99 - CELL-FREE RNA SIGNATURES OF KAWASAKI DISEASE SUBGROUPS

BACKGROUND

KD is heterogeneous in terms of manifestations and patient outcomes. Recent work has uncovered four distinct subgroups of KD using clinical features and clinical laboratory measurements (PMID37598693). The four subgroups are characterized by KD patients with elevated liver involvement (subgroup 1: liver), high band neutrophil counts (subgroup 2: band), high rates of cervical lymphadenopathy (subgroup 3: node), and young age (subgroup 4: young).

Plasma cell-free RNA (cfRNA) is released from live cells via active excretion or from dead/dying cells resident in solid tissues or circulation. Thus, cfRNA informs about cell and tissue damage and immune dynamics. The cell type of origin (CTO) fractions of cfRNA can be estimated using bulk RNA-seq deconvolution algorithms with scRNA-seq atlases as a reference. Previously, we have shown that cfRNA is informative as a biomarker of disease and reflects organ/tissue damage in inflammatory conditions (PMID37279751, PMID36711999) . We characterized the cfRNA signatures of KD subgroups and explored de novo clustering of KD samples using cfRNA sequencing data.

METHODS

We sequenced plasma cfRNA from 98 patients classified into KD subgroups as previously defined (liver, n=25; band, n=25; node, n=25; young, n=23) (PMID37598693). We performed differential abundance analysis using DESeq2 for biomarker discovery and gene pathway analysis. Sample CTO estimates were determined via deconvolution using BayesPrism with the Tabula Sapiens scRNA-seq atlas as a reference (PMID35469013, PMID35549404).

RESULTS

Pairwise comparisons of cfRNA profiles between sample groups identified differentially abundant transcripts (DATs) in each comparison, with the most DATs present between the node and young subgroups. Furthermore, we observed significant differences in the CTO profiles between subgroups, including unique patterns that matched clinical manifestations. For example, we observed elevated hepatocyte derived cfRNA in the liver subgroup and elevated plasmablast derived cfRNA in the node subgroup.

Next, we performed de novo clustering of the KD samples using only the cfRNA data. We clustered samples using hierarchical, correlation-based clustering. The resulting clusters were unique in terms of cell types of origin and transcript abundance.

CONCLUSION

Plasma cfRNA analysis supports the hypothesis of different KD subgroups. Furthermore, we showed that cfRNA can be used to independently subgroup samples, potentially providing additional information on patient heterogeneity. Future work will further characterize the cfRNA-based subgroups and combine cfRNA, clinical, and laboratory measurements for new subgroup analysis.