Limits...
An integrative genomics approach identifies Hypoxia Inducible Factor-1 (HIF-1)-target genes that form the core response to hypoxia.

Benita Y, Kikuchi H, Smith AD, Zhang MQ, Chung DC, Xavier RJ - Nucleic Acids Res. (2009)

Bottom Line: The proximal promoters of these genes were then analyzed for the presence of conserved HIF-1-binding sites.We present experimental validation for ANKRD37 as a novel HIF-1-target gene.Together these analyses demonstrate the potential to recover novel HIF-1-target genes and the discovery of mammalian-regulatory elements operative in the context of microarray data sets.

View Article: PubMed Central - PubMed

Affiliation: Center for Computational and Integrative Biology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA.

ABSTRACT
The transcription factor Hypoxia-inducible factor 1 (HIF-1) plays a central role in the transcriptional response to oxygen flux. To gain insight into the molecular pathways regulated by HIF-1, it is essential to identify the downstream-target genes. We report here a strategy to identify HIF-1-target genes based on an integrative genomic approach combining computational strategies and experimental validation. To identify HIF-1-target genes microarrays data sets were used to rank genes based on their differential response to hypoxia. The proximal promoters of these genes were then analyzed for the presence of conserved HIF-1-binding sites. Genes were scored and ranked based on their response to hypoxia and their HIF-binding site score. Using this strategy we recovered 41% of the previously confirmed HIF-1-target genes that responded to hypoxia in the microarrays and provide a catalogue of predicted HIF-1 targets. We present experimental validation for ANKRD37 as a novel HIF-1-target gene. Together these analyses demonstrate the potential to recover novel HIF-1-target genes and the discovery of mammalian-regulatory elements operative in the context of microarray data sets.

Show MeSH

Related in: MedlinePlus

Prediction strategy for identifying HIF-1-target genes. Candidate genes that respond to hypoxia were first identified by microarrays. Each data set was subjected to a computational analysis in which HIF-binding sites were detected in proximal promoters. Each gene was scored for its response to hypoxia and for the best HIF-binding site. No cutoff was set for determining HIF-target genes. Finally, all genes were ranked.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2724271&req=5

Figure 1: Prediction strategy for identifying HIF-1-target genes. Candidate genes that respond to hypoxia were first identified by microarrays. Each data set was subjected to a computational analysis in which HIF-binding sites were detected in proximal promoters. Each gene was scored for its response to hypoxia and for the best HIF-binding site. No cutoff was set for determining HIF-target genes. Finally, all genes were ranked.

Mentions: Proximal promoters were defined from 700 bases upstream to 300 bases downstream of the transcription start site (TSS) of each transcript. Initially, HIF-1 PWMs were used to identify binding sites; however, we observed a too small overlap between the three HIF-1 PWMs for any given P-value and conservation filtering criteria. Differences in PWM length (12–18 bp) and in the sequences flanking the consensus binding site were found to introduce a bias in HIF-1-binding-site prediction since calculation of P-values and conservation consider the entire length of the matrix and all bases defined within. To avoid this bias all HIF consensus-binding sites sequences (RCGTG) were extracted from each promoter and tested for conservation in other species (see ‘Materials and Methods’ section). A shift from A to G and vice versa in the first position was allowed for the purpose of conservation. Next, the software site-cons in CREAD was employed to calculate the level of conservation at each binding site compared to the flanking 100 bases. A site was predicted as a HIF-binding site if it was conserved in at least 10 species (including human) or if conserved in four species with a site-cons P < 0.05. When the HIF consensus site is preceded by a C, it may form the E-box-binding site CACGTG which is a known binding site for several bHLH transcription factors including HIF-1. In 43% of the E-box sites other TFs were predicted to bind at the same location as HIF-1, compared to only 2% of the non-Ebox sites. The transcription factors that are most commonly predicted at the same location as HIF are ARNT (HIF-1β; as homodimer), AHR:ARNT, MYC, MAX, MYCN, CLOCK:ARNTL and USF1 (Supplementary Figure S2). We only accepted genes with at least one non-Ebox HIF site as HIF-target genes. Throughout this strategy our filtering criteria were very stringent in order to obtain a catalogue of high confidence at the expense of not recovering some known HIF targets. Each gene was scored as follows:The hypoxia/normoxia ratio was inverted and assigned a negative value for genes that were downregulated in hypoxia. Each gene was assigned two ranks, one based on the highest scoring HIF site and one on the hypoxia score. Finally, the sum of both ranks was calculated and all genes were ranked based on that value. For instance, the gene BNIP3, a bona fide HIF-1 target, was ranked 515 by HIF site score and 15 by hypoxia score. Using the sum of ranks BNIP3 was ranked 56 among all genes. This scoring scheme is biased towards genes that respond to hypoxia in multiple cells. While some genes could be HIF targets in only one cell, we reason that a gene that responds to hypoxia in more than one cell is more likely to be a HIF target. This strategy is summarized in Figure 1. Despite the stringent selection criteria of this approach we were able to identify a HIF-binding site in 35/86 promoters of known target genes. A list of previously identified known HIF-1-target genes is shown in Supplementary Table S2.Figure 1.


An integrative genomics approach identifies Hypoxia Inducible Factor-1 (HIF-1)-target genes that form the core response to hypoxia.

Benita Y, Kikuchi H, Smith AD, Zhang MQ, Chung DC, Xavier RJ - Nucleic Acids Res. (2009)

Prediction strategy for identifying HIF-1-target genes. Candidate genes that respond to hypoxia were first identified by microarrays. Each data set was subjected to a computational analysis in which HIF-binding sites were detected in proximal promoters. Each gene was scored for its response to hypoxia and for the best HIF-binding site. No cutoff was set for determining HIF-target genes. Finally, all genes were ranked.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2724271&req=5

Figure 1: Prediction strategy for identifying HIF-1-target genes. Candidate genes that respond to hypoxia were first identified by microarrays. Each data set was subjected to a computational analysis in which HIF-binding sites were detected in proximal promoters. Each gene was scored for its response to hypoxia and for the best HIF-binding site. No cutoff was set for determining HIF-target genes. Finally, all genes were ranked.
Mentions: Proximal promoters were defined from 700 bases upstream to 300 bases downstream of the transcription start site (TSS) of each transcript. Initially, HIF-1 PWMs were used to identify binding sites; however, we observed a too small overlap between the three HIF-1 PWMs for any given P-value and conservation filtering criteria. Differences in PWM length (12–18 bp) and in the sequences flanking the consensus binding site were found to introduce a bias in HIF-1-binding-site prediction since calculation of P-values and conservation consider the entire length of the matrix and all bases defined within. To avoid this bias all HIF consensus-binding sites sequences (RCGTG) were extracted from each promoter and tested for conservation in other species (see ‘Materials and Methods’ section). A shift from A to G and vice versa in the first position was allowed for the purpose of conservation. Next, the software site-cons in CREAD was employed to calculate the level of conservation at each binding site compared to the flanking 100 bases. A site was predicted as a HIF-binding site if it was conserved in at least 10 species (including human) or if conserved in four species with a site-cons P < 0.05. When the HIF consensus site is preceded by a C, it may form the E-box-binding site CACGTG which is a known binding site for several bHLH transcription factors including HIF-1. In 43% of the E-box sites other TFs were predicted to bind at the same location as HIF-1, compared to only 2% of the non-Ebox sites. The transcription factors that are most commonly predicted at the same location as HIF are ARNT (HIF-1β; as homodimer), AHR:ARNT, MYC, MAX, MYCN, CLOCK:ARNTL and USF1 (Supplementary Figure S2). We only accepted genes with at least one non-Ebox HIF site as HIF-target genes. Throughout this strategy our filtering criteria were very stringent in order to obtain a catalogue of high confidence at the expense of not recovering some known HIF targets. Each gene was scored as follows:The hypoxia/normoxia ratio was inverted and assigned a negative value for genes that were downregulated in hypoxia. Each gene was assigned two ranks, one based on the highest scoring HIF site and one on the hypoxia score. Finally, the sum of both ranks was calculated and all genes were ranked based on that value. For instance, the gene BNIP3, a bona fide HIF-1 target, was ranked 515 by HIF site score and 15 by hypoxia score. Using the sum of ranks BNIP3 was ranked 56 among all genes. This scoring scheme is biased towards genes that respond to hypoxia in multiple cells. While some genes could be HIF targets in only one cell, we reason that a gene that responds to hypoxia in more than one cell is more likely to be a HIF target. This strategy is summarized in Figure 1. Despite the stringent selection criteria of this approach we were able to identify a HIF-binding site in 35/86 promoters of known target genes. A list of previously identified known HIF-1-target genes is shown in Supplementary Table S2.Figure 1.

Bottom Line: The proximal promoters of these genes were then analyzed for the presence of conserved HIF-1-binding sites.We present experimental validation for ANKRD37 as a novel HIF-1-target gene.Together these analyses demonstrate the potential to recover novel HIF-1-target genes and the discovery of mammalian-regulatory elements operative in the context of microarray data sets.

View Article: PubMed Central - PubMed

Affiliation: Center for Computational and Integrative Biology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, USA.

ABSTRACT
The transcription factor Hypoxia-inducible factor 1 (HIF-1) plays a central role in the transcriptional response to oxygen flux. To gain insight into the molecular pathways regulated by HIF-1, it is essential to identify the downstream-target genes. We report here a strategy to identify HIF-1-target genes based on an integrative genomic approach combining computational strategies and experimental validation. To identify HIF-1-target genes microarrays data sets were used to rank genes based on their differential response to hypoxia. The proximal promoters of these genes were then analyzed for the presence of conserved HIF-1-binding sites. Genes were scored and ranked based on their response to hypoxia and their HIF-binding site score. Using this strategy we recovered 41% of the previously confirmed HIF-1-target genes that responded to hypoxia in the microarrays and provide a catalogue of predicted HIF-1 targets. We present experimental validation for ANKRD37 as a novel HIF-1-target gene. Together these analyses demonstrate the potential to recover novel HIF-1-target genes and the discovery of mammalian-regulatory elements operative in the context of microarray data sets.

Show MeSH
Related in: MedlinePlus