Phosphorylation of 17β-hydroxysteroid dehydrogenase 13 at serine 33 attenuates nonalcoholic fatty liver disease in mice

Abstract

17β-hydroxysteroid dehydrogenase-13 is a hepatocyte-specific, lipid droplet-associated protein. A common loss-of-function variant of HSD17B13 (rs72613567: TA) protects patients against non-alcoholic fatty liver disease with underlying mechanism incompletely understood. In the present study, we identify the serine 33 of 17β-HSD13 as an evolutionally conserved PKA target site and its phosphorylation facilitates lipolysis by promoting its interaction with ATGL on lipid droplets. Targeted mutation of Ser33 to Ala (S33A) decreases ATGL-dependent lipolysis in cultured hepatocytes by reducing CGI-58-mediated ATGL activation. Importantly, a transgenic knock-in mouse strain carrying the HSD17B13 S33A mutation (HSD17B13^33A/A) spontaneously develops hepatic steatosis with reduced lipolysis and increased inflammation. Moreover, Hsd17B13^33A/A mice are more susceptible to high-fat diet-induced nonalcoholic steatohepatitis. Finally, we find reproterol, a potential 17β-HSD13 modulator and FDA-approved drug, confers a protection against nonalcoholic steatohepatitis via PKA-mediated Ser33 phosphorylation of 17β-HSD13. Therefore, targeting the Ser33 phosphorylation site could represent a potential approach to treat NASH.

Fig. 1. PKA phosphorylates 17β-HSD13 at serine 33 residue.

a The N terminal amino acid sequences of 17β-HSD13 from homo sapiens, rattus, mus, bos taurus and pongo abelii are aligned. A conserved serine 33 residue (Ser33) which is predicted to be a potential PKA phosphorylation site was identified among species. b HEK293 cells were transfected with a GFP-tagged 17β-HSD13 expression vector for 24 h, then treated with 10 μM forskolin for 30 min, followed by immunoprecipitation (IP) with anti-GFP-beads. The IP product was loaded and stained with Coomassie brilliant blue (CBB). c The major band in b was cut and digested. Phosphor-peptides were detected by LC-MS/MS. Only the serine 33 was found to be phosphorylated. Data were reproduced in three independent experiments. d, e HEK293 cells were transfected with an expression vector carrying a full-length myc-tagged 17β-HSD13 wild-type (WT), S33A mutant or S33E mutant, followed by IP with an anti-myc antibody. 10 μM forskolin (d) or 10 μM 6-BNZ-cAMP-AM (e) was added 30 min before lysis, respectively. Proteins retained on sepharose were blotted with an anti-phospho-(Ser/Thr) PKA substrate antibody. f HEK293 cells were transfected with a full-length myc-tagged 17β-HSD13 WT or S33A mutant, followed by IP with an anti-myc antibody. Proteins retained on sepharose were incubated with an active recombinant PKA Cα subunit for the kinase assay and then blotted with an anti-phospho-(Ser/Thr) PKA substrate antibody or anti-myc antibody. g, h HEK293 cells were transfected with a full-length GFP-tagged 17β-HSD13 WT, S33A or S61A mutant for 24 h and then pretreated with forskolin for 30 min with or without H89 for 1 h as indicated, followed by IP with an anti-GFP antibody. Proteins retained on sepharose were then blotted with an anti-phospho-(Ser/Thr) PKA substrate (p-PKA-sub). n = 2 biologically independent experiments for b–h.

Fig. 2. S33A mutant cells exhibit larger lipid droplets with reduced lipolysis and fatty acid oxidation.

a Huh7 cells were infected with lentiviruses carrying GFP, GFP-tagged 17β-HSD13 WT or GFP-tagged S33A mutant. Oil red O staining (scale bar, 50 μm) and confocal images (scale bar, 5 μm) were analyzed. Lipid droplets were stained by Nile red (red) and the nucleus was stained by DAPI (blue). LD size (b) and TG content (c) were analyzed. GFP: n = 4; WT: n  =  4; S33A: n  =  4 biologically independent cells. d N-SIM-3D reconstruction of the subcellular localization of 17β-HSD13 WT and 17β-HSD13 S33A protein in Huh7 cells. Lipid droplets were stained with Nile red (red) and 17β-HSD13 WT and 17β-HSD13 S33A proteins were visualized by green fluoresce. e, f Huh7 cells were stably infected with the lentiviruses carrying a full-length GFP (GFP), GFP-tagged 17β-HSD13 WT (WT) or GFP-tagged 17β-HSD13 S33A mutant (S33A). Mitochondrial respiration was analyzed in real-time using the Seahorse XF24 Extracellular Flux Analyzer. The oxygen consumption rate (OCR) (e) at different stages of respiration, basal and spare respiratory capacity (f) were measured as described in the “Methods” section. Data presented are the mean ± SEM. One-way ANOVA with Bonferroni post hoc analysis was performed. GFP: n = 5; WT: n = 4; S33A: n = 4 biologically independent cells. g–i Lipolysis was induced by forskolin treatment (10 μM) for 24 h in Huh7 cells with stable expression of GFP, 17β-HSD13 WT (WT) or GFP-tagged 17β-HSD13 S33A mutant (S33A). Oil red O staining was performed (g) and the LD sizes (h) were analyzed. i Glycerol levels in culture medium were measured. GFP: n = 5; WT: n = 4; S33A: n = 4 biologically independent cells. j–l Oxygen consumption rate (OCR) in the presence of exogenous palmitic acid in WT and the S33A mutant Huh7 cells. Advanced ORC was measured in real-time using the Seahorse® metabolic flux analyzer as described in the section of “Methods”. j The OCR curves. The times of addition of etomoxir (ETO), oligomycin (oligo), FCCP, and rotenone + antimycin A (Rot/AA) were indicated at the top. The OCRs in the presence of palmitate-BSA (PA) and ETO were displayed. k Acute response due to the addition of ETO under basal condition. l Maximal respiration due to the addition of FCCP. Medium-WT-PA: n = 5; Medium-S33A-PA: n = 4; ETO-WT-PA: n = 5; ETO-S33A-PA: n = 4 biologically independent cells. Data represent mean ± SEM; Two-tailed student’s t test was performed for b, c; One-way ANOVA with Bonferroni post hoc analysis was performed for e–l.

Fig. 3. Targeted knock-in of the 17β-HSD13 S33A mutation drives NAFLD pathogenesis.

a Schematic representation of the gene targeting strategy of 17β-HSD13 S33A. Homologous recombination was used to replace exon 1 with the corresponding mutant sequence, resulting in a substitution of alanine for serine 33 in 17β-HSD13. b Representative liver gross view, H&E staining, Oil red O staining, and immunohistochemistry (IHC) pictures in the Hsd17b13^33+/+(WT) and Hsd17b13^33A/A (S33A) fed a chow at the age of 5 months; Scale bar, 100μm. WT: n = 5; S33A: n = 5 biologically independent animals. c Transmission electron microscopy (TEM) analysis of lipid droplets (LDs) in the livers of WT and S33A mice. Nu nucleus; mi mitochondria; RER Rough endoplasmic reticulum. Scale bar = 2 μm (upper panels); 1 μm (lower panels); WT: n = 3; S33A: n = 3 biologically independent animals. Serum triglyceride (TG) and cholesterol (TC) levels (d), liver TG and TC contents (e), serum ALT and AST levels (f), serum insulin (g) and serum non-esterified fatty acid (NEFA) (h) were analyzed in the S33A mutant mice compared with their WT littermates at the age of 5 months. WT: n = 9; S33A: n = 10 biologically independent animals. Glucose tolerance test (GTT) (i), insulin tolerance test (ITT) (j), and pyruvate tolerance test (k) were performed at the age of 14–15 weeks. AUC, the area under the curve. WT: n = 5; S33A: n = 5 biologically independent animals. l Lipidomics analysis was performed in the livers from WT and the S33A mutant mice. The expression levels of various lipids species were displayed by heat map.WT: n = 6; S33A: n = 6 biologically independent animals. Data represent mean ± SEM; Significance was calculated by two-tailed student’s t test (d–h). Two-way ANOVA with Bonferroni post hoc analysis was performed for i–k.

Fig. 4. 17β-HSD13 physically interacts with ATGL on lipid droplet.

a Representative Nile red staining of purified LDs from WT and the S33A mutant mice. LD size distribution (b) and average LD size (c) were analyzed. WT: n = 5; S33A: n = 5 biologically independent animals. Quantitative RT-PCR analysis showing the expression of genes involved in lipid synthesis (d), lipid transport (e), and lipid oxidation (f) in the Hsd17b13^+/+ and Hsd17b13^33A/A mice. WT: n = 5; S33A: n = 5 biologically independent animals. g Immunoblot assay of hepatic ATGL, PNPLA3, and CGI-58 protein expression showed no difference between two genotypes. Quantitative results were shown below. WT: n = 5; S33A: n = 5 biologically independent animals. h Immunoblot assay of purified LDs from the livers of the Hsd17b13^+/+ and Hsd17b13^33A/A mice fed a chow diet (ND). A total of 2 μg of LD protein was immunoblotted for 17β-HSD13, ADRP, ATGL, PNPLA3, and CGI-58 protein expression. Silver staining was used as a loading control. Quantitative results were also shown. WT: n = 4; S33A: n = 4 biologically independent animals. i, j Co-IP experiments examining the interaction between 17β-HSD13 and ATGL. An expression vector carrying a GFP-tagged-17β-HSD13, myc-tagged 17β-HSD13 or mcherry-tagged ATGL (MC-ATGL) was transfected into 293T cells and cell lysates were immunoprecipitated using indicated beads, respectively. The immune precipitates were examined by immunoblotting using specific antibodies. k GST-pull down experiment showed 17β-HSD13 can directly bind to ATGL in vitro. Lysates of 293T cells overexpressed with mCherry-ATGL (MC-ATGL) (1 μg) were incubated with Glutathione-Sepharose beads precoated with equal amount (1 μg) of GST-17β-HSD13 or GST. The proteins pulled down by the GST-fusion proteins were analyzed by western blot. n = 2 biologically independent experiments for i–k. Data represent mean ± SEM; Two-way ANOVA with Bonferroni post hoc analysis was performed for b, (d–h), or two-tailed student’s t test (c).

Fig. 5. 17β-HSD13 S33A mutant enhances its interaction with ATGL.

a Huh7 cells were stably infected with the GFP-tagged 17β-HSD13 WT, GFP-tagged S33A mutant and MC-ATGL lentiviruses. 150μM oleic acid (OA) was loaded for 16 h. The cells were fixed and LDs were stained with Lipi-blue, an LD tracker. The yellow dots indicate co-localization (Scale bar = 5 μm). The zoomed pictures from each merged image were shown (Scale bar = 2 μm). n  = 2 biologically independent experiments. b The GFP-tagged 17β-HSD13 WT, S33A and MC-ATGL plasmids were transfected into 293T cells as indicated and cell lysates were immunoprecipitated using anti-GFP-beads. The immunoprecipitates were examined by immunoblotting using anti-GFP and mCherry antibodies. n = 2 biologically independent experiments. c The porcupine plots of PC1 of the 17β-HSD13 WT and 17β-HSD13 S33A system. ATGL was shown in green and 17β-HSD13^S33A shown in cyan. d Distances of center of mass along with the time traces of the simulations against the 17β-HSD13 WT and 17β-HSD13 S33A system are shown as black and red curves, respectively. e RMSD along with the time traces of the simulations against the 17β-HSD13 WT and 17β-HSD13 S33A system are shown as black and red curves, respectively. f The snapshots of the MD simulation of 17β-HSD13 WT-ATGL with the local structure around the S33 in 17β-HSD13 WT (left) and the A33 in 17β-HSD13 S33A (right) at 100 ns. g The expression vectors carrying Flag-CGI-58, MC-ATGL, GFP, GFP-tagged 17β-HSD13, and GFP-tagged 17β-HSD13 S33A were transfected into 293T cells, with or without forskolin treatment for 30 min. Cell lysates were immunoprecipitated using anti-flag beads. ATGL in the immune precipitates was examined by immunoblotting using an anti-mCherry antibody. GFP represents the protein levels of the 17β-HSD13 WT and S33A mutant. n = 3 biologically independent experiments. Data represent mean ± SEM; One-way ANOVA with Bonferroni post hoc analysis was performed for g.

Fig. 6. Targeted mutation of the Ser33 residue in 17β-HSD13 promotes NASH development.

Representative macroscopic view (a), liver gross view (b), H&E staining, Oil red O staining, and Masson’s staining (c) in the Hsd17b13^33+/+(WT) and Hsd17b13^33A/A (S33A) mice fed a high-fat diet for 3 months. Scale bar, 100μm. WT: n = 5; S33A: n = 5 biologically independent animals. d Liver TG and TC contents in HFD-treated WT and the S33A mutant mice. WT: n = 9; S33A: n = 10 biologically independent animals. The NAS score (e) and fibrotic area (%) (f) in WT and the S33A mutant mice. WT: n = 5; S33A: n = 5 biologically independent animals. Serum ALT and AST levels (g), and serum NEFA and TG levels (h) were increased in the S33A mutant mice compared with their WT littermates. WT: n = 9 for g, n = 7 for h; S33A: n = 10 for g, n = 7 for h biologically independent animals. i, j Representative immunohistochemical staining of inflammation markers (F4/80 and CD68) and fibrotic marker (α-SMA) (i). The semi-quantification data of F4/80 and CD68 (j). WT: n = 5; S33A: n = 5 biologically independent animals. k–n Global Hsd17b13 gene deficient mice (KO) were fed with a HFD for 3 months. At the end of second month with HFD, an AAV9 empty vector (AAV-vector, v), an AAV carrying a full-length of WT 17β-HSD13 (AAV-WT, WT) or the S33A mutant (AAV-S33A, S33A) was injected via the tail vein. Representative liver H&E staining, Oil red O staining and Masson’s staining were displayed (k). Serum ALT and AST levels (l), the NAS score (m) and hepatic fibrotic area (%) (n) of mice were analyzed. AAV-V: n = 5; AAV-WT: n = 5; AAV-S33A: n = 5 biologically independent animals. Data represent mean ± SEM; Two-tailed student’s t test was performed for d–h, (j); One-way ANOVA with Bonferroni post hoc analysis was performed for l–n.

Fig. 7. Primary hepatocytes of the Hsd17b13^33A/A mice are resistant to lipolysis via reducing ATGL function.

a Primary hepatocytes from the Hsd17b13^+/+ (WT) and Hsd17b13^33A/A (S33A) mice were isolated and stained for LDs with Oil Red O. At basal condition, the hepatocytes of the S33A mice exhibit larger LD size than that in the WT mice. Triglyceride contents (b) were higher, while the glycerol levels (c) were lower in culture medium of primary hepatocytes isolated from the S33A mice. WT: n =  3; S33A: n = 3 biologically independent cells. d Confocal micrographs showing the LDs (red) and nuclei (blue) in primary hepatocytes isolated from two genotypes treated with atglistatin, an ATGL inhibitor (ATGLi, 10 μM) and/or forskolin (10μM) for 24 h. e Changes of LD areas in WT and S33A mutant primary hepatocytes treated with atglistatin and/or forskolin. In WT cells, forskolin significantly reduced the LD area, which can be abolished by atglistatin treatment. In contrast, in the S33A mutant cells, both forskolin and atglistatin had little effect on the LD area. WT: n = 3; S33A: n = 3 biologically independent cells. Glycerol contents in culture medium (f) and triglyceride (TAG) contents in the primary hepatocytes (g) in d were examined after 24 h treatment. WT: n = 3; S33A: n = 3 biologically independent cells. Data represent mean ± SEM; Two-tailed student’s t test was performed for b, c or Two-way ANOVA with Bonferroni post hoc analysis was performed for e–g.

Fig. 8. Pharmacological induction of PKA-mediated phosphorylation of 17β-HSD13 at Ser33 protects against NASH.

a The chemical structure of reproterol. b Primary hepatocytes were cultured and treated with 10 μm reproterol (Rep) with or without the PKA inhibitor H89. Phosphorylated ATGL (S406), phosphorylated HSL (S660), and total HSL were immunoblotted. Forskolin (Fsk) was used as a positive control for PKA activation. c HEK293 cells were transfected with a full-length myc-tagged 17β-HSD13 WT, S33A mutant, and S33E mutant followed by IP with an anti-myc antibody. 10 μM reproterol or DMSO was added 30 min before cell lysis. Proteins retained on sepharose were blotted with an anti-phospho-(Ser/Thr) PKA substrate antibody. d Reproterol treatment increases the phosphorylation of 17β-HSD13 at serine 33 residue via PKA in vivo (see “Methods” for detail). Liver lysates were used for IP assay with an anti-myc antibody. Proteins retained on sepharose were blotted with an anti-phospho-(Ser/Thr) PKA substrate antibody. Red arrowheads indicate the phosphorylated 17β-HSD13 as a PKA-substrate. e–p Mice were fed with an HFD at 6 weeks of age for 16 weeks. Ten weeks after HFD treatment, mice began to receive reproterol treatment via intragastric administration at the dosage of 5 mg/kg body weight daily for 6 weeks. Body weight were recorded every week (e). HF-R, mice receiving both HF diet and reproterol; HF-Ctrl, mice only receiving HFD. HF-Ctrl: n = 5; HF-R: n = 5 biologically independent animals. Representative Oil red O, H&E, and Masson staining of the livers of HF-Ctrl and HF-R (f) and the NAS score (g) and fibrotic area (%) (h) in HF-Ctrl and HF-R group were shown. (scale bar, 50 μm) Liver TG (i) and TC content (j), body weight (k), and liver weight (l) in HF-Ctrl and HF-R group were measured. Expression of genes related with lipid synthesis (m), lipid transport (n), and lipid oxidation (o) was assessed. Protein expression levels of SREBP-1, SREPB-2, 17β-HSD13, ATGL, and β-actin in HF-Ctrl and HF-R groups were analyzed using immunoblot assay and quantitated (p). q Proposed mechanism by which 17β-HSD13 regulates LD lipolysis. Free fatty acids (FFAs) from extracellular and intracellular sources are packaged into triglycerides and stored in LDs. The S33-dephosphorylated 17β-HSD13 binds tightly to ATGL on the surface of LDs and sequesters CGI-58 to reduce ATGL lipolytic activity, leading to hepatocyte lipid accumulation. In contrast, upon PKA activation, the Ser33 residue of 17β-HSD13 is phosphorylated, which allows more physical interaction between ATGL and CGI-58 to increase ATGL activity, thereby reducing hepatocyte lipotoxicity. Data represent mean ± SEM; Significance was calculated by two-tailed student’s t test (g–l) or two-way ANOVA with Bonferroni post hoc analysis was performed for e, (m–p).