Wild-type C57BL/6 mice were purchased from Guangdong Zhiyuan Biomedical Technology Co., Ltd., and TLR7^-/- C57BL/6 mice (B6.129S1-Tlr7^tm1Flv/J) were purchased from The Jackson Laboratory. All mice were fed in a specific pathogen-free micro-environment at the Experimental Animal Centre, Guangzhou Medical University. The animal study was approved by the Institutional Animal Care and Use Committee of Guangzhou Medical University (2012–11).

The Plasmodium strain used is Plasmodium yoelii NSM (P. yoelii NSM) and infected red blood cells (iRBCs) were preserved at -80°C. The frozen iRBCs were thawed in a 37°C water bath and injected into 4-week-old mice intraperitoneally and the fresh iRBCs were injected into 6 or 8-week-old experimental female mice intraperitoneally (1×10⁶ iRBCs per mouse). The uninfected and infected mice were sacrificed by cervical dislocation under anesthesia.

The scheme of R848 treatment was that mice were injected on the day before infection and 2, 5, 8, and 11 days after infection with 100 μg of R848 (Sigma) respectively. For in vivo blocking of IFN-γ, mice were injected with 100μg of anti-mouse IFN-γ (XMG1.2, BioX Cell) at 3, 6, 9, and 12 days after infection respectively.

Blood was collected from the orbit and placed into EDTA-containing tubes and serum was isolated from the blood that was preserved in the coagulant-containing tubes. The blood and serum samples were detected in the Laboratory Department of the Sixth People’s Hospital of Panyu District (Guangdong Province, China). The indexes of Hematology were detected by the Sysmex XN hematology analyzer, and the indexes of iron metabolism were detected by the Beckman Coulter AU5800 analyzer.

The spleen harvested from the mouse abdomen was ground by a sterile syringe piston and filtered with a 100-μm cell strainer. Bone marrow was obtained by flushing the femurs with RPMI 1640 medium (Gibco). Then, the cell suspension was collected and RBCs were removed by RBC lysis buffer. The remaining cells were washed by HBSS (Hank’s balanced salt solution) and resuspended in 1640 medium supplemented with 10% fetal bovine serum (FBS) (Gibco), penicillin (50 U/mL, Gibco), and streptomycin (50 mg/mL, Gibco) finally. The cells were washed with phosphate-buffered saline (PBS) and blocked with CD16/CD32 antibodies (Biolegend) for 10 minutes at 4°C and then stained with fluorescent antibodies (Biolegend) including PerCP/Cyanine5.5 anti-mouse CD71(RI7217), FITC anti-mouse TER-119/Erythroid Cells (TER-119), Alexa Fluor^® 700 anti-mouse CD45 (I3/2.3), PE/Cyanine7 anti-mouse CD45 (I3/2.3), PE/Cyanine7 anti-mouse/human CD11b (M1/70), APC/Cyanine7 anti-mouse F4/80 (BM8), PE anti-mouse TLR7 (A94B10), PE anti-mouse CD3ϵ (145-2C11), PerCP/Cyanine5.5 anti-mouse CD19 (6D5), PE/Cyanine7 anti-mouse NK-1.1 (PK136), APC anti-mouse IFN-γ (XMG1.2). Detection of CD45⁺ Ter119^- cells and CD45^- Ter119⁺ cells was performed before the lysis of RBCs. To detect intracellular cytokine IFN-γ, the cells were first stimulated with PMA (20 ng/ml, Sigma) and ionomycin (1 µg/ml, Sigma) at 37°C, 5% CO₂ for 5 h and incubated with brefeldin A (10 µg/ml, Sigma) for final 4 h to inhibit cytokines secretion. After that, the cells were stained with surface markers and fixed with fixation and permeabilization solution (BD Biosciences) for 20 min at 4°C in the dark and then permeabilized for 30 minutes at 4°C in saponin-containing buffer (BD Biosciences). IFN-γ fluorescent antibody was added to stain cells after washing with saponin-containing buffer. Annexin V-FITC (Lianke Bio) was added and incubated with cells for 10 min at room temperature. Cells were detected by flow cytometry (Beckman Coulter) and results were analyzed with the software CytoExpert 2.3 (Beckman Coulter) or FlowJo v10 (Flow Jo, LLC).

Splenocytes and RAW264.7 cells were preserved in TRIzol reagent (Invitrogen) for RNA extraction by using chloroform, isopropanol, and 75% ethanol. 1 μg of total RNA was transcribed to cDNA by using HiScript^® III RT SuperMix for qPCR (Vazyme) and quantitative real-time PCR (qRT-PCR) was performed by using ChamQ Universal SYBR qPCR Master Mix (Vazyme) and measured by CFX96 Real-Time PCR Detection System (Bio-Rad). Data were normalized to housekeeping gene β-actin. The primer sequences were: IFNg, CAAGTGGCATAGATGTGGAAGA (forward) and GTTGACCTCAAACTTGGCAATAC (reverse); Hmox1, TGATGGCTTCCTTGTACCATATC (forward) and CTCGTGGAGACGCTTTACATAG (reverse); Spic, CCAGAGCCCTGAGGAATTATG (forward) and TGCCGTGAGCATAGTGATTT (reverse); Slc40a1, TGGAACTCTATGGAAACAGCCT (forward) and TGGCATTCTTATCCACCCAGT (reverse); β-actin, CCGTAAAGACCTCTATGCCAAC (forward) and GGGTGTAAAACGCAGCTCAGTA (reverse).

Splenocytes from uninfected and infected mice were obtained according to the protocol mentioned above and CD45⁺ cells were sorted by FACS (Beckman MoFlo). After the viability detection and cell count, the RNA expression profile in each cell was detected by 10x Genomics Chromium Single Cell 3’s platform in YUANXIN Biotechnology CO., LTD (Guangzhou, China). First, the single cell was barcoded and underwent reverse transcription in oil droplets for preparation of the cDNA library by the Chromium Single Cell 3′Library & Gel Bead Kit v3. Then, libraries were sequenced on 8 lanes of Illumina Nova seq6000 using paired-end 150 bp.

All data processing and analysis were performed in YUANXIN Biotechnology CO., LTD. We used CellRanger (version 3.1.0), aligned reads on the GRCm38 reference genome for mouse and generated unique molecular identifier gene expression profiles for every single cell under standard sequencing quality threshold (default parameters). Low-quality cells were removed for downstream analysis when they met the following criteria for retaining cells: (1) ≥50,000 sequence reads; (2) ≥40% of reads uniquely aligned to the genome; (3) ≥40% of these reads mapping to RefSeq annotated exons. For further filtration, undesirable cells were removed under the criteria: (1) expressed less than 200 and more than 6000; (2) mitochondrial genome transcript was higher than 10%; (3) genes expressed in less than 3 cells; (4) more than 50000 UMI counts. Finally, 13790 cells and 18795 genes remained in data from the uninfected mouse, and 11829 cells and 19333 genes remained in data from the infected mouse.

The final filtered gene expression data matrix was normalized by using the “NormalizeData” function with default setting and 4000 highly variable genes were chosen through the “FindVariableGenes” function to be centered and scaled via the “ScaleData” function. Then, principal component analysis (PCA) was performed on the 4000 genes by the “RunPCA” function, and dimensional reduction was performed through Canonical correlation analysis (CCA) in Seurat. The “FindClusters” function was used to cluster cells and the clustered cells were shown in a uniform manifold approximation and projection (UMAP) plot.

The “FindMarkers” function of Seurat was used to identify differentially expressed genes (DEGs) by using “Wilcox” test methods and Bonferroni correction. Significant DEGs with P value (p_val) ≤ 0.05 and log processed average fold change (avg_log2FC) ≥ 0.25 were further analyzed and visualized. GO analysis and KEGG pathway enrichment analysis of these significant DEGs were performed by using the clusterProfiler package.

IRBCs were isolated from the blood that was collected from infected mice with 30% or higher parasitemia. Briefly, blood was centrifuged at 2500 rpm, 4°C for 5 minutes to remove the plasma and the remaining RBCs were washed with PBS. Then, the whole RBCs were resuspended in 1 ml PBS and overlayed onto 2 ml percoll gradient (50% and 70%) in a 15 ml centrifuge tube to centrifuge at 300 g, 4°C, ± 2 acceleration for 30 minutes. IRBCs were collected from the interface and washed with PBS. Lysates of iRBCs were harvested by three freeze-thaw cycles in liquid nitrogen and a 37°C water bath.

RAW264.7 cells were cultured in DMEM medium supplemented with 10% FBS (Gibco), penicillin (50 U/mL, Gibco), and streptomycin (50 mg/mL, Gibco) at 37°C and 5% CO₂. 2×10⁵ cells were inoculated in 48 well plates on the day before stimulation, and cells were stimulated with PBS, R848 (2 µg/ml, Sigma), mouse recombinant IFN-γ (200 ng/ml, Novoprotein), iRBCs (iRBCs: cells = 3: 1), R848 plus iRBCs, IFN-γ plus iRBCs respectively for 20 h.

According to the manufacturer’s instructions, IFN-γ levels of serum and supernatant were measured by using Mouse IFN-γ ELISA Set (BD), serum erythroferrone was measured by using Mouse Erythroferrone ELISA Kit (Bio-Swamp), and serum hepcidin was measured by using Mouse Hepcidin ELISA Kit (Sangon Biotech). Results were read at 450 nm by using a microplate reader (Moder ELX- 800, BioTek).

Uninfected RBCs (uRBCs) and iRBCs were labeled with CellTrace CFSE (Invitrogen) according to the manufacturer’s instructions. RAW264.7 cells or splenocytes were cocultured with labeled uRBCs or iRBCs for 5 h and were detected by flow cytometry.

For detection of Fe²⁺ in cells, splenocytes or RAW264.7 cells were incubated with FerroOrange (1µM, DOJINDO) following either infection in vivo or stimulation in vitro at 37°C in dark. Cells were washed twice with PBS and then stained with surface markers if necessary. For detection of lipid peroxidation in cells, RAW264.7 cells were incubated with C11-BODIPY 581/591 (2 µM, Invitrogen) following stimulation in vitro at 37°C in dark, then, cells were washed twice with PBS containing 1% FBS. To measure reactive oxygen species (ROS) level in cells, RAW264.7 cells were incubated with DCFH-DA (10 µM, Beyotime) following stimulation in vitro at 37°C in dark, then, cells were washed twice with PBS containing 1% FBS. Cells were immediately detected by flow cytometry at last.

Cells collected from the spleen, bone marrow, or blood were washed twice with PBS containing 1% BSA and stained with CD45, TER-119, and CD71 fluorescent antibodies. Then, cells were incubated with Hoechst 34580 (0.5 µg/ml, BD) at room temperature in dark for 30 min to be measured by flow cytometry. In CD45^- Ter119⁺ cells, CD71^- indicated mature and enucleated erythroid cells, and CD71⁺ indicated less mature erythroid progenitor cells that may contain DNA molecules. Hoechst, a DNA dye with cell- permeability, could stain parasite and erythroblast DNA. In CD45^- Ter119⁺ CD71^- population, uninfected mature erythroid cells without DNA molecules are Hoechst^neg (CD45^- Ter119⁺ CD71^- Hoechst^neg), infected mature erythroid cells that only contain parasite DNA are Hoechst^pos (CD45^- Ter119⁺ CD71^- Hoechst^pos). In CD45^- Ter119⁺ CD71⁺ population, uninfected and relatively mature erythroid progenitor cells without DNA molecules are Hoechst^neg (CD45^- Ter119⁺ CD71⁺ Hoechst^neg), uninfected and less mature erythroid progenitor cells that only contain erythroblast DNA are Hoechst^low (CD45^- Ter119⁺ CD71⁺ Hoechst^low), infected and less mature erythroid progenitor cells that contain erythroblast and parasite DNA are Hoechst^hi (CD45^- Ter119⁺ CD71⁺ Hoechst^hi).

8 µm sections of the spleen were washed with distilled water before staining. The sections were stained with Perls’ Prussian Blue for 40 min and then stained with Neutral Red for 2 min. After staining, hemosiderin or trivalent iron (Fe³⁺) is blue, and the nucleus and other tissues are red. Stained sections were scanned by a Digital Section Scanner (Leica). The percentage of blue area in the section is calculated by using ImageJ (V1.8.0).

Unpaired t-tests test was used for comparisons between the two groups with equal variance and normal distributions and ANOVA with Sidak multiple comparisons test was used for comparisons between more than two groups with normal distributions and equal variance, unless differently specified. When various groups have unequal variance, ANOVA with Tamhane T2 was used to analyze. Statistical differences were defined as P < 0.05. All data were analyzed in GraphPad Prism (v9).

Our previous finding indicates that infection of P. yoelii NSM could stimulate a T cell-mediated inflammatory response in C57BL/6 mice, and unlike P. yoelii XL, P. yoelii NSM usually induces a non-lethal infection in C57BL/6 mice (31), which is an appropriate model for an investigation into malaria-induced splenic erythropoiesis. After being infected with P. yoelii NSM, parasitemia, the number of RBCs, spleen weight, and body weight were monitored in infected mice for 28 days, showing the most serious condition at 12-20 day post-infection (dpi.). At 16 dpi., while parasitemia (26.36%), spleen weight (1.01 g), and the ratio of spleen weight over body weight (0.0555), were the highest relative to uninfected status the number of RBCs in the blood (3.775×10¹²/L) was the lowest relative to uninfected status ( Figures 1A–D ). To examine changes in splenic cellularity during the serious period, flow cytometry was used to distinguish immune cells (CD45⁺ Ter119^-) and more mature developing erythroid cells (CD45^- Ter119⁺) in the spleen ( Figure 1E ), and the results showed that the percentage of splenic CD45⁺ Ter119^- cells decreased but the percentage of splenic CD45^- Ter119⁺ cells increased at 16 dpi. ( Figures 1F, G ). The absolute number of splenic CD45⁺ Ter119^- cells and splenic CD45^- Ter119⁺ cells increased dramatically ( Figures 1H, I ). CD71, a transferrin receptor expressed by erythroid progenitor cells, declines as the development of erythroid progenitor cells and is lost on the surface of mature erythrocytes (32). About 70% of splenic CD45^- Ter119⁺ cells in spleens that were observed in infected mice were marked with CD71 ( Figure 1J ). These findings suggested that erythroid cells accumulated in the spleen of P. yoelii NSM-infected mice, which is likely a key contributor to splenomegaly, and that the accumulation of erythroid cells in the spleen is a potential outcome of extramedullary splenic erythropoiesis. Therefore, 12-20 dpi is an appropriate time point to investigate the regulation of the extramedullary splenic erythropoiesis, and the following experiments were performed at this time point.

As inflammation is a critical factor in stress erythropoiesis, the increased level of TLR7 expression in splenic CD45⁺ cells after the infection ( Figures 2A, B ) triggers our interest in studying the role of the innate immune response induced by TLR7 in extramedullary splenic erythropoiesis. Wild-type and TLR7 ^-/- mice were infected with P. yoelii NSM and we found that the parasitemia of TLR7 ^-/- infected mice seemed more severe ( Supplementary Figure S1A ). Meanwhile, TLR7 ^-/- infected mice exhibited more severe malaria-induced anemia. The amount of RBC in the blood was reduced more dramatically in TLR7 ^-/- infected mice, although there was no difference in hemoglobin and hematocrit between these two groups of infected mice ( Supplementary Figures S1B-D ). Besides, mean corpuscular volume and red blood cell distribution width were higher but mean corpuscular hemoglobin concentration was lower in TLR7 ^-/- infected mice ( Supplementary Figures S1E-H ). Similar to wild-type mice, the percentage of splenic CD45⁺ Ter119^- cells was decreased and the percentage of splenic CD45^- Ter119⁺ cells was increased in TLR7 ^-/- mice ( Figures 2C-E ). The absolute number of CD45⁺ Ter119^- cells or CD45^- Ter119⁺ cells in the spleen increased, but the absolute number of CD45^- Ter119⁺ cells was larger than CD45⁺ Ter119^- cells ( Figures 2F, G ), which was consistent with that the expansion of erythroid cells primarily contributed to splenomegaly. The absolute number of CD45^- Ter119⁺ cells in TLR7 ^-/- mice was less than in wild-type mice ( Figure 2G ). For understanding which stage of erythroid progenitor cells was affected, we distinguished the serial differentiation of ProE (CD71^high Ter119^low), EryA (CD71^high Ter119^high FSC^high), EryB (CD71^high Ter119^high FSC^low) and EryC (CD71^low Ter119^high FSC^low) based on the expression level of Ter119 and CD71 and cell volume ( Figure 2H , Supplementary Figure S2 ) (19, 32). After the infection, the percentage of splenic ProE or splenic EryA in wild-type mice was higher than in TLR7 ^-/- mice ( Figures 2I, J ), however, the percentage of splenic EryB in wild-type mice and TLR7 ^-/- mice were similar ( Figure 2K ), resulting in a lower percentage of splenic EryC in wild-type mice than in TLR7 ^-/- mice ( Figure 2L ).

Similarly, the erythropoiesis response was stimulated in bone marrow (BM) after the infection in our model. The percentage of BM CD45⁺ Ter119^- cells was decreased and the percentage of BM CD45^- Ter119⁺ cells was increased after the infection ( Supplementary Figures S3A-C ). The percentage of BM ProE or BM EryA increased after the infection ( Supplementary Figures S3D, E ). The percentage of BM EryB was not altered after the infection ( Supplementary Figure S3F ). The percentage of BM EryC was reduced after the infection ( Supplementary Figure S3G ). Nevertheless, we did not observe any difference in bone marrow erythropoiesis response between wild-type and TLR7 ^-/- mice following the infection. PS exposure at the cell surface of RBCs could be induced by Plasmodium and facilitate phagocytosis of RBCs by macrophages that display a series of receptors for PS, which is considered to be a secondary mechanism of removal of RBCs (23, 33, 34). An increase in the percentage of annexin V⁺ in BM CD45^- CD71⁺ cells (erythroid progenitor cells) was observed after the infection ( Supplementary Figure S3H ), which could not be observed in the spleen ( Figure 2M ), indicating that splenic erythropoiesis is more efficient than bone marrow erythropoiesis. Additionally, the expression of TLR7 could not be detected in CD45^- cells ( Supplementary Figure S3I ). These demonstrated that TLR7 modulates splenic erythropoiesis but not bone marrow erythropoiesis through an indirect way after the infection.

Wild-type infected mice were treated with R848 to highlight the impact of TLR7 on splenic erythropoiesis. Treatment of R848 did not affect the percentage of splenic CD45⁺ Ter119^- cells or splenic CD45^- Ter119⁺ cells ( Supplementary Figures S4A-C ). Though treatment of R848 did not regulate the absolute number of splenic CD45⁺ Ter119^- cells, it significantly increased the absolute number of splenic CD45^- Ter119⁺ cells ( Supplementary Figures S4D, E ). Moreover, the percentage of ProE was increased by treatment of R848 ( Supplementary Figure S4F ), but EryA, EryB, and EryC were not the targets regulated by R848 ( Supplementary Figures S4G-I ).

The spleen is responsible for the clearance of iRBCs (35). To investigate whether iRBCs that were sequestered in the spleen contributed to the increased percentage of CD45^- Ter119⁺ cells, the percentage of infected CD45^- Ter119⁺ cells were measured according to Hoechst 34580 fluorescence in cells ( Supplementary Figure S5A ) (36, 37). Most of the infected CD45^- Ter119⁺ cells in the spleen or bone marrow were marked with CD71, which could not be observed in the blood ( Supplementary Figure S5A ). Infected CD45^- Ter119⁺ cells were just a small portion of erythroid cells in the spleen or bone marrow, although they were higher than the percentage in the blood ( Supplementary Figures S5B, C ). It suggested that the erythroid progenitor cells in the spleen were newly formed. Therefore, extramedullary splenic erythropoiesis was induced after the infection of P. yoelii NSM.

It’s reported that the TLR7 activation could stimulate Th1 cell response, inducing a higher percentage of IFN-γ⁺ CD4⁺ T cells in the spleen at 15 dpi. in P. chabaudi-infected mice (38). Given that, we hypothesized that the production of IFN-γ could be influenced by TLR7 signaling after Plasmodium infection. In our experiments, an increase in IFNg mRNA in splenocytes by Plasmodium infection was abolished by the knockout of TLR7 ( Figure 3A ). The production of IFN-γ in different splenic populations including immune cells (CD45⁺), T cells (CD45⁺, CD3⁺), NK cells (CD45⁺, CD3^-, NK1.1⁺), NKT cells (CD45⁺, CD3⁺, NK1.1⁺) and macrophages (CD45⁺, CD11b⁺, F4/80⁺) were measured by flow cytometry ( Figure 3B ). TLR7 ^-/- mice displayed a lower level of the percentage of splenic CD45⁺ IFN-γ⁺ cells than wild-type mice after the infection, which was a consequence of reduced production of IFN-γ in splenic T cells ( Figures 3C-G ). Additionally, wild-type infected mice exhibited a higher level of serum IFN-γ than TLR7 ^-/- infected mice ( Figure 3H ).To demonstrate that iRBCs components can activate TLR7 to induce the production of IFN-γ, uRBCs lysates, iRBCs lysates, and R848 that was positive control were used to stimulate splenocytes from wild-type and TLR7 ^-/- uninfected mice for 48 h in vitro respectively. R848 and iRBCs lysates only stimulated splenocytes from wild-type mice to produce a great amount of IFN-γ ( Figure 3I ). All of these indicated that IFN-γ response is a downstream event of TLR7 activation.

Single-cell RNA sequencing of splenic CD45⁺ cells from wild-type uninfected and infected mice was performed to find the mediators that IFN-γ utilizes to regulate extramedullary splenic erythropoiesis. Splenic CD45⁺ cells were classified into 12 clusters including B cells, CD4⁺ T cells, CD8⁺ T cells, DCs, erythrocytes, macrophages, NK cells, γδT cells/NKT cells, neutrophils, plasma cells, proliferating B cells and proliferating T cells and all clusters were displayed on UMAP plot ( Figure 4A ). Consistent with the results detected by flow cytometry ( Figure 3B ), high expression of IFN-γ was observed in T cells, NK cells, and NKT cells after the infection ( Figure 4B ). IFNGR is a heterodimeric complex that is comprised of IFNGR1 and IFNGR2. Macrophages were identified as the cluster expressing high levels of IFNGR1 and IFNGR2 ( Figures 4C, D ), then we investigated if macrophages responded to IFN-γ and were regulated by IFN-γ after the infection. There were 474 differentially expressed genes including 213 upregulated genes and 261 downregulated genes that were found in splenic macrophages from wild-type infected mice compared with wild-type uninfected mice ( Figures 4E, G ). To identify the biological function of upregulated genes, GO enrichment analysis was performed and revealed that some of the upregulated genes involved in “respond to IFN-gamma”, “iron ion homeostasis” and “cellular iron ion homeostasis” pathways ( Supplementary Figures S6A-C ). All the upregulated genes involved in “iron ion homeostasis” or “cellular iron ion homeostasis” pathways were labeled ( Figure 4E ), and the heatmap showed that macrophages from wild-type infected mice expressed higher levels of these genes ( Figure 4F ). After the infection, serum iron, transferrin saturation (TS), and total iron binding capacity (TIBC) were increased, however, there was no difference between wild-type and TLR7 ^-/- infected mice ( Supplementary Figures S6D-F ). Hepcidin, a negative regulator of iron export by suppressing ferroportin, and erythroferrone, a positive regulator of iron export by inhibiting hepcidin synthesis, are involved in the iron metabolism but serum erythroferrone and hepcidin were not altered after the infection ( Supplementary Figures S6G, H ). Notably, after the infection, iron was accumulated and stored in the spleen, especially in wild-type infected mice ( Figures 5A, B ), which may be a major source of iron for splenic erythropoiesis in our model. Besides, Fe²⁺ content of macrophages or erythroid cells was detected with FerroOrange. After the infection, the iron content of newly produced splenic ProE was not changed ( Figure 5C ) but the iron content of the CD45^- Ter119⁺ population (including EryA, EryB, and EryC) was increased ( Figure 5D ). And the iron content of the CD45^- Ter119⁺ population was more abundant in wild-type infected mice than in TLR7 ^-/- infected mice ( Figure 5D ). Meanwhile, the accumulation of iron in red pulp macrophage (RPM) or splenic macrophage was more significantly increased in wild-type infected mice than in TLR7 ^-/- infected mice ( Figures 5E, F ). These findings indicated a hint that upregulated iron metabolism in splenic macrophages is a possible link between IFN-γ and extramedullary splenic erythropoiesis.

RAW264.7 were cocultured with iRBCs to illustrate the regulation of macrophages by IFN-γ. Before RAW264.7 were cocultured with CFSE-labelled uRBCs or iRBCs, the efficiency of labeling in uRBCs or iRBCs was assessed. All uRBCs and iRBCs were labeled with CFSE successfully ( Supplementary Figure S7A ) and iRBCs were more sensitive to being phagocytosed by RAW264.7 than uRBCs in vitro ( Supplementary Figures S7B, C ). When cocultured with iRBCs, IFN-γ could significantly enhance the phagocytosis of iRBCs by RAW264.7 ( Figures 6A, B ). After phagocytosis, iRBCs are digested by ROS and hydrolytic enzymes, and then heme is released from hemoglobin (39). Subsequently, heme oxygenase-1 (HO-1) cleaves heme to produce equimolar amounts of iron, carbon monoxide, and biliverdin (39). Therefore, Fe²⁺ was increased in RAW264.7 after phagocytosis of iRBCs, which could be more significant with the stimulation of IFN-γ ( Figures 6C, D ). Ferroptosis, a type of regulated necrosis dependent on iron, could be indicated by the levels of annexin V, ROS, and lipid peroxidation, and ferroptosis of RPM was induced by increased phagocytosis of RBCs in the murine model of transfusion (40). In our model, macrophages were responsible to phagocytose iRBCs. IRBCs, but not uRBCs, could induce a high level of annexin V, ROS, and lipid peroxidation in RAW264.7 in vitro, indicating the ferroptosis of RAW264.7 ( Supplementary Figures S7D-I ). It’s likely that ferroptosis of RAW264.7 was induced by phagocytosis of iRBCs and the accumulation of iron. As IFN-γ augmented the phagocytosis of iRBCs by RAW264.7, higher levels of annexin V and lipid peroxidation were observed in RAW264.7 ( Figures 6E–J ). However, the level of ROS in RAW264.7 was not increased by enhanced phagocytosis of iRBCs ( Figures 6F, I ). HO-1 that cleaves heme to release iron, Spic that is a transcription factor to promote Slc40a1 expression, and Slc40a1 that is the sole mammalian non-heme iron exporter, were associated with iron metabolism and upregulated in macrophages after the infection ( Figures 4E, F ). The relative mRNA levels of HO-1, Spic, and Slc40a1 were measured to highlight the iron metabolism in RAW264.7 after phagocytosis. Treated with IFN-γ, relative mRNA levels of HO-1 and Slc40a1 in RAW264.7 that were cocultured with iRBCs were increased, and the relative mRNA level of Spic was decreased because IFN-γ can suppress the expression of Spic ( Figures 6K-M ). These findings indicated that IFN-γ enhances phagocytosis of iRBCs by RAW264.7 in vitro, which may facilitate iron metabolism subsequently.

To emphasize the impact of IFN-γ on the splenic extramedullary erythropoiesis in wild-type infected mice, neutralization of IFN-γ in vivo was performed after the infection. Neutralization of IFN-γ did not affect the percentage of splenic CD45⁺ Ter119^- cells or splenic CD45^- Ter119⁺ cells in wild-type infected mice ( Figures 7A-C ). And neutralization of IFN-γ did not alter the absolute number of splenic CD45⁺ Ter119^- cells or CD45^- Ter119⁺ cells ( Figures 7D, E ). Contrary to the treatment of R848, neutralization of IFN-γ reduced the percentage of ProE in wild-type infected mice ( Figure 7F ), while the percentages of EryA, EryB, and EryC were not altered ( Figures 7G-I ). Previous reports suggested that IFN-γ could enhance the phagocytosis of RBCs by macrophages (29, 30). We also confirmed that IFN-γ could enhance phagocytosis of iRBCs by RAW264.7 in vitro ( Figures 6A, B ). Associatively, after neutralization of IFN-γ, serum IFN-γ was reduced significantly ( Figure 7J ), and a slight increase in the percentage of infected CD45^- Ter119⁺ cells in the spleen was observed ( Figures 7K, L ). Given that phagocytosis of RBCs is of great importance in iron recycling and iron homeostasis (41), Perls’ Prussian blue staining was performed to assess the accumulation of iron in the spleen. The result showed that the accumulation of iron in the spleen decreased after the neutralization of IFN-γ ( Figures 7M, N ). All of these suggested that the neutralization of IFN-γ significantly reduced the iron accumulation in the spleen and slightly hampered the extramedullary splenic erythropoiesis.

Splenic macrophages expanded dramatically after the infection and displayed a higher level of TLR7 expression ( Supplementary Figures S8A-D ), indicating that macrophages responded to TLR7 ligand as well. Given the effect of Myd88 on erythrophagocytosis (14), TLR7 signaling may regulate the functionality of macrophages. Compared with splenic macrophages from TLR7 ^-/- mice, splenic macrophages from wild-type mice phagocytosed iRBCs more powerfully in vitro ( Supplementary Figures S8E, F ). Additionally, R848 could significantly enhance the phagocytosis of iRBCs by RAW264.7 in vitro ( Supplementary Figures S9A, B ). However, stimulation with R848 could not alter the level of annexin V, ROS, and lipid peroxidation in RAW264.7 that cocultured with iRBCs ( Supplementary Figures S9C-H ). And stimulation with R848 did not alter HO-1, Spic, and Slc40a1 expression in RAW264.7 that cocultured with iRBCs ( Supplementary Figures S9I-K ). These suggested that TLR7 signaling could also promote the phagocytosis of iRBCs in macrophages and RAW264.7.