Single-cell transcriptomics reveal heat shock protein dysregulation in severe SARS-CoV-2–associated pediatric encephalopathy

Abstract

Severe acute encephalopathy/encephalitis (AE) associated with SARS-CoV-2 has been increasingly reported since the emergence of the Omicron variant. Several pediatric cases have shown the development of acute fulminant cerebral edema (AFCE) or hemorrhagic shock encephalopathy syndrome (HSES), which are linked to high morbidity and mortality. However, the underlying pathogenic mechanisms remain unclear. We performed single-cell RNA sequencing of peripheral blood mononuclear cells from a pediatric patient with SARS-CoV-2-associated AE presenting with AFCE/HSES and compared the data with those from two patients with mild AE, one patient with febrile seizures due to non-SARS-CoV-2 pathogens, and publicly available pediatric COVID-19 datasets without neurological complications. During the acute phase, we observed a prominent expansion of B-cell populations, including distinct activated B-cell clusters. Cell–cell communication analysis identified macrophage migration inhibitory factor signaling, although it was not specific to SARS-CoV-2-associated AE. Notably, heat shock protein genes, particularly HSPA1A and HSPB1, were selectively upregulated across multiple immune cell types only in severe SARS-CoV-2-associated AE. Enzyme-linked immunosorbent assay confirmed significantly elevated plasma and serum protein levels of HSPA1A and HSPB1 during the acute phase. These findings highlight HSPA1A and HSPB1 as potential biomarkers of severe SARS-CoV-2-associated AE and suggest a pathogenic possible role for stress-response pathways.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-41827-2.

Introduction

Acute encephalitis and encephalopathy (AEs) in children are clinical syndromes defined by altered mental status lasting more than 24 h, often accompanied by seizures¹. The majority of pediatric AEs are associated with viral infections. Although encephalitis involves direct inflammation of the brain parenchyma and encephalopathy primarily results from cerebral edema, distinguishing between them is challenging when direct viral invasion is not evident¹. Viruses commonly linked to virus-associated AEs include influenza virus^2,3, human herpesvirus 6B (HHV-6B)^4,5, rotavirus⁶, and respiratory syncytial virus⁷. Clinical presentations range from relatively mild subtypes with favorable outcomes—such as mild encephalopathy with reversible splenial lesions (MERS)—to severe, life-threatening forms such as acute necrotizing encephalopathy (ANE). Since the emergence of the SARS-CoV-2 Omicron variant, reports of pediatric COVID-19-associated AEs have increased substantially⁸. A recent nationwide survey in Japan identified 103 cases of pediatric SARS-CoV-2-associated AE between January 2020 and November 2022, with an estimated incidence of approximately 1–2 per 100,000 pediatric COVID-19 cases⁹. Similar to other virus-associated AEs, these cases encompass various clinical subtypes. Notably, acute fulminant cerebral edema (AFCE) and hemorrhagic shock encephalopathy syndrome (HSES)—both associated with poor neurological outcomes—represented 13.6% of all reported SARS-CoV-2-associated AEs, a higher proportion than typically observed for other viral etiologies⁹.

Severe AE subtypes, including ANE, AFCE, and HSES, possibly involve cytokine storm-driven hyperinflammation¹⁰ and blood–brain barrier (BBB) disruption¹¹; however, the precise pathogenesis remains poorly defined. Early differentiation of severe cases is particularly difficult, and currently no definitive, evidence-based treatments exist. Therefore, identifying robust genetic or protein biomarkers for early risk assessment, along with elucidating the underlying mechanisms, is critical for improving outcomes.

Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool to dissect cellular heterogeneity and molecular pathways in infectious and inflammatory diseases. Although this technology has significantly advanced our understanding of COVID-19^12,13 and multisystem inflammatory syndrome in children (MIS-C)¹⁴, scRNA-seq studies investigating SARS-CoV-2-associated AEs are currently lacking.

Here, we report an infantile case of severe SARS-CoV-2-associated AE presenting with AFCE/HSES. We performed scRNA-seq on peripheral blood mononuclear cells (PBMCs) collected during the acute phase and conducted comparative transcriptomic analyses with samples obtained from the same patient during the later phase of illness (hereafter referred to as the “post phase”), as well as from patients with mild AE and febrile seizures associated with other viral infections.

Methods

Patients and sample collection

Three pediatric patients with AE (patients E1–E3), one pediatric patient with febrile seizure (patient F1), and two healthy adult donors (H1 and H2) were enrolled (Table 1). AE was diagnosed based on acute onset of impaired consciousness lasting more than 24 h in the absence of drug toxicity or metabolic disorders¹⁰. Pathogenic viruses identified included influenza A virus (E1), SARS-CoV-2 (E2), Epstein–Barr virus (EBV; E3), and HHV-6 (F1). PBMC samples were collected during the acute phase (1–7 days from onset) and the post phase (≥ 7–111 days from onset).

Table 1.

Patient characteristics.

Patient ID	Age	Sex	Diagnosis	Pathogen	Clinical classification	Interval onset to 1st sample (days)	Interval onset to 2nd sample (days)	Treatment	Outcome
E1	9 y	F	Encephalopathy	Influenza A (H3)	Mild	2	7	Steroid pulse	Recovered
E2	10 mo	F	Encephalopathy	SARS-CoV-2	AFCE/HSES	1	7	Induced normothermia	Deceased
E3	7 y	F	Encephalopathy	EBV	Mild	7	111	Steroid pulse	Recovered
F1	1 y	M	Febrile seizure	HHV-6	–	1	18	–	Recovered
H1	31 y	M	Healthy	–	–	–	–	–	–
H2	38 y	F	Healthy	–	–	–	–	–	–

AFCE, acute fulminant cerebral edema; EBV, Epstein–Barr virus; F, female; HHV-6, human herpesvirus 6; HSES, hemorrhagic shock encephalopathy syndrome; M, male; mo, months; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; y, years; —, not applicable.

Single-cell RNA library preparation and sequencing

ScRNA-seq library preparation and analysis were conducted as previously described¹⁵. PBMCs were isolated using Ficoll–Paque density-gradient centrifugation, resuspended in cell cryopreservation solution, and stored at − 80 °C until use. After thawing, cells were washed in phosphate-buffered saline containing 1% bovine serum albumin, and dead cells were removed with the Dead Cell Removal Kit (Miltenyi Biotec, Bergisch Gladbach, Germany). Select samples were tagged with TotalSeq™-B antibodies (BioLegend, San Diego, CA, USA) for sample multiplexing. All processed samples demonstrated ≥ 75% viability.

Single-cell suspensions were barcoded using the Chromium platform (10x Genomics, Pleasanton, CA, USA). Libraries were generated with the Chromium Single Cell 3′ Reagent Kits v3.1 (Single/Dual Index) with or without Feature Barcode technology for cell-surface protein detection, following manufacturer instructions. Sequencing was performed on an Illumina HiSeq X Ten or NovaSeq X Plus system (Macrogen Japan Corp., Tokyo, Japan).

Single-cell RNA-seq processing and analysis

Sequencing reads were aligned to the human GRCh38 reference genome using Cell Ranger v5.0.1 (10x Genomics) with the cellranger count pipeline. Quality filtering was performed using Seurat v5.2.1 (Satija Lab, New York Genome Center)¹⁶, excluding cells with > 25% mitochondrial RNA content or < 200 or > 5,000 unique detected transcripts. Sample demultiplexing for hashed datasets was conducted using the HTODemux function. Following quality control, 1,158–10,310 cells were retained per sample, yielding a total of 54,793 cells for analysis (Table 2). Data integration was performed using reciprocal principal component analysis (RPCA) as recommended in the Seurat workflow. The top 30 principal components (PCs) were used for clustering and visualization via uniform manifold approximation and projection (UMAP).

Table 2.

Single-cell RNA sequencing datasets analyzed in this study.

Patient ID	Age	Sex	Diagnosis	Pathogen	Phase	Cells captured	Source
E1	9 y	F	Encephalopathy	Influenza A (H3)	Acute	7,942	This study
					Post	8,884
E2	10 mo	F	Encephalopathy	SARS-CoV-2	Acute	7,415
					Post	7,727
E3	7 y	F	Encephalopathy	EBV	Acute	1,266
					Post	1,158
F1	1 y	M	Febrile seizure	HHV-6	Acute	4,500
					Post	10,310
H1	31 y	M	Healthy	–	–	3,385
H2	38 y	F	Healthy	–	–	2,206
P5	3 mo	M	COVID-19	SARS-CoV-2	Acute	7,509	GSE167029
P8	3 mo	M	COVID-19	SARS-CoV-2	Acute	6,382
P12	1 y	M	COVID-19	SARS-CoV-2	Acute	5,175
P13	2 mo	F	COVID-19	SARS-CoV-2	Acute	6,667
C27	7 y	M	Healthy	–	–	8,405
C28	5 y	M	Healthy	–	–	12,762
C32	3 y	F	Healthy	–	–	8,247

EBV, Epstein-Barr virus; F, female; HHV-6, human herpesvirus 6; M, male; mo, months; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; y, years; —, not applicable.

External scRNA-seq dataset

scRNA-seq PBMC data from pediatric patients with COVID-19 without neurological complication and from healthy pediatric controls were obtained from Gene Expression Omnibus (GEO), accession number GSE167029¹⁴. Raw feature-count matrices from four COVID-19 patients (P5, P8, P12, P13) and three healthy controls (C27, C28, C32) were filtered and integrated within the study dataset for comparative analysis (Table 2). Additionally, we extracted PBMC scRNA-seq data of six patients with MIS‑C under 10 years of age (P17, P20, P31, P33, P39, and P43) included in the GSE167026 dataset (Table S1) to perform a comparative analysis, as MIS-C represents a SARS‑CoV‑2-related, post‑infectious autoimmune inflammatory condition in the pediatric population.

Gene ontology enrichment and cell–cell interaction analyses

Differentially expressed genes (DEGs) were identified using the FindMarkers function in Seurat. Statistical significance was determined using the Wilcoxon rank-sum test. Genes with adjusted p < 0.01 and log2 fold-change (log2FC) > 1.0 were considered significant and subjected to gene ontology enrichment analysis using Metascape (http://metascape.org)¹⁷. Cell–cell communication networks were inferred using CellChat v2.1.2¹⁸ based on curated ligand–receptor interaction datasets.

Enzyme-linked immunosorbent assay (ELISA)

Plasma or serum samples obtained from patients during the acute phase of AE were stored at − 30 °C until analysis. Plasma from three pediatric healthy controls (HC1–HC3) was included as comparator samples. Plasma concentrations of heat shock proteins HSPA1A and HSPB1 were quantified using the Human HSP70 Sandwich ELISA Kit (Proteintech, Rosemont, IL, USA) and High-Sensitivity ELISA Kit for Hsp27 (Cloud-Clone Corp., Katy, TX, USA), respectively. All assays were performed according to the manufacturers’ protocols.

Results

Patient characteristics

Clinical characteristics of all enrolled patients are summarized in Table 1 and Table S2. Patient E2 presented with severe SARS-CoV-2–associated AE and was clinically diagnosed with AFCE/HSES. The patient developed status epilepticus and circulatory shock requiring intensive care management. Despite aggressive treatment, diffuse cerebral edema rapidly progressed to brain herniation, and the patient ultimately died. Acute-phase PBMCs were collected on day 1 of disease onset, and post-phase samples were collected on day 7. Patient E1 presented with AE associated with influenza A (H3) infection and received high-dose steroid pulse therapy, recovering fully without neurological sequelae. Patient E3 developed AE following EBV infection, was treated with steroid pulse therapy, and also recovered without sequelae. Patient F1, who served as a non-AE disease control, had a typical simple febrile seizure triggered by HHV-6 infection and exhibited full recovery.

Comparison of single-cell transcriptomes among AE, febrile seizure, pediatric COVID-19, and healthy controls

We integrated a total of 109,940 PBMC transcriptomes by combining our dataset with publicly available scRNA-seq data (Table 2) and performed unsupervised clustering, yielding 19 distinct cell clusters (Fig. S1a). Cell-type identities were assigned based on canonical marker gene expression and annotation using the Azimuth reference workflow¹⁹(Fig. S1b, S1c). Comparison of cellular composition across samples (Fig. 1a, b, S1d) demonstrated a marked expansion of B-cell populations during the acute phase of the severe SARS-CoV-2–associated AE case (patient E2). B cells accounted for 48.31% of PBMCs at onset (acute phase) and 5.85% during the later phase of illness (post phase). Notably, two B-cell subsets, corresponding to clusters 12 and 14, were prominently increased during the acute phase (cluster 12: 5.56%, cluster 14: 7.58%), whereas these clusters were scarce or absent in other samples.

Fig. 1.

Expansion of B-cell clusters and differentially expressed genes in the acute phase of severe SARS-CoV-2–associated AE (patient E2). (a) Cluster distribution across sample types. (b) UMAP visualization of PBMCs from acute and post-phase samples of patient E2. (c) Volcano plot showing differentially expressed genes (DEGs) in B-cell clusters between E2_acute and E2_post. HSPA1A and HSPB1 genes are highlighted and annotated for clarity. (d) Dot plot of upregulated genes in each B-cell cluster. (e) Gene ontology enrichment analysis of highly expressed genes in B-cell cluster 14, colored by p-value. AE, acute encephalopathy/encephalitis; cDC, classical dendritic cell; DEGs, differentially expressed genes; EBV, Epstein–Barr virus; fluA, influenza A; FS, febrile seizure; HC, healthy control; HSPC, hematopoietic stem and progenitor cell; Mono, monocyte; pDC, plasmacytoid dendritic cell; PBMC, peripheral blood mononuclear cell; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; UMAP, uniform manifold approximation and projection.

Activated B-cell cluster (cluster 14) emerged during the acute phase of severe SARS-CoV-2–associated AE

Gene expression profiles of B-cell clusters in the severe SARS-CoV-2–associated AE case (E2_acute) were compared with those in the post phase (E2_post) (Fig. 1c). E2_acute B cells exhibited marked upregulation of type I interferon-responsive genes, including IFI6, ISG15, IFI44L, and IFITM1, along with increased expression of HSP genes such as HSPA1A and HSPB1. When compared with acute-phase samples from patients E1/E3 (mild) and F1 (FS) and public pediatric COVID-19 datasets (GSE167029_COVID-19), E2_acute demonstrated a significantly stronger induction (p < 0.001, average log2FC > 5) of HSPA1A and HSPB1 gene expression (Fig. S2a–c). Pathway enrichment analysis of genes upregulated in the E2_acute B-cell population revealed significant enrichment of the “response to unfolded protein” pathway (GO:0006986; Log10(p) = − 9.93), distinguishing it from other patient groups (Fig. S3). Among the four identified B-cell clusters, transcriptional activation was most pronounced in cluster 14, which was selectively expanded during the acute phase (Fig. 1d). Genes upregulated in cluster 14 were highly enriched for antiviral pathways, including “response to virus” (GO:0009615; Log10(p) = − 41.98) and “response to type I interferon” (GO:0034340; Log10(p) = − 19.70) (Fig. 1e). These data suggest that cluster 14 represents an activated B-cell subset engaged in robust antiviral and cellular stress responses unique to severe SARS-CoV-2–associated AE.

Macrophage migration inhibitory factor (MIF) signaling contributed to recruitment and activation of cluster 14

To investigate signaling pathways potentially regulating cluster 14 dynamics, cell–cell interaction analysis was performed using CellChat in the E2_acute dataset. Interactions involving receptors expressed in cluster 14 were extracted (Fig. 2a). Among all inferred interaction pathways, MIF signaling emerged as the most dominant contributor. The principal sources of MIF were FCGR3A⁺ monocytes (cluster 9) and T-cell subsets, including CD4⁺ T cells (clusters 0 and 6) and CD8⁺ T/γδT/natural killer (NK) cells (clusters 3, 4, and 8) (Fig. 2b). Although MIF expression was also elevated during the febrile seizure acute phase, the most pronounced induction was observed in the E2_acute sample, particularly within the CD4⁺ T-cell and CD8⁺ T/γδT/NK-cell clusters (Fig. 2c, d). These results indicate that heightened MIF signaling may contribute to the selective expansion and activation of cluster 14 B cells in severe SARS-CoV-2–associated AE.

Fig. 2.

Cell-cell interaction analysis identifies MIF signaling as a major pathway in the acute phase of severe SARS-CoV-2–associated AE (patient E2). (a) Heatmap of predicted ligand–receptor interactions involving receptors expressed in B-cell cluster 14 in E2_acute. The colored bar plots on the top represent the total signaling strength of each cluster, obtained by summarizing all signaling pathways displayed in the heatmap. The grey bar plots on the right indicate the total signaling strength of each signaling pathway, calculated by summarizing all clusters shown in the heatmap using CellChat. (b) Heatmap visualization of the MIF signaling network in E2_acute. The colored bar plots on the top represent the receiver contribution strength of each cluster, and the colored bar plots on the right represent the sender contribution strength. (c) Violin plots of MIF gene expression in CD4⁺ T-cell clusters across sample groups. (d) Violin plots of MIF gene expression in CD8⁺ T/γδ T/NK-cell clusters across sample groups. Severe_SARS‑CoV‑2+, FS, mild, and HC correspond to E2_acute, F1_acute, E1/E3_acute, and H1/H2/GSE167029_HC, respectively. AE, acute encephalopathy/encephalitis; FS, febrile seizure; HC, healthy control; MIF, macrophage migration inhibitory factor; Mono, monocyte; NK, natural killer; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

Widespread upregulation of HSPA1A and HSPB1 during the acute phase of severe SARS-CoV-2–associated AE

To further characterize the differential expression of HSP genes, we evaluated their expression across non-B-cell populations (Fig. S4). Notably, elevated HSP gene expression in E2_acute was not restricted to B cells but was observed across multiple immune cell types when compared with all the other groups. In particular, monocytes in E2_acute showed robust upregulation of HSPA1A and HSPB1, as well as additional HSP family members such as HSPH1, HSPD1, and HSP90AA1. Among the commonly induced HSP-related genes, HSPA1A and HSPB1 were most consistently and prominently upregulated across immune subsets. Comparative analysis across sample groups confirmed that HSPA1A and HSPB1 expression was specifically and significantly elevated during the acute phase of severe SARS-CoV-2–associated AE (Fig. 3a–h). In contrast, expression levels were low in acute-phase samples from mild AE and febrile seizures and COVID-19 cases without neurological involvement, as well as in MIS-C cases (Fig. S5), indicating fundamental differences in the underlying pathophysiology between severe SARS-CoV-2-associated AE and MIS‑C, which emerges as a delayed post‑infectious hyperinflammatory syndrome.

Fig. 3.

HSPA1A and HSPB1 are broadly upregulated across immune cell types and elevated in plasma during the acute phase of severe SARS-CoV-2–associated AE (patient E2). Violin plots of HSPA1A gene expression in (a) B cell, (b) monocyte, (c) CD4⁺ T cell, and (d) CD8⁺ T/γδT/NK cell clusters across sample groups; and HSPB1 gene expression in (e) B cell, (f) monocyte, (g) CD4⁺ T cell, and (h) CD8⁺ T/γδT/NK cell clusters. Severe_SARS‑CoV‑2+, FS, mild, and HC correspond to E2_acute, F1_acute, E1/E3_acute, and H1/H2/GSE167029_HC, respectively. (i,j) ELISA quantification of plasma/serum HSPA1A (i) and HSPB1 (j) levels in AE patients (E1–E3) and pediatric healthy controls (HC1–HC3). AE, acute encephalopathy/encephalitis; ELISA, enzyme-linked immunosorbent assay; FS, febrile seizure; HC, healthy control; HSP, heat shock protein; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

To validate these observations at the protein level, we quantified plasma/serum concentrations of HSPA1A and HSPB1 using ELISA on stored samples from patients E1–E3 and from healthy pediatric controls (HC1–3) (Fig. 3i, j). Consistent with the transcriptomic data, HSPA1A and HSPB1 protein levels were markedly elevated in E2_acute, but low in E2_post and in samples from mild AE, and healthy children. Together, these findings indicate that HSPA1A and HSPB1 are selectively upregulated at the both the transcript and protein levels during the acute phase of severe SARS-CoV-2–associated AE, highlighting their potential as biomarkers associated with disease severity.

Discussion

Since the emergence of the SARS-CoV-2 Omicron variant, reports of pediatric AE have risen substantially⁸. However, to the best of our knowledge, this study is the first to apply scRNA-seq to SARS-CoV-2–associated AE in children. Importantly, PBMCs were collected on day 1 of symptom onset, capturing the hyperacute phase and offering valuable insights into early immunopathological events.

Virus-associated AEs can follow various infections, including SARS-CoV-2, and severe subtypes characterized by hyperinflammation—such as ANE, AFCE, and HSES—are associated with poor neurological outcomes⁹. Although cytokine storm and BBB disruption have been proposed as mechanisms^10,11, the molecular drivers remain incompletely understood. Elevated cytokines, including interleukin 6 (IL-6), tumor necrosis factor-α (TNF-α), and MIF, have been reported in severe cases^20–22; however, how these are regulated is unclear.

In our severe case (E2_acute), we observed strong activation of the MIF signaling network, with upregulated MIF expression in monocytes and T/NK cells. MIF stimulates production of proinflammatory cytokines such as IL-6 and TNF-α, promotes immune cell activation, and has been implicated in cytokine storms during severe COVID-19^23,24. MIF can also promote B-cell mobilization²⁵, consistent with the expansion of an activated B-cell population (cluster 14) in E2_acute. However, a similar increase in MIF expression was detected in F1_acute—without the development of AFCE/HSES—indicating that MIF activation alone does not fully explain the severe phenotype.

Distinct from the febrile seizure case, patient E2 exhibited specific and widespread upregulation of HSP genes, including HSPA1A and HSPB1. HSPs are stress-induced chaperones that can protect host cells but may also facilitate viral replication, including SARS-CoV-2^26–28. Increased expression of HSP70 and HSP90 have been associated with severe COVID-19²⁹, aligning with our findings. Beyond chaperone activity, HSPs serve as endogenous danger signals, activating Toll-like receptors (e.g., TLR2. TLR4) and leading to proinflammatory cytokine release^30,31. HSP90 is also a client protein of MIF³², and its stabilization may potentiate MIF-mediated inflammatory responses. Together, these findings support a model in which excessive HSP upregulation amplifies immune activation during SARS-CoV-2–associated AE.

HSPs have also been implicated in neuropathology. HSPA1A is upregulated in severe neonatal hypoxic-ischemic encephalopathy³³, and inhibition of HSP90 mitigates BBB disruption and inflammation in intracerebral hemorrhage models³⁴. These observations suggest that elevated HSPs may contribute to BBB breakdown and neuroinflammation—key pathological features in AFCE and HSES. Consistent with transcriptional findings, plasma levels of HSPA1A and HSPB1 were markedly elevated at disease onset in E2_acute but not in mild AE or healthy controls. Taken together, our results suggest that HSP dysregulation play a mechanistic role in severe SARS-CoV-2–associated AE and that HSPA1A and HSPB1 represent promising early biomarkers for identifying high-risk cases.

This study has some limitations. First, the scRNA-seq analysis was performed on a single case of severe SARS-CoV-2–associated AE, underscoring the need for validation in larger patient cohorts. It remains unclear whether the observed HSP upregulation is specific to SARS-CoV-2–associated AE or represents a generalized feature of severe, cytokine storm–driven AEs caused by other viral pathogens. Comparative studies, including febrile seizures or mild neurological cases associated with SARS-CoV-2, are needed to clarify disease specificity. Second, healthy pediatric samples were not available in our own cohort, and in‑house controls consisted only of adults. Because immune profiles differ by age, we incorporated additional samples from publicly available datasets that included healthy pediatric PBMC data to enable age‑appropriate comparisons. Nevertheless, age heterogeneity across datasets remains a potential confounder. Third, although assessing circulating cytokine levels—particularly MIF—would provide important complementary insights, limited sample volumes, especially from infants and young children, constrained the scope of protein analyses. Finally, the mechanistic role of HSPs in AE pathogenesis remains to be determined. Experimental studies using relevant in vitro or in vivo models will be essential to elucidate how HSP dysregulation may contribute to immune activation, BBB disruption, and neurological injury.

Conclusions

In this study, we performed single-cell transcriptomic profiling of PBMCs from pediatric patients with virus-associated AE, including a hyperacute sample from a child with severe SARS-CoV-2–associated AE. We identified marked upregulation of HSP genes, particularly HSPA1A and HSPB1, at both transcript and protein levels in the severe case, distinguishing it from mild AE and healthy controls. These findings suggest that HSP dysregulation may contribute to the immunopathogenesis of severe SARS-CoV-2–associated AE and that HSPA1A and HSPB1 may serve as candidate early biomarkers to identify high-risk patients. Our findings demonstrate the utility of scRNA-seq in dissecting immune alterations in pediatric AE and highlight the need for further mechanistic and clinical validation studies.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: T.S. and J.K.; methodology: T.S., Y.S., H.K., and J.K.; investigation: T.S., Y.S., Y.F., and J.K.; data curation: T.S.; formal analysis: T.S.; software: T.S. and Y.S.; visualization: T.S.; supervision: J.K. and Y.T.; resources: M.S., Y.K., and T.Y.; writing—original draft: T.S.; writing—review and editing: T.S., Y.S., Y.F., K.I., M.Y., Y.K., Y.T., H.K., and J.K.; project administration: J.K. and Y.T.; funding acquisition: T.S., Y.S., Y.T., and J.K.

Funding

This work was supported in part by the Japan Society for the Promotion of Science (JSPS) KAKENHI (Grant Numbers 21K20756 to T.S. and 21K077478 to J.K.); the Mitsubishi Foundation Research Grant 2020 to J.K.; the Kawano Pediatric Foundation to J.K.; and the Japan Agency for Medical Research and Development (AMED) (Grant Number JP256f0137010 to Y.S. and Y.T.).

Data availability

The scRNA-seq datasets generated in this study are available in the DNA Data Bank of Japan Sequence Read Archive (DDBJ SRA, https://www.ddbj.nig.ac.jp/dra/index.html) under accession numbers DRR500111 (pooled E3_acute/post samples) and DRR787061–DRR787066 (E1/E2/F1_acute/post samples).

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval

The study protocol was approved by the Institutional Review Board of Nagoya University Hospital (IRB No. 2020 − 0645). All methods were performed in accordance with the Declaration of Helsinki and the Ethical Guidelines for Medical and Biological Research Involving Human Subjects in Japan.

Consent to participate

Informed consent was obtained from all patients or their legal guardians.

Consent for publication

Written informed consent for publication of the data presented in this study was obtained from all patients or their legal guardians.