The lipid-sensor TREM2 aggravates disease in a model of LCMV-induced hepatitis

Lipid metabolism is increasingly being appreciated to affect immunoregulation, inflammation and pathology. In this study we found that mice infected with lymphocytic choriomeningitis virus (LCMV) exhibit global perturbations of circulating serum lipids. Mice lacking the lipid-sensing surface receptor triggering receptor expressed on myeloid cells 2 (Trem2 −/−) were protected from LCMV-induced hepatitis and showed improved virus control despite comparable virus-specific T cell responses. Non-hematopoietic expression of TREM2 was found to be responsible for aggravated hepatitis, indicating a novel role for TREM2 in the non-myeloid compartment. These results suggest a link between virus-perturbed lipids and TREM2 that modulates liver pathogenesis upon viral infection. Targeted interventions of this immunoregulatory axis may ameliorate tissue pathology in hepatitis.

TREM2 deficiency in macrophages, dendritic cells and microglia leads to increased production of pro-inflammatory cytokines upon stimulation with toll-like receptor ligands [49][50][51][52] . Likewise, Trem2 −/− bone marrow-derived macrophages and peritoneal macrophages produce increased levels of TNFα and IL-6 upon bacterial stimulation 43,53 . Yet, there is evidence that TREM2 deficiencies can also attenuate inflammatory responses in macrophages and dendritic cells 43,54,55 as well as in a brain injury model 56 , indicating that the function of TREM2 may vary in a context-dependent manner. Loss-of-function mutations in the human TREM2 gene have been linked to the fatal Nasu-Hakola disease, which is associated with bone fractures, progressive dementia and epileptic seizures 57,58 . Recent genome-wide association studies identified the missense mutation R47H in TREM2 to confer an increased risk for Alzheimer's disease (AD) 38,59,60 . These insights instigated increased interest in the biology of TREM2 with a primary focus on the central nervous system and bone tissue.
In this study we employed the model of lymphocytic choriomeningitis virus (LCMV), which represents a benchmark model of acute and chronic viral infection and results in hepatitis as a consequence of immunopathological mechanisms [61][62][63][64] . We found that infection with LCMV perturbs the composition of circulating lipid species in wild type (WT) mice, which prompted us to investigate the role of the recently described lipid-sensing receptor TREM2. By using Trem2 −/− mice we could identify a novel disease-aggravating role of non-hematopoietically-expressed TREM2 in virus-induced liver pathology.

Viral infection perturbs the serum composition of metabolites.
To investigate the changes in serum metabolites after virus infection, we infected WT mice with LCMV strain WE 61,63 . Infected mice exhibited elevated serum concentrations of the clinical hallmark parameters of hepatitis, alanine aminotransferase (ALT) and aspartate aminotransferase (AST) (Fig. 1A) and elevated serum concentrations of cholesterol (Fig. 1B). Next, we wanted to assess systemic changes of serum metabolites upon viral infection by a targeted metabolomics approach. We infected WT mice with LCMV and collected sera before and during the peak of hepatitis. Quantification of 188 metabolites was performed by mass-spectrometry unveiling an enrichment of altered glycerophospholipid and sphingolipid species in the serum of infected mice on day 8 post infection (Fig. 1C, Table S1). Other metabolites such as acylcarnitines, amino acids and biogenic amines showed less drastic changes (Fig. 1C, Table S1). Of the quantified 91 different glycerophospholipids and 15 sphingolipids, we observed a complex pattern of up-and downregulated individual lipid species at both two and eight days post infection (Fig. 1D,  Fig. S1A). These lipids are a class of metabolites with intricate regulatory mechanisms and multiple known roles including the involvement in hepatocellular death and liver diseases [65][66][67] . Together, this data shows that LCMV infection perturbs the global serum composition of lipids including the upregulation of numerous species of glycerophospholipids and sphingolipids.
Lack of TREM2 ameliorates liver damage upon viral infection. TREM2 was recently revealed to be a lipid-sensing receptor that binds glycerophospholipids and sphingomyelin 38 . Indeed, addition of lipids extracted from sera of infected mice tended to increase the activity of Ppar-δ in WT but not in Trem2 −/− cells (Fig. S1B) 43 . Similarly, individual lipid species, which were significantly upregulated in the sera of mice infected eight days previously, led to increased reporter activation in a TREM2-dependent manner (Fig. S1C).
After observing differential regulation of systemic lipid species after LCMV infection and their effects on a TREM2-dependent reporter system, we hypothesized a potential role for TREM2 in LCMV hepatitis. We infected Trem2 −/− and WT mice and observed that mice lacking TREM2 lost less body weight compared to infected WT mice ( Fig. 2A). This was associated with reduced serum concentrations of ALT and AST in infected Trem2 −/− compared to WT mice (Fig. 2B,C). Moreover, Trem2 −/− mice exhibited lower serum concentrations of two alternative parameters for liver damage, alkaline phosphatase (AP) and bilirubin, compared to WT mice upon viral infection (Fig. 2D,E). In line with these findings, a histological assessment of liver damage revealed a lower pathological score (Fig. 2F) and fewer cleaved Caspase 3 + cells (Fig. 2G) in infected Trem2 −/− mice compared to WT mice. Yet, serum concentrations of the kidney-related parameters blood urea nitrogen and creatinine were comparable between infected WT and Trem2 −/− mice (Fig. S2A,B), suggesting that the differences in tissue damage were not generalized to all organs. We also performed a metabolic profiling of Trem2 −/− mice on day 8 post infection and found significant changes in the lipid profiles compared to WT mice, likely reflecting the extent of pathology (Table S1). In summary, the lack of TREM2 resulted in an unexpected amelioration of LCMV-induced liver damage.

Trem2 −/− mice show improved virus clearance. To investigate viral propagation in WT and
Trem2 −/− mice, we determined the infectious viral loads by focus-forming assay ( Fig. 3A-C). Both genotypes had comparable viral loads in the serum until day six (Fig. 3A). At later time points, however, Trem2 −/− mice exhibited reduced viral loads in serum (Fig. 3A), liver (Fig. 3B) and spleen (Fig. 3C) compared to WT mice. These findings were corroborated by immunohistochemical staining of liver sections, whereby reduced amounts of the viral nucleoprotein (NP) in Trem2 −/− mice compared to WT mice were detected at 8 days post infection (Fig. 3D). Hepatocytes and Kupffer cells are the most abundant cell types in the liver and represent the primary target of LCMV in this organ 61 . We thus quantified the signal of viral NP in these cell populations and detected reduced numbers of NP + hepatocytes and NP + Kupffer cells in Trem2 −/− liver sections compared to WT mice (Fig. 3E,F). The lack of the cellular receptor TREM2 per se may affect viral entry or replication. We therefore infected primary bone marrow-derived macrophages (BMDMs) from both genotypes but found similar viral growth curves in WT and Trem2 −/− BMDMs (Fig. S3). In summary, Trem2 −/− mice show improved viral clearance compared to WT mice.
WT and Trem2 −/− mice mount comparable T cell responses. Virus control in the LCMV model is dependent on T cells and the damage observed in LCMV-induced hepatitis is largely attributed to CD8 T cells 63 . To assess T cell responses in our model, we infected WT and Trem2 −/− mice with LCMV and enumerated CD8 + effector memory (EM) and CD4 + EM T cells as well as viral antigen-specific CD8 and CD4 T cell responses from spleen and liver tissue by intracellular cytokine staining using flow cytometry (Fig. 4, Fig. S4). Infected WT and Trem2 −/− mice had comparable percentage of EM T cells (Fig. S4), and indistinguishable CD8 T cell numbers of IFNγ single producers and IFNγ TNFα double producers upon peptide stimulation for the viral epitopes GP 33-41 , NP 396-404 and GP 276-286 in both liver and spleen (Fig. 4A,B). This experiment indicated that WT and Trem2 −/− mice mount similar CD8 T cell responses upon viral infection. Likewise, we assessed the cytokine , serum from uninfected mice and mice that were infected with LCMV 8 days previously was used (n = 4 mice per group). (C) Fold enrichment of the metabolite classes (glycerophospolipids, sphingolipids, biogenic amines, amino acids, acylcarnitines) was calculated using the relative proportion (number of significant metabolites from a specific metabolite class divided by the number of all significant metabolites measured) normalized to the number of measured metabolites of a specific metabolite class divided by the number of all measured metabolites. Members of the acylcarnitine class did not pass the threshold of fold-change >1.5 and p-value < 0.05. (D) Bubble plot of individual glycerophospholipids (circle) and sphingolipids (triangle) depicting fold-change (x-axis), p-value (y-axis) and the absolute change (size of symbols). Statistical significance of changed individual metabolites in sera of infected vs. uninfected mice was calculated with unpaired t-test (C,D). Red symbols in (D) reflect metabolites with a fold-change >1.5 and a p-value < 0.05.
Together, we found that LCMV-infected WT and Trem2 −/− mice mount comparable virus-specific CD8 and CD4 T cell responses in spleen and liver, suggesting that the observed differences in liver pathology and virus clearance are unlikely to be driven by T cell proliferation, migration, cytokine production or cytotoxicity.
We also profiled γ/δ T cells, NK cells, NKT cells, neutrophils and macrophages in spleen and liver of WT and Trem2 −/− mice upon infection by flow cytometry (Fig. S6). This analysis revealed largely comparable numbers of cell populations albeit we noticed an increase of NK and NKT cells in Trem2 −/− mice compared to WT mice on day 4 post infection ( Fig. S6B,C,G,H).

Comparable cytokine responses in WT and Trem2 −/− mice upon LCMV infection. In order to
investigate the induction of cytokines before the peak of LCMV-induced hepatitis, we quantified protein concentrations of chemokine (C-X-C motif) ligand 1 (CXCL1), interleukin 6 (IL-6) and interferon alpha (IFNα) in sera of WT and Trem2 −/− mice by ELISA (Fig. S7). This revealed comparable levels of IL-6 and CXCL1 in WT and Trem2 −/− . Yet, Trem2 −/− mice exhibited reduced levels of IFNα compared to WT mice upon LCMV infection.

Non-hematopoietic TREM2 aggravates LCMV induced hepatitis.
To determine the cellular compartment relevant for the observed TREM2 dependent liver pathology, we performed a reciprocal bone marrow transfer (Fig. 5). As a subset of Kupffer cells is radio-resistant 69 , we additionally treated the chimeric mice with liposomal clodronate to ablate remaining Kupffer cells in irradiated mice as demonstrated previously 62,69 . Reconstitution of bone marrow was confirmed by flow cytometry (Fig. S8). Subsequently, the four groups of chimeric mice were infected with LCMV and serum concentrations of ALT, AST, AP and bilirubin were determined ( Fig. 5A-D). These results revealed reduced liver damage in WT -> Trem2 −/− and Trem2 −/− -> Trem2 −/− chimeric mice, consistent with a loss of TREM2 in radio-resistant cells. In line with these results, chimeric mice lacking non-hematopoietically expressed TREM2 showed reduced viral loads in hepatocytes (Fig. 5E-G). TREM2 is expressed in macrophages but not at detectable levels in hepatocytes (Fig. S9A,B), which may suggest the involvement of another non-hematopoietic cell population. Collectively, these experiments show that loss of TREM2 in the non-hematopoietic compartment results in ameliorated viral hepatitis.

Discussion
In this study, we report a novel pathogenic role for TREM2 that results in aggravated viral hepatitis. Mice lacking the lipid-sensor TREM2 and WT mice were analyzed using the model of LCMV-induced hepatitis, which is largely T-cell dependent 63 . Interestingly, we found reduced liver pathology in Trem2 −/− mice compared to WT mice despite comparable virus-specific T cell responses. This may be due to accelerated virus clearance in the Trem2 −/− mice and reduced virus spread from Kupffer cells to hepatocytes between day 4 and 8 post infection 61 . Of note, we found increased levels of NK and NKT cells in Trem2 −/− mice compared to WT mice preceding the onset of liver pathology. This could indicate an involvement of these cell populations, which have been implicated in the control of LCMV 70,71 . We also detected reduced levels of serum IFNα in Trem2 −/− mice upon infection. It is tempting to speculate that IFNα which is known to impact not only the antiviral cellular state but also the interactions between NK and T cells 72 , could provide additional cues for the underlying molecular mechanisms that govern the observed virus-induced TREM2-dependent phenotype.
TREM2 appears to exert highly specific effects in a cell-type specific and possibly pathogen-specific manner. The discovery that TREM2 is involved in Alzheimer's disease led to a strong focus on its role in the central nervous system (CNS) 59,60 . However, many other studies suggest that TREM2, in addition to its important role in the CNS, plays an immunomodulatory role during infection. For example, TREM2 has been shown to promote the clearance of Pseudomonas aeruginosa keratitis 73 and injection of bone marrow cells over-expressing TREM2 improved bacterial phagocytosis in a model of cecal ligation and puncture 74 . In contrast, lack of TREM2 did not impact the outcome of bacterial peritonitis with Escherichia coli 53 and TREM2 was found to aggravate pneumococcal pneumonia and respiratory infection with Sendai virus 43,75 .
Until now, TREM2's effects were mainly attributed to its expression in macrophages, monocytes and dendritic cells. Thus, it is interesting that our results indicate that TREM2 expression in non-hematopoietic cells is involved in the observed pathology. TREM2 was not detectable in primary hepatocytes and we, thus, hypothesize that a non-hematopoietic cell population other than hepatocytes is exerting the detrimental effects of TREM2 observed in our model of viral hepatitis. Further studies will need to address the potential involvement of radio-resistant liver-resident cell populations such as hepatic stellate cells and liver sinusoidal endothelial cells. In addition, we cannot rule out neither systemic effects from other cell types outside the liver such as TREM2-expressing adipocytes 76 , nor a role for the soluble form of TREM2 75 .
The seminal study of Wang et al. showed that TREM2 binds to phosphatidylcholines and sphingomyelins, assigning a lipid-sensing role to this surface receptor 38 . Interestingly, these are the lipid species we found altered in mice upon infection with LCMV. TREM2 may also sense apoptotic cells through exposed phosphatidylserines 38 . Additional studies implicated ApoE, which plays an important role in lipoprotein metabolism and in immune-related processes and diseases [77][78][79] , as another ligand for TREM2 34,35,37 . ApoE −/− mice infected with LCMV develop exacerbated liver damage 80 , which may suggest the involvement of ApoE in the TREM2-dependent liver pathology. More recently, Trem1 −/− mice were reported to be protected from LCMV-induced hepatitis despite normal CD8 T cell responses 81 . TREM1, whose physiological ligands include peptidoglycan recognition protein 1 and likely others 82 , was found to be mostly expressed in neutrophils and suggested to modulate their inflammatory activity. Our bone marrow chimera experiments indicate that TREM2 is unlikely to play a role in neutrophils, which is supported by the notion that TREM1 is typically expressed by activated neutrophils 83, 84 while TREM2 is not. Thus, the fact that both Trem2 −/− and Trem1 −/− mice show a protective phenotype in viral hepatitis indicates distinct cell type-specific and non-redundant roles of these TREM family members in liver pathology.
The technological advancements in mass-spectrometry based lipidomics enable to map changes in the composition of circulating lipids during infection. Several studies have reported variations in serum levels of specific sphingolipids in chronic HBV and HCV patients and associated them with disease progression such as fibrosis and hepatic steatosis or antiviral treatment outcome 25,[28][29][30][31] . Systematic efforts will be needed to study the biology of individual lipid species in the context of liver pathology and other diseases. It will be crucial to investigate which species are released into the serum by active secretion and which may represent danger-associated molecular patterns from cellular debris at a given time, through which receptors these lipids are sensed by cells and how signal transduction contributes to inflammation. In-depth knowledge about lipid metabolism, perturbations, sensing and signaling in vivo will allow to identify disease-relevant dynamic patterns and assign prognostic value. It is understood that LCMV only shares certain features of immunobiology with HBV and HCV 19,85,86 , and lipid metabolism is known to differ considerably between mouse and man 87,88 . Yet, this versatile experimental system may provide valuable contributions to characterize infection-induced global changes in serum metabolites and dissect the affected downstream signaling cascades. Of note, Trem2 −/− mice subjected to the concanavalin A-induced acute hepatitis model revealed elevated serum transaminases similar to WT mice, which may indicate that the TREM2-dependent pathology requires context-dependent parameters such as time kinetic, antigen specificity or distinct inflammatory milieus found in viral infections.
Our study revealed global changes in circulating lipids in the sera of LCMV-infected mice and identified a novel detrimental role for the non-hematopoietically expressed lipid-sensor TREM2 in a model of viral hepatitis. This may suggest novel ways to ameliorate virus-induced liver pathologies. Numerous additional functional links between infection-induced changes in serum metabolites and immunoregulation are expected to be uncovered 4 , which will allow novel insights into translational immmunometabolism of inflammatory and infectious diseases.  For the generation of chimeric mice, bone marrow was harvested by flushing the femurs and tibias of donor mice as previously described 90 . Recipient mice (WT and Trem2 −/− ) were exposed to a single dose of γ-irradiation (9 Gy) and reconstitution of bone marrow cells was achieved by injecting 2 × 10 6 of the freshly isolated cells from WT or Trem2 −/− donor mice. Four weeks later, chimeric mice were treated with 200 μg of anti-CD90 antibody intraperitoneally (i.p.), to deplete any remaining peripheral T cells from the recipient mouse. Another three weeks later, the mice received 100 μl of liposomal clodronate (ClodronateLiposomes.com) i.p., per 10 g of body weight to deplete any remaining macrophages from the recipient. > 4 weeks later mice were used for infection experiments.

Mouse infections.
Mice were infected intravenously (i.v.) with 0.3-1 × 10 6 focus forming units of LCMV strain WE. LCMV WE was propagated on L929 cells (European Collection of Authenticated Cell Cultures (ECACC) Catalogue No.: 85011425). LCMV titers in serum and organs were determined by immunological focus forming assay as previously described 62 albeit modified by using Vero cells (originally obtained from Daniel D. Pinschewer, University of Basel).
Targeted Metabolomics. Targeted metabolomics was performed using the AbsoluteIDQ p180 kit (Biocrates Life Sciences AG, Innsbruck, Austria). The kit allows the identification and quantitation of metabolites by LC-and flow injection analysis (FIA)-MRM. All serum samples were stored at −80 °C until being analyzed on an AB SCIEX QTrap 4000 mass spectrometer using an Agilent 1200 RR HPLC system, which were operated with Analyst 1.6.2 (AB SCIEX). The chromatographic column was obtained from Biocrates. The serum samples and additional blanks, calibration standards and quality controls were prepared according to the user manual. The experiments were validated with the supplied software (MetIDQ, Version 5-4-8-DB100-Boron-2607, Biocrates Life Sciences, Innsbruck, Austria). Quality controls were analyzed every 20 th sample. Data was analyzed and visualized with R Studio. For metabolites that were below the limit of detection (LOD), the LOD was used for calculating fold changes and p-values.

Extraction of serum lipids and pure lipid species.
Total lipids were extracted from sera from uninfected mice, and mice infected with LCMV WE eight days previously as follows; Briefly, 100 µl of serum was mixed with 150 µl of 0.5 M KH 2 PO 4 , 750 µl of chloroform and 250 µl of methanol. The mixture was then vortexed for two minutes and then centrifuged for five minutes at 500× g at room temperature. The lower lipid phase was collected and stored at −80 °C, before being used to stimulate the cells 1:20.
The pure lipid species 1,2-dipalmitoleoyl-sn-glycero-3-phosphocholine ( Ppar-δ reporter assay. For a dual reporter assay (Promega E1960), MEFs were seeded at 2 × 10 5 cells/ml, were transfected with PPRE luciferase promoter constructs (provided by Nikolina Papac, Medical University of Vienna) and a vector encoding PPAR-δ (provided by Ajay Chawla, University of California, San Francisco) using Lipofectamine LTX (Sigma Aldrich). On the next day, transfected cells were transferred to a 96-well plate at a density of 5 × 10 4 /well and stimulated with either extracted lipids of uninfected or infected WT animals (see above) or pure lipid species (see above). Luciferase activity was measured after 48 h using the Dual-Luciferase Reporter Assay System according to the manufacturer's instructions (Promega). pRenilla luciferase (Promega) activity was used to control the transfection efficiency. Background was measured in non transfected control cells and subtracted. All obtained luciferase values were normalized to renilla and plotted as fold change to untreated control cells.
Cell Culture. Primary hepatocyte isolation was conducted as previously described 62 . Bone marrow-derived macrophages (BMDMs) were generated as previously described 43  Histology. Liver specimens were fixed for 24 hours at room temperature in 4% formaldehyde-PBS solution and paraffin embedded. After cutting, sections (3-4 μm) were deparaffinized using xylene and ethanol, stained with hematoxylin and eosin and a trained pathologist analyzed (blinded) liver pathology using a scoring system, considering lobular and portal inflammation as well as endothelitis and thrombus formation. Each parameter Scientific RepoRts | 7: 11289 | DOI:10.1038/s41598-017-10637-y was scored from 0 to 4 points, with 0 representing absent, 1 very mild, 2 mild, 3 moderate and 4 severe. For staining of cleaved caspase 3 and LCMV NP, antigen-retrieval was performed using citrate buffer pH6.0 (Vector laboratories) followed by blocking of endogenous peroxidase with 2% H 2 O 2 and blocking steps using an avidin biotin blocking kit (Vector laboratories) and rabbit serum. Thereafter sections were incubated over night in either anti-cleaved caspase 3 antibody (Cell signaling) or anti-LCMV nucleoprotein rabbit serum. On the next day, the slides were washed with PBS and incubated with biotinylated swine anti-rabbit IgG (Dako) and developed using the Vectastain ABC kit (Vector laboratories) and DAB-HRP conjugate (Vector laboratories). After counterstaining with hematoxylin and embedding, the sections were analyzed by light microscopy. From each mouse 10 random pictures were taken with a 20-fold magnification and positive cells/field were counted by eye. For the LCMV positive cells, large polyglonal hepatocytes were distinguished from small spindle-shaped infected cells, which were considered as Kupffer cells 61 .
For flow cytometric analysis of different immune cell subsets, livers were cut into small pieces. After incubation in 0.05% collagenase/dispase and 0.01% trypsin inhibitor at 37 °C, liver samples were pressed through a 40 µm cell strainer, centrifuged at 800 g for 5-10 min at 4 °C and resuspended in RPMI 1640. The suspension was then overlayed onto a 33% (vol/vol) Percoll Solution (15 ml) and centrifuged for 30 min at room temperature with no brake. The supernatant was aspirated and red blood cell lysis was performed. After further washing and centrifugation steps, leukocytes were incubated with the following antibodies and analysed by flow cytometry: In vivo Cytotoxicity Assay. Cytotoxic activity of CD8 + T cells was measured as previously described 91 .
Concanavalin A-induced acute hepatitis model. Mice were injected intraperitoneally (i.p.) with either concanavalin A (10 mg/kg, Sigma L7647) or PBS as control. Serum was collected at 24 hours post treatment.

Statistical analyses.
Results are shown as line graph or bar graph with the mean + /− the standard error of the mean as indicated. Statistical differences between experimental groups were calculated by either unpaired t-test, One-way ANOVA or Two-way ANOVA with Bonferroni correction as indicated in the respective figure legends. Significant p-values were calculated as follows; *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 or ****p ≤ 0.0001. Graphs and statistical tests were done with GraphPad Prism 6.