Single-cell profiling reveals immunoregulation of artemisinin on CD8⁺GZMB⁺ T cells via JAK2-STAT3 in malaria-infected mice

Abstract

Malaria, a pervasive and devastating disease, is characterized by systemic complications and dysregulated host immune responses to Plasmodium infection. Artemisinin derivatives, particularly artesunate (ART), are a cornerstone in malaria treatment strategies. Although the precise immunomodulatory mechanisms remain unclear, ART not only kills parasites but also impacts host immune homeostasis. In this study, we employed single-cell RNA sequencing to characterize the cellular landscape of 241,837 cells from multiple murine tissues (including liver, spleen, and peripheral blood) upon Plasmodium berghei ANKA (PbA) infection and after ART treatment. Meanwhile, we observed significant transcriptomic shifts across diverse immune cell types, with the liver exhibiting the most pronounced changes in response to PbA infection. Notably, CD8⁺GZMB⁺ T lymphocytes, characterized by elevated cytokine and cytotoxic module scores, play a pivotal role in driving hepatic injury. Furthermore, ART modulated this pathogenic subtype via the JAK2-STAT3 pathway, reducing its frequency and mitigating its inflammatory response. Our research provides a valuable dataset resource for exploring malaria immunopathogenesis and elucidates a novel immunoregulatory mechanism of ART within the infected host.

Graphical abstract

Public summary

• •Single-cell transcriptomics landscape of the peripheral blood, liver, and spleen of malaria-infected mice.
• •The heterogeneous immune responses across tissue and cell types before and after artemisinin treatment.
• •CD8⁺GZMB⁺ T cells exert an inflammatory response that drives liver injury during Plasmodium berghei infection.
• •Artemisinin modulates the cytotoxicity of CD8⁺GZMB⁺ T cells via the JAK2-STAT3 pathway.

Introduction

Malaria, caused by Plasmodium parasites, remains a global health burden with millions of infections and hundreds of thousands of deaths annually.¹ The disease progresses through distinct stages, including an asymptomatic liver stage and symptomatic erythrocytic stage, each associated with stage-specific pathological changes.² Reflecting its complex life cycle, Plasmodium infection commonly triggers systemic complications and severe immune responses, leading to multiple dysfunctions in the host. While an effective immune response is essential for parasite clearance and host protection, excessive immune activation and inflammatory cytokine release can induce severe damage, particularly in the brain, liver, and kidney.^3,4 Infected erythrocytes circulate through multiple organs or tissues, inducing local cytokine production (e.g., TNF-α) and immune cell-mediated cytotoxicity (e.g., by T cells or natural killer cells [T/NK], neutrophils [Neutro]), which exacerbates organ or tissue damage.^5,6 Despite these insights, the intricate host-parasite interplay remains incompletely understood, hindering therapeutic target identification and vaccine development. Addressing these gaps requires comprehensive characterization of immune response patterns across various organs or tissues during infection and treatment.

Over the past decade, the widespread use of artemisinin and its derivatives, such as artesunate (ART) and dihydroartemisinin (DHA), has significantly reduced malaria incidence, transmission, and mortality in endemic regions.^2,7 Previous studies have consistently demonstrated ART’s mechanism of action and protein target profiles against malaria parasites using chemoproteomics strategies.^8,9 Beyond direct antiparasitic effects, ART and DHA possess immunomodulatory properties in contexts including autoimmune diseases and sepsis-induced immune dysfunction.^9,10,11 We therefore hypothesized that ART plays a dual role in malaria treatment for Plasmodium-infected hosts: direct parasite killing and host immunomodulation. To verify this, we employed single-cell RNA sequencing (scRNA-seq) to elucidate the cellular and molecular mechanisms underlying ART’s immunoregulatory effects in a murine model with Plasmodium infection.

scRNA-seq captures transcriptomes at single-cell resolution, making it invaluable for exploring cellular heterogeneity and the complex interactions among pathogens, the host immune system, and anti-malarial drugs.^12,13,14 This technology has been applied to decipher the immune microenvironment in infectious diseases such as sepsis,¹⁵ tuberculosis,¹⁶ and COVID-19.^17,18,19 In malaria research, scRNA-seq has been used to construct Plasmodium life cycle atlas,^20,21 elucidate host-parasite interaction,^22,23 and identify disease biomarkers.^24,25 However, most studies have focused on in vitro models (such as hepatocytes [Hep] and red blood cells [RBCs]) or individual infected tissues in rodent models,²⁶ overlooking systemic host response. A comprehensive immune landscape across multiple tissues is therefore needed to dissect the transcriptomic alterations, and to pinpoint sources of inflammatory cascades driving pathogenesis during Plasmodium infection. Moreover, profiling cellular responses of the infected host to ART treatment may reveal critical pharmacological mechanisms and guide therapeutic strategies.

In this study, we utilized scRNA-seq to systematically characterize the cellular composition and immunological dynamics in peripheral blood (PB), liver (LV), and spleen (SP) tissues in C57BL/6J mice following Plasmodium berghei ANKA (PbA) infection and post-ART treatment. Our objectives include: (1) to map innate and adaptive immune cell landscapes, (2) to explore multi-tissue cellular response heterogeneity during malaria infection, and (3) to elucidate ART’s immunoregulatory mechanisms.

Materials and methods

Detailed materials and methods are included in the supplemental information.

Results

Therapeutic and immunoregulatory effect of ART in PbA-infected mice

As outlined in Figure 1A, we employed a PbA-infected mouse model in C57BL/6J mice to profile the immune cell landscape of multi-tissues under PbA infection and following ART treatment. Following the malaria treatment guidelines,²⁷ PbA-infected mice received intraperitoneal administration of ART (10 mg/kg) for 3 consecutive days, starting on day 5 post-infection. All experimental groups (control, model, and ART) were humanely euthanized on day 8 for comprehensive analysis.

Figure 1.

The single-cell cell atlas of murine multi-tissues during PbA infection and ART treatment

(A) The schematic workflow depicts the experimental design, including the time point of PbA infection and ART treatment, sample collection from PB, LV, and SP tissues for scRNA-seq experiment, and the initial bioinformatics analysis in this study.

(B–D) The uniform manifold approximation and projection (UMAP) plots of 241,837 individual cells are colored and annotated by the main cell types (B), 3 immune-relevant tissue types (C), and 3 group types (D). Each dot represents an individual cell.

(E) The Sankey plots show the relative proportion of group and main cell types across the PB (left panel), LV (middle panel), and SP (right panel).

Compared with the control group, mice in the model group exhibited significant weight reduction commencing at day 5, while ART treatment partially mitigated this weight loss (Figure S1A). Parasitemia analysis demonstrated a progressive increase in the model group from day 3 post-inoculation, contrasting sharply with the significant reduction observed in the ART group from day 5 onward (Figure S1B). Pathological assessment revealed marked hepatosplenomegaly in the model group, evidenced by significantly elevated coefficients for both LV and SP compared with the control group (Figures S1C and S1D). Histopathological examination of LV tissues in normal mice demonstrated well-organized hepatic plates radiating from central veins, while PbA-infected mice exhibited architectural distortion, characterized by central vein deformation, sinusoidal dilation, edema, hyperemia, inflammatory cell infiltration, and hemozoin deposition. Such pathological changes were alleviated by ART treatment. Splenic architecture in control mice maintained normal cellular organization, whereas disrupted white pulp architecture, increased red pulp cellularity, and indistinct marginal zones were observed in both model and ART groups (Figure S1E).

Serum biochemical analysis revealed significant elevation of aspartate aminotransferase and alanine aminotransferase activities, coupled with reduced albumin and total protein levels in PbA-infected mice, consistent with hepatic dysfunction.²⁸ ART treatment effectively reversed most of these hepatocyte injury markers and improved synthetic function (Figure S1F). In both the LV and SP, inflammatory cytokines (TNF-α, IL-6, and IL-1β)²⁹ were significantly upregulated during infection, and subsequently downregulated after ART treatment (Figures S1G and S1H). These results collectively demonstrate that ART exerts potent anti-inflammatory and immunoregulatory effects in the PbA-infected mice.

Construction of a multi-tissue cellular atlas in PbA infection and ART treatment via single-cell profiling

We performed scRNA-seq to profile the cellular transcriptomes of PB, LV, and SP from control, model, and ART groups (three biological replicates per tissue-group combination). We isolated and processed 296,833 cells using the 10x Genomics Chromium platform. After stringent quality control, a total of 241,837 high-quality single cells were retained for downstream analysis. On average, each sample contained 8,956.92 cells, with a median sequencing depth of 6,764.62 unique molecular identifiers and 1,719.76 genes per cell (Table S1). The final dataset comprised 81,441 cells from PB, 96,722 from LV, and 63,674 from SP, distributed across 87,611 control, 67,048 model, and 87,178 ART cells (Figures S2A–S2F). The cell counts per sample ranged from 3,358 (PM2) to 13,747 (LC2) (Figure S2G).

As visualized in uniform manifold approximation and projection, we annotated 11 major cell types based on the canonical markers: RBC (Hba-a1⁺), Hep (Alb⁺), endothelial cell (Endo, Kdr⁺), T/NK (Cd3e⁺Nkg7⁺), B lymphocyte (B cell, Cd79a⁺), plasma cell (Plasma, Igkc⁺), platelet (Pf4⁺), monocyte or macrophage (Mono/Macro, Cd68⁺), Neutro (S100a9⁺), dendritic cell (DC, Irf8⁺), and mast cell (Cpa3⁺) (Figures 1B and S3A). Differentially expressed genes (DEGs) analysis (adjusted_p_value < 0.05, avg_log2FC ≥ 0.25) and Gene Ontology (GO) enrichment revealed cell-type-specific biological processes, such as regulation of T cell activation in T/NK and erythrocyte homeostasis in RBC (Figure S3B). The cellular numbers ranged from 429 (mast cells) to 54,960 (B cells) (Figure S3C).

Principal-component analysis demonstrated that samples primarily clustered by tissue origin, with LV samples distinct from PB and SP samples (Figure S3D). Within LV, model and ART groups clustered more closely than with the control group. Most cell types lacked strong tissue and group specificity, except Endo and Hep (predominantly localized in LV) and RBC (enriched in PB and SP) (Figures 1C and 1D). Among major immune populations (Mono/Macro, T/NK, and B cells), the transcriptional pattern of B cells showed high correlation across samples, while Mono/Macro and T/NK cells displayed pronounced tissue-group-specific transcriptional heterogeneity, underscoring the context-dependent regulation of these two cell types (Figure S3E).

The relative abundances of each cell type across different tissues and groups are illustrated in Figure 1E. As anticipated, the proportion of RBCs in PB decreased upon malaria infection but rebounded after ART treatment. Conversely, Mono/Macro and T/NK cells in the LV were significantly expanded in model mice and partially reduced post-ART treatment. Similar dynamic redistribution occurred for T/NK and B cells in SP, highlighting the dynamic changes of immune cell in response to PbA infection and ART therapy.

Characterizing the heterogeneous immune response across tissues during PbA infection

Using our integrated scRNA-seq dataset, we analyzed immune response heterogeneity across the three tissues types upon PbA infection. DEGs analysis between the model and control groups revealed transcriptomic shifts for each immune cell type. As illustrated in Figure 2A, LV-resident T/NK and Mono/Macro harbored higher numbers of upregulated DEGs, while Neutro in PB exhibited predominant downregulation in DEGs, indicating that dynamic alteration occurred in these cell types upon PbA infection. We further investigated the shared and tissue-specific upregulated DEGs in Mono/Macro and T/NK across the three tissues. In both types, we found that the LV displayed the highest number of tissue-specific upregulated DEGs than others (n = 325 for Mono/Macro and n = 562 for T/NK), underscoring its unique role in the immune response to PbA infection (Figures 2B and 2C).

Figure 2.

Characterizing common and specific responses in different tissues upon PbA infection

(A) The radar chart shows the number of upregulated DEGs (left) and downregulated DEGs between model vs. control across three tissues.

(B and C) The upset plot shows the intersection of upregulated DEGs between model vs. control from three tissues in Mono/Macro (B) and T/NK cells (C).

(D) The vlnplot shows the relative expression of Ifitm1 and Ifitm2 in Mono/Macro between control and model groups across three tissues.

(E) The vlnplot shows the relative expression of Gzmk and Gzmb in T/NK between control and model groups across three tissues.

(F) The dotplot shows the GO enrichment of upregulated DEGs between model vs. control across three tissues in Mono/Macro (top panel) and T/NK (bottom panel).

(G) The heatmap indicates the relative abundance level of various cytokines and chemokines in interstitial fluid derived from the control and model groups across three tissues (n = 6). ∗∗∗∗p < 0.0001.

In Mono/Macro, interferon-related genes Ifitm1, Ifitm2, and Fc receptor-related genes Fcgr1 and Fcer1g were consistently upregulated among the 3 tissues (Figure 2B). In T/NK, there were 90 DEGs elevated together across 3 tissues upon PbA infection, including granzyme (GZMB) coding genes (Gzmk, Gzmb), and S100 calcium-binding protein (S100a4, S100a6) (Figure 2C). Notably, the elevated expression of Ifitim1 and Ifitim2 in Mono/Macro was observed across all three tissues in the model group, indicating a systemic inflammation (Figure 2D). Similarly, Gzmk and Gzmb expression in T/NK were significantly upregulated in the model group compared with the control group, with the highest abundance observed in LV (Figure 2E). Functional enrichment analysis revealed that hepatic T/NK cells specifically upregulated pathways related to cytokine binding, C-C chemokine receptor activity, and MHC class I protein binding, while T/NK cells from PB and SP were enriched in molecular function inhibitor activity, and extracellular matrix binding. In Mono/Macro, immune receptor activity, ubiquitin protein ligase binding, and MHC class I protein binding were consistently upregulated across all tissues, with Toll-like receptor binding uniquely activated in LV (Figure 2F).

Similar analyses extended to other immune cell types revealed that PB-derived platelets and SP-derived plasma harbored the highest numbers of upregulated DEGs, accounting for 98.61% and 85.23% of total upregulated DEGs, respectively (Figures S4A–S4D). However, these cell types did not activate canonical immune pathways, suggesting limited roles in direct defense against PbA infection. In contrast, Neutro and B cells exhibited a higher proportion of hepatic-specific DEGs (33.4% and 38.66%, respectively), associated with the regulation of responses to biotic stimuli and interferon-β or interferon-γ (IFN-γ) signaling (Figures S4E–S4H). These cell types also upregulated defense responses to bacteria and interferon signaling in both PB and SP.

Next, luminex cytokine profiling confirmed LV as the primary immunological hub. Multi-cytokine analysis demonstrated that most of the cytokines and chemokines, such as IFN-γ, GZMB, and C-C motif chemokine ligand 3, were enriched in the LV of the model group, while no significant difference was noted in PB and SP between the model and control groups (Figure 2G). Collectively, these findings highlight the critical roles of T/NK, Mono/Macro, and Neutro in driving tissue-specific immune responses during PbA infection. The LV particularly exhibits a robust and heterogeneous immune activation, warranting further investigation into the transcriptomic and functional states of these cell types.

Subtype analysis of myeloid cell and lymphocyte populations across tissues and groups

To resolve immune cell heterogeneity, we performed subset analysis and re-clustering of myeloid and lymphocyte populations, exploring their functional states and proportional changes across different tissue-group combinations. Myeloid cells were classified into six major subtypes: plasmacytoid DC (pDC, Siglech⁺), mast cell (Cpa3⁺), Mono/Macro (Cd68⁺), activated neutrophil (Neutro_Act, S100a8⁺S100a9⁺), homeostasis neutrophil (Neutro_Home, Camp⁺Ltf⁺), and platelet (Pf4⁺) (Figures 3A and 3B). In particular, given the tissue-resident characteristics and functional diversity of Mono/Macro, we further subdivided them into tissue-specific subtypes based on classical markers (Figures S5A and S5B).^18,30,31 A total of 20 Mono/Macro subtypes were annotated, including 6 circulating subtypes (Mono_C1-C5 and megakaryocytes), 8 hepatic subtypes (liver resident macrophages [Kupffer_C1-C4] and liver capsule macrophages [LCM_C1-C4]), and 6 splenic subtypes (SPM_C1-C4) (Figures 3B and S5C). Notably, Mono_C1_Cd14, represented the classical monocyte characterized by the expression of Cd14. Kupffer_C3_Itgax and LCM_C2_Itgax exhibited high expression level of Itgax, a marker of monocyte-derived macrophages inhibiting hepatic dissemination of bacterial infection.^32,33

Figure 3.

Single-cell transcriptome atlas of myeloid cells from three tissues upon PbA infection and ART treatment

(A) The UMAP plots of main myeloid cell types, labeled by distinguishing colors (left panel) and Mono/Macro subpopulations in three tissues (right panel).

(B) The violin plot shows the expression levels of the relative markers across all myeloid cell subtypes in the PB (top panel), LV (middle panel), and SP (bottom panel).

(C–E) (C) The bubble plot shows the cellular proportion of cell subtypes across the group-tissue pairs, and the dot size represents the cellular proportion, colored by lymphocyte subtypes, (D and E) IF staining of CD68⁺ITGAX⁺ cells from control, model, and ART groups in SP (D) and LV tissues (E). Scale bars, 10 μm.

(F and G) IF staining of C1QA⁺ITGAX⁺ cells (F) and C1QA⁺ CCR2⁺ cells (G) in LV tissues across control, model, and ART groups. Scale bars, 10 μm. ∗∗∗p < 0.001, ∗∗p < 0.01, and ∗p < 0.05.

The cellular numbers of myeloid subtypes ranged from 19 (SPM_C6_Ccl5) to 9,399 (Kupffer_C3_Itgax) (Figure S5D), with the majority of cells being from the model and ART groups (Figure S5E). Proportional changes in myeloid subtypes revealed distinct responses to PbA infection and ART treatment. The proportions of Neutro_Act and Mono_C3_S100a8 in PB increased during infection but decreased after ART treatment, whereas Mono_C1_Cd14 and Mono_C5_Fcgr4 gradually increased during infection and post-treatment (Figure 3C). In LV, Kupffer_C3_Itgax, and LCM_C2_Itgax cells, nearly absent in the control group, were enriched following PbA infection and remained elevated after ART treatment. Conversely, Kupffer_C4_Ccr2 increased in the model group but decreased after ART treatment. The proportion of SPM_C4_Itgax in SP was higher in the model group compared with control but showed no reduction post-treatment. Immunofluorescence (IF) experiment and quantitative analysis corroborated these findings, demonstrating consistent trends with the scRNA-seq data in the proportions of SPM_C4_Itgax (CD68⁺ITGAX⁺) in SP (Figure 3D), as well as LCM_C2_Itgax (CD68⁺ITGAX⁺), Kupffer_C3_Itgax (C1QA⁺ITGAX⁺), and Kupffer_C4_Ccr2 (C1QA⁺CCR2⁺) in LV (Figures 3E–3G).

In parallel, the re-clustering of lymphocytes delineated 23 subpopulations by canonical markers, encompassing 5 CD4⁺ T cell subtypes (Cd4_T_C1-C5), 5 CD8⁺ T cell subtypes (Cd8_T_C1-C5), 6 B cell subtypes (B cell_C1-C6), 6 plasma cell subtypes (Plasma_C1-C6), and NK cell (Figures 4A, 4B, and S6A–S6C). Notably, Cd4_T _C3_Cxcr5, characterized by elevated expression of Cxcr5, Tox, and Tox2, was characterized as T follicular helper cells.^34,35,36 Similarly, Cd4_T_C5_Cxcr3 with enriched Cxcr3 and Ifng was regarded as T helper 1 cells.^37,38 Likewise, Cd8_T_C2_Gzmb exhibiting high levels of GZMB-relative genes such as Gzmb and Gzmk³⁹ was defined as cytotoxic cells, a potent pro-inflammatory CD8⁺ T cell subtype.⁴⁰

Figure 4.

Single-cell transcriptomics analysis of lymphocytes from three tissues upon PbA infection and ART treatment

(A) The UMAP plots of lymphocytes annotated by main cell types, labeled by distinguishing colors (left panel), and T/NK, B cell, and plasma subtypes (right panel).

(B) The violin plot shows the expression levels of the respective selected markers across all lymphocyte subtypes.

(C) The bubble plot shows the cellular proportion of cell subtypes across the group-tissue pairs, and the dot size represents the cellular proportion colored by lymphocyte subtypes.

(D) IF staining of CD4⁺CXCR5⁺ T cells in SP tissues of control, model, and ART groups. Scale bars, 10 μm.

(E–G) IF staining of CD4⁺CXCR3⁺ T cells (E), CD8⁺GZMB⁺ T cells (F), and CD8⁺Ki67⁺ T cells (G) in LV tissues across control, model, and ART groups. Scale bars, 10 μm. ∗∗∗p < 0.001, ∗∗p < 0.01, and ∗p < 0.05.

The cellular numbers of lymphocyte subtypes ranged from 219 (Plasma_C6_Igha) to 26,857 (B cell_C1_Zfas1) (Figure S6D), with the majority of cells being from the model and ART groups (Figure S6E). The proportion of Cd4_T_C2_Foxp3 in PB decreased upon PbA infection but increased after ART treatment. Conversely, the proportions of Cd4_T_C5_Cxcr3, Cd8_T_C2_Gzmb, and Cd8_T_C3_Mki67 increased in the PbA-infected mice and decreased after ART treatment. Similar trends were observed in LV for Cd4_T_C1_Lef1, Cd4_T_C2_Foxp3, Cd8_T_C2_Gzmb, and Cd8_T_C3_Mki67, with the abundance of Cd4_T_C5_Cxcr3 being enriched post-PbA infection and ART treatment compared with controls. Meanwhile, the proportion of splenetic Cd8_T_C2_Gzmb and Cd8_T_C3_Mki67 underwent a substantial increase upon PbA infection and ART treatment. IF analysis demonstrated that relative abundance of the Cd4_T_C3_Cxcr5 (CD4⁺CXCR5⁺) in SP (Figure 4D), and Cd4_T_C5_Cxcr3 (CD4⁺CXCR3⁺), Cd8_T_C2_Gzmb (CD8⁺GZMB⁺), and Cd8_T_C3_Mki67 (CD8⁺Ki67⁺) in LV were consistent with our analysis results (Figures 4E–4G). These results highlight the tissue-specific dynamics of myeloid and lymphocyte subsets during PbA infection and ART treatment, providing insights into their functional roles and potential contributions to immune regulation.

Rewriting cell-cell interaction networks during PbA infection and ART treatment

We hypothesized that the divergent immune responses might arise from distinct intracellular communication among various cell types across the 3 tissue types. To verify this, we inferred the cell-cell interaction (CCI) patterns of diverse cell populations (10 types in PB, 10 in SP, and 11 in LV) across different tissue and group types (Figure S7A). Focusing on 6 representative cell types, we evaluated the upregulated CCI intensities between the model vs. control and model vs. ART groups (Figure 5A). As for PB, we observed heightened CCI intensities between Mono/Macro and DC, as well as between DC and T/NK cells in the model group, which were absent in the ART group. Additionally, T/NK in LV and Mono/Macro in SP exhibited enhanced intercellular communication with other cell types, highlighting their pivotal roles within the CCI networks during PbA infection. Notably, most of these elevated interaction intensities were attenuated following ART treatment, suggesting a regulatory role of ART in modulating intercellular communication.

Figure 5.

Deciphering the cell-cell interaction networks upon PbA infection and after ART treatment

(A) The heatmap demonstrates differential cell-cell interaction numbers between model vs. control and model vs. ART across the PB (left panel), LV (middle panel), and SP (right panel).

(B) The upset plot shows the intersecting differential interaction pathways between model vs. control and model vs. ART across the PB, LV, and SP.

(C) The bubble plot shows the communication probability of ligand-receptor pairs among various cell types across the PB, LV, and SP.

(D) The chordal graph showing the cell-to-cell interaction strength of Tnf-Tnfrsf1b among the control, model and the ART groups, colored by each cell type, the thickness degree according to the interaction strength between sender and receiver cell.

(E) The heatmap shows the relative expression level of ligand genes in various source sender cell types (left panel) and the corresponding receptor genes in various source receiver cell types in LV.

(F) The Immunohistochemical staining of MHC-I on formalin-fixed slides in LV tissues from control, model, and ART groups. Scale bar, 10 μm.

(G) IF staining of F4/80 (the marker of mature macrophages) and CD8 (the marker of CD8⁺ T cells) in LV tissues from control, model, and ART groups. Scale bar, 10 μm.

Subsequent analysis of CCI numbers across different tissues and groups revealed that LV exhibited significantly more CCI types and interactions compared with PB and SP (Figure 5B). We then identified the upregulation of several signaling pathways, including TNF, CXCL, and ITGAL-ITGB2, exclusively in LV of the model group (Figure 5C). The upregulation of ICAM-1 and TNF ligand-receptor (L-R) pairs was particularly pronounced among T/NK cells, Mono/Macro, and Neutro in the model group across all three tissues, with the most significant changes observed in the LV. Given the established roles of TNF and ICAM-1 in driving inflammatory responses,^41,42 these findings suggest that inflammatory reactions were robustly activated following PbA infection and subsequently mitigated by ART treatment. As for TNF signaling, we found that the interactions between Neutro, mast cells, Mono/Macro, and T/NK from the model group contained higher cellular interaction intensities than other cell types (Figure S7B). For specific L-R pairs, such as Tnf-Tnfrsf1b, we found that Neutro, mast cells, and Mono/Macro manifested higher molecular interaction intensity with T/NK (Figure 5D). These results demonstrate that Neutro, mast cells, and Mono/Macro mediate the CD8⁺ T cells’ activity and function in LV.

Similarly, the antigen presentation modules (e.g., MHC-I) and adhesion molecules (e.g., ICAM) were upregulated in the LV and PB of the model group compared with the control and ART groups. In LV, we identified increased communication probabilities of the H2-Q7-Cd8b1 and H2-Q4-Cd8b1 L-R pairs from Mono/Macro or Neutro to T/NK cells in the model group, suggesting enhanced activation of antigen processing and presentation pathways in the LV during PbA infection. Furthermore, L-R gene expression patterns across the three tissues revealed the upregulation of Tnf in Neutro-Act cells of the model group, while H2-Q4 and H2-Q7 were elevated in LCM_C2_Itgax, SPM_C6_Ccl5, and Neutro-Act in both the model and ART groups (Figures 5E, S7C, and S7D).

These results collectively indicated that the activation of the MHC-I pathway and inflammatory responses were heightened in myeloid and lymphocyte subtypes within LV. Immunohistochemical experiments further confirmed the activation of the MHC-I pathway in LV (Figure 5F), and IF assay revealed co-localization of F4/80⁺ macrophages and CD8⁺ T cells in LV of PbA-infected mice (Figure 5G). Together, these findings suggest that antigen processing and presentation pathways were activated following PbA infection and subsequently downregulated by ART treatment, potentially contributing to the divergent immune interaction networks across different tissues.

Identification of CD8⁺GZMB⁺ T cells upon PbA infection and ART treatment

Following the annotation of immune cell subtypes and investigation of cell-cell interaction, we focused on identifying critical cell subsets and exploring potential sources of cytokine production and cytotoxicity among myeloid and lymphocyte subpopulations. To elucidate transcriptomic changes during PbA infection and recovery post-ART treatment, we analyzed the distribution of DEGs across immune cell types, comparing the model group with the control group (model vs. control) and the model group to the ART group (model vs. ART) (Table S2).

As illustrated in Figure S8A, the Kupffer_C4_Ccr2 and LCM_C3_Ccr2 subpopulations exhibited a higher number of upregulated model vs. control DEGs while LCM_C2_Itgax and Kupffer_C3_Itgax showed more upregulated DEGs in the model vs. ART comparison. Notably, Kupffer_C4_Ccr2 and LCM_C3_Ccr2 contained a greater overlap in upregulated DEGs numbers between two comparison groups, suggesting that the upregulated DEGs of these cell types during PbA infection can be reversed after ART treatment and return to the normal state. To further assess the functional activities of myeloid subtypes, we evaluated their inflammatory and cytokine module scores. Neutro_Act and Kupffer_C4_Ccr2 exhibited elevated inflammatory scores, while mast cells and Neutro_Act showed enhanced cytokine scores compared with other myeloid subtypes (Figures S8B and S8C). These findings aligned with previous reports implicating Neutro and other effector cells in hepatic damage during anti-malarial immune responses.⁵

As for lymphocytes, the Cd8_T_C2_Gzmb (CD8⁺GZMB⁺ T cell) subtype contained higher numbers of upregulated DEGs in both model vs. control and model vs. ART comparisons with significant overlap (Figure 6A). Besides, Cd8_T_C2_Gzmb and NK also exhibited elevated cytokine and cytotoxic module scores than other subtypes (Figures 6B and 6C). Specifically, Cd8_T_C2_Gzmb cells in the model group showed increased cytotoxic module scores relative to the control group, which were subsequently reduced following ART treatment (Figure 6D). Furthermore, cytotoxicity-related genes in Cd8_T_C2_Gzmb were highly expressed in the model group within LV and SP, compared with the control and ART groups (Figure 6E).

Figure 6.

ART exerts an immunoregulatory effect on CD8⁺GZMB⁺ T cell subsets in PbA-infected mice

(A) Bubble plots show the number of upregulated DEGs in various lymphocyte types (x axis: model vs. control; y axis: model vs. ART, dot size indicates the overlapping number between two comparisons for each cell subtype).

(B and C) Violin plot showing the distribution of cytotoxic module scores (B) and cytokines module scores (C) among different lymphocyte immune cell subtypes.

(D) Violin plot showing cytotoxic scores of Cd8_T_C2_Gzmb subtype in three tissues across three groups.

(E) Heatmap shows the relative expression of cytotoxic gene set of three groups across three tissues in the Cd8_T_C2_Gzmb subtype.

(F) The trajectory inference traces a path based on pesudotime (top), cell states (middle), and group types (bottom).

(G) The heatmap reveals the relative gene expression level of four gene sets at two branches based on pseudotime trajectory, combined with corresponding GO enrichment entries for each gene cluster.

(H) The workflow chart depicts the experimental design, including flow cytometry, WB, and qPCR or enzyme-linked immunosorbent assay assays.

(I) Representative flow cytometry plots show the proportion of CD8⁺GZMB⁺ T cells from the control, model, and ART groups in LV.

(J) Bar plot shows the relative proportion of CD8⁺GZMB⁺ subtypes in T cells from each group and tissue type.

(K and L) The boxplot indicates the expression level of IFN-γ and GZMB in CD8⁺ T cells derived from the LV (K) and SP (L) tissues among the control, model, and ART groups (n = 6).

(M) The histogram indicates the expression level of IFN-γ and GZMB in CD8⁺ T cells derived from the LV tissues after in vitro among the control, model, ART (5 μM), ART (10 μM), ART (20 μM) groups (n = 6). ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

To explore the regulatory effect of ART on Cd8_T_C2_Gzmb during PbA infection, we performed pseudotime analysis to explore the transcriptional alterations. The trajectory of this subset was divided into three states: cells from the control and model groups primarily occupied state 1 and state 3, while ART-treated cells were predominantly found in state 1 and state 2 (Figure 6F). Branched expression analysis modeling revealed two distinct branches: branch 1 (state 1 to state 3) and branch 2 (state 1 to state 2). GO enrichment analysis indicated that branch 1 was associated with DNA replication, T cell receptor signaling, and T cell proliferation, suggesting that the activation and proliferation of Cd8_T_C2_Gzmb to exert cytotoxic effects against the pathogen. In contrast, enriched pathways of branch 2 were related to immune effector regulation and immune cell homeostasis, highlighting ART’s immunoregulatory role in modulating Cd8_T_C2_Gzmb activity (Figure 6G). These findings suggest that Cd8_T_C2_Gzmb plays a pivotal role during PbA infection, and ART treatment may regulate its cytotoxicity.

To validate these findings, CD8⁺GZMB⁺ T cells were isolated from PB, LV, and SP using flow cytometry and further analyzed by western blot (WB), real-time quantitative PCR (qPCR), and enzyme-linked immunosorbent assay (Figure 6H). Flow cytometry revealed a significant increase in the proportion of CD8⁺GZMB⁺ T cells in the model group across all tissues, with a notable reduction observed only in hepatic CD8⁺GZMB⁺ T cells post-ART treatment (Figures 6I, 6J, S8D, and S8E), consistent with the cellular proportion results in the scRNA-seq dataset (Figure 4C). Functional assays evaluating IFN-γ and GZMB expression in CD8⁺ T cells from LV and SP demonstrated elevated levels in the model group, indicating enhanced cytotoxic activity during PbA infection. ART treatment significantly reduced IFN-γ and GZMB expression in LV-derived CD8⁺ T cells, but not in SP-derived cells, suggesting tissue-specific immunoregulatory effects of ART (Figures 6K and 6L). Additionally, in vitro assays confirmed that ART directly inhibits the activation and cytotoxic function of CD8⁺ T cells, as evidenced by reduced GZMB expression and IFN-γ release following ART treatment (Figure 6M). These results collectively underscore the critical role of CD8⁺GZMB⁺ T cells in PbA infection and highlight ART’s ability to modulate their cytotoxic activity, particularly in LV.

ART modulates the cytotoxicity of CD8⁺GZMB⁺ T cells via the JAK2-STAT3 pathway

Transcription factor analysis of Cd8_T_C2_Gzmb in the scRNA-seq dataset demonstrated that the regulon activities of Stat3 and Lrf8 were elevated in the model group compared with the control group, but decreased following ART treatment (Figure 7A). The upregulation of Stat3 activity is known to promote immune response programs in T lymphocytes and other immune cells, contributing to inflammatory injury.⁴³ These findings suggest that Stat3 may serve as a potential therapeutic target for modulating the host immune response during PbA infection, and that ART may mitigate cytotoxicity through the suppression of Stat3. In contrast, the regulatory scores of Klf2 and Foxp1 exhibited an opposite trend. Notably, Foxp1 has been implicated in the immunomodulation of T cells, highlighting the immune imbalance induced by PbA infection. Interestingly, these transcriptional alterations were predominantly observed in LV rather than in PB or SP (Figure 7B), aligning with the more pronounced immune responses in the LV of PbA-infected mice (Figures 2A and 2G).

Figure 7.

ART treatment regulated the cytotoxicity of CD8⁺GZMB⁺ T cells via the JAK2-STAT3 signaling pathway

(A) Network visualization of core regulatory transcription factors (TFs) and its targeted differential expressed genes between model vs. control in CD8⁺GZMB⁺ T cells.

(B) The heatmap shows the relative regulon activity score of core regulatory TFs across control, model, and ART groups across three tissues.

(C) Gene set enrichment analysis plot demonstrates that ART treatment downregulated the JAK-STAT signaling pathway in CD8⁺GZMB⁺ T cells from the ART group compared with the model group.

(D) The expression of p-JAK2, JAK2, p-STAT3, and STAT3, proteins by WB assay, and quantitative statistics of the fold change of p-JAK2/JAK2 and p-STAT3/STAT3 protein expression levels of CD8⁺GZMB⁺ T cells among the three groups.

(E) The heatmap shows the relative expression of STAT3-regulated gene sets in Cd8_T_C2_Gzmb cells across control, model, and ART groups in LV.

(F) Bar plots showing the expression levels of Ccl3, Ccl4, Prdm1, and Gzmb in CD8⁺GZMB⁺ T cells of different groups from the LV tissues (n = 5).

(G) The expression of p-JAK2, JAK2, p-STAT3, and STAT3, proteins by WB assay, and quantitative statistics of the fold change of p-JAK2/JAK2 and p-STAT3/STAT3 protein expression levels of CD8⁺ T cells from LV among the three groups after ART treatment in vitro (n = 3). ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05.

To further determine whether ART regulates the transcriptional activity of Cd8_T_C2_Gzmb cells through the Stat3 pathway, we performed gene set enrichment analysis based on DEGs of ART vs. model, finding that the JAK-STAT signaling pathway in Cd8_T_C2_Gzmb was downregulated following ART treatment (NES = −1.6983, p = 0.0163) (Figure 7C). These findings were further validated by flow-sorted CD8⁺GZMB⁺ T cell subsets from LV tissues. WB analysis revealed that the expression levels of JAK2 and STAT3 remained relatively unchanged across the three groups, while the phosphorylation levels of JAK2 and STAT3 markedly increased in the model group compared with the other groups, indicating the activation of the JAK2-STAT3 signaling pathway during PbA infection (Figure 7D).

Additionally, we assessed the expression of STAT3-regulated gene sets in the LV-derived Cd8_T_C2_Gzmb cells, observing that most of the genes were upregulated in the model group compared with the control and ART groups (Figure 7E). qPCR results further confirmed that that Prdm1, Ccl3, Ccl4, and Gzmb were upregulated in LV-derived Cd8_T_C2_Gzmb cells in the model group compared with the other two groups. As shown in these results, ART reduces the cytotoxicity of Cd8_T_C2_Gzmb cells through JAK2-STAT3 signaling (Figure 7F). Moreover, the phosphorylation levels of JAK2 and STAT3 in CD8⁺ T cells post-ART-treatment in vitro were also significantly reduced, suggesting that ART directly affects the activity of the JAK2-STAT3 signaling pathway (Figure 7G).

Collectively, these results demonstrate that ART likely attenuates the cytotoxic function of CD8⁺ T cells by suppressing the JAK2-STAT3 signaling pathway. This regulatory mechanism underscores the potential of ART to modulate immune responses during PbA infection, particularly by targeting the transcriptional activity of Cd8_T_C2_Gzmb cells in LV.

Discussion

Plasmodium parasites cause tissue-specific immunopathology during the blood stage of infection, leading primarily to localized dysfunction in affected tissues. Thus, investigating immune response across multiple tissues during Plasmodium infection is crucial for revealing the immune escape or inhibition of Plasmodium, exploring the host immune homeostasis disorder, and developing effective malaria vaccines.^44,45 Numerous in vivo and in vitro studies have elucidated the activation mechanisms and protein targets of ART against Plasmodium parasites.^8,9,46,47 In addition to the direct parasite killing effect, previous studies have reported the immune regulatory effects of ART on autoimmune disease and immune dysfunction in the host.^11,48,49 We have thus hypothesized that ART might exert similar immunoregulatory functions in a Plasmodium infection murine model. In vivo experiments confirmed that ART significantly reduced parasite burden, as expected. More importantly, ART treatment alleviated immune response dysfunction and inflammatory states within LV and SP. These findings suggest that, beyond its direct parasiticidal effect, ART may modulate host inflammatory responses during PbA infection. Consequently, we further investigated the immunoregulatory role and underlying molecular mechanisms of ART in malaria-infected hosts.

Previous studies have primarily focused on specific immune cell types, such as B cells,⁵⁰ CD4⁺ memory T cells,²⁶ CD11b⁺Ly6C^hi Mono,⁵¹ Macro,⁵² and Neutro⁵ from individual tissues or organs, and the depletion or inhibition of these immune cells upon PbA infection confers protection against inflammatory injury. In the current study, by constructing a single-cell atlas from multiple tissues and group types, we comprehensively profiled the immune cellular identities including 25 myeloid and 23 lymphocyte subtypes. Our study characterizes the cellular heterogeneity that has been masked in previous experimental studies using microarray chips, bulk transcriptome sequencing, and other techniques.^53,54 The annotation results revealed the proportional changes in multiple tissues following malaria infection and ART treatment, identified immune cell subsets critical for PbA infection-induced injury that are targeted by ART, and provided valuable data resources and a theoretical basis for subsequent research.

To systematically reveal the underpinning regulatory mechanism of ART on multiple tissues of malaria-infected hosts at the single-cell level, we investigated the transcriptomic patterns and functional dynamics of various lymphocytes and myeloid cells. Among them, the CD8⁺GZMB⁺ T cells were identified as a critical lymphocyte subtype with higher cytokine and cytotoxicity module scores than other subtypes. Previous studies have reported that this cell subtype can cause multi-tissue damage via releasing GZMB in diseases such as rheumatoid arthritis, neuroinflammation, and acute myocardial infarction. Depleting or inhibiting CD8⁺GZMB⁺ T cells reduces apoptosis in injured tissues, decreases inflammation, limits tissue damage, and promotes functional recovery.^55,56,57 Another study has indicated that CD8⁺ T cells releasing GZMB induced neuronal apoptosis and contributed to nerve damage during cerebral malaria,⁵⁸ and Kaminski et al. found that CD8⁺GZMB⁺ T cells are enriched in PB of children with severe Plasmodium falciparum malaria,³⁹ supporting our results that CD8⁺GZMB⁺ T cells displayed increased abundance during PbA infection in the murine model. These observations collectively suggest that this subtype plays a critical role in malaria pathogenesis. Our investigation further revealed that ART effectively reduced the proportion of CD8⁺GZMB⁺ T cells, mitigated malaria symptoms, and restored immune cell homeostasis. Additionally, a previous study has shown that ART can suppress GZMB expression in T cells, aligning with our experimental findings⁵⁹ and suggesting that ART treatment exerts an immunoregulatory effect on this pathogenic subtype.

In terms of signaling pathways, we found that the JAK2-STAT3 pathway was over-activated in the CD8⁺GZMB⁺ T cell subtype during PbA infection, mediating the activation of the immune response program and producing inflammatory damage. Regarding the pathways identified in our CCI analysis, several studies have reported that TNF activates the JAK2-STAT3 signaling cascade via TNFR1, highlighting its critical role in JAK2-STAT3 pathway activation.^60,61,62 Our result further demonstrated that Neutro, mast cells, and Mono/Macro cells might be involved in regulating the function of CD8⁺ T cells. However, elucidating the mechanistic details of the interactions between Neutro or Mono/Macro and CD8⁺ T cells, and how these interactions lead to JAK2/STAT3 pathway activation, extends beyond the scope of this study and warrants future investigation and ex vivo validation.

Previous studies have established that the JAK2-STAT3 pathway mediates the expression of key mediators in various cancers and inflammatory processes, and its dysregulation is associated with numerous cancers and autoimmune diseases.⁶³ Within the tumor ecosystem, STAT3 over-activation plays a critical role in inhibiting immune activation regulators and promoting immunosuppressive factor production. Consequently, the US Food and Drug Administration has approved several direct or indirect STAT3 inhibitors for cancer treatment and immune regulation.⁶⁴ Regarding autoimmune diseases, STAT3 signal activation in CD8⁺ T cells, plasma cells, dendritic cells, and other immune cells has been reported to be related to the development of autoimmune diseases. In recent years, studies have found that ART and its derivatives can modulate the JAK-STAT pathway. For example, DHA can relieve arthritis symptoms in mice by acting on the HIF-1α and JAK-STAT signaling pathways,⁶⁵ while artemisinin and its derivatives inhibit the STAT3 pathway to exert an anti-tumor mechanism.⁶⁶ Our results demonstrate that ART regulates the cytotoxic activity of CD8⁺GZMB⁺ T cells by inhibiting the JAK2-STAT3 signaling pathway. However, the specific molecular targets of ART within this pathway and the resultant upstream/downstream alterations require further in-depth investigation and functional validation.

Although there are important discoveries revealed by our study, several aspects warrant future investigation. Firstly, Plasmodium infection can often infect the host’s brain, bone marrow, kidney, and other organs/tissues, causing sequential damage. This study mainly focuses on the targeted PB, LV, and SP; subsequent research should include other tissues such as the brain and kidneys for a more systematic evaluation of pathology and treatment effects. Secondly, although the PbA-infected mouse is well established for studying malaria pathology and drug efficacy, significant immunological differences exist between mice and humans. Therefore, it is essential to validate whether the functions of immune cell populations observed in mice such as CD8⁺GZMB⁺ T cells, can be recapitulated in human patients using clinical samples.⁶⁷ Thirdly, other cutting-edge techniques such as single-cell immune profiling T cell receptors and spatial transcriptomics^68,69,70 would be valuable for uncovering the intricate mechanism governing the functionality, activation, and clonality of CD8⁺GZMB⁺ T cells. Moreover, advanced methods such as surface-enhanced Raman spectroscopy show promise for drug discovery, particularly in understanding the interaction patterns of drug-protein targets.^71,72

In summary, our study elucidates the triadic interplay among malaria pathogens, the host immune system, and anti-malarial drugs. We identify critical immune cell populations and their functional characteristics responsive to ART treatment, providing novel insights into ART’s immunoregulatory mechanisms during malaria infection. These findings significantly advance our understanding of ART’s mode of action within the host and inform efforts toward more effective malaria treatment and eradication.

Resource availability

Materials availability

This study did not generate new unique materials/reagents.

Data and code availability

The single-cell transcriptomics datasets reported in this paper have been deposited in the OMIX, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (https://ngdc.cncb.ac.cn/omix: accession no. OMIX001839). The main analysis scripts used in this study are available at https://github.com/Nino5105/MO_scRNA-seq-analysis-code.

Funding and acknowledgments

We gratefully acknowledge the financial support from the National Natural Science Foundation of China (82521104, 82274182, 82441001), the Scientific and Technological Innovation Project of the China Academy of Chinese Medical Sciences (CI2023D003, CI2023E005TS05), the CACMS Innovation Fund (CI2023E002, ZG2024001-05), the National Key Research and Development Program of China (2020YFA0908000), the Establishment of Sino-Austria “Belt and Road” Joint Laboratory on Traditional Chinese Medicine for Severe Infectious Diseases and Joint Research (2020YFE0205100), the Innovation Team and Talents Cultivation Program of the National Administration of Traditional Chinese Medicine (no. ZYYZDXK-2023230), the Shenzhen Medical Research Funds (B2302051), the Distinguished Expert Project of Sichuan Province Tianfu Scholar (CW202002), and the Fundamental Research Funds for the Central Public Welfare Research Institutes (ZZ13-ZD-07, ZZ18-ND-10, ZZ17-ND-10, ZZ15-ND-10, ZZ14-ND-010, ZZ14-FL-010, ZZ14-YQ-050). We also acknowledge Tianshu Zhou from the University of Hong Kong and Yi Zhang from Zhejiang University for their important suggestions and valuable technical support.

Author contributions

J.G.W. conceived and supervised the study. Y.Y.T. supervised the study. Y.G. and W.Z. contributed to project conception, supervision, and manuscript revision. J.Y.C. contributed to study design, data analysis, and manuscript writing. P.G contributed to data analysis and manuscript writing. X.L.H. and Y.W.H. contributed to experimental validation and manuscript writing. W.Z., Y.X.L., J.Y.W., and X.H.L. contributed to data collection and experimental validation. Y.M.B., L.N.C., and C.W. contributed to data collection. G.Q.C. and X.Z. contributed to data analysis. Y.K.W., F.L.L., and C.C.X. contributed to manuscript revision. J.J.O. and Y.Q.W. contributed to experimental validation. All authors contributed to the manuscript and approved the final version.

Declaration of interests

The authors declare no competing interests.

Published Online: August 14, 2025

Lead contact website

https://scholar.google.com.sg/citations?user=ufrLUNUAAAAJ&hl=en.