CXCR4high megakaryocytes regulate host-defense immunity against bacterial pathogens

Megakaryocytes (MKs) continuously produce platelets to support hemostasis and form a niche for hematopoietic stem cell maintenance in the bone marrow. MKs are also involved in inflammatory responses; however, the mechanism remains poorly understood. Using single-cell sequencing, we identified a CXCR4 highly expressed MK subpopulation, which exhibited both MK-specific and immune characteristics. CXCR4high MKs interacted with myeloid cells to promote their migration and stimulate the bacterial phagocytosis of macrophages and neutrophils by producing TNFα and IL-6. CXCR4high MKs were also capable of phagocytosis, processing, and presenting antigens to activate T cells. Furthermore, CXCR4high MKs also egressed circulation and infiltrated into the spleen, liver, and lung upon bacterial infection. Ablation of MKs suppressed the innate immune response and T cell activation to impair the anti-bacterial effects in mice under the Listeria monocytogenes challenge. Using hematopoietic stem/progenitor cell lineage-tracing mouse lines, we show that CXCR4high MKs were generated from infection-induced emergency megakaryopoiesis in response to bacterial infection. Overall, we identify the CXCR4high MKs, which regulate host-defense immune response against bacterial infection.

MK-biased hematopoietic stem cells (HSCs) induce emergency megakaryopoiesis to actively generate MKs upon acute inflammation, which can efficiently replenish the platelet loss during inflammatory insult (Haas et al., 2015). Studies suggested that MKs might participate in immune responses independent of their platelet generation role (Cunin and Nigrovic, 2019). MKs express multiple immune receptors, such as IgG Fc receptors and toll-like receptors (TLRs), enabling them to sense inflammation directly (Cunin and Nigrovic, 2019). Mature MKs also express major histocompatibility complex (MHC) to activate antigen-specific CD8 + T cells and enhance CD4 + T cells and Th17 cell responses through stimulating antigen processing (Finkielsztein et al., 2015;Pariser et al., 2021;Zufferey et al., 2017). Furthermore, MKs release multiple cytokines and chemokines to influence immune cells. For example, MKs produce IL-1α and IL-1β to promote arthritis susceptibility in mice resistant to arthritis (Cunin et al., 2017) and produce CXCL1 and CXCL2 to promote neutrophil efflux from the bone marrow (Köhler et al., 2011). Lung MKs contribute to thrombosis (Lefrançais et al., 2017) and, more interestingly, participate in immune responses (Pariser et al., 2021), although the relationship between lung MKs and bone marrow circulating MKs (Nishimura et al., 2015) remains unexplored. Furthermore, the recent single-cell atlas shows that MKs are heterogeneous and contain subpopulations that express multiple immune genes and are involved in inflammation response Pariser et al., 2021;Sun et al., 2021;Yeung et al., 2020). Here, by combining scRNA-seq with functional assays, we identified a CXCR4 high MK population, which was generated by infectioninduced emergency megakaryopoiesis, and stimulated innate immunity against bacterial infection.

Single-cell atlas identifies an immune-modulatory subpopulation of MKs
We applied droplet-based scRNA-seq with CD41 + forward scatter (FSC) high bone marrow MKs to explore the MK heterogeneity ( Figure 1A; Figure 1-figure supplement 1A,B). To enrich accurate MKs, we further performed transcriptomic profile analysis in the phenotypically enriched MKs (Yeung et al., 2020). Our scRNA-seq successfully detected 5368 high-quality cells (Figure 1-figure supplement 1B,C), in which one MK cluster (1712 cells) and six immune cell clusters (3656 cells) were annotated according to their gene profile ( Figure 1-figure supplement 1D-G) and the alignment with published scRNA-seq data (Almanzar et al., 2020;Hamey et al., 2021;Pariser et al., 2021;Xie et al., 2020;Yeung et al., 2020). Our annotated MKs were similar to MKs but distinct to immune cells, including myeloid progenitors, basophils, neutrophils, monocytes, dendritic cells, macrophages, B cells, and T cells, in an integrated scRNA-seq analysis platform (Figure 1-figure supplement 2). Therefore, we re-clustered the transcriptionally enriched 1712 MKs into five subpopulations, termed MK1 to MK5 ( Figure 1B; Figure 1-figure supplement 3A,B), which were further confirmed by the integrated scRNA-seq analysis platform to rule out the potential immune cell contamination (Pariser et al., 2021;Xie et al., 2020;Yeung et al., 2020;Figure 1-figure supplement 3C,D). We noticed that mature MKs with huge sizes were captured at a relatively low rate, potentially due to the limitation in current techniques in cell purification and single-cell preparation Sun et al., 2021).
Enriched signature genes by Gene Ontology exhibited that MK1 and MK2 highly expressed nuclear division, DNA replication and repair genes for endomitosis ( Figure 1C,D). MK3 enriched blood coagulation and thrombosis genes for platelet generation ( Figure 1C    and immune MKs (Ccl3,Cd52,Selplg,Sell,Adam8) Figure 1figure supplement 5B). We also noticed that MK5 highly expressed Cxcr4 than other MK subpopulations ( Figure 1H,I), although most MKs express CXCR4 (Hamada et al., 1998;Figure 1-figure supplement 6A). To confirm this, we found that CXCR4 high MKs expressed MK markers (  , we further explored the role of CXCR4 high MKs, which enriched MK5, in regulating myeloid immune cells, in regulating the innate immunity function of myeloid cells against pathogens. We challenged mice with Listeria (L.) monocytogenes, a Gram-positive facultative intracellular bacterium (Bishop and Hinrichs, 1987;Edelson and Unanue, 2000), which induce myelopoiesis (Eash et al., 2009;Figure 2-figure supplement 1A, B). Interestingly, we noticed that CXCR4 high MKs were more dramatically associated with myeloid cells in the bone marrow of mice 3 days after L. monocytogenes infection, which was a significant increase than the association between myeloid cells and CXCR4 low MKs or the association between randomly placed myeloid cells and CXCR4 high MKs (Figure 2A,B). The myeloid cell-CXCR4 high MK association (mean distance 15.36 μm) was significantly closer than the myeloid cell-CXCR4 low MKs association ( Figure 2C; mean distance 25.62 μm, p=7.0 × 10 -4 by KS test), and the association between randomly placed myeloid cells and CXCR4 high MKs [35.37 μm, p (μ<15.36)=1.8 × 10 -10 ] in the bone marrow of mice 3 days after L. monocytogenes infection. Whereas the observed mean distance of myeloid cells to CXCR4 low MKs (25.62 μm) is not different from random simulations [27.76 μm, p (μ<25.62)=0.14] ( Figure 2C). This suggested that the increased association between myeloid cell-CXCR4 high MK may not be due to the infection-induced expansion of myeloid cells. Furthermore, we did not observe a significant association between myeloid cells and MKs during homeostasis ( Figure 2-figure supplement 1C,D). We also noticed that bone marrow myeloid cells were preferably adjacent to the CXCR4 high MK-blood vessel intersection in mice 3 days after L. monocytogenes enriched gene number of each term. (E-F) UMAP visualization (E) and statistical analysis (F) of cytokine score (left), inflammatory score (middle) and chemokine score (right) in MK1 to 5. (G) Dotplots of significant cytokine ligand (source) -receptor (target) interactions between MKs and immune cells discovered. The color indicates the means of the receptor-ligand pairs between two cell types and bubble size indicates p values. Mon, monocytes; MΦ, macrophages (Dong et al., 2020); DC, dendritic cells; Neu, neutrophils; MP, myeloid progenitors; T, T cells; B, B cells. (H-I) Violin plot (H) and feature plot (I) of selected signature genes of MK5. Red lines in (H) indicate the median gene expression. Repeated-measures one-way ANOVA followed by Dunnett's test for multiple comparisons in (F), ǂ ǂ <0.01, ǂ ǂ ǂ p<0.001.
The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Signature genes of MK1 to MK5.
Source data 2. The means and p value of the average expression level of interacting molecule 1 in cluster 1 and interacting molecule 2 in cluster 2 by CellPhoneDB.       To explore how CXCR4 high MKs regulate myeloid cells, we interestingly found that CXCR4 high MKs, but not CXCR4 low MKs, effectively promoted myeloid cell mobilization in our transwell assays ( Figure 2D). Furthermore, we asked whether CXCR4 high MKs regulate myeloid cell function against pathogens. To this aim, we incubated purified CXCR4 low MKs and CXCR4 high MKs with neutrophils or macrophages for bacterial phagocytosis analysis. We found that CXCR4 high MKs, but not CXCR4 low MKs, efficiently enhanced the bacterial phagocytosis of neutrophils and macrophages ( Figure  Our scRNA-seq also exhibited that the high expression of Cxcr4 was positively correlated with immune cell-stimulating cytokines, such as Ccl6, Tnf, and Il6 (Li et al., 2018;Rothe et al., 1993;Shapouri-Moghaddam et al., 2018) in MKs ( Figure 2I). In line with this, CXCR4 high MKs had higher TNFα and IL-6 protein levels than CXCR4 low MKs ( Figure 2J; Figure 2-figure supplement 2C,D). The TNFα and IL-6 levels in CXCR4 high MKs were comparable to macrophages from mice 3 days after L. monocytogenes infection (Figure 2-figure supplement 2E), which are known as the primary cellular source of TNFα and IL-6 upon infection (Shapouri-Moghaddam et al., 2018). These observations suggested that CXCR4 high MKs might stimulate myeloid cell phagocytosis by producing TNFα and IL-6. Indeed, anti-TNFα and anti-IL-6 blocking antibodies compromised the role of CXCR4 high MKs in stimulating bacterial phagocytosis of neutrophils and macrophages ( Figure 2K-N).

CXCR4 high MKs stimulate host-defense immunity against bacterial pathogens
To explore the in vivo role of MKs upon L. monocytogenes infection in mice, we employed Pf4 Cre ; Rosa26 fs-iDTR mice, in which MKs were rendered sensitive to diphtheria toxin (DT) (Zhao et al., 2014;Figure 3A and B). MK ablation increased the number of hematopoietic stem and progenitor cells and myelopoiesis in the bone marrow upon infection (Figure 3-figure supplement 1A-D). Notably, MK ablation dramatically increased the bacterial burdens in the liver and spleen 3 days after L. monocytogenes infection ( Figure 3C). We also found that MK ablation reduced the number of myeloid cells, including monocytes, macrophages, dendritic cells (DCs), and neutrophils, in the liver and spleen ( Figure 3D,E; Figure 3-figure supplement 1E,F), suggesting the role of MKs in promoting myeloid cells against pathogens. We further investigated whether MKs regulate adaptative immunity against pathogen infection. Interestingly, we noticed that CXCR4 high MKs were able to phagocytose bacteria and presented the ovalbumin (OVA) antigens on their surface via MHC-I ( Figure 3F,G). Furthermore, cells normalized to Ctrl (without MKs in the lower chambers) as indicated by transwell assays. Mon, monocytes; MΦ, macrophages; DC, dendritic cells; Neu, neutrophils. (E-F) Representative images (E) and quantification (F) of neutrophil phagocytosis capacity with or without MK co-culture as indicated. CD11b, red; E. coli, green; DAPI, blue. Ctrl, neutrophil without MK co-culture. (G-H) Representative images (G) and quantification (H) of macrophage phagocytosis capacity with or without MK co-culture as indicated. F4/80, red; E. coli, green; DAPI, blue. Ctrl, macrophage without MK co-culture. (I) Spearman correlation analysis between expression profiles of Cxcr4 and feature genes in MK subpopulations. (J) Quantification of TNFα + and IL-6 + cells in CXCR4 low MKs and CXCR4 high MKs from control mice or mice 3 days after L. monocytogenes infection. (K-L) Quantification of neutrophil (K) and macrophage (L) phagocytosis with or without CXCR4 high MK co-culture in the absence or presence of anti-TNFα or anti-IL-6 neutralizing antibodies. (M-N) Quantification of the phagocytosis abilities by neutrophils (M) and macrophages (N) with or without CXCR4 low MK or CXCR4 high MK co-culture in the absence or presence of anti-TNFα or anti-IL-6 neutralizing antibodies by flow cytometry. Ctrl, neutrophils (M) or macrophages (N) without MKs co-culture. Scale bars, 20 μm (A, E, G). Data represent mean ± s.e.m (D) and boxplots show medians, first and third quartiles (F, H, K-L). Repeatedmeasures one-way ANOVA followed by Dunnett's test for multiple comparisons in (D, M, N), ǂ p<0.05, ǂ ǂ p<0.01, n.s., not significant. A two-sample KS test was performed to assess statistically significant (C, F, H, K, L), n.s., not significant. Paired Student's t-test was performed to assess statistical significance (J), # p<0.05, ## p<0.01, n.s., not significant.
The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Distance of actual myeloid cells to the closest CXCR4 low and CXR4 high MKs 3 days after L. monocytogenes infection.
Source data 2. Distance of randomly positioned myeloid cells to the closest CXCR4 low and CXR4 high MKs 3 days after L. monocytogenes infection.
Source data 3. Mean distance of randomly positioned myeloid cells to the closest CXCR4 low and CXR4 high MKs 3 days after L. monocytogenes infection from 500 times simulations.    OVA antigens presented by CXCR4 high MKs activated OT-I CD8 + T cells ( Figure 3H) and B3Z T cells ( Figure 3-figure supplement 2), a T cell hybridoma which expresses TCR that specifically recognizes OVA (Karttunen et al., 1992). We challenged Pf4 Cre ; Rosa26 fs-iDTR mice with OVA-expressing recombinant microbe (L. monocytogenes-OVA). Seven days after L. monocytogenes-OVA infection, splenocytes from control or MK ablated mice were re-stimulated with OVA peptide in vitro to assess OVA-specific T cell activation ( Figure 3I). Notably, MK ablation dramatically reduced the number of CD4 + IFNγ + Th1, CD4 + IL4 + Th2, and CD8 + cytotoxic T lymphocytes but did not impact the total number of CD4 + T cells and CD8 + T cells ( Figure 3J). These observations demonstrated that MKs regulate host-defense immunity against L. monocytogenes infection. To explore whether CXCR4 high MKs contribute to the immune response against bacterial pathogens, we infused the purified CXCR4 high MKs and CXCR4 low MKs into MK ablation mice during L. monocytogenes infection ( Figure 3K). Notably, we found that the infusion with CXCR4 high MKs, but not CXCR4 low MKs, partially rescued the bacterial clearance defect in MK ablation mice ( Figure 3L). This is potentially due to the reduced platelets known for regulating immune responses (Semple et al., 2011).

Bacterial infection stimulates the migration of CXCR4 high MKs
High Cxcr4 expression indicated that CXCR4 high MKs might migrate between the bone marrow microenvironment and circulation in response to infection (Suraneni et al., 2018). In line with this, our spatial distribution analysis showed that ~80% of MKs directly contacted blood vessels 3 days after L. monocytogenes infection, which was much higher than in control mice (~40%) ( Figure 4A (Bruns et al., 2014;Itkin et al., 2016;Zhao et al., 2014)  To further explore the dynamic migration of MKs upon pathogen infection, we adapted an ex vivo real-time imaging method to trace MK migration in the bone marrow (Xie et al., 2009). Using Pf4 Cre ; Rosa26 fs-tdTomato mice and ex vivo live imaging approach, we observed that small tdTomato + MKs rapidly migrated into sinusoids without rupture or platelet release upon infection ( Figure 4F, Figure 4-video 1). In contrast, MKs with large sizes showed much slower migration ( Figure 4video 1). Additionally, CXCR4 high MKs were decreased in the bone marrow 3 days after L. monocytogenes infection but with a similar proliferation and apoptosis rate compared to CXCR4 low MKs ( Figure  2). Furthermore, we found that L. monocytogenes infection decreased the expression of CXCL12, the ligand of CXCR4 (Sugiyama et al., 2006), in bone marrow but increased CXCL12 expression in the lung, liver, and spleen ( Figure 4-figure supplement 3), suggesting that the distinguished CXCL12 levels between tissues might drive the migration of CXCR4 high MKs between tissues. In line with this, CXCR4 high MKs were increased in the peripheral blood and organs, including the liver, spleen, and lung 3 days after L. monocytogenes infection without an alternation of cell cycle and apoptosis, whereas CXCR4 low MKs did not differ except for a slight increase in the liver ( Figure 4G; Figure  CXCR4 low or CXCR4 high MK transfused. Scale bars, 20μm. Data represent mean ± s.e.m. Repeated-measures one-way ANOVA followed by Dunnett's test for multiple comparisons in (F), ǂ ǂ ǂ p<0.001. Paired Student's t-test was performed to assess statistical significance (G), # p<0.05, n.s., not significant. Twotailed Student's t-test was performed to assess statistical significance except (F), * p<0.05, ** p<0.01, *** p<0.001, n.s., not significant.
The online version of this article includes the following figure supplement(s) for figure 3:    To further explore how MKs migrate between organs during bacterial infection in vivo, we employed Pf4 Cre ; Rosa26 fs-tdTomato , and Pf4 Cre ; cell membrane-localized tdTomato cell membranelocalized EGFP (Rosa26 fs-mTmG ) mice in which Tomato or cell membrane-localized EGFP (mGFP) were exclusively expressed in MK lineage (Tiedt et al., 2007). mGFP expressing MKs or Tomato expressing CXCR4 high MKs were increased in the liver and spleen 3 days after L. monocytogenes infection ( Figure 4I; Figure 4-figure supplement 4E-H), whereas Tomato expressing CXCR4 low MKs did not change (Figure 4-figure supplement 4I). To further confirm the tissue infiltration of MKs upon infection, we intravenously injected membrane-localized tdTomato (mTomato) expressing bone marrow cells from Rosa26R fs-mTmG mice into control recipients or recipients infected with L. monocytogenes 1 day before mTomato + cell perfusion ( Figure 4J). We found that 2 days after mTomato + cell perfusion, engrafted mTomato + CXCR4 high MKs more efficiently infiltrated into the liver (92.1%) and spleen (92.5%); by contrast, most mTomato + CXCR4 low MKs (66.7%) migrated to the bone marrow ( Figure 4K).
As the lung is an important site for platelet generation (Lefrançais et al., 2017), we aligned our MK sc-RNAseq data with lung MKs (Pariser et al., 2021;Yeung et al., 2020), and found that MK5, MK4, and MK3 showed similar gene profiles with lung MKs ( Figure 4L). Moreover, MK5 enriched more inflammatory pathway genes, antigen processing, and presentation pathway after L. monocytogenes infection, which enabled MK5 to achieve a more similar transcriptional profile as the lung MKs than normal MK5 (Figure 4-figure supplement 5). Interestingly, we found that engrafted Tomato + MKs (from Pf4 Cre ; Rosa26 fs-tdTomato mice) more efficiently infiltrated the lungs in the infected recipients as extravascular MKs than in the control recipients ( Figure 4M-O and Figure 4-figure supplement 6).
control and 103 infected MKs) in the bone marrow of control mice or mice at 3 days after L. monocytogenes infection. (C) Comparison of the distance between CXCR4 low or CXCR4 high MKs and endothelial cells (Ecs) in the bone marrow of mice 3 days after L. monocytogenes infection (n=68 CXCR4 low and 78 CXCR4 high MKs). CD41 (blue), CXCR4 (red), and EMCN (green). Yellow arrows indicate CXCR4 high MKs, while white arrowheads indicate CXCR4 low MKs. (D) Representative immunofluorescent staining images (left) and quantification (right) of CXCR4 (red) labeled MKs (blue) egressed into sinusoids (green) upon infection (n=46 CXCR4 low MKs and 69 CXCR4 high MKs in 4 biological replicates). Yellow arrows indicate CXCR4 high MKs. (E) Comparison of the distance between HSCs to MKs (n=96 control and 127 infected HSCs, p=0.21 by two-sample KS test) in the bone marrow of control mice or mice at 3 days after L. monocytogenes infection. (F) Visualization of MK migration (red, arrowhead) into sinusoids (green) by live imaging in the bone marrow of Pf4 Cre+ ; Rosa26 fs-tdTomato+/mice 24 hr after L. monocytogenes infection (Movie S1). 'S' indicates sinusoid and dashed lines demarcate the border of sinusoids. (G) Quantification of CXCR4 high MKs in bone marrow, peripheral blood, liver, spleen, and lung of control mice and mice 3 days after L. monocytogenes infection. (H) Quantification of CXCR4 levels on CXCR4 low MKs treated with IFN-γ, LPS or L. monocytogenes for 4 hr compared to CXCR4 high MKs. (I) Quantification of tdTomato + CXCR4 high MKs in the liver and spleen from control mice or mice 3 days after L. monocytogenes infection by flow cytometry. (J) Schema of mtdTomato + bone marrow (from R26R mTmG mice) cell perfusion in control and L. monocytogenes infected recipients. (K) The percentage of CXCR4 high mtdTomato + MKs and CXCR4 low mtdTomato + MKs in bone marrow, liver, and spleen of control or infected recipients were analyzed 2 days after mtdTomato + bone marrow cells were perfused. (L) Radar chart showing transcriptomic similarities of bone marrow MK subpopulations with reported BM and lung MK datasets (Pariser et al., 2021;Yeung et al., 2020). (M) Schema for transfer experiments using tdTomato + MKs from Pf4 Cre+ ; Rosa26 fs-tdTomato+/mice into control recipients or recipients 1 day following L. monocytogenes infection. (N-O) Representative images (N) and quantification by flow cytometry (O) of tdTomato + MKs in the lung of control or infected recipients 2 days after cell perfusion (n=3 mice). Arrows indicate tdTomato + MKs in the lung. Scale bars without indicated, 20 μm. Data represent mean ± s.e.m. A two-sample KS test was performed to assess statistical significance in (E). Repeated-measures one-way ANOVA followed by Dunnett's test for multiple comparisons in (H), ǂ ǂ p<0.01, ǂ ǂ ǂ p<0.001. Two-tailed Student's t-test was performed to assess statistical significance except (E, H), * p<0.05, ** p<0.01, *** p<0.001, n.s., not significant.
The online version of this article includes the following video, source data, and figure supplement(s) for figure 4: Source data 1. Distance of HSCs to MKs in the bone marrow from control mice or mice 3 days after L. monocytogenes infection.       Acute inflammation-induced emergency megakaryopoiesis generates CXCR4 high MKs upon infection Infection-induced emergency megakaryopoiesis compensates the platelet consumption (Verschoor et al., 2011). Consistently, we observed that MKs were ruptured in the bone marrow 3 days after L. monocytogenes infection to recover the reduced platelets post-L. monocytogenes infection (Couldwell and Machlus, 2019;Nishimura et al., 2015;Figure 5A-C). However, CXCR4 high MKs were increased at 18 hr after L. monocytogenes infection and substantially declined at 72 hr in bone marrow, whereas CXCR4 low MKs remained unchanged upon infection ( Figure 5D). As MK-committed HSCs drive infection-induced emergency megakaryopoiesis (Haas et al., 2015), we asked whether emergency megakaryopoiesis also generates CXCR4 high MKs to participate in the host-defense response. To this aim, we employed Scl CreER ; Rosa26 fs-tdTomato mice (Göthert et al., 2005) to monitor the HSPC derived emergency megakaryopoiesis upon bacterial infection. Eighteen hours after tamoxifen recombining Tomato in HSPCs and L. monocytogenes infection ( Figure 5E), we observed that Tomato + HSPCs derived Tomato + CXCR4 high MKs rapidly increased in the bone marrow, similar to the platelet-generating MKs (tdTomato + CXCR4 low MKs) (Figure 5F), without a noticeable rise of hematopoietic progenitors ( Figure 5G). Overall, our observations indicated that CXCR4 high MKs might be generated by emergency megakaryopoiesis to stimulate pathogen defense.

Discussion
MKs participate in megakaryocyte maturation, platelet activation, and potentially influence neutrophils and the adaptive immune cells (Cunin and Nigrovic, 2019). Accordingly, MKs prevent the spread of dengue virus infection by enhancing the type 1 interferons pathway in murine and clinical biospecimens (Campbell et al., 2019) and contribute to cytokine storms in severe COVID-19 patients (Bernardes et al., 2020;Ren et al., 2021;Stephenson et al., 2021). MKs were reported to express multiple inflammation receptors, such as Fcγ receptors (Markovic et al., 1995), Toll-like receptors (Beaulieu et al., 2011;Ward et al., 2005), interleukin receptors (Navarro et al., 1991;Yang et al., 2000), and IFN receptors (Negrotto et al., 2011), which might enable MKs to receive inflammation signals and express cytokines. Recent scRNA-seq studies suggested the existence of MK subpopulations for inflammation responses Pariser et al., 2021;Sun et al., 2021;Wang et al., 2021a). Here, we identified that MK5 has both MK and immune cell characteristics for platelet generation and immune responses. More importantly, we demonstrated that CXCR4 high MKs recruited and stimulated innate myeloid cells by producing TNFα and IL-6, for bacterial phagocytosis. Furthermore, CXCR4 high MKs had the ability for antigen processing and antigen presentation capacity, which suggested that CXCR4 high MKs might contribute to the regulation of adaptive immune function. This is consistent with a previous observation that lung MKs are able to process and present antigens (Pariser et al., 2021). Our data suggested that CXCR4 high MKs might contribute to the regulation of adaptive immune function. However, as the distinction between CXCR4 high MKs and CXCR4 low MKs is not entirely objective, additional markers are warranted to further enrich CXCR4 high MKs.
We observed that MK ablation increased HSPCS and myeloid granulocyte/macrophage progenitor (GMP) in bone marrow under bacterial infection, which is consistent with 5-FU stress (Hérault et al., 2017). However, increased GMP only increased myeloid cells in the bone marrow but not in other organs, which further supported the role of CXCR4 high MKs in promoting the migration of myeloid cells. Normal HSC to MK development takes 11-12 days in humans and 4 days in mice; However, emergency megakaryopoiesis takes less than a day to generate MKs upon inflammation stress (Couldwell and Machlus, 2019;Liu et al., 2021;Sun et al., 2021; Figure 5D). Previously, researchers believed that emergency megakaryopoiesis mainly contributes to the replenishment of damaged platelets upon acute inflammation (Haas et al., 2015). We found that inflammation signals could not upregulate CXCR4 in CXCR4 low MKs in vitro, although we cannot entirely exclude the plasticity of MKs in vivo. Our data showed that CXCR4 high MKs might be generated from the emergency megakaryopoiesis, instead of CXCR4 low MKs, to facilitate host-defense responses against bacterial infection.
A recent report showed that the lung is a reservoir of MKs for platelet production (Lefrançais et al., 2017). Other works also indicate that lung MKs share a similar transcriptional profile with lung DCs and participate in pathogen infection (Boilard and Machlus, 2021;Pariser et al., 2021). However, the correspondence between MKs in the lung and bone marrow remains unexplored. Neonatal lung MKs lack the immune molecules in adult lung MKs (Pariser et al., 2021), which indicates that lung MKs might have distinct developmental origins. Similarly, MKs are observed to egress and migrate to the pulmonary capillary under stresses (Davis et al., 1997). Our works suggested lung MKs might migrate from bone marrow upon infection challenges, although more detailed investigations are warranted in future studies.

Continued on next page
Key resources

Cell line
B3Z hybridoma T cells were kindly gifted by Dr. Nilabh Shastri (Johns Hopkins University). This cell line was verified to be mycoplasma free by EZdetect PCR Kit for Mycoplasma Detection (HiMedia).

Bacteria and infections
Listeria (L.) monocytogenes infection was performed as described with minor modifications (Edelson and Unanue, 2000;Verschoor et al., 2011). In brief, wild-type L. monocytogenes strain 10,403 S grown to exponential phase at 37 °C in TSB media was injected intravenously at a dose of 2500 colony-forming units (CFUs) to determine spleen and liver bacterial burdens 3 days after infection. Recombinant L. monocytogenes expressing the chicken ovalbumin peptide (OVA 257-264 ) (L.m. -OVA 257-264 ) was injected intravenously at a dose of 2500 CFUs to determine activated spleen T cells 7 days after infection. Escherichia (E.) coli wild-type strain 85,344 expressing GFP was constructed as previously described (Feng et al., 2020). GFP-labeled E. coli was grown to exponential phase at 37 °C in LB media and washed with PBS before being suspended for phagocytosis assays.

Flow cytometry and cell sorting
Bone marrow cells were isolated from mouse femora and tibiae as previously reported (Jiang et al., 2018). Splenocytes were mechanically dissociated in PBS with 2% FBS. Peripheral blood was collected from the retro-orbital sinus and anticoagulated by K2-EDTA. Those three kinds of cells then underwent red blood cell lysis for 5 min using 0.16 M ammonium chloride solution. Liver cells were mechanically dissociated and lysed using 0.16 M ammonium chloride solution, followed by gradient sedimentation using a density reagent (LTS1077, TBD Science) following the manufacturer's instruction. Cell sorting was performed using a cell sorter (MoFlo Astrios, Beckman Coulter) with a 100 μm nozzle at a speed of around 5000 cells s -1 . For intracellular cytokine staining, cells were pretreated with Brefeldin-A (BFA, 10 μg ml -1 ) for 4 hr at 37℃ before staining. For MK antigen presentation detection, MKs were co-culture with 100 μg ml -1 soluble full-length OVA for 24 hr before staining. For IFNγ, LPS and L. monocytogenes treatment, cells were co-culture with 10 ng ml -1 IFN-γ or 30 μg ml -1 LPS for 4, 18, or 24 hr, or 10 6 L. monocytogenes for 4 hr in a 37℃ incubator before staining. Cell analysis was performed on either one of the flow cytometers (Attune NxT, Thermo Fisher; Cytek AURORA, Aurora).

Single-cell library construction and sequencing
Sorted CD41 + FSC high single cells from four mice of a control MK group and an MK group from mice 3 days upon L. monocytogenes infection each were processed through the Chromium Single Cell Platform using the Chromium Single Cell 3' Library and Gel Bead Kit v3 (PN-1000075, 10 x Genomics) and the Chromium Single Cell B Chip Kit (PN-1000074, 10 x Genomics) as the manufacturer's protocol. In brief, over 7000 cells were loaded onto the Chromium instrument to generate single-cell barcoded droplets. Cells were lysed and barcoded reverse transcription of RNA occurred. The library was prepared by following amplification, fragmentation, adaptor, and index attachment then sequenced on an Illumina NovaSeq platform.

scRNA-seq processing
The scRNA-seq reads were aligned to the mm10 reference genomes, and unique molecular identifier (UMI) counts were obtained by Cell Ranger 3.0.2. Normalization, dimensionality reduction, and clustering were performed with the Seurat 3.0 R package (Butler et al., 2018). For the control and Listeria (L.) monocytogenes infection group, we loaded one 10 x Genomics well each and detected 5663 and 5948 cells that passed the Cell Ranger pipeline, respectively. To ruled out low quality cells, cells with >12% mitochondrial content or <200 detected genes were excluded with Seurat function subset ( percent. mt <12 & nFeature_RNA >200). We ruled out doublets with default parameters of Doublet-Decon R package, and 54 control cells and 939 L. monocytogenes infected cells were excluded. Following the standard procedure in Seurat's pipeline, we identified 3272 MKs from control mice (1712 MKs) and mice with L. monocytogenes infection (1560 MKs) (3897 and 3449 immune cells were discarded, respectively) in combination with MetaNeighbor method. Preprocessed dataset normalization was performed by dividing the UMI counts per gene by the total UMI counts in the corresponding cell and log-transforming before scaling and centering. SCT normalization was performed with the script: object <-SCTransform(object, vars. to. regress = "" percent. mt"", verbose = FALSE). Signature genes of each cluster were obtained using the Seurat function FindMarkers with Wilcox test with fold change >1.5 and p value <0.05 after clustering. Heatmaps, individual UMAP plots, and violin plots were generated by the Seurat functions in conjunction with ggplot2 and pheatmap R packages. Similarities and UMAP projection between our scRNA-seq data and published datasets GSE152574 (Yeung et al., 2020), GSE158358 (Pariser et al., 2021), GSE137540 (Xie et al., 2020), GSE128074 (Hamey et al., 2021), or GSE132042 (Almanzar et al., 2020) were conducted by MetaNeighbor R package (Crow et al., 2018), iMAP. py and Symphony R package (Kang et al., 2021). iMAP integration was performed using the default parameters except n_top_genes = 2000, min_genes = 0, min_cells = 0, and n_epochs = 100 before doing dimensionality reduction using Uniform Manifold Approximation and Projection method (UMAP, n_neighbors = 30, n_pca = 30). Radar charts were generated with JavaScript written by Nadieh Bremer (https://www.visualcinnamon.com/). Euclidean distances denote the distances between the centroid of each cluster.
Correlations were calculated based on normalized RNA values, with the function cor and the parameter 'method = "spearman"'. Multiple testing correction using the function cor. test with the parameter "method = "spearman" and it was applied for Cxcr4 expression correlations. We calculated the similarities between MK1 to 5 with the published MK, immune cell, and myeloid progenitor datasets (Almanzar et al., 2020;Hamey et al., 2021;Pariser et al., 2021;Xie et al., 2020;Yeung et al., 2020) using scmap R package (Kiselev et al., 2018). Default parameters and 1000 features were used and threshold >0 was set. Cell-type matches are selected based on the highest value of similarities and the second-highest value which is not 0.01 less than the highest value across all reference cell types.
Cytokine, inflammatory, chemokine, and antigen processing and presentation scores were evaluated with the AddModuleScore function of Seurat using genes from KEGG pathway ko04060, cytokine-cytokine receptor interaction; GO:0006954, inflammatory response; chemokine ligands from CellPhoneDB. mouse  and GO:0019882, antigen processing and presentation.
Interaction analysis of MKs and immune cells were conducted by CellPhoneDB ) (transformed to human orthologous genes Davidson et al., 2020) and CellChat R package . Only interactions involving cytokines were shown. Gene Ontology (GO) analysis was performed using clusterProfiler R package (Yu et al., 2012) and visualized using enrichplot R package (Yu, 2019).
Gene set enrichment analysis (GSEA) was performed using gsea R package (Subramanian et al., 2005) and visualized using enrichplot R package. Gene lists were pre-ranked by the fold change values of the differential expression analysis using Seurat function FindMarkers. Gene sets for GSEA were obtained from GO database (GO:0002367, cytokine production involved in immune response; GO:0006954, inflammatory response; GO:0008009, chemokine activity; GO:0022409, positive regulation of cell-cell adhesion; GO:0002275, myeloid cell activation involved in immune response).
Gene set variation analysis (GSVA) was performed using GSVA R package (Hänzelmann et al., 2013). GSVA was performed to calculate GSVA score of indicated pathway genes in single cell datasets with the whole protein encoding genes after log normalization of expression values. Gene sets for GSVA were obtained from GO database (GO:0022409, positive regulation of cell-cell adhesion; GO:0002275, myeloid cell activation involved in immune response; GO:0002367, cytokine production involved in immune response; GO:0007596, blood coagulation; GO:0019882, antigen processing and presentation; GO:0034340: response to type I interferon; GO:0034341: response to interferongamma; GO:0045088, regulation of innate immune response; GO:0042742, defense response to bacterium; GO:0002819, regulation of adaptive immune response; GO:1903708, positive regulation of hemopoiesis).

Lung cells preparation for flow cytometry
Lungs were removed and digested as described with minor modifications (Lefrançais et al., 2017). In brief, removed lungs were placed in 1.5 ml tubes, minced with scissors, and digested with 1 ml digestion buffer (HBSS with 1 mg ml -1 collagenase D, 0.1 mg ml -1 DNase I, 25 mM HEPES, 2 mM L-glutamine, and 2% FBS) for 30 min at 37℃ before filtration through a 100 μm cell strainer and red blood cell lysis for 5 min. Samples were then filtered through 70 μm strainers and resuspended for subsequent surface marker staining for flow cytometry.

Megakaryocyte ablation induction
Pf4 Cre mice were mated with the Rosa26 fs-iDTR line to generate Pf4 Cre ; Rosa26 fs-iDTR mice. Diphtheria toxin (DT, Sigma-Aldrich) was injected intraperitoneally every day at a dose of 40 ng g -1 bodyweight into Pf4 Cre+ ; Rosa26 fs-iDTR+/mice and their cre negative counterparts to induce megakaryocyte ablation as indicated.

BrdU incorporation assay
5-Bromo-2-deoxyuridine (BrdU) was administered at a single dose of 125 mg kg -1 body mass by intraperitoneal injection. Whole bone marrow cells were collected 12 hr later and incubated with anti-CD41 and anti-CXCR4 for 1 hr. Cells were washed and then fixed with 4% PFA at 4 °C overnight. Cells were then permeabilized with 0.5% TritonX-100 for 15 min at room temperature and incubated with 1 mg ml -1 DNase I (Roche) for 1 hr at 37 °C followed by incubating with anti-BrdU (BU20A, eBioscience) for 1 hr at room temperature before being analyzed.

Annexin V binding assay
For Annexin V binding assay, bone marrow cells were incubated with cell surface markers for 1 hr at 4 °C and then washed with PBS before being resuspended with Annexin V binding buffer (Biolegend). Cells were then incubated with FITC Annexin V (Biolegend) for 15 min at room temperature in dark, and then 300 μl Annexin V binding buffer was added to each tube. Cells were analyzed by a flow cytometer.

Transwell migration
Transmigration assays were performed on a transwell with a pore size of 5 μm (Biofil). CXCR4 low MKs or CXCR4 high MKs from bone marrow were sorted (5000 cells per well) from control mice and added to the lower chamber with 600 μl IMDM (Thermo Fisher) plus 10% FBS (Gibco). Peripheral blood cells were collected as described in the 'Flow cytometry and cell sorting' section. 6×10 5 peripheral blood cells were resuspended in 100 μl RPMI 1640 (Gibco) plus 10% FBS and added to the upper insert to continue for 2-hr incubation at 37 °C, 5% CO 2 . Cells in the lower chamber were harvested, washed with PBS once, and resuspended with 100 μl PBS for staining and FACS counting.

Phagocytosis
Bone-marrow-derived macrophages (BMDM) from C57BL/6 mice at 6-8 weeks of age were differentiated from bone marrow precursors with minor modifications (Minutti et al., 2019). In brief, bone marrow cells were isolated and propagated for 7 days in DMEM without sodium pyruvate or HEPES (Gibco), containing 20% FBS (Gibco), 30% supernatants of L929 conditioned media, and 1% Pen/ Strep (Hyclone) at 37 °C. Macrophage phagocytosis assays were performed on a transwell plate with a pore size of 3 μm (Biofil) as described with modifications (Sharif et al., 2014). Briefly, attached cells were replated into 24-well plates, 5×10 4 cells per well, on glass coverslips for 24 hr culture. Then 5000 sorted CXCR4 low MKs or CXCR4 high MKs were added in the upper inserts and placed onto macrophages chambers for additional 16 hr incubation without or with 2 μg ml -1 TNFα neutralizing antibody (R023, Sino Biological; 1:100) or 2 μg ml -1 IL-6 neutralizing antibody (MP5-20F3, Biolegend) at 37 °C, 5% CO 2 . The upper inserts were discarded and macrophages were washed with PBS without antibiotics and incubated with 10 5 GFP-labeled E. coli for 2 hr at 37 °C, 5% CO 2 . Cells were washed three times with PBS and incubated with DMEM without sodium pyruvate or HEPES (Gibco) with gentamycin (50 μg ml -1 ) for 30 min at 37 °C, 5% CO 2 to remove adherent bacteria. Cells were then detected by flow cytometry or fixed by cold methanol for 15 min and blocked with 10% BSA overnight, followed by incubation with F4/80 (BM8, eBioscience; 1:100) for 2 hr at room temperature before being quantified using a spinning disk confocal microscope (Dragonfly, Andor).
For neutrophil phagocytosis, CD11b + Gr1 + Ly6cneutrophils were sorted from the spleen and propagated in RPMI 1640 (Gibco) containing 10% FBS. Neutrophil phagocytosis was performed as described in macrophage phagocytosis assay, except cells were sedimented for 30 min and fixed on glass coverslips after incubated with GFP-E. coli and gentamycin. The capacity of phagocytosis was evaluated by flow cytometry or by fluorescence intensity of GFP-E. coli. using the confocal microscope within macrophages and neutrophils.
For megakaryocyte phagocytosis, CXCR4 low and CXCR4 high MKs were sorted from the bone marrow and propagated in RPMI 1640 (Gibco) containing 10% FBS without antibiotics and incubated with 10 5 GFP-labeled E. coli for 2 hr at 37 °C, 5% CO 2 . Cells were washed three times with PBS and incubated with DMEM without sodium pyruvate or HEPES (Gibco) with gentamycin (50 μg ml -1 ) for 30 min at 37 °C, 5% CO 2 to remove adherent bacteria. Cells were then detected by flow cytometry.

Bone marrow ex vivo live imaging
Pf4 Cre+ ; Rosa26 fs-tdTomato+/mice were infected with L. monocytogenes for 24 hr. FITC-Dextran (average mol wt 2000000, Sigma-Aldrich) was injected intravenously at a dose of 1.25 mg per mouse before being sacrificed. The ends of the femur below the end of the marrow cavity were cut. The bone marrow plug was gently flushed out of the end of the bone with a 21-gauge blunt needle not to break up the marrow plug. Bone marrow was flushed integrally and fixed onto a glass slide in a chamber, rinsed with RPMI 1640 (Gibco), and covered slightly with a coverslip. The integrity of the vascular structure in the bone marrow was observed and warranted through FITC-Dextran inflorescence before capturing images. Confocal images were obtained every minute on the spinning-disk confocal microscope (Dragonfly, Andor) and analyzed using Imaris 9.0 software (Oxford Instruments).
In vitro MK culture, MK size, and proplatelet formation measurement MKs were sorted using a cell sorter (MoFlo Astrios, Beckman Coulter) and cultured in 24-well plates in SFEM (Stem Cell Technologies) plus 100 ng ml -1 mTPO (Novoprotein) and 1% Pen/Strep (Hyclone), and incubated at 37 °C, 5% CO 2 for 4 days. Images were taken by a Nikon Ts2R microscope equipped with a Nikon DS-Ri2 camera. Cell size and proplatelet formation were measured on day 3 or day 5 post-cultured, respectively, using Nikon NIS-Elements BR.
For in vivo CXCR4 high MK function assay in MK ablation mice, DT was intraperitoneally injected every day for 5 days. On the second and fourth days, 2×10 5 sorted wild-type CXCR4 high MKs or CXCR4 low MKs were intravenously injected into indicated groups. PBS or 2500 CFUs of L. monocytogenes as previously described were injected intravenously on the third day. Spleen and liver were harvested 3 days after infection to determine the bacterial burdens as described.

T cell reactivation in vitro
Splenocytes (1×10 6 cells well -1 ) from control or MK ablated mice after 7 days L.m.-OVA infection were re-stimulated for 4 hr in vitro with OVA peptide (10 μM) in the presence of Brefeldin-A (BFA, 10 μg ml -1 ). Activated T cells were then analyzed by a flow cytometer.
For MK-induced T cell activation, 3×10 4 MK subpopulations for each sample were sorted and co-cultured with 100 μg ml -1 soluble full-length OVA for 24 hr, then co-cultured with 6×10 4 OT-I CD8 + T cells or B3Z T cells (Karttunen et al., 1992) for 48 hr at 37 °C in a 5% CO 2 incubator as described (Zufferey et al., 2017). OT-I T cell activation was detected by measuring intracellular IL-2 levels. B3Z T cell activation was detected using β-galactosidase Assay Kit (RG0036, Beyotime). Bone marrowderived dendritic cells (DCs) were adopted as positive controls for T cell activation assay. To obtain bone marrow-derived DCs, isolated bone marrow cells were cultured in RPMI 1640 with 10 ng ml -1 of GM-CSF and 10 ng ml -1 of IL-4 as described (Roney, 2019).

Computational modeling of random myeloid cell localization
We have performed randomized simulations as in previous reports (Bruns et al., 2014;Jiang et al., 2018) in Python. Images of a 400 μm × 400 μm bone marrow region with CXCR4 high and CXCR4 low MKs, in which background staining was removed, were used to generate MKs onto which 200 myeloid cells were randomly placed, consistent with an average density of 200 myeloid cells per field. Each simulated run placed 200 random myeloid cells (mean diameter 5 μm) was repeated 500 times. The shortest Euclidean distance was calculated for each myeloid cell to CXCR4 high or CXCR4 low MKs. Random and observed distance distributions were compared using the modified nonparametric twodimensional (2D) KS test as described (Bruns et al., 2014;Jiang et al., 2018).

Statistical analyses
Data are presented as means ± s.e.m or presented medians, first and third quartiles. For phagocytosis assay and MK size measurement, data were analyzed by a one-dimensional KS test. Differences were considered statistically significant if p<0.05. For the comparison of three-dimensional distances, a two-dimensional KS test was used. The difference was considered statistically significant if p<0.05. For multiple comparisons analysis, data were analyzed by repeated-measures one-way analysis of variance (ANOVA) followed by Dunnett's test. Differences were considered statistically significant if p<0.05. ǂ p<0.05, ǂ ǂ p<0.01, ǂ ǂ ǂ p<0.001, n.s., not significant. For pairs of measurements, data were analyzed by paired Student's t-test. Differences were considered statistically significant if p<0.05. # p<0.05, ## p<0.01, ### p<0.001, n.s., not significant. For other experiments except for scRNA-seq analysis, data were analyzed by a two-tailed Student's t-test. Differences were considered statistically significant if p<0.05. * p<0.05, ** p<0.01, *** p<0.001, n.s., not significant. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Additional files
The following dataset was generated: Author (