| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Nutritional Sciences and Toxicology, University of California, Berkeley, California
2 Gene Array Technology Center, Department of Medicine
3 Neurogenomic Laboratory, Pain Research Center, Department of Anesthesiology, Perioperative and Pain Medicine, Brigham and Women’s Hospital
4 Harvard Medical School, Boston, Massachusetts 02115
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
opioids; cDNA microarray; regression analysis; cluster analysis; spinal cord; striatum; skeletal protein
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
MATERIAL AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
All animal protocols were approved by the Animal Care Committee at Harvard Medical School, Boston, MA.
Analgesia
Analgesia was assessed by the tail immersion assay at a water temperature of 52°C as described elsewhere (35). The mouse tail was immersed in the heated water. The latency to a rapid flick of the tail was recorded as the withdrawal latency. After drug administration, we used a maximum cutoff of 15 s to minimize tissue damage.
Naloxone-Precipitated Jumping
Naloxone-precipitated jumping was determined by observation of the animal in a 3-liter glass cylinder containing 2–3 cm of bedding for 15 min following naloxone injection as described elsewhere (34). To confirm that even a single morphine injection produces an acute withdrawal reaction (28), the separate group of mice was injected with naloxone (15 mg/kg ip) 2 h following morphine administration. The response to naloxone was considered positive if the mouse jumped at least four times.
Tissue Dissection and RNA Isolation
To prepare the total RNA for array hybridization, morphine-injected (M group) or saline-injected (S group) mice were killed 30 or 120 min after injection, and morphine+ naloxone-injected mice (N group) were killed 30 min after injection. Thus there were five groups of at least six mice each: M30, M120, S30, S120, and N30. After decapitation, the brain was quickly removed and placed in mouse brain matrix (TedPella) with division 1 mm, cooled on ice. The prefrontal cortex was removed at the level of about 2.0 mm; the distances are given from the Bregma line according to the stereotaxic atlas of mouse brain (7). The next section cut at 0.5 mm (just in front of the optical chiasma) was used to dissect the medial striatum with nucleus accumbens by a diagonal cut just above the anterior commissure. The spinal cord was removed under hydraulic pressure using a 10-ml syringe with saline (21), and the lumbar enlargement was saved. Tissues were frozen and kept at -80°C until RNA isolation. (We have found that it is better to keep the brain tissue rather than extracted RNA; even at -80°C, extracted RNA slowly deteriorates.) The tissue was homogenized at high speed in Trizol (Life Sciences), and total RNA was isolated according to the manufacturer’s instructions. The concentration of total RNA was measured by ultraviolet (UV) spectrophotometry at 260/280 nm, and RNA quality was assessed by electrophoresis on a 1% agarose gel. Only those samples with a 260/280 ratio more than 1.9 and no signs of degradation based on agarose electrophoresis were used for analysis.
RNase Protection Assay
RNase protection assay was done as described elsewhere (1) with some modifications. In brief, 32P-labeled antisense riboprobes were synthesized with T7 or SP6 RNA polymerase from linearized plasmids and purified by electrophoresis on polyacrylamide gels. Specific activity of the labeled probes was 
1–3 x 108 cpm/µg. The aliquots of the sample (2 µg of total RNA) were incubated with a molar excess of 32P-labeled antisense RNA (105 cpm) for 16–20 h at 55°C in 30 µl of the hybridization buffer (80% formamide, 40 mM PIPES, 1 mM EDTA, and 400 mM NaCl, pH 6.7). The nonhybridized antisense RNA was digested with RNase A and T1. The samples were treated with SDS + proteinase K (0.6% and 290 µg/ml, respectively) for 15 min at 37°C followed by phenol:chloroform:isoamyl alcohol extraction and ethanol precipitation; these were resuspended in the 80% formamide, 10 mM Tris, and 10 mM EDTA, pH 7.0, heat denaturated, and electrophoresed on a polyacrylamide gel (5% polyacrylamide and 7% urea). Gels were exposed to film (Biomax, Kodak) and quantified by image analysis software (Imaging Research)
RNA Labeling
Total RNA (10–60 µg, depending on type of experiments) was labeled with a corresponding fluorescent nucleotide (Cy3- or Cy5-labeled dUTP) as described elsewhere (6), with some modifications. In brief, total RNA was mixed with oligo dT18 primers (20 µM) and incubated at 72°C for 5 min and then quickly cooled on ice. In some experiments, before heating, control RNA from Escherichia coli with poly-A linker (pGIBS-Phe from ATCC) was added at molar ratios between 1:100 and 1:50,000 (assuming a 1.0-kb average length for mRNAs) to evaluate the efficiency of reverse transcription. Unlabeled nucleotides (dCTP, dATP, and dGTP) in final concentrations of 0.5 mM, 200 U of reverse transcriptase (RNase H-) and its corresponding buffer, and labeled dUTP in final concentration 0.2 mM were added. The reaction mixture was incubated for 5 min at 25°C and then for 1 h at 42°C in the dark; 200 U more of reverse transcriptase (RNase H-) was added, and the reaction continued for another 50 min. After incubation, the sample was heated at 95°C for 3 min to denature cDNA/mRNA hybrids and chilled on ice immediately. RNA was hydrolyzed by the addition of 5 µl of 0.5 M EDTA and 5 µl of 0.2 M NaOH, incubated at 60°C for 10 min, and cooled to room temperature; 5 µl of 1 M Tris·HCl (pH 7.5) and 5 µl of 0.2 N HCl were added to neutralize the reaction. The sample was purified from unlabeled nucleotide on a Sephadex G50 minicolumn, and the eluate was kept frozen at -20°C until hybridization.
Hybridization to Array
DNA microarray analysis of gene expression was done essentially as described (6) with some modifications. In brief, for each experiment, fluorescent cDNA probes were prepared from two samples and labeled with Cy5- or Cy3-dUTP. In this study we have used Cy5-dUTP to label samples from morphine-injected animals and Cy3-dUTP for samples from saline- or morphine+naloxone-injected mice, to enable comparison of the samples on the same slide. Cy3- and Cy5-labeled cDNAs were mixed and dried on Speedvac (Savant, Farmingdale, NY) and dissolved in hybridization buffer (Tris·HCl, 20 mM, 0.04% Denhardt’s solution, 0.02% salmon sperm DNA, 0.01% poly-A DNA, 20% dextran sulfate, 0.01% SDS, and 50% formamide). Prior to application to the array, the solution was heated at 85°C for 4 min, chilled on ice, and centrifuged at 14,000 g for 1 min. Hybridization was performed in a humidified chamber. Twenty microliters of hybridization solution was added to each DNA array and covered with a 22-mm coverslip. Slides were placed in a humidified chamber and incubated overnight at 48°C. Following hybridization, coverslips were removed in a glass staining dish containing 2x SSC with 0.1% SDS. Slides were washed twice for 4 min with the same solution, washed in 1x SSC for 10 min at 65°C, rinsed twice with 0.1x SSC for 2 min, rinsed with water twice for 1 min, and quickly dried by filtered air. Arrays were usually immediately scanned with a GMS 418 scanner (Affymetrix); however, we found that an array may be kept for 2–3 days in the dark without loss of fluorescence intensity.
Preparation of cDNA Array
cDNA fragment preparation.
The cDNA array contains 1,667 cDNA fragments and 61 spots used as a negative control (3 bacterial cDNA in duplicate and the spots with primers used for fragment amplification, buffer, and other solutions used in fragment preparation). Some cDNA fragments were different sequences of the same gene. About 1,100 of the cDNA fragments were prepared from a sequenced mouse brain cDNA library arrayed in 96-well plates (kindly provided by Dr. D. Beier, Genomic Center, Brigham and Women’s Hospital). About 600 fragments were prepared by subtractive cloning (32). Using a representational difference analysis (17), we prepared libraries from cortex, striatum, medulla oblongata, spinal cord, and thalamus enriched in genes altered during tolerance and abstinence. The original methodology was modified to increase the probability of cloning genes that are slightly different in target and driver populations. Amplification of PCR fragments from the plasmid vector employed a T7 and SP6 primer pair or a T3 and T7 primer pair. The PCR reaction contained dNTP mix (1 mM), primer (0.5 µm), 1–2 ng of plasmid, and 1 U of Taq polymerase for 100 µl of reaction. Plasmids were amplified in a 96-well plate. Cycling protocol was as follows: 4 h 95°C-1 cycle, 45 min at 56°C; 2.5 h at 73°C, 50 min at 95°C; 38 cycles. The PCR reaction product was purified by ethanol precipitation, and the presence of fragments was verified by agarose gel electrophoresis. The concentration of fragments after PCR was 0.1–0.2 µg/µl.
cDNA arraying.
Amplified cDNA fragments were arrayed on a poly-L-lysine-coated slide with GMS 417 (Affymetrix). After arraying, slides were rehydrated, dried, UV cross-linked, and blocked as described (6). Arrays were kept at room temperature and used within 2 mo after preparation.
Data Analysis
Scanning parameters (light intensity and detection levels) were set such that the brightest spots are just below saturation. Images were analyzed with ArrayVision (Imaging Research), and fluorescence intensity values (along with quality control parameters) were stored in a database. Spots with obvious defects were tagged and excluded from subsequent analyses. Nontagged array elements were considered positive when the fluorescence intensity in each channel was greater than 1.9–3 times the local background (the threshold ratio for each slide was chosen such that negative control spots were at the background level). After background subtraction, the two channels were normalized to channel-specific median signal intensity. This approach compensates for possible biases in intensity resulting from inefficiencies in sample extraction and labeling and is less sensitive to outliers than normalization by total signal intensity. Fluorescence intensity values were log-transformed (base 2) and stored in a single table. The replicates of identical samples were analyzed for homogeneity (two-way nonparametric ANOVA by Friedman) and then averaged for regression analysis. The analysis of data was based on the combination of two procedures: regression and self-organizing map (SOM) clustering. This regression/clustering approach combines the statistical power of regression analysis and the exploratory strength of SOM-based clustering. Up- and downregulated genes were detected by regression analysis (18) with robust and resistant linear regression (MM-estimator) and simultaneous prediction confidence intervals (SPCIs) (20). The basic assumption of this model is that genes whose expression levels significantly deviate from the regression line (outliers) are the most likely to be up- or downregulated. The model is based on the fact that intensity values in two channels for a given set of spots are highly correlated (18, 26), if there are no print-tip and/or spatial effects (5). Since the standard error of replicates is smaller than intensity value range, we can use the regression analysis to estimate the confidence interval (4, 18). The MM-estimator was used to build linear regression, since an ordinary least-squared linear regression approach is sensitive to outliers itself. The SPCIs were used to provide a desirable confidence level across the whole range of intensity values and to estimate P values for up- and downregulated genes.
The results of regression analysis were combined with SOM clustering. SOM was applied by the weighted pair-group method with centered average and Pearson correlation implemented in the program xCluster (kindly provided by Gavin Sherlock, Stanford University). The clusters were modified with the subset of genes selected by regression analysis so that P values for the differentially expressed genes are less than or equal to 0.05, at least in one pair used to build the corresponding scatter plot for regression analysis.
To organize selected genes according to the intracellular pathways involved, we used a context-based information search and graphical tools (GeneSphere, Boston, MA).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
The use of amplified RNA (by reverse transcribing RNA using T7 polymerase promoter as a primer) may decrease the required amount of total RNA to 0.5 µg. However, it distorted the RNA expression profile: after amplification, the level of expression of at least 450 genes out of 1,667 was altered more than twofold. Amplification also affected linearity and decreased the dynamic range of detection (data not shown). Therefore, in this experiment we have not used amplification.
In a separate experiment, we tested the detection range of the array using different concentrations of pGIBS-Phe from E. coli with an attached poly-A tail. In a range of 1:100 to 1:50,000, the intensity of signal was linearly dependent on the amount of RNA used (r = 0.94, P < 0.05).
There is not yet a theoretically and experimentally verified best design for two-channel array experiments, particularly where multiple comparisons are involved. Currently there are two variants, one based on the cross-reference sample, which is labeled with one dye, while experimental and control samples are labeled with another dye, and a second design, that also includes two dyes, one for experimental and another for control samples. It seems that the first type of design poses significant difficulties for the analysis of nervous tissue in which different brain areas are involved. The great variability of gene expression between brain areas makes a reference sample practically useless (32). Therefore, in this study there were Cy5-labeled cDNA (in triplicate) from morphine-injected mice and Cy3-labeled cDNA from saline- and morphine+ naloxone-injected animals (in duplicate), for a total of 24 samples hybridized to 12 slides. Because this design does not automatically compensate for the difference between Cy3 and Cy5 properties, we independently normalized the data from the Cy5 and Cy3 channels prior to further analysis on the basis of median intensity. The replicates were homogenous according to two-way nonparametric ANOVA by Friedman and thus, for further analysis, were averaged. After removing all genes that lacked at least one value (to perform logarithmic transformation), the final set contained 522 rows and 10 columns corresponding to five groups (S30, N30, S120, M30, and M120) and two brain areas [5 from medial striatum (s) and 5 from lumbar part of the spinal cord (d)].
Regression Analysis
We used scatter plots and regression analysis to detect up- and downregulated genes in different groups within each brain area. The common practice of selecting up- and downregulated genes based on the predetermined fold differences in the expression assumes that variance of ratios is a constant for the microarray of interest. However, because of between-slide variability, a ratio value, e.g., 2, may be statistically significant for one slide and nonsignificant for another; therefore, the use of fold difference without appropriate statistical analysis may be misleading. One way to solve this problem is to collect more replicates and to apply ANOVA and a multiple comparison procedure. The successful use of replication and estimates for variation of intensity to reduce misclassification rates has been described in literature (5, 13, 16). MM-estimator, a robust and outlier-resistant method of regression analysis (see MATERIALS AND METHODS, Figs. 2 and 3) used in this study, is less sensitive to data distribution and more economical (compare with use of multiple replicates) but less accurate from a statistical point of view. Nevertheless, it enables us to determine the statistical significance of the observed differences in gene expression. Using the SPCIs (see MATERIALS AND METHODS), we were able to estimate the probability that a given gene is differentially expressed, i.e., that the gene is up- or downregulated. A total of 12 pairs were compared on the basis of three factors: time (30 and 120 min after injection), drug (S, M, and N groups), and brain area (s, medial striatum; and d, spinal cord; Fig. 2). Significance of regression coefficients varied for different scatter plot pairs but in all cases P < 0.0001 at least, except for scatter plot dM120 vs. dS120 where P = 0.0127.
|
|
0.05. The 100 gene fragments that had such a P value in at least one pair used to build a scatter plot for regression analysis were selected (Fig. 4). The number of unique genes is fewer, since some fragments were different sequences of the same gene (see below). As expected, there is an inversely proportional correlation between the ratio magnitude and its statistical significance; i.e., the higher the ratio the less its P value. However, as mentioned above, there is no single threshold to apply to the entire data set. Thus, for gene g1194, the ratio = 2.68 from the scatter plot sM30 vs. sS30 and the ratio = 1.3 from the scatter plot sM30 vs. sN30 have P values in the same range (between 0.01 and 0.05), whereas for gene g1191 the ratio = 1.94 from the scatter plot dM30 vs. dN30 is more statistically significant (0.001 < P 0.01) than the ratio = 3.43 from the scatter plot sM30 vs. sS30 (0.01 < P 0.05). Moreover, the ratio = 0.72 (g1547, 0.01 < P 0.05; Fig. 4) is usually considered too small to be an indicator of alteration in gene expression, and the important difference may be lost. The approach based on MM-estimator and SPCIs is able to detect these genes as differentially expressed, since it takes into account variability characterized by residual error of linear regression and specific to a given comparison. The fact that 80% of all selected genes were significantly altered in more than one pair further confirms the exploratory power of the approach. However, the interpretation of the results even for this relatively simple experimental design is constrained by the fact that we did not take explicitly into account the variability between replicates in expression value specific for a given gene. We use a relatively high P value (P = 0.05) for gene selection; therefore, our analysis may have high misclassification rates (i.e., the relative numbers of false-positive cases may be high). However, it enables us to complement the regression analysis used to detect differentially expressed genes with cluster analysis (SOM), which by itself has no statistical value and often may lead to spurious conclusions.
|
|
|
|
|
|
-actin, cofilin, SPARC, and csk, which negatively regulates actin filament formation), microtubule interacting proteins (tubulin alpha, stathmin), and a number of genes whose products are involved in secretory vesicle formation and cell adhesion (Table 1). We include in this group the sodium channel ß1-subunit, since it may function as an adhesion molecule (19). Based on this classification and analysis of available publications, we tried to reconstruct the pathway modulated by opioid receptor activation. There is no exact match with known intracellular pathways, and the closest match is a new "isotype" of integrin-activated skeletal protein regulatory pathway (Fig. 9).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
However, the reliability of microarray data is still uncertain, and the interpretation of the results is confounded by potential methodological problems. Therefore, we used a new procedure based on regression analysis and clustering to analyze the gene expression data. The regression analysis provides a statistical approach to selecting up- and downregulated genes, whereas SOM-based cluster analysis adds the power of pattern recognition to select groups of coregulated genes. This hybrid procedure (clustering followed by regression analysis) seems to combine the advantages of both methods. However, this simple and empirically useful approach requires further theoretical analysis to understand its validity and limitations. Data preprocessing plays an important role in our approach: to safeguard against spurious results, we included in the analysis only spots that were positive in all samples. In this way useful information can be lost, since most variable genes in the low-intensity area are removed. A possible alternative of including in the analysis even highly variable genes (a compendium approach as described in Ref. 11) requires many more replicates and may not be practical for certain experiments. By using the hybrid approach (regression clustering), we detected two predictable effects of morphine on gene expression: inhibition and activation. Both of these effects are probably mediated by specific opioid receptors, since naloxone coinjection prevents them. The unexpected finding was that only 9 genes were upregulated and 45 genes were downregulated. Currently we do not have a simple explanation for this phenomenon. It might be a consequence of the general inhibitory effect of opioids on metabolic function, since it is known that morphine decreases temperature, respiration, heart rate, and heat production. Whether the alteration of these physiological parameters is directly connected to the metabolic rate of neurons is not known but seems unlikely, since neurons are generally protected from fluctuation of metabolic rates. The changes in mitochondrial protein expression may still be secondary to the general depressant effect of morphine on metabolism. The observed changes, however, were prevented by the opioid receptor antagonist naloxone and limited to the specific subset of genes that belong to distinct intracellular pathways. It was also reported that a single morphine injection decreased mRNA turnover in the brain but not in the liver (9). The relatively large number of altered genes is not an argument in favor of general metabolic inhibition, since part of the gene fragments on array were preselected based on their potential changes following morphine administration, and 75% of all altered genes belong to this group.
The activation of the opioid receptor inhibits cAMP levels and ion currents via pertussis toxin-sensitive and -insensitive Gi proteins. The decrease of cAMP and/or Ca2+ concentrations may inhibit gene transcription, but it is unclear how to explain the particular effect of morphine observed in this study. We do not know whether any agent that decreases the level of cAMP or Ca2+ will produce a similar effect, nor are we sure how information to induce a specific gene expression pattern in response to extracellular receptor activation is transmitted and coded. The ratio of cAMP, diacylglycerol, and ion concentrations may define the specific alterations in gene expression in response to opioid receptor activation. Yet, we cannot exclude other more sophisticated mechanisms. At this point there is not sufficient sequence information to decide whether the altered genes have similar promoter regions, but certainly the data might be reanalyzed at a later time.
Morphine decreases the expression of multiple genes following repeated administration (22), but there are limited data regarding its acute effect. Nevertheless, these data confirm morphine’s ability to alter mRNA levels even following a single injection (2, 3, 8, 14, 15, 33).
Functionally, the altered genes can be divided into three groups (Table 1). One includes proteins involved in mitochondrial functions: oxidative phosphorylation, glycolysis, and transport into mitochondria. Another group are the proteins involved in broadly defined cytoskeletal functions: synaptic vesicle formation from the Golgi complex and its fusion to membrane, axon growth, and adhesion. The third more heterogeneous group consists of regulatory proteins. The observed changes in gene expression are consistent with the fact that activation of the opioid receptor reduces neuromediator secretion, axon formation, and protein synthesis (9), consistent with proapoptotic and antiproliferative activity of opioids (10). It was also reported that even a single morphine injection inhibits axon regeneration (27) and diminishes the structural integrity of cytoplasm in the hypothalamus (31). Moreover, the incubation of opioids with kidney cells induces transitory reorganization of actin filaments (24) as early as 15 min after addition. There were no changes in the level of ß-actin mRNA, which is consistent with our data, but the levels of -actin or cofilin were not measured (24).
It is less clear how changes in the changes in gene expression related to the analgesic, locomotor, and other pharmacological effects of opioids. However, this is beyond the scope of the current project and would require not only the measurement of the level of corresponding proteins but also intensive pharmacological analysis to confirm whether the changes in mRNA level are consequence or mediator of the opioid effects on neuronal activity.
It is accepted that these opioid effects are mediated by inhibition of neuromediator release via a decrease of cAMP concentration or Ca2+ current. However, downstream events are less clear. From our data the following downstream pathway is feasible: opioids attenuate the formation of synaptic vesicles from Golgi apparatus, prevent turnover of microtubules and actin filaments, and induce redistribution of adhesion proteins on the cell surface. In this way opioids may prevent the fusion of intracellular vesicles with the extracellular membrane and decrease the release of neurotransmitters. There are few data to support this hypothesis but at least one of the adhesion proteins, SPARC, potentiates morphine’s effects on locomotor activity following direct administration into the amygdala (12). In any way, the data indicate that expression of cytoskeletal and mitochondrial genes may be sensitive indicators of drug effects and may be used to find subtle distinctions among different opioid analogs and to detect the specific "isotypes" of regulatory pathways that mediate drug activity. The present findings also help to focus the future experiment on the new protein subset that would be difficult to select by other methods. It is noteworthy that observed alterations of gene expression were detected with use of a limited number of gene fragments that, however, were preselected according to their response to morphine administration.
The difference between the two areas studied was not dramatic. In part it may be explained by the fact that we measured alterations that occur simultaneously in many neurons. Further studies will show whether this is true for other brain areas and other drugs. In the striatum, however, the maximal effect was observed at 30 min after morphine administration vs. 2 h in the spinal cord. It is not clear why the effect of morphine occurs first in the striatum but lasts longer in the spinal cord (Figs. 5 and 6). There are no data indicating different pharmacokinetics in the brain and the spinal cord following systemic morphine injection. It remains to be proved that these lasting changes in the lumbar enlargement of the spinal cord are relevant to the development of acute dependence on or involved in the chronic effect of morphine.
In conclusion, the expression profile analysis proved to be able to detect changes in gene expression following single morphine injection. Moreover, these changes are mediated by a specific opioid receptor and might be useful in clarifying the effect of a known substance by providing insight into the regulatory pathways that otherwise evade detection. It seems that this approach may not only complement the pharmacological analysis of the drugs but may actually be a primary method to describe the properties of new substances, in particular those with neurotropic activity, and provide targets for traditional pharmacological and neurochemical experiments. However, to unleash its full potential will require further development of data analysis procedures and more comprehensive coverage of the genome.
|
ACKNOWLEDGMENTS |
|---|
This work was supported by grants from the Anesthesia Foundation of the University of Wisconsin to R. Y. Yukhananov and by National Institutes of Health Grant GM-42466 to G. J. Crosby.
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: R. Y. Yukhananov, Anesthesia Research, MRB 611, Dept. of Anesthesia, Brigham and Women’s Hospital, Harvard Medical School, 75 Francis St., Boston, MA 02115 (e-mail: ryyukhan{at}zeus.bwh.harvard.edu).
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIAL AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Fluck, C. Dapp, S. Schmutz, E. Wit, and H. Hoppeler Transcriptional profiling of tissue plasticity: role of shifts in gene expression and technical limitations J Appl Physiol, August 1, 2005; 99(2): 397 - 413. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M.-N. Giraud, M. Fluck, C. Zuppinger, and T. M. Suter Expressional reprogramming of survival pathways in rat cardiocytes by neuregulin-1{beta} J Appl Physiol, July 1, 2005; 99(1): 313 - 322. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Fluck, S. Schmutz, M. Wittwer, H. Hoppeler, and D. Desplanches Transcriptional reprogramming during reloading of atrophied rat soleus muscle Am J Physiol Regulatory Integrative Comp Physiol, July 1, 2005; 289(1): R4 - R14. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
1-tubulin (NM_011653, bp 918–1225), H-2D(q) antigen (X52915, bp 1887–1991)] were transcribed from corresponding plasmids. The gels were exposed to film, and the images were quantified with image analysis software. Data are presented as an average ratio (log2) from four gels ± SE. Array data are presented as intensity ratio morphine/saline (see 