Studies in lymphocytes and fibroblasts have shown that resting cells are not in a passive state resulting simply from the lack of stimulation. Pajalunga et al. [12] summarized cell cycle exit as "a shift in the balance between positive and negative regulators of proliferation in favour of the latter". Indeed, resting cells were found to express sets of up-regulated and down-regulated genes that maintained an active state of non-division [1,12,13,15,16]. A group of genes required for cell cycle exit and the maintenance of cell quiescence in human fibroblasts following serum deprivation has been recently identified [17]. Comparative studies of transcriptional profiles of resting and stimulated lymphocytes [1,18] have shown that T lymphocyte quiescence is actively maintained by products of a set of genes, including TOB and KLF (LKLF, GKLF and BKLF) and FOXO families of transcription factors. Other candidate genes involved in lymphocyte quiescence, including TSC-22 and Dyrk1, have been identified in a mouse model among genes highly expressed in resting lymphocytes and down-regulated after T or B cell activation [1].

Five cell lines were subjected to IL-2 deprivation and microarray analysis, three primary, IL-2-dependent T lymphoblast lines derived from three different donors (denoted "6", "43" and "j") and two immortalised T cell lines (denoted line 5 and S9), examined in three biological replicates each. Altogether, 18 microarray hybridisations were performed.

In the first step, unsupervised Principal Component Analysis was performed to assess the reliability of the experiments. Biological replicates were shown to cluster together. The gene expression profile of all the analysed cell lines was shown to be affected by IL-2 withdrawal, but the major variability of the gene expression profile in the analysed dataset was found between primary and immortalised cell lines (unpublished data). Differences between primary and immortalised cells were much greater than those resulting from IL-2 withdrawal. Therefore, our analysis was carried out in a block design, and changes in gene expression following IL-2 withdrawal were assessed separately in primary and immortalised cell lines. By means of the parametric non-corrected t-test we selected 158 differentially expressed genes at p < 0.001 (see Additional File 1: Table A1 for genes differentiating between samples before and after IL-2 withdrawal, selected by a non-paired analysis). The probability of a random selection of such a number of genes was low (0.03 in a global test) which confirmed the reliability of the observed differences. A false discovery rate for the selected genes ranged from 0.003 to 0.12. The hierarchical clustering of the 158 differentially expressed probe-sets obtained in this analysis resulted in an ideal discrimination between the samples before and after IL-2 withdrawal, with a rather uniform pattern of expression (Figure ​(Figure11).

Having confirmed the statistical significance of differences related to IL-2 withdrawal, we have selected the most consistently changed genes in all five different cell lines analysed, by employing a paired analysis (of the corresponding cells before and after IL-2 withdrawal). The univariate non-corrected permutation test revealed 166 probe-sets changed significantly at P < 0.001 (while in a global test, a borderline significance of P = 0.0625 was obtained). A hundred and eight of these genes were repressed after growth factor withdrawal, and the remaining 58 were up-regulated (see Additional File 2: Table A2 for genes differentiating between samples before and after IL-2 withdrawal, selected by a paired sample analysis). Out of those, DUSP6, OSM, CISH, SOCS2, LIF and IL13 were found to be the most prominently down-regulated (more than 20-fold), and the expression of EMB, SOX4, RPS24, WNT5B, SLC1A4, SQSTM1, SUMF1, TMEM1, STK38, LAMP2 and C17orf42 was found to be the most increased (2-fold to 3.5-fold change).

We carried out an external validation of the results obtained in the microarray analysis by a qRT-PCR. For validation we used the same cell lines and we chose 8 genes from the list of 158 most significantly changed transcripts, and a further 4 genes, known for their interesting function, with lower significance of differences. Most of the qRT-PCR results confirmed the microarray findings (Table ​(Table1):1): 7 of 8 top genes and 3 of 4 less significant transcripts were confirmed in a qRT-PCR experiment, usually with a similar fold-change.

To select transcripts consistently changed upon IL-2 withdrawal, we combined the two lists of 166 and 158 significant genes, obtained in the paired and non-paired analyses, respectively. We identified a set of 53 genes comprised of 13 up-regulated and 40 down-regulated genes, which we designated a "T lymphocyte cell cycle exit signature" (Table ​(Table22).

The up-regulated genes identified after IL-2 deprivation in both primary and immortalised cells included SQSTM1, ECOP, YY1AP1, RPS24, TMEM1, LRRC8D, C1orf63, ASAH1, SLC25A46 and MIA3, some of which have been associated with cellular growth control.

The SQSTM1 protein, previously known as p62 protein, interacts with the ubiquitinated proteins to mediate their clearance and is an important scaffold molecule in the RANK-NF-kappaB signalling pathway [20]. In addition, while the p62 protein is also a ligand for Lck kinase of the Src protein-tyrosine kinase family that is involved in T-cell receptor-dependent activation [21], Köller et al. [22] stress the role of Lck in transient T cell unresponsiveness, mediated by a selective IL-2 deficiency. Thus, SQSTM1 may play an important role in cell growth, especially because it may facilitate cell survival through signalling cascades for example those that result in Akt activation, by interaction with protein kinase C, as shown in human neuronal cells [23].

Interestingly, the product of the ECOP gene, which was found here to be up-regulated, has also been previously shown to be involved in NF-kappa B-related regulation of cellular growth. ECOP (EGFR-Co-amplified and over-expressed protein) has been implicated as a key regulator in the NF-kappaB signalling, and it has been postulated that high-level, amplification-mediated ECOP expression, such as that occurring in tumours with amplified EGFR, could contribute to resistance to apoptosis [24].

The identified genes may as well be associated with other cell proliferation pathways. For example, a product of the YYAP1 gene, called YY1-associated protein, a co-activator of the YY1 (Yin Yang 1) transcription factor (YYAP1), had been initially identified as a long splice variant of HCC (hepatocellular carcinoma)-specific protein encoded by the HCCA2 gene (HCC-associated gene 2). An over-expression of HCCA2 has been found to lead to a cell cycle arrest at G0/G1 phase and to an inhibition of cell proliferation [25].

For the other up-regulated genes distinguished in this study, we could not identify links to the known mechanisms of cell cycle arrest, but some of the genes have been previously associated with different cancers. An over-expression of ASAH1 has been found in different malignant tumours, including prostate cancer, head and neck squamous cell cancers and T-cell large granular lymphocyte (LGL) leukaemia [26-28]. An inhibition of acid ceramidase, the product of ASAH1, the over-expression of which in cancer cells has been implicated in drug resistance, is suggested as an efficient and promising novel treatment strategy [29,30]. The ribosomal protein gene RPS24 has been associated with hepatocellular carcinoma [14]. The MIA3 (TANGO) protein has been identified as a tumour suppressor in malignant melanoma and in colon and hepatocellular carcinomas [31,32].

Out of genes encoding for the suggested crucial regulators of T lymphocyte quiescence, such as TOB and members of the KLF and FOXO transcription factors' families, as well as TSC-22 and Dyrk1, which had previously been found to be overexpressed in resting T cells [1,18,33-35], we only found the FOXO3 gene to be slightly, but significantly up-regulated after IL-2 withdrawal. This discrepancy might be explained by differences between the different cell models employed (mouse vs. human) and by different experimental settings. On the other hand, the activity of these transcription factors may be regulated posttranscriptionally, so that their mRNA levels may not necessarily reflect their activity. Thus, the expression of some of the known FOXO target genes was examined and found to be affected following IL-2 withdrawal. IL7R, the FOXO1 target gene was up to 4-fold overexpressed in the primary T lymphocytes, but not in immortalised T cells. CDKN1B (p27), another FOXO target in human, mouse and C. elegans models [1,34-37], encoding a cell-cycle regulatory protein, presented slightly, but significantly increased expression. Similarly, Pink1, a recently identified FOXO target, the product of which helps to protect lymphocytes from apoptosis after growth factor deprivation [38], was among genes of significantly elevated expression in our model of cell-cycle exit. Studies of Murphy [36] on C. elegans model pointed at metallothionein genes as possible FOXO targets. In the present study, several metallothionein genes (MT1H, MT1X, MT1G, MT2A, MT1M, MT1E, MT1F) were overexpressed (2.6 to 3.8 times) following IL-2 withdrawal. In addition, we show here, that Cyclin D expression, reported to be downregulated by FOXO and KLF [1,35], decreased after IL-2 withdrawal. However, some genes reported by other research groups as human FOXO targets, such as p130 gene, Cyclin G2 [37], or IGF1R, BCL-2BP, or Sestrin 1 found to be FOXO3a-inducible genes in mouse T cells [34], as well as most of the KLF targets summarised by Tsachanis [16] and Yusuf [33], were unaffected by IL-2 deprivation in our experimental model.

A set of genes identified as significantly down-regulated 8 hours following IL-2 withdrawal included PIM1, BCL2, IL-8, HBEGF, DUSP6, OSM, CISH, SOCS2, LIF and IL13. It is notable that some of the identified genes, namely PIM1, BCL2, DUSP6, CISH and SOCS2 were previously demonstrated to be the IL-2 target genes or the regulators of IL-2 signalling [39-41]. In addition, it is clear that PIM1, CISH and suppressors of cytokine signalling (SOCS) are implicated in the biological actions of IL-2 [42], and that IL-2 was found to regulate mRNA levels of SOCS2, CISH, DUSP5 and DUSP6 [41]. It has also been demonstrated that OSM can quickly up-regulate CISH [43]. The coordinated repression of OSM and CISH that we showed in T cells following IL-2 withdrawal is consistent with these observations.

To perform a functional analysis of the up- and down-regulated genes, the 158 genes selected in the non-paired comparison between IL-2-deprived and non-deprived lymphocytes were first analysed by Gene Ontology (GO) Categories (see Additional File 3. Table A3 for gene ontology analysis of genes selected in a non-paired sample comparison between IL-2-deprived and non-deprived lymphocytes). The number of genes changed in each category was compared with the expected quantity of genes and the greatest enrichment was found in two molecular function (MF) classes, GO:0016868, intramolecular transferase activity, phosphotransferases; (observed/expected ratio = 43.9) and GO:0005126, hematopoietin/interferon-class (D200-domain) cytokine receptor binding, (observed/expected ratio = 26.3) and two biological process (BP) classes, GO:0007259, JAK-STAT cascade (observed/expected ratio = 16.7) and GO:0009968, negative regulation of signal transduction, (observed/expected ratio = 6.8). Secondly, a gene set analysis was performed by comparing the overall significance of gene groups defined by GO categories. Out of the 2348 GO classes 317 were found to be significantly affected by IL-2 withdrawal, as shown by at least one of the three tests used to assess the significance of the differences. The selected GO categories are shown in Table ​Table3.3. The LS/KS permutation tests, which find gene sets that have more differentially expressed genes among the classes than expected by chance, identified 206 significant gene-sets. Efron-Tibshirani's test, which uses 'maxmean' statistics to identify differentially expressed gene-sets, found 231 significant gene-sets (under 200 permutations). Eleven out of the 215 cellular component categories, 69 of 545 molecular function (MF) categories and 237 of 1588 biological process (BP) categories were significant. Thirty-nine GO categories with p < 0.005 in all of the three applied gene group significance tests were identified (see Additional File 4: Table A4 for overall significance of expression changes following IL-2 deprivation in gene groups defined by GO categories).

The same analytical procedure applied to the database of biological pathways (Biocarta, human database) resulted in 54 significantly up-regulated gene sets. Selected pathways affected by IL-2 withdrawal are shown in Table ​Table4.4. Both analyses, GO and Biocarta, confirmed the consistent up-regulation of the genes belonging to the JAK-STAT pathway.

Our data demonstrate for the first time that IL-2 withdrawal induces a coordinate repression of the same set of genes that have been found to be induced during T cell activation as IL-2 targets [44,45]. A similar "symmetry" has been previously shown in human fibroblasts, where approximately half of the genes of the early response to serum stimulation were correspondingly repressed at the cell cycle exit after mitogen deprivation [17].

A few previous studies have shown that some mRNAs that are expressed in resting T and B lymphocytes become repressed following cell activation, and it was suggested that cell quiescence is under an active transcriptional control [1]. Similarly, as shown recently, exit from the cell cycle of human fibroblasts is under control of a "quiescence program" dependent upon a set of genes that enforce the non-dividing state, and ensure the reversibility of the cell cycle arrest [15,17]. Our data support the idea of cell cycle arrest as an active state, controlled by some up-regulated genes.

While in vitro T lymphocytes deprived of a growth factor would eventually undergo apoptosis, in vivo a small proportion of antigen-activated cells would exit to G₀ and become quiescent. Cellular quiescence is thought to be an indispensable state for the maintenance of lymphocyte homeostasis following immune response, and therefore it is an important barrier against tumorigenesis [46-48]. Some autoimmune and chronic inflammatory disorders involving an excessive lymphocyte proliferation, were found to be associated with increased risks of lymphoma [44,45]. Thus, it is very likely that a deregulated expression of genes responsible for cellular quiescence may contribute to the development of some lymphoid malignancies.

To assess whether T cells adopt quiescent, memory cell phenotype upon IL-2 withdrawal in our model, we compared the gene expression patterns obtained with those described by other researchers as characterising memory vs. naïve vs. effector T cells. Two molecules known to facilitate T cell homing to lymphoid tissues, SELL (CD62L) and CCR7 are expressed predominantly by naïve and memory T cells [49-52], but CCR7 expression seems not to be affected by growth factor withdrawal [52]. In our model of cell cycle exit the SELL gene was up-regulated in primary T cell lines, while CCR7, was unchanged in primary T cells and slightly down-regulated in immortalised T cells after IL-2 deprivation. Out of other genes identified by Holmes et al. [49] as up-regulated in memory T cells, T cells deprived of IL-2 overexpressed SMAD3, TRIM66 and NAP1L3. There was also an increased expression of IL7R, thought to be expressed in memory [53] and naïve [50,52] T cells, but this was found not significant by the qRT-PCR (Table ​(Table1)1) when analysed for all T cells, possibly because this was truth only in primary T cells. The CD58 (LFA3) gene, encoding an adhesion molecule known to be expressed on activated and memory T cells [54], was indeed highly expressed in actively growing cells and consistently up-regulated after IL-2 withdrawal in our experimental system.

Holmes et al. [49] reported Granzyme B and KLRD1 expression to be induced in activated T cells. Our experiments showed that growth factor-deprived T cells presented decreased Granzyme B expression, while KLRD1 expression was slightly upregulated, but only in primary T cells. A hierarchical clustering of some genes outlined by Holmes et al. [49] is shown in an Additional file 5: Figure A1.

Haining et al. [55] demonstrated a gene expression profile common for CD8 and CD4 memory T cells. We showed that some of the genes of this profile were significantly changed at cell cycle exit; there was an increased expression of S100A4, S100A11, ANXA1, ANXA2, CRIP1 and CASP1, which, according to Haining et al. [55], are up-regulated in memory T cells, and a reduced expression of Granzyme A, downregulated in memory cells [55].

In summary, T cells withdrawn from cell cycle gained some crucial characteristics of memory T cells. Nevertheless, the cell cycle exit signature that we describe here is different from memory T cell gene expression signatures published so far.