Thrombocytopenia Independently Leads to Changes in Monocyte Immune Function

Abstract

Background:

While platelets have well-studied hemostatic functions, platelets are immune cells that circulate at the interface between the vascular wall and white blood cells. The physiologic implications of these constant transient interactions are poorly understood. Activated platelets induce and amplify immune responses, but platelets may also maintain immune homeostasis in healthy conditions, including maintaining vascular integrity and T helper cell differentiation, meaning platelets are central to both immune responses and immune quiescence. Clinical data has shown an association between low platelet counts (thrombocytopenia) and immune dysfunction in sepsis and ECMO patients, further implicating platelets as more holistic immune regulators, but studies of platelet immune functions in non-disease contexts has had very limited study.

Methods:

We used in vivo models of thrombocytopenia and in vitro models of platelet and monocyte interactions, as well as RNA-seq and ATAC-seq, to mechanistically determine how resting platelet and monocyte interactions immune program monocytes.

Results:

Circulating platelets and monocytes interact in a CD47 dependent manner to regulate monocyte metabolism, histone methylation, and gene expression. Resting platelet-monocyte interactions limit toll-like receptor (TLR) signaling responses in healthy conditions in an innate immune training like manner. In both human sepsis patients and mouse sepsis models, thrombocytopenia exacerbated monocyte immune dysfunction, including increased cytokine production.

Conclusion:

Thrombocytopenia immune programs monocytes in manner that may lead to immune dysfunction in the context of sepsis. This is the first demonstration that sterile, endogenous cell interactions between resting platelets and monocytes regulate monocyte metabolism and pathogen responses, demonstrating platelets to be immune rheostats in both health and disease.

Summary.

Platelets regulate monocyte immune programming in a manner that is dependent on CD47 interactions that induces monocyte metabolism, epigenetic remodeling and immune programming. Therefore, a decline in platelet count leads to monocyte immune dysfunction.

Graphical Abstract

Introduction

Platelets are an integral part of pathogen and sterile tissue injury responses, including directly engulfing and killing bacteria, releasing bactericidal mediators, and regulating innate and adaptive immune cell differentiation/polarization and trafficking^1–4. Platelets also regulate immune cell functions and differentiation in non-diseased conditions, as evidenced by platelet deficient mice having increased T helper 17 (Th17) type of CD4⁺ T cells ⁵, mice lacking platelet-derived β2M having more reparative monocytes in basal conditions ^6,7, and thrombocytopenic mice having a loss in vascular integrity ⁸. These studies suggest that while activated platelets promote inflammatory responses to tissue injury or infection, resting platelets maintain immune quiescence in healthy conditions. Platelets therefore function as an immune rheostat to set the tone of the inflammatory environment.

The immune system has evolved to maintain tissue and vascular homeostasis but activate when faced with pathogen challenges or tissue injury. This requires the integration of many cells, including platelets and monocytes. Activated monocytes release inflammatory cytokines that trigger immune responses. This must be tightly regulated as dysregulated monocyte responses contribute to a “cytokine storm” and exaggerated inflammation, tissue damage and potentially mortality ⁹. The cellular and molecular mechanisms that regulate monocyte immune phenotypes are complex and include roles for platelets. Monocytes and macrophages also adapt to prior stimulation and respond differently to a second challenge to either enhance or limit the secondary immune responses, a process called innate immune training¹⁰. Metabolic reprogramming leading to epigenetic changes is an underlying mechanism for innate immune training and modified second stimuli responses¹¹. However, whether endogenous cell interactions mediate innate immune training has not been explored.

Sepsis is caused by dysregulated host responses to systemic infection that can lead to multi-organ failure^12,13. Dysregulated host responses result in a profound activation of cytokine networks, including elevated IL-6 and TNFα. Increased IL-6 is associated with increased sepsis severity and is a diagnostic and prognostic sepsis biomarker^14–17. A human platelet count below 150,000/μl is defined as thrombocytopenia and is common in patients with sepsis. Sepsis thrombocytopenia is thought to be due to infection driven platelet activation and consumption. The platelet count is included in ICU severity of sepsis illness scoring¹⁸ and there is an inverse correlation between the degree of thrombocytopenia and increased cytokine activation, organ dysfunction, and mortality ^19–22. Thrombocytopenia is also seen in about 20% of acute coronary syndrome (ACS) patients, up to 50% of patients put on extracorporeal membrane oxygenation (EMCO), as well as in infectious diseases such as malaria, and is similarly associated with increased risk of inflammatory complications and mortality in each condition ^23–28. Despite abundant clinical evidence that a decreased platelet count is predictive of inflammation severity, there is little mechanistic understanding of how thrombocytopenia may worsen immune driven disease outcomes.

We now demonstrate that resting platelets maintain monocyte immune quiescence in healthy conditions, and that thrombocytopenia independently leads to monocyte dysregulation. Using mouse models, we found that thrombocytopenia exacerbated monocyte inflammatory responses and immune dysregulation in a platelet CD47 interaction dependent manner, that limits monocyte inflammatory responses to toll-like receptor (TLR) agonists. Resting platelet expressed CD47 interactions with monocytes activate AKT/mTOR signaling and glycolysis, leading to histone methylation, chromatin remodeling, and changes in monocyte TLR responses. This study also represents the first demonstration of endogenous cell mediated innate immune programming that limits inflammation in healthy conditions.

Methods

Standard methodologies are detailed in supplemental materials.

Sequencing data GEO Submission Accession numbers: GSE256083, GSE256084, GSE256085

Results

Thrombocytopenia changes monocyte phenotypes

Our past studies demonstrated that genetically thrombocytopenic mice (TPOR^−/−) had more Th17 type of CD4⁺ T cells²⁹. To begin to determine whether thrombocytopenia affects innate immune cells, we isolated circulating neutrophils and monocytes from WT and TPOR^−/− mice and measured Il6, Cxcl1 and Tnfa expression. Circulating monocytes, but not neutrophils, from TPOR^−/− mice expressed more Il6 and Cxcl1 and produced more IL-6 and CXCL1 in response to LPS, with no changes in bone marrow (BM) monocyte gene expression (Supplemental Fig 1A). To demonstrate a more direct effect of acute thrombocytopenia on innate immune cells, we used a PF4-Cre × diphtheria toxin receptor (DTR)^fl/fl (PF4-DTR) mouse model³⁰. Upon DT treatment megakaryocytes in PF4-DTR mice undergo apoptosis leading to a decline in platelet count by d1 post-DT, a platelet nadir at d5, and a rebound by d7 (Fig 1A). DT treatment changed platelet counts, but not other cell counts (Supplemental Fig 1B), and on both d3 and d5 plasma from DT treated WT and PF4-DTR mice had similar HMGB1 and S100A8/A9, indicating that DT induced thrombocytopenia does not increase DAMPs (Supplemental Fig 1C). We isolated circulating neutrophils or monocytes from control and thrombocytopenic mice on d5 post-DT when each were at least two generations post-DT, and quantified Il6, Cxcl1, and Tnfa mRNA, as well as their induced production of each cytokine in response to LPS. Circulating neutrophils from control and thrombocytopenic mice had similar Il6, Cxcl1 and Tnfa expression and production (Fig 1B and Supplemental Fig 1D). Circulating monocytes isolated from platelet depleted mice had increased Il6 and Cxcl1 and secreted more IL-6 and CXCL-1 in response to LPS, with no significant differences in BM monocytes (Fig 1 B–C and Supplemental Fig 1E, unstimulated monocytes in any conditions did not produce detectable cytokines by ELISA so data not shown). Similarly, circulating monocytes isolated from mice made thrombocytopenic using anti-GPIbα antibody exhibited increased Il6 and Cxcl1 compared to IgG treated controls. (Supplemental Fig 1F). To further demonstrate a platelet number dependent effect, we isolated monocytes from WT and PF4-DTR mice on d10 post-DT when mice were no longer thrombocytopenic. Prior thrombocytopenic and control mice had similar monocyte Il6 and Tnfa and LPS mediated cytokine secretion (Supplementary Fig 1G–H). Together, these data indicate that normal platelet numbers limit the expression and production of at least some monocyte proinflammatory cytokines.

Figure 1.

Thrombocytopenia alters circulating monocyte gene expression. A) Acute model of thrombocytopenia. PF4cre-DTRflox mice or WT control mice were treated with diphtheria toxin (DT) (400 ng/mouse) and platelet counts determined on multiple days post-DT. n=4 mice in each group. B) Circulating neutrophils, circulating and BM monocytes were isolated from control and thrombocytopenic mice and gene expression determined by RT-qPCR, normalized to Actb; n=independent neutrophils isolated from 4 mice; n=independent monocytes isolated from 6 mice. C) Isolated circulating monocytes from WT and thrombocytopenic mice were treated with LPS ex vivo and 24 hrs later, IL-6 and CXCL-1 were determined by ELISA; n=6 mice in each group. D-E) Circulating monocyte mRNA was isolated on d5 post-DT from control and thrombocytopenic mice for RNA-seq; n=3 in each group. D) GO analysis of up and down regulated genes indicated increased inflammatory and decreased metabolism gene expression in thrombocytopenic mouse monocytes, E) confirmed by RT-qPCR; n=6 mice in each group. F) Glycolytic rate assays were performed on circulating monocytes from WT and DT-induced thrombocytopenic mice; n=4 WT mice; n=8 thrombocytopenic mice. Data were represented as mean±SEM. Statistics: unpaired, 2-tailed Student’s t test in B (circulating and BM monocytes), C and E (Cxcl1, Il1b, Cxcl2, Pfkl, Tpi1, Ldha and Hif1a); Welch t test in E (Cxcl5); Mann-Whitney test in B (circulating neutrophils), E (Il6) and F.

To gain a more comprehensive understanding of circulating monocyte thrombocytopenia associated changes we performed RNA-seq on circulating monocytes from control and PF4-DTR mice on d5 post-DT and found different monocyte gene expression patterns (Supplemental Fig 1I–K). Pathway analysis indicated that monocytes from thrombocytopenic mice had higher expression of genes associated with LPS/bacteria responses and inflammation, decreased expression of genes in glycolysis/glucose metabolism, cell proliferation, and changes in multiple cytokines and immune molecules (Fig 1D). Select inflammatory genes and glycolysis associated genes were validated by qRT-PCR (Fig 1E). To functionally validate that thrombocytopenia altered monocyte metabolism, Seahorse glycolytic rate assays were performed on circulating monocytes from control and thrombocytopenic mice. After isolation and washing, circulating monocytes from TPOR^−/− and thrombocytopenic PF4-DTR mice had very few platelets attached to them (~12%), similar to WT mice (Supplemental Fig 1L). Monocytes isolated from both TPOR^−/− mice (Supplemental Fig 1M) and DT-induced thrombocytopenic mice (Fig 1F and Supplemental Fig 1N) had lower extracellular acidification rate (ECAR) in both basal conditions (basal glycolysis) and after rotenone/antimycin A treatment (compensatory glycolysis) compared to controls. These results demonstrated that resting platelet interactions with monocytes regulate monocyte immune and metabolic programming.

To mechanistically explore this concept and the potential mechanisms, we performed overnight co-cultures of platelet naïve mouse BM monocytes with resting platelets or control media at multiple platelet:monocyte ratios, before washing monocytes to remove platelets and stimulating the monocytes with LPS. After co-culture, about 90% of monocytes had platelet aggregates, but gentle washing reduced this to less than 10% (Supplemental Fig 2A). Co-culture with resting platelets did not change monocyte survival (Supplemental Fig 2B). Prior monocyte-platelet co-culture limited LPS induced IL-6 and TNFα in a platelet to monocyte ratio dependent manner (Fig 2A and Supplemental Fig 2C), with similar results using bone marrow derived macrophages (BMDMs- bone marrow isolated monocytes differentiated to macrophages in vitro) (Fig 2A and Supplemental Fig 2C). Human CD14⁺CD16⁻ circulating monocytes (Fig 2A and Supplemental Fig 2C) and THP-1 cells (Supplemental Fig 2D) also had less LPS induced IL-6 and TNFα after overnight co-culture with human platelets in a ratio dependent manner. An ex vivo flow-and-cone system was used to recapitulate hemodynamic-like continuous flow³¹. Steady-flow with platelets and monocytes resulted in similar pre-wash platelet-monocyte aggregates as static conditions, and similarly reduced LPS induced IL-6 (Supplemental Fig 2E). Based on these data, all subsequent platelet-monocyte co-cultures were performed in static conditions at 50:1 platelet:monocyte ratio.

Figure 2.

Resting platelets change monocyte gene expression in a contact dependent manner. A) Resting platelets limited monocyte and BMDM cytokine production in a platelet ratio dependent manner. Primary mouse BM monocytes, BMDMs or human CD14⁺CD16⁻ monocytes were co-incubated overnight with platelets at multiple platelet:monocyte ratios and cells then washed to remove platelets before LPS stimulation. IL-6 was quantified 24 hrs later; n=6 replicates from 3 mice; n=4 replicates from 1 human blood donor. B) Monocytes were cultured with platelets directly or separated in a transwell chamber overnight and monocytes then LPS stimulated; n=4 replicates from 2 mice. C-D) Monocytes were incubated overnight in control buffer or with platelets. Platelets were removed and mRNA isolated for RNA-seq or monocytes were LPS stimulated and 4 hrs later mRNA isolated for RNA-seq. C) GO analysis of up and down regulated genes in platelet cultured monocytes before LPS stimulation and D) confirmation of select metabolism and immune related genes after incubation in control or platelet co-culture conditions; n=6 replicates from 2 mice. E) CXCL-1 and CCL-2 production were determined in monocytes cultured with control media or platelets then LPS stimulated. n=3–4 replicates from 2 mice. F) Monocytes/BMDMs were cultured with control media or platelets overnight and platelets washed away before TLR2 agonist, TLR9 agonist or IL-1β stimulation. IL-6 was determined by ELISA 24hs later; n=4–6 replicates from 2 mice. Data were represented as mean±SEM. Statistics: One-way ANOVA followed by Tukey’s multiple comparisons in A (monocytes and BMDMs); Kruskal-Wallis followed by Conover-Iman post hoc test with Holm-sidak corrections in A (human monocytes) and B; Unpaired, 2-tailed Student’s t test in D and F (TLR2 and TLR9 agonists); Mann-Whitney test in E and F (IL-1β).

Platelets interact with immune cells through both contact and secreted molecule mediated mechanisms^1,32,33. To determine whether resting platelet regulation of monocyte LPS responses is contact dependent, transwell chambers were used to separate platelets from either BM monocytes or BMDMs in the lower chamber overnight. When platelets and monocytes/BMDMs were physically separated, platelets no longer limited LPS induced IL-6 (Fig 2B and Supplemental 2F). Furthermore, overnight incubation with platelet releasates or platelet conditioned media did not limit monocyte IL-6 responses to LPS (Supplemental Fig 2G), highlighting a direct contact dependent mechanism. Platelets that were activated and then washed to remove platelet releasates still limited LPS responses, further demonstrating a contact dependent mechanism of immune quiescence, while also indicating that activated platelet derived products induce different, more activating monocyte responses (Supplemental Fig 2G).

To determine whether the gene changes in thrombocytopenic mice were directly platelet mediated, we performed RNA-Seq on monocytes incubated in control media (NC) or with platelets overnight (PLT) in vitro, as well as 4 hrs after LPS stimulation. Platelet treated monocytes had different gene expression patterns compared to control media treated monocytes, both before and after LPS (Supplemental Fig 2H). Resting platelets increased gene pathways associated with glucose metabolism/glycolysis and immune regulatory genes, while gene pathways suppressed by platelets included responses to bacteria origin/LPS, (Fig 2C), confirming a direct platelet dependent mechanism, consistent with LPS stimulation (Supplemental Fig 2I). Select immune and glycolysis gene changes were confirmed by qRT-PCR (Fig 2D). We functionally confirmed select changes, including resting platelet interactions limiting LPS induced CXCL1, but increasing CCL2, IL-10 (Fig 2E and Supplemental 2J), and monocyte proliferation (Supplemental Fig 2K). Consistent with the in vivo RNA-seq, prior platelet co-culture attenuated monocyte TLR agonist responses, including to Pam2CSK4 (TLR2 agonist) and CpG (TLR9 agonist), but had little effects on IL1β induced CXCL1, indicating resting platelet mediated immune regulation is skewed to regulating TLR responses (Fig 2F and Supplemental Fig 2L). Platelet and monocyte co-cultures limited Ifnb expression (Supplemental Fig 2M). These data support the concept that resting platelet interactions genetically program monocytes.

Thrombocytopenia in sepsis independently increases monocyte cytokine responses in a platelet CD47 dependent manner

Thrombocytopenia is a clinical biomarker for sepsis immune dysfunction, but a direct association between thrombocytopenia and monocyte dysfunction has not been explored. We therefore isolated sepsis patient plasma at the time of sepsis diagnosis (Supplemental Table 1), measured IL-6 and IL-8 by ELISA, and correlated with platelet counts. IL-6 and IL-8 negatively correlated with platelet counts (Fig 3A). To correlate more directly with monocyte derived cytokines, we isolated circulating monocytes from sepsis patients at the time of diagnosis, as well as from those admitted with non-septic illness (Supplemental Tables 2–3), and measured IL6, IL8 (Fig 3B) and TNFA (Supplemental Fig 3A) by qRT-PCR. Each negatively correlated with platelet counts, demonstrating a direct correlation between thrombocytopenia and increased monocyte cytokines in a disease context.

Figure 3.

Thrombocytopenia increased bacteria associated immune activation in vivo. A) Plasma IL-6 and IL-8 inversely correlated with platelet counts in septic patients. Plasma IL-6 and IL-8 were determined by ELISA and correlation to the platelet count was determined by linear regression followed by Pearson correlation analysis. n=12 individual septic blood donors. B) Circulating monocyte IL6 and IL8 inversely correlated with platelet counts in both septic and non-septic human patients. Monocyte IL6 and IL8 relative to the platelet counts were determined by linear regression followed by Pearson correlation analysis. n=20 (sepsis patient); n=7 (non-septic patient). C) Thrombocytopenic mice or WT control mice were treated with LPS (20 mg/kg) i.p. on d5 post-DT and 4 hrs later plasma IL-6, CXCL-1 and TNFα were determined by ELISA. n=6 mice in each group. D) Male and female thrombocytopenic mice and WT control mice were injected with cecal slurry (CS) i.p. (0.8 mg/g) and 4 hrs later plasma IL-6 and TNFα were determined by ELISA. n=5 mice in each group. E) Monocytes/macrophages were depleted with chlodronate liposomes on d4 and d5 post-DT and 4 hrs after the second dose, mice were LPS treated and 4 hrs later plasma IL-6 and TNFα were measured by ELISA. n=4 mice in each group. Data were represented as mean±SEM. Statistics: linear regression followed by Pearson correlation analysis in A and B; Unpaired, 2-tailed Student’s t test in C (IL6 and CXCL-1); Welch t test in C (TNFα); 2-way aligned rank transformation (ART) ANOVA with Sidak corrections in D and E.

To demonstrate mechanistic associations between thrombocytopenia and the potential for immune dysfunction, WT and thrombocytopenic PF4-DTR mice were given LPS (20 mg/kg) i.p. on d5 post-DT when all circulating innate immune cells were platelet naïve. All mice developed hypothermia as early as 2 hours post-LPS. Thrombocytopenic mice had a more dramatic early drop in temperature compared to WT mice and were similar 24h after LPS (Supplemental Fig 3B). Thrombocytopenic mice had a greater mortality compared to WT mice (Supplemental Fig 3C). Lethality in sepsis is associated with severe systemic inflammation. Therefore, we examined the plasma inflammatory cytokine levels in WT and thrombocytopenic mice. Plasma IL-6, CXCL1 and TNFα were all increased in thrombocytopenic mice 4 hrs post-LPS (Fig 3C). We observed a decreasing trend in plasma levels of CCL2 and IL-10 in thrombocytopenic mice but no significant difference was seen in plasma IFN-γ level (Supplemental Fig 3D). Similarly, in antibody mediated platelet depletion mouse model, mice treated with anti-GPIbα Ab 2d prior had increased IL-6 and TNFα production in response to LPS challenge (Supplemental Fig 3E). Using a polymicrobial i.p. cecal slurry (CS) sepsis model, both male and female PF4-DTR thrombocytopenic mice had increased plasma IL-6 and TNFα 4 hrs after CS (Fig 3D). TPOR^−/− also had increased plasma cytokines in response to CS (Supplemental Fig 3F). When monocytes/macrophages were depleted using liposomal clodronate prior to LPS challenge (Supplemental Fig 3G), the exaggerated inflammatory responses in thrombocytopenic mice were attenuated demonstrating monocyte/macrophage dependent thrombocytopenic responses (Fig 3E).

Resting platelets constitutively express surface receptors, glycoproteins, and integrins, and activated platelets increase the expression of P-selectin and CD40L that facilitate interactions with monocytes ³³. P-selectin or CD40L blocking antibodies had no effects on resting platelet inhibition of LPS induced monocyte IL-6 (Supplemental Fig 4A). The integrin blocker RGDS (Fig 4A) and blocking antibodies to PSGL-1 and Mac-1 complex (CD11b and CD18) also had no effects (Supplemental Fig 4A). We also verified that platelet induced monocyte responses were not dependent on phosphatidylserine (PS), thrombopoietin (elevated during thrombocytopenia) or TGFβ (Supplemental Fig 4A). Platelets constitutively express CD47, and anti-CD47 antibody blocked the platelet dependent effect on LPS induced monocyte IL-6 (Fig 4A). Similarly, monocytes incubated with CD47^−/− platelets did not have attenuated LPS responses (Fig 4A), despite similar platelet-monocyte aggregates (Supplemental Fig 4B). CD47 is an immunoglobulin (Ig) superfamily membrane protein expressed on hematopoietic and non-hematopoietic cells. It serves as a receptor for thrombospondin (TSP) and is a ligand for the SIRPα receptor^34–36. CD47 is most recognized for delivering ‘don’t eat me’ signals that regulate red blood cell (RBC) clearance. However, CD47 has cell signaling roles, CD47^−/− mice have no changes in platelet count ³⁷ (Supplemental Figure 4C, WT, TPOR^−/− and CD47^−/− CBC), and RBC-monocyte co-culture did not limit LPS induced IL-6, indicating that platelet CD47 may primarily be a signaling molecule (Fig 4B). CD47^−/− platelets have normal surface β3 integrin and GPIbα level compared to WT platelets (Supplemental Fig 4D). Circulating monocytes, not BM monocytes isolated from mice treated with CD47 blocking antibody had increased basal Il6 and secreted more IL-6 in response to LPS ex vivo compared to IgG treated mice (Fig 4C and Supplemental Fig 4E). These data were recapitulated in circulating monocytes from CD47^−/− mice (Supplemental Fig 4F). Mice treated with anti-CD47 antibody prior to in vivo LPS challenge also had increased trends of plasma IL-6 and TNFα compared to IgG treated controls (Supplemental Fig 4G). To demonstrate platelet CD47 specificity in vivo, thrombocytopenic mice were transfused with either saline, washed WT platelets, or CD47^−/− platelets on d4 post-DT, and given LPS on d5. Increased plasma IL-6 and TNFα in thrombocytopenic mice was largely reversed by the transfusion of WT, but not CD47^−/− platelets (Fig 4D). These results suggest that resting platelet regulation of monocyte LPS responses is platelet CD47 dependent.

Figure 4.

Platelet CD47 mediates monocyte immune programming. A) Monocytes were incubated with control media, control platelets, RGDS treated platelets, anti-CD47 Ab treated platelets, or with CD47^−/− platelets overnight and then LPS stimulated. n=6 replicates from 2 mice. B) Monocytes were incubated with control media, control platelets or RBCs overnight and then LPS treated. n=4 replicates from 2 mice. C) Mice were treated with anti-CD47 Ab or control IgG on d0, d2, and d4. On d5, peripheral monocytes were isolated for qRT-PCR or cultured and stimulated with LPS for 24 hs ex vivo. n=5–6 mice in each group D) Thrombocytopenic mice were transfused with saline, WT platelets or CD47^−/− platelets. 24 hrs later, mice were treated with LPS and plasma IL-6 and TNFα were quantified 4 hrs later. n=4 mice in each group. E) WT or CD47^−/− monocytes were incubated with either WT or CD47^−/− platelets overnight and LPS stimulated after platelets removed. n=6 replicates from 3 mice. F-G) Resting platelets limit macrophage phagocytosis and increase monocyte migration to chemokines. F) BMDMs were treated with platelets overnight and then co-cultured with latex bead-rabbit FITC-IgG for 2 hrs. Phagocytosis was quantified by flow cytometry. n=6 replicates from 2 mice. G) Platelet pre-treated monocytes were seeded in the top of trans-well chamber and CCL2 in the bottom chamber. 24 hrs later, migrated monocytes were stained with crystal violet and quantified under microscope. 20 random fields were assessed per group; n=4 replicates from 2 mice. The representative images were chosen based on their quality and to most accurately reflect the group average across all the available data. Data were represented as mean±SEM. Statistics: One-way ANOVA followed by Tukey’s multiple comparisons test in A, E and F; Kruskal-Wallis followed by Conover-Iman post hoc test with Holm-sidak corrections in B, D and G; Unpaired, 2-tailed Student’s t test in C (circulating monocyte gene expression); Mann-Whitney test in C (LPS treated ex vivo); Scale bar, 200μm.

Most studies of CD47 interactions have focused on SIRPα, but neither SIRPα blocking antibody nor SHP (downstream effector of SIRPα) inhibition (PTP I inhibitor) prior to platelet-monocyte co-culture changed platelet mediated LPS signaling inhibition (Supplemental Fig 4H). CD47 can form homotypic interactions with CD47 on other cells to elicit signal transduction, but much less is known about the associated signaling^38,39. WT platelet mediated IL-6 inhibition was limited when co-cultured with CD47^−/− monocytes, indicating an essential role for monocyte CD47 in platelet mediated LPS tolerance (Fig 4E). These data suggest that platelet-monocyte CD47 homotypic interactions are crucial to platelet mediated suppression of monocyte LPS responses.

Cytokine secretion is only one monocyte function, so we determined whether resting platelets broadly regulate monocyte and macrophage functions in a CD47 dependent manner, including phagocytosis and trafficking. BMDMs were incubated with control buffer, WT, or CD47^−/− platelets overnight and washed prior to incubation with latex bead FITC-IgG. WT, but not CD47^−/− platelets, reduced BMDM phagocytosis (Fig 4F). To explore monocyte trafficking, monocytes were co-cultured with buffer, WT or CD47^−/− platelets overnight prior to being washed and added to the top of a transwell chamber with CCL2 or CCL5 in the lower chamber overnight. Both WT and CD47^−/− platelets increased monocyte migration to the lower membrane (Fig 4G and Supplementary Fig 4I) demonstrating that resting platelets regulate cytokine secretion and phagocytosis in a CD47 dependent manner, but not monocyte trafficking.

Resting platelet and monocyte interactions immune program monocytes by regulating histone methylation

Innate immune cells, including monocytes and macrophages, can durably adapt to prior stimuli and mount greater (trained immunity) or suppressed (innate immune tolerance) responses to secondary stimulation with the same or a different agonist many days after the first stimuli ^{10,11,40–43}. To determine whether platelets durably alter monocyte and macrophage LPS responses, monocytes or BMDMs were left in culture media for 48 hrs after platelet removal prior to stimulation. Prior platelet interaction limited monocyte and BMDM IL-6 and increased proliferation 48 hrs later (Fig 5A and Supplemental Fig 5A), and similar results were found using RAW cells (mouse macrophage line) up to 6d after platelet removal (Fig 5A). Glycolysis dependent epigenetic programming and changes in chromatin organization form the molecular basis of durable innate immune training and tolerance¹⁰. Trained immunity is contingent upon inflammatory gene regions having more open/accessible chromatin, and immune tolerance/quiescence on gene regions having more condensed chromatin ^40,44. ATAC-Seq to measure chromatin accessibility across the genome on circulating monocytes from DT treated WT and thrombocytopenic mice on d5 post-DT demonstrated that thrombocytopenic mouse monocytes had a distinct chromatin configuration, with more open chromatin regions (Fig 5B, Principal component analysis (PCA) and Supplemental Fig 5B). The more accessible chromatin regions were associated with the regulation of cell adhesion, inflammatory responses, positive regulation of defense responses, or responses to LPS/TLR related pathways (Fig 5C), consistent with both RNA-seq and functional data. Among the 40 genes related to LPS response/TLR pathways identified in RNA-Seq, 30, including Il6, Cxcl1 and Il1b, had chromatin regions that were more accessible in monocytes from thrombocytopenic mice, while Tnfa was unchanged compared to controls, consistent with prior data (Fig 5D). This suggests that chromatin remodeling is at least in part responsible for platelet mediated monocyte immune programming.

Figure 5.

Resting platelets durably program monocytes. A) Platelet mediated effects on monocyte/macrophage LPS responses are durable over many days. Monocytes were incubated with platelets overnight and washed to remove platelets. Cells were then LPS stimulated 48 hrs later. IL-6 production was determined 24hs later and for proliferation, cells were stained with CFSE before platelet-co-culture. 24 hs after LPS treatment, proliferation was assessed by CFSE dye dilution using flow cytometry. Similarly, platelets limited RAW cell LPS induced IL-6 up to 6d after platelet removal. n=4 replicates from 3 mice. B-D) Circulating monocytes were isolated on d5 post-DT from control and thrombocytopenic mice for ATAC-Seq. n=3 mice in each group. B) PCA plot showed monocytes from control and thrombocytopenic mice had different chromatin configurations. C) GO analysis on more accessible monocyte DNA regions from thrombocytopenic mice were related to inflammatory and bacteria responses. D) Venn diagram of the intersection between altered LPS response genes in RNA-Seq and ATAC-Seq from thrombocytopenic mouse monocytes compared to WT monocytes. Heatmap showing normalized accessibility fold change (relative to WT monocyte chromatin accessibility) of those 30 LPS response associated gene regions. Data were represented as mean±SEM. Statistics: Mann-Whitney test in A (LPS stimulated 48h later); 2-way aligned rank transformation (ART) ANOVA with Sidak corrections in A (LPS stimulated 4d or 6d later).

To functionally determine how resting platelet interactions with monocytes may lead to histone modifications, platelet-monocyte co-cultures were performed in the presence of histone deacetylases (HDACs) (SAHA, LAQ824), histone demethylases (pargyline) or histone methyltransferases inhibitors (MTA). Only MTA limited platelet mediated monocyte LPS responses (Fig 6A and Supplemental Fig 5C), indicating that histone methylation mediates platelet driven immune programming. H3K9me3 and H3K27me3 are histone methylation modifications associated with repression of immune-related genes ^45,46. Circulating monocytes isolated from thrombocytopenic and CD47^−/− mice had less H3K9me3 and H3K27me3 compared to WT mouse monocytes (Fig 6B), and monocyte H3K9me3 and H3K27me3 were increased more by WT platelets than CD47^−/− platelets co-culture in vitro (Fig 6C). In contrast, histone H3K9 acetylation (H3K9ac) was similar in each condition (Fig 6B–C). These data indicate that resting platelet interactions increase histone methylation in a CD47 dependent manner.

Figure 6.

Resting platelets program monocytes through histone methylation. A) Monocytes were treated with inhibitor of HDACs (SAHA), histone demethylase inhibitor (pargyline) or histone methyltransferase inhibitor (MTA) prior to co-culture with platelets overnight and then platelets washed away and LPS stimulated. n=4 replicates from 2 mice. B) Immunoblots of H3K9me3, H3K27me3 and H3K9ac in circulating monocytes from CD47^−/−, thrombocytopenic and WT mice. n=3 mice in each group. C) Immunoblots of H3K9me3, H3K27me3 and H3K9ac in monocytes cultured with control media, WT platelets or CD47^−/− platelets in vitro overnight. n=3 replicates from 3 mice. Data were represented as mean±SEM. Statistics: 2-way aligned rank transformation (ART) ANOVA with Sidak corrections in A; Kruskal-Wallis followed by Conover-Iman post hoc test with Holm-sidak corrections in B and C.

Metabolic programming regulates innate immune training through epigenetic modifications that are in part dependent on metabolite regulation of histone methylation and demethylation ^47–49. Co-culture of WT platelets and monocytes increased monocyte glycolytic rates, but this was limited with CD47^−/− platelets and CD47 antibody blocking (Fig 7A and Supplemental Fig 5D). Similar to thrombocytopenic conditions, circulating monocytes isolated from CD47^−/− mice displayed lower glycolytic rates compared to WT monocytes. (Supplemental Fig 5E). AKT/mTOR signaling is a metabolic environment sensor that regulates glycolysis, innate immune cell development, and cytokine production ^50–52. WT, but not CD47^−/− platelets, increased phosphorylation of both AKT and mTOR (Fig 7B). When AKT was inhibited with MK2206 and mTOR with rapamycin, platelet induced changes in LPS responses were blunted (Fig 7C). Platelet mediated monocyte histone methylation were also blocked by MK2206 and Rapamycin (Fig 7D), demonstrating a functional relationship between platelet CD47, AKT/mTOR, and histone modifications. Inhibition of monocyte glycolysis using either glucose-free media (Fig 8A) or 2-deoxy-glucose (2-DG) (Supplemental Fig 5F) also blunted platelet mediated histone methylation and LPS responses (Fig 8A–8C) demonstrating a crucial role for platelet mediated glycolysis induced histone methylation in monocyte immune programming.

Figure 7.

Platelet mediated metabolic programming leads to monocyte histone methylation in a platelet CD47 dependent manner. A) Glycolytic rate assays were performed on monocytes incubated with control media, WT platelets, CD47 blocking Ab treated, or CD47^−/− platelets overnight. n=6 replicates from 6 mice. B) Immunoblots of phosphorylated AKT and mTOR in monocytes following overnight culture with control buffer, WT or CD47^−/− platelets. n=3 replicates from 3 mice. C) Monocytes were treated with mTOR or AKT inhibitors prior to platelet co-culture, platelets washed away and monocytes were treated with LPS. n=3 replicates from 2 mice. D) Immunoblots of H3K9me3, H3K27me3 and H3K9ac in monocytes treated with inhibitors of AKT (MK2206) or mTOR (Rapamycin) prior to co-culture with platelets in vitro. n=3 replicates from 4 mice. Data were represented as mean±SEM. Statistics: One-way ANOVA followed by Tukey’s multiple comparisons test in A; Kruskal-Wallis followed by Conover-Iman post hoc test with Holm-sidak corrections in B and C; 2-way aligned rank transformation (ART) ANOVA with Sidak corrections in D.

Figure 8.

Platelet mediated monocyte glycolysis is essential for monocyte immune programming. A) Immunoblots of H3K9me3, H3K27me3 and H3K9ac in monocytes cultured with platelets under control media or glucose-free media. n=3 replicates from 2 mice. B-C) Monocytes were treated with B) 2-DG prior to platelet co-culture or C) co-culture performed in glucose-free media for 24 hrs and then LPS treated. n=3 replicates from 2 mice. Data were represented as mean±SEM. Statistics: 2-way aligned rank transformation (ART) ANOVA with Sidak corrections in A, B and C.

Together, these data demonstrate that platelets limit monocyte TLR responses in healthy conditions, in a manner that is dependent on CD47 signaling induced metabolism and epigenetic programming (Graphical Abstract).

Discussion

We have discovered that circulating platelet-monocyte CD47 interactions activate AKT/mTOR signaling and glycolysis, which leads to histone methylation and gene expression changes that limit monocyte responses to TLR agonists (Graphical Abstract). Therefore, a decline in platelet count may promote immune dysfunction (Graphical Abstract). This makes physiologic sense: thrombocytopenia is indicative of significant vascular injury or infection and signals the need for heightened immune responses, making platelet numbers an immune biosensor. These data lead us to propose the novel concept that activated platelets induce immune cell activation, while a decline in platelet count provides a physiologic cue of the need to heighten immune responses, so thrombocytopenia leads to a propensity for immune dysfunction in the context of an infectious or inflammatory insult.

This study not only provides novel insights into prior unknown roles for platelets in immune homeostasis, but it also represents a novel discovery of endogenous cell based innate immune training. In the context of sepsis, this means that infection induced platelet activation early post-infection drives monocyte activation that is exacerbated by a decline in platelet count. Monocytes are consumed early in infection but continue to be rapidly produced from the bone marrow, meaning thrombocytopenia amplifies the inflammatory priming of monocytes as they rapidly emerge from the marrow⁵³. Strategies to limit sepsis associated thrombocytopenia may help to limit immune dysfunction. Thrombocytopenia is also seen in about 20% of acute coronary syndrome (ACS) patients, up to 50% of patients put on EMCO, as well as in other infectious diseases such as malaria, and is similarly associated with increased risk of inflammatory complications and mortality ^23–28. Results of this study may be more broadly generalizable and lead to a deeper mechanistic understanding of immune dysfunction.

Innate immune training continues to be conceptually defined. This is the first demonstration that platelets regulate monocyte metabolism and gene programming in healthy conditions and demonstrates a sterile cell-cell interaction function for platelets that is mechanistically similar to pathogen driven immune training. Monocyte metabolic activity can either heighten or limit responses to a second challenge. How specific metabolites regulate epigenetic modifications, and as a result specific immune responses, is less well understood. It is in part dependent on regulating the activities of histone acetylation and methylation modifiers, but specific metabolites that may lead to these changes downstream of platelet CD47 activation of AKT/mTOR is an area of active and future investigation. This and our past studies have demonstrated that thrombocytopenia immune programs CD4⁺ T cells and monocytes, but platelets do not appear to directly regulate basal neutrophil functions. Thrombocytopenia also leads to a decline in vascular integrity, but whether it is dependent on platelets genetically programming endothelial cells is yet to be defined.

CD47 is most studied for delivering ‘don’t eat me’ signals, and a decline in RBC CD47 signals the clearance of aged RBCs. CD47 was once referred to as integrin associated protein (IAP) because it is found in an integrin complex, but integrin blocking did not affect platelet induced LPS tolerance. Incubating monocytes with RBCs did not alter LPS responses, indicating that platelet CD47 has unique functions that RBC CD47 does not. CD47 complexes vary between cell types and CD47 can form a cell membrane complex with FasR, CD14, TLR4, VEGFR, CD36 and integrins, as well as associating with lipid rafts and GPCRs ^54,55. CD47 can also be heavily glycosylated affecting its protein interactions and signaling potential, and the cytoplasmic tail of CD47 is alternatively spliced in different cell types. Monocytes treated with purified recombinant CD47 protein showed a slight decrease in IL-6 production (Supplemental Fig 5G) further indicating cell specific CD47 complexes and modifications may account for platelet, but not RBC, CD47 modulation of monocyte immune tolerance. Much further study is needed to define these unique platelet CD47 signaling functions.

Our data clearly demonstrates that thrombocytopenia makes sepsis associated inflammation worse. However, it does not necessarily mean platelet transfusions would be beneficial in sepsis. Platelet activation is noted early in sepsis, and as platelets release their granule contents it promotes vascular inflammation and immune activation. Sepsis can then lead to thrombocytopenia and our data indicates that as monocytes emerge into a platelet deficient environment they produce more inflammatory cytokines, perhaps making sepsis worse. A platelet transfusion at that time point may not be helpful if it is associated with platelet activation. Rather more direct therapies, such as granule deficient platelets or direct monocyte CD47 mediated signaling may be beneficial. Some studies have noted platelet derived products such as prostaglandins (PGE2) may be beneficial in sepsis ⁵⁶. This is likely dependent on the concentrations and the local in vivo environment as PGE2 can have differential signaling based on its concentration dependent receptor binding. This again points to the timing and disease context of platelet functions.

We have now discovered a novel mechanism for resting platelet regulation of immune homeostasis. These data represent the first demonstration of sterile, endogenous cell interactions that lead to innate immune programming and may lead to therapeutic strategies to better treat sepsis immune dysfunction.

Novelty and Significance.

What is Known:

• Platelets are not only mediators of thrombosis, they are also part of the immune system.

• Thrombocytopenia is a complication of many diseases, including sepsis, and is a clinical indicator of a worse prognosis.

What new information does this article contribute:

• Thrombocytopenia, independent of the cause, leads to changes in monocyte immune programming in a CD47 dependent manner.

• Resting platelet interactions with monocytes induces glycolysis dependent changes in monocyte histone methylation that changes TLR responses and limits inflammation.

In addition to their well-studied hemostasis functions, platelets function as immune cells. How activated platelets induce and amplify immune responses has received an increasing amount of research attention, and studies have also shown that normal platelet counts are needed to maintain immune and vascular homeostasis in healthy conditions. We discovered that circulating monocytes, but not neutrophils, from thrombocytopenic mice proliferated less ex vivo and produced more inflammatory cytokines in response to toll like receptor (TLR) agonists, and thrombocytopenic mice had more immune dysfunction in sepsis models. We also found that in clinical samples from sepsis patients, a lower platelet count correlated with greater monocyte IL8 and IL6 release. Resting platelets express CD47 that interacts with circulating monocytes to induce monocyte glycolysis and changes in histone methylation, thus genetically modifying monocytes to a more immune quiescent agonist response phenotype. A decline in platelet number therefore independently leads to a loss of this ‘brake’ on immune activation, and increased sepsis immune dysfunction.

Nonstandard Abbreviations and Acronyms

TLR

Toll-like receptor

Th17

T helper 17

ACS

Acute coronary syndrome

EMCO

Extracorporeal membrane oxygenation

TPOR

Thrombopoietin receptor

BM

Bone marrow

DTR

Diphtheria toxin receptor

PF4

Platelet factor 4

ECAR

Extracellular acidification rate

OCR

Oxygen consumption rate

BMDM

Bone marrow derived macrophage

LPS

Lipopolysaccharides

CS

Cecal slurry

PLT

Platelet

RBC

Red blood cell

CBC

Complete blood count

2-DG

2-deoxy-glucose

PGE2

Prostaglandin E2