Effect of PPARα and/or CREB3L3 ablation on fasting plasma metabolites
To study the potential interaction between PPARα and CREB3L3 in metabolic regulation in the fasted state, we first performed basic metabolic measurements in wild-type, PPARα−/−, CREB3L3−/−, and combined PPARα/CREB3L3−/− mice after 16 h of fasting. Plasma triglyceride levels were markedly elevated in the CREB3L3−/−, but not in the PPARα−/− mice (Fig. 1a). The hypertriglyceridemia in CREB3L3−/− mice was improved by the simultaneous ablation of PPARα, suggesting functional antagonism between PPARα and CREB3L3 in plasma triglyceride regulation (Fig. 1a). As previously shown [9], PPARα ablation significantly increased plasma non-esterified fatty acid (NEFA) levels (Fig. 1b), and decreased β-hydroxybutyrate levels (Fig. 1c). In agreement with our previous report [19], NEFA and β-hydroxybutyrate levels were elevated in CREB3L3−/− mice, while levels in PPARα/CREB3L3−/− mice were similar to those in PPARα−/− mice (Fig. 1b and c), suggesting a dominant effect of PPARα ablation. Interestingly, liver triglyceride levels were elevated in both PPARα−/− and CREB3L3−/− mice compared with wild-type mice and were highest in the combined PPARα/CREB3L3−/− mice (Fig. 1d). Plasma FGF21 levels were dramatically lower in PPARα−/−, CREB3L3−/−, and PPARα/CREB3L3−/− mice as compared with wild-type mice (Fig. 1e). These data indicate the pronounced impact of PPARα and CREB3L3 deficiency on metabolic regulation during fasting.
Effects of PPARα and CREB3L3 ablation on hepatic gene expression in the fasted state are largely independent
To study the potential interaction between PPARα and CREB3L3 in hepatic gene regulation in the fasted state, whole genome expression analysis was performed on liver samples of the four groups of mice after 16 h of fasting. To study the magnitude of the effect of PPARα and CREB3L3 ablation on liver gene expression, we performed Volcano plot analysis. Interestingly, the effects of PPARα ablation were much more pronounced as compared to CREB3L3 ablation (Fig. 2a). The combined ablation of PPARα and CREB3L3 had the most significant effect on gene regulation, pointing to a potential additive or synergistic effect of PPARα and CREB3L3 ablation on hepatic gene expression. Analysis of the number of significantly changed genes showed that in the fasted state, loss of PPARα altered the expression of 1097 genes, of which 553 genes were upregulated and 544 genes were downregulated (Fig. 2b). Loss of CREB3L3 altered expression of 312 genes, of which 134 genes were upregulated and 178 genes were downregulated. Combined loss of PPARα and CREB3L3 altered the expression of 1917 genes, of which 1064 genes were upregulated and 853 genes were downregulated (Fig. 2b). The fact that the number of significantly changed genes in the combined PPARα/CREB3L3−/− mice exceeds the sum of significantly changed genes in the PPARα−/− and CREB3L3−/− mice suggests a modest synergistic effect of PPARα and CREB3L3 deficiency on hepatic gene regulation, dominated by PPARα.
To study the potential similarity between the effect of PPARα and CREB3L3 deficiency on liver gene expression, we performed principle component analysis (Fig. 2c) and hierarchical clustering (Fig. 2d). Both analyses indicated the separate clustering of the four experimental groups, with limited variation between the individual mice in each group, and the more pronounced effect of PPARα deficiency compared to CREB3L3 deficiency. In addition, both analyses showed that the whole genome effects of combined PPARα/CREB3L3 deficiency are largely taken up by PPARα deficiency (Fig. 2c, d).
Hierarchical biclustering of samples and genes visualized in a heatmap further supported the conclusions reached above, showing the much more pronounced effect of PPARα deficiency on hepatic gene expression and the more significant contribution of PPARα towards the effect of combined PPARα/CREB3L3 deficiency (Fig. 2e). The heat map also shows that for certain genes, the effects of PPARα and CREB3L3 deficiency are additive and seemingly independent, whereas for other genes the effect of PPARα and CREB3L3 deficiency appears to be synergistic and thus dependent.
A limited number of genes is commonly downregulated by PPARα and CREB3L3 deficiency in liver during fasting
The Venn diagram of significantly downregulated genes (Fig. 3a) and scatter plot analysis (Fig. 3b) confirmed that, in general, the effects of PPARα and CREB3L3 deficiency are disparate, with only a limited number of genes showing similar regulation in PPARα−/− and CREB3L3−/− mice. The list of 34 genes that were significantly downregulated in livers of PPARα−/−, CREB3L3−/−, and PPARα/CREB3L3−/− mice is presented in Additional file 1: Table S1. This list includes Fgf21, which is known to be under dual control of PPARα and CREB3L3, and Mfsd2a, a gene involved in lysophospholipid transport that is known to be under control of PPARα but not CREB3L3 [26, 27]. To examine whether any of these 34 genes may be directly regulated by PPARα, we determined the effect of the PPARα agonist Wy-14,643 on gene expression in the liver (Fig. 3c). To examine whether any of these 34 genes may be directly regulated by CREB3L3, we determined the effect of adenoviral-mediated CREB3L3 overexpression on gene expression in the liver (Fig. 3c). The results suggest that several of the 34 genes may be direct targets of both PPARα and CREB3L3, as they are markedly upregulated by PPARα and CREB3L3 activation. These genes include Fgf21, Mfsd2a, Xrcc3, Suclg1, Tmem184a and Sel1l3.
To further explore the differential impact of PPARα and CREB3L3 deficiency on hepatic gene expression, the top 40 most significantly downregulated genes in each condition (PPARα−/−, CREB3L3−/−, and PPARα/CREB3L3−/−) were taken and visualized in a heatmap (Fig. 4). The top 40 list of most significantly downregulated genes in the CREB3L3−/− mice contains the known CREB3L3 targets Cidec, Apoa4, and Fgf21. A relatively large portion of the genes downregulated in the CREB3L3−/− mice were also downregulated in the PPARα−/− mice and, especially, in the combined PPARα/CREB3L3−/− mice. For PPARα, only a very small portion of the genes downregulated in the PPARα−/− mice were also downregulated in the CREB3L3−/− mice, the exception being Fgf21, Mfsd2a, and Mtnr1a. Other typical PPARα target genes such as Retsat, Cy4a14, Plin5, Fabp1, Acaa1b and Ehhadh were exclusively downregulated in the PPARα−/− mice. For nearly all genes shown, the downregulation in the PPARα−/− mice was copied in the combined PPARα/CREB3L3−/− mice. The top 40 list of most significantly downregulated genes in the PPARα/CREB3L3−/− mice represents a combination of genes mainly controlled by PPARα (Vnn1, Cyp4a14, Krt23, Slc27a1), by CREB3L3 (Cidec, Fabp2), or both (Fgf21, Mfsd2a, Mtnr1a) (Fig. 4).
A limited number of genes has been identified as direct or putative target of CREB3L3. The expression levels of these genes are illustrated in Fig. 5, showing that CREB3L3 deficiency leads to significant downregulation of genes involved in lipoprotein metabolism (Apoc2, Apoa5, Apoa1, Scarb1), lipid storage (Cidec), fatty acid binding (Fabp2), fatty acid desaturation and elongation (Fads1, Fads2, Elovl2, Elovl5), gluconeogenesis (Pck1, G6pc), and fatty acid oxidation (Cpt1a). Most of these genes were not or minimally affected by PPARα deficiency. Together, these data indicate that in the fasted state CREB3L3 and PPARα regulate different sets of genes, with some notable exceptions, suggesting that the transcription factors largely operate independently.
CREB3L3 deficiency leads to downregulation of genesets related to lipoprotein and lipid transport
To gain more insight into the functional differences between PPARα and CREB3L3 deficiency, we compared the effects of PPARα and CREB3L3 deficiency at the level of pathways using geneset enrichment analysis (Fig. 6a). Deficiency of PPARα led to the downregulation of numerous genesets that are known to be controlled by PPARα, mainly representing genesets related to peroxisomal and mitochondrial fatty acid catabolism and the electron transport chain. By contrast, deficiency of CREB3L3 led to the downregulation of genesets related to lipoprotein and lipid transport, as well as several genesets connected to immunity (Fig. 6b). At the pathway level, minimal overlap was observed between the effect of PPARα and CREB3L3 deficiency (Fig. 6c). In fact, out of 98 genesets that were significantly downregulated in PPARα−/− mice, only one geneset, named branched chain amino acid catabolism, was also downregulated in the CREB3L3−/− mice (Fig. 6c). The commonly enriched genes within the geneset branched chain amino acid catabolism included Auh, Hibch, Hibadh, Acad8, Ivd, and Hsd17B10.
Consistent with the notion that the effects of combined PPARα/CREB3L3 deficiency are largely taken up by PPARα deficiency, the far majority of genesets downregulated in the combined PPARα/CREB3L3−/− mice were also downregulated in the PPARα−/− mice (Fig. 6d). Indeed, the enrichment scores of the most highly downregulated genesets in the combined PPARα/CREB3L3−/− mice were very similar in the single PPARα−/− mice, suggesting that the functional impact of combined PPARα/CREB3L3−/− deficiency is mostly accounted for by deficiency of PPARα. The exception were two genesets related to lipoprotein and lipid transport, which had similar enrichment scores in the combined PPARα/CREB3L3−/− mice and single CREB3L3−/− mice (Fig. 6d), suggesting that the regulation of these two genesets is driven by CREB3L3 deficiency.
Deficiency of PPARα led to the upregulation of genesets related to the unfolded protein response and inflammatory signalling (Additional file 1: Figure S1A). By contrast, deficiency of CREB3L3 led to upregulation of genesets related to cholesterol synthesis and protein translation (Additional file 1: Figure S1B). Consistent with this result, genes involved in cholesterol metabolism feature prominently among the top 40 most highly upregulated genes in CREB3L3−/− mice (Additional file 1: Figure S1C).
Overall, the above analyses indicate that the effects of PPARα and CREB3L3 deficiency on hepatic gene expression during fasting are very distinct. Only a limited number of genes is under regulation of both PPARα and CREB3L3. The PPARα/CREB3L3−/− mice reflect the combined effect of especially PPARα and to a lesser extent CREB3L3 deficiency, showing a minor degree of synergism.
Effect of PPARα and/or CREB3L3 deficiency on plasma metabolites during ketogenic diet
To further explore the cooperativity between PPARα and CREB3L3 in hepatic gene regulation, we compared the effect of PPARα and CREB3L3 deficiency under the condition of a ketogenic diet. Previously, this diet was shown to provoke a pronounced hepatic phenotype in CREB3L3−/− mice, characterized by hepatomegaly and signs of steatohepatitis [19, 25]. No difference in bodyweight between the four genotypes was observed before the start of the study (Fig. 7a). Four days of ketogenic diet induced pronounced weight loss in all groups, which was most pronounced in the PPARα−/− mice and combined PPARα/CREB3L3−/− mice (Fig. 7b). Interestingly, compared to the wild-type mice, the liver to body weight ratio was modestly increased in the PPARα−/− mice and combined PPARα/CREB3L3−/− mice, yet was highest in the CREB3L3−/− mice, suggesting hepatomegaly (Fig. 7c) [19, 25]. Compared to the other three groups, CREB3L3−/− mice fed a ketogenic diet for 4 days also exhibited markedly elevated plasma alanine aminotransferase (ALT) activity (Fig. 7d), suggesting liver damage. Plasma ALT levels were below 30 IU/L in all groups before starting the ketogenic diet (not shown). Elevated plasma ALT was accompanied by elevated liver and plasma triglycerides in CREB3L3−/− mice (Fig. 7e,f). These parameters were also increased in the PPARα−/− mice. Plasma FGF21 levels followed a very different pattern and were about 50% decreased in the CREB3L3−/− mice, more than 90% decreased in the PPARα−/− mice, and nearly 99% decreased in the PPARα/CREB3L3−/− mice (Fig. 7g). Overall, these data are in line with a previous report [25].
To study the magnitude of the effect of PPARα and CREB3L3 deficiency during ketogenic diet on liver gene expression, we performed Volcano plot analysis (Fig. 8a). In contrast to what was observed in the fasted state, the effects of CREB3L3 deficiency during ketogenic diet were more pronounced as compared to PPARα deficiency. Strikingly, the effect of combined deficiency of PPARα and CREB3L3 on hepatic gene expression was less pronounced as compared to deficiency of only CREB3L3. Analysis of the number of significantly changed genes showed that loss of CREB3L3 altered the expression of 5878 genes, of which 3490 genes were upregulated and 2388 genes were downregulated (Fig. 8b). Loss of PPARα altered expression of 2843 genes, of which 1616 genes were upregulated and 1227 genes were downregulated. Combined loss of PPARα and CREB3L3 altered the expression of 3707 genes, of which 1996 genes were upregulated and 1711 genes were downregulated. These observations indicate that deficiency of PPARα mitigates the effect of CREB3L3 deficiency on hepatic gene expression.
Effects of CREB3L3 deficiency on hepatic gene expression during ketogenic diet are dependent on PPARα
To study the similarity between the three different genetic models in liver gene expression, we performed principle component analysis (Fig. 8c) and hierarchical clustering (Fig. 8d). Principle component analysis and hierarchical clustering of samples showed that the CREB3L3−/− mice formed a distinct cluster, underscoring the profound effect of CREB3L3 deficiency on hepatic gene expression during ketogenic diet. Surprisingly, the PPARα−/− mice and combined PPARα/CREB3L3−/− mice clustered together and were very distinct from the CREB3L3−/− mice. Hierarchical biclustering of samples and genes visualized in a heatmap further confirmed that at the level of hepatic gene expression, the PPARα−/− mice and combined PPARα/CREB3L3−/− mice were nearly indistinguishable, whereas the CREB3L3−/− mice showed a very different gene expression profile (Fig. 8e). These data thus show that deficiency of CREB3L3 has no effect on hepatic gene expression in the absence of PPARα, indicating that the major liver phenotype triggered by CREB3L3 deficiency during ketogenic diet is dependent on PPARα.
Scatter plot analysis confirmed that the effects of PPARα and CREB3L3 deficiency on hepatic gene expression are very dissimilar, whereas the effect of PPARα deficiency and combined PPARα/CREB3L3 deficiency are similar (Additional file 1: Figure S2A). Venn diagram of significantly changed genes confirmed that deficiency of CREB3L3 leads to the up- and downregulation of a large set of genes that are not affected in the PPARα−/− or PPARα/CREB3L3−/− mice (Additional file 1: Figure S2B).
Induction of mitogenic genes in CREB3L3−/− mice during ketogenic diet is mediated by PPARα
To obtain more insight into the functional pathways affected by CREB3L3 deficiency on ketogenic diet, we performed geneset enrichment analysis. Surprisingly, many of the most highly downregulated genesets represented pathways of fatty acid and/or amino acid metabolism, including peroxisome, PPARα targets, and fatty acid degradation (Fig. 9a). Enrichment scores for these latter genesets were similar in the PPARα−/− and PPARα/CREB3L3−/− mice (Fig. 9a). A heatmap of the geneset PPARα targets shows the consistent downregulation of PPARα target genes across the 3 groups of mice (Fig. 9b). These data suggest that CREB3L3 deficiency, as well as PPARα deficiency and combined PPARα/CREB3L3 deficiency, leads to reduced PPARα activity. In line with these data, the PPARα mRNA expression level was markedly reduced in the CREB3L3−/− mice (Fig. 9c).
Geneset enrichment analysis also underscored the dramatic effect of CREB3L3 deficiency on hepatic gene expression. Indeed, 538 genesets met the statistical significance cut-off of FDR q-value< 0.05, covering numerous biological processes, including immunity, cellular stress pathways, and DNA/RNA-related processes (not shown). Intriguingly, the 20 most upregulated genesets were all related to cell cycle/mitosis (Fig. 10a). Enrichment scores for these genesets were much lower in the PPARα−/− and PPARα/CREB3L3−/− mice, indicating the selective induction of cell cycle/mitosis-related genes in the CREB3L3−/− mice (Fig. 10a). A heatmap of the most enriched genes within the geneset Cell.Cycle.Mitotic demonstrates the pronounced upregulation of cell cycle genes in the CREB3L3−/− mice (Fig. 10b). Strikingly, the upregulation is completely abolished upon additional deficiency of PPARα, suggesting that PPARα mediates the induction of cell cycle genes in CREB3L3−/− mice on ketogenic diet (Fig. 10b). Consistent with the upregulation of cell cycle upon CREB3L3 deficiency, many of the most highly induced genes in the CREB3L3−/− mice on ketogenic diet were related to cell cycle (Fig. 10c). Again, the upregulation of these genes was almost completely abolished in the PPARα−/− mice.
In line with the known mitogenic effect of PPARα activation on hepatocyte proliferation, pharmacological activation of PPARα in vivo has been shown to cause the induction of numerous genes and proteins involved in cell cycle control [28], which is specifically mediated by mouse PPARα and not human PPARα [29]. Previously, we found that treating mice with the specific PPARα agonist Wy-14,643 markedly induced numerous genesets related to cell cycle [2]. A heatmap of the most highly enriched genes in the geneset Mitotic.M.M.G1 phase underscores the marked induction of cell cycle-related genes by Wy-14,643, which is entirely PPARα dependent (Fig. 10d). Strikingly, most of these genes are also highly upregulated in the CREB3L3−/− mice on ketogenic diet, which again is entirely PPARα dependent (Fig. 10d), indicating that the pronounced upregulation of the cell cycle in CREB3L3−/− mice is mediated by PPARα. Taken together, these data indicate that CREB3L3 deficiency uncouples the hepatoproliferative and lipid metabolic effects of PPARα.