STING-Licensed Macrophages Prime Type I IFN Production by Plasmacytoid Dendritic Cells in the Bone Marrow during Severe Plasmodium yoelii Malaria

Malaria remains a global health burden causing significant morbidity, yet the mechanisms underlying disease outcomes and protection are poorly understood. Herein, we analyzed the peripheral blood of a unique cohort of Malawian children with severe malaria, and performed a comprehensive overview of blood leukocytes and inflammatory mediators present in these patients. We reveal robust immune cell activation, notably of CD14+ inflammatory monocytes, NK cells and plasmacytoid dendritic cells (pDCs) that is associated with very high inflammation. Using the Plasmodium yoelii 17X YM surrogate mouse model of lethal malaria, we report a comparable pattern of immune cell activation and inflammation and found that type I IFN represents a key checkpoint for disease outcomes. Compared to wild type mice, mice lacking the type I interferon (IFN) receptor exhibited a significant decrease in immune cell activation and inflammatory response, ultimately surviving the infection. We demonstrate that pDCs were the major producers of systemic type I IFN in the bone marrow and the blood of infected mice, via TLR7/MyD88-mediated recognition of Plasmodium parasites. This robust type I IFN production required priming of pDCs by CD169+ macrophages undergoing activation upon STING-mediated sensing of parasites in the bone marrow. pDCs and macrophages displayed prolonged interactions in this compartment in infected mice as visualized by intravital microscopy. Altogether our findings describe a novel mechanism of pDC activation in vivo and precise stepwise cell/cell interactions taking place during severe malaria that contribute to immune cell activation and inflammation, and subsequent disease outcomes.


Introduction
Malaria still remains a significant global health problem worldwide with 214 million people infected (range 149-303 million, WHO 2015), and half a million deaths each year [1,2]. Death occurs as a result of severe malaria, a multi-systemic disorder resulting from blood stage Plasmodium infection that is associated with significant morbidity and high case fatality rates despite antimalarial treatment [1,3]. It primarily affects young children in endemic areas and travelers, both lacking exposure-induced immunity. Severe malaria includes cerebral malaria (CM), severe anemia and metabolic acidosis and is characterized by extensive parasitized erythrocyte sequestration in CM and subsequent organ dysfunction with a heightened and dysregulated immune response that remains poorly understood [4]. Lack of a highly effective vaccines and emergence of artemisinin resistance limits malaria control measures. Thus additional efforts to understand severe disease with the goal of improving clinical outcomes remains a very high priority [5,6].
Severe malaria patients exhibit high levels of serum pro-inflammatory cytokines (TNF, IL-1β, IL-6, IL-8, IL-12, IFNγ) and chemokines (CCL2, CCL5, CXCL9, CXCL10), and lower levels of regulatory cytokines (IL-10, TGFβ, PGE2) [7] compared to those with mild and asymptomatic malaria [8][9][10][11][12][13]. Yet the mechanisms driving the high inflammatory states in severe malaria are not well understood. The type I interferon (IFN) cytokine is particularly interesting as it is released very rapidly upon sensing of microbial products and exhibits broad immunomodulatory properties [14,15]. Upregulation of the type I IFN pathway has been associated with mild malaria patients compared to severe disease patients [16] and polymorphisms in the type I IFN receptor with severe disease outcomes [17,18]. Age-associated resistance against severe malaria infection-both for P. falciparum and P. vivax-and asymptomatic infection, directly correlates with high levels of regulatory cytokines [7,10]. In this context, a role for Foxp3 + CD4 + regulatory T cells (Tregs) as well as non-Foxp3 + activated/memory (CD45RO + ) CD4 + T cells in regulating malaria-induced inflammation and subsequent immunopathology is supported by many studies in human patients and in surrogate mouse models [19][20][21][22][23][24]. NK and γδ T cell activation is increased in children suffering from CM compared to other forms of malaria, as is the proportion of activated/memory T cells [25]. Inflammatory CCR2 + CD14 + monocytes mediate antibody-dependent cellular inhibition, and their higher proportion was correlated to lower blood parasitemia in acute uncomplicated malaria patients [9]. While most cited studies have focused on one or another immune cell type, there is only relatively limited information that exists on a global analysis of the leukocyte response taking place in the blood of patients undergoing severe malaria.
Initiation of a systemic immune response that follows the sensing of blood Plasmodium parasites is a poorly understood process that involves multiple pathways triggered by parasitederived molecules and byproducts, and a subsequent rapid inflammatory response. Major Plasmodium parasite molecules sensed by the host immune system include glycophosphatidylinositol (GPI), haemozoin crystals and nucleic acids. GPI anchors trigger TLR2/TLR1 and TLR2/ TLR6 heterodimers -and to some extent TLR4-, leading to the production of pro-inflammatory cytokines (TNF, IL-1β) via MyD88 [26,27], reactive nitrogen and oxygen species [28,29], and the upregulation of adhesion molecules (ICAM1, VCAM1) [30]. Haemozoin crystals result from haemoglobin digestion by blood stage Plasmodium parasites which stimulate massive secretion of IL-1β, TNF and chemokines. Haemozoin alone [31,32] or complexed to parasite DNA [33] triggers TLR9, ultimately leading to phago-lysosomal destabilization and cytosolic sensing of parasite DNA via AIM2 and haemozoin via NLRP3 inflammasomes inside phagocytes [34]. The STING pathway has also been implicated in the recognition of AT-rich stem loop motifs of Plasmodium DNA inside the cytosol of macrophages in vitro, driving TBK1/ IRF3-mediated secretion of type I IFN [35]. These studies were primarily conducted on macrophages cell lines in vitro, on ex vivo isolated or in vitro grown myeloid cells, and/or made use of parasite-derived or synthetic ligands. Other reports analyzed the role of these pathways using various surrogate mouse models of malaria and mice lacking the tested pathways [35][36][37]. Most commonly used models in these studies included non-lethal acute P. yoelii (Py) (17XNL) or chronic P. chabaudi (Pc), as well as lethal Py (17X YM) and P berghei (ANKA) respectively as models of high parasitemia and of CM. Along with the human reports, results from these investigations suggested that these sensing receptors and pathways promote high inflammatory response to the parasite and often poor outcomes. For instance, mice lacking MyD88, the common adaptor protein for many Plasmodium-stimulated TLRs, exhibited diminished blood inflammatory cytokines (IL-6, IL-12, IFNα, IFNγ, TNF, CCL2) in several of the murine models, leading to improved resistance to these infections [38,39]. Interestingly, single TLR deficiencies, such as in TLR2, TLR4, TLR9 or TLR7, only led to mild differences compared to WT mice, suggesting that multiple sensing mechanisms converge in mediating host resistance to the parasite. Similarly to MyD88 -/mice, mice lacking a functional cytosolic adaptor protein STING, which controls multiple cytosolic sensing receptors driving the induction of type I IFN, exhibited increased resistance to P. berghei infection [35]. Consistent with the potential importance of type I IFN, lack of type I IFN production in mice deficient for the type I IFN transcriptional regulators IRF3 and IRF7 [35], or lack of type I IFN signaling in mice disrupted for the IFNα/β receptor, enabled the survival of P. berghei infected mice [17,37]. Production of type I IFN was also suggested to involve TLR9/ and TLR7/MyD88, and IRF7 in splenic macrophages and pDCs, promoting enhanced early immune cell activation and pro-inflammatory response (IL-12, IFNγ, TNF) in P. chabaudi infection [36,40,41]. Yet none of the mice lacking these sensors or type I IFN signaling exhibited substantial clinical phenotypes in non-lethal models, consistent with the idea that clinical outcomes are regulated by the nature of Plasmodium infection, e.g., mouse genetic background and strain of Plasmodium.
Classical DCs and pDCs [42], macrophages [36,43] and CCR2 + monocytes [44] are important contributors to Plasmodium-induced inflammation, through secretion of type I IFN and other inflammatory mediators (IL-12, TNF, IL-6) and chemokines, production of antimicrobial effector proteins (reactive oxygen and nitrogen species) and activation of robust T cell responses [45]. For instance, early type I IFN signaling in DCs activated following P. berghei infection, suppressed their ability to prime strong protective "Th1" CD4 + T cell responses [37,46], preventing the control of infection and host survival.
While above referenced studies have clearly highlighted key innate sensing pathways and cytokines, and the implication of specific subsets of cells in detecting Plasmodium parasites in vivo, the precise initiating steps which include the tissue, the cell subsets and the cell-intrinsic sensing mechanisms and their relative contribution remains to be further defined. The spleen for instance, represents an essential organ in the course of blood stage malaria infections, both in humans and mice, largely because of its central function in inducing immune responses, monitoring blood through the removal of damaged RBCs and iron recycling, and its ability to produce emergency extramedullary haematopoeisis [47]. Splenectomized patients, similarly to that of non-lethal murine malaria, exhibit higher parasitemia, yet in lethal P. yoelii host resistance is improved, suggesting a deleterious contribution of spleen-derived inflammation. Whether other organs and cells play a key role during blood stage malaria has not yet been thoroughly addressed. The systemic nature of this infection would support the hypothesis that many organs and myeloid cells residing in these tissues are likely to contribute to shape infection outcomes. Recent reports analyzing changes in the bone marrow hematopoiesis during P. chabaudi infection illustrated the existence of major modifications occurring in this compartment, with a temporary loss of myeloid-erythroid precursors and the common lymphoid progenitor, and the emergence of a new, infection-induced population of atypical progenitor cells (IL-7Rα + c-Kit hi ) with lymphoid and myeloid potential [48]. Whether the bone marrow plays additional roles in initiating the immune response against the parasite requires further investigations.
In summary, multiple cellular sources for key inflammatory factors such as type I IFN have been proposed, and these notably include macrophages, DCs and pDCs [36,41,49]. However, little is known as far as (i) which immune cells first sense infected iRBCs and/or parasitederived immunogens and (ii) which cell-intrinsic molecular pathways are involved. The precise cell/cell interactions taking place during active blood stage malaria infection in vivo, the relative contributions of the various cell subsets to these key processes, and the sequences of events that occur and in which tissues have not been determined. Such information is acutely lacking and is absolutely essential to gain understanding of this important global infectious disease, and to possibly design novel therapeutic strategies.
In the current study we have made use of lethal P. yoelii 17X YM as surrogate model of severe malaria to investigate these relevant mechanistic questions. Two major murine models of severe malaria, P. yoelii and P. berghei, recapitulating some of the key immunological features reported in patients experiencing severe malaria, are available, yet substantially less information is available on Py-infected mice. We first conducted an extensive phenotypic and functional analysis of immune cells and inflammatory factors on all peripheral blood leukocytes isolated from a unique cohort of Malawian children undergoing severe malaria. A parallel analysis conducted on the blood of Py-infected mice, largely reflected the high inflammatory state and immune cell activation observed in the human patients. We show this broad activation of leukocytes, in particular Ly6C hi monocytes, NK cells and pDCs is accounted for by an early systemic burst of type I IFN. Type I IFN is acting as a key initiating cytokine orchestrating the robust inflammatory syndrome and the lethal outcomes. We define (i) pDC as the major producers of systemic type I IFN in the bone marrow and the blood, (ii) pDC activation requirements by macrophages in the bone marrow and their physical interactions, and (iii) the Plasmodium parasite sensing mechanisms that occur via TLR7/MyD88 in pDCs and STING in macrophages.

Results
Parallel systemic inflammation and innate immune cell activation in the blood of patients and mice with severe malaria To gain insights into the immune response associated with severe human malaria, we analyzed inflammatory mediators and immune cells in the peripheral blood of 37 Malawian children with WHO defined cerebral malaria (CM), enrolled in the Blantyre Malaria Project during the 2013 transmission season (S1A Fig). The median age for the cohort was 43 months; patients exhibited detectable parasitemia (~70x10 3 parasites/μl), anemia (23% hematocrit) and primarily manifested retinopathy positive CM (73%). Plasma and peripheral blood mononuclear cells (PBMCs) were drawn for analysis upon admission and 30 day convalescent blood samples among survivors. Plasma levels of inflammatory cytokines (IFNα, IFNγ, TNF, IL-6) and chemokines (CCL2, CCL3, CCL4, CXCL10) were significantly higher during severe malaria compared to convalescence ( Fig 1A). Using as a surrogate model of severe malaria, wild type (WT) C57BL/6 (B6) mice were inoculated with 2x10 5 iRBCs of the lethal Plasmodium yoelii (Py) 17X YM strain. Mice succumbed within 5-9 days (d) post infection with very high blood parasitemia (~80% iRBC)(S1B Fig), and we report a comparable blood inflammatory mediator profile and increase to that of human patients (Fig 1B and S1A Table), consistent with the development of severe signs of malaria and subsequent death (d 4.5 and after, S1B Fig). This comparative analysis underlines a common set of elevated blood inflammatory cytokines (IFNα, IFNγ, TNF, IL-6) and chemokines (CCL2, CCL3, CXCL10) in patients and mice. Yet, the use of the mouse model further revealed an early burst (d 1.5) of blood inflammatory cytokines including type I IFN (IFNα, IFNβ) and chemokines (S1C and S1D Fig) that are likely involved in early myeloid effector cell recruitment to the blood and the initiation of systemic inflammation.
Twelve color flow-cytometry analysis on the PBMCs of these patients (S1E and S1F Fig) revealed a higher proportion of inflammatory (CD14 + ) monocytes [9,50] and lower frequencies of pDCs [51] but no significant differences for myeloid DCs or patrolling (CD14 dim ) monocytes during CM compared to follow up (S1G Fig), a trend that was also noted in the mouse model (S1B Table). Expression of the costimulatory molecule CD86, MHC class II (MHC-II) HLA-DR and the integrin CD11c were decreased on all blood myeloid cells except on pDCs, yet the type I IFN responsive marker BST2 was always upregulated (Fig 1C and 1D and S1H and S1I Fig). Cell surface levels of the CD40 costimulatory molecule and the ICAM-1 adhesin were also increased on the various monocyte subsets. NK and T cells exhibited activated phenotypes (CD69 hi , CD62 lo , CD45RO + ) and actively proliferated (Ki67 + ) (Fig 1E and  S1J Fig). In Py-infected mice too, inflammatory (Ly6C + ) monocytes, pDCs and NK cells underwent robust activation compared to that of uninfected controls (Fig 1F and S1K Fig). Specifically, Ly6C + monocytes upregulated cell surface expression of CD86, MHC-II, CD11c, F4/80, ICAM-1 and the IFN-responsive markers BST2, Sca-1 and IL-15Rα. Increased cell surface CD86 and BST2 was also measured on pDCs, and NK cells upregulated CD69, became cytolytic (granzyme B + , NKG2D + ), and secreted IFNγ. While expression of some of these markers (CD86, MHC-II, CD11c) on inflammatory monocytes (CD14 + /Ly6C + ) followed a different trend in the patients than in the mouse model at early time points (d 1.5), their expression was significantly down-modulated at later times (d 4.5, Fig 1F) when mice exhibited severe clinical symptoms, a physiological situation closely related to that of the patients at enrollment (S1C Table). Thus, in summary, several functional markers related to cell activation, adhesion, antigen presentation, costimulation, differentiation, proliferation, effector functions and type I IFN were strongly upregulated on the various immune cell subsets during severe malaria in human patients and in the surrogate mouse model. This is particularly noticeable for CD14 + /Ly6C + monocytes, NK cells and pDCs. The current results draw an interesting parallel between the high level and the nature of the inflammatory mediators, and the activated immune cell subsets found in the peripheral blood of patients with severe malaria and the Py YM surrogate mouse model.

Type I IFN signaling promotes lethal outcomes and immune activation in severe murine malaria
Since type I IFN is among the first cytokines detected (Fig 1B), and represents a key mediator of immune cell activation in human and murine malaria [16, 35-37, 41, 52], we hypothesized that induction of type I IFN may be an essential checkpoint controlling the severity of infection and associated inflammation and immune cell activation. We inoculated mice lacking the type I IFN receptor (Ifnar1 -/-) or control WT B6 mice with Py YM, and monitored survival and blood parasitemia (Fig 2A). While WT mice succumbed 5-8 d post infection, the majority of Ifnar1 -/mice survived (~80%) and controlled parasite growth, as was also reported for Ifnar1 -/mice inoculated with the P. berghei ANKA mouse model of severe malaria [35,37]. As hypothesized, the levels of inflammatory mediators (IL-6, CCL2, CCL3, CXCL1, CXCL10 and IFNα as expected) were lower in the blood and the spleen of Ifnar1 -/mice compared to WT counterparts at all times post-infection, except for IFNγ ( Fig 2B). Consistent with these findings, a significant reduction of early immune leukocyte recruitment (factor of~1. [2][3] to the blood of infected Ifnar1 -/mice was also noted, which was accounted for by direct type I IFN signaling to the various innate immune cells subsets as shown in Ifnar1 -/-/WT mixed chimeras ( Fig 2C and S2A Fig). The activation of inflammatory monocytes (CD86 + , CD11c + , Sca-1 + , IL-15Rα + , BST2 + ), NK cells (CD69 + , NKG2D + , Granzyme B + ) and pDCs (CD86 + , BST2 + ) was decreased by~20-85% in Ifnar1 -/mice ( Fig 2D, filled bars and S2B Fig). These results extended to Ifnar1 -/-/WT chimeras, demonstrating that direct type I IFN signaling to these cell subsets mediates their activation (Fig 2D, open bars and S2B Fig). Overall, type I IFN production and signaling during severe blood stage malaria represents a key checkpoint of inflammation and immune cell activation, particularly for inflammatory monocytes and NK cells, a finding that correlates to disease outcomes.
Plasmacytoid dendritic cells produce immune-activating type I IFN in the bone marrow and blood of Py-infected mice We next sought to define which cells produced type I IFN during Py YM infection and their tissue distribution. We utilized Ifnb-Yfp +/+ reporter mice expressing the yellow fluorescent protein YFP under the control of the IFNβ promoter [53], and monitored Ifnb gene expression (e.g., YFP + cells) by immune cells in the blood, bone marrow, spleen and liver of Py YM infected mice (Fig 3A and 3B). Staining of YFP + cells in infected mice with cell-surface lineage that all YFP + cells in the various tissues were BST2 hi SiglecH hi CD11b low B220 int CD11c int , defining them as plasmacytoid DCs (pDCs). YFP + pDCs were present predominantly in the bone marrow and blood of infected mice, with far fewer in the spleen. At least 20 times more YFP + pDCs were detected in the bone marrow (factor of~10 per femur) compared to the other compartments, which was consistent with higher levels of quantified IFNα and IFNβ ( Fig 3C and S3B Fig). Thus the bone marrow contains the vast majority of IFNα/β-producing pDCs and their kinetics of appearance suggests that production of type I IFN occurs in this compartment as early as 24 hours post-infection, peaking at 36-42 hours before waning. Of note, pDC frequencies increased in the blood while decreasing in the bone marrow, consistent with the idea that following activation in the bone marrow, pDCs emigrate to the blood ( Fig 3D). In line with this possibility, YFP + pDCs underwent stronger upregulation of cell-surface CD86, BST2 and ICAM-1 compared to their YFP neg counterparts ( Fig 3E).
To further confirm that pDCs are the predominant source of type I IFN in Py infection, we next conducted loss of function experiments in BDCA2 depleter mice (Bdca2-Dtr +/-), which express the simian diphtheria toxin receptor (DTR) under the human BDCA2 pDC-specific promoter [54]. Diphtheria toxin (DT)-treated Bdca2-Dtr +/mice were inoculated with Py YM, and blood and bone marrow levels of IFNα and IFNβ were measured 1.5 d later (Fig 3F and  S3C Fig). Production of both IFNα and IFNβ in these mice was abrogated compared to WT counterparts, further establishing pDCs as the major producers of systemic type I IFN during severe malaria. Selective elimination of the pDCs also impaired blood recruitment and activation of inflammatory monocytes and NK cells ( Fig 3G and S3D Fig), altogether supporting the idea that pDC-derived type I IFN control immune cell activation. Yet survival of pDC-depleted mice was only slightly enhanced (S3E Fig) suggesting pDC-derived type I IFN partially contributed to disease severity in the course of severe murine Py malaria.
In addition, both CCR5 and CXCR3 chemotactic receptors (but not CCR2), which are expressed on mature bone marrow pDCs, and can contribute to pDC migration from the bone marrow to the blood [55,56], underwent cell-surface downregulation in infected mice ( Fig 3H  and S3F Fig). Taken together, these results provide evidence that pDCs are the primary detectable cell type responsible for secretion of systemic type I IFN during severe malaria, with the bulk of production occurring in the bone marrow (also likely the site of initial activation, See kinetics Fig 3C) and in the blood.

MyD88 and STING control systemic type I IFN production during severe malaria
Both the MyD88 and the STING sensing pathways may contribute to type I IFN production in response to Plasmodium parasite-derived molecules [35,36,41]. To define whether and to which extent any of these sensing pathways control systemic type I IFN secretion during severe malaria in vivo, we first inoculated Myd88 -/-, Sting Gt/Gt or control WT mice with Py YM, and measured blood and bone marrow IFNα/β levels at its peak (d 1.5) of production ( Fig 4A and  S4A Fig). While as reported earlier, a significant increase in IFNα and IFNβ was detected in the blood of Py-infected WT mice, only levels close to that of uninfected mice, i.e., at the limit of detection of the assay, could be measured in Myd88 -/and Sting Gt/Gt mice. We next infected the Ifnb-Yfp +/+ reporter mice crossed to either Myd88 -/-, Sting Gt/Gt or control WT mice, and monitored expression of the Ifnb gene, e.g., YFP + cells, by immune cells -all YFP + cells were pDCsin the blood, bone marrow and spleen (Fig 4B and S4B Fig). In line with IFNα/β protein levels, YFP + pDCs were undetectable in Myd88 -/-Ifnb-Yfp +/+ mice and reduced by over two-third in Sting Gt/Gt Ifnb-Yfp +/+ mice, compared to WT controls. Consistent with the role of type I IFN in activating the immune system during Py infection (Fig 2), the recruitment and maturation of blood Ly6C + monocytes, NK cells and pDCs, was strongly impaired in Myd88 -/mice (Figs 4C and S4C). These differences remained modest in Sting Gt/Gt compared to WT mice, suggesting either that the low frequency of IFNβ-expressing pDCs (YFP + ) -and likely of secreted IFNαproduced in Py-infected Sting Gt/Gt mice was sufficient to induce immune cell activation or that other, type I IFN-independent STING-dependent signals, influence this process. Altogether, these results establish the importance of both MyD88 and STING sensing pathways for systemic pDC-derived type I IFN production during severe blood stage malaria.

Plasmacytoid dendritic cells produce type I IFN via TLR7/MyD88 during malaria in vivo
MyD88 and STING, and pDCs accounted for systemic type I IFN production in severe malaria (Fig 4). To determine whether either pathway is active in pDCs, we generated mixed bone marrow chimeras reconstituted with either Myd88 -/or Sting Gt/Gt and WT donor cells, all expressing the Ifnb-Yfp +/+ reporter transgene, and monitored the presence of YFP + pDCs following Py-infection (Fig 5). Lack of MyD88 but not STING in pDCs abrogated transcription of the Ifnb gene in Py-infected chimeras (Fig 5A). Since TLR7 and TLR9 are functionally active in pDCs upstream of MyD88 [57], we tested their implication in MyD88-dependent production of type I IFN. YFP + Tlr7 -/-pDCs were neither detected in Tlr7 -/-/WT Ifnβ-Yfp +/+ mixed chimeras nor in Tlr7 -/-Ifnb-Yfp +/+ mice, further establishing that lack of TLR7 in pDCs fully prevents Ifnb gene expression (Fig 5A and 5B). Likewise, production of IFNα in the blood and bone marrow of Py-infected Tlr7 -/mice was largely abrogated compared to WT mice ( Fig 5C). Unexpectedly [31, 33, 34], we did not detect any significant differences in IFNα levels in the blood or bone marrow of Py-infected Tlr9 -/mice, ruling out a role for TLR9 in this model. Of note, the frequency of YFP + pDCs and associated levels of type I IFN, were consistently lower in the chimera mouse models compared to non-irradiated counterparts (See Fig 3C versus  Fig 5A).
To further link pDC production of type I IFN via these pathways to immune cell activation, we generated additional mixed chimeras reconstituted with Bdca2-Dtr +/and either Tlr7 -/-, Tlr9 -/or Sting Gt/Gt bone marrow (Fig 5D and S5 Fig). In these mice, WT pDCs are eliminated following DT injection while pDCs lacking either of the above pathways remain. As predicted from type I IFN levels (Figs 4A and 5C), blood inflammatory monocyte and NK cell activation were impaired when pDCs lacked TLR7 but not TLR9 or STING (Fig 5D). Altogether, these data establish pDCs as key producers of type I IFN in the course of severe malaria, through the TLR7/MyD88 pathway. Moreover, pDC-secreted type I IFN directly promoted immune cell activation that correlates with disease severity. CD169 + macrophages prime pDCs to secrete type I IFN STING contributes to systemic type I IFN production during Py infection, but this is extrinsic to pDCs (Figs 4 and 5A). Prior work in human macrophage cell lines proposed that STING senses AT-rich stem-loop motifs from Plasmodium parasite DNA [35], therefore we hypothesized that STING triggering inside macrophages may provide low levels of type I IFN that could "prime" pDCs to further secrete high amounts of type I IFN. We used CD169 depleter mice in which tissue-resident CD169 + macrophages (F4/80 hi ) can be selectively eliminated after DT injection [58], and inoculated DT-treated Cd169-Dtr +/-Ifnb-Yfp +/+ or control WT Ifnb-Yfp +/+ reporter mice with Py YM (Fig 6A and 6B). Only very low frequencies of YFP + pDCs were detected in the blood, bone marrow and spleen of DT-treated CD169 depleter mice, and correspondingly low levels of IFNα, demonstrating that CD169 + macrophages were required for pDCs to produce type I IFN during this infection. Of note, total pDC frequencies remained either comparable or increased (bone marrow) in DT-treated versus untreated Cd169-Dtr +/mice, ruling out direct depletion of pDCs in this model (S6A Fig). In line with our prior findings, elimination of CD169 + macrophages also prevented blood immune cell activation, specifically that of Ly6C + monocytes, NK cells and pDCs (Fig 6C). We confirmed these results using clodronate liposome-treated Ifnb-Yfp +/+ mice, and DT-treated LysM-Cre +/-Rosa26-iDtr +/mice, as different means to deplete tissue-resident macrophages (S6B and S6C Fig).
CD169 + macrophages were reported to be largely radio-resistant [59]. Consistent with this fact, the production of IFNβ by pDCs (YFP + ) was abrogated in Py-infected DT-treated Cd169-Dtr +/but not WT mice reconstituted with Ifnb-Yfp +/+ WT bone marrow cells (Fig 6D). Thus the majority of IFNβ-producing DCs (YFP + ) (>75%) undergo "priming" by radio- resistant CD169 + macrophages. To further link STING-mediated activation and CD169 + macrophages, we next asked whether STING in the radio-resistant compartment was essential to promote type I IFN production by pDCs. WT or Sting Gt/Gt mice were lethally irradiated and reconstituted with either Sting Gt/Gt or WT Ifnb-Yfp +/+ reporter bone-marrow cells, and the presence of YFP + pDCs was monitored 1.5 d post Py YM infection ( Fig 6E). As hypothesized, STING inside the radio-resistant compartment was required for pDCs to secrete type I IFN. STING is a key inducer of type I IFN [60], and consistent with the idea that STING-dependent production of type I IFN is an essential orchestrator of pDC "priming", Ifnar1 -/-pDCs exhibited impaired activation compared to WT pDCs (Fig 2D). Thus collectively these results support a model in which STING-activated CD169 + macrophages prime pDCs -possibly through very low levels of secreted type I IFN-which synergizes with pDC-intrinsic TLR7/MyD88-mediated activation.
pDCs and CD169 + macrophages interact in the bone marrow of malariainfected mice Since CD169 + macrophages control pDC activation, we then asked whether pDCs interacted with CD169 + macrophages in the bone marrow. We used intravital time-lapse two-photon imaging to visualize pDCs and CD169 + macrophages located within the tibia of live, Pyinfected or control uninfected mice. Taking advantage of Ptcra-Gfp + reporter mice [61] in which GFP is expressed almost exclusively by bone marrow pDCs (>95%), we injected fluorescently-labeled (phycoerythrin, PE) anti-CD169 mAb allowing for tracking of both pDCs and CD169 + macrophages in vivo (Fig 7A-7D and S1 and S2 Movies). In uninfected mice, pDC exhibited a surveying and "hopping" behavior, with transient interactions with CD169 + macrophages. In stark contrast, the vast majority of pDCs in infected mice arrested (Fig 7A and 7B), maintained sustained interactions (Fig 7C) and elongated and clustered with CD169 + macrophages (Fig 7D). This latter behavior was also observed for YFP + pDCs in Py-infected Ifnb-Yfp +/+ mice (S3 Movie). Interestingly, the majority of CD169 + macrophages were lost between 24 to 36 hours post infection (Fig 7E), which correlates with the kinetics of type I IFN production by pDCs, and supports the idea that macrophages may die and release their content to activate pDCs. In line with this possibility, we also observed lower numbers of CD169 + macrophages in the bone marrow of Py-infected mice (1.5 d) by intravital imaging. The remaining CD169 + macrophages adopted a different morphology, were less "dendritic" in shape and we noted increased PE + cell debris, consistent with cell death (S1, S2 and S3 Movies). Thus, in the course of Py infection, pDCs arrested, clustered and contacted CD169 + bone marrow macrophages for prolonged periods of time, a result supporting the idea that CD169 + macrophages license pDCs to produce high levels of type I IFN, possibly via STING-induced type I IFN and/ or releasing intracellular materials (Fig 8).

Discussion
This study reveals novel cellular and molecular mechanisms that control the initiation of the early immune response during severe malaria infection. We characterized a highly activated and inflammatory state using a mouse model that recapitulates features of the immune response found in humans with severe malaria, and we identify broad and robust activation of multiple immune cell populations. We reveal that the antiviral cytokine type I IFN is a major orchestrator of these processes during blood stage Plasmodium infection, leading to enhanced inflammatory (Ly6C hi /CD14 + ) monocyte and NK cell activation in the blood of infected mice. We establish that pDCs produce most of the systemic immune activating type I IFN in both the bone marrow and the blood of infected mice, following TLR7/MyD88-sensing of Plasmodium parasites. pDCs only secreted type I IFN upon CD169 + macrophage "priming" in the bone marrow of infected mice. Macrophage licensing of pDCs requires STING sensing of parasites and both cell subsets establish stable long-lasting interactions, as visualized by intravital microscopy. Altogether, our data suggest a model in which STING-activated CD169 + macrophages may deliver low levels of type I IFN and release parasite RNA to interacting pDCs in the bone marrow of malaria-infected mice, leading to the massive production of type I IFN by the pDCs.
Severe infection with P. falciparum continues to result in high morbidity and mortality, despite robust antimalarial therapy. The clinical manifestations and pathology are well described and field studies have shown that severe malaria is associated with high inflammation, yet the mechanisms of immune dysregulation resulting in such high inflammation remain poorly understood. Adjunct measures to reduce the inflammatory syndrome and improve outcomes in severe malaria have been unsuccessful to date, and thus further studies to define the basis of immune dysregulation are needed [62]. In this work, we provide a comprehensive characterization of immune cells and inflammatory status in a cohort of pediatric severe malaria patients. We report high plasma levels of proinflammatory cytokines and chemokines (IFNγ, TNF, IL-6, CCL2, CCL3, CCL4, CXCL10) in severe malaria patients compared to baseline uninfected follow-up, which matched that of Py infection in mice in which we could conduct a temporal analysis. Only few studies have performed functional immunophenotyping during severe human malaria [21,25], and most of them focused on specific subsets of immune cells, namely NK cells, γδT cells and CD4 + T cells. Our study examined both myeloid and lymphoid cell compartments to provide a global analysis of blood immune leukocytes during severe human malaria. Data reveal both expansion and activation of inflammatory monocytes (CD14 + ), which express an exquisite machinery to sense pathogen-derived molecules, produce inflammatory cytokines and contribute to systemic inflammation [63]. We also found diminished pDC blood frequencies in severe malaria patients as reported before [51], yet our data further reveal that pDCs were activated. pDCs represent by far the major type I IFN-producing cells in vivo [57] in response to viral infections, during autoimmunity and even bacterial infection, through TLR7/ and TLR9/MyD88 sensing of nucleic acids. pDCs are capable of secreting type I IFN in response to Pf schizonts in vitro [51] and they can produce type I IFN in the P. chabaudi mouse model [36], both results consistent with our data.
While NK cell blood proportions did not differ in acutely infected versus follow-up patients, these cells were strongly activated in severe malaria patients and in the Py mouse model, in line with prior in vitro reports [64][65][66]. Likewise and as also suggested before [67], a slight decrease in the blood proportion of CD8 + and CD4 + T cells was observed in severe malaria patients and both subsets of T lymphocytes underwent robust activation, proliferation and differentiation.
Collectively, our data draw a parallel between the immune status of a cohort of patients with severe malaria and the surrogate P. yoelii 17X YM mouse model. Blood immune cells, notably CCR2 + monocytes, NK cells and pDCs underwent robust activation and common inflammatory mediators were noted. A similar expression pattern of cell-surface activation markers between severe malaria patients and the mouse model could be found, notably at later time points (d 4.5) post infection. While this remains speculative, it may suggest that the initiating events driving severe immune activation and inflammation, and poor clinical outcomes are comparable to that of human patients.
Recent reports have dissected fine sensing mechanisms of Plasmodium parasites. At liver stage, parasite RNA is detected by the cytosolic sensor Mda5 inside hepatocytes, which initiates the host immune response [68]. During blood stage, TLR7 via MyD88 may act as an early sensor of several Plasmodium species [41]. In vitro data have also reported that "AT-rich" stem loop motifs of Plasmodium DNA are recognized via the STING cytosolic pathway [35]. The current results make a link between these findings, and reveal the relative importance of both TLR7/MyD88 and STING sensing pathways during severe malaria infection in vivo, and in which cell subsets they act. We provide novel evidence that the bone marrow, a site undergoing major functional changes during this infection [48,69], may also represent an important site of initiation of the innate immune response against this parasite, a finding that has not been previously documented since most reports have focused on the spleen, and to some extent the blood. Intravital imaging of pDCs and CD169 + macrophages directly in the bone marrow of infected mice -which we are the first to report-further supports this interpretation. While pDCs exhibit a surveying and "hopping" type of behavior from one CD169 + macrophage to another in bone marrow and at steady state, their arrest and prolonged clustering around the macrophages after the infection are reminiscent to that of T or NK T cells after they "see" their cognate antigen [70,71].
Since both CD169 + macrophages and the STING pathways in the radio-resistant compartment are required to activate the type I IFN response, we concluded that STING exerts its function inside CD169 + macrophages. Such interpretation is consistent with STING sensing of Plasmodium AT-rich motifs in macrophage cell lines in vitro [35]. A major consequence of STING triggering leads to the production of type I IFN [60] and although we could not detect YFP + CD169 + macrophages in Ifnb-Yfp + reporter mice, it is likely that low levels of CD169 + macrophage-derived type I IFN in the bone marrow is among key signals to "prime" pDCs. Given that levels of IFNα/β are comparably reduced in Sting Gt/Gt and in Myd88 -/mice, it may also be that other, non-type I IFN dependent signals downstream of STING may drive immune cell activation.
pDCs also need to recognize parasites in a TLR7/MyD88-dependent manner in order to produce type I IFN, which suggests the requirement for several signals -at least two-to achieve their full type I IFN-producing capacity, through cell-intrinsic TLR7 and cell-extrinsic STING. Selective depletion of CD169 + macrophages largely prevents subsequent activation of immune effector cells while lacking STING only partially impairs (by~20-50%) immune activation. One likely possibility is that macrophages provide other signals to the pDCs, notably other non-RNA parasite-derived (or not) ligands and/or cytokines. Macrophages can scavenge iRBCs and may undergo subsequent activation and pyroptotis [34], ultimately releasing their cytosolic contents to interacting pDCs. The concomitant loss of CD169 + macrophages with their morphological changes observed by intravital microscopy in the bone marrow between 24 and 36 hours post Py infection is consistent with such interpretation. Another non-exclusive possibility could be that macrophages deliver signals to pDCs by secretion of exosomes-containing parasite RNA. pDCs may indeed be using such indirect mechanisms of activation as suggested in the case of viral infections [72][73][74].
The model we favor proposes that pDCs produce systemic type I IFN which orchestrates immune activation and contributes to deleterious outcomes during lethal malaria. Yet while selective depletion of pDCs prevents massive immune cell activation in vivo, it only has a modest impact on mice survival contrary to that of Ifnar1 -/mice, suggesting that low levels of type I IFN may be produced and consumed locally by many other cell types. Disrupting type I IFN signaling is also known to promote enhanced IFNγ production [75], a cytokine exhibiting protective effects during malaria [76], and this may account for these seemingly discrepant findings. Another possible interpretation is that IFNβ, in contrast to all other IFNα isoforms, have a unique non-Jak/STAT pathological signaling feature through IFNAR1, independent of IFNAR2 (the high affinity receptor for IFNα/β) as reported in a model of LPS-induced sepsis [77]. If true, disrupting IFNAR2 should not phenocopy the survival advantage of Ifnar1 -/mice but that of the WT counterparts. Thus clearly, further investigations will be needed to resolve these important issues in future work.
An additional intriguing difference with regards to current literature relates to the P. berghei murine model of CM, in which type I IFN signaling to DCs was shown to prevent Ly6C hi monocyte and cDC activation, and subsequent optimal priming of protective Th1 CD4 + T cell responses while its disruption enabled better control of parasite infection and host survival [37]. Despite similar outcomes, the proposed mechanisms have to be very different, which is likely accounted for by the use of very distinct surrogate models of severe malaria in which host resistance or death are not regulated by the same processes. For instance, in this model and in contrast to Py YM infection, the burst of systemic type I IFN occurs much later (day 4 versus day 1.5), parasitemia stay very low (1.5% versus~60% iRBCs by day 4), and significant differences in MHC class II expression on WT versus Ifnar1 -/monocytes are reported, altogether supporting the idea that immune cells are exposed to very different milieus both at early and late times post infection. Yet, these results overall underline the importance and the need for further investigations of the role of type I IFN in host resistance during malaria.

Materials and Methods
Human studies Patient population. To characterize immune responses associated with severe human malaria we examined Malawian children with CM, enrolled in the Blantyre Malaria Research Project (BMP), an ongoing longitudinal study of CM, during the 2013 transmission season [78]. Children between the ages of 6 months and 12 years with P.falciparum infection and a Blantyre Coma Score <3 and parental consent were enrolled, and those with a positive blood or CSF bacterial culture were excluded. Whole blood samples were collected at enrollment into the BMP, and at 30 days after discharge. A fundoscopic exam was performed to determine the presence (Ret + CM) or absence (Ret -CM) of malarial retinopathy [78]. Clinical characteristics and laboratory data were extracted from the study database.
Sample preparation. Six mLs of blood was collected from enrolled patients, and PBMCs were isolated using Ficoll separation within 6 hours of blood draw. Isolated PBMCs were resuspended in FCS/10% DMSO and immediately transferred to a Mr. Frosty freezing container and put in -80°C freezer for 24 hours, before being transferred and stored in liquid nitrogen and shipped to Albert Einstein College of Medicine for analysis.

Ethic Statements
Human research. IRB approvals were obtained from the Albert Einstein College of Medicine, Michigan State University and The University of Malawi College of Medicine Research and Ethics Committee. Written informed consent from the parent or guardian of any child participant was obtained on their behalf.
Animal research. This study was carried out in strict accordance with the recommendations by the animal use committees at the Albert Einstein College of Medicine. The institution is accredited by the "American Association for the Use of Laboratory Animals" (DHEW Publication No. (NIH) 78-23, Revised 1978), and accepts as mandatory the NIH "Principles for the Use of Animals". All efforts were made to minimize suffering and provide humane treatment to the animals included in the study. Approved study protocol is 20141202.

Generation of bone marrow chimera mice
WT B6 mice received a lethal dose of irradiation (1,200 rads), and were immediately reconstituted with 1-2x10 6 bone marrow cells isolated from flushed femurs from indicated donor mice, and antibiotic treated for 2 weeks. Reconstituting target ratio (post irradiation) and actual ratio (post-reconstitution in blood and spleen, 6-8 weeks later) of bone marrow donor cells are specified for each experiment legend and associated supplementary figures. We usually determined mixed chimera ratios to minimize the impact of cell-specific depletion and of having too high a proportion of knockout versus WT hematopoietic-derived cells (like in the case of WT/ Myd88 -/chimeras).
Plasmodium infections, measure of blood parasitemia and mice survival Infections. Plasmodium yoelii (17X), clone YM parasites (stock MRA-755) were obtained from the Malaria Research and Reference Reagent Resource Center as part of the BEI Resources Repository, NIAID, NIH (Manassas, VA; Deposited by D. Walliker). P. yoelii YMinfected red blood cells (iRBCs) from a frozen stock (stored in liquid nitrogen in Alsever's solution and 10% glycerol) were intraperitoneally (i.p.) injected into one WT B6 mouse and grown for~4 days. When parasitemia reached 2-5%, 2x10 5 P. yoelii YM iRBCs were infected intravenously into each experimental mouse, unless otherwise indicated.
Parasitemia. Blood parasitemia was determined by flow cytometry as described [86]. One μL of blood was obtained by cutting the tip of the tail with a sterile razor, and blood was deposited in 200μl of 0.025% glutaraldehyde in phosphate buffered saline (PBS) containing 1mM EDTA in a 96-well plate. Fixed red blood cells were washed with PBS and then permeabilized with 0.25% Triton X-100 (Sigma) in PBS for 5 min. After centrifugation, red blood cells were incubated in 1mg/mL RNAse A (Sigma) for 30 min at room temperature. Red blood cells were then stained with 0.5 μM of the YOYO-1 dye (Invitrogen) for 30 min at room temperature and directly analyzed on a BD FACSCanto II (Becton Dickinson, CA), equipped with a 488-nm argon laser, using a 530/30 emission filter to detect YOYO-1 fluorescence. Red blood cells were gated based on forward and side scatter, and parasitemia was determined as the frequency of YOYO-1 positive cells among all cells. Eye-counting of Giemsa-stained blood smears by microscopy confirmed that determination of parasitemia using the above method gave comparable results.
Mice survival. Mice were monitored for survival twice a day after day 4, in the morning and evening.

Preparation of cell suspensions for flow cytometry (FACS) analysis
Spleens were dissociated on nylon meshes (100μm) and incubated at 37°C for 20 min in HBSS medium (Gibco) containing 4,000 U/ml of collagenase I (Gibco) and 0.1 mg/ml of DNase I (Roche), and red blood cells (RBC) further lysed with NH 4 Cl buffer (0.83% vol/vol). Blood was harvested into heparin tubes and red blood cells lysed as above. Cell suspensions were further used for the different analyses (see below). Bone marrow was harvested from the femur by cutting off the ends of the femur, centrifuging the femur at 16000 rcf using an Eppendorf bench top centrifuge, and resuspending the resulting cell pellet in media, mixing until a single cell suspension was achieved.

Cell staining for FACS analysis
Mice. Cell suspensions prepared as above were incubated with 2.4G2 antibody for 15 min on ice and further stained with specified antibodies cocktail (See S1 Table) in FACS buffer (PBS 1% FCS, 2mM EDTA, 0.02% sodium azide). For intracellular IFNγ staining, cells were incubated for 3-4 hrs at 37°C, 5% CO 2 in RPMI 1640 (Invitrogen) 5% FCS, Golgi Plug (BD), fixed in IC fixation buffer (eBioscience) for 15 min at 4°C, and permeabilized for 30 min in 1X Perm/Wash buffer (eBioscience) containing indicated intracellular marker(s). For intracellular Granzyme B staining, unstimulated cells were fixed with Foxp3/Transcription Factor Fixation/ Permeabilization Buffer (eBioscience) for 15 min at 4°C, and permeabilized for 30 min in 1X Perm/Wash buffer (eBioscience) containing anti-Granzyme B. For determining absolute numbers of cells, we have made use of counting beads (SpheroTech) as internal standard added in cells isolated from either one femur or one spleen (20,000 beads) or from a given volume of blood (3960 beads for 120 μl of blood).
Human. Frozen Ficoll-separated human PBMCs (in FCS/10% DMSO) were thawed quickly in a 37°C water bath, until only a small ice pellet remained, transferred to 12mLs of room temperature RPMI 1640 (Invitrogen) 5% FCS, and immediately washed. Cells were then resuspended in FACS buffer for staining with specified antibody combinations. Intracellular Tbet and Ki67 staining were done as described for Granzyme B. Stained cells were collected either on a FACS BD LSR-II or Aria III. A list of all mAbs and sera used in the study with full information is provided in S1 Table. Data were analyzed using FlowJo version 9.6.2 (TriStar) on G5 Macintosh computers.

Measure of cytokine and chemokine contents
Mice. Spleens from immunized mice were harvested, snap-frozen and stored at -80°C until further processing. Spleens were then thawed in 150μl of lysis buffer (150mM NaCl, 40mM Tris pH 7.4) containing a 1mM PMSF and a cocktail of protease inhibitors (Promega), homogenized with a douncer, and supernatant frozen at -80°C, thawed and centrifuged to obtain cleared supernatants. For bone marrow, content was harvested as described above for FACS analysis, resuspended in 150μl of lysis buffer, centrifuged, and supernatant was frozen at -80°C. Serum was obtained by collecting blood in a unheparanized tube, allowing clotting for 15 minutes at room temperature before centrifugation for 20 min at 16000g, 4°C. Separated serum was aliquoted and stored frozen at -80°C. Dosing of indicated cytokines/chemokines was performed using FlowCytomix Multiplex Kits (ebioscience) on an LSRII flow cytometer and according to the manufacturer protocol. Dosing of IFNα and IFNβ was done using Veri-Kine Mouse Interferon ELISA Kits (PBL Assay Science).

Two-photon preparation, imaging and analysis
In vivo imaging of bone marrow was conducted as described previously [87]. Briefly, mice were anesthetized and the medial region of the tibial bone was surgically exposed, and carefully thinned to 100 micron thickness using a microdrill. Tissue was stabilized and mice were kept warm during imaging. Images were collected on an Olympus FV-1000MPE upright laser scanning microscope using a MaiTai DeepSee Laser (Spectraphysics). To visualize CD169 cells, mice were intravenously injected with 5μg of anti-CD169 (SER4) antibody conjugated to PE (BD Pharmingen), 16 hours prior to imaging [88]. All procedures were approved by the Institutional Animal Care and Use Committee.
Analysis. 4D image sets were analyzed on Volocity 6.3 (Improvision). For cell-contact duration between pDCs and CD169 + cells, analysis was conducted by tallying time durations for all pDCs in the field in contact with CD169 + myeloid cells. Data was pooled from 3 or more mice from independent experiments. Supplementary movies were assembled using After Effects© (Adobe).

Statistics
Statistical significance was calculated using an unpaired or a paired (human study) Student t test with GraphPad Prism software and two-tailed P values are given as: (