Bilirubin ameliorates murine atherosclerosis through inhibiting cholesterol synthesis and reshaping the immune system

Animal experiments in our study were approved by the Institutional Committee for the Use and Care of Laboratory Animals of Guangzhou Medical University (2017-014) and performed in accordance with the guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes or the NIH guidelines. 8-week-old male ApoE^−/− mice (C57BL/6.129P2-APOE/J, Beijing Vital River Laboratory Animal Technology Co., Ltd) were divided into two subgroups. One group (Bilirubin, n = 14) received an intraperitoneal injection of bilirubin (20 mg/kg/d, B4126, Sigma) every other day for 10 weeks. Mice injected with the vehicle served as a control (Control, n = 8). Both groups started the western-type diet supplemented with 21% fat (wt/wt) and 0.15% cholesterol (wt/wt), simultaneously receiving the injection of bilirubin or vehicle. All mice were maintained on these diets for 10 weeks. The body weight, diet, and drinking of mice were monitored for a continuous week. The mice were euthanized with an overdose of pentobarbital sodium (200 mg/Kg) for tissue collection.

Peripheral blood was collected into a heparin-coated tube by cardiac puncture and centrifuged to isolate plasma. Total bilirubin (T-BIL) in the plasma was determined by the vanadic acid oxidation method. Liver enzymes, including alanine aminotransferase (ALT) and aspartate aminotransferase (AST), were measured by the ultra-violet lactate and malate dehydrogenase methods. Total cholesterol and triacylglycerol levels were measured by an oxidase method. High-density cholesterol (HDL-C) and low-density cholesterol were measured by a direct method. The total bilirubin assay kit was from FUJIFILM Wako pure Chemical Corporation, Japan. The kits for glucose, liver enzymes, lipids, and lipoproteins were from Kehua Bio-Engineering, China. All the spectrophotometric methods were performed according to the manufacture’s instructions using an autoanalyzer.

The heart was perfused with phosphate-buffered saline. Whole aorta from the heart, extending 5–10 mm after bifurcation of the iliac, including the brachiocephalic trunk, left common carotid artery, and subclavian arteries, were dissected free of fat tissues and opened longitudinally before being fixed in formalin solution at room temperature (RT) for 24 h. After staining with Oil Red O (ORO), aortae were pinned on a black wax plate and photographed under stereoscopy at standardized magnification and illumination. The ORO positive area and entire aorta tree area were quantified using Image J software. Aorta lesion areas were expressed as a percentage of the whole aorta tree area. The aortic root and brachiocephalic trunk were serially sectioned into 5 μm cryosections. The first section of the aortic sinus was harvested and collected in pairs on glass slides when all three aortic valves became visible in the lumen of the aorta. Neutral lipids deposition in the lesion was visualized by hematoxylin and eosin (H&E) or ORO staining, respectively. The collagen of the lesion was detected by Masson's Trichrome staining. The liver was dissected and weighed, expressed as a ratio of liver weight to body weight. Liver sections were collected and performed H&E or ORO staining for histological and lipid deposition analysis.

The contents of macrophages and smooth muscle cells in aortic sinus lesions were determined by staining with anti–Galectin 3(ab53082) and anti-α-SMA (ab5694) and detecting with Alexa-conjugated antibodies (ab150073 and ab150080). Nuclei were counterstained with DAPI. All antibodies were purchased from Abcam.

Hepatic total cholesterol and free cholesterol concentration were measured using a tissue total cholesterol assay kit (E1015; Applygen Technologies, Inc.) and free cholesterol assay kit (E1016; Applygen Technologies, Inc.) following the manufacturer's instructions. Cholesteryl ester was quantified by subtracting the free cholesterol values from the total cholesterol value.

Human normal hepatocyte LO2 (ATCC) was cultured in Dulbecco's Modified Eagle's Medium (DMEM) containing 10% fetal bovine serum (FBS; Gibco), 1 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. Hepatocellular carcinoma cell line HepG2 (ATCC) was cultured in RPMI-1640 medium (Gibco, Carlsbad, CA, USA) supplemented with 10% FBS, 1 mM L-glutamine, and 100 U/ml penicillin, and 100 μg/ml streptomycin. Both LO2 and HepG2 were maintained at 37 °C in a humidified atmosphere of 5% CO2. For the dose-dependent assay, cells subcultured in the 6-well plate were treated with various concentrations of bilirubin (0, 3, 6 and 12 μM) for 24 h. For the time-dependent studies, cells were treated with 12 μM bilirubin for 0, 6, 12 and 24 h.

Total RNA from cells was extracted with TRIzol reagent (Invitrogen) and reverse-transcribed using PrimeScript™ RT reagent Kit with gDNA Eraser (Takara) according to the manufacturer’s instructions. Real-time PCR was performed on a StemOne™ Real-Tim PCR System (Applied Biosystem, USA) following TB Green™ Premix Ex Taq ™ (Tli RNaseH Plus, Takara) protocol. The primer pairs were used: 18S, forward 5′-CCCAGTAAGTGCGGGTCATAA-3′, reverse 5′- CCGAGGGCCTCACTAAACC-3′; mouse HMGCR forward 5′-TGATTGACCTTTCCAGAGCAAG-3′, reverse 5′- CTAAAATTGCCATTCCACGAGC-3′. Gene expression was calculated based on the −ΔΔCt method. The relative level of mRNA expression was determined using as reference the geometric mean of human/mice 18S (ThermoFisher Scientific).

Cells or tissues were collected and homogenized in ice-cold RIPA buffer. Lysates were maintained constant agitation for 2 h at 4 °C. Gently aspirate the supernatant to a prechilled tube after the centrifugation at 13,200 rpm for 10 min at 4 °C. The protein level of lysates was determined by the BCA protein assay. An aliquot of proteins was separated by SDS-PAGE and then transferred to the immunoblot PVDF membrane. The immunoblots were incubated with primary antibodies at 4 °C overnight, followed by HRP-conjugated secondary antibodies (1:5000) at RT for 1 h. The signals of targeted proteins were visualized by employing the ECL detection system (Amersham). Primary antibodies used for immunoblots were rabbit polyclonal antibodies against HMGCR (ab174830, diluted 1:1000) and mouse monoclonal antibody against GAPDH (ab8245, diluted 1:1000). All antibodies were purchased from Abcam.

Cells were collected and treated with IP lysis buffer containing protease inhibitor cocktails at 4 °C for 30 min. Cell lysates were centrifugated at 13,200 rpm for 10 min at 4 °C to remove cell aggregates. 50 μl of protein extract were dissolved in a DTT-containing loading buffer as an input sample. The remaining protein extract was adjusted to a concentration of 2 μg/μl of Co-IP dilution buffer. 500 μl of protein extract were immunoprecipitated with magnetic beads conjugated with indicated antibodies at RT for 1–2 h in a roller. The beads were washed with prechilled Co-IP washing buffer three times and then heated in 2 × loading buffer for 10 min at 70 °C. After centrifugation at 12,000 rpm for 10 min at 4 °C, the beads were collected by inserting the protein samples in a magnetic stack for 10 min at RT. The well-prepared protein samples were analyzed by western blot. The immunoblots were probed with rabbit polyclonal antibody against HMGCR (ab174830, diluted 1:1000), rabbit polyclonal antibody against ubiquitin (ab179434, diluted 1:5000), and mouse monoclonal antibody against GAPDH (ab8245, diluted 1:1000). All antibodies were purchased from Abcam.

To evaluate protein half-life, we treated LO2 and HepG2 cells with or without bilirubin (12 μM) in the presence of cycloheximide at a final concentration of 50 µg/ml to block protein synthesis for the indicated time (0, 6, 12, and 24 h). Protein samples were subjected to western blot to analyze the expression of HMGCR.

The peripheral blood was collected from the heart after the injection of pentobarbital sodium. After mice were sacrificed, the spleen was isolated and spleen cells suspension was separated by gently crushing the spleen. Cells from peripheral blood or spleen were fixed with Fix I buffer (Fluidigm) for 10 min at RT. After washing with cold PBS, red blood cell lysis buffer was used to remove erythrocytes from peripheral blood or spleen cells.

Peripheral blood or spleen cells were barcoded separately with premade combinations of six different palladium isotopes using the Cell-ID 20-Plex Pd Barcoding Kit (Fluidigm) for 30 min at RT. The samples were washed twice with cell staining buffer and then pooled together. The pooled samples were incubated with anti-CD16/32 (FcR III/II, Biolegend, CA, USA) for 10 min at RT to lower nonspecific antibody binding. These cells were washed twice with cell staining buffer and stained with a cocktail of 26 metal isotope-conjugated antibodies (Additional file 1: Table S1) for 30 min at RT. Stained cells were washed and stained with Intercalator-Ir (Fluidigm) at 4 °C overnight. After washing with cell staining buffer and ddH₂O, cells were resuspended in ultrapure water supplemented with EQ Four Element Calibration Beads (Fluidigm) and analyzed using Helios mass cytometer (Fluidigm).

The resulting flow cytometry standard (FCS) files generated by mass cytometry were normalized, randomized and debarcoded using the CyTOF software (Fluidigm). The debarcoded files were uploaded to cytobank.cn, an online platform for cytometry-based single cell analysis. The single cell data without debris and doublets were used for high-dimensional analyses, such as viSNE and FlowSOM.

The Shapiro–Wilk test was used to determine whether the data was normally distributed. A parametric test (Student’s t test) was used to determine the statistical significance between 2 groups if the data is normally distributed, otherwise, a nonparametric test (Mann–Whitney test) was used. Correlation analyses were performed using Person’s correlation coefficient. Error bars represent the mean ± standard deviation (SD). p < 0.05 was regarded as statistically significant.

To investigate whether bilirubin can protect atherogenesis in vivo, we used a murine model in which ApoE^−/− mice were fed on a western-type (high fat) diet. Mice received an intraperitoneal injection of bilirubin or vehicle every other day for 10 weeks. There was no significant difference between the average body weight of the two groups of mice throughout the 10 weeks of intervention. (Fig. 1A). In comparison with a vehicle control group, exogenous bilirubin administered resulted in a fourfold increase in total plasma bilirubin (Fig. 1B). Treatment with bilirubin reduced plasma glucose, total cholesterol, and LDL cholesterol levels in mice. (Fig. 1C and D), whereas both groups had similar levels of triglycerides and high-density lipoprotein (HDL) cholesterol. Furthermore, multiple correlations were found between glucose, total cholesterol, LDL, HDL, triglyceride, and total bilirubin. (Fig. 1E). For example, total bilirubin levels were negatively and significantly correlated with the levels of glucose and LDL levels in the peripheral blood. The LDL level was positively and significantly correlated with the total cholesterol level. Mice with higher levels of glucose also had higher levels of total cholesterol and LDL. There were significant and positive correlations between the concentration of triglyceride and total cholesterol or HDL (Fig. 1E). Considering the long-term exposure, we also investigated the bilirubin toxicity to the liver through histological and biochemical analysis. Bilirubin-treated mice displayed a downward trend in a lower hepatic enzyme activity (AST and ALT) than control mice (Additional file 1: Fig. S1B). But both groups showed no change in liver weight/body weight. (Additional file 1: Fig. S1A). These findings illustrate that a mildly elevated bilirubin reduced the risks of developing atherosclerosis without causing liver toxicity, suggesting a protective role of bilirubin in atherogenesis.

In vivo, bilirubin demonstrated a risk-lowering effect, thus we investigated whether it affects atherosclerotic lesion formation. Morphometric assays were used to measure the aortae and the aortic sinus lesion formation after 10 weeks of intervention. In both groups of mice, lesions were present throughout the aorta, but lesions were reduced in mice treated with bilirubin, as determined by en face ORO staining. (Fig. 2A). The bilirubin treatment significantly decreased the size and lipid load of the aortic sinus plaque and the aortic sinus lesion area. (Fig. 2B). Mice treated with bilirubin also showed reduced brachiocephalic arteries (Fig. 2C). Additionally, bilirubin treatment promoted collagen deposition (Fig. 2B) and α-SMA ( +) smooth muscle cell accumulation without affecting the distribution of Galectin-3( +) macrophages (Fig. 2D). The liver plays a crucial role in lipid metabolism. In the liver, cholesterol is derived from dietary, de novo synthesis, and transport from extrahepatic tissues. In mice fed a western diet for a long time, excessive lipid accumulates in the liver, increasing the risk of atherosclerosis and liver steatosis. H&E and ORO staining showed that bilirubin-treated mice had less hepatic steatosis than vehicle-treated mice (Fig. 2E). Next, The liver was then analyzed using an enzymatic method to determine the content of total cholesterol, cholesterol ester, and free cholesterol. The bilirubin treatment decreased liver total cholesterol and cholesterol ester concentration, but did not affect free cholesterol (Fig. 2F). As a result, bilirubin significantly inhibits the formation of atherosclerotic plaques and favors the phenotype of stable plaque.

The liver is the primary organ that synthesizes cholesterol and around 20–25% of total daily cholesterol production occurs here. Since HMGCR is the key rate-limiting enzyme for endogenous cholesterol synthesis, we examined HMGCR expression in vivo and in vitro. HMGCR expression in the liver was significantly reduced in ApoE^−/− mice treated with bilirubin compared to vehicle-treated mice (Fig. 3A and Additional file 1: S2A). In some cases, two bands of HMGCR have been detected in our and other studies [32, 33]. The upper one is the HMGCR band and the lower one is the non-specific band. The expression of HMGCR in the liver was also decreased in BALB/c mice treated with bilirubin for three consecutive days (Additional file 1: Fig. S2B). In two hepatic cell lines, bilirubin did not decrease the mRNA expression of HMGCR in vitro (Fig. 3B), but decreased the protein level of HMGCR in a time- and dose-dependent manner (Fig. 3C and D), suggesting that bilirubin may affect HMGCR expression at the translational or post-translational level. After treating two hepatic cell lines with cycloheximide (CHX), we found that bilirubin treatment decreased protein synthesis (Fig. 3E). Considering that HMGCR is degraded through ubiquitination at lysine 89 and 248 [34], we further evaluated the effect of bilirubin on the ubiquitination of HMGCR. Bilirubin increased HMGCR ubiquitination in HepG2 hepatic cells treated with MG132, a proteasome inhibitor (Fig. 3F). Furthermore, bortezomib, a potent protease inhibitor, reversed the decrease in HMGCR levels induced by bilirubin in two hepatic cell lines (Fig. 3G). These results indicate that bilirubin reduces HMGCR level mainly by promoting its ubiquitination and degradation, consequently reducing endogenous cholesterol synthesis.

Since bilirubin significantly improves atherosclerosis which is closely associated with the immune response, it is likely that bilirubin also has an effect on the immune system. Thus, we used a high-throughput single cell technology, mass cytometry [31, 35], to simultaneously measure the expression of 26 markers on immune cells from the spleen or peripheral blood at the single cell level. ViSNE analysis of CD45⁺ cells from the spleen was performed to visualize these high-dimensional mass cytometry data in two dimensions. According to the expression of 13 surface markers on the cells displayed on the viSNE map (Fig. 4A), 18 major immune cell populations were identified and gated (Fig. 4B). In each gated population, the expression of 13 markers was the same as the phenotypes of the indicated types of immune cells (Fig. 4C). Next, we compared the proportions of all these cell populations and found that after bilirubin treatment, the percentages of CD11b^low dendritic cells (CD11b^lowDCs) significantly increased, while the proportion CD11b^highDCs markedly decreased (Fig. 4D). In addition, the percentage of CD4 T cells in the spleen T cells was significantly reduced, while the proportion of CD8 T cells increased (Fig. 4D), suggesting that bilirubin induced a change in spleen T cell composition. Interestingly, the proportion of CD4 T cells among spleen T cells was strongly, significantly, and positively related to the LDL and (total cholesterol) TCHO concentration in the blood, while CD8T was strongly, significantly, and negatively related to the LDL and TCHO concentration in blood bilirubin (Fig. 4E), indicating that spleen CD4 T cells are positively associated with atherosclerosis. The proportion of CD11b^highDC that decreased by bilirubin showed a strong and negatively relationship with total bilirubin (Fig. 4E). Moreover, the proportion of CD8 T cells in spleen T cells, as well as the percentages of CD11b^lowDC, were significantly and negatively associated with the LDL and TCHO concentrations in the blood (Fig. 4E), implicating that these immune cells may favor the improvement of atherosclerosis.

T cells can influence the process of atherosclerosis by regulating their anti-inflammatory and pro-inflammatory functions and their functions are mainly determined by the direction of their differentiation. T cells are highly heterogeneous and contain various subtypes of differentiated T cells. Thus, we used 12 T cell-related markers to systematically analyze the spleen T cell. After viSNE analysis, heterogeneity in markers was assessed at the single-cell level. From the viSNE map of T cells, we also observed a huge heterogeneity of the T-cell compartments (Additional file 1: Fig. S3A). FlowSOM, an unsupervised cluster analysis for visualizing and interpreting cytometry data, was used on these T cells to map T cell phenotypes exhaustively and to further define any exclusive or significantly altered T cell clusters. 30 T cell clusters containing cells with similar phenotypes were identified and visualized on the viSNE map (Fig. 5A). The expression profiles of all the T cell clusters were visualized in a heatmap (Fig. 5B) and this approach led to the identification of 12 CD4⁺ clusters, one double-positive cluster, eight CD8⁺ clusters, five double-negative phenotypes, one natural killer (NK) T cluster and three TCRβ⁻ clusters (Fig. 5B). Among these 30 T cell clusters, proportions of T5 (CD44⁺CD62L⁻CXCR3⁺CD4T), which mainly contains effector memory (em) Th1 cells, and T16 (Ly6C⁺CD44⁺CD62L⁺CD8T), which mainly contains central memory Ly6C⁺CD8 T (Tcm) cells, were significantly increased after bilirubin treatment (Fig. 5C). In addition, these two clusters, together with T23 which was increased with a trend close to significance, had a mutual, significant, and negative relationship with the concentrations of LDL and TCHO in the blood (Fig. 5D and E). Furthermore, T6 and T13 proportions decreased with a trend close to significance after bilirubin treatment (Fig. 5C). Bilirubin did not affect the proportion of T1 (CD3⁺CD4⁺CD25⁺CD127⁻Ly6C⁺) and T2 (CD3⁺CD4⁺CD25⁺CD127⁻Ly6C⁻) clusters which mainly contain regulatory T (Treg) cells (Additional file 1: Fig. S3B). These data indicate that bilirubin also can regulate the composition of spleen T cells which are related to the improvement of atherosclerosis.

After bilirubin administration, the immune cells in the peripheral blood were also comprehensively analyzed using mass cytometry. viSNE was also used to identify the cell populations (Fig. 6A). 19 immune cell populations were obtained and the expression of their phenotypic markers was confirmed (Fig. 6B and C). The percentages of these 19 cell populations were then compared. Proportions of DC, Ly6C⁺NK, and all NK cells were significantly reduced in the peripheral blood of mice treated with bilirubin (Fig. 6D). The percentages of polymorphonuclear (PMN)-MDSC, neutrophils, and all MDSC significantly increased (Fig. 6D) and were strongly, significantly, and positively correlated with bilirubin concentration (Fig. 6E and F), implicating that bilirubin treatment may directly increase these cells. In addition, the proportion of all MDSC was significantly and negatively correlated with the concentrations of LDL and TCHO in the blood (Additional file 1: Fig. S4A). Considering that higher levels of LDL and TCHO are closely associated with atherogenesis, MDSCs increase by bilirubin may contribute to the improvement of atherogenesis though regulating the levels of LDL and TCHO. Moreover, proportions of Ly6C⁺NK and all NK cells showed a significant negative correlation with blood bilirubin level (Fig. 6G), suggesting that the reduction of these NK cells may be directly caused by bilirubin administration. The proportion of Ly6C⁺NK cells correlated positively and significantly correlated with the level of LDL in the blood (Additional file 1: Fig. S4B). Furthermore, B cells were negatively correlated with the concentration of bilirubin and LDL, and positively associated with TCHO concentration (Fig. 6F and Additional file 1: Fig. S4C). These data propose that bilirubin affects the MDSC and NK cells which are closely associated with the improved atherosclerotic blood lipid profiles.

Peripheral blood T cell composition after bilirubin treatment was also analyzed using viSNE (Additional file 1: Fig. S5A). 29 T cell clusters, including 13 CD4⁺ clusters, eight CD8⁺ clusters, four NKT clusters, two double-negative phenotypes, and two TCRβ⁻ clusters were identified (Fig. 7A and B). Treg cell clusters, T1 and T2, were not changed by bilirubin treatment (Additional file 1: Fig. S5B). T7 (CD44⁺CCR4⁺CD4 T), T9 (CD44⁺CCR4⁺CCR6⁺CD4 T), T13 (CD44⁻CD62L⁺CD4 T), and T17 (Ly6C⁺CD44⁻CD62L⁺CD8 T) were significantly decreased, whereas T20 (Ly6C⁻CD44⁻CD62L⁻CD8 T) was increased after bilirubin treatment (Fig. 7C). Additionally, proportions of two TCRβ⁻ clusters, T28 and T29, were also changed by bilirubin with a trend close to significance (Additional file 1: Fig. S5C). The percentages of T9 and T28 showed a negative and significant correlation with the total bilirubin level, whereas T20 and T29 showed a positive correlation with total bilirubin (Fig. 7D and E and Additional file 1: S5C). There was a positive and significant correlation between the proportions of T7, T9, T13, and T17 and the amount of LDL in the blood. (Fig. 7F). Except for T17, the other clusters were also positively correlated with TCHO concentration (Additional file 1: Fig. S5D). These results suggest that bilirubin can induce changes of T cell subclusters and that these changed T cells also have a strong correlation with the improvement of atherosclerosis.