Results from the present study are the first to demonstrate genome-wide changes in gene expression mediated specifically by PPARβ/δ in mouse keratinocytes. Keratinocytes provide an outstanding model to examine gene expression mediated by PPARβ/δ because of the constitutively high expression of the receptor . Comparison of wild-type and Pparβ/δ-null keratinocytes revealed that PPARβ/δ constitutively regulates expression of 482 genes. Further, activation of PPARβ/δ altered expression of 99 genes that required PPARβ/δ. The ultimate phenotype resulting from ligand activation of PPARβ/δ in keratinocytes cannot be determined based solely on the gene expression profiles obtained from these studies but they are consistent with previous studies showing that activating PPARβ/δ in keratinocytes and/or skin promotes terminal differentiation, inhibits cell proliferation, improves barrier function and inhibits inflammation [11, 32–35]. Thus, it is also worth noting that activating PPARβ/δ can induce terminal differentiation and modulate lipid metabolism in keratinocytes coincident with changes in the regulation of target genes observed in these studies, consistent with previous work [32, 36, 37]. Expression of Pdpk1, Ilk, or Pten was not altered by ligand activation of PPARβ/δ in the present studies, which is also in line with a previous study , but in contrast to another study . Increased expression of Hb-Egf and many pro-inflammatory genes was also reported in human keratinocytes and skin of transgenic mice over-expressing human PPARβ/δ, respectively, following activation with a PPARβ/δ ligand [40, 41]. However, changes in expression of these genes were not observed in the present studies. Further work is needed to determine the reason for these inconsistencies.
Genes were categorized into eight distinct response types representing essentially every combination of transcriptional activation and/or repression in response to exogenous ligand and/or PPARβ/δ disruption. Interestingly, the type I and II responses that occur independent of exogenous ligand were the most commonly observed (79%). There are two existing models that could explain this regulation. First, PPARβ/δ can dynamically occupy chromatin in association with co-repressors leading to repression of gene expression . This model explains why type I genes are induced in Pparβ/δ-null keratinocytes in the absence of exogenous ligand. Whether an endogenous ligand(s) is part of this complex remains uncertain. ChIP-seq analysis confirmed promoter occupancy of PPARβ/δ in some genes exhibiting a type I response. The second model is that the PPARβ/δ-RXR heterodimer in association with co-activators and RNA polymerase II is dynamically binding chromatin as a result of the presence of an endogenous ligand or ligands, and available chromatin binding sites as a result of the activities of DNA modifying enzymes [15–18]. Thus, when PPARβ/δ is disrupted, the expression of these type II genes is reduced. This is in agreement with the promoter occupancy of PPARβ/δ detected by ChIP-seq for some genes exhibiting a type II response. The fact that occupancy of PPARβ/δ was not detected for all of the type I and II genes is likely due to: 1) the dynamic fluid nature of PPARβ/δ binding with chromatin, which has multiple levels of regulation, 2) the fact that PPREs for these genes could exist in chromatin at distant sites not in close proximity to the TSS [15–18], or 3) local chromatin status at some PPREs may preclude binding at those sites . Of the functional categories of genes that exhibited type I or II responses, approximately 50% of these categories were altered by both response types, suggesting common factors that induce these changes are somewhat independent of the direction of the regulation observed (e.g. repression or activation).
Type III and IV responses that were dependent on exogenous ligand activation, were the next most common type of gene regulation observed (16%). These responses were characterized by a lack of constitutive regulation by PPARβ/δ, but regulatory changes arising after treatment with exogenous ligand. Type III responses are consistent with dynamic nuclear receptor-mediated up-regulation of target gene expression [15–18]. Indeed, less promoter occupancy of PPARβ/δ was found in the absence of exogenous ligand and increased promoter occupancy of PPARβ/δ was noted following ligand activation. It is curious to note that expression of type II target genes was not increased by an exogenous ligand, since it appears that constitutive expression of type II genes is driven by an endogenous ligand(s). In contrast, constitutive expression of type III response genes did not appear to be altered by an endogenous ligand, and was only increased by an exogenous ligand. There are at least two mechanisms that could explain these differences. First, the presence of an endogenous ligand in relatively high concentration (or at sufficient concentration to saturate the receptor) that also exhibits higher affinity for PPARβ/δ as compared to GW0742 could explain why an exogenous ligand does not further increase expression of a type II response gene. This would suggest that there are different endogenous ligands that cause differential conformational changes in the receptor following ligand activation and/or differences in the recruitment of co-effector proteins to the complex to modulate specific subsets of PPARβ/δ target genes. Alternatively, the concentration of an exogenous ligand could be greater than an endogenous ligand and sufficient to activate the receptor, and/or have higher affinity for PPARβ/δ than an endogenous ligand (e.g. the endogenous ligand is in low concentration and unable to activate PPARβ/δ and/or the endogenous ligand has low affinity). This scenario also implies that the ligand could uniquely alter receptor conformation and/or recruitment of co-effector proteins to the transcriptional complex as compared to that which occurs in response to an endogenous ligand. The data from the present studies also indicate that PPARβ/δ does not influence expression of type IV genes in the absence of exogenous ligand, but that ligand activation causes recruitment of PPARβ/δ to regions near the TSS and this increase in occupancy is associated with repression of gene expression for 28 genes. Whereas repression of gene expression observed for type I response genes could be mediated by dynamic occupancy of PPARβ/δ in the absence of exogenous ligands in complex with co-repressors as suggested by other studies , the precise mechanism that underlies this exogenous ligand-dependent effect is uncertain and requires further studies. In general, there was a lack of overlap between the functional categories of regulatory pathways modulated by type I/II as compared to type III/IV response genes. This supports the view that PPARβ/δ is integrated into distinct transcriptional response pathways mediated by distinct mechanisms that may be influenced by more than one endogenous ligand. The fact that PPARβ/δ has a relatively large ligand binding domain as compared to other PPARs  and nuclear receptors supports the idea that PPARβ/δ can accommodate more than one endogenous ligand. Moreover, this analysis also suggested that PPARβ/δ can tightly bind more than one type of fatty acid, a feature that interfered with solving the crystal structure of the ligand binding domain , consistent with the hypothesis that more than one endogenous ligand exists for PPARβ/δ.
Mixed responses were observed for a much smaller cohort of PPARβ/δ target genes (5%). Type V response genes exhibited responses that were similar to those found with type I and III genes because in the absence of exogenous ligand and PPARβ/δ, expression was higher yet ligand activation of PPARβ/δ caused an increase in expression. Conversely, type VI response genes exhibited responses that were similar to those found with type II and IV genes because in the absence of exogenous ligand and PPARβ/δ, expression was lower and ligand activation of PPARβ/δ caused a decrease in expression. One mechanism that explains these response types is that a ligand-mediated switch occurs between repression and activation (type V) or vice-versa (type VI) . For example, when PPARβ/δ is disrupted, expression of Angptl4 is enhanced because PPARβ/δ represses expression, whereas ligand activation of PPARβ/δ increases expression of Angptl4. This phenomenon was inversed but also observed with type VI genes, consistent with a ligand-mediated release of an activating complex. However, it is important to point out that constitutive expression Angptl4 and other type V response genes were typically much lower than the levels observed following activation by an exogenous ligand. This is consistent with a model where PPARβ/δ represses without an endogenous ligand and that binding of ligand to the receptor causes recruitment of transcriptional proteins and increased transcription after forming a transcriptional complex that increases transcription and/or outcompetes the repressive complex for repression. Alternatively, an exogenous ligand could alter the ratio of repressive PPARβ/δ complexes to activated PPARβ/δ complexes limiting the availability of PPARβ/δ to form repressive complexes on chromatin. Results from the analysis of type V genes in the present studies is in contrast to previous work by others showing comparable levels of expression of genes following either PPARβ/δ knockdown or activation by an exogenous ligand in human WPMY1 myofibroblast cells , for genes exhibiting similar regulation as observed in the present studies. This could reflect differences in: 1) the concentration of available ligand, 2) chromatin structure near the Angptl4 gene, and/or 3) co-effector proteins recruited to alter chromatin structure. The single type VII response gene exhibited a response that was similar to that found with type II and III genes because in the absence of exogenous ligand and PPARβ/δ, expression was modestly lower and ligand activation of PPARβ/δ with GW0742 caused an increase in expression. The type VII response is consistent with a model whereby an endogenous ligand drives constitutive expression and an exogenous ligand increases this expression through mechanisms described above for type II and III response types. In contrast, type VIII response genes exhibited a response that was similar to those found with type I and IV genes because in the absence of exogenous ligand and PPARβ/δ, expression was modestly higher and ligand activation of PPARβ/δ with GW0742 caused a decrease in expression. The type VIII response is consistent with a model whereby constitutive expression is repressed and exogenous ligand also causes repression through mechanisms described above for type I and IV response types.
There are several possible explanations for the lack of detecting PPARβ/δ occupancy near the TSS of the different response type genes. It is possible that PPARβ/δ regulates these genes in regions further away than ± 10 kb from the TSS, or that regulation is mediated by another direct PPARβ/δ target gene that in turn directly regulates the target genes. This mechanism is likely for the type VII response gene because promoter occupancy of PPARβ/δ was not detected within 10 kb of the TSS. Alternatively, PPARβ/δ could occupy the regulatory regions of the response genes, but the ChIP may not be sensitive enough to effectively pull down chromatin with bound PPARβ/δ, which could be influenced by the relative ability of the antibody to bind with PPARβ/δ. The relative antibody binding to PPARβ/δ occupying chromatin could be impaired if: 1) conformational changes resulting from different co-effector molecules bound to PPARβ/δ are present, limiting access of the receptor to the antibody, 2) PPARβ/δ is indirectly bound to chromatin as part of a larger regulatory complex, 3) proteins are bound to PPARβ/δ as a result of the crosslinking step of the ChIP, and/or 4) the residence time of PPARβ/δ is too brief to register a signal in the ChIP-seq analysis because the receptor is rapidly exchanging with chromatin as observed with many other receptors [15–18].
PPARβ/δ target genes were recently identified from microarray analysis of human myofibroblast-like cells following either siRNA knockdown of PPARβ/δ or ligand activation of PPARβ/δ in these cells . Adhikary and colleagues identified 595 genes that were regulated by a PPARβ/δ ligand in human WPMY1 cells , whereas only 130 genes were specifically regulated by ligand activation of PPARβ/δ in mouse keratinocytes in the present study (Additional file 3: Table S4). Of the 130 genes that were regulated by ligand activation of PPARβ/δ in mouse keratinocytes, 24 (19%, Additional file 4: Table S5) were also regulated by ligand activation of PPARβ/δ in human WPMY1 cells . Adhikary and colleagues also identified 3704 genes that were regulated following knockdown of PPARβ/δ in human WPMY1 cells . In contrast, 482 genes were differentially regulated in the absence of PPARβ/δ expression in mouse keratinocytes in the present study. Of the 482 genes that were differentially regulated by PPARβ/δ in mouse keratinocytes as detected by comparing wild-type and Pparβ/δ-null cells, 79 (16%, Additional file 4: Table S5) were also regulated by disrupting expression of PPARβ/δ in human WPMY1 cells . Thus, of the 612 genes that were regulated by PPARβ/δ in mouse primary keratinocytes, 103 of these genes were also regulated by PPARβ/δ in WPMY1 cells (Additional file 4: Table S5). For approximately 50% of these genes, the response type exhibited was identical between mouse primary keratinocytes and WPMY1 cells (Additional file 4: Table S5). Many interchanges of response types were observed between the remaining genes but interchanges from type II to type I, type III to type I, and type I to type II were slightly more commonly noted in mouse primary keratinocytes as compared to WPMY1 cell (Additional file 4: Table S5). Collectively, these observations demonstrate that PPARβ/δ regulates some common sets of genes in human and mouse cells, but that there can also be differences in the molecular targets and the types of regulation observed. These differences might be due to the presence or absence of one or more endogenous ligand, differences in accessibility to regulatory regions of chromatin, species differences in the sequences of binding motifs, differences in the approach used to delete/knockdown PPARβ/δ (e.g. genetic versus siRNA), and/or possible species differences in the expression levels of the three PPARs between the two cell types.
The present studies identified eight different, PPARβ/δ-dependent response types of genes in mouse keratinocytes. In contrast, others characterized only three different PPARβ/δ-dependent response types in human WPMY1 cells . To facilitate comparisons with the present study, data from Adhikary  was re-examined and indeed, all eight different response types were evident in these data. In mouse keratinocytes, constitutive expression of 482 (79%) target genes was regulated by PPARβ/δ through type I and II responses. In human WPMY1 cells, 2728 genes (84%) were also regulated through either a type I or type II response . Type III and IV responses observed following activation of PPARβ/δ with an exogenous ligand were observed for 99 genes (16%) in mouse keratinocytes while 345 target genes (11%) were regulated by similar mechanisms in human WPMY1 cells . Lastly, types V-VIII responses were observed for 31 genes (11%) in mouse keratinocytes, whereas 186 genes (6%) were modulated similarly in WPMY1 cells . These data suggest that regulation of gene expression in mouse keratinocytes and human WPMY1 cells is likely mediated in large part by the one or more endogenous ligand, and that activation of PPARβ/δ with an exogenous ligand modulates expression of a relatively smaller set of genes as compared to those that are regulated by PPARβ/δ endogenously.
ChIP-seq analysis revealed new insight into the functional role of PPARβ/δ in the regulation of gene expression in keratinocytes. PPARβ/δ is constitutively enriched on chromatin on chromosomes 7, 9, 11 and 17, and in response to exogenous ligand activation, PPARβ/δ is enriched on chromatin located on chromosomes 2, 4, 7 and 11. These findings indicate an important role for PPARβ/δ in regulating genes encoded on these chromosomes. Interestingly, while ChIP-seq demonstrated that PPARβ/δ was present near ~6700 genes in the mouse genome, only 203 of these genes were regulated by PPARβ/δ when compared with microarray analysis. This is likely due in part to the presence of PPARβ/δ in intronic sequences that may or may not be functional. One mechanism that might explain the occupancy of PPARβ/δ on chromatin not associated with genes that were found to be regulated based on microarray analysis is that PPARβ/δ may require the presence of other transcription factors or signaling molecules in order to modulate gene expression. For example, the aryl hydrocarbon receptor (AHR) occupies the interleukin 6 (IL6) promoter but does not modulate expression of IL6 unless NF-kB becomes activated through IL1β-dependent signaling . This type of signaling paradigm has not been examined to date for PPARβ/δ, and results from the present studies suggest this possibility could exist and should be examined in future studies. This suggests that while 612 genes were regulated specifically by PPARβ/δ, many of these changes appear to be mediated by mechanisms that are secondary to effects induced by the direct target genes. Alternatively, it remains possible that there are more direct target genes, but the regulatory elements are more distal than ± 10 kb from TSS.
ChIP-seq analysis provided a unique opportunity to begin to examine the molecular mechanisms by which PPARβ/δ differentially regulates gene expression. A consensus PPRE was derived from these analyses and was comparable to PPREs identified for other PPARβ/δ target genes [13, 29]. No consistent differences in the PPRE sequences were identified that were able to distinguish between the different response types. However, the presence of binding site motifs for other transcription factors was observed that distinguished between effects observed with and without exogenous ligand. For example, the ETS binding sites were commonly present near the PPRE of genes that were modulated in the presence or absence of exogenous ligand, whereas the CREB/ATF/AP1 binding sites were commonly present near the PPRE of genes that were modulated only in the presence of exogenous ligand. Interestingly, different patterns of consensus binding sites of various transcription factors were also noted near the PPRE of genes that exhibited types I, II and III responses, but not for the other five response types. Of particular interest, is the novel finding that PPARβ/δ cooperates with ATF4 in modulating expression of some target genes. This suggests that PPARβ/δ requires cooperation with other transcription factors to specifically regulate subsets of genes. A similar phenomenon has also been found for other transcription factors [45–47]. While the physiological role of ATF4 in modulating PPARβ/δ-dependent gene expression and function requires further investigation, this type of interaction might begin to explain some of the complex regulation associated with the dynamic and fluid nature of nuclear receptor binding with chromatin.