sTREM2 is a plasma biomarker for human NASH and promotes hepatocyte lipid accumulation

Abstract

Background:

Pathogenetic mechanisms of the progression of NAFL to advanced NASH coupled with potential noninvasive biomarkers and novel therapeutic targets are active areas of investigation. The recent finding that increased plasma levels of a protein shed by myeloid cells —soluble Triggering Receptor Expressed on Myeloid cells 2 (sTREM2) —may be a biomarker for NASH has received much interest. We aimed to test sTREM2 as a biomarker for human NASH and investigate the role of sTREM2 in the pathogenesis of NASH.

Methods:

We conducted studies in both humans (comparing patients with NASH vs. NAFL) and in mice (comparing different mouse models of NASH) involving measurements of TREM2 gene and protein expression levels in the liver as well as circulating sTREM2 levels in plasma. We investigated the pathogenetic role of sTREM2 in hepatic steatosis using primary hepatocytes and bone marrow derived macrophages.

Results:

RNA sequencing analysis of livers from patients with NASH or NAFL as well as livers from 2 mouse models of NASH revealed elevated TREM2 expression in patients/mice with NASH as compared with NAFL. Plasma levels of sTREM2 were significantly higher in a well-characterized cohort of patients with biopsy-proven NASH versus NAFL (area under receiver-operating curve 0.807). Mechanistic studies revealed that cocultures of primary hepatocytes and macrophages with an impaired ability to shed sTREM2 resulted in reduced hepatocyte lipid droplet formation on palmitate stimulation, an effect that was counteracted by the addition of exogenous sTREM2 chimeric protein. Conversely, exogenous sTREM2 chimeric protein increased lipid droplet formation, triglyceride content, and expression of the lipid transporter CD36 in hepatocytes. Furthermore, inhibition of CD36 markedly attenuated sTREM2-induced lipid droplet formation in mouse primary hepatocytes.

Conclusions:

Elevated levels of sTREM2 due to TREM2 shedding may directly contribute to the pathogenesis of NAFLD by promoting hepatocyte lipid accumulation, as well as serving as a biomarker for distinguishing patients with NASH versus NAFL. Further investigation of sTREM2 as a clinically useful diagnostic biomarker and of the therapeutic effects of targeting sTREM2 in NASH is warranted.

INTRODUCTION

NAFLD is the most common chronic liver disease in the world, affecting up to 30% of the adult population.^1,2 NAFLD may progress to fibrosing NASH, cirrhosis, HCC, and end-stage liver disease.^3–5 In addition to its primary hepatic manifestations, NAFLD is associated with extrahepatic manifestations, especially cardiometabolic complications.^6,7

The molecular mechanisms driving the development of advanced fibrosing NASH and cirrhosis in a relatively small subset of persons with NAFLD are incompletely understood. To date, there are no Food and Drug Administration-approved pharmacotherapies for NASH. Furthermore, with an increasing number of patients developing NASH-related end-stage liver disease, there is a pressing need to develop noninvasive NASH biomarkers for screening, diagnosis, prognostication, and selection of patients for treatment and monitoring. This is particularly true as liver biopsy, the gold standard diagnostic test for histologic NASH, is limited by its invasive nature, poor patient acceptability, and sampling variability.

Triggering receptor expressed on myeloid cells 2 (TREM2) is an innate immune receptor expressed by myeloid cells, including macrophages (as well as KC in the liver), microglia, dendritic cells, and osteoclasts.⁸ TREM2 is a transmembrane receptor that contains an immunoglobulin-like domain present in the extracellular region and a cytoplasmic tail that lacks any signal transduction or trafficking motifs.⁹ TREM2 is postulated to bind ligands by its extracellular immunoglobulin V-type domain and to transduce intracellular signals through its association with TYRO protein tyrosine kinase binding protein, which recruits the spleen tyrosine kinase through its cytosolic immunoreceptor tyrosine-based activation motifs that regulate myeloid cell functions.⁸ The ligands of TREM2 encompass a wide array of molecules, such as LDL and apolipoprotein E,¹⁰ and relevant to NASH, TREM2 has the ability to sense and recognize the molecular structure of lipids.¹¹ We and others showed that full-length, membrane-anchored TREM2 is proteolytically shed by ADAM17 in myeloid cells, releasing the soluble form of TREM2 (sTREM2) into the circulation.¹²

TREM2 is expressed on circulating myeloid cells, which are recruited to sites of inflammation to participate in innate immune regulation along with the resident immune cells. In the liver, TREM2 is expressed by KCs, the liver’s “resident” macrophages. TREM2-positive transcriptional signatures have been identified in aortic macrophages in lesions of atherosclerosis, in microglia in brain tissue from individuals with Alzheimer disease,^13,14 and in fatty livers of mice fed a high-fat (HF) diet, and other models of acute and chronic liver injury.^15–18

The role of TREM2 and sTREM2 in the pathogenesis of NAFLD and NASH is an area of active investigation. Hendrikx et al recently showed elevated levels of plasma sTREM2 in patients with NASH and cirrhosis,¹⁸ while Chandran et al suggested that plasma sTREM2 may serve as a noninvasive biomarker of elevated liver stiffness suggestive of NASH.¹⁹ Hou et al showed that TREM2 transgenic mice were protected from HF diet-induced hepatic steatosis and that macrophage TREM2 deficiency resulted in more severe hepatic steatosis.²⁰ Sharif et al demonstrated that TREM2 deficiency in nonhematopoietic tissues contributed to severe hepatic steatosis under HF diet feeding.²¹ Moreover, Hendrikx et al demonstrated that hematopoietic Trem2 deficiency caused defective lipid handling and extracellular matrix remodeling, resulting in exacerbated steatohepatitis, cell death, and fibrosis.¹⁸ Wang et al²² and Liebold et al²³ suggested that TREM2-positive macrophages have a protective function in NASH by processing and removing apoptotic hepatocytes.

We aimed to consider separately the effects of full-length, membrane-bound TREM2 versus those of sTREM2, which is released after the cleavage of membrane-bound TREM2, and to further test sTREM2 as a circulating biomarker of human NASH. We confirmed that hepatic expression of TREM2 is elevated in humans and mice with NASH versus NAFL. We demonstrated that plasma sTREM2 levels could distinguish well-characterized patients with biopsy-proven NASH versus NAFL, thus being a potentially useful biomarker in at-risk persons. Furthermore, our mechanistic findings in primary mouse hepatocytes suggested that sTREM2 plays a direct role in promoting lipid accumulation in these cells through a CD36-dependent increase in fatty acid uptake.

METHODS

We conducted a series of studies in both humans (comparing patients with NASH vs. NAFL) and in mice (comparing different mouse models of NASH) that involved measurements of TREM2 gene and protein expression levels in the liver as well as circulating sTREM2 levels in plasma. Furthermore, we determined the pathological role of sTREM2 in hepatic steatosis by using mouse primary hepatocytes and bone marrow-derived macrophages (BMDMs).

Additional details of the materials and methods used can be found in the Supplemental Methods, http://links.lww.com/HC9/A534.

Human subjects: Patients with biopsy-proven NASH versus NAFL

Patient data, serum specimens, and liver tissue were obtained from a biorepository at Veterans Affairs Puget Sound Health care System, used in multiple prior NASH studies.^24–28 This biorepository prospectively recruited patients undergoing clinically indicated liver biopsies and stored liver tissue that was flash-frozen in liquid nitrogen immediately after liver biopsy. We identified patients with NAFLD based on histological hepatic steatosis on liver biopsy in the absence of HCV (negative serum HCV antibody and HCV RNA), HBV (negative serum HBV surface Ag), excessive alcohol consumption (dedicated alcohol questionnaire administered on the day of the liver biopsy), iron overload (hepatic stain and serum iron markers), or markers of autoimmune liver diseases. Liver slides were prospectively reviewed by an expert hepatopathologist who scored the grade of steatosis,^1–3 inflammation,(0–3) ballooning degeneration (0–2), and the stage of fibrosis (0–4) according to the system proposed by Kleiner et al²⁹. From this biorepository, we randomly selected patients with NAFLD who either had histological “NAFL” or “fibrosing NASH” defined as follows:

1. “NAFL”³⁰ (n = 16): Defined by histological steatosis grades 1–3, inflammation grade 0–1, fibrosis stage 0, and ballooning degeneration grade 0.

2. “Fibrosing NASH” (n = 26): Defined by histological steatosis grades 1–3, inflammation grade 1-3, fibrosis stage 1–3, and ballooning degeneration grade 1–2. We purposefully selected patients with NASH who had fibrosis because fibrosis is the histological feature most strongly associated with adverse long-term outcomes in patients with NAFLD.^31,32

Blood specimens were collected from all patients after an overnight fast (12 h) on the day of the liver biopsy and used to measure aspartate aminotransferase, alanine aminotransferase, total cholesterol, LDL, triglycerides, and platelet count. Relevant clinical history was confirmed from medical records, and body weight and height were measured to calculate the body mass index.

All study participants provided informed consent. The study was approved by the Institutional Review Board at Veterans Affairs Puget Sound Health care System.

Mouse models of experimental NASH

We studied 2 different mouse models of experimental steatohepatitis. In the first mouse model, adult male C57BL/6J mice (Jackson Laboratory) were assigned for 24 weeks to either a HF diet (15% fat, wt/wt) or to a HF (15%) and high-cholesterol (HFHC) diet supplemented with 0.75% dietary cholesterol (n = 10–11 per group). Cocoa butter, which contains ~60% saturated fat, was the source of the extra fat in this diet (Supplemental Table 1, http://links.lww.com/HC9/A534). We demonstrated that C57BL/6J mice on this HFHC diet developed fibrosing steatohepatitis (NASH) while mice on the HF diet without added cholesterol developed “simple” hepatic steatosis (NAFL) without substantial inflammation or fibrosis.³³

In a second mouse model, adult male LDL receptor-deficient (Ldlr ^-/-) mice (8–10 weeks of age, n = 8–10 per group) on the C57BL/6J background were fed a rodent laboratory standard diet or HF, high-sucrose diet with added 0.15% dietary cholesterol (high-fat, high-sucrose [HFHS] diet)^34,35 for 20 weeks. The HFHS diet provides 59% of total calories from fat and 36.6% as carbohydrates. Lard, which contains ~60% saturated fat, was the source of the extra fat in this diet (Supplemental Table 1, http://links.lww.com/HC9/A534). As compared to normal chow, the HFHS diet increases insulin resistance, adipose tissue inflammation, chronic systemic inflammation, atherosclerosis, and hepatic steatosis and inflammation in Ldlr ^-/- mice.^34–37 Ldlr ^-/- mice were used as the second model because they exhibit a lipoprotein profile more similar to humans than wild-type mice.

At the end of the experiments, mice underwent phlebotomy and were euthanized by cervical dislocation following isoflurane anesthesia, and their livers were harvested after perfusion for studies as outlined below.

All animal studies were approved by the Animal Care and Use Committee of the University of Washington or the Veterans Affairs Puget Sound Health Care System.

Cell culture experiments of hepatocytes and bone marrow-derived macrophages

BMDMs were isolated and cultured as described.³⁸ After 7 days, BMDMs either from wild-type (WT), Lyz2 ^Cre/Cre ; Adam17 ^wt/wt or Lyz2 ^Cre/Cre ; Adam17 ^fl/fl mice were cocultured with primary hepatocytes in transwell plates in the presence or absence of recombinant sTREM2-Fc chimeric protein (100 ng/ml, R&D systems # 1729-T2-050) or recombinant Fc (IgG1) control protein (100 ng/ml, G&P Biosciences # FCL0029). Lipid accumulation was induced in primary hepatocytes by incubation with palmitate (0.25 mM, 18 h) conjugated with albumin as described.³⁹ Briefly, palmitate was first dissolved in NaOH (100 mmol/l) and conjugated with fatty acid–free albumin (MilliporeSigma) at a molar ratio of 3:1 (palmitate/albumin). Lipid accumulation in hepatocytes was determined by staining with the neutral lipid dye BODIPY 493/503 (5 µg/ml, Invitrogen) or a perilipin 2 antibody (Novus Biologicals, Cat # NB110-40877) and quantification of the positive area by ImageJ. Negative controls are shown in the Supplemental methods, http://links.lww.com/HC9/A534.

In additional experiments, isolated mouse primary hepatocytes were treated with recombinant sTREM2-Fc chimeric protein (100 ng/ml) for 18 hours under serum-free conditions, and lipid accumulation in hepatocytes was determined by quantification of BODIPY-positive area.

Hepatic gene expression studies by RNA sequencing (RNA-seq)

For RNA-seq, total RNA was isolated from frozen human livers from the VA biorepository and mouse livers using RNeasy kits (Qiagen) with on-column DNase digestion. RNA (0.5 ng) was reverse transcribed into full-length amplified cDNA. Dual-index, single-read sequencing of pooled libraries was carried out on a HiSeq2500 sequencer (Illumina) with a target depth of 5 million reads per sample. Gene counts were generated, and QC and metrics analysis was performed.

ELISA for human and mouse sTREM2

Human or murine sTREM2 levels were determined by ELISA on serum/plasma specimens as described by Ewers et al and Piccio et al, respectively.^40,41 Murine sTREM2 was also measured in gallbladder bile samples collected from the C57BL/6J mouse model by ELISA.

Immunohistochemistry for TREM2 in human and mouse liver tissue

Human biopsy samples and liver sections from mice were immunostained for TREM2, as described.²⁰ Negative isotype controls of the same concentrations were used as controls (Supplemental Fig. 1, http://links.lww.com/HC9/A534). All photos were taken by fluorescence microscopy (Nikon A1R confocal microscope) using a UV filter at a magnification of × 60. The percentage of positive staining was calculated using Image J analysis software.

Statistical analysis

All animal and cell culture data are presented as mean ± SEM. In all experiments, the sample size (n) represents the number of individual mice or human samples. The statistical parameters (n, mean, SEM, and statistical tests used) can be found within the figure legends. Tests for normality (Shapiro-Wilk) and equal variance (Brown–Forsythe) were performed for each of the datasets. To define differences between 2 datasets, 2-tailed unpaired nonparametric Mann-Whitney U test was used. Comparison of multiple datasets was done using 1-way ANOVA with Tukey’s post hoc test. The criterion for significance was set at p < 0.05. Statistical analyses were performed using GraphPad Prism, version 8.4.3 (San Diego, CA). For RNA-seq, gene expression differences were evaluated by limma analysis,⁴² and genes significant at 5% false discovery rate were considered to be differentially expressed.

RESULTS

Characteristics of patients with NAFL and NASH

Patients with NASH (n = 26) and simple steatosis (n = 16) had similar mean age (~51 y) and mean body mass index (~34 kg/m²) and were predominantly male (81% vs. 94%), as expected in a Veteran population (Table 1). Compared to patients with NAFL, those with NASH had more advanced hepatic fibrosis (fibrosis stage 1/2/3/4 in 52%/20%/16%/4%, respectively, vs. 100% with stage 0 in the NAFL group), more advanced inflammation (inflammation grade 1/2/3 in 42%/54%/4%, respectively, vs. 100% with grade 0 or 1 in the NAFL group), consistent with the histological criteria we used to define NAFL and NASH. Compared to patients with NAFL, those with NASH tended to be more likely to have diabetes (62% vs. 25%) and had higher mean serum transaminase levels (Table 1).

TABLE 1.

Comparison of patients with NASH versus NAFL with respect to histological, clinical, and demographic characteristics

Clinical and demographic characteristics	NAFL N = 16	NASH N = 26
Age, mean (SD)	51.3 (12.8)	50.4 (10.0)
Male, n (%)	15 (94)	21 (81)
Race/ethnicity
Non-Hispanic White	16 (100)	19 (73)
Non-Hispanic Black	0 (0)	1 (4)
Hispanic	0 (0)	2 (8)
Other	0 (0)	4 (15)
BMI, kg/m2 (SD)	33.9 (5.8)	33.6 (5.2)
Diabetes, n (%)	4 (25)	16 (62)
Fasting blood levels, mean (SD)
ALT (U/L)	48.9 (24.2)	98.5 (39.3)a
AST (U/L)	36.6 (26.0)	58.3 (25.7)b
Total cholesterol (mg/dl)	198.2 (46.4)	206.8 (39.6)
LDL-C (mg/dl)	104.3 (30.0)	116.4 (41.9)
HDL-C (mg/dl)	43.6 (20.7)	39.4 (11.9)
Triglyceride (mg/dl)	269.2 (256.0)	264.3 (99.9)
Platelet count (K/μl)	230.9 (49.2)	227.4 (60.4)
Hepatic histology, n (%)
Steatosis gradea
0	0 (0)	0 (0)
1	9 (56)	1 (4)
2	6 (38)	16 (62)
3	1 (6)	9 (34)
Inflammation gradea
0	3 (19)	0 (0)
1	13 (81)	11 (42)
2	0 (0)	14 (54)
3	0 (0)	1 (4)
Ballooning degenerationa
0	11 (69)	1 (4)
1	5 (31)	11 (42)
2	0 (0)	14 (54)
Fibrosis stagea
0	16 (100)	2 (8)
1	0 (0)	13 (52)
2	0 (0)	5 (20)
3	0 (0)	4 (16)
4	0 (0)	1 (4)

Notes: Data are expressed as mean, and numbers within parenthesis represent either SD or percentage of patients.

^a p<0.01 different between NASH and simple steatosis.

^b p<0.05.

p-value from unpaired 2-tailed Student t-test.

Characteristics of mouse models with NAFL and NASH

Mice on an HFHC diet for 24 weeks developed fibrosing NASH, characterized by histological evidence of fibrosis and a 10-fold greater Sirius red-positive area (mean 3% vs. 0.3%) measured as a marker of fibrosis, histological evidence of inflammation, and advanced steatosis together with elevated plasma transaminases, cholesterol, and triglyceride levels. In contrast, mice on an HF diet developed NAFL, with advanced steatosis but no fibrosis and limited inflammation (Table 2).

TABLE 2.

Comparison of the histological and laboratory characteristics of mice fed with HF or HFHC diet

Genetic Strain	Wild-type	Wild-type
Genetic background	C57BL/6	C57BL/6
Diet	High-fat 0% Cholesterol HF	High-fat 0.75% Cholesterol HFHC
Duration	24 weeks	24 weeks
Number of mice	11	10
Body weight (g)	44.5 ± 2.6	47.5 ± 4.9
Liver weight (g)	2.7 ± 0.6	4.3 ± 1.1a
Liver weight/body weight (%)	6.0 ± 1.0	9.0 ± 1.5a
Food consumption (g/mouse/day)	3.2 ± 0.3	2.9 ± 0.3+
Plasma levels (fasting)
ALT (U/L)	220.5 ± 54.6	486.5 ± 103.7a
AST (U/L)	233.3 ± 90.1	352.7 ± 76.1a
ALP (U/L)	119.8 ± 26.3	148.6 ± 24.4+
Cholesterol (mg/dl)	187± 50	332 ± 39a
Triglyceride (mg/dl)	55 ± 9	70 ± 5a
Glucose (mg/dl)	325 ± 1.4	315 ± 44
Insulin (ng/ml)	2.7 ± 2.6	4.5 ± 2.5
HOMA-IR	44.1 ± 31.5	63.7 ± 33.4
HMW-adiponectin (mg/ml)	4.18 ± 1.33	6.75 ± 3.66
Hepatic histology, n (%)
Steatosis grade
0
1
2	3 (27)	3 (30)
3	8 (73)	7 (70)
Inflammation gradea
0	3 (27)	0
1	7 (64)	6 (60)
2	1 (9)	3 (30)
3	0	1 (10)
Fibrosis stagea
0	11 (100)	0
1	0	10 (100)
2	0	0
3	0	0
Sirius Red staining (fibrosis), % area	0.30 ± 0.28	2.96 ± 2.1a

Notes: Data are expressed as mean ± SD, number within parenthesis represents the percentage of mice.

^a p-value from unpaired 2-tailed Student’s t-test (p<0.05, different between HF and HFHC).

Abbreviations: ALP, alkaline phosphatase; HF, high-fat; HFHC, high-fat high-cholesterol; HOMA-IR, homeostatic model assessment for insulin resistance.

We also studied Ldlr ^-/- mice, which develop insulin resistance with a human-like lipoprotein profile on feeding an HFHS diet. Ldlr ^-/- mice on an HFHS diet for 20 weeks developed hepatic steatosis, mild hepatic inflammation, and elevated levels of transaminases (Supplemental Fig. 1C, http://links.lww.com/HC9/A534), suggesting early stages of NASH. Ldlr ^-/- mice on normal chow had essentially normal liver histology with no steatosis, inflammation, or fibrosis (Table 3).

TABLE 3.

Comparison of the histological and laboratory characteristics of mice fed with chow or high-fat, high-sucrose diet

Genetic background	C57BL/6J (Ldlr-/-)	C57BL/6J (Ldlr-/-)
Diet	Normal chow	HFHS
Duration	20 wks	20 wks
Number of mice	7–9	9
Body weight	28.3 ± 1.2	52.02 ± 1.2a
Plasma levels (fasting)
Glucose	226 ± 13.5	278.2 ± 16.0a
AST (pg/ml)	454.3 ± 100.2	934.2 ± 75.1a
Cholesterol (mg/dl)	221.1 ± 13.74	1462.1 ± 56.74a
Triglyceride (mg/dl)	338.1 ± 13.74	1176.9 ± 15.14a
Insulin (ng/ml)	0.56 ± 0.164	5.51 ± 2.16a
Hepatic histology, n (%)
Steatosis gradea
0	7 (100)	0
1	0	0
2	0	9 (100)
3	0	0
Inflammation gradea
0	7 (100)	4 (44)
1	0	5 (56)
2	0	0
3	0	0
Fibrosis stage, n (%)
0	7 (100)	6 (67)
1	0	3 (33)
2	0	0
3	0	0

^a Notes: Data are expressed as mean ± SEM, number within parenthesis represents the percentage of mice. p-value from unpaired 2-tailed nonparametric Mann-Whitney test (p<0.05, difference between HF and HFHS). HFHS diet with added 0.15% dietary cholesterol. Statistical analysis using Student t-test.

Abbreviations: HF, high fat; HFHS, high-fat, high-sucrose.

Transcriptomic analysis demonstrates higher hepatic TREM2 mRNA levels in NASH versus NAFL, in both humans and mice

We compared the global gene expression profiles in liver tissue in patients and mice with NASH versus NAFL, with particular attention to gene transcripts related to fibrosis, inflammation, and myeloid cells. Gene transcripts related to fibrosis (COL1A1, COL1A2, MMP2, MMP9, COL3A1, COL5A1, COL5A3, COL14A1), inflammation (IL8, CD109, P2RX7, CCL20, DFNA5), and myeloid cells (TREM2, ITGAM) were significantly higher in patients with NASH as compared to the liver from patients with NAFL (Figure 1A). Also, in mice, hepatic gene transcripts related to fibrosis (Col1a1, Col1a2, Col3a1, Col4a1, Col4a2, Col5a1, Col5a2, Col6a1), inflammation (Tnfa, Ccl2, Tlr1, Saa1, Gasdmd, Il7r, Casp4, Casp6), and myeloid cells (Trem2, Itgam, Cx3cr1, Clec4f, Ccr2, Cd44, Cd68, Tyrobp) were higher in mice on an HFHC diet (NASH) than in mice on an HF diet (NAFL) (Figure 1B). Therefore, in both humans and mice, TREM2 gene expression was higher in the setting of NASH compared to simple steatosis (Figure 1C). Extended lists of gene expression profiles based on the RNA-seq results are shown in Supplemental Table 2–7, http://links.lww.com/HC9/A534. In addition, induction of hepatic gene expression of Trem2 was observed in Ldlr ^-/- mice on HFHS as compared to the chow diet (Figure 1D), demonstrating that hepatic Trem2 mRNA levels are also induced in a model of mild NASH as compared to normal liver. Tyrobp, a Trem2 + macrophage-related marker and the putative ligand of Trem2, showed increased expression at the RNA level in mice with NASH compared to NAFL which was not observed in the human analysis.

FIGURE 1.

Similar transcriptomic changes in the liver from patients with NASH and in a mouse model of NASH. (A) Heat map of gene clusters related to fibrosis and inflammation in the liver from patients with NAFL or NASH. (B) Heat map of gene clusters related to fibrosis and inflammation in the liver from HF and HFHC fed mice. (C) Venn diagram showing the overlap of significantly modified genes related to myeloid populations in patients with NASH and HFHC fed (NASH) mouse liver derived from the RNA-seq experiments shown in A and B. Red color indicates upregulation of genes and blue color indicates downregulation of genes in the liver from patients with NASH as compared to patients with NAFL. (D) Ldlr ^-/- mice were fed either chow (n = 9) or HFHS diet (n = 9) for 22 weeks, and liver Trem2 mRNA levels were determined by RT-PCR. For RNA-seq, gene expression differences were evaluated by limma analysis, and genes significant at 5% FDR were considered to be differentially expressed (A–C). Statistical analysis was performed by unpaired 2-tailed nonparametric Mann-Whitney test (D) *p < 0.05. Data are shown as mean ± SEM. Abbreviations: FDR, false discovery rate; HF, high-fat; HFHC, high-fat, high-cholesterol; HFHS, high-fat, high-sucrose.

Hepatic TREM2 protein immunoreactivity is higher in NASH versus NAFL in both humans and mice

Patients with NASH exhibited higher hepatic TREM2 immunoreactivity compared to patients with NAFL (Figure 2A). Immunostaining analysis also revealed that TREM2 in the liver was mainly expressed on CD68 + macrophages/KCs, as expected (Figure 2A). Similarly, hepatic TREM2 immunoreactivity was higher in mice fed an HFHC diet (NASH) than mice fed an HF diet (NAFL) (Figure 2B) and also in Ldlr ^-/- mice on HFHS diet versus normal chow (Figure 2C). These data indicate that consistent with Trem2 gene expression, experimental NASH resulted in elevated hepatic TREM2 immunoreactivity as compared to simple steatosis and that TREM2 is expressed primarily in macrophages/KCs.

FIGURE 2.

Liver TREM2 immunoreactivity is elevated in patients and mouse models of NASH (A) Liver sections from both patients with NAFL and NASH were immunostained for TREM2 (red), CD68 (green), and DAPI (blue, nucleus stain). Confocal images at × 60 were taken with a Nikon A1 inverted fluorescent microscope, and TREM2⁺ area was quantified using ImageJ. Individual CD68⁺ TREM2⁺ macrophages are depicted at a higher magnification in the inserts. (B) C57BL/6 mice were fed either HF (n = 11) or HFHC (n = 10) diet for 24 weeks. Liver sections were used to immunostain TREM2 (red) and DAPI (blue). Images at × 20 were taken with a Keyence fluorescence microscope (BZ-X800), and TREM2 + area was quantified using image J. (C) Ldlr ^-/- mice were fed either chow (n = 7) or HFHS diet (n = 9) for 20 weeks and liver sections were immunostained for TREM2 (red) and DAPI. Images at × 20 were taken with a Keyence fluorescence microscope (BZ-X800), and TREM2⁺ area was quantified using image J. Statistical analysis was performed by unpaired 2-tailed nonparametric Mann-Whitney test (A–C) *p<0.05. Data are shown as mean ± SEM. Scale bar: 50 μm (A) or 20 μm (B–C). Abbreviations: HF, high-fat; HFHC, high-fat, high-cholesterol; HFHS, high-fat, high-sucrose; TREM2, Triggering Receptor Expressed on Myeloid cells 2.

Circulating sTREM2 levels are higher in mice and patients with NASH versus NAFL

The protease ADAM17 mediates proteolytic cleavage (shedding) of full-length membrane-anchored TREM2, which generates sTREM2.¹² sTREM2 levels in serum were significantly higher in patients with histologically proven NASH (mean ± SEM; 1051 pg/ml ± 134) versus those with NAFL (mean 477 pg/ml ± 84) (Figure 3A). The area under the receiver-operating characteristic curve was 0.807 for discriminating NASH versus NAFL in this population. A threshold level of sTREM2 ≥ 986 pg/ml had 56.5% sensitivity and 92.9% specificity for NASH versus NAFL (Figure 3B). Mice on an HFHC diet or HFHS diet had significantly higher gene expression of hepatic Adam17 as compared with mice fed an HF diet or normal chow, respectively (Figure 3C-3D). Consistent with higher expression of Adam17, circulating levels of sTREM2 were also significantly higher in HFHS-fed mice as compared with mice fed chow diet (Figure 3E). For the first time, sTrem2 was measured in bile collected from the gallbladder taken at the time of liver removal. The bile collected from the mice on an HFHC diet had significantly higher levels of sTrem2 compared to the control diet (Figure 3F). In other related experiments, we compared sTrem2 levels in both the plasma and gallbladder bile of the same mice, and this showed a strong positive correlation (Pearson’s r = 0.7647) between higher plasma sTrem2 and bile sTrem2 levels (Data not shown).

FIGURE 3.

Circulating soluble TREM2 levels are elevated in mice and patients with NASH. (A) Serum samples from NAFLD (n = 15) and NASH (n = 22) patients were analyzed for soluble TREM2 by ELISA. (B), ROC curves for diagnostic accuracy for NASH using serum levels of sTREM2. (C) C57BL/6 mice were fed either HF (n = 11) or HFHC (n = 10) diet for 24 weeks, and liver Adam17 mRNA levels were determined by RNA sequencing. (D) Ldlr ^-/- mice were fed either chow (n = 9) or HFHS diet (n = 10) for 20 weeks, and liver Adam17 mRNA levels were determined by RT-PCR. (E) Serum samples from mice fed chow (n =9) or HFHS diet (n = 6) were analyzed for soluble TREM2 by ELISA. Statistical analysis was performed by unpaired 2-tailed nonparametric Mann-Whitney test. (A, C–E). Data are shown as mean ± SEM. (F) Gallbladder bile samples from mice fed chow, HF, or HFHC diet for 3 months were analyzed for sTREM2 by ELISA. Statistical analysis was performed by unpaired t-test, and data are shown as mean ± SD. Abbreviations: HF, high-fat; HFHC, high-fat, high-cholesterol; HFHS, high-fat, high-sucrose; sTREM2, soluble TREM2; TREM2, Triggering Receptor Expressed on Myeloid cells 2.

sTREM2 increases lipid accumulation in primary hepatocytes through CD36

To investigate if sTREM2 plays a causative role in fatty liver disease, we used ADAM17-deficient macrophages, which exhibit a deficit in TREM2 shedding and increased membrane-anchored TREM2 expression.¹² First, to determine the effect of impaired shedding of sTREM2 in vitro, BMDMs from WT or Adam17 ^-/- mice were cocultured with primary mouse hepatocytes isolated in transwell plates. WT cultures supplemented with palmitate conjugated to albumin in serum-free conditions, which is known to activate ADAM17-mediated sTREM2 shedding,⁴³ showed increased hepatic lipid droplet content (Figure 4A). However, palmitate-induced lipid droplet formation was markedly reduced when WT hepatocytes were cocultured with ADAM17^-/- macrophages (Figure 4A). To further test the idea that sTREM2 generated from hepatic myeloid cells exerts detrimental effects in hepatocytes, we stimulated mouse primary hepatocytes with recombinant sTREM2 protein. The extracellular domain of TREM2, spanning the N-terminal amino acids 19-168 fused with IgG-Fc, was used to mimic the soluble form of TREM2. The addition of recombinant exogenous sTREM2 protein to the macrophage-hepatocyte coculture system rescued the effect of ADAM17^-/- macrophages and increased lipid droplet formation in mouse primary hepatocytes to levels seen in cocultures with WT BMDMs (Figure 4B), suggesting that the effect of the ADAM17^-/- macrophages on hepatocyte lipid droplet formation was in fact due to a lack of sTREM2 shedding.

FIGURE 4.

Soluble TREM2 increases lipid droplets and triglyceride content in hepatocytes. (A) BMDMs from WT or ADAM17-deficient mice were cocultured with primary hepatocytes in transwell plates. Cultures were added with BSA or BSA-conjugated palmitate (0.25 mM) for 18 hours in serum-free conditions. Lipid accumulation in hepatocytes was determined by quantification of BODIPY-positive area by ImageJ (n = 4-6). (B) Coculture of ADAM17-deficient BMDMs and hepatocytes stimulated with sTREM2 in transwell plates in the presence of palmitate (0.25mM) for 18 hours. Lipid accumulation in hepatocytes was determined by quantification of BODIPY-positive area by ImageJ (n = 4). (C) Representative images and quantification of BODIPY-positive area or (D) perilipin-2 positive area in mouse primary hepatocytes after treatment with sTREM2 for 18 hours (n = 4) in the presence of palmitate (0.25 mM). (E) Triglyceride levels in mouse primary hepatocytes after treatment with sTREM2 for 18 hours (n = 4) in the presence of palmitate (0.25 mM). Statistical analysis was performed by 1-way ANOVA with Tukey multiple comparisons tests (A–B) or unpaired 2-tailed nonparametric Mann-Whitney test (C–E). Data are shown as mean ± SEM. Abbreviations: BMDM, bone marrow-derived macrophages; BSA, bovine serum albumin; sTREM2, soluble TREM2; TREM2, Triggering Receptor Expressed on Myeloid cells 2; WT, wild-type.

In the absence of macrophages, after an 18-hour incubation, sTREM2 significantly increased lipid droplet accumulation in hepatocytes in the presence of palmitate (Figure 4C). Increased immunoreactivity of the lipid droplet protein perilipin 2 (Figure 4D) and increased intracellular triglyceride content (Figure 4E) confirmed the increase in lipid content in sTREM2-treated primary hepatocytes. Importantly, we found that the presence of sTREM2-Fc fusion protein, but not Fc alone, significantly enhanced the lipid droplet accumulation of primary hepatocytes (Supplemental Fig. 2A, http://links.lww.com/HC9/A534), indicating that the effect of sTREM2-Fc was due to TREM2 and not to the Fc portion of the chimeric protein. Additionally, the effect of sTREM2 on lipid droplets was specific for active sTREM2, as the effect was lost on heat inactivation (Supplemental Fig. 2A, http://links.lww.com/HC9/A534).

CD36, which belongs to the class B scavenger receptor family, is a transmembrane glycoprotein that serves as a facilitator of lipid transport, including fatty acid uptake. We observed that accumulation of lipids in sTREM2-treated hepatocytes was associated with increased mRNA (Figure 5A) and protein levels (Figure 5B) of CD36 without significant changes in the lipogenic genes, Srebf1c, and Fasn (Supplemental Fig. 2B, http://links.lww.com/HC9/A534). Similarly, hepatic mRNA expression of CD36 was higher in Ldlr ^-/- mice fed an HFHS diet than in mice fed a chow diet (Figure 5C). Lastly, we inhibited CD36 activity in mouse primary hepatocytes by using sulfosuccinimidyl oleate⁴⁴ to investigate whether CD36 plays a causative role in sTREM2-induced lipid droplet formation. Our data demonstrated that inhibition of CD36 activity by sulfosuccinimidyl oleate prevents the effect of sTREM2 on BODIPY-positive lipid accumulation (Figure 5D). Taken together, these results suggest that in mouse hepatocytes, sTREM2-promoted lipid accumulation is mediated by CD36.

FIGURE 5.

sTREM2 increases lipid droplets by CD36-dependent manner in hepatocytes. (A) Mouse primary hepatocytes were treated with sTREM2 for 18hr and mRNA expression of CD36 normalized to 18S was determined (n = 4). This shows increased CD36 mRNA levels in sTrem2-treated hepatocytes. (B) Representative image and quantification of CD36 protein levels normalized to beta-actin in hepatocytes treated with sTREM2 (n = 4). This shows increased CD36 protein levels in sTrem2-treated hepatocytes. C. Ldlr ^-/- mice were fed either chow (n = 9) or HFHS diet (n = 10) for 20 weeks, and liver Cd36 mRNA levels were determined by RT-PCR. (D) Mouse primary hepatocytes were pretreated with SSO (25 μM), an inhibitor of CD36. After 30 minutes, cells were treated with sTREM2 with palmitate for 18 hours. Lipid accumulation in hepatocytes was determined by quantification of BODIPY-positive area by ImageJ (n = 4). (E) Working model. Increased ADAM17-mediated shedding of cell surface TREM2 from hepatic macrophages under conditions of Western diet/obesity leads to elevated soluble TREM2 (sTREM2). Elevated levels of sTREM2 increase hepatocyte triglyceride accumulation by increasing CD36-mediated fatty acid uptake, which can aggravate NAFLD progression. Statistical analysis was performed by unpaired two-tailed nonparametric Mann-Whitney test (A–C). Data are shown as mean ± SEM. Abbreviations: HFHC, high-fat, high-cholesterol; HFHS, high-fat, high-sucrose; SSO, sulfosuccinimidyl oleate; sTREM2, soluble TREM2; TREM2, Triggering Receptor Expressed on Myeloid cells 2.

DISCUSSION

In this series of animal and human experiments, we demonstrate that hepatic TREM2 expression is increased in NASH and that circulating sTREM2 is a promising noninvasive plasma-based biomarker that can distinguish NASH from NAFL in at-risk individuals. Furthermore, we demonstrate that sTREM2 increases lipid accumulation in hepatocytes through CD36 and may thereby directly contribute to the progression of hepatic steatosis and NASH. Further investigation of the therapeutic effect of inhibiting sTREM2 or the shedding of TREM2 to sTREM2 in NASH is warranted.

The pathogenesis of NASH is incompletely understood, and there is also an urgent need for novel blood-based, noninvasive biomarkers that may be used to screen for NASH, either alone or as part of biomarker panels or prediction models. Indeed, gaps in our understanding of NASH pathogenesis and progression, combined with insufficient or unreliable screening and surveillance modalities, contribute to delays in the diagnosis and lack of noninvasive screening tests or approved pharmacotherapies for NASH. Our data from a well-characterized prospective biorepository of patients undergoing clinically indicated liver biopsy suggest that serum sTREM2 levels can differentiate histological NASH from NAFL and may be a promising, novel noninvasive biomarker. Two recent independent cohorts from Austria and Denmark also reported very good performance characteristics of circulating sTREM2 as a biomarker for NASH.^18,19 We believe the results of sTREM2 so far are promising enough to merit larger, prospective validation studies of its use as a clinical biomarker. Furthermore, our in vitro results suggest that not only is sTREM2 a potential biomarker of advanced steatohepatitis, but it may also drive the development of hepatic steatosis.

Although others observed that hepatic Trem2 expression is elevated under various pathophysiological conditions,^16,18,19 the exact molecular trigger and its role are not completely understood. We demonstrated that hepatic TREM2 expression and immunoreactivity were both consistently higher in the setting of NASH compared with NAFL in humans as well as in 2 different mouse models. In addition, the RNA-seq analysis in the liver from humans and mouse models revealed increases in myeloid cell markers. In line with our observations, Hendrikx et al recently suggested that elevated Trem2 expression in the liver is mainly due to recruitment of Trem2-expressing myeloid cells.¹⁸ Taken together with prior findings, our results raise the possibility that the elevated expression of TREM2 in fatty liver disease is due to an increased recruitment of circulating myeloid cells rather than originating from resident macrophages/KCs. However, mechanisms that increase the recruitment of circulating myeloid cells that might become TREM2-positive in the liver remain to be identified.

Our understanding of the role of TREM2 and sTREM2 in NAFLD/NASH is evolving very quickly. Hendrikx et al reported that recruited Trem2-expressing macrophages appeared to be protective in nature as hematopoietic TREM2 deficiency caused defective lipid handling and extracellular matrix remodeling, resulting in exacerbated steatohepatitis, cell death, and fibrosis in methionine and choline-deficient diet-induced NASH in mice.¹⁸ Wang et al²² and Liebold et al²³ suggested that TREM2-positive macrophages, also known as NASH-associated macrophages or Scar-associated macrophages, exercise a protective function by processing and removing apoptotic hepatocytes. Our findings extend our understanding of the role of TREM2/sTREM2 by suggesting that cleavage of TREM2 from macrophages may promote NASH in 2 distinct ways: first, by reducing TREM2-positive macrophages which have protective functions as postulated^18,22,23 and second, by releasing sTREM2, which we demonstrate may then induce lipid accumulation in hepatocytes (Figure 5E). Our results show that the deficiency of ADAM17, which prevents sTREM2 generation by shedding, ameliorates hepatocyte lipid droplet accumulation, while the addition of exogenous sTREM2 rescues this protective effect of Adam17^-/- macrophages. Furthermore, the treatment of hepatocytes with recombinant sTREM2 protein results in lipid droplet and triglyceride accumulation, which appears to be mediated by a CD36-dependent pathway. Figure 5E demonstrates our hypothesized working model whereby obesity and/or a Western diet may increase ADAM17, resulting in cleavage of membrane-bound TREM2 from macrophages and release of sTREM2, which promote steatosis and steatohepatitis. Taken together, our results suggest that inhibiting the shedding of TREM2 into sTREM2 might be a promising therapeutic approach for NAFLD/NASH.

Much more needs to be elucidated to properly understand the role of TREM2/sTREM2 in NASH and design rational therapeutic targets, and we acknowledge the limitations of the current study. First, to investigate the contribution and functional role of sTREM2 versus full-length TREM2 in NASH would require mice expressing an uncleavable TREM2 mutant that lacks sTREM2. Second, testing the effect of a TREM2 antibody,^45,46 which can prevent TREM2 shedding, or the use of a recombinant anti-TREM2 antibody^47,48 in the development of NASH, would be important. Third, we did not specifically investigate the hypothesis that increased plasma sTREM2 levels in NASH may be due to resistance to the actions of sTREM2. Fourth, although our data strongly suggest that sTREM2 can induce lipid droplet accumulation through a CD36-mediated increase in fatty acid uptake, we acknowledge that the inhibitor of CD36 that we employed (sulfosuccinimidyl oleate) may not be completely specific for CD36. Thus, the molecular mechanisms and other functional properties of sTREM2 in the liver still need to be identified. Along this line, it is well known that as a ligand, sTREM2 binds to a number of receptors, affecting signal transduction pathways to modulate inflammation.⁴⁹

In conclusion, our results in a well-characterized cohort of patients with biopsy-proven NASH versus NAFL confirm and validate the findings of 2 simultaneous independent cohorts showing sTREM2 as a promising serum-based biomarker of NASH. We also demonstrate for the first time that sTREM2 may play a causative role in the pathogenesis of NASH by inducing lipid accumulation in hepatocytes. Therefore, inhibiting sTREM2 or the shedding of TREM2 to sTREM2 may represent therapeutic targets in NASH. Further investigations are warranted to elucidate the interplay between membrane-bound TREM2 and sTREM2 in the pathogenesis of NAFLD and NASH.

Disclaimer: The contents do not represent the views of the US Department of Veterans Affairs or the United States Government.

Supplementary Material

hc9-7-e0265-s001.docx ^{(1.2MB, docx)}