Patient characteristics

A total of 29 pediatric participants were enrolled in the scRNA-seq analysis of this study, comprising 17 patients with EBV-HLH (HLH group), 9 patients with IM group, and 3 healthy volunteers (HV group). The diagnosis of EBV-IM was based on the presence of at least three of the following six clinical features: fever, cervical lymphadenopathy, pharyngitis or tonsillitis, hepatomegaly, splenomegaly, or eyelid edema, along with detectable EBV DNA in peripheral blood. EBV-HLH was diagnosed according to the HLH-2004 diagnostic criteria²⁰ in conjunction with EBV DNA positivity in peripheral blood. All samples for single-cell sequencing were collected within 0–3 days after admission (acute phase, prior to treatment).

The median plasma EBV DNA levels were 1.31 × 10³ copies/mL (range: 0.63 × 10³–2.41 × 10³) in the HV group, 1.50 × 10⁴ copies/mL (range: 1.05 × 10³–1.02 × 10⁵) in the IM group, and markedly elevated in the HLH group at 3.84 × 10⁵ copies/mL (range: 1.73 × 10³–9.50 × 10⁶). Clinically, EBV-HLH patients exhibited hallmark features of HLH, including persistent high-grade fever and cytopenia. Bone marrow hemophagocytosis was observed in 88.2% of cases. In 41.2% cases, reduced NK cell activity was observed, while all cases exhibited elevated levels of soluble CD25 (sCD25). Among the 17 EBV-HLH patients, 15 (88.2%) achieved complete remission (CR) following treatment, while 2 (11.8%) were classified as non-responders (NR). Analysis of EBV-infected lymphocyte subsets in all 17 EBV-HLH cases revealed diverse infection patterns: 2 patients exhibited EBV infection restricted to B cells, 4 had dual positivity in B and NK cells, and 11 demonstrated EBV presence across T, B, and NK cell subsets.

Identifying the single-cell transcriptional landscape

We obtained a transcriptome dataset comprising 216,817 cells after excluding low-quality cells. Of these, 25,091 cells (11.57%) were from healthy volunteers (HV), 56,752 cells (26.18%) were from individuals with IM condition, and 134,974 cells (62.25%) were from HLH condition patients. After normalizing for read depth and mitochondrial read counts, we compiled a high-quality, batch-free dataset suitable for comparative analysis. Principal component analysis was then conducted (Fig. 1a). Canonical gene markers were employed to classify the transcriptomes into 10 major cell types (Fig. 1b–d and Supplementary Fig. 1a, b). These cell types include CD4⁺ T cells, CD8⁺ T cells, γδ T cells, NK cells, NKT cells, B cells, plasma B cells, monocytes, classical dendritic cells (cDC), and plasmacytoid dendritic cells (pDC).

To compare cell distribution across different groups, we calculated Ro/e values and cell proportions for the 10 major cell types (Fig. 1e and Supplementary Fig. 1c–f). Compared to the HV group, the proportion of CD8⁺ T cells increased in both the IM and HLH groups, correlating with the activation of CD8⁺ T cells by the immune system in response to EBV infection. In contrast, γδ T cells, B cells, cDC, and pDC were predominantly enriched in the HV group. Plasma cells were enriched in the IM group but decreased in the HLH group, suggesting a potential weakening of humoral immunity in patients with HLH.

To investigate the pathogenic immune response to EBV-HLH, termed the “inflammatory factor storm,” we evaluated the functional module scores of cytokines (Fig. 1f) and inflammatory responses (Fig. 1g) in the major cell types. The UMAP plot indicated that high expression of cytokine-related genes was mainly concentrated in T cells and NK cells, with monocytes being the central cell population mediating inflammatory responses.

Overall, this finding suggests that EBV-HLH exhibits a unique transcriptional profile, characterized by both shared and distinct patterns of immune cell dysfunction when compared to IM and healthy control groups.

Activation of NF-κB signaling in T cells of HLH

To elucidate the heterogeneity among individual T cell subpopulations across the three clinical states, we re-clustered all T cells from PBMCs and performed unsupervised clustering, yielding a total of 13 clusters (Fig. 2a and Supplementary Fig. 2a). The top 10 differentially expressed genes (DEGs) for each cell cluster are depicted in Fig. 2b and Supplementary Fig. 2b. Based on canonical T cell markers (Fig. 2c and Supplementary Fig. 2c), these clusters were categorized into three CD4⁺ T cell clusters (CD4), nine CD8⁺ T cell clusters (CD8), and one γδ T cell cluster (T_C13_GDT_TRDV2; TRDV2, TRGV9, TRGC1, TRDC).

Functional scoring results (Supplementary Fig. 2d) identified three naïve T cell subgroups highly expressing naïve gene features (TCF7, SELL, CCR7, LEF1): T_C01_CD4_CCR7, T_C02_CD4_TCF7, and T_C04_CD8_CCR7 clusters. All CD8⁺ T cell subgroups, except for T_C04_CD8_CCR7, highly expressed cytotoxic effector molecules (GZMB, NKG7, GNLY, PRF1). Among these, T_C05_CD8_STMN1, T_C06_CD8_MKI67, and T_C07_CD8_UBE2C exhibited high expression of proliferation markers. T_C08_CD8_RGS1 and T_C12_CD8_HAVCR2 also highly expressed cytokine genes (IFNG, TNF, CXCR3, CCL3, CCL4) and exhaustion markers (HAVCR2, CTLA4, LAG3, PDCD1, LAYN), consistent with the functional scoring results (Supplementary Fig. 2d). The T_C03_CD4_MAF cluster had relatively high cytotoxic and exhaustion scores (Supplementary Fig. 2d), suggesting these cells might be CD4⁺ CTLs. Additionally, the T_C02_CD4_TCF7 and T_C03_CD4_MAF clusters highly expressed Treg markers (FOXP3, IL2RA, TNFRSF4) (Fig. 2c). However, these clusters had a low proportion of positive cells (Supplementary Fig. 2c), which may be due to the mixing of CD4⁺ T cells with a small fraction of Treg cells.

Next, the degree of cellular enrichment was quantified based on the Ro/e value of each cell cluster (Fig. 2d and Supplementary Fig. 2e). CD8⁺ CTL clusters, especially the T_C05_CD8_STMN1, T_C06_CD8_MKI67, and T_C07_CD8_UBE2C, were most enriched in the IM group (Supplementary Fig. 2f, g). In contrast, the T_C08_CD8_RGS1 and T_C12_CD8_HAVCR2 cell clusters, which have high cytokine and exhaustion scores, were enriched in the HLH group. As expected, the exhaustion score of CD8 + T cells in the HLH group was significantly higher than that in the IM group (Supplementary Fig. 2h). This may be related to the insufficient antiviral capacity and over-activation of cytokine production in T cells in HLH.

To further investigate the differential transcriptomic alterations in T cells (excluding γδ T and NKT cells) within children with HLH, we compared the DEGs of T cells in HLH patients to those in HV or IM patients and conducted enrichment analyses. Compared to HV (Fig. 2e), the upregulated genes in HLH were mainly enriched in processes, such as oxidative phosphorylation, mitochondrial electron transport, cytochrome c to oxygen, cell killing, regulation of canonical NF-κB signal transduction, and positive regulation of cytokine production. In contrast to IM (Fig. 2f), the upregulated genes in HLH were involved in cellular response to cytokine stimulus, leukocyte migration, regulation of apoptotic signaling pathways, and regulation of canonical NF-κB signal transduction.

We further screened for genes specifically upregulated or downregulated in the HLH group. Genes involved in the positive regulation of canonical NF-κB signal transduction (NFKBIA, S100A12, TNFSF10, TRIM22, TMED4) and the JAK-STAT signaling pathway (CISH, PIM1, SOCS3) were upregulated in the HLH group (Fig. 3a). The activation of NF-κB signaling pathways can lead to increased cytokine production²¹. Consistent with the single-cell sequencing results, flow cytometry analysis confirmed that phosphorylated p65 levels in T cells were elevated in the HLH group compared to the IM group (Fig. 3b), indicating the activation of the canonical NF-κB pathway in EBV-HLH. As expected, most T cell subgroups had the highest cytokine scores in the HLH group (Fig. 3c). Genes involved in the regulation of vesicle-mediated transport, regulation of receptor-mediated endocytosis, and positive regulation of cell adhesion were specifically downregulated in the HLH group (Fig. 3d). The downregulation of these functions could be a potential mechanism leading to the impaired immune response of T cells in HLH.

HLH samples were further stratified into two subgroups based on clinical testing for EBV infection in T cells: the EBV-T group (patients with EBV-infected T cells) and the non-EBV-T group (patients without EBV-infected T cells) (Fig. 3e). Comparative transcriptomic analysis revealed substantial functional heterogeneity in T cell gene expression profiles between these two groups. In the EBV-T group, upregulated genes were predominantly enriched in innate immune-related pathways, including antiviral responses, type I interferon signaling, ribosome assembly, and cytoplasmic translation. Additionally, pathways involved in leukocyte migration, cytokine production, and NF-κB signaling were significantly activated. In contrast, downregulated genes in the EBV-T group were associated with adaptive immune processes such as T cell activation, antigen receptor signaling, cytotoxicity, antigen presentation, and lymphocyte proliferation. The concurrent activation of innate immune responses and suppression of adaptive immunity in EBV-infected T cells reflects a paradoxical immune state that may underlie the susceptibility to cytokine storm in a subset of EBV-HLH patients.

SH2D1A, ADA, and RAB27A are genes known to be involved in the pathogenesis of HLH^4,22,23,24. These genes were highly expressed in the IM group but downregulated in the EBV-T group (Fig. 3f). This pattern suggests that direct EBV infection of T cells may suppress the expression of key genes involved in T cell activation (SH2D1A), purine metabolism (ADA), and antigen processing or presentation (RAB27A, SLAMF7), thereby impairing antiviral immune responses and facilitating viral persistence. In contrast, expression levels of these genes in the non-EBV-T group closely resembled those observed in the IM group. Notably, CD81 expression followed a graded pattern, with high expression in the IM group, intermediate levels in the non-EBV-T group, and markedly reduced expression in the EBV-T group. As a tetraspanin involved in immune synapse formation, cell adhesion, and signal transduction²⁵, reduced CD81 expression might compromise T cell activation and regulation, thereby promoting uncontrolled inflammation. Furthermore, EBV infection could exacerbate this effect by further suppressing CD81 expression in T cells.

In the HLH-D (the deceased HLH sample) group, the upregulated genes were primarily enriched in NF-κB signaling activation, cytokine storm, and intrinsic apoptosis pathways (Fig. 3g). Conversely, the downregulated genes were mainly enriched in oxidative phosphorylation, positive regulation of leukocyte activation, regulation of hematopoiesis, and positive regulation of binding, suggesting impaired metabolic function and dysregulated T cell activation in the HLH-D group.

In addition, compared to the HV group, ANXA1 expression was elevated in the IM group compared to healthy volunteers (HV), but notably downregulated in the HLH group (Fig. 3d), with further reduction observed in the HLH-D subgroup (Fig. 3h). These findings suggest that, as an endogenous anti-inflammatory molecule^26,27,28, ANXA1 expression levels may correlate with disease severity and prognosis.

Activation of classical inflammatory signaling in NK cells of HLH

All NK cells (excluding NKT cells) were re-clustered, resulting in five distinct clusters (Fig. 4a and Supplementary Fig. 3a). The top 10 DEGs for each cluster are presented in Fig. 4b and Supplementary Fig. 3b. NK cells are traditionally categorized into two main groups based on CD56 (NCAM1) and CD16 (FCGR3A) expression levels: CD56^brightCD16^low and CD56^dimCD16^{high 29}. Based on functional gene expression (Fig. 4c), we identified one CD56^brightCD16^low cell cluster (NK_C03_XCL2) and two CD56^dimCD16^high cell clusters (NK_C01_GZMH and NK_C02_AKR1C3). The remaining two clusters (NK_C04_STMN1 and NK_C05_MKI67) exhibited high proliferative activity. Consistent with previous studies²⁹, CD56^brightCD16^low C03_XCL2 NK cells highly expressed cytokine-related genes, such as LTB and IL2RA. The two CD56^dimCD16^high NK cell clusters highly expressed cytotoxic effector genes, including PRF1 and several granzymes (GZMB, GZMA, GZMH), except for GZMK, which was expressed only in CD56^brightCD16^low C03_XCL2 NK cells. CD56^dimCD16^high C01_GZMH NK cells with high expression of KLRC2 (NKG2C) were identified as adaptive NK cells.

Compared to healthy volunteers (Fig. 4d), the upregulated DEGs in HLH were primarily involved in oxidative phosphorylation, response to type II interferon, cell killing, and positive regulation of the inflammatory response. In contrast to the IM group (Fig. 4e), the upregulated genes in HLH were enriched in pathways including cell activation, granulocyte chemotaxis, granzyme-mediated programmed cell death signaling, cellular response to cytokine stimulus, and regulation of apoptotic signaling. These findings support that HLH involves different pathway activation.

A more detailed analysis identified genes specifically upregulated or downregulated in the HLH group. Markers of reactive oxygen species and oxidative stress (GSTP1, LDHA, HSPB1, PRDX5, PGD) were abnormally upregulated in HLH (Fig. 5a), indicating that NK cells in HLH patients might be experiencing increased oxidative stress. Similarly, genes associated with the canonical NF-κB pathway were upregulated, and this was corroborated in clinical samples by elevated phosphorylation levels of p65 in the HLH group compared to the IM group (Fig. 5b).

Among the downregulated genes (Fig. 5c), we identified RASGRP1, a gene previously associated with EBV-HLH³⁰. Genes involved in processes, such as positive regulation of leukocyte cell-cell adhesion, defense response, regulation of cell killing, and immunological synapse formation (CD81, CD160, CD6, DOCK8, ITGA4), were also downregulated, potentially impairing binding between NK cells and target cells and weakening NK cell-mediated immune surveillance and cytotoxic functions. Additionally, compared to the HV group, DUSP1 was upregulated in the IM group, but downregulated in HLH. DUSP1 downregulation can lead to sustained activation of the MAPK pathway³¹, mediating a more intense or prolonged inflammatory response³². Other genes involved in the negative regulation of the MAPK cascade, including DUSP2, NCOR1, STK38, PDCD4, and ITCH, were also downregulated in the HLH group.

The HLH group was further stratified into EBV-NK (patients with EBV-infected NK cells) and non-EBV-NK (patients without EBV-infected NK cells) subgroups, based on clinical testing for EBV infection in NK cells. Comparative analysis revealed that EBV infection altered both metabolic and immunological profiles of NK cells between the two groups (Fig. 5d). NK cells in the EBV-NK group exhibited signs of metabolic reprogramming, characterized by upregulation of genes involved in ATP metabolism and oxidative phosphorylation. This metabolic activation was accompanied by enhanced expression of genes related to cytotoxicity, cytokine production, and positive regulation of immune defense. However, this activated state coexisted with features of immune dysfunction. Specifically, genes associated with negative regulation of NK cell-mediated cytotoxicity, immune suppression, and apoptotic signaling were also enriched, implying that EBV may facilitate immune evasion by functionally impairing NK cells. Downregulated genes in the EBV-NK group were enriched in pathways related to lymphocyte differentiation, antigen presentation, cellular activation, and cytotoxic effector function.

Furthermore, LYST, a known HLH-associated gene^4,5, exhibited comparable expression in HV, IM, and non-EBV-NK groups but was downregulated in the EBV-NK group (Fig. 5e). Similarly, DUSP1, the negative regulator of MAPK signaling, was upregulated in the IM and non-EBV-NK groups but downregulated in the EBV-NK group (Fig. 5f). DUSP1 expression was significantly lower in the HLH-D group compared to the HLH-L group (Fig. 5g), and markedly higher in the IM group. These patterns suggest that reduced DUSP1 expression may contribute to sustained MAPK activation and could be associated with disease severity and poor prognosis in EBV-HLH.

The NF-κB and MAPK pathways are key regulators of inflammatory responses, and their dysregulation is closely associated with excessive cytokine release in hyperinflammatory conditions^33,34. In EBV infection, the latent membrane protein LMP1 activates NF-κB through a TRAF2/5-dependent mechanism, suppresses SH2D1A (SAP) expression, and promotes secretion of Th1-type cytokines such as TNF-α and IFN-γ³⁵. Concurrently, EBV triggers the MAPK/p38 pathway to promote the production of IL-6 and IL-8³⁶. These findings highlight the therapeutic potential of targeting NF-κB and MAPK signaling to mitigate cytokine storms in EBV-HLH (Fig. 5h) and related hyperinflammatory syndromes.

Reduced memory B and plasma cells in HLH

We re-clustered all B cells and performed unsupervised dimensionality reduction clustering, resulting in five distinct subpopulations (Supplementary Fig. 4a). Each cell cluster is characterized by the top 10 DEGs (Supplementary Fig. 4b). B_C01_TCL1A and B_C02_IGLC2 highly expressed naive B cell marker genes (IGHD, CR2, CCR7, FCER2), whereas B_C03_AIM2 and B_C04_CD1C exhibited memory B cell features (CD27, CD1C). B_C05_FCRL3 aligned with activated naive B cell characteristics, showing high expression of IGHD, ITGAX, TBX21, FCRL5, and CD19, but low expression of CD24, CD38, CD27, and CR2. Additionally, B_C01_TCL1A, B_C02_IGLC2, and B_C05_FCRL3 also highly expressed interferon-stimulated genes (ISGs) (Supplementary Fig. 4c). The composition of B cells in HV and IM groups was similar; however, in HLH, the proportion of memory B cells was reduced, and plasma cells were absent (Supplementary Fig. 4d). This alteration may weaken the immune clearance capability against EBV and other pathogens in HLH patients.

Compared to HV, the upregulated DEGs in HLH were involved in various biological processes, including response to viruses, positive regulation of cytokine production, regulation of type I interferon production, regulation of canonical NF-κB signal transduction, response to type II interferon, inflammatory response, and apoptotic signaling pathways (Supplementary Fig. 4e). Compared to IM (Supplementary Fig. 4f), the upregulated genes in HLH were mainly enriched in processes such as negative regulation of the immune system, response to type II interferon, positive regulation of cytokine production, regulation of canonical NF-κB signal transduction, positive regulation of programmed cell death, negative regulation of cell population proliferation, regulation of IL4 production, and regulation of type I interferon production.

Further analysis identified genes specifically upregulated or downregulated in the HLH group. Genes involved in the positive regulation of canonical NF-κB signal transduction (BST2, CCR7, S100A12, TNFSF10, TRIM22, PIM2, SHISA5) and in the negative regulation of cell population proliferation (B2M, GSTP1, IFIT3, TYROBP, H2AC6, IFITM1, NMI, PIM2, EAF2) were upregulated in the HLH group (Supplementary Fig. 4g). As expected, we observed elevated levels of p65 phosphorylation in B cells from the HLH group, indicating a higher degree of NF-κB pathway activation compared to the IM group (Supplementary Fig. 4h). Meanwhile, the SYK gene, involved in leukocyte differentiation, B cell receptor signaling pathway, and positive regulation of cell adhesion, was downregulated in the HLH group (Supplementary Fig. 4i). SYK encodes a tyrosine kinase, a critical component in BCR activation³⁷. Thus, the downregulation of SYK could impede BCR signal transduction, potentially limiting B cell activation and function, and reducing the body’s antiviral capability.

Abnormally enhanced MHC class I antigen presentation of DCs in HLH

The proportion of cDC in the HLH group was reduced compared to the HV and IM groups (Supplementary Fig. 1f). Differential gene expression analysis of cDC between HLH and HV or IM groups, followed by enrichment analysis (Supplementary Fig. 5a, b), showed that genes upregulated in HLH versus HV were primarily associated with antiviral defense and innate immune activation pathways, including viral response, interferon-gamma response, cytokine production, and MHC class I presentation. This indicates that residual cDC in HLH are in a highly activated state during viral infection and may drive CD8⁺ T cell activation via MHC class I pathways. Concurrently, the upregulation of apoptotic signaling pathways might partially explain the reduced cDC numbers.

In contrast, downregulated genes in HLH versus HV were enriched in pathways related to cytoplasmic translation, ribosome assembly, MHC class II antigen presentation, and cellular homeostasis. When comparing HLH to IM, cDC in HLH showed upregulation of genes involved in antigen presentation, while downregulated genes were enriched in defense response, cytokine production, and energy metabolism. Additionally, TNF production in cDC was lower in HLH compared to IM. These transcriptional changes suggest alterations in both activation and metabolic profiles of cDC in the HLH group.

pDC were consistently reduced in both IM and HLH groups. As primary producers of type I interferons (IFN-I), pDC depletion likely impairs antiviral immunity. Comparative analysis of DEGs revealed that, in HLH versus HV, upregulated genes in pDC were enriched in antigen processing and presentation (particularly via MHC class I), IFN-γ response, oxidative phosphorylation, ATP metabolism, and T cell activation (Supplementary Fig. 5c). These findings indicate that residual pDC in HLH are metabolically active and engaged in cell-mediated immune responses, potentially contributing to CD8⁺ T cell hyperactivation.

Similarly, in HLH versus IM, upregulated genes were also associated with MHC class I antigen presentation and cytotoxic T cell regulation (Supplementary Fig. 5d). In contrast, genes downregulated in HLH versus HV were linked to histone modification, myeloid differentiation, IFN-β production, and dendritic cell development, suggesting impaired pDC maturation and reduced antiviral signaling.

In summary, DCs in HLH display enhanced MHC class I antigen presentation, promoting CD8⁺ T cell overactivation, while concurrent suppression of protein synthesis and activation of apoptotic programs may drive DCs exhaustion, further compromising immune regulation.

Identifying monocytes as key effector cells in the inflammatory storm

We also re-clustered all monocytes from PBMCs and performed unsupervised dimensionality reduction clustering, resulting in 10 subpopulations (Fig. 6a and Supplementary Fig. 6a). Each cell cluster was characterized by the top 10 ranked genes in expression levels (Fig. 6b and Supplementary Fig. 6b). Based on the expression levels of CD14 and CD16 (FCGR3A) (Fig. 6c and Supplementary Fig. 6c), seven subpopulations (C01 ~ C07) were identified as classical monocytes (CD14⁺, FCGR3A⁻); Mono_C10_FCGR3A as non-classical monocytes (CD14⁺, FCGR3A⁺⁺); and Mono_C08_HLA-DQA1 and Mono_C09_C1QB as intermediate monocytes (CD14⁺, FCGR3A⁺). Both Mono_C08_HLA-DQA1 and Mono_C09_C1QB highly expressed HLA-II class genes.

Mono_C08_HLA-DQA1 was specifically enriched in the IM group, with no upregulation observed in HLH compared to HV (Fig. 6d). Functional enrichment analysis of this cluster revealed that its upregulated genes were primarily associated with MHC class II antigen presentation pathways, including “antigen processing and presentation of exogenous peptide antigens” and “MHC class II protein complex assembly” (Fig. 6e). In contrast, downregulated genes were associated with humoral immune responses, microbial defense, and positive regulation of cytokine and inflammatory signaling (Fig. 6e). These findings suggest that monocytes in IM adopt an antigen-presenting phenotype, supporting MHC class II–mediated T cell activation while limiting inflammatory responses—potentially contributing to the self-limiting course of IM. In contrast, monocytes in HLH display a pro-inflammatory profile that may promote cytokine storm development.

Compared to HV (Fig. 7a), the upregulated DEGs in HLH participated in cell activation, positive regulation of response to external stimuli, positive regulation of cytokine production, and positive regulation of canonical NF-κB signal transduction. Compared to IM (Fig. 7b), the upregulated DEGs in HLH were enriched in pathways, such as inflammatory response, cellular response to cytokine stimulus, antigen processing and presentation, regulation of dendritic cell differentiation, and regulation of the viral life cycle.

Further analysis identified genes specifically upregulated or downregulated in the HLH group, revealing differences compared to HV and IM groups. IDO1 was upregulated in HLH, participating in pathways including negative regulation of leukocyte activation, negative regulation of cell-cell adhesion, inflammatory response, and positive regulation of cytokine production (Fig. 7c). This indicates its role in the pathogenesis of HLH. Additionally, genes involved in the positive regulation of the MAPK cascade and canonical NF-κB signal transduction were also upregulated in HLH. In contrast, genes involved in the negative regulation of the MAPK cascade (DUSP1, PTPRC, RGS2, TLR4) were downregulated in HLH (Fig. 7d).

Enrichment analysis results showed that monocytes in HLH exhibited abnormally activated NF-κB signaling, MAPK signaling, and inflammatory responses. Consistent with transcriptomic data, phosphorylated p65 levels were elevated in HLH monocytes compared to IM, indicating activation of the canonical NF-κB pathway (Fig. 7e). Furthermore, Ucell scoring indicated that Mono_C06_CD14_IDO1 had the highest inflammation score (Fig. 7f), and this subgroup was enriched in the HLH group (Fig. 6d and Supplementary Fig. 6d). Moreover, across all monocyte subpopulations, inflammation scores in HLH were higher than in HV and IM groups (Fig. 7g).

Collectively, these findings indicate that monocytes may function as effector cells and contribute significantly to the inflammatory cytokine storm observed in HLH.

IDO1⁺ monocyte with specific activation revealed by trajectory analysis

To investigate the transitions between monocyte subpopulations, we utilized Monocle2 for pseudotime trajectory analysis³⁸. This approach allowed us to arrange monocytes along a trajectory based on their expression profiles and transition dynamics. Monocytes were categorized into five cell groups, each representing a different state of differentiation, based on their gene expression patterns. According to the primary branching points, monocytes differentiated into two distinct branches, termed cell fate 1 and cell fate 2 (Fig. 8a and Supplementary Fig. 7a). The tissue distribution of these two fates differed significantly (Fig. 8b and Supplementary Fig. 7b, c), with cell fate 1 predominantly found in HLH, and Mono_C06_CD14_IDO1 cells clustering at the end of fate 1, indicating disease specificity of this differentiation path.

Subsequent analysis focused on gene expression patterns and pathways associated with changes along fate 1 (Fig. 8c). We observed that certain genes, including IDO1 and IL18, were gradually upregulated along fate 1. These genes were associated with inflammatory responses and the positive regulation of cytokine production. In contrast, some genes that were downregulated along fate 1 were related to endocytosis (ITGB1) and the negative regulation of the MAPK cascade (DUSP1, DUSP7). Additionally, genes responding to interferon (IFI27, IRF7, IFITM1, ISG15, IRF1, GBP1, IFNGR2) were upregulated along fate 1, suggesting that cells differentiating along fate 1 are influenced by interferon stimulation, exhibiting pro-inflammatory characteristics with reduced phagocytic ability.

We then predicted transcription regulators for cells in fate 1, identifying the top 10 transcription factors with the highest activity scores: MAFG, GTF2B, SOX6, TCF7L1, ETV7, IRF7, ZNF713, ZNF846, KLF5, and ZNF32 (Fig. 8d and Supplementary Fig. 7d). We defined the cell subgroup prior to the branch as Mono_S01, the subgroup on branch 2 as Mono_S02, and the subgroup on branch 1 as Mono_S03 (Fig. 8e). Comparing with the inflammation score plot (Fig. 8f), cells with high inflammation scores were primarily concentrated in the Mono_S03 subgroup, which was enriched in the HLH group (Fig. 8g and Supplementary Fig. 7e). The bubble chart illustrated the characteristic genes of Mono_S03. Notably, IDO1 was highly expressed in Mono_S03, while it was almost not expressed in the other two groups, making it a distinct marker for the Mono_S03 cluster (Fig. 8h).

Validation of IDO1⁺ monocytes and L-kynurenine as new biomarkers for HLH

To verify the potential of IDO1⁺ monocyte populations as biomarkers for HLH, we conducted flow cytometry analysis on fresh peripheral blood samples collected from a validation cohort. This cohort included 6 healthy controls (HV), 4 pediatric patients with infectious mononucleosis (IM), 7 newly diagnosed EBV-HLH pediatric patients (HLH), and 3 paired EBV-HLH pediatric patients who achieved complete remission (CR) after standard treatment (HLH-T). The samples of HLH-T group were collected within 4–6 weeks after standard treatment. The remaining samples were collected within 0–3 days after admission (acute phase, prior to treatment). The flow cytometry analysis revealed that traditional monocyte markers, CD14 and CD16, were unable to differentiate HLH patients from HVs and IM patients. However, there was a higher proportion of IDO1⁺ monocytes in HLH patients compared to the other groups. There was no significant difference in the proportion of IDO1⁺ monocytes among HV, IM, and HLH-T groups (Fig. 9a). This suggests that the level of IDO1⁺ monocytes increases during the acute phase of HLH and decreases to levels comparable to healthy children after treatment. The analysis of paired samples also supported this conclusion (Fig. 9b). Therefore, IDO1⁺ monocytes can serve as biomarkers for the diagnosis and treatment of HLH.

To assess whether IDO1 expression reflects general monocyte activation or represents a distinct feature of EBV-HLH, we analyzed its association with established activation markers (HLA-DR, CD163, and CX3CR1) using flow cytometry in a new validation cohort comprising IM and EBV-HLH samples. A significantly higher proportion of IDO1⁺ monocytes was observed in the HLH group compared to the IM group. In contrast, HLA-DR expression was uniformly high in both groups, with no significant differences in the proportions of HLA-DR⁺, CD163⁺, or CX3CR1⁺ monocytes (Fig. 9c). Further analysis within the HLH group revealed no significant differences in HLA-DR⁺, CD163⁺, or CX3CR1⁺ populations between IDO1⁺ and IDO1⁻ monocytes. Similarly, the frequency of IDO1⁺ cells did not vary across subgroups defined by the expression of these markers (Fig. 9d). These results suggest that IDO1 upregulation in HLH occurs independently of conventional monocyte activation states. The flow gating strategy and FACS plots for staining were in Supplementary Fig. 8.

IDO1 is the key rate-limiting enzyme that metabolizes tryptophan into kynurenic acid³⁹. To further investigate this, we collected serum samples for targeted tryptophan metabolomics sequencing. Consistent with the results from single-cell sequencing and flow cytometry, the concentration levels of L-kynurenine, the direct metabolite of IDO1, were specifically elevated in the HLH group and reduced in the HLH-T group (Fig. 9e). We further investigated in vitro whether kynurenine (KYN) affects the inflammatory responses of monocytes and cytotoxic T lymphocytes (CTLs). We conducted endogenous enhancement (via IDO1 overexpression) and exogenous kynurenine (KYN) supplementation treatments in THP1 and U937. qRT-PCR analysis confirmed that KYN significantly upregulated pro-inflammatory cytokine expression in these monocytes (Fig. 9f). Similarly, exogenous KYN stimulation enhanced pro-inflammatory responses in primary T cells and NK92 cells (Fig. 9g).

Recognition of cellular communication in monocytes

To gain deeper insight into the cellular communication between monocytes and other immune cells, we utilized the CellphoneDB algorithm to investigate interactions among cell populations within the HV and HLH groups. This analysis revealed that interactions between monocyte populations and DCs, NK cells, NKT cells, and CD8⁺ T cells were particularly abundant, with the highest number of interactions occurring between monocytes and DCs (Supplementary Fig. 9a).

Significant changes were observed in the interactions between cells in the HV and HLH groups. Compared to HV, the ligand-receptor interactions in monocytes were increased in HLH patients. In HLH, monocytes highly expressed receptors, such as CCR1, TNFRSF1A, TNFRSF1B, and TNFRSF14, which can receive stimulation from various cytokines secreted by other cell types. Additionally, monocytes in HLH highly expressed cytokine ligands, such as CXCL10, CXCL8, CCL3, TNF, and TNFSF10, which can stimulate corresponding receptors on other cell types (Fig. 10a–c and Supplementary Fig. 9b, c).

Overall, these results indicate the enhanced communication between monocytes and other immune cells during EBV-HLH progression.