Identification of Genomic Targets of Transcription Factor Aebp1 and its role in Survival of Glioma Cells Free

Glioblastoma multiforme (GBM) is the most common and malignant form of primary tumor of CNS in adults, which is characterized by a median survival of less than a year. The prognostic behavior of GBMs is rather poor and hence there have been efforts to identify molecular signatures and also to discover new biomarkers for characterizing different types and stages of GBMs (1–4). GBM is broadly classified into primary and secondary GBM (WHO), each one arising through distinct genetic pathways. Primary GBM arises de novo and is frequently associated with amplification and/or overexpression of EGFR and PTEN deletion combined with INK4A/ARF and CDKN2A losses and MDM2 amplification (5). However, secondary GBM often exhibits P53 mutations, PDGF/PDGFR overexpression, RB loss, and CDK4 amplifications (6). Recent studies have shown, however, that there is an overlapping spectrum of mutations in these 2 types of GBM (7, 8). In one of our earlier studies we had found AEBP1 expression to be upregulated in primary GBMs as opposed to progressive secondary GBMs (9). Aebp1 was originally identified as a transcriptional repressor that binds to adipocyte enhancer 1 (AE-1 element) located in the proximal promoter region of the adipose P2 gene, which codes for adipocyte specific fatty acid binding protein 4 (FABP4; ref. 10). Aebp1 is also overexpressed in transgenic mouse probasin-Neu (ERBB2) induced advanced prostate cancer (11). However, the exact role of AEBP1 in tumorigenesis is not clear and hence we set out to identify the genomic targets of this transcription factor to understand its biology in the cellular context. Toward this direction we have undertaken a detailed study to analyze the Aebp1 genomic targets by transcriptome profiling of AEBP1 downregulated U87MG cells and its role in cell proliferation, growth, and survival.

U87MG and U138MG cells (ATCC) were grown in Eagle's Minimal Essential Medium supplemented with 10% FBS (Sigma-Aldrich). Cells were transfected with 100 nmol/L siRNA pool targeted against AEBP1 (Dharmacon Inc.). Quantitative real-time PCR (qRT-PCR) was done using Eva Green (Biorad) on a Biorad iQ5 cycler. Downregulation of AEBP1 was assessed by qRT-PCR and Western blot analysis. All the primers used in this study are listed in Supplementary Table S1.

Total RNA was isolated from 4 independent scrambled siRNA and 5 test siRNA–AEBP1 transfected U87MG cells and hybridized to Affymetrix U133Plus 2.0 gene chip that queried 47,000 genes. The results were analyzed initially using Gene-Chip operating software and the data were subsequently processed using ArrayAssist (Agilent) to statistically analyze changes in gene expression. qRT-PCR validation for 114 genes was done using Eva Green (Biorad) on a Biorad iQ5 cycler. ChIP assays were done according to the previously described method (12). Briefly, log-phase U87MG cells were fixed with formaldehyde and chromatin was sonicated to generate an average length of 200 to 800 base pairs. After preclearing, the chromatin solution was incubated with affinity-purified rabbit polyclonal Aebp1 antibody (SantaCruz) or purified rabbit IgG antibody. The abundance of genomic DNA containing a promoter was determined by PCR amplification using sequence-specific primer pairs flanking Aebp1 binding site identified through position weighted matrix analysis within −1 kb promoters. For ChIP-chip analysis, amplified immunoprecipitated DNA and input genomic DNA was labeled with Cy5 and Cy3 fluorophores respectively, using random primer labeling kit (Invitrogen Corp.). Five micrograms each of immunoprecipitated and genomic DNA was combined along with human Cot-1 DNA and hybridized to each of the Agilent human promoter tiling array (2 × 224) containing 474,393 probes excluding control features.

Gene expression data was normalized using PLIER algorithm in ArrayAssist (Agilent) and expression changes were filtered at >1.5-fold between experiments. Genes were considered significantly perturbed at a p-value of ≤0.05. The method of Benjamini and Hochberg (13) for false discovery rate was set to 0.05 using R software (14). These genes were then subjected to an unsupervised 2-way average linkage hierarchical cluster analysis with uncentered correlation as similarity metric using Cluster 3.0 software (15). Java Tree view version 1.1.4 was used to visualize structure of the data (16). Functional annotation was done using Gene Ontology database and DAVID Ease software (17–19) on differentially regulated genes. Pathway enrichment analysis was done using Genotypic Technologies Biointerpretor tool. A p-value cut-off of 0.05 was used to identify significant enrichment pathway categories.

Raw intensity data were generated using Feature extraction software v 10.5.1.1. Feature extracted data were analyzed using DNA Analytics software from Agilent (hg18 build). Data were normalized using Median Blanks subtraction to exclude the probes having negative intensities, intraarray median normalization to remove dye bias and interarray median normalization to bring the distribution uniform across replicates. The significantly enriched genes were detected based on the statistical p-value using Whitehead Per Array Neighborhood Model. False discovery rate analysis (13) was then applied to 11,659 enriched genes for Aebp1 promoter occupancy using Bioconductor R software (14). Peaks were considered significant at a P-value ≤0.05. Two biological replicate experiments were carried out for Aebp1 occupancy analysis.

Promoter sequences (−1 kb) of perturbed genes were retrieved using 3 major databases viz., transcription regulatory element database (TRED), eukaryotic promoter database (EPD), and UCSC genome browser (20–22). De novo motif discovery was done using CisFinder algorithm (23) to identify motifs in most enriched sequences by ChIP experiments. Position frequency matrices were estimated from counts of n-mer words with and without gaps and clustered to generate nonredundant sets of motifs. Web logo was used to construct sequence logos (24). To test the validity of motifs predicted from ChIP-chip data, we built control data set of 5810 random sequences each of approximately 50mer length from human, using RSAT tools (25) and motif analysis were conducted for these random sequences.

Pearson's correlation coefficients were calculated between all replicates in gene expression and promoter tiling arrays using R statistical computing (14).

The DNA binding sites of 25 transcription factors, which were perturbed upon AEBP1 gene silencing and also validated by qRT-PCR, were mined from the literature. These binding sites were searched in −1 kb promoters of each of the transcription factor genes and transcription factor gene network among these 25 transcription factors was generated. The heat maps were constructed using Java Tree view software (16).

FABP4 promoter (−200 to +21) was amplified from genomic DNA and cloned into the XhoI and HindIII sites of basic pGL3-promoter vector (Promega Corp.). Mutant motif promoters were generated by substituting G for A and C for T and vice versa (Supplementary Table S1). Two micrograms of various reporter constructs were cotransfected in U87MG cells with 200 ng of pCMVβ (that expresses the β-galactosidase gene under the control of CMV promoter) as transfection control. After 24 hours of transfection relative light units was measured in a Luminometer (Berthold detection systems). β-Galactosidase activity was measured by fluorometric assay and used to normalize transfection efficiency.

U87MG and U138MG cells were plated in 96-well plates and were transfected every 60 and 36 hours, respectively, with siRNA pool designed against AEBP1 or nontargeting scrambled siRNA. To assess the effect of the AEBP1 gene silencing, cells were treated with MTT (3-[4-5 dimethyl thaizole-2-yl]2-5 diphenyltetrazolium bromide; Sigma-Aldrich) for 4 hours and the formazan crystals formed by metabolically active cells was solubilized and measured in a spectrophotometer at 550 nm. Colony suppression was done on U87MG or U138MG cells (0.5 × 10⁶) by transfection with 2 μg of control shRNA vector (Open Biosystems) or shRNA designed against AEBP1. Forty-eight hours posttransfection, puromycin selection was done for >2 weeks. Resistant colonies were stained with crystal violet solution and photographed. Apoptosis was assayed using FITC-Annexin V-PI (Invitrogen Corp.) and APO-BrdU kit (Becton Dickinsion) following manufacturers protocol.

We had observed earlier that AEBP1 was upregulated in the majority of primary GBM tumor samples (9). Here, we have used a complementary approach wherein we have suppressed endogenous AEBP1 expression in U87MG cells, an astrocytoma cell line, to gain an insight toward understanding the biological role of AEBP1. We found that 100 nmol/L of siRNA pool brought about significant downregulation of AEBP1 (>90%) as against mock (scrambled siRNA) treated cells without affecting the expression of human β-actin and GAPDH (Fig. 1A). Downregulation of AEBP1 was also observed at the protein level (Fig. 1B). Global gene expression profile of U87MG cells after mock transfection or transient silencing of AEBP1 was determined by using human U133 plus 2 array from Affymetrix. The correlation coefficient analysis of the expression data revealed that results are comparable between replicates (Supplementary Fig. S1A). A flow diagram of different steps of our analysis is shown in Fig. 1C. We observed perturbation of expression in 734 genes at more than 1.5-fold change at a P-value of ≤0.05 (Supplementary Table S2) of which 326 genes were upregulated and 408 genes downregulated. These genes were sorted by expression ratios; median centered and then subjected to hierarchical cluster analysis (Fig. 1D, downregulated and Fig. 1E, upregulated). Functional categorization revealed a diverse set of GO biological processes that were statistically significant. Enriched categories included cell proliferation, cell cycle, cell differentiation, apoptosis, transcription, protein and ion binding, signaling, and ubiquitin related (Fig. 1F). A list of the most altered genes based on gene ontology is given in Table 1.

To elucidate transcriptional targets of Aebp1, we retrieved well-annotated and characterized promoter sequences of these 734 genes. Among them 65 genes were unannotated leaving behind 669 genes, which formed the basis for our further analysis. To identify targets of Aebp1, we started with Aebp1 DNA binding site based on the previous literature (26). Aebp1 binds to AE-1 sequence (−168 to +21) of aP2 gene that was originally identified by Hunt and colleagues (27) (Supplementary Fig. S2A). Ro and Roncari (26) also showed the importance of (G-139) in the aP2 promoter for binding of positive and negative factors to the AE-1 element. PPARγ and LXRa are also reported as targets of AEBP1 (28). This was done on the sequence length of 35 base pairs. When we reanalyzed these sequences, we found that nucleotides matching with AE-1 element were scattered in the case of PPARγ. The nucleotides that matched most with PPARγ promoter were “AGAA” starting at −644 to −641 and “AGAAATTT” at (−631 to −624). We checked the presence of this motif within the ChIP DNA obtained from Aebp1 chromatin immunoprecipitation using the flanking primer pairs to GAAAT present in the FABP4 promoter. We could observe enrichment of this fragment in the ChIP DNA of FABP4 (Supplementary Fig. S2B). Analysis of the promoter sequence (−1 kb) of brain-specific FABP7 and its subsequent enrichment in ChIP PCR confirmed that it also contains the GAAAT sequence (Supplementary Fig. S2C and S2D). Based on all these observations we predicted that “GAAAT” is the probable Aebp1 binding site. Using this information we searched for the presence of this motif in the −1 kb upstream sequences of the transcription start sites in all the 669 genes. Among these, 442 genes had this predicted AE-1 element, whereas 227 genes did not have this element (Fig. 1C). There were a total of 863 predicted motifs in these 442 promoters.

Our next effort was to experimentally show the occupancy of Aebp1 in the promoter sequences. To address this question we employed the ChIP-chip technique using Agilent human promoter tiling array. The Aebp1 bound immunoprecipitated DNA was hybridized to the tiling array in replicates. Peak detection algorithm of DNA Analytics detected robust peaks of probe signal corresponding to the binding events. The correlation analysis showed that results are reproducible between replicates (Supplementary Fig. S1B). The chromosome wise occupancy of Aebp1 is shown in Supplementary Fig. S3(A–X). We detected 11,659 genes as target sites of Aebp1 occupancy. These genes were further subjected to FDR analysis (13) to minimize false positives. A total of 5810 genes were subsequently identified following this exercise (Fig. 1C). Binding sites predicted using CisFinder algorithm for these genes are documented in Supplementary Table S3 along with their frequencies and enrichment ratios. Further to rule out any nonspecificity, we retrieved random sequences of 5810 genes using RSAT tools (25). Motif analysis of these random sequences did not predict real motifs (GAAAT/TTTCT) as shown in Supplementary Table S4. Of the 669 genes that were modulated upon silencing of AEBP1 gene, 185 genes overlapped with the ChIP-chip enriched gene list. The list of these congruent genes seen both in microarray and tiling arrays are given in Supplementary Table S5 and their location on individual chromosomes analyzed by human genome tool in UCSC genome browser (22) is presented in Supplementary Fig. S4. The congruent genes between microarray and promoter tiling array analysis were divided into 3 categories based on their probe location as within the promoter, inside the gene or downstream from transcription start site (Fig. 2A). A total of 49.73% of these genes showed binding in the proximal promoter region, whereas 50.27% show binding in the genes downstream of the transcription start site. It is not uncommon that transcription factor binding sites are observed within the coding regions (29). Fig. 2B shows the distribution of best probes for top scoring target genes as a function of their distance from the transcription start site. We also observed that 227 genes that were differentially regulated upon AEBP1 silencing did not possess the predicted binding motifs.

As discussed previously, FABP4 and FABP7 are known targets of Aebp1 containing the AE-1 element that possess GAAAT motif. Hence when we scanned the enriched peaks of the congruent target genes and queried for highly represented binding site using CisFinder algorithm, GAAAT containing motif was enriched in these genes. The resulting position weight matrix (Supplementary Fig. S5A) was used to generate sequence logo (Fig. 2C). Some of the genes like PCDHA2, zinc finger proteins, EPHA4, CA3, and CSRP2BP had this site represented more than 4 times within their −1 kb promoters. Those perturbed genes derived from the microarray data that did not have predicted AE-1 elements in their proximal promoters [227] was further compared with congruent targets [185]. This comparison yielded 63 genes for which we further analyzed for an alternate consensus motif.We found a TTTCTTTAT motif to be highly enriched in these genes, the position weight matrix and logo of which is represented in Supplementary Fig. S5B and S2D, respectively. Interestingly, AE-1 element in the promoter region of both FABP4 and FABP7 contained part (TTTCT) of the alternate consensus motif, more so in FABP4 where both motifs were present in tandem. It was noteworthy that some congruent target genes also had this motif present along with GAAAT, albeit noncontiguously in their −1 kb promoter. The distribution of these 2 enriched motifs in the promoters of the congruent 185 genes is represented in Fig. 2E.

The motifs identified above by bioinformatics approach were further experimentally confirmed by Luciferase reporter assay with wild-type and mutant Aebp1 binding motifs in the FABP4 promoter region containing a continuous stretch of GAAAT and TTTCT (Fig. 2F). U87MG cells were cotransfected with β-galactosidase construct to normalize for transfection efficiency. This analysis showed that GAAAT site in particular was responsible for AEBP1 promoter function. Mutant GAAAT motif showed considerable reduction in luciferase activity in comparison to TTTCT motif mutation. A double mutant motif of GAAATTTCT showed almost negligible promoter activity compared with the wild-type control.

To experimentally confirm the validity of our ChIP-chip data, we carried out ChIP-PCR analysis of randomly selected 48 genes belonging to varied gene ontology (Fig. 2G). Flanking primers pairs were designed for the computationally predicted occupancy in the promoters of these genes (Supplementary Table S1). As can be seen in Fig. 2G, the promoter sequences are indeed enriched in the Aebp1 ChIP-DNA as compared with preimmune control sample. We also amplified some of the genes that did not have GAAAT sequence for Aebp1 binding (TP53INP1, ERBB2IP, DMTF, and SIVA-1). These genes showed no amplification in the immunoprecipitated DNA sample.

As discussed earlier, among the 734 differentially regulated Aebp1 target genes (Supplementary Table S2) we noticed several of the genes belonged to varied pathways; 27 related to cell cycle, 13 to differentiation, 27 to proliferation, and 21 to apoptosis. The quantitative expression pattern of these genes following siRNA-mediated downregulation of AEBP1 were further validated by real-time PCR analysis from 3 independent biological replicates as shown in Fig. 3A to D. Some important molecules related to cell cycle such as CDC20, CDC25C that promote mitosis (30, 31) are downregulated, whereas CDK6 and MDM2 are upregulated. It is also noteworthy that TP53, a tumor suppressor protein, is upregulated upon AEBP1 silencing. Again, ITGB1 as well as FZD8 involved in the process of differentiation and cancer are upregulated whereas NGEF is downregulated. The expressions of IRS1, EGFR, IL4R, PDGFB, and NRAS that are widely implicated in promoting proliferation in cancer cells upon dysregulation of growth factor signaling are increased. DNA replication related gene PRIM2A is downregulated and so is the growth index gene MKI67. Interestingly, apoptotic regulators TNFAIP3, TNFAIP8, TNFFRSF10D, TNFSF14, and BIRC5 show increase in their expression. Cellular hypoxia is a hallmark of cancer and ARNT that rescues cells from such a condition is downregulated upon AEBP1 gene downregulation.

It is also interesting to note that a large number of transcription factors are differentially regulated upon AEBP1 gene silencing (Fig. 1F) in addition to the genes affecting growth phenotype. We also carried out real-time PCR analyses of these 25 transcription factors that were differentially regulated in our microarray experiment (Supplementary Table S2), which are presented in Fig. 3E. Among these, transcription factors FOXJ3, ASCL2, ELF5, CLOCK, NFATC1, EPAS1, NFE2L3, and TBX15 do contain AE 1 binding element in their promoters and therefore are differentially regulated upon AEBP1 gene silencing. However, we observe alteration in expression of other transcription factors as well. It is quite likely that these transcription factors might cross-regulate each other by binding to their cognate DNA binding sites present in the promoters of the transcription factor genes. Based on the bioinformatics analysis, we constructed a transcription factor network of the differentially regulated transcription factor genes, which is represented in Fig. 3F. We found that some of these genes also possess binding sites for many of the transcription factors. Prominent among them are FOXJ3, ASCL2, CLOCK, ELF5, and NFATC1. The functional relevance of this transcription factor network in the biology of AEBP1 and more so in gliomagenesis needs further investigation.

Because many of the genes associated with cell growth and apoptosis were perturbed in AEBP1 silenced cells, we examined the growth promoting potential of Aebp1in 2 glioma cell lines, U87MG, and U138MG cells. For this purpose, AEBP1 was silenced using 100 nmol/L siRNA pool. Maximum downregulation was observed at 72 and 48 hours posttransfection for U87MG and U138MG cells, respectively (Fig. 4A). Hence, siRNA pool was replenished every 60 hours in case of U87MG and 36 hours in the case of U138MG cell line for over a period of 9 days and assessed for cellular proliferation by the MTT assay. There was no apparent change in cell viability during the first 4 days of posttransfection. However, from 5th day onward there was a reduction in cell viability suggesting that silencing of AEBP1 resulted in loss of proliferative potential or cell death (Fig. 4B and C). Further colony suppression assay was done by transfecting AEBP1-shRNA construct in U87MG and U138MG cell lines. In comparison to mock shRNA treated cells, most of the AEBP1 silenced cells do not survive to form colonies after >2 weeks of puromycin selection (Fig. 4D).

As described earlier, silencing of AEBP1 expression resulted in loss of cellular viability, which prompted us to explore the mechanism of this phenotype. For this purpose, U87MG and U138MG glioma cells were synchronized by serum deprivation and after siRNA mediated downregulation, cells were analyzed for Annexin V/PI staining [Fig. 5A (U87MG) and B(U138MG)] as well as for DNA fragmentation by TUNEL assay [Fig. 6A (U87MG) and B (U138MG)]. Annexin staining because of phosphatidyl serine externalization was observed in both the cell lines from 5th day of postsilencing. The percentage of cells showing early apoptosis were 10.83%, 15.45%, and 22.24% in U87MG cells on 5th, 7th, and 9th day, respectively. At same time points, the early apoptotic cells were 11.91%, 11.07%, and 7.2% in U138MG cells. Similarly, late apoptotic cells were 2.82%, 12.16%, and 15.23% for U87MG cells and 10.54%, 20.20%, and 38.32% in U138MG cells at same time points postsilencing (Fig. 5A and B). Both FITC and propidium iodide staining of late apoptotic cells can also be visualized on 7th and 9th day, respectively. In TUNEL assay, we observed considerable DNA fragmentation in both the cell lines upon AEBP1 gene silencing. For U87MG cells, the percentage was 12.96% on 5th day of postsilencing, which increased to 44.80% on 9th day and in the case of U138MG cells, the percentage of 15.09% on 5th day increased to 50.33% on the 9th day of postsilencing (Fig. 6A and B). Thus, these results clearly show that AEBP1 downregulation drives the glioma cell lines toward apoptosis.

Aebp1 is a transcription factor that has been studied in great detail toward its role in regulating adipogenesis by its repressive action on aP2 promoter (26). Our observation showing an upregulation of AEBP1 in primary GBM (9) prompted us to undertake a genome wide approach to identify the targets in an effort to understand the various pathways that are intimately influenced by Aebp1. A total number of 734 genes were found to be perturbed under AEBP1-silenced condition in comparison to mock siRNA treated U87MG cells. A large number of genes that were differentially regulated belonged to categories such as cell cycle, differentiation, proliferation, apoptosis, and transcription regulators. We followed up this study with ChIP-chip analysis to determine the occupancy of Aebp1 on the promoter sequences using Agilent human promoter tiling array. Initially, 11,659 genes were picked up for occupancy of Aebp1. However, on application of false discovery rate analysis (13), the list was reduced to 5810 genomic loci. Interestingly, only 185 genes were congruent between the gene list of differentially regulated genes and those with Aebp1 occupancy in their promoter region.

Another effort of this work was to identify a consensus Aebp1 binding element in the promoters of the identified gene list. A systematic bioinformatics approach has indeed identified 863 GAAAT motifs present in a total number of 442 genes. We also found another motif TTTCT highly enriched in the genes that did not contain the GAAAT motif within their promoters. Some of the targets (118 genes) containing GAAAT motif also had TTTCT consensus motif. This bioinformatics/in silico identification of these putative Aebp1 binding elements were experimentally confirmed by functional promoter assays and it is evident that GAAAT plays a major role in Aebp1 mediated promoter activity on FABP4 promoter (Fig. 2F). The sequence TTTCT present in tandem to GAAAT seems to have an additive role in the promoter function. Mutation in both of these sites resulted in a drastic abrogation of the promoter function.

To interpret the biological significance of Aebp1 in glioma, we undertook a systematic and comprehensive analysis of the gene expression data. Apart from differential regulation of genes related to cell cycle, differentiation, proliferation, apoptosis, and transcription regulators as mentioned previously; we also retrieved genes belonging to signaling pathways such as insulin, MAPK, WNT, TGF-β that are widely implicated in gliomagenesis. Some of the important genes that are affected upon AEBP1 downregulation and have established role in glioma are growth factors EGFR, PDGFB, (5, 32) and growth factor signaling intermediates like NRAS, IRS1, and IL4R (33–35). Although, we observed enrichment of genes pertaining to activation of NF-κB and PI3Kinase signaling, upon depletion of AEBP1 in U87MG and U138MG cell lines, we did not observe any apparent change in either localization or expression of NF-κB (data not shown). Aebp1 is also known to interact with PI3Kinase inhibitor Pten and promote the process of differentiation in preadipocytes (36). PTEN is often deleted in primary GBM, both the cell lines employed in this study were PTEN mutant (data not shown). Therefore, it is unlikely that this interaction has any significance in gliomagenesis. Apart from signaling, our analysis also yielded an array of molecules related to cell cycle. Cell-cycle regulator CDC20 that is often upregulated in several malignancies (30) and indirectly regulated by p53, is downregulated under AEBP1 downregulated condition, whereas p53 itself is upregulated. Interestingly CDC20 has also been associated with prognostic behavior of glioma patients (37). Again CDK6, an established marker associated with GBM (38, 39), is also affected by AEBP1 downregulation. Other established prognostic markers that have been widely used in GBM such as survin/BIRC5 (40) and MKI67 (41) are also seen in our list of differentially regulated genes. Genes like ITGB1, FZD8, and NGEF belonging to the ontology of differentiation and involved in GBM are also targets of Aebp1. Amplification of MDM2 (5) and hypermethylation of genes such as TES, CDH13 have been reported in glioma (42, 43). Differential expression of TPD52 and CDKN2C in glioma is also reported in literature (44, 45). A wide array of genes involved in the initiation, progression or maintenance of other cancers such as RELN, BIRC3, TNFAIP3, and TNFAIP8 (46–49) can also seen to be differentially regulated upon AEBP1 downregulation. Genes associated with hypoxia namely ARNT, BIRC3, and CTGF were also identified as AEBP1 targets in this study. Another important observation made in this study is that when we compared the perturbed gene list identified upon AEBP1 silencing and those genes, which are perturbed in human primary and secondary glioma tumor samples (9), we observed that EGFR, MDM2, B2M, TEGT, UACA (primary), and CAMK2D (secondary) to be represented in both the gene lists (Table 1). Moreover, EGFR, MDM2, CAMK2D, and UACA have GAAAT motif in their −1 kb promoter regions while B2M and TEGT have TTTCT motif in their −1 kb promoters.

To gain an insight into the functional role of Aebp1 occupancy on the promoter of differentially regulated genes, we compared ChIP-chip data with gene expression data of siRNA-mediated downregulated AEBP1 in U87MG cells. Although 669 annotated genes were affected by AEBP1 downregulation, Aebp1 occupied promoters of only 185 genes. It is important to note that modulation of gene expression may not be because of cis-binding of the transcription factor but its binding kilo bases away from transcription start site can also affect changes in its expression (29). Furthermore, the lack of an absolute correlation between gene expression data and ChIP-chip may be also because of lack of total depletion of Aebp1 by siRNA-mediated silencing. Even low concentrations of Aebp1 can suffice in such downregulated conditions. Also there may be multiple sites of other transcription factors as well as coactivators located far apart, which need to coordinate to propel transcription. Hence, all the 5810 genes whose promoters are occupied by Aebp1 may not be poised for transcription. Thus, it is not surprising that mere promoter occupancy of Aebp1 has not translated into transcriptional perturbation at many of these gene loci. A similar interpretation was proposed in the case of genomic promoter occupancy of RUNX2 in osteosarcoma cells (50).

AEBP1 (−/−) null mice has been earlier studied in the context of adipose tissue metabolism and it is reported that AEBP1 −/− mice display slow growth and suppressed survival with 75% embryonic lethality (36). In this context, perturbation of genes mainly involved in growth-related ontologies under AEBP1 downregulated condition, prompted us to probe the role of AEBP1 in cellular survival. Experiments on cell proliferation and colony suppression assay provided direct evidence for such a role of Aebp1 in U87MG and U138MG glioma cells. Moreover, these cells were directed toward apoptosis as monitored by annexinV staining and DNA fragmentation. An important question that we are presently addressing is what are the key molecule(s) and the pathway that leads to loss of cell viability and initiation of apoptosis? In summary, a comprehensive analysis of the genomic targets of Aebp1 carried out in the present study should form a basic framework for further experimentation on the biological role of AEBP1 in gliomagenesis.

No potential conflicts of interest were disclosed.

This work was supported by grants from NMITLI program of Council of Scientific and Industrial Research (CSIR) and Department of Biotechnology (DBT), New Delhi. MRSR is a JC Bose Fellow of Department of Science and Technology. The gene expression and ChIP-chip data from this study have been submitted to Gene Expression Omnibus (GEO) http://www.ncbi.nlm.nih.gov/geo/under accession no: GSE18892.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.