| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
,
* Sealy Center for Cancer Cell Biology, University of Texas Medical Branch, Galveston, Texas 77555
Department of Human Biological Chemistry and Genetics, University of Texas Medical Branch, Galveston, Texas 77555
Bioinformatics Program, University of Texas Medical Branch, Galveston, Texas 77555
|| Department of Pharmacology and Toxicology, University of Texas Medical Branch, Galveston, Texas 77555
 |
ABSTRACT |
|---|
 TOP ABSTRACT I. IntroductionII. ResultsIII. DiscussionIV. Materials and MethodsACKNOWLEDGEMENTS |
|---|
|
I. Introduction |
|---|
|
TOPABSTRACTI. IntroductionII. ResultsIII. DiscussionIV. Materials and MethodsACKNOWLEDGEMENTS |
|---|
Considerable complexity can be imagined — and, indeed, has been demonstrated — in the interactions between the glucocorticoid receptor and the glucocorticoid response elements with which it interacts. The molecular details of these interactions have been deduced in a number of cases and are actively studied in many laboratories. However, many important questions remain to be addressed concerning the role of glucocorticoid-regulated gene expression in cellular physiology. At the genomic level, it would be of interest to know whether direct interaction of the glucocorticoid receptor with a simple response element, along the model of the MMTV long terminal repeat, is the predominant mechanism for gene regulation, or whether composite-type interactions, involving indirect association with the promoter via other DNA-binding proteins, are more common among primary target promoters. It would be of interest to know whether all primary targets are regulated in all cell types, an idea that seems unlikely. If not, then what features of the promoters predict cell-specific versus ubiquitous responsiveness?
These questions require a complete definition of the primary glucocorticoid targets, ideally in several different cellular and physiological contexts. In addition, a detailed knowledge of the promoter structures of all these target genes will be essential for a complete understanding of how glucocorticoids regulate different genes under different circumstances. We are near to having some of this information in hand. High-throughput gene-profiling technology, although in its infancy, holds the promise of being able to identify all the glucocorticoid target genes. Recent release of a draft of the mouse genomic sequence provides a first opportunity to examine the promoters of primary glucocorticoid response genes.
We have undertaken an initial analysis of glucocorticoid-regulated gene expression in murine T-lymphoma cells. Affymetrix Gene Chips® were used to measure expression of 12,422 genes in cell cycle-arrested cells exposed to dexamethasone (dex) in the presence and absence of cycloheximide. A statistical evaluation of the data was carried out to estimate the level of confidence that one might have in using this technology to identify target genes. The Celera® mouse genomic sequence was used to discover the promoters for the target genes and to ascertain to what extent these promoters contain potential glucocorticoid receptor binding sites. The results we have obtained are provocative but must be considered only a first step in the process of defining the entire repertoire of glucocorticoid-regulated genes in murine T-lymphoma cells. The clearest lesson that we have learned is that the technology is adequate to provide some initial insight into many of the questions raised above. However, several significant limitations in the theory and technology of genomic analysis must be overcome before a more-complete understanding will emerge.
|
II. Results |
|---|
|
TOPABSTRACTI. IntroductionII. ResultsIII. DiscussionIV. Materials and MethodsACKNOWLEDGEMENTS |
|---|
This protocol was repeated three times; the individual experiments are identified as G49, G95, and G116. Within a given experiment, C49 refers to the control sample from the G49 experiment, X116 to the cycloheximide-treated sample for the G116 experiment, and so on.
The ability to identify large numbers of glucocorticoid-regulated genes depends upon the reproducibility of the analytical system. Consequently, our initial efforts focused upon evaluation of the extent to which reproducible results were obtained from replicate experiments. Our first objective was to identify and analyze genes that were not regulated by glucocorticoids. We used two Affymetrix parameters to identify such genes. The Affymetrix MicroArray Suite software (MAS 4.0) reports a number of parameters related to fluorescent intensity of hybridization of labeled RNAs. Average Difference corresponds more or less to the intensity of the signal, whereas Absolute Call is derived from an algorithm that purportedly designates individual RNAs as present, marginal, or absent. These parameters are used in reference to approximately 12,500 probesets, each of which corresponds to a known gene or expressed sequence tag (EST) sequence printed onto the murine MG-U74A gene chips.
An average of 5533 probesets (standard deviation (SD) = 594) were scored as present on each of the 12 chips, with a range of 4232 probesets present (DX95) to 6255 probesets present (D116). We initially excluded probesets that were scored as absent in all four chips from a given experiment (G49, G95, or G116). Spotfire was used to identify probesets that were increased by > 2.0 in dex-treated samples or decreased by > 0.5 in dex-treated samples. Probesets that conformed to these conditions in all three experiments were segregated as potential glucocorticoid-regulated genes. The remaining data were then sorted for probesets that were present on all 12 chips. The result was a dataset of 3170 probesets that were scored as positive on all 12 chips and did not appear to be reproducibly induced or repressed by glucocorticoids. These 3170 probesets formed our control dataset, which we analyzed to determine the amount of random variation in average difference for a given probeset in replicate analyses.
Figure 1A contains the results of three-dimensional (X,Y,Z) linear regression analysis in which the average difference (i.e., roughly the intensity of the hybridization signal) was plotted for each probeset in the control dataset of 3170 probesets from three control samples (C49, C95, and C116). As can be seen from the correlation coefficients (r2), the data exhibited a high degree of correspondence to a linear relationship, with very little scatter around the regression line for the data. This outcome indicates a high degree of reproducibility. More specifically, the data indicate that the average difference measured for a given probeset in one control experiment has a strong predictive value for the average difference of the same probeset in a replicate control experiment. The same relationship was obtained within the dex-treated, cycloheximide-treated, and dex + cycloheximide-treated samples (data not shown).
|
2.5-fold. We also calculated the probability that a control probeset would vary by a given amount in C versus D and X versus DX in any two of three experiments (open circles) or in all three experiments (filled triangles). As can be observed from visual examination of the curves in Figure 1C, the probability that a control probeset will vary in C versus D and X versus DX in two of three experiments approaches zero around 1.8-fold change, whereas the probability is nearly zero that a control probeset will vary by > 1.5-fold in C versus D and X versus DX in all three experiments (filled triangles).
This kind of analysis predicts that when one analyzes an experiment (e.g., G49 or G95) consisting of four chips with average datasets of 5000–6000 probesets present, perhaps 50–100 probesets will vary by > 2.0-fold in both dex and dex + cycloheximide-treated samples. The probability that the same probeset will yield the same outcome, as the result of random variation, in two separate experiments (e.g., that D49/C49 > 2.0 AND DX49/X49 > 2.0 AND D95/C95 > 2.0 AND DX95/X95 > 2.0 or D49/C49 < 0.5 AND DX49/X49 < 0.5 AND D95/C95 < 0.5 AND DX95/X95 < 0.5) would be (0.096)4 or about 8.5 x 10-5. This probability predicts that something on the order of one probeset will, as a result of random variation, change by > 2-fold in both dex-treated and dex + cycloheximide-treated samples in two independent experiment. The probability that a given probeset will behave in this fashion in three independent experiments, as a result of random variation, is 7 x 10-9, far less than one probeset per dataset of 5000 probesets present.
This initial evaluation of the data from three experiments of four chips each suggests two important considerations in the design and interpretation of experiments of this sort. Initially, comparing three sets each of control data — dex, cycloheximide, and dex + cycloheximide-treated — indicates that the results are highly reproducible. In this regard, it should be kept in mind that these three experiments were done over a 6-month period. The second consideration that one must deal with in this kind of experiment is the degree of random variation. Some sense of the degree of random variation must be made if one is to have confidence that the number of replicate experiments is sufficiently great to allow statistically significant conclusions to be drawn. In our experience, random variation, defined as the probability that a given probeset will vary by > 2.0-fold in a pair of control experiments, varies from 14% to 4%, depending on the cells and the conditions under which they are analyzed.
B. IDENTIFICATION AND ANALYSIS OF GLUCOCORTICOID-REGULATED PROBESETS
Our initial analysis of these data was performed using Affymetrix MAS 4.0 software, which has several problems that we needed to overcome. We developed a query, illustrated in Figure 2, to accommodate present/absent calls and negative average differences. Primary-induced probesets were identified as those in which, for a given experiment, C > 0 AND X > 0 AND D/C > 2.0 AND DX/X > 2.0 AND the probeset was scored as present in the D sample AND the probeset was scored as present in the DX sample. The rationale was that we would reject a probeset that appeared to be induced (D/C > 2) if it was scored as absent in the samples in which it was supposed to be induced (D = absent). Using this query, we identified 41 probesets that were induced in each of three experiments. We constructed an additional query (illustrated in Figure 2) to identify probesets that had positive average difference in the D and DX samples (i.e., D > 0 AND DX > 0) AND were scored as present in D and DX AND had negative average differences in C or X (i.e., C < 0 or X < 0). The query was constructed to allow any combination of these conditions. Using this query, we identified three additional probesets that have negative average differences in control and cycloheximide-treated samples but were induced by dex in the presence and absence of cycloheximide in each of three experiments. Thus, we identified 44 probesets that were induced by > 2.0-fold in both the presence and absence of cycloheximide in each of three experiments.
|
50 bp of 5' untranslated region (UTR) in the corresponding human cDNA and the ATG start codon must be in exon 1 of the human gene. In addition, we accepted only those genes that contained four or more contiguous, ordered exons including exon 1. Finally, we accepted only promoters with no string of NNN(N)x (with x being a variable and indeterminate number) between the putative GRE and the transcriptional start site. If no GRE was found and there was no string of NNN(N)x within 4 kb of the transcriptional start site, the promoter was scored as having no GRE. We have attempted to emphasize stringency in our initial analysis, at the expense of excluding some genes that clearly contained potential receptor binding sites but were otherwise ambiguous in their sequence or organization in the Celera draft database.
Forty-four genes were induced by glucocorticoids in each of three independent experiments. Of these, only eight genes corresponded to the strict genomic criteria that we defined to verify promoter structure. Five of these promoters were associated with low-abundance mRNAs: Src-suppressible C kinase (i.e., protein kinase C) substrate (SSeCKS), acid phosphatase 5, RhoB, eIF2a kinase, and phosphatidic acid phosphatase (Figure 3A). Three promoters were associated with high-abundance mRNAs: L29441, 70zpep, and TDAG8 (Figure 3B). TaqMan® probes and primer sets were designed for these eight genes and mRNA abundance was assayed by real-time polymerase chain reaction (PCR). Two internal standards were used, beta-actin and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Triplicate samples were analyzed using the 
|gDCt approach that yields mRNA abundance relative to the internal standard. As shown in Figure 3, the mRNAs corresponding to all eight genes were induced by dex in the presence and absence of cycloheximide.
|
|
|
|
III. Discussion |
|---|
|
TOPABSTRACTI. IntroductionII. ResultsIII. DiscussionIV. Materials and MethodsACKNOWLEDGEMENTS |
|---|
0.1. Likewise, the probability that DX/X > 2.0 or < 0.5 is 0.1. Since these probabilities are multiplicative, the probability that D/C > 2.0 and DX/X > 2.0 or D/C < 0.5 and DX/X < 0.5 is 0.01 for a single experiment. The probability that this condition will prevail for two experiments is 0.1 x 0.1 x 0.1 x 0.1, or 1 x 10-4. For three independent experiments, the probability would be 1 x 10-4 x 0.1 x 0.1 or 1 x 10-6. So, from a statistical perspective, we have a very high degree of confidence that the genes that we have identified will behave in a reproducible fashion. Many of the genes that we have identified contain canonical glucocorticoid response elements within 4 kb of the presumptive transcriptional start site. Five out of eight of the genes that were induced by > 2-fold contained such sequences, whereas two of eight control genes, selected at random, contained presumptive receptor binding sites. If we accept promoters that contain long strings of Ns, we have detected potential receptor binding sites in nine of 13 target genes. More will be said about the reliability of this analysis but one must have a reasonable degree of confidence in assigning as a target any gene that is induced > 2-fold in three independent experiments and contains a canonical glucocorticoid response element within the promoter. For those genes, the majority, for which promoter sequence is unavailable, we must rely on the very high degree of reproducibility of the data. We have confirmed eight out of eight genes by quantitative reverse transcription (qRT)-PCR but this is not a practical approach to screening expression of hundreds of genes. Therefore, it will be necessary to rely heavily on the gene chip data.
At this point, it is appropriate to comment on the Celera mouse genomic database. In our experience, this database must be approached with considerable caution. The filters that we have applied require that the genomic sequence must contain a minimum of four exons in sequence and must be devoid of long strings of unassigned bases (identified by the letter N, hence strings of Ns) within 4 kb of the presumptive start site. Slightly less than 20% of the sequences that we have identified meet these criteria. We feel quite confident in those promoters that we have been able to analyze but we were disappointed that we could not generate a larger and more-reliable dataset. The majority (five of eight) of the induced genes contain MMTV-like glucocorticoid receptor binding sequences. Four of the six presumptive GREs were upstream of the transcriptional start site, none closer than -1 kb, and two were downstream at +1.3 and +3.5 kb, respectively. The GRE sequences were more or less equally associated with the + and - strands of the gene. GRE sequences were found at a much-lower frequency among control genes (2/8). One control gene, nucleolin, contained two GREs downstream of the transcriptional start site and the CD80 gene contained a GRE at -648 bp. We have previously shown that nucleolin expression is inhibited by glucocorticoids in P1798 cells (Suzuki et al, 1992), almost certainly due to a delayed, secondary effect that would not be apparent under the conditions used in the present study. Glucocorticoids do not affect CD80 expression in dendritic cells (Vieira et al, 1998). It will be interesting to determine why these genes are not induced by glucocorticoids. However, it is clear from our analysis that the presence of the sequence ACAnnnTGTnCT is not sufficient to convey induction by glucocorticoids. It remains to be proven that the presumptive response elements that we have identified in five of the eight glucocorticoid-induced genes actually mediate the response. Although we were somewhat surprised that such a high percentage of our glucocorticoid-induced genes contain GREs, we feel that this observation must be interpreted with caution, since it remains to be seen whether this kind of correlation will be maintained as we refine the analysis to include more promoters. We are particularly interested in analyzing promoters that are repressed by glucocorticoids. Unfortunately, we could identify only three such promoters, using the criteria defined previously. None of these contained GREs; however, the sample size is not sufficient for any conclusion to be drawn from this result. For the present, there is not much more that can be made of the database.
Several interesting points can be made with respect to the genes identified in Table I. Two of these, rhoB and the putative G protein-coupled receptor TDAG8, are known to be involved in apoptosis (Choi et al, 1996; Liu et al, 2001) and have been reported to be induced by glucocorticoids in other cell lines (Choi et al, 1996; Koukouritaki et al, 1999). Apoptosis is the normal fate of glucocorticoid-treated T cells. P1798 cells are unusual in that they do not undergo apoptosis when treated with glucocorticoids in medium containing serum growth factors, although such cells die rapidly when treated with glucocorticoids in serum-free medium (Thompson, 1991). We are in the process of examining gene expression profiles in G1-arrested cells exposed to dex in serum-free medium. It is possible that we may, by comparing genomic responses to glucocorticoids in the presence and absence of serum, identify downstream targets of serum growth factors that attenuate the apoptotic response. Such principles could prove to be important therapeutic targets to increase the sensitivity of malignant T cells to glucocorticoid-mediated apoptosis.
One of the target genes in Table I encodes SSeCKS. SSeCKS is a cytoplasmic scaffolding protein (Wassler et al, 2001) involved in nuclear cytoplasmic trafficking of D-type cyclins (Lin et al., 2000). Overexpression of SSeCKS causes G1 arrest (Lin et al, 2000), which is the cellular phenotype of glucocorticoid-treated P1798 cells (Thompson, 1991; Rhee et al, 1995). Glucocorticoids also induce a phospholipid phosphatase, which may be an important target for glucocorticoids during lung maturation (Snyder et al, 1981) and in hepatocytes (Pittner et al, 1985).
Phosphaditic acid phosphatase generates diacylglycerol, the activator of classical and novel isozymes of protein kinase C (PKC). The prototypic classical member of this family is PKC alpha (PKC
), which was induced by glucocorticoids (data not shown). PKC is known to be involved in proliferative control and may be responsible for activation of SSeCKS (Lin et al, 1996). Our data hint at potential cross-talk between protein kinase cascades and nuclear hormone receptor signaling pathways, whereby glucocorticoids stimulate transcription of 1) a phosphatidic acid phosphatase, thereby increasing diacylglycerol synthesis; 2) PKC, which is stimulated by diacylglycerol; and 3) SSeCKS, which, when phosphorylated by PKC, causes cytoplasmic sequestration of D-type cyclins and G1 arrest. This mechanism remains to be proved. It will be necessary to confirm that the relevant proteins are induced but abundant evidence exists that cross-talk between activator protein-1 (AP-1), a critical PKC target, and the glucocorticoid receptor is important in glucocorticoid signaling (Miner and Yamamoto, 1991; Herrlich, 2001). Our data suggest that there may be significant cross-talk between the glucocorticoid and PKC signaling pathways upstream of AP-1.
Kofler and coworkers have published a similar analysis of glucocorticoid regulation of gene expression in proliferating and cell cycle-arrested human CCRF-CEM cells (Tonko et al, 2001). There are several very important differences between their experimental approach and ours. They used Incyte chips, which contain significantly fewer probes than the Affymetrix chips (7074 versus about 12,500). They used a human cell line that responds much more slowly to glucocorticoids and undergoes apoptosis when treated with dex for long periods of time. Although they analyzed genes that were rapidly induced in G1-arrested cells, they did not use cycloheximide to block secondary effects. So, it would not be surprising to find that the genes that they identified are not identical to those we have identified. They identified only eight genes that were either induced or repressed when glucocorticoids were added to both asynchronous and G1-arrested cultures of CCRF-CEM cells. Presumably, these would include — but probably not be limited to — primary transcriptional targets in these cells. Not one of these eight genes was identified in our analysis. Furthermore, only one of the genes listed in Table I (acid phosphatase 5) was identified in any of their analyses.
Brad Thompson recently completed an analysis of CCRF-CEM cells. A preliminary comparison of his results indicates that there are a few genes that are induced in both P1798 and CCRF-CEM cells. However, it seems clear CCRF-CEM cells are very different from P1798 T-lymphoma cells in their glucocorticoid response. Perhaps one should not be surprised, in light of the very different glucocorticoid response phenotypes of these two cell lines: CEM cells die rather slowly but continue to proliferate to a considerable extent in dex, whereas P1798 cells immediately withdraw from the cell cycle and do not die when treated with dex in the presence of fetal bovine serum. However, this result would appear to speak to the question of whether or not there is a subset of primary transcriptional targets that always is regulated in every cell type. The answer to this question awaits additional analysis of glucocorticoid target genes in cell lines and primary cells. The data presented here are an initial step in this direction and represent only a subset of the data that will be required to define the transcriptional targets of the glucocorticoid receptor.
|
IV. Materials and Methods |
|---|
|
TOPABSTRACTI. IntroductionII. ResultsIII. DiscussionIV. Materials and MethodsACKNOWLEDGEMENTS |
|---|
B. GENE CHIP® ANALYSIS
First-strand cDNA synthesis was performed using total RNA (10–25 µg), a T7-(dT)24 oligomer (5' GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG-dT24 3') and SuperScript II reverse transcriptase (Invitrogen). The T7 promoter, introduced during first-strand cDNA synthesis, directed the synthesis of cRNA using bacteriophage T7 RNA polymerase. The cRNAs were labeled with biotin during the T7 transcripton. Biotin-labeled target RNAs were fragmented to a mean size of 200 bases to facilitate their hybridization to probe sequences on the Gene Chip® (Affymetrix) array. Each target RNA sample initially was hybridized to a test array. This array contains a set of probes representing genes commonly expressed in the majority of cells (e.g., actin, GAPDH, hexokinase, 5S rRNA, B1/B2 repetitive elements). Test arrays confirmed the successful labeling of the target RNAs and precluded the use of degraded or nonrepresentative target RNA samples.
Hybridization was performed at 45°C for 6 hours in 0.1 M morpholenoethane sulfonic acid (MES), pH 6.6, 1 M sodium chloride, 0.02 M ethylenediaminetetraacetic acid (EDTA), and 0.01% Tween 20. Four prokaryotic genes (bio B, bio C, and bio D from the E. coli biotin synthesis pathway and cre, the recombinase gene from P1 bacteriophage) were added to the hybridization cocktail as internal controls. These control RNAs were used to normalize expression levels between experiments. Because they are added at varying copy numbers (Bio B, 1.5 pM; Bio C, 5 pM; Bio D, 25 pM; cre, 100 pM), they may be used to estimate relative abundance of RNA transcripts in the sample. Arrays were washed using both nonstringent (1 M sodium chloride (NaCl), 25°C) and stringent (1 M NaCl, 50°C) conditions prior to staining with phycoerythrin streptavidin (10 µg/ml). Gene Chip® arrays were scanned using a Gene Array Scanner (Hewlett Packard) and analyzed using Affymetrix MicroArray Suite 4.0 software.
C. DATA ANALYSIS
Data from individual chips were analyzed separately, using a combination of programs. However, chip-to-chip comparison via MAS 4.0 was not used, since this approach limits comparisons to pairs of samples. Initially, data were imported into Excel files. Affymetrix controls were removed and absolute calls were converted to a numerical value (absent = 0, marginal = 1, present = 2) to facilitate quantitative assessment of presence or absence in multiple samples. The data from three replicates of four chips each were combined in a single file, which was queried using SQL as follows.
For induced probesets from an individual experiment (e.g., G49, G95, or G116):
SELECT *
FROM g49
WHERE (c49 > 0 and x49 > 0 and d49/c49 > = 2 and dx49/x49 > = 2 and d49ac = 2 and dx49ac = 2)
or (c49 < 0 and d49 < 0 and x49 > 0 and dx49/x49 > = 2 and d49ac = 2 and dx49ac = 2)
or (c49 < 0 and d49 > 0 and x49 < 0 and dx49 > 0 and d49ac = 2 and dx49ac = 2)
or (c49 > 0 and x49 < 0 and dx49 > 0 and d49/c49 > = 2 and d49ac = 2 and dx49ac = 2); The following query was used to combine induced probesets in all three experiments:
SELECT *
FROM allthree
WHERE probeset in (select probeset from g49twoac2new)
and probeset in (select probeset from g95twoac2new)
and probeset in (select probeset from g116twoac2new); (Note: g49twoac2new is the result of first query.) The following query was used to select repressed probesets from individual experiments:
SELECT *
FROM g49
WHERE (c49 > 0 and c49ac = 2 and x49 > 0 and x49ac = 2 and d49 < = 0.5*c49 and dx49 < = 0.5*x49);
D. REAL-TIME PCR
Applied Biosystems assays-by-design 20x assay mix of primers and TaqMan® MGB probes (FAMTM dye-labeled) were prepared for all target genes and mouse beta-actin. Primers were designed to span exon-exon junctions, to not detect genomic DNA. All primers and probes sequences were subject to BLAST search against the Celera mouse genome to confirm specificity. TaqMan® rodent GAPDH with VICTMdye-labeled probe also was used as an internal control. The sequences of primers and probes of these genes may be obtained by contacting Huiping Guo (huiguo@utmb.edu). TaqMan® one-step RT-PCR master mix reagent kit was used. A validation experiment was performed to test the efficiency of the target amplification and the efficiency of the reference amplification for all primers and probes. All absolute values of the slope of log input RNA versus CT were < 0.1. Separate tubes (singleplex) one-step RT-PCR was performed using 5 ng of RNA. The cycling parameters for one-step PCR were RT 48°C for 30 minutes, AmpliTaq activation 95°C for 10 minutes, denaturation 95°C for 15 seconds, and annealing/extension 60°C for 1 minute (repeat 40 times) on ABI7700. Triplicate CT values were analyzed using the comparative CT(CT) method, as described by the manufacturer.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTI. IntroductionII. ResultsIII. DiscussionIV. Materials and MethodsACKNOWLEDGEMENTS |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S. Patel, Z. F. Xi, E. Y. Seo, D. McGaughey, and J. A. Segre Klf4 and corticosteroids activate an overlapping set of transcriptional targets to accelerate in utero epidermal barrier acquisition PNAS, December 5, 2006; 103(49): 18668 - 18673. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J.-C. Wang, M. K. Derynck, D. F. Nonaka, D. B. Khodabakhsh, C. Haqq, and K. R. Yamamoto From The Cover: Chromatin immunoprecipitation (ChIP) scanning identifies primary glucocorticoid receptor target genes PNAS, November 2, 2004; 101(44): 15603 - 15608. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
