There are numerous forkhead transcription factors encoded in the human genome, but we know little about how many of these function (reviewed in reference 18
). In this study, we identified the binding sites occupied by the forkhead family member FOXK2 and demonstrate that one important function for FOXK2 is in promoting signal-dependent gene expression changes through the AP-1 transcription factor complex.
The possible association with AP-1 transcription factors was revealed by the overrepresentation of the consensus AP-1 binding motif TGA(G/C)TCA within FOXK2 binding regions identified by ChIP-seq in U2OS cells. This association is further supported by the large overlaps with ChIP-seq data for the AP-1 components FOS and JUN, even though the latter experiments were conducted with a different cell type. The recruitment of FOS and JUN to FOXK2 binding regions in U2OS cells was subsequently validated, and FOXK2 is instrumental in this process (). Importantly, this association with AP-1 binding has functional consequences, as depletion of FOXK2 has a large effect on the AP-1-mediated gene expression program (A). It is currently unclear how FOXK2 promotes AP-1 recruitment, although there are several possible mechanisms that are not mutually exclusive. The observation that FOXK2 and JUN can interact suggests that cooperativity at the level of DNA binding is possible through additive protein-protein interactions. However, unlike other cooperatively acting transcription factor pairs like RUNX1-ETS1 (22
), no strict spacing between FOXK2 and AP-1 binding sites is apparent, which would impose stereospecific constraints on the interaction surfaces provided by each transcription factor. An alternative might be interaction following looping of intervening DNA. Another forkhead transcription factor, FOXA1, has been shown to function as a pioneer factor that facilitates access to chromatin, thereby permitting other transcription factors such as the estrogen and androgen receptors to bind to the exposed DNA (5
). It is possible that FOXK2 might also work in this manner through its forkhead DNA binding domain, and its significant association with open chromatin () hints at such a possibility.
At this point, it is not clear what the specificity determinants are for AP-1 transcription factor recruitment and why only a subset of sites are apparently specified by FOXK2. For example, there are numerous AP-1 complex subunits and other bZIP proteins that can potentially bind to the TGA(G/C)TCA motif. Furthermore, the AP-1 motif is also overrepresented in binding regions occupied by FOXA1 in MCF7 cells (30
), and close associations are seen between FOXL2 binding and the occurrence of AP-1 motifs (26
). Importantly, the AP-1 motifs that cooccur with FOXA1 binding events show only an 8.2% overlap with those cooccurring with FOXK2, which suggests potentially redundant and specific actions for these forkhead transcription factors in this context. In the redundant mode, it is likely that due to their widely differing overall sequences beyond the forkhead DNA binding domain, the key aspect of FOX function is to provide a suitable chromatin environment, rather than cooperativity at the level of transcriptional output. Other transcription factors have also been shown to cooperate with AP-1 activity, such as NFAT on a subset of promoters (21
), and more globally, coassociations have been observed with the T-cell factor (TCF)–β-catenin complex (3
). Thus, more-complex networks exist that converge on AP-1 transcription factors exist, and they likely contribute to subpartitioning the action of signaling through AP-1 and thereby elicit unique gene expression responses to different cellular environments.
In general, the number of binding sites identified for forkhead transcription factors is high, and this is also the case for FOXK2, where there are likely tens of thousands of binding regions. Importantly, most of the features we have identified using the high-confidence data set of 8,600 binding regions derived from the intersect of two experiments can also be found within the 59,531 binding regions found in the best of our two experiments, albeit usually with lower frequency. Interestingly, however, the frequency of cooccurrence of AP-1 binding motifs in these regions increased substantially (from 19% to 26%; see Fig. S3E in the supplemental material), further emphasizing the intimate relationship between FOXK2 and AP-1 binding. It is not clear why there are so many binding regions and whether they are all functionally significant. Furthermore, FOXK2 binding regions are usually located many kilobases away from the TSS of the nearest gene (A). This distal location might reflect the need to retain active chromatin regions away from the proximal promoters and hence permit accessibility by AP-1 transcription factors. This distribution is also similar to that seen with many other transcription factors, including other forkhead transcription factors (e.g., FOXA1, FOXA2, FOXA3) (30
) but differs from other transcription factors like GABPα (43
), which exhibits strong promoter-proximal binding. It is currently unclear what the underlying mechanistic differences are between transcription factors with broad genomic distributions compared to those which act in a more promoter-proximal manner.
It is possible that in different cell types, different forkhead transcription factors bind to the same sites depending on their relative abundance. Indeed, the fact that the core sequence GTAAACA is identified as a binding motif for many forkhead transcription factors hints at such a possibility. However, the overlap in binding regions occupied by FOXK2 and other forkhead proteins studied thus far is relatively low (E), suggesting that functional redundancy at the level of DNA binding is not particularly widespread. Further specificity might be embedded in additional combinatorial interactions or in regions flanking the core GTAAACA motif. In the case of FOXK2, the sequence WWWGTAAACAWG (7% of all such motifs in the genome) is more enriched than the core GTAAACA motif (0.8% of all such motifs in the genome) in its binding regions. Further studies are needed to further explore the basis to binding specificity generation.
The biological function of FOXK2 and mechanism of action are unclear. However, our study provides important advances in addressing these key issues. First, gene ontology studies suggest important roles of FOXK2 in controlling the expression of genes associated with cellular signaling pathways, transcriptional control, apoptosis, and cell movement (F; see Fig. S5 and Table S3 in the supplemental material). All these processes are responsive to a changing cellular environment, and hence to signal-dependent gene expression changes, such as those elicited by the AP-1 transcription factors (reviewed in reference 20
). Indeed, FoxK has been shown to be important in TGFβ signaling in flies (6
), which is consistent with our observation that TGFβ signaling appears as one of the most enriched GO terms for genes associated with FOXK2 binding regions. However, in flies, transforming growth factor β (TGFβ) signaling acts upstream, rather than downstream. Importantly though, genetic studies indicate that FoxK in flies works in a combinatorial manner with Dfos/AP-1, suggesting that the association of FOXK transcription factors with AP-1 signaling is evolutionarily conserved. Thus, FOXK2 appears to have a broad role in potentially controlling the expression of gene expression programs which underpin some of the central decision points that a cell makes. It is unclear whether FOXK2 is an activator or a repressor protein as previously suggested (33
). The latter would be consistent with the observation made on the closely related mouse homologue of FOXK1, as current evidence suggests that this is also a repressor (45
). Our overexpression studies are supportive of a repressive role for FOXK2, as no significant increases in gene expression were observed upon FOXK2 expression (B). However, our siRNA depletion studies suggest a more bivalent role for FOXK2 as a repressor and activator depending on the gene target and signaling conditions, as significant changes in the up- and downregulation of gene expression were observed upon FOXK2 loss (A and A). In this scenario, one likely role for FOXK2 is to maintain the chromatin in an open state to facilitate either transcriptional repressive or activating events to occur.
In summary, our study provides novel insight into FOXK2 function and its association with gene regulatory events on a genome-wide scale. Little was previously known about this transcription factor, but its functional association with AP-1-mediated gene regulatory events demonstrates an important role in coordinating signal-dependent gene expression changes.