Adipose tissue macrophage infiltration and hepatocyte stress increase GDF-15 throughout development of obesity to MASH

Abstract

Plasma growth differentiation factor-15 (GDF-15) levels increase with obesity and metabolic dysfunction-associated steatotic liver disease (MASLD) but the underlying mechanism remains poorly defined. Using male mouse models of obesity and MASLD, and biopsies from carefully-characterized patients regarding obesity, type 2 diabetes (T2D) and MASLD status, we identify adipose tissue (AT) as the key source of GDF-15 at onset of obesity and T2D, followed by liver during the progression towards metabolic dysfunction-associated steatohepatitis (MASH). Obesity and T2D increase GDF15 expression in AT through the accumulation of macrophages, which are the main immune cells expressing GDF15. Inactivation of Gdf15 in macrophages reduces plasma GDF-15 concentrations and exacerbates obesity in mice. During MASH development, Gdf15 expression additionally increases in hepatocytes through stress-induced TFEB and DDIT3 signaling. Together, these results demonstrate a dual contribution of AT and liver to GDF-15 production in metabolic diseases and identify potential therapeutic targets to raise endogenous GDF-15 levels.

Fig. 1. GDF-15 is produced by AT in obesity and T2D.

A–C Mice were fed a chow (n = 9) or a HFD (n = 11) for 12 weeks. A Plasma GDF-15 concentrations. B Gdf15 mRNA expression levels in tissues. C Correlation between plasma GDF-15 concentrations and Gdf15 mRNA expression levels. D–I Analysis of paired subcutaneous AT (SAT) and visceral AT (VAT) (n = 42) or liver (n = 46) biopsies from patients. D, E GDF15 mRNA expression levels according to obesity status. F Correlation between GDF15 mRNA expression levels and clinical parameters. G–I GDF15 mRNA expression levels according to T2D status. Data are shown as mean ± SEM. P values calculated by two-tailed Mann–Whitney test (A, E), 2-way ANOVA followed by Sidak’s multiple comparisons test (B, D), two-tailed Spearman correlation (C, F) or Kruskal–Wallis test followed by Dunn’s multiple comparisons test (G–I). *P < 0.05; **P < 0.01; ***P < 0.001; FC fold change. Source data are provided as a Source Data file.

Fig. 2. GDF-15 is produced by liver in MASH.

A Correlation between liver parameters and GDF15 mRNA expression levels in paired SAT and VAT (n = 42) or liver (n = 46). B GDF15 mRNA expression levels in liver according to MASLD status (n = 46). C GDF15 mRNA expression levels in liver of patients with obesity measured by microarray according to MASLD status (n = 840). D Correlation between clinical parameters and GDF15 mRNA expression levels in liver of patients with obesity measured by microarray (n = 840). E Plasma GDF-15 concentrations (n = 23/group). F GDF15 mRNA expression levels in liver measured by microarray according to conventional disease progression (n = 797). G ROC curve of GDF15 mRNA expression levels in liver measured by microarray to predict MASL (vs no MASLD) or MASH (vs no MASLD & MASL). H ROC curve of GDF15 mRNA expression levels in liver measured by microarray to predict steatosis, inflammation or ballooning (≥1 vs 0). I, J Mice were fed a chow (n = 8) or a CDAA diet (n = 12) for 8 weeks. I Gdf15 mRNA expression levels in tissues. J Plasma GDF-15 concentrations. K, L Mice were fed a chow (n = 8) or a HFSCD (n = 12) for 24 weeks. K Gdf15 mRNA expression levels in tissues. L Plasma GDF-15 concentrations. M Correlation between plasma GDF-15 concentrations and Gdf15 mRNA expression levels in tissues according to diet (n = 20/diet). Data are shown as mean ± SEM or SD (C, F). P values calculated by two-tailed Spearman correlation (A, D, M), Kruskal–Wallis test followed by two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli (B, E), Kruskal–Wallis test followed by Dunn’s multiple comparisons test (C, F), ROC analysis (G, H), 2-way ANOVA followed by Sidak’s multiple comparisons test (I, K) or two-tailed Mann–Whitney test (J, L). *P < 0.05; **P < 0.01; ***P < 0.001; FC fold change, AUC area under the curve. Source data are provided as a Source Data file.

Fig. 3. Macrophages express high levels of GDF15 in mouse epiAT.

A Representative immunohistochemistry of epiAT sections from HFD-fed mice stained with anti-GDF-15 antibody. Scale bars, 100 µm. B Gdf15 mRNA expression levels in adipocytes and sorted SVF from epiAT of mice on HFD feeding for 12 weeks (n = 8). C Gdf15 mRNA expression levels in the main immune cell populations sorted from several tissues of naive mice (n = 4). D Correlation between mRNA expression levels of macrophage or lymphocyte markers in AT and Gdf15 mRNA expression levels or plasma GDF-15 concentrations in mice on chow or HFD feeding (n = 20). E Correlation between immune cell composition in AT determined by flow cytometry and GDF15 mRNA expression levels in the corresponding tissue or plasma GDF-15 concentrations in mice on chow or HFD feeding (n = 19). F Mice were fed a chow (n = 5) or a HFD (n = 7) for 12 weeks and epiAT were processed for cell sorting. Gdf15 mRNA expression levels were measured in adipocytes and sorted SVF. G–I Mouse blood monocytes were differentiated in monocyte-derived macrophages (MDM) by stimulation with M-CSF for 7 days (n = 3-4). G Gdf15 mRNA expression levels in freshly sorted monocytes (T0) or in fully differentiated MDM (D7) (n = 4). H GDF-15 protein levels detected by western blot in freshly sorted monocytes or in fully differentiated MDM (n = 3). Nonspecific bands are indicated with a *. Uncropped blot in Source Data. I Secretion of GDF-15 in supernatant (SN) during the first (T0-D1) and the last (D6-D7) 24 h (n = 4). Data are shown as mean ± SEM. P values calculated by Friedman test followed by two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli (B), two-tailed Spearman correlation (D, E), 2-way ANOVA followed by Sidak’s multiple comparisons test (F) or two-tailed Mann–Whitney test (G, I). *P < 0.05; **P < 0.01; ***P < 0.001; AU arbitrary unit, SVF stromal vascular fraction, ADSCs adipose-derived stem cells, FC fold change. Source data are provided as a Source Data file.

Fig. 4. ATMs are the main source of GDF-15 in human AT.

A GDF15 mRNA expression levels in adipocytes and SVF isolated from VAT of patients with obesity (n = 15). Connected dots represent fractions from the same patient. B Representative immunohistochemistry of VAT sections from patients with obesity stained with anti-GDF-15 antibody. Scale bars, 100 µm. C GDF15 mRNA expression levels in sorted SVF from VAT of patients with obesity (n = 24). D Representative immunofluorescence staining of VAT section from a patient with obesity stained with anti-GDF-15 and anti-CD68 antibodies. Scale bars, 30 µm. E Correlation between GDF15 mRNA expression levels and mRNA expression levels of macrophage or lymphocyte markers in paired SAT and VAT from patients (n = 42). Data are shown as mean ± SEM. P values calculated by two-tailed Wilcoxon matched-pairs signed rank test (A), Kruskal–Wallis test followed by Dunn’s multiple comparisons test (C) or two-tailed Spearman correlation (E). *P < 0.05; **P < 0.01; ***P < 0.001; AU arbitrary unit, SVF stromal vascular fraction, ADSCs adipose-derived stem cells. Source data are provided as a Source Data file.

Fig. 5. Macrophage infiltration regulates GDF-15 production by AT.

A, B Mice were fed a chow or a HFD for 12 weeks followed by a single intraperitoneal injection of anti-CD115 antibody or isotype control (n = 8 chow and 12 HFD/antibody). Mice were sacrificed after 4 days. A ATM content determined by flow cytometry. B Gdf15 mRNA expression levels in AT. C–F Mice were fed a chow (n = 6) or a HFD (n = 16) for 18 weeks followed by intraperitoneal injection of siGdf15-GeRPs or siCtrl-GeRPs for 5 consecutive days (n = 8/siRNA-GeRPs). Mice were sacrificed 24 h after the last injection. C Gdf15 mRNA expression levels in adipocytes or sorted SVF from epiAT (n = 5/group). D Gdf15 mRNA expression levels in epiAT. E Difference in plasma GDF-15 concentrations after 3 days or at sacrifice (D5) compared to T0. F Difference in body weight after 3 days or at sacrifice (D5) compared to T0. G–K After bone marrow transplantation (BMT) from Gdf15^+/+ or Gdf15^−/− mice, reconstituted mice were fed a chow or a HFD for 12 weeks (n = 5 chow and 11 HFD/genotype). G Evolution of body weight. Statistical analysis is only shown for the factor “genotype” (Gdf15^+/+ or Gdf15^-/-) calculated by 2-way ANOVA. H Body weight gain (12th week minus T0). I Gdf15 mRNA expression levels in epiAT. J Plasma GDF-15 concentrations. K Gdf15 mRNA expression levels in liver. Data are shown as mean ± SEM. P values calculated by 2-way ANOVA followed by Holm-Sidak’s (A, G, J) or Sidak’s (B, C, E, F, H, I, K) multiple comparisons test or two-tailed Mann–Whitney test (D). AU, arbitrary unit; FC, fold change. Source data are provided as a Source Data file.

Fig. 6. MASH-related stress increases GDF15 expression in hepatocytes.

Mice were fed a HFD for 12 weeks (A) (n = 40) or CDAA diet for 4 weeks (B) (n = 16) followed by injection of anti-CD115 antibody or isotype control. Gdf15 mRNA expression was measured in liver. C Correlation between GDF15 mRNA expression levels and mRNA expression levels of macrophage/lymphocyte markers in liver. D Gdf15 mRNA expression levels in hepatocytes and non-parenchymal cells (NPCs) from naive mice (n = 8). Connected dots represent fractions from the same mouse. E Gdf15 mRNA expression levels in hepatocytes from mice on different diet feeding (n = 36). F, G Mice were fed a CDAA diet for 4 weeks and tail vein injected with invivofectamine-siRNA complex prior sacrifice (n = 28). F Gdf15 mRNA expression levels in liver. G Plasma GDF-15 concentrations. H Correlation between mRNA expression levels of GDF-15 regulators and GDF15 mRNA expression levels, steatosis, inflammation or ballooning in liver of patients with obesity (n = 840). I GSEA of TFEB, EGR1 and DDIT3 (CHOP) signatures in liver of patients with obesity according to MASLD parameters (n = 840). J Correlation between mRNA expression levels of TFs regulating Gdf15 with Gdf15 mRNA expression levels in liver from mice on different diet feeding (n = 20/diet). K GSEA of stress signatures in liver of patients with obesity according to MASLD parameters (n = 840). Ten pathways for each type of stress are illustrated. Full analysis in Supplementary Fig. 9. L, M Mice were fed a CDAA diet for 4 weeks and tail vein injected with invivofectamine-siRNA complex prior sacrifice (n = 31). L Gdf15 mRNA expression levels in liver. M Plasma GDF-15 concentrations. Data are shown as mean ± SEM. P values calculated by 2-way ANOVA followed by Holm-Sidak’s multiple comparisons test (A, E–G, L, M), Kruskal–Wallis test followed by Dunn’s multiple comparisons test (B), two-tailed Spearman correlation (C, H, J) or two-tailed Wilcoxon matched-pairs signed rank test (D). *P < 0.05; **P < 0.01; ***P < 0.001; FC, fold change; AU, arbitrary unit. Source data are provided as a Source Data file.