GH-Dependent STAT5 Signaling Plays an Important Role in Hepatic Lipid Metabolism

Abstract

GH deficiency is known to be clinically associated with a high incidence of nonalcoholic fatty liver disease, and this can be reversed by GH administration. Here we investigated the mechanistic basis for this phenomenon using engineered male mice lacking different signaling elements of the GH receptor, hepatic stat5a/b^−/− mice and a mouse hepatoma line. We found deficient GH-dependent signal transducer and activator of transcription (STAT)-5 signaling correlates with steatosis, and through microarray analysis, quantitative PCR, and chromatin immunoprecipitation, identified putative targets of STAT5 signaling responsible for the steatosis seen on a normal diet. These targets were verified with liver-specific stat5a/b deletion in vivo, and in vitro we show that dominant-negative (DN) STAT5 increases lipid uptake in a mouse hepatoma line. Because loss of STAT5 signaling results in elevated STAT1 and STAT3 activity and intracellular lipid accumulation, we have used DN-STAT5a/b, DN-STAT1, constitutively active (CA)-STAT3, or addition of oleate/palmitate in the hepatoma line to assign which of these apply to individual targets in STAT5 signaling deficiency. These findings and published mouse models of steatosis enable us to propose elevated cd36, pparγ, and pgc1α/β expression as primary instigators of the steatosis along with elevated fatty acid synthase, lipoprotein lipase, and very low-density lipoprotein receptor expression. Decreased fgf21 and insig2 expression may also contribute. In conclusion, despite normal plasma free fatty acids and minimal obesity, absent GH activation leads to steatosis because activated STAT5 prevents hepatic steatosis. These results raise the possibility of low-dose GH treatment for nonalcoholic fatty liver disease.

Hepatic lipid metabolism is a complex process closely regulated at both the transcriptional and posttranscriptional levels (1, 2) by insulin and glucagon as well as lipid ligands [sterols and fatty acids (FAs)], which directly activate transcription factors/nuclear receptors. The latter include peroxisomal proliferator-activated receptor (PPAR)-α (NR1C1), PPARγ (NR1C3), liver X receptor a (NR1H3), the retinoid X receptor, the sterol response element binding protein (SREBP), the carbohydrate response element binding protein and hepatic nuclear factor 4a (NR2A1). Together with the glucocorticoid and thyroid hormone receptors, these regulate the key processes of FA uptake, storage, oxidation, and secretion.

The role of GH in hepatic lipid metabolism is currently proposed to be the opposite of its lipolytic actions in adipose tissue, i.e. it is thought to induce triglyceride (TG) uptake and storage by increasing lipoprotein lipase (LPL)/hepatic lipase expression (3). However, GH excess transgenic mice actually have decreased hepatic TG content (4), and humans with GH deficiency have fatty infiltration of the liver more frequently than those with normal GH secretion (5, 6). Furthermore, patients with nonalcoholic fatty liver disease (NAFLD) have lower plasma GH levels than controls (7). Cessation of childhood GH therapy on reaching final height is associated with the development of NAFLD in 29% of patients when surveyed on average 10 yr after ceasing GH replacement (8). Indeed, a clinical survey of glucocorticoid and thyroid hormone replaced hypopituitary individuals with NAFLD revealed that liver disease develops relatively quickly (mean 6.4 yr after diagnosis of pituitary dysfunction), with 60% of those biopsied displaying cirrhosis and 14% requiring liver transplantation or resulting in death from liver-related causes (5). Instituting GH replacement therapy for 6 months in a GH-deficient adult with NAFLD resulted in the reversal of NAFLD, concomitant with a reduction in markers of oxidative stress (9). In this case only a modest reduction in visceral fat (5.7%) was observed, whereas steatosis was reduced from 33 to 7%, fibrosis was reduced, and ballooning disappeared.

GH receptor (GHR) loss of function mutations (Laron dwarfism) also manifest NAFLD in adults, particularly males (four of five displaying NAFLD). Although these patients are obese, the presence or absence of fatty liver is not associated with degree of obesity or homeostasis model assessment insulin resistance index (10) and chronic replacement of Laron patients with IGF-I does not influence the NAFLD status. Unusually, serum adiponectin in Laron adults was reported to be elevated, rather than decreased, as also seen in GHR null mice (11). Although GHR null mice were reported not to be steatotic (11, 12), liver-specific deletion of the GHR results in marked steatosis, with elevated free fatty acid (FFA) and insulin resistance, which was not corrected by IGF-I infusion over 2 wk, despite normalization of serum IGF-I (13). Although this appears to support a role for GH in hepatic lipid metabolism, the elevated plasma GH resulting from the loss of hepatic IGF-I in these mice increased the release of FAs from the adipose reserves of these mice, producing a situation resembling high-fat feeding, leading to hepatic steatosis. Other mouse models are needed to uncover a direct role of GH in hepatic metabolism and reveal the signaling processes involved.

Signaling by the GHR is complex and involves the activation of Janus kinase (JAK)-2-signal transducer and activator of transcription (STAT)-5/STAT3/ERK/phosphatidylinositol 3-kinase (PI-3 kinase)/Akt pathways as well as the Src/ERK pathway (14, 15). The kinetics of activation of these pathways, particularly the activation of STAT5b, is critically dependent on the pulse pattern of GH secretion and is modulated by tyrosine phosphatases and suppressors of cytokine signaling (16, 17). A wide range of transcription factors are regulated by these pathways, including several of the hepatic nuclear factors, and these, along with STAT5b and its antagonist Bcl6, regulate the sexually dimorphic expression of hundreds of genes in the rodent liver (17, 18). The difficulty in associating particular signaling pathways with complex physiological processes such as growth and obesity led us to create lines of mice with loss-of-function mutations in different signaling domains of the GHR to implicate particular pathways in these processes (19, 20). In particular, we created mice with partial loss of STAT5 activation ability (truncated at residue 569 plus Y539/F and Y545/F), total loss of STAT5 activation (truncated at 391), and loss of STAT5 and JAK2 activation (Box1^−/−, 4 Pro/Ala) and have compared these with GHR null mice also lacking Src/ERK activation ability (19, 20). Because the mutant receptors are similarly disabled in adipose tissue, they do not enhance adipocyte lipolysis in the presence of elevated plasma GH. Using these mice, we aimed to identify the signaling elements used by GH to prevent hepatic steatosis, and because the study implicated STAT5, we verified our conclusions both in vitro and with hepatic stat5a/b-deleted mice.

Materials and Methods

Animals

Animals were housed and treated with ethics approval from the University of Queensland Animal Ethics Committee. Water and feed pellets were available ad libitum under a 12 h light, 12-h dark cycle at 20–22 C. Animals passed standard virus screens quarterly. Stat5a/b floxed mice were kindly provided by Lothar Hennighausen (National Institutes of Health, Bethesda, MD) and genotyped as described elsewhere (21), with backcrossing into C57BL/6J background for eight generations before being crossed with C57BL/6J albumin promoter-driven Cre transgenic mice from Jackson Laboratories (JAX® Mice and Services, Bar Harbor, ME). Mice with partial or total loss of STAT5 or JAK2 activation by GH are described elsewhere (19, 20) and have very low plasma IGF-I levels with marked impairment of postnatal growth. The mouse ghr mutants used in this study were the ghr-391 (truncated at cytoplasmic residue 391, not able to activate STAT5, but able to activate JAK2, PI-3 kinase, and ERK, ghr-box1^−/−) with four key proline residues in the JAK2 binding box 1 sequence mutated to alanine (unable to activate JAK2, STAT5, or PI-3 kinase but able to activate ERK), and ghr^−/−, total deletion of ghr expression (12). In all comparisons for GHR mutants, littermate wt controls of similar age were used.

High-fat diet

Male mice at 8 wk age were fed a normal control diet [4.6% fat (wt/wt), meat-free rat and mouse diet; Specialty Feeds, Glen Forrest, Western Australia] or a high-fat diet [22.6% fat (wt/wt), SF01-028; Specialty Feeds, Glen Forest, WA] for 9 wk ad libitum.

Lipid analyses

Hepatic triglyceride accumulation was determined by saponification in ethanolic KOH as described (22). Glycerol content was determined after neutralization with MgCl₂ using a serum triglyceride determination kit (Sigma, St. Louis, MO. Hepatic lipid accumulation was also assessed histologically by oil red O staining of frozen liver sections as reported elsewhere (23).

Plasma analysis

Plasma from fasted mice was tested for lipid profile and liver enzymes by the University of Queensland Diagnostic Services. Plasma adiponectin and leptin were measured with Linco kits (Linco Research, St. Charles, MO) and insulin with Mercodia ultrasensitive mouse insulin ELISA (Mercodia, Uppsala, Sweden).

Hepatic triglyceride secretion rate

The hepatic triglyceride secretion rate was determined by measuring the increase of plasma triglyceride concentration after an ip injection of Poloxamer 407 (Sigma; 1000 mg/kg body weight) as described elsewhere (24). Blood samples were collected immediately before injection and at 3 h and 6 h afterward from the tail vein. Triglyceride content of plasma was determined using a serum triglyceride determination kit (Sigma). The TG secretion rate (micromoles triglycerides per hour per kilogram body weight) was calculated as the increase in plasma TG concentration per hour multiplied by the plasma volume (3.5% of the body weight) and divided by the body weight.

Illumina microarray analysis

Liver tissue was collected from three 16-wk-old male wt, ghr-391, and ghr^−/− mice starved overnight (fasted array) or four 16-wk-old male wt, ghr-391, and ghr^−/− mice without fasting (fed array). RNA was extracted using a RNeasy (QIAGEN, Hilden, Germany) minikit, analyzed for quality, and then biotin-labeled amplified RNA was synthesized and hybridized to Illumina mouse WG6 version 1.1 array (San Diego, CA) as described previously (20), with expression analysis by GeneSpring GX (Agilent Technologies, Santa Clara CA) and DAVID (http://david.abcc.ncifcrf.gov) for Gene Ontology. Raw data from fasted animals are deposited at GEO GSE11396.

Transfection of hepatocytes with STAT5 and STAT1 expression constructs

Murine AML-12 hepatoma cells were cultured in DMEM/F12 containing 10% Serum Supreme (Cambrex, Walkersville, MD), insulin/transferrin/selenium, and dexamethasone (0.4 μg/ml). Cells were transfected with dominant-negative (DN) STAT5 (STAT5-DN) or DN-STAT1 using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). After an overnight rest, cells were washed in PBS and media were replaced with DMEM/F12 containing 10% Serum Supreme. After 24 h cells were washed with PBS and harvested into Trizol reagent (Invitrogen).

In vitro model of hepatic lipid accumulation

AML-12 cells were cultured in 50% F-12 Kaighn’s 50% Leibovitz’s L-15 media with 7% Serum Supreme. One millimole of lipid was added (2:1 ratio of oleate and palmitate in culture media with 1% BSA) for 24 h. Cells were washed in PBS and harvested into Trizol reagent. AML-12-GHR cells transfected with STAT5 DN were cultured in 0.5 mm lipid with 1% BSA for 24 h and lipid uptake visualized using oil red O staining as described (23). Quantification of staining was performed using Image J (National Institutes of Health, Bethesda, MD).

Quantitative PCR (qPCR)

Liver tissue or cells were homogenized in Trizol reagent and RNA as per the manufacturer’s instructions. Deoxyribonuclease-treated RNA was reverse transcribed using Superscript III (Invitrogen Australia Pty Limited, Victoria, Australia) and used for qPCR with Sybr Green technology (Applied Biosystems Inc., Foster City, CA). Primer sequences are given in Supplemental Table 1, published on The Endocrine Society’s Journals Online web site at http://endo.endojournals.org. Analysis was performed by calculating the change in cycle threshold between glyceraldehyde-3-phosphate dehydrogenase and the gene of interest and expressed as fold change relative to wild-type (wt) animals or vehicle-treated cells.

Chromatin immunoprecipitation (ChIP)

ChIP was performed according to Barclay et al. (25) using the following primers:

Primer set promoter 1A, forward, 5′-GCAAGTGAAGATGAGTGAACAG-3′, reverse, 5′-TCATCGTGGGAGCTCTAGCA-3′; primer set promoter 1B, forward, 5′-GGAGCAACTGGTGGATGGTT-3′, reverse, 5′-CAGTGTAAAGGCAGACAAATTTTT-3′; primer set promoter C, forward, 5′-CCTGGAGTTGGCGAGAAA-3′, reverse, 5′-TTCCGATCACAGCCCATTC-3′ (Supplemental Fig. 3).

Statistical analysis

Multiple comparisons were undertaken by ANOVA, with Tukey’s post hoc test. Comparisons of two groups used an unpaired Student’s t test.

Results

Livers of mice lacking ability to generate phospho-STAT5 via GHR are steatotic with normal serum lipids

Although not evidently obese by 4 months age, homozygous male mice lines lacking ability to activate STAT5 in response to GH (ghr-391, Box1^−/−, and ghr^−/−) were found to be markedly steatotic on a normal control diet as visualized by hepatic oil red O staining and tissue TG analysis (Fig. 1, A and B). Steatosis was exacerbated by feeding the mice a moderately high-fat [22.6% (wt/wt)] diet for 9 wk, although this also elevated wt hepatic TG so that the differences among mice strains were no longer significant, with the exception of the ghr-Box1^−/−. This is suggestive of an effect of GH-induced STAT5 activation on the uptake of lipids. Because the steatosis in ghr^−/− mice was similar to mice with the deletion of STAT5 signaling but JAK2, PI-3 kinase, and ERK signaling intact (ghr-391), we focused on STAT5 signal impairment as the likely cause of steatosis.

Fig. 1.

Histological and biochemical evidence for steatosis correlating with loss of GH-STAT5 signaling (A) oil red O staining of normal diet (ND) and high-fat diet (HFD) liver sections, 4 months of age showing steatosis in ND ghr-391, ghr-box1^−/−, and ghr^−/− mice and increased steatosis with HFD. B, Liver TG content of wt, ghr391, and ghr^−/− mice (n = 6) for ND and HFD. C, Liver TG content for stat5a/b^−/− mice at 4 months of age, normal diet (n = 6). ***, P < 0.001 vs. normal littermates.

Steatosis can lead to NAFLD, and hematoxylin and eosin staining in the 4-month ghr-391 and ghr^−/− mice revealed morphology suggestive of ballooning, a hallmark of advanced steatosis, particularly after a high-fat diet (Fig. 2A). This effect was more severe in the ghr-391 at 13 months of age, with advanced steatosis and moderate ballooning evident, whereas wt mice showed only minor ballooning (Fig. 2B). Serum liver enzymes [alanine transaminase (ALT) and aspartate transaminase (AST)] showed a trend to increase compared with wt control mice, but significance was not reached in ALT levels in the ghr-391 and the ghr^−/− mice until 13 months. Similarly, AST levels in the ghr^−/− mice became significantly different from wt mice by 13 months of age (Fig. 2C).

Fig. 2.

Hepatic pathology in mutant mice. A, Hematoxylin/eosin stained liver sections at 4 months of age, for normal diet (ND)-fed or 9-wk high-fat diet (HFD)-fed mice. No pathology is apparent in the ND-fed mice, whereas on the HFD, ballooning (indicative of hepatocyte apoptosis associated with increased cell size) and inflammatory foci (with granulocyte invasion) were seen in one ghr-391 mouse and in all the ghr^−/− mice. B, By 13 months of age on ND, severe steatosis and moderate ballooning were apparent in the ghr-391 mice, whereas wt mice showed only moderate steatosis with minor ballooning. C, Liver function markers ALT and AST for GHR mutant mice serum (n = 3–5) and (D) for stat5a/b^−/− mice (n = 5). *, P < 0.05, **, P < 0.01, ***, P < 0.001 vs. control littermates.

Mice lacking hepatic STAT5a/b have fatty livers with normal serum lipids

To confirm the role of phospho-STAT5 deficiency in steatosis, we examined hepatic stat5a/b^−/− mice for steatosis. By 4 months age, there was a marked (3-fold) increase in hepatic TG content in these mice (Fig. 1C), accompanied by a marked increase in ALT and AST suggestive of liver damage (Fig. 2D). These mice also displayed elevated plasma TG (Fig. 3A), not seen with whole-body deletion of GHR-activated STAT5 (Fig. 3B). Thus, the phenotype in the hepatic stat5a/b^−/− mice is more pronounced than the ghr-391 and ghr^−/− mice, although steatosis is a common feature.

Fig. 3.

Plasma hormones and metabolites in GHR mutant and stat5a/b^−/− mice. Plasma TG in stat5a/b^−/− mice and control littermate mice (A), plasma TG (B), plasma FFAs (C), plasma cholesterol (D), plasma insulin (E), plasma adiponectin (F), and plasma leptin (G) in STAT5 impaired ghr mutants and littermate controls are shown (in all cases, mean ± sem, n = 6). *, P < 0.05, by ANOVA; **, P < 0.01 by ANOVA.

Plasma metabolite and hormone levels are normal in STAT5-impaired ghr mutants

At 4 months age, ghr-391 and ghr^−/− mice have plasma TG, FFAs, cholesterol, and insulin in the normal range (Fig. 3, B–E). Plasma adiponectin is significantly elevated in the ghr^−/− mice but not in the ghr-391 mice (Fig. 3F), and leptin is increased in the ghr-391 mice but not the ghr^−/− mice (Fig. 3G). The latter differences may be indicative of non-STAT5 effects (e.g. PI-3 kinase or ERK actions) because both these mice lack GHR-induced STAT5 activation.

Illumina Microarray and qPCR verification

To elucidate the basis for steatosis, transcript analyses of STAT5-impaired ghr-391 and ghr^−/− mice were undertaken in both the fed and fasted state by Illumina Microarray (20). Gene transcripts that were altered 2-fold or more relative to wt are shown in Supplemental Table 2, A and B. Of particular interest were genes relating to lipid uptake and transport, including cd36, lpl, cpt1, and vldlr, genes relating to lipid synthesis including pparg, insig2, scd2, and fasn, and those relating to lipid oxidation including cyp4a14, cpt1, pgc1a, and pgc1b. No changes in srebp1c transcript were observed, indicating that enhanced insulin action is not involved here.

Transcript changes were verified using qPCR in ghr-391 and ghr^−/− mice (Supplemental Fig. 1) and in hepatic stat5a/b^−/− mice (Supplemental Fig. 2). Cd36, pparg, cpt1, lpl, vldlr, pgc1a, insig2, fgf21, and scd2 were all confirmed as altered in the livers of all mutant mice, whereas fasn, cyp4a14, and pgc1b were found to be altered only in the ghr-391 and ghr^−/− mice. Because of their importance, cd36 and pparg transcript changes were verified at both the RNA level and the protein level by immunoblot in STAT5-impaired ghr mutants (Fig. 4, A–C). Similar increases in cd36 and pparg expression were seen in the hepatic stat5a/b^−/− mice (Fig. 4D). Collectively these transcript and protein alterations suggest that loss of GHR-induced STAT5 activation could promote lipid uptake and intracellular transport, lipid synthesis, and β-oxidation.

Fig. 4.

PPARγ and CD36 transcript and protein are increased in STAT5-impaired GHR mutants and hepatic stat5a/b^−/− mice. Pparγ and cd36 transcript expression (A) and protein (B and C) in GHR mutants and Pparγ and cd36 transcript expression in hepatic stat5a/b^−/− mice (D) are shown. Mean ± sem, n = 4 for transcript and for protein (representative images of Western blots shown for individual mice). *, P < 0.05, **, P < 0.01, ***, P < 0.001 by ANOVA.

Regulation of CD36 by STAT5

Cd36 expression was altered in the ghr-391, ghr^−/−, and hepatic stat5a/b^−/− mice. The role of STAT5 in cd36 expression was investigated by ChIP analysis of liver tissue from wt and ghr-391 mice at 4 months of age. Primer sets were designed to span the three previously identified promoters, 1A, 1B, and C. STAT5 occupancy on liver-expressed promoter C (26) but not promoters A and B was significantly reduced in ghr-391 livers compared with wt livers (Fig. 5A), which correlates to increased cd36 expression in these mice in accordance with the negative regulation of cd36 by STAT5 reported here. Based on the position of the ChIP primers for promoter C, the most proximal consensus STAT5 element (TTC AGG GAA) is 655 bp from the ATG within the first intron of promoter C. There remains the possibility that the consensus HNF6 site at −705 from the transcription start site is a positive element for cd36 expression, and STAT5 is antagonizing its action by direct binding or binding to a nearby nonconsensus STAT5 sequence, as shown by Delesque-Touchard et al. (27) for the cyp2c12 promoter. To establish the effect of reduced STAT5 binding to promoter C on cd36 transcripts in these mice, promoter-specific qPCR primers were designed and tested on cDNA from ghr-391, ghr^−/− and hepatic stat5a/b^−/− liver (Fig. 5, B and C). Interestingly these mice showed increased expression from promoter A as well as C, despite loss of STAT5 binding being restricted to promoter C. Promoter B showed a weak but significant increase only with the ghr^−/− mice, implying it is not regulated by STAT5. Increased expression of promoter A for both ghr^−/− and ghr-391 mice suggests that a more distal repressive STAT5 site exists for promoter A or that increased PPARγ may be responsible for induction.

Fig. 5.

A, In vivo ChIP assay demonstrating loss of STAT5b binding to the CD36 promoter C in the livers of ghr-391 mice lacking the ability of GH to activate STAT5 (n = 4). *, P < 0.05. B, Promoter-specific transcript expression in STAT5-impaired ghr mutant mice. C, Promoter specific transcript expression in hepatic stat5a/b^−/− mice.

Increased activation of STAT1 and STAT3 in mutant mice with decreased GHR-induced STAT5 activation

The microarray analysis shows significantly decreased cis and socs2 expression and a strong trend to decreased socs3. As noted by Hosui and Hennighausen (28), this can lead to increased basal STAT1 and STAT3 activation by other cytokines because of decreased feedback by these suppressor of cytokine signaling (SOCS). Increased STAT1 or STAT3 activity could contribute to expression of steatosis related genes, so we measured phospho-STAT1 and -3 protein levels in 4-month livers by immunoblot. Figure 6A shows that phospho-STAT1 is significantly elevated 2- to 3-fold relative to wt littermates, and Fig. 6B shows that phospho-STAT3 is 2- to 4-fold elevated in these livers. Interestingly, STAT1and STAT3 proteins are also significantly elevated, which may be a further effect of loss of SOCS. However, the overall levels of phospho-STATs, being the biologically active form of the proteins, are of most relevance here.

Fig. 6.

Loss of GH-induced STAT5 activation leads to an increase in STAT1 and STAT3 phosphorylation. STAT1 (A) and STAT3 (B) phosphorylation is increased in mice lacking STAT5 activation by GH, but so is the protein expression. Sixteen-hour fasted mice (n = 6) (mean ± sem) are shown. *, P < 0.05, **, P < 0.01, ***, P < 0.001 by ANOVA.

Direct effect of STAT5 on lipid uptake

Given the complexity of the in vivo models of STAT5 depletion because of concurrent STAT1 and STAT3 activation, the direct effect of STAT5 inactivation on lipid accumulation was investigated in vitro. Mouse hepatic AML-12 cells were transfected with STAT5-DN in the presence of low-level exogenous lipid, and lipid accumulation was measured by oil red O staining. Compared with control cells, STAT5-DN caused increased lipid uptake (Fig. 7A). This effect can be attributed to lipid uptake rather than lipogenesis because it was not seen in the absence of exogenous lipids (not shown). In addition, the STAT5-impaired ghr mutants show normal hepatic triglyceride release (Fig. 7B).

Fig. 7.

A, Lipid uptake in the AML-12 cells is significantly enhanced by disabling STAT5 signaling with transfection of DN-STAT5 (n = 4). *P < 0.05. This also supports the AML-12 model as a means of assigning roles for lipid excess in transcript regulation. B, Hepatic triglyceride production rate in GHR mutant mice 0–3 h and 3–6 h after administration of Poloxamer 407 as described in Materials and Methods is unchanged compared with wt littermate controls (n = 4–6). C–E, Altered transcript expression in AML-12 hepatocytes of target genes identified in mice lacking GH activated STAT5 using STAT5-DN (C), STAT1-DN (D), and oleate/palmitate addition (E) as described in Materials and Methods.

In vitro validation of STAT-5, STAT-1, and lipid-regulated genes

Expression of STAT5-DN in AML-12 hepatocytes resulted in the direct up-regulation of only cd36, fasn, and cyp4a14 of the transcripts previously verified by qPCR (Fig. 7C). Given the increased STAT1 activation seen in the STAT5-depleted mutant mice, putative STAT1-regulated genes were examined by transfecting STAT1-DN into AML-12-cells. Pparg, lpl, vldlr, and pgc1a only were found to be directly down-regulated by STAT1-DN (Fig. 7D). There was no effect seen on candidate metabolic gene expression after transfection of STAT3-constitutively active (CA) into these cells (data not shown). Because ghr-391, ghr^−/− and hepatic stat5a/b^−/− mice display hepatic lipid accumulation, an in vitro model of steatosis in AML-12 was developed, involving addition of oleate/palmitate. Of the qPCR-verified genes, fasn, fgf21, gck, and insig2 were all regulated by lipid accumulation in a similar manner to the mice lines (Fig. 7E). Some genes could not be validated as STAT5, STAT1, or lipid regulated, and this may be attributed to limitations in the in vitro model. Overexpression of transcription factors cannot mimic the in vivo environment completely with regard to the temporal recruitment of cofactors, which may be essential for the expression of some genes.

Discussion

In this study we sought to determine the molecular basis for the steatosis seen clinically and in rodent models as a consequence of GH deficiency. To do this, we used our panel of male mice homozygously expressing GHR mutants with deletions of signaling elements in the receptor cytoplasmic domain and because these indicated STAT5 deficiency as key, hepatic stat5a/b-deleted mice. Microarray analysis and qPCR of liver was then used to identify putative target genes. We supported these in vivo studies with an in vitro GH-responsive murine hepatocyte line.

Steatosis was evident on a normal diet in the absence of GH-induced STAT5 activation (ghr-391, ghr-box1^−/−, and ghr^−/−). The key role of loss of STAT5 activation in GH-deficient steatosis was verified with hepatic stat5 a/b-deleted mice, which also showed marked triglyceride accumulation, confirming the earlier study of Cui et al. (29). The higher TG content of these livers is likely to be a result of the lipolytic action of elevated GH secretion in response to lowered hepatic IGF-I output (29). This situation also applies to the liver-specific GH receptor null mice (13), whereby enhanced GH-dependent lipolysis will drive hepatic lipid uptake in a similar manner to high-fat diet. This is not the situation in our GHR mutant mice (19, 20) because their enlarged adipose tissue is also refractory to the lipolytic actions of GH so that we are able to gauge more realistically hepatic-specific GH actions on lipid metabolism.

We sought a mechanistic basis for the steatosis with Illumina array analysis of the transcriptomes of fed and fasted mutant mice. The large number of transcripts regulated by GH pulses in the male mouse (17, 18) and the complexities of hepatic sexually dimorphic metabolism make this a difficult task. However, the most straightforward explanation of the steatosis based on the transcript profiling is that removal of GH-activated STAT5 leads to increased lipid uptake and synthesis associated with increased PPARγ activity and expression. This is supported by our demonstration that DN-STAT5 expression in the AML-12 hepatocyte line results in increased uptake and storage of FFAs, mirroring the in vivo results. The steatosis is not a consequence of increased food consumption because this is reduced (not shown) and not a consequence of elevated lipolysis because plasma FFAs are normal, and adipose depots are enlarged in the absence of GH-induced lipolysis in the GHR mutants (19). We cannot exclude decreased very low-density lipoprotein (VLDL) secretion as an additional mechanism (30) but have no microarray evidence for it (e.g. increased mttp transcript) and could measure no change in hepatic TG secretion in hepatic stat5^−/− mice. Concomitantly with the increased lipid uptake, lipid oxidation is increased [hence, the lowered respiratory exchange ratio (data not shown) and increased cpt1 and cyp4a14 expression]. This is presumably driven by increased polyunsaturated fatty acid (PUFA) uptake and relief of STAT5-dependent suppression of PPARα activity (31).

While we identified a number of genes regulating lipid synthesis and uptake as altered in GH-dependent or absolute phospho-STAT5 deficiency (Supplemental Table 2, A and B), less than half of these could be shown to be direct STAT5 targets in vitro using DN-STAT5 in the AML-12 hepatoma line. Identified STAT5 targets included FA synthase (fasn), previously reported to be repressed by STAT5a and GH in adipocytes (32, 33), and cd36, which we show here has GH-inducible STAT5 binding to proximal promoter C by in vivo ChIP. Several other key lipid-regulating genes that differed according to STAT5 generation in mice mutants but were unaffected by DN-STAT5 in vitro were shown to be significantly down-regulated by DN-STAT1 in vitro (PPARγ, Pgc1α, lpl, and vldl). This is concordant with the 3-fold elevated phospho-STAT1 that we observe in the livers of these mutant mice (Fig. 6), possibly as a result of the decline in SOCS and CIS expression. Although phospho-STAT3 is also elevated 3- to 4-fold in the liver of these GHR mutant mice, none of the identified lipid-related genes were altered in expression in vitro with constitutively active CA-STAT3 (34). We note that the elevated phospho-STAT1 and STAT-3 is largely a result of increased protein expression, which was observed previously by Cui et al. (29). We have previously reported that SOCS3 regulates STAT3 protein levels (25), and phospho-STAT1 protein has been reported to increase STAT1 transcript and protein (35), consistent with our findings.

Because of the steatosis evident in these livers, we investigated the response of lipid-metabolism genes to oleate/palmitate in vitro and found robust decreases in fgf21, gck, and insig2 and an increase in fasn expression. Fgf21 transcript, which is strikingly decreased in GHR mutants and stat5a/b nulls, was also decreased by lipid supplementation in vitro but not affected by DN-STAT5 or DN-STAT1. As well as suppressing sensitivity to GH action during starvation, Fibroblast growth factor (FGF)-21 is able to reverse hepatic steatosis and increase energy expenditure (36). Although deficiency of FGF21 is an appealing basis for steatosis, the profile of lipogenic genes altered by FGF21 injection (36) does not correlate with the changes we observed. Thus, we found no change in scd1, elovl6, and dgat1, all of which are decreased by FGF21 administration. Nevertheless, it is possible that FGF21 deficiency does contribute to steatosis here, particularly by increasing SREBP1/2 activity. Similarly, the decrease in insig2 transcript seen with loss of STAT5 signaling in the mutant mice would be expected to result in increased lipid synthesis through relief of SREBP1/2 inhibition, yet the absence of changes in SREBP target genes scd1, elovl6, and dgat1 suggest that it is not a major mechanism for the steatosis.

Central to our putative model for the origins of steatosis in GH deficiency (Fig. 8) is an elevation of hepatic PPARγ1 activity, which would increase lipid uptake and synthesis. Adenovirus-mediated overexpression of PPARγ1 in mouse liver is known to cause steatosis (37), whereas liver-specific deletion of PPARγ reduced steatosis in lipoatrophic AZIP mice (38). The beneficial actions of thiazolidonediones in hepatic steatosis are thus likely to be a consequence of their actions on adipose tissue and muscle. Our STAT5-DN data showing no effect on PPARγ expression in vitro and the published literature suggest the 3-fold elevation in PPARγ transcripts in our mice is not a direct consequence of loss of STAT5 inhibition, but rather a consequence of elevated phospho-STAT1 (because CA-STAT1 increases PPARγ1 transcript in vitro) and, potentially, elevated stearoyl-CoA-desaturase-2, which enhances PPARγ transcript expression in preadipocytes (39). Scd2 is not normally expressed in adult liver but is the major form for monounsaturated FA synthesis in fetal liver and may be elevated as a consequence of elevated cholesterol and saturated fats in these steatotic livers (40). In addition, increased 17α-hydroxypregnenolone could be expected to result from the sharp decrease in 3β-hsd expression, and this sterol is a ligand for the pregnane X receptor (PXR), which is reported to up-regulate PPARγ expression (41).

Fig. 8.

Putative model for steatosis in GH deficiency based on transcript changes in mouse models of deficient STAT5 activation. This shows mechanisms that could account for enhanced lipid uptake and synthesis regulated directly and indirectly by STAT5. See text of Discussion for details.

As well as increased PPARγ expression, a number of factors will contribute to enhanced PPARγ activity. Elevated PUFA and acyl-PUFA resulting from enhanced CD36-mediated lipid uptake would be expected to activate PPARγ (42), and increased expression of the PPARγ coactivators pgc1α and -1β would strengthen this activation. In addition, the elevated Rdh6 expression would increase availability of 9-cis retinoic acid, the ligand for the retinoid X receptor heterodimeric partner of PPARγ (40). Finally, the antagonism between active STAT5 and PPARα mentioned above applies to PPARγ (31), further increasing PPARγ activity in the absence of phospho-STAT5.

Together with de novo synthesis of FFAs, hepatic FFA uptake also plays an important role in promoting NAFLD (43). Thus, increased hepatic expression of cd36 is seen in diet-induced obesity associated with increased FA uptake, TG content, and steatohepatitis. Exogenous adenoviral expression of cd36 in the liver (5-fold increase) recapitulates the increased FA uptake and elevated hepatic TG even in lean mice (44). Although increased lipid uptake via increased cd36 expression can result from PPARγ activation (40), other factors directly regulate cd36 expression, including (as we show here), STAT5. DN-STAT5 induces cd36 in vitro, and STAT5 binds to proximal promoter C of CD36 in wt mice by ChIP but does not bind in mutants lacking active STAT5a/b. In vivo, promoter A expression is also increased in the absence of phospho-STAT5, potentially as a result of upstream inhibitory STAT5 elements or elevation of PPARγ activity. Ligand induced activation of PXR (see above) would also be predicted to increase cd36 expression through PXR/DR-3 elements in the CD36 promoter in a liver-specific manner (45).

Lpl is expressed at a very low level in normal liver but is markedly elevated in morbidly obese patients with NAFLD and decreased after bariatric surgery. Antibody neutralization studies indicated LPL activity is similar to hepatic lipase activity in these patients (46). This elevation confirmed a report of increased lpl expression in NAFLD in which hepatic fat content correlated with lpl expression (47). Likewise in mice, liver-specific overexpression of lpl results in hepatic steatosis (48). Here we found robust induction of lpl transcript expression in both ghr-391 and hepatic stat5a/b^−/− mice. This appears tissue specific because we did not observe altered lpl expression in the adipose or muscle of these GHR mutants (not shown). In vitro analysis suggests that increased lpl expression is a result of elevated STAT1 activity. LPL is responsible for hydrolyzing the TAG in circulating VLDL and chylomicrons, facilitating FFA uptake, and we also found vldl receptor expression markedly increased in both ghr-391 and hepatic stat5a/b^−/− mice. Given that VLDL is taken up by hepatic low-density lipoprotein receptor, which shows unaltered expression in our mice, and that the VLDL receptor is normally restricted to peripheral tissues or macrophages, it is difficult to assess the significance of this conjoint elevation. However, adenoviral-mediated hepatic expression of vldlr resulted in a marked increase in hepatic intermediate-density lipoprotein uptake and a 50% reduction in plasma cholesterol in LDLR null mice (49). It may be that hepatic postprandial chylomicron uptake is facilitated by the cooperation of LPL and VLDL receptor in GH deficiency (50).

Mice lacking GH-induced STAT5 activation did not display features of steatohepatitis at 4 months age. Even when fed a moderately high-fat diet, hepatocyte ballooning and inflammatory foci were minimally elevated in comparison with wild-type animals. Liver function markers were not significantly increased at 4 months in the mutants, but they were at 13 months age on a normal diet. We could not find evidence of fibrosis, even in these 13-month-old mice, presumably because a second hit is needed, such as serial CCl4 injections in mice. Indeed, Hosui et al. (51) were able to show very recently that fibrosis and TGFβ protein expression resulting from CCl4 injection are enhanced in hepatic stat5a/b^−/− mice, with cancer development in a portion of these.

In conclusion, the STAT5 mutant mice lines studied here have provided a rational basis for understanding the key role of loss of STAT5 signaling in GH-deficient steatosis, and our model is described in Fig. 8. Considering the key role for GH-activated STAT5 in the prevention of hepatic steatosis, low-dose GH treatment may warrant investigation as a novel therapeutic option for patients with NAFLD, which has been successful on at least one occasion (9).

Acknowledgments

We thank Dr. Agnes Lichanska for her valuable input in the early stages of this project.

This work was supported by National Health and Medical Research Council (Australia) Grant 401668 (to M.J.W.).

Disclosure Summary: None of the authors have a financial conflict of interest regarding any company to disclose.

^*

J.L.B. and C.N.N. contributed equally to this work.

Abbreviations

 
• ALT

  Alanine transaminase

 
• AST

  aspartate transaminase

 
• CA

  constitutively active

 
• ChIP

  chromatin immunoprecipitation

 
• DN

  dominant negative

 
• FA

  fatty acid

 
• FFA

  free fatty acid

 
• FGF

  fibroblast growth factor

 
• GHR

  growth hormone receptor

 
• JAK

  Janus kinase

 
• LPL

  lipoprotein lipase

 
• NAFLD

  nonalcoholic fatty liver disease

 
• PI-3 kinase

  phosphatidylinositol 3-kinase

 
• PPAR

  peroxisomal proliferator-activated receptor

 
• PUFA

  polyunsaturated fatty acid

 
• PXR

  pregnane X receptor

 
• qPCR

  quantitative PCR

 
• SOCS

  suppressor of cytokine signaling

 
• SREBP

  sterol response element binding protein

 
• STAT

  signal transducer and activator of transcription

 
• TG

  triglyceride

 
• VLDL

  very low-density lipoprotein

 
• wt

  wild type