Social jetlag elicits fatty liver via perturbed circulating prolactin rhythm-mediated circadian remodeling of hepatic lipid metabolism

The clinical cohort study was a case–control study, in which serum PRL levels of control and SJL subjects (n = 125 and 72, respectively) from Nanjing Drum Tower Hospital were analyzed. Liver specimens were collected for RNA-sequencing (RNA-seq). The detailed information can be found in Additional file 1: Methods. The protocol of the present study conformed to the guidelines of the Declaration of Helsinki and was approved by the Ethics Committee of Nanjing Drum Tower Hospital, Nanjing University Medical School (2021-388-01). All participants signed informed consent before study inclusion.

Female C57BL/6 J (n = 130) and Prl-knockout (Prl^−/−) mice (n = 15) were assigned to normal light cycle and jetlag conditions. All animal studies were conducted in adherence to the guidelines of the Animal Ethics Committee of Nanjing Drum Tower Hospital (2021AE01044). Histopathology, metabolic parameters, and gene expression analysis were performed as described in Additional file 1: Methods.

HepG2, MMQ, and 293 T cells were obtained from the Cell Bank of the Chinese Academy of Sciences. Primary hepatocytes were isolated from female wild-type C57BL/6 or Prl^−/− mice. Detailed methods of cell culture and molecular experiments are given in Additional file 1: Methods.

Human and mouse liver tissue were subjected to RNA-seq. The ZeitZeiger algorithm was applied to infer the circadian features of human liver gene expressions [14]. JTK_CYCLE was used to analyze the periodicity of mice liver gene expressions [15]. The details are provided in Additional file 1: Methods.

Clinical data were expressed with mean ± standard deviation (SD) for normally distributed continuous variables or median (interquartile range, IQR) for non-normally distributed variables, with normality assessed using Shapiro–Wilk test; categorical variables are expressed as n (%). For animal and cell experiments, data were shown as scatter dot plots with mean. Statistical comparisons employed Student’s t-test (parametric) or Mann–Whitney U test (non-parametric) for clinical continuous variables, χ² test or Fisher’s exact test for categorical data, while experimental data used unpaired two-tailed t-tests (two groups) or one-way ANOVA with Tukey’s post hoc test (multiple groups). Advanced analyses [correlation, principal component analysis (PCA), mediation] were performed after verifying model assumptions. All statistical analyses were conducted using SPSS 22.0 and R 4.2.1, with figures generated in GraphPad Prism 9. A two-sided P value < 0.05 was considered statistically significant.

With respect to the role of pituitary hormones in SJL-associated MASLD, serum PRL, ACTH, TSH, LH, FSH, and GH levels at 8:00, 16:00, and 24:00 were assessed in both the control and SJL groups. PRL secretion in the control group exhibited circadian fluctuations, with the nadir observed in the afternoon and the zenith at 24:00. However, PRL levels at the 3 time points in subjects with SJL ≥ 1 h were remarkably lower than in those with SJL < 1 h [7.48 (5.75, 9.23) μg/L vs. 11.28 (8.53, 14.59) μg/L, 6.74 (5.48, 8.80) μg/L vs. 9.84 (7.47, 14.07) μg/L, 6.82 (5.58, 7.93) μg/L vs. 11.29 (8.81, 16.17) μg/L; P < 0.001], while other pituitary hormones at the 3 time points showed no significant differences between the two groups (Additional file 1: Table S1). PRL levels at the 3 time points were inversely associated with the absolute value of SJL (r = −0.36, −0.32, −0.47, respectively; P < 0.05) and the histopathological degree of steatosis (r = −0.29, −0.34, −0.31, respectively; P < 0.05). Other hormones exhibited no correlation with either absolute SJL values or steatosis grade (Additional file 1: Fig. S1b). Circadian analysis was further implemented and revealed a significant reduction in the amplitude of PRL in subjects with SJL ≥ 1 h, whereas the amplitude and phase of ACTH, TSH, LH, FSH, and GH were not significantly different between the two groups (Additional file 1: Table S1). The sinus model of serum PRL in control and SJL subjects was fitted using the 3-time point rhythm prediction method with the values for amplitude, mesor, and phase. The equation was f(t) = 12.55 + 3.55 × sin(x + 0.06) in control subjects, and f(t) = 7.30 + 1.95 × sin(x − 0.12) in SJL subjects. PCA nalysis based on clinical variables separated MASLD patients from controls, with components 1 (PC1), 2 (PC2), and 3 (PC3) accounting for 77.3% of the variation between samples. PC1, which accounted for the greatest amount of variance (42.1%), included variables of PRL levels from 3-time points, liver enzymes, together with BMI and TG levels (Fig. 1d). Finally, to quantify the effect of PRL rhythm in SJL associated MASLD, mediation analysis was carried out. After adjusting for age, BMI, and HbA1c, mediation analysis showed that SJL-induced MASLD was partially mediated by decreased PRL amplitude, in which the total effect was 0.0022 (95% CI 0.0013–0.0031, P < 0.001), and the proportion of mediation effect was 28.98%, P < 0.001 (Fig. 1e). These results suggested that the altered PRL rhythm mediated the association between SJL and MASLD.

At 16 weeks, serum ALT, AST, and TG levels were remarkably elevated in the JL group compared to the NC group (P < 0.01, Fig. 2b). Liver magnetic resonance imaging (MRI) scanning documented a higher liver fat fraction in the JL group than the NC group (Fig. 2c). Substantiating these findings, Oil Red O staining, H&E staining, and electron microscopy confirmed greater lipid accumulation in hepatocytes of JL mice relative to NC mice (Fig. 2d). Circadian analysis revealed that in NC mice, serum PRL levels exhibited significant rhythmicity with a peak at ZT12, whereas in JL group mice, PRL levels were reduced at all time points, particularly at ZT12 (P < 0.001, Fig. 2e). Importantly, using the CircaCompare algorithm, rhythmicity test showed that in mice under NC conditions, serum PRL levels oscillated strikingly during 24 h (P < 0.05), however, this rhythmicity was significantly dampened in JL mice (P > 0.05; Additional file 1: Fig. S2e). These results suggested that jetlag induced fatty liver and disrupted the rhythmic secretion of PRL in mice. We also examined hepatic mRNA levels of key enzymes involved in hepatic lipogenesis such as Cidea, fatty acid synthase (Fasn), Pparγ, and Acc. We found that gene expressions of Cidea, Fasn, and Acc were remarkably increased in the liver of JL mice, while Pparγ expressions were not statistically different between the two groups (Additional file 1: Fig. S2f).

To disrupt endogenous PRL secretion, C57BL/6 mice under normal light cycle (NC) or jetlag shifts conditions (JL) were intraperitoneally injected with PRL inhibitor bromocriptine mesylate (BRC, 5 mg/kg) daily under (NC-BRC) and for 16 weeks. The results indicated that both NC-BRC mice and JL-BRC mice developed mild hepatic steatosis compared to NC mice, yet NC-BRC and JL-BRC mice showed similar extent of liver steatosis (Fig. 2f).

To further explore the role of PRL in maintaining liver homeostasis, global Prl^−/− mice were generated. Compared with wild-type mice, Prl^−/− mice exhibited increased glucose levels at 0, 30, and 60 min during GTT, and higher concentrations of glucose at 0, 15, and 120 min during ITT (Additional file 1: Fig. S2g). Then Prl^−/− mice were subjected to a normal light cycle (Prl^−/−-NC) or jetlag shifts (Prl^−/−-JL) for 16 weeks. Prl^−/− mice exhibited increased serum ALT, AST, and TG levels (Fig. 2g), higher liver fat content as assessed by MRI (Fig. 2h) and histological analysis (Fig. 2i), relative to wild-type mice under a normal light cycle (WT-NC). Meanwhile, Prl^−/−-JL mice exhibited similar serum ALT, AST, and TG levels (Fig. 2g), as well as hepatic lipid content relative to Prl^−/−-NC mice (Fig. 2h, i). These findings suggested that jetlad did not aggravate hepatic steatosis in the absence of PRL.

To delineate the binding site of RORα on the Prl promoter, an electrophoretic mobility shift assay (EMSA) was conducted using human nonfunctional pituitary adenoma, which was diagnosed based on normal hormone levels and negative immunohistochemical staining. Labeled PRL probe was used as negative control (lane 1), and specific binding of nuclear extract to the wild-type probe (RORE, 5’-AGGTCA-3’) was observed (lane 4). Moreover, a 100-fold excess of unlabeled wild-type probe competitively inhibited binding (lane 2), whereas unlabeled probes with mutated RORE failed to compete with this binding (lane 3) (Fig. 3f). Next, qRT-PCR analysis revealed that the circadian pattern of Rorα and Prl mRNA expression in the pituitary tissue of C57BL/6 mice synchronized, exhibiting a circadian surge at ZT12. However, jetlag significantly suppressed Rorα and Prl mRNA expression at ZT12 (Fig. 3g). Finally, we sought to evaluate the effect of SJL on the binding between RORα and Prl promoter region using ChIP-qPCR assay in NC and JL mice pituitary tissues. We observed that RORα was recruited to the Prl promoter at all time points in pituitary tissues from NC mice, however, a significant reduction in this recruitment was observed in the JL group at ZT12 (Fig. 3h). These data suggest that Prl expression is transcriptionally regulated by RORα, a process that is compromised by jetlag.

To comprehend the molecular profiles implicated in SJL-associated MASLD, we conducted RNA-seq analysis on liver tissues of 197 subjects. We employed two distinct strategies to analyze this dataset. The first strategy was to analyze differential expression genes (DEGs) of 197 subjects (125 control and 72 SJL subjects), then the DEGs were subjected to Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. The second strategy involved using the ZeitZeiger algorithm [14], a novel machine-learning tool, to predict the time of gene expression data in control and SJL subjects.

Next, the ZeitZeiger algorithm was employed to analyze RNA-seq data from 125 control and 72 SJL subjects. This machine learning approach derives predictors that map gene abundance to observed time, representing the real-time of receiving liver biopsy [14]. Subsequently, the liver gene expression of subjects in the control group was used as the training dataset to acquire predictors. To assess the accuracy of the predictors, the mean absolute error (MAE) was used, which represents the mean of the differences between the observed and predicted time of each sample. The MAE was 1.30 h in control subjects, indicating relatively accurate prediction performance in the training dataset (Fig. 4d). However, the MAE was 2.02 h in SJL subjects (testing dataset) (Fig. 4e), indicating that the predicted internal time of the gene expression pattern in the liver was different from the observed time in SJL subjects. Specifically, with respect to canonical genes involved in pathways of PRL signaling, circadian rhythm, and NAFLD, the fitted curve of the predicted time and the observed time demonstrated that the oscillation of gene expression in these pathways was altered in the liver of SJL subjects (Fig. 4f).

Subsequently, microarray data from liver tissues of male mice maintained in a normal light cycle were analyzed using the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database (GSE52333↗) [17]. Processed with the JTK algorithm, rhythmic genes were found to be enriched in the PRL signaling pathway, circadian rhythm, and lipid metabolism processes (including fatty acid biosynthesis, AMPK signaling pathway, peroxisome-proliferator-activated receptors signaling pathway) (P < 0.05; Additional file 1: Fig. S6). Additionally, we examined the rhythmicity of downstream pathways of other pituitary hormones in the liver from our data and GSE52333↗. The results indicated that the TSH receptor, LH receptor, FSH receptor, GH receptor, and glucocorticoid receptor pathways did not significantly oscillate in the liver in both datasets (Additional file 1: Figs. S5, S6).

Finally, we observed a substantial change in circadian gene expression profiles in JL mice livers compared to NC. A total of 884 genes exhibited oscillation exclusively in the liver of the JL group. Moreover, the circadian rhythmicity of genes involved in circadian rhythm, PRL signaling pathway, and lipid metabolism processes (including fatty acid biosynthesis, AMPK signaling pathway, peroxisome-proliferator-activated receptors signaling pathway) was significantly attenuated (Rhythmicity test P > 0.05), while genes associated with NAFLD gained oscillation and were significantly enriched (P = 0.01; Fig. 5a, Additional file 1: Fig. S7). These results confirmed the rhythmicity of genes involved in circadian rhythm, PRL signaling pathway, and lipid metabolism processes, while jetlag dampened cyclic levels of these genes and induced periodic expression of genes linked with MASLD.

To illustrate the association between the 3 common oscillating pathways mentioned under NC conditions, enriched genes from these transcripts (8 genes from the PRL signaling pathway, 13 genes from circadian rhythm, and 36 genes from lipid metabolism processes) were imported into the STRING database to obtain the protein–protein interaction (PPI) network. The PPI network, comprising 57 nodes and 272 edges, demonstrated that PRL signaling was highly interconnected with circadian rhythm and lipid metabolism (P < 1 × 10^–16). Next, these genes were imported into Cytoscape software and clustered into functional communities. Subsequently, 3 clusters of oscillating pathways were identified with a default cutoff on the closeness CytoHubba. The top 10 identified hub genes were Reverbα from circadian rhythm, and Fasn, acyl-coenzyme A oxidase 1 (Acox1), sterol regulatory element binding transcription factor 1 (Srebf1), and glucose-6-phosphatase catalytic (G6pc) from lipid metabolism, as well as serine/threonine kinase 1 (Akt1), cyclin D1 (Ccnd1), forkhead box O3 (Foxo3), signal transducer and activator of transcription 3 (Stat3), and mitogen-activated protein kinase14 (Mapk14) from the PRL signaling pathway (Fig. 5b).

Overall, the hepatic PRL signaling pathway played a crucial physiologic role in bridging circadian rhythm and lipid metabolism under normal light cycle; while under jetlag, genes involved in the “NAFLD” pathway were highly enriched.

As serum PRL levels were notably suppressed at ZT12, we further analyzed the RNA-seq data from the livers of NC and JL mice at ZT12. A total of 855 DEGs were identified, consisting of 562 up-regulated and 293 down-regulated genes between the NC and JL groups. With a significance level of P < 0.05, up-regulated DEGs were enriched in 18 KEGG pathways, including pathways related to NAFLD and circadian rhythm. Down-regulated DEGs were enriched in 45 pathways, including those related to the PRL signaling pathway and circadian rhythm (Additional file 1: Figs. S8, S9a, b). After that, the mechanisms involved in the desynchronization of hepatic PRL signaling in SJL-induced fatty liver were explored. Pathways of the PRL signaling pathway and circadian rhythm were significantly enriched in both human and mouse transcriptomes and were further analyzed. Overlapping genes that were commonly annotated in the PRL signaling pathway of both human and mouse datasets were identified: Stat3, Mapk12, Ccnd1, Tnfrsf11a, and Slc2a2 (Fig. 5c).

To elucidate the therapeutic impact of PRL rhythmicity on jetlag-induced hepatic steatosis, C57BL/6 mice were categorized into the normal light cycle (NC-vehicle) and jetlag shifts (JL-vehicle) groups for 16 weeks. Subsequently, mice in the JL group were intraperitoneally administered PRL (1 mg/kg) (JL-PRL) at ZT0 (JL-PRL0) or ZT12 (JL-PRL12) for 2 weeks. In GTT, glucose levels at each time point were significantly increased in JL mice compared to NC mice, in which glucose levels of JL mice at 0, 15, 30, and 120 min were significantly restored in JL-PRL0 mice, and were significantly restored in JL-PRL12 mice at each time point. At 15, 30, 60, and 120 min, JL mice that received PRL treatment at ZT12 exhibited remarkably decreased glucose levels than mice that received PRL treatment at ZT0. During ITT, JL mice displayed a marked increment of glucose levels across 5 time points than NC mice. PRL intervention at ZT0 significantly decreased glucose levels at 0, 60, and 120 min of JL mice, and PRL intervention at ZT12 significantly decreased glucose levels at 0, 30, 60, and 120 min of JL mice. Compared with JL-PRL0 mice, JL-PRL12 mice showed a pronounced reduction of glucose levels at 0 and 120 min (Additional file 1: Fig. S9d). At the end of the experiment, JL-PRL mice exhibited lower serum ALT, AST, and TG levels, as well as decreased hepatic fat deposition, as revealed by MRI, Oil Red O staining, H&E staining, and electron microscopy compared with the JL-vehicle group, and this effect was more pronounced when PRL was administrated at ZT12 (Fig. 6d–f). mRNA levels of lipogenic genes were detected and the results showed that decreased Ccnd1 mRNA levels in the liver of jetlag mice were up-regulated after PRL treatment, and a more obvious up-regulation was seen in JL-PRL12 compared to JL-PRL0. In addition, elevated expressions of Fasn, Acc, and Cidea in the liver of jetlag mice were markedly reversed in the JL-PRL group, in which PRL treatment at ZT12 decreased hepatic Acc, Fasn, and Cidea expressions to a greater extent compared with PRL treatment at ZT0. However, Pparγ levels were not significantly altered after PRL treatment (Fig. 6g, Additional file 1: Fig. S9e). Western blotting analysis also confirmed that down-regulated levels of CCND1, and up-regulated FASN and ACC protein levels in the liver of jetlag mice were reversed after PRL treatment, particularly at ZT12 (Fig. 6h).