The hepatic integrated stress response suppresses the somatotroph axis to control liver damage in nonalcoholic fatty liver disease

Abstract

Nonalcoholic fatty liver disease (NAFLD) can be ameliorated by calorie restriction, which leads to the suppressed somatotroph axis. Paradoxically, the suppressed somatotroph axis is associated with patients with NAFLD and is correlated with the severity of fibrosis. How the somatotroph axis becomes dysregulated and whether the repressed somatotroph axis impacts liver damage during the progression of NAFLD are unclear. Here, we identify a regulatory branch of the hepatic integrated stress response (ISR), which represses the somatotroph axis in hepatocytes through ATF3, resulting in enhanced cell survival and reduced cell proliferation. In mouse models of NAFLD, the ISR represses the somatotroph axis, leading to reduced apoptosis and inflammation but decreased hepatocyte proliferation and exacerbated fibrosis in the liver. NAD⁺ repletion reduces the ISR, rescues the dysregulated somatotroph axis, and alleviates NAFLD. These results establish that the hepatic ISR suppresses the somatotroph axis to control cell fate decisions and liver damage in NAFLD.

Keywords: ATF3; CP: Metabolism; ER stress; IGF-1; SIRT7; aging; integrated stress response; nonalcoholic fatty liver disease; nonalcoholic steatohepatitis; sirtuin; somatotroph axis.

Figure 1.. A mouse model of NAFLD with the suppressed somatotroph axis

(A) Single-cell RNA sequencing of the livers of WT and SIRT7^−/− mice using the 10x Genomics Chromium platform. Uniform manifold approximation and projection (UMAP) clustering of single-cell transcriptomes (3,270 cells from WT and 8,340 cells from SIRT7^−/− mice) colored by cell type. n = 3 mice. (B and C) Pathway analysis for the biological function of differentially expressed genes in hepatocyte 1 (pericentral) and hepatocyte 2 (periportal) of the livers of WT and SIRT7^−/− mice. n = 3 mice. (D–I) Quantitative real-time PCR analyses for the mRNA levels of the indicated genes in the livers of SIRT7^−/− mice and wild-type controls. GAPDH was used as an internal control. n = 9–13 mice. (J and K) Violin plots comparing log-normalized expression values of IGF-1 in hepatocyte 1 (pericentral) and hepatocyte 2 (periportal) in the livers of WT and SIRT7^−/− mice. Each dot represents the gene expression levels in one cell. Wilcoxon rank-sum test. n = 3 mice. (L) ELISA quantification of plasma levels of IGF-1 in SIRT7^−/− mice and wild-type controls. n = 8 mice. (M and N) Western analyses (M) and quantification (N) of phosphorylated Akt in the livers of SIRT7^−/− mice and wild-type controls. n = 3 mice. Error bars represent standard errors. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S1 and S2.

Figure 2.. Hepatic ER stress suppresses the somatotroph axis autonomously

Comparison of wild-type and SIRT7^−/− mice with or without Myc knockdown mediated by AAV8-mediated gene delivery. Mice were analyzed 4 weeks after viral infection. (A and B) Western analyses (A) and quantification (B) for phosphorylated eIF2α in the livers. n = 3 mice. (C–G) Quantitative real-time PCR analyses for the mRNA levels of the indicated genes in the livers. GAPDH was used as an internal control. n = 4–5 mice. (H) ELISA analyses of plasma levels of IGF-1. n = 4 mice. (I and J) Western analyses (I) and quantification (J) for phosphorylated Akt in the livers. n = 3 mice. Error bars represent standard errors. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S3.

Figure 3.. Hepatic ER stress and the ISR suppress the somatotroph axis by inducing ATF3

(A–C) ChIP with ATF3 antibody followed by quantitative real-time PCR showing ATF3 occupancy at the gene promoters of IGFBP3 and IGF1R in Hepa 1–6 cells. Tubulin was used as a negative control. n = 2. (D) Western blots showing ATF3 expression in stable ATF3 knockdown Hepa 1–6 cells using shRNA. (E and F) ChIP with ATF3 antibody followed by quantitative real-time PCR showing reduced ATF3 occupancy at the gene promoters of IGFBP3 and IGF1R in ATF3 knockdown Hepa 1–6 cells. n = 2. (G) Western analyses of GHR and ATF3 in control and ATF3 knockdown Hepa 1–6 cells with or without tunicamycin induction. (H) Proliferation of stable ATF3 knockdown Hepa 1–6 cells and control cells. n = 3. (I) Annexin V staining of ATF3 knockdown and control Hepa 1–6 cells with or without tunicamycin induction was analyzed with flow cytometry. n = 3. (J) Quantitative real-time PCR analyses of mRNA levels of ATF3 in the livers of SIRT7^−/− mice and wild-type mice with or without Myc knockdown mediated by AAV8-mediated gene delivery. Mice were analyzed 4 weeks after viral infection. n = 4 mice. (K–P) Comparison of SIRT7^−/− mice and wild-type mice with or without ATF3 knockdown mediated by AAV8-mediated gene delivery. Mice were analyzed 4 weeks after viral infection. (K–M) Quantitative real-time PCR analyses of mRNA levels of indicated genes in the livers. GAPDH was used as an internal control. n = 4–5 mice. (N) Elisa analyses of plasma levels of IGF-1. n=4–5 mice. (O and P) Western analyses (O) and quantification (P) for phosphorylated Akt in the livers. n = 3 mice. Error bars represent standard errors. *p < 0.05; **p < 0.01; ***p < 0.001; ns p > 0.05. See also Figure S4.

Figure 4.. Suppression of the somatotroph axis controls liver damage in NAFLD

(A–E) Liver sections stained for Ki67, TUNEL, CD68, and Sirius red (A) and their quantifications (B–E) for SIRT7^−/− mice and wild-type mice with or without ATF3 knockdown mediated by AAV8-mediated gene delivery. Mice were analyzed 4 weeks after viral infection. n = 4–5 mice. Scale bar: 100 μm. (F) A proposed model. Hepatic ER stress and the ISR induce ATF3 expression and the suppression of the somatotroph axis, leading to reduced hepatocyte death, liver damage, and inflammation, while reducing hepatocyte proliferation and parenchymal repair, resulting in compensatory fibrosis. Error bars represent standard errors. *p < 0.05; **p < 0.01; ***p < 0.001; ns p > 0.05. See also Figure S5.

Figure 5.. Suppression of the somatotroph axis controls liver damage in mice fed a CD-HFD

Comparison of wild-type mice with or without ATF3 knockdown in the livers fed a chow diet or a CD-HFD for 8 weeks. (A and B) Western analyses (A) and quantification (B) of ATF3 in the livers. n = 3 mice. (C) ELISA analyses of plasma levels of IGF-1. n = 8 mice. (D–H) Liver sections stained for Ki67, TUNEL, CD68, and Sirius red (D) and their quantifications (E–H). n = 5–6 mice. Scale bars: 200 (Ki67), 100 (TUNEL, Sirius red), and 50 μm (CD68). Error bars represent standard errors. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S6.

Figure 6.. IGF-1 controls liver damage in NAFLD

(A–C) Comparison of wild-type and SIRT7^−/− mice treated with or without IGF-1 for 4 weeks. Data shown are liver sections stained for CD68 and Sirius red (A) and their quantifications (B and C). n = 7 mice. Scale bar: 100 μm. (D–F) Comparison of wild-type mice fed a CD-HFD for 3 weeks followed by treatment with or without IGF-1 for 4 weeks. Data shown are liver sections stained for CD68 and Sirius red (D) and their quantifications (E and F). n = 6–8 mice (E) and 7 mice (F). Scale bar: 100 μm. Error bars represent standard errors. *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 7.. NAD⁺ repletion ameliorates hepatic ER stress, dysregulated somatotroph axis, and liver damage in NAFLD

Comparison of mice fed a chow diet or a CD-HFD for 3 weeks followed by treatment with or without 78c for 4 weeks. (A–C) Western analyses (A) and quantification (B and C) for phosphorylated eIF2α and ATF3 in the livers. n = 3–4 mice. (D and E) Quantitative real-time PCR analyses for the mRNA levels of indicated genes in the livers. GAPDH was used as an internal control. n = 5–8 mice. (F) ELISA analyses of plasma levels of IGF-1. n = 5–8 mice. (G) Liver triglyceride quantification. n = 5–8 mice. (H and I) Quantitative real-time PCR analyses for the mRNA levels of the indicated genes in the livers. GAPDH was used as an internal control. n = 5–8 mice. (J–L) Liver sections stained for CD68 and Sirius red (J) and their quantifications (K and L). n = 5–8 mice. Scale bars: 100 (CD68) and 200 μm (Sirius red). Error bars represent standard errors. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S7.