|Home | About | Journals | Submit | Contact Us | Français|
Chronopharmacology is an important but under-explored aspect of therapeutics. Rhythmic variations in biological processes can influence drug action, including pharmacodynamic responses, due to circadian variations in the availability or functioning of drug targets. We hypothesized that global gene expression analysis can be useful in the identification of circadian regulated genes involved in drug action. Circadian variations in gene expression in rat liver were explored using Affymetrix gene arrays. A rich time series involving animals analyzed at 18 time points within the 24 hour cycle was generated. Of the more than 15,000 probe sets on these arrays, 265 exhibited oscillations with a 24 hour frequency. Cluster analysis yielded 5 distinct circadian clusters, with approximately two-thirds of the transcripts reaching maximum expression during the animal’s dark/active period. Of the 265 probe sets, 107 of potential therapeutic importance were identified. The expression levels of clock genes were also investigated in this study. Five clock genes exhibited circadian variation in liver, and data suggest that these genes may also be regulated by corticosteroids.
Virtually all organisms have biological rhythms associated with the light-dark cycle (Badiu, 2003; Oishi et al., 2003; Murphy, 2005; Ueda et al., 2005). In mammals, rhythms exist at all levels of organization from the organismal to the cellular. The central orchestrators of these rhythms are paired suprachiasmatic nuclei (SCN) in the anterior part of the hypothalamus which receive direct input by way of the retinohypothalamic tract. However, the existence of diurnal and nocturnal animals and the ability of animals to shift with changes in the light-dark cycle indicate that beyond the SCN the rhythms are not slaves to the presence or absence of light (Dardente et al., 2002; Challet et al., 2003). In addition to receiving inputs from the retina, the SCN also receives inputs from forebrain areas that modulate the downstream influences of the SCN. Of particular importance to rhythmicity in peripheral tissues are outputs from the SCN which are directed to other parts of the hypothalamus that regulate both anterior and posterior pituitary hormones, as well as the autonomic nervous system. In addition, behavioral adjuncts associated with the exigencies of life such as feeding and activity can impact rhythmicity downstream of the SCN.
In the SCN the circadian clock involves an autoregulatory negative feedback loop of gene expression (Dardente et al., 2002; Challet et al., 2003; Ueda et al., 2005). Its basic elements are several transcription factors including CLOCK (Circadian Locomotor Output Cycles Kaput) and BMAL1 (Aryl Hydrocarbon Receptor Nuclear Translocator-like) which heterodimerize and enhance the expression of PERIOD (PER) and CRYPYOCHROME (CRY). In turn, PER and CRY heterodimerize and repress the expression of CLOCK and BMAL1. The core system is entrained to the light/dark cycle with CLOCK:BMAL being high during the light and PER:CRY being high during the dark (Ueda et al., 2005). In addition to the core transcription factors, there are additional transcription factors that add flexibility and adaptability to the central clock.
The central clock anticipates the change in photoperiods, preparing the animal for the upcoming period of activity and feeding, regardless of whether that period is in light or dark. The input from the SCN to the regulation of both pituitary hormones and the autonomic nervous system impart rhythmicity to peripheral tissues. However, this is further complicated by more diffuse behavior-related factors which alter systemic demands. Many of the transcription factors involved in regulating the central clock are also expressed in peripheral tissues. However, their regulation is complicated by variations in ancillary factors. The existence of both diurnal and nocturnal mammals and the phenomena of phase shifting by food restriction illustrate both the complexity and flexibility in peripheral rhythmicity. Its intrinsic nature is illustrated by the observations that rhythmic behavior with a periodicity of approximately 24 hours can be induced in a variety of cells in culture (Ueda et al., 2005).
The hypothalmic/pituitary/adrenal axis (HPA) is of particular importance to the active feeding period. Its effector hormones, glucocorticoids, are high during the light period in diurnal animals and during the dark period in nocturnal animals (Dardente et al., 2002). The mechanism for most glucocorticoid effects involves modulation in the amount of specific mRNAs (Almon et al., 2007). By virtue of their circadian periodicity, glucocorticoids are effectors of many circadian changes in gene expression.
The liver expresses an unusually large and diverse repertoire of genes (Almon et al., 2007). The nature of the processes carried out by liver suggests that many of its expressed genes should be under circadian control either directly or indirectly. In the present report we describe the use of Affymetrix arrays to analyze the livers of rats maintained on a strict light/dark regimen consisting of 12 hours light/12 hours dark with three animals sacrificed at 18 time points during the 24 hour period. This rich time series allowed us to group genes into five relatively discrete circadian clusters. Circadian responsive genes were also examined within the context of glucocorticoid regulation and their response to exogenous corticosteroids.
The probe sets that were found to have an oscillation with a 24 hour frequency had their identities confirmed when possible using the Basic Local Alignment and Search Tool (BLAST). This information was used to parse the genes into 13 functional groups. Functional groups were then analyzed by clusters to determine the distribution of these functions in circadian time.
Dysregulation of aspects of liver function are associated with a variety of common pathologies. As a result, liver functions are commonly targeted by drugs. It has been long recognized that rhythmic variations in biological processes can affect therapeutics, including absorption/distribution, excretion, protein binding, and response (Reinberg, 1992; Labrecque et al., 1995; Smolensky et al., 1999). Therefore the circadian regulation of gene expression was also examined and discussed within the context of drug targeting, with emphasis of cholesterol/bile acid synthesis, cancer chemotherapeutics, and translation and protein processing.
Fifty-four normal (150–175 g) male Wistar rats were purchased in two separate batches of 27 from Harlan Sprague-Dawley Inc. (Indianapolis, IN, USA) and experiments were initiated at body weights between 225 and 275 g. Animals were housed and allowed to acclimatize in a constant-temperature environment (22°C) equipped with a 12-h light/dark cycle. Twenty-seven rats (Group I) were acclimatized for 2 weeks prior to study to a normal light/dark cycle, where lights went on at 8 AM and off at 8 PM. The onset of the light period was considered as time zero. The other 27 rats (Group II) were acclimatized for 2 weeks prior to study to a reversed light/dark cycle, where lights went on at 8 PM and off at 8 AM. Rats in Group I were killed on three successive days at 0.25, 1, 2, 4, 6, 8, 10, 11, and 11.75 hr after lights on to capture the light period. Rats in Group II were killed on three successive days at 12.25, 13, 14, 16, 18, 20, 22, 23, and 23.75 h after lights on to capture the dark period. Animals sacrificed at the same time on successive days were treated as triplicate measurements. Because normal rats were used, minimal animal handling with least possible environmental disturbances was employed to minimize stress. Night vision goggles were used to carry out animal procedures conducted in the dark period. At sacrifice, rats were weighed, anesthetized by ketamine/xylazine, and sacrificed by aortic exsanguination. Blood was drawn from the abdominal aortic artery into syringes using ethylenediaminetetraacetic acid (4mM final concentration) as anticoagulant. Plasma was harvested from blood by centrifugation (2000 × g, 15 minutes, 4°C) and frozen at minus 80°C until analyzed for corticosterone. Livers were excised and frozen in liquid nitrogen immediately after sacrifice and stored at minus 80°C until RNA preparation. Both acute and chronic MPL dosing experiments have been previously published (Almon et al., 2007). In brief, populations of adrenalectamized male Wistar rats were given doses of the synthetic glucocorticoid, methylprednisolone (MPL). In the acute experiment the animals were given a single bolus dose (50 mg/kg) of MPL and were sacrificed at 16 times over a 72 hour period following dosing. In the chronic experiment, the animals received a constant infusion of 0.3 mg/kg/h MPL via Alzet osmotic pumps and were sacrificed at 10 times over a 168 hour period. All rats had free access to rat chow and 0.9% saline drinking water. Our research protocol adheres to the Principles of Laboratory Animal Care (NIH publication 85-23, revised in 1985) and was approved by the University at Buffalo Institutional Animal Care and Use Committee.
Plasma corticosterone concentrations were determined by a sensitive normal-phase high-performance liquid chromatography (HPLC) method as previously described (Haughey and Jusko, 1988). The limit of quantitation was 10 ng/ml. The interday and intraday coefficients of variation (CV) were less than 10%.
Liver samples from each animal were ground into a fine powder in a mortar cooled by liquid nitrogen and 100 mg was added to 1 ml of pre-chilled Trizol Reagent (InVitrogen, Carlsbad CA). Total RNA extractions were carried out according to manufacturer’s directions and were further purified by passage through RNeasy mini-columns (QIAGEN, Valencia, CA) according to manufacturer’s protocols for RNA clean-up. Final RNA preparations were resuspended in RNase-free water and stored at −80°C. The RNAs were quantified spectrophotometrically, and purity and integrity assessed by agarose gel electrophoresis. All samples exhibited 260/280 absorbance ratios of approximately 2.0, and all showed intact ribosomal 28S and 18S RNA bands in an approximate ratio of 2:1 as visualized by ethidium bromide staining. Isolated RNAs from each liver sample was used to prepare the hybridization targets according to manufacturer’s protocols. The biotinylated cRNAs were hybridized to 54 individual Affymetrix GeneChips Rat Genome 230A (Affymetrix, Inc., Santa Clara, CA), which contained 15,967 probe sets. The 230A chip was used in the chronic infusion experiment as well allowing direct comparison between the two experiments. The 230A gene chips contain over 7,000 more probe sets more than the ones used (U34A) in our earlier muscle bolus dose MPL study (Almon et al., 2005). The high reproducibility of in situ synthesis of oligonucleotide chips allows accurate comparison of signals generated by samples hybridized to separate arrays. This data set has been submitted to GEO (GSE8988).
As detailed above, animals were sacrificed at precise times on three successive days to obtain data points for the light period and three successive days to obtain data points for the dark period. Animals sacrificed at the same time on different days were treated as three replicates for that time to construct a 24 hr light:dark cycle. In order to obtain a clear picture of an entire cycle, two 24 hr periods were concatenated to obtain a 48 hr period which allowed visualization of rhythms that spanned the dark/light and dark/light transitions.
A non-linear curve fit using MATLAB was conducted which fitted a sinus function [A*sin(Bt + c)] to the data including the replicates. Genes that could be curve fitted with a R2 correlation of greater than 0.8 were kept. This curve fitting approach enabled use of replicate information instead of depending on the ensemble average necessary with Fourier transforms or Lombs Scargle methods. This approach is viable due to our relatively large number of time samples. This dataset was then loaded into a data mining program, GeneSpring 7.0 (Silicon Genetics, Redwood City, CA), and we normalized the value of each probe set on each chip to the average of that probe set on all chips. In order to identify genes with similar patterns of oscillation within the daily cycle we applied Quality Threshold Clustering (QT) in GeneSpring using Pearson’s correlation as the similarity measurement.
It is assumed that genes whose expression levels are part of the circadian rhythm will show one full oscillation every twenty-four hours. However, in light of possible ultradian or infradian cycles within the data, we have utilized a much more general model of periodic signals given as [A*sin(Bt + c)]. A non-linear curve fit was conducted which fitted a sinusoid to the data including the replicates. We identified 265 probe sets which fit the model [A*sin(Bt + c)] with a R2 correlation greater than 0.8. With this more general model, we found that all genes which showed a high level of correlation had similar values for B and showed one full cycle over 24 hours. This suggests that while there may exist ultradian or infradian signals, they were not evident in the experimental dataset, perhaps due to the sampling strategy employed in the experimental design. Using GeneSpring, we normalized the value of each probe set on each chip to the average of that probe set on all chips such that the expression pattern of all probe sets oscillated approximately around 1. There appear to be two major patterns as illustrated in Figure 1, with one pattern reflecting maximum expression during the light/inactive period while the others reach a maximum during the dark/active period. However, oscillations in expression have more discrete relationships to the light/dark periods. In order to group genes with similar patterns within the daily cycle we applied QT Clustering, yielding 5 clusters. Figure 2 shows these five clusters with the centriod (average of all of the genes in each cluster) highlighted in white. Approximately two-thirds of the probe sets reach maximum expression during the dark/active period. Supplementary Tables 1–5 (available online) provide a detailed list of all genes in each cluster, including Probe Set ID, Accession Number, Pearson’s correlation coefficient with the centroid, gene symbol, gene name, and gene function. Corticosterone reaches its maximum plasma concentration in these animals at hour 13.3 (Figure 3).
Three basic categories of transcription factors have been associated with control of circadian oscillation in gene expression. The first are those that participate through E-Box binding, the second through DBP/E4BP4 binding elements (D-boxes), and the last through RevErbA/ROR binding elements (RREs). We examined the chip for probe sets for transcription factors previously identified as involved in regulation of circadian patterns (Ueda et al., 2005). The 230A chip contained probe sets for 16 out of 17 of these transcription factors. However, only five (PER2, BMAL1b, DBP, Nr1d1, and Nr1d2) showed distinct circadian oscillation. Figure 4 shows the circadian patterns of these five genes in relation to the light/inactive and dark/active periods. Consistent with the literature on the core clock in the SCN, BMAL1b reaches a maximum during the light period while PER2 reaches a maximum during the dark period. The chip did contain probe sets for Clock, PER1, PER 3, CRY2, Bhlhb2 (Basic Helix-Loop-Helix Domain-Containing Protein, Class B 2, Dec1); Bhlhb3 (Basic Helix-Loop-Helix Domain-Containing Protein, Class B 3, Dec2, Sharp1); NFIL3A (Nuclear Factor, Interleukin 3 Regulated, E4BP4); RORA (RAR-Related Orphan Receptor A, RAR-Related Orphan Receptor Alpha, RZR-Alpha, RZRA, Retinoic Acid-Binding Receptor Alpha); RORB (RAR-Related Orphan Receptor B, RAR-Related Orphan Receptor Beta, RZR-Beta, RZRB, Retinoic Acid-Binding Receptor Beta) and RORC (RAR-Related Orphan Receptor C, RAR-Related Orphan Receptor Gamma, RORG, RZR-Gamma, RZRG, Retinoic Acid-Binding Receptor Gamma). We visually inspected the signals for these circadian related genes. The objective was to ascertain if there was a signal that just did not oscillate with a 24 hour frequency or if the signal was either not present or too low to be measured by the probe set. Signal intensities for Bhlhb2, NFIL3A, Clock, RORC, and RORB were sufficiently strong to indicate that they were expressed in the tissue even though they did not have a circadian rhythm. For the remainder of the probe sets, the signal was very low indicating that either they are not expressed in the tissue or that the probe set was not adequate to measure their presence.
Previously we conducted two time series experiments in which cohorts of rats were given MPL either as a single bolus dose or chronic infusion and livers analyzed by gene arrays (Almon et al., 2005; Almon et al., 2007). Because a major regulator of circadian rhythms is the HPA axis, these arrays were examined for clock genes. Figure 5 and and66 show the acute and chronic profiles for both PER2 and BMAL1, two major clock genes. In the acute profile both respond to the single dose with a transient oscillation. With chronic infusion, both genes begin to oscillate but after about 48 hours all points for BMAL1 shows enhanced expression while all points for PER2 shows down-regulation. Three additional clock genes, DBP, nr1d1 and nr1d2, all have acute and chronic profiles similar to PER2 with the chronic profiles being consistently down-regulated after 48 hrs (Figures 7–9).
Although the data suggest that BMAL1 may be continuously up regulated after 48 hours and that PER2, DBP, nr1d1 and nr1d2 may be continuously down regulated after 48 hours, this conclusion may be an artifact of sampling times. If one simply ignores the 36 hour point between 24 hours and 48 hours the conclusion of continuous up or down regulation extends back 24 hours. An alternative possibility is that all continue to oscillate with BMAL1 being out of phase with PER2, DBP, nr1d1 and nr1d2.
Using extensive literature searches and domain knowledge we parsed all of the genes for which there were probe sets with a 24 hour frequency of oscillation into functional groups. We were able to identify genes that corresponded to all but eleven of the 265 probe sets. In all cases we attempted to classify the gene with respect to its function in the liver. This categorization is not perfect because some genes can fit into more than one group. For example, we placed interleukin 32 (IL32) and Kruppel-like factor 13 (Klf13) in the Immune Related group, whereas they could also have been placed in the Signaling and Transcription Regulation groups respectively. The thirteen functional groups are as follows: Bile Acid/Cholesterol Biosynthesis; Cell Cycle/Apoptosis; Translation/Protein Processing; Cytoskeleton; Carbohydrate/Glucose Metabolism; Immune Related; Lipid Metabolism; Mitochondrial; Protein Degradation; Signaling; Small Molecule Metabolism; Transcription Regulation; and Other. Table 1 shows the relevant information for each probe set in each functional grouping along with the cluster to which it belongs. As can be seen in Figure 2, Clusters 1 and 2 reach maxima during the light period, Cluster 3 reaches a maximum very close to the transition between light and dark while Clusters 4 and 5 reach maxima during the dark period. The most highly populated functional group is Translation/Protein Processing which contains 65 probe sets. Interestingly 55 of the probe sets are in Clusters 4 and 5 with maxima during the dark period. In contrast, the second most populated functional group is Cell Cycle/Apoptosis with 35 probe sets that are distributed almost equally between Clusters 1 and 2 with maxima during the light period and Clusters 4 and 5 with maxima during the dark period. The next two most populated groups are Lipid Metabolism and Transcription Regulation with 21 probe sets each. Lipid Metabolism has several probe sets in Cluster 3 with most of the remainder in Clusters 4 and 5. This pattern suggests that the system begins to anticipate the active dark period during the end of the light inactive period. Transcription Regulation shows no anticipation but Clusters 4 and 5 clearly dominate. The next two most populated groups are Bile Acid/Cholesterol Biosynthesis and Cytoskeleton, with 16 and 15 probe sets respectively. Quite clearly, Cluster 4 contains major enzymes involved in cholesterol biosynthesis while the production and movement of bile acids seems much more distributed. The remaining functional groups, Signaling (14), Carbohydrate/Glucose metabolism (13), Small Molecule Metabolism (12), Mitochondrial (12), Immune Related (11) and Protein Degradation (6) also have distributions that indicate functional significance during different times of the circadian cycle. The Other category contains 24 probe sets but 11 of these could not be assigned a function.
Three of the functional groups presented in Table 1 were unusually rich in potential drug targets and biomarkers. These functional groups of genes were examined more closely to identify current or potential drug targets and biomarkers, exploring the premise that the use of a drug or measurements of biomarkers may be optimized by taking advantage of circadian variations in the associated gene targets.
Enzymes associated with the synthesis of both cholesterol and bile acids are all in Cluster 4 which has a maximum expression four hours into the animal’s dark/active period. This is not particularly surprising since rats, being nocturnal, are active and ingest food during the dark period, thus requiring bile acids during this time. Notable in this cluster are two probe sets for the enzyme HMG-CoA reductase which is the target for statin cholesterol lowering drugs (Stacpoole et al., 1987; Staels, 2006), as well as Sqle, another potential target for hypocholesterolemic drugs (Chugh et al., 2003), and Cyp7a1, which is the rate-limiting enzyme in the conversion of cholesterol to bile acids and is inhibited by fibrates, a class of hypolipidemic drugs (Post et al., 2001). Clusters 1 and 2 (lights on +3 hr and +6 hr, respectively) contain genes that are important to bile acid flow. For example, Cluster 2 contains Abcb11 which mediates the elimination of cytotoxic bile salts from liver cells to bile, and therefore plays a critical role in the generation of bile flow. A variety of drugs inhibit this export pump which can cause drug-induced intrahepatic cholestasis, one of the major causes of hepatotoxicity (Carlton et al., 2004).
Of these 35 genes, almost all are either cancer chemotherapeutic targets or biomarkers relevant to prognosis. They are distributed throughout the light/inactive and dark/active periods and as such are found in all five clusters. Clusters 1 and 2 both peak in the light period. Cluster 1 is particularly rich in both chemotherapeutic targets and biomarkers, including beta tubulin (the main target of paclitaxel) (Tommasi et al., 2007), TXR1 (whose up-regulation impedes taxane-induced apoptosis in tumor cells) (van Amerongen and Berns, 2006), DAPK1 (whose lack of or low levels of expression is associated with highly aggressive metastatic tumors and is also a prognostic marker for disease recurrence) (Fraser and Hupp, 2007), and reprimo (a candidate tumor-suppressor gene whose aberrant methylation is associated with various cancers) (Takahashi et al., 2005). Similar relationships to cancer can be found in the remaining seven genes in Cluster 1. Cluster 2 contains five genes related to apoptosis, including several Bcl-2-binding proteins (Erkan et al., 2005; Zhao et al., 2005). Cluster 3 which peaks close to the light/dark transition contains only one gene, SHMT1, whose polymorphism is related to methotrexate resistance (de Jonge et al., 2005). Clusters 4 and 5 both peak during the dark period. Cluster 4 contains five genes relevant to the control of cell cycle and apoptosis, including Bnip3, whose down-regulation is associated with increased resistance to both 5-fluoro-uracil and gemcitabine (Erkan et al., 2005). Cluster 5 contains six genes. Of particular import is ODC1, the first enzyme in polyamine biosynthesis. Many chemotherapeutic strategies involve inhibition of polyamine biosynthesis and ODC plays a significant role in many of these, which include direct inhibitors of the enzyme often in combination with polyamine uptake inhibitors (Basuroy and Gerner, 2006). In addition, there are therapeutic approaches seeking to silence the ODC gene (Nakazawa et al., 2007). ODC is also a prognostic indicator, with treatment outcome being inversely related to tumor content (Basuroy and Gerner, 2006). This cluster also contains several other genes involved with DNA repair and thus potential targets for cancer therapies.
In this last functional group we included all genes directly associated with both translation such as ribosomal proteins and protein processing such as chaperonins which are large molecular assemblies that assist protein folding to the native state. Inhibitors of chaperonins are being assessed as chemotherapeutic agents while enhancers of chaperonin activity are under investigation because misfolded proteins are responsible for a variety of diseases (Fenton and Horwich, 2003; Murphy, 2005; Powers and Workman, 2006; Zheng and Yenari, 2006). Fifty-five of the 65 genes are concentrated in Clusters 4 and 5 which peak in the dark/active period. Cluster 4 includes Hsp70 and three of its partner proteins: Dnaja1; Dnaja2; and Hsj2 (Zheng and Yenari, 2006). It also contains several genes associated with ribosomal synthesis and assembly, one of which, nucleolin, is currently under investigation as a drug target (Sakita-Suto et al., 2007). A nucleolin antisense oligonucleotide is being studied for inhibition of tumor cell proliferation. Cluster 5 is even richer in genes associated with translation and protein processing. Among these are transcripts for 14 proteins with chaperonin activity including Hsp90, the most abundant molecular chaperone in eukaryotic cells and a major focus for drug development (Powers and Workman, 2006). It also contains many genes associated with ribosomes including five transcripts for proteins that are part of the 60S ribosomal subunit. In addition there are transcripts for RNA helicases, several proteins involved in mRNA processing and EIF4A3 (Chan et al., 2004). What is clear from these data is that, for the most part, protein synthesis and processing take place during the dark when the animal is active. However, there are a limited number of transcripts that reach a maximum at other times. For example, the only gene in Cluster 3 is FKBP5 which is both a potential drug target and a biomarker. Its isomerase activity is inhibited by FK506 (tacrolimus), a macrolide immunosuppressant. Allelic variations in the FKBP5 gene are associated with depression and response to antidepressants (Binder et al., 2004). This relationship to depression seems to be related to activity of the HPA axis. Therefore it is probably relevant that FKBP5 reaches an expression maximum at a time very close to when circulating corticosterone peaks.
This report describes an analysis of circadian rhythms of mRNA expression in the liver of adult male rats. Animals were sacrificed at nine times during a 12 hour light period and nine corresponding times during a 12 hour dark period. Liver RNAs from each of the 54 animals were applied to individual Affymetrix GeneChips (RAE230A). Analysis yielded 265 probe sets with a 24 hour frequency. Because of the richness of this dataset, we were able to apply QT clustering and identified 5 groups with maxima at different times during the cycle. Two peaked during the light period, two during the dark period, and one very close to the light to dark transition. Approximately two-thirds of the probe sets reach maximum expression during the active (dark) period. The chip in several cases contained more than one probe set for the same gene. In 14 out of 15 instances, all probe sets for the same gene sorted to the same cluster. The single exception was Pvrl2 which sorted to both Clusters 1 and 2. The correlation coefficient of the probe set in Cluster 1 (1375216_at) was 0.84 while the correlation coefficient of the probe set in Cluster 2 (1370345_at) was also 0.84. Both have amongst the lowest correlations with the centroids of their respective clusters.
Regulation of the central clock involves a number of other transcription factors that may be expressed in peripheral tissues. The array used here contained probe sets for PER1, PER2 and, PER3. However, only PER2 showed significant expression and circadian oscillation while PER1 and PER3 had very low signals. Similarly, probe sets for CRY2 and Bhlhb3 also had very low signals. A very low signal can be due to either the lack of expression of the gene or inadequacy of the probe set to measure the signal. In contrast, the chip contained probe sets for Bhlhb2, NFIL3A, Clock, RORC and RORB, and these signals were reasonably strong but without oscillation. It has been reported that at least in some tissues, Clock is expressed at tonic levels and that cycling is due to the rhymicity of its heterodimeric partner BMAL (Reddy et al., 2005). Of the remaining transcription factors that have been implicated in regulation of circadian changes in gene expression only PER2, BMAL DBP, Nr1d1, and Nr1d2 showed a pattern of circadian oscillation. PER2 was in Cluster 5 during the dark period while BMAL was in Cluster 1 during the light period. DBP, Nr1d1 and Nr1d2 are all in Cluster 3.
The fact that BMAL1, PER2, DBP, Nr1d1 and Nr1d2 all begin to oscillate in response to acute MPL dosing suggest that they are all glucocorticoid sensitive either directly or indirectly. The observation that following chronic dosing an initial oscillation of BMAL occurs followed by what appears to be continuous up-regulation while PER2, DBP, Nr1d1 and Nr1d2 shows oscillation followed by what appears to be continuous down-regulation is potentially informative. However, the apparent continuous up- or down-regulation of the genes may actually be an artifact of sampling times. Just as reasonable an interpretation of the results is that all five genes continue to oscillate throughout the infusion period with BMAL being out of phase with PER2, DBP, Nr1d1 and Nr1d2.
Because synthetic glucocorticoids are a widely used class of drugs, we compared circadian regulated gene expression with those directly regulated by corticosteroids. These datasets together allowed us to address two basic but related questions. The first is: do all genes that respond to corticosteroids have circadian rhythms? The second is: do all genes with circadian rhythms respond to corticosteroid? The answer to both questions is no. Seventy-seven of the genes identified were both circadian and MPL responsive. The fact that all genes that respond to MPL are not circadian and that all genes with circadian rhythms do not respond to MPL suggests that there exist some diversity in mediating mechanisms. This result is consistent with previously described observations comparing our acute and chronic profiles (Almon et al., 2007).
If an animal is diurnal, changes in mRNA expression near the end of the dark period begins to prepare the animal for the activity and feeding time. Similarly, changes during the end of the light period prepare the diurnal animal for inactivity and rest. Rats are nocturnal and cycling of gene expression in peripheral tissues like the liver is reversed relative to humans who are essentially diurnal. We explored the results with a focus of potential chronotherapeutic insight.
We identified several genes transcripts that are closely associated with hypocholesterolemic drug strategies. Among these are transcripts for HMG-CoA reductase, the statin target, Sqle, involved in cholesterol synthesis, and Cyp7a1 which is the fibrate target in conversion of cholesterol to bile acids. These transcripts are in Cluster 4 which reaches a maximum 4 hr into the animal’s dark/active period. The current practice of having patients take statins before they go to bed (Staels, 2006) would seem inappropriate since available data indicates that in humans, HMG-CoA reductase has a maximum expression at about 10 AM (Harwood et al., 1987; Stacpoole et al., 1987). In those experiments the investigators were directly measuring enzymatic activity in serially drawn mononuclear leukocyte. The assumption was that activity in mononuclear leukocyte mirrors activity in the liver. In contrast, whole body cholesterol biosynthesis has been reported to peak between midnight and 3 AM (Parker et al., 1982). In those experiments, the investigators used plasma mevalonic acid as a biomarker for whole body cholesterol biosynthesis. Mevalonic acid is the direct product of HMG-CoA reductase. In addition, urinary mevalonic acid has become an indicator of the effectiveness of statin drug treatment (Hiramatsu et al., 1998). However, our data is consistent with the data of Harwood et al. indicating the enzyme reaches its peak during the animal’s active feeding period which in the case of rats would be during the dark period as opposed to the light period in humans. The presence of squalene epoxidase in Cluster 4 further reinforces the validity of these observations. What is confusing is why plasma and urine mevalonic acid peak during the inactive period in humans. Most of the mevalonic acid synthesized in the liver is used for cholesterol and then bile acid biosynthesis. However, the preponderance of mevalonic acid in circulation is metabolized by the kidney with the primary products being squaline and lanosterol (Raskin and Siperstein, 1974). The reason that the timing of the use of statins is important may have less to do with efficacy in lowering cholesterol and more to do with the toxic side effect associated with destabilization of muscle membranes and the development of rhabdomyolysis.
Of the transcripts with circadian rhythms, 35 were for proteins related to cell cycle and apoptosis. In contrast to the cholesterol synthesis related genes, the genes in this functional group are distributed in all five clusters. A relationship between circadian rhythms and cell cycle is well established and our data simply confirms and elaborates on the observations of others. However, the exploitation of these observations in cancer chemotherapy is not straightforward. In cancer therapeutics the important consideration is outcomes, which is based on the balance between toxic and therapeutic effects. If all dividing cells are entrained the same way to the circadian rhythm, then attaining an optimum balance is more complicated. However, some evidence is available that at least in some cases cancer cells have altered organization of the cell cycle relative to the circadian rhythm (Canaple et al., 2003; Garcia-Saenz et al., 2006). To the degree that this is true then knowledge of the circadian expression of drug targets in normal cells may provide a basis for reducing toxicity. Adding to the complexity are the observations that endogenous circadian rhythms are often disrupted in cancer patients.
Of the 265 circadian transcripts, 65 are associated with translation and protein processing. Out of these, only 8 reach a maximum during the light/inactive period. Cluster 3, which reaches a maximum shortly after the transition, contains only one transcript, FKBP5. FKBP5 is associated with glucocorticoid signaling (Binder et al., 2004) and reaches a maximum expression very close to the maximum of the corticosterone circadian rhythm. Clusters 4 and 5 contain the remaining 55 transcripts. Prominent among these are 20 chaperone related proteins. Both inhibiting and enhancing chaperone activities are evolving drug strategies. An important set of drugs in this category are geldanamycin derivatives which inhibit Hsp90 causing the degradation of proteins involved in a large variety of cellular processes from cell cycle and apoptosis to angiogenesis. Because misfolded proteins are associated with several diseases there are a variety of approaches being developed to enhance the activity of Hsp90 and other chaperonins (Powers and Workman, 2006). Two particularly interesting areas are chaperone-mediated enzyme enhancement and gene therapy. What is also clear is that protein synthesis related expression occurs primarily in Cluster 5. The fact that six proteins that are part of the 60S ribosomal subunit are co-expressed in Cluster 5 tends to validate our results. Proteins that work together are expressed together.
With the burgeoning development of antisense oligonucleotides it is probable that more transcripts will become drug targets. Timing will be an important aspect in the use of antisense technology when applied to transcripts with circadian rhythms.
This dataset was developed at the Children’s National Medical Center under the auspices of a grant from the National Heart, Lung, and Blood Institute (NHLBI)/NIH Programs in Genomic Applications HL 66614 (Eric P. Hoffman, PI). We would like to acknowledge the technical assistance of Ms. Nancy Pyszczynski and Ms. Suzette Mis.
This work was supported by grant GM 24211 from the National Institute of General Medical Sciences, NIH, Bethesda, MD. EY and IPA also acknowledge support from NSF grant 0519563 and EPA grant GAD R 832721-010.