A Permissive Role for Phosphatidylinositol 3-Kinase in the Stat5- mediated Expression of Cyclin D2 by the Interleukin-2 Receptor*

The interleukin-2 (IL-2) receptor promotes T cell proliferation in part by inducing the expression of D-type cyclins, which enable cells to progress from the G1 to S phase of the cell cycle. We previously showed that the IL-2 receptor induces expression of cyclin D2 by activating the transcription factor Stat5, which binds directly and immediately to a site upstream of the cyclin D2 promoter. We show here that subsequent transcription of the cyclin D2 gene occurs by a delayed, cycloheximide-sensitive mechanism, which implies the involvement of additional regulatory mechanisms. The transcription factor c-Myc is induced by Stat5 and is reported to bind to two E box motifs in the cyclin D2 promoter. However, in IL-2-stimulated T cells, c-Myc does not appear to be involved in cyclin D2 induction, since we found that these two E boxes are preferentially bound by USF-1 and USF-2 and, moreover, are dispensable for cyclin D2 promoter activity. Instead, we found that Stat5 activates the phosphatidylinositol 3-kinase (PI3 kinase) pathway by a delayed, cycloheximide-sensitive mechanism and that PI3 kinase activity is essential for the induction of cyclin D2 by Stat5. Chromatin immunoprecipitation experiments revealed that PI3 kinase is required for the optimal binding of RNA polymerase II to the promoters of cyclin D2 as well as other genes. Our results reveal a novel link between PI3 kinase and RNA polymerase II promoter binding activity and demonstrate discrete, coordinated roles for the PI3 kinase and Stat5 pathways in cyclin D2 transcription.

A hallmark feature of the immune response is the ability of resting T cells to undergo rapid clonal expansion in response to antigen stimulation. Optimal proliferation of T cells requires signals from cytokines such as interleukin-2 (IL-2) 1 (1), which promotes G 1 to S phase cell cycle progression. Critical downstream events in this process are the increased expression of Dand E-type cyclins, which bind and up-regulate the activity of cyclin-dependent kinases 2, 4, and 6 (2). These complexes phosphorylate retinoblastoma family proteins, thereby causing the derepression of E2F transcription factors, which in turn promote cell cycle progression (3). Of the three D-type cyclin genes present in mammals, cyclin D2 and cyclin D3 are up-regulated in response to IL-2 in T cells (4). However, the molecular mechanisms that regulate expression of these genes remain poorly defined.
The IL-2 receptor (IL-2R) is a trimeric complex of three transmembrane proteins, termed ␣, ␤, and ␥ c (1). Whereas the ␣ subunit functions primarily to increase ligand affinity, signaling is initiated by the IL-2 induced heterodimerization of the ␤ and ␥ c subunits (5,6). This results in the activation of the Jak1 and Jak3 tyrosine kinases, which are associated with the intracellular domains of the ␤ and ␥ c subunits, respectively (7,8). Once activated, the Jak kinases phosphorylate tyrosine residues on the IL-2R␤ chain, which then serve as docking sites for the recruitment of downstream signaling effectors.
Earlier studies by our group and others have delineated two distinct proliferative signaling pathways downstream of the IL-2R (9 -12). One is mediated by the adaptor molecule Shc, which binds to phosphorylated tyrosine 338 of IL-2R␤ (13). The other is mediated by the transcription factor Stat5, which associates with phosphorylated tyrosines 392 and 510 of IL-2R␤ (9). Although the molecular signaling mechanisms of these two pathways contrast considerably, each can independently generate a proliferative signal that is characterized by the induction of several promitogenic genes, including c-myc, cyclin D2, cyclin D3, cyclin E, bcl-2, and bcl-x L (11,12,14).
The Shc pathway is composed of at least two major branches: the Ras/Erk pathway, which is activated through a Shc-Grb2-Sos complex, and the phosphatidylinositol 3-kinase (PI3 kinase) pathway, which is activated through a Shc-Grb2-Gab2 complex (15)(16)(17)(18)(19)(20). The Erk pathway has well established roles in several aspects of cell cycle control (21) but is not required for the IL-2R-mediated induction of D-and E-type cyclins (14). In contrast, PI3 kinase is essential for the induction of cyclin D2 and cyclin D3 by the Shc pathway in T cells (14), and in fibroblasts this pathway is both necessary and sufficient for the induction of cyclin D1 (22,23). In addition to its role in inducing D-type cyclin expression, the PI3 kinase pathway enhances cyclin-dependent kinase function by down-regulating activity of the cyclin-dependent kinase inhibitor p27 Kip1 (24 -27).
Stat5 is also clearly implicated in D-type cyclin induction. T cells from Stat5a/b-deficient mice fail to express cyclin D2 or cyclin D3 in response to anti-CD3 or IL-2 treatment (28). In the pancreatic ␤-cell line INS-1, a dominant negative form of Stat5 partially inhibited growth hormone-mediated expression of cyclin D2 (29), whereas a constitutively active version of Stat5 induced cyclin D2 expression (30). We have previously characterized a Stat5 binding element in the cyclin D2 promoter that is responsive to IL-2R signaling in T cells (31). This site is adjacent to an Sp1 binding site, and the binding of both Stat5 and Sp1 is required to achieve full activity of the promoter. A Stat5 binding element has also been identified in the cyclin D1 promoter (32).
Although the Stat5 and PI3 kinase pathways have each been implicated in the regulation of D-type cyclin expression, the mechanism by which these two signals are integrated remains undefined. In the present study, we use mutant IL-2 receptors to demonstrate that Stat5 activation results in a late wave of PI3 kinase pathway activity. Studies with PI3 kinase inhibitors demonstrate that PI3 kinase is required for cyclin D2 induction by Stat5. This requirement does not pertain to the activation of Stat5 itself but rather to the ability of RNA polymerase II to bind to the cyclin D2 promoter. Therefore, the Stat5 and PI3 kinase pathways play distinct, coordinated roles in the IL-2Rmediated induction of cyclin D2.
Linearized plasmids were electroporated into cells, and stably transfected subclones were selected at limiting dilution for G418 resistance. Receptor expression was assessed by flow cytometry with antibodies to human GM-CSFR␣ or -␤ c (sc-458 and sc-457; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) or human IL-2R␤ (18745B; BD Biosciences). Recombinant human GM-CSF was purchased from Amgen (Thousand Oaks, CA). LY294002 and wortmannin were purchased from Calbiochem.
Reporter Gene Assays-Construction of the Ϫ1624 D2-Luc reporter gene has been described (31). Splice overlap extension PCR (34) was used to mutate the CACGTG motif within E box 1 (Ϫ1596) and/or E box 2 (Ϫ1425) to ACATGT. Luciferase reporter gene assays were performed as described (31).
Cross-linked chromatin was thawed and sonicated six times for 30 s each, resulting in chromatin fragments averaging 400 -800 bp in size. Immunoprecipitations were conducted as previously described (35). Antibodies to Stat5 (sc-835) and RNA polymerase II (sc-899) were purchased from Santa Cruz Biotechnology. Immunoprecipitated DNA was extracted and subjected to duplex PCR using the murine ␤-globin gene as a neutral, nonenriched reference. PCR cycle numbers were chosen to provide a linear range of amplification. Results were visualized on acrylamide gels and quantified by phosphor screen imaging analysis. Enrichment values were calculated as the ratio of gene signal to ␤-globin signal divided by the ratio of gene signal to ␤-globin signal for nonimmunoprecipitated samples (input DNA). Primer sequences used for Stat5 ChIP PCR were as follows (5Ј to 3Ј, forward/reverse): cyclin D2, CGAGCCATTTCCTAGAAAGC/AGGGCTAGCAGCTGTTCCTC; CIS, AGGGTCCCTGCACTTCAATA/TTGGACCGAGAAGGAAACAC; bcl-x, GTCGCCGGAGATAGATTTGA/ACCAGCCAGCTTTGAATTGT.

Stat5-mediated Cyclin D2 Expression Is
Delayed and Requires de Novo Protein Synthesis-Studies were performed in the murine IL-2-dependent CD8 ϩ T cell line CTLL-2. Chimeric GM-CSF/IL-2 receptors were used where indicated to study distinct signaling pathways by receptor mutagenesis. As previously described, these receptors are composed of the extracellular ligand binding domain of the GM-CSF receptor ␤ and ␣ subunits fused to the transmembrane and intracellular signaling domains of the IL-2 receptor ␤ and ␥ subunits, respectively, to yield ␤␤wt and ␣␥wt ( Fig. 1A) (6,12). Treatment of cells expressing ␤␤wt and ␣␥wt with GM-CSF produces an IL-2-like signal. To study the Stat5 signal in the absence of other IL-2R␤-dependent signals such as Shc, we utilized ␤␤⌬325ϩY510, a minimal form of ␤␤wt that contains Jak binding sites and a single tyrosine residue (Tyr 510 ) that mediates a robust proliferative signal through Stat5 (12,14).
Previous work by our group has identified cyclin D2 as a direct Stat5 target gene (31). However, despite the fact that Stat5 is activated within minutes of receptor ligation, Northern blot analysis revealed that cyclin D2 mRNA is not induced by either the wild type IL-2R or ␤␤⌬325ϩY510 receptor signal until at least 3 h after stimulation (Fig. 1, B and C). These kinetics contrast with the relatively rapid induction of other Stat5 target genes, including c-myc, CIS, and bcl-x L (Fig. 1C) (12,36,37). To determine whether the slow induction of cyclin D2 reflects a requirement for de novo protein synthesis, we applied the translation inhibitor cycloheximide to cells expressing ␤␤⌬325ϩY510. Cycloheximide treatment did not affect Stat5 phosphorylation or the induction of c-myc, CIS, or bcl-x L , which is consistent with their classification as immediate early genes. By contrast, cycloheximide completely inhibited the in-duction of cyclin D2. These results suggest that, unlike other Stat5 target genes, cyclin D2 expression requires additional factors that are cycloheximide-sensitive, presumably because they are either labile or newly synthesized upon receptor activation.

IL-2-induced Cyclin D2 Transcription Is Not Mediated through Promoter-proximal E Box Motifs-Previous studies
have suggested a role for c-Myc in the transcriptional activation of the cyclin D2 gene (38,39). In these studies, activated Myc-Max complexes were found to bind two conserved E box motifs located upstream of the cyclin D2 translational start site. Because c-myc mRNA and protein expression are strongly induced by Stat5 within 1-3 h of IL-2 stimulation ( Fig. 1C and data not shown), we evaluated whether c-Myc might represent the cycloheximide-sensitive factor required for cyclin D2 induction by ␤␤⌬325ϩY510.
In the human cyclin D2 gene, the E boxes reportedly bound by Myc-Max are located at positions Ϫ1596 bp (E box 1) and Ϫ1425 bp (E box 2) relative to the translational start site (38,40). We have previously shown via deletional analysis of a human cyclin D2 luciferase reporter gene that the region be-tween Ϫ1624 and Ϫ1303 bp is not required for IL-2-inducible reporter gene activity (31). Because the two conserved E boxes are located within this region of the promoter, this would suggest that they are not important for the IL-2-mediated induction of cyclin D2. However, in light of the aforementioned studies, we reevaluated the functional activity of these E boxes by directly mutating them in a full-length cyclin D2-luciferase reporter construct. Specifically, we created reporter constructs containing the upstream 1624-bp region of the human cyclin D2 gene with either both E boxes intact, E box 1 mutated only, E box 2 mutated only, or both E boxes mutated. As a negative control, we used a reporter construct containing only the first 444 bp upstream of the cyclin D2 gene, which we have previously shown to be nonresponsive to IL-2 (31). Consistent with prior results, after transient transfection into CTLL-2 cells, the full-length, wild type cyclin D2-luciferase reporter gene was induced ϳ3-fold in response to IL-2, whereas the reporter gene with the Ϫ444 deletion showed no significant induction ( Fig.  2A). The three constructs containing singly or doubly mutated E boxes were induced to the same level as the full-length, wild type reporter gene. Therefore, neither E box 1 nor E box 2 appears to contribute significantly to the IL-2-mediated induction of cyclin D2 in T cells.
To directly test whether Myc-Max complexes bind to these E boxes, we conducted EMSAs using DNA probes representing the two E box sequences. Both E boxes were constitutively bound by a factor that was present in unstimulated cells and maintained throughout 6 h of IL-2 stimulation (Fig. 2B). The addition of antibodies to c-Myc or Max did not result in the supershifting of the E box probes (data not shown), but antibodies to the transcription factors USF-1 or USF-2, which are both known to bind E box motifs (41), eliminated the E box 2 EMSA band at all time points (Fig. 2B). Thus, in T cells, c-Myc does not appear to contribute to cyclin D2 induction by the IL-2R, given that E box 1 and E box 2 are not required for inducible reporter gene activity and appear to constitutively bind the transcription factors USF-1 and USF-2 rather than Myc-Max or Mad-Max complexes.
Stat5 Induces Delayed Activation of the PI3 Kinase Pathway-Since it appeared unlikely that c-Myc was a critical coregulator of cyclin D2 induction, we considered a possible role for the PI3 kinase pathway. This hypothesis stemmed from earlier reports that PI3 kinase is involved in D-type cyclin expression (14,(22)(23)(24) and that Stat5 and PI3 kinase interact in response to IL-3 signaling (42). Indeed, our own studies have demonstrated the ability of PI3 kinase to potentiate Stat5mediated proliferation in T cells (14).
We and others have previously shown that rapid activation of the PI3 kinase pathway by the IL-2R occurs through the Shc pathway and that signals from the Jak kinases and Stat5 are not capable of mediating this event (14,20). However, this issue has not been adequately addressed at later time points, in particular the 3-6 h time period during which cyclin D2 induction occurs. As shown previously, stimulation of the endogenous wild type IL-2R induced phosphorylation of the PI3 kinase effectors Akt and p70 S6 kinase (43-45) within 30 min, and this persisted for at least 6 h (Fig. 3A) (14). Although the ␤␤⌬325ϩY510 receptor did not induce rapid phosphorylation of Akt and p70 S6 kinase , it consistently induced a modest level of phosphorylation of these kinases after 3 h of stimulation (Fig.  3A). Similar results were obtained in the IL-3-dependent pro-B cell line Ba/F3, where signaling through ␤⌬325ϩY510 (a mutated IL-2 receptor with the same intracellular modifications as ␤␤⌬325ϩY510) (12, 31) caused a late wave of Akt and p70 S6 kinase phosphorylation like that seen in CTLL-2 cells (Fig.  3B). This was in contrast to the endogenous IL-3 receptor, which, like the IL-2 receptor, can directly activate PI3 kinase through its association with Shc (20). Thus, in the absence of a Shc signal, the IL-2R can induce a late wave of PI3 kinase activity, which could conceivably facilitate cyclin D2 induction by Stat5.
To determine which IL-2R signaling components are required for this late wave of PI3 kinase activity, we utilized ␤␤⌬325ϩF510, a mutated form of ␤␤⌬325ϩY510 in which the tyrosine residue necessary for Stat5 binding has been pointmutated to phenylalanine (Fig. 1A) (12). This receptor can still activate Jak1 and Jak3 but is unable to activate Stat5, as demonstrated by phosphospecific Western blot analysis of these molecules (Fig. 3A). ␤␤⌬325ϩF510 failed to induce Akt or FIG. 3. Stat5 mediates a late wave of PI3 kinase pathway activity. A, CTLL-2 subclones expressing ␤␤⌬325ϩY510 or ␤␤⌬325ϩF510 were deprived of cytokines for 8 h and then stimulated with IL-2 (100 units/ml) or GM-CSF (GM) (100 ng/ml) for the indicated times. Western blot analysis of cytoplasmic extracts was performed to assess phosphorylation of Jak1, Stat5, and the PI3 kinase effectors, Akt and p70 S6 kinase . B, Ba/F3 subclones expressing ␤⌬325ϩY510 were deprived of cytokines for 8 h and then stimulated with either IL-3 (10% final concentration of WEHI-3 conditioned medium) or IL-2 (1000 units/ml). Western blot analysis of cytoplasmic extracts was performed to assess phosphorylation of Stat5, Akt, and p70 S6 kinase as in A. C, CTLL-2 subclones expressing ␤␤wt or ␤␤⌬325ϩY510 were treated as in A. Where indicated, 20 g/ml cycloheximide (CHX) was added to cells 30 min prior to cytokine stimulation. Data shown are representative results from multiple independent subclones. p70 S6 kinase phosphorylation throughout 6 h of stimulation. Thus, the late wave of PI3 kinase activity induced by ␤␤⌬325ϩY510 is not mediated by the Jak kinases alone but instead appears to require Stat5. The addition of cycloheximide to cells completely blocked the late wave of PI3 kinase activity mediated by ␤␤⌬325ϩY510, whereas the rapid Shc-dependent activity mediated by the wild type IL-2R signal was unaffected (Fig. 3C). Thus, activation of the PI3 kinase pathway by ␤␤⌬325ϩY510 requires de novo protein synthesis, which suggests that Stat5 might mediate this process by inducing the expression of an unknown signaling intermediary.
PI3 Kinase Is Required for Optimal Stat5-mediated Gene Induction-Having demonstrated that the PI3 kinase pathway is active during the time period in which cyclin D2 is induced by ␤␤⌬325ϩY510, we tested the potential role of PI3 kinase in Stat5-mediated gene induction using pharmacologic inhibitors. Treatment of cells with the PI3 kinase inhibitor LY294002 (LY) completely blocked PI3 kinase pathway activation by ␤␤⌬325ϩY510 as well as ␤␤wt, as evidenced by a lack of Akt and p70 S6 kinase phosphorylation (Fig. 4A). This was accompanied by a complete inhibition of ␤␤⌬325ϩY510-mediated cyclin D2 mRNA induction, whereas c-myc and bcl-x L expression were significantly but not completely blocked (Fig. 4B). In contrast, LY treatment had only a marginal effect on CIS expression. Similar results were achieved with the PI3 kinase inhibitor wortmannin (data not shown). These results suggest that PI3 kinase activity is required to varying degrees for the optimal induction of Stat5 target genes.
PI3 Kinase Is Not Required for Optimal Stat5 Phosphorylation, Nuclear Translocation, or DNA Binding-We investigated possible mechanisms by which the PI3 kinase pathway might contribute to the induction of Stat5 target genes. As shown in Fig. 5, the extent of ␤␤⌬325ϩY510-mediated Stat5 phosphorylation, either in the cytoplasmic or nuclear compartments, was not adversely affected by treatment with LY. Moreover, LY did not inhibit the nuclear translocation of Stat5 in response to receptor stimulation (Fig. 5). To assess Stat5 DNA binding function, EMSA was performed using DNA probes corresponding to defined Stat5 binding sites from the cyclin D2, CIS, and bcl-x genes (31, 36, 37) (Fig. 6A). All three probes were inducibly bound by Stat5 upon stimulation of either the endogenous IL-2R or ␤␤⌬325ϩY510 receptors. The addition of a 10-fold excess of cold competitor probe eliminated Stat5 binding to the radiolabeled probe, and the addition of a Stat5 antibody supershifted the DNA-protein complex, thereby demonstrating the specificity of Stat5 binding to these probes. Importantly, treatment of cells with LY did not inhibit the binding of Stat5 to any of these probes, thereby indicating that PI3 kinase is not required for Stat5 DNA binding activity.
We have previously demonstrated an important role for Sp1 in the IL-2-mediated induction of cyclin D2 (31). Therefore, we assessed the effect of LY on Sp1 binding to the human cyclin D2 promoter. EMSA experiments revealed that, like Stat5, Sp1 is not dependent on the PI3 kinase pathway for optimal DNA binding activity (data not shown).
To investigate whether PI3 kinase signaling is required for binding of Stat5 to native chromatin, we performed Stat5 ChIP assays. PCRs with primer sets specific to the known Stat5 binding sites of the cyclin D2, CIS, and bcl-x promoters were used to detect the relative amounts of gene enrichment obtained after Stat5 immunoprecipitation of formaldehyde-crosslinked chromatin (Fig. 6, B and C). Treatment of CTLL-2 cells with IL-2 resulted in the increased enrichment of cyclin D2, CIS, and bcl-x genomic DNA in Stat5-immunoprecipitated chromatin, indicating an increased level of Stat5 binding to these genomic sites (Fig. 6, B and C). Importantly, LY treatment did not decrease the level of enrichment of any of these genes, thereby demonstrating that PI3 kinase is not required for optimal Stat5 binding in the context of native chromatin. Indeed, in some experiments, LY treatment appeared to mod- Where indicated, LY (50 M) was added to cells 30 min prior to cytokine stimulation. Cytoplasmic and nuclear extracts were analyzed by Western blotting for phospho-Stat5 and total Stat5 content. Jak1 Western blots are provided as a control for protein loading. Nuclear Stat5 is represented by two isoforms of distinct molecular weight due to proteolytic cleavage by a nuclear protease (46). Data shown are representative results from multiple independent subclones. estly enhance the amount of Stat5 bound to the cyclin D2 and CIS genes, but these increases were neither statistically significant (Fig. 6C) nor consistently observed.
PI3 Kinase Is Required for Optimal Binding of RNA Polymerase II to the Cyclin D2 Promoter-Since PI3 kinase activity is not required for Stat5 phosphorylation, translocation, or DNA binding, we used ChIP to investigate a possible role for PI3 kinase in RNA polymerase II (RNA pol II) recruitment to Stat5 target genes. As shown in Fig. 7, the level of cyclin D2 enrichment in RNA pol II immunoprecipitates is relatively high in unstimulated cells, indicating a high basal level of RNA pol II binding to this promoter. The level of RNA pol II binding remained largely unchanged after 1 h of receptor stimulation and in some experiments waned after 6 h, although this was not consistently seen. A similar pattern of RNA pol II binding was observed with the bcl-x gene, although the overall level of enrichment was less than that seen with cyclin D2. In contrast, the level of RNA pol II binding to the CIS promoter increased sharply at 1 h and then waned at 6 h, indicating a significant degree of IL-2-inducible RNA pol II binding to this gene.
Treatment of cells with LY caused a substantial decrease in the amount of RNA pol II bound to the cyclin D2 promoter in both unstimulated and stimulated conditions (Fig. 7). Thus, the PI3 kinase pathway appears to play a role in facilitating RNA pol II binding to the cyclin D2 promoter. LY treatment also decreased the amount of RNA pol II bound to the CIS, bcl-x, and GAPDH promoters, again in both unstimulated and stimulated conditions. The indiscriminate effect of LY on all of the genes studied, in both unstimulated and stimulated conditions, did not reflect a technical problem, since LY did not inhibit gene enrichment in the Stat5 ChIP experiments (Fig. 6, B and  C). Instead, it appears that PI3 kinase activity is required for optimal binding of RNA pol II to a broad set of genes, including targets of Stat5 (cyclin D2, CIS, and bcl-x) and genes not regulated by Stat5 (GAPDH).
Since the inhibition of PI3 kinase decreased RNA pol II binding to several genes, we addressed the issue of whether increased PI3 kinase activity would lead to increased RNA pol II binding levels. RNA pol II ChIP assays were performed on cells stimulated through the endogenous IL-2R, which, unlike ␤␤⌬325ϩY510, strongly and rapidly activates the PI3 kinase pathway (Fig. 3A). Thus, if induced PI3 kinase activity is capable of modulating RNA pol II binding, then a change in RNA pol II binding activity should be evident during IL-2R stimulation. As shown in Fig. 7C, IL-2 stimulation of the endogenous IL-2R did not result in a significant change in RNA pol II binding to the cyclin D2 gene, thereby indicating that increased PI3 kinase activity, as well as other signals from the wild type IL-2R, is not sufficient to increase RNA pol II binding activity to this gene. Thus, our data demonstrate that by facilitating RNA pol II promoter binding, PI3 kinase plays a permissive role in regulating cyclin D2 expression by Stat5.

DISCUSSION
The present study provides new insights into the regulation of cyclin D2 expression by the Stat5 and PI3 kinase pathways in T cells. The IL-2R-mediated activation of Stat5 is a rapid event that leads to the immediate early expression of several Stat5 target genes, including c-myc, CIS, and bcl-x L . Whereas cyclin D2 is also a direct target of Stat5, it is induced more slowly and requires de novo protein synthesis, implying that additional factors downstream or parallel to the Stat5 pathway are required for complete gene activation. In this report, we demonstrate that the PI3 kinase pathway represents one such factor. Through the use of pharmacologic inhibitors, we showed that PI3 kinase activity is required for the optimal induction of a subset of Stat5 target genes, including cyclin D2, c-myc, and bcl-x L . PI3 kinase is not required for intrinsic Stat5 function, since the PI3 kinase inhibitor LY does not inhibit IL-2R-mediated Stat5 phosphorylation, nuclear translocation, or target gene binding. Instead, we found that the PI3 kinase pathway facilitates RNA pol II binding to the promoters of multiple genes, including cyclin D2, and thus plays a permissive role in gene induction by Stat5.
Previous reports by Bouchard et al. (38,39) showed that activated c-Myc induces cyclin D2 expression by directly binding to two conserved E box motifs in the cyclin D2 promoter. Because c-myc mRNA and protein are strongly induced by Stat5 in T cells, we hypothesized that c-Myc might be a critical link between Stat5 activation and cyclin D2 induction. However, our experiments failed to demonstrate binding of c-Myc to either of these E boxes, although we observed strong constitutive binding by the transcription factors USF-1 and USF-2. More importantly, we found these E boxes to be dispensable for the IL-2-induced activity of a cyclin D2 reporter gene. There are several possible explanations for the discordance between our results and prior studies. Whereas Bouchard's group used fi-broblasts and promyeloid cells, our studies were performed in T lymphocytes. Moreover, the Bouchard study relied largely on overexpressed and conditionally active forms of c-Myc, whereas our study involved physiologic levels of endogenous c-Myc induced by the IL-2 receptor. Whereas our data do not completely exclude the possibility that c-Myc may play some role in cyclin D2 induction, they clearly indicate that the two E boxes defined by Bouchard et al. (38,39) are not critical for this process in T cells.
Rapid activation of the PI3 kinase pathway by the IL-2R is mediated by the adapter protein Shc (20). However, we have now shown through receptor mutagenesis that, in the absence of Shc signaling, a late wave of PI3 kinase activity is induced through the Stat5 activation site of IL-2R␤. Moreover, like cyclin D2 induction, this late wave of PI3 kinase activity requires de novo protein synthesis, thus suggesting the involvement of one or more Stat5 target genes. Preliminary experiments have been initiated to identify Stat5-inducible factors that may be responsible for PI3 kinase activation. Conditioned medium from ␤␤⌬325ϩY510-stimulated cells failed to induce Akt phosphorylation when added to cytokine-starved CTLL-2 parental cells, thereby suggesting that the relevant factor is not secreted but rather acts intracellularly (data not shown). Continuing studies utilizing gene expression arrays may help to identify the factor(s) underlying this novel signaling event.
It is tempting to propose that the late wave of PI3 kinase activity induced by Stat5 is the critical late acting factor required for cyclin D2 induction. Indeed, treatment of cells with the PI3 kinase inhibitor LY or wortmannin inhibited Stat5mediated induction of cyclin D2. However, this hypothesis is inconsistent with our demonstration that the wild type IL-2R signal, which rapidly induces both Stat5 and PI3 kinase activity, nevertheless also induces cyclin D2 expression with delayed kinetics. In addition, we have previously shown that forced activation of the PI3 kinase pathway through a tamoxifen-regulated PI3 kinase construct does not affect the magnitude or kinetics of Stat5-mediated cyclin D2 induction by ␤␤⌬325ϩY510 (14). Thus, whereas PI3 kinase is clearly a complementary signal required for the optimal induction of cyclin D2 by Stat5, there must be at least one other late acting factor involved in the regulation of cyclin D2 induction by the IL-2R.
We have shown that the PI3 kinase pathway is required for the optimal Stat5-mediated induction of c-myc, cyclin D2, and bcl-x L . We have also shown previously that the PI3 kinase pathway is required for the optimal induction of c-myc, cyclin D2, cyclin D3, cyclin E, and bcl-x L by the Shc signaling branch of the IL-2R (14). Thus, the PI3 kinase pathway appears to play a role in providing a permissive environment for the induction of several promitogenic genes, regardless of the specific signals that trigger their induction. We propose that the permissive role played by the PI3 kinase pathway in promitogenic gene induction at least partially reflects its role in facilitating RNA pol II binding to the promoters of these genes.
Intriguingly, we found that even unstimulated CTLL-2 cells are sensitive to the effects of LY on RNA pol II promoter binding (Fig. 7). This implies that these T cells have a basal level of PI3 kinase activity that sustains constitutive RNA pol II promoter binding. Consistent with this, a low level of basal p70 S6 kinase phosphorylation, which can be eliminated by the addition of LY or rapamycin, is often evident in these cells (data not shown). Given the large number of stimuli that activate the PI3 kinase pathway in T cells, it seems reasonable to speculate that primary T cells would also exhibit some degree of PI3 kinase activity in the absence of IL-2 or other cytokines, and this might sustain basal RNA pol II promoter binding. Indeed, this could represent one mechanism by which co-stimulatory molecules such as CD28 and ICOS, which also activate the PI3 kinase pathway, might prime T cells for cytokine signaling (47).
Our studies with LY showed that the PI3 kinase pathway is also required for optimal RNA pol II binding at the promoter for GAPDH, a housekeeping gene that is not regulated by Stat5. Consistent with this finding, LY treatment caused a modest reduction in GAPDH mRNA expression (Fig. 4B). Thus, the requirement of PI3 kinase for RNA pol II recruitment appears to extend beyond IL-2 target genes and may represent a more general theme in transcription.
LY treatment substantially reduced the amount of RNA pol II binding at the CIS promoter, yet it only caused a slight inhibition of mRNA expression (Figs. 4B and 7). This implies that, for genes such as CIS, other regulatory mechanisms must compensate for reduced RNA pol II binding in the presence of LY. Intriguingly, the CIS gene is further distinguished from cyclin D2 and bcl-x by demonstrating a high level of inducible RNA pol II binding upon ␤␤⌬325ϩY510 receptor stimulation. One could speculate that the recruitment of RNA pol II to the CIS promoter is the major mode of transcriptional regulation for this gene. By contrast, cyclin D2 and bcl-x exhibit basal RNA pol II binding that does not increase with receptor stimulation (Fig. 7). Such a pattern of RNA pol II binding may be indicative of transcriptional regulation through other mechanisms such as promoter release of RNA pol II complexes or transcriptional elongation. Thus, if the PI3 kinase pathway plays a role in transcriptional elongation as well as RNA pol II loading, then genes such as cyclin D2, c-myc, and bcl-x that rely on both of these mechanisms might be more dependent on PI3 kinase pathway signaling than a gene like CIS, which may be primarily regulated through RNA pol II recruitment alone. Preliminary studies have indicated that the PI3 kinase pathway is indeed required for optimal transcriptional elongation of the c-myc gene in response to IL-2 signaling (data not shown). Further studies are required to clarify the role of the PI3 kinase pathway on transcriptional elongation of cyclin D2 and other promitogenic genes.
Although we were able to demonstrate a role for the PI3 kinase pathway in the binding of RNA pol II to the promoters of several promitogenic genes, the molecular mechanism by which this occurs remains unknown. Further insight into transcriptional control by the PI3 kinase pathway might be gained from studying other components of the transcriptional preinitiation complex as well as local chromatin configuration (48,49). Intriguingly, recent studies have identified nuclear isoforms of PI3 kinase that can be regulated by external stimuli such as growth factor receptors (50). The significance of nuclear PI3 kinases was highlighted by a report that linked nuclear PI3 kinase activity to cyclin D1 expression (51). Other studies have shown that nuclear PI3 kinases localize with subnuclear structures known as nuclear speckles, which contain elements of both transcriptional and pre-mRNA processing machinery (50,52,53). One could speculate that PI3 kinase and its phosphoinositide products may be involved in the arrangement of ordered nuclear structures that facilitate RNA pol II binding to promoters, as shown here, as well as subsequent RNA pol II complex formation, transcriptional elongation, and RNA splicing.