Pancreatic Sirtuin 3 Deficiency Promotes Hepatic Steatosis by Enhancing 5-Hydroxytryptamine Synthesis in Mice With Diet-Induced Obesity

Although nonalcoholic fatty liver disease (NAFLD) is prevalent in people with type 2 diabetes (T2D) (1–4), the exact mechanisms underpinning the association between NAFLD and T2D remain obscure. Insulin resistance is a prominent feature in many patients with NAFLD and T2D and is likely one of the important mechanisms linking NAFLD and T2D (5–7). In people with obesity and insulin resistance, excessive free fatty acids (FFA) are released from adipose tissue as a result of increased lipolysis. Increased FFA facilitates triglyceride (TG) synthesis in the hepatocytes to induce hepatic lipogenesis, causing steatosis, accelerating hepatic insulin resistance and T2D (8). Steatosis worsens hepatic insulin sensitivity, resulting in a vicious cycle that exacerbates dysregulation of glucose and lipid metabolism.

Molecules secreted from β-cells or hepatocytes under metabolic stress may circulate through the blood and modulate the function of each other. Bile acids serve as hormone-like molecules and convey the hepatic glucose sensing signals to maintain insulin secretory capacity of the β-cells (9). Hepatocytes from a liver insulin receptor knockout mouse secreted SERPINB1, a protease inhibitor, to cause adaptive β-cell proliferation (10). Pancreatic β-cell-derived miRNA29 was predominantly delivered to the liver, contributing to the development of insulin resistance (11). Likewise, Zinc ions released from the β-cell during stimulation with glucose have been suggested to influence hepatic insulin sensitivity (12). Based on these premises, it is plausible that cross talk exists between β-cells and hepatocytes through circulating biomarkers to explain the high prevalence of NAFLD in people with T2D.

We previously reported that sirtuin 3 (SIRT3), an NAD⁺-dependent deacetylase abundantly expressed in the mitochondria, protects β-cells from lipotoxicity by antagonizing oxidative stress by using Sirt3 global knockout mice (13). Here, we aimed to investigate the specific role of SIRT3 in the pancreatic β-cells by establishing pancreatic β-cell–selective Sirt3 knockout (Sirt3^f/f;Cre/+) mice. We also previously demonstrated that SIRT3 protein expression in mouse pancreatic islets was downregulated after 8 weeks of high-fat diet (HFD) feeding (13). Similarly, 16 weeks of HFD reduced SIRT3 activity in mice and increased mitochondrial protein oxidation in liver (14). These findings suggest that SIRT3 may act as a metabolic sensor to mediate cross talk between organs. Against this background, we further hypothesized that SIRT3 deficiency in β-cells may induce changes in liver by modulating the secretion of factors into the circulation.

Pancreatic β-cell–selective Sirt3 knockout (Ins-Cre^+/−; Sirt3^flox/flox [herein named Sirt3^f/f;Cre/+]) mice were generated by crossing of Sirt3^flox/flox (herein named Sirt3^f/f) mice (originally generated in J.A.’s laboratory [15]) with mice expressing Cre recombinase–driven rat insulin promoter (B6.Cg-Tg (Ins2-cre) 25Mgn/J; The Jackson Laboratory) and housed in pathogen-free conditions with a 12-h light-dark cycle. The male Sirt3^f/f;Cre/+ and Sirt3^f/f mice were fed with standard diet (STD) (Select, 50IF/6F; LabDiet) or HFD (60% cal from fat; Research Diets). Animal Experimental Ethics Committee of The Chinese University of Hong Kong approved the procedures (14/101/MIS and 19/245/MIS). Oral glucose tolerance tests (OGTT) and insulin tolerance tests (ITT) were conducted as previously described (13). The glucose and insulin dosages were 2 g/kg and 0.75 IU/kg body wt, respectively.

We dissolved Collagenase P (Roche, Switzerland) in Hanks’ balanced salt solution HBSS (Invitrogen) to obtain 0.8 mg/mL collagenase solution. Ice-cold collagenase solution (around 3 mL) was injected into the pancreas of anesthetized mice through the joint site of the hepatic duct and the cystic duct of the common bile duct. Pancreas was removed and infused in another 3 mL of 0.8 mg/mL ice-cold collagenase solution in a centrifuge tube. Pancreas was then transferred to a 37°C water bath and gently shaken (200 rpm) for 8 min. Pancreas was shaken vigorously for 10 s at the end of the digestion. Digested tissues were washed with 10 mL HBSS three times and filtered with 500 µm mesh. Tissues were resuspended with HBSS and gradient centrifuged with Histopaque 1119, 1083, and 1077 (Sigma-Aldrich). Islet layer was poured into a 70-µm cell strainer. Finally, the strainer was turned upside down and rinsed with RPMI medium on a petri dish for capture of the islets. Islets were cultured in RPMI medium supplemented with 10% FBS, as well as 100 units/mL penicillin and 0.1 mg/mL streptomycin (Invitrogen), for further treatment. Primary mouse hepatocytes were isolated as previously described (16) and cultured in Medium 199 with 100 units/mL penicillin and 0.1 mg/mL streptomycin (Invitrogen). MIN6 cells were cultured in a DMEM medium supplemented with 15% (v/v) FBS, 100 units/mL penicillin, 0.1 mg/mL streptomycin, and 110 μmol/L 2-mercaptoethanol (Invitrogen). HepG2 cells were cultured in an RPMI medium supplemented with 10% (v/v) FBS, 100 units/mL penicillin, and 0.1 mg/mL streptomycin. Palmitic acid (PA), BSA, N-acetyl cysteine (NAC), 5-hydroxytryptamine (5-HT), p-chlorophenylalanine (PCPA), and H89 were from Sigma-Aldrich. LP533401 (LP) was purchased from Cayman Chemical. Sarpogrelate hydrochloride was from Biochempartner (China).

Total RNA was isolated with an RNeasy Plus Mini kit (QIAGEN, Germany) according to the manufacturer’s instructions. The integrity of the total RNA was assessed with an Agilent 2100 Bioanalyzer System (Agilent Technologies) and an Agilent RNA 6000 Nano Kit (Agilent Technologies). Samples with an RNA integrity number >8 were selected for use. RNA sequencing (RNA-Seq) and data analysis were performed by Novogene Bioinformatic Technology (Beijing, China). Briefly, libraries were constructed using 1 μg total RNA and prepared with a TruSeq RNA Sample Prep kit (Illumina). Sequencing was performed with an Illumina NextSeq500 to generate 100 base pair paired-end reads. Quality trimmed reads were mapped onto the mouse genome (mm10) by TopHat2 (17), and gene expression in terms of fragments per kilobase of transcript per million mapped reads of genes overlapping with gene annotations in Ensembl release 96 was calculated by Cufflinks (18). Differentially expressed transcripts were detected with use of the R/Bioconductor package DESEq (19). HFD-fed Sirt3^f/f;Cre/+ mice (n = 3) were compared with HFD-fed Sirt3^f/f littermates (n = 3), and representative significantly less enriched gene sets with nominal P value of <0.05 are presented.

The glucose-stimulated insulin secretion (GSIS) study was similar to that previously described (13). Insulin level and glucagon level were measured with an ELISA kit (Millipore, Germany). 5-HT was analyzed by 5-HT ELISA kit (LDN, Germany).

RT-PCR was conducted with the SYBR Green kit (Promega) on Applied Biosystems 7900HT Fast Real-Time PCR System. Data were normalized to 18S rRNA or B-actin (Actb) in mice and Actb in human in each sample. Primer sequences are described in Supplementary Tables 4 and 5. Protein (10–30 μg) was loaded into each well of 10–15% SDS-PAGE gel. Primary antibodies rabbit anti-FASN (1:1,000), rabbit anti–histone H3 (1:1,000), and rabbit anti-GAPDH (1:5,000) were purchased from Cell Signaling Technology, and rabbit anti-SIRT3 (1:1,000) was purchased from Abcepta, while rabbit anti-H3K9Ac (1:2,000) and mouse anti-H4K16Ac (1:2,000) were purchased from PTM Biolabs and rabbit anti-SREBP1 (1:1,000) was purchased from Santa Cruz Biotechnology; horseradish peroxidase–linked anti-rabbit and anti-mouse IgG (1:2,000; Cell Signaling Technology) were used as a secondary antibody. Protein bands were developed by Immobilon Western Chemiluminescent HRP Substrate (Millipore).

We conducted tissue and cell immunofluorescence and immunohistochemistry staining as previously described (13). Guinea pig anti-insulin (1:700; Dako), rabbit anti-glucagon (1:300; Abcam, U.K.), rabbit anti–5-HT (1:4,000; ImmunoStar), and rabbit anti-SIRT3 (1:500; Abcepta) antibodies were applied to the fixed and embedded sections overnight at 4°C, followed by incubation with secondary antibodies. The β-cell area (%) was calculated as sum of insulin staining–positive area relative to total pancreas area.

Hematoxylin-eosin (H-E) solution (Sigma-Aldrich) was applied to the deparaffinized and rehydrated sections and incubated for 4 min, followed by rinsing of the slides in distilled (d)H₂O and application of acid alcohol to remove excess stain, incubation with Eosin Y solution (Sigma-Aldrich) for 2 min, and rinsing of dH₂O followed by dehydration in alcohol and mounting.

Immersed deparaffinized and rehydrated slides were treated with periodic acid solution (Sigma-Aldrich) for 5 min and rinsing of slides in dH₂O. The slides were immersed in Schiff’s Reagent (Sigma-Aldrich) for 15 min and washed in running tap water for 5 min, followed by counterstain of slides in H-E solution (Sigma-Aldrich) for 90 s. The slide was rinsed in running tap water followed by dehydration in alcohol and mounting.

After deparaffinization and rehydration, slides were placed in preheated (56–64°C) Bouin’s fluid (Sigma-Aldrich) for 60 min and cooled down for 10 min. This was followed by staining of mixed equal parts of Weigert’s iron hematoxylin solution A and B (Sigma-Aldrich) for 5 min and Biebrich scarlet-acid fuchsin solution (Sigma-Aldrich) for 15 min and differentiation in phosphomolybdic-phosphotungstic acid solution (Sigma-Aldrich) for ∼10–15 min until collagen was not red. Without rinsing, aniline blue solution was applied to slide for 5–10 min and 1% acetic acid solution (VWR, Radnor, PA) for 3–5 min, followed by rinsing of slide in running tap water, dehydration in alcohol, and mounting.

After being fixed in 2% paraformaldehyde for 20 min, primary mouse hepatocytes were added with BODIPY 493/503 (Thermo Fisher Scientific) and Hoechst 33342 (Thermo Fisher Scientific) working solution and incubated for 1 h at room temperature.

For knocking down mouse Sirt3, three specific shRNA (sh-Sirt3-1, 5′-CCCGTACCCTGAAGCCATCTTTGAACCCGTACCCTGAAGCCATCTTTGAA-3′; sh-Sirt3-2, 5′- TCGCTTTGGCAGATCTGCTACTCAT-3′; and sh-Sirt3-3, 5′-GGACACAAGAACTGCTGGATCTTAT-3′) lentiviruses targeting Sirt3 and a control lentivirus-expressed scrambled negative shRNA (5′-TTCTCCGAACGT GTCACGTAA-3′) were generated by Hanbio Biotechnology Co., Ltd. (Shanghai, China). The stable overexpression cell line establishment was similar to that previously described (13).

We used Triglyceride and Cholesterol LiquiColor kits (Stanbio) for serum TG and serum total and HDL cholesterol measurement. For serum AST activity and hepatic and cell TG measurement, we used assay kits from Cayman Chemical. For FFA measurement, we used an assay kit from Abcam.

Lysates from ∼3 × 10⁷ MIN6 cells were fixed with 1% formaldehyde. Three independent replicates of fixed cells per group were pooled and sonicated to an average length of 200–1,000 base pairs using a Vibra-Cell sonicator (Sonics & Materials). Chromatin was precleared with protein G magnetic beads (Bio-Rad Laboratories) for 1 h at 4°C and immunoprecipitated with antibodies against H3K9Ac and H4K16Ac from PTM Biolabs, or IgG from Cell Signaling Technology overnight at 4°C. Immune complexes were retrieved with use of protein G magnetic beads (Bio-Rad Laboratories) for 2 h at 4°C. The beads were subsequently washed and eluted prior to DNA purification using phenol:chloroform extraction. Chromatin immunoprecipitation (ChIP) and input DNA were quantified. DNA pellets were analyzed by RT-PCR by use of primers (described in Supplementary Table 6) directed to the Tph1 and Tph2 promoter.

Data are expressed as means ± SD. Differences between groups were analyzed by Student t test or one-way ANOVA followed by Bonferroni post hoc test where appropriate. Statistical testing was performed using GraphPad Prism 6 software (GraphPad Software). Differences with P values of <0.05 were considered statistically significant for all tests.

The data sets generated and analyzed during the current study are available from the corresponding author upon reasonable request. The RNA-Seq data have been deposited at the Gene Expression Omnibus (GEO) website: https://www.ncbi.nlm.nih.gov/geo/ (accession number GSE156345).

Tissue-selective deletion was validated with a minimal expression of SIRT3 protein in islets without affecting other tissues (Supplementary Fig. 1).

Both Sirt3^f/f;Cre/+ and Sirt3^f/f littermates, aged 4–6 weeks, were fed with STD and HFD for 36 weeks (Fig. 1A). Although fasting blood glucose (FBG) of HFD-fed Sirt3^f/f;Cre/+ mice was lower than that for HFD-fed Sirt3^f/f mice at the 16th, 20th, and 24th weeks, it started to ascend faster than in the HFD-fed Sirt3^f/f group after the 24th week (Supplementary Fig. 2A). No difference was found in FBG between STD-fed Sirt3^f/f and Sirt3^f/f;Cre/+ mice. Body weight increments were similar between Sirt3^f/f and Sirt3^f/f;Cre/+ mice, regardless of the types of diet (Supplementary Fig. 2B). Glucose tolerance was impaired and insulin level was lowered in Sirt3^f/f;Cre/+ mice compared with Sirt3^f/f mice upon HFD feeding but not upon STD feeding (Fig. 1B and C). ITT results showed better insulin sensitivity in the Sirt3^f/f;Cre/+ group under both diets (Fig. 1D).

We also found less adaptive hypertrophy of islets with SIRT3 deletion upon HFD feeding (Fig. 1E and F). In vitro experiments on islets isolated from STD-fed Sirt3^f/f;Cre/+ and Sirt3^f/f mice showed that PA further increased insulin secretion under high glucose conditions in islet from Sirt3^f/f mice, and this was inhibited in islets from Sirt3^f/f;Cre/+ mice (Fig. 1G).

Histological examination of the liver revealed that HFD-fed Sirt3^f/f;Cre/+ mice had more severe hepatic steatosis shown by H-E staining and more glycogen accumulation shown by periodic acid Schiff staining, without any obvious hepatic inflammation, fibrosis, or cirrhosis shown by H-E and trichrome staining (Fig. 2A and B). Adipose cell size in both inguinal and epididymal adipose tissue was increased slightly in HFD-fed Sirt3^f/f;Cre/+ compared with HFD-Sirt3^f/f mice (Supplementary Fig. 3). No significant difference was found in serum AST activity between HFD-fed Sirt3^f/f;Cre/+ and Sirt3^f/f mice (Fig. 2C).

The expressions of genes involved in lipogenesis, rather than genes for fatty acid uptake, fatty acid oxidation, and glucose metabolism, were generally increased in HFD-fed Sirt3^f/f;Cre/+ mice in comparison with HFD-fed Sirt3^f/f mice (Fig. 2D). Protein levels of full-length and mature SREBP1 ((flSREBP1 and mSREBP1, respectively) and fatty acid synthase (FASN) were elevated in HFD-fed Sirt3^f/f;Cre/+ mice compared with HFD-fed Sirt3^f/f mice (Fig. 2E).

To further explore the underlying mechanism for the regulatory role of SIRT3 in β-cells, we conducted RNA-Seq on the isolated islets from these four groups of mice. Three biological replicates were performed for each group, which exhibited good reproducibility (Supplementary Table 1). For visualization of transcriptomic differences, a heat map was generated showing log-twofold change >1 (Fig. 3A) (gene list shown in Supplementary File).

In comparison with findings in the STD-fed Sirt3^f/f group, 269 upregulated and 71 downregulated differentially expressed genes (DEGs) were identified in the STD-fed Sirt3^f/f;Cre/+ group (Fig. 3B). Similarly, 324 upregulated and 156 downregulated DEGs were identified in the HFD-fed Sirt3^f/f;Cre/+ compared with the HFD-fed Sirt3^f/f group (Fig. 3B and C). Nine DEGs in the STD-fed Sirt3^f/f and HFD-fed Sirt3^f/f groups did not overlap with aforementioned DEGs in the Sirt3^f/f;Cre/+ and Sirt3^f/f groups (Fig. 3B). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed that 264 of 324 upregulated DEGs in HFD-fed Sirt3^f/f;Cre/+ mice had defined functions. They were enriched in the categories including cytokine-cytokine receptor interaction, chemokine and tumor necrosis factor (TNF) signaling pathway, Jak-STAT signaling pathway, and tryptophan metabolism (Fig. 3D). These changes indicated activation of inflammatory changes and stress signaling pathways in β-cells due to SIRT3 deletion under HFD feeding in these mice. Notably, the rate-limiting enzymes for 5-HT synthesis, tryptophan hydroxylase 1 (Tph1), was the most significantly upregulated gene, followed by guanylate binding protein 11 (Gbp11)—which catalyze the conversion of GTP to GDP (Supplementary Table 2). Similarly, 150 of 156 downregulated DEGs in HFD-fed Sirt3^f/f;Cre/+ mice have defined functions. Among these, 67 (44%) were involved in biological processes including calcium signaling pathway, hormone secretion and synthesis, and cGMP-PKG signaling pathway (Fig. 3E). These changes suggested possible suppression of normal cellular function in β-cells due to SIRT3 deletion. Two of the 10 most significantly downregulated genes, Nrk and B3gat1, are catalytically active. Eight other genes were related to different biological function categories (Supplementary Table 3).

RNA-Seq indicated upregulation of tryptophan metabolism and 5-HT synthesis pathway in β-cells due to SIRT3 deletion (Fig. 4A). We validated the RNA expression of Tph1 and Tph2, which was upregulated ∼400-fold and 10-fold, respectively, in islets from Sirt3^f/f;Cre/+ mice compared with Sirt3^f/f mice (Fig. 4B). Immunostaining demonstrated that 5-HT was highly expressed in Sirt3 knockout β-cells (Fig. 4C). 5-HT content in islets was lowered by the TPH1 inhibitor LP in a concentration-dependent manner (Supplementary Fig. 4).

Although the role of SIRT3 is widely studied in mitochondria, several studies found its existence in nucleus. We confirmed that SIRT3 was localized to the nucleus of MIN6 cells, shown by immunofluorescence staining (Fig. 4D and E). In MIN6 cells, knockdown of Sirt3 increased, whereas overexpression of Sirt3 reduced two histone acetylation marks, acetyl-Lys 9 of histone H3 (H3K9Ac) and acetyl-Lys 16 of histone H4 (H4K16Ac). These suggested that SIRT3 might regulate gene transcriptions by deacetylating these histone marks as previously reported (20) (Fig. 4F). More importantly, the binding of H4K16Ac, rather than H3K9Ac (data not shown), to the promoter regions of the Tph1 and Tph2 genes was confirmed by ChIP. This was enhanced by Sirt3 downregulation and reduced by overexpression (Fig. 4G).

By searching Ensembl genome browser (https://asia.ensembl.org/ [accessed May 2020]), we identified binding motifs of CREBP and signal transducers and activators of transcription (STAT5) at the promoter regions of Tph1 and Tph2, as previously reported (1–3). We demonstrated binding of CREBP and STAT5 to both Tph1 and Tph2 promoters after Sirt3 knockdown in MIN6 cells (Fig. 4H).

Elevated pancreatic β-cell–derived 5-HT caused by SIRT3 deficiency may act as an endocrine signal circulating through the portal vein to the liver, resulting in hepatic lipid dysregulation. We detected the lipogenic effect of Sirt3^f/f;Cre/+ and Sirt3^f/f mice islets–derived conditioned medium to facilitate PA-induced lipid accumulation in mouse primary hepatocytes indicated by BODIPY 493/503, with pretreatment of LP to islets and posttreatment of LP to hepatocytes (Fig. 5A). To validate whether 5-HT exerts prosteatotic effect via 5-HT receptor 2A (HTR2A) as previously reported (21), we also applied selective HTR2A antagonist sarpogrelate to hepatocytes before adding the conditioned medium. With the exposure of conditioned medium from Sirt3^f/f;Cre/+ mice islets, lipid accumulation was significantly increased compared with that in the cells incubated with medium from Sirt3^f/f islets, and both were reversed by pretreatment with LP and sarpogrelate—rather than posttreatment of LP to only hepatocytes (Fig. 5B). These indicated that islet-derived 5-HT might be enhancing lipogenic pathways in hepatocytes, a key process during steatosis, via HTR2A.

The effect of 5-HT on both basal and PA-induced upregulation of lipogenic genes in HepG2 cells and primary mouse hepatocytes was studied. At the concentration of 10 μmol/L, most of the lipogenic genes were upregulated in the presence of PA, including the most important transcription factor for lipogenesis, Srebp1, and its downstream targets acetyl-CoA carboxylase (Acaca), fatty acid synthase (Fasn), stearoyl-CoA desaturase 1 (Scd), and glycerol-3-phosphate acyltransferase 1 (Gpam). Peroxisome proliferator–activated γ (Pparg), another key factor for lipogenesis and adipogenesis, was only upregulated by 5-HT in the absence of PA (Fig. 5C and Supplementary Fig. 5). Importantly, 5-HT also increased protein levels of SREBP1 and FASN in HepG2 (Fig. 5D). Sarpogrelate treatment counteracted the effect of 5-HT on lipid accumulation and lipogenic gene expression (Fig. 5E and F and Supplementary Fig. 6).

We performed intraperitoneal injection of 75 mg/kg body wt TPH1-specific inhibitor PCPA to HFD-fed Sirt3^f/f;Cre/+ and Sirt3^f/f mice for 4 weeks for further investigation of whether blocking 5-HT signaling ameliorates HFD-induced hepatic steatosis of Sirt3^f/f;Cre/+ mice.

As an endocrine signal, 5-HT should be detectable in the portal vein of HFD-fed Sirt3^f/f;Cre/+ mice. Indeed, 5-HT level was twofold higher in portal vein of HFD-fed Sirt3^f/f;Cre/+ mice compared with HFD-fed Sirt3^f/f mice (Fig. 6A). PCPA reduced 5-HT expression in islets and 5-HT serum concentration in heart and locally in the portal vein of both Sirt3^f/f and Sirt3^f/f;Cre/+ mice (Fig. 6A and B). Although PCPA-treated mice consumed slightly more food compared with the nontreated control group (Supplementary Fig. 7A), both the Sirt3^f/f and Sirt3^f/f;Cre/+ groups were more resistant to HFD-induced body weight increase (Supplementary Fig. 7B). Moreover, PCPA protected against HFD-induced hepatic steatosis in both groups as shown by liver weight measurement and H-E staining (Fig. 6C and D). Liver damage, indicated by AST level, was reduced slightly by PCPA in Sirt3^f/f;Cre/+ mice (Fig. 6E). Hepatic TG content was also reduced after PCPA treatment, indicating inhibition of lipid accumulation (Fig. 6F). This was confirmed by the reversal of the increase in lipogenic genes in the liver (Fig. 6G). Serum TG and total cholesterol level were slightly but not significantly reduced by PCPA in Sirt3^f/f;Cre/+ mice (Fig. 6H). Serum FFA level was higher in PCPA-treated Sirt3^f/f;Cre/+ mice, despite that fact that no significant difference was found between nontreated HFD-fed Sirt3^f/f and Sirt3^f/f;Cre/+ mice (Fig. 6H).

Although we did not find any improvement of glucose tolerance in PCPA-treated mice, they were more insulin sensitive than nontreated mice (Supplementary Fig. 7C–E). PCPA increased the FBG level of Sirt3^f/f;Cre/+ mice and reduced that of Sirt3^f/f mice (Supplementary Fig. 7F). A lower fasting glucagon level was found in HFD-fed Sirt3^f/f;Cre/+ mice compared with HFD-fed Sirt3^f/f mice, which was elevated by PCPA treatment (Supplementary Fig. 7G), implicating an additional role of PCPA in regulating glucagon level to modulate glucose homeostasis of Sirt3^f/f;Cre/+ mice and Sirt3^f/f mice.

The close associations between T2D and NAFLD suggest possible cross talk between β-cells and hepatocytes. For the first time, we demonstrate activation of 5-HT pathway associated with hepatic lipogenesis following selective deletion of SIRT3 in β-cells in mice under HFD conditions. Using MIN6 cells, we have demonstrated that knockdown of Sirt3 by shRNA (sh-Sirt3) increased, whereas overexpression of Sirt3 reduced H3K9Ac and H4K16Ac. We also confirmed the nuclear SIRT3 localization in MIN6 cells and the binding of H4K16Ac, CREBP, and STAT5 to the promoter regions of the Tph1 and Tph2 genes by ChIP in sh-Sirt3 MIN6 cells. Collectively, these data support that the SIRT3 protein might function in nucleus as histone deacetylase to cause changes in multiple genes implicated in hepatic lipogenesis via 5-HT pathway.

The location and precise roles of SIRT3 have been a matter of controversy. Early studies had suggested that SIRT3 resides in the mitochondria and functions as NAD⁺-dependent deacetylase (22–25). Scher et al. (20) first demonstrated that the nuclear flSIRT3 protein is a histone deacetylase, which is enzymatically active and deacetylates H3K9 and H4K16 to repress gene transcription. More recently, other researchers reported the role or mechanism of SIRT3 inside the nucleus (26–30).

5-HT is a monoamine neurotransmitter that modulates central and peripheral functions. 5-HT is synthesized from essential amino acid tryptophan and regulated by the enzyme TPH (31–33). The two distinct isoforms of TPH, TPH1, and TPH2 show mutually exclusive tissue expression patterns, with TPH1 expressed in peripheral nonneuronal tissues and TPH2 in the neurons of the central and enteric nervous system (34). Since 5-HT cannot cross the blood-brain barrier, centrally and peripherally produced 5-HT have distinct functions (35). Most of the peripheral 5-HT is synthesized by TPH1 in enterochromaffin cells of the gut. Once 5-HT is released into the blood circulation, the majority of 5-HT is taken up and sequestered into platelets, whereas the rest of 5-HT enters the systemic circulation and reaches peripheral tissues in free form (36).

In pancreatic β-cells, 5-HT has been known to reside within the insulin granule for more than four decades (37–39). Recent studies (40–43) have rekindled research in 5-HT by showing that, during conditions with heightened metabolic demands such as during pregnancy or HFD feeding, 5-HT production is increased in the islet, which may induce insulin secretion to maintain glucose homeostasis. 5-HT can also act as an autocrine signal to enhance β-cell function and increase β-cell mass during insulin-resistant states (38,39,44). 5-HT also acts as a paracrine signal to regulate the functions of other endocrine cells in the islet such as α-cells to produce glucagon (44,45). In mice, HFD feeding increased Tph1 expression, with increased 5-HT level in gut, accompanied by increased 5-HT concentration in the portal blood (21). Both gut-selective Tph1 knockout mice and liver-selective HTR2a knockout mice are resistant to HFD-induced hepatic steatosis. These results suggested a local role of 5-HT on hepatic lipid storage, without affecting systemic energy homeostasis (21). Recently, 5-HT was also found to induce lipogenesis in adipose tissues through HTR2A (46).

Sirt3^f/f;Cre/+ mice developed, besides impaired pancreatic β-cell function, more severe hepatic steatosis upon HFD feeding. Due to increased TPH1 expression in islet from Sirt3^f/f;Cre/+ mice and given the possible roles of 5-HT in lipid metabolism (47,48), we reasoned that 5-HT might be a key endocrine signal to mediate steatosis in HFD-fed Sirt3^f/f;Cre/+ mice. In support of this notion, we found twofold-higher 5-HT level in the portal vein of HFD-fed Sirt3^f/f;Cre/+ compared with Sirt3^f/f mice, suggesting that 5-HT might circulate to the liver and promote lipogenic gene expression. While gut-derived 5-HT might provide the basal value in portal vein, the difference between levels in HFD-fed Sirt3^f/f;Cre/+ and Sirt3^f/f mice likely reflected the β-cell–derived 5-HT, activated by Sirt3-selective knockout in β-cells. We also showed that treatment with PCPA, 5-HT receptor antagonist, reversed the phenotype. Collectively, these data support the notion of the endocrine effect of 5-HT in mediating signals from islet to liver in the cross talk between islet cells and hepatocytes.

We acknowledge that PCPA might also inhibit gut-derived 5-HT synthesis to arrest hepatic lipogenesis in HFD-fed Sirt3^f/f;Cre/+ mice. However, we have not managed to find a practical way to specifically inhibit 5-HT in β-cells. That said, the elevated amount of 5-HT derived from islets from Sirt3^f/f;Cre/+ compared with Sirt3^f/f mice suggested that more signals were coming from Sirt3^f/f;Cre/+ mice to worsen the severity of hepatic steatosis. After treatment with PCPA for 4 weeks, fat accumulation in the liver was largely eliminated in both Sirt3^f/f;Cre/+ and Sirt3^f/f mice. Our results suggest that PCPA might block 5-HT synthesis from β-cells to exert its antisteatosis role.

HFD caused metabolic stresses and induced hepatic steatosis either by increasing de novo lipogenesis and fatty acid uptake or by decreasing fatty acid oxidation (49). The alteration of lipid balance can also be induced by insulin resistance with upregulation of lipogenic gene expressions and increased FFA flux to promote lipogenesis in the liver (49). In our study, apart from hepatic steatosis, we also found enlarged adipose cell size in inguinal and epididymal adipose tissue from HFD-fed Sirt3^f/f;Cre/+ compared with HFD-fed Sirt3^f/f mice. We examined gene expression changes related to fatty acid intake, fatty acid oxidation, and glucose metabolism in mouse liver from Sirt3^f/f and Sirt3^f/f;Cre/+ mice, fed under STD and HFD conditions. We found that the genes related to de novo lipogenesis were activated to a larger extent compared with those related to fatty acid uptake, fatty acid oxidation, and glucose metabolism in the liver. These changes occurred without changes in body weight and FFA level in comparisons between HFD-fed Sirt3^f/f and Sirt3^f/f;Cre/+ mice. Four weeks’ treatment with PCPA caused weight loss in both groups, which possibly reflected the direct effect of 5-HT on adipose tissue as previously reported (46,48).

5-HT released from human pancreatic β-cells inhibits glucagon secretion from α-cells (45). Consistent with this finding, we also observed a lower fasting glucagon level in Sirt3^f/f;Cre/+ mice compared with Sirt3^f/f mice when fed with HFD, which might explain the lower FBG level in the former group. PCPA treatment did not ameliorate impairment of glucose tolerance of Sirt3^f/f;Cre/+ mice. While PCPA-treated Sirt3^f/f;Cre/+ mice showed increased FBG, Sirt3^f/f mice showed reduced FBG. It is possible that PCPA, by inhibiting 5-HT synthesis, may ameliorate the inhibitory effect of 5-HT on glucagon secretion in Sirt3^f/f;Cre/+ mice. This may also explain the low FBG level in Sirt3^f/f;Cre/+ mice in the presence of insulin deficiency. By contrast, PCPA reduced FBG in Sirt3^f/f mice by improving insulin sensitivity in Sirt3^f/f mice.

In conclusion, in the context of HFD feeding, SIRT3 deficiency in β-cells results in impaired adaptive hypertrophy of islets and GSIS. This is accompanied by upregulation of TPH1, which acts as an endocrine signal to cause hepatic steatosis (Fig. 7). These results provide novel insights regarding the complexity of islet function and the cross talk between islet and hepatocytes in energy homeostasis.

Funding. This study is supported by the General Research Fund of the Research Grant Council, the Hong Kong SAR Government (project reference: RGC ECS 24122318 and GRF 14109519). G.A.R. was supported by a Wellcome Trust Investigator Award (212625/Z/18/Z), MRC Programme grants (MR/R022259/1, MR/J0003042/1, MR/L020149/1) and Experimental Challenge Grant (DIVA, MR/L02036X/1), MRC (MR/N00275X/1), Diabetes UK (BDA/11/0004210, BDA/15/0005275, BDA 16/0005485), and Imperial Confidence in Concept (ICiC) grants.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.C.N.C., A.C.K.C., A.P.S.K., X.M., and X.Y.T. contributed to the study concept and design. A.P.S.K., X.M., G.A.R., and X.Y.T. contributed to the data analysis and the writing of the manuscript. X.M., D.M., H.C., B.F., W.K.K.W., C.C.H., and H.M.L. conducted the in vitro and in vivo studies. K.S. and J.A. contributed to the generation of the Sirt3^f/f;Cre/+ mouse model. All authors contributed to the critical review of the manuscript and approved the final version of the manuscript. X.M., X.Y.T., and A.P.S.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 21st Diabetes and Cardiovascular Risk Factors - East Meets West Symposium, Hong Kong, China, 28–29 September 2019 and at the 55th Annual Meeting of the European Association for the Study of Diabetes, 16–20 September 2019, Barcelona, Spain.