During starvation, nutrient metabolism should adapt in order to preserve life in the absence of caloric intake. The initial response to fasting is hormonal regulation (e.g. glucagon, catecholamine) to maintain the blood glucose level. Soon thereafter, the glucose insufficiency triggers catabolic reactions such as breakdown of glycogen stored in liver and muscle, fat from adipose tissue, and amino acid from skeletal muscle. As a result, fatty acids are released into the systemic circulation to feed peripheral tissues and liver. Liver assimilates fatty acids and generates ketone bodies, which covers two-thirds of the caloric requirement of the brain during the period of fasting (Cahill Jr, 2006). Overall, anabolic reactions regulated by insulin signaling are shifted to catabolism (e.g. fatty acid oxidization in mitochondria and peroxisomes).

The fatty acid oxidation and ketone body producing enzymes in the liver are induced mainly by peroxisome proliferator-activated receptor alpha (PPARα) (Rakhshandehroo et al., 2010), while fatty acid/sterol synthetic enzymes regulated by sterol regulatory element-binding transcription factor (SREBF)1c and SREBF2 are all reduced during fasting (Hakvoort et al., 2011; Schupp et al., 2013; Zhang et al., 2011). As lipid oxidation requires many oxido-reductive enzymes, the cytochrome P450 family enzymes and antioxidant proteins are also induced during starvation. Along with these fasting responses, Tis21/Btg2/Pc3 is induced in liver, muscle, and adipose tissues in mouse models in response to starvation (Kim et al., 2014; Schupp et al., 2013).

TPA-Inducible Sequence 21 (TIS21) has been reported as one of the primary response genes induced by the tumor promoter, 12-O-tetradecanoylphorbol-13-acetate (TPA), in mice (Fletcher et al., 1991). TIS21 expression is differential in mouse thymic carcinoma tissues and human lung cancer cells compared with those of the normal tissues (Lim et al., 1995), suggesting its role as a potential tumor suppressor in the thymus and lung. B-cell translocation gene 2 (BTG2) and PC3 are ortholog genes of TIS21 in human, and rat, respectively (Bradbury et al., 1991; Rouault et al., 1996), that regulates cell proliferation, apoptosis and cellular senescence in p53-dependent and -independent manners (Lim, 2006). Loss of BTG2 expression has been reported in various tumors (e.g. prostate, kidney, and breast) and its anti-carcinogenic mechanisms are evidenced in in vivo knockout mice and in vitro cell culture analyses (Choi et al., 2016; Lim, 2006; Mao et al., 2015; Park et al., 2008).

Recent reports suggested that TIS21 activates gluconeogenesis. The overexpression of BTG2 in AML-12 immortalized mouse liver cells increases glucose output by inducing the transcription of gluconeogenic genes. The same study documented the same results in a mouse model (Hwang et al., 2012; Kim et al., 2014). However, there no study has addressed the regulation mechanism of lipid metabolism by the TIS21 gene under fasting condition.

In this paper, we report the regulation of lipid metabolism by the TIS21 gene using the TIS21-knockout (TIS21-KO) mouse model by employing microarray analysis. The fatty acid oxidation and synthesis pathways were significantly affected in the TIS21-KO mice.

Animal

All mice used for the present study were maintained under the SPF condition at the Ajou University Animal Care Center under constant temperature and constant humidity with light/dark cycle 12/12 h, starting at 7 o’clock in the morning. WT and TIS21-KO mice were generated previously in the C57BL/6 background (Park et al., 2004). All mice were maintained ad libitum before fasting experiment for the indicated duration (20–48 h). All of the animal procedures were followed by Ajou University Institutional Review Board.

Microarray & data processing

Total cellular RNAs were extracted from livers of the WT and TIS21-KO adult male mice (10 week-old, male) and fetal livers of the 13.5 day-old embryos using TRIzol (Invitrogen, USA). To minimize variation of metabolism and to synchronize metabolic status in the mice, all mice were fasted for 20 h (from 5:00 pm to next day 1:00 pm), and the fetal livers derived from female mice (n > 3) were pooled before RNA extraction. These samples were processed and hybridized on Agilent SurePrint G3 Mouse GE 8X 60K array covered by the Gasket 8-plex slide (Agilent technologies, USA), then data were processed and analyzed using GeneSpring GX12.6 (Agilent technologies, USA). The methods and data processing procedures are described elsewhere (Lee et al., 2012), and raw data is available on NCBI’s Gene Expression Omnibus (GEO) as GSE105772.

Data analysis

After selecting the differentially expressed genes (DEGs; ± 1.5 fold, p < 0.05) between the TIS21-KO vs. WT mice, the data was analyzed by DAVID functional annotation tool to find out highly enriched gene ontologies (GOs) (Huang et al., 2008; 2009). The whole DEGs were listed in Table 1, and arbitrarily segmented based on their transcription amplitudes (Table 2). In addition, the relative expression folds were analyzed by performing Gene Set Enrichment Analysis (GSEA) (Mootha et al., 2003; Subramanian et al., 2005). To search the upstream transcription factors for the DEGs, enriched transcription factor binding site (TFBS) of the affected genes were queried to DAVID functional annotation based on the UCSC-TFBS database. The metabolic pathways were illustrated by Cytoscape 3.2.1 based on KEGG pathways (Kanehisa and Goto, 2000; Kanehisa et al., 2016; Shannon et al., 2003).

RNA extraction and Real-time quantitative PCR

Total cellular RNAs were isolated with RNAiso plus^™ (Takara, Japan) according to the manufacturer’s instruction. Then 1.0 μg of total RNAs were reverse-transcribed with Oligo(dT)₁₈ by using PrimeScript reverse transcriptase (Takara, Japan), and the cDNAs were amplified with 2X qPCR mix (Nanohelix, South Korea). qPCR reactions were performed by CFX96 Touch^™ Real-Time PCR Detection System (BioRad, USA). Sequences of the primer sets are listed on the Supplementary Table S1.

Quantification of triglyceride and cholesterol contents in liver

Lipid contents were extracted according the method described (Folch et al., 1957; Newberry et al., 2003); liver homogenates isolated from the 10 week old male mice were prepared in 10 volumes of 1× PBS by using Dounce homogenizer, and protein concentrations were determined by Bradford method. The homogenates (300 μl) were added to the mixture of chloroform-methanol (2:1 v/v, 5 ml) plus 0.1% sulfuric acid (0.5 ml), and then incubated with agitation for ~2 h at room temperature. The organic phase was dried by SpeedVac under low vacuum. Extracted lipids were re-suspended in 1× PBS (300 μl) with 2% Triton X-100, and the triglyceride and cholesterol levels were measured by colorimetric assay in the Ajou University Hospital. Final contents were normalized based on the protein concentration of the initial homogenates.

Statistical analysis

Data were analyzed with two-sided Student t-test, and the p-values < 0.05 were considered as significant between the two comparing groups.

TIS21 gene knockout changes metabolic profiles in mouse liver after fasting stimulation

To investigate any effect of TIS21 on nutrient metabolism, TIS21-KO and wild type (WT) male mice were fasted for 20 h before sacrifice and the gene expression profiles were evaluated by microarray analysis. Furthermore, 13.5 day old embryos were evacuated from the womb to examine the activity of the TIS21 gene in the fetal stage. mRNAs isolated from livers of the ‘KO-adult’, ‘WT-adult’, ‘KO-fetus’, and ‘WT-fetus’ were screened, and the effects of the TIS21 gene after 20 h of fasting were compared with that of the WT, e.g. KO-adult vs. WT-adult, and KO-fetus vs. WT-fetus (Fig. 1A). Among the 55,681 probes provided by the microarray chip, 26,988 probes were positive at least in one group. By applying a cutoff threshold ± 2-fold with p < 0.05 difference, 1,109 genes and 280 genes were differentially expressed in the KO-adult and the KO-fetus, respectively. Expression of hepatic Tis21 mRNA was 6 times higher in the fetal stage than the adult stage (Supplementary Fig. S1A), which is consistent with a previous report of observations using in situ hybridization of whole embryos (Lim, 2006). Other adult- or fetal-stage selective genes (>±10 times in adult/fetus) were 98.2% identical between WT and TIS21-KO mice (Supplementary Fig. S1B). Among those, 90% of the genes were positively-correlated and 8.2% were negatively-correlated with the reference gene set reported previously (Li et al., 2012).

When we evaluated the differentially expressed genes (DEGs) between the WT and TIS21-KO mice using the DAVID web-based program with a less stringent criterion (over ±1.5-fold, KO/WT with p < 0.05), 1,680 genes were upregulated and 1,006 genes were downregulated in adult mice livers (Fig. 1B). The expression of Tis21 was significantly increased after 20 h fasting in adult mice (Fig. 1C), suggesting that TIS21 gene may have a role in liver of mice in response to fasting stimulation.

Metabolism-regulating genes require TIS21 expression under fasting condition

To further investigate effect of TIS21-KO under fasting condition, the DEGs were analyzed by DAVID and the GO terms are listed in Table 1. In the biological process, metabolism of xenobiotics by cytochrome P450, immune response, and steroid biosynthetic process were prominent, and microsome and lysosome component was upregulated, and glutathione transferase activity was high in the molecular function. On the other hand, genes in the PPAR signaling pathway and fatty acid metabolism were significantly reduced in the TIS21-KO than the WT in response to fasting stimulation (Fig. 1B, Table 1). When the DEGs were ranked based on the amplitude of each gene expression from the highest to the lowest, the top 5% of the genes were clustered in the metabolism of xenobiotics and fatty acyl metabolic process (Fig. 1B, Table 2). The top 30% of the genes in the middle rank exhibited immune response regulation, such as antigen processing and presentation along with complement and coagulation cascades. Genes regulating cell division cycle and signaling pathways were expressed much less. Although the intensity of an mRNA transcript does not always directly reflect its protein level, our data was enough to assume a major role of TIS21 gene in the liver under starvation. The collective data strongly suggest that xenobiotic metabolism and fatty acid oxidation might require TIS21 gene expression under starved condition.

PPARα-target gene expression is reduced in the TIS21-KO mice under fasting

To investigate the metabolic pathways that are significantly changed in the TIS21-KO mice under fasting condition, the DEGs were further analyzed by gene set enrichment analysis (GSEA) and UCSC-TFBS analysis. As shown in Figs. 2A and 2B, gene expression under SREBF1/2 was significantly increased, as opposed to lack of the genes regulated by PPARα in the TIS21-KO mice (Supplementary Fig. S2). Indeed, expression of Fasn, Scd1, and Scd2 regulating fatty acid synthesis was significantly increased, in contrast, fatty acid oxidation genes in mitochondria and peroxisome were clearly downregulated in KO mice. Overall metabolic process observed in liver of the TIS21-KO mice are depicted in Fig. 2C, which show increased SREBF1/2 targets and reduced PPARa target gene expression. The scheme also revealed that glucose metabolic pathway such as glycolysis/gluconeogenesis or tricarboxylic acid cycle, were unaffected in the TIS21-KO liver with 20 h fasting. The DEG changes between the KO and WT were validated by RT-qPCR analyses (Figs. 2D–2E). When the promoters of DEGs were examined by the UCSC-TFBS analysis, PPARα was the most frequently observed transcription factor interacting with promoters of the DEGs, followed by STAT5B and NF-κB (Table 3, with selections from the transcription factors shown in Supplementary Fig. S3). The observations were the same in both adult and fetal livers. In summary, TIS21-KO liver increased the expression of fatty acid/cholesterol synthesis enzymes (SREBF1/2 targets), whereas fatty acid degrading enzymes (PPARα targets) were significantly reduced in the TIS21-KO liver compared to those of the WT under fasting condition.

Starvation-induced fatty change is diffuse in WT liver, but focal in TIS21-KO mice

When the fat deposition in liver was evaluated by Oil red O staining, 24 h fasting significantly increased lipid deposition in the livers of both WT and TIS21-KO mice compared to those of non-fasted mice (Fig. 3A), and the fatty change was accompanied by the increase of triglyceride (TG) and cholesterol after 24 h fasting (Fig. 3E). However, the increase was not significantly different between the WT and TIS21-KO mice. Immunohistochemisty analysis by hematoxylin and eosin staining revealed that 24 h fasting significantly induced microvesicular fatty change in both male and female WT mice (Figs. 3C and 3D). The fat distribution was diffuse and homogeneous in the WT, whereas it was focal in TIS21-KO liver at the centrilobular zone of both male and female mice (Supplementary Table S2 and Supplementary Fig. S4A). Oil red O staining in female liver showed that lipid deposition was more severe in the TIS21-KO mice than that of the WT under fed condition (compare a and b panels, Fig. 3B). We speculate that fasting-induced uneven or patchy like change and whitish mottling in the liver (Supplementary Fig. S4B) may reflect the focal fat deposition in TIS21-KO mice at the centrilobular area (Supplementary Table S2). In summary, there was no pathologic change in liver of TIS21-KO mice, except starvation-induced focal fat distribution described above.

Cathepsin E expression is highest in the DEGs of adult and fetal livers in TIS21-KO mice

To figure out the effect of TIS21 gene knockout on liver development, the DEGs (KO vs. WT) obtained from both adult and fetal livers were examined and then selected the commonly increased 9 genes and the decreased 12 genes in the TIS21-KO mice compared to the WT were determined (Figs. 1A and 4A). Using RT-qPCR, the highest expression of cathepsin E (Ctse) and the lowest expression of glutaredoxin 2 (Glrx2) were proven in the TIS21-KO mice compared to WT mice (Fig. 4B and Supplementary Table S3). Expression of cathepsin E was significantly increased not only in the adult and fetal livers but also in other lymphoid organs, such as thymus and mature splenocytes (Fig. 4C).

In the present study, the effect of TIS21 gene on energy metabolism has been investigated using microarray analysis in adult and fetal livers dissected from the whole body of TIS21-KO mice after overnight fasting stimulation. The liver was the focus of study because it is the center of metabolism and since TIS21 is highly expressed in fetal liver (Lim, 2006; Terra et al., 2008). By analyzing gene ontology terms, we found that TIS21-KO significantly increased xenobiotic metabolism by cytochrome P450 and immune response rather than the regulation of cell division cycle after overnight fasting (Tables 1 and 2). In addition, the PPARα signaling pathway and fatty acid catabolism were significantly inhibited in the TIS21-KO mice compared to those of the WT mice (Fig. 2, Table 3). GSEA revealed that fatty acid/steroid synthesis and fatty acid oxidation were inversely regulated in the TIS21-KO mice compared with WT mice under fasting condition. Increased Oil-red O staining in the female KO mice (Fig. 3B) might be concordant with the GSEA data shown in Fig. 2A. The data strongly suggests that TIS21 might be essential in energy metabolism upon starvation by regulating lipid metabolism.

Starvation stimulates autophagy in the liver (Komatsu et al., 2005; Mizushima et al., 2004; Rabinowitz and White, 2010), and the response requires nutrient sensing nuclear receptors such as PPARα (Lee et al., 2014) Hence, the suppression of PPARα pathway in TIS21-KO mice can be linked to suppression of the autophagy reaction. We previously described that mechanistic target of rapamycin (mTOR) signaling is constitutively activated in TIS21KO-mouse embryonal fibroblasts in response to estradiol (Kim et al., 2008). Therefore, we speculate that the autophagy reaction can be reduced in liver of TIS21-KO mice under fasting condition. In C57BL/6 mice, 24 h-fasting significantly reduced body weight (13%) compared to that of mice that were fed, and the weight loss was larger in WT mice than in KO mice. The smaller weight loss under the starvation condition was also evident in the weight of organs such as the liver and gonadal white adipose tissues, of the KO mice than the WT mice (Supplementary Fig. S5). This observation indirectly supports the failure of autophagy and lipid catabolism in the TIS21-KO mice compared to the WT. In the TIS21-KO mice, compromised lipid catabolism may lead to failure of energy generation after starvation. However, regulation of gluconeogenesis through Nur77 and blood glucose level (Kim et al., 2014) were not significantly changed in the TIS21-KO mice compared with that of WT after 24 h fasting (data not shown). The collective data imply the role of TIS21 expression in maintaining energy level under stressful conditions by regulating fatty acid oxidation in the liver.

Cathepsin E (Ctse) is as a member of the aspartate endopeptidase localized in the late endosome or plasma membrane, and mostly within antigen-presenting cells (Yamamoto et al., 2012). It is highly expressed in gastrointestinal mucosa and lymphoid organs including the fetal liver, but obviously not in adult liver (Chlabicz et al., 2011; Kageyama et al., 1998). Presently, Ctse expression was very high in adult liver as well as lymphoid organs of TIS21-KO mice (Fig. 4C). Therefore, we can suggest a significant role of the TIS21 gene in regulating Ctse expression in lymphoid organs and adult liver during starvation. Although we cannot suggest a reason yet, a plausible mechanism is that activation of cytochrome P450 regulating genes may produce ligands for constitutive androstane receptor/retinoid X receptor heterodimer (CAR/RXRα) (Page et al., 2007) which may stimulate Ctse expression in TIS21-KO mice. How does TIS21^/BTG2/PC3 regulate PPARα signaling under fasting condition? It has been reported that PPARα shares RXRα as a heterodimeric nuclear receptor complex and TIS21 binds to nuclear receptor via two LXXLL motifs (Hu et al., 2011; Passeri et al., 2006), which may indicate a possible clue in the regulation of Ctse and β-oxidative genes at the downstream of PPARα. Moreover, PGC1α, a co-activator of PPARα, expression was also downregulated in the TIS21-KO mice. Thus TIS21 knockout-mediated downregulation of PPARα activity and Ctse expression need to further investigate during fasting.

On the other hand, expression of glutaredoxin 2 was also significantly reduced in liver of TIS21-KO mice compared to WT mice (Fig. 4B), and the focal fatty change in livers of TIS21-KO mice (Supplementary Fig. S4 and Supplementary Table S2) may reflect deficit of anti-oxidative capacity during overnight fasting. The assumption can be supported by previous reports that the expression of TIS21^/BTG2 downregulates reactive oxygen species (ROS) generation (Choi and Lim, 2013; Lim et al., 2012), that glutaredoxin 2 plays a protective role against hydrogen peroxide-induced damage and that its loss sensitizes cells to oxidative stress and apoptosis (Wu et al., 2010). In addition, TIS21 stimulates manganese superoxide dismutase (SOD2) transcription by nuclear factor-kappa B via crosstalk with phosphoinositol-3-kinase-AKT1 activation (Sundaramoorthy et al., 2013).

Taken together, the data indicate that the expression of the TIS21 gene protects cells and tissues from the fasting-induced ROS generation. In conclusion, we strongly suggest that TIS21^/BTG2 as an essential gene for anti-oxidative role against ROS and for regulation of fatty acid β-oxidation under stressful conditions, such as fasting and starvation.