Juvenile tilapia of uniform size were purchased from a local farm (Wuhan, Hubei Province, China). They were kept in a recyclable system for 2 weeks to acclimate the experimental conditions.

Three diets were produced with dietary phosphorus addition in the form of NaH₂PO₄·2H₂O at the inclusion levels of 0 (low-phosphorus diet, LPD), 0.53 g/100g (middle-phosphorus diet, MPD) and 1.05 g/100g (high-phosphorus diet, HPD) (Supplementary Table 2). The Pi supplemental levels were determined according to previous study (29, 30). When formulating the diets, dry feedstuffs were ground, and weighed and mixed for 20 min. Next, NaH₂PO₄·2H₂O was added, and they were mixed thoroughly for another 20 min. Then, we added the water and mixed them until a dough was formed. Finally, the dough was passed through the pelletizer with the 2·0-mm die in diameter. The diets were oven-dried until the moisture was about 10%. They were kept in the −20 °C. Final dietary phosphorus levels were measured, and the values were 1.21 g/100g (low-phosphorus diet, LPD), 1.75 g/100g (middle-phosphorus diet, MPD) and 2.66 g/100g (high-phosphorus diet, HPD).

After acclimating the experimental conditions, 225 tilapia (8.89 ± 0.01g, mean ± S.E.M) were randomly divided into nine circular fiberglass tanks (90 cm height, 80 cm diameter; 300 L water volume). Each experimental diet was assigned into three tanks randomly. The fish were fed to satiation twice a day (08:00 and 16:00, respectively) for 10 weeks. The experiment was conducted at natural photoperiod (approximately 12 h of light and 12 h of darkness). During the experiment, the parameters of the water quality in the tanks were followed below: water temperature 28.0–30.5°C, dissolved oxygen 6.10–6.52 mg/L, pH 7.05–7.59 and NH₄-N 0.05–0.08 mg/L. The water temperature was measured with thermometer. Dissolved oxygen was measured with portable DO instrument. pH was measured with portable pH meter, and NH₄-N with Nessler reagent spectrophotometry (GB/T 7479-1987). After the 10-week experiment, before collecting the samples, tilapia was fasted for 24 h. All sampled tilapia were anesthetized using MS-222 (100 mg/L water). Then, the liver tissues from ten fish were obtained from each tank, immediately frozen in liquid nitrogen and stored at −80°C for the RNA and protein isolation. The liver of another 3 fish was collected in each tank. They were fixed in 2.5% paraformaldehyde and 10% neutral buffered formalin for the histological, histochemical and ultrastructural analysis, respectively. Another 3 fish were collected and their livers were obtained for analysis of TG content and enzyme activities. The livers of another 6 fish were frozen in the liquid nitrogen and stored at −80°C for subsequent analysis.

Commercial kits were used to determine TG (A110-1-1, Nanjing Jiancheng Bioengineering Institute, Nanjing, China, detection limit: 0–9.04 mmol/L) and NEFA contents (A042-1-1, Nanjing Jiancheng Bioengineering Institute, Nanjing, China, detection limit: 0.01–2.0 mmol/L) according to the manufacturer's instructions, respectively. Phosphorus content of diets and the liver was determined using the molybdovanadate method (spectrophotometry) as described previously (31). Fatty acid synthase (FAS) activity was measured according to the method of Chang et al. (32) and Chakrabarty (33), 6-phosphogluconate dehydrogenase (6PGD) and glucose 6-phosphate dehydrogenase (G6PD) were determined by the method of Barroso et al. (34), malic enzyme (ME) by Ochoa [Ochoa (35)], isocitrate dehydrogenase (ICDH) according to the method of Bernt and Bergmeyer (36) and carnitine palmitoyltransferase 1 (CPT1) was analyzed using the method of Bieber and Fiol (37). The amount of enzyme was defined that converts 1 μM of the substrate to product per minute at 30 °C. The one unit of enzyme activity was expressed as units per milligram of soluble protein, and Bradford Protein Assay Kit (W042-1-1, Nanjing Jiancheng Bioengineering Institute, Nanjing, China, detection limit: 20–2000 μg/ml) was used to determine soluble protein concentration.

ORO staining was conducted according to Spisni et al. (38). H&E staining was conducted according to Woods and Ellis (39). In total 10 fields from each sample were examined randomly to quantify the relative areas of hepatic lipid droplets in ORO and vacuoles in H&E staining by the Image J software. TEM observation has been described by recent publications (40).

The qPCR assays were performed to quantify the mRNA expression of genes, based on these in our previous research (41). In brief, total RNA was isolated using in TRIzol^TM reagent (15596018, Invitrogen, USA), and transcribed into the cDNA with reverse transcription kit (4368813, Invitrogen, USA). qPCR assays were carried out in a 10 μL reaction system, 2 × SYBR^®Premix Ex Taq^TM (TaKaRa) 5 μL, 10 mM each of forward and reverse primers 0.2 μL, 0.6 μL diluted cDNA template and 4 μL double distilled H₂O. The specific primer sequences for the qPCR analysis are listed in Supplementary Table 3. Seven housekeeping genes (b2m, rpl7, hprt, β-actin, ubce, 18s rRNA and tuba) were selected to screen out two stable genes as the endogenous controls, based on the analysis of the geNorm. The methods 2^−ΔΔCt was used to calculate the relative mRNA expression of each gene.

Western blotting analysis was conducted to analyze the protein expression according to our recent publications (42). In brief, tilapia liver tissues and hepatocytes were washed in the 1 X PBS. They were placed in the RIPA lysis buffer (G3424, GBCBIO Technologies, Guangdong, China) and oscillated in physical and ultrasonic systems, respectively. Then, they were kept on the ice for 30 min and then centrifuged at 12000 rpm at 4°C for 10 min. We used the BCA assay (G3522-2, GBCBIO Technologies, Guangdong, China. The detection limit: 50–2000 μg/ml) to quantify protein. Next, 15% SDS–polyacrylamide gel was used to separate proteins (40 μg), which were transferred to PVDF membranes. Then, the blots were blocked with the 8% (w/v) skimmed milk in TBST for 2 h and washed thrice in the TBST buffer for 5 min each time. They were then incubated overnight with the specific primary antibodies at 4°C. These specific primary antibodies included the rabbit anti-AMPKα1(1:1000, ET1608-40, HUABIO, Hangzhou, China), anti-phospho-AMPKα1(1:500, ET1612-72, HUABIO, Hangzhou, China), anti-Beclin1 (1:2000, A7353, ABclonal, Wuhan, China), anti-phospho-Beclin1 (1:1000, AP1254, ABclonal, Wuhan, China), anti-SQSTM1/p62 (1:500, #5114, Cell Signaling Technology, Boston, USA), anti-LC3B (1:1000, ab51520, Abcam, MA, USA), anti-GAPDH (1:10000, #2118, Cell Signaling Technology, Boston, USA ), anti-GFP tag (1:2000, AE012, ABclonal, Wuhan, China) and 6 × His-tag (1:2000, 66005-1-Ig, Proteintech, Wuhan, China). Next, membranes were incubated with the corresponding secondary antibodies: the HRP-conjugated antirabbit IgG antibody (7074, Cell Signaling Technology, Boston, USA), HRP-294 conjugated mouse antirabbit IgG (5127, Cell Signaling Technology, Boston, USA), or HRP-conjugated antirabbit IgG antibody (light chain-specific) (93702, Cell Signaling Technology, Boston, USA). Finally, the membranes were visualized with the enhanced chemiluminescences and quantified by the Image-Pro Plus 6.0.

To explore the mechanism of dietary phosphorus on lipid metabolism, we isolated the primary hepatocytes from the tilapia and cultured them as described previously (43). Approximately 300 male tilapia were from the same original batch as used in Experiment I (8.89 ± 0.01g, mean ± S.E.M). We designed two treatments: the control (without extra Pi addition), 3.0 mmol/L Pi in the form of Na₂HPO₄·12H₂O/NaH₂PO₄·2H₂O. The Pi concentrations were used based on previous studies (44) and our pilot trials, which did not affect cell viability. We designed specific small interfering RNA (siRNA) for knock-downing the ampkα1 and beclin1 genes to determine the effect of the AMPK/Beclin1 pathway on lipid metabolism in tilapia. Tilapia hepatocytes were incubated with the inhibitor chloroquine (CQ) of the autophagy-lysosomal pathway (CQ; C6628, MilliporeSigma, MA, USA) to investigate whether and how autophagy mediated dietary phosphorus-induced lipolysis. The primary hepatocytes were incubated in the control or Pi group in the L-15 medium for 48 h with or without 2-h pre-incubation with 5μM CQ.

We used the 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2-H-tetrazolium bromide (MTT) assay to test the cell viability. The protocols were based on the publication (45).

To identify the phosphorylation sites of Beclin1 and AMPKα1, we constructed the vectors of Beclin1 and AMPKα1 expression according to our publication (46). The open reading frames (ORFs) of Beclin1 and AMPKα1 sequences were subcloned into the pcDNA3.1 (+) vector with the GFP-tag and the His-tag, respectively. They were named GFP-Beclin1 and His-AMPKα1, respectively. We produced the mutations of AMPKα1 serine (S) 496 to alanine (A) in the His-AMPKα1 plasmid and mutations of Beclin1 serine (S) 90 to alanine (A) in the GFP-Beclin1 plasmid by the Mut Express II Fast Mutagenesis Kit (C112-02, Vazyme, Piscataway, NJ, USA). The transient transfection of the plasmids into the HEK 293T cells was performed using Lipo293^TM transfection reagent (C0521, Beyotime, Shanghai, China). For Beclin1 and AMPKα1-knockdown experiment, the qPCR analysis was used to determine the knockdown efficiencies of these siRNA sequences. The siRNA sequences were given in Supplementary Table 3. Cell transfections was operated based on the protocols (46). Cells were treated for 48h to investigate the potential mechanism that AMPK and Beclin1 mediated phosphorus-induced lipolysis and lipophagy.

The Bodipy 493/503 staining was performed by our published methods (42). At first, tilapia hepatocytes were stained after Pi treatments, and then, flow cytometry or laser scanning confocal microscopy (Leica) were used to analyze and image, respectively. The microtubule-associated protein 1 light chain 3B (LC3B), Lyso-Tracker Red and LDs were stained to investigate their colocalization. Tilapia hepatocytes were fixed with 4% formaldehyde, and then the cells were blocked for 1 h in 5% BSA. Next, the hepatocytes were incubated with the anti-LC3B at 4°C overnight and then incubated with the goat anti-rabbit IgG H&L secondary antibody. The images were captured with the Leica laser scanning confocal microscope (Germany), and the fluorescence intensity was quantified by the software Image J.

The immunoprecipitation analysis was performed based on these protocols (47). The hepatocytes were lysed in the NP40 cell lysis buffer (p0013F, Beyotime, Shanghai, China) with phosphorylase inhibitor (P1082, Beyotime, Shanghai, China). Then, the anti-GFP tag or anti-His tag was added to the cell lysate overnight at 4 °C and the protein A/G beads (P2012, Beyotime, Shanghai, China) were added. Finally, the immunocomplexes were washed by NP40 buffer. The western blot was used to measure the protein expression.

The survival was not influenced significantly by dietary phosphorus (Supplementary Table 4). Weight gain (WG), specific growth rate (SGR), condition factor (CF) and feed intake (FI) increased, but feed conversion rate (FCR) reduced with increasing dietary phosphorus levels (P ≤ 0.05). Hepatosomatic index (HSI) showed no significant differences among the three treatments.

The total hepatic phosphorus content increased with dietary phosphorus levels (P ≤ 0.05) (Supplementary Figure 1). Compared to the LPD group, the HPD group tended to reduce hepatic TG content, as shown in ORO and H&E staining, and analysis of TG content (P ≤ 0.05) (Figures 1A–E). The HPD group also reduced the activities of lipogenic enzymes (FAS, G6PD, ICDH and ME) (P ≤ 0.05), but increased the activity of lipolytic enzyme CPT1 (P ≤ 0.05); the HPD group also reduced the mRNA expression of lipogenic genes (accα, g6pd, icdh, me and srebp-1) (P ≤ 0.05), but increased the mRNA expression of the lipolytic gene pparα (peroxisome proliferator-activated receptor α) (P ≤ 0.05) (Figures 1F,G). These results indicated that dietary phosphorus addition reduced TG content, suppressed lipogenesis and activated lipolysis in the liver of tilapia.

Since the study pointed out that autophagy regulates lipid metabolism (12), we determine effects of dietary phosphorus addition on the autophagic activity. Compared to the LPD group, HPD group increased autophagosome formation and reduced the number of LDs (Figure 2A). Compared to the LPD group, HPD group up-regulated the mRNA levels of beclin1, atg1a, atg1b, atg101, atg13, atg3, atg4b, atg4d, atg5, atg7, atg8a and atg8b (P ≤ 0.05) (Figure 2B), increased the protein levels of LC3B-II and the phosphorylation level of Beclin1 (P ≤ 0.05) (Figures 2C,D), and up-regulated the mRNA expression of many fatty acid β-oxidation genes (acads, acadm, acadvl, acadsb, acox1, acox3, hadh and hadhb) (P ≤ 0.05) (Figure 2E), and increased NEFA content (P ≤ 0.05) (Figure 2F). Thus, these results demonstrate that HPD enhanced autophagic activity and increased fatty acid β-oxidation.

To assess the effect of HPD on AMPK pathway, we analyzed the expression of the key genes and proteins related to the AMPK pathway (Figure 3A). Compared to the LPD group, the HPD group increased mRNA levels of the ampkα1, ampkα2, ampkβ1, ampkβ2, ampkγ1 and ampkγ2 (P ≤ 0.05) (Figure 3A), and increased the AMPKα1 phosphorylation level (P ≤ 0.05) (Figures 3B,C). These indicated that dietary phosphorus addition activated AMPK pathway in the liver of tilapia.

To elucidate the mechanisms of Pi influencing lipid metabolism of tilapia, tilapia hepatocytes were isolated. The MTT assay showed that Pi concentration of 0–3 mM did not adversely influence hepatocyte viability (P ≤ 0.05) (Supplementary Figure 2A). Thus, we chose 3 mM for our study. Compared with the control, Pi treatment significantly reduced TG content of hepatocytes, as shown by the analysis of TG contents, and by the flow cytometric analysis and the confocal microscopy after Bodipy 493/503 staining (P ≤ 0.05) (Supplementary Figures 2B–E). These data suggested that Pi incubation promoted lipid degradation of tilapia hepatocytes.

To determine whether Pi activated the AMPK pathway and accordingly induced autophagy, the effect of Pi on the AMPK pathway in tilapia hepatocytes was investigated. Compared to the control, Pi treatment up-regulated mRNA abundances of ampkα1, ampkα2, ampkβ1, ampkβ2, ampkγ1 and ampkγ2 (P ≤ 0.05) (Figure 4A), and increased the p-AMPKα1 protein levels (P ≤ 0.05) (Figures 4B,C). These indicated that Pi activated AMPK pathway and enhanced AMPK phosphorylation levels in tilapia hepatocytes.

In order to confirm whether Pi incubation activated autophagy, we determined the protein levels of several autophogic markers (Beclin1, p-Beclin1, LC3B-II and p62). Compared to the control, Pi incubation increased the protein level of the p-Beclin1 and LC3B-II, and reduced the p62 expression (P ≤ 0.05) (Figures 5A,B). Lyso-Tracker and acridine orange (AO) staining demonstrated that Pi increased the autolysosomal formation (red dots), indicating that Pi incubation activated autophagic flux (P ≤ 0.05) (Figures 5C–H).

To confirm whether phosphorus-induced autophagy was linked with the LD degradation, the double immunofluorescences were conducted. The LC3 puncta were linked with the autophagosome formation (48), and their colocalization demonstrated a direct association between LDs and autophagosomes (12). Herein, Pi incubation increased the amounts of LDs (green) bound to LC3 puncta (red) and autolysomes (red) (Figures 6A,B), indicating that phosphorus promoted the autophagosome formation, which helps sequester LDs for degradation. TEM observation confirmed that the LDs integrated with the autolysosomes after Pi incubation, indicating that Pi induced lipophagy (Figure 6C). Singh et al. (12) pointed out that the lipophagy promoted the lipolysis and FA β-oxidation. Accordingly, we tested mRNA expression of genes relevant to FA β-oxidation. Compared to the control, Pi incubation increased the mRNA levels of fatty acid β-oxidation genes (acadm, acadvl, acadl, acox3, hadh and hadhb) (P ≤ 0.05) (Figure 6D). Thus, our data suggested that Pi treatments activated the lipophagy and facilitated mitochondrial fatty acid β-oxidation, and according reduced TG deposition.

To address potential roles of autophagy in Pi-induced lipid degradation and metabolism, we first used CQ, a pharmacological inhibitor of autophagy. CQ pre-incubation alleviated the Pi-induced decrease of p62 protein expression and aggravated the Pi-induced increment of LC3B-II protein expression (P ≤ 0.05) (Figures 6E,F). Meanwhile, CQ pretreatment alleviated the Pi-induced decrease of TG content (P ≤ 0.05) (Figure 6G), and attenuated Pi-induced increment of NEFA content (P ≤ 0.05) (Figure 6H). Taken together, these results proved that lipophagy mediated the Pi-induced lipid degradation in tilapia hepatocytes, and thus showed a causal link between phosphorus-induced lipophagy and LD degradation.

We investigated the possibility of whether Beclin1 was required for Pi-induced lipophagy by knockdown experiment. The siRNA-774 was chosen from three sequences because of its capacity for inhibiting beclin1 expression (P ≤ 0.05) (Supplementary Figure 3). Compared to the control, flow cytometric analysis after LysoTracker and AO staining demonstrated that Beclin1-knockdown alleviated the Pi-induced increment of their fluorescence density (P ≤ 0.05) (Figures 7A–D). The confocal microscopic analysis indicated that Beclin1 knockdown alleviated the Pi-induced activation of autophagy (Figure 7E). Moreover, compared to the control, Beclin1 knockdown alleviated the Pi-induced increase of the LC3B-II and p-Beclin1 protein expression, and alleviated the Pi-induced decline of the p62 protein expression (P ≤ 0.05) (Figures 7F,G). Meantime, compared to the control, Beclin1 knockdown alleviated the Pi-induced reduction of TG content (P ≤ 0.05) (Figure 7H), which was confirmed by the flow cytometry and confocal microscopy after the Bodipy 493/503 staining (P ≤ 0.05) (Figures 7I–K). These findings indicated that Beclin1 was required for Pi-induced lipophagy and lipolysis.

To establish a direct link between AMPKα1 and Beclin1 in lipophagy, genetic inhibition of ampkα1 by siRNA was used. The siRNA-1374 was chosen from three sequences because of its capability for inhibiting the ampkα1 mRNA expression (P ≤ 0.05) (Supplementary Figure 4). Compared to the control, the ampkα1 knockdown inhibited AMPKα1 protein expression (P ≤ 0.05) (Supplementary Figures 5A,B). Moreover, flow cytometric analysis and confocal microscopy after Lyso-Tracker and AO staining demonstrated that AMPKα1 knockdown alleviated the Pi-induced increase of autophagic flux (P ≤ 0.05) (Figures 8A–E). AMPKα1 knockdown also alleviated Pi-induced increment of the protein expression of the LC3B-II and p-Beclin1, and alleviated the Pi-induced reduction of p62 protein expression. Compared to the control, AMPKα1 knockdown significantly reduced Beclin1 protein expression (P ≤ 0.05) (Figures 8F,G). AMPKα1 knockdown also alleviated the Pi-induced reduction of TG content (P ≤ 0.05) (Figure 8H). The results were further proved by the flow cytometric analysis and the confocal microscopic observation after Bodipy 493/503 staining (P ≤ 0.05) (Figures 8I–K). These indicated that AMPKα1 was required for Beclin1 phosphorylation in lipophagy.

Since HPD increased phosphorylation levels of Beclin1 and AMPKα1, further we investigated the interaction between the two proteins. We aligned the protein sequences of AMPKα1 and Beclin1 among the species, and found that Beclin1 serine 90 and AMPKα1 serine 496 were conserved evolutionarily from fish to mammals (Figures 9A,B). The overexpression plasmid of the tilapia beclin1 and ampkα1 genes were constructed and transfected into HEK 293T cells to investigate their interaction. We found that AMPKα1 coprecipitated with Beclin1 (Figure 9C). To further validate these findings, site mutations were introduced into the ampkα1 and beclin1 genes. The S496A mutants of AMPKα1 reduced the Beclin1 phosphorylation level, indicating that tilapia Beclin1 protein could be phosphorylated at the S90 site. Similarly, the S90A mutants of Beclin1 reduced the phosphorylation level of AMPKα1 (Figures 9D,E). Compared to the control, Pi incubation increased the coprecipitation between the Beclin1 and AMPKα1 (Figure 9F). Therefore, these results indicated that Beclin1 could be phosphorylated at the site of S90 by AMPKα1, which supported the hypothesis that AMPKα1-mediated Beclin1 phosphorylation activated lipophagy and promoted lipid degradation.

In summary, we identified an innovative mechanism of coordinated hepatic lipid metabolism mediated by phosphorus-induced lipophagy (Figure 10). HPD activated AMPKα1 via phosphorylating serine residue 496. The activated AMPK phosphorylated beclin1 on Ser90, and led to lipophagy and mitochondrial fatty acid β-oxidation, which promoted lipolysis and reduced lipid deposition.