Seeking the genes responsible for the build up over time to GC-mediated apoptosis, we have used microarrays to uncover genetic "signatures". By sequentially screening transcripts regulated in common between various sub-types of sensitive, resistant, and resistant-restored-to-sensitive leukemic human cell lines and even non-malignant mouse thymocytes, we have discovered a small list of genes regulated in common among all GC-sensitive conditions. We believe these genes to be at the core of the apoptotic signature to GCs. As might be expected, cell type-specific subsets were found, and thymocytes showed fewer genes regulated in common with the GC-sensitive malignant cells. Among the signatory genes we find BCL2-like 11 apoptosis facilitator (BCL2L11, Bim), Down syndrome critical region gene 1 (DSCR1), Thioredoxin-interacting protein (TXNIP, VDUP1), DNA-damage-inducible transcript 4 (DDIT4, RTP801, REDD1, or dig2), Nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha (NFKBIA), FK506 binding protein 5 (FKBP5), and exclusive to the examined human leukemias GC-induced leucine zipper (TSC22D3, DSIPI, or GILZ), all of which have been investigated for possible roles in GC-stimulated apoptosis [22, 25–31]. These prior studies show both anti- and pro-apoptotic actions for many of these genes studied in isolation, but no study thus far has been able to link particular genes together to fill in all the critical gaps in the apoptotic mechanism. Our results suggest that this is because these genes work as part of a network and only in the context of the particular cellular network in which they are found can their action be fully understood. Further experiments manipulating more than one gene simultaneously will be necessary to close these gaps in such networks. Nonetheless, the fact that a relatively small set of genes are implicated in several leukemic lineages speaks strongly to the universality of the key genes systems at work in GC-dependent cell death.
Core profile genes have been shown relevant to GC-dependent apoptosis
Several genes in the set shared by all the GC-sensitive leukemic cells tested (see Table 2 and Additional file 2) have been demonstrated to be important for GC-dependent apoptosis. These genes include BCL2L11/Bim, DDIT4, DSCR1, TXNIP, NFKBIA, FKBP5, and TSC22D3/DSIPI/GILZ. In 1999, Bouillet et al.  observed that lymphocytes derived from Bim knockout mice were partially resistant to Dex-mediated apoptosis. Subsequently, we noted that Bim was upregulated by GC exclusively in two Dex-sensitive CEM clones . Independently, this finding was confirmed in CEM cells and extended to mouse lymphoma lines . Since then, Bim induction in response to GCs has been shown in chronic lymphocytic leukemia cells destined for apoptosis  and in thymocytes . Recently, our finding that GCs activate p38 MAPK in both human and mouse lymphoid cells  has been extended to show an interaction between GC-mediated p38 MAPK stimulation and the expression of Bim, linking these two pro-apoptotic pathways in context to the GC signal transduction system as well as enzymatic protein activity and transcription . We have demonstrated by qRT-PCR and confirm here by microarray that CVZ, which causes apoptosis in CEM-C1–15 cells, induces Bim in them, whereas Dex, to which they are resistant, does not . It has also been shown that Bim regulation converges with the PKA pathway  consistent with our findings of Bim induction in the CEM-C1–15 cells restored to Dex-sensitive by FSK activation of that pathway. The rise in Bim, which occurs late in the lag phase, may be one of the final precipitating factors initiating apoptosis.
Our data indicated that DSCR1, like Bim, was induced in all GC-sensitive conditions in both human and mouse indicating its importance in the apoptotic program regulated by GCs. DSCR1 antagonizes the biological effects of the phosphatase calcineurin by interacting with catalytic subunit A and disrupting its downstream signaling pathways . Data from glioblastoma cells suggests that in some cell types at least, DSCR1 may also stabilize IkappaB alpha, by an action independent of calcineurin .
TXNIP and DDIT4 regulate the redox state within the cell [40, 41]. This effect was tested and the induction of TXNIP with the subsequent reduction in thioredoxin activity in response to Dex was confirmed, in WEHI7.2 cells . Our comparative study indicates TXNIP to be regulated in all cells and conditions tested, but induction by Dex in the Dex-resistant cell line was very weak (Table 3).
NFKBIA antagonizes the action of transcription factor NFκB. We found NFKBIA to be induced by GCs in all systems tested, but like TXNIP, only weakly by Dex in resistant C1–15 cells. Inhibition of NFκB has been correlated with restoration of GC-sensitivity to resistant human monocytic/macrophage THP1 and U937 cell lines . Several reports have suggested the use of the inductive response of the immunophilin FKBP5 (also known as FKBP51) as a diagnostic tool for GC response in peripheral blood mononuclear, T- and B-lymphocytes, and the highly Dex-resistant lymphoblastoid cell line IM-9 [42–44]. We call attention to the fact that Dex strongly induced FKBP5 in the resistant C1–15 cells. Thus FKBP5 induction alone is not always an indicator of apoptotic sensitivity to GC.
TSC22D3/DSIPI/GILZ was highly induced in all the leukemic cells tested, regardless of apoptotic sensitivity or resistance. Thus, as with FKBP5, GILZ induction alone cannot be relied upon as an indicator of apoptotic sensitivity to GCs.
The final gene sets that we have discovered are almost certainly incomplete, limited as they are by the numbers of genes represented on the smallest microarray chips employed. Thus our gene profile sets, while valid, may be expanded by additional studies with more complete arrays on larger chips. Nevertheless, the gene sets we have identified provide an initial tool that can be tested as prognostic indicators and almost certainly offer greater reliability than a single gene, e.g. FKBP5 or GILZ.
Coincidence of regulated genes in varied leukemic cell types may provide the basis for prognostic screening
Therefore, we propose that our data offers the basis for developing a prognostic panel to screen relevant patients for the likelihood of an apoptotic response to GC. In at least some leukemias, response to GC is also predictive of response to overall chemotherapy [19, 20]. Our work identifies the commonality of certain genes regulated by GCs in the apoptotically sensitive state of several leukemias, regardless of lineage. We present in Additional file 3 the list of 191, (83 induced, symbol plus, 108 repressed, symbol minus) genes regulated in common and specific to each list from the CEM signature (see Additional file 1) and the multiple cell line signature (see Additional file 2).
Though low cellular GR levels or GR mutations may account for a fraction of GR-resistant leukemias, the majority of GC-resistant cases are GR-positive [45–47]. In one study, involving xenografts derived from pediatric ALL patient biopsies, the failure of Bim to induce upon exposure to GC was associated with the resistant phenotype . Combining this marker with the additional genes we have identified should strengthen the correlations.
The differences in GC-regulated genes found between the various types of leukemic cells examined may also be of importance. Examination of Additional files 4, 5, 6, 7, 8 and 9 shows that groups of genes are GC-controlled specifically in the adult B-cell ALL RS4;11, pediatric AML Kasumi-1, or pediatric B-cell ALL SUP-B15 lines. Since we were only able to examine one example of each of these types, it would be premature to conclude that the regulated genes found here as cell line-specific are representative of GC-apoptotic sensitivity in all leukemias of that class. We provide here in the additional files the comparisons that reveal the seemingly cell type-specific regulated genes, for the use of others and for the basis of wider-ranging experiments.
Shared GC-dependent gene regulation in mouse thymocytes and GC-sensitive leukemias speaks to a conserved pathway for GC-evoked apoptosis in lymphoid and other cells of hematologic origin. Since the "mouse chip" we employed only contains 6,000 probes, its breadth of sampling is limited compared to the human chips used. Even so, a sub-set of regulated genes common to the GC-sensitive human leukemic cell lines was also found to be regulated in the thymocytes. This indicates a remarkable underlying fundamental basis for the GC-dependent pre-apoptotic pathway considering the many differences between the systems: species, cells proliferating or not, oncogenetically transformed or not, lymphoid or myeloid. Finding several genes in the leukemic cells that are also regulated in the thymocyte system argues strongly for an essential basic network involved in GC-dependent apoptosis.