cDNA Microarray Gene Expression Profiling of Hedgehog Signaling Pathway Inhibition in Human Colon Cancer Cells

Background Hedgehog (HH) signaling plays a critical role in normal cellular processes, in normal mammalian gastrointestinal development and differentiation, and in oncogenesis and maintenance of the malignant phenotype in a variety of human cancers. Increasing evidence further implicates the involvement of HH signaling in oncogenesis and metastatic behavior of colon cancers. However, genomic approaches to elucidate the role of HH signaling in cancers in general are lacking, and data derived on HH signaling in colon cancer is extremely limited. Methodology/Principal Findings To identify unique downstream targets of the GLI genes, the transcriptional regulators of HH signaling, in the context of colon carcinoma, we employed a small molecule inhibitor of both GLI1 and GLI2, GANT61, in two human colon cancer cell lines, HT29 and GC3/c1. Cell cycle analysis demonstrated accumulation of GANT61-treated cells at the G1/S boundary. cDNA microarray gene expression profiling of 18,401 genes identified Differentially Expressed Genes (DEGs) both common and unique to HT29 and GC3/c1. Analyses using GenomeStudio (statistics), Matlab (heat map), Ingenuity (canonical pathway analysis), or by qRT-PCR, identified p21Cip1 (CDKN1A) and p15Ink4b (CDKN2B), which play a role in the G1/S checkpoint, as up-regulated genes at the G1/S boundary. Genes that determine further cell cycle progression at G1/S including E2F2, CYCLIN E2 (CCNE2), CDC25A and CDK2, and genes that regulate passage of cells through G2/M (CYCLIN A2 [CCNA2], CDC25C, CYCLIN B2 [CCNB2], CDC20 and CDC2 [CDK1], were down-regulated. In addition, novel genes involved in stress response, DNA damage response, DNA replication and DNA repair were identified following inhibition of HH signaling. Conclusions/Significance This study identifies genes that are involved in HH-dependent cellular proliferation in colon cancer cells, and following its inhibition, genes that regulate cell cycle progression and events downstream of the G1/S boundary.


Introduction
Hedgehog (HH) signaling plays a critical role in a variety of normal cellular processes. It is pivotal in embryogenesis, regulation of the epithelial-to-mesenchymal transition, the patterning of a diverse range of vertebrate structures in a variety of organs, maintenance of adult tissue homeostasis, tissue repair, cellular proliferation, and in cell survival [1,2,3,4,5,6,7,8,9]. The canonical HH pathway is also critical to normal mammalian gastrointestinal development, where it is involved in the coordinate regulation of differentiation of normal intestinal villi [10,11,12]. Thus, in the normal gastrointestinal tract, HH ligands are induced in the differentiated cells around the villous surface, generating a negative feedback loop to inhibit canonical WNT signaling in the basal cells of the crypt, thereby protecting differentiated cells from the proliferative effects of WNT [13]. Activation of the canonical HH signaling pathway comprises the binding of HH ligands to the membrane receptor Patched (PTCH1), which becomes internalized leading to the activation of the signaling molecule Smoothened (SMO) via release from PTC-mediated suppression. SMO activates the final arbiter of HH signaling, the GLI family of transcription factors that bind to the GAC-CACCCA-like consensus binding element in promoter sequences to transcriptionally regulate HH target genes [3,14,15]. GLI1 and GLI2, the transcriptional activators of HH signaling, possess distinct as well as overlapping functions that involve activator (GLI1 and GLI2) or repressor (GLI2) activities [16]; however, their roles in the regulation of HH-driven cellular proliferation, survival or cell death processes are poorly understood. Historically, GLI1 has been considered the most reliable marker of HH pathway activity, however GLI2 appears to be the primary activator of HH signaling, with GLI1 as a transcriptional target of GLI2 [3,7], leading to augmentation of HH signaling both quantitatively as well as qualitatively [16]. An important feature of GLI proteins is that their biological activity is context-dependent, influenced by the cellular environment [17,18].
Activation of the canonical HH signaling cascade is aberrantly activated and well known to play a critical role in oncogenesis and maintenance of the malignant phenotype in several types of human cancers. Such activation involves amplification of GLI1 or GLI2, mutations in PTC or SMO, or dysregulated gene expression [3,4]; these malignant cells are also sensitive to the small molecule inhibitor that targets SMO, cyclopamine [4,19,20,21,22,23]. Colon carcinomas are thought to derive from constitutive activation of WNT signaling by mutation of the APC or b-CATENIN genes, while the involvement of the HH signaling pathway is not as clear. In gastrointestinal malignancies, transcriptional up-regulation of HH ligands has been identified as the predominant activator of HH signaling in these diseases (reviewed in [3]). In addition, there is emerging evidence that HH signaling is involved in colorectal carcinogenesis [24,25], colon carcinoma stem cell self renewal, and in the metastatic behavior of advanced colon cancers [26]. However, genomic approaches to elucidate the role of HH signaling in cancers in general are lacking, regulatory genes downstream of GLI1 and GLI2 that function in cellular proliferation, survival, and maintenance of the malignant HH phenotype remain incompletely characterized [5], and data derived on HH signaling in colon cancer is extremely limited. Cellular proliferation is driven by progression of cells through the cell cycle consisting of sequential passage through G1, S, G2 and M phases. Cyclin-dependent kinases (CDKs) associate with cyclins to drive the cell cycle machinery [27,28]. Thus, CDK2 associates with CYCLIN E at the G1/S transition and with CYCLIN A during S-phase, CDK4 and CDK6 bind to CYCLIN D during progression at G1/S, while CDC2 complexes with CYCLIN A at G2, and with CYCLIN B during the G2/M transition. CDC25 family members also regulate cell cycle progression through dephosphorylation of the CDKs [29,30]. CDK inhibitors, including p21 Cip1 [29,31] and p15 Ink4b [ [32], bind to cyclin-CDK complexes during the cell cycle transition, in particular at G1/S (p21 Cip1 , p15 Ink4b ; [32]) and G2/M (p21 Cip1 ; [29]), and can also induce cell cycle arrest at the G1/S boundary following cytostatic signals through functional inhibition of cyclin-CDK complexes. The E2F family of transcription factors also regulates the expression of genes required for the G1/S transition, in particular genes involved in the activation of the DNA replication machinery, and DNA repair [33].
cDNA microarray technology has provided the ability to study the expression of thousands of genes simultaneously, and is an important tool in the dissection of signal transduction pathways. For the HH signaling cascade, HH/GLI target gene expression has been examined following EGF stimulation [6] or inducible GLI1 [16] or GLI2 [9] gene activation in human keratinocytes, or in GLI1-induced cell transformation [7]. To identify unique downstream targets of the GLI genes that function in cellular proliferation in the context of colon carcinoma, we employed a small molecule inhibitor of both GLI1 and GLI2, GANT61, identified in a cell-based small molecule screen for inhibitors of GLI1-mediated transcription [34]. GANT61 acts in the nucleus to block GLI1 function, inhibits both GLI1-and GLI2-mediated transcription, and demonstrates a high degree of selectivity for HH/GLI signaling [34]. Thus, GANT61 acts downstream of cyclopamine (targeting SMO) to inhibit the final determinants of HH transcriptional regulation.
In two human colon carcinoma cell lines, HT29 and GC3/c1, inhibiting the HH signaling pathway using GANT61 decreased expression of GLI1, GLI2 and the HH ligand receptor, PTCH1, and inhibited proliferation by inducing cellular accumulation at the G1/S boundary 24 hr after treatment, determined by flow cytometric analysis. On further detailed analysis using cDNA microarray gene expression profiling and quantitative Real-Time PCR, p21 Cip1 (CDKN1A) and p15 Ink4b (CDKN2B), that can elicit the G1/S checkpoint, were up-regulated, while genes that further determine entry from G1 to S-phase including E2F2, CYCLIN E2 (CCNE2), CDC25A and CDK2 were decreased in expression. Concomitant with decreased G1 to S-phase progression, decreased expression of CYCLIN A2 (CCNA2), CDC25C, CYCLIN B2 (CCNB2), CDC20 and CDC2 (CDK1), that regulate the passage of cells through G2/M were also demonstrated. Additional novel genes that are involved in stress response, and the response to DNA damage, not previously identified following termination of HH signaling in human cancer cells, include the early response genes DDIT2 (GADD45G), DDIT3 (GADD153), DDIT4 (REDD1), PPP1R15A (GADD34) and ATF3 that were significantly up-regulated. Genes involved in DNA synthesis and repair (TYMS, TOP2A, TK1, POLE, POLE2), and additional novel genes involved in S-phase progression or DNA damage responses that were significantly down-regulated, include KIAA0101 (p15 [PAF]), Replication Factor C variants 2, 3, 4, 5, CDT1, the E2F transcription factors CDCA4 and TFDP1, MDC1, FANCD2, PCNA, and the genes involved in DNA repair, RAD51C (XRCC3), RAD54B, RAD51 and HELLS. This study has therefore identified genes that are regulated during the termination of HH-dependent cellular proliferation and survival in colon cancer cells, and involves genes associated with G1/S-phase arrest, DNA damage and stress responses.

GANT61 induces down-regulated expression of GLI genes and accumulation of human colon carcinoma cells at G1/S
In HT29 and GC3/c1 cells treated with GANT61 (20 mM) for up to 48 hr, expression of the target genes GLI1 and GLI2 were both down-regulated, and also the HH ligand receptor PTCH1, as determined by qRT-PCR ( Figure 1A). Subsequently, HT29 or GC3/c1 cells were treated, in duplicate, with GANT61 (20 mM) followed by PI staining and flow cytometric analysis for the determination of cell cycle distribution between G1, S and G2/M phases ( Figure 1B). In both cell lines, cells accumulated in G1, 24 hr after treatment with GANT61. In GANT61-treated HT29 cells, a 20% increase in G1-phase cells was associated with a corresponding decrease in cells within the G2/M phase (14%) and in S-phase (5%), consistent with a G1/S checkpoint arrest. In GC3/c1 cells, an 8% increase in G1-phase cells at 24 hr after GANT61 treatment also corresponded with a similar reduction in cells in the G2/M phase of the cell cycle ( Figure 1B).
cDNA microarray analyses identify changes in gene expression both common and unique to HT29 and GC3/c1 following GANT61 treatment To delineate the changes in gene expression in HT29 and GC3/c1 human colon carcinoma cell lines in response to treatment with the GLI1/GLI2 antagonist, GANT61, the expression of 18,401 human genes was profiled in control cells treated with vehicle (0.2% DMSO) and in cells treated with GANT61 (20 mM) for 24 hr. Genes with a False Discovery Rate (FDR)-adjusted p-value of ,0.001 and fold change .1.5 were considered Differentially Expressed Genes (DEGs) induced by GANT61 relative to the vehicle control, of which 1,368 genes were differentially expressed in HT29, and 1,002 genes in GC3/c1 cells ( Figure 2). 755 genes or 558 genes were up-regulated, and 613 or 444 genes were down-regulated, in HT29 and GC3/c1 cells, respectively. 763 and 397 genes were differentially expressed and unique to HT29 or GC3/c1, respectively. Of the 763 DEGs unique to HT29, 459 (60%) were up-regulated and 304 (40%) were down-regulated. Similarly, of the 397 DEGs unique to GC3/ c1, 262 (66%) were up-regulated and 135 (34%), down-regulated. In contrast, 605 genes representing 3.4% of all genes were differentially expressed that were common to both cell lines; of these, 296 were up-regulated (49%), and 309 (51%) were downregulated ( Figure 2). All genes common to both HT29 and GC3/ c1 that were significantly up-regulated or down-regulated (p,0.001) are listed in Tables 1 and 2, respectively.

Modulation of canonical signaling pathways following inhibition of HH signaling
Genes with significant changes in expression following GANT61 treatment were assigned to different canonical signaling pathways and subjected to Ingenuity Pathway Analysis (IPA), where the resulting 1,368 DEGs in HT29 and 1,002 DEGs in GC3/c1 were mapped to networks defined by the IPA database ( Figure 3). For the mapped DEGs including both up-and down-regulated genes, the 15 most significantly altered canonical pathways in HT29 demonstrated -log(p-value) ranging from 2.045 to 9.025, and in GC3/c1 from 2.32 to 7.509. Of the 15 pathways involving genes significantly down-regulated, 12 were common to both cell lines. The 3 common pathways with the greatest differential down-regulated expression include genes involved in the DNA damage response, cell cycle checkpoint control, and mitosis. Other pathways down-regulated involved the G1/S and G2/M DNA damage checkpoints, DNA precursor metabolism, and cell signaling involving different pathways including those involved in cancers, which also demonstrate 3 signatures unique to either HT29 or GC3/c1 (Figure 3). Of the 15 pathways involving genes that are the most significantly upregulated, 8 are common to both HT29 and GC3/c1, and 7 represent unique pathways for each cell line, demonstrating more diversity in patterns of up-regulated gene expression ( Figure 3). The up-regulated pathways common to both cell lines include the metabolism-related such as steroids, pyruvate, glycolysis glutathione, or glycerolipid, and not directly related to the control of cellular proliferation.
Genes that demonstrate differential fold change patterns in GANT61-treated HT29 and GC3/c1 cells Fold change patterns of most highly DEGs in HT29 and GC3/ c1 were selected, analyzed and displayed in a heat map to evaluate and compare similarity and differences in differential expression between the two cell lines treated with GANT61 ( Figure 4). In addition to genes with diverse functions that are not directly related to HH-dependent proliferation, up-regulated genes that influence the G1/S transition and subsequent cell cycle progression, and that are common to both cell lines, include CDKN1A, and the DNA-damage-inducible transcripts 3 and 4 (DDIT3 and DDIT4). A considerably greater number of genes involved in cellular proliferation and cell cycle transition through the G1/S boundary, S-phase progression, and the G2/M transition, were significantly down-regulated in expression, and common to both cell lines. These include CDC6 (involved at G1/S), three genes that drive entry into and passage through S-phase (CCNE2, E2F2) and G2 (CCNA2), genes involved in DNA replication and repair (TYMS, POLE, TOP2A, TK1, POLE2), and two genes that regulate mitosis (AURKB, CDC20; Figure 4, asterisks).
Differentially expressed genes involved in the G1/S and G2/M transitions To further evaluate the genes involved in control of cell cycle progression in human colon carcinoma cells following GANT61 treatment, 10 genes involved in the G1/S or G2/M transitions, identified by IPA, were selected for further examination. Genes required at the G1/S boundary for G1/S transition, or for the induction of a G1/S checkpoint following cytostatic signals,      (Table 3).
To determine the robustness of cDNA microarray gene expression profiling following treatment of HT29 and GC3/c1 cells with GANT61 (20 mM) for 24 hr, qRT-PCR was employed to determine changes in expression of the selected group of 10 DEGs determined from the cDNA microarrays. The genes involved and primers synthesized are shown in Table 4. qRT-PCR was performed on cDNA generated using total RNA independently isolated from GANT61-treated HT29 and GC3/c1 cells for 0 hr, 16 hr, 24 hr, 38 hr and 48 hr after treatment. GAPDH was used to normalize all qRT-PCR data. Genes determined by qRT-PCR included the expression of E2F2, CCNE2, CDC25A and CDK2 at G1/S, which were downregulated, up-regulation of CDKN1A and CDKN2B at G1/S, and down-regulation of CCNA2, CDC25C, CCNB2, and CDK1 at G2/M ( Figure 5). The up-regulated or down-regulated changes      in gene expression following GANT61 treatment and determined by cDNA microarray profiling, were confirmed by qRT-PCR ( Figure 5).

Discussion
The HH signaling pathway is activated in a variety of human cancers following mutations in genes that regulate canonical HH signaling, including the receptor PTC, and the HH signaling molecule, SMO, and can also be activated via transcriptional upregulation of the HH ligands (reviewed in [3]). This pathway is becoming of increasing importance due to gaining insight into its prominent role in many developmental processes, and in the maintenance of the malignant phenotype in a wide variety of human cancers, whose growth has been found to be prevented by selective inhibition of constitutive HH pathway activity [14,35,36]. Tumors of the brain, prostate, skin, pancreas, and kidney have demonstrated the requirement for HH-GLI signaling, and have responded to inhibition of the HH signaling target molecule SMO by cyclopamine or SMOshRNA [4,19,20,21,22,23].
The transcriptional activators in HH signaling comprise members of the GLI family of transcription factors, GLI1 and GLI2, which have both distinct as well as overlapping functions [16]. Activation of the GLI proteins is an intricate process that involves modifications and interactions of a number of positive and negative pathway regulators and is not fully understood [1,14,35]. Target genes regulated by the HH signaling pathway differ between tissues and cell types, as well as being influenced by the presence or absence of regulatory factors co-expressed with GLI proteins that eventually determine the transcriptional programs activated by HH signaling [6,37]. Thus, oncogenic signaling pathways converge on canonical HH signaling at the level of the GLI transcription factors and additionally on target genes downstream of GLI1 and GLI2 to further drive the HH signaling pathway in cellular survival in malignancies [6,7,20,38,39,40]. The HH signaling phenotype is therefore significantly influenced and ultimately determined by the co-expression of additional regulatory factors, and hence by the cellular context of gene expression.
HH signaling plays a role in the differentiation program of normal intestinal villi [11,12,41], and it has been suggested recently that human colon cancer epithelial cells display a HH-GLI signaling axis in the process of carcinogenesis [24,25]. Expression of HH-GLI pathway components was consistently  demonstrated in an analysis of 40 primary human colon carcinomas and tumors metastatic to the liver [26], consistent with findings of previous investigators [25,42,43]. Thus, using qRT-PCR, the expression of GLI1, PTCH1, GLI2 and SHH was determined in all human colon carcinomas examined. The requirement for both GLI1 and GLI2 for sustained proliferation and survival of human colon carcinoma cell lines in vitro, including HT29, was demonstrated using siRNA technology [26].
In addition, knockdown of SMO by SMOshRNA prevented the growth of HT29 cells in SCID mice, while wt HT29 subcutaneous xenografts responded to cyclopamine by reduction in tumor volume [26]. Thus, canonical activation of GLI1 and GLI2 via SMO is important for the survival and proliferation of human colon carcinoma cells in vivo.
In the current study, the function of both GLI1 and GLI2 downstream of SMO was inhibited in the presence of GANT61, a small molecule inhibitor that was identified from a cell-based screen to specifically inhibit GLI1-mediated transcription, but that also inhibited the function of GLI2 [34]. This agent was selected to specifically inhibit the final arbiters of HH signaling, the GLI transcription factors, in elucidation of the downstream target genes that determine HH-dependent proliferation in human colon carcinoma cells. Two cell lines, well characterized in our laboratories, HT29 and GC3/c1, were treated with GANT61 (20 mM) for 24 hr, and the expression of GLI1, GLI2 and PTCH1 mRNA was down-regulated. Further, the effects on cellular proliferation as determined by the distribution of cells within the cell cycle and flow cytometric analysis demonstrated accumulation of cells in G1 following treatment, with a concomitant decrease of cells from the G2/M compartment, and in the case of HT29, also from S-phase, suggesting the induction of a G1/S checkpoint.
HT29 and GC3/c1 cells were subsequently treated with GANT61 (20 mM) for 24 hr, RNA was extracted, and changes in gene expression were determined by Illumina cDNA microarray profiling. Following statistical analyses, 1,368 genes in HT29 and 1,002 genes in GC3/c1, were determined to be significantly modulated by GANT61 treatment (FC.1.5; p,0.001). For genes that were up-regulated in expression, 296 genes were common to both cell lines, and for down-regulated genes, 309 genes were common to both cell lines. The blockade of cells at the G1/S boundary is evidenced by up-regulated expression of p21 Cip1 and p15 Ink4b that in part regulate the G1/S transition. p15 Ink4b is a member of the Ink4 family of CDK inhibitors, is induced in response to cytostatic signals [32,44], and complexes with CYCLIN D/CDK4 or CYCLIN D/CDK6 to mediate G1-phase arrest at the G1/S transition in certain systems [30,32]. p21 Cip1 can bind a broad range of cyclin-CDK complexes, with a preference for those containing CDK2 (reviewed in [31,45], and during a normal cell cycle, facilitates active cyclin-CDK complex formation to promote cell proliferation. However when overexpressed, p21 Cip1 forms an inhibitory complex with CYCLIN E/ CDK2, leading to G1-and consequently S-phase arrest, thereby forming the G1/S checkpoint [46,47]. The scheduled timing of expression of CYCLINS E, A and B that drive cell cycle progression, is reflected in the major changes in the phases of the cell cycle. CYCLIN E is expressed at maximal levels in cells undergoing the G1 to S transition, declining during S-phase progression, such that G2/M cells are CYCLIN E-negative (reviewed in [30]). CYCLIN A is expressed in late G1, demonstrates pronounced expression during S-phase, and increases as the cells advance towards G2, with degradation in early mitosis; B type cyclins begin to be expressed in late S-phase, and drive the cells though G2-and M-phases of the cell cycle [30]. CDK2 controls the G1/S transition by complexing with CYCLIN E, and is activated by CDC25A, which dephosphorylates CDK2 [48]. CDC2 controls cellular entry into mitosis at the G2/M transition, thereby forming complexes with CYCLINS A and B, is activated by CDC25C, and is down-regulated in late M-phase [29]. All of these genes are down-regulated in expression in response to the inhibition of GLI1/GLI2 function (Table 2). Thus, signals elicited to promote cellular accumulation at G1/S lead to the repression of genes that regulate further cell cycle progression, and p21 Cip1 expression has been shown to deplete the expression of genes that regulate DNA replication and repair, and mitosis [49,50,51]. In the current study these genes include CDC6 (active at the G1/S transition and essential for the initiation of DNA replication), TYMS, TOP2A, TK1, POLE and POLE2 (S-phase), and AURKB and CDC20 (mitosis [52,53]), determined by heat map analysis. A schematic representation of the genes involved in GANT61-induced inhibition of cell cycle progression at G1/S, Sphase progression, and regulation during G2-and M-phases, identified from cDNA microarrays, heat map analysis, and by qRT-PCR, is shown in Figure 6, and involves 5 of the 12 common signaling pathways determined by IPA analysis.
In summary, we have compared gene expression profiles in two human colon carcinoma cell lines after targeting the function of the transcriptional regulators of HH signaling, GLI1 and GLI2, using the small molecule inhibitor GANT61. Data are consistent with accumulation of cells at the G1/S boundary, as evidenced from flow cytometric analysis, cDNA microarray gene profiling, and qRT-PCR. GANT61-treated cells demonstrated up-regulat- Figure 3. The top 15 canonical signaling pathways influenced by inhibition of GLI1/GLI2 function in HT29 and GC3/c1 cells. The top 15 canonical signaling pathways, determined by IPA, that were significantly up-regulated or down-regulated by GANT61 treatment in HT29 and GC3/ c1 cells, are shown. The 1,368 DEGs in HT29 and 1,002 DEGs in GC3/c1 were mapped to the IPA-defined network. The significance p-values that determine the probability that the association between the genes in the dataset and the canonical pathway is by chance alone were calculated by Fisher's exact test, and are expressed as -log (p-value). A. Pathways with enriched down-regulated genes. B. Pathways with enriched up-regulated genes. Blue: Pathways common to both HT29 and GC3/c1. Red: Pathways unique to either HT29 or GC3/c1. Yellow squares: Ratio of the number of DEGs that map to a specific canonical pathway divided by total number of genes that make up that pathway. doi:10.1371/journal.pone.0013054.g003 ed expression of the CDK inhibitors p21 Cip1 and p15 Ink4b that function at the G1/S boundary, and down-regulated expression of additional key genes that determine the G1/S transition, initiation of DNA replication, S-phase progression, DNA repair, and subsequent transition through the G2/M phases. Inhibition of the transcriptional regulation of HH signaling in human colon carcinoma cells therefore directly involves genes that regulate cell cycle transition through G1/S, cell cycle progression, and proliferation, and genes involved in stress-induced and DNA damage responses.

Materials and Methods
Human colon carcinoma cell lines HT29 was purchased from ATCC (Manassas, VA), while GC3/ c1 was established in culture by our group from a human colon adenocarcinoma xenograft model [54]; both cell lines express mutant p53 alleles. Cell lines were maintained in the presence of folate-free RPMI 1640 medium containing 10% dFBS and 80 nM [6RS] 5-methyltetrahydrofolate.

Flow cytometric analysis
HT29 and GC3/c1 cells were plated at a density of 100,000 cells/well in six-well plates. After overnight attachment, cells were treated with GANT61 (20mM; Enzo Life Sciences, Germany) or vehicle control (DMSO, 0.2%), in duplicate, for 24 hr, followed by washing 61 with PBS, trypsinization, and centrifugation. Cells were fixed with 70% ethanol at RT, 20-30 min, stored at 220uC overnight, centrifuged for 5 min at 2006g to remove ethanol, and washed 61 in PBS. The cells were resuspended in PBS, and low molecular weight DNA was extracted using DNA extraction buffer (0.2M Na 2 HPO 4 and 0.1M citric acid, pH 7.8) for 5 min. The extracted DNA was centrifuged and resuspended in DNA staining solution containing propidium iodide (50 mg) and DNAse-free RNAse (2 mg), and allowed to incubate for 30 min in the dark at RT. Distribution of cells throughout the cell cycle was analyzed using a FACSCalibur flow cytometer, and data were analyzed using CellQuest software.

RNA isolation
Cells were seeded at 7610 6 cells/10 cm plate overnight at <60% confluency, and subsequently treated with either vehicle control (0.2% DMSO) or GANT61 (20 mM), in duplicate, for 24 hr. Cells were subsequently harvested and RNA was isolated using the RNeasyH Mini Kit (Qiagen, Valencia, CA) following the manufacturer's protocol. Integrity of the RNA was determined by spectrophotometry and electrophoresis.
cDNA microarray analysis RNA (250 ng) was reverse transcribed into cRNA and biotin-UTP labeled using the IlluminaH TotalPrep RNA Amplification Kit (Ambion, Applied Biosystems, Foster City, CA) according to the manufacturer's protocol. cRNA was quantified using a nanodrop spectrophotometer, and the cRNA quality (size distribution) further analyzed on a 1% agarose gel. Biotinylated cRNAs were hybridized to the Illumina Human-ref8 V3.0 BeadChip (Illumina, San Diego, CA) that represents 18,401 genes, using standard protocols. The arrays were washed and subsequently scanned using an Illumina BeadArray Reader.
Raw signal intensities of gene expression data were processed and analyzed using GenomeStudio (Illumina), with background subtraction and average normalization performed on the average signal intensities, and p-values were calculated. Raw data were exported from GenomeStudio into Excel (available as Data S1 and Data S2), and the fold change in gene expression calculated, by dividing the average gene expression signal intensity of treated samples (GANT61) with that of the vehicle control (DSMO). Differential gene expression (DEG) analysis between control and GANT61-treated samples, was subsequently conducted using the Illumina customer model, which applies multiple testing corrections that determines the false discovery rate (FDR; [55]. Thus, FDR-adjusted differential scores and p-values for each gene/probe set between treated and control samples were generated. Genes with a FDR-adjusted p-value of p,0.001 and fold change $1.5 were considered to be DEGs and were subjected to Venn analysis and Ingenuity Pathway analysis. specific roles in G1/S transition, S-phase progression, DNA replication or repair, or regulation of the G2-or M-phase transitions. Fold changes of all down-regulated DEGs and all but one up-regulated DEG are #8 (central color spectrum bar). doi:10.1371/journal.pone.0013054.g004 Table 3. Gene expression changes from cDNA arrays at G1/S and G2/M. These differentially expressed genes were, uploaded and mapped to the library of canonical pathways of the Ingenuity Pathways Analysis (IPA) database for pathway analysis (IngenuityH Systems, http://www.ingenuity.com, Mountain View, CA). The mapped datasets containing either up-regulated or down-regulated genes in HT29 or GC3/c1 cells with corresponding expression values were subjected to core analysis that includes overall canonical pathway enrichment analysis. The significance of enrichment of genes mapped to different canonical pathways was calculated by the Fischer's exact test (p-value) to determine the probability that the association between the genes and the canonical pathway could be explained by chance alone. Further, the ratio between the number of identified genes in a particular pathway and total number of genes that make up that pathway provides an estimation of the extent of pathway involvement. The enriched canonical pathways were ranked by 2log (p-value) as shown in a histogram of pathway vs. -log (p-value). Datasets containing both up-regulated and down-regulated genes were also analyzed in selected pathways pertaining to cell cycle progression at G1/S and G2/M. Fold change patterns of most highly DEGs in HT29 and GC3/c1 were compared by Heat map analysis using Matlab (Mathworks) software. DEGs analyzed and displayed in the heat map demonstrated a differential expression p-value of p,0.001 between control and GANT61 treated cells in both cell lines.

Quantitative Real-Time PCR (qRT-PCR)
The expression levels of selected genes identified by cDNA microarray expression profiling at 24 hr following GANT61 treatment, were validated by qRT-PCR. Thus, HT29 or GC3/ c1 cells were either untreated (vehicle control, 0.2% DMSO), or treated with GANT61 (20 mM) for 0 hr, 16 hr, 24 hr, 38 hr or 48 hr at 37uC, dissolved in DMSO-containing medium. Total RNA (1 mg) was employed to prepare cDNA via reverse transcription using the iScript Select cDNA Synthesis Kit (Biorad) Reverse Transcription System according to manufacturers instructions and analyzed using an Applied Biosystems 7500 PCR Detection System (Applied Biosystems Inc.). All amplifications were primed by pairs of chemically synthesized 18-to 24-mer oligonucleotides designed using freely available primer design software (Primer-BLAST, NCBI) to generate target amplicons of 50-547 bp. All reactions were performed in a final volume of 15 ml. qRT-PCR reaction conditions were as follows: activation at 95uC for 10 min with 40 cycles of denaturation at 95uC for 15 s, primer annealing and extension at 60uC for 1 min and ramping back to 95uC. Melt curve analysis of all samples was routinely performed to ascertain that only the expected products had been generated. A fluorescence reading determined the extent of amplification at the end of each cycle. mRNA expression levels of target genes were normalized to the expression of glyceraldehyde phosphate dehydrogenase (GAPDH) and quantified using the comparative CT method [56]. Q-RT-PCR for each gene was   Table 2. Data represent the mean6SD of 4 determinations, and GAPDH was used to normalize the relative mRNA levels. doi:10.1371/journal.pone.0013054.g005 Figure 6. Schematic representation of genes involved in GANT61-induced inhibition of cell cycle progression. From cDNA microarray, heat map, and qRT-PCR analyses, genes involved at different phases of the cell cycle including the G1/S transition, and progression through S-G2and M-phases, are shown. The genes identified include CDK inhibitors, members of the CDK and CDC families, cyclins, genes involved in DNA replication and repair, and genes that regulate the mitotic spindle, and involve 5 of the 12 common signaling pathways determined by IPA analysis. Red: Up-regulated genes. Green: Down-regulated genes. Light shadeRdark shade, increasing differential expression. doi:10.1371/journal.pone.0013054.g006 determined in duplicate, and each experiment was repeated at least twice.

Supporting Information
Data S1 HT29 Raw Microarray Data.