Abstract
Glucocorticoids play a role in regulation of T lymphocytes homeostasis and development. In particular, glucocorticoid treatment induces massive apoptosis of CD4+CD8+ double-positive (DP) thymocytes. This effect is due to many mechanisms, mainly driven by modulation of gene transcription. To find out which genes are modulated, we analyzed DP thymocytes treated for 3 h with dexamethasone (a synthetic glucocorticoid) by global gene expression profiling. Results indicate modulation of 163 genes, also confirmed by either RNase protection assay or real-time polymerase chain reaction. In particular, dexamethasone caused down-regulation of genes promoting DP thymocyte survival (e.g., Notch1, suppressor of cytokine signaling 1, and inhibitor of DNA binding 3) or modulation of genes activating cell death through the ceramide pathway (UDP-glucose ceramide glucosyltransferase, sphingosine 1-phosphate phosphatase, dihydroceramide desaturase, isoform 1, and G protein-coupled receptor 65) or through the mitochondrial machinery. Among the latter, there are Bcl-2 family members (Bim, Bfl-1, Bcl-xL, and Bcl-xβ), genes involved in the control of redox status (thioredoxin reductase, thioredoxin reductase inhibitor, and NADP+-dependent isocitrate dehydrogenase) and genes belonging to Tis11 family that are involved in mRNA stability. Our study suggests that dexamethasone treatment of DP thymocytes modulates several genes belonging to apoptosis-related systems that can contribute to their apoptosis.
Regulation of T cell survival is important as a physiological mechanism involved in determining the development of the immune response. Moreover, it is well accepted that apoptosis plays a relevant role in the thymus, where massive cell death occurs in continuous selection process (Vacchio and Ashwell, 2000; Jondal et al., 2004; Lepine et al., 2005). Among different signals and stimuli, glucocorticoid hormones (GCs) have been shown to regulate apoptosis of T cells and thymocytes (Kong et al., 2002), among which CD4+CD8+ double-positive thymocytes (DPTs) are the major target of GC-induced apoptosis (Vacchio and Ashwell, 2000; Kong et al., 2002).
Apoptosis involves binding of GCs to their receptor, whose activation causes genomic and nongenomic effects (Goulding, 2004; Lepine et al., 2005). In thymocytes, modulation of gene transcription is crucial for the proapoptotic effect so that inhibition of mRNA or protein synthesis inhibits apoptosis (Distelhorst, 2002; Lepine et al., 2005). In the attempt to find out those genes whose modulation plays a crucial role in the GC-induced apoptosis, some studies have focused their attention only on a single gene modulated by GC treatment [such as Bim and G protein-coupled receptor 65 (Gpr65)], whose overexpression has a proapoptotic effect (Tosa et al., 2003; Wang et al., 2003a). However, despite the valuable meaning of these studies, other studies using mice in which those genes were individually ablated demonstrated that these single genes were dispensable for the GC-induced apoptosis (Bouillet et al., 1999; Radu et al., 2006). On the contrary, the forced overexpression of antiapoptotic genes (such as Bcl-2 and Notch1) protects from GC-induced apoptosis, and their elimination causes an increased sensitivity to GC-induced apoptosis (Ma et al., 1995; Izon et al., 2002), possibly suggesting that the balance between pro- and antiapoptotic genes plays a role in the response to GCs (network hypothesis).
In the very recent past, studies using microarray techniques focused on cell lines undergoing apoptosis following GC treatment. By this way, it was possible to study the GC effects following several hours and some days of treatment, identifying genes potently modulated by GCs. However, it is likely that cell lines respond differently from untransformed cells to GC treatment.
Two more studies used microarray techniques to describe GC effects on primary culture of tumoral cells undergoing apoptosis but reached quite different conclusions (Chauhan et al., 2002; Schmidt et al., 2005). Indeed, a meta-analysis of all microarray studies dealing with GC-induced apoptosis revealed that only few genes appeared in three or more systems (Schmidt et al., 2004), strongly suggesting that the effects of GC treatment are different, depending on differentiation and functional status of the cells, and level of expression of other transcription factors and apoptosis regulators, so that different apoptotic genes are important in a cell subpopulation but not in others.
For these reasons, we studied the modulation of gene transcription following GC treatment using exclusively DPTs, a prototype of GC-sensitive cells. In particular, in this study, we show the results of a global gene expression profiling of DPTs undergone to short-time treatment with Dex, a synthetic GC. Upon statistical validation, data bank queries, and evaluation of known gene functions in T lymphocytes, we were able to group modulated genes in five systems that could cooperate in the induction of apoptosis in DPTs.
Materials and Methods
DPT Purification. Thymi from 4-week-old C3H/HeN mice were teased in culture medium (RPMI supplemented with 10% heat-inactivated FCS, 100 μg/ml streptomycin, 10 mM HEPES, 0.1% nonessential amino acids, 1 mM sodium pyruvate, and 50 μM 2-mercapto-ethanol). After staining cells with FITC-conjugated anti-CD8 mAb (BD Biosciences PharMingen, San Diego, CA) and MACS MultiSort anti-FITC microbeads (Miltenyi Biotec Inc., Auburn, CA), CD8+ cells were isolated by using the MidiMACS LS+ positive selection column. Before elution of fixed cells, columns were washed twice. To wash away microbeads from selected cells, eluted cells were incubated with MACS MultiSort release reagent (Miltenyi Biotec Inc.) following manufacturer's instructions. Then, CD8+ cells were stained with anti-CD4 microbeads and passed through a MidiMACS LS+ positive selection column as above described. DPTs were considered pure population when cells resulted in >98% CD4+CD8+ cells upon flow cytometry.
Dex Treatment and Apoptosis Analysis. DPTs were plated in six-well plates in 3 ml of the culture medium at a concentration of 5 × 106 cell/ml together with 10–7 M Dex or a corresponding volume of phosphate-buffered saline. DPTs were kept at 37°C, 5% CO2 in a humidified atmosphere for 1, 3, 6, or 9 h. Apoptosis was analyzed with FACScan flow cytometry (BD Biosciences) and LYSIS II software after treating cells with hypotonic propidium iodide (Cifone et al., 1999). Phenotyping of treated thymocytes was performed with FACScan flow cytometry (BD Biosciences) and LYSIS II software after staining with phosphatidylethanolamine-conjugated anti-CD4 and FITC-conjugated anti-CD8 mAb (BD Biosciences).
Preparation of Complementary RNA (cRNA) and Chip Hybridization. The cRNA from 3-h-treated (medium or Dex) DPTs, to be used in hybridization of GeneChip (Affymetrix, Santa Clara, CA), was synthesized starting from total RNA. Briefly, preparation includes: synthesis of double-stranded DNA, purification of double-stranded DNA, synthesis of full-length cRNA, and nonenzymatic fragmentation of cRNA. Total RNA was isolated with TRIzol LS reagent (Invitrogen, Carlsbad, CA) following the manufacturer's instructions.
Double-stranded DNA was synthesized with the Superscript Choice System kit (18090-019; Invitrogen). Briefly, total RNA was used in a reverse transcription reaction to synthesize cDNA with a primer containing 24 thymidine residues and T7 RNA polymerase promoter sequences (T7-dT). T7-dT primer (1 μl) was added to 10 μg of total RNA and DEPC-H2O to a final volume of 11 μl. Then samples were heated for 10 min at 70°C, spun, and kept on ice. A second mix (4 μl of 5× first strand cDNA buffer, 2 μl of 0.1 M dithiothreitol, and 1 μl of 10 mM dNTP mix) was added to the sample and then heated at 42°C for 2 min. After spinning, 2 μl of SSII RT enzyme was added, and samples (20-μl final volume) were kept at 42°C for 1 h. To synthesize double-stranded cDNA, 150 μl of a third mix (91 μl of DEPC-H2O, 30 μl of 5× second strand reaction buffer, 3 μl of 10 mM dNTP mix, 1 μl of 10 U/μl DNA ligase, 1 μl of 10 U/μl DNA polymerase I, and 1 μl of 2 U/μl RNase H) was added, and samples were incubated for 2 h at 16°C. After the addition of 10 U of T4 DNA polymerase (2 μl), incubation was prolonged for 5 min further. Finally, reaction was stopped by adding 10 μl of 0.5 M EDTA.
Resulting double-strand cDNA was cleaned up with an equal volume of phenol/chloroform/isoamyl alcohol (25:24:1) and precipitated upon centrifugation (20 min) using 0.5 volumes of 7.5 M NH4Ac and 2.5 volumes of 100% ethanol. Pellet was washed with ice-cold 80% ethanol (500 μl). Dry pellet was resuspended in 12 μl of DEPC-H2O. cDNA was checked for appropriate length through agarose gel electrophoresis.
Synthesis of cRNA was performed with the Bioarray High Yield RNA Transcription Kit (Enzo Diagnostics, New York, NY). At first, we made a mix containing 4 μl of 10× High Yield reaction buffer, 4 μl of 10× biotin-labeled ribonucleotides, 4 μl of 10× dithiothreitol, 4 μl of 10× RNase inhibitor mix, 2 μl of 20× T7 RNA polymerase, and 17 μl of DEPC-H2O. The mix was added to 5 μl of double-stranded cDNA (40-μl final volume), and samples were heated at 37°C for 6 h. Then, RNAeasy spin columns (QIAGEN, Valencia, CA) were used to clean up cRNA following the manufacturer's instructions, and cRNA was first checked for quality through agarose gel electrophoresis and then quantified at the spectrophotometer. The quantification of cRNA was adjusted with the rule: adjusted cRNA yield = microgram of cRNA measured – 4.2.
The nonenzymatic fragmentation of cRNA was performed by mixing 8 μl of 5× fragmentation buffer (2.5 μl of 1 M Tris acetate, pH 8.1, 256 mg of MgOAc, 392 mg of KOAc, 8-μl final volume with DEPC-H2O) with 12 μl of DEPC-H2O and 20 μl of cRNA (1 μg/μl) and incubation at 94°C for 35 min.
Hybridization of 5 μg of cRNA, appropriate washing, and reading of the MGU74Av2 GeneChip was performed using Affymetrix devices following manufacturer's instructions in the Biopolo laboratories (Milan). In the same laboratories, the data from GeneChip reading were processed using MAS 5.0 software in Excel sheets with expression and modulation data (Microsoft, Redmond, WA).
Data and Statistical Analysis. Two independent cRNA from 3-h cultured DPTs (samples A and B) and two from 3-h 10–7 M Dex-treated DPTs (samples C and D) were analyzed using MAS 5.0 software (Affymetrix) and significance analysis of microarrays (SAM; Stanford University, Stanford, CA) (Tusher et al., 2001). Only genes resulting modulated in both methods were considered. Moreover, cut-off values of 1.7-fold for up-regulation and of 1.4-fold for down-regulation were adopted. A detail of statistical analysis is reported below.
Through MAS 5.0 Software, upon scaling, sample A was compared with sample C (comparison 1), sample B with C (comparison 2), sample A with D (comparison 3), and sample B with D (comparison 4) in the Biopolo laboratories. For each comparison, the software provides the fold change between untreated and treated samples (expressed as logarithm on 2 basis – log 2) and the absolute difference (I, increase; MI, marginal increase; NC, no change; MD, marginal decrease; D, decrease). We considered modulated the probe sets (i.e., genes) concordant (only I + MI or D + MD) with NC label no more than once in the four comparisons. Such a method gave 506 modulated genes (see Supplemental Data Fig. 1).
Through SAM Software, upon scaling, the expression data (called “average difference” by MAS 5.0 software) of each probe set in each array (A–D) were considered. As input information, it was specified that A and B expression data derived from untreated samples and C and D expression data from Dex-treated samples. Comparison was performed using the two classes unpaired data analysis, choosing a Δ of 0.88 (corresponding to a false discovery rate of 30.7%). As a consequence of the high false discovery rate, probe sets resulting modulated were 2013 (see Supplemental Data Fig. 1). However, the SAM analysis was useful to discard 121 probe sets accepted by MAS 5.0 analysis (Supplemental Data Fig. 1).
To evaluate the value of the increased or decreased expression, the mean of the fold modulation (in the four comparisons) deriving from MAS 5.0 software expressed as log2 was calculated. Then, the antilog of the absolute log 2 value was calculated. To further decrease the false positive rate and discard probe sets whose modulation was too little to represent a modulation with a biological meaning, only probe sets with a mean increase change equal or higher than 1.7-fold or a mean decrease change equal or higher than 1.4-fold were considered for further analysis. A lower cut-off for decreased gene expression was chosen considering the short incubation time and the half-life of several mRNA species, far beyond 3 h. For example, a gene with an mRNA half-life of 6 h and whose transcription has been completely inhibited by GC treatment shows a theoretic decreased expression of approximately 1.4-fold. On the contrary, increased expression is much less dependent on mRNA half-life. Our reasoning is somehow confirmed by data we obtained. In fact, probe sets with an increased expression equal or more than 2-fold were 51, whereas probe sets with a decreased expression equal or more than 2-fold were only 10. The use of fold change cut-off discarded a further 209 probe sets of 385 probe sets resulting modulated.
Gominer software (Zeeberg et al., 2003) was used to investigate the meaning of gene resulting modulated and participating to a similar function or cell structure. All gene information was related to each other using a compiled PERL program.
Real-Time PCR. Total RNA was isolated with TRIzol LS reagent (Invitrogen), following the manufacturer's instruction. cDNA was prepared by using 2.5 μg of Moloney murine leukemia virus reverse transcriptase (Invitrogen), and at the end of the reaction, 80 μl of water was added.
Real-time PCR was performed in a Chromo4 thermal cycler (Bio-Rad, Hercules, CA). Briefly, the investigated gene and the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (Gapdh) were amplified in the same tube in the presence of a 5-carboxyfluorescein-labeled TaqMan probe (investigated gene) and a VIC-labeled TaqMan probe (Gapdh). Primers, probes, and master mix were purchased from Applied Biosystems (TaqMan gene expression assay; Applied Biosystems, Foster City, CA). Reaction was performed in 20 μl following the manufacturer's instructions. To minimize variability, every time point was investigated with four replicates, and the amplification of two independent treatments was performed. The ΔΔCt method was used to determine modulation of the genes of interest (Livak and Schmittgen, 2001).
RNase Protection Assay. To perform RNase protection assay (RPA), the following probes were prepared: Bcl-x (U10101, 105–504 bp), Bim (NM_207680, 611–797 bp), Bfl-1 (NM_009742, 198–435 bp), and Bcl-x to investigate Bcl-x alternative isoforms (U51279, 445–919 bp). Moreover, probe for the housekeeping gene Gapdh (Ambion, Austin, TX) was used. DNA-free RNA was prepared according to the manufacturer's instructions (Ambion). RPA was performed using the RPAIII kit (Ambion) as previously described (Vecchini et al., 2005). Counts per minute from protected fragment and the picture presented in Fig. 3 were obtained using Instant Imager autoradiography system (PerkinElmer Life and Analytical Sciences, Boston, MA).
Results
Dex Modulates the Expression of 163 Genes in DPTs within the First 3 h of Contact. Thymocytes are very sensitive to GC treatment, and DPTs are the most sensitive thymus subpopulation to GC-induced apoptosis (Vacchio and Ashwell, 2000; Jondal et al., 2004). In fact, following in vitro Dex treatment, the percentage of DPTs decreased, whereas percentage of the other populations showed a relative increase (Fig. 1A). Therefore, to study the proapoptotic effect of Dex, we used DPTs.
It is well known that Dex-induced apoptosis requires gene transcription and protein synthesis (Distelhorst, 2002; Lepine et al., 2005). In fact, when an inhibitor of RNA synthesis (actinomycin D) was added to the thymocyte culture together with Dex, apoptosis was completely prevented (Fig. 1B). Apoptosis inhibition occurred when actinomycin D was added 2 but not 3 h after the addition of Dex (Fig. 1B), suggesting most of the mRNA expression sufficient for the induction of apoptosis is regulated within the first 3 h of Dex-cell contact. Therefore, we studied DPTs treated for 3 h with Dex comparing gene expression of Dex- and medium-treated DPTs.
Gene expression of DPTs was evaluated by using Affymetrix GeneChip microarray. Data were analyzed by using both the output of the MAS 5.0 software and by comparing the expression data through SAM; the probe sets (i.e., genes and gene families) selected by both analyses were considered modulated. By these criteria, 385 probe sets resulted to be significantly modulated (Supplemental Data Fig. 1 and Supplemental Data Table 1). However, when a cut-off for up-regulation was set to 1.7 factor and for down-regulation to 1.4 factor, only 176 probe sets (corresponding to 163 genes) resulted in being regulated, among which 59 were up-regulated and 104 were down-regulated (Supplemental Data Table 2). Presently, for 126 of those genes, a biological meaning has been conferred by the Gene Ontology Consortium (GO), as suggested by the assignment of one or more GO identification numbers (ID). On the basis of GO ID, the modulated genes can be grouped for their presence in one or more cellular components (Table 1), as having one or more molecular functions (Table 2) and as participating in one or more biological processes (Table 3). By interrogating GO database through GoMiner software, we evaluated how many genes, belonging to each GO ID, were modulated by Dex in comparison with all genes that were included in the same GO ID and were investigated with the GeneChip we used. Then, we evaluated whether the number of genes modulated in each GO ID was above the expected value and calculated the relative enrichment as shown in Tables 1, 2, 3. In particular, genes present in each cellular component were modulated by Dex treatment with a certain preference for cytoskeletal proteins and proteins localized in the nucleoplasm and nuclear membrane (Table 1). Surprisingly, mitochondrial proteins were less frequently modulated. Furthermore, Dex modulated preferentially the expression of genes encoding enzymes (including those with oxidoreductase activity), transporter with particular reference to the solute carrier family members, protein characterized by regulatory activity, and transcription factors (Table 2). In addition, Dex modulated genes involved in steroid metabolism, ion transport, microtubule-based processes, protein processing, signaling, and transcription regulation (Table 3). As a consequence, genes involved in cell proliferation, development, differentiation, and apoptosis resulted in wide modulation (Table 3). Our attention was focused on genes whose modulation contributes the explanation of the proapoptotic effect of Dex in DPTs, chosen on the basis of GO ID and on the function of modulated genes in T cells and thymocytes. By this way, we identified 25 genes belonging to apoptosis-related systems, active in GC-treated DPTs and potentially responsible for GC-induced apoptosis (Table 4).
Up-Regulation of Genes with Proapoptotic Function and Down-Regulation of Genes with a Protective Activity: Alteration in the Balance of Gene Expression of the Bcl-2 Family Members. The Bcl-2 family members are crucial for cell survival and death in several tissues. In DPTs, Bcl-xL and Bim are the main players since Bcl-xL is the dominant prosurvival member (Ma et al., 1995), and Bim is required for apoptosis of autoreactive thymocytes (Bouillet et al., 2002).
In Dex-treated DPTs, we found Bim and Bcl-x gene expression significantly up- and down-regulated, respectively. Moreover, Bfl-1/A1 was also slightly but significantly down-regulated (1.2-fold, under the cut-off) (Table 4). To confirm these results, we performed an RNase protection assay on Dex-treated DPTs (1–6 h) using Bim, Bcl-x, and Bfl1/A1 probes. For each probe, the ratio between the values expressed as counts per minute of the fragment protected by the chosen probe and that protected by Gapdh, a housekeeping gene that is not modulated in our system, was compared with untreated control. Results shown in Fig. 2 demonstrate that, upon Dex treatment, level of expression of Bim powerfully increases, whereas levels of Bcl-xL and Bfl1/A1 decrease significantly, confirming the GeneChip results.
It is known that Bcl-x gene is expressed in five isoforms, three of which (Bcl-xL, Bcl-xγ, and Bcl-xΔTM) are antiapoptotic, and two (Bcl-xS and Bcl-xβ) are proapoptotic (Yang et al., 1997). Since we found that in DPTs, level of Bcl-xβ expression is quite high, whereas that of Bcl-xS is low (data not shown), we investigated the modulation of Bcl-xβ and the antiapoptotic Bcl-x isoforms by RNase protection assay using a probe that detects the antiapoptotic splice variants into one single fragment and Bcl-xβ into another one. Then, the ratio between the protected fragment of Bcl-xβ and antiapoptotic splice variants was evaluated in untreated and Dex-treated samples. Figure 3 shows that the ratio Bcl-xβ/antiapoptotic splice variants after 1 to 6 h of Dex treatment increases. Thus, Dex modulates both Bcl-x gene and splice variant expression.
Alteration in Expression of Redox Status-Regulating Genes. One of the mechanisms responsible for GC-induced death of thymocytes has been identified in increased presence of reactive oxygen species (ROS) (Lepine et al., 2005). According to our study, three genes participating to the control of ROS are modulated in Dex-treated DPTs, thus potentially contributing to increased ROS presence. In particular, expression of thioredoxin reductase (Txnrd1) and NADP+-dependent isocitrate dehydrogenase decreased, and expression of Txnrd1 inhibitor (Txnip) increased (2.1-fold) (Table 4).
To confirm data from GeneChip array, we performed real-time PCR of selected genes, such as Txnrd1 and Txnip, also evaluating different treatment time points (1, 3, 6, and 9 h). For normalization of reactions and RNA quality, we used the housekeeping gene Gapdh that was amplified contemporary to the investigated gene. Real-time PCR confirmed the regulation of both Txnrd1 and Txnip. Txnrd1 reached the maximum decrease level after 6 h of Dex treatment (Fig. 4). On the contrary, Txnip kept on increasing, but if we consider that even medium treatment increased the expression of Txnip, the maximum effect of Dex is seen after 3 h (Fig. 4). The role in the redox imbalance of Txnrd1, Txnip, and NADP+-dependent isocitrate dehydrogenase is summarized in Fig. 5.
Alteration in Expression of Sphingomyelin Pathway Genes. It is well known that ceramide and sphingosine are involved in the GC-induced apoptosis of thymocytes (Cifone et al., 1999; Lepine et al., 2004, 2005). We found that three enzymes participating in the sphingomyelin pathway are modulated in Dex-treated DPTs, thus potentially contributing to an increased level of ceramide and sphingosine. In particular, UDP-glucose ceramide glucosyltransferase (Ugcg), which decreases the intracellular ceramide levels via glycosylation of ceramide, was down-regulated by a 1.9 factor (Table 4). Real-time PCR demonstrated that level of Ugcg increased during Dex treatment (Fig. 4). However, since Ugcg expression increased even more in untreated DPTs during medium treatment, the overall effect of Dex treatment is Ugcg down-regulation (Fig. 4).
Expression of sphingosine 1-phosphate phosphatase (Sgpp1), an enzyme catalyzing the transformation of sphingosine-1-phosphate (a lipid with antiapoptotic activity) in sphingosine (Lepine et al., 2005), was increased by a 1.8 factor. Real-time PCR shown in Fig. 4 demonstrates that increased mRNA levels of Sgpp1 was evident already after 1-h treatment and reached more than 8-fold increase at the 9-h time point. The increased intracellular concentration of ceramide seen during thymocyte apoptosis may be also favored by the significant but slight up-regulation (1.4-fold, under the cut-off) of dihydroceramide desaturase, isoform 1 participating to the de novo synthesis of ceramide (Table 4).
Finally, we found an increased expression of Gpr65 (or Tdag8), a proton-sensing and lysolipid-sensitive receptor (Radu et al., 2006), that may contribute to the apoptosis caused by the activation of ceramide pathway. In fact, its ligand psycosine is a sphingosine metabolite. The role of Ugcg, Sgpp1, dihydroceramide desaturase, isoform 1, and Gpr65 in favoring the proapoptotic activity of the ceramide pathway products is summarized in Fig. 5.
Modulation of Genes Involved in the Control of mRNA Stability and Protein Synthesis. The immediate early protein Tis11d (Zfp36l2) is a member of Tis11 family, responsible for apoptotic cell death that occurs starting from the mitochondrial death machinery and involved in mRNA destabilization (Ciais et al., 2004). Following Dex treatment, Tis11d was up-regulated 2.9- or 6.1-fold depending on the set probe considered (Table 4). Real-time PCR shown in Fig. 4 demonstrates that the increased mRNA levels of Tis11d were already evident after 1 h of treatment and kept on increasing during the treatment. The overall effect of Dex is even higher if we consider that Tis11d is down-regulated following incubation with medium alone. On the contrary, Tis11 is slightly down-regulated as confirmed by the overall effect of Dex investigated by real-time PCR (Fig. 4).
The regulated in development and damage response 1 gene (Rpt801, Ddit4, or Redd1) is an inhibitor of mTOR, a serine/threonine kinase working as a central regulator of protein synthesis (Brugarolas et al., 2004). Rpt801 was up-regulated in treated DPTs (Table 4) and in other T cells (Wang et al., 2003b). Functional meaning of Rpt801 up-regulation is uncertain. In fact, although its up-regulation upon stresses such as hypoxia and GC treatment makes likely the conclusion that RPT801 is a proapoptotic gene (Brugarolas et al., 2004), other data suggest that it may play an antiapoptotic role (Schwarzer et al., 2005).
Regulation of Genes Involved in Maturation of DPTs. Several genes participating in maturation and differentiation of DPTs demonstrated a decreased expression in Dex-treated DPTs (Table 4). Among them, we found: runt-related transcription factor 1 (or AML1), whose functional knocking makes thymocytes more sensitive to anti-CD3-induced apoptosis; the transcription factor Ets2, whose functional knocking makes thymocytes more sensitive to glucocorticoid-induced apoptosis; the suppressor of cytokine signaling 1 (Socs1), whose lack determines a decreased development and an increased apoptosis of thymocytes (Starr et al., 1998). Moreover, glucocorticoid treatment may directly and indirectly decrease the activity of other antiapoptotic genes (NF-κB, inhibitor of DNA binding 3, and Notch1) as briefly summarized below.
The expression of tumor necrosis factor α-induced protein 3 (or A20), a specific inhibitor of NF-κB, increased 2-fold. As a consequence, the activity of NF-κB (a powerful antiapoptotic factor) can decrease.
The expression of the inhibitor of DNA binding 3 (Id3), a repressor transcription factor crucial in thymocyte development and fate (Rivera et al., 2000), decreased. Real-time PCR demonstrated that the level of Id3 powerfully decreased after 1 h of contact and peaked after 3 h, reaching a 16-fold decrease (Fig. 4), which is in line with, but quantitatively different from, the value obtained with the GeneChip array (Table 4). Decreased expression of Id3 may be further amplified by the increased expression of dual-specificity phosphatase 2 (Dusp2, PacI), a negative modulator of ERK pathway involved in the control of Id3 (Bain et al., 2001). Real-time PCR confirmed overexpression of Dusp2 that peaked at the 6-h time point. Figure 5 summarizes the effect of Dex treatment on Id3-induced apoptosis.
The Notch pathway participates in lineage commitment, maturation, and survival in thymus and inhibits GC-induced apoptosis upon overexpression (Izon et al., 2002). Therefore, down-regulation of Notch1 and glycoprotein 130, a receptor subunit potentiating Notch1 pathway, seems particularly relevant for induction of DPT apoptosis (Fig. 5). Real-time PCR experiments confirmed the GeneChip data and demonstrated that Dex-induced Notch1 down-regulation peaked as early as 1 h of contact (Fig. 4).
Treatment with Dex Regulates the Expression of Several Genes Playing a Rescuing Activity in Dex-Treated DPTs. Although Dex treatment has a proapoptotic effect on DPTs, it regulates expression of genes potentially counteracting GC-induced apoptosis. First of all activity of GC receptor is regulated through overexpression (3.6-fold) of FKBP5 immunophilin, an inhibitor of the interaction between GCs and their receptor (Table 5). This effect was countered, at least in part, by the increase of GC receptor itself (2.0-fold) (Table 4). Moreover, modulation of several genes potentially exerting antiapoptotic effects seems to counteract the proapoptotic effect of Dex (Table 5). Of note is the modulation of promyelocytic leukemia zinc finger gene (or Zbtb16), a transcription repressor promoting stem cell growth. It resulted powerfully up-regulated following DPT treatment (10.7-fold) and virtually absent in untreated DPTs (not shown). Another gene with antiapoptotic functions is the interleukin-7 receptor gene (Franchimont et al., 2002) that resulted up-regulated (2.3-fold) in Dex-treated DPTs. We did not perform real-time PCR to confirm the modulation of the above-reported genes and of those listed in Table 5 since the regulation of most of them has been already described by other studies, cited in Supplemental Data Table 4.
Discussion
GCs mediate DPT's fate mainly through modulation of gene transcription as suggested by the inhibition of apoptosis during simultaneous exposure to Dex and actinomycin D (Fig. 1) (Distelhorst, 2002). Our analysis, indeed, revealed several apoptosis-related genes whose transcription is modulated in DPTs following Dex contact for 3 h (Fig. 4). This time is sufficient to trigger apoptosis. In fact, actinomycin D does not inhibit apoptosis when added 3 h after Dex. Three hours of treatment is short enough to minimize the possibility of indirect transcriptional effects. Therefore, we suppose that most of the described genes are directly regulated by Dex. Moreover, at this time, the apoptosis process has not been executed yet, and the cell components, including mRNA species, are still intact, avoiding a technical bias.
Approximately one-fifth of the modulated genes can be grouped into five systems that seem to contribute to Dex-induced apoptosis of DPTs (Table 4; Fig. 5). First of all, nine genes participating in lineage commitment, maturation, and survival of thymocytes are regulated. Even if the modulation tends toward normal levels after 6 to 9 h of treatment (Id3 and Notch1) or is not impressive from the quantitative point of view (e.g., Socs1 and glycoprotein 130), the overall decrease of antiapoptotic messages during the first hours of Dex contact could make DPTs more sensitive to proapoptotic signals, thus favoring apoptosis.
We also described the modulation of several genes somehow related to mitochondrion: Bcl-2 and Tis11 family members and genes involved in ROS detoxification. These genes may be responsible for the early mitochondrial alteration leading to the death decision (Lepine et al., 2005). Another group of Dex-regulated genes participates in the ceramide pathways and modulates apoptosis without being linked to mitochondrial machinery (Cifone et al., 1999; Lepine et al., 2004).
We were able to link the above-described genes to apoptosis on the basis of what is known so far, but it is very likely that future functional studies will shed light on other genes modulated by Dex in DPTs (Supplemental Data Table 2) contributing to their apoptosis. For example, the proteins of cytoskeleton and those involved in protein depolymerization, including ubiquitin cycle, are frequently modulated genes (Tables 1 and 3), representing new fields of investigation.
Data concerning Bcl-2 family may suggest that the GC effect on thymocytes is due to an overall imbalance more than to the regulation of a single gene. In support of this hypothesis is a study demonstrating that mature T cells of Bim-deficient mice are resistant to SEB-induced deletion (Hildeman et al., 2002). This observation suggested that Bim modulation was crucial for the induction of cell death in this system. However, following SEB injection, Vβ8+ wild-type cells expressed levels of Bim similar to those of untreated cells, demonstrating that regulation of Bim expression is not involved in SEB-induced apoptosis. On the contrary, decreased levels of Bcl-2 were detected, suggesting that Bcl-2 down-regulation facilitates the proapoptotic effect of Bim in those cells (Hildeman et al., 2002).
Concerning GC effects on T cells, despite Bim modulation may be crucial for GC-induced apoptosis, as already reported (Wang et al., 2003a), Bim-deficient thymocytes are only partially resistant to Dex (Bouillet et al., 1999), suggesting that Bim modulation alone is not sufficient to explain thymocyte apoptosis. The hypothesis of imbalance of Bcl-2 family members (Bcl-2, Bcl-xL, and Bax) has been proposed to explain the GC-induced apoptosis in granule cells of the hippocampus (Almeida et al., 2000), which is in line with our data but suggests that GC-induced gene regulation is peculiar to each tissue.
We also demonstrate that the ratio between the Bcl-x antiapoptotic splice variants and the proapoptotic splice variant Bcl-xβ changes upon Dex treatment. Modulation of Bcl-x gene splice variants has been described in endometrial cells treated with progestins, where it plays a determinant protective role (Pecci et al., 1997), suggesting that splice variant modulation is a possible mechanism for the induction of apoptosis.
Treatment of DPTs with Dex has a clear proapoptotic effect, as discussed. Nevertheless, some studies, including ours, demonstrate that GCs regulate the expression of genes counteracting apoptosis (Table 5) (D'Adamio et al., 1997; Franchimont et al., 2002; Schmidt et al., 2005). This paradoxical phenomenon needs to be explained. In some cases, it is possible that genes do not play the same role in thymocytes and other cells, including lymphocytes. In fact, the functional role of a protein may differ in cells from various origins and within the same cells in different microenvironmental contexts, depending on the cell state of activation/differentiation. In particular, cell death may follow activating stimuli as can be seen in the activation-induced cell death of mature T lymphocytes. This may apply to Dex-treated DPTs in which several proapoptotic systems have been triggered. A known example of a different role played by a gene involved in the activation process is represented by Socs1, whose lack means higher activation in mature T cells and higher apoptosis level in thymocytes (Starr et al., 1998; Chong et al., 2005). Moreover, several lines of evidence suggest that GCs cover a multifaceted function in thymus development, promoting thymocyte expansion more than apoptosis when present at physiological concentrations (Jondal et al., 2004). Therefore, it can be hypothesized that those genes whose modulation seems to have an antiapoptotic action may favor thymocyte survival when low concentrations of GCs are present in the context of the thymic stroma.
The list of genes modulated by GCs has been continuously updated so far, and hundreds of genes are known to be modulated by GCs (Smith and Herschman, 2004; Ploner et al., 2005). In addition, approximately one-third of the apoptosis-related genes that we found to be modulated in Dex-treated DPTs were already known to be regulated by GCs (see Tables 4 and 5). Considering the number of studies dealing with GCs and T cells, the finding of several genes never described as modulated may sound bizarre if we hypothesize that the mechanism of action of GC is similar in different cell types. However, this hypothesis may be questionable. In fact, composition and proportion of individual isoforms of GC receptor expressed in particular cellular contexts may change GC receptor function (Zhou and Cidlowski, 2005). Moreover, we know that the GC receptor modulates gene transcription by several ways, including protein-protein interaction with other transcription factors and coregulators (Reichardt et al., 1998; Zhou and Cidlowski, 2005). Therefore, it is likely that the effect of GC treatment is different according to the functional status of the cells, determined, in turn, by the type and the amount of transcription factors and coactivators that are active at a specific time. This is clearly exemplified by cells, such as endometrial cells, in which GC treatment exerts an antiapoptotic effect (Pecci et al., 1997).
This reasoning may explain, at least in part, why different studies using global gene profiling, including ours, found different modulated genes. In our opinion, this depends on the cells used and their activation and/or differentiation status more than on technical bias. In fact, in the attempt to find the core gene pattern responsible for the proapoptotic effect of GC in lymphocytes, we tested some primary cultures from human malignant lymphocytes and a murine thymus-derived cell line. However, analysis of data revealed that only a few modulated genes were common to the different systems tested (data not shown).
Another reason explaining discrepancies between our study and others is the choice of a short time of GC-cell contact. In fact, although some genes increase their levels of modulation along with the time of contact, others (e.g., Id3 or Notch1) recover to normal expression levels soon after modulation. Since the death machinery is activated during the first hours of GC-DPT contact, these genes play a role similar to those increasingly modulated but are barely considered by the studies using a long time of treatment. Furthermore, studying cells treated for several hours with GCs leads to the identification of genes that are modulated by the transcription factors regulated by GC (i.e., studying secondary effects of GC). Finally, we used medium-treated cells as control, instead of untreated cells. In our opinion, this is the best way to discriminate the GC effects from those of the stress occurring before and during incubation. If an untreated sample is used as a control, one may run the risk of underestimating the results (see Tis11d; Fig. 4) or even reach wrong conclusions (see Ugcg, apparently up-regulated upon Dex treatment; Fig. 4).
In the recent and less recent past, several groups (including ours) have spent their efforts in the attempt to find the crucial proapoptotic event (including increase or decrease of gene expression) responsible for the GC-induced apoptosis. Our study indicates that several genes, participating in different cellular functions, are modulated by the GC treatment. Even if it does not formally demonstrate the network hypothesis, it is in accordance with it, further suggesting that several genes cooperate in inducing GC-induced apoptosis of T cells.
Footnotes
-
This work was supported by the Associazione Italiana Ricerca sul Cancro.
-
R.B. and G.N. contributed equally to this work.
-
The data discussed in the article have been deposited in the National Center for Biotechnology Information's Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE 5463.
-
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
-
doi:10.1124/jpet.106.108480.
-
ABBREVIATIONS: GC, glucocorticoid hormone; DPT, CD4+CD8+ double-positive thymocyte; Gpr65, G protein-coupled receptor 65; Dex, dexamethasone; FITC, fluorescein isothiocyanate; cRNA, complementary RNA; DEPC, diethyl pyrocarbonate; SAM, significance analysis of microarrays; PCR, polymerase chain reaction; Gapdh, glyceraldehyde-3-phosphate dehydrogenase; RPA, RNase protection assay; GO, Gene Ontology Consortium; ID, identification number(s); ROS, reactive oxygen species; Txnrd1, thioredoxin reductase; Txnip, thioredoxin reductase inhibitor; Ugcg, UDP-glucose ceramide glucosyltransferase; Sgpp1, sphingosine 1-phosphate phosphatase; Socs1, suppressor of cytokine signaling 1; NF, nuclear factor; Rpt801, regulated in development and damage response 1 gene; Id3, inhibitor of DNA binding 3; Dusp2, dual-specificity phosphatase 2; SEB, staphylococcal enterotoxin B.
-
↵ The online version of this article (available at http://jpet.aspetjournals.org) contains supplemental material.
- Received May 28, 2006.
- Accepted August 15, 2006.
- The American Society for Pharmacology and Experimental Therapeutics