z-scores.² The variable, (Ao Z max), which was the largest dimension of the AoR at any timepoint during the first year after disease onset, was used for the continuous variable analysis. For subjects who were not seen within the first year after disease onset, we used measurement of the AoR from the first available echocardiogram.

Genotyping

Genomic DNA from whole blood or mouth wash samples was extracted as previously described.⁹. Haplotype-tagging SNPs (n=164), selected from Hapmap Caucasian populations (Supplemental Table 3) with minor allele frequencies ≥ 0.1, from 15 genes in the TGF-β signaling pathway (Figure 1) were genotyped using a custom Illumina™ Oligo Pool Assay.⁹ All SNPs met quality control criteria with genotype call rate ≥ 93% and no deviation from Hardy-Weinberg equilibrium in controls and parents of KD subjects.

Microarray analysis

Two different types of microarray platforms (Stanford University Lymphochip, 19 KD subjects, and Agilent G4112F Human Whole Genome Oligo Arrays (design #14850), 20 KD subjects) were used to study two independent cohorts of KD subjects.^{21, 23} (Supplemental Tables 2 and 4, Supplemental Figure 1, Supplemental Methods).

RT-PCR

To validate the microarray results, transcript abundance levels were measured by RT-PCR for SMAD1, SMAD3, ACVRL1, FURIN, TGFBR3 and TGFBR2 for 14 KD subjects with acute and convalescent paired whole blood RNA samples (Supplemental Table 2, Supplemental Figure 1, Supplemental Methods). Relative abundance of the target transcripts was normalized to the expression level of the house keeping gene, TATA box binding protein-associated factor, RNA polymerase I, B (TAF1B), as previously described.²¹

ELISA

The concentrations of TGFB2 in paired acute and convalescent plasma samples were compared in 18 KD subjects (Quantikine, R&D system) (Supplemental Table 2, Supplemental Figure 1).

Statistical Methods

The transmission disequilibrium test (TDT) was performed as previously described¹⁰ Case-control association studies based on disease and CA status were analyzed using the general linear model (Supplemental Methods). Bonferroni correction for multiple comparisons was used. Association analyses between genetic variants and Z max, gene expression and protein levels were performed using paired and unpaired non-parametric tests (Supplemental Methods). Analysis was performed using the SNPassoc and DGCgenetics packages in the R software (version 2.6.2, http://www.r-project.org/). Pathway and gene-based analyses on Cohort 1 (Dutch case-control) and Cohort 2 (451 trios) were applied as previously described²⁴ with some modification (Supplemental Methods). Haplotype associations were analyzed using a moving window approach (window length from 2 to 7 SNPs). Score statistics were computed by PLINK²⁵ to test associations between the haplotypes and various traits. LD plots were made using Haploview²⁶ with genotyping data from Cohort 3 and 4.

Results

TGF-β signaling pathway and KD susceptibility

Case-control analysis of the Dutch subjects (Cohort 1: 128 KD vs. 159 control subjects) identified SNPs in 6 genes in the pathway that modestly influenced susceptibility to KD (nominal p 0.0031-0.047) (Supplemental Table 5). The significance of genetic variation in 3 of these 6 genes (SMAD3, TGFB2, and TGFBR2) was replicated in the TDT analysis of the independent US/UK/Aus subjects (Cohort 2: 451 trios) that identified 7 different SNPs associated with KD susceptibility (Table 2, Figure 3). Five of the 7 SNPs were in SMAD3, with the most significant result for an intronic SNP (A/G, rs4776338) for which 272 of the 453 heterozygous parents transmitted the G allele to their affected offspring (nominal p=0.00002, OR 1.50, 95%CI 1.25-1.81, corrected p=0.003).

Table 2. TDT analysis of genetic variants and KD susceptibility in US/UK/Australian trios (n=451)^*.

Chr.	Gene	rs number	Allele 1/2	Allele 2			OR	95%CI		p†
				Frequency	Transmitted	Untransmitted
1	TGFB2	rs2796817	A/C	0.13	139	98	1.42	1.10	1.84	0.0092
3	TGFBR2	rs11466480	A/C	0.03	12	27	0.44	0.23	0.88	0.024
15	SMAD3	rs4776338	A/G	0.42	272	181	1.50	1.25	1.81	0.000022
		rs7162912	C/A	0.31	229	174	1.32	1.08	1.60	0.0071
		rs12901071	A/G	0.29	230	169	1.36	1.12	1.66	0.0026
		rs1438386	A/G	0.41	190	237	0.80	0.66	0.97	0.026
		rs6494633	G/A	0.43	174	219	0.80	0.65	0.97	0.026

^*Results are shown only for SNPs with a nominal p-value <0.05

^†p values <0.0003 remained significant after Bonferroni correction

OR: odds ratio, CI: confidence interval

Figure 3.

Significant SNP locations for TGFB2 (A), TGFBR2 (B) ans SMAD3 (C) Arrows show the location of significant SNPs genotyped in this study. Gene structure and the location of SNPs are shown: boxes= exons and 3′ and 5′ untranslated regions;. Underlined text highlights SNPs with nominal p<0.01. Letter code above SNP rs# refers to type of analysis and cohort: CC: susceptibility in Cohort 1 (Supplemental Table 5), TDT: susceptibility in Cohort 2 (Table 2), CAA3 and CAA4: association with CAA in Cohort 3 and Cohort 4, respectively (Supplemental Table 6), CAZ: association with CA Z-worst (Supplemental Table 7), AoR; association with AoR Z-worst (Supplemental Table 8), IVIG; association with IVIG treatment response (Supplemental Table 9). LD maps are shown for SNPs of interest.

TGF-β signaling pathway and coronary artery outcome

Genetic variation in TGFB2, TGFBR2, and SMAD3 consistently influenced coronary artery outcome in 2 independent, non-overlapping cohorts: Cohort 3 from the UK, Australia, and the Netherlands (CAA-: n=362, CAA+: n=73) and Cohort 4 from the US (CAA-: n=186, CAA+: n=51) (Supplemental Table 6). Although the associated SNPs in Cohort 3 and 4 were different, many of the SNPs co-localized to the first intron of each of the 3 genes (TGFB2, TGFBR2 and SMAD3) (Figure 3). Slight differences in LD between Cohorts 3 and 4 were detected in these regions (Supplemental Figure 2 A-F) that may account for the variability in the implicated SNPs.

We analyzed the association between genetic variants and CA Z max from a subset of Cohort 4 (US, n=176). Significant differences in the distribution of CA Z max for 11 SNPs in 4 genes in the pathway (Supplemental Table 7) were revealed. Whether the analysis was performed using Zmax (n=176) or using the presence or absence of aneurysms in the complete Cohort 4 (176 + 61= 237), 3 SNPs in TGFB2 (rs10482751, rs2027567, rs12029576) and 2 SNPs in SMAD3 (rs12910698, rs4776339) were consistently associated. (Supplemental Table 6 and 7, Figure 3).

TGF-β signaling pathway and aortic root dimension

The maximal internal diameter for the aortic root normalized for body surface area (Ao Z max) was available for a subset of the US subjects (n=98) (Supplemental Table 1). Twenty SNPs in 8 genes in the pathway, including TGFB2, TGFBR2 and SMAD3, were significantly associated with Ao Z max (Supplemental Table 8). One SNP (rs9310940) within TGFBR2 and one SNP (rs12901071) within SMAD3 were significant in both the analysis of CA outcome and the analysis of AoR dilatation (Figure 3).

Association with genetic variants and IVIG treatment response in the US subjects

Case-control analysis of treatment response as a function of genotype was performed for the US subjects (IVIG-resistant n=46, IVIG-responsive n=147) (Supplemental Table 9). The same 3 genes (TGFB2, TGFBR2 and SMAD3) that were associated with KD susceptibility, coronary artery aneurysm formation, and aortic root dilatation were again associated with IVIG treatment response.

Haplotype analysis for TGFB2, TGFBR2 and SMAD3

Three genes were consistently associated with KD susceptibility and outcome in different cohorts and the associated SNPs clustered in discrete intronic regions (Figure 3 A-C). To identify haplotypes associated with susceptibility and CAA, we analyzed Cohorts 1 and 2 (susceptibility) and Cohorts 3+4 (CAA) for the three genes (TGFB2, TGFBR2 and SMAD3) that were consistently implicated in the single SNP analysis. Haplotypes with frequencies greater than 1% and nominal p values <0.01 are shown in Supplemental Figure 3. panel A-C. Significant haplotype blocks in TGFB2 and SMAD3 were detected in Cohort 1 (case-control) and were replicated in Cohort 2 (TDT) (Supplemental Figure 3 A-C). However, only for rs4846476 in TGFB2 did the p value dramatically increase when compared to the single SNP analysis (p= 0.00061 vs.0.013, respectively), suggesting that the other significant haplotypes mostly reflected the effect of genetic variation already detected in the single SNP analysis. In the haplotype analysis for CAA+ vs CAA-, no haplotype exceeded the significance of the individual SNPs (data not shown).

Pathway analysis

There is an increasing recognition that genetic contribution to disease may operate through a combined effect of multiple genes in a biological pathway. Analysis of the cumulative variation of 15 genes in the TGF-β pathway in Cohorts 1 and 2 showed a significant association of the pathway with susceptibility (P= 0.00065) (Supplemental Table 10). Gene-based analysis on the combined dataset identified TGFB2 (P=0.006), TGFBR2 (p=0.08), ACVRL1 (P=0.04), SMAD3 (p=0.06) and FURIN (P=0.01) as most highly associated with susceptibility.

TGF-β pathway transcript abundance and plasma levels in acute and convalescent KD

To discover differences in transcript abundance levels of genes in the TGF-β pathway between acute and convalescent KD samples, we analyzed two independent microarray experiments each with 19 paired samples (Lymphochip array: 2 subjects had CAA and 2 had IVIG-resistance; Agilent array: 1 subject had CAA and 8 had IVIG-resistance). We found differential transcript abundance for 17 genes (Lymphochip array) and 19 genes (Agilent array) in the TGF-β pathway (17 genes were common between two platforms, Supplemental Table 4). Transcript abundance levels for SMAD3 and TGFBR3 were significantly lower during the acute compared to convalescent stage on both platforms; transcript abundance for ACVRL1, SMAD1, and FURIN were higher during the acute phase on both array platforms (Table 3, Supplemental Figure 4). After Bonferroni correction, all transcripts on the Lymphochip array except for SMAD3 remained significant. All transcripts on the Agilent array were under the q value of FDR. To validate the microarray results, we performed RT-PCR on cDNA for transcripts of ACVRL1, SMAD1, FURIN, SMAD3, and TGFBR3. Increased transcript levels for ACVRL1 (p=0.02), SMAD1 (p=0.001) and decreased TGFBR3 (p=0.02) were documented during the acute phase (Figure 4A, B, D). FURIN was detected in only three of 14 acute whole blood RNA samples and in none of the convalescent samples. To test whether the change in expression levels was simply a function of neutrophil number, we examined the correlation between ACVRL1, SMAD1, SMAD3, TGFBR3 and FURIN gene expression levels and found only weak associations: r² ≤ 0.36 for absolute neutrophil count (Agilent array) and r² ≤ 0.58 for percent neutrophil change (Lymphochip array). Therefore, the dynamic changes in gene expression levels could not be explained simply by differences in cell populations in the peripheral blood. Genotype data were available for 14 of 15 subjects whose gene expression levels were measured by RT-PCR, (Supplemental Figure 1). We analyzed the association between changes in transcript levels and genotype for the most significant SNPs in ACVRL1 rs11169953, SMAD1 rs6537355 and SMAD3 rs4776338 (Figure4E-G). No significant association was detected in this small number of subjects. Plasma concentrations of TGF-β2 during the acute phase were lower compared to the convalescent phase (p=0.002) (Figure 5) (n=18, 3 had CAA and 5 were IVIG-resistant). For the 13 subjects whose genotypes were available, we analyzed the association between changes of TGFB2 plasma levels and genetic variants of rs12029576, the most significantly associated SNP in TGFB2 (p=0.0011 Supplemental Table 6). No significant association was detected in this small number of subjects.

Table 3. Transcript abundance patterns for differentially expressed genes in the TGF-β pathway^*.

Gene	Transcript abundance	Lymphochip array n=19		Agilent array n=19§
		p†	Fold change	p‡	Fold change
ACVRL1	↑Acute, ↓Conv.	0.0024	1.34	4.29E-08	2.03
SMAD1		0.00027	1.25	3.84E-06	2.37
FURIN		0.0012	1.37	6.27E-07	2.38
SMAD3	↓Acute, ↑Conv.	0.015	0.91	9.59E-07	0.6
TGFBR3		0.001	0.58	0.0007	0.47

^*Results are shown only for SNPs with a nominal p-value <0.05

^†pvalues <0.0029 remained significant after Bonferroni correction

^‡All p values < q values of FDR

^§only 19 pairs were available for analysis

Figure 4.

(A-D) RT-PCR analysis of transcript abundance in whole blood from KD patients during the acute and convalescent phase and association between genotype and gene expression. Relative transcript abundance levels were normalized by TAF1B for ACVRL1 (A), SMAD1 (B), SMAD3 (C) and TGFBR3 (D). Solid line: subjects with normal coronary arteries, Dashed line: subjects with coronary artery aneurysms. p value by paired t-test. (E-G) Association between acute and convalescent gene expression levels and ACVRL1 rs11169953 (risk allele= G, GG: n=7, GA: n=5, AA: n=2) (E), SMAD1 rs6537355 (risk allele=A, AA: n=9, AG: n=5, no subject had GG genotype). (F), SMAD3 rs4776338 (risk allele= G, AA: n=6, AG: n=5, GG: n=3) (G). p value by Kruskal-Wallis test (E, G) and Mann Whitney test (F).

Figure 5.

(A) Plasma levels of TGF-β2 in KD patients during the acute and convalescent phase. Solid line: subjects with normal coronary arteries, Dashed line: subjects with coronary artery aneurysms. p value by paired t-test. (B) Association between acute and convalescent plasma levels of TGF-β2 and rs12029576 in TGF-β2 (risk allele= C, AA: n=7, AC: n=6, no subject had CC genotype). p value by Mann Whitney test.

Discussion

Several lines of evidence presented here suggest that the TGF-β signaling pathway plays an important role in KD pathogenesis and that genetic variation in three genes in the pathway (TGFB2, TGFBR2 and SMAD3) influence KD susceptibility, CAA formation, AoR dilatation and IVIG treatment response (Figure 3, Table 2, and Supplemental Tables 5-9). This is the first genetic analysis to find an association between coronary artery and aortic root Z-scores and genotype. Genetic variants associated with KD phenotypes clustered in discrete regions based on analyses of both single SNPs and haplotypes. On the basis of the analysis of the Venter genome, individuals of European descent are likely to carry 200 to 500 non-synonymous, rare (minor allele frequency <5%) and/or novel variants that affect protein function.^{27, 28} Sequencing the implicated regions of the three genes, TGFB2, TGFBR2 and SMAD3, may identify novel or rare genetic variants associated with KD pathogenesis. The significance of genetic variation in this pathway was further highlighted by the finding of dynamic changes in transcript abundance and plasma protein levels during the course of the illness. An analysis of the dynamic change in transcript or protein levels as a function of genotype failed to show an association, although the sample size was very small. Future studies with a bigger sample size will be needed to more definitively address the affect of these polymorphisms on gene expression.

Each of the three genes identified as important by our genetic studies has a plausible role in KD pathogenesis. The first gene highlighted by our genetic analyses, TGFB2, encodes one of the three isoforms of TGF-β and is expressed in both a tissue-specific and a developmentally regulated fashion. Only TGFβ2 is expressed and required for endocardial cushion cell transformation in the mouse²⁹ and Tgfb2 knockout mice suffer from a variety of cardiovascular anomalies.^30-32 Plasma levels of TGF-β2 in our KD subjects were low in the acute compared to the convalescent phase. Similar findings have been reported for TGF-β1 plasma levels and transcript abundance in acute KD but were not measured in this study.^33-35

The second gene highlighted by our study was TGFBR2, which was associated with KD susceptibility, CAA formation, AoR dilatation, and response to IVIG treatment. Genetic variation in this gene is linked to several aneurysm syndromes including Loey-Dietz and abdominal aortic aneurysm.^{36, 37} Immunohistochemistry of histologically normal and aneurysmal coronary arteries 3-12 years after disease onset demonstrated TGF-β1, TGFBR1 and TGFBR2 in intimal or medial smooth muscle cells.^{38, 39} This was not seen in age-matched autopsy tissues from children dying of other conditions, suggesting the importance of this pathway in arterial remodeling late after KD.

The third gene, SMAD3, has an essential role in down-regulating T-cells and increasing expression of FoxP3, an essential step in the differentiation of regulatory T cells.⁴⁰ Imbalance of pro-inflammatory Th17 and regulatory T-cells has been reported in acute coronary syndrome.⁴¹ Recent data from our laboratory documented the emergence of peripherally induced regulatory T-cells in acute KD subjects, suggesting that this pathway may be important in disease recovery.¹⁴ SMAD3 also has an important role in cardiovascular remodeling and fibrosis after injury, both active processes in the recovery phase from KD.⁴²

We recognize several limitations to our study. First, we only genotyped tagging SNPs with a minor allele frequency of at least 10%. Thus, other rare, but important, genetic variants would not have been detected by our methods. Secondly, although we have identified genetic variants that influence disease susceptibility and outcome, none of these intronic variants have a known function. The clustering of significant SNPs in introns suggests the possibility of regulatory elements that might be encoded for by these regions. MicroRNA coding sequences often map to intronic or intergenic regions near the genes that they regulate. miR141 is a negative regulator of TGFB2 expression.⁴³ Using the miRBase (http://www.mirbase.org/) (Release 14: Sept 2009), we were unable to locate candidate microRNA coding sequences in these three genes. Although this is the first analysis to find an association between coronary artery and aortic root Z-scores and genotype, the number of subjects with CAA and AoR dilatation was limited and our findings will need to be validated in independent cohorts. Our cohorts may be biased toward cases with CAA because of sample availability (CAA 73/362 (20%), and 51/186 (27%) in cohort 3 and 4). This may have affected our analysis of KD susceptibility. This issue can only be clarified by testing our observations in larger cohorts of subjects without CAA. KD is highly prevalent in Asian populations and our studies only address genetic variation in subjects of European descent. Future studies must also address the mechanism by which these variants exert their effect. Studies of T-cell activation and differentiation of regulatory T-cells in cultured peripheral blood mononuclear cells from individuals of known genotype will contribute to our understanding of how modulation of the TGF-β pathway impacts KD pathogenesis.

Our studies support the importance of the TGF-β signaling pathway in KD pathogenesis.

Recently, Onouchi et al in collaboration with our group discovered a functional polymorphism in inositol 1,4,5-triphosphate 3-kinase C that increases calcineurin/NFAT signaling and leads to increased levels of IL2 and T-cell activation in KD subjects.¹² Genetic variation in the TGF-β pathway may lead to an imbalance of pro-inflammatory and regulatory T-cells by affecting Foxp3 expression mediated through Smad3 and NFAT.⁴⁰ The TGF-β pathway is also important in inflammation and tissue remodeling mediated by endothelial cells, fibroblasts, and smooth muscle cells. Investigation of the association between the calcineurin/NFAT and TGF-β pathways in T-cells and immunohistochemical study of tissue will further our understanding these pathways in patients with KD. Both of these pathways have important potential for pharmacologic intervention with the availability of calcineurin inhibitors (eg. cyclosporine and tacrolimus) and modulators of the TGF-β signaling pathway (eg. angiotensin receptor blockers). Therefore, further study of these pathways may lead to important new therapeutic intervention strategies for KD patients.