Glucocorticoid Receptor α Isoform-Selective Regulation of Antiapoptotic Genes in Osteosarcoma Cells: A New Mechanism for Glucocorticoid Resistance

Abstract

Glucocorticoids regulate a variety of physiological processes and are commonly used to treat disorders of inflammation, autoimmune diseases, and cancer. Glucocorticoid action is predominantly mediated through the classic glucocorticoid receptor (GR)α isoform. Recent data suggest that the mature GRα mRNA is translated into multiple N-terminal isoforms that have distinct biochemical properties and gene regulatory profiles. Interestingly, osteosarcoma cells stably expressing the GRα-D translational isoform are unique in that they are resistant to glucocorticoid-induced apoptosis. In this study, we investigate whether GRα isoform-specific differences in the regulation of antiapoptotic genes contribute to this resistant phenotype. We now show that GRα-D, unlike the other receptor isoforms, does not inhibit the activity of a nuclear factor κB (NF-κB)-responsive reporter gene and does not efficiently repress either the transcription or protein production of the antiapoptotic genes Bcl-xL, cellular inhibitor of apoptosis protein 1, and survivin. The inability of GRα-D to down-regulate the expression of these genes appears to be associated with a diminished interaction between GRα-D and NF-κB that is observed in cells, but not in vitro, and likely reflects the sequestration of GRα-D in the nucleus. Deletion of the GRα N-terminal amino acids 98–335 also results in a nuclear resident GR, which fails to interact with NF-κB in cells and promote apoptosis in response to glucocorticoids. These data suggest that the N-terminal translational isoforms of GRα selectively regulate antiapoptotic genes and that the GRα-D isoform may contribute to the resistance of certain cancer cells to glucocorticoid-induced apoptosis.

Glucocorticoids elicit many physiological, pathological, and pharmacological responses that are mediated by the ubiquitously expressed glucocorticoid receptor (GR). The GR is a member of the nuclear receptor superfamily of ligand-activated transcription factors and maps to human chromosome 5 (5q31–32) (1, 2). Although there is only one GR gene that has been identified to date, multiple phases of processing yield numerous GR proteins that have distinguishing characteristics (3–5). The α-isoform of the GR (GRα) is the prototypic, most well-studied protein product of the GR gene. GRα is organized into modules, which contain an N-terminal transactivation domain, a central DNA-binding domain (DBD), and a C-terminal ligand-binding domain (Fig. 1A). In the classic model of glucocorticoid signaling, ligand-binding stimulates the translocation of GRα from the cytoplasm to the nuclear compartment (6). In the nucleus, GRα regulates the expression of genes involved in diverse process, such as immunity, inflammation, metabolism, behavior, and cell proliferation and survival. GRα directly activates or represses transcription through binding specific hormone response elements and recruiting coactivators or corepressors, respectively (7). GRα can also modulate gene transcription independent of its DNA-binding activity by interacting with other chromatin-bound transcription factors (8, 9), most notably nuclear factor κB (NF-κB) (10).

Fig. 1.

Translational isoforms of the GR. A, GRα is a modular protein that contains an N-terminal transactivation domain (NTD), a central DBD, and a C-terminal ligand-binding domain (LBD). The numbers indicate amino acid positions, and the shaded area corresponding to amino acids 187-227 represents the functional core region of the NTD, τ1 (34). B, The GRα translational isoforms GRα-A, GRα-B, GRα-C1, GRα-C2, GRα-C3, GRα-D1, GRα-D2, and GRα-D3 are generated through initiation of translation at internal start codons corresponding to methionines 27, 86, 90, 98, 316, 331, and 336, respectively. The asterisks that represent these internal methionine residues were mutated to isoleucines to generate osteosarcoma cell lines that express only one of the translational isoforms. C, The GRα Δ98-335 deletion mutant is missing amino acids 98-335, a region common to the GRα-A, GRα-B, and GRα-C isoforms but absent in GRα-D.

The NF-κB heterodimer is typically composed of p65 and p50 and is sequestered predominantly in the cytoplasm in an inactive state by inhibitory κB (IκB) (11), although recent studies have also found low levels of NF-κB subunits in the nucleus of quiescent cells (12, 13). Cytokine-induced stimulation of the NF-κB signal transduction pathway leads to the activation of IκB kinase (14), which results in the phosphorylation and degradation of IκB (15) followed by nuclear translocation of p65/p50 heterodimer (16). Similar to GRα, NF-κB regulates the expression of genes involved in immunity, inflammation, and cancer. In fact, the clinically relevant immunosuppressive and antiinflammatory actions of glucocorticoids are attributed to GRα interaction with and functional antagonism of NF-κB at endogenous promoters (17, 18). In certain cell types, GRα and NF-κB also exhibit opposing actions on programmed cell death (apoptosis). For example, constitutive NF-κB activity contributes to the antiapoptotic and tumorigenic phenotype of leukemia, lymphoma, and myeloma cells (19–21), whereas GRα activation stimulates apoptosis and tumor regression in these cancerous cells (22, 23). Furthermore, glucocorticoid-induced apoptosis of immature thymocytes is associated with down-regulation of NF-κB transcriptional activity (24). Thus, GRα-mediated functional antagonism of NF-κB-driven gene expression represents a potential mechanism for glucocorticoid-induced apoptosis in certain cell types.

GR variants are known to be generated through alternative splicing of primary GR mRNA (25–30). Recently, alternative initiation of translation has emerged as another mechanism for generating distinct subtypes of GRα (31, 32). In addition to the initial start codon that corresponds to methionine 1, internal start codons are also capable of directing translation of the mature GRα transcript, resulting in the production of eight GRα isoforms that differ only in the length of the N terminus (Fig. 1B) (31). These translational isoforms have been designated GRα-A, GRα-B, GRα-C1, GRα-C2, GRα-C3 (GRα-C), GRα-D1, GRα-D2, and GRα-D3 (GRα-D). To understand the relative contribution of the translational isoforms to the composite function of GRα, these N-terminal variants were individually expressed in the GR-null U-2 OS osteosarcoma cell line (31, 32). Microarray analysis revealed that the translational isoforms exhibit significant differences in their genomic response to glucocorticoids. In addition to regulating a common set of genes, the GRα variants each regulate a unique set of genes (31), consistent with the proposed role for the N terminus in transactivation and coregulator binding (33–36). The shortest isoform, GRα-D, is missing approximately 80% of the N terminus and exhibits several major differences in GRα function when compared with the other translational subtypes. In contrast to GRα-wild type (GRα-wt), GRα-A, GRα-B, and GRα-C, the GRα-D isoform is predominantly localized to the nucleus in the absence of hormone (31) and fails to mediate glucocorticoid-induced apoptosis (32). However, the molecular mechanisms that contribute to the apoptotic resistant phenotype of U-2 OS GRα-D cells are unclear. Data presented here highlight a novel role for a previously undescribed region of the GRα N terminus in the repression of antiapoptotic genes in osteosarcoma cells.

Results

GRα isoform-selective inhibition of apoptosis and NF-κB signaling

Glucocorticoids are known to have cell type-specific effects on survival. For example, glucocorticoids induce cell death in monocytes and lymphocytes yet protect against cell death in lung epithelial cells and hepatocytes (37, 38). Recent data suggest that the effect of glucocorticoids on cell survival varies even within the same cell type and is dependent on the repertoire of GRα isoform expression (32). To examine this phenomenon further, we evaluated the role of GRα translational isoforms (Fig. 1B) in glucocorticoid-induced cell death. Osteosarcoma cell lines that express only one of the GRα N-terminal variants were treated with the synthetic glucocorticoid dexamethasone (dex), and cell death was determined by flow cytometric analysis of propidium iodide (PI)-stained cells. In agreement with previous studies, we show that the GR-null cell line (U-off) is resistant to dex-induced cell death (Fig. 2A, upper panel). Stable expression of GRα-wt, GRα-A, GRα-B, or GRα-C renders osteosarcoma cells sensitive to the cell killing-effect of dex. However, osteosarcoma cells stably expressing GRα-D resemble U-off cells in that they are resistant to dex-induced cell death despite having the ability to bind ligand and induce gene expression (31, 32). To determine whether the isoform-selective effects on cell death occurred by apoptotic mechanisms, we also assessed caspase 3/7 activity. As shown in Fig. 2A, lower panel, caspase 3/7 activity was significantly increased in the GRα-wt, GRα-A, GRα-B, and GRα-C cell lines but not in the U-off or GRα-D expressing cells. These data suggest that glucocorticoid-induced apoptosis is dependent not only on the presence of GRα but also on the subtype of GRα expressed. Furthermore, these data imply that alterations in the translational control of GRα mRNA expression within one particular type of cell may significantly impact the glucocorticoid response.

Fig. 2.

GRα isoform-selective regulation of cell death and NF-κB activity. A, Osteosarcoma cell lines were treated with vehicle control or 100 nm dex for 24 h. Cells were collected and evaluated for PI staining (upper panel) and caspase 3/7 activity (lower panel) by flow cytometry. B, Osteosarcoma cell lines were transiently transfected with the NF-κB firefly luciferase (pGL2–3XMHC) and Renilla luciferase (pGL3-hRL) reporter plasmids. After treatment with vehicle control or 100 nm dex for 16 h, the cells were lysed and analyzed for luciferase activity. Firefly luciferase light units were normalized to Renilla luciferase light units and reported as fold change relative to control for each cell line. C, Western blot analysis of GRα, p65, and IκBα levels in the osteosarcoma cell lines. The level of β-actin was used as an internal loading control. D, Osteosarcoma cell lines were treated with vehicle (Con) or 100 nm dex for 1 h before nuclear/cytosolic fractionation. Cytosolic (C) or nuclear (N) extracts were then blotted with anti-GRα, anti-p65, and anti-β-actin antibodies, respectively. **, P < 0.01; ***, P < 0.001.

Glucocorticoid-induced apoptosis in immature thymocytes is associated with repression of NF-κB transcriptional activity (24). Therefore, we sought to determine whether GRα isoform-selective antagonism of the NF-κB pathway contributes to the lack of apoptosis induced by dex in GRα-D-expressing osteosarcoma cells. As a first step, the cell lines were transfected with an NF-κB reporter plasmid that contains three copies of the NF-κB response element from the major histocompatability complex (MHC) promoter fused to firefly luciferase. After treatment with 100 nm dex, NF-κB-driven luciferase activity was measured. Dex significantly inhibited basal NF-κB activity in osteosarcoma cells expressing GRα-wt, GRα-A, GRα-B, and GRα-C (Fig. 2B). In contrast, dex had no effect on NF-κB activity in receptor-negative U-off or osteosarcoma cells expressing the GRα-D isoform. The lack of effect in the GRα-D cells cannot be explained by alterations in the total cellular levels of GRα, p65, or IκBα (Fig. 2C) nor by changes in the nuclear/cytoplasmic distribution of p65 (Fig. 2D), because no major differences in protein were observed between the different cell lines. Thus, these data show that dex antagonizes basal NF-κB activity in osteosarcoma cells through a mechanism that depends on GRα. Based on the differential effect of GRα-C and GRα-D on NF-κB transcriptional activity, we conclude that the N-terminal GRα region containing amino acids 98-335 is likely important for dex-induced NF-κB antagonism. Taken together, these studies highlight a novel role for the GRα N terminus in regulating both apoptosis and NF-κB signaling. Furthermore, these data implicate a potential role for the GRα-D translational isoform in resistance to glucocorticoid-induced apoptosis.

GRα isoform-selective down-regulation of Bcl-xL, cellular inhibitor of apoptosis protein 1 (cIAP1), and survivin

The decision for a cell to undergo apoptosis is often determined by changes in the cellular balance of pro- and antiapoptotic proteins, leading us to investigate the impact of the GRα N-terminal isoforms on dex-mediated repression of constitutively expressed antiapoptotic genes Bcl-xL, cIAP1, and survivin. In the following experiments, osteosarcoma cells expressing GRα-wt and GRα-C were used to represent the group of cells that are sensitive to glucocorticoid-induced apoptosis, and the results in these cells were compared with results in the glucocorticoid-resistant U-off and GRα-D-expressing cells. Parallel experiments were also performed in the glucocorticoid-sensitive GRα-A and GRα-B cells, and the results were similar to those in GRα-wt cells (data not shown).

Osteosarcoma cells were treated with dex for 3, 6, and 12 h, and the relative quantity of Bcl-xL, cIAP1, and survivin mRNA was measured by real-time PCR (Fig. 3). Dex significantly inhibited Bcl-xL mRNA expression in the GRα-wt- and GRα-C-expressing cells in a time-dependent manner (Fig. 3A). The GR-null U-off cells were completely resistant to dex-induced down-regulation of Bcl-xL mRNA at all time points analyzed, whereas dex had only a small inhibitory effect on Bcl-xL mRNA in cells expressing GRα-D at the 12-h time point. Dex significantly inhibited the expression of cIAP1 mRNA in all GRα isoform-positive cells at all time points (Fig. 3B). However, GRα-wt and GRα-C were more effective mediators of dex-induced cIAP1 inhibition at 12 h when compared with GRα-D. Survivin mRNA expression was significantly down-regulated by dex in GRα-C cells at 12 h (Fig. 3C). Although a trend was apparent in GRα-wt cells, there was no significant effect of dex on survivin mRNA levels in U-off, GRαwt-, or GRα-D-expressing cell lines. Consistent with the real-time PCR data, dex significantly down-regulated Bcl-xL, cIAP1, and survivin protein expression at 24 and 48 h in GRα-wt- and GRα-C-expressing osteosarcoma cells (Fig. 4, A–E). In marked contrast, glucocorticoids had no significant effect on Bcl-xL, cIAP1, or survivin protein levels in U-off and GRα-D cells (Fig. 4, A–E). Together, these data show that the GRα-D isoform, unlike GRα-wt and GRα-C, does not efficiently mediate the repression of Bcl-xL, cIAP1, or survivin expression. These results suggest that GRα isoform-selective repression of antiapoptotic genes may contribute to the distinct apoptotic phenotypes of the osteosarcoma GRα cell lines.

Fig. 3.

GRα isoform-selective inhibition of antiapoptotic gene expression. Osteosarcoma cell lines were treated with vehicle control or 100 nm dex for the indicated time periods. After isolation of total RNA, the quantity of Bcl-xL (A), cIAP1 (B), and survivin (C) mRNA was measured by real-time PCR. The signal obtained from each gene was normalized to total RNA and reported as fold change relative to control for each cell line. **, P < 0.01; ***, P < 0.001.

Fig. 4.

GRα isoform-selective inhibition of antiapoptotic protein expression. A and B, Western blot analysis of Bcl-xL, cIAP1, and survivin levels in osteosarcoma cell lines after a 24-h (A) or a 48-h (B) exposure to vehicle control or 100 nm dex. The level of β-actin was used as an internal loading control. C–E, Densitometric analysis of Western blottings from three independent experiments. The level of Bcl-xL (C), cIAP1 (D), and survivin (E) was normalized to the level of β-actin and reported as fold change relative to control for each cell line. **, P < 0.01; ***, P < 0.001.

Recruitment of GRα isoforms to Bcl-xL, cIAP1, and survivin promoters

Several mechanisms have been identified for GRα-mediated repression of target genes, including the activated receptor binding directly to poorly-defined negative GR response elements or tethering itself to other chromatin-bound transcription factors, such as NF-κB. Therefore, we investigated the effect of glucocorticoids on protein assembly at the promoters of these antiapoptotic genes by performing chromatin immunoprecipitation (ChIP) assays with primers targeting NF-κB response elements that were identified by sequence analysis or previously described in the literature (39, 40). Dex treatment had no effect on the recruitment of the p65 subunit of NF-κB to the Bcl-xL, cIAP1, and survivin promoters regardless of the GRα cell line examined (Fig. 5, A and B). However, both GRα-wt and GRα-C were recruited to all three promoters in a glucocorticoid-dependent manner (Fig. 5, C and D). In contrast, no enrichment was observed for GRα-D at the promoters of these antiapoptotic genes (Fig. 5, C and D). Failure of GRα-D to be recruited to these genes is not due to a general inability of this isoform to bind DNA, because we have previously shown by ChIP that GRα-D is recruited to a variety of glucocorticoid-induced genes at levels similar to that measured for GRα-wt (32). We also evaluated the association of p65 with these promoters under basal conditions and observed a small enrichment compared with nonimmune IgG controls (Supplemental Fig. 1, published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org), consistent with several recent studies showing NF-κB occupancy of target gene promoters in the absence of proinflammatory stimuli (12, 13). Collectively, these results suggest that the dex-mediated repression of Bcl-xL, cIAP1, and survivin involves GRα recruitment to these promoters. Moreover, the GRα isoform-selective nature of promoter recruitment is consistent with the inability of GRα-D to down-regulate the expression of these antiapoptotic genes.

Fig. 5.

Selective recruitment of GRα isoforms to the promoters of antiapoptotic genes. A and C, ChIP assay. Osteosarcoma cell lines were treated with vehicle control or 100 nm dex 3 h before isolation of chromatin. Chromatin-associated p65 and GRα were immunoprecipitated with anti-p65 (A) and anti-GRα (C) antibodies, respectively. End-point PCR analysis was performed using primers specific for the Bcl-xL, cIAP1, and survivin promoters. Input DNA and PCR products were resolved on 2% agarose gels. B and D, Densitometric analysis of agarose gels from three independent experiments. The amount of Bcl-xL, cIAP1, and survivin DNA immunoprecipitated with the p65 (B) or GRα (D) antibody was normalized to input and reported as fold change relative to control for each cell line. **, P < 0.01; ***, P < 0.001.

GRα isoform-selective interaction with p65

Previous studies have shown that GRα physically associates with p65 (18, 41). Therefore, the ability of the translational subtypes of GRα to associate with p65 in osteosarcoma cells was evaluated by coimmunoprecipitation studies (Fig. 6). The p65 subunit of NF-κB associated with GRα-wt in the presence and absence of dex. The p65 subunit also associated with the GRα-C translational isoform in the presence and absence of dex, and the extent of the interaction was comparable with the GRα-wt-p65 interaction. However, the ability of GRα-D to associate with p65 in these cells was significantly compromised compared with GRα-wt and GRα-C. We further investigated these GRα isoform-selective interactions with p65 in vitro by employing a cell-free coimmunoprecipitation assay. In contrast to our finding in cells, in vitro translated GRα-wt and GRα-D associated with p65 to a similar extent (Fig. 6D). These results indicate that the cellular environment and/or location may contribute to the relatively weak interaction between GRα-D and p65, and thus provide a mechanism responsible for the impaired ability of this receptor isoform to repress expression of the Bcl-xL, cIAP1, and survivin genes.

Fig. 6.

Interaction between GRα-D and p65. A–C, Osteosarcoma cells were transiently transfected with the p65 expression vector and FLAG-tagged GRα-wt, GRα-C, or GRα-D expression vector. After treatment with vehicle control or 100 nm dex for 1 h, the cells were lysed and FLAG-tagged GRα proteins were immunoprecipitated with the anti-FLAG M2-agarose resin. The top panels of Western blottings show the amount of p65 coimmunoprecipitated with GRα-C in comparison with GRα-wt (A) and the amount of p65 coimmunoprecipitated with GRα-D in comparison with GRα-wt (B). The bottom panels of Western blottings show the total amount of p65 and FLAG-tagged GRα proteins expressed in the cell extracts. C, Densitometric analysis of Western blottings from three independent experiments. The level of p65 that coimmunoprecipitated with FLAG-tagged GRα proteins was normalized to the amount of GRα protein immunoprecipitated and total amount of p65 expressed. The data are reported as percentage of GRα-wt control. **, P < 0.01. D, In vitro translated ³⁵S-labeled GRα-wt (left panel) or GRα-D (right panel) were incubated with unlabeled p65 and immunoprecipitated with either IgG, p65, or GR antibodies as indicated. The labeled GRα-wt and GRα-D coimmunoprecipitated with p65.

Previous studies have demonstrated that the GRα translational isoforms differ with respect to subcellular localization (31), and GRα-D is unique among the isoforms in that it resides primarily in the nucleus of cells independent of glucocorticoid treatment. Therefore, differences in cellular compartmentalization may account for the decreased interaction between p65 and GRα-D. As a first step in determining whether nuclear sequestration of GRα-D plays a role in decreased association with p65, subcellular localization studies were performed in the GRα cell lines. As expected, GRα-wt and GRα-C were predominantly cytoplasmic in the absence of hormone, whereas GRα-D was predominantly localized to the nuclear compartment (Fig. 7). Dex stimulated the nuclear translocation of GRα-wt and GRα-C but had no discernable effect on the localization of GRα-D. Although the majority of p65 resided in the cytoplasm of these cells in the presence and absence of dex, a small amount appeared to be diffusely distributed in the nucleus. Similar results were also observed by cell fractionation studies for the GRα isoforms and p65 subunit of NF-κB (Fig. 2D). Thus, GRα-wt and GRα-C are contained within the same intracellular compartment as the majority of p65 before glucocorticoid treatment, whereas GRα-D is physically separated from the predominant portion of p65, suggesting that cytoplasmic colocalization may be necessary for maximal association of GRα and p65. Interestingly, a similar requirement for both GRα and NF-κB to reside in the cytoplasm for optimal cross talk to occur between these two signaling pathways has been suggested by several other recent studies (42, 43).

Fig. 7.

Colocalization of p65 with GRα isoforms. U-off (A), GRα-wt (B), GRα-C (C), and GRα-D (D) osteosarcoma cell lines were treated with vehicle control or 100 nm dex for 1 h and processed for immunocytochemistry. Cells were incubated overnight with primary anti-GRα and anti-p65 antibodies. After incubation with the appropriate secondary antibodies and DAPI nuclear stain, confocal images were generated on the Zeiss LSM510-UV meta.

Osteosarcoma cells expressing GRα Δ98-335 are resistant to glucocorticoid-induced cell death

Amino acids 98-335 are common to the GRα-A, GRα-B, and GRα-C isoforms but absent in GRα-D (Fig. 1B), suggesting that this region of GRα confers the isoform-selective differences in receptor subcellular distribution, interaction with p65, antagonism of NF-κB signaling, and glucocorticoid-induced apoptosis. To define the contribution of these amino acids to GR function, we generated the receptor mutant Δ98-335, in which amino acids 98-335 were deleted from the GRα-wt isoform (Fig. 1C). When expressed in osteosarcoma cells, Δ98-335 was functional, because it activated a glucocorticoid-responsive reporter gene after dex treatment (data not shown). Additionally, the Δ98-335 mutant localized predominantly to the nucleus of cells both in the absence and presence of dex (Fig. 8A). Thus, Δ98-335 is indistinguishable from GRα-D with respect to subcellular distribution. Moreover, Δ98-335 was similar to GRα-D in that the mutant displayed a relatively weak interaction with p65 (Fig. 8, B and C). The decreased ability of Δ98-335 to interact with p65 was also associated with resistance to dex-induced cell death (Fig. 8D). These data suggest that the absence of amino acids 98-335 renders the GRα-D translational isoform resistant to the cell-killing effect of dex by compromising its capacity to interact with p65 and antagonize antiapoptotic gene expression.

Fig. 8.

The GRα N-terminal amino acids 98-335 are important for receptor signal transduction. A, Osteosarcoma Δ98-335 cells were treated with vehicle control or 100 nm dex for 1 h and processed for immunocytochemistry. Cells were incubated overnight with primary anti-GRα and anti-p65 antibodies. After incubation with the appropriate secondary antibodies and DAPI nuclear stain, confocal images were generated on the Zeiss LSM510-UV meta. B, Osteosarcoma cells were transiently transfected with the p65 expression vector and FLAG-tagged GRα-wt or GRα Δ98-335 expression vector. After treatment with vehicle control or 100 nm dex for 1 h, the cells were lysed, and FLAG-tagged GRα proteins were immunoprecipitated with the anti-FLAG M2-agarose resin. The top panel of Western blottings shows the amount of p65 coimmunoprecipitated with Δ98-335 in comparison with GRα-wt. The bottom panel of Western blottings shows the total amount of p65 and FLAG-tagged GRα proteins expressed in the cell extracts. C, Densitometric analysis of Western blottings from three independent experiments. The level of p65 that coimmunoprecipitated with FLAG-tagged GRα proteins was normalized to the amount of GRα protein immunoprecipitated and total amount of p65 expressed. The data are reported as percentage of GRα-wt control. D, Osteosarcoma cell lines were treated with vehicle control or 100 nm dex for 24 h. Cells were collected and incubated in 10 μg/ml PI before flow cytometric analysis of PI positive cells. Data are reported as the number of PI positive cells per total number of cells analyzed. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Discussion

The data presented in this manuscript suggest that the N-terminal GRα translational isoforms have selective effects on the expression of antiapoptotic genes and cell survival. In contrast to GRα-wt, GRα-A, GRα-B, and GRα-C, the GRα-D isoform does not mediate glucocorticoid-induced cell death in osteosarcoma cells (Fig. 2A). In addition, GRα-D does not efficiently inhibit basal NF-κB reporter gene activity (Fig. 2B) nor repress expression of the antiapoptotic genes Bcl-xL, cIAP1, and survivin (Figs. 3 and 4). Further comparisons among the GRα translational isoforms revealed deficits in the ability of GRα-D to interact with the p65 subunit of NF-κB in cells (Fig. 6) and to be recruited to the promoters of the three antiapoptotic genes (Fig. 5). The unique nuclear compartmentalization of GRα-D that results in its physical separation from the majority of p65 in the cytoplasm (Fig. 7) likely contributes to the reduced association between GRα-D and p65. Mutational studies confirm the importance of the N-terminal amino acids 98-335, which are absent in GRα-D, for receptor subcellular localization, interaction with p65, and efficient induction of apoptosis by glucocorticoids (Fig. 8). Taken together, these data highlight a novel role for the GRα N terminus in modulating NF-κB signaling and cell death and suggests that GRα-D expression may be a contributing factor to glucocorticoid resistance.

In living cells, there is often a balance between pro- and antiapoptotic proteins whose established purpose is to regulate the cell death machinery. In response to glucocorticoids, the GRα-A, GRα-B, and GRα-C isoforms inhibit the expression of the antiapoptotic genes Bcl-xL, cIAP1, and survivin, and this repression disrupts the normal balance of pro- and antiapoptotic factors and creates a microenvironment that is conducive to apoptosis. In contrast, the glucocorticoid-activated GRα-D isoform does not effectively repress the transcription of these antiapoptotic genes; consequently, the cells survive glucocorticoid treatment. The net result is resistance of GRα-D-expressing cells to glucocorticoid-induced cell death. The impaired ability of nuclear-localized GRα-D to cross talk with the NF-κB pathway is consistent with previous studies showing that cytoplasmic colocalization is required not only for optimal association of GRα with p65 (43) but also for GRα-mediated repression of NF-κB via the catalytic subunit of protein kinase A (42). Moreover, amino acids residing in the GRα N-terminal domain that are missing from GRα-D have been shown to be necessary for full antagonism of NF-κB (44). Our finding in a cell-free system that GRα-D has the capacity to bind p65 at levels comparable with GRα-wt implicates the subcellular localization of GRα-D as playing an important role in its diminished interaction with and repression of p65 in cells. Nevertheless, we cannot exclude the possibility that the N-terminal deletion in GRα-D is, by itself, a contributing factor to the reduced effects on NF-κB, perhaps through alterations in the ability of this isoform to be posttranslationally modified and/or interact with various coregulators and cofactors.

Nuclear import and export coordinately regulate GRα nucleocytoplasmic shuttling and therefore impact GRα function (45). GRα contains two nuclear localization signals (NL1 and NL2), a nuclear retention signal and a nuclear export signal. All of these domains have been mapped toward the C-terminal region of the GRα protein. NL1 and NL2 are found within amino acids 479-506 and amino acids 526-777, respectively (46, 47); whereas the more recently identified nuclear retention signal is found within the hinge region of GRα and closely overlaps with the NL1 (48). The nuclear export signal is located between the zinc-binding loops of the DBD and is comprised of amino acids 442-456 (45). The data presented here also highlight a novel role for the N-terminal amino acids 98-335 in determining the subcellular localization of GRα. The deletion of these amino acids in GRα-D may alter the receptor conformation and constitutively expose NL1 or NL2 or prevent this isoform from interacting with other proteins necessary for its cytoplasmic distribution.

Although inherited forms of glucocorticoid resistance are due to mutations or polymorphisms in the GR gene (49, 50), acquired resistance to glucocorticoid therapy may occur in the absence of aberrant GR nucleotide sequence (51). The wild-type GR gene is normally processed into multiple mature protein subtypes that regulate both common and unique sets of genes (31, 32). Therefore, at any given time, the profile of GR signaling will be a function of the cellular repertoire of receptor isoforms and their composite actions (4, 52). Up-regulation of alternatively spliced GR variants, such as GRβ, GRγ, GR-P, or GR-A, has been associated with decreased cellular responsiveness to glucocorticoids and/or glucocorticoid resistance in vivo (28, 29, 53–58). Data presented here and elsewhere suggest that alternative usage of GRα mRNA start codons may also significantly impact cellular responsiveness to glucocorticoids (31, 32). Specifically, these data implicate a role for GRα-D in resistance to glucocorticoid-induced cell death (32). Resistance to glucocorticoid-induced cell death limits the therapeutic benefit of GR agonists in cancer chemotherapeutic regimens and is positively correlated with poor prognosis (59, 60). Therefore, it is of clinical importance to establish a detailed understanding of GR gene processing and the contribution of individual GR subtypes to composite GR action. Future research will ideally elucidate the molecular triggers involved in alternative codon usage in the GRα mRNA transcript. Of note, the relative levels of the GRα subtypes expressed in the human brain change during development and the aging process (61), and treatment of differentiated skeletal muscle cells with the selective estrogen-related receptor β/γ agonist was shown recently to induce a significant and selective increase in the GRα-D isoform (62). The triggers or switching factors controlling GRα isoform expression represent potential pharmacological targets for combating glucocorticoid resistance.

Materials and Methods

Reagents and antibodies

The synthetic glucocorticoid, dex (1,4-pregnadien-9α-fluoro-16α-methyl-11β,17,21-triol-3,20-dione), was purchased from Steraloids (Newport, RI), and fetal calf serum (FCS) was obtained from HyClone (Logan, UT). The REDTaq ReadyMix PCR Reaction Mix with MgCl₂, HEPES, penicillin-streptomycin, anti-FLAG M2-agarose resin, mouse IgG-agarose, protein A-sepharose, and anti-FLAG antibodies (monoclonal and polyclonal) were purchased from Sigma (St. Louis, MO). DMEM/F-12, glutamine, geneticin, hygromycin B, PI, goat antirabbit Alexa Fluor 488, and goat antimouse Alexa Fluor 594 were obtained from Invitrogen (Carlsbad, CA). The anti-IκBα and anti-Bcl-xL antibodies were from Cell Signaling Technology, Inc. (Danvers, MA), anti-p65 polyclonal antibodies were from Cell Signaling Technology, Inc. (Danvers, MA) and Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), and antisurvivin antibody was from Assay Designs, Inc. (Ann Arbor, MI). The monoclonal anti-p65 antibody was obtained from Zymed Laboratories, Inc. (San Francisco, CA). The anti-β-actin and anti-cIAP1 antibodies were purchased from Millipore (Billerica, MA). The generation of the rabbit polyclonal anti-GR 57 antibody was described previously (63).

Plasmids

FLAG-tagged constructs containing cDNA for GRα, GRα-C3, or GRα-D3 were generated by PCR and cloned into the pcDNA3.1/Zeo (+) vector (Invitrogen) using BamHI and XhoI restriction sites. The forward primers used for PCR are as follows: 5′-CTA GGA TCC GCC ACC ATG GAT TAC AAG GAT GAC GAC GAT AAG GAC TCC AAA GAA TCA TTA ACT CCT GGT AGA G-3′ (F1) for FLAG-GRα, 5′-CTA GGA TCC GCC ACC ATG GAT TAC AAG GAT GAC GAC GAT AAG GGA AAT GAC CTG GGA TTC CCA CAG CAG GGC-3′ for FLAG-GRα-C3, and 5′-CTA GGA TCC GCC ACC ATG GAT TAC AAG GAT GAC GAC GAT AAG AAT ACA GCA TCC CTT TCT CAA CAG CAG GAT C-3′ for FLAG-GRα-D3. The reverse primer 5′-CCG CTC GAG TCA CTT TTG ATG AAA CAG AAG-3′ was used in all reactions. The FLAG-tagged GRα Δ98-335 construct (Fig. 1C) was generated by sequential PCR and cloned into pcDNA3.1/Zeo (+) using BamHI and XhoI restriction sites. The primers F1 and 5′-GTT GAG AAA GGG ATG CTG TAT TCA TCA CTT TTG TTT CTG TCT CTC C-3′ (R1) were used to generate the 5′ end of FLAG-GRα Δ98-335 cDNA (piece A), and the primers 5′-GGA GAG ACA GAA ACA AAA GTG ATG AAT ACA GCA TCC CTT TCT CAA C-3′ (F2) and 5′-CCG CTC GAG TCA CTT TTG ATG AAA CAG AAG TTT TTT GAT ATT TCC ATT TGA ATA TTT TGG TAT CTG ATT GGT GAT G-3′ (R2) were used to generate the 3′ end (piece B). Pieces A and B were joined in a final PCR that contained primers F1 and R2. GRα Δ98-335 cDNA was subcloned into the pTRE2-hygro vector (CLONTECH Laboratories, Inc., Mountain View, CA) using MluI and EcoRV restriction sites and the following primers: 5′-GAT ACG CGT CTG ATG GAC TCC AAA GA-3′ and 5′-CCG GAT ATC TCA CTT TTG ATG AAA CAG-3′. All cloning products were verified through sequencing by the DNA Sequencing Core at the National Institute of Environmental Health Sciences.

Cell culture and stable cell lines

The parental U-2 OS osteosarcoma cell line was obtained from American Type Culture Collection (Manassas, VA) and maintained in DMEM/F-12 supplemented with 10% FCS, 2 mm l-glutamine, 50 U/ml penicillin, and 0.05 mg/ml streptomycin. U-off cells were maintained in DMEM/F-12 supplemented with 10% FCS, 2 mm l-glutamine, 200 μg/ml geneticin, 50 U/ml penicillin, and 0.05 mg/ml streptomycin. U-2 OS cells expressing GRα, GRα-A, GRα-B, GRα-C3, GRα-D3, and GRα Δ98-335 were maintained in DMEM/F-12 supplemented with 10% FCS, 2 mm l-glutamine, 200 μg/ml geneticin, 200 μg/ml hygromycin, 50 U/ml penicillin, and 0.05 mg/ml streptomycin. Establishment of the U-off cell line and the U-2 OS cell lines expressing GRα, GRα-A, GRα-B, GRα-C3, and GRα-D3 was described previously (31). The U-2 OS cell line expressing GRα Δ98-335 was generated in accordance with the previously published protocol (31). Briefly, U-off cells were plated in 150-mm dishes at a density of 3 × 10⁶. Cells were transfected with 38 μg of the pTRE2-hygro-GRα Δ98-335 vector using the TransIT-LT1 transfection reagent (Mirus Bio, Madison, WI). Twenty-four hours after transfection, cells were replated in selection media containing hygromycin. Selection media was replaced every 3 d.

Flow cytometry analysis

Glucocorticoid-induced cell death was evaluated by measuring both PI staining and caspase 3/7 activity using flow cytometry. Cells plated in six-well dishes were treated with 100 nm dex. After 24 h, floating as well as adherent cells were collected and incubated with a fluorescent aspartic acid-glutamic acid-valine-aspartic acid substrate (Caspatag from Millipore, Temecula, CA) for 1 h at 37 C in 5% CO₂ following the manufacturer's protocol. The cells were then washed twice with buffer provided and placed in a final volume of 500 μl. PI (1 μg/ml; Molecular Probes, Eugene, OR) was then added, and the cells were analyzed on a FACSort flow cytometer (Becton-Dickinson, San Jose, CA). Untreated, viable cells had a normal distribution of low caspase fluorescence, low PI fluorescence. Cells were considered caspase positive if their fluorescence was higher than normal cells but still PI negative. The percentage of dead cells was determined by high PI fluorescence, indicating a compromised plasma membrane.

NF-κB reporter assays

The effect of dex and the GRα isoforms on basal NF-κB activity was analyzed by an NF-κB reporter assay. Cells were plated in 100-mm dishes at a density of 2 ×10⁶ cells/dish. The next day, the TransIT-LT1 transfection reagent was used to cotransfect cells with 5 μg of the pGL3-hRL Renilla luciferase reporter (Promega, Madison, WI) and 10 μg of the pGL2–3XMHC-firefly luciferase NF-κB reporter (18). Twenty-four hours after transfection, cells were split into 48-well dishes at a density of 75,000 cells/well. Cells were grown in charcoal-dextran stripped FCS for 24 h before treatment with either vehicle or 100 nm dex for 16 h. The cells were washed in PBS and prepared for the luciferase activity assay according to the Dual-Luciferase Reporter Assay System protocol (Promega). In each experiment, luciferase activity was measured in triplicate and averaged. Firefly luciferase light units were normalized to Renilla luciferase light units and expressed as fold change relative to control for each cell line. Remaining samples were resolved on NuPAGE 4–12% Bis-Tris gels (Invitrogen) and analyzed by Western blotting.

Real-time PCR analysis

Quantitative PCR was used to determine the effect of dex and the GRα isoforms on Bcl-xL, cIAP1, and survivin mRNA. Cells were plated in 100-mm dishes at a density of 1 × 10⁵ cells/dish. After 24 h, growth media were replaced with media containing charcoal-dextran stripped FCS. The cells were grown for an additional 24 h before treating with either vehicle or 100 nm dex for the indicated time periods. Total RNA was isolated using the RNeasy mini kit and the ribonuclease-free deoxyribonuclease set from QIAGEN (Valencia, CA). The quantity of mRNA was measured using a TaqMan one-step RT-PCR procedure on the Prism 7900HT thermocycler (Applied Biosystems, Foster City, CA). Reaction mixtures were as previously described (32). Predesigned primer-probe sets for Bcl-xL, cIAP1, and survivin were obtained from Applied Biosystems. At least three independent RNA samples were collected at each time point and analyzed in triplicate. The signal obtained from each gene was normalized to total RNA and reported as fold change relative to control for each cell line.

ChIP assays

ChIP assays were performed using the Magna ChIP A Chromatin Immunoprecipitation kit (Millipore) according to the manufacturer's protocol. Briefly, cells were plated in 150-mm dishes at a density of 2 × 10⁶ cells/dish. After 24 h, growth media was replaced with serum-free media. The cells were grown for an additional 24 h before treatment with either vehicle or 100 nm dex for 3 h. The cells were fixed in 1% formaldehyde and harvested in lysis buffer containing protease inhibitors. The nuclear contents were then sonicated using a Branson Sonifier 150 at setting 4 with 10-sec pulses, four times on ice. Chromatin was immunoprecipitated using 4 μl of polyclonal antibodies to either GR or p65. The total amount of input DNA for each immunoprecipitation was 35 μg. After elution of protein:DNA complexes and DNA purification, end-point PCR analysis was performed. PCR contained 1× REDTaq Ready Mix PCR Reaction Mix with MgCl₂, 2–3 μl purified DNA, and 0.5 μm of each of the following primers: 5′-CCT CTC CCG ACC TGT GAT AC-3′ and 5′-CCC CCG TCT TCT CCG AAA TG-3′ for Bcl-xL, 5′-CAT TGG AAC AGT AGA CCT CCT G-3′ and 5′-ATC CGA TGC TTG CTT CTC TCT G-3′ for cIAP1, and 5′-GTT CTT TGA AAG CAG TCG AG-3′ and 5′-TCA AAT CTG GCG GTT AAT GG-3′ for survivin. Input DNA and PCR products were resolved on 2% agarose gels. For assessment of p65 binding to the Bcl-xL, cIAP1, and survivin promoters under basal conditions, sheared chromatin was precleared with rabbit IgG and protein A agarose/salmon sperm DNA and then immunoprecipitated overnight with 4 μg of rabbit IgG or p65 antibodies. Immunoprecipitated and input DNA were analyzed by real-time PCR using the following primer/probe sets: forward primer 5′-TCCTCTCCCGACCTGTGATACAAA-3′, probe 5′-TGCACCTGCCTGCCTTTGCCTA-3′, and reverse primer 5′-CACCACCTACATTCAAATCCGCCT-3′ for Bcl-xL; forward primer 5′-TGGGAAATGGTTCAGGGTCTTGGA-3′, probe 5′-CGGCTCCCTAATTAAGTGGCTTGCTA-3′, and reverse primer 5′-TTTGCCCGTTGAATCCGATGCTTG-3′ for cIAP1; and forward primer 5′-TAGGCTGCAGGACTTACTGTTGGT-3′, probe 5′-TGCTTTGCGAAGGGAAAGGAGGAGTT-3′, and reverse primer 5′-TGGGCCACTACCGTGATAAGAAGA-3′ for survivin. The primers used target previously described NF-κB response elements for Bcl-xL and cIAP1 (39, 40) or a putative NF-κB response element (5′-TGGGCTTTCCC-3′) located 678 nucleotides upstream from exon1 of the survivin gene and identified using the JASPAR CORE database (64).

Western blottings

Cells were washed twice in cold PBS and lysed for 60 min at 4 C in Triton X-100 lysis buffer [50 mm Tris (pH 7.4), 150 mm sodium chloride, 2 mm EDTA, and 1% Triton X-100] supplemented with protease inhibitor cocktail tablets (Roche, Indianapolis, IN). The extracts were clarified by centrifugation at 16,000 × g for 15 min at 4 C. Protein concentrations were determined by using the bicinchoninic acid protein quantification kit (Pierce, Rockford, IL). Extracts were diluted in sample buffer (NuPAGE LDS sample buffer supplemented with NuPAGE sample reducing agent; Invitrogen) and heated to 70 C for 10 min. Samples containing 10–50 μg protein were resolved on NuPAGE 4–12% Bis-Tris gels and transferred to nitrocellulose membranes. Membranes were incubated with primary antibody overnight at 4 C. All antibodies were diluted 1:1000 with the exception of the anti-β-actin antibody, which was diluted 1:20,000. After washing, membranes were incubated with secondary antibody (1:10,000) for 1 h at room temperature (RT). Blots were developed using the Amersham enhanced chemiluminescence Western blotting detection reagents (GE Healthcare Bio-Sciences Corp., Piscataway, NJ). Densitometry and quantification were performed by using National Institutes of Health ImageJ software. The data from three independent experiments were normalized to β-actin and reported as fold change relative to control for each cell line.

Nuclear/cytosolic fractionation

Cells were plated in 100-mm dishes at a density of 2 × 10⁶ cells/dish and grown in charcoal-dextran stripped FCS for 24 h. The cells were then treated with 100 nm dex or vehicle for 1 h. The cells were washed once with PBS and then harvested by scraping in 4 ml PBS containing protease inhibitors (Complete Mini; Roche, Mannheim, Germany). The cells were pelleted by centrifugation at 600 × g for 5 min at 4 C and subsequently fractionated into a cytosolic protein extract and a nuclear protein extract using the Nuclear/Cytosol Fractionation kit from Biovision Research Products (Mountain View, CA) following the manufacturer's protocol. Samples containing 15 μg of protein extract as determined using the Bio-Rad Protein assay (Bio-Rad, Hercules, CA) were then run on precast 4–20% Tris-Glycine gels (Invitrogen), transferred to nitrocellulose membranes, and blotted as described.

Immunoprecipitations

For immunoprecipitation of GR-p65 complexes, U-2 OS parental cells were plated in 100-mm dishes at a density of 1.5 × 10⁶ cells/dish. The next day, the TransIT-LT1 reagent was used to cotransfect cells with 2.5 μg of the p65 expression vector (pCMV-p65) (18) and 2.5 μg of the FLAG-tagged GRα, GRα-C3, GRα-D3, or GRα Δ98-335 expression vector. After 24 h, the transfection media were removed, and the cells were grown for an additional 24 h. Cells were then treated with either vehicle or 100 nm dex for 1 h before extracts were collected. Cells were washed twice in cold PBS and lysed for 60 min at 4 C in Triton X-100 lysis buffer containing protease inhibitors with or without 100 nm dex. The extracts were clarified by centrifugation at 16,000 × g for 15 min at 4 C. Cellular proteins (179 μg) were precleared with mouse IgG-agarose for 2 h (4 C) before immunoprecipitating overnight (4 C) with the anti-FLAG M2-agarose resin. The resin was washed four times with Triton X-100 lysis buffer in the presence or absence of 100 nm dex, and immunoprecipitates were eluted in sample buffer by heating to 70 C for 10 min. Cellular lysates and immunoprecipitates were resolved on NuPAGE 4–12% Bis-Tris gels before Western blotting. Densitometry and quantification of three independent experiments were performed by using National Institutes of Health ImageJ software. The amount of coimmunoprecipitated protein was normalized to input as well as the amount of immunoprecipitated FLAG-tagged GR. Data are reported as percent of GRα-wt control.

In vitro coimmunoprecipitation assay

Coimmunoprecipitation assays were carried out as described (65). Briefly, GRα-wt, GRα-D, or p65 proteins were in vitro translated using the TNT T7 Quick Coupled Transcription/Translation System (Promega) with the addition of either Tran³⁵S-label 70% l-methionine (GRα-wt and GRα-D) (MP Biomedicals, Solon, OH) or 1 mm methionine (p65) (Promega) to the programmed lysate. An equal volume of p65 was then combined with either GRα-wt or GRα-D, and 100 nm dex was added to each tube. The treated lysates were incubated on ice for 2 h, then incubated for 30 min at RT. A 10-fold volume of cross-link buffer (20 mm HEPES, 50 mm KCl, 2.5 mm MgCl₂, and 1 mm dithiothreitol) was added per sample, and the samples were incubated at RT for 15 min to equilibrate. Dithiobis succinmidyl proprionate (Thermo Scientific, Rockford, IL) was added to a final concentration of 2.5 mm, and the samples were further incubated at RT for 30 min to cross-link the proteins. The cross-link reagent was quenched by the addition of ethanolamine (Sigma-Aldrich, St. Louis, MO) to a final concentration of 0.1 m and incubated for 15 min at RT to insure the quenching was complete. The samples were diluted to a final volume of 1 ml in immunoprecipitation buffer [10 mm Tris-Cl (pH 7.4), 150 mm NaCl, 2 mm EDTA, 0.5% deoxycholate, 0.5% Nonidet P-40, and 0.1 m ethanolamine] with the addition of protease inhibitors (Complete Mini; Roche). To reduce nonspecific binding, the samples were precleared for 20 min at RT with normal rabbit IgG (Millipore) and Protein A Sepharose (Sigma-Aldrich). The Protein A Sepharose was pelleted, and the cross-linked proteins were immunoprecipitated overnight at 4 C with either normal rabbit IgG, anti-p65 (Santa Cruz Biotechnology, Inc.), or anti-GR no. 57. The next day, Protein A Sepharose was added and the samples further incubated at 4 C for 1 h. The samples were then washed two times with 1 ml IP buffer. Laemmli sample buffer was added and the proteins dissociated from the beads by boiling each sample for 5 min. The samples were subjected to SDS-PAGE on 4–20% Tris-Glycine gels and visualized by autoradiography.

Immunocytochemistry

Cells were plated in 35-mm glass-bottom dishes at a density of 2 × 10⁵ cells/dish. After 24 h, growth media were replaced with serum-free media. The cells were grown for an additional 24 h before treatment with either vehicle or 100 nm dex for 1 h. The cells were washed with PBS, fixed with 3.7% formaldehyde, permeabilized with 0.2% Triton X-100, and blocked with 5% goat serum for 1 h at RT. The cells were then incubated overnight at 4 C with the anti-GR 57 rabbit polyclonal antibody (1:1000) and the anti-p65 mouse monoclonal antibody (1:250). After washing three times with PBS, the cells were incubated with the goat antirabbit (Alexa Fluor 488) and goat antimouse (Alexa Fluor 594) secondary antibodies for 3 h at RT. The secondary antibodies were diluted 1:2000 in 5% goat serum. The cells were incubated with the 4′,6-diamidino-2-phenylindole (DAPI) nuclear stain for 10 min and subsequently washed three times with PBS. Confocal images were taken on a Zeiss LSM510-UV meta (Carl Zeiss, Inc., Oberkochen, Germany) using a Plan-NeoFluar 40×/1.3 Oil differential interference contrast objective.