PPARα Association With Sirt1 Suppresses Cardiac Fatty Acid Metabolism in the Failing Heart

Abstract

Background

Heart failure (HF) is accompanied by changes in cardiac metabolism characterized by reduced fatty acid (FA) utilization. However, the underlying mechanism and its causative involvement in the progression of HF are poorly understood. The peroxisome proliferator activated receptor-α (PPARα)/retinoid X receptor (RXR) heterodimer promotes transcription of genes involved in FA metabolism through binding to the PPAR response element, called direct repeat 1 (DR1). Silent information regulator 1 (Sirt1) is a histone deacetylase which interacts with PPARα.

Methods and Results

To investigate the role of PPARα in the impaired FA utilization observed during HF, genetically altered mice were subjected to pressure overload (PO). The DNA binding of PPARα, RXRα and Sirt1 to DR1 was evaluated with oligonucleotide pull-down and chromatin immunoprecipitation assays. Although the binding of PPARα to DR1 was enhanced in response to PO, that of RXRα was attenuated. Sirt1 competes with RXRα to dimerize with PPARα, thereby suppressing FA utilization in the failing heart. DR1 sequence analysis indicated that the typical DR1 sequence favors PPARα/RXRα heterodimerization, whereas the switch from RXRα to Sirt1 takes place on degenerate DR1s. Sirt1 bound to PPARα through a region homologous to the PPARα binding domain in RXRα. A short peptide corresponding to the RXRα domain not only inhibited the interaction between PPARα and Sirt1 but also improved FA metabolism and ameliorated cardiac dysfunction.

Conclusions

A change in the heterodimeric partner of PPARα from RXRα to Sirt1 is responsible for the impaired FA metabolism and cardiac dysfunction in the failing heart.

The healthy heart produces and consumes a large amount of energy in order to maintain cardiac output. To satisfy the high energy demand, cardiomyocytes harbor a large number of mitochondria and utilize fatty acids (FAs) as their primary substrate for energy production. FAs are catabolized to acetyl-CoA through β-oxidation in mitochondria and peroxisomes. The resultant acetyl-CoA is then utilized for ATP production through oxidative phosphorylation in the tricarboxylic acid (TCA) cycle and the electron transport chain (ETC) in mitochondria. However, the FA utilization capability declines in the failing heart¹. Associated with this decline is a broad downregulation of metabolic genes throughout the FA consumption pathway, from FA uptake to the ETC². Nevertheless, neither the mechanism responsible for this downregulation nor its functional significance is well understood.

Peroxisome proliferator activated receptor-α (PPARα), a member of the PPAR nuclear receptor subfamily, plays an important role in cardiac FA utilization³. PPARα heterodimerizes with retinoid X receptor (RXR) to activate transcription through binding to the PPAR response element (PPRE), also called direct repeat 1 (DR1). The consensus sequence of the PPRE/DR1 comprises 2 AGGTCA sequences separated by 1 nucleotide, AGGTCANAGGTCA, with PPARα and RXR binding to the 5’ and 3’ AGGTCA sequences, respectively⁴. PPARα/RXR target genes harboring a functional PPRE/DR1 are enriched in the upstream steps of the FA metabolic pathway, including FA uptake, transport and β-oxidation³. Loss of PPARα in mice results in impaired PPARα target gene expression, FA utilization and contractile function, indicating that PPARα is essential for cardiac function and FA utilization in the healthy heart⁵. However, overexpression of PPARα in the heart induces a mild failing heart phenotype⁶. Thus, PPARα may also play a role in the pathological development of heart failure (HF).

Silent information regulator 1 (Sirt1) is a nicotinamide adenine dinucleotide (NAD)–dependent protein deacetylase that plays an important role in metabolic adaptation to starvation conditions⁷. Sirt1 is a functional binding partner of both PPARα and PPARγ^8–10. However, modulation of PPAR function by Sirt1 is complex. PPARα target gene expression and FA utilization are enhanced by Sirt1 in the liver⁹. Sirt1 also protects against hypertrophy and metabolic dysregulation through PPARα activation in cultured cardiomyocytes¹¹. However, Sirt1 promotes PPARα-induced failing heart phenotypes in vivo¹². Likewise, Sirt1 inhibits PPARγ in white adipocytes, leading to release of FAs during starvation⁸, but enhances its function in brown adipocytes¹⁰.

Sirt1 is upregulated during HF induced by pressure overload (PO), and exacerbates the failing heart phenotypes^13–15. Furthermore, cardiac-specific overexpression of Sirt1 exacerbates cardiac dysfunction under PO conditions¹⁴. We showed previously that PPARα recruits Sirt1 to the estrogen related receptor (ERR) response element (ERRE) and suppresses ERR target gene expression in the failing heart¹². PPARα can bind to the ERRE because its consensus sequence, TNAAGGTCA, contains a core hexad binding sequence for PPARα as a monomeric unit. ERR targets are enriched in the mitochondrial TCA cycle and ETC, and PPARα thus suppresses genes mediating the downstream metabolic pathway of FA consumption. This raises the question of whether PPARα is also responsible for downregulation of canonical PPARα target genes harboring a PPRE/DR1, namely, those mediating the upstream FA metabolism, in the failing heart. We here show that PPARα is required not only for PPAR target gene expression in the healthy heart but also for gene suppression in the failing heart. We propose that inhibition of PPARαSirt1 heterodimerization has therapeutic potential for HF treatment.

Methods

Animal experiments

Cardiac specific PPARαtransgenic, PPARα knockout and cardiac specific Sirt1 knockout mice have been described previously¹². All procedures involving animals were performed in accordance with protocols approved by Rutgers Biomedical and Health Sciences.

Quantitative RT-PCR

Total RNA was prepared from left ventricles using the RNeasy Fibrous Tissue Mini Kit (Qiagen), and cDNA was then generated using M-MLV Reverse transcriptase (Promega). Real-time RT-PCR was performed using Maxima SYBR Green qPCR master mix (Fermentas).

In vitro binding assays

Recombinant Glutathione S-transferase (GST)-fused proteins and double-stranded DNA were incubated at 4° C for 2 hours. Precipitation was performed with streptavidin beads or glutathione beads.

Human heart samples

Heart lysates from explanted hearts were obtained from patients who received heart transplants and from age-matched donors. All patients or their families expressed their willingness to participate through an informed consent form. Detailed information regarding the human samples was provided previously¹⁶.

Statistical analysis

Detailed statistical methods are descried in supplemental information. P<0.05 was defined as statistically significant and indicated by a filled asterisk. NS=not significant. All error bars represent S.E.M. The Biostatistics Laboratory of Rutgers New Jersey Medical School independently reviewed all data sets and applied the appropriate statistical methods.

Results

PPRE/DR1 sequence affects PPARα/RXR heterodimerization and target gene expression

To examine the role of PPARα in PPARα target gene expression in the failing heart, we employed a murine model of HF in which surgical constriction of the aorta (transverse aortic constriction (TAC)) induces PO, leading to cardiac dysfunction and metabolic remodeling¹. PPARα homozygous and heterozygous knockout (PPARα−/ − and PPARα+/ −) mice were subjected to 4 weeks of TAC, while PPARα-overexpressing (Tg-PPARα mice were subjected to 2 weeks of TAC due to their high mortality rate (Supplementary Figure 1). As expected, many PPAR target genes were downregulated in PPARα −/ − mice even at baseline, confirming that PPARα is essential for their basal gene expression (Figure 1A and Supplementary Figure 2A). Conversely, several PPAR targets, such as Acox1, Cd36, Cpt1b, Acsl3, Cpt1a, Hmgcs2, Pcx and Plin2, were upregulated in Tg-PPARα mice and were not significantly suppressed below the wild type (WT) baseline even in the presence of PO. Interestingly, however, some PPAR targets were not upregulated in Tg-PPARα at baseline and were downregulated by TAC in WT mice, including Cpt2, Ech1, Fabp3, Mcad, Vldlr, Abcd2, Acaa1b, Acsl1, Fatp1, Mlycd, Slc22a5, Txnip and Lpl. This TAC-induced downregulation was partly normalized in PPARα+/ −, whereas it was exacerbated in Tg-PPARα mice (Figure 1A, 1B, Supplementary Figure 2A and 2B). Thus, PPARα negatively regulates a subset of PPAR target genes under PO.

Figure 1.

PPAR target genes with PPREs/DR1s that only weakly promote PPARα/RXR heterodimerization are negatively regulated by PPARα. After 4 (A) and 2 weeks (B) of TAC, qPCR was performed. At least 5 mice were used for each group. (C) Sequences of double-stranded oligonucleotides used, including the consensus sequence of PPRE/DR1 (DR1) and the consensus PPRE/DR1 with mutations in either the 3’ hexad (DR1m3’) or both hexads (DR1m5’3’). Intrinsic PPRE/DR1 sequences located in PPAR target genes are indicated by gene name. 5’ PPARα and 3’ RXR binding hexad sequences which have no more than 1 nucleotide mismatch from the reported consensus RGSWVANAGGTCA¹⁷ (R=A or G, S=G or C, W=A or T, V=C, G or A) are shown in red, and those with 2 or more nucleotide mismatches are shown in blue. (D) Biotin-labeled DNA comprising the indicated PPREs/DR1s was incubated with recombinant PPARα. DNA-bound PPARα was pulled down with streptavidin beads. (E) PPREs/DR1s differ in their abilities to promote PPARα/RXR heterodimerization. GST-PPARα was incubated with RXR and oligonucleotides comprising DR1s derived from the indicated genes and the consensus DR1.

PPARα regulates expression of its targets primarily through PPRE/DR1. In order to understand the mechanism by which PPARα differentially regulates its downstream targets, we examined the sequences of known PPREs/DR1s. Although the consensus sequence of the PPRE/DR1 has been reported as RGSWVANAGGTCA¹⁷ (R=A or G, S=G or C, W=A or T, V=C, G or A), the actual nucleotide sequence exhibits considerable variation from gene to gene. We hypothesized that the diversity in the PPRE/DR1 sequences is a critical factor in the differential responses of PPARα targets. Of the 9 PPAR target genes we examined that were upregulated under basal conditions in Tg-PPARα, all 9 harbored well-conserved typical PPRE/DR1 sequences with no more than one nucleotide mismatch from the reported consensus in each 5’ and 3’ hexad sequence (Figure 1C and Supplementary Figure 2B and 2C), which we designated as perfect DR1s. On the other hand, 11 out of 14 PPAR target genes that were not significantly upregulated in Tg-PPARα harbored unique PPRE/DR1 sequences comprising one well-conserved AGGTCA sequence with no more than one nucleotide mismatch from the consensus and one relatively degenerate hexad sequence with more than one nucleotide mismatch, which we designated as imperfect DR1s. Fisher’s exact test indicates that the perfect DR1 correlates with upregulation, whereas the imperfect DR1 correlates with either no change or downregulation of PPARα targets in Tg-PPARα mice under basal conditions (p=0.014) (Supplementary Figure 2D). Several PPAR target genes, including Acaa1b, Ehhadh, Fatp1 and Slc22a5, harbor both perfect and imperfect DR1s (Supplementary Figure 2C) and were therefore included in both the perfect DR1 and imperfect DR1 groups. The number of genes harboring both perfect and imperfect DR1s is indicated in parentheses (Supplementary Figure 2D and 2E).

We next investigated whether PO differentially regulates PPARα targets with perfect and imperfect DR1s. Of the 13 PPAR targets suppressed under PO in WT mice, 11 harbored an imperfect DR1. In contrast, all 10 of the PPAR target genes not significantly suppressed under PO harbored perfect DR1s. The perfect DR1 correlates with resistance against PO-induced downregulation, whereas the imperfect DR1 correlates with PO-induced downregulation (p=0.0047) (Supplementary Figure 2E). Because the downregulation of the imperfect DR1 group of genes was partly normalized in PPARα+/ − , these results suggest that PPARαnegatively regulates target genes harboring an imperfect DR1, whereas genes with a perfect DR1 are resistant to PPARα-mediated downregulation during PO.

RXR is an essential heterodimerization partner for PPARα-mediated transcription and directly binds to the 3’ hexad sequence of the PPRE/DR1. We hypothesized that the extent of similarity of an individual PPRE/DR1 to the perfect DR1 affects the level of PPARα/RXR heterodimerization, which, in turn, may determine the response of each PPARα target. To test this hypothesis, recombinant PPARα and RXRα were incubated with double-stranded DNA comprising intrinsic DR1s. PPARα bound to all DR1s we tested (Figure 1D). Recruitment of RXRα to PPARα was readily observed in the presence of either the DR1 consensus sequence (DR1) or perfect DR1s derived from genes upregulated in Tg-PPARα, including Acox1, Cd36 and Cpt1b. However, it was much weaker in the presence of imperfect DR1s derived from genes downregulated during PO, including Cpt2, Mcad and Vldlr (Figure 1E). Thus, intrinsic DR1s differ in their abilities to promote PPARα/RXR heterodimerization. Specifically, perfect DR1s with 2 typical AGGTCA-like sequences strongly recruit both PPARα and RXR, whereas imperfect DR1s strongly recruit PPARα but only weakly recruit RXR. These results suggest that the extent of nucleotide mismatch from the consensus in the DR1 affects the level of PPARα/RXR heterodimerization, which, in turn, may affect the pattern of gene expression in response to PO or in Tg-PPARα.

Functional significance of imperfect DR1 group of genes in FA utilization and cardiac function in the failing heart

To investigate the functional significance of the imperfect DR1 group of genes, FA utilization and cardiac function were examined in PPARα gene-manipulated mice. As expected, PPARα −/ − mice showed impaired FA utilization both at baseline and under PO. However, PO-induced impairment of FA utilization was attenuated in PPARα+/ − mice (Figure 2A), in which PO-induced downregulation of the imperfect DR1 group of genes was normalized (Figure 1A and Supplementary Figure 2A). PO induced cardiac systolic dysfunction, characterized by a decrease in the LV ejection fraction, in WT mice. However, the ejection fraction under PO was normalized in PPARα+/ − mice, while PPARα −/ − mice exhibited cardiac dysfunction even at baseline (Figure 2B). PO-induced lung congestion and LV hypertrophy were normalized in PPARα+/ −, but not in PPARα −/ − mice (Figure 2C and 2D). Both the impairment of FA utilization (Figure 2E) and cardiac dysfunction (Figure 2F to 2H) induced by PO were exacerbated in Tg-PPARα mice, in which the imperfect DR1, but not the perfect DR1, group of genes was further suppressed (Figure 1B and Supplementary Figure 2B). Even without PO, there was a trend towards reduced FA utilization in Tg-PPARα mice. Indeed, many genes belonging to the imperfect DR1 group showed a trend of downregulation. These results suggest that suppression of the imperfect DR1 group of genes results in a decline in FA utilization and cardiac function.

Figure 2.

PPARα is essential for FA oxidation and cardiac function at baseline, but promotes a decline in FA oxidation and cardiac dysfunction during PO. Cardiac FA oxidation activity (A and E), ejection fraction (B and F), lung weight/body weight (C and G) and LV weight/body weight (D and H) were examined after 4 (A–D) and 2 (E–H) weeks of TAC. The numbers of mice in each experimental group were: 6–17 (A), 13–20 (B), 14–23 (C–D), 7–9 (E), 7–15 (F), and 14–32 (G–H).

An essential role of Sirt1 in PPARα-mediated gene suppression

As in PPARα-mediated ERR target gene suppression¹², Sirt1 may have a role in PPARα-mediated suppression of PPARα target genes harboring an imperfect DR1. Sirt1 was increased, whereas RXR was decreased, in response to PO in the mouse heart (Figure 3A). Similarly, there was a trend towards upregulation of Sirt1 and downregulation of RXR in failing human hearts (Figure 3B and Supplementary Figure 3). Furthermore, Sirt1 binding to PPARα increased, whereas RXR binding to PPARα decreased, during PO in the mouse heart (Figure 3C and Supplementary Figure 4). In vitro binding assays demonstrated that Sirt1 binds to PPARα more strongly than to RXRαsuggesting that Sirt1 interacts with PPARα directly rather than through RXR (Figure 3D). Sirt1 does not have a typical DNA binding domain. To investigate whether PPARα recruits Sirt1 to DNA α recombinant PPARα and Sirt1 were incubated with biotin-labeled double-stranded DNA containing 3 repeats of the PPAR binding sequence (AGGTCA) separated by 7 bp spacers. Sirt1 was pulled down with the AGGTCA sequences in a PPARα-dependent manner, suggesting that Sirt1 is recruited to DNA through PPARα(Figure 3E) and that neither RXR nor the perfect DR1 element is necessary for PPARα DNA binding and Sirt1 recruitment. Chromatin immunoprecipitation (ChIP) assays showed that, during PO, Sirt1 was recruited to imperfect DR1s, such as in the Mcad and Vldlr promoters, in WT but not in PPARα −/ − mice, suggesting that PPARα recruits Sirt1 to imperfect DR1s during PO in vivo (Figure 3F). Furthermore, PO-induced downregulation of the PPAR target genes harboring imperfect DR1s was partly normalized in Sirt1+/− mice (Figure 3G). Reporter gene assays showed that co-overexpression of PPARα and Sirt1 significantly suppressed reporter activity driven by an imperfect DR1, such as those in the Mcad and Vldlr promoters, whereas PPARα-induced suppression was normalized by knockdown of Sirt1 (Figure 3I). Thus, Sirt1 recruited to imperfect DR1s by PPARα appears to be crucial for PPARα-mediated transcriptional suppression. To investigate whether the imperfect DR1 is important for suppression by PPARα/Sirt1, the degenerate hexad sequences of intrinsic promoters were mutated to typical AGGTCA sequences to form perfect DR1s (Mcad(DR1) and Vldlr(DR1)) (Figure 3H). PPARα/Sirt1 failed to suppress reporter activity driven by these sequences (Figure 3J). These results suggest that PPARα/Sirt1 suppresses transcription through direct interaction with an imperfect DR1, but that the perfect DR1 resists PPARα/Sirt1-induced suppression.

Figure 3.

Sirt1 plays an important role in PPARα-mediated transcriptional suppression. (A) Expression levels of Sirt1, RXR and PPARα in mouse hearts. (B) Expression levels of Sirt1, RXR and PPARα in healthy (Donor) and failing (Recipient) human hearts. (C) Reciprocal binding of RXR and Sirt1 to PPARα in the heart. Tg-Flag-PPARα mice were subjected to TAC. The Flag-PPARα was immunoprecipitated with anti-Flag antibody. (D) Direct interaction between PPARα and Sirt1. Recombinant Sirt1 was incubated with GST-fused PPARα and RXRα. (E) PPARα recruits Sirt1 to the DNA. Recombinant PPARα and Sirt1 were incubated with biotin-labeled DNA containing 3 repeats of the PPARα binding sequence (AGGTCA) or a mutant sequence (GAATCA). (F–G) Sirt1 recruited by PPARα suppresses gene expression. (F) PPARα is required for recruitment of Sirt1 to imperfect DR1s in the heart in vivo. ChIP assays were performed with WT and PPARα −/ − hearts subjected to TAC (n=6–10). (G) Sirt1 suppresses PPAR target genes harboring imperfect DR1s. qPCR was performed using cDNA prepared from Sirt1+/− mice (n=5–8). (H) A schematic representation of the PPRE reporter constructs. (I–J) Luciferase assays were performed (n=6–12). (I) Reporter gene activity driven by intrinsic imperfect DR1s was suppressed by PPARα/Sirt1. (J) PPARα failed to suppress perfect DR1 reporter activity.

To further evaluate whether PPARα and Sirt1 suppress a subset of the known PPARα targets, and, if so, whether the sequence of the PPRE/DR1 contributes to this mechanism, we conducted a microarray analysis using transgenic mice with combined cardiac-specific overexpression of both PPARα and Sirt1 (Tg-PPARα/Sirt1)¹². We examined 33 PPARα target genes that were upregulated or downregulated more than 1.2 fold in Tg-PPARα/Sirt1 mice and have been reported or suggested to contain a PPRE. Although many known PPARα target genes were upregulated, a subset of the known targets of PPARα was downregulated in Tg-PPARα/Sirt1 (Figure 4A). Among the 33 PPARα targets examined, 13 of the15 genes that were upregulated in Tg-PPARα/Sirt1 mice harbored perfect DR1s (Figure 4B and 4C), while 16 of the 18 genes that were downregulated in Tg-PPARα/Sirt1 mice harbored imperfect DR1s (Figure 4B and 4D). Thus, the perfect DR1 correlates with upregulation, whereas the imperfect DR1 correlates with downregulation, of PPARα targets (p=0.0028) (Figure 4E). These results suggest that co-upregulation of PPARα and Sirt1 suppresses PPARα target genes harboring imperfect DR1s but not those with perfect DR1s.

Figure 4.

Intrinsic PPRE/DR1 sequences that are upregulated or downregulated in cardiac-specific Tg-PPARα/Sirt1 mice. (A) Heat map of PPAR target gene expression. (B) The reported PPRE/DR1 sequences harbored by the indicated PPAR target genes. 5’ PPARα and 3’ RXR binding hexad sequences which have no more than 1 nucleotide mismatch from the reported consensus RGSWVANAGGTCA (R=A or G, S=G or C, W=A or T, V=C, G or A) are shown in red, and those with more than 1 nucleotide mismatch are light blue. (C) The sequence logo generated by the PPRE/DR1 sequences of genes that are upregulated in Tg-PPARα/Sirt1 mice. (D) The sequence logo generated by the PPRE/DR1 sequences of downregulated genes. The logos were generated using PPRE/DR1 sequences that have a conserved AGGTCA sequence in either the 5’ (Top) or 3’ (Bottom) hexad element in the DR1. (E) Contingency table of PPAR target gene expression in Tg-PPARα/Sirt1. PPAR target genes with a 1.2 fold or greater increase or decrease in expression in the bigenic mice are defined as upregulated and downregulated genes. The number of genes counted as both perfect and imperfect is indicated in parentheses.

Switching PPARα heterodimerization partner from RXR to Sirt1 suppresses FA utilization and cardiac function

Given that the perfect DR1 more effectively recruits RXR to PPARα (Figure 1E), RXR may prevent transcriptional suppression by PPARα/Sirt1. Forced expression of RXRα normalized the reporter gene activity suppressed by PPARα/Sirt1 (Figure 5A), suggesting that decreased heterodimerization of RXR with PPARα and recruitment of Sirt1 to PPARα promote the transcriptional suppression in the failing heart. In vitro binding assays showed that RXRα inhibited the interaction between Sirt1 and PPARα in a dose-dependent manner (Figure 5B), whereas Sirt1 did not significantly inhibit the interaction between RXRα and PPARα(Supplementary Figure 5). Thus, Sirt1 and RXRα appear to compete for heterodimerization with PPARα, with RXRα having a higher affinity for PPARα than Sirt1. Incubation with a perfect DR1, but not with an imperfect DR1 from the Mcad promoter (DR1(Mcad)) or a mutated DR1 (DR1m5’3’), enhanced both PPARα/RXRαheterodimerization and dissociation of Sirt1 from PPARα (Figure 5C). These results suggest that RXR competitively inhibits the interaction between Sirt1 and PPARα, and that Sirt1 recruitment is inhibited on a perfect DR1 due to enhancement of PPARα/RXR heterodimerization. As shown in Figure 5D, co-occupancy of PPARα/Sirt1 on imperfect DR1s, such as those in Mcad and Vldlr, was increased under PO conditions, whereas that of PPARα/RXR was decreased. Furthermore, deacetylation of histone H3 at lysine 9 by Sirt1 generally leads to suppression of transcriptional activity and, under PO, acetylation of histone H3 at lysine 9 was decreased in the imperfect DR1 promoter regions. These results suggest that, during PO, PPARα/RXR is replaced by PPARα/Sirt1, which, in turn, suppresses transcription in correlation with histone deacetylation.

Figure 5.

Sirt1 and RXR are competitive heterodimerization partners of PPARα. (A) RXR counteracts PPARα/Sirt1-mediated transcriptional suppression on the imperfect DR1. The indicated reporter genes and expression vectors were transfected into cultured cardiomyocytes (n=5–12). (B) Sirt1 and RXRα compete for interaction with PPARα. The indicated recombinant proteins were incubated with GST-PPARα. (C) The indicated recombinant proteins and double-stranded DNA were incubated with GST-PPARα. (D) The occupancies of PPARα/Sirt1, PPARα/RXR and Histone H3 acetylated at lysine 9 (AcK9HH3) on Mcad and Vldlr promoters. ChIP and Re-ChIP were performed with the indicated antibodies (n=3–12). (E) Sirt1 possesses regions homologous to the RXRαDBD). (F) Sirt1 binds to PPARα through the RXR-homologous regions (Left). RXRαDBD) inhibits interaction between PPARα and Sirt1 (Right). (G) RXRα(DBD) inhibits interaction between PPARα and Sirt1 in myocytes. (H) PPARα/Sirt1-induced transcriptional suppression on imperfect DR1s derived from the Mcad and Vldlr promoters is attenuated by RXRαDBD) (n=6). (I) RXRαDBD) ameliorates PO-induced cardiac dysfunction. Adenovirus vectors encoding LacZ (Control) and RXRα(DBD) were injected into the heart and the mice were subjected to 4 weeks of TAC. FA oxidation activity (Left) and ejection fraction (Right) (n=6–8).

The competitive binding of RXR and Sirt1 to PPARα could be due to structural similarity between the proteins. A comparison of the protein sequences of Sirt1 and RXRα showed that Sirt1 contains regions homologous to the DNA binding domain of RXRα (RXRα(DBD)) that binds to PPAR (Figure 5E)¹⁸. PPARα was effectively pulled down with Sirt1(184–409), which includes the regions homologous to RXR, but not with Sirt1(199–379), which lacks the RXR-homologous regions (Figure 5F). Conversely, RXRα(DBD) inhibited the interaction of PPARα with Sirt1 (Figure 5F and 5G). Thus, the RXR-homologous regions in Sirt1 are binding sites for PPARα. RXRα(DBD) also attenuated PPARα/Sirt1-induced transcriptional suppression (Figure 5H). Furthermore, PO-induced impairment of FA utilization and cardiac dysfunction were partially rescued by RXRα(DBD) expressed via adenovirus injection into the heart (Figure 5I and Supplementary Figure 6). Thus, disruption of the PPARα/Sirt1 dimer ameliorates impairment of FA utilization and cardiac dysfunction.

Discussion

We here show a transcriptional regulatory mechanism leading to downregulation of PPARα target gene expression and impairment of FA utilization during HF. PPARα typically heterodimerizes with RXRα, thereby promoting transcription of PPARα targets. However, PPARα can also heterodimerize with Sirt1, which suppresses transcription of PPARα targets. Using Sirt1 heterozygous knockout mice and a mini-gene approach that interferes with the specific interaction between the RXR-homologous region of Sirt1 and PPARα, we show that there is a switch from PPARα/RXR to PPARα/Sirt1 during PO, leading to downregulation of genes controlled by imperfect DR1s and inhibition of FA utilization.

Downregulation of genes involved in FA utilization is commonly observed during HF, and the substrate switch from FAs to glucose has a great impact on cardiac metabolism, including ATP production, oxidative stress and mitochondrial function¹⁹. Previous works have suggested that reduced levels of PPARα primarily explain the downregulation of genes involved in FA metabolism^20–22. However, neither decreased nuclear PPARα expression nor downregulation of certain well-established targets of PPARα such as Cpt1b is uniformly observed in the failing heart^{12, 23}. The PPARα level may vary depending upon the severity and the timing of PO. Thus, alternative mechanisms may mediate the downregulation of genes involved in FA metabolism, in a context-dependent manner. Our results provide an alternative mechanism through which a subset of genes controlled by PPARα and involved in FA utilization are downregulated without obvious downregulation of PPARα but instead through a mechanism dependent upon Sirt1, which is upregulated in failing hearts. We have previously reported that acetylation of PPARα is not detectable even in Sirt1−/ − mice¹². We speculate that the level of PPARα acetylation may be below the detection limit or that Sirt1 may not be a major deacetylase of PPARα. However, whether Sirt1 deacetylates PPARα, and if so, whether deacetylation of PPARα rather than histone deacetylation is involved in the suppression of PPARα targets remain to be elucidated.

Our previous work showed that the PPARα/Sirt1 complex suppresses ERR target genes involved in the mitochondrial ETC and the TCA cycle through direct interaction with the ERRE during fasting and PO¹². The consensus sequence of the ERRE is TNAAGGTCA, providing a binding sequence for monomeric PPARα which recruits Sirt1, but lacking an additional AGGTCA-like sequence for RXR binding. We proposed that this mechanism may allow the heart to reduce nutrient consumption in the presence of energy stress. Here, we extended this observation from the ERRE to the PPRE/DR1 and showed that PPARα/Sirt1 binding to a subset of DR1s plays an important role in mediating metabolic remodeling during HF. Importantly, this study demonstrates for the first time that 1) the dimerization partner of PPARα on a given DR1 can switch from RXR to Sirt1 during HF, depending upon the DNA sequence of the DR1, 2) the switch causes a change from upregulation to downregulation of the genes involved in FA utilization, the metabolic hallmark of HF, and 3) PPARα/Sirt1-mediated transcriptional suppression promotes both impaired FA utilization and cardiac dysfunction in response to PO.

Although Sirt1 and RXRα are competitive heterodimerization partners of PPARα, RXRα has a stronger affinity for PPARα than Sirt1. This suggests that PPARα preferentially heterodimerizes with RXR rather than Sirt1 and that Sirt1 can bind to PPARαonly when PPARα exists in excess, making available PPARα not dimerized with RXR. This may explain why PO-induced decreases in FA oxidation are normalized by heterozygous downregulation of PPARα(Figure 2B). Partial knockdown of PPARα in PPARα+/ − mice preferentially reduces PPARα/Sirt1 heterodimerization rather than PPARα/RXR, which prevents transcriptional suppression of genes involved in FA utilization.

Although never explicitly stated, close inspection of previously published reports led us to speculate that the function of PPARs is partially controlled by the sequence of their DNA binding sites. ChIP-on-Chip and ChIP-sequencing analyses showed that, in fact, single isolated AGGTCA and imperfect DR1 sequences comprise the majority of DNA binding sites for all PPARs^{24, 25}, suggesting that the perfect DR1 type of sequence is not necessarily the predominant binding site for PPARs in vivo and that the diversity of DNA binding sites may explain the functional diversity of PPARs. For example, suppression of the phospholipid transfer protein gene by PPARα is mediated through isolated AGGTCA sequences in the promoter²⁶. We here show the critical involvement of Sirt1 in mediating the molecular mechanism by which PPARα downregulates its targets through DR1. Importantly, Sirt1 and RXR both bind to PPARα through homologous amino acid stretches. This property allows Sirt1 to displace RXR as the binding partner of PPARα especially when RXR is downregulated and where RXR has a low affinity for binding to the DR1, as in the case of imperfect DR1s.

Our data suggest that the PPARα/Sirt1 heterodimer primarily forms on imperfect DR1s because of the weak binding of RXR to the imperfect DR1s. In contrast, since perfect DR1s provide dedicated and more secure binding sites for both PPARα and RXR, replacement of RXR with Sirt1 rarely takes place at these sites (Figure 6). We propose that upregulation of Sirt1 and parallel downregulation of RXR during pathological hypertrophy induces a switch from PPARα/RXR to PPARα/Sirt1 on imperfect DR1s, thereby causing downregulation of many known PPARα target genes that mediate FA utilization. This model well explains why PPAR targets harboring an imperfect DR1, but not a perfect DR1, are suppressed during PO. It should be noted that there are several PPAR targets harboring perfect DR1s that are suppressed during PO, including Fatp1, Hmgsc2 and Lpl (Supplementary Figure 2). We speculate that these PPAR targets may also harbor imperfect DR1s that have not been reported. Further investigation is needed to clarify this issue.

Figure 6.

Schematic representation of transcriptional regulation on the imperfect DR1. PPARα/RXR and PPARα/Sirt1 dimers facilitate transcriptional activation and suppression, respectively.

Although the decline in FA metabolism is a hallmark of the metabolic change that takes place during HF, whether the decline in FA utilization is adaptive or maladaptive in the failing heart has been debated²⁷. We here show that both the decreased FA oxidation and cardiac dysfunction during PO were ameliorated in the presence of a peptide harboring the RXRα(DBD) that disrupts PPARα/Sirt1 dimerization (Figure 5F). Our results not only show the importance of competitive binding of Sirt1 and RXR to PPARα in mediating cardiac dysfunction in response to PO, but also suggest that downregulation of FA utilization through this Sirt1-dependent mechanism plays a causative role in mediating PO-induced HF.

In conclusion, PPARα targets regulated by imperfect DR1s are flexibly regulated by dimer transition between PPARα/RXR and PPARα/Sirt1, leading to metabolic remodeling in the heart in response to PO. Preventing PPARα/Sirt1 and/or restoring PPARα/RXR heterodimerization may have therapeutic potential in combating HF.