|Home | About | Journals | Submit | Contact Us | Français|
The p53-family of proteins regulates expression of target genes that promote cell cycle arrest and apoptosis, which may be linked to cellular growth control as well as tumor suppression. Within the p53-family, p53 and TA-p73 have hepatic-specific functions in development and tumor suppression. Here, we determined TA-p73-interactions with chromatin in adult mouse liver and found Forkhead factor Foxo3 as one of 158 gene targets. Global profiling of hepatic gene expression in regenerating versus quiescent liver revealed specific, functional categories of genes regulated over time of regeneration. Foxo3 is the most responsive gene among transcription factors with altered expression during regenerative, cellular proliferation. p53 and TA-p73 bind a Foxo3 p53-response element and maintain active expression in quiescent liver. During regeneration of liver, binding of p53 and TA-p73, recruitment of acetyltransferase p300 and an active chromatin structure of Foxo3 are disrupted, alongside loss of Foxo3 expression. Consistent with the loss of Foxo3 transcriptional activation, a decrease of histone activation marks (H3K4me2, H3K14Ac and H4Ac) at Foxo3 p53RE was detected following partial hepatectomy in mice. These parameters of Foxo3 regulation are reestablished at completion of liver growth and regeneration, supporting a temporary suspension of p53 and TA-p73 regulatory functions in normal cells during tissue regeneration. p53 and TA-p73-dependent activation of Foxo3 is also observed in mouse embryonic fibroblasts and in mouse hepatoma cells overexpressing p53, TA-p73α and TA-p73β isoforms.
p53 and p73 directly bind and activate expression of Foxo3 gene in adult mouse liver and murine cell lines. p53, TA-p73, and p300 binding and Foxo3 expression decrease during liver regeneration, suggesting a critical growth control mechanism mediated by these transcription factors in vivo.
Tumor suppressors p53 and p73 are members of a family of proteins with both unique and shared primary functions as transcription factors in mammalian cells (1). The p53+/−; p73+/− mice develop hepatocellular carcinoma (HCC) at 5-7 months of age, suggesting a pivotal and cooperative role for p53 and p73 in regulation of hepatic gene expression (2). Approximately 90% of p53+/−;p73+/− mice with HCC have loss of heterozygocity in Trp73, further emphasizing the importance of tissue-specific functions of p73 in liver (2).
Studies of cancer cell lines, mouse models, and patient samples clearly establish that loss of p53 and p73 functions is causative in tumor development (2); however, much less is known about the status and functions of p53 and p73 in normal, quiescent tissues in the absence of cellular stress. Our previous studies show that p53 protein levels are developmentally regulated in mouse liver, as p53 is undetectable in newborn mice but increases within two weeks and maintained throughout adulthood (3). Both p53 and p73 bind to the p53 response element (p53RE) of Afp in liver, targeting co-repressor proteins and repressive histone modifications to chromatin at the p53RE and Afp transcription start site (TSS), and repressing Afp within 2-3 weeks of age (4, 5).
Tumor suppressors p53 and TA-p73 regulate cell death, cell cycle, and senescence through transcriptional activation or repression of target genes (6). These processes are highly regulated during regeneration of liver when mature, quiescent hepatocytes re-enter cell cycle, proliferate and grow, in efforts to reestablish liver mass after surgical or chemical removal of liver tissue (7). To uncover potential liver-specific targets of p53 family members, we combined chromatin immunoprecipitation (ChIP) of TA-p73 in adult mouse liver and hybridization to microarrays (ChIP/chip), with determinations of global expression during liver regeneration in response to partial hepatectomy (PH). We found a highly restricted number of TA-p73-target genes that changed expression during liver regeneration, and identified the Forkhead box transcription factor Foxo3 as a new target gene of p53 and TA-p73 in normal quiescent liver.
FoxO3 is a bona fide tumor suppressor that regulates expression of genes that inhibit cell cycle and activate apoptosis (8). Ectopic expression of transcription factor E2F1 was shown to activate Foxo3-driven reporter expression, but direct regulation of endogenous Foxo3 transcription has not been demonstrated (9). Interestingly, FoxO3 protein can directly bind to promoters of other FoxO family members and activate expression of FOXO1 and FOXO4; however, despite high homology between FOXO genes, FoxO3 protein fails to bind and activate FOXO3 (10). Thus, the mechanisms of transcriptional regulation of FOXO3 remain to be identified.
Our work shows that p53 and TA-p73 bind to the p53RE of the endogenous Foxo3 gene in adult mouse liver and recruit acetyltransferase p300 to activate chromatin structure and expression of Foxo3. In response to PH, binding of p53, TA-p73, and p300 to the Foxo3 p53RE is lost and Foxo3 expression is decreased, during the proliferative stage of liver regeneration, and restoration occurs with recovery of liver mass. Our findings establish a direct regulatory link between p53, TA-p73 and FoxO transcription factors, which are growth suppressors in normal tissues: an axis of homeostasis in hepatic cells that is temporarily disrupted during regeneration of liver.
PH (70% removal of total liver) or Sham control surgery was performed using isoflurane anesthesia (Cold Spring Harb. Protoc.; 2006; doi:10.1101/pdb.prot4384). 5-7 C57Bl6/Sv129 mice, 2 months of age, were used for each experimental condition according to the MD Anderson Cancer Center Institutional Animal Care and Use Committee guidelines. Mice were sacrificed 1, 2, 3, 4, or 7 days following surgeries; remnant liver tissue was harvested, flash-frozen, and processed for RNA and ChIP analyses.
Hepa1-6 cells were obtained from ATCC and cultured under standard, suggested conditions. The plasmid encoding wild type HA-tagged p53 was previously described (11). Plasmids encoding HA-TA-p73α and HA-TA-p73β were gifts of G. Melino (4). Cells were transfected with 4 μg of total DNA/100-mm plate using Lipofectamine (Invitrogen). Val5 mouse embryonic fibroblasts (MEFs) stably expressing temperature-sensitive p53 R135V mutant were kindly provided by M. Murphy (12).
Chromatin immunoprecipitation (ChIP) was performed on liver tissue lysate from 2-month-old C57Bl6/Sv129 mice with TA-p73 antibody (4). TA-p73 antibody-enriched DNA and Input genomic DNA were differentially labeled and hybridized to Agilent promoter array representing 1-2 60-mer oligomeric probes within -5.5kb and +2.5kb of 17,000 genes or predicted gene regulatory regions of the mouse genome. Ligation-mediated amplification was used prior to labeling and hybridization; amplified material was shown to have TA-p73-interaction sites by analysis for Afp and Cdkn1a binding.
Liver tissue was collected at T=0, 0.5, 1, 2, 4, 24, 38, and 48 hours post-surgery. 5 μg of total RNA, from each individual animal, were analyzed by Affymetrix MGU74 (T = 0.5, 1, 2, 4 hours) and Affymetrix 430.2 (T = 24, 38, 48 hours) gene arrays, 12,488 and 45,000 probe sets respectively, for mouse genes and expressed sequence tags. Variation between arrays using the same RNA sample in a quality control check revealed less than 2% difference. Data quality control was done with Affymetrix Microarray Suite (MAS) 5.0 and normalized by Robust Multichip Analysis (RMA) software. Genes with a negative or positive fold change of 1.5 times or more between T=0 and indicated post-PH time points, with a significance cut-off of p-value < .005 (MGU74) or < 0.0001 (430.2), were further analyzed. These microarray data are available on the GEO database (accession number GSE20427) at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE20427. ChIP/chip and microarray data sets were annotated and analyzed by Ingenuity Pathway Analysis (IPA, Ingenuity® Systems, www.ingenuity.com), DAVID Bioinformatics Resources (13) and Panther Classification System (14).
Chromatin lysate was precleared and incubated overnight with the following specific antibodies for ChIP: histone H3 (Abcam), H3K4me2 (Active Motif), H3K9Ac (Active Motif), H3K14Ac (Upstate/Millipore), H4Ac (Upstate/Millipore), p53 (Novocastra), TA-p73 (Santa Cruz), p300 (Santa Cruz), and normal sheep IgG (Upstate/Millipore). To analyze antibody-protein-bound DNA, primers for real-time quantitative PCR (qPCR), and TaqMan (Applied Biosystems, ABI, Foster City, CA) were used (Supplementary Table 5). The percentage of input DNA bound was calculated by dCt methodology and averaged over at least three experiments.
Primers and reverse transcription-PCR determinations of RNA expression were as previously described (4). Briefly, total RNA from homogenized mouse liver was extracted with TRIzol reagent (Invitrogen) according to the manufacturer's instructions. cDNA was obtained by reverse transcription of 2 ug of RNA using the SuperScript system (Invitrogen). Real-time PCR primers for indicated genes are in Supplementary Table 5.
Immunoblotting of whole-cell and nuclear lysates was performed using standard SDS-PAGE methodology. Nuclear extracts from liver tissue, collected at 2 days following Sham and PH surgery, were prepared as described (5). The primary antibodies used are: p53 (Santa Cruz), TA-p73 (Santa Cruz), HA-tag (Cell Signaling), Foxo3 (Cell Signaling), and β-actin (Santa Cruz).
Results are expressed as means standard error of the mean (SEM). Statistical analyses were performed by t-test; p values <0.05 were considered significant.
We performed ChIP of liver tissue from 2-month-old wt mice, with antibody against TA-p73, and isolated TA-p73-bound chromatin fragments for ChIP/chip. We uncovered 158 genes, among the arrayed promoters, as potential targets of TA-p73-binding activity in liver (Supplementary Table 1). In a separate set of experiments, we determined genes that change expression after PH. Functional annotations of genes among the 158 TA-p73 target genes (Figure 1A) and those that changed expression in response to PH (Figure 1B and 1C) were performed using IPA. The analysis yielded 12 categories of function, with cancer, cell death, and cell proliferation receiving the highest number of hits with the lowest p-value, as determined by IPA (Figure 1). As expected, TA-p73-bound genes in the quiescent liver also included development, inflammatory response, cell signaling, and cell cycle categories (Figure 1A) (15, 16). Gene ontology analysis using DAVID and Panther bioinformatics tools yielded similar categories, including additional ones, such as sensory perception and transcription (Supplementary Figure 1A and 2A). Of these gene targets, none were previously reported as p73-regulated and 8 are known targets of p53 (e.g, Annexin A1 (17), Notch1 (18), and Bai1 (19), Supplementary Table 1). Thus, many of the TA-p73 target genes identified by ChIP/chip are not known as p53/p73-regulated and potentially define a specific set of TA-p73-regulated genes in quiescent liver.
Two-thirds PH promotes proliferation of liver cells and rapid growth of the remaining liver tissue, resulting in complete restoration of organ mass in approximately 7 days following PH in mice. We collected liver tissue from mice at T=0, 0.5, 1, 2, 4, 24, 38, and 48 hours following PH for microarray analysis of gene expression. We found 434 genes that changed expression between 0.5 – 4 h after surgery (Supplementary Table 2), and 3807 genes that changed expression between 24 - 48 h following PH, compared to T=0 (Supplementary Table 3).
Consistent with previous observations utilizing more limited expression analyses (20, 21), our microarray analysis of 0.5-4 h post-PH livers showed a significant increase in expression of ‘early response’ genes, e.g., C/EBPb, Jun, Myc, Tnfrsf1a, Hif1a, Atf3, Ets2 (Supplementary Table 2). Several responsive genes, not previously reported as regulated in the context of liver regeneration, included pluripotency regulator Klf4, transcription factors Mxi1 and Sin3a, as well as anti-apoptotic Bcl2 family member Bcl2l1 (Supplementary Table 2). Functional annotations of genes that changed expression in response to PH at 0.5-4 h included categories also identified for TA-p73-bound genes (Figure 1B, Supplementary Figure 1B and 2B).
Similar to TA-p73-bound genes in quiescent liver, genes that changed expression (either increasing or decreasing) during 24-48 hours of liver regeneration were associated with cancer, cell death, and cell proliferation (Figure 1C). Cell signaling and inflammatory response, represented among TA-p73-bound genes, had less than 1% hits among genes that changed expression at 24-48 hours following PH, suggesting unique functions for TA-p73 in liver that are not executed during 24-48 hours of regeneration.
Compared to earlier time points, genes that changed expression in regenerating liver at 24-48 hours post-PH had a significant increase in targets within cell cycle and DNA replication categories (Figure 1C, Supplementary Figure 1C and 2C). Direct comparison of gene IDs, from microarray analysis of genes with altered expression 24-48 hours post-PH to TA-p73-bound genes, yielded seventeen TA-p73-bound genes up- or down-regulated in response to PH. This list included a group of transcription factors, Dmtf1, Foxo3, and Nfatc3, as well as cell cycle regulators and plasma membrane receptors (Supplementary Table 4).
We hypothesized that the select group of TA-p73-bound genes, which displays altered expression during liver regeneration, may offer further clues regarding liver-specific gene targets of both p53 and TA-p73. TA-p73 can bind to the same consensus site, simultaneously with p53, at the p53RE of hepatic gene Afp (4). We analyzed TA-p73-bound genes, uncovered by ChIP/chip, as potential p53-regulated gene-targets in normal, quiescent liver. Using a published algorithm to find p53 consensus sites (22), we mapped potential, shared p53 and TA-p73 (p53/p73) binding sites upstream of four TA-p73-bound genes that changed expression during 24-48 h of LR: Foxo3, Jak1, Pea15, and Tuba1 (Supplementary Table 4 and Supplementary Figure 3). Binding of p53 and TA-p73 was observed for all examined genes at identified p53REs, thus confirming that putative targets uncovered by TA-p73 ChIP/chip may be bound by both p53 and TA-p73 in quiescent liver in vivo (Figure 2). Afp p53RE served as a positive control for p53/p73 binding in quiescent liver, whereas upstream regions of Alb and Brn3B genes served as negative controls for p53 and TA-p73 binding (4, 23). Taken together, these results suggest that p53 and TA-p73 activate or repress target genes in quiescent liver, and that regulatory activities of p53 and TA-p73 change during liver regeneration.
Among the seventeen TA-p73 gene-targets revealed by ChIP/chip, Foxo3 had the most significant change in expression in response to PH and strong p73-binding (Supplementary Table 4). We found a p53 consensus site at -3.7 kb upstream of the TSS of Foxo3, as well as several other potential p53 binding sites within the second and third introns (Figure 3A). We detected binding of both p53 and TA-p73 to the p53RE at -3.7 kb upstream of Foxo3 (Figure 3B). To confirm the specificity of p53/p73 binding to the Foxo3 p53RE, we used primers for a region that contains no p53REs, located at -2.0 kb upstream of Foxo3 TSS, and saw background levels of interaction (non-specific region, n.s., Figure 3A and 3B).
TA-p73 compensates for loss of p53, by binding to the Afp p53RE in the absence of p53 (4), and promotes a delayed but significant reduction of Afp expression in liver by 4-months of age in p53−/− mice (23). We performed ChIP from liver tissue collected from p53−/− mice at 2 months of age and found that TA-p73 binds the p53RE of Foxo3 in the absence of p53 (Figure 3C). Thus, both p53 and TA-p73 regulate transcription of Foxo3 in adult mouse liver at T=0.
Based on known functions of FoxO3 as a tumor suppressor, we hypothesized that p53 and TA-p73 act as positive regulators of Foxo3 at the level of transcription. We determined levels of Foxo3 mRNA, isolated from liver tissue collected from p53+/−, p53−/−, and p73+/− mice, in comparison to wt littermates, and observed a significant decrease of Foxo3 expression in p53−/− and p73+/− mice (Figure 4A).
Transcription of Trp73 from multiple promoters, together with alternative mRNA splicing, results in at least 28 isoforms of p73 (24). We performed transient transfection of a mouse hepatoma-derived cell line Hepa1-6 (25), with plasmids that express “transactivating” TA-p73 isoforms, HA-TA-p73α and HA-TA-p73β or HA-p53. A significant increase in endogenous Foxo3 mRNA levels is observed in cells expressing p53 or TA-p73 (Figure 4B). Immunoblotting with antibodies against p53 and all TA-p73 isoforms (Figure 4B, lower panel) shows that HA-TA-p73β is expressed at a lower level than HA-TA-p73α, but induction of endogenous Foxo3 expression is comparable (Figure 4A). This is consistent with increased transcriptional activity previously reported for TA-p73β, which lacks a previously identified, repressive SAM domain, versus other TA-p73 isoforms (26-28).
To establish cause-and-effect in direct transcriptional regulation of Foxo3 by p53, we used immortalized MEFs that express a temperature-sensitive p53 conformational mutant: p53val135 (Val5MEFs, Figure 4C) (12). In this model system, Val5MEFs incubated at a restrictive temperature (37°C) have only cytoplasmic-localized p53, p53val135, which is unable to regulate target gene expression. At the permissive temperature of 32°C, p53val135 assumes a wt conformation and moves to the nucleus to activate or repress its target genes, including endogenous Foxo3 (Figure 4C). Together, these results demonstrate that endogenous Foxo3 is activated by p53 and TA-p73 in mouse liver, as well as by nuclear translocation of p53 or ectopic expression of p53 or TA-p73.
Our analysis of global gene expression levels (Supplementary Table 2 and 3) suggested that Foxo3 expression is decreased at 24-48 hours following PH. We determined if loss of Foxo3 expression in regenerating liver, occurs as a result of decreased p53/p73 binding to chromatin at the p53RE of Foxo3. We performed ChIP analysis of liver tissue, collected at 1, 2, 4 and 7 days after PH and Sham surgeries, with antibodies that recognize p53 and TA-p73. Chromatin interaction of p53 at the Foxo3 p53RE was dramatically reduced at day 1 and 2 after PH, accompanied by an equally significant reduction in TA-p73 binding (Figure 5A). Binding of both p53 and TA-p73 is partially restored at days 4 and 7 of liver regeneration (Figure 5A), but is not equivalent to the level of binding observed in Sham (Figure 5B), suggesting that regulatory mechanisms in addition to those mediated by p53 and TA-p73 may activate Foxo3.
Microarray analysis of early time points (0.5 – 4 h) showed no significant change in Foxo3 expression (Supplementary Table 2), compared to T=0; a significant decrease in Foxo3 expression was observed in liver collected at 24, 38, and 48 hours following PH (Supplementary Table 3). This result suggests that loss of Foxo3 expression occurs specifically during cell cycle G1-S-G2 transition. We performed sets of PH and Sham surgeries on 2-month-old wt mice, collecting liver at 1, 2, 3, 4, and 7 days. We observed a significant decrease in Foxo3 mRNA levels between 1-3 days after PH, compared to T=0, with lowest Foxo3 expression on day 2 (Figure 6A). FoxO3 protein levels were also reduced in hepatic nuclei on day 2 following PH (Supplementary Figure 4), suggesting that transcriptional functions of FoxO3 are inhibited during the proliferative stage of regeneration. Importantly, several known target genes of FoxO3 changed expression during day 1 and 2 of liver regeneration (Bcl6, Ccnd1, Ccng2, Sirt1, and Sod2, Supplementary Table 3) (29). Foxo3 expression gradually returned to T=0 level on day 4 following PH (Figure 6A), when final adjustments of regenerating liver tissue restores a normal liver/body weight index by 7 days in mice (data not shown). No significant difference in Foxo3 expression is observed following Sham surgeries, compared to T=0 (Figure 6B).
Gene expression is associated with active chromatin structure, e.g. dimethylation of histone H3 at lysine 4 (H3K4me2) and acetylation of H3K9, H3K14, and several lysines of histone H4. Using ChIP analysis of liver tissue collected 1 day after PH and Sham surgeries, we tested whether histone modifications, associated with active chromatin, decrease with loss of p53 and TA-p73 binding at the Foxo3 p53RE region. We observed decreased H3K4me2 and H3K14Ac (Figure 7A) and a decrease of global H4 acetylation (Supplementary Figure 5) without a significant change in H3K9 acetylation (data not shown), at the Foxo3 p53RE, in regenerating liver on day 1, compared to day 4 and Sham. While several methylatransferases and acetyltransferases modify H3K4, H3K14, and H4 residues, p53 and p73 are known to recruit CBP/p300 (KAT3A/KAT3B) to activate transcription of target genes (30, 31). Using an antibody against p300, we observed a significant decrease in p300 binding to the Foxo3 p53RE region of chromatin, in quiescent liver of p53−/− mice (Figure 7B). Binding of p300 to the Foxo3 p53RE decreased even further in regenerating liver of wt mice at 1 day post-PH (Figure 7C), suggesting that both p53 and p73 contribute to the recruitment of p300 to activate expression of Foxo3 in quiescent liver. Recruitment of p300 was restored at day 4, following PH, alongside p53 and p73.
Genome-wide evaluation of p53- and p73-bound sequences by ChIP/chip analysis, using cultured cells, showed that 72% of p53-bound sites are also bound by p73 (32). In the current study, we uncovered 158 genes bound by TA-p73 in normal quiescent liver cells. Ten genes were known p53-targets with the remainder not previously connected to p53/p73-mediated transcriptional regulation. In a recent review, Riley et al. identified stringent criteria for bona fide p53 target genes, and generated a list of 129 genes containing at least one p53RE per gene (33). In our ChIP/chip experiments, we identified a similar number of TA-p73 target genes, and confirmed binding of both p53 and TA-p73 to the p53REs of four target genes, Foxo3, Jak1, Pea15, and Tuba1a. Annotation of the 158 hepatic target genes yielded physiological categories of established p53/p73 functions, e.g., cancer, cell death, and cell proliferation. Other TA-p73 bound genes, associated with inflammatory response and development, are supported by the phenotype of p73−/− mice, which display severe developmental and inflammatory defects and die soon after birth (15).
Our work identifies the Foxo3 gene as a new target of p53 and TA-p73 in quiescent mouse liver. Loss of FoxO3-mediated regulation of transcription is directly linked to increased proliferation and tumorigenesis in human cancers (8, 34); however, mechanisms that control Foxo3 itself remain poorly defined. We suggest that p53, p73, and FoxO3 function along the same axis in a tumor suppressor network, since many target genes of FoxO3 are also known p53/p73 targets, e.g., Cdkn1a (p21), Cdkn1b (p27), and Gadd45a. In turn, functions of FoxO3, as a transcription factor, augment p53 pro-apoptotic activity (35). Previous studies suggest that p53 and FOXO3 proteins interact as a protein complex (36). Whether feed-forward regulation of anti-proliferative genes by a p53-FoxO3 complex in quiescent liver is disrupted during regeneration, due to decreased levels of FoxO3 protein and loss of p53-target gene interactions, requires further study. Overall, our results establish a direct, linear pathway between p53-family members and FoxO tumor suppressors in normal tissues.
Despite many studies of p53/p73 target genes, mechanisms of endogenous p53 and TA-p73-mediated transcriptional regulation are understood for a very limited number of target genes in vivo. In quiescent liver cells, we found p53-dependent recruitment of histone acetyltransferase p300 at the Foxo3 p53RE, as a mechanism of histone modification and gene activation in vivo and in agreement with direct interaction of p53/p73 with p300 observed in cultured cells (30, 31). Binding of p300 at the Foxo3 p53RE is decreased but not lost in liver of p53−/− mice, suggesting that endogenous p73, independently of p53, relies on similar co-regulators of transcription and partially compensates for loss of p53. Recruitment of p300 and activation of Foxo3 expression significantly decrease during day 1 of liver regeneration, when both p53 and p73 show loss of binding to the Foxo3 p53RE, providing additional evidence for p53/p73-mediated recruitment of p300 as a critical mechanism to activate expression of the Foxo3 gene in quiescent liver. Thus, we conclude that both p53 and TA-p73 lose the ability to bind and regulate expression of specific hepatic genes during G0-G1-S transition.
Importantly, p53/p73 binding to the Foxo3 p53RE is restored when liver regeneration is complete, leading to activation of Foxo3 expression as hepatocytes re-enter G0. Similarly, p53 binding to Afp is restored at 7 days following PH (5), but transcriptional activity of p53 is not required for termination of liver regeneration (37). TA-p73 compensates for lack of p53, in repression of Afp during liver development (4), and similar mechanisms may be in play during regeneration of hepatocytes in p53−/− mice. This hypothesis cannot be tested in p73−/− mice, as these mice succumb to developmental and inflammatory defects soon after birth (15). As a bona fide tumor suppressor, FoxO3 is aligned with p53 and p73 in regulating transcription in normal tissues. Their functions in surveillance of normal cells are temporarily disrupted during liver regeneration; mechanisms that restore their regulatory functions are of considerable interest for future studies.
We are grateful to members of our laboratories for helpful discussions, and to Jyothi Paniyadi, Customer Support Scientist, Ingenuity Systems, Inc. for her guidance in using IPA. We also thank Scott Ochsner for the GEO deposition of the microarray data.
Financial support: This work was supported by funds from the National Institutes of Health DK078024 to MCB, AG028865 to GJD, and in part by the NCI Cancer Center Support Grant to the University of Texas M.D. Anderson Cancer Center, and P30 DK056338 to the Functional Genomics Core of the Texas Medical Center Digestive Disease Center. SK received support from the William Randolph Hearst Foundation and a Sylvan Rodriguez Scholarship of the Cancer Answers Foundation.