• ERRα;
  • PGC-1α;
  • PPAR;
  • pyruvate dehydrogenase kinase;
  • skeletal muscle


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References

Pyruvate dehydrogenase kinase 4 (PDK4) is a key regulatory enzyme involved in switching the energy source from glucose to fatty acids in response to physiological conditions. Transcription of the PDK4 gene is activated by fasting or by the administration of a PPARα ligand in a tissue-specific manner. Here, we show that the two mechanisms are independent, and that ERRα is directly involved in PPARα-independent transcriptional activation of the PDK4 gene with PGC-1α as a specific partner. This conclusion is based on the following evidence. First, detailed mutation analyses of the cloned PDK4 gene promoter sequence identified a possible ERRα-binding motif as the PGC-1α responsive element. Second, overexpression of ERRα by cotransfection enhanced, and the knockout of it by shRNAs diminished, PGC-1α-dependent activation. Third, specific binding of ERRα to the identified PGC-1α responsive sequence was confirmed by the electrophoresis mobility shift assay. Finally, cell-type-specific responsiveness to PGC-1α was observed and this could be explained by differences in the expression levels of ERRα, however, ectopic expression of ERRα in poorly responsive cells did not restore PGC-1α responsiveness, indicating that ERRα is necessary, but not sufficient for the response.


Dulbecco's modified Eagle's medium


estrogen related receptor α


forkhead box class O


green fluorescent protein


glucocorticoid receptor


pyruvate dehydrogenase kinase


pyruvate dehydrogenase kinase 4


peroxisome proliferator-activated receptor γ co-activator 1α


peroxisome proliferator-activated receptor α


peroxisome proliferator response element


poly(vinylidene difluoride)


retinoid X receptor

Metabolic switching of oxidative fuel use from glucose and three-carbon compounds to fatty acids is a key adaptive mechanism to maintain blood glucose levels during starvation [1,2]. To limit the irreversible decarboxylation of pyruvate to acetyl-CoA, pyruvate dehydrogenase kinases (PDKs) are induced by starvation and phosphorylate the pyruvate dehydrogenase complex that catalyzes decarboxylation. Of four isoforms, PDK4 has been reported to be the most highly inducible [3]. It is known that PDK4 is induced in the liver and muscles of mice by fasting [4,5] and also by the administration of a peroxisome proliferator-activated receptor α (PPARα) ligand [6].

We reported previously that hypolipidemic drug fibrates, PPARα ligands, rapidly induce PDK4 mRNA in various mouse tissues [7]. This is in stark contrast to the relatively slow induction of the mRNAs for fatty acid utilization which are mostly limited to the liver. The observation that fibrates rapidly and efficiently induced mRNA at the whole-body level prompted us to propose a metabolic-switching hypothesis as the first step in the hypolipidemic effect of fibrates. Rapid and general inactivation of the pyruvate dehydrogenase complex by fibrate-induced PDK4 is followed by metabolic switching to limit oxidative fuel to fatty acids, thus enhancing fatty acid utilization. This is the driving force for fatty acid uptake from the blood, and uptake is further accelerated by induction of the enzymes involved in fatty acid utilization and repression of the inhibitor apoCIII for fatty acid-producing lipoprotein lipase. A metabolic-switching hypothesis predicts that these changes will contribute sequentially to the rapid and sustained hypolipidemic effect of fibrates [7].

The induction of PDK4 by fasting has been studied from the viewpoint of the action of insulin in inhibiting the ability of dexamethasone to stimulate PDK4 gene expression [8]. The mechanism responsible for the insulin inhibition of glucocorticoid-stimulated human PDK4 gene expression was extensively analyzed by Harris and co-workers using human hepatoma HepG2 cells [9]. They proposed a model in which glucocorticoid activates PDK4 gene expression via the glucocorticoid-responsive element located −820 from the transcriptional start site. Activation occurs in a glucocorticoid receptor-dependent manner in cooperation with forkhead box class O (FOXOs) bound at three sites in the promoter. Insulin activates protein kinase B to phosphorylate FOXOs [10,11], leading to the inhibition of glucocorticoid-stimulated PDK4 gene expression.

It is not known whether the regulatory mechanism for PDK4 gene expression in skeletal muscle is the same as in liver. Nor is the relationship between the PPARα-dependent mechanism and other physiological responses currently understood. The unique tissue distribution and kinetic characteristics of the PDK isoforms suggest the existence of tissue-specific regulation by PDKs [4,12], but the detailed mechanisms have not been elucidated. Sugden and co-workers reported that the PPARα pathway and other pathways that induce PDK4 were parallel or overlapping, and that there was, at least in part, a fatty acid-dependent, PPARα-independent pathway in liver and cardiac muscle [13,14]. Consistent with these studies, we report a PPARα-independent pathway that induces PDK4 in skeletal muscle, and we demonstrate that the orphan nuclear receptor ERRα is directly involved in the mechanism by recruiting PGC-1α as a coactivator.


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References

Activation of the mouse PDK4 gene promoter by PGC-1α independent of PPARα

PDK4 mRNA is induced in the liver and skeletal muscle of mice by fasting or by the administration of a PPARα ligand [6,7], and PPARα ligand-induced expression has been reproduced in cultured cells such as rat hepatoma Fao cells and mouse myotubles differentiated from myoblast C2C12 cells [15]. To analyze the molecular mechanism of transcriptional activation of the PDK4 gene, we cloned the mouse 2.3 kb promoter sequence into a luciferase plasmid and performed a reporter gene assay. The cloned promoter did not respond to a PPARα ligand with or without cotransfection of PPARα and retinoid X receptor (RXRα) expression plasmids in the cells (e.g. rat hepatoma Faco, mouse myoblast C2C12 or Africa green monkey kidney CV-1). During the search for coactivators to enhance PPARα-dependent transcriptional activation, we found that PGC-1α stimulated expression of the reporter gene containing the PDK4 promoter in CV-1 cells (Fig. 1A). A synergistic effect of the PPARα/RXRα pair and PGC-1α was seen on activation of an artificial promoter containing a tandem repeat of the typical peroxisome proliferator response element (PPRE) of the rat acyl-CoA oxidase gene promoter (positive control) [16], whereas there was little effect of the PPARα/RXRα pair on activation of the PDK4 gene promoter with or without PGC-1α, although expression was stimulated by PGC-1α alone. To understand the physiological importance of PPARα/RXRα pair-independent and PGC-1α-dependent transcriptional activation of the PDK4 gene, induction of PDK4 mRNA in PPARα-null mice was compared with that in wild-type mice under different conditions. Northern blot analysis of total RNA from liver and skeletal muscle revealed that PDK4 mRNA was largely induced by two days of fasting, even in the muscle of PPARα-null mice (Fig. 1B). A PPARδ-specific ligand GW501516 had little effect on PDK4 expression, suggesting that the PDK4 gene is a target for PPARα, but not PPARδ in muscle; this is in contrast to a previous report using cultured cells [17]. Thus, there are at least two pathways to activate transcription of the PDK4 gene in muscle; one depends on PPARα/RXRα and the other does not. The PGC-1α-dependent but PPARα/RXRα-independent pathway observed in the in vitro study above (Fig. 1A) may correspond to the fasting-induced pathway in vivo (Fig. 1B).


Figure 1. Expression of the mouse PDK4 gene is activated in PPARα-dependent and -independent ways. (A) PGC-1α activates the PDK4 gene promoter in the absence of PPARα/RXRα in CV-1 cells. The 2.3 kb mouse PDK4 gene promoter sequence was cloned into the pGL3-basic firefly luciferase reporter plasmid and a reporter gene assay was carried out in CV-1 cells with or without expression plasmids of PPARα, RXRα and PGC-1α. The AOx PPRE promoter was used as a positive control for PPARα ligand responsiveness and pRL-TK-Rellina luciferase plasmid was cotransfected for the internal control of transfection. Transfected cells were cultured with (closed bar) or without (open bar) the PPARα ligand Wy14,643 for 48 h. Mean relative luciferase activities are shown (± SD). The results shown are representative of more than three independent experiments performed in triplicate. (B) Fasting induced PDK4 mRNA in the skeletal muscles of both wild-type and PPARα knockout mice. Wild-type (+/+) and PPARα knockout mice were fed control diet, containing 0.2% bezafibrate (PPARα agonist) or 0.05% GW501516 (PPARδ agonist), or were fasted for 2 days. Total RNA from skeletal muscle was analyzed by northern blotting using cDNAs for PDK4 and CTE1.

Download figure to PowerPoint

Localization of the PGC-1α responsive regions in the mouse PDK4 gene promoter

To identify the partner of PGC-1α and activate mouse PDK4 gene expression, we first analyzed the PGC-1α responsive regions in the 2.3 kb promoter. The reporter assay, using a series of upstream deletion mutants, indicated that the sequence between −391 and −386 (the transcriptional start site is +1, according to the NCBI Accession no. NT_039340.5, GI:82802238) is important for PGC-1α activation (Fig. 2A). In addition to this region, the reporter assay, using a series of internal deletion mutants, suggested that two regions between −434/−370 and −370/−300 had variable effects on the activation (Fig. 2B). To clarify the relationship among these three regions, reporter plasmids containing the fragments −400/−370, −369/−338 and −337/−301 in all combinations were constructed and their responsiveness to PGC-1α examined (Fig. 2C). All plasmids containing the fragment −400/−370 responded markedly to PGC-1α, but plasmids without the fragment responded poorly. Addition of fragment −369/−338 increased both basal and PGC-1α-activated expression levels in all four constructs, whereas fragment −337/−301 had little effect on expression levels in any combination. From these results, we concluded that the region necessary for PGC-1α responsiveness in the cloned PDK4 promoter is around −390 from the transcriptional start site, and the region −369/−338 contributes to the basal activity. There is a possible Sp1-binding site in the region −369/−338 (Fig. 3A).


Figure 2. Localization of the region for PGC-1α responsiveness in the 2.3 kb promoter of the mouse PDK4 gene. (A) 5′-Upstream deletions. Luciferase reporter plasmids with serial deletions of the 5′-upstream sequence of the promoter were constructed and their responsiveness to PGC-1α was examined by cotransfection with (closed bar) and without (open bar) PGC-1α. (B) Internal deletions. Luciferase reporter plasmids with various internal deletions were constructed and their responsiveness to PGC-1α was examined as in (A). (C) Combinations of three regions. The three regions (−400/−370, −369/−337 and −336/−301) identified as having some effects on the responsiveness to PGC-1α were arranged in every combination in the upstream of the pGL3-promoter vector and their responsiveness to PGC-1α was examined in CV-1 cells, as in (A). Nucleotides are numbered relative to the transcription initiation site (+1). The mean relative luciferase activities measured at 48 h after transfection are shown (± SD). Results shown are representative of three independent experiments performed in triplicate.

Download figure to PowerPoint


Figure 3. The ERRα, but not the FOXO1, binding site in the PDK4 promoter is necessary for the responsiveness to PGC-1α. (A) Alignment of the promoter sequences of the PDK4 genes in humans (hgPDK4), mice (mgPDK4) and rats (rgPDK4). The nucleotide sequences conserved among three species are shown in characters on the black background and gaps that were introduced to maximize homology are shown by dashes. Nucleotides are numbered relative to the transcription initiation site (+1). Possible binding sites for ERRα, FOXO1 and Sp1 are underlined and labeled. (B) Effect of mutations in the FOXO1 binding site on the responsiveness to PGC-1α. The possible FOXO1 binding site (TAAACAAGGA, boxed) in the −391 promoter was changed to (TGGGGAAGGA) as described in Experimental procedures and its effect on the responsiveness was examined by the reporter assay using cotransfection with (closed bar) and without (open bar) PGC-1α in CV-1 cells. (C) Effect of mutations in the ERRα binding site on the responsive to PGC-1α. The possible ERRα binding site (TGACATT, boxed) in the −398 promoter was changed to GTACATT (mut1) or TGGACGT (mut2) and the responsiveness was examined as in (B). (D) Effect of mutations in the ERRα binding site in the context of the 2.3 kb promoter. The −398 promoter fragments containing mut1, mut2 or wild-type sequences were inserted into the wild-type 2.3 kb promoter between the SacII (−427) and HindIII sites, producing −2.3 k m-mut1, −2.3 k m-mut2 and −2.3 k m-WT plasmids, respectively. Their responsiveness to PGC-1α was examined by comparing with that of the −2.3 k WT promoter. Mean relative luciferase activities measured at 48 h after transfection are shown (± SD). Results shown are representative of two independent experiments performed in triplicate.

Download figure to PowerPoint

Mutation analysis of the putative transcription factor binding sites

Comparison of the reported PDK4 promoter sequences in humans, mouse and rat [18] (Fig. 3A) confirmed that the identified region for PGC-1α responsiveness (around −390) is conserved among the three species, and an in silico search for possible transcription factors to bind to this and adjacent regions suggest two candidates, ERRα[19] for the exact region and FOXO1 for the adjacent region (Fig. 3A), both of which have been reported to be partners for PGC-1α in several genes. Furthermore, Furuyama et al. [20] suggested the involvement of FOXO1 in the transcriptional activation of the PDK4 gene. Therefore, we first examined the possibility of FOXO1 being the PGC-1α partner in the activation by using a mutant plasmid constructed to lack FOXO1-binding ability (Fig. 3B). A reporter assay using wild-type or mutant promoter plasmid showed no significant difference between the two, indicating that FOXO1 present in CV-1 cells did not participate in PGC-1α-dependent activation of the truncated promoter. We then examined ERRα with mutant plasmids, as shown in Fig. 3C, using a reporter gene assay. Substitution or deletion of the ERRα-binding site abolished the responsiveness to PGC-1α. This site is not an accessory but rather is essential because substitutions of the motif in the 2.3 kb promoter abolished PGC-1α-dependent activation (Fig. 3D). This result indicated that the exact ERRα-binding site was essential for PGC-1α-dependent activation of the truncated promoter and suggested that ERRα present in CV-1 cells was involved.

ERRα is a partner of PGC-1α in the activation of PDK4 gene expression

To confirm that ERRα is a partner of PGC-1α in transcriptional activation of the cloned PDK4 promoter, the effect of variable expression levels of ERRα on PGC-1α-dependent activation was examined by overexpressing and then knocking out ERRα. Overexpression of ERRα by the cotransfection of increasing amounts of its expression plasmid in CV-1 cells did not show enhancement but rather inhibition of the PGC-1α-dependent activation of the truncated promoter (not shown). However, the same experiment, using C2C12 cells showed some increase in transcription enhancement (Fig. 4A). The excess expression of ERRα in C2C12 cells showed a noticeable, but relatively small, enhancement, which together with the negative result obtained for CV-1 cells, suggests that the ratio of ERRα to other transcriptional factors is critical to establish the activation complex. To further define the essential role of ERRα in activation, RNA interference was used to specifically knockout ERRα in CV-1 cells. A mixture of three RNAi sequences almost completely inhibited PGC-1α-dependent transcriptional activation, although any one of the sequences alone did not significantly reduce activation (Fig. 4B). In agreement with this, a reduction in ERRα protein was seen using western blotting in CV-1 cells transfected with three shRNA plasmid DNAs (Fig. 4C). Insufficient reduction (45% of control) in total ERRα protein levels with almost complete inhibition of PGC-1α-dependent transcriptional activation was explained by the incomplete transfection efficiency of the plasmid DNAs under the conditions used, as suggested by a control experiment using a green fluorescent protein (GFP) plasmid (Fig. 4D).


Figure 4. Effects of overexpression or knockout of ERRα on the responsiveness of the promoter to PGC-1α. (A) A dose-dependent effect of the overexpression of ERRα on the responsiveness in C2C12 cells. Increasing amounts (0.001, 0.01 and 0.1 µg·well−1) of mouse ERRα expression plasmid were transfected and the responsiveness to PGC-1α was examined by cotransfection with (closed bar) and without (open bar) PGC-1α. Asterisks denote significant differences (*P < 0.05, **P < 0.01, t-test). (B) Effect of antisense knockout ERRα on the responsiveness in CV-1 cells. CV-1 cells were transfected with one or a mixture of three ERRα-specific shRNA plasmids (A, B, C) or with a mutated control shRNA plasmid (negative) plus the −391 promoter plasmid with (closed bar) or without (open bar) PGC-1α. Total amounts of the shRNA plasmid DNA were adjusted to 150 ng·well−1. Mean relative luciferase activities measured at 48 h after transfection are shown (± SD). Results shown are representative of two independent experiments performed in triplicate. (C) Western blot analysis of the reduced levels of ERRα in cells transfected with a mixture of three plasmids (all) but not with a control plasmid (N). The band intensity on the X-ray film was quantified using NIH image j software and the relative amounts are given as percent of control. (D) Transfection efficiency was evaluated by introducing a CMV-GFP plasmid under the same conditions. Transfected cells were examined using confocal fluorescence microscopy.

Download figure to PowerPoint

ERRα binds to the PGC-1α-responsive region in the PDK4 promoter

To show that ERRα binds to the possible responsive element, we performed a gel-shift assay using oligonucleotides corresponding to −396/−366 with or without mutation and a nuclear extract from CV-1 cells. A specific shifted band was detected when the oligonucleotide with the wild-type sequence was used as the labeled probe, but not when the mutated oligonucleotides were used (Fig. 5A), and the shifted band was readily out-competed only by the oligonucleotide with the wild-type sequence (Fig. 5B). Furthermore, the specific shifted band disappeared when the extract was preincubated with anti-ERRα IgG, but not when preincubated with control IgG (Fig. 5C). These results demonstrate that ERRα present in CV-1 cells specifically bound to the responsive element in the promoter region of the mouse PDK4 gene.


Figure 5. Electrophoresis mobility shift assay (EMSA) at sequence −396 to −366 of the PDK4 promoter. Double-stranded oligonucleotides corresponding to the DNA sequence of the PDK4 promoter region −396 to −366 were used as the wild-type (WT) probe in EMSA. (A) Binding assay using labeled WT and mutant probes. WT and mutant probes (Fig. 3C) were labeled with [32P]dCTP[αP] and incubated with 3, 6 or 12 µg nuclear extract prepared from CV-1 cells. The arrow indicates the specific band of interest and F indicates free probes. (B) Cold competition assay using excess amounts of cold WT and mutant oligonucleotides. We incubated 20 fmole of labeled WT probe with 10 µg of nuclear extract protein with or without nonlabeled competitors. Competition was performed at 50- and 500-fold molar excess of WT or mutant oligonucleotides, as indicated above the appropriate lanes. The arrow indicates the specific band of interest and F indicates free probes. (C) Effect of anti-ERRα IgG on specific binding. EMSA assay with WT probe was performed in the presence of specific anti-ERRα IgG (5 µg) or control rabbit IgG (5 µg). The arrow indicates the specific band of interest and F indicates free probes.

Download figure to PowerPoint

An abundance of ERRα in cells explains a cell-type-specific response of the PDK4 promoter to PGC-1α

In this study, we noticed that PGC-1α-dependent enhancement of transcriptional activation of the truncated PDK4 promoter was seen when CV-1 and C2C12, but not Fao, cells were used. A representative result to compare the responsiveness to PGC-1α among the three cell lines is shown in Fig. 6A. To examine the possibility that the distinction was caused by differences in the levels of endogenous ERRα, western blot analysis of nuclear extracts from the three cell lines was performed using an antibody against the peptide corresponding to the conserved amino acid sequence of ERRα among the three species (amino acids 339–364, SVHIEDAEAVEQLREALHEALLEYEA of human ERRα). As shown in Fig. 6B, the ERRα expression level was significantly lower in Fao cells than in CV-1 and C2C12 cells, and the amounts of ERRα roughly corresponded to the gel-shift activities when the assay was performed using nuclear extracts from the three cell lines and the probe described in Fig. 5 (Fig. 6C). Thus, the unresponsiveness of the truncated PDK4 promoter in Fao cells to PGC-1α can be explained by a low abundance of endogenous ERRα, although this does not exclude the possibility that another essential factor for PGC-1α-dependent transcriptional activation is also missing in Fao cells.


Figure 6. Cell-type-specific response of the −530 PDK4 promoter to PGC-1α. (A) A reporter gene assay was performed with the −530 promoter in the presence (closed bar) or absence (open bar) of PGC-1α in three cell lines, CV-1, C2C12 and Fao cells, as in Figs 1–4. Mean relative luciferase activities measured at 48 h after transfection are shown (± SD). Results shown are representative of three independent experiments performed in triplicate. (B) Western blot analysis was performed to determine the expression levels of ERRα in three cell lines, CV-1, C2C12 and Fao cells. The same amounts of nuclear extracts (50 µg protein) were analyzed by western blotting using anti-ERRα serum. The arrow indicates the specific band and NS indicates a nonspecific band according to the manufacturer's catalog. (C) EMSA was performed to determine the relative band shift activities in the nuclear extracts prepared from three cell lines. The probe and the conditions were as in Fig. 5. The arrow indicates the specific band of interest.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References

We identified ERRα as an essential partner of PGC-1α in the PPARα-independent activation of the mouse PDK4 gene promoter in C2C12 myotube cells. In addition to its previously known function in regulating the amounts of energy-creating enzymes [21], PGC-1α seems to play an essential role in metabolic switching by collaborating with the muscle-abundant transcription factor ERRα.

During preparation of this manuscript, Ma et al. reported activation of the rat PDK4 gene by PGC-1α[18]. Using an adenovirus expressing PGC-1α, they showed that it activates PDK4 gene transcription in rat hepatocytes and ventricular myocytes. Further in vitro analyses were carried out in human hepatoma HepG2 cells and HNF4 was identified as a transcription factor that stimulates PDK4 promoter activity. However, HNF4 is not directly involved in the recruitment of PGC-1α, and Ma et al. suggested that FOXO1 might participate in the recruitment, based on the fact that a FOXO1-binding site is conserved among humans, mice and rats.

Our results on the PGC-1α-responsive site in the 2.3 kb mouse PDK4 gene promoter obtained using mutation analyses do not support speculation that FOXO1 is a partner of PGC-1α in this region (Fig. 3B), but they show that the partner is ERRα, as stated above. Our data do not contradict the model proposed by Kwon et al. that multiple FOXO1s are involved in the regulation with glucocorticoid receptor (GR) by binding to different regions [9]. We concentrated on the identification of a direct PGC-1α partner and did not search for other independent factors. So, the possible independent involvement of HNF4 from PGC-1α on the gene activation remains, and at least two mechanisms are able to explain the independent involvement of ERRα and HNF4. One explanation is that the two factors act on the promoter in the same cell, each factor responding to a different signal. The other explanation is that a given cell type preferentially uses one factor rather than the other. We think that the latter is more likely because the expression profile of ERRα in various tissues differs from that of HNF4 [22]. In particular, it is known that the expression level of ERRα is high, whereas that of HNF4 is low, in skeletal muscle, and vice versa in liver [23,24]. Ma et al. analyzed the mechanism using hepatoma cells and suggested that HNF4 will stimulate PDK4 gene expression, at least in liver [18]. We analyzed it using kidney CV-1, myoblast and hepatoma cells, and concluded that ERRα enhances PDK4 expression with PGC-1α in myoblast but not hepatoma cells. Consequently, our model in skeletal muscle and the model proposed by Ma et al. for liver [18] seem to be compatible, as shown in Fig. 7. Both mechanisms may operate in either tissue to varying degrees, depending on the expression levels of ERRα and HNF4. FOXO1 is not included in the model because we have not directly studied its involvement in the mechanism. To understand the overall regulation of PDK4 expression, it is essential to study the relationship between the ERRα/PGC-1α and the FOXO1/GR pathways in detail.


Figure 7. Proposed model depicting the cell-type-specific response of the PDK4 promoter to PGC-1α in skeletal muscle and liver. (A) In skeletal muscle, PGC-1α is specifically recruited by ERRα and the complex activates transcription (this study). The pathway to activate ERRα expression is taken from Sato et al.[35]. (B) In liver, PGC-1α and HNF4 independently activate transcription. The partner of PGC-1α is not yet identified. These liver-specific pathways are taken from Ma et al. [18].

Download figure to PowerPoint

The attempt to identify the ERRα-regulated pathway using primary myocytes overexpressing ERRα conducted by Huss et al. [25] did not assign PDK4 as a gene that is upregulated by ERRα. Instead, pyruvate dehydrogenase phosphatase I, which reverses the action of PDK4 and activates pyruvate dehydrogenase complex, was identified as a being upregulated by ERRα. Although Huss et al.'s approach did not identify all the genes directly regulated by ERRα, they did identify a subset of PGC-1α target genes involved in multiple energy-production pathways. This may be because the difference in cell types (cardiac myocytes vs. skeletal myoblast) was critical and/or that transcriptional control of PDK4 gene expression is such a complex process that some complementary effects cancelled out the enhanced expression. Contrary to the overexpressing approach, Luo et al. generated ERRα-null mice and reported the differentially expressed genes only in the adipose tissue because of their lean phenotype [26]. Interestingly, PDK4 expression was increased by 1.8-fold in the adipose tissue of ERRα-null mice. The expression of several genes directly linked to lipid and energy metabolism was also altered, resulting in a defect in triglyceride synthesis. Luo et al. suggested that the loss of ERRα might interfere with other nuclear signaling pathways. If so, the direct cause of their result showing increased PDK4 expression in the adipose tissue may not be the loss of ERRα. Thus, both the forced expression and loss-of-function studies using animal models to understand the in vivo role of ERRα are not straightforward, and the results reported by other groups do not necessarily conflict with our conclusion.

We also noticed that the PGC-1α and ERRα pathway leading to PDK4 activation was not simple. During the ERRα overexpressing experiment (Fig. 4A), dose-dependent enhancement of PGC-1α-dependent activation of the PDK promoter was observed only in C2C12 cells. Although the lack of responsiveness to PGC-1α in hepatoma Fao cells could be explained by the undetectable level of endogenous ERRα (Fig. 6), the positive effect of the cotransfected ERRα expression plasmid was not observed at various doses (not shown). It is conceivable that the expression level is important in reproducing the essential functions of a regulatory protein interacting with many other effector proteins. Furthermore, there is the possibility that another unidentified muscle-specific factor is needed to activate the PGC-1α and ERRα pathway. Thus, we conclude that ERRα is necessary but not sufficient for the PGC-1 responsiveness of PDK4 gene expression in the muscle.

Wide and essential roles of PGC-1α in regulating the metabolism, including the maintenance of glucose, lipid and energy homeostasis, have been revealed by recent studies [27]. In skeletal muscle, the most characterized signals including PGC-1α would be the MEF2 pathway, which activates MEF2-dependent genes [28]. In fact, we first examined the possibility that MEF2 was involved in PGC-1α-dependent activation of the PDK4 promoter. However, cotransfection of the MEF2 expression plasmid had a dose-dependent inhibitory effect on PGC-1α-dependent activation. We now think that the ERRα pathway leading to the activation of the PDK4 is independent of the MEF2 pathway and speculate that the inhibition was caused by competition for PGC-1α between the two factors. Further possible inhibitory interactions between ERRα and PGC-1α suggested by Ichida et al. [23] may also be involved in the complex regulation. Nevertheless, the two independent pathways use PGC-1α as a common and essential coactivator and each pathway has its own physiological significance. Therefore, there must be some mechanism to discriminate between the two pathways and we again emphasize the possibility that unidentified factors are essential to separately activate these PGC-1α pathways.

Wende et al. [29] reported that the PGC-1α-mediated transcriptional activation of the mouse PDK4 gene is independent of PPARα and FOXO1, but it is dependent on ERRα. Their conclusion is apparently the same as ours, but they suggested a different binding site for ERRα (AAGGACA/TGTCCTT; bp −357 to −352) from the one that we have identified (TGACATT; bp −389 to −383) (Fig. 3) and both sequences do not perfectly matched with the consensus sequence, TGACCTT [19]. In contrast to our direct demonstration, Wende et al. used a long DNA fragment containing both possible sites in their binding studies. Their result showing the loss of the responsiveness to PGC-1α by the mutation in the suggested region can be explained by our result that the region is important for basal activity (Fig. 2C). Not ERRα, but an unidentified factor, should bind to the site suggested and may play an important role in basal transcription.

In this study, we characterized the PPARα-independent transcriptional activation of the mouse PDK4 gene and concluded that it is activated via the ERRα/PGC-1α pathway. We speculate that this pathway mediates the signals produced by fasting to induce PDK4 in the skeletal muscle. However, there is no direct evidence to indicate that the ERRα/PGC-1α pathway is responsible for transcriptional activation of the PDK4 gene in muscle in response to fasting. Further studies to directly link the result obtained using cultured cells and those that occur in animals are necessary to understand the in vivo mechanism and its physiological significance. In addition, characterization of PPARα-dependent activation of the PDK4 gene and its relation to other activation pathways remain important issues.

Experimental procedures

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References


PPARα ligands, Wy14,643 (4-chloro-6-(2,3-xylidino)-2-pyrimidinyl–thio) acetic acid and bezafibrate were purchased from Tokyo-Kasei (Tokyo, Japan) and Sigma (St. Louis, MO), respectively. The PPARδ-specific ligand, GW501516 [30], was synthesized by and received as a gift from Nippon Chemiphar (Misato, Japan). Dual-Luciferase Reporter Assay System and pGL reporter vectors were purchased from Promega (Madison, WI). T4 DNA polymerase, Klenow fragment, random primer labeling kit and restriction enzymes were purchased from TaKaRa Bio (Osaka, Japan). KOD polymerase was purchased from TOYOBO (Osaka, Japan). [32P]dCTP[αP] (3000 Ci·mmol−1) was purchased from Muromachi Chemicals (Tokyo, Japan). Anti-ERRα serum against the peptide corresponding to amino acids 339–364 of human ERRα was purchased from Upstate (Lake Placid, NY). Super Signal West Pico Chemiluminescent Substrate was purchased from Pierce (Rockford, IL). Expression plasmids of shRNA for ERRα were purchased from ORIGENE (Rockville, MD). The shRNA sequences cloned in the plasmids are: shRNA-A, GCATGCTCAAGGAGGGAGTGC; shRNA-B, GGCAGTCGCTGGAGGCCCCCG; shRNA-C, CTGCCCCTTGCAAGCCATAAC.

RNA isolation and northern blot analyses

Total RNA from the skeletal muscle of mice fed with a control diet or one containing 0.2% (w/w) bezafibrate, or 0.05% (w/w) GW501516 was isolated and analyzed essentially as described previously using an Express Hyb hybridization solution (Clontech, Palo Alto, CA) [31]. The cDNA probes used to detect PDK4, cytosolic thioesterase 1 (CTE1) and cyclophilin mRNAs were as described previously [15,32].

Plasmid construction

The mammalian expression plasmids for PPARα and RXRα were as described previously [33]. Expression plasmids for PGC-1α and ERRα were obtained by PCR cloning, as described previously [31], using the primer pairs, 5′-AATTCGGCTCGAGGTTGCCTGCATGAGT-3′ and 5′-TGGAAATGCGGCCGCTTGAAGGGGT-3′ for PGC-1α, and 5′-GGCAAGCTTGCCTGCCCGCCAGCCCTG-3′ and 5′-GCAGATATCGGCACTGCCCAGACTCCAGC-3′ for ERRα. The PCR product was digested using XhoI and NotI, and cloned into the XhoI/NotI site of pOPRSVI/MCS vector (Stratagene, La Jolla, CA). The PCR product was blunted with KOD polymerase and cloned into HindIII/EcoRI sites of pCMX vector after blunting. Their identities were confirmed by sequencing.

Gel mobility shift assay

Nuclear extract was prepared from culture cells as described previously [34]. Double-stranded DNA fragments corresponding to −396/−366WT, −396/−366mut1 and −396/−366mut2 (Fig. 5A) were prepared by annealing the oligonucleotides having additional GG dinucleotides at the 5′-end (5′-GGGAATGCGTGACATTGAGATGGCTCTGGAGT-3′, 5′-GGGAATGTAGTACATTGAGATGGCTCTGGAGT-3′ and 5′-GGGAATGCGTGGACGTGAGATGGCTCTGGAGT-3′) and their complementary oligonucleotides with additional GG dinuleotides at the 5′-end, respectively. The 3′-terminal dinucleotides CC were labeled by a fill-in reaction with [32P] dCTP[αP] using the Klenow fragment. Gel mobility shift assays were performed as described previously [16]. The gels were dried and analyzedusing an image analyzer, Typhoon (Amersham).

Cloning of the mouse PDK4 gene promoter and construction of reporter plasmid

The 2.3 kb mouse PDK4 promoter sequence was amplified by PCR using mouse genomic DNA and the primer pair, 5′-GCATCGGATGATGCCTGAAAGGAA-3′ and 5′-GTCGAGAGGGAGCAAGTGCGTTGT-3′, and cloned into the HincII site of pUC18 after blunting the ends with KOD polymerase. After confirmation of its identity by sequencing, the 2.3 kb promoter fragment was cut out by digestion with KpnI and HindIII, and cloned into the KpnI/HindIII site of the pGL3-basic vector (Promega). The deletion mutants of the mouse PDK4 promoters were created by PCR using the left primers with the KpnI site and the right primers with the HindIII site. PCR products were doubly digested with KpnI and HindIII and inserted between the KpnI and HindIII sites of the pGL3-basic vector. Reporter plasmids of −398mut1 and −398mut2 (Fig. 3) were created with PCR, using the left mutant primers, 5′-CCCGGTACCAGGAATGTAGTACATTGAGAT-3′ for −398mut1 and 5′-CCCGGTACCAGGAATGCGTGGACGTGAGAT-3′ for −398mut2. These PCR products were digested with HindIII and KpnI and inserted between the KpnI and HindIII site of the pGL3-basic vector. The 2.3 kb full-length promoter plasmids containing mut1 or mut2 mutation were constructed by removing the SacII (−427) and HindIII fragment from the wild-type 2.3 kb promoter vector and inserting the KpnI (−398) and HindIII fragment from the −398 promoter plasmid containing mut1 or mut2 mutation after blunting the cohesive ends of the SacII and KpnI sites using T4 DNA polymerase. For a wild-type control plasmid, the plasmid containing the wild-type KpnI (−398)/HindIII fragment inserted between the SacII (−427) and HindIII site was constructed and designated as −2.3 k m-WT (Fig. 3D). The FOXO1 mutant reporter plasmid was constructed starting from the 391 bp PDK4 promoter sequence with a PCR-assisted method using a site-directed mutagenesis kit, Mutant-Super Expression KMTM (TaKaRa Bio) and the mutant primer of the sequence 5′-GGCTCTGGAGTTGTAGGGGAGGACAAGTC-3′.

Construction of internal deletion mutants of reporter plasmids

Δ−370/−300 plasmid (Fig. 2B) was created as follows. The −530/−370 fragment was obtained by PCR using the primer pair, 5′-CCCGGTACCTTGCTTTCGAGGTCCAATGGC-3′ for the left primer and 5′-AAAGGGCCCAGAGCCATCTCAATGTCACGC-3′ for the right primer. This PCR product and −530 bp reporter plasmid were digested with KpnI and ApaI, and these fragments were ligated to obtain the Δ−370/−300 plasmid. The Δ−500/−369 and Δ−500/−434 plasmids were created as follows. First, we created the Δ−500/−300 reporter plasmid. Oligonucleotides, 5′-CTTGCTTTCGAGGTCCAATGGCAAAGAAGGGCC-3′ and 5′-CTTCTTTGCCATTGGACCTCGAAAGCAAGGTAC-3′, were annealed and inserted into the −530 bp reporter plasmid that had been digested with KpnI and ApaI. Promoter fragments of −369 and −434 were obtained by PCR cloning using the left primer 5′-AAAGGGCCCAGTTGTAAACAAGGACAAGTC-3′ for −369 and 5′-AAAGGGCCCTCCGCGGTGAGATTCTTGGAA-3′ for −434 and the right primer used to clone the 2.3 kb promoter sequence above. These PCR products and the Δ−500/−300 reporter plasmid were ligated after digestion with HindIII and ApaI. The Δ−300/−80 reporter plasmid was constructed by inserting the −80 promoter sequence obtained by PCR using the left primer, 5′-AAAGGGCCCGTGGAGAGGGCGTGGAGGTAG-3′ and the right primer as above into the ApaI and HindIII site of the −530 bp reporter plasmid after digestion with the same restriction enzymes.

Construction of reporter plasmids having combinations of three regions

The −400/−301 region of the PDK4 promoter was separated into three regions, −400/−370, −369/−337 and −336/−301 (Fig. 2C). Oligonucleotides corresponding to these regions were synthesized with the upstream SalI site and the downstream XhoI site. These oligonucleotide sets were (5′-TCGACTAGGAATGCGTGACATTGAGATGGCTCTGGC-3′ and 5′-TCGAGCCAGAGCCATCTCAATGTCACGCATTCCTAG-3′) for −400/−370 (5′-TCGACAGTTGTAAACAAGGACAAGTCTGGGCGGGCCTGC-3′ and 5′-TCGAGCAGGCCCGCCCAGACTTGTCCTTGTTTACAACTG-3′) for −369/−337, and (5′-TCGACAAGTCCTAGCGACCTGGGATCTATCCCAAGGGGGC-3′ and 5′-TCGAGCCCCCTTGGGATAGATCCCAGGTCGCTAGGACTTG-3′) for −336/−301. All the reporter plasmids containing every combination of three regions were constructed by using these fragments and restriction enzymes. The organization, orientation of the fragments and sequences of these inserts in the reporter plasmids were confirmed by sequencing.

Cell culture and DNA transfection

CV-1 and C2C12 cells were cultured at 37 °C and 5% CO2 in Dulbecco's modified Eagle's medium (DMEM, Gibco, Grand Island, NY) containing 10% fetal bovine serum. Fao cells were cultured at 37 °C and 10% CO2 in F-12 coon's medium containing 5% fetal bovine serum. C2C12 was differentiated using DMEM containing 2% horse serum [16,35]. Cells were transferred into DMEM containing 2% charcoal-treated fetal bovine serum after transfection. The cells cultured in 24-well plates were transfected with 100 ng·well−1 of reporter plasmid, 1–100 ng·well−1 expression plasmid, 5 ng·well−1 of pRL-TK internal control plasmid and 0–300 ng·well−1 of salmon sperm carrier DNA to adjust the total amount of DNA to be transfected according to the assay conditions. The DNA mixtures were complexed with LipofectAmine (Invitrogen, Carlsbad, CA) with or without PLUS reagent according to the cell type. The shRNA expression plasmid DNA mixtures were added at 50–150 ng·well−1. For the negative control of shRNAs, the pRS plasmid was used.

Luciferase assay

Cells transfected with DNA mixtures were cultured for 48 h. After washing with NaCl/Pi, cells were solubilized with Passive Lysis Buffer (Promega) and assayed for both Firefly and Renilla luciferase activities using a Dual-Liciferase Reporter Assay System (Promega) according to the manufacturer's instructions [35].

Western blot analysis

Nuclear extracts prepared from culture cells were analyzed on 10% SDS-polyacrylamide gels. The separated proteins were transferred to a poly(vinylidene difluride) (PVDF) membrane. The blot was then probed with anti-ERRα sera at 1/500 dilution. Anti-(rabbit IgG) labeled with horseradish peroxidase was used as the secondary antibody, and the blot was developed using a SuperSignal West Pico Chemiluminescent Substrate (Pierce). The blot was exposed to X-ray film and PVDF membrane after exposure was stained with Coomassie Brilliant Blue for the proteins (loading control).


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References

We thank Dr H. Urabe for discussion. This study was supported by the Meiyaku Open Research Project.


  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Experimental procedures
  6. Acknowledgements
  7. References
  • 1
    Reed LJ & Hackert ML (1990) Structure–function relationships in dihydrolipoamide acyltransferases. J Biol Chem 265, 89718974.
  • 2
    Sugden MC, Bulmer K & Holness MJ (2001) Fuel-sensing mechanisms integrating lipid and carbohydrate utilization. Biochem Soc Trans 29, 272278.
  • 3
    Wu P, Sato J, Zhao Y, Jaskiewicz J, Popov KM & Harris RA (1998) Starvation and diabetes increase the amount of pyruvate dehydrogenase kinase isoenzyme 4 in rat heart. Biochem J 329, 197201.
  • 4
    Bowker-Kinley MM, Davis WI, Wu P, Harris RA & Popov KM (1998) Evidence for existence of tissue-specific regulation of the mammalian pyruvate dehydrogenase complex. Biochem J 329, 191196.
  • 5
    Wu P, Blair PV, Sato J, Jaskiewicz J, Popov KM & Harris RA (2000) Starvation increases the amount of pyruvate dehydrogenase kinase in several mammalian tissues. Arch Biochem Biophys 381, 17.
  • 6
    Wu P, Inskeep K, Bowker-Kinley MM, Popov KM & Harris RA (1999) Mechanism responsible for inactivation of skeletal muscle pyruvate dehydrogenase complex in starvation and diabetes. Diabetes 48, 15931599.
  • 7
    Motojima K (2002) A metabolic switching hypothesis for the first step in the hypolipidemic effects of fibrates. Biol Pharm Bull 25, 15091511.
  • 8
    Huang B, Wu P, Bowker-Kinley MM & Harris RA (2002) Regulation of pyruvate dehydrogenase kinase expression by peroxisome proliferator-activated receptor-α ligands, glucocorticoids, and insulin. Diabetes 51, 276283.
  • 9
    Kwon HS, Huang B, Unterman TG & Harris RA (2004) Protein kinase B-α inhibits human pyruvate dehydrogenase kinase-4 gene induction by dexamethasone through inactivation of FOXO transcription factors. Diabetes 53, 899910.
  • 10
    Biggs WH 3rd, Meisenhelder J, Hunter T, Cavenee WK & Arden KC (1999) Protein kinase B/Akt-mediated phosphorylation promotes nuclear exclusion of the winged helix transcription factor FKHR1. Proc Natl Acad Sci USA 96, 74217426.
  • 11
    Kops GJ, de Ruiter ND, De Vries-Smits AM, Powell DR, Bos JL & Burgering BM (1999) Direct control of the Forkhead transcription factor AFX by protein kinase B. Nature 398, 630634.
  • 12
    Sugden MC, Kraus A, Harris RA & Holness MJ (2000) Fibre-type specific modification of the activity and regulation of skeletal muscle pyruvate dehydrogenase kinase (PDK) by prolonged starvation and refeeding is associated with targeted regulation of PDK isoenzyme 4 expression. Biochem J 346, 651657.
  • 13
    Sugden MC, Bulmer K, Gibbons GF, Knight BL & Holness MJ (2002) Peroxisome-proliferator-activated receptor-α (PPARα) deficiency leads to dysregulation of hepatic lipid and carbohydrate metabolism by fatty acids and insulin. Biochem J 364, 361368.
  • 14
    Holness MJ, Smith ND, Bulmer K, Hopkins T, Gibbons GF & Sugden MC (2002) Evaluation of the role of peroxisome-proliferator-activated receptor-α in the regulation of cardiac pyruvate dehydrogenase kinase 4 protein expression in response to starvation, high-fat feeding and hyperthyroidism. Biochem J 364, 687694.
  • 15
    Motojima K & Seto K (2003) Fibrates and statins rapidly and synergistically induce pyruvate dehydrogenase kinase 4 mRNA in the liver and muscles of mice. Biol Pharm Bull 26, 954958.
  • 16
    Kawabe K, Saegusa H, Seto K, Urabe H & Motojima K (2005) Peroxisome proliferator-activated receptor α and its response element are required but not sufficient for transcriptional activation of the mouse heart-type fatty acid binding protein gene. Int J Biochem Cell Biol 37, 15341546.
  • 17
    Tanaka T, Yamamoto J, Iwasaki S, Asaba H, Hamura H, Ikeda Y, Watanabe M, Magoori K, Ioka RX, Tachibana K et al. (2003) Activation of peroxisome proliferator-activated receptor δ induces fatty acid β-oxidation in skeletal muscle and attenuates metabolic syndrome. Proc Natl Acad Sci USA 100, 1592415929.
  • 18
    Ma K, Zhang Y, Elam MB, Cook GA & Park EA (2005) Cloning of the rat pyruvate dehydrogenase kinase 4 gene promoter: activation of pyruvate dehydrogenase kinase 4 by the peroxisome proliferator-activated receptor γ coactivator. J Biol Chem 280, 2952529532.
  • 19
    Xie X, Lu J, Kulbokas EJ, Golub TR, Mootha V, Lindblad-Toh K, Lander ES & Kellis M (2005) Systematic discovery of regulatory motifs in human promoters and 3′ UTRs by comparison of several mammals. Nature 434, 338345.
  • 20
    Furuyama T, Kitayama K, Yamashita H & Mori N (2003) Forkhead transcription factor FOXO1 (FKHR)-dependent induction of PDK4 gene expression in skeletal muscle during energy deprivation. Biochem J 375, 365371.
  • 21
    Sano M & Schneider MD (2005) Energizer: PGC-1α keeps the heart going. Cell Metab 1, 216218.
  • 22
    Heard DJ, Norby PL, Holloway J & Vissing H (2000) Human ERRγ, a third member of the estrogen receptor-related receptor (ERR) subfamily of orphan nuclear receptors: tissue-specific isoforms are expressed during development and in the adult. Mol Endocrinol 14, 382392.
  • 23
    Ichida M, Nemoto S & Finkel T (2002) Identification of a specific molecular repressor of the peroxisome proliferator-activated receptor γ coactivator-1α (PGC-1α). J Biol Chem 277, 5099150995.
  • 24
    Nakhei H, Lingott A, Lemm I & Ryffel GU (1998) An alternative splice variant of the tissue specific transcription factor HNF4α predominates in undifferentiated murine cell types. Nucleic Acids Res 26, 497504.
  • 25
    Huss JM, Kopp RP & Kelly DP (2002) Peroxisome proliferator-activated receptor coactivator-1α (PGC-1α) coactivates the cardiac-enriched nuclear receptors estrogen-related receptor-α and -γ. Identification of novel leucine-rich interaction motif within PGC-1α. J Biol Chem 277, 4026540274.
  • 26
    Luo J, Sladek R, Carrier J, Bader JA, Richard D & Giguere V (2003) Reduced fat mass in mice lacking orphan nuclear receptor estrogen-related receptor α. Mol Cell Biol 23, 79477956.
  • 27
    Lin J, Handschin C & Spiegelman BM (2005) Metabolic control through the PGC-1 family of transcription coactivators. Cell Metab 1, 361370.
  • 28
    Lin J, Wu H, Tarr PT, Zhang CY, Wu Z, Boss O, Michael LF, Puigserver P, Isotani E, Olson EN et al. (2002) Transcriptional co-activator PGC-1α drives the formation of slow-twitch muscle fibres. Nature 418, 797801.
  • 29
    Wende AR, Huss JM, Schaeffer PJ, Giguere V & Kelly DP (2005) PGC-1α coactivates PDK4 gene expression via the orphan nuclear receptor ERRα: a mechanism for transcriptional control of muscle glucose metabolism. Mol Cell Biol 25, 1068410694.
  • 30
    Oliver WR Jr, Shenk JL, Snaith MR, Russell CS, Plunket KD, Bodkin NL, Lewis MC, Winegar DA, Sznaidman ML, Lambert MH et al. (2001) A selective peroxisome proliferator-activated receptor δ agonist promotes reverse cholesterol transport. Proc Natl Acad Sci USA 98, 53065311.
  • 31
    Motojima K, Passilly P, Peters JM, Gonzalez FJ & Latruffe N (1998) Expression of putative fatty acid transporter genes are regulated by peroxisome proliferator-activated receptorα and γ activators in a tissue- and inducer-specific manner. J Biol Chem 273, 1671016714.
  • 32
    Motojima K (2000) Differential effects of PPARα activators on induction of ectopic expression of tissue-specific fatty acid binding protein genes in the mouse liver. Int J Biochem Cell Biol 32, 10851092.
  • 33
    Fujishiro K, Fukui Y, Sato O, Kawabe K, Seto K & Motojima K (2002) Analysis of tissue-specific and PPARα-dependent induction of FABP gene expression in the mouse liver by an in vivo DNA electroporation method. Mol Cell Biochem 239, 165172.
  • 34
    Gorski K, Carneiro M & Schibler U (1986) Tissue-specific in vitro transcription from the mouse albumin promoter. Cell 47, 767776.
  • 35
    Sato O, Kuriki C, Fukui Y & Motojima K (2002) Dual promoter structure of mouse and human fatty acid translocase/CD36 genes and unique transcriptional activation by peroxisome proliferator-activated receptor α and γ ligands. J Biol Chem 277, 1570315711.
  • 36
    Mootha VK, Handschin C, Arlow D, Xie X, St Pierre J, Sihag S, Yang W, Altshuler D, Puigserver P, Patterson N et al. (2004) Errα and Gabpa/b specify PGC-1α-dependent oxidative phosphorylation gene expression that is altered in diabetic muscle. Proc Natl Acad Sci USA 101, 65706575.