Serum biochemistry measurement kits (ALT, AST, ALP, and TBILI) were purchased from Shanghai Kehua Bio-Engineering Co., Ltd. (China). Serum TBA measurement kits were purchased from Nanjing Jiancheng Bioengineering Institute (China). All other solvents and regents were of analytical or HPLC grade when appropriate.

Vps33b hepatic knockout mice (Vps33b^flox/flox, alb-cre) and wild-type mice (Vps33b^flox/flox) with a C57BL/6 genetic background were obtained from Junling Liu’s laboratory (Shanghai Jiao Tong University School of Medicine, China). Mice were housed in stainless exhaust-ventilated closed-system cages in a specific-pathogen-free environment. Mice were maintained under a standard 12 h light/12 h dark cycle with water and a normal diet provided ad libitum. Animal experiments were performed in accordance with the guidelines of the Institutional Animal Care and Use Committee at Sun Yat-sen University (Guangzhou, China).

The Vps33b floxed allele was generated as previous reported (Wang et al., 2018). Mice were bred by crossing Vps33b^flox/flox, alb-cre and Vps33b^flox/flox mice. A 1–2 mm portion of mouse tail was cut for genotyping at the age of 3 weeks. A one-step mouse genotyping kit was purchased from Vazyme Biotech Co., Ltd (China). The supernatant of tail lysis buffer was added into the PCR system with 2x Taq Plus Master Mix (Dye Plus), primers and DEPC water. The PCR procedure were performed by agarose gel electrophoresis. Vps33b^flox/flox, alb-cre mice showed two bands at 300 and 606 bp, while Vps33b^flox/flox mice only had only a 606 bp band. Primer sequence of genotyping are listed as follows: Cre-reverse primer 5′-ATTTGCCTGCATTACCGGTCG-3′ and Cre-forward primer 5′-CAGCATTGCTGTCACTTGGTC-3′; Vps33b-floxed allele gene primer A1 5′-CTGACTAGGGGAGGAGTAGAAGGT-3′, A2 5′-GTATCACTGAGTCACACACATCCA-3′ and A3 5′-ATAGAGACGTTAGCAATTCGATCC-3′.

Vps33b^flox/flox, alb-cre mice were sacrificed at 3–4 months of age, and age-matched Vps33b^flox/flox mice were used as wild-type controls. All procedures were undertaken with the approval of the Institutional Animal Care and Use Committee at Sun Yat-sen University (Guangzhou, China). Serum was obtained after centrifugation of the blood at 3000 rpm for 10 min at RT. Bile was transferred into a 1.5 mL Eppendorf tube and weighed. Livers, intestines and feces were collected for further study respectively.

Serum activities of ALT, AST, ALP, TBILI, and TBA were measured by commercially available kits on an automatic biochemical analyzer.

Samples preparation for metabolomic analysis were performed according to our previously reported methods with some slight modifications (Chen P. et al., 2017; Zeng et al., 2017; Zhang et al., 2017b). Ten microliters serum and five microliters bile were mixed with 67% aqueous acetonitrile in distilled water to remove the protein, followed by centrifugation at 18000 g for 20 min at 4°C to obtain the supernatant. Liver and intestinal tissues were homogenized with 50% aqueous acetonitrile in distilled water and centrifuged at 18000 g for 20 min at 4°C to precipitate the protein. Twenty microliters of feces-PBS homogenate (20 μg per 400 μL) were vortexed with 67% ACN, followed by centrifugation to obtain the supernatant.

The obtained supernatant was transferred to an UPLC vial. Five microliter aliquots of metabolic samples were injected into the UPLC-ESI-Q Exactive interfered system (Dionex Corporation, Sunnyvale, CA, United States, Thermo Fisher Scientific, Waltham, MA, United States). ACQUITY UPLC BEH C18 column 1.7 μm (2.1^∗50 mm, Waters Corporation, Milford, MA, United States) was used to performed chromatography separation. Colum temperature was 60°C. The mobile phase consists of solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile) with the flow rate of 0.5 mL/min. The gradient program was as follows: 0 min 5% B, 0.5 min 5% B, 1 min 35% B, 6 min 60% B, 8 min 95% B, 9 min 95% B, 10 min 5% B, 10.5 min 5% B.

Electrospray negative ionization mode was used for analysis. The spray voltage was set to 3.5 kV. Capillary and auxiliary gas heater temperatures were set at 325 and 350°C respectively. Nitrogen was used as both the sheath gas at a flow rate of 60 arbitrary units and the auxiliary gas at a flow rate of 20 arbitrary units. Target SIM method which included the accurate Q/Z of bile acids was used to get higher sensitivity. The mass spectral data were aligned using SIEVE 2.2 (Thermo Fisher Scientific, Waltham, MA, United States). Then the extracted component was prepared for further analysis.

Multivariate data analysis was performed using SIMCA 13.0 software (Umetrics, Kinnelon, NJ, United States). Principal components analysis (PCA) and supervised orthogonal partial least squares discriminate analysis (OPLS-DA) models were used to analyze the data of tissues samples, such as the serum, liver, bile, intestine, and feces. Further identification of bile acids was conducted by comparing the retention time and fragmentation patterns with authentic standards according to our previously reported method (Chen P. et al., 2017).

Methyl tert butyl ether (MTBE) method (Matyash et al., 2008) was used to extract lipids from the serum and liver tissues. Twenty microliters serum was vortexed by adding prechill methanol, MTBE, and ultrapure water successively for 30 s respectively. Then the mixture was centrifuged at 3000 × rpm for 10 min. The supernatant was obtained for vacuum drying. Twenty micrograms liver tissues were homogenized with PBS and lipids were extracted as we mentioned before (Li et al., 2018). Before injection, samples were re-suspended in 200 μL mixture of methanol/isopropanol (1:1, v/v) and centrifuged at 18,000 × g for 5 min at 4°C.

Chromatography separation was performed using an Ascentis Express C18 2.7 μm column (100 mm × 2.1 mm, Sigma-Aldrich, St. Louis, MO, United States) on a Thermo Scientific Dionex Ultimate 3000 UPLC-ESI-Q Exactive system. The chromatograpy conditions, including the consistent of mobile phase and linear gradient, were similar with our previous report (Li et al., 2018). Mass spectrometry was performed with electrospray positive (ESI⁺) and negative (ESI^-) ionization modes. The main parameters for MS/MS included the following parameters: AGC target 1e⁵, maximum IT 65 ms, isolation window 1.2 m/z, normalized collision energy 25, 35 eV in positive mode, 20, 30, and 40 eV in negative mode, apex trigger 5–10 s, and dynamic exclusion 10.0 s. Ionization conditions were operated at a spray voltage of 3.5 kV and a capillary temperature of 300°C.

Lipidomic data processing was performed according to our previous report (Zhang et al., 2017a) using Lipid Search software (Thermo Scientific, San Jose, CA, United States).

RNA isolation and qRT-PCR analysis of hepatic gene mRNA expression level was performed as described previously (Chen P. et al., 2017). The primer sequences were obtained from Primer Bank and synthesized by Thermo Fisher Scientific. Primer sequences were listed in Supplementary Table 1.

Liver total protein extraction and western blot analysis were performed as described previously. Bolts were incubated with primary antibody against BSEP (F-6) (Santa Cruz Biotechnology, Santa Cruz, CA, United States), E-CADHERIN1 (Gentex Corporation, Zeeland) and GAPDH (Cell Signaling Technologies, Danvers, MA, United States).

Liver specimens were fixed in 10% formalin solution and processed routinely for paraffin embedding. Sections (4 μm thick) were deparaffinized and then stained with hematoxylin and eosin solutions (H&E) and examined under a light microscope (Nikon 80i, Japan). Immunohistological staining was performed with a primary antibody against CLAUDIN-1 (Abcam, San Francisco, CA, United States).

Each group was consisted of six animals. All values are expressed as the means ± SD. Statistical analysis was performed by unpaired Students’ t-test or Mann–Whitney U-test with Prism 6 (GraphPad Software Inc., San Diego, CA, United States) or SPSS statistical software. A p-value of less than 0.05 was considered statistically significant.

Compared to the age-matched male wild-type mice, hepatic Vps33b-depleted mice fed a controlled diet were identified for Vps33b depletion efficiency in the liver. Vps33b mRNA and protein levels were significantly decreased in Vps33b^flox/flox, alb-cre mice compared with those in the Vps33b^flox/flox mice (Supplementary Figures 1A–C).

Furthermore, no significant differences in body weight and liver/body weight ratio were observed between Vps33b^flox/flox and Vps33b^flox/flox, alb-cre mice (Supplementary Figures 2A,B). Although no significant differences in bile volume were observed between Vps33b^flox/flox and Vps33b^flox/flox, alb-cre mice, Vps33b^flox/flox, alb-cre mice displayed a distinctly smaller gall bladder. In addition, serum biochemistry analysis showed that ALP and serum total bilirubin levels in Vps33b^flox/flox, alb-cre mice were higher than those in Vps33b^flox/flox mice (Figure 1E,F). Liver damage indexes, such as serum ALT and AST levels, were also elevated in Vps33b^flox/flox, alb-cre mice compared with those in Vps33b^flox/flox mice (Figure 1). In addition, histological analysis showed hepatocyte degeneration, necrosis and ductal proliferation around the portal area in Vps33b^flox/flox, alb-cre mice (Figure 2). These results indicated that hepatic Vps33b knockout mice displayed cholestasis and slight liver injury compared to Vps33b^flox/flox mice.

Metabolomics analysis was performed to examine dynamic changes of endogenous metabolites in the serum, liver, bile, intestine and feces. Unsupervised PCA was used to analyze the data sets from Vps33b^flox/flox mice and Vps33b^flox/flox, alb-cre mice. Each point represented a mouse in their group. Significant separation of serum, liver and bile samples was observed in the PCA model, as shown by the distribution of Vps33b^flox/flox and Vps33b^flox/flox alb-cre mice in different quadrants, indicating a significant difference in endogenous metabolomes between the two groups of mice (Figure 3A–C). However, both the intestine and feces maintained a cross-distribution of metabolites in the PCA score plot (Figure 3D,E). These results were in accordance with the trend observed in the serum biochemistry. We further performed targeted metabolomics analysis of bile acids to gain a full understanding of how hepatic Vps33b depletion influences bile acid homeostasis. The extracted ions from mass spectrum of mice samples were confirmed by comparing with authentic bile acids standard according to retention time and MS/MS (Supplementary Figure 3). Bile acids were classified as conjugated and unconjugated types, and their relative amount in the tissues of mice was calculated (Figure 3F–O).

TCA and T-beta MCA consisted of a majority of conjugated bile acids in mouse tissue samples. Both were elevated in serum and reduced in bile by hepatic depletion of Vps33b. T-beta MCA was increased in the livers of Vps33b^flox/flox, alb-cre mice, while TCA was prone to reduction compared with those in the Vps33b^flox/flox mice. These major conjugated bile acid levels in the intestine were not changed by Vps33b deficiency in the liver. Overall, combined with other conjugated bile acids, such as TCDCA, TDCA and THDCA, conjugated bile acids were mostly increased in the serum of Vps33b^flox/flox, alb-cre mice, while decreased in the liver and bile compared with those in the Vps33b^flox/flox mice. However, unconjugated bile acids were generally reduced in the serum, liver and bile of Vps33b^flox/flox, alb-cre mice. Furthermore, hepatic Vps33b depletion seemed to have no effect on bile acid distribution between the intestine and feces (Supplementary Figure 4). In summary, the bile acid distribution pattern was clearly changed in hepatic Vps33b-depleted mice compared with that in the Vps33b^flox/flox mice, demonstrating that Vps33b plays an important role in sustaining bile acid homeostasis.

According to the disrupted bile acid metabolomics pattern in Vps33b^flox/flox, alb-cre mice, we detected several bile acid homeostasis-related gene expression levels in the liver (Figure 4). Cyp7a1 converts cholesterol to 7a-hydroxycheolesterol, which is the rate-limiting enzyme of bile acid formation in the liver. The mRNA expression level of Cyp7a1 was downregulated in Vps33b^flox/flox, alb-cre mice compared to that in Vps33b^flox/flox mice. However, Cyp450s enzymes, such as Cyp2b10 and Cyp3a11, which function in the formation of hydrophilic bile acids, were significantly upregulated in the Vps33b^flox/flox, alb-cre mice.

In addition, mRNA expression of hepatic canalicular membrane transporters (Bsep and Mdr2) were upregulated in Vps33b^flox/flox, alb-cre mice, while Abcg5/8 expression was downregulated compared to those in Vps33b^flox/flox mice. Another canalicular transporter, Mrp2, was unchanged in Vps33b^flox/flox, alb-cre mice. In addition, sinusoidal membrane multidrug resistance protein bile acid efflux transporters (Mrp3 and Mrp4) were not significantly regulated by hepatic Vps33b depletion. However, bile acid reuptake transporters (Oatp1b1 and Ntcp) were significantly downregulated in Vps33b^flox/flox, alb-cre mice (Figure 4A).

Vps33b function was related to hepatocyte polarity, which is vital for cell junction formation and specific protein localization. Furthermore, we performed immunohistochemistry staining of the bile acid export pump BSEP and measured the levels of representative cell junction proteins (E-CADHERIN1, CLAUDIN-1). As shown in Figure 4A,B, BSEP mRNA and protein levels were significantly increased in Vps33b^flox/flox, alb-cre mice compared to those in Vps33b^flox/flox mice. Moreover, BSEP distribution features were subdued in Vps33b^flox/flox, alb-cre mice, which were no longer localized along the specific side of hepatocytes. Livers express different kinds of cell tight junction proteins, and we chose E-CADHERIN1 and CLAUDIN-1 for detection. Increased E-CADHERIN1 levels and decreased CLAUDIN-1 levels were observed in Vps33b^flox/flox, alb-cre mice compared with those in the Vps33b^flox/flox mice (Figure 4B,D). Collectively, these data indicated that Vps33b^flox/flox, alb-cre mice displayed disrupted bile acid homeostasis in accordance with an altered targeted-bile acid metabolomics pattern compared to that in Vps33b^flox/flox mice. Irregular BSEP localization and altered cell tight junction expression suggested impaired hepatocyte polarity caused by Vps33b hepatic depletion.

Because lipid metabolism is integrally connected to bile acid metabolism, we speculated whether the lipid metabolite pattern was also affected by hepatic Vps33b deletion. Upon UPLC-ESI-HRMS-based lipidomics analysis, we measured serum and liver samples of Vps33b^flox/flox, alb-cre mice and Vps33b^flox/flox mice to reveal the lipid profiles within specific lipid species.

Principal component analysis of serum and livers showed a clear difference between Vps33b^flox/flox and Vps33b^flox/flox, alb-cre mice (Figure 5A,E). Liver scatter plots for Vps33b^flox/flox mice were in the second quadrants, while scatter plots for Vps33b^flox/flox, alb-cre mice were mainly located in the third and fourth quadrants (Figure 5E). To identify the lipids that contribute to the unambiguous separation between Vps33b^flox/flox, alb-cre mice and Vps33b^flox/flox mice, Lipid Search software was used to identify lipid molecular species based on the accurate mass values of each ion and MS2 pattern. The outstanding matched mass spectrums were shown in Supplementary Figure 5. An OPLS-DA score plot (Figure 5B,F) and s-plot (Figure 5C,G) was performed to screen the specific lipids altered in Vps33b^flox/flox, alb-cre mice among all identified lipid molecular species. VIP values > 0.8 and absolute p(corr) value > 0.6 are highlighted in green and red, respectively, in the s-plot (Figure 5D,H). Using the FDR (false discovery rate) test, screened specific lipid species of serum and livers with p-values < 0.05 are shown in the bar graph (Figure 6, 7). It is notable that many lipid species were altered in the serum and liver of hepatic Vps33b-depleted mice according to normalized heatmap pictures (Supplementary Figures 6, 7). TGs were the most abundant lipid species that were altered significantly in serum. Interestingly, regardless of the number of carbons or double bonds in the fatty acid chain, all species of TGs were significantly increased in serum but decreased in the liver of Vps33b^flox/flox, alb-cre mice compared with those in the Vps33b^flox/flox mice (Figure 6H–J, 7I). SMs were another type of lipid that exhibited a similar altered trend as TGs, in which the number of SM lipids was apparently less than that of TGs (Figure 6B, 7D). In addition, hepatic Vps33b depletion induced an upregulation of ceramides and a downregulation of PE in both serum and liver compared with those of Vps33b^flox/flox mice (Figure 6A,E, 7E). Lipids, such as PI, PS and LPC, were significantly increased in the serum of Vps33b^flox/flox, alb-cre mice (Figure 6C,D,F). However, some lipids were significantly altered with a different variation trend in Vps33b^flox/flox, alb-cre mice, such as PC in serum and PI, PC, and PS in livers (Figure 6G, 7C,F,H). Moreover, PA and CL were increased, and PG was decreased in the livers of Vps33b^flox/flox, alb-cre mice, which were not significantly changed in serum compared with those in the Vps33b^flox/flox mice.

In addition, we detected the expression of lipogenesis-related genes and specific lipid metabolism-related genes. As shown in Figure 8, fatty acid uptake genes were significantly decreased in the livers of Vps33b^flox/flox, alb-cre mice compared with those in the Vps33b^flox/flox mice, while FA synthesis-related genes displayed an inconsistent change, as Fas and Acc1 increased and Scd1 decreased. TG metabolism-related genes were reduced, especially Pnpla2, in accordance with the decreased level of TG in the livers of Vps33b^flox/flox, alb-cre mice. Meanwhile, CL-related genes were decreased as their amount was also lower in the livers of Vps33b^flox/flox, alb-cre mice compared with those in the Vps33b^flox/flox mice. Lipidomics analysis indicated that hepatic Vps33b regulated the lipid pattern of serum and livers in mice. Without Vps33b expression in the liver, the lipid species were distinct from Vps33b^flox/flox mice, which might contribute to the pathology of cholestasis.