To examine the role of histone acetylation on the differentiation state of mouse ES cells, we employed the histone deacetylase inhibitor TSA. We first tested the effect of different TSA concentrations on the mouse CGR8 ES cell line cultivated in the presence of LIF. When 10-50 nM concentrations were used, we observed morphological changes that depended on the concentration and duration of the treatment (not shown). Figure 1a shows a phase contrast morphology and alkaline phosphatase staining (ALP) of mES cells subjected to 50 nM TSA for 12 h. Control cells form round and compact colonies, which stain > 90% positive for ALP. After treatment with 50 nM TSA for 12 h, 70% of the colonies are disrupted and the cells become flattened and negative for ALP; the other 30% show a loose morphology containing a mixed population of ALP-positive and -negative cells.

To identify possible molecular changes underlying the TSA-induced phenotypic transformation, we analyzed the expression of Nanog, one of the master regulators of ES cell identity. Nanog is down-regulated by TSA in a dose-dependent manner, with a concentration of 50 nM having maximal effect (Figure S1A in Additional data file 1). Figure 1b shows that Nanog mRNA levels decline very rapidly starting within the first 2 h of treatment with 50 nM TSA and are minimal by 4 h. Nanog protein levels drop with slower kinetics (Figure 1b). Similar to Nanog, Oct4 and Sox2 mRNA (Figure S2 in Additional data file 1) and protein levels (Figure S1B in Additional data file 1) were reduced during TSA treatment.

In order to test if the rapid Nanog suppression is due to transcriptional silencing, we examined the effect of TSA on the activity of different Nanog promoter fragments cloned in front of the luciferase reporter gene (Figure 1c). Two Nanog promoter fragments extending 966 or 220 bp upstream from the transcriptional start site are both suppressed by TSA (Figure 1c, -966, -220). The proximal promoter has higher activity, possibly because it is deprived of negative regulatory elements that reside within upstream regions. Interestingly, when the Nanog enhancer, located 5 kb upstream of the transcription start site and containing a Nanog auto-regulatory site 28, is fused to the -966 promoter (Figure 1c, enh/-966), it produces a stronger element that is more robustly repressed by TSA than either of the two promoter fragments (Figure 1c). These findings suggest that loss of Nanog expression after TSA treatment is due to transcriptional repression. Moreover, it appears that the effect is mediated by both the proximal promoter, which harbors a composite Oct-4/Sox2 binding site known to regulate Nanog expression 2930, and the distal enhancer where Nanog binding sites reside 28.

The observed morphological effects of TSA and Nanog repression might be indicative of a global and rapid assault on the self-renewal capacity of ES cells, with possibly a simultaneous launch of differentiation. To test this idea, we performed Affymetrix microarray analysis using RNA samples isolated after TSA treatment of ES cells for 6 h to gauge early events and 12 h to test putative secondary effects. The data show extensive changes in the ES transcriptome (Additional data file 2), suggesting that a large fraction of the genome was transcriptionally reprogrammed. Of the differentially expressed genes in TSA-treated compared to control ES cells, 792 genes were down-regulated and 1,376 up-regulated at the 2-fold change cut-off value (Additional data file 2). The gene chip data were validated by real time RT-PCR for 20 selected genes (Figure S3 in Additional data file 1).

A selection of up- and down-regulated genes is presented in Table 1. The category of down-regulated genes contains those encoding important regulators of pluripotency, including Nanog, Sall4, Klf4, Sox2 and Oct4, and other genes typical of the undifferentiated state, such as Rex1/Zfp42, FoxD3, Gdf3, Nr0b1, Eras, Rif1, Tbx3 and Esrrb 2. It also contains genes encoding a group of chromatin and transcription regulators, such as the histone acetyltransferase PCAF, the H3K9 methyl transferase Suv39, the H3K9 demethylases Jmjd1a and Jmjd2c, the H3K27 demethylase Utx, the Polycomb factors Bmi1, Cbx5, Suz12 and Eed (Table 1) and the transforming growth factor-β/activin signaling pathway members Inhbb, Gdf3 and Lefty2. Among the early and strongly TSA-supressed genes are Sall1, Gli2 and Klf2, which have not been previously associated with regulation of pluripotency and may be novel candidates.

In the category of up-regulated transcripts, we detected: genes of the neural lineage, such as Hoxa1, Hoxb13, Nnat, and Mbp; genes of the hematopoietic lineage, for example, Mlf1; vascular and neuronal differentiation related genes like Pdgfrβ; a group of genes encoding histones, including the differentiation-specific histone H1f0; and genes encoding the connective tissue growth factor Ctgf and the endothelial-specific receptor Edg3. In addition, TSA activates the immediate-early response genes Egr1, Fos and JunB, which have been associated with cell proliferation, differentiation, transformation and apoptosis.

To narrow down the genes under study, we focused on the most significant changes and chose to analyze genes with expression changes equal to or greater than four-fold. Using this gene list we performed a hierarchical clustering in order to uncover genes that respond similarly to TSA treatment, pointing to a possible functional interconnection (Figure 2a; Figure S4 in Additional data file 1). This analysis showed the existence of two major clusters of down-regulated (cluster 1) and up-regulated (cluster 2) genes and unveiled a further division of each cluster into two subclusters (Additional data file 3). Subclusters 1b (117 transcripts) and 2a (111 transcripts) show major changes at 6 h whereas subclusters 1a (60 transcripts) and 2b (112 transcripts) do so at 12 h. To functionally categorize the gene clusters and subclusters, we used the Database for Annotation, Visualization, and Integrated Discovery (DAVID) 31 to obtain Gene Ontology annotations for the category of 'biological process' (Table 2). Down-regulated transcripts (subclusters 1a and 1b) contain genes that fall in the categories of metabolism (Cbr3, Tdh, Enpp3, Cacna1a, Cul1), development/morphogenesis (Nanog, Nr0b1, Sall1, Gli2, Lefty1, Lefty2) and growth (Gdf3, Gja1, Socs2, Inhbb). In addition, genes from subcluster 1b fall in the category of transcription (Tcfap4, Gtf2I, Ubtf, Suv39h1). Up-regulated genes from the early-induced subcluster 2a participate in neural system development (Sema4f, Hoxa1, Stxbp1), whereas subcluster 2b (induced after 12 h) contains genes that take part in angiogenesis and hemopoiesis. Additionally, subcluster 2a members participate in cell organization/biogenesis (CenpJ, Tubb2a, Sept4) and intracellular signaling (Edg3, Rnd1, Mknk2), while subcluster 2b members have been implicated in metabolism (Hsdl2, Tgm2, Pygl), chromosome organization, that is, nucleosome and chromatin assembly/disassembly, (H1h2bf, H1h2bc, H12bp, H2h2be, H1fx) and antigen processing (H2-T3, H2-K1, CD74).

An overview of the various gene function categories is shown as pie charts for clusters 1 and 2 in Figure 2b. Comparison of the pie chart representations between up- and down-regulated genes reveals some interesting differences. For example, regulators of cell cycle, cell growth and transcription are represented in the cluster of down-regulated genes, suggesting a strong effect of TSA treatment on the self-renewal machinery. On the other hand, signaling and adhesion molecules become evident in the pie chart of induced genes, indicating the appearance of new response mechanisms to environmental cues. A full list of genes from each subcluster and the corresponding biological process annotations are included in Additional data files 3-7.

To examine how gene expression modulation caused by TSA corresponds to changes taking place during the 'natural' in vitro differentiation process, we placed ES cells in hanging drops to form embryoid bodies (EBs) and allowed them to differentiate without LIF. We then analyzed the mRNA levels of six genes from Table 1 that are strongly affected, negatively or positively, by TSA. These genes were those encoding Nanog, the spalt homologue Sall1, the orphan nuclear receptor Nr0b1, the vascular and neuronal differentiation related receptor Pdgfrβ, and the hematopoietic lineage switch gene Mlf1 and the homeotic Hoxa1 gene. Figure 3a shows that TSA causes a very rapid repression of Nanog, Nr0b1 and Sall1, a rapid but more gradual induction of Pdgfrβ and Hoxa1, and a late increase of Mlf1 mRNA levels.

During embryoid body differentiation, suppression or induction of individual genes takes place in distinct time frames. To compare the TSA effect with EB differentiation, days 4 and 8 were chosen as the most indicative even though gene expression changes were observed earlier (EBs days 2 and 3). Nanog, Nr0b1 and Sall1 were also found to be repressed during EB formation whereas Pdgfrβ, Mlf1 and Hoxa1 genes were activated. (Figure 3b). Therefore, the expression of these genes in embryoid bodies is regulated in the same direction as after TSA treatment, albeit at much slower paces (Figure 3b). This was true for the majority of genes checked so far.

We next asked if the differentiation imposed by TSA is reversible. To answer this question, ES cells were treated with TSA for 12 h and then cultured for an additional 6 and 12 h without TSA (ES re). We observed that the morphological changes induced by TSA were gradually reversed with the emergence of compact colonies (approximately 70% of the control), which are indicative of the undifferentiated state (Figure 1a, ES re). These colonies stain weakly for ALP in their center (Figure 1a). RT-PCR analysis demonstrated that the expression of Nanog, Mlf1, Hoxa1 and Pdgfrβ was fully reversed to pre-treatment, undifferentiated ES cell levels (Figure 3a). In contrast, the mRNA levels of Nr0b1 and Sall1 did not recover fully. In agreement with the reduction of Nanog mRNA, FACS analysis showed that the population of cells expressing Nanog above control antibody levels was reduced by TSA and then recovered upon release from TSA treatment (Figure S5 in Additional data file 1). Further analysis of three well characterized pluripotency factors, Oct4, Sox2 and Zfp42/Rex1, revealed that the mRNA levels of Oct4 and Sox2 but not of Zfp42/Rex1 were restored after TSA removal (Figure S2 in Additional data file 1). In conclusion, although the morphological changes that are caused by TSA treatment are largely reversed, expression of individual genes can undergo either fully or partially reversible alterations. The partial return to the ES cell phenotype suggests that histone deacetylase inhibition alone does not appear to fully commit ES cells to differentiation.

In order to examine the competence of the 'ES re' cells to contribute to the three germ layers, we placed them in hanging drops and observed that they formed EBs of normal morphology. We then checked for expression of markers of the three germ layers, endoderm (Sox17), mesoderm (Brachyury) and ectoderm (Mash1). As shown in Figure 4, Sox 17 and Mash1 followed the same expression patterns as in wild-type cells whereas Brachyury was up-regulated one day later than in wild-type cells. These results show that 'ES re' cells are still capable of acquiring either one of the three cell fates.

To gain insight into the molecular mechanisms whereby TSA triggers stem cell differentiation, we analyzed the dynamics of histone H3 modifications after TSA treatment and compared them to changes taking place during EB formation. In both cases, we observed an increase in the global amounts of two activatory modifications, H3 acetylation and K4 trimethylation and a decrease in repressive K27 trimethylation (Figure 5). These results show that TSA treatment instigates a global enhancement of activation-linked epigenetic marks that also appears during the natural ES cell differentiation process.

However, analysis of individual genes using ChIP assays uncovered a complex, gene-specific pattern of histone modifications (Figure 5). In this experiment, we first examined the histone modifications on the promoters of three genes that were up-regulated by TSA, namely Pdgfrβ, Mlf1 and Hoxa1 (Figure 3a). During activation of Pdgfrβ and Mlf1 by TSA, we detected an increase in H3 acetylation and H3K4 trimethylation and a decrease in H3K27 trimethylation (Figure 6a). Analysis during EB formation (Figure 6b) gave similar results. In contrast, we found that the neural lineage gene Hoxa1 had significant concurrent activatory (K4) and suppressive (K27) methylations in the undifferentiated state (in agreement with the bivalent structure model 20). Upon activation by TSA, we observed a reduction in K27 trimethylation levels and maintenance of K4 trimethylation and H3 acetylation levels. Thus, activation of Hoxa1 expression after TSA treatment relies on loss of suppressive modifications. Hoxa1 expression during EB differentiation is correlated with a transient increase in H3 acetylation, matching the Hoxa1 maximal activation (Figure 3a), and a transient decrease in K27 trimethylation (Figure 6b). In contrast to this finding, another study 32 has reported that TSA-induced activation of Hoxa1 is not correlated with a drop in K27 trimethylation. However, in that case, ChIP analysis was performed on a retinoic acid-regulated enhancer at the 3' end of the gene as opposed to our analysis, which is based on the 5' promoter region, proximal to the transcriptional start site.

ChIP analysis was also performed for genes down-regulated by TSA treatment. Along with Nanog mRNA repression, we detected a gradual decrease of H3 acetylation and an abrupt drop in H3K4 trimethylation of its promoter without an increase in the repressive H3K27 trimethylation (Figure 6a). During EB differentiation, Nanog inactivation similarly correlates with a decrease in both H3 acetylation and H3K4 trimethylation, but in this case gene suppression is also accompanied by an increase in H3K27 trimethylation (Figure 6b). Unlike Nanog, repression of Nr0b1 was accompanied by a robust increase in H3K27 trimethylation, and no significant decrease of either acetylation or H3K4 trimethylation (Figure 6a). In EBs, H3K27 trimethylation was also increased but both activatory modifications were reduced. Finally, the Sall1 promoter represented an intermediate situation of repression, connected to a strong decrease in acetylation, a decrease in H3K4 trimethylation and a rise in H3K27 trimethylation (Figure 6a). Similar changes accompany Sall1 deactivation during EB differentiation.

Trimethylation of K27 is catalyzed by Enhancer of Zeste 2 (Ezh2), a methyl-transferase component of the PRC2 complex. Employing ChIP assays, we confirmed the recruitment of Ezh2 on Nanog, Sall1 and Nr0b1 promoters (Figure 7a), in agreement with the appearance of K27 trimethylation during suppression either by TSA or in EBs (Figure 6a,b). Gene activation or repression of all six genes in both TSA-treated ES cells and EBs correlates with a respective increase or decrease in promoter-bound RNA polymerase II (Figure 6a,b). These results indicate that the observed chromatin modifications correlate well with the expected recruitment of transcriptional regulators and enzymes to the corresponding gene promoters.

Correlating ChIP data with changes in mRNA levels, it appears that H3K27 trimethylation might predispose individual genes for stable repression. For example, Sall1 and Nr0b1, which show an increase in H3K27 trimethylation following addition of TSA (Figure 6a), do not regain full expression after release from TSA (Figure 3a). Nanog, on the other hand, which is not marked by H3K27 trimethylation when repressed, regains full expression following TSA release. To further strengthen this idea, we prepared chromatin samples from cells treated with TSA for 12 h and then released for 6 and 12 h. ChIP analysis shows that H3K27 trimethylation induced by TSA is maintained on the Nr0b1 promoter and, partially, on the Sall1 promoter even after TSA removal (Figure 7b), in agreement with the irreversible repression of the two genes (Figure 3a). Collectively, by analyzing six different genes during TSA treatment and EB formation, we have encountered similar but gene-specific combinations of promoter chromatin modifications that correlate with expression state.

CGR8 ES cells were cultivated in GMEM (10% fetal bovine serum, 1,000 units LIF (ESGRO-Chemicon Temecula, CA, USA)). Antibodies were from Upstate (Upstate/Millipore, Billerica, MA, USA) (AcH3, 3mK4H3, 3mK27H3, Ezh2), Chemicon (Nanog) and Santa Cruz (Santa Cruz, CA, USA) (His, RNAP II). TSA was from Sigma (Sigma, Saint Louis, MO, USA).

CGR8 cells were fixed with 100% methanol and stained with a solution of 1 mg/ml Fast Red TR salt TM (Sigma) and 200 μg/ml Napthol AS-MX phosphate (Sigma) in 0.1 M Tris pH 9.2.

Nanog promoter/enhancer fragments were cloned upstream of the luciferase reporter gene in the pGL3-basic vector (Promega Madison, WI, USA). A -966/+50 fragment was obtained by PCR and was cloned in the pGL3 vector. Primers used were: forward, 5'-AGCACAAGGACTGATCGG-3', reverse, 5'-GCAGCCTTCCCACAGAAAG-3'. The enhancer fragment was obtained by PCR and was cloned in front of the -966 fragment in the pGL3 vector. Primers used were: forward, 5'-ATATAGGTACCCCCCTCCCCCACCTGTCCC-3', reverse, 5'-TATATGCTAGCG GCCACATAGCCTTAAGT-3'. For -220/+50 construction, -966/+50 was digested with HindIII and the excised fragment was cloned in the pGL3 vector. CGR8 cells were transfected using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). The full length Nanog cDNA plasmid was kindly provided by P Savatier.

ES cells and EBs were fixed with 1% formaldehyde for 10 minutes at room temperature and the reaction was quenched by adding glycine to a final concentration of 0.125 M. ES cells were washed once with ice-cold phosphate-buffered saline (PBS), harvested and washed two more times. EBs were also washed three times with ice-cold PBS. ES cells and EBs were resuspended in lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl pH 8.0, 1 mM phenylmethylsulphonyl fluoride (PMSF), 1 ml/10⁶ cells) and incubated on ice for 10 minutes. The suspension was sonicated 5 times for 1 minute each and 10 μl samples were analyzed by gel electrophoresis (1.5% agarose). Properly sonicated samples were centrifuged at 14,000 rpm, 4°C for 15 minutes and the supernatant was stored at -80°C. We kept 10 μl from each sample as input and 50 μl were immunoprecipitated with 5 μg of relevant antibodies in RIPA buffer (1% Triton X-100, 0.1% deoxycholate (DOC), 140 mM NaCl, 1 mM PMSF) overnight at 4°C under rotation. Protein G beads were incubated in the same conditions with 100 μg/ml sonicated salmon sperm DNA and 1 μg/ml bovine serum albumin in RIPA buffer. Blocked beads and immunoprecipitated samples were combined next day and were incubated under rotation for 3 h at 4°C. The immunoprecipitates were then washed 7 times with RIPA wash buffer (1% Triton-X, 0.1% DOC, 0.1% SDS, 500 mM NaCl, 1 mM PMSF). Input samples and beads were resuspended in 100 μl of 10 mM Tris-HCl pH 8.0, 1 mM EDTA (TE) buffer supplemented with 0.5% SDS and proteinase K to a final concentration of 200 μg/ml and incubated for 3 h at 55°C and overnight at 65°C. The next day samples were phenol-chloroform extracted and ethanol immunoprecipitated with NaOAc and 20 mg of glycogen as a carrier. DNAs from input and immunoprecipitate pellets were resuspended in 50 μl and 250 μl of TE buffer, respectively. The DNA content was analyzed using real-time PCR (5 μl/20 μl reaction). The primers used were: Nanog promoter sense, 5'-CTTACTAAGTAGCCCAGTC-3'; Nanog promoter antisense, 5'-GTTTATACACGGTTCTTT-3'; Nr0b1 promoter sense, 5'-AGTTGGAACAGAGCCCTAAC-3'; Nr0b1 promoter antisense, 5'-GCCTTTGGTTGAATGTG-3'; Sall1 promoter sense, 5'-TGCGACATGGGTCCTGAG-3'; Sall1 promoter antisense, 5'-AATTCTGGAGCGCCTTTGAGT-3'; Hoxa1 promoter sense, 5'-GAGCGCGCGTCACCTACAC-3'; Hoxa1 promoter antisense, 5'-CTGAGCCGCCTGCGAAAGTT-3'; Pdgfrb promoter sense, 5'-GCAGGCAGGAGACTGACGA-3'; Pdgfrb promoter antisense, 5'-AGTCCCGGCTACCCTATCTGG-3'; Mlf1 promoter sense, 5'-TGCCATAGCAGCCGAGCGAT-3'; Mlf1 promoter antisense, 5'-GCTTGACGCAGGCCGTTTC-3'.

RNA was prepared using the Trizol reagent (GIBCO-BRL, Invitrogen Corp., Carlsbad, CA, USA). RT reactions were performed with Moloney murine leukemia virus from Finnzymes (Finnzymes, Espoo, Finland). SYBR Green I and Opticon monitor system from MJ Research (MJ Research, Waltham, MA USA) were used for real-time PCR reactions. Primers used for RT-PCR reactions were: Nanog sense, 5'-CGCTGCTCCGCTCCATAACT-3'; Nanog antisense, 5'-GCGCATGGCTTTCCCTAGTG-3'; Nr0b1 sense, 5'-CTGGTGTGCAGCGTCTGA-3'; Nr0b1 antisense, 5'-GTGTTGGTCTCCGGATCTC-3'; Sall1 sense, 5'-AGTTCTCCCAGGAGGCGAGGTG-3'; Sall1 antisense, 5'-GGTTGGCAGATGTTCGTAAAGT-3'; Hoxa1 sense, 5'-GGTCAACCCAACGCAGTG-3; Hoxa1 antisense, 5'-TGCTTCATGCGGCGATT-3'; Pdgfrβ sense, 5'-GACTACCTGCACCGGAACA-3'; Pdgfrβ antisense, 5'-GGGACTCAATGTCTGCGTATT-3'; Mlf1 sense, 5'-GAACCCATAATCGTCGAG-3'; Mlf1 antisense, 3'-CTTCGGGTTTGAGTTGAG-3'; Sox17 sense, 5'-CTCTGCCCTGCCGGGATGG-3'; Sox17 antisense, 5'-AATGTCGGGGTAGTTGCAATA-3'; Brachyury sense, 5'-GCGAGCTGGGTGGATGTAGA-3'; Brachyury antisense, 5'-CAAGGCGGCACAAGACTAAGTC-3; Mash1 sense, 5'-CCACCATCTCCCCCAACTA-3'; Mash1 antisense, 5'-CTGGGCTAAGAGGGTCGTAGG-3'; Hprt sense, 5'-CTCCTCAGACCGCTTTTTG-3'; Hprt antisense, 5'-TCCTCGGCATAATGATTAGG-3'.

CGR8 cell extracts were prepared using a lysis buffer containing 50 mM Tris 8.0, 170 mM NaCl, 50 mM NaF, 0.5% NP-40, 1 mM PMSF. For histone extraction, cells were harvested, washed three times with ice-cold PBS supplemented with 5 mM Na butyrate and lysed with a Triton extraction buffer (PBS supplemented with 0.5% Triton, 2 mM PMSF, 0.02% NaN3) at a cell density of 10⁷ cells/ml. The lysate was centrifuged at 2,000 rpm for 10 minutes at 4°C, and the pellet was washed twice with a half Triton extraction buffer volume and then resuspended in 0.2 N HCl at a cell density of 4 × 10⁷ cells/ml. Histones were acid extracted overnight at 4°C. The next day samples were centrifuged at 2,000 rpm for 10 minutes at 4°C and the supernatant was dialyzed against CH₃COOH and H₂O. Samples were kept at -80°C with 20% glycerol.

Total RNA from ES cells was isolated using the RNeasy Mini Kit from QIAGEN (QIAGEN GmbH, Hilden, Germany) and treated with RNase-free DNase I (5 U/100 μg of nucleic acids, Sigma). Biotinylated cRNA was prepared according to the standard Affymetrix protocol. In brief, double-stranded cDNA was synthesized from 10 μg total RNA using the SuperScript Choice System from Invitrogen and the Affymetrix T7-(dT)₂₄ primer, which contains a T7 RNA polymerase promoter attached to a poly-dT sequence. Following a phenol/chloroform extraction and ethanol precipitation, the cDNA was transcribed into biotin-labeled cRNA using the Retic Lysate IVT™ kit (Ambion Inc., Woodward Austin, TX, USA). cRNA products were purified using the RNeasy kit (QIAGEN) and fragmented to an average size of 30-50 bases according to Affymetrix recommendations. Fragmented cRNA (15 μg) was used to hybridize the Mouse Genome 430 2.0 Array for 16 h at 45°C. The arrays were washed and stained in the Affymetrix Fluidics Station 400 and scanned using the Hewlett-Packard GeneArray Scanner G2500A.

The image data were analyzed with the GeneChip^® Operating Software (GCOS) using Affymetrix default analysis settings. After a quality control test, arrays were normalized by log scale robust multi-array analysis (RMA) 44.

A parametric ANOVA (F-test) was performed. The false discovery rate of the resulting test-set was calculated using the Benjamini Hochberg procedure 45. An f-test p-value < 10^-2 corresponding to a false discovery rate of 4.066 × 10^-2 and a fold change > 2 was used in order to select transcripts appearing in Additional data file 2 (792 down-regulated and 1,376 up-regulated transcripts). For the hierarchical clustering, a f-test p-value < 10^-3, corresponding to a false discovery rate of 7.833 × 10^-3 and a fold change > 4 between the control and TSA treatment at 6 h or 12 h was used to identify and restrict the number of differentially expressed genes (fold change > 4, 458 probe sets). We then performed hierarchical clustering of the above 458 probesets to identify genes that respond similarly to the various experimental conditions. The cluster analysis was done using cluster version 2.11 46 applying mean-centering and normalization of genes and arrays before average linkage clustering with uncentered correlation. The expression profiles of differentially expressed transcripts were discriminated into four groups according to the expression profile of the hierarchical clustering.

Functional annotation of transcripts differentially expressed ≥ 4-fold was done using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) 31 to obtain Gene Ontology annotations for the category of 'biological process'. Transcript redundancies were removed based on statistical analysis and PCR validation data.