Multisystem inflammatory syndrome in children (MIS-C), emerging during the COVID-19 pandemic, presents with similarities to Kawasaki disease (KD). Both are rare but severe pediatric conditions characterized by inflammation. MIS-C, often appearing 4-6 weeks post SARS-CoV-2 exposure, involves fever, rash, conjunctival injection, gastrointestinal issues, shock, and elevated inflammation markers. Myocardial dysfunction and coronary arterial dilation can mimic those seen in KD, an acute inflammatory disorder, the leading cause of pediatric acquired heart disease in developed nations. The triggers for KD remain unclear, with evidence pointing to various infectious or environmental stimuli. Distinguishing MIS-C from KD and other conditions poses a challenge, and early diagnostic and prognostic markers are needed to guide treatment decisions. Previous research has highlighted differences in the inflammatory responses between MIS-C and KD concerning T cell subsets and autoimmunity. However, limitations such as medication effects on MIS-C subjects, lack of contemporaneous healthy controls, and absence of concurrent MIS-C and KD studies in some analyses have hindered a comprehensive understanding. This study utilized viral pandemic (ViP) signatures—a 166-gene signature and a 20-gene severe (s)ViP subset—previously developed from analyzing diverse viral pandemic datasets. These signatures, trained on past pandemic data, were applied to investigate both shared and unique features in MIS-C and KD, providing a framework to understand the host immune response in these conditions and their relationship to COVID-19.

Several studies have attempted to characterize the immunopathogenesis of MIS-C, often focusing on the differences from KD. Gruber et al. and Consiglio et al. showed variations in the inflammatory response between MIS-C and KD concerning T cell subsets. Vella et al. and Ramaswamy et al. further supported these findings, noting skewed memory T cell repertoires and autoimmunity in severe MIS-C. Carter et al. identified activated CD4⁺CCR7⁺ T cells and γδ T cell subsets in MIS-C, absent in KD studies. While these studies provide insight, limitations include medication influence on samples, lack of healthy controls, absence of concurrent MIS-C and KD comparisons, and limited validation. The authors’ previous work established a 166-gene ViP signature and a 20-gene severe-ViP subset that predicted COVID-19 severity. These signatures were shown to capture an invariant host response across viral pandemics, making them suitable for comparing the immune responses of MIS-C and KD without requiring further training on these specific diseases.

This study used a multi-omics approach integrating gene expression data from whole blood and serum cytokine arrays. The researchers utilized the previously established ViP and sViP signatures to analyze publicly available KD datasets and newly recruited KD and MIS-C cohorts. The study included age-matched healthy pediatric controls and febrile controls (with non-KD/MIS-C infections). All samples were collected before treatment initiation. The ViP and sViP signatures were computed by normalizing gene expression values using a modified Z-score approach and the StepMiner algorithm to obtain Boolean values. The Boolean Equivalent Correlated Clusters (BECC) analysis was used to identify functionally related gene sets. Several publicly available microarray and RNA-Seq datasets were downloaded from the NCBI GEO database. Data normalization was performed using RMA for Affymetrix platforms and TPM for RNA-Seq platforms. The StepMiner algorithm identified step-wise transitions in time-series data, converting gene expression levels into Boolean values (high/low). Boolean analysis determined the relationships between gene expression levels. Heatmaps, hierarchical agglomerative clustering, PCA, and differential expression analysis using DESeq2 were performed. Statistical analysis included t-tests, ROC-AUC calculations, and gene set enrichment analysis (GSEA). For serum cytokine analysis, the V-PLEX Custom Human Biomarkers platform from MSD was used. Immunohistochemistry was conducted on heart tissue from a fatal KD case. The study adhered to ethical guidelines, with IRB approval and informed consent.

The ViP and sViP signatures were upregulated in KD samples compared to healthy controls, showing a strong association with acute KD and reduced expression during convalescence. The sViP signature consistently outperformed the ViP signature in classification accuracy. The signatures also tracked treatment response to IVIG and IVIG/IVMP combination therapy. Analysis in two independent KD cohorts confirmed the induction of ViP signatures in acute KD and their downregulation during convalescence, with the sViP signature showing superior classification. The sViP signature also correlated with the development of giant coronary artery aneurysms (CAA) in KD. In MIS-C patients, both ViP and sViP signatures were significantly upregulated compared to acute KD and healthy controls. The sViP signature effectively classified severe MIS-C cases (with myocardial dysfunction) from mild-moderate cases. The 13-transcript KD-specific signature failed to distinguish between MIS-C and KD. Cytokine analysis showed that TNFα, IFNγ, IL10, IL8, and IL1β were elevated to a greater extent in MIS-C than KD, suggesting potential therapeutic targets. MIS-C patients showed more pronounced thrombocytopenia and eosinopenia than KD patients. Thrombocytopenia and eosinopenia were negatively correlated with ViP, sViP, and IL15/IL15RA scores in MIS-C. The sViP signature correlated with reduced left ventricular ejection fraction (LVEF) in MIS-C but not with ViP or IL15/IL15RA scores. Immunohistochemistry on a fatal KD case revealed IL15 and IL15RA expression in cardiomyocytes and coronary arterioles. The ViP signatures were induced in various infectious diseases but not in immunosuppressed states or all autoimmune diseases, further supporting their role in identifying conditions with infectious triggers.

This study demonstrates that MIS-C and KD share a fundamental IL15/IL15RA-centric cytokine storm, indicating common proximal immunopathogenesis pathways despite differing clinical manifestations. The ViP signatures effectively identify a spectrum of disease severity, placing MIS-C further along this spectrum than KD. The study highlights key distinguishing features of MIS-C, including more pronounced thrombocytopenia and eosinopenia, and a negative correlation of these parameters with IL15 levels. The findings suggest a potential for repurposing FDA-approved therapeutics targeting TNFα and IL1β pathways for MIS-C treatment. The study also reveals that the IL15/IL15RA pathway may have different target end organs in these conditions, with implications for understanding cardiac involvement in both diseases. Although both conditions share some clinical symptoms and the underlying immune response, the study identifies specific markers that differ in intensity between MIS-C and KD, providing valuable insights for diagnosis and treatment.

This study reveals significant similarities and differences in the host immune response between MIS-C and KD using AI-guided gene signatures. Both diseases share an IL15/IL15RA-centric cytokine storm but differ in severity and key clinical/laboratory parameters. The sViP signature proves useful in identifying severe MIS-C. The findings suggest potential therapeutic strategies for MIS-C, and highlight the need for further research into the distinct immunopathogenesis of these diseases. Future studies with larger MIS-C cohorts and access to cardiac tissues are crucial for further validation and insights.

The study's main limitation is the relatively small sample size of the MIS-C cohort (n=12) and limited available public datasets for independent validation. The analysis of cardiac tissue was restricted to one fatal KD case, limiting the generalizability of those findings to all KD patients. Further research is needed to confirm findings with larger and more diverse cohorts and explore the implications for individual patient care.