Neutrophil-derived S100A8/A9 impairs megakaryocyte maturation in immune thrombocytopenia

Abstract

Immune thrombocytopenia (ITP) features autoantibody-mediated platelet clearance, but how inflammatory mediators impair thrombopoiesis remains less defined. Using single-cell and spatial transcriptomics, we identify a neutrophil–megakaryocyte axis in ITP marrow in which neutrophil-derived S100A8/A9 engages TLR4 and activates JNK/c-Jun signaling. This pathway represses the megakaryocyte master regulator GATA1 and constrains maturation. In primary CD34⁺-derived megakaryocyte cultures, ITP plasma reduces polyploidization, maturation-marker expression, and platelet-like particle release; tasquinimod treatment or TLR4 inhibition partially restores these endpoints. In active and passive ITP mouse models, tasquinimod improves platelet recovery, normalizes marrow megakaryocyte abundance, and dampens MAPK signatures. Together, these complementary data position neutrophils as contributors to thrombopoietic failure in ITP, connect marrow inflammation to defective platelet production, and nominate the S100A8/A9–TLR4–JNK/c-Jun axis as a therapeutic target. Modulating this pathway partially restores megakaryopoiesis and alleviates thrombocytopenia in vivo, supporting pharmacologic targeting of S100A8/A9–TLR4 signaling as a potential adjunct to ITP therapy.

Immune thrombocytopenia (ITP) is often attributed to peripheral platelet clearance, but how marrow inflammation suppresses platelet production remains incompletely understood. The authors here show that neutrophil-derived S100A8/A9 activates TLR4–JNK/c-Jun signaling to repress GATA1 in megakaryocytes, impairing megakaryocyte maturation and platelet production.

Introduction

Immune thrombocytopenia (ITP) is an autoimmune disorder marked by persistently low platelet counts (<100 × 10⁹/L) with bleeding manifestations (cutaneous purpura, mucosal, or life-threatening hemorrhage)^1,2. Although ITP can occur at any age, adult-onset disease is typically chronic, whereas childhood ITP often presents acutely and may resolve spontaneously^3,4. First-line therapies—high-dose corticosteroids, intravenous immunoglobulin (IVIG), and anti-D immunoglobulin—with thrombopoietin receptor agonists (eltrombopag, romiplostim) for relapsed or refractory cases^5–7 leave a substantial subset with persistent or relapsing thrombocytopenia^8,9. Pathophysiology involves humoral and cellular immunity (autoantibodies to platelet antigens, cytotoxic T-cell activity, dysregulated cytokines)^10,11. However, most work has emphasized peripheral platelet clearance^12,13 rather than how bone-marrow inflammatory cues—potentially from macrophages, T cells, or neutrophils—impair megakaryocyte (MK) maturation. Emerging data in autoimmunity suggest neutrophil-derived alarmins (e.g., S100 proteins) as candidate mediators for a “neutrophil–megakaryocyte” axis in ITP^14,15. Recent single-cell RNA-seq profiling of bone-marrow CD34⁺ hematopoietic stem and progenitor cells (HSPCs) in ITP revealed broad progenitor-level transcriptomic alterations and impaired megakaryopoiesis, including a reduction of CD9⁺ and HES1⁺ cells within Lin⁻CD34⁺CD45RA⁻ HSPCs and diminished megakaryocytic differentiation potential¹⁶.

Among alarmins, S100A8/A9 has gained prominence for modulating inflammation via receptors such as Toll-like receptor 4 (TLR4)¹⁷. Prior hematologic reports of S100A8/A9, however, derive from immune thrombotic thrombocytopenic purpura (iTTP)¹⁸, myelodysplastic syndromes (MDS)¹⁹, and multiple myeloma (MM)²⁰—contexts with pathogenesis distinct from ITP (e.g., ADAMTS13 deficiency–driven microvascular thrombosis in iTTP; clonal hematopoiesis in MDS; tumor–stroma interactions in MM)—and none evaluated whether S100A8/A9 perturbs MK maturation or platelet production in ITP. Against this backdrop, whether and how S100A8/A9-driven signaling directly disrupts MK maturation and thrombopoiesis in ITP remains unknown, defining a critical gap this study addresses.

Neutrophils actively sculpt the MK niche: intravital imaging shows they can “pluck” MKs to accelerate proplatelet formation and platelet release²¹. They transit MK cytoplasm via emperipolesis, with short- and long-dwell forms that may reflect distinct functional states, and can transfer membrane to nascent platelets^22,23. Inflammatory cues couple innate immunity to thrombopoiesis: interleukin-1α can trigger rapid rupture-type thrombopoiesis during acute demand²⁴, whereas acute innate activation can conversely impair megakaryopoiesis²⁵. These observations position neutrophil alarmins—particularly S100A8/A9—as plausible brakes on MK maturation in ITP.

Here, we test whether neutrophil-derived S100A8/A9 links inflammatory remodeling of the ITP marrow niche to defective megakaryopoiesis. Integrating single-cell and spatial transcriptomics with primary megakaryocyte assays and ITP mouse models, we identify neutrophils as a major S100A8/A9 source and show that S100A8/A9–TLR4 signaling activates JNK/c-Jun, represses GATA1, and limits megakaryocyte maturation. S100A8/A9 antagonism or TLR4 inhibition partially restores maturation endpoints in primary cultures, and tasquinimod improves platelet recovery in active and passive ITP models. These findings connect innate inflammatory cues in the bone-marrow niche to impaired platelet production and nominate S100A8/A9–TLR4–JNK/c-Jun signaling as a targetable pathway in ITP. This framework supports adjunctive strategies that restore thrombopoiesis while complementing therapies directed at peripheral platelet clearance.

Results

Neutrophils dominate marrow S100A8/A9 in ITP

We first performed scRNA-seq on bone marrow samples from patients with ITP (n = 6) and age-matched healthy controls (n = 5), aiming to characterize the heterogeneity of hematopoietic cells in ITP at the single-cell level. Approximately 10,000 bone marrow–derived cells were captured per sample using a droplet-based microfluidic platform for library construction. To ensure data quality, we systematically examined nFeature_RNA, nCount_RNA, and percent.mt (Supplementary Fig. 1A). After filtering out cells with extreme metrics, we tested clustering resolutions from 0.2 to 1.2, ultimately choosing 0.5 to balance cluster granularity and interpretability (Supplementary Fig. 1B). Under these conditions, we could clearly identify distinct subpopulations including monocytes, macrophages, CD4 + T cells, NK cells, neutrophils, B cells, erythrocytes, CD8+ T cells, plasma cells, platelets, dendritic cells (DCs), megakaryocytes, and hematopoietic stem cells (HSCs) (Fig. 1A). Further quantitative analysis showed a marked increase in the proportion of neutrophils (3.4% to 6.9%) in ITP bone marrow, with a mild decrease in megakaryocytes (0.74% to 0.67%) (Fig. 1B, Supplementary Fig. 1F), suggesting that granulocyte–platelet homeostasis might be disrupted²⁶.

Fig. 1. Single-cell transcriptome analysis of the neutrophil–megakaryocyte axis in ITP.

A UMAP of bone-marrow scRNA-seq (ITP n = 6; HD n = 5). Each dot is a cell colored by inferred cell type (monocytes, macrophages, CD4⁺ T cells, NK cells, neutrophils, B cells, erythrocytes, CD8⁺ T cells, plasma cells, platelets, dendritic cells, megakaryocytes, hematopoietic stem cells). B Proportion of each cell type per donor in HD and ITP. C Nebulosa density maps of S100A8 and S100A9 expression across bone-marrow cells in HD and ITP from the same cohort (ITP n = 6; HD n = 5). Peripheral-blood validation: D S100A8 mRNA, E S100A9 mRNA, and F plasma S100A8/A9 by ELISA (HD n = 20; ITP n = 48). Each dot represents one individual; bars show mean ± SD. Statistical analysis: two-sided unpaired Student’s t-test (HD vs ITP). Technical replicates are detailed in Methods. Boxplots of S100A9 expression from GEO datasets: G GSE56232, H GSE46922, I GSE23754; group labels follow the original study annotations. Heatmaps of differentially expressed genes (DEGs) in (J) neutrophils and (K) megakaryocytes from the scRNA-seq cohort (ITP n = 6; HD n = 5). Rows include S100A8/S100A9 and GATA1; colors denote relative expression (z-score). L KEGG pathway enrichment in megakaryocytes from the scRNA-seq cohort; top enriched pathways include MAPK (BH-FDR < 0.05). M Network of putative ligand–receptor interactions among HSCs, neutrophils, and megakaryocytes inferred from the scRNA-seq dataset (subclusters labeled numerically; method and thresholds in Methods). N Schematic of the working model: neutrophil-derived S100A8/A9 and the candidate receptor TLR4 on megakaryocytes. O Visium spatial maps for three bone-marrow slices (BM1–3; GSE269875). Columns show H&E, module score maps for neutrophils, megakaryocytes (MKs), and S100A8/A9, and a merged overlay. Hotspots denote the top 20% of module scores. Insets report MK→neutrophil nearest-neighbor permutation results (ΔNN = NN(obs) − NN(rand), μm; one-sided p_close, n = 1,000; ΔNN < 0 indicates closer-than-random). P For each slice, bars show closeness gain = NN(rand) − NN(obs) (μm) for MKs vs neutrophils and MKs vs S100A8/A9 (positive = closer than random), with p_close (n = 1000). Right-hand dumbbells display paired NN(rand) and NN(obs). Distances were converted to μm using the Visium calibration (55 μm per spot_diameter_fullres). Statistics: For GEO comparisons in (G–I), significance was assessed using the Kruskal–Wallis test followed by Dunn’s multiple-comparisons test (as applicable to the group structure in each dataset). Differential expression and pathway enrichment results were adjusted using Benjamini–Hochberg FDR, where indicated (e.g., panel L). For spatial proximity in (O, P), one-sided label-permutation tests were used (n = 1,000 permutations), as described. Statistics: For (D–F), bars show mean ± SD; each dot represents one biologically independent donor. Statistical analysis was performed using a two-sided unpaired Student’s t-test. For (G–I), box plots show the median, 25th–75th percentiles, and whiskers extending to minimum and maximum values; statistical analysis was performed using the Kruskal–Wallis test followed by Dunn’s multiple-comparisons test or two-sided Mann–Whitney test, as appropriate to the dataset structure. For (O, P), spatial proximity was assessed using one-sided label-permutation tests with 1000 permutations; ΔNN and P values are shown in the panels. For differential expression and pathway enrichment, Benjamini–Hochberg FDR correction was applied.

On this basis, we performed differential gene expression analysis on the integrated single-cell data. Density plots generated using the Nebulosa package (Fig. 1C, Supplementary Fig. 1C) indicated significant upregulation of S100A8 and S100A9 in bone marrow cells of the ITP group, with the most pronounced increase occurring in neutrophils. This finding aligns with our observation of increased neutrophil counts, indicating that neutrophils in ITP patients not only expand in number but also upregulate key inflammatory mediators, suggesting a potentially important pathogenic role in the disease.

To validate this observation, we analyzed peripheral blood samples from 20 healthy volunteers and 48 ITP patients (whose baseline characteristics are summarized in Supplementary Table 1), finding elevated mRNA levels of S100A8 and S100A9 in ITP (Fig. 1D, E). Circulating S100A8/A9 protein levels were also significantly increased (423.3 ± 153.4 ng/mL vs. 220.5 ± 115.0 ng/mL; Fig. 1F), corroborating the single-cell results. Further investigation using bone-marrow aspirates (ITP n = 10; HD n = 10), both S100A8 and S100A9 transcripts were higher in ITP (qPCR, 2^-ΔCt normalized to GAPDH; Supplementary Fig. 1G, H). Consistently, S100A8/A9 protein in bone-marrow aspirate supernatant was higher in ITP (mean ± SD: 450.3 ± 70.1 vs 250.1 ± 59.8 ng/mL; Supplementary Fig. 1I; test as indicated). We re-analyzed three independent PBMC transcriptome cohorts spanning active, recovery, and chronic ITP (GEO: GSE56232²⁷, GSE46922²⁸, GSE23754²⁹); across cohorts, S100A9 declined during recovery relative to active disease but remained elevated in chronic ITP versus healthy controls (Fig. 1G–I). The results showed that S100A9 significantly decreased in patients during the recovery phase, whereas it remained high in chronic ITP, supporting a potential role for S100A9 in persistent or severe ITP. It also revealed a strong inverse correlation between plasma S100A8/A9 protein levels and platelet counts. Patients with higher circulating S100A8/A9 exhibited markedly lower platelet numbers. Spearman correlation analysis confirmed a significant negative association (r = −0.71, p < 0.05) (Supplementary Fig. 1J), supporting the clinical relevance of S100A8/A9 elevation in the context of ITP-associated thrombocytopenia.

A subset analysis focused on neutrophils and megakaryocytes revealed marked upregulation of S100A8 and S100A9 in ITP neutrophils, whereas the critical megakaryocyte transcription factor GATA1 was significantly downregulated (Fig. 1J, K). KEGG pathway enrichment analysis indicated a notable dysregulation of MAPK signaling in ITP megakaryocytes (Fig. 1L), suggesting that this stress-response pathway may play an important role in impaired platelet production. Further ligand–receptor interaction analyses (Fig. 1M, N) suggested that S100A8/A9 secreted by neutrophils could bind TLR4 on megakaryocytes, potentially exacerbating inflammation and hindering platelet production. Notably, western blot analysis of bone marrow samples showed no significant differences in TLR4 protein levels between the ITP group and healthy controls (Supplementary Fig. 1D, E), indicating that the abnormally high level of S100A8/A9 is consistent with ligand excess being a major contributor to megakaryocyte dysregulation in ITP. Taken together, this previously unrecognized “neutrophil–megakaryocyte” paracrine axis provides deeper molecular insights into the pathogenesis of ITP, suggesting that neutrophils can influence megakaryocyte function via the S100A8/A9-mediated TLR4 pathway, thereby disrupting the balance of platelet production.

Across three independent human bone-marrow Visium slices, neutrophil- and S100A8/A9-high regions coincide spatially with MK-high regions (Fig. 1O, columns: H&E, Neutrophils, MKs, S100A8/A9, and merged overlays). Nearest-neighbor permutation tests using top 20% hotspots showed that MK hotspots are significantly closer to neutrophil hotspots than expected by chance: ΔNN (obs − rand) = −9.2 μm (BM sample 1), −50.9 μm (BM sample 2), and −20.2 μm (BM sample 3), all p_close <0.01. MK hotspots were likewise closer to S100A8/A9 hotspots (BM sample 3, ΔNN = −19.6 μm, p_close <0.01). Figure 1P summarizes the effect sizes as closeness gain (positive across slices for both MK→Neutrophils and MK → S100A8/A9) and shows paired NN(rand)→NN(obs) differences on dumbbell plots. These spatial statistics from an external dataset independently support co-localization of neutrophils/S100A8/A9 with MKs in the bone-marrow niche.

ITP plasma impairs MK maturation via S100A8/A9–TLR4

We began by isolating CD34⁺ hematopoietic stem/progenitor cells from healthy donor bone marrow, enriching them via microbead separation and subsequently culturing in a thrombopoietin (TPO), stem cell factor (SCF), and IL-3 supplemented system³⁰ (Fig. 2A). After 4 days, the culture medium was replaced with either healthy control (HC) or ITP patient plasma to examine how disease-associated factors might shape megakaryocyte development. Notably, megakaryocytes grown in ITP bone marrow–derived supernatant displayed elevated S100A8 and S100A9 mRNA levels without any appreciable change in TLR4 transcripts (Fig. 2B–D). Concordantly, the S100A8/A9 protein concentration was markedly higher in ITP samples (Fig. 2E). These findings implicate soluble ITP-specific mediators in driving S100A8/A9 overexpression, potentially contributing to abnormal megakaryocyte maturation.

Fig. 2. In-vitro CD34⁺ culture defines S100A8/A9–TLR4–dependent effects on megakaryocyte (MK) maturation.

A In-vitro differentiation timeline. Schematic of the CD34⁺ → MK culture showing Day-4 addition of pooled HC or ITP plasma and the Day-4/Day-7 dosing of ± tasquinimod/IAXO-102 (see Methods). qPCR/ELISA measurements of S100A8 (B), S100A9 (C), TLR4 (D), and S100A8/A9 protein (E) in MK cultures exposed to HD vs ITP plasma (biological n = 3 per group). F Representative IF images of CD41 (magenta) and DAPI (blue) after 7 days (biological n = 3; 5 fields per sample). Scale bar, 50 μm. G DNA-content histograms (propidium iodide) of MK ploidy distributions (biological n = 3). Quantification of CD41 mean fluorescence intensity (H) and ≥16 N MKs (I) across indicated conditions (biological n = 3; points represent donor/culture means). J Transmission electron micrographs of MKs (biological n = 5; 10 cells analyzed per group). Scale bar, 10 μm. K PLP gating and representative frequencies. Flow cytometry plots of CD41 vs PI within the <1 µm submicron gate (FSC^low/SSC^low; see Methods). PLPs were operationally defined as CD41⁺ submicron events with PI⁻ (membrane-integrity indicator). L Early apoptosis of megakaryocytes. Annexin V/PI analysis of CD41⁺ megakaryocytes showing the proportion of early apoptotic cells (Annexin V⁺/PI⁻) under HC or ITP plasma ± tasquinimod (biological n = 3). (Annexin V⁺/PI⁺ events considered non-viable/late apoptosis and not included in the early-apoptosis metric). M Representative adhesion images: equal numbers of viable CD41⁺ MKs were seeded on fibronectin-coated plates, incubated 45 min at 37 °C, washed, and imaged in predefined fields under blinded conditions (biological n = 3). Scale bar, 10 μm. N Quantification: adherent MKs per field. Identical seeding across conditions permits direct comparison between HC and ITP (± tasquinimod). Dots denote individual fields; open circles, per-replicate means; boxes, median and IQR. Statistics were performed at the biological-replicate level (two-sided Wilcoxon for two-group comparisons; Kruskal–Wallis with Dunn’s post-hoc otherwise); p values are indicated. O Transwell migration assay showing diminished migratory capacity of megakaryocytes under ITP conditions, which is improved by tasquinimod. (biological n = 3). Scale bar, 10 μm. Statistics: For (B–E), two-sided unpaired Student’s t-test (HD vs ITP). For multi-group comparisons in (H), (I), (L) and (O), one-way ANOVA with Tukey’s multiple-comparisons test. For (N), statistical testing was performed at the biological-replicate level as described (two-sided Wilcoxon rank-sum test for two-group comparisons; Kruskal–Wallis with Dunn’s multiple-comparisons test otherwise). n and error bars are as indicated in each panel. Statistics: For (B–E), bars show mean ± SD; n = 3 biologically independent cultures per group; two-sided unpaired Student’s t-test. For (H), bars show mean ± SEM with individual independent cultures overlaid; n = 3 independent differentiations; one-way ANOVA with Tukey’s multiple-comparisons test. For (I), grouped bars show independent DNA-content peak-window gate frequencies; n = 3 independent differentiations; high-ploidy fractions were compared using one-way ANOVA with Tukey’s multiple-comparisons test. For (L), bars show mean ± SEM; n = 3 independent differentiations; one-way ANOVA with Tukey’s multiple-comparisons test. For (N), bars show mean ± SEM; n = 3 independent cultures; one-way ANOVA with Tukey’s multiple-comparisons test.

To dissect the functional impact of the S100A8/A9–TLR4 axis, we supplemented both HC- and ITP-plasma cultures with tasquinimod (a quinoline-3-carboxamide reported to bind S100A9 and disrupt S100A8/A9 receptor engagement) or IAXO-102 (a TLR4 inhibitor). Megakaryocytes exposed to ITP plasma alone exhibited significantly reduced CD41 (GPIIb) staining (Fig. 2F, H), an indicator of megakaryocytes. Both tasquinimod and IAXO-102 substantially restored CD41 intensity, suggesting that inhibiting S100A8/A9 or TLR4 pathways can ameliorate ITP-related defects in megakaryopoiesis. Consistent with this observation, ploidy analysis further revealed that ITP plasma increased the proportion of 2 N cells while reducing higher ploidy (16 N), but these abnormalities were reversed upon tasquinimod or IAXO-102 treatment (Fig. 2G, I). In parallel, surface marker analyses demonstrated that ITP plasma downregulated CD41, CD42b, and CD61 expression, as well as the double-positive CD41⁺CD42b⁺ fraction, whereas tasquinimod treatment or TLR4 restored marker-high populations (Supplementary Fig 2A, S2E). Transmission EM of CD34⁺-derived MKs exposed to ITP plasma showed a poorly developed demarcation membrane system (DMS) with reduced α-granules. Addition of tasquinimod restored ultrastructural hallmarks of maturation, including expanded DMS and increased α-granule content (Fig. 2J), consistent with quantitative analyses in Supplementary Fig. 2Q, R.

Functionally, we found that ITP plasma markedly depressed platelet-like particle (PLP) release and diminished megakaryocyte apoptotic regulation. To address the potential interference from nucleated contaminants in PLP gating, we incorporated PI staining into the CD41 flow-cytometry panel to exclude membrane-compromised events and debris. Representative plots (Fig. 2K) showed that, under HC plasma conditions, the proportion of CD41⁺ PI⁻ events were 65.5% in the DMSO group and increased to 72.9% and 70.9% with tasquinimod and IAXO-102, respectively. In ITP plasma cultures, this fraction dropped to 49.9% but was significantly rescued to 61.9% and 63.1% by tasquinimod and IAXO-102, a trend confirmed across independent replicates (Supplementary Fig. 2G). These data corroborate our earlier CD41-only measurements and confirm that the observed effects are not attributable to nucleated debris or apoptotic fragments. In parallel, Annexin V/PI staining (early apoptosis defined as Annexin V⁺/PI⁻, i.e., phosphatidylserine exposure with intact membrane integrity) showed that ITP plasma reduced the proportion of CD41⁺ MKs in the early-apoptotic compartment (Fig. 2L). We therefore interpret this as altered apoptotic regulation in MKs under ITP plasma conditions rather than a measure of overall cell death, whereas tasquinimod treatment restored early apoptosis to near-physiological levels. This observation is consistent with reports that a subset of ITP plasmas suppresses MK apoptosis in vitro³⁰, while other studies have reported different patterns depending on cohorts and assays. Together, these results indicate that blocking the S100A8/A9–TLR4 axis not only rescues anucleate PLP production but also normalizes apoptotic regulation in megakaryocytes under ITP conditions. Moreover, ITP plasma also impaired integrin-mediated processes, reducing both adhesion to the culture substrate (Fig. 2M, N) and transwell migration (Figs. 2O,  S2F). In each case, tasquinimod counteracted the disease-associated deficits, affirming that S100A8/A9 is a critical driver of megakaryocyte dysfunction in ITP. Taken together, our data highlight how aberrant S100A8/A9–TLR4 signaling contributes to multiple facets of megakaryocyte pathology—ranging from reduced surface marker expression and impaired high-ploidy attainment to diminished platelet production, altered apoptosis, adhesion, and migration. These findings underscore the potential therapeutic value of targeting the S100A8/A9–TLR4 axis to normalize megakaryocyte function in ITP. We next examined the impact of graded rS100A8/A9 or Tasquinimod concentrations on megakaryocyte (MK) maturation. Flow cytometric DNA ploidy analysis revealed a progressive reduction in high-ploidy MKs with increasing rS100A8/A9 concentrations, with the most pronounced shift at ≥0.5 µg/mL (Supplementary Fig. 2H, S). Consistent with this, surface CD41 expression decreased in a dose-dependent manner, reaching a plateau at ≥0.5 µg/mL (Supplementary Fig. 2I, K, T). Conversely, Tasquinimod treatment of ITP plasma–exposed MKs restored CD41 expression in a concentration-dependent manner, with maximal effect achieved at ≥ 1 µM (Supplementary Fig. 2J, L, U). These data define functional thresholds for rS100A8/A9 (0.5 µg/mL) and Tasquinimod (1 µM) beyond which further changes in MK maturation marker expression are not observed.

ITP plasma impaired MK maturation, whereas selective removal of S100A8/A9 partially restored it. DNA-content histograms revealed a left-shift of ploidy under ITP plasma, with enrichment of 2 N/4 N and loss of high-ploidy ( ≥ 8 N) peaks compared with HC plasma. S100A8/A9 immunodepletion of ITP plasma recovered higher-ploidy populations toward the HC pattern (Supplementary Fig. 2M, V). Surface CD41 was reduced by ITP plasma and increased after S100A8/A9 removal, as shown by overlaid CD41-APC histograms of live MKs (Supplementary Fig. 2N, W). Functionally, ITP plasma lowered PLP generation, whereas S100A8/A9 immunodepletion rescued PLP output (representative bivariate plots in Supplementary Fig. 2O; quantification in Supplementary Fig. 2P, n = 3 independent cultures). Together, these data support S100A8/A9 as a key inhibitory signal in ITP plasma that constrains MK polyploidization, lineage maturation, and platelet-like particle release.

ITP plasma diverts MK maturation trajectories

We first leveraged an in vitro culture system to obtain enough megakaryocytes (MKs) for single-cell RNA sequencing, as primary ITP bone marrow often yields low MK counts (Figs. 3A, S3A). Cultures were supplemented with equal-volume pooled plasma from 10 ITP donors or equal-volume pooled plasma from 10 healthy donors (HD); a tasquinimod arm was included where indicated. Guided by established criteria³¹, we annotated five MK subsets: Immature MK, Cycling MK, Niche-supporting MK, Immune MK, and Thrombopoiesis-biased (T-B) MK. Under ITP plasma, Immature and Cycling MK populations expanded at the expense of T-B MK (Fig. 3B).

Fig. 3. Single-cell transcriptomics reveals distinct megakaryocyte (MK) subsets and impaired differentiation under ITP plasma.

A UMAP plots of in vitro–generated megakaryocytes (MKs) cultured with equal-volume pooled HD plasma (10 donors), equal-volume pooled ITP plasma (10 donors), or ITP plasma plus tasquinimod (right), annotated into five subsets: Immature MK, Cycling MK, Niche-supporting MK, Immune MK, and Thrombopoiesis-biased (T-B) MK. B Proportions of MK subsets across conditions. C Volcano plots of subset/condition differential expression with selected genes labeled; testing framework and thresholds are described in Methods. D RNA-velocity streamlines overlaid on UMAP for each condition. E Pseudotime density distributions shown per subset and condition. F Pseudotime-ordered heatmap of representative MK markers (e.g., CD34, GATA1, GP1BA, ITGA2B, ITGB3, S100A8, S100A9). G Dot plots of selected MK genes across subsets and conditions. H KEGG pathway enrichment within MK subsets; MAPK pathway is among the enriched terms (BH-FDR < 0.05). Statistics: Differential expression analyses in (C) were performed using Wilcoxon rank-sum tests with Benjamini–Hochberg FDR correction (thresholds as indicated). KEGG/pathway enrichment in (H) was assessed using BH-FDR (<0.05). Statistics: For (C), differentially expressed genes were identified using Wilcoxon rank-sum tests with Benjamini–Hochberg FDR correction; adjusted P values are reported where applicable. For RNA-velocity transition analyses in (D, E), uncertainty was estimated by bootstrap resampling and significance by label-permutation tests, as described in Methods. For GSEA in (H), pathway enrichment significance was assessed using Benjamini–Hochberg FDR correction; NES and FDR q values are shown. No additional inferential statistics were applied to descriptive UMAP, pseudotime-density, or heatmap panels.

Cluster-level differential expression further revealed that, in MEP (megakaryocyte–erythroid progenitor), Immature MK, and T-B MK subsets, S100A8/A9 transcripts were lower despite elevated extracellular S100A8/A9, consistent with—but not proving—a negative feedback response (Fig. 3C). Representative pseudo-bulk contrasts (ITP vs HD) gave log₂FC(S100A8) = − 5.6 (q = 0.012) and log₂FC(S100A9) = − 5.7 (q = 0.019) in MEP; and −5.5 (q = 0.049) and −5.6 (q = 0.058) in T-B MK. In parallel, PF4 was reduced in Immature MK (log₂FC = − 5.5, q = 0.008), whereas GATA1 was diminished in Cycling MK (log₂FC = − 2.4, q = 0.022) but remained high in T-B MK, indicating limited platelet-release capacity in the Cycling compartment and a poised thrombopoietic state in T-B MK^32,33.

RNA velocity analysis corroborated these patterns and provided directionality metrics (Supplementary Fig. 3B, D). Under ITP, the probability of Immature→Cycling transitions increased to 0.46 (bootstrap 95% CI 0.41–0.50) versus 0.28 (0.24–0.33) under HD (Δ=+0.18; permutation p = 0.004), whereas Immature→T-B transitions decreased to 0.12 (0.09–0.15) versus 0.31 (0.26–0.35) under HD (Δ=−0.19; permutation p = 0.003). Tasquinimod shifted these flows toward maturation—Immature→T-B = 0.27 (0.23–0.31) and Immature→Cycling = 0.30 (0.26–0.34)—with deltas relative to ITP of +0.15 (0.09–0.21) and −0.16 (−0.22 to −0.09), respectively (permutation p ≤ 0.006). Consistently, the unspliced RNA fraction decreased with tasquinimod (0.35 ± 0.02 vs 0.42 ± 0.03 under ITP; batch-level paired t-test p = 0.011), indicating reduced transcriptional stress and a shift toward maturation. Although the overall proportion of T-B MKs changed only modestly under ITP, trajectory-level analyses revealed a maturation detour: RNA velocity showed reduced Immature→T-B flux and increased Immature→Cycling flux (Δ=−0.19 and +0.18; permutation p ≤ 0.006), and pseudotime mapping indicated fewer cells reaching late T-B states. T-B MKs retained a poised thrombopoietic signature (higher GATA1/ITGA2B), and the global unspliced-RNA fraction decreased with tasquinimod, together indicating impaired progression that is not captured by proportions alone.

Pseudotime mapping underscored the impaired trajectory under ITP (Fig. 3E, Supplementary Fig. 3C), with Immature MKs preferentially diverted to a Cycling fate and fewer cells reaching late T-B states. Marker dynamics aligned with maturation status: S100A8/A9 peaked mid-pseudotime and declined in late MKs, while GATA1 and ITGA2B (CD41) increased along paths that successfully reached a mature phenotype (Fig. 3F, G).

Finally, pathway analysis provided mechanistic support. In Immature MK under ITP, MAPK signaling was enriched by GSEA with NES = 2.10, FDR q = 0.004 (Fig. 3H), with core genes including FOS, JUN, DUSP1, MAP3K8. Tasquinimod reduced MAPK involvement (ITP+tasquinimod vs ITP: NES = 1.08, FDR q = 0.27), serving as functional “reverse” evidence consistent with the proposed chain (blocking S100A8/A9–TLR4 → dampened aberrant MAPK → partial restoration of maturation). Directionally, these single-cell results are concordant with bone-marrow spatial analyses from public datasets (neutrophil–MK proximity and elevated local S100A8/A9 module scores near MK neighborhoods).

Taken together, the batch-level reproducibility with confidence intervals, the cautious negative-feedback signature of S100A8/A9 transcripts, the quantified velocity directionality with uncertainty, and the NES/FDR-resolved pathway shifts—together with pharmacological reversal by tasquinimod—support a model in which S100A8/A9-driven signaling perturbs MK maturation and reduces platelet-producing capacity in ITP.

JNK/c-Jun represses GATA1 without JAK-STAT activation

To clarify how S100A8/A9–TLR4 signaling disrupts megakaryocyte (MK) homeostasis in ITP, we first examined the MAPK pathway branches p38 and JNK^34,35. Western blot analyses revealed no appreciable differences in TLR4/β-actin protein levels or phospho-p38/p38 ratios across tested conditions (Fig. 4A–C). By contrast, ITP plasma or exogenous recombinant S100A8/A9 significantly elevated phospho-JNK/JNK (Fig. 4D) and phospho-c-Jun/c-Jun (Fig. 4E), aligning with a more pronounced activation of the JNK/c-Jun pathway. Concomitantly, GATA1 expression was markedly diminished (Fig. 4F), indicative of impaired MK maturation³⁶. Notably, Tasquinimod abrogated these disease-associated changes by lowering JNK/c-Jun phosphorylation and restoring GATA1 levels (Fig. 4G). Collectively, these data underscore the significance of JNK/c-Jun, rather than p38, in mediating MK dysfunction downstream of the S100A8/A9–TLR4 axis in ITP. To further assess potential involvement of the JAK-STAT pathway, we examined the mRNA expression of three canonical STAT target genes, SOCS1, BCL-XL, and PIM1, in megakaryocytes from ITP patients and healthy controls (HC) (Supplementary Fig. 4D–F). No significant differences were detected between ITP and HC in any of these genes, consistent with the absence of measurable JAK-STAT activation in our phospho-flow and western blot assays. For analysis of p-JNK dose–response and wash-out, rS100A8/A9 induced a concentration-dependent increase in p-JNK signal (Thr183/Tyr185) in MKs, plateauing at ≥1 µg/mL. In the wash-out group, p-JNK levels largely returned to baseline after 24 h in fresh medium, indicating that S100A8/A9-induced JNK activation is reversible (Supplementary Fig. 4G).

Fig. 4. S100A8/A9–TLR4 activates JNK/c-Jun and represses GATA1 in megakaryocytes.

A Representative immunoblots for TLR4, phospho-p38/total p38, phospho-JNK/total JNK, phospho-c-Jun/total c-Jun, and GATA1 in MK cultures exposed to HC or ITP plasma, with/without recombinant S100A8/A9 and/or tasquinimod (biological n = 3). Quantification of immunoblots: B TLR4/β-actin; C p-p38/p38; D p-JNK/JNK; E p-c-Jun/c-Jun; F GATA1/β-actin; G composite index GATA1/p-c-Jun across conditions (biological n = 3). H c-Jun Cut&Tag transcription start site (TSS) profile and heatmap showing strong enrichment around TSSs. I Genomic annotation of c-Jun Cut&Tag peaks (n = 3932) by feature class (promoter, UTRs, exons/introns, intergenic, etc.). J IGV tracks at the GATA1 locus displaying c-Jun occupancy in Meg-01 (c-Jun vs IgG control). K HOMER motif analysis highlighting a GATA-like motif enriched in c-Jun peaks (similarity score and P value indicated). L GATA1 promoter–driven luciferase activity with/without c-Jun overexpression (OE); values normalized to Renilla and vector control. (M) qPCR of GATA1 mRNA following c-Jun OE or shRNA knockdown (sh#1, sh#2) in the indicated system. N Immunoblots for phospho-c-Jun, c-Jun, and GATA1 after OE or knockdown. Densitometry of O phospho-c-Jun/β-actin and P GATA1/β-actin corresponding to N (biological n = 3). Q Map of the GATA1 promoter (approx. −1.5 kb to +0.3 kb) indicating three predicted high-scoring c-Jun binding sites (sites 1–3) selected for perturbation. R In-silico promoter predictions (Basenji) showing Δ predicted GATA1 expression after single or combinatorial disruption of c-Jun sites 1–3. Bars show mean ± SEM with individual biological replicates overlaid. Statistics: For immunoblot quantifications (B–G) and densitometry in (O, P), one-way ANOVA with Tukey’s multiple-comparisons test. For luciferase in (L), two-sided unpaired Student’s t-test (vector vs c-Jun OE). For qPCR in (M) and related comparisons across ≥3 groups, one-way ANOVA with Tukey’s multiple-comparisons test. n and error bars are as indicated. Statistics: For western blot densitometry in (B–G) and (O, P), bars show mean ± SD from n = 3 biologically independent experiments; statistical analysis was performed using one-way ANOVA with Tukey’s multiple-comparisons test unless otherwise indicated. For reporter and qPCR assays in (L–M), bars show mean ± SD from n = 3 independent experiments; statistical analysis was performed using one-way ANOVA with Tukey’s multiple-comparisons test. For predicted-expression analysis in (R), bars show mean ± SD from n = [X] independent model predictions/replicate simulations, and comparisons were performed using [insert exact test]. Exact P values are provided in Source Data. For Cut&Tag peak annotation, motif, and IGV panels (H–K), statistical metrics are shown in the panels or described in Methods.

To further delineate the role of c-Jun in regulating MK gene networks, we performed Cut&Tag analysis in Meg-01 cells using a c-Jun–specific antibody. The resulting TSS plot and heatmap (Fig. 4H) demonstrated robust c-Jun enrichment around transcription start sites, whereas broader annotation (Fig. 4I) revealed that most of the 3,932 peaks were located near promoters (1–2 kb upstream of the TSS) or in distal intergenic regions—consistent with c-Jun’s promoter-binding function. Parallel RNA-seq of in vitro–cultured MKs (exposed to ITP vs. healthy control plasma) showed reduced GATA1 transcript levels and heightened S100A9 expression in ITP MKs (Fig. S4A). Importantly, intersecting these differentially expressed genes with c-Jun Cut&Tag peaks confirmed that GATA1 fell within the overlapping set (Fig. S4B). Further KEGG pathway enrichment revealed that the MAPK pathway was significantly upregulated in the ITP group (Fig. S4C), implicating MAPK/c-Jun–driven transcriptional reprogramming in ITP-associated MK pathology. IGV visualization of Cut&Tag peaks (Fig. 4J) further pinpointed robust c-Jun occupancy within the GATA1 promoter region, and HOMER motif analysis indicated an 83% similarity (p < 0.001) to the verified GATA1 binding motif (Fig. 4K). Functional assays demonstrated that overexpressing c-Jun or co-transfecting a c-Jun expression vector diminished GATA1 promoter–driven luciferase activity (Fig. 4L), whereas RNA interference–mediated knockdown of c-Jun elevated GATA1 transcription (Fig. 4M). Western blot assays confirmed these effects at the protein level: enhanced c-Jun expression suppressed GATA1, whereas c-Jun knockdown rescued it (Fig. 4N–P). Using the JASPAR database³⁷, we identified multiple potential c-Jun binding sites within the GATA1 promoter (see Supplementary Information), then selected the three with the highest scores (labeled 1, 2, and 3) for further analysis (Fig. 4Q). By leveraging the deep learning model Basenji³⁸, we performed in silico “knockouts” of three high-score c-Jun binding sites. Disrupting sites 2 and 3 significantly elevated the predicted GATA1 expression, while site 1 had a milder effect (Fig. 4R). Taken together, these findings reveal that the S100A8/A9–TLR4–JNK/c-Jun cascade site-selective occupancy within the GATA1 regulatory region, thereby impairing MK maturation and exacerbating thrombocytopenia in ITP.

TLR4 mediates ITP plasma–induced MK defects

To complement pharmacologic inhibition and test receptor specificity, we performed siRNA-mediated knockdown of TLR4 (siTLR4) or AGER/RAGE (siRAGE) in primary CD34⁺-derived megakaryocytes cultured with HC or ITP plasma. Flow cytometry showed that ITP plasma reduced the frequency of CD41⁺ and CD61⁺ cells, which was partially rescued by siTLR4 but not by siRAGE (Supplementary Fig. 4I–L). Consistently, PI DNA-content analysis revealed an ITP plasma–induced left shift in ploidy distribution (enrichment of 2 N/4 N with depletion of high-ploidy populations), and siTLR4 restored polyploidization toward the HC pattern, whereas siRAGE did not rescue (Supplementary Fig. 4M, N). At the signaling level, siTLR4 reduced p-cJun/c-Jun and p-JNK/JNK and increased GATA1 abundance under ITP plasma conditions (Supplementary Fig. 4O, P), supporting a predominantly TLR4-dependent pathway in this primary MK system.

Because the genome-wide c-Jun occupancy map was generated in MEG-01 cells, we next asked whether c-Jun binding within the GATA1 regulatory region could also be detected in a primary human megakaryocyte system. We performed c-Jun ChIP–qPCR in primary CD34⁺-derived megakaryocytes cultured with pooled healthy-control plasma, pooled ITP plasma, or ITP plasma plus tasquinimod. At the originally interrogated GATA1 amplicon, c-Jun signal was detectable above the pooled IgG background, whereas the GAPDH control locus did not show selective enrichment over its corresponding IgG reference. However, c-Jun occupancy at this single amplicon was reduced under ITP plasma and restored toward baseline by tasquinimod, indicating that the originally tested region did not capture a uniform ITP-associated increase across the GATA1 promoter region (Supplementary Fig. 4Q). We therefore extended the analysis to three predicted c-Jun-binding regions within the GATA1 regulatory region in primary megakaryocytes. Notably, c-Jun occupancy showed a site-selective pattern rather than a uniform change across the locus: Site2 displayed the strongest ITP-associated enrichment and was reversed by tasquinimod, whereas Site1 showed only modest changes and Site3 remained largely unchanged (Supplementary Fig. 4R). Together, these data support a model in which S100A8/A9–TLR4–JNK signaling represses GATA1 through preferential c-Jun occupancy at a discrete regulatory site, rather than through uniformly increased binding at every promoter-proximal amplicon.

Tasquinimod rescues MKs in active ITP

Active immune thrombocytopenia (ITP) was modeled in C57BL/6 RAG1^−/− mice via adoptive transfer of autoreactive splenocytes isolated from Itgb3a^−/− donors (Fig. 5A). Repeated platelet transfusions to Itgb3a^−/− mice induced high-titer anti-platelet autoantibodies³⁹, which, upon transfer into irradiated RAG1^−/− recipients, led to a pronounced and persistent thrombocytopenia resembling human ITP. Monitoring daily platelet counts revealed a substantial, prolonged decrease in the ITP group relative to irradiated-only controls. By contrast, administration of Tasquinimod accelerated platelet recovery in both ITP and irradiation-only mice (Fig. 5B), indicating that S100A8/A9 antagonism can mitigate thrombocytopenia in this active ITP context.

Fig. 5. In-vivo readouts of platelet counts, MK histology, and single-cell profiles in the active ITP model.

A Experimental schematic: active ITP induced by transfer of anti-platelet splenocytes (Itgb3a^−/− donors) into irradiated RAG1^−/− recipients; tasquinimod dosing as indicated. B Platelet counts over time for indicated groups (biological n per group as above). C Representative CD41/DAPI immunofluorescence staining of bone-marrow sections from the indicated groups. CD41 staining is shown in magenta and DAPI nuclear counterstaining in blue. White arrows indicate megakaryocytes. Scale bar, 50 µm. D Representative H&E staining of bone-marrow and spleen sections from the indicated groups. Black arrows indicate megakaryocytes. Scale bars, 50 µm. E Bone-marrow MKs per field (biological n = 3; 5 fields/mouse). F Ratio of bone-marrow to spleen MK counts per mouse (biological n = 3). G Plasma S100A8/A9 concentrations by ELISA at Day 7 (biological n = 10; 3 technical wells/sample). H White blood cell counts at Day 7 (biological n = 10). I Collagen-induced platelet aggregation traces (PRP/whole blood as indicated; biological n = 3; 3 technical runs/sample). J Hemoglobin levels at Day 7 (biological n = 10). K Tail bleeding time (biological n = 10). Assay conditions are described in Methods (temperature, cutoff time). L UMAPs of bone-marrow scRNA-seq from ITP and ITP + tasquinimod groups. M Fraction of MKs among bone-marrow cells: ITP, 0.75%; ITP + tasquinimod, 1.07% (computed at the cell level; group summaries at the mouse level). N Gene set enrichment analysis (GSEA) comparing MKs from ITP + tasquinimod vs ITP; MAPK pathway among the terms with FDR q < 0.05. Statistics: For platelet-count time courses in (B), two-way repeated-measures ANOVA with Sidak’s multiple-comparisons test. For single–time-point multi-group comparisons in (E–H) and (J), one-way ANOVA with Tukey’s multiple-comparisons test. For tail bleeding time in (K), Kruskal–Wallis test with Dunn’s multiple-comparisons test. For MK fraction comparisons in (M), two-sided unpaired Student’s t-test at the mouse level. GSEA significance in (N) was assessed using BH-FDR (q < 0.05). n and error bars are as indicated. Statistics: For platelet-count time courses in (B), data are mean ± SEM; n = 10 biologically independent mice per group; two-way repeated-measures ANOVA with Sidak’s multiple-comparisons test. For (E–H, J), bars show mean ± SEM with individual mice overlaid; n = 3 mice per group for histology and n = 10 mice per group for plasma/CBC measurements, as indicated in Source Data; one-way ANOVA with Tukey’s multiple-comparisons test. For tail bleeding in (K), bars show mean ± SEM with individual mice overlaid; n = 10 mice per group; Kruskal–Wallis test with Dunn’s multiple-comparisons test. Exact P values are provided in the Source Data file.

Subsequent histological assessments demonstrated markedly reduced megakaryocyte (MK) counts in the bone marrow (BM) of ITP mice, while Tasquinimod restored MK density to near-normal levels (Fig. 5C, D). Quantitative analyses (Figs. 5E, F) further confirmed enhanced MK numbers in the BM and attenuated extramedullary hematopoiesis in the spleen. Concurrently, plasma S100A8/A9 concentrations were significantly elevated in ITP mice but normalized upon Tasquinimod treatment (Fig. 5G), whereas leukocyte (WBC) counts, and hemoglobin (Hb) levels remained largely unaffected (Figs. 5H, J). In addition, the tail-bleeding assay (Fig. 5K) demonstrated that ITP mice exhibited prolonged hemostatic times—an indicator of compromised platelet function—yet these abnormalities were reversed by Tasquinimod.

Interestingly, collagen-induced platelet aggregation assays (Fig. 5I) showed no appreciable differences between ITP and control groups across a range of collagen concentrations, implying that platelet reactivity to this stimulus remained intact. Clot retraction (Supplementary Fig. 5M) and platelet spreading (Supplementary Fig. 5N) were also unaltered, suggesting that contractile and adhesive functions of platelets were preserved despite thrombocytopenia. Finally, scRNA-seq of BM cells from ITP mice before and after Tasquinimod treatment (Fig. 5L) revealed an increased megakaryocyte fraction (from 0.75% to 1.07%) in Tasquinimod-treated mice (Fig. 5M), aligning with the observed improvement in platelet counts. Gene set enrichment analysis (GSEA) showed significant downregulation of MAPK signaling in megakaryocytes (Fig. 5N), echoing previous reports that inhibiting S100A8/A9–TLR4–MAPK pathways can foster MK maturation and augment platelet production.

Tasquinimod mitigates passive ITP

We also established a passive ITP model in wild-type C57BL/6 mice by injecting an anti-mouse CD42b antibody (0.1 µg/g, intravenous) on Day 0 and Day 1, thereby targeting platelet surface glycoproteins and inducing acute thrombocytopenia. Beginning on Day −1, platelet counts were measured daily to monitor disease progression. Tasquinimod (1 mg/kg) was administered on Days 1, 2, and 3, and mice were observed until Day 17. This experimental timeline provided an opportunity to evaluate whether Tasquinimod could counteract platelet destruction mediated by anti-GPIbα antibodies.

Platelet counts revealed a stark contrast between the passive ITP group and irradiated-only controls (Supplementary Fig. 5B). ITP mice experienced a significantly delayed recovery in thrombocytopenia, consistent with heightened antibody-mediated platelet clearance. We next tested whether JAK-STAT inhibition affects platelet recovery in ITP. In both active (Supplementary Fig. 5C) and passive (Supplementary Fig. 5D) ITP mouse models, ruxolitinib did not accelerate platelet count restoration, with recovery curves overlapping untreated ITP controls. These findings align with our signaling data showing no detectable JAK-STAT activation in ITP megakaryocytes.

To clarify the role of neutrophils and S100A8/A9 in platelet recovery, we applied rS100A8/A9 supplementation and/or anti-Ly6G–mediated neutrophil depletion in active and passive ITP mouse models. In active ITP mice, rS100A8/A9 administration further suppressed platelet counts and delayed recovery compared to ITP controls, whereas anti-Ly6G treatment moderately improved platelet rebound (Supplementary Fig. 5E). In passive ITP mice, rS100A8/A9 also prolonged thrombocytopenia, and neutrophil depletion partially reversed this effect (Supplementary Fig. 5F). Bone marrow histology showed that rS100A8/A9 reduced MK numbers, while anti-Ly6G restored MK density toward baseline. These findings support a pathogenic role for neutrophil-derived S100A8/A9 in impairing MK maturation and delaying platelet recovery in both active and passive ITP.

Notably, Tasquinimod-treated ITP mice displayed a marked rebound in platelet counts, suggesting that inhibiting the S100A8/A9–TLR4 axis may restore platelet homeostasis under passive ITP conditions. Histological analyses of bone marrow (BM) sections by CD41 immunofluorescence (DAPI nuclear counterstain) and H&E staining (Supplementary Fig. 5G, H) showed a marked reduction in megakaryocyte (MK) numbers in untreated ITP mice, which was effectively reversed by tasquinimod. These qualitative observations were corroborated by quantitative readouts—BM MK counts and the BM-to-spleen MK ratio (Supplementary Fig. 5I, J). Moreover, MKs visible in the spleens of ITP mice (white arrows) indicated extramedullary hematopoiesis, which was mitigated with tasquinimod treatment.

Interestingly, in contrast to findings from the active ITP model, plasma S100A8/A9 levels did not differ significantly among the experimental groups in this passive ITP setting (Supplementary Fig. 5K). Nonetheless, prolonged tail-bleeding times in ITP mice (Supplementary Fig. 5L) signaled impaired hemostatic function, which was normalized by Tasquinimod treatment. These observations underscore that even if circulating S100A8/A9 levels remain unchanged, pharmacological blockade of the S100A8/A9–TLR4 pathway can ameliorate platelet destruction and restore effective clot formation in passive ITP.

Discussion

Our data define a bone-marrow “neutrophil–megakaryocyte” axis in immune thrombocytopenia (ITP) whereby neutrophil-derived S100A8/A9 engages TLR4 on megakaryocytes (MKs), activates a JNK/c-Jun cascade that represses GATA1, and constrains thrombopoiesis (Supplementary Fig. 6). Single-cell RNA-seq reveals enrichment of neutrophils with a reduced MK proportion, and both CD34⁺-derived MK cultures and active/passive ITP mouse models converge on the same mechanism. Pharmacologic blockade—using tasquinimod or TLR4 inhibitors—partially restores platelet production, underscoring therapeutic tractability within the broader immunoinflammatory framework of ITP^40,41.

S100A8/A9 shapes inflammatory microenvironments⁴², yet its roles in megakaryopoiesis and thrombopoiesis have been insufficiently explored^43,44. Our data support that S100A8/A9 signals through TLR4^45,46. By examining S100A8/A9 in the context of ITP, our study reveals that these proinflammatory alarmins actively disrupt megakaryocyte maturation rather than merely playing a passive role. We further demonstrate that overactivated neutrophils release S100A8/A9 to promote TLR4-dependent paracrine effects, impairing megakaryocyte adhesion and migration. Moreover, S100A8/A9-mediated dysregulation functions within a broader “inflammatory–thrombopoietic” loop: inflammatory mediators induce abnormal megakaryocyte maturation, whereas platelets and megakaryocytic cells themselves can release or sense inflammatory signals. While S100A8/A9 has been implicated in cardiovascular disease sepsis^47,48 and sepsis^49–51, our current findings elucidate the underlying mechanisms by which these alarmins modulate megakaryocyte biology in an autoimmune setting.

Prior work has reported heterogeneous effects of ITP plasma/sera on MK apoptosis depending on patient subsets, MK maturation stage, and readouts. For example, Yang et al. reported that many ITP plasmas reduced Annexin V/PI-defined MK apoptosis and were associated with altered TRAIL/Bcl-xL signaling, whereas ultrastructural analyses of bone marrow biopsies identified MKs with apoptotic/para-apoptotic features in ITP³⁰. Other in-vitro studies observed minimal changes in phosphatidylserine exposure under certain conditions⁵². In this context, our findings support altered apoptotic regulation in MKs under ITP plasma conditions in our experimental system.

From a therapeutic standpoint, our finding that S100A8/A9 blockade can partially restore megakaryocyte function opens new avenues for mitigating pathologic thrombocytopenia. These findings establish ~0.5 µg/mL as the in-vitro functional threshold for S100A8/A9, above which MK maturation defects plateau, and ~1 µM as the effective rescue threshold for tasquinimod. Prior work shows that tasquinimod binds S100A9 with high affinity and, in the presence of divalent cations such as Zn²⁺, prevents S100A8/A9 from engaging its immune receptors (notably TLR4/RAGE)⁵³. Clinically, plasma S100A8/A9 ≥ 0.5 µg/mL predicts significant thrombopoietic failure, aligning the mechanistic threshold with a clinically actionable cut-point. Together, these data suggest that therapies targeting S100A8/A9 or its downstream signaling could be prioritized in patients exceeding this threshold, enabling biomarker-guided intervention strategies. Preserved platelet aggregation, clot retraction, and spreading in ITP indicate that contractile and adhesive functions remain intact, arguing against increased consumption as the primary cause of thrombocytopenia. These findings, together with the megakaryocyte maturation defects observed, support impaired platelet production as the dominant mechanism. Although TLR4 antagonists have largely been explored in sepsis and systemic inflammatory disorders, our data suggest that small-molecule inhibitors or monoclonal antibodies targeting S100A8/A9 may correct specific hematopoietic deficits in ITP. Future studies will be needed to clarify how S100A8/A9-mediated TLR4 signaling integrates with other bone marrow inflammatory pathways, and to assess whether these therapies can be safely translated into clinical practice.

While tasquinimod is widely used as a pharmacologic probe for S100A9/S100A8/A9 signaling and has been reported to disrupt S100A9/A8/A9 engagement with receptors such as TLR4 and RAGE, it also exhibits pleiotropic activities, including broader immunomodulatory effects and reported interactions with additional pathways (e.g., HDAC4 binding and thrombospondin-1 regulation)⁵⁴. Accordingly, we interpret the tasquinimod data as supportive pharmacologic evidence consistent with involvement of the S100A8/A9–TLR4 axis, rather than definitive proof of ligand- or cell-type specificity. We therefore place greater weight on orthogonal, megakaryocyte-focused evidence in this study—ligand gain- and loss-of-function (recombinant S100A8/A9 and immunodepletion), phenocopy by TLR4 inhibition, and concordant attenuation of downstream JNK/c-Jun signaling across readouts. Future work using receptor-specific blocking antibodies and/or genetic perturbation in megakaryocytes will further refine receptor specificity and help disentangle cell-intrinsic from systemic contributions.

Western blot and Cut&Tag experiments revealed that the S100A8/A9–TLR4 axis predominantly signals through the JNK/c-Jun branch of the MAPK pathway, whereas p38 activity remains largely unchanged. Specifically, consistent with c-Jun-dependent repression of GATA1, it causes a deficiency in megakaryocyte (MK) maturation and impairs platelet production. This mechanism aligns with previous observations that c-Jun–mediated transcriptional reprogramming can inhibit hematopoietic cell differentiation^55–57, thereby connecting excessive S100A8/A9–TLR4 signaling to a critical bottleneck in thrombopoiesis. Our findings complement prior single-cell studies in ITP that focused on upstream hematopoiesis. Liu et al. performed scRNA-seq of bone-marrow CD34⁺ HSPCs and showed that gene expression programs, cell–cell interaction features, and transcriptional regulatory networks were altered in ITP HSPCs, with prominent changes in immune progenitor compartments. They further reported a reduction of CD9⁺ and HES1⁺ cells within Lin⁻CD34⁺CD45RA⁻ HSPCs by flow cytometry and demonstrated that the megakaryocytic differentiation propensity of CD9⁺ HSPCs observed in healthy controls was diminished in ITP. Moreover, co-culture with pre-B cells from ITP patients impaired the differentiation of CD9⁺ HSPCs toward the megakaryocytic lineage, and their analyses subdivided megakaryocytic progenitors into multiple subclusters with ITP-associated DEGs enriched in a subset featuring dual immunomodulatory and platelet-generation functions¹⁶. While this work highlights progenitor-level contributions to thrombopoietic defects, our study interrogates the whole bone-marrow ecosystem (including mature neutrophils and megakaryocytes) and supported by spatial co-localization and functional perturbations, pinpoints neutrophil-derived S100A8/A9 as an inflammatory cue that impairs MK maturation through the TLR4–JNK/c-Jun–GATA1 pathway. Together, these studies support a multi-level model in which both altered hematopoietic programs at the HSPC stage and inflammatory niche signaling converge to limit effective megakaryopoiesis in ITP. Lin et al. reported that S100A8/A9 engages TLR4 to activate STAT5, particularly in a thrombopoietin-rich milieu, expanding megakaryocytic populations and supporting tumor progression in multiple myeloma²⁰. By contrast, in ITP, we delineate a distinct axis whereby neutrophil-derived S100A8/A9 activates JNK/c-Jun downstream of TLR4 to repress GATA1, leading to a block in megakaryocyte maturation and reduced platelet production. These divergent effects likely reflect disease context and cytokine environment—malignant, TPO-rich marrow versus immune-mediated inflammation. Consistent with this, tasquinimod improved thrombocytopenia in our ITP models rather than suppressing thrombopoiesis. We therefore propose a context-dependent bifurcation downstream of TLR4, with STAT5-biased signaling in MM and JNK/c-Jun–dominant signaling in ITP; the molecular switches governing this branch choice warrant future study.

Consistent with these molecular events, our scRNAseq analyses (including RNA velocity and pseudotime) showed that overabundant S100A8/A9 prevents MKs from progressing beyond earlier developmental stages, directing them to remain in Immature MK and Cycling MK subsets rather than differentiating into the platelet-producing phenotype. This blockade is associated with reduced expression of surface markers (such as CD41, CD42b, and CD61), impaired polyploidization, and diminished platelet release, highlighting the broader spectrum of functional disruption that arises when heightened proinflammatory signaling converges on megakaryopoiesis.

Beyond the insight into disease mechanisms, our findings underscore the potential for translating this work into clinical applications. Notably, blocking the S100A8/A9–TLR4 axis via tasquinimod or TLR4 antagonists improved platelet counts and restored megakaryocyte (MK) maturation in our ITP models, suggesting that targeting these proteins could be a viable avenue for therapy. Tasquinimod (also known as ABR-215050) has already shown benefit in other inflammatory and autoimmune contexts⁵³, indicating that small-molecule inhibitors of S100A8/A9 or TLR4 may hold broad therapeutic promise. Beyond inhibiting this alarmin-receptor interaction, our data also point to potential strategies focused on the downstream JNK/c-Jun pathway. Given c-Jun’s role in suppressing GATA1 expression, using JNK inhibitors or devising means to enhance GATA1 activity could pave the way for additional therapeutic approaches aimed at correcting the profound megakaryocytic impairment observed in ITP. These interventions may help address refractory cases of thrombocytopenia while improving overall safety profiles, given that off-target concerns associated with broad immunomodulators could potentially be circumvented by more specific blockade of the S100A8/A9–TLR4–JNK axis. Further preclinical and clinical trials will be required to evaluate efficacy and safety, but this dual-target concept—blunting upstream inflammatory triggers and restoring essential transcriptional regulators—may represent a significant step forward in managing ITP and related autoimmune cytopenias.

Collectively, these findings inform ongoing efforts to refine ITP management by proposing a novel therapeutic strategy aimed at the newly recognized “neutrophil–megakaryocyte” axis rather than relying solely on conventional immunosuppression or thrombopoietin receptor agonists. Current standard therapies—high-dose corticosteroids, IVIG, rituximab, and TPO receptor agonists (eltrombopag, romiplostim)—are effective for many patients^4,58,59, yet a subset remains refractory to these approaches or experiences recurrent relapses^7,60. By contrast, inhibiting S100A8/A9–TLR4 signaling may address distinct pathogenic events beyond the scope of purely immunosuppressive measures, specifically targeting abnormal neutrophil expansion and the resulting megakaryocyte dysfunction. Future clinical trials will be required to confirm the efficacy, tolerability, and safety of such interventions in humans. However, if validated, these therapies could expand the therapeutic arsenal for ITP, offering an alternative option for individuals who do not adequately respond to current standard-of-care regimens. Our findings highlight a clear distinction between the source and the signal in the neutrophil–megakaryocyte axis. Bone-marrow neutrophils act as the sustained, localized source of S100A8/A9, generating high concentrations within the marrow niche that impair megakaryocyte maturation. Once the S100A8/A9 heterodimer is present systemically—whether produced endogenously or delivered exogenously—it is sufficient to reproduce the thrombopoietic defects, rendering the presence of neutrophils dispensable as downstream effectors. This “source versus signal” framework suggests that therapeutic strategies do not necessarily require depletion of neutrophils but could instead focus on neutralizing the pathogenic S100A8/A9 signal or blocking its receptor-mediated effects, offering broader applicability even in patients with normal neutrophil counts.

In parallel, our observations also underscore that aberrant S100A8/A9 elevations can be maintained for a longer duration in more advanced or persistent disease, such as chronic or severe ITP, aligning with observations from prior work indicating that sustained inflammatory signals characterize difficult-to-treat cases^61,62. Analyses of public datasets also reveal a notable decrease in S100A9 expression during remission, mirroring the clinical and laboratory improvements typically observed in milder disease states. These findings underscore the notion that S100A8/A9-mediated pathways may be closely tied to ITP severity and chronicity, providing an immunologic signature that correlates with distinct disease phases.

Despite this emerging insight, it remains to be established whether the S100A8/A9–TLR4 axis operates similarly in other clinically relevant subtypes, such as pediatric or pregnancy-associated ITP. Heterogeneity in immune context, hormonal environment, and hematopoietic biology could all modulate the impact of excessive neutrophil- and MK-derived alarmins in these populations. As such, further investigations focused on these subgroups will be necessary to confirm our observations and refine precision-targeted therapies across the full spectrum of ITP presentations.

Looking ahead, several lines of investigation may help translate our findings into concrete clinical advances. First, a prospective clinical trial evaluating S100A8/A9 antagonists or TLR4 inhibitors in ITP patients at various disease stages would define safety, tolerability, and efficacy profiles, as well as clarify optimal timing of these interventions—particularly in relapsing or refractory populations⁶³. Identifying rational combination strategies with existing therapies (thrombopoietin receptor agonists, immunosuppressants) could bolster platelet recovery while potentially minimizing overall immunosuppression⁶⁴. Such combined regimens might synergize by disrupting complementary pathways that sustain pathological thrombocytopenia.

At the mechanistic level, leveraging single-cell multi-omics (ATAC-seq, proteomics) should expand on the scRNA-seq data presented here to deliver a comprehensive view of the bone marrow immune microenvironment in ITP^65,66. These approaches would enable detailed interrogation of multicellular interactions—including how abnormal neutrophil or T-cell subsets shape megakaryocyte function—and highlight cell-specific regulators of platelet production⁶⁷. Finally, although our work establishes S100A8/A9 as a key disruptor of megakaryocyte maturation and platelet output, the broader S100 family warrants attention, given their known roles in amplifying inflammation and modulating hematopoiesis^68,69. Further research on these related alarmins could reveal yet unrecognized targets for preserving and restoring healthy megakaryopoiesis across a range of autoimmune cytopenias.

In conclusion, our study provides compelling evidence across single-cell analyses and multiple in vivo and in vitro models that the “neutrophil–megakaryocyte” axis constitutes a previously underappreciated pathogenic driver in ITP. Mechanistically, we demonstrate that S100A8/A9 secreted by neutrophils stimulates TLR4–JNK/c-Jun signaling in megakaryocytes, ultimately downregulating the crucial transcription factor GATA1 and thereby obstructing megakaryocyte maturation. This novel framework not only illuminates how abnormal neutrophil activity can derail thrombopoiesis in ITP but also points toward opportunities for therapeutic intervention. By targeting the S100A8/A9–TLR4–JNK/c-Jun pathway, our data suggest a viable route for improving platelet recovery in refractory or otherwise difficult-to-treat ITP cases, offering new directions for both fundamental immune-hematological research and future clinical strategies.

Limitations of Study

Despite involving multiple patients and experimental models in this study, a broader, multicenter clinical dataset is needed to confirm the generalizability of these findings. Additionally, although we focused on the S100A8/A9–TLR4–JNK/c-Jun pathway, other signaling pathways, such as p38 or NF-κB, could also play collaborative or compensatory roles. Our data identify MAPK signaling as the predominant pathway linking S100A8/A9–TLR4 activation to GATA1 suppression in ITP megakaryocytes, while canonical JAK-STAT activation was not detectable. We acknowledge the well-established MAPK ↔ JAK-STAT crosstalk in hematopoiesis and propose that GATA1 deficiency may sensitize megakaryocytes to JAK-STAT dysregulation under specific cytokine contexts, such as elevated thrombopoietin or IFN-γ. Thus, MAPK appears to be the primary, but not exclusive, effector in this S100A8/A9-driven circuit, and alternative pathways warrant exploration in future studies. A related limitation is that bulk serum S100A8/A9 measurements—especially in mice with fixed sampling schedules—may underestimate biologically relevant exposure at the bone-marrow niche due to species/assay differences and rapid distribution/clearance; local concentrations at megakaryocyte proximity are likely more informative. For Single-cell profiling, an apparently stable T-B MK proportion does not contradict a maturation blockade, because reduced inflow and longer dwell within the T-B state can preserve proportions while halting progression; accordingly, trajectory metrics (velocity, pseudotime) and functional markers are the more appropriate readouts of maturation dynamics. Plasma S100A8/A9 may underestimate marrow-niche exposure, and that model/kinetic/assay factors contribute to the smaller systemic signal in mice. Further exploration of other immune cells in the megakaryocyte niche, including macrophages and T cells, may provide a more comprehensive understanding of immune dysregulation in this context. Finally, while tasquinimod is already used in other disease settings, its translation to ITP therapy demands extensive pharmacokinetic and safety assessments. Potential side effects associated with long-term administration must also be thoroughly evaluated before clinical application can be considered.

Methods

Study cohort and diagnostic criteria

This study enrolled 48 adult patients (21 males, 27 females; age range 26–67 years, median 53) with primary immune thrombocytopenia (ITP) and 20 healthy controls at the First Affiliated Hospital of Soochow University, after written informed consent was obtained from all participants. ITP diagnosis followed both the Chinese guidelines for adult primary immune thrombocytopenia (2020 edition) and international standards, requiring persistently low platelet counts (<100 × 10⁹/L) on repeated CBCs, normal or elevated megakaryocyte counts on bone marrow examination but with signs of platelet maturation disruption, and typically no morphological abnormalities in peripheral RBCs or WBCs. Patients generally lacked overt splenomegaly, and secondary causes of thrombocytopenia (such as myelodysplastic syndrome, leukemia, autoimmune diseases, pregnancy-related thrombocytopenia, or drug-induced thrombocytopenia) were excluded. Inclusion criteria specified adults (≥18 years) with confirmed primary ITP and no confounding conditions, while exclusion criteria encompassed coexisting hematologic malignancies, infectious/autoimmune disorders, pregnancy, or unrelated immunosuppressive therapy. Clinically, patients presented with classical ITP manifestations (cutaneous purpura, mucosal bleeding), and many had chronic or persistent ITP requiring diagnostic bone marrow analysis. Autoantibody status was assessed by MAIPA and flow cytometry targeting anti-GPIIb/IIIa, anti-GPIb/IX, and anti-GPV. Bleeding symptoms were scored using the ITP Bleeding Assessment Tool (ITP-BAT), with severity defined as mild (ITP-BAT 0–2), moderate (3–5), or severe ( ≥ 6). This study was an observational case-control biospecimen study, not a clinical trial. Human participants were enrolled for the collection and analysis of peripheral blood and bone-marrow samples, and no participant was prospectively assigned to any treatment, diagnostic strategy, or other health-related intervention as part of this study. All pharmacologic or biologic interventions, including tasquinimod, IAXO-102, recombinant S100A8/A9, ruxolitinib, and anti-Ly6G, were performed only in ex vivo/in vitro systems or in mouse models. All procedures involving human participants or human biological materials were performed in accordance with the Declaration of Helsinki and were approved by the Medical Ethics Committee of The First Affiliated Hospital of Soochow University (approval/report no. 2022 Ethics Review Approval No. 176). Written informed consent was obtained from all patients and healthy donors before sample procurement and data collection.

Sample collection procedure

Bone marrow aspirations (1–2 mL) were obtained under sterile conditions from the posterior iliac crest using local anesthesia, collected in anticoagulant (heparin), and processed within 4 hours for cell isolation and subsequent assays (single-cell RNA-sequencing, flow cytometry). Peripheral blood samples (up to 5 mL) were drawn from an antecubital vein or an alternative site into EDTA tubes, with part of each sample used for routine complete blood counts (CBC) and the remainder for research experiments (flow cytometry, plasma cytokine evaluation). When needed, platelet-poor or platelet-rich plasma was prepared by differential centrifugation for further analyses. All human biological materials were managed according to institutional biosafety protocols; leftover specimens not required for clinical diagnostics were used for research in accordance with IRB approval and participant informed consent.

Cell lines

The human megakaryocytic cell line (Meg-01, ATCC #CRL-2021) was maintained in RPMI 1640 medium (Gibco, #11875093) supplemented with 10% fetal bovine serum (Gibco, #10099141) at 37 °C in a humidified 5% CO₂ incubator. Cells were routinely tested for mycoplasma contamination and authenticated by short tandem repeat (STR) profiling. All cultures were passaged according to the supplier’s guidelines and used for in vitro assays as described. Chromatin profiling (c-Jun Cut&Tag) and c-Jun perturbation (OE/knockdown, GATA1-promoter luciferase) were performed in MEG-01 owing to input and transfection constraints of primary polyploid MKs. All receptor-proximal signaling (TLR4 levels, p-JNK, p-c-Jun) and GATA1 readouts were measured and replicated in CD34⁺-derived primary MKs, and functional phenotypes were established in primary systems.

Primary cell culture

Primary human CD34⁺ hematopoietic progenitor cells (HPCs) were obtained from healthy adult donors (or from ITP patient bone marrow where indicated) following approval by the Medical Ethics Committee of The First Affiliated Hospital of Soochow University and written informed consent. Cells were enriched using immunomagnetic selection and then cultured in serum-free expansion medium (SFEM; STEMCELL Technologies, #09650) supplemented with thrombopoietin (TPO, 100 ng/mL, PeproTech, 300-18-10UG), stem cell factor (SCF, 100 ng/mL, PeproTech, 300-07-10UG), and interleukin-3 (IL-3, 10 ng/mL, PeproTech, 200-03-10UG) at 37 °C in a 5% CO₂ humidified incubator. Cultures were monitored for cell density and viability, and used for downstream assays (megakaryocytic differentiation, functional analyses) as described.

Mouse strains, husbandry, and experimental allocation

All animal procedures were approved by the Institutional Animal Care and Use Committee / Animal Ethics Committee of Soochow University and were performed in accordance with relevant national and institutional guidelines for the care and use of laboratory animals. Mice were maintained in a specific pathogen-free (SPF) barrier facility and were not housed under germ-free conditions. Animals were kept in individually ventilated cages with sterilized bedding and nesting material under a 12-h light/dark cycle, controlled ambient temperature of 20–25 °C, relative humidity of 40–70%, and ad libitum access to standard chow and autoclaved water.

Experimental and control mice were maintained in the same animal room under identical husbandry conditions. Within each experiment, mice were randomized to treatment groups and cage-balanced across racks and cages. Where compatible, age-, sex-, and strain-matched mice assigned to experimental and control groups were co-housed before randomization. After treatment allocation or ITP induction, mice were housed by experimental group when necessary to prevent cross-exposure to administered antibodies, recombinant proteins, or drugs, while maintaining the same cage density and environmental conditions across groups. Genetically distinct colonies, including Rag1^−/− recipient mice and Itgb3^−/− donor mice, were bred and maintained separately by genotype and were not co-housed across strains. No animals were housed in germ-free isolators.

Mice were humanely euthanized at the indicated experimental endpoints according to the approved institutional animal protocol. Euthanasia was performed by gradual-fill CO₂ inhalation using compressed CO₂ at a displacement rate of 30–70% of the chamber volume per minute, followed by cervical dislocation as a secondary method to ensure death. Death was confirmed by the absence of respiration and heartbeat before bone marrow, spleen, and other tissues were collected.

All animal experiments used Mus musculus. Unless otherwise specified, mice were 6–8 weeks old at the start of experiments. Sex- and age-matched mice were used within each experiment, and animals were randomly allocated to treatment groups using computer-generated randomization lists. The biological replicate and experimental unit for all in vivo analyses was an individual mouse. Exact animal numbers, sex distribution, strain/genotype, and age for each experiment are provided below and in the relevant figure legends.

Mice were monitored at least once daily throughout the experiments and at least twice daily during the expected thrombocytopenic phase or after antibody, recombinant protein, or drug administration. Humane endpoint criteria were predefined in the approved institutional animal protocol. Mice were removed from the study and humanely euthanized if they reached any of the following criteria: body-weight loss exceeding 20% from baseline, body condition score below 2/5, severe lethargy or immobility, inability to access food or water, labored breathing, persistent or uncontrolled bleeding, severe dehydration, severe wound or infection, hypothermia, or any moribund condition. All experiments were conducted in accordance with these approved humane endpoint criteria.

Single-cell RNA sequencing

For scRNA-seq analyses, we obtained 1–2 mL of bone marrow (BM) aspirates from 6 ITP patients and 6 age-matched healthy donors (HD), isolating mononuclear cells (MNCs) via Ficoll gradient centrifugation (450 g, 26 °C, 30 min). One HD sample did not meet pre-established QC criteria and was excluded; thus, the final scRNA-seq cohort included 6 ITP donors and 5 HD donors. To minimize hemodilution, all BM aspirates were obtained from the posterior iliac crest as first-pull draws and processed within 4 h under a single SOP. Sampling, RBC lysis, cell isolation, library chemistry, and sequencing depth were identical between ITP and HD cohorts. After discarding the supernatant, RBC lysis was performed, if necessary, followed by PBS washes (with 0.04% BSA). Viability (>80%) was confirmed (Trypan Blue), and cells were resuspended at ~700–1200 cells/μL for droplet-based scRNA-seq (Chromium Next GEM Single Cell 3ʹ Kit v3.1, 10x Genomics). Approximately 7,000–10,000 cells per sample were encapsulated and processed on a Chromium Controller; barcoded cDNA was then PCR-amplified, fragmented, and indexed to create sequencing libraries. Quality control was performed with Qubit and Bioanalyzer/TapeStation before pooling for high-throughput, paired-end 150-bp (PE150) sequencing on an Illumina NovaSeq 6000. Reads (~50,000 per cell) were aligned to the human GRCh38 reference genome using Cell Ranger (v9.0.1), which also demultiplexed and generated raw gene–cell count matrices. Downstream analysis employed the Seurat package (v5.0.1), filtering cells with nFeature_RNA of 500–6,000, nCount_RNA < 30,000, and percent.mt <5%. Log normalization and HVG selection were followed by PCA, UMAP, and clustering (FindClusters at resolution ~0.5). Differential expression was identified (FindAllMarkers; Wilcoxon rank-sum test, FDR < 0.01). Where indicated, pseudotime inference (Monocle) was applied to clarify lineage trajectories in megakaryocyte differentiation. RNA velocity was estimated in scVelo using the dynamical model (tl.recover_dynamics, tl.velocity(mode = “dynamical”), tl.velocity_graph) and projected onto the same UMAP used for clustering. Pathway enrichment used clusterProfiler (v4.8.3) with org.Hs.eg.db (v3.18) and KEGG release 110.0 (April 1, 2024); background was all genes tested; significance by Benjamini–Hochberg q < 0.05⁷⁰.

Spatial transcriptomics analysis

Human bone-marrow Visium data (GEO: GSE269875) were downloaded and three slices (hHBM1–3) were processed in “Seurat v5.0.1”. Raw 10x files and image assets (H&E) were organized in the standard Visium layout, low-resolution H\&E images were attached, and only in-tissue spots were retained. Counts were normalized and scaled with ‘NormalizeData()’, ‘FindVariableFeatures()’, and ‘ScaleData()’. Per-spot module scores were computed via ‘AddModuleScore()’ for neutrophils (CSF3R, MMP8, LCN2, CXCR2, S100A8, S100A9), megakaryocytes (MKs; PF4, PPBP, ITGA2B, GP9, TUBB1, ITGB3, GP1BB, GP6), and S100A8/A9 (S100A8, S100A9) using case-insensitive gene matching; spatial maps were rendered with ‘SpatialFeaturePlot()’ with display limits clipped to q02–q98. Hotspots were defined as the top 20% of module scores per slice. Using MK hotspots as queries, we calculated the mean nearest-neighbor distance to neutrophil or S100A8/A9 hotspots in full-resolution pixel coordinates and assessed significance with a label-permutation test (1,000 permutations of the reference hotspot mask) to obtain the null distribution of mean distances. We report ΔNN = NN (obs) − NN (rand), where negative values indicate closer-than-random proximity, alongside the one-sided p-value for the “closer” alternative. Pixel distances were converted to micrometers using the Visium calibration. For Fig. 1P, we additionally show closeness gain = NN (rand) − NN (obs), where positive values denote stronger proximity.

In vitro megakaryocyte culture from human CD34⁺ cells

Human hematopoietic progenitor cells (HPCs) were obtained from healthy donor bone marrow aspirates or G-CSF–mobilized peripheral blood after informed consent and Institutional Review Board approval. Following Ficoll gradient centrifugation to isolate mononuclear cells (MNCs), CD34⁺ cells were enriched using an anti-human CD34 microbead kit (EasySep Human CD34 Positive Selection Kit, STEMCELL Technologies, #17897), and the purified fraction ( ≥ 85% CD34⁺ by flow cytometry) was either resuspended in serum-free expansion medium (SFEM) for immediate use or cryopreserved. For megakaryocyte (MK) differentiation, CD34⁺ cells (2 × 10⁵ cells/mL) were seeded in 24-well plates (500 μL per well) using a serum-free medium with thrombopoietin (TPO, 100 ng/mL), stem cell factor (SCF, 100 ng/mL), and interleukin-3 (IL-3, 10 ng/mL). On Day 4, the culture volume was increased to 1 mL per well with the same cytokines, and from Day 7 onward, half-medium exchanges or total refreshes were performed every 3–4 days until Day 10–12 to obtain late-stage MKs. To simulate ITP, 1% (v/v) patient-derived plasma or healthy control plasma—each prepared by pooling samples from 10 individual donors—was added from Day 0 to Day 4, and this concentration was maintained in each subsequent medium refresh. Where indicated, pharmacological agents were introduced once MK commitment had begun: the S100A8/A9 antagonist Tasquinimod (Selleckchem, S7617, 2 μM) or the TLR4 inhibitor IAXO-102 (Selleckchem, S0002, 10 μM) was added on Day 4 or Day 7 and refreshed at each medium change (schematic in Fig. 2A). End-point assessments included flow cytometric analysis of MK surface markers (CD41/GPIIb, BioLegend, #984504; CD42b/GPIbα, BioLegend, #303903; CD61/GPIIIa, BioLegend, #336405), ploidy distribution (Propidium iodide, Invitrogen, R37169) (2 N, 4 N, 8 N, ≥16 N), and platelet-like particle (PLP) release, measured by collecting culture supernatants, performing low-speed centrifugation to remove larger cells, then pelleting the PLPs at higher speed ( ~ 3000 rpm), and analyzing CD41+ events by flow cytometry. On Day 8, cultures were treated with recombinant human S100A8/A9 (R&D Systems, #8226-S8, 0–3 µg/mL) for 1 h for p-JNK analysis, 72 h for CD41 expression, or 11 days for DNA ploidy. p-JNK and CD41 were quantified by flow cytometry (APC–anti-CD41), and ploidy was determined by propidium iodide staining. MKs were also incubated with 5% ITP plasma in the presence of tasquinimod (0–3 µM) for 72 h. CD41 expression was measured by flow cytometry, and rescue was expressed relative to ITP plasma alone.

Mouse platelet isolation (for function readouts)

Terminal cardiac puncture blood was collected into syringes preloaded with 3.8% sodium citrate (anticoagulant: blood = 1:9), mixed gently, and processed within 60 minutes; unless noted, centrifugations are reported as relative centrifugal force (×g) using a swing-bucket rotor at room temperature (20–25 °C) with the brake off. Platelet-rich plasma (PRP) was prepared by centrifugation at 100 × g for 8 min, and the supernatant was transferred carefully; if residual erythrocytes or leukocytes were visible, a clarifying spin at 100×g for 5 min was performed. For light-transmission aggregometry (LTA), PRP counts were adjusted to ~2.5 × 10⁸ platelets/mL, and matched platelet-poor plasma (PPP; see below) from the same mouse was used to set 100% transmission; assays were run at 37 °C and 1000 rpm with collagen at the indicated final concentrations. When washed platelets were required (e.g., spreading or clot-retraction protocols), platelets were pelleted from PRP at 800×g for 10 min and gently resuspended in Ca²⁺/Mg²⁺-free modified Tyrode’s buffer (137 mM NaCl, 2.7 mM KCl, 12 mM NaHCO₃, 0.36 mM NaH₂PO₄, 1 mM MgCl₂, 5.6 mM glucose, 10 mM HEPES, 0.35% BSA, pH 7.35–7.40) supplemented, when indicated, with prostacyclin (PGI₂, 0.5 µM) and apyrase (0.02 U/mL) to limit spontaneous activation; suspensions rested 30 min at room temperature, after which CaCl₂ and MgCl₂ were restored to 1 mM and inhibitors omitted immediately prior to stimulation or seeding on fibrinogen. Quality control included automated platelet counts, flow-cytometric confirmation of residual leukocytes <1 per 10⁶ platelets (CD45), and baseline activation <5% CD62P⁺ before agonist addition.

Plasma preparation (PPP, PFP) and handling

For megakaryocyte culture and ELISA, human research plasma was prepared as platelet-free plasma (PFP); whole blood collected in K₂-EDTA tubes was processed within 60 min by a first spin at 2500 × g for 15 min to obtain PPP, followed by clarification at 10 000 × g for 10 min to yield PFP with residual platelets <1 × 10⁴/µL. Aliquots (0.25–0.5 mL) were stored at −80 °C, thawed gently at 37 °C, mixed, and clarified again at 10 000 × g for 5 min immediately before use; no sample was subjected to more than one freeze–thaw cycle. For in-vitro experiments, equal-volume pooled PFP from 10 ITP donors or 10 healthy donors was used at 1% (v/v) unless otherwise stated (5% in specified dose–response/rescue assays); the culture medium contains physiological Ca²⁺/Mg²⁺, which mitigates trace EDTA chelation at these dilutions. For platelet function assays in mice, the plasma diluent/blank was matched citrated PPP prepared from the same draw (2500 × g, 15 min; repeat if needed). When bone-marrow supernatants were analyzed, aspirates collected in heparin were spun at 500×g for 10 min, clarified at 10 000 × g for 10 min, stored at −80 °C, and used after a single thaw. All temperatures, times, and RCF settings were recorded to ensure batch-to-batch consistency and reproducibility.

Ultrastructural quantification of MKs

MKs were performed on transmission electron micrographs of CD34⁺-derived megakaryocytes using Fiji/ImageJ under blinded conditions. For each biological replicate, intact MK cross-sections were sampled from predefined, systematically spaced fields; truncated or poorly preserved sections were excluded, and pixel-to-length calibration used the embedded scale bar. The cytoplasmic reference area (A_cytoplasm) was obtained by outlining the cell boundary and subtracting the nucleus. Demarcation membrane system (DMS) cisternae/lumina were manually segmented to yield total DMS area (A_DMS); the primary readout was the DMS area fraction A_DMS/A_cytoplasm. α-granules were quantified by threshold-based particle analysis within the cytoplasm with size/circularity filters (typically 80–1000 nm²; 0.2–1.0) followed by manual review, and expressed as number per μm² cytoplasm. Measurements were performed by two blinded operators with discrepancies resolved by consensus. Single-cell values were averaged within each biological replicate, and group comparisons were conducted at the replicate level using two-sided Wilcoxon rank-sum tests unless otherwise stated; n (replicates and cells per replicate) is reported in the figure legends.

Flow cytometry

Flow cytometric analyses were performed on instruments such as the BD FACSCanto II or ACEA NovoCyte, with panels tailored to species and experimental needs. For human megakaryocytes (MKs), antibodies included anti-human CD41a (APC), CD42b (FITC), CD61 (PE), and annexin V/PI for apoptosis detection. Gating strategies began with forward scatter (FSC) vs. side scatter (SSC) to exclude debris, followed by FSC-H vs. FSC-A to remove doublets. For all flow-cytometric assays, propidium iodide (PI; Invitrogen, R37169) (1 µg/mL, 10 min on ice) was used as the sole exclusion dye for live/dead gating. DAPI was never used in live-cell flow-cytometry experiments. In confocal microscopy, DAPI was applied only after 4% paraformaldehyde fixation as a nuclear counterstain. MKs were generally identified as CD41⁺CD42b⁺ or CD41⁺CD61⁺ populations. FlowJo (v10.6) or NovoExpress software was used for compensation, gating, and FCS data export, while GraphPad Prism or R was employed for statistical analysis and graphing. For functional assessments, cells were harvested, washed in PBS, and stained with anti-human CD41-APC, CD42b-FITC, or CD61-PE for 15–20 minutes at room temperature in the dark, then washed and analyzed to assess surface marker expression. DNA-content analysis involved labeling CD41+ cells with PI after RNase treatment. Manual gates were placed around visible 2 N, 4 N, 8 N, and ≥16 N PI/DNA-content peak windows and were applied consistently across conditions within each experiment. These gates were analyzed as independent peak-window readouts rather than as an exhaustive mutually exclusive partition of all DNA-content events; therefore, peak-window frequencies were interpreted separately and were not required to sum to 100%. For PLP quantification with nuclear exclusion, PI was incorporated into the CD41 staining panel to identify and exclude membrane-compromised events. To validate DNA content of the gated population, we flow-sorted CD41⁺PI⁻ submicron events and performed post-sort CD41/DAPI immunofluorescence, showing near-background DAPI signals comparable to peripheral platelets (Supplementary Fig. 2X). Following surface staining with APC–anti-human CD41a, cells were incubated with PI (1 µg/mL, 5 min at room temperature, protected from light) immediately prior to acquisition. PLPs were defined as CD41⁺ PI⁻ events within the submicron gate (FSC^low/SSC^low) calibrated using size-standard beads (Megamix, BioCytex). The percentage of PLPs was calculated relative to the total events within the acquisition gate. This gating refinement ensured that only intact, anucleate platelet-like particles were included in the analysis, eliminating potential contamination from small, nucleated cells or apoptotic fragments. For analysis of p-JNK dose–response and wash-out, cells were fixed with 2% paraformaldehyde for 10 min at room temperature, permeabilized with 90% methanol on ice for 30 min, and stained with PE-conjugated anti–phospho-SAPK/JNK (Thr183/Tyr185) (G9) Mouse mAb (Cell Signaling Technology, #5755, 1:50 dilution) according to the manufacturer’s protocol. For DNA-content/ploidy analysis, CD41+ singlet megakaryocytes were analyzed on PI/DNA-content histograms. Manual gates were placed around visible 2 N, 4 N, 8 N, and ≥16 N peak regions using the HC/control sample as a reference and then applied consistently across conditions within each experiment. These gates were treated as independent peak-window readouts rather than as an exhaustive mutually exclusive partition of all DNA-content events. Boundary events in valley/shoulder regions between adjacent peaks could remain unassigned or fall within adjacent manually drawn peak-window gates; therefore, individual peak-window frequencies were analyzed separately and were not required to sum to 100%.

Cell adhesion, migration, and apoptosis assays

Immediately before seeding, viable CD41⁺ megakaryocytes (MKs) were enumerated, and equal numbers were plated per well onto fibronectin-coated plates (Sigma, F2006, 10 μg/mL). After 45 minutes at 37 °C, non-adherent cells were gently washed off with pre-warmed buffer. Adherent MKs were imaged and counted in predefined fields by a blinded observer. The primary readout was adherent MKs per field; for transparency, the same data are also provided in the Source Data, normalized to cells plated per well (% adherent of plated MKs). Transwell inserts (8-μm pores, 24-well) were used with stromal cell–derived factor-1α (SDF-1α, PeproTech, #300-28A-10UG) in the lower chamber as chemoattractant. Equal numbers of MKs were loaded into the upper chamber in serum-free medium. After incubation (typically 3 h, 37 °C), non-migrated cells were removed from the upper surface. Cells on the lower surface were fixed and stained (or quantified by fluorescence if pre-labeled) and counted in predefined fields by a blinded observer. Apoptosis was assessed by flow cytometry using an Annexin V–APC/PI kit (BioLegend, #640932). Gates were defined as Annexin V⁺/PI⁻ (early apoptosis) and Annexin V⁺/PI⁺ (cells with compromised membranes: late apoptosis/secondary necrosis); MKs were identified as CD41⁺ with standard FSC/SSC gating.

S100A8/A9 immunodepletion from ITP plasma

Pooled ITP plasma was pre-cleared with isotype IgG–Protein A/G magnetic beads (Thermo Pierce, #88802, 30 min), then incubated with anti-human S100A8/S100A9 heterodimer mAb (R&D Systems, MAB45701, clone 1099 F) immobilized and BS³-crosslinked on Protein A/G beads. Incubations were done in PBS containing 2 mM CaCl₂ for 45–60 min; supernatant was collected on a magnetic rack and, when needed, subjected to a second round to improve yield. Depletion efficiency was quantified by calprotectin ELISA and routinely exceeded 80%. Mock-depleted plasma (isotype-coupled beads) was used during method development.

siRNA knockdown in primary megakaryocytes

Primary human CD34⁺ progenitors were differentiated into megakaryocytes (MKs) as described above. At differentiation day 7, MKs were transfected with a non-targeting control siRNA (siCtrl) or siRNAs targeting TLR4 (siTLR4) or AGER/RAGE (siRAGE) according to the manufacturer’s instructions. TLR4 knockdown was validated by immunoblotting.

For flow cytometry, MKs were stained with anti-CD41 and/or anti-CD61 and analyzed after doublet exclusion; the frequency of CD41⁺ and CD61⁺ events was reported as Freq. of parent (%). For DNA ploidy analysis, cells were ethanol-fixed/permeabilized, treated with RNase A, and stained with propidium iodide (PI) for DNA-content measurement by flow cytometry. PI staining provides a stoichiometric readout of DNA content and is widely used for cell cycle/ploidy analysis; ploidy gates were set on the 2 N peak and applied to quantify 2 N, 4 N, 8 N and ≥16 N populations within the CD41⁺ gate.

Active ITP model in mice

Adult Itgb3a^−/− mice (JAX, #004669, 6–8 weeks old) were repeatedly immunized with wild-type platelets (1.0–1.5 × 10⁹ platelets per injection) via the tail vein on a weekly or biweekly schedule for 4–6 weeks to induce high-titer anti-platelet autoantibodies. Once elevated anti-platelet IgG levels were confirmed by flow cytometry, donor mice were euthanized, spleens were removed under aseptic conditions, and single-cell suspensions were prepared by passing the spleens through a 200-μm nylon mesh into ice-cold PBS, followed by RBC lysis and subsequent washes. Cells were then resuspended in RPMI-1640 at 5 × 10⁷–1 × 10⁸ cells/mL for immediate transfer into sublethally irradiated (200–400 cGy) RAG1^−/− mice (JAX, #002216, 6–8 weeks old) via tail vein injection (5 × 10⁶–1 × 10⁷ splenocytes in 100–200 μL PBS) within 4–6 hours post-irradiation. These recipient mice typically developed marked thrombocytopenia mimicking ITP about 1–2 weeks later. At that point, animals were randomly assigned to receive Tasquinimod (an S100A8/A9 antagonist) dissolved in a suitable vehicle (0.2% CMC), administered at 1 mg/kg or other optimized doses by intraperitoneal injection, daily or every other day for 3–7 days. Vehicle-only controls received the same volume of CMC or PBS without the active compound.

Platelet counting and tail bleeding assay

Peripheral blood samples (10–30 μL) were collected from the orbital plexus or via tail snip at baseline and at designated intervals (daily or every 2–3 days). These samples were diluted (1:20 or 1:50) in an isotonic buffer containing EDTA (or an RBC lysis buffer) and analyzed using an automated hematology analyzer or by flow cytometry–based platelet counting. For the tail bleeding assay, a standardized 2–3 mm segment of the tail tip was excised, and the tail was submerged in isotonic saline at 37 °C. Bleeding time was recorded until hemostasis was achieved (no visible bleeding for ≥30 seconds). If bleeding persisted beyond 10–15 minutes, gentle pressure was applied to ensure animal welfare. Using these approaches, the active ITP mouse model was utilized to investigate the pathophysiological role of S100A8/A9 in immune-mediated thrombocytopenia and to evaluate the therapeutic potential of tasquinimod in restoring platelet counts and normal hemostatic function.

Passive ITP model

Six- to eight-week-old wild-type C57BL/6 mice (JAX, #000664, 6–8 weeks old) of both sexes were maintained under SPF conditions and used to induce passive immune thrombocytopenia (ITP). A rat anti-mouse CD42b (GPIbα) monoclonal antibody (0.1 µg/g) in sterile PBS was administered intravenously (tail vein or retro-orbital) on Day 0 (100–200 µL), followed by a second injection of the same dose on Day 1 to sustain platelet depletion. Blood samples (10–30 µL) were taken at baseline (Day −1) and daily thereafter (Days 0–5 or as indicated), diluted (1:20–1:50) in an isotonic buffer containing EDTA or RBC lysis buffer, and analyzed via an automated hematology analyzer or flow cytometry to confirm a significant drop in platelet counts (often <50% of baseline) within 24–48 hours. Starting on Day 1 or Day 2, mice were treated once daily for 3–5 days with tasquinimod (1 mg/kg) intraperitoneal injection; control animals received equivalent volumes of vehicle (0.2% carboxymethyl cellulose or PBS).

Neutrophil depletion and recombinant S100A8/A9 administration in ITP mouse models

For neutrophil depletion, mice received intraperitoneal injections of anti-Ly6G monoclonal antibody (clone 1A8, Bio X Cell, #BE0075-1, 200 µg per mouse) or isotype control on Day −1 and every 3 days thereafter until the experimental endpoint. For recombinant S100A8/A9 (rS100A8/A9) treatment, purified murine S100A8/A9 heterodimer (R&D Systems, #8916-S8) was administered intraperitoneally at 50 µg per mouse daily for the indicated duration, beginning on Day 0 after ITP induction.

Histology and immunofluorescence

At designated endpoints (Day 5 or Day 7 in ITP models), mice were euthanized, and femurs/tibias (for bone marrow) as well as spleens were harvested under aseptic conditions. The collected tissues were fixed in 4% paraformaldehyde (PFA) for 12–24 hours at room temperature or 4 °C; for bone marrow, long bones underwent EDTA-based decalcification for 1–2 weeks at 4 °C with regular buffer changes. Following fixation and decalcification, tissues were dehydrated through ascending ethanol concentrations, cleared in xylene, and then embedded in paraffin. Paraffin blocks were sectioned at 5–8 μm using a rotary microtome, and sections were floated on a water bath before being mounted onto glass slides. For hematoxylin and eosin (H&E) staining, slides were deparaffinized in xylene, rehydrated, stained with hematoxylin (optionally differentiated in acidic ethanol), blued in alkaline solution, counterstained with eosin, dehydrated, and mounted. Immunofluorescence staining involved deparaffinization, rehydration, heat-mediated antigen retrieval (in citrate or EDTA buffer), and blocking (with 5% normal serum or 3% BSA). Primary antibodies (anti-CD41) were applied overnight at 4 °C, followed by washes, incubation with fluorophore-conjugated secondary antibodies, and nuclear counterstaining with DAPI (Invitrogen, D1306) before mounting. Stained slides were examined by light microscopy (H&E) or fluorescence/confocal microscopy (immunofluorescence), and images were captured under standardized conditions. Quantification included counting megakaryocytes per high-power field, assessing their size and distribution, evaluating extramedullary hematopoiesis in the spleen, and analyzing any relevant immunofluorescent signals (CD41 expression, Abcam, ab134131) using ImageJ or comparable software. This integrated approach enables the detailed examination of tissue-level changes in megakaryocyte biology and extramedullary hematopoiesis under various conditions, including ITP and pharmacological interventions.

Real-time qPCR / RT-PCR analysis

Total RNA was extracted from cultured megakaryocytes, Meg-01 cells, or primary bone marrow mononuclear cells (BMMCs) using TRIzol (Invitrogen, #15596026) in accordance with the manufacturer’s instructions. RNA concentration and purity were evaluated via a NanoDrop spectrophotometer (Thermo Fisher), and samples with A260/280 ratios of 1.8–2.1 were deemed suitable. Reverse transcription was performed using 1 µg of total RNA with a 5× All-In-One RT Master Mix (ABM) or a similar kit, under thermal conditions of 25 °C for 10 minutes, 42 °C for 15 minutes, and 85 °C for 5 minutes; the resulting cDNA was stored at −20 °C. Primer sequences for target genes included S100A8 (Forward: ATGCCGTCTACAGGGATGAC; Reverse: ACTGAGGACACTCGGTCTCTA), S100A9 (Forward: GCACCCAGACACCCTGAACCA; Reverse: TGTGTCCAGGTCCTCCATGATG), TLR4 (Forward: AGCTTGGTGGTGACTATCCATTGA; Reverse: AGGGCTTGAATGGAGGCAC), and GATA1 (Forward: TCGGCAAGAAAAGGGCAAG; Reverse: TTGCAAATTCATCCTGTGGTTCT), with GAPDH (Forward: GTCTCCTCTGACTTCAACAGCG; Reverse: ACCACCCTGTTGCTGTAGCCAA) serving as a housekeeping control. Real-time qPCR was conducted with 2× SYBR Green qPCR Master Mix (Thermo Fisher, # K0221) on an Applied Biosystems instrument using a 20-µL reaction volume comprising 10 µL SYBR mix, 1 µL cDNA, 0.8 µL forward primer, 0.8 µL reverse primer, and nuclease-free water. The cycling protocol typically consisted of an initial 95 °C hold for 30 seconds, followed by 40 cycles of 95 °C for 5 seconds and 60 °C for 30 seconds. Relative gene expression levels were calculated by the 2^−ΔΔCt method, normalizing each target to GAPDH or β-actin.

Western blot analysis

Cells (Meg-01 lines, primary megakaryocytes, or bone marrow mononuclear cells [BMMCs]) were harvested, washed in PBS, and lysed on ice for 30 minutes in RIPA buffer (Beyotime, P0013B) supplemented with 1× protease and 1× phosphatase inhibitor cocktails (CST, #5871). Lysates were sonicated briefly (two 5-second pulses) and centrifuged at 12,000 × g for 5 minutes at 4 °C. The supernatant was collected, and protein concentration was measured using a BCA assay (Thermo Fisher, # A55860). Equal protein amounts (20–40 µg) were then mixed with 5× SDS loading buffer, boiled at 95–100 °C for 5 minutes, and stored at −20 °C. Samples were resolved on 8%–12% SDS-PAGE gels and transferred onto PVDF membranes. Membranes were blocked in 5% nonfat milk or BSA in TBST for 1 hour at room temperature and incubated overnight at 4 °C with primary antibodies against, TLR4 (CST, #14358, #38519), JNK (CST, #9252), pJNK (CST, #9251), c-Jun (CST, #9165), p-c-Jun (CST, #3270), and GATA1 (CST, #3535). Housekeeping proteins (β-actin, CST, #4970) served as loading controls. After washing, membranes were incubated for 1 hour with HRP-conjugated secondary antibodies (goat anti-rabbit, CST, #7074 or goat anti-mouse, CST, #7076), washed again, and developed using an ECL reagent. The resulting bands were visualized on film or a digital imaging system, and their intensities were quantified with ImageJ software, normalized to β-actin (CST, #4970) or GAPDH (CST, #2118).

S100A8/A9 ELISA analysis

To measure secreted S100A8/A9 levels, an ELISA was utilized according to the manufacturer’s instructions. Culture supernatants, plasma, or bone marrow aspirates were collected and appropriately diluted in the provided assay buffer before being added to 96-well plates pre-coated with capture antibodies specific for S100A8/A9 (Sigma-Aldrich, # RAB0730). Following incubation and washing steps, detection antibodies and substrate solution were applied, and the absorbance at 450 nm was measured using a microplate reader (BioTek, # Epoch 2). Sample concentrations were calculated by comparing the absorbance values to a standard curve generated with known concentrations of recombinant S100A8/A9.

Cut&Tag for c-Jun binding analysis

Approximately 1–2 × 10⁵ Meg-01 stable cells were collected and processed using a standard Cut&Tag protocol (pA-Tn5 Transposase from EpiCypher or an equivalent published method) to map c-Jun binding sites. Briefly, nuclei were isolated and permeabilized, followed by incubation with an anti–c-Jun primary antibody and a secondary pA-Tn5 enzyme (EpiCypher, #15-1117), allowing targeted tagmentation of c-Jun–associated chromatin regions. The resulting DNA fragments were PCR-amplified, size-selected, and sequenced on an Illumina platform using paired-end reads (50–75 cycles). Reads were aligned to the appropriate reference genome (GRCh38 for human) using Bowtie2, and peak calling was performed with MACS2. Downstream annotation of peaks (ChIPseeker) and motif analyses (HOMER) were conducted to identify c-Jun–occupied regulatory elements. Particular attention was given to regions near the GATA1 promoter and enhancer to evaluate potential functional interactions.

Chromatin immunoprecipitation (ChIP)–qPCR

Primary human megakaryocytes were generated from CD34⁺ hematopoietic progenitor cells as described above and cultured with pooled healthy-control plasma or pooled ITP plasma (1% v/v; 10 donors pooled per group). Where indicated, tasquinimod (2 μM) was added during MK commitment and refreshed with each medium change. At the indicated time point, cells were crosslinked with 1% formaldehyde for 10 min at room temperature and quenched with 125 mM glycine. Nuclei were isolated, and chromatin was sonicated to an average size of ~200–500 bp. Equal amounts of chromatin were immunoprecipitated overnight at 4 °C with an anti–c-Jun antibody or species-matched IgG control, followed by capture using Protein A/G magnetic beads. After stringent washing, chromatin was eluted, reverse-crosslinked at 65 °C, treated with RNase A and proteinase K, and DNA was purified. qPCR was performed using SYBR Green chemistry with primers targeting the GATA1 promoter region and a control genomic region (GAPDH). Enrichment was calculated using the percent input method with correction for input dilution, and IgG ChIP served as the negative immunoprecipitation control (SimpleChIP Plus Sonication Chromatin IP Kit, CST, #56383).

Ligand–receptor interaction and pathway analyses

For scRNA-seq data, tools such as CellPhoneDB or Seurat’s built-in methods were utilized to identify potential ligand–receptor pairs between neutrophils and megakaryocytes, exemplified by the S100A8/A9–TLR4 interaction. Interactions were deemed significant based on high average expression and p-values below 0.05 (or adjusted q-values below 0.05). In addition, KEGG or Gene Ontology (GO) enrichment analyses were performed on differentially expressed genes to investigate relevant signaling pathways, highlighting processes such as MAPK, JNK, and NF-κB.

Platelet aggregation, clot retraction, and spreading assays

Washed platelets were stimulated with collagen (2, 3, or 5 µg/mL) in a light transmission aggregometer. Clot retraction was monitored in platelet-rich plasma clotted with thrombin (Sigma, T1063) in the presence of the indicated plasma and/or tasquinimod and photographed at 30- and 120-min. Platelet spreading on fibrinogen-coated coverslips was performed under the same treatment conditions, fixed, and stained with phalloidin to visualize the actin cytoskeleton.

Quantification and statistical analysis

All statistical analyses were performed using GraphPad Prism (version 9.0) and R (version 4.3.3). Unless otherwise specified, all tests were two-sided. Data are presented as mean ± SD, mean ± SEM, or median with interquartile range (IQR), as indicated in the corresponding figure legends. The exact sample size (n) for each experiment, the definition of biological replicates (e.g., independent donors, mice, or independent differentiations), and any technical replicates (e.g., ELISA wells) are reported in the figure legends and Source Data.

For comparisons between two independent groups, we used a two-sided unpaired Student’s t-test; for paired/repeated measurements on the same biological replicate, we used a two-sided paired t-test. For comparisons among ≥3 groups, we used one-way ANOVA followed by Tukey’s multiple-comparisons test; when multiple groups were compared against a single control group (e.g., dose–response assays), we used one-way ANOVA followed by Dunnett’s multiple-comparisons test. For experiments involving two factors (e.g., treatment × time), we used two-way (repeated-measures where appropriate) ANOVA followed by Sidak’s or Bonferroni multiple-comparisons test, as specified in the relevant figure legend. When nonparametric testing was required, we used Mann–Whitney U test (two groups), the Wilcoxon matched-pairs signed-rank test (paired), or the Kruskal–Wallis test followed by Dunn’s multiple-comparisons test (≥3 groups), as indicated.

Correlation analyses used Spearman correlation. For scRNA-seq differential expression, we used Wilcoxon rank-sum tests with Benjamini–Hochberg false discovery rate (BH-FDR) correction; pathway enrichment and GSEA significance were assessed using BH-FDR. Spatial nearest-neighbor proximity was assessed using one-sided label-permutation tests (1,000 permutations), as described in Methods. Exact statistical tests, multiple-comparison corrections, and P-value reporting for each figure are provided in the corresponding figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.Q. conceived the study, designed and performed experiments, analyzed data, and wrote the manuscript. J.Y., M.Z., Y.Z., H.H., and X.S. performed experiments and analyzed data; X.S. conducted extensive additional experiments during the revision. Z.Z. and X.L. contributed to clinical data collection and curation. Y.T. and F.Y. contributed to data analysis and interpretation. Z.L., Y.H., and D.W. contributed to study conception and design, supervised the work, and revised the manuscript. All authors reviewed and approved the final manuscript.

Peer review

Peer review information

Nature Communications thanks Tamam Bakchoul and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the National Natural Science Foundation of China (grants 82230005, 82200133, 82570174 and 82200137), the Elite Talent Reserve Program of the First Affiliated Hospital of Soochow University and Jiangsu Province of China (BE2021645), the Translational Research Grant of NCRCH (2021ZKMA01 and 2021ZKQA01), the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), and Beijing Bethune Charitable Foundation (XYZL-022, 2024-YJ-156- J-071).

Data availability

The single-cell RNA-sequencing data generated in this study have been deposited in the Gene Expression Omnibus under accession code GSE275370. The publicly available datasets used or reanalyzed in this study are available in the Gene Expression Omnibus under accession codes GSE56232, GSE46922, GSE23754, and GSE269875. De-identified numerical data underlying the graphs in the main and Supplementary Figs. are provided in the Source Data file. Additional processed data supporting the findings of this study are available within the Article, Supplementary Information, and Source Data file. Individual-level raw clinical or medical-record data that are not included in the Source Data file are protected by participant privacy and institutional ethics restrictions and are not publicly available. Requests for de-identified data access for academic research may be directed to the corresponding author Yue Han, and will be reviewed according to institutional ethics and data-governance requirements. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74774-7.