Corticosteroids, either alone or in conjunction with transcription factors/hormones affect a wide array of genes responsible for regulating various physiological processes. Recently oligonucleotide microarrays were successfully used in our lab to examine these diverse effects of corticosteroids from two sets of animal studies, which allowed us to evaluate changes in an enormous number of genes in various tissues simultaneously after acute (Almon et al. 2004
; Almon et al. 2005c
; Jin et al. 2003
) and chronic MPL. While such evaluations are quite helpful in determining which genes are affected by steroids in rats, another important aspect of such experiments is specific disease/physiological process based evaluation of the data as demonstrated by Almon et al. who examined the effects of MPL on insulin resistance specific genes (Almon et al. 2005b
) in rat muscle. In this report, we investigated the dynamics of various genes related to the greater urea cycle, an important biochemical pathway present in mammalian livers for nitrogen disposal. Many of these enzymes are known to be highly regulated by CS in the liver (Christowitz et al. 1981
; de Groot et al. 1984
; Gautier et al. 1985
The dynamics of many of these genes following acute dosing were previously reported from our lab (Jin et al. 2003
). However simultaneous analyses of both the acute and chronic dataset revealed more complex gene regulations which were not apparent after only single high-dose MPL.
and Class 4A
were the simplest regulation patterns apparent amongst all the classes. TAT, which facilitates the transfer of an amino group from amino acid breakdown to the urea cycle and is also one of the most highly studied biomarkers, could be classified as one of the Class 1
genes. Both TAT and glucocorticoid receptor (Class 4A)
dynamics had extensively been studied in our lab (both at mRNA and protein levels) and five generations of highly mechanistic, quantitative PK/PD models were developed for MPL at various dose levels (Ramakrishnan et al. 2002
; Sun et al. 1998a
; Sun et al. 1998b
). Thus the presence of these two genes in our database not only had importance with respect to their involvement in the urea cycle, but also allowed us to compare the results from the microarrays to previous measurements by quantitative Northern Hybridization. For TAT the correlations (R) between these two techniques were quite high for both acute (0.68) and chronic (0.87) dosing while for GR the correlation was less than 0.5 which could be attributed to the much lower abundance of this message, the presence of greater variability in these data, and the more limited downward range of observed changes.
The other Class 1
genes were aspAT, with a similar function as TAT, ASS, one of the five urea cycle enzymes and MDH, a citric acid cycle enzyme that catalyzes the conversion of malate into oxaloacetate, thus facilitating recycling of the urea cycle by-product fumerate, which are shown in the urea cycle general schematic in . The dynamic profile for aspAT (), known to be regulated by CS (Garlatti et al. 1996
; Pave-Preux et al. 1988
) resembled TAT, however, incorporation of a capacity-limited stimulation by DR
) seemed to provide the best fit for the data.
The action of CS on ASS and MDH are known to be secondary (Husson et al. 2003
; Recupero et al. 1986
; Ulbright and Snodgrass, 1993
). Bourgeois et al. reported dose-dependent increase in ASS mRNA expression in cultured rat hepatocytes which could be completely abolished by actinomycin D indicating involvement of the synthesis of some secondary factors in this regulation (Bourgeois et al. 1997
). The MDH has also been reported to be secondarily regulated by CS by potentiating the binding capacity of T3 receptors since CS action on MDH in thyroidectomized animals was minimal (Recupero et al. 1983
; Recupero et al. 1986
). However, our current in vivo
datasets precluded addition of such secondary factors and therefore for simplicity Model B was used to capture both these enzyme inductions. For both ASS and MDH, however, the general patterns of the stimulation were quite different than TAT or aspAT. While the acute time profiles for ASS and MDH yielded a somewhat delayed and prolonged stimulation (lower degradation rates, 0.15 and 0.08 hr−1
compared to TAT and aspAT, 0.506 and 0.81 hr−1
), chronic infusion profiles for these two genes did not follow the apparent tolerance phenomenon visible in the chronic DR
) profile (). Rather the message seemed to remain up-regulated during the entire period of seven days () indicating somewhat higher sensitivity to DR
) compared to aspAT or TAT. The estimated values of the SC50
parameters for these two genes (Model B) agreed with this observation, almost 15–20 fold (3.2–3.5 nM) lower than the steady-state DR
) after chronic MPL (57 nM) compared to the higher estimated value for aspAT (191.5 nM).
Reports in the literature show CS-mediated direct or indirect increases in all three genes of Class 2,
i.e. ASL, CEBP-β and ODC (Chowdhury et al. 1996
; Hirvonen et al. 1988
; Matsuno et al. 1996
; Nebes and Morris, 1988
). Although our acute dataset predominantly showed up-regulation for these genes, chronic profiles indicated biphasic responses, especially for ODC and CEBP-beta. The initial increase in these messages was followed by a down-regulation up to or close to their baseline followed by a delayed increase. This was similar to the Class 1
genes suggesting a possible existence of a negative feedback regulator limiting the degree of increase of the message during chronic infusion of MPL. Models with a secondary mediator arising from the message itself regulating the initial increase would probably be most mechanistic; however, model convergence could not be achieved with such a model. ASL, one of the five urea-cycle enzymes known to be secondarily regulated by CS, showed similar profile as TAT or aspAT, however incorporation of the biosignal (Model C) was necessary for satisfactory fitting of the data.
An important transcription factor CEBP-β, contains a GRE in its promoter region and has been postulated to be one of the primary mediators of glucocorticoid effects on urea cycle enzymes in the liver, especially arginase and CPS. This chain of effects, i.e. of stimulation of CEBP-β by CS and in turn delayed stimulation of ARG and CPS by CEBP-β could be nicely captured for the acute dataset (results not shown); however, this model could not be implemented for the chronic dataset where the regulation seemed to be more complex, perhaps due to the involvement of multiple regulatory factors.
3 contains the largest number of genes among all the classes (including three of the five main urea cycle enzymes) predominantly showing a biphasic response, usually a rapid down-regulation followed by an up-regulation, which could generally be described by Model D or E. Such a phenomenon is not surprising for CS since they may induce or inhibit the same gene depending on other regulatory proteins/transcription factors that are already present on the DNA control regions (Alberts et al. 2001
). Since the up-regulation was somewhat delayed for all the genes, both models assume the increase is mediated by a biosignal.
Although most of the reports in the literature suggest an inhibitory effect of CS on OTC in fetal as well as in adult rats (Gautier et al. 1985
; Gautier et al. 1977
; Ulbright and Snodgrass, 1993
), we observed a subtle and delayed up-regulation in this gene after both acute and chronic MPL. As has been mentioned in the previous section, the coordinated regulation of CS on CPS (the first enzyme of the ornithine cycle, converts ammonia, originating from amino acid degradation, into urea) or arginase (the last enzyme of the cycle) via CEBP-β could not be applied to the chronic data-set and thus these two genes were modeled independently of CEBP-β. Arginase and GDH seemed to show saturability in stimulation by the biosignal similar to a few of Class 1
genes. Although this phenomenon could be described by Model E, the precision of the estimated parameters was poor perhaps because of over-parameterization of the model and inadequate data thus limiting the capability of the estimation, especially the capacity-limited parameters such as SC50
Class 4B genes, unlike Class 3 genes, do not show any delayed stimulation after acute MPL, however chronic dosing predominantly showed delayed up-regulation (glutaminase, involved in interconversion of two amino acids, glutamine and glutamate) or a biphasic response (carbonic anhydrase, responsible to form the bicarbonate ion from carbon dioxide from the Kreb’s cycle for the first step of urea cycle) suggesting differential regulation between acute and chronic dosing. Not much information is available regarding CS regulation of these two genes in the literature; however, initial investigation of the acute dataset showing down-regulation only was surprising to us owing to the fact that CS treatment seemed to increase most of the genes related to the urea cycle either instantaneously or with a delay, which could be clearly observed only with chronic dosing for these two genes.
An important component of our analysis of both acute and chronic dosing of MPL was the necessity of including a scaling factor for the chronic dataset. Since this factor was only included in the chronic dataset (refer to Eq. 17), a value of greater than 1.0 signified greater sensitivity for the chips for the chronic dataset. Two different chips were used for the two treatment groups and data generated from them may not always produce concordant results. In general signal values tend to be higher for the majority of RAE-230A probe sets relative to the corresponding probe sets on RG-U34 suggesting a need for a scaling factor to accommodate this discordance. It was suggested by Affymetrix that the newer 230A chip is most likely to outperform the RG-U34 probe set when both magnitude of signal and responsiveness to a biologically diverse tissue panel are evaluated. The 230A chips were superior in 51% of the probe sets, 5.7% of the probes seemed to have higher signals in RG-U34 chips, and less than 50% of the probes had equivalent signals.
Consistent with the physiological role of glucocorticoids in stimulating gluconeogenesis/urea cycle activity, most urea cycle genes exhibited a general enhanced expression following CS treatment. However, this rich time series data set illustrates that examination of changes in gene expression at a single time following drug administration can be misleading as to magnitude or even direction of change. For example, if a gene such as OTC which exhibits a complex pattern is examined at an early time following acute administration, one would conclude down-regulation by CS. On the other hand, if later time points were examined up-regulation is evident. This study also illustrates that time of exposure to drug (i.e. acute versus chronic treatment) is an important consideration to interpretation of the effects of CS on gene expression. Chronic exposure to CS in some cases results in more complex biphasic patterns of expression not evident following acute exposure, indicating probable secondary factors in control of expression. PD modeling can reveal insights into the complexities of mechanisms of gene regulation beyond simple up- or down-regulation by the drug.
In conclusion, endogenous and exogenous corticosteroids affect many physiological processes. Studying each pathway individually is more rigorous and quantitative using techniques such as Quantitative Real-Time PCR, however, it is more time consuming and laborious. Although challenges in proper analysis and understanding of microarray results have emerged since the technology evolved rapidly over the years, we were able to use this plethora of data to understand and analyze the underlying mechanism of various biological processes by using our mechanism-based PD models. The proposed models, while often premised on obvious mechanisms and supported by literature findings, also present testable hypotheses that can be examined in more extensive studies of specific genes.