Functional diversification of paralogous transcription factors can arise either by mutation in cis
-regulatory elements or changes in the coding sequence of the proteins 
. We examined the acquisition of novel biochemical functions by Foxa1 and Foxa2, the closest paralogs in the Foxa subfamily of winged-helix transcription factors, in the adult liver. Genome-wide location analysis (ChIP-Seq) revealed that Foxa1 and Foxa2 have unique targets in addition to many common ones, indicative of diverged function, which is also reflected in the divergent effects on the liver transcriptome by ablation of either factor. These results are consistent with an early study by Lai and colleagues who found that Foxa1 and Foxa2 (Hnf-3α and Hnf-3β) have different affinities for the two binding sites in the promoter of the TTR gene in vitro
. Differences in DNA binding by the Foxa paralogs could be the result of a handful of divergent residues in the forkhead
domain itself, and a few amino acid residues located outside the DNA binding domain that are targeted by post-translational modifications. Among seven amino acids in the winged-helix DNA binding domain that differ between Foxa1 and Foxa2, five are conserved between Foxa1 and the remaining Foxa family member Foxa3 (Figure S1
), indicating that Foxa1 and Foxa3 represent the ancient precursor gene, while Foxa2 has acquired new mutations at those positions.
Recently, Kohler and Cirillo have reported that acetylation of Foxa1 by p300 attenuates binding of Foxa1 to DNA 
. Multiple putative acetylation sites, identified by in silico
analysis, are divergent between Foxa1 and Foxa2 and likely contribute to their specific DNA binding properties. Additionally, a recent study implicated Foxa2 as a substrate for DNA-dependent protein kinase (DNA-PK), which targets serine 283 
. A mutation of that residue to alanine resulted in a protein with greater affinity for sequence-specific DNA-binding. Interestingly, Foxa1 has an alanine rather than a serine at position 283, evidence for another functional diversification of the two proteins. Cirillo and colleagues showed that the C-terminal domain of Foxa1 also enhances DNA-binding of the protein to the albumin enhancer 
. While Foxa1 and Foxa2 share strong sequence conservation in their transactivation domains within the C-terminus, the remainder of the C-terminal domain is quite divergent between the two paralogs.
Binding of Foxa2 to its targets in the adult liver has been studied previously 
. We reported that Foxa2 is required for normal bile acid homeostasis and a cluster of categories with genes involved in lipid and steroid metabolism was identified as bound by Foxa2 in vivo
. The previously reported data sets 
were comprised of all sites, including those bound by Foxa1 as well. We also showed that deletion of Foxa2 in hepatocytes affects expression of hundreds of genes in mice fed a standard diet and thousands of genes in mice on a cholic acid-enriched diet, demonstrating that Foxa1 cannot compensate for the loss of its paralog 
. Here, we found that Foxa2-only sites are also associated with genes important to lipid metabolism and contain a medium-strength forkhead
consensus, as well as motifs for liver-enriched transcription factors and nuclear receptors, and AT-rich motifs, including homeodomain transcription factors, Nfil3, and Hmga1. Hence, specialization of Foxa paralogs in the adult liver has resulted in Foxa2 acquiring a specific role in coordinating the transcriptional regulatory network that controls bile acid and lipid metabolism.
The function of Foxa1 has been studied primarily in a variety of cancer cell lines 
and, together with Foxa2, during embryonic development 
. Binding of Foxa1 to its targets was shown to be required for chromatin-association of androgen receptor (AR) in prostate cancer cells 
, and the estrogen receptor (ER) and retinoic acid receptor (RAR), in breast cancer cell lines 
. Foxa1 was also implicated in cell cycle regulation in tumor-derived cells 
, but a mechanistic model was not established. We found that Foxa1-only sites are enriched for binding sites of p53, a tumor suppressor that activates target genes that induce cell cycle arrest, apoptosis, or senescence. A single such composite Foxa1/p53 site was previously characterized in the promoter of the alpha-fetoprotein (Afp
) gene 
, which is expressed during development and repressed in the adult hepatocyte. We found that a p53 motif is prevalent in numerous Foxa1-only targets in the adult liver and validated that p53 binds these sequences. In addition, we detected other motifs corresponding to factors that interact with p53 (Klf12 (repressor of Tfap2a), Tfap2a, and Smads) or compete with p53 for binding (Hic1) 
. This is a novel function of Foxa1, which has implications for the role of Foxa1 in cell cycle progression and cancer.
A study by Gao and Matusik (personal communication) to detect potential DNA binding complexes occupying the TS2 regulatory element of the probasin gene, bound by Foxa1 in the prostate 
, identified poly (ADP-ribose) polymerase (Parp1) as a protein also interacting with this region. Parp1, a chromatin-associated enzyme, is involved in regulation of numerous processes, including proliferation, recovery from DNA damage, and tumor transformation. Parp1 modulates stability of p53 in unstressed cells 
and interacts with androgen receptor (AR) 
, retinoic acid receptor (RAR) 
, proteins functionally linked with Foxa1, and may interact with Foxa1 itself. These data support a regulatory network unique to the Foxa1 paralog, one that functions in cellular growth and genome stability.
In summary, the transcriptional regulators Foxa1 and Foxa2 share a significant fraction of cis-regulatory elements that contain a high-affinity forkhead binding site and regulate genes essential in development and those implicated in etiology of diabetes. It is possible that gene regulatory networks of important disease susceptibility genes have redundant modules and closely related paralogs that compensate for each other, as occurs in developmental regulation. However, while Foxa1 retains the more ancient role of regulating proliferation and growth by influencing DNA binding of p53, Foxa2 has acquired mutations in its DNA binding domain and a new role in the hepatocyte, regulating genes involved in lipid metabolism. In this instance, it is more advantageous for duplicated paralogs to perform different functions and transduce the many different physiological signals propagated through differentiated tissues. We propose that this functional diversification of the Foxa paralogs contributed to the maintenance of both genes during evolution.