We obtained the single-cell RNA sequencing (ScRNA-seq) data from the GEO public database (https://www.ncbi.nlm.nih.gov/geo/) with accession number GSE168732 (27), which includes PBMC from six KD patients sampled at two time points (acute pre-treatment and 24 h post-IVIG treatment), together with three healthy donors. Thus, the dataset comprises a total of 15 samples, and each sample was processed as an individual Seurat object. The extracted data were in the form of a gene-barcode matrix of (unique molecular identifier) UMI counts. Seurat (v5.0.1) was used (28, 29) for quality control, normalisation, dimensional reduction, batch effect removal, clustering, and visualization. Each sample was converted into a Seurat object, and cells from each sample having counts between 2000 and 60000 having less than 5% mitochondrial genes were selected for further downstream analysis. After filtering low-quality cells, they were log-normalised using the NormalizeData function of Seurat, and the top 2000 variable features were selected for dimensional reduction. All the 15 Seurat objects were integrated using the IntegrateData function in the Seurat package to eliminate any batch effects present in the samples. Canonical correlation analysis (CCA) was used to identify the anchors, and these anchors were used to correct the technical difference between the samples. Later, this integrated Seurat object containing 86074 cells from healthy controls, acute untreated KD patients, and IVIG-treated patients was used for further analysis. After scaling the integrated matrix, the uniform manifold approximation and projection (UMAP) was performed using the top 30 dimensions that emerged from the principal component analysis (PCA). Meanwhile, cell clusters were found using the nearest neighbour graph-based clustering technique on the PCA-reduced data. To obtain the specific innate immune cell subset within PBMC, the resolution was adjusted to 1 and expression of canonical marker genes was then used to annotate cell clusters. The SingleR (v2.8.0) (30) was also used to annotate rare cell populations with the help of the Monnocle immune database. Standard single-cell QC plots (distribution of UMIs, mitochondrial gene percentage, clustering UMAP/t-SNE) and dot plots showcasing marker genes used for annotating innate immune cell subsets were presented in Supplementary Figures S1–S5.

A list of core autophagy genes and selective autophagy was curated based on previously published literature (31). To compare the average expression levels of core autophagy and selective autophagy genes across different subsets (e.g., cell types or clusters) within the control, diseased, and IVIG-treated conditions, the average expression data were visualised using the pheatmap (v1.0.12) to generate heatmaps (32). These heatmaps provided a visual comparison of the expression levels of core autophagy genes across different conditions and subsets. The pheatmap function was configured with clustering to highlight similarity in the patterns of gene expression. The expression levels of core autophagy genes were further visualised using box plots to depict the distribution of gene expression across different conditions and immune cell subsets. Significant differences, as determined by the two-sided t-tests, were annotated on the plots.

For each cell type, differential expression between treatment groups was assessed using the MAST (Model-based Analysis of Single-cell Transcriptomics) algorithm (v1.32.0) implemented in Seurat. This method accounts for the bimodal distribution and sparsity typical of single-cell data by incorporating cellular detection rate as a covariate in the linear hurdle model. Genes expressed in fewer than 10% of cells were excluded to reduce noise. Differentially expressed genes (DEGs) were defined using thresholds of |log2 fold change| ≥ 0.25 and adjusted p value < 0.05 after Benjamini–Hochberg correction. The resulting DEGs were used for examining the expression of autophagy-related genes across different treatment conditions in innate immune cell subsets.

To evaluate coordinated immune remodeling, Spearman correlation analysis was performed on per-sample cell-type proportions. Correlations were calculated separately for healthy control, untreated KD, and IVIG-treated groups. Correlation matrices were visualized as hierarchical clustering heatmaps using the pheatmap R package. Positive and negative correlations were interpreted as coordinated expansion or reciprocal changes between innate immune cell subsets.

Cell–cell communication analysis was performed using the CellChat R package (v2.2.0) (33) on the integrated single-cell RNA-sequencing dataset of innate immune cells from healthy controls, untreated KD patients, and IVIG-treated patients. Cells were annotated into major innate immune populations, including classical, intermediate, and non-classical monocytes, natural killer (NK) cells, low-density neutrophils, and Vδ2 γδ T cells. For each condition, CellChat objects were constructed independently using the default human ligand–receptor interaction database. Communication probabilities were inferred based on the expression of known ligand–receptor pairs, and statistically significant interactions were identified using permutation testing. Interaction strength was summarized at both the cell–cell level and the pathway level.

To compare global communication patterns across conditions, we visualized (i) heatmaps representing the number of interactions between sender and receiver cell types, (ii) network diagrams illustrating the directionality and strength of intercellular signalling, and (iii) bubble plots highlighting condition-specific ligand–receptor interactions enriched in KD. Plots were generated using standard CellChat visualization functions and CCPlotR R package v(1.4.0) (34).

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted using differentially expressed genes for each innate immune cell subset using clusterProfiler R package (v4.16.0) (35). Enrichment analyses focused on autophagy-, mitophagy-, and vesicular trafficking–related pathways. Dot plots were used to visualize enriched pathways, with dot size representing gene ratios and color indicating adjusted p values.

Pseudotime analysis was performed exclusively on monocyte subsets using monocle3 R package (v1.4.26) (36) (classical, intermediate, and non-classical monocytes) to infer continuous cell-state transitions. A trajectory inference method compatible with Seurat objects was applied to order cells along a pseudotime axis. Pseudotime distributions were compared across conditions to assess disease- and treatment-associated shifts in monocyte state trajectories.

Rabbit MAb to β-actin (Clone 13E5, HRP-conjugated), affinity purified antibody to LC3B (#2775), and HRP-linked anti-rabbit IgG (#7074) from Cell Signaling Technology (Ozyme, Saint Quentin Yvelines, France) were used for immunoblotting. Plasma-derived human serum albumin (HSA) was from Laboratoire Française de Biotechnologies (Les Ulis, France).

The IVIG products used for the mechanistic studies were Sandoglobulin®, Gamunex®, Privigen® and Octagam®. IVIG preparations were dialyzed against large volumes of phosphate-buffered saline (PBS) followed by RPMI-1640 at 4 °C for 18 hrs before use. Fc fragments of IVIG were obtained from Dr M. C. Bonnet (Institut Mérieux, Lyon, France) (37). The purity of the Fc fragment was verified by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

PBMC from healthy donors were isolated by Ficoll density gradient centrifugation of buffy bags purchased from Centre Necker-Cabanel, Etablissement Français du Sang, Paris, France (Institut National de la Santé et de la Recherche-EFS ethical committee convention 15/EFS/012; 18/EFS/033) (26). Human CD14 MicroBeads (Miltenyi) were used to isolate monocytes from the PBMC.

PBMC and monocytes (0.5 million cells/ml) were cultured in RPMI-1640 supplemented with 10% fetal calf serum, and 1% penicillin-streptomycin. PBMC were treated in vitro with different IVIG products (25 mg/ml) or equimolar concentrations (0.15 mM) of Fc fragment of IVIG or HSA (as an irrelevant protein control) until the indicated time points. To explore the role of C-type lectin receptors (CLRs), monocytes were treated with the calcium chelating agent ethylenediamine tetra acetic acid (EDTA, 0.5 mM) for one hour before culture with IVIG (25 mg/ml) for 24 hrs.

PBMC or monocytes from healthy donors were treated in vitro, pelleted, and lysed in RIPA buffer (Sigma) supplemented with protease and phosphatase inhibitors (Roche Biotech) on ice for 30 min. Protein concentrations in whole-cell lysates were quantified, and equal amounts from each sample were resolved by SDS-PAGE, followed by transfer onto polyvinylidene fluoride (PVDF) membranes using a wet western blotting method. Membranes were blocked, incubated with primary antibodies and incubated with HRP-conjugated anti-rabbit IgG secondary antibody as described previously (26). SuperSignal™ West Dura Extended Duration substrate (Thermo Fisher Scientific, Courtaboeuf, France) was used for developing the blots. β-actin served as a loading control. myImageAnalysis software v2.0 (Thermo Fisher Scientific) was used for the analyses of Western blot images. Full western blot images are presented in Supplementary Figure S6.

As described in the figure legends, the studies were conducted using PBMC or monocytes from multiple independent donors. Data on the expression of autophagy genes were compared using two-sided Wilcoxon rank sum tests for pairwise comparison in R using ggpubr R package (0.6.0). Immunoblot data were analysed by one-way analysis of variance (ANOVA) with Dunnett’s multiple comparison test by using Prism 8 (GraphPad Software Inc., CA).

We examined the distribution of innate immune cell subsets across healthy individuals (C), untreated KD patients (KD), and IVIG-treated patients (T) using single-cell transcriptomic data. Distinct alterations in cell proportions were observed between disease and treatment conditions (Figure 1).

Classical monocytes were markedly expanded in KD patients compared to healthy controls, but their frequency declined following IVIG treatment, approaching baseline levels. A similar trend was evident for intermediate monocytes, which were elevated in untreated patients and reduced post-IVIG treatment. Non-classical monocytes were also increased in patients but were nearly absent after IVIG therapy, suggesting a selective suppression of this pro-inflammatory subset. Low-density neutrophils, which were undetectable in controls, accumulated prominently in KD and were decreased following IVIG therapy, consistent with their role in systemic inflammation. In contrast, NK cells displayed the opposite pattern: they were reduced in KD patients but partially restored upon treatment, highlighting an immunoregulatory effect of IVIG. Finally, Vδ2+γδ T cells were diminished in disease compared with controls, but their frequency increased following treatment, indicating recovery of this protective lymphocyte subset.

Together, these findings, in line with previous studies (38–40), revealed a pronounced reshaping of the innate immune landscape in KD, characterized by expansion of inflammatory monocytes and neutrophils, and a concomitant loss of NK and γδ T cells. IVIG therapy reverses these alterations.

To determine whether changes in innate immune cell frequencies occurred in a coordinated manner, we performed Spearman correlation analysis of per-patient cell-type proportions across conditions. In healthy controls, correlations among innate immune subsets were generally weak, consistent with independent variation of immune compartments under homeostatic conditions (Figure 2A).

In contrast, KD patients exhibited pronounced restructuring of correlation patterns, characterized by strong positive correlations among low-density neutrophils, classical monocytes, and intermediate monocytes, alongside strong negative correlations between these myeloid populations and cytotoxic innate lymphocytes, including NK cells and Vδ2 γδ T cells (Figure 2B). These findings indicate coordinated expansion of inflammatory myeloid subsets coupled with reciprocal contraction of cytotoxic innate populations in KD.

Notably, IVIG-treated KD patients displayed partial normalization of these relationships, with attenuation of both positive myeloid correlations and negative correlations with NK and Vδ2+γδ T cells (Figure 2C), suggesting restoration of immune network balance following treatment.

Together, these analyses demonstrate that KD is associated with a coordinated remodeling of the innate immune landscape, characterized not only by altered cell-type proportions but also by disease-specific restructuring of inter-cellular relationships. The emergence of tightly coupled myeloid expansion and reciprocal suppression of cytotoxic innate lymphocytes underscores a polarized inflammatory network in KD, which is partially reversed following IVIG treatment. These findings provide systems-level evidence that KD involves functional reorganization of innate immunity rather than isolated changes in individual cell subsets.

Heatmap visualization revealed substantial differences in cell–cell communication patterns across healthy, acute untreated KD, and IVIG-treated patients (Figure 3A). In healthy controls, interactions among innate immune subsets were relatively balanced, with moderate communication involving monocyte subsets and NK cells. In contrast, KD patients showed a marked increase in interaction density, particularly involving classical and intermediate monocytes, low-density neutrophils, and NK cells, indicating heightened inflammatory crosstalk during disease. Several monocyte-centered interactions were selectively amplified in KD, consistent with aberrant activation of the innate immune compartment. Network visualization further highlighted condition-specific differences in communication architecture (Figure 3B). In healthy controls, signaling networks were relatively sparse and evenly distributed across cell types. In KD, however, monocyte subsets emerged as dominant signaling hubs, with strong outgoing and incoming interactions connecting monocytes to neutrophils, NK cells, and γδ T cells.

Further, bubble plot analysis identified specific ligand–receptor interactions that were significantly enriched in KD patients compared with healthy and IVIG-treated patients (Figure 3C). These interactions predominantly involved monocyte-derived inflammatory signaling, including pathways associated with resistin, annexin, and TLR-related signaling, which are known upstream regulators of pro-inflammatory cytokines such as TNF, IL-1β, and IL-6.

The strength and frequency of these KD-specific interactions were reduced or redistributed following IVIG treatment, supporting a role for IVIG in dampening excessive innate immune activation and restoring communication toward a homeostatic state.

Our previous work has illustrated the induction of canonical macroautophagy by IVIG in peripheral blood inflammatory innate immune cells such as monocytes, dendritic cells, and M1 macrophages, and also in treated inflammatory myopathy patients (26). Therefore, by using single-cell transcriptomic data, we first investigated the induction of autophagy genes in innate immune cell subsets of IVIG-treated KD patients.

Among the monocyte subsets, classical monocytes exhibited increased expression of GABARAP, BECN1, ATG2B, and MAP1LC3B2 and other macroautophagy genes following 24 hours of IVIG therapy, with a particularly significant upregulation of GABARAP, consistent with its known induction in macroautophagy-activated cells (31) (Figure 4A). Intermediate monocytes exhibited elevated expression of majority of the autophagy-related genes with significant upregulation of ATG7 and UVRAG, which are implicated in the extension of autophagosome and its fusion with lysosome during macroautophagy (41) (Figure 4B). Of note, in non-classical monocytes, IVIG significantly enhanced the expression of RUBCN and ATG16L2 (Figure 4C). RUBCN is known to be negative regulator of canonical autophagy and plays a critical role in LC3 induced phagocytosis (LAP), which relies on components of the autophagy machinery particularly, ATG16L2 for processing the extracellular cargo (42).

NK cells also exhibited upregulation of core macroautophagy genes following IVIG therapy, with a significant increase in RB1CC1 and ATG16L2 (Figure 5A). Low density neutrophils, which are activated in inflammatory conditions such as KD and contribute to inflammation, showed increased autophagy gene expression after IVIG therapy (Figure 5B). Similarly, γδ T cells displayed induction of autophagy with elevated expression of autophagy genes and significant upregulation of RB1CC1 (Figure 5C).

Overall, these results demonstrate that IVIG induces macroautophagy across diverse innate immune cell subsets in KD patients, thereby validating our previous findings and extending them to an inflammatory disease setting.

To investigate if IVIG therapy induces distinct forms of selective autophagy in different innate immune cell subsets, we analyzed the gene expression of key selective autophagy cargo receptors and adaptor proteins (Figures 6, 7). This approach enabled us to delineate the specific selective autophagy pathways triggered in different innate immune cells following IVIG therapy.

In classical monocytes, we observed significantly increased expression of NIPSNAP2 and CCT2, which mediate the selective degradation of damaged or depolarized mitochondria and protein aggregates, respectively (43, 44) (Figure 6A). In intermediate monocytes, IVIG therapy led to significant upregulation of mitophagy-related genes such as OPTN and BNIP3L, which participate in parkin-induced mitophagy by binding to ubiquitin chains on damaged mitochondrial membranes and recruiting autophagy proteins (Figure 6B). Non-classical monocytes showed significant induction of BCL2L13, a mitophagy cargo receptor that promotes DNM1L-mediated mitochondrial fission (45) (Figure 6C).

In NK cells, IVIG therapy significantly increased the expression of OPTN, a mitophagy adaptor, and NCOA4, a ferritin cargo receptor that promotes ferritinophagy and the release of free iron into the cytosol (Figure 7A). Low density neutrophils displayed broad upregulation of cargo receptors, including a marked increase in LGALS3, a xenophagy receptor that targets cytosolic pathogens for autophagic degradation (46) (Figure 7B). γδ T cells exhibited enhanced expression of multiple cargo proteins, with a notable increase in TECPR1, which facilitates the clearance of aggregated proteins (47, 48) (Figure 7C).

Thus, in addition to inducing macroautophagy and LC3-associated phagocytosis (LAP), IVIG therapy also triggers diverse selective autophagy pathways across innate immune cell subsets.

To investigate functional autophagy reprogramming of innate immune populations in IVIG-treated KD patients compared to acute untreated KD patients, we performed Gene Ontology (GO) and KEGG pathway enrichment analyses using differentially expressed genes from each cell subset (Figure 8). Pathway enrichment analysis revealed widespread but cell-type-specific activation of autophagy-related programs across innate immune populations. Classical and intermediate monocytes displayed strong enrichment of macroautophagy, mitophagy, and autophagosome-associated pathways (Figures 8A, B), consistent with heightened metabolic and inflammatory activity. Non-classical monocytes showed enrichment of pathways related to regulation of autophagosome maturation and vesicular trafficking, indicating a more regulatory autophagy profile (Figure 8C).

In addition to monocytes, low-density neutrophils exhibited pronounced enrichment of mitophagy and xenophagy pathways, together with innate immune signaling pathways (Figure 8E), suggesting integration of mitochondrial quality control with inflammatory responses. NK cells and γδ T cells also demonstrated enrichment of autophagy- and mitophagy-related pathways, albeit with lower gene ratios (Figures 8D, F), consistent with metabolic adaptation and functional reprogramming of cytotoxic innate lymphocytes. Collectively, these results indicate that autophagy-related processes are broadly engaged across innate immune subsets, with distinct pathway usage reflecting cell-type-specific functional states.

Pseudotime analysis of monocyte subsets showed that monocytes from KD patients were preferentially distributed toward later pseudotime states, indicating altered monocyte activation compared with controls (Figure 9A). Following IVIG treatment, monocyte subsets shifted toward earlier pseudotime values, approaching control-like distributions and suggesting normalization of monocyte state trajectories.

Consistent with this shift, autophagy gene signature scores were increased in monocytes from IVIG-treated KD patients across classical, intermediate, and non-classical subsets (Figure 9B). Importantly, increased autophagy in monocytes from IVIG-treated KD patients was associated with the restoration of pseudotime distributions toward homeostatic states. Together, these results indicate that IVIG-induced autophagy is linked to the normalization of monocyte phenotypes in KD.

While these data strongly suggested engagement of the autophagy machinery at the transcriptional level, they did not provide temporal resolution regarding when autophagy is initiated by IVIG or whether its activation is formulation-specific. Defining the timing of autophagy induction is particularly important for identifying optimal windows for downstream mechanistic studies and for distinguishing early autophagy signaling events from later-stage cellular responses such as cargo degradation or cytokine modulation.

To address this, we investigated the kinetics of IVIG-induced autophagy in vitro in PBMC isolated from healthy individuals by analyzing LC3-II protein levels. Cells were treated with IVIG (Privigen; 25 mg/ml) for different time points (1, 3, 6, 12, and 24 hours), followed by treatment with bafilomycin (50 nM) for 45 minutes. We observed that IVIG significantly increased LC3-II levels by 6 hours, which peaked at 12 hours and remained elevated at 24 hours (Figure 10).

IVIG products exhibit variability in plasma source, IgG purification methods, stabilizing agents, and formulation (49, 50). We therefore asked whether IVIG induces autophagy irrespective of the therapeutic preparation. To address this, we compared the ability of different IVIG formulations: Sandoglobulin^®, Gamunex^®, Privigen^®, and Octagam^® to induce autophagy. Treatment of PBMC in vitro with each IVIG preparation (25 mg/ml, 24 hours) resulted in a comparable upregulation of LC3-II proteins, indicating that autophagy induction is not restricted to a specific IVIG formulation but rather represents a universal phenomenon independent of source or manufacturing process (Figure 11).

IgG consists of an F(ab’)₂ fragment and an Fc fragment. Some mechanisms of IVIG are mediated through the F(ab’)₂ region, while others depend on the Fc portion (15–24). Our previous data showed that F(ab’)₂ fragments of IVIG can induce autophagy in PBMC (26). To determine whether Fc fragments also contribute, we exposed PBMC from healthy donors in vitro to either IVIG (Privigen; 25 mg/ml) or equimolar concentrations of Fc fragments (9 mg/ml) for 24 hours. Immunoblot analysis of LC3-II levels revealed that Fc fragments are dispensable for autophagy induction (Figure 12A).

Glycosylation of IgG, particularly sialylation of the F(ab’)₂ or Fc fragments, is critical for several IVIG effector functions, at least partly via glycoprotein-binding cell surface receptors of the calcium-dependent C-type lectin family (51–54). However, sialylation-independent effects of IVIG have also been reported (55–58). Consistent with this, our experiments with desialylated IVIG confirmed that sialylation is not required for autophagy induction (26). This does not, however, exclude the potential involvement of other C-type lectin receptors (CLRs) in a sialylation-independent manner (59). To examine this, we treated monocytes with the calcium chelating agent EDTA (0.5 mM) for one hour before culturing the PBMC with IVIG (Privigen; 25 mg/ml) for 24 hrs. We found that CLR inhibition had no repercussion on IVIG-induced autophagy and LC3-II levels were comparably upregulated in IVIG-treated cells with or without prior treatment of PBMC with EDTA, indicating that CLRs are not implicated in IVIG-induced autophagy in monocytes (Figure 12B).