Casein Kinase 1 (cid:1) Interacts with Retinoid X Receptor and Interferes with Agonist-induced Apoptosis*

Agonists of retinoid X receptors (RXRs), which include the natural 9- cis -retinoic acid and synthetic analogs, are potent inducers of growth arrest and apoptosis in some cancer cells. As such, they are being used in clinical trials for the treatment and prevention of solid tumors and are used to treat cutaneous T cell lymphoma. How-ever, the molecular mechanisms that underlie the anti-cancer effects of RXR agonists remain unclear. Here, we show that a novel pro-apoptotic pathway that is induced by RXR agonist is negatively regulated by casein kinase 1 (cid:1) (CK1 (cid:1) ). CK1 (cid:1) associates with RXR in an agonist-de-pendent manner and phosphorylates RXR. The ability of an RXR agonist to recruit CK1 (cid:1) to a complex with RXR in cells correlates inversely with its ability to inhibit growth. Remarkably, depletion of CK1 (cid:1) in resistant cells renders them susceptible to RXR agonist-induced growth inhibition and apoptosis. Our study shows that CK1 (cid:1)

Retinoids (natural retinoic acids and their synthetic derivatives) exert biological effects by regulating gene transcription through two classes of nuclear receptors, the retinoic acid receptors (RAR 1 ␣, -␤, and -␥) and the retinoid X receptors (RXR␣, -␤, and -␥) (1,2). RXRs are unique among members of the nuclear receptor superfamily in their ability to heterodimerize with several other nuclear receptors. RXR agonists (the natural 9-cis-retinoic acid (3,4) and its synthetic analogs) are potent inducers of growth arrest and apoptosis in a variety of malignant cells, such as lung and breast carcinomas, lymphomas, and leukemias (5)(6)(7)(8)(9)(10)(11)(12)(13)(14). However, despite the ubiquitous expression of RXRs, many malignant cells are insensitive to RXR agonists, and it is, therefore, important to identify cellular mechanisms that mediate this resistance.
RXRs associate with various proteins that play important roles in regulating cell survival. Nerve growth factor (NGF) induces the phosphorylation of the orphan nuclear receptor NGFI-B (Nur77 or TR3), which is heterodimerized with RXR, leading to translocation of the NGFI-B⅐RXR complex out of the nucleus (15). In the presence of apoptotic stimuli, NGFI-B is translocated from the nucleus to mitochondria, causing cytochrome c release and apoptosis (16,17). RXR␣ binds insulinlike growth factor-binding protein-3 and is necessary for the induction of apoptosis by insulin-like growth factor-binding protein-3 or RXR agonists in cancer cells (18). We have shown that RXR agonists affect the proliferation of cells by promoting interaction of RXR␣ with ␤-catenin and the degradation of both of these molecules (19). The variety of roles that RXRs can play in regulating cell survival leads to the possibility that tumor cells may circumvent RXR-driven apoptosis.
Here we demonstrate the requirement of both RXR agonist and RXR protein for driving the pro-apoptotic and growth inhibitory effects in cells and identify a protective mechanism for the RXR action in which an RXR agonist-induced apoptotic pathway is negatively regulated by the protein kinase casein kinase 1␣ (CK1␣).
Cell Culture, Recombinant Protein Expression, and Transactivation Assay-Human embryonic kidney (HEK293), HeLa, CV-1, and Jurkat cells were from ATCC and grown according to ATCC instructions. DT40 cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum. HircRXR cells, the rat fibroblasts that stably express the human insulin receptor and RXR, were kindly provided by Dr. Jyoti Kusari.
The effects of AGN194204 on the growth of cell lines were measured by counting viable cells after treatment with various doses for 3-6 days. Data are presented as a percentage of control (without treatment).
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
For the transactivation assay, CV-1 cells were seeded at a concentration of 1.5 ϫ 10 5 cells per well in 6-well plates. After overnight culture, cells were transfected as described in the legend to Fig. 4. All transfections contained a ␤-galactosidase expression vector (Promega) to correct transfection efficiency. The pRXRE-Luc is the report plasmid that contains five tandem repeats of a 35-base pair sequence (DR-1) from the promoter of the mouse CRBP-II gene (21) inserted immediately upstream of thymidine kinase-luciferase. pRARE-Luc contains three copies of DR-5 (22). 24 h after transfection, the cells were incubated for another 16 h in medium containing 0.5% charcoal-treated serum along with the ligand and then harvested for the measurement of ␤-galactosidase activity and luciferase activity using a system and protocol from Promega. Relative luciferase activity was normalized to the ␤-galactosidase activity to account for transfection efficiency. Triplicate wells were used for each experimental group.
Small Interfering RNA (siRNA)-siRNA experiments were carried out as described previously (23). The previously described siRNA oligonucleotides for CK1␣ and CK1⑀ together with plasmid DNAs were transfected into cells with LipofectAMINE (24). To generate HEK293RXR and Jurkat cells with stable suppression of the expression of CK1␣, the same sequences of double-strand DNA oligonucleotides were inserted into pSiren-RetroQ (BD Biosciences), and lines were created according to the manufacturer's protocols.

Immunoprecipitation, Immunoblot Analysis, and Flow Cytometric
Analysis-Immunoprecipitations were performed at 4°C by incubating clarified cell extracts with the antibodies (2-5 g/ml) and protein A/Gagarose beads for 4 h or overnight. Cell extracts or immunoprecipitates were subjected to separation on SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and immunoblotted with specific antibodies. The targeted proteins were detected using enhanced chemiluminescence immunodetection (Amersham Biosciences). DNA content analysis by fluorescence-activated cell sorter was performed on FACScan TM (BD Biosciences) as described (25).
Protein Kinase Assay and Phosphoamino Acid Analysis-Kinase reactions were initiated by adding 2 l of 50 M [␥-32 P]ATP (10 Ci) to the immunoprecipitated complexes that were suspended in 15 l of kinase assay buffer (30 mM Tris-HCl, pH 7.4, and 10 mM MgCl 2 ). Following incubation at 30°C for 20 min, reactions were terminated by adding 6 l of 4ϫ SDS sample loading buffer, and mixtures were resolved onto SDS-PAGE. Incorporation of 32 P into proteins was determined by autoradiography. Phosphoamino acid analysis was performed on thin layer cellulose plates using the Hunter thin layer electrophoresis system as described (26).

RXR Agonist Induces Growth Inhibition and Apoptosis in
Cells that Overexpress RXR␣-To examine the mechanism of action of RXR on cell growth and apoptosis, chicken B lymphoma DT40 and human T lymphoma Jurkat cells that stably overexpress the RXR␣ protein (DT40RXR and Jurkat RXR) were generated (Fig. 1a). These cells were dramatically more sensitive than parental cells to growth inhibition by the RXR agonist AGN194204 (Fig. 1b), and AGN194204 induced a sig-nificant level of apoptosis in these cells (Fig. 1c) but not in the parental cells (data not shown). Similarly, caspase-3 and -9 were activated by AGN194204 in a dose-and time-dependent manner in DT40RXR cells (Fig. 1d) and not in DT40 cells (data not shown). By contrast, H 2 O 2 activated caspase-3 and -9 in both DT40RXR and DT40 cells (data not shown). Agonistinduced growth inhibition and caspase activation in DT40RXR cells both were blocked by the RXR antagonist AGN195393, and this agent had no effect by itself (Fig. 1e). These data show that both the RXR protein and its agonist are required for growth arrest and apoptosis in DT40 and Jurkat cells.
RXR␣ Associates with a Protein Kinase in the Presence of RXR-specific Ligands-Studies using HEK293 cells revealed a protein kinase activity that co-immunoprecipitated with overexpressed FLAG-RXR␣ in an agonist-dependent manner and that phosphorylated RXR␣ and a 160-kDa protein in the same complex (Fig. 2a). Phosphorylation of RXRs in response to the activation of various pathways by extracellular stimuli has been described before (27)(28)(29). However, the phosphorylation of RXR and the interaction between a protein kinase and RXR that is induced by RXR agonist have never been reported. Agonist treatment of cells recruited an activated kinase to the RXR complex rather than activating an already associated kinase, because RXR immunocomplexes from untreated HEK293 cells did not show kinase activity when treated subsequently with AGN194204 (data not shown). AGN194204 recruited the kinase in a dose-dependent manner (Fig. 2b). Furthermore, several other RXR agonists (but neither an RXR antagonist nor an RAR agonist (TTNPB)) recruited the kinase activity (data not shown). Also, cotransfection of HEK293 cells with RAR␣ did not alter the specificity or efficacy of the AGN194204 response (Fig. 2c). Thus, the agonist-mediated recruitment of kinase activity is an RXR-driven response that is not mediated by classical RXR⅐RAR heterodimers.
The RXR-associated Kinase Is Identified as CK1␣-Phosphoamino acid analysis demonstrated that the co-immunoprecipitated kinase phosphorylates RXR␣ primarily on Ser residues (Fig. 3a). The kinase of interest appears to be constitutively active as inhibition or stimulation of several protein kinase cascades had no effect on the RXR-associated kinase activity (data not shown). The A/B domain of human RXR␣ contains 26 Ser residues, and analysis using a program from Protein Kinase Resources (University of California, San Diego) (30) yielded many potential RXR kinases. CK1␣ was of particular interest because of its known constitutive activity (31). Indeed, immunoblot analysis demonstrated the presence of CK1␣ in the RXR immunoprecipitates from lysates of cells treated with AGN194204 but not in immunoprecipitates from untreated cells (Fig. 3b). By contrast, CK1⑀, CK1␦, and CK2 were not detected in any of the immunoprecipitates, although they were clearly present in the cell lysates (data not shown). Several RXR deletion mutants were transfected into HEK293 cells, and the corresponding immunoprecipitates were examined for in vitro kinase activity and for the presence of CK1␣ (Fig. 3c). All of the immunoprecipitates that had kinase activity contained CK1␣. Moreover, an intact ligand-binding region (E domain) is both essential and sufficient for interaction of RXR with CK1␣, and only those RXR mutants with intact A/B domains were phosphorylated by the associated CK1␣.
To confirm that the RXR-associated kinase is CK1␣, siRNA was used to significantly reduce the level of expression of CK1␣ in HEK293 cells without affecting levels of CK1⑀ and other proteins (RXR␣ and ␤-actin) (Fig. 3d). Reducing the level of CK1␣ dramatically reduced the RXR-associated kinase activity, whereas a reduction of the expression of CK1⑀ had no effect (Fig. 3d). Also, recombinant CK1 effectively phosphorylated RXR␣ in vitro even in the absence of AGN194204 (data not shown). These data indicate that the RXR agonist functions in vivo to recruit CK1␣ to the RXR complex.
CK1␣ Is Not Required for Transcriptional Activities of RXR Homodimers or RXR⅐RAR Heterodimers-The transcriptional activity of certain hormone nuclear receptors and other transcription factors can be modulated by phosphorylation (32)(33)(34)(35). We investigated whether changing the expression levels of CK1␣ affects the transcriptional activity of RXR. Increased or decreased expression of CK1␣ had no effect on either RXR agonist-induced transactivation of RXR homodimers or RAR agonist-induced transactivation of RXR⅐RAR heterodimers (Fig. 4). These data indicate that the direct transcriptional roles of RXR are independent of CK1␣. Phosphorylation of RXR by activated c-Jun NH 2 -terminal kinases has also been reported to not affect the transcriptional functions of RXR (28).
Correlation of RXR Agonist-induced Growth Arrest and the Presence of CK1␣ Kinase Activity in RXR Immunoprecipitates-To provide insight into the biological consequence of RXR/CK1␣ interaction, we examined cell lines that had in- FIG. 3. Identification of the RXR-associated kinase as CK1␣. a, phosphoamino acid analysis of the phosphorylated RXR. Phosphorylated RXR was separated on SDS-PAGE, transferred onto a polyvinylidene difluoride membrane, detected by autoradiography, and cut out for hydrolysis with acid (6 N HCl). The 32 P-labeled phosphorylated amino acids, together with standard Ser(P), Thr(P), and Tyr(P), were separated by two-dimensional cellulose thin layer electrophoresis and detected by autoradiography. Ser(P), Thr(P), and Tyr(P) were visualized by ninhydrin staining and are indicated by arrows. b, AGN194204-dependent pull down of CK1␣ with RXR␣. The total cell lysates from each of the transfected cells were blotted with anti-FLAG antibody to check for expression of the RXR␣ mutants (top panel). d, depletion of CK1␣ by siRNA abolishes the kinase activity in RXR complexes. HEK293 cells were transfected with FLAG-RXR␣ (Control) or plus sense strand RNA (S-CK1␣ and S-CK1⑀) as negative controls or double-strand interfering RNA (SiRNA-CK1␣ and SiRNA-CK1⑀) to decrease the levels of CK1␣ or CK1⑀ proteins. After 48 h, the cells were treated with vehicle (Ϫ) or AGN194204 (ϩ) as indicated. Cell lysates were prepared, and the kinase activity in the RXR immunoprecipitated complexes was assayed (bottom panel). The protein levels of CK1␣, CK1⑀, FLAG-RXR, and ␤-actin (as an internal control) were checked by immunoblotting the total cell lysates with specific antibodies as shown (top three panels).
creased RXR expression. In some instances, these had enhanced sensitivity to RXR agonist-induced growth inhibition and apoptosis, whereas other RXR overexpressing cells were resistant (Fig. 5a). We investigated whether the association of RXR with CK1␣ governs RXR agonist-induced growth inhibition in these cell lines. RXR/CK1␣ interaction was measured using the in vitro kinase assay and by immunoblot analysis. As shown in Fig. 5c, RXR immunoprecipitates from the cells that were insensitive to inhibition of growth by RXR agonist (HEK293RXR and HeLaRXR) had kinase activity. Kinase activity was not immunoprecipitated with RXR in the AGN194204-sensitive cells (DT40RXR, JurkatRXR, and HircRXR) even though these cells expressed CK1␣ (Fig. 5b). These results imply that the RXR complex that mediates the growth inhibitory effects of RXR agonists is inactive in the presence of CK1␣.
Reducing the Level of CK1␣ in Cells Increases Efficacy of RXR Agonist-induced Cell Growth Inhibition and Apoptosis-We tested the possibility that RXR agonist-resistant cells may be rendered sensitive to growth inhibition by down-regulating the expression of CK1␣. siRNA was used to reduce the expression of CK1␣ in HEK293RXR cells (Fig. 6a, left panel), and this led to a concomitant reduction in RXR-associated kinase activity (Fig. 6a, middle panel). Importantly, the cells with decreased levels of CK1␣ were then sensitive to growth inhibition by AGN194204 (Fig. 6a, right panel). Similarly, native Jurkat cells that do not overexpress RXR are insensitive to growth inhibition by AGN194204 and became sensitive when CK1␣ levels were decreased (Fig. 6b, first two panels). In addition, AGN194204 treatment of these CK1␣-suppressed Jurkat cells resulted in the activation of caspases and the induction of apoptosis (Fig. 6b, last two panels).
RXRs play key roles in regulation of gene transcription by forming heterodimers with many other ligand-activated nuclear receptors. There have also been suggestions that RXRs have novel physiological functions besides the regulation of gene transcription (15)(16)(17)(18)(19). Casein kinase 1 was one of the first protein kinases discovered, yet its function and regulation remain poorly understood. CK1 represents a family of second messenger-independent Ser/Thr protein kinases. In mammals, CK1␣, ␤, ␥, ␦, and ⑀ have been identified and cloned, and each isoform appears to have different roles. CK1␣ is ubiquitously expressed, appears to be constitutively active (31), and has recently been identified as an essential component of the complex that controls ␤-catenin phosphorylation and degradation. Depletion of CK1␣ causes ␤-catenin protein accumulation in a manner similar to that caused by the lack of the functional tumor suppressor protein adenomotous polyposis coli (24).
We have clearly shown that RXR interacts in vivo with CK1␣ and that this interaction interferes with RXR agonist-driven inhibition of cell growth and induction of apoptosis. Interaction of CK1␣ with RXR does not affect the activity of RXR in regulation of gene transcription. Recently, Tanaka et al. (36) reported that in RXR agonist-unresponsive MDA-MB-231 cells, RXR␣ is located in interchromatin granule clusters and that forced nucleoplasmic expression of RXR renders these cells sensitive to induction of apoptosis by RXR agonist. These findings suggest that altered localization of RXR␣ coincides with loss of RXR agonist responsiveness. Interestingly, CK1␣ is also found in a multiprotein complex in interchromatin granule clusters (31,37). Therefore, it is possible that RXR agonistmediated apoptosis is regulated by a process of sequestration of phosphorylated RXR⅐CK1␣ into interchromatin granule clusters. It would be very interesting to investigate whether the phosphorylated RXR⅐CK1␣ complex is sequestered in interchromatin granule clusters. The exact molecular mechanisms by which RXR/CK1␣ interaction regulates cell growth and apoptosis are interesting subjects for future study.
Importantly, our findings add new insights into the novel mechanisms of RXR action and are of significant therapeutic importance in that inhibiting CK1␣ function can enhance the anti-cancer activities of RXR agonists. FIG. 6. Reducing levels of CK1␣ in cells increases sensitivity to RXR agonist-induced apoptosis. a, the effects of the depletion of CK1␣ in HEK293RXR cells. The cell line HEK293RXR-pSupCK1␣ was generated by stably expressing the CK1␣ RNA interference vector. The left panel shows the levels of CK1␣ in this cell line as compared with the parental line HEK293RXR, the middle panel shows the kinase activity of RXR immunoprecipitates from these cells, and the right panel shows growth inhibition by AGN194204 (10 Ϫ7 M, 4 days). b, the effects of depletion of CK1␣ in Jurkat cells. The Jurkat-pSupCK1␣ cell line was generated by stably expressing the CK1␣ RNA interference vector. The first panel shows the levels of CK1␣ in Jurkat and Jurkat-pSupCK1␣, the second panel shows growth inhibition by AGN194204 (10 Ϫ7 M, 4 days), the third panel shows activation of caspase-3 and -8 as revealed by immunoblotting for cleavages of procaspases, and the fourth panel shows the percentages of apoptotic cells after treatment with different doses of AGN194204.