Comparative analysis of acute and chronic corticosteroid pharmacogenomic effects in rat liver: Transcriptional dynamics and regulatory structures
© Nguyen et al; licensee BioMed Central Ltd. 2010
Received: 4 February 2010
Accepted: 14 October 2010
Published: 14 October 2010
Comprehensively understanding corticosteroid pharmacogenomic effects is an essential step towards an insight into the underlying molecular mechanisms for both beneficial and detrimental clinical effects. Nevertheless, even in a single tissue different methods of corticosteroid administration can induce different patterns of expression and regulatory control structures. Therefore, rich in vivo datasets of pharmacological time-series with two dosing regimens sampled from rat liver are examined for temporal patterns of changes in gene expression and their regulatory commonalities.
The study addresses two issues, including (1) identifying significant transcriptional modules coupled with dynamic expression patterns and (2) predicting relevant common transcriptional controls to better understand the underlying mechanisms of corticosteroid adverse effects. Following the orientation of meta-analysis, an extended computational approach that explores the concept of agreement matrix from consensus clustering has been proposed with the aims of identifying gene clusters that share common expression patterns across multiple dosing regimens as well as handling challenges in the analysis of microarray data from heterogeneous sources, e.g. different platforms and time-grids in this study. Six significant transcriptional modules coupled with typical patterns of expression have been identified. Functional analysis reveals that virtually all enriched functions (gene ontologies, pathways) in these modules are shown to be related to metabolic processes, implying the importance of these modules in adverse effects under the administration of corticosteroids. Relevant putative transcriptional regulators (e.g. RXRF, FKHD, SP1F) are also predicted to provide another source of information towards better understanding the complexities of expression patterns and the underlying regulatory mechanisms of those modules.
We have proposed a framework to identify significant coexpressed clusters of genes across multiple conditions experimented from different microarray platforms, time-grids, and also tissues if applicable. Analysis on rich in vivo datasets of corticosteroid time-series yielded significant insights into the pharmacogenomic effects of corticosteroids, especially the relevance to metabolic side-effects. This has been illustrated through enriched metabolic functions in those transcriptional modules and the presence of GRE binding motifs in those enriched pathways, providing significant modules for further analysis on pharmacogenomic corticosteroid effects.
Glucocorticoids (GC) are a class of steroid hormones present in almost every animal cell, playing a central role in a wide range of physiological responses . Because of their potent anti-inflammatory and immunosuppressive effects, synthetic glucocorticoids referred as corticosteroids (CS) (e.g. methylprednisolone - MPL) have been used widely in pharmacology as a therapeutic option for a wide range of autoimmune and inflammatory diseases [2, 3]. However, beneficial effects are derived from magnifying the physiological actions of endogenous glucocorticoids, causing a variety of side effects following long-term treatment with this class of drugs e.g. hyperglycemia, dyslipidemia, arteriosclerosis, muscle wasting, and osteoporosis [4–7]. The physiological and pharmacological effects of corticosteroids are complex and manifest themselves with expression changes of many genes across multiple tissues [8–10]. It has been observed that even in a single tissue different dosing regimens of CS administration can induce different patterns of expression [11–13]. As such genes with similar expression profiles under acute CS administration may not exhibit similar expression patterns during continuous infusion, pointing to the possibility of alternative regulatory mechanisms. Therefore, a better understanding of corticosteroid pharmacogenomic effects from multiple dosing regimens are very valuable not only to reveal the transcriptional dynamics under different patterns of input perturbations but also to provide an insight into the underlying molecular mechanisms of action, for both beneficial and detrimental effects, and thus for the optimization of clinical therapies.
It has been noted that genes affected by CS include both immunosuppressive genes, mostly associated with therapeutic effects, and metabolic genes often associated with adverse effects whose regulation is mainly controlled by glucocorticoid receptor gene mediated pathways . Unbound CS binds with cytosolic free glucocorticoid receptors (GR) releasing it from the heat shock complex allowing dimerization and translocation into the nucleus where it binds to glucocorticoid response element (GRE) of the target genes, leading to enhancement or inhibition of the target gene expression. As a result, long-term treatment with corticosteroids results in sustained up- or down-regulation of numerous genes, leading to a new steady state which might be the basis for occurrence of adverse effects. However, it has also been noted that chronic infusion of CS causes a sustained down-regulation of the receptor (mRNA and thus protein) [14, 15]. While several alternative mechanisms have been proposed [16–18] it is still not understood why drug effects remain strong although GR mRNA is down-regulated to the point of almost being eliminated. A plausible explanation is that besides direct regulation through GRE binding sites in the 5' regulatory regions of genes, there are changes in expression that are also the indirect results of effects due to changes in expression of transcription factors (TFs) that act as secondary biosignals directly or indirectly modulating the transcription of genes [15, 19, 20]. Thus, along with identification of expression patterns, predicted regulatory control structures are also an essential source of information towards understanding corticosteroid effects.
In this study we address the question as to whether (1) significant transcriptional modules coupled with complex patterns of mRNA changes across multiple dosing regimens of corticosteroids and (2) their common regulatory controls can be computationally identified. Hypothetically, transcriptional modules that are significantly coexpressed under different dosing regimens will be important gene clusters for further analysis towards better understanding of both beneficial and adverse effects of corticosteroids, especially the metabolic side-effects since these patterns are survived under a long-term treatment of corticosteroids. The hypothesis explored here is that if two or more genes have the same temporal expression profiles in response to different dosing regimens, they are more likely to share some common regulatory mechanisms. The liver was selected because of its major role in both the physiological efficacious and adverse effects of corticosteroids e.g. altering the expression of serum proteins that regulate immune/inflammatory responses , enhancing the expression of liver enzymes involved in metabolic effects (gluconeogenesis and lipid metabolism) .
However, rich in vivo datasets of pharmacological time-series across multiple dosing regimens are often obtained from different microarray platforms and time-sets [11, 23], leading to a problematical issue for computational analysis [24–26]. As an example, in a study comparing normal and chronic lymphocytic leukemia B-cells, Wang et al.  identified only 9 differentially expressed genes across all three studies, when combining results from three different platforms, while there are at least 1,172 differentially expressed genes in each individual platform. In general, there are two important issues relevant to the analysis of data derived from different platforms: (i) genes may be present in one platform but not in the other, and (ii) genes present on both platforms may not be represented by the same probes. Since different microarray platforms do not contain the same probesets, and even do not have a similar hardware design and sample processing protocols, standard analyses may not yield comparable expression level quantifications across platforms, leading to many challenges for computational models aiming at the analysis of microarray data from heterogeneous sources [25, 28, 29].
A number of approaches have been proposed and are generally classified into two main categories: (1) integrate raw expression profiles from different studies into one dataset so that available computational models can be directly applied, and (2) develop and/or utilize a unitless statistic as a primary analysis and then combine the result through a meta-level analysis. The former category can be further divided into two sub-classes, namely combining raw data through a normalization and/or transformation procedure [30–33] and pooling raw information from common probes that can be mapped to the same Unigene clusters or full-length mRNA transcripts [34–37]. However, these approaches are not general enough to make data from different platforms fully compatible [25, 38]. Since combining data across different platforms remains a serious challenge, meta-analysis - the second category - has been identified as a more popular technique in order to combine results, and thus data, from a number of independent studies [39, 40]. The assumption here is that while the raw expression levels from different platforms may not be comparable, the results of the primary analysis should be. However, almost all prior studies has focused on the discovery of genes that are differentially expressed in conjunction with standard models such as effect size models [41–43], Bayesian models [44, 45].
Consequently, in order to identify significant clusters of genes that share common expression patterns across multiple dosing regimens, we extend our prior study  in the aspect of (i) producing an agreement matrix (AM) that describes the agreement levels of co-expression of genes across multiple conditions and (ii) successively searching clusterable subsets to infer all such gene clusters. The approach follows the concept of meta-analysis to avoid the limitation of incompatible data across multiple datasets from different platforms (also different tissues, time-grids, as well as lab-protocols when applicable). The unitless statistic, expressing the confidence level of co-expression is the agreement level of cluster assignments drawn from multiple clustering runs. There remain a number of open critical issues associated with a single clustering run (e.g. the input number of clusters [47, 48], the biases and assumptions of distance metrics and/or clustering methods , cluster significance ), and thus consensus clustering coupled with the examination of AM distribution has been designed with the aims of reducing aforementioned limitations [46, 51]. Once the AM is obtained for each condition independently (e.g. each dosing regimen in this case), an average agreement matrix is calculated to estimate the confidence levels of coexpression between genes across multiple conditions, thus combining data from different datasets into a single input for the next analysis. For the analysis at the meta-level, we extend the selection and clustering processes (also proposed in ) to identify all possible clusters of genes that are highly coexpressed with the average AM above as the input. As such these clusters of genes will share common patterns of expression across multiple dosing regimens. Additionally, due to the selection of all possible patterns of expression several clusters may have similar expression patterns and thus we also provide a heuristic as an optional procedure to merge such similar clusters based on a criterion of maximizing the total homogeneity and separation of selected clusters. Subsequently, we analyze promoter regions of genes in every cluster in order to predict putative transcriptional regulators, aiming at providing another source of information towards better understanding those complex patterns of expression and the underlying regulatory mechanisms of corticosteroid effects.
Our results demonstrate that the proposed computational approach is highly effective on both synthetic and real data. When applying the approach to real time-series datasets (acute/chronic corticosteroid administration [11, 23]), selected patterns of transcriptional responses are enriched in a biological sense with relevant putative-regulatory controls and significant metabolic pathways in each pattern. Computational results are further validated predicated upon literature evidence.
A number of synthetic datasets from the open literature are utilized to assess our approach for finding common sets of genes that are highly coexpressed across multiple conditions. Specifically, we used a series of four high-noise 20-timepoint sine-format synthetic datasets with different number of replicates at each time-point (1, 3, 4, and 20 respectively) from [52, 53]. Each dataset contain five separate sets with 400 genes allocated equally in 6 classes, each of which contains the same list of genes but has different patterns across five conditions. For each set, in the first step the data are generated according to an artificial pattern Φ(i, t, l) which shows the values for gene i at time-point t in cluster l; four of six clusters follow the sine function i.e. Φ (i, t, l) = sin(2πt/10 - wl) (wl is some random phase shift between 0 and 2π), and the other two follow the non-periodic linear function (Φ (i, t, 5) = t/20 and Φ(i, t, 6) = -t/20), i = 1,...,400, t = 1,...,20, l = 1,...,4. In the second step, let x(i, t, r) be the error-added value for gene i, time-point t and repeat; x(i, t, r) is randomly drawn from a normal distribution N(μ, σ) where μ is the value of the synthetic pattern Φ(i, t, l) and σ is equal to λσit (σit is randomly extracted from measurement errors observed in the yeast galactose data  and λ is the multiplicative factor that controls the noise level). High-noise synthetic data are generated with λ = 6 . Datasets are downloaded from the links provided in synthetic data in Additional File 1.
Acute corticosteroid administration
Forty-seven male ADX Wistar rats weighting from 225 to 250 g underwent right jugular vein cannulation under light ether anesthesia 1 day before the study . Forty-three rats were injected with a single intravenous bolus dose of methylprednisolone (MPL) of 50 mg/kg. Animals were sacrificed by exsanguinations under anesthesia and liver samples were harvested at 0.25, 0.5, 0.75, 1, 2, 4, 5, 5.5, 6, 7, 8, 12, 18, 30, 48, and 72 after dosing. The sampling time points were selected based on preliminary studies describing GR dynamics and enzyme induction in liver. Four untreated rats were sacrificed at random times and nominally considered as 0 h controls. The gene expression was obtained via the Affymetrix RG-U34A array which consists of 8,799 probesets. The data are publicly available through the GEO Omnibus Database under the accession number GDS253.
Chronic corticosteroid administration
In a similar experiment model, forty rats were given 0.3 mg/kg/hr infusions of MPL over 168 h via an Azlet pump . The pump drug solutions were prepared for each rat based on its predose body weight. Animals were sacrificed at various times up to 7 days; specifically the time-points included are 6, 10, 13, 18, 24, 36, 48, 72, 96, and 168 h. A control group of four animals was implanted with a saline-filled pump and killed at various times throughout the 7-day study period. Unlike the previous experiment, the microarray platform for this dataset is the RAE230A which consists of 15,923 probesets. The data are publicly available through the GEO Omnibus Database under the accession number GDS972.
All protocols followed the Principles of Laboratory Animal Care (Institute of Laboratory Animal Resources, 1996) and were approved by the University at Buffalo Institutional Animal Care and Use Committee.
2. Identifying critical transcriptional modules
The pre-processing step
Each dataset is pre-filtered to identify differentially expressed probesets. Since we would like to identify gene clusters with common expression patterns across multiple conditions, input datasets must contain the same set of genes. Thus using the respective platform information, probesets in each dataset are mapped to a list of genes and then the intersection across those gene lists is evaluated to extract a common set of genes which are differentially expressed across multiple conditions (i.e. datasets). However, genes are sometimes characterized by multiple probesets whose expression profiles may be similar or sometimes different, but not identical. These probesets can be considered as replicates of expression profiles for a gene and thus taking an average expression profiles across all these probesets to characterize for the expression profile of the gene may lose useful potential information. Therefore, from the common set of genes we re-map genes to corresponding probesets in each dataset with the respective platform before starting the analysis.
Construction of the agreement matrix
In addition to the various clustering methods that were utilized, different distance metrics (Euclidean, Pearson correlation, and Manhattan ) are also explored in order to attenuate the biases associated with individual distance metrics. In our implementation, we are using hierarchical clustering (hclust), divisive analysis clustering (diana), fuzzy analysis clustering (fanny), partitioning around medoid (pam) with Pearson correlation and Manhattan metric, k-means (kmeans), fuzzy c-means (cmeans), self-organizing map (som), and model-based clustering (mclust) with Euclidean metric as the core clustering methods (supported by R packages) [56–61]. Since clustering results are highly dependent on the input number of clusters (nc), the sensitivity of the AM as a function of nc was examined to find a 'suggestive' number of clusters (nc*) for each particular dataset. After identifying nc* based on the procedure in our prior work , all clustering runs are repeated with nc* to produce the final AM for further analysis (see more details in ).
As a result, we obtain an agreement matrix whose entries exhibit a quantity that shows how confident genes are coexpressed. This will be the input for the selection and clustering process.
Selection and clustering
With the hypothesis that the more clusterable the data is the more biologically relevant it is, we applied our previously proposed procedure to select a more 'hypothetically clusterable' subset from the entire set of genes . The main hypothesis underlying the selection is that AM entries associated with genes at the 'hypothetical' core of an expression pattern (or a cluster) will be consistently grouped together over multiple clustering runs. This should be manifested by high corresponding values in the AM, whereas genes belonging to the 'hypothetical' core of two clearly distinct clusters are associated with consistently low AM entries. On the contrary, cluster assignments associated with genes at cluster boundaries or between clusters would be very sensitive to the method used and thus they would have relatively moderate agreement levels with other genes. As a result, with a user-defined confidence level δ genes associated with moderate AM entries () are eliminated to produce a more 'clusterable' subset of genes (δ = 70% in this study). The process starts removing genes associated with the highest number of moderate AM entries and then updates the AM for the next loop until no moderate AM entry exists. The corresponding subset of genes is now considered as a 'hypothetically clusterable' subset since any two genes are highly coexpressed or non-coexpressed with the confidence level at least δ. Subsequently, we used the concept of consensus clustering [51, 62, 63] to divide the subset of genes into a number of clusters by applying the hierarchical clustering with the selected AM as input data. The algorithm starts with every gene filling a cluster and then grouping two clusters into a new one for each loop so that any pair of genes belonging to a new cluster always has an agreement level greater than or equal to δ. The iteration is stopped when no more new cluster is formed (see more details in ).
Merging similar patterns
where C is the current set of selected clusters and n is the current number of clusters; H k (C p ) is the homogeneity of cluster Cp in condition k and S k (C p , C q ) is the separation between cluster Cp and Cq in condition k; sim(g ki , g kj ) and dis(g ki , g kj ) are the average similarity and dissimilarity (or distance) respectively between all probesets of gene i and gene j in condition k. Similarity is measured by the Pearson correlation coefficient and dissimilarity is estimated by the Pearson correlation distance.
3. Predicting putative transcriptional regulators
Promoters of genes are extracted from a rich database of promoter information with a default length (500 bp upstream and 100 bp downstream of the transcription start site) if there is no experimentally defined length as suggested by Genomatix . In order to identify putative transcriptional regulators, we explore the basic underlying assumption of comparative genomics which states that functional regions evolve in a constrained fashion and thus at a lower rate than non-functional regions [65, 66]. It implies that conserved regions in a set of orthologous sequences survive due to their special functional implications i.e. TFBSs located on these conserved regions will be more promising as functional binding sites and thus associated TF families are more relevant to our context. Therefore, each promoter is characterized by a set of promoters from orthologous genes of other vertebrate species, if available (e.g. Homo sapiens, Mus musculus, Macaca mulatta, Pan troglodytes, Equus caballus, Bos Taurus, Gallus gallus, etc.). To be consistent in the search for conserved regions on promoter sequences in order to identify putative transcription factor binding sites (TFBSs) we eliminate those that do not consist of more than two orthologous promoters.
Putative functional binding sites
We next apply MatInspector  to scan for all physical TFBSs and only those that overlap with the conserved regions selected above are kept for further analysis. We used a common core similarity 0.75 and utilized the optimal matrix similarity threshold for each position weight matrix (a corresponding profile of TFBSs) suggested from MatBase, Genomatix  which ensure that a minimum number of matches are found in non-regulatory sequences i.e. the false positive matches is minimized. However, a gene may have multiple alternative promoters  and virtually in all cases, it is not known which promoter of the gene is activated. Therefore, all putative TFBSs detected from all alternative promoters of a gene are considered as candidates to infer putative transcriptional regulators for the gene. Subsequently, we estimate the common level of each candidate above in each corresponding module and select those TFBSs present more than a common threshold (70% in this study) (Figure 5b). Associated TF families with those selected TFBSs are inferred and considered as transcriptional regulators for corresponding transcriptional modules.
Results and Discussion
Method evaluation on synthetic data
Effectiveness of the approach on synthetic data
Number of selected genes
Number of clusters
Accuracy* (Adjusted Rand-index)
Effectiveness of the approach on synthetic data
# of sel. genes
# of clusters +
# of sel. genes
# of clusters
# of sel. genes
# of clusters
In general, this approach selects a smaller number of genes with an equal or greater number of cluster structures, resulting in lower accuracy. As an example, in each set of dataset 4 there are two cluster structures that are not clearly identifiable. As a result, a single clustering methods (even consensus clustering) may fail to properly separate them in each set, leading to the situation where the intersection between clusters from set to set divides those cluster structures into many sub-clusters with a small number of genes. On the contrary, by taking the average of the co-expression levels across multiple sets, the relationship of whether two genes are coexpressed across multiple conditions can be recovered. Consequently, our proposed approach is more advantage, resulting in a final highly correct classification as illustrated in Table 1. Furthermore, since this simpler alternative approach produces many resulting clusters, we also attempted to apply the proposed merging process to reduce the number of clusters as well as improve the accuracy if applicable. However, its testing performance is still not as high as that of our proposed approach although we do not apply the merging process for the proposed approach in this test. Additionally, the alternative approach is highly sensitive with the initial number of clusters. For instance, when we constantly set nc* = 7 and test on dataset 3, without the merging option our approach still recovers the correct number of cluster structures with high accuracy: (number of selected genes, number of clusters, accuracy) = (386, 6, 100%) whereas 'pam' approach yields (366, 13, 87.8%), 'mclust' provides (360, 11, 82.3%), and 'consclust' does (351, 7, 98.3%). Since this information is not available for real datasets, the more sensitive with it the less robust the approach is. Therefore, by taking the average of the co-expression levels between two genes across multiple datasets, our proposed approach provides more robust results.
Acute vs. Chronic CS administration
For the analysis of corticosteroid administration, the pre-processing step (Figure 1) is performed to provide corresponding mapped datasets. The datasets are first filtered for differentially expressed probesets using ANOVA technique (p-value < 0.05) implemented in R  and also customized by our previous work for easy uses . 2,920 probesets in the acute and 4,361 probesets in the chronic are selected for further analysis. To obtain the common set of genes across two conditions, these probesets are mapped into sets of genes based on the corresponding platform information. 2,920 differentially expressed probesets in the acute are mapped into a set of 2,340 genes and 4,361 probesets in the chronic are mapped into another set of 4,076 genes. The intersection of these two gene-sets yields 967 genes in common for both dosing regimens. From this common gene set, the re-mapping process subsequently returns a corresponding set of 1,314 probesets for the acute and a set of 1,112 probesets for the chronic data. All datasets (including synthetic data) are pre-processed with the model in our previous study to estimate the 'true' expression profiles that are integrated with potential information in replicates instead of simply taking the average expression profiles . The suggestive number of clusters nc* for both datasets is 7.
Characterization of significant transcriptional modules
Number of genes
Expression pattern in acute*
Expression pattern in chronic*
Similar to module 2 is the transcriptional dynamics exhibited by transcriptional module 4 (71 genes) characterized by an early induction with a maximum at 5.5 h in the acute and 18 h in the chronic. A typical pattern with down regulation for both acute and chronic administration is illustrated by transcriptional module 5 (14 genes). However, genes in the acute regimen exhibited a fluctuated repression with a maximum at around 8 h and then followed by an induction to return to the baseline as late as 72 h. Meanwhile, genes in the chronic regimen characterized a pattern with a slightly transient up-regulation followed by a sustained down-regulation and eventual convergence to a new steady state in the presence of the drug. The last transcriptional module (54 genes) has a similar acute pattern of expression with two phases of regulation as that of module 2. However, in the chronic regimen after falling to a value below the baseline (~24 h) this set of genes was further sustained a slight suppression.
While comparing these expression patterns, we observe that modules 2, 4, and 6 have similar expression patterns in acute (2 & 6) or chronic (2 & 4) with a slight difference in the other dosing regimen (e.g. 2 & 6 in chronic, 2 & 4 in acute). Although the difference is not large enough to be intuitively recognized, the merging process could not merge them together, implying that the difference is significant. Furthermore, the separation of these expression patterns is also reinforced with different functional characteristics which will be illustrated below. In summary, selected transcriptional modules exhibit a number of typical expression patterns under corticosteroid administration. The pattern can be simply expressed as an up- or down- regulation or as a more complex one with two phases of regulation plus some fluctuation (see expression patterns in Additional File 2).
Putative transcriptional regulators of critical transcriptional modules
It has been widely accepted that after corticosteroids bind to cytosolic glucocorticoid receptors (GR), the activated steroid-receptor complex is rapidly translocated into the nucleus where it can alter the expression of target genes. However, the drug seems to be cleared within about 6 h following a bolus injection, suggesting that the mRNA levels of CS-target genes will return to the base line after that . In the contrast, the drug will reach and remain to a stable steady state after 6 h in the chronic administration. Yet, the GR is greatly diminished in response to corticosteroids [14, 15, 19, 20], suggesting that the mRNA levels of CS-target genes in the chronic regimen should also return to the base line. This mechanism is corresponding to the first-phase regulation of target genes. However, almost all chronic patterns involve two phases of regulation and some (module 3 & 5) are only half-phase patterns i.e. persistent up or down without returning to the baseline. These complexities in expression patterns of CS-target genes can be explained by a number of possibilities previous studies have shown [11, 12], including multiple GR isoforms, multiple GREs with different affinities to the drug receptor complex, or some other receptors that can mediate the effect of corticosteroids and thus affected genes in this case can reach a new steady state in the presence of the drug (e.g. module 3 & 5).
However, another possibility is a mechanism that results in the regulation of secondary biosignals which transcription factors are the most potential factors. After affected by corticosteroids, they in turn further modulate the expression of glucocorticoid-regulated genes as a continuing cascade of events that were initiated by the drug. As a result, this possibility suggests a possible interpretation of the complexities in expression changes of multiple CS-target genes with the second phase of regulation (e.g. module 1, 2, 4, and 6). In order to reveal some underlying regulatory mechanism of these selected transcriptional modules, we start analyzing the promoter regions of genes to search for significant putative transcriptional regulators as well as possible relationships of regulation. The hypothesis we explore here is that if two or more genes have similar temporal profiles in response to multiple dosing regimens, they are more likely to share some common regulatory mechanisms.
However, it has been widely recognized that genes affected by CS include both immunosuppressive genes and metabolic genes. Upon the identification of putative transcriptional regulators, their relevance to immune response is demonstrated based on current literature evidence. Specifically, nine among the 29 recognized ETS transcription factors are known to regulate genes involved in immunity ; forkhead transcription factors (FKHD) play a major role in the control of apoptosis ; and especially CREB has been showed as an essential factor for interactions of glucocorticoid receptors to mediate gene expression [78, 79]. A number of others are overlapped with earlier in silico studies e.g. E2FF, EGRF, HOXF, NKXH, SP1F . However, given that the experiment of corticosteroid administration has been studied on normal rats, the relevance to adverse effects may be more important than the relevance to immune response. In fact, almost all enriched functions (gene ontologies, pathways) in these transcriptional modules are relevant to metabolic side-effects (see discussion below). Also, due to this reason NFkB and Ap-1 families widely considered as factors involved in inflammation are not present as direct transcriptional regulators for these sets of genes. Furthermore, we identify a number of transcriptional regulators known to be critical factors in metabolic syndrome including obesity, dyslipidemia, hypertension, insulin resistance, etc. e.g. RXRF , FKHD , SP1F . For instance, the deletion of RXR in mouse liver results in abnormalities of all metabolic pathways regulated by retinoid X receptors heterodimers ; FoxOs, members of FKHD family, are able to increase hepatic glucose production, decrease insulin secretion, and affect glucose or lipid metabolism .
Functional characterization of critical transcriptional modules
Connecting CS transcriptional modules to enriched gene ontology terms (p-value < 0.0001)
Gene Ontology Terms*
Amino acid, compound, organic acid
Cofactor, coenzyme, vitamin, heme, ion
Nucleotide, nucleic acid binding
Cellular catabolic process
Catalytic, oxidoreductase activity
Transmembrane transporter activity
Protein-RNA complex assembly
RNA splicing, processing
Structural molecule activity
Connecting CS transcriptional modules to enriched biological pathways (p-value < 0.01)
Enriched biological pathways
Glycine, serine and threonine metabolism(rno00260)
Bisphenol A degradation(rno00363)
Bile acid biosynthesis(rno00120)
Arachidonic acid metabolism(rno00590)
Pantothenate and CoA biosynthesis(rno00770)
Valine, leucine and isoleucine degradation(rno00280)
Selenoamino acid metabolism(rno00450)
Alanine and aspartate metabolism(rno00252)
Arginine and proline metabolism(rno00330)
Androgen and estrogen metabolism(rno00150)
Androgen and estrogen metabolism(rno00150)
Starch and sucrose metabolism(rno00500)
Urea cycle and metabolism of amino groups(rno00220)
Pentose and glucuronate interconversions(rno00040)
TGF-beta signaling pathway(rno04350)
Wnt signaling pathway(rno04310)
Using ArrayTrack, we also searched for enriched pathways in these transcriptional modules (p-value < 0.01). A large proportion of significant pathways selected in each module are metabolic pathways of amino acid metabolism or biosynthesis, providing another support that selected transcriptional modules are critical and able to capture metabolic side effects for further analysis. Table 5 shows significant pathways in each transcription module.
It is generally accepted that expression levels of many CS-affected genes are mediated through the binding motifs, called GREs - glucocorticoid response elements, on their control regions. We thus examine the presence of this binding site on the promoter of genes in each of the enriched pathways in order to assess the possible effect of GRE of metabolic functions. However, such GREs are short (5-9 bp) and fairly degenerate, leading to matches occurring by chance alone thus not implying any kind of functionality. In order to address this issue, after extracting gene promoters from the Genomatix database we identified conserved regions across sets of orthologous promoters. As a result, those matches located on these conserved regions would be more reliable estimates of functional binding sites.
Although it is currently believed that GREs are composed of two hexamers with a three-nucleotide random-hinge region in between, the general consensus is that towards one hexamer, namely TGTTCT . We therefore search for this motif on conserved promoter regions across orthologous promoters of the selected genes. The results are shown in Table 5 and detailed information is provided in additional files in functional characterization in Additional File 3. In general, almost all metabolic pathways contain genes with the GRE binding sites, implying that these genes are more likely to be directly regulated by the complex between corticosteroids and glucocorticoid receptors. Additionally, we also examine how frequently the GRE binding sites are present on the control regions of all selected genes (315 genes). Furthermore, we determined that given a background set of 2,000 randomly selected genes, the frequency of GREs in a set of genes is similar to that in the random set (~20%), implying that not all genes in those modules are directly regulated by the drug and that the presence of GRE binding sites on the control regions of genes in enriched pathways is very significant and not random.
In summary, we have proposed a systematic computational approach that can identify critical transcriptional modules coupled with their common regulatory controls under the CS administration. The approach provides a framework to handle challenging issues related to different platforms, time-grids, genes with multiple probesets, and also different tissues if applicable. Even if the datasets across multiple conditions are present on the same platform, time-grid and tissue, the approach is still useful since genes contain multiple probesets and estimation of a single gene profile by taking the average across these probesets may lose some useful information. However, the analysis may be limited due to the small common set of genes across different platforms.
The computational effectiveness of the approach has been demonstrated on synthetic data. When applying to real time-series datasets, the approach not only yields critical transcriptional modules but also provides an insight into the complexities of regulation of expression patterns. These complexities are further analyzed by techniques in promoter analysis and functional analysis to deduce useful information of transcriptional regulators and enriched metabolic pathways, providing a better understanding towards regulatory mechanisms and adverse pharmacogenomic effects of corticosteroids.
TTN and IPA acknowledge financial support from the NIH under grant GM082974 and the EPA under grant GAD R 832721-010. RRA, DCD and WJJ acknowledge financial support from the NIH under grant GM 2421.
- Rhen T, Cidlowski JA: Antiinflammatory action of glucocorticoids--new mechanisms for old drugs. N Engl J Med 2005, 353(16):1711–1723. 10.1056/NEJMra050541View ArticlePubMedGoogle Scholar
- Barnes PJ: Corticosteroid effects on cell signalling. Eur Respir J 2006, 27(2):413–426. 10.1183/09031936.06.00125404View ArticlePubMedGoogle Scholar
- Baxter JD: Advances in glucocorticoid therapy. Adv Intern Med 2000, 45: 317–349.PubMedGoogle Scholar
- Bialas MC, Routledge PA: Adverse effects of corticosteroids. Adverse Drug React Toxicol Rev 1998, 17(4):227–235.PubMedGoogle Scholar
- Frauman AG: An overview of the adverse reactions to adrenal corticosteroids. Adverse Drug React Toxicol Rev 1996, 15(4):203–206.PubMedGoogle Scholar
- Schacke H, Docke WD, Asadullah K: Mechanisms involved in the side effects of glucocorticoids. Pharmacol Ther 2002, 96(1):23–43. 10.1016/S0163-7258(02)00297-8View ArticlePubMedGoogle Scholar
- Locsey L, Asztalos L, Kincses Z, Gyorfi F, Berczi C: Dyslipidaemia and hyperlipidaemia following renal transplantation. Int Urol Nephrol 1996, 28(3):419–430. 10.1007/BF02550506View ArticlePubMedGoogle Scholar
- Almon RR, Dubois DC, Jin JY, Jusko WJ: Pharmacogenomic responses of rat liver to methylprednisolone: an approach to mining a rich microarray time series. Aaps J 2005, 7(1):E156–194. 10.1208/aapsj070117View ArticlePubMedPubMed CentralGoogle Scholar
- Almon RR, DuBois DC, Piel WH, Jusko WJ: The genomic response of skeletal muscle to methylprednisolone using microarrays: tailoring data mining to the structure of the pharmacogenomic time series. Pharmacogenomics 2004, 5(5):525–552. 10.1517/146224126.96.36.1995View ArticlePubMedPubMed CentralGoogle Scholar
- Almon RR, Lai W, DuBois DC, Jusko WJ: Corticosteroid-regulated genes in rat kidney: mining time series array data. Am J Physiol Endocrinol Metab 2005, 289(5):E870–882. 10.1152/ajpendo.00196.2005View ArticlePubMedPubMed CentralGoogle Scholar
- Almon RR, DuBois DC, Jusko WJ: A microarray analysis of the temporal response of liver to methylprednisolone: a comparative analysis of two dosing regimens. Endocrinology 2007, 148(5):2209–2225. 10.1210/en.2006-0790View ArticlePubMedPubMed CentralGoogle Scholar
- Almon RR, DuBois DC, Yao Z, Hoffman EP, Ghimbovschi S, Jusko WJ: Microarray analysis of the temporal response of skeletal muscle to methylprednisolone: comparative analysis of two dosing regimens. Physiol Genomics 2007, 30(3):282–299. 10.1152/physiolgenomics.00242.2006View ArticlePubMedPubMed CentralGoogle Scholar
- Yao Z, Hoffman EP, Ghimbovschi S, Dubois DC, Almon RR, Jusko WJ: Mathematical modeling of corticosteroid pharmacogenomics in rat muscle following acute and chronic methylprednisolone dosing. Mol Pharm 2008, 5(2):328–339. 10.1021/mp700094sView ArticlePubMedPubMed CentralGoogle Scholar
- Ramakrishnan R, DuBois DC, Almon RR, Pyszczynski NA, Jusko WJ: Pharmacodynamics and pharmacogenomics of methylprednisolone during 7-day infusions in rats. J Pharmacol Exp Ther 2002, 300(1):245–256. 10.1124/jpet.300.1.245View ArticlePubMedGoogle Scholar
- Sun YN, DuBois DC, Almon RR, Jusko WJ: Fourth-generation model for corticosteroid pharmacodynamics: a model for methylprednisolone effects on receptor/gene-mediated glucocorticoid receptor down-regulation and tyrosine aminotransferase induction in rat liver. J Pharmacokinet Biopharm 1998, 26(3):289–317.View ArticlePubMedGoogle Scholar
- Dong Y, Poellinger L, Gustafsson JA, Okret S: Regulation of glucocorticoid receptor expression: evidence for transcriptional and posttranslational mechanisms. Mol Endocrinol 1988, 2(12):1256–1264. 10.1210/mend-2-12-1256View ArticlePubMedGoogle Scholar
- Oakley RH, Cidlowski JA: Homologous down regulation of the glucocorticoid receptor: the molecular machinery. Crit Rev Eukaryot Gene Expr 1993, 3(2):63–88.PubMedGoogle Scholar
- Vedeckis WV, Ali M, Allen HR: Regulation of glucocorticoid receptor protein and mRNA levels. Cancer Res 1989, 49(8):2295s-2302s.PubMedGoogle Scholar
- Almon RR, DuBois DC, Brandenburg EH, Shi W, Zhang S, Straubinger RM, Jusko WJ: Pharmacodynamics and pharmacogenomics of diverse receptor-mediated effects of methylprednisolone in rats using microarray analysis. J Pharmacokinet Pharmacodyn 2002, 29(2):103–129. 10.1023/A:1019762323576View ArticlePubMedGoogle Scholar
- Sun YN, DuBois DC, Almon RR, Pyszczynski NA, Jusko WJ: Dose-dependence and repeated-dose studies for receptor/gene-mediated pharmacodynamics of methylprednisolone on glucocorticoid receptor down-regulation and tyrosine aminotransferase induction in rat liver. J Pharmacokinet Biopharm 1998, 26(6):619–648. 10.1023/A:1020746822634View ArticlePubMedGoogle Scholar
- Morand EF, Leech M: Glucocorticoid regulation of inflammation: the plot thickens. Inflamm Res 1999, 48(11):557–560. 10.1007/s000110050503View ArticlePubMedGoogle Scholar
- Andrews RC, Walker BR: Glucocorticoids and insulin resistance: old hormones, new targets. Clin Sci (Lond) 1999, 96(5):513–523. 10.1042/CS19980388View ArticleGoogle Scholar
- Jin JY, Almon RR, DuBois DC, Jusko WJ: Modeling of corticosteroid pharmacogenomics in rat liver using gene microarrays. J Pharmacol Exp Ther 2003, 307(1):93–109. 10.1124/jpet.103.053256View ArticlePubMedGoogle Scholar
- Hardiman G: Microarray platforms--comparisons and contrasts. Pharmacogenomics 2004, 5(5):487–502. 10.1517/146224188.8.131.527View ArticlePubMedGoogle Scholar
- Jarvinen AK, Hautaniemi S, Edgren H, Auvinen P, Saarela J, Kallioniemi OP, Monni O: Are data from different gene expression microarray platforms comparable? Genomics 2004, 83(6):1164–1168. 10.1016/j.ygeno.2004.01.004View ArticlePubMedGoogle Scholar
- Pedotti P, t Hoen PA, Vreugdenhil E, Schenk GJ, Vossen RH, Ariyurek Y, de Hollander M, Kuiper R, van Ommen GJ, den Dunnen JT, et al.: Can subtle changes in gene expression be consistently detected with different microarray platforms? BMC Genomics 2008, 9: 124. 10.1186/1471-2164-9-124View ArticlePubMedPubMed CentralGoogle Scholar
- Wang J, Coombes KR, Highsmith WE, Keating MJ, Abruzzo LV: Differences in gene expression between B-cell chronic lymphocytic leukemia and normal B cells: a meta-analysis of three microarray studies. Bioinformatics 2004, 20(17):3166–3178. 10.1093/bioinformatics/bth381View ArticlePubMedGoogle Scholar
- Morris JS, Yin G, Baggerly KA, Wu C, Zhang L: Pooling information across different studies and oligonucleotide microarray chip types to identify prognostic genes for lung cancer. In Methods of Microarray Data Analysis IV. New York: Springer-Verlag; 2005:51–664. full_textView ArticleGoogle Scholar
- Irizarry RA, Warren D, Spencer F, Kim IF, Biswal S, Frank BC, Gabrielson E, Garcia JG, Geoghegan J, Germino G, et al.: Multiple-laboratory comparison of microarray platforms. Nat Methods 2005, 2(5):345–350. 10.1038/nmeth756View ArticlePubMedGoogle Scholar
- Jiang H, Deng Y, Chen HS, Tao L, Sha Q, Chen J, Tsai CJ, Zhang S: Joint analysis of two microarray gene-expression data sets to select lung adenocarcinoma marker genes. BMC Bioinformatics 2004, 5: 81. 10.1186/1471-2105-5-81View ArticlePubMedPubMed CentralGoogle Scholar
- Kim KY, Ki DH, Jeong HJ, Jeung HC, Chung HC, Rha SY: Novel and simple transformation algorithm for combining microarray data sets. BMC Bioinformatics 2007, 8: 218. 10.1186/1471-2105-8-218View ArticlePubMedPubMed CentralGoogle Scholar
- Park T, Yi SG, Shin YK, Lee S: Combining multiple microarrays in the presence of controlling variables. Bioinformatics 2006, 22(14):1682–1689. 10.1093/bioinformatics/btl183View ArticlePubMedGoogle Scholar
- Shabalin AA, Tjelmeland H, Fan C, Perou CM, Nobel AB: Merging two gene-expression studies via cross-platform normalization. Bioinformatics 2008, 24(9):1154–1160. 10.1093/bioinformatics/btn083View ArticlePubMedGoogle Scholar
- Carter SL, Eklund AC, Mecham BH, Kohane IS, Szallasi Z: Redefinition of Affymetrix probe sets by sequence overlap with cDNA microarray probes reduces cross-platform inconsistencies in cancer-associated gene expression measurements. BMC Bioinformatics 2005, 6: 107. 10.1186/1471-2105-6-107View ArticlePubMedPubMed CentralGoogle Scholar
- Lu J, Lee JC, Salit ML, Cam MC: Transcript-based redefinition of grouped oligonucleotide probe sets using AceView: high-resolution annotation for microarrays. BMC Bioinformatics 2007, 8: 108. 10.1186/1471-2105-8-108View ArticlePubMedPubMed CentralGoogle Scholar
- Mecham BH, Klus GT, Strovel J, Augustus M, Byrne D, Bozso P, Wetmore DZ, Mariani TJ, Kohane IS, Szallasi Z: Sequence-matched probes produce increased cross-platform consistency and more reproducible biological results in microarray-based gene expression measurements. Nucleic Acids Res 2004, 32(9):e74. 10.1093/nar/gnh071View ArticlePubMedPubMed CentralGoogle Scholar
- Morris JS, Wu C, Coombes KR, Baggerly KA, Wang J, Zhang L: Alternative probeset definitions for combining microarray data across studies using different versions of affymetrix oligonucleotide arrays. In Meta-Analysis in Genetics. New York: Chapman-Hall; 2006:1–214.Google Scholar
- Kuo WP, Jenssen TK, Butte AJ, Ohno-Machado L, Kohane IS: Analysis of matched mRNA measurements from two different microarray technologies. Bioinformatics 2002, 18(3):405–412. 10.1093/bioinformatics/18.3.405View ArticlePubMedGoogle Scholar
- Ramasamy A, Mondry A, Holmes CC, Altman DG: Key issues in conducting a meta-analysis of gene expression microarray datasets. PLoS Med 2008, 5(9):e184. 10.1371/journal.pmed.0050184View ArticlePubMedPubMed CentralGoogle Scholar
- Ghosh D, Barette TR, Rhodes D, Chinnaiyan AM: Statistical issues and methods for meta-analysis of microarray data: a case study in prostate cancer. Funct Integr Genomics 2003, 3(4):180–188. 10.1007/s10142-003-0087-5View ArticlePubMedGoogle Scholar
- Choi JK, Yu U, Kim S, Yoo OJ: Combining multiple microarray studies and modeling interstudy variation. Bioinformatics 2003, 19(Suppl 1):i84–90. 10.1093/bioinformatics/btg1010View ArticlePubMedGoogle Scholar
- Hu P, Greenwood CM, Beyene J: Integrative analysis of multiple gene expression profiles with quality-adjusted effect size models. BMC Bioinformatics 2005, 6: 128. 10.1186/1471-2105-6-128View ArticlePubMedPubMed CentralGoogle Scholar
- Stevens JR, Doerge RW: Combining Affymetrix microarray results. BMC Bioinformatics 2005, 6: 57. 10.1186/1471-2105-6-57View ArticlePubMedPubMed CentralGoogle Scholar
- Conlon EM, Song JJ, Liu A: Bayesian meta-analysis models for microarray data: a comparative study. BMC Bioinformatics 2007, 8: 80. 10.1186/1471-2105-8-80View ArticlePubMedPubMed CentralGoogle Scholar
- Liang Y, Kelemen A: Bayesian models and meta analysis for multiple tissue gene expression data following corticosteroid administration. BMC Bioinformatics 2008, 9: 354. 10.1186/1471-2105-9-354View ArticlePubMedPubMed CentralGoogle Scholar
- Nguyen TT, Nowakowski RS, Androulakis IP: Unsupervised selection of highly coexpressed and noncoexpressed genes using a consensus clustering approach. Omics 2009, 13(3):219–237. 10.1089/omi.2008.0074View ArticlePubMedGoogle Scholar
- Monti STP, Mesirov J, Golub T: Consensus Clustering: A Resampling-Based Method for Class Discovery and Visualization of Gene Expression Microarray Data. Mach Learn 2003, 52: 91–118. 10.1023/A:1023949509487View ArticleGoogle Scholar
- Yan M, Ye K: Determining the number of clusters using the weighted gap statistic. Biometrics 2007, 63(4):1031–1037. 10.1111/j.1541-0420.2007.00784.xView ArticlePubMedGoogle Scholar
- Belacel N, Wang Q, Cuperlovic-Culf M: Clustering methods for microarray gene expression data. OMICS 2006, 10(4):507–531. 10.1089/omi.2006.10.507View ArticlePubMedGoogle Scholar
- Munneke B, Schlauch KA, Simonsen KL, Beavis WD, Doerge RW: Adding confidence to gene expression clustering. Genetics 2005, 170(4):2003–2011. 10.1534/genetics.104.031500View ArticlePubMedPubMed CentralGoogle Scholar
- Strehl A, Ghosh J: Cluster Ensembles A Knowledge Reuse Framework for Combining Multiple Partitions. Journal on Machine Learning Research 2002, 3: 583–617. 10.1162/153244303321897735Google Scholar
- Medvedovic M, Yeung KY, Bumgarner RE: Bayesian mixture model based clustering of replicated microarray data. Bioinformatics 2004, 20(8):1222–1232. 10.1093/bioinformatics/bth068View ArticlePubMedGoogle Scholar
- Yeung KY, Medvedovic M, Bumgarner RE: Clustering gene-expression data with repeated measurements. Genome Biol 2003, 4(5):R34. 10.1186/gb-2003-4-5-r34View ArticlePubMedPubMed CentralGoogle Scholar
- Ideker TTV, Ranish JA, Christmas R, Buhler J, Eng JK, Bumgarner RE, Goodlett DR, Aebersold R, Hood L: Integrated genomic and proteomic analyses of a systemically perturbed metabolic network. Science 2001, 292: 929–934. 10.1126/science.292.5518.929View ArticlePubMedGoogle Scholar
- Gibbons FD, Roth FP: Judging the quality of gene expression-based clustering methods using gene annotation. Genome Res 2002, 12(10):1574–1581. 10.1101/gr.397002View ArticlePubMedPubMed CentralGoogle Scholar
- Dimitriadou E, Hornik K, Leisch F, Meyer D, Weingessel A: e1071: Misc Functions of the Department of Statistics. R packages 2006.Google Scholar
- Fraley A: mclust: Model-Based Clustering/Normal Mixture Modeling. R packages 2007.Google Scholar
- Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, Ellis B, Gautier L, Ge Y, Gentry J, et al.: Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 2004, 5(10):R80. 10.1186/gb-2004-5-10-r80View ArticlePubMedPubMed CentralGoogle Scholar
- Ihaka R, Gentleman R: R: A Language for Data Analysis and Graphics. J Comp Graphical Statistics 1996, 5(3):299–314. [http://www.R-project.org] 10.2307/1390807Google Scholar
- Maechler M, Rousseeuw P, Struyf A, Hubert M: cluster: Cluster Analysis Basics and Extensions. R packages 2005.Google Scholar
- Yan J: som: Self-Organizing Map. R packages 2004.Google Scholar
- Laderas T, McWeeney S: Consensus framework for exploring microarray data using multiple clustering methods. Omics 2007, 11(1):116–128. 10.1089/omi.2006.0008View ArticlePubMedGoogle Scholar
- Swift S, Tucker A, Vinciotti V, Martin N, Orengo C, Liu X, Kellam P: Consensus clustering and functional interpretation of gene-expression data. Genome Biol 2004, 5(11):R94. 10.1186/gb-2004-5-11-r94View ArticlePubMedPubMed CentralGoogle Scholar
- Doniger SW, Huh J, Fay JC: Identification of functional transcription factor binding sites using closely related Saccharomyces species. Genome Res 2005, 15(5):701–709. 10.1101/gr.3578205View ArticlePubMedPubMed CentralGoogle Scholar
- Hardison RC: Conserved noncoding sequences are reliable guides to regulatory elements. Trends Genet 2000, 16(9):369–372. 10.1016/S0168-9525(00)02081-3View ArticlePubMedGoogle Scholar
- Morgenstern B: DIALIGN 2: improvement of the segment-to-segment approach to multiple sequence alignment. Bioinformatics 1999, 15(3):211–218. 10.1093/bioinformatics/15.3.211View ArticlePubMedGoogle Scholar
- Schmollinger M, Nieselt K, Kaufmann M, Morgenstern B: DIALIGN P: fast pair-wise and multiple sequence alignment using parallel processors. BMC Bioinformatics 2004, 5: 128. 10.1186/1471-2105-5-128View ArticlePubMedPubMed CentralGoogle Scholar
- Pollard DA, Bergman CM, Stoye J, Celniker SE, Eisen MB: Benchmarking tools for the alignment of functional noncoding DNA. BMC Bioinformatics 2004, 5: 6. 10.1186/1471-2105-5-6View ArticlePubMedPubMed CentralGoogle Scholar
- Cartharius K, Frech K, Grote K, Klocke B, Haltmeier M, Klingenhoff A, Frisch M, Bayerlein M, Werner T: MatInspector and beyond: promoter analysis based on transcription factor binding sites. Bioinformatics 2005, 21(13):2933–2942. 10.1093/bioinformatics/bti473View ArticlePubMedGoogle Scholar
- Singer GA, Wu J, Yan P, Plass C, Huang TH, Davuluri RV: Genome-wide analysis of alternative promoters of human genes using a custom promoter tiling array. BMC Genomics 2008, 9: 349. 10.1186/1471-2164-9-349View ArticlePubMedPubMed CentralGoogle Scholar
- Hubert L, Arabie P: Comparing partitions. J Classification 1985, 2(1):193–218. 10.1007/BF01908075View ArticleGoogle Scholar
- Nguyen TT, Almon RR, DuBois DC, Jusko WJ, Androulakis IP: Importance of replication in analyzing time-series gene expression data: Corticosteroid dynamics and circadian patterns in rat liver. BMC Bioinformatics 2010. (accepted) (accepted)Google Scholar
- Pavlidis P: Using ANOVA for gene selection from microarray studies of the nervous system. Methods 2003, 31(4):282–289. 10.1016/S1046-2023(03)00157-9View ArticlePubMedGoogle Scholar
- Rodriguez-Caso C, Medina MA, Sole RV: Topology, tinkering and evolution of the human transcription factor network. Febs J 2005, 272(24):6423–6434. 10.1111/j.1742-4658.2005.05041.xView ArticlePubMedGoogle Scholar
- Gallant S, Gilkeson G: ETS transcription factors and regulation of immunity. Arch Immunol Ther Exp (Warsz) 2006, 54(3):149–163. 10.1007/s00005-006-0017-zView ArticleGoogle Scholar
- Coffer PJ, Burgering BM: Forkhead-box transcription factors and their role in the immune system. Nat Rev Immunol 2004, 4(11):889–899. 10.1038/nri1488View ArticlePubMedGoogle Scholar
- McKay LI, Cidlowski JA: CBP (CREB binding protein) integrates NF-kappaB (nuclear factor-kappaB) and glucocorticoid receptor physical interactions and antagonism. Mol Endocrinol 2000, 14(8):1222–1234. 10.1210/me.14.8.1222PubMedGoogle Scholar
- Sulser F: The role of CREB and other transcription factors in the pharmacotherapy and etiology of depression. Ann Med 2002, 34(5):348–356. 10.1080/078538902320772106View ArticlePubMedGoogle Scholar
- Hutton JJ, Jegga AG, Kong S, Gupta A, Ebert C, Williams S, Katz JD, Aronow BJ: Microarray and comparative genomics-based identification of genes and gene regulatory regions of the mouse immune system. BMC Genomics 2004, 5(1):82. 10.1186/1471-2164-5-82View ArticlePubMedPubMed CentralGoogle Scholar
- Shulman AI, Mangelsdorf DJ: Retinoid x receptor heterodimers in the metabolic syndrome. N Engl J Med 2005, 353(6):604–615. 10.1056/NEJMra043590View ArticlePubMedGoogle Scholar
- Nakae J, Oki M, Cao Y: The FoxO transcription factors and metabolic regulation. FEBS Lett 2008, 582(1):54–67. 10.1016/j.febslet.2007.11.025View ArticlePubMedGoogle Scholar
- Solomon SS, Majumdar G, Martinez-Hernandez A, Raghow R: A critical role of Sp1 transcription factor in regulating gene expression in response to insulin and other hormones. Life Sci 2008, 83(9–10):305–312. 10.1016/j.lfs.2008.06.024View ArticlePubMedGoogle Scholar
- Wan YJ, An D, Cai Y, Repa JJ, Hung-Po T, Flores M, Postic C, Magnuson MA, Chen J, Chien KR, et al.: Hepatocyte-specific mutation establishes retinoid X receptor alpha as a heterodimeric integrator of multiple physiological processes in the liver. Mol Cell Biol 2000, 20(12):4436–4444. 10.1128/MCB.20.12.4436-4444.2000View ArticlePubMedPubMed CentralGoogle Scholar
- Tong W, Cao X, Harris S, Sun H, Fang H, Fuscoe J, Harris A, Hong H, Xie Q, Perkins R, et al.: ArrayTrack--supporting toxicogenomic research at the U.S. Food and Drug Administration National Center for Toxicological Research. Environ Health Perspect 2003, 111(15):1819–1826.View ArticlePubMedPubMed CentralGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.