Fos-induced osteosarcoma growth causes a cachexia-like phenotype in mice and correlates with high Fgf21 serum levels

Abstract

Background

The dimeric AP-1 transcription factor has been described as a regulator of bone metabolism, but also of adipocyte biology. More specifically, its family members adopt specific functions in the regulation of adipocyte differentiation and adipose tissue hypoxia, but also in controlling lipolysis, energy expenditure and insulin sensitivity.

Methods

In this study, we analyzed the role of the AP-1 family member Fos, whose ubiquitous over-expression in transgenic mice has been described to cause osteosarcoma formation, on adipose tissue, glucose and lipid metabolism. More specifically, we analyzed the metabolic phenotype of the respective mouse model by histological analyses, monitoring gene expression, determination of serum parameters and ex vivo adipogenesis assays.

Results

We show that FosTg mice additionally display an age-dependent loss of white adipose tissue, which is associated with reduced adipocyte size. Quantitative real-time PCR analyses of white adipose tissue as well as in vitro studies using adipocyte-derived stromal cells excluded a cell-autonomous defect in adipocyte differentiation. However, FosTg mice displayed low circulating glucose levels along with increased glucose tolerance and insulin sensitivity. By proteomic analysis we identified elevated serum levels of fibroblast growth factor 21 (Fgf21) in FosTg mice which could explain the observed reduced hepatic lipogenesis and white adipose tissue mass. Most importantly, we demonstrated that this metabolic phenotype is dependent on osteosarcoma formation by crossing the FosTg mice with mice deficient for Rsk2, a kinase necessary for tumor growth. Moreover, we could show a direct correlation between tumor burden and Fgf21 serum levels in our mouse model.

Conclusion

In summary, our data identify Fgf21 as a potential regulator of glucose and lipid metabolism in Fos transgenic osteosarcoma-bearing mice, thereby supporting its essential function in the regulation of the metabolic phenotype associated with cancer cachexia.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40170-025-00417-y.

Introduction

The transcription factor AP-1 is formed by heterodimerization of one of the four Fos family members (Fos, FosB, Fra1 and Fra2) with one of the three Jun family members (Jun, JunB and JunD) or by homodimerization of members of the Jun family. The expression and activity of the family members are regulated by a variety of extracellular stimuli including growth factors, pro- and anti-apoptotic agents or differentiation signals. Consequently, AP-1 has been implicated in the regulation of cell proliferation and apoptosis and it is involved in the differentiation of various cell types as well as in tumor development [1]. Genetic analysis using gain- or loss-of-function experiments in mice, further revealed the key function played by each AP-1 member in skeletal development or in the regulation of bone remodeling necessary for post-developmental bone maintenance [2]. Specifically, gain-of-function of any of the Fos members has generated phenotypes affecting bone-forming cells, i.e. osteoblasts. For instance, the main consequence of an ubiquitously overexpressed short isoform of FosB (ΔFosB), of Fra1 (encoded by Fosl1) or of Fra2 (encoded by Fosl2) is increased osteoblast differentiation leading to osteosclerosis [3–5]. The function of Fra1 and Fra2 as physiological regulators of bone formation has been confirmed by the analysis of mouse deficiency models, which developed osteopenia due to decreased osteoblast function [5, 6].

In contrast to other AP-1 members, ubiquitous over-expression of Fos (originally called c-Fos) in transgenic mice leads to osteosarcoma formation due to the transformation of chondro-osteoblastic cells, a phenocopy of the tumorigenic activity of the viral oncogene v-fos [7, 8]. The relevance of Fos in tumor formation is confirmed by its increased expression in human osteosarcomas [9, 10]. In H2kb-FosTg mice, Fos transgene expression occurs postnatally in a tissue-specific manner. However, tumor formation is restricted to bone tissue where transgene expression starts at 2–3 weeks of age, leading to osteosarcoma formation in all bones with 100% penetrance [7].

Cancer cachexia, affecting approximately 80% of cancer patients, is a catabolic syndrome, characterized by a progressive loss of muscle mass, fat mass and body weight, that cannot be completely reversed by nutritional supplementation [11]. During tumor-associated anorexia-cachexia, Fos induction was detected in the forebrain [12]. Furthermore, elevated IL-6 levels associated with increased Fos expression regulating cancer cachexia were detected in the area postrema [13]. In addition, the AP-1 signaling cascade has been implicated in muscle wasting [14, 15], in particular as persistent Fos expression in muscle progenitor cells disrupts myogenic differentiation [16], while blocking AP-1 restores muscle mass in tumor-bearing rats [15, 17]. Finally, Fos has been described to be induced when stimulated by growth differentiation factor 15 (GDF15), a factor implicated in cachexia [18].

Fgf21 is a recently discovered factor that is proposed as a prognostic biomarker associated with a poor prognosis in various cancers [19–21]. Its expression has been described to be increased in several types of cancer, but the effect seems tumor-, context- and dose-dependent. For instance, Fgf21 has been described to inhibit pancreatic cancer cell proliferation [22] and pancreatic inflammation [23] and furthermore delayed liver tumor development [24]. In other models, Fgf21 secretion by tumors promoted the growth of lung cancer cell lines [25] as well as tumor progression [20, 26]. Originally however, Fgf21 has been described as an endocrine factor primarily produced in the liver, where its expression is increased in response to starvation. Nonetheless, Fgf21 is also expressed in pancreas, white adipose tissue and muscle and it plays a crucial role in regulating energy expenditure, glucose and lipid metabolism. Gain- and loss-of-function studies, as well as pharmacological administration of Fgf21 in mice have additionally demonstrated its function as a regulator of insulin sensitivity and adipose tissue mass [27–29].

In the present study, we investigated the effects of ubiquitous Fos over-expression in mice, which causes osteosarcoma formation, on adipose tissue, glucose and lipid metabolism. Our analyses demonstrate a decreased adipose tissue mass and an increased glucose tolerance and insulin sensitivity in FosTg mice. Proteomic analysis identified Fgf21 as a central mediator that can cause the metabolic alterations associated with the cachexia-like phenotype. Importantly, we could show a correlation between Fgf21 levels and tumor burden, indicating that Fgf21 has an essential function in the regulation of metabolic effects associated with cancer cachexia.

Materials and methods

Mice

The FosTg mice (H2-c-fos-LTR), described by Ruther et al. [30], were crossed in a C57BL/6 background. RSK2 knockout mice were previously described [31, 32]. Genotyping was performed using the following primers: H2cfos up 5’-AGT CTG GCC TGC GGG TCT CT-3’, H2cfos down 5’-GTC GGC TGG GGA ATG GTA GTA GG-3’, Rsk2 up 5´-TTG TTG GTT TAC TTT CTT TCG GTC TG-3´ and Rsk2 down 5’-AAG ATG ATT GCT TTG CTT AGT TTA-3’. Mice were kept under standard diet (1328P, Altromin Spezialfutter GmbH&Co. KG) ad libitum in a specific pathogen-free environment with a 12 h light/dark cycle and 20–24 °C ambient temperature. A maximum of 6 mice was housed in individually ventilated cages. All animal experiments were approved by the local animal facility and by the ‘Amt für Gesundheit und Verbraucherschutz’ (Org_633, Org_984, N22/059).

Tolerance tests

To determine the glucose tolerance, overnight fasted mice were intraperitoneally injected with 2 mg/g body weight of glucose. For insulin tolerance tests, mice fasted for 6 h were injected with 0.75U insulin. Blood glucose levels were measured by incision of the tail vein using a glucometer (Ascentia Elite, Bayer). These animal experiments were approved by the local animal facility and by the ‘Amt für Gesundheit und Verbraucherschutz’ (G16/073).

RNA analyses

RNA was isolated using Trifast Reagent (Peqlab) according to the manufacturer’s instructions and RNA was reversed transcribed using the Verso cDNA kit (Thermo Scientific). Q-PCR analyses were performed using predesigned TaqMan gene expression assays (Thermo Scientific) or SybrGreen (Applied Biosystems) with self-designed primers in a StepOnePlus system (Thermo Scientific). Results were normalized to the level of GapDH (adipose tissue, bone tissue, tumor and cell culture experiments) or Hprt (liver).

Protein analyses

Serum protein concentrations were quantified by Elisas: leptin and insulin (CrystalChem), adiponectin (Invitrogen), Fgf21 (R&D) and osteocalcin (AlfaAesar). Triglycerides and non-esterified fatty acids were quantified by colorimetric assays (Sigma and Wako Diagostics, respectively).

Total protein and nuclear protein extracts were isolated as previously described [33, 34]. In brief, total protein was extracted by incubation of cells with RIPA buffer (150 mM NaCl, 2 mM EDTA, 10 mM sodium phosphate, 1% NP-40, 1% sodium desoxycholat, 0.1% SDS) for 15 min, followed by 10 min of centrifugation at 16.000 rpm and 4 °C. For nuclear extracts, protein was isolated by addition of 450 µl buffer A (50 ml consist of: 500 µl HEPES (1 M), 250 µl KCl (2 M), 10 µl EDTA (0.5 M), 50 µl EGTA (0.1 M), 37.5 µl spermidine (1 M), 15 µl spermine (0.5 M), 50 µl DTT (1 M), 250 µl PMSF (0.1 M), 500 µl sodium molybdate (1 M), 250 µl sodium vanadate (10 mM) and protease inhibitor cocktail), incubation on ice for 15 min, followed by addition of 25 µl of 10% NP 40. Vortexed samples were centrifuged (30 s, 12.000 x g, 4 °C) and incubation of the pelleted nuclei with buffer B (for 50 ml: 1 ml HEPES (1 M), 4 ml NaCl (5 M), 100 µl EDTA (0.5 M), 500 µl EGTA (0.1 M), 50 µl DTT (1 M), 250 µl PMSF (0.1 M), 500 µl sodium molybdate (1 M), 250 µl sodium vanadate (10 mM), and protease inhibitor cocktail) for 15 min at 4 °C, followed by a second centrifugation step (5 min, 12.000 x g, 4 °C). Determination of protein concentration was performed by Bradford (Bio-Rad). For Western blotting, 12% SDS polyacrylamide gels and a nitrocellulose membrane were used, and membranes were incubated with an anti-phospho-p44/42 MAP Kinase (Thr202/Tyr204), an anti-phospho-CREB (Ser133), an anti-phospho-mTOR (Ser2448), an anti-phospho-Jak1 (Tyr1034/1035), an anti-phospho-Stat3 (Ser727), an anti-phospho-Smad2 (Ser465/467), an anti-mTOR, an anti-Jak1, an anti-Stat3, an anti-Smad2, an anti-ß-actin or an anti-α-tubulin antibody (all Cell Signaling). Quantification was performed using ImageJ2 software.

Histological analysis

Adipose tissue and liver samples were fixed in 3.7% Formafix overnight and embedded in paraffin. 1.5 μm sections (adipose tissue) and 3 μm sections (liver) were stained with hematoxylin and eosin according to standard protocols. Adipocyte size was measured in 3–6 fields at a 20x magnification.

Sections of the pancreas (1 μm) were unmasked (citrate buffer, 20 min, 92 °C), followed by staining with an anti-insulin antibody (Santa Cruz). The insulin positive areas and total pancreas surface were measured using osteomeasure software (OsteoMetrics).

Contact x-ray of formalin-fixed skeletons was performed using a Faxitron X-ray cabinet at 35 kV for 2 s. For histology of vertebral bodies (L2-L5), skeletons were fixed in 3.7% Formafix for two days, dehydrated in ascending alcohol concentrations and embedded in methylmethacrylate. 4 μm sections were stained by von Kossa/van Giesson according to standard protocols. Static histomorphometry was performed using BioQuant software (Bioquant Image Analysis Corporation).

Cell culture

ADSCs were isolated as described in Schwabe et al., 2016 [35]. Adipocyte differentiation was induced by addition of insulin (5 µg/ml), IBMX (500 µM) and dexamethasone (1 µM) to the medium of confluent cells (α-MEM, 10% FCS, 1% penicillin/streptomycin). For Red Oil O staining, cells were fixed in 3.7% Formafix for 15 min and stained with 0.3% Red Oil O in 60% isopropanol for 2 h. For the growth curve, 4 × 10⁴ cells/well were plated in 6 well plates and over a seven-day period, trypsinized cells were counted. For stimulation experiments, cells were stimulated with dexamethasone (10 µM) or serum starved (0.5% FBS) and stimulated with insulin (5 μm/ml) or IBMX (500 µM).

Primary osteoblasts were isolated by digestion of calvariae of newborn mice as described in Jochum et al., 2000 [3]. Osteoblast differentiation was induced by addition of ß-glycerophosphate (10 mM) and ascorbate (50 µg/ml) to the medium (α-MEM, 10% FCS, 1% penicillin/streptomycin) of confluent cells. For Alizarin Red staining, cells were fixed in 90% ethanol (10 min), followed by staining with Alizarin Red (40 mM, pH 4.2, 10 min). To quantify the staining, cells were incubated with 10% acetic acid (30 min room temperature, followed by 5 min at 85 °C). After centrifugation (13.000 rpm, 10 min), ammonium hydroxide solution was added at a ratio of 1:8 and the optical density was measured at 405 nm.

Proteomic analysis

For protein profiling pooled serum samples of at least 5 wt and FosTg mice were used. Experiments were performed by Sciomics GmbH (Heidelberg, Germany). In brief, extracted proteins were labeled with fluorescent dyes (ScioDye 1 and ScioDye 2) and the abundance of proteins was analyzed in a dual-color approach using scioDiscover antibody array. The proteins were defined as differentially expressed for ǀlog₂FCǀ > 0.5. Pathway annotations were performed using String (https://string-db.org).

Study design and statistical analysis

For the experiments, littermates were used as controls. Mice were assigned randomly to the groups and samples were analyzed in a blinded fashion. All data are the mean ± s.e.m. Statistical significance was determined by two-tailed Student’s t-test, Mann-Whitney-U-test, one-way ANOVA or 2-way ANOVA with Bonferroni’s comparisons test using GraphPad Prism Software (GraphPad Software Inc.). *P < 0.05, ** P < 0.01, *** P < 0.001, ****P < 0.0001.

Results

Decreased body weight and decreased white adipose tissue mass in Fos overexpressing mice

Adult Fos transgenic mice (thereafter termed FosTg mice) appeared smaller and leaner than their wild-type littermates (Fig. 1A). We therefore compared their age-dependent weight gain from 8 to 22 weeks of age. Decreased body weight was constantly found in the FosTg mice (Fig. 1B), suggesting decreased adipose tissue mass. To verify this, we isolated various fat pads from wild-type and transgenic mice in order to compare their size. A reduced size of all fat pads was observed in the FosTg mice (Fig. 1C and Fig. S1A). By quantifying these differences, we detected a clear decrease in the mass of all white adipose tissues (WAT) (Fig. 1D-F), as well as of brown adipose tissue (BAT) (Fig. 1G), in mice over-expressing Fos. Thus, similar to mice over-expressing Fra1 or ΔFosB [4, 36], Fos transgenic mice display a significantly decreased body weight associated with generally decreased adipose tissue mass.

Fig. 1.

General loss of adipose tissue mass in Fos transgenic mice. (A) Representative images of 16-week-old male wild-type (wt) and Fos transgenic (FosTg) littermates. (B) Body weight of male wt and FosTg mice at the indicated ages. (C) Representative images of isolated adipose tissues from 16-week-old male wt and FosTg mice. (D) Epididymal white adipose tissue (WAT) weight of male wt and FosTg mice at the indicated ages. (E-G) Weight of subcutaneous posterior WAT (E), of the perirenal WAT (F) and of the brown adipose tissue (BAT) (G) of 16-week-old male mice. Data represent mean ± s.e.m. (n ≥ 4). Asterisks indicate statistically significant differences (****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05)

Fos overexpression leads to reduced adipocyte size, but does not globally affect the expression of adipocyte markers in WAT

We histologically analyzed the epididymal WAT of FosTg mice at the age of 8 and 16–20 weeks and performed histomorphometric quantification of the adipocyte diameter in WAT, a surrogate measurement for maturation and lipid storage of the cells. As expected, increased cell size was observed in wild-type mice during aging. However, although the size repartition of the cells of 8-week-old mice was similar in wild-type and FosTg mice, the increased cell size that was normally observed during aging was abolished in the FosTg mice (Fig. 2A-C).

Fig. 2.

Reduced adipocyte size, but no change in the expression of adipocyte markers in WAT of FosTg mice. (A-C) Representative pictures of H&E-stained sections (A) and repartition curves of the adipocyte diameter of epididymal WAT from 8-week-old (B) and 16-20-week-old males (C) of the indicated genotypes. (D-G) Q-PCR analysis for expression of the Fos transgene (D), endogenous (E) and total Fos (F) and quantification of Fos protein levels (G) in epididymal WAT of 16-week-old mice. (H-M) Q-PCR analysis for expression of markers for white adipocytes in epididymal WAT of 16-week-old mice. Data represent mean ± s.e.m. (n ≥ 5). Asterisks indicate statistically significant differences (****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05)

The cellular phenotype observed in WAT of Fos overexpressing mice suggested that the reduced adipose tissue mass might be associated with decreased maturation of adipocytes, a phenotype reminiscent of the adipose tissue characteristics caused by overexpression of Fra1, a member of the Fos family also known to be a Fos target gene [36, 37]. Therefore, we analyzed the expression levels of Fra1 and other AP-1 members in WAT of wild-type and FosTg mice. As expected, and in agreement with its ubiquitous expression, the mRNA encoded by the transgene (FosTg) was found to be expressed in WAT of transgenic mice (Fig. 2D). Although expression of the transgene did not significantly affect expression of Fosl2 (Fra2) or of any Jun family members, a significant repression of Fosl1 (Fra1) was observed (Fig. S1B). We next performed Q-PCR analyses of endogenous and total Fos. We observed a non-significantly reduced expression of endogenous Fos in FosTg WAT, leading to similar total Fos expression levels in wild-type and FosTg mice (Fig. 2E-F). The absence of increased Fos expression was confirmed at the protein level (Fig. 2G). Nevertheless, to exclude a potential cell-autonomous mechanism explaining the fat phenotype of FosTg mice, we performed Q-PCR analyses for adipocyte differentiation and activity markers. No clear reduction was found when quantifying the expression of transcription factors regulating early differentiation (Cebpb and Cebpd) or maturation (Cebpa and Pparg) of adipocytes (Fig. 2H-K). Markers characterizing differentiated adipocytes (Fabp4 and Glut4) were also unaffected (Fig. 2L-M). These findings suggested that white adipocyte differentiation and function are not globally affected in our Fos transgenic model.

The Fos-induced adipose tissue phenotype is not cell-autonomous

We isolated adipose-derived stem cells (ADSCs) from adult mice to further omit the potential cell-autonomous effect of Fos overexpression on white adipocyte differentiation in vitro. First, we established cellular growth curves that did not reveal any difference, thus excluding a proliferation or apoptosis phenotype (Fig. S2A). We next compared the response to adipogenic stimulation of the ADSCs isolated from wild-type and FosTg mice. Here, we first studied the response of the cells to IBMX by analyzing the phosphorylation of Creb. A comparable level of Creb phosphorylation was observed using Western blotting following IBMX stimulation (Fig. S2B). Similarly, we demonstrated the functionality of the insulin receptor in ADSCs isolated from wild-type and FosTg mice by monitoring the activation of ERK in response to short term stimulation by insulin. A similar level and kinetic of ERK phosphorylation were observed by Western blot (Fig. S2C). Finally, a similarly increased expression of Gilz1 and Per1, two typical glucocorticoid target genes, were observed after dexamethasone treatment, thereby demonstrating a normal response of adipocyte progenitor cells to glucocorticoids (Fig. S2D and E). Thus, ADSCs isolated from FosTg mice are normally responsive to adipogenic stimuli in vitro and are therefore expected to normally differentiate into adipocytes.

To test this hypothesis, we next induced adipogenesis by treating confluent cells with insulin, IBMX and dexamethasone. The transgene was expressed in cells isolated from FosTg mice and its level of expression was not affected by adipogenic stimulation (Fig. 3A). We next analyzed endogenous and total Fos expression levels. Endogenous Fos displayed a transient peak at day 3 of adipogenic differentiation in both genotypes, indicating the importance of Fos in early adipogenesis (Fig. 3B). However, total Fos expression levels were similar in wild-type and FosTg mice (Fig. 3C). To prove the absence of an effect mediated by the transgene, we performed Red Oil O staining to demonstrate the efficacy of adipocyte differentiation and did not detect any qualitative difference between cultures from wild-type and FosTg mice (Fig. 3D). This observation was confirmed by quantitative analysis, which demonstrated that there was no difference in the spontaneous adipocyte differentiation in the absence of stimulation (Fig. 3E). Importantly, the cells responded with a similar kinetic and amplitude to the adipogenic stimulation (Fig. 3F). These data were further verified by Q-PCR analysis for expression of adipocyte markers, including Pparg2, Cebpa, Glut4 and Fabp4, as well as adipokines which were all unaffected by expression of the transgene (Fig. S3A to H).

Fig. 3.

The fat phenotype of Fos transgenic mice is not caused by a cell-autonomous defect in adipocyte differentiation. (A-C) Q-PCR analysis for expression of the Fos transgene (A), endogenous Fos (B) and total Fos (C) in wt and FosTg ADSCs during the course of adipocyte differentiation. (D) Representative images of wt and FosTg ADSCs after Red oil O staining (in red) at day 0 (confluency) or 12 days after inducing the adipocyte differentiation. (E) Quantitative analysis of the Red oil O staining of ADSCs isolated from subcutaneous WAT of wild-type (wt) and FosTg mice at different time points in cultures without adipogenic stimulation. (F) Quantitative analysis of the Red oil O staining at different time points in cultures with adipogenic stimulation. Data represent mean + s.e.m. (n ≥ 3 independent cell isolations)

As ADSCs are adipose tissue resident mesenchymal stem cells with the capacity to differentiate into osteoblasts [38], we also performed osteogenic differentiation of the ADSCs, which demonstrated the normal capacity of cells isolated from the FosTg mice to mineralize in response to stimulation with β-glycero-phosphate and ascorbic acid (Fig. S4A and B). Likewise, no change was observed when quantifying bone parameters in tumor-free areas of bones from FosTg mice (Fig. S4C and D). Thus, the decreased WAT mass in FosTg mice is independent of a cell-autonomous defect affecting mesenchymal progenitors and thus adipogenesis, as well as of a switch from adipogenesis to osteoblastogenesis.

Decreased circulating glucose but normal insulin levels in FosTg mice

Since the fat phenotype in FosTg mice was not explained by a cell-autonomous defect in adipogenesis, we analyzed for potential variations in the systemic regulation of adipose tissues related to the known interaction of osteoblasts and adipocytes. We therefore measured the circulating levels of hormones that have been shown to couple bone and adipose tissue metabolism in 16-week-old mice. In agreement with their lean phenotype, leptin, a hormone produced by mature adipocytes to control satiety, was decreased in the FosTg mice and was not sensitive to food deprivation (Fig. 4A). In contrast, levels of osteocalcin, a hormone produced by osteoblasts, were found increased in sera of FosTg mice (Fig. 4B), probably as a result of its production by the tumors. As osteocalcin has been shown to induce the production of insulin via an increased proliferation of pancreatic β-cells, we performed insulin immunostaining in order to quantify the average volume of β-cell islets on pancreas sections, a commonly used surrogate method to evaluate proliferation of these cells. No difference in the average surface occupied by insulin-positive cells was observed when comparing the two genotypes (Fig. 4C). Likewise, no increase in the circulating levels of insulin could be detected and a physiologically increased insulin secretion was observed in the blood of FosTg mice following food deprivation (Fig. 4D). However, serum glucose levels were significantly lower in the FosTg mice after feeding, and the expected lowering of glucose was not observed after food deprivation (Fig. 4E). However, the expression level of the adipokines resistin and Il6, that were examined in this context, were not affected in white adipose tissue (Fig. S5A-B).

Fig. 4.

Decreased serum glucose levels in FosTg mice. (A) ELISA-based quantification of circulating leptin concentrations in 16-week-old mice fed (+) or starved for 6 h (-). (B) Circulating levels of total osteocalcin. (C) Representative images showing immunostaining for insulin in the pancreas of 16-week-old mice and quantification of the β-cell area. (D-E) Circulating levels of insulin (D) and glucose (E) in fed (+) and starved (-) mice. Data represent mean ± s.e.m. (n ≥ 5). Asterisks indicate statistically significant differences (***P < 0.001; **P < 0.01; *P < 0.05)

Increased serum Fgf21 concentrations in Fos transgenic mice

To assess the mechanism responsible for the metabolic phenotype of FosTg mice, we employed a proteomic approach. Thereby, the abundance of 1400 proteins was determined in pooled murine serum samples of wt and FosTg mice. Functional annotation of factors using STRING showed pathways related to cancer and TNF signaling, as well as PI3K-Akt/MAPK/Ras/Rap1 signaling as main pathways affected when focusing on differentially expressed secreted factors in the osteosarcoma-developing FosTg mice compared to control (Fig. 5A). By analyzing this dataset, we found an increased abundance of cachexia-inducing factors, namely Tnfsf10, Csf1, Tgfa and Hgf in FosTg mice. Furthermore, the analysis revealed an increased level of Fgf21, a factor described to be elevated in patients with cachexia [39], and of its target gene adiponectin in the sera of FosTg mice (Fig. 5B). To confirm the results of the antibody microarray analysis, we applied ELISA and qRT-PCR. Despite an increased expression of adiponectin in adipose tissue on RNA level, its serum levels were not significantly elevated (Fig. 5C-D). Nevertheless, we observed increased concentrations of Fgf21 in the serum of FosTg mice (Fig. 5E). To address the tissue specificity of Fgf21 expression we compared the RNA expression levels in white and brown adipose tissue, liver and bone/tumor of wild-type and FosTg mice. In wild-type mice, Fgf21 showed the highest expression in the liver, which is described as primary site of Fgf21 production (Fig. 5F). In FosTg mice, Fgf21 RNA levels were significantly elevated in BAT, liver and tumor with high expression levels in liver and tumor tissue, emphasizing the importance of these organs for Fgf21 expression (Fig. S6A-D and Fig. 5G).

Fig. 5.

Increased serum Fgf21 concentrations in Fos transgenic mice. (A-B) Results of the antibody microarray analysis: Selection of most affected pathways showing the percentage of proteins within each pathway (A). Heat map of differentially expressed secreted factors related to cachexia in wt compared to FosTg mice (B). (C) ELISA-based quantification of circulating adiponectin concentrations. (D) Quantitative Real-time PCR analysis of Adipoq expression in the white adipose tissue of 16-week-old mice. (E) Circulating levels of Fgf21 in wt compared to FosTg mice. (F-G) qRT-PCR analyses of Fgf21 expression in white and brown adipose tissue, liver and bone samples of wild-type mice (F), and of white and brown adipose tissue, liver and tumor samples in FosTg mice (G). Data represent mean ± s.e.m. (n ≥ 3). Asterisks indicate statistically significant differences (**P < 0.01; *P < 0.05)

Increased insulin sensitivity in FosTg mice

Cachexia is characterized by a reduced muscle mass [40]. Furthermore, Fgf21, that was shown to induce skeletal muscle atrophy, has been described to play a role in cachexia [41, 42]. Therefore, we measured the weight of the quadriceps, which was found to be reduced in FosTg mice (Fig. 6A). Similarly, the heart weight was decreased in mice overexpressing Fos compared to control (Fig. 6B). However, the reduction in muscle mass as well as of the total body weight of Fos transgenic mice was not caused by anorexia, as we could exclude a decreased food consumption by comparing food intake in wild-type and FosTg mice at the ages of 8 and 16 weeks (Fig. 6C).

Fig. 6.

Increased glucose tolerance and insulin sensitivity in FosTg mice. (A-B) Quadriceps weight (A) and heart weight (B) of 16-week-old male wt and FosTg littermates. (C) Quantification of food intake in 8- and 16-week-old wt and FosTg mice. (D) Representative pictures of H&E stained sections of the subcutaneous white adipose tissue from 16-week-old males of the indicated genotype. (E-G) Q-PCR analysis for expression of Fgf21 (E) and Ucp1 (F) and Cidea (G) in the subcutaneous adipose tissue of FosTg mice relative to wt. (H) Glucose tolerance test in 16-week-old wt and FosTg mice. (I) Insulin tolerance test in 16-week-old wt and FosTg mice. Data represent mean ± s.e.m. (n ≥ 4). Asterisks indicate statistically significant differences (****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05)

As in cachexia, an increase of pro-inflammatory cytokines in the liver is typically described, we performed Q-PCR analyses of Tnfa, Il1b, Il6 and S100a9 expression. These genes seem to be elevated in the liver of FosTg mice, with Tnfa and Il1b being significantly increased (Fig. S7A-D).

Next, we analyzed the subcutaneous white adipose tissue of FosTg mice for symptoms of cachexia. Histological examination of the subcutaneous white adipose tissue, which is the typical site of adipocyte browning, showed, similar to the gonadal adipose tissue, adipocytes with a reduced cell size (Fig. 6D). Importantly, while Fgf21 expression is not significantly elevated, we observed an increased expression level of the thermogenic gene Ucp1 and Cidea, markers that are associated with WAT browning, leading to the increased energy expenditure observed in cachexia (Fig. 6E-G).

As Fgf21 was described to regulate insulin sensitivity [27], glucose and insulin tolerance tests were performed. Here, FosTg mice showed an increased rate of glucose clearance compared to wild-type controls (Fig. 6H). Furthermore, Fos overexpression was found to enhance insulin sensitivity (Fig. 6I).

Taken together, FosTg mice show multiple symptoms of cachexia, that have been described to be regulated via Fgf21, namely reduced muscle mass, signs of browning of subcutaneous white adipose tissue and increased insulin sensitivity.

Decreased lipogenesis in the liver of FosTg mice

Fgf21 was further shown to decrease lipogenesis and to increase gluconeogenesis and fatty acid oxidation in the liver [43]. Therefore, we evaluated the glucose and lipid metabolism in the liver of FosTg mice compared to control. We confirmed Fos transgene expression in the liver of transgenic mice (Fig. 7A). While endogenous Fos levels remained unchanged, total Fos expression was significantly higher in FosTg mice compared to wild-type animals as demonstrated at RNA and at protein level (Fig. 7A-B and Fig. S8A). Despite the observed adipose tissue phenotype, no histological evidence of liver steatosis was observed (Fig. 7C-D). In addition, we did not observe a significant difference in the glycogen content in the liver despite tumor formation in FosTg mice (Fig. 7E). Next, we analyzed the serum level of non-esterified fatty acids and triglycerides. In accordance with the effects of Fgf21 in the liver, the level of non-esterified fatty acids was not affected, while the level of triglycerides was strongly reduced in FosTg compared to control mice (Fig. 7F-G). Transcriptional regulators of enzymes involved in ß-oxidation (Ppara and Cpt1a) were not differentially expressed in FosTg mice (Fig. 7H). Importantly however, several enzymes regulating lipogenesis and elongation of long-chain fatty acids, namely Thrsp, Acly, Fasn and Elovl6 showed significantly decreased expression levels (Fig. 7H and Fig. S8B).

Fig. 7.

Impaired lipogenesis in the liver of Fos transgenic mice. (A) Q-PCR analysis for expression of the Fos transgene, endogenous Fos and total Fos in the liver of 16-week-old wt and FosTg mice. (B) Quantification of Fos protein levels in the liver of FosTg mice compared to wild-type. (C) Representative pictures of H&E stained sections of the liver from 16-week-old males of the indicated genotype. (D) Liver weight of wt and FosTg mice. (E) Glycogen content in the liver of 16-week-old wt and FosTg males. (F) Serum levels of non-esterified fatty acid and (G) triglycerides in the serum of 16-week-old wt and FosTg mice. (H) Q-PCR analysis for expression of genes controlling lipogenesis in the liver. (I) Quantification of Western Blot analyses for mTor, Jak1, Stat3 and Smad2 phosphorylation in liver samples of wt and FosTg mice. Tubulin detection was used as loading control. Data represent mean ± s.e.m. (n ≥ 6). Asterisks indicate statistically significant differences (****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05)

Mechanistically, we analyzed the activation of the Jak-Stat-pathway that is induced by Il6, for which the circulating levels correlate with the development of cachexia [44]. We neither observed an increased phosphorylation of Jak1 nor of mTor. In addition, the Stat3 phosphorylation levels were significantly decreased, possibly reflecting a compensatory effect of Fgf21. However, Smad2, which is activated by Tgf-ß, showed increased phosphorylation in FosTg liver samples compared to control samples (Fig. 7I and Fig. S8C).

Taken together, these data strongly suggest that the reduced adipose tissue mass in FosTg mice can be explained by increased serum concentrations of Fgf21, leading to reduced lipogenesis in the liver.

Reducing Fos-induced tumor burden leads to reduced Fgf21 levels and an ameliorated adipose tissue phenotype

To determine whether the metabolic phenotype developing in FosTg mice is a consequence of tumor growth, we genetically inactivated the Ribosomal S6 kinase (Rsk2) in Fos transgenic mice (FosTg;Rsk2^−/y mouse line) in order to reduce tumor development [32]. First, we confirmed our original observation that inactivating Rsk2 in FosTg mice strongly decreases tumor growth (Fig. 8A). We next compared the weight parameters of FosTg and FosTg;Rsk2^−/y mice to the two parental lines (wt and Rsk2^−/y), respectively. At the age of 16 weeks, body weight of FosTg;Rsk2^−/y animals was less decreased than the body weight of FosTg mice compared to the respective controls (Fig. 8B). Likewise, the fat pad weight was also reduced in FosTg mice lacking Rsk2, however significantly less pronounced than in FosTg mice compared to wt (Fig. 8C). Histological analyses of the epididymal WAT of Rsk2^−/y compared to FosTg;Rsk2^−/y mice at the age of 16 weeks suggested a reduced cell size in the white adipose tissue of FosTg;Rsk2^−/y mice (Fig. 8D). Quantification showed, that while the repartition curve of the adipocyte diameter in the Rsk2^−/y animals was comparable to wild-type mice, a decreased cell size was also evident in the FosTg;Rsk2^−/y mice, that was, nevertheless, less pronounced than in FosTg mice (Fig. 8E). Thus, the fat phenotype observed in the Fos transgenic mice is associated with the tumor growth, thereby suggesting that the metabolic phenotype observed in these mice is linked to tumor cachexia.

Fig. 8.

The adipose tissue phenotype of Fos transgenic mice is caused by osteosarcoma formation. (A) Representative X-rays of 16-week-old FosTg and Rsk2-deficient FosTg males (FosTg;Rsk2^−/y). (B) Body weight of 16-week-old FosTg and FosTg;Rsk2^−/y males relative to the respective parental line (wt or Rsk2^−/y). (C) Epididymal white adipose tissue (WAT) weight relative to the respective parental line in the same mice. (D) Representative pictures of H&E-stained sections of epididymal WAT from 16-week-old males of the indicated genotypes. (E) Repartition curves of the adipocyte diameter in epididymal WAT of 16 to 20-week-old males of the indicated genotype. (F) Elisa-based quantification of serum Fgf21 levels in FosTg and FosTg;Rsk2^−/y relative to the respective parental line (wt or Rsk2^−/y). (G) Correlation between serum Fgf21 levels and tumor area per vertebral body of FosTg mice. Data represent mean ± s.e.m. (n ≥ 6). Asterisks indicate statistically significant differences (****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05)

Next, we measured the concentration of circulating Fgf21. In accordance with the association of reduced adipose tissue mass and tumor burden, Fgf21 serum levels are less increased in FosTg; Rsk2^−/y mice than in FosTg mice compared to the respective control (Rsk2^−/y mice and wt mice, respectively) (Fig. 8F). In contrast, Tnfa, that was reported to be regulated by Fos [45] and potentially responsible for transmitting the effect, was not affected (Fig. S9). Finally, we plotted Fgf21 serum levels of FosTg mice to the tumor area in the vertebral bodies of the same mice. Importantly, we observed a direct correlation between serum Fgf21 levels and tumor burden in the respective mice (Fig. 8G).

Therefore, these data uncover a new phenotype of transgenic mice over-expressing Fos, that in addition to osteosarcoma development show a reduced adipose tissue mass but also develop an increased glucose tolerance and insulin sensitivity that can be explained by an increased concentration of circulating Fgf21. Most importantly, serum levels of Fgf21 are correlating with tumor burden in FosTg mice.

Discussion

The analyses of loss- and gain-of-function of AP-1 transcription factor family members in mice showed not only the important role of AP-1 in the regulation of bone mass and cancer development, but also as cell-autonomous and/or systemic regulators of fat metabolism and adipose tissue mass [4, 36, 46–49]. In this publication, we show that a further member of the AP-1 transcription factor family, namely Fos, that, when overexpressed, causes osteosarcoma formation [7], is regulating lipid metabolism and insulin tolerance. Mechanistically, we found an elevated Fgf21 level in the serum of FosTg mice compared to control animals that can explain the phenotype of reduced lipogenesis in the liver and a reduced fat content in FosTg mice.

Fos transgenic mice show a reduced white adipose tissue mass during aging associated with a reduced adipocyte size. This adipose tissue phenotype does not seem to be caused by a cell-autonomous mechanism, unlike what was observed in mice overexpressing Fra1 [36]. A cell-autonomous effect is also contributing to the phenotype described for mice with an adipocyte-specific deletion of Fra2 as well as for DeltaFosB overexpressing mice [47, 48]. The non-cell-autonomous nature of the fat phenotype in FosTg mice is supported by several arguments: First, the expression levels of adipocyte differentiation and maturation markers were unchanged in adipose tissue of FosTg mice. Second, overexpression of Fos did not block the differentiation capacity of adipose derived stem cells in vitro. And third, common signaling pathways and gene induction were not affected in stimulation experiments.

In our study, we analyzed H2kb-FosTg mice, a mouse model where Fos is ubiquitously over-expressed under the H2Kb promoter [7, 30], including expression in white adipose tissue, but also in liver. A further site of transgene overexpression is bone tissue, leading to spontaneous osteosarcoma formation [50]. Given the ubiquitous over-expression of Fos, we performed serum antibody microarray analyses where we found Fgf21 levels to be increased, a result that we confirmed by ELISA as well as on RNA level. The described effects of increased Fgf21 concentrations, namely improved insulin sensitivity and glucose clearance as well as decreased lipogenesis in the liver, mirrored the phenotype of FosTg mice.

Fgf21 expression in the white adipose tissue was described to be regulated by Pparg, a transcription factor activated by the AP-1 member Fos [51, 52]. While we found Pparg expression to be slightly elevated in white adipose tissue, and therefore cannot exclude the Ppar signaling pathway to mediate the metabolic phenotype of FosTg mice, we could not observe an increase in the RNA level of Ppara in the liver, that is known to regulate Fgf21 expression in this tissue [53]. Therefore, we suppose a Ppar-independent regulation of Fgf21 by Fos. Indeed, a direct regulation of Fgf21 expression by Fos can be suggested, given that Xiao et al. also showed the occurrence of an AP-1 binding site in the promoter of Fgf21, that is bound by the AP-1 member cJun, thereby regulating adipose tissue mass by directly controlling Fgf21 expression [54]. Furthermore, Zhang et al. demonstrated in vitro that AP-1 activation is required for the chemotherapeutic drug cisplatin to induce Fgf21 expression and that TPA, a phorbol ester that activates AP-1 signaling, can induce Fgf21 expression, effects that are normalized by pre-treatment with an AP-1 inhibitor [55]. However, if the AP-1 member Fos directly induces the expression of Fgf21 and therefore, whether the relationship between Fos and Fgf21 is causal or correlational has not yet been described and has to be further investigated.

Furthermore, in the FosTg mouse model, we observe a phenotype resembling the cachexia syndrome, which can develop in advanced stages of cancer. The state of chronic inflammation in cachexia, leads, for example via proinflammatory cytokines like Il6, to an activation of the Jak/Stat3 signaling pathway, causing muscle atrophy [56, 57]. Inhibitors targeting Jak1/2 have furthermore been described to block adipocyte lipolysis in cachexia [58]. In addition, Fgf21 was described to be regulated via Jak/Stat signaling by different cytokines [59, 60]. TNFa was additionally described to play a crucial role in cachexia and for the AP-1 member JunB, it has been demonstrated that adipose-specific deletion enhances thermogenesis [61] and that JunB-ko mice show a reduced adipose tissue mass that was explained by an increased lipolytic rate accompanied by an increased level of TNFa [46]. Increased Il6 levels could not be detected by proteomic analysis, and even though we observed an elevated level of Pnpla2, an enzyme catalyzing the first step of hydrolysis of triglycerides, in the white adipose tissue of FosTg mice (data not shown), Tnfa, while increased on RNA level in the liver, was not elevated in the serum of FosTg mice. Still, we observe a reduced phosphorylation of Stat3, possibly caused by a feedback mechanism of Fgf21 [62]. However, we found several additional cachexia-inducing factors, possibly responsible for the induction of the cachexia-like phenotype, to be upregulated in FosTg mice. Another aspect to consider is that AP-1 action has been shown to depend on the dimer composition [63, 64], and it was described that the combination of AP-1 family members is differentially regulating lipid metabolism [51]. Thus, we suppose that the composition of the AP-1 dimer can systemically regulate fat metabolism and further research would be needed to establish the specific function of Fos in the context of cachexia and Fgf21 induction.

The main limitation of our study is, that with our analyses we cannot exclude that a factor different from Fgf21 may influence fat mass and liver metabolism in FosTg mice. Therefore, a follow-up study should aim to directly test whether Fgf21 mediates the observed adipose tissue and liver phenotypes using Fgf21 knock-out studies or inhibitor treatments. Furthermore, Fgf21 has been associated with improved adipocyte function, systemic glucose and lipid clearance and energy expenditure. Therefore, we cannot exclude that Fos exerts its effects on adipose tissue via systemic metabolic alterations independent of the cachexic phenotype and a more detailed assessment of lipid and glucose metabolism should be performed in a model overexpressing Fos specifically in adipocytes. It is also important to state that, as metabolic alterations are caused by tumor development, it is possible that the adipose tissue phenotype of FosTg mice is a mere consequence of osteosarcoma progression [65]. Thus, we could show, by analyzing Rsk2-deficient Fos transgenic mice, that the reduced adipose tissue mass is linked to tumor development. Lack of Rsk2 in these mice led to a diminished tumor growth [32] causing a reduced body weight, adipose tissue mass and adipocyte phenotype intermediate to the phenotype of FosTg mice compared to wild-type mice. Furthermore, we could show that the level of Fgf21 is dependent on tumor size. In this context it is important to note, that Fgf21 is described to be secreted from several organs under stress conditions [66], and it is especially noteworthy that - while we observed an increased expression level of Fgf21 in the liver, the main physiological site of Fgf21 production - we found that Fgf21 expression is even further upregulated in osteosarcoma tissue. Therefore, while we proof an increase in Fgf21 production in FosTg mice, we do not demonstrate which site of expression is responsible for the observed metabolic phenotype.

Overall, cachexia is associated with a poor prognosis and increased mortality [67]. It has far-reaching negative effects on the quality of life and the treatment of cancer patients by impairing the tolerability and efficacy of cancer therapies [68–70]. Cachexia represents a major challenge in oncology treatment. A multimodal approach is recommended, which includes nutrition, physical activity and pharmacological interventions [71]. The detected elevated Fgf21 levels in patients with cardiac cachexia and cachexic geriatric patients [39, 72, 73] supports the use of Fgf21 as a biomarker for tumors with poor prognosis. On the other hand, as Fgf21 has been described to exert protective functions in cancer and cachexia [74], the elevated levels of Fgf21 might in our model reflect a compensatory response in FosTg mice to limit the loss of adipose tissue mass caused by tumor-induced cachexia and to ensure stable energy supply. This counteraction could be supported by the following reasons: (a) Fgf21 expression is induced during starvation affecting various tissues to overcome the deprivation status and (b) inhibition of Fgf21 was described to lead to a reduced muscle loss in cachexia [42]. Thus, better understanding the interaction of factors, such as Fos, Fgf21 and Ppars in the regulation of insulin, lipid metabolism and cancer associated cachexia can contribute to the identification of novel therapeutic targets for the treatment of cancer-associated metabolic syndromes.

In conclusion, our study not only shows that Fos is an important factor for the regulation of adipose tissue mass, glucose tolerance and insulin sensitivity as well as hepatic lipogenesis, but also that ubiquitous overexpression of Fos causes tumor-dependent Fgf21 activation. Thus, we suggest that the induction of Fgf21 expression by Fos-induced tumor formation is the key event modulating the metabolic phenotype associated with cancer cachexia.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

JL, AC and CB acquired, analyzed and interpreted the data. MN and TK contributed to data acquisition. JL, TS and MA designed the study and wrote the manuscript. All authors read and approved the final manuscript.

Funding

Open Access funding enabled and organized by Projekt DEAL. This project has received funding from the Deutsche Forschungsgemeinschaft (LU 2515/2 − 1).

Data availability

The dataset that supports the findings of this study is available from the corresponding author upon request.

Declarations

Ethical approval

All animal experiments were approved by the local animal facility and by the ‘Amt für Gesundheit und Verbraucherschutz’ (Org_633, Org_984, N22/059, G16/073).

Competing interests

The authors declare no competing interests.