TREM2 expression is upregulated in monocytes/macrophages and is associated with disease severity in sepsis. Sepsis patients who met the diagnostic criteria for sepsis on the ICU admission day were enrolled in this study. To identify critical regulatory genes in sepsis, RNA-Seq was performed on peripheral blood mononuclear cells (PBMCs) of sepsis patients and healthy controls. Cluster analysis revealed an upregulation of inflammation-related genes in sepsis patients, including genes encoding inflammatory cytokines (Tnfa, Il6, Il1a, Il1b), chemokines (Ccl3, Ccl4, Cxcl1, Cxcl2, Cxcl8), immune receptors such as TREM family receptors (Trem1, Trem2, Treml2, Treml4), Toll-like receptors (Tlr1, Tlr2, Tlr4, Tlr5, Tlr6, Tlr8, Tlr9), and NOD-like receptors (Nlrp3, Nlrc4, Nlrp12), while antiinflammatory factors such as Il4, Treml2, and Foxp3 were downregulated (Figure 1A). As the predominant cell subsets driving inflammation in sepsis, monocytes/macrophages initiate the inflammatory responses via surface or intracellular receptors (2). Among the various receptors, we observed that TREM2, a receptor constitutively expressed on myeloid cells, was markedly upregulated in monocytes of sepsis patients compared with healthy controls (Figure 1B and Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/JCI159400DS1). To validate these observations in vivo, we established a cecal ligation and puncture (CLP) polymicrobial sepsis mouse model and assessed the expression pattern of TREM2. Consistent with the observations from human samples, TREM2 expression in CD11b+F4/80+ macrophages was markedly upregulated in the peritoneal lavage fluids (PLFs), spleen, liver, and lung of septic mice (Figure 1C and Supplemental Figure 2A). Since macrophages in mouse peritoneal lavage are made up of 2 subsets including large peritoneal macrophages (LPMs) (F4/80hiMHC-IIlo) and small peritoneal macrophages (SPMs) (F4/80loMHC-IIhi) (34), we further analyzed TREM2 expression in these 2 subsets. Results showed that TREM2 was predominantly upregulated in LPMs following CLP challenge (Supplemental Figure 2, B and C). Additionally, we established an endotoxemia model via LPS injection and a bacterial sepsis model by Pseudomonas aeruginosa (PA) infection to determine TREM2 expression in macrophages. As expected, TREM2 expression in macrophages was continuously increased in PLF, liver, and lung after LPS injection or PA infection (Supplemental Figure 2D). These findings demonstrated in vivo that TREM2 expression in macrophages was upregulated in sepsis. Overall, we observed an increased expression of TREM2 in monocytes/macrophages in both sepsis patients and mice, suggesting a correlation of TREM2 with sepsis progression.

TREM2 expression is upregulated in monocytes/macrophages and associated with disease severity in sepsis. (A and B) RNA-Seq of healthy controls (n = 10) and sepsis patients (n = 10) was performed. (A) Heatmap of markedly altered genes related to inflammation is shown. (B) PBMCs were isolated from healthy controls (n = 45) and sepsis patients (n = 54), respectively. TREM2 expression on CD11b+CD14+ monocytes was determined by flow cytometry. (C) CLP mouse model was established and TREM2 expression in CD11b+ F4/80+ macrophages at day 1, day 3, and day 5 after infection was assessed in PLFs, spleen, liver, and lung by flow cytometry. (D) Single-cell sequencing data from the lung of CLP sepsis mice were analyzed, and violin plots for the expression of Trem2, Ms4a3, Ms4a7, Spp1, Cd81, and Cd63 in TREM2– and TREM2+ macrophage clusters are shown. (E) The correlations of the percentages of TREM2+ monocytes with CRP, total bilirubin, BUN, and ALT levels were analyzed in sepsis patients (n = 54). (F) PBMCs were collected from sepsis patients (n = 15) on the ICU admission day (day 0) and 1, 3, 5, and 7 days after treatment. TREM2 expression on monocytes was detected and serum CRP levels were displayed. (G) The correlations of the percentages of TREM2+ monocytes with serum glucose and triglyceride concentrations were analyzed in sepsis patients (n = 54). Unpaired Student’s t test was performed (B). One-way ANOVA was employed (C and F). Spearman’s correlation analysis was used (E and G). Data are represented as means ± SEM from 3 independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001.

To investigate the characteristics of TREM2-expressing macrophages in sepsis, we analyzed the transcriptional profiles of TREM2+ and TREM2– macrophages from previously reported single-cell RNA-Seq data on sepsis (35). The analysis showed that TREM2+ macrophages displayed hallmark features of macrophages originated from circulating monocytes, characterized by the high expression of genes Ly6c2, Lyz2, Cd68, Ms4a3, and Ms4a7 (23) (Figure 1D and Supplemental Figure 3A). Further examination of gene modules revealed the high transcriptional expressions of Spp1, Lgals1, Lgals3, Apoe, Cd9, and Cd63, which are markers for LAMs (23), in TREM2+ macrophages (Figure 1D and Supplemental Figure 3A). Moreover, genes involved in phagocytosis (Mrc1, C1qa, C1qb, C1qc), chemotaxis (Ccl2, Ccl7), and inflammatory response (Hmgb1, Hmgb1, Hmgn2) were also highly expressed in TREM2+ subsets (Supplemental Figure 3A). These findings indicate TREM2+ macrophages as a group of LAMs with proinflammatory properties. Consistently, in vivo experiments demonstrated that CD63 but not CD9 was upregulated in TREM2+ macrophages in the spleen, liver, and lung of CLP-induced septic mice (Supplemental Figure 3B). Likewise, CD63 was also largely induced in TREM2+ monocytes of sepsis patients (Supplemental Figure 3C). Furthermore, we found that TREM2+ macrophages exhibited higher lipid uptake and storage abilities compared with TREM2– macrophages (Supplemental Figure 3D). Collectively, these results suggest TREM2+ macrophages as a subset of induced LAMs with proinflammatory properties under the condition of sepsis.

Subsequently, to explore the differential diagnostic potential of TREM2 expression in sepsis, we divided patients into groups based on the pathogen species and analyzed TREM2 expression levels. Nevertheless, TREM2 expression was uniformly upregulated across all sepsis patients with no significant differences observed among groups (Supplemental Figure 4A). To assess the association between TREM2 and sepsis progression, we next analyzed the correlations between TREM2 expression and laboratory diagnostic markers indicative of disease severity of sepsis patients. Notably, positive correlations were observed between TREM2 expression and the inflammatory marker C-reactive protein (CRP), as well as organ damage indicators including total bilirubin, blood urea nitrogen (BUN), and alanine transaminase (ALT) (Figure 1E). Furthermore, we collected a series of blood samples from sepsis patients on the day of ICU admission (day 0, patients are diagnosed as having sepsis and admitted to ICU on the same day) and 1, 3, 5, and 7 days after treatment (days 1, 3, 5, and 7, respectively) to monitor the dynamic changes of TREM2 expression. As expected, TREM2 expression decreased in parallel with the gradual decline of CRP levels (Figure 1F), suggesting a strong association of TREM2 with the severity of sepsis patients.

Since hyperglycemia and impaired FAO are implicated in the pathogenesis of sepsis and contribute to the mortality of sepsis patients (20), we next investigated the correlations of TREM2 with serum glucose and triglyceride levels. Results showed a positive correlation between TREM2 expression and serum triglyceride levels, but not glucose levels, in sepsis patients (Figure 1G), further suggesting a link between TREM2 and lipid metabolism. The cytokine storm mediated by innate immune cells, especially myeloid cells, is a hallmark of sepsis. To determine whether triglyceride or glucose levels are associated with TREM2-mediated cytokine regulation in sepsis, we further analyzed the correlations between these metabolic parameters and inflammatory cytokines produced by TREM2+ monocytes. Unsurprisingly, the levels of IL-1β, TNF-α, and IL-6 produced by TREM2+ monocytes were positively correlated with serum triglyceride concentrations of sepsis patients (Supplemental Figure 4B). However, no significant associations were observed between glucose levels and the amounts of TNF-α, IL-1β, or IL-6 (Supplemental Figure 4C). Besides, no correlation was found between IL-10 produced by TREM2+ monocytes and either triglyceride or glucose levels in sepsis patients (Supplemental Figure 4, B and C). These findings demonstrated that TREM2 expression in monocytes was markedly elevated and was associated with the disease severity of sepsis. Meanwhile, TREM2+ monocytes/macrophages displayed a lipid-associated and inflammatory phenotype in the context of sepsis.

TREM2 knockout in macrophage alleviates sepsis-induced inflammation and organ damage. To investigate the role of TREM2 in sepsis in vivo, we employed WT and TREM2-knockout (TREM2–/–) mice to establish sepsis mouse models and compared the symptoms induced by sepsis. We first compared the survival rates of WT and TREM2–/– mice. Results showed that TREM2 knockout reduced the mortality in the CLP model (Figure 2A). Sepsis is characterized by excessive inflammation, cytokine storm, and organ damage, so we next assessed the levels of inflammation and organ injuries in WT and TREM2–/– mice. In line with the improved survival rates, attenuated lung injuries and reduced lung inflammatory infiltration were observed in TREM2–/– mice, while WT mice showed more alveolar collapse, thickened alveolar walls, and aggravated lung inflammation (Figure 2B). In addition, TREM2 knockout also led to reduced liver and kidney injuries caused by sepsis (Supplemental Figure 5, A and B). To assess the impact of TREM2 on the recruitment of inflammatory cells, we analyzed the percentage of infiltrated inflammatory cells and observed reduced neutrophil and macrophage infiltration in the lung of TREM2–/– sepsis mice (Figure 2C). Furthermore, we measured the levels of proinflammatory cytokines in WT and TREM2–/– mice. Results showed that macrophages from TREM2–/– mice produced lower amounts of IL-6, IL-1β, and TNF-α compared with those from WT mice (Figure 2D). Correspondingly, overall levels of IL-1β, IL-6, and TNF-α in serum, lung, and liver supernatants were decreased after the knockout of TREM2 (Figure 2E). Based on the above results, we demonstrated that TREM2 knockout ameliorated sepsis-induced mortality, inflammation, and organ damage. Finally, we tested serum levels of clinical indexes for human sepsis evaluation in mice to comprehensively determine the in vivo effects of TREM2 during sepsis. As expected, the levels of sepsis-associated indicators including ALT, CRP, BUN, and creatinine (CREA2) were lower in TREM2–/– mice (Supplemental Figure 5C). To further confirm the role of TREM2 in acute inflammation in vivo, we established an LPS endotoxemia model and a PA-induced bacterial sepsis model. Consistent with the observations from the CLP model, knockout of TREM2 reduced the mortality in both LPS and PA models (Supplemental Figure 6, A and B). Moreover, TREM2 deficiency led to a decrease of serum IL-6 levels in a dose- and time-dependent manner following LPS treatment in vivo (Supplemental Figure 6C). In addition, IL-1β, TNF-α, and IL-6 levels were also downregulated in TREM2–/– mice after the stimulation of TLR3 ligand poly(I:C) (Supplemental Figure 6D). These findings revealed the proinflammatory role of TREM2 in acute inflammation induced by TLR ligation or bacterial infection.

TREM2 knockout in macrophages alleviates sepsis-induced inflammation and organ damage. (A–E) CLP sepsis mouse model was established in WT and TREM2–/– mice. (A) Survival rates were observed. (B) Lung injuries and inflammatory cell infiltration were evaluated by H&E staining 24 hours later. (C) Lung neutrophil (Neu) and macrophage (Mφ) proportions were examined by flow cytometry 24 and 72 hours later. (D) Levels of IL-6, IL-1β, and TNF-α produced by CD11b+F4/80+ macrophages in liver and lung were determined by flow cytometry 12 hours after CLP. (E) IL-1β, IL-6, and TNF-α levels in serum and lung or liver suspension were detected by ELISA at 24 hours after challenge. (F–J) CLP sepsis mouse model was established in Lyz2Cre and TREM2fl/fl Lyz2Cre mice. (F) Survival rates were recorded. (G) Structural damage of lung tissue was evaluated by H&E staining 24 hours later. (H) The percentages of neutrophils and macrophages in lung were determined by flow cytometry 24 and 72 hours later. (I) Levels of macrophage-derived IL-6, IL-1β, and TNF-α in liver and lung were detected by flow cytometry 12 hours after CLP. (J) IL-1β, IL-6, and TNF-α levels in serum, lung, and liver supernatant were detected by ELISA 24 hours later. Log rank (Mantel-Cox) test was adopted to compare significance (A and F). One-way ANOVA was employed (B–E and G–J). Data are represented as means ± SEM from at least 3 independent experiments. Scale bars: 50 μm. *P < 0.05; **P < 0.01.

Since the elevated expression of TREM2 was observed in monocytes/macrophages during sepsis (Figure 1), we next generated TREM2 conditional knockout mice (TREM2fl/flLyz2Cre), in which TREM2 was specifically deleted in macrophages, to explore whether TREM2 exerted functions in sepsis via macrophages. Results showed that TREM2fl/flLyz2Cre mice displayed lower mortality compared with TREM2fl/fl mice after CLP challenge (Figure 2F). In the meanwhile, reduced lung structural damage (Figure 2G) and less infiltration of macrophages and neutrophils were observed in TREM2fl/flLyz2Cre mice (Figure 2H). Furthermore, the specific deficiency of TREM2 in macrophages decreased the production of IL-6, IL-1β, and TNF-α in the lung of sepsis mice (Figure 2I). Similarly, lower levels of IL-1β, IL-6, and TNF-α in serum, lung, and liver were observed in TREM2fl/flLyz2Cre mice (Figure 2J). In addition, liver and kidney damage as well as sepsis severity indicators ALT, AST, BUN, and CREA2 were reduced in TREM2fl/flLyz2Cre mice (Supplemental Figure 7, A–C).

To further determine whether TREM2 directly influenced the outcome of sepsis, we transferred sorted TREM2+ and TREM2– monocytes from CD45.1 mice into CD45.2 recipient mice, followed by CLP challenge (Supplemental Figure 8A). We first assessed the stability of TREM2 expression in monocytes after transfer and found that approximately 99% of CD45.1+ monocytes maintained the TREM2+ phenotype at 24 hours after CLP challenge (Supplemental Figure 8, B and C). Meanwhile, about 26% of TREM2– CD45.1+ monocytes converted to TREM2+ monocytes following sepsis induction (Supplemental Figure 8C), indicating that sepsis induced TREM2 expression in monocytes. Furthermore, the transfer of TREM2+ monocytes accelerated the mortality of sepsis mice compared with TREM2– monocytes, further confirming the proinflammatory role of TREM2 in sepsis (Supplemental Figure 8D).

Since effective bacterial clearance is crucial to preventing sepsis, we then explored the role of TREM2 in bacterial clearance. We sorted TREM2+ versus TREM2– macrophages from sepsis mice and found that TREM2+ macrophages displayed an impaired bacterial killing activity compared with TREM2– macrophages after PA infection (Supplemental Figure 9A). Consistently, TREM2 knockout reduced the intracellular bacterial burden of PA (Supplemental Figure 9B). Furthermore, in vivo results showed that the bacterial counts were markedly decreased in the lung and spleen of TREM2fl/flLyz2Cre mice after PA infection (Supplemental Figure 9C). These data indicated that TREM2 suppressed bacterial clearance of macrophages in PA-induced bacterial sepsis. Collectively, we investigated the in vivo role of TREM2 in sepsis and demonstrated that TREM2 deficiency protected mice from sepsis.

TREM2 deficiency promotes FAO of macrophage in sepsis. FAO is a critical metabolic process regulating inflammation during sepsis, and impaired FAO has been considered as a contributor to sepsis-associated organ damage and mortality (36). During the analysis of RNA-Seq data, we observed an increase in the expression of genes encoding ATP-binding cassette transporters (Abca1, Abca2, Abca7, Abcd1, Abcg1) and lipid-associated receptors (Cd63, Ldlr, Vldlr), as well as disturbed FA metabolism in sepsis patients compared with healthy controls (Figure 3, A and B). Notably, genes involved in the FAO process, including peroxisome proliferator-activated receptor (Ppargc1a, Ppara) and rate-limiting enzyme (Cpt1c), were markedly downregulated (Figure 3A). We found that TREM2 expression in monocytes was positively correlated with triglyceride concentration in sepsis patients (Figure 1G). Consistently, we further observed that TREM2 expression was upregulated, while FAO rate-limiting enzyme carnitine palmitoyl transferase I (CPTI) and the regulator PCC-1α were downregulated in the monocytes of sepsis patients (Figure 3C). To explore the connection of TREM2 with FAO in sepsis, we isolated monocytes from the peripheral blood of healthy controls and sepsis patients and treated monocytes with recombinant TREM2-Fc protein to block TREM2 signaling, followed by the detection of FAO-related regulators. In the process of FAO, CD36 acts as an internalization receptor for FA uptake. CPTI is the rate-limiting enzyme of FAO and is responsible for the transport of long-chain FAs into mitochondria, while CPTII is in charge of the disassociation of l-carnitine and the release of FAs (10). Following treatment with TREM2-Fc protein, the expressions of CD36, CPTI, and CPTII in monocytes were increased in sepsis patients but not in healthy controls (Figure 3, C–E), indicating an enhancement of FAO after TREM2 blockade during sepsis.

TREM2 deficiency promotes FAO of macrophages in sepsis. (A and B) RNA-Seq of healthy controls (n = 10) and sepsis patients (n = 10) was performed. (A) Heatmap of altered genes involved in the FA metabolism was shown. (B) Flowchart of FA metabolism is displayed. Upregulated genes in sepsis patients are marked as red and downregulated genes are marked as green. (C) Monocytes were isolated from healthy controls and sepsis patients. Western blot was performed to detect the expression of TREM2, CPTI, PGC-1α, and β-actin. (D) PBMCs were isolated from sepsis patients or healthy controls and treated with recombinant TREM2-Fc protein (4 μg/mL) and IgG-Fc for 12 hours, followed by the detection of the expression levels of CD36, CPTI, and CPTII in CD11b+CD14+ monocytes by flow cytometry. (E–H) CLP mouse model was established. (E and F) Twelve hours later, serum triglyceride levels in WT versus TREM2–/– mice (E) or Lyz2Cre versus TREM2fl/fl Lyz2Cre mice (F) were detected. (G and H) Lipid droplets in the liver of WT versus TREM2–/– mice (G) or Lyz2Cre versus TREM2fl/fl Lyz2Cre mice (H) were assessed by oil red O staining 24 hours later. (I and J) CLP mouse model was established in WT and TREM2–/– mice. Peritoneal (I) or splenic (J) macrophages were isolated and the rates of FAO were determined by Seahorse XF Extracellular Flux Analyzers. (K) BMDMs were isolated and stimulated with LPS (1μg/ml) for 12 hours. Then the rate of FAO was determined. Paired Student’s t test was performed (D). Unpaired Student’s t test was used in C. One-way ANOVA was employed (E–H). Two-way ANOVA was used to analyze significance (I–K). Data are represented as means ± SEM from at least 3 independent experiments. Scale bars: 50 μm. *P < 0.05; **P < 0.01; ***P < 0.001.

To investigate the impact of TREM2 on macrophage FAO in vivo, we established a CLP mouse model with WT and TREM2-deficient mice, and detected triglyceride levels in serum and liver at first. Consistent with the positive correlation between monocyte TREM2 expression and serum triglyceride concentration in sepsis patients, both systematic and macrophage-conditional knockout of TREM2 resulted in decreased serum triglyceride levels in sepsis mice (Figure 3, E and F). Meanwhile, lipid accumulation in the liver of TREM2–/– and TREM2fl/flLyz2Cre mice was also reduced, as indicated by less lipid droplets stained as red (Figure 3, G and H). In addition, in vitro assay also showed that FA uptake and lipid droplets were reduced in TREM2–/– macrophages (Supplemental Figure 10, A and B). We further examined the expressions of rate-limiting enzyme CPTI and related molecules PPARα and PPARγ as well as its cofactors PGC-1α and PGC-1β to determine FAO levels in WT and TREM2–/– sepsis mice. As expected, elevated expressions of CPTI, PPARα, PPARγ, PGC-1α, and PGC-1β were observed in the liver and lung of TREM2–/– mice (Supplemental Figure 10C), indicating the increased FAO rates after TREM2 deficiency. We also assessed the glycolysis levels in liver and lung by measuring the expression of glycolysis rate-limiting enzymes hexokinase 2 (HK2) and pyruvate kinase M2 (PKM2), but no differences were found between WT and TREM2–/– mice (Supplemental Figure 10D). To further elucidate the effect of TREM2 on macrophage FAO, we isolated peritoneal macrophages (pMφ) and splenic macrophages from sepsis mice to evaluate their FAO rates ex vivo. Results showed that TREM2-deficient macrophages exhibited enhanced FAO rates compared with WT macrophages (Figure 3, I and J). Besides, glycolysis rates of WT and TREM2–/– macrophages were assessed and no differences were observed (Supplemental Figure 11A). Moreover, we isolated bone marrow–derived macrophages (BMDMs) from WT and TREM2–/– mice for in vitro explorations. As expected, FAO rates were increased in TREM2–/– BMDMs after LPS stimulation (Figure 3K), while limited differences in glycolysis were observed (Supplemental Figure 11B). These results indicated that TREM2 inhibited macrophage FAO and TREM2 knockout alleviated impaired FAO in sepsis mice.

Inhibition of FAO abolished the improved sepsis symptoms induced by TREM2 deficiency. Since FAO is impaired in sepsis and TREM2 deficiency could alleviate sepsis and improve macrophage FAO, we next explored whether TREM2 regulated sepsis-induced inflammation and organ damage through affecting FAO. We generated conditional CPTIfl/fl Lyz2Cre mice, in which CPTI is specifically deleted in macrophages, by crossing CPTIfl/fl mice with Lyz2Cre mice. We then established a CLP sepsis mouse model with CPTIfl/fl and CPTIfl/fl Lyz2Cre mice following the treatment of TREM2 blocking Ab or control IgG Ab. Results showed that mice receiving TREM2 Ab had lower mortality than mice treated with IgG control after CLP challenge (Figure 4A). However, when CPTI was knocked out in macrophages, the survival rate of TREM2-blocked mice dropped to a level similar to that of CPTIfl/fl mice receiving IgG control (Figure 4A). Moreover, CPTI knockout in macrophages also markedly increased the levels of proinflammatory cytokines and indicators for organ injury, counteracting the effects of TREM2 blockade in CLP sepsis mice (Figure 4, B and C). Subsequently, we generated CPTIfl/fl TREM2fl/fl Lyz2Cre double-knockout mice, in which both CPTI and TREM2 are specifically deficient in macrophages, to further investigate the effects of TREM2 and CPTI on sepsis. Consistent with the findings in TREM2 blocking Ab-treated mice, further knockout of CPTI in macrophages abolished the improved survival rate due to TREM2 deficiency (Figure 4D). Meanwhile, CPTI knockout exacerbated lung, liver, and kidney injuries, which were ameliorated in TREM2fl/fl Lyz2Cre mice (Figure 4E and Supplemental Figure 12). In addition, we observed more lipid droplets in the liver of CPTIfl/fl TREM2fl/fl Lyz2Cre mice than in TREM2fl/fl Lyz2Cre mice (Figure 4F). Furthermore, levels of proinflammatory cytokines and organ injury indicators were also elevated in CPTIfl/fl TREM2fl/fl Lyz2Cre mice compared with TREM2fl/fl Lyz2Cre mice (Figure 4, G and H). These results indicated that TREM2 deficiency alleviated sepsis through enhancing FAO.

Inhibition of FAO abolishes the improved sepsis symptoms regulated by TREM2. (A–C) CPTIfl/fl and CPTIfl/flLyz2Cre mice were treated with anti-TREM2 blocking Ab (150mg/kg) or IgG isotype control for 2 hours, followed by the establishment of CLP model. (A) The survival rates were observed. (B) IL-1β, IL-6, and TNF-α levels in serum were determined by ELISA at 24 hours after CLP challenge. (C) The serum biochemical indexes including ALT, CREA2, and BUN were detected 24 hours later. (D–H) CLP model was established in TREM2fl/flLyzCre, CPTIfl/flLyz2Cre, CPTIfl/flTREM2fl/flLyz2Cre, and LyzCre control mice respectively. (D) The survival rates were observed. (E) H&E staining was performed to assess the lung injuries and inflammatory cell infiltration 24 hours later. (F) Lipid droplets in liver were assessed by oil red O staining 24 hours later. (G) Serum IL-1β, IL-6, and TNF-α levels were detected by ELISA 24 hours later. (H) ALT, BUN, and CREA2 concentrations in serum were detected 24 hours later. Log rank (Mantel-Cox) test was adopted to compare significance (A and D). One-way ANOVA was employed (B, C and E–H). Data are represented as means ± SEM from at least 3 independent experiments. Scale bars: 50 μm. *P < 0.05; **P < 0.01; ***P < 0.001.

Since TREM2 absence increased the resistance to sepsis via restoring FAO, we next investigated whether there were synergistic effects between TREM2 blockade and l-carnitine supplementation, which can help with the transport of FAs into mitochondria to fuel FAO and has been reported to be advantageous to reducing mortality in sepsis (18, 19). Surprisingly, we found that TREM2 blockade markedly improved the survival rate of sepsis mice, and l-carnitine administration did not further increase the survival rate of TREM2 Ab–treated mice (Supplemental Figure 13A). Meanwhile, both TREM2 blocking Ab and l-carnitine supplementation displayed protective effects on lung, liver, and renal damage, but the combination failed to show better effects (Supplemental Figure 13, B–D). Moreover, levels of IL-6, TNF-α, and IL-1β were not further reduced after l-carnitine supplementation on the basis of TREM2 blockade (Supplemental Figure 13E). Similar results were observed in serum levels of ALT, CRP, BUN, and CREA2 (Supplemental Figure 13F). These findings demonstrated that TREM2 blockade had a comparably beneficial effect with l-carnitine supplementation, which may provide support for developing sepsis treatment strategies.

TREM2 regulates macrophage FAO through BTK kinase. Next, to elucidate the mechanism underlying TREM2-mediated FAO regulation, we isolated WT and TREM2–/– pMφ and investigated the involved signaling pathways. We assessed the levels of rate-limiting enzyme CPTI in WT and TREM2–/– pMφ at first. Following LPS stimulation, CPTI expression was decreased, while PKM2 and HK2 were upregulated in macrophages (Figure 5A), which is consistent with previous reports (12, 37). Notably, TREM2 knockout increased the expression of CPTI, but had no effect on HK2 and PKM2 expression (Figure 5A), in line with in vivo data. In addition, we also observed elevated expression of FAO-related molecules, including PGC-1α, PGC-1β, and PPARα, in TREM2–/– pMφ (Figure 5B). These results indicated that FAO was enhanced in LPS-stimulated macrophages after TREM2 deficiency. We then explored the effects of TREM2 on FAO-related signaling pathways. It is known that adenosine monophosphate activated protein kinase (AMPK) signal and signal transducers and activators of transcription 6 (STAT6) are crucial for the FAO process (38, 39). Unsurprisingly, increased phosphorylation levels of AMPK and STAT6 were detected in TREM2–/– pMφ after LPS treatment (Figure 5, C and D). As a critical kinase regulating signal transmission of TREM family members in myeloid cells (40, 41), BTK participates in the regulation of lipid uptake, lipid accumulation, and oxidative stress (42–44). To explore whether BTK was involved in TREM2-mediated FAO regulation, we examined the phosphorylation of BTK and found an increase of BTK phosphorylation in TREM2–/– pMφ (Figure 5D). Furthermore, TREM2-Fc treatment increased the phosphorylation of BTK in sorted monocytes from sepsis patients but not healthy donors (Supplemental Figure 14A). To further assess the involvement of BTK in TREM2-regulated FAO pathways, we inhibited BTK activity with small molecular inhibitors and evaluated FAO changes in WT and TREM2–/– pMφ. As expected, increased expressions of CPTI and PGC1α, as well as phosphorylation of AMPK and STAT6, were substantially suppressed in TREM2–/– pMφ after the use of BTK inhibitors LFM-A13 and ibrutinib (Figure 5E). Meanwhile, LFM-A13 and ibrutinib treatment also markedly reduced the elevated FAO rate in TREM2–/– pMφ (Figure 5F). In addition, BMDMs were employed to confirm the regulatory effect of BTK on TREM2 signaling. Similarly, the enhanced FAO rate in TREM2–/– BMDMs was suppressed by LFM-A13 and ibrutinib (Figure 5G). Moreover, ibrutinib treatment also abolished the upregulated expression of CPTI induced by TREM2-Fc in monocytes from sepsis patients (Supplemental Figure 14B). These findings indicated that BTK was engaged in TREM2-mediated FAO modulation. To further confirm the role of BTK in TREM2-mediated inflammation, we detected proinflammatory cytokine levels following BTK inhibition in WT and TREM2–/– pMφ in vitro. Consistently, results showed that LFM-A13 and ibrutinib treatment abolished the differences of IL-1β and IL-6 in WT and TREM2–/– pMφ (Figure 5H).

TREM2 regulates macrophage FAO through BTK kinase. (A and B) WT and TREM2–/– pMφ was stimulated with LPS (1 μg/ml) for indicated times. (A) The expressions of CPTI, HK2, and PKM2 were detected by Western blot. (B) The expressions of FAO-related regulators were determined by Western blot. (C) Phosphorylation and total levels of AMPKα were determined. (D) Phosphorylation and total levels of BTK and STAT6 were compared among groups by Western blot. (E and F) WT and TREM2–/– pMφ cells were treated with BTK inhibitor LFM-A13 (1 μM) or ibrutinib (1 μM) for 1 hour, followed by stimulation with LPS (1 μg/ml) for 12 hours. (E) The expressions of FAO rate-limiting enzyme CPTI and associated molecules PGC-1, as well as the phosphorylation and total levels of AMPKα and STAT6, were measured. (F) The FAO rate was determined. (G) WT and TREM2–/– BMDM cells were treated with LFM-A13 (1 μM) or ibrutinib (1 μM) for 1 hour. Then LPS (1 μg/ml) was added for additional stimulation for 12 hours. The FAO rate was tested by Seahorse XF Extracellular Flux Analyzers. (H) WT and TREM2–/– pMφ cells were pretreated with LFM-A13 or ibrutinib for 1 hour and stimulated with LPS for 12 hours. Relative mRNA expressions of IL-1β and IL-6 were detected by quantitative real-time PCR. Two-way ANOVA was used to analyze significance (F and G). One-way ANOVA was employed (H). Data are represented as means ± SEM from at least 3 independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001.

To further clarify the influence of BTK on TREM2-mediated effects in vivo, we treated Lyz2Cre and TREM2fl/fl Lyz2Cre mice with ibrutinib and subsequently established a CLP sepsis model to investigate the potential role of BTK. After the use of ibrutinib, the survival rate of TREM2fl/fl Lyz2Cre mice rapidly decreased to a level similar to that of Lyz2Cre mice in the vehicle group (Supplemental Figure 15A). Meanwhile, alleviated lung injury observed in TREM2fl/fl Lyz2Cre mice disappeared following ibrutinib treatment (Supplemental Figure 15B). In addition, BTK inhibition with ibrutinib also substantially abolished the decreased levels of proinflammatory cytokines caused by TREM2 deficiency (Supplemental Figure 15C). Furthermore, ALT, CRP, BUN, and CREA2 levels in TREM2fl/fl Lyz2Cre mice were substantially increased following ibrutinib treatment (Supplemental Figure 15D). These results suggest that TREM2 exerted function in sepsis through BTK kinase. Collectively, these observations demonstrate that TREM2 inhibited macrophage FAO to regulate inflammation via BTK kinase.

TREM2 inhibits BTK-mediated FAO via recruiting SHP1. We demonstrated that TREM2 inhibited FAO through suppressing the phosphorylation of BTK. To explore whether TREM2 directly interacted with BTK or other FAO regulators, we conducted coimmunoprecipitation (Co-IP) experiments between TREM2 and BTK, PGC-1α, or PGC-1β. However, no interactions were observed between TREM2 and these regulators (Supplemental Figure 16, A and B). Consequently, we investigated the mechanism by which TREM2 suppressed BTK phosphorylation. Tyrosine phosphatases (PTPs) are a class of enzymes that exist in various immune cells and function as negative regulators of signal transduction by inhibiting the phosphorylation of kinases. Among those, SHP1, SHP2, and SHIP1 are the most common PTPs in myeloid cells and exert inhibitory effects on diverse signaling pathways (45). In particular, BTK has been reported to be a substrate of SHP1 and SHIP1 (46, 47). To explore whether TREM2 recruited these PTPs, we performed co-IP assay to test the interactions of TREM2 with SHP1, SHP2, and SHIP1. Results showed that TREM2 specifically interacted with SHP1, but not SHP2 or SHIP1 (Figure 6A). Furthermore, endogenous co-IP confirmed the binding of TREM2 with SHP1 in pMφ cells (Figure 6B). To determine whether TREM2 inhibited BTK phosphorylation through SHP1, we first examined the phosphorylation of SHP1, a basic step during its activation, in LPS-challenged WT and TREM2–/– pMφ. Results showed that SHP1 phosphorylation was decreased in TREM2-deficient pMφ (Figure 6C), indicating that TREM2 may recruit SHP1 to inhibit the phosphorylation of BTK. To verify this hypothesis, we overexpressed TREM2 in primary BMDMs and treated cells with SHP1 inhibitors to detect BTK phosphorylation. As expected, TREM2 overexpression reduced the phosphorylation level of BTK, and the use of PTP inhibitor and NSC87877 markedly restored BTK phosphorylation (Figure 6D), indicating the involvement of SHP1 in TREM2-BTK signal transduction.

TREM2 inhibits BTK-mediated FAO via recruiting SHP1. (A) Constructed plasmids were transfected into 293T cells. The interactions of TREM2 with SHP1, SHP2, and SHIP1 were determined by co-IP and Western blot 48 hours later. (B) PMφ cells were stimulated with LPS (1 μg/ml) for 12 hours and immunoprecipitated with IgG or TREM2 Ab to determine the binding between TREM2 and SHP1. (C) WT and TREM2–/– PMφ cells were stimulated with LPS (1 μg/ml) for indicated times. The phosphorylation and total levels of SHP1 were determined by Western blot. (D) TREM2 plasmid was transfected into BMDMs. Forty-eight hours later, BMDMs were pretreated with PTP inhibitor II (1 μM) or NSC87877 (1 μM) for 1 hour, followed by the treatment of LPS (1 μg/ml). BTK phosphorylation was assessed 12 hours later. (E) TREM2, SHP1, and DAP12 plasmids were transfected into 293T cells. Forty-eight hours later, co-IP assay was performed to determine the interaction between TREM2 and SHP1. (F) WT and DAP12-deficient (DAP12–/–) pMφ cells were treated with LPS (1μg/ml) for 12 hours and immunoprecipitated with IgG or TREM2 Abs. The binding among TREM2, DAP12, and SHP1 was detected by Western blot. (G) Plasmids expressing TREM2 lacking extracellular domain (ΔExtra) or transmembrane plus cytoplasmic domain (ΔTrans-cyto) and expressing SHP1 were transfected into 293T cells. Forty-eight hours after transfection, co-IP was performed. (H) TREM2 plasmid was transfected into 293T cells with full-length SHP1, N terminal-SH2 domain (N-SH), C-terminal SH2 (C-SH), or PTPase domain of SHP1, respectively, and the interactions of TREM2 with these domains were determined by co-IP after 48 hours. (I) PTPase domain or PTPase domain containing R352A, K356A, R358A, N359A, Y536A, or Y564A mutations were transfected into 293T cells with TREM2 plasmid and the interactions were determined 48 hours after transfection.

Since DNAX activating protein of 12 kD (DAP12) is a well-known adaptor by which TREM2 transmits signals, we investigated the role of DAP12 in the interaction between TREM2 and SHP1. Surprisingly, the binding of TREM2 with SHP1 in 293T cells was detected regardless of the presence of DAP12 (Figure 6E). Furthermore, even in DAP12-deficient pMφ, the interaction between TREM2 and SHP1 was still observed (Figure 6F), suggesting that TREM2 may bind to SHP1 in a DAP12-independent manner. As a membrane receptor, TREM2 contains an Ig-like extracellular domain, a transmembrane domain, and a short cytoplasmic tail (48). To explore which domain of TREM2 was responsible for the interaction with SHP1, we constructed plasmids expressing TREM2 proteins lacking either the extracellular domain (ΔExtra) or transmembrane plus cytoplasmic domains (ΔTrans-cyto) and detected their binding to SHP1. Results showed that the interaction between TREM2 and SHP1 disappeared when the transmembrane and cytoplasmic domains were absent, suggesting that TREM2 bound to SHP1 through the transmembrane and cytoplasmic domains (Figure 6G). Consistent with this, structural analysis indicated that 2 negatively charged regions within the intracellular tail of TREM2 were capable of binding to 2 positively charged regions of SHP1 (Supplemental Figure 17). Furthermore, we investigated the domain of SHP1 involved in the interaction with TREM2. SHP1 contains 3 domains, including N-SH2, C-SH2, and PTPase domain (49). We constructed plasmids expressing each of these domains and evaluated their binding capacity with TREM2. Results showed that the PTPase domain but not SH2 domains directly bound to TREM2 (Figure 6H). In addition, we explored the key amino acid residues required for the interaction between TREM2 and the PTPase domain. Several amino acid sites, including 352 Arg, 356 Lys, 358 Arg, 359 Asn, 536 Tyr, and 564 Tyr, were selected for mutation based on the structural analysis of the PTPase domain of SHP1 (50–52). Results showed that the mutations of 352 Arg to Ala (R352A) and 359 Asn to Ala (N359A) substantially abolished the binding between TREM2 and the PTPase domain (Figure 6I). Overall, these findings demonstrated that TREM2 recruited the PTPase domain of SHP1 dependent on the 352 Arg and 359 Asn residues within SHP1, thereby inhibiting the phosphorylation of BTK kinase. Collectively, these results indicated that TREM2 inhibited BTK-mediated FAO via recruiting SHP1 to suppress BTK phosphorylation.